Edited by Gernot Friedbacher and Henning Bubert Surface and Thin Film Analysis
Related Titles Watts, J. F., Wolstenholme, J.
An Introduction to SIMS for Surface and Thin Film Analysis 2011 ISBN: 978-0-470-09132-6
Guo, J. (ed.).
X-Rays in Nanoscience Spectroscopy, Spectromicroscopy, and Scattering Techniques 2010 ISBN: 978-3-527-32288-6
Birkholz, M.
Thin Film Analysis by X-Ray Scattering 2006 ISBN: 978-3-527-31052-4
Bordo, V. G., Rubahn, H.-G.
Optics and Spectroscopy at Surfaces and Interfaces 2005 ISBN: 978-3-527-40560-2
Edited by Gernot Friedbacher and Henning Bubert
Surface and Thin Film Analysis A Compendium of Principles, Instrumentation, and Applications
Second, Completely Revised and Enlarged Edition
The Editors Prof. Dr. Gernot Friedbacher Institute of Chemical Technology and Analytics Getreidemarkt 9 /164 1060 Vienna Austria Dr. Henning Bubert Augsburger Weg 51 59439 Holzwickede Germany
All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate. Library of Congress Card No.: applied for British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at
. © 2011 Wiley-VCH Verlag & Co. KGaA, Boschstr. 12, 69469 Weinheim, Germany All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Composition Toppan Best-set Premedia Ltd., Hong Kong Printing and Binding Cover Design Adam Design, Weinheim Printed in the Federal Republic of Germany Printed on acid-free paper ISBN: 978-3-527-32047-9
V
Contents Preface to the First Edition XVII Preface to the Second Edition XIX List of Contributors XXI
1
Introduction 1 John C. Rivière and Henning Bubert
Part One 2 2.1 2.2 2.2.1 2.2.2 2.2.3 2.2.4 2.2.5 2.3 2.4 2.4.1 2.4.2 2.4.3 2.5 2.6 2.6.1 2.6.2 2.6.3 2.6.4 2.6.5 2.6.6
Electron Detection 7
X-Ray Photoelectron Spectroscopy (XPS) 9 Henning Bubert, John C. Rivière, and Wolfgang S.M. Werner Principles 9 Instrumentation 12 Vacuum Requirements 12 X-Ray Sources 13 Synchrotron Radiation 16 Electron Energy Analyzers 16 Spatial Resolution 18 Spectral Information and Chemical Shifts 19 Quantification, Depth Profiling, and Imaging 21 Quantification 21 Depth Profiling 23 Imaging 26 The Auger Parameter 27 Applications 28 Catalysis 28 Polymers 30 Corrosion and Passivation 31 Adhesion 32 Superconductors 34 Semiconductors 35
VI
Contents
2.7
Ultraviolet Photoelectron Spectroscopy (UPS) 38 References 39
3
Auger Electron Spectroscopy (AES) 43 Henning Bubert, John C. Rivière, and Wolfgang S.M. Werner Principles 43 Instrumentation 44 Vacuum Requirements 44 Electron Sources 44 Electron-Energy Analyzers 45 Spectral Information 47 Quantification and Depth Profiling 51 Quantification 51 Depth Profiling 53 Applications 54 Grain Boundary Segregation 54 Semiconductor Technology 56 Thin Films and Interfaces 58 Surface Segregation 58 Scanning Auger Microscopy (SAM) 61 References 64
3.1 3.2 3.2.1 3.2.2 3.2.3 3.3 3.4 3.4.1 3.4.2 3.5 3.5.1 3.5.2 3.5.3 3.5.4 3.6
4
4.1 4.2 4.3 4.3.1 4.3.2 4.3.3 4.4 4.5 4.6
5 5.1 5.2 5.3 5.3.1 5.3.2 5.3.3 5.4
Electron Energy-Loss Spectroscopy (EELS) and Energy-Filtering Transmission Electron Microscopy (EFTEM) 67 Reinhard Schneider Principles 68 Instrumentation 70 Qualitative Spectral Information 72 Low-Loss Excitations 74 Ionization Losses 77 Fine Structures 79 Quantification 83 Imaging of Element Distribution 85 Summary 88 References 89 Low-Energy Electron Diffraction (LEED) 93 Georg Held Principles and History 93 Instrumentation 94 Qualitative Information 96 LEED Pattern 96 Spot Profile Analysis 100 Applications and Restrictions 100 Quantitative Structural Information 101
Contents
5.4.1 5.4.2 5.4.3 5.4.4 5.5 5.5.1 5.5.2
Principles 101 Experimental Techniques 102 Computer Programs 104 Applications and Restrictions 105 Low-Energy Electron Microscopy 106 Principles of Operation 106 Applications and Restrictions 108 References 108
6
Other Electron-Detecting Techniques 111 John C. Rivière Ion (Excited) Auger Electron Spectroscopy (IAES) 111 Ion Neutralization Spectroscopy (INS) 111 Inelastic Electron Tunneling Spectroscopy (IETS) 112 Reference 113
6.1 6.2 6.3
Part Two 7 7.1 7.2 7.2.1 7.2.2 7.2.2.1 7.2.2.2 7.3 7.4 7.5 7.5.1 7.5.2 7.5.3 7.5.4 7.5.5 7.5.6 7.5.7
8 8.1 8.1.1 8.1.2 8.1.3
Ion Detection 115
Static Secondary Ion Mass Spectrometry (SSIMS) 117 Heinrich F. Arlinghaus Principles 117 Instrumentation 119 Ion Sources 119 Mass Analyzers 120 Quadrupole Mass Spectrometers 120 Time-of-Flight Mass Spectrometry (TOF-MS) 121 Quantification 123 Spectral Information 125 Applications 127 Oxide Films 128 Interfaces 128 Polymers 131 Biosensors 133 Surface Reactions 134 Imaging 135 Ultra-Shallow Depth Profiling 137 References 138 Dynamic Secondary Ion Mass Spectrometry (SIMS) 141 Herbert Hutter Principles 141 Compensation of Preferential Sputtering 141 Atomic Mixing 142 Implantation of Primary Ions 142
VII
VIII
Contents
8.1.4 8.1.5 8.1.6 8.2 8.2.1 8.2.1.1 8.2.2 8.2.2.1 8.2.2.2 8.3 8.4 8.4.1 8.4.2 8.4.3 8.4.4 8.4.4.1 8.4.4.2 8.4.4.3 8.5 8.6 8.6.1 8.6.2 8.7 8.7.1 8.7.2 8.8 8.9 8.9.1 8.9.2 8.9.3
Crater Bottom Roughening 142 Sputter-Induced Roughness 142 Charging Effects 142 Instrumentation 143 Ion Sources 143 Duoplasmatron 144 Mass Analyzer 144 Magnetic Sector Field 144 Detector 145 Spectral Information 146 Quantification 147 Relative Sensitivity Factors 147 Implantation Standards 147 Metal Ceside (MCs+) Ions 148 Theoretical Models 148 Electron Tunneling Model 148 Broken Bond Model 148 Local Thermodynamic Equilibrium LTE 148 Mass Spectra 149 Depth Profiles 149 Dual-Beam Technique for TOF-SIMS Instruments Molecular Depth Profiles 152 Imaging 152 Scanning SIMS 152 Direct Imaging Mode 153 Three-Dimensional (3-D)-SIMS 154 Applications 156 Implantation Profiles 156 Layer Analysis 157 3-D Trace Element Distribution 158 References 159
9
Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS) 161 Michael Kopnarski and Holger Jenett Introduction 161 General Principles of SNMS 162 Postionization via Electron Impact 163 Suppression of Residual Gas and Secondary Ions Instrumentation and Methods 166 Electron Beam SNMS 166 Plasma SNMS 167 Spectral Information and Quantification 170 Element Depth Profiling 172
9.1 9.2 9.2.1 9.2.2 9.3 9.3.1 9.3.2 9.4 9.5
152
164
Contents
9.6
Applications 174 References 175
10
Laser Secondary Neutral Mass Spectrometry (Laser-SNMS) Heinrich F. Arlinghaus Principles 179 Nonresonant Laser-SNMS 179 Resonant Laser-SNMS 179 Experimental Set-Up 180 Ionization Schemes 181 Instrumentation 182 Spectral Information 183 Quantification 183 Applications 184 Nonresonant Laser-SNMS 184 Resonant Laser-SNMS 186 References 189
10.1 10.1.1 10.1.2 10.1.3 10.1.4 10.2 10.3 10.4 10.5 10.5.1 10.5.2
11 11.1 11.2 11.3 11.4 11.5 11.6 11.6.1 11.6.2 11.6.3 11.6.4 11.7 11.8
12 12.1 12.2 12.3 12.3.1 12.3.2 12.4 12.5
Rutherford Backscattering Spectroscopy (RBS) Leopold Palmetshofer Introduction 191 Principles 191 Instrumentation 194 Spectral Information 194 Quantification 196 Figures of Merit 197 Mass Resolution 197 Sensitivity 198 Depth Resolution 198 Accuracy 198 Applications 198 Related Techniques 201 References 201 Low-Energy Ion Scattering (LEIS) Peter Bauer Principles 203 Instrumentation 206 LEIS Information 208 Energy Information 208 Yield Information 208 Quantification 211 Applications of LEIS 211 References 214
203
191
179
IX
X
Contents
13 13.1 13.2 13.3 13.4 13.5 13.6 13.7
14 14.1 14.2 14.3 14.4
15 15.1 15.2 15.2.1 15.2.2 15.2.3 15.3 15.3.1 15.3.1.1 15.3.1.2 15.3.2 15.3.2.1 15.3.2.2 15.3.2.3
16 16.1 16.1.1 16.1.2 16.2 16.3
Elastic Recoil Detection Analysis (ERDA) Oswald Benka Introduction 217 Fundamentals 218 Particle Identification Methods 220 Equipment 222 Data Analysis 223 Sensitivity and Depth Resolution 223 Applications 224 References 226
217
Nuclear Reaction Analysis (NRA) 229 Oswald Benka Introduction 229 Principles 231 Equipment and Depth Resolution 232 Applications 234 References 236 Field Ion Microscopy (FIM) and Atom Probe (AP) 237 Yuri Suchorski and Wolfgang Drachsel Introduction 237 Principles and Instrumentation 239 Field Ion Microscopy 239 Time-of-Flight Atom Probe Techniques 242 Field Ion Appearance Energy Spectroscopy 246 Applications 248 FIM Applications 248 FIM in Catalysis 248 Fluctuation-Induced Effects 249 Applications of AP Techniques 252 Applications of TOF-AP Techniques 252 PFDMS Applications 254 FIAES Applications 255 References 257 Other Ion-Detecting Techniques 261 John C. Rivière Desorption Methods 261 Electron-Stimulated Desorption (ESD) and ESD Ion Angular Distribution (ESDIAD) 261 Thermal Desorption Spectroscopy (TDS) 262 Glow-Discharge Mass Spectroscopy (GD-MS) 263 Fast-Atom Bombardment Mass Spectroscopy (FABMS) 263 References 264
Contents
Part Three 17 17.1 17.2 17.3 17.4 17.5 17.5.1 17.5.2 17.5.2.1 17.5.2.2
17.5.2.3 17.5.2.4
18 18.1 18.2 18.3 18.4 18.5 18.6
19
19.1 19.1.1 19.1.2 19.1.3 19.1.4 19.2 19.2.1 19.2.2 19.3 19.3.1 19.3.2 19.3.3
Photon Detection 265
Total-Reflection X-Ray Fluorescence (TXRF) Analysis 267 Laszlo Fabry, Siegfried Pahlke, and Burkhard Beckhoff Principles 267 Instrumentation 269 Spectral Information 275 Quantification 276 Applications 277 Particulate and Film-Type Surface Contamination 277 Semiconductors 278 Synchrotron Radiation-Based Techniques 280 Depth Profiling by TXRF and by Grazing Incidence XRF (GIXRF) for the Characterization of Nanolayers and Ultra-Shallow Junctions 283 Vapor-Phase Decomposition (VPD) and Droplet Collection 285 Vapor-Phase Treatment (VPT) and Total Reflection X-Ray Fluorescence Analysis 287 References 288 Energy-Dispersive X-Ray Spectroscopy (EDXS) 293 Reinhard Schneider Principles 293 Practical Aspects of X-Ray Microanalysis and Instrumentation Qualitative Spectral Information 303 Quantification 304 Imaging of Element Distribution 306 Summary 308 References 309 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies 311 P. Neil Gibson Principles 311 The Grazing Incidence X-Ray Geometry 312 Grazing Incidence X-Ray Reflectivity (GXRR) 314 Glancing Angle X-Ray Diffraction 314 ReflEXAFS 316 Experimental Techniques and Data Analysis 317 Grazing Incidence X-Ray Reflectivity (GXRR) 318 Grazing Incidence Asymmetric Bragg (GIAB) Diffraction Applications 321 Grazing Incidence X-Ray Reflectivity (GXRR) 321 Grazing Incidence Asymmetric Bragg (GIAB) Diffraction Grazing Incidence X-Ray Scattering (GIXS) 324
319
323
295
XI
XII
Contents
19.3.4
ReflEXAFS 325 References 326
20
Glow Discharge Optical Emission Spectroscopy (GD-OES) Volker Hoffmann and Alfred Quentmeier Principles 329 Instrumentation 330 Glow Discharge Sources 330 Spectrometer 334 Signal Acquisition 334 Spectral Information 335 Quantification 336 Depth Profiling 337 Applications 339 dc GD Sources 340 rf GD Sources 340 References 342
20.1 20.2 20.2.1 20.2.2 20.2.3 20.3 20.4 20.5 20.6 20.6.1 20.6.2
21 21.1 21.2 21.2.1 21.2.2 21.3 21.4 21.5
22 22.1 22.2 22.3 22.4 22.5
23 23.1 23.2 23.3 23.3.1 23.3.2
Surface Analysis by Laser Ablation 345 Roland Hergenröder and Michail Bolshov Introduction 345 Instrumentation 346 Types of Laser 346 Different Schemes of Laser Ablation 347 Depth Profiling 348 Near-Field Ablation 354 Conclusion 354 References 355 Ion Beam Spectrochemical Analysis (IBSCA) Volker Rupertus Principles 357 Instrumentation 358 Spectral and Analytical Information 360 Quantitative Analysis by IBSCA 361 Applications 363 References 366
357
Reflection Absorption IR Spectroscopy (RAIRS) 367 Karsten Hinrichs Instrumentation 367 Principles 368 Applications 369 RAIRS 369 ATR and SEIRA 372
329
Contents
23.4
Related Techniques 374 References 374
24
Surface Raman Spectroscopy 377 Wieland Hill and Bernhard Lendl Principles 377 Surface-Enhanced Raman Scattering (SERS) 378 Instrumentation 380 Spectral Information 382 Quantification 383 Applications 383 Unenhanced Raman Spectroscopy at Smooth Surfaces 383 Porous Materials 385 Surface-Enhanced Raman Spectroscopy (SERS) 386 Near-Field Raman Spectroscopy 387 Nonlinear Optical Spectroscopy 387 Sum Frequency Generation (SFG) Spectroscopy 387 Coherent Anti-Stokes Raman Scattering (CARS) 389 Stimulated Femtosecond Raman Scattering (SFRS) 389 Spatially Offset Raman Spectroscopy (SORS) 390 References 390
24.1 24.2 24.3 24.4 24.5 24.6 24.6.1 24.6.2 24.6.3 24.6.4 24.7 24.7.1 24.7.2 24.7.3 24.7.4
25 25.1 25.2 25.3 25.3.1 25.3.2
26 26.1 26.2 26.2.1 26.2.2 26.3 26.4 26.4.1 26.4.1.1 26.4.1.2 26.4.1.3 26.4.1.4
UV-VIS-IR Ellipsometry (ELL) 393 Bernd Gruska and Karsten Hinrichs Principles 393 Instrumentation 395 Applications 398 UV-Vis-NIR Spectral Region 398 Infrared Ellipsometry 400 References 405 Sum Frequency Generation (SFG) Spectroscopy 407 Günther Rupprechter and Athula Bandara Introduction to SFG Spectroscopy 407 SFG Theory 410 SFG Signal Intensity and Lineshape 412 Determining the Number Density of Molecules from SFG Signal Intensity 413 SFG Instrumentation and Operation Modes 414 Applications of SFG Spectroscopy and Selected Case Studies 417 SFG Spectroscopy on Solid Surfaces and Solid–Gas Interfaces 417 SFG Spectroscopy under UHV Conditions 417 Polarization-Dependent SFG Spectroscopy 419 SFG Spectroscopy under Near-Atmospheric Gas Pressure 420 SFG Spectroscopy on Supported Metal Nanoparticles 421
XIII
XIV
Contents
26.4.1.5 Time-Resolved (Pump-Probe) and Broadband SFG Spectroscopy 423 26.4.1.6 SFG Spectroscopy on Colloidal Nanoparticles and Powder Materials 427 26.4.2 SFG Spectroscopy on Solid–Liquid Interfaces 428 26.4.3 SFG Spectroscopy on Polymer and Biomaterial Interfaces 428 26.4.4 SFG Spectroscopy at Liquid–Gas and Liquid–Liquid Interfaces 429 26.5 Conclusion 430 References 430 27 27.1 27.1.1 27.2
Other Photon-Detecting Techniques 437 John C. Rivière Appearance Potential Methods 437 Soft X-Ray Appearance Potential Spectroscopy (SXAPS) 437 Inverse Photoemission Spectroscopy (IPES) and Bremsstrahlung Isochromat Spectroscopy (BIS) 437
Part Four
Scanning Probe Microscopy 439
28
Introduction 441 Gernot Friedbacher References 442
29
Atomic Force Microscopy (AFM) 443 Gernot Friedbacher Principles 443 Further Modes of AFM Operation 446 Friction Force Microscopy (FFM) 446 Young’s Modulus Microscopy (YMM) or Force Modulation Microscopy (FMM) 447 Phase Imaging 447 Force–Distance Curve Measurements 447 Pulsed Force Mode AFM 448 Harmonic Imaging and Torsional Resonance Mode 449 Instrumentation 452 Applications 455 References 462
29.1 29.2 29.2.1 29.2.2 29.2.3 29.2.4 29.2.5 29.2.6 29.3 29.4
30 30.1 30.2 30.3 30.4
Scanning Tunneling Microscopy (STM) 465 Gernot Friedbacher Principles 465 Instrumentation 467 Lateral and Spectroscopic Information 468 Applications 470 References 479
Contents
31
Scanning Near-Field Optical Microscopy (SNOM) 481 Marc Richter and Volker Deckert 31.1 Introduction 481 31.2 Instrumentation and Operation 482 31.2.1 Basic Set-Up 482 31.2.2 Variations of SNOM 483 31.2.3 Scanning and Feedback Techniques 484 31.2.4 Tip Fabrication 485 31.2.4.1 Taper Formation 486 31.2.4.2 Coating Deposition and Aperture Formation 486 31.2.4.3 Advanced Tip Fabrication 487 31.3 SNOM Applications 488 31.3.1 Fluorescence 488 31.3.2 Near-Field Raman Spectroscopy 490 31.3.3 SNOM-IR-Spectroscopy 492 31.4 Outlook 493 References 493
Appendices 499 Appendix A Summary and Comparison of Techniques 501 Appendix B Surface and Thin-Film Analytical Equipment Suppliers 507
Index
519
XV
XVII
Preface to the First Edition The surface of a solid interacts with its environment. It may be changed by the surrounding medium either unintentionally (for example, by corrosion) or intentionally due to technological demands. Intentional changes are made in order to refine or protect surfaces, that is, to generate new surface properties. Such surface changes can be made, for instance, by ion implantation, deposition of thin films, epitaxially grown layers, and other procedures. In all these cases, it is necessary to analyze the surface, the layer or system of layers, the grain boundaries, or other interfaces in order to control the process which finally meets the technological requirements for a purposefully changed surface. A wealth of analytical methods is available to the analyst, and the choice of the method appropriate for the solution of his problem requires a basic knowledge on the methods, techniques, and procedures of surface and thin film analysis. Therefore, the goal of this book is to give the analyst – whether a newcomer wishing to acquaint with new methods or a materials analyst seeking information on methods that are not available in his own laboratory – a clue about the principles, instrumentation, and applications of the methods, techniques, and procedures of surface and thin film analysis. The first step into this direction was the chapter Surface and Thin Film Analysis of Ullmann’s Encyclopedia of Industrial Chemistry (Vol. B6, Wiley-VCH, Weinheim 2002), in which practitioners give a brief outline of various important methods. The present book is based on that chapter. It has essentially been extended by new sections dealing with electron energy loss spectroscopy (EELS), low-energy electron diffraction (LEED), elastic recoil detection analysis (ERDA), nuclear reaction analysis (NRA), energy dispersive X-ray spectroscopy (EDXS), X-ray diffraction (XRD), surface analysis by laser ablation (LA), and ion-beam spectrochemical analysis (IBSCA). Thus, the book now comprises the most important methods and should help the analyst to make decisions on the proper choice of methods for a given problem. Except for atomic force microscopy (AFM) and scanning tunneling microscopy (STM), microscopic methods, as essential as they are for the characterization of surfaces, are only briefly discussed when combined with a spectroscopic method. Methods of only limited importance for the solution of very special problems, or without availability of commercial equipment, are not considered or
XVIII
Preface to the First Edition
only briefly mentioned in the sections entitled Other Electron/Ion/Photon Detecting Techniques. Furthermore, the objective was not to issue a voluminous book, but a clearly arranged one outlining the basic principles and major applications of important methods of surface and thin film analysis. For more detailed information on any of these topics, the reader is referred to the special literature given in the references. The editors are gratefully indebted to all contributors who were ready to redirect time from their research, educational, and private activities in order to contribute to this book. They also wish to thank Mrs Silke Kittel for her tireless help in developing our editorial ideas. Autumn 2001
Henning Bubert Holger Jenett
XIX
Preface to the Second Edition The first edition of this book was very well received on the market and, after becoming “out-of-print”, a variety of ideas was discussed to produce a second edition. It became clear to us very quickly that, instead of an unchanged reprint of the first edition, the opportunity should be taken to update the information in the book and to add new chapters based on feedback from our readers. Fortunately, all authors of the first edition immediately supported this idea, though some were no longer available to actively contribute to the revisions due to changes in their professional careers. Almost all chapters of this book have been thoroughly revised, taking into consideration new developments on the described methods as well as valuable feedback from the First Edition. Although a complete collection of surface analytical techniques would be beyond the scope of a compendium such as this, new chapters on field ion microscopy (FIM) and atom probe (AP), sum frequency generation (SFG), and scanning near-field optical microscopy (SNOM) have been added. With regard to Appendix B the point must be addressed that, due to a rapidly changing market that is characterized by the frequent takeover of one company (or of their subsidiaries) by another, it became rather difficult to produce a compilation that was fully consistent with regard to the names of brands, branches, and company owners. However, the given internet addresses should serve to guide readers to the desired information and contacts to their local distributors. The editors would like to thank all authors for revising and updating their chapters from the First Edition of the book, and all new authors for writing the new chapters and for revising some of the chapters already in existence. To those authors who were unable to revise their chapters themselves, we are certainly indebted that they agreed to a revision of their chapters by new authors. Without this consent between “old” and “new” authors the revision of this book would not have been possible. Finally, we would like to thank Dr. Manfred Köhl and Mrs. Lesley Belfit from Wiley-VCH for their continued support to move this book project forward, as well as Mrs. Bernadette Cabo for the helpful and pleasant communication during the production process. April 2011
Gernot Friedbacher Henning Bubert
XXI
List of Contributors Heinrich F. Arlinghaus Westfälische Wilhelms-Universität Münster Physikalisches Institut Wilhelm-Klemm-Str. 10 48149 Münster Germany
Oswald Benka Johannes Kepler Universität Linz Institut für Experimentalphysik Altenbergerstr. 69 4040 Linz Austria
Athula Bandara University of Peradeniya Department of Chemistry Peradeniya (20400) Sri Lanka
Michail Bolshov Russian Academy of Sciences Institute of Spectroscopy Fizicheskaja street 5 142092 Troitsk, Moscow Region Russia
Peter Bauer Johannes Kepler Universität Linz Institut für Experimentalphysik Altenbergerstr. 69 4020 Linz Austria
Henning Bubert Leibniz-Institut für Analytische Wissenschaften - ISAS - e.V. Otto-Hahn-Str. 6b 44227 Dortmund Germany
Burkhard Beckhoff Physikalisch-Technische Bundesanstalt (PTB) X-ray Spectrometry Abbestr. 2-12 10587 Berlin Germany
Volker Deckert Friedrich Schiller Universität Jena Institut für Physikalische Chemie Helmholtzweg 4 07743 Jena Germany and IPHT Institut für Photonische Technologien e.V. Albert-Einstein-Str. 9 07745 Jena Germany
XXII
List of Contributors
Wolfgang Drachsel Technische Universität Wien Institut für Materialchemie Getreidemarkt 9 1060 Wien Austria
Roland Hergenröder Leibniz-Institut für Analytische Wissenschaften - ISAS - e.V. Otto-Hahn-Str. 6b 44227 Dortmund Germany
Laszlo Fabry Wacker Chemie AG Johannes-Hess-Str. 24 84489 Burghausen Germany
Wieland Hill LIOS Technology GmbH Schanzenstr. 39 51063 Köln Germany
Gernot Friedbacher Technische Universität Wien Institut für Chemische Technologien und Analytik Getreidemarkt 9/164-IAC 1060 Wien Austria
Karsten Hinrichs Leibniz-Institut für Analytische Wissenschaften - ISAS - e.V. Department Berlin Albert-Einstein-Str. 9 12489 Berlin Germany
P. Neil Gibson European Commission – Joint Research Centre Institute for Health and Consumer Protection TP 500 21027 Ispra, VA Italy
Volker Hoffmann Leibniz-Institut für Festkörper- und Werkstoffforschung IFW Institut für Komplexe Materialien Helmholtzstr. 20 01069 Dresden Germany
Bernd Gruska SENTECH Instruments GmbH Schwarzschildstr. 2 12489 Berlin Germany
Herbert Hutter Technische Universität Wien Institut für Chemische Technologien und Analytik Getreidemarkt 9/164-IAC 1060 Wien Austria
Georg Held University of Reading Department of Chemistry Whiteknights P.O. Box 224 Reading, Berkshire RG6 6AD UK
Holger Jenett Albrecht-Dürer-Gymnasium Heinitzstr. 73 58097 Hagen Germany
List of Contributors
Michael Kopnarski Institut für Oberflächen- und Schichtanalytik IFOS GmbH Trippstadter Str. 120 67663 Kaiserslautern Germany Bernhard Lendl Technische Universität Wien Institut für Chemische Technologien und Analytik Getreidemarkt 9/164-UPA 1060 Wien Austria Siegfried Pahlke Analytical Consulting Keltenstr. 7 84375 Kirchdorf am Inn Germany Leopold Palmetshofer Johannes Kepler Universität Linz Institut für Halbleiter- und Festkörperphysik Altenbergerstr. 69 4040 Linz Austria Alfred Quentmeier Zum Paradies 2a 34516 Vöhl Germany Marc Richter IPHT Institut für Photonische Technologien e.V. Albert-Einstein-Str. 9 07745 Jena Germany
John C. Rivière Oxford University Begbroke Science Park (OUBSP) Department of Materials Sandy Lane Yarnton Kidlington OX5 1PF UK Volker Rupertus SCHOTT AG Corporate Research & Technology Development Process Technology and Characterization Hattenbergstr. 10 55122 Mainz Germany Günther Rupprechter Technische Universität Wien Institut für Materialchemie Getreidemarkt 9 1060 Wien Austria Reinhard Schneider Karlsruher Institut für Technologie (KIT) Laboratorium für Elektronenmikroskopie Engesserstr. 7 76131 Karlsruhe Germany Yuri Suchorski Technische Universität Wien Institut für Materialchemie Getreidemarkt 9 1060 Wien Austria
XXIII
XXIV
List of Contributors
Wolfgang S.M. Werner Technische Universität Wien Institut für Angewandte Physik Wiedner Hauptstr. 8 1040 Wien Austria
1
1 Introduction John C. Rivière and Henning Bubert
Wherever the properties of a solid surface are important, it is also important to have the means to measure those properties. The surfaces of solids play an overriding part in a remarkably large number of processes, phenomena, and materials of technological importance. These include: catalysis; corrosion, passivation, and rusting; adhesion; tribology, friction, and wear; brittle fracture of metals and ceramics; microelectronics; composites; surface treatments of polymers and plastics; protective coatings; superconductors; and solid-surface reactions of all types with gases, liquids, or other solids. The surfaces in question are not always external; processes occurring at inner surfaces such as interfaces and grain boundaries are often just as critical to the behavior of the material. In all of the above examples, the nature of a process or of the behavior of a material can be understood completely only if information about both the surface composition (i.e., the types of atoms present and their concentrations) and the surface chemistry (i.e., the chemical states of the atoms) is available. Furthermore, knowledge of the arrangement of surface atoms (i.e., the surface structure) is also necessary. First of all, what is meant by a solid surface? Ideally, the surface should be defined as the plane at which the solid terminates – that is, the last atom layer before the adjacent phase (vacuum, vapor, liquid, or another solid) begins. Unfortunately such a definition is impractical, because the effect of termination extends into the solid beyond the outermost atom layer. Indeed, the current definition is based on that knowledge, and the surface is thus regarded as consisting of that number of atom layers over which the effect of termination of the solid decays until bulk properties are reached. In practice, this decay distance is of the order of 5–20 nm. By a fortunate coincidence, the depth into the solid from which information is provided by the techniques described here matches the above definition of a surface in many cases. These techniques are, therefore, surface-specific; in other words, the information they provide comes only from that very shallow depth of a few atom layers. Other techniques can be surface-sensitive, in that they would normally be regarded as techniques for bulk analysis, but have sufficient sensitivity for certain elements that can be analyzed only if they are present on the surface. Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
2
1 Introduction
Why should surfaces be so important? The answer is twofold. First, the properties of surface atoms are usually different from those of the same atoms in the bulk; and second, because in any interaction of a solid with another phase the surface atoms are the first to be encountered. Even at the surface of a perfect single crystal the surface atoms behave differently from those in the bulk, simply because they do not have the same number of nearest neighbors; their electronic distributions are altered, and hence their reactivity. Their structural arrangement is often also different. When the surface of a polycrystalline or glassy multielemental solid is considered – such as that of an alloy or a chemical compound – the situation can be very complex. The processes of preparation or fabrication can produce a material, the surface composition of which is quite different from that of the bulk, in terms of both constituent and impurity elements. Subsequent treatment (e.g., thermal and chemical) will almost certainly change the surface composition to something different again. The surface is highly unlikely to be smooth, and roughness at both the micro and macro level can be present, leading to the likelihood that many surface atoms will be situated at corners and edges and on protuberances (i.e., in positions of increased reactivity). Surfaces exposed to the atmosphere, which include many of those of technological interest, will acquire a contaminant layer that is one to two atom layers thick, containing principally carbon and oxygen but also other impurities present in the local environment. Atmospheric exposure might also cause oxidation. Because of all these possibilities, the surface region must be considered as a separate entity, effectively a separate quasi-two-dimensional (2-D) phase overlaying the normal bulk phase. Analysis of the properties of such a quasi phase necessitates the use of techniques in which the information provided originates only or largely within the phase – that is, the surface-specific techniques described in this volume. Nearly all these techniques involve interrogation of the surface with a particle probe. The function of the probe is to excite surface atoms into states giving rise to the emission of one or more of a variety of secondary particles such as electrons, photons, ions, and neutrals. Since the primary particles used in the probing beam can also be electrons or photons, or ions or neutrals, many separate techniques are possible, each based on a different primary–secondary particle combination. Most of these possibilities have now been established, but in fact not all the resulting techniques are of general application – some due to the restricted or specialized nature of the information obtained, and others due to difficult experimental requirements. In this book, therefore, most space is devoted to those surface analytical techniques that are widely applied and readily available commercially, whereas much briefer descriptions are provided of some others, the use of which is less common but which – under appropriate circumstances, particularly in basic research – can provide vital information. Since the various types of particle can appear in both primary excitation and secondary emission, most authors and reviewers have found it convenient to group the techniques in a matrix, in which the rows refer to the nature of the exciting particle and the columns to the nature of the emitted particle. Such a matrix of techniques is provided in Table 1.1, which uses widely accepted acronyms. The
1 Introduction Table 1.1 Surface-specific analytical techniques using particle or photon excitation. The acronyms (see Listing 1.1) printed in bold are those used for methods discussed in more detail in this book.
Excitationa)
Detection
Ions, neutrals, A+, A−, A0
−
Electrons, e e−
AES EELS EFTEM LEED
SAM
A+, A−, A0
ESD
ESDIAD
hv
EDXS SXAPS IPES
Photons, hv
IAES INS
XPS
UPS
TXRF XRD LA RAIRS SHG ELL SNOM
LIBS SERS SFG
RHEED SIMS GDMS RBS ERDA
SNMS FABMS LEIS NRA
GD-OES IBSCA BIS
a) Some of the techniques in Table 1.1 have angle-resolved variants, with the prefix AR (e.g., ARUPS), or use Fourier-transform methods, with the prefix FT (e.g., FT-RAIRS). Table 1.2
Surface-specific analytical techniques using non-particle excitation.
Detection
Excitation Heat, kT
High electrical field, F
A+
TDS
FIM
A−
TDS
e− (Displacement)
Mechanical force
AP
IETS STM, STS AFM
meanings of the acronyms, together with some of the alternatives that have appeared in the literature, are provided in Listing 1.1. A few techniques cannot be classified according to the nature of the exciting particle, because they do not employ primary particles but depend instead on the application either of heat or a high electric field. These techniques are listed in Table 1.2.
3
4
1 Introduction
Listing 1.1. Meanings of the surface analysis acronyms, and their alternatives, that appear in Tables 1.1. and 1.2. 1.
Electron Excitation AES, Auger electron spectroscopy BIS, Bremsstrahlung isochromat spectroscopy (or ILS, ionization loss spectroscopy) EDXS, Energy-dispersive X-ray spectroscopy EELS, Electron energy loss spectroscopy EFTEM, Energy-filtered transmission electron microscopy ESD, Electron-stimulated desorption (or EID, electron-induced desorption) ESDIAD, Electron-stimulated desorption ion angular distribution IPES, Inverse photoemission spectroscopy LEED, Low-energy electron diffraction RHEED, Reflection high-energy electron diffraction SXAPS, Soft X-ray appearance potential spectroscopy (or APS, appearance potential spectroscopy) SAM, Scanning Auger microscopy
2. Ion Excitation ERDA, Elastic recoil detection analysis GDMS, Glow discharge mass spectrometry GD-OES, Glow discharge optical emission spectroscopy IAES, Ion (excited) Auger electron spectroscopy IBSCA, Ion beam spectrochemical analysis (or SCANIIR, surface composition by analysis of neutral and ion impact radiation or BLE, bombardmentinduced light emission) INS, Ion neutralization spectroscopy LEIS, Low-energy ion scattering (or ISS, Ion-scattering spectroscopy) NRA, Nuclear reaction analysis RBS, Rutherford back-scattering spectroscopy (or HEIS, high-energy ion scattering) SIMS, Secondary-ion mass spectrometry (SSIMS, static secondary-ion mass spectrometry) (DSIMS, dynamic secondary-ion mass spectrometry)
1 Introduction
SNMS, Secondary neutral mass spectrometry 3. Photon Excitation ELL, Ellipsometry LA, Laser ablation LIBS, Laser-induced breakdown spectroscopy (or LIPS, Laser-induced plasma spectroscopy) RAIRS, Reflection–absorption infrared spectroscopy (or IRRAS, infrared reflection–absorption spectroscopy, or IRAS, infrared absorption spectroscopy, or ERIRS, external reflection infrared spectroscopy) SERS, Surface-enhanced Raman scattering SFG, Sum frequency generation SHG, (optical) Second harmonic generation SNOM, Scanning near-field optical microscopy TXRF, Total reflection X-ray fluorescence analysis UPS, Ultraviolet photoelectron spectroscopy XPS, X-ray photoelectron spectroscopy (or ESCA, electron spectroscopy for chemical analysis) XRD, X-ray diffraction 4. Neutral Excitation FABMS, Fast-atom bombardment mass spectrometry 5. Thermal Excitation TDS, Thermal desorption spectroscopy 6. High-Field Excitation AP, Atom probe FIM, Field ion microscopy IETS, Inelastic electron tunneling spectroscopy STM, Scanning tunneling microscopy STS, Scanning tunneling spectroscopy 7. Mechanical Force AFM, Atomic force microscopy
5
7
Part One Electron Detection
9
2 X-Ray Photoelectron Spectroscopy (XPS) Henning Bubert, John C. Rivière, and Wolfgang S.M. Werner
X-ray photoelectron spectroscopy (XPS) is one of the most widely used surface analytical techniques, and is therefore described here in more detail than any of the other techniques. At its inception by Siegbahn and coworkers [1], it was called ESCA (electron spectroscopy for chemical analysis), but the name ESCA is now considered too general, as many surface electron spectroscopies exist, and the name given to each one must be precise. Nevertheless, the name ESCA is still used in many places, particularly in industrial laboratories and their publications. Briefly, the reasons for the popularity of XPS are the exceptional combination of compositional and chemical information that it provides, its ease of operation, and the ready availability of commercial equipment.
2.1 Principles
The surface to be analyzed is irradiated with soft X-ray photons. When a photon of energy hν interacts with an electron in a level with binding energy EB (EB is the energy EK of the K-shell in Figure 2.1), the entire photon energy is transferred to the electron, with the result that a photoelectron is ejected with kinetic energy E kin = hv − EB − ΦS
(2.1)
where ΦS is a small, almost constant, work-function term. Obviously, hν must be greater than EB. The ejected electron may come from a core level or from the occupied portion of the valence band, but in XPS most attention is focused on the electrons in core levels. As no two elements share the same set of electronic binding energies, measurement of the photoelectron kinetic energies enables elemental analysis. In addition, Equation 2.1 indicates that any changes in EB are reflected in Ekin, which means that changes in the chemical environment of an atom can be followed by monitoring changes in the photoelectron energies, leading to the provision of chemical information. In principle, XPS can be used to analyze all elements in the Periodic Table; however, in general Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
10
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.1 Schematic diagram of electron emission processes in solids. Left: Auger process; Right: photoelectron emission process. Electrons involved in the emission processes are indicated by open circles.
hydrogen and helium cannot be detected due to the low cross-section of interaction. Although XPS is concerned principally with photoelectrons and their kinetic energies, the ejection of electrons by other processes also occurs. An ejected photoelectron leaves behind a core hole in the atom. The sequence of events following the creation of the core hole is shown schematically in Figure 2.1. In the example, the hole has been created in the K shell, giving rise to a photoelectron whose kinetic energy would be (hν − EK). This hole can be filled, for example, by an electronic transition from the L1 shell (Figure 2.1, left). The energy EK − EL1 associated with the transition can then either be dissipated as a characteristic X-ray photon or given up to an electron in the same or a higher shell, in this example the L23 shell. The second of these possibilities is called the Auger process after its discoverer [2], and the resulting ejected electron is called an Auger electron and has an energy given by E kin (KL1L23 ) = EK − EL1 − EL23 − E inter (L1L23 ) + ER − ΦS
(2.2)
where Einter(L1L23) is the interaction energy between the holes in the L1 and L23 shells, and ER is the sum of the intra- and extra-atomic relaxation energies. X-ray photon emission (i.e., X-ray fluorescence) and Auger electron emission are obviously competing processes, but for the shallow core levels involved in XPS and Auger electron spectroscopy (AES) the Auger process is far more likely. Thus, in all X-ray photoelectron spectra, features appear due to both photoemission and Auger emission. In XPS, the Auger features can be useful but are not central to the technique, whereas in AES (see Chapter 3 and Equation 2.2) they form the basis of the technique. At this point, the nomenclature used in XPS and AES should be explained. In XPS, the spectroscopic notation is used, and in AES the X-ray notation. The two are equivalent, the different usages having arisen for historical reasons, but the differentiation is a convenient one. They are both based on the so-called j − j coupling
2.1 Principles Table 2.1
Spectroscopic and X-ray notations.
Quantum numbers n 1 2 2 2 3 3 3 3 3
l 0 0 1 1 0 1 1 2 2 etc.
j 1/2 1/2 1/2 3/2 1/2 1/2 3/2 3/2 5/2
Spectroscopic state
X-ray state
1s 2s 2p1/2 2p3/2 3s 3p1/2 3p3/2 3d3/2 3d5/2 etc.
K L1 L2 L3 M1 M2 M3 M4 M5 etc.
scheme describing the orbital motion of an electron around an atomic nucleus, in which the total angular momentum of an electron is found by summing vectorially the individual electron spin and angular momenta. Thus, if l is the electronic angular momentum quantum number and s the electronic spin momentum quantum number, the total angular momentum for each electron is given by j = l + s. Since l can take the values 0, 1, 2, 3, 4, … and s = ± 21 , clearly j = 21 , 23 , 25 , etc. The principal quantum number n can take values 1, 2, 3, 4, … . In spectroscopic notation, states with l = 0, 1, 2, 3, … are designated s, p, d, f, … , respectively, and the letter is preceded by the number n; the j values are then appended as suffixes. Therefore, one obtains 1s, 2s, 2p1/2, 2p3/2, 3s, 3p1/2, 3p3/2, etc. In X-ray notation, states with n = 1, 2, 3, 4, … are designated K, L, M, N, … , respectively, and states with various combinations of l = 0, 1, 2, 3, … and j = 21 , 23 , 25 are appended as the suffixes 1, 2, 3, 4, … . In this way, one arrives at K, L1, L2, L3, M1, M2, M3, etc. The equivalence of the two notations is set out in Table 2.1. In X-ray notation, the Auger transition shown in Figure 2.1 would therefore be labeled KL1L23. In this coupling scheme, six Auger transitions would be possible in the KLL series. Obviously, many other series are possible (e.g., KLM, LMM, MNN). These are discussed more fully in Chapter 3. The reasons why techniques such as XPS and AES, which involve measurement of the energies of ejected electrons, are so surface-specific, should be examined. An electron with kinetic energy E moving through a solid matrix M has a probability of traveling a certain distance before losing all or part of its energy as a result of an inelastic collision. Based on that probability, the average distance traveled before such a collision is known as the inelastic mean free path (IMFP) λM(E). The IMFP is a function only of M and of E. Figure 2.2 shows a compilation of measurements of λ made by Seah and Dench [3], in terms of atomic monolayers as a function of kinetic energy. Note that both λ and energy scales are logarithmic. The important consequence of the dependence of λ on kinetic energy is that in the ranges of secondary electron kinetic energies used in XPS and AES, the values of
11
12
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.2 Compilation by Seah and Dench [3] of measurements of inelastic mean free path as a function of electron kinetic energy. The solid line is a least-squares fit.
λ are very small. In XPS, for example, typical energy ranges are 250–1500 eV, corresponding to a range of λ from about four to eight monolayers, while in AES, the energy range is typically 20 to 1000 eV, in which case λ would range from about two to six monolayers. What this means in practice is that if the photoelectron or the Auger electron is to escape into a vacuum and be detected, it must originate at or very near the surface of the solid. This is the reason why the electron spectroscopic techniques are surface-specific. Furthermore, the inelastic mean free path is of paramount importance for quantifying XPS-data. Presently, it is most commonly accepted that the calculations, and in particular the semiempirical formula for the IMFP, by Tanuma, Powell, and Penn, constitute a reliable source for this quantity [4].
2.2 Instrumentation 2.2.1 Vacuum Requirements
Electron spectroscopic techniques require a vacuum on the order of 10−8 Pa for their operation. This requirement arises from the extreme surface specificity of these techniques, as discussed in Section 2.1. With sampling depths of only a few atom layers, and elemental sensitivities down to 10−5 atom layers (i.e., one atom of a particular element in 105 other atoms in an atomic layer), the techniques are clearly very sensitive to surface contamination, most of which comes from the residual gases in the vacuum system. According to gas kinetic theory, for sufficient
2.2 Instrumentation
time to be available to make a surface-analytical measurement on a surface that has just been prepared or exposed, before contamination from the gas phase interferes, the base pressure should be 10−8 Pa or lower – that is, in the region of ultrahigh vacuum (UHV). The requirement for UHV conditions imposes restrictions on the types of material that can be used for the construction of surface analytical systems, or inside the systems, because UHV can be achieved only by accelerating the rate of removal of gas molecules from internal surfaces by raising the temperature of the entire system (i.e., by baking). Typical baking conditions are 150–200 °C for several hours. Inside the system, any material is permissible that does not produce volatile components either during normal operation or during baking. For example, brass (which contains the volatile metal zinc) cannot be used. The principal construction material is stainless steel, with mu-metal (76% Ni, 5% Cu, 2% Cr) used occasionally where magnetic screening is needed (e.g., around electron energy analyzers). For the same reasons, metal seals – not elastomers – are used for the demountable joints between individual components; the sealing material normally used is pure copper, although gold is sometimes employed. Other materials that may be used between ambient atmosphere and UHV are borosilicate glass or quartz for windows, and alumina for electrical insulation for current or voltage connections. 2.2.2 X-Ray Sources
The most important consideration in choosing an X-ray source for XPS is energy resolution. Equation 2.1 gives the relationship between the kinetic energy of the photoelectron, the energy of the X-ray photon, and the binding energy of the core electron. Since the energy spread – or linewidth – of an electron in a core level is very small, the linewidth of the photoelectron energy depends on the linewidth of the source, if no undue broadening is introduced instrumentally. In XPS, the analyst devotes much effort to extracting chemical information by means of detailed study of individual elemental photoelectron spectra. Such a study needs an energy resolution better than 1.0 eV, if subtle chemical effects are to be identified. Thus, the linewidth of the X-ray source should be significantly smaller than 1.0 eV, if the resolution required is not to be limited by the source itself. Other considerations are that the source material – which forms a target for high-energy electron bombardment leading to the production of X-rays – should be a good conductor to allow the rapid removal of heat, and it should also be compatible with UHV. Table 2.2 lists the energies and linewidths of the characteristic X-ray lines from a few possible candidate materials. In practice, MgKα and AlKα are the two used universally, due to their line energies and width and their simple use as anode materials. For the efficient production of X-rays by electron bombardment, exciting electron energies that are at least an order of magnitude higher than the line energies
13
14
2 X-Ray Photoelectron Spectroscopy (XPS) Table 2.2 Energies and linewidths of some characteristic low-energy X-ray lines.
Line
Energy (eV)
Width (eV)
YMζ ZrMζ NbMζ MgKα AlKα SiKα YLα ZrLα
132.3 151.4 171.4 1253.6 1486.6 1739.5 1922.6 2042.4
0.47 0.77 1.21 0.70 0.85 1.00 1.50 1.70
Table 2.3 Satellite lines of magnesium and aluminum [5].
X-ray line
Kα′ Kα3 Kα4 Kα5 Kα6 Kβ
Mg
Al
Separation from Kα1,2 (eV)
Relative intensity, % (Kα1,2 = 100%)
Separation from Kα1,2 (eV)
Relative intensity, % (Kα1,2 = 100%)
4.5 8.4 10.0 17.3 20.5 48.0
1.0 9.2 5.1 0.8 0.5 2.0
5.6 9.6 11.5 19.8 23.4 70.0
1.0 7.8 3.3 0.4 0.3 2.0
must be used, so that in Mg and Al sources accelerating potentials of 15 kV are employed. Modern sources are designed with dual anodes; in this case, one anode face is coated with magnesium and the other with aluminum, and with two filaments, one for each face. Thus, a switch from one type of X-irradiation to the other can be made very quickly. To protect the sample from stray electrons from the anode, from heating effects, and from possible contamination by the source enclosure, a thin (ca. 2 μm) window of aluminum foil is interposed between the anode and the sample. For optimum X-ray photon flux on the surface (i.e., optimum sensitivity), the anode must be brought as close to the sample as possible, which means in practice a distance of ≈2 cm. The entire X-ray source is therefore retractable by means of a bellows and a screw mechanism. The X-radiation from magnesium and aluminum sources is quite complex. The principal Kα lines are in fact unresolved doublets and should correctly be labeled Kα1,2. Besides the Kα1,2 lines, a series of further lines – so-called “satellite lines” – also exist, of which the more important ones are Kα3,4. The energy separations of the satellite lines for Mg and Al and their intensities relative to Kα1,2 are listed in Table 2.3.
2.2 Instrumentation
Figure 2.3 Schematic of X-ray monochromatization to remove satellites (S), eliminate
Bremsstrahlung background (B), and separate the Al Kα1,2 doublet. X denotes the characteristic Al Kα only. Illustration courtesy of Kratos Analytical.
The removal of satellites, elimination of the Bremsstrahlung background, and separation of the Kα1,2 doublet can be achieved by monochromatization, as shown schematically in Figure 2.3. The X-ray source is positioned at one point on a spherical surface, called a Rowland sphere, and a quartz crystal is placed at another point. X-rays from the source are diffracted from the quartz, and by placing the sample at the correct point on the Rowland sphere, the Kα1 component can be selectively focused on it. Quartz is a very convenient diffracting medium for AlKα, because the spacing between the 1010 planes is exactly half the wavelength of the X-radiation. Since the width of the AlKα1 line is <0.4 eV, the energy dispersion needed around the surface of the sphere implies that the Rowland sphere should have a diameter of at least 0.5 m. Although an XPS spectrum will be much “cleaner” when a monochromator is used, because satellites and background have been removed, the photon flux at the sample is much lower than that from an unmonochromatized source operating at the same power. Against this must be set the greatly improved signal-to-background level in a monochromatized spectrum. It should also be mentioned that monochromatized X-ray sources open up the possibility of imaging by raster-scanning the electron beam across the anode, which allows scanning of the X-ray spot across the sample surface. This feature is implemented, for example, in the Quantum 2000 instrument developed by Physical Electronics (Eden Prairie, USA) and the Theta Probe instrument developed by Thermo Fisher Scientific Inc. (East Grinstead, UK).
15
16
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.4 Normalized spectrum of photon energies emitted from a synchrotron. λc = wavelength characteristic of the individual synchrotron.
2.2.3 Synchrotron Radiation
The discrete line sources described above for XPS are perfectly adequate for most applications, but some types of analysis require that the source be tunable (i.e., that the exciting energy be variable). The reason for this is to allow the photoionization cross-section of the core levels of a particular element or group of elements to be varied, which is particularly useful when dealing with multielement semiconductors. Tunable radiation can be obtained from a synchrotron. In a synchrotron, electrons are accelerated to near-relativistic velocities and constrained magnetically into circular paths. When a charged particle is accelerated, it emits radiation, and when the near-relativistic electrons are forced into curved paths, they emit photons over a continuous spectrum. The general shape of the spectrum is shown in Figure 2.4. For a synchrotron with an energy of several gigaelectronvolts and a radius of some tens of meters, the energy of the emitted photons near the maximum is of the order of 1 keV (i.e., ideal for XPS). As can be seen from the universal curve, plenty of usable intensity exists down into the ultraviolet (UV) region. With suitable monochromators on the output to select a particular wavelength, the photon energy can be tuned continuously from about 20 to 500 eV. The available intensities are comparable to those from conventional line sources. The main advantage here is the small beam diameter of the synchrotron radiation, which allows a high spatial resolution. Furthermore, polarized radiation for probing electron spin properties in magnetic materials is available. 2.2.4 Electron Energy Analyzers
In electron spectroscopic techniques – among which XPS is the most important – analysis of the energies of electrons ejected from a surface is central. Nowadays, the concentric hemispherical analyzer (CHA) is universally employed.
2.2 Instrumentation
Figure 2.5 Scheme of a concentric hemispherical analyzer [6].
Energy analyzers cannot be discussed without a discussion of energy resolution, which is defined in two ways:
• •
Absolute resolution is defined as ΔE, the full width at half-maximum (FWHM) of a chosen peak. Relative resolution is defined as the ratio R of ΔE to the kinetic energy E of the peak energy position (usually its centroid); that is, R = ΔE/E.
Thus, absolute resolution is independent of peak position, but relative resolution can be specified only by reference to a particular kinetic energy. In XPS, closely spaced peaks at any point in the energy range must be resolved, and this requires the same absolute resolution at all energies. The CHA is shown in schematic cross-section in Figure 2.5 [6]. Two hemispheres of radii r1 (inner) and r2 (outer) are positioned concentrically. Potentials −V1 and −V2 are applied to the inner and outer hemispheres, respectively, where V2 is greater than V1. The source S and the focus F are in the same plane as the center of curvature, and r0 is the radius of the equipotential surface between the hemispheres. If electrons of energy E = eV0 are injected at S along the equipotential surface, they will be focused at F if V2 − V1 = V0 (r2 r1 − r1 r2 )
(2.3)
If electrons are injected not exactly along the equipotential surface, but with an angular spread Δα about the correct direction, then the energy resolution is given by ΔE E = (wS + wF ) 4r0 + (δα)2
(2.4)
where wS and wF are the respective widths of the entrance and exit slits. In most instruments, for convenience in construction, wS = wF = w, whereupon the resolution becomes ΔE E = w 2r0 + (δα)2
(2.5)
In XPS, the photoelectrons are retarded to a constant energy, called the pass energy, as they approach the entrance slit. If this were not done, Equation 2.5 shows that to achieve an absolute resolution of 1 eV at the maximum kinetic energy of about 1500 eV (with AlKα radiation), and with a slit width of 2 mm, would require an analyzer with an average radius of about 300 cm, which is impracticable. Pass
17
18
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.6 Typical configuration in an XPS spectrometer.
energies are selected in the range 20–100 eV for XPS, which allows the analyzer to be built with a radius of 10–15 cm. Modern XPS spectrometers employ a lens system on the input to the CHA, which transfers an image of the analyzed area on the sample surface to the entrance slit of the analyzer. The detector system on the output of the CHA consists of several single channeltrons or a channel plate. Such a spectrometer is shown schematically in Figure 2.6. Besides imaging capabilities, this arrangement also opens up the possibility of simultaneous detection of photoelectrons emitted from the sample at various angles – that is, originating from different depths depending on the inelastic mean free path. This option for depth profiling is frequently called parallel angle-resolved XPS (indicating parallel data acquisition for various angles); this is in contrast to conventional angle-resolved XPS, where the emission angle of the detected photoelectrons is changed sequentially during analysis by tilting the sample. An example of such an arrangement is that of the Theta Probe instrument developed by Thermo Fisher Scientific Inc. (East Grinstead, UK). 2.2.5 Spatial Resolution
The principal disadvantage of conventional XPS was the lack of spatial resolution; the spectral information was derived from an analyzed area of several square millimeters, and was therefore an average of the compositional and chemical analyzes from that area. However, many technological samples are inhomogeneous on a scale much smaller than that of conventional XPS analysis, and obtaining chemical information on the same scale as the inhomogeneities would be very desirable.
2.3 Spectral Information and Chemical Shifts
Improved XPS spatial resolution has been achieved by using the focusing properties of the CHA described above, and this led to the development of the ESCASCOPE, which was further developed into the ESCALAB 250 (Thermo Fisher Scientific Inc. [formerly VG Scientific], East Grinstead, UK), which acts as an XPS microscope. Here, a Fourier-transformed image is generated at the entrance slit of the CHA by introducing a third lens into the transfer lens system. The CHA disperses this image energetically, so that a Fourier-transformed image then also exists at the exit slit, but only for the energy selected by the setting of the CHA. This image is then inverted into a real image, which can be detected in the plane of a channel plate, allowing a spatial resolution of about 3 μm to be obtained. A further development using the focusing properties of the transfer lens system has been realized in the AXIS instruments (Kratos Analytical, Manchester, UK). Here, the principal plane of the first lens is moved towards the specimen by introducing a magnetic lens below the specimen holder. The spatial resolution is improved to about 15 μm without any loss of intensity, due to the simultaneous enhancement of the angle of acceptance. Such a magnetic lens has also been introduced into the Escalab instrument, from Thermo Fisher Scientific Inc.. A third development is that of the Quantum 2000 (Physical Electronics, Eden Prairie, USA). In this variation, high-energy electrons from an electron gun are scanned over the surface of an Al anode, thereby generating a scanning X-ray beam. This beam is then focused onto the specimen surface by an ellipsoidal monochromator and scans the specimen surface with the same frequency as that of the electron beam. The X-ray beam is approximately 10 μm in diameter, and the scanned area is 1.4 × 1.4 mm2. The thrust of development is toward ever-better spatial resolution, which means the smallest possible spot size on the sample compatible with adequate signal-tonoise ratio (SNR), for acquisition within a reasonable length of time.
2.3 Spectral Information and Chemical Shifts
Figure 2.7 shows a wide-scan or survey XPS spectrum – that is, one recorded over a wide range of energies (in this case, 1000 eV). The radiation used was unmonochromatized AlKα, at 1486.6 eV, and the surface is that of almost clean copper. Whilst such a spectrum reveals the major, or primary, features to be found, in order to investigate minor or more detailed features the spectra are acquired over much more restricted energy ranges, and at better energy resolution; the latter are termed narrow-scan spectra. The primary features in Figure 2.7 are peaks arising from excitation of core level electrons according to Equation 2.1. At the low-energy end, two intense peaks are found at about 553 and 533 eV, corresponding to photoelectrons from the 2p3/2 and 2p1/2 levels of copper, respectively, the separation of 20 eV being the spin-orbit splitting. At the high kinetic energy end, there are three other copper photoelectron peaks at about 1363, 1409, and 1485 eV, which arise from the 3s, 3p,
19
20
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.7 Wide-scan spectrum from almost clean copper, recorded with AlKα radiation.
and 3d levels, respectively. The other primary features associated with copper are the three Auger peaks L3M4,5M4,5, L3M2,3M4,5, and L3M2,3M2,3 at 919, 838, and 768 eV, respectively. As pointed out in Section 2.1, the creation of a core hole by any means of excitation can lead to ejection of an Auger electron so that, in XPS, Auger features make a significant contribution to the spectrum. If Auger and photoelectron peaks happen to overlap in any spectrum, they can always be separated by changing the excitation (e.g., from AlKα to MgKα, or vice versa), because the Auger peaks are invariant in energy (Equation 2.2), whereas the photoelectron peaks must shift with the energy of the exciting photons according to Equation 2.1. In addition to primary features due to copper in Figure 2.7, there are small photoelectron peaks at kinetic energies of 955 and 1204 eV that arise from the oxygen and carbon 1s levels, respectively, due to the presence of some contamination on the surface. Secondary features are X-ray satellite and ghost lines, surface and bulk plasmon energy loss features, shake-up lines, multiplet splitting, shakeoff lines and asymmetries due to asymmetric core levels [7]. Chemical shift is the observed shift in energy of a photoelectron peak from a particular element when the chemical state of that element changes. When an atom enters into combination with another atom or group of atoms, an alteration occurs in the valence electron density, which can be positive or negative according to whether charge is accepted or donated, and this causes a consequent alteration in the electrostatic potential affecting the core electrons. Therefore, the binding energies of the core electrons change, giving rise (according to Equation 2.1) to shifts in the corresponding photoelectron peaks. Tabulation of the chemical shifts experienced by any one element in a series of pure compounds of that element thus enables its chemical state to be identified during analysis of unknown samples. Many such tabulations have appeared, and a major collection of them can be found in Ref. [8] and the NIST X-ray photoelectron spectroscopy database [9]. The identification of chemical states in this way is the principal advantage of XPS over other surface analytical techniques.
2.4 Quantification, Depth Profiling, and Imaging
Figure 2.8 Example of a chemical shift in the Sn 3d peak, exhibited by a very thin layer of Sn
oxide on Sn metal. (a) Spectrum after linear background subtraction; (b) Spectrum resolved into its respective components. (a) Snn+; (b) Sn0.
An example of a spectrum exhibiting a chemical shift is that of the tin 3d peaks in Figure 2.8. Here, a thin layer of oxide on the metallic tin surface allows photoelectrons from both the underlying metal and the oxide to appear together. Resolution of the doublet 3d5/2, 3d3/2 into their components, from the metal (Sn0) and from the oxide Snn+, is shown in Figure 2.8b; the shift in this case is 1.6–1.7 eV. Curve resolution is an operation that can be performed routinely in data-processing systems associated with photoelectron spectrometers.
2.4 Quantification, Depth Profiling, and Imaging 2.4.1 Quantification
If an X-ray photon of energy hν ionizes a core level X in an atom of element A in a solid, the photoelectron current IA from X in A is
21
22
2 X-Ray Photoelectron Spectroscopy (XPS)
I A ( X ) = Kσ A (hv, X )β A (hv, X )N A λM (E A )cosθ
(2.6)
where σA (hν, X) is the photoelectric cross-section for ionization of X (binding energy EB) by photons of energy hν, βA(hν, X) is the asymmetry parameter for emission from X by excitation with photons of energy hν, N A the atomic density of A averaged over depth of analysis, λM (EA) the inelastic mean free path in matrix M containing A, at kinetic energy EA, where EA = hν − EB, and θ is the angle of emission to the surface normal. K is a constant of proportionality that contains fixed operational parameters such as incident X-ray flux, transmission of the analyzer at kinetic energy EA, and efficiency of the detector at kinetic energy EA, kept fixed during any one analysis. Values of the total cross-section σA for AlKα radiation, relative to the carbon 1s level, have been calculated by Scofield [10] and Trzhaskoskaya et al. [11, 12], and of the asymmetry parameter βA by Reilman et al. [13]. Both, theoretical and semiempirical expressions exist that allow estimates of the required atomic densities to be made conveniently. For the inelastic mean free path (IMFP) (see also Figure 2.2), several such relationships have been reported [3, 4, 14–17]. Tanuma, Powell, and Penn have developed a formula commonly referred to as TPP-2M which is frequently used to estimate the IMFP [4]:
λ i = E {E p2 [β ln(γ E ) − C E + D E 2 ]}
(2.7)
Here, λi is the IMFP, E is the kinetic energy, and E p = N v ρ/M is the generalized plasmon energy (where Nv is the number of valence electrons [per atom or molecule], ρ is the density of the material and M is the atomic mass). The parameters β, γ, C, and D can be determined by fitting IMPF values obtained from linear response theory to Equation 2.7. The following empirical relationships for these quantities are recommended [4]:
β = −0.100 + 0.944(E p2 + Eg2 )−1/2 + 0.069ρ 0.1 γ = 0.191ρ −0.50 C = 1.97 − 0.91U D = 53.4 − 20.8U U = N v ρ/M = E p2 / 829.4 In principle, therefore, the surface concentration of an element can be calculated from the intensity of a particular photoelectron emission, according to Equation 2.6. In practice, the method of relative sensitivity factors is in common use. If spectra were recorded from reference samples of pure elements A and B on the same spectrometer, and the corresponding line intensities are I A∞ and IB∞, respectively, then Equation 2.6 can be written as I A I A∞ ⎛ λM (E A )λB(EB ) ⎞ ⎛ RB∞ ⎞ ⎛ N AN B∞ ⎞ = ⎟ ⎜ ⎟⎜ IB IB∞ ⎜⎝ λM (EB )λ A (E A )⎠⎟ ⎝ RA∞ ⎠ ⎝ N BN A∞ ⎠
(2.8)
where RA∞ and RB∞ are factors for the reference samples, depending on the roughness of the surfaces, and N A∞ and N B∞ are the atomic densities of the reference
2.4 Quantification, Depth Profiling, and Imaging
samples. From the above equation the ratio NA/NB can be obtained from the measured intensity ratio IA/IB. However, the accuracy of quantitative results is only as good as the accuracy of the parameters and of the intensity values IA and IB. These intensities are simply the areas under the line peaks after background subtraction. Three different procedures are commonly in use: straight line subtraction; Shirley’s method; and Tougaard’s method. In the last-mentioned method, the measured spectrum is deconvoluted into the primary (true) excited spectrum by taking into account that the emitted electrons have suffered energy loss due to scattering while traveling through the solid. A further critical point concerns the intensities I∞ obtained from spectra of the pure elements. Calculated and experimentally determined values can diverge considerably, and the best data sets for I∞ measured on pure reference samples still show a scatter of up to 10%. The use of an internal standard or of a simultaneously measured external standard seems to be the most successful way to decrease the inaccuracy to below 10%. (For a more detailed discussion of background subtraction and quantification, see Ref. [18].) 2.4.2 Depth Profiling
Very often, in addition to analytical information about the original surface, information is required about the distribution of chemical composition to depths considerably greater than the inelastic mean free path. Nondestructive methods, such as the variation of emission angle or variation of exciting photon energy are possible, but these are limited in practice to an evaluation of chemical profile within depths of only approximately 5 nm. Obtaining information from greater depths requires destructive methods, of which the most universally used is removal of the surface by ion bombardment, also called “sputtering.” A combination of removal of the surface by sputtering and of analysis by XPS, AES, or SSIMS, is termed depth profiling. Instrumentally, this is performed by means of ion guns which are described elsewhere (see Chapter 7; Section 7.2.1). Depth profiling in XPS is usually performed with a noble gas to minimize any chemical effects. In compounds, preferential sputtering – for example, of O from TiO2 surfaces and of Be from Cu–Be alloys – may occur. This is easily comprehensible in terms of the different surface-binding energies and different momenta transferred to atoms with different masses. Preferential sputtering frequently results in a chemically modified surface, even if the primary ion from the ion gun is not chemically reactive after its (immediate) neutralization in the solid. The depth resolution achievable during profiling depends on many variables, and the reader is referred to the comprehensive discussion in Refs [18, 19]. In XPS, chemical information as a function of depth is acquired comparatively slowly in a stepwise fashion, by using alternate cycles of sputtering and analysis. Examples of profiles through oxide films on pure iron and on Fe–12Cr–1Mo alloy are shown in Figure 2.9, in which the respective contributions from the metallic and oxide components of the iron and chromium spectra have been quantified [20a]. In these examples, the oxide films were only ≈5 nm thick on iron and ≈3 nm
23
24
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.9 Depth profiles of thin (3–5 nm) oxide films [20a] on: (a) Pure iron; (b) Fe–12Cr–
1Mo alloy.
thick on the alloy; in this case, the depth profile sputtering was performed without any sample rotation. It should be mentioned that roughening of the sample can be significantly reduced by rotating the sample during sputtering [20b,c]. In this way, a higher depth resolution and sharper depth profiles can be achieved. Common practice in depth profiling is to use positively charged argon ions at energies between 0.5 and 10 keV, focused into a beam of 2–5 μm diameter, which is then rastered over the area to be profiled. The area bombarded by the primary ion beam, which is typically finely focused and rastered, should be significantly greater than the analyzed area, so that only the flat bottom of the sputtered crater is analyzed. Ion current densities are variable between 5 and 50 μA cm−2. The
2.4 Quantification, Depth Profiling, and Imaging
uniformity of sputtering – and therefore the depth resolution of the profile – can be improved by rotating the sample during bombardment. During the sputtering process, many artifacts are introduced that can affect quantification; some of these derive from the nature of the sample itself (e.g., roughness, crystalline structure, phase distribution, electrical conductivity), while others are radiation-induced (e.g., atomic mixing, preferential sputtering, compound reduction, diffusion and segregation, sputter-induced topography). This is an extensive subject in its own right; a comprehensive discussion is given in Ref. [21]. The state of the art for data evaluation of complex depth profiles is the use of factor analysis. Here, the acquired data can be compiled in a two-dimensional (2-D) data matrix in such a manner that the n intensity values N(E) or, in the derivative mode dN(E)/dE, respectively, of a spectrum recorded in the ith of a total of m sputter cycles are written in the ith column of the data matrix D. For the purpose of factor analysis, it then becomes necessary that the (n × m)-dimensional data matrix D can be expressed as a product of two matrices, that is, the (n × k)dimensional spectrum matrix R and the (k × m)-dimensional concentration matrix C, in which R contains the spectra of k components in k columns, and C contains the concentrations of the respective m sputter cycles in k rows: that is D = R ⋅C + E
(2.9)
Here, E is an error matrix taking errors of measurement (e.g., random noise) into consideration. The term component describes such chemical or physical states whose spectra cannot be generated by a linear combination of the other components. Thus, components can be elements, chemical compounds (stoichiometric or nonstoichiometric) or even states induced by physical processes, provided that the spectra differ significantly, for example, in line shapes or line shifts. Since the introduction of factor analysis for evaluation of depth profile data by Gaarenstroom [22], many reports have been made on the topic [23–30]. Today, this technique of data analysis is a standard tool in most software packages for surface analysis with electron beams. With the help of factor analysis, three results can be achieved directly:
• • •
The number k of relevant components which are necessary to create the data matrix. The so-called abstract spectrum matrix R* and the abstract concentration matrix C*. A new data matrix D*, which can be generated from R* and C* by leaving out the so-called “noise components.”
If the spectra of the k relevant components are known, the abstract matrices can be transformed into interpretable spectrum and concentration matrices, respectively, and as a final result, the required component depth profile is obtained. If one or more of the spectra of the k relevant components is not known, a quantified component depth profile cannot be constructed.
25
26
2 X-Ray Photoelectron Spectroscopy (XPS)
2.4.3 Imaging
As already mentioned in Section 2.2.5, ESCALAB acts as an XPS microscope, when the result is termed an elemental map or image. An example of chemical state imaging is given in Figure 2.10 [31a]. In this case, the material is a contaminated fluoropolymer, the contamination being in the form of dark spots 10–80 μm in size. Both images in Figure 2.10 are based on the C 1s photoelectron peak, but of different chemical states – that is, different binding energies. The upper state is in that contribution arising from carbon bound to fluorine, and the lower state from carbon bound to carbon (i.e., graphitic in nature). The complementarity of the images indicates that the contamination is graphitic in nature. If the analyzer is set to accept electrons of an energy characteristic of a particular element, and the incident X-ray beam is rastered over the surface to be analyzed, then a visual display, the intensity of which is modulated by the peak intensity,
(a)
(b)
Figure 2.10 C 1s chemical state images obtained with an ESCASCOPE, from a contaminated
fluoropolymer [31a], based on: (a) the contribution from C–F bonding, and (b) the contribution from C–C bonding.
2.5 The Auger Parameter
will correspond to the distribution of that element over the surface. The result is also an image, and this technique is standard in modern instruments from various manufacturers. When the whole sample surface is irradiated by the exciting X-rays, an image can be obtained in a different way. The spot accepted by the transfer lens system in front of the input of the CHA is rastered by introducing deflector plates in front of the lens system. Again, only electrons of a characteristic energy can pass the analyzer. This variation can be found in the AXIS series of instruments. Tonner et al. have recorded scanning XPS micrographs at the Advanced Light Source synchrotron radiation center of Lawrence Berkeley National Laboratory [7]. These authors investigated a polished and sputter-cleaned surface of mineral ilmenite with the nominal composition FeTiO3, and used the Fe 3p and Ti 3p lines for imaging; thus, a spatial resolution of about 0.25 μm was achieved. However, by using Fresnel zone plates for focusing the X-ray beam from a synchrotron source, a lateral resolution of 100 nm or better could be achieved [31b]. Care must be taken in interpreting the intensity distribution, since the electron intensity depends not only on the local concentration of the element, but also on the topography, because surface roughness can affect the inelastic background underneath a spectral line. Therefore, elemental maps are customarily presented as variations of the ratio of peak intensity divided by the magnitude of the background on one or both sides of the line, a procedure which can be carried out easily by a computer.
2.5 The Auger Parameter
In Section 2.3, it was shown that in an XPS spectrum, X-ray excited Auger peaks are often as prominent as the photoelectron peaks themselves. For many elements, the chemical shifts in Auger peaks are actually greater than the shifts in photoelectron peaks. The two shifts can be combined in a very useful quantity called the Auger Parameter α*, first used by Wagner [32] and defined in its modified form [33] as: α* = E kin ( X Y Z ) + EB(W )
(2.10)
where Ekin(X Y Z) is the kinetic energy of the Auger transition X Y Z, and EB(W) is the binding energy of an electron in level W. The Auger and photoelectron peak energies must be measured in the same spectrum. The parameter α* is independent of photon energy, and has the great advantage of being independent of any surface charging; it is, therefore, measurable for all types of material, from metals to insulators. Extensive tabulations of α* have been produced, based on a large number of standard materials; the most reliable of these can be found in Ref. [34] and the NIST database [9]. As defined in Equation 2.10, α* is purely empirical; any prominent and conveniently situated Auger and photoelectron peaks can be used. If α* is defined slightly
27
28
2 X-Ray Photoelectron Spectroscopy (XPS)
more rigorously, by replacing EB(W) by EB(X) in Equation 2.10, it is related closely to a quantity called the extra-atomic relaxation energy, ERea , because Δα* can be shown to equal 2⋅ ΔERea ( X ). In other words, the change in Auger parameter between two chemical states is equal to twice the change in the extra-atomic relaxation energies in the two states, associated with the hole in the core level X. Thus, α* has an important physical basis as well as being useful analytically.
2.6 Applications
XPS has been applied in virtually every area in which the properties of surfaces are important. The most prominent areas can be inferred from conferences on surface analysis, such as the European Conference on Applied Surface and Interface Analysis (ECASIA) (e.g., Refs [35, 36]), which is held every two years. These areas are: adhesion; biomaterials; catalysis; ceramics and glasses; corrosion; environmental problems; magnetic materials; metals; micro- and optoelectronics; nanomaterials; polymers and composite materials; superconductors; thin films and coatings; and tribology and wear. Furthermore, the proceedings of these conferences provide a representative survey of current surface-analytical problems and studies. A few selected examples from the areas mentioned above are listed below. 2.6.1 Catalysis
An understanding of the basic mechanisms of catalysis has been one of the major aims of XPS from its inception. XPS is among the most frequently used techniques in catalysis [37], and to this end the technique has been used in two ways: (i) to study “real” catalysts closely resembling those used industrially; and (ii) to study “model” catalysts, where the number of parameters has been reduced by the use of well-characterized crystal surfaces. For industrial catalysts, the chemical information obtainable from XPS is all important, as the course and efficiency of the catalyzed reaction will be governed by the states of oxidation of the elements at the surface. Hence, XPS is used to analyze both elemental and chemical composition as a function of bulk composition, of surface loading where the active material is deposited on a support, of oxidation and reduction treatments, and of the time in the reactor under operating conditions. Figure 2.11 demonstrates the effect on the chemical state of molybdenum in the surface of Mo–Sm–O catalysts of various compositions [38]. For the highest concentration of molybdenum (spectrum d), the Mo 3d doublet is very similar to that of MoO3, but at the samarium-rich composition (spectrum a) it is more like that of a molybdate or a mixed oxide. The effect of successive reduction treatments on a cobalt–molybdenum–alumina catalyst is shown in the changes in
2.6 Applications
Figure 2.11 The Mo 3d XPS spectra of model catalysts in the Mo–Sm–O system [38]. The Mo : Sm ratios are (a) 0.25; (b) 1.33; (c) 4.00; and (d) 8.00.
the Mo 3d spectra of Figure 2.12 [39]. In this case, the air-fired catalyst in spectrum a contains entirely MoVI at the surface, but with repeated hydrogen reduction the spectrum changes radically and can be curve-resolved into MoVI, MoV, and MoIV contributions. The chemical state of the surface in spectrum f contains approximately equal proportions of these three oxidation states. Pijpers et al. [40] have reviewed the application of XPS in the field of catalysis and polymers. Other applications of XPS to catalytic problems deal with the selective catalytic reduction of NOx on Pt- and Co-loaded zeolites. Although the Al 2p line (Al from zeolite) and Pt 4f line interfere strongly, the two oxidation states Pt0 and PtII can be distinguished after careful curve fitting [41]. The influence of Zn deposition on Cu(111) surfaces in methanol synthesis by the hydrogenation of CO2 shows that Zn creates sites that stabilize the formate intermediate, and thus promote the hydrogenation process [42]. Further publications deal with methane oxidation by various layered rock-salt-type oxides [43]; the poisoning of vanadia in VOx/TiO2 by K2O, leading to a lower reduction capability of vanadia due to the formation of VV [44]; and the interaction of SO2 with Cu, Cu2O, and CuO showing temperature-dependent SO2 absorption or sulfide formation [45].
29
30
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.12 Changes in the Mo 3d XPS spectrum from a cobalt–molybdenum–alumina
catalyst during successive reduction treatments in hydrogen at 500 °C [39]. (a) Air-fired catalyst; (b) Reduction time 15 min; (c) 50 min; (d) 60 min; (e) 120 min; (f) 200 min.
2.6.2 Polymers
Polymers have been the most prominent objects of investigation since XPS became commercially available during the late 1960s. XPS is one of the most widely applied methods for characterizing polymer surfaces. Indeed, “bibles” of highresolution XP spectra of more than a hundred different organic polymers have been published by Beamson and Briggs [46], Briggs [47], and Christ [48]. A crucial problem here is the X-ray degradation that takes place in a more or less pronounced manner on virtually all polymer surfaces. Consequently, caution must be exercised when investigating polymer surface reactions and in quantification; in practice, this means reducing the radiation time of the exciting X-rays.
2.6 Applications
Since polymers are typically nonconductive, sample charging can occur and must be compensated for carefully by using, for example, a low-energy electron flood gun, to avoid lineshape distortion and misinterpretation of the measurements. The photoelectron line of main interest is C 1s, as different bonding environments of the carbon atom lead to very small chemical shifts of this line. Therefore, high-resolution XPS is required, and monochromatic radiation should be applied in order to avoid overlaps with satellite lines. Many reports are made each year dealing with the analysis of polymers, and a few examples are provided here. Butyl rubbers are mainly used in tire production, and to study both the influence of different polymers on the surface chemistry and the adhesion of bromobutyl blends as a function of temperature, valence-band XPS was applied to a variety of pure polymers and different butyl–elastomer blends. These investigations showed the value of valence-band XPS for polymers, which are not easy to analyze when using core-level XPS [49]. In a study dealing with the enhancement of adhesion strength of the interface Cu–polyimide, polyimide surfaces were subjected to a plasma treatment, and the modified surfaces then studied using TOF-SIMS/XPS [50]. Many review articles have highlighted specific problems in polymer analysis, including polymer membranes [51], polymeric biomedical and optoelectronic applications [52], and the status of XPS instrumentation in polymer analysis [53]. 2.6.3 Corrosion and Passivation
Besides the application of XPS in catalysis, the study of corrosion mechanisms and corrosion products is another main area of application. Special attention must be devoted to artifacts as a consequence of X-ray irradiation. For example, the reduction of metal oxides (e.g., CuO → Cu2O) can take place, loosely bound water or hydrates may be desorbed into the spectrometer vacuum, and decomposition of hydroxides may occur. Thorough investigations are usually supported by other surface-analytical and/or microscopic techniques such as AFM (see Chapter 29), which is gaining increasing importance. Corrosion products that are formed as thin layers on metal surfaces in aqueous or gaseous environments, along with the nature and stability of passive and protective films on metals and alloys, have also been major areas of XPS application. For this, XPS has been used in two ways – one in which materials that have corroded or passivated in the natural environment are analyzed, and another in which wellcharacterized (usually pure metal) surfaces are studied after their exposure to controlled conditions. More effort has been devoted to study of the corrosion and passivation properties of Fe–Cr–Ni alloys, such as stainless steel and other transitionmetal alloys, than to any other metallic system [40, 42, 54, 55]. The type of spectral information obtainable from an Fe–Cr alloy of technical origin, carrying an oxide and contaminant film after corrosion, is shown schematically in Figure 2.13 [56].
31
32
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.13 Schematic of type of information obtainable from XPS spectra from an Fe–Cr
alloy with an oxide film underneath a contaminant film [56].
Corrosion-inhibiting organic films on steel were studied by using XPS [57], while investigations on inhibiting films formed in sea water [58] demonstrated their stability; that is, no pitting corrosion was observed. Protective layers (frequently TiN), when deposited onto different materials are of major interest. Investigations of TiN coatings exposed to SO2-containing aggressive environments of different humidities have shown that the most aggressive atmosphere can penetrate through pores towards the interface TiN–Fe, and cause corrosion [59]. The corrosion behavior of sandwich-structured TiN–Ti layers was also investigated [60]. The biocorrosion of stainless steel occurs due to exopolymer-producing bacteria, it having been shown that Fe accumulates in the biofilm [61]. The influence of bacteria on the corrosion behavior of a Mo metal surface was also investigated [62]. These last two investigations have highlighted a new field of research in which XPS can be employed successfully. XPS has also been applied to the study of the corrosion of glasses [63], of polymer coatings on steel [64], of tooth-filling materials [65], and also to the investigation of the role of surface hydroxyl groups of oxide films on metals [66] or other passive films. 2.6.4 Adhesion
The adhesion of metal and ink to a polymer, as well as the adhesion of paint and coating to a metal, are of vital importance in several technologies. In the aircraft industry, aluminum-to-aluminum adhesion is employed. In this case, the strength
2.6 Applications
Figure 2.14 The C 1s XPS spectra recorded by Chin-An Chang et al. [67] from perfluoroalkoxy polymer (PFA). (a) Before deposition; (b) Deposition of copper; (c) Deposition of
chromium; (d) Deposition of titanium. For titanium deposition a strong carbide bond was formed, as shown by the structure at 282–285 eV.
and durability of an adhesive bond depend absolutely on the way in which the adhesive compound interacts with the surfaces to which it is supposed to adhere, which in turn often involves the pretreatment of surfaces to render them more reactive. The nature and degree of the reactivity are functions of the chemical states of the adhering surfaces; these states can be monitored by using XPS. In electronic packaging technology, metal–polymer adhesion has been important for many years, with many studies having been devoted to the interactions between metals and polymers. A good example was the study of Chin-An Chang et al. [67], who deposited 200 nm-thick films of copper, chromium, titanium, gold, and aluminum onto various fluorocarbon polymer films, and found that the adhesion in terms of peel strength was highest for titanium. Figure 2.14 shows the C 1s spectra after metal deposition; these revealed that for titanium, and to a lesser extent for chromium, a strong carbide bond was formed, as evidenced by the peak at 282–285 eV. With the other metals, however, a very low adhesion was found, and no carbide-like peaks occurred. The path of failure of an adhesive joint can provide information concerning the mechanism of failure, if an analysis of the elemental and chemical composition can be carried out along the path. Several authors have performed such analyses
33
34
2 X-Ray Photoelectron Spectroscopy (XPS)
by loading the adhesive joint until it fractures, and then using XPS to analyze each side of the fracture. The complementary use of XPS and TOF-SIMS was demonstrated by Fitzpatrick et al. and by Watts et al. [68, 69], who investigated the failure of hot-dipped galvanized steel pretreated with a zinc phosphate conversion coating and joined with epoxy. The observed electrochemical activity was found to be due to degradation at the zinc–polymer interface, and this was responsible for the ingress of aggressive substances. Furthermore, Wolany et al. [70] investigated the adhesion of untreated and plasma-modified polyimide by using XPS and static SIMS. These authors showed that both the number and type of functional groups available for adhering bonds were enhanced by plasma treatment, and that the adhesion strength for Cu deposition was improved. These results were also corroborated by using SIMS. Other topics studied with XPS have included: the effects of thermal treatment on the morphology and adhesion of the interface between Au and the polymer trimethylcyclohexane–polycarbonate [71]; the compositions of the surfaces and interfaces of plasma-modified Cu–PTFE and Au–PTFE, as well as the surface structure and the improvement of adhesion [72]; the influence of excimer laser irradiation of a polymer on the adhesion of metallic overlayers [73]; and the behavior of a Co-rich binder phase of WC–Co hard metal and the deposition of diamond thereon [74]. 2.6.5 Superconductors
Since the discovery of superconducting oxides in 1986, many reports have been made describing the application of XPS to superconductors. One reason for this frenzied activity has been the quest to identify the precise mechanism of superconductivity in these materials. One essential piece of information concerns the chemical state of the constituent elements (particularly of copper) which is common to all of these materials. Some theories have predicted that small, but significant, amounts of the Cu3+ state should occur in superconductors, although other models have disagreed. A comparison by Steiner et al. [75] of the Cu 2p3/2 spectra from compounds containing copper as Cu0 (metal), Cu+, Cu2+, and Cu3+, with the spectra from the superconducting oxide YBa2Cu3O7 and from the base material La2CuO4, showed conclusively that the Cu3+ contribution to the superconducting mechanism must be negligible. Some of these comparative spectra are shown in Figure 2.15. The positions of the Cu 2p3/2 peak in YBa2Cu3O7 and La2CuO4, along with the satellite structure, are very similar to that in CuO (i.e., Cu2+, rather than in NaCuO2, which contains Cu3+). Much effort in superconductor research has been devoted to fabricating thin films with the same superconducting properties as the bulk material, and to preparing suitable electrical contacts to the surfaces of superconductors. XPS has been of great use in both areas through its ability to monitor both composition and chemical state during the processes either of thin superconducting film formation, or of the deposition of thin conducting and semiconducting films onto a
2.6 Applications
Figure 2.15 Comparison of the Cu 2p3/2 and
satellite XPS spectra from several copper compounds, with the spectrum from the superconducting oxide YBa2Cu3O7 [75]. (a)
CuO, Γ = 3.25 eV; (b) La2CuO4, Γ = 3.30 eV; (c) YBa2Cu3O7, Γ = 3.20 eV; (d) NaCuO2, Γ = 1.60 eV; (e) Cu2+ satellites. Γ is the peak width.
superconducting surface. Typical of the latter studies was that of Ziegler et al. [76], who deposited increasing amounts of silicon onto epitaxial films of YBa2Cu3O7−x. Figure 2.16 shows the Si 2p and Ba 4d spectra during the deposition of silicon and after heating the films in oxygen at 250 °C once a 6 nm layer of silicon had accumulated. The Ba 4d spectrum at the bottom of the figure can be resolved into three 4d5/2,3/2 doublets, corresponding – with increasing energy – to barium in the bulk superconductor, in the surface of the superconductor, and as carbonate. With increasing silicon thickness the barium spectrum attenuates but does not change in character, though silicon does not appear in the spectrum until almost 5 nm of silicon has been deposited, which suggests that migration has occurred into the superconductor at room temperature. At a silicon thickness of 6 nm, the Si 2p spectrum can be separated into suboxide contributions, indicating the reaction of silicon with the YBa2Cu3O7−x surface. After heating in oxygen, the Si 2p spectrum changes to that characteristic of SiO2 and silicates, while the Ba 4d spectrum also changes by increasing in intensity and conforming mostly to that expected of a barium silicate. As a result of the latter changes, the superconducting properties of the film were destroyed. The Y 3d and Cu 2p spectra established that yttrium and copper oxides were also formed. 2.6.6 Semiconductors
XPS is an important technique in semiconductor technology. In particular, angleresolved XPS has become popular in the recent past for measuring oxide thickness and composition. In principle, the angular distribution of photoelectrons emitted from layered samples carries information on the compositional depth profile. The reason is that the surface sensitivity of XPS varies with the (off-normal) polar
35
36
2 X-Ray Photoelectron Spectroscopy (XPS)
Figure 2.16 The Si 2p and Ba 4d XPS spectra
[76] from (a) the clean surface of the superconducting oxide YBa2Cu3O7−x; (b–e) The same surface after increasing deposi-
tions of (b) 0.8 nm Si; (c) 2.8 nm Si; (d) 4.4 nm Si; (e) 6.0 nm Si; (f) Deposited surface after heating to 250 °C in oxygen (8 × 10−3 mbar).
emission angle: for grazing emission, most of the signal comes from near-surface layers, whereas for perpendicular emission a larger depth is probed by the outgoing photoelectrons. Conversion of the measured angular distribution into a compositional depth profile is cumbersome since, mathematically speaking, such a procedure belongs to a class of ill-posed problems. In other words, there are limits to the amount of information that can be extracted from such data [77]. Experimental data are often analyzed on the basis of a simple approach in which the elastic scattering of electrons at the peak energy is ignored [78, 79]. Within this model, it is found that the intensity ratio of elemental silicon and silicon bound to oxygen in a carbon-contaminated Si/SiO2 sample is related to the oxide thickness dox as: dox I 1 ⎛ ⎞ × = ln ⎜ 1 + C ox (θ )⎟ ⎝ L cosθ Iel ⎠
(2.11)
Here, θ is the polar emission angle, L is a characteristic length for electron attenuation, and C is a constant depending on various quantities describing the emission
2.6 Applications
process. Elastic scattering effects and other phenomena that are not intrinsically accounted for in the approach of Equation 2.11, are sometimes incorporated in the characteristic length L and the constant C. Thus, the slope of a plot of the right-hand side of Equation 2.11 against 1/cosθ gives the oxide thickness in units of L. One reason why the analysis based on Equation 2.11 is so attractive is that the thickness of the carbonaceous contamination layer does not appear in it. In other words, the derived oxide-layer thickness for an arbitrary thickness of the carbonaceous contamination layer is independent of the latter. On the other hand, for emission angles exceeding 60 ° the effects of elastic scattering become quite significant, and the experimental points in Figure 2.17 then deviate from the straight line expected from Equation 2.11. The data shown as open symbols in Figure 2.17 were acquired from experiments with a Si wafer and analyzed, using the above procedure, by Seah and White [80] for emission angles <60 °. The analysis of these authors gave a value of 39 Å for the mean thickness of the oxide layer. These data have been compared with results obtained using the SESSA simulation software [81], indicated by the solid (elastic scattering included) and dashed (elastic scattering excluded) lines in Figure 2.17. The results were obtained for a 36 Å-thick oxide layer that gave the best consistency between the simulation using SESSA and the experimental data published by Seah and White. It is seen that when elastic scattering is taken into account, the data points for emission angles exceeding 60 ° are also reproduced quite well by the simulation results, and the resulting oxide thickness is in reasonable agreement with the assessment of Seah and White.
Figure 2.17 A plot of ln(1+ CISiO2 / ISi ) as a
function of 1/cosθ. Open symbols: experimental data of Seah and White [80]; dashed line: straight line approximation; solid line:
results of simulations by the NIST database SRD-100 (SESSA [81]) in which elastic electron scattering is taken into account. Reproduced from Ref. [79].
37
38
2 X-Ray Photoelectron Spectroscopy (XPS)
2.7 Ultraviolet Photoelectron Spectroscopy (UPS)
In its principle of operation, UPS is similar to XPS, in that a surface is irradiated with photons, after which the energies of the ejected photoelectrons are analyzed. The physical relationship involved is the same as that of Equation 2.1 for XPS, but in UPS the energy hν of the exciting photons is much lower as the photons are derived from a gaseous discharge that produces hard UV radiation. The gas normally used is helium which, depending on pressure conditions in the discharge, will provide line sources of energy 21.21 eV (He I) or 40.82 eV (He II) with very narrow linewidths (≈20 meV). Due to these low exciting energies, the binding energies EB that appear in Equation 2.1 refer not to core levels, as in XPS, but rather to shallow valence band levels and to other shallow levels, such as those of adsorbates, near the valence band. For energy analysis, the CHA is again used, but the pass energies employed for UPS are much lower than in XPS as a better energy resolution is required. Although UPS is not an analytical technique in the sense that it can provide quantified elemental concentrations on a surface as can XPS and AES, it is a powerful tool in basic research and is another of the group of most frequently used techniques. UPS is used in two ways:
•
Angle-integrated UPS has been employed extensively to study the interactions of gaseous molecules with surfaces, usually in combination with other techniques such as XPS, AES, LEED, and thermal desorption spectroscopy (TDS).
•
In angle-resolved UPS (ARUPS), the energy analyzer – a CHA of small radius – is mounted mechanically in such a way that it can traverse the space around the sample in both polar and azimuthal directions, while the angle of photon incidence is kept constant. The angle of incidence is then changed by rotation of the sample. If the sample is a single crystal with a clean surface, the dispersion of spectral features (i.e., their movement along the energy axis and their increase or decrease in magnitude) can be used theoretically to plot the band structure in the surface region. The changes in surface-associated band structure during reaction with gaseous molecules can then be followed, leading to additional information about the nature of the reaction.
Figure 2.18 Fermi energy scan of a Cu(001) surface [84]. High symmetry points and
directions and the surface Brillouin zone (white square) are also indicated.
References
In general it can be said that, compared to XPS, UPS is used much more in basic research for determining the electronic band structure of solids [82], and less importantly for general analytical purposes. A review of 2-D ARUPS for characterization of the valence band and for Fermi surface mapping can be found in Ref. [83]. As an example, Figure 2.18 shows a Fermi energy scan of a Cu(001) surface [84], where the image depicts the angular distribution of the intensity of photoelectrons collected in a very narrow energy window close to the Fermi level. High symmetry points and directions and the surface Brillouin zone (the white square) are also indicated.
References 1 Siegbahn, K., Nordling, C., Fahlman, A., Nordberg, R., Hamrin, K., Hedman, J., Johansson, G., Bergmark, T., Karlsson, S.-E., Lindgreen, I., and Lindberg, B. (1967) ESCA: Atomic, Molecular, and Solid State Structure Studied by Means of Electron Spectroscopy, Almqvist and Wiksells, Uppsala. 2 Auger, P. (1925) J. Phys. Radium, 6, 205–209. 3 Seah, M.P. and Dench, W.A. (1979) Surf. Interface Anal., 1, 2–11. 4 Tanuma, S., Powell, C.J., and Penn, D.R. (1994) Surf. Interface Anal., 21, 165–176. 5 Briggs, D. and Rivière, J.C. (1990) Practical Surface Analysis (eds D. Briggs and M. Seah), 2nd edn, vol. 1, John Wiley & Sons, Ltd, Chichester, pp. 85–141. 6 Seah, M.P. (1989) Methods of Surface Analysis (ed. J.M. Walls), Cambridge University Press, Cambridge. 7 Tonner, B.P., Droubay, T., Denlinger, J., Meyer-Ilse, W., Warwick, T., Rothe, J., Kneedler, E., Pecher, K., Nealson, K., and Grundl, T. (1999) Surf. Interface Anal., 27, 247–258. 8 Seah, M.P. and Briggs, D. (1990) Practical Surface Analysis (eds D. Briggs and M. Seah), 2nd edn, vol. 1, John Wiley & Sons, Ltd, Chichester, pp. 1–18. 9 Werner, W.S.M., Smekal, W., and Powell, C.J. (2005) National Institute of Standards and Technology, Database SRD 100, Simulation of Electron Spectra for Surface Analysis (SESSA), Gaithersburg, MD. Available at: http:// srdata.nist.gov.
10 Scofield, J.H. (1976) J. Electron Spectrosc. Relat. Phenom., 8, 129–137. 11 Trzhaskoskaya, M.B., Nefedov, V.I., and Yarzhemsky, V.G. (2003) At. Data Nucl. Data Tables, 82, 257–311. 12 Trzhaskoskaya, M.B., Nefedov, V.I., and Yarzhemsky, V.G. (2001) At. Data Nucl. Data Tables, 77, 97–159. 13 Reilman, R.F., Msezane, A., and Manson, S.T. (1976) J. Electron Spectrosc. Relat. Phenom., 8, 389–394. 14 Tanuma, S., Powell, C.J., and Penn, D.R. (1993) Surf. Interface Anal., 20, 77–89. 15 Tanuma, S., Powell, C.J., and Penn, D.R. (1991) Surf. Interface Anal., 17, 927–939. 16 Tanuma, S., Powell, C.J., and Penn, D.R. (1991) Surf. Interface Anal., 17, 911–926. 17 Tanuma, S., Powell, C.J., and Penn, D.R. (1987) Surf. Sci., 192, L849–L857. 18 Werner, W.S.M. (2003) Surface Analysis by Auger and X-Ray Photoelectron Spectroscopy (eds D. Briggs and J. Grant), IM Publications, Chichester, p. 235. 19 Seah, M.P. (1990) Practical Surface Analysis (eds D. Briggs and M.P. Seah), 2nd edn, vol. 1, John Wiley & Sons, Ltd, Chichester, pp. 201–255. 20 (a) Brüsch, P., Müller, K., Atrens, A., and Neff, H. (1985) Appl. Phys. A, 38, 1–18. (b) Zalar, A. and Hofmann, S. (1993) Appl. Surf. Sci., 68 (3), 361–367. (c) Hofmann, S., Zalar, A., Cirlin, E.H., Vajo, J.J., Mathieu, H.J., and Panjan, P. (1993) Surf. Interface Anal., 20 (8), 621–626.
39
40
2 X-Ray Photoelectron Spectroscopy (XPS) 21 Hofmann, S. (1990) Practical Surface Analysis (eds D. Briggs and M.P. Seah), 2nd edn, vol. 1, John Wiley & Sons, Inc., Chichester, pp. 143–199. 22 Gaarenstroom, S.W. (1979) J. Vac. Sci. Technol., 16, 600–604. 23 Hofmann, S. and Steffen, J. (1989) Surf. Interface Anal., 14, 59–65. 24 Bubert, H. and Jenett, H. (1992) Fresenius J. Anal. Chem., 335, 643–647. 25 Gatts, C., Zalar, A., Hofmann, S., and Rühle, M. (1995) Surf. Interface Anal., 23, 809–814. 26 Card, J., Testoni, A.L., and Le Tarte, L.A. (1995) Surf. Interface Anal., 23, 495–505. 27 Do, T. and McIntyre, N.S. (1999) Surf. Interface Anal., 27, 1037–1045. 28 Reiche, R., Oswald, S., Wetzig, K., Dobler, M., Reuther, H., and Walterfang, M. (2000) Nucl. Instrum. Methods Phys. Res. B, 160, 397–407. 29 Morohashi, T., Hoshi, T., Nikaido, H., and Kudo, M. (1996) Appl. Surf. Sci., 100/101, 84–88. 30 Wenham, M.J.G., Barkshire, I.R., Prutton, M., Roberts, R.H., and Wilkinson, D.K. (1995) Surf. Interface Anal., 23, 858–872. 31 (a) Coxon, P., Krizek, J., Humpherson, M., and Wardell, I.R.M. (1990) J. Electron Spectrosc. Relat. Phenom., 52, 821–836. (b) Margaritondo, G. (2003) Surface Analysis by Auger and X-Ray Photoelectron Spectroscopy (eds D. Briggs and J. Grant), IM Publications, Chichester, p. 733. 32 Wagner, C.D. (1975) Faraday Discuss. Chem. Soc., 60, 291–300. 33 Wagner, C.D., Gale, L.H., and Raymond, R.H. (1979) Anal. Chem., 51, 466–482. 34 Wagner, C.D. (1990) Practical Surface Analysis (eds D. Briggs and M. Seah), 2nd edn, vol. 1, John Wiley & Sons, Ltd, Chichester, pp. 595–634. 35 Proceedings of the 12th European Conference on Applications of Surface and Interface Analysis, 9–14 September 2007, Brussels, Belgium (2008) Surf. Interface Anal. 40, 1. 36 Proceedings of the 11th European Conference on Applications of Surface and Interface Analysis, 25–30 September 2005, Vienna, Austria (2006) Surf. Interface Anal. 38, 1.
37 Chorkendorff, I. and Niemantsverdriet, J.W. (2003) Concepts of Modern Catalysis and Kinetics, Wiley-VCH Verlag GmbH, Weinheim. 38 López Nieto, J.M., Cortés Corberán, V., and Fierro, J.L.G. (1991) Surf. Interface Anal., 17, 940–946. 39 Patterson, T.A., Carver, J.C., Leyden, D.E., and Hercules, D.M. (1976) J. Phys. Chem., 80, 1700–1708. 40 Paul Pijpers, A. and Meier, R.J. (1999) Chem. Soc. Rev., 28, 233–238. 41 Boix, A. and Fierro, J.L.G. (1999) Surf. Interface Anal., 27, 1107–1113. 42 Fujitani, T., Nakamura, I., Ueno, S., Uchijiama, T., and Nakamura, J. (1997) Appl. Surf. Sci., 121/122, 583–586. 43 Miyazaki, T., Doi, T., Kato, M., Miyake, T., and Matsuura, I. (1997) Appl. Surf. Sci., 121/122, 492–495. 44 Robba, D., Casale, C., Notaro, M., and Ori, D.M. (1997) ECASIA 97 (eds I. Olefjord, L. Nyborg, and D. Briggs), John Wiley & Sons, Ltd, Chichester, pp. 169–172. 45 Galtayries, A., Grimblot, J., and Bonnelle, J.-P. (1996) Surf. Interface Anal., 24, 345–354. 46 Beamson, G. and Briggs, G. (1992) High Resolution XPS of Organic Polymers, John Wiley & Sons, Ltd, Chichester. 47 Briggs, D. (1998) Surface Analysis of Polymers by XPS and Static SIMS, Cambridge University Press, Cambridge. 48 Christ, B.V. (2000) Handbook of Monochromatic XPS Spectra: Polymers and Polymers Damaged by X-Rays, John Wiley & Sons, Ltd, Chichester. 49 Galuska, A.A. (1999) Surf. Interface Anal., 27, 889–896. 50 Wolany, D., Fladung, T., Duda, L., Lee, J.W., Gantenfort, T., Wiedmann, L., and Benninghoven, A. (1999) Surf. Interface Anal., 27, 609–617. 51 Marchand-Brynaert, J. (1999) Surfactant Sci. Ser., 79, 91–124. 52 Dai, L. and Albert, A.W.H. (2000) J. Phys. Chem. B, 104, 1891–1915. 53 Fulghum, J.E. (1999) J. Electron Spectrosc. Relat. Phenom., 100, 331–355. 54 Asami, K. and Hashimoto, K. (1987) Langmuir, 3, 897–904. 55 Ramanauskas, R. (1999) Appl. Surf. Sci., 153, 53–64.
References 56 Pajonk, G. and Bubert, H. (1999) Fresenius J. Anal. Chem., 365, 236–243. 57 Bentiss, F., Traisnel, M., Gengembre, L., and Lagrenee, M. (1999) Appl. Surf. Sci., 152, 237–249. 58 Qiu, J.H. and Chua, P.H. (1999) Surf. Interface Anal., 28, 119–122. 59 Marco, J.F., Agudelo, A.C., Gancedo, J.R., and Hanzel, D. (1998) Surf. Interface Anal., 26, 667–673. 60 Marco, J.F., Agudelo, A.C., Gancedo, J.R., and Hanzel, D. (1999) Surf. Interface Anal., 27, 71–75. 61 Johansson, L.-S. and Saastamoinen, T. (1999) Appl. Surf. Sci., 144/145, 244–248. 62 Chen, G. and Clayton, C.R. (1999) Surf. Interface Anal., 27, 230–235. 63 Sprenger, D., Bach, H., Meisel, W., and Guetlich, P. (1993) Surf. Interface Anal., 20, 796–802. 64 Idla, K., Talo, A., Niemi, H.E.-M., Forsen, O., and Ylasaari, S. (1997) Surf. Interface Anal., 25, 837–854. 65 Ziegler, C., Reusch, B., and GeisGerstorfer, J. (1998) Fresenius J. Anal. Chem., 361, 547–553. 66 McCafferty, E. (1999) Proc. Electrochem. Soc., 98 (17), 42–55. 67 Chang, C.-A., Kim, Y.-K., and Schrott, A.G. (1990) J. Vac. Sci. Technol. A, 8, 3304–3309. 68 Fitzpatrick, M.F. and Watts, J.F. (1999) Surf. Interface Anal., 27, 705–715. 69 Watts, J.F., Fitzpatrick, M.F., and Carney, T.J. (1997) ECASIA97, John Wiley & Sons, Ltd, Chichester, pp. 113–116. 70 Wolany, D., Fladung, T., Duda, L., Lee, J.W., Gantenfort, T., Wiedmann, L., and Benninghoven, A. (1999) Surf. Interface Anal., 27, 609–617.
71 Bechtolsheim, C.V., Zaporojtchenko, V., and Faupel, F. (1999) Appl. Surf. Sci., 151, 119–128. 72 Zhang, J., Cui, C.Q., Lim, T.B., Kang, E.-T., and Neoh, K.G. (1999) Surf. Interface Anal., 28, 235–239. 73 Horn, H., Beil, S., Wesner, D.A., Weichenhain, R., and Kreutz, E.W. (1999) Nucl. Instrum. Methods Phys. Res., Sect. B, 151, 279–284. 74 Capelli, E., Orlando, S., Pinzari, F., Napoli, A., and Kaciulis, S. (1999) Appl. Surf. Sci., 138/139, 376–382. 75 Steiner, P., Kinsinger, V., Sander, I., Siegwart, B., Huefner, S., Politis, C., Hoppe, R., and Mueller, H.P. (1987) Z. Phys. B, 67, 497–502. 76 Ziegler, C., Baudenbacher, F., Karl, H., Kinder, H., and Göpel, W. (1991) Fresenius J. Anal. Chem., 341, 308–313. 77 Cumpson, P. (1995) J. Electron Spectrosc. Relat. Phenom., 73, 25. 78 Werner, W.S.M. (2005) Surf. Interface Anal., 37, 846. 79 Smekal, W., Werner, W.S.M., and Powell, C.J. (2005) Surf. Interface Anal., 37, 1059. 80 Seah, M.P. and White, R. (2002) Surf. Interface Anal., 33, 960. 81 Werner, W.S.M., Smekal, W., and Powell, C.J. (2005) Simulation of Electron Spectra for Surface Analysis, National Institute for Standards and Technology (NIST), Gaithersburg, MD, US. 82 Hüfner, S. (1995) Photoelectron Spectroscopy, Springer-Verlag. 83 Daimon, H. and Matsui, F. (2006) Progr. Surf. Sci., 81, 367–386. 84 Aebi, P., Osterwalder, J., Fasel, R., Naumovic, D., and Schlapbach, L. (1994) Surf. Sci., 307–309, 917–921.
41
43
3 Auger Electron Spectroscopy (AES) Henning Bubert, John C. Rivière, and Wolfgang S.M. Werner
Together with XPS, AES is among the most widely used surface-analytical techniques. As an accepted surface technique, AES actually predates XPS by two to three years, because the potential of XPS as a surface-specific technique was not recognized immediately by the surface-science community. Although pioneering studies have been conducted by Harris [1] and by Weber and Peria [2], the technique as it is known today is basically the same as that established by Palmberg et al. [3].
3.1 Principles
The surface to be analyzed is irradiated with a beam of electrons of sufficient energy, typically in the range 2–10 keV, to ionize one or more core levels in surface atoms. After ionization, the atom can relax by either of the two processes described in Section 2.1 for XPS – the ejection of a characteristic X-ray photon (fluorescence) or the ejection of an Auger electron. Although these are competing processes, for shallow core levels (EB < 2 keV) the probability of the Auger process is far higher. The Auger process is described schematically in Figure 2.1 (left side), which points out that the final state of the atom is doubly ionized. Equation 2.2 shows that the Auger energy is a function of atomic energy levels only. Because no two elements have the same set of atomic binding energies, an analysis of Auger energies provides elemental identification. Even if levels L1 and L23 are in the valence band of the solid, an analysis is still possible because the dominant term in Equation 2.2 is always the binding energy of K, the initially ionized level. Considering that heavy elements have more levels than just K and L, Equation 2.2 also indicates that the heavier the element, the more numerous are the possible Auger transitions. Fortunately, there are large differences between the probabilities of different Auger transitions so that, even for the heaviest elements, only a few intense transitions occur and analysis is still possible.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
44
3 Auger Electron Spectroscopy (AES)
In principle, chemical as well as elemental information should be available in AES, because the binding energies appearing in Equation 2.2 are subject to the same chemical shifts as measured in XPS. In practice, because the binding energies of three levels are involved, extracting chemical information from Auger spectra is usually difficult, although significant progress has been made in data-processing methods (Section 2.4.2), which have yielded valuable chemical information. The reasons why AES is a surface-specific technique have been given in Chapter 2, with reference to Figure 2.2. The normal range of kinetic energies recorded in an AES spectrum would typically be from 20 to 1000 eV, corresponding to inelastic mean free path values of two to six monolayers. The nomenclature used in AES has also been mentioned in Chapter 2. The Auger transition in which initial ionization occurs in level X, followed by the filling of X by an electron from Y and the ejection of an electron from Z, would therefore be labeled XYZ. In this rather restricted scheme, one would thus find in the KLL series the six possible transitions KL1L1, KL1L2, KL1L3, KL2L2, KL2L3, and KL3L3. Other combinations could be written for other series such as the LMM, MNN, and so on.
3.2 Instrumentation 3.2.1 Vacuum Requirements
The same considerations discussed in Section 2.1 for XPS apply to AES. 3.2.2 Electron Sources
The energy of the exciting electrons does not enter into Equation 2.2 for the Auger energy, unlike that of the exciting X-ray photon in XPS. Thus, the energy spread in the electron beam is irrelevant. However, the actual energy of the primary electrons is relevant, because the cross-section for ionization by electron impact depends on that energy. For an electron in a core level of binding energy EB, this dependence increases steeply with primary energies above EB, passes through a maximum of approximately three- to fivefold EB, and then decreases to a fairly constant plateau. This is illustrated in Figure 3.1, from Casnati et al. [4], in which various theoretical and experimental cross-sections are shown for ionization of the Ni–K shell. The implication of this dependence is that, for the efficient ionization of a particular core level, the primary energy should be approximately fivefold the binding energy of an electron in that level. Since in most samples for analysis, several elements are found in the surface region with core-level binding energies extending over a large range, the choice of a primary energy sufficiently high to ionize all of the core levels efficiently is advisable. Hence, for conventional AES,
3.2 Instrumentation
Figure 3.1 Electron-impact ionization
cross-section for the Ni–K shell, as a function of reduced electron energy U, U = Ep/EK, where Ep is the primary electron energy and
EK the binding energy of the K shell. (a) Experimental points; (b) Semi-empirical or theoretical curves.
primary energies are typically in the range 3–10 keV. In scanning Auger microscopy (SAM), much higher primary energies are used, in the range of 25–50 keV, because in that version of AES the electron spot on the sample must be as small as possible, and the focusing required can be effected only at such energies. For the primary energy range from 3–10 keV used in conventional AES, the electron emitter is thermionic, usually an LaB6 or tungsten cathode, and focusing of the electron beam is performed electrostatically. Typically, such an electron source would be able to provide a spot size on the specimen of approximately 0.5 mm at 10 keV and a beam current of approximately 10−8 A. Although the beam can normally be rastered over the specimen surface, such a source would not be regarded as adequate for SAM. Sources for SAM can be of either the thermionic or the field-emission type, the latter being particularly advantageous in achieving minimum spot sizes, because it is a high-brightness source. Focusing is performed electromagnetically, and for optimum performance a modern electron gun for SAM would be capable of a spot size of approximately 10 nm at 30 keV and a beam current of approximately 10−10 Å. The beam is rastered over the surface at scan rates variable up to television scan rates, and the dimensions of the scanned area can also be varied. 3.2.3 Electron-Energy Analyzers
An electron-energy analyzer often applied in AES is the cylindrical mirror analyzer (CMA). In the CMA shown schematically in Figure 3.2, two cylinders of radii r1 (inner) and r2 (outer) are accurately positioned coaxially. Annular entrance and exit apertures are cut in the inner cylinder, and a deflection potential –V is applied between the cylinders. Electrons leaving the sample surface at the source S with a particular energy E on passing into the CMA via the entrance aperture are then deflected back through the exit aperture to the focus F. For the special case in
45
46
3 Auger Electron Spectroscopy (AES)
Figure 3.2 Schematic diagram of a cylindrical mirror analyzer (CMA).
which the acceptance angle α = 42.3 °, the first-order aberrations vanish, and the CMA becomes a second-order focusing device. The relationship between the electron energy E and the deflection potential is then: E = 1.31 eV/ ln (r2/r1 )
(3.1)
As in the CHA, therefore, scanning the deflection potential –V and recording the signal as a function of electron energy provide the distribution in energy of electrons leaving the sample surface. If the angular spread of the acceptance angle is Δα, the relative energy resolution of a CMA is: ΔE/E = 0.18 w/r1 + 1.39 (Δα)3
(3.2)
where w is the effective slit width. For a typical angular spread of Δα ≈ 6 °, w can be replaced by the source size (i.e., the area on the sample from which electrons are accepted). The source size is very small, because of the focused primary beam, and the important property of the CMA is, therefore, its very high transmission, arising from the large solid angle of acceptance; for Δα ≈ 6 °, the transmission is approximately 14%. Conventionally, Auger spectra have been presented in the differentiated energy distribution form dN(E)/dE, rather than in the undifferentiated form N(E) used in XPS. Differentiating enhances the visibility of Auger features, as is demonstrated in Figure 3.3, in which the differentiated spectrum of boron is shown above the corresponding N(E) spectrum. The N(E) distribution is nowadays recorded digitally, and differentiation is performed by computer. With the rapid improvement in the signal-to-noise characteristics of detection equipment, however, the trend in presenting Auger spectra is increasingly toward the undifferentiated N(E) distribution, which also offers significant benefits for quantification of the specimen’s composition. Note from Figure 3.3 that, although the true position of the boron KLL Auger peak in the N(E) spectrum is at 167 eV, the position in the dN(E)/dE spectrum is
3.3 Spectral Information
Figure 3.3 Secondary electron distribution of boron. (a) In the N(E) mode; (b)–(d) In the
dN(E)/dE mode, where the differentiation reveals the plasmon satellites.
taken for purely conventional reasons to be that of the negative minimum, that is, at 175 eV.
3.3 Spectral Information
As stated above, the range of primary energies typically used in conventional AES is 3–10 keV, and the ionization cross-section passes through a maximum at threeto fivefold the binding energy of an electron in the ionized core level. The core levels that can be ionized efficiently are, therefore, limited. Thus, K-shell ionization efficiency is adequate up to Z ≈ 14 (silicon), L3 up to Z ≈ 38 (strontium), M5 up to Z ≈ 76 (osmium), and so on (for the much higher primary energies used in SAM, the ranges are correspondingly extended). This means that in various regions of the Periodic Table, characteristic Auger transitions occur that are most prominent under typical operating conditions. Figures 3.4–3.6 show some of these prominent Auger features, all recorded using a CMA, in the differential distribution [5]. Examples of the KLL Auger series are shown in Figure 3.4, of the LMM series in Figure 3.5, and of the MNN series in Figure 3.6. In the KLL series, the most prominent feature is a single peak arising from the KL2,3L2,3 transition, with minor features from other Auger transitions to lower energy. The LMM series is characterized by a triplet arising from L3M2,3M2,3, L3M2,3M4,5, and L3M4,5M4,5, transitions in ascending order of kinetic energy, with another intense peak, the M2,3M4,5M4,5, appearing at very low energy (e.g., 50 eV
47
48
3 Auger Electron Spectroscopy (AES)
Figure 3.4 KLL Auger series characteristic of light elements [5].
Figure 3.5 LMM and MMM Auger series in the middle of the first series of transition elements [5].
3.3 Spectral Information
Figure 3.6 MNN Auger series in the second series of transition elements [5].
for iron). Note that the AES transitions involving electrons in valence band levels are often written with a “V” rather than the full symbol of the level. Thus, L3M2,3M4,5 and L3M4,5M4,5 would often appear as L3M2,3V and L3VV, respectively, and similarly M2,3M4,5M4,5, as M2,3VV. In Figure 3.5 the increase in the intensity of the L3VV peak relative to the other two, upon going from chromium to iron, is due to the progressive increase in the electron density in the valence band. The characteristic doublet seen in the MNN series arises from the M4,5N4,5N4,5 transitions, in which the doublet separation is that of the core levels M4 and M5. Chemical effects are quite commonly observed in Auger spectra, but are difficult to interpret compared with those in XPS, because additional core levels are involved in the Auger process. Some examples of the changes to be seen in the KLL spectrum of carbon in different chemical environments are given in
49
50
3 Auger Electron Spectroscopy (AES)
Figure 3.7 Examples of the effect of different chemical states on the KLL Auger spectrum of carbon [6] (SiC* and graphite* denote Ar+-bombarded surfaces of SiC and graphite, respectively).
Figure 3.7 [6]. Such spectra are typical components of data matrices (see Section 2.4.2) derived from AES depth profiles. Associated with prominent features in an Auger spectrum are plasmon energy loss features that are also found associated with photoelectron peaks in an XPS spectrum. Plasmon energy losses arise from excitation of modes of collective oscillation of the conduction electrons by outgoing secondary electrons of sufficient energy. Successive plasmon losses suffered by the backscattered primary electrons can be seen in both the N(E) and the dN(E)/dE spectra of Figure 3.3 for boron. The same magnitude of plasmon loss, ∼27 eV, is also associated with the KLL
3.4 Quantification and Depth Profiling
Auger peak of boron. Plasmon losses are also present in Figures 3.5–3.7, but they are difficult to disentangle from minor Auger features. Another type of loss feature not often recorded in Auger spectra can also be seen in Figure 3.3, in the dN(E)/dE spectrum at 810 eV. This type arises from core-level ionization, and forms the basis for the technique of core electron energy-loss spectroscopy (CEELS). A primary electron interacting with an electron in a core level can cause excitation either to an unoccupied continuum state, leading to complete ionization, or to localized or partly localized final states, if available. Since the primary electron does not need to lose all its energy in the interaction, the loss feature appears as a step in the secondary electron spectrum, with a tail decreasing progressively towards lower energies. Differentiation, as in Figure 3.3, accentuates the visibility. In that figure, the loss at 810 eV corresponds to the K-shell ionization edge of boron at approximately 190 eV.
3.4 Quantification and Depth Profiling 3.4.1 Quantification
For a pure element in a homogeneous sample, the Auger intensity into a solid angle dΩ is given by I AXYZ = k × I 0 × γ AXYZ × n AX × σ AX (E 0 ) × sec α × [1 + rA (E AX ,E 0,α )] dΩ ⎞ × N A × Q (E AXYZ ) × λ A (E AXYZ ) × cos Θ × ⎛ ⎝ 4π ⎠
(3.3)
where γ AXYZ is the probability that the core hole in level X is filled through an XYZ Auger process, n AX is the number of electrons in level X, σ AX (E 0 ) is the ionization cross-section for an electron in level X for incoming electrons of energy E 0 at an angle α relative to the surface normal, rA (E AX ,E 0,α ) gives the additional ionization due to backscattering of the incident electrons, N A the atomic density of A atoms, Q (E AXYZ ) takes into account the elastic scattering of Auger electrons on their way to the surface, λ A (E AXYZ ) is the inelastic mean free path at the Auger emission energy, and Θ is the angle of emission of the detected electrons from the surface normal. k is a proportionality constant containing fixed operating conditions, for example the transmission of the analyzer at the kinetic energy EA, the efficiency of the detector at the kinetic energy of the Auger electrons, and the probability of the Auger transition XYZ which is given by 1 − ω (where ω is the fluorescence yield). The cross-section for electron impact ionization has already been mentioned in Section 3.2.2 in connection with electron sources, and a variety of experimental and theoretical cross-sections have been shown in Figure 3.1 for the particular case of the K-shell of nickel. The expression for the cross-section derived by Casnati et al. [4] gives reasonably good agreement with experiment; the earlier expression of Gryzinski [7] is also useful.
51
52
3 Auger Electron Spectroscopy (AES)
For the inelastic mean free path, Equation 2.7 given for XPS also applies. Ichimura and Shimizu [8] have performed extensive Monte Carlo calculations of the backscattering factor as a function of primary beam energy and angle of incidence, of atomic number and of binding energy. A selection of their results for Ep = 10 keV at normal incidence is shown in Figure 3.8. Recently, Jablonski and coworkers [9] have revisited this subject and published improved empirical data for the backscattering factor derived from Monte Carlo calculations. As for XPS, the average surface concentration N A can, in principle, be calculated by measurement of the Auger current, according to Equation 3.3. Figure 3.9 shows
Figure 3.8 Values of the backscattering factor R (i.e., 1 + r) calculated as a function of atomic
number for an electron beam of 10 keV at normal incidence by Ichimura et al. [8]. (Curve a) EB = 0.1 keV; (curve b) EB = 0.5 keV; (curve c) EB = 1.0 keV; (curve d) EB = 2.0 keV.
Figure 3.9 Ratio between experimental and theoretical Auger sensitivities [10].
3.4 Quantification and Depth Profiling
the ratio between experimental and theoretical sensitivities [10]. It can clearly be seen that deviations of more than a factor of 2 occur. Given the fact that these results hold true for pure elements, this indicates clearly that, for practical purposes, the quantification of AES data without standards is unacceptable, in contrast to XPS where an accuracy of approximately 10% can be achieved. Thus, relative sensitivity factors are generally used. If the Auger current I A∞ for the same transition XYZ in a standard of pure A is measured under the same experimental conditions as in the analysis of A in M, then the ratio of the atomic concentrations is N A I A × (1 + rA ) × l A × R A∞ = N A∞ I A∞ × (1 + rM ) × lM × RM
(3.4)
because the ionization cross-section is the same for both standard and sample. Then, as in XPS, assuming nearly equal roughness factors R A∞ ≈ RM, Equation 3.4 can be reduced to: N A = SA × I A
(3.5)
where: SA =
N A∞ × (1 + rA ) × l A I A∞ × (1 + rM ) × lM
(3.6)
SA is the relative sensitivity factor. Normally, values of SA are derived either empirically or semi-empirically. Tables of such sensitivity factors have been published by Payling [11] and by Mroczkowski and Lichtman [12] for differential Auger spectra, and for the direct, undifferentiated, spectra by Sato et al. [13]. A comprehensive discussion of the assumptions and simplifications involved, and the corrections that should be applied, is given elsewhere [14]. Although modern Auger spectrometers detect the spectra in the undifferentiated mode, many researchers still evaluate the spectra in the subsequently electronically differentiated mode – that is, using peak-to-peak heights – so that any changes in peak shape that occur as a result of changes in chemical bonding are not taken into account. It should also be mentioned that software routines for quantification of AES raw data are normally supplied with the instruments, and sensitivity factors can also be found in the various manufacturers’ handbooks (see e.g., Ref. [15]). 3.4.2 Depth Profiling
As in XPS, elemental distributions near the surface – but to depths greater than the analytical depth resolution – are frequently required when using AES. Again, the universally employed technique is that of progressive removal of the surface by ion sputtering. Again, one must be aware that the electron spectroscopic sampling depth corresponds roughly to the depth of the layer modified by preferential sputtering effects, and again the edges of the sputter crater should not contribute to the signal.
53
54
3 Auger Electron Spectroscopy (AES)
A variation on depth profiling that can be performed by modern scanning Auger instruments is to program the incident electron beam to jump from one preselected position on a surface to each of many others in turn, with multiplexing at each position; this is called multiple point analysis. Sets of elemental maps acquired after each sputtering step or each period of continuous sputtering can be digitally processed to derive a three-dimensional (3-D) analysis of a selected micro volume in the nanometer range. It should be mentioned, however, that depth profiling with AES is very time consuming in combination with SAM, due to the long time periods (sometimes up to several days) required for sputtering to reasonable depths.
3.5 Applications
Like XPS, the application of AES has been very widespread, particularly during the earlier years of its existence. More recently, the technique has been applied increasingly to those problem areas that require the high spatial resolution that AES can provide in the nanometer range and XPS, currently, cannot. Since data acquisition in AES is faster than in XPS, it is also employed widely in routine quality control by surface analysis of random samples from production lines of, for example, integrated circuits. In the semiconductor industry, in particular, SIMS is a competing method. Note that AES and XPS on the one hand and SIMS/ SNMS on the other hand, when both are in depth-profiling mode, are complementary. The former is based on signals from the sputter-modified surface, and the latter from the flux of sputtered particles. 3.5.1 Grain Boundary Segregation
One of the original applications of AES – and still one of the most important – is the analysis of grain boundaries in metals and ceramics. Very small amounts of impurity or dopant elements in the bulk material can migrate under appropriate temperature conditions to the boundaries of the grain structure and accumulate there. In this way, the concentration of minor elements at the grain boundaries can become much higher than in the bulk, and the cohesive energy of the boundaries can be so altered that the material becomes brittle. Knowledge of the nature of the segregating elements and their grain boundary concentrations as a function of temperature can be used to modify fabrication conditions, and thereby improve the strength performance of the material in service. Goretzki [16] has discussed the importance and analysis of internal surfaces such as grain boundaries and phase boundaries, and given several examples. He emphasizes that, in order to avoid ambiguity in interpretation of the analysis, the internal surfaces must be exposed by fracturing the material under ultra-high
3.5 Applications
Figure 3.10 Auger spectra of fracture surfaces of a 12%-Cr steel [16]. (a) Embrittled state; (b) Unembrittled state.
vacuum (UHV) conditions inside the electron spectrometer. If this is not done, atmospheric contamination would immediately change the surface condition irrevocably. One such example is given in Figure 3.10, in which Auger spectra in the differential mode from fracture surfaces in the embrittled (Figure 3.10a) and unembrittled (Figure 3.10b) states of a 12%-Cr steel containing phosphorus as an impurity, are compared. Embrittlement occurs when the steel is held at 400–600 °C for a long time. Note in the spectrum from the embrittled material that not only is the phosphorus concentration greatly enhanced at the grain boundary, but the amounts of chromium and nickel, relative to iron, are also increased. By setting the analyzer energy first to that of the phosphorus peak, and then to that of the principal chromium peak, Auger maps could be recorded showing the distribution of phosphorus and chromium over the fracture surface. The two maps showed a high correlation of the two elements, indicating an association of phosphorus and chromium at the grain boundary, possibly as a compound. Other recently published studies of the same steel have also reported enhanced phosphorus levels at the grain boundaries of the embrittled material [17, 18]. Similar results have been obtained for other high-chromium steel alloys [19, 20]. Reichl et al. [21] have investigated the temperature-dependence of the surface segregation of S and N in α-Fe. Initially S diffusion occurred to the grain boundaries, but then it displaced N at the surface at higher temperatures. Other applications in the study of grain boundary segregation in metals have included the effects of boron, zirconium, and aluminum on Ni3Al intermetallics [22], the segregation of phosphorus and molybdenum to grain boundaries in the superalloy Nimonic PE 16 as a function of aging treatment [23],
55
56
3 Auger Electron Spectroscopy (AES)
and the enrichment of chromium carbide and chromium–iron carbide at the grain boundaries in steel [24]. The application of AES to zirconia ceramics has been reported by Moser et al. [25]. Elemental maps of Al and Si demonstrated the grain boundary segregation of small impurities of silica and alumina in these ceramics. 3.5.2 Semiconductor Technology
With ever-increasing miniaturization of integrated circuits goes a need for the ability to analyze ever smaller areas, so that the integrity of fabrication at any point on a circuit can be checked. This checking process includes not only microdetails, such as the continuity of components in the form of thin films and establishment of the correct elemental proportions in contacts, diffusion barriers, and Schottky barriers, but also the effectiveness of macro treatments such as surface cleaning. In all these types of analysis, depth profiling is used extensively because, in general, elemental compositional information is required, and not information about the chemical state. As mentioned in Section 2.4.2, chemical state information tends to become blurred or destroyed by ion bombardment. As reviewed in more detail in Ref. [26], the application of AES in semiconductor technology is particularly important for dopant depth profiling, defect localization, and the characterization of metal contacts, capping layers, barrier layers, and etch stop layers. It should also be mentioned that analysis with AES can be aided by shaping the sample with the focused ion beam (FIB) technique. This can be used, for instance, to remove surrounding material from via (vertical interconnect access) holes in order to analyze the undisturbed base of the via by AES. Several materials have been prescribed as diffusion barriers to prevent one circuit material from diffusing into another (often Si) at the temperatures needed for the preparation of certain components, such as oxide films in integrated circuits. Zalar et al. [27] studied interdiffusion at an Al2O3/Ti interface by covering Si(111) with a thin film of TiN as a diffusion barrier, then with a 45 nm-thick layer of polycrystalline Ti, and finally with 55 nm of amorphous Al2O3. These samples were heated to different temperatures at a rate of 40 °C min−1, and the interdiffusion processes investigated by AES depth profiling. Figure 3.11a–c shows that oxygen from Al2O3, and to a lesser extent also Al, migrated into the Ti layer. (The “N + Ti” intensity in the Ti layer is that of Ti only, whereas in the neighboring TiN layer the “N + Ti” intensity is that of Ti and N due to peak overlapping.) Control measurements indicated that the oxygen did not arise from the external gas phase. In the interface region a new phase consisting of a-Ti3Al was formed, identified by using selected-area electron diffraction (SAED). Such AES investigations are normally performed in combination with other surface-analytical techniques – in this case with XPS, XRD, TEM, and electron diffraction. Other studies dealing with the analysis of CVD WxN films as barriers for Cu metallization have been reported by Hua Li et al. [28], who used AES and RBS, AFM, XRD, and TEM.
3.5 Applications
Figure 3.11 AES sputter-depth profiles of the α-Al2O3–Ti thin-film structure on a smooth Si
substrate covered with a TiN thin film diffusion barrier. (a) As-deposited; (b) After heating to 500 °C; (c) After heating to 580 °C [27].
The nature of the metallic contacts to compound semiconductors has also been an interesting problem. Ohmic contacts must be attached to semiconducting surfaces, otherwise circuits could not be fabricated. During fabrication, however, excessive interdiffusion must not occur, otherwise the electrical characteristics of the semiconductor could not be maintained. Among the many studies of the
57
58
3 Auger Electron Spectroscopy (AES)
reactions of metallizing alloys with compound semiconductors is that of Roberts et al. [29], who investigated the W/TiN/Ti/Si contact structure using a multispectral Auger microscope (MULSAM) [30] and TEM. An Auger electron spectrometer with high spatial resolution imaging capability has been developed especially for the detection of those small particles and defects that are particularly problematic for ultra large-scale integrated (ULSI) electronics. It has enabled the inspection of wafers up to 200 mm in diameter [31]. AES has also been applied to study preferential sputtering of TiSix which is formed in low-resistivity conductor films in ULSI devices [32], and the electromigration behavior of Au–Ag films on SiO2 using AES, XPS, and AFM [33]. 3.5.3 Thin Films and Interfaces
The nature of the interface formed between very thin metallic films and substrates of various types has been studied extensively by AES, just as it has by XPS. The extent of interaction and interdiffusion can be established by AES with depth profiling, although chemical information is normally absent, but the development of the interface is usually studied by continuous recording of Auger spectra during film deposition. Zhu et al. [34] deposited Ti or Cr layers on diamond samples, following which the Ti–diamond or Cr–diamond specimens were annealed at 600 °C for 4 h during metallization. Figure 3.12a and b shows the depth profiles before and after annealing a Ti/diamond specimen, while Figures 3.12c and d show the Ti LM1M4 Auger line shapes in the N(E) mode and the C-KLL line shape in the differentiated dN(E)/ dE mode at different depths. The unstructured Ti signal at the surface can be attributed to TiO2 due to a smooth oxidation of the surface. However, the greater part of the layer consists of TiC, the formation of which is recognizable from the strongly structured carbon line that changes its shape with depth, gradually becoming similar to that of diamond. These features are consistent with the presence of TiC, indicating complete metallization. Other interesting thin-film studies using AES have included the growth of platinum on TiO2- and SrO-terminated (100) SrTiO3 single-crystal substrates [35], of epitaxial niobium films on (110) TiO2 [36], the interaction of copper with a (0001) rhenium surface [37], and the characterization of radiofrequency (rf) sputtered TiN films on stainless steel [38]. 3.5.4 Surface Segregation
The surface composition of alloys, compounds, and intermetallic materials can change profoundly during heat treatment, as a result of the segregation of one or more constituents and impurity elements. Surface enhancement can also occur as a result of chemical forces during oxidation or other reactions. Although the mechanism of surface segregation is similar to that of grain boundary segregation
3.5 Applications
Figure 3.12 Depth profiles of a Ti–diamond layer (a) before and (b) after annealing at 600 °C for 4 h [34]. Auger line shapes of (c) Ti-LM1M4 and (d) C-KLL at various depths.
(as discussed above), at the free surface the material is much more likely to be lost through reaction, evaporation, and so on, so that excessive continued segregation can change the bulk properties as well as those of the surface. Nearly all such studies have been performed on metals and alloys. Both, AES and LEED have often been applied together to the study of surface segregation. On occasion, the segregation of metallic and nonmetallic elements on surfaces can result in the formation of surface compounds, as demonstrated by Uebing [39, 40]. In this case, a series of Fe–15%Cr alloys containing either 30 ppm nitrogen, 20 ppm carbon, or 20 ppm sulfur was heated to 630–800 °C; the typical Auger spectra shown in Figure 3.13 were obtained. Each spectrum shows that the amounts of both chromium and the nonmetal were greater than would be expected if the bulk concentration were to be maintained. The fine structure of the nitrogen and carbon KLL spectra is characteristic of nitride and carbide, respectively. Thus,
59
60
3 Auger Electron Spectroscopy (AES)
Figure 3.13 Formation of surface compounds on Fe–15%-Cr alloys by cosegregation of
chromium and a nonmetallic element [39]. (a) Nitride; (b) Carbide; (c) Sulfide.
the cosegregation of chromium with nitrogen and with carbon has led to the formation of surface nitrides and carbides. Song et al. [41] have investigated surface segregation in Ni0.65Nb0.35 up to 500 °C and in an atmosphere of 1 atm O2. Below 100 °C, Nb migrated to the surface, with the formation of Nb2O5, while at temperatures ≥300 °C the migration of Ni to the surface resulted in Ni and NiO. This inverse segregation can be explained by the different diffusion rates of O and Ni in the Nb2O5 layer formed at the surface. Other investigations involving AES, often with depth profiling, have dealt with the surface segregation of Ag in Al–4.2% Ag [42], of Sn in Cu and the formation of a superficial Sn–Cu alloy [43], of Mg in Al–Mg alloy [44], and of Sb in Fe–4% Sb alloy [45]. When using depth profiling, it was necessary to differentiate between segregation (which is a bulk property) and any artifacts introduced by preferential sputtering.
3.6 Scanning Auger Microscopy (SAM)
3.6 Scanning Auger Microscopy (SAM)
If an incident electron beam of sufficient energy for AES is rastered over a surface in a manner similar to that in scanning electron microscopy (SEM), and if the analyzer is set to accept electrons of Auger energies characteristic of a particular element, then an elemental map or image can be obtained, similar to that using XPS in the Quantum 2000 (see Section 2.2.5). This is illustrated in Figure 3.14 [46]. The surface was that of a fractured compact of SiC to which boron and carbon had been added to aid the sintering process. The aim of the analysis was to establish the uniformity of distribution of the additives and the presence or absence of impurities. The Auger maps show not only a very nonuniform distribution of boron (Figure 3.14a) but also a strong correlation of boron with sodium (Figure 3.14c), and weaker correlation of boron with potassium (Figure 3.14b). Point analyses for points A and B marked on the images also reveal the presence of sulfur and calcium in some areas. In this case, the sintering process had not been optimized. The quality of elemental maps obtained from SAM can be enhanced significantly by various image processing techniques [47]. As an example, Figure 3.15 shows how scatter diagrams can be used to enhance the boundaries between different phases, which are blurred in the raw images due to edge artifacts introduced by the sample’s topography. In this example, a 40 nm-thick Al film on Si is shown. Briefly, the scatter diagram in Figure 3.15a depicts a distribution of the number of pixels (visualized by the brightness) for a given pair of Auger signal intensities for Al and Si. Two clusters of pixels are clearly visible, those with high Al signal and low Si signal (upper left) and those with high Si signal and low Al signal (lower right). Between these two clusters a region of lower numbers of pixels exhibiting significant signal intensities of both Al and Si is apparent. From this scattering diagram, different regions can be selected and converted into an enhanced false color image of the different phases in the sample (Figure 3.15b) [48]. An application of scanning Auger microscopy in material science is shown in Figure 3.16, which depicts a false color Auger element map of Fe, Cr, and Si at a metal–oxide interface [49]. An enrichment of Cr (red) and Si (green) at the grain boundaries can be observed. Figure 3.17 shows how the combination of X-ray element mapping with Auger element mapping has been used to study pitting corrosion in stainless steel at MnS inclusions [50]. It can be seen that the two techniques deliver different distributions of manganese, which can be explained by the higher surface specificity of the Auger technique. From that it can be concluded that manganese was present both in the deeper regions of the pit and on the very surface of the sample. Similarly, the distribution of iron revealed that that element was present only in deeper regions of the pit, and that the surface was free from iron. In combination with similar maps of other elements, it could be concluded that the surface of the pit consisted solely of MnCl2.
61
62
3 Auger Electron Spectroscopy (AES)
Figure 3.14 SAM map of fractured SiC after sintering with B added [46]. (a–d) Elemental
maps of boron, potassium, sodium, and oxygen, respectively; (e, f) Point analyses at points A and B, respectively.
3.6 Scanning Auger Microscopy (SAM)
Figure 3.15 (a) Scatter diagram of a 40 nm-thick Al film on Si; (b) False color image of the elements Al (blue, region w1 in the scatter diagram), Si (yellow, region w3 in
the scatter diagram), and of the correlation line between the two regions (red, region w2 in the scatter diagram) [48].
Figure 3.16 False color Auger element maps of iron (blue), chromium (red), and silicon (green) at the grain boundaries of a metal–oxide interface. Image size: 40 μm × 30 μm [49].
Dedicated scanning Auger microscopes with high spatial resolution have been developed with the aim of achieving the smallest possible spot size while retaining an acceptable Auger signal-to-noise ratio (SNR) and a tolerable acquisition time. For this purpose, sources with a high brightness are needed. The electron guns used for SAM employ electromagnetic focusing. However, because the focusing of electrons of low energy into a small spot is difficult to achieve, the beam energies used in SAM are in the range of 25–50 keV, with beam currents in the order of 1 nA or less, and a minimum spot size of approximately 10 nm.
63
64
3 Auger Electron Spectroscopy (AES)
Figure 3.17 (a) Auger element map and (b) X-ray element map of manganese at a corrosion
pit in stainless steel [50].
References 1 Harris, L.A. (1968) J. Appl. Phys., 39, 1419–1427, 1428–1431. 2 Weber, R.E. and Peria, W.T. (1967) J. Appl. Phys., 38, 4355–4358. 3 Palmberg, P.W., Bohm, G.K., and Tracy, J.C. (1969) Appl. Phys. Lett., 15, 254–255. 4 Casnati, E., Tartari, A., and Baraldi, C. (1982) J. Phys. B, 15, 155–167. 5 Davis, L.E., MacDonald, N.C., Palmberg, P.W., Riach, G.E., and Weber, R.E. (1976) Handbook of Auger Electron Spectroscopy, Physical Electronic Industries, Minnesota. 6 (a) Kny, E. (1980) J. Vac. Sci. Technol., 17, 658–660. (b) Kleefeld, J. and Levinson, L.L. (1979) Thin Solid Films, 64, 389–393. (c) Smith, M.A. and Levinson, L.L. (1977) Phys. Rev. B Solid State, 16, 1365–1369. 7 Gryzinski, M. (1965) Phys. Rev., 138 (A), 305–321. 8 Ichimura, S. and Shimizu, R. (1981) Surf. Sci., 112, 386–408. 9 Jablonski, A. (2009) Surf. Sci., 603, 2047–2056. 10 Matthew, J.A.D. (2006) Scanning Auger Electron Spectroscopy (eds M. Prutton and
11 12
13
14
15 16 17
M. El Gomati), John Wiley & Sons, Chichester, p. 15. Payling, R. (1985) J. Electron Spectrosc. Relat. Phenom., 36, 99–104. Mroczkowski, S. and Lichtman, D. (1985) J. Vac. Sci. Technol. A, 3, 1860–1865. Sato, T., Nagasawa, Y., Sekine, T., and Sakai, Y. (1989) Surf. Interface Anal., 14, 787–793. (a) Seah, M.P. (1986) Surf. Interface Anal., 9, 85–98. (b) Werner, W.S.M., Smekal, W., and Powell, C.J. (2005) National Institute of Standards and Technology, Database SRD 100, Simulation of Electron Spectra for Surface Analysis (SESSA), Gaithersburg, MD. Available at: http://srdata.nist.gov. Thermo Fisher Scientific Inc. (2009) XPS and Auger Handbook, East Grinstead. Goretzki, H. (1987) Fresenius J. Anal. Chem., 329, 180–189. Wild, R.K. and Hickey, J. (1997) ECASIA 97 (eds I. Olefjord, L. Nyborg, and D. Briggs), John Wiley & Sons, Ltd, Chichester, pp. 643–646.
References 18 Saidi, D., Souami, N., Oulladj, M., Negache, M., and Lebaili, S. (1997) ECASIA 97 (eds I. Olefjord, L. Nyborg, and D. Briggs), John Wiley & Sons, Ltd, Chichester, pp. 683–686. 19 Vatter, I.A. and Titchmarsh, J.M. (1997) Surf. Interface Anal., 25, 760–776. 20 Norell, M. and Nyborg, L. (1997) ECASIA 97 (eds I. Olefjord, L. Nyborg, and D. Briggs), John Wiley & Sons, Ltd, Chichester, pp. 687–690. 21 Reichl, B.M., Eisl, M.M., Weis, T., Hutter, H., and Stoeri, H. (1995) Fresenius J. Anal. Chem., 353, 762. 22 Chuang, T.H. (1991) Mater. Sci. Eng., Struct. Mater. Prop. Microstruct. Process., A141 (2), 169–178. 23 Nettleship, D.J. and Wild, R.K. (1990) Surf. Interface Anal., 16, 552–558. 24 Bukaluk, A. (1999) Appl. Surf. Sci., 144/145, 395–398. 25 Moser, E.M., Metzger, M., and Gauckler, L.J. (1995) Fresenius J. Anal. Chem., 353, 684–689. 26 Gondran, C.F.H. (2006) Scanning Auger Electron Spectroscopy (eds M. Prutton and M. El Gomati), John Wiley & Sons, Ltd, Chichester, p. 295. 27 Zalar, A., Baretzky, B., Dettenwanger, F., Rühle, M., and Panjan, P. (1998) Surf. Interface Anal., 26, 861–867. 28 Li, H., Jin, S., Bender, H., Lanckmans, F., Heyvaert, I., Maex, K., and Froyen, L. (2000) J. Vac. Sci. Technol. B, 18, 242–251. 29 Roberts, R.H., Prutton, M., Wilkinson, D.K., Barkshire, I.R., Hill, C.J., Pearson, P.J., Augustus, P.D., Skinner, D.K., and Stribley, K. (1998) Surf. Interface Anal., 26, 461–470. 30 Greenwood, J.C., El Gomati, M.M., and Prutton, M. (1989) EMAG-MICRO 89, vol. 98, Institute of Physics, London, p. 547. 31 Fillmore, D.K. and Krasinski, H.A. (1998) Surf. Interface Anal., 26, 109–112. 32 Palacio, C. (1999) Surf. Interface Anal., 27, 1092–1097.
33 Cao, L., Shi, F., Song, W., and Zhu, Y. (1999) Surf. Interface Anal., 28, 258–263. 34 Zhu, Y., Yao, W., Zheng, B., and Cao, L. (1999) Surf. Interface Anal., 28, 254–257. 35 Polli, A.D., Wagner, T., Gemming, T., and Rühle, M. (2000) Surf. Sci., 448, 279–289. 36 Marien, J., Wagner, T., Duscher, G., Koch, A., and Rühle, M. (2000) Surf. Sci., 446, 219–228. 37 Wagner, R., Schlatterbeck, D., and Christmann, K. (1999) Surf. Sci., 440, 231–251. 38 Kang, T.W. and Kim, T.W. (1991) Appl. Surf. Sci., 150, 190–194. 39 Uebing, C. (1990) Surf. Sci., 225, 97–106. 40 Uebing, C. (1997) ECASIA 97 (eds I. Olefjord, L. Nyborg, and D. Briggs), John Wiley & Sons, Ltd, Chichester, pp. 58–67. 41 Song, Z., Tan, D., He, F., and Bao, X. (1999) Appl. Surf. Sci., 137, 142–149. 42 Lee, H.K., Hyland, R.W. Jr, Aaronson, H.I., and Wynblatt, P.P. (1998) Surf. Sci., 408, 288–299. 43 Contini, G., Castro, V.D., Motta, N., and Sgarlata, A. (1998) Surf. Sci., 405, L 509–L 513. 44 van Agterveld, D.T.L., Palasantzas, G., and De Hosson, J.T.M. (1999) Appl. Surf. Sci., 152, 250–258. 45 Clauberg, E., Uebing, C., and Grabke, H.J. (1999) Appl. Surf. Sci., 143, 206–214. 46 Hamminger, R., Grathwohl, G., and Thümmler, F. (1983) J. Mater. Sci., 18, 353–364. 47 Prutton, M. (2006) Scanning Auger Electron Spectroscopy (eds M. Prutton and M. El Gomati), John Wiley & Sons, Ltd, Chichester, p. 201. 48 El Gomati, M.M., Peacock, D.C., Prutton, M., and Walker, C.G. (1987) J. Microsc., 147, 149. 49 Wild, R.K. (1985) Spectrochim. Acta, 40B, 827. 50 Baker, M.A. and Castle, J.E. (1993) Corros. Sci., 34, 667.
65
67
4 Electron Energy-Loss Spectroscopy (EELS) and Energy-Filtering Transmission Electron Microscopy (EFTEM) Reinhard Schneider
When, in a transmission electron microscope (TEM), high-energy electrons go through a thin specimen, there is a certain probability of both inelastic and elastic scattering; that is, the electrons can always lose energy when passing through a sample. This probability is given by the so-called inelastic scattering cross-section, which essentially depends on the energy of the electrons (and, therefore, on the high voltage applied to accelerate them), the thickness of the specimen, and on the atomic number of the atom excited (see Refs [1–3]). In conventional TEM imaging, the inelastically scattered electrons are unwanted, because they strike the image plane away from the correct image point (defocusing); this results in a diffuse background and a reduced image contrast. However, having lost energy these electrons can be used to obtain detailed information about the chemical composition of the material through which they are transmitted, its electronic structure, the chemical bonding of the elements inside, and their nearest-neighbor distances. The physical method used to measure the energy loss of electrons scattered into a specific angular range is termed electron energy-loss spectroscopy (EELS). This method can be used in analytical transmission electron microscopy (AEM) for chemical analysis at high lateral resolution, and also in reflection arrangements in surface-science physics. In the latter case, a specimen surface is bombarded with electrons of relatively low energy (from only ∼100 eV up to ∼3 keV), and the energy and angular distribution of the reflected beam are analyzed. The instrumentation necessary for reflection EELS (REELS) is usually combined with Auger electron spectroscopy (see Chapter 3), since an Auger spectrometer can usually also be operated as an energy-loss spectrometer. The following discussion is restricted to EELS in combination with TEM and scanning transmission electron microscopy (STEM), and only a brief overview of the fundamentals, experimental technique, and information gained from EEL spectra is provided. For a thorough understanding of the theoretical background and peculiarities of the EELS experiment, and its application, further reading of appropriate textbooks (e.g., Refs [4–9]) and original reports is recommended.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
68
4 EELS and EFTEM
Likewise, in order to simplify matters a degree of experience in electron microscopy is assumed; such details can be obtained elsewhere [6–10]. It should be noted that for TEM at accelerating voltages of 100–400 kV, the specimen thickness must be of the order of 10–100 nm, which requires dedicated preparation techniques (see Refs [6, 9, 11]). Nowadays, in addition to conventional TEM specimen preparation – which includes mechanical grinding and polishing, dimpling, and finally ion thinning – the use of focused ion beam (FIB) instruments [12, 13] represents a powerful tool that enables the fabrication of extended thin regions with homogeneous thickness.
4.1 Principles
Among electron-optical techniques, such as scanning electron microscopy (SEM), electron-probe microanalysis (EPMA) and TEM, there will always be a wealth of signals caused by interactions between the primary electrons and the target. These interactions can be used for materials characterization via imaging, diffraction, and chemical analysis. The different interaction processes for an electron-transparent crystalline specimen, using TEM, are shown schematically in Figure 4.1. Often, the specimen is illuminated by means of a focused electron probe so as to obtain the signals from different regions. Owing to the crystalline structure, diffracted beams also accompany the directly transmitted beam; subsequently, the energy of the transmitted electrons can be analyzed by means of an electron energy-loss spectrometer, with the electron-suffered losses yielding chemical information. Likewise, Auger electrons generated as secondary particles of the inelasti-
Figure 4.1 Schematic diagram of the interactions between high-energy electrons and matter
in TEM.
4.1 Principles
Figure 4.2 Atomistic view of interaction processes between incident high-energy electrons
and electrons of an individual atom.
cally scattered primary electrons are important for microanalysis (see Chapter 3). Secondary electrons that escape from the surface after elastic or inelastic scattering enable imaging of the topography if the electron probe is scanned across the specimen, with elastically backscattered electrons yielding images that demonstrate materials contrast. Photonic signals are also excited (including X-rays), in a process which is complementary to the emission of Auger electrons, and cathodoluminescence (CL), as a result of corresponding electronic transitions. In a closer view (Figure 4.2), the following interactions occur on an atomic scale when a material is hit by electrons. First, in addition to elastic scattering, inelastic scattering can also be observed as a result of a direct interaction between the primary electrons of energy E0 and electrons in inner shells and the valence or conduction band of the atoms in a solid. The energy transferred to the excited electron can be measured as an energy loss, E, of the transmitted electron, which reduces its kinetic energy to E0 − E. The tightly bound core electrons or the more loosely bound valence electrons may be excited to higher unoccupied energy states in the conduction band, or even be set free into the vacuum. As a consequence a hole is generated and the atom becomes ionized. After a certain dwell time of excitation, the system relaxes and the hole state in the originally excited level is filled with an electron from an outer shell. This process is accompanied by secondary processes – the emission either of Auger electrons or X-rays – which yield additional information about the chemical composition of the specimen under investigation. A full description of all the imaginable interaction processes is possible only by using quantum-mechanical physics, and will not be presented here (for details, see Refs [1, 2]).
69
70
4 EELS and EFTEM
4.2 Instrumentation
Entirely different electron spectrometers have been constructed to be used as energy analyzers combined with TEM. Owing to chromatic aberration, even an electron-optical round lens could be used to separate electrons, passing through it, with respect to their energy; however, because of such a small dispersion this is of no practical importance. Well-known types of energy analyzer include the electrostatic Möllenstedt energy analyzer [14], the Wien filter [15, 16], which is based on the combined use of crossed electrostatic and magnetic fields, single magnetic sector fields or prisms [17–22], respectively, magnetic prisms combined with an electron mirror (Castaing–Henry filter) [23–26], and systems employing magnetic prisms, such as the Ω filter [27–31] and the so-called Mandoline filter [32]. Wien and Ω filters are also used as monochromators in state-of-the-art TEM systems (e.g., FEI TITAN 80-300; ZEISS Libra; 200 CRISP) to decrease the energy distribution of the primary electrons, and offering an energy resolution of the order of 0.1 to 0.3 eV, in dependence of the beam current needed. Electrostatic filter systems, including those with electron mirrors, demonstrate certain disadvantages compared to the magnetic variety, predominantly because of the need for high-voltage connections and possible contamination of the electrodes; both effects lead to system instability. In contrast, a current of only a few amperes is required to build up a magnetic field acting as an energy analyzer, because of the Lorentz force. In addition, the electron-optical properties of magnetic sector fields have undergone intensive study (e.g., Refs [7, 33]), and their imaging aberrations can often be corrected without great difficulty; moreover, they are also simpler to construct. Magnetic energy analyzers are, therefore, commonly applied to EELS in combination with TEM/STEM. The literature includes many reports of self-made magnetic EEL spectrometers (e.g., Refs [17–31]) and, indeed, such systems have also been commercially available for many years. Although several companies have developed such spectrometers, the 90° magnetic prisms created by Gatan (parallel-detection EELS model 666 and the model Enfina) are the only units offered for TEM/STEM. In addition, because they can be easily attached to a transmission microscope below the camera chamber, these prisms are widely used. In general, a magnetic sector field has electron-optical properties comparable with those of a round lens – that is, first-order imaging of a spectrometer object point into a point in the image plane is realized. The spectrometer object can be a cross-over of the incident beam, or an intermediate image of the ray-path in the TEM. In the image plane, electrons that differ in their energies are separated owing to the dispersive action of the homogeneous magnetic field. In former times, EEL spectra generated in this way were sequentially recorded by scanning them across a slit, with the corresponding technique termed serial-detection EELS (SEELS). SEELS can be performed by continuously increasing the excitation of the magnet, or by using an additional magnetic dipole field. In order to minimize imaging aberrations – which must not be neglected because they particularly deter-
4.2 Instrumentation
Figure 4.3 Self-constructed magnetic-prism spectrometer for a TEM/STEM system.
(a) Schematic diagram of set-up; (b) Photograph of the system with the prism opened.
mine the energy resolution achievable – the best solution is to increase the accelerating voltage of the microscope in stepwise fashion. Hence, when the electron optics is well aligned for the primary electrons the path of the energy-loss electrons will also be optimized. Following the development of detectors with local sensitivity, such as photodiode arrays or charge-coupled devices (CCDs), the parallel recording of spectra has become feasible, and the respective spectrometers are referred to as parallel-detection EELS (PEELS). Because a PEELS system gathers the whole energy-loss spectrum simultaneously, it is much more efficient than SEELS. Consequently, either spectra of a better signal-to-noise ratio (SNR) can be recorded or the acquisition time can be kept short. The latter criterion is particularly important for electron-sensitive materials, such as polymers and biological samples. An example of a serial-recording EEL spectrometer is shown in Figure 4.3; this features a magnetic prism system which was constructed for a TEM/STEM of the type JEOL JEM 100S [34, 35]. Its second-order aberrations are corrected by curved pole-piece boundaries, an additional field clamp, and two extra hexapoles acting as stigmators. The electron beam can be adjusted relative to the optical axis by using several deflection coils. A magnetic round lens is positioned just in front of the prism to adapt to the different imaging and diffraction modes of the microscope. The energy analyzer is followed by a double-projective lens providing a rotation-free magnified image of the spectrum. A scintillator–photomultiplier combination is used as electron detector, and the EEL spectra are recorded serially by increasing the high voltage of the TEM stepwise. The detector output signal is recorded by means of a multichannel analyzer or personal computer for subsequent processing. Operation of a thermionic electron source results in an energy resolution of approximately 1.5 eV.
71
72
4 EELS and EFTEM
In principle, energy-analyzer systems can be designed such that their electronoptical properties do not limit the energy resolution attainable. That is, their intrinsic energy resolution is much better than the energy width of the primary electron beam, which is of the order of approximately 1.5–2.5 eV for a tungsten hairpin cathode, approximately 1 eV for a LaB6 cathode, approximately 0.7 eV for a Schottky field emitter, and 0.3–0.5 eV for a pure cold-field emitter. With regard to the instrumentation, some of the energy-analyzer systems mentioned above, namely the Castaing–Henry filter, the Ω filter and the Mandoline filter, are not only spectrometers but also image-forming systems enabling energyfiltered imaging or diffraction with TEM – that is, with stationary illumination of the specimen. Thus, they enable the formation of a two-dimensional (2-D) image or diffraction pattern using energy-loss electrons within a specified energy range. In addition to the contrast enhancement obtained by filtering out all of the inelastically scattered electrons, this also enables the imaging of element distribution. Nevertheless, for a STEM system equipped solely with an EEL spectrometer, energy-filtered imaging is possible simply by recording the electrons of a certain energy loss passing the slit while scanning the electron probe over the specimen in a rectangular field. All of the energy-filter systems referred to above are incorporated into the electron-microscope column, usually between the diffraction lens and the following intermedium lens. They are, therefore, known as in-column filters, which are also commercially available in the form of different TEM systems, for example, as the Castaing–Henry filter in the ZEISS EM 902 and as Ω filters in the LEO1) EM 912, EM 922 (e.g., Ref [36]) and in the JEOL JEM 2010 FEF and 3010 FEF [37]. The operation of only a single magnetic prism, which is not part of the column, can also be extended to energy-filtered TEM. A corresponding post-column filter is offered by Gatan; this is the well-known and widespread Gatan imaging filter (GIF), which can be attached to any TEM/STEM system. The two concepts – the in-column and post-column filters – are compared schematically in Figure 4.4. It should be noted that the post-spectrometer optics of the GIF consists of multipoles only (viz., quadrupoles), preferably for imaging, and hexapoles for correcting image aberrations. The in-column filter shown in Figure 4.4 might be that of the EM 922 Omega or JEM 2010 FEF microscopes, respectively.
4.3 Qualitative Spectral Information
The EEL spectrum measured is a plot of the intensity of the inelastically scattered electrons as a function of energy loss at a certain spectrometer collection angle, which is defined by an aperture in front of the energy analyzer. Typically, the spectrum can be divided into two regions: (i) the low-loss region, ranging from 1) These instruments are now sold by the Zeiss company, as the brand name LEO no longer exists.
4.3 Qualitative Spectral Information
Figure 4.4 Arrangement of in-column (left) and post-column imaging energy filters (right) in
TEM [6].
0 eV to approximately 50 eV; and (ii) the high-energy region lying beyond that (50 to approximately 2000 eV). The upper limit of energy loss is determined by the signal intensity, which generally decreases exponentially with increasing energy loss. This may be much higher than 2 keV, particularly when a PEELS is used, because of its better detection efficiency. The characteristic losses that provide information on the elemental composition arise from the ionization of inner shells. These appear as edges superimposed on a background caused by low-loss excitations (plasmons, single-valence electron excitations), or by the tail of an ionization edge of lower energy. The reason for the appearance of edges instead of peaks is that, usually, not only the threshold energy is transferred to the inner-shell electrons, but also some additional kinetic energy. A spectrum of silicon carbide is shown as an example of a typical EEL spectrum in Figure 4.5. Here, the spectral features are correlated to the electronic structure of this semiconductor material. In the low-loss region, besides the zero-loss peak of highest intensity there is only one relatively sharp plasmon peak at approximately 21 eV, whilst on its high-energy side the signal intensity falls dramatically. In order to make visible the characteristic ionization edges following in the region of higher losses (viz., the Si-L23 and C-K edges), the vertical scale must be expanded by a factor of a few hundred. Owing to the high-energy resolution of approximately 1 eV, fine structures can be observed near the edge onset, especially for the Si-L23 edge. Such energy-loss near-edge structures (ELNES) occur because the individual inner-shell electron can be excited from its ground state into unoccupied states in the conduction band (cf. Section 4.3.3).
73
74
4 EELS and EFTEM
Figure 4.5 EEL spectrum of silicon carbide and corresponding energy-level scheme.
4.3.1 Low-Loss Excitations
In the low-loss region, the predominant feature is the zero-loss peak that itself contains no useful analytical information. This is also termed the elastic peak, because mainly elastically scattered electrons contribute to it, as well as those transmitted by the thin specimen without interaction. Strictly speaking, however, the term zero-loss peak is incorrect, because electrons of very low energy losses (typically 100 meV and lower) caused by the excitation of phonon vibrations also form part of that peak. On account of the energy width of the primary electrons, and of the finite energy resolution of the spectrometer system used, they are not resolved on the high-energy tail of the peak. Even by adapting a monochromator to the illumination system of the TEM unit to reduce the primary energy width would barely allow their detection. Commonly, the full-width at half maximum (FWHM) of the zero-loss peak is accepted as a measure of the energy resolution of an EEL spectrometer. The region from approximately 5 to 50 eV contains mostly Gaussian-shaped peaks, which might be of different origins [38]. For materials with free electrons – such as metals – these clearly derive from the collective oscillations (plasmons) of these electrons. If, however, the specimen is composed of a material in which free electrons are not abundant, an alternative mechanism is present which
4.3 Qualitative Spectral Information
involves the transfer of energy to single valence electrons. Hence, this effect of single-electron excitation is rather important for solids such as semiconductors and insulators, which also generate plasmon-like peaks in the range of 15–30 eV. In addition, both interband and intraband transitions can also cause extra resonances in that energy region. Plasmon oscillations are excited not only in the bulk of the specimen but also on the surface. The proportion of surface plasmons increases when the specimen becomes thinner, and their excitation energy is approximately half that of the corresponding volume plasmon. Analyzing the plasmon energy, Epl, is a local probe of the density of free electrons, Ne, because of the dependence E pl ∝ N e1/2 [39]. In special cases this phenomenon can be applied to materials analysis; this has been demonstrated on Al−Mg and Al−Zn alloys [40, 41] for which the displacement of plasmon energy enables phase identification. More recently, the region close to the right of the zero-loss peak has become of increasing interest in so-called band-gap spectroscopy [42], which directly yields data on electronic properties as, for example, shown for (In,Ga,Al)N compounds [43]. When inelastic electron scattering is treated in terms of particle collisions, the individual events are independent and obey Poisson statistics. Yet, because the number of scattering processes increases with the specimen thickness, t, the incident electron can undergo multiple losses. Thus, the thicker the specimen, the more interactions generating plasmon oscillations the incident electron will have, and this will result in the occurrence of equidistant peaks of multiples of the plasmon energy. The dependence of plural scattering on specimen thickness can be used to measure the thickness in the region illuminated by the electron beam. It can be shown that the relative thickness t/λ is given by t ⎛ Itot ⎞ = ln ⎜ ⎟ ⎝ I0 ⎠ l where λ is the total mean free path for all inelastic scattering, I0 is the intensity integrated under the zero-loss peak, and Itot that under the whole spectrum recorded up to approximately 200 eV. The total inelastic mean free path, λ, must be known to determine the absolute thickness, t. Corresponding λ values can be calculated when the composition of the material is known [4]. Figure 4.6a shows a typical low-loss spectrum taken from boron nitride (BN). The structure of BN is similar to that of graphite (i.e., sp2-hybridized carbon), and for this reason the low-loss features are quite similar and comprise a distinct plasmon peak at approximately 27 eV that is attributed to collective excitations of both π and σ electrons, whereas the small peak at 7 eV derives only from π electrons. Besides the original spectrum, the zero-loss peak and the low-loss part derived by deconvolution are also shown in the figure. Thus, by calculating the ratio of the signal intensities Itot and I0, a relative specimen thickness t/λPl of approximately unity was found. Owing to this specimen thickness there is slight indication of a second plasmon.
75
76
4 EELS and EFTEM
Figure 4.6 Low-loss spectra of (a) BN and thickness determination, and (b) CaTiSiO5 with outer-shell ionization edges.
That even low-loss spectra can be complex in appearance is recognizable from that of the silicate CaTiSiO5, as shown in Figure 4.6b. Here, the problem arises of the superposition of regular plasmon or single-valence electron excitations, and excitations of tightly bound outer-shell electrons with extremely low binding energy. Respective ionization edges are the Ca-M23 edge at 25 eV and the Ti-M23 edge at 35 eV, which appear partly overlapped. The peaks at about 15 and 22 eV are presumably caused by interband transitions and plasmons, respectively. It should be noted that low-loss spectra are basically connected to the optical properties of materials. This is because, for small scattering angles, the energydifferential cross-section dσ/dE – in other words, the intensity of the EEL spectrum measured – is directly proportional to Im{−1/ε(E,q)} [4]. Here, ε = ε1 + iε2 is the complex dielectric function, E the energy loss, and q the momentum vector. Owing to the comparison to optics (|q| = 0), the above-quoted proportionality is fulfilled if the spectrum has been recorded with a reasonably small collection aperture. When Im(−1/ε) is gathered, its real part can be determined by the Kramers–Kronig analysis [4] and subsequently such optical quantities as refraction index, absorption coefficient, and reflectivity may be assessed.
4.3 Qualitative Spectral Information
4.3.2 Ionization Losses
In general, EELS is particularly sensitive to light elements, because the ionization cross-sections increase with decreasing atomic number. Because of this, energyloss spectroscopy and energy-dispersive X-ray spectroscopy (EDXS) complement each other reasonably well, as the detection sensitivity of EDXS behaves in an opposite manner (see Chapter 18). Most of the transmitted electrons, having suffered energy losses via ionization and plasmon excitations can, moreover, be readily collected owing to their strong forward scattering (see Ref. [6]); rather, they are scattered into an angular range of typically 10−2 rad. Thus, the features that are most important for materials analysis are the inner-shell edges, because the ionization energies are characteristic for the chemical elements. Consequently, nearly all elements can principally be detected, and in the energy range up to 2 keV there appear series of K edges for Li to Si, L23 edges for Al to Sr, and M45 edges for Rb to Os (see Ref. [44]). The specific shape of the edge varies with the edge type (K, L, M, etc.), and depends heavily on the electronic structure and chemical bonding. Usually, K shells arising from 1s to 2p transitions have a so-called hydrogenic-like shape which is characterized by a steep signal increase at the threshold energy. L-shell edges (p→s and p→d transitions) occur either as rounded profiles or nearly hydrogenic-like with sharp, so-called white lines at the edge onset. The latter can be observed for elements of atomic number 19 ≤ Z ≤ 28 and 38 ≤ Z ≤ 46. In EELS, the term “white line” is used in analogy to X-ray absorption spectroscopy, where the features were historically first observed as severe bright lines on photographs. M edges tend to appear as delayed signals, although they also show up white lines when empty d or f states are present. An example of a high-energy loss spectrum is shown in Figure 4.7; this is an EEL spectrum in the range from 400 to 900 eV, taken from a thin region of a BaTiO3 particle at 200 keV primary energy. The ionization edges of all three constituting elements are visible, namely the Ti-L23, the O-K, and the Ba-M45 edges. Whilst the K edge of oxygen is sawtooth-shaped, the two others have more or less
Figure 4.7
EEL spectrum of BaTiO3, exhibiting ionization losses.
77
78
4 EELS and EFTEM
well-resolved peaks near the edge onsets, indicating individual energy levels (L3 and L2 for Ti, M5 and M4 for Ba). In particular, the Ba-M45 edge demonstrates the effect of white lines caused here by the excitation from initial 3d states to unfilled 4f states. In general, besides the gross edge profiles, pronounced fine structures can also occur in the region of the very edge onset up to approximately 30 eV beyond it. These so-called ELNES are caused by solid-state effects (cf. Section 4.3.3). In Figure 4.7 the O-K edge, in particular, shows such ELNES detail, and additional weak modulations of the signal intensity reaching some 100 eV above the edge threshold can also be seen. These oscillations, which are referred to as extended energy-loss fine structures (EXELFS), provide information on the number of nearest neighbors and their distances. Each inner-shell edge has a high-energy tail, which is the background signal on which the next edge is superimposed. In front of the first edge occurring in an EEL spectrum, the background contribution comes from low-loss excitations of relatively high intensity. Thus, several different interaction phenomena give rise to the background signal and it is, therefore, impossible to calculate it ab initio. It was, however, proven experimentally that for a limited energy range the background intensity, IB, can be fitted by a power law: IB = A ⋅ E − r where E is the energy loss and both A and r are constants. Both of the fit parameters vary with the mass thickness of the material investigated, the accelerating voltage, and the acceptance angle of the spectrometer. Typical values of the exponent r range from 2 to 5, whereas A can vary dramatically. EELS analyses can be complicated, or even fail, if the specimen is so thick that intense multiple scattering occurs. For spectra recorded from extremely thick regions (a few 100 nm), the characteristic ionization edges vanish in the high background. Thus, the inner-shell edge measured often does not represent single but rather multiple scattering resulting from the combined excitations of innershell electrons and valence electrons. This behavior is shown schematically in Figure 4.8, where the effect of the convolution of a hydrogenic-like edge with the low-loss spectrum is indicated. The single-scattering curve of an edge can be gained from the EEL spectrum measured by deconvoluting it by means of the corresponding low-loss spectrum. Fast Fourier routines are preferably applied to spectrum processing, because the EEL spectra involve many data points that are equally spaced in energy. Deconvolution is possible by use of either the Fourier-log or Fourier-ratio method, and by multiple least-squares fitting (as described in greater detail elsewhere [4, 6]). An example of the determination of the singlescattering profile via the Fourier-ratio method – that is, for the Ti-L23 edge taken from a thin Ti film – is shown in the lower part of Figure 4.8. On the left, the original measured and deconvoluted spectra are compared. Whilst it is clear that the near-edge region remains almost unaffected by multiple scattering, there is however a strong contribution of multiple scattering starting at approximately 15 eV above the edge onset.
4.3 Qualitative Spectral Information
Figure 4.8 Convolution of inner-shell losses and low-loss features, schematically (above) and
as observed experimentally (below) for the Ti-L23 edge.
4.3.3 Fine Structures
The occurrence of fine structures has already been noted in the sections on spectral information and ionization losses (Sections 4.3 and 4.3.2). In the following text, some principal considerations are made regarding the physical background and possible applications of both types of respective feature – that is, near-edge and extended energy-loss fine structures (ELNES/EXELFS). A wealth of more detailed information on their use is available in both textbooks [4, 6] and monographs [45–47]. In Figure 4.9a, the Al-K edge (onset at about 1560 eV) recorded from Al4C3 is shown, and the specific energy ranges of the appearance of ELNES and EXELFS are marked. In the ELNES region reaching from the edge onset up to approximately 30 eV above, there are only two distinct peaks visible, at 1566 and 1578 eV. EXELFS, which are much weaker in intensity, start at approximately 1590 eV and extend over the whole high-energy tail of the edge shown. Usually, EXELFS modulations can be observed up to some 100 eV beyond the ionization threshold. Near-edge structures superimposed on the general atomic edge profile are caused by transitions of inner-shell electrons of a particular atom to the lowest
79
80
4 EELS and EFTEM
Figure 4.9 Fine structures of inner-shell edges. (a) Al-K edge of Al4C3 exhibiting ELNES and
EXELFS; (b) O-K ELNES of different titanates and Ti-containing silicates.
unoccupied states in the solid. Thus, ELNES probes the density of empty states above the Fermi level at the site of the excited atom. This local density of states (DOS) is related not only to the atom excited but also to the arrangement, number, and types of atoms in its environment that determine the charge transfer and chemical bonding. All of these peculiarities influence such ELNES details as the number, energy width, and intensity of individual peaks, and also the threshold energy of the ionization edge. It is mainly the first coordination shell which affects the ELNES features observed. Hence, near-edge structures represent a so-called coordination fingerprint which, for the combination of EELS and STEM/TEM, enables an analysis of the chemical bonding of a particular compound with high lateral resolution. A practical approach to phase identification by ELNES analysis is to compare the fine-structure features recorded from the unknown phase with those measured from reference materials of known structure and stoichiometry (the fingerprint method). Many reports have been made containing a multitude of data on the several materials systems available; the EEL spectra published in reference books [44, 48] are also helpful in this context. It should be noted that a comprehensive ELNES study is possible only by comparing experimentally observed structures with those calculated [45–47]. This represents an extra field of investigation, and different procedures based on molecular orbital approaches [49–51], multiple-scattering theory [52, 53], or band-structure calculations [54, 55] can be used to compute the densities of electronic states in the valence and conduction bands. Because near-edge structures reflect the nearest chemical neighborhood of an atomic species in a solid, for single-crystalline materials they can also depend on the orientation of the specimen. A classical example is graphite, for which the height of the pre-peak of the C–K edge caused by transitions from 1s states into unoccupied π* states is strongly orientation-dependent [56]. The π* peak intensity reaches a maximum when the incident electron beam is parallel to the c-axis of
4.3 Qualitative Spectral Information
graphite. Because of the structural similarity of boron nitride to graphite, its B-K edge behaves in the same manner (cf. Figure 4.12a). Thus, such orientation dependence yields additional information on the anisotropy of chemical bonding and band structure. If these effects must be avoided, however, it is recommended that the ELNES data are collected from polycrystalline (or even amorphous) samples, if available, of the individual material under investigation. Neglecting theoretical treatments of ELNES analyses, a practical application result is presented in Figure 4.9b in which the O-K ELNES of different titanates and Ti-containing silicates are compared [57]. The fine structures shown were acquired from reference materials at an energy resolution of approximately 0.8 eV, using a PEELS (Gatan Model 666) attached to a TEM/STEM Philips CM20 FEG running at 200 keV, with a thermally assisted field-emission gun. All ELNES features were obtained from the raw data after sharpening the resolution (deconvolution with the zero-loss peak), subtracting the background (power-law fit), and subsequently deconvoluting with the low-loss spectrum in order to obtain the single-scattering profile. The O-K ELNES is extremely sensitive to the nearestneighbors surroundings, which is also known from other titanates [58]. The Ba6Ti17O40 phase can be distinguished from BaTiO3 by its near-edge details at the very edge onset, and by those lying in the range from 540 to 550 eV. The O-K ELNES of the silicate fresnoite (Ba2TiSi2O8) is most different – its onset is shifted to higher energies and delayed. There is, moreover, only one broad maximum at approximately 542 eV. Such pronounced differences in the ELNES can be used for phase identification, as demonstrated in Figure 4.10. In this case, Figure 4.10a shows a TEM bright-field image of a BaTiO3 ceramic processed via liquid phase sintering at about 1350 °C, with an excess of 1 mol% TiO2. Between two BaTiO3 grains an unknown phase was found, where EELS (cf. Figure 4.10b) clearly indicates the presence of Ba2TiSi2O8. In the series of spectra recorded along a line of approximately 50 nm
Figure 4.10 Chemical-bond analysis of an additional phase in a BaTiO3 ceramic using
differences in the O-K ELNES.
81
82
4 EELS and EFTEM
Figure 4.11 Characteristics of the Co-L23 ELNES of CoO as a function of energy width of primary electrons. (a) Standard Tecnai F20 (ΔE = 0.7 eV); (b) F20 with improved high tension and high-resolution Gatan imaging
filter (HR-GIF) (ΔE = 0.5 eV); (c) Monochromized F20 and HR-GIF (ΔE = 0.2 eV). Illustration courtesy of B. Freitag, FEI Electron Optics.
length in the STEM mode, the local transition from BaTiO3 to Ba2TiSi2O8 is clearly visible from the change of the O-K ELNES. Differences in the Ti-L23 edge are also apparent. The Ba2TiSi2O8 detected is presumably a reaction product introduced by the impurity SiO2. For a detailed study of the bonding and atomic configuration of elements in chemical compounds by means of ELNES, the use of electron microscopes equipped with monochromators or cold field-emission cathodes, respectively, is recommended, because they enable a much better energy resolution (see also Section 4.2). This is demonstrated in Figure 4.11, which shows the results of measurements of the Co-L23 ELNES of cobalt monoxide (CoO) for different experimental set-ups (B. Freitag, personal communication; [59]). In each case, a Tecnai F20 microscope with a field-emission source (Schottky emitter) was used, but the energy width of the primary electrons had been reduced from originally 0.7 eV over 0.5 eV to 0.2 eV (FWHM of zero-loss peak) by improving the high-tension stability and the use of a high-resolution version of a Gatan imaging filter, and finally the additional application of a Wien filter as monochromator in combination with the field emitter. If the Co-L23 spectra are recorded with the monochromized TEM, much more spectral features can be revealed, in particular at the L3 white line. Here, a more or less broad shoulder in the spectrum taken on the conventional system (Figure 4.11, spectrum a) changes into a fourfold split structure (Figure 4.11, spectrum c), which has also been found by X-ray absorption spectroscopy [60]. Finally, it should be noted that, analogous to the EXAFS effect observed in X-ray absorption [61, 62], EXELFS are weak modulations (cf. Figure 4.9a), still observable a few hundred eV away from the edge onset, and occur because the excited electron wave has approximately the same wavelength as the atomic spacing of the substance transmitted. Consequently, the wave can be diffracted by neighboring
4.4 Quantification
atoms, and can return to interfere with the outgoing wave. The interference can be either constructive or destructive, yielding an oscillating part of the intensity following the ionization edge. Standard EXAFS theory can be used to interpret extended fine structures in EELS [63]. Thus, EXELFS analysis enables interatomic distances between the first-neighbor atoms to be measured, yielding the so-called radial distribution function (RDF). For an ideal single crystal, the RDF would consist of a series of delta functions which can be attributed to discrete values of shell radii given by the crystal structure. Even bond lengths in a particular direction can be determined when the collection angle of the transmitted electrons and the orientation of the specimen are precisely chosen.
4.4 Quantification
In addition to the qualitative analysis of nearly all elements of the Periodic Table, EEL spectra also enable determination of the concentration of a single element which is part of the transmitted volume and hence give rise to a corresponding ionization edge. As in all comparable spectroscopic techniques, for quantification the net edge signal – which is related to the number N of excited atoms –must be extracted from the raw data measured. The net intensity Ik of the kth ionization shell of an individual element is directly connected to this number, N, multiplied by the partial cross-section of ionization σk(Δ,α) and the intensity I0 of the incident electron beam; that is: Ik (Δ,a ) = Ns k (Δ,a )I 0 where Δ is the energy window for integrating the signal intensity and α the collection semi-angle of the spectrometer. Different approaches are used to calculate partial cross-sections. Calculations which make use of atomic wave functions are usually based either on a hydrogenic model [64, 65] or a Hartree–Slater model [66]. Neither model, however, can be used to predict the effects resulting from electronic transitions to bound states. The Hartree–Slater cross-sections are in very good agreement with experimental findings, although substantial computational effort is required. The results of calculations by use of the hydrogenic model agree with the Hartree–Slater values to an accuracy of better than 20%, which is within the experimental error. The background signal, IB, contributing to the kth inner-shell edge must be subtracted from the total energy-loss intensity to obtain the signal, Ik, of the ionization loss itself. As already mentioned in Section 4.3.2, the background often follows a power law (IB = A·E−r). This fit can particularly be used to extrapolate the background in the higher-loss region; for inner-shell losses at approximately 100 eV and below, the use of a polynomial fit is sometimes more suitable. The procedure of background extrapolation and subtraction is demonstrated for the B-K edge of BN in Figure 4.12a where, in addition to the recorded edge profile, the extrapolated background and the net edge are shown.
83
84
4 EELS and EFTEM
Figure 4.12 Basic procedure for EELS quantification exercised for BN. (a) Background
extrapolation and subtraction; (b) Determination of net counts in a certain energy window and calculation of corresponding partial cross-sections.
The absolute number of atoms per unit area making the kth ionization edge occur is given by: N=
I k ( Δ, a ) s k ( Δ, a ) ⋅ I 0
Often, only the concentration ratio of two elements A and B is of interest; this is, therefore: N A I A ( Δ, a ) ⋅ s B ( Δ, a ) = N B IB ( Δ , a ) ⋅ s A ( Δ , a ) This type of estimation of the relative concentration is the most widely used method for quantitative EELS analysis. It is advantageous because the dependence on the primary electron current, I0, is canceled out; this is not easily determined in a transmission electron microscope under suitable analytical conditions. Furthermore, in comparison with other methods – for example, AES and EDXS – there is no need for additional terms to correct for any atomic number effects, the specific yield of the interaction process, absorption, and so on. Figure 4.12b shows the graphical results from a complete quantification procedure for BN, including the determination of the net B-K and N-K edges and the calculation of their partial cross-sections (Hartree–Slater model), yielding a B/N ratio of 0.97 ± 0.16 at Δ = 70 eV and α = 10 mrad. The accuracy achievable by the ratio method amounts to approximately ±5–10 atom% when ionization edges of the same type are used – that is, only K edges or only L edges – whereas the error in quantification increases to ±15–20 atom% for the use of dissimilar edges. However, an improvement of the quantification accuracy up to about 1 atom % is possible if standards are used. It is of no use denying that energy-loss spectroscopy cannot usually be used to detect traces of elements in a matrix. The attainable detection sensitivity is highly
4.5 Imaging of Element Distribution
dependent on the specific elemental composition of the material to be analyzed. For example, the minimum detectable mass fraction amounts to atom% levels for the identification of B, N, O, and so on, in a thin Si specimen, because the corresponding ionization edges ride on the high-energy tail of the Si-L23 edge which limits the sensitivity [67]. In contrast to that, it is possible to find solely approximately 0.1 atom% of Li (onset at 55 eV) or Al (73 eV) in silicon; this corresponds to an absolute number of few thousand atoms. Under special circumstances (a high primary current of the order of 1.6 × 105 A s cm−2, parallel-recording spectrometer), EELS seems, nevertheless, to be sensitive to a few single atoms – for example, the detection of Fe atoms in a 10 nm-thick carbon film [68, 69].
4.5 Imaging of Element Distribution
The combination of EELS and TEM affords different experimental facilities for the imaging of element distributions (see also Section 4.2), depending on the electron optics of the microscope and the particular type of spectrometer. The physical fundamentals, experimental set-up, and application possibilities of elementspecific imaging, such as energy-filtered TEM (EFTEM), are dealt with elsewhere (see Refs [4, 6, 7]). In the following, only a brief introduction and some experimental examples are provided. When the electron microscope can be operated in STEM mode, two-dimensional (2-D) energy-selective images can easily be gathered by applying the signal of a serial-detection spectrometer to the brightness control of a cathode-ray tube. In general, for an element map three individual images must be taken: two background images (also termed pre-edge images), and one post-edge image. Hence, this procedure – which is referred to as the three-window method (see Refs [4, 6, 7]) – enables the determination of the net contribution of an inner-shell edge to image brightness after background extrapolation and subtraction. The image processing required is quite similar to the handling of an ionization edge in EELS to obtain the net edge profile, although local differences between low-loss spectra can also be used for element-specific imaging. In the case of Al layers on SiO2, for example, in cross-section the Al-containing regions appear bright because of a plasmon energy of approximately 15 eV compared to 23 eV for silicon oxide for imaging with a 5 eV wide energy-selecting slit centered at 15 eV. When a PEELS is run on a STEM system, the element distribution can be visualized along a line by recording series of spectra (see Figure 4.11b). For STEM, the lateral resolution is essentially fixed by the diameter of the electron probe used. Thus, subnanometer – or even atomic – resolution can be obtained for scanning microscopes equipped with field-emitter electron sources, making the combination of EELS and STEM a sophisticated means of interface analysis [70]. The attachment of an imaging filter to a transmission microscope enables energy-filtered TEM to be performed. In this case, the method used to generate an elemental map is the same as that described above (the three-window
85
86
4 EELS and EFTEM
Figure 4.13 Energy-filtered TEM (three-
window method) of the carbon distribution in the region of Cu-filled carbon nanospheres. C-K edge and choice of energy
windows (above), individual energy-filtered images together with the resulting carbon map and the corresponding TEM bright-field image (below).
technique), except that each of the three individual images is recorded simultaneously, while the specimen is illuminated with a parallel stationary electron beam. In Figure 4.13, an example is provided for mapping the C distribution in the region of carbon nanospheres filled with copper particles of approximately 50 nm in diameter. These multishell graphitic carbon nanospheres were synthesized by pyrolysis using Cu-phthalocyanine as the precursor material [71]. TEM samples were prepared by dispersing the nanospheres in ethanol and nebulizing this dispersion on a Cu grid with holey carbon film. For EFTEM, a microscope LEO 922 Omega was used at 200 kV accelerating voltage. The experimental conditions for
4.5 Imaging of Element Distribution
Figure 4.14 EFTEM images and corresponding line profiles of oxidized AlAs/(Al,Ga)As
multilayers.
the three-window technique regarding the choice of energy windows are illustrated in a corresponding C-K edge spectrum (Figure 4.13, top panel). Pre-edge images were recorded for 10 s at approximately 235 eV and 265 eV, and the post-edge image at 312 eV. After extrapolating and subtracting the background image from the post-edge image, the resultant map clearly showed the carbon distribution (Figure 4.13, lower panels). EELS line profiles can also be gained from EFTEM images a posteriori, by digitally defining a line of interest and representing the element-specific signal along the line chosen. For materials systems having layers of homogeneous thickness, the signal can be integrated in the direction parallel to the interfaces, thus yielding an improved SNR. This is shown in Figure 4.14 for a cross-section specimen of AlAs/(Al,Ga)As multilayers deposited on a (001)GaAs substrate by metal–organic vapor-phase epitaxy [72]. Here, the AlAs layers in particular are oxidized by exposure to air, which is imaged in the elemental maps and corresponding line profiles. An especially dedicated technique is the so-called spectrum-image method [73], which comprises a series of spectra measured in a STEM not along a line but rather in a rectangular scanned field. This results in a data cube in which the upper plane is the scanned x–y area and the third axis is the energy-loss spectrum. Complementarily, the image spectrum means a series of energy-filtered TEM images taken at such small energy increments that for every point in the x–y plane of view, a full spectrum can be reconstructed afterwards. Although both the spectrum– image and image–spectrum methods provide a comprehensive insight into the specimen, both can be limited by the large data capacity required and the extended time needed for data acquisition and off-line processing. Energy filtering enables the imaging not only of the distribution of an element, but also of its chemical bonding when ELNES features are used. Corresponding chemical-bond maps can be obtained in both experimental arrangements – that is, the combined SEELS/STEM or imaging filter and TEM. In each case, the slit width of the spectrometer system must be so small that a single fine-structure detail contributes to the signal. In order to obtain reliable results, the three-window method must normally be used, however.
87
88
4 EELS and EFTEM
Figure 4.15 Chemical-bond mapping. (a) Comparison of EEL spectra recorded from matrix
and SiC-fiber with the window for energy filtering shown; (b) Map of oxidic-bound Si.
Chemical-bond mapping has been practiced on an SiC(Nicalon)-fiber reinforced borosilicate glass (Duran) of high SiO2 content, as illustrated in Figure 4.15 [74]. The differences in the Si-L23 ELNES (Figure 4.15a) were used to image the oxidicbound silicon in the interfacial region between the fiber and the matrix, where an additional C-rich reaction layer is present. Individual STEM images were taken with a 3 eV slit width at approximately 95 eV (background B), at 102 eV (first dominant peak of the Si-L23 ELNES of SiC), and at 107 eV (first dominant peak of the Si-L23 ELNES of SiO2). The 3-D plot of the image intensity in Figure 4.15b is obtained by subtraction of the partial image recorded at 102 eV from that at 107 eV – that is, the value of the resulting signal is higher for regions with strongly oxidized silicon. In Figure 4.15b, the zone of minimum signal in the center corresponds to the carbon layer. On the fiber side of the carbon layer a region is visible where the signal has a local maximum hinting at oxidic-bound Si. The mean signal level is higher in the matrix (right) owing to the presence of SiO2.
4.6 Summary
In recent years, EELS, in conjunction with TEM/STEM, has become a powerful means for analyzing materials, particularly interlayers and interfaces in nanostructured materials, at a lateral resolution of about 1 nm or even better. This analytical technique yields both qualitative and quantitative information on specimen composition, for which its thickness must be of the order of some 10 nm, requiring a major effort of sample preparation. In addition, the chemical bonding of an element detected within the transmitted volume can be characterized by ELNES. Investigating low-loss spectra in more detail at high-energy resolution (valence and band-gap spectroscopy) enables the determination of electronic and optical properties of materials. The element distribution can be imaged along a line, or
References
in 2-D fashion using energy-filtered (S)TEM, where subnanometer resolution is achievable. In the future, the limits of both energy resolution and lateral resolution will be pushed further. Notably, projects with the names of SESAME [32, 75] and TEAM [76, 77] are currently in progress, with the aim of developing aberrationcorrected, high-resolution TEM and energy-filter systems with resolution values below 1 Å and 0.2 eV.
References 1 Bethe, H. (1930) Ann. Phys., 5, 325. 2 Inokuti, M. (1971) Rev. Mod. Phys., 43, 297. 3 Schattschneider, P. (1986) Fundamentals of Inelastic Electron Scattering, SpringerVerlag, Vienna, New York. 4 Egerton, R.F. (1996) Electron Energy-Loss Spectroscopy in the Electron Microscope, 2nd edn, Plenum Press, New York, London. 5 Hren, J.J. and Goldstein, J.I. (eds) (1979) Introduction to Analytical Electron Microscopy, Plenum Press, New York. 6 Williams, D.B. and Carter, C.B. (1996) Transmission Electron Microscopy – A Textbook for Materials Science, Plenum Press, New York, London. 7 Reimer, L. (ed.) (1995) Energy-Filtering Transmission Electron Microscopy, Springer-Verlag, Berlin. 8 Reimer, L. (1989) Transmission Electron Microscopy: Physics of Image Formation and Microanalysis, Springer-Verlag, Berlin. 9 Bethge, H. and Heydenreich, J. (eds) (1987) Electron Microscopy in Solid State Physics, Elsevier. 10 Amelincks, S., van Dyck, D., van Landuyt, J., and van Tendeloo, G. (eds) (1997) Electron Microscopy: Principles and Fundamentals, VCH Verlagsgesellschaft mbH, Weinheim. 11 Anderson, R.M. and Walck, S.D. (eds) (1997) Specimen Preparation for Transmission Electron Microscopy of Materials IV, Materials Research Society, Pittsburgh. 12 Gianuzzi, L.A. and Stevie, F.A. (eds) (2005) Introduction to Focused Ion Beams: Instrumentation, Theory, Techniques and Practice, Springer Verlag GmbH & Co., Berlin.
13 Orloff, J., Swanson, L., and Utlaut, M. (2003) High Resolution Focused Ion Beams: FIB and Its Applications, Kluwer Academic/Plenum Publishers, New York. 14 Metherell, A.J.F. and Cook, R.F. (1972) Optik, 34, 535. 15 Kramer, J., van Zuylen, P., and Hardy, D.F. (1974) Proceedings 8th International Congress on Electron Microscopy, Canberra, vol. 1, p. 372. 16 Batson, P.E. (1984) Proceedings 42nd EMSA Meeting, San Francisco, p. 558. 17 Wittry, D.B. (1976) J. Phys. D, 2, 1757. 18 Crewe, A.V., Isaacson, M., and Johnson, D. (1971) Rev. Sci. Instrum., 42, 411. 19 Pearce-Percy, H.T. (1976) J. Phys. D, 9 (135), 515. 20 Egerton, R.F. and Lyman, C.E. (1976) Developments in Electron Microscopy and Analysis (ed. J.A. Venables), Academic Press, London, New York, p. 35. 21 Engel, W. (1978) Proceedings 9th International Congress on Electron Microscopy, Toronto, vol. 1, p. 48. 22 Andrew, J.W., Ottensmeyer, F.P., and Mortell, E. (1978) Proceedings 9th International Congress on Electron Microscopy, Toronto, vol. 1, p. 40. 23 Castaing, R. and Henry, L. (1964) J. Microsc., 3, 133. 24 Jouffrey, B. (1972) Proceedings 5th European Conference on Electron Microscopy, Manchester, p. 190. 25 Henkelman, R.M. and Ottensmeyer, F.P. (1974) J. Microsc., 102, 79. 26 Ichinokawa, T. (1968) Jpn. J. Appl. Phys., 7, 799. 27 Rose, H. and Plies, E. (1974) Optik, 40, 336. 28 Zanchi, G., Perez, J.P., and Sevely, J. (1975) Optik, 43, 495.
89
90
4 EELS and EFTEM 29 Pearce-Percy, H.T., Krahl, D., and Jäger, J. (1976) Proceedings European Conference on Electron Microscopy, Jerusalem, p. 348. 30 Krahl, D., Herrmann, K.-H., and Kunath, W. (1978) Proceedings 9th International Congress on Electron Microscopy, Toronto, vol. 1, p. 42. 31 Zanchi, G., Sevely, J., and Jouffrey, B. (1977) J. Microsc. Spectrosc. Electron., 2, 95. 32 Rose, H. (1999) Ultramicroscopy, 78, 13. 33 Hawkes, P.W. and Kasper, E. (1996) Principles of Electron Optics, vol. 2, Academic Press, London. 34 Rechner, W. (1981) Veröff. 10. Tagung Elektronenmikroskopie, Leipzig, S. 254. 35 Schneider, R. and Rechner, W. (1992) Mikrochim. Acta, 12 (Suppl.) 197. 36 Probst, W., Benner, G., and Mayer, J. (1993) Inst. Phys. Conf. Ser., 130, 295. 37 Tsuno, K., Kaneyama, T., Honda, T., Ishida, Y., Tsuda, K., Terauchi, M., and Tanaka, M. (1998) Proceedings International Congress on Electron Microscopy, Cancun, vol. I, p. 253. 38 Raether, H. (1980) Excitations of Plasmons and Interband Transitions by Electrons, Springer-Verlag, Berlin. 39 Ritchie, R.H. (1957) Phys. Rev., 106, 874. 40 Spalding, D.R. and Metherell, A.J.F. (1968) Philos. Mag., 18, 41. 41 Cook, R.F. and Cundy, S.L. (1969) Philos. Mag., 20, 665. 42 Schattschneider, P. and Jouffrey, B. (1995) Energy-Filtering Transmission Electron Microscopy (ed. L. Reimer), Springer-Verlag, Berlin, Heidelberg, p. 151. 43 Brockt, G. and Lakner, H. (1998) Proceedings International Congress on Electron Microscopy, Cancun., vol. III, p. 625. 44 Ahn, C.C. and Krivanek, O.L. (1983) EELS Atlas – A Reference Guide of Electron Energy Loss Spectra Covering All Stable Elements, GATAN Inc. 45 Rez, P. (1992) Transmission Electron Energy Loss Spectrometry in Materials Science (eds M.M. Disko, C.C. Ahn, and B. Fultz), The Minerals, Metals & Materials Society, Warrendale, p. 107. 46 Hofer, F. (1995) Energy-Filtering Transmission Electron Microscopy (ed. L. Reimer), Springer-Verlag, Berlin, p. 225.
47 Brydson, R., Sauer, H., and Engel, W. (1992) Transmission Electron Energy Loss Spectrometry in Materials Science (eds M.M. Disko, C.C. Ahn, and B. Fultz), The Minerals, Metals & Materials Society, Warrendale, p. 131. 48 Reimer, L., Zepke, U., Schulze-Hillert, S., Ross-Messemer, M., Probst, W., and Weimer, E. (1992) EELS Spectroscopy – A Reference Handbook of Standard Data for Identification and Interpretation of Electron Energy Loss Spectra and for Generation of Electron Spectroscopic Images, Carl Zeiss, Electron Optics Division, Oberkochen. 49 Johnson, K.H. (1973) Advances in Quantum Chemistry, vol. 7 (ed. P.O. Loewdin), Academic Press, New York, p. 143. 50 Dill, D. and Dehmer, J.L. (1974) J. Chem. Phys., 61, 692. 51 Wurth, W. and Stöhr, J. (1990) Vacuum, 41, 237. 52 Durham, P.J. (1982) Computer Phys. Commun., 25, 193. 53 Vvedensky, D.D., Saldin, D.K., and Pendry, J.B. (1986) Computer Phys. Commun., 40, 421. 54 Jansen, R.W. and Sankey, O.F. (1987) Phys. Rev. B, 36, 6520. 55 Weng, X., Rez, P., and Sankey, O.F. (1989) Phys. Rev. B, 40, 5694. 56 Leapman, R.D. and Silcox, J. (1979) Phys. Rev. Lett., 42, 1361. 57 Schneider, R., Syrowatka, F., Röder, A., Abicht, H.P., Völtzke, D., and Woltersdorf, J. (1998) Proceedings 11th European Congress on Electron Microscopy, Brussels, p. I-287. 58 Brydson, R., Sauer, H., Engel, W., and Hofer, F. (1992) J. Phys. Condens. Matter, 4, 3429. 59 Mitterbauer, C., Kothleitner, G., Grogger, W., Zandbergen, H., Freitag, B., Tiemeijer, P., and Hofer, F. (2003) Ultramicroscopy, 96, 469. 60 de Groot, F.M.F., Abbate, M., van Elp, J., Sawatzky, G.A., Ma, Y.J., Chen, C.T., and Sette, F. (1993) J. Phys. Condens. Matter, 5, 2277. 61 Colliex, C. (1984) Advances in Optical and Electron Microscopy (eds R. Barer and V.E. Cosslett), Academic Press, London.
References 62 Teo, B.K. and Joy, D.C. (eds) (1981) EXAFS Spectroscopy, Techniques and Applications, Plenum Press, New York. 63 Stern, E.A. (1988) X-Ray Absorption (eds D.C. Koningsberger and R. Prins), John Wiley & Sons, Inc., New York, p. 3. 64 Egerton, R.F. (1975) Philos. Mag., 31, 199. 65 Egerton, R.F. (1979) Ultramicroscopy, 4, 169. 66 Ahn, C.C. and Rez, P. (1985) Ultramicroscopy, 17, 105. 67 Joy, D.C. (1986) Principles of Analytical Electron Microscopy (eds D.C. Joy, A.D. Romig Jr, and J.I. Goldstein), Plenum Press, New York, London, p. 249. 68 Egerton, R.F. and Leapman, R.D. (1995) Energy-Filtering Transmission Electron Microscopy (ed. L. Reimer), SpringerVerlag, Berlin, Heidelberg, New York, p. 269. 69 Leapman, R.D. and Andrews, S.B. (1991) J. Microsc., 161, 3.
70 Pennycook, S.J., Jesson, D.E., Browning, N.D., and Chisholm, M.F. (1994) Mikrochim. Acta, 114/115, 195. 71 Schaper, A.K., Hou, H., Greiner, A., Schneider, R., and Phillipp, F. (2004) Appl. Phys. A, 78, 73. 72 Zorn, M., Knigge, A., Zeimer, U., Klein, A., Kissel, H., Weyers, M., and Tränkle, G. (2003) J. Cryst. Growth, 248, 186. 73 Jeanguillaume, C. and Colliex, C. (1989) Ultramicroscopy, 28, 252. 74 Schneider, R., Woltersdorf, J., and Röder, A. (1994) Mikrochim. Acta, 114/115, 545. 75 Sigle, W. (2000) Proceedings 12th European Congress on Electron Microscopy, vol. II (eds L. Frank and F. Ciampor), Czechoslovak Society for Electron Microscopy, Brno, p. 101. 76 Dahmen, U. (2007) Microsc. Microanal., 13 (Suppl. 2), 1150 CD. 77 Kabius, B., Hartel, P., Haider, M., Müller, H., Uhlemann, S., Loebau, U., Zach, J., and Rose, H. (2009) J. Electron Microsc., 58, 147.
91
93
5 Low-Energy Electron Diffraction (LEED) Georg Held
5.1 Principles and History
When Davisson and Germer reported, in 1927, that the elastic scatting of lowenergy electrons from well-ordered surfaces leads to diffraction spots similar to those observed in X-ray diffraction [1–3], this was the first experimental proof of the wave nature of electrons. A few years earlier, in 1923, De Broglie had postulated that electrons have a wavelength (in Ångströms) of
λ e = h/me v = (150 eV/E kin )1 2
(5.1)
and a corresponding wave vector of the length k = 2π /λ e
(5.2)
where h is Planck’s constant, me the electron mass, v the velocity, and Ekin the kinetic energy of the electron. Already, Davisson and Germer realized that the diffraction of low-energy electrons (also termed low-energy electron diffraction; LEED) could be used to determine the structure of single crystal surfaces, in analogy to X-ray diffraction, if their kinetic energy was between 40 and 500 eV – that is, if their wavelength ranged between 0.5 and 2 Å. Due to their small inelastic mean free path (IMFP) of only a few Ångstroms (typically <10 Å), electrons in this energy range sample only the topmost atomic layers of a surface and are, therefore, better suited for the analysis of surface geometries than X-ray photons, which have a much larger mean free path (typically of a few micrometers). However, unlike for photon diffraction, multiple scattering plays an important role in the diffraction process of electrons at solid surfaces. Therefore, the analysis of LEED data with respect to the exact positions of atoms at the surface is somewhat more complicated, and requires fully dynamical quantum mechanical scattering calculations. The use of LEED for surface analysis became important when sufficiently large single crystals became available for surface studies. Initially, the technique was used only for the qualitative characterization of surface ordering and quantitative determination of the two-dimensional (2-D) surface lattice parameters (e.g., superstructures, see below). Information regarding the positions of the atoms in the Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
94
5 Low-Energy Electron Diffraction (LEED)
surface is hidden in the energy-dependence of the diffraction spot intensities, the so-called LEED I–V, or I(E), curves. During the late 1960s, computer programs became available which could perform fully dynamical scattering calculations for simple surface geometries. The comparison of such theoretical I–V curves for a set of model geometries with experimental data allows determination of the atomic positions within the surface by trial and error. With the subsequent immense growth of available computer power and speed, LEED I–V structure determination could be applied to a large number of increasingly complex surface geometries. This has led to LEED becoming a standard technique for modern surface crystallography. For further details concerning the history, experimental set-up, and theoretical approaches of LEED, the reader is referred to the books of Pendry [4], Van Hove and Tong [5], Van Hove, Weinberg and Chan [6], and Clarke [7]. Indeed, the content of the present chapter relies extensively on this material.
5.2 Instrumentation
The standard modern LEED optics is of the “rear view” type, as shown schematically in Figure 5.1. The incident electron beam, accelerated by the potential V0, is emitted from the electron gun behind the hemispherical fluorescent screen which is made of glass, and hits the sample through a hole in the screen. Typically, the electron beam has a current of about 1 μA, and illuminates an area of 1 mm2. The surface is in the center of the hemisphere, so that all backdiffracted electrons travel
Figure 5.1 Schematic diagram of a four-grid LEED display system [6].
5.2 Instrumentation
towards the LEED screen on radial trajectories. Before the electrons hit the screen they must pass a retarding field energy analyzer; this typically consists of four (or three) hemispherical grids that are concentric with the screen, with each containing a central hole through which the electron gun is inserted. The first grid (nearest to the sample) is connected to earth ground, as is the sample, so as to provide an essentially field-free region between the sample and the first grid. This minimizes any undesirable electrostatic deflection of diffracted electrons. A suitable negative potential, –(V0 – ΔV), is applied to the second and third (only second) grids (the so-called suppressor grids) to allow only those electrons to be transmitted to the fluorescent screen whose energy is within eΔV of the elastically scattered electrons. The fourth (third) grid is usually grounded in order to reduce field penetration of the screen voltage to the suppressor grids. The screen voltage is of the order of 5–6 kV; this provides the electrons with sufficient energy to make the diffraction pattern visible on the fluorescent screen. Older designs have opaque screens, which allow observation only from behind the sample, where the sample holder obstructs the view. However, all modern instruments have transparent fluorescent screens that allow the pattern to be observed through a view-port from behind the screen. With this design, only the electron gun assembly (diameter <15 mm) limits the view slightly. During recent years, the so-called MCP-LEED systems, which have position-sensitive “multi-channel plate” (MCP) electron multipliers between the grids and the fluorescent screen, have become commercially available for applications that require low incident electron currents. The main aim of this approach is to avoid beam damage (see below) or charging of the insulating samples. These systems can be operated with electron currents as low as 1 nA. Typical LEED systems have diameters of around 140 mm, so that they can be fitted into a 150 mm (inner diameter) flange. To date, the usual way of recording the LEED pattern is to use a video camera with suitable image-processing software. In older systems, movable Faraday cups (FCs) were used, which were able to detect the electron current directly, but because of the long data acquisition times and problems of transferring motion into ultra-high vacuum (UHV) these systems are now rarely used. In fact, only one widely used modern type of LEED optic, which is designed for accurate spot profile analysis, employs a FC arrangement in combination with a “channeltron” electron detector (the SPA-LEED; see Figure 5.2) [8]. In this design, the FC is held in a fixed position and the grids are replaced by round apertures which define the lateral resolution of the system. Instead of the position of the electron detector, the angle of the incoming electron beam is varied through an electrostatic octupole lens, thereby deflecting the desired part of the LEED pattern towards the detector, without the need for any moving parts. SPA-LEED systems have been specially designed to have a high resolution in k-space; this means that features in the LEED pattern (e.g., spot splitting, spot broadening) can be observed and measured, which correspond to large typical distances (step separation, island size) on the surface. The largest resolvable length on the surface is called the transfer width, and this characterizes the LEED instrument [9]. Typical values of the transfer width are around 150 Å for conventional rear view LEED systems, and up to 1000 Å for SPA-LEED systems.
95
96
5 Low-Energy Electron Diffraction (LEED)
Figure 5.2 Schematic set-up of the SPA-LEED system [8].
5.3 Qualitative Information
The most direct information obtained from LEED concerns the periodicity and the degree of order within the transfer width of the surface under investigation. Such data can be gathered by visual inspection of the diffraction pattern, and/or through relatively simple mathematical transformations of the spot profiles. 5.3.1 LEED Pattern
Because low-energy electrons do not penetrate into the crystal bulk far enough to experience its three-dimensional (3-D) periodicity, the diffraction pattern is determined by the 2-D surface periodicity described by the lattice vectors a1 and a2, which are parallel to the surface plane. A general lattice point within the surface plane is an integer multiple of these lattice vectors: R = n1 a 1 + n 2 a 2
(5.3)
The 2-D Bragg condition leads to the definition of reciprocal lattice vectors a1* and a2* which fulfill the set of equations: a1 ⋅ a1* = a2 ⋅ a2* = 2π
(5.4a)
a1 ⋅ a2* = a2 ⋅ a1* = 0
(5.4b)
These reciprocal lattice vectors, which have units of Å−1 and are also parallel to the surface, define the LEED pattern in k-space. Each diffraction spot corresponds to the sum of integer multiples of a1* and a2*. g(n 1,n 2) = n1a1* + n2 a2*
(5.5)
The integer numbers (n1,n2) are used as indices to label the spot. The parallel component of the corresponding wave vector is: k||,(n 1,n 2) = k||,0 + g(n 1,n 2)
(5.6)
5.3 Qualitative Information
where k||,0 is the parallel component of the wave vector, k, of the incoming electron beam. The length of k is (4πmeEkin/h2)1/2 for both the incoming and elastically scattered electrons. The vertical component, kz, of the electrons backdiffracted into spot (n1,n2) is defined by energy conservation: 2 12 k z,(n 1,n 2) = (4 πm eE kin/h 2 − k||,( n 1,n 2 ) )
(5.7)
This equation also limits the number of observable LEED spots, since the expression inside the brackets must be greater than zero. With increasing electron energy, the number of LEED spots increases while the polar emission angle with respect to the specular spot (0,0) decreases for each spot. The specular spot does not change its position if the angle of incidence is kept constant. Figure 5.3 shows examples of common surface unit cells and the corresponding LEED patterns. In many cases the periodicity of the surface lattice is larger than expected for the bulk-truncated surface of a given crystal, due to adsorption or reconstruction. This leads to additional (superstructure) spots in the LEED pattern, for which fractional indices are used. The lattice vectors b1 and b2 of such superstructures can be expressed as multiples of the (1 × 1) lattice vectors a1 and a2: b1 = m11a1 + m12 a2
(5.8a)
b2 = m21a1 + m22 a2
(5.8b)
where the numbers mij are the coefficients of the superstructure matrix, which is a straightforward way of characterizing the superstructure. The positions and indices of the additional LEED spots can be calculated directly from this matrix [6] according to the formulae: b1* = (m11m22 − m12m21 )−1(m22 a1* − m21a 2* )
(5.9a)
b2* = (m11m22 − m12m21 )−1( − m12 a1* − m11a 2* )
(5.9b)
The fractional indices of the superstructure spots are multiples of the prefactors of a1* and a2* in the above equations. The area of the superstructure unit cell, A, in units of the (1 × 1) unit cell area can also easily be calculated from the coefficients of the superstructure matrix: A = (m11m22 − m12m21 )
(5.10)
Another, less general notation according to Wood [10] specifies the lengths of the vectors b1 and b2 in units of a1 and a2 together with the rotation angle α between b1 and a1 (only specified if non-zero): p/c (|b1|/|a1| ¥ |b2 |/|a2 |) Rα
(5.11)
where “p” indicates a “primitive” and “c” a “centered” surface unit cell. Examples are “p(2 × 2)”, “p( 3 × 3 ) R30°”, and “c(2 × 2).” This notation is not applicable in all cases, but it is more frequently used than the matrix notation because it is shorter. Figure 5.4 shows examples of common superstructures with the corresponding matrix and Wood notations.
97
98
5 Low-Energy Electron Diffraction (LEED)
Figure 5.3 Direct (left column) and reciprocal (right column) lattices for the five 2-D Bravais
lattices [6].
5.3 Qualitative Information
Figure 5.4 Direct and reciprocal lattices of a series of commonly occurring 2-D super-lattices.
Open circles correspond to the ideal (1 × 1) surface structure, while filled circles represent adatoms in the direct lattice and fractional-order spots in the reciprocal lattice [6].
99
100
5 Low-Energy Electron Diffraction (LEED)
5.3.2 Spot Profile Analysis
While the spot positions carry information about the size of the surface unit cell, the shapes and widths of the spots – that is, the spot profiles – are influenced by the long-range arrangement and order of the unit cells at the surface. If vertical displacements of the surface unit cells are involved (steps, facets), the spot profiles do change as a function of electron energy. If all surface unit cells are in the same plane (within the transfer width of the LEED optics), then the spot profile is constant with energy. A periodic arrangement of equal steps at the surface leads to a spot splitting at certain energies from which the step height can be determined directly. For a statistical arrangement of steps, the analysis of energy-dependent changes in the spot profiles allows in many cases the determination of the mean step height and a characterization of the step distribution [11]. Facets lead to extra spots which move in k|| upon changes of the kinetic energy. Point defects, static disorder, and thermally induced displacements lead to an increase of the background intensity between the spots. Depending on the correlation between the scatterers, the background is either homogeneous (no correlation) or structured (correlation). If the coherently ordered surface areas (islands, domains) are smaller than the transfer width of the LEED system and at the same vertical height, then the width of these areas, Δw, is directly related to the width of the LEED spots in k-space, Δk||: Δk|| = 2π/Δw
(5.12)
This relation holds for each direction parallel to the surface independently. It is particularly useful for determining the size of adsorbate islands which lead to extra superstructure spots. A good introduction (in German) into spot profile analysis is given by Henzler and Göpel in Ref [12]. 5.3.3 Applications and Restrictions
The visual inspection of the LEED pattern is a very good and fast method for qualitative characterization of the surface under investigation; consequently, LEED optics form part of most UHV systems dedicated to the analysis of single crystal surfaces. As a result, a large number of LEED patterns with and without superstructures have been reported in the literature (e.g., Ref. [13]). The observation of sharp LEED spots indicates long-range ordering in the topmost layers. The absence of such spots is often a sign of surface roughness (e.g., after ion bombardment) or of a thick amorphous layer (e.g., water, hydrocarbons) covering the surface. The appearance of a superstructure is also very useful in characterizing the surface after adsorption or heat treatments. Numerous examples have been reported as to how superstructures can be used to identify certain adsorbate states, oxidation, carbide formation, and alloying. As these effects depend very much on the reactiv-
5.4 Quantitative Structural Information
ity and surface structure, such information is best found in the original literature concerning the particular surface. The size of the surface unit cell can be deduced exactly for a given superstructure according to Equation 5.10. Thus, the relative coverage of an adsorbate inducing this superstructure can often also be determined by LEED, since there must be an integer number of adsorbate atoms/ molecules per unit cell. Determination of the surface morphology (domain size, step distribution, etc.) by LEED spot profile analysis always takes into account the whole surface area hit by the incoming electron beam. It therefore uses a much larger statistical sample than any microscopic method (e.g., STM, SEM, TEM) could deal with and allows, therefore, a more reliable determination of average surface properties such as roughness or step density. Spot profile analysis has been applied to flat metal surfaces in order to study the development of domain sizes during phase transitions in adsorbate layers [14, 15]. Changes in the step density and/or distribution or faceting upon adsorption or annealing have also been studied using LEED on stepped and nominally flat metal and semiconductor surfaces [16–19]. For these experiments, either conventional LEED systems with a fluorescent screen or FC systems with small apertures (e.g., a movable FC or a SPA-LEED system) have been used. Another advantage of LEED relates to the fact that it is a contact-free analysis method, and allows access to the surface while the data are recorded. This fact has been used to analyze surfaces in situ during adsorption, epitaxy, or sublimation in order to study dynamical restructuring processes [20–22]. Nonetheless, all LEED data analysis must rely on assumptions and model distributions, and is therefore not really direct. Microscopic methods have the advantage of directly delivering the shape of the surface (domains, terraces, etc.), without any prior assumptions having to be made.
5.4 Quantitative Structural Information
The analysis of the LEED pattern or the spot profiles does not provide any quantitative information concerning the position of the atoms within the surface unit cell. This type of information is hidden in the energy dependence of the spot intensities, the so-called LEED I–V curves. 5.4.1 Principles
The spot intensity variations are caused by the interference of electrons scattered from different atomic layers parallel to the surface. For an infinite penetration depth this would impose a third Bragg condition for kz and, hence, for Ekin (cf. Equation 5.7) with sharp intensity maxima for certain energies and zero intensity between them. As the penetration depth is very small, the backscattered
101
102
5 Low-Energy Electron Diffraction (LEED)
electrons interact with only a few layers of atoms, which causes broad maxima and nonzero intensities in the intermediate energy regimes. Due to the influence of the atomic scattering potentials and multiple scattering shifts with respect to the Bragg peaks, additional peaks are observed in the I–V curves. All of these effects are included in modern LEED computer programs, which perform fully dynamical quantum mechanical scattering calculations. These programs deliver a set of I–V curves, which would be expected for a given user-specified surface geometry. Because multiple scattering dominates the electron diffraction process at low energies, there is no easy way of determining the surface geometry directly, such as the Patterson function in X-ray crystallography. Instead, I–V curves must be calculated for a large number of model geometries and compared with the experimental I–V curves. Their agreement is then quantified by the means of a reliability factor (R factor). Such R factors can be defined in several ways [6], with Pendry’s R factor, RP, being the most common [23]. By convention, RP is 0 when the agreement is perfect, 1 for uncorrelated sets of I–V curves, and 2 for completely anticorrelated curves (when each maximum of one curve coincides with a minimum of the other). Usually, automated search procedures are used, which modify the model geometries to be tested according to the R factor values achieved by the preceding geometries in order to find an R factor minimum within the set of geometric parameters to be optimized. The search strategies are either conventional downhill-oriented algorithms (simplex method, Powell’s method, Marquard’s algorithm [24]), which usually identify only the nearest local minimum, or stochastic Monte-Carlo-type methods (simulated annealing [24], genetic algorithm [25]) which, in principle, should always find the global minimum. A set of experimental and theoretical I–V curves for benzene on Ru(0001) [26] is depicted in Figure 5.5, showing a typical degree of agreement achieved in a structure optimization for organic molecules. Usually, the collection of LEED I–V curves requires single crystal surfaces with long-range order in the topmost layers. However, structural information can also be obtained in a similar way for certain adsorbate-covered surfaces without longrange order, when the energy dependence of the diffusely scattered intensity is analyzed (Diffuse LEED [27, 28]), or from the I–V curves of integer-order spots, which are still observed even if the adsorbate layer is not ordered. In both cases, however, the data analysis must assume that the local adsorption geometry is the same for all adsorbates [29, 30]. 5.4.2 Experimental Techniques
The standard experimental set-up for collecting LEED I–V curves uses a video camera for recording images of the fluorescent screen for each energy (Video LEED) [31]. When a conventional video camera is used, the rate at which images are collected is fixed at 50–60 Hz. In this case, several images (typically between
5.4 Quantitative Structural Information
Figure 5.5 Examples of experimental (solid lines) and calculated (dashed lines) LEED I–V
curves for the p( 7 × 7 )R19 superstructure of benzene on Ru(0001). The individual RP factors are indicated in the figure [26].
eight and 256) must be averaged in order to obtain a satisfying signal-to-noise ratio (SNR), especially for weak superstructure spots. Newer systems often use slow-scan CCD cameras, which can perform the averaging directly on the cooled CCD chip (exposure times up to several seconds) and therefore avoid multiple readout noise [32]. The images are either stored on a computer hard disk and analyzed later, or the spot intensities are extracted online using special software. Usually, the intensity is averaged within a given area at the spot position and the averaged local background outside this area is subtracted from this intensity. Commercially available data-acquisition software also controls the electron energy and records the electron beam current, which is needed to normalize the spot intensities. Older experimental set-ups used FCs with small apertures mounted on goniometers, which could be moved around the sample in order to collect the backscattered electron current directly; alternatively, spot photometers were used, which could be directed at one diffraction spot on the fluorescent screen while the energy was varied. Data collection is mostly carried out at normal incidence of the primary electron beam. Under these conditions, those spots which are related by the symmetry of the surface have identical I–V curves. Therefore, by ensuring that the I–V curves of equivalent spots are identical, normal incidence conditions can be adjusted to within a few tenths of a degree.
103
104
5 Low-Energy Electron Diffraction (LEED)
5.4.3 Computer Programs
Most computer programs currently used for calculating LEED I–V curves are based on the multiple scattering algorithms outlined in the early works of Pendry, Tong, and Van Hove [4, 5]. The most common way of calculating LEED I–V curves is to subdivide the crystal into atomic layers parallel to the surface. All possible multiple scattering paths inside each layer are added up first and combined in a layer diffraction matrix. The total backdiffracted intensity for each LEED spot is then calculated in a second step, by combining these layer diffraction matrices in a way that includes all remaining multiple scattering paths between the layers. The amount of computer time needed for calculating a set of I–V curves for one model geometry depends on the number of atoms per surface unit cell, and also on the number of spots within the LEED pattern. In both cases, the dependence is cubic, and therefore the time requirements will vary quite substantially, from a few seconds up to several hours depending on the complexity of the surface geometry. The computational effort can be reduced significantly by making use of rotational or mirror symmetries at the surface. Some of the more recent program versions use the Tensor LEED approximation developed by Rous and Pendry, which can save a substantial amount of computer time [33–35]. In Tensor LEED, the amplitudes Ag(0) of all escaping electron waves (spots) are first calculated in the conventional manner, as described above for a reference geometry. The derivatives of these amplitudes δAg/δri with respect to small displacements of each atom i in this reference geometry, are then calculated. These derivatives are the constituents of the “Tensor.” The wave amplitude for a modified model geometry, where atom i is displaced by the vector Δri, is then approximately given by: Ag (Δri ) = Ag (0) + (δAg/δri )Δri
(5.13)
In this way, a simple summation over all displaced atoms avoids a new multiple scattering calculation and is, therefore, several orders of magnitude faster than a conventional LEED calculation, especially for large surface unit cells with many atoms. Obviously, Tensor LEED requires much more storage space and overhead than a conventional LEED calculation, as the derivatives must be stored for each atom at each energy. Moreover, the applicability is limited to geometries in the neighborhood of the reference structure with maximum displacements of a few tenths of an Ångström. When automated searches are implemented, the LEED I–V calculation and the R factor comparison are called by the master search program, either as a subroutine or as a separate program through the operating system. In all cases, the LEED I–V calculations represent the most time-consuming parts of the searches, and the main criterion for selecting search algorithms is to reduce the number of trial geometries. Currently, there are no commercially available computer programs for LEED I–V structure determinations. However, most programs – together with
5.4 Quantitative Structural Information
their documentation – can be downloaded from the internet or are distributed by their authors [36]. 5.4.4 Applications and Restrictions
To date, the structure analysis by LEED I–V represents the most accurate and reliable way of determining the atomic positions at surfaces. Especially, the fact that LEED is sensitive to the atomic positions not only in the topmost layer but also down to several layers below the surface, makes it an ideal tool for studying spontaneous or adsorbate-induced surface reconstructions. General restrictions are that LEED usually requires ordered and conducting surfaces, otherwise charging would distort the LEED pattern. The time required for calculating the I–V curves and the number of trial geometries are two factors which limit the complexity of accessible surface structures on the computational side; the density of LEED spots on the fluorescent screen is an experimental factor limiting the size of unit cells that can be studied. Although LEED is a nondestructive method, electron-induced damage to the sample can be problematic, especially when organic molecules are studied. As a rule of thumb, a conventional LEED system with a beam current of 1 μA and a spot cross-section of 1 mm2 leads to an exposure of the order of one electron per molecule and second. If I–V curves are recorded over an energy range of 250 eV in steps of 1 eV, with a data collection time of 1 s per step, this leads to a total exposure of 250 electrons per molecule. Cases have been reported, however, where an exposure of the order of one electron per molecule can cause significant damage to a molecular layer [37]. Beam damage can be reduced by using low-current LEED systems, and by scanning the sample during data acquisition. As the LEED theory was initially developed for close-packed clean metal surfaces, these are the most reliably determined surface structures, leading to RP factors often below 0.1, which is of the order of the agreement between two experimental sets of I–V curves. In these cases, the error bars for the atomic coordinates are as small as 0.01 Å, when the total energy range of I–V curves is large enough (>1500 eV). A good overview on state-of-the-art LEED structure determinations of clean metal surfaces and further references may be found in two articles by Heinz et al. [38, 39]. For more open adsorbate-covered and/or reconstructed surfaces, certain approximations used in the standard programs are less accurate, which in turn leads to higher RP factor values. For simple superstructures of mono-atomic adsorbates or small molecules on metal surfaces, RP factors between 0.1 and 0.2 can be expected. However, for large molecules, which often adsorb in complex superstructures, the optimum RP factor values may be as large as 0.3. As the error margins scale with the optimum R factor value, the accuracy of atomic coordinates in these latter cases is smaller, with error bars up to 0.1 Å. (a review describing structure determinations of molecular adsorbates was produced by Over [40].) In general, the accuracy is higher for coordinates perpendicular to the surface than for lateral coordinates.
105
106
5 Low-Energy Electron Diffraction (LEED)
Due to the minimization of the number of dangling bonds, semiconductor surfaces often show large displacements of the surface atoms from their bulk lattice positions. As a consequence, these surfaces are also very open and the agreement is more in the range of RP factor values around 0.2. The structure determination of semiconductor surfaces has been reviewed by Kahn [41]. A fairly complete listing of all surface geometries determined to date by using LEED can be found in the NIST surface structure database [13].
5.5 Low-Energy Electron Microscopy 5.5.1 Principles of Operation
The technique of low-energy electron microscopy (LEEM) is based on imaging electron microscopy, the principle being similar to that of TEM. The important differences between the techniques are that, in LEEM, the microscope detects reflected electrons and the incoming electrons are slowed down to low energies in the LEED range just before they hit the sample. The schematic diagram in Figure 5.6 shows the main components of the LEEM instrument, as developed by Bauer and Telieps during the 1960s and 1970s [42, 43]. The electrons are emitted from a conventional electron gun at an energy of typically 20 keV, and pass through the focusing optics consisting of condenser and deflectors. The magnetic beam splitter bends the electron beam towards the sample such that, while passing through the objective lens, the electrons are slowed down to energies in the LEED range. After interacting with the surface, the reflected electrons are accelerated
Figure 5.6 Schematic diagram of a low-energy electron microscope (according to Ref. [42]).
5.5 Low-Energy Electron Microscopy
Figure 5.7 (a, c) Dark-field images recorded
using the spots of two different domains of the p(5 × 1) superstructure on Ir{100}, as indicated in the LEED pattern in panel (b).
Regions appearing bright are covered by the respective domain. Images recorded at the Nanospectroscopy LEEM/PEEM instrument of Sincrotrone Elettra, Trieste, Italy.
back to their initial high energy through the same objective lens. The outgoing electrons are bent away from the electron gun in the beam splitter and focused onto the detector screen by a series of magnetic lenses. This lens system can be operated in two modes, displaying either the real space image or the diffraction pattern of the surface. In diffraction mode, the instrument provides essentially the same information as a LEED system with a very small incoming beam diameter. This mode is, therefore, often termed “micro-LEED.” The size of the LEED beam can be limited to diameters of a few micrometers by using the beam aperture in the electro-optical path on the left-hand side of Figure 5.6. Because no electron gun or sample is obstructing the view, the complete diffraction pattern can be observed, as shown in panel b of Figure 5.7. The small size of the LEED beam allows a LEED I–V analysis to be performed even on samples which are inhomogeneous at a scale of several micrometers [44]. In image mode, the instrument serves as an electron microscope, which records the intensity of the backscattered electrons. Commercial LEEM instruments reach spatial resolutions of 5 nm, with a theoretical limit of about 3 nm; the maximum field of view is about 50 μm. The intensity of elastically backscattered electrons is only high enough to be detected if their energy is in the LEED range, typically below 100 eV for LEEM. For higher energies, in the keV range, forward scattering dominates and the reflected intensity is too low. The two main contrast mechanisms for low-energy electrons are differences in the work function [detected in “mirror electron microscopy” (MEM) mode with kinetic energies below 5 eV] and interference effects. Different surface structures will lead to different I–V characteristics and, hence, to different backscattered intensities at a given electron energy. These are usually recorded in the “bright-field mode,” which uses the specular beam (or (0,0) spot). If the surface has areas covered with long-range ordered surface structures, the “dark-field image” mode can be used, in which the intensity of a higher-order diffraction spot is recorded. The spot used for imaging
107
108
5 Low-Energy Electron Diffraction (LEED)
is selected by adjusting the contrast aperture, which is placed in the “diffraction plane” of the detector path on the right-hand side of Figure 5.6. In dark-field mode, only those parts of the surface which have a superstructure that produces the selected diffraction spot will lead to measurable intensity; all other areas of the image will remain dark. Figure 5.7a and c show dark-field images of the same area of a Ir{100} surface, which is covered by two types of p(5 × 1) superstructures rotated by 90 ° with respect to each other, as seen from the diffraction pattern in Figure 5.7b. Spots from different rotational domains were used for Figures 5.7a and c; hence the patches appearing dark on panel a appear bright in panel c. The images also illustrate another contrast mechanism, namely phase contrast, which causes the dark stripes along steps on the surface. This is due to the interference between electron waves backdiffracted from atoms at different heights on both sides of the step. This contrast is only visible within a thin strip of a few nanometers from each step, over which the electron waves are coherent. 5.5.2 Applications and Restrictions
Most applications of LEEM are related to studies of pattern formation in the upper nanometer range, growth modes, and phase transitions. Low-energy electrons have a penetration depth of only a few atomic layers, and are therefore very surfacesensitive. The sensitivity towards differences in the crystallographic nature of different parts of the surface is an advantage over most other microscopies with similar spatial resolution, for example, SEM. LEEM is, therefore, particularly suited for studying structural surface phase transitions. On the other hand, LEEM has a much lower depth of focus than SEM and can, therefore only be used for flat samples. As for conventional LEED, only conducting samples can be used. The problem of beam damage may be more serious than for conventional LEED as the sample is typically exposed to a larger current density than in conventional LEED. All commercial instruments operate in UHV and allow heating and cooling of the sample while the measurements are carried out. A variant of the same type of instrument with an added energy filter can be used to record the spatial distribution of photoelectrons, if the sample is illuminated by a suitable UV or X-ray source. This is known as photoelectron emission microscopy (PEEM) [45].
References 1 Davisson, C. and Germer, L.H. (1927) Nature, 119, 558. 2 Davisson, C. and Germer, L.H. (1927) Phys. Rev., 29, 908. 3 Davisson, C. and Germer, L.H. (1927) Phys. Rev., 30, 705.
4 Pendry, J.B. (1974) Low Energy Electron Diffraction, Academic Press, London. 5 Van Hove, M.A. and Tong, S.Y. (1979) Surface Crystallography by LEED, Springer, Berlin.
References 6 Van Hove, M.A., Weinberg, W.H., and Chan, C.-M. (1986) Low-Energy Electron Diffraction, Springer, Berlin. 7 Clarke, L.J. (1985) Surface Crystallography – An Introduction to Low Energy Electron Diffraction, John Wiley & Sons, Ltd, Chichester. 8 Scheithauer, U., Meyer, G., and Henzler, M. (1986) Surf. Sci., 178, 441. 9 Park, R.L., Houston, J.E., and Schreiner, D.G. (1971) Rev. Sci. Instrum., 42, 60. 10 Wood, E.A. (1964) J. Appl. Phys., 35, 1306. 11 Henzler, M. (1977) Electron Spectroscopy for Surface Analysis (ed. H. Ibach), Springer, Berlin, p. 117. 12 Henzler, M. and Göpel, W. (1991) Oberflächenphysik Des Festkörpers, Teubner, Stuttgart. 13 Watson, P.R., Van Hove, M.A., and Hermann, K. (2002) NIST Surface Structure Database (SSD), Version 5.0, NIST, Gaithersburg. 14 Pfnür, H., and Piercy, P. (1989) Phys. Rev. B, 40, 2515. 15 Pfnür, H., and Piercy, P. (1990) Phys. Rev. B, 41, 582. 16 Hoogers, G. and King, D.A. (1993) Surf. Sci., 286, 306. 17 Bauer, E., Poppa, H., and Visvanath, Y. (1976) Surf. Sci., 58, 517. 18 Zhang, C., Van Hove, M.A., and Somorjai, G.A. (1985) Surf. Sci., 149, 326. 19 Fölsch, S., Barjenbruch, U., and Henzler, M. (1989) Thin Solid Films, 123, 172. 20 Neureiter, H., Spranger, S., Schneider, M., Winkler, U., Sokolowski, M., and Umbach, E. (1997) Surf. Sci., 186, 388. 21 Sokolowski, M., Neureiter, H., Schneider, M., Umbach, E., and Tatarenko, S. (1998) Appl. Surf. Sci., 71, 123–124. 22 Neureiter, H., Schinzer, S., Sokolowski, M., and Umbach, E. (1999) J. Crystal Growth, 93, 201–202. 23 Pendry, J.B. (1980) J. Phys. C, 13, 937. 24 Press, W.H., Flannery, B.P., Teukolsky, S.A., and Vetterling, W.T. (1988) Numerical Recipes in C, Cambridge University Press, Cambridge.
25 Goldberg, D.E. (1989) Genetic Algorithms in Search, Optimization and Machine Learning, Addison-Wesley, Reading, Mass. 26 Braun, W., Held, G., Steinrück, H.-P., Stellwag, C., and Menzel, D. (2001) Surf. Sci., 475, 18. 27 Heinz, K., Starke, U., and Bothe, F. (1991) Surf. Sci. Lett., 243, L70. 28 Heinz, K., Starke, U., Van Hove, M.A., and Somorjai, G.A. (1992) Surf. Sci., 57, 261. 29 Poon, H.C., Weinert, M., Saldin, D.K., Stacchiola, D., Zheng, T., and Tysoe, W.T. (2004) Phys. Rev. B, 69, 035401. 30 Braun, W. and Held, G. (2005) Surf. Sci., 594, 203. 31 Lang, E., Heilmann, P., Hanke, G., Heinz, K., and Müller, K. (1979) Appl. Phys., 19, 287. 32 Held, G., Uremovic, S., Stellwag, C., and Menzel, D. (1996) Rev. Sci. Instrum., 67, 378. 33 Rous, P.J. and Pendry, J.B. (1989) Surf. Sci., 219, 355. 34 Rous, P.J. and Pendry, J.B. (1989) Surf. Sci., 219, 373. 35 Rous, P.J. (1992) Prog. Surf. Sci., 39, 3. 36 LEED I–V program packages are distributed by: Held, G. http:// www.reading.ac.uk/chemistry/about/ staff/g-held.asp; Heinz, K. [email protected]; Van Hove, M.A. http://www.ap.cityu.edu.hk/ personal-website/Van-Hove. htm#Download_software 37 Faradzhev, N.S., Kostov, K.L., Feulner, P., Madey, T.E., and Menzel, D. (2005) Chem. Phys. Lett., 415, 165. 38 Heinz, K. (1994) Surf. Sci., 299/300, 433. 39 Heinz, K., Müller, S., and Hammer, L. (1999) J. Phys. Condens. Matter, 11, 9437. 40 Over, H. (1998) Prog. Surf. Sci., 58, 249. 41 Kahn, A. (1994) Surf. Sci., 299/300, 469. 42 Bauer, E. (1994) Rep. Prog. Phys., 57, 895. 43 Bauer, E. (1998) Surf. Rev. Lett., 5, 1275. 44 Figuera, J., Puerta, J.M., Cerda, J.I., Gabaly, F.E., and McCarty, K.F. (2006) Surf. Sci., 600, L105. 45 Bauer, E. (2001) J. Electron Spectr. Relat. Phenom., 975, 114–116.
109
111
6 Other Electron-Detecting Techniques John C. Rivière
6.1 Ion (Excited) Auger Electron Spectroscopy (IAES)
Auger emission after the creation of a core hole by electron or photon irradiation has been described under AES and XPS. Incident ions can also create core holes, so that Auger emission as a result of ion irradiation can also occur, giving rise to the technique of IAES. The ions used are normally noble gas ions, but protons and α-particles have also occasionally been used. In addition to the normal Auger features found in AES and XPS spectra, peaks are found in IAES spectra arising from Auger transitions apparently occurring in atoms or clusters sputtered from the surface. These peaks do not always coincide with those found in gas-phase excitation, and others are found that are not present in gas-phase measurements. Because of the complexity of the spectra, IAES cannot be used directly as an analytical technique, but it is very useful in basic physics experiments used to study Auger processes occurring in excited atoms. A review of the technique can be found in Ref [1].
6.2 Ion Neutralization Spectroscopy (INS)
Although, because of experimental difficulties, the technique of INS is probably the least used of those described here, it is also one of the physically most interesting. Ions of He+ of a chosen low energy in the range 5–10 eV approach a metal surface, and within an interaction distance of a fraction of a nanometer form ion– atom pairs with the nearest surface atoms. The excited quasi-molecule thus formed can de-excite by Auger neutralization. If unfilled levels in the ion fall outside the range of filled levels of the solid, as for He+, an Auger process can occur in which an electron from the valence band of the solid fills the core hole in the ion and the excess energy is given up to another valence electron, which is then ejected. Since either of the valence electrons can come from anywhere within the valence band, the observed energy spectrum reflects the local density of states at the solid Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
112
6 Other Electron-Detecting Techniques
surface, but is a self-convolution of the LDOS and of transition probabilities across the valence band. The technique of INS requires complex mathematical processing for deconvolution of the spectra.
6.3 Inelastic Electron Tunneling Spectroscopy (IETS)
IETS is unique in that it is entirely surface-specific, but does not require a vacuum environment and has almost never been performed in UHV. Its principle is relatively simple. If two metals are separated by a thin (∼3 nm) insulating layer, and a voltage is applied across them, electrons can tunnel from one to the other through the insulator, and if no energy is lost by the electrons, the process is known as elastic tunneling. If, however, discrete impurity states occur in the interfaces between the metal and the insulator, or molecules exist in such an interface that have characteristic vibrational energies, then the tunneling electron can give up some of this energy either to one of the states or to the vibrational mode, before reaching the other metal; this is termed inelastic tunneling. Obviously, the applied voltage must be greater than the state or vibrational energies. If the current across the metal–insulator–metal sandwich is recorded as a function of the applied voltage, the current increases as the threshold for each state or vibrational mode is crossed. The increases in current are in fact very small, and for improved detectability the current is double-differentiated with respect to voltage, thereby providing – in effect – a vibrational spectrum that can be compared directly with free-molecule IR and Raman spectra. The metal–insulator–metal sandwich is known as a tunnel junction, and its preparation is all-important. The standard junction consists of an aluminum strip about 60–80 nm thick and 0.5–1.0 mm wide, deposited in a very good vacuum onto scrupulously clean glass or ceramic. The surface of the strip is then oxidized either thermally or by glow discharge; the resultant oxide layer is extremely uniform and approximately 3 nm thick. The introduction of adsorbed molecules onto the oxide layer – or “doping” as it is called – is then effected either by immersion in a solution followed by spinning to remove excess fluid, or by coating from the gas phase. The final stage in preparation is deposition of the second metal (invariably lead) of the sandwich. This deposition is performed in a second vacuum system (i.e., not that used for aluminum deposition), the final thickness of lead being approximately 300 nm, and the width of the lead strip being the same as that of the aluminum strip. The reason for using lead is that all IETS measurements are conducted at liquid helium temperature (4.2 K) to optimize the energy resolution by reducing the contribution of thermal broadening to the linewidth, and lead is superconducting at that temperature. Normally, not one but several junctions are fabricated simultaneously, because the likelihood of junction failure always exists. After fabrication, the electrical connections are made to the metal electrodes, the junctions are dropped into liquid helium, and the measurements are commenced immediately.
Reference
Although insulators other than aluminum oxide have been tried, aluminum is still used almost universally, because it is easy to evaporate and forms a limiting oxide layer of high uniformity. Although being restricted to the adsorption of molecules onto aluminum oxide might appear to be a disadvantage of the technique, aluminum oxide is very important in many technical fields. In fact, many catalysts are supported on alumina in various forms, as are sensors; in addition, the properties of the oxide film on aluminum metal are of the greatest interest in adhesion and protection.
Reference 1 Grant, J.T. (2003) Surface Analysis by Auger and X-Ray Photoelectron Spectroscopy (eds D. Briggs and J. Grant), IM Publications, Chichester, pp. 763–774.
113
115
Part Two Ion Detection
117
7 Static Secondary Ion Mass Spectrometry (SSIMS) Heinrich F. Arlinghaus
Static secondary ion mass spectrometry (SSIMS) emerged as a technique of potential importance for surface analysis as a consequence of the work of Benninghoven and his group in Münster [1] during the late 1960s and early 1970s – that is, at about the same time as AES. The prefix “static” was added to distinguish the technique from “dynamic” SIMS, the difference between the two lying in the incident (or primary) ion current densities used, of the order of <1 nA cm−2 for SSIMS but much higher for dynamic SIMS. With this low primary ion dose density, spectral data can be generated on a time scale which is very short compared to the lifetime of the surface layer (typically surface destruction of <1%). Together with XPS and AES, SSIMS ranks as one of the principal surface analytical techniques. Since its sensitivity for elements greatly exceeds that of the other two techniques and much chemical information is available, the use of SSIMS is rapidly expanding in many fields of application.
7.1 Principles
A beam of positive ions bombards a surface, leading to interactions that cause the emission of a variety of types of secondary particle, including secondary electrons, Auger electrons, photons, neutrals, excited neutrals, positive secondary ions, and negative secondary ions. SSIMS is concerned with the last two of these, positive and negative secondary ions. The emitted ions are analyzed in a mass spectrometer, resulting in positive or negative mass spectra that consist of parent and fragment peaks characteristic of the surface. The peaks can be identified as arising from the substrate material itself, from contaminations and impurities on the surface, or from deliberately introduced species adsorbed onto the surface. When a heavy energetic particle such as an argon ion (typically 1 to 15 keV) hits a surface, it will not be stopped short by the first layer of atoms but rather will continue to move into the surface until it comes to a halt as a result of energy lost in atomic and electronic scattering. Along its way, the ion displaces some atoms from their original positions in the solid structure which, as they recoil, displace Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
118
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.1 Schematic diagram of the sputtering process.
other atoms; these displace additional atoms, and so on, resulting in a complex sequence of collisions. Depending on the energy absorbed in an individual collision, some atoms are permanently displaced from their normal positions, whereas others return elastically after temporary displacement. This sequence, termed a collision cascade, is illustrated schematically, along with some other processes, in Figure 7.1. As the cascade spreads out from the path of the primary ion, its effect can eventually reach back to the surface layer, and with enough kinetic energy bonds can be broken so that material leaves the surface in form of neutral or ionized atoms or clusters. The resulting secondary particle emission is of low energy (peak of energy distribution of 5–10 eV), with over 95% of the secondary particles originating from the top two monolayers of the solid. Using a mass spectrometer, these ionized surface particles can be analyzed directly, resulting in very high sensitivity. The theoretical aspect of the collision cascade is given by Sigmund’s collision cascade model [2], which explains much about the factors involved in the removal of material from a surface by sputtering, although it cannot predict the extent of positive or negative ionization of the material. A unified theory of secondary ion formation in SSIMS does not yet exist, although many models have been proposed for the process. Since the secondary ion yield – that is, the probabilities of ion formation – can vary by several orders of magnitude for the same element in different matrices, or for different elements across the Periodic Table, the lack of any means of predicting ion yield in a given situation will limit both interpretation and quantification. Although some theoretical and semi-theoretical models have enjoyed some predictive success for a very restricted range of experimental data, they fail when extended further. The most important theoretical models are briefly addressed in Section 8.1; further information is given elsewhere [3].
7.2 Instrumentation
7.2 Instrumentation
The instrumentation for SSIMS can be divided into two sections: (i) the primary ion source in which the primary ions are generated, transported, and focused towards the sample; and (ii) the mass analyzer in which sputtered secondary ions are extracted, mass separated, and detected. 7.2.1 Ion Sources
Several ion sources are particularly suited for SSIMS. The first produces positive noble gas ions (usually argon) either by electron impact (EI) or in a plasma created by a discharge (see Figure 8.18 in Section 8.2). The ions are then extracted from the source region, accelerated to the chosen energy, and focused in an electrostatic ion-optical column. More recently, it has been shown that the use of primary polya+ , or Arx+ , can enhance the molecular tomic or cluster ions, for example, SF5+, Bi +x , C60 secondary ion yield by several orders of magnitude [4, 5]. Argon ion guns (see also Section 8.2.1) of the EI type operate in the energy range 0.2 to 10 keV with a 0.1–1 μA dc current, providing optimum spot sizes on a surface of approximately 10 μm. Discharge-type argon ion guns, often using the duoplasmatron arrangement, operate at up to 15 keV, with currents variable up to 20 μA and, depending on the particular focusing lens system used, providing spot sizes down to approximately 1 μm. Cesium ions are also sometimes used to enhance the secondary ion yield of negative elemental ions and that of some polymer fragments [6]. These are produced by surface ionization with an extraction technique similar to that of EI sources. Another type of ion gun produces positive ions from a liquid metal (e.g., gallium) in the manner shown schematically in Figure 7.2 [7]. A fine needle (f, tip radius ∼5 μm) of refractory metal passes through a capillary tube (d) into a reservoir of liquid metal (e). The liquid is drawn up through the tube over the needle tip by capillary action. The application of 10–30 keV between the needle and a nearby extractor electrode (b) draws the liquid up into a cusp (Taylor cone) as a result of the combined forces of surface tension and electrostatic stress. Ions are formed in the region just above the cusp and then accelerated through an aperture in the extractor and focused through an ion optical column onto the sample surface. Gallium is often used as the liquid metal, because its melting point is 30 °C, and it thus remains liquid at room temperature. Today, mainly Bi is used for SSIMS, but it requires heating of the reservoir. Because ion production occurs in a very small volume, liquid metal ion sources (LMIS) are very bright (∼1010 A m−2 Sr−1). As a result, the ion beam can be focused to a fine spot, resulting in a spot size of 0.2 μm at 8–10 keV; this is reduced to approximately 20 nm at 30 keV. All ion gun optical columns are provided with deflection plates for scanning the ion beam over areas adjustable from a few square micrometers to many square
119
120
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.2 Schematic diagram of the design and operation of a liquid-metal ion source (LMIS) [7]. (a) Metal ions; (b) Extractor; (c) Liquid–metal film; (d) Capillary tube; (e) Liquid metal; (f) Needle.
millimeters. These have been adapted for pulsing by the introduction of additional deflection plates which rapidly sweep the beam across an aperture. By applying an ion beam bunching technique, ion pulses less than 1 ns wide can be produced. 7.2.2 Mass Analyzers
In order to minimize surface damage, static SIMS mass spectrometers should be as efficient as possible for detecting the total yield of secondary ions from a surface. Likewise, to be able to separate elemental from molecular species, and molecular species from each other, the mass resolution usually given as the mass m divided by the separable mass Δm, should be very high. With this in mind, two types of mass spectrometer have been used – in early studies mainly quadrupole mass filters and, today, time-of-flight (TOF) mass spectrometers. 7.2.2.1 Quadrupole Mass Spectrometers Quadrupole mass spectrometers have been in use for many years as residual gas analyzers (with an ionizing hot filament) and in desorption studies and SSIMS. They consist of four circular rods, or poles, arranged equally spaced in a rectangular array and exactly coaxial. This arrangement, as depicted by Krauss and Gruen [8], is shown in Figure 7.3. Two voltages are applied to the rods – a dc voltage (Udc) and an rf voltage (Urf ≈ Uo cos ωt). When an ion with a certain mass-to-charge ratio, m/q, enters the space between the rods it is accelerated by the electrostatic field and, for a particular combination of dc and rf voltages, the ion has a stable
7.2 Instrumentation
Figure 7.3 Experimental arrangement used
by Krauss and Gruen for SSIMS [8]. A quadrupole mass spectrometer was used for mass analysis and a retarding-field analyzer for prior energy selection. (a) Ion gun; (b–d) Lenses 1–3; (e) Quadrupole mass spectrom-
eter; (f) Charge-detecting electron multiplier; (g) Quadrupole power supply; (h) Pulse amplifier; (i) ISO amplifier; (j) Rate meter; (k) Multichannel analyzer; (l) Mass programmer; (m) Sawtooth generator; (n) Exit hole for neutrals.
trajectory and passes to a detector. For other combinations of voltages, the trajectory diverges rapidly and the ion is lost, either as a result of hitting one of the poles or by passing between them to another part of the system. The mass resolution is governed by the dimensions of the mass spectrometer, the accuracy of its construction, and the stability and reproducibility of the ramped voltage. The quadrupole mass spectrometer is compact, does not require magnets, and is entirely ultrahigh vacuum compatible – hence its popularity. It does have disadvantages, however, in that its transmission (typically <1%) is very low and decreases with increasing mass number. It is, furthermore, a scanning instrument enabling only the sequential transmission of ions, all other ions being discarded; the information loss is, therefore, very high. 7.2.2.2 Time-of-Flight Mass Spectrometry (TOF-MS) Since the quadrupole mass spectrometer is unsuitable for the analysis of large molecules on surfaces or of heavy metals and alloys, the TOF mass spectrometer was developed for use in SSIMS by Benninghoven and coworkers [9]. In a TOF mass analyzer (Figure 7.4), all sputtered ions are accelerated to a given potential V (2–8 keV), so that all ions have the same kinetic energy. The ions are then allowed to drift through a field-free drift path of a given length L before striking the detector. According to the equation (mL2)/(2t2) = qUo, light ions travel the fixed distance through the flight tube more rapidly than identically charged heavy ions. Thus, measurement of the time, t, of ions with mass-to-charge ratio, m/q, provides
121
122
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.4 Schematic diagram of the imaging time-of-flight SSIMS system used at the University of Münster, Germany.
a simple means of mass analysis with t2 = (mL2)/(2qUo) ∝ m/q. Since a very welldefined start time is required for flight time measurement, the primary ion gun must be operated in a pulsed mode to enable delivery of discrete primary-ion packages [10]. Electric fields (e.g., ion mirrors [10, 11] or electrical sectors [12, 13]) are used in the drift path to compensate for different incident energies and angular distributions of the secondary ions. For good mass resolution (m/Δm ≈ 10 000), the flight path must be sufficiently long (1–1.5 m), and very sophisticated highfrequency pulsing and counting systems must be employed to time the flight of the ion to within a tenth of a nanosecond. One great advantage of TOF-MS is its capacity to provide simultaneous detection of all masses of the same polarity. Raw data acquisition [14] enables the reconstruction of TOF spectra for any ion species as a function of depth and lateral position. The original TOF design [9] used pulsed beams of argon ions, but commercial development of the LMIS has led to a significant extension of the capabilities of the TOF system. The principle of LMIS operation enables the beam of isotopically enriched 69Ga+ or Bi +x ions to be focused to a probe of 50 nm minimum diameter,
7.3 Quantification
while being pulsed at frequencies up to 50 kHz and rastered at the same time. Operating an LMIS at beam currents of 10–100 pA makes it possible to perform scanning SSIMS and to produce secondary ion images from all masses. Pulsing of the ion source is achieved by rapid deflection of the ion beam across a small aperture, the pulse length being variable between 1 and 50 ns. The flight path length is 1.5 m, part of which includes an energy compensator (ion mirror) to ensure that all ions of the same mass, but of different energies, arrive at the same time. Secondary electron detectors (SEDs) enable the generation of topographical images produced by ion-induced secondary electrons. A low-energy (≤20 eV) electron flood gun can be used for charge compensation, if an insulating sample is analyzed. A temperature-controlled sample holder enables the acquisition of SSIMS data as a function of surface temperature (temperature-programmed SIMS, TPSIMS). With such a TOF-imaging SSIMS instrument, the useful mass range is extended beyond 10 000 amu; the mass resolution, m/Δm, is ∼10 000 with simultaneous detection of all masses, and within each image all masses can be detected. The number of data generated in a short time is enormous, and very sophisticated data-acquisition systems are required to handle and process the data.
7.3 Quantification
By bombarding a surface consisting of species A with primary ions, the surface coverage of A is reduced. Particles of A can be removed by desorption, by driving them into a deeper layer or, for molecular species, by fragmentation. The ratio of the number of sputtered particles to the number of primary ions is given by the disappearance yield YD(A): YD (A) =
No. of sputtered A No. of primary ions
(7.1)
A quantity related to the disappearance yield for a particle of a species which covers a solid surface with a surface density ϑ(A), is the disappearance cross-section, σD(A):
σ D (A) =
YD (A) ϑ (A)
(7.2)
where σD(A) corresponds to the average area damaged by one primary ion impact. The change of the surface density ϑ(A) with increasing primary ion fluence FPI is given by Equation 7.3:
ϑ (A, FPI ) = ϑ (A)0 exp(−σ D (A)FPI )
(7.3)
Some particles sputtered from the surface are neutral, whereas others are charged. Molecular particles can be emitted either as intact molecules or fragmented. The probability of the desorption of A into the emission channel X qi is given by the transformation probability P(A → X qi ):
123
124
7 Static Secondary Ion Mass Spectrometry (SSIMS)
P(A − X qi ) =
No. of emitted particles X qi No. of sputtered A
(7.4)
where i is an index distinguishing between the different emission channels and q is the charge of the particle Xi with q = 0, ±1, ±2, ±3, … The transformation probability for ionization cannot be measured easily, because it is highly dependent on matrix and concentration. This is, in fact, the principal obstacle to achieving proper quantification in SSIMS (“matrix effect”). The average number of emitted particles X qi per incident primary ion is given by the secondary yield Y(q) (X qi ): Y (q)(X qi ) =
No. of particles X qi No. of primary ions
(7.5)
For the detection of sputtered ions, the transmission T(X qi ) of the mass spectrometer and the detection probability D(X qi ) must be taken into account. Their product gives the probability of detecting one sputtered particle. For a TOF mass spectrometer, both T(X qi ) and D(X qi ) are typically 10–50%. The number of detected particles per incident primary ion is given by the detected yield Y(X qi ): Y(X qi ) =
No. of detected X qi No. of primary ions
(7.6)
An important quantity is the so-called useful yield Yu (X qi (A)): Yu (X qi (A)) =
No. of detected X qi No. of sputtered A
(7.7)
The useful yield provides an overall measure of the extent to which the sputtered material is used for analysis. It is a quantity employed to estimate the sensitivity of the mass spectrometric method. Values of Yu (X qi (A)) for elements typically range from 10−6 to 10−2 in TOF-SIMS. The number of sputtered particles A per incident primary ion (the “sputtering yield”) can be measured from elemental and multielemental standards under different operational conditions and can, therefore, by judicious interpolation between standards, be estimated with reasonable accuracy for the material being analyzed. Since measuring A can be problematic, quantification is normally performed using relative sensitivity factor (RSF) methods. If a species A on the surface is detected by the ion X qi , the ratio of the detected ion current I A (X qi ) to the primary ion current IPI and the surface density ϑ(A) is called the practical sensitivity factor S p (X qi (A)): S p (X qi (A)) =
I A (X qi ) ϑ (A) ⋅ IPI
(7.8)
The ratio of two practical sensitivity factors is called the RSF S(X qi (A), X qi (B)), which is independent of the primary ion current IPI:
7.4 Spectral Information
S(X qi (A), X qi ′′ (B)) =
S(X qi (A)) I A (X qi ) ϑ (B) = ⋅ S(X qi ′′ (B)) IB (X qi ′′ ) ϑ (A)
(7.9)
By using RSFs it is possible to determine surface densities of other species, if the surface density of the reference species ϑ(B) is known. Although the RSF contains matrix-dependent quantities, their variations are damped to some extent by virtue of taking ratios, and in practice the RSF is assumed constant for low concentrations of A (e.g., <1 atom%). It can be evaluated from measurements on a well-characterized set of standards containing A in known dilute concentrations, although the accuracy of the method is not as high as in laser-SNMS and XPS.
7.4 Spectral Information
A SSIMS spectrum, like any other mass spectrum, consists of a series of peaks of different intensity – that is, ion current – occurring at certain mass numbers. The masses can be allocated on the basis of atomic or molecular mass-to-charge ratio. Many of the more prominent secondary ions from metal and semiconductor surfaces are singly charged atomic ions, which makes the allocation of mass numbers slightly easier. Masses can be identified as arising either from the substrate material itself, from deliberately introduced molecular or other species on the surface, or from contaminations and impurities on the surface. Complications in allocation often arise from isotopic effects. Although some elements have only one principal isotope, for many others the natural isotopic abundance can make identification difficult. Figure 7.5 shows the positive SSIMS spectrum of a silicon wafer, illustrating both the allocation of peaks and potential isobaric problems. SSIMS reveals many impurities on the surface, particularly hydrocarbons, for which it is especially sensitive. The spectrum also demonstrates reduction of isobaric interference by high-mass resolution. For reasons discussed in Section 7.3, the peak heights cannot be taken to be directly proportional to the concentrations on the surface, and standards must be used to quantify trace elements. The relationship between what is recorded in a SSIMS spectrum and the chemical state of the surface is not as straightforward as in XPS and AES (Chapters 2 and 3). Due to the large number of molecular ions that occur in any SSIMS spectrum from a multicomponent surface – for example, during the study of a surface reaction – much chemical information is obviously available in SSIMS, potentially more than in XPS. The problem in using the information from a molecular ion lies in the uncertainty of knowing whether or not the molecule represents the surface composition. For some materials, such as polymers, the clusters observed are definitely characteristic of the material, as seen in Figure 7.6 [15], in which the SSIMS spectrum from polystyrene measured using TOF-MS contains peaks spaced at regular 104-mass-unit intervals, corresponding to the polymeric repeat
125
126
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.5 High-mass-resolution TOF-SIMS spectrum of a contaminated Si wafer (m/Δm = 15 000).
Figure 7.6 Positive SSIMS spectrum of polystyrene (Mn = 5100, Mn,calc = 4500, Mn,corr = 5600)
from TOF-MS [15].
unit. SSIMS measurements have also proven to be a very powerful tool for the characterization of additives in plastic materials. On the other hand – for example, during oxidation studies – various metal–oxygen clusters are usually found in both positive and negative secondary ion spectra, the relationship of which to the actual chemical composition is problematic. SSIMS spectra recorded after the interaction of oxygen with a half-monolayer of potassium adsorbed onto a silver substrate are shown in Figure 7.7 [16]. Here, the positive and negative spectra both show evidence of KiOj clusters, which the authors were unable to relate to the surface condition, although quite clearly much chemical information is available. On occa-
7.5 Applications
Figure 7.7 SSIMS spectra after interaction of oxygen with a half-monolayer of potassium
adsorbed onto silver, at 400 K [16]. (A) Positive SSIMS spectrum; (b) Negative SSIMS spectrum.
sion, the use of pure and well-characterized standards may be helpful when interpreting complex spectra. In conclusion, SSIMS spectra provide not only evidence of all the elements present, but also detailed insight into the molecular composition. Quasimolecular ions can be desorbed intact up to 15 000 u, depending on the particular molecule [17], and on whether an effective mechanism of ionization is present. Larger molecules can be identified from fragment peak patterns which are characteristic of the particular molecules. If the identity of the material being analyzed is completely unknown, then spectral interpretation can be accomplished by comparing the major peaks in the spectrum with those in a library of standard spectra.
7.5 Applications
In general, SSIMS tends not to be used for the quantification of surface composition due to all of the uncertainties described above. Rather, its application has been more qualitative in nature, with emphasis on its advantages of high surface specificity, very high sensitivity for certain elements and molecules, the multiplicity of chemical information, and high spatial resolution in the TOF imaging mode. SSIMS can be used to determine the surface composition, and the surface contaminants present, as a function of lateral position for all types of material, including metals, alloys, semiconductors, oxides, carbides, ceramics, glasses, organic molecular species, organometallic compounds, biomolecules, and polymers. SSIMS can also be used to study surface phenomena such as adhesion, chemisorption and physisorption, wettability, wear and tear, lubrication, chemical reactivity, corrosion, surface diffusion, and segregation. As a common analytical tool, SSIMS
127
128
7 Static Secondary Ion Mass Spectrometry (SSIMS)
is used for many applications, including coatings, composite materials, micro- and nanoelectronics, heterogeneous catalysis, biosensors, combinatorial chemistry, pharmaceuticals, medical implants, chromatographies, sensors, painting, environment control and protection, and failure analysis. 7.5.1 Oxide Films
Compared with XPS and AES, the higher surface specificity of SSIMS (one to two monolayers compared to two to eight monolayers) can be useful for the more precise determination of the chemistry of an outer surface. Although, based on details of the O1s spectrum, XPS could provide the information that OH and oxide were present on a surface, and from the C1s spectrum that hydrocarbons and carbides were present, only SSIMS could be used to identify the particular hydroxide or hydrocarbons. In the growth of oxide films for different purposes – such as passivation or anodization – such information is valuable because it provides a guide to the quality of the film and the nature of the growth process. One important area in which the properties of very thin films play a crucial role is the adhesion of aluminum to aluminum in the aircraft industry. In order for bonds of the required strength to form, the aluminum surfaces must be pretreated by anodization, which creates not only the correct chemical conditions for adhesion but also the correct structure in the resulting oxide film. Films formed by anodizing aluminum in phosphoric acid under ac conditions have been studied by Treverton et al. [18], using both SSIMS and XPS. Knowledge both of the chemistry of the outer monolayers in terms of aluminum oxides and hydroxides, and of the nature and extent of any contamination introduced by the anodizing process, was important. The positive and negative SSIMS spectra obtained after anodizing for 3 and 5 s at 20 V and 50 Hz are compared in Figure 7.8. Both sets of spectra show that the longer anodizing treatment resulted in a significant reduction in hydrocarbon contamination (peaks at 39, 41, 43, and 45 amu in the positive spectrum, and 25, 41, and 45 amu in the negative), and marked increases in oxide- and hydroxide-associated cluster ions at 44, 61, 70, and 87 amu (positive) and 43 and 59 amu (negative), respectively. At higher masses, peaks allocated to Al2O4H−, Al2O3−, and Al3O5− were also identified. Residual PO2− and PO3− from the anodizing bath were reduced as a result of the longer anodizing time. This low level of contaminants is important, because the adhesive coating can then wet the surface more effectively, leading to an improved adhesion. In addition, the change from AlO+ (mass 43) to AlOH+ (mass 44) in aluminum chemistry as a result of changing from 3- to 5-s anodizing is a significant effect that could not have been established by XPS. 7.5.2 Interfaces
The protection of steel surfaces with paint depends significantly on the chemistry of the paint–metal interface. The system has many variables, because the metal
7.5 Applications
Figure 7.8 SSIMS spectra of oxide films on aluminum after anodization in phosphoric acid
for 3 or 5 s [18]. (a) Positive SSIMS spectrum, 3 s; (b) Positive SSIMS spectrum, 5 s; (c) Negative SSIMS spectrum, 3 s; (d) Negative SSIMS spectrum, 5 s.
surface is usually pretreated in a variety of ways, including galvanizing and phosphating. Indeed, there are probably several interfaces of importance, and corrosion protection might be a function of the conditions at all of these. As one of the final steps in motor car production, visual checks are performed to determine the quality of the paint layers. If defects are observed, the car will
129
130
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.9 Video images, ion-induced secondary electron images, and mass-resolved ion images of a defect in car paint. The white signal corresponds to high intensities, and black signal to low intensities [19].
Figure 7.10
Positive TOF-SIMS spectra of the defective area and from a spot remote from the
defect [19].
not be processed further, and paint repair is necessary. The analysis of such a defect (a crater in a red paint coat) is shown in Figures 7.9 and 7.10 [19]. The upper sequence of Figure 7.9 depicts a series of optical images of the defect for different fields of view (the right image is approximately 150 × 150 μm2), with paint remnants being observed in the center of the crater. The second row
7.5 Applications
of Figure 7.9 shows ion-induced secondary electron images taken from the crater with a highly focused liquid metal ion beam. The images clearly depict the crater topography. Figure 7.10 shows the secondary ion spectra recorded from within the defect and its surroundings. The spectrum from the defective area clearly shows the presence of a polluting perfluorinated polyether structure (the Cx Fy+ peaks and the peak at mass 47 amu (CFO+) are diagnostic of the polyether structure). The third sequence in Figure 7.9 shows a mass-resolved secondary ion image recorded from the crater. On the far left, the lateral distribution of the uncharacteristic hydrocarbon C3F5+ can be seen. The signal is enhanced in the crater due to an increased sputtering yield. Topographic information can be obtained from hydrocarbon images. The next image shows the lateral distribution of the silicone oil, which is incorporated into the paint as a smoothing agent. It can be seen that silicone oil is detected only from the paint itself, and not in the crater bottom; thus, the silicone oil can be excluded as the cause of the defect. The final three images show the lateral distributions of several Cx Fy+ fragments, indicating unambiguously that the perfluorinated polyether is found exclusively in the crater. Later, it was found that the polluting lubricant droplets had originated from the transport belts used in the production, and had fallen into the paint bath so as to prevent adhesion of the paint to the metal. It was concluded that the high sensitivity of SSIMS in the detection of submonolayer coverage of organic species makes it an extremely powerful tool for solving such interface problems. 7.5.3 Polymers
The treatment of polymer surfaces to improve their wetting, water repulsion, and adhesive properties is now a standard procedure. The treatment is designed to change the chemistry of the outermost groups in the polymer chain, without affecting bulk polymer properties. Any study of the effects of treatment therefore will require a technique that is specific mostly to the outer atomic layers, which explains why SSIMS is extensively used in this area. Plasma etching is a favored form of surface treatment. Depending on the conditions of the plasma discharge – that is, the nature of the discharge gas, the gas pressure, rate of gas flow, discharge power and frequency, and substrate polarity – plasma etching can alter the chemical characteristics of a surface over a wide range. A plasma discharge is, by nature, highly controllable and reproducible. The effects of plasma treatment on the surface of polytetrafluoroethylene (PTFE) have been studied by Morra et al. [20] by using a combination of SSIMS and XPS. These authors found, as is apparent from Figure 7.11, that not only did the plasma change the surface chemistry, but the surface emission fragments themselves led to a temporary change in the plasma discharge conditions. Thus, the untreated (Figure 7.11a) and 15-min oxygen plasma-treated (Figure 7.11c) positiveion spectra are virtually identical, whereas the 0.5-min oxygen plasma-treated
131
132
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.11 Positive SSIMS spectra of PTFE [20]. (a) Untreated; (b) 0.5 min oxygen plasma
treatment; (c) 15 min oxygen plasma treatment.
spectrum (Figure 7.11b) is quite different. The initial break-up of the PTFE surface produced hydrocarbon fragments which reacted with the oxygen to produce water vapor, thus changing the local nature of the discharge gas. The important deduction to be made here is that one of the vital conditions in plasma treatment is the length of the contact time. In this instance, the surface was unchanged if the contact time was too long. Other SSIMS studies of polymer surfaces have included perfluorinated polyether [21], low-density polyethylene [22], poly(ethylene terephthalate) [23], and the oxidation of polyetheretherketone [24].
7.5 Applications
7.5.4 Biosensors
Biosensor chip technology has been a subject of growing interest for a variety of applications. In particular, DNA-sequencing chips based on the method of sequencing by hybridization are becoming increasingly important. For this, unknown DNA sequences, which are typically labeled with radioactive or fluorescent markers, are hybridized to known complementary short DNA sequences called oligodeoxynucleotides (ODNs), which have been immobilized on a solid surface [25–27]. This method has become invaluable for clinical diagnostics, the sequencing of cDNAs or the partial sequencing of clones, DNA and RNA sequencing, gene polymorphism studies, and the identification of expressed genes. The use of peptide nucleic acid (PNA) hybridization biosensor chips [26, 27] would even enable the sequencing of unlabeled DNA. PNA is a synthesized DNA analog in which both the phosphate and deoxyribose of the DNA backbone are replaced by polyamides, resulting in a phosphate-free PNA backbone. This enables detection of the presence of DNA from a phosphate signal. Figure 7.12 depicts the TOF-SIMS spectra obtained from ODN and PNA immobilized on silanized silicon wafers. The spectra show clearly that the masses
Figure 7.12 Negative TOF-SIMS spectra obtained from oligodeoxynucleotides (ODN) (left) and PNA (right) [26].
133
134
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.13 TP-SIMS measurements of an ODN (c = 0.1 mm) and PNA (c = 0.1 mm, curve
labeled as PNA) TTTTCCCTCTCTC-sequence [26].
corresponding to PO2− and PO3− provide the best correlation of the presence of ODN, enabling their use for precise distinction between ODN and PNA. The CF3− and C2O2F3− peaks seen in the PNA spectra represent trifluoroacetic acid, which was part of the PNA solution. Deprotonated (Cyt-H)− and (Thy-H)− signals of the bases cytosine and thymine are observed for both immobilized PNA and ODN sequences, and can be used to detect the presence of these bases. Figure 7.13 shows the thermal stability of immobilized ODN and PNA. The signal for the Thy- and Cyt-bases obtained with temperature-programmed (TP)SIMS begins to decrease at approximately 150 °C for ODN and at 200 °C for PNA. This variance is caused by the different strengths of binding between the bases and the sugar–phosphate and peptide backbones, respectively. The data show that SSIMS can be used as a tool to characterize the different steps in the production of biosensors, or even for sequencing. Similarly, SSIMS can be used to solve a variety of problems in bioanalytical chemistry, including the screening of combinatorial libraries and the characterization of Langmuir– Blodgett layers. 7.5.5 Surface Reactions
Due to the inherently destructive nature of ion bombardment, the use of SSIMS alone in the study of the reactions of surfaces with gases and vapor must be viewed with caution, although in combination with other surface techniques it can provide valuable additional information. The parallel techniques are most often XPS, TDS, and LEED, and the complementary information required from SSIMS normally
7.5 Applications
Figure 7.14 Positive SSIMS spectrum from Ru (001) after exposure to 0.5 monolayer thiophene at 95 K [28].
refers to the nature of molecules on surfaces and with which other atoms, if any, they are combined. A typical SSIMS spectrum of an organic molecule adsorbed on a surface is that of thiophene on ruthenium at 95 K, as shown in Figure 7.14 [28]. Here, the exposure was 0.5 Langmuir only (i.e., 5 × 10−7 torr s = 37 Pa s), and the principal positive ion peaks are those from ruthenium, consisting of a series of seven isotopic peaks around 102 amu. Ruthenium–thiophene complex fragments are, however, found at ca. 186 and 160 amu; each has the same complicated isotopic pattern, indicating that interaction between the metal and the thiophene occurred even at 95 K. In addition, thiophene and protonated thiophene peaks are observed at 84 and 85 amu, respectively, with the implication that no dissociation of the thiophene had occurred. The smaller masses are those of hydrocarbon fragments of different chain lengths. SSIMS has also been used to study the adsorption of propene on ruthenium [29], the decomposition of ammonia on silicon [30], and the decomposition of methane thiol on nickel [31]. 7.5.6 Imaging
SSIMS has been used in the TOF-SSIMS imaging mode to study very thin layers of organic materials [32–36], polymeric insulating materials [37], and carbon fiber
135
136
7 Static Secondary Ion Mass Spectrometry (SSIMS)
Figure 7.15 Secondary ion images of blends melt-mixed for 20 min (top) and 40 min (bottom).
Images A and B were obtained by adding the mass-resolved images pertaining to the masses listed below each image. Also shown are the overlays of A (black) and B (white) [39].
and composite fracture surfaces [38]. In these studies, a spatial resolution of approximately 80 nm in mass-resolved images was achieved. Figure 7.15 shows several mass-resolved images of polymer blends [Nylon 6 (Ny6) and bisphenol A polycarbonate (PC)], mixed for 20 and 40 min, respectively [39]. The reactive blending of polymers and exchange reactions that can occur during the melt–mixing process have attracted considerable interest from many research groups. The reactive mixing of immiscible polymers can lead to a partial compatibilization of the two components with properties advantageous for technological applications. It is evident that the 20-min sample is quite inhomogeneous, with large distinct domains of different composition, whereas for the 40-min sample the distribution is more homogeneous. A detailed spectral analysis of the different imaging regions supported the view that reactive mixing and compatibilization of Ny6/PC blends occurred via the incorporation of PC in the nylon phase. TOF-SIMS imaging can also be used to detect extremely small particles on a variety of substrates, for example, silicon wafers or photographic films [40]. Images of silver bromide and silver chloride crystals, demonstrating the high spatial resolution (50 nm) possible with TOF-SIMS, are shown in Figure 7.16. Current developments aim to reach a spatial resolution for chemical mapping of approximately 20 nm, or even 10 nm.
7.5 Applications
Figure 7.16 High-resolution TOF-SIMS images of silver bromide and silver chloride crystals.
Figure 7.17 Multielement depth profile of a B layer in Si performed with 600 eV SF5+
sputtering [41].
7.5.7 Ultra-Shallow Depth Profiling
TOF-SIMS has also proved to be a very powerful tool for ultra-shallow depth profiling, having the advantage of simultaneously detecting all elements of interest. The dual-beam mode [41], in particular (see Section 8.2.1), enables an optimized depth resolution, because the sputtering conditions can be independently optimized. Figure 7.17 shows an ultra-shallow TOF-SIMS depth profile of a 100-eV B-implant in Si, capped with 17.3 nm Si. The measurement was performed with 600-eV
137
138
7 Static Secondary Ion Mass Spectrometry (SSIMS)
SF5+-sputtering and with O2-flooding. The original wafer surface, into which the B was implanted, is indicated by the maxima of the alkali- and C-signals. Due to these contaminants, a minimum is observed in the 30Si-signal. The dynamic range of the B-profile is more than 3.5 decades, and the depth resolution ≤0.5 nm.
References 1 Benninghoven, A. (1969) Z. Phys., 220, 159–180. 2 Sigmund, P. (1977) Sputtering processes: Collision cascades and spikes, in Inelastic Ion-Surface Collisions (eds N.H. Tolk, J.C. Tully, W. Heiland, and C.W. White), Academic Press, New York, pp. 121–152. 3 Rivière, J.C. (1990) Surface Analytical Techniques, Oxford University Press, Oxford. 4 Stapel, D., Brox, O., and Benninghoven, A. (1999) Appl. Surf. Sci., 140, 156–167. 5 Heile, A., Lipinsky, D., Wehbe, N., Delcorte, A., Bertrand, P., Felten, A., Houssiau, L., Pireaux, J.-J., De Mondt, R., Van Vaeck, L., and Arlinghaus, H.F. (2008) Appl. Surf. Sci., 255, 941–943. 6 Benninghoven, A., Rüdennauer, F.G., and Werner, H.W. (1987) Secondary Ion Mass Spectrometry, John Wiley & Sons, Inc., New York. 7 Prewett, P.D. and Jefferies, D.K. (1980) J. Phys. D, 13, 1747–1755. 8 Krauss, A.R. and Gruen, D.M. (1977) Appl. Phys., 14, 89–97. 9 Steffens, P., Niehuis, E., Friese, T., Greifendorf, D., and Benninghoven, A. (1985) J. Vac. Sci. Technol. A, 3, 1322–1325. 10 Niehuis, E., Heller, T., Feld, H., and Benninghoven, A. (1987) J. Vac. Sci. Technol. A, 5, 1243–1246. 11 Karataev, V.I., Mamyrin, B.A., and Shmikk, D.V. (1972) Sov. Phys. Techn. Phys., 16, 1177–1179. 12 Schueler, B.W. (1992) Microsc. Microanal. Microstruct., 3, 119–139. 13 Sakurai, T., Matsuo, T., and Matsuda, H. (1985) Int. J. Mass. Spectrom. Ion Phys., 63, 273–287. 14 Iltgen, K., Bendel, C., Niehuis, E., and Benninghoven, A. (1997) J. Vac. Sci. Technol. A, 15, 460–464.
15 Bletsos, I.V., Hercules, D.M., Benninghoven, A., and Greifendorf, D. (1986) TOF-SIMS of Polymers in the range of M/Z = 500 to 5000, in SIMS V (eds A. Benninghoven, R.J. Colton, D.S. Simons, and H.W. Werner), SpringerVerlag, Berlin, pp. 538–541. 16 McBreen, P.H., Moore, S., Adnot, A., and Roy, D. (1988) Surf. Sci., 194, L112–L118. 17 van Leyen, D., Hagenhoff, B., Niehuis, E., Benninghoven, A., Bletsos, I.V., and Hercules, D.M. (1989) J. Vac. Sci. Technol. A, 7, 1790–1794. 18 Treverton, J.A., Ball, J., Johnson, D., Vickerman, J.C., and West, R.H. (1990) Surf. Interface Anal., 15, 369–376. 19 Hagenhoff, B. and Rading, D. (1998) Ion Beam Techniques: Surface Mass Spectrometry, in Handbook of Surface and Interface Analysis (eds J.C. Rivière and S. Myhra), Marcel Dekker, New York, pp. 209–253. 20 Morra, M., Occhiello, E., and Garbassi, F. (1990) Surf. Interface Anal., 16, 412–417. 21 Fowler, D.E., Johnson, R.D., van Leyen, D., and Benninghoven, A. (1991) Surf. Interface Anal., 17, 125–136. 22 Leggett, G.J., Briggs, D., and Vickermnan, J.C. (1991) Surf. Interface Anal., 17, 737–744. 23 Ramsden, W.D. (1991) Surf. Interface Anal., 17, 793–802. 24 Pawson, D.J., Ameen, A.P., and Short, R.D. (1992) Surf. Interface Anal., 18, 13–22. 25 Arlinghaus, H.F., Kwoka, M.N., and Jacobson, K.B. (1997) Anal. Chem., 69, 3747–3753. 26 Arlinghaus, H.F., Höppener, C., and Drexler, J. (2000) TOF-SIMS Characterization of DNA and PNA Biosensor Chips, in SIMS XII
References
27
28 29
30
31 32 33 34
35
36
(eds A. Benninghoven, P. Bertrand, H.N. Migeon, and H.W. Werner), Elsevier Science, Amsterdam, pp. 951–954. Hellweg, S., Jacob, A., Hoheisel, J.D., Grehl, T., and Arlinghaus, H.F. (2006) Appl. Surf. Sci., 252, 6742–6745. Cocco, R.A. and Tatarchuk, B.J. (1989) Surf. Sci., 218, 127–146. Sakakini, B.H., Ransley, I.A., Oduoza, C.F., Vickerman, J.C., and Chesters, M.A. (1992) Surf. Sci., 271, 227–236. Zhou, X.-L., Flores, C.R., and White, J.M. (1992) Surf. Sci. Lett., 268, L267–L273. Castro, M.E. and White, J.M. (1991) Surf. Sci., 257, 22–32. Briggs, D. and Hearn, M.J. (1988) Surf. Interface Anal., 13, 181–185. Benninghoven, A. (1994) Angew. Chem. Int. Ed. Engl., 33, 1023–1043. Benninghoven, A., Hagenhoff, B., and Niehuis, E. (1993) Anal. Chem., 65, 630A–640A. Bourdos, N., Kolmer, F., Benninghoven, A., Sieber, M., and Galla, H.-J. (2000) Langmuir, 16, 1481–1484. Arlinghaus, H.F. (2008) Appl. Surf. Sci., 255, 1058–1063.
37 Briggs, D., Hearn, M.J., Fletcher, I.W., Waugh, A.R., and McIntosh, B.J. (1990) Surf. Interface Anal., 15, 62–65. 38 Hearn, M.J. and Briggs, D. (1991) Surf. Interface Anal., 17, 421–429. 39 Licciardello, A., Puglisi, C., Lipinsky, D., Niehuis, E., and Benninghoven, A. (1997) Time-of-Flight SIMS Characterization of Nylon-6/ Polycarbonate Blends, in SIMS X (eds A. Benninghoven, B. Hagenhoff, and H.W. Werner), John Wiley & Sons, Inc., New York, pp. 711–714. 40 Kollmer, F., Kamischke, R., Ostendorf, R., Schnieders, A., Kim, C.Y., Lee, J.W., and Benninghoven, A. (2000) TOF-SIMS and Laser-SNMS Analysis of sub-μm Particles, in SIMS XII (eds A. Benninghoven, P. Bertrand, H.N. Migeon, and H.W. Werner), Elsevier Science, Amsterdam, pp. 329–332. 41 Iltgen, K., Benninghoven, A., and Niehuis, E. (1998) TOF-SIMS Depth Profiling with Optimized Depth Resolution, in SIMS XI (eds G. Gillen, R. Lareau, J. Bennett, and F. Stevie), John Wiley & Sons, Inc., New York, pp. 367–370.
139
141
8 Dynamic Secondary Ion Mass Spectrometry (SIMS) Herbert Hutter
Today, dynamic secondary ion mass spectrometry (SIMS) is a standard technique for the measurement of trace elements in semiconductors, high-performance materials, coatings, and minerals. The main advantages of the method are the excellent detection limit of less than 1 ppm (part per million) for all elements, the isotopic sensitivity, the inherent possibility of measuring depth profiles, and the fast direct imaging and three-dimensional (3-D) species distribution detection capability. The first SIMS instrument was built by Herzog and Viehboeck [1] in Vienna during the 1940s. In the early 1960s, and almost simultaneously with Casting and Slozian [2] in Paris, Herzog and Liebl [3] built the first sophisticated SIMS instrument. The acronym SIMS was first used by Benninghoven in 1970 [4].
8.1 Principles
Bombardment of the sample with a dose of high-energy primary ions (1–20 keV) results in destruction of the initial surface and surface-adjacent regions (see also Section 7.1). If the primary ion dose is higher than 1011 ions mm−2, the assumption of an initial intact surface is no longer valid, and a sputter equilibrium is reached at a depth greater than the implantation depth of the primary ions. The permanent bombardment of the sample with primary ions leads to various effects described in the following sections. 8.1.1 Compensation of Preferential Sputtering
The species with the lower sputter yield is enriched at the surface; this effect – which is termed preferential sputtering – complicates Auger measurements, for example. Such an enrichment also compensates the different sputter yields, so that in dynamic SIMS the sputtered atoms have the same composition as the sample.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
142
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
8.1.2 Atomic Mixing
Depending on their mass, energy, and impact angle, the primary ions reach a mean depth until they are stopped by several collisions with sample atoms. Furthermore, the sample atoms are moved from their initial locations, and this results in a mixing of the surface-adjacent region, thus limiting the depth resolution. The use of low primary ion energies and flat bombardment angles can reduce this effect, however (see also Section 7.1). 8.1.3 Implantation of Primary Ions
The primary ions are implanted in the sample, and thereby influence the chemical constitution. When using energies in the region of 20 keV, the implantation depth is about 30 nm. The sputter coefficient – the ratio of secondary to primary particles (not ions) – is energy-dependent and has a maximum in the order of 10 keV. 8.1.4 Crater Bottom Roughening
Depth resolution is also limited by roughening of the crater under ion bombardment, due to the different sputter coefficients of the different crystal orientations. The sputter coefficients of single crystals can vary by a factor of two, depending on the orientation. On polycrystalline samples, the depth resolution is reduced with increasing depth due to roughening of the surface. 8.1.5 Sputter-Induced Roughness
The measurement of the depth profiles of single crystal semiconductors (notably in the analysis of Group II-VI semiconductors) with Cs+ ions results in a roughening of the surface. Although various models have been discussed for many years, there is to date no generally accepted explanation for this effect. This effect, and the corresponding reduction in depth resolution, can be avoided by rotating the sample during the measurement, however [5, 6]. 8.1.6 Charging Effects
The electrostatic potential of nonconducting samples and layers is changed due to bombardment with charged atoms, and the emission of secondary electrons and ions. Usually, this results in a positive charging of the sample and a drift of the secondary ion energy. On good insulators this charging reaches such high values that the primary ions can no longer reach the surface, and the analysis is impos-
8.2 Instrumentation
sible. To compensate for this effect, however, low-energy electrons can be directed to the surface (this is termed an electron shower).
8.2 Instrumentation
Typically, SIMS instruments consist of two fundamental parts: one unit produces the primary particles for bombardment of the sample surface, while the second unit analyzes and detects the particles emitted from the surface. In SIMS the primary particles are ions (primary ions) that impact the sample surface, leading to the emission of neutrals and also ions – the so-called secondary ions. The secondary ions originate from the sample material, and are analyzed on the basis of their mass. Depending on the total intensity, beam diameter, and type of ions, a variety of different ion sources are used. The use of several mass analyzers, each with different characteristics, enables a broad field of application. 8.2.1 Ion Sources
Bombardment with reactive species increases the ionization yield and, therefore, the sensitivity. Electropositive elements are more sensitive under oxygen bombardment, and electronegative elements under cesium bombardment (see Section 8.3). Therefore, in dynamic SIMS Cs+ or O2+ ions are normally used for sample sputtering. Due to the lower melting point of gallium – and therefore of the higher brightness of a Ga+ source – this type of primary ion is also used for fine focus imaging applications, although the sensitivity is lower. The Cs+ and Ga+ liquid metal ion sources (LMIS) are described in Section 7.2.1. O2+ beams are produced in an electron beam ion source (EBIS), or in a duoplasmatron ion source (Figure 8.1).
Figure 8.1 Scheme of a duoplasmatron ion source.
143
144
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
8.2.1.1 Duoplasmatron For bombardment with argon or oxygen, gas discharge ion sources are used. In particular, the use of O2+ ions is common due to the chemical enhancement effect (see Section 8.3). With a duoplasmatron, an ion beam of several μA intensity can be generated, while the diameter of the beam is limited typically to 0.5 μm. The achievable density of the beam (ions·μm−2·s−1) is an inherent property of the ion source, the so-called brightness. Thus, any reduction in the beam diameter is related to a reduction in the total intensity of the beam current. 8.2.2 Mass Analyzer
Three types of mass analyzer are commonly used in SIMS instruments, namely the magnetic sector field, the quadrupole, and the time-of-flight (TOF) analyzer. The quadrupole and TOF analyzers are described in Section 7.2.2. 8.2.2.1 Magnetic Sector Field In a magnetic field B, an ion with the velocity v and the charge q experiences a centripedal force, the Lorenz force F: F = q ⋅ ( v × B)
This force must be balanced by the centrifugal force of the ion mv2/r, where r is the radius of curvature of the trajectory. Hence, mv 2 = q ⋅ ( v × B) r Due to this force, particles with an energy E = mv2/2 are held on circular paths with the radius r: r=
2E(m/q) B
Secondary ions are accelerated from the surface to a certain energy. After passing an entrance slit, the different masses are forced to different radial paths in a homogeneous magnetic field. As only one mass is able to pass the exit slit and reach the detector, it is possible, by scanning the magnetic field, to detect all of the masses sequentially. The mass can be calculated by: m B2r 2 = 2E q The formation probability of singly ionized atoms is much higher than that of higher charged particles, so q can be set to 1. In commercial instruments, two forms are realized (currently, only one commercial producer of magnetic sector field instruments exists):
8.2 Instrumentation
Figure 8.2 Scheme of a magnetic sector instrument with stigmatic secondary ion optics
consisting of an electrostatic analyzer (ESA) and a magnet sector field.
Figure 8.3 Scheme of a magnetic sector instrument with simultaneous mass detection.
•
The direct imaging magnetic sector mass analyzer (Figure 8.2) has the unique property that the secondary ion optics is stigmatic. This means that the surface is directly projected to the analyzer, at all points simultaneously.
•
The Mattauch–Herzog geometry (Figure 8.3) allows the detection of several masses simultaneously, and is therefore ideal for scanning instruments [7]. The detectors (up to five in number) are adjusted mechanically to the masses of interest. Thus, within a certain mass range it is possible to detect for example, all isotopes of one element simultaneously. This allows a rapid, sensitive, and precise measurement of images of different masses to calculate the isotope ratios. The only commercial instrument of this type uses a coaxial optical design of the ion gun, and secondary ion extraction. As a consequence, a lateral resolution down to 50 nm is possible.
8.2.2.2 Detector In order to measure the total ion intensities, electron multipliers are used. Since, over a count rate of 2 × 106 s−1 they are saturated, a Faraday cup (FC) is used for
145
146
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
high signals. In electron multipliers, the intensity is determined by digital counting of individual ions, while FCs deliver a current which is proportional to the intensity (or counting rate). The amplification of the output of the FC must be adjusted in such a way that it is identical to the output of an electron multiplier for a given signal intensity. This is particularly important to avoid any sudden jumps in depth profiles on switching between the FC cup and the electron multiplier. For measurement of the local ion intensities in direct imaging mode, detection with laterally resolved single ion sensitivity is necessary. Depending on the sensitivity of the detector receiving the signal from the channel plate, either single- or double-channel plates are used. Currently, two laterally resolving detectors are in use:
•
The CCD camera: For standard CCD cameras a double channel plate (amplification >106) is necessary, whereas for high-sensitivity cameras (sensitivity >10−3 lux, cooled or with internal amplification) a single channel plate is sufficient. Controlling the channel plate’s high voltage (= amplification) opens up a high dynamic range [8].
•
The resistive anode encoder (RAE): This detector has the advantage that the single ion events are detected digitally. Therefore, it delivers quantitative results, independent of any local differences in the amplification of the channel plate. However, one disadvantage is that the count rate is limited to 200 000.
8.3 Spectral Information
The element sensitivity is determined by the ionization probability of the sputtered atoms, with the probability in turn being influenced by the chemical state of the surface. In dynamic SIMS, Cs+ or O2+ ions are used for sample bombardment, because these elements cause an increase in the ionization probability, the socalled chemical enhancement effect. The surface-adjacent region is partially oxidized in the case of O2+ bombardment, and during the sputter process the chemical bonding of the oxides is broken. As the binding electron has a higher affinity to oxygen, the (metallic) binding partner will leave the surface positively ionized. Bombardment with Cs+ results in a coverage of the surface with Cs, which in turn reduces the work function. This results in a higher probability that one electron from the conducting band will tunnel from the surface to the leaving atom during the sputter process. This leads to a higher rate of negatively charged ions, especially of electronegative elements. The effects of so-called “reactive sputtering” with O2+ and Cs+ have been described on a theoretical basis, and verified experimentally. Unfortunately, due to the complex surface structures of real samples, there is currently no quantification algorithm available based on physical models.
8.4 Quantification
8.4 Quantification
As the ionization probability depends heavily on the matrix, it is necessary to apply standards. 8.4.1 Relative Sensitivity Factors
Clearly, the most accurate – and therefore the most popular – quantitative method is to analyze the unknown sample together with a similar sample of known content. Generally, the relationship between measured intensity and content of each sample is defined via the relative sensitivity factor (RSF): Iel c = RSF ⋅ el Iref c ref Here, the reference element (ref) specifies a matrix element, while the other element (el) indicates the analyzed element. By using a standard sample with a known concentration, the RSF can be determined. By using standards, a relative precision down to 1 rel.% is possible. 8.4.2 Implantation Standards
The production of homogeneous solid-state standards is costly; however, dynamic SIMS has the advantage that nonhomogeneous ion implantation standards can also be used. By knowing the implantation dose, the RSF can be calculated by summing up the signal along a depth profile. Therefore, it is necessary to calculate the RSF, using the mean intensity ratio indicated by the sum in the following equation: RSF = fref cref ρ NL d fel QT M Cyc Iel (i) Iref (i)
f ref ⋅ c ref ⋅ ρ ⋅ N L ⋅ d ⋅ f el ⋅ Q T ⋅ M ⋅ Cyc
Cyc
∑I i =1
I el (i ) ref (i )
isotope abundance of the reference element concentration of the reference element (at/at) matrix density (g/cm3) Avogadro’s number (= 6.022 × 1023) depth of the crater (cm) isotope abundance of the element implanted dose of the element (atoms cm−2) molecular mass of the matrix number of measured cycles signal intensity of the element (counts s−1) of cycle i signal intensity of the reference (counts s−1) of cycle i
147
148
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
For the quantification of multilayers, each layer must be calculated separately. In particular, when the layers are oxides quantification is not possible due to the strong variation of the ionization probability in the different layers. Modern materials have a complex 3-D inner structure with many different phases, and for these samples quantification is not possible. However, in many cases technologists are interested only in the relative differences between samples of known bulk concentrations or in the element distribution, which makes SIMS also a valuable tool for this purpose. 8.4.3 Metal Ceside (MCs+) Ions
Under Cs+ bombardment, the matrix effect can be significantly reduced by using the MCs+ ion signals for quantification. The detection limit is reduced by two or more orders of magnitude, but for some cases even standard free quantification is reported [9]. One disadvantage is that MCs+ ions have high masses, and in this mass region interferences are more likely. 8.4.4 Theoretical Models
Due to the complex nature of the surface, a satisfactory theoretical description of the ionization process is not yet possible. However, different ionization mechanisms have been proposed. 8.4.4.1 Electron Tunneling Model Several models based on a quantum mechanical concept have been introduced. One suggestion is that one electron of the conduction band tunnels to the leaving atom, or vice versa. The tunneling probability depends on the ionization potential of the sputtered element, the velocity of the atom (available time for the tunneling process), and on the work function of the metal (adiabatic surface ionization, the Schroeer model [10]). 8.4.4.2 Broken Bond Model The broken bond model was developed especially to describe the ionization process under oxygen bombardment. In this model it is assumed that an oxide layer is formed on the surface; the binding electrons then remain at the oxygen, resulting in the emission of positively ionized atoms [11]. 8.4.4.3 Local Thermodynamic Equilibrium LTE The basis of this model (which has only historic importance) is that, under ion bombardment, a surface-adjacent plasma is generated in which the sputtered atoms are ionized [12]. The plasma should be under local equilibrium, so that the Eggert–Saha equation for determining the ionization probability can be used. In this case, the important parameter is the plasma temperature; by knowing the
8.6 Mass Spectra
concentration of one element, the plasma temperature can be determined and used to calculate the concentration of other species in the plasma. Although it is known that the prerequisites for a local thermal equilibrium are not really fulfilled (e.g., due to the long period of time between individual sputter events or unrealistically high temperatures determined by the model), at least semiquantitative results can be achieved with this model. This must, however, be explained rather in terms of empirical fitting, because many phenomena can be fitted empirically by exponential equations, such as the Eggert–Saha equation with free fit parameters.
8.5 Mass Spectra
Scanning from a low mass to a high mass in discrete steps picks up mass spectra. Magnetic sector field instruments have mass resolutions up to m/Δm = 20 000, while quadrupole instruments are limited to a mass resolution of approximately m/Δm = 500. Both types of instrument provide a mass range up to 500. High mass resolution is required to separate mass interferences of molecular and atomic ions. Due to the mass defect of the binding energy of the nucleus, atomic ions have a slightly smaller mass than the corresponding molecular ions, and typically a mass resolution between 5000 and 10 000 is necessary to resolve the two. There is, however, a second method available by which mass interferences can be avoided. Molecular ions have a different energy distribution compared to atomic ions, namely they have a lower portion at higher energies. Thus, by energy filtering – that is, by the gating of low-energy ions – the molecular ions can be suppressed. In practice, energy filtering is performed by placing an adjustable energy slit between the electrostatic and magnetic analyzer, and by shifting the energy distribution of the ions across that slit by applying a voltage offset. Due to a reduction in the energy acceptance of the spectrometer, this method results in a reduction of the detection limit. Figure 8.4 shows that energy filtering greatly suppresses the molecular ions, whereas the atomic ions are almost unaffected. The 63Cu+ signal is not visible without energy filtering, because many molecular interferences are present, such as 51V12C.
8.6 Depth Profiles
The measurement of depth profiles is based on detection of the masses of interest during sputter removal of the sample material. However, the evaluation of these experiments has certain limitations:
•
Depth resolution is limited by atomic mixing, the roughness of the initial sample surface, and the roughening processes that occur during sputtering.
149
150
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
Figure 8.4 Mass spectra of a high-speed steel. (a) Primary ions: O2+, measurement without energy filtering, electropositive elements are sensitive, many molecular ions are visible; (b) Same parameters as (a) but energy filtering with an offset of 300 V was
used. The molecular ion intensities are significantly reduced; (c) Primary ions: Cs+, energy filtering with an offset of 300 V was used. Electronegative elements are more sensitive.
8.6 Depth Profiles
Figure 8.5 Depth profile of a passivation layer on high-purity chromium. The 18O layer is on
top, the 16O layer is located at the interface to the metal.
•
The determination of the depth scale by the measurement of the crater depth may be hampered by the fact that multilayer systems consist of layers with different sputter yields.
•
The different chemical constitutions of the layers may also cause different RSF factors for a given element in different layers. These effects may necessitate determination of the depth scale and RSF for each layer individually.
SIMS depth profiles have no significant background; only the elements H, C, and O have a background due to residual gas. This background depends only on the gas pressure in the sample chamber, and not on the primary ion beam intensity. In order to test the influence of the signal background due to residual gas, the ion beam can be reduced for several cycles. All signals should be reduced in the same way, except for those caused by residual gas. To illustrate the potential of isotope sensitivity of SIMS, Figure 8.5 shows a depth profile of a passivation layer on high-purity chromium. The layer was produced by a two-step experiment: (i) the sample was oxidized for 30 min under air with natural isotope concentrations; and (ii) the gas was changed to a mixture of 18 O and 15N. The profile shows that the 18O layer is located on the top, indicating that the chromium had diffused to the surface, where it had reacted with the oxygen. The high dynamic range of SIMS forced the use of logarithmic scales over six or seven orders of magnitude, and this point should be borne in mind when interpreting depth profiles. For example, slight concentration variations in diffusion profiles will appear almost as straight lines on a logarithmic scale. The diffusion of oxygen through the layer to the interface at the metallic chromium was slower by a factor of 50. The enrichment of carbon at the interfaces was caused by contamination due to the gas change.
151
152
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
8.6.1 Dual-Beam Technique for TOF-SIMS Instruments
For common depth profiling, TOF instruments are not recommended due to the so-called “working cycle.” As only 0.1% of the total measurement time is used for bombardment by primary ions, it takes a long time to sputter into deeper regions. However, by using a second ion gun, and working anticyclically to the measuring beam without an applied accelerating potential, substantial sample material can be removed. One advantage of this technique is the fact that sputtering for measurement and sputtering for material removal are not coupled, and therefore the beam diameter, energy, mass, and angle of the sputter gun can be optimized independently from the measuring gun. For example, by sputtering with 200 eV primary ions a depth resolution better than 0.5 nm can be achieved. The extraction field for accelerating the secondary ions towards the TOF mass analyzer is only applied when the measurement ion beam is active. Therefore, the region over the sample is field-free during sputtering. This fact enables an efficient charge compensation for insulating samples by low-energy electrons [13]. One disadvantage of this technique is the fact that atoms sputtered during the material removal phase are not analyzed, and this reduces the detection power. Even by detecting all of the masses, this disadvantage can be only partially reduced, especially in implantation profiles where only one mass is of interest. 8.6.2 Molecular Depth Profiles
The differentiation between dynamic and static SIMS has become increasingly blurred during the past few years. The use of extremely low-energy sputter ions (200 eV, O2+, Cs+) or polyatomic ions such as C60+ (Fullerene) [14] or even Ar1000 clusters [15] enables the sputtering of sample material, without completely destroying the bonds of organic materials. This opens up the possibility of overcoming the static SIMS limit and reaching lower detection limits, especially for imaging organic or even biological samples. Currently, the depth profiling of thin organic layers is already possible for many systems [16].
8.7 Imaging 8.7.1 Scanning SIMS
In scanning mode, the image is measured sequentially point-by-point. As the lateral resolution of the element mapping in scanning SIMS depends only on the primary beam diameter, mainly liquid metal ion sources (LMIS) are used. In this
8.7 Imaging
Figure 8.6 Scanning SIMS image of carbide
structure of a high-speed steel [19]. In the vanadium (V) distribution, different phases (VC and V2C) are visible by the different intensities of the V signal. The Al distribution shows the shell structure of the nonmetallic
impurities (NMI) acting as nuclei for condensation of the hard carbide phases during casting of the material. Measurement time: 10 min, image size: 64 μm × 64 μm, primary ions: O2+, primary ion energy: 5.5 keV.
way, beam diameters down to 50 nm with high total intensities of 1 nA can be reached. The beam diameter of oxygen is limited to 0.5 μm by the use of duoplasmatrons. Nevertheless, for mapping electropositive elements this procedure must be selected in order to make use of the chemical enhancement effect. The counting time of one pixel, and the number of pixels, determine the measurement time for one image. In the case of a low concentration of the elements of interest, the Poisson statistics of the small integer values of the secondary ions determines the required measurement time. A reduction in the measurement time for element distributions is made possible by the simultaneous detection of several masses. This can be achieved only by using a magnetic sector field spectrometer with the Mattauch–Herzog geometry (see Figure 8.3), and the parallel detection of up to seven masses by using mechanically adjusted electron multipliers. Such simultaneous detection also increases the precision of the measurement of isotopic ratios, especially for small particles, which are sputtered off rather quickly, thus limiting the applicable signal acquisition time. Initially, this approach was used to determine small inclusions in rocks, meteorites [17], and dust sampled from suspected nuclear production facilities. However, during the past few years this technique has been used increasingly by microbiologists to measure the isotopic ratios of entire single cells nourished with isotope-enriched nutrients [18]. An example of imaging with scanning SIMS is shown in Figure 8.6, which demonstrates the carbide structure of a high-speed steel sample [19]. 8.7.2 Direct Imaging Mode
In direct imaging mode, all pixels (picture elements) of the element distribution are detected simultaneously by the use of a stigmatic secondary ion optics. The
153
154
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
lateral resolution of the image is not influenced by the primary ion beam diameter; rather, the resolution is determined by the energy distribution of the secondary ions. The energy (and angle) distribution results in a chromatic aberration of the first electrostatic lens, the emission lens [20]. The resolution can only be increased by limiting the energy (or angle) acceptance of the instrument, although this of course reduces the transmission and, therefore, the detection limit. In practice, this effect limits the lateral resolution to about 1 μm. A resolution down to 0.5 μm is possible only for the main components. In practice, the quality of the image is also reduced by the use of high mass resolution and energy offset, and therefore in many cases it is not possible to avoid mass interferences. However, by using image processing tools to classify the mappings of the different masses, it is possible to determine the element distributions [21]. For secondary ion detection a laterally resolving detector is necessary. In the first step, a cannel plate for amplification is used; the secondary electrons of the output of this device are then accelerated either to a fluorescent screen or to a resistive anode. In the case of a fluorescent screen, the image is picked up by a CCD camera and summed up frame-by-frame using a computer. This system has the advantage of unlimited secondary ion intensities, although compared to the digital detection of the resistive anode decoder artifacts may be introduced by lateral variation of the sensitivity and nonlinearities with respect to the detected intensity. The main advantage of the imaging mode is the rapid data acquisition. Likewise, due to the fact that all pixels are projected and detected simultaneously, the measurement time for one distribution is significantly lower. Images of a passivation layer on a niobium alloy are shown as an example of the direct imaging mode in Figure 8.7 [22].
8.8 Three-Dimensional (3-D)-SIMS
3D-SIMS is a further development combining depth analysis and imaging analysis. In imaging mode (microscope mode), the acquisition time for single ion distributions is low (1–5 s), which allows cyclic image acquisition for different masses during the measurement of a normal depth profile. Thus, measurements of large – and therefore representative – volumes (150 × 150 × 10 μm3) are possible in a short time (1 h) [23]. In scanning mode (microprobe mode), the sequential detection of the single pixels (picture elements) and voxels (volume elements) results in long measuring times. Therefore, in practice only small volumes (10 × 10 × 1 μm3) can be measured [24]. The distribution of vanadium (V) in a high-speed steel sample is shown as an example in Figure 8.8 [25].
8.8 Three-Dimensional (3-D)-SIMS
Figure 8.7 Direct-imaging mode SIMS image
of a passivation layer on a niobium alloy [22]. An enrichment of boron at the interface is clearly observable. This is particularly important, because such boron distributions cannot be determined with for example,
electron probe X-ray microanalysis (EPXMA) due to the low sensitivity of this element. Measurement time: 10 s, image diameter: 150 μm, primary ions: O2+, primary ion energy: 5.5 keV.
Figure 8.8 3-D-SIMS measurement of the V distribution in high-speed steel. Dimension:
150 μm × 150 μm × 10 μm. (a) Block view; (b) Isosurface view. Primary ions: O2+, primary ion energy: 5.5 keV, primary ion intensity: 2 μA.
155
156
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
8.9 Applications 8.9.1 Implantation Profiles
Ion implantation is the common method for doping semiconductors. For this process, the concentrations of the dopants (mostly B and P) are very low, and a dynamic range of more than five orders of magnitude is often necessary. The measurement of P is more difficult than that of B, mainly due to the mass interference of the 30SiH and 31P ions, high mass resolution of Δm/m = 5000 or energy filtering with an energy offset of 300 V is necessary. In order to achieve high concentrations in deeper regions while keeping the damage low, channeling is exploited for ion implantation. In this case, the Si wafer is bombarded in a high-index crystal direction. However, due to the production of small amounts of defects, the channeling effect can be utilized only to an upper limit of implantation concentrations. The depth profile in Figure 8.9 shows the typical channeling implantation profiles generated only with low doses. It can be observed that, with an increasing ion dose, the channeling tail in the profiles becomes less pronounced. This can be explained by the reduced channeling due to an increasing amorphization of surface-adjacent zones with increasing ion dose. With correct calibration using homogeneous wafers, these channeling experiments can also be used to determine the formation of defects and amorphization in semiconductor production.
Figure 8.9 Depth profiles of P implantation in Si. Primary ions: Cs+, primary ion energy:
14.5 keV.
8.9 Applications
8.9.2 Layer Analysis
The benefits provided by surface-coating techniques have the common thread to increase productivity and to reduce costs [26]. CrN provides excellent material properties to meet these requirements, including a high hardness, superior oxidation stability, a high corrosion resistance, a low friction coefficient, and good adhesion characteristics. Due to these outstanding properties, CrN is often applied for the machining of iron-based materials and wear parts. Although CrN has extraordinary material properties, further improvement is desirable. For this reason, it is very important to have not only exact but also complete information concerning the presence of any desired or undesired trace elements within CrN-layers. Figure 8.10 shows a depth profile of a CrN–layer, where an enrichment of Ti and Al at the interface to the substrate is visible and the signal of Si is decreasing towards the surface in the layer. Although, at the first glance, the tail of the Si profile into the CrN-layer suggests that Si diffuses into that layer, this cannot be concluded unequivocally from this depth profile alone. Figure 8.11 shows the 3-DSIMS distributions of the elements Si, Al, Ti, and Cr in the CrN-layer shown previously in Figure 8.10. In this case, the Cr distribution is shown upside-down in order to make the interface visible. It can be seen that the interface is not smooth, and this is the reason for the slowly decreasing Cr-signal (into the Si substrate) in Figure 8.10. Furthermore, it can be seen in the 3-D distributions that Ti and Al are present in the form of particles, which explains their comparatively high signal in the CrN-layer in the profile in Figure 8.10. From the depth profile in Figure 8.10 alone, it cannot be deduced whether the profile shape is a result of diffusion
Figure 8.10 Depth profile of a CrN-layer deposited on Si. Primary ions: O2+, primary ion
energy: 5.5 keV, primary ion intensity 2 μA, measuring time: 1 h.
157
158
8 Dynamic Secondary Ion Mass Spectrometry (SIMS)
Figure 8.11 3-D depth profiles of Si, Al, Ti, and Cr in a CrN-layer on Si a substrate. Data are
from the same measurement as those in Figure 8.10. The CrN-layer thickness is 2.3 μm. Diameter of the analyzed area 150 μm. The Cr distribution is shown upside-down.
Figure 8.12 3-D element distributions of powder-metallurgically produced steel. Lateral size of the cube: 150 μm × 150 μm. Primary ions and intensity: (a) 0.1 μA Cs+; (b) 2 μA O2+, depth: (a) 1 μm, (b) 11 μm.
of elements, or due to geometric effects. This example demonstrates the power of 3-D-SIMS to significantly enhance the information content of depth profiles. 8.9.3 3-D Trace Element Distribution
Steel alloyed with ∼1% Al shows a higher resistance against oxidation at high temperatures. Hot isostatically pressed steels (HIP) exhibit a better macro- and microscopic homogeneity of the distribution of all elements. A combination of these advantages should result in a higher performance of high-speed steels. Before conducting the HIP process, the steel particles are covered with an Al2O3layer [27], which is shown still to exist in the compacted material by 3-D-SIMS measurements (Figure 8.12). The distribution of Ca in steel, as an example of a trace element with a concentration in the ppm range, is shown in Figure 8.12b.
References
References 1 Herzog, R.F.K. and Viehböck, F.P. (1949) Phys. Rev., 76, 855L. 2 Casting, R. and Slodzian, G. (1962) J. Microsc., 1, 395. 3 Herzog, R.F.K. and Liebl, H. (1963) J. Phys., 34, 2893. 4 Benninghoven, A. (1970) Z. Phys., 239, 403. 5 Zalar, A. and Hofmann, S. (1993) Appl. Surf. Sci., 68 (3), 361–367. 6 Hofmann, S., Zalar, A., Cirlin, E.H., Vajo, J.J., Mathieu, H.J., and Panjan, P. (1993) Surf. Interface Anal., 20 (8), 621–626. 7 Mattauch, J. and Herzog, R. (1934) Z. Phys., 89, 786. 8 Hutter, H. and Grassrbauer, M. (1992) Mikrochim. Acta, 107, 137. 9 Schroeer, J.M., Gnaser, H., and Oechsner, H. (1994) SIMS IX-Proceedings of the Yokohama Conference, John Wiley & Sons, p. 394. 10 Schroeer, J.M., Rodin, T., and Bradley, R. (1973) Surf. Sci., 34, 511. 11 Sroubek, Z. (1983) Phys. Rev., B25, 604b. 12 Anderson, C.A. and Hinthorne, J.R. (1973) Anal. Chem., 45, 1421. 13 Hagenhoff, B., Van Leyen, D., Niehuis, E., and Benninghoven, A. (1989) J. Vac. Sci. Technol. A: Vac. Surf. Films, 7 (5), 3056–3064. 14 Wucher, A. and Winograd, N. (2010) Anal. Bioanal. Chem., 396 (1), 105–114. 15 Lee, J.L.S., Ninomiya, S., Matsuo, J., Gilmore, I.S., Seah, M.P., and Shard, A.G. (2010) Anal. Chem., 82, 98–105. 16 Shard, A.G., Green, F.M., and Gilmore, I.S. (2008) Appl. Surf. Sci., 255 (4), 962–965.
17 Zinner, E., Amari, S., Guinness, R., Jennings, C., Mertz, A.F., Nguyen, A.N., Gallino, R., Hoppe, P., Lugaro, M., Nittler, L., and Lewis, R. (2007) Geochim. Cosmochim. Acta, 71 (19), 4786–4813. 18 Wagner, M. (2009) Annu. Rev. Microbiol., 63, 411–429. 19 Brunner, C.H., Hutter, H., Piplits, K., Gritsch, M., Pöckl, G., and Grasserbauer, M. (1998) Fresenius J. Anal. Chem., 361, 667–671. 20 Benninghoven, A., Ruedenauer, F.G., and Werner, H.W. (1987) Secondary Ion Mass Spectrometry, John Wiley & Sons, Inc., New York, p. 351. 21 Stubbings, T., Wolkenstein, M., and Hutter, H. (1999) J. Trace Microprobe Tech., 17 (1), 1–16. 22 Gritsch, M., Piplits, K., Barbist, R., and Wilhartitz, P. (2000) Mikrochim. Acta, 133, 89–93. 23 Grasserbauer, M. and Hutter, H. (1994) SIMS IX - Proceedings of the Yokohama Conference, John Wiley & Sons, Ltd, p. 545. 24 Ruedenauer, F.G. and Steiger, F.G. (1987) SIMS VI - Proceedings, John Wiley & Sons, Ltd, p. 361. 25 Hutter, H., Novikow, K., and Gammer, K. (2001) Appl. Surf. Sci., 179, 161–166. 26 Jarms, C., Stock, H.-R., and Mayr, P. (1998) Surf. Coat. Tech., 108–109 (1), 206–210. 27 Pollak, C., Kriszt, B., and Hutter, H. (2000) Mikrochim. Acta, 133, 261–266.
159
161
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS) Michael Kopnarski and Holger Jenett
9.1 Introduction
Since the early days of sputtering it has been known that, for most cases, the flux of particles ejected from a solid under ion bombardment consists predominantly of sputtered neutrals (SN)s [1]. Thus, in order to enhance sensitivity the pioneers of surface investigation by mass spectrometry applied the process of postionization [2]. Very shortly afterwards, however, due to the rather limited postionization efficiency when using electron beam methods, and also because of difficulties with parasitic ion and residual gas suppression, the direct detection of secondary ions provided an advantage over using the neutral component of the sputtered particle flux. As a consequence, sophisticated SIMS instruments were developed with, prior to SNMS, SIMS being introduced into the field of surfaceand thin-film analysis. In contrast, a serious limitation of SIMS – the so-called matrix effect – emerged whereby, for a certain concentration of an element X, the secondary ion yield could vary by orders of magnitude as a function of the surface contamination and matrix composition (see e.g., Ref. [3]). Indeed, in many cases – and especially with unknown or technical samples that had changing matrices – this would hamper quantification, or even make it impossible. Subsequently, during the early 1970s, H. Oechsner [4] developed a strategy to improve quantification by using the effective postionization of the neutral particle flux via the electron component of a low-pressure noble gas plasma. In addition to the high postionization efficiency, this special postionization arrangement provided two further advantages: (i) an almost perfect suppression of the secondary ions; and (ii) the possibility of a lowenergy primary bombardment which featured excellent depth resolution for sputter depth profiling. In the meantime, electron beam and laser SNMS had also been developed for postionization purposes, and had expanded into the field of applied surface- and thin-film analysis. The latter technique is described fully in Chapter 10 of this
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
162
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
book; however, the details of both electron impact techniques – namely plasma and electron beam SNMS – will be discussed in the following subsections. The main advantages of SNMS are: Compared to XPS and AES sputter depth profiling: The detection power of SNMS (or SIMS) is higher by several orders of magnitude. Under the conditions of sputter equilibrium, the sputtered particle flux represents the true stoichiometry and is not altered because of preferential sputtering effects, as is the case for the surface stoichiometry observed in sputter depth profile measurements when using electron spectroscopic methods [5]. Compared to glow discharge (GD)-OES (and -MS, if used for depth profiling): SNMS provides a somewhat better depth resolution (1 nm range); highfrequency (HF)-plasma SNMS hardly suffers from molecule formation in the plasma gas (Ar), as do the GD techniques, in which argides (molecular ions containing Ar) are formed because of the crucial role of the heavy particle interaction (e.g., Penning ionization) that these techniques employ instead of electron impact ionization. This not only requires higher working pressures for the discharge, but also gives rise to the backscattering of sputtered particles to the sputter bombarded area. This, in turn, limits the dynamic range obtainable in sputter depth profiling with GD-OES or GD-MS. Compared to SIMS: Sputtering produces predominantly neutral atoms; for most of the elements in the Periodic Table the typical secondary ion yield is between 10−2 and 10−5. Even more crucial, the probability of a sputtered particle to be emitted as an ion can vary widely with surface composition. This is not true for the SN. For example, the variation of secondary ion formation probability from 10−2 to 10−5 produces a <1% change in the SN yield. As a result, matrix effects as known from SIMS are avoided with SNMS, although minor influences of the matrix may still be present.
9.2 General Principles of SNMS
The basic components of the SNMS method are (Figure 9.1):
• • • •
Sputtering of the sample surface by energetic particle bombardment (typically Ar+ ions, energy range from 100 eV up to several keV). Transfer of the sputter-released neutrals into a charged state via electron impact ionization. Suppression of residual gas and secondary ions. Mass analysis (usually by quadrupole mass filters) and single ion detection.
Of prime importance here are the postionization step and the suppression of residual gas and secondary ions; hence, these will be described from a general point of view in the following subsections.
9.2 General Principles of SNMS
Figure 9.1 Scheme of the basic components of secondary neutral mass spectrometry (SNMS).
9.2.1 Postionization via Electron Impact
The postionization process of the sputter-released neutrals is the “heart” of any SNMS experiment. Let us consider a beam of sputtered particles with the density n x0 of the neutrals X0 which interact with electrons of density ne. The density of ions dn x+ (for simplicity, only singly charged particles shall be considered) built during a time interval dt depends on the particle densities, the electron speed ve, and on the electron impact ionization process X0+e−→X++2e−, as described by the corresponding cross-section σ(ve): dn x+ = n x0 ⋅ n e ⋅ ve ⋅ σ (ve ) ⋅ dt
(9.1)
With the electron current density je = ne· ve and the ionization collision frequency fc = je·σ(ve) the postionization probability α x0 after an interaction time Δt1) is calculated according to Equation 9.2:
α x0 =
Δn x+ = f c ⋅ Δt n x0
(9.2)
For non-monoenergetic electrons, this leads to the differential electron current density dje dE and the collision frequency can be calculated from: 1) It is assumed that Δt is small enough so that a x0 << 1 and, hence, n x0 does not change during Δt.
163
164
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
Figure 9.2 Ionization function for electron impact ionization of Cu and differential electron current densities for typical e-gas and e-beam arrangements. The resulting collision frequencies fc depend on the overlap of the respective density distribution and the ionization function.
∞
fc =
dje (E ) ⋅ σ (E )dE dE 0
∫
(9.3)
Figure 9.2 shows, as a function of the electron energy, the cross-sections for the ionization of Cu atoms by electron impact (dotted line, left-hand scale [6],) together with the typical electron current density distributions available in representative electron-beam (e-beam) and electron-gas (e-gas) arrangements (straight lines, right-hand scale). The current density in the e-beam is limited by space charge effects. Even if magnetic guidance is used, the postionization probability cannot be increased because of space charge trapping of the postionized sputtered particles [7]. The frequency of ionizing collisions fc is calculated according to Equation 9.3. Due to the higher current densities in the space charge-compensated plasma, the ionization frequency in the dense electron gas is, in comparison to the e-beam, more than an order of magnitude higher (Figure 9.2). 9.2.2 Suppression of Residual Gas and Secondary Ions
Sputter ion bombardment creates not only neutrals but also positively and negatively charged secondary ions. Hence, every SNMS arrangement must be capable of discriminating between both types of secondary ions and the neutrals’ signal. Positive ions – which have the same charge sign as the postionized particles – are usually suppressed prior to postionization via a retarding electric field directly above the sample surface. This field accelerates the positive primary ions towards
9.2 General Principles of SNMS
the sample, and retains the positive secondary ions at the surface. In the case of electron beam SNMS, suppression of the positive secondary ions is accomplished by a retarding or deflection field between the sample and ionizer, although a combination of both may also be used. Negative secondary ions can easily be removed from the flux of postionized (positive) particles by an electrical sector field following postionization. Besides the secondary ions, residual gas particles ionized in the postionization volume may also contribute to the background signals. Energy dispersive stages in the ion optical arrangement are used to prevent such influences, this being possible because the sputtered atoms gain an average energy of some eV from the sputtering process. Although, the sputter energy distribution is somehow broadened by the electrical potential distribution across the ionization volume, and partially overlaps with the energy distribution of the ionized residual gas particles, it is possible to separate this background from the signal with a discrimination ratio better than 104 (see Figure 9.3).
Figure 9.3 SNMS spectra of a Cu sample containing trace amounts of P and Fe. (a)
Bombardment voltage UDBM = 1 kV, primary ion current Ip = 500 μA; (b) As panel (a), but with suppression of residual gas signals; (c) As panel (b), but UDBM = 0 V (no sputtering).
165
166
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
Figure 9.4 SNMS-intensities (relative to the matrix signal) of various residual gas elements observed during sputtering of a pure Si surface. For high bombarding voltages (UDBM ≥ 1 kV; i.e., high sputter removal rates), the contribution of the 12C and the 16O intensity amounts to only some
10 ppm of the matrix signal, in contrast to the corresponding signal fraction for low sputter removal rates (UDBM ≈ 100 V). The mass number 14 signal is superimposed by a contribution originating from sputtered and doubly postionized Si atoms restricting the detection limit for N to the per-million range.
The efficiency of the residual gas suppression is crucial with respect to the detection limits for species occurring in the ambient gas, such as H, C, N, or O. It should be noted that there is no possibility of inhibiting the contributions from residual gas elements which, in a first step, are adsorbed at the sample surface and subsequently sputtered together with the target material. Therefore, the signal-to-background ratio for these species will vary depending on the actual sputter rate, as shown in Figure 9.4.
9.3 Instrumentation and Methods 9.3.1 Electron Beam SNMS
The basic principle of e-beam SNMS, as introduced by Lipinsky et al. in 1985 [8], is simple (Figure 9.5). As in SIMS, the sample is sputtered with a focused keV ion beam, while SN postionization is accomplished by an e-beam accelerated between a filament and an anode. The applied electron energy, Ee ≈ 30–70 eV, is higher than the range of first ionization potentials (IP) of the elements (4–24 eV). Positive secondary ions and the suppression of residual gas ions is achieved with electrostatic lenses before SN postionization and energy filtering, respectively. The electron current in the ionization region, which is a few millimeters in length, is in
9.3 Instrumentation and Methods
Figure 9.5 Schematic diagram of an electron beam SNMS set-up. After Ref. [9].
the lower mA range. The velocities of the sputtered neutrals lie in the range of some 105 cm s−1, and, hence, their dwell time Δt in the ionizing e-beam amounts to some microseconds. According to Equation 9.2, Δt must be multiplied by the corresponding collision frequency fc, around 0.1 kHz (see Figure 9.2) to determine the ionization probability α x0 (e-beam) as being in the range of about 10−4. Instead of the simple lens, as depicted in Figure 9.5, an electrostatic energy analyzer can serve for the separation of postionized SN (1–10 eV) from residual gas signals (<10 meV). A mass analysis is performed by means of a quadrupole mass analyzer, and pulse counting by means of a dynode multiplier. A quadrupole enables swift switching between mass settings, thus enabling continuous data acquisition for many elements, even at high sputter rates within thin layers. In contrast to the other SNMS techniques, e-beam SNMS modules can easily be added to many of the UHV-based surface-analysis instruments currently in use. 9.3.2 Plasma SNMS
A schematic diagram of the basic set-up of HF-plasma (e-gas) SNMS [10] is shown in Figure 9.6. Here, an inductively coupled, low-pressure, HF noble gas plasma serves for sputtering using the ion component, and for electron impact postionization using the Maxwellian gas of plasma electrons. The rectangular plasma vessel is completely mounted inside the UHV chamber, while a constant Ar pressure of 0.2 ± 0.1 Pa is maintained by means of a Piezo valve located between the highpressure supply and the plasma chamber. The HF power of approximately 150 W is supplied from an HF generator by a single-turn coil around the plasma chamber. By using a rectangular pair of Helmholtz coils for the right-handed wave, the skin effect is avoided by the superposition of a relatively weak static magnetic field in the mT range, which allows the establishment of an electron cyclotron wave resonance [11]. In the so-called direct bombardment mode (DBM) [12], a negative target potential UDBM is applied with respect to the reference electrode, which is usually
167
168
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
Figure 9.6 Schematic diagram of the HF-plasma SNMS set-up (INA-X). The optional primary sputter ion source can be used for the application of the so-called SBM mode of SNMS or, when the plasma is switched off, for SIMS measurements.
grounded. The difference between the plasma potential ΦPl, some 10 eV positive with respect to the reference electrode, and the sample potential determined by UDBM (see Figure 9.6) accelerates the Ar-ions arriving at the plasma boundary towards the sample, and allows for sputtering with ion current densities of about 1 mA cm−2, even at low bombarding energies of about 100 eV. At the same time, positive secondary ions (SI+) are completely retained at the sample surface, and cannot be mixed-up with the postionized neutrals (SN+). If, depending on the plasma density ne and the electron temperature Te, the distance between the sample surface and the sample shielding aperture is fitted appropriately to UDBM, then sputter erosion will occur with almost perfect lateral homogeneity – that is, with nanometer depth resolution on a spot which is 2–5 mm in diameter [13]. For comparison, this is similar to removing the sand grains layer by layer from a beach volleyball field! The plasma also delivers the electron gas at electron temperatures in the range of Te = 5 × 104–105 K and electron densities up to ne = 1011 cm−3 [14]. With a traveling distance through the plasma of some centimeters, the dwell time Δt of the sputtered neutrals in the ionizing e-gas is approximately 10 μs. Again, the collision frequency fc can be estimated from Figure 9.2. With fc ≥ 1 kHz according to Equation 9.2 for plasma SNMS, the ionization probability α x0 lies in the range of some 10−2. According to Ref. [10], the commercially available INA-X type spectrometer (IFOS, Kaiserslautern and SPECS, Berlin) [15] collects the postionized particles
9.3 Instrumentation and Methods
under an oblique take-off angle, and transfers them through a spherical 45° energy filter towards the quadrupole mass spectrometer. Space-charge effects at the entrance electrode of the ion optical system can be largely reduced by setting its voltage around ΦPl. Electrons from the e-gas can leak into the entrance region and partly compensate the positive space charge. The energy filter is used to eliminate negative secondary ions (SI−) and to discriminate between the residual gas and sputtered particles (see Section 9.2.2). For normal bombardment, the sampling of postionized neutrals under an oblique angle is advantageous for the following reason [16]. The composition of the neutral flux sputtered under 30–40° does not depend greatly on the bombarding energy (in contrast to the particle flux sputtered in the normal direction), and therefore the relative sensitivity factors (see Section 9.4) are kept almost constant, particularly when the bombarding energy is reduced [15]. In the so-called high-frequency mode (HFM) of SNMS, electrically insulating materials can be analyzed in plasma SNMS by applying a square-wave HF voltage in the 100 kHz range to the sample (Figure 9.7) [17]. Dielectric charge transfer at the start of a period shifts the surface potential to the amplitude UHFM applied. Consequently, the Ar+ ions are attracted from the plasma and sputter the surface until the end of Δt−. The potential increase ΔUs = 1–100 V caused by their charge is then converted to a positive absolute ΔUs which is reduced to less than 1 V within <0.1 μs by the plasma electron current. In the majority of cases, quadrupoles are used for mass filtering. Recently, a novel plasma SNMS instrument was developed which is equipped with a time-offlight (TOF) mass spectrometer [18]. Similarly, detection limits in the nmol mol−1 range have been achieved, using a dedicated plasma SNMS instrument attached to a double-focusing magnetic sector mass spectrometer [19].
Figure 9.7 The principle of high-frequency mode (HFM) SNMS. Variation of potentials
(versus reference electrode of the plasma) with time t: --- applied UHFM, — surface potential US, UPl,W plasma potential.
169
170
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
9.4 Spectral Information and Quantification
The spectral information obtained from e-beam and plasma SNMS, both of which are dynamic sputtering methods, is similar to that obtained with dynamic SIMS. However, because the sputtered molecule flux is often comparatively small, and the working pressure of the plasma is sufficiently low that heavy particle interaction in the gas phase (which possibly could create molecular ions) is avoided, the emphasis lies on elemental analysis based on mostly atomic spectra. Only in cases of strong binding forces and large mass differences are binary molecules formed from the main constituents to a major degree and must be considered for quantification [20]. Molecular signals also might add restricted speciation information [17, 21–23]. Multiple SN ionization occurs rarely, usually because of insufficient electron energy and the low probability of a second electron impact. As a consequence, e-beam and plasma SNMS spectra are characterized by a small amount of interferences, and are relatively easily interpreted. Figure 9.8 shows a typical HF-plasma SN+ mass spectrum as obtained from a TiAl alloy. In this case, the quadrupole was operated typically with the absolute mass resolution ΔM ≈ 1 throughout the spectrum. As mentioned above, molecule intensities are in the percentage range of the parent peaks. Doubly ionized species (Al2+, Ti2+) appear with intensities approximately three orders of magnitude lower than the corresponding singly post-ionized particles (of only the main constituents, for reasons of intensity), and adsorbed or implanted plasma gas contamination gives rise to a few other low-intensity peaks.
Figure 9.8 HF-plasma SNMS survey spectrum of a powder-metallurgically produced TiAl alloy sample. Signals in brackets are derived from adsorbed and re-sputtered plasma gas contamination.
9.4 Spectral Information and Quantification
Following Oechsner [24], the measured SN+ intensity Ix of a sputtered species X may be factorized by: I X = IP ⋅ YX ⋅ (1 − α x± ) ⋅ α x0 ⋅ GX ⋅ TX
(9.4)
where IP is the primary ion particle current (ions per second); YX is the partial sputtering yield of X, α x± is the probability that X is emitted as a positive or negative secondary ion, and TX is the combined spectrometer transmission and detector response probability for the postionized X particles (for quadrupole instruments, this is typically 10−4). The accepted fraction of the sputtered neutral flux GX is in the range of 10−3 or 10−2 for the plasma and the e-beam arrangement, respectively. Using the α x0 values calculated in Section 9.2 and neglecting α x±, it is possible to estimate the absolute detection factor (useful yield) SX = α x0·GX·TX for both arrangements: Sxplasma ≈ 10 −9 and Sxe−beam ≈ 10 −10. Equation 9.4 is valid for any species X. Neglecting molecular signals, however, X stands for elements and atoms only. Then, after attaining sputter equilibrium, the total sputter yield Ytot is connected to the atomic concentration cx and the partial sputter yield via YX = Ytot·cx and, hence: i =n
Ytot =
i =n
∑Y = I ∑ S 1
Ii
(9.5)
i
P i =1
i =1
i
where i = 1 ... n sample elements. Hence, relating the Si to a reference value Sref, the concentration cx of an element X can be determined from the measured SNMS intensities: −1
⎡ i = n ⎛ Sref ⎞ ⎤ (9.6) ⋅ ⎢ ⎜ Ii ⎟⎥ ⎣ i =1 ⎝ Si ⎠ ⎦ For quantification according to Equation 9.6, the relative sensitivity factors S(ref)X = (SX/Sref) can be determined from standard samples with known concentration ratios (cref/cx)st via:
∑
c x = IX ⋅
Sref SX
S(ref )X =
I X ⎛ c ref ⎞ ⋅⎜ ⎟ Iref ⎝ c x ⎠ st
(9.7)
The different postionization probabilities, α M0 (for M = Ti and Al on one hand and Cr and Nb on the other, both pairs being present at the same concentrations), are the main reason that the respective signal intensities in Figure 9.8 are similar, but not equal. Figure 9.9 shows that experimental relative sensitivity factors vary approximately with theoretical α x0. The measured relative sensitivity factors [25] and the calculated postionization probabilities [26], with few exceptions (Zn and Mo), agree within a factor of 3. Both, the inaccuracy of the theory (ionization crosssections based on the Lotz formula [27] have been used) and experimental influences (e.g., possible variation of the ratio of Gx and Tx values, or the disregard of molecular contributions) must be taken into account. Bearing this in mind, the agreement is satisfactory. Except for C, O, N, and F, the S(Fe),X values lie within two orders of magnitude. It is this narrow range of S(ref)x variation, compared to SIMS,
171
172
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
Figure 9.9 Experimental, Fe-related sensitivity factors S(Fe)X determined for 29 elements X from plasma SNMS measurements under standard conditions [25]. Eighteen standard materials, including high-alloy steels, brass, and Cu-, Zn-, Pb-, Ni-, and Al-based alloys, have been used. In
addition, the S(Fe)X values for O and F are shown in brackets (W. Bock, personal communication). The measured sensitivity factors are plotted versus the relative post-ionization probabilities α(0Fe ),x calculated for the measuring conditions [26].
the virtual absence of any strong matrix effects, and the ease of quantification which has made SNMS famous.
9.5 Element Depth Profiling
Figure 9.10 provides an example of the depth resolution routinely achieved with both e-beam and HF-plasma SNMS on suitable samples. If Equation 9.5 is applicable, the actual sputter depth z(t) associated with the sputter time t can be calculated according to: t
z(t ) =
IP Ytot (t ′) ⋅ ⋅ dt ′ As 0 n(t ′)
∫
(9.8a)
where AS is the sputter-bombarded area. As a first approximation, the timedependent total particle density n can be obtained from the densities ni of the respective bulk materials by [28]: i =n
n(t ) =
∑c ⋅n i
i =1
i
(9.8b)
9.5 Element Depth Profiling
Figure 9.10 HF-plasma SNMS concentration depth profile of a Ti/TiSi2-layer structure on SiC.
The thin C interlayer becomes clearly visible; the according sputter time is also shown. Note that the sputter rate is not constant. Data taken from Ref. [15]
Figure 9.11 HF-plasma SNMS sputter time profile (HFM) of a multilayer system consisting of five double layers, 100 nm SiO2 + 100 nm Si3N4, each, on a glass substrate. Illustration courtesy of V.-D. Hodoroaba, BAM Berlin, and Schott Glas, Mainz, Germany.
For plasma SNMS, an approach using the measured weight loss before and after depth profiling is also feasible [29, 30]. Figure 9.11 shows that, during the entire sputter time, a good depth resolution is maintained, although the micrometer-thick layer system consists of electrically insulating materials.
173
174
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS)
9.6 Applications
Figure 9.12 exemplifies the use of HF-plasma SNMS in materials research [31]. Here, the intermetallic compound γ-TiAl, to which 2 mol.% Cr has been added in order to increase ductility, has been heated to 1070 K for less than 4 s only, cooled down again (Figure 9.12a), and then kept at this high temperature for 8 h (Figure 9.12b). Initially, an oxide layer of about 150 nm thickness was formed, with the Al2O3-enriched region topmost. In Figure 9.12b, it can be seen that following this initial oxide layer formation, nitrogen began to penetrate to the interface between the oxide and the alloy, thus oxidizing the latter by the formation of nitrides. Furthermore, N-enrichment at the interface and the results of the tracer isotope experiments described in Ref. [31] suggested that nitride had been continuously converted to the oxide between the oxide layer and the nitride-enriched interface layer. The data in Figure 9.12b also showed that Ti had diffused outwards, forming a topmost TiO2-rich layer. The applications of e-beam and HF-plasma SNMS have been reported in the following areas: aerosol particles [32]; X-ray mirrors [33, 34]; ceramics and hard
Figure 9.12 HF-plasma SNMS sputter depth profiles (DBM) of Ti-48AI-2Cr (atom%) samples
having been oxidized at 800 °C in air for: (a) <4 s; and (b) 8 h.
References
coatings [35–39]; glasses [40]; interface reactions [41]; ion implantations [42]; molecular beam epitaxy (MBE) layers [43]; multilayer systems [44]; ohmic contacts [45]; organic additives [46]; perovskite-type and superconducting layers [47]; steel [48, 49]; surface deposition [50]; subsurface diffusion [51]; sensors [52–54]; soil [55]; and thermal barrier coatings [56]. Recently, high-resolution depth profiling with plasma SNMS was also used for the investigation of various electrodeposited thin-film structures [57–62], Sb diffusion in amorphous silicon [63], failure mechanisms at Ta and TaO diffusion barriers [64], doped perovskites [62, 65], as well as for the evaluation of solid-state phases in Zn–Ni coatings and the quantitative detection of temperature-induced chemical state reactions for ohmic contacts in microelectronic devices based on silicon carbide [66].
References 1 Honig, R. (1958) J. Appl. Phys., 29, 549. 2 Smith, A.J., Cambey, L.A., and Marshall, D.J. (1963) J. Appl. Phys., 34, 2489. 3 Benninghoven, A. and Mueller, A. (1972) Phys. Lett., 40A, 169. 4 Oechsner, H. and Gerhard, W. (1972) Phys. Lett., 40A, 211. 5 Albers, T.H., Neumann, M., Lipinsky, D., Wiedmann, L., and Benninghoven, A. (1994) Surf. Interf. Anal., 22, 9. 6 Kopnarski, M., Oechsner, H., and Halden, T. (1991) Proceedings, 17th International Conference of Physical Electronic and Atomic Collisions (XVII ICPEAC) (eds I.E. McCarthy, W.R. MacGilliway, and M.C. Standage), Griffith University, Qld, p. 176. 7 Jede, R., Ganschow, O., and Kaiser, U. (1992) Practical Surface Analysis, 2nd edn, vol. 2 (eds D. Briggs and M.P. Seah), John Wiley & Sons, Ltd, Chichester. 8 Lipinsky, D., Jede, R., Ganschow, O., and Benninghoven, A. (1985) J. Vac. Sci. Technol. A, 3, 2007–2017. 9 Colligon, J.S., Kheyrandish, H., and Walls, J.M. (1991) Thin Solid Films, 200, 293–300. 10 Oechsner, H. and Müller, M. (1999) J. Vac. Sci. Technol. A, 17, 3401–3405. 11 Oechsner, H. (2009) Vacuum, 83, 727–731. 12 Müller, K.H. and Oechsner, H. (1983) Mikrochim. Acta, 10, 51–60. 13 Stumpe, E., Oechsner, H., and Schoof, H. (1979) Appl. Phys., 20, 55.
14 Wucher, A. (1988) J. Vac. Sci. Technol. A, 6, 2293–2298. 15 Jorzick, J., Lösch, J., Kopnarski, M., and Oechsner, H. (2004) Appl. Phys. A, 78, 655–658. 16 Wucher, A. (1988) J. Vac. Sci. Technol. A, 6, 2316–2318. 17 Oechsner, H. (1995) Int. J. Mass. Spectrom. Ion Processes, 143, 271–282. 18 Kopnarski, M., Lösch, J., and Simeonov, L. (2009) Anal. Biol. Chem., 393, 1913–1916. 19 Gnaser, H., Schneider, H.-J., and Oechsner, H. (1998) Nucl. Instrum. Methods Phys. Res. B, 136–138, 1023–1027. 20 Oechsner, H. (1985) The Physics of Ionized Gases (SPIG XII) (eds M.M. Popovic´ and P. Krstic´), World Scientific Publishing, Singapore, pp. 571–598. 21 Kopnarski, M., Wucher, A., and Oechsner, H. (1987) Secondary Ion Mass Spectrometry (SIMS VI) (eds A.M. Huber and M.W. Werner), Springer-Verlag, Berlin, pp. 849–852. 22 Goschnick, J., Sommer, M., and Ache, H.J. (1995) Fresenius J. Anal. Chem., 353, 548–551. 23 Sommer, M. and Goschnick, J. (2005) Appl. Surf. Sci., 252, 252–256. 24 Oechsner, H. (1984) Thin Film and Depth Profile Analysis (ed. H. Oechsner), Springer, New York. 25 Wucher, A., Novak, F., and Reuter, W. (1988) J. Vac. Sci. Technol. A, 6, 2265.
175
176
9 Electron-Impact (EI) Secondary Neutral Mass Spectrometry (SNMS) 26 Wucher, A. (1988) J. Vac. Sci. Technol. A, 6, 2287. 27 Lotz, W. (1970) Z. Phys., 232, 101. 28 Wucher, A. and Oechsner, H. (1989) Fresenius J. Anal. Chem., 333, 470. 29 Bock, W., Kopnarski, M., and Oechsner, H. (1998) Vak. Forsch. Prax., 1, 21–28. 30 Bentz, J.W.G., Goschnik, J., Schuricht, J., and Ache, H.J. (1995) Fresenius J. Anal. Chem., 353, 559–564. 31 Jenett, H., Sunderkötter, J.D., and Stroosnijder, M.F. (1997) Fresenius J. Anal. Chem., 358, 225–229. 32 Goschnick, J., Natzeck, C., and Sommer, M. (1999) Appl. Surf. Sci., 144-145, 201–207. 33 Hopfe, S., Kallis, N., Mai, H., Pompe, W., Scholz, R., Völlmar, S., Wehner, B., and Weissbrot, P. (1993) Fresenius J. Anal. Chem., 346, 14–22. 34 Martin, R.W., Kim, T., Burns, D., Watson, I.M., Dawson, M.D., Krauss, T.F., Marsh, J.H., De la Rue, R.M., Romani, S., and Kheyrandish, H. (1999) Phys. Status Solidi A, 176, 67–71. 35 Dreer, S. (1999) Fresenius J. Anal. Chem., 365, 85–95. 36 Galesic, I. and Kolbesen, B.O. (1999) Fresenius J. Anal. Chem., 365, 199–202. 37 Tharigen, T., Lippöld, G., Riede, V., Lorenz, M., Koivusaari, K.J., Lorenz, D., Mosch, S., Grau, P., Hesse, R., Streubel, P., and Szargan, R. (1999) Thin Solid Films, 348, 103–113. 38 Witthaut, M., Cremer, R., von Richthofen, A., and Neuschütz, D. (1998) Fresenius J. Anal. Chem., 361, 639–641. 39 Boerner, H., Jenett, H., and Hodo-roaba, V.-D. (1998) Fresenius J. Anal. Chem., 361, 590–591. 40 Hueber, B., Frischat, G.H., Maldener, A., Dersch, O., and Rauch, F. (1999) Non-Cryst. Solids, 256-257, 130–134. 41 Getto, R., Freytag, J., Kopnarski, M., and Oechsner, H. (1999) Mater. Sci. Eng. B, 61–62, 270–274. 42 Kruijer, S., Nikolov, O., Keune, W., Reuther, H., Weber, S., and Scherrer, S. (1998) J. Appl. Phys., 84, 6570–6581. 43 Seryogin, G.A., Nikishin, S.A., Temkin, H., Schlaf, R., Sharp, L.I., Wen, Y.C., Parkinson, B., Elyukhin, V.A., Kudriavtsev, Yu.A., Mintairov, A.M.,
44
45
46
47
48
49 50
51
52
53
54
55 56
57
58
Faleev, N.N., and Baidakova, M.V. (1998) J. Vac. Sci. Technol. B, 16, 1456–1458. Simmonds, M.C., Savan, A., van Swygenhoven, H., and Pfluger, E. (1999) Thin Solid Films, 354, 59–65. Vogt, A., Simon, A., Hartnagel, H.L., Schikora, J., Buschmann, V., Rodewald, M., Fuess, H., Fascko, S., Koerdt, C., and Kurz, H. (1998) J. Appl. Phys., 83, 7715–7719. Goschnick, J., Maeder, U., and Sommer, M. (1998) Fresenius J. Anal. Chem., 361, 707–709. Rebane, J.A., Gorbenko, O.Yu., Suslov, S.G., Yakovlev, N.V., Korsakov, I.E., Amelichev, V.A., and Tretyakov, Yu.D. (1997) Thin Solid Films, 302, 140–146. Degueldre, C., Buckley, D., Dran, J.C., and Schenker, E. (1998) J. Nucl. Mater., 252, 22–27. Sommer, D. and Essing, A. (1999) Rep. Eur. Comm., EUR 18864, 1–126. Dang, T.A., Frisk, T.A., Grossman, M.W., and Peters, C.H. (1999) J. Electrochem. Soc., 146, 3896–3902. Baumann, I., Cusso, F., Herreros, B., Holzbrecher, H., Paulus, H., Schaefer, K., and Sohler, W. (1999) Appl. Phys. A, 68, 321–324. Miloshova, M., Bychkov, E., Tsegel-nik, V., Strykanov, V., Klewe-Nebenius, H., Bruns, M., Hoffmann, W., Papet, P., Sarradin, J., Pradel, A., and Ribes, M. (1999) Sensors Actuators, B7, 171–178. Vasiliev, R.B., Rumyantseva, M.N., Podguzova, S.E., Ryzhikov, A.S., Ryabova, L.I., and Gaskov, A.M. (1999) Mater. Sci. Eng. B, 57, 241–246. Goschnick, J., Natzeck, C., Sommer, M., and Zudock, F. (1998) Thin Solid Films, 332, 215–219. Goschnick, J. and Sommer, M. (1998) Fresenius J. Anal. Chem., 361, 704–707. Nickel, H., Quadakkers, W.J., and Singheiser, L. (2002) Anal. Bioanal. Chem., 374, 581–587. Csik, A., Vad, K., Langer, G.A., Katona, G.L., Tóth-Kádár, E., and Péter, L. (2010) Vacuum, 84, 141–143. Bartók, A., Csik, A., Vad, K., Molnár, G., and Tóth-Kádár, E. (2009) J. Electrochem. Soc., 156, D253–D260.
References 59 Csik, A., Vad, K., Tóth-Kádár, E., and Péter, L. (2009) Electrochem. Commun., 11, 1289–1291. 60 Katona, G.L., Berényi, Z., Péter, L., and Vad, K. (2008) Vacuum, 82, 270–273. 61 Péter, L., Katona, G.L., Berényi, Z., Vad, K., Langer, G.A., Tóth-Kádár, E., Pádár, J., Pogány, L., and Bakonyi, I. (2007) Electrochim. Acta, 53, 837–845. 62 Kálmán, V., Csik, A., and Langer, A. (2009) Spectrosc. Eur., 21, 13–16. 63 Csik, A., Langer, G.A., Erdélyi, G., Beke, D.L., Erdelyi, Z., and Vad, K. (2008) Vacuum, 82, 257–260.
64 Lakatos, A., Csik, A., Langer, G.A., Erdelyi, G., Katona, G.L., Daroczi, L., Vad, K., Toth, J., and Beke, D.L. (2010) Vacuum, 84, 130–133. 65 Vad, K., Hakl, J., Csik, A., Mészáros, S., Kis-Varga, M., Langer, G.A., Pallinger, Á., and Bódog, M. (2010) Vacuum, 84, 144–146. 66 Oechsner, H., Getto, R., and Kopnarski, M. (2009) J. Appl. Physics, 105, 063523.
177
179
10 Laser Secondary Neutral Mass Spectrometry (Laser-SNMS) Heinrich F. Arlinghaus
10.1 Principles 10.1.1 Nonresonant Laser-SNMS
Surface analysis by nonresonant (NR) laser secondary neutral mass spectrometry (laser-SNMS) [1–5] has been used to improve ionization efficiency while retaining the advantages of probing the neutral component. In NR-laser-SNMS, an intense laser beam is used to ionize, nonselectively, all of the atoms and molecules within the volume intersected by the laser beam (Figure 10.1b). With sufficient laser power density it is possible to saturate the ionization process. For NR-laser-SNMS, adequate power densities are typically achieved in a small volume only at the focus of the laser beam, although this limits sensitivity and leads to problems with quantification due to differences between the effective ionization volumes of different elements. Whilst the nonresonant postionization technique provides rapid, multielement, and molecular survey measurements with significantly improved ionization efficiency over SIMS, it still suffers from isobaric interferences. 10.1.2 Resonant Laser-SNMS
Resonant (R-) laser-SNMS [6–11] has almost all the advantages of SIMS, e-SNMS, and NR-laser-SNMS, with the additional advantage of using a resonance laser ionization process which selectively and efficiently ionizes the desired elemental species over a relatively large volume (Figure 10.1c). For over 80% of the elements in the Periodic Table, R-laser-SNMS has almost unity ionization efficiency over a large volume, so the overall efficiency is greater than that of NR-laser-SNMS. Quantification is also simpler, because the unsaturated volume (where ionization is incomplete) is insignificant compared to the saturated volume. This leads to consistently high sensitivity and great accuracy in R-laser-SNMS measurements. In particular, the extremely high selectivity prevents almost all isobaric and Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
180
10 Laser Secondary Neutral Mass Spectrometry (Laser-SNMS)
Figure 10.1 Comparison of TOF-SIMS and laser-SNMS.
molecular interferences, thus improving the accuracy of the data and simplifying its interpretation without a need for very high-resolution mass spectrometry. Unfortunately, the main drawbacks that accompany this immunity to interference are that only one element can be analyzed at a time, and that low-repetition rate pulsed lasers impose a low duty cycle. The use of multiple lasers enables the simultaneous ionization of multiple elements, but this is practically limited to two to three elements. Advances in diode-pumped Nd:YAG lasers and optical parametric oscillators (OPOs) will enable increased repetition rates of over 1000 Hz from the currently available 30–100 Hz. 10.1.3 Experimental Set-Up
In laser-SNMS experiments, the sample is bombarded with a finely focused pulsed primary ion beam (see Section 7.2.1). During the time that the primary ion pulse is striking the target, the voltages on the extraction electrodes are set so that the electric fields retard any positive secondary ions. Following the primary ion pulse, the voltages are then switched so that ions formed by laser postionization of the neutral particles in the sputtered cloud have an energy which is significantly different from that of the secondary ions formed in the sputtering process. The ions are analyzed using time-of-flight mass spectrometry (TOF-MS) with an electrostatic energy analyzer or a reflectron that discriminates against the secondary ions. The use of TOF-MS enables the simultaneous measurement of all photoions in each laser shot. Laser-SNMS imaging is performed either by scanning the ion beam over the sample or by translating the sample under a fixed ion beam. If laser ionization mass spectrometry (LMIS) is used for sputtering, nanoanalysis down to 50 nm can be achieved. Charge compensation for insulator analysis is possible using pulsed low-energy electrons, which are introduced during the time interval between sput-
10.1 Principles
Figure 10.2 Schemes of laser postionization steps.
tering pulses. Laser-SNMS depth profiles are obtained by sputtering the sample with a continuous ion beam and taking data with a pulsed ion beam in the center of the crater (dual-beam mode; see Section 8.2.1). In laser-SNMS, as in other microprobe techniques (see Section 8.8), the combination of two-dimensional (2D) imaging with depth-profile capability enables visualization of the threedimensional (3-D) distribution of elements. However, the extremely high overall efficiency of R-laser SNMS (typically 3–8% useful yield for most elements) enables the measurement of low concentrations of analytes in much smaller volumes, thus enabling a higher lateral and depth resolution in these 3-D visualizations. This capability could contribute greatly to the study of both dopant and contaminant distributions in semiconductor devices and trace elements in biological samples. 10.1.4 Ionization Schemes
As illustrated in Figure 10.2, several laser schemes can be used to ionize elements and molecules. In Figure 10.2, scheme (a) represents nonresonant ionization. Since the ionization cross-section is very low, a very high laser intensity is required to saturate the ionization process. Scheme (b) shows a simple single-resonance scheme; this is the simplest – but not necessarily the most desirable – scheme for resonant postionization. Cross-sections for (photo-)ionization are much lower than those for resonant excitation steps. The photoionization laser intensity must, therefore, be proportionately higher to ionize the atoms efficiently. High-intensity visible or UV laser beams can cause nonresonant ionization of interfering species, but this is rarely a problem with high-intensity infrared (IR) beams. Whenever selectivity is an issue, therefore, IR wavelengths for the photoionization step are preferable. Scheme (c) is the most versatile of the schemes shown, and provides the best performance for metals. However, H, He, C, N, O, F, Ne, P, S, Cl, Ar, Br,
181
182
10 Laser Secondary Neutral Mass Spectrometry (Laser-SNMS)
Kr, I, Xe, and Rn cannot be analyzed by using scheme (c) with commercial laser systems and frequency up-conversion in nonlinear crystals. Resonance ionization scheme (d) can be used when an autoionizing structure occurs in the ionization continuum. As the cross-section for autoionization is significantly larger than for normal photoionization, the intensity of the ionizing laser can be reduced, and this enables the use of visible wavelengths without extensive nonresonant ionization of interfering species. Scheme (e) is used when even the lowest-energy-excited states of the analyte element require vacuum UV wavelengths. This differs from scheme (c) only in the experimental method used to generate the wavelength for the first transition. For wavelengths below approximately 189 nm, either four-wave mixing or anti-Stokes Raman must be used, and this makes scheme (e) the most experimentally complex of the resonance ionization schemes. With this technique, it is possible to measure the amount of 81Kr (a radioactive isotope with approximately 1 part in 1018 natural abundance in the atmosphere) in a water sample after several isotope-enrichment steps [12]. Scheme (f) is generally used when the energy of the excited state structure is too high for schemes (b), (c), or (d), and the experimental complexity of scheme (e) is to be avoided. Examples are C, N, O, P, S, Cl, Br, Kr, I, and Xe. The two-photon transition used in scheme (f) is a secondorder process, which is why its cross-section is quite low compared to the onephoton transitions described in the other schemes. This means that high-intensity light is needed, and focusing of the laser beam is generally required. In fact, the intensity of the light required for efficient two-photon excitation is usually sufficient to ionize the excited state, and therefore scheme (f) often requires only a single wavelength. However, this experimental simplicity is counterbalanced by the fact that the high-intensity light can cause nonresonant ionization processes, which may lead to isobaric interferences when a mass spectrometer with low mass resolution is used. Isotope shifts for most elements are small in comparison to the bandwidth of the pulsed lasers used in resonance ionization experiments, and thus all the isotopes of the analyte will be essentially resonant with the laser. In this case, isotopic analysis is achieved by using a mass spectrometer. TOF mass spectrometers are especially well-suited for the isotopic analysis of ions produced by pulsed resonance ionization lasers, because all of the ions are detected in each pulse.
10.2 Instrumentation
A versatile laser-SNMS instrument consists of a versatile microfocus ion gun, a sputtering ion gun, a liquid metal ion gun, a pulsed flood electron gun, a resonant laser system consisting of a pulsed Nd:YAG laser pumping two dye lasers, a nonresonant laser system consisting of a high-power excimer or Nd:YAG laser, a computer-controlled high-resolution sample manipulator on which samples can be cooled or heated, a video and electron-imaging system, a vacuum lock for sample introduction, and a TOF mass spectrometer.
10.4 Quantification
10.3 Spectral Information
Compared with electron impact on molecules, laser photoionization can induce the fragmentation of molecules to a much lesser extent by optimizing the laser settings intensity, pulse width, and wavelength. Especially, the use of a fs-laser for postionization can significantly reduce molecular fragmentation. Consequently, more chemical information is available from laser-SNMS spectra. On the other hand, significant multiple ionization of atoms typically occurs when high laser power is used to achieve high postionization rates.
10.4 Quantification
Only a portion of the neutral particles present in the ionization volume (IV) of the laser beam is available for ionization. For this description, the geometric yield YIV (X 0i ) is introduced (for notation, see Section 7.3): YIV (X 0i ) =
No. of X 0i in the ionization volume No. of X 0i
(10.1)
Postionization occurs with the probability α(X 0i → X ⊕j ) (here. the sign ⊕ is used to denote a photoion):
α(X i0 → X ⊕j ) =
No. of generated photoions X ⊕j No. of X i0 in the interaction volume
(10.2)
Molecular particles can dissociate during the ionization process. This is taken into account by the different indices i and j. For postionized particles, it is possible to define a generalized transformation probability P ⊕ (A → X ⊕j ): P ⊕ (A → X ⊕j ) =
No. of generated photoions X ⊕j No. of sputtered A
(10.3)
P ⊕ (A → X ⊕j ) is given by Equation 10.4: P ⊕ (A → X ⊕j ) =
∑ P(A → X )⋅Y 0 i
IV
(X i0 ) ⋅ α(X i0 → X ⊕j )
(10.4)
i
The summation over i is necessary, because different precursors can be transformed into the same photoion X ⊕j by fragmentation. For detection of the generated ionic particles, the transmission of the mass spectrometer and the detection probability are considered in the same way as discussed in Section 7.3. An important quantity is, again, the useful yield Yu (X i⊕ (A)): Yu (X i⊕ (A)) =
No. of detected X i⊕ No. of sputtered A
(10.5)
183
184
10 Laser Secondary Neutral Mass Spectrometry (Laser-SNMS)
Values of Yu (X ⊕i (A)) for elements typically range from 10−4 to 10−2 in NR-laserSNMS, and from 10−2 to 10−1 in R-laser-SNMS. If the experimental conditions are not well known, the concentration of A can also be quantified by using the relative sensitivity factor (RSF) method (Equations 7.8 and 7.9 in Section 7.3).
10.5 Applications 10.5.1 Nonresonant Laser-SNMS
Figure 10.3 [4] shows a typical NR-laser-SNMS spectrum of a contaminated Si wafer under low-dose sputtering conditions (primary ion fluence <1014 cm−2). For postionization, the radiation of an excimer laser operating at a wavelength of
Figure 10.3 Nonresonant laser-SNMS spectrum of a Si wafer contaminated with a variety of
transition metals and hydrocarbons [4].
10.5 Applications NR-laser-SNMS: Relative sensitivity factors S (Me, ∑Si) and detection limits DL for metals on Si wafer surfaces.
Table 10.1
Element
Al Ti V Cr Fe Co Ni Cu Ga As Mo W
DL (atoms cm−2)
S (M, ∑Si) 193 nm
248 nm
193 nm
0.9 1.9 2.1 2.9 1.6 1.1 0.9 1.2 4.4 1.9 1.8 5.1
0.9 9 42 28 31 6.9 14 24 0.4 10 32 39
8 5 3 3 5 5 1 5 5 3 3 5
× 109 × 109 × 108 × 109 × 109 × 108 × 109 × 109 × 108 × 108 × 108 × 108
248 nm 3 5 3 3 5 5 1 5 8 3 3 5
× × × × × × × × × × × ×
1010 109 108 109 109 108 109 109 109 108 108 108
248 nm was focused into the sputtered cloud of particles (intensity 1010 W cm−2). Since at 248 nm the ionization cross-section of Si is relatively low, the intensity of the matrix signals (Si, SiO) is lower than expected. Following the use of a highresolution TOF mass spectrometry (mass resolution m/Δm = 3000), the metal contaminants (Mn 20 ppm, Ni 130 ppm, Cu 500 ppm, and Zn 400 ppm) are clearly separated from the molecular signals at the same integral mass. The ultimate detection limits (DLs) of NR-laser-SNMS are given by the amount of sample material available and by the useful yield. If the outer monolayer is completely consumed within an area of 10−4 cm2, and if the useful yield is 10−2–10−3, then a surface coverage of approximately 107 atoms cm−2 can be detected by at least 10 ions. This limit can only be reached, however, if the corresponding ion signal is free from background and from mass interference. In Table 10.1 are listed the relative sensitivity factors S (M+, Si+ + SiO+) with respect to the sum of the matrix signals Si+ and SiO+, and the DL determined for several metals on Si wafers by nonresonant postionization with 248 nm and 193 nm light [4]. Due to the low matrix-dependence of laser-SNMS, the same sensitivity can be achieved for the detection of trace metals in matrices such as polymers or large biomolecules, as for semiconductor matrices. Element mapping with nonresonant laser-SNMS can be used to investigate the structure of electronic devices and to locate defects and microcontaminants [13]. Typical SNMS maps for a GaAs test pattern are shown in Figure 10.4. In the subscript of each map, the maximum number of counts obtained in one pixel is given. These images were acquired using a 25 keV Ga+ liquid metal ion source with a spot size of approximately 150–200 nm. For the given images, only 1.5% of a monolayer was consumed – thus, the term “static SNMS” is appropriate.
185
186
10 Laser Secondary Neutral Mass Spectrometry (Laser-SNMS)
Figure 10.4 Nonresonant laser-SNMS mapping of a contact test structure on GaAs. Field of
view 40 × 40 μm2 [13].
10.5.2 Resonant Laser-SNMS
Figure 10.5 shows the Cu depth profiles obtained from silicon and 0.8 μm silicon oxide on silicon samples. For this experiment, half a silicon wafer without an SiO2 layer and half an SiO2-covered Si wafer were implanted simultaneously, with 5 × 1014 63Cu atoms cm−2 at 150 keV. For the analysis, both samples were cooled to 107 K to reduce migration effects, and depth profiles taken under the same experimental conditions. The integrated Cu signal for both samples – that is, the Cu sensitivity factor for Cu in silicon and for Cu in silicon oxide – was almost the same (<3% difference, which is due to the differences in sputter yield), thus demonstrating the matrix independence and quantification accuracy possible with R-laser-SNMS. The dynamic range for 63Cu is better than six orders of magnitude. Due to the limited mass discrimination of the implanter, 2% 65Cu is present. At the SiO2–Si interface a higher concentration of Cu was observed. The isotopic ratio at this pile-up near the interface and below is closer to that of natural copper, indicating that the sample was contaminated with Cu when the oxide was grown. The Cu profile in Si was broadened due to channeling effects that occur in crystalline silicon, but not in amorphous silicon oxide [14]. The lowest measured signal corresponds to a concentration of 400 ppt, which is still not the detection limit. Figure 10.6 shows the R-laser-SNMS images of copper concentration in the region of Te and Cd inclusions in a CdZnTe (CZT) film. It is clear from these
10.5 Applications
Figure 10.5 Depth profile of 63Cu in Si and of 63Cu and 65Cu in SiO2 [14].
Figure 10.6 R-laser-SNMS images of copper around (a) tellurium and (b) cadmium inclusions [15].
images that the copper concentration is approximately tenfold higher in the Te inclusion (peak concentration ca. 50 ppb) than in the surrounding CZT matrix, and approximately 500-fold higher in the Cd inclusion. These data support the theory that the copper migrates preferentially to Te/Cd second-phase regions inside the CZT matrix [15]. However, further experiments must be conducted to verify whether the solubility of Cu is much higher in Cd than in Te. Resonant as well as nonresonant laser-SNMS can also be used to locate and quantify pharmaceutical products in tissue, with subcellular resolution and sub-ppm detection limits [16–18].
187
188
10 Laser Secondary Neutral Mass Spectrometry (Laser-SNMS)
Figure 10.7 118Sn image of Sn-labeled DNA hybridized to six oligonucleotide spots; three complimentary and three noncomplementary spots.
R-laser-SNMS can also be applied to detect Sn-labeled DNA at positively hybridized and unhybridized sites on a DNA biosensor chip which can be used, for example, for the diagnosis of genetic disease and cancer [19, 20]. DNA diagnostics using sequencing by hybridization (SBH) involve the binding of a small (typically 18–20-mer) oligonucleotide to a chip that might be composed of glass, silicon, gold, or platinum. By employing the polymerase chain reaction (PCR), fragments of the genomic DNA can be produced, each with an attached label. The size of such fragments can vary from a few dozen to several hundred or several thousand nucleotides. Figure 10.7 shows an R-laser-SNMS image of 118Sn obtained from a DNA biosensor chip. For this experiment, two different 17-mer oligonucleotides were bound to platinum circles on a silicon wafer (three locations each). Enriched 118Sn-labeled DNA, which was completely complementary to one of the oligonucleotides and noncomplementary to the other, was hybridized to the chip. Following hybridization, the sample was washed to remove any unhybridized probes. The image shows three peaks at the complementary DNA sites, with the discrimination between hybridized and unhybridized sites being better than 100. A single point on the top center spot was analyzed twice before the full image was taken, which led to a slight reduction of the signal. The same technique can be readily applied to a variety of problems in bioanalytical chemistry, such as the screening of combinatorial libraries.
References
References 1 Becker, C.H. and Gillen, K.T. (1984) Anal. Chem., 56, 1671–1674. 2 Becker, C.H. (1987) J. Vac. Sci. Technol. A, 5, 1181–1183. 3 Pallix, J.B., Becker, C.H., and Newman, N. (1988) J. Vac. Sci. Technol. A, 6, 1049–1052. 4 Terhorst, M., Möllers, R., Schnieders, A., Niehuis, E., and Benninghoven, A. (1994) Elemental Surface Analysis of Si Wafers by Time-of-Flight Mass Spectrometry of Laser Postionized Sputtered Neutrals, in SIMS IX (eds A. Benninghoven, Y. Nihei, R. Shimizu, and H.W. Werner), John Wiley & Sons, Ltd, Chichester, pp. 434–437. 5 Peterson, R.E., Nair, A., Dambach, S., Arlinghaus, H.F., and Tyler, B.J. (2006) Appl. Surf. Sci., 252, 7006–7009. 6 Thonnard, N., Parks, J.E., Willis, R.D., Moore, L.J., and Arlinghaus, H.F. (1989) Surf. Interface Anal., 14, 751–759. 7 Arlinghaus, H.F., Spaar, M.T., and Thonnard, N. (1990) J. Vac. Sci. Technol. A, 8, 2318–2322. 8 Arlinghaus, H.F., Spaar, M.T., Thonnard, N., McMahon, A.W., Tanigaki, T., Shichi, H., and Holioway, P.H. (1993) J. Vac. Sci. Technol. A, 11, 2317–2323. 9 Pappas, D.L., Hrubowchak, D.M., Ervin, M.H., and Winograd, N. (1989) Science, 243, 64–66. 10 Pellin, M.J., Young, C.E., Calaway, W.F., Whitten, J.E., Gruen, D.M., Blum, J.B., Hutcheon, I.D., and Wasserburg, G.J. (1990) Phil. Trans. R. Soc. Lond. A, 333, 133–146. 11 Downey, S.W., Emfrson, A.B., and Kopf, R.F. (1991) Inst. Phys. Conf. Ser., 114 (Resonance Ionization Spectroscopy), 401–404.
12 Thonnard, N., Willis, R.D., Wright, M.C., Davis, W.A., and Lehman, B.E. (1987) Nucl. Instrum. Methods Phys. Res. B, 29, 398–406. 13 Zehnpfenning, J., Niehuis, E., Cramer, H.-G., Heller, T., Möllers, R., Terhorst, M., Schnieders, A., Zhang, Z., and Benninghoven, A. (1994) HighResolution Surface Imaging and Microanalysis of Electronic Materials by TOF-SIMS and Laser-SNMS, in SIMS IX (eds A. Benninghoven, Y. Nihei, R. Shimizu, and H.W. Werner), Chichester, pp. 561–564. 14 Arlinghaus, H.F. and Joyner, C.F. (1996) J. Vac. Sci. Technol. B, 14, 294–300. 15 Arlinghaus, H.F., Joyner, C.F., Tower, J., and Sen, S. (1997) Ultra-Sensitive and Quantitative Profiling and Imaging of Dopants and Impurities in Semiconductors, in SIMS X (eds A. Benninghoven, B. Hagenhoff, and H.W. Werner), John Wiley & Sons, Inc., New York, pp. 463–466. 16 Arlinghaus, H.F., Spaar, M.T., Switzer, R.C., and Kablaka, G.W. (1997) Anal. Chem., 69, 3169–3176. 17 Arlinghaus, H.F., Kriegeskotte, C., Fartmann, M., Wittig, A., Sauerwein, W., and Lipinsky, D. (2006) Appl. Surf. Sci., 252, 6941–6948. 18 Kriegeskotte, C., Cantz., Haberland, J., Zibert, A., Haier, J., Köhler, G., Schöler, H.R., Schmidt, H.H.-J., and Arlinghaus, H.F. (2009) J. Mass Spectron, 44, 1417–1422. 19 Arlinghaus, H.F., Kwoka, M.P., Guo, X.Q., and Jacobson, K.B. (1997) Anal. Chem., 69, 1510–1517. 20 Arlinghaus, H.F., Kwoka, M.P., and Jacobson, K.B. (1997) Anal. Chem., 69, 3747–3753.
189
191
11 Rutherford Backscattering Spectroscopy (RBS) Leopold Palmetshofer
11.1 Introduction
Rutherford backscattering spectroscopy (RBS) is one of the most frequently used techniques for the quantitative analysis of the composition, thickness, and depth profiles of thin solid films or solid samples near the surface region. RBS was first introduced during the 1960s, and has since evolved into a major materialscharacterization technique. Whilst the number and range of applications of RBS are enormous, because of its quantitative features the method often also serves as a standard for other techniques.
11.2 Principles
In RBS, a beam of monoenergetic ions, usually H+ or He+ of typical energy 0.5 to 2.5 MeV, is directed at a target, and the energies of the ions which are scattered backwards are analyzed. In the backscattering collision, energy is transferred from the impinging particle to the stationary target atom. The energy ratio between the projectile energy E1 after collision and the energy E0 before collision, derived from binary collision theory, is [1, 2]: ⎡ (M22 − M12 sin2 q )1 / 2 + M1 cosq ⎤ E1 =K = ⎢ ⎥ E0 M 2 + M1 ⎢⎣ ⎥⎦
2
(11.1)
The energy ratio E1/E0, called the kinematic factor K, shows that the energy after scattering depends only on the mass M1 of the projectile, the mass M2 of the target atom, and the scattering angle θ (i.e., the angle between the incident and scattered beams). If M1, E0, and θ are known, then M2 may be determined and the target element identified. For direct backscattering through 180°, the lowest value of the energy ratio is given by:
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
192
11 Rutherford Backscattering Spectroscopy (RBS)
E 1 ⎛ M 2 − M1 ⎞ =⎜ ⎟ E 0 ⎝ M 2 + M1 ⎠
2
(11.2)
If M1 = M2, the incident particle is at rest after a central collision, and all the energy will be transferred to the target atom. For target atoms with M2 < M1 no backscattering occurs. The energy of the backscattered ion is given by Equation 11.1 only for scattering by an atom at the surface of the target. In RBS, however, the ion beam penetrates the target, and an ion may be backscattered by target atoms at any point along its path. In the energy region used for RBS the ion trajectory is a straight line (apart from the backscattering collision), along which the ion loses energy primarily through excitation and ionization of atomic electrons (electronic energy loss). The energy loss per unit path length, dE/dx, is called the stopping power. These additional energy losses broaden the peak to be observed in an RBS spectrum of a thin film. The situation is illustrated in Figure 11.1a, in which a thin film of Ni is shown deposited on a Si substrate. Only those particles scattered from the front surface
Figure 11.1 Schematic backscattering spectra for MeV He+ ions incident on a 100-nm Ni film
on Si (a) and after reaction to form Ni2Si (b). The depth scales are indicated below the energy axes [1].
11.2 Principles
of the Ni film will have an energy given by the kinematic equation (Equation 11.1), E1 = KNiE0. As the particles traverse the solid, they lose energy along the incident path. Hence, particles scattered from a Ni atom at the Si–Ni interface will have an energy less than KNiE0. On the outward path the particles again lose energy, but on emerging from the surface the particles scattered at the interface will have a total energy difference of ΔE compared to particles scattered at the surface, and this results in a broad peak in the backscattering spectrum. The peak width ΔE is related to the thickness of the Ni film (see below). For Si atoms, the kinematic factor is smaller, KSi < KNi, because of their lower mass. Backscattered particles from Si atoms appear at lower energies in the spectrum. Because no Si atoms are at the surface, the energy spectrum produced by scattering from Si starts at an energy lower than KSiE0, and then extends to zero energy, because the Si atoms form the substrate with an effectively infinite depth. For a surface layer consisting of atoms of a lower mass than the substrate atoms, the peak arising from the surface layer will merge with the broad continuum and appear as a small feature on top of it. For surface analysis by RBS, the conditions must be such that the mass of surface atoms is considerably higher than the mass of substrate atoms, if the peak from surface atoms is to be completely resolved. Only under these circumstances is a high sensitivity of up to ∼10−3 atomic layers possible. No chemical information can be obtained, however. So far, it has been tacitly assumed that all the target atoms are equally visible for the projectiles. In a single crystal, the atomic rows and planes can guide energetic ions along the channels between rows and planes, so that the ion beam penetrates deeply into the crystal; this effect is known as channeling [1, 3]. Channeling occurs when the ion beam is carefully aligned with a major symmetry direction of the single crystal. Figure 11.2 shows a side view of this process, in which most of the ion beam is steered through the channels formed by the strings of atoms. Channeled particles cannot approach close enough to the atomic nuclei to undergo large-angle Rutherford scattering, and hence the scattering is drastically reduced by a factor of approximately 100. Atoms at the surface produce a surface peak in the backscattering spectrum, because the ion beam is scattered from the surface atoms with the same intensity as from a random array of atoms. The second atom in each string of atoms is completely shadowed in a perfect rigid
Figure 11.2 Schematic diagram of particle trajectories undergoing scattering at the surface and channeling within the crystal. The depth scale is compressed relative to the width of the channel in order to display the trajectories [1].
193
194
11 Rutherford Backscattering Spectroscopy (RBS)
lattice, because of the “shadow cone” formed by the ion trajectories. Thus, channeling enormously improves the sensitivity of RBS to atoms at the surface.
11.3 Instrumentation
Ion beams suitable for RBS are produced in particle accelerators. The ions are first produced in an ion source, then accelerated, mass- and charge-selected, and finally transported to the specimen. A general-purpose ion source is the duoplasmatron, which can produce ions of both polarities. For producing MeV ion beams, the most widely used device historically has been the Van de Graaff electrostatic accelerator, either single-ended or double-ended (tandem Van de Graaff), and some of these are still in use today. Modern-day accelerators are built from capacitively coupled power supplies of the Cockroft–Walton type (Tandetrons or Singletrons), the main advantage of which is that they have no moving parts. Detailed information on currently available accelerator technology is provided elsewhere [4]. Accelerators normally produce ions in several charge states, and even a multiplicity of species. In order to extract a beam suitable for materials analysis, the beam is first passed into the field of an analyzing magnet. A mass- and chargeselected beam then enters a UHV environment via differentially pumped apertures, and is steered to the target by electrostatic or magnetic lenses. The beam size at the target is, typically, 1 mm2. Small beam spots, if desired, can be obtained by using suitable lens systems. Selected area analysis using small spots is usually achieved by specimen manipulation. For general analysis, the ions strike the specimen at normal incidence. When structural information is sought (RBS + channeling), the specimen is mounted on a multiple axis manipulator, and can be rotated about the point of ion impact in order to vary the angle of incidence about the channeling directions. The scattering angle for optimum mass resolution would be 180 ° (a certain ΔM2 gives the largest change in K when θ = 180 °, Equation 11.1), but because of detector size, in practice θ ≈ 170° is chosen. The detection of backscattered ions is usually performed using a solid-state detector, either a silicon barrier detector or a passivated implanted planar silicon (PIPS) detector. For conventional RBS with He+ or H+ ions at 1–2 MeV, the PIPS detector has the better energy resolution (ca. 10 keV compared to 15 keV for surface barrier detectors [5]). The detector signals, which are highly proportional to the energy of the incident particle, are amplified and assorted in energy in a multichannel analyzer.
11.4 Spectral Information
An RBS spectrum contains quantitative information about the depth distribution of elements in a surface layer – that is, the mass of the scattering atoms, the com-
11.4 Spectral Information
position of the surface layer, the depth of scattering atoms, and the thickness of the surface layer. Evaluation of the mass of the target atoms from the energy of backscattered particles has been described above. An example of the nature of the compositional information obtainable from a RBS spectrum is based on Figure 11.1b which shows, schematically, a Ni film on Si reacted to form Ni2Si. Following the reaction, the Ni signal ΔENi has spread slightly, because of the presence of Si atoms contributing to the energy loss. The Si signal has a step at the energy KSiE0 corresponding to Si in the Ni2Si. The hatched peak areas in the figure are a measure of the number of particles scattered by Si or Ni atoms. As the probability of a collision between the projectile and a target atom is easily obtained (Equation 11.5), the number of Si and Ni atoms in the target and hence the composition NNi/NSi can be calculated. By knowing the composition of a layer it is possible to establish a depth scale for the distribution of an element, or to measure the layer thickness from the energy of the scattered particles. This depends on the energy loss of the projectile on its inward and outward paths, as described in Section 11.2. The energy difference, ΔE, for a particle scattered at the surface and a particle scattered at a depth x is given by: ΔE = [S]x
(11.3)
where [S], called the energy loss factor, depends on the stopping power on the inward and outward paths, the kinematic factor, K, and the orientation of the sample both to the incident beam and to the detector direction [2]. In Figure 11.1 the depth scales are indicated. RBS with channeling can be used to detect and measure the thickness of an amorphous layer on an otherwise crystalline substrate. The method is commonly used to obtain structural information about the damaged surface layer of ion-implanted semiconductors, and also to study removal of the damage on annealing. An example is shown in Figure 11.3 for GaAs [6]. If the incident ion beam is directed onto the GaAs crystal in a 〈110〉 channeling direction, the backscattering yield is drastically reduced (spectrum d) when compared to the yield when the beam is incident in a nonchanneling or “random” direction (spectrum a). Note the small surface peak in the channeled spectrum. Ion implantation with 120 keV Si+ ions to a dose of 5 × 1015 cm−2 produced an amorphous surface layer in GaAs. Because no channeling is possible in this layer, the intensity of the backscattered signal under channeling conditions increases, and is as high as for the random direction. Ions passing through the amorphous layer experience the crystalline nature of GaAs, which results in a reduced backscattering yield (spectrum b). From the width of the broad peak in curve b, the thickness of the amorphous layer is found to be ∼140 nm. After annealing at 950 °C, the intensity of the backscattering yield under channeling conditions falls substantially (spectrum c). The latter spectrum closely resembles that of crystalline GaAs, and indicates that the implantation damage has been almost completely annealed out.
195
196
11 Rutherford Backscattering Spectroscopy (RBS)
Figure 11.3 RBS spectra of GaAs implanted with Si (120 keV, 5 × 1015 cm−2) before and after
annealing at 950 °C. The uppermost spectrum is taken in a random direction, the others are taken in a channeling direction [6]. For details, see the text.
One of the most fascinating applications of channeling RBS is the study of lattice locations of impurity atoms. By measuring the angular dependence of the backscattering yield of the impurity and host atoms around three independent channeling axes, it is possible to calculate the position of the impurity. Details of this technique have been reported elsewhere [3].
11.5 Quantification
For an ion beam with the total number Q of ions impinging on a thin film, the number, QA, of particles backscattered from atoms of type A and registered in the detector (also called yield, YA), is given by: Q A = YA = QN Aσ A ΔΩ
(11.4)
where NA is the areal density of atoms A in the film (atoms cm−2), σA the differential scattering cross-section (cm2 sr−1), and ΔΩ the solid angle of the detector. The differential scattering cross-section describes the probability of a projectile to be scattered by a target atom through an angle θ into a solid angle dΩ centered about θ. If the interaction potential between the particle (M1, Z1) and the target atom
11.6 Figures of Merit
(M2, Z2) during scattering is given by the Coulomb potential, the cross-section is given by the Rutherford formula: 2 2 2 2 ⎛ Z1Z2e 2 ⎞ 4 ⎡⎣(M2 − M1 sin q ) + M2 cosq ⎤⎦ s (E ,q ) = ⎜ 1/2 ⎝ 4E ⎟⎠ M2 sin 4 q (M22 − M12 sin2 q ) 1/2
2
(11.5)
Equation 11.5 is given in cgs units, which are still commonly used for ion-beam analysis, while for practical calculations the number e2 ≈ 1.44 × 10−13 MeV cm is useful. For standard RBS with 1–2 MeV He+ ions, the use of the Rutherford crosssection is justified (which gives the technique its name). Deviations occur at both higher and lower energies [2]. For a particular primary ion and fixed experimental conditions, the scattering cross-section is a function only of the mass and atomic number of the scattering atom, and can be calculated. Thus, in principle, all of the terms in Equation 11.4, except the required NA, are known. RBS is, therefore, an absolute method that does not require the use of standards. For a compound film AmBn, the composition can be calculated from Equation 11.4 to be: n N B Q B s A (E ,q ) = = m N A Q A s B (E ,q )
(11.6)
Note that this ratio depends only on the ratio of measured yields, QA/QB, and knowledge of the cross-section ratio, σA/σB. The difficult-to-measure quantities Q and ΔΩ have canceled. For targets containing several elements which might produce overlapping peaks, the RBS spectra are analyzed by using computer simulations. The energy spectrum of the backscattered particles is calculated for the actual experimental conditions and an assumed target composition, after which the target composition is altered (either manually or by using a least-squares fitting procedure) until the calculated and measured spectra are closely matched. One widely used RBS analysis program is the RUMP code [7], although a recent addition to the growing list of analysis programs has been the “IBA Data Furnace.” This fitting code automatically extracts the elemental depth profiles from RBS data, without any user intervention [8].
11.6 Figures of Merit 11.6.1 Mass Resolution
Differentiating Equation 11.1 for fixed θ gives: ⎛ dK ⎞ ΔE1 = E 0 ⎜ ΔM2 ⎝ dM2 ⎟⎠
(11.7)
197
198
11 Rutherford Backscattering Spectroscopy (RBS)
If ΔE1 is set equal to δE (the minimum energy separation that can be resolved experimentally), then the mass resolution is given by ΔM2. For fixed δE, the mass resolution depends on E0, M1, and M2; it improves with increasing analysis beam energy E0 and mass M1 (11.2). The energy resolution δE is limited by detector resolution. 11.6.2 Sensitivity
The sensitivity obviously depends on the scattering cross-section, which in turn varies with the mass and energy of the incoming ions (see Equations 11.4 and 11.5). Experimentally, it may be limited by pulse pile-up in the detector, however. The detection limit for a thin layer of heavy elements on or in light substrates is approximately 10−12 atoms cm−2. The sensitivity for light elements is poor because of the underlying spectrum from heavier substrate atoms. 11.6.3 Depth Resolution
In depth profiling it is important to distinguish between accessible depth and depth resolution, both of which depend on the same factors (see Equation 11.3). Small energy loss factors result in a large profiling depth, but a poor depth resolution (for a given detector resolution). A typical depth resolution for conventional RBS is ∼10 nm, although this can be improved by tilting the sample normal relative to the ion beam (the “grazing angle technique”). 11.6.4 Accuracy
The accuracy in RBS results is approximately 3% for areal densities, and better than 1% for stoichiometric ratios. These high accuracies are only obtained when all relevant quantities are measured or evaluated carefully. Pitfalls which often prevent RBS from achieving its full accuracy are described in Ref. [9]. A calibration can be achieved by measuring standards obtained by either implanting into or depositing onto a light element (silicon) a known amount of a much heavier element (e.g., Ta or Sb).
11.7 Applications
Because RBS is a major technique for the analysis of thin solid films and surface layers, the number of reported applications is enormous, with approximately 1000 publications describing the use of RBS, either alone or with other analytical
11.7 Applications
Figure 11.4 RBS spectra of buried α-FeSi2 (solid line) and β-FeSi2 (dotted line) layers taken in
a random direction. The 〈111〉 channeling spectrum (circles) refers to the β-FeSi2 layer [10].
techniques, appearing annually. Most of these applications are in the field of semiconductor technology; for instance ion-implantation damage and damage annealing, implantation profiles, multilayer systems grown by molecular beam epitaxy (MBE) or chemical vapor deposition (CVD), silicide formation, and diffusion barriers for contacts are among the topics investigated routinely using RBS. A typical application example is the formation of a buried β-FeSi2 layer in Si (Figure 11.4 [10]). This system is favorable for RBS, because Fe is much heavier than Si. The FeSi2 was synthesized by implanting Fe to a high dose (2.7 × 1017 cm−2) into Si and performing a high-temperature annealing. FeSi2 occurs as different phases, which complicates formation of the low-temperature semiconducting βphase. Whilst α-FeSi2 is formed after a very brief annealing (10 s) at 1150 °C, the transformation to β-FeSi2 is achieved following long-term annealing (17 h) at 800 °C, a temperature well below the α–β transition. The RBS spectra for the process are shown in Figure 11.4. Here, the Si yield starts at the surface energy KSiE0 and has a large dip at lower energies, whereas the Fe peak starts at energies below KFeE0, indicating that the FeSi2 layer is buried below the Si surface. The difference between Fe signal heights for the α- and β-phases indicates a change in density during the phase transformation. It is estimated that the α-phase contains structural vacancies of ∼17%. The channeling spectrum of the β-phase shows that the buried layer is crystallographically aligned or slightly off-oriented to Si(111). Both, RBS and channeling are extremely useful for the characterization of epitaxial layers, an example being the analysis of a Si1–xGex/Si strained layer superlattice [11]. In this case, four pairs of layers, each approximately 40 nm thick, were grown by MBE on a (100)Si substrate although, because of the lattice mismatch between Si1–xGex (x ≈ 0.2) and Si, the Si1–xGex layers were strained.
199
200
11 Rutherford Backscattering Spectroscopy (RBS)
Figure 11.5 RBS spectra of 2.07 MeV He+ ions backscattered from a Si1–xGex/Si strained layer
superlattice. The solid line refers to a random direction; the other spectra are taken along a variety of channeling directions [11].
Figure 11.5 shows the RBS spectra in random and channeling directions; the four pairs of layers are clearly seen in both the Ge and Si yield. The spectrum for channeling in 〈100〉 direction is typical, whereas a large dechanneling is observed in the 〈110〉 direction, because of lattice distortion at the interface Si1–xGex/Si. (The strained Si1–xGex layer grows such that the in-plane lattice constant is the same as that of Si, whereas the perpendicular lattice constant is larger; the 〈100〉 channels are, therefore, rather undistorted and diagonal 〈110〉 channels are distorted by a small angle.) The large dechanneling yield is a direct consequence of the lattice strain. The distortion can be measured as a small angular shift in angular scans at energies corresponding to specified Si and Si1–xGex layers. By applying highresolution RBS, even the intermixing of Si and Ge during epitaxial growth has been studied [12]. Other fields for RBS analysis include optical and dielectric materials, hard and protective coatings, superconductors, and magnetic materials. RBS, as an extremely useful method for applied materials science, has also contributed to basic science, with one famous example being that of surface melting [13]. Figure 11.6 shows the normalized backscattering yields of H+ from Pb near the melting point in the 〈101〉 direction (double alignment). The surface peak increases with increasing temperature, and develops into a broad peak at 600.5 K. At 600.8 K, just above the melting temperature (600.7 K), the random spectrum is obtained, as would be expected for a bulk liquid. The broad peak below the melting temperature (spectrum d) can be fitted by assuming a disordered liquid surface layer (M) plus a partially ordered transition layer (I) between the liquid and the crystalline substrate. Whilst these were the first direct measurements of surface melting, later experiments showed surface melting to be highly dependent on the orientation of the surface.
References
Figure 11.6 Normalized backscattering yields of H+ ions from Pb near the melting point, with the incident beam and scattered beam directed along 〈101〉 crystal axes (double alignment). Spectrum a, 295 K; spectrum b, 506 K; spectrum c,
561 K; spectrum d, 600.5 K; spectrum e, 600.8 K. Spectrum d is fitted by a sum of contributions M, from a liquid surface layer, and I, from a partially ordered transition layer [13].
11.8 Related Techniques
Recent years have witnessed an increasing use of medium-energy ions and of heavier ions to optimize certain features of backscattering analyses. Two such examples include:
•
Medium-energy ion scattering (MEIS), which employs ions with energies at 100–500 keV, together with electrostatic analyzers or time-of-flight (TOF) detectors. MEIS offers a very shallow analysis depth, with resolution up to monolayers [14].
•
Heavy ion backscattering spectroscopy (HIBS) employs ions such as 12C, 16O, or 35Cl, together with TOF detectors to obtain extremely high sensitivities (∼109 atom cm−2) for trace analysis [15].
References 1 Feldman, L.C. and Mayer, J.W. (1986) Fundamentals of Surface and Thin Film Analysis, North Holland, New York. 2 Leavitt, J.A., McIntyre, L.C., and Weller, M.R. (1995) Backscattering spectrometry, in Handbook of Modern Ion Beam Materials Analysis (eds J.R. Tesmer and
M. Nastasi), Materials Research Society, Pittsburgh, pp. 37–81. 3 Swanson, M.L. (1995) Channeling, in Handbook of Modern Ion Beam Materials Analysis (eds J.R. Tesmer and M. Nastasi), Materials Research Society, Pittsburgh, pp. 231–300.
201
202
11 Rutherford Backscattering Spectroscopy (RBS) 4 Humphries, S. (1986) Principles of Charged Particle Acceleration, John Wiley & Sons, Inc., New York. 5 Steinbauer, E., Bauer, P., Geretschläger, M., Bortels, G., Biersack, J.P., and Burger, P. (1994) Nucl. Instrum. Methods B, 85, 642–649. 6 Bhattacharya, R.S. and Pronko, P.P. (1982) Appl. Phys. Lett., 40, 890–892. 7 (a)Doolittle, L.R. (1985) Nucl. Instrum. Methods B, 9, 344–351; (b)(1986) Nucl. Instrum. Methods B 15, 227–231. 8 Jeynes, C., Barradas, N.P., Marriott, P.K., Boudreault, G., Jenkin, M., Wendler, E., and Webb, R.P. (2003) J. Phys. D: Appl. Phys., 36, R97–R126; available at http:// www.ee.surrey.ac.uk/IBC/ndf (accessed 10 February 2009). 9 Davies, J.A., Lennard, W.N., and Mitchell, I.V. (1995) Pitfalls in ion beam analysis, in Handbook of Modern Ion
10 11
12 13 14
15
Beam Materials Analysis (eds J.R. Tesmer and M. Nastasi), Materials Research Society, Pittsburgh, pp. 343–363. Mantl, S. (1993) Nucl. Instrum. Methods B, 80/81, 895–900. Parikh, N.R., Sandhu, G.S., Yu, N., Chu, W.K., Jackman, T.E., Baribeau, J.M., and Houghton, D.C. (1988) Thin Solid Films, 163, 455–460. Nakajima, K., Konishi, A., and Kimura, K. (1999) Phys. Rev. Lett., 83, 1802–1805. Frenken, J.W.M. and van der Veen, J.F. (1985) Phys. Rev. Lett., 54, 134–137. Vrijmoeth, J., Zagwin, P.M., Frenken, J.W.M., and van der Veen, J.F. (1991) Phys. Rev. Lett., 67, 1134–1137. Banks, J.C., Doyle, B.L., Knapp, J.A., Werho, D., Gregory, R.B., Anthony, M., Hurd, T.Q., and Diebold, A.C. (1998) Nucl. Instrum. Methods B, 136–138, 1223–1228.
203
12 Low-Energy Ion Scattering (LEIS) Peter Bauer
12.1 Principles
In principle, a low-energy ion scattering (LEIS) experiment is like RBS at low energies: an ion beam is sent to a target at a large angle with respect to the surface plane, and backscattered projectiles are detected at a large scattering angle [1, 2]. In a typical LEIS experiment, an electrostatic analyzer is used and therefore only positive ions are detected. Consequently, also the singly charged ion fraction P+ – that is, the yield of singly charged ions divided by the total backscattered yield – has to be included when calculating the ion yield Q A+ , backscattered from an atom of species A: Q A+ = Q 0N A
ds A + P TeΔW dW
(12.1)
Here, Q0 is the number of incident projectiles, NA the areal density of atoms A in the surface (atoms cm−2), dσA/dΩ the differential scattering cross section, T and ΔΩ the transmission function and the solid angle of the spectrometer, respectively, and ε is the detection efficiency of the ion detector. The scattering cross-section is considerably different from the Rutherford crosssection, as the distance of closest approach Rmin is rather large at low energies. Thus, electronic screening of the interaction between the nuclei is important. The screened scattering potential V(r) reads: V (r ) = VC (r ) + Ve (r ) = VC (r ) ⋅ F (r / a )
(12.2)
where VC and Ve are the potential of target nucleus and target electrons, respectively, and Φ(r/a) is the screening function, which describes how the Coulomb potential is weakened by electronic screening. Φ is a function of the reduced distance, r/a, where the screening length a is characteristic of the ion-target combination. Within the Thomas–Fermi–Molière model [3] and using Firsov’s definition of a, screening function and screening length are given by the following equations:
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
204
12 Low-Energy Ion Scattering (LEIS) 3 r c ir Φ ⎛⎜ ⎞⎟ = ∑ δ i exp ⎛⎜ − ⎞⎟ ⎝ a ⎠ i =1 ⎝ a⎠
0.8852 ⋅ a0
a = aF = 3
(
Z1 + Z2
)
2
(12.3a) (12.3b)
with a0 = 0.529 Å (Bohr’s radius), and the constants δ1 = 0.35, δ2 = 0.55, δ3 = 0.1, c1 = 0.3, c2 = 1.2, and c3 = 6. Thus, the influence of screening is negligible for small distance (Φ → 1 for r → 0), and limits the interaction to distances comparable to the screening length (Φ → 0 for r >> aF). It has been known, however, for many years that the screening function is not fully appropriate in the complete range of scattering angles and ion energies. One way to modify the screening function is to use a modified screening length a, obtained from aF by a multiplicative factor ca: a = c a ⋅ aF A suggestion for ca was given in Ref. [4]. Experimentally, the scattering potential can be deduced from the analysis of angular scans obtained for single crystalline targets, but even for a simple system such as He ions and Cu as a sample material, largely varying values for the screening length have been reported (see Ref. [5] and references therein). From the screened potential, the scattering angle θ(b) that corresponds to a given impact parameter b is obtained by solving the scattering integral – that is, the general relationship θ(b) for an arbitrary central potential [6]. From θ(b), the differential scattering cross-section can easily be obtained: ds b db = dW sin q dq
(12.4)
In Equation 12.1, the major unknown is the charge fraction P+, which today is still the subject of theoretical investigations [7]. The state-of-the-art knowledge is as follows. When noble gas ions are used as projectiles, the detection of backscattered positive ions leads to the excellent surface sensitivity that is typical for LEIS, at least at sufficiently low projectile energies – that is, below the reionization threshold. When only two charge states (0, +1) are of relevance, there are two contributions to P+: 1)
Projectile ions that have escaped neutralization along their trajectory and in the close collision.
2)
Projectile ions which were neutralized on their incoming path, reionized in the surface collision, and remained in the charged state when leaving the surface: + + P + = Psurv , in ⋅ (1 − PCIN ) Psurv , out + + + (1 − Psurv , in ) ⋅ PCIR ⋅ Psurv , out
(12.5)
12.1 Principles + + Here, Psurv ,in and Psurv ,out are the probabilities for the ion to evite neutralization on the incoming and outgoing paths, respectively, PCIN and PCIR are the probabilities for neutralization and reionization in a close collision. PCIN and PCIR depend heavily on the distance of closest approach in the surface collision, and on the detailed electronic structure of projectile and target atom, and become important at energies above a certain threshold energy, Eth [8]. In the reionization regime, the information depth of LEIS is largely increased due to the contribution of projectiles that are backscattered in deeper layers and reionized in a final collision with an atom close to the surface when leaving the sample [9]. At sufficiently low energies (E < Eth), the survivals usually dominate the charge fraction. As long as only one neutralization process dominates – that is, Auger + , is obtained from the processes for He+ – the surviving probability of the ion, Psurv transition rate of the Auger process (transitions per second), Γ(t):
⎡ ∞ ⎤ + Psurv = exp ⎢ − ∫ dtΓ (t )⎥ ⎣ 0 ⎦
(12.6)
Further evaluation of Equation 12.6 is based on the following assumptions: (i) backscattering of the projectile is by a single binary collision with a surface atom; and (ii) that neutralization occurs via the tunneling of a (delocalized) target conduction electron to the ion. With this, the time integral can be replaced by an integration over the projectile–surface distance, z, by use of the identity dt = dz/v⊥. Here, v⊥ denotes the component of the ion velocity perpendicular to the surface, and it will be different for the incoming and the outgoing parts of the trajectory, vin,⊥ and vout,⊥, due to different directions and to energy transfer in the binary collision. ⎡ ∞ ⎤ 1 + Psurv Γ (z)⎥ ,i = exp ⎢ − ∫ dz ⎣ zmin vi ,⊥ ⎦
(12.7)
+ Psurv is obtained as the product of the surviving probabilities along the incoming + + + = Psurv and the outgoing part of the trajectory, Psurv ,in ⋅ Psurv ,out. With the additional + assumption that vin,⊥ and vout,⊥ are independent of z, Psurv is obtained as
v ⎤ ⎡ v ⎡ v ⎤ + Psurv ≈ exp ⎢ − c − c ⎥ ≡ exp ⎢ − c ⎥ ⎣ v⊥ ⎦ ⎣ vin,⊥ vout ,⊥ ⎦
(12.8)
In Equation 12.8, vc ≡ ∫Γ(z)dz is a characteristic velocity, illustrating the neutralization efficiency: large values for vc correspond to strong neutralization. From Equation 12.8 we may draw the following conclusions: 1) 2)
At high energies (small values of 1/v⊥) the ion fraction is high, as there is only very little time available for neutralization. For a given geometry, the ion fraction decreases exponentially with the inverse of the velocity.
These theoretical predictions were verified experimentally for numerous target materials [10]. In Figure 12.1, P+ is shown for He+ scattered from Cu(110) (full symbols), Cu(100) (open asterisks and triangles) and polycrystalline Cu (crosses
205
12 Low-Energy Ion Scattering (LEIS)
1 0.8
+
He → Cu
+
0.6
Ion fraction P
206
0.4 v
0.2
c
Cu(110): Cu(100):
0.1 0.08
Cu-poly:
0.06 0.0
2.0 keV 1.5 keV 2.0 keV 1.9 keV 1.5 keV Markin et. al. Draxler et. al. –6
4.0x10
(11 0)
=1 .18 x10 5 m/ s
v
c
v c
–6
8.0x10
(1 00 )
po ly
=
1. 92 x1 0
–5
1.2x10
=1 .62 x1 5 0 m/ s 5 m /s
–5
1.6x10
1/v⊥ (s/m) Figure 12.1 Ion fraction P+ of He+ scattered
from Cu(110) (filled symbols), Cu(100) (open asterisks and triangles) and polycrystalline Cu (crosses and open circles) in the Auger neutralization regime, as a function of
1/v⊥ = 1/v⊥,in + 1/v⊥,out. Also shown are single exponential fits (see Equation 12.8) with characteristic velocity values as indicated in the figure ([9]).
and open circles) in the Auger neutralization regime [9]. Note that in Figure 12.1, the dependence of P+ on 1/v⊥ is perfectly fulfilled. There is, however, a pronounced difference in the neutralization efficiency between Cu atoms in different crystal orientations. In terms of quantitative surface analysis, this might be called a physical matrix effect, and indicates that care must be taken in quantitative surface composition analysis.
12.2 Instrumentation
When comparing the RBS and LEIS set-ups, the following differences are apparent: RBS
LEIS
Accelerator Deflection magnet Silicon-detector H+, He+ ions
Ion source (<10 keV) Wien filter Electrostatic spectrometer or time-of-flight (TOF) spectrometer Noble gas ions (He+, Ne+)
Due to its lower beam energy, a LEIS ion source is much more compact than an accelerator for MeV ions, although it has the same purpose, namely to provide a beam of ions with well-defined energy and mass. In LEIS, the latter demand is
12.2 Instrumentation
Figure 12.2 Schematic diagram of the TOF-LEIS set-up analysis of the charge of light ions scattered from surface atoms (ACOLISSA) [11].
fulfilled by a Wien filter that selects ions of one specific mass and charge state, as does the deflection magnet in RBS. In LEIS, the energies of the ions are too low for passivated implanted planar silicon (PIPS) detectors, and would lead to a signal-to-noise ratio (SNR) less than unity. Therefore, a stack of microchannel plates is used to detect the ions in LEIS. The detection efficiency ε of the stop detector (microchannel plates) is included in Equation 12.1. The energy selection is carried out by using either an electrostatic analyzer or a TOF spectrometer. Each type of spectrometer has both, advantages and disadvantages: In a TOF measurement ([11], see Figure 12.2), short packets of ions, <30 ns long, are cut out of the primary beam by means of an electrostatic chopper, and the beam intensity is sufficiently low to warrant that, at most, one projectile out of a packet is registered by the stop detector. The flight time from the sample to the stop detector is measured, which provides information on the final velocity of the scattered projectile for a given path length. The ions and neutrals can be separated by post-acceleration of the ions. When the ions and neutrals are detected, the spectrum intensity increases by orders of magnitude, such that depth-dependent information is obtained (similar to RBS; see Chapter 11) with excellent depth resolution [12]. To analyze the surface structures of single crystals, channeling and blocking may be used to regain surface sensitivity.
207
208
12 Low-Energy Ion Scattering (LEIS)
A conventional electrostatic spectrometer functions only with ions, and is therefore surface-sensitive. However, it has a low efficiency when only those ions in a narrow energy window around a well-defined pass energy are detected simultaneously; the pass energy must be scanned in order to acquire the spectrum. The detection efficiency can be increased by orders of magnitude if a double-toroidal energy analyzer is used in combination with a position sensitive detector; such a system will measure simultaneously an energy spectrum of the backscattered ions in a larger energy window. This system, which has been developed at Eindhoven University [13], allows static LEIS measurements to be made (without noticeable damage), even at polymer surfaces for which the sputter rate is very high even under LEIS conditions.
12.3 LEIS Information
The energy spectrum in LEIS contains two types of information: energy and ion yield. 12.3.1 Energy Information
Until now, only the typical LEIS experiment has been considered – that is, with large angles of incidence and of exit with respect to the surface plane. Under these conditions, a quantitative composition analysis is generally possible, since the ion–target interaction can be considered as a binary collision, when the quantification is carried out correctly (see Section 12.4). Surface composition analysis by LEIS is based on the use of noble gas ions as projectiles, and utilizing the superb surface sensitivity of LEIS in the Auger neutralization regime. A consequence of this surface sensitivity is that the LEIS energy spectrum consists of lines, one per element, if the masses differ sufficiently. The lines are narrow, as inelastic energy losses play only a minor role. Thus, the information acquired on the atomic species present is deduced from the energy of the backscattered ions, which may be converted to the mass of the scattering center (Figure 12.3 [14]). In this figure, the mass range is shown to depend on the projectile mass, where LEIS is sensitive. 12.3.2 Yield Information
As may be seen in Equation 12.1, the backscattered ion yield in LEIS is a measure of the concentration of a certain element in the surface, if the other quantities are known. This is generally the case: the scattering cross-section may be calculated along the line of Equations 12.3 and 12.4, or acquired from tabulated data. For a TOF system, the transmission probability T is unity, and the solid angles ΔΩ+ and ΔΩ0 for ions and neutrals, respectively are equal; the only nontrivial quantity that
12.3 LEIS Information
Figure 12.3 LEIS spectra obtained from an Os/Ru top layer dispenser cathode with 3 keV He, Ne, and Ar projectiles, respectively. The He spectrum demonstrates the absence of O
on an undamaged cathode surface. The insert in the He spectrum was obtained for a cathode exposed to 20 Langmuir oxygen at room temperature [14].
must be determined carefully is the detection efficiency, ε. For an ideal electrostatic spectrometer, the transmission probability T is proportional to E when the pass energy is varied (as long as the detection efficiency is energy-independent); the solid angle ΔΩ of the spectrometer can be assumed to be constant; ε must be determined independently. In the absence of any physical or chemical matrix effects, and for fixed experimental conditions (constant primary energy, constant geometry, …), Equation 12.1 may be simplified to Q A+ = hA c A
(12.9)
Here, ηA is the sensitivity factor which contains all element specific quantities, and cA is the concentration of element A at the surface. Note that these conditions are best met in the reionization regime, that is, at E >> Eth, where surface-dependent Auger neutralization effects play a minor role. In Figure 12.4, the LEIS spectra are shown for two completely different types of Al2O3 samples – α-alumina (sapphire) and γ-alumina (a powder with a high specific surface area); both materials show very similar results after thermal treatment at 400 °C [15]. The lowering of the Al signal in γ-alumina was ascribed to shielding by hydroxyl groups which are typical adsorbates on γ-alumina.
209
210
12 Low-Energy Ion Scattering (LEIS)
Figure 12.4 LEIS obtained from different types of alumina with 3 keV He ions [15].
Figure 12.5 Different experimental geometries for low-energy ion spectroscopies.
Figure 12.6 Schematic illustration of projectile trajectories, showing focusing collisions when
the projectiles impinge under a critical angle αc [16].
When scattered intensities for neutrals and ions are measured in angular scans – that is, for variation of the polar angle or the azimuth – information on atomic structure may be obtained for a given single crystalline surface (see Figure 12.5). Both, LEIS and ICISS make use of the shadow cone, and vary the scattering geometry so that deeper layer atoms are either hidden in the shadow of a first layer atom, or experience enhanced ion flux density when the primary beam impinges at the critical angle (see Figure 12.6).
12.5 Applications of LEIS
12.4 Quantification
The usefulness of Equation 12.9 depends crucially on the question of whether, or not, the sensitivity factor ηA depends on the physical properties of the surface (physical matrix effects) or on the presence of other elements in the surface (chemical matrix effects). In general, it has been shown experimentally that neutralization depends only on the atomic number of the scattering center, and that matrix effects occur only rarely (see Table 9 in Ref. [1]). An instructive example of this is the neutralization of He by Al in the pure metal, and in alumina. The slopes of the neutralization curves are shown to be the same in both cases – that is, the matrix effects are absent [17]. Since for Al the reionization threshold Eth = 300 eV, it is clear from Figure 12.4 that the probabilities for collision-induced processes are very insensitive to details of the band structure, thereby facilitating quantitative surface composition analysis, albeit at the expense of a reduced surface sensitivity (see above). From a practical point of view, it is more convenient to measure intensity ratios rather than absolute intensities. Thus, for example, polycrystalline Cu may serve as a reference material, relative to which the ion intensities backscattered from the atoms of the surface under consideration are measured: Q A+ hAc A = + Q Cu hCu
(12.10)
Another example where quantitative surface composition analysis is possible for a nontrivial surface is shown in Figure 12.7, where for the systems Ta + O and Nb + O adsorption the ion signal from the metal is shown as a function of the ion signal from O. In this binary case, the following concentration equations hold: c Ta + c O = 1
(12.11a)
c Nb + c O = 1
(12.11b)
Inserting Equation 12.11 into Equation 12.10 yields: + Q O+ = hO − Q Ta
hO hTa
(12.12)
From Equation 12.12, one expects a linear relationship between QO and QTa, QNb and, indeed, this is observed experimentally ([18], see Figure 12.7).
12.5 Applications of LEIS
Applications of LEIS are widespread, going far beyond the needs of a typical surface science laboratory, since LEIS is capable of providing information also on
211
212
12 Low-Energy Ion Scattering (LEIS)
Figure 12.7 Peak intensities of Nb and Ta versus the oxygen peak intensity for adsorption of
oxygen on the pure metals [18].
Figure 12.8 (a) LEIS spectra of the 63
Cu/68ZnO/SiO2 catalyst obtained with 5-keV Ne+ ions for different doses (see insert, spectra are shifted vertically for clarity). Catalyst reduction temperature 700 K. Solid
lines: fitted Gauss peaks [19]; (b) The relative coverage of Cu and ZnO on the silica supported catalyst, reduced at 700 K, as a function of the ion dose [19].
12.5 Applications of LEIS
Figure 12.9 (a) Energy-converted TOF-LEIS spectra N(Ef) recorded for Au films deposited on
B using 1.2-keV He+ ions. The inset shows the ion fraction (after post-acceleration) [12]; (b) Comparison of experimental spectra N(Ef) with Monte-Carlo simulations of thin Au films [12].
insulating samples and on very rough structures (e.g., catalysts). The main problem with rough insulating surfaces is that they require charge neutralization (flooding by thermal electrons which neutralize the charging of the surface by incoming ions and emitted energetic electrons). An ambitious goal is to characterize an unsupported catalyst, since in this case the surface is extremely rough and the target deteriorates rapidly under bombardment: energy deposition will lead to an enormous erosion, as the substrate cannot lose the energy deposited as a consequence of the low heat conductivity. Hence, static LEIS conditions must be applied in order to provide information on the surface alone. Figure 12.8a shows a series of LEIS spectra obtained with 5-keV Ne+ ions on a 63Cu/68ZnO/SiO2 catalyst, reduced at 700 K [19]. These LEIS spectra were
213
214
12 Low-Energy Ion Scattering (LEIS)
obtained at three different ion doses, namely 3 × 1014, 3.41 × 1015 and 8.67 × 1015 Ne cm−2, respectively. Due to the use of isotopically enriched Cu and Zn, and of Ne+ ions as projectiles, the Cu and Zn can clearly be separated in the LEIS spectrum. A strong dose dependence is seen. Figure 12.8b shows the dose-dependent surface concentrations of Cu and Zn. At low doses (<1.5 × 1014 Ne cm−2), the Zn concentration remained constant, whereas the Cu concentration was increased. At these low doses, the hydroxyl layer on top of the catalyst was sputtered. The Zn signal remained constant despite removal of the adsorbate, which indicated that at the virgin surface the Zn concentration was even higher. As a final application, TOF-LEIS for quantitative characterization of the growth mode of Au deposited on B is presented. Information on the fraction of the surface that is covered with Au (“filling factor,” f ) is deduced from the comparison of the spectrum height of scattered neutrals H0 to the height of a simulated spectrum, Hs for a known number of projectiles (see Figure 12.9), with f = H0/Hs. Information on the average cluster height 〈dAu〉 is obtained from the energy distribution of scattered neutrals and comparison with computer simulation (Figure 12.9). Very good depth resolution is obtained at low ion energy, where the relative time resolution is excellent.
References 1 Brongersma, H.H., Draxler, M., de Ridder, M., and Bauer, P. (2007) Surf. Sci. Rep., 62, 63–109. 2 Niehus, H., Heiland, W., and Taglauer, E. (1993) Surf. Sci. Rep., 17, 213–303. 3 Feldman, L.C. and Mayer, J.W. (1986) Fundamentals of Surface and Thin Film Analysis, North-Holland, Amsterdam. 4 O’Connor, J. and Biersack, J.P. (1986) Nucl. Instrum. Methods B, 15, 14. 5 Primetzhofer, D., Markin, S.N., Draxler, M., Beikler, R., Taglauer, E., and Bauer, P. (2008) Surf. Sci., 602, 2921. 6 Parilis, E.S., Kishinevsky, L.M., Turnaev, N.Y.U., Baklitzky, B.E., Umarov, F.F., Verleger, V.K.H., Nizhnaya, S.I., and Bitensky, I.S. (1993) Atomic Collisions on Atomic Surfaces, North-Holland, Amsterdam. 7 Goldberg, E.C., Monreal, R., Flores, F., Brongersma, H.H., and Bauer, P. (1999) Surf. Sci., 440, L875–L880. 8 Tsuneyuki, S. and Tsukada, M. (1986) Phys. Rev. B, 34, 5758–5768. 9 Primetzhofer, D., Markin, S.N., Juaristi, J.I., Taglauer, E., and Bauer, P. (2008) Phys. Rev. Lett., 100, 213201.
10 van den Oetelaar, L.C.A., Mikhailov, S.N., and Brongersma, H.H. (1994) Nucl. Instrum. Methods B, 93, 210–214. 11 Draxler, M., Markin, S.N., Ermolov, S.N., Schmid, K., Hesch, C., Gruber, R., Poschacher, A., Bergsmann, M., and Bauer, P. (2004) Vacuum, 73, 39. 12 Primetzhofer, D., Markin, S.N., Zeppenfeld, P., Bauer, P., Prusa, S., Kolibal, M., and Sikola, T. (2008) Appl. Phys. Lett., 92, 011929. 13 Brongersma, H.H., Groenen, P.A.C., and Jacobs, J.-P. (1994) Science of Ceramic Interfaces II (ed. J. Nowotny), Elsevier, pp. 113–182. 14 Denier van der Gon, A.W., Cortenraad, R., Jansen, W.P.A., Reijme, M.A., and Brongersma, H.H. (2000) Nucl. Instrum. Methods B, 56–64, 161–163. 15 van Leerdam, G.C. Surface analysis of catalysts by low-energy ion scattering, PhD thesis, Eindhoven University of Technology, The Netherlands, 1991. 16 Krauss, A., Auciello, O., and Schultz, J.A. (1995) MRS Bull., 20, 18–23. 17 Jacobs, J.-P. Catalytically-active oxides studied by low-energy ion scattering,
References PhD Thesis, Technical University Eindhoven, 1995. 18 van den Oetelaar, L.C.A., Jacobs, J.-P., Mietus, M.J., Brongersma, H.H., Semenov, V.N., and Glebovsky, V.G.
(1993) Appl. Surf. Sci., 70/71, 79–84. 19 Viitanen, M.M., Jansen, W.P.A., van Welzenis, R., and Brongersma, H.H. (1999) J. Phys. Chem. B, 103, 6025–6029.
215
217
13 Elastic Recoil Detection Analysis (ERDA) Oswald Benka
13.1 Introduction
From methods based on the elastic scattering of high-energy ions, quantitative information would be expected regarding the depth distribution of elements in the surface region of solids. In Rutherford backscattering spectrometry (RBS), projectiles scattered by angles larger than 90 ° are analyzed. As projectiles with mass M1 can only be backscattered from a target atom with mass M2 if M1 < M2, light projectiles such as protons and He ions are usually used in RBS. High backscattered energies and large backscattering cross-sections are found for heavy target atoms. RBS is, therefore, well-suited for the analysis of heavy target elements, though its sensitivity for light elements is poor. In elastic recoil detection analysis (ERDA), the target atoms – which are recoiled in the forward direction by projectiles with energies in the MeV range – are analyzed, whereas the scattered projectiles are not analyzed. Ecuyer et al. [1] described this method for the first time during the mid-1970s. Since the sensitivity of ERDA is approximately the same for all target atoms, the technique is mainly used for light element profiling, which is hardly possible by using RBS. An important application of ERDA is hydrogen profiling, as information is provided that cannot be obtained from RBS and other standard techniques such as Auger and photoelectron spectroscopy. As a result of momentum conservation in forward scattering, both scattered projectiles and recoiled target atoms emerge from the sample in the forward direction, and will be detected simultaneously in an ERDA experiment. As the energy spectra of recoiled atoms and of scattered projectiles overlap, the measured spectrum is very complex and a useful evaluation is usually not possible unless additional information is obtained simultaneously for particle identification. This allows the ERDA spectra to be split up into contributions of individual recoiled elements and of scattered projectiles. For particle identification, two methods are usually applied: (i) the (ΔE − E) method; and (ii) the time-of-flight (TOF) method. The (ΔE − E) method uses the specific energy loss ΔE of the detected particles in a thin solid or gas layer to distinguish between different ion species with the same Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
218
13 Elastic Recoil Detection Analysis (ERDA)
energy E; this yields the atomic number (“effective charge”) of the particles. The TOF method uses the velocity to distinguish between particles of the same energy; this yields the particle masses (see below). ERDA, like RBS, is based on the following physical concepts:
• • •
The kinematic factors describe the energy transfer from the projectile to target atoms in an elastic two body collision. The differential scattering cross-section provides the probability of the scattering event occurring. The stopping power gives the average energy loss of projectiles and recoiled atoms as they traverse the sample, and define the depth scale.
By applying these concepts, individual recoil energy spectra can be transformed quantitatively into the corresponding concentration–depth profiles. The depth from which recoils originate correlates with the energy of these recoils. The concentration of analyzed atoms at this depth is obtained from the measured intensity of recoils at this energy. Analytical expressions which provide the depth profile as a function of the measured energy spectrum are less straightforward than for RBS, because the energy loss of both, projectiles and recoil atoms, must be taken into account. Thus, measured energy spectra are most conveniently evaluated by use of computer codes. Currently, several programs are available that are capable of calculating the energy spectra of projectiles and recoils for a given sample and experimental set-up (geometry). In ERDA, different regimes have been developed with a broad range of projectiles and energies, which can be separated generally into three groups:
•
Light projectiles (He, C ions) with energies between 2 and 10 MeV using the (ΔE − E) method, for the analysis of very light elements, mainly hydrogen.
•
Medium-heavy projectiles (Cl) with energies of approximately 30 MeV using TOF, for the analysis of light elements.
•
Heavy projectiles (Au) with energies larger than 100 MeV using mainly the (ΔE − E) method, to analyze a broad range of light and intermediate elements.
The sensitivity and depth resolution of ERDA depend on the type of projectile, on the type of particle, and on energy measurement. Due to the broad range of particles and methods used, general statements regarding sensitivity and depth resolution are hardly possible. Some reviews of ERDA techniques are available, however [2–4].
13.2 Fundamentals
In ERDA, the particle yield is measured in forward scattering geometry – that is, at angles of detection <90 ° relative to the beam. The typical scattering geometry
13.2 Fundamentals
Figure 13.1 Schematic diagram of the ERDA geometry.
is shown in Figure 13.1. Projectiles impinge at an angle of incidence α between the ion beam and the sample surface on a target. Recoils and scattered projectiles, which leave the sample at an exit angle β relative to the sample surface, are observed at a recoil and scattering angle θ. When a projectile of mass M1, energy El, and atomic number Z1 collides with a target atom of mass M2 and atomic number Z2, it will transfer the energy E2 to the target atom at a recoil angle θ, which is given by: E2 = K RE1
(13.1)
where KR is the kinematic factor for elastic recoil. This can be derived from laws of conservation of energy and momentum to be: KR =
4M1M2 cos2 θ (M1 + M2 )2
(13.2)
The projectiles which are also scattered with a scattering angle θ will have the energy: ES = K SE1
(13.3)
where Ks is the kinematic factor for elastic scattering: ⎡ (M 2 − M12 sin2 θ )1/ 2 + M1 cos θ ⎤ KS = ⎢ 2 ⎥⎦ M1 + M 2 ⎣
2
(13.4)
Both kinematic factors KR and KS are functions of the mass ratio M2/M1 and the recoil and scattering angle θ. Because the recoil kinematic factor is symmetric in M2,1 M1 = the masses M1 and M2, target atoms with masses M2,1 and M2,2 with M1 M2,2 will have the same recoil energy. The largest recoil energy is found for target atoms with M2 = M1 and is given by E2max = E1 cos2 θ . The scattered projectiles will have a higher energy than recoils for M2 > M1. If, therefore, in the surface region of a sample elements are present with atomic masses larger than that of the projectile, the scattered projectiles will have the highest detected energy. When M2 < M1 for all masses M2, the recoil atoms with highest masses will have the highest energy. It is also important to remember that there is a critical angle for scattering. No scattered projectiles are found for scattering angles, θ, larger than arc-sin (M2/M1), and this has stimulated the use of heavy projectiles in ERDA. A depth scale can be obtained from the energy of recoiled ions. If ions recoiled from a depth x are lower in energy by ΔE compared with ions recoiled from the
219
220
13 Elastic Recoil Detection Analysis (ERDA)
surface, a simple relationship between ΔE and x can be found for thin layers, when constant stopping power is assumed: ΔE = xNε R
(13.5)
where N is the atomic density and εR is the recoil stopping cross-section factor of the target:
εR = K R
ε in ε + out sin α sin β
(13.6)
Here, εin and εout are the stopping cross-sections of the incident projectiles and of the recoiled atoms. The recoil cross-section (in cm2) for Rutherford scattering is: 1 ⎡ Z Z (M + M2 ) ⎤ σ R = 5.18 × 10 −27 ⎢ 1 2 1 ⎥ M2E1 ⎣ ⎦ cos3 θ 2
(13.7)
for projectiles of energy E1, given in MeV. For atomic masses M2 << M1 the recoil cross-section is almost independent of the atomic number, because the cross-section becomes proportional to (Z2/M2)2 and the ratio Z2/M2 is close to 0.5 for all elements. ERDA with heavy projectiles thus has the advantage of almost constant sensitivity for all elements. Only for hydrogen is the ratio Z2/M2 equal to 1; hence, the intensity of hydrogen recoils is enhanced by roughly a factor of four. For the quantitative evaluation of ERDA energy spectra, considerable deviations of recoil cross-sections from the Rutherford cross-section (Equation 13.7) must be taken into account. Light projectiles with high energy can penetrate the Coulomb barrier of the recoil atom; the nuclear interaction generally leads to a cross-section that is larger than σR (see Equation 13.7). For example, the H recoil cross-section for MeV 4He projectiles is increased by approximately a factor of three. Bozoian et al. [5] described a means of estimating, for any ion-target system, the critical projectile energy above which deviations from the Rutherford cross-sections due to nuclear interaction are expected. In this non-Rutherford regime, measured cross-sections should be used in the spectrum evaluation (e.g., H recoil crosssections for 4He impact have been reported by several groups [6–9]). Deviations from Rutherford cross-sections are also found for heavy projectiles at lower impact energies, when the projectile can bind inner shell electrons which screen the nuclear charge. These deviations are usually small and can easily be taken into account by use of a theoretical correction [10].
13.3 Particle Identification Methods
As mentioned above, particle identification is achieved using either the energyloss measurement (the ΔE − E method) or the velocity measurement (TOF method).
13.3 Particle Identification Methods
In the (ΔE − E) method, those particles with the same energy, E, are identified by their energy loss, ΔE, in a thin solid or gas layer in front of the energy detector. This method can be used for particle energies E > EBr, where EBr is the Bragg energy – that is, the energy of maximum electronic stopping power. For particles with a fixed energy E < EBr, the stopping power depends only weakly on the atomic number of the ion, because in this regime stopping is roughly proportional to the velocity of the ion and, at a given energy, a heavier ion will have a lower velocity. Hence, particle identification by ΔE measurement is hardly possible at energies below EBr, which is approximately 100 keV for protons, approximately 0.5 MeV for He, 2.7 MeV for C, 7 MeV for O, and approximately 25 MeV for Si ions. For small accelerators (1–10 MeV ion energy), therefore, the (ΔE − E) method is applicable to the identification of very light elements only (mainly H), while for large accelerators and ion energies >100 MeV it enables the analysis of elements up to Cu. The simplest arrangement for (ΔE − E) measurement employs a stopper foil in front of the E-detector, which is usually a surface barrier detector (SBD) [11–13]. This system is mainly used for H profiling with low-energy light ions. The thickness of the foil must be such that H ions are transmitted through the foil (with a certain energy loss), whereas scattered projectiles and heavier recoiled atoms are stopped within the foil. A considerable improvement over the stopper foil method is the use of electron emission for particle identification [14]. In this case, when ions pass the foil, electrons are emitted from both surfaces, with the number of emitted electrons, Ne, being roughly proportional to the ΔE deposited in the foil by the ions. Thus, particles can be identified by measuring the number of electrons emitted from a set of thin foils in front of the SBD. As the emitted electrons originate from a layer close to the surface, foils as thin as possible can be used, so that energy loss and energy loss straggling are almost negligible. This results in a better depth resolution compared to the conventional stopper-foil technique. For ERDA arrangements using high-energy heavy projectiles, gas telescope detectors are mainly used for ΔE and E measurements [3, 15, 16]. For this, the individual particles ionize gas atoms in an ionization chamber, while the emitted electrons are accelerated by an electrical drift field (which is usually aligned perpendicular to the particle path) and detected by means of a split anode. The shorter section of the anode next to the entrance window yields the energy ΔE deposited in the front region next to the window; the remaining energy (E – ΔE) is then measured in the second part of the anode. The total energy, E, is obtained from the total charge accumulated in both sections of the anode. The second part of the ionization chamber, which measures the energy (E – ΔE), can be replaced by an SBD [17], and the first part – which measures the energy loss ΔE – by a transmission SBD [18, 19]. When SBDs are used to measure heavy ions, radiation damage of the detector by the ions must be taken into account. In TOF systems, the particle energies are usually determined by SBDs, in addition to particle velocities being obtained with a TOF set-up which primarily measures the time required by a particle to pass the distance between two thin foils 0.5–1 m apart [20, 21]. The first foil delivers a start signal, and the second a stop signal. Although the stop signal can also be obtained from the SBD, foils usually
221
222
13 Elastic Recoil Detection Analysis (ERDA)
provide better timing signals. The timing signal from the foils is obtained from electrons emitted when the particles pass the foil. Usually, carbon foils are used, because these are the most stable ultrathin foils. The emitted electrons are focused to a microchannel plate, which produces the time signal. The efficiency for the detection of light particles (hydrogen) is low, due to the low probability of electron emission from the foil [22]. Particle identification by the TOF method can also be achieved at particle energies below the Bragg energy, EBr, in contrast to (ΔE − E) systems. For low-energy particles the particle energy – which is a measure of the emission depth – can sometimes be calculated more precisely from the measured TOF than directly from the SBD spectrum [23]. Under these circumstances, the SBD is used to identify particles with the same TOF but different energy.
13.4 Equipment
In ERDA, the projectiles are high-energy ions produced by an ion accelerator. The lowest ion energies used for ERDA are several MeV for the detection of very light elements (mainly hydrogen), whereas for the analysis of a wider mass range of elements higher projectile energies are favorable. Higher projectile energies also increase the information depth. The largest projectile energies used in ERDA are hundreds of MeV for measuring the light-element composition in surface layers. Most of the ion accelerators used are of the tandem van de Graaff type. The pressure in the scattering chamber should be held below approximately 10−4 Pa (high vacuum). Lower pressures in the UHV range are usually not necessary, because ERDA is not surface-sensitive. The energies of recoil atoms and scattered projectiles are usually measured with solid-state SBDs. In order to identify particles with the same energy but different atomic numbers, an additional quantity (TOF, ΔE, or Ne) must be measured along with the energy. Usually, both quantities (energy and the identification quantity) are then stored in a two-dimensional (2-D) multichannel analyzer [14]; only for the simplest case (ERDA with a stopper foil) can a one-dimensional (1-D) multichannel analyzer be used. A high efficiency can be achieved by using detectors with a large solid angle. In this case, the recoils are measured with recoil angles in an interval, Δθ, around a mean recoil angle θ. The different recoil angles will lead to different recoil energies; this effect – termed kinematic line broadening – can be minimized by the use of a position-sensitive energy detector. The energies for particles with recoil angles θ + Δθ can then be corrected to the nominal energy at the recoil angle θ [3, 24]. The exit angle, β, and the angle of incidence, α, of the beam (Figure 13.1) determine the depth resolution and information depth. Small angles increase the depth resolution, but reduce the depth probed. In order to optimize both quantities, recoil angles θ between 30 ° and 45 ° are often used, with α = β = θ/2.
13.6 Sensitivity and Depth Resolution
13.5 Data Analysis
For particle identification, individual energy spectra of the different recoil elements and of scattered projectiles are obtained from the stored 2-D multichannel analyzer data. For a given recoil element of a given energy, the corresponding emission depth is deduced from the energy, and the concentration of the element obtained from the number of particles detected. A simple evaluation of the energy spectra can be made by dividing the target into thin slabs, and then assuming a constant energy of the projectiles within one slab to calculate the recoil crosssections (including deviations from the Rutherford values for light elements) and the energy losses [2]. A more precise and rapid evaluation of the individual energy spectra can be carried out by using available computer programs, which simulate the energy spectra of recoils and scattered projectiles for a given target composition and experimental arrangement. In these programs, SIMNRA [25] or DEPTH [26], nonRutherford cross-sections can also be used. The multiple scattering of incident projectiles and recoils in the target can also be taken into account in the DEPTH program. Multiple scattering becomes important for analysis of deeper layers in ERDA geometries with small recoil angles and low energies. In order to make the evaluation procedure both simple and efficient, the energy spectra for a variety of target compositions may be simulated and compared to measured spectra, the aim being to determine which target composition provides the best agreement between the simulated and measured spectra.
13.6 Sensitivity and Depth Resolution
Reasonable estimates of ultimate sensitivity and depth resolution in ERDA are difficult to make, due to the large range of projectiles and energies (from He ions of several MeV up to 200-MeV Au ions), and the use of different detection systems. In addition, the stability of the sample under irradiation (which, of course, depends on the target material) is also important when discussing sensitivity and detection limits. The sensitivity is determined mainly by the recoil cross-section, the solid angle of the detector system, and the total number of incident projectiles which can be used for analysis. For H profiling with light-MeV projectiles, a mean detection limit of approximately 1014 hydrogen atoms cm−2 in a near-surface region can be deduced from measurements of implanted H in Si [27]. Similar detection limits are obtained for the analysis of light impurities (C, N, and O) by 200-MeV Au projectiles [3]. By using position-sensitive detectors, it is possible to obtain large solid angles and, consequently, a high sensitivity with minor deterioration of the depth resolution due to kinematic broadening [3, 24].
223
224
13 Elastic Recoil Detection Analysis (ERDA)
TOF detector systems usually have smaller solid angles and sensitivity than (ΔE − E) systems, due to the long TOF system in front of the energy detector and the limited size of the stop detector. They also have worse detection limits for very light elements (hydrogen), due to the low probability of obtaining start and stop signals for particles of very low atomic number [22]. The depth resolution of ERDA is mainly determined by the energy resolution of the detector system, the scattering geometry, and the type of projectile and recoil. The depth resolution also depends on the depth analyzed due to energy straggling and multiple scattering. When the relative importance of different contributions to the depth resolution were studied for some specific ERDA arrangements [11, 13], the depth resolution at the surface was found to be determined mainly by the energy resolution of the detector, the recoil geometry, and the type of projectile. In contrast, in deeper layers the depth resolution was mainly defined by energy loss straggling and multiple scattering of projectiles and recoils. The optimum resolution is obtained at the surface. For H-profiling with MeV He ions, using a standard SBD, approximately 40 nm depth resolution is obtained [27]. For 5 MeV Ne projectiles impinging on a polymer sample, a depth resolution down to 8 nm was obtained for surface hydrogen in an arrangement optimized for depth resolution [13]. For heavy-ion projectiles and (ΔE − E) telescope detectors, depth resolutions of the order of several nanometers can be obtained in surface layers [3]. For layers as deep as several 100 nm, the resolution decreases to approximately 50 nm [28]. TOF systems with good timing signals enable good depth resolution of few nanometers, when the energy spectra are obtained from the timing signals [29]. TOF systems have the advantage of a larger information depth for most elements compared to (ΔE − E) systems [30]. Extremely high depth resolution corresponding to atomic layer resolution can be obtained, if the ion beam and detector system are optimized. Monolayer resolution has, for example, been realized for 60 MeV iodine projectiles impinging on HOPG graphite, when using a special magnetic spectrograph for energy determination [31].
13.7 Applications
As an example of depth profiling, hydrogen implanted into Si is shown in Figure 13.2 [27]. Here, the measured energy spectra of H recoils are shown for the impact of 6-MeV C ions. H identification was achieved with the (ΔE − E) technique and the use of ion-induced electron emission [14], and the energy scale was converted to a depth scale for comparison with simulated spectra. In order to analyze the data, the ERDA spectra were simulated by the code SIMNRA, for which an H depth profile calculated by the TRIM program was used as input. In Figure 13.2, the surface contamination peak can clearly be separated from the implanted H peak. Spectra are provided for different recoil angles, and the depth resolution can be estimated from the width of the surface peak. At low recoil angles, the depth resolution is better, whereas at large recoil angles the sensitivity is higher. A good
13.7 Applications
Figure 13.2 H depth profile of an H-implanted Si sample obtained with 6-MeV C projectile ions for different recoil angles θ. q gives the charge of the incident ions. The
experimental depth profiles (full line) are compared with simulated spectra (dashed line = SIMNRA, dotted line = DEPTH) [27].
Figure 13.3 Two-dimensional spectrum showing dependence of TOF on energy for a multilayer sample and impact of 120 MeV Kr ions [21].
agreement was obtained between the measured and calculated depth profiles. The number of implanted H atoms was found by this measurement to be 1.35 × 1016 atoms cm−2. As a second example, results from a TOF-ERDA measurement for a multielement sample are shown in Figure 13.3 [21]. The sample consists of different
225
226
13 Elastic Recoil Detection Analysis (ERDA)
Figure 13.4 Energy spectra of O and Al, calculated from the measured TOF of Figure 13.3.
Experimental results are compared with a simulation (SIMNRA) [21].
metal–metal oxide layers on a boron silicate glass, and the projectiles were 120MeV Kr ions. Clearly, many different recoil ions can be separated from the most intense line, produced by the scattered projectiles. Figure 13.4 shows the energy spectra for O and Al recoils calculated from the measured TOF spectra, together with simulated spectra using the SIMNRA code. The concentration and thickness of the O and Al layers are obtained from the simulations.
References 1 L’Ecuyer, J., Brassard, C., Cardinal, C., Chabbat, J., Deschenes, L., and Labrie, J. (1976) J. Appl. Phys., 47, 381. 2 Barbour, J. and Doyle, B. (1995) Elastic recoil detection: ERD, in Handbook of Modern Ion Beam Analysis (eds J. Tesmer and M. Nastasi), Materials Research Society, Pittsburgh, p. 83. 3 Assmann, W., Huber, H., Steinhausen, C., Dobler, M., Glückler, H., and Weidinger, A. (1994) Nucl. Instrum. Methods B, 89, 131. 4 Habraken, F. (1992) Nucl. Instrum. Methods B, 68, 181.
5 Bozoian, M., Hubbard, K., and Nastasi, M. (1990) Nucl. Instrum. Methods B, 51, 311. 6 Baglian, J., Kellok, A., Crockett, M., and Shih, A. (1992) Nucl. Instrum. Methods B, 64, 469. 7 Kido, Y., Miyauchi, S., Takeda, O., Nagayama, Y., Sato, M., and Kusao, K. (1993) Nucl. Instrum. Methods B, 82, 474. 8 Quillet, V., Abel, F., and Schott, M. (1993) Nucl. Instrum. Methods B, 83, 47. 9 Kim, C., Kim, S., and Choi, H. (1999) Nucl. Instrum. Methods B, 155, 229.
References 10 Andersen, H., Besenbacher, F., Loftager, P., and Moller, W. (1980) Phys. Rev. A, 21, 1891. 11 Nagata, S., Yamaguchi, S., Fujino, Y., Hori, Y., Sugiyama, N., and Kamada, K. (1985) Nucl. Instrum. Methods B, 6, 533. 12 Genzer, J., Rothman, J., and Composto, R. (1994) Nucl. Instrum. Methods B, 86, 345. 13 Ermer, H., Pfaff, O., Straub, W., Geoghean, M., and Brenn, R. (1998) Nucl. Instrum. Methods B, 134, 237. 14 Steinbauer, E., Benka, O., and Steinbatz, M. (1998) Nucl. Instrum. Methods B, 136, 695. 15 Behrooz, A., Headrick, R., Seiberling, L., and Zurmühle, R. (1987) Nucl. Instrum. Methods B, 28, 108. 16 Elliman, R., Timmers, H., Ophel, T., Weijers, T., Wielunski, L., and Harding, G. (2000) Nucl. Instrum. Methods B, 161–163, 231. 17 Stoquert, J., Guillaume, G., Hagi-Ali, M., Grob, J., Ganter, C., and Siffert, P. (1989) Nucl. Instrum. Methods B, 44, 184. 18 Anold Bik, W., deLaat, C., and Habraken, F. (1992) Nucl. Instrum. Methods B, 64, 832. 19 Wielunski, M., Mayer, M., Behrisch, R., Roth, J., and Scherzer, A. (1997) Nucl. Instrum. Methods B, 122, 133. 20 Thomas, J., Fallavier, M., and Ziani, A. (1986) Nucl. Instrum. Methods B, 15, 443. 21 Bohne, W., Röhrich, J., and Röschert, G. (1998) Nucl. Instrum. Methods B, 136–138, 633.
22 Zhang, Y., Whitlow, H., Winzell, T., Bubb, I., Sajavaara, T., Arstila, K., and Keinonen, J. (1999) Nucl. Instrum. Methods B, 149, 477. 23 Johnston, P., El Bouanani, M., Stannard, W., Bubb, I., Cohen, D., Dytlewski, N., and Siegele, R. (1998) Nucl. Instrum. Methods B, 136–138, 669. 24 Timmers, H., Elliman, R., Palmer, G., Ophel, T., and O’Connor, D. (1998) Nucl. Instrum. Methods B, 136–138, 611. 25 Mayer, M. (1997) Technical Report IPP9/113, Max-Planck-Institut für Plasmaphysik, Garching (Germany). 26 Szilagyi, E., Paszti, F., and Amsel, G. (1995) Nucl. Instrum. Methods B, 100, 103. 27 Bogdanovic-Radovic, I., Steinbauer, E., and Benka, O. (2000) Nucl. Instrum. Methods B, 170, 163. 28 Elliman, R., Timmers, H., Palmer, G., and Ophel, T. (1998) Nucl. Instrum. Methods B, 136–138, 649. 29 Kim, Y., Kim, J., Choi, H., Kim, G., and Woo, H. (1998) Nucl. Instrum. Methods B, 136–138, 724. 30 Grigull, S., Kreissig, U., Huber, H., and Assmann, W. (1997) Nucl. Instrum. Methods B, 132, 709. 31 Dollinger, G., Frey, C., Bergmaier, A., and Faestermann, T. (1998) Nucl. Instrum. Methods B, 136–138, 603.
227
229
14 Nuclear Reaction Analysis (NRA)1) Oswald Benka
14.1 Introduction
Nuclear reaction analysis (NRA) is a technique for the analysis of surface layers using light ion beams with energies in the range 1 to 10 MeV. When an ion “a” hits an atom “A,” it might cause a nuclear reaction. An intermediate excited nucleus I* is then formed, which decays to the ground state “B” by the emission of a particle “b.” The emitted particle “b” may be a γ photon, or an ion (H+, He2+), and will have a characteristic energy determined by the Q-value of the reaction. Such a reaction is described in an abbreviated form as A(a,b)B. For a constant number of incident projectiles the number of emitted reaction products “b” depends on the number of target atoms “A”; therefore, it is possible to derive the number of atoms “A” present in an analyzed sample from the measured number of reaction products “b.” If the mean life-time of the excited state I* is reasonably long, the reaction products can be measured after irradiation, and this technique is termed charged-particle activation analysis (by analogy with neutron activation analysis). If the mean lifetime is small, or emission is prompt, the reaction products are measured during ion bombardment; this technique is then termed “nuclear reaction analysis”. As a projectile which hits a target nucleus must overcome the Coulomb barrier to excite a nuclear reaction, only light projectiles with energies in the MeV range are used in NRA for the analysis of light elements, taking advantage of the low Coulomb barrier. The probability of a nuclear reaction, which is given by the corresponding cross-section, depends on the energy and type of projectile and on the type of target atom, and there is no simple means of predicting its magnitude. For proton and He projectiles with energies in the MeV range, few elements have 1) In most nuclear reactions A(a,b)B, the emitted radiation “b” used for analysis consists of charged particles (e.g., α-particles and protons). This is why NRA is covered in this section on ion detection. The emitted radiation can, however, also consist
of γ-rays or neutrons. In accordance with the chosen structure of this book, we ask for the reader’s understanding of our omission of an additional NRA chapter dealing with these γ-ray- and neutron-producing reactions.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
230
14 Nuclear Reaction Analysis (NRA)
nuclear reactions with reasonable probability for analysis. Therefore, NRA is not a technique to probe the composition of an unknown sample by use of high-energy ions, such as particle-induced X-ray emission (PIXE) or Rutherford backscattering spectroscopy (RBS). Rather, NRA is a powerful technique for quantifying the concentration or depth profile of a certain element in a sample for which the qualitative chemical composition is known. In addition, NRA offers a sensitive means of measuring the depth profiles of individual isotopes of an element, which cannot be ascertained by using techniques such as RBS, AES, or XPS. As the cross-sections for nuclear reactions are usually lower than those for the elastic scattering of projectiles used in RBS or in elastic recoil detection analysis (ERDA), higher currents must be used to obtain comparably high intensity in NRA as in RBS or ERDA. In addition, a possible modification of the target composition as a result of irradiation must be considered. Nuclear reaction cross-sections are also usually not available in analytical form for the direct evaluation of measured data. Concentrations are, therefore, often obtained by comparison of the measured data with results from standard samples of known concentration. The experimental requirements for NRA can be divided roughly into two groups, depending on the emitted reaction products – gamma photons or particles. If the emitted γ photons are used for analysis, this method is also referred to as particleinduced gamma emission (PIGE). In this method, the gamma detector that measures the intensity of the photons is usually outside the scattering chamber. Only the concentration of a specific isotope in a thin layer can be deduced from the number of photons. To obtain information also concerning the depth distribution, the number of gamma photons must be measured as a function of the beam energy, E0, for an equal number of impinging projectiles; this is termed the excitation curve N(E0). For the analysis of emitted particles, solid-state surface barrier detectors (SBDs) are used inside the scattering chamber to measure the number and energy of the reaction products. Stopper foils are used to prevent scattered projectiles from reaching the detector. Depth profiles can be obtained from the energy spectra, since the reaction products emitted in deeper layers will have less energy than those emitted from the surface. The concentration in the corresponding layer can be determined from the intensity of reaction products with a certain energy. The cross-section curve σ(E) indicates the dependence of the nuclear crosssection on the projectile energy, E. The measured energy spectra of emitted particles or the excitation curve N(E0) will depend on the depth profile N(x) of the analyzed isotope, and also on the cross-section curve σ(E(x)), where E(x) gives the energy of the projectiles at a depth x. Evaluation of the depth profile N(x) from measured energy spectra or excitation curves often requires a tedious evaluation procedure, if the cross-section curve has a complex structure. However, it can be simplified for two special types of behavior of the cross-section curve:
•
If σ(E) is constant – that is, independent of the projectile energy in an energy interval ΔE – the measured energy spectrum of the reaction products directly reflects the concentration profile in a depth interval corresponding to ΔE.
14.2 Principles
•
If the cross-section curve has a strong resonance at an energy ER – that is, it is large in a small energy interval around ER and negligibly small for all other energies – the measured excitation curve will directly reflect the concentration profile.
An excellent review has been produced on the various techniques of NRA [1]; reviews of depth profiling with narrow resonances are also available [2, 3].
14.2 Principles
For a nonresonant nuclear reaction with emission of an ion, a depth scale can be obtained from the measured energy of the emitted ions. If ions emitted from depth x are lower in energy by ΔE than ions emitted from the surface, then a relationship between ΔE and x can be identified, similar to RBS and ERDA analysis: ΔE = xNε nr
(14.1)
where N is the atomic density of the sample and εnr is the nuclear reaction stopping cross-section factor, given by:
ε nr = αε in +
ε out cos ϕ
(14.2)
when the projectiles impinge perpendicularly on the surface and the reaction products are emitted at an angle φ relative to the surface normal. εin and εout are the stopping cross-sections for the incoming projectiles and the emitted ions. The reaction factor α weights the energy loss of the incident projectiles, in the same way as the kinematic factor in RBS, with:
α=
d Er dEp
(14.3)
where Ep is the projectile energy when it produces a nuclear reaction and Er is the energy of the reaction product. The reaction factor α can be calculated from the reaction kinematics [1], and depends on the mass of projectile, the reaction product, the reacting atom, the Q-value of the reaction, and the angle of emission. For light nuclei and backward emission, α can become negative; in this case the contributions of deeper layers extend the spectrum towards higher energies, and the depth resolution may be worsened. For slowly changing reaction cross-sections, the depth profile of the analyzed isotope can be calculated by using the reaction cross-sections (by analogy with RBS), for which the scattering cross-section is the corresponding quantity. In order to obtain a simple estimate of the depth profile, the target can be divided into thin slabs, with constant cross-section and stopping power within each slab. Here, the standard procedure is to simulate energy spectra for different target compositions, by using a computer program [4, 5], and to deduce the depth profile by the comparison of simulated and measured spectra.
231
232
14 Nuclear Reaction Analysis (NRA)
When, in NRA, energy spectra of emitted particles are analyzed, a sufficiently thick foil in front of the detector is normally used to absorb the scattered projectiles. This reduces the depth resolution of NRA, because of energy loss straggling of the reaction products in the foil. When, in NRA, resonances are used, and the depth profile of an isotope A is obtained from the excitation curve N(E0), the reaction depth x is given by the requirement that projectiles incident with energy E0 are slowed down to the resonance energy ER at x, which leads to: E0
x=
∫ N(m ε
ER
1 dE ( ) (1 − m )ε B(E )) E + A
(14.4)
for a sample with atomic density N and composition AmB(1−m), where m is the atomic fraction of the isotope A and εA and εB are the stopping cross-sections of the constituents A and B. For m << 1 and for thin layers, the stopping crosssections can be assumed to be constant, and Equation 14.4 reduces to: x = (E 0 − ER )
1 Nε
(14.5)
where ε is the stopping cross-section of the sample at the mean projectile energy, (E0 + ER)/2. The easiest way to obtain the concentration m of the isotope A is to use a reference sample (“standard”), that contains the isotope A with a known atomic fraction f, for comparison. Knowing the NRA yield of the standard YSt and its stopping cross-section εSt, the concentration m in the sample can easily be evaluated from the yield YA of the sample for the same projectile energy, taking the different stopping cross-sections into account: m=
f YA ε B YSt εSt + f YA (ε B − ε A )
(14.6)
For m << 1, Equation 14.6 can again be approximated by: m= f
YA ε YSt εSt
(14.7)
where ε is the stopping cross-section of the sample.
14.3 Equipment and Depth Resolution
Nuclear reactions are excited when projectile energies are typically in the MeV range. Medium size ion-accelerators are, therefore, necessary to obtain these projectile energies. Protons and α projectiles, which are typical projectiles in other ion-beam analysis techniques as RBS or PIXE, have few useful nuclear reactions. Deuteron beams excite many more nuclear reactions, but their use instead of standard beams is more hazardous due to efficient neutron
14.3 Equipment and Depth Resolution
production. Strict safety rules are necessary when high-energy deuteron beams are used. The main requirement for accelerators used in resonance NRA are reasonably good energy resolution and the possibility of changing the beam energy easily. The beam energy is usually increased stepwise by adjusting the magnetic field used to select the energy and to stabilize the terminal voltage. More sophisticated energy-scanning systems [6, 7] have, however, been developed to change the beam energy while keeping the analyzing magnet constant. The nuclear reaction products are usually measured in a high-vacuum scattering chamber. At resonance NRA, where the beam energy is varied, the number of incident projectiles must be known at each energy. A good beam-current measurement and integration system is necessary to determine the total incident beam charge. In NRA depth profiling using the energy spectra of ion reaction products, an absorber foil is normally used to filter out the scattered projectiles [1]. However, the main disadvantage of this method is that the reaction products will suffer energy straggling when they pass the foil, resulting in a degradation of the depth resolution. The main limiting factors of depth resolution are:
• • •
The energy resolution of the detector. Kinematic broadening due to the solid angle of the detector. Energy straggling of the incident beam and of the reaction products in the sample and in the absorber foil [8].
Which of these contributions dominates depends on the nuclear reaction, the energy of the projectiles, the analyzed depth, and the geometry of the equipment used. The depth resolution in resonance NRA close to the surface is mainly determined by the stopping power of the projectiles in the target. For deeper layers, there are different contributions:
• • • •
energy loss straggling; energy resolution of the beam; resonance width; and Doppler broadening due to vibrations of the target atoms [1–3].
Doppler broadening is usually a small contribution, which becomes important only for the nuclear reactions of heavy projectiles and light target atoms [3, 9] . Close to the surface of the sample, energy loss straggling can be neglected and for narrow resonances a high-depth resolution can be obtained for beams with good energy resolution (typically 100 eV). The corresponding depth resolutions for proton and He projectiles are in the range of several nanometers [1]. H depth profiling is best achieved with a 15N beam which has a resonance at 6.5 MeV [1, 3]. For H depth profiles close to the surface, the depth resolution is limited by Doppler broadening; the corresponding depth resolution in Si is approximately 25 nm [1, 9]. The depth resolution in deeper layers is mainly determined by energyloss straggling. Depth resolution in NRA is, therefore, similar to that for the other high-energy ion-beam profiling techniques, such as RBS.
233
234
14 Nuclear Reaction Analysis (NRA)
In resonance NRA with gamma-emission, the emitted gamma rays are usually measured outside the scattering chamber by means of an NaI or a Ge(Li) detector. NaI scintillation counters are used when high efficiency is needed and energy resolution is not critical. This is often the case when few light elements are present in the sample and only one well-isolated resonance is measured. BGO (bismuth– germanium oxide) detectors are similar to NaI detectors, but are more efficient [10]. Solid-state Ge(Li) detectors are used when high gamma-energy resolution is necessary to distinguish between adjacent gamma lines of nuclear reactions from different target elements. If an element present at a low concentration is to be analyzed, then a high sensitivity is required, and in order to achieve this the background radiation must be minimized [11]. This can be achieved by careful shielding of the gamma detector, and also by reducing the gamma rays produced by the beam in the beam line, for example, on beam collimators in front of the target.
14.4 Applications
The major application of NRA is determination of the concentration of selected light elements as a function of depth in thin surface layers. The accessible depth is typically several microns, with a depth resolution of approximately 10 to 100 nm. Useful nuclear reactions, together with their applications, have been reviewed [1].
Figure 14.1 RBS spectrum for impact of 3 MeV protons and NRA proton spectrum for impact
of 2 MeV deuterons. The full line gives the fit result (SIMNRA) of a four-layer simulation. The thickness is expressed in 1015 atoms cm−2 and the element content in atom% [13].
14.4 Applications
Figure 14.2 (a) Dependence of measured yield on ΔE for impact of (6.385 + ΔE) MeV N15
ions on an SiO2–Si sample and for different fluence intervals; (b) Deconvoluted H concentration profiles corresponding to panel (a) [14].
For protons with energies in the MeV range (p,γ), reactions for the isotopes 7Li, N, 18O, and 19F can be used for depth profiling, where only 7Li, and 19F are the most abundant isotopes of the natural elements. With MeV deuteron projectiles, many nuclear reactions can be used for NRA (for a review of this topic, see Ref. [12]). Deuterons, for example, have useful nuclear reactions with the isotopes 12C, 14 N, and 16O, which are abundant in natural elements and have no useful reactions with other projectiles. An example [13] is shown in Figure 14.1, where the constituent light elements in a Cu patina of an historical object were determined with an external 2 MeV deuteron beam. The proton energy spectrum from different (d,p) reactions was measured; the different reactions are denoted by the reacting target atom and by the number of proton reactions, according to different final states of the target atom; p0 corresponds to the transition with highest proton energy. The scattered deuterons were stopped in an absorber foil. In addition, an RBS spectrum was measured to analyze Cu using a 3 MeV proton beam. Figure 14.1 also shows simulated spectra fitted to the measured spectra, together with 15
235
236
14 Nuclear Reaction Analysis (NRA)
the concentration profiles, when the patina was separated into four layers of different thickness. Within each layer, constant concentrations of elements were assumed, and the layer thickness was expressed in 1015 atoms cm−2. NRA with gamma-emission is mainly used for the depth profiling of medium light isotopes, notably N, F, Na, Mg, Al, and Si. Some useful reactions, with detection limits, have been described by Elkes et al. [11]. An important application of NRA with gamma-emission is the determination of H depth profiles, as hydrogen is probably the most common elemental contamination in thin-film materials, and is invisible to many analytical methods. Although, in principle, all inverse p-induced nuclear reactions can be used, the 15N reaction is most often employed. Whilst this has a strong narrow resonance at 6.385 MeV (laboratory energy), the main problem is to obtain a good 15N ion beam, because the 15N abundance in natural N is below 1% and N does not form a stable negative ion as is necessary in tandem accelerators. An application [14] of H depth profiling is shown in Figure 14.2, where the gamma yield as a function of projectile energy is shown in Figure 14.2a. Here, the projectile energy Ep = Er + ΔE, where Er is the resonance energy (6.385 MeV) and ΔE is the abscissa. The sample is an SiO2/Si layer structure, and gamma yield measurements are given for different amounts of N irradiation. Figure 14.2b shows the deconvoluted hydrogen concentration profiles corresponding to the measurements of Figure 14.2a. It can be seen that an increasing 15N fluence, Φ, degrades the original shape of the H profile: H is enriched to a depth of 173 nm, which corresponds to the depth of the SiO2/Si transition region.
References 1 Tesmer, J. and Nastasi, M. (eds) (1985) Handbook of Modern Ion Beam Materials Analysis, Materials Research Society, Pittsburgh, pp. 139–204. 2 Maurel, B., Amsel, G., and Nadai, J. (1982) Nucl. Instrum. Methods, 197, 1. 3 Lanford, W. (1992) Nucl. Instrum. Methods B, 66, 65. 4 Vickridge, I. and Amsel, G. (1990) Nucl. Instrum. Methods B, 45, 6. 5 Mayer, M. (1997) Technical Report IPP9/113, Max-Planck-Institut für Plasmaphysik, Garching (Germany). 6 Amsel, G., d’Artemare, E., Batistig, G., Girard, E., Gosset, L., and Reveresz, P. (1998) Nucl. Instrum. Methods B, 136–138, 545. 7 Meier, J. and Richter, F. (1990) Nucl. Instrum. Methods B, 47, 303. 8 Sagura, A., Kamada, K., and Yamaguchi, S. (1988) Nucl. Instrum. Methods B, 34, 465.
9 Hartmann, B., Kalbitzer, S., and Behar, M. (1995) Nucl. Instrum. Methods B, 103, 494. 10 Kuhn, D., Rauch, F., and Baumann, H. (1990) Nucl. Instrum. Methods B, 45, 252. 11 Elkes, Z., Kiss, A., Gyürky, G., Somorjai, E., and Uzonyi, I. (1999) Nucl. Instrum. Methods B, 158, 209. 12 Vickridge, I. (1988) Nucl. Instrum. Methods B, 34, 470. 13 Ioannidou, E., Bourgarit, D., Galligaro, T., Dran, J., Dubus, M., Salomon, J., and Walter, P. (2000) Nucl. Instrum. Methods B, 161–163, 730. 14 Ecker, K., Krauser, J., Weidinger, A., Weise, H., and Maser, K. (2000) Nucl. Instrum. Methods B, 161–163, 682.
237
15 Field Ion Microscopy (FIM) and Atom Probe (AP) Yuri Suchorski and Wolfgang Drachsel
15.1 Introduction
Information concerning the structure and chemical identity of surface and subsurface species on an atomic scale is essential for surface chemistry. The application of structure-sensitive surface analyses, such as scanning tunneling microscopy (STM), atomic force microcopy (AFM) and related tools provides in situ surfacevisualization [1, 2], but suffers from the principle-caused limitations such as a relatively poor time-resolution. In turn, STM-based element identification methods, such as scanning atom probe (SAP, [3]) do not yet allow operation under realistic conditions. In turn , the parallel imaging techniques based on the point projection principle (Figure 15.1a), such as field ion microscopy (FIM), field emission microscopy (FEM) and lithium field desorption microscopy (Li-FDM), allow the in situ imaging of dynamic surface processes on an atomic-level scale, with a time-resolution that is limited only by the data collection rate [5, 8]. These techniques utilize those particles [e.g., electrons in FEM, metal (Li) ions in Li-FDM, or gas ions in FIM] that are radially emitted from the surface in a high electrostatic field. The sample studied using the above-described techniques is the apex of a needle-shaped specimen (the field emitter tip). This exhibits a heterogeneous surface formed by differently oriented nanofacets, which allows it to serve as a suitable model for a catalytic particle of comparable dimensions (Figure 15.1c). However, in contrast to a catalyst particle, the tip surface can be prepared in a reproducible manner by field evaporation, and subsequently characterized with atomic resolution by imaging in the FIM. This technique, as well as the FEM and Li-FDM can be used then to visualize in situ catalytic reactions such as CO oxidation or NO reduction on the platinum metal nanofacets laterally resolved on the nanoscale [9–11]. One serious limitation of field emission-based microscopies has been their inability to identify the chemical identity of the individual surface atoms. However, a breakthrough was achieved in 1967 by using a combination of a probe-hole FIM with a mass spectrometer having single particle sensitivity. This process was named by its inventor, E.W. Müller, as atom probe field ion microscopy (APFIM) Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
238
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
(a) polarized gas atom accommodated gas atom (111)
field adsorbed gas atom
+ zc F ri
(b)
field ion
t
(100)
screen
r
R
imaging particle specimen (tip) spot imaged surface site
(c) (111)
(100)
(d)
z e· F·
Ve
Ι Φ
Ev(F >0)
Fermi sea zcrit
Figure 15.1 Principles of the field emission-
based microscopies. (a) Geometry of the point projection. The surface of the specimen apex is projected on a screen by imaging particles with a magnification M ≈ R/r. Imaging particles are electrons (in FEM), imaging gas ions (in FIM), or Li+ ions (in Li-FDM). The top inset shows a ball model of the apex of a spherical fcc tip [4], and illustrates the mechanism of ion formation in
z
Ev(F = 0)
FIM. Imaging gas ions are generated over the protruding surface atoms covered by field adsorbed gas atoms. Field-ionized atoms are replaced by highly mobile accommodated gas atoms (see text); (b) Ne+ field ion image of a [111]-oriented Rh field emitter tip [6]; (c) Metal aggregate on an oxide substrate [7]; (d) Potential energy diagram for field ionization over a metal surface.
[12]. During FIM operation, an atomic site of interest could be selected by placing it (by adjusting the position of the tip) over a probe hole in the image screen. Pulsed-field evaporation was then used to extract the chosen surface atom and pass it through the hole into the spectrometer. Subsequent time-of-flight (TOF) spectroscopy allowed the measurement of one-dimensional (1-D) “in-depth” compositional maps (atom probe, AP). A three-dimensional (3-D) nanometer-scale reconstruction of a sample composition was then made possible by moving the sample relative to the probe hole. Later, the need to move the tip was avoided by developing the imaging atom probe (IAP) [13, 14], where the ions emitted from the surface were analyzed using a two-dimensional (2-D) detector. The application
15.2 Principles and Instrumentation
of short laser pulses (as first applied by Drachsel [15], Kellogg, and Tsong [16]), in combination with the local electrode atom probe (LEAP®, [17]) led to the atom probe becoming the ultimate analytic tool, where the 3-D position of each detected specimen atom and its identity – including all of the isotopes – could be resolved. In this way, the first part of the ingenious prediction of the principal inventor of these techniques, E.W. Müller, was realized to a full extent: “Further experimental work will be centered around two objectives: One is the straightforward application of the atom-probe FIM as a microanalytical tool of ultimate sensitivity. The chemical identity and the location with respect to the lattice structure of impurities, segregations, precipitates, and alloy constituents are immediate goals” [18]. Mainly due to the efforts of research groups at Oxford University, the University of Rouen, and Imago Scientific Instruments, commercial devices were developed and produced, and are today widely used in materials science [19, 20]. The second part of Müller’s prediction: “The second aim of atom-probe research will be to shed new light on the complex situation at specific atomic sites of the surface. … where the surface-gas interaction data carry a new dimension of reliability by the identification of the particles involved” [18], that is, the application of AP techniques to in situ surface chemistry and catalysis, has, unfortunately, been realized only partially, despite significant experimental effort having been made during the past decades. Correspondingly, the relevant literature has been dominated by pure materials science applications such as metallurgy. In turn, surface chemistry applications of FIM/AP instruments as well as other atom-probe like techniques combined with FIM such as pulsed field desorption mass spectrometry (PFDMS, [21]) and field ion appearance energy spectroscopy (FIAES, [22, 23]) have attracted much less attention. The aim of this chapter is to account, at least partially, for this imbalance and to discuss surface chemistry and analytics, together with conventional AP applications.
15.2 Principles and Instrumentation 15.2.1 Field Ion Microscopy
The field ion microscope is a point projection microscope, invented by E.W. Müller in 1951 [24], which permitted for the first time true atomic resolution that enabled the visualization of individual atoms as early as 1955 [25]. Since then, FIM has contributed significantly to surface sciences, and in particular to materials research
239
240
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
[26, 27], notably with regard to studies of the behavior of individual atoms on metal surfaces [28, 29]. Whilst a vast number of reports have been dedicated to FIM, in the following subsections the principles are only briefly summarized. Corresponding applications in surface chemistry are detailed in Section 15.3. In FIM, a sharp specimen tip (Figure 15.1a) with a radius of curvature in the sub-micrometer range is positioned a few centimeters away from an imageintensifier (usually a micro-channel-plate/phosphor-screen assembly). The tip is exposed to a high positive voltage in the presence of an imaging gas at 10−3–10−5 mbar. In a sufficiently high field (tens of V nm−1) the atoms or molecules of the imaging gas (usually He, Ne, or a reactive gas such as oxygen) become polarized, and are attracted to the specimen surface where they accommodate at cryogenic temperatures, losing their kinetic energy in a thermal accommodation process [30]. The accommodated gas atoms follow the local field gradient across the surface, arrive at local field maxima above the prominent surface atoms and become (in the case of the normally used He or Ne as imaging gas) field-adsorbed onto protruding surface atoms. The following accommodated gas atoms become steadied over those field-adsorbed species, whereby the local field distribution plays a decisive role [31–35]. Steadied gas atoms, being free and at rest, are then field-ionized by electron tunneling through the field adsorbed atoms into the specimen [30]. A collision-induced exchange process between the field-adsorbed and “arriving” accommodated gas atoms is also possible shortly prior to the fieldionization event [35]. At the instance of field ionization, the gas atom is localized at the critical distance zcrit (see inset in Figure 15.1a) to the specimen surface. This distance is necessary to allow lifting of the atom’s valence electron level over the Fermi level of the specimen by the applied field. Tunneling at distances smaller than zcrit is not possible, as this would mean a tunneling into occupied states below the Fermi level (Figure 15.1d). The gas ions formed at (or beyond) zcrit are accelerated in the applied field projecting in the narrow ion beams the surface atoms on a screen (Figure 15.1b). An extremely high degree of spatial localization of the imaging gas atoms at the instant of field ionization provides atomic resolution in the FIM images. The essential role of the field-adsorbed layer of the imaging gas atoms for atomic resolution in FIM is illustrated for Ne gas atoms ionized over a Rh tip in Figure 15.2 [34]. By increasing the temperature of the tip, or enhancing the field over the best image field (BIF), the field-adsorbed layer is depleted and the following observations for noble imaging gas atoms such as Ne or He are made:
• • •
•
The field ion image becomes blurred. The field ionization rate at the probed surface sites decreases drastically. The value of field ion appearance energy (FIAE) derived from the field ion retardation curves decreases and becomes equal to the ionization energy of the free gas atom (the meaning of the FIAE is discussed in Section 15.2.3). The full-width at half-maximum (FWHM) of the field ion energy distribution increases significantly.
15.2 Principles and Instrumentation
Figure 15.2 Role of the field-adsorbed Ne atoms in localized field ionization. (a) Ne field ion image of a [111]-oriented Rh emitter kept at 79 K and at 35 V nm−1 (field-adsorbed Ne layer is present). The area encircled in white represents the probe-hole. The locally measured Ne+ retardation curve N(δ) and the differentiated curve, dN/dσ are shown on the right; (b) The same as in panel (a), at 107 K (field-adsorbed Ne layer is depleted). It is observed that: (i) the field ion image
becomes blurred; (ii) the field ionization rate at the probed surface sites decreases drastically; (iii) the value of field ion appearance energy derived from the field ion retardation curves decreases and becomes equal to the ionization energy of the free gas atom; and (iv) the full-width at halfmaximum of the field ion energy distribution increases significantly (from 1.45 V to 2.08 V). Adapted from Ref. [34].
241
242
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
A wide range of conducting samples can be studied using FIM, notably metals including alloys, semiconductors, and even oxides [27–31]. The natural limit of FIM applicability is the field evaporation process (i.e., field desorption of the specimen lattice atoms), in which the surface atoms of the specimen are field-ionized and removed from the surface. This process is used for controlled fine shaping and cleaning of the tip, although it can also lead to destruction of the specimen when the imaging field is equal to or higher than the onset of field evaporation of the particular sample. The values of the evaporation field strength have been listed [30]. For details of the field desorption and field ionization processes the reader is referred elsewhere [29–31, 35–37]. FIM operation with reactive imaging gases, such as O2, CO, and N2, occurs mainly by field desorption of the adsorbed gas molecules, which are subsequently replaced by a diffusion supply from the tip shank where the gas molecules are adsorbed [38–40]. Thus, a FIM with a reactive imaging gas is basically a field desorption microscope. It is also possible to obtain images with metal ions that have been field-emitted directly from the specimen surface. In both possible cases – whether imaging by the field evaporation of specimen atoms used as imaging species, or by field desorption where the adsorbed metal atoms are exploited – the corresponding technique is termed field desorption microscopy (FDM) [41]. Whereas, in the first case the imaging process is rather instable due to evaporation of the specimen during the visualization [42], in the second case – and especially when using Li+ ions supplied by diffusion from the specimen shank – stable images comparable to those obtained with FIM can be obtained [43]. 15.2.2 Time-of-Flight Atom Probe Techniques
As noted in Section 15.1, an AP is a combination of a FIM system and a mass spectrometer capable of single ion detection. The sample is imaged with atomic resolution by FIM, and the selected surface area is positioned over the probe-hole in the screen (Figure 15.3). The surface atoms are field-evaporated from the entire tip surface by nanosecond high-voltage pulses, and those atoms projected into the probe-hole of a few millimeters in size (corresponding to few nanometers on the specimen surface) are detected and identified by TOF mass spectrometry (timeof-flight atom probe, TOFAP). The inherent shortcoming of such an AP is the limited mass resolution, m/Δm, of about 250 due to the energy deficits of fieldevaporated ions caused by varying the field during the pulsed high-voltage evaporation process [44]. However, this can be significantly improved by using energy-compensating lenses (Poschenrieder- or reflectron-type [45, 46]); indeed, such energy-compensated devices are the most widely used AP instruments, mainly for metallurgical applications. Another AP version with a magnetic sector field for mass-to-charge separation of ions (magnetic sector field atom probe, MSAP, Figure 15.4) was proposed in 1968 [47] and improved in 1974 [48]. Such devices are widely used for field-ion energy distribution measurements [49] and for FIAES [22, 23] (see below).
15.2 Principles and Instrumentation
Figure 15.3 Principles of the atom probe. (a) A straight TOF configuration. The specimen surface is imaged by FIM, field ion micrographs are recorded with a CCD-camera. Surface atoms are field-evaporated (with or without pulsed laser irradiation), and those passing through probe hole are identified by TOF measurement. For surface reaction studies, reactive gases are introduced, surface species are field-desorbed with laser pulse assistance, and local ion rates are mass separated by a Wien filter and recorded with a multichannel analyzer (MCA). Adapted from Ref. [44]; (b) Local electrode atom
probe (LEAP®) configuration. A substrate supporting microtip specimens is placed at μm distance to a cone-shaped local electrode, and can be laterally translated relative to the local electrode (scanning mode). Both, electrical and laser pulsing can be applied. The position-sensitive detector consists of a micro-channel plate (MCP) combined with a delay-line detector. The MCP provides TOF data and the delay-line detector reads out the x- and y-values of each arriving atom. From these data, the atomic 3-D composition of the specimen can be reconstructed.
In the conventional 1-D-AP, the elemental mapping of the tip surface can be obtained by consecutive changing the tip position laterally to the screen, scanning quasi the tip surface by the projection of the probe hole. This laborious procedure may be avoided, however, by using the imaging atom probe (IAP), as developed by Panitz in 1973 [13, 14]. In this device, single ions are field-evaporated from tip surface and detected without a probe hole by a detector consisting of two spherical
243
244
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
Figure 15.4 Principle of FIAES. (a) Mass-to-
charge resolved retarding potential analysis. Field-desorbed surface species (e.g., CO+ ions) laterally selected by a probe-hole are separated using magnetic sector field, pass a retarding potential analyzer, and are detected by a single-particle detector. Adapted from Ref. [8]; (b) Energy diagram. The appearance energy A1hkl represents the total energy of a
singly charged ion (e.g., Ne+) created over a particular (hkl) surface site at the critical distance zcrit. Value EF is the Fermi energy of the emitter (tip) and the retarder (collector), respectively, the potential difference Ur – Ue provides δon, the onset voltage at which the ion is just collected. The inset shows schematically a retardation curve N(δ). Adapted from Ref. [35].
microchannel plates (chevron design) placed relatively near to the tip (∼10 cm); this allows identical radial flight distances from any position on the tip surface to the detector. The instrument can be operated either as a conventional FIM or in TOF mode, where time gating (switching-on the detector gain only during the arrival time of the selected species), in combination with video-recording of the visual signal, permits elemental mapping of the surface. The use of laser pulses of nanosecond duration to initiate the photodissociation of adsorbed molecules on the tip surface [15, 50, 51] led to the introduction of the pulsed laser atom probe (PLAP) [16]. In this instrument, the laser-assisted field evaporation is exploited to remove single atoms from the tip surface. In this mode, the high field applied to the specimen is held constant on a level which is insufficient for field evaporation at the present (cryogenic) temperature. The laser pulse causes a steep increase in the surface temperature, suddenly rendering the applied field sufficient for field evaporation. By using this approach, a mass resolution of 4000 was achieved with 300 ps laser pulses [52].
15.2 Principles and Instrumentation
During the past few years, the availability of laser systems capable of producing pulses shorter than 1 ps with high intensity (<1012 W m−2) has led to their stimulated application in PLAP [53], and also to intensive discussions regarding the mechanism of pulsed laser evaporation [54–56]. The need to consider direct photoionization, and the role of the electric field of the photon pulse was noted, whereas the degree of the contribution of these effects to the evaporation mechanism has not yet been established. Despite the lack of full understanding of the details of laser pulsed evaporation, the first commercial PLAPs with ultrashort laser pulsing became available in 2006 (produced by Imago Scientific Instruments Corporation), and these have been widely used since that time [19]. The AP methods described above are basically destructive; that is, complete surface layers are field-evaporated during the operation, even when only a selected nanometer-sized area is analyzed in depth in a TOFAP. Thus, a “true” 3-D analysis of the specimen is not possible, neither with TOFAP nor with IAP instruments, as only a small fraction of the total number of atoms is analyzed. For a “true” 3-D analysis, the detection of the TOF and of the position of each single atom field evaporating from the surface is necessary. Although the first suggestion to perform this task, by combining the fiber optics of an IAP with photomultipliers, was made in 1975 [57], the procedure has become technically possible only since the 1980s. The corresponding technique of 3-D-AP is also referred to as atom probe tomography (APT), in analogy to conventional tomography, which produces 3-D images from consecutive slices. The different versions of APT vary by their differing types of position-sensitive single ion detectors, whereas a relative charge either on a wedge-and-stripe (position-sensitive AP, PoSAP [58, 59]), or on an array of anodes (tomographic AP, TAP [60, 61]), can be measured. Another approach is based on optical recording, either directly from the phosphor screen by using a CCDcamera (optical AP, OAP) [62, 63], or combining this with a linearly segmented anode (optical tomographic AP, OTAP) [64]. Today, the experimental challenges caused by difficulties in the detection of the near-simultaneous arrivals of ions (the problem of “multiple hits”) are resolved by use of delay-line detectors, which provide a 1000 × 1000 pixel resolution at ion rates up to 500 000 counts per second [19]. The problems associated with the collecting and processing of 3-D data, as well as the corresponding solutions, have been comprehensively reviewed by Kelly and Miller [65] and by Seidman [66]. These articles also include detailed descriptions of the specimen preparation techniques required, including those for focused-beam methods. The next significant improvement in the AP technique resulted from an idea to microminiaturize the counterelectrode, and to place it in immediate proximity (μm-scale) to the specimen [67], as in the SAP [3]. This configuration allowed a reduction in the applied voltages, and the pulse repetition rates up to 250 kHz at a pulse duration of 2 ns could be achieved. In this instrument (LEAP [19]) the ions, which had been field-evaporated from the specimen, passed the local electrode within tens of picoseconds, and thus were much less influenced by temporal variations of the field. As a result, a mass resolution could be achieved sufficient to separate the individual isotopes of all the elements [19]. A combination of the
245
246
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
LEAP design (that provides a high mass resolution, a high data collection rate, and a wide field of view) with laser pulsing appears to be the most progressive AP concept to date. Details of other commercial developments of laser-pulsed instruments, and recent improvements in instrumentation (e.g., the large-angle reflection compensator, LARCTM [68]), are provided by the respective manufacturer source [19] or in recent reviews [65, 66]. One particular version of a TOF-based AP method is that of pulsed-field desorption mass spectrometry (PFDMS), which is used mainly in surface chemistry studies. The method and instrumentation for this technique were systematically developed by the research group of Block at the Fritz Haber Institute, Berlin, in order to target the kinetics of surface reactions. As for the AP, in PFDMS field ions produced by high-field pulses are analyzed using TOF-mass spectrometry, the difference being that the high field necessary for field desorption is applied not as an additional short pulse to a tip being at high positive potential (as is usual in a conventional AP), but rather as an entire voltage pulse (up to 50 kV) applied to a counterelectrode [69, 70]. This means that, during the time between pulses, the processes on the surfaces may occur field-free, thus avoiding or minimizing any field effects on the reaction process. Moreover, by applying a varying positive bias to the tip, the field effects can be studied in a controlled manner [71, 72]. Another important peculiarity of PFDMS is the small distance (≤0.1 mm) between the counterelectrode and the specimen. This allows, in combination with a negative pulse application to the counterelectrode, the detection of weakly bound species with still reasonable energy deficits [21a]. In this sense, the PFDMS – to a certain degree – was seen as the forerunner of a local AP where negative pulses could also be applied to the counterelectrode placed at μm-distance from the specimen. 15.2.3 Field Ion Appearance Energy Spectroscopy
Whilst the above-described techniques confirmed Müller’s visions of the AP as a microanalytical tool of ultimate sensitivity, in surface chemistry it is not only the location(s) and identity of specimen atoms that are of interest, but also the surface–gas interaction data, such as binding energies. As shown at an early stage by Müller and Bahadur [73], thermodynamic information regarding the energetics of field ionization processes at the surface can be derived by measuring the onset of the field ion energy distribution. Historically, FIAES instrumentation has grown from the technique of field ion mass spectrometry (FIMS) [74], in which field-desorbed surface species with different masses are selected by probe-hole and mass-to-charge ratio, and analyzed by using a magnetic sector field. Corresponding instruments are often similar to the MSAP, devices and differ solely by the presence of an energy analyzer (Figure 15.4a), although linear arrangements (without a magnetic field) are also possible [22, 23]. In an FIAES instrument with a magnetic mass-to-charge separation, a 60 ° magnetic sector field is normally used, ending in a retarding potential analyzer with a single particle detector. The current of ions reaching the detector is then measured as a function of the voltage δon, applied between the retarder electrode
15.2 Principles and Instrumentation
and emitter to vary the height of the retarding electrostatic barrier. The appearance energy A, as introduced by Goldenfeld et al. [75], is derived for a singly charged ion: A1 = Fret − ed on
(15.1)
where Φret is the work function of the retarder, e is the charge of the ion, and δon is the onset voltage for ion collection (δon = Ur – Ue; see explanation in Figure 15.4b). In principle, A1 is the energy to remove an electron of a given molecule (ground state) to create an ion at its surface position. By conducting a closed thermionic cycle, the appearance energy for a field-desorbed molecule amounts to [76]: A1 = I + H(Floc ) − Q (Floc )
(15.2)
if temperature-dependent terms are neglected. In this expression, I represents the ionization potential of the ionized species, H(Floc) is the adsorption energy at the local field, and Q(Floc) the activation energy for field desorption of this species. An extended discussion of the energy terms involved in field ion appearance spectroscopic measurements is provided in Refs [22, 23]. The concept of the field ion appearance energy is closely related to the term critical energy deficit (as the atom/ molecule will be ionized at a certain distance from the emitter surface, zcrit, it does not experience the whole acceleration potential of the emitter; i.e., it has an energy deficit) [31]. Crucial to the use of FIAES in surface chemistry was the possibility to calibrate the work function of the retarder in situ, for example, from the onset voltage for rare gas ions (e.g., Ne) by using Equation 15.2. Unfortunately, it was not possible initially to measure the absolute A1 values inclusive of the field-dependent terms; rather, these terms were first measured during the 1990s [23]. It was also found that, at elevated temperatures (T > 100 K), when the field-adsorbed imaging gas layer was depleted, the field-dependent terms for noble gas atoms such as Ne or Ar did not contribute essentially (thus, A1 ≡ I) [23, 34]. Since that time, the work function of the retarder has been determined routinely in situ from the known ionization potential of a noble test gas as Φret = eδtest gas + Itest gas. By replacing the test gas with a reactive gas (such as CO), its appearance energy could be obtained from the calibrated Φret and measured δCO values. The first successful measurements of such type date back to 1995 [77]. By using Equations 15.1 and 15.2, the field dependence of H(F) – Q(F) can be obtained from the measured δon values. The Q(F) value can be estimated from the temperature dependence of the ion rate, thus the binding energy H(F) from chosen atomic surface sites is available. This procedure can also be used during a reaction to provide in situ data regarding the binding energy of the reactants. Also, aside of the surface chemistry, the method can be used to determine local electrostatic fields over individual atoms [23], to distinguish between the localized and delocalized field ionization (see Figure 15.2, [6]), and to determine the binding energy of metal adsorbates [78] and/or even the sublimation energy of metals by directly measuring the appearance energy of field-evaporated atoms [79]. These rather physical applications have been summarized in two reviews [35, 36], while details of the surface chemistry applications are discussed in Section 15.3.2.1.
247
248
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
15.3 Applications 15.3.1 FIM Applications
The most important applications of FIM can be found in materials science, especially in metallurgy, where the segregation of impurities, grain–boundary segregation, the decomposition of alloys, the role of lattice defects, and other surface effects have been studied on an atomic scale [26, 27, 29–31]. Independently, a major contribution to the surface sciences has been accomplished by FIM, especially in studying the diffusion processes and behavior of individual atoms on metal surfaces at a much earlier stage [28–31, 80] than when such studies became routine for STM [81, 82]. The first evidences of a substrate-mediated indirect interaction of adatoms were obtained by FIM for Re adatoms on W(110) [83], as well as for W–Pd and Re–Pd and Re–Pd atom pairs on W(110) [84, 85]. A compilation of this early development, including the dipole–dipole and elastically mediated nonoscillatory interactions, has been comprehensively reviewed [86]. The first experiments on the directional walking of adsorbed atoms in an externally applied field gradient [87], and on the field desorption of single adatoms by a voltage pulse [88], were the direct predecessors of atom manipulation in STM [89]. 15.3.1.1 FIM in Catalysis In turn, the main applications of FIM in surface chemistry and catalysis were established not before the “STM-era,” but rather simultaneously with corresponding STM applications, when it was realized that a parallel imaging principle had essential advantages for visualizing surface reactions. This was subsequently and convincingly demonstrated via photoemission electron microscopy (PEEM) studies on the macroscopic (∼μm) scale. Various spatial–temporal phenomena as target patterns, island formation, spirals, standing waves, and so on, were observed, and have been summarized in reviews [90–92]. Since the early 1990s, it has been shown that such phenomena can be also imaged in situ by using FIM/FEM, and with nanometer-resolution using platinum-metals tip specimens. Many such results have been recently reviewed [8]. Some important questions concerning the surface reactions can be answered via FIM/FEM studies:
• • •
The role of the applied electric field on the reaction mechanism [93]. The influence of the structural heterogeneity of the field emitter tip surface (dynamic coupling effects between the individual planes) [94]. The role of fluctuations on the reaction kinetics and possible fluctuationinduced deviations from the behavior, as predicted by macroscopic (meanfield) rate laws [95, 96].
To prove experimentally the effect of the imaging electrostatic field (∼4 V·nm−1 in FEM, ∼12 V nm−1 in FIM), the kinetic phase diagram of CO oxidation on Pt was obtained by FEM and FIM for the same sample. In FIM, the different ionization
15.3 Applications
probabilities for oxygen (used as the imaging gas) over the oxygen- and CO-covered surface [97] and in FEM the local work function variations created the imagecontrast, respectively. This allowed identification of the catalytically active state, as well as measurement of the kinetic phase diagram by monitoring the local field emission current (or image brightness) while adjusting the external parameters such as pCO, pO2, and T. Figure 15.5a shows the kinetic phase diagrams obtained with both, FIM and FEM, for the same Pt tip and the same oxygen partial pressure of 4 × 10−4 Torr. Since additional FEM measurements performed with pulsed high-voltage supply and varying the duty cycles factor showed no influence of the duration of the pulses on the catalytic CO oxidation [93], the FEM results can be interpreted as quasifield-free. In turn, the phase diagram obtained by FIM appeared as shifted significantly to lower pCO values in comparison to the field-free one. The physical reason for this effect was revealed by FIAES measurements, and is described in the following subsection. Figure 15.5b illustrates the local and “global” oscillatory behavior of the reaction on the tip apex shown in Figure 15.5a. Although different facets are exhibited, the whole tip area undergoes synchronous transitions from a CO-covered to an oxygencovered surface, and vice versa, at the transition points (bifurcation points) of the bistability region (compare the variation of the local and total image intensity in Figure 15.5b). The comparison of FEM measurements with data for catalytic CO oxidation on a macroscopic Pt(110) surface reveals the influence of the structural heterogeneity of the field emitter tip surface: the single crystal studies by Eiswirth et al. (open circles in Figure 15.5a; [98]) and Moldenhauer (filled circles; [99]) suggest the position for the bifurcation point, which is significantly shifted with respect to the FEM results. These differences can be directly attributed to the size of the system and to dynamic coupling effects of the {110} facets with neighboring orientations. 15.3.1.2 Fluctuation-Induced Effects Nanofacets on the specimen apex exhibit a prime example of nanosized systems where a significant influence of fluctuations is potentially expected, because of the 1/N1/2 (where N is the number of fluctuating particles, e.g., reacting molecules) scaling of the relative amplitude of fluctuations. As the numbers of reacting particles on a nanofacet varying the range 102 to 103, deviations from the behavior predicted by macroscopic rate laws may be observed, such as fluctuation-induced transitions in a bistable system [100]. The “parallel” imaging principle of the FEM/ FIM makes these techniques naturally suitable for monitoring fluctuations: the processes proceeding on different surface regions can be analyzed simultaneously, with the time resolution being limited simply by varying the recording speed of the video-technique used. Digitizing and sophisticated processing of the corresponding video-images allow the reaction dynamics within the so-called “virtual probe-holes” (regions of interest which are selected on digitized video-images and which correspond to chosen surface regions of a few nm2) arbitrarily located on the surface to be studied [101].
249
250
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
Figure 15.5 (a) Kinetic phase diagrams for
CO oxidation on Pt nanofacets determined with FEM (F0 = 4 V nm−1) and FIM (F0 = 12 V nm−1). The shaded areas mark the existence ranges for the oscillations. Below their crossing point, the boundary lines for instabilities enclose a region of bistability. The open circles are Pt(110) single crystal data indicating the existence range for oscillations; the bistability range of reaction on Pt(110) at T = 349 K is indicated by filled circles. The inset shows the geometry of the
[111]-oriented Pt tip used in these experiments (clean Pt surface imaged at 78 K with FIM using Ne as imaging gas, F0 = 35 V nm−1); the central (111) and one of three {110} planes are indicated. Adapted from Ref. [93]; (b) Oscillatory behavior of CO oxidation on the Pt tip apex displayed in the inset of panel (a). Upper curve: total FEM image intensity; lower curve: local FEM intensity of the (110) plane. T = 398 K, pCO = 6.2 × 10−6 Torr, pO2 = 4.3 × 10−4 Torr. Reproduced with permission from Ref. [93].
15.3 Applications
Figure 15.6 Fluctuation-induced transitions in catalytic CO oxidation on a Pt field emitter tip. Time series of the local (2 × 2 nm2) FEM brightness (panel a) and respective
probability distributions (panel b). The inset shows a FEM image with marked rectangular area where the local FEM brightness was analyzed. Adapted from Ref. [95].
Figure 15.6 shows the local time series recorded for a small 2 × 2 nm2 area in the vicinity of the (110) facet of a Pt-tip. The small amplitude fluctuations caused mainly by the CO diffusion, and characterized by a Gaussian probability distribution, were observed in the monostable range on the inactive branch of the reaction. This is illustrated by the time series a (Figure 15.6a) and a corresponding probability distribution (Figure 15.6b). Similar time series, but with a somewhat higher magnitude of fluctuations, were also observed in the monostable range on the active branch (curves b in Figure 15.6). In the bistable range of the reaction, the probability distribution becomes broad and asymmetric, and turns even into a bimodal probability distribution (curves c and d in Figure 15.6, correspondingly). This bimodal distribution evidences the fluctuation-induced transitions between the two possible states in the bistability region [95]. More detailed studies using the Haar–Wavelet analysis detected that such fluctuation-induced transitions would occur spatially correlated over the flat facet surface (the above-mentioned region in the vicinity of Pt(110) was identified as a small (331) facet) [102], and are confined by the first atomic step due to the disturbed diffusional coupling over the stepped surface. It should be noted that this effect cannot be predicted within the mean-field theory. The Monte-Carlo simulations, which naturally incorporate the stochastic nature of the modeled processes, can in turn predict such an effect, provided that the modeled system is sufficiently small and is in close proximity to a bifurcation (critical) point where bistability vanishes and fluctuations diverge [103]. The noise-induced transitions in the CO oxidation detected first on the Pt nanofacets allowed an explanation of the vanishing bistability during CO oxidation on a Pd model catalyst consisting of small Pd particles [104]. For the model system
251
252
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
consisting of particles of ∼500 nm in size, a pronounced bistability is observed which vanishes when the particle size becomes smaller than 6 nm (approximately the size of an individual facet of the field emitter tip). The fluctuation-induced transitions, which occur in an unsynchronized way, lead to the apparent disappearance of the bistability as observed by an averaging method (the total CO2 rate for many particles was measured in Ref. [104]). Another promising FIM application is to monitor the surface diffusion by the “virtual probe-hole” technique. This concept, that the local image brightness fluctuations in the recorded FIM (FEM or Li-FDM) video images can provide the surface diffusion parameters, is based on Onsager’s hypothesis regarding the applicability of macroscopic diffusion laws to the microscopic density fluctuations [105]. It also takes into account the finding that such brightness fluctuations reflect microscopic concentration deviations in an apparently homogeneous adsorbed layer, and can thus be analyzed via the density fluctuation method, as proposed by Kleint and Gomer [106, 107]. Digital image analysis permits the utilization of the full potential of the “parallel”- imaging principle of FIM (as of any field emission-based microscope), and allows the simultaneous analysis of processes on any chosen surface regions of only a few nm2 in size, in contrast to the “real probe-hole” measurements, where the emission from just one or two regions (in a “two probe-holes” experiment [108]) can be measured. In recent applications of the “virtual probe-hole,” the anisotropy of Li diffusion on the spatially inhomogeneous Pt surface (along and across atomic steps) was directly proven, and a slowing effect of the coadsorbed CO on Li diffusion was observed [5]. 15.3.2 Applications of AP Techniques 15.3.2.1 Applications of TOF-AP Techniques Standard up-to-date TOF-AP applications in the materials sciences are mainly related to the atom-by-atom 3-D reconstruction of a specimen. These applications are also important for catalysis, as knowledge concerning the surface and subsurface composition will allow insight into the details of the reaction mechanism, and especially into the role of subsurface species such as oxygen, which are especially important for catalytic reactions. The application of short laser pulses (<10 ns) also opens up additional possibilities for in situ studies of “volatile” short-living adspecies (with a residence time <1 μs), such as H3O(+) in the hydrogen oxidation reaction. Figure 15.7 shows the results of such a study during the H2 oxidation on a Pt surface. These results allowed the detection and explanation of the additional field-induced reaction routes in this reaction. Further details of this study are available elsewhere [8]. Recently, a 3-D-AP system was combined with a versatile gas reaction cell, thus enabling the high-temperature (≤873 K) handling of metal specimens in reactive gases mixtures under atmospheric pressure (∼1 bar). After reactive treatment, the specimens could be immediately transferred into a UHV chamber of an energycompensated 3-D-AP equipped with a delay line detector [109]. The first results
15.3 Applications
Figure 15.7 Laser-induced desorption and TOF AP study of H2 oxidation on Pt. Field
dependence of the ion yield for H3O·H2O+ (䉱), H3O+ (•), H2O+ (䊏), O2 (䉲). A TOF spectrum measured at 19 V nm−1 is shown in the inset. Reproduced with permission from Ref. [8].
obtained in such a combined instrument, which was termed the catalytic atom probe (CAP), allowed the detection of chemical segregation in Pt−Rh alloys caused by NO, O2, and N2O exposition [110]. It should be noted here that such lateral transport processes within the first few atomic layers could barely be detected and quantified by other techniques. The further application of this approach to ternary Pt−Rh−Ru and Pt−Rh−Ir alloys has revealed how heavily the resulting surface composition after gas–surface interaction depends on both, the exact mix of elements and the surface crystallography [111]. Applying the same instrument to a Au−Pd alloy in reactive environments (NO), the authors of Ref. [112] were able to govern the surface enrichment with Pd over a wide range by chemical/thermal treatment, thus producing a desired Au/Pd ratio on the surface. The surface designed in this way could then be assessed in a catalytic reaction. The AP technique presents unique possibilities in studying the initial stages of metal oxidation. The pioneering studies in this area were conducted by Kellogg [113], who investigated the nucleation and growth stages of oxide formation on Rh specimens, using imaging AP. A quantitative AP analysis of samples heated in oxygen (at 1 Torr) showed that stoichiometric Rh2O3 is formed at temperatures as low as 500 K, while Rh suboxides are formed at lower temperatures. The activation energy for the process was determined from the temperature dependence of the surface oxygen accumulation rate to be 4–5 kcal mol−1, which was in good agreement with later measurements [114]. These measurements also provided an explanation of the mechanism of self-sustained oscillations in CO oxidation on Rh [115]. Recently, the new generation of 3-D-AP (which allow both voltage and laser pulsing) was applied to study the oxidation mechanisms of Zr [116]. Surprisingly,
253
254
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
ZrO1−x was detected at the initial stages of oxidation (ZrO2 is the only known stable oxide of zirconium). Additionally, the segregation of hydrogen to the metal–oxide interface and a distinct ZrH phase were observed in this study. The study of kinetics of the room-temperature oxidation of zirconium showed that the ZrO layer seems to be nonprotective over a longer time (up to 1 h). Again, these results could barely be obtained with any other current surface-analysis technique. 15.3.2.2 PFDMS Applications The kinetic reaction studies using PFDMS are based on the observation of the formation of adsorbed layers from the gas phase during the time tR between any two consecutive field pulses (Figure 15.8). Varying tR over a range from 100 μs to a few seconds allows the kinetics of the adsorption and reaction processes (the
Figure 15.8 Evaluation of kinetic data from
PFDMS measurements. (a) Applied field pulses with different repetition frequencies (b) corresponding evolution of surface
concentrations of the impinching species and (c) reaction products. Relaxation times τ for the processes in the adsorbed layer can be obtained. Illustration courtesy of N. Kruse.
15.3 Applications
initial stages of adsorption prevail at short tR, the reaction intermediates and products at longer tR) to be monitored. A prime example of the power of PFDMS in surface chemistry is the formation of Ni(CO)x during the reaction of CO with Ni. By varying tR between 10−4 and 10−1 s it was found that, at 3 × 10−6 mbar and at 300 K, Ni(CO)2 and Ni(CO)3 require an approximately 10-fold longer period to achieve their constant coverages than the (refilling) CO adlayer (10 and 1 ms, correspondingly); in other words, the rate-determining step of carbolynation is associated with Ni(CO)2 formation [21, 117]. The final product Ni(CO)4 undergoes, in turn, immediate thermal desorption without being readsorbed onto the Ni surface. These insights allowed the formulation of a reaction mechanism that involved the formation of Ni(CO)2 by adding a terrace-adsorbed CO to an Ni kink site covered by CO: Ni(CO)kink + COad ↔ Ni(CO)2,ad + *kink, accompanied by breaking of a Ni−Ni bond at kink sites. The formation of Ni(CO)2,ad is followed by the rapid formation of Ni(CO)3,ad and finally, of Ni(CO)4, that is immediately desorbed. This reaction route appears to be generally valid for stepped metal surfaces: PFDMS results with Rh, Ru, Pd, and Ni have demonstrated the formation of mobile Me(CO)x,ad [21, 117–119]. Another interesting example is a PFDMS study of the kinetics of thermal desorption of NO from a stepped Pt(111) nanofacet [69]. In this case, the variation of tR allowed determination of the dwell time of NOad at different temperatures, while the temperature dependence of the dwell time permits, in turn, a determination of the activation energy of thermodesorption from a stepped Pt(111) surface. The obtained value of 139 kJ mol−1 was in a good agreement with results obtained with molecular beam relaxation spectroscopy [120, 121]. Such an agreement demonstrates the applicability of PFDMS to the thermally equilibrated adlayers. Due to short field pulses and the absence of a steady field, the field effect seems to be rather negligible in such studies [122, 123]. In turn, an applied steady field would allow a direct insight into field-dependent processes, such as field-dependent methanol decomposition on Rh, where the field-stabilized intermediates were observed. 15.3.2.3 FIAES Applications The main strength of FIAES is the ability to obtain locally the binding energy of adsorbed atoms or molecules on chosen surface sites, by directly measuring the appearance energy of the corresponding field ions [78, 124]. The binding energy of Li adatoms on W(111) was, for example, obtained from the energy analysis of locally desorbed Li+ ions [125], and in good agreement with known thermodesorption data [36]. The binding energy values of Rh/Rh(110) obtained in this way agreed well with the sublimation energy of Rh [79]. The application of this method to the reactants of a surface reaction allows not only the determination of their binding energy, but also the field-dependence of this energy. This, in turn, can shed light on the role of perturbances of local electron density distributions as they occur, for example, on the oxide/metal interface in supported catalysts [126]. Figure 15.9 shows the experimentally measured field-dependences of the binding energy for the CO and O2 molecules on chosen step sites of the Pt(111) nanofacet
255
256
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
Figure 15.9 FIAES of reactants in CO
oxidation on Pt. (a) O2+ retardation curves measured at 11.9 V nm−1 from O-covered Pt(111) step sites at T = 79 K. The dashed line indicates the ionization energy for a free CO molecule. The Ar retardation curve used for calibration of Φret is also shown; (b) Data
for H-Q as a function of applied field F0. The upper (lower) results have been obtained from CO+ (O2+ ) field ion appearance energies measured at 79 K for stepped Pt-surfaces covered with COad (Oad). Reproduced with permission from Ref. [127].
References
[127]. The plots show that the binding energy of adsorbed CO increases quite strongly with the electric field strength, whereas adsorbed O2 displays a relatively weak dependence on the applied field, in agreement with the self-consistent calculations by Kreuzer and Wang [77]. These results provide an explanation of the field-effect in the CO oxidation on Pt, as described in Section 15.3.1.1. The field effect on the binding strength of adsorbed species results from the additional charge transfer caused by a field-induced redistribution of the electronic charge at the surface; the “positive” field applied in FIM reduces the electronic density outside the geometric surface plane, as compared to the field-free case [35]. The changes in the binding energies of the molecular adsorption states of O2 and CO on Pt shift the adsorption/desorption equilibria of the molecular O2 species and of CO. Enhanced binding strength of CO modifies the supply rate of CO into the reaction zone, whereas the considerably lower binding energy of O2 molecules leads to an insignificant additional supply of O2 molecules under reaction conditions. Both effects, taken together, lead to a shift of the FIM diagram towards smaller pCO values in comparison to the (quasi-field-free) FEM-diagram. It should be noted, that weakly bound (field-stabilized) highly mobile CO species detected with FIAES under UHV conditions are apparently present on the catalyst surface under high-pressure conditions revealed by the comparison of high-field and high-pressure studies [128]. In this way, the high-field experiments allow the mimicking of high-coverage adsorption states on individual metal nanoparticles of a catalyst, thus bridging (at least partially) both the materials and the pressure gap.
References 1 Wintterlin, J., Völkening, S., Janssens, T.V., Zambelli, T., and Ertl, G. (1997) Science, 278, 1931. 2 Hendriksen, B.L.M., Bobaru, S.C., and Frenken, J.W.M. (2005) Top. Catal., 36, 43. 3 Nishikawa, O. and Kimoto, M. (1994) Appl. Surf. Sci., 76/77, 424. 4 Herrmann, K. (2010) Gallery of Surfaces. Available at: http://www. fhi-berlin.mpg.de/th/personal/ hermann/pictures.html (accessed 3 November 2010). 5 Suchorski, Y., Beben, J., Frac, A., Medvedev, V., and Weiss, H. (2007) Surf. Interface Anal., 39, 161, and references therein. 6 Suchorski, Y., Schmidt, W.A., and Block, J.H. (1993) Appl. Surf. Sci., 67, 124. 7 Freund, H.-J., Bäumer, M., Libuda, J., Kuhlenbeck, H., Al-Shamery, K., and Hamann, H. (1998) Cryst. Res. Technol., 33, 977.
8 Suchorski, Y. and Drachsel, W. (2007) Top. Catal., 46, 201 and references therein. 9 Gorodetski, V.V., Drachsel, W., and Block, J.H. (1993) Catal Lett., 19, 223. 10 Medvedev, V.K., Suchorski, Y., and Block, J.H. (1995) Surf. Sci., 343, 169. 11 Voss, C., and Kruse, N. (1995) Appl. Surf. Sci., 87/88, 127. 12 Müller, E.W., Panitz, J.A., and McLane, S.B. (1968) Rev. Sci. Instrum., 39, 83. 13 Panitz, J.A. (1974) Rev. Sci. Instrum, 44, 1034. 14 Panitz, J.A. (1978) Progr. Surf. Sci., 8, 219. 15 Drachsel, W., Nishigaki, S., and Block, J.H. (1980) Int. J. Mass Spectrom. Ion. Phys., 32, 333. 16 Kellog, G.L. and Tsong, T.T. (1980) J. Appl. Phys., 51, 1184. 17 Kelly, T.F., Kamus, P.P., Larson, D.J., Holzman, L.M., and Bajikar, S.S. (1996) Ultramicroscopy, 62, 29.
257
258
15 Field Ion Microscopy (FIM) and Atom Probe (AP) 18 Müller, E.W. (1970) Naturwissenschaften, 67, 222. 19 IMAGO Scientific Instruments Corporation, Madison, WI. Available at: http://www.imago.com/ (accessed 10 September 2009). 20 Cameca Sa, Paris, France. Available at: http://www.cameca.fr (accessed 10 September 2010). 21 (a) Kruse, N. (1999) Mater. Sci. Eng. A, 270, 75; (b) Kruse, N. (2001) Ultramicroscopy, 89, 51. 22 Ernst, N. (1993) Appl. Surf. Sci., 67, 82. 23 Schmidt, W.A., Ernst, N., and Suchorski, Y. (1993) Appl. Surf. Sci., 67, 101. 24 Müller, E.W. (1951) Z. Physik, 131, 136. 25 (a) Müller, E.W. (1956) Z. Naturforsch., 11a, 88; (b) Müller, E.W. (1956) J. Appl. Phys. 27, 474. 26 Miller, M.K., Cerezo, A., Hetherington, M.G., and Smith, G.D.W. (1996) Atom Probe Field Ion Microscopy, Oxford University Press, Oxford. 27 Seidman, D.N. (1976) Radiation Damage in Metals (eds N.L. Pterson and S.D. Harkness), Am. Soc. Metals, Metals Park, OH. 28 Ernst, N. and Ehrlich, G. (1986) Microscopic Methods in Metals (ed. U. Gonser), Springer-Verlag, Berlin. 29 Tsong, T.T. (1988) Rep. Progr. Phys., 51, 759. 30 Müller, E.W. and Tsong, T.T. (1969) Field Ion Microscopy, Principles and Applications, American Elsevier, New York. 31 Tsong, T.T. (1990) Atom-Probe Field Ion Microscopy, Field Ion Emission and Surfaces and Interfaces at Atomic Resolution, Cambridge University Press, Cambridge, New York. 32 (a) Rendulic, K.D. (1971) Surf. Sci., 28, 285–298; (b) Rendulic, K.D. (1973) Surf. Sci., 34, 581. 33 Ernst, N. and Block, J.H. (1982) Surf. Sci., 117, 561. 34 Schmidt, W.A., Suchorski, Y., and Block, J.H. (1994) Surf. Sci., 301, 52. 35 Suchorski, Y., Schmidt, W.A., Ernst, N., Block, J.H., and Kreuzer, H.J. (1995) Progr. Surf. Sci., 48, 121. 36 Suchorski, Y., Ernst, N., Schmidt, W.A., Medvedev, V.K., Kreuzer, H.J., and
37 38
39 40
41 42 43
44
45 46
47 48 49 50
51
52 53
54
55
56
Wang, R.L.C. (1996) Progr. Surf. Sci., 53, 135. Gomer, R. (1994) Surf. Sci., 299/300, 129 and references therein. Suchorski, Y., Schmidt, W.A., Block, J.H., and Medvedev, V.K. (1995) Surf. Sci., 331/333, 277. Suchorski, Y. (1998) Ultramicroscopy, 73, 139. Suchorski, Y., Medvedev, V.K., and Block, J.H. (1996) Appl. Surf. Sci., 94/95, 217. Walko, R.J. and Müller, E.W. (1972) Phys. Status Solidi, A9, K9. Waugh, A.R., Boyes, E.D., and Southon, M.J. (1976) Surf. Sci., 61, 109. Medvedev, V.K., Suchorski, Y., and Block, J.H. (1994) Ultramicroscopy, 53, 27. Block, J.H., Dirks, J., and Drachsel, W. (1992) Rev. Lat. Am. Met. Mater., 11, 39. Müller, E.W. and Krishnaswamy, S.V. (1974) Rev. Sci. Instrum., 45, 1053. Drachsel, W.D., Alvensleben, L., and Melmed, A.J. (1989) J. Phys., 50-C8, 541. Barofsky, D.F. and Müller, E.W. (1968) Surf. Sci., 10, 177. Müller, E.W., and Sakurai, T. (1974) J. Vac. Sci. Technol., 11, 878. Ernst, N., Bozdech, G., and Block, J.H. (1979) Surf. Sci., 80, 645. Tsong, T.T., Block, J.H., Nagasaka, M., and Viswanathan, B. (1976) J. Chem. Phys., 65, 2469. Nishigaki, S., Drachsel, W., and Block, J.H. (1979) Surf. Sci., 87, 389. Liu, J., Wu, C.-W., and Tsong, T.T. (1991) Surf. Sci., 246, 157. Gault, B., Vurpillot, F., Bostel, A., Menand, A., and Deconihout, B. (2005) Appl. Phys. Lett., 86, 094101. Gault, B., Vurpillot, F., Vella, A., Gilbert, M., Menand, A., Blavette, D., and Deconihout, B. (2006) Rev. Sci. Instrum., 77, 043705. Vella, A., Vurpillot, F., Gault, B., Menand, A., and Deconihout, B. (2006) Phys. Rev. B, 73, 165416. Cerezo, A., Smith, G.D.W., and Clifton, P.H. (2006) Appl. Phys. Lett., 88, 154103.
References 57 Panitz, J.A. (1975) US Patent No. 3,868,507. 58 Cerezo, A. and Smith, G.D.W. (1990) European Patent No. 0,231,247. 59 Cerezo, A., Godfrey, T.J., and Smith, G.D.W. (1988) Rev. Sci. Instrum., 59, 862. 60 Bostel, A., Blavette, D., Menand, A., and Sarrau, A. (1989) J. Phys. Paris, 50, C8501. 61 Deconihout, B., Bostel, A., Menand, A., Sarrau, J.M., Bouet, M., Chambreland, S., and Blavette, D. (1993) Appl. Surf. Sci., 67, 444. 62 Miller, M.K. (1991) Surf. Sci., 246, 428. 63 Leisch, M. (1995) Appl. Surf. Sci., 87/88, 318. 64 Deconihout, B., Renaud, L., Da Costa, G., Bouet, M., Bostel, A., and Blavette, D. (1998) Ultramicroscopy, 73, 253. 65 Kelly, T.F. and Miller, M.K. (2007) Rev. Sci. Instrum., 78, 031101. 66 Seidman, D.N. (2007) Annu. Rev. Mater. Res., 37, 127. 67 Kelly, T.F., Camus, P.P., Larson, D.J., Holzman, L.M., and Bajikar, S.S. (1996) Ultramicroscopy, 62, 29. 68 Panayi, P. (2006) Great Britain Patent Application No. GB2426120A. 69 Kruse, N., Abend, G., and Block, J.H. (1988) J. Chem. Phys., 88, 1307. 70 Hess, T.R., Cocke, D.L., Abend, G., and Block, J.H. (1996) Appl. Surf. Sci., 94/95, 238. 71 Chuah, G.-K., Kruse, N., Block, J.H., and Abend, G. (1987) J. Phys., 48, C6–493. 72 Kruse, N. and Visart de Bocarmé, T. (2007) Handbook of Heterogeneous Catalysis (eds G. Ertl, H. Knözinger, J. Weitkamp, and F. Schüth), Wiley-VCH Verlag GmbH. 73 Müller, E.W. and Bahadur, K. (1956) Phys. Rev., 102, 634. 74 Block, J.H. and Chanderna, A.W. (1975) Methods and Phenomena (ed. A.W. Chanderna), vol. 1, Elsevier, Amsterdam, p. 379. 75 Goldenfeld, I.V., Korostyshewsky, I.Z., and Mischanchuk, B.G. (1974) Int. J. Mass Spectrom. Ion. Phys., 13, 297. 76 Forbes, R.G. (1976) Surf. Sci., 61, 221. 77 Schmidt, W.A., Suchorski, Y., Block, J.H., Kreuzer, H.J., and Wang, R.L.C. (1995) Surf. Sci., 326, 243.
78 Suchorski, Y., Medvedev, V.K., Block, J.H., Wang, R.L., and Kreuzer, H.J. (1996) Phys. Rev. B, 53, 4109. 79 Schmidt, W.A. and Ernst, N. (1994) Vacuum, 45, 255. 80 Ehrlich, G. (1966) J. Chem. Phys., 44, 1050. 81 Barth, J.V. (2000) Surf. Sci. Rep., 40, 75. 82 Hwang, I.S., et al. (2001) J. Phys. Chem. Solids, 62, 1655. 83 Tsong, T.T. (1973) Phys. Rev. Lett., 31, 1207. 84 Fink, H.-W., Faulian, K., and Bauer, E. (1980) Phys. Rev. Lett., 44, 1008. 85 Watanabe, F. and Ehrlich, G. (1991) J. Chem. Phys., 95, 6075. 86 Braun, O.M. and Medvedev, V.K. (1989) Sov. Phys. Usp., 32, 328. 87 Tsong, T.T. and Kellog, G.L. (1975) Phys. Rev. B, 12, 1343. 88 Tsong, T.T. (2006) Physics Today, March, 31–37 and references therein. 89 Eigler, D.M. and Schweizer, E.K. (1990) Nature, 344, 524. 90 Ertl, G. (2008) Angew. Chem. Int. Ed., 47, 3524–3535. 91 Ertl, G. (2008) Handbook of Heterogeneous Catalysis, 2nd edn (eds G. Ertl, H. Knözinger, F. Schüth, and J. Weitkamp), Wiley-VCH Verlag GmbH, Weinheim, pp. 1492–1516. 92 Ertl, G. (2002) Faraday Discuss., 121, 1. 93 Suchorski, Y., Imbihl, R., and Medvedev, V.K. (1998) Surf. Sci., 401, 392. 94 Gorodetskii, V., Lauterbach, J., Rotermund, H.H., Block, J.H., and Ertl, G. (1994) Nature, 370, 276. 95 Suchorski, Y., Beben, J., James, E.W., Evans, J.W., and Imbihl, R. (1999) Phys. Rev. Lett., 82, 1907. 96 Suchorski, Y., Beben, J., and Imbihl, R. (1998) Progr. Surf. Sci, 59, 343. 97 Kreuzer, H.J. and Wang, R.L.C. (1997) Z. Phys. Chem., 202, 127. 98 Eiswirth, M., Möller, P., Wetzel, K., Imbihl, R., and Ertl, G. (1989) J. Chem. Phys., 90, 510. 99 Moldenhauer, S. (1996) In-situ FT-IR Spektroskopie der katalytischen CO-Oxidation auf Platinmetallen, PhD Thesis, Freie Universität, Berlin. 100 Horsthemke, W. and Lefever, R. (1983) Noise-Induced Transitions: Theory and
259
260
15 Field Ion Microscopy (FIM) and Atom Probe (AP)
101 102
103
104
105 106 107 108 109
110 111 112
113 114
Applications in Physics, Chemistry, and Biology, Springer Series in Synergetics, 1st edn, vol. 15, Springer. Suchorski, Y. and Beben, J. (2003) Progr. Surf. Sci., 74, 3. Suchorski, Y., Beben, J., and Weiss, H. (2004) Symposium on Surface Science 3s’04, St. Christoph Am Arlberg, Austria, Contributions (eds F. Aumayr and P. Varga), p. 145. Suchorski, Y., Beben, J., Imbihl, R., James, E.W., Liu, D.-J., and Evans, J.W. (2001) Phys. Rev. B, 63, 165417. Johánek, V., Laurin, M., Grant, A.W., Kasemo, B., Henry, C.R., and Libuda, J. (2004) Science, 304, 1639. Onsager, L. (1931) Phys. Rev., 38, 2265108. Kleint, C.H. and Gasse, H.-J. (1960) Z. Naturforsch. A, 15, 87. Gomer, R. (1973) Surf. Sci., 38, 373. Beben, J., Kleint, C., and Meclewski, R. (1989) Surf. Sci., 213, 438. Bagot, P.A.J., Visart de Bocarmé, T., Cerezo, A., and Smith, G.D.W. (2006) Surface Sci., 600, 3028. Bagot, P.A.J., Cerezo, A., and Smith, G.D.W. (2007) Surf. Sci., 601, 2245. Bagot, P.A.J., Cerezo, A., and Smith, G.D.W. (2008) Surf. Sci., 602, 1381. Visart de Bocarmé, T., Moors, M., Kruse, N., Atanasov, I.S., Hou, M., Cerezo, A., and Smith, G.D.W. (2009) Ultramicroscopy, 109, 619. Kellogg, G.L. (1985) Phys. Rev. Lett., 54, 82. Medvedev, V.K., Suchorski, Y., Voss, C., Visart de Bocarmé, T., Bär, T.,
115 116
117 118
119 120 121 122
123 124
125 126
127
128
and Kruse, N. (1998) Langmuir, 14, 6151. Medvedev, V., Suchorski, Y., and Block, J. (1995) Surf. Sci., 343, 169. Hudson, D., Cerezo, A., and Smith, G.D.W. (2009) Ultramicroscopy, 109, 667. Liang, D.B., Abend, G., Block, J.H., and Kruse, N. (1983) Surf. Sci., 126, 392. Visart de Bocarmé, T., Thoi-Dai, C., and Kruse, N. (2006) Top. Catal., 39, 111 and references therein. Kruse, N. (1986) Surf. Sci., 178, 820. Lin, T.H. and Somorjai, G.A. (1981) Surf. Sci., 107, 573. Campbell, C.T., Ertl, G., and Segner, J. (1982) Surf. Sci., 115, 309. Block, J.H., Cocke, D.L., and Kruse, N. (1997) Handbook of Heterogeneous Catalysis (eds G. Ertl, H. Knözinger, and G. Weitkamp). Kruse, N. (1996) Bull. Soc. Chim. Belg., 105, 57. Medvedev, V., Kulik, V., Chernyi, V., and Suchorski, Y. (1997) Vacuum, 48, 341. Suchorski, Y., Medvedev, V., and Block, J. (1995) Phys. Rev. B, 51, 4734. Suchorski, Y., Wrobel, R., Becker, S., and Weiss, H. (2008) J. Phys. Chem. C, 112, 20012. Sieben, B., Suchorski, Y., Bozdech, G., and Ernst, N. (1997) Z. Phys. Chem., 202, 103. Suchorski, Y., Drachsel, W., and Rupprechter, G. (2009) Ultramicroscopy, 109, 430.
261
16 Other Ion-Detecting Techniques John C. Rivière
16.1 Desorption Methods 16.1.1 Electron-Stimulated Desorption (ESD) and ESD Ion Angular Distribution (ESDIAD)
The electron irradiation of a surface, particularly one covered with one or more adsorbed species, can give rise to many types of secondary particles, including positive and negative ions. In ESD and ESDIAD, the surface is irradiated with electrons of energies in the range of 100–1000 eV, and the ejected positive-ion currents of selected species are measured in a mass spectrometer. If the angular distribution of the secondary ions is also measured, either by display on a screen or by using position-sensitive detection, then electron-stimulated desorption (ESD) becomes electron-stimulated desorption ion angular distribution (ESDIAD). Although the electron desorption techniques are not used, and probably cannot be used, for compositional analysis, they provide valuable information about the nature of electronic interactions leading to the breakage of bonds and, in the angleresolved form, about the geometry of surface molecules and the orientation of broken bonds. The primary electrons do not, at the energies employed, succeed in breaking molecular or surface-to-molecule bonds, or in knocking ions out of the surface directly, but the process is one of initial electronic excitation. An electron is absorbed by a surface–adsorbate complex or an adsorbed molecule itself, leading to excitation to an excited state by a Franck–Condon process. If the excited state is antibonding and the molecule or radical is already far enough from the surface, then desorption can occur. Since a return to the ground state after excitation is a much more probable process, the cross-sections for ion desorption are low, between 10−20 and 10−23 cm2. If core-level ionization is involved in the initial interaction with the incident electron, rather than valence levels – as in Franck– Condon-type excitation – a desorption mechanism based on Auger decay has been proposed. The core level left behind after interatomic Auger decay creates a positive ion which is then expelled by the repulsive Madelung potential.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
262
16 Other Ion-Detecting Techniques
The most common ions observed as a result of electron-stimulated desorption are atomic (e.g., H+,O+,F+), but molecular ions such as OH+, CO+, H2O+, and CO2+ can also be found in significant quantities after adsorption of H2O, CO, CO2, and so on. Substrate metallic ions have never been observed, which means that ESD is not applicable to the surface compositional analysis of solid materials. The most important application of ESD in its angularly resolved form, ESDIAD, is in determining the structure and mode of adsorption of adsorbed species. This is because the ejection of positive ions in ESD is not isotropic. Instead, the ions are desorbed along specific directions only, characterized by the orientation of the molecular bonds that are broken by electron excitation. 16.1.2 Thermal Desorption Spectroscopy (TDS)
TDS, which sometimes is referred to as temperature-programmed desorption (TPD), is simple in principle. A gas or mixture of gases is adsorbed onto a clean metal foil for a chosen time. Then, after the gas is pumped away, the foil is heated – at a strictly linear rate – to a high temperature, during which time the current of a particular ion or group of ions is monitored as a function of temperature. The ion masses are selected in a quadrupole mass spectrometer. As the binding energy thresholds of the adsorbed species on the surface are crossed, the peaks in the desorbed ion current appear at characteristic temperatures. From the characteristic temperatures and the shape of the desorption peak above the threshold, the activation energies for desorption can be obtained, along with information about the nature of the desorption process. The mass spectrum from the mass spectrometer, of course, provides information about the species that actually occur on the surface after adsorption. Although simple in principle, experimental artifacts that are possible in TDS must be avoided. Thus, ions accepted by the mass spectrometer must originate only from the surface of the foil, and the temperature distribution across the foil should be uniform in order to avoid the overlapping of desorption processes occurring at different temperatures. To ensure that these experimental requirements are met, the angle of acceptance into the spectrometer is restricted by placing a drift tube with an aperture between the foil and the spectrometer, while the foil itself is usually in the form of a long thin ribbon, with only its center section contributing to the detected ions. In addition, the heating rate must be sufficiently fast so that the desorbed species accepted by the mass spectrometer is characteristic of the desorption process, but not so fast that a sudden pressure increase will occur around the foil. With correct experimental procedure, TDS is straightforward to use and has been applied extensively in basic experiments concerned with the nature of reactions between pure gases and clean solid surfaces. Most of these applications have been catalysis-related (i.e., performed on surfaces acting as models for catalysts), and TDS has always been used in combination with other techniques, such as UPS, ELS, AES, and LEED. To a certain extent TDS is quantifiable, in that the area
16.3 Fast-Atom Bombardment Mass Spectroscopy (FABMS)
under a desorption peak is proportional to the number of ions of that species desorbed in that temperature range, although measurement of the area is not always easy if several processes overlap.
16.2 Glow-Discharge Mass Spectroscopy (GD-MS)
Similarly to GD-OES (see Chapter 20), in GD-MS the surface of the material to be analyzed is eroded by making it the cathode in a glow discharge; again, it is the flux of sputtered neutral atoms that is used to obtain the analytical information. When the neutrals enter the discharge, many are ionized. Whereas in GD-OES the information is provided by spectroscopic analysis of the light emitted during the decay of the excited neutrals, in GD-MS the ions are extracted from the discharge chamber via differential pumping stages into a region at a much lower pressure, where they are mass-analyzed with a magnetic sector mass spectrometer. Because the flux of neutrals emitted from a sputtered surface is always several orders of magnitude greater than the fluxes of positive and negative ions, the sensitivity of GD-MS is high, with detection limits below 1 ng g−1. Although the erosion rate of the surface is also high, such sensitivity allows the compositions of even the outer surface layers to be analyzed. However, the principal applications of the technique have been in the analyses of multilayered materials and of trace elements in semiconductors [1–3]. In principle, all elements in the Periodic Table can be analyzed, and, compared to GD-OES, with interpretation being easier than for the line-rich emission spectra used in the latter technique. Quantification of the mass spectra follows the same principles as in SNMS (see Chapter 9), from which so-called depth profiles can be calculated using relative or absolute sensitivity factors [4]. GD-MS has been available in a commercial instrument, the VG9000 (Thermo Fisher Scientific, East Grinstead, UK), for many years, and that instrument has been used in a variety of applications. For a detailed discussion of its performance, see the review in Ref. [5]. On the other hand, the literature is also full of descriptions of analyses carried out either with modified sources in the commercial instrument [6, 7], or with homebuilt sources coupled to quadrupole [4], sector field [8], or TOF [9] mass spectrometers. Reference [1] contains a review of the subject.
16.3 Fast-Atom Bombardment Mass Spectroscopy (FABMS)
Fast-atom bombardment mass spectrometry (FABMS) is very similar to SSIMS in practice, the only difference being that instead of using positively charged ions as the primary probe, a beam of energetic neutral atoms is used. Secondary ions are emitted, as in SSIMS, and analyzed in a mass spectrometer, usually of the quadrupole type. The beam of fast atoms is produced by passing a beam of ions through
263
264
16 Other Ion-Detecting Techniques
a charge-transfer cell, which consists of a small volume filled with argon to a pressure of approximately 100 Pa. Charge transfer occurs by resonance between fast argon ions and argon atoms with thermal energy, with 15–20% efficiency if geometric and pressure conditions are optimized. Residual ions are removed by electrostatic deflection. FABMS has two advantages over SSIMS, both arising from the use of neutral rather than charged particles. First, little or no surface charging of insulating materials occurs, so organic materials such as polymers can be analyzed without the need to employ auxiliary electron irradiation to neutralize surface charge. Second, the extent of beam damage to a surface, for the same particle flux, is much lower using FABMS than SSIMS, thus enabling materials such as inorganic compounds, glasses, and polymers to be analyzed with less worry about damage introducing ambiguity into the analysis.
References 1 Jakubowski, N., Dorka, R., Steers, E., and Tempez, A. (2007) J. Anal. Atom. Spectrom., 22, 722–735. 2 Marcus, R.K. (ed.) (1993) Glow Discharge Spectroscopies, Plenum Press, New York. 3 Barshick, C.M., Duckworth, D.C., and Smith, D.H. (eds) (2000) Practical Spectroscopy Series, vol. 23, Marcel Dekker, New York. 4 Jakubowski, N. and Stuewer, D. (1992) J. Anal. Atom. Spectrom., 7, 951–958. 5 Becker, J.S. and Dietze, H.-J. (2000) Int. J. Mass. Spectrom., 197, 1–35.
6 Venzago, C. and Weigert, M. (1994) Fresenius J. Anal. Chem., 350, 303–309. 7 Raith, A., Hutton, R.C., and Huneke, J.C. (1993) J. Anal. Atom. Spectrom., 8, 867–874. 8 Jakubowski, N., Feldmann, I., and Stuewer, D. (1997) J. Anal. Atom. Spectrom., 12, 151–157. 9 Yang, C., Mohill, M., and Harrison, W.W. (2000) J. Anal. Atom. Spectrom., 15, 1255–1260.
265
Part Three Photon Detection
267
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis Laszlo Fabry, Siegfried Pahlke, and Burkhard Beckhoff
17.1 Principles
Less than two decades after the first practical application of X-ray fluorescence [1], Moseley established the relation between the reciprocal wavelength of emitted characteristic X-ray fluorescence (XRF) and the atomic number Z of the elements [2, 3]. However, the XRF arrangement was not sensitive enough for trace or ultratrace analyses. In 1923, Compton reported that the reflectivity of a flat target increases below a critical angle of 0.1° under conditions of total X-ray reflection (TXRF). The high reflectivity of the sample support reduced the spectral background of the support and improved the detection limit (DL) down to picogram levels during the early 1970s, when Yoneda and Horiuchi applied the principle of TXRF mainly to ultratrace elemental microanalyses of biological samples [4] (for further details on the history of TXRF, see Refs [5–7]). Since then, TXRF has become a standard tool for surface and subsurface microanalyses [8–12]. In 1983, Becker reported on the angular dependence of X-ray fluorescence intensities in the range of total reflection [13]. Further developments in the field of TXRF are improved detection limits [14] in combination with subtle surface-preparation techniques [15, 16], analyte concentrations extended even to ultratraces (pg) of light elements such as Al [17], the speciation of different chemical states [18], and novel optical arrangements [19] and X-ray sources [20, 21]. For information on the history, principles, and future trends in XRF, the reader is referred to recent books [22, 23], to a handbook by Klockenkämper [24], and to current reviews [25–32]. The basis of XRF analysis is the photoelectric effect and the subsequent emission of X-ray photons characteristic for the analyte atoms in the sample. The element composition can be quantified by the relative intensities of the individual element peaks. In a TXRF arrangement, however, the primary beam interacts with the surface analyte atoms as a standing wave field (Figure 17.1) and excites surface atoms rather than the bulk, in contrast to conventional XRF. The emitted intensity is directly proportional to the intensity of the standing wave field. In the
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
268
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.1 Interference of the incoming and
reflected X-ray waves in the triangular region above a flat and thick reflecting substrate. The strength of the electromagnetic field is
represented on the gray scale with instantaneous crests (white) and troughs (black). During the course of time, the pattern moves from left to right [24].
Figure 17.2 Incident, reflected, and refracted beam at the interface of two media 1 and 2. (a)
Medium 2 is optically denser than medium 1 (n2 > n1); (b) The situation is reversed (n1 > n2) [24].
subsurface, the primary beam generates an evanescent wave field. Hence, the characteristic fluorescence of analyte atoms is a response to these varying excitations by the standing and evanescent wave fields, and its dependence upon the glancing angle, can be exploited for stratigraphy of parallel-layered structures and for distinguishing particulate from film-type surface contamination. In X-ray spectrometry, attenuation, deflection, and interference must be considered. Attenuation is described by the well-known Lambert–Beer law, and the mass attenuation coefficient as given for conventional XRF. In general, X-ray energies do not exceed 100 keV for practical applications. The primary beam is mainly attenuated by the photoelectric effect. Scattering – both elastic (Rayleigh) and inelastic (Compton) – represents a minor contribution to attenuation at energies below 100 keV. The reflection and refraction of X-rays follow the laws of optics (Figure 17.2), where the glancing angles of incidence (Φ1) and reflection (Φ1*) are equal. The respective refraction angles follow Snell’s law for different phase velocities v in medium 1 and 2: V2COSF 1 = V1COSF 2
(17.1)
17.2 Instrumentation
Figure 17.3 Signal intensity of a thick, flat,
and smooth Si substrate (——), calculated for an impinging MoKα beam. In addition, the reflectivity R (·······) is shown as a function of the glancing angle Φ. Below Φcrit = 0.102°,
total reflection occurs with a stepwise increase in reflectivity and a stepwise decrease in signal intensity. The oblique dashed line represents the intensity from a rough Si substrate [24].
When the glancing angle Φ2 of the refractive beam becomes zero, the refractive beam runs tangential to the boundary surface. Below the critical glancing angle of incidence, reflectivity increases to 100%, and the penetration depth – and hence the information depth – is limited to a few nanometers. Reflection below the critical angle is defined as total reflection. The critical angle Φcrit depends on the photon energy E and the atomic mass M of the medium, its atomic number Z, and its density (Equation 17.2). The signal intensity is also a function of surface roughness. Φ crit ≈
1.65 Z ∗ ρ ∗ E M
(17.2)
At Φcrit the penetration depth approaches a minimum, particularly in the case of reflective surfaces, such as chemo-mechanically polished Si. Total reflection disappears on rough surfaces (Figure 17.3). Below Φcrit, the penetration depth is in the range of a few nanometers (Figure 17.4).
17.2 Instrumentation
The construction of a TXRF system, including the X-ray source, energy-dispersive detector, and pulse-processing electronics, is similar to that of conventional XRF. In addition, the geometric arrangement must provide for total reflection of a
269
270
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.4 Penetration depth zn of X-rays impinging on silicon with varying glancing angle
Φ1. The curves were calculated for three different photon energies. The dashed vertical line represents the respective critical angle [24].
Figure 17.5 Scheme of a TXRF system equipped with a fine-focused anode, doubly curved
multilayer monochromator and silicon drift detector (SDD).
monochromatic primary beam. The totally reflected beam interferes with the incident primary beam; such interference causes the formation of standing waves above the surface of a homogeneous sample, as depicted in Figure 17.1, or within a multiple-layered sample. A part of the primary beam fades away in an evanescent wave field in the bulk or substrate [33]. A TXRF instrument (Figure 17.5) consists of an X-ray source, a low-pass filter, collimator, monochromator, sample holder, detector, and an electronic registration unit. Various target materials (e.g., Mo, W, or Mo–W alloys) are used in finefocus, sealed-anode X-ray tubes powered by a water-cooled 3.5-kW generator. The maximum permissible power for high-Z anodes is about 2–3 kW. During their lifetime of 2000–4000 h, the anodes gradually lose intensity; however, rotating anodes are warranted for 2000 h of operation at higher operating powers of up to
17.2 Instrumentation
30 kW and intense water cooling (15 l min−1 instead of 5 l min−1 as normally used for operation at lower power). Nowadays, several low-power (∼50 W) TXRF instrumentations, even with tabletop arrangements, have become available from different manufacturers, including Bruker AXS microanalysis [34], Ital Structures, and Rigaku. In contrast to highpower configurations, these set-ups are optimized for quality control and researchrelated applications in certain biomedical, environmental [35], forensic, nutritional, and pharmaceutical fields. In this case, a double reflector is used as low-pass filter [10]. Tunable monochromators in combination with alloy anodes are also commercially available. For sensitive-surface analyses, monochromators are necessary; these are constructed from LiF, highly oriented pyrolytic graphite (HOPG), or multilayers of W–Si, W–C, or Mo–B4C, which provide for intense and monochromatic radiation, preferentially WLβ or MoKα. The sample positioning device must operate at high geometric reproducibility in all three dimensions, with automated sample load and unload stations. The sample must be situated in an evacuated or He-flushed chamber in order to avoid the absorption being disturbed by air. The energy-dispersive (EDX) solid-state detector (SSD; Figures 17.6 and 17.7) is composed of a lithium-drifted Si crystal [Si(Li)]. Between the thin p- and n-type layers lies a high-resistivity Si crystal of centimeter dimensions, while the front and end planes of the crystal are coated with Au and serve as electrodes. The crystal, when cooled to 77 K with liquid nitrogen, represents a p–i–n diode (Figure 17.7). An incident X-ray photon with an energy above 2 keV causes a cascade of photoelectron–hole pairs into which the initial photon energy is completely converted. Due to the reverse bias, electrons rapidly drift to the positively polarized n-type electrode, and holes to the negative p-type electrode. The number of
Figure 17.6 Sectional drawing of the front end of an SSD with a grooved Si(Li) crystal. Crystal and preamplifier are connected to a cooled copper rod and shielded by a case with an end cap and Be window [24, 36].
271
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis Isolating layer (Si crystal)
To field effect transistor Au-layer 20 nm n-layer 100 nm
–500 V
~4 mm
272
Electrons Holes p-layer 100 nm 4–12 mm
Au-layer 20 nm
hn Figure 17.7 Semiconductor detector
operated as a pin diode with a reverse voltage or bias. An incident X-ray photon ultimately produces a series of electron–hole pairs. These are swept out by the bias field of
−500 V: electrons in the direction of the n-layer, and holes in the direction of the p-layer. Thus, a small charge pulse is produced. After Ref. [37].
electron–hole pairs is proportional to the energy of the impinging photon; hence, the charge pulse is a measure of the characteristic photon energy. Finally, the X-ray spectrum is established by a multichannel analyzer which distributes the pulses according to their height into a series of channels, each of which is reserved for a small range of pulse heights (i.e., energies), and adds them up to give the measured intensity. Optimal resolution – that is, low full-width at half-maximum (FWHM) – is a trade-off between a high count rate – that is, a low dead-time – and good spectral resolution. The new generation of silicon drift detectors (SDD)s [38], as depicted in Figures 17.8–17.10, is an optimal choice for TXRF, GIXRF and XRF analysis, as well as for XRD in both research and routine investigations. The warranted energy resolution of current SDDs ranges from 125 eV to 150 eV FWHM at Mn–Kα; that is, at a photon energy of 5.89 keV. The active detector area is in the range between 7 mm2 and 100 mm2. Currently, state-of-the-art detection systems enable high-throughput capacities up to 106 s−1. At the highest count-rates, a careful online treatment of pile-up events is required. The peak to background ratio (P/B) can achieve levels of up to 20 000, which is comparable to the respective performance of Si(Li) detectors. An integrated thermoelectric cooling (Peltier) allows operation without any need for liquid nitrogen; however, for specific operation conditions, double-stage Peltier cooling systems are available. An automatic temperature monitor system
17.2 Instrumentation
Figure 17.8 Schematic view of a cylindrical silicon drift detector (SDD) with external FET [37]. Negative voltages (−20 V up to −130 V) are applied at the inner and outer ring and at the negative −65 V back contact. X-rays enter the detector at the homogeneous entrance
window (back contact), and are absorbed within the depleted silicon. The generated electrons are then drifted by the applied voltage field towards the readout anode (see Figure 17.9).
Figure 17.9 Principle of the SDD. X-rays generate electron–hole pairs when absorbed within the depleted silicon. The holes drift toward the back contact of the rings, and the electrons drift along the potential minimum defined by the ring voltages and the back voltage towards the readout anode. The drift
time is defined by the distance from the point of interaction to the readout anode. During drift of the electrons the electron cloud spreads, due to diffusion processes. Thus, the signal rise time is increased for events that occur at a longer distance from the anode.
273
274
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.10 (a) The silicon drift detector (SDD). Electronic block diagram of modular analytical X-ray acquisition system (AXAS) with SDD and external FET; (b) 100 mm2 SDD module with open integrated AXAS electronic system; (c) 100 mm2 SDD with Be
window (left), open module (middle) and encapsulated SDD (right); (d) SDD “multi-array” assembled by seven single 100 mm2 SDD components: open module (left), encapsulated module (right).
is integrated in the Peltier cooling stage. Usually, beryllium or polymer windows are used, but ultra-thin polymer entrance windows are optionally available for low-energy measurements. The radiation hardness is in a range of 1012 photons per sensitive area. The commercially available integrated electronic analytical X-ray acquisition systems are either analog (A) or digital (D); both use a pulsed reset amplifier, but only a ± 12 V input is required. The length of the detector finger is optionally up to 300 mm.
17.3 Spectral Information
A physical model for the spectral response behavior of SDDs [39] has been successfully developed and allows for advanced TXRF spectra deconvolution routines, which are already employed for conventional detectors. Hence, the use of SDDs in industrial TXRF arrangements, such as for wafer surface contamination control [40], is possible. When varying the angle of incidence or the distance between the detector and the illuminated spot on the sample, the angular characteristics of the fluorescence radiation detected by the SDD may change, potentially resulting in a modification of the spectral response behavior. At photon energies above about 12 keV, the transmittance of detected radiation becomes relevant in SDDs, due to the finite thickness of the active volume. This can lead to spectral artifacts, depending on the lay-out of the specific SDD.
17.3 Spectral Information
The spectral information acquired from TXRF spectra is similar to that from conventional XRF-EDX spectra. However, the analyst must always be aware of potential spectral interferences that cannot be resolved with standard EDX detectors [17]. Spurious peaks can be assigned to escape peaks in energy-dispersive spectra. An escape peak emerges when intense fluorescence of sufficiently high energy (giving rise to the mother peak) impinges on the SSD, expelling, as usual, photoelectrons from the Si core shells. Excited Si atoms emit X-ray photons that are usually reabsorbed by the crystal. Photoelectrons generated by these photons initiate additional electron–hole cascades (i.e., charge pulses) in the same way as the sample fluorescence. However, photons close to the SSD crystal surface also escape without producing a pulse, thus carrying off the difference between the energy of the mother peak and the generated core photoelectron. This “escaped” energy gives rise to a separate, low-energy daughter peak. Some elements for which the Kα peaks interfere with escape peaks of other elements in SSDs or SDDs are listed in Table 17.1. The most reliable way to include such daughter peaks, and detector-created background contributions, in the TXRF spectra evaluation is
Table 17.1 Elements where the Kα peaks interfere with escape peaks of other elements in SSDs or SDDs [17].
Element
Escape peaks
V Cr Mn Fe Co Ni Cu
CoK α, MnK β, DyL α, HoL α, ErL α, GdL β CoK α, FeK β, ErL α, TmL α, TbL β, DyL β NiK α, CoK β, YbL α, LuL α, HfL α, HoL β CuK α, WL α, TaL α, HfL α, NiK β, TmL β, ErL β ZnK α, WL α, CuK β, YbL β, LuL β, ReL α, OsL α GaK α, PtL α, IrL α, TaL β, HfL β ZnK β, AuL α, WL β, HgL α, GeK α
275
276
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
by means of response functions based on a physical modeling of all relevant processes within an SSD [41, 42]. Spurious peaks can also appear by energy doubling if the charge collection induced by the first fluorescence photon is not complete when a second photon from the same fluorescence source impinges. Then, the detector registers twice the number of charges, and a sum-up peak results. Sum-up peaks can be reduced by a pulse pile-up rejector (PUR). Contamination along the beam path – for example, Fe in the detector window (Be) – may also result in spurious peaks and limit the detection capability for certain elements [43]. In addition, it is equally worthwhile to coat the beam collimating diaphragms with appropriate elements – for example, Si coatings in the case of ultra-trace analysis on Si wafers – in order to prevent any undesired secondary fluorescence radiation from reaching the SSD. Spectral interferences in TXRF cannot be neglected. In the case of wafer cleaning processes, the interference of Al-Kα fluorescence radiation at 1.487 KeV with Br-L fluorescence radiation at 1.499 KeV is of real practical relevance, because the HCl used for etching processes is normally contaminated with Br. Consequently, Br must be also analyzed by its K-fluorescence line at 11.91 KeV, or at least controlled by using independent ultratrace analytical methods, such as ICP-MS, ICP-MS-HR, AAS, and laser diode AAS (LD-AAS).
17.4 Quantification
Reliable quantification is based on peak-search software that combines peak localization, peak identification, and element deduction. Element deduction means that, for unambiguous detection, at least two of the principal peaks be detected for each analyte of interest. In trace analysis, often only the strongest α peaks can be detected, and special attention must be paid to interfering satellites and spurious peaks. An effective alternative to the spectra evaluation by single fluorescence lines is to use multiplets composed of all the fluorescence lines originating from specific inner-shell ionization. The shape of these multiplets is based on the energetic positions and transition probabilities of the respective fluorescence lines [44]. By using an optically flat plate of a pure element such as Ni, TXRF systems can be calibrated absolutely [45]: I X = BX c X
(17.3)
where IX is the background-corrected net intensity of the principal peak of analyte X, BX is a proportionality factor relative to the absolute sensitivity of the standard reference (e.g., a Ni plate), and cX is the concentration of X. Multielement analyses are based on known relative sensitivities, S: Sj =
I j/C j Sref I ref /C ref
(17.4)
17.5 Applications
where the suffix “ref” denotes reference values. Values of relative sensitivities can be determined by theoretical calculations including, for example, mass absorption [46]. The intensity of the fluorescence radiation is angle-dependent. Because matrix effects are absent in the case of minute sample amounts (<1 μg cm−2), quantification by internal standardization is possible. Equation 17.4 holds correspondingly. As the bulk-type response curve also depends on surface roughness [47], all reference materials must be carefully investigated by angle scan prior to use. The angle-scan characteristics of the sample – that is, recording the fluorescence intensity at more than one glancing angle near Φcrit – should not deviate from the reference. These measurements must be carried out under similar optical conditions. With monochromatic AuLβ (11.44 keV) excitation, the variation of Cu depth distribution in monocrystalline Si with time was studied by means of TXRF [48]. Such and similar effects must be carefully investigated when using reference samples of unknown behavior, and in this respect complementary analytical techniques may be helpful. However, the propagation of errors must be taken into consideration and, moreover, the calibration must cover the concentration range of interest [49], because the accuracy of the reference concentrations will greatly affect the accuracy of the results [50, 51]. Typically, an external calibration with reference materials can yield deviations of 3% to 20% from the true value [52]. However, the reliable preparation of controlled spiking has been reported [53]. For general issues of external calibration of TXRF systems, see Refs [54, 55]; for the requirements and perspectives of reference-free quantitation employing table-top TXRF instruments, see Ref. [56].
17.5 Applications
TXRF is an ideal tool for microanalysis [24], since it has a broad range of linearity (1013–108 atoms cm−2), and is extremely surface-sensitive and matrix-independent. However, the effective intensity of the exciting radiation may be altered due to different X-ray standing wave field strengths [57] when moving from one substrate matrix to another. TXRF can be applied to a very wide variety of different organic and inorganic samples, such as water, pure chemicals, oils, body fluids and tissues, and suspended matter, down to the picogram range. 17.5.1 Particulate and Film-Type Surface Contamination
The penetration depth of impinging X-rays is very limited under the conditions of total reflection. For light substrates such as Si, quartz, and poly(methyl methacrylate), the spectral background of TXRF is up to six orders of magnitude smaller than that of XRF. The fluorescence signal arises from the uppermost
277
278
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.11 Fluorescence signal of small particles or thin films deposited on a silicon substrate as sample carrier. The intensity was calculated for particles, thin films, or sections of different thickness but equal
mass of the analyte, and plotted against the glancing angle, Φ. A MoKα beam was used for excitation. Particles or films more than 100 nm thick show double intensity below the critical angle of 0.1° [24].
layer; therefore, the scattered intensity – that is, the background – is lower than in a conventional XRF arrangement. If metallic smear, organic tissues, or thin (<100 nm) dried residues of a solution, suspension, or dispersion are present on a substrate within the standing wave front, then strong oscillation of signal intensity will result (Figure 17.11). Yet, such oscillations will vanish when the grain size or film thickness is increased to about 1000 nm [58]. One can clearly distinguish thick-film from particulate analytes by means of angle characteristics (cf. spectra b and c in Figure 17.12). Although, the particulate and thin-layer angle characteristics do not differ at large angles [59, 60], the surface roughness crucially modifies the peak height [47] and may simulate a stratified structure [61]. 17.5.2 Semiconductors
Metrology and contamination analysis in particular have been decisive factors for profitable semiconductor production [62]. The semiconductor applications of TXRF date back to the late 1980s, when they were first introduced by Eichinger et al. [63]. A review on the subject can found in Ref. [64]. Due to its high sensitivity, wide linear range, facile spectrum deconvolution, and ease of calibration, TXRF has rapidly become the “economic workhorse” for monitoring cleaning efficiency and front-end wafer cleanliness [65, 66]. Likewise, TXRF provides reliable qualitycontrol data that can be easily integrated into a data-management system [67, 68].
17.5 Applications
Figure 17.12 Characteristic intensity profiles
for three different types of contamination on a wafer surface. (a) Bulk; (b) Residue on a surface “particulate contamination type”; (c)
Surface layer (“thin-film contamination type”). The critical angle Φcrit is determined by total reflection at the substrate. Reproduced from Ref. [24], after Ref. [59].
Figure 17.13 Statistical process control chart for TXRF measurement systems. The sensitivity
is monitored by daily calibrations with a Ni standard reference sample. When the X-ray intensity sinks below 75% of the original intensity, the tube must be replaced.
Automatic operation with repeated one-point calibration every 8 h, as well as automatic data management and transfer, has led to abundant applications of TXRF (Figure 17.13). The detection limits for various elements by TXRF on Si wafers are shown in Figure 17.14. To improve the reliability of straight-TXRF for the rapid wholesurface contamination control on a wafer, sweeping techniques have been developed and validated [69].
279
280
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.14 Limits of detection (3σ) for VPD-TXRF measurements on 200- and 300-mm Si
wafers.
17.5.2.1 Synchrotron Radiation-Based Techniques Synchrotron radiation (SR) provides a bright and horizontally polarized X-ray excitation that is tightly collimated and drastically reduces both the elastic Rayleigh and inelastic Compton scattering at the silicon substrate, due to its linear polarization. Developments in the field of SR-TXRF and extreme-ultraviolet (EUV) lithography have promised an improved sensitivity down to the range of less than 107 atoms cm−2 [14, 70–73]. At present, the practical applications of SR-TXRF are focused on the fast mapping of wafers up to a diameter of 300 mm, and reference measurement capabilities in the range of 105–107 atoms cm−2 [74]. The lowest detection limits for transition metals were determined when taking advantage of the extremely high flux of SR, and detecting the K-fluorescence radiation by using a wavelength-dispersive Johansson spectrometer [75]. For a given count-rate capacity of an SSD in a TXRF arrangement, SR improves the detection limits as compared to an X-ray tube-based instrument. This is due to the fact that a higher flux of the incident radiation can be used for the specimen excitation, due to the reduced spectral scattering contributions. In addition, the high spectral purity of current SR beamline monochromators further reduces the spectral background below the fluorescence lines of interest. These advantages have led to TXRF installations in the conventional hard X-ray range at both the Stanford Synchrotron Radiation Lightsource (SSRL) [76] and European Synchrotron Radiation Facility (ESRF) [72] synchrotron radiation facilities, each of which is permanently equipped with local cleanrooms to enable ultra-trace analysis on the surfaces of up to 300-mm wafers. SR – and, in particular, the radiation created by undulators at third-generation synchrotron radiation facilities – offers a tremendously high flux also in the “soft X-ray regime,” as opposed to radiation provided by X-ray tubes. Thus, even “light” elements such as B, C, N, O, and F can be effectively excited and detected, despite their very low fluorescence yields (i.e., the probability of the emission of a fluores-
17.5 Applications
cence photon instead of an Auger electron). Currently, the Physikalisch-Technische Bundesanstalt (PTB), Germany’s national institute for metrology, employs TXRF instrumentation for 200 mm and 300 mm Si wafers [56], including an equipment front-end module (EFEM) and a local clean-room environment (Figure 17.15a) at the electron storage ring BESSY II operated by the Berliner ElektronenSpeicherring-Gesellschaft für Synchrotronstrahlung. In this way, the entire surface of a wafer can be probed by straight-TXRF. Moreover, as the angle of incidence can be varied between 0° and 45°, grazing-incidence XRF (GIXRF) and conventional XRF are also accessible, allowing subsurface elemental profiles or nanolayers, respectively, to be investigated. The excitation radiation for this set-up is provided by a plane-grating monochromator (PGM) for undulator radiation. At BESSY II, PTB is used to absolutely calibrate photodiodes, to record the flux of the incident or reflected beam, and SSDs for various applications in X-ray spectrometry [44, 77]. The use of absolutely calibrated instrumentation and well-known radiation sources, such as the PGM, allows for the “reference-free quantitation” of surface contamination by TXRF analysis [56, 78]. Apart from the experimental parameters, such as the incident radiant power or flux, the effective solid angle of detection, and the response behavior and counting efficiency of the SSD, the accuracy of the knowledge on the relevant atomic parameters [79] (e.g., the photo-electric crosssections, fluorescence yields, and transition probabilities) determines the relative uncertainty of the analytical results. The main advantage of such a reference-free approach is that contamination can be reliably evaluated on different wafer substrates, including novel materials or stratified structures, by calculating the strength of the X-ray standing wave field [57], which determines the effective excitation intensity. Thus, it is no longer necessary to use calibration standards similar to the wafer sample being analyzed. Typically, TXRF spectra (Figure 17.15b) are deconvoluted at PTB with detector response functions with respect to relevant fluorescence lines, bremsstrahlung and resonant Raman scattering (RRS) background, and Rayleigh and Compton scattering contributions. The crosssections of RRS, a resonant inelastic scattering process at inner-shell electrons, could be determined for Si [80] using a reference-free approach in a type of conventional XRF beam geometry. Due to the energetic tunability of SR, near-edge X-ray absorption fine structure (NEXAFS) [81–85] can be combined with TXRF analysis to contribute to the speciation of contamination by light or organic compounds at extremely low levels [56, 86] or for heavier compounds [87]. This is due to the fact that different compounds or functional groups are often associated with resonant transitions from occupied bound to unoccupied bound, or to continuum states at different characteristic photon energies. Both, micro- and nanoparticles [88, 89], when deposited on wafer surfaces, can be likewise analyzed and speciated, demonstrating the potential of this method also for semiconductor-oriented applications. TXRFNEXAFS also allows for the speciation of the surface and near-surface region of nanolayers consisting of light elements [90, 91] or of novel materials [92, 93]. Highresolution X-ray emission techniques [94] based, for example, on spherical grating
281
282
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.15 (a) PTB’s TXRF instrumentation for 200-mm and 300-mm Si wafers at the PGM beamline for undulator radiation; (b) Soft X-ray TXRF spectrum (solid line) of a spin-coated 200-mm Si wafer deconvoluted with detector response functions (dotted)
with respect to C, Ti, O, F, Fe, Ni, Cu, Na, and Al fluorescence lines, bremsstrahlung background (dashed), RRS (dashed) and scattering contributions. The excitation energy was 1622 eV and the angle of incidence 0.9°.
17.5 Applications
spectrometers [95] can complementarily contribute to the speciation of nanolayers. Varying the angle of incidence above the critical angle of total-reflection – that is, entering the GIXRF [96] regime characterized by oscillations of the X-ray standing wave field below the surface – even buried nanolayers [97, 98] or elemental depth profiles of light elements in photoresists [99] can be probed in the soft X-ray range. The higher penetration and information depths of photons, as compared to electrons, thus enables a technique for the nondestructive speciation of deeply buried nanolayers [98] and interfaces that is complementary to, for example, XPS. By combining TXRF and GIXRF, both the surface contamination and the nanolayer thickness of nearly vertical sidewalls [100] on three-dimensional (3-D) trench structures on wafers can be determined, thus providing an addition tool for elemental depth profiling and an extension of reference-free XRF analysis with respect to nanolayer characterization [101, 102]. 17.5.2.2 Depth Profiling by TXRF and by Grazing Incidence XRF (GIXRF) for the Characterization of Nanolayers and Ultra-Shallow Junctions Buried layers are important components of microcircuits, and TXRF is a sensitive microanalytical tool for their inspection (Figure 17.16). On a thick substrate, the angle characteristics is a function of layer thickness [103, 104]. First-principle calculations have been reported on model-free TXRF of stratified structures [105]. Shallow doping profiles such as ultra-shallow junctions (USJs), and particularly those of As, require nanoscale information on dopant distribution. Although SIMS can be reliably applied for layers below 5 nm depth, the technique of choice for
Figure 17.16 Fluorescence intensity from
layers buried in a thick substrate. The intensity as a function of the glancing angle was calculated for layers of different thickness, but with a constant area density of
the analyte. Silicon was assumed as substrate and MoKα X-rays as primary beam. Total reflection occurs below 0.1°. Without total reflection, the dashed horizontal line would be valid in all cases [24].
283
284
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.17 Angular dependence of the fluorescence radiation emitted from a Co-layered Si substrate. The CoKα intensity is plotted semilogarithmically for layers of different thickness (curve parameter in mm). The maxima for the ultrathin Co layers are
localized at the critical angle of Si (dashed vertical line). These are shifted to the critical angle of Co (dotted vertical line) if the layer is more than 10 mm thick. Reproduced from Ref. [24], after Ref. [54].
the upper subsurface is angle-dependent TXRF [106, 107]. The use of angle-scan TXRF allows stratified microstructures to be accurately analyzed for element composition, layer thickness, and density (Figure 17.17) [54]. A combination of TXRF with layer-by-layer chemical etching provides reproducible results on the stratigraphy of metallic implants [16, 108–110]. Secondary ion mass spectrometry (SIMS) is a well-known technique for determining profile shapes, junction depths, and doses of dopants in Si wafers. Whereas, the depth resolution and quantification in the deeper regions of USJs can be sufficiently resolved by SIMS, the effects induced by the native oxide-induced matrix are limiting factors for this technique, if the dose and surface-near distributions in the first nanometers are to be determined in a reliable manner. Therefore, a more complete characterization of the USJ requires the application of different complementary techniques, such as medium energy ion scattering, scanning transmission electron microscopy (STEM), and spectroscopic ellipsometry, as well as grazing incidence X-ray fluorescence (GIXRF) analysis, as demanded by advanced microelectronics technological nodes. The development of the abovementioned techniques, and an understanding of their performance as a complementary method in USJ analysis, represents one of the tasks of research activity within the EC-funded project ANNA (European Integrated Activity of Excellence and Networking for Nano- and Micro-Electronics Analysis), which deals also with
17.5 Applications
inorganic and organic contamination and nanolayer characterization [111–113]. Within the frame of this activity, the X-ray group of the Vienna Atominstitut has developed GIXRF as a laboratory source-driven tool for USJ analysis. In this case, the technique relies on the measurement of fluorescence signals at various incidence angles of the exciting X-ray beam. As the penetration of the primary X-rays becomes larger as a function of increasing incidence angles, GIXRF provides information on the total implanted dose in the substrate material. Since GIXRF is very sensitive in the first few nanometers, it delivers satisfactory results for sample depths, where SIMS has its drawbacks. Complementary investigations employing synchrotron radiation are performed at PTB in the soft X-ray range, providing access to reference-free GIXRF investigations not only of B- and As-USJ junctions [111] but also of buried nanolayers [98]. 17.5.2.3 Vapor-Phase Decomposition (VPD) and Droplet Collection Vapor-phase decomposition and collection (see Figures 17.18–17.20) is a standardized method of silicon wafer surface analysis [12]. For this, the native oxide on wafer surfaces reacts readily with isothermally distilled HF vapor, and forms small droplets on the hydrophobic wafer surface at room temperature [116]. These small droplets can be collected with a scanning droplet. Ultimately, the scanned, accumulated droplet will contain all of the dissolved contamination, and is then dried on a concentrated spot (diameter ca. 150 μm) and measured against the blank droplet residue of the scanning solution (Figure 17.19) [117–119]. VPD-TXRF has
Figure 17.18 (a) Wafer surface preparation
system (WSPS). The first prototype of an automated VPD system for wafer diameters from 100–300 mm [14]; (b) Fully automated
“state-of-the-art” WSPS for semiconductor industry in an integrated mini-environment equipment system, clean room of class 1.
285
286
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
Figure 17.19 Dried VPD droplets. (a) Worst case: the VPD solution exploded under rapid drying with an infrared lamp, the droplet size was a few mm; (b) Best case (WSPS): the VPD solution was dried under controlled conditions with vacuum and carrier gas [49].
Figure 17.20 Detection limit of TXRF for the
residues of aqueous solutions as a function of the atomic number of the analyte element. Three excitations modes were used: (a) W tube; 50 kV; Ni filter, cutoff 35 keV; (b) Mo tube; 50 kV; Mo filter, cutoff 20 keV; (c) W
tube, 25 kV, Cu filter. The three curves to the left were determined by the detection of K peaks; the three curves to the right, by that of L peaks. Reproduced from Ref. [24], after Refs. [114, 115].
been carefully evaluated against standardized surface analytical methods. Great care must be taken to use reliable reference materials [120–122]. The accuracy of VPD-TXRF depends largely on the reliability of the reference material, and on the reproducibility of the droplet-drying procedure (Figure 17.19) [124, 125]. Mass absorption sets the upper limit of reliable analyses at 1014 atoms cm−2 [126]. VPD-TXRF is also a straightforward technique for use in interface analysis [127, 128]. Automated VPD equipment (Figure 17.18) improves both the detection limit (upper range of 107 atoms cm−2) and the reliability (by >50%) of the VPD-TXRF measurement [15]. Current research is focused on issues of sample holders [129, 130] and light-element detection capabilities [131–133].
17.5 Applications
Figure 17.21 Vapor-phase decomposition (VPD) versus vapor-phase treatment (VPT).
Schematic flow diagram of VPD-TXRF and VPD-TXRF wafer sample preparation technique. After Ref. [123].
17.5.2.4 Vapor-Phase Treatment (VPT) and Total Reflection X-Ray Fluorescence Analysis Vapor-phase-treatment (VPT), in combination with total reflection X-ray fluorescence analysis (TXRF), collectively termed VPT-TXRF, has recently been developed (Figure 17.21). This is a new, practical method for the quantification of metallic wafer contaminations with concentrations as low as 109 atoms cm−1 [123]. VPT can change the form of metallic impurities on the surface from thin-film type to particulate type; as a consequence the detected X-ray intensity is increased, because of the inherent nature of TXRF, namely, the angular dependence of fluorescence yield. A high sensitivity was achieved within a practical measuring time of 500 s, while maintaining the original surface contamination distribution of the wafers [123]. A mapping capability has also been developed for VPT-TXRF, as for sweeping TXRF [134–136]. In order to provide proper quantification spurious peaks due to Bragg reflections must be taken into consideration [137–139].
287
288
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis
References 1 Pupin, M.I. (1896) Science, 3, 538–544. 2 Moseley, H.G.J. (1913) Philos. Mag., 26, 1024. 3 Moseley, H.G.J. (1914) Philos. Mag., 27, 703. 4 Yoneda, Y. and Horiuchi, T. (1971) Rev. Sci. Instrum., 42, 1069. 5 Wobrauschek, P., Kregsamer, P., Streli, C., and Aiginger, H. (1991) X-Ray Spectrom., 20, 23–28. 6 Kregsamer, P. (1991) Spectrochim. Acta B, 46, 1333–1340. 7 Aiginger, H. (1991) Spectrochim. Acta B, 46, 1313–1321. 8 Aiginger, H. and Wobrauschek, P. (1974) Nucl. Instrum. Methods, 114, 157. 9 Knoth, J. and Schwenke, H. (1978) Fresenius J. Anal. Chem., 291, 200. 10 Schwenke, H. and Knoth, J. (1982) Nucl. Instrum. Methods, 193, 239. 11 Semiconductor Equipment and Materials International (July 2003) SEMI Standards E 45. 12 Semiconductor Equipment and Materials International (September 1998) SEMI Standards M 33. 13 Becker, R.S., Golovchenko, J.A., and Patel, J.R. (1983) Phys. Rev. Lett., 50, 153. 14 Wobrauschek, P., Görgl, R., Kregsamer, P., Streli, C., Pahlke, S., Fabry, L., Haller, M., Knöchel, A., and Radtke, M. (1997) Spectrochim. Acta Part B, 52, 901–906. 15 Fabry, L., Pahlke, S., Kotz, L., Wobrauschek, P., Streli, C., and Eichinger, P. (1999) Fresenius J. Anal. Chem., 363, 98–102. 16 Klockenkämper, R. and von Bohlen, A. (1999) Spectrochim. Acta B, 54, 1385–1392. 17 Streli, C., Kregsamer, P., Wobrauschek, P., Gatterbauer, H., Pianetta, P., Pahlke, S., Fabry, L., Palmetshofer, L., and Schmeling, M. (1999) Spectrochim. Acta B, 54, 1433–1441. 18 Xiao, X., Hayakawa, S., Goshi, Y., and Oshima, M. (1998) Anal. Sci., 14, 1139–1144. 19 de Bokx, P.K., Kidd, S.J., Wiener, G., Urbach, H.P., de Gendt, S., Mertens, P.W., and Heyns, M.M. (1998)
20
21
22
23
24
25
26
27 28 29
30 31 32 33
34
Proceedings of the Electrochemical Society, Vols 98-91. Silicon Mater. Sci. Technol., 2, 1511–1523. Schwenke, H. (1999) GKSS, Geesthacht, personal communication. Proceedings GMM Workshop, Erlangen 23–24 February 2000, Gesellschaft für Mikroelektronik VDE/VDI. Wobrauschek, P. and Streli, C. (1999) Advanced Light Sources, ADLIS, TU Wien, Vienna, pp. 284–300. Tsuji, K., Injuk, J., and van Grieken, R. (2004) X-Ray Spectrometry: Recent Technological Advances, John Wiley & Sons, Inc.. Beckhoff, B., Kanngießer, B., Langhoff, N., Wedell, R., and Wolff, H. (2006) Handbook of Practical X-Ray Fluorescence Analysis, Springer. Klockenkämper, R. (1997) Total Reflection X-Ray Analysis, John Wiley & Sons, Inc., New York. Potts, P.J., Ellis, A.T., Kregsamer, P., Streli, C., West, M., and Wobrauschek, P. (1999) J. Anal. Atom. Spectrom., 14, 1773–1799. Wobrauschek, P., Streli, C., and Selin-Lindgren, E. (2010) Energy Dispersive, X-Ray Fluorescence Analysis, in Encyclopedia of Analytical Chemistry, (ed. R.A. Meyers), John Wiley & Sons, Chichester; doi: 10.1002/9780470027318.pub2. Stoev, K.N. and Sakurai, K. (1999) Spectrochim. Acta B, 54, 41–82. Pahlke, S. (2003) Spectrochim. Acta B, 58, 2025–2038. Hellin, D., De Gendt, S., Valckx, N., Mertens, P.W., and Vinckier, C. (2006) Spectrochim. Acta B, 61, 496–514. Klockenkämper, R. (2006) Spectrochim. Acta B, 61, 1082–1090. Streli, C. (2006) Appl. Spectrosc. Rev., 41, 473–489. Wobrauschek, P. (2007) X-Ray Spectrom., 36, 289–300. Yakushiji, K., Ohkawa, S., Yoshinaga, A., and Harada, J. (1993) Jpn. J. Appl. Phys., 32, 4750–4751. Waldschläger, U. (2006) Spectrochim. Acta B, 61, 1115–1118.
References 35 Stosnach, H. (2006) Spectrochim. Acta B, 61, 1141–1145. 36 Williams, K.L. (1987) An Introduction to X-Ray Spectrometry, Allen & Unwin, London. 37 Strüder, L., Lechner, P., and Leutenegger, P. (1998) Naturwissenschaften, 85, 539–543. 38 Longoni, A. and Fiorini, C. (2006) X-ray detectors and signal processing, in Handbook of Practical X-Ray Fluorescence Analysis (eds B. Beckhoff, B. Kanngießer, N. Langhoff, R. Wedell, and H. Wolff), Springer, pp. 203–261. 39 Eggert, T., Boslau, O., Kemmer, J., Pahlke, A., and Wiest, F. (2006) Nucl. Instrum. Meth. A, 568, 1–11. 40 Pahlke, S., Meirer, F., Wobrauschek, P., Streli, C., Westphal, G.P., and Mantler, C. (2006) Spectrochim. Acta B, 61, 1110–1114. 41 Lowe, B.G. (2000) Nucl. Instrum. Meth. A, 439, 247–261. 42 Scholze, F. and Procop, M. (2001) X-Ray Spectrom., 30, 69–76. 43 Yakushiji, K., Ohkawa, S., Yoshinaga, A., and Harada, J. (1995) Anal. Sci., 11, 505–511. 44 Scholze, F., Beckhoff, B., Kolbe, M., Krumrey, M., Müller, M., and Ulm, G. (2006) Microchim. Acta, 155, 275–278. 45 Knoth, J., Schwenke, H., and Eichinger, P. (1994) Ultra-Clean Processing (UCPSS 94, IMEC) (eds M. Heyns, M. Heuris, and P. Mertens), ACCO, Leuven, p. 107. 46 Tao, G.Y., Zhuo, S.J., Ji, A., Norrish, K., Fazey, P., and Senff, U.E. (1998) X-Ray Spectrom., 27, 357–366. 47 Tsuji, K., Yamada, T., Utaka, T., and Hirokawa, K. (1995) J. Appl. Phys., 78, 969–973. 48 Mori, Y. and Shimanoe, K. (1996) Anal. Sci., 12, 277–279. 49 Fabry, L., Pahlke, S., and Kotz, L. (1996) Fresenius J. Anal. Chem., 354, 266–270. 50 Mertens, P., DeGendt, S., and Kenis, K. (1996) Calibration Accuracy of Different Atomika TXRF Instruments, IMEC, Leuven. 51 Rink, I. and Thewissen, H. (1999) Spectrochim. Acta B, 54, 1427–1431. 52 Mori, Y., Shimanoe, K., and Sakon, T. (1995) Anal. Sci., 11, 499.
53 Hölzl, R., Range, K.J., Fabry, L., and Huber, D. (1999) J. Electrochem. Soc., 146, 2245. 54 Diebold, A.C. (1996) J. Vac. Sci. Technol. A, 14, 1919. 55 Hockett, R.S. (1995) Anal. Sci., 11, 511. 56 Beckhoff, B., Fliegauf, R., Kolbe, M., Müller, M., Weser, J., and Ulm, G. (2007) Anal. Chem., 79, 7873–7882. 57 Windt, D. (1998) Comput. Phys., 12, 360–370. 58 Schwenke, H., Knoth, J., and Weisbrod, U. (1991) X-Ray Spectrom., 20, 277. 59 Weisbrod, U., Gutschke, R., Knoth, J., and Schwenke, H. (1991) Fresenius J. Anal. Chem., 341, 83. 60 Berneike, W. (1993) Spectrochim. Acta B, 48, 269–275. 61 Schwenke, H., Gutschke, R., Knoth, J., and Kock, M. (1992) Appl. Phys. A, 54, 460. 62 Fabry, L., and Matsushita, Y. (1998) Proceedings of the Electrochemical Society, Vols 98–91, Silicon Mater. Sci. Technol., 2, 1459. 63 Eichinger, P., Rath, H.-J., and Schwenke, H. (1989) Semiconductor Fabrication: Technology and Metrology, ASTM STP 990 (ed. D.C. Gupta), ASTM, p. 305. 64 Hockett, R.S. (1994) Adv. X-Ray Anal., 37, 565–575. 65 Penka, V. and Hub, W. (1989) Fresenius J. Anal. Chem., 333, 586–589. 66 Sees, J.A. and Hall, L.H. (1995) J. Electrochem. Soc., 142, 1238–1241. 67 Fabry, L. (1996) Accred. Qual. Assur., 1, 99. 68 Fabry, L., Köster, L., Pahlke, S., Kotz, L., and Hage, J. (1996) IEEE Trans. Semicond. Manuf., 9, 428. 69 Mori, Y., Uemura, K., and Iizuka, Y. (2002) Anal. Chem., 74, 1104–1110. 70 Streli, C., Wobrauschek, P., Bauer, V., Kregsamer, P., Görgl, R., Pianetta, P., Ryon, R., Pahlke, S., and Fabry, L. (1997) Spectrochim. Acta B, 52, 861–872. 71 Wobrauschek, P., Kregsamer, P., Ladisich, W., Streli, C., Pahlke, S., Fabry, L., Garbe, S., Haller, M., Knöchel, A., and Radtke, M. (1995) Nucl. Instrum. Methods A, 363, 619–620.
289
290
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis 72 Comin, F., Navizet, M., Mangiagalli, P., and Aposolo, G. (1999) Nucl. Instrum. Methods B, 150, 538–542. 73 Binder, H. (2000) M + W Zander Facility Engineering GmbH, BadenWürttemberg, Germany. Available at: www.mw-zander.com. 74 McHugo, S.A., Thompson, A.C., Flink, C., Weber, E.R., Lambie, G., Gunion, B., MacDowell, A., Celestre, R., Padmore, H.A., and Hussain, Z. (2000) J. Cryst. Growth, 210, 395–400. 75 Sakurai, K., Eba, H., Inoue, K., and Yagi, N. (2002) Anal. Chem., 74, 4532–4535. 76 Pianetta, P., Takaura, N., Brennan, S., Tompkins, W., Laderman, S.S., Fischer-Colbrie, A., Shimazaki, A., Miyazaki, K., Madden, M., Wherry, D.C., and Kortright, J.B. (1995) Rev. Sci. Instrum., 66, 1293–1297. 77 Krumrey, M., Gerlach, M., Scholze, F., and Ulm, G. (2006) Nucl. Instrum. Methods A, 568, 364–368. 78 Beckhoff, B. (2008) J. Anal. Atom. Spectrom., 23, 845–853. 79 Zschornack, G.H. (2007) Handbook of X-Ray Data, Springer. 80 Müller, M., Beckhoff, B., Ulm, G., and Kanngießer, B. (2006) Phys. Rev. A, 74, 012702. 81 Stöhr, J. (1992) NEXAFS Spectroscopy, vol. 25, Springer Series in Surface Sciences, Springer. 82 Rehr, J.J. and Albers, R.C. (2000) Rev. Mod. Phys., 72, 621–654. 83 de Groot, F. (2001) Chem. Rev., 101, 1779–1808. 84 de Groot, F. (2005) Coordination Chem. Rev., 249, 31–63. 85 Rehr, J.J. (2006) Radiat. Phys. Chem., 75, 1547–1558. 86 Pepponi, G., Beckhoff, B., Ehmann, T., Ulm, G., Streli, C., Fabry, L., Pahlke, S., and Wobrauschek, P. (2003) Spectrochim. Acta B, 58, 2245–2253. 87 Meirer, F., Streli, C., Pepponi, G., Wobrauschek, P., Zaitz, M.A., Horntrich, C., and Falkenberg, G. (2008) Surf. Interface Anal., 40, 1571–1576. 88 Török, S., Osan, J., Beckhoff, B., and Ulm, G. (2004) Powd. Diffract. J., 19, 81–86.
89 Osan, J., Török, S., Beckhoff, B., Ulm, G., Hwang, H., Ro, C.-U., Abete, C., and Fuoco, R. (2006) Atmosph. Environm., 40, 4691–4702. 90 Hoffmann, P., Baake, O., Beckhoff, B., Ensinger, W., Fainer, N., Klein, A., Kosinova, M., Pollakowski, B., Trunova, V., Ulm, G., and Weser, J. (2007) Nucl. Instrum. Methods A, 575, 78–84. 91 Baake, O., Hoffmann, P.S., Klein, A., Pollakowski, B., Beckhoff, B., Ensinger, W., Kosinova, M., Fainer, N., Sulyaeva, V.S., and Trunova, V. (2008) X-Ray Spectrom., 38, 68–73. 92 Danel, A., Nolot, E., Veillerot, M., Olivier, S., Decorps, T., Calvo-Muñoz, M.L., Hartmann, J.M., Lhostis, S., Kohno, H., Yamagami, M., and Geoffroy, C. (2008) Spectrochim. Acta B, 63, 1365–1369. 93 Sparks, C.M., Lanee, S., Beebe, M., and Page, M. (2008) Spectrochim. Acta B, 63, 1351–1354. 94 Reinhardt, F., Beckhoff, B., Eba, H., Kanngießer, B., Kolbe, M., Mizusawa, M., Müller, M., Pollakowski, B., Sakurai, K., and Ulm, G. (2009) Anal. Chem., 81, 1770–1776. 95 Müller, M., Beckhoff, B., Fliegauf, R., and Kanngießer, B. (2009) Phys. Rev. A, 79, 032503. 96 de Boer, D.K.G. (1991) Phys. Rev. B, 44, 498–511. 97 Beckhoff, B., Fliegauf, R., Hönicke, P., Kolbe, M., Müller, M., Pollakowski, B., Weser, J., and Ulm, G. (2007) ECS Transactions, 11 (3), 273–279. 98 Pollakowski, B., Beckhoff, B., Reinhardt, F., Braun, S., and Gawlitza, P. (2008) Phys. Rev. B, 77, 235408. 99 Conard, T., List, S., Claes, M., and Beckhoff, B. (2008) Solid State Phenomena, 134, 281–284. 100 Hönicke, P., Beckhoff, B., Kolbe, M., List, S., Conard, T., and Struyff, H. (2008) Spectrochim. Acta B, 63, 1359–1364. 101 Kolbe, M., Beckhoff, B., Krumrey, M., and Ulm, G. (2005) Spectrochim. Acta B, 60, 505–510. 102 Baake, O., Hoffmann, P.S., Flege, S., Ortner, H.M., Gottschalk, S., Berky, W., Balogh, A.G., Ensinger, W., Beckhoff, B., Kolbe, M., Gerlach, M., Pollakowski,
References
103
104
105
106
107
108
109
110 111
112
113
114
B., Weser, J., Ulm, G., Haschke, M., Blokhina, E., Peter, M., Porta, D., and Heck, M. (2008) Anal. Bioanal. Chem., 393, 623–634. Schwenke, H., Berneike, W., Knoth, J., and Weisbrod, U. (1989) Adv. X-Ray Anal., 32, 105. Schwenke, H., and Knoth, J. (1993) Handbook on X-Ray Spectrometry (eds R. van Grieken and A. Markowicz), Dekker, New York, p. 453. Dane, A.D., van Sprang, H.A., and Buydens, L.M.C. (1999) Anal. Chem., 71, 4580–4586. Schwenke, H., Knoth, J., Fabry, L., Pahlke, S., Scholz, R., and Frey, L. (1997) J. Electrochem. Soc., 144, 3979–3983. Pepponi, G., Streli, C., Wobrauschek, P., Zoeger, N., Luening, K., Pianetta, P., Giubertoni, D., Barozzi, M., and Bersani, M. (2004) Spectrochim. Acta B, 59, 1243–1249. Klockenkämper, R., von Bohlen, A., Becker, H.W., and Palmetshofer, L. (1999) Surf. Interface Anal., 27, 1003–1008. Klockenkämper, R. and von Bohlen, A. (1999) J. Anal. Atom. Spectrom., 14, 571–576. Klockenkämper, R. and von Bohlen, A. (1999) Anal. Commun., 36, 27–29. Nutsch, A., Beckhoff, B., Altmann, R., van den Berg, J.A., Giubertoni, D., Hönicke, P., Bersani, M., Leibold, A., Meirer, F., Müller, M., Pepponi, G., Otto, M., Petrik, P., Reading, M., Pfitzner, L., and Ryssel, H. (2009) Solid State Phenomena, 145–146, 97–100. Beckhoff, B., Nutsch, A., Altmann, R., Borionetti, G., Pello, C., Polignano, M.L., Codegoni, D., Grasso, S., Cazzini, E., Bersani, M., Lazzeri, P., Gennaro, S., Kolbe, M., Müller, M., Kregsamer, P., and Posch, F. (2009) Solid State Phenomena, 145–146, 101–104. Lommel, M., Hönicke, P., Kolbe, M., Müller, M., Reinhardt, F., Möbus, P., Mankel, E., Beckhoff, B., and Kolbesen, B.O. (2009) Diffusion and Defect Data – Solid State Data, Pt. B: Solid State Phenomena, 145–146, 169–172. Reus, U., Freitag, K., Haase, A., and Alexandre, J.F. (1989) Spectra 2000, 143, 42–46.
115 Prange, A. and Schwenke, H. (1992) Adv. X-Ray Anal., 35B, 899. 116 Shiraiwa, T., Fujino, N., Sumita, S., and Tanizoe, Y. (1989) Semiconductor Fabrication: Technology and Metrology, ASTM STP 990 (ed. D.C. Gupta), ASTM, p. 314. 117 Huber, A., Rath, H.J., Eichinger, P., Bauer, T., Kotz, L., and Staudigl, R. (1988) Proceedings of the Electrochemical Society, PV 88–20, Diagn. Tech. Semicond. Mater. Devices, 109–112. 118 Neumann, C. and Eichinger, P. (1991) Spectrochim. Acta B, 46, 1369–1377. 119 Buhrer, G. (1999) Spectrochim. Acta B, 54, 1399. 120 Diebold, A.C., Maillot, P., Gordon, M., Baylis, J., Chacon, J., Witowski, R., Arlinghaus, H.F., Knapp, J.A., and Doyle, B.L. (1992) J. Vac. Sci. Technol. A, 10, 2945–2952. 121 Metz, S., Kilian, G., Mainka, G., Angelkort, C., Fester, A., and Kolbesen, B.O. (1997) Proceedings of the Electrochemical Society, PV, 97–22, Cryst. Defects Contamin.: Impact Control Device Manufact., II, 458–467. 122 Fabry, L., Pahlke, S., Kotz, L., Schemmel, E., and Berneike, W. (1993) Proceedings of the Electrochemical Society, PV, 93–15, Cryst. Defects Contamin., 232–239. 123 Shimazaki, A., Miyazaki, K., Matsumura, T., and Ito, S. (2006) Jap. J. Appl. Phys., 45, 9037–9043. 124 Fabry, L., Pahlke, S., and Kotz, L. (1995) Adv. X-Ray Chem. Anal. Jpn., 27, 345. 125 Kondo, H., Ryuta, J., Morita, E., Yoshimi, T., and Shimanuki, Y. (1992) Jpn. J. Appl. Phys., 31 (Part 2 No. 1A/B), L11. 126 Klockenkämper, R. and von Bohlen, A. (1989) Spectrochim. Acta B, 48, 461. 127 Rink, I. and Thewissen, H. (1999) Spectrochim. Acta B, 54, 1427–1431. 128 Matsumura, M., Sakoda, T., and Nishioka, Y. (1998) Jpn. J. Appl. Phys., 37, 5963–5964. 129 Yamaguchi, H., Itoh, S., Igarashi, S., Naitoh, K., and Hasegawa, R. (1998) Anal. Sci., 14, 909.
291
292
17 Total-Reflection X-Ray Fluorescence (TXRF) Analysis 130 Yamaguchi, H., Itoh, S., Igarashi, S., Naitoh, K., and Hasegawa, R. (1999) Bunseki Kagaku, 48, 855L. 131 Yamagami, M., Nonoguchi, M., Yamada, T., Shoji, T., Utaka, T., Nomura, S., Taniguchi, K., Wakita, H., and Ikeda, S. (1999) X-Ray Spectrom., 26, 451. 132 Yamagami, M., Nonoguchi, M., Yamada, T., Shoji, T., Utaka, T., Nomura, S., Taniguchi, K., Wakita, H., and Ikeda, S. (1999) Bunseki Kagaku, 48, 1005. 133 Vereecke, G., Schaekers, M., Verstraete, K., Arnauts, S., Heyns, M.M., and Plante, W. (2000) J. Electrochem. Soc., 147, 1499. 134 Mori, Y., Uemura, K., Kohno, H., Yamagami, M., Yamada, T., Shimizu,
135
136
137
138
139
K., Onizuka, Y., and Iizuka, Y. (2004) Spectrochim. Acta B, 59, 1277–1282. Danel, A., Cabuil, N., Lardin, T., Despois, D., Veillerot, M., Geoffroy, C., Yamagami, M., and Kohno, H. (2008) Spectrochim. Acta B, 63, 1375–1381. Borde, Y., Danel, A., Roche, A., and Veillerot, M. (2008) Spectrochim. Acta B, 63, 1370–1374. Yakushiji, K., Ohkawa, S., Yoshinaga, A., and Harada, J. (1993) Jpn. J. Appl. Phys., 32, 1191–1196. Yakushiji, K., Ohkawa, S., Yoshinaga, A., and Harada, J. (1994) Jpn. J. Appl. Phys., 33, 1130–1135. Liou, B.W. and Lee, C.L. (1999) Chin. J. Phys., 37, 623–630.
293
18 Energy-Dispersive X-Ray Spectroscopy (EDXS) Reinhard Schneider
Physically, the generation of X-rays is often a secondary process that is preceded by the ionization of an atom. There are, therefore, several possibilities for X-ray generation, depending on the type of the exciting medium; examples include neutral particles or charged particles such as electrons and ions, and also highenergy photons – that is, X-rays themselves. Thus, the occurrence of X-rays can be observed in all electron-optical instruments where samples are bombarded by electrons with energies sufficiently high to ionize matter (cf. Chapter 4). The following discussion is restricted to the use of X-rays for microanalysis in transmission electron microscopy (TEM); however, as X-ray spectroscopy is widely used also in scanning electron microscopy (SEM), some fundamental considerations are made on the combined application of these techniques. In general, two main X-ray spectroscopic methods can be distinguished, depending on the physical property to be measured, namely the wavelength or the energy of the emitted X-rays. The corresponding methods are wavelength-dispersive X-ray spectroscopy (WDXS) and energy-dispersive X-ray spectroscopy (EDXS), where WDXS is used almost exclusively in conjunction with SEM. The following text provides an introduction to the fundamentals, instrumentation, and application of EDXS. Because of its importance, peculiarities, and application possibilities, a wealth of monographs (see, e.g., Refs [1–7]) and original reports have been devoted to EDXS/WDXS, while many also relate to TEM/SEM. The fundamentals and principles of electron microscopy itself, including such details as electron–target interaction, electron optics, and image as well as contrast generation, are presented elsewhere.
18.1 Principles
When an atom has been ionized by electron bombardment – for instance, by the transition of a K-shell electron into vacuum – then after a finite dwell time the hole generated can be filled by an electron of the L shell (cf. Chapter 4, Figure 4.2. There is a certain probability that the energy set free by this process will lead to Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
294
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
the excitation of characteristic X-rays, viz. Kα radiation for the transition described. If an M-shell electron falls into the hole state, the emission of so-called Kβ X-rays occurs. One process which competes with the emission of X-rays is the emission of Auger electrons, where the sum of both the fluorescence yield of X-rays and the probability of Auger electron emission is always equal to 100%. The fluorescence yield of X-rays increases with atomic number Z, while the portion of emitted Auger electrons behaves in the opposite manner. For ionization of the L-shell and filling of the remaining hole by electrons from the M- or N-shells, the emission of Lα or Lβ radiation, respectively, is observed. For heavier elements, electrons from the more outer shells can also occupy the holes, although not all conceivable transitions between the existing electronic energy levels are allowed. The physically allowed transitions are determined by the following quantum-mechanical selection rules (cf., e.g., [8]): 1) 2)
Δn > 0, that is, L → K, M → K transitions but not, for example, L3 → L2 Δl = ±1 and Δj = −1, 0, +1,
where n is the principal quantum number, l the angular momentum quantum number, and j = l ± s is that of the total angular momentum – that is, the combination of orbital momentum and spin. These so-called dipole selection rules enable the prediction of all possible transitions between electronic levels, and thus also of the corresponding X-ray lines. This is illustrated in Figure 18.1, where some
Figure 18.1 Allowed electronic transitions to the K-shell and corresponding X-ray lines after
ionization of an atom (two forbidden transitions are also shown as dashed lines).
18.2 Practical Aspects of X-Ray Microanalysis and Instrumentation
allowed transitions contributing to the K-series of X-rays are shown as solid lines, whereas forbidden transitions are drawn as dashed lines. In accordance with spin-orbit coupling [9], the higher energy levels are split, enabling additional transition processes. Thus, for the Kα radiation two individual Kα1 and Kα2 lines can be distinguished which are attributed to transitions from 2p3/2 and 2p1/2 levels to the 1s1/2 state, because Δl = ±1 and Δj = 0, ±1. In practice, because of an average energy resolution of approximately 130 eV of EDXS detectors, this splitting of energy levels is only measurable for heavier elements, whereas WDXS can usually be used to resolve the individual sub-lines (see Figure 18.7). The Kα1/Kα2 line splitting amounts to approximately 2 eV for potassium (Z = 19), 75 eV for yttrium (Z = 39), and 1.815 keV for gold (Z = 79). The energies of the X-ray quanta of the K- and L-series can be calculated by use of the equations [10]: E(Kξ(n )) = E A(n +1) − EK = hcR(Z − 1)2 (1/n 2 − 1) and E(Lξ(n )) = E A(n +2) − EL = hcR(Z − 7.4)2 (1/n 2 − 1/22 ) where ξ(n) = α, β, γ, … , A(n) = K, L, M, … , and n = 1, 2, 3, … , R is the Rydberg constant, h is Planck’s constant, and c is the velocity of light. In these approximations for the K-series, the value 1 is subtracted from the atomic number Z to correct for the screening of the nuclei by the remaining K-shell electron. For the L-series, the screening effect of the two K-shell electrons and the seven remaining L-shell electrons must be taken into consideration by subtracting 7.4. The energy E of the characteristic X-rays within a given series of lines – that is, Kα, Kβ, etc. – increases regularly with the atomic number Z. This dependence is referred to as Moseley’s law of X-ray emission: E ∝ (Z − 1) . In an electron-excited X-ray spectrum, the discrete X-ray lines are superimposed on a continuous background; this is the well-known bremsstrahlung continuum ranging from 0 to the primary energy E0 of the electrons. The reason for this continuum is that, because of the fundamental laws of electrodynamics, electrons emit X-rays when they are decelerated in the Coulomb field of an atom. As a result, the upper energy limit of X-ray quanta is identical with the primary electron energy.
18.2 Practical Aspects of X-Ray Microanalysis and Instrumentation
For the electron-beam illumination of an object, the spatial extension of the volume from which a specific interaction signal is gained depends heavily on the material itself, as well as on the diameter of the electron beam, the primary
295
296
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
Figure 18.2 Beam broadening in Si as a function of sample thickness at 100 keV primary
electron energy. (a) Bulk specimen; (b) 200 nm thickness; (c) 50 nm thickness.
electron energy, and the specimen thickness. In bulk materials, many more interaction processes occur between incoming electrons and target atoms in comparison with a thin film. Each individual interaction process can change the direction of the electron path within the specimen, leading to a broadening of the average diameter of the electron beam. In general, electron propagation directly determines the distribution of the energy within the target and, therefore, also the size of the interaction or excitation volume, because the absorbed energy promotes the generation of secondary information carriers like secondary electrons, electron– hole pairs, and X-rays. The thicker the specimen, the more extended is the interaction volume, essentially limiting the achievable lateral resolution. Consequently, for thin films of approximately 10 nm thickness the resolution can be much better than for massive samples. The phenomena of beam broadening as a function of specimen thickness are illustrated in Figure 18.2; here, each figure represents 200 electron trajectories in silicon calculated by Monte Carlo simulations (cf., e.g., Refs [7, 11–13]) for 100 keV primary energy, where an infinitesimally small electron probe is assumed to enter the surface. In massive Si, the electrons suffer a large number of elastic and inelastic interactions during their passage through the material, until they are finally completely stopped. The resulting penetration depth of the electrons is approximately 50 μm, whilst in the perpendicular direction their ranges are nearly twice that, owing to backscattering. This situation meets the conditions present in SEM and electron probe microanalysis (EPMA). For the foil specimens applicable in TEM/STEM, the beam-broadening effect is usually defined by the diameter of the disc at the exit surface containing 90% of the emerging electrons [12, 13]. In Figure 18.2, this diameter is approximately 1 nm for the 50 nm-thick foil, and reaches as
18.2 Practical Aspects of X-Ray Microanalysis and Instrumentation
much as 12 nm for 200 nm-thick foil. Based on these considerations, it is clear that analytical information of extremely high lateral resolution can be obtained only for very thin specimens in TEM/STEM. Before the development of semiconductor detectors opened the field of EDXS during the late 1960s, crystal-spectrometer arrangements were used widely to measure the intensity of emitted X-rays as a function of their wavelength. Such wavelength-dispersive X-ray spectrometers use the reflections of X-rays from a known crystal, which can be described by Bragg’s law (see also Section 19.1.3) nλ = 2d sin θ where λ is the wavelength of the X-rays, θ the scattering angle, n the order of reflection, and d the interplanar spacing of the single crystal chosen. The point of the electron probe on the specimen, the surface of the analyzing crystal, and the detector slit are placed on the circumference of a focusing circle, also called the Rowland circle (see Refs [14–16] and Section 2.2.2). The movement of the crystal and the detector is mechanically coupled, such that the detector forms an angle θ with the crystal surface while it moves an angular amount 2θ. The wavelength, λ, measured is given by the relationship:
λ=
d L R
where R is the radius of the Rowland circle and L the variable distance between the specimen and the analyzing crystal. The experimental arrangement for both WDXS and EDXS attachable to SEM/EPMA and STEM/TEM is shown schematically in Figure 18.3.
Figure 18.3 Schematic diagram of spectrometer arrangements for wavelength-dispersive and energy-dispersive X-ray spectroscopy (WDXS/EDXS) in electron microscopy.
297
298
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
One major disadvantage of the WDXS arrangement is the precise mechanical adjustment needed to meet the focusing conditions. Consequently, WDX spectroscopy can be complicated or even fail in the case of samples with high surface roughness. In order to cover the range of wavelengths of interest, a set of three or four crystals which differ in their lattice spacing are usually employed. For example, for an angular scanning, θ = 15–65 °, LiF crystals cover a wavelength range from 0.35 to 0.1 nm, which is equal to X-ray energies of 3.5–12.5 keV. X-Ray detection is usually achieved by use of a sealed proportional counter for shorter wavelengths, and a gas-flow counter for longer wavelengths (>0.7 nm). Owing to the particular type of X-ray detection, different experimental requirements must be met for interfacing an EDXS system to an electron microscope, compared to those of a WDXS arrangement (cf. Figure 18.3). Because of the specific geometry and dimensions of the specimen chambers, different experimental arrangements can be determined for SEM and TEM/STEM; the free space for inserting a semiconductor detector enabling EDXS is limited for the latter case. The situation is much more complicated when a conventional WDXS system needs to be adapted to (S)TEM, because the specimen is usually placed in between the upper and lower pole pieces of the objective lens (typical gap dimensions of 1–5 mm). In addition, the operation of a crystal spectrometer with its many moveable parts that generate vibrations would disturb the TEM function. It is for these reasons that transmission electron microscopes are usually not equipped with Rowland-circle spectrometers, but rather dedicated analytical scanning electron microscopes, which also enable STEM imaging, and EPMA. Moreover, recent developments of multicapillary X-ray (MCX) lenses [17] and flat-field grazingincidence optics [18] have enabled WDXS to be performed in combination with TEM/STEM. Because of the high-energy resolution of approximately 4 eV for MCX-WDXS and 0.6 eV and better for WDXS performed with grazing-incidence optics, a detailed thin-film phase analysis is possible; for the latter, the study of chemical bonding phenomena may even be included. For combined EDXS/(S) TEM, the detector should be attached to the microscope in an orthogonal position relative to the tilt axis of the specimen; this enables alignment of interface structures of cross-section specimens parallel to the detector axis, and hence permits X-ray analysis at a high lateral resolution. It is important to know the exact position and angle of the front side of the detector relative to region of the specimen contributing to the X-ray signal measured. The angle between the plane of the specimen and the detector axis is called the X-ray take-off angle, α. For a detector arrangement in the plane of the specimen and perpendicular to it, the take-off angle is simply the goniometer tilt. The ideal configuration is, however, a positive take-off angle (Figure 18.3), providing X-ray detection without any tilting of the specimen. In TEM, EDXS detectors are differentiated into two groups according to the take-off angle: (i) low take-off systems with approximately 10–20 °; and (ii) high take-off systems, in which the angle may reach up to 70 °. A second important setting which determines the detector efficiency is that of the collection angle, Ω; this is the solid angle subtended by the front face of the detector. In practice, this angle is given by the approximation [7, 17].
18.2 Practical Aspects of X-Ray Microanalysis and Instrumentation
Figure 18.4 Principal set-up of an X-ray detector with Li-drifted Si crystal.
Ω≈
A s2
where A is the active area of the detector and s the distance from specimen to the detector. For a detector with 30 mm2 area in 15 mm distance from the specimen – which are typical values for EDXS/(S)TEM – the collection angle is approximately 0.13 sr. In EDXS, the collection angles range from approximately 0.01 to 0.2 sr, and are thus more than one order in magnitude wider than for WDXS. As mentioned above, EDXS is based on the use of a semiconductor detector, which can be a lithium-drifted Si crystal [a so-called Si(Li) detector] or a Ge crystal of high purity, respectively (see Refs [5–7, 16–20]). For Si(Li), the lithium is introduced to compensate for impurity effects in Si, thus producing an intrinsic zone inside the crystal that is effective for X-ray detection. In general, inside the detector material electron–hole pairs are generated by X-ray excitation (Figure 18.4). The energy necessary to produce one electron–hole pair is approximately 3.8 eV for Si(Li), but only 2.9 eV for Ge. Because the energy of X-rays can be several keV and higher – depending on the primary electron energy and on the bombarded material – a single photon can generate thousands of electron–hole pairs. Thus, the number of electrons and holes is a direct measure of the energy of the incoming X-rays. The electron–hole pairs are separated by the effect of an electric field caused by a negative bias of approximately 0.5–1 keV applied across the Si crystal. Gold layers deposited on the front and back of the crystal serve as electrical contacts. The resulting electrical pulses are amplified, analyzed for pulse height, and finally stored in a multichannel analyzer (MCA) or computer. The detector crystal is cooled with liquid nitrogen to minimize the portion of thermally activated electron–hole pairs, to reduce the noise level of the associated field-effect transistor acting as a preamplifier and, in the case of Si(Li), to prevent the diffusion of Li. Thus, whilst in the past, the Dewar flasks required for cooling have been characteristic of these X-ray detector assemblies (see Figure 18.5a), Peltier cooling is today being applied on an increasing basis.
299
300
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
EDXS instrumentation for TEM and SEM. (a) EDXS detector (Kevex) with Si(Li) crystal attached to a 200 kV TEM/STEM of the type Hitachi H-8110; (b) Si(Li) detector
Figure 18.5
(Noran Instruments) on an SEM LEO 1530 (Gemini); (c) Si drift-chamber detector XFlash 5030 T for TEM. Illustration courtesy of Bruker AXS Microanalysis GmbH [23].
The photographic images shown in Figure 18.5 should provide an impression of the set-up of EDXS systems in practice. Figure 18.5a shows the column of a 200 kV TEM/STEM Hitachi H-8110, from the section of the objective lens (below) up to the segment containing the electron gun and accelerator (above), with a high take-off angle spectrometer [Si(Li) detector] attached to the microscope. In Figure 18.5b, the arrangement of an Si(Li) detector on an SEM can be seen. In contrast, Figure 18.5c shows an energy-dispersive X-ray spectrometer based on the use of a thermoelectrically cooled silicon drift detector (SDD) [21, 22], which can be installed on a TEM/STEM (see also Section 17.2). This system is particularly suited for handling high count rates, and thus enabling the acquisition of element maps at up to fourfold the speed of conventional EDXS detectors. Today, many instrument manufacturers offer EDXS and WDXS systems that can be attached to commercially available electron microscopes. The details of companies supplying surface and thin-film analytical equipment are listed in Appendix 2.
18.2 Practical Aspects of X-Ray Microanalysis and Instrumentation
Figure 18.6 Fraction of transmitted X-rays calculated for EDXS detectors with windowless Si(Li) and Ge crystals, and for different windows.
In order to prevent the condensation of hydrocarbons (and even ice formation) from the microscope environment onto the cold surfaces of the Si(Li) or Ge detector crystals, the latter are typically sealed in a pre-pumped tube. A window located in front of the crystal will then permit the transmission of X-rays, although if the electron microscope has a particularly good oil-free vacuum the X-ray detectors can be used without windows. Historically, the first such windows to be used were prepared from sheets of beryllium, typically about 12 μm thick or even thicker. Such high window thickness led to X-rays with energies <1 keV being strongly absorbed, thereby preventing the analysis of light elements such as B, C, N, and O. In an attempt to overcome this problem, ultrathin windows composed of lessabsorbing advanced materials have been developed, including very thin carbon (diamond-like) or boron nitride windows. The dependence of the absorption effect on the type of detector window is shown in Figure 18.6, where the transmitted fraction of X-rays is plotted against the photon energy for both the Si(Li) and the Ge crystals. In this case, detectors with and without windows (Be and diamondcoated with Al) were employed. Besides the influence of the window used, the individual curves also demonstrated the absorption edges of the detector material itself (i.e., Si and Ge), respectively. Whereas, the transmission of low-energy X-rays is insignificantly higher for Si(Li) compared to Ge, the efficiency of the Si(Li) detectors decreases remarkably above 20 keV. For the high-energy range, the application of Ge crystals is more useful, as this enables the detection of X-rays up to approximately 50 keV and higher, with 100% efficiency. The dependence of the portion of transmitted X-rays on the type of window is clearly visible; the windowless configurations, and partly also those with Al-coated diamond windows, prove to be feasible for the detection of light-element X-rays. The energy resolution of an X-ray detector is experimentally defined by the fullwidth at half maximum (FWHM) of the Mn-Kα line. The FWHM, in eV, can also be calculated by use of the relationship: FWHM = ( Δn 2 + 2.352 FεE + Δc 2 )
1/2
301
302
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
where Δn is the electronic noise contribution, Δc is that of incomplete charge collection, F is the Fano factor1), ε the energy to create an electron–hole pair, and E the initial X-ray energy [24]. Owing to the lower ionization energy for Ge, the number of charge carriers is approximately 25% higher than that in Si. Because, in addition, the Fano factor for Ge is approximately 0.06 compared to 0.087 for Si, Ge detectors will intrinsically afford a better energy resolution. The resolution is also dependent on the area of the detector surface – a broadening is observed with increasing area. It should, moreover, be noted that it is also a function of output count rates and the setting of the pulse processor; the best resolution values are obtained at rates less than 3000 counts per second, and at medium pulse processor times [25]. Typically, Si(Li) detectors have an average energy resolution of about 140 eV measured at the Mn-Kα line, whereas that of Ge detectors is approximately 125 eV. The resolution of Si drift detectors is also about 125 eV, but this is not limited by the count rate (rates up to 1 MHz can be handled). The use of microcalorimeter EDXS devices [26] offers energy-dispersive X-ray analyses with approximately 10 eV energy resolution, although their operating expense is relatively high in comparison to the other EDXS techniques. Because of the limited energy resolution in EDX spectra, an overlap of peaks can often occur, depending on the composition of the material to be analyzed. The situation is much improved in WDXS, for which the energy resolution is approximately 10 eV and better. This is demonstrated in Figure 18.7, in which the WDX and EDX spectra recorded from BaTiO3 are compared [7]. Here, WDXS enables an easier resolution of the Ba-Lα and Ti-Kα lines; this is impossible by EDXS. In addition, for WDXS the peak-to-background ratio is better. However, owing to the low X-ray collection efficiency compared with EDXS, the wavelength-dispersive
Figure 18.7 A WDX spectrum of BaTiO3 plotted against energy and compared with the
corresponding EDX spectrum [7]. 1) The Fano factor is defined as F = σ 2 W/μW , where σW and μW are the variance and mean of a signal in a given time window W.
18.3 Qualitative Spectral Information
technique is commonly not combined with TEM/STEM. Some additional advantages of the EDXS technique are the compact and stable instrumentation, its ease of use, and the short acquisition times as a result of parallel detection.
18.3 Qualitative Spectral Information
In EDX spectra, the X-ray intensity is usually plotted against energy. Typically, the spectra consist of several approximately Gaussian-shaped peaks characteristic of the elements present in the transmitted volume. The characteristic peaks are superimposed on a background (bremsstrahlung continuum) of distinctly low intensity, if the specimen is very thin. Most chemical elements can be identified using EDXS, the only limit being whether the particular type of detector window registers soft X-rays of the light elements (see Section 18.2). Thus, detectors equipped with Be windows enable the detection only of elements of atomic number Z ≥ 11 (Na), whereas detectors with ultrathin light-element windows can even be used for the analysis of boron (Z = 5). Problems in element detection may also arise as a result of peak overlap caused by the restricted energy resolution, as discussed above. An EDX spectrum typical of thin-film analysis in TEM/(S)TEM is shown in Figure 18.8. This was obtained from a polycrystalline TiC/ZrO2 ceramic by using an Si(Li) detector at 100 keV primary electron energy. For spectrum recording, the electron probe of approximately 1 nm diameter was focused on the triple junction between the grains in the STEM mode (Figure 18.8a). Besides the elements expected for the material under investigation (viz. Ti and Zr), Si, Fe, and Co were also detected, hinting at the presence of a (Fe,Co) silicide as an impurity. For ceramic materials it is known that such foreign phases are predominantly present in triple regions and in interlayers between the individual grains.
Figure 18.8 Typical X-ray spectra. (a) STEM bright-field image of a polycrystalline ZrO2/TiC ceramic with a triple junction; (b) corresponding EDX spectrum.
303
304
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
Figure 18.9 Artifacts in energy-dispersive X-ray spectra. Occurrence of (a) escape and (b)
sum peaks.
Depending on the measurement conditions chosen, there is some probability that artifacts will contribute to an EDX spectrum. These comprise effects that intrinsically are produced by signal detection (notably in the form of escape peaks) and internal fluorescence peaks and, in addition, those of signal processing, such as pile-up peaks. Escape peaks appear because an X-ray photon entering the Si(Li) detector can lose part of its energy by exciting the Si to fluoresce Si-Kα radiation (1.74 keV). Thus, an extra peak is produced at an energy of the characteristic peak reduced by 1.74 keV. In Figure 18.9a this phenomenon is demonstrated for the Cu-Kα line, where an escape peak of low intensity is also observed for the Cu-Kβ line. Of course, Ge crystals also generate escape peaks, but several peaks (Ge-Kα and Ge-Lα) can occur because of the electronic structure of Ge. The software for processing EDX spectra commonly uses routines to correct for escape peaks – their intensities are usually added to the original characteristic peaks. As a consequence of the internal fluorescence a corresponding peak can be seen in the spectrum at high count rates and long acquisition times. When the number of incoming X-ray photons is too high (>10 000 counts per second), the signal electronics can no longer differentiate between two single photons of a certain energy, and this leads to the occurrence of so-called pile-up or sum peaks, respectively, with exactly twice of the energy of a characteristic X-ray line. This is shown in Figure 18.9b for the Al-Kα peak.
18.4 Quantification
As it is usual in all spectroscopic methods, for quantification the original signal characteristic of an individual element inside the excited volume that can be correlated with the content of this element must be extracted from the raw-data
18.4 Quantification
spectrum. Hence, quantitative EDXS analysis involves determination of the background contribution, background subtraction, and counting the net intensities of the characteristic X-ray peaks. There are a number of different possible ways to fit the background curve underlying the whole EDX spectrum. For thin-foil analysis, it is an almost horizontal line of rather low intensity; consequently, one procedure used to determine the background signal below an X-ray peak is to assume a straightline contribution that is then subtracted. Another method of background extrapolation comprises averaging the background intensities in two windows of identical width just in front of and beyond the characteristic peak. It is also possible to model the background by using Kramers’ law [27], which defines the number NBack of bremsstrahlung photons as a function of the photon energy E: N Back = CZ
E0 − E E
where Z is the mean atomic number of the material and E0 is the primary energy of the electrons. The factor C, which must be found by modeling, includes the original Kramers’ constant and corrections for the efficiency of collection and processing by the detector and the effects of absorption of the excited X-rays within the specimen. It should be noted that additional procedures are available, such as top-hat filtering, which can also be applied to background elimination [28]. In Figure 18.10, an EDX spectrum of a layered Al/Cu sample is shown with the modeled background marked by a solid line. After background subtraction, the net X-ray peaks are usually fitted by Gaussian peaks and their intensities determined by integration. The procedure commonly used to quantify EDX spectra was originally outlined by Castaing [29], although for the general situation of investigating bulk materials. To a good approximation, it can be assumed that the concentration CSp of an element present in an unknown sample is related to the concentration CSt of the same element in a standard specimen by
Figure 18.10 Quantification of EDX spectra (background fit and subtraction).
305
306
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
CSp ISp =K CSt ISt where the corresponding X-ray intensities ISp of ISp emerging from the sample or standard, respectively, must be measured under identical experimental conditions. The factor K accounts for the different X-ray intensities generated inside a material and measured by the detector for both the standard and the specimen under investigation. Three contributions to K, which must be considered, are usually termed the ZAF correction (Z = atomic number, A = absorption, F = fluorescence) (e.g., Refs [5–7, 11, 30–37]). In detail, this procedure involves:
•
Z – correction for the different inelastic scattering properties introduced by differences between the mean atomic numbers of the specimen of interest and the standard.
• •
A – correction for differences between X-ray absorption. F – correction for corresponding X-ray fluorescence differences.
For an electron-transparent specimen the absorption and fluorescence correction parts can often be neglected; this is the so-called thin-film criterion introduced by Cliff and Lorimer [38]. Thus, for a thin specimen containing two elements A and B yielding the net X-ray intensities IA of IB, the concentration ratio reduces to: CA I = kAB A CB IB where the atomic-number correction factor, kAB, is also called Cliff–Lorimer factor. This factor varies with the accelerating voltage, and depends on the particular microscope and EDXS detector system used. The kAB factors needed for quantification are either calculated from first principles or determined experimentally by the use of reference materials (note: the latter procedure is more accurate). In order to obtain quantitative results of high accuracy and reliability, the influence of absorption and fluorescence effects should, however, also be taken into account for thin films; this procedure is described in detail elsewhere (see Refs [7, 39–41]). It should be noted that the minimum element concentration detectable by EDXS is approximately 0.01–1 atom%; this value depends on the element to be analyzed in a matrix, and is a function of instrumental settings determining the peak-to-background ratio.
18.5 Imaging of Element Distribution
The combination of EDXS and SEM/STEM enables the imaging of element distribution either along a line or two-dimensionally in a rectangular field of view. In general, line-profile analyses and element maps can also be obtained by combined SEM/WDXS. Thus, facilities for generating a fine electron probe and scanning it are essential to obtain information on element distributions. Usually, before
18.5 Imaging of Element Distribution
mapping of the element distribution, a spectrum of the whole area of interest must be acquired; the peaks of the elements sought are then selected by defining the energy windows. When the electron probe is scanned, the signals from the different windows are stored simultaneously on a computer, creating a line profile or even building up element-specific images. Careful interpretation of these elementselective profiles and images is sometimes advisable, because artifacts can arise from thickness variations, peak overlap, and the effects of absorption and fluorescence of X-rays. The lateral resolution achievable in EDXS line-profile analysis and maps is strongly determined by such factors as the size of the S(T)EM probe, and the thickness and composition (mean atomic number) of the specimen. For very thin specimens investigated in TEM/STEM systems, resolution values of several nanometers and better are attainable when electron probes of 1 nm in diameter or even smaller are used. In EDXS, the so-called spectrum-image method [42] can also be employed. Here, a series of spectra is taken from a scanned rectangular field, resulting in a data cube with its upper plane as the scanned x-y area and the third axis as the X-ray spectrum. Comprehensive information regarding the chemical composition and element distribution is extractable from this data set by subsequent processing. An example of interface analysis by EDXS line profiling at high lateral resolution is given in Figure 18.11. This is of particular importance, because the distribution of light elements such as carbon, nitrogen, and oxygen is also revealed. This was made possible by means of an Si(Li) EDXS detector (Kevex) with an ultrathin window attached to a dedicated STEM HB 501, from Vacuum Generators, with a cold field-emission cathode. The TEM bright-field image (Figure 18.11, top) shows an interfacial region of a fiber-reinforced SiC composite including the carbon fiber with a pyrolytically
Figure 18.11 EDXS line-profile analysis across the interfacial region of a C-fiber reinforced SiC
composite and corresponding TEM bright-field image (inset).
307
308
18 Energy-Dispersive X-Ray Spectroscopy (EDXS)
Figure 18.12 Imaging of element distribution by X-ray mapping. (a) Cross-section STEM bright-field image of an (In,Ga)As quantum wire (QWI) in InP; (b) The corresponding As-Kα map.
deposited graphitic layer, approximately 30 nm thick, and the SiC matrix [43]. EDXS line profiles were taken across this layer system in the STEM mode (1 nm probe diameter, 100 kV accelerating voltage). The individual line profiles (Figure 18.11, bottom) furnished detailed insight into the particular processes occurring during processing of the composite. Nitrogen was detected inside the graphitic fiber coating; this was introduced from the environment during chemical vapor deposition. The profile of the Si signal from the matrix in the direction of the fiber indicates, moreover, that silicon diffused towards the fiber. The local maximum of the Si content in the region of the graphitic fiber coating can be correlated to the formation of a compound, probably of the type SixNy or SixCyNz. Figure 18.12 is an example of X-ray mapping of an (In,Ga)As quantum wire structure using a TEM/STEM Philips CM20 equipped with a thermally assisted field-emitter and a Ge EDXS detector (Tracor Northern) [44]. The cross-section STEM bright-field image (Figure 18.12a) exhibits the crescent-like shape of this wire, which was grown on (001)InP by metal-organic chemical vapor deposition (MOCVD) [45]. Its width is approximately 240 nm, whereas the extension in the growth direction varies from approximately 10 to 20 nm. Despite of the low signalto-noise ratio of the corresponding As-Kα map the position of the quantum wire is clearly revealed.
18.6 Summary
The combined use of EDXS and TEM/STEM is a routine method of analytical electron microscopy that enables both qualitative and quantitative chemical analy-
References
sis of interfaces and interlayers, with high lateral resolution. A resolution of a few nanometers can be achieved, depending, on particular, on the size of the electron probe and on specimen thickness. Because EDXS is an especially highly sensitive means for detecting elements of medium and high atomic number, it is a useful complement to electron energy-loss spectroscopy (EELS), which is predominantly sensitive to light elements. In the STEM mode, the element distribution can be imaged as line profile or element map by use of characteristic X-ray lines.
References 1 Reimer, L. (1989) Transmission Electron Microscopy: Physics of Image Formation and Microanalysis, Springer-Verlag, Berlin. 2 Reimer, L. (1998) Scanning Electron Microscopy: Physics of Image Formation and Microanalysis, Springer-Verlag, Berlin. 3 Bethge, H. and Heydenreich, J. (eds) (1987) Electron Microscopy in Solid State Physics, Elsevier. 4 Amelincks, S., van Dyck, D., van Landuyt, J., and van Tendeloo, G. (eds) (1997) Electron Microscopy: Principles and Fundamentals, VCH Verlagsgesellschaft mbH, Weinheim. 5 Hren, J.J. and Goldstein, J.I. (eds) (1979) Introduction to Analytical Electron Microscopy, Plenum Press, New York. 6 Williams, D.B. (1984) Practical Analytical Electron Microscopy in Materials Science, Verlag Chemie International, Weinheim. 7 Williams, D.B. and Carter, C.B. (1996) Transmission Electron Microscopy – A Textbook for Materials Science, Plenum Press, New York. 8 Agarwal, B.K. (1991) X-Ray Spectroscopy, Springer-Verlag, Berlin. 9 Azaroff, L.V. (1974) X-Ray Spectroscopy, McGraw-Hill Book Company, New York. 10 Bearden, J.A. (1967) Rev. Mod. Phys., 39, 78. 11 Heinrich, K.F.J., Newbury, D.E., and Yakowitz, H. (1975) NBS Special Publication, vol. 460, U.S. Department of Commerce, Washington, D.C.. 12 Joy, D.C. (1995) Monte Carlo Modeling for Electron Microscopy and Microanalysis, Oxford University Press, New York. 13 Werner, U. and Johansen, H. (1987) Electron Microscopy in Solid State Physics
14 15
16
17 18 19
20
21 22
23 24
25
26 27
(eds H. Bethge and J. Heydenreich), Elsevier, p. 170. Johansson, T. (1933) Z. Phys., 82, 507. Bertin, E.P. (1975) Principles and Practice of X-Ray Spectrometric Analysis, Plenum Press, New York. Heinrich, K.F.J. (1981) Electron Beam X-Ray Microanalysis, Van Nostrand, New York. Tanaka, M., Takeguchi, M., and Furuya, K. (2008) Surf. Interface Anal., 40, 1684. Terauchi, M. and Kawana, M. (2006) Ultramicroscopy, 106, 1069. Joy, D.C., Romig, A.D., and Goldstein, J.I. (eds) (1986) Principles of Analytical Electron Microscopy, Plenum Press, New York. Reed, S.J.B. (1975) Electron Microprobe Analysis, Cambridge University Press, London. Gatti, E. and Rehak, P. (1984) Nucl. Instrum. Methods A, 225, 608. Strüder, L., Meidinger, N., Stötter, D., Kemmer, J., Lechner, P., Leutenegger, P., Soltau, H., Eggert, F., Rohde, M., and Schülein, T. (1999) Microsc. Microanal., 4, 622. Falke, M. and von Platen, A. (2009) Imag. Microsc., 11, 35. Lifshin, E., Ciccarelli, M.F., and Bolon, R.B. (1975) Practical Scanning Electron Microscopy (eds J.I. Goldstein and H. Yakowitz), Plenum Press, New York, p. 363. White, T.J., Cousens, D.R., and Auchterlonie, G.J. (1991) J. Microsc., 162, 379. Newbury, D.E. (2003) e-J. Surf. Sci. Nanotechnol., 1, 130. Kramers, H.A. (1923) Philos. Mag., 46, 836.
309
310
18 Energy-Dispersive X-Ray Spectroscopy (EDXS) 28 Statham, P.J. (2002) J. Res. Natl Inst. Stand. Technol., 107, 531. 29 Castaing, R. (1951) Application of electron probes to local chemical and crystallographic analysis, PhD thesis, University of Paris. (English translation by P. Duwez and D.B. Wittry, California Institute of Technology, 1955). 30 Duncamb, P. and Reed, S.J.B. (1968) The Calculation of Stopping Power and Backscatter Effects in Electron Probe Microanalysis (ed. K.F.J. Heinrich), NBS Special Publ. 298, Washington. 31 Philibert, J. (1962) Proceedings, Symposium on X-Ray Optics and Microanalysis, Stanford, p. 379. 32 Reed, S.B.J. (1965) Br. J. Appl. Phys., 16, 913. 33 Ziebold, T.V. and Ogilvie, R.E. (1964) Anal. Chem., 36, 323. 34 Philibert, J. and Tixier, R. (1968) in Quantitative Electron Probe Microanalysis (ed. K.F.J. Heinrich), NBS Special Publication 298, National Bureau of Standards, US Department of Commerce, Washington, DC, p. 13. 35 Beamon, D.R. and Isasi, J.A. (1970) Anal. Chem., 42, 1540.
36 Weinke, H.H., Malissa, H., Jr, Kluger, F., and Kiesel, W. (1974) Microchim. Acta Suppl., 5, 233. 37 Büchner, A.R. and Stienen, J.P.M. (1975) Microchim. Acta Suppl., 6, 227. 38 Cliff, G. and Lorimer, G.W. (1975) J. Microsc., 103, 203. 39 Horita, Z., Sano, T., and Nemoto, M. (1987) Ultramicroscopy, 21, 271. 40 van Cappellen, E. (1990) Microsc. Microanal. Microstruct., 1, 1. 41 van Cappellen, E. and Doukhan, J.C. (1994) Ultramicroscopy, 53, 343. 42 Jeanguillaume, C. and Colliex, C. (1989) Ultramicroscopy, 28, 252. 43 Schneider, R. and Woltersdorf, J. (1993) Proceedings, Analytical Transmission Electron Microscopy in Materials Science – Fundamentals and Techniques (eds J. Heydenreich and W. Neumann), Elbe Druckerei, Wittenberg. 44 Schneider, R., Kirmse, H., Hähnert, I., and Neumann, W. (1999) Fresenius J. Anal. Chem., 365, 217. 45 Kappelt, M., Grundmann, M., Krost, A., Türck, V., and Bimberg, D. (1996) Appl. Phys. Lett., 68, 3596.
311
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies P. Neil Gibson
19.1 Principles
X-ray diffraction (XRD) represents the classical method for determining the crystalline structure of solid materials. The basic principle of XRD – Bragg’s law – was expounded early during the twentieth century, and subsequently supplemented by theories regarding the detailed interpretation of XRD patterns. Over the past several decades, the instrumentation of XRD has evolved continuously, with more powerful and collimated X-ray beams and different diffraction geometries opening up possibilities for structural analysis in many different applications. Today, sophisticated software packages facilitate studies into determining the structure of very complex crystallographic systems. As research into surface modification has led to increasing numbers of applications in industry – especially of thin films – “grazing incidence,” or “glancing angle” geometries have been developed in order to render XRD more surface-sensitive. Currently, many XRD equipment manufacturers offer systems, or system attachments for “glancing angle” or “grazing incidence” XRD (GAXRD, GIXRD, GIXD, or GXRD). The aim of this chapter is to provide a broad outline of the two major grazing incidence XRD geometries, and to identify the differences between these and the standard θ−2θ diffraction method. The technique of grazing incidence X-ray reflectivity (GXRR, GIXR, GXR, or XRR) is also described, as the related theory is of central relevance here and is often applied on the same systems used for GAXRD. In order to complete the picture concerning grazing incidence X-ray structural analysis, a particular variant of X-ray absorption spectroscopy is briefly mentioned. However, this technique can only be effectively applied at synchrotron radiation sources, and a full description of the method and the enormous range of applications would require a lengthy article in itself. In this chapter GXRR and one of the two principal glancing angle diffraction techniques are described in more detail, as these are normally applied on laboratory X-ray systems. Since theoretical aspects are not
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
312
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
expounded in detail here, the reader is referred to the appropriate bibliography for further information. 19.1.1 The Grazing Incidence X-Ray Geometry
The basic concept of the grazing incidence X-ray geometry is straightforward. In order to reduce the penetration of X-rays into a surface, and thus limit the depth from which information is gathered, one simply reduces the angle of incidence ϕ of the beam onto the sample surface. At angles of incidence higher than a few degrees, the intensity I of X-rays within a material as a function of distance z from the surface is given by I = I0e-μz/sin(ϕ), where I0 is the intensity at the surface, and the absorption coefficient μ is dependent on the composition and density of the material. The depth at which the intensity has fallen to 1/e of its value at the surface is often termed the “1/e penetration depth” (z1/e), and is given by z1/e = sin (φ ) μ
(19.1)
At angles of incidence of less than a few degrees, the above relations are not valid. The index of refraction of solid and liquid materials at typical X-ray wavelengths is slightly less than unity; thus, at low angles of incidence refraction becomes significant, and below a critical angle ϕc total external reflection occurs [1], such that the effective penetration of the evanescent wave into the material decreases to a few nanometers. This only occurs for extremely flat and smooth surfaces; the case of rough surfaces is discussed below. The complex index of refraction n of a material is usually written n = 1 – δ – iβ, where δ and β are given by
δ = (λ 2e 2/2π mc 2 )
∑ N (Z + f ′) i
i
i
i
and
β = (λ 2e 2/2π mc 2 )
∑ N ( f ′′) i
i
i
Here, e and m are the charge and mass of the electron, λ is the X-ray wavelength, c is the velocity of light, Ni is the atomic number density of atoms of type i, Zi is the corresponding atomic number, and f′ and f″ are the real and imaginary parts of the anomalous dispersion correction to the atomic scattering factor. The absorption coefficient is related to β by μ = βλ/4π and ϕc is related to δ by ϕc ≅ (2δ)1/2. The equation for the reflection coefficient of a flat surface in air at an incident angle ϕ is given by I (φ − A )2 + B2 (19.2) R (φ ) = r (φ ) = I0 (φ + A )2 + B2 where 2 2 A 2 = ⎡⎣(φ 2 − 2δ ) + 4β 2 ⎤⎦
and
1/2
+ (φ 2 − 2δ )
19.1 Principles
Figure 19.1 Variation of the reflection coefficient and penetration depth for X-rays of 1.5405 (Å) incident on a perfectly flat silicon surface.
2 2B2 = ⎡⎣(φ 2 − 2δ ) + 4 β 2 ⎤⎦
1/ 2
− (φ 2 − 2δ )
Below angles of incidence of a few degrees, the 1/e penetration depth is given by z1/e ≅ λ 4 πB (for low angles of incidence)
(19.3)
Figure 19.1 shows how R and z1/e vary for a perfectly flat silicon surface. The rapid change in penetration depth near the critical angle shows how X-ray techniques become very surface sensitive at very low incident angles, and can theoretically be used to study extremely thin films, including atomic monolayers. However, for diffraction studies, the intensity of the diffracted beam is drastically reduced at incident angles below ϕc, and the range of materials that can be studied in this way with conventional sealed X-ray sources is limited. The reduction in intensity is related to the significantly limited dimensions of the incident beam, the greatly reduced volume of diffracting material, and the fact that most of the incident beam is specularly reflected. For very flat surfaces, it should be noted that the intensity of the evanescent wave (parallel to the surface) is increased by a factor of about four at an incident angle of exactly ϕc, with respect to its intensity at higher angles. Despite this, only powerful synchrotron sources or high-intensity laboratory sources may be used effectively for surface-sensitive diffraction studies at such low angles on most materials. To study the structure of the first few atomic layers of a material, the sample must also be exceptionally flat and smooth – single crystal semiconductor wafers are ideal. Some examples of such studies are given below. For slightly rough surfaces – for example, metals polished mechanically to a mirror finish, with an rms roughness of tens to hundreds of nanometers – the intensity of the reflected beam below ϕc is lower, and as the roughness increases so the reflectivity decreases. The non-reflected part of the beam penetrates the surface. It is difficult to assign a particular formula for calculating average penetration depth at this point, as it will be sensitive to details of the surface topography.
313
314
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
However, it is likely that for the purposes of diffraction analysis, Equation 19.1 rather than Equation 19.3 can be taken as a more accurate estimate in most cases, even if a significant fraction of the beam is being reflected. If an analysis is made of the reflected beam – as is the case for ReflEXAFS (see below) – then the surface sensitivity will still be rather good even for slightly rough surfaces, though poorly reflecting surfaces introduce significant experimental difficulties for such analysis and are best avoided. 19.1.2 Grazing Incidence X-Ray Reflectivity (GXRR)
It has been seen that partial reflection of an X-ray beam occurs from a surface at angles of incidence above ϕc, with the intensity dropping off rapidly with ϕ. In fact, partial or total reflection occurs at the interface between any two materials with different refractive indices. For a layered material these reflected beams can combine to form an interference pattern, an analysis of which can provide extremely accurate information concerning film thickness. In addition, the reflectivity coefficient at each interface depends on its roughness, as well the difference in composition and density on each side of the interface. Thus, the technique of GXRR – that is, the analysis of curves of R versus ϕ – can provide information on all these parameters. Kiessig [2, 3] was the first to note interference effects in X-ray reflectivity, and indeed interference fringes seen in GXRR curves are often referred to as “Kiessig fringes.” Although Kiessig derived a formula for the position of the interference maxima as a function of the film thickness, this is not normally used in GXRR analysis nowadays, it having been replaced by full mathematical models combined with least-squares fitting techniques that can be run on modern computers. The basic theory for modern GXRR analysis was developed by Parratt in 1954 [4]. 19.1.3 Glancing Angle X-Ray Diffraction
For details about general XRD theory, the reader is referred to the plethora of information on the subject that has been produced over the past few decades. Here, attention is focused only on the application of the technique in the glancing angle geometry, and the appropriate considerations that must be made. The basic aim of a diffraction measurement is to detect diffraction peaks from the material under investigation. Diffraction only occurs when the incident angle of the X-ray beam on a set of planes is such that (for a particular wavelength) Bragg’s law is satisfied and constructive interference occurs. The diffracted beam from that set of crystalline planes is at the same angle to those planes as the incident angle. Reversing this argument, it is immediately apparent that any diffraction peak will arise only from crystals or crystallites oriented in a particular manner. In the standard XRD geometry – also known as the θ−2θ, or Bragg-Brentano, geometry – the incident angle on the sample surface is maintained at the same angle
19.1 Principles
Figure 19.2 Schematic diagrams of three diffraction geometries. (a) θ-2θ; (b) GIAB; (c) GIXS.
as the detector. Thus, any crystalline planes detected will lie parallel to the surface of the sample. This is not the case for the two main glancing angle X-ray geometries – the grazing incidence angle asymmetric Bragg (GIAB) geometry [5] and the grazing incidence X-ray scattering (GIXS) geometry [6]. These are shown schematically in Figure 19.2, together with the standard θ−2θ geometry. It is worth mentioning in passing the asymmetric parafocusing Seeman–Bohlin geometry [7], which was originally developed as a surface-sensitive diffraction technique but is now not generally employed. This technique maintains a parafocusing geometry at low angles of incidence by moving the detector around a focusing circle defined by the X-ray source and the sample surface, the latter being at a tangent to the focusing circle. However, as the focusing circle increases in radius as the angle of incidence is lowered, the technique cannot be used below about ϕ = 5 °. A description is first given of the GIAB geometry, which is that most often applied using standard laboratory sources, and which is most widely available commercially. The above discussion has made it clear that the angle of the crystalline planes that contribute to any particular diffraction peak are inclined with respect to the sample surface, and that the angle of inclination changes as the detector is scanned. This makes the GIAB geometry much more suitable for studying polycrystalline surfaces than for studying single-crystal surfaces (e.g., epitaxial
315
316
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
films), since for single-crystal studies the orientation of the sample with respect to the incident beam is an important variable. Where polycrystalline surfaces with no texture (or preferential orientation of the crystallites) are under investigation, GIAB will produce a diffraction pattern very similar to that of identical material in bulk form examined by θ−2θ diffraction, apart from slight differences in peak intensities due to different absorption factors. However, thin-film materials are usually textured to some extent, and often very strong preferential orientations are apparent. In these cases, very different peak intensities will be recorded, and experimental techniques and procedures that can be employed in such cases are described below. With regard to penetration depths, the GIAB geometry applied at just above the critical angle for total external reflection is in the region of two orders of magnitude more surface-sensitive than the standard θ−2θ diffraction method, although it should be remembered that this occurs together with a significant loss in intensity due to the restricted dimensions of the incident beam, the lower penetration depth, and the nonfocusing geometry. At angles of incidence below ϕc, the penetration depth is only a few nanometers for very smooth surfaces, and most materials can only be effectively studied in this way with high-intensity sources. In contrast to the classical θ−2θ and GIAB geometries, the GIXS geometry is ideal for studying sets of crystalline planes that are perpendicular to the sample surface. Ideally, the incident angle and the detector elevation angle should both be set at the critical angle of the surface material. In this way, the refracted X-ray beam travels parallel to the surface and the surface evanescent wave intensity is maximized. The restrictions of the GIXS geometry greatly reduce the beam intensity at the detector, and thus GIXS is totally inappropriate for studying randomly oriented polycrystalline surfaces with a laboratory X-ray source. However, it can be used on strongly textured polycrystalline surfaces for determining texture or strain, although such measurements may be made in other ways. Where GIXS really comes into its own is in the study of single crystal surfaces and epitaxial films. With typical sealed laboratory sources, only strongly scattering materials can be studied in the GIXS geometry and it is desirable that several nanometers of material are present to achieve a good signal-to-noise ratio. With GIXS facilities at synchrotron radiation sources the structure of fractions of monolayers can be studied, even of relatively weakly scattering materials. 19.1.4 ReflEXAFS
The technique of extended X-ray absorption fine structure (EXAFS) [8, 9] is based on the analysis of the variations in the absorption coefficient of a material observed in a range of several hundred eV above the absorption edge of one of the atomic components of the material. Since the incident energy must be scanned in an EXAFS experiment, a high intensity over a wide energy range is required; hence, EXAFS is normally only applied at synchrotron radiation sources, where it is now
19.2 Experimental Techniques and Data Analysis
a commonplace technique. When an inner shell electron is ejected from an atom, the outgoing photoelectron has a wavelength that depends on its energy. Scattering from surrounding atoms occurs and effectively leads to constructive or destructive interference effects that modulate the measured absorption coefficient as the X-ray energy is varied. Analysis of these modulations involves background subtraction followed by analysis with an appropriate computer program in order to extract information concerning the distance and type of the surrounding atoms. EXAFS is a very powerful technique for local structural analysis around particular atomic species of a sample. It is often combined with analysis of the near-edge region, this being termed X-ray absorption near edge structure (XANES) or near-edge X-ray absorption fine structure (NEXAFS), whereby information is obtained on the oxidation state and the local environmental symmetry of the atomic species under investigation. Further information on X-ray absorption spectroscopy may be found in the extensive literature available on the subject. In order to render EXAFS surface-sensitive, two main methods are employed:
•
The current due to surface photoelectron yield may be monitored [10]. Since the escape depth of the photoelectrons is only a few nanometers, this is effectively the region from which information is obtained. This technique – which usually is referred to as surface EXAFS (SEXAFS), is normally performed with the sample in a vacuum chamber.
•
The technique may be employed in air or, indeed, in a special environmental chamber by using a grazing incidence geometry [11, 12]. Typically, an incident angle of ϕc/2 is used; analysis is then made of the intensity of the reflected beam as a function of incident beam energy. This technique is termed reflection EXAFS (ReflEXAFS). The reflected beam contains information only from the surface of the sample, even in the case of slightly rough surfaces. The technique can also be applied by monitoring the X-ray fluorescence instead of the reflected intensity [13]. This has advantages in some situations, but care should be taken if the sample is slightly rough, as the fluorescence detection method will rapidly lose its surface sensitivity as the reflectivity coefficient decreases.
19.2 Experimental Techniques and Data Analysis
In this section only GXRR and GIAB are considered, these being techniques that can be applied using sealed laboratory X-ray sources, and that are available commercially from several suppliers of X-ray equipment. Both, GIXS and ReflEXAFS are rather more specialized and need to be carefully studied before being utilized. In addition, they usually require high-intensity sources and are normally only employed at synchrotron radiation facilities, although rotating anode laboratory sources and even sealed tubes can be intense enough for some GIXS applications.
317
318
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
19.2.1 Grazing Incidence X-Ray Reflectivity (GXRR)
GXRR can only be applied to surfaces that are extremely smooth and flat; in fact, an rms-roughness of greater than a few nanometers will render the technique unusable. The technique is thus usually applied only to films deposited onto single-crystal silicon substrates, float glass, or similar suitable substrates, where the effective roughness can range from zero to a few Angstroms. It may also be applied to liquid surfaces. Due to absorption in the material under examination and/or the instrumental resolution of the measuring apparatus, for film thickness evaluation GXRR can only be applied to thin films or multilayers of less than a few hundred nanometers thickness. Thus, the range of applications is limited, and sample preparation is of critical importance. However, the information gained from GXRR is difficult or impossible to derive using other techniques, as the examples outlined below illustrate. Several experimental methods can be employed for GXRR. A very generalized schematic set-up is shown in Figure 19.3. For many applications, it is of great importance that the intensity of the reflected beam can be measured over several orders of magnitude; consequently, the background radiation must be reduced to an absolute minimum. On dedicated GXRR systems a monochromator is normally used in the incident beam to remove unwanted X-ray wavelengths and, in combination with slits, to collimate the incident beam. Often, the Kα1 emission line of a Cu anode source (1.5405 Å) is employed. Two concentric goniometers are generally used, with the sample mounted on one goniometer and the detector on the other, or alternatively with the source on one and the detector on the other. A slit is placed in front of the detector to define the angle of the reflected beam. Alternatively, a second monochromator/slit arrangement can be used for this purpose, with the advantage of further reducing unwanted background radiation. Special evacuated beam tubes may be inserted to reduce air scatter of the beam. Careful alignment of the whole system is required in order to ensure that the results are reliable. Any distortion of the curve will make analysis very difficult or unreliable. On several systems a somewhat less rigorous beam “conditioning” is employed. Although these systems may be completely adequate for most studies, they may
Figure 19.3 Schematic diagram of the experimental arrangement for GXRR measurements.
Beam conditioning optics may include slits, monochromators, Goebel mirrors, and evacuated beam tubes.
19.2 Experimental Techniques and Data Analysis
not achieve a good enough performance for particular applications. Monochromation may be limited to a monochromator crystal placed before the detector, and may allow both the Kα1 and Kα2 wavelengths to reach the detector. On at least one commercially available system an interesting solution for achieving both collimation and sample alignment has been used, with a simple “knife-edge” being positioned near the sample center. No incident beam monochromator is used, which has the advantage of simplifying alignment, in particular for small samples. An important development in recent years has been the use of Goebel mirrors on some systems to achieve high intensity, better collimation, and some degree of monochromation. As with any other expensive scientific instrument, before investing in a GXRR system it is important to consider its suitability for the types of sample to be investigated. GXRR data analysis is usually based on the theory developed by Parratt [4], who derived a recursive formula based on the Fresnel equation and Snell’s law of refraction, for the reflection coefficient of multilayered surfaces. This, however, could only be applied to perfectly flat surfaces or multilayers. Following the later investigations of Cowley and Ryan [14] and of Névot and Croce [15, 16], roughness could also be taken into account in the analysis. The approach of Cowley and Ryan was to apply a roughness-related Debye–Waller type factor to the Fresnel reflection coefficient from each interface. Névot and Croce, on the other hand, applied a correction factor based on a transition of optical parameters according to a Gaussian distribution of the roughness at each interface. Such a transition of optical parameters may be simulated in computer models by generating a series of sublayers according to an error function. This approach is usually the most accurate and realistic way of modeling roughness, but requires more computational power. However, it is easily implemented on modern computers and is often used nowadays. A good overview of theoretical models for GXRR is available in Ref. [17]. 19.2.2 Grazing Incidence Asymmetric Bragg (GIAB) Diffraction
The above description of roughness and penetration depth makes it clear that ultrasmooth samples are not necessary for GIAB, and great improvements in surface sensitivity may be achieved even with rather rough surfaces. Very rough surfaces will cause some experimental problems and uncertainty, however, as will curved surfaces, as the incident angle will not be well-defined and alignment may prove difficult. Samples therefore should be flat over an area of at least a few cm2 and not too rough – preferably they should appear at least slightly shiny. As mentioned above, the optimum incident angle and beam dimensions depend on the sample under investigation. A rough calculation should be made of the absorption coefficient of the material, so the incident angle can be set to achieve the desired penetration depth. The beam dimensions can then be adjusted to maximize the X-ray “footprint” on the sample, bearing in mind that the beam should only hit the surface of the sample, and the footprint size should be less than the aperture of both the Soller slit (this is a stack of parallel metal plates for collimating the
319
320
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
X-ray beam) and the detector active area (see Figure 19.2b) in order to avoid distortion of the collected XRD pattern. Such distortion may occur since the Soller slit and detector “see” more of the sample at low 2θ angles than they do at high 2θ angles. It should be borne in mind that, at low incident angles on very smooth surfaces, a correction should be made to the measured 2θ value [5] to account for refraction. Special experimental procedures are necessary for samples that exhibit a strong texture, or preferred crystallite orientation. If the texture is very strong it may occur that no crystalline planes are aligned with respect to the incident beam to satisfy the Bragg condition, and hence no diffraction peaks from the surface layer are observed. In such cases, the best procedure is to attempt a θ−2θ scan, as a very strong texture will render the diffraction peak or peaks from the preferentially oriented planes intense enough to be detected, even from rather thin films. However, this may fail if the film is too thin, or if the preferred orientation is associated with non-allowed diffraction peaks. An example of the latter case would be a preferred 111 orientation of a bcc metal such as iron or chromium, though the 222 peak should appear in any case, if the diffractometer can run to a high enough 2θ angle in the θ−2θ geometry. If the preferred orientation cannot be determined in the θ−2θ geometry, then the only procedure left is to use different incident angles in order to try to find an asymmetric geometry where the Bragg condition is satisfied for some set of planes. If the surface crystalline structure is not known, then this must be done by trial and error. If the probable crystalline structure is known, it is not difficult to calculate the appropriate angle of incidence for different sets of planes under the assumption of a particular preferred orientation. Weak texture in a surface layer will manifest itself in different peak intensity ratios than those of a randomly oriented material. Such cases can prove to be the most difficult with regard to the identification of the direction of preferred orientation, especially if the film is too thin for detection at high angles of incidence. It is possible [18] to make a calculation of how the relative peak intensities will be modified at glancing angles under an assumption of a certain direction and spread of orientations, though this is not always straightforward. The GIAB geometry is a non-focusing geometry, and requires a Soller slit between the sample and detector; however, this leads to a poorer instrumental resolution than that of most θ−2θ diffraction systems. A typical GIAB system may have a resolution of 0.1–0.2°. For simple phase identification this causes no real difficulty; however, if a line profile analysis is used to determine crystallite size or average microstrain [19, 20], then care should be taken in drawing conclusions, especially where the line broadening is of the same order of magnitude as the instrumental broadening. The often noisy nature of GIAB patterns only serves to compound the problem, since separating crystallite size from microstrain broadening requires reliable determination of peak shape as well as width. The very broad peaks associated with very fine-grained nanocrystalline materials, where the line broadening is much higher than the instrumental resolution or microstrain broadening, can usually be used to determine average crystallite size.
19.3 Applications
Macrostrain is often observed in modified surfaces such as deposited thin films or corrosion layers. This results from compressive or tensile stress in the plane of the sample surface, and causes shifts in diffraction peak positions. Such stresses can be easily analyzed by using standard techniques, if the surface layer is thick enough to detect a few diffraction peaks at high angles of incidence. If the film is too thin, these techniques cannot be used and the analysis can only be performed by assuming an unstrained set of lattice parameters and calculating the stress from the peak shifts, taking into account the angle of the detected sets of planes with respect to the surface (see discussion above). If the assumed unstrained lattice parameters are incorrect, not all peaks will give the same values. It should be borne in mind that, due to stoichiometry or impurity effects, modified surface films often have unstrained lattice parameters that are different from the same materials in bulk form. In addition, the thin film mechanical properties (Young’s modulus and Poisson ratio) may differ from those of bulk materials. Where pronounced texture and stress are present at the same time, analysis can be particularly difficult. A final complication with XRD of modified surfaces and thin films is that the crystalline structure, lattice parameters, strain and/or preferred orientation may vary as a function of distance into the surface. This may occur for a variety of reasons, such as the basic process of textured film growth, or the presence of a compositional gradient. Where this is suspected, it is worthwhile taking scans at several low angles of incidence in order to ensure that data interpretation is correct, bearing in mind that the X-rays penetrate into the surface according to the Equation 19.1 or Equation 19.3, so that information is gathered from a range of depths in each scan. The XRD data may be very difficult to interpret in some complicated cases.
19.3 Applications
In this section, examples of the application of the various techniques described are presented. The applications described have been selected simply to illustrate the wide range of areas of use, and in general early rather than more recent examples have been included. The details of many other examples are available in the literature. 19.3.1 Grazing Incidence X-Ray Reflectivity (GXRR)
X-ray reflectivity is now used routinely for surface and thin-film analyses. Applications include the simple determination of film thickness, multilayer structure, surface density determination, surface and buried interface roughness evaluation, structural studies of Langmuir–Blodgett films, and studies of ion-beam or thermal mixing. The first GXRR studies of Kiessig [2, 3] showed how film thickness could be determined from X-ray reflectometry measurements. In Parratt’s studies [4],
321
322
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
the theory was set out on how to apply the technique to multilayer samples in order to extract an electron density profile of the surface region. In 1959, X-ray reflectometry was used to determine the density of copper films [21], and in 1960 the annealing and oxidation of evaporated films of Cu, Ni, Ge, and Se was studied [22]. However, the technique remained relatively obscure until further investigations were conducted during the 1970s [15, 16], whereby surface and interfacial roughness could be taken into account in the analysis procedures. It was at this point that the potential of GXRR became apparent and the technique began to attract the attention of various research groups. In 1979, the diffusional alloying of Ag−Al thin film couples was studied [23], and the results supported a particular model for the alloying process and allowed the calculation of the atomic diffusion coefficients. In 1987, Cowley and Ryan studied the growth of thermal oxides on silicon wafers and presented their model for incorporating interfacial roughness effects into GXRR theory [14]. In 1988, Le Boité et al. studied the ion-beam mixing of Ni/Au and Ni/Pt multilayers [24], and in a separate report compared GXRR with Rutherford back-scattering as a method for studying ion-beam mixing of thin films [25]. An example of GXRR patterns of Ti/BN multilayers before and after ion-beam mixing [26] is shown in Figure 19.4. These curves can be analyzed using appropriate computer programs to monitor how the ion beam causes interfacial mixing and surface sputtering. GXRR has been used regularly to assess the quality or performance of multilayer optics for X-ray or X-UV systems. In 1988, Névot et al. [27] performed such a study on W/C multilayer stacks of up to forty periods of between 3 nm and 6 nm.
Figure 19.4 GXRR patterns of Ti/BN multilayers before and after ion-beam mixing with
various ion fluences (intensity ranges shifted for clarity).
19.3 Applications
Perhaps the most useful and potentially important application of GXRR is in the study of certain surface modifications that are difficult to structurally assess using other methods. Such systems include Langmuir–Blodgett films, extremely thin polymer layers, and the grafting of biomolecules onto surfaces in order to control their interaction with the environment. Examples of the former are studies of manganese stearate deposited on silicon wafers [28], of lead stearate films [29] and arachidic acid monolayers [30] directly on water, and of amphiphilic cyclodextrins on silicon [31]. The surface grafting of bioactive molecules such as heparin onto the surface of bioimplants in order to improve biocompatibility, or perhaps to form specialized sensors, has great potential for healthcare and sensor technologies, while ultrathin layers of various types have many applications in optics and electronics. GXRR is one of the few methods by which the structure of such surfaces can be studied. Many examples can be found in recent literature of the application of GXRR in such areas, where the deposition or grafting of extremely thin layers is of importance in creating nanostructured surfaces and thereby controlling their properties and their interaction characteristics with the environment. 19.3.2 Grazing Incidence Asymmetric Bragg (GIAB) Diffraction
Perhaps the most common use of a GIAB glancing angle XRD system is in the study of the structure of deposited solid films. It may also be employed in a variety of other studies, including surface corrosion and ion-beam modification. The technique was first described in 1987 by Lim et al. [5], who employed the high intensity of a synchrotron radiation source to apply the system at incident angles near and below ϕc for studying thin iron oxide layers; these authors also proposed a method for correcting the diffraction patterns for refractive index effects. GIAB studies of sputtered iron oxide layers had also been reported using a laboratory rotating anode system [32]. Subsequently, application of the technique at angles just above ϕc in order to maximize the diffracted intensity while maintaining reasonable surface sensitivity with sealed laboratory sources became more widespread. As an example of the application of the technique for studying thin corrosion layers, the case can be cited of yttrium ion-implanted NiCr that had been subjected to oxidation in air for 8 min at 700 °C [18, 33]. Standard θ-2θ diffraction patterns on this sample revealed diffraction peaks only from the face-centered cubic alloy substrate. Figure 19.5 shows GIAB patterns taken on the sample at four angles of incidence. At the higher angle of 0.5° oxide peaks indicating the presence of both Y2O3 and Cr2O3 near the surface were observed, whilst at 0.4° the relative intensities of the Y2O3 and Cr2O3 peaks changed. Reduction of the incident angle to 0.3° or even 0.2° revealed that the hexagonal Cr2O3 was present as a surface layer, while the Y2O3 was embedded in the alloy. Such low incident angles are actually below the critical angle for total external reflection of Cr2O3; however, the fact that the sample was mechanically polished before oxidation would have allowed a rather
323
324
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
Figure 19.5 GIAB depth profiling of yttrium ion-implanted NiCr that had been oxidized in air
for 8 min at 700 °C.
deeper penetration than for a perfectly smooth sample, thus increasing the diffracted intensity. Despite this, the reduction in diffracted intensity with decreasing incident angle was very apparent. The spectra shown required 15 h to collect. Analysis of the relative peak intensities revealed that the Cr2O3 had a pronounced 001-preferred orientation [18]. A subsequent Auger depth profile of the sample revealed that the Cr2O3 surface layer had a thickness of only a few hundred Angstroms. GIAB studies of sputtered thin films of various compositions for tribological applications have been reported [34–36]. GIAB has been used for studying the structure of very thin CdS layers (deposited by chemical bath deposition) for photovoltaic applications and, in combination with θ-2θ diffraction, allowed the identification of their polytype structure [37]. Glancing angle diffraction in the GIAB geometry is now extensively used for many different applications; the details of many other examples are available in the literature. 19.3.3 Grazing Incidence X-Ray Scattering (GIXS)
The first description of the GIXS technique was that of Marra et al [6], who presented a structural study of the interface between a GaAs single-crystal substrate and an epitaxial Al layer grown by molecular beam epitaxy (MBE). In this study, a 60 kW rotating anode source was used (GIXS normally requires high-intensity sources). Today, many synchrotron radiation sources have dedicated GIXS facilities, sometimes combined with specialized deposition chambers so that the growth
19.3 Applications
of thin epitaxial films can be studied in situ. A variety of such studies have been reported, with analysis even at the initial submonolayer growth stage; many of these are studies of MBE film growth. An example of a GIXS study of monolayers formed by a different mechanism is the analysis by Samant et al. of Pb monolayers on Ag(111) and Au(111) electrode/electrolyte interfaces [38]. Another important application of GIXS is in the structural analysis of reconstructed surfaces; examples of such studies include the analysis of Ge(001) and Au(110) surfaces [39, 40]. A GIXS analysis of the melting of Pb monolayers on Cu(110) surfaces has been reported [41]. GIXS analysis of the Si–SiO2 interface was reported by Fuoss et al. [42], and several other examples of GIXS applications have been described by Segmüller [43]. GIXS diffraction studies of Langmuir– Blodgett films [29], and also of oriented polycrystalline Cr2O3 layers [44], have been reported (the latter study employed a sealed laboratory X-ray tube). The details of many other examples of GIXS surface diffraction analysis are available in the literature. 19.3.4 ReflEXAFS
ReflEXAFS can often be used to acquire near-surface structural information that cannot be obtained using other techniques. As with EXAFS, ReflEXAFS is particularly suited for studying the local atomic structure around particular atomic species in noncrystalline environments. However, it is also widely applied for the analysis of nanocrystalline materials and for studying the initial stages of crystallization at surfaces or interfaces. The ReflEXAFS technique was first proposed by Barchewitz [11] and, following several reports during the early 1980s [12, 45–47], the method became an established (although rather exotic) characterization technique. Today, many synchrotron radiation sources have beamlines dedicated to ReflEXAFS experiments. ReflEXAFS studies of the passivation of Ni electrodes were reported by Bosio et al. [48]. Studies of the effect of oxygen on interfacial reactions in Al/Ni bilayers were made by Chen and Heald in 1989 [49], and the technique was applied to the field of long-term storage of nuclear waste by analysis of the local structure around uranium in leached waste-containing borosilicate glasses [50]. ReflEXAFS has also been used to study the oxidation of stainless steel [51] and to investigate the socalled “active element effect” in chromia-forming alloys by analysis of the local atomic structure around yttrium and chromium during the initial stages of oxidation of yttrium ion-implanted NiCr and Cr [52, 53]. A fluorescence detection of surface EXAFS was described by Heald et al. in 1984 [13]; an example of the application of this technique is the study of the local atomic structure around iron dispersed in polymer coatings on steel [54]. As with the other techniques described in this chapter, only a few examples of the huge variety of applications of ReflEXAFS have been mentioned here, and the details of many more are available in the scientific literature.
325
326
19 Grazing Incidence X-Ray Methods for Near-Surface Structural Studies
References 1 Compton, A.H. (1923) Philos. Mag., 45, 1121–1131. 2 Kiessig, H. (1931) Ann. Physik, 10, 715–768. 3 Kiessig, H. (1931) Ann. Physik, 10, 769–788. 4 Parratt, L.G. (1954) Phys. Rev., 95, 359–369. 5 Lim, G., Parrish, W., Ortiz, C., Bellotto, M., and Hart, M. (1987) J. Mater. Res., 2, 471–477. 6 Marra, W.C., Eisenberger, P., and Cho, A.Y. (1979) J. Appl. Phys., 50, 6927–6933. 7 Feder, R. and Berry, B.S. (1970) J. Appl. Crystallogr., 3, 372–379. 8 Lee, P.A. and Pendry, J.B. (1975) Phys. Rev. B, 11 (8), 2795–2811. 9 Eisenberger, P. and Kincaid, B.M. (1978) Science, 200, 1441–1447. 10 Eisenberger, P., Citrin, P., Hewitt, R., and Kincaid, B. (1981) CRC Crit. Rev. Solid State Mater. Sci., 10 (2), 191–207. 11 Barchewitz, R., Cremonese-Visicato, M., and Onori, G. (1978) J. Phys. C, 11, 4439–4445. 12 Fox, R. and Gurman, S.J. (1980) J. Phys. C, 13, L249–L253. 13 Heald, S.M., Keller, W., and Stern, E.A. (1984) Phys. Lett., 103A (3), 155–158. 14 Cowley, R.A. and Ryan, T.W. (1987) J. Phys. D, 20, 61–68. 15 Croce, P., and Névot, L. (1976) Revue Phys. Appl., 11, 113–125. 16 Névot, L. and Croce, P. (1980) Revue Phys. Appl., 15, 761–779. 17 Stoev, K. and Sakurai, K. (1997) Rigaku J., 14 (2), 22–37. 18 Crabb, T.A. (1993) Advanced surfacesensitive X-ray techniques: Application to the study of the initial stages of high temperature corrosion of metals and alloys. PhD thesis, University of Strathclyde, UK. 19 De Keijser, T.H., Mittemeijer, E.J., and Rozendaal, H.C.F. (1983) J. Appl. Crystallogr., 16, 309–316. 20 Enzo, S., Polizzi, S., and Benedetti, A. (1985) Z. Kristallographie, 170, 275–287. 21 Wainfan, N., Scott, N.J., and Parratt, L.G. (1959) J. Appl. Phys., 30, 1604–1609. 22 Wainfan, N. and Parratt, L.G. (1960) J. Appl. Phys., 31, 1331–1337.
23 Wagendristel, A., Bangert, H., and Tonsern, W. (1979) Surf. Sci., 86, 68–74. 24 Le Boité, M.G., Traverse, A., Névot, L., Pardo, B., and Corno, J. (1988) J. Mater. Res., 3 (6), 1089–1096. 25 Le Boité, M.G., Traverse, A., Névot, L., Pardo, B., and Corno, J. (1988) Nucl. Instrum. Methods B, 29, 653–660. 26 Raggio, M.E. (1994) Utilizzazione della spettroscopia di riflettanza a raggi X per lo studio di superfici modificate da bombardamento ionico. Degree thesis, Università degli Studi di Milano, Italy. 27 Névot, L., Pardo, B., and Corno, J. (1988) Revue Phys. Appl., 23, 1675–1686. 28 Pomerantz, M. and Segmüller, A. (1980) Thin Solid Films, 68, 33–45. 29 Bosio, L., Benattar, J.J., and Rieutord, F. (1987) Revue Phys. Appl., 22, 775–778. 30 Kjaer, K., Als-Nielsen, J., Helm, C.A., Tippman-Krayer, P., and Möhwald, H. (1989) J. Phys. Chem., 93, 3200–3206. 31 Schalchli, A., Benattar, J.J., Tchoreloff, P., Zhang, P., and Coleman, A.W. (1993) Langmuir, 9, 1968–1970. 32 Huang, T.C. and York, B.R. (1987) Appl. Phys. Lett., 50 (7), 389–391. 33 Crabb, T.A., Gibson, P.N., and McAlpine, E. (1993) Corros. Sci., 34 (9), 1541–1550. 34 Rickerby, D.G., Gibson, P.N., Gissler, W., and Haupt, J. (1992) Thin Solid Films, 209, 155–160. 35 Gissler, W. and Gibson, P.N. (1996) Ceram. Int., 22, 335–340. 36 Gissler, W., Baker, M.A., Haupt, J., Gibson, P.N., Gilmore, R., and Mollart, T.P. (1997) Diamond Films Technol., 7 (3), 165–180. 37 Gibson, P.N., Özsan, M.E., Lincot, D., Cowache, P., and Summa, D. (2000) Thin Solid Films, 361–362, 34–40. 38 Samant, M.G., Toney, M.F., Borges, G.L., Blum, L., and Melroy, O.R. (1988) J. Phys. Chem., 92, 220–225. 39 Eisenberger, P. and Marra, W.C. (1981) Phys. Rev. Lett., 46 (16), 1081–1084. 40 Robinson, I.K. (1983) Phys. Rev. Lett., 50 (15), 1145–1148. 41 Marra, W.C., Fuoss, P.H., and Eisenberger, P.E. (1982) Phys. Rev. Lett., 49 (16), 1169–1172.
References 42 Fuoss, P.H., Norton, L.J., Brennan, S., and Fischer-Colbrie, A. (1988) Phys. Rev. Lett., 60 (7), 600–603. 43 Segmüller, A. (1987) Thin Solid Films, 154, 33–42. 44 Gibson, P.N., Crabb, T.A., McAlpine, E., and Falcone, R. (1993) Mater. Sci. Forum, 126–128, 595–598. 45 Martens, G. and Rabe, P. (1980) Phys. Status Solidi, 58, 415–424. 46 Martens, G. and Rabe, P. (1981) J. Phys. C, 14, 1523–1534. 47 Goulon, J., Goulon-Ginet, C., Cortes, R., and Dubois, J.M. (1982) J. Physique, 43, 539–548. 48 Bosio, L., Cortes, R., Defrain, A., and Froment, M. (1984) J. Electroanal. Chem., 180, 265–271.
49 Chen, H. and Heald, S.M. (1989) Solid State Ionics, 32/33, 924–929. 50 Thornley, F.R., Barrett, N.T., Greaves, G.N., and Antonini, G.M. (1986) J. Phys. C, 19, L563–L569. 51 Barrett, N.T., Gibson, P.N., Greaves, G.N., Mackle, P., Roberts, K.J., and Sacchi, M. (1989) J. Phys. D, 22, 542–546. 52 Gibson, P.N. and Crabb, T.A. (1995) Nucl. Instrum. Methods Phys. Res. B, 97, 495–498. 53 Cristóbal, M.J., Gibson, P.N., and Stroosnijder, M.F. (1996) Corros. Sci., 38 (6), 805–822. 54 Pizzini, S., Roberts, K.J., Dring, I.S., Moreland, P.J., Oldman, R.J., and Robinson, J. (1992) J. Mater. Chem., 2 (1), 49–55.
327
329
20 Glow Discharge Optical Emission Spectroscopy (GD-OES) Volker Hoffmann and Alfred Quentmeier
Today, GD-OES is becoming one of the most important techniques for the direct analysis of solids, at their surface, through their depth, and in their bulk [1–4]. The ease and speed of operation, multielement capability (including light elements such as Li, Be, B, and C, and even the gases such as H, O, N, Cl), high sensitivity, accuracy of analysis and breadth of application make GD-OES a very versatile and powerful technique. Following the basic development of an efficient discharge cell by Grimm [5–7], the application of GD-OES began during the 1960s with the dc operation for the bulk analysis of metals. Later, the technique was used most frequently for the analysis of thick layers in the micrometer range, such as Zn or hard coatings. The introduction of the radiofrequency (rf) mode during the late 1980s [8, 9] led to an expansion of the application on the analysis of nonconducting materials. GD-OES has proven its capability, in particular, for the rapid and reliable analysis of metallic coatings, oxide scales, diffusion and alloy layers, and thus can be used successfully for the quality control of technical products during different stages of heat treatment, for example nitriding or nitrocarburizing steps, hard coatings, or physical vapor deposition (PVD) processes. Furthermore, an increasing number of applications for the analysis of layers in the nanometer range are today being reported [10–12]. Indeed, even the characterization of monolayers has now been proven possible [13].
20.1 Principles
Glow discharges as used in this context are low-pressure electrical discharges in a noble gas at pressures ranging from 100 to 1000 Pa [14]. It is well known that the glow discharge in the discharge chamber comprises several alternating dark and luminous zones. The complicated structure can be altered, however, depending on the geometry of the chamber, in particular the distance between the two electrodes. In general, when the electrode distance is relatively long, the positive column (the bright zone near the anode, often with striations) is the most Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
330
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
prominent. It is found that, when the space between the electrodes is reduced, the positive column shrinks and finally disappears, whereas the negative glow (the bright zone near the cathode) and the cathode dark space are hardly affected. Both the negative glow and cathode dark space are essential to maintain the discharge and cannot be omitted. Under these conditions the discharge is called an “obstructed” glow and can be realized when the inter-electrode separation is just a few times the thickness of the cathode dark space. All GD-OES instruments operate under obstructed conditions, with the analytical sample of interest serving as the cathode. During the discharge, the cathode is bombarded by positive ions and atoms mainly from the noble gas. This bombardment induces surface erosion called cathodic sputtering; in this case, atoms, electrons, and ions are removed from the surface of the material that constitutes the cathode. The sputtered material then participates in different processes occurring in the discharge, the result being the emission of photons corresponding to the characteristic spectral lines of the elements present within the plasma. By this means, the glow discharge furnishes information on the elemental composition of the analytical material being removed layer-by-layer from the cathode. The sputtering rates vary from 0.1 nm s−1 up to more than 100 nm s−1, depending on the cathode material and the excitation conditions used. GD-OES thus enables the rapid and detailed characterization of the near-surface layer, interfaces, and bulk composition, if the spectral line intensities of the elements of interest are recorded simultaneously.
20.2 Instrumentation 20.2.1 Glow Discharge Sources
A conductive material is commonly analyzed with a constant potential across cathode and anode (dc mode), whereas both conductive and nonconductive materials can be analyzed if the potential varies at radiofrequencies (rf mode). Both principles of excitation can be applied either continuously or in a pulsed mode. Only pulsed rf generators are available commercially, with repetition rates up to 10 kHz. All commercially used GD-OES sources available to date are based on the Grimm principle, originally designed for dc operation. Different modifications were later investigated to run this source also in the rf mode. The principle of the Grimm-type source is illustrated in Figure 20.1. Here, the anode is no longer a flat plate but rather a hollow tube, the annular front of which is spaced by only 0.1–0.2 mm from the surface of the sample. This restricts the discharge to a sample area which is equal to the inner aperture of the anode tube (typically 2–8 mm). When the discharge current is sufficiently high – that is, in the so-called “abnormal” glow regime – the whole area is covered by the discharge, which results in
20.2 Instrumentation
Figure 20.1 Diagram of a typical glow discharge source used for dc GD-OES analysis.
very efficient cathodic sputtering. The sample is mounted against the discharge chamber and so can be removed very easily. In order to maintain the low-pressure conditions required, the sample must be vacuum-tight and seal the discharge chamber by means of an O-ring. Normally, flat samples with sufficiently low surface roughness are used, although samples with other regular shapes (rods or wires) can also be used with special sample holders and modified discharge chambers. Irregular and nonvacuum-tight samples can be used if they are located within a special cup, which serves as a vacuum seal. Because of the short distance between the outer diameter of the anode tube and the inner diameter of the cathode plate (0.1–0.2 mm), sputtered sample material which is deposited partially in the gap between anode and cathode will cause an electric short-circuit after some time of operation. Therefore, the maximum layer thickness accessible by depth profiling is restricted typically to 0.1–0.2 mm. After each analysis, the deposited material inside the anode tube and in front of the anode is removed by a reamer and flushed out with a pressure surge of the discharge gas. The amount of eroded sample material transported by the gas flow and deposited in the ring slit is comparatively small, but difficult to remove. Therefore, modern GD-OES sources are equipped with an additional ceramic insert (mostly a ring) in the cathode to prevent sparking or short-circuits. The glow discharge sources are evacuated, usually by pre-vacuum pumps, although dry scroll pumps are preferred in order to avoid contamination with
331
332
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
backstreaming pump oil, as can occur with rotary pumps [10]. On occasion, a turbo molecular pump is used to reduce the contamination at the surfaces of source and sample before the discharge gas is introduced. Monitoring not only the characteristic pumping-down time but also the pressure increase curves allows the characterization of the state of the vacuum system and the vacuum integrity of the loaded sample [15]. During his early source developments, Grimm changed from a more simple vacuum system with a single pump [5] which evacuated the narrow space between the anode tube and the front (cathode) plate, to a differential pumping scheme with two evacuation lines [6, 7]. In this case, one pump was used to maintain the low gas pressure in the ring slit between cathode plate and anode and between the front of the anode tube and the sample. This is one of the key features of the original Grimm design that was missing in some other sources, which have also a hollow tube anode, but no reduced pressure between the anode and cathode. Due to the high breakdown voltage of gases (cf. the left side of the Paschen curve [16]), the combination of low pressure and small distances keeps the plasma away from the walls and thus from unwelcome sputtering of chamber material. The second pumping line maintains the operating pressure in the main source volume, and sucks about 90% of the total gas flow (≈0.1–1 l min−1). This permits stable, fast, and reproducible pressure regulation. For this pumping regime single and double pump systems are available on the market. A noble gas (mostly argon) is flushed continuously through the chamber, with the pressure inside the source during sputtering typically ranging from 100 to 1000 Pa and regulated by a mass flow controller or a fast valve. The pressure is monitored with a Pirani or a combined Pirani/Penning gage, respectively, if turbo-pump systems are employed. In general, a reamer is used to press the sample against the sealing ring, and a block with water cooling is optionally positioned between the sample and the reamer. There are also sources with water cooling for the anode and/or cathode. The optical radiation of the excited sample atoms is observed end-on through a window with the required transmittance. MgF2 is most commonly used as window material. Besides the conventional Grimm-type dc source, which has dominated the GD-OES scene for approximately thirty years, other discharge sources are well known. Among these are various boosted sources which employ either an additional electrode to achieve a secondary discharge [17], or a magnetic field [18–22] or microwave power [23–28] to enhance the efficiency of excitation and thus the analytical sensitivity. However, none of these sources has yet been applied to surface or depth-profile analysis, nor is commercially available. The introduction of rf-powered sources has extended the capability of GD-OES to nonconductors [29, 30]. This is of major importance for surface and depthprofile analysis, because there exists a multitude of technically and industrially important nonconductive coating materials (e.g., painted coatings and glasses) which are extremely difficult to analyze using any other technique. Continuous discharge and sputtering of the sample are accomplished by the negative dc self-bias voltage, or offset, which is acquired at the surface of the insulator without any external dc potential being supplied. This phenomenon is
20.2 Instrumentation
explained by the different mobilities of the positive ions and negative electrons in the discharge region, which reverse their direction of movement toward the sample with every half-cycle of the bipolar rf potential. The positive charge of the insulator surface during bombardment with ions is easily neutralized and shifted to negative charge by electrons in the next half-cycle; this results in the observed steady-state negative offset potential which enables the effective sputtering of the sample by positive ions. The dc bias voltage mentioned above is generated within only a few rf cycles at the surface of an insulator. Although conducting samples can also be analyzed with rf mode, the electrical reaction depends on the rf system used, and especially on the size of the coupling capacitor of the generator. One practical drawback of the rf sources is the influence of sample thickness. Basically, the thickness of a dielectric sample directly affects the voltage drop between front and backside of the sample. The sample acts as capacitor, whose capacity decreases and resistance increases with sample thickness, leading to reduced sputtering rates and analytical signals. However, this difficulty can be overcome by using a new rf source design, based on the Grimm device including current and voltage probes [31]. Of special interest are thin conductive layers which cover the surface of an insulator. These demonstrate an opposite effect when compared to thick dielectric samples; this effect can be exploited [32] and was investigated by Therese et al. [33]. Although, in the past several rf sources of different design have been developed (cf. compilation [34]), very few of them have become commercially available. Those which have been produced are all based on the Grimm design and allow both rf and dc operation. The first practical application of rf GD-OES was reported by Passetemps [8], whereby a nonconducting insert was placed into the former cathode plate and the rf applied from the reverse side of the sample – a technique referred to as back coupling. Partial or total replacement of the conductive parts of the cathode plate is essential to avoid high capacitive currents. This is especially important if “matched generators” are used, where the output impedance (50 Ω) of a fixedfrequency generator (typically running at 13.56 MHz) is adapted to the high impedance of the GD source (several kΩ) by a so-called “matchbox.” The modification of the former conductive cathode plate is useful also in case of free-running generators [35] without matching system, where changes in the plasma load are compensated for by a small frequency shift. Since high capacitive currents may cause heating of the cables and connectors, and also possibly disturb the measurement of plasma current [31], low frequencies (e.g., 3.4 MHz and 6.78 MHz) are preferred in freerunning systems, where the high rf voltage is applied directly by the generator. Mitchell et al. [36] modified the cathode plate in such a way that a conductive ring is used to supply the rf voltage from the front side; this method is known as front coupling. Conductive layers on insulating substrates thus can be analyzed by using rf GD-OES, without any problem. Moreover, there are other techniques, which combine both coupling types and include an optional dielectric material between the electrodes and the sample [37]. It has been reported that with this design version, the plasma parameters of the discharge are almost independent of the size and position of the sample [31].
333
334
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
In matched systems, the power is measured by a reflectometer between the generator and the matchbox; therefore, power losses in the matchbox are included, and must be subtracted [38]. 20.2.2 Spectrometer
Apart from the glow discharge source with its associated gas and power supplies, vacuum pumps, and controls, the optical emission spectrometer is the most important part of a GD-OES instrument. To make use of the analytical capability of the discharge source, the spectrometer must have sufficient spectral resolving power and an adequate spectral range. Depending on the application, a commercial spectrometer can cover the wavelength range from 110 to 800 nm, which incorporates the most sensitive lines of all elements including the light elements (Li, Be, B, C, etc.) and the gases (H, N, O, Cl). A high spectral resolution is required to avoid spectral interference with lines of other elements, and in particular of the discharge gas, argon. The most widely used spectrometer designs are the Paschen–Runge polychromator configuration for simultaneous spectrometers, and the Czerny–Turner monochromator for sequential analyzers. In the Paschen–Runge mount, the concave grating focuses the spectral line intensities of preselected elements onto fixed exit slits, which are positioned on the Rowland circle. Up to 64 elements are detected simultaneously in this way. In the Czerny–Turner mount, quasimonochromatic radiation is observed at the exit slit. Then, by turning the plane grating around its axis it is possible to scan a spectral range and to profile an emission line, including background radiation and neighboring spectral lines. Beside these well-known classical designs, nowadays a new spectrometer configuration has come into focus which is equipped with an échelle grating. As the échelle grating is used at large angles of incidence and diffraction, a high angular dispersion is available. The multitude of overlapping diffraction orders may be separated by an additional prism in the optical path, which is oriented approximately perpendicular to the dispersion of the grating. This results in the typical pattern of a two- dimensional (2-D), high-resolution échelle spectrum, which can be evaluated by using a solid-state detector of adequate size. 20.2.3 Signal Acquisition
The optical radiation passing the exit slit(s) of a conventional spectrometer is recorded by a detector system. Today, the best-known detector is still the photomultiplier tube (PMT), which is implemented in most commercial spectrometers. In the common polychromator system, each detection channel is equipped with a PMT and a corresponding high-voltage supply. Each supply can be adjusted individually to afford appropriate sensitivity for each element required to match the analytical conditions. Because of the linear characteristics of the PMT, which
20.3 Spectral Information
is operated in analog mode, the output signal is proportional to the radiation power of the spectral line recorded with each detection channel. The analog output signals are amplified, converted into digital information, and transferred to a PC for further data processing. For depth profiling and continuous sample erosion, it is necessary to use a fast A/D converter, because the acquisition rate must be high (≈1 kHz) to ensure sufficient depth resolution. The rapid development in the performance of the charge-coupled device (CCD) and related detection techniques – for example, the charge-injection-device (CID) – makes these detector systems very attractive, especially when combined with the well-known compact échelle spectrometer. This configuration results in a small multichannel spectrometer with a high resolving power [39]. When CIDs are coupled with échelle spectrometers, the systems have very desirable properties, such as a high sensitivity and a wide dynamic range with very low read noise and almost nonexistent dark current. The random access of individual detector sites enables the flexible selection of spectral lines, whereas the new collective readout mode will promise a faster readout and improved signal-to-noise ratios (SNRs) required for depth-profile analysis [40].
20.3 Spectral Information
The emission process in a glow discharge source depends on several obvious factors: the number of atoms in the plasma; the availability of energetic particles such as high-energy electrons and metastable argon atoms which provide the excitation energy by numerous collision processes; and the excitation cross-sections of the competing energy levels of the elements. It can be estimated that the number of argon atoms in the discharge exceeds the number of sputtered sample atoms by a factor of approximately 103. Because only a small fraction of sputtered atoms is charged and energetically excited, the discharge plasma is (for a given sample material) dominated mainly by argon atoms. This explains why calibration curves obtained with GD-OES are usually linear over a wide range of elemental concentration, and matrix effects are less important. The separation of sputtering and excitation in space and time is also essential for the low matrix effects in GD-OES. The light-emitting material from the sample is highly dissolved in argon, and provides no more information about former neighbors in the solid sample. The operating conditions of the discharge – that is, low pressure and low power input – result in low kinetic gas temperatures of approximately 500–1000 K, and emission lines with small physical widths. The spectral information obtained from the glow discharge source is, therefore, highly specific and selective. With constant excitation conditions, the spectral line intensities are proportional to the number densities of sputtered atoms in the plasma and, hence, the top element concentration in the sample used. To maintain reproducible excitation conditions in the glow discharge source, the working conditions (e.g., argon pressure, dc current or rf power) are carefully controlled.
335
336
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
The voltage-current characteristic, which describes the excitation conditions, depends on the pressure and the matrix. Therefore, at different matrices it is only possible to keep two discharge parameters constant. It turns out that the pressure has the lowest influence on the light emission and therefore the voltage and current are kept constant, in dc mode. In the rf mode the measurement of the plasma current is difficult and therefore other parameters, like constant power and constant voltage, are preferred.
20.4 Quantification
Due to the complex nature of the discharge conditions, GD-OES is a comparative analytical method, and standard reference materials must be used to establish a unique relationship between the measured line intensities and the elemental concentrations. In quantitative bulk analysis, which has been developed to very high standards, calibration is performed with a set of calibration samples of composition similar to the unknown samples. Normally, a major element is used as reference and the internal standard method is applied. However, this approach is not generally applicable in depth-profile analysis, because the different layers encountered in a depth profile often comprise widely different types of material, which means that a common reference element is not available. The quantification algorithm most commonly used in dc GD-OES depth profiling is based on the concept of constant emission yield [41–43] Rik, according to the observation that the emitted light of element i at the spectral line k per sputtered mass of element i per time (i.e., emission yield) is an almost matrix- and pressureindependent constant for each element, if the source is operated under constant excitation conditions of voltage and current. In this approach, the observed line intensity Iik is described by the concentration ci of element i in the sample j, and by the sputtering rate qj, which is the total sputtered mass of sample j per time: Iik = c i q jRik
(20.1)
The sputtered mass δmi of element i during the time increment δt is described by: δmi = Iik δt/Rik
(20.2)
The emission yield Rik must be determined independently for each spectral line, and therefore the sputtering rates qj and the line intensities Iik must be measured for a variety of different standard bulk samples with known concentrations ci. In practice, both sputtering rates and emission yields are influenced by the working conditions of the discharge, but are constant at constant electrical discharge parameters. According to Boumans [44], the sputtering rate qj depends only on voltage U and linear on current I, but not on pressure: q j = CQ I(U − U 0 )
(20.3)
The threshold voltage U0 is typically about 300 V, and CQ is a material-specific constant that is related to the probability of a sample atom being ejected during
20.5 Depth Profiling
the sputter process. Systematic variation of the discharge voltage U and current I leads to the empirical intensity expression [45]: Iik = K ik c iCQ I Ak f k (U )
(20.4)
where Kik is an atomic- and instrument-dependent constant characteristic of the spectral line k of element i, Ak is a matrix-independent constant characteristic of the spectral line k, and fk(U) is a polynomial of degree 1–3, also characteristic of the spectral line k. The most important reason for the success of the emission yield technique is that the total sputtered mass is easily determined by summing over all the elements present in each depth segment. Thus, the time-dependent concentration of all elements can be calculated as c i = δmi /Σδmi
(20.5)
A model of the density ρ in dependence on the concentrations ci of the elements i now permits the estimation of the sputtered depth: δz = Σδmi /( rA )
(20.6)
In this manner, intensities are transformed into concentrations or coating weight, and time transforms into depth. This emission yield approach has proven to be extremely successful for a large and increasing number of applications, and is currently implemented in all commercial GD-OES depth-profiling instruments. In contrast to the dc operation, in the rf mode two electrical discharge parameters, which characterize the matrix-dependent impedance of the plasma, cannot be measured in all cases as accurately as is necessary for analytical application. It has, however, been shown that the concept of matrix-independent emission yields can be used for quantitative depth-profile analysis with rf GD-OES, if the measurements are performed at constant impedance [46] or corrections for the variation of the impedance are included in the quantification algorithm [47].
20.5 Depth Profiling
The primary information obtained in GD-OES depth-profile measurements is the relative intensity from the elemental detection channels as a function of sputtering time. The intensity–time curves obtained for different elements can be converted into concentration–depth curves by applying Equations 20.1–20.6. The depth is determined from the sputtered mass, which is the quantity obtained by the emission-yield technique, while the density of the composite material is calculated using an approach based on a weighted average of the density of pure elements. This method provides very accurate results for all types of metal alloys, but tends to be less accurate for compounds which contain light elements (oxides, nitrides, carbides, etc.). Here, an in situ depth measurement system [48] has great advantage but, mainly for commercial reasons, is not yet available on the market. In general, the accuracy, precision, and reproducibility of quantitative GD-OES depth-profile analyses are assured on the basis of inter-laboratory round-robin tests
337
338
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
Figure 20.2 Depth (temporal) profile obtained on a multilayer coating produced by plasma
vapor deposition (PVD) using optimized rf-glow discharge conditions. Layer thickness: 10 × (300 nm CrN/300 nm CrAlN)/100 nm Cr/steel [59].
[49, 50], by use of a layered certified reference material [51–53], and evaluated by comparison with wet chemical analysis [50] or other surface techniques [54, 55]. The information depth achieved by using the GD technique is determined, in principle, by the depth of penetration of the incident ions, which is in the range of a few nanometers at the relatively low energies employed in the discharge. The practical depth resolution is, however, always smaller and determined by several effects which are introduced by the sputter process (preferential sputtering, atomic mixing), the sample properties (surface roughness, polycrystalline structure) and, most importantly, by the nonuniform erosion of the sample material [56]. Most of these effects are inherent also in other techniques which use sputtering for surface and depth-profile analysis. In practice, the depth resolution obtained on technical surfaces is roughly proportional to the sputtered depth, and usually deteriorates from the nanometer-range for near-surface layers to the micron-range at depths of several micrometers [57, 58]. Excellent depth resolution is realized on multilayer coatings (cf. Figure 20.2), as produced by vacuum evaporation or by PVD [59]. The lateral resolution of the continuous GD-OES technique, on the other hand, is restricted by the size of the sputtered area of the sample surface (usually 2–8 mm diameter) [60], and is much larger than for other surface techniques. The lateral and depth resolution of GD-OES are, however, usually adequate for the rapid quantitative determination of the elemental composition of technical surfaces. The application of pulsed discharges improved the lateral resolution of GD-OES [61, 62], which may be used in the future to discriminate the crater edge effect. Both, pulse length and frequency have also a direct influence on the crater shape. However, pulsed discharges today are mainly used to reduce the stress for thermally sensitive samples, such as layers on glass.
20.6 Applications
20.6 Applications
Glow discharges as used in GD-OES offer exciting analytical capabilities: shortterm sample-to-sample reproducibility (typically 0.1–1%) for the determination of major and minor elements [63], and detection limits for most elements in the range of 0.1 to 10 μg g−1. These figures of merit make glow discharges very powerful atomization and excitation sources which have found wide analytical and technical application. Besides the field of spectrochemistry, where GD has gained renewed interest in bulk analysis [64–67], it is applied extensively as an ion source in glow discharge mass spectrometry (GDMS) [68–71]. Those applications involving depth-profile analysis are the economically most important use of GD-OES today, and will probably become even more important in the future [72]. In comparison with other surface analytical techniques, GD-OES has the advantage of easy operation, rapid in-depth analysis, wide depth range, and sensitivity to all elements in the Periodic Table. For these reasons, GD-OES depth profiling is of primary interest for industrial use, especially for the quality control of largescale technical products. The numerous applications [3] cover the field of different coating technologies of metallic surfaces, the investigation of corrosion effects, and reactive diffusion processes in layer–bulk interfaces. Some typical applications of the GD-OES technique for depth-profile analysis of conducting and of nonconducting materials are summarized in Table 20.1. Table 20.1
Some typical applications of GD-OES depth-profile analysis.
Coating type
Surface treatment
Reference(s)
Metallic coatings
Galvanized coatings on steel Electroplated Zn coating on steel Electroplated ZnNi coating on steel Hot-dipped Zn coating on steel AlZn alloy coating on steel Nitriding, carburizing TiN,TiC (PVD) layers Oxide scale on alloyed steel Painted automotive components SiC on Si3N4 ceramics SiO2/MgO coating on steel Anodic alumina films on aluminum C-Co-Cr layers on a computer hard-disk Ni-P-plated aluminum Cu/Cr-Ni superlattice material on silicon Cr/Al multilayers on steel Ti/Al multilayers on aluminum alloy Cu on Si Cr/Ti on Si Brass layer on glass
[73–75] [49, 73, 76–79] [49, 76, 79, 80] [49, 76] [74, 75, 80] [72, 81] [52, 73, 82–84] [80, 85] [74, 77, 86, 87] [35] [88] [89, 90] [72, 91, 92] [72, 92–94] [52, 88, 95] [59] [51] [96–99] [100, 101] [102]
Hard coatings Oxide scales Polymer coatings Ceramic coatings Thin layers
Coated glass
339
340
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
Figure 20.3 Quantitative depth profile of TiN-coated steel [83].
20.6.1 dc GD Sources
In the past, GD-OES with dc sources was mainly used for steel analysis, and the qualitative characterization of surfaces of cold-rolled and hot-rolled steels, to control the enrichment or depletion of minors and traces during the process of manufacturing or the formation of oxide scales [103–105]. Zinc-based coatings are currently of great industrial importance in building materials and materials for car manufacture. Due to the wide industrial application of hot-dip galvanizing and electroplating technology, the rapid and reliable determination of the thickness of the zinc coatings (typically ≈5–30 μm) and the amounts of major elements in these layers [45, 75, 106, 107] represents a major challenge to surface- and depthprofiling methods. Today, GD-OES has proven its capability in this field and is accepted as a standard method for the quantitative depth-profile analysis of zincbased metallic coatings [49]. Other typical applications of GD-OES include the analysis of surface-hardened steel after the process of nitriding or carburizing [81], and after different hard-coating processes, for example of TiN layers by PVD, chemical vapor deposition (CVD), or particle-aided chemical vapor deposition (PACVD) [3] (Figure 20.3). In addition to the analysis of relatively thick metallic coatings and diffusion profiles, dc GD-OES has also been successfully applied to the analysis of thin protective layers (thickness typically <100 nm), for example, phosphate and chromate layers on steel [45]. 20.6.2 rf GD Sources
If the rf source is applied to the analysis of conducting bulk samples, its figures of merit are very similar to those of the dc source [108]. This is also shown by comparative depth-profile analyses of commercial coatings on steel [74, 109]. The capability of the rf source is, however, unsurpassed in the analysis of poorly or
20.6 Applications
Figure 20.4 Quantitative depth profile of a commercial pre-painted, metal-coated steel. (a) As recorded; (b) Quantitative [74].
nonconducting materials, for example anodic alumina films [89, 90], CVD-coated tool steels [83], composite materials such as ceramic-coated steel [88], coated glass surfaces [102], and polymer coatings [74, 77, 86]. These coatings are used for automotive body parts, and consist of a number of distinct polymer layers on a metallic substrate. The total thickness of the paint layers is typically more than 100 μm. An example of a quantitative depth profile on pre-painted, metal-coated steel is shown in Figure 20.4. In conclusion, GD-OES is a very versatile analytical technique which is still in a state of rapid technical development [110, 111]. In particular, the introduction of rf sources for nonconductive materials has opened up new areas of application. The further development of more advanced techniques, for example, pulsed glow
341
342
20 Glow Discharge Optical Emission Spectroscopy (GD-OES)
discharge operation combined with time-gated detection [112], is likely to improve the analytical capabilities of GD-OES in the near future.
References 1 Marcus, K.R. (1993) Glow Discharge Spectroscopies, Plenum Press, New York. 2 Marcus, K.R. and Broekaert, J.A.C. (2003) Glow Discharge Plasmas in Analytical Spectroscopy, John Wiley & Sons, Ltd, Chichester. 3 Payling, R., Jones, D.G., and Bengtson, A. (1997) Glow Discharge Optical Emission Spectrometry, John Wiley & Sons, Inc., New York, Weinheim, Brisbane, Singapore, Toronto. 4 Nelis, T. and Payling, R. (2003) Glow Discharge Optical Emission Spectroscopy, A Practical Guide, Royal Society of Chemistry, Cambridge, UK. 5 Grimm, W. (June 1967) DE 1 589 389. 6 Grimm, W. (1967) Naturwissenschaften, 54, 586. 7 Grimm, W. (1968) Spectrochim. Acta B, 23, 443. 8 Chevrier, M. and Passetemps, R. (June 1987) FR 2 616 545 – A1. 9 Marcus, K.R. (1989) US 0407 030 A1. 10 Angeli, J., Bengtson, A., Bogaerts, A., Hoffmann, V., Hodoroaba, V.-D., and Steers, E.B.M. (2003) J. Anal. Atom. Spectrom., 18, 670. 11 Hoffmann, V., Dorka, R., Wilken, L., Hodoroaba, V.-D., and Wetzig, K. (2003) Surf. Interface Anal., 35, 575. 12 Pisonero, J., Fernández, B., Pereiro, R., Bordel, N., and Sanz-Medel, A. (2006) Trends Anal. Chem., 25, 11. 13 Shimizu, K., Payling, R., Habazaki, H., Skeldon, P., and Thompson, G.E. (2004) J. Anal. Atom. Spectrom., 19, 692. 14 Chapman, B. (1980) Glow Discharge Processes – Sputtering and Plasma Etching, John Wiley & Sons, Ltd, pp. 77–138. 15 Klemm, D., Hoffmann, V., and Edelmann, C. (2008) Vakuum in Forschung und Praxis, 20, 30. 16 Paschen, F. (1889) Ann. Phys. Chem., 37, 69. 17 Lowe, R.P. (1976) Spectrochim. Acta B, 31, 257. 18 Cho, W.B., Kim, M.Y., and Kim, H.J. (2000) Microchem.. J., 64, 41.
19 Chen, M., Ren, J., Ma, H., and Zhang, G. (1997) Spectrochim. Acta B, 52, 1161. 20 Heintz, M.J., Galley, P.J., and Hieftje, G.M. (1994) Spectrochim. Acta, 8, 745. 21 Kruger, R.A., Bombelka, R.M., and Laqua, K. (1980) Spectrochim. Acta, 35B, 581. 22 Kruger, R.A., Bombelka, R.M., and Laqua, K. (1980) Spectrochim. Acta, 35B, 589. 23 Su, Y., Yang, P., Chen, D., Zhang, Z., Zhou, Z., Wang, X., and Huang, B. (1997) J. Anal. At. Spectrom., 12, 817. 24 Duan, Y., Li, Y., Du, Z., Jin, Q., and Olivares, J.A. (1996) Appl. Spectrosc., 8, 977. 25 Leis, F. and Steers, E.B.M. (1996) Fresenius J. Anal. Chem., 7/8, 873. 26 Steers, E.B.M. and Leis, F. (1989) J. Anal. Atom. Spectrom., 4, 199. 27 Leis, F., Broekaert, J.A.C., and Laqua, K. (1987) Spectrochim. Acta B, 42, 1169. 28 Ülgen, A., Gökman, A., and Leis, F. (2001) J. Anal. Atom. Spectrom., 16, 1154. 29 Winchester, M.R. and Payling, R. (2004) Spectrochim. Acta B, 59, 607. 30 Pisonero, J., Costa-Fernández, J.M., Pereiro, R., Bordel, N., and Sanz-Medel, A. (2004) Anal. Bioanal. Chem., 379, 617. 31 Wilken, L., Hoffmann, V., Uhlemann, H.-J., Siegel, H., and Wetzig, K. (2003) J. Anal. Atom. Spectrom., 18, 646. 32 Fernández, B., Martín, A., Bordel, N., Pereiro, R., and Sanz-Medel, A. (2006) Anal. Bioanal. Chem., 384, 876. 33 Therese, L., Ghalem, Z., Guillot, P., and Belenguer, P. (2006) Anal. Bioanal. Chem., 386, 163. 34 Winchester, M.R. and Marcus, K.R. (1990) J. Anal. Atom. Spectrom., 5, 575. 35 Hoffmann, V., Uhlemann, H.-J., Präßler, F., Wetzig, K., and Birus, D. (1996) Fresenius J. Anal. Chem., 355, 826. 36 Mitchell, J.C. (July 1993) US 5408315. 37 Analytis, M. (October 2000) DE 19915582 A1.
References 38 Wilken, L., Hoffmann, V., and Wetzig, K. (2007) Spectrochim. Acta B, 62, 1085. 39 Becker-Ross, H., Florek, S., Franken, H., Radziuk, B., and Zeiher, M. (2000) J. Anal. Atom. Spectrom., 15, 851. 40 Pennebaker, F.M., Jones, D.A., Gresham, C.A., Williams, R.H., Simon, R.E., Schappert, M.F., and Denton, M.B. (1998) J. Anal. Atom. Spectrom., 13, 821. 41 Bengtson, A., Eklund, A., Lundholm, M., and Saric, A. (1990) J. Anal. Atom. Spectrom., 5, 563. 42 Weiss, Z. (2006) Spectrochim. Acta B, 61, 121. 43 Bengtson, A. and Nelis, T. (2006) Anal. Bioanal. Chem., 385 (3), 568. 44 Boumans, P.W.J.M. (1972) Anal. Chem., 44, 1219. 45 Bengtson, A. (1994) Spectrochim. Acta B, 49, 411. 46 Wilken, L., Hoffmann, V., and Wetzig, K. (2005) Anal. Bioanal. Chem., 383, 424. 47 Bengtson, A. and Hänström, S. (1998) J. Anal. Atom. Spectrom., 13, 437. 48 Wilken, L., Hoffmann, V., and Wetzig, K. (2003) J. Anal. Atom. Spectrom., 18, 1133. 49 Bengtson, A., Hänström, S., Lo Piccolo, E., Zacchetti, N., Meilland, R., and Hocquaux, H. (1999) Surf. Interface Anal., 27, 743. 50 Weiss, Z. and Smid, P. (2000) J. Anal. Atom. Spectrom., 15, 1485. 51 Winchester, M.R. and Beck, U. (1999) Surf. Interface Anal., 27, 930. 52 Hodoroaba, V.-D., Hoffmann, V., Steers, E.B.M., Griepentrog, M., Dück, A., and Beck, U. (2006) J. Anal. Atom. Spectrom., 21, 74. 53 Hodoroaba, V.-D., Klemm, D., Reinholz, U., Strub, E., Röhrich, J., Bohne, W., Hoffmann, V., and Wetzig, K. (2008) J. Anal. Atom. Spectrom., 23, 460. 54 Ives, M., Lewis, D.B., and Lehmberg, C. (1997) Surf. Interface Anal., 25, 191. 55 Shimizu, K., Habazaki, H., Skeldon, P., Thompson, G.E., and Wood, G.C. (1999) Surf. Interface Anal., 27, 1046. 56 Oswald, S., Hoffmann, V., and Ehrlich, G. (1994) Spectrochim. Acta B, 49, 1123. 57 Quentmeier, A. (1994) J. Anal. Atom. Spectrom., 9, 355.
58 Shimizu, K., Habazaki, H., Skeldon, P., Thompson, G.E., and Wood, G.C. (1999) Surf. Interface Anal., 27, 950. 59 Hodoroaba, V.-D. and Wirth, T. (1999) J. Anal. Atom. Spectrom., 14, 1533. 60 Hoffmann, V. and Ehrlich, G. (1995) Spectrochim. Acta B, 50, 607. 61 Gamez, G., Ray, S.J., Andrade, F.J., Webb, M.R., and Hieftje, G.M. (2007) Anal. Chem., 79, 1317. 62 Webb, M.R., Hoffmann, V., and Hieftje, G.M. (2006) Spectrochim. Acta B, 61, 1279. 63 Weiss, Z. (2001) J. Anal. Atom. Spectrom., 16, 1275. 64 Harville, T.R. and Marcus, K.R. (1995) Anal. Chem., 7, 1271. 65 Winchester, M.R. and Maul, C.L. (1997) J. Anal. Atom. Spectrom., 12, 1297. 66 Winchester, M.R. (1998) J. Anal. Atom. Spectrom., 13, 235. 67 Marshall, K.A. (1999) J. Anal. Atom. Spectrom., 14, 923. 68 Harrison, W.W., Hess, K.R., Marcus, K.R., and King, F.L. (1986) Anal. Chem., 58, 341A. 69 Jakubowski, N., Stuewer, D., and Toelg, G. (1986) Int. J. Mass Spectrom. Ion. Proc., 71, 183. 70 Hoffmann, V., Kasik, M., Robinson, P.K., and Venzago, C. (2005) Anal. Bioanal. Chem., 381, 173. 71 Matschat, R., Hinrichs, J., and Kipphardt, H. (2006) Anal. Bioanal. Chem., 386, 125. 72 Nelis, T. and Pallosi, J. (2006) Appl. Spectrosc. Rev., 41, 227. 73 Payling, R., Michler, J., and Aeberhard, M. (2002) Surf. Interface Anal., 33, 472. 74 Payling, R., Jones, D.G., and Gower, S.A. (1993) Surf. Interface Anal., 20, 959. 75 Payling, R. and Jones, D.G. (1993) Surf. Interface Anal., 20, 787. 76 Perez, C., Pereiro-Garcia, R., BordelGarcia, N., and Medel, A.S. (2000) J. Anal. Atom. Spectrom., 15, 1247. 77 Marcus, K.R. (2000) J. Anal. Atom. Spectrom., 15, 1271. 78 Bengtson, A. (1996) J. Anal. Atom. Spectrom., 11, 829. 79 Pisonero, J., Pérez, C., Pereiro, R., Bordel, N., and Sanz-Medel, A. (2001) J. Anal. Atom. Spectrom., 16, 370.
343
344
20 Glow Discharge Optical Emission Spectroscopy (GD-OES) 80 Xhoffer, C. and Dillen, H. (2003) J. Anal. Atom. Spectrom., 18, 576. 81 Rose, E. and Mayr, P. (1989) Mikrochim. Acta, 97, 197. 82 Böhm, H. (1997) Hard coatings, in Glow Discharge Optical Emission Spectrometry (eds R. Payling, D.G. Jones, and A. Bengtson), John Wiley & Sons, Ltd, pp. 676–687. 83 Payling, R., Aeberhard, M., and Delfosse, D. (2001) J. Anal. Atom. Spectrom., 16, 50. 84 Böhm, H. (1994) Oberflächen Werkstoffe., 4, 8. 85 Nickel, H., Fischer, W., Guntur, D., and Naoumidis, A. (1992) J. Anal. Atom. Spectrom., 7, 239. 86 Fernandez, M., Bordel, N., Pereiro, R., and Sanz-Medel, A. (1997) J. Anal. Atom. Spectrom., 12, 1209. 87 Bengtson, A. (2003) J. Anal. Atom. Spectrom., 18, 1066. 88 Hartenstein, M.L., Christopher, S.J., and Marcus, K.R. (1999) J. Anal. Atom. Spectrom., 14, 1039. 89 Shimizu, K., Habazaki, H., Skeldon, P., Thompson, G.E., and Marcus, K.R. (2001) Surf. Interface Anal., 31, 869. 90 Shimizu, K., Habazaki, H., Skeldon, P., Thompson, G.E., and Wood, G.C. (2000) Surf. Interface Anal., 29, 155. 91 Yang, C., Ingeneri, K., Mohill, M.L., and Harrison, W.W. (2000) J. Anal. Atom. Spectrom., 15, 73. 92 Shimizu, K., Habazaki, H., Skeldon, P., and Thompson, G.E. (2003) Spectrochim. Acta B, 58, 1573. 93 Shimizu, K., Habazaki, H., Skeldon, P., Thompson, G.E., and Wood, G.C. (2000) Surf. Interface Anal., 29, 151. 94 Luesaiwong, W. and Marcus, K.R. (2004) J. Anal. Atom. Spectrom., 19, 345. 95 Präßler, F., Hoffmann, V., Schumann, J., and Wetzig, K. (1995) J. Anal. Atom. Spectrom., 9, 677. 96 Klemm, D., Stangl, M., Peeva, A., Hoffmann, V., Wetzig, K., and Eckert, J. (2008) Surf. Interface Anal., 40, 418.
97 Stangl, M., Acker, J., Hoffmann, V., Gruner, W., and Wetzig, K. (2006) Microchim. Acta, 156, 159. 98 Stangl, M., Acker, J., Dittel, V., Gruner, W., Hoffmann, V., and Wetzig, K. (2005) Microelectr. Eng., 82, 189. 99 Stangl, M., Dittel, V., Acker, J., Hoffmann, V., Gruner, W., Strehle, S., and Wetzig, K. (2005) Appl. Surf. Sci., 252, 158. 100 Galindo, E.R., Fornies, E., and Albella, J.M. (2005) J. Anal. Atom. Spectrom., 20, 1116. 101 Galindo, E.R., Fornies, E., and Albella, J.M. (2005) J. Anal. Atom. Spectrom., 20, 1108. 102 Marcus, K.R. (1996) J. Anal. Atom. Spectrom., 9, 821. 103 Koch, K.H., Kretschmer, M., and Grunenberg, D. (1983) Mikrochim. Acta, 2, 225. 104 Berneron, M.R. and Charbonnier, J.C. (1981) Surf. Interface Anal., 3, 134. 105 Janssen, E. (1980) Mater. Sci. Eng., 42, 309. 106 Biele, P. (1994) Der Einfluß von Aluminium und Eisen auf die Zinkbadmetallurgie und die Schichtlegierungsbildung beim kontinuierlichen Feuerverzinken von Stahlband, PhD thesis, RWTH Aachen. 107 Blumenthal, G., Schneider, W., Plieth, W., Baunack, S., and Hoffmann, V. (1999) Mo. Galvanotechnik, 10, 10. 108 Pan, X., Hu, B., Ye, Y., and Marcus, K.R. (1998) J. Anal. Atom. Spectrom., 13, 1159. 109 Pérez, C., Pereiro, R., Bordel, N., and Sanz-Medel, A. (2000) J. Anal. Atom. Spectrom., 15, 67. 110 Jakubowski, N., Dorka, R., Steers, E.B.M., and Tempez, A. (2007) J. Anal. Atom. Spectrom., 22, 722. 111 Jakubowski, N., Steers, E.B.M., and Tempez, A. (2007) J. Anal. Atom. Spectrom., 22, 715. 112 Bengtson, A., Yang, C., and Harrison, W.W. (2000) J. Anal. Atom. Spectrom., 9, 1279.
345
21 Surface Analysis by Laser Ablation Roland Hergenröder and Michail Bolshov
21.1 Introduction
Laser ablation (LA) is a popular and widespread sampling technique for direct solid-sample analysis. Several excellent monographs and review articles on the fundamentals of LA and its analytical applications have been published during the past three decades [1–5]. In contrast with most of the techniques used for surface analysis, LA is a “destructive” technique in which focused laser radiation is used to release material from a solid sample. The main goal of LA is analysis of the elemental composition of a solid sample. The material is removed by evaporation, or a type of phase explosion, from the overheated spot of a sample. Evidently, under such conditions information on the structural composition of a sample surface and the nature of the binding of the components is lost. During the past three decades, the majority of effort has been invested in the development of LA as a technique for bulk analysis. Typical values of the masses ablated in a single laser shot vary from several tens of nanograms to a few micrograms, depending on the type of sample, which corresponds to the thickness of an ablated layer of several hundreds of nanometers to a few micrometers. These probing depths of LA are much larger than for most typical surface analytical techniques, such as SIMS, SNMS, and TXRF. The first results from LA analysis of thin multicomponent and multilayer samples, with better depth resolution than that achieved hitherto, were reported only relatively recently. The advantages of LA are now well-known: no sample preparation is needed; both conducting and nonconducting samples of arbitrary structure can be analyzed directly; a spatial resolution of up to a few microns can be obtained; highvacuum conditions are not required; rapid simultaneous multielement analysis is possible; and it is possible to obtain complete analytical information with a single laser pulse. A brief overview of the potential and limitations of LA is provided in this chapter.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
346
21 Surface Analysis by Laser Ablation
21.2 Instrumentation 21.2.1 Types of Laser
To date, powerful lasers with picosecond to nanosecond pulse duration have normally been used for the ablation of material from a solid sample. The very first results from the application of lasers with femtosecond pulse duration were reported only quite recently. The ablation thresholds vary within a rather wide interval of laser fluences of 0.1 to 10 J cm−2, depending on the type of a sample, the wavelength of the laser, and the pulse duration. Different advanced laser systems have been tested for LA: 1) 2) 3)
Nd:YAG laser (1064 nm) with its harmonics (532 nm, 354.7 nm, 266 nm, 213 nm), Excimer lasers XeCl (308 nm), KrF (248 nm), ArF (193 nm), F2 (157 nm), and Ti:Sapphire laser (800 nm) with a pulse duration around 100 fs with its higher harmonics (400 nm, 266 nm).
Furthermore, specialized highly computerized systems (i.e., New Wave, Cetac, Geolas) based on the above-mentioned laser types that combine beam delivery, eye-saving viewing optics, and motorized XYZ-stages and sampling chambers are available. The systems are also provided with the appropriate gas tubing for the transport of ablated material into an ICP-OES/MS. These complete solutions are developed as specialized systems for coupling with the ICP customer’s systems. One very important characteristic of laser radiation is the beam shape. To date, most LA experiments have been performed with Gaussian laser beams. Lasers with uniform distribution of the beam cross-section have been used only recently to achieve high lateral and depth resolution. Specially designed beam homogenizers must be used for this purpose [6]. Today, most systems operate with a uniformly illuminated, image aperture, and consequently a uniform energy distribution is projected onto the sample surfaces. A clear advantage of such optics compared to a focusing system is that the fluence – that is, the energy per area – is held constant when the laser spot size on the sample is changed. A disadvantage, however, is the need for stronger and more expensive lasers. Although, in principle, lasers with all wavelengths can be used for reproducible and accurate sampling, the experimental conditions – such as the laser intensity and pulse length, the buffer gas and buffer gas pressure, and the delay and duration of the gate width for data acquisition – must each be optimized for a specific excitation wavelength. Recently, there has been a trend toward using UV-lasers, because of a greater absorption of the UV-radiation by most solid materials and a lesser absorption of the UV-laser radiation by the plasma above a sample surface. Both methods enable a more efficient coupling of the laser energy to the solid sample. Consequently, the lasers most widely used nowadays for LA are the fourth
21.2 Instrumentation
harmonic of Nd:YAG (266 nm) and ArF (197 nm), KrF (248 nm), and XeCl (308 nm) excimer lasers. The disadvantage of lasers with nanosecond–picosecond pulse duration for depth profiling is the predominantly thermal character of the ablation process [7]. For metals, the irradiated spot is melted and much of the material is evaporated from the melt. Melting of the sample causes the modification and mixing of different layers, followed by changes of phase composition during material evaporation (preferential volatilization) and bulk resolidification [8, 9]; this reduces the lateral and depth resolution of LA-based techniques. If a laser pulse of sub-picosecond duration is used, deposition of the laser energy to the sample occurs so rapidly that the thermal diffusion length is determined by the diffusion of hot electrons before they transfer the energy to the lattice of the solid sample. Less-pronounced thermal diffusion provides a better lateral and depth resolution, and is the basis of the successful application of femtosecond pulses in material processing and microstructuring [10, 11]. 21.2.2 Different Schemes of Laser Ablation
Different analytical techniques are used to detect the elemental composition of the solid samples. The simplest is the direct detection of emission from the plasma of the ablated material formed above a sample surface; this technique is generally referred to as laser-induced breakdown/plasma spectroscopy (LIBS or LIPS). A strong continuous background radiation from the hot plasma plume does not enable detection of atomic and ionic lines of specific elements during the first few hundred nanoseconds of plasma evolution. However, a reasonable signal-to-noise ratio (SNR) can be achieved for the measurement of atomic and ionic spectral line intensities by optimization of the experimental conditions, namely the type of laser and laser intensity, the type and pressure of buffer gas, and the time delay and distance from the sample surface for data acquisition. Detection limits (DLs) in the range of mg g−1 to μg g−1 have been achieved for the direct analysis of metals, glasses, and ceramics by using LIBS techniques for a variety of trace elements. Yet, better DLs were achieved when LA was followed by detection of the ablated atoms or ions of a selected analyte by laser atomic absorption spectroscopy (LAAS) or by laser-induced fluorescence (LIF) spectroscopy. Although, matrix effects are significant and the detection limits are inferior to those for more complex hyphenated techniques, LIBS has one major advantage, namely that it enables rapid online or remote analysis in an industrial environment, including the characterization of solid samples in hazardous zones. The use of fibers for both excitation and emission collection also enables the construction of a very robust system that allows on-line quality control. The intrinsic drawback of LIBS is a short duration (less than a few hundred microseconds) and the strongly nonstationary conditions of a laser plume. Much higher sensitivity has been realized by transporting the ablated material into secondary atomic reservoirs, such as a microwave-induced plasma (MIP) or an inductively coupled plasma (ICP). Owing to the much longer residence time of ablated
347
348
21 Surface Analysis by Laser Ablation
atoms and ions in a stationary MIP (typically several milliseconds, compared to at most a hundred microseconds in a laser plume), and because of the additional excitation of the radiating upper levels in the low-pressure plasma, the line intensities of atoms and ions are greatly enhanced. Due to these factors, the DLs of LA–MIP have been improved by one to two orders of magnitude compared to LIBS [12]. The sensitivity, accuracy, and precision of solid-sample analysis have been greatly improved by coupling LA with ICP-OES/MS. In this case, the ablated species are transported by means of a carrier gas (usually argon) into the plasma torch. Further atomization, excitation, and ionization of the ablated species in the stationary, hot plasma result in a dramatic increase in the sensitivity of the detection of radiation (LA-ICP-OES) or of the detection of ions (LA-ICP-MS). DLs in the sub-mg g−1 to ng g−1 range for different materials (metals, glasses, polymers) have been realized when using LA-ICP-OES. Even higher sensitivity was achieved for LA-ICP-MS, due to the high efficiency of ion collection and detection. Under optimum ablation conditions (choice of gas, diameter of the crater, flow rate of the carrier gas), DLs in the ng g−1 to pg g−1 range are now routinely realized for most elements of the Periodic Table. For example, the 3σ DLs listed in the LSX-200 prospectus vary from 10 ppt for Ho to approximately 14 ppb for Fe. The list of DLs includes 31 elements, 26 of which can be detected with DLs in sub-ppb range. As a result of its attractive characteristics, LA-ICP-MS is currently being used for a large variety of applications, for example, in situ trace-element analysis of glasses, geological samples, metals, ceramics, polymers, and atmospheric particulate material. The fingerprinting (characterization of the traceelement composition) of diamonds, gold, glass, and steel by using LA-ICP-MS was reported to be very useful for provenance determination and for forensic purposes. Today, LA-ICP-MS is accepted as a most powerful technique for direct analysis of solid samples. The analytical capabilities of LIBS and LA-MIP-OES were recently noticeably improved by the use of an advanced detection scheme based on an échelle spectrometer, combined with a high-sensitivity intensified charge-coupled device (ICCD) detector. This combination enables the simultaneous detection of a large spectral range, from the UV to the near-infrared, in a single laser shot. It also enables estimation of the temperature of a laser plume by constructing Boltzmann plots and correction for plasma temperature variations. The advantages of this technique are: (i) complete sample analysis in a single laser shot; (ii) improved accuracy and precision; and (iii) the possibility to detect sample inhomogeneities [12]. A lateral resolution of a few micrometers can be achieved with focused laser beams in the UV and visible spectral ranges.
21.3 Depth Profiling
Multilayer coatings of different composition and thickness are widely used in the materials sciences, and in the production of high-technology materials. The single-
21.3 Depth Profiling
or multi-component thin layers significantly improve the important characteristics of the materials with, for example, specific properties. Improvement of the technology of such advanced materials requires appropriate methods for surface and depth-profile analysis in the nanometer to micrometer range. Most of the methods used for surface analysis are reviewed elsewhere in this book. For example, X-ray photoelectron spectroscopy (XPS) and secondary ion mass spectrometry (SIMS) are widely used for the characterization of nanometerthin layers. Neither technique can be applied directly for the depth profiling of thicker layers (a few micrometers thick), nor hardly applied for rapid on-line process analysis, however. An alternative technique used for depth profiling is that of dc- or rf-glow discharge (GD) sputtering, followed by detection of the sputtered material by optical emission spectrometry (GD-OES) or mass spectrometry (GDMS) [13–15]. GD sputtering enables a low sputtering rate and provides a depth resolution of approximately 10 nm. The limitations of GD include: (i) a poor lateral resolution (at best a few millimeters); (ii) specific requirements on the form and dimensions of the sample; and (iii) a need for low-pressure conditions for sample sputtering. Laser ablation has been proven to be a reasonable alternative technique for direct solid sampling [16–21]. The potential of LA-based techniques for the depth profiling of coated and multilayer samples has been reported recently; notably, the depth profiling of zinccoated steels by LIBS has been demonstrated [22]. In this case, an XeCl excimer laser with 28 ns pulse duration and variable pulse energy was used for ablation. The emission of the laser plume was then monitored using a Czerny–Turner grating spectrometer with a CCD two-dimensional (2-D) detector. The dependence of the intensities of the Zn and Fe lines on the number of laser shots applied to the same spot was measured, and the depth profile of the Zn coating constructed by using the estimated ablation rate per laser shot. In order to obtain the true Zn–Fe profile, the measured intensities of both analytes were normalized to the sum of the line intensities. The LIBS profile thus obtained correlated very well with the GD-OES profile of the same sample. Both profiles are shown in Figure 21.1. The ablation rate of approximately 8 nm per shot was estimated on the basis of the number of shots required to reach the Zn–Fe interface located 12 μm beneath the sample surface. The depth-profile analysis of titanium-based coatings on steel samples has been performed using LA-ICP-MS [22]. For this, a commercially available ArF excimer laser ablation system was used in combination with an ICP time-of-flight (TOF) MS. The laser pulse energy was varied over the range of 64 to 132 mJ, while the repetition rate was varied from 1 to 10 Hz. In these studies, the laser beam was imaged by the optical system onto the sample surface so as to produce a crater 120 μm in diameter. The laser beam with a flat-top intensity distribution was used to avoid material mixing from different layers. The flat morphology of the crater bottom obtained with the uniform laser beam is readily apparent in Figure 21.2. The lateral resolution of the system can be changed by realignment of the imaging optics; in this way, craters approximately 10 μm wide with a similar flat morphology can be obtained.
349
350
21 Surface Analysis by Laser Ablation
Figure 21.1 (a) GD-OES and (b) LIBS emission profiles of Zn (䊏) and Fe (ⵧ) in zinc-coated
steel [19].
21.3 Depth Profiling
Figure 21.2 The crater after two shots on TiN (120 μm crater diameter) [21].
Figure 21.3 Drill profiles through the two TiN coatings of 6.4 μm and 2.7 μm thickness (crater 120 μm, 100 mJ pulse energy, 3 Hz repetition rate). The titanium signal is given on the left y-axis; the chromium signal on the right y-axis [21].
Different TiN-, TiC-, and TiAlN-based single layer coatings on steel alloyed with Cr, Ni, Mn, and WC were prepared by use of the cold vapor deposition technique. The thickness of the coatings obtained ranged from 2.7 to 6.4 μm. The transient signals from 53Cr+, 52Cr+, and 182W+ isotopes were used as the indicators of the steel substrate, while 48Ti+ and 48Ti16O+ were used as the indicators of the coating layer. Examples of the temporal profiles of the Ti and Cr isotopes for two samples with coatings of different thickness are shown in Figure 21.3. Here, the TOF temporal profiles of Ti and matrix component Cr were shown to be well separated, demonstrating the depth resolution of the LA system used. For the given laser fluence, the drill time through the coating and the peak area of the TOF signals of the Ti isotopes were linearly proportional to the thickness of the coating. Results from the first experiments on the application of the femtosecond laser for depth profiling have been reported [22, 23]. Two types of multicomponent samples were investigated: (i) multilayer Cu–Ag coatings on a Si substrate, and (ii) TiN–TiAlN on an Fe substrate. The Cu–Ag multilayer samples were prepared as double and triple sandwiches of alternating Cu and Ag layers, with the thickness
351
352
21 Surface Analysis by Laser Ablation
of each copper and silver layer being approximately 600 nm. The TiN/TiAlN samples comprised five TiN–TiAlN double layers on an iron substrate, with the thickness of each TiN and TiAlN layer being equal, at 280 nm. The nitrogen concentration was held constant throughout the layers. Two detection techniques were then tested: LIBS for the Cu–Ag–Si samples; and LA–TOF–MS for the TiN–TiAlN samples. A commercial fs-laser (CPA-10; Clark-MXR, MI, USA) was used for ablation. The parameters used for the laser output pulses were: central wavelength 775 nm; pulse energy ∼0.5 mJ; pulse duration 170–200 fs; and repetition rate from single pulse operation up to 10 Hz. In these experiments a laser with a Gaussian beam profile was used due to the lack of commercial beam homogenizers for femtosecond lasers. The depth profiling of the Cu–Ag–Si samples was performed with the laser beam focused onto the sample surface to a spot of approximately 30–40 μm. In order to obtain the best depth resolution, the laser fluence was maintained near the 1 J cm−2 level, close to the threshold of the LIBS detection scheme. The intensity profiles of Cu and Ag emission lines are shown in Figure 21.4. The individual layers of Cu and Ag were definitely resolved, although the contrast of the normalized intensities in the successive layer was smoothed due to the bell-shaped profile of the laser beam. The estimated ablation rate for Cu–Ag–Si sandwiches varied between 15 and 30 nm per laser shot, depending on the beam fluence. The “soft” ablation of the TiN–TiAlN samples by the low fluence laser beam was performed by using LA-TOF-MS. Due to the greater sensitivity of this technique compared to direct LIBS, a lower laser fluence of approximately 0.3–0.4 J cm−2 was used. One of the depth profiles, obtained by use of femtosecond LA-TOF-MS, is
Figure 21.4 Depth profile of a triple Cu–Ag sandwich (the laser fluence is 1 J cm−2; 10 pulses
accumulation). The dashed lines indicate different layers [22] .
21.3 Depth Profiling
Figure 21.5 Depth profile of TiN–TiAlN–Fe sample. Normalization to the sum of all the major components is used (laser fluence 0.35 J cm−2) [22] .
Figure 21.6 SEM image of TiN–TiAlN–Fe sample after 100 laser shots (laser fluence 0.35 J cm−2) [22] .
shown in Figure 21.5. Each 280 nm-thick layer was ablated by approximately 20–25 pulses, which resulted in an average ablation rate of 11–14 nm per pulse. The ablation rate was low enough for the resolution of all layers. The crater surfaces obtained in the LA-TOF-MS experiment on the TiN–TiAlN– Fe sample were remarkably smooth, and clearly demonstrated the Gaussian intensity distribution of the laser beam. Figure 21.6 shows a SEM image of the crater after 100 laser pulses (fluence 0.35 J cm−2); the crater can be seen as symmetrical and bell-shaped, and there was no significant distortion of the single layers.
353
354
21 Surface Analysis by Laser Ablation
Figure 21.6 is an excellent demonstration of the potential of femtosecond laser ablation, if the laser beam has a flat-top, rather than Gaussian, intensity profile.
21.4 Near-Field Ablation
A very recent development in laser ablation has been the attempt to achieve sub-μm lateral resolution by combining laser ablation/desorption with near-field optical microscopy [24–28]. For this, the laser energy is delivered to the surface either by a tapered optical fiber with a tip diameter of 100 nm or smaller, or by light scattering at a nanometer metal tip. In order to achieve subwavelength resolution, the fiber or tip must be in close proximity (∼10 nm) to the surface. Initially, ICP-MS and direct MS measurements have been demonstrated.
21.5 Conclusion
As noted in Section 21.1, LA was designed primarily as a technique for direct sampling in the bulk analysis of solid samples. The main advantages of the technique are the possibility to ablate all types of solid material (metals, isolators, glasses, crystals, minerals ceramics, etc.), a lack of any need for special requirements regarding the form of the sample, a high lateral resolution down to a few microns, and the possibility of remote sampling using fiber optics. Different hyphenated techniques can be used for detection of the elemental composition of the ablated material. Among these, LIBS is the most flexible, technically the simplest, the cheapest, and enables a moderate sensitivity with DLs in the 10 to 100 ppm range. A much higher sensitivity can be realized by combining LA with the ICP-OES-MS detector; indeed, very low DLs (in the ppb to ppt range) have been achieved in the routine analysis of a wide variety of samples with LA-ICP-MS. Today, LA-ICP-OES/MS instruments for the direct analysis of solid samples are available commercially. With this increased sensitivity, the amount of ablated material required for analysis has been significantly reduced, and this has transformed LA into an almost nondestructive technique. This latter characteristic is especially important for such applications as product certification and the analysis of works of art and gemstones. Initial results have confirmed the high potential of LA-based hyphenated techniques for the depth profiling of coatings and multilayer samples. These techniques can also be used as complementary methods to other surface-analysis techniques. Perhaps the most reasonable application of LA for depth profiling would be the range from a few tens of nanometers to a few tens of microns, as this is difficult to analyze using other techniques such as SIMS, SNMS, TXRF, and GD-OES-MS. Both, the lateral and depth resolution of LA can be improved by the use of femtosecond lasers and near-field ablation.
References
References 1 Moenke-Blankenburg, L. (1993) Spectrochim. Acta Rev., 15, 1–37. 2 Darke, S.A. and Tyson, J.F. (1993) J. Anal. Atom. Spectrom., 8, 145–209. 3 Russo, R.E. (1995) Appl. Spectrosc., 49 (9), A14–A28. 4 Hergenröder, R., Samek, O., and Hommes, V. (2006) Mass Spectrom. Rev., 25, 551–572. 5 Pisonero, J. and Günther, D. (2008) Mass Spectrom. Rev., 27, 609–623. 6 Günter, D., Frischknecht, R., Heinrich, C., and Kahlert, H. (1997) J. Anal. Atom. Spectrom., 12, 939. 7 Amoruso, S., Bruzzese, R., Spinelli, N., and Velotta, R. (1999) J. Phys. B, 32, R131. 8 Cromwell, E. and Arrowsmith, P. (1995) Appl. Spectrosc., 49, 1652. 9 Hergenröder, R. (2006) J. Anal. Atom. Spectrom., 21, 505–516. 10 Chichkov, B.N., Momma, S., von Alvensleben, F., and Tünnermann, A. (1996) Appl. Phys. A, 63, 109–115. 11 Tönshoff, H.K., von Alvensleben, F., Ostendorf, A., Kamlage, G., and Nolte, S. (1999) Int. J. Electr. Machining, 4, 1–6. 12 Bauer, H., Leis, F., and Niemax, K. (1998) Spectrochim. Acta B, 53, 1815. 13 Bengtson, A. and Lundholm, M. (1988) J. Anal. Atom. Spectrom., 3, 879. 14 Jakubowski, N. and Stuewer, D. (1992) J. Anal. At. Spectrom., 7, 951. 15 Oswald, S., Hoffmann, V., and Ehrlich, G. (1994) Spectrochim. Acta B, 49, 1123. 16 Anderson, D., McLeod, C., English, T., and Trevor, A. (1995) Appl. Spectrosc., 49, 691.
17 Kanicky, V., Novotny, I., Musil, J., and Mermet, J. (1997) Appl. Spectrosc., 51, 1037. 18 Kanicky, V., Musil, J., Novotny, I., and Mermet, J. (1997) Appl. Spectrosc., 51, 1042. 19 Vadillo, J., Garcia, C., Palanco, S., and Lazerna, J. (1998) J. Anal. Atom. Spectrom., 13, 793. 20 Garcia, C., Corral, M., Vadillo, J., and Lazerna, J. (2000) Appl. Spectrosc., 54, 1027. 21 Bleiner, D., Plotnikov, A., Vogt, C., Wetzig, K., and Günter, D. (2000) Fresenius J. Anal. Chem., 368, 221. 22 Margetic, V., Bolshov, M., Stockhaus, A., Niemax, K., and Hergenröder, R. (2001) J. Anal. Atom. Spectr., 16, 616. 23 Mateo, M.P., Garcia, C.C., and Hergenröder, R. (2007) Anal. Chem., 79, 4908. 24 Hwang, D., Ryu, S.G., Misra, N., Jeon, H., and Grigoropoulos C.P. (2009) Appl. Phys. A, 96, 289. 25 Becker, J.S., Gorbunoff, A., Zoriy, M., Izmer, A., and Kayser, M. (2006) J. Anal. Atom. Spectrom., 21, 19. 26 Samek, O., Kurowski, A., Kittel, S., Kukhlevsky, S., and Hergenröder, R. (2005) Spectrochim. Acta Part B, 60, 1225. 27 Stöckle, R., Setz, P., Deckert, V., Lippert, T., Wokaun, A., and Zenobi, R. (2001) Anal. Chem., 73, 1399. 28 Schmitz, T.A., Gamez, G., Setz, P.D., and Zenobi, L.Z.R. (2008) Anal. Chem., 80, 6537.
355
357
22 Ion Beam Spectrochemical Analysis (IBSCA) Volker Rupertus
Ion beam spectrochemical analysis (IBSCA) is a sputtering-based surface analytical technique similar to SIMS/SNMS. In the case of IBSCA, the emitted radiation of excited sputtered secondary neutrals or ions is detected. IBSCA, which was developed parallel to SIMS during the 1960s and early 1970s [1–3], is also known as SCANIIR (surface composition by analysis of neutral and ion impact radiation) or BLE (beam-induced light emission). For widespread use as a common analytical technique, the detection of quantifiable signals is strictly necessary. This is a principal problem that IBSCA shares with SIMS, namely that application to the analysis of metals and semiconductors has shown the photon yield to be highly dependent on the surface oxygen content and oxygen partial pressure [4–7]. In contrast, in the case of oxidic samples – for example glass or glass ceramic – the matrix dependence usually proves negligible, so that IBSCA can be used for stoichiometric quantification [8, 9]. IBSCA is, therefore, used mainly as an analytical tool in combination with other methods (e.g., SIMS) for the analysis of highly insulating surfaces and thin films. Most applications deal with the depth profiling of multilayer structures on glass substrates, as well as with the characterization of the near-surface structures of glass samples. The latter often differ from the bulk stoichiometry, due to interaction(s) with the environment.
22.1 Principles
As is typical for a sputter-based erosion technique, an ion beam of mostly positive noble gas ions (e.g., Ar+) penetrates the surface. During the ion beam sputter processes of a nonelemental solid, many physical and chemical processes are initiated; these lead to a change in stoichiometry in the near-surface region by interaction of the primary ions with the atoms of the solid, then induced solid–solid atom interactions and, finally, surface erosion by emission of secondary particles [10– 12]. The ion beam-induced emission processes include the emission of electrons and surface particles (atoms or molecules) in a charged, neutral, and sometimes Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
358
22 Ion Beam Spectrochemical Analysis (IBSCA)
excited state. Some theoretical aspects of the principal mechanism of producing excited sputtered particles have been discussed in the literature; these include the “electron-transfer-model” [13–15], the “level-crossing model” [6], and the “LTEmodel” [16, 17]. The fluence of the sputtered particles is representative for the elemental composition of the near-surface region. Analysis of the emitted particle flux during the sputtering process enables calculation of quantitative concentration depth profiles of the elemental components. For the sputter erosion process, energetic ions (0.5–10 keV) with current densities of 10 to 100 μA cm−2 are commonly used. The lifetime of the excited states of the sputtered ions and neutrals ranges from 10−13 to 10−14 s, so that the de-excitation processes are located up to 2 cm above the sample surface. The emitted photons can be detected in the visible range (250–900 nm) by using transfer optics made from fused silica. For the detection of UV-lines, special optics and detection systems are required (lenses made from CaF2, high-vacuum-pumped optical spectrographs, etc.). The radiation from atoms, ions, or molecules sputtered in an excited state, as well as the emission of the luminescence excited within the solid in the case of glass, can be analyzed by means of optical multichannel analysis (OMA) or CCD-cameras. The analysis of the sputtered species becomes possible by coordinating the peak at a measured wavelength λ with the wavelength-dependent listing of possible energetic transitions for all elements in spectral tables.
22.2 Instrumentation
For IBSCA analysis, a standard high-vacuum (or better, a UHV) equipment with a turbomolecular pump and a residual gas pressure of less than 10−7 mbar is necessary. As can be seen in Figure 22.1, the optical detection system, which consists
Figure 22.1 Scheme of an IBSCA-measurement equipment which is mostly combined with a mass spectrometer (SIMS or SNMS).
22.2 Instrumentation
of a transfer optics, a spectrograph, and a lateral-sensitive detector, is often combined with a quadrupole mass spectrometer to analyze secondary sputtered particles (ions or postionized neutrals). The typical ion sources are using noble gases (usually Ar). The ionization process operates either by electron impact or within a plasma created by a discharge, followed by extraction of the created ions out of the formation space. The ions are then accelerated and focused with two or more electrostatic lenses. These ion guns are normally operated to produce ions of 0.5–10 keV energy at currents between 1 and 10 μA (in the case of a duoplasmatron, up to 20 μA). The chosen spot size varies between 100 μm and 5 mm in diameter. In the case of insulator analysis, an electron gun is also necessary to compensate the positive ion current at the sample surface. Two operation types are typical:
•
For lower electron current densities a flood gun can be employed which produces directly low-energy electrons of 1–20 eV.
•
For higher electron current densities of up to 300 μA cm−2, standard electron guns are used which emit electrons of 0.5–2 keV. The outcoming electrons are focused to a metal or carbon mask which covers the sample rim (see Figure 22.2). By correctly selecting the primary electron energy, a secondary electron yield δ > 1.5 at perpendicular incidence can be achieved (e.g., δmax(Ta) = 1.3 at Ep = 0.6 keV). An additional increase of the SE-yield by a factor of two can be reached using bombardment angles of 50–60° normal to the sample surface [18]. With this procedure, low-energetic secondary electrons are created with an energetic distribution around the maximum of about 2 eV in the case of Ta [19]. These low-energetic electrons follow directly the attractive field forces at the sample surface and compensate the slight positive charging caused by the ion impact.
The penetration depth of the incoming SE-electrons is comparable to the penetration depth of the primary ions, so that no local charging effects like a capacitor
Figure 22.2 Scheme of the compensation mechanism to avoid charging of the sample surface when primary energetic ions and electrons are used.
359
360
22 Ion Beam Spectrochemical Analysis (IBSCA)
will build up, and therefore charging-induced migration/diffusion in the surfaceadjacent region will be suppressed. The photon detection system consists of a transfer optics (a few lenses) made from fused silica to achieve an improved transmission of down to 250 nm. For the near-UV region, crystal materials such as CaF2 must be used, as well as a vacuumpumped spectrograph. The transfer optics focus the space volume in front of the penetrated sample surface to the entrance slits of a grating spectrograph, typically in Czerny–Turner arrangements (two mirrors and one grating). Depending on the grating used (120–2400 lines mm−1), the resolution can be varied between 0.5 nm and 0.05 nm. The transmitted photons are registered with an optical multichannel analyzer or with a CCD-camera, so that the intensity can be detected with a high lateral resolution. The detection system is computer-controlled, so that the sputter time-dependent change of the spectra is stored and sputter depth profiles can subsequently be computed.
22.3 Spectral and Analytical Information
An IBSCA-spectrum (Figure 22.3) consists of many peaks in the visible range (250–900 nm), with each peak being related to an electron de-excitation process of a sputtered particle from a higher to a lower state; in the case of the more dominant peaks, this will be to the ground state. In principle, two major types of peak family exist:
• •
Type I: where the photons are emitted from excited sputtered secondary neutrals. Type II: where the photons are emitted from excited sputtered secondary ions (singly charged).
Figure 22.3 IBSCA-spectrum of SK16 optical glass bombarded by 5 keV Ar+.
22.4 Quantitative Analysis by IBSCA
The peaks can be identified by using spectra tables [20]. The variety of allowed electronic transitions causes many peaks in the visible range which, on occasion, are energetically so close together that deconvolution of neighboring peaks is hampered. Therefore, the correct choice of lines to be used is particularly important for a quantitative interpretation of the analytical results of multicomponent samples. A second source of failure is the determination of the background of the spectra, which depends on complete photon emission from the sample (solidstate luminescence). During the measurement – and especially in the case of depth profiling – the background level can change, and therefore a correct background deduction is important for quantitative analysis. The peak resolution depends on the grid size used in the spectrograph, which also determines the detection interval (e.g., 250–600 nm). A better peak resolution is directly correlated with a smaller detection interval, and therefore the grid size must be optimized for each range of application. In Figure 22.3 the main components of SK16 (Ba, B, Si, Al, Na) are visible. Clearly, more than one line for each element, excited neutral (label: I) and ion (label: II), is detectable in the depicted detection range. Only oxygen, with the highest concentration (≈60 atom%) of all SK16 components, is not detectable, because the main radiation lines are located in the vacuum-UV range. The relationship between the recorded intensity and the actual surface concentration/chemical state of the surface is not as straightforward as in XPS or AES. As in the case of peak intensity in SIMS or SNMS, the height of the detected peaks depends more or less on the chemical matrix of the actual sample surface.
22.4 Quantitative Analysis by IBSCA
The intensity Iki (λA) of a spectral emission line, which is characteristic of a sputtered element or molecule A, is calculated with Iki (l A ) = IP ⋅ YA ⋅ hki ⋅ a ki /e0 Iki (λA) is the intensity of the radiative recombination of an electron of a species A from a higher energy level k to the lower level i; IP is the primary ion current; λA is the wavelength of the transition from an excited state k to the ground state i; YA is the partial sputtering yield of the sputtered species A; ηki
is the detection factor of the optical system which quantitatively takes into account the aperture and the quantum efficiency of the optical detection system (electrons/quantum) for λA;
361
362
22 Ion Beam Spectrochemical Analysis (IBSCA)
αki is the excitation coefficient (or photon yield), which is defined as the ratio of the number of atoms of the element A showing this transition to the total number of the sputtered atoms; and e0
is the electron charge.
The radiation factor αki is determined by
α ki = where
Nk ⋅ ε ki ⋅ P Nk + Ni
Nk is the ratio of the excited sputtered secondary particles of species A; Nk + Ni
εki is the transition probability from level k to level i; and P is the probability that the deexcitation is a radiative process (P ∼ 1 for insulating samples [6]). Similar to other sputter-based techniques, a sensitivity factor can be determined: DAλ = ηki ⋅ α ki so that the intensity is given by: Iki (λ A ) = IP ⋅ YA ⋅ DAλ Taking atomic sputtering into account, the ratio of particles emitted as molecules is negligible and the partial sputtering yield in sputter equilibrium can be determined by YA = c Ab ⋅ Ytot where YA is the partial sputtering yield for element A; c Ab is the bulk concentration of element A; and Ytot is the total sputtering yield of the sample, where Ytot = Taking into account that mined by c Ab =
∑c
b A
∑Y . A
A
= 1, the bulk concentration of element A is deter-
A
Iki (λ A )/ DAλ
∑(I (λ )/D ) ki
j
λ j
j
By using relative sensitivity factors, reference measurements on standard samples are not necessary. The ratio of two different elemental concentrations in one sample is determined by: c Ab I A DB = ⋅ c Bb IB DA The ratio DB/DA is a so-called relative sensitivity factor, D. This ratio is mostly related to one element, for example, the element for insulating samples, silicon,
22.5 Applications
Figure 22.4 Comparison of relative intensities measured with IBSCA, resp. SNMS. (parameters: sputter equilibrium after 5 keV Ar+-bombardment; samples: oxidic glasses with varying content of Na (0.1–7.4 atom%) and Pb (4.4–22.1 atom%).
which is one of the main components of glasses. Using the equation that the sum of the concentration of all elements is equal to 1, the bulk concentrations can be directly determined by the measured intensities and the known D-factors, if all components of the sample are known. The linearity of the detected intensity and the flux of the sputtered neutrals has been demonstrated for silicate glasses in the case of IBSCA and SNMS [9]. For SNMS, the lower matrix dependence is shown for a variety of samples [21]. A comparison of normalized SNMS and IBSCA signals for Na and Pb as prominent components of optical glasses shows that a fairly good linear dependence exists (Figure 22.4). From these results it can be postulated that a fixed portion of sputtered secondary neutrals is emitted in an excited state in the case of oxidic glasses. Such linearities can only be determined for similar matrices, which limits the use of D-factors to sample systems that are similar to the reference sample system employed for the D-factor determination.
22.5 Applications
In most cases the depth profile mode of IBSCA is used in thin-film analysis routine. A typical application is depicted in Figure 22.5, where a fixed spectral range is monitored during the sputter erosion of the surface so that many IBSCA spectra are stored during the sample analysis. This example shows how the single IBSCA spectrum changes during the sputter erosion of an antireflecting coating on soda-lime glass (SiO2–TiO2–SiO2/TiO2– substrate). The intensity variation of single peaks, for example, Si or Ti, during the sputter erosion of the multilayer becomes visible; sputter depth profiles can be computed from this type of measurement (see Figure 22.6).
363
364
22 Ion Beam Spectrochemical Analysis (IBSCA)
Figure 22.5 IBSCA spectra of an antireflecting coating on soda-lime glass (SiO2–TiO2–SiO2/
TiO2–substrate) obtained at 5 keV Ar+-bombardment.
The lower matrix sensitivity in comparison to SIMS is demonstrated in Figure 22.6. With both methods, a TiO2–SiO2–TiO2 multilayer coating on soda-lime glass was analyzed, with a sol–gel technique, followed by a tempering process, being employed for the coating process. During this tempering treatment, a diffusion process of alkaline species of the substrate material into the multilayer is initiated. A comparison of the positive SIMS profile with the IBSCA-depth profile shows similar characteristics of the main components Ti and Si, but drastic differences in the shape of the Na-signal. In the case of SIMS, intensity maxima at the SiO2/ TiO2 interface, as well as at the surface area, are detected which are 20%, respectively, 60% higher than the bulk intensity of the substrate which is the Na-source. A quite different behavior is visible in the IBSCA profile, however (Figure 22.6b), where the Na-intensity shows also maxima at the same positions, but only with local character. This Na profile demonstrates the diffusion process from the glass substrate to the surface, and indicates that SiO2 behaves as a diffusion barrier so that the interface from TiO2 to SiO2 shows an enrichment of Na, because the diffusion coefficient in SiO2 is lower than in TiO2. The IBSCA–Na-profile provides realistic information concerning the relative concentration in comparison to the bulk of the glass substrate, which was supported by additional XPS-measurements. The enlargement of the Na signal in the case of SIMS provides clear evidence of the influence of the chemical matrix. At the surface, as well as at the single layer interfaces, a chemical environment is identified that increases the ionization probability. Consequently, IBSCA can be regarded as a sensitive method that can be combined with SIMS to obtain a better interpretation of thin oxide film multilayer structures. A second set of examples deals with the analysis of near-surface regions of glasses that normally demonstrate so-called “altered” or “leached” layers. The altered layer is found for soda-lime glasses as well as for many glasses used for optical applications. The change in surface composition is caused by weathering
22.5 Applications
Figure 22.6 (a) SIMS and (b) IBSCA depth profiles of an interference layer system used for automotive application (TiO2–SiO2–TiO2–soda-lime glass) obtained at 5 keV Ar+ bombardment.
effects, and indicates an ion exchange of alkaline species from the glass site and OH from the environment. This hydration effect leads to a change in the glassy network, affecting the physical parameters of the altered layer in comparison to the bulk glass. The strong influence on the chemical matrix is demonstrated in Figure 22.7, where the SIMS signal Si+ shows a different behavior compared to the IBSCA Si I-signal. This fact indicates the presence of H and/or OH in the topmost region that increases the ionization probability for Si+, and leads to a depletion in Li as a result of ion exchange (Li against H). The increase in H-content in the surface-adjacent region was also detected using nuclear
365
366
22 Ion Beam Spectrochemical Analysis (IBSCA)
Figure 22.7 SIMS and IBSCA depth profiles of the altered layer region of a lithium alumosili-
cate (LAS) glass ceramic (data taken at 5 keV Ar+ bombardment).
reaction analysis (NRA) [22], which supports the fact that SIMS and IBSCA react in quite different ways to slight chemical matrix changes. This effect can be used as a fingerprint for an increase in H, respectively OH, in the near-surface region of oxidic glasses when both methods are employed.
References 1 Goutte, R., Guillaud, C., Javelas, R., and Meriaux, J.P. (1967) Optik, 26, 574. 2 White, C.W., Simms, D.L., and Tolk, N.H. (1972) Science, 177, 481. 3 Tolk, N.H., Tsong, I.S., and White, C.W. (1977) Anal. Chem., 49, 1617. 4 Blaise, G. (1976) Surf. Sci., 60, 65. 5 Thomas, G.E. (1979) Surf. Sci., 90, 381. 6 Tsong, I.S.T. (1981) Inelastic ParticleSurface Collisions (eds E. Taglauer and W. Heiland), Springer, Heidelberg. 7 Weg, W.F.v.d. and Rol, P.K. (1965) Nucl. Instrum. Methods, 38, 274. 8 Bach, H. (1989) Glass Technol., 30, 75. 9 Rupertus, V., Rothhaar, U., Köpfer, P., Lorenz, A., and Oechsner, H. (1993) Vakuum in der Praxis, 3, 183. 10 Behrisch, R. (1983) Sputtering by Particle Bombardment II, Springer, Heidelberg. 11 Behrisch, R. and Eckstein, W. (2007) Sputtering by Particle Bombardment, Springer Verlag, Heidelberg. 12 Benninghoven, A., Rüdenauer, F.G., and Werner, H.W. (1987) Secondary Ion Mass
13 14 15 16 17 18 19 20
21
22
Spectrometry, John Wiley & Sons, Inc., New York. Weg, W.F.v.d. and Rol, P.K. (1965) Nucl. Instrum. Methods, 38, 274. Ghose, D., Brinkmann, U., and Hippler, R. (1997) Phys. Rev. B, 55, 13989. Thomas, G.E. (1979) Surf. Sci., 90, 381. Martin, P.J. and MacDonald, R.J. (1977) Surf. Sci., 62, 551. Andersen, C.A. and Hinthorne, J.R. (1973) Anal. Chem., 45, 1421. Bauer, H.E. and Seiler, H. (1984) Scan. Electron Microsc., III, 1081. Kollath, R. (1956) Handuch der Physik XXI, Springer Verlag, Berlin. Zaidel, A.N., Prokofev, V.K., Raiskii, S.M., Slavnyi, V.A., and Shreider, E.Y. (1970) Tables of Spectral Lines, IFI, Plenum, New Yoek. Oechsner, H. (1984) Thin Film and Depth Profile Analysis, Springer Verlag, Heidelberg. Dersch, O., Laube, M., Rauch, F., and Becker, O. (1999) Glastechn. Ber. Glass Sci. Technol., 72 (10), 329.
367
23 Reflection Absorption IR Spectroscopy (RAIRS) Karsten Hinrichs
Reflection absorption infrared spectroscopy (RAIRS) is a well-established vibrational spectroscopic method that is frequently applied within the research areas of chemistry, physics, and biology [1–34]. The technique is used to identify and characterize the chemical state, molecular structure, anisotropy, and conductivity of thin films adsorbed onto metallic and nonmetallic surfaces. Vibrational spectra are used as characteristic fingerprints for adsorbate molecules, adsorption configurations, and structures. The method is also referred to as different acronyms, including RAIRS, FT-IRAS, IRRAS, and ERIRS (external reflection infrared spectroscopy), all of which refer to the same principle of measurement using IR radiation and a reflection absorption geometry. Typically, this technique involves a single external reflection at a sample covered substrate. The sample absorbs radiation at material specific frequencies of free carriers, electronic transitions, or vibrational bands, and a corresponding absorption occurs in the spectrum of the reflected radiation. For this technique, substrates are chosen that do not produce disturbing absorption bands in the spectral range of interest.
23.1 Instrumentation
In the IR spectral range, Fourier-transform (FT) interferometers are often used. When compared to the use of dispersive spectrometers, FTIR enables a higher optical throughput and multiplex advantage at equivalent high spectral resolution. However, in the face of technological developments in detector sensitivity and focal plane arrays, the use of dispersive spectrometry has during recent years been the subject of much discussion. In the near-IR region, a tungsten-filament lamp is conventionally used as a source of radiation, whereas a “glowbar” is used in the mid-IR region and a mercury-arc lamp is used in the far-IR region. A synchrotron – a special radiation source which offers a high brilliance throughout the whole IR region – has been applied for in situ measurements during the anodic oxidation of copper in aqueous solution environments [6].
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
368
23 Reflection Absorption IR Spectroscopy (RAIRS)
RAIRS is a nondestructive IR technique with special versatility; it does not require the vacuum conditions as needed for electron spectroscopic methods, and is therefore – at least in principle – applicable to growth processes [7]. Moreover, by using a polarization modulation technique, the surfaces in a gas phase can be investigated. The method can also be used to discriminate between anisotropic near-surface absorption and isotropic absorption in a gas phase [8].
23.2 Principles
Conventionally, RAIRS has been used for both the qualitative and quantitative characterization of adsorbed molecules or films on mirror-like (metallic) substrates [1]. During the past decade, the applicability of RAIRS to the quantitative analysis of adsorbates on nonmetallic surfaces (e.g., semiconductors, glass [3] and water [10]) has also been proven. The classical three-phase model for a thin isotropic adsorbate layer on a metallic surface was developed by Greenler [1, 9], and calculations for the model have since been extended to include the description of anisotropic layers on dielectric substrates [11–13]. A review of the interpretation of the IR spectra of thin films is also available [5]. The dielectric properties of the substrate and adsorbate determine the optimum conditions under which the RAIRS experiment should be performed. The excitation of vibrational adsorbate modes on metals is governed by a so-called “surface selection rule” [1]. In this case, the amplitude and phase of the reflected radiation depend on the direction of the electric field vector, which is composed of one component parallel to the reflection plane (p-polarized), and a second component normal to the reflection plane (s-polarized). On considering Fresnel’s equations, only the p-polarized radiation shows a significantly resultant electromagnetic field close to the metallic surface for high incident angles of radiation near a grazing incidence (at ca. 80°). In contrast, the mean square electric field of radiation polarized in the surface plane is negligible due to the 180° phase shift after reflection at the metallic surface. Hence, only those vibrations with dipole components perpendicular to the surface can be efficiently excited. With p-polarized radiation and incident angles near grazing incidence, an approximate 25-fold increase in sensitivity can be achieved in comparison to transmission experiments [1]. However, this increase is reduced to about 17-fold in a more realistic experimental situation, when the spread of incident angles of approximately ±5° occurs at about 85°. RAIRS spectra contain absorption band structures, related to electronic transitions and vibrations of the bulk, the surface, a film, or adsorbed molecules. In reflectance spectroscopy, the absorbance is usually determined by calculating the – log (RS/R0), where RS represents the reflectance from the adsorbate-covered substrate, and R0 is the reflectance from the bare substrate. For thin films with strong dipole oscillators, the Berreman effect – which can lead to an additional spectral feature in the reflectance – must be considered (see Chapter 25). The
23.3 Applications
frequencies, intensities, full widths at half-maximum and band-shapes in the absorption spectrum yield information related to the adsorption states, chemical environments, ordering effects, and vibrational coupling. For films on nonmetallic substrates (semiconductors, dielectrics), the situation is much more complex when, in contrast to metallic surfaces, both the parallel and perpendicular vibrational components of the adsorbate can be detected. The sign and intensity of the RAIRS-bands depend heavily on the angle of incidence, the polarization of radiation, the film thickness, and the orientation of vibrational transition moments [3].
23.3 Applications 23.3.1 RAIRS
RAIRS is routinely used in the area of chemically modified surfaces, with a variety of surface systems in electrochemistry [14], polymer research [2, 15], catalysis [1, 8], self-assembled monolayers [3, 4, 30, 33], and protein adsorption [4, 16, 31, 32] having been investigated. Beside studies of adsorbates on metal and semiconductor surfaces, considerable interest has been shown in the structural characterization of monolayers at the air–water interface [10, 17]. Likewise, experiments at the gas–water interface have been carried out, with examples in biochemical research to characterize membrane proteins [16] or to determine the structures of other biological molecules [18]. Both, experimental and simulated RAIRS spectra of a single monolayer poly-γ-benzyl-L-glutamate (PGB) and a synthetic peptide K(LK)7 at the air–water interface are depicted in Figure 23.1 [18]. In the simulations, the different secondary structures of both molecules and anisotropic optical constants were taken into account, with the PGB spectra containing significant features resulting from amide I, amide II, and ester bands in the α-helix structure. For K(LK)7, the simulations revealed molecules lying in antiparallel β-sheets on the water surface, and the frequency of the amide I mode (1622 cm−1) provided information concerning the interactions between the antiparallel β-sheets and the substrate. The knowledge of such optical data is of great interest in determining the main secondary structures (α-helix, β-sheets) present in proteins and polypeptides [18]. Another application for RAIRS is to determine the average tilt angles of molecules on a surface. In comparison to calculations of relative intensities of absorption bands [3], an average tilt angle of the hydrocarbon chains of approximately 10° towards the surface normal was deduced. The relevant IR reflection spectra of an octadecylsiloxane (ODS) monolayer on a silicon substrate [3] for both s- and p-polarized radiation at different incident angles are shown in Figure 23.2. Here, the absorption bands are assigned to fundamental symmetric and asymmetric CH2
369
370
23 Reflection Absorption IR Spectroscopy (RAIRS)
Figure 23.1 Experimental and simulated polarization-modulated-IRRAS spectra of a single
monolayer (a) PGB and (b) K(LK)7 at the air–water interface. The surface pressure was 20 mN m−2. Reproduced with permission from Ref. [18]; © 2000, American Chemical Society.
stretching vibrations, and the spectra change significantly as a function of the incident angle of radiation and the direction of polarization. For p-polarization, the sign of RAIRS bands is changed by crossing the Brewster angle at 73°. In addition to the structure and orientation of molecules, reaction kinetics – for example, of alkoxides on Cu(111) surfaces [19] – may also be investigated. The line shapes and intensities of vibrational bands in reflection spectra from thin films are usually different from those in transmission spectra. Both, calculated and measured spectra for isotropic poly(methyl methacrylate) (PMMA) films of different thicknesses on glassy carbon are shown in Figure 23.3. Here, the spectra obtained with incident light with s- and p-polarization are completely different for the three films. Notably, in each spectrum the band shapes were dis-
23.3 Applications
Figure 23.2 IR reflection spectra from of an
ODS monolayer on silicon for s- and p-polarized radiation at different incident angles θ. Symmetric (s), asymmetric (as),
in-plane (ip) and out-of plane (op) CH2/CH3 vibrations are marked. Reproduced with permission from Ref. [3]; © 1997, Society for Applied Spectroscopy.
torted compared to conventional transmission experiments, which measure the material dispersed in a KBr matrix [2]. The observed absorbance maxima were approximately 10 cm−1 higher in energy than the maximum of 1731 cm−1 in conventional transmission experiments [2]. It was concluded that all band shape distortions had resulted from “optical effects,” which meant all effects had caused changes in the recorded spectrum, except those that had resulted from surface perturbations and changes in structure or chemical bonding. Optical effects are taken into account by optical theory [5, 11, 13, 20]. The use of rapid and sensitive measurement procedures including photoelastic modulators can be employed for the analysis of processes, for example, in electrochemistry. Polarization-modulated PM-FT-IRRAS experiments of unilamellar vesicles of 1,2 dioyl-sn-glycero-3-phosphatidyl choline (DOPC) on Au(111) were used to show that the field-driven transformation of the film involves changes in hydration, orientation, and conformation in the polar head groups. The study results showed that changes in the packing and tilt of the aryl chains are consequences of the head-group rearrangements. Figure 23.4 shows the dependence of the angle between the transition dipole moment and the electric field vector for the DOPC bilayer on the Au(111) electrode for the C–O[P] stretching bands at (a) 1050 cm−1 and (b) 1070 cm−1 [29].
371
372
23 Reflection Absorption IR Spectroscopy (RAIRS)
Figure 23.3 Experimental and calculated
(dashed line) RAIR-spectra for poly(ethyl methacrylate) films for film thicknesses of (3270 ± 100) nm, (362 ± 30) nm, and (78 ±
15) nm. (a) p-polarized light is incident at 60 °; (b) s-polarized light is incident at 60 °. Reproduced with permission from Ref. [2]; © 1988, American Chemical Society.
23.3.2 ATR and SEIRA
To complete this treatment, it is necessary to discuss some possibilities which allow the detection sensitivity to be increased. Typically, for this purpose either certain measurement geometries in, for example, attenuated total reflection (ATR), or substrates and modifications which enable the surface-enhanced infrared absorption (SEIRA) effect, are used. Notably, either internal or external ATR configurations can be used. For example, monolayers of small molecules on semiconductors can be detected, when the detection sensitivity is increased by employing multiple internal reflections [21, 22]. However, the edges of samples must first be beveled in order to couple the radiation in and out. This method was first applied to investigations of the morphology of H-terminated Si surfaces upon etching [21],
23.3 Applications
Dependence of the angle between the transition dipole moment and the electric field vector for a DOPC bilayer on a Au(111) electrode for the C–O[P] stretching
Figure 23.4
bands at (a) 1050 cm−1 and (b) 1070 cm−1. Reproduced with permission from Ref. [29]; © 2003, Biophysical Society.
and also of the surface chemistry of metalorganic chemical vapor deposition (MOCVD), which is the technique of choice when fabricating Group III/V compound semiconductor devices [23]. When using the SEIRA effect, the films to be analyzed are typically deposited on metallic island substrates, or vice versa. The use of the metallic islands leads to an enhancement effect for the observed vibrational band amplitudes and, dependent on the effective dielectric function of the island film, to specific vibrational band shapes. This enhanced IR absorption is well known as SEIRA effect, and different models have been discussed for its explanation [24–28]. SEIRA spectroscopy may also serve as in situ method for studies of structural changes in thin organic films [34].
373
374
23 Reflection Absorption IR Spectroscopy (RAIRS)
To conclude, RAIRS – which offers a high spectral resolution – is a very versatile nondestructive optical technique that is independent of a vacuum environment. Vibrational spectra may also serve as characteristic fingerprints for adsorbate molecules, adsorption configurations, and structures on metallic and dielectric substrates, the extension of which to include dielectric substrates continues to open new areas of application in polymer and biochemical research.
23.4 Related Techniques
RAIRS is often used as alternative method to HREELS (see Chapter 4), which serves as a standard technique for investigating dipole active vibrations but has a lower spectral resolution (about 20 cm−1 on semiconductor surfaces). As a complementary vibration spectroscopic technique, Raman spectroscopy (see Chapter 24) may be used; however, when the film thickness and degree of polarization are to be determined from IR spectra, IR ellipsometry may offer an alternative approach (see Section 25.3.2). In IR microscopy, grazing incidence reflection (GIR) objectives can be attached in order to increase the lateral resolution for the measurement of thin films on metallic substrates. As a nonlinear optical vibrational spectroscopic technique, sum frequency generation (SFG) (see Chapter 26) has been applied successfully to detect structural changes in thin films.
References 1 Hayden, B.E. (1987) Vibrational Spectroscopy of Molecules on Surfaces, Plenum Press, New York, London, pp. 267–344. 2 Porter, M.D. (1988) Anal. Chem., 60, A1143–A1155. 3 Brunner, H., Mayer, U., and Hoffmann, H. (1997) Appl. Spectrosc., 51, 209–217. 4 Tengvall, P., Lundström, I., and Lindberg, B. (1998) Biomaterials, 19, 407–442. 5 Tolstoy, V.P., Chernyshova, I.V., and Skryshevsky, V.A. (eds) (2003) Handbook of Infrared Spectroscopy of Ultrathin Films, John Wiley & Sons, Inc., Hoboken, New Jersey. 6 Melendres, C.A., Bowmaker, G.A., Leger, J.M., and Beden, B. (1998) J. Electroanl. Chem., 499, 215–218. 7 Bauer, G. and Richter, W. (1996) Optical Characterisation of Epitaxial Semiconductor Layers’ “IRRAS”, Springer, New York, Berlin, Heidelberg, pp. 60–62.
8 Beitel, G.A., Laskov, A., Oosterbeek, H., and Kuipers, E.W. (1996) J. Phys. Chem., 100, 12494–12502. 9 Greenler, R.G. (1966) J. Chem. Phys., 44, 310–314. 10 Flach, C.R., Gericke, A., and Mendelsohn, R.J. (1997) J. Phys. Chem. B, 101, 58–65. 11 Parikh, A.N. and Allara, D.L. (1992) J. Chem. Phys., 96, 927–945. 12 Buffeteau, T., Blaudez, D., Pere, E., and Desbat, B. (1999) J. Phys. Chem. B, 103, 5020–5027. 13 Yamamoto, K. and Ishida, H. (1994) Appl. Spectrosc., 48, 775–787. 14 Christensen, P., and Hamnett, A. (2000) Electrochim. Acta, 45, 2443–2459. 15 Sammon, C., Yarwood, J., and Everall, N. (2000) Polym. Degrad. Stab., 67, 149–158. 16 Lavoie, H., Gallant, J., Grandbois, M., Blaudez, D., Desbat, B., Boucher, F., and Salesse, C. (1999) Mater. Sci. Eng. C, 10, 147–154.
References 17 Urai, Y., Ohe, C., Itoh, K., Yoshida, M., Iimura, K.-, and Kato, T. (2000) Langmuir, 16, 3920–3926. 18 Buffeteau, T., Le Calvez, E., Castano, S., Desbat, B., Blaudez, D., and Dufourcq, J. (2000) J. Phys. Chem. B, 104, 4537–4544. 19 Street, S.C. and Gellman, A.J. (1996) J. Phys. Chem., 100, 8338–8348. 20 Mielczarski, J.A. (1993) J. Phys. Chem., 97, 2649–2663. 21 Dumas, P., Chabal, Y.J., and Jakob, P. (1992) Surf. Sci., 269/270, 867–878. 22 Hicks, R.F., Qi, H., Fu, Q., Han, B.-K., and Li, L. (1999) J. Chem. Phys., 110, 10498–10508. 23 Fu, Q., Li, L., Begarney, M.J., Han, B.-K., Law, D.C., and Hicks, R.F. (1999) J. Phys. IV France, 9, Pr8-3. 24 Röseler, A., and Korte, E.H. (1998) Thin Solid Films, 313–314, 732. 25 Nishikawa, Y., Fujiwara, K., Ataka, K.-I., and Osawa, M. (1993) Anal. Chem., 65, 556. 26 Osawa, M. (2001) Surface-enhanced infrared absorption, in Near-Field Optics and Surface Plasmon Polaritons, Topics in Applied Physics, Springer, Berlin/ Heidelberg, vol. 81, pp. 163–187.
27 Priebe, A., Sinther, M., Fahsold, G., and Pucci, A. (2003) J. Chem. Phys., 119, 4887. 28 Aroca, R.F., Ross, D., and Domingo, C. (2004) Appl. Spectrosc., 58, A325. 29 Zawisza, I., Lachenwitzer, A., Zamlynny, V., Horswell, S.L., Goddard, J.D., and Lipkowski, J. (2003) Biophys. J., 85, 4055–4075. 30 Love, J.C., Wolfe, D.B., Haasch, R., Chabinyc, M.L., Paul, K.E., Whitesides, G.M., and Nuzzo, R.G. (2003) J. Am. Chem. Soc., 125, 2597–2609. 31 Lopes, D.H.J., Meister, A., Gohlke, A., Hauser, A., Blume, A., and Winter, R. (2007) Biophys. J., 93, 3132–3141. 32 Meister, A., Nicolini, C., Waldmann, H., Kuhlmann, J., Kerth, A., Winter, R., and Blume, A. (2006) Biophys. J., 91, 1388–1401. 33 Tinazli, A., Tang, J.L., Valiokas, R., Picuric, S., Lata, S., Piehler, J., Liedberg, B., and Tampe, R. (2005) Chem. Eur. J., 11, 5249–5259. 34 Jiang, X., Ataka, K., and Heberle, J. (2008) J. Phys. Chem C, 112, 813–819.
375
377
24 Surface Raman Spectroscopy Wieland Hill and Bernhard Lendl
24.1 Principles
Raman scattering [1] is one of the phenomena that occur when electromagnetic radiation interacts with a molecule or a crystal. As a result of the inelastic scattering process, photons lose energy, which excites molecular or lattice vibrations (Stokes scattering), or can even gain energy (anti-Stokes scattering) when the transition starts from an excited vibrational state and the final state is of lower energy (e.g., the ground state). When a sample is irradiated with monochromatic light, the scattered light contains a spectrum of wavelengths in which the intensity peaks are shifted from the excitation wavelength by energy-equivalent amounts corresponding to that for the excitation of the molecular vibration modes or crystal phonons. These vibrational (Raman) spectra are highly specific for the material studied, and can be used for the unambiguous identification of substances, the characterization of interactions with the surface, and also for quantitative purposes. Since the Raman cross-section of molecules is usually low, intense light sources and low-noise detectors must be used, and high sensitivities – as required for surface analysis – are difficult to achieve. Different approaches, both singly and in combination, enable the detection of Raman spectroscopy bands from surfaces:
•
Ultrasensitive equipment: In recent years, all components of Raman equipment (laser, sampling optics, filtering, monochromator, and detector) have been clearly improved. This has led to an enormous increase in sensitivity, as well as shortened analysis times, and has enabled the direct observation of adsorbed molecules with carefully optimized instruments without the need for further enhancement or resonance effects.
•
Large specific surface area: Porous materials can have a large proportion of surface atoms – their surface area within a typical sampling volume of 103 mm3 can reach 105 mm2, which is approximately 103-fold larger than for a smooth surface crossing the same volume. These effects lead to clearly increased
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
378
24 Surface Raman Spectroscopy
Raman intensities of surface species, and also to improved intensity ratios of surface and bulk Raman bands.
•
Resonant excitation: Excitation by a laser, which is resonant with an electronic transition of the material under investigation, can increase the Raman crosssection by a factor of approximately 102–106. The transitions – and thus the resonance wavelengths – are specific for the substances. Resonance excitation thus leads to selectivity that can be useful for suppressing bulk bands, but can also complicate the detection of mixtures of substances with different absorption spectra.
•
Surface-enhancement: Electromagnetic and chemical effects can enhance the Raman intensities from substances in close proximity to appropriate metal surfaces by several orders of magnitude. It is generally accepted that electromagnetic effects dominate. Closely packed nanoparticles are needed to create hot spots with maximum SERS enhancement [2].
•
Other considerations: Another issue in Raman spectroscopy is the superposition of the weak Raman spectrum by the more intense fluorescence of the substances or their impurities. This problem can be overcome by the use of longer excitation wavelengths, which cannot excite electronic transitions. Although this concept works well for many samples, near-infrared (NIR) excitation wavelengths result in lower sensitivity in comparison with visible excitation, due to the noise of detectors working in this range and due to the I−4 dependence of the Raman scattering intensity. Alternatively, in order to overcome the problem with fluorescence, UV excitation can be used. In this case, the Raman spectra are compressed to a wavelength range of a few nanometers only, and thus the spectrum occurs prior to the onset of fluorescence at longer wavelengths.
Raman spectroscopy does not require a vacuum and is, in principle, applicable to all optically accessible samples – even inclusions in glass or minerals, substances in aqueous solutions, or substances packaged in transparent glass or plastics can be analyzed. Small particles down to diameters below 1 μm can also be characterized by correct focusing of the exciting laser beam. Single-cell analysis and the identification of bacteria based on Raman spectra have also been demonstrated. As transmission is not required for backscattering measurements, strongly scattering materials can also be investigated.
24.2 Surface-Enhanced Raman Scattering (SERS)
The application of Raman spectroscopy in surface analysis is limited by the low scattering cross-section of the molecules, and the correspondingly low intensities of their Raman bands. This situation can be substantially improved by use of socalled surface-enhanced Raman scattering (SERS) [3, 4]. The Raman scattering of
24.2 Surface-Enhanced Raman Scattering (SERS)
Figure 24.1 Schematic diagram of a SERS-active substrate and the measurement arrangement. Alumina nanoparticles are deposited on a glass surface and produce the required roughness. A thin silver layer is
evaporated onto the nanoparticles and serves for the enhancement. Organic molecules adsorbed on the silver surface can be detected by irradiation with a laser and collecting the Raman-scattered light.
molecules adsorbed onto rough metal surfaces or small metal particles (Figure 24.1) can be increased by six, or more, orders of magnitude by electromagnetic and chemical effects, whereby electromagnetic effects dominate [2]. This large enhancement enables the easy analysis of molecular monolayers at appropriate surfaces; even 1% of a monolayer has been observed and the detection of single molecules on optimum metal particles was reported [5]. Single-molecule detection was made possible by a combination of resonance Raman (RR) and SERS, referred to as surface-enhanced resonance Raman scattering (SERRS). The first SERS experiments were performed with electrochemically roughened electrodes and metal colloids, and many other types of suitable SERS substrates are known. Examples include metal island films, metal films over nanoparticles or rough substrates, gratings, and sputter-deposited metal particles. SERS is usually restricted to specially prepared “active” surfaces. A broad range of other surfaces also becomes accessible to SERS spectroscopy following special preparation techniques:
•
Metal island films form spontaneously during slow evaporation of the metal onto supports with low adhesion. The metal islands can produce the surfaceenhancement required for Raman characterization of the surface of the support.
•
Appropriate metal particles can be sputter-deposited on almost any support that can be exposed to vacuum.
•
Ultrathin metallic, semiconductor, insulator, or organic overlayers can be deposited on SERS-active metal surfaces.
For overlayer thicknesses of a few atomic or molecular layers, the supporting metal can produce surface-enhanced fields at the surface of the overlayer. The composition and structure of the overlayer surface can then be analyzed using SERS spectroscopy [6].
379
380
24 Surface Raman Spectroscopy
24.3 Instrumentation
The Raman scattering effect was first observed in 1928 by C.V. Raman, who used sunlight and complementary filters [1]. Later, mercury vapor lamps were used for the illumination of samples. Current Raman spectrometers use lasers as monochromatic light sources, with different types of lasers being used, depending on the wavelength, sensitivity, and spectral resolution required. Gas lasers such as Ar+ and HeNe lasers have been used quite commonly due to their narrow, welldefined spectral lines, stability, and beam quality. High-intensity Ar+ ion lasers, in particular, are expensive to maintain, because they consume much electrical power and cooling water. Diode lasers are more convenient; powerful diode lasers for Raman spectroscopy usually emit red wavelengths that are suitable for avoiding fluorescence and fit the sensitivity range of silicon detectors. Diode lasers must be frequency-stabilized, because their emission frequency is not defined by narrow transitions. Spontaneous emission background and low-intensity side-bands must, furthermore, be filtered out when diode lasers are used for high-sensitivity Raman measurements. Solid-state lasers emitting in the NIR (e.g., Nd : YAG lasers) are used in combination with Fourier-transform spectrometers. UV lasers are also used. It should be stressed that diode and solid-state lasers are becoming increasingly important compared to gas lasers. The sample optics focuses the laser onto the sample and collects the Raman scattered light. In the widespread-backscattering or 180° arrangement, a single objective serves for both, focusing the laser and collecting the scattered light. This makes measurements easy, as only the distance between the sample and objective has then to be adjusted. As transparency of the whole sample is not needed with a backscattering arrangement, Raman spectroscopy is well suited for in situ measurements at interfaces between gases, liquids, and solids with at least one transparent phase. Elastically scattered laser light must be removed in front of the spectrometer, because it is several orders of magnitude more intense than the Raman-scattered radiation, and would otherwise produce stray radiation at the detector. Holographic notch filters are normally used for this purpose, due to their high transmission of required wavelengths and strong suppression of unwanted wavelengths. A new generation of currently available notch filters allows the recording of spectra down to wavenumbers of 10 cm−1. However, it should be noted that, due to aging problems with notch filters, edge filters are a popular alternative. Premonochromators are preferable for measurements at low wavenumbers and for use with varying laser wavelengths. Spectral analysis of the Raman scattered light is achieved by using either dispersive or Fourier-transform (FT) spectrometers. Dispersive spectrometers decompose the radiation according to the wavelength by means of a grating monochromator (Figure 24.2), whereas FT spectrometers calculate the spectrum from an interference pattern (interferogram) produced by an interferometer with a variable pathlength within one of the interferometer arms. Monochromators are preferable in
24.3 Instrumentation
Figure 24.2 Example of sensitive Raman equipment. The band pass filter (BP) cleans the laser radiation. The high NA objective lens L1 focuses the laser on the sample and collects the Raman-scattered radiation within a large solid angle. The band-rejection filter (BR) blocks elastically scattered radiation.
Imaging of the sample on the entrance slit (S) of the monochromator reduces stray radiation. Within the monochromator, the Raman-scattered radiation is dispersed by a phase grating (G) and finally focused on the CCD multichannel detector by lens L4.
the UV/visible spectral range. Compact and robust monochromators without moving parts can be used in rough environments, whereas larger high-resolution devices are mainly used for research purposes. In combination with array detectors, imaging monochromators can employ the multichannel advantage, and thus enable large signal-to-noise ratios (SNRs) for short measurement times. FT spectrometers also afford high SNRs due to their throughput advantage. However, the noise generated by their NIR detectors is considerably greater than that of detectors operating in the visible region. The major advantages of FT spectrometers are their strong suppression of fluorescence (as a result of using long-wavelength excitation) and variable spectral resolution. Charge-coupled device (CCD) detectors are preferred in the visible spectral range, due to their low dark current, high quantum efficiency, and multichannel capability. However, they must be cooled either thermoelectrically or by use of liquid nitrogen for high-sensitivity applications, such as surface spectroscopy. As the sensitivity of CCD detectors decreases considerably at wavelengths >1 μm, owing to the band-gap of the silicon material, other semiconductor detectors with a larger dark current (e.g., germanium or InGaAs) must be used for NIR Raman spectroscopy. Different accessories are available for Raman spectroscopy. For example, confocal microscopes afford a lateral resolution of approximately 1 μm and depth resolution down to approximately 2 μm. Such depth resolution is especially important for surface and interface spectroscopies, because it helps to eliminate the Raman intensities from the bulk phases. Fiber optics can also be used to guide the exciting
381
382
24 Surface Raman Spectroscopy
laser light to the sample and the Raman scattered light to the spectrometer; this greatly facilitates changing the samples and also eliminates the hazards of freely propagating laser light. Special sample cells have been designed for measurement at high or low temperatures and at high pressures, and for laser multipass through gaseous samples. Recently developed variations of Raman spectroscopy, such as spatial offset Raman scattering (SORS), even allow the provision of sample information on subsurface layers at a depth of some millimeters; examples include medical studies for bone analysis or breast cancer detection [7]. It should also be mentioned that transmission Raman spectroscopy has recently attracted renewed attention for the analysis of bulk objects such as tablets [8]. Moreover, handheld portable units utilizing a 785 nm laser are also commercially available, for use in identifying chemicals onsite. Systems for stand-off detection are also currently under development [9].
24.4 Spectral Information
For large molecules, at least, Raman spectra contain numerous bands that cannot always be assigned completely to particular vibrational modes. However, the majority of bands can, when measured with appropriate spectral resolution, enable the unambiguous identification of substances by comparing the spectral pattern (“fingerprint”) with those of reference spectra, if the latter are available. Several bands in characteristic spectral regions can definitely be assigned to molecular or crystal vibrations, especially if information on the chemical composition of the sample is available. For example, these bands enable the identification of aromatic or aliphatic hydrogen, carbonyls, thiols, double or triple bonds, heterocycles and other groups, or skeletal components. The vibrations of weakly polar and even symmetrically bonded atoms usually result in intense Raman bands, because the Raman activity is connected with changes in molecular polarizability during the vibration. This is a major difference from IR spectroscopy, in which bands from vibrations causing changes of the dipole moment dominate the spectrum. The Raman bands of carbon–carbon bonds are easily observable; strong vibrational coupling along carbon chains and symmetry selection rules cause these bands to be sensitive markers of substitution at aromatic rings, and for conformation of the carbon backbone. Structural isomers usually have very different spectra and can, therefore, be easily distinguished. Polarization effects represent another feature of Raman spectroscopy that improves the assignment of bands, and also enables the determination of molecular orientation. Analysis of the polarized and nonpolarized bands of isotropic phases enables determination of the symmetry of the respective vibrations. For aligned molecules in crystals or at surfaces, it is possible to measure the dependence of up to six independent Raman spectra on the polarization and direction of propagation of incident and scattered light relative to the molecular or crystal axes.
24.6 Applications
24.5 Quantification
In Raman spectroscopy the intensity of scattered radiation depends not only on the polarizability of a given vibrational mode and concentration of the analyte molecules, but also on the optical properties of the sample and the adjustment of the instrument, particularly the laser power. Therefore, absolute Raman intensities are not inherently a very accurate measure of concentration. These intensities are, of course, useful for quantification under well-defined experimental conditions and for well characterized samples; otherwise, relative intensities should be used instead. Raman bands of the major component, the solvent, or another component of known concentration can be used as internal standards. For isotropic phases, the intensity ratios of Raman bands of the analyte and the reference compound depend linearly on the concentration ratio over a wide concentration range and are, therefore, very well-suited for quantification. Changes of temperature and the refractive index of the sample can, however, influence Raman intensities, and the band positions can be shifted by different solvation at higher concentrations or in different solvent mixtures. Although these effects can be negligible, they must be considered if high accuracy is required. Chemometric computations might be required for quantification of complex mixtures. These computations perform much better than, for example in NIR spectroscopy, due to the usually narrow and well-resolved Raman bands. Quantification at surfaces is more difficult, because the Raman intensities depend not only on the surface concentration but also on the orientation of the Raman scatterers and the (usually unknown) refractive index of the surface layer. However, if noticeable changes of orientation and refractive index can be excluded, the Raman intensities are roughly proportional to the surface concentration, and intensity ratios with a reference substance at the surface will provide quite accurate concentration data. Further information on quantitative Raman scattering is available in an excellent review [10].
24.6 Applications 24.6.1 Unenhanced Raman Spectroscopy at Smooth Surfaces
Recent developments in Raman equipment have led to a considerable increase in sensitivity, and this has enabled the monitoring of reactions of organic monolayers on glassy carbon [11] and diamond surfaces, as well as analysis of the structure of Langmuir–Blodgett monolayers, without any enhancement effects. Although such unenhanced surface-Raman spectroscopy is expected to be applicable to a variety of technically or scientifically important surfaces and interfaces, it does
383
384
24 Surface Raman Spectroscopy
Figure 24.3 Raman spectra of (a)
nitrophenyl-modified and (b) untreated glassy carbon, the difference between spectra (a) and (b) (c), and the reference spectrum of solid 4-nitrobiphenyl (d). The strongest surface bands at approximately 1336 and
1586 cm−1 might be affected by subtraction of closely neighboring, much stronger bands of the carbon bulk. Five further marked surface bands are clearly visible in the difference spectrum (c) [11].
nevertheless require careful optimization of the apparatus, data treatment, and sample preparation. Nitrophenyl groups bonded covalently to glassy carbon and graphite surfaces have been detected and characterized using unenhanced Raman spectroscopy in combination with voltammetry and XPS [11]. The difference spectra from glassy carbon with and without nitrophenyl modification contained several Raman bands from the nitrophenyl group, with a comparatively large SNR (Figure 24.3). Electrochemical modification of the adsorbed monolayer was observed spectrally, because this led to clear changes in the Raman spectrum. The utilization of resonance effects can facilitate unenhanced Raman measurement of surfaces, and render the technique more versatile. For instance, a fluorescein derivative and another dye were used as resonant Raman scattering labels for hydroxyl and carbonyl groups on glassy carbon surfaces. The labels were bonded covalently to the surface, their fluorescence was quenched by the carbon surface, and their resonance Raman spectra could be observed at surface coverages
24.6 Applications
of approximately 1%. These labels enabled access to changes in surface coverage by C–OH and C=O with acidic or alkaline pretreatment [12]. Raman spectroscopy has also been used to determine the structure of Fe(CN)36−/ 4 − and its reaction products in chromate conversion coatings on an aluminum aircraft alloy, and to monitor coating growth rates and their dependence on the coating bath composition. The Raman spectra of the coatings contained bands of Berlin green, a Fe3+–CN–Fe3+ polymer, and Fe(CN)3− 6 physisorbed onto Cr(OH)3. The results obtained from different coating baths indicated a redox-mediation action for Fe(CN)3− 6 as the mechanism of acceleration of coating formation [13]. 24.6.2 Porous Materials
The surfaces of porous materials, such as catalysts, molecular sieves, or adsorbents, are much more readily accessible than smooth surfaces to Raman spectroscopy, because larger amounts of adsorbed substance can be placed within the laser focus, thus contributing to the scattering process. Raman spectroscopy has provided information on catalytically active transition metal oxide species (e.g., V, Nb, Cr, Mo, W, and Re) present on the surface of different oxide supports (e.g., alumina, titania, zirconia, niobia, and silica). The structures of the surface metal oxide species were reflected in the terminal M=O and bridging M–O–M vibrations. The location of the surface metal oxide species on the oxide supports was determined by monitoring the specific surface hydroxyls of the support that were being titrated. The surface coverage of metal oxide species on the oxide supports could be quantitatively obtained, because at monolayer coverage all of the reactive surface hydroxyls were titrated and additional metal oxide resulted in the formation of crystalline metal oxide particles. The nature of surface Lewis and Brønsted acid sites in supported metal oxide catalysts has been determined by adsorbing probe molecules such as pyridine. Information on the behavior of the surface metal oxide species during catalytic reactions has been provided by in situ characterization studies [14]. Investigations of synergistic effects were performed on 18O-exchange and the reoxidation of SbOx–O3 physical mixtures. In this case, peroxide species were detected on working methanecoupling catalysts at temperatures up to 1070 K, and identified as centers for the activation of methane [15]. The diffusion, location, and interactions of guests in zeolite frameworks has been studied using both in situ Raman spectroscopy and Raman microscopy. For example, the location and orientation of crown ethers used as templates in the synthesis of faujasite polymorphs has been studied in the framework they helped to form [16]. The polarized Raman spectra of p-nitroaniline molecules adsorbed into the channels of AlPO4-5 molecular sieves revealed their physical state and orientation; those molecules within the channels formed either a phase of head-to-tail chains similar to that in the solid crystalline substance, with a characteristic ω3 band at 1282 cm−1, or a second phase characterized by a similarly strong band at about 1295 cm−1. This second phase consisted of weakly
385
386
24 Surface Raman Spectroscopy
interacting molecules in a pseudo-quinonoid state similar to that of molten p-nitroaniline [17]. 24.6.3 Surface-Enhanced Raman Spectroscopy (SERS)
Since silver, gold, and copper electrodes are easily activated for SERS by roughening by the use of reduction–oxidation cycles, SERS has been widely applied in electrochemistry to monitor the adsorption, orientation, and reactions of molecules at those electrodes in situ. Special cells for SERS spectroelectrochemistry have been manufactured from chemically resistant materials, and with a working electrode accessible to the laser radiation. The versatility of such a cell has been demonstrated in electrochemical reactions of corrosive, moisture-sensitive materials such as oxyhalide electrolytes [18]. Langmuir–Blodgett (LB) films and self-assembled monolayers (SAMs) deposited on metal surfaces have been studied using SERS spectroscopy in several investigations. For example, mono- and bilayers of phospholipids and cholesterol deposited on a rutile prism with a silver coating have been analyzed in contact with water. The study results showed that, in these models of biological membranes, the second layer modified the fluidity of the first monolayer, and revealed the conformation of the polar head close to the silver [19]. The coating of metal SERS substrates with organic receptor layers renders SERS suitable for highly sensitive chemo-optical sensors for organic trace analysis in water and air [20, 21]. The combination of selective adsorption by the receptor with measurement of molecule-specific SERS spectra can lead to considerable reductions in the cross-sensitivities. Inclusion complexation with immobilized calix[4] arene receptors led to clearly decreased detection limits for aromatic compounds without groups that could attach to metals. SERS bands of the receptor were used as internal standards for the surface concentration of the aromatic compounds; subsequently, the intensity ratios of the aromatic compound and receptor bands provided a measure of the concentration of the solution of the aromatic compounds, with a dynamic range of two orders of magnitude. The enthalpy of the adsorption of chlorobenzene was determined from the temperature-dependence of the SERS intensities [20]. SERS substrates with bare metal surfaces irreversibly adsorb thio-organics (Figure 24.4) and other compounds, and can thus serve for the detection and identification of very low gas or solution concentrations of these substances [22]. SERS is especially well suited to the analysis of traces of gases, because it combines the measurement of surface concentration with an extremely high sensitivity. A monolayer in a typical focus of a laser with a diameter of 10 μm has a mass in the range of 10 femtograms! Even smaller amounts of substance are easily detectable, because 1% of a monolayer in a region 1 μm in diameter results in SERS of sufficient intensity. SERS has also been applied as a sensitive, molecule-specific detection method in chromatography, for example, thin-layer, liquid, and gas chromatography (for a review of this subject, see Ref. [23]).
24.7 Nonlinear Optical Spectroscopy
Figure 24.4 (a) Raman spectrum of methyl mercaptan and (b) SERS spectrum of methyl mercaptide, formed by adsorption of the mercaptan onto a silver surface. The surface reaction is proven by the disappearance of the S–H stretching and bending bands at 2575 cm−1 and 806 cm−1, respectively. The
Raman shift of the C–S stretching band at approximately 700 cm−1 is reduced during adsorption by withdrawal of electron density from the C–S due to bonding to the silver. The symmetric methyl stretching appears above 2900 cm−1 [22].
Selective SERS tags based on SERS reporter molecules attached to a SERS-active nanoparticle have been recently developed for different forms of biochemical assay. The simultaneous quantification of several differently labeled SERS tags has provided an improved selectivity over conventional fluorescence-based assays [24, 25]. 24.6.4 Near-Field Raman Spectroscopy
Raman scattering excited by near-fields close to either a small aperture of a tapered optical fiber, a single small field-enhancing particle, or an accordingly prepared tip of a scanning probe microscope, enables analysis with a spatial resolution beyond the diffraction limit. This technique enables not only the characterization of lateral structures with nanometer dimensions, but also the measurement of very thin layers without a strong background from the bulk. Initial investigations demonstrated the detection of dye and C60 molecules on glass and metal surfaces [26–28] (Figure 24.5). Further information on this subject is available in Chapter 31.
24.7 Nonlinear Optical Spectroscopy 24.7.1 Sum Frequency Generation (SFG) Spectroscopy
Unlike linear optical effects – such as absorption, reflection, and scattering – secondorder nonlinear optical effects are inherently specific for surfaces and interfaces.
387
388
24 Surface Raman Spectroscopy
Figure 24.5 Comparison of resonance
Raman spectra with and without tip enhancement for 0.5 monolayers of brilliant cresyl blue on a smooth gold film. The tip strongly increased the total Raman intensity,
when positioned at a distance of 1 nm from the sample surface. Several other bands were made visible as a result of the tip enhancement [28].
These effects, namely second harmonic generation (SHG) and sum frequency generation (SFG), are dipole-forbidden in the bulk of centrosymmetric media. In the investigation of isotropic phases – such as liquids, gases, and amorphous solids, in particular – signals arise exclusively from the surface or interface region, where the symmetry is disrupted. Nonlinear optics are applicable in situ, without the need for a vacuum, and the time response is rapid. Surface SHG [29] produces frequency-doubled radiation from a single pulsed laser beam, while the intensity, polarization dependence, and rotational anisotropy of the SHG provide information regarding the surface concentration and orientation of the adsorbed molecules, and on the symmetry of surface structures. SHG has been used successfully for the analysis of adsorption kinetics and ordering effects at surfaces and interfaces, the reconstruction of solid surfaces and other surface phase transitions, and potential-induced phenomena at electrode surfaces. For example, orientation measurements were used to probe the intermolecular structure at air–methanol, air–water, and alkane–water interfaces, and within mono- and multilayer molecular films. Time-resolved investigations have revealed the orientational dynamics at liquid–liquid, liquid–solid, liquid–air, and air–solid interfaces [29]. SFG [30, 31] employs both visible and IR lasers for the generation of their sum frequency. Tuning the IR within a certain spectral range enables the monitoring of molecular vibrations of adsorbed molecules, with surface selectivity. SFG includes the capabilities of SHG and can, in addition, be used to identify molecules
24.7 Nonlinear Optical Spectroscopy
and their structure on the surface by analyzing the vibrational modes. SFG has also been used to observe surfactants at liquid surfaces and interfaces, and the ordering of interfacial water molecules. Further information on SFG is available in Chapter 26. 24.7.2 Coherent Anti-Stokes Raman Scattering (CARS)
CARS, which is probably the most commonly used nonlinear Raman technique [32], overcomes several limitations of spontaneous Raman spectroscopy, namely the requirements for a high average power and the amount of time required to collect a spectrum, especially from dynamic or living systems [33]. Although, currently, many instrumental variations of CARS are available, in its simplest form CARS spectroscopy requires three laser sources: (i) a pump laser to create a virtual state, as per spontaneous Raman scattering; (ii) a second laser having a frequency equal to that which would be scattered in spontaneous Stokes Raman, which creates an excited vibrational state; and (iii) a third laser to create a second virtual state. These are often referred to in terms of the “pump field,” “Stokes field,” and “probe field,” respectively. Scattering from the second virtual state to the ground state gives rise to the CARS signal [32–34]. In most CARS experiments, the pump and probe frequencies are the same. As with spontaneous Raman, CARS signals are also thought to be enhanced when close to metal spheres (as considered theoretically by Chew et al. [35]). The generation of surface plasmons has been utilized in the analysis of a range of analytes, including carbon nanotubes and copper phthalocyanine thin films [36]. It should be noted, however, that inherent in the CARS signal is the existence of the nonresonant background contributions which mix coherently with the resonant signal. These complications may lead to a red shift in the spectral peak, thus complicating spectral assignments. Although suppression of the background nonresonant contribution can be achieved, it is not a trivial process [37]. Further information on CARS is available elsewhere [38]. 24.7.3 Stimulated Femtosecond Raman Scattering (SFRS)
In SFRS, a narrow, intense, 1- to 3-ps, pump frequency laser, ωr, and a weak, broad, 30–50 fs probe, ωc (usually obtained from white light), overlap at the sample location. Stimulated processes will occur in the sample if the frequency component of the probe, minus that of the pump, is equal to the frequency of a vibration (|ωc − ωr| = ωv). When the difference is positive/negative, a photon with the frequency ωc is annihilated/generated. As the probe is spectrally broad, all Raman resonances of a sample are probed at once, manifesting themselves as modulations in the white-light spectrum from which the Raman spectrum can be retrieved [39–41]. Like CARS, again FSRS offers advantages over spontaneous Raman in terms of increased speed of data collection and fluorescence rejection; it also has an advantage over CARS in that it generates distortion-free spectra that are free
389
390
24 Surface Raman Spectroscopy
from nonresonant contributions [39]. Further information on SFRS is available elsewhere [42, 43]. 24.7.4 Spatially Offset Raman Spectroscopy (SORS)
SORS is most often employed to probe subsurface layers in turbid media, or to penetrate through packaging layers to interrogate the substances below. In a SORS set-up, Raman scattered photons are collected from a position offset from the point of incidence from the probing laser beam. At different offset distances, the contributions from the surface and deeper layers will differ; that is, the smaller the offset distance, the higher the contribution of the top layer to the spectrum. Moving away from the probe laser, the contribution of the top layer will diminish rapidly, and Raman-scattered photons from deeper layers will be preferentially recorded, as these are likely to have traveled laterally through the sample before being emitted from the sample. By measuring at a number of defined offsets, combined with a multivariate analysis, the spectrum of each layer can be elucidated [44]. To date, SORS has been applied to many analytical problems where the sample is contained in bottles or packets, or coated with a substance not of interest. This allows the direct, noninvasive measurement of, for example, liquid explosives [45] and tablets in blister packets [46]. The noninvasive nature of SORS also opens up potential biological applications to identify tissue components and layers [47].
References 1 Raman, C.V. and Krishnan, K.S. (1928) Nature, 121, 501. 2 Kneipp, K., Moskovits, M., and Kneipp, H. (eds) (2006) Surface-enhanced Raman scattering – Physics and applications, in Topics in Applied Physics, vol. 103, Springer, Heidelberg, Berlin. 3 Fleischman, M., Hendra, J.P., and McQuillan, A.J. (1974) Chem. Phys. Lett., 26, 163. 4 Jeanmaire, D.J. and Van Duyne, R.P. (1977) J. Electroanal. Chem., 84, 1. 5 Nie, S. and Emroy, S.R. (1997) Science, 275, 1102. 6 Weaver, M.J., Zou, S., and Chan, H.Y.H. (2000) Anal. Chem., 72, A38. 7 Matousek, P. and Stone, N. (2009) Analyst, 134 (6), 1058–1066. 8 Aina, A., Hargreaves, M.D., Matousek, P., and Burley, J.C. (2010) Analyst, 135 (9), 2328–2333.
9 Hobro, A.J. and Lendl, B. (2009) Trends Anal. Chem., 28, 1235–1242. 10 Pelletier, M.J. (2003) Appl. Spectrosc., 57, A20–A42. 11 Liu, Y.-C. and McCreery, R.L. (1995) J. Am. Chem. Soc., 117, 11254. 12 Ray, K.R. and McCreery, R.L. (1999) J. Electroanal. Chem., 469, 150. 13 Xia, L. and McCreery, R.L. (1999) J. Electrochem. Soc., 146, 3696. 14 Wachs, I.E. (1996) Catal. Today, 27, 437. 15 Knozinger, H. (1996) Catal. Today, 32, 70. 16 Bremard, C. and Bougeard, D. (1995) Adv. Mater., 7, 10. 17 Hill, W., Marlow, F., and Kornatowski, J. (1994) Appl. Spectrosc., 48, 224. 18 Mosier-Boss, P.A., Newbery, R., Szpak, S., Lieberman, S.H., and Rovang, J.W. (1996) Anal. Chem., 68, 3277.
References 19 Souto, J., Aroca, R., and Desaja, J.A. (1994) J. Raman Spectrosc., 25, 435. 20 Hill, W., Wehling, B., Gibbs, C.G., Gutsche, C.D., and Klockow, D. (1995) Anal. Chem., 67, 3187. 21 Hill, W., Fallourd, V., and Klockow, D. (1999) J. Phys. Chem. B, 103, 4707. 22 Hill, W., Wehling, B., and Klockow, D. (1999) Appl. Spectrosc., 53, 547. 23 Hobro, A.J. and Lendl, B. (2010) SERS and separation science, in Surface Enhanced Raman Spectroscopy – Analytical, Biophysical, and Life Science Applications (ed. S. Schluecker), Wiley-VCH Verlag GmbH, Weinheim, pp. 155–169. 24 Graham, D., Faulds, K., Thompson, D., McKenzie, F., Stokes, R., Dalton, C., Stevenson, R., Alexander, J., Garside, P., and McFarlane, E. (2009) Biochem. Soc. Trans., 37, 697–701. 25 Stevenson, R., Faulds, K., and Graham, D. (2010) Quantitative DNA analysis using surface-enhanced resonance Raman scattering, in Surface Enhanced Raman Spectroscopy – Analytical, Biophysical, and Life Science Applications (ed. S. Schluecker), Wiley-VCH Verlag GmbH, Weinheim, pp. 241–259. 26 Emory, S.R. and Nie, S. (1997) Anal. Chem., 69, 2631. 27 Stöckle, R.M., Suh, Y.D., Deckert, V., and Zenobi, R. (2000) Chem. Phys. Lett., 318, 131. 28 Pettinger, B., Picardi, G., Schuster, R., and Ertl, G. (2000) Electrochemistry, 68, 942. 29 Bloembergen, N. and Pershan, P.S. (1962) Phys. Rev., 128, 606. 30 Du, Q., Freysz, E., and Shen, Y.R. (1994) Science, 264, 826. 31 Richmond, G.L. (1997) Anal. Chem., 69, A536. 32 Smith, E. and Dent, G. (eds) (2005) Modern Raman Spectroscopy – A Practical Approach, John Wiley & Sons, Ltd, Chichester.
33 Cheng, J.-X. and Xie, X.S. (2004) J. Phys. Chem. B, 108, 827–840. 34 Tolles, W.M., Nibler, J.W., McDonald, J.R., and Harvey, A.B. (1977) Appl. Spectrosc., 31, 253–271. 35 Chew, H., Wang, S.-D., and Kerker, M.J. (1984) Opt. Soc. Am. B., 1, 56–66. 36 Baltog, I., Baibarac, M., and Lefrant, S. (2005) J. Opt. A – Pure Appl. Opt., 7, 632–639. 37 Fang, C., Lu, F., Zheng, W., and Huang, Z. (2010) Opt. Express, 18, 15714–15724. 38 Müller, M. and Zumbusch, A. (2007) Chem. Phys. Chem., 8, 2156–2170. 39 Ploetz, E., Laimgruber, S., Berner, S., Zinth, W., and Gilch, P. (2007) Appl. Phys. B Lasers Opt., 87, 389–393. 40 Sun, Z., Lu, J., Zhang, D.H., and Lee, S.-Y. (2008) J. Chem. Phys., 128, 144114. 41 Lee, S.-Y., Zhang, D., McCamant, D.W., Kukura, P., and Mathies, R.A. (2004) J. Chem. Phys., 121, 3623–3642. 42 Kukura, P., Yoon, S., and Mathies, R.A. (2006) Anal. Chem., 78, 5952–5959. 43 Kukura, P., McCamant, D.W., and Mathies, R.A. (2007) Annu. Rev. Chem., 58, 461–468. 44 Matousek, P., Clark, I.P., Draper, E.R.C., Morris, M.D., Goodship, A.E., Everall, N., Towrie, M., Finney, W.F., and Parker, A.W. (2005) Appl. Spectrosc., 59, 393–400. 45 Eliasson, C., Macleod, N.A., and Matousek, P. (2007) Anal. Chem., 79, 8185–8189. 46 Ricci, C., Eliasson, C., MacLeod, N.A., Newton, P.N., Matousek, P., and Kazarian, S.G. (2007) Anal. Bioanal. Chem., 389, 1525–1532. 47 Stone, N., Baker, R., Prieto, C.H., and Matousek, P. (2008) Progress in Biomedical Optics and Imaging – Proceedings of SPIE 6853, art. no. 68530N.
391
393
25 UV-VIS-IR Ellipsometry (ELL) Bernd Gruska and Karsten Hinrichs
25.1 Principles
According to a widely accepted definition, ellipsometry is the art of measuring and analyzing the elliptical polarization of light [1]. Typically, optical simulations are used to correlate the measured ellipsometric parameters with the sample properties, such as film thickness and complex refractive index. The complex refractive index is given by the sum of the real part n and the imaginary part k (where k is defined as extinction coefficient or index of absorption). In particular, ellipsometry [2, 3] is a sensitive method for the analysis of single films, layer stacks, substrate materials, rough surfaces, interfaces, material gradients, as well as mixtures of different materials. Films with thicknesses of between 0.1 nm and 200 μm can be investigated, depending on the spectral range used and the homogeneity of the films. Thicknesses less than 1 μm can be determined with a sensitivity better than 0.01 nm, while thicknesses in the micrometer range typically are analyzed with a sensitivity better than 1 nm. The refractive index of a film or a substrate material can be determined with a sensitivity better than 5 × 10−4, which is the highest available value for noninvasive optical measuring methods, especially for thin films. The extinction coefficient can be revealed with nearly the same sensitivity, which corresponds to a lower limit of 10–100 cm−1 for the absorption coefficient of the material. The number of measurable layers of a stack is limited only by the optical contrast between the different layers. In practice, stacks of ten layers and more can be analyzed by ellipsometry. The further advantages of ellipsometry compared to other metrological methods are the noninvasive and nondestructive characteristics of the optical method, the low energy input into the sample, the direct measurement of the dielectric function of materials, and the ability to characterize any type of optically transparent material in different environments (e.g., liquid, atmosphere, vacuum). The principle of ellipsometric measurement is shown schematically in Figure 25.1. In this case, linearly polarized light is incident on a sample surface, which Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
394
25 UV-VIS-IR Ellipsometry (ELL)
Figure 25.1 Schematic representation of the ellipsometric principle. The incident light is
linearly polarized; the elliptically polarized reflected light is described by the relation between the measured quantities (Ψ, Δ) [2–5].
must be flat and sufficiently reflecting; the reflected light then becomes elliptically polarized due to the sample’s properties. The parameters measured by ellipsometry are: tanΨ, the amplitude ratio of the resolved components of the electric field vector of the reflected light parallel to and perpendicular to the plane of incidence; and Δ, the phase difference of the two components. A more sensitive and unambiguous determination of the phase angle Δ in the whole data range between 0 and 360 ° is possible by using a retarder introducing a defined additional phase shift. The properties of the sample are correlated with the reflection coefficients, and therefore with the complex ratio ρ of the Fresnel reflection coefficients parallel to (rp) and perpendicular to (rs) the incident plane-polarized electric field vector:
ρ=
rp rs
(25.1)
The fundamental equation of ellipsometry [5]
ρ = tan Ψ ⋅ e iΔ
(25.2)
describes the correlation between the measured quantities Ψ, Δ, and the sample properties. In principle, this complex equation can be solved analytically only for pure substrates; more complex sample structures require an optical modeling of the sample and a fit of the calculated ratio ρ to the measured quantities Ψ and Δ. The film thickness, refractive index, extinction coefficient, and other parameters are calculated on the basis of an optical model. The choice of this model is a fundamental assumption for determination of correct values of the calculated parameters.
25.2 Instrumentation
The thickness of a film influences the interference of light waves reflected from the front and reverse side of the film, and hence the reflectance. Therefore, the thickness of an absorbing film can only be measured while there is still a contribution of reverse-side reflection to the reflectance of the sample. In particular, for metallic layers the typical measurable thicknesses are then less than 50 nm. Ellipsometric measurements depend on the incident angle ϕ due to the angle dependence of the reflection coefficients rs and rp. Large differences between both quantities are found for angles above 50°, whereby rp has a local minimum at the so-called Brewster angle. This angle is between 50° and 60° for visible-transparent materials, and greater than 70° for most absorbing materials. Ellipsometric measurements close to the Brewster angle of the substrate are especially sensitive for very thin layers on top of the substrate. Multiple-angle measurements are suitable for confirmation of optical models and the determined properties. The two measured ellipsometric angles Ψ and Δ, at a fixed wavelength and a fixed angle of incidence, enable the calculation of a maximum of two other parameters, for example, the film thickness and the refractive index of a transparent layer. Multiple-angle measurements increase the number of measured quantities and, in turn, the number of parameters which can be determined for a specific sample. However, in this case the number of measurable parameters will also not be typically greater than four or five. Spectroscopic ellipsometry further increases the number of measurable parameters, and allows calculation of the dispersion of the refractive index and the extinction coefficients of materials. It also enables the analysis of more complex sample structures (e.g., multilayer stacks, interfaces, material gradients, material compositions) on the basis of parameterized dielectric functions, which drastically reduce the number of unknown parameters compared to the number of measured ellipsometric angles. Spectroscopic ellipsometry is sensitive to the dielectric functions of the different materials. The combination with other compositional analytical techniques such as AES or XPS and with scanning microscopy, can provide information about the vertical structure and add valuable additional information for creating a suitable optical model of an unknown complex sample structure.
25.2 Instrumentation
The basic configuration of an ellipsometer is the following: light source–polarizer– sample–polarizer–detector. Advanced configurations use a phase-shift device (a retarder) located before or behind the sample. There are many possibilities in which a specific ellipsometer can be realized. The main differences involve in the way in which the light is polarized, and the detection of the polarization state of the reflected light. The light source may be either monochromatic [lasers, white light (Xe arc lamp, deuterium lamp, halogen lamp) with a monochromator] or chromatic [white light, silicon carbide rods (glowbars; mainly for IR ellipsometry)].
395
396
25 UV-VIS-IR Ellipsometry (ELL)
Photomultiplier tubes (vacuum UV), Si photodiodes (UV-Vis), Ge or InGaAs photodiodes (NIR), and MCT (MIR) or DTGS detectors (MIR) are used as broadband detectors. These must be operated with a monochromator if a white-light source is used. Fast spectroscopic ellipsometers use photometers and diode arrays as detectors; this allows the simultaneous detection of multiple wavelengths rather than a sequential detection. State-of-the-art ellipsometry in the IR spectral region (ca. 0.8–100 μm wavelength) employs modulated white light from an FT-IR instrument and broadband detectors to measure the reflected light; the intensity spectra are then generated by Fourier transformation of the detected signal. Ellipsometric measurements are performed at fixed angles of incidence, with each ellipsometer configuration capable of being operated at a discrete-wavelength or spectroscopic (i.e., variable angle of incidence mode). Null ellipsometers are one of the oldest configurations, in which a rotating polarizer and a rotating compensator are used to change the polarization state of the incident light, while a rotating polarizer (analyzer) is used to analyze the reflected light. The angular position of the optical elements is changed until the light intensity on the detector is zero. Although this configuration is one of the most accurate, it is also the slowest – even if automated. Rotating-element ellipsometers have a continuously rotating element (10–40 Hz frequency of rotation), which allows automated acquisition and faster measurements. Today, most of the instruments used operate in this configuration, with either a polarizer (rotating polarizer ellipsometer, RPE) or an analyzer (rotating analyzer ellipsometer, RAE) being used as the rotating element. The detector signal is measured as function of time, and is then Fourier-analyzed to obtain the ellipsometric angles Ψ and Δ. The RPE configuration has the advantage that any polarization sensitivity of the detector [photomultiplier tube (PMT), monochromator] does not influence the measurement. The polarization state of the detected light is fixed by the fixed analyzer position; the disadvantage of this is the sensitivity to residual polarized light of the light beam that is incident on the rotating polarizer. The RAE configuration has the advantage of being insensitive to any residual polarization of the light source (laser, white-light source with monochromator), but errors may occur due to polarization sensitivity of the detector (PMT). With polarization-insensitive Si photodiodes as detectors, this configuration is most often used for discrete-wavelength ellipsometers. Typically, single-wavelength ellipsometers are equipped with a quarter-wave plate (compensator) which can be moved into the light beam. This allows the linear polarized light to be changed into circularly polarized (in general, elliptically polarized) light after passing through the quarter-wave plate. Consequently, the measured ellipsometric angle Δ is shifted by 90°, and Δ can be determined over the whole data range from 0° to 360°, with almost the same accuracy; the major uncertainty when Δ is close to 0° and 180° is removed, however. Spectroscopic ellipsometers that measure Δ over the whole data range require an achromatic compensator with a phase shift close to 90° over a large spectral range. The compensator can be located between the polarizer and the sample, or between the sample and the analyzer.
25.2 Instrumentation
Step-scan polarizer/analyzer (SSP/SSA) and continuously rotating element configurations are used for diode-array or FT-IR-based ellipsometers. In the SSP or SSA modes, complete intensity spectra are measured at a number of angle positions of the rotating polarizer/analyzer. In this case, the acquisition time will vary between some seconds (UV-Vis) and several minutes (IR), depending on the intensity of the incident beam and the reflectance of the sample. However, the measurements are made rapidly compared to the monochromator-based configurations, and are very accurate with regard to the exact positioning of the rotating element. The continuously rotating element configuration is used for very fast spectroscopic ellipsometers measuring complete ellipsometric spectra in less than 100 ms. For such instruments, the intensity at each wavelength (pixel of the diode array) is accumulated over a 45° revolution angle of the rotating element. Continuously rotating compensator ellipsometers have a fixed setting of polarizer and analyzer. The advantage of this configuration is the insensitivity to residual polarization of the light source or a polarization sensitivity of the detector. A possible disadvantage, however, is the influence of a small misalignment of the rotating achromatic compensator on the performance of the instrument. Recently developed research instruments have two rotating compensators that allow the measurement of all elements of the Mueller matrix (Mueller matrix ellipsometry) [3, 6]. Polarization modulation ellipsometers employ a photoelastic modulator to modulate the polarization state of the incident beam. The polarizer and analyzer are at fixed azimuthal positions during the measurement, and a Fourier analysis of the time-dependent signal provides the ellipsometric parameters Δ and Ψ. The advantage of this configuration is the very high measuring speed (typical modulation frequency of ca. 50 kHz) for discrete wavelengths, and the lack of any moving parts. Moreover, the ellipsometric angle Δ can be measured over the whole range of 0° to 360°. Unfortunately, the necessary frequent calibration of the instrument, owing to the high temperature dependence of the polarization modulator, limits the measurement time. The IR ellipsometer consists of a combination of a Fourier-transform spectrometer (FTS) and a photometric ellipsometer. One of the two polarizers (the analyzer) is moved step-by-step in four or more azimuths, because the spectrum must be constant during the scan of the FTS. From these spectra, the tanΨ and cosΔ spectra are calculated. In this case, only Δ is determined in the range of 0°–180°; thereby, the measurement accuracy is very low for Δ values close to 0 and 180°. This problem can be overcome, however, by using a retarder (compensator) with a phase shift of approximately 90° in a second measurement. Thereby, cosΔ and sinΔ are measured independently with full phase information [7]. In a second type of IR ellipsometer, a dynamic retarder consisting of a photoelastic modulator (PEM) replaces the static unit. The PEM produces a sinusoidal phase shift of approximately 40 kHz, and supplies the detector exit with signals of the ground frequency and the second harmonic. The ellipsometric spectra can then be determined from these two frequencies and two settings of the polarizer and PEM [8]. This ellipsometer system is mainly used to conduct rapid and relative measurements.
397
398
25 UV-VIS-IR Ellipsometry (ELL)
25.3 Applications 25.3.1 UV-Vis-NIR Spectral Region
Spectroscopic ellipsometry in the UV-Vis-NIR region is mainly used for the characterization of new materials (e.g., photoresists, polymers, low-k dielectrics, semiconductors, composite materials), the analysis of multilayer stacks (e.g., low-k and antireflection (AR) layer stacks on architectural glass, flat-panel applications, photovoltaic applications, silicon IC technology, optoelectronic devices, photonic devices), the qualification of a large number of deposition processes for production and R&D applications, as well as the in situ monitoring of growth and deposition processes. Both, spectral range and speed of the measurement are key features which recommend the use of spectroscopic ellipsometers for R&D and production surveillance. The existence of spectroscopic ellipsometers in the spectral range of 140 nm to 2.5 μm allows for many applications in metrology. Ellipsometry in the vacuum-UV (<190 nm) allows the analysis of materials for next-generation lithography (photoresists, AR coatings). The short wavelengths increase the sensitivity of ellipsometric measurements of ultrathin films (<10 nm). New prospects are expected for the analysis of thin metallic and dielectric layers. In the UV, most of the interesting materials – such as Si, polysilicon, transparent conductive oxides (TCO), SiGe, GaAs, and other semiconductor materials –are strongly absorbent; this permits surface-sensitive measurements. Band gap, surface roughness, native oxide covering, material composition, and structural properties can be analyzed. The Vis–NIR spectral range – within which many materials (dielectrics, semiconductors, polymers) are transparent – is mainly used to measure the thickness of single films and layer stacks. The analysis of complex layer stacks often requires knowledge of the dielectric function of each material in the stack, or parameterized dispersion models to reduce the number of unknown parameters. The Cauchy formula is often used for transparent layers, while the Drude–Lorentz formula works well for metals, TCOs, and IR-active materials. The Leng formula [9] can be used for semiconductor materials, and the Tauc–Lorentz formula [10] works best for amorphous materials. The general Gaussian-broadened polynomial superposition (GBPS) parametric dispersion model [11] can be broadly applied to most materials, including crystalline and amorphous semiconductors, metals, and organics. Figure 25.2 shows the measured ellipsometric spectra of a-Si/SixNy multilayer stack deposited on GaAs. Here, four layers of SixNy and three layers of a-Si were alternately deposited, with the dispersion of each layer material being measured on single films. Film thicknesses of 78 nm for a-Si and 145 nm for SixNy were determined for the different layers of the stack by using appropriate parameterized dispersion models for each material. Other typical multilayer applications are dielectric stacks on silicon, spectral selective mirror stacks, and Bragg reflectors for laser devices.
25.3 Applications
Figure 25.2 Measured ellipsometric spectra of an a-Si/SiN multilayer stack (seven layers)
on GaAs.
Figure 25.3
Measured and calculated ellipsometric spectra from a 2.5 μm-thick SiO2 layer on Si.
Ellipsometric measurements in the NIR are often used for the determination of thickness and refractive index of thicker dielectric films and semiconductor stacks based on InP and GaAs, which are transparent in this spectral region. The wavelength and film thickness should not differ much, so that a highly sensitive ellipsometric measurement of film thickness and refractive index is obtained. Figure 25.3 shows the measured and fitted Δ-spectra of a 2.5 μm-thick SiO2 layer on silicon in a spectral range of 350 nm to 2.3 μm wavelength. The higher sensitivity in the NIR spectral range is clear. The growing number of applications of 1.55 μm semiconductor lasers continues to force the application of NIR ellipsometers to control and to monitor AR coatings on laser facets. A further application of NIR ellipsometry – the analysis of TCO films – plays an important role, especially in modern photovoltaic applications. The absorption of such conductive films starts typically in the NIR between 1 μm and 2 μm
399
400
25 UV-VIS-IR Ellipsometry (ELL)
Figure 25.4 Dispersion of refractive index n and absorption index k of ZnO : Al.
wavelength. Figure 25.4 shows the optical constants of an ZnO : Al film, where the increase in the absorption index above 1 μm wavelength can clearly be seen. Ellipsometric measurements in the NIR spectral range can also be applied to determine the sheet resistance of such films. Laterally inhomogeneous films and patterned structures of microelectronic and optoelectronic applications require small measuring spots. Today, measurements in 50 μm × 50 μm areas are standard for microspot spectroscopic ellipsometers used in production lines. Areas more than tenfold smaller can be analyzed by using discrete-wavelength ellipsometers equipped with laser light sources. 25.3.2 Infrared Ellipsometry
Interdisciplinary problems in the development of new materials, polymer research, biotechnology, nanotechnology, and solar cell research require methods for the structural analysis of thin films. In the recent years, IR ellipsometry has been proven to be well-suited for providing multifold structural information of thin films for various technological applications [3, 4, 12–14]. Infrared spectroscopic ellipsometry (IRSE), which is often carried out in the mid-infrared (MIR) range (400 to 5000 cm−1), is a technique for the measurement and analysis of the elliptical polarization of reflected or transmitted IR radiation from a sample. Besides the identification of materials (e.g., molecular groups and molecules), sample properties (e.g., conductivity, anisotropy, molecular orientation), it is also possible to derive film thicknesses and optical constants (refractive index and absorption index) from the measured parameters. If the spectral range is extended to the farinfrared (FIR, below 400 cm−1), often inorganic samples are studied, for example,
25.3 Applications
with respect to their phonon properties, superconducting behavior, metal–insulator transitions, and ferroelectric properties [15, 16]. Compared to standard IR transmission and reflection techniques (see Chapter 23), the ellipsometric experiment has the advantage that it provides at least two independent measurement parameters that facilitate optical interpretations and the quantification of inhomogeneities (e.g., roughness, variations in density or thickness). In the MIR, the ellipsometric parameters Ψ and Δ are influenced in particular by the absorptions of free carriers and characteristic bands caused by molecule vibrations or phonons. For thin films on semiconductors or isolators, the vibrational band shapes depend heavily on the average orientation of the corresponding transition dipole moments [12]. In contrast to this finding for metallic substrates, due to the so-called surface “selection rule” (see Chapter 23), only transmissionlike absorption bands are observed in tanΨ spectra of thin-film samples. For quantitative interpretation, the measured ellipsometric spectra typically are interpreted within optical simulations. In the macroscopic description, the dielectric function is thereby often represented by the sum of the high-frequency dielectric constant, contributions of free carriers, and the contributions of vibrational bands. In many cases, the vibrational absorptions of organic thin films in the MIR spectral range can be approximated by Lorentzian oscillators [12]. With regard to the vibrational contributions, two types of resonances can be distinguished in the IR: strong and weak resonances [14]. Hereby, the strong or weak resonance depends on the density and strengths of the single oscillators. Strong resonance is characterized by a region where the refractive index is lower than unity for wavenumbers higher than the oscillator frequency. This type of oscillator is mostly found for inorganic compounds with reststrahlen bands, such as silicon oxide (the n, k spectra of a 7 nm-thick silicon oxide film are shown in Figure 25.5).
Figure 25.5 n, k spectra as determined from a 7 nm-thick silicon oxide film as an example of a strong oscillator.
401
402
25 UV-VIS-IR Ellipsometry (ELL)
Figure 25.6 (a) n,k spectra of poly(vinyl
chloride) (PVC) as an example of a weak oscillator. The constants were revealed by a best fit of the ellipsometric spectra of a thick
single layer on a gold-coated glass substrate (PVC: d = (104.1 ± 0.5) nm and n∞ = 1.546; n∞ values are taken from Vis ellipsometry) [17].
In the ellipsometric spectra of thin oxide films on silicon with thicknesses up to approximately 100 nm, in addition to a band at the frequency position of the vibrational oscillators, the Berreman effect is also found near the position of n = k and n < 1, with a shift to higher wavenumbers relative to the oscillator frequency. In contrast, the refractive index of weak resonances is always greater than 1, as is typically found for many organic compounds. The optical constants of poly(vinyl chloride) (PVC) are shown in Figure 25.6. As can be seen in the spectrum for this compound, the refractive index in the shown spectral range is greater than 1, and the band amplitudes are much weaker than for silicon oxide, as shown in Figure 25.5. Double and buried layers can also be detected by using IR ellipsometry [14]. In the special case that the film material is characterized by reststrahlen bands, the detection sensitivity is increased. As discussed above, the response of strong oscillators is observed at the position of the Berreman effect (Figure 25.7). In this case, the SiOx absorption at around 1065 cm−1 produces a strong peak at about 1250 cm−1 (Figure 25.7). For the SiO2/Si3N4 layer sample, additionally a Si3N4 related Berreman mode at about 1135 cm−1 is found (Figure 25.7). However, current improvements in detector and interferometer technology also permit the study of ultrathin organic films with IR spectroscopic ellipsometry. Besides the molecular identification that is essential when investigating chemical processes, typical applications include the characterization of ordering, molecular tilt angles, adsorption, or etching characteristics. For organic silicon technology, in situ IR ellipsometry has recently become available that also allows the chemical
25.3 Applications
Figure 25.7 IR ellipsometric spectra of a very thin SiO2/Si3N4 double layer on silicon. A
scheme of the sample is shown at the right [13].
Figure 25.8 pH-dependent switching of PAA-mix-P2VP brush [18].
characterization of ultrathin films in liquid environments. As an example, a study of the switching behavior of stimuli-responsive mixed polymer brushes (d = 11 nm) is shown in Figure 25.8. In this case, the structural and chemical changes in a thin polymer brush layer were identified from the analysis of IR ellipsometric tanΨ spectra while the pH was varied; the spectra recorded at different pH-values,
403
404
25 UV-VIS-IR Ellipsometry (ELL)
as shown in Figure 25.8, were referenced to the spectrum with pH 6.5 [18]. The bands in Figure 25.8 at 1720 cm−1, 1630 cm−1, 1558 cm−1, and 1414 cm−1 were assigned to the symmetric stretching of the carbonyl group, the vibration of the protonated pyridine ring, and the asymmetric and symmetric stretching vibration of the dissociated carboxylate groups, respectively. The spectral changes could be explained by the responsive behavior (protonation and deprotonation reaction) of the polyacrylic acid/poly-2-vinylpyridine (PAA-mix-P2VP) brush in aqueous solutions with the pH ranging from 2 to 10. The systematic change in pH confirmed the reversible switching behavior of the PAA-mix-P2VP brush between three different states: swollen P2VP and compact PAA chains at pH 2; a compact “P2VP… PAA” complex at pH 6.5; and swollen PAA and compact P2VP chains at pH 10. The thickness at the differently swollen states of the polymer brush can be determined using Vis ellipsometry [19]. Today, the R&D of responsive materials such as functional polymer films and surfaces is strongly motivated by their high technological potential for adhesives, coatings, separation, microfluidic devices, surface transport, sensors, responsive colloids, and biomedical applications. In concluding this section on IR ellipsometry, it should be noted that it is normally possible to investigate very thin films down to the subnanometer range by the use of IR wavelengths, which are much larger than the thickness of the film (by a factor of 10 000). The typically probed spots in the laboratory are above 25 mm2 at monolayer sensitivity. The use of brilliant light sources (e.g., synchrotron radiation) is helpful in studies with monolayer sensitivity at higher lateral resolution (250–1000 μm). The combined application of UV, Vis, and IR ellipsometry can provide optical constants over the entire spectral range from UV to Vis, even for anisotropic thin films [20]. As an example, the optical constants of guanine are shown in Figure 25.9.
Figure 25.9 Ansiotropic optical constants of a 84 nm-thick guanine film [20].
References
References 1 (a)Rothen, A. (1964) Symposium Proceedings on Ellipsometry in the Measurement of Surfaces and Thin Films (eds E. Passaglia, R.R. Stromberg, and J. Kruger), National Bureau of Standards, Miscellaneous Publication 256, pp. 7–21; (b)Rothen, A. (1945) Rev. Sci. Instrum., 16, 26. 2 Tompkins, H.G. and Irene, E.A. (eds) (2005) Handbook of Ellipsometry, William Andrew Publishing Inc., Norwich, New York. 3 Fujiwara, H. (2007) Spectroscopic Ellipsometry, Principles and Applications, John Wiley & Sons, Ltd, Chichester, UK. 4 Smith, D.M. and Chughtai, A.R. (2007) Polymer optical and dielectric properties through vibrational spectroscopy, in Vibrational Spectrocopy of Polymers (eds N.J. Everall, J.M. Chalmers, and P.R. Griffiths), John Wiley & Sons, Ltd, Chichester, pp. 237–254. 5 Azzam, R.M.A. and Bashara, N.M. (1977) Ellipsometry and Polarized Light, North Holland, Amsterdam, p. 287. 6 Lee, J., Koh, T., and Collins, R.W. (2000) Opt. Lett., 25, 1573–1575. 7 Röseler, A. (1993) Thin Solid Films, 234, 307–313. 8 Canillas, A., Pascual, E., and Drevillon, B. (1993) Rev. Sci. Instrum., 64, 2153–2159. 9 Leng, J., Opsal, J., Chu, H., Senko, M., and Aspnes, D.E. (1998) Thin Solid Films, 313–314, 132. 10 Jellison, G.E. Jr and Modine, F.A. (1996) Appl. Phys. Lett., 69, 371.
11 Johs, B., Woollam, J.A., Herzinger, C.M., Hilfiker, J., Synowicki, R., and Bungay, C.L. (1999) Crit. Rev. Opt. Sci. Technol., CR72, 29–58. 12 Hinrichs, K., Gensch, M., and Esser, N. (2005) Appl. Spectrosc., 59, 272A–282A. 13 Röseler, A. and Korte, E.H. (2002) Infrared spectroscopic ellipsometry, in Handbook of Vibrational Spectroscopy, vol. 2 (eds P.R. Griffiths and J. Chalmers), John Wiley & Sons, Ltd, Chichester, UK. 14 Röseler, A. (2005) Handbook of Ellipsometry (H.G. Tompkins and E.A. Irene), William Andrew Publishing Inc., Norwich, New York. 15 Sirenko, A.A., Bernhard, C., Golnik, A., Clark, A.M., Hao, J.H., Si, W.D., and Xi, X.X. (2000) Nature, 404, 373–376. 16 Bernhard, C., Boris, A.V., Kovaleva, N.N., Khaliullin, G., Pimenov, A.V., Yu, L., Chen, D.P., Lin, C.T., and Keimer, B. (2004) Phys. Rev. Lett., 93, 167003. 17 Hinrichs, K. and Eichhorn, K.-J. (2007) Spectrosc. Eur., 19, 11–14. 18 Mikhaylova, Y., Ionov, L., Rappich, J., Gensch, M., Esser, N., Minko, S., Eichhorn, K.-J., Stamm, M., and Hinrichs, K. (2007) Anal. Chem., 79, 7676–7682. 19 Houbenov, N., Minko, S., and Stamm, M. (2003) Macromolecules, 36, 5897–5901. 20 Hinrichs, K., Silaghi, S.D., Cobet, C., Esser, N., and Zahn, D.R.T. (2005) Phys. Status Solidi B, 242, 2681–2697.
405
407
26 Sum Frequency Generation (SFG) Spectroscopy Günther Rupprechter and Athula Bandara
Infrared (IR)-visible sum frequency generation (SFG) laser spectroscopy is an interface-specific nonlinear optical method that provides vibrational spectra of molecules located at surfaces or interfaces. Provided that the interface is accessible to light, SFG allows the identification of chemical species, molecular densities, and molecular orientations, with sub-picosecond time resolution. As a result of its selection rules, SFG is typically not sensitive to bulk or isotropic phases, which makes the method highly interface-specific. SFG can thus be utilized to investigate buried interfaces in gaseous or liquid environments.
26.1 Introduction to SFG Spectroscopy
The ability to obtain detailed knowledge of the interaction of molecules with surfaces and interfaces has been of major importance for many areas of chemistry, physics, materials sciences, and biology. Typical examples include heterogeneous-, environmental- and photo-catalysis, sensors, microelectronic devices, fuel cells, corrosion protection coatings, self-cleaning/antibacterial surfaces, and membranes. Any study of surface/interface processes relies critically on the availability and quality of surface/interface-specific characterization methods. The number of molecules at an interface between two phases is clearly much smaller than that of the molecules in the bulk phases. Consider, for example, a solid surface (1 cm2) in contact with a gas (1 cm3; ambient conditions) or liquid (1 cm3); in this case, only about 1015 molecules will be adsorbed at the interface whereas the gas and liquid phases will comprise ∼1019 and ∼1023 molecules, respectively. Most spectroscopic methods will, therefore, yield bulk (volume) spectra of the involved media. A surface/interface-sensitive analysis method should therefore:
• •
Exhibit a detection limit well below 1015 molecules cm−2 (i.e., submonolayer sensitivity). In the most favorable case, exhibit inherent surface or interface specificity, so that bulk contributions to the spectra are avoided.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
408
26 Sum Frequency Generation (SFG) Spectroscopy
•
Exhibit compatibility with various sample media/environments (e.g., ambient gas pressure, liquid phase, high temperature).
Surface analytical techniques based on the scattering of electrons, atoms, or ions clearly satisfy the first two of these requirements [1, 2]. Such methods include lowenergy electron diffraction (LEED), reflection high-energy electron diffraction (RHEED), and He atom scattering (HAS), which provide information on surface structure/crystallography, as well as Auger electron spectroscopy (AES), X-ray photoemission spectroscopy (XPS), and secondary ion mass spectrometry (SIMS), which provide information on the chemical composition of surfaces. Highresolution electron energy-loss spectroscopy (HREELS) is used to probe the vibrational properties of surfaces and of the molecules adsorbed onto them. However, whilst the small mean free path of the electrons, atoms, and ions makes these methods surface-sensitive it also requires an (ultra)high vacuum (UHV) environment which will make such methods incompatible with gas pressures above 10−3 mbar, or with liquid phase. Hence, the above-mentioned techniques can typically not be operated under realistic conditions of technical processes (ambient or higher pressure, solvents, etc). In contrast, the large mean free path of photon-in–photonout techniques, such as IR, Raman or X-ray diffraction (XRD), makes them compatible with various environments. However, these methods are not surface-specific; that is, it is necessary to cope with strong contributions of the bulk phase (these can be minimized, perhaps by using grazing incidence angles, but not fully removed). Second-order nonlinear optical processes combine the benefits of surfacespecificity and compatibility with various sample environments. Both, second harmonic generation (SHG) and SFG are insensitive to gaseous or liquid bulk phases, such that only molecules at a surface or interface are detected (the underlying physical principles are described in Section 26.2). Consequently, SHG and SFG are inherently interface-specific and can probe any “buried” interface that is accessible by light. SHG, converting for example, 1064 nm to 532 nm light, is typically used to probe the kinetics and dynamics of adsorption, desorption, diffusion, surface melting, and phase transitions [3–5]. By tuning the laser wavelength, SHG provides information on surface electronic states [3, 4, 6–9]. However, SHG has no molecular selectivity [5] and so cannot be used for identifying surface species. Hence, in the following sections attention will be focused on the spectroscopic variant, SFG. SFG allows the vibrational spectra of adsorbed and reacting molecules at surfaces/interfaces to be obtained, whereas bulk phase molecules do not contribute to the SFG vibrational spectrum. The applications of SFG include solid–gas, solid–liquid, liquid–gas and liquid–liquid interfaces [10–31]. As a laser-based method, SFG allows advantage to be taken of the light polarization to carry out an orientation analysis of molecules [32–36]. Furthermore, pulsed lasers have enabled the performance of time-resolved measurements to provide access to the dynamics of surface and interface processes at sub-picosecond time scales [37–46]. Before turning to SFG theory, instrumentation, and modes of operation, it would be pertinent to discuss a simple example of SFG spectroscopy, such as the
26.1 Introduction to SFG Spectroscopy
Figure 26.1 (a) IRAS spectrum of 50 mbar CO on Pd(111) at 300 K. Signals of adsorbed CO around 2100 cm−1 are obscured by the rotational absorption spectrum of the CO
gas phase; (b) SFG spectra of CO on Pd(111) with 100 mbar at 300 K and 1 bar at 190 K. Adapted in part, with permission, from Ref. [20 ]; © 2007, Elsevier.
adsorption of carbon monoxide onto metal surfaces. Carbon monoxide is involved in many large-scale catalytic processes (e.g., methanol synthesis and methanation, Fischer–Tropsch synthesis, water-gas shift, CO oxidation for pollution control) [1, 47, 48], which explains why this system has attracted such wide attention in the past. CO is also frequently used for dispersion measurements of heterogeneous catalysts – that is, using CO as a probe molecule to “titrate” the number and nature of catalytic atoms/sites [49]. An infrared reflection absorption spectrum (IRAS) of a Pd(111) single crystal surface in contact with 50 mbar CO at 300 K is shown in Figure 26.1a. The IR spectrum contains contributions of CO molecules in the gas phase, as well as of surface-adsorbed CO. However, the signal of surface-adsorbed terminal (on-top) CO at ∼2060 cm−1 is almost obscured by the rotational absorption spectrum of the CO gas phase. For higher gas pressure (higher CO coverage), other metals (e.g., Pt, Rh), or for metal nanoparticles, the on-top species is typically characterized by wavenumbers around 2100 cm−1, and such species are fully obscured by the gasphase signal. The main benefit of SFG spectroscopy is that it allows selective monitoring of the adsorbed surface species, despite the presence of a gas phase (which is not SFG-active). Figure 26.1b (lower trace) shows an SFG spectrum of 100 mbar CO on a Pd(111) surface, displaying only the surface-adsorbed CO molecules, bonded to single atoms (on-top) and threefold hollow sites of the Pd surface (the shift is due to a higher CO coverage). Note that there is no gas-phase contribution to the vibrational SFG spectrum (the same is true for the spectrum at 1 bar CO; details of this are discussed below). It is also important to note that the mechanism of a molecule–surface interaction is clearly governed by the exact atomic and
409
410
26 Sum Frequency Generation (SFG) Spectroscopy
electronic structure of a solid surface. Surface atomic structures may deviate from simple truncations of the well-known bulk solid structures (surface relaxation and reconstruction) [1, 50], or they may change depending on the gaseous/liquid environment, with pronounced effects on the physico-chemical properties of the surfaces/interfaces. This highlights the importance of in situ surface analysis under “real world” conditions. In the following sections, first the theory and technical aspects of SFG spectroscopy are described, illustrating its versatility as a unique technique for surface and interface analysis. A number of selected case studies of solid–gas, solid–liquid, liquid–gas, and liquid–liquid interfaces are then presented. The focus is, however, on solid metal and oxide surfaces (in the form of single crystals, thin films, supported nanoparticles) relevant to heterogeneous catalysis. Other sections are devoted to electrochemistry and the “life sciences”; that is, polymer and biological surfaces/interfaces in gaseous and liquid environment.
26.2 SFG Theory
SFG spectroscopy utilizes the second-order nonlinear optical process of SFG; that is, two light waves at different frequencies interact in a medium characterized by a second-order nonlinear susceptibility tensor χ(2), and generate a wave at the sum of their frequencies (ωSFG = ω1 + ω2) [51]. Because this nonlinear process typically produces only a small signal, high incident light intensities – that is, pulsed lasers – are required. In order to acquire an SFG vibrational spectrum of molecules adsorbed onto a solid surface, two (e.g., picosecond) laser pulses are spatially and temporally overlapped on the sample (Figure 26.2a). One input beam is in the visible range at fixed frequency (ωvis), and the second beam is tunable in the mid-IR region (ωIR) to probe the vibrational modes of the surface species. In a simplified picture, when the IR beam is tuned through a vibrational resonance of the adsorbate, it induces a vibrational transition from the ground state to an excited state. Simultaneously, the visible beam induces a transition to a higher-energy virtual state through an anti-Stokes Raman process (cf. the energy scheme in Figure 26.2b). When the high-energy virtual state relaxes, light is generated at a frequency that is the sum of the frequencies of two incident optical fields (ωSFG = ωIR + ωvis), resulting in a signal in the visible region. By tuning the wavelength of the IR beam and monitoring the intensity of the SFG output, an adsorbate vibrational spectrum is obtained as a plot of the SFG intensity against the IR wavenumber (see Figure 26.1b). As mentioned, the power of SFG lies in its surface-specificity. According to the selection rules discussed below, only vibrational modes that simultaneously satisfy both IR and Raman selection rules are SFG-active. Therefore, SFG is not allowed in media with inversion symmetry (such as in the centrosymmetric bulk of a solid or in an isotropic gas or liquid phase), but has a finite value only at an interface where the inversion symmetry is broken. The underlying physical principles are discussed in the following.
26.2 SFG Theory
Figure 26.2 Schematic illustration of vibrational IR-visible sum frequency generation (SFG). In case of a vibrational resonance of an adsorbed molecule, visible
light is generated at a frequency that is the sum of the frequencies of two incident optical fields (ωSFG = ωIR + ωvis); (b) Schematic energy level diagram.
The interaction between light and matter induces a polarization, p. In the electric–dipole approximation, the microscopic polarization p and the electric field E are related through molecular (hyper)polarizability tensors α, β, and γ (Equation 26.1) [51–53]. p = αE + βEE + γ EEE + … .
(26.1)
The polarization produced by conventional light sources scales linearly with the electric field E (coefficient α; polarizability), whereas irradiation with intense laser light induces significant nonlinear polarization, characterized by the higher-order hyperpolarizability tensors β and γ. However, macroscopic amounts of substances are typically investigated (involving large numbers of molecules). The macroscopic polarizability tensors χ(n) are obtained by multiplying the ensemble/orientation averages (denoted “< >”) of the corresponding molecular (hyper)polarizabilities with the number of molecules N (Equation 26.2)
χ (1) = N <α >, χ (2) = N <β >, χ (3) = N <γ >
(26.2)
and the macroscopic polarization P can thus be written as P = χ (1)E + χ (2)EE + χ (3)EEE + … .
(26.3)
The magnitude of the local electric field E is described as E(ω ; r , t ) = E 0 (ω ){exp{i(ωt − kr )} + c.c.} = 2E 0 (ω )cos(ωt − kr )
(26.4)
where r is the position vector, t is the time, E0(ω) is the amplitude vector, k is the wave vector, and c.c. represents the complex conjugate. For SFG spectroscopy, the second-order polarization is most important, because E is (still) too small for the first-order term to become relevant, whereas third- and higher-order terms are almost negligible due to small hyperpolarizabilities. In the following, attention will be focused on the second-order (nonlinear) polarization P(2)
411
412
26 Sum Frequency Generation (SFG) Spectroscopy
P (2) = χ (2)EE
(26.5)
The electric field at a position r, irradiated by two light waves with different frequencies ω1 and ω2, can be expressed by the vector sum of the two electric fields: E(r , t ) = E(w 1 ; r , t ) + E(w 2; r , t )
(26.6)
Substituting Equations 26.6 for E in Equations 26.5 yields a second-order polarization P(2), comprising polarizations oscillating at ω1 + ω2 (SFG) and ω1 − ω2 (difference frequency generation, DFG), in addition to polarizations oscillating at zero frequency (optical rectification), 2ω1 (SHG of ω1), and 2ω2 (SHG of ω2). For the specific case of SFG by IR and visible light (IR-visible SFG), the nonlinear polarization P(2) is P (2)(ωSFG = ω IR + ω vis ) = χ (2)E(ω IR )E(ω vis )
(26.7)
As shown below, when the frequency/energy of the IR beam is in resonance with a vibrational transition of a surface/interface species, the SFG intensity is resonantly enhanced giving rise to a signal. IR-vis SFG has thus become a versatile tool for vibrational spectroscopy of molecules located at surfaces and interfaces, utilizing frequency-tunable or broadband IR and visible lasers [13, 15, 17–20, 22]. However, as mentioned previously, not all vibrational modes are SFG active (see below). 26.2.1 SFG Signal Intensity and Lineshape
The intensity of the generated sum frequency output is proportional to the square of P(2) – that is, to the absolute square of the effective second-order susceptibility (2 ) of the surface/interface, denoted χ s , and to the intensities of the two incident IR and visible beams. ISFG (ωIR ) ∝| χ (s2)(ω IR )|2 Ι (ω IR )Ι (ω vis )
(26.8)
Whereas I(ωIR) and I(ωvis) “simply” depend on the sources (lasers) of the IR and (2 ) visible radiation, the surface/interface properties are characterized by χ s , which has two components. Let us start with the resonant nonlinear susceptibility χ R(2) originating from the vibrational resonances:
χ R(2) (ω IR ) =
∑ω q
AR(q ) IR − ω q + iΓ q
(26.9)
where AR(q), ωq, Γq, and ωIR are the resonant amplitude, resonance frequency, and damping constant (homogeneous linewidth 2Γq = FWHM) of the qth vibrationally resonant mode, and the IR laser frequency, respectively. (2 ) The term χ R incorporates the resonance condition (ωIR – ωq), and as the IR beam is tuned through vibrational resonances of surface species, χ R(2) and thus ISFG reach a maximum. The amplitude of the vibrationally resonant susceptibility AR(q) is determined by the adsorbate concentration (number density N) and the product
26.2 SFG Theory
of the IR and Raman transition moments of the vibration (Tq , Mq; δρ is the population difference between the vibrational ground and excited state). AR(q ) ∝ NTqMqδr
(26.10)
Equation 26.10 represents the selection rule for the SFG process. In order to generate a sum frequency emission, the excited vibrational mode must be both IR- and Raman-active. Therefore, SFG is not allowed in media with inversion symmetry (due to the “Rule of mutual exclusion”). SFG, or any other second-order nonlinear optical process, therefore does not occur in (isotropic) gases, liquids, or centrosymmetric solids. The SFG signal is hence dominated by the vibrational modes of the molecules at the surface/interface where the inversion symmetry is broken, making SFG inherently surface/interface-specific. Nevertheless, as with most other spectroscopic methods, SFG must cope with a signal background. In addition to the resonantly-enhanced nonlinear susceptibil(2 ) ity χ R (related to surface vibrations), the surface/interface itself generates a (2 ) vibrational-nonresonant SFG signal, characterized by χ NR . A nonresonant SFG signal is for example, produced when a surface electronic state is excited by either the visible or SFG light [54]. However, in many cases the visible or SFG light frequencies are far from the resonances of the surface, and the nonresonant response of the surface can thus be modeled by a frequency-independent nonreso(2 ) nant susceptibility χ NR (cf. Equation 26.11; assuming that this also incorporates nonresonant contributions of higher-order components). Consequently, χ s(2)(ωIR ) can be expressed as (2 ) χ s(2)(ω IR ) = χ R(2)(ω IR ) + χ NR =
∑ω q
AR ( q ) + ANRe iφ ω q + iΓ q
IR −
(26.11)
where ANR is the amplitude of the vibrationally nonresonant susceptibility, and ϕ represents its phase relative to the resonant term. Interference between the resonant and the nonresonant parts governs the SFG (2 ) lineshape [54]. In case χ NR << χ R(2), nearly symmetric lineshapes are obtained, with an increase of the total SFG intensity at the resonance frequency (cf. SFG spectrum of on-top and bridge-bonded CO on Pd(111) in Figure 26.1b; ANR amounts to only ∼5% of the resonant CO amplitude). However, as apparent from Equation 26.11, (2 ) when χ NR ≈ χ R(2), the lineshape of an SFG-resonance is primarily determined by the phase-difference ϕ between the resonant and nonresonant signal. The phasedifference is a result of the different physical origin of vibrational-resonant and (2 ) (Equation nonresonant SFG fields/signals. Depending on ϕ, the sum of χ R(2) + χ NR 26.11) may change sign when ωIR “moves across” ωq, which may give rise to an asymmetry or even a signal decrease at the resonance frequency. 26.2.2 Determining the Number Density of Molecules from SFG Signal Intensity
Experimental SFG spectra are fitted according to Equations 26.8 and 26.11 which allow the accurate extraction of resonance positions, amplitudes, and linewidths
413
414
26 Sum Frequency Generation (SFG) Spectroscopy
[54]. Since AR(q) is proportional to the adsorbate concentration (number density N) the molecular density can be estimated, assuming that the IR and Raman transition moments of the vibration do not change. However, the latter two terms are usually different for different adsorbate species (e.g., bridging versus terminally bonded CO [55]), and may even depend on coverage [36, 54]. Therefore, the SFG signal cannot be easily correlated with a molecular concentration and a direct quantitative analysis is typically difficult. Nevertheless, attempts to place SFG on a quantitative basis are increasing [54, 56–59]. In any case, given that the nonresonant background is small, it may be useful to estimate the molecular concentration N from the SFG intensity. According to Equations 26.8–26.10, ISFG is proportional to N2. Vice versa, N is proportional to ISFG , and an estimate can be obtained by multiplying the square-root of the peak intensity with the HWHM, Γq. N ∝ ISFG ⋅ G q = ISFG ⋅ G q2
(26.12)
It should be noted that the product of ISFG and the band width 2Γq, which approximates the SFG peak area, is not proportional to N.
26.3 SFG Instrumentation and Operation Modes
The main components of an SFG spectrometer include:
• • •
A laser system to generate the incident visible and IR beams (ps or fs pulses) A sample preparation/manipulation stage (preferentially a spectroscopy cell with controlled environment). A detection system for the generated SF signal.
Most frequently, neodymium–yttrium-aluminum-garnet (Nd : YAG), and titanium– sapphire (Ti : Sa) lasers have been used [16, 34, 44, 60–62]. The operational scheme of a typical laboratory-type “scanning” SFG spectrometer is shown in Figure 26.3. The output of a Nd : YAG picosecond laser (1064 nm, ∼30 mJ per pulse, ∼25 ps, 10–50 Hz) is partly converted to 532 nm light by SHG using for example, a KDP (potassium dihydrogen phosphate) crystal. About 200 μJ per pulse of 532 nm light is used as a visible (green) beam in the SFG experiment; the 1064 nm and remaining 532 nm beams are mixed in an optical parametric generator/amplifier (OPG/ OPA) to generate tunable IR light (3−6 μm, ca. 200 μJ per pulse, resolution ∼3 cm−1) in a difference frequency generation (DFG) stage [34]. A delay line allows optimization of the temporal overlap of IR and visible pulses, and there are also devices [half-plates, polarizers (Glan–Taylor prisms)] to adjust the intensity and the polarization of the visible and IR beams. The use of different DFG crystals (e.g., AgGaS2, AgGaSe2) or even synchrotron radiation [63] allows spectroscopy to be carried out over a range of ∼1000 to 4000 cm−1. In order to differentiate the (weak) SFG signal from the (strong) reflected visible beam, a dedicated detection system is required which combines spatial, spectral,
26.3 SFG Instrumentation and Operation Modes
Figure 26.3 Schematic illustration of an SFG
spectrometer based on a Nd : YAG picosecond laser system, together with a crosssection of an SFG-compatible UHV-high-pressure reaction cell. The cells is coupled to a ultrahigh vacuum (UHV) sample preparation and analysis system (not
shown), allowing SFG spectroscopy to be performed in the pressure range from 10−13 to 1 bar. A single crystal is mounted onto the sample holder, which is inserted into three differentially pumped spring-loaded Teflon seals. Adapted in part, with permission, from Ref. [61]; © 2001, Springer.
and temporal filtering [29]. The IR and visible beams have different incidence angles, ∼55 ° and 50° respectively, producing an SFG signal that lies in between them and that can be (spatially) separated by an aperture (cf. Figure 26.2; in Figure 26.3 the reflected visible and IR beams are not shown for clarity). The SFG signal is then further (spectrally) filtered by an edge filter and a monochromator (both remove the remaining 532 nm light and allow only the SFG light to pass) before reaching a photomultiplier, the signal of which is directed to a gated boxcar integrator (triggered by the laser for temporal filtering) and sent to a PC via an A/D interface. SFG spectroscopic measurements can be carried out in different modes, such as (conventional) scanning SFG, broadband SFG, time-resolved (pump-probe) SFG, or polarization-dependent SFG. The schemes shown in Figure 26.4 illustrate each of these modes. In the case of scanning SFG (Figure 26.4a), the IR energy is tuned stepwise over the range of interest (which makes this mode rather slow). To speed-up the spectral acquisition, “broadband” SFG takes advantage of
415
416
26 Sum Frequency Generation (SFG) Spectroscopy
Figure 26.4 Scheme of different modes of operation of SFG. (a) Scanning SFG; (b) Broad-
band SFG; (c) Pump-probe SFG; (d) Polarization-dependent SFG.
ultrashort – and thus spectrally broad– IR laser pulses (e.g., 150 fs; width ∼150 cm−1) (Figure 26.4b), covering an IR region of interest that may be centered around C–O or C–H stretching vibrations ( [20, 41, 44, 46] and references therein). The broadband IR pulse is overlapped with a narrowband visible pulse (e.g., 7 ps; width 2 cm−1), but only that part of the IR spectrum which is in resonance with a vibrational transition will be upconverted to generate a sum frequency signal. The broadband approach thus allows the capture of an entire SFG vibrational spectrum in a few single laser shots, without tuning of the IR wavelength (the SFG spectrum is dispersed and recorded using a CCD-camera). For time-resolved “pump-probe” SFG spectroscopy (Figure 26.4c), the surface species are first excited by, for example, an intense near-IR (ps or fs) laser pulse (“pump”; that sets the time to “zero”), followed by a time-delayed weak SFG (IR + vis) “probe” that monitors the changes in the vibrational properties of the adsorbate–substrate complex. By varying the delay time between the pump and probe, the effect of the excitation can be examined. For example, excitation (pump) may allow the detection of surface reaction intermediates that are too short-lived for steady-state scanning SFG (see Section 26.4.1.5 below for details). However, by taking “snapshots” of the transient vibrational spectrum at different delay times, the lifetime of transient species can be determined. A prerequisite for such measurements is, of course, that the surface species relax in between the pump– probe pulses (e.g., for measurements at 1 kHz, the adsorbate–substrate system must relax within 1 ms). The use of lasers prompts advantage to be taken of light polarization. SFG is frequently carried out in ppp geometry (i.e., by detecting a p (parallel)-polarized SFG signal produced by a p-polarized visible and a p-polarized IR beam). This combination is typically used for adsorption/reaction investigations of metal surfaces, because it produces the most intense adsorbate SFG signal [14, 36]. However, other polarization combinations can also be used, for example, ssp (s-polarized
26.4 Applications of SFG Spectroscopy and Selected Case Studies
SFG, s-polarized visible and p-polarized IR; Figure 26.4d). For metals, the IR beam is always p-polarized, because the light field of an s-polarized IR beam is screened by the conduction electrons of metal surfaces. Comparison of the signal intensities for different polarization combinations (e.g., Ippp versus Issp) can then be utilized to obtain information about the molecular orientation of surface species [34, 36, 64–66]. Although, until now only SFG spectroscopy in reflection mode – which is frequently utilized – has been discussed, other geometries can also be used, such as transmission and total internal reflection using prisms (see Figure 26.9a) [18, 67–69]. Applications of SFG microscopy have also been reported, imaging the spatial distribution of selected SFG vibrational bands, thus combining the analytical ability of IR with the spatial resolution of visible light microscopy [70–74].
26.4 Applications of SFG Spectroscopy and Selected Case Studies
Applications of SFG include solid–gas, solid–liquid, solid–solid, liquid–gas (air), and liquid–liquid interfaces, with each combination having been reviewed in detail [12, 13, 17, 18, 20, 22, 25, 30]. In this context, the expression “surface” is used when molecules are adsorbed onto solids in an ultra-high vacuum (UHV) environment, whereas “interface” is rather used for solids in contact with a gas phase or liquid phase. In the following subsections, selected case studies illustrating the versatility of SFG are discussed. 26.4.1 SFG Spectroscopy on Solid Surfaces and Solid–Gas Interfaces
Early applications of SFG on solid surfaces were dedicated to examining the adsorption of small molecules in a UHV environment ( [10, 20, 75] and references therein). Shortly afterwards, time-resolved and polarization-dependent spectroscopies were carried out [39, 66], and the surface-specificity of SFG was utilized to monitor surface-adsorbed molecules at elevated (ambient) gas pressure, or during catalytic reactions [14, 15]. 26.4.1.1 SFG Spectroscopy under UHV Conditions The first surface science studies of SFG examined the adsorption of small molecules such as CO, NO, or C2H4 on metal single-crystal surfaces [15, 76–81]. The SFG vibrational spectra provided an insight into the adsorption configuration and adsorption sites, for example the relative population of hollow/bridge/on-top bonded CO or di-σ/π-bonded ethylene. Both, coverage-dependent and temperaturedependent measurements yielded information on adsorbate–adsorbate interactions and adsorbate binding energies. As an example, Figure 26.5a shows the SFG spectra of CO adsorption on a Pt(111) single crystal surface [81, 82]. For CO on Pt(111), vibrational features were
417
418
26 Sum Frequency Generation (SFG) Spectroscopy
Figure 26.5 Polarization-dependent SFG
spectra (ppp and ssp) of CO (a) and NO (b) adsorbed on Pt(111) at 10−6 mbar and 300 K. The resulting CO and NO adsorption geometries are displayed schematically. For details, see text. Adapted in part, with permission, from Ref. [81]; © 2000, Oldenburg Wissenschaftsverlag;
(c and d) Corresponding ppp and ssp SFG spectra of CO (c) and NO (d) adsorbed onto NiO(111)/Ni(111) at 140 K and 10−7 mbar. Models of CO and NO adsorbed onto the reconstructed inclined NiO(111) microfacets are also shown. Adapted in part, with permission, from Ref. [34]; © 1997, Elsevier.
observed at 2090 cm−1 and ∼1830 cm−1, and assigned to terminally bound (on-top) CO and bridge-bonded CO, respectively. The assignment was based on comparison with IR spectra, and similar SFG spectra were acquired for CO on Pd(111), Rh(111), Ni(111), and Ru(110) [15, 34, 37, 41, 55, 83–85]. The SFG intensity of multiple-coordinated (bridge, hollow) CO is typically smaller than that of on-top
26.4 Applications of SFG Spectroscopy and Selected Case Studies
CO, as observed for several noble metal surfaces. This was attributed not only to lower Raman cross-sections but also to screening and intermolecular coupling effects, specifically affecting SFG [34, 37, 54, 59, 65, 81, 83]. 26.4.1.2 Polarization-Dependent SFG Spectroscopy The upper SFG spectrum in Figure 26.5a (as with all spectra discussed to this point) were acquired in so-called ppp polarization combination; that is, all beams (SFG, visible, IR) were p-polarized. Whilst this typically yields the strongest signal, the additional use of other polarization combinations allows conclusions to be drawn on the molecular orientation of the adsorbed molecules [36]. For metal surfaces, the IR laser beam must be always p-polarized, because the surface electric field of an s-polarized IR laser beam is screened by the conduction electrons of the metal. In contrast, the visible beam can be either s- or p-polarized, because the surface electric field is less efficiently screened by the conduction electrons due to the lower dielectric constants of metals in the visible region. The polarization of the SFG output can be either s- or p-polarized, leading to the ppp and ssp polarization combinations (ssp refers to a s-polarized SFG, s-polarized visible and ppolarized IR; psp and spp do not yield a signal on metals [36]). The polarization combinations can be selected by adjusting suitable polarizers of the spectrometer (cf. Figure 26.3). The ppp and ssp configurations are illustrated in Figure 26.4d. Very briefly (and also somewhat simplified), it can be stated that an SFG signal is generated only when the electric fields of the visible and IR lights have a component parallel to the bond axis. Consequently, ppp spectra detect molecules with molecular axes parallel or inclined to the surface normal, whereas ssp spectra are mainly sensitive to tilted molecules. Consequently, by comparing the signal intensities Ippp and Issp of ppp and ssp spectra, respectively, it is possible to deduce the molecular orientation (tilt angles) of molecules (flat-lying molecules with molecular axes parallel to the surface – that is, perpendicular to the surface normal – cannot be detected due to the metal surface selection rule [86]). The analysis of polarization-dependent SFG spectra is quite complex, and further details are available elsewhere [36, 64, 81, 87]. A comparison of ppp and ssp spectra for CO and NO adsorbed onto Pt(111) clearly illustrates the different binding geometries of the molecules on the surface (Figures 26.5a and b). For CO on Pt(111) in Figure 26.5a, the ppp spectra showed a vibrational feature of on-top CO (2090 cm−1) and bridge CO (∼1830 cm−1). In the ssp polarization configuration neither peak was observed, which indicated that the C–O molecular axis of both CO species was oriented parallel to the surface normal (in other words, CO was adsorbed perpendicularly onto the surface). For CO on Pd(111) [36] and Ni(111) [34], an upright orientation of on-top and bridged CO was also reported. The polarization-dependent spectra of NO adsorption on Pt(111), as reported by Metka et al. [81], are shown in Figure 26.5b. The stretching mode of NO is detected both in the ppp and ssp spectra, indicating that the NO had been adsorbed in a tilted geometry (at high coverage). In fact, by analyzing the intensities of the ppp and ssp signals, a tilt angle of about 20° was determined, in good agreement with results obtained using NEXAFS [88].
419
420
26 Sum Frequency Generation (SFG) Spectroscopy
Bandara et al. [34, 37, 39] have utilized the polarization dependence of the SFG intensity to determine the orientation of CO and NO adsorbed on a thin (∼2 nm) NiO(111) film grown on Ni(111) [34]. The SFG spectra of CO and NO, respectively, adsorbed at 130 K (background pressure 1 × 10−7 mbar) for both polarization combinations are shown in Figure 26.5c and d. For CO, a vibrational band appeared at 2144 cm−1, both for ppp and ssp polarization combination. This indicates – in contrast to the corresponding measurements on metallic surfaces – that CO adopts a tilted geometry on NiO(111). In the case of NO on NiO(111), the N–O stretching band at 1800 cm−1 was observed only in the ppp spectra, which indicated that the N–O bond must be aligned parallel to the surface normal, again in contrast to measurements on metallic surfaces. The results for NiO can be better understood by considering that the NiO(111) surface is unstable upon heating, and restructures into a (2 × 2) structure with trigonal NiO(100) microfacets [89] (see models in Figure 26.5c and d). When CO is adsorbed perpendicularly on the microfacets (on Ni2+), it is tilted with respect to the underlying Ni(111) surface and thus appears in the ssp spectrum. On the other hand, NO is adsorbed tilted on the inclined facets, such that the overall orientation of the N–O bond axis is close to parallel to the surface normal of the Ni(111) substrate. As mentioned previously, the analysis of polarization-dependent SFG spectra is complex [64, 81, 90]; further details of the CO/Pd(111) system are available elsewhere [36, 91]. Polarization-dependent SFG spectra on solid surfaces were also reported for larger entities, such as isotopically (deuterium) labeled formic acid DCOOD [38, 39, 66]. For DCOOD on Ni(111), a band at 2196 cm−1, characteristic of the C–D bond of bridging bidentate formate, appeared in ppp spectra only and was absent in ssp; this pointed to a C–D bond orientation parallel to the surface normal (cf. Figure 26.7b). 26.4.1.3 SFG Spectroscopy under Near-Atmospheric Gas Pressure Solid–gas interactions are key processes in heterogeneous catalysis. Traditionally, metal–adsorbate interactions at solid surfaces have been studied using surfacesensitive techniques that require a UHV environment [1, 92, 93]. Unfortunately, this led to the “pressure gap” controversy – that is, whether results obtained under UHV are relevant to “real” catalysis at ambient (or even higher) gas pressures. Consequently, by applying SFG to ambient pressure systems, investigations into catalytic reactions conducted at transition metal single crystals and supported metal nanoparticles helped to bridge this gap [14, 15, 20, 77–79, 94, 95]. Before turning attention to SFG spectroscopy during catalytic reactions, it would be of interest first to discuss CO adsorption at mbar gas pressures. The major interest in CO adsorption stems from the use of CO as a probe molecule to titrate surface sites and determine metal dispersion, from the involvement of CO in many catalytic reactions, and from its relevance to catalyst poisoning. The use of surface-specific SFG allowed previous CO adsorption studies conducted under UHV to be extended to ambient pressure [55, 96, 97]. The adsorption of CO onto Pt(111) and Pd(111) were investigated by several research groups, all of whom focused on the pressure- and temperature-dependence of the CO adsorbate layer
26.4 Applications of SFG Spectroscopy and Selected Case Studies
coverage and structure [20, 61, 76, 95, 98–102]. As inferred from Equation 26.8, when SFG is performed at a high gas pressure, a careful normalization of the SFG signal intensity by the effective IR surface intensity is required to compensate for IR gas-phase absorption [20, 61]. Figure 26.6a shows a “high-pressure” SFG spectrum for CO adsorption on Pd(111) [55, 97]. The same adsorption configurations as those known from UHV studies were observed to exhibit hollow, bridge, and on-top CO (for a detailed description of the various CO adsorbate layers, see Ref. [20]). Although “highpressure species” were absent, differences were still apparent between the UHV and ambient pressure studies, although these were rather related to differences in the CO surface coverage and, thus, in the relative population of the various sites. For example, at low CO coverage the hollow sites were preferred, whereas at higher coverage the bridge and on-top sites were also populated. Figure 26.6b shows the pressure-dependence of CO adsorption on stepped/ionbombarded Pd(111). At 10−6 mbar CO and 190 K, SFG detected a peak at ∼1990 cm−1, which originated from CO adsorption on steps/defects; this was in addition to bridging (1955 cm−1) and on-top (∼2090 cm−1) CO that was also observed on “smooth” Pd(111) (Figure 26.6a). However, at high pressure (>1 mbar), only threefold-hollow (∼1890 cm−1) and on-top (∼2100 cm−1) CO were observed on both “smooth” and “defect-rich” Pd(111), characterizing a (2 × 2) CO saturation structure (coverage 0.75 monolayer) [55, 97, 103]. Note: this does not indicate that the steps on the “defect-rich” surface were removed by high-pressure CO (because this species reappeared upon decreasing the pressure); rather, it indicates that on an extended single-crystal surface intermolecular coupling at high CO coverage no longer allows observation of the molecules adsorbed at step sites [20, 55]. 26.4.1.4 SFG Spectroscopy on Supported Metal Nanoparticles Figure 26.6c shows the SFG spectra of CO adsorbed onto Pd nanoparticles (mean diameter ∼6 nm) grown in UHV on a thin Al2O3 film (∼1 nm) on NiAl(110) [55, 96, 97, 104]. These model catalysts were intended to bridge the “materials gap” between studies on single crystals and supported nanoparticles. Scanning tunneling microscopy (STM) images of the cubo-octahedral Pd nanoparticles, and of the top facet of a single particle, are shown in the inset [105]. Based on singlecrystal measurements, IRAS spectroscopy [106] and density functional theory (DFT) [104], the SFG peaks observed for Pd nanoparticles were assigned to CO adsorbed on threefold hollow sites (1895 cm−1), bridge-bonded CO at particle edges (1990 cm−1) and particle terraces (shoulder at 1950 cm−1), and to linearly (on-top) bonded CO (2100 cm−1). The species are indicated in Figure 26.6c (see also Refs [54, 55] for a more detailed description of relative intensities and lineshapes). It appears that there are differences between the high-pressure single-crystal and nanoparticle spectra [55, 96]. In particular, the band at 1990 cm−1, which characterizes CO at steps or particle edges, was absent at high pressure on Pd single crystals, but was observed for Pd nanoparticles. As mentioned previously, intermolecular CO coupling at high coverage obscures this species on an extended single-crystal surface, whereas on Pd nanoparticles such a “masking” effect did not occur. In
421
422
26 Sum Frequency Generation (SFG) Spectroscopy
Figure 26.6 SFG spectra of CO on (a)
smooth Pd(111), (b) defect-rich (rough/ stepped) Pd(111), and (c) Pd nanoparticles deposited on thin Al2O3 film on NiAl(110), at pressures of 10−6 mbar and 1 bar at 190 K; (d) High-pressure SFG spectrum of CO : H2 mixture on Pd/Al2O3 at 500 K. The peaks characterizing the various binding geometries of CO on threefold-hollow, bridge, and on-top Pd sites are indicated. Pronounced differences occur between Pd nanoparticles and single crystals, demonstrating the need
to carry out model catalytic studies on supported nanoparticles. Adapted in part, with permission, from Refs [97, 108]; © 2002, 2005, Elsevier. The STM images of a Pd–Al2O3/NiAl(110) model catalyst illustrate the cubo-octahedral shape of the Pd particles (upper image 100 × 100 nm2), and the (111) orientation of the particle top facet (lower image 1.8 × 1.6 nm2). Adapted in part, with permission, from Ref. [105]; © 2007, Elsevier.
26.4 Applications of SFG Spectroscopy and Selected Case Studies
contrast, the edge-bonded CO species was rather enhanced by intensity transfer from CO bound at the particle facets to edge-bonded CO [106]. This strongly suggests that, whenever possible, catalytic model studies should be performed on nanoparticle model catalysts. In subsequent studies, SFG was employed to monitor the coadsorption of CO and hydrogen under UHV, and atmospheric pressure CO hydrogenation, both on Pd nanoparticles and Pd(111) (Figure 26.6d). For the high-pressure/hightemperature reaction, the relative populations of hollow and on-top sites differed from those observed under UHV, and indicated a partly disordered CO overlayer and/or surface roughening [20, 107, 108]. Because the coverage of the CO layer under reaction conditions was quite high (ca. 0.5 monolayer) – a coverage which prevents dissociative hydrogen adsorption under UHV – the disorder/roughening seems essential to allow for CO–H reaction under the conditions of technical catalysis. Pd nanoparticles supported on Al2O3/NiAl(110) and Pd(111) have also been utilized to examine methanol oxidation (partial oxidation to formaldehyde CH2O, or full oxidation to CO2) [20, 109–111]. Under the applied experimental conditions, the adsorbed CO was the only surface species observed (spectra similar to Figure 26.6c and thus not shown), whereas other reaction intermediates were too shortlived for SFG detection. Nevertheless, the in situ studies revealed a marked difference between nanoscale and macroscopic catalyst surfaces; whereas Pd nanoparticles were partially oxidized during the reaction, Pd(111) remained metallic, again pointing to the need to perform fundamental studies on realistic nanoscale models [20, 94, 111, 112]. Other catalytic reactions studied by SFG include the hydrogenation and dehydrogenation of hydrocarbons, CO oxidation, and methanol synthesis and oxidation on Pt, Pd, Rh, Ru, Co, Cu, Au surfaces (see Refs [20, 75] and references therein). 26.4.1.5 Time-Resolved (Pump-Probe) and Broadband SFG Spectroscopy Although scanning SFG spectroscopy is capable of identifying interface species and their orientation, it cannot be used to detect reaction intermediates with lifetimes that are too short for steady-state measurements. However, the application of pump-probe techniques allows time-resolved SFG spectroscopy to be performed which provides access to reaction dynamics (Figure 26.4c) [43–46, 85, 113]. By combining time-resolved and polarization-dependent SFG, Bandara et al. have obtained detailed insight into the elementary processes of formic acid decomposition on thin (∼1–2 nm) NiO(111) films grown on Ni(111) [35, 38–40, 66, 114]. Upon exposure of NiO(111)/Ni(111) to isotopically labeled formic acid (3 × 10−7 mbar DCOOD) at 400 K, surface-adsorbed formates were detected, with a C–D stretch (νCD) vibration at 2160 cm−1, characterizing (chelating) bidentate (Figure 26.7a). Increasing the temperature induced the dehydrogenation and dehydration of formate, producing CO2 and D2, and CO and D2O, respectively [115–117]. This led to a continuous decrease in the intensity of the 2160 cm−1 band, although only limited information on intermediate species was provided under steady state. In contrast, when a near-IR (10 mJ per pulse at 1064 nm) picosecond laser pulse
423
424
26 Sum Frequency Generation (SFG) Spectroscopy
Figure 26.7 (a) Time-dependent and polarization-dependent SFG spectra acquired during decomposition of deuterated formic acid (DCOOD) on NiO(111)/Ni(111) at 400 K (DCOOD; pressure of 3 × 10−7 mbar; delay times of −100, 0, and 100 ps). Upon transient thermal excitation by an IR pump (10 mJ per pulse), bidentate formate is partly converted to monodentate formate, an effect that cannot be observed by isothermal SFG spectroscopy. A comparison of ppp and ssp SFG spectra indicates a tilted orientation of
the monodentate C–D bond; (b) Decomposition of DCOOD on Ni(111) at 320 K (5 × 10−7 mbar DCOOD). Observation of the monodentate C=O stretching mode confirmed the suggested mechanism. Dashed lines represent the experimental data, whereas solid lines represent the fitted components. Molecular configurations deduced from the spectra are included. Adapted in part, with permission, from Ref. [66]; © 1998, American Chemical Society and Ref. [39]; © 1999, Springer.
(“pump”) was used to induce an instantaneous temperature “jump” of about 300 K [39, 66], the stable bidentate formate transformed to a (unstable) monodentate formate, the latter being characterized by a C–D stretch vibration at 2190 cm−1 (Figure 26.7a; delay time 0 ps). Varying the delay time between the near-IR pump and the SFG probe indicated that the effect of the temperature jump vanished within ∼125 ps; that is, the spectra for delay times >125 ps were identical to spectra
26.4 Applications of SFG Spectroscopy and Selected Case Studies
without near-IR excitation (cf. a negative delay time of −100 ps when the probe arrived before the pump). The C=O band of monodentate formate could not be observed due to a parallel orientation to the metallic Ni(111) substrate. The time-resolved measurements of formic acid decomposition (Figure 26.7a) were further combined with polarization-dependent SFG; this meant that the same type of pump-probe spectra would be acquired for the ppp and ssp configurations [38, 39]. As mentioned previously, ppp detects bonds that are both parallel with and inclined to the surface normal, whereas ssp has an enhanced sensitivity for tilted bonds. Consequently, the higher intensity of the monodentate 2190 cm−1 peak in ssp (Figure 26.7a) indicates a tilted C–D bond, and thus further supports the proposed transformation of a stable surface bidentate formate to an unstable monodentate species, preceding formic acid decomposition. The very weak 2160 cm−1 peak of bidentate in ssp indicates that the bidentate C–D bond is oriented close to the surface normal [38, 39]. Studies of formic acid decomposition on Ni(111) revealed a similar mechanism on the metallic surface [39, 40, 114]. Although detection of the C–D stretch mode of monodentate formate (∼2190 cm−1) on Ni(111) was prevented by the overlap with the stronger SFG band of bridging bidentate formate (2196 cm−1), the monodentate species was evident from the observation of its C=O stretch at 1718 cm−1 (Figure 26.7b). No SFG signals were observed in ssp. Nevertheless, the time-dependent intensity changes in ppp – that is, the reduction of the C–D stretch intensity of bridging bidentate and, in particular, the observation of monodentate formate around 0 ps delay time – clearly illustrate the bidentate–monodentate transformation. Observation of the monodentate C=O band for Ni(111), and the absence for NiO(111)/Ni(111), are due to the different molecular orientations of monodentate (C=O bond inclined and parallel to the Ni substrate, respectively; cf. Figure 26.7). Subsequent DFT studies of formic acid decomposition confirmed the proposed reaction scheme [118]. It should be noted that these experiments were carried out by scanning SFG at a repetition rate of 10 Hz (i.e., one spectrum takes ∼10–20 min to record). However, combining the pump-probe approach with broadband SFG speeds up the spectral acquisition (kHz repetition rates reduce the acquisition time to seconds/minutes) by overlapping an ultrashort spectrally broad IR laser pulse (e.g., 150 fs; width ∼150 cm−1) with a narrowband visible pulse (e.g., 7 ps; width 2 cm−1) (Figure 26.8a). Only that part of the IR spectrum that is in resonance with a molecular vibrational transition will be upconverted to generate an SFG signal. The SFG spectrum is then dispersed by a spectrograph and recorded with a CCD-camera; in this way, an entire SFG spectrum can be acquired within a few single laser shots. For timeresolved “pump-probe” SFG (Figure 26.8b), the surface species are again first excited by an intense laser pulse (“pump”; that sets the time to “zero”), followed by a time-delayed weak SFG (broadband IR + vis) probe that monitors the changes induced in the vibrational properties of the adsorbate–substrate system. Varying the delay time between the pump and probe then reveals the effect of the excitation. Further details of this method are available elsewhere [20, 43–46, 85, 113, 119–122]; an example is discussed here for purposes of illustration (Figure 26.8b).
425
426
26 Sum Frequency Generation (SFG) Spectroscopy
Figure 26.8 (a) Broadband IR-visible SFG
spectroscopy. Spectrum of an IR femtosecond laser pulse and resonant SFG signal of CO adsorbed on Ru(0001); (b) Time-resolved SFG spectroscopy of CO on Ru(0001) at 100 K using a “pump–probe” SFG technique. Excitation by the pump pulse leads to a
transient temperature rise which downshifts the frequency of the CO-stretch vibration (resulting from anharmonic coupling of the CO-stretch to a frustrated translational mode: see text). Adapted in part, with permission, from Refs [41, 113]; © 2000 American Physical Society.
Using broadband pump-probe SFG, Wolf, Bonn, Ertl, and coworkers [41, 113, 119] examined CO adsorbed onto Ru(0001). After photoexcitation with an intense near-IR femtosecond-laser pulse (“pump”), a time-delayed weak SFG “probe” was employed to take “snapshots” of the transient vibrational spectrum at different delay times (Figure 26.8b). The transient temperature rise at the surface due to excitation by the pump pulse led to a downshift in frequency of the CO-stretch vibration, which can be explained by the model of Persson et al. [123]: With increasing temperature, the stretching frequency of on-top CO shifts to lower vibrational frequency due to anharmonic coupling of the CO-stretch vibration to a frustrated
26.4 Applications of SFG Spectroscopy and Selected Case Studies
translational mode. An increase in the amplitude of the excited frustrated translation leads to a displacement of the CO molecule from its top-position towards a bridge site. At longer pump-probe delays, the heating effect is reduced and the CO frequency returns to the regular value. Although no time-resolved or polarization-dependent measurements have been performed under high (mbar) gas pressure to date, the geometry of adsorbed species or catalytic intermediates may well be coverage- (pressure-) and temperaturedependent, and may change during the course of a catalytic reaction. 26.4.1.6 SFG Spectroscopy on Colloidal Nanoparticles and Powder Materials The SFG studies discussed above were performed on well-defined planar surfaces, such as UHV-prepared single crystals, thin films, and supported nanoparticles. For technical industrial-grade (powder) materials, vibrational IR spectroscopy using small in situ cells minimizing gas-phase absorption has long been employed, though the application of polarization-dependent and time-resolved SFG would have additional advantages. The strong light-scattering ability of powder materials typically prevents SFG spectroscopy, although new approaches have recently been demonstrated. Wet-chemically synthesized powders or shape-controlled colloidal nanoparticles were coated onto (surface modified) single-crystal CaF2, sapphire or quartz prisms; subsequently, by using total internal reflection (TIR) geometry, much higher signals than by (standard) external reflection could be obtained (Figure 26.9a) [68, 69, 124]. Kweskin et al. [124] deposited a monolayer of as-prepared poly(vinylpyrrolidone) (PVP)-capped cubic Pt nanoparticles on a single crystal sapphire prism (Figure 26.9a). Based on transmission electron microscopy (TEM) imaging, the Pt nanoparticles had a mean size of ∼7 nm, with a number density of ∼7 × 1011
Figure 26.9 (a) Illustration of the set-up for SFG spectroscopy on colloidal Pt nanoparticles deposited on a single-crystal sapphire prism. The measurement was carried out in total internal reflection (TIR) geometry; (b) SFG spectra (ssp polarization) of CO adsorbed on PVP-capped cubic Pt nanoparticles at 295 K (flowing CO), before and after
removal of water and part of the polymer capping by oxidation in O2 at 373 K (3 h), affecting the peak position and intensity. The inset shows a transmission electron microscopy image of the Pt nanoparticle monolayer. Adapted in part, with permission, from Ref. [124]; © 2006, American Chemical Society.
427
428
26 Sum Frequency Generation (SFG) Spectroscopy
particles cm−2 (Figure 26.9b). Subsequent measurements using atomic force microscopy (AFM) indicated a film thickness of ∼11 nm, and thus a polymer coating thickness of ∼2 nm. In Figure 26.9b, the room temperature SFG spectrum of CO adsorbed on as-prepared PVP-capped cubic Pt nanoparticles is compared to the spectrum obtained after oxygen treatment at 373 K (3 h), which removed hydrogenbonded water and part of the PVP-capping. This increased the intensity of the CO peak and shifted it to a higher wavenumber (characteristic of smoother Pt surfaces). The Pt particle surface is clearly accessible to gas molecules, as corroborated by IR spectroscopy [125], and also performs CO oxidation. If the TIR approach were to be coupled with femtosecond lasers, then intriguing time-resolved, surfacespecific SFG studies on technological materials might become feasible. 26.4.2 SFG Spectroscopy on Solid–Liquid Interfaces
SFG spectroscopy has also been applied to examine solid–liquid interfaces, with special attention being directed towards electrochemistry. The electrochemical interface is affected by the applied (variable) electric field and the presence of the (bulk) liquid phase. A combination of SFG with voltammetry has been used widely to obtain in situ information on the (adsorption), oxidation, and reduction of adsorbates at the electrode–electrolyte interface [26, 30, 52, 126–146]. For example, SFG was employed to detect CO, CN− (∼2050–2120 cm−1), and SCN− anions at the surface of metal electrodes [4, 126–129, 144, 146–148]. In many cases, asymmetric SFG lineshapes were observed, which resulted from the applied potential that changes the interference of the (electronic) nonresonant term with the vibrational resonant SFG signal. Other electrochemical systems studied with SFG have included H/Pt (important for hydrogen evolution), CO/Pt, and methanol/Pt. For hydrogen adsorption onto Pt surfaces under acidic conditions, SFG was capable of differentiating between the vibrational bands of platinum dihydride structures and of (terminal) Pt–H species [149, 150]. Studies of CO/Pt under electrochemical conditions have contributed to an understanding of CO oxidation and poisoning [151]. In particular, the effect of the applied potential on adsorbate orientation under electrochemical conditions was examined [57, 152–155]. For “simple” adsorbates (such as CO and CN−) at electrochemical interfaces, pump–probe approaches have been utilized to study vibrational relaxation [133, 156]. Chou et al. [144] employed SFG to investigate potential-induced structural changes of acetonitrile adsorbed onto a Pt(111) electrode. In this case, the SFG spectra indicated that acetonitrile had responded to the electrode potential and reoriented with the C–C bond perpendicular to the electrode surface. 26.4.3 SFG Spectroscopy on Polymer and Biomaterial Interfaces
SFG has been applied to study self-assembled monolayers and polymers on metal, semiconductor, and insulator (solid) surfaces, in gaseous (air) or liquid environ-
26.4 Applications of SFG Spectroscopy and Selected Case Studies
ments [23, 157–160]. Both, IR and/or Raman vibrational spectra of polymer films of micrometer thickness have provided only bulk spectra, whereas SFG selectively detects the conformation of the hydrocarbon chains in the surface layer. Because a material’s properties (e.g., adhesion or hydrophilicity/hydrophobicity) are often governed by the surface – which may be restructured when the environment is changed – only SFG can provide the relevant information. Although such systems can be classified as solid–gas or solid–liquid interfaces (similar to the “inorganic” samples described above), a separate section is devoted to polymer and biomaterial interfaces here in order to account for their “organic nature.” Both, Zhang et al. [157, 161] and Gracias et al. [162] have investigated the surface structures of polyethylene and polypropylene; the surface glass transition temperature (Tg) of polypropylene was found equivalent to the bulk Tg [162]. When Gautam et al. [163] reported the details of studies of the structure and melting transition temperatures of alkyl-side chain acrylate polymers at air and solid interfaces. the SFG polarization-dependence suggested a structural rearrangement at the polymer–air interface, but not at the polymer–solid interface. SFG studies were also performed on polymeric biomaterials that are commonly used in bioimplants. For example, Zhang et al. [164] showed that polyurethane with poly(dimethylsiloxane) (PDMS) end groups changed its interfacial structure depending on the environment (e.g., in air and liquids). Likewise, an examination of hydrogel surfaces [165–167] indicated changes in the molecular orientation of poly(2-hydroxyethyl methacrylate) (pHEMA) in various environments. This complex situation was magnified when the adsorption of biomolecules on polymers was studied; examples included the coverage-dependent surface structures of fibrinogen, lysozyme, and bovine serum albumin (BSA) on deuterated polystyrene, partially deuterated polystyrene, and silica surfaces [168–170]. The adsorption of amino acids and small model polypeptides at interfaces, as well as molecular interactions at single planar lipid bilayers, have also been investigated (see Refs [171, 172]). 26.4.4 SFG Spectroscopy at Liquid–Gas and Liquid–Liquid Interfaces
Liquid surfaces and interfaces each play key roles in many physical, chemical, and biological processes [13]. The first application of SFG to liquid surfaces was reported for the methanol–vapor interface [173]; in this case, the SFG spectra acquired in the C–H stretching frequency range detected the CH3 groups at the liquid surface, and provided molecular level information on the interface structure. When the liquid–vapor interfaces of longer alcohols (C2–C8) were subsequently studied [174], the alcohol molecules showed substantial ordering at the liquid– vapor interface, with their O–H bonds pointing inward into the liquid and forming a hydrogen-bonded network; in contrast, the alkane chains were shown to stick out from the surface (Figure 26.10). SFG has also been used to examine O–H stretch vibrations at the water–vapor interface [175, 176]. Both, polarization-dependent and time-resolved SFG spectra showed that orientation of the O–H bond varied on a ∼1 ps time scale [151, 154,
429
430
26 Sum Frequency Generation (SFG) Spectroscopy
Figure 26.10 SFG spectroscopy at the alcohol–vapor interface [174]. The normal alcohols show substantial ordering at the liquid–vapor interface. The OH groups extend into the liquid, whereas the alkyl chains are oriented away from the liquid.
155, 177–179]. Various other liquid surfaces, including chiral liquids [27, 180, 181], aqueous solutions [182], surfactant layers, and Langmuir films [183–187], have been examined using SFG. Typically, molecular solutes and ion complexes penetrate the top monolayer of the aqueous–air interface and displace water from the interface. Polarization-dependent SFG detected a tilted geometry for interfacial ammonia–water complexes (C3 axis tilted ∼38° from normal) [182, 188]. Studies of liquid–liquid interfaces may be hampered by the absorption of IR light in the liquids, and consequently fewer studies have been reported [13, 189–192]. Example of these studies were the application of SFG to investigate oil–water and CCl4– water interfaces [191, 192].
26.5 Conclusion
Vibrational SFG spectroscopy, as an inherently interface-specific method, has contributed significantly to the present understanding of the chemistry and physics at solid–gas, solid–liquid, liquid–gas, and liquid–liquid interfaces, with implications for surface adsorption and surface reactions relevant to heterogeneous catalysis, electrochemistry, polymer science, and biocompatibility. SFG can be operated in a wide variety of environments (gaseous or liquid), specifically to detect “buried” interface species. Although, in many cases designed and well-defined nanostructured surfaces have been investigated, the first important steps of applying SFG to technological materials have been successfully mastered. The extension of conventional scanning SFG spectroscopy, with the added benefits of time-resolved (pump-probe broadband) and polarization-dependent operation modes, will surely lead to a fascinating new science of nanoscale interface materials.
References 1 Somorjai, G.A. (1994) Introduction to Surface Chemistry and Catalysis, John Wiley & Sons, Inc., New York. 2 Somorjai, G.A. and Rupprechter, G. (1998) J. Chem. Educ., 75, 161.
3 Sipe, J.E., Mizrahi, V., and Stegeman, G.I. (1987) Phys. Rev. B, 35, 9091. 4 Guyot-Sionnest, P. and Tadjeddine, A. (1990) J. Chem. Phys., 92, 734.
References 5 Corn, R.M. and Higgins, D.A. (1994) Chem. Rev., 94, 107. 6 Heinz, T.F., Chen, C.K., Ricard, D., and Shen, Y.R. (1982) Phys. Rev. Lett., 48, 478. 7 Guyot-Sionnest, P., Chen, W., and Shen, Y.R. (1986) Phys. Rev. B, 33, 8254. 8 Guyot-Sionnest, P., Chen, W., and Shen, Y.R. (1987) Phys. Rev. B, 35, 4420. 9 Guyot-Sionnest, P. and Shen, Y.R. (1988) Phys. Rev. B, 38, 7985. 10 Hall, R.B., Russell, J.N., Miragliotta, J., and Rabinowitz, P.R. (1990) Sum Frequency Generation Spectroscopy, in Chemistry and Physics of Solid Surfaces (Springer Series in Surface Science) (eds R. Vanselow and R. Howe), Springer, Berlin, p. 87. 11 Shen, Y.R. (1984) Principles of Nonlinear Optics, John Wiley & Sons, Inc., New York. 12 Shen, Y.R. (1994) Proc. Int. Sch. Phys., 120, 139. 13 Eisenthal, K.B. (1996) Chem. Rev., 96, 1343. 14 Cremer, P.S., Su, X., Shen, Y.R., and Somorjai, G.A. (1998) J. Mol. Catal., 131, 225. 15 Somorjai, G.A. and Rupprechter, G. (1999) J. Phys. Chem. B, 103, 1623. 16 Tadjeddine, A. and Peremans, A. (1996) Surf. Sci., 368, 377. 17 Richmond, G. (2001) Annu. Rev. Phys. Chem., 52, 357. 18 Buck, M. and Himmelhaus, M. (2001) J. Vac. Sci. Technol. A, 19, 2717. 19 Williams, C.T. and Beattie, D.A. (2002) Surf. Sci., 500, 545. 20 Rupprechter, G. (2007) Adv. Catal., 51, 133. 21 Shen, Y.R. (1989) Annu. Rev. Phys. Chem., 40, 327. 22 Shen, Y.R. (1994) Surf. Sci., 299–300, 551. 23 Chen, P., Westerberg, S., Kung, K.Y., Zhu, J., Grunes, J., and Somorjai, G.A. (2002) Appl. Catal. A, 229, 147. 24 Somorjai, G.A., York, R.L., Butcher, D., and Park, J.Y. (2007) Phys. Chem. Chem. Phys., 9, 3500. 25 Somorjai, G.A. (1996) Chem. Rev., 96, 1223. 26 Vidal, F. and Tadjeddine, A. (2005) Rep. Prog. Phys., 68, 1095.
27 Eisenthal, K.B. (1993) Account. Chem. Res., 26, 636. 28 Rupprechter, G., Kaichev, V.V., Morkel, M., Unterhalt, H., and Bukhtiyarov, V.I. (2004) Appl. Surf. Sci., 235, 26. 29 Rupprechter, G., Dellwig, T., Unterhalt, H., and Freund, H.-J. (2001) J. Phys. Chem. B, 105, 3797. 30 Tadjeddine, A. and Peremans, A. (eds) (1998) Advances in Spectroscopy, John Wiley & Sons, Inc., New York. 31 Dai, H.L. and Ho, W. (1995) Laser Spectroscopy and Photochemistry on Metal Surfaces Part 1, World Scientific. 32 Zhang, D., Gutov, J., and Eisenthal, K.B. (1994) J. Phys. Chem., 98, 13729. 33 Hirose, C., Yamamoto, H., Akamatsu, N., and Domen, K. (1993) J. Phys. Chem., 97, 10064. 34 Bandara, A., Dobashi, S., Onda, K., Kubota, J., Wada, A., Domen, K., Hirose, C., and Kano, S. (1997) Surf. Sci., 387, 312. 35 Bandara, A., Kubota, J., Onda, K., Wada, A., Domen, K., and Hirose, C. (1999) Surf. Sci., 427–428, 331. 36 Galletto, P., Unterhalt, H., and Rupprechter, G. (2003) Chem. Phys. Lett., 367, 785. 37 Bandara, A., Katano, S., Kubota, J., Onda, K., Wada, A., Domen, K., and Hirose, C. (1998) Chem. Phys. Lett., 290, 261. 38 Bandara, A., Kubota, J., Onda, K., Wada, A., Domen, K., and Hirose, C. (1999) Surf. Sci., 433–435, 83. 39 Bandara, A., Kubota, J., Onda, K., Wada, A., Domen, K., and Hirose, C. (1999) Appl. Phys. B., 68, 573. 40 Hirose, C., Bandara, A., Katano, S., Kubota, J., Wada, A., and Domen, K. (1999) Appl. Phys. B, 68, 559. 41 Bonn, M., Hess, C., Funk, S., Miners, J.H., Persson, B.N.J., Wolf, M., and Ertl, G. (2000) Phys. Rev. Lett., 84, 4653. 42 Hess, C., Funk, S., Bonn, M., Denzler, D.N., Wolf, M., and Ertl, G. (2000) Appl. Phys. A, 71, 477. 43 Ueba, H. (1997) Prog. Surf. Sci., 55, 115. 44 Richter, L.J., Petralli-Mallow, T.P., and Stephenson, J.C. (1998) Opt. Lett., 23, 1594.
431
432
26 Sum Frequency Generation (SFG) Spectroscopy 45 Symonds, J.P.R., Arnolds, H., Zhang, V.L., Fukutani, K., and King, D.A. (2004) J. Chem. Phys., 120, 7158. 46 Roeterdink, W.G., Aarts, J.F.M., Kleyn, A.W., and Bonn, M. (2004) J. Phys. Chem. B, 108, 14491. 47 Ertl, G., Knözinger, H., Schüth, F., and Weitkamp, J. (2008) Handbook of Heterogeneous Catalysis, Wiley-VCH Verlag GmbH, Weinheim. 48 Chorkendorff, I. and Niemantsverdriet, J.W. (2003) Concepts of Modern Catalysis and Kinetics, Wiley-VCH Verlag GmbH, Weinheim. 49 Gruber, H.L. (1962) J. Phys. Chem., 66, 48. 50 Freund, H.-J. and Goodman, D.W. (2008) Ultrathin Oxide Films, in Handbook of Heterogeneous Catalysis (eds G. Ertl, H. Knözinger, F. Schüth, and J. Weitkamp), Wiley-VCH Verlag GmbH, Weinheim. 51 Shen, Y.R. (1989) Nature, 337, 519. 52 Bain, C.D. (1995) J. Chem. Soc., Faraday Trans., 91, 1281. 53 Hirose, C., Akamatsu, N., and Domen, K. (1992) J. Chem. Phys., 96, 997. 54 Morkel, M., Unterhalt, H., Klüner, T., Rupprechter, G., and Freund, H.-J. (2005) Surf. Sci., 586, 146. 55 Unterhalt, H., Rupprechter, G., and Freund, H.-J. (2002) J. Phys. Chem. B, 106, 356. 56 Guyot-Sionnest, P. (2005) Surf. Sci., 585, 1. 57 Lu, G.Q., Lagutchev, A., Dlott, D.D., and Wieckowski, A. (2005) Surf. Sci., 585, 3. 58 Höbel, F., Bandara, A., Rupprechter, G., and Freund, H.-J. (2006) Surf. Sci., 600, 963. 59 Backus, E.H.G. and Bonn, M. (2005) Chem. Phys. Lett., 412, 152. 60 van der Ham, W.F., Vrehen, Q.H.F., and Eliel, E.R. (1996) Opt. Lett., 21, 1448. 61 Rupprechter, G., Dellwig, T., Unterhalt, H., and Freund, H.-J. (2001) Top. Catal., 15, 19. 62 Rupprechter, G. (2001) Phys. Chem. Chem. Phys., 3, 4621. 63 Williams, C.T., Yang, Y., and Bain, C.D. (1999) Catal. Lett., 61, 7.
64 Hunt, J.H., Guyot-Sionnest, P., and Shen, Y.R. (1987) Chem. Phys. Lett., 133, 189. 65 Baldelli, S., Markovic, N., Ross, P.N., Shen, Y.R., and Somorjai, G.A. (1999) J. Phys. Chem. B, 103, 8920. 66 Bandara, A., Kubota, J., Onda, K., Wada, A., Kano, S.S., Domen, K., and Hirose, C. (1998) J. Phys. Chem. B, 102, 5951. 67 Shen, Y.R. (1999) Appl. Phys. B, 68, 295. 68 Tourillon, G., Dreesen, L., Volcke, C., Sartenaer, Y., Thiry, P.A., and Peremans, A. (2007) Nanotechnology, 18, 415301. 69 Yeganeh, M.S., Dougal, S.M., and Silbernagel, B.G. (2006) Langmuir, 22, 637. 70 Flörsheimer, M. (1999) Phys. Status Solidi A, 173, 15. 71 Flörsheimer, M., Brillert, C., and Fuchs, H. (1999) Mater. Sci. Eng. C, 8–9, 335. 72 Kuhnke, K., Hoffmann, D.M.P., Wu, X.C., Bittner, A.M., and Kern, K. (2003) Appl. Phys. Lett., 83, 3830. 73 Schaller, R.D., Johnson, J.C., Wilson, K.R., Lee, L.F., Haber, L.H., and Saykally, R.J. (2002) J. Phys. Chem. B, 106, 5143. 74 Humbert, B., Grausem, J., Burneau, A., Spajer, M., and Tadjeddine, A. (2001) Appl. Phys. Lett., 78, 135. 75 Somorjai, G.A. and McCrea, K.R. (2000) Adv. Catal., 45, 385. 76 Su, X., Cremer, P.S., Shen, Y.R., and Somorjai, G.A. (1996) Phys. Rev. Lett., 77, 3858. 77 Cremer, P.S., McIntyre, B.J., Salmeron, M., Shen, Y.R., and Somorjai, G.A. (1995) Catal. Lett., 34, 11. 78 Cremer, P.S., Su, X., Shen, Y.R., and Somorjai, G.A. (1996) J. Am. Chem. Soc., 118, 2942. 79 Cremer, P.S., Stanners, C., Niemantsverdriet, J.W., Shen, Y.R., and Somorjai, G.A. (1995) Surf. Sci., 328, 111. 80 Somorjai, G.A. (1995) Surf. Sci., 335, 10. 81 Metka, U., Schweitzer, M.G., Volpp, H.R., Wolfrum, J., and Warnatz, J. (2000) Z. Phys. Chem., 214, 865. 82 Volpp, H.-R. and Wolfrum, J. (2001) Sum-frequency generation (SFG) vibrational spectroscopy as a means for
References
83
84
85
86 87 88
89
90
91
92 93 94 95
96
97
98
the investigation of catalytic combustion, in Applied Combustion Diagnostics (eds K. Kohse-Höinghaus and J.B. Jeffries), Taylor & Francis, New York. Katano, S., Bandara, A., Kubota, J., Onda, K., Wada, A., Domen, K., and Hirose, C. (1999) Surf. Sci., 427–428, 337. Pery, T., Schweitzer, M.G., Volpp, H.-R., Wolfrum, J., Ciossu, L., Deutschmann, O., and Warnatz, J. (2002) Proceedings of the Combustion Institute, The Combustion Institute, p. 973. Arnolds, H., Symonds, J.P.R., Zhang, V.L., and King, D.A. (2003) Rev. Sci. Instrum., 74, 3943. Hoffmann, F.M. (1983) Surf. Sci. Rep., 3, 103. Chen, Z., Gracias, D.H., and Somorjai, G.A. (1999) Appl. Phys. B, 68, 549. Esch, F., Greber, T., Kennou, S., Siokou, A., Ladas, S., and Imbihl, R. (1996) Catal. Lett., 38, 165. Schönnenbeck, M., Cappus, D., Klinkmann, J., Freund, H.-J., Petterson, L.G.M., and Bagus, P.S. (1996) Surf. Sci., 347, 337. Chen, Z., Gracias, D.H., and Somorjai, G.A. (1999) Appl. Phys. B, 68, 549. Rupprechter, G. (2004) Annual Reports on the Progress of Chemistry, Section C (Physical Chemistry Issued by: G. Webb (RSC), vol. 100, Royal Chemical Society Press, Cambridge, UK, p. 237. Ertl, G. (1994) Surf. Sci., 299/300, 742. Ertl, G. and Freund, H.-J. (1999) Phys. Today, 52, 32. Rupprechter, G. (2007) Catal. Today, 126, 3. Su, X., Cremer, P.S., Shen, Y.R., and Somorjai, G.A. (1997) J. Am. Chem. Soc., 119, 3994. Dellwig, T., Rupprechter, G., Unterhalt, H., and Freund, H.-J. (2000) Phys. Rev. Lett., 85, 776. Rupprechter, G., Unterhalt, H., Morkel, M., Galletto, P., Hu, L., and Freund, H.-J. (2002) Surf. Sci., 502–503, 109. Härle, H., Lehnert, A., Metka, U., Volpp, H.R., Willms, L., and Wolfrum, J. (1998) Chem. Phys. Lett., 293, 26.
99 Härle, H., Lehnert, A., Metka, U., Volpp, H.R., Willms, L., and Wolfrum, J. (1999) Appl. Phys. B, 68, 567. 100 Kung, K.Y., Chen, P., Wei, F., Shen, Y.R., and Somorjai, G.A. (2000) Surf. Sci., 463, 627. 101 Rupprechter, G., Dellwig, T., Unterhalt, H., and Freund, H.-J. (2001) J. Phys. Chem. B, 105, 3797. 102 Rupprechter, G., Unterhalt, H., Morkel, M., Galletto, P., Dellwig, T., and Freund, H.-J. (2003) Vacuum, 71, 83. 103 Kuhn, W.K., Szanyi, J., and Goodman, D.W. (1992) Surf. Sci. Lett., 274, L611. 104 Yudanov, I.V., Sahnoun, R., Neyman, K.M., Rösch, N., Hoffmann, J., Schauermann, S., Johánek, V., Unterhalt, H., Rupprechter, G., Libuda, J., and Freund, H.-J. (2003) J. Phys. Chem. B, 107, 255. 105 Napetschnig, E., Schmid, M., and Varga, P. (2007) Surf. Sci., 601, 3233. 106 Wolter, K., Seiferth, O., Kuhlenbeck, H., Bäumer, M., and Freund, H.-J. (1998) Surf. Sci., 399, 190. 107 Morkel, M., Rupprechter, G., and Freund, H.-J. (2003) J. Chem. Phys., 119, 10853. 108 Morkel, M., Rupprechter, G., and Freund, H.-J. (2005) Surf. Sci., 588, 209. 109 Borasio, M., Rodríguez de la Fuente, O., Rupprechter, G., and Freund, H.-J. (2005) J. Phys. Chem. B Lett., 109, 17791. 110 Borasio, M. (2006) Polarization modulation infrared reflection absorption spectroscopy on Pd model catalysts at elevated pressure. PhD thesis, Free University Berlin. 111 Bäumer, M., Libuda, J., Neyman, K.M., Rösch, N., Rupprechter, G., and Freund, H.-J. (2007) Phys. Chem. Chem. Phys., 9, 3541. 112 Rupprechter, G. and Weilach, C. (2008) J. Phys. Condens. Matter, 20, 184020. 113 Hess, C., Wolf, M., and Bonn, M. (2000) Phys. Rev. Lett., 85, 4341. 114 Domen, K., Bandara, A., Kubota, J., Onda, K., Wada, A., and Hirose, C. (1999) Surf. Sci., 427-428, 349. 115 Bandara, A., Kubota, J., Wada, A., Domen, K., and Hirose, C. (1996) J. Phys. Chem., 100, 14962.
433
434
26 Sum Frequency Generation (SFG) Spectroscopy 116 Bandara, A., Kubota, J., Wada, A., Domen, K., and Hirose, C. (1996) Surf. Sci., 364, L580. 117 Kubota, J., Bandara, A., Wada, A., and Domen, K. (1996) Surf. Sci., 368, 361. 118 Miura, T., Kobayashi, H., and Domen, K. (2001) J. Phys. Chem. B, 105, 10001. 119 Funk, S., Bonn, M., Denzler, D.N., Hess, C., Wolf, M., and Ertl, G. (2000) J. Chem. Phys., 112, 9888. 120 Bandara, A., Kano, S.S., Onda, K., Katano, S., Kubota, J., Domen, K., Hirose, C., and Wada, A. (2002) Bull. Chem. Soc. Jpn, 75, 1125. 121 Roke, S., Schins, J., Müller, M., and Bonn, M. (2003) Phys. Rev. Lett., 90, 128101. 122 Bourguignon, B., Carrez, S., Dragnea, B., and Dubost, H. (1998) Surf. Sci., 418, 171. 123 Persson, B.N.J., Hoffmann, F.M., and Ryberg, R. (1989) Phys. Rev. B., 34, 2266. 124 Kweskin, S.J., Rioux, R.M., Habas, S.E., Komvopoulos, K., Yang, P., and Somorjai, G.A. (2006) J. Phys. Chem. B, 110, 15920. 125 Borchert, H., Fenske, D., Kolny-Olesiak, J., Parisi, J., Al-Shamery, K., and Bäumer, M. (2007) Angew. Chem. Int. Ed., 46, 2923. 126 Tadjeddine, A., LeRille, A., Pluchery, O., Vidal, F., Zheng, W.Q., and Peremans, A. (1999) Phys. Status Solidi A, 175, 89. 127 Guyot-Sionnest, P. and Tadjeddine, A. (1990) J. Chem. Phys., 92, 734. 128 Guyot-Sionnest, P. and Tadjeddine, A. (1990) Chem. Phys. Lett., 172, 341. 129 Tadjeddine, A. and Guyot-Sionnest, P. (1991) Electrochim. Acta, 36, 1849. 130 Daum, W., Friedrich, K.A., Klunker, C., Knabben, D., Stimming, U., and Ibach, H. (1994) Appl. Phys. B, 59, 553. 131 Tadjeddine, A. and Peremans, A. (1996) J. Electroanal. Chem., 409, 115. 132 Tadjeddine, A., Peremans, A., LeRille, A., Zheng, W., Tadjeddine, M., and Flament, J.P. (1996) J. Chem. Soc., Faraday Trans., 92, 3823. 133 Tadjeddine, A., Peremans, A., and Guyot-Sionnest, P. (1995) Surf. Sci., 335, 210.
134 LeRille, A., Tadjeddine, A., Zheng, W.Q., and Peremans, A. (1997) Chem. Phys. Lett., 271, 95. 135 Vidal, F., Busson, B., and Tadjeddine, A. (2004) Electrochim. Acta, 49, 3637. 136 LeRille, A. and Tadjeddine, A. (1999) J. Electroanal. Chem., 467, 238. 137 Pluchery, O., Zheng, W.Q., Marin, T., and Tadjeddine, A. (1999) Phys. Status Solidi A. 138 Peremans, A. and Tadjeddine, A. (1995) J. Chem. Phys., 103, 7197. 139 Peremans, A., Tadjeddine, A., and Guyot-Sionnest, P. (1995) Surf. Sci., 335, 210. 140 Peremans, A., Caudano, Y., Thiry, P.A., Dumas, P., Zheng, W.Q., LeRille, A., and Tadjeddine, A. (1997) Phys. Rev. Lett., 78, 2999. 141 Peremans, A. and Tadjeddine, A. (1995) J. Electroanal. Chem., 395, 313. 142 Browmaker, G.A., Leger, J.-M., LeRille, A., Melendres, A., and Tadjeddine, A. (1998) J. Chem. Soc., Faraday Trans., 94, 1309. 143 Caudano, Y., Peremans, A., Thiry, P.A., Dumas, P., and Tadjeddine, A. (1997) Surf. Sci., 377–379, 1071. 144 Chou, K.C., Markovic, N.M., Kim, J., Ross, P.N., and Somorjai, G.A. (2003) J. Phys. Chem. B, 107, 1840. 145 Kim, J., Chou, K.C., and Somorjai, G.A. (2002) J. Phys. Chem. B, 106, 9198. 146 Lahann, J., Mitragotri, S., Tran, T.N., Kaido, H., Sundaram, J., Choi, I.S., Hoffer, S., Somorjai, G.A., and Langer, R. (2003) Science, 299, 371. 147 Chou, K.C., Westerberg, S., Shen, Y.R., Ross, P.N., and Somorjai, G.A. (2004) Phys. Rev. B, 69, 153413. 148 Guyot-Sionnest, P., Dumas, P., Chabal, Y.J., and Higashi, G.S. (1990) Phys. Rev. Lett., 64, 2156. 149 Zheng, W.Q. and Tadjeddine, A. (2003) J. Chem. Phys., 119, 13096. 150 Zheng, W.Q., Pluchery, O., and Tadjeddine, A. (2002) Surf. Sci., 502–503, 490. 151 Allen, H.C., Raymond, E.A., and Richmond, G.L. (2001) J. Phys. Chem. A, 105, 1649. 152 Baldelli, S., Mailhot, G., Ross, P., and Somorjai, G.A. (2001) J. Am. Chem. Soc., 123, 7697.
References 153 Baldelli, S., Mailhot, G., Ross, P.N., Shen, Y.R., and Somorjai, G.A. (2001) J. Phys. Chem. B, 105, 654. 154 Schnitzer, C., Baldelli, S., and Shultz, M.J. (1999) Chem. Phys. Lett., 313, 416. 155 Schnitzer, C., Baldelli, S., and Shultz, M.J. (2000) J. Phys. Chem. B, 104, 585. 156 Matranga, C. and Guyot-Sionnest, P. (2001) Chem. Phys. Lett., 340, 39. 157 Zhang, D., Ward, R.S., Shen, Y.R., and Somorjai, G.A. (1997) J. Phys. Chem. B, 101, 9060. 158 Wang, J., Woodcock, S.E., Buck, S.M., Chen, C.Y., and Chen, Z. (2001) J. Am. Chem. Soc., 123, 9470. 159 Wang, J., Chen, C.Y., Buck, S.M., and Chen, Z. (2001) J. Phys. Chem. B, 105, 12118. 160 Kim, J., Koffas, T.S., Lawrence, C.C., and Somorjai, G.A. (2004) Langmuir, 20, 4640. 161 Zhang, D., Shen, Y.R., and Somorjai, G.A. (1997) Chem. Phys. Lett., 281, 394. 162 Gracias, D.H., Zhang, D., Lianos, L., Ibach, W., Shen, Y.R., and Somorjai, G.A. (1999) Chem. Phys. Lett., 245, 277. 163 Gautam, K.S. and Dhinojwala, A. (2002) Phys. Rev. Lett., 88, 145501. 164 Zhang, D., Gracias, D.H., Ward, R., Gaukler, M., Tian, Y., Shen, Y.R., and Somorjai, G.A. (1998) J. Phys. Chem. B, 102, 6225. 165 Chen, Q., Zhang, D., Somorjai, G.A., and Bertozzi, C. (1999) J. Am. Chem. Soc., 121, 446. 166 Kweskin, S.J., Komvopoulos, K., and Somorjai, G.A. (2005) J. Phys. Chem. B, 109, 23415. 167 Kweskin, S.J., Komvopoulos, K., and Somorjai, G.A. (2006) Appl. Phys. Lett., 88, 134105. 168 Mermut, O., Phillips, D.C., York, R.L., McCrea, K.R., Ward, R.S., and Somorjai, G.A. (2006) J. Am. Chem. Soc., 128, 3598. 169 Kim, J. and Somorjai, G.A. (2003) J. Am. Chem. Soc., 125, 3150. 170 Amitay-Sadovsky, E., Komvopoulos, K., Tian, Y., and Somorjai, G.A. (2002) Appl. Phys. Lett., 80, 1829. 171 Wang, J., Paszti, Z., Clarke, M.L., Chen, X., and Chen, Z. (2007) J. Phys. Chem. B, 111, 6088. 172 Chen, X., Tang, H., Even, M.A., Wang, J., Tew, G.N., and Chen, Z.
173 174
175
176 177 178 179
180
181
182
183
184 185 186
187 188 189
190 191
192
(2006) J. Am. Chem. Soc., 128, 2711. Superfine, R., Huang, J.Y., and Shen, Y.R. (1991) Phys. Rev. Lett., 66, 1066. Stanners, C.D., Du, Q., Chin, R.P., Cremer, P.S., Somorjai, G.A., and Shen, Y.R. (1995) Chem. Phys. Lett., 232, 407. Du, Q., Superfine, R., Freysz, E., and Shen, Y.R. (1993) Phys. Rev. Lett., 70, 2313. Du, Q., Freysz, E., and Shen, Y.R. (1994) Science, 264, 826. Wei, X. and Shen, Y.R. (2001) Phys. Rev. Lett., 86, 4799. Simonelli, D., Baldelli, S., and Schultz, M.J. (1998) Chem. Phys. Lett., 298, 400. Brown, M.G., Raymond, E.A., Allen, H.C., Scatena, L.F., and Richmond, G.L. (2000) J. Phys. Chem. A, 104, 10220. Allen, H.C., Gragson, D.E., and Richmond, G.L. (1999) J. Phys. Chem. B, 103, 660. Belkin, M.A., Kulakov, T.A., Ernst, K.-H., Yan, L., and Shen, Y.R. (2000) Phys. Rev. Lett., 85, 4474. Schultz, M.J., Baldelli, S., Schnitzer, C., and Simonelli, D. (2002) J. Phys. Chem. B, 106, 5313. Guyot-Sionnest, P., Hunt, J.H., and Shen, Y.R. (1987) Phys. Rev. Lett., 59, 1597. Wolfrum, K. and Laubereau, A. (1994) Chem. Phys. Lett., 228, 83. Braun, R., Casson, B., and Bain, C.D. (1995) Chem. Phys. Lett., 245, 326. Bell, G.R., Bain, C.D., Li, Z.X., Thomas, R.K., Duffy, D.C., and Penfold, J. (1997) J. Am. Chem. Soc., 119, 10227. Casson, B.D. and Bain, C.D. (1999) J. Am. Chem. Soc., 121, 2615. Simonelli, D. and Shultz, M.J. (2000) J. Chem. Phys., 112, 6804. Messmer, M.C., Conboy, J.C., and Richmond, G.L. (1995) J. Am. Chem. Soc., 117, 8039. Scatena, L.F., Brown, G., and Richmond, G. (2001) Science, 292, 908. Knock, M.M., Bell, G.R., Hill, E.K., Turner, H.J., and Bain, C.D. (2003) J. Phys. Chem. B, 107, 10801. McFearin, C.L., Beaman, D.K., Moore, F.G., and Richmond, G.L. (2009) J. Phys. Chem. C, 113, 1171.
435
437
27 Other Photon-Detecting Techniques John C. Rivière
27.1 Appearance Potential Methods
Appearance potential methods all depend on detecting the threshold of ionization of a shallow core level and the fine structure near the threshold. They differ only in the way in which detection is performed. In all of these methods, the primary electron energy is ramped upward from near-zero to whatever is appropriate for the sample material, while the primary current to the sample is kept constant. As the incident energy is increased, it passes through successive thresholds for ionization of core levels of atoms in the surface. An ionized core level, as discussed earlier, can recombine by emission either of a characteristic X-ray photon or of an Auger electron. 27.1.1 Soft X-Ray Appearance Potential Spectroscopy (SXAPS)
In SXAPS, the X-ray photons emitted by the sample are detected, normally by allowing them to strike a photosensitive surface from which photoelectrons are collected. However, with the advent of X-ray detectors of increased sensitivity they may also be detected directly. Above the X-ray emission threshold from a particular core level the excitation probability is a function of the densities of unoccupied electronic states. As two electrons are involved – the incident and the excited – the shape of the spectral structure is proportional to the self-convolution of the unoccupied state densities.
27.2 Inverse Photoemission Spectroscopy (IPES) and Bremsstrahlung Isochromat Spectroscopy (BIS)
The irradiation of a surface with electrons leads to the emission not only of X-ray photons of energies characteristic of the material, but also of a continuous Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
438
27 Other Photon-Detecting Techniques
background of photons called Bremsstrahlung radiation. If a detector is set to detect only those background photons of a particular energy, and the primary electron energy is ramped upwards from zero, then the variations in photon flux should mirror variations in the densities of unoccupied electronic states. This process is termed inverse photoemission, because it is clearly the opposite of ordinary photoemission, as observed in XPS and UPS. Although the name inverse photoemission spectroscopy (IPES) should apply to all forms of the inverse photoemission technique, it is in fact confined by usage to the form where the energy at which the photons are detected is in the UV region, typically with hν = 9.7 eV. If, on the other hand, a crystal monochromator normally used to provide monochromatized X-rays for XPS is used in reverse, with a detector placed where the X-ray source is in Figure 2.3, the detection is then at the X-ray energy hν = 1486.6 eV, and the technique is referred to as Bremsstrahlung isochromat spectroscopy (BIS). Since IPES maps the densities of unoccupied states, it is related to other techniques that perform the same function (e.g., STS and SXAPS). When used in conjunction with a technique that maps the densities of occupied surface states, such as UPS or energy loss spectroscopy (ELS), a continuous spectrum of state density from occupied to unoccupied can be obtained. Just as in UPS, in which angular resolution enables the elucidation of the three-dimensional (3-D) occupied band structure, so too in IPES does angular resolution enable mapping of the 3-D unoccupied band structure; this latter version is known as K-resolved IPES (KRIPES).
439
Part Four Scanning Probe Microscopy
441
28 Introduction Gernot Friedbacher
The invention of scanning tunneling microscopy (STM) by Binnig and Rohrer in 1981 [1, 2], triggered a tremendous development of scanning probe microscopy (SPM) techniques [3–5]. The basic principle that all scanning probe microscopes have in common (see Figure 28.1) involves a sharp probe (e.g., tip, optical fiber, pipette) being raster-scanned across a sample surface by means of piezoelectric translators, while a certain signal is recorded by the probe for every single image point. Due to the fact that the size of the probe’s apex and the distance between probe and sample surface can be smaller than the wavelength of the analytical signal (e.g., electrons, electromagnetic radiation), information can be obtained in the so-called near-field; thus, the resolution is no longer diffraction-limited as given by the wavelength of the signal. As the measured signal constitutes local information of the surface below the probe, the term local probe technique is often referred to. The most important perspective of this concept is to use the local signal for monitoring the distance between the probe and the surface; thus, topographical information – in principle, down to atomic resolution – can be obtained in real space. Moreover, the possibility to position sharp probes above a surface with Ångstrom and sub-Ångstrom precision has opened up the exciting potential to access local spectroscopic information, even when confined to spots as small as a single atom. Although many scanning probe microscopy techniques – incorporating a wide variety of signal-generation mechanisms – has emerged during recent years, atomic force microscopy (AFM) and STM are clearly the most important, having been used in a wide range of applications in science and technology. Nonetheless, the development of scanning near-field optical microscopy (SNOM) (see Chapter 31) has opened a promising perspective for the chemical characterization of surfaces by vibrational spectroscopy, with high lateral resolution.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
442
28 Introduction
Figure 28.1 The general scheme of scanning probe microscopy. A sharp probe (e.g., tip,
optical fiber, pipette) is raster-scanned across a sample surface by means of piezoelectric translators, while a certain signal is recorded by the probe for every single image point.
References 1 Binnig, G., Rohrer, H., Gerber, C., and Weibel, E. (1982) Phys. Rev. Lett., 49, 57–61. 2 Bonnell, D.A. (ed.) (2000) Scanning Tunneling Microscopy and Spectroscopy. Theory, Techniques, and Applications, Wiley-VCH Verlag GmbH, New York. 3 Wiesendanger, R. (ed.) (1994) Scanning Probe Microscopy and Spectroscopy. Methods
and Applications, Cambridge University Press, Cambridge. 4 Colton, R.J., Engel, A., Frommer, J.E., Gaub, H.E., Gewirth, A.A., Guckenberger, R., Rabe, J., Heckl, W.M., and Parkinson, B. (eds) (1988) Procedures in Scanning Probe Microscopies, John Wiley & Sons, Ltd, Chichester. 5 Bhushan, B. (ed.) (2008) Nanotribology and Nanomechanics, Springer, Berlin.
443
29 Atomic Force Microscopy (AFM) Gernot Friedbacher
29.1 Principles
Atomic force microscopy (AFM), as invented in 1985 by Binnig, Quate, and Gerber [1–4], is a member of the group of scanning force microscopy (SFM) techniques, all of which are based on the measurement of different forces (e.g., attractive, repulsive, magnetic, electrostatic, van der Waals) between a sharp tip and the sample surface. Imaging is accomplished by measuring the interaction force via deflection of a soft cantilever while raster-scanning the tip across the surface (Figure 29.1). Although, various types of forces are encountered when a tip approaches the sample surface, signal generation in AFM is essentially based on interatomic repulsive forces which are of an extreme short-range nature (Figure 29.2). In the ideal case, the tip can be imagined to terminate with a single atom, which means that direct contact of the tip with the sample surface is confined to an extremely small area. As a consequence, there is always an interatomic repulsive force at this small contact area due to penetration of the electronic shells of the tip and substrate atoms. Since the interatomic repulsive force is influenced by the total electron density around an atom – and not just by particular states of energy, as in scanning tunneling microscopy (STM) – this force can be used to map the topography of the surface down to atomic dimensions. In addition to this short-range force, long-range forces – such as coulomb forces between charges, dipole–dipole-interactions, polarization forces, van der Waals dispersion forces, and capillary forces due to adsorbate films between the tip and substrate – all of which can be either attractive or repulsive, are also encountered (Figure 29.2). Although, both types of forces contribute to the total force acting on the probing tip, only the varying repulsive interatomic force will allow the high-resolution imaging of surfaces, because of its extreme short-range nature. As the long-range attractive forces pull the tip towards the sample surface, they give rise to an increase in the local repulsive force, which may cause a deterioration in the experimental conditions. This may lead, for instance, to artifactual results due to sample deformation or, in the worst case, the sample may be destroyed by the tip scratching across the surface. Thus, it is important to minimize those long-range Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
444
29 Atomic Force Microscopy (AFM)
Figure 29.1 The principle of AFM. The sample symbolized by the circles is scanned by means
of a piezoelectric translator. The piezo crystal and the oscillator is only needed for tapping mode operation.
Figure 29.2 Schematic view of forces encountered when the tip touches the sample surface.
White circles symbolize tip atoms, gray circles symbolize sample atoms.
forces (e.g., by imaging under liquids) in order to achieve very low repulsive forces (nano-Newton and less) in the contact area between tip and sample. This is especially important when investigating soft materials, which may easily be deformed or destroyed by the load of the tip. The interaction between the tip and sample can be described by force–distance curves. The schematic force–distance curve shown in Figure 29.3 indicates how the force changes when the sample surface approaches the tip. At large separations, there is no interaction and the observed force is zero (straight line between points 1 and 2, if it is assumed that there are no electrostatic charging forces). At point 2, the tip jumps into contact due to attractive van der Waals interactions. Then, as the sample is moved further towards the tip, the total force acting on the cantilever becomes repulsive. When the sample is retracted again, the force is
29.1 Principles
Figure 29.3 Force–distance curve depicting the interaction of the AFM tip with the sample surface. The operation range of the AFM is between points 3 and 4. For an explanation, see the text.
reduced along the line from points 3 to 4. Below the zero force line in the diagram, the net force acting on the cantilever becomes attractive, because the tip is held at the surface due to adhesion. At point 4 the adhesion force and the cantilever load are just balanced, and the tip flips off the surface when the sample is further retracted. For AFM measurements (in contact mode) the force can be set along the curve between points 3 and 4, but preferably closer to 4 in order to minimize the contact force. The value of the pull-off force can be reduced significantly by imaging under liquids, due to an elimination of the capillary condensation forces which pull the tip towards the sample. One option to image surfaces under more gentle conditions is to use tapping mode AFM [5, 6], whereby the cantilever with the tip is driven near its resonance frequency (typically 200–300 kHz) by means of a piezo oscillator (see Figure 29.1). Thus, only intermittent contact between the tip and sample occurs during oscillation of the cantilever. In tapping mode AFM, the topographic information is retrieved from the amplitude signal of the oscillating cantilever. The advantage of this technique is that, due to the intermittent contact, any lateral shear forces can be eliminated, and the scratching of soft samples or the removal of any loosely bound surface features (e.g., particles, thin films) can thus be avoided. Techniques like tapping mode AFM, where the cantilever is driven at a certain oscillation frequency, may be referred to as dynamic force microscopy (DFM), operated in “AC-mode.” In contrast, the original concept of AFM is based on the measurement of the quasi-static deflection of the cantilever, which can be expressed by the term “DC-mode.” In dynamic techniques such as tapping mode AFM, the motion of the cantilever is determined by the amplitude, the phase, and the frequency. Interaction of the tip with the sample surface can be described as a damping of the cantilever oscillation, where the response time of the cantilever motion towards varying damping will depend on the quality factor of the cantilever. This quality factor can be expressed as ω0/Δω, where ω0 is the resonance frequency and Δω the bandwidth of the resonance peak. Normally, high quality factors are advantageous in AFM
445
446
29 Atomic Force Microscopy (AFM)
measurements due to a better signal-to-noise ratio (SNR) and increased sensitivity, as the minimum detectable force gradient is inversely proportional to the square root of the quality factor. For free oscillations in vacuum, the quality factors are usually high; however, in air and under liquids they can be less than 100 and 10, respectively. The damping responsible for this reduction in the quality factor can be electronically compensated, and this has been introduced into the relevant literature as “Q-control” [2, 7]. In this technique, the effective quality factor of the cantilever oscillation can be changed through feedback of the detector’s amplitude signal after passing a variable gain amplifier and a variable phase shifter. In many cases, image contrast can be significantly enhanced by the application of Q-control. AFMs can be operated in two different modes:
•
In constant force mode (or constant amplitude mode in the case of tapping mode AFM), the cantilever deflection – and thus the force – is kept constant by readjusting the sample in a vertical direction, following the topographic features on the surface. In this mode comparatively large and rough sample areas can be imaged without destroying the tip and/or sample surface (“tip crash”), but the scan rates must be kept comparatively low in order to allow the feedback system to respond to height changes.
•
In constant height mode, the vertical position of the sample is held constant and the varying deflection of the cantilever recorded. In this mode, higher scan rates can be achieved; whilst this is advantageous for eliminating thermal drifts in high-resolution imaging, large scan sizes – particularly on rough surfaces – should be avoided because “tip crashes” are possible.
29.2 Further Modes of AFM Operation 29.2.1 Friction Force Microscopy (FFM)
Since in AFM the tip is permanently in contact with the sample surface while scanning, there are also shear forces acting on the tip which lead to torsion of the cantilever. This can be exploited analytically in FFM (often also referred to as lateral force microscopy; LFM). In practical terms, the lateral forces can be measured with regular AFM instrumentation. The detection of torsion of the cantilever can be achieved by means of a quadruple photodiode (see also Section 29.3), which allows the simultaneous detection of cantilever deflection and torsion. This allows the topography and lateral force images to be recorded simultaneously, and also bears the potential of assigning material-specific information (friction) to single features seen in the topography images [2, 8]. Since the friction behavior between the surface and the tip depends not only on the chemical nature of the sample surface but also on the tip surface, friction patterns in the images taken with chemically modified tips can strongly enhance the information content of the
29.2 Further Modes of AFM Operation
technique with respect to chemical surface composition. The corresponding technique, which is referred to as chemical force microscopy (CFM) [9], has been used to image hydrophilic and hydrophobic regions on organic thin films. An alternative to FFM, as introduced by Mate et al. [8], is that of modulated lateral force microscopy (MLFM) [10–13], whereby a small periodic lateral movement between the tip and sample is applied, and the lateral force is then measured using a lockin technique. This allows improvements to be made with regard to crosstalk between normal and lateral deflection, distance offsets between forward and backward scanning, and SNR; it also permits quantification to be conducted. 29.2.2 Young’s Modulus Microscopy (YMM) or Force Modulation Microscopy (FMM)
Dynamic modes of AFM operation can also be performed while the tip is in permanent contact with the sample surface, a concept introduced by Maivald et al. [14]. In this case, the sample is vertically modulated at a certain amplitude Δz; this also drives the tip, which is in contact with the sample, at a certain amplitude Δz′. Due to deformation of the sample by the load of the tip, Δz′ is smaller than Δz, and only if the sample were to be indefinitely stiff would Δz′ equal Δz. Thus, the difference between the two values provides a measure of the elasticity of the sample. Similarly to modulating the sample in a vertical direction, an oscillating force can also be applied on the cantilever. This approach has been used successfully for example, when imaging carbon fibers in a polymer matrix [14] and molecular chains in a polymer [15, 16], based on the different elastic properties of the sample. This technique is of particular interest when studying phase separation phenomena, for example. 29.2.3 Phase Imaging
In phase imaging or phase detection imaging (PDI), which can be performed simultaneously with tapping mode AFM, the phase lag between the piezo oscillation driving the cantilever and the actually measured responding oscillation of the cantilever is recorded. This phase lag depends heavily on the tip–surface interaction, which in turn affects the damping of the oscillation and the energy transfer from the tapping tip into the sample [17]. Although a general interpretation of phase images is difficult, it has been shown that phase imaging can reveal materialspecific contrast either directly or indirectly in relation to adhesion and elasticity [18, 19]. 29.2.4 Force–Distance Curve Measurements
Force–distance curves were described above to demonstrate the range of operation of a conventional AFM (see also Figure 29.3 and corresponding explanation). At
447
448
29 Atomic Force Microscopy (AFM)
this point it should be mentioned that, besides optimization of the imaging force for AFM measurements, force–distance curves can also be used to obtain information concerning interactions between the tip and the sample surface [2, 20–22]. For example, the pull-off force observed in the force–distance curves provides a measure of the adhesion between the tip and the sample. Unfortunately, however, the value of the pull-off force depends largely on the medium in which measurements are performed [21], and interpretation is not straightforward. Moreover, force–distance curves usually show large deviations from the schematic shape drawn in Figure 29.3 (see e.g., Ref. [21]). Specifically, the straight line at a force equal to zero drawn for the noncontact region is only a simplification. Due to long-range interactions (e.g., van der Waals, electrostatic), experimental force– distance curves show specific features in this region, and this has been employed, for example, when studying electrostatic interactions in electrolytes. Besides utilizing force–distance measurements to study adhesion phenomena or long-range force interactions, such force–distance measurements can also be used to exert forces which are high enough to produce small indents in a sample surface. Such nanoindentation measurements can reveal mechanical sample properties such as hardness or elasticity, on a nanometer scale. For instance, the size and structure of such indents can be characterized by subsequent AFM-imaging, thereby delivering information on the material’s hardness. The degree of elastic deformation can also be determined by the measurement of residual indents through load– displacement curves [2]. 29.2.5 Pulsed Force Mode AFM
As indicated above, there has always been a desire to image topography and to monitor material-specific properties simultaneously, using AFM, and indeed many attempts to achieve such a goal have been made. The force–distance curves provide a wealth of information on the force interaction between the tip and the sample, which can in turn be used to identify the material-related properties of a sample under investigation. However, the lateral mapping of a surface by means of force–distance curves – though possible in principle – is rather time-consuming and thus not satisfactory for real-time imaging along with surface topography. In contrast, in tapping-mode AFM the phase images can be easily recorded simultaneously with the topography. The amplitude and the phase shift depend on many parameters such as adhesion, elasticity, viscosity, or long-range forces introduced by surface charges. Unfortunately, the separation of these parameters from each other, simply by relying on two measured quantities, is not possible, and therefore the interpretation of phase images in terms of individual physical or chemical properties is difficult, or not possible. In order to overcome the problems described above, pulsed force mode AFM (PFM-AFM) has been introduced [23]. In this technique, the vertical offset of the piezo tube is modulated with a sine voltage at a frequency of 100 Hz to 5 kHz with an amplitude between 10 and 500 nm. During each oscillation cycle the tip jumps
29.2 Further Modes of AFM Operation
Figure 29.4 Force-versus-time curve. 1, Point of maximum repulsive force; 2, Free cantilever after damping of resonant oscillation; 3a and 3b, slope of increasing and decreasing repulsive force, respectively; 4, jump-off point at minimum force. Reproduced from Ref. [23].
in and out of contact with the sample, delivering a full force-versus-distance curve (Figure 29.4). In order to reduce the amount of data to be recorded and processed, only characteristic and analytically useful features of this curve are extracted by special electronic circuits. The maximum repulsive force (point 1 in Figure 29.4) at the point of the largest indentation of the tip into the sample during each oscillation cycle can be adjusted by a vertical offset of the piezo tube, which can be recorded as a topographic image of the sample. The offset of the so-called zero force (point 2 in Figure 29.4), where the tip is not in contact with the surface, can be used to obtain information on for example, long-range electrostatic forces which could map the distribution of surface charges. The elastic properties of the sample can be obtained from the slope of the force curve before and after the point of maximum force. In order to determine that slope, points 3a and 3b are also captured by the PFM-electronics. Finally, the minimum force at the jump-off point (point 4) is also captured, which corresponds to the adhesive force between the tip and the sample. The different signals can be recorded simultaneously, thus opening up the opportunity to make direct comparisons between topography and the elasticity, adhesion, and charging phenomena, with high lateral resolution. 29.2.6 Harmonic Imaging and Torsional Resonance Mode
As noted above, force–distance curves provide a wealth of information regarding tip–sample interaction, and thus also regarding the various properties of the
449
450
29 Atomic Force Microscopy (AFM)
sample. Unfortunately, the recording of such curves as described in Section 29.2.4 proceeds at low operation speeds (on the order of 1 second per measurement), which is unsuitable for real-time imaging. In contrast to force–distance measurements, phase imaging supplies information concerning tip–sample interaction during imaging; however, interpretation of the phase images in terms of the material’s properties is difficult as only two parameters of the oscillation (amplitude and phase) are used for signal generation. This does not take into account any deviations from the sinusoidal shape. Consequently, in order to fully reconstruct a force–distance cycle the shape of a full tapping oscillation cycle must be measured accurately. Although, in principle, this can be achieved by recording higher harmonics [24], for the flexural oscillations in tapping mode the SNRs of higher harmonic vibrations are insufficient for practical purposes, and the frequency response of the cantilever will depend on experimental parameters that cannot easily be controlled. In order to overcome such problems, torsional vibrations of the cantilever have been exploited for signal generation [25, 26]. One means of generating torsional vibrations is the so-called torsional harmonic cantilever (THC) (Figure 29.5) [27–29]. In the THC, the tip is located off the long axis of the cantilever, so that a torque is generated on the cantilever by tip–sample interaction forces that occur during tapping mode oscillation. The resultant torsional deflection can then be measured through the horizontal motion of the laser spot on the photodetector. The torsional vibration signal is then used to calculate the timeresolved tip–sample interaction forces, while the flexural signal, which is recorded simultaneously, is used to follow the topography of the sample. The advantage of torsional vibrations are demonstrated in Figure 29.6, where the vibrational spectra have been recorded applying a drive frequency of 51.2 kHz. It can be seen clearly that, in the torsional spectrum (Figure 29.6b), higher harmonics up to 1 MHz can be observed with a very good SNR, whereas in the flexural spectrum (Figure 29.6a) higher harmonics above 300 kHz can no longer be used due to a bad SNR. The time-resolved deflection signal during one oscillation cycle, both for the flexural and the torsional motion, are shown in Figure 29.7. At the lowest point in the flexural signal, the tip hits the surface and the resultant torque
Figure 29.5 Schematic illustration of the
torsional harmonic cantilever (THC). The tip is located asymmetrically with respect to the cantilever axis, resulting in a torque
oscillation around the axis upon tapping of the tip on the sample surface. F, force. Reproduced from Ref. [27].
29.2 Further Modes of AFM Operation
Figure 29.6 Flexural (a) and torsional (b) vibration spectra of a THC cantilever tapping on a polystyrene sample. Reproduced from Ref. [27].
Figure 29.7 Time-resolved detector signal during one oscillation cycle both for the flexural (black curve) and the torsional (gray curve) motion of a THC cantilever tapping on a polystyrene sample. Reproduced from Ref. [27].
451
452
29 Atomic Force Microscopy (AFM)
on the cantilever leads to an increase in the torsional signal (marked by an arrow). Following impact, decaying torsional oscillations are observed, from which raw data the force–distance curves (see also Section 29.2.4) can be calculated during real-time imaging. Based on these curves, material properties such as stiffness and surface adhesion can be calculated and depicted as image information, along with the topographical image. This approach has already been applied successfully when investigating the mechanical properties of polymers at different temperatures.
29.3 Instrumentation
From the general principles described above, the following basic components of the atomic force microscope can be identified:
• • • • •
A sharp tip mounted on a soft cantilever. A detection system for measuring the deflection of the cantilever. A piezoelectric translator to move the probe relative to the sample. A feedback system to keep the deflection constant by height readjustment of the probe. An imaging system to convert the single data points into an image.
The tips and cantilevers must fulfill a number of requirements. Typically, the tip should be very sharp with a small radius of curvature and a high aspect ratio in order to be able to trace fine details on the surface. The cantilever must be softer than the bonds between the atoms of the sample, in order to achieve a measurable deflection without causing any destructive displacement of the surface atoms. Moreover, the resonance frequency should be high in order to minimize noise due to coupling with low-frequency noise (e.g., vibrations of buildings) from the environment. Low spring constants and high resonance frequencies can only be achieved by reducing the mass, and thus by making the cantilevers very small. Today, a wide variety of cantilevers is available commercially, with most produced via lithographic multistep processes [30, 31] that are common in microelectronic technology. Usually, doped single-crystalline silicon is used as the cantilever material, although cantilevers made from other materials such as silicon nitride are also available. For standard tapping mode AFM applications, silicon cantilevers with a length of about 100 μm, a tip height of 10–15 μm, a resonance frequency of approximately 300 kHz, and a force constant of approximately 40 N m−1 are used (Figure 29.8). The nominal radius of curvature of the tips is typically between 5 and 10 nm. For special applications, cantilevers with force constants ranging from less than 0.1 up to several hundred N m−1, resonance frequencies from 10 up to almost 1000 kHz, different geometries, and different coatings (e.g., diamond, Cr, Al), are available. Such microfabricated cantilevers allow imaging with forces typically in the nano-Newton and sub-nano-Newton range. Note: These forces are about 10 000- to 100 000-fold lower than the force of gravity introduced by a fly (1 mg) sitting on a surface.
29.3 Instrumentation
Figure 29.8 Image of a single crystal silicon AFM cantilever taken with a secondary electron
microscope.
Figure 29.9 Schematic view of the deflection sensing system as used in the NanoScope® AFM (Bruker AXS [formerly Veeco], Santa Barbara, USA). The deflection of the
cantilever is amplified by a laser beam focused on the rear of the cantilever and reflected towards a split photodiode detector.
The force acting on the tip is sensed by measuring the deflection of the cantilever. While many options (e.g., tunneling current, interferometry, capacitance, laser diode feedback) [4] for measuring this deflection have been described, the most important option is the optical lever technique [32], as depicted in Figure 29.9. In this technique, a laser beam is focused and positioned on the rear of the cantilever, and the reflected laser beam is then directed towards a photodiode consisting of four segments. As the cantilever deflection – and thus the mirror plane for the laser beam – changes (for example, due to the sample topography), the position of the laser beam on the photodiode changes in the vertical direction, which in turn leads to a change in the difference signal between the two upper and the two lower segments. Consequently, this difference signal will serve as a sensitive measure for the cantilever deflection, with a resolution of 0.01 nm. Similarly, the difference signal between the left and right segments can be used to monitor the torsional bending of the cantilever, which is required for FFM (see Section 29.2.1) or torsional resonance imaging (see Section 29.2.6).
453
454
29 Atomic Force Microscopy (AFM)
Figure 29.10 Schematic illustration of a single tube piezo scanner. Motion of the upper cylinder face can be achieved by applying voltages to electrodes located at different regions of the tube. Reproduced from Ref. [33].
Usually, in AFM the position of the tip is fixed and the sample is raster-scanned. Following a manual coarse approach with fine-threaded screws (usually assisted by a step motor), the subsequent motion of the sample is carried out with a piezo translator made from piezo ceramics such as lead zirconate titanate (PZT). Although, in principle, piezo tripods could be used to move the sample in all three directions of space, single tube scanners are used almost exclusively. These scanners consist of a single cylindrical tube of piezo ceramics coated with metal electrodes through which the voltages are applied to control the piezo material (Figure 29.10) [33]. The maximum scan range (from 1 μm up to more than 100 μm) of such piezo tubes is basically determined by their length. Although, single-tube scanners are more difficult to calibrate, they can be built more rigidly with higher resonance frequencies, and are thus less sensitive towards vibrational perturbations. In contrast to many other surface-analytical techniques, such as scanning electron microscopy (SEM), AFM does not require a vacuum, and therefore it can be operated under ambient conditions, allowing the direct observation of processes at solid–gas and solid–liquid interfaces. The latter can be accomplished by means of a liquid cell, as shown schematically in Figure 29.11. The cell is formed by the sample at the bottom, a glass cover (which holds the cantilever) at the top, and a silicone O-ring seal in between. Studies with these liquid cells can also be performed under potential control, which provides valuable opportunities for electrochemistry [34–38]. Moreover, imaging under liquids provides the possibility to protect sensitive surfaces by in situ preparation and imaging under an inert fluid [39]. Imaging under a controlled environment can also be performed in the gas
29.4 Applications
Figure 29.11 Cross-sectional view of a liquid cell for in situ AFM measurements of surface processes. For details, see the text.
phase by means of environmental chambers that are constructed as Plexiglas hoods around the measurement cell. The ability to flood the hood with a variety of gases is especially useful when measurements are to be made under a defined humidity. For measurements at controlled temperatures, small heating elements and piezo coolers can be mounted on top of the piezo scanner, allowing investigations to be conducted over a temperature range of approximately −20 °C to over 200 °C. Finally, it should be noted that various combinations of AFM with measurements of other signals, such as current (conductive AFM [40, 41]) or temperature (scanning thermal microscopy [42, 43]) are also possible, though the details of these techniques are beyond the scope of this book.
29.4 Applications
The vast analytical potential of AFM is based on several properties, summarized as follows. Atomic resolution can be achieved, and imaging performed on areas greater than 100 μm × 100 μm. This allows overview images to be obtained that can be used to zoom in on details with high resolution, without changing the sample or the instrumental set-up, that might make it impossible to re-identify the identical sample position. An image of a PVD gold film on silicon is shown as an example in Figure 29.12a; this provides a good overview regarding the size and distribution of single crystallites, which typically are 100 nm in diameter [44]. The root-meansquare (rms)-roughness (which reflects the standard deviation of all height values within a considered area) determined from the image was 3 nm. Based on X-ray diffraction (XRD) measurements, it is known that the surface of the gold film is mainly a (111)-surface. The atomic resolution image of the position marked with the arrow in Figure 29.12a revealed a Au(111) surface for that specific crystallite
455
456
29 Atomic Force Microscopy (AFM)
Figure 29.12 (a) AFM image of a PVD gold film on silicon. Image size: 1 μm × 1 μm, depth scale: 30 nm from black to white; (b) Atomic resolution image of the crystallite marked by the arrow in panel (a). Image size: 4 nm × 4 nm, depth scale: 1 nm from black to white.
(Figure 29.12b). Although the corrugation in one lattice direction was more prominent (this may have been caused by convolution with an asymmetric tip), a hexagonal symmetry was observed and the spacings were in good agreement with the expected values (0.29 nm). One further important advantage of AFM is the fact that images contain direct depth information, which makes the technique a valuable metrological tool. An AFM image of a structured photoresist on a silicon surface is shown in Figure 29.13a, as an example, where the pattern observed has been produced using ion lithography. A cross-sectional profile through the image in Figure 29.13a is shown in Figure 29.13b; this shows that the AFM data can be conveniently analyzed using software, in order to measure the heights of the surface features in direct fashion. In addition to conducting samples, insulators may also be readily imaged using AFM, without a need for them to be coated with a conductive film. This, in turn, not only greatly facilitates sample preparation but also avoids the introduction of artifacts during the coating process. Thus, a wide variety of organic [45, 46], biological [47–51], and also inorganic insulators [52] can be studied in a straightforward manner. As an example, Figure 29.14 shows the tapping mode AFM images of a sensitive glass surface with a high level of alkali oxides, both before exposure (Figure 29.14a) and after 5 min of exposure (Figure 29.14b) to humid air (relative humidity 50%) [53]. Whilst under dry conditions, a flat cleavage surface with an rms-roughness of <1 nm was observed, exposure to humidity led to the formation of corrosion products with diameters of approximately 50 nm and heights of up
29.4 Applications
Figure 29.13 (a) AFM image of a photoresist layer on silicon structured by ion lithography.
Image size: 4 μm × 4 μm; (b) Cross-sectional profile through the AFM image shown in panel (a).
Tapping mode AFM images of cleaved glass surface under (a) dry nitrogen and (b) after 5 min of exposure to humid air (relative humidity 50%). Images sizes: 5 μm × 5 μm. Figure 29.14
to 5 nm. It should be emphasized here that the capability of tapping mode AFM to image such sensitive structures on the low-nanometer scale under ambient conditions is unique, given the fact that other high-resolution imaging techniques such as SEM require conductive samples and a vacuum. The high instability of the observed corrosion products was clearly revealed by investigations with conventional contact AFM, where the tip is in permanent contact with the surface during scanning. Here, instead of elevated structures formed by corrosion, the formation of holes was observed due to removal of the corrosion products by the AFM tip (see Figure 29.15); the corrosion products were shown to be piled up by the AFM tip at the right-hand edge of the image. One important potential role for AFM in surface chemistry relates to its capability to perform in situ measurements under liquids and in air, thus providing oppor-
457
458
29 Atomic Force Microscopy (AFM)
Figure 29.15 Contact AFM image of cleaved glass surface after 4 h exposure to humid air (relative humidity 50%). The crater in the center of the image has been produced by removal of corrosion products by the
scanning AFM tip. The removed material is piled up at the right-hand side of the image. Image size: 13 μm × 13 μm. Reproduced from Ref. [53].
Figure 29.16 Series of sequential in situ AFM images of the growth of ODS on silicon. The numbers indicate the adsorption time in minutes. Images sizes: 20 μm × 20 μm.
tunities for the direct investigation of surface processes such as corrosion, crystal growth, thin-film deposition, or electrochemical processes at electrode surfaces [34–37, 39, 54–56]. As an example of in situ AFM imaging, Figure 29.16 depicts the deposition of an octadecylsiloxane (ODS) monolayer on silicon. Such selfassembled monolayers (SAMs) can be obtained through the deposition of octadecyltrichlorosilane from toluene [57]. Clearly, large islands with diameters of typically 5 μm are deposited on the surface right from the beginning of the growth experiment; this indicates that, under the given conditions, such ordered structures had already existed in the deposition solution. However, when the uncovered surface area could no longer supply sufficient space for the deposition of further
29.4 Applications
larger islands, continuing film formation was able to proceed only by the growth of already-deposited islands through the attachment of smaller units, a process which is comparatively slow. Although, during the course of time, the growing islands will coalesce, it may take 1 h or more before full coverage of the surface is reached. It is also possible to evaluate surface coverage on a quantitative basis from these AFM images. As AFM is capable of revealing the structure of sub-monolayer islands in such deposition experiments with sub-micrometer resolution, the technique has contributed significantly to the present understanding of the respective growth mechanisms; this is important when controlling film growth through varying deposition parameters such as the water content of the precursor solution, or the temperature [57–60]. As an example, Figure 29.17 shows a series of AFM
Figure 29.17 AFM images of ODS films on silicon produced by adsorption of OTS from
toluene at various temperatures. (a) 10 °C; (b) 23 °C; (c) 25 °C; (d) 30 °C. Images sizes: (a–c) 25 μm × 25 μm; (d) 5 μm × 5 μm. Reproduced from Ref. [58].
459
460
29 Atomic Force Microscopy (AFM)
AFM images of the growth of ODS on silicon at a temperature of 20 °C and an adsorption time of 15 s. The water concentration of the adsorption solution was (a) 13 and (b) 17 mmol l−1. Image sizes: 5 μm × 5 μm. Reproduced from Ref. [59].
Figure 29.18
images of the growth of an ODS film on silicon at various temperatures. At low temperature, large ordered islands with a typical dentritic shape were observed, but with increasing temperature the islands became smaller until, at 30 °C, island growth was no longer observed under the given conditions [58]. The growth behavior of the films is governed not only by the temperature but also by the water content of the adsorption solution. AFM images of the growth of an ODS film on silicon using octadecyltrichlorosilane (OTS) dissolved in toluene as a precursor solution, are shown in Figure 29.18 [59]. This experiment was performed at two different concentrations (13 and 17 mmol l−1) of water in the deposition solution, and a temperature of 20 °C. It is clear that, at a higher water content, larger ordered islands of organic film were formed on the surface (Figure 29.18b). At a temperature of 30 °C and a water concentration of 15 mmol l−1 the growth of ordered islands could no longer be achieved (Figure 29.19a). An ordered island growth was achieved, however, even at elevated temperatures if the water concentration in the adsorption solution was increased simultaneously (to 18 mmol l−1 in this case) (Figure 29.19b). Further investigations of the adsorption solution itself showed the formation of ordered islands on the substrate surface to be closely related to the existence of ordered species (micelles) in the solution, which in turn is affected both by the temperature and the water content [59]. In fact, adsorption experiments conducted at ≤10 °C showed the existence of spherical dots (see Figure 29.20) that could be related to the micelles observed in solution by dynamic light scattering. These structures are observed by AFM only at low temperatures, since at higher temperatures they coalesce to form the typical dentritic larger islands [58]. AFM has also been used to investigate the influence of the substrate surface on the growth mechanism of SAMs. An AFM image of a sub-monolayer ODS film on a silicon surface previously rendered hydrophobic with trimethylchlorosilane (TMCS), is shown in Figure 29.21 [61]. Compared
29.4 Applications
Figure 29.19 AFM images of the growth of ODS on silicon at a temperature of 30 °C and an
adsorption time of 10 s. The water concentration of the adsorption solution was (a) 15 and (b) 18 mmol l−1. Image sizes: 5 μm × 5 μm. Reproduced from Ref. [59].
Figure 29.20 AFM images of ODS on silicon
adsorbed at a temperature of 10 °C and an adsorption time of 10 s. The water concentration of the adsorption solution was 15 mmol l−1. (a) Besides large fractally shaped
islands, smaller dot-like features can be observed. Images size: 5 μm × 5 μm; (b) Zoom of lower left region marked in panel (a). Images size: 1 μm × 1 μm. Reproduced from Ref. [58].
to silicon substrates with reactive hydroxyl groups – which are needed to form covalent siloxane bonds with the film molecules – the growth on this deactivated substrate was much slower, and spherical particles with very smooth edges were formed. This effect was explained by a higher mobility of the film molecules, as covalent siloxane bonds could not be formed on this substrate.
461
462
29 Atomic Force Microscopy (AFM)
Figure 29.21 AFM image of ODS film on silicon (adsorption time 4 h) obtained after deactivation of the silicon substrate with TMCS for 30 min. Image size: 5 μm × 5 μm. Reproduced from Ref. [61].
References 1 Colton, R.J., Engel, A., Frommer, J.E., Gaub, H.E., Gewirth, A.A., Guckenberger, R., Rabe, J., Heckl, W.M., and Parkinson, B. (eds) (1988) Procedures in Scanning Probe Microscopies, John Wiley & Sons, Ltd, Chichester. 2 Bhushan, B. (ed.) (2008) Nanotribology and Nanomechanics, Springer, Berlin. 3 Binnig, G., Quate, C.F., and Gerber, C. (1986) Phys. Rev. Lett., 56, 930–933. 4 Sarid, D. (1991) Scanning Force Microscopy, Oxford University Press, New York. 5 Zhong, Q., Inniss, D., Kjoller, K., and Elings, V.B. (1993) Surf. Sci., 290, L688–L692. 6 Hansma, P.K., Cleveland, J.P., Radmacher, M., Walters, D.A., Hillner, P.E., Bezanilla, M., Fritz, M., Vie, D., Hansma, H.G., Prater, C.B., Massie, J., Fukunaga, L., Gurley, J., and Elings, V. (1994) Appl. Phys. Lett., 64, 1738–1740. 7 Schirmeisen, A., Anczykowski, B., and Fuchs, H. (2007) Handbook of Nanotechnology, 2nd edn, Springer, pp. 737–765. 8 Mate, C.M., McClelland, G.M., Erlandsson, R., and Chiang, S. (1987) Phys. Rev. Lett., 59, 1942–1945.
9 Frisbie, C.D., Rozsnyai, L.F., Noy, A., Wrighton, M.S., and Lieber, C.M. (1994) Science, 265, 2071–2074. 10 Goeddenhenrich, T., Mueller, S., and Heiden, C. (1994) Rev. Sci. Instrum., 65, 2870–2873. 11 Colchero, J., Luna, M., and Baro, A.M. (1996) Appl. Phys. Lett., 68, 2896–2898. 12 Mazeran, P.E. and Loubet, J.L. (1999) Tribol. Lett., 7, 199–212. 13 Piétrement, O., Beaudoin, J.L., and Troyon, M. (1999) Tribol. Lett., 7, 213–220. 14 Maivald, P., Butt, H.J., Gould, S.A.C., Prater, C.B., Drake, B., Gurley, G., Elings, V.B., and Hansma, P.K. (1991) Nanotechnology, 2, 103–106. 15 Kimura, K., Kobayashi, K., Yamada, H., and Matsushige, K. (2008) Nanotechnology, 19, 065701/1–065701/5. 16 Kimura, K., Kobayashi, K., Yamada, H., and Matsushige, K. (2007) Nanotechnology, 18, 305504/1– 305504/6. 17 Anczykowski, B., Gotsmann, B., Fuchs, H., Cleveland, J.P., and Elings, V.B. (1999) Appl. Surf. Sci., 140, 376–382. 18 Schmitz, I., Schreiner, M., Friedbacher, G., and Grasserbauer, M. (1997) Appl. Surf. Sci., 115, 190–198.
References 19 Ciccotti, M., George, M., Ranieri, V., Wondraczek, L., and Marliere, C. (2008) J. Non-Cryst. Solids, 354, 564–568. 20 Burnham, N.A. and Colton, R.J. (1989) J. Vac. Sci. Technol. A, 7, 2906–2913. 21 Weisenhorn, A.L., Maivald, P., Butt, H.J., and Hansma, P.K. (1992) Phys. Rev. B Condens. Matter Mater. Phys., 45, 11226–11232. 22 Hoh, J.H., Cleveland, J.P., Prater, C.B., Revel, J.P., and Hansma, P.K. (1992) J. Am. Chem. Soc., 114, 4917–4918. 23 Rosa-Zeiser, A., Weilandt, E., Hild, S., and Marti, O. (1997) Meas. Sci. Technol., 8, 1333–1338. 24 Sahin, O., Quate, C.F., Solgaard, O., and Atalar, A. (2004) Phys. Rev. B Condens. Matter Mater. Phys., 69, 165416/1–165416/9. 25 Kasai, T., Bhushan, B., Huang, L., and Su, C. (2004) Nanotechnology, 15, 731–742. 26 Bhushan, B. and Kasai, T. (2004) Nanotechnology, 15, 923–935. 27 Sahin, O., Magonov, S., Su, C., Quate, C.F., and Solgaard, O. (2007) Nat. Nanotechnol., 2, 507–514. 28 Sahin, O., and Erina, N. (2008) Nanotechnology, 19, 445717/1–445717/9. 29 Sahin, O. (2008) Phys. Rev. B Condens. Matter Mater. Phys., 77, 115405/1– 115405/6. 30 Albrecht, T.R., Akamine, S., Carver, T.E., and Quate, C.F. (1990) J. Vac. Sci. Technol. A, 8, 3386–3396. 31 Wolter, O., Bayer, T., and Greschner, J. (1991) J. Vac. Sci. Technol. B, 9, 1353–1357. 32 Meyer, G. and Amer, N.M. (1990) Appl. Phys. Lett., 56, 2100–2101. 33 Veeco Instruments Inc. (2006) MultiMode V SPM Instruction Manual. 34 Manne, S., Hansma, P.K., Massie, J., Elings, V.B., and Gewirth, A.A. (1991) Science, 251, 183–186. 35 Gewirth, A.A. (1992) Am. Inst. Physics Conf. Proc., 241, 253–261. 36 Martin, F.A., Bataillon, C., and Cousty, J. (2008) Corros. Sci., 50, 84–92. 37 Dale, S.E.C., Bending, S.J., and Peter, L.M. (2009) Langmuir, 25, 11228–11231. 38 Hirai, N., Kubo, S., and Magara, K. (2009) J. Power Sources, 191, 97–102.
39 Prohaska, T., Friedbacher, G., Grasserbauer, M., Nickel, H., Loesch, R., and Schlapp, W. (1995) Anal. Chem., 67, 1530–1535. 40 Mativetsky, J.M., Palma, M., and Samori, P. (2008) Top. Curr. Chem., 285, 157–202. 41 Benstetter, G., Biberger, R., and Liu, D. (2009) Thin Solid Films, 517, 5100–5105. 42 Cramer, R.M., Heiderhoff, R., and Balk, L.J. (1999) Diamond Relat. Mater., 8, 1581–1586. 43 Meckenstock, R. (2008) Rev. Sci. Instrum., 79, 041101/1–041101/29. 44 Friedbacher, G., Prohaska, T., and Grasserbauer, M. (1994) Microchimica Acta, 113, 179–202. 45 Frommer, J. (1992) Angew. Chem., 104, 1325–1357. 46 Fuchs, H. (1993) J. Mol. Struct., 292, 29–47. 47 Hansma, H.G. (1996) J. Vac. Sci. Technol. B, 14, 1390–1394. 48 Kasas, S., Thomson, N.H., Smith, B.L., Hansma, P.K., Miklossy, J., and Hansma, H.G. (1997) Int. J. Imaging Syst. Technol., 8, 151–161. 49 Parot, P., Dufrene Yves, F., Hinterdorfer, P., Grimellec, C.L., Navajas, D., Pellequer, J.-L., and Scheuring, S. (2007) J. Mol. Recognit., 20, 418–431. 50 Goldsbury Claire, S., Scheuring, S., and Kreplak, L. (2009) Curr. Protoc. Protein Sci., Chapter 17, 17.7.1–17.7.19. 51 Raghavachari, M., Kottke-Marchant, K., and Marchant, R.E. (2000) Thromb. Res., 98, 351–358. 52 Magonov, S.N. (1993) Appl. Spectrosc. Rev., 28, 1–121. 53 Schmitz, I., Schreiner, M., Friedbacher, G., and Grasserbauer, M. (1997) Anal. Chem., 69, 1012–1018. 54 Hillner, P.E., Manne, S., Gratz, A.J., and Hansma, P.K. (1992) Ultramicroscopy, 42–44, 1387–1393. 55 Hillner, P.E., Gratz, A.J., Manne, S., and Hansma, P.K. (1992) Geology, 20, 359–362. 56 Schmitz, I., Prohaska, T., Friedbacher, G., Schreiner, M., and Grasserbauer, M. (1995) Fresenius J. Anal. Chem., 353, 666–669. 57 Leitner, T., Friedbacher, G., Vallant, T., Brunner, H., Mayer, U., and Hoffmann, H. (2000) Mikrochim. Acta, 133, 331–336.
463
464
29 Atomic Force Microscopy (AFM) 58 Glaser, A., Foisner, J., Friedbacher, G., and Hoffmann, H. (2004) Anal. Bioanal. Chem., 379, 653–657. 59 Glaser, A., Foisner, J., Hoffmann, H., and Friedbacher, G. (2004) Langmuir, 20, 5599–5604.
60 Rill, C., Glaser, A., Foisner, J., Hoffmann, H., and Friedbacher, G. (2005) Langmuir, 21, 6289–6295. 61 Foisner, J., Glaser, A., Leitner, T., Hoffmann, H., and Friedbacher, G. (2004) Langmuir, 20, 2701–2706.
465
30 Scanning Tunneling Microscopy (STM) Gernot Friedbacher
30.1 Principles
In STM [1, 2], the microscope’s probe is a sharp metal tip which is scanned across a conducting surface at distances in the order of typically 1 nm (Figure 30.1). A bias voltage of typically a few millivolts is applied between the tip and the sample, leading to a tunneling current in the order of a few nanoamperes. The situation that occurs when a voltage is applied between two metal electrodes separated by a very small distance is shown in Figure 30.2 [3]. If a voltage V is applied, then the Fermi levels EF are shifted against each other by an energy e × V, where e is the electrostatic charge of an electron. Due to the energy barrier between the two metals, which is in the order of the metals’ work functions Φ, at small voltages a conventional current cannot flow. However, due to the fact that the electrons can be described by wavefunctions decaying exponentially across the energy barrier, the probability of finding an electron on the opposite side of the barrier is not zero. This phenomenon is known as the quantum mechanical tunneling effect. As a consequence, electrons can flow from occupied states on the negative electrode to unoccupied states on the positive electrode, whereby at small voltages they do not change their energy; this is described as elastic tunneling. The dependence of the tunneling current IT on the applied voltage VT and the tip–surface separation s is given by : IT ≈ VT exp(− AΦ 1/ 2 s )
(30.1)
where A is a constant given by A = 2((2me) )/h (me is the mass of an electron). Due to this exponential relationship, the tunneling current is an extremely sensitive measure to control the separation. For example, a variation of the separation by 0.1 nm changes the tunneling current by a factor of approximately 10. This means that, in principle, variations in the separation of 0.001 nm can be monitored by keeping the current constant within a few percent. However, the tunneling current depends not only on the tip–surface separation; it is also influenced by the electronic structure of the surface. This is important when interpreting STM 1/2
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
466
30 Scanning Tunneling Microscopy (STM)
Figure 30.1 The operating principle of STM.
Figure 30.2 Energy level diagram of a tunneling junction [3].
images on the atomic scale, because in that case, in contrast to the AFM, in general the image no longer simply shows the topography of the surface but rather a contour map of the local density of states near the Fermi level (LDOSE ) as long as small voltages in the millivolt range are applied. For larger voltages of up to a few volts, the density of other electronic states can be imaged. At this point it should be noted that the relationships stated above are only an approximation, assuming that the tip and the applied voltage do not influence the electronic structure of the surface. Furthermore, for an exact description, three-dimensional (3-D) tunneling must be considered in order to take into account the complex spatial structure of the electronic states (wavefunctions) involved [4]. F
30.2 Instrumentation
The scanning tunneling microscope can be operated in two ways:
•
In constant current mode, the tip is scanned across the sample surface at a fixed bias voltage, while the tunneling current is held constant by readjusting the tip in vertical direction following the topographic features (or, on the atomic level, the local density of states) on the surface. In this mode, comparatively large and rough sample areas can be imaged without destroying tip and/or sample surface (“tip crash”). However, the scan rates must be kept comparatively low in order to allow the feedback system to respond to height changes. For images taken in constant current mode at a certain bias voltage, the term constant current topography (CCT) is occasionally used.
•
In constant height mode, the vertical position of the tip is kept constant and the varying tunneling current is recorded. In this mode higher scan rates can be achieved which is advantageous for eliminating thermal drifts in highresolution imaging. Large scan sizes should be avoided, however, because “tip crashes” are possible.
30.2 Instrumentation
The STM system, as shown schematically in Figure 30.3, consists of the following basic components [5]:
• • •
A sharp metal tip. A piezoelectric translator to move the tip relative to the sample. Control electronics for applying the bias voltage and measuring the tunneling current.
Figure 30.3 Scheme of an STM system [5].
467
468
30 Scanning Tunneling Microscopy (STM)
• •
A feedback system to keep the tunneling current constant by height readjustment of the tip. An imaging system to convert the single data points into an image.
The tips used for STM experiments should be sharp and stable. Chemical stability can be achieved by using a noble metal, while mechanical rigidity can be acquired by using short wires. Alloys of Pt and Ir are frequently used for the fabrication of STM tips, which can be produced in a surprisingly simple manner by cutting a metal wire with conventional cutting tools. Due to their high chemical stability, such Pt/Ir tips are well-suited for atomic resolution experiments on flat samples. However, due to their low aspect ratio near the apex they fail to trace steep features and narrow trenches. Consequently, electrolytically etched tungsten tips with higher aspect ratios are occasionally used as an alternative, although these are less stable against oxidation. Usually, in STM the position of the sample is fixed and the tip is raster-scanned. As in AFM, following a manual coarse approach with fine-threaded screws, motion of the tip is performed with a piezo translator made from piezo ceramics such as lead zirconate titanate (PZT), which can be either a piezo tripod or a single tube scanner (see also Chapter 29.3). Although physical studies of the electronic structure of surfaces must be conducted under UHV conditions in order to guarantee clean uncontaminated samples, the technique does not require vacuum for its operation. Thus, in situ observation of the processes at solid–gas and solid–liquid interfaces is also possible. Such an approach has been used, for instance, to observe corrosion and electrode processes directly, with atomic resolution [2, 6–9]. STM measurements can also be performed over a very wide range of temperatures, from less than 1 K to more than 1000 K [10–13]. Variable-temperature STM has been used, for example, to investigate the temperature-dependent adsorption of oxygen on Si (111)-7 × 7 [11] from room temperature down to 30 K, or the annealing behavior of epitaxial TiN(001) layers on MgO(001) at 1053 K [12].
30.3 Lateral and Spectroscopic Information
STM images can only be interpreted in terms of surface topography for surface structures with dimensions well above the atomic scale. In general, images taken in constant current mode deliver contour maps of the local density of states. If the polarity of the sample is negative, then states in the valence band are imaged. For a positive polarity of the sample the distribution of electronic states in the conduction band can be recorded. In case different chemical species are present on the surface, the image contrast is further influenced by the varying effective barrier height (work function) at different positions. As an example, Figure 30.4 shows an STM image of a Si(111) surface with one-third of a silver monolayer on top of it [14]. The high contrast between the silicon surface and the silver islands does
30.3 Lateral and Spectroscopic Information
Figure 30.4 Atomically resolved STM image of silicon covered with one-third of a monolayer of silver. Image size: 16 nm × 16 nm. From Ref. [14].
not represent the real height of the islands, but it appears exaggerated due to the lower work function in the silver regions. Such local differences in the effective barrier height can be directly imaged by modulating the tip in the vertical direction and recording dI/ds which, according to Equation (30.1) is proportional to Φ 1/2 × I. STM can also yield spectroscopic information by recording dI/dU curves at fixed positions; this technique, which is referred to as scanning tunneling spectroscopy (STS), provides the opportunity to perform electron spectroscopy with a resolution down to a single atom. Another option for obtaining spectroscopic information is the simultaneous recording of STM images at various bias voltages. This can be accomplished by performing point spectroscopy at every image point, or by modulating the bias voltage while scanning. Among published reports, the simultaneous recording of images at various bias voltages has been termed current imaging tunneling spectroscopy (CITS) [15]. Wouda et al. [16] have used the term “chemically resolved STM” for measurements revealing the chemical surface composition of a Pt50Rh50(100) single crystal with atomic resolution (Figure 30.5). The results obtained revealed a surface enrichment of Pt atoms and a preferential population of Pt atoms at the step edges of the crystal surface, without any long-range ordering effects. Olson et al. [17] have controlled the chemical contrast of STM images with Au tips chemically modified with self-assembled monolayers (SAMs) of 4-mercaptopyridine. The results obtained by these authors showed that interaction of the lone pair of the pyridine with hydrogen of surface carboxylic groups would enhance the tunneling current in these regions. Spin-polarized STM (SPSTM) [18, 19], which requires a ferromagnetic tip for its operation, can be used to probe surface magnetism and complex spin phenomena on the nanometer scale [20, 21]. Other variants of STM have involved electromagnetic radiation, whereby two basic concepts can be distinguished: (i) the measurement of photons generated in the tunneling gap by inelastic tunneling
469
470
30 Scanning Tunneling Microscopy (STM)
Figure 30.5 Atomically resolved STM image of a Pt50Rh50(100) surface with chemical contrast. The bright spots are the Rh atoms. Due to preferential segregation, the surface consists of 69% Pt atoms and 31% Rh atoms. Image size: 20 nm × 20 nm. From Ref. [16].
effects [22, 23], which is termed photon emission scanning tunneling microscopy (PESTM); and (ii) the generation of electrical currents by irradiation of the tunneling gap [24–27], which is known as photon-assisted scanning tunneling microscopy (PASTM). Previously, PESTM has been used to study the optical and structural properties of AlGaAs/GaAs quantum well systems [28], and to stimulate photon emission from the surface of a thin MgO film grown on Mo(001) [29]. Likewise, PASTM has been used to study a wide variety of photon-induced phenomena such as thermal expansion, the build-up of thermovoltages, photoninduced voltages and currents on semiconductors, and photoconduction. A review of these techniques is provided in Ref. [30]. In order to investigate buried interfaces and semiconductor heterostructures, STM and scanning tunneling spectroscopy has also been performed on crosssectional surfaces; such an approach has been termed cross-sectional STM (XSTM) [31, 32]. Finally, it should be noted that STM and STS experiments can also be performed in the inelastic tunneling regime, thus providing the opportunity to investigate vibrational modes of molecules as well as optical sample properties [33–35]. Further information on STM-related spectroscopic techniques is available elsewhere in reviews [10, 35, 36].
30.4 Applications
In similar fashion to AFM, one major advantage of STM is the fact that a wide range of scan sizes, from more than 100 μm down to the atomic level, can be covered in a single experiment. Moreover, due to the fact that the information is obtained in real space, local defects (e.g., mono-atomic defects, steps, dislocations)
30.4 Applications
can also be investigated. Clearly, this represents a major advantage compared to diffraction methods that rely on extended periodic structures, and thus provide only averaged information. As an example, the growth of Ce on a Rh(111) surface should be briefly described [37]. Atomically resolved STM images of 0.1 monolayer (ML) Ce deposited on Rh(111) at room temperature are shown in Figure 30.6. Notably, Figure 30.6a shows the clustering of the Ce atoms, where the clusters have the shape of pinwheels. No long-range order could be observed, and this was confirmed with LEED measurements. It is assumed that these clusters are pinned by defects. The image in Figure 30.6b shows that Ce atoms also form a seam along the step edges of the Rh surface; subsequently, with increasing surface coverage of up to about 0.25 ML, a (2 × 2) superstructure is formed (not shown). Upon further increase of the surface coverage, this superstructure disappears until at coverage of 1 ML only a rough unstructured surface with a few 0.5 nm deep pits can be observed (Figure 30.7). How these surfaces change upon annealing at 250 °C is clearly shown in Figure 30.8 where, on the surface with 0.05 ML Ce the clusters are seen to completely disappear while the Ce atoms become randomly embedded in the Rh surface (Figure 30.8a). At a surface coverage of 0.1 ML small areas with a local (2 × 2) superstructure and islands with the same superstructure are formed; this can be explained by the expulsion of Rh atoms from the surface due to alloy formation (Figure 30.8b). At a surface coverage of 0.25 ML the entire surface is covered by the (2 × 2) surface alloy and the (2 × 2) alloy islands (Figure 30.8c). The ability to obtain spectroscopic information leads to STM being a very valuable tool for studying surface processes on the atomic scale. A good example, highlighting several figures of merit of STM, was introduced during the early days of STM by the group of Avouris [38–40], and details of this are described in the following. Figure 30.9 shows an STM image of the unoccupied states of a Si(111)7 × 7 surface exposed to 0.2 L (Langmuir) of O2 at a temperature of 300 K. Besides the characteristic pattern of the 7 × 7 reconstruction of the Si(111) surface, pronounced dark and bright spots were observed as a result of the exposure to oxygen.
Figure 30.6 STM images of 0.1 ML Ce deposited on Rh(111) at room temperature. (a) Pinwheel-shaped clusters of Ce can be seen. Image size: 7 nm × 7 nm; (b) Seam of Ce atoms at the step edge of the Rh(111) surface. Image size: 4.5 nm × 7.5 nm. From Ref. [37].
471
472
30 Scanning Tunneling Microscopy (STM)
Figure 30.7 STM image of 1 ML Ce deposited on Rh(111) at room temperature. The inset
shows a section over a pit. Image size: 100 nm × 100 nm. From Ref. [37].
Figure 30.8 STM images of (a) 0.05 ML, (b) 0.1 ML, and (c) 0.25 ML of Ce on Rh(111) deposited at room temperature and annealed at 250 °C. Image size: 20 nm × 20 nm. From Ref. [37].
In order to clarify the mechanistic origin of these spots, tunneling spectra at various atomic sites in the 7 × 7 unit cell (Figure 30.10) have been recorded and compared to UPS spectra and theoretical results from tight-binding slab calculations for various adsorption models of oxygen. First, Figure 30.10 shows that the different sites can be clearly distinguished by the respective tunneling spectra, whereas photoemission spectra deliver only averaged information on the electronic structure of the surface. The spectra labeled A, B, and C have been obtained on sites which are not changed by the adsorption of oxygen, whereas the spectra labeled D, E, and F have been obtained on oxygen-induced dark, bright, and perturbed (gray) adatom sites, respectively. Comparisons of these spectra with theoretical calculations, and a thorough evaluation including UPS spectra for different
30.4 Applications
Figure 30.9 Atomically resolved STM image of a Si(111)-7 × 7 surface exposed to 0.2 L of O2 at 300 K. The sample bias voltage was +2 V. Dark and bright sites generated by oxygen exposure are marked with A and B, respectively [39].
Figure 30.10 (a) STM image of a region of
an O2-exposed Si(111)-7 × 7 surface. The sites labeled with A, B, and C have not been changed by exposure to oxygen. Sites D, E,
and F correspond to oxygen-induced dark, bright, and perturbed (gray) sites, respectively; (b) Tunneling spectra corresponding to the different sites [39].
473
474
30 Scanning Tunneling Microscopy (STM)
Figure 30.11 (a) STM image of surface oxide
on a Rh(111) surface. The arrow marks a position where a tip-induced defect has been produced; (b) STM image of the same area
during reduction with hydrogen at a pressure of 10−8 mbar. Images sizes: 70 nm × 70 nm. From Ref. [41].
temperatures and oxygen pressures, have led to the conclusion that the bright spots (E) are silicon atoms that have maintained their dangling bond but are perturbed by a neighboring oxygen atom inserted in the back bond of the adatom. The dark spots have been interpreted to be oxygen atoms saturating the dangling bond of a silicon atom which has another oxygen inserted in its back bond. Although this study was conducted at a rather early stage in the development of scanning tunneling spectroscopy, it provides an excellent example of the atomresolved investigation of surface chemistry with STM. A further example for studying the mechanism of a surface reaction with STM and STS is the reduction of surface oxide on Rh(111) with hydrogen at room temperature [41]. The surface before reduction is shown in Figure 30.11a, where the entire surface is covered with a three-layer oxide, which can be identified by its hexagonal moiré pattern forming a (9 × 9) superstructure on the Rh(111) surface. Moreover, monolayer islands can be observed which are formed by an excess of Rh, due to the fact that the density of Rh atoms in the oxide is approximately 20% lower than in the metal. The arrow in Figure 30.11a marks a position where a tip-induced defect has been produced. Figure 30.11b shows the same surface during reduction with hydrogen at a pressure of 10−8 mbar. It can be seen that reduction – as observed by the larger dark area – begins on the left side of the defect. In the reduced area, dark holes can be observed which are formed due to the lower density of Rh atoms in the oxide compared to the pure metal. In order to distinguish between oxide and reduced areas in large-scan images, STM dI/dV spectroscopy has been carried out, using a lock-in amplifier (Figure 30.12). Figure 30.12a shows a topographic image of the surface, while Figure 30.12b shows the result of lock-in dI/dV spectroscopy of the same area. The reduced areas appear brighter due to a lower density of states at a gap voltage of about 2 V, which can
30.4 Applications
Figure 30.12 (a) STM image of surface oxide on Rh(111) after reduction with hydrogen at a
pressure of 10−8 mbar. Image size: 1.5 μm × 1.5 μm; (b) Lock-in spectroscopy dI/dV image of the same area; (c) tunneling spectra at the positions indicated in panel (b). From Ref. [41].
475
476
30 Scanning Tunneling Microscopy (STM)
Figure 30.13 Series of in situ high-speed STM images (video-STM) of Cu dissolution on a Cu(100) surface under 0.01 M HCl and 2 μM CuSO4 at a potential of −9 mVCu/Cu(I). Image sizes: 19 nm × 19 nm. From Ref. [8].
be seen in the tunneling spectrum in Figure 30.12c. It can be observed that reduction begins almost exclusively in the stepped areas acting as defects for initializing the reduction process. In order to demonstrate also the capabilities of STM for in situ imaging under ambient conditions, the dissolution and electrodeposition of Cu on Cu(100) in HCl should be described [8]. Figure 30.13 shows a series of high-speed STM (video-STM) images of a Cu(100) surface under 0.01 M HCl and 2 μM CuSO4 at a potential of −9 mV. The whole series covers a time interval of only 1 s. The atomic structure visible in the images represents an ordered c(2 × 2) Cl adlayer; moreover, a monolayer step can be seen in the images. Dissolution can be seen to take place via the removal of Cu atoms from the monolayer step, but only along the [010] direction and not along the [001] direction. Figure 30.14 shows another series of STM images which indicate clearly that the removal of individual terraces takes place along the [010] direction, but the width of the terraces does not change along the [001] direction. This has been explained by an anisotropy at the steps introduced by the Cl adlayer. At double-layer steps, such an anisotropic dissolution behavior was not observed, which may also be explained with a model for the structure of the Cl adlayer on double-layer terraces. The dissolution of atomic rows at the steps has been found to begin predominantly at the outer corner atoms of the terraces. However, it can be seen in Figure 30.15 that the removal of such a corner atom is usually not followed immediately by a proceeding dissolution of the whole atomic row; rather, the redeposition of that atom frequently takes place. Although, in principle, STM is limited to conductive samples, it has been applied not only to metals and semiconductors but also to thin films of insulators supported on a conductive substrate. A recent review on this topic is provided in Ref. [35].
30.4 Applications
Figure 30.14 Series of in situ high-speed STM images (video-STM) of Cu dissolution on a
Cu(100) surface under 0.01 M HCl and 70 μM CuSO4 at a potential of +5 mVCu/Cu(I). Image sizes: 6.5 nm × 9.5 nm. From Ref. [8].
Figure 30.15 Series of in situ high-speed STM images (video-STM) of Cu dissolution on a
Cu(100) surface under 0.01 M HCl at a potential of −230 mVSCE. Image sizes: 2.5 nm × 2.5 nm. Fluctuation of a Cu atom at the corner of a terrace can be observed. From Ref. [8].
Figure 30.16 STM image of a Violet Lander
molecule adsorbed at a (311) step on a Cu(211) surface. The four well-resolved bumps correspond to the 3,5-(di-tert-butyl)-
phenyl (TBP) groups of the molecule. The planar polyaromatic board of the molecule is not imaged. Image size: 6.8 nm × 6.8 nm. From Ref. [45].
STM has also been used successfully to study the ordered layers of organic molecules on various substrates, and even the structure of individual molecules [42–45]. Figure 30.16 shows, for instance, an STM image of an individual dye molecule (Violet Lander) adsorbed at a (311) step on a Cu(211) surface [45]. As a further example, Figure 30.17 shows double-decker cerium alkoxyl-substituted phthalocyanine (Pc) complexes adsorbed onto highly oriented pyrolytic graphite
477
478
30 Scanning Tunneling Microscopy (STM)
Figure 30.17 STM image of (a) (C8OPc)2Ce and (b) (C12OPc)2Ce on HOPG. Image sizes:
50 nm × 50 nm. From Ref. [42].
Figure 30.18 High-resolution STM image of (a) C8OPc and (b) (C8OPc)2Ce on HOPG. Image
sizes: 10 nm × 10 nm; (c and d) Corresponding CPK (Corey, Pauling, and Koltun) models of the structures seen in panels (a) and (b), respectively. From Ref. [42].
(HOPG) [42]. In the case of a (C8OPc)2Ce complex, both fourfold and threefold symmetrical domains are observed (Figure 30.17a), whereas (C12OPc)2Ce predominantly forms threefold symmetrical structures (Figure 30.17b). At higher resolution, the intramolecular structure of the adsorbed molecules can also be resolved (Figure 30.18). Here, it can be seen that for both C8OPc and (C8OPc)2Ce, complex square-shaped molecules are arranged in a fourfold symmetry on the surface.
References
However, a different orientation of the molecules is observed for the complex (Figure 30.18b), because the top ligands are rotated approximately 45 ° against the bottom ligands adsorbed on the substrate in the same orientation as the C8OPc molecules (Figure 30.18a).
References 1 Binnig, G., Rohrer, H., Gerber, C., and Weibel, E. (1982) Phys. Rev. Lett., 49, 57–61. 2 Bonnell, D.A. (ed.) (2000) Scanning Tunneling Microscopy and Spectroscopy. Theory, Techniques, and Applications, Wiley-VCH Verlag GmbH, New York. 3 Van der Walle, G.F.A., Van Loenen, E.J., Elswijk, H.B., and Hoeven, A.J. (1991) Scanning tunneling microscopy (STM), in Analysis of Microelectronic Materials and Devices (eds M. Grasserbauer and H.W. Werner), John Wiley & Sons, Ltd, Chichester, pp. 657–678. 4 Hansma, P.K. and Tersoff, J. (1987) J. Appl. Phys., 61, R1–R23. 5 Golovchenko, J.A. (1986) Science, 232, 48–53. 6 Magnussen, O.M., Hotlos, J., Beitel, G., Kolb, D.M., and Behm, R.J. (1991) J. Vac. Sci. Technol. B, 9, 969–975. 7 Magnussen, O.M., Polewska, W., Zitzler, L., and Behm, R.J. (2002) Faraday Discuss., 121, 43–52. 8 Polewska, W., Behm, R.J., and Magnussen, O.M. (2003) Electrochim. Acta, 48, 2915–2921. 9 Tansel, T. and Magnussen, O.M. (2006) Phys. Rev. Lett., 96, 026101/1–026101/4. 10 Fischer, O., Kugler, M., Maggio-Aprile, I., and Berthod, C. (2007) Rev. Mod. Phys., 79, 353–419. 11 Mayne, A.J., Rose, F., Comtet, G., Hellner, L., and Dujardin, G. (2003) Surf. Sci., 528, 132–137. 12 Kodambaka, S., Petrova, V., Vailionis, A., Petrov, I., and Greene, J.E. (2003) Surf. Sci., 526, 85–96. 13 Morrow, S.L., Luttrell, T., Carter, A., and Batzill, M. (2009) Surf. Sci., 603, L78–L81. 14 Neddermeyer, H. (1989) Trends Anal. Chem., 8, 230–235. 15 Hamers, R.J., Tromp, R.M., and Demuth, J.E. (1986) Phys. Rev. Lett., 56, 1972–1975.
16 Wouda, P.T., Nieuwenhuys, B.E., Schmid, M., and Varga, P. (1996) Surf. Sci., 359, 17–22. 17 Olson, J.A. and Buehlmann, P. (2003) Anal. Chem., 75, 1089–1093. 18 Alvarado, S.F. (1988) J. Appl. Phys., 64, 5931–5933. 19 Wiesendanger, R., Guentherodt, H.J., Guentherodt, G., Gambino, R.J., and Ruf, R. (1990) Phys. Rev. Lett., 65, 247–250. 20 Heinze, S., Bode, M., Kubetzka, A., Pietzsch, O., Nie, X., Blugel, S., and Wiesendanger, R. (2000) Science, 288, 1805–1808. 21 Wulfhekel, W. and Kirschner, J. (2007) Annu. Rev. Mater. Res., 37, 69–91. 22 Gimzewski, J.K., Reihl, B., Coombs, J.H., and Schlittler, R.R. (1988) Z. Phys. B Condens. Matter, 72, 497–501. 23 Berndt, R., Schlittler, R.R., and Gimzewski, J.K. (1991) J. Vac. Sci. Technol. B, 9, 573–577. 24 Voelcker, M., Krieger, W., Suzuki, T., and Walther, H. (1991) J. Vac. Sci. Technol. B, 9, 541–544. 25 Kuk, Y., Becker, R.S., Silverman, P.J., and Kochanski, G.P. (1991) J. Vac. Sci. Technol. B, 9, 545–550. 26 Akari, S., Lux-Steiner, M.C., Voegt, M., Stachel, M., and Dransfeld, K. (1991) J. Vac. Sci. Technol. B, 9, 561–563. 27 Cahill, D.G. and Hamers, R.J. (1991) J. Vac. Sci. Technol. B, 9, 564–567. 28 Ropers, C., Wenderoth, M., Winking, L., Reusch, T.C.G., Erdmann, M., Ulbrich, R.G., Grochol, M., Grosse, F., Zimmermann, R., Malzer, S., and Dohler, G.H. (2007) Phys. Rev. B Condens. Matter Mater. Phys., 75, 115317/1–115317/7. 29 Benia, H.M., Myrach, P., and Nilius, N. (2008) New J. Phys., 10. 30 Grafstrom, S. (2002) J. Appl. Phys., 91, 1717–1753.
479
480
30 Scanning Tunneling Microscopy (STM) 31 Feenstra, R.M. (1999) Physica B, 273–274, 796–802. 32 Cobley, R.J., Teng, K.S., Maffeis, T.G.G., and Wilks, S.P. (2006) Surf. Sci., 600, 2857–2859. 33 Lorente, N. and Persson, M. (2000) Phys. Rev. Lett., 85, 2997–3000. 34 Ho, W. (2002) J. Chem. Phys., 117, 11033–11061. 35 Nilius, N. (2009) Surf. Sci. Rep., 64, 595–659. 36 Zandvliet, H.J.W., and van Houselt, A. (2009) Annu. Rev. Anal. Chem., 2, 37–55. 37 Napetschnig, E., Schmid, M., and Varga, P. (2004) Surf. Sci., 556, 1–10. 38 Avouris, P. (1990) J. Phys. Chem., 94, 2246–2256. 39 Avouris, P., Lyo, I.W., and Bozso, F. (1991) J. Vac. Sci. Technol. B, 9, 424–430.
40 Avouris, P. and Lyo, I.W. (1992) Am. Inst. Physics Conf. Proc., 241, 283–297. 41 Klikovits, J., Schmid, M., Gustafson, J., Mikkelsen, A., Resta, A., Lundgren, E., Andersen, J.N., and Varga, P. (2006) J. Phys. Chem. B, 110, 9966–9975. 42 Miyake, K., Fukuta, M., Asakawa, M., Hori, Y., Ikeda, T., and Shimizu, T. (2009) J. Am. Chem. Soc., 131, 17808–17813. 43 MacLeod, J.M., Ivasenko, O., Fu, C., Taerum, T., Rosei, F., and Perepichka, D.F. (2009) J. Am. Chem. Soc., 131, 16844–16850. 44 Godlewski, S., Tekiel, A., Budzioch, J., Gourdon, A., Prauzner-Bechcicki, J.S., and Szymonski, M. (2009) ChemPhysChem, 10, 3278–3284. 45 Savio, L., Gross, L., Rieder, K.-H., Gourdon, A., Joachim, C., and Moresco, F. (2006) Chem. Phys. Lett., 428, 331–337.
481
31 Scanning Near-Field Optical Microscopy (SNOM) Marc Richter and Volker Deckert
31.1 Introduction
Traditional optical microscopy has been extraordinarily popular for over 300 years, largely because it is inexpensive, fast, easy to use, and can be applied to the investigation of samples under ambient conditions. In recent years, advances in optical microscopy have been undertaken to provide high-contrast imaging of transparent samples using fluorescence, dark-field, and phase-contrast techniques [1]. However, the optical resolution of today’s microscopes is limited to approximately half the optical wavelength (e.g., ∼300 nm for visible light). This fundamental diffraction limit of resolution was postulated by Abbe [2] (Equation 31.1), and is very similar to the Rayleigh criterion: d=
λ 2πNA
(31.1)
where d is the smallest resolvable distance between two objects, λ is the wavelength of light and NA is the numerical aperture of the lens. During the 1920s, E.H. Synge proposed an approach to overcome the diffraction limit in optical microscopy by performing measurements in the optical near field of the source of the sample [3, 4]. Some decades passed until this technique became reality, however, at which time it became known as scanning near-field optical microscopy (SNOM or NSOM). Synge’s central concept and the foundations for the early SNOM methods consisted of using aperture probes featuring a subwavelength-sized hole in an opaque screen. For this, the aperture probe was brought into close vicinity to the sample to utilize near-field effects (Figure 31.1). With this arrangement the lateral resolution is determined only by the size of the light source and the distance to the sample. The near-field approach requires exquisite nanometer-range distance control between the probe and the sample. The development of scanning probe microscopy (SPM) [5–8] during the 1980s subsequently enabled the practical application of SNOM. The first modern aperture probes, as demonstrated simultaneously
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
482
31 Scanning Near-Field Optical Microscopy (SNOM)
Figure 31.1 The resolution of a traditional optical microscope is defined by the diffraction
limit. By performing measurements in the optical near-field of the sample the diffraction limit can be overcome.
by Pohl [9] and Lewis [10], were used to achieve a spatial resolution of approximately λ/20. Since then, the SNOM technique has been further developed and, by combination with fluorescence, Raman, and infrared (IR) spectroscopy, has become a powerful and versatile tool for a host of advanced analytical applications. The different approaches to near-field resolution are discussed, and applications of today’s SNOM instruments presented, in the following sections.
31.2 Instrumentation and Operation 31.2.1 Basic Set-Up
Although SNOM configurations can appear quite different, they nevertheless share common elements. A standard SNOM set-up consists of an illumination unit, a collection and adjustment module, and a detection unit (Figure 31.2). As for most applications, it is important to couple as much light as possible into the small aperture; typically, lasers are used as light sources as they can be very well focused and also allow the polarization of the light to be easily controlled. The light is coupled into the near-field aperture, and to obtain near-field resolution it is necessary to control the distance between the aperture and sample within a range of <10 nm. In most instruments this is achieved by using atomic force microscopy (AFM), although electron tunneling distance control has also been used. Today, most near-field apertures are based on tapered glass fibers, or directly on specifically manufactured AFM cantilevers. In general, the light source, optical probe and AFM module can be considered to form the illumination unit (part a in Figure 31.2). For practical reasons, the distance control module (e.g., the AFM unit) is placed in most cases on an inverted microscope which, when combined
31.2 Instrumentation and Operation
Figure 31.2 Basic SNOM arrangement. Illumination unit (a), collection unit (b), and detection module (c).
with a module for visual inspection, represents the collection unit (part b in Figure 31.2), which can have various configurations. The detection module (part c in Figure 31.2) incorporates either photomultipliers or photodiodes if monochromatic light is being investigated, or a CCD-camera coupled to a spectrograph if the spectral distribution of light is to be detected. 31.2.2 Variations of SNOM
Two major variants of optical near-field set-up exist, namely the “aperture” or the “scattering” (apertureless) modes. The principle of the aperture-type instruments has been already described above, but in the case of the scattering type near-field instruments a subwavelength-sized feature (edge, particle, etc.) acts as an active source of radiation. Most often, these probes are illuminated with a diffractionlimited laser beam, and irradiate either at a shifted wavelength or show certain field-enhancement properties, such that the resolution capabilities are very similar or even better than those of the aperture-type SNOM instruments. Hence, the choice of which set-up is to be used is determined by the application. When using aperture tips, transmission and reflection modes are equally applicable as illumination/collection modes (Figure 31.3a–c). In the first mode, the transmitted and reflected signals are collected by external optics, whereas in the second mode the illumination and collection are accomplished through the same aperture. Furthermore, photon scanning tunneling is made possible by probing the evanescent field on the surface of a prism (Figure 31.3d) [11]. Scattering-type probes can be used in a similar fashion (see Figure 31.3e and f).
483
484
31 Scanning Near-Field Optical Microscopy (SNOM)
Figure 31.3 Different SNOM configurations.
(a–d) Aperture-based SNOM. (a) Transmission mode. Illumination through an aperture, with the transmitted signal collected by external far-field optics; (b) Illumination and collection mode. Illumination and signal collected through the same aperture; (c) Reflection mode. Illumination through the aperture, with the reflected signal collected by external far-field optics; (d) Photon
scanning tunneling mode (PSTM). Total internal reflection generates an evanescent field at the surface of a prism and is collected through the aperture; (e and f) Scattering-based SNOM. (e) Scattering evanescent excitation mode; (f) Scatteringtype SNOM. Illumination and collection through the same far-field optics. In both modes, the tip creates an enhancement and is used as a subwavelength light source.
31.2.3 Scanning and Feedback Techniques
By using the scattering mode it is feasible to produce optical images of the sample with resolution similar to that obtained using AFM. By using silicon tips, several types of imaging method can be used for scanning in x and y directions to determine the topography (z-direction) of the sample. Because in contact AFM mode the sample is constantly touched by the tip, it is advisable to apply this only to hard and flat surfaces in order to avoid sample damage; hence, the various methods of noncontact AFM analysis are more popular. Early noncontact SNOM systems relied on electron tunneling as the feedback signal [9, 11], but this approach is limited to only electrically conductive samples. Consequently, for the majority of life sciences applications the so-called “tapping” or “intermittent” mode is used, where the probe oscillates vertically or horizontally over the sample, contacting the sample surface only for an instant. A change in the oscillation frequency, phase, or amplitude provides the feedback signal, which can be used to maintain the probe at a specified average distance. This type of measurement is suitable for applications under environmental conditions, or in liquids. Typically, the oscillation of the tip is measured by the deflection of a laser beam, which is focused on
31.2 Instrumentation and Operation
Figure 31.4 Scheme for laser-based AFM feedback signal detection.
the tip. This signal is then detected by a photodiode and provides the feedback input for the piezoelectric height control of the tip (Figure 31.4). This feedback system [12] relies on a broadband laser diode. When used for near-field optical detection, such an arrangement will frequently interfere with the actual measurement, but to overcome this limitation a nonoptical so-called “shear force” feedback based on a quartz tuning fork was developed [13]. In this technique, the tip is mounted on a quartz tuning fork and excited at the resonance frequency of the system. Both, the amplitude and phase of this oscillation can be monitored and used for the distance control, as the approach of the tip to the surface leads to a damped amplitude and a phase shift [14, 15]. A combination of multiple factors are responsible for this damping, including mechanical friction, adhesive forces, or contact interaction caused by a thin film of water [15–18]. Due to the kinetic energy of the oscillating tip, changes in the phase appear more rapidly than those of the amplitude [19]. Thus, it may be advantageous to use the phase shift, or a combination of phase and amplitude, for fast shear force feedback systems. 31.2.4 Tip Fabrication
Both, aperture and apertureless tips have their advantages and disadvantages. When using aperture probes the throughput is limited by the size of the aperture, whereas apertureless probes create problems for distinguishing near-field signals from the background. For both techniques, the controlled and reproducible fabrication of the tips remains the major challenge for scanning near-field microscopy. While the micro fabrication of near-field tips has been improved throughout the years [20, 21], their application depends heavily on the specific optical layout of the instrument, and they cannot be used in standard AFMs. Hence, for many experiments the probes must be fabricated to meet the demands of the given application. High-quality tips for aperture SNOM are characterized by a welldefined taper structure, large cone angles, high brightness, high optical damage threshold, a dense metal coating [22, 23], and the need for high material purity.
485
486
31 Scanning Near-Field Optical Microscopy (SNOM)
The general procedure for the fabrication of fiber-based optical probes can be divided into two main steps: (i) the fabrication of the taper structure ending in a sharp apex; and (ii) tip coating with a metal layer. 31.2.4.1 Taper Formation In order to fabricate taper structures, two common methods are used:
•
The heating and pulling method is based on the adiabatic tapering of an optical silica fiber by heating with a CO2-laser or a heating filament for controlled local melting while pulling the fiber [24, 25]. The resultant quality of the tip depends on many parameters, including temperature, heating time, dimensions of the heating area, and pulling adjustments. The advantage of this method is a smooth glass surface, which improves the quality of the subsequent metal coating. Additionally, the cone features a flat facet at the apex, which is formed when the fiber ultimately breaks; this simplifies the shaping of an aperture during evaporation. A disadvantage of this method is the difficulty to produce tips with large cone angles. Correspondingly, the optical transmission is generally lower for the given aperture size.
•
Chemical etching of optical fibers in hydrofluoric acid (HF) with an organic overlayer remains the most popular method (the Turner method [26]). This technique is self-terminating, while forming a taper along the etchant meniscus height [27]. By using different overlayers, probes with varying cone angles can be fabricated in a laboratory-scale “mass production,” and with correspondingly large transmission coefficients. One disadvantage of Turner’s etching method is that temperature variations, vibrations, and turbulences at the liquid interface during etching greatly affect the tip quality. Under these conditions microscopic surface roughness, and asymmetrical tip shapes result in poor quality metal coatings with pinholes. These fabrication defects greatly affect the near field optic (NFO) properties. To overcome these problems, tube etching was introduced [23, 28], where the coating polymer of the optical fiber is not removed (Figure 31.5a and b). This HF-resistant coating provides a jacket to encircle the fiber with etchant, and greatly increases the tolerance to environmental perturbations. In this method convective HF currents cause the formation of a taper; this results in reproducible and high-quality tips, with the large cone angles and smooth surfaces that are required for the preparation of a light-tight metal coating [29].
31.2.4.2 Coating Deposition and Aperture Formation After taper formation, evaporation can be used to coat the tips with metal (Al, Au, or Ag). Due to the small skin depth requirement in the optical region, aluminum is usually the metal of choice, although silver [30] and gold [31, 32] have also been used and show good opacity. Additionally, layers of chromium, cobalt, and nickel can be applied between the glass surface and the metal coating to significantly aid adhesion and increase the optical destruction threshold [29].
31.2 Instrumentation and Operation
Figure 31.5 (a, b) Taper formation by tube etching; (c) Evaporation of aluminum onto a tilted and rotating fiber tip.
A common technique used to form the subwavelength-sized aperture at the tip apex is based on geometrical shadowing (Figure 31.5c) [19, 25] where, by adjusting the angle of evaporation, the dimension of the metal-free aperture can be controlled. In practice, the minimal coating thickness is limited to the fact that aluminum tends to aggregate in small grains; thereby, a compromise is required between the coating thickness (>20 nm) and light throughput. As an alternative, the controlled all solid-state electrolysis (CASSE) technique can be used to form apertures with predetermined small diameters and small rims [33]. Here, a probe which is completely coated with silver is brought into feedbackcontrolled contact with amorphous silver metaphosphate-iodide (AgPO3 : AgI), and acts as the solid electrolyte while becoming demetallized. 31.2.4.3 Advanced Tip Fabrication The milling and polishing of a completely coated fiber tip, using a focused ion beam (FIB), is currently the best and also most elaborate method of producing flat and well-defined circular apertures [34–36]. Moreover, it is possible to use FIB machining to create an optical monopole antenna of controlled length and arbitrarily placed on the end face of a standard aperture probe [37]. The FIB can also be applied to generate an aperture in a scattering-type probe. These micro-machined aperture tips are similar to AFM probes and can be operated in an AFM system, thus combining the advantages of the aperture and apertureless types (i.e., using an AFM tip) [38, 39]. Both, uncoated and coated AFM tips can be used for SNOM. When coating the AFM tips, evaporation is used for vapor deposition of the metal (e.g., Ag or Au). A disadvantage of this coating method is that the size of the metal nanoparticles cannot be controlled, and this may result in variable levels of tip functionalization. To illustrate the associated drawbacks of these fabrication imperfections, consider an asymmetrically aggregated metal grain at the apex of the tip, which can cause displacement between the optical and the topographical images. Alternatively, apertures can be formed using an opaque metal screen at the apex of the tip [40],
487
488
31 Scanning Near-Field Optical Microscopy (SNOM)
Figure 31.6 (a) SEM image of an AFM tip with attached Si nanowire; (b) Magnification of the
tip region with attached nanowire with the gold droplet. Adapted from Ref. [41].
or AFM tips can be decorated with silicon nanowires on which hemispherical gold droplets are attached (Figure 31.6 [41]).
31.3 SNOM Applications 31.3.1 Fluorescence
The most frequently used spectroscopic technique in combination with SNOM is fluorescence imaging. This is especially the case for biological applications, and this has become a standard method with the potential for imaging the distribution of fluorescent labels at the single molecule level [42–44]. Without significant technical investment, it is now possible to perform single-molecule detection by applying wide-field or confocal microscopy [45] under a wide range of ambient conditions, within flow cells and even with temperatures ranging from cryogenic to ambient. Indeed, a large number of articles contain images of living cells produced by confocal microscopy [46] [47]. The disadvantage of all
31.3 SNOM Applications
far-field methods is the fact that the spatial resolution is restricted by the diffraction limit [48]. SNOM, when combined with fluorescence, can be used to overcome the diffraction limit and also to offer several other advantages. First, the background emission is reduced by minimizing the sample volume [49]. Second, SNOM provides high sensitivity for the three-dimensional (3-D) orientation and rotational dynamics of molecules [50]. Furthermore, SNOM can be used to map the distribution of fluorescent proteins on a membrane while mapping the topography, simultaneously [51]. Apart from these advantages, the required specifications such as superior quality near-field probes, tip–sample distance control, or vibration isolation render a SNOM experiments quite complex. Hence, compared to the huge number of farfield fluorescence experiments, near-field fluorescence data are comparatively rare. Nevertheless – and particularly in biological applications – SNOM has been applied to many different targets. For instance, it is possible to study fluorescent-labeled chromosomes [52], as well as DNA [50, 53, 54] and purified fluorescent proteins [55, 56]. The targeting of analytes within the sample is also feasible by using fluorescence resonance energy transfer (FRET) techniques [57]. By detecting the energy transfer between two fluorophores (Figure 31.7), FRET can be used to identify calcium [58], copper [59], pH [60, 61], cyclic adenosine monophosphate (cAMP) [62], and cyclic guanosine monophosphate (cGMP) [63]. Another technique, fluorescence correlation spectroscopy (FCS) [64, 65], can be used to characterize chemically well-defined lipid bilayer models for biomembrane research [46] and peptide modulation [66].
Figure 31.7 Energy diagram of fluorescence
resonance energy transfer (FRET). Excitation of the donor molecule to a higher energy level. The energy of the donor is nonradia-
tively transferred to an acceptor by FRET, and excites the acceptor to a higher energy level. Finally, the energy is released by the emission of light.
489
490
31 Scanning Near-Field Optical Microscopy (SNOM)
31.3.2 Near-Field Raman Spectroscopy
Scanning near-field optics can also be used for vibrational spectroscopy that, in general, provides molecular specificity without the need for dye labels. The first demonstration of the combination of near-field optics and Raman spectroscopy was presented by Jahnke et al. and Smith et al. [67, 68]. Although the technique suffers from low Raman scattering cross-sections, it is possible to obtain chemical information even from unknown substances. Near-field Raman has also been used for several medical applications, including the detection of cancerous tissues, osteoporosis, and to analyze adamantine [69]. The particularly high spatial resolution orthogonal to the aperture SNOM/Raman technique (“depth resolution”) has been utilized to investigate liquid–liquid interfaces with submicrometer resolution [70, 71]. A common method to increase Raman scattering uses rough noble metal surfaces. This enhancement effect was first noted by Fleischmann et al. [72], and confirmed by other groups [73, 74]. Today, surface-enhanced Raman scattering (SERS) is in common use (for details, see Refs [19, 75–79]). As SNOM-SERS can achieve detection limits down to the single-molecule level [79–82], the scope of applications is considerable, ranging from classical chemical analysis [83] to biological [80, 84, 85], and medical applications [86]. The main benefits of SERS and surface enhanced resonance Raman scattering (SERRS), which is an extended application of SERS [87], are the low detection limit and wide range of applications. Despite these advantages, the SERS of surfaces has suffered from intrinsic inhomogeneities of the metal surface (see Figure 31.8) of the SERS substrates, which has resulted in variability of the signal enhancement [77]. One unique SERS-based technique is that of tip-enhanced Raman scattering (TERS) [89], where the signal enhancement area is limited to the end of a scanning tip, in order to increase the measurement reproducibility and contrast [90–92]. This method uses the enhancement effect of a single noble metal nanoparticle, as discovered by Nie and Emory [93]. The enhancement factors may be almost as high as in SERS, typically 106 to 108. The main advantage of TERS is that the signal enhancement is generated always by the same particle; consequently, different measurements performed using the same probe are comparable with respect to the signal enhancement. Currently, TERS is used for the structural analysis of inorganic ions [94], inorganic films [95, 96], and organic samples [97–101]. A combination of TERS with single-walled carbon nanotube (SWNT)-coated tips (Figure 31.9), as demonstrated by Hartschuh and coworkers [90, 102], can be used for the analysis of inorganic and organic samples [103] and even proteins [104]. Here, the use of nanotubes for sample storage enables a spatial resolution better than 30 nm. Nowadays, TERS has reached a spatial resolution even below 10 nm according to the small size of the probe [90, 92, 105–107]. This resolution comes close to the requirements for direct DNA sequencing [108]. Another interesting application of TERS is the use of dyes to analyze inorganic samples [109–111]. In this case, the dye acts as an additional feature to enhance and enforce the Raman signal.
31.3 SNOM Applications
Figure 31.8 (a) AFM image of an inhomogeneous silver film. The height varies from back-
ground (black) to 65 nm (white); (b) Selected AFM line scan of the same film as indicated in panel (a). Adapted from Ref. [88].
During recent years, the spatial resolution has greatly increased to enable the detection of single nanocrystal DNA-base molecules [112]. In fact, a differentiation between DNA bases and also between desoxynucleotides is possible [85]. Beyond these applications, TERS can be further applied to study the properties of biological cells [113–117]. Another promising SNOM application is its combination with coherent antistokes Raman scattering (CARS) [118–120]. In fact, a combination of tipbased signal enhancement and CARS (TE-CARS) microscopy can lead to dramatic reductions in the acquisition times while maintaining a high lateral resolution [121–123]. The fact that TE-CARS produces a high signal-to-noise ratio (SNR) obviates the need for a high-performance detector and, by varying the excitation
491
492
31 Scanning Near-Field Optical Microscopy (SNOM)
Figure 31.9 (a) Near-field Raman (G′ band), and (b) topographic images of single-walled carbon nanotubes (SWNTs). Line scans taken along the indicated dashed lines in the Raman (c) and topographic (d) images. Illustration courtesy of Prof. A. Hartschuh [90].
wavelength, specific vibrational states in several molecular formations can be generated. The spatial resolution of about 15 nm can also be used for singlemolecule detection and to visualize DNA network structures [121]. 31.3.3 SNOM-IR-Spectroscopy
Infrared (IR) spectroscopy is a widely used method for the identification of molecules, based on their vibrational spectra. Combinations of IR spectroscopy and SNOM have also been developed, and have included surface-enhanced infrared absorption (SEIRA) [124, 125] and tip-enhanced infrared absorption (TEIRA) [126]. By using noble metal nanoparticles (Ag, Au [127]), an enhanced spatial resolution of λ/600 can be achieved [128] and the monolayer-sensitive IR imaging of DNA stripes is possible [129]. Although several combinations of IR and SNOM have been developed [130], the technique remains problematic due to adsorption effects when using silica fibers. A further development of IR/SNOM has been scanning near-field infrared microscopy (SNIM), as developed by Platkov and coworkers [131]. This method employs an etched silver halide fiber (AgClBr) as a probe with
References
a wide transparency window in the mid-IR region. The development of SNIM continues to pave the way for IR to reach spatial resolution at the nanometer scale. In summary, these near-field techniques provide a rich toolbox for extremely high lateral resolution and sensitive molecular spectroscopy.
31.4 Outlook
Today, SNOM techniques represent powerful tools for ultra-high-resolution investigations where the capabilities of standard optical microscopes fail. These techniques have the added advantage that topographic and spectroscopic information can be obtained simultaneously. With SNOM being used today in a wide range of life sciences applications, the high spatial resolution imaging capabilities have been exploited to analyze the structure and function of proteins, and even to sequence DNA [85] and RNA [108]. Clearly, further technological improvements are to be expected in the near future, and even more detailed investigations can be anticipated. For example, Wilson and colleagues have already described the detection of cyanobacteria within geological samples via SERS [132], whilst with continuously improving positioning technology, protein detection in extraterrestrial environments might be imagined. Currently, the primary disadvantage of all SNOM techniques is that they are restricted to the surface of a sample. However, the combination of SNOM with spatially offset Raman spectroscopy (SORS) represents an interesting concept [133, 134] whereby spatially resolved subsurface analysis should at least be feasible (see also Section 24.7.4). In addition to research applications, the use of SNOM within an industrial environment is also envisaged. The ever-decreasing size of components in the semiconductor industry requires ultrahigh-resolution probing techniques to obtain chemical defect information beyond the capabilities of SEM analysis. In this case, near-field microscopy might offer an alternative approach, in combination with other established analytical tools.
References 1 Murphy, D.B. (2001) Fundamentals of Light Microscopy and Electronic Imaging, John Wiley & Sons, Inc., New York. 2 Abbe, E. (1873) Archiv für mikroskopische Anatomie, 9, 413–468. 3 Synge, E. (1928) Philos. Mag., 6, 356–362. 4 Synge, E. (1931) Philos. Mag., 11, 65–81. 5 Binnig, G., Quate, C., and Gerber, C. (1986) Phys. Rev. Lett., 56, 930–933.
6 Binnig, G. and Rohrer, H. (1982) Helv. Phys. Acta, 55, 726–735. 7 Binnig, G. and Rohrer, H. (1983) Surf. Sci., 126, 236–244. 8 Binnig, G. and Rohrer, H. (1986) IBM J. Res. Dev., 30, 355–369. 9 Pohl, D., Denk, W., and Lanz, M. (1984) Appl. Phys. Lett., 44, 651–653. 10 Lewis, A., Isaacson, M., Haroontunian, A., and Muray, A. (1983) Ultramicroscopy, 13, 227–231.
493
494
31 Scanning Near-Field Optical Microscopy (SNOM) 11 Ferrell, T., Warmack, R., and Reddicke, R. (1991) United States Patent 5018865. 12 Meyer, G. and Amer, N. (1988) Appl. Phys. Lett., 53, 1045–1047. 13 Karrai, K. and Grober, R. (1995) Ultramicroscopy, 61, 197–205. 14 Karrai, K. and Grober, R. (1995) Appl. Phys. Lett., 66, 1842–1844. 15 Toledo-Crow, R., Yang, P., Chen, Y., and Vaeziravani, M. (1992) Appl. Phys. Lett., 60, 2957–2959. 16 Davy, S., Spajer, M., and Courjon, D. (1998) Appl. Phys. Lett., 73, 2594–2596. 17 Durkan, C. and Shvets, I. (1996) J. Appl. Phys., 80, 5659–5664. 18 Okajima, T. and Hirotsu, S. (1997) Appl. Phys. Lett., 71, 545–547. 19 Hecht, B., Sick, B., Wild, U., Deckert, V., Zenobi, R., Martin, O., and Pohl, D. (2000) J. Chem. Phys., 112, 7761–7774. 20 Munster, S., Werner, S., Mihalcea, C., Scholz, W., and Oesterschulze, E. (1997) J. Microsc. Oxf., 186, 17–22. 21 Noell, W. (1997) Appl. Phys. Lett., 70, 1236–1238. 22 Burgos, P., Lu, Z., Ianoul, A., Hnatovsky, C., Viriot, M., Johnston, L., and Taylor, R. (2003) J. Microsc. Oxf., 211, 37–47. 23 Stöckle, R., Fokas, C., Deckert, V., Zenobi, R., Sick, B., Hecht, B., and Wild, U. (1999) Appl. Phys. Lett., 75, 160–162. 24 Betzig, E., Trautman, J., Harris, T., Weiner, J., and Kostelak, R. (1991) Science, 251, 1468–1470. 25 Valaskovic, G.A. (1995) Appl. Opt., 34, 1215–1228. 26 Turner, D. (1984) United States Patent 4,469,554. 27 Hoffmann, P., Dutoit, B., and Salathe, R. (1995) Ultramicroscopy, 61, 165–170. 28 Lambelet, P. (1998) Appl. Opt., 37, 7289–7292. 29 Stöckle, R., Schaller, N., Deckert, V., Fokas, C., and Zenobi, R. (1999) J. Microsc., 194, 378–382. 30 Deckert, V., Zeisel, D., Zenobi, R., and Vo-Dinh, T. (1998) Anal. Chem., 70, 2646–2650. 31 Hosaka, S., Hosoki, S., Hasegawa, T., Koyanagi, H., Shintani, T., and Miyamoto, M. (1995) J. Vac. Sci. Technol. B, 13, 2813–2818.
32 Kalkbrenner, T., Ramstein, M., Mlynek, J., and Sandoghdar, V. (2001) J. Microsc. Oxf., 202, 72–76. 33 Bouhelier, A., Toquant, J., Tamaru, H., Guntherodt, H., Pohl, D., and Schider, G. (2001) Appl. Phys. Lett., 79, 683–685. 34 Lacoste, T., Huser, T., Prioli, R., and Heinzelmann, H. (1998) Ultramicroscopy, 71, 333–340. 35 Oshikane, Y., Hara, S., Matsuda, T., Inoue, H., Nakano, M., and Kataoka, T. (2008) Surf. Interface Anal., 40, 1054–1058. 36 Veerman, J., Otter, A., Kuipers, L., and van Hulst, N. (1998) Appl. Phys. Lett., 72, 3115–3117. 37 Taminiau, T., Segerink, F., Moerland, R., Kuipers, L., and van Hulst, N. (2007) J. Opt. A Pure Appl. Opt., 9, S315–S321. 38 Grabiec, P., Radojewski, J., Zaborowski, M., Domanski, K., Schenkel, T., and Rangelow, I. (2004) J. Vac. Sci. Technol. B, 22, 16–21. 39 Radojewski, J. and Grabiec, P. (2003) Mater. Sci. Pol., 21, 321–332. 40 Kwak, A.M., Onuta, T.-D., Amarie, D., Potyrailo, R., Stein, B., Jacobson, S., Schaich, W.L., and Dragnea, B. (2004) J. Phys. Chem. B, 108, 13607–13612. 41 Becker, M., Sivakov, V., Andra, G., Geiger, R., Schreiber, J., Hoffmann, S., Michler, J., Milenin, A., Werner, P., and Christiansen, S. (2007) Nano Lett., 7, 75–80. 42 Betzig, E. and Chichester, R. (1993) Science, 262, 1422–1425. 43 Trabesinger, W., Kramer, A., Kreiter, M., Hecht, B., and Wild, U. (2003) J. Microsc. Oxf., 209, 249–253. 44 van Hulst, N., Veerman, J., GarciaParajo, M., and Kuipers, L. (2000) J. Chem. Phys., 112, 7799–7810. 45 Pawley, J. (1995) Handbook of Biological Confocal Microscopy, Springer. 46 Korlach, J., Schwille, P., Webb, W., and Feigenson, G. (1999) Proc. Natl Acad. Sci. USA, 96, 8461–8466. 47 Chan, J., Fore, S., Wachsman-Hogiu, S., and Huser, T. (2008) Laser Photonics Rev., 2, 325. 48 Laurence, T. and Weiss, S. (2003) Science, 299, 667–668. 49 de Lange, F., Cambi, A., Huijbens, R., de Bakker, B., Rensen, W., Garcia-
References
50
51 52
53
54 55
56
57 58
59 60
61 62
63
64 65 66
Parajo, M., van Hulst, N., and Figdor, C. (2001) J. Cell Sci., 114, 4153–4160. Garcia-Parajo, M., Veerman, J., Ruiter, A., and van Hulst, N. (1998) Ultramicroscopy, 71, 311–319. Zhong, L., Liao, W., Wang, X., and Cai, J. (2008) Colloids Surf. A, 313, 642–646. Moers, M., Kalle, W., Ruiter, A., Wiegant, J., Raap, A., Greve, J., De Grooth, B., and van Hulst, N. (1996) J. Microsc., 182, 40–45. Ha, T., Enderle, T., Ogletree, D., Chemla, D., Selvin, P., and Weiss, S. (1996) Proc. Natl Acad. Sci. USA, 93, 6264–6268. Spring, B. and Clegg, R. (2007) J. Phys. Chem. B, 111, 10040–10052. Garcia-Parajo, M., Segers-Nolten, G., Veerman, J., Greve, J., and van Hulst, N. (2000) Proc. Natl Acad. Sci. USA, 97, 7237–7242. Garcia-Parajo, M., Veerman, J., Segers-Nolten, G., De Grooth, B., Greve, J., and van Hulst, N. (1999) Cytometry, 36, 239–246. Vickery, S. and Dunn, R. (1999) Biophys. J., 76, 1812–1818. Miyawaki, A., Llopis, J., Heim, R., McCaffery, J., Adams, J., Ikura, M., and Tsien, R. (1997) Nature, 388, 882–887. White, B. and Holcombe, J. (2007) Talanta, 71, 2015–2020. Awaji, T., Hirasawa, A., Shirakawa, H., Tsujimoto, G., and Miyazaki, S. (2001) Biochem. Biophys. Res. Commun., 289, 457–462. Hong, S., Ahn, C., Huh, J., and Jo, W. (2006) Macromolecules, 39, 7694–7700. Zaccolo, M., De Giorgi, F., Cho, C., Feng, L., Knapp, T., Negulescu, P., Taylor, S., Tsien, R., and Pozzan, T. (2000) Nat. Cell Biol., 2, 25–29. Honda, A., Adams, S., Sawyer, C., Lev-Ram, V., Tsien, R., and Dostmann, W. (2001) Proc. Natl Acad. Sci. USA, 98, 2437–2442. Kahya, N. and Schwille, P. (2006) Mol. Membr. Biol., 23, 29–39. Widengren, J., Mets, U., and Rigler, R. (1995) J. Phys. Chem., 99, 13368–13379. Chowdhury, P., Wang, W., Lavender, S., Bunagan, M., Klemke, J., Tang, J., Saven, J., Cooperman, B., and Gai, F. (2007) J. Mol. Biol., 369, 462–473.
67 Jahncke, C., Paesler, M., and Hallen, H. (1995) Appl. Phys. Lett., 67, 2483–2485. 68 Smith, D., Webster, S., Ayad, M., Evans, S., Fogherty, D., and Batchelder, D. (1995) Ultramicroscopy, 61, 247–252. 69 Griffiths, J. (2007) Anal. Chem., 79, 3975–3978. 70 De Serio, M., Bader, A., Heule, M., Zenobi, R., and Deckert, V. (2003) Chem. Phys. Lett., 380, 47–53. 71 De Serio, M., Mohapatra, H., Zenobi, R., and Deckert, V. (2006) Chem. Phys. Lett., 417, 452–456. 72 Fleischmann, M., Hendra, P., and McQuillan, A.J. (1974) Chem. Phys. Lett., 26, 163–166. 73 Albrecht, M. and Creighton, J. (1977) J. Am. Chem. Soc., 99, 5215–5217. 74 Jeanmaire, D. and Vanduyne, R. (1977) J. Electroanal. Chem., 84, 1–20. 75 Bahns, J., Imre, A., Vlasko-Vlasov, V., Pearson, J., Hiller, J., Chen, L., and Welp, U. (2007) Appl. Phys. Lett., 91, 081104-1–081104-3. 76 Heinzelmann, H. and Pohl, D. (1994) Appl. Phys. A Mater. Sci. Processing, 59, 89–101. 77 Micic, M., Klymyshyn, N., and Lu, H. (2004) J. Phys. Chem. B, 108, 2939–2947. 78 Renger, J., Grafström, S., Eng, L., and Deckert, V. (2004) J. Opt. Soc. Am. A, 21, 1362–1367. 79 Etchegoin, P.G., and Le Ru, E.C. (2008) Phys. Chem. Chem. Phys., 10, 6079–6089. 80 Hering, K., Cialla, D., Ackermann, K., Dorfer, T., Moller, R., Schneidewind, H., Mattheis, R., Fritzsche, W., Rosch, P., and Popp, J. (2008) Anal. Bioanal. Chem., 390, 113–124. 81 Kneipp, J., Kneipp, H., and Kneipp, K. (2008) Chem. Soc. Rev., 37, 1052–1060. 82 Kneipp, K., Wang, Y., Kneipp, H., Perelman, L., Itzkan, I., Dasari, R., and Feld, M. (1997) Phys. Rev. Lett., 78, 1667–1670. 83 Fokas, C. and Deckert, V. (2002) Appl. Spectrosc., 56, 192–199. 84 Hudson, S.D. and Chumanov, G. (2009) Anal. Bioanal. Chem., 394, 679–686. 85 Rasmussen, A. and Deckert, V. (2006) J. Raman Spectrosc., 37, 311–317.
495
496
31 Scanning Near-Field Optical Microscopy (SNOM) 86 Shah, N., Lyandres, O., Walsh, J., Glucksberg, M., and Duyne, R. (2007) Anal. Chem., 79, 6927–6932. 87 Douglas, P., McCarney, K., Graham, D., and Smith, W. (2007) Analyst, 132, 865–867. 88 Vogel, E., Kiefer, W., Deckert, V., and Zeisel, D. (1998) J. Raman Spectrosc., 29, 693–702. 89 Bailo, E. and Deckert, V. (2008) Chem. Soc. Rev., 37, 921. 90 Hartschuh, A., Sanchez, E., Xie, X., and Novotny, L. (2003) Phys. Rev. Lett., 90, 095503-1–095503-4. 91 Rasmussen, A., and Deckert, V. (2005) Anal. Bioanal. Chem., 381, 165–172. 92 Stöckle, R., Suh, Y., Deckert, V., and Zenobi, R. (2000) Chem. Phys. Lett., 318, 131–136. 93 Nie, S. and Emory, S. (1997) Science, 275, 1102–1106. 94 Pettinger, B., Picardi, G., Schuster, R., and Ertl, G. (2003) J. Electroanal. Chem., 554, 293–299. 95 Matsui, R., Verma, P., Ichimura, T., Inouye, Y., and Kawata, S. (2007) Appl. Phys. Lett., 90, 061906-1–061906-3. 96 Tarun, A., Hayazawa, N., Motohashi, M., and Kawata, S. (2008) Rev. Sci. Instrum., 79, 013706. 97 Chen, J., Yang, W., Dick, K., Deppert, K., Xu, H., Samuelson, L., and Xu, H. (2008) Appl. Phys. Lett.., 92, 093110. 98 Deckert-Gaudig, T., Bailo, E., and Deckert, V. (2009) Phys. Chem. Chem. Phys., 11, 7360–7362. 99 Deckert-Gaudig, T. and Deckert, V. (2009) J. Raman Spectrosc., 40, 1446–1451. 100 Deckert-Gaudig, T., Rauls, E., and Deckert, V. (2010) J. Phys. Chem. C, 114, 7412–7420. 101 Verma, P., Yamada, K., Watanabe, H., Inouye, Y., and Kawata, S. (2006) Phys. Rev. B, 73, 045416-1–045416-6. 102 Hartschuh, A., Pedrosa, H., Novotny, L., and Krauss, T. (2003) Science, 301, 1354–1356. 103 Saito, Y., Yanagi, K., Hayazawa, N., Ishitobi, H., Ono, A., Kataura, H., and Kawata, S. (2006) Jpn. J. Appl. Phys. Part 1, 45, 9286–9289. 104 Zhang, Y., Kanungo, M., Ho, A., Freimuth, P., van der Lelie, D., Chen,
105
106
107
108 109
110
111
112
113
114
115
116
117
118 119
120
121
M., Khamis, S., Datta, S., Johnson, A., Misewich, J., and Wong, S. (2007) Nano Lett., 7, 3086–3091. Hayazawa, N., Inouye, Y., Sekkat, Z., and Kawata, S. (2000) Opt. Commun., 183, 333–336. Pettinger, B., Picardi, G., Schuster, R., and Ertl, G. (2002) Single Mol., 5, 285–294. Yeo, B.-S., Maedler, S., Schmid, T., Zhang, W., and Zenobi, R. (2008) J. Phys. Chem. C, 112, 4867–4873. Bailo, E. and Deckert, V. (2008) Angew. Chem. Int. Ed., 47, 1–5. Pan, D., Klymyshyn, N., Hu, D., and Lu, H. (2006) Appl. Phys. Lett., 88, 093212-1–093121-3. Pettinger, B., Ren, B., Picardi, G., Schuster, R., and Ertl, G. (2004) Phys. Rev. Lett., 92, 096101-1–096010-4. Pettinger, B., Ren, B., Picardi, G., Schuster, R., and Ertl, G. (2005) J. Raman Spectrosc., 36, 541–550. Watanabe, H., Ishida, Y., Hayazawa, N., Inouye, Y., and Kawata, S. (2004) Phys. Rev. B, 69, 155418-1–155418-11. Budich, C., Neugebauer, U., Popp, J., and Deckert, V. (2008) J. Microsc., 229, 533–539. Cialla, D., Deckert-Gaudig, T., Budich, C., Laue, M., Moeller, R., Naumann, D., Deckert, V., and Popp, J. (2009) J. Raman Spectrosc., 40, 240–243. Neugebauer, U., Rosch, P., Schmitt, M., Popp, J., Julien, C., Rasmussen, A., Budich, C., and Deckert, V. (2006) ChemPhysChem, 7, 1428–1430. Neugebauer, U., Schmid, U., Baumann, K., Ziebuhr, W., Kozitskaya, S., Deckert, V., Schmitt, M., and Popp, J. (2007) ChemPhysChem, 8, 124–137. Schmid, T., Messmer, A., Yeo, B.-S., Zhang, W., and Zenobi, R. (2008) Anal. Bioanal. Chem., 10, 1907–1916. Dudovich, N., Oron, D., and Silberberg, Y. (2002) Nature, 418, 512–514. Kachynski, A., Kuzmin, A., Prasad, P., and Smalyukh, I.I. (2007) Appl. Phys. Lett., 91, 151905-1–151905-3. Tolles, W., Nibler, J., McDonald, J., and Harvey, A. (1977) Appl. Spectrosc., 31, 253–271. Ichimura, T., Hayazawa, N., Hashimoto, M., Inouye, Y., and
References
122
123 124
125
126 127
128
Kawata, S. (2004) Phys. Rev. Lett., 92, 220801. Kawata, S., Ichimura, T., Hayazawa, N., Inouye, Y., and Hashimoto, M. (2004) J. Nonlinear Opt. Phys. Mater., 13, 593–599. Kawata, S. and Verma, P. (2006) Chimia, 60, A770–A776. Kundu, J., Le, F., Nordlander, P., and Halas, N.J. (2008) Chem. Phys. Lett., 452, 115–119. Roseler, A. and Korte, E. (1998) Fresenius J. Anal. Chem., 362, 51–57. Knoll, B. and Keilmann, F. (1999) Nature, 399, 134–137. Hillenbrand, R., Taubner, T., and Keilmann, F. (2002) Nature, 418, 159–162. Lahrech, A., Bachelot, R., Gleyzes, P., and Boccara, A. (1996) Opt. Lett., 21, 1315–1317.
129 Akhremitchev, B., Sun, Y., Stebounova, L., and Walker, G. (2002) Langmuir, 18, 5325–5328. 130 Cricenti, A., Generosi, R., Luce, M., Perfetti, P., Sanghera, J.S., Aggarwa, I.D., Tolk, N.H., Vobornik, D., Margaritondo, G., Piston, D.W., Manni, V., Grimaldi, S., Lisi, A., and Rieti, S. (2008) J. Microsc. Oxf., 229, 551–554. 131 Platkov, M., Tsun, A., Nagli, L., and Katzir, A. (2006) Rev. Sci. Instrum., 77, 126103-1–126103-4. 132 Wilson, R., Monoghan, P., Bowden, S., Parnell, J., and Cooper, J. (2007) Anal. Chem., 79, 7036–7041. 133 Macleod, N.A., Goodship, A., Parker, A.W., and Matousek, P. (2008) Anal. Chem., 80, 8146–8152. 134 Matousek, P., Clark, I., Draper, E., Morris, M., Goodship, A., Everall, N., Towrie, M., Finney, W., and Parker, A. (2005) Appl. Spectrosc., 59, 393–400.
497
499
Appendices
501
Appendix A
Summary and Comparison of Techniques
Important surface and thin film analytical techniques are summarized and compared in Table A.1.
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
502
Typical figures of merit of techniques for surface and thin film analysis.
Technique
Primary probe
Elemental range
Type of information
Depth of information
Lateral resolution
Sensitivity (atom%)
Ease of quantification
Insulator analysis
Destructive
UHV environment
XPS
X-ray photons (Al or Mg Kα)
All except H, He
Elemental, chemical
4–8 monolayers
100 μm (standard), 3 μm (imaging)
0.1
Good
Yes
Sometimes
Yes
UPS
UV photons (He I, He II)
(Adsorbed molecules)
Chemical (valence band)
2–10 monolayers
1 mm
1
NA
Yes
Sometimes
Yes
AES
e− (3–10 keV)
All except H, He
Elemental, (chemical)
2–6 monolayers
20 nm
0.2
Moderate
No
Often
Yes
SAM
e− (20–50 keV)
All except H, He
Elemental, (chemical)
2–6 monolayers
5 nm
0.2
Moderate
No
Often
Yes
EELS
e− (60– 1000 keV)
All (H, He only hardly)
Elemental, chemical
(None)
1 nm
0.1–1
Moderate
Yes
Yes
No
EFTEM
e− (60– 1000 keV)
All except H, He
Elemental, chemical
(None)
≤1 nm
0.1–1
Moderate
Yes
Yes
No
LEED
e−
All
Crystallographic
≈5 monolayers
0.01 nm
NA
NA
No
No
Yes
SSIMS
Ar+, Ga+, and other ions
All
Chemical
1–2 monolayers
50 nm
10−6–10−2
Poor
(Yes)a)
(yes)
Yes
Appendix A Summary and Comparison of Techniques
Table A.1
Ar+, Ga+, and other ions
All
Chemical
1–2 monolayers
50 μm 20 nm (imaging)
10−7–10−1
Poor
(Yes)a)
Yes
Yes
SNMS e−-beam
Ar+, Cs+
All
Elemental
1–2 monolayers
5 μm
10−2–1
Moderateb)
Yes (with rf)
Yes
Yes
SNMS plasma
Ar+, Kr+
All
Elemental
1–2 monolayers
1 mm
10−3–10−1
Moderateb)
Yes
Yes
Yes
SNMS laser
Ar+, Ga+
All
Chemical, elemental
1–2 monolayers
50 nm
10−7–1
Moderateb)
Yes
Yes
Yes
RBS
H+, He+
High Z on low Z
Elemental, depth distribution
1 monolayer - 2 μm
5 μm– 1 mm
10−6
Good
Yes
No
No
LEIS
He+, Ne+
All except H, He
Elemental
1–2 monolayers
≈ 1 mm
10−4
Good
Yes
Not with He+
Yes
ERDA
He+ – Au+
Low Z
Elemental, Depth distribution
< μm
1 mm
1–10−4
Good
Yes
Yes
No
NRA
H+, D+, N+
Low Z
Isotopic, Depth distribution
< μm
10–100 μm
1–10−4
Good
Yes
Yes
No
FIM
High field
Conducting samples
Topographic
1 monolayer
1 atom
(1 atom)
NA
No
Yes
Yes
AP
High field
Conducting samples
Topographic
1 monolayer
1 atom
1 atom
Good
No
Yes
Yes
(Continued)
Appendix A Summary and Comparison of Techniques
DSIMS
503
504
Technique
Primary probe
TXRF
X-ray photons
EDXS
e− (few keV1000 keV)
GAXRD
Elemental range
Type of information
Depth of information
Lateral resolution
Sensitivity (atom%)
Ease of quantification
Insulator analysis
Destructive
UHV environment
Elemental
1 monolayer
(none)
10−4–10−1
Good
Yes
No
No
All Z ≥ 5
Elemental, (chemical)
(None)
≈ 5 nm
0.1–1
Good
Yes
Yes
No
X-ray photons
All
Crystalline structure
Monolayers to microns
ca. 5 mm
NA
NA
Yes
No
No
GXRR
X-ray photons
All
Thickness Density Roughness
0–400 nm
ca. 1 cm
NA
NA
Yes
No
No
ReflEXAFS
X-ray photons
All
Local atomic environment
ca. 2 nm
ca. 1 cm
10−2–10−4
NA
Yes
No
No
GD-OES
A+ (Ar+)
All
Elemental, depth profile
10 nm
none
10−6
Moderate
Yes
Yes
No
LA-OES
UV-Vis photon
All
Elemental, (depth profile)
≈0.1–1 μm
≈10 μm
10−3–10−2
Moderate
Yes
Yes
No
IBSCA
Ar+ (2–20 keV)
All
Elemental
0.3 nm
100 μm
10−4
Moderate
Yes
Yes
Yes
RAIRS
IR photons
Adsorbed molecules
Chemical (vibrational)
1–2 monolayers
≈1 mm
10−4
NA
No
No
No
Appendix A Summary and Comparison of Techniques
Table A.1 (Continued)
Laser photons
Adsorbed molecules
Chemical (vibrational)
10–100 nm
≈ 0.5 mm
10−6
Moderate
Yes
No
Yes
ELL
UV/Vis/IR photons
Index of reflection
Chemical
1 nm
≈ 1 mm
1 monolayer
Poor
(Yes)
No
No
SFG
Vis/IR laser photons
Adsorbed molecules
Chemical (vibrational)
1 monolayer
≈ 1 μm
1 monolayer
Moderate
Yes
No
No
AFM
Mechanical force
All
Topographic
1 monolayer
1 atom
(1 atom)
NA
Yes
No
No
STM STS
High field
All
Topographic, chemical
1 monolayer
1 atom
(1 atom)
NA
No
No
No
SNOM
Vis/IR photons
All
Topographic, chemical
1 monolayer
≈ 10 nm
(1 molecule) when fluorescent
NA
Yes
No
No
NA = Not applicable. a) With e−-charge compensation. b) Good with matrix-matched standards.
Appendix A Summary and Comparison of Techniques
SERS
505
507
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Many companies supply surface and thin-film analytical equipment, a few of them provide several techniques, but most specialize in just one or two. Only suppliers of the equipment that is discussed in more detail are included in the list below. It should be noted that, as the market in this field is rapidly changing, the authors cannot guarantee completeness, and errors cannot be excluded.
Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
ABB Bomem Inc. 585 Charest East, Suite 300 Québec, QC G1K 9H4 Canada www.abb.com/analytical
–
–
RAIRS
–
Accurion GmbH Stresemannstraße30 37079 Göttingen Germany www.accurion.com
–
–
ELL
–
Agilent Technologies Inc. 5301 Stevens Creek Blvd. Santa Clara, CA 95051 USA www.agilent.com
–
–
–
AFM
(Continued)
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
508
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
AIST-NT Inc. 353 Bel Marin Keys Blvd Suite 7-8 Novato, CA 94949 USA www.aist-nt.com
–
–
Raman
STM, AFM
AMETEK Inc. 37 N. Valley Road, Building 4 P.O. Box 1764 Paoli, PA 19301 USA www.ametek.com
–
SIMS
EDXS, WDXS, XRF
–
Angstrom Advanced Inc. 50 Braintree Hill Park Suite 201 Braintree, MA 02184 USA www.angstromadvanced.com
–
–
ELL
–
Asylum Research 6310 Hollister Ave Santa Barbara, CA 93117 USA www.asylumresearch.com
–
–
–
AFM
Atomic Force F&E GmbH Hauptstraße 161 68259 Mannheim Germany www.atomicforce.com
–
–
–
AFM
Beaglehole Instruments Ltd 32 Salamanca Road Kelburn Wellington, 6012 New Zealand www.beaglehole.com
–
–
ELL
–
BioTools Inc. 17546 Bee Line Highway (SR 710) Jupiter, FL 33458 USA www.btools.com
–
–
Raman
–
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
Bourevestnik Inc. Malookhtinsky pr., 68 St Petersburg, 195112 Russia www.bourevestnik.com
–
–
GAXRDGXRR
–
BRUKER AXS Inc. 5465 East Cheryl Parkway Madison, WI 53711–5373 USA www.bruker-axs.com
–
–
EDXS, TXRF, XRD
STM, AFM
Bruker Optics Inc. 19 Fortune Drive Manning Park Billerica, MA 01821–3991 USA www.bruker.com
–
–
RAIRS, Raman
–
CAMECA SAS 29, quai des Grésillons 92622 Gennevilliers Cedex, France www.cameca.com
–
SIMS, FIM, AP
–
–
Carl Zeiss NTS GmbH Carl-Zeiss-Straße 56 73447 Oberkochen Germany www.smt.zeiss.com/leo
EELS, EFTEM
–
–
–
CETAC Technologies 14306 Industrial Rd Omaha, NE 68144 USA www.cetac.com
–
–
Laser ablation
–
ChemIcon Inc. 7301 Penn Avenue Pittsburgh, PA 15208 USA www.chemimage.com
–
–
RAIRS, Raman
–
Coherent Inc. 5100 Patrick Henry Drive Santa Clara, CA 95054 USA www.coherent.com
–
–
Laser ablation
–
(Continued)
509
510
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
Crystal Logic Inc. 10573 W. Pico Blvd PMB 106 Los Angeles, CA 90064–2348 USA www.xtallogic.com
–
–
GAXRDGXRR
–
DeltaNu 5452 Aerospace Drive Laramie, WY 82070 USA www.deltanu.com
–
–
Raman
–
Digilab Inc. 84 October Hill Road Holliston, MA 01746 USA www.digilabglobal.com
–
–
Raman
–
DME Nanotechnologie GmbH Am Listholze 82 30177 Hannover Germany www.dme-spm.com
–
–
–
STM, AFM
EDAX Inc. 91 McKee Drive Mahwah, NJ 07430 USA www.edax.com
–
–
EDXS, WDXS, XRF
–
Elmitec Elektronenmikroskopie GmbH Albrecht-von-Groddeck-Straße 3 38678 Clausthal-Zellerfeld Germany www.elmitec.de
LEEM, PEEM
–
–
–
ESI 13900 NW Science Park Drive Portland, OR 97229-5497 USA www.esi.com
–
–
Laser ablation
–
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
FEI 5350 NE Dawson Creek Drive Hillsboro, OR 97124 USA www.fei.com
EFTEM
–
–
–
VG Scienta Ltd Maunsell Road St Leonards-on-Sea East Sussex, TN38 9NN United Kingdom www.vgscienta.com
XPS
–
–
–
Gatan Inc. 5794 W. Las Positas Blvd Pleasonton, CA 94588 USA www.gatan.com
EELS, EFTEM
–
–
–
GE, Measurement & Control Solutions Robert Bosch Straße 3 50354 Hürth Germany www.gesensinginspection.com
–
–
GAXRDGXRR
–
Hiden Analytical Ltd 420 Europa Boulevard Warrington, WA5 5UN United Kingdom www.hiden.co.uk
–
SIMS, SNMS
–
–
HORIBA Jobin Yvon S.A.S. 16–18, rue du Canal 91165 Longjumeau Cedex France www.horiba.com
–
GDMS
GD-OES, XRF, RAIRS, Raman, ELL
–
INEL Ltd Suite 5 Enterprise Center Shrivenham 100 Business Park Wiltshire, SN6 8TZ United Kingdom www.inel-xrd.com
–
–
GAXRDGXRR
–
(Continued)
511
512
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
ION-TOF GmbH Heisenbergstraße 15 48149 Münster Germany www.ion-tof.com
–
SIMS, LaserSNMS, LEIS
–
AFM
J.A. Woollam Co. 645 M Street, Suite 102 Lincoln, NE 68508-2243 USA www.jawoollam.com
–
–
ELL
–
Jasco Inc. 28600 Mary’s Court, Easton, MD 21601 USA www.jascoinc.com
–
–
Raman
–
Jeol USA Inc. 11 Dearborn Road Peabody, MA 01960 USA www.jeol.com
XPS, AES, EELS, EFTEM
–
EDXS, WDXS
STM, AFM
Jordan Valley Semiconductors Inc. 8601 Cross Park Drive, Suite 200 Austin, TX 78754-4578 USA www.jvsemi.com
–
–
GAXRDGXRR
–
JPK Instruments AG Bouchéstraße 12 Haus 2, Aufgang C 12435 Berlin Germany www.jpk.com
–
–
–
AFM
Kaiser Optical Systems Inc. 371 Parkland Plaza Ann Arbor, MI 48103 USA www.kosi.com
–
–
Raman
–
KETEK GmbH Hofer Straße 3 81737 München Germany www.ketek.net
–
–
TXRF, XRF
–
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
Kleindiek Nanotechnik GmbH Aspenhaustraße 25 72770 Reutlingen Germany www.nanotechnik.com
–
–
–
AFM
Kratos Analytical Ltd Wharfside Trafford Wharf Road Manchester, M17 1GP United Kingdom www.kratos.com
XPS, AES
–
–
–
Lambda Photometrics Lambda House Batford Mill Harpenden Herts, AL5 5BZ United Kingdom www.lambdaphoto.co.uk
–
–
–
STM, AFM, SNOM
LECO Corporation 3000 Lakeview Avenue St Joseph, MI 49085 USA www.leco.com
–
–
GD-OES
–
LK Technologies Inc. 3910 Roll Avenue Bloomington, IN 47403 USA www.lktech.com
AES, LEED
–
–
–
Nanonics Imaging Ltd Manhat Technology Park, Malcha Jerusalem, 91487 Israel www.nanonics.co.il
–
–
Raman
AFM, SNOM
Nanosurf AG TENUM Zentrum Grammetstraße 14 4410 Liestal Switzerland www.nanosurf.com
–
–
–
STM, AFM
(Continued)
513
514
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
National Electrostatics Corp. 7540 Graber Road P. O. Box 620310 Middleton, WI 53562–0310 USA www.pelletron.com
–
RBS
–
–
NT-MDT Co. Building 100, Zelenograd Moscow, 124482 Russia www.ntmdt.com
–
–
–
STM, AFM, SNOM
Ocean Optics GmbH Maybachstraße 11 73760 Ostfildern Germany www.oceanoptics.com
–
–
Raman
–
OCI Vacuum Microengineering Inc. 340 Saskatoon Street London, Ontario N5W 4R3 Canada www.ocivm.com
AES, LEED
–
–
–
Omicron NanoTechnology GmbH Limburger Straße 75 65232 Taunusstein Germany www.omicron.de
XPS, AES, UPS, LEED, PEEM
LEIS
–
STM, AFM
omt - optische messtechnik gmbh Hörvelsinger Weg 6 89081 Ulm Germany www.omt-instruments.com
–
–
ELL
–
PANalytical B.V. Lelyweg 1 7602 EA Almelo The Netherlands www.panalytical.com
–
–
TXRF, GAXRDGXRR
–
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
Park Systems Inc. 3040 Olcott Street Santa Clara, CA 95054 USA www.parkafm.com
–
–
–
AFM
PerkinElmer 940 Winter Street Waltham, MA 02451 USA www.perkinelmer.com
–
–
RAIRS, Raman
–
Physical Electronics Inc. 18725 Lake Drive East, Chanhassen, MN 55317 USA www.phi.com
XPS, AES, UPS
SIMS
–
–
Princeton Gamma-Tech Instruments Inc. 303C College Road East Princeton, NJ 08540 USA www.pgt.com
–
–
EDXS, WDXS
–
Princeton Instruments 3660 Quakerbridge Road Trenton, NJ 08619 USA www.princetoninstruments.com
–
–
Raman
–
Renishaw plc New Mills Wotton-under-Edge Gloucestershire, GL12 8JR United Kingdom www.renishaw.com
–
–
Raman
–
RHK Technology Inc. 1050 East Maple Road Troy, MI 48083 USA www.rhk-tech.com
XPS, LEED
–
–
AFM, STM
(Continued)
515
516
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
Rigaku/USA Inc. 9009 New Trails Drive The Woodlands, TX 77381 USA www.rigaku.com
–
–
TXRF, GAXRDGXRR
–
Sentech Instruments GmbH Schwarzschildstraße 2 12489 Berlin Germany www.sentech.com
–
–
ELL
–
SOPRALAB 55, avenue de l’Europe 92400 COURBEVOIE France www.sopra-sa.com
–
–
ELL
–
Specs Surface Nano Analysis GmbH Voltastraße 5 13355 Berlin Germany www.specs.de
XPS, AES, UPS, LEED, LEEM, PEEM
SIMS, SNMS, LEIS
–
STM
SPECTRUMA Analytik GmbH Fabrikzeile 21 95028 Hof Germany www.spectruma.de
–
–
GD-OES
–
Staib Instruments GmbH Hagenaustraße 22 85416 Langenbach Germany www.staibinstruments.com
AES, RHEED
–
–
–
STOE and CIE GmbH Hilpertstraße 10 64295 Darmstadt Germany www.stoe.com
–
–
GAXRDGXRR
–
Appendix B Surface and Thin-Film Analytical Equipment Suppliers Supplier
Technique Electron detection
Ion detection
Photon detection
Scanning probe microscopy
The AFM Workshop 2667 E. 28th Street #518 Signal Hill, CA 90755 USA www.afmworkshop.com
–
–
–
AFM
Thermo Fisher Scientific Inc. The Birches Industrial Estate Imberhorne Lane East Grinstead West Sussex, RH19 1UB United Kingdom www.thermoscientific.com
XPS, AES, UPS, HREELS
SIMS, SNMS, GDMS, FIM, AP, LEIS
BIS, EDXS, WDXS, XRF, RAIRS, Raman
STM, AFM
VSW Atomtech Ltd Nursery Road North Leigh Business Park Witney, Oxfordshire, OX29 6SN United Kingdom www.vsw.co.uk
XPS, AES, UPS, EELS
LEIS
BIS
–
Waterloo Digital Electronics Division of WDE Inc. 279 Weber Street N. Waterloo, Ontario N2J 3H8 Canada www.onramp.tuscaloosa.al.us
–
–
ELL
–
WITec Wissenschaftliche Instrumente und Technologie GmbH Lise-Meitner-Straße 6 89081 Ulm, Germany www.witec.de
–
–
Raman
AFM, SNOM
517
519
Index
a absolute resolution 17 accuracy – in EELS 83, 84 – in R-laser-SNMS measurements 179, 180 – in Rutherford backscattering spectroscopy 198 – in Total-Reflection X-Ray Fluorescence (TXRF) Analysis 277, 281 – of atomic coordinates in superstructures 105 ACOLISSA 207 acronyms 4, 5 adhesion, X-ray photoelectron spectroscopy and 32–34 AES, see Auger electron spectroscopy AFM, see atomic force microscopy aluminum 113 – energy spectrum of 226 – satellite lines of 14 – SSIMS spectra of oxide films on 128, 129 aluminum oxide 113 angle-integrated UPS 38 angle-resolved UPS 38 angle-resolved XPS 35 angle-scan TXRF 284 APFIM, see atom probe field ion microscopy appearance potential methods 437, 438 APT, see atom probe tomography argon 119, 144 ARUPS, see angle-resolved UPS atom probe – applications 252–257 – principles of operation 243 atom probe field ion microscopy 237 atom probe tomography 245 atomic force microscopy 3, 441, 443–464, 482
– applications 455–461 – constant force mode 446 – constant height mode 446 – force–distance curve measurements 447, 448 – force modulation microscopy 447 – forces 443, 444 – friction force microscopy 446, 447 – growth of ODS on silicon 459, 460, 461, 462 – harmonic imaging and torsional resonance mode 449–452 – instrumentation 452–455 – laser-based feedback signal detection 485 – liquid cell for measurements of surface processes 454 – phase imaging 447 – photoresist layer on silicon 456, 457 – principles 443–446 – pulsed force mode 448, 449 – PVD gold film on silicon 456 – quality factor 445, 446 – in situ measurements 457, 458 – tapping mode 445, 456, 457 – view of a liquid cell for measurements of surface processes 455 – Young’s modulus microscopy 447 atomic mixing 142 atomic-number correction factor 306 ATR, see attenuated total reflection attenuated total reflection 372–374 attenuation, in XRF 268 Auger electron spectroscopy 3, 43–65 – applications 54–60 – depth profiling 53, 54 – electron energy analyzers 45–47 – electron sources 44, 45 – and grain boundary segregation 54–56 – instrumentation 44–47
Surface and Thin Film Analysis: A Compendium of Principles, Instrumentation, and Applications, Second Edition. Edited by Gernot Friedbacher, Henning Bubert. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2011 by Wiley-VCH Verlag GmbH & Co. KGaA.
520
Index – interfaces 58 – nomenclature 10, 11 – principles 43, 44 – quantification 51–53 – and semiconductors 56–58 – spectral information 47–51 – and surface segregation 58–60 – thin films 58 – vacuum requirements 44 Auger energy 43 Auger parameter 27, 28 Auger process 10 Auger sensitivities, ratio between experimental and theoretical 52, 53
b back coupling 333 backscattering factor 52 backscattering yield, of H+ ions from Pb 200, 201 BaTiO3, EEL spectrum 77 beam broadening 296 beam-induced light emission, see ion beam spectrochemical analysis benzene on Ru(0001), experimental and calculated I–V curves of 102, 103 Berreman effect 368, 402 biomaterial interfaces, SFG spectroscopy on 428, 429 biosensors 133, 134 BIS, see Bremsstrahlung isochromat spectroscopy BLE, see ion beam spectrochemical analysis boron nitride – EELS quantification 83, 84 – low-loss spectrum 75, 76 Bragg energy 221 bremsstrahlung continuum 295 bremsstrahlung, elimination of 15 bremsstrahlung isochromat spectroscopy 3, 437, 438 bremsstrahlung radiation 438 Brewster angle 395 bright-field image 307, 308 broken bond model 148
c cantilevers 452, 453 CAP, see catalytic atom probe car paint defects 129, 130 carbon, effect of different chemical states on the KLL Auger spectrum of 49, 50 CARS, see coherent anti-stokes Raman scattering
CASSE technique, see controlled all solidstate electrolysis technique catalysis – field ion microscopy in 248, 249 – X-ray photoelectron spectroscopy and 28, 29, 30 catalytic atom probe 253 cathodic sputtering 330 CaTiSiO5, low-loss spectrum 76 Cauchy formula 398 CCD, see charge-coupled device CCT, see constant current topography ceramic coatings 339 cesium 119, 148 CFM, see chemical force microscopy CHA, see concentric hemispherical analyzer channeling 193 charge-coupled device 335, 381 charge-coupled device camera 146 charge-injection device 335 charging effects 142, 143 chemical-bond mapping 88 chemical enhancement effect 146 chemical force microscopy 447 chemical shift 20 chemical state imaging 26 CID, see charge-injection device CITS, see current imaging tunneling spectroscopy Cliff–Lorimer factor 306 CMA, see cylindrical mirror analyzer CO adsorption 420, 421 – onto Pd nanoparticles 421–423 coated glass 339 coatings 339 coherent anti-stokes Raman scattering 389, 491 collision cascade 118 colloidal nanoparticles, SFG spectroscopy on 427, 428 complex refractive index 393 component depth profile 25 components 25 computer programs – for calculating LEED I–V curves 104, 105 – elastic recoil detection analysis 223 concentration–depth profiles 218 concentration matrix 25 concentric hemispherical analyzer 16, 17 constant current mode 467 constant current topography 467 constant emission yield 336 constant height mode 467
Index continuously rotating compensator ellipsometers 397 controlled all solid-state electrolysis technique 487 copper, wide-scan XPS spectrum 19, 20 corrosion, X-ray photoelectron spectroscopy and 31, 32 crater bottom roughening 142 CrN 157 cross-section curve 230 cross-sectional STM 470 current imaging tunneling spectroscopy 469 cylindrical mirror analyzer 45, 46 Czerny–Turner monochromator 334
d DEPTH 223 depth profiling 149–152 – Auger electron spectroscopy 53, 54 – Cu–Ag–Si 352 – of Cu in Si 186, 187 – of Cu in SiO2 186, 187 – dual-beam technique for TOF-SIMS instruments 152 – Fe–Cr–Mo alloy 24 – glow-discharge optical emission spectroscopy 338, 339 – molecular depth profiles 152 – of P implantation in Si 156 – of passivation layer on high-purity chromium 151 – pure iron 24 – surface analysis by laser ablation 348–354 – by TXRF and GIXRF 283–285 – X-ray photoelectron spectroscopy and 23–25 depth resolution – elastic recoil detection analysis 223, 224 – nuclear reaction analysis 232–234 – Rutherford backscattering spectroscopy 198 – secondary neutral mass spectrometry 162 detected yield 124 detection limits – in ERDA 223, 224 – Glow-Discharge Mass Spectroscopy (GDMS) 263 – of NR-laser-SNMS 185 – using LIBS/LIPS techniques 347, 348 – using TXRF 279, 280 – for metals on Si wafer surfaces 185 deuteron beams 232 DFM, see dynamic force microscopy
differential scattering cross-section 218 diffraction of low-energy electrons, see lowenergy electron diffraction direct imaging magnetic sector mass analyzer 145 direct imaging mode 153, 154 disappearance cross-section 123 disappearance yield 123 DNA – detection of Sn-labeled DNA 188 – single nanocrystal DNA-base molecules 491 doppler broadening 233 Drude–Lorentz formula 398 dual-beam technique, for TOF-SIMS instruments 152 duoplasmatron 143, 144, 194 dynamic force microscopy 445 dynamic secondary ion mass spectrometry 117, 141–159 – applications 156–158 – atomic mixing 142 – charging effects 142, 143 – compensation of preferential sputtering 141 – crater bottom roughening 142 – imaging 152–154 – implantation of primary ions 142 – implantation profiles 156 – implantation standards 147, 148 – instrumentation 143–146 – ion sources 143, 144 – layer analysis 157, 158 – mass analyzers 144–146 – molecular depth profiles 152 – principles 141–143 – quantification 147–149 – spectral information 146 – sputter-induced roughness 142 – 3-D trace element distribution 158 – theoretical ionization models 148, 149
e EBIS, see electron beam ion source échelle spectrometer 348 EDX SSD, see energy-dispersive solid-state detector EDXS, see energy-dispersive X-ray spectroscopy EELS, see electron energy-loss spectroscopy EFTEM, see energy-filtered TEM EI SNMS, see electron-impact secondary neutral mass spectrometry elastic peak 74
521
522
Index elastic recoil detection analysis 3, 217–227 – applications 224–226 – data analysis 223 – ΔE–E method (energy-loss measurement) 217, 218 – depth resolution 223, 224 – energy spectra of O and Al recoils 226 – equipment 222 – fundamentals 218–220 – heavy projectiles 218 – light projectiles 218 – medium-heavy projectiles 218 – particle identification methods 220–222 – sensitivity 223, 224 – time-of-flight method (velocity measurement) 217, 218 elastic tunneling 112, 465 election-stimulated desorption 3 electron beam ion source 143 electron beam SNMS 166, 167 electron energy analyzers – Auger electron spectroscopy 45–47 – XPS 16–18 electron energy-loss spectroscopy 3, 67, 309 – combination with TEM 85 – imaging of element distribution 85–88 – instrumentation 70–72 – ionization losses 77, 78 – low-loss excitations 74–76 – principles 68, 69 – qualitative spectral information 72, 83 – quantification 83–85 electron-impact secondary neutral mass spectrometry 161–177 – see also secondary neutral mass spectrometry electron spectroscopy for chemical analysis 9 electron-stimulated desorption 261, 262 electron-stimulated desorption ion angular distribution 261, 262 electron tunneling model 148 element depth profiling, secondary neutral mass spectrometry 172, 173 element distributions, EELS and TEM imaging of 85–88 elemental map 26 ELL, see UV-VIS-IR ellipsometry ellipsometers 395–397 – achromatic compensators 396 – continuously rotating compensators 397
– null ellipsometers 396 – polarization modulation ellipsometers 397 – rotating-element ellipsometers 396 ellipsometry 393–405 – fundamental equation of 394 ELNES, see energy-loss near-edge structures energy analyzers 70 energy-dispersive solid-state detector 271 energy-dispersive X-ray spectroscopy 3, 77, 293–310 – artifacts in spectra 304 – imaging of element distribution 306–308 – line profiling 307 – principles 293–295 – qualitative spectral information 303, 304 – quantification 304–306 – spectrometer 297 – X-ray microanalysis and instrumentation 295–303 energy-filtered TEM 3, 85–87 energy-loss measurement 221 energy-loss near-edge structures 73 energy resolution – of detectors 295, 298, 301–303 – elastic recoil detection analysis 224 – in electron energy-loss spectroscopy 70–74 – in NRA depth profiling 233, 234 – of X-ray detector 301 – in X-ray photoelectron spectroscopy 16, 17 ERDA, see elastic recoil detection analysis ERIRS, see reflection absorption IR spectroscopy ESCA, see electron spectroscopy for chemical analysis ESCA-SCOPE 19 ESCALAB 250 19, 26 escape peaks 275 ESD, see electron-stimulated desorption ESD ion angular distribution 3 ESDIAD, see ESD ion angular distribution ESRF, see European Synchrotron Radiation Facility European Synchrotron Radiation Facility 280 EXAFS, see extended X-ray absorption fine structure Excimer lasers 346 EXELFS, see extended energy-loss fine structures extended energy-loss fine structures 78 extended X-ray absorption fine structure 316, 317
Index external reflection infrared spectroscopy, see reflection absorption IR spectroscopy extinction coefficient 393
friction force microscopy 446, 447 front coupling 333 FT-IRAS, see reflection absorption IR spectroscopy
f FABMS, see fast-atom bombardment mass spectroscopy factor analysis 25 Faraday cups 95, 145, 146 fast-atom bombardment mass spectroscopy 3, 263, 264 FCS, see fluorescence correlation spectroscopy FDM, see field desorption microscopy Fe–Cr alloys, Auger spectra 59, 60 FFM, see friction force microscopy FIAES, see field ion appearance energy spectroscopy FIB technique, see focused ion beam technique field desorption microscopy 242 field ion appearance energy spectroscopy 246, 247 – applications 255–257 – principles 244 – of reactants in CO oxidation on Pt 255, 256 field ion mass spectrometry 246 field ion microscopy 3, 237–260 – applications 248–252 – in catalysis 248, 249 – fluctuation-induced effects 249–252 – instrumentation 239–242 – principles 239–242 field ionization 240, 241 FIM, see field ion microscopy FIMS, see field ion mass spectrometry fine structures 79–83 fluorescence 304, 306, 378, 380 – in SNOM 488, 489 – Total-Reflection X-Ray 275–278, 280–285 – yield 51, 281, 294 fluorescence correlation spectroscopy 489 fluorescence resonance energy transfer 489 FMM, see force modulation microscopy focused ion beam technique 56, 487 force–distance curves 444, 445, 447, 448 force modulation microscopy 447 formic acid, time-resolved measurements for decomposition of 424, 425 FRET, see fluorescence resonance energy transfer
g GaAs – ELL spectra of a-Si/SiN multilayer stack on 398, 399 – nonresonant laser-SNMS mapping of a contact test structure on 185, 186 – RBS spectra of 195, 196 gallium 119 gas telescope detectors 221 Gatan imaging filter 72 Gaussian-broadened polynomial superposition parametric dispersion model 398 GD-MS, see glow-discharge mass spectrometry GD-OES, see glow-discharge optical emission spectroscopy geometric yield 183 geometrical shadowing 487 GIAB diffraction, see grazing incidence asymmetric Bragg diffraction GIAB geometry, see grazing incidence angle asymmetric Bragg geometry GIR, see grazing incidence reflection GIXRF, see grazing incidence X-ray reflectivity GIXS, see grazing incidence X-ray scattering glancing angle X-ray diffraction 314–316 glow-discharge mass spectrometry 3, 263, 339 glow-discharge optical emission spectroscopy 3, 329–344 – applications 339–342 – dc GD sources 340 – depth profiling 336–338 – glow discharge sources 330–334 – instrumentation 330–335 – principles 329, 330 – quantification 336, 337 – rf GD sources 340–342 – signal acquisition 334, 335 – spectral information 335 – spectrometer 334 glow discharge sputtering 349 Goebel mirrors 319 grain boundary segregation 54–56 graphite 477, 478 grazing incidence angle asymmetric Bragg geometry 315
523
524
Index grazing incidence asymmetric Bragg diffraction 319–321 – applications 323, 324 grazing incidence reflection 374 grazing incidence X-ray methods for nearsurface structural studies 311–327 – applications 321–325 – experimental techniques and data analysis 317–321 – grazing incidence X-ray geometry 312–314 – principles 311–317 grazing incidence X-ray reflectivity 281, 314, 318, 319 – applications 321–323 – depth profiling 283–285 grazing incidence X-ray scattering 315, 316, 324, 325 Grimm principle 330 guanine, optical constants 404 GXRR, see grazing incidence X-ray reflectivity
h H depth profiles 233, 236 hard coatings 339 harmonic imaging 449–452 HAS, see He atom scattering He atom scattering 408 heavy ion backscattering spectroscopy 201 HIBS, see heavy ion backscattering spectroscopy high-resolution electron energy-loss spectroscopy 374, 408 highly oriented pyrolytic graphite 477, 478 HOPG, see highly oriented pyrolytic graphite hot isostatically pressed steels 158 HREELS, see high-resolution electron energy-loss spectroscopy
i I–V curves, LEED 102, 103 IAES, see ion (excited) Auger electron spectroscopy IBSCA, see ion beam spectrochemical analysis ICCD detector, see intensified chargecoupled device detector IETS, see inelastic electron tunneling spectroscopy imaging – direct imaging mode 153, 154 – dynamic secondary ion mass spectrometry 152–154
– static secondary ion mass spectrometry 135, 136, 137 – XPS 26, 27 imaging atom probe 243 IMFP, see inelastic mean free path impact collision ion-scattering spectroscopy, structure analysis 210 implantation profiles 156 implantation standards 147, 148 in-column filters 72 inelastic electron tunneling spectroscopy 3, 112, 113 inelastic mean free path 11, 12 inelastic scattering cross-section 67 inelastic tunneling 112 infrared ellipsometer 397 infrared ellipsometry 374, 400–404 infrared reflection absorption spectrum 409 infrared spectroscopic ellipsometry 400 infrared spectroscopy 492 INS, see ion neutralization spectroscopy instrumentation – atomic force microscopy 452–455 – Auger electron spectroscopy 44–47 – dynamic secondary ion mass spectrometry 143–146 – electron energy-loss spectroscopy 70–72 – field ion microscopy 239–242 – glow-discharge optical emission spectroscopy 330–335 – ion beam spectrochemical analysis 358–360 – laser-secondary neutral mass spectrometry 182 – low-energy electron diffraction 94, 95, 96 – low-energy ion scattering 206–208 – near-field optical microscopy 482–488 – reflection absorption IR spectroscopy 367, 368 – Rutherford backscattering spectroscopy 194, 206 – scanning tunneling microscopy 467, 468 – static secondary ion mass spectrometry 119–123 – sum frequency generation spectroscopy 414–417 – surface analysis by laser ablation 346–348 – surface Raman spectroscopy 380–382 – total-reflection X-ray fluorescence analysis 269–275
Index – UV-VIS-IR ellipsometry 395–397 – X-ray photoelectron spectroscopy 12–19 intensified charge-coupled device detector 348 intensity of a spectral emission line, IBSCA 361 interfaces 128–131 interference fringes 314 inverse photoemission spectroscopy 3, 437, 438 ion beam spectrochemical analysis 3, 357–366 – applications 363–366 – depth profile of altered layer region of a LAS glass ceramic 365, 366 – instrumentation 358–360 – principles 357, 358 – quantitative analysis 361–363 – spectra of an antireflecting coating soda-lime glass 363, 364 – spectral and analytical information 360, 361 ion bombardment 23 ion (excited) Auger electron spectroscopy 3, 111 ion neutralization spectroscopy 111, 112 ion sources 359 ion sputtering 23, 53 IPES, see inverse photoemission spectroscopy IR spectroscopy, see infrared spectroscopy IRAS, see infrared reflection absorption spectrum IRRAS, see reflection absorption IR spectroscopy IRSE, see infrared spectroscopic ellipsometry
k K-resolved IPES 438 Kiessig fringes 314 kinematic factor 191, 193, 218 kinematic line broadening 222 KLL Auger series 47, 48 Kramers’ law 305 KRIPES, see K-resolved IPES
l LA, see laser ablation LAAS, see laser atomic absorption spectroscopy Langmuir–Blodgett films 134, 321, 323, 325, 386
laser ablation 3 – advantages of 354 – surface analysis by 345–355 laser atomic absorption spectroscopy 347 laser-induced breakdown spectroscopy 347 laser-induced fluorescence spectroscopy 3, 347 laser-induced plasma spectroscopy 347 laser ionization mass spectrometry 180 laser-secondary neutral mass spectrometry 3, 179–189 – applications 184–188 – experimental set-up 180, 181 – instrumentation 182 – ionization schemes 181, 182 – nonresonant laser-secondary neutral mass spectrometry 184, 185, 186 – principles 179–182 – quantification 183, 184 – resonant laser-SNMS 179, 180, 186–188 – spectral information 183 lasers – beam shape 346 – Excimer lasers 346 – Nd : YAG laser 346, 414 – titanium-sapphire laser 414 lateral force microscopy 446 lattices – direct 98, 99 – reciprocal 98, 99 layer analysis 157, 158 LEED, see low-energy electron diffraction LEEM, see low-energy electron microscopy LEIS, see low-energy ion scattering LFM, see lateral force microscopy LIBS, see laser-induced breakdown spectroscopy LIF spectroscopy, see laser-induced fluorescence spectroscopy LIPS, see laser-induced plasma spectroscopy liquid–gas interfaces, SFG spectroscopy at 429, 430 liquid–liquid interfaces, SFG spectroscopy at 429, 430 liquid metal ion sources 119, 120, 152 LMIS, see liquid metal ion sources local thermodynamic equilibrium 148, 149 low-energy electron diffraction 3, 93–106, 408 – applications and restrictions 100, 101, 105, 106 – computer programs 104, 105 – experimental techniques 102, 103
525
526
Index – four-grid display system 94 – history 93, 94 – instrumentation 94, 95, 96 – LEED I–V curves 101–103 – LEED pattern 95–97 – principles 93, 94, 101, 102 – qualitative information 96–101 – quantitative structural information 101–106 – spot profile analysis 100 low-energy electron microscopy 106–108 – applications and restrictions 108 – principles of operation 106–108 low-energy ion scattering 203–215 – alumina 209, 210 – applications 211–214 – chemical analysis 210 – energy information from spectrum 208 – instrumentation 206–208 – principles 203–206 – quantification 211, 212 – spectra of Cu/ZnO/SiO2 catalyst 212, 213 – structure analysis 210 – yield information 208–210 low-energy X-ray lines – energies 14 – linewidths 14
m macrostrain 321 magnesium, satellite lines of 14 magnetic sector field 144, 145 mass analyzers 120–123 – detector 145, 146 – dynamic secondary ion mass spectrometry 144–146 – magnetic sector field 144, 145 – quadrupole mass spectrometers 120, 121 – time-of-flight mass spectrometry 121–123 mass attenuation coefficient 268 mass resolution 120–123, 125, 149, 154, 182, 185, 242, 245 – Rutherford backscattering spectroscopy 197, 198 mass spectra 149 – of high-speed steel 149, 150 Mattauch–Herzog geometry 145 medium-energy ion scattering 201 MEIS, see medium-energy ion scattering 4-mercaptopyridine 469 metal ceside ions 148 metal island films 379 metallic coatings 339
metalorganic chemical vapor deposition 373 MeV He+ ions, backscattering spectra 192 mid-infrared range 400, 401 MLFM, see modulated lateral force microscopy MOCVD, see metalorganic chemical vapor deposition modulated lateral force microscopy 447 molecular depth profiles 152 monochromatization 15 Müller, E.W. 237 MULSAM, see multi-spectral Auger microscope multi-spectral Auger microscope 58 multicapillary X-ray lenses 298 multiple point analysis 54
n nanolayers, characterization of 283–285 NanoScope® AFM 453 narrow-scan spectra 19 Nd : YAG laser 346, 414 near-edge X-ray absorption fine structure 281, 317 near-field ablation 354 near-field optical microscopy 481–497 – advanced tip fabrication 487, 488 – basic set-up 482, 483 – coating deposition and aperture formation 486, 487 – instrumentation and operation 482–488 – scanning and feedback techniques 484, 485 – SNOM applications 488–493 – SNOM variations 483, 484 – taper formation 486 – tip fabrication 485–488 near-field Raman spectroscopy 387, 490–492 near-infrared, ellipsometric measurements in 399 near-infrared excitation wavelengths 378 neodymium-yttrium-aluminum-garnet laser 346, 414 NEXAFS, see near-edge X-ray absorption fine structure Ni on Si substrate, backscattering spectra for MeV He+ ions on 192 nomenclature, in electron spectroscopy 10 nonlinear optical spectroscopy 387–390 – coherent anti-stokes Raman scattering 389
Index – spatially offset Raman spectroscopy 390 – stimulated femtosecond Raman scattering 389, 390 – sum frequency generation spectroscopy 387–389 nonresonant laser-secondary neutral mass spectrometry 179, 180, 184, 185, 186 – element mapping 185, 186 nonresonant susceptibility 413 NRA, see nuclear reaction analysis NSOM, see scanning near-field optical microscopy nuclear reaction analysis 3, 229–236 – applications 234–236 – depth resolution 232–234 – equipment 232–234 – with gamma-emission 236 – principles 231, 232 null ellipsometers 396
o octadecylsiloxane 458, 459, 460, 461, 462 octadecyltrichlorosilane 460 ODNs, see oligodeoxynucleotides ODS, see octadecylsiloxane oligodeoxynucleotides 133, 134 optical constants 404 optical lever technique 453 optical microscopy, near-field 481–497 oxide films 128 oxide scales 339 oxygen, energy spectrum of 226
p parallel angle-resolved XPS 18 parallel-detection EELS 71 particle-induced gamma emission 230 particle probes 2 particulate and film-type surface contamination 277, 278 Paschen–Runge polychromator configuration 334 passivated implanted planar silicon detector 194 passivation, X-ray photoelectron spectroscopy and 31, 32 PASTM, see photon-assisted scanning tunneling microscopy PDI, see phase detection imaging PEELS, see parallel-detection EELS PEEM, see photoelectron emission microscopy PEM, see photoelastic modulator peptide nucleic acid 133, 134
PESTM, see photon emission scanning tunneling microscopy PFDMS, see pulsed-field desorption mass spectrometry PGM, see plane-grating monochromator phase detection imaging 447 phase imaging 447 pHEMA, see poly(2-hydroxyethyl methacrylate) photoelastic modulator 397 photoelectron emission microscopy 108 photoelectron emission process 10 photomultiplier tube 334 photon-assisted scanning tunneling microscopy 470 photon emission scanning tunneling microscopy 470 photon energies, normalized XPS spectrum of 16 photon scanning tunneling 483 Physikalisch-Technische Bundesanstalt, TXRF instrumentation 281, 282 PIGE, see particle-induced gamma emission PIPS detector, see passivated implanted planar silicon detector plane-grating monochromator 281 PLAP, see pulsed laser atom probe plasma etching 131 plasma SNMS 167–169 plasmon energy loss 50, 51 plasmons 75 PMT, see photomultiplier tube PNA, see peptide nucleic acid polarization 382 polarization-dependent SFG spectroscopy 419, 420 – ppp polarization 419 – on solid surfaces 420 – ssp polarization 419 polarization modulation ellipsometers 397 poly(2-hydroxyethyl methacrylate) 429 polyethylene 429 polymer brushes, pH-dependent switching of 403, 404 polymer coatings 339 polymers – mass-resolved images of 136 – SFG spectroscopy on 428, 429 – treatment of surfaces 131, 132 – X-ray photoelectron spectroscopy and 30, 31 poly(methyl methacrylate) 370 polypropylene 429 polystyrene, SSIMS spectrum 125, 126
527
528
Index polytetrafluoroethylene 131, 132 poly(vinyl chloride), optical constants 402 positive noble gas ions 119 postionization 161 – via electron impact 163, 164 powder materials, SFG spectroscopy on 427, 428 ppp polarization 419, 425 preferential sputtering 141 primary ions, implantation of 142 probes 2 projectiles 232 protons 232 Pt(111), SFG spectra of CO adsorption on single crystal surface of 417, 418 PTB, see Physikalisch-Technische Bundesanstalt PTFE, see polytetrafluoroethylene pulsed-field desorption mass spectrometry 246 – applications 254, 255 pulsed force mode AFM 448, 449 pulsed laser atom probe 244, 245 pump-probe SFG, broad band and 423–427 PVC, see poly(vinyl chloride)
q quadruple photodiode 446 quadrupole mass spectrometers 120, 121 Quantum 2000 19 quantum mechanical tunneling effect 465 quartz 15
r radial distribution function 83 RAE, see resistive anode encoder RAIRS, see reflection absorption IR spectroscopy Raman scattering 377 Raman spectra – of methyl mercaptan 387 – of nitrophenyl-modified and untreated glassy carbon 384 Raman spectroscopy 374 Rayleigh criterion 481 RBS, see Rutherford backscattering spectroscopy reactive sputtering 146 reflection absorption IR spectroscopy 3, 367–375 – applications 369–371 – instrumentation 367, 368 – principles 368, 369
– related techniques 374 – spectra for poly(ethyl methacrylate) 370, 372 reflection EXAFS 316, 317 – applications 325 reflection high-energy electron diffraction 3, 408 ReflEXAFS, see reflection EXAFS refractive index 393 relative resolution 17 relative sensitivity factor 124, 125, 147, 362 – for metals on Si wafer surfaces 185 repulsive force 449 resistive anode encoder 146 resolution – depth, see depth resolution – energy, see energy resolution – mass 197, 198 – spatial 18, 19 resonance NRA 234 resonant laser-SNMS 179, 180, 186–188 – of copper around tellurium and cadmium inclusions 186, 187 rf-powered sources 332 Rh(111), reduction of surface oxide on 474 RHEED, see reflection high-energy electron diffraction rotating-element ellipsometers 396 Rowland sphere 15 RSF, see relative sensitivity factor Rutherford backscattering spectroscopy 3, 191–202, 217 – accuracy 198 – applications 198–200 – depth resolution 198 – instrumentation 194, 206 – mass resolution 197, 198 – principles 191–194 – quantification 196, 197 – sensitivity 198 – spectra of GaAs implanted with Si 195, 196 – spectral information 194–196
s SAED, see selected-area electron diffraction SAM, see scanning Auger microscopy sample thickness 333 satellite lines – of magnesium and aluminum 14 – removal of 15 SBD, see solid-state surface barrier detectors
Index SCANIIR, see ion beam spectrochemical analysis scanning atom probe 237 scanning Auger microscopy 3, 45, 61–63 scanning near-field infrared microscopy 492 scanning near-field optical microscopy 3, 441, 481 – applications 488–493 – disadvantages 493 – fluorescence 488, 489 – near-field Raman spectroscopy 490–492 – outlook 493 – variations of 483, 484 scanning near-field optical microscopy-IR spectroscopy 492, 493 scanning probe microscopy 441, 481 scanning SFG 415 scanning SIMS 152, 153 scanning tunneling microscopy 3, 441, 442, 465–480 – applications 470–479 – cerium alkoxyl-substituted phthalocyanine complex on HOPG 477, 478 – constant current mode 467 – constant height mode 467 – Cu dissolution on a Cu(100) surface 476, 477 – growth of Ce on a Rh(111) surface 471, 472 – instrumentation 467, 468 – lateral and spectroscopic information 468–470 – principles 465–467 – Si(111)-7X7 surface exposed to O2 471, 473 – Si(111) surface 468, 469 – spin-polarized 469 – surface oxide on Rh(111) surface 474, 475 – Violet Lander molecule adsorbed on Cu(211) surface 477 scanning tunneling spectroscopy 469 screened scattering potential 203 second harmonic generation 3, 388, 408 secondary electron detectors 123 secondary ion mass spectrometry 284, 349, 408 – depth profile of altered layer region of a LAS glass ceramic 365, 366 – dynamic, see dynamic secondary ion mass spectrometry – three-dimensional 154, 155 secondary ions 164–166
secondary neutral mass spectrometry – advantages of 162 – applications 174, 175 – depth resolution 162 – detection power 162 – electron beam 166, 167 – element depth profiling 172, 173 – HF-plasma SNMS 174 – instrumentation and methods 166–169 – plasma SNMS 167–169 – postionization via electron impact 163, 164 – principles 162–166 – spectra of a Cu sample containing trace amounts of P and Fe 165 – spectral information and quantification 170–172 – suppression of residual gas and secondary ions 164–166 secondary yield 124 SEELS, see serial-detection EELS Seeman–Bohlin geometry 315 SEIRA, see surface-enhanced infrared absorption effect; surface-enhanced Raman spectroscopy selected-area electron diffraction 56 self-assembled monolayers 386, 458, 469 semiconductors 278–287 – Auger electron spectroscopy and 56–58 – depth profiling by TXRF and GIXRF 283–285 – synchrotron radiation-based techniques 280–283 – vapor-phase decomposition and droplet collection 285–287 – vapor-phase treatment and total reflection X-ray fluorescence analysis 287 – X-ray photoelectron spectroscopy and 35–37 sensitivity 53, 71, 77, 84, 85, 124, 143, 146, 218, 223, 224, 348, 380, 381 – relative sensitivity factors 147, 171, 172, 185, 263, 362, 393, 396 – Rutherford backscattering spectroscopy 198 – surface 204, 208, 209, 211 – ultimate 239, 246 sequencing by hybridization 188 serial-detection EELS 70 serial-recording EEL spectrometer 71 SERRS, see surface-enhanced resonance Raman scattering SERS, see surface-enhanced Raman scattering
529
530
Index SEXAFS, see surface EXAFS SFG, see sum frequency generation SFG spectroscopy, see sum frequency generation spectroscopy SFRS, see stimulated femtosecond Raman scattering SHG, see second harmonic generation SiGe/Si strained layer superlattice 199, 200 signal-to-noise ratios 63, 71, 103, 207, 381 silicon – ellipsometric spectra from SiO2 layer on 399 – formation of a buried β-FeSi2 layer in 199 – nonresonant laser-SNMA spectrum of 184 – SSIMS spectrum 125, 126 silicon carbide, EEL spectrum 73, 74 silicon drift detectors 272, 273, 274 SIMNRA 223, 224 SIMS, see secondary ion mass spectrometry single tube piezo scanner 454 single-walled carbon nanotube-coated tips 490 SiO2/Si3N4 double layer on silicon, IR ellipsometric spectra 402, 403 Sn-labeled DNA, R-laser-SNMS image of 188 SNIM, see scanning near-field infrared microscopy SNMS, see laser-secondary neutral mass spectrometry SNOM, see scanning near-field optical microscopy SNOM-IR, see scanning near-field optical microscopy-IR spectroscopy soft X-ray appearance potential spectroscopy 3, 437 solid–liquid interfaces, SFG spectroscopy on 428 solid-state surface barrier detectors 230 solid surface, definition of 1 SORS, see spatial offset Raman scattering; spatially offset Raman spectroscopy SPA-LEED system 95, 96 spatial offset Raman scattering 382 spatial resolution, X-ray photoelectron spectroscopy 18, 19 spatially offset Raman spectroscopy 390, 493 spectroscopic notation 10, 11 spectrum-image method 87, 307 spectrum matrix 25 spin-polarized STM 469
SPM, see scanning probe microscopy spot profile analysis 100 sputter-induced roughness 142 sputtering 23, 53, 118, 141, 162 – cathodic 330 – glow discharge 349 – reactive 146 SSA, see step-scan analyzer mode SSDs, see silicon drift detectors SSIMS, see static secondary ion mass spectrometry SSP, see step-scan polarizer mode ssp polarization 425 SSRL, see Stanford Synchrotron Radiation Lightsource stainless steel, study of pitting corrosion at MnS inclusion 61–63, 64 Stanford Synchrotron Radiation Lightsource 280 static secondary ion mass spectrometry 3, 117–139 – applications 127–138 – biosensors 133, 134 – imaging 135, 136, 137 – instrumentation 119–123 – interfaces 128–131 – ion sources 119, 120 – mass analyzers 120–123 – oxide films 128 – polymers 131, 132 – principles 117, 118 – quadrupole mass spectrometers 120, 121 – quantification 123–125 – spectral information 125–127 – surface reactions 134, 135 – time-of-flight mass spectrometry 121–123 – ultra-shallow depth profiling 137, 138 steel – Auger spectra of fracture surfaces of 12%-Cr steel 55 – 3-D trace element distribution 158 step-scan analyzer mode 397 step-scan polarizer mode 397 stimulated femtosecond Raman scattering 389, 390 STM, see scanning tunneling microscopy stopping power 218 STS, see scanning tunneling spectroscopy sum frequency generation 3, 374, 388, 408 sum frequency generation spectroscopy 387–389, 407–435 – at the alcohol–vapor interface 429, 430 – applications 417–430
Index – broad band 416 – broad band IR-visible 426 – on colloidal nanoparticles and powder materials 427, 428 – on colloidal Pt nanoparticles 427 – instrumentation and operation modes 414–417 – introduction 407–410 – at liquid–gas and liquid–liquid interfaces 429, 430 – measurements 415 – modes of 415, 416 – under near-atmospheric gas pressure 420, 421 – number density of molecules from signal intensity 413, 414 – polarization-dependent 416, 417, 419, 420 – on polymer and biomaterial interfaces 428, 429 – pump-probe 416 – second-order polarization 411 – signal intensity and lineshape 412, 413 – on solid–liquid interfaces 428 – on solid surfaces and solid–gas interfaces 417–428 – on supported metal nanoparticles 421–423 – theory 410–414 – time-resolved and broadband 423–427 – under UHV conditions 417–419 – vibrational IR-visible 410, 411 super-lattices 99 superconductors, X-ray photoelectron spectroscopy and 34, 35 supported metal nanoparticles, SFG spectroscopy on 421–423 surface analysis acronyms 4, 5 surface analysis by laser ablation 345–355 – depth profiling 348–354 – instrumentation 346–348 – laser types 346, 347 – near-field ablation 354 – schemes 347, 348 surface barrier detector 221 surface composition by analysis of neutral and ion impact radiation, see ion beam spectrochemical analysis surface-enhanced infrared absorption 492 surface-enhanced infrared absorption effect 372–374 surface-enhanced Raman scattering 378, 379 – spectra of methyl mercaptide 387
surface-enhanced Raman spectroscopy 3, 386, 387, 490 surface-enhanced resonance Raman scattering 379, 490 surface EXAFS 317 surface melting 200 surface Raman spectroscopy 377–391 – applications 383–387 – instrumentation 380–382 – porous materials 385, 386 – principles 377, 378 – quantification 383 – resonant excitation 378 – specific surface area 377, 378 – spectral information 382 – surface-enhancement 378 – ultrasensitive equipment 377 – unenhanced Raman spectroscopy at smooth surfaces 383–385 surface reactions 126, 134, 135, 246, 248, 316 surface removal 23 surface segregation, Auger electron spectroscopy and 58–60 surface-specific analytical techniques – using non-particle excitation 3 – using particle or photon excitation 3 surfaces, importance of 2 survey XPS spectrum 19, 20 switching behavior of stimuli-responsive mixed polymer brushes 403 SWNT-coated tips, see single-walled carbon nanotube-coated tips SXAPS, see soft X-ray appearance potential spectroscopy synchrotron 367 synchrotron radiation 16 synchrotron radiation-based techniques 280–283 Synge, E.H. 481
t TCO films, analysis of 399 TDS, see thermal desorption spectroscopy TEIRA, see tip-enhanced infrared absorption TEM, see transmission electron microscopy TEM/STEM Hitachi H-8110 300 temperature-programmed desorption, see thermal desorption spectroscopy Tensor LEED approximation 104 TERS, see tip-enhanced Raman scattering THC, see torsional harmonic cantilever
531
532
Index thermal desorption spectroscopy 3, 262, 263 thin-film criterion 306 thin films, and interfaces 58 thin layers 339 three-dimensional SIMS 154, 155 Ti-diamond layer, depth profiles 58, 59 tilt angles of molecules on a surface 369 time-of-flight atom probe techniques 242–246 – applications of 252–254 time-of-flight ERDA 225 time-of-flight LEIS, for Au films deposited on B 213, 214 time-of-flight mass spectrometry 121–123, 136, 180 – of silver bromide and silver chloride crystals 137 time-of-flight SIMS 180 time-of-flight SIMS instruments, dual-beam technique for 152 tin, chemical shift 21 TiN-coated steel, quantitative depth profile 340 TiN–TiAlN–Fe sample, crater surfaces 353 TiO2–SiO2–TiO2 (TST-mirror) – IBSCA depth profile 364, 365 – SIMS depth profile 364, 365 tip-based signal enhancement 491 tip-enhanced infrared absorption 492 tip-enhanced Raman scattering 490 titanium-based coatings, depth profile analysis 349 titanium-sapphire laser 346, 414 TOF-ERDA, see time-of-flight ERDA TOF-LEIS, see time-of-flight LEIS TOF-MS, see time-of-flight mass spectrometry TOF-SIMS, see time-of-flight SIMS torsional harmonic cantilever 450, 451 torsional resonance mode 449–452 total-reflection X-ray fluorescence analysis 3, 267–292 – applications 277–287 – depth profiling by 283–285 – detection limits for various elements on Si wafers 279, 280 – instrumentation 269–275 – particulate and film-type surface contamination 277, 278 – principles 267–269 – quantification 276, 277 – semiconductors 278–287
– spectral information 275, 276 – synchrotron radiation-based techniques 280–283 – vapor-phase decomposition and droplet collection 285–287 – vapor-phase treatment and 287 transformation probability 123 transition elements – LLM Auger series 47, 48 – MNN Auger series 47, 48, 49 transmission electron microscopy 67, 68 – instrumentation 70–72 – interactions between high-energy electrons and matter in 68 transmission Raman spectroscopy 382 TRIM program 224 tube etching 486 tunneling current 465 tunneling junction, energy level diagram 466 Turner’s etching method 486 TXRF analysis, see total-reflection X-ray fluorescence analysis
u ultra-shallow depth profiling 137, 138 ultra-shallow junctions, characterization of 283–285 ultrahigh vacuum 13 ultraviolet photoelectron spectroscopy 3, 38, 39 unenhanced Raman spectroscopy at smooth surfaces 383–385 UPS, see ultraviolet photoelectron spectroscopy useful yield 124, 183 UV-VIS-IR ellipsometry 3, 393–405 – applications 398–404 – infrared ellipsometry 400–404 – instrumentation 395–397 – principles 393–395 – UV-VIS-NIR spectral region 398–400
v Van de Graaff electrostatic accelerator 194 vapor-phase decomposition 287 – and droplet collection 285–287 vapor-phase treatment 287 – and total reflection X-ray fluorescence analysis 287 velocities of particles 221 VG9000 263 Violet Lander molecule adsorbed on Cu(211) surface 477
Index Vis–NIR spectral range 398 VPD, see vapor-phase decomposition VPT, see vapor-phase treatment
w W/TiN/Ti/Si contact structure 58 wafer surface preparation system 285 wavelength-dispersive X-ray spectroscopy 293 – spectrometer 297 – spectrum of BaTiO3 302 WDXS, see wavelength-dispersive X-ray spectroscopy white lines 77 wide-scan XPS spectrum 19, 20 WSPS, see wafer surface preparation system
x X-ray absorption near edge structure 317 X-ray detector, energy resolution 301 X-ray detector with Li-drifted Si crystal 299 X-ray diffraction 3 – glancing angle 314–316 X-ray microanalysis and instrumentation 295–303 X-ray notation 10, 11 X-ray photoelectron spectroscopy 3, 9–41, 349 – and adhesion 32–34 – applications 28–37 – Auger parameter 27, 28 – and catalysis 28, 29, 30 – chemical shifts 19–21 – configuration in a spectrometer 18 – and corrosion 31, 32 – depth profiling 23–25
– electron energy analyzers 16–18 – energy resolution 17 – imaging 26, 27 – instrumentation 12–19 – nomenclature 10, 11 – and passivation 31, 32 – and polymers 30, 31 – principles 9–12 – quantification 21–23 – and semiconductors 35–37 – spatial resolution 18, 19 – spectral information 19–21 – spectrometers 18 – and superconductors 34, 35 – synchrotron radiation 16 – ultrahigh vacuum 13 – vacuum requirements 12, 13 – X-ray sources 13–15 X-ray take-off angle 298 XANES, see X-ray absorption near edge structure XPS, see X-ray photoelectron spectroscopy XRD, see X-ray diffraction XSTM, see cross-sectional STM
y YMM, see Young’s modulus microscopy Young’s modulus microscopy 447
z ZAF correction 306 zero force 449 zero-loss peak 74 zinc-coated steel, GD-OES and LIBS emission profiles of Zn and Fe in 349, 350 ZnO : Al film, optical constants 400
533