Springer Series in

Surface Sciences
Editor: Robert Gomer 25

Springer-Verlag Berlin Heidelberg GmbH

Springer Series in Surface Sciences
Editors: G. Ertl, R. Gomer and D. L. Mills Managing Editor: H.K.V. Latsch

Physisorption Kinetics 20 Scanning Tunneling Microscopy I

By H. J. Kreuzer. Z. W. Gortel General Principles and Applications to Clean
2 The Structure of Surfaces and Adsorbate-Covered Surfaces
Editors: M. A. Van Hove, S. Y. Tong Editors: H.-J. Gtintherodt, R. Wiesendanger
2nd Edition
3 Dynamical Phenomena at Surfaces, Interfaces
and Superlattices 21 Surface Phonons
Editors: F. Nizzoli, K.-H. Rieder, R. F. Willis Editors: W. Kress, F. W. de Wette

4 Desorption Induced by Electronic Transitions, 22 Chemistry and Physics of Solid Surfaces VIII
DIET II Editors: R. Vanselow, R. Howe
Editors: W. Brenig, D. Menzel 23 Surface Analysis Methods in Materials
5 Chemistry and Physics of Solid Surfaces VI Science
Editors: R. Vanselow, R. Howe Editors: D. J. O'Connor, B. A. Sexton,
R. SI. C. Smart
6 Low-Energy Electron Diffraction
Experiment, Theory 24 The Structure of Surfaces III
and Surface Stmcture Determination Editors: S. Y. Tong, M. A. Van Hove,
By M. A. Van Hove, W. H. Weinberg, C.-M. Chan K. Takayanagi, X. D. Xie

7 Electronic Phenomena in Adsorption 25 NEXAFS Spectroscopy

and Catalysis By J. StOhr
By V. F. Kiselev, O. V. Krylov 26 Semiconductor Surfaces and Interfaces
8 Kinetics ofinterface Reactions ByW. Monch
Editors: M. Gmnze, H. J. Kreuzer 2nd Edition

9 Adsorption and Catalysis on Transition Metals 27 Helium Atom Scattering from Surfaces
and Their Oxides Editor: E. Hulpke
By V. F. Kiselev, O. V. Krylov 28 Scanning Tunneling Microscopy II
10 Chemistry and Physics of Solid Surfaces VII Further Applications
Editors: R. Vanselow. R. Howe and Related Scanning Techniques
Editors: R. Wiesendanger, H.-J. Giintherodt
II The Structure of Surfaces II 2nd Edition
Editors: J. F. van der Veen, M. A. Van Hove
29 Scanning Tunneling Microscopy III
12 Diffusion at Interfaces: Microscopic Concepts Theory of STM
Editors: M. Gmnze, H. J. Kreuzer, J. J. Weimer and Related Scanning Probe Methods
13 Desorption Induced by Electronic Transitions, Editors: R. Wiesendanger, H.-J. Gtintherodt
DIET III 2nd Edition
Editors: R. H. Stulen, M. L. Knotek 30 Concepts in Surface Physics
14 Solvay Conference on Surface Science By M. C. Desjonqueres, D. Spanjaard*)
Editor: F. W. de Wette 31 Desorption Induced by Electronic Transitions,
15 Surfaces and Interfaces of Solids DIET V
By H. Ltith*) Editors: A. R. Burns, E. B. Stechel,
16 Atomic and Electronic Structure of Surfaces D. R. Jennison
Theoretical Foundations 32 Scanning Tunneling Microscopy
By M. Lannoo, P. Friedel and its Application
17 Adhesion and Friction By C. Bai
Editors: M. Gmnze, H. J. Kreuzer 33 Adsorption on Ordered Surfaces
18 Auger Spectroscopy and Electronic Structure of Ionic Solids and Thin Films
Editors: G. Cubiotti, G. Mondio, K. Wandelt Editors: H.-J. Freund, E. Umbach

19 Desorption Induced by Electronic Transitions, 34 Surface Reactions

Editor: R. J. Madix
DIET IV
Editors: G. Betz, P. Varga 35 Applications of Synchrotron Radiation
High-Resolution Studies of Molecules
and Molecular Adsorbates on Surfaces
*) Available as a textbook Editor: W. Eberhardt
Joachim Stohr

NEXAFS
Spectroscopy

With 177 Figures

Dr. Joachim StOhr
IBM Almaden Research Center, 650 Harry Road,
San Jose, CA 95120-6099, USA

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

Professor Robert Gomer, Ph.D.

The James Franck Institute, The University of Chicago, 5640 Ellis Avenue,
Chicago, IL 60637, USA

Professor Douglas L. Mills, Ph.D.

Department of Physics, University of California,
Irvine, CA 92717, USA

Managing Editor: Dr.-Ing. Helmut K. V. Lotsch

Springer-Verlag, TIergartenstrasse 17,
D-69121 Heidelberg, Germany

First Edition 1992

Corrected Printing 1996
ISBN 978-3-642-08113-2 ISBN 978-3-662-02853-7 (eBook)
DOI 10.1007/978-3-662-02853-7

Library of Congress Cataloging-in-Publication Data applied for

Die Deutsche Bibliothek - CIP-Einheitsaufnahme.
StOhr, Joachim: NEXAFS spectroscopy 1 Joachim StOhr. - Corr. 2. printing. -
Berlin; Heidelberg; New York; Barcelona; Budapest; Hong Kong; London; Milan; Paris;
Santa Clara; Singapore; Tokyo : Springer, 1996
(Springer series in surface sciences; 25)
ISBN 978-3-642-08113-2
NE:GT
This work is subject to copyright. All rights are reserved, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data
banks. Duplication of this publication or parts thereof is permitted only under the provisions
of the German Copyright Law of September 9, 1965, in its current version, and permission for
use must always be obtained from Springer-Verlag. Violations are liable for prosecution under
the German Copyright Law.
© Springer-Verlag Berlin Heidelberg, 1992
Originally published by Springer-Verlag Berlin Heidelberg New York in 1992
Softcover reprint of the hardcover I st edition 1992
The use of general descriptive names, registered names, trademarks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
Typesetting: Macmillan India Ltd., India
SPIN 10535950 54/3144-5 4 3 2 1 0 - Printed on acid-free paper
 To Megan
All too often the thought of you had to substitute for your presence.
Yet, what a lucky man I am to have a daughter with so much life!
Preface

The purpose of this book is the development of the principles and experimental
techniques underlying near edge X-ray absorption fine structure (NEXAFS)
spectroscopy and the demonstration of the power of the technique for the study
of the electronic and crystallographic structure of low-Z molecules bonded to
surfaces. Low-Z molecules are defined as those consisting of hydrogen, carbon,
nitrogen, oxygen and/or fluorine atoms, which are particularly important in
surface chemistry. This book is the first comprehensive treatment of the subject
and presents a unified picture of theoretical and experimental concepts and
results. It develops all concepts from an elementary level and is suitable for
students and researchers without extensive prior knowledge in X-ray absorption
spectroscopy. On the other hand, it discusses state-of-the-art instrumentation,
analysis techniques, and experimental and theoretical results and is therefore
also suited for the advanced spectroscopist. The spectra of free molecules are
discussed first, since their understanding provides the basis for understanding
spectra of molecules bonded to surfaces, the main topic of the book. The
connection to spectra of polymeric molecules is also made. The book may
therefore be of interest not only to surface scientists but also to researchers
studying free molecules or polymers. The various molecular adsorption systems
studied by NEXAFS are tabulated. Future scientific opportunities making use
of the NEXAFS technique in conjunction with advanced synchrotron radiation
sources are also discussed. These range from element-specific microscopy stud-
ies of solid surfaces to studies of molecular conformations at liquid surfaces.

Portola Valley, CA J. Stohr

January 1991
Acknowledgements

I started writing this book in November 1987 and thought I could finish it in a year. Little
did I know ... , most notably, that I knew too little. If nothing else, it's been worthwhile
writing this book for all that I have learned!
The Stanford Synchrotron Radiation Laboratory (SSRL) has played a crucial role in
my own research and in the development of NEXAFS spectroscopy. I have fond
memories of the excitement that hovered in the air in the early years of the Laboratory
and I consider myself lucky to have been a part of it. I also remember the frustrations
associated with doing synchrotron radiation research ... the endless hours of waiting for
the beam. It certainly has been a love-hate relationship (more love)! It is only fitting to
start the acknowledgements by thanking the SSRL staff who made the Laboratory hum.
Artie Bienenstock deserves a special word ofthanks because he made it possible for me to
start my own research program during my years on the SSRL staff (1977-1981)-rumor
has it, at the expense of some outside users.
Over the last ten years I have benefited from the help of many people, and I need to
mention a few. It was Dave Shirley who first threw the idea at me in 1976 to do
Absorption Spectroscopy for Chemical Analysis (ASCA). Except for a few measurements
on polymer films the idea did not go far in my postdoc days at Lawrenae Berkeley
Laboratory. It did, however, stimulate my interest in X-ray absorption spectroscopy and
later led to my involvement with SEXAFS and NEXAFS. Rolf Jaeger was the key person
who did the first NEXAFS experiments with me in 1980. What an exciting time and what
great collaboration it was!
In my EXXON years, 1981-1985, I met the late Earl Muetterties who taught me that
in chemistry there are other molecules besides CO, and so his student Allen Johnson and
I began the study of organic molecules. I remember Allen drawing structural formulas for
me at SSRL and teaching me about molecules like cyciooctatetraene that I had never
heard of. I also learned more chemistry and surface chemistry through interactions with
John Horsley and John Gland. John Horsley deepened my understanding ofthe K-shell
spectra of molecules. John Gland taught me about desulfurization and other catalytic
reactions of molecules with surfaces. During 1983/84 I had the pleasure of working with
Francesco Sette for a memorable year. We both lived near Brookhaven National
Laboratory to do experiments at the National Synchrotron Light Source but ironically
had to fly to California to see the light (at SSRL). Our brainstorming sessions about the
spectra were great fun, and Francesco often awed me with his scientific knowledge. The
lack of photons at Brookhaven also had its good side, it left me enough spare time to
learn how to windsurf on the Long Island Sound. During this time I also started to
collaborate with Adam Hitchcock, a collaboration that has continued to this day. After
reading this book it will be easy for the reader to judge the depth of this collaboration.
Adam has not only been a great source of information for me but he always shared his
knowledge and unpublished data in a truly scientific spirit.
After "coming home" to California and starting work at IBM in 1985 I had the
pleasure of ciosely working with Duane Outka. Duane was so independent and efficient
that I could have spent my winters skiing at Lake Tahoe. Instead, I only went skiing
occasionally and wrote a lot of papers with Duane. At this point I need to mention the
X Acknowledgements

special collaboration I have enjoyed since about 1982 with Bob Madix. It has involved
many students, of whom Duane, Paul Stevens and Jeff Solomon became true "syn-
chrotronjocks". My collaboration with Bob has been so successful because our interests,
at least initially, were quite complementary. He was after specific problems in surface
chemistry and my main goal was the development of new synchrotron techniques. I know
that at least I succeeded in learning and benefiting a great deal from him. The same is true
for my more recent collaboration with Cindy Friend. She and her students Jeff Roberts
and Albert Liu played a big role in teaching me the surface chemistry of organic
molecules.
When I realized that I didn't understand enough theory to write this book (I still may
not), I asked John Horsley, who in the meantime had also moved to Silicon Valley, to
give me his XIX multiple scattering code. John shared his programs without hesitation and
I am deeply indebted to him not only for his programs but for his constant advice and
teaching, all the way to critically reading some of the chapters of this book. The XIX
calculations were carried out by Wilfried Wurth, who, like previously Rolf Jaeger, had
been a student of Dietrich Menzel. As such he knew his science. I was fortunate that
Wilfried allowed me to look over his shoulder and he taught me while cranking out the
spectra of a large number of molecules. Alex Bradshaw deserves a "thank you" for
arguing with me and in the end teaching me about symmetry selection rules in NEXAFS.
A special word of thanks is due to my long time collaborator and friend Klaus
Baberschke, who as a visitor took part in the first NEXAFS experiment in 1980, and in
the meantime has established his own highly successful program in X-ray absorption
spectroscopy. Klaus and I have had our share offriendly arguments and discussions over
the years and I have learned a great deal from him, his students, and associates, especially
Dimitri Arvanitis. When beam time was scarce or non-available at SSRL he allowed me
to collaborate with him at BESSY. Some of these were the easiest experiments I ever
did - on the telephone.
I also need to express some thoughts about the IBM Almaden Research Center
where, besides my own home, this book was written. The laboratory is not only located in
beautiful surroundings, but it is a great place to do science. The following anecdote shows
why I have loved working there. I had been trying to dose benzoic acid onto the Ag(11O)
surface at SSRL but, because of its low vapor pressure (it is a solid), I had been
unsuccessful in dosing it without contamination. At IBM I mentioned my frustrations to
my boss George Castro, who advised me to sublime it a few times to clean it up. An hour
later he came by my office with a glass flask and said "Here, I sublimed some for you."
George also shared my vision about the importance of synchrotron radiation research
and he made the difference where it counts - getting the money to do it! At Almaden I
learned about polymers from Tom Russell and Jerry Swalen, and I greatly benefited from
stimulating discussions with the numerous surface scientists at the lab. Some noteworthy
conversations that left me truly breathless took place while running through the hills with
Dick Brundle, Hugh Brown and Charlie Rettner. During one of those runs, after I had
just become a manager and did not think that I could finish this book without a
coauthor, Charlie told me how much more respect he had for authors who did it on their
own. That decided it - how could I disappoint Charlie? The task of finishing this book
was greatly aided by my secretary Toni Vanderwege, who in many ways ran the
Department for me and gave me time to think and write.
Finally, a word of thanks and appreciation to Renee, who supported me in the
endeavor (or should I say ordeal) of writing this book. When my motivation dropped
toward the end she encouraged me not to succumb to this all-too-human tendency. It's
certainly harder to finish than to start, but what a feeling it is to be done!
Contents

1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Why Another Spectroscopy and Another Book? . . . . . . . . 1
1.2 Development of the NEXAFS Technique . . . . . . . . . . . . 4

2. Theory of Inner Shell Excitation Spectra . . . . . . . . . . . . 8

2.1 Description of the X-Ray Absorption Process. . . . . . 8
2.1.1 The X-Ray Absorption Cross Section . . . . . . 8
2.1.2 Optical Oscillator Strength and Sum Rules . . . . . .. 11
2.2 Time Scales in Inner-Shell Excitations . . . . . . . . . . . . .. 13
2.2.1 Electron and Hole Lifetimes . . . . . . . . . . . . . . . . 13
2.2.2 Separation of Electronic and Nuclear Degrees
of Freedom. . . . . . . . . . . . . . . . . . . . . . . . .. 14
2.3 The Electronic Ground State. . . . . . . . . . . . . . . . . . . . 15
2.3.1 The Hartree-Fock Method. . . . . . . . . . . . . . . . . 15
2.3.2 Roothaan-Hall and Semiempirical Methods. . . . . . . 19
2.4 Transition Energies . . . . . . . . . . . . . . . . . . . . . . . .. 20
2.4.1 Koopmans'Theorem.................... 21
2.4.2 The Transition State Method . . . . . . . . . . . . . .. 23
2.4.3 Localized Versus Delocalized Core Hole. . . . . . . .. 26
2.5 Transition Intensities . . . . . . . . . . . . . . . . . . . . . . .. 26
2.5.1 One-Electron Versus Multi-Electron Transitions. . .. 27
2.5.2 Effects of Nuclear Vibrations . . . . . . . . . . . . . .. 27
2.5.3 The Sudden Approximation . . . . . . . . . . . . . . .. 28
2.5.4 Adiabatic Versus Sudden Excitation . . . . . . . . . .. 29
2.6 Bound Versus Continuum Final States. . . . . . . . . . . . .. 30
2.6.1 Improved Virtual Orbitals . . . . . . . . . . . . . . . .. 30
2.6.2 Continuum Final States . . . . . . . . . . . . . . . . .. 31
2.7 The Xa Multiple Scattering Method. . . . . . . . . . . . . .. 34
2.7.1 Introduction to the Method. . . . . . . . . . . . . . .. 34
2.7.2 Exchange Potential and Latter Tail. . . . . . . . . . .. 34
2.7.3 Muffin Tin Potential . . . . . . . . . . . . . . . . . . .. 35
2.7.4 Multiple Scattering Wavefunctions . . . . . . . . . . .. 36
2.7.5 Transition Energies. . . . . . . . . . . . . . . . . . . .. 37
2.7.6 Practical Procedures for Calculation
of K-Shell Spectra. . . . . . . . . . . . . . . . . . . . .. 38
XII Contents

2.8 Ab Initio Stieltjes-Tchebycheff Molecular Orbital Method.. 39

2.8.1 Introduction to the Method. . . . . . . . . . . . . . .. 39
2.8.2 Calculational Procedure . . . . . . . . . . . . . . . . .. 39
2.8.3 Stieltjes-Tchebycheff Orbitals . . . . . . . . . . . . . .. 41
2.8.4 Feshbach-Fano Method . . . . . . . . . . . . . . . . .. 42
2.9 Shell-by-Shell Multiple Scattering Method. . . . . . . . . . .. 43
2.10 Approximations Leading to the EXAFS Equation. . . . . .. 44

3. Symmetry and Molecular Orbitals. . . . . . . . . . . . . . . . . . . 48

3.1 Origin and Labelling of Molecular Orbitals. . . . . . . . . . 48
3.2 Some Molecular Orbitals and Irreducible Representations . 49
3.2.1 Diatomics and Linear Triatomics. . . . . . . . . . . .. 49
3.2.2 Hydrogen Fluoride, Water, Ammonia, and Methane. 53
3.3 Molecular Orbitals, Equivalent Orbitals and Hybrid Orbitals 55
3.3.1 Molecular Orbital Versus Valence Bond Theory. . 55
3.3.2 Ionization Potentials in Methane . . . . . . . . . . . .. 58
3.3.3 Bonding in Ethane, Ethylene, and Acetylene . . . . .. 59
3.4 Interactions Between Localized Orbitals: Conjugation. . . .. 61
3.4.1 First and Second Order Perturbation Treatment. . .. 62
3.4.2 Interactions in Chain-Like Hydrocarbons. . . . . . .. 64
3.5 Splitting of Antibonding Orbitals
Due to Bond-Bond Interactions . . . . . . . . . . . . . . . . . 67
3.5.1 The Linear Combination of Bond Orbitals Method.. 67
3.5.2 Application to (J and n Bonds in Hydrocarbons . . .. 67
3.6 Orbital Orientation, Symmetry, and the Dipole Selection Rule 69
3.6.1 Orbital Orientation and Angular Dependence
of the Dipole Matrix Element . . . . . . . . . . . . . .. 69
3.6.2 Group Theory and the Dipole Selection Rule . . . . . 72
3.6.3 Applications of Group Theoretical Selection Rules .. 73
3.7 Spin-Dependent Excitations . . . . . . . . . . . . . . . 75

4. Experimental and Calculated K-Shell Spectra

of Simple Free Molecules . . . . . . . . . . . . . . .. . . . . . . . . .. 79
4.1 Experimental Methods: The ISEELS Techniques . . . . . . . 79
4.2 Characteristic Resonances in K-Shell Spectra. . . . 83
4.2.1 Overview........................... 83
4.2.2 n* Resonances . . . . . . . . . . . . . . . . . . . . . . .. 88
4.2.3 Rydberg and Mixed Valence/Rydberg Resonances. .. 90
4.2.4 (J* Shape Resonances. . . . . . . . . . . . . . . . . . .. 93
4.2.5 Multi-Electron Features . . . . . . . . . . . . . . . . .. 95
4.2.6 Correlation of Multiple Scattering and Molecular
Orbital Calculations: The N2 Molecule . . . . . . . .. 97
4.2.7 Molecular Orbitals and Resonances
of Simple Hydrocarbons. . . . . . . . . . . . . . . . .. 102
4.2.8 Exchange Splitting in the Oxygen Molecule. 104
4.3 Systematics of Resonance Positions . . . . . . . . . . . . . . . 106
 Contents XIII

5. Principles, Techniques, and Instrumentation of NEXAFS . . . . .. 114

5.1 Achieving Adsorbate Sensitivity. . . . . . . . . . . . 114
5.2 Electron Yield Detection. . . . . . . . . . . . . . . . 118
5.2.1 Principles..................... 118
5.2.2 Quantitative Description of Electron Yield . . . . . 122
5.2.3 Adsorbate Versus Substrate Signal . . . . . . . . . . .. 127
5.2.4 Experimental Details and Detectors . . . . . . . . . .. 130
5.3 Fluorescence Yield Detection . . . . . . . . . . . . . . . . . .. 133
5.3.1 Absorption and Scattering of Soft X-Rays. . . . . . .. 133
5.3.2 X-Ray Reflection and Diffuse Scattering. . . . . . . .. 137
5.3.3 Adsorbate Fluorescent Signal and Substrate Background 139
5.3.4 Practical Scheme for Suppression of Background Signal 141
5.3.5 Experimental Details and Detectors . . . . . . . . . .. 145
5.4 Comparison of Detection Techniques. . . . . . . . . . . . . .. 149
5.5 Normalization and Background Corrections . . . . . . . . .. 154
5.5.1 General Considerations. . . . . . . . . . . . . . . . . .. 154
5.5.2 Normalization by a Reference Monitor . . . . . . . .. 156
5.5.3 Division by the Clean Sample Spectrum. . . . . . . . 158
5.5.4 Subtraction of the Clean Sample Spectrum. . . . . . 160
6. Spectra of Condensed, Chemisorbed, and Polymeric Molecules:
An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.1 From Free to Chemisorbed Molecules. . . . . . . . . . . . 162
6.1.1 Influence of Extra-Molecular Interactions
on K-Shell Spectra . . . . . . . . . . . . . . . . . . . 162
6.1.2 X-Ray Polarization and Molecular Orientation 169
6.2 Chemisorbed Atoms Versus Molecules. . . . . . . . . . . . .. 172
6.2.1 The NEXAFS Region. . . . . . . . . . . . . . . . . . .. 172
6.2.2 The SEXAFS Region . . . . . . . . . . . . . . . . . . 176
6.3 The Building Block Approach for Large Molecules 179
6.3.1 Assembly of Pseudodiatomics . . . . . . . . . . . . . 179
6.3.2 Experimental Examples: Free, Adsorbed,
and Polymeric Molecules . . . . . . . . . . . . . . . . . 179
6.3.3 Theoretical Foundation of the Building Block Picture 183
6.4 Limitations of the Building Block Picture. . . . . . . . . . .. 185
6.4.1 Effects of Conjugation . . . . . . . . . . . . . . . . . .. 185
6.4.2 n* Bond-Bond Interactions . . . . . . . . . . . . . . .. 185
6.4.3 a* Bond-Bond Interactions. . . . . . . . . . . . . . .. 190
6.4.4 Aromatic Rings: Benzene and Related Molecules ... 199
6.4.5 Crystalline Solids . . . . . . . . . . . . . . . . . . . . .. 202
6.4.6 Effects of Core Hole Localization. . . . . . . . . . . .. 205
6.5 Assembly of Functional Groups to Macromolecules. . . . .. 207
7. Analysis of K-Shell Excitation Spectra by Curve Fitting . . . . . .. 211
7.1 The Need for a Quantitative Analysis . . . . . . . . . . . . .. 211
7.1.1 Curve Fitting of Original Spectra. . . . . . . . . . . .. 211
7.1.2 Curve Fitting of Difference Spectra. . . . . . . . . . .. 212
XIV Contents

7.2 Lineshapes of NEXAFS Resonances. . . . . . . . . . . . . .. 213

7.2.1 Gaussian, Lorentzian and Voigt Functions. . . . . .. 213
7.2.2 Asymmetric Gaussian and Lorentzian Lineshapes . .. 214
7.2.3 Giant Resonance Lineshape . . . . . . . . . . . . . . .. 219
7.2.4 Giant Resonance Versus Asymmetric Gaussian
Lineshapes . . . . . . . . . . . . . . . . . . . . . . . . . 221
7.3 Lineshapes of NEXAFS Steps. . . . . . . . . . . . . 222
7.3.1 Origin of Steps. . . . . . . . . . . . . . . . . . 222
7.3.2 Gaussian and Lorentzian Shaped Steps . . . 223
7.4 Examples of Steps. . . . . . . . . . . . . . . . . . . . 225
7.4.1 Continuum Steps for Free Molecules. . . . . . . . . . 225
7.4.2 Continuum Steps for Condensed Molecules,
Polymers, and Solids . . . . . . . . . . . . . . . . . . 228
7.4.3 Steps for Physisorbed and Chemisorbed Molecules 231
8. a* Resonance Position and Bond Length . . . . . . . . . . . . . . .. 239
8.1 Theoretical Predictions and the Search for a Correlation . .. 239
8.2 Predictions by Scattering Theory . . . . . . . . . . 242
8.3 Predictions by Molecular Orbital Theory . . . . . 245
8.4 Empirical Correlation for Simple Free Molecules 249
8.4.1 The Energy Reference Question. . . . . . . 250
8.4.2 Bonds Involving High-Z Atoms. . . . . . . . . . . . 252
8.5 Correlation for Large Molecules . . . . . . . . . . . . . . . .. 255
8.5.1 Chain-Like Hydrocarbons with Alternating Bonds .. 255
8.5.2 Non-aromatic Hydrocarbon Rings . . . . . . . . . . . 258
8.5.3 a-Conjugated and Aromatic Molecules . . . . . . 260
8.6 Extension to Condensed, Physisorbed,
and Chemisorbed Molecules. . . . . . . . . . . . . . . . . 264
8.6.1 Chemical Shifts and Resonance Positions. . . . . . .. 265
8.6.2 Weakly Adsorbed Molecules. . . . . . . . . . . . . . .. 267
8.6.3 Strongly Adsorbed Molecules Without Bonding Shifts 269
8.6.4 Strongly Adsorbed Molecules with Bonding Shifts. . 271
8.6.5 General Rules, Comments and the Use of Standards 272
9. The Angular Dependence of Resonance Intensities. . . . . . . . . . 276
9.1 Classification of Molecules. . . . . . . . . . . . . . . . . . 276
9.2 Resonance Intensities for Elliptically Polarized X-Rays . . .. 277
9.3 Angular Dependence of the Transition Matrix Elements . .. 279
9.4 Effect of Substrate Symmetry . . . . . . . . . . . . . . . . . .. 283
9.4.1 General Considerations. . . . . . . . . . . . . . 283
9.4.2 Twofold or Higher Substrate Symmetry. . . . . . . .. 283
9.4.3 Threefold or Higher Substrate Symmetry . . . . . . .. 284
9.5 Intensity Plots for n* and a* Vector Orbitals . . . . . . . . . 285
9.6 Intensity Plots for n* and a* Orbitals in a Plane ... . . .. 287
9.7 An Example: The n* Resonance Intensity in Graphite ... 288
9.8 Angular Dependence of Intensities in Difference Spectra . .. 290
 Contents XV

10. Selected Applications of NEXAFS. . . . . . . . . . . . . . . . . . .. 292

10.1 What Can We Hope to Learn? . . . . . . . . . . . . . . . . .. 292
10.2 CO on Pt(111) and the Effects of Na and H2 . . . . . . . . .. 295
10.2.1 CO/Pt(111) and CO/Na/Pt(111) . . . . . . . . . . .. 295
10.2.2 CO/Pt(l11) in the Presence of H2 Gas. . . . . . . .. 299
10.3 Molecular Oxygen on Pt(111) and Ag(110) . . . . . . . . . .. 301
10.3.1 Physisorbed Versus Chemisorbed 02 on Pt(l11) . .. 302
10.3.2 Chemisorbed 02 on Ag(llO) . . . . . . . . . . . . . .. 306
10.3.3 Hybridization and Bond Length. . . . . . . . . . . .. 307
10.4 The Bonding of Simple Hydrocarbons on Metals . . . . . .. 309
10.4.1 Bonding and Orientation of C 2 on Ag(110) . . . . .. 310
10.4.2 Di-sigma Bonded C 2H 4 on Pt(111). . . . . . . . . .. 315
10.5 The Bonding of Phenyl-Ring-Based Molecules
to Metal Surfaces . . . . . . . . . . . . . . . . . . . . . . . . .. 317
10.5.1 General Comments . . . . . . . . . . . . . . . . . . .. 317
10.5.2 Benzene........................... 319
10.5.3 Phenyl Thiolate and Phenoxide . . . . . . . . . . . .. 322
10.5.4 Pyridine........................... 325
10.5.5 Reaction Intermediates: Benzyne. . . . . . . . . . . .. 327
10.6 Thiophene on Pt (111) and Poly thiophene on Pt . . . . . . .. 329
10.7 Langmuir-Blodgett Chains on Si(111) . . . . . . . . . . . . .. 333
10.7.1 Experimental Results . . . . . . . . . . . . . . . . . .. 334
10.7.2 Chain Tilt, Intra-Chain Bonding,
and Origin of Resonances . . . . . . . . . . . . . . . . 337
10.7.3 Analysis of Difference Spectra . . . . . . . . . . . . .. 337
10.7.4 Curve Fits of Original Spectra. . . . . . . . . . . . .. 338
10.7.5 Structural Results . . . . . . . . . . . . . . . . . . . .. 340
11. A Look into the Future . . . . . . . .. . . . . . . . . . . . . . . . .. 342
11.1 Micro-NEXAFS.......................... 342
11.2 Liquid Surfaces. . . . . . . . . . . . . . . . . . . . . . . . . .. 343
11.3 Time-Resolved Studies. . . . . . . . . . . . . . . . . . . . . .. 344
11.4 Monosized Cluster Ions. . . . . . . . . . . . . . . . . . . . .. 345
11.5 Molecular Subgroups in Complex Environments . . . . . .. 347
Appendices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
A. Derivation of the EXAFS Equation. . . . . . . . . . . . . . . . . . 349
B. Chemisorbed Molecules Studied by NEXAFS. . . . . . . . . . . . 359
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Listing of Illustrated K-Shell Spectra . . . . . . . . . . . . . . . . . . . . . . 401
1. Introduction

In this first chapter we outline the reasons for writing a book about the near
edge X-ray absorption fine structure, or NEXAFS, technique, and explain why
the application of the technique to the study of molecules bonded to surfaces is
particularly important and interesting. A historical perspective of the develop-
ment of NEXAFS spectroscopy is also given.

1.1 Why Another Spectroscopy and Another Book?

Our century has been marked by the success of X-ray diffraction in revealing the
atomic structure of matter. Some of the most remarkable examples of this
success are the detailed structure determinations of complex macromolecules
that are the basis of life. The protein myoglobin, for example, which transports
oxygen in muscle, consists of a long twisted polypeptide chain in the form of a
so-called oc helix, as illustrated in Fig. 1.1. Especially remarkable is that the
landmark structure determinations shown in Fig. Ll and that of the well-known
molecule of heredity, deoxyribonucleic acid, or DNA, were accomplished in the
1950s [Ll]. At a time when the structures of macromolecules of such complexity
have been known for more than 30 years it would seem a trivial task to
determine the local geometry of a simple diatomic like O 2 bonded to the surface
of a material. Unfortunately and very surprisingly this is not the case. Surface
crystallography is still in a development stage and the answers to many
conceptually simple problems, such as the local structural geometry of a
chemisorbed molecule, still evade us. This is largely due to the fact that surface
science problems deal with a low abundance of species and in many cases a lack
of periodicity, which renders the tool of choice for bulk crystallography, X-ray
diffraction, of limited value. In fact, to date not a single structural determination
of a chemisorbed molecule has been performed using this technique.
In general, only a handful of structural determinations of molecules on
surfaces have been carried out, mostly by means of low energy electron diffrac-
tion (LEED) [1.2]. This is a striking deficit in light of the importance of the
problem. After all, water and our atmosphere consist of molecules and therefore
molecule-surface interactions are ever-present phenomena in our world, in fact,
they are the origin of life. For example, the bonding of O 2 to the heme group
(Fig. Ll) in myoglobin and hemoglobin is the basis of oxygen transport in
 N
Helical thread Backbone of N, Entire helix High resolution model
with a-carbon a-C and carbonyl
atoms C atoms :-
go
3Q.
c:
~
o·
:>

oM ~ N

Heme group

sA

Fig. 1.1. Microscopic structure of Ct: helix, an important structural motif in many proteins, and of the
protein myoglobin, taken from the textbook by Stryer [1.1]. The tightly coiled Ct: helix was proposed
by Pauling and Corey in 1951 as one possible conformation of polypeptide chains. It underlies the
structures of the oxygen carriers myoglobin and hemoglobin whose structures were solved by
Kendrewand Perutz, respectively, in the late I 950s. On the right is shown a high resolution model of
myoglobin which contains the oxygen-binding heme group. The O 2 molecules are bound in a bent,
end-on orientation to the iron atom in the center of the heme group shown at the bottom right
 1.1 Why Another Spectroscopy and Another Book? 3

muscle and blood [Ll]. The interactions of molecules with surfaces also play
key roles in modern technology. They underlie the multibillion dollar petroleum
industry serving the world's needs in gasoline and energy production. Here the
interaction of oil, which can be envisioned as large molecules consisting of
carbon and hydrogen atoms, with the surface of metal particles, so-called
catalysts, is used to process oil into high octane gasoline [1.3, 4]. Another
example is the important Haber-Bosch process, where hydrogen from water and
nitrogen from the atmosphere are converted into ammonia by an iron-based
catalyst [1.5]. Ammonia and its derivatives are widely used as fertilizers and for
the production of explosives, plastics, vitamins, and drugs. In high technology
areas, molecule-surface interactions are important in the preparation of semi-
conductor materials by reactive etching processes and they determine the
adhesion of thin polymer coatings used as insulators in the packaging process of
electronic devices [1.6] and as protective and lubricating overcoats on magnetic
recording media. In a more general and basic sense, the interaction of molecules
with surfaces creates fascinating molecular transformations and reactions whose
understanding is the essence of chemistry.
This book describes a new surface science technique which appears capable
of solving some of the problems in surface chemistry mentioned above. The near
edge X-ray absorption fine structure or, for short, NEXAFS technique was
developed in the 1980s with the goal of elucidating the structure of molecules
bonded to surfaces, in particular low-Z molecules. The term "low-Z molecules"
refers to mostly organic molecules containing important atomic building blocks
such as hydrogen, carbon, nitrogen, oxygen and fluorine. It is these molecules
that, for the technological reasons mentioned above, are of special interest in
surface science. NEXAFS is no panacea for all structural problems concerning
molecules bonded to surfaces but, as most techniques, has specific strengths and
weaknesses.
NEXAFS selects a specific atomic species through its K-edge and probes its
bonds to intra-molecular and, to a lesser degree, extra-molecular (i.e., surface
atoms) neighbors. Among the capabilities are: the ability to detect the presence
of specific bonds in molecules (e.g., C-C, C=C, C=C, and C-H bonds in
hydrocarbons), the determination of the lengths of these intra-molecular bonds
and the derivation of the precise orientation of molecules and functional groups
on surfaces or in solids. By comparison of spectra for free and chemisorbed
molecules NEXAFS can also reveal which orbitals are involved in the chemi-
sorption bond. To first order it gives no information on the detailed atomic
adsorbate-substrate registry. This part of the problem can, however, be solved
with the related and complementary surface extended X-ray absorption fine
structure (SEXAFS) technique [1.7, 8]. Nevertheless, NEXAFS by itself is a new
powerful surface structural tool which not only provides important information
but does so with conceptual and experimental ease.
The purpose of this book is to give a thorough and comprehensive review of
the principles, experimental techniques, and analysis methods of NEXAFS.
Such a review appears timely because the major development phase ofNEXAFS
4 1. Introduction

lies behind us and the commissioning and construction of new synchrotron

radiation facilities around the world will enable an increasing number of
scientists to use the technique. Besides discussing the NEXAFS technique itself,
this volume also presents several applications of NEXAFS for the study of
molecules bonded to surfaces (Chap. 10). Emphasis is given to the character-
ization of the behavior of different classes of molecules upon chemisorption. The
book would not be complete without a look at the prospects of novel scientific
applications of the technique, which is provided in the last chapter. The various
molecular chemisorption systems studied by NEXAFS are tabulated in
Appendix B.

1.2 Development of the NEXAFS Technique

After revealing the shell structure of atoms at the beginning of this century,
X-ray absorption spectroscopy was first used in the 1920s for structural invest-
igations of matter. The observed fine structure near the absorption edges was
first explained in general terms by a theory of Kossel [1.9] and for many years
was referred to as "Kossel structure". In contrast, the structure extending for
hundreds of electron volts past the edge was called "Kronig structure" after the
scientist [1.10] who germinated its theoretical explanation. The latter is, of
course, what is now called extended X-ray absorption fine structure (EXAFS)
[1.11].
During the 1970s, when EXAFS was developed into a powerful structural
tool, the near edge structure which was automatically recorded with every
EXAFS spectrum was largely discarded as too complicated. One exception was
the near edge structure oflow-Z molecules whose K-shells with binding energies
in the 250-750 eV range (carbon: 285 eV, nitrogen: 400 eV, oxygen: 535 eV, and
fluorine: 685 eV) could be conveniently excited by electron energy loss spectro-
scopy [1.12, 13], as discussed in detail in Chap. 4. The high quality K-shell
excitation spectrum ofN2 recorded with inner shell electron energy loss spectro-
scopy (ISEELS) [1.14] served as the input for the first quantitative calculation of
the near edge structure of a molecule, performed by Dehmer and Dill in 1975
[1.15]. Since then significant progress in the understanding of the near edge
structure in molecules, inorganic complexes, biological systems, crystalline and
disordered solids, and chemisorbed atoms and molecules has been made
[1.16-18], and we are now in a position to use it as a structural tool.
In recent years the Kronig near edge structure has been mostly referred to as
the X-ray absorption near edge structure, XANES, [1.19] or the near edge X-ray
absorption fine structure, NEXAFS [1.20]. The term XANES is more com-
monly used for solids and inorganic complexes while NEXAFS is used more in
conjunction with surfaces. In fact, the name NEXAFS was created in part
because it rhymes with SEXAFS, the surface version of EXAFS [1.7, 8], and it
constitutes the first part of a SEXAFS spectrum. Alternatively, one can think of
 1.2 Development of the NEXAFS Technique 5

"N" EXAFS as "Not" EXAFS. In the following we shall use the term NEXAFS
specifically for K -shell excitation spectra of low-Z molecules adsorbed on
surfaces.
The general importance oflow-Z molecules is a consequence of their natural
abundance and their unique chemical bonding properties leading to myriads of
different molecules and organic compounds. The nature of the intramolecular
bonds also puts the low-Z molecules into a class of their own with respect to
their near edge structure. First, the strong directionality of the covalent bonds
between low-Z atoms coupled with the polarized nature of synchrotron radi-
ation leads to strongly polarization-dependent K-shell spectra for oriented
molecules. Secondly, the characteristically short bond lengths (1.1-1.5 A), the
strong dependence of the bond length on the hybridization of the bond, and the
large backscattering amplitude of low energy electrons from low-Z atoms
combine to create K-shell spectra with prominent, structure-sensitive reson-
ances in the first 30 eV above the edge.
This important fact is demonstrated in Fig. 1.2 for the C K-edge. The
molecules benzene (C6H6) and cyclohexane (C 6H 12 ) as well as the correspond-
ing bulk compounds graphite and diamond all exhibit large resonance-like
structures in the first 30 eV above the K-shell threshold (~285 eV) which are
significantly larger than the higher energy EXAFS oscillations. These reson-
ance-like structures are the topic of this book. As we shall see, near edge
structures are very sensitive to the intramolecular and chemisorptive bonding
and, most importantly, follow simple rules. These semiempirical rules, which
were aided or retroactively confirmed by theory, allow a theory-independent

Cyclohexane
Cs H12

Fig. 1.2. Carbon K-shell X-ray absorption

Diamond
spectra of solid benzene and cydohexane, con-
densed onto a substrate at 80 K, and of single
crystal diamond and highly oriented pyrolytic
graphite. All spectra were recorded in geo-
metries which eliminate polarization depend-
ent effects [1.21-23]. Note pronounced NEX-
AFS structure compared to the weaker EX-
Photon Energy (eV) AFS structure at higher energies
6 1. Introduction

interpretation of the observed near edge structures or resonances in most cases.

This makes the study oflow-Z molecules adsorbed on surfaces possibly the most
important application of near edge X-ray absorption spectroscopy.
The first NEXAFS spectrum of a chemisorbed molecule [CO on Ni(l00)]
was recorded in 1980 at the Stanford Synchrotron Radiation Laboratory and
published in conjunction with data for NO on Ni(l00) in 1981 [1.24]. The
motivation behind the first study was a comparison of the K-shell near edge
structures of chemisorbed CO and NO recorded by electron versus ion yield
detection [1.25]. Important prerequisites for the first NEXAFS measurements
were the availability of high-intensity monochromatic soft X-ray (250-1000 eV)
synchrotron radiation [1.26, 27] and the previous development of electron yield
detection techniques for SEXAFS measurements [1.26, 28-31]. The 1981 publi-
cation focused on the polarization dependence of the near edge resonances,
which was used to determine the molecular orientation on the surface. A more
detailed study of the spectra of CO, NO, and N2 on Ni(l00) published in 1982
[1.20] derived the equations for the angular dependence of the resonances
following the theory of Davenport [1.32] and Wallace and Dill [1.33] and
discussed the energetics of the observed NEXAFS resonances in terms of those
of the free molecules recorded earlier with ISEELS [1.12, 13].
In 1983 it was realized through the study of CO, formic acid (HCOOH), and
methanol (CH 3 0H) on Cu(l00) [1.34] that the position of the so-called "u shape
resonance" was a sensitive measure of the intramolecular bond length and in
conjunction with the existence and intensity of the lower energy "n resonance"
revealed the bond hybridization. With this study the sensitivity of NEXAFS to
the intramolecular structure and bonding was established. In the same year
NEXAFS was applied for the first time to the study of aromatic rings such as
benzene (C 6 H 6 ) and pyridine (CsHsN) chemisorbed on Pt(111) [1.35] and it
was demonstrated that the orientation of such molecules can be readily
determined.
The close correspondence between the spectra for gas and chemisorbed
molecules led to a closer inspection of the correlation between u resonance
position and bond length in 1984. Using the large data base available for the K-
shell ISEELS spectra of low-Z molecules [1.13] a simple empirical correlation
was established [1.36, 37] and extended to chemisorbed hydrocarbons by Stohr
et al. [1.38]. In particular, these authors pointed out that, because of the linear
nature of the correlation, bond lengths could be simply obtained from the
spectra by measuring with a ruler the separation of the u resonance position
from the onset of absorption.
The years 1984-1986 saw the use of NEXAFS to solve several important
surface science problems, like the desulfurization process of thiophene on
Pt(lll) [1.39], the structural transformation of ethylene to ethylidine on Pt(lll)
[1.40,41], the Na-induced bonding and bond length changes of CO on Pt(111)
[1.42], the observation of novel intermediates from methanethiol (CH 3 SH)
decomposition on Pt(111) [1.43], and the coverage dependent phase transition
of pyridine on Ag(lll) [1.44]. The latter study is especially noteworthy in that it
 1.2 Development of the NEXAFS Technique 7

was carried out in real time during continuous molecular adsorption on the
surface and therefore sets the stage for kinetic studies by means of NEXAFS.
An important development in NEXAFS instrumentation occurred during
1985-1986 with the first demonstration of fluorescence detection, first at the
sulfur K-edge (2470eV) [1.45, 46] and then at the carbon K-edge (285eV) [1.47,
48]. This enabled NEXAFS studies at ambient pressure under reaction condi-
tions [1.49].
One major development since 1986 lies in the application of NEXAFS and
also ISEELS [1.50-58] to increasingly more complex molecules, including thin
polymer films on surfaces [1.59-69] and the understanding of such spectra
through detailed calculations [1.70, 71]. The message emerging from these
studies is that the spectra of complex molecules can often be explained in a
simple building block picture by the superposition of the spectra of diatomic or
larger functional subgroups [1.71-73].
Another significant advance, made possible by the development of novel soft
X-ray monochromators [1.74, 75], is the study of condensed, physisorbed, and
chemisorbed molecules with high ( ~ 50 me V) energy resolution revealing vibra-
tional fine structure [1.76, 77]. Such studies also aided the identification of
magnetic effects in K -shell spectra, e.g., of the large exchange splitting of the (1 *
resonance in the O 2 molecule [1.78].
From all these studies it has become clear that NEXAFS is applicable to the
study of a large range of molecular systems, and in the following chapters the
principles, techniques, and applications of NEXAFS spectroscopy will be
presented.
2. Theory of Inner Shell Excitation Spectra

Below we review the fundamental aspects underlying the calculation of X-ray

absorption spectra. We emphasize approximations commonly made for the
calculation of transition energies and transition intensities and, in particular,
discuss two theoretical techniques which have been successfully applied for the
calculation of molecular X-ray absorption spectra.

2.1 Description of the X-Ray Absorption Process

2.1.1 The X-Ray Absorption Cross Section

The X-ray absorption cross section (Ix of an atom or molecule is defined as the
number of electrons excited per unit time divided by the number of incident
photons per unit time per unit area [2.1]. It therefore has the dimension
(length)2 and is usually given in units of cm 2 or barn (1 cm 2 = 1024 barn). The
cross section can be calculated from Fermi's "Golden Rule" for the transition
probability per unit time Pif from a state Ii) to a state If) driven by a harmonic
time-dependent perturbation V(t) = Ve- icot

(2.1)

where (!f(E) is the energy density of final states. In the case of a K-shell
excitation Pif is the number of electrons excited per unit time from the 1s shell
to a final state If) which, in principle, can be a bound or continuum state. Since
the term "cross section" usually refers to photoionization let us at this point
assume that If) describes a continuum state corresponding to a free electron in
the potential of an ion. The connection to bound final states, corresponding to
an excited electron trapped in the potential of an overall neutral atom or
molecule, will be made later.
In order to obtain the total X-ray absorption cross section at a given photon
energy we have to sum over all shells with binding energies less than hw.
However, since we are mainly interested in the partial absorption cross section
of inner shells and since all electrons in the outer shells have a smooth cross
section at inner-shell excitation energies, we shall for simplicity ignore the
underlying cross section contribution from the outer shells. The inner shell
 2.1 Description of the X-Ray Absorption Process 9

excitation is produced by an electromagnetic wave with electric field vector E

and vector potential A. It is common to work in the Coulomb gauge [2.2],
where the two quantities are related according to

E= _~ aA . (2.2)
e at
We can write the vector potential in the form of a plane electromagnetic
wave of wave vector k, frequency w, and unit vector e,

A = eAo cos(k. x - wt) = e ~o (ei(k'x - wt) +e- i(k·x - wt)) • (2.3)

The wave vector magnitude is related to the photon energy hw and the X-ray
wavelength A according to Ikl = hw/he = 2n/A. Equations (2.2) and (2.3) show
that A and E are collinear in space and their magnitudes are related by
Eo = Aow/e. The photon flux Fph associated with this plane wave, i.e., the
number of photons per unit time per unit area, is given by the energy flux of the
electromagnetic field [2.2] divided by the photon energy

A~w E~e
F h =--=--· (2.4)
p 8nhe 8nhw

In accordance with our earlier definition the X-ray absorption cross section is
then obtained as

(2.5)

Let us now evaluate Pif. The dominant perturbative term describing the
interaction of spinless particles of charge - e and mass m with an electromag-
netic field is given by [2.3 - 5]

e
V(t)=-A'p, (2.6)
me
where P = L,Pi is the sum of the linear momentum operators of the electrons.
Substituting (2.6) and (2.3) into (2.1), and realizing that only the time dependent
term e- irot in (2.3) causes transitions that absorb energy (the other term would
induce emissions ifthe atom had an inner shell hole), we obtain for the transition
probability per unit time

(2.7)

We can simplify (2.7) by retaining only the first term in the expansion of the
exponential, the dipole approximation. This important approximation assumes
k' x ~ 1 or Ixl ~ A/2n, where A is the X-ray wavelength. For excitation of the
10 2. Theory of Inner Shell Excitation Spectra

oxygen K -shell at hw = 550 eV the wavelength is A./2n = 3.6 A and Ix I, charac-

terizing the K-shell diameter, can be estimated from the Bohr radius ao = 0.53 A
and the atomic number Z as Ixl ~ 2ao/Z = 0.13 A, so that the dipole approx-
imation is well satisfied. We then obtain the final result for the X-ray absorption
cross section [2.1, 5J,
4n 2 h 2 e2 1
Ux "";-2--h -h
mew
l<fl e ·pli)1 2 (>,(E). (2.8)

The expression for the energy density of final states, (>,(E), depends on the
normalization of the wavefunctions. Bound states have unit normalization while
continuum states are normalized to a Dirac delta function in the kinetic energy
E of the photoelectron [2.6]. Although it is known that the plane wave
description of the photoelectron is oversimplified and does not give accurate
cross sections, especially near threshold [2.7J, let us use it, nevertheless, to
illustrate the normalization of continuum states and the dimension of UX' For
simplicity we evaluate (2.8) in a one-electron model (Sect. 2.5.3). We write the
photoelectron wavefunction as

If> = L!/2 exp(iq'r), (2.9)

where q is the wavevector of the photoemitted electron and the function is

volume-normalized in a large cubic box of length L. This volume normalization
of the free electron wavefunction is similar to that ofthe initial core state Ii> such
that the dipole matrix element in (2.8) simply has the dimension of p2. For the
plane wave given by (2.9) the expression for the density of final states per unit
energy is given by the familiar expression [2.8J
L3 mq
(>f(E) = (2n)3 h2 . (2.10)

A quick check of the dimension of U x according to (2.8) verifies that the cross
section is indeed given in units of an area. The first term 4n 2 h 2 /m 2 has the
dimension (energy x length 2 /mass), the second e2 /hc = 1/137.04 is the dimen-
sionless fine structure constant, the third l/hw and the last (>,(E) have the
dimension (l/energy), and the matrix element has the dimension of p2, or
(energy x mass).
Other forms of the dipole matrix element in (2.8) can be obtained by use of
operator equivalents for the total linear momentum operator [2.9J
'h im(E f - Ea ih 17V (2.11)
P = mv = -1 J7 = r = ",
h Ef - E;
where v and r are the sum of electron velocities and positions, J7 is the sum
of the gradient operators and V is the potential energy represented by the
electron-nuclear attraction and electron--electron repulsion terms in the Hamil-
tonian of the atom or molecule. Sometimes the matrix element is written using
 2.1 Description of the X-Ray Absorption Process 11

the "dipole operator" p. = er. The dipole matrix element with one of the first
three operators is usually referred to as the "velocity" form, the matrix element
with the position operator is termed the "length" form and the last expression is
called the "acceleration" form. In principle, hw = E f - E;, but it should be
remembered that hw is an experimental quantity, while E f and E; are, in
principle, eigenvalues of an approximate Hamiltonian. Similarly, the matrix
elements for different operators are only identical if the wavefunctions are exact
eigenstates of the total electronic Hamiltonian. For approximate eigenfunctions
the matrix elements differ according to the quality of the wavefunctions in
different regions of space [2.9].

2.1.2 Optical Oscillator Strength and Sum Rules

For discussion of transitions to bound states and intensities of resonances it is
convenient to define a dimensionless quantity, the optical oscillator strength f
which is related to the X-ray absorption cross section according to [2.10]

df
(1'.(E) = c dE' (2.12)

where C = 2n 2 e 2 h/mc = 1.1 x 10 2 MbeV. Since fis the energy integral of the
cross section, the oscillator strength is a measure of the intensity of a resonance
and the intensities of bound state transitions are typically quoted as an
"f number",
2
f I<fle·pli)1 2
= -h-
mw
• (2.13)

In comparing calculated bound and continuum intensities and in their

comparison to measured spectra, care has to be taken in properly normalizing
them to each other. Since continuum structures are typically broader than
the instrumental resolution the calculated continuum cross sections can be
directly compared to experiment. The calculated oscillator strengths for bound
state transitions can be converted to cross sections by use of (2.12) and
introduction of an energy density [2.1, 10, 11]

2n 2 e 2 h
(1'x(E) = f (1b(E) , (2.14)
mc

where (1b(E) is the bound state energy density of final states. As the excitation
energy approaches the ionization potential (IP) the bound state cross section
merges into the continuum cross section through ever closer-spaced Rydberg
states as illustrated in Fig. 2.1a for K-shell excitation in helium gas [2.12]. At the
IP, (2.14) is formally identical to the continuum cross section given by (2.8) if we
set l?b(E) = (1f(E). In a cross section versus energy plot the height of a bound
state resonance is given by (1'. and the width by 1/l?b such that the intensity or
12 2. Theory of Inner Shell Excitation Spectra

Fig. 2.1a, b. Illustration of the Is -+ np Rydberg ser-

104 (a) 1s-np IP ies in helium and argon gas, merging into the ioniza-
tion continuum at the Is ionization potential (IP). (a)
He 24.6eV High resolution (FWHM :::: 20meV) electron energy
.~
m
103
"0 c:
loss spectrum of 25keV electrons, recorded in the
0>2
~ c: dipole excitation limit of small momentum transfer,
2- near the K-shell excitation threshold (IP = 24.6eV)
co g 10 2 in He gas [2.12]. (b) X-ray absorption spectrum of
~!:
u
0>
Ar gas near the K-shell threshold (IP = [Link])
W 10 [2.13, 14]. Also shown as dashed curves is a fit using
Lorentzians to describe the Rydberg resonances
at the IP. In both spectra the widths of the Ryd-
berg peaks originate mainly from the final state
I lifetime, the instrumental broadening being small.
(b) I
[Link] Note logarithmic intensity scale in (a) and linear scale
c: I
0 Ar in (b)
I
~c I
o~ ._
0>
I
m u
.n .- I
ct::::
0> 0.5
>-0
taO
a:I
x

0
-6 -4 -2 0 2
Excitation Energy - IP (eV)

area of the peak is equal to the oscillator strength. Discrete transitions are often
very narrow and the measured peak height depends on the lineshape and width
of the peak. This is clearly revealed by a comparison of Rydberg transitions in
the helium and argon [2.13] K -shell spectra shown in Fig. 2.1. In both spectra
the width of the discrete resonances is mostly determined by the Lorentzian
lifetime width ofthe electronic final state, the Gaussian instrumental broadening
being small. The larger lifetime width for argon [2.15,16] thus leads to a reduced
Rydberg peak height relative to the continuum step at the IP. Comparison of
experiment with theory therefore necessitates conversion of the calculated
oscillator strength into an appropriately shaped peak keeping the area of the
peak, i.e., the oscillator strength, constant.
The discrete and continuum oscillator strengths satisfy some important sum
rules [2.10, 17, 18], the best known of which is the Thomas-Reiche-Kuhn sum
rule. It states that for a given electron in an atom or molecule the sum of the
oscillator strengths of all transitions to all other states, discrete and continuous,
occupied and unoccupied, is unity. It then follows that the total oscillator
strength for the electronic excitation of an atom or molecule is equal to the
number of electrons N in the atom or molecule, i.e.,

(2.15)
 2.2 Time Scales in Inner-Shell Excitations 13

The first term in (2.15) represents the intensities of the discrete resonances below
the IP, i.e., absorption edge, arising from the various electronic shells of the
atom or molecule while the second term is the intensity above the IP. Hence if
one plots df(E)/dE as a function of energy E, including the discrete part of the
spectrum, the area under the curve is proportional to the number of electrons in
the atomic or molecular system. Equation (2.15) is also approximately valid for a
given subshell of an atom, e.g., the K-shell. In this case the integration in (2.15)
extends over energies above the respective IP of the subshell, after subtraction of
the background from lower binding energy shells, and N is the number of
electrons in the subshell. Wheeler and Bearden [2.17] have considered the K-
shell oscillator strengths of low-Z atoms and pointed out that, in practice, N is
less than the expected value of 2 because excitations to occupied valence orbitals
are excluded by the Pauli principle. Deviations from the values expected from
the sum rule have also been noted by other authors [2.19, 20].

2.2 Time Scales in Inner-Shell Excitations

The theoretical description of the X-ray absorption process in atoms and

molecules requires knowledge of electronic energies and wavefunctions. For an
explanation of the most detailed features nuclear vibrations must also be
considered. Before entering into a mathematical treatment let us first consider
the time scales involved in electronic excitations and nuclear motions. Sudden
vertical electronic excitation caused by the electric field of a photon occurs in the
time that it takes the photon to travel the diameter d of the inner shell. With
c = 3 X 10 18 A/s and d ~ 0.1 A for the K-shell of oxygen we can estimate that the
excitation takes place in about 10 - 17 to 10- 18 s.

2.2.1 Electron and Hole Lifetimes

The excitation leads to a molecule in a nonequilibrium state subject to electronic
rearrangements and nuclear motions. Resonances in inner shell excitation
spectra correspond to an excitation from an initial to a resonant final state. The
full width of the resonance r is determined by the lifetime of the final state 't
according to the Heisenberg uncertainty principle r ~ h/'t. In discussing X-ray
absorption spectra it is convenient to separate 't into two contributions, the
resonance "lifetime" of the excited electron within the molecular potential 't e and
the lifetime of the inner-shell hole state 'th, according to

1 1 1
-=-+-. (2.16)
't 't e 'th

Note that the separation of the core hole and electron lifetimes used in (2.16)
relies on a somewhat simplified model. For example, there is evidence that the
14 2. Theory of Inner Shell Excitation Spectra

core hole lifetime in a given molecule depends on the excitation energy

[2.21, 22].
The inner-shell hole lifetime can be estimated in two ways. Using high
resolution X-ray absorption spectra, we can estimate 'h from the measured
lifetime width r of a bound state resonance since such a transition is character-
ized by 'e = 00. The measured width r is related to the hole lifetime by the
uncertainty relation 'hr ~ h = 6.6 x 1O- 16 eVs. Alternatively, we can use
photoemission spectroscopy [2.23]. In the limit of negligible instrumental
broadening the width of the inner-shell photoemission peak, i.e., the measured
kinetic energy distribution of the inner-shell photoelectron, is also determined
'h
by according to the above uncertainty relation.
Hole lifetime widths of low-Z atoms for the K and L3 shells have been
estimated by Parratt and Brown [2.14,24] and Krause and Oliver [2.15,16] and
are shown in Fig. 2.2. Note that the bound state lifetime widths correspond to
the total hole lifetime according to l/'h = l/'a + 1/,J where for low-Z atoms the
Auger lifetime 'a is much shorter than the fluorescence lifetime, J [2.15]. If we
use a hole lifetime width of r ~ 0.1 eV for the K-shell of a nitrogen atom
we obtain a characteristic lifetime in the 1O- 15 _1O- 14 s range. In contrast
to bound state resonances, the width of continuum resonances is determined by
'e' the lifetime of the resonantly trapped electron. Using a value of 10eV for
a characteristic continuum resonance width we obtain a lifetime in the
10- 17_10- 16 s range. Hence the longer lived core hole will influence the trapped
photoelectron.

Natural Width 01 Core Levels

Mo

:> 1.0
~
.c
'0
3
-.; Fig. 2.2. Full width at half maximum
>
for the K and L3 levels of the lighter
'"
..J
elements as a function of photon en-
I
I
ergy as estimated by Parratt [2.14]
I
I
I
and plotted by Brown [2.24]. The
I
dashed lines are semiempirical values
0.01 ~"""--'---'--J._"--..l...--'---'----'_'--"""--'----'
o 400 800 1200 1600 2000 2400 taken from the tables of Krause and
Photon Energy (eV) Oliver [2.15, 16]

2.2.2 Separation of Electronic and Nuclear Degrees of Freedom

Compared to the total lifetime of the electronic final state, the vibrational
motions of the nuclei are slow because of their large mass (the ratio of proton
and electron masses is 1836) and typically occur in approximately 10- 13 s. This
 2.3 The Electronic Ground State 15

allows separation of the electronic and nuclear degrees of freedom, the

Born-Oppenheimer approximation [2.25], and solution of the Schrodinger equa-
tion for a fixed geometry (usually the ground-state equilibrium geometry) of the
molecule. Furthermore, the Franck-Condon principle [2.26] can be employed,
which states that the internuclear distance can be assumed to be constant during
the fast electronic excitation process.
We shall now consider electronic wavefunctions and energies needed to
calculate the X-ray absorption spectrum. They are, of course, obtained as
solutions of the Schrodinger equation for the molecular potential. Of particular
interest are resonances in the cross section near the K-shell excitation threshold.
Such resonances arise at certain transition energies which correspond to the
promotion of a Is core electron from the molecular ground state to a bound or
quasi-bound excited state. Calculation of these spectral resonance features
according to (2.8) requires the following: (1) the wavefunction of the molecule
in its ground state, (2) the wavefunctions of the bound and/or quasi-bound
(continuum) final states, and (3) the energy differences between the ground state
and the various resonant excited states. The wavefunctions determine the
transition intensities via the dipole matrix element while the energy differences
give the transition energies. Below we discuss how the necessary energies and
wavefunctions can be obtained. We shall start by discussing the electronic
ground state and then later deal with the more complex aspects of excited states.

2.3 The Electronic Ground State

2.3.1 The Hartree-Fock Method
According to the Born-Oppenheimer approximation, the total wavefunction for
a molecule consisting of N electrons and P nuclei can be separated into a
product of an electronic and a nuclear part,
(2.17)

and the Schrodinger equation for the electrons is

(2.18)

Here the Hamiltonian, in which we ignore relativistic and spin-orbit effects, is

written in an internal coordinate system such that the nuclear kinetic energy
term does not enter and is given by

+ L L zzlie
P P
I
2
(2.19}
1=111>1 Rill
16 2. Theory of Inner Shell Excitation Spectra

The different sums are the electron kinetic energy, the electron-nuclear attrac-
tion, the electron-electron repulsion, and the nuclear-nuclear repulsion terms,
and m is the electronic mass, e the electronic charge, and Z,e the charge ofthe lth
nucleus. The denominators rin = Iri - Rnl, rik = Iri - rkl, and R'n = IR, - Rnl
specify distances between electronic (rJ and/or nuclear (Ri) coordinates. The
solutions lJIe and Eo of(2.18) are parametric functions of the nuclear positions. If
the Schrodinger equation is solved as a function of the internuclear separations,
Eo = Eo(R'n) can provide the potential in which the nuclei move. The solution
to a second Schrodinger equation for the nuclear motion then yields vibrational
energies EVib and wavefunctions ifJvib(R) for the molecule as shown in Fig. 2.3.
For example, for a diatomic molecule the total energy as a function of inter-
nuclear distance, Eo(R), often approximates the form for a harmonic oscillator.
Then

(2.20)
where Wvib is the classical vibration frequency and v = 0, 1, 2, ... is the vibra-
tional quantum number. Rotational splittings are typically too small to be
resolvable in K-shell spectra. For excitations of core electrons the equilibrium
internuclear distance in the electronic ground and excited states may be slightly
different as shown in Fig. 2.3. According to the Franck-Condon principle, in a
vertical transition only those vibrational levels vf of the excited electronic state
can be reached whose eigenfunctions have nonvanishing overlap with the
wavefunction of the lowest vibrational state of the electronic ground state Vi = O.
A common starting point for obtaining reasonably accurate wavefunctions
for electrons in ground states of atoms and molecules is the Hartree-Fock (HF)
self-consistent-field (SCF) method [2.27]. The HF equations are derived from the
quantum mechanical variation principle. lJIe is approximated by a single Slater
determinant IJI of N orthonormal one-electron spin-orbitals <Pi [2.28],

(2.21)

If the expectation value of the total energy,

(2.22)

is calculated from an approximate solution IJI of the SchrOdinger equation then

Eo is always greater than the exact ground state energy Eo for the Hamiltonian.
The difference

(2.23)

is called the correlation energy and reflects the fact that correlation, i.e., the
dependence of the motion of one electron on the position of a second electron,
cannot be represented by a simple product of functions of individual electron
 2.3 The Electronic Ground State 17

a::
a
w
>-
Ol
Q;
c:
w
iii
~
o
Q)

Cl.

Internuclear Distance R

Fig. 2.3. Illustration of the Franck-Condon principle. Through X-ray absorption a molecule X Y
undergoes an electronic transition from an initial state (the K-shell) to a bound final state (a Rydberg
or unfilled valence orbital). For the initial and final electronic states the potential in which the nuclei
move is given by the energies Eo( R) obtained from the Schrodinger equation (2.18) for the molecular
ground (XY) and the core excited (XY*) states, respectively, as a function of the internuclear
separation R. In case A the potential for the ground and excited states have a minimum at the same
internuclear distance R, while for the excited state in case B the minimum falls at a larger R value.
The Franck-Condon principle states that during an electronic transition the internuclear distance
can be assumed to be constant. Therefore only those vibrational states Evib ( vf) of the excited
electronic state can be reached in a "vertical" electronic transition from the vibrational ground state
Evib ( Vi = 0) whose wavefunctions I/Ivib( vf, R) have a finite overlap with those of the electronic
ground state I/Ivib(V i = 0, R). The overlap integral is the Franck-Condon factor given by (2.49). The
range of internuclear distances spanned by the ground state vibrational wavefunction is shown
shaded. For case A, maximum wavefunction overlap occurs for the vf = 0 excited state, while for the
higher vibrational states the overlap is nearly zero, owing to similar positive and negative
contributions. For case B, the overlap is largest for the first excited vibrational state vf = 1 with
finite, but smaller, contributions from vf = 0 and vf = 2. Note that vibrational splittings in X-ray
absorption are only observed iQ cases where the electronic lifetime r, given by (2.16), is sufficiently
long that the lifetime width is smaller than the separation between the vibrational levels of the
excited state

coordinates as given by (2.21) [2.29]. If by variation of the 4J/s the total energy
Eo is minimized we obtain N Hartree-Fock equations. The equation for the ith
electron at site x 1 with spin orbital 4Ji( x d is given by

(2.24)
18 2. Theory of Inner Shell Excitation Spectra

For a molecule consisting of P different nuclei, the Fock operator F is given by

[2.30]

(2.25)

where the Coulomb operator Jk , defined by the operation

(2.26)

gives the average repulsive Coulombic potential energy between all the electrons
in the system and the electron at position Xl' and the exchange operator, defined
as

(2.27)

takes account of the spin correlation according to the Pauli principle. This latter
term is a consequence of the tendency of electrons with like spins to avoid one
another, lowering, on the average, the Coulomb repulsion between them. The
wavefunctions depend on spin such that the terms in (2.27) vanish unless tPi and
tPk correspond to spins in the same direction.
For spectroscopists, it is a special form of the HF equations, namely the
diagonal or canonic form, which is offundamental importance. Fortunately it is
always possible to find a unitary transformation of the orbitals tPi under which
the Fock operator, the total wavefunction P, and the total energy Eo are
invariant, such that the right-hand side of (2.24) is brought into diagonal form
[2.31]. This yields the canonical HF equations

(2.28)

The orbitals tPi are called canonical orbitals and they have an orbital energy Ei.
In practice, the HF equations are solved in the canonical form and for this
reason we shall use the term "molecular orbitals" or "MOs" in the sense of
canonical molecular orbitals. The significance of the canonical HF equations for
the interpretation of spectroscopic data will be discussed in more detail later.
Since Jk and Kk depend on all tPi'S, the integrodifferential equations (2.28)
have to be solved self-consistently. This involves the evaluation of matrix
elements of the Fock operator,

N
= Hii + L (J ik -
k=l
K ik ) , (2.29)
 2.3 The Electronic Ground State 19

where we have introduced the following integrals:

(2.30)

2
lik = JJ<p~(Xd<P:(X2) ~
r 12
<Pt(X 2)<Pi(x 1 )dx2dx 1 , (2.31)

and
2
Kik = JJ<p~(Xd<P:(X2) ~ <Pi(X 2)<Pk(xddx2dx 1 • (2.32)
r 12

lik is called the two-electron Coulomb integral and Kik the two-electron exchange
integral. The iteration process is terminated when the difference in total energy

(2.33)

between two consecutive iterations is less than some threshold value. Note that,
because of the factor 1/2 in front of the Coulomb and exchange integrals in
(2.33), Eo cannot be obtained from a sum over the canonical eigenvalues Ei given
by (2.29), i.e., from photoemission binding energies (see below).

2.3.2 Roothaan-Hall and Semiempirical Methods

The most time-consuming part of a HF-SCF calculation is the evaluation of the
two-electron Coulomb and exchange integrals defined above. It is apparent that
the practicality of the HF method depends on the complexity (e.g., the types of
mathematical functions) and number of the spin-orbitals <Pi' For all but the
smallest molecules the numerical integration of the Fock matrix elements, which
needs to be carried out at each cycle of the iterative procedure, becomes very
impractical. For this reason, a method due to Roothaan [2.32] and Hall [2.33] is
used. In the Roothaan-Hall HF method the molecular orbitals <p;'s are ex-
panded in a basis set of M atomic orbitals X, according to
M
<Pi = L ailX,'
1= 1
(2.34)

The atomic orbitals, themselves solutions of the HF equations for the

individual atoms, are regarded asfixed and the variation in the HF procedure is
performed on the expansion coefficients ail' Thus the set of integrodifferential
equations of the Hartree-Fock formalism is reduced to a set of linear equations.
The physical justification for the above linear combination of atomic orbitals or
LCAO method is two-fold. Firstly, in the limit that the molecule is pulled apart
the MOs must converge smoothly into the set of atomic orbitals of the atoms.
20 2. Theory of Inner Shell Excitation Spectra

Secondly, the functional form of the MOs near the nuclei must be similar to the
functional form of the atomic orbitals of the individual atoms. In the limit of a
complete basis set the molecular orbitals found by this method approach those
obtained by the rigorous HF method.
It is clear that an important part of the Roothaan-Hall HF method is the
selection of a finite basis set which is flexible enough to accurately describe all
possible molecular orbitals [2.34J. In practice, the following types of basis sets
are used. The simplest, the minimal basis set, consists of one type of basis
function for each type of atomic orbital that is occupied by electrons in the
ground state of the atoms. As basis functions Slater-type orbitals or Gaussian-
type orbitals are usually used [2.27, 35]. A minimal basis set typically yields only
a qualitative picture ofthe molecular properties. A better description is provided
by use of a double-zeta [2.36J or triple-zeta basis set which consists of two or
three variationally independent basis functions for each occupied atomic orbital,
i.e., two or three times as many functions as in the minimal basis. Sometimes a
different set of basis functions is used for the inner (core) and outer (valence)
orbitals or for occupied and virtual (empty) orbitals. In particular, for the
description of virtual orbitals polarization and diffuse functions may be added.
Polarization functions provide higher angular momenta than needed for the
ground state of the individual atoms, e.g., p functions for hydrogen or d
functions for the second row elements boron through fluorine, and diffuse
functions have small exponential coefficients and therefore a large spatial extent.
For large molecules and large basis sets even the Roothaan-Hall HF method
leads to very time consuming calculations because approximately n4 two-
electron integrals of the form
2
JJxi(xdx;(X2)~
r 12
x,(x 2)xk(x 1 )dx 2dx 1 , (2.35)

need to be calculated, where n is the number of terms in the basis. One approach
to overcome this problem is to use empirical data to obtain some ofthe integrals
and ignore most of the rest, as done in the semiempirical methods such as
MINDO, CNDO, etc. [2.37, 38]. However, the accuracy of such methods is
often questionable and for many cases no good empirical parameters are
available. For this reason we shall not discuss or use semiempirical calculations
here.

2.4 Transition Energies

In the previous section we discussed techniques for obtaining the electronic

ground state energy and the Is initial state wavefunction or orbital. We shall
now take the next step and consider the energetics of electronic excitations. At
 2.4 Transition Energies 21

this point we shall restrict ourselves to bound state transitions in which the atom
or molecule is not ionized. The case of ionization, i.e., transitions to continuum
states, is discussed later. Before we continue, another point needs to be made. In
the following we shall often use the terms "states" and "orbitals". In principle, a
"state" is an observable, while an "orbital" is a theoretical construction. We shall
use the terms in this sense, noting that in quantum mechanics orbital configura-
tions are a convenient way of describing electronic states.
An important quantity in the discussion of K -shell excitation spectra is the
Is ionization potential, defined as the minimum energy necessary to excite a Is
electron to the continuum of states above the vacuum level. The ionization
potential IP(i), or binding energy Eb(i), associated with a particular electron i in
an atom or a molecule is conveniently measured by photoelectron spectroscopy
(PES) [2.23] as the difference between the exciting photon energy lim and the
kinetic energy of the photoelectron E kin , i.e., the energy position of the peak in
the PES spectrum,

IP(i) = Eb(i) = lim - Ekin • (2.36)

In K-shell X-ray absorption, the Is IP is the threshold energy (usually

characterized by an absorption step) for transitions to continuum states. In
contrast to PES, which directly measures the "free" photoelectron at photon
energies larger than the IP, the lowest energy transitions observed in X-ray
absorption spectroscopy occur at photon energies less than the IP. These bound
state transitions involve final states below the vacuum level. Such states arise
naturally as solutions of the Schrodinger equation for atoms and molecules and
they correspond to the well-known Rydberg orbitals which give rise to the sharp
pre-edge peaks in Fig. 2.1. Similarly, in core excited molecules unfilled or empty
(virtual) molecular orbitals can have orbital energies below the vacuum level
and transitions to such orbitals also require excitation energies less than the Is
IP (see later).

2.4.1 Koopmans' Theorem

The question arises how one may calculate the IP or the transition energies
associated with bound state transitions. In principle, this is straightforward if we
ignore electron correlation. All we need to calculate is the difference in total
energy between the initial (ground) and final (excited) electronic states according
to (2.33). For a bound state transition to a Rydberg state, an electron is removed
from the Is shell and inserted in a Rydberg orbital in the final state calculation;
for the calculation of the IP the final state is assumed to be that of an ion with an
electron missing from the K-shell. This scheme, usually referred to as ASCF, is
cumbersome and, in practice, approximations are employed which yield suffi-
ciently accurate results in a single calculation. These are discussed below.
The most famous approximation, Koopmans' theorem, links the IPs or
binding energies measured for valence and core electrons by PES to the orbital
22 2. Theory of Inner Shell Excitation Spectra

energies Ei calculated by means of the HF equation (2.28) for specific atomic or

molecular orbitals. It states that the binding energy or IP of an electron i is equal
to the negative of its orbital energy, E i • Koopmans' theorem is derived by
considering the difference in total energy between the ionized species (E et ) with
an electron missing from orbital i and the neutral species (Eo), which by
definition is the ionization potential IP(i). In contrast to the rigorous L\SCF
approach, however, Koopmans' approximation assumes that upon excitation,
i.e., removal of an electron from some orbital fiJi' none of the other fiJkS change,
i.e., they remain frozen. Then by use of (2.29) and (2.33) the difference in total
energy, i.e., the ionization potential, can be shown to satisfy the relation

IP(i) = Eet - Eo = - Ei . (2.37)

This result also follows logically from the definition of Ei (2.25, 28) as the sum of
the kinetic energy, the potential energy in the nuclear field and the average
interaction energy with all other electrons. For bound states, Ei is negative such
that the IP and the binding energy are positive.
For K-shell excitations of the second row atoms the frozen-orbital assump-
tion made in the derivation of Koopmans' theorem is inadequate and the
predicted binding energies err by 10-20eV [2.39-41]. To account for this error
it is common to introduce a relaxation energy correction L\ER in the comparison
of Ei with experimental binding energies. Other shortcomings of this approach
are the neglect of relativistic effects, accounted for by a correction Er [2.42] and
of correlation effects, except those introduced by the Pauli exclusion principle
(2.27). The correlation energy correction L\Ecor accounts for the difference in
energy stabilization of the final and initial states through configuration inter-
action [2.29], which would require description ofthe states by a linear combina-
tion of Slater determinants. Finally, for open shell systems multiplet structure
cannot be ignored and it is accounted for by a term L\Em' which can be
calculated from tabulated Slater integrals [2.43]. We can then write the follow-
ing expression linking experimental binding energies and theoretical quantities:

(2.38)

For the K-shells of carbon, nitrogen and oxygen the various corrections
have been calculated and their approximate values are L\ER ~ 15eV [2.39-41],
L\Ecor ~ 1 eV [2.29, 41], and L\Er ~ O.2eV [2.42]. For the open shell molecules
NO and O 2 the multiplet splitting of the Is photoemission peak is 2L\Em ~ 1 eV
[2.44]. We see that the relaxation correction is especially large and must be
accounted for in the calculation of K -shell spectra.
One method for including relaxation uses the so-called equivalent cores
approximation [2.45]. Here, it is assumed that outer electrons are affected by
ionization of a core electron in essentially the same way as they would be if the
nuclear charge were increased by one unit, i.e., by the "equivalent core" of the
next element in the periodic table.
 2.4 Transition Energies 23

2.4.2 The Transition State Method

A particularly important approximation in conjunction with calculations of K-
shell spectra by means of the X a multiple scattering method [2.46-48], dis-
cussed below, is the transition state or transition operator method, introduced by
Slater [2.46, 49], in which relaxation effects are treated by second order
perturbation theory. The technique was later generalized by Goscinski et al.
[2.50] and used in conjunction with conventional Hartree-Fock calculations
and by Williams et al. [2.51] to include higher-order perturbation terms.
To illustrate how the transition state method works, let us define occupation
numbers {n i } = (nl' n2 , ••• ,nN) for the N spin-orbitals <Pi' Following (2.33), the
total energy can be written as a function of the occupation numbers,
_ N 1 N N
EO(nl' ... ,nN) = if:l
niH jj +"2 if:l kf:l nink(Jik - K ik )

p P ZZ 2
+
,=L n>'
L 'R'ne n (2.39)
1

Let us now consider ionization of an electron from orbital <p,. In the initial
ground state the occupation numbers are {nil = (nl = 1, ... n, = 1, ... nN = 1),
while the final excited state is characterized by {ni} = (nl = 1, ... n, =
0, ... nN = 1). If we assume that there is no relaxation, the binding energy Eb(l)
is given by Koopmans' value,

(2.40)

Since all passive electron orbitals remain frozen as n, is changed, it is apparent

that in this approximation Eo, from (2.39), is a linear function of n, and therefore
(2.40) is equivalent to

Eb(l) = -E, = _ oEo(n,) . (2.41)

on,
Now let us allow for relaxation such that the total energy is minimized for
any excited state, characterized by a change in the occupation n, of the "active"
orbital <p,. We shall not only allow the extreme cases n, = 0 and n, = 1 but also
fictitious states described by a fractional occupation number. In the conven-
tional ~SCF formulation the relaxed binding energy, which according to (2.38)
is given by Eb(l) = - E, - LJER , is obtained as

(2.42)

where we have explicitly indicated with a superscript that all total energies are
calculated by allowing relaxation of the passive electrons. It is clear that in this
case the total energy is no longer a linear function of n,. For example, the
Coulomb and exchange integrals in (2.39) will change as n, changes. Let us
assume, however, that the total energy can be expressed by inclusion of the next
24 2. Theory of Inner Shell Excitation Spectra

higher term as a second order polynomial in n,. If we plot E~ as a function of n,

we see from (2.42) that the relaxed binding energy Eb(l) is given by the difference
of the values of the curve at n, = 0 and n 1 = 1, or alternatively, by the slope of
the chord connecting these two points on the curve. But for a parabola the slope
of a straight line between two points on the curve is equal to the slope of the
curve itself at midpoint. Thus in our case we obtain the simple result,

oEo(n,) I (2.43)
on, nr = 1/2 '

which accounts for relaxation to second order. In fact, higher-order corrections

can also be accounted for by taking the derivative in (2.43) at a different n, value
[2.51]. In practice, the relaxed binding energies are not obtained from (2.43) but
directly from the Hartree-Fock equations written in the transition operator form
[2.50]
h2 Z e2
vi - L _"- + L
P N
FT = - -2 nk ( Jk - K k ), (2.44)
m 11=1 rill k=l

where {nd = (nl = 1, .. n, = 1/2, .. nN = 1). In this method, the relaxed IP or

binding energy of a specific orbital is thus obtained in a single calculation by
adjusting the conventional Fock operator (2.25) to involve a fractional occupa-
tion number of that orbital.
The above discussion of the transition state method specifically referred to
the calculation of IPs or binding energies. In X-ray absorption, however, the
excited electron may remain bound in a Rydberg or unfilled molecular orbital
and the question arises as to how one calculates transition energies for such
"two level" cases, consisting of a hole in the Is orbital and an electron in an
"outer" orbital. In analogy to the "one level" case, the transition operator in
(2.44) is used with half an electron removed from the lower and half an electron
added to the upper orbital [2.52]. The transition energies are then simply the
differences of the "orbital energies" of the two levels. This approach, however,
involves a new calculation for each "outer" orbital.
To avoid this complication and to allow both bound and continuum state
calculations to be performed with the same potential, calculations may be
performed ignoring the excited electron and modifying only the oc~upation of
the Is orbital by one half. The effect of this simplification is illustrated in Fig. 2.4
for the case of the C K-shell excitation in CO, using an Xa. multiple scattering
formalism (see below) [2.53]. Here we have plotted the "orbital energies"
obtained by solving the X a. HF equations in the transition state form (2.44),
using the occupation numbers indicated in the figure. In case A, we have
calculated the IP (295.2 eV) of the C Is or 2a orbital in CO by removing half an
electron from the 2a orbital. We have also indicated the "orbital energy"
obtained for the empty 2n * orbital (- 7.9 eV) in the figure. In a second
calculation, case B, we have carried out a "two level" transition state calculation
by removing half an electron from the 2a and adding half an electron to the
 2.4 Transition Energies 25

A B
Fig. 2.4. XCl multiple scattering transition state calcu-
Vacuum Level
lation for the 20" and 21t· "orbital energies" in carbon
monoxide [2.53]. In case A the occupation number of the
20" orbital, i.e., the carbon Is orbital, was reduced by half
t 2.." an electron in the Fock operator (2.44), relative to the
>
~ I
ground state, and the resulting 20" "orbital energy" repre-
I sents the relaxed binding energy or IP (295.2eV) of that
>-E' I orbital. The 20" -+ 21t· transition energy, E~ = 287.3eV
I
'"c:
w I calculated for this case is close to that, E. = 285.7eV,
I calculated in the proper "two level" transition state for-
~
:e ~
I malism, case B, where half an electron has been removed
p I from the 20" and half an electron added to the 21t· orbital.
3 IP
I
IE~ E.. This justifies that, in practice, calculations are usually
en performed without adjusting the occupation number of
.g -280 I
I the upper orbital
·in
c:
I
I
~ I
I ~(2a)1Y,
o(.)
I

~(2a)1Y,

empty 2n* orbital. This calculation thus yields the Is -+ 2n* transition energy
(285.7 eV) as the difference between the two "orbital energies", - 3.1 eV for the
2n* and - 288.8 eV for the 20" orbitals. Note that in this case only the difference
in "orbital energies", not the energies themselves, are meaningful.
Thus in an attempt to properly account for the IP and the 20" -+ 2n*
transition energy one would use the IP = 295.2eV calculated for case A and the
transition energy E" = 285.7 eV calculated for case B. Comparison with experi-
ment shows that the calculated IP is 1.0eV lower and the calculated transition
energy 1.6eV lower than the measured values [2.54]. The most important point
about Fig. 2.4, however, is that the calculation for case A predicts the 20" -+ 2n *
transition energy within 1.6eV of the value calculated for case B (the proper
method), and therefore it justifies the use of scheme A for the calculation of K-
shell excitation spectra. In fact, in our example, the E" value calculated for case
A, labeled E~ in the figure, is fortuitously predicted to be exactly the experi-
mental value of 287.3 eV.
A final word of caution concerns the position of the 2n * orbital below the
vacuum level in the final, core excited state. This cannot be directly obtained .
from either calculation A or B in Fig. 2.4 but can be derived indirectly as the
difference between the 20" IP calculated for case A (295.2eV) and the 20" -+ 2n*
transition energy (285.7eV) calculated for case B. The value of -9.5eV ob-
tained thus is in good accord with rigorous ASCF calculations [2.55, 56] and the
9.7 eV 2n* binding energy of the equivalent core molecule NO [2.57]. A direct
way of deriving this energy by a transition state calculation would be to remove
a full electron from the 20" orbital and to allow half an electron to occupy the 2n *
orbital. This yields a transition state 2n* "orbital energy" of - 9.2eV [2.53].
26 2. Theory of Inner Shell Excitation Spectra

2.4.3 Localized Versus Delocalized Core Hole

Another question which needs to be addressed concerns the localized nature of
the core hole, the concomitant reduction of the molecular symmetry, and the
effect on the transition energies. This has been extensively discussed in the
literature [2.58-62], especially for molecules with several equivalent cores, e.g.,
benzene [2.63]. As shown by Cederbaum and Domcke [2.60], in principle,
localized as well as delocalized core hole calculations may provide similar
answers for the energies of molecular states because the relaxation energy of a
localized core hole is equal to the sum of relaxation energy and the change of
correlation energy due to relaxation of a delocalized core hole. In addition,
wavefunctions within the framework of the nuclear symmetry can be construc-
ted using either MOs for delocalized core hole states or a linear combination of
MOs for localized hole states.
In practice, it is easier to carry out localized calculations since the inclusion
of correlation effects requires sophisticated configuration interaction calcu-
lations. The difference between complete localization and delocalization of the
core hole during excitation into the a* shape resonance has been discussed by
Dill et al. [2.64] for the N2 molecule. For the calculation of the a* shape
resonance the lifetime of the trapped electron is shorter than that of the core hole
and therefore the hole will live long enough to influence the resonant photo-
electron. It was found that the cross section is relatively insensitive to the
treatment of the hole. However, the photoelectron asymmetry parameter was
found to depend on the localized or delocalized nature of the hole.

2.5 Transition Intensities

Let us now consider the wavefunctions needed to calculate X-ray absorption

cross sections according to (2.8). For molecules, the wa vefunctions Ii) and If)
and the momentum operator p will be given in a coordinate system fixed along
certain symmetry axes of the molecule; the matrix element depends on the
orientation of the X-ray polarization direction e in the molecular coordinate
system. While the effect of X-ray polarization will be discussed in detail in
Chap. 9, let us here for simplicity assume that the X-rays are linearly polarized
and the molecules are randomly oriented with respect to the polarization
direction e. The X-ray absorption cross section is then obtained by an average
over all molecular directions, and from (2.8) we obtain

Ux = C (}~~) ~ {1<fIPxli)1 2 + l<flp li)1 2 + l<fIPzli)1 2 }

y

= C (}~~) ~ l<flpli)1 2 , (2.45)

where we have defined the constant C = 4n 2/ie 21m2 c. In the following we shall
 2.5 Transition Intensities 27

for simplicity discuss only the dipole matrix element,

Dif = </lpli> . (2.46)

2.5.1 One-Electron Versus Multi-Electron Transitions

We must know what wavefunctions to use in the evaluation of the matrix
element. Electronic excitations are, in general, a multi-electron process since they
affect the total electronic configuration of the atom or molecule, and re-
membering that P = LPk is the sum of the momentum operators of the indi-
vidual electrons, see (2.6), we find that the matrix element in its general form is
given by

I
Dif = ( 'l'f(N) kt/k l'l'i(N) ) , (2.47)

where 'l'i(N) and 'l'f(N) are the Slater determinants for the initial ground state
and final excited state of the N-electron system, respectively. In particular,
'l'f(N) is a Slater determinant where the spin-orbitallPk that has lost an electron
has been replaced by a new spin-orbital characterizing the excited electron or by
a continuum function, and the other spin-orbitals have been allowed to adjust to
that change. To account for correlation effects it is insufficient to use a single
Slater determinant. If we allow electrons to occupy "excited" or virtual orbitals,
an infinite number of determinants can be formed and a "better" wavefunction
can be obtained by expansion in a series of such Slater determinants. This model
is called configuration interaction (el), and although it has a small effect on
binding energies as revealed by the relatively small correlation correction AEcoro
it is of importance for the calculation of the intensities of satellite peaks [2.41],
also called "shake-up", "many-electron", or "correlation" peaks [2.65]. In the
following we shall assume for simplicity that a single Slater determinant can be
used for the general description of the initial and final electronic states.

2.5.2 Effects of Nuclear Vibrations

In the calculation of the transition matrix element we can include the effects of
nuclear vibrations through the wavefunctions t/I~ib and t/ltib of the vibrational
states associated with the initial and final electronic potential (Fig. 2.3). The
total initial and final state wavefunctions, including the effects of vibrations, can
be written as a product of the electronic and vibrational functions, and neglect-
ing the influence of the perturbing X-ray radiation on the internuclear distance
R we can write

D!:fb =( I
'l'f(N)t/ltib(R) ktl Pkl t/I~ib(R)'l'i(N))
= ( 'l'f(N) Ittl Pkl'l'i(N)) <t/ltib(R)It/I~ib(R». (2.48)
28 2. Theory of Inner Shell Excitation Spectra

The squared overlap integral between the initial and final vibrational
wavefunctions

(2.49)

is the Franck-Condon factor which determines the intensities of the vibrational

peaks (Fig. 2.3). The intensities of electronic transitions shown between the
Vi = 0 electronic and vibrational ground state and the vibrational states vf of the
excited electronic state are determined by the overlap integral (2.49) over the
range of internuclear distances of Vi = O. We shall see below that in X-ray
absorption only resonances arising from bound state transitions are sufficiently
narrow that vibrational splittings may be resolved. The effect of the interplay
between the core hole lifetime and the vibrational time period for the correct
description of vibrational effects in core excited spectra is still the subject of
discussion in the literature [2.66-68]. In the following discussion of the dipole
matrix element Dif' we shall not continue to write down the vibrational overlap
term but it is understood that the full matrix element D!!;b defined by (2.48) does
contain such a term.

2.5.3 The Sudden Approximation

A great simplification in the calculation of the matrix element is achieved by a
fundamental assumption, known as the sudden approximation. Its essence lies in
assuming a strongly "one-electron" character of the electronic transition and
that the primary excitation event is rapid or "sudden" with respect to the
relaxation times of the other "passive" electrons. This allows separation of
the initial and final state wavefunctions into an "active" one-electron and a
"passive" multi-electron part, and the transition matrix element becomes

Dif = ( IJ'f(N) I kt1 Pk IlJ'i(N) )

= <xf lp114>1><lJ'f(N - 1)llJ'i(N - 1» + .... (2.50)

Here 4>1 is the orbital from which an electron is excited, Xf the wavefunction
of the excited electron, and the functions lJ'i(N - 1) and IJ'f(N - 1) are the
passive N - 1 electron remainders in the initial and final states. They are
(N - I)-dimensional Slater determinants in which one row and column corres-
ponding to 4>1 or Xf have been deleted [2.41]. In (2.50) we have omitted higher-
order terms which are associated with excitations of one or more of the "passive"
electrons [2.41]. In the sudden approximation we thus obtain as the leading
term a one-electron matrix element and an overlap integral between the passive
(N - I)-electron wavefunctions in the initial and final states. For the overlap
integral to be nonzero, the two functions lJ'i(N - 1) and IJ'f(N - 1) must
correspond to the same atomic symmetry or irreducible representation of the
symmetry point group of the molecule, the so-called monopole selection rule. The
 2.5 Transition Intensities 29

overlap term is always less than but close to 1 ( ~ 0.7-0.9) [2.69] and reflects the
intensity lost from the "main" one-electron transition by multi-electron effects.
Details of the various multi-electron excitation mechanisms and intensities
are contained in the higher-order terms omitted from (2.50). The form of this
equation suggests that there is a sum rule for the balance of the "one-electron"
and "multi-electron" intensities in the sudden approximation. The sum rule
states that the sum of the one- and multi-electron peak intensities for a given initial
state is equal to the peak intensity calculated by use of an unrelaxed final state
wavefunction. In the latter the matrix element stems from omission of the
passive-electron terms in (2.50), viz.

(2.51)

where the momentum operator is associated with the "active" electron. The
cross sections calculated using (2.51), called the active electron approximation,
are larger than those obtained in photoemission spectroscopy from the "main
peak" intensity alone, neglecting the satellite intensities [2.70].
In X-ray absorption spectroscopy two cases need to be distinguished. If only
the average cross section is of interest, e.g., all resonance effects are eliminated by
a smoothing procedure, then the measured cross section is well predicted by the
one-electron unrelaxed cross section since the measurement inherently sums
over all one- and multi-electron excitations. On the other hand, near edge
resonances or the EXAFS structure [2.71] are reduced in size by multi-electron
effects, similar to photoemission. The reason is that only those photoelectrons
which correspond to the "main peak" in photoemission are coherent and
therefore give rise to scattering resonances, while the electrons created in multi-
electron events are incoherent because of their smeared energy distribution.
A second sum rule, derived by Manne and Aberg [2.72], should also be
mentioned, and is best thought of in conjunction with core-level photoemission
spectra. In Koopmans' approximation all multi-electron effects are ignored and
the dipole matrix element describing the excitation of an electron in orbital 4Ji
has the simplest possible form given by (2.51). In this case a single photoemission
peak is observed corresponding to a binding energy Eb(i) = - E i • If we allow
passive electrons to relax and additional multi-electron excitations to occur, the
photoemission spectrum will exhibit a main "one-electron" line at a lower
binding energy Eb(i) = - E; - ..[Link] and higher binding energy multi-electron
satellites, i.e., shake-up and shake-off features. The sum rule states that there
exists a "lever arm" relationship between the one- and multi-electron peaks such
that the center of gravity of all photoemissionfeatures is at the Koopmans' binding
energy.

2.5.4 Adiabatic Versus Sudden Excitation

Qualification of the applicability of the sudden approximation in X-ray absorp-
tion is necessary. X-ray photoemission and EXAFS typically employ excitation
energies well above the IPs of the core electrons of interest and therefore the
30 2. Theory of Inner Shell Excitation Spectra

sudden approximation is justified. In near edge X-ray absorption spectroscopy,

the excited electron is either trapped in a bound state or leaves the atom or
molecule slowly. The appropriate physical picture, at least for transition ener-
gies close to the IP, is that the excited atom lowers its energy by slowly adjusting
to the effective atomic potential in an instantaneous, self-consistent way, the so-
called adiabatic excitation limit. The multi-electron statellites do not exist or are
of reduced intensity. In contrast to the second sum rule given above, which is
only valid in the sudden limit, the measured binding energy of the "main line" is
shifted from Koopman's value by a relaxation correction but "lever arm" multi-
electron intensities are absent.
Thus we have the remarkable situation that the IP measured in the sudden
limit by X-ray photoemission, i.e., the binding energy of the "main line", is identical
to the IP measured in the adiabatic limit by X-ray absorption, i.e., the position of
the "absorption edge". This has been carefully checked by comparing IPs
measured by photoemission with ISEELS results [2.73, 74], where the IP was
obtained by extrapolation of the Rydberg structures to an "absorption edge",
indicating the continuum onset as shown in Fig. 2.1. Further independent proof
is provided by zero kinetic energy electron spectra [2.75, 76] recorded with
tuneable synchrotron radiation where the IP in the adiabatic limit is directly
revealed as a pronounced peak.
To date the transition from "adiabatic" to "sudden" excitation is still not
well understood but it has been observed that satellite intensities are already
significant close to threshold [2.75-79] and the sudden limit appears to be
reached relatively fast, approximately within the first 10eV above the IP [2.77,
80,81]. Hence, in near edge X-ray absorption spectra multi-electron structures
are absent below the IP but they may be present above the IP, similar to
photoemission.

2.6 Bound Versus Continuum Final States

2.6.1 Improved Virtual Orbitals

For the calculation of K-shell spectra it is necessary to utilize techniques which
can account for transitions to bound and continuum final states. The energies
and wavefunctions describing the filled molecular orbitals of the N-electron
ground state of the molecule are obtained by solving the Hartree-Fock or
Roothaan-Hall HF equations as discussed earlier. Although solutions are also
obtained for empty or virtual orbitals, these do not represent an accurate
description of the excited or final state orbitals which are of interest in electronic
excitations. The final state of an electronic transition is described by the removal
of an electron from an inner orbital fjJi and addition of this electron to an outer
orbital fjJ f. In principle, the orbital fjJ f and its eigenvalue Ef should therefore be
obtained from a separate calculation for the excited electronic configuration. In
practice, this requires a new calculation for each final state fjJ f' a rather
 2.6 Bound Versus Continuum Final States 31

cumbersome task, and it is therefore common to use an approximate procedure

suggested by Hunt and Goddard [2.82]. In this method one neglects the
contribution of the excited outer shell electron to the molecular potential which
is then simply given by that of the ion, i.e., the (N - I)-electron system with an
electron missing in orbital41i. The Fock operator, see (2.25), in this, the so-called
static exchange, approximation is given by

(2.52)

where the summation is over doubly occupied orbitals "'k' not the previously
used singly occupied spin-orbitals 41k = "'kak (a k = ± 1/2). In deriving (2.52) we
have also made the restricted-Hartree-Fock assumption that all doubly occu-
pied orbitals are identical in their space parts such that there are exactly twice as
many nonzero Coulomb as exchange integrals. The orbital "'iis only singly
occupied and the sign of the exchange term depends on the singlet ( + K i ) or
triplet (- K i ) nature of the excited state. The empty orbitals '"f obtained as
solutions of F ex are commonly referred to as improved virtual orbitals and closely
correspond to the optical orbitals referred to in the experimental spectroscopy
literature.
Improved virtual orbitals have lower orbital energies compared to the
virtual orbitals obtained by Hartree-Fock ground state calculations (Sect. 2.1.1),
resulting from the additional Coulomb attraction of the excited (N - 1)-
compared to the ground-state N-electron system. Thus "unbound" virtual
orbitals with positive energy may correspond to "bound", negative-energy,
improved virtual orbitals. Note that excited electronic states may also be
described by the equivalent core model mentioned earlier, where the excited
atom with atomic number Z is replaced by a Z + 1 atom [2.63, 83, 84]. It should
also be noted that, in practice, the downward shift of the ground state virtual
orbitals caused by the core hole is often similar for the various orbitals such that
the relative ordering and intensities predicted by a ground state calculation may
be rather accurate.
By use of the ground-state Is core orbital and excited-state improved virtual
orbitals, it is straightforward to calculate the oscillator strengths or cross
sections of the discrete transitions according to (2.13) and (2.14). Such calcu-
lations may furthermore be carried out with great accuracy by using determin-
antal initial state and final state wavefunctions in the evaluation of the dipole
matrix element (2.47), or even include the effects of configuration interactions
[2.63,85].

2.6.2 Continuum Final States

In comparison, the calculation of the continuum cross section for molecules is
more difficult. At first sight this is not apparent since, in principle, virtual or
improved virtual orbitals with positive energy could be used, i.e., states which
32 2. Theory of Inner Shell Excitation Spectra

Orbital
Energy (eV)
co
15
..-,
"-;',
~, ~
N
'.
-~ '
3u~

, 10

5
2"'

50'

1"

40' ,.'-., Q.
t:,·_· c~~
':~,. :,;/

- 30

-35

30'

-45
Fig.2.S. Correlation between orbital energies and molecular orbitals for N2 and CO. The energies
of the filled and empty (indicated by an asterisk) orbitals are taken from Roothaan-Hall,
Hartree-Fock self-consistent field calculations with a Gaussian double-zeta basis set by Snyder and
Basch [2.36] ; the molecular orbital contour plots are taken from semiempirical calculations by
Jorgensen and Salem [2,86]

energetically overlap with the continuum of states above the vacuum level.
Examples are given in Fig. 2.5 for the simple diatomic molecules N 2 and CO,
where the orbital energies have been taken from Snyder and Basch's book
Molecular Wave Functions and Properties [2.36] and the orbital contours from
Jorgensen and Salem's The Organic Chemists Book of Orbitals [2.86]. Thus for
the two pictured cases one would expect resonances in the K-shell excitation
spectra corresponding to transitions to the virtual In: and 30': orbitals for N2
and the 2n* and 60'* orbitals for CO.
We have already encountered the 20' ---t 2n* excitation for CO in Fig. 2.4 and
have seen that for the core excited molecule the 2n * orbital is lowered below the
vacuum level, resulting in a bound state transition. On the other hand, we shall
 2.6 Bound Versus Continuum Final States 33

see below that the 1s -+ 60'* and 1s -+ 30': transition energies exceed the IP and
the corresponding resonances lie in the continuum. While MOs are useful in
visualizing transitions to quasi-bound resonant states embedded in the con-
tinuum, they do not represent a proper description of a continuum state suitable
for a cross section calculation. There are several reasons for this inadequacy.
First, the orbitals obtained from conventional bound state calculations obey the
orthonormality condition
(2.53)

while proper continuum or ionization states satisfy a Dirac delta-function

orthonormality in the kinetic energy E of the photoelectron [2.6], viz.,

(2.54)

Secondly, bound state and continuum functions satisfy different boundary

conditions in the asymptotic limit r -+ 00, where continuum functions approach
nearly pure plane wave character [2.87]. Finally, bound state functions cannot
account for the infinite degeneracy of the continuum at any energy E. Therefore
(improved) virtual orbitals do not provide stable and reliable solutions at
positive energies, and cannot be used without further development for the
calculation of ionization cross sections.
The difficulty in properly describing the continuum wavefunction of the
photoelectron for molecules lies in the nonlocal and noncentral nature of the
potential of the parent molecular ion. Ideally, one would use scattering theory
with appropriate boundary conditions and partial wave expansion techniques
[2.88] for the description of the motion of a low energy unbound electron in a
realistic potential of the molecular ion, but this approach has been impeded by
computational difficulties. For small molecules, progress has been made recently
in the construction of one-electron continuum functions which are solutions of
the static-exchange Hartree-Fock equation, fulfilling the proper normalization
and asymptotic boundary conditions for a free photoelectron [2.89]. Although
this technique may, in fact, be the technique of choice in the future we give brief
mention to it here.
Below we shall discuss two other techniques which have been extensively
used in the past for the calculation of K-shell spectra; much of our understand-
ing of the spectra is based on the results obtained with these techniques. The
first, so-called XtX multiple scattering (MS) method, [2.90,91] uses a simplific-
ation of the molecular potential, the so-called "muffin-tin" form, such that the
final state can be treated by customary partial wave expansions of scattering
theory [2.88]. The second, so-called Stieltjes-Tchebycheff (ST) moment theory
method, is based on first-principles molecular calculations. It constructs the
ionization cross section, using the theory of moments, from a large but finite
basis of eigenfunctions obtained by conventional Roothaan-Hall HF bound
state methods [2.92]. Below we shall outline the two approaches.
34 2. Theory of Inner Shell Excitation Spectra

2.7 The X~ Multiple Scattering Method

2.7.1 Introduction to the Method

The self-consistent field Xa-MS method, developed by Slater [2.46] and Johnson
[2.47] in the late 1960s, has been employed successfully for the calculation of the
electronic structure of molecules and clusters [2.48, 93-97]. The method is
capable of performing calculations for relatively large clusters (~50 atoms) by
employing two basic approximations: (i) the exchange potential in the
Hartree-Fock Hamiltonian is approximated by an average potential, deter-
mined by the total charge density, and (ii) the overall cluster potential is
approximated by a muffin-tin form. For this simplified form of the potential the
wavefunctions can be joined continuously throughout the various regions of the
cluster via multiply-scattered-wave theory (hence the name "multiple
scattering"). Originally the technique was used for bound electronic states but
following an early suggestion by Johnson [2.98] the method was extended by
Dill and Dehmer [2.99] and Davenport [2.91] in the mid 1970s to the description
of continuum final states, i.e., excited electronic states characterized by the
motion of a free photolectron in the field of an atomic or molecular ion.

2.7.2 Exchange Potential and Latter Tail

Let us briefly discuss the different steps carried out in Xa-MS calculations. First,
the one-electron Schr6dinger equation is solved for each individual atom or ion
which is part of the cluster. The calculation of the numerous two-electron
integrals is avoided by expressing both the Coulomb and exchange potentials
directly in terms of the total charge density of the other electrons. For the
Coulomb term this is straightforward but it is more difficult for the exchange
term. Following a density functional formalism introduced by Slater [2.30],
Gaspar [2.100], and Kohn and Sham [2.101], the exchange term given by (2.27)
can be written in the so-called "X a" form,
N
L Kkllp;(r»
k=l
= Vxa(r)llp;(r» . (2.55)

The exchange potential is given by

(2.56)

where
N
Q(r) = L nk¢t(r)¢k(r) (2.57)
k=l

is the local electron density (both spins) and {nd represents the occupation
 2.7 The XQ( Multiple Scattering Method 35

numbers of the various orbitals. The parameter IX is chosen on the basis of

atomic calculations by Schwarz [2.102]. Since IX is found to be close to 0.7 for all
but the lightest atoms, sometimes IX =0.7 is used for all atoms. In practice, the
charge densities calculated by Herman and Skillman [2.42], who used IX = 1, are
taken as the initial estimate for an atomic self-consistent-field calculation with
Schwarz's value of IX.
The failure of the XIX potential to correctly describe the Coulomb attraction
between the electron at large distance and the positive ion is sometimes
amended by applying a "Latter tail" correction [2.103]. This is simply done by
substituting the Coulomb potential e2 /r between the excited electron and the
positive molecular ion for the calculated potential VIII (see below) when the
magnitude of the latter drops below that of the Coulomb potential. Other forms
of the exchange potential can also be used, like the Hara exchange [2.104] which
is based on the free electron gas approximation [2.105]. Further refinements
include many-body effects, as in the complex Hedin-Lunqvist [2.106, 107]
exchange potential.

2.7.3 Muffin Tin Potential

In the second step the cluster is divided into regions as illustrated for a simple
diatomic molecule in Fig. 2.6. Each atom is first surrounded by a sphere with
radius bi and then the entire cluster is surrounded in a close-packed fashion by
an outer sphere with radius bo centered on the molecular center of mass and
creating regions Ii' II and III, respectively. The potential at an arbitrary point r
is then obtained as a superposition of the individual potentials of the atoms
centered at positions R i ,

V(r) = L V~(ri) + V~<z(rJ .

i
(2.58)

Here we have explicitly indicated that the local potentials are a sum of the
total (electron-nuclear plus electron--electron) Coulomb potential V~(ri) and
exchange potential V~ (rJ [compare (2.25) and (2.44)], and we have defined

Fig. 2.6. Partitioning of the molecular field for a hetero-

nuclear diatomic molecule. The corresponding muffin-tin
potential is spherically symmetric in regions 11 ,1 2 and III
and constant in region II. The coordinate vectors'l and'2
'j
are centered on the atoms, while, = R j + is centered on
the molecular center of gravity. IRti + IR21 is the inter-
nuclear distance
36 2. Theory of Inner Shell Excitation Spectra

r i = r - Ri (Fig. 2.6). The potentials in regions Ii are then expressed as an

expansion of V(r) in a series of spherical harmonics, using theorems described
by Liiwdin [2.108]. By retaining only the first (l = m = 0) spherically symmetric
term V'(r;) in regions Ii' the potential assumes a "muffin-tin" form. In region II
the potential is averaged to a constant value, often referred to as the "inner
potential" in the literature. In region III a spherical average is taken with respect
to the center of the molecule, resulting in the potential VIII (r).
The relative sphere radii around the atoms, i.e., their ratios, are typically
chosen, following a suggestion by Norman [2.109], to be proportional to the
volumes containing the atomic number of electrons. With the ratio of radii fixed,
the spheres around individual atoms are chosen to minimize the region II
between the spheres where the potential is least well defined. Often the spheres
are made to touch but they may be chosen to overlap. Wurth and Stohr [2.53]
have suggested that for the calculation of K-shell spectra the overlap may be
chosen in a semiempirical fashion. By performing a transition state calculation
with half an electron removed from the Is shell the overlap is adjusted such that
the calculated Is "orbital energy" relative to the vacuum level matches the
experimental IP.

2.7.4 Multiple Scattering Wavefunctions

In the third step the wavefunctions are constructed. In region Ii single

center wavefunctions are written in terms of solutions RI(E, ri) of the radial
Schrodinger equation with the potential V'(r i ), real spherical harmonics Yi(ri),
and to-be-determined partial wave coefficients cl. m according to

I/I(ri) = L CLmR,(E,ri ) Yi(r i ) .

I.m
(2.59)

To speed up the molecular calculation, symmetry adapted functions are used.

The functions are chosen such that they form a basis for one of the irreducible
representations of the molecular point group, either the nuclear point group
symmetry or a broken symmetry, e.g., assuming a localized core hole. In
practice, a computer program by Cook [2.110] can be used which determines the
appropriate basis functions automatically, with only the molecular geometry as
input.
The wavefunctions in region III also have the form given by (2.59) but are
centered on the molecular center and use radial functions which are solutions of
the SchrOdinger equation with the potential VIII(r). In region II a multicenter
partial-wave representation is used where the radial functions are expressed in
terms of spherical Bessel, Hankel and Neuman functions about the molecular
center and the atomic positions [2.47].
By use of expansion theorems the multicenter wavefunctions can be re-
written in terms of single-center wavefunctions around the molecular center and
distances between the centers. Because there must be outer sphere basis func-
 2.7 The XC( Multiple Scattering Method 37

tions for every irreducible representation the maximum angular momentum

quantum number, I, should be greater in the outer sphere region than on the
atomic centers. All regional wavefunctions and their derivatives are required to
be continuous at all spherical boundaries. These boundary conditions allow
mapping of a given partial wave near an atomic nucleus into the asymptotic
partial waves in region III [2.48] and lead to explicit relations between partial-
wave coefficients. These relations may be written as linear homogeneous equa-
tions which constitute the secular equations of the Xa-MS method.
In discussing the results from Xa-MS calculations it is important to dis-
tinguish between the local angular momentum states around an atom (region I)
and the partial wave label in the asymptotic molecular wavefunction (region III).
In X-ray absorption from a given initial state only those final states are allowed
in the "atomic-like" environment of the local region I which obey the dipole
selection rule ~I = ± 1, i.e., for an s initial state only the p-wave component of
the final state is allowed. On leaving the excited atom in region I the photo-
electron is scattered by the anisotropic molecular potential into the entire range
of angular momentum states contributing to a given, e.g., the (J, ionization
channel. The resonances calculated by MS theory are typically discussed in
terms of specific resonating 1channels which are the asymptotic labels in region
III, i.e., the corresponding wavefunctions are expanded about the center of the
molecule.
Of special importance for the understanding of the scattering processes are
the so-called eigenchannel wavefunctions obtained by diagonalizing the K -matrix
[2.99, 111, 112]. The partial-wave character of these functions remains un-
changed by scattering processes in the molecular potential and they can there-
fore be used to represent the final state in a manner similar to contour plots used
in conjunction with molecular orbitals [2.112]. These concepts will be discussed
in more detail for the N2 K-shell spectrum in Section 4.2.6.

2.7.5 Transition Energies

The orbital energies are obtained by finding the zeros of a secular determinant
[2.48]. Since the elements of the determinant contain the energy as a parameter
one must calculate the value of the determinant for a series of closely spaced
energies and ascertain where it changes sign. For core states the determination
of the orbital energies is simplified significantly by limiting the potential to the
appropriate atomic sphere. To obtain ionization potentials of resonable accur-
acy the transition state method discussed in Sect. 2.4.2 is used. The excitation
energy for a given electronic transition is obtained from a calculation in which
half an electron is removed from the K-shell. In principle, for bound-state
transitions half an electron should be added to the unoccupied orbital, with the
transition energy being the difference in orbital energies of the fractionally
occupied orbitals. However, this scheme requires a new calculation for every
excited state. In practice, the virtual orbitals may be left unoccupied since this
does not affect the transition energies significantly, as discussed in Sect. 2.4.2.
38 2. Theory of Inner Shell Excitation Spectra

2.7.6 Practical Procedures for Calculation of K-Shell Spectra

The set of wavefunctions obtained for the occupied orbitals after the first pass
leads to an electronic charge density from which a new potential of the form
given by (2.58) is generated. This potential is spherically averaged in regions Ii
and III and volume-averaged to a constant value in region II. An iteration is
then started by using a suitable combination of this new potential and the
original one to calculate new wavefunctions and eigenvalues. The entire nu-
merical procedure is repeated until self-consistency of the charge densities and
potentials is achieved.
In Fig. 2.7 is plotted the fully converged self-consistent potential V(r), (2.58),
calculated for the N 2 molecule in its electronic ground state under the assump-
tion of touching muffin-tin spheres. The potential is plotted along the inter-
nuclear axis and includes the nuclear positions. Note that the potential is
discontinuous at the sphere boundaries. We have also indicated the value
(-28.1 eV) of the "inner potential" VII. The small kink in the potential near Irl
= 1.3 A arises from attaching the "Latter tail" Coulomb potential at this point.
It can be seen that the "Latter tail" falls off more gradually with Irl than the
extrapolated XIX potential. Above we have deliberately used the symbollrl for
the distance from the molecular center to make clear that Irl is not a spherical
coordinate. If expressed in the spherical coordinate r the potential would look
quite different and contain a potential barrier arising from the centrifugal
potential term h2l (l + 1)/(2mr2) as discussed in Sect. 4.2.1. The sensitivity of the
calculated K-shell spectrum to the choice of the potential will be discussed in
Sect. 4.2.6.

Xa Potential for N2

III Fig. 2.7. Construction of muffin-tin

spheres and resulting self-consistent mo-
lecular potential for N 2' (2.58), as a func-
I tion of the distance Irl from the center of
I
I the molecule, calculated by the XIX-MS
Latter tail - I I 1 - Latter tail technique. The potential is plotted along
o I I I the internuclear axis, and as shown in the
I I I top part of the figure we have chosen
I
> -20 __ y:~_ I _J __: touching spheres with a radius around the
atomic centers of 0,55 A and an outer
~ I
,-: -40 I sphere radius around the molecular center
>" I
I
of 1.l0A. We have used maximum angular
iii momenta of Imax = 4 relative to the atomic
E., -60 I
centers and Imax = 7 relative to the center of
I
(5
Cl. I the molecule. The inner potential VII is
-80 I
I
calculated to be -[Link], The change of
the potential near Irl = 1.3 A is associated
- 100 '---:_2:-..J....-_~1_,,-I~~-"-~--'---2=----'

Distance from Molecular Center, r" (AI

with the "Latter tail" which is substituted
for the XIX potential at larger distances
 2.8 Ab Initio Stieltjes-Tchebycheff Molecular Orbital Method 39

The X-ray absorption cross section is calculated according to (2.8) and (2.14)
by use of the XIX-MS wavefunctions for continuum and bound states, respect-
ively. Rather than using the momentum operator form of (2.8), in practice, the
acceleration form given by (2.11) is used, as suggested by Davenport [2.113] and
Noodleman [2.114]. This has the advantage that VV = 0 in the interstitial region
II and the matrix element is simply the sum of terms over the atomic and outer
sphere regions. Bound-state transitions which occur at discrete energies and
transitions to continuum states require separate calculations but they can be
joined continuously by utilizing (2.14).
In this volume many calculated K-shell spectra will be presented. The
following computational strategy was used. (1) A broken molecular symmetry is
employed with half an electron in the Is orbital of a particular atom of the
molecule. (2) The sphere overlap is chosen such that the Is orbital energy
calculated by the transition state method matches the experimental IP. (3) The
same transition state potential is used for bound and continuum states. (4) The
calculated bound state oscillator strength is converted into a cross section by use
of (2.14). A Gaussian (Lorentzian) line shape function is used whose width
represents the instrumental (lifetime) resolution and whose area is the oscillator
strength. (5) The calculated continuum cross section is convoluted with the same
Gaussian or Lorentzian function used for the discrete part of the spectrum.
To date the XIX-MS method has been extensively applied to the calculation of
K-shell excitation spectra of simple and complex molecules and several reviews
of the technique itself [2.111,115] and the results obtained [2.53,96,97,116] are
available.

2.8 Ab Initio Stieltjes-Tchebycheff Molecular Orbital Method

2.8.1 Introduction to the Method

The second theoretical method used for the description of K -shell spectra of
molecules, developed by Langhoff et al. [2.92, 117-120], uses the concepts of
valence molecular orbitals (MOs), developed by Mulliken [2.121], and conven-
tional Roothaan-Hall HF computational techniques. By use of the Stieltjes-
Tchebycheff moment-theory techniques the continuum cross section is obtained
by smoothing discrete cross sections calculated by use of a finite number of
ground state and excited state MOs. This technique thus uses a highly accurate
molecular potential, in contrast to the XIX-MS technique, but avoids the difficult-
ies associated with the construction of proper continuum functions by use of
improved virtual orbitals and a smoothing procedure.

2.8.2 Calculational Procedure

Calculations begin by solving the Hartree-Fock equations for the ground state
electronic configuration. The resulting canonical Fock orbitals are then used to
40 2. Theory of Inner Shell Excitation Spectra

construct the Fock operator in the static exchange form (2.52), appropriate for
excitation of an electron from a selected orbital o/i' The two calculations yield
the eigenvalue Ei and eigenfunction o/i of the initial state orbital of the active
electron and M improved virtual orbitals 0/ j and the corresponding negative
(bound) and positive ("unbound") energies E j ' Typically, in the binding energy
calculation of the core state, relaxation is not explicitly included but is allowed
for by using experimentally determined 1s IPs rather than Koopmans' eigen-
values E i •
Since a fairly extended array of virtual orbitals is required to account for the
densely spaced Rydberg orbitals and the "unbound" orbitals above the vacuum
level, a large set of compact and diffuse basis functions is chosen for expansion of
the improved virtual orbitals while the occupied orbitals are described by a
modest number of valence-like basis functions. The set of M improved virtual
orbitals which are chosen to be Lebesgue square-integrable (so-called L 2 )
eigenfunctions and their eigenvalues form a pseudo-spectrum {o/ j' E j;
f = 1, ... ,M} from which discrete cross section values, compare (2.8, 14), can
be calculated according to

(2.60)

where hW ij = Ej - E i .
The pseudo spectrum identifies physically distinct final states which are
expected to dominate the photoabsorption cross section. It is evident, however,
that the wavefunctions do not satisfy the normalization and asymptotic bound-
ary condition of proper continuum states. Although, in general, the improved
virtual orbitals comprising the pseudo spectrum can be obtained by some linear
combination of continuum functions, it is not necessary to explicitly specify the
continuum functions in computing reliable ionization cress sections. Instead,
they are obtained by smoothing the discrete cross sections given by (2.60) by use
of theorems from the theory of moments [2.122, 123]. According to one theorem
the cross section is uniquely determined by its spectral moments,

J(hw)-kaAhw)d(hw) .
00

11k = (2.61)
IP

It is found that for a sufficiently large number M of improved virtual orbitals

the moments of(2.61) for k:::5 M/2 can be approximated by those obtained from
the discrete cross section according to

(2.62)

The Stieltjes-Tchebycheff approach is then used for the construction of the

ionization cross section from a finite number of moments. In principle, this
 2.8 Ab Initio Stieltjes-Tchebycheff Molecular Orbital Method 41

method corresponds to a finite-energy-interval averaging of the pseudospectrum

but in contrast to step-like "bin-smoothing" it produces a smooth oscillator
strength density with resonance-like structures at the appropriate discrete
transition energies. For a detailed discussion of the technique the reader is
referred to reviews by Langhoff [2.118-120].

2.8.3 Stieltjes-Tchebychelf Orbitals

Because the ST-MO technique uses improved virtual orbitals for the construc-
tion of the ionization cross section it is not surprising that the technique
identifies particular virtual orbitals that are associated with strong continuum
resonances. The orbital nature of continuum scattering functions is most
elegantly established by pictorial examination of so-called "dipole interaction
prepared states" or Stieltjes-Tchebycheff orbitals [2.120, 124-126], which allow
convenient graphical illustrations similar to bound state orbitals. The ST
orbitals closely correspond to the eigenchannel wavefunctions discussed earlier
in the context of XIX-MS calculations. They are constructed to specifically
identify the resonant channels in the eigenchannel functions, thus avoiding the
explicit construction of all the degenerate eigenchannel solutions, i.e., of the
entire spectrum of continuum functions. Note that these functions are construc-
ted by using the cross section as a normalization factor, and therefore this
procedure does not provide an alternative method for the construction of the
cross section itself.
Figure 2.8 compares the outer MOs of nitric oxide (NO) calculated by the
Hartree-Fock method in a minimal Gaussian basis set [2.119] with the ST
orbitals corresponding to ionization of the 50" MO at various excitation energies
e. At e = 25.2eV the ST orbital reveals a remarkable similarity (except for an
arbitrary phase reversal) to the empty 60"* MO obtained with the HF minimal
basis set calculation, indicating that the resonance observed experimentally at
this excitation energy [2.127, 128] arises from a 60"* orbital embedded in the
continuum. Note that the difference in orbital energies between the 50" and 60"*
orbitals of 39.2eV, corresponding to the 50" --+ 60"* transition energy, obtained
with the ground state HF calculation is much greater due to the neglect of final
state relaxation than the resonant excitation energy of ~ 25 eV obtained with
the static exchange Hamiltonian used for the ST calculation.
The connection between continuum shape resonances and minimal basis
set MOs has also been examined by Thiel et al. [2.129, 130] who suggested that
one may view this resonance as a two-step, one-electron process where a K-shell
electron is excited to an unoccupied virtual orbital of 0" symmetry and then
decays into the continuum as a photoelectron. The unified description of bound
state and continuum resonances as transitions to well-defined molecular final
states emerging from ST-MO calculations has many virtues for our understand-
ing of the overall characteristics of K-shell excitation spectra, and we shall use
this concept throughout.
42 2. Theory of Inner Shell Excitation Spectra

Minimal Basis set MO's 5a-Ea Stieltjes Orbitals

NO

Fig. 2.8. On the left are shown molecular orbitals for NO constructed in a minimal Cartesian
Gaussian basis set [2.119, 126). Calculated orbital energies are also shown. The orbitals are plotted
in a 3.2 A x 8.5 A plane that includes the internuclear axis, with N on the left and 0 on the right. The
6cr* orbital is empty, the 2n* orbital filled with only one electron and all other orbitals are occupied.
On the right are shown Stieltjes orbitals and excitation energies appropriate for 5cr -+ ecr ionization
of NO. The Stieltjes orbitals represent the scattering functions of an electron excited from the 5cr
orbital, shown at the bottom, with a binding energy of 16.5 eV. At an excitation energy of about
25 eV, corresponding to a resonance energy of about 8.5 eV above the vacuum level, the Stieltjes
orbital reveals a remarkable similarity, except for an arbitrary phase change, to the 6cr* orbital
(dashed arrow)

2.8.4 Feshbach-Fano Method

Finally we briefly mention the development of a new but related method
of calculating resonant molecular continuum cross sections, developed by
Winstead and Langhoff [2.131, 132J and reviewed by Gil et al. [2.133]. This
method uses the fact that, in many cases, continuum resonances can be under-
stood as arising from the interaction of a discrete state with nonresonant
 2.9 Shell-by-Shell Multiple Scattering Method 43

continuum channels, as originally pointed out by Feshbach [2.134] and Fano

and Cooper [2.10, 135]. The Feshbach-Fano method yields a convenient
parameterization of the photoionization cross section profile, the parameters
being (1) a nonresonant background cross section, expressed as the square of a
dipole matrix element between the initial core state and degenerate background
scattering states, (2) an overlap parameter, which characterizes the interaction of
the discrete state with the degenerate background scattering states, and (3) a
strongly energy dependent profile junction, dependent on energy-width and
energy-shift functions. This method is readily adapted to the quantitative
calculation of molecular (1* resonance cross sections by use of minimal basis
set (1* orbitals for the "discrete state(s)") and approximating the "background
states" by linear combinations of a large but finite number of L 2 wavefunctions
[2.131, 132, 136]. The success of such calculations underlines the relevance of
minimum basis set MOs to the understanding of (1* continuum resonances.

2.9 Shell-by-Shell Multiple Scattering Method

We briefly discuss below a third method that has been employed for the
calculation of near edge spectra of various solids, molecular complexes, and for
chemisorbed atoms on surfaces. Since the method has been reviewed in detail
elsewhere [2.137] and because of its limited applicability for the calculation of
NEXAFS spectra of chemisorbed molecules, a rather short account will be
given.
The particular multiple scattering method of interest was introduced by
Durham et al. in 1981 [2.138, 139] and is formulated in a real-space cluster
model. The code was later updated with emphasis on surfaces by Vvedensky et
al. [2.140]. The essence of this theoretical approach is the exact formulation,
within the one-electron approximation, of the scattering process of the photo-
electron excited on the central atom by the surrounding neighbors with in-
finitely many scattering events (multiple scattering) included. These events may
not only occur in a back scattering geometry as in the EXAFS approximation
but at any angle, and thus the photoelectron wave may bounce off numerous
atoms before returning to the origin. Scattering is considered in a shell-by-shell
manner, where a "shell" consists of atoms which have similar but not necessarily
identical distances from the origin and may even contain different atoms. Both,
intra- and inter-shell scattering is taken into account. By adding shells, the
results may be checked for convergence with respect to the cluster size and large
clusters consisting of more than 50 atoms can be handled on a personal
computer.
The crucial part of the calculation is the construction of a realistic potential
which reflects not only the core but also the valence charge distribution. Since
the shape of the potential near the vacuum level, which is largely determined by
the valence electrons, greatly influences the scattering processes of a low-energy
44 2. Theory of Inner Shell Excitation Spectra

photoelectron, it is highly desirable to use self-consistent procedures for the

calculation of the potential. Unfortunately, at present, self-consistent calcu-
lations have not yet been carried out and the potentials for the adsorbate system
under investigation are typically constructed by using a "model compound",
either an adsorbate system or bulk compound of well-known structure con-
sisting of the same atomic species. For this model compound the potentials are
chosen such that the calculation accurately reflects the measured NEXAFS
spectrum.
Because of the lack of self-consistency, at present the method is deemed
of insufficient reliability for the calculation of K -shell spectra of free and chemi-
sorbed molecules. We shall, however, use results obtained with this method in
Sect. 6.2.1, in the discussion of the qualitative differences between the K-shell
excitation spectra of chemisorbed atoms and molecules.

2.10 Approximations Leading to the EXAFS Equation

During the last twenty years it has been not the near-edge but the extended X-ray
absorption fine structure, or EXAFS, that has received the most attention
[2.141]. The reason is the close and simple link between the EXAFS oscillations
and the local crystallographic structure about the central excited atom. Al-
though EXAFS is not the topic of this book, it is obviously closely linked with
NEXAFS in that the NEXAFS and EXAFS regions smoothly join together in
the range 30-50eV above the absorption edge. We therefore need to examine
how the scattering processes of the photoelectron change with increasing energy.
Also, in the high-excitation-energy region certain approximations can be made
which lead to an analytical expression for the EXAFS structure. The interested
reader is referred to Appendix A where the EXAFS equation is derived. In this
case, it is possible to follow all physical processes using relatively simple
mathematical concepts and the treatise in Appendix A greatly helps in under-
standing the scattered-wave or partial wave concepts outlined earlier in the
discussion of the XO(-MS technique.
Here we shall simply state the final result for the EXAFS equation and
discuss the link between the NEXAFS and EXAFS. The polarization dependent
EXAFS signal as a function of the photoelectron wavevector k is given by
(A.47). If we group all atoms j with the same atomic number Z around the
central excited atom into shells i of equal distance R i , then the EXAFS signal is
given by

3 cos 2 ()..
+ 215 1 + Pd.
2 2
x(k) = - ~ kR2 IJ F i (k)e- 2t1 ,k e- 2R .;).,(k)sin(2kR i (2.63)
.,] i

Here cos 2 (}ij is a polarization dependent factor, Fi(k) the back scattering ampli-
tude characteristic of a neighbor with atomic number Z, exp( - 2o} P) the
 2.10 Approximations Leading to the EXAFS Equation 45

Debye-Waller-like term accounting for thermal vibrations, exp[ - 2R;/A;(k)]

the term accounting for the finite inelastic mean free path of the photoelec-
tron, and the sinusoidal term contains the distance and scattering phaseshift
dependence.
Let us review the approximations and limitations of (2.63), ignoring the
vibrational and inelastic terms. Three fundamental approximations were made:
(1) Neglect of multiple scattering, (2) point scattering at the neighbor atom, and
(3) use of the asymptotic limit for spherical waves. We shall discuss these
approximations in turn.
As the kinetic energy of the excited photoelectron becomes large relative to
its interaction energy with the surrounding atoms (a few eV) the scattering by
the neighbor atoms is largely determined by their atomic-like core potentials.
We can then use the approximation that the scattering off the neighbors is a
simple sum of their isolated contributions. In this single scattering model,
multiple scattering between several centers is ignored, and all the physics is
contained in the proper description of the final state wavefunction for a
diatomic, as done in Appendix A. Larger systems can be treated using a building
block principle. The essence of the single scattering approximation therefore lies
in the fact that at sufficiently high photoelectron kinetic energy the scattering
processes off neighboring atoms are dominated by the core rather than the
valence electron potential, and the scattering cross sections are peaked in the
forward and backward directions [2.142]. For this reason scattering phaseshifts
in the EXAFS region can be derived from atomic potentials. The charge overlap
between bonded atoms is either ignored and retroactively incorporated into the
analysis by use of an adjustable inner potential [2.141], or in a better ap-
proximation [2.143, 144] is accounted for by use of a non-self-consistent muffin-
tin potential which allows the determination of the inner potential, i.e., the zero
of energy of the EXAFS scale.
Because all multiple scattering path lengths are substantially longer than the
single scattering path length to a nearest neighbor atom and back, multiple
scattering can be ignored in the EXAFS analysis of the nearest neighbor shell
structure. Bunker and Stern [2.145] have argued that multiple scattering is only
important in the EXAFS region if an intervening atom lies approximately
collinear with the central atom and the backscattering atom. In this case
forward scattering needs to be considered in addition to backscattering, and the
photoelectron wave may be focused by the intervening atom onto the next
nearest neighbor atom. A significant change in amplitude and phase of the
EXAFS signal may result [2.146]. Bunker and Stern pointed out that in
exceptional cases where the bond length becomes short ( < 1.6 A) and there is no
center of inversion symmetry, other types of multiple scattering effects could also
be important in the EXAFS region.
Evidence for large-angle multiple scattering events in the EXAFS region has
been found for low-Z atoms chemisorbed onto metal surfaces by Arvanitis et al.
[2.147]. In this case single scattering contributions from higher neighbor shells
are weak and large angle multiple scattering events within the first neighbor
46 2. Theory of Inner Shell Excitation Spectra

shell may be detected since they involve comparable path lengths and ampli-
tudes to the former. Thus care has to be exercised in the interpretation of all
peaks in the Fourier transform at distances larger than the dominant nearest
neighbor shell peak.
For low-Z molecules, multiple scattering within the molecule leads to large
resonances in the near-edge region. These resonances, which are the main topic
of this volume, can be thought of as a trapping of the photoelectron wave by the
molecular potential which is largely determined by the valence electron charge
distribution. In particular, the so-called a* shape resonances discussed in Sect.
4.2.4 may be understood as back and forth scattering of the photoelectron wave
between two atoms. Although the higher energy EXAFS oscillations have a
close connection to the a * resonance as discussed in Sect. 8.2, they can
nevertheless be accounted for by a single scattering theory [2.148]. One simply
needs to include curved wave effects and the charge overlap caused by the
intramolecular bonding, as in the theory of Rehr et al. [2.143, 144].
It is easy to understand why approximations (2) and (3), which formally lead
to the plane-wave limit [2.149-151], become better with increasing kinetic
energy of the photoelectron. The higher the kinetic energy, the more the
photoelectron will penetrate the core of the neighbor atom and the scattering
will be increasingly dominated by core electrons and the nuclear potential.
Hence point scattering becomes a better approximation. With increasing energy
and therefore wavevector k all mathematical expressionsf(kr) encountered in
the derivation in Appendix A become closer to their asymptotic kr -+ 00 value.
Thus, spherical waves, in general given by Hankel functions, approach their
asymptotic limit in the EXAFS range.
The two approximations (2) and (3) lead to two problems. First, as shown by
Schaich [2.152] and Gurman et al. [2.153], in a proper spherical wave treatment,
the EXAFS phase shifts and scattering amplitudes become dependent on the
bond length R. This can be seen from (A.30), which only in the asymptotic limit
can be factored into an r'-independent amplitude and phase scattering function.
In practice, the R-dependence is taken into account either by use of standards or
properly calculated phases and amplitudes [2.154, 155].
Second, one has to neglect a depolarization term [2.156, 157]. By neglecting
the finite size of the scattering center B, this term which is proportional to
e,
sin 2 is absent. The origin of the depolarization term is easy to understand.
Assume that the E vector is aligned perpendicular to the internuclear axis R of a
diatomic. The photoelectron dipole pattern on the excited atom A which is
directed along the E vector will then also be perpendicular to the internuclear
axis. Thus a point-like neighbor atom B will lie in the nodal plane of the dipole
pattern and no electrons can scatter off it. For a finite-sized atom (potential) at
B, however, photoelectrons which are emitted at finite angles with respect to the
internuclear axis can scatter off atom B and therefore produce EXAFS. The
depolarization term was first observed, but not specificallY identified, in the XIX-
MS calculation for the N2 molecule by Dehmer and Dill [2.B], who found
pronounced EXAFS oscillations in the 7t ionization channel, which by definition
 2.10 Approximations Leading to the EXAFS Equation 47

is the EXAFS contribution for E perpendicular to the internuclear axis. Al-

though this depolarization contribution will be small compared to the max-
imum contribution (the "(J contribution") when Ell R, it may nevertheless cause
problems in analyzing amplitude ratios [2.157] in polarization-dependent
SEXAFS studies [2.158].
It is interesting to note that the depolarization term is of fundamental
importance at low photoelectron energies, i.e., in the NEXAFS range. In this
case the low-energy electron can be effectively scattered by the molecular
potential, which is largely determined by the valence charge distribution, even
when emitted perpendicular to the internuclear axis, and pronounced reson-
ances, e.g., n* resonances, are observed. Since in NEXAFS the observed
resonances are associated with molecular orbitals of well-defined symmetry, the
contributions from final states of different symmetry are naturally isolated, as
long as the resonances are separated in energy.
3. Symmetry and Molecular Orbitals

In this chapter the rudiments of molecular orbital theory are reviewed in the
light of their importance for the understanding of the resonances which domin-
ate molecular X-ray absorption spectra. Particular emphasis is given to the
symmetry classification of orbitals and how it relates to the polarization
dependent dipole selection rule that governs X-ray absorption.

3.1 Origin and Labelling of Molecular Orbitals

Although the calculation of the detailed resonance positions and intensities of

molecular K-shell excitation spectra requires initial and final state wave-
functions obtained by solving the Schrodinger equation, there are some simple
symmetry classifications of resonances which arise from the one-to-one corres-
pondence between resonances and molecular orbitals. This symmetry labeling
not only facilitates the understanding of the K -shell spectra of free molecules
discussed in the following section but it is particularly useful in describing the
dependence of resonance intensities on X-ray polarization, as discussed in
Chap. 9.
The close correlation between K -shell resonances and specific molecular
orbital final states revealed by Stieltjes-Tchebycheff MO calculations [3.1-5J
makes it important to develop an understanding of MO concepts [3.6-12]. The
simplest symmetry classification for low-Z molecules is in terms of a and n
orbitals [3.8]. For K-shell excitations the initial Is state is always of.a symmetry
while the final states allowed by the dipole selection rule have to contain an
atomic p-orbital component, and may be of a or n symmetry. In this volume we
are mainly concerned with low-Z molecules, composed of the atoms H, C, N, 0
and F. Bonds between these atoms involve the Is electrons of H and/or the 2s
and 2p electrons of the second row atoms. The composition of the molecular
wavefunction is a linear combination of atomic orbitals (LCAO) according to
(2.34).
If for the description of diatomics we choose a coordinate system with its z-
axis along the internuclear axis, a valence orbitals originate from atomic 2s
and/or 2pz orbitals and n orbitals arise from 2px and 2py atomic orbitals [3.6, 8,
10]. According to perturbation theory the degree of mixing of the 2s and 2pz
orbitals depends on the energy separation of the 2s and 2p orbitals which
 3.2 Some Molecular Orbitals and Irreducible Representations 49

increases from about 2 eV in Li to about 22 eV in F, and therefore the mixing is

weaker in F 2 than N 2. In the presence of hydrogen bonds the H Is orbital may
mix, if allowed by symmetry, with the 2s and 2p orbitals of the second row
atoms. Important examples are hydrocarbons, in which bonding and anti-
bonding MOs are formed from the Is orbital of hydrogen (Eb(ls) = 1 Rydberg
or 13.6 eV) and the 2s and 2p orbitals of carbon (Eb(2p) ~ 14 eV and
Eb(2s) ~ 23 eV [3.13J).
For diatomics, a orbitals are symmetric with respect to reflection through a
plane containing the symmetry axis of the molecule and 1t orbitals are antisym-
metric. In poly atomic molecules, the molecular orbitals involve atomic overlap,
which is generally more extended, but the resulting orbitals can still be labeled
either a, symmetric, or 1t, antisymmetric, with respect to a local symmetry plane
[3.8]. For a orbitals or bonds there is no nodal surface that contains the bond
axis, while a 1t bond has one nodal surface or plane containing the bond axis.
The a bond is always present while the existence of a 1t bond depends on the
atomic constituents of the molecule and the number of valence electrons
available for bonding. The a and 1t manifolds each contain both bonding and
anti bonding orbitals corresponding to in-phase and out-of-phase orbital ampli-
tudes on adjacent, bonding atoms.
For K-shell spectra, it is the nonfilled antibonding orbitals which are of
importance, since the observed resonances correspond to dipole allowed trans-
itions of a Is core electron to 1t and a antibonding orbitals. For homonuclear
molecules, orbitals are given a second label g (gerade) or u (ungerade) indicating
the inversion symmetry through the center of the molecule. The wavefunction of
g orbitals is symmetric, that of u orbitals antisymmetric with respect to inver-
sion. For complex molecules it is appropriate to use group theory and designate
the MOs by their irreducible representations, rather than by the simple a and 1t
label [3.11, 14]. Below we shall meet examples of how MOs for molecules of
increasing complexity are constructed by the LCAO scheme under group-
theoretical symmetry constraints.

3.2 Some Molecular Orbitals and Irreducible Representations

3.2.1 Diatomics and Linear Triatomics

The simplest MOs are those of diatomic molecules. Figure 3.1 illustrates how
atomic orbitals combine to produce molecular orbitals using the diatomic
homonuclear N2 molecule as an example [3.10]. We again choose a coordinate
system with the z-axis pointing along the internuclear axis. If we ignore the
mixing of the 2s and 2pz atomic orbitals, the interaction of the 2pz atomic
orbitals on each N atom leads to a bonding (3ag ) and antibonding (3a u ) MO of a
symmetry. The 2px and 2py orbitals on each N atom form two orthogonal but
energetically equivalent 1t bonds characterized by the bonding (l1tu) and anti-
50 3. Symmetry and Molecular Orbitals

Atom 1 Molecule Atom 2 Fig. 3.1. Diagram of the formation of molecular

3<7 u orbitals from atomic orbitals in a homonuclear
diatomic molecule consisting of second row atoms
(i.e., Li2 through F2). The asymmetric spacing of
the four highest (J orbitals about the 2p and 2s
centers of gravity is due to a mixing of the 2s and
2pz atomic orbitals as discussed in the text. In
17T 9
~
particular, the shown order of the 311, and lnu

~~::5
x 2p .~'i'l'l 2p x orbitals reverses for molecules heavier than N2
y 3<79 Y [3.15]. Also the occupation changes across the
z ~*-'~ ..... Z
" ..:::::.~

17T u
/-:/
'l
series. For Li2 the 211u is the lowest unoccupied
MO, for N2 the In. orbital. Note in particular
25 ......... 25 that 02 has two electrons with parallel spins
2<7u
(Hund's rule) in the two In. orbitals and is there-
fore paramagnetic. (From [3.10])

2<79

1<7u
15 . . . . . . .. . ....... ::. 15
... .....
1<79

bonding (Ing) orbitals. Combination of the two 2s orbitals leads to a bonding

(20"g) and anti bonding (20"u) combination. Because the Is atomic orbitals are so
localized the lowest MO in N2 is a nearly degenerate 100g and 100u combination.
The bonding 0" orbitals are gerade with respect to inversion through the center of
the molecule and the antibonding 0" orbitals ungerade. For n orbitals the reverse
is true. Consideration of the mixing of the 2s and 2pz atomic orbitals leads to an
asymmetric spacing of the energy levels of 0" symmetry around their atomic
centers of gravity (Fig. 3.1), resulting from a repulsion between the mainly 2pz
and mainly 2s derived MOs, respectively. N2 is said to be triple bonded by one 0"
(the 20"g is the most strongly bonding MO) and two n bonds.
The schematic energy level diagram shown in Fig. 3.1 describes the bonding
of molecules ranging from Li2 to F 2' With increasing Z of the second row atoms
the 2s-2p energy separation increases and therefore the asymmetric spacing of
the 0" orbitals decreases. This leads to a different ordering of the 30"g and 1nu
orbitals depending on whether the molecule is lighter or heavier than N 2. Also,
of course, the electron occupation of the MOs changes through the second row
series. For example, the Ing orbital is empty for N 2, halffull for O 2 and filled for
F 2' In the following we shall denote empty orbitals with an asterisk label, e.g.,
n;
1 for the N 2 molecule.
Because carbon, nitrogen, and oxygen have similar orbital energies for the
atomic 2s and 2p orbitals, MO theory predicts very similar molecular orbitals
and orbital energies for N2 and isoelectronic CO. This is supported by experi-
ment [3.16, 17J and ab initio theory [3.18J and a comparison of the calculated
orbital energies [3.18J is shown in Fig. 2.5, together with the respective orbital
 3.2 Some Molecular Orbitals and Irreducible Representations 51

contours. The MOs of CO cannot be labeled by g and u since there is no

inversion symmetry in the molecule, i.e., the atomic contributions from carbon
and oxygen to the MOs are different. For example, because oxygen is more
electronegative than carbon the bonding MOs tend to have more oxygen, the
anti bonding MOs more carbon character (Fig. 2.5). The Is orbital of oxygen is
the 10", that of carbon the 20" MO.
So far we have constructed MOs without the explicit use of symmetry
although we have, at one point, made use of it. In considering the mixing of the
2s and 2p atomic orbitals in the diatomics we have only allowed the 2s and 2pz
orbitals to mix, not the 2s and 2px and 2py orbitals. This is a result of symmetry
which is most elegantly handled by group theory. The simplest way to take
symmetry into account in constructing MOs is to use the irreducible representa-
tions and character tables [3.11, 14] for the point group of interest. In the
literature, capital letters are typically used for the irreducible representations of
the point groups and small letters for the labeling of orbitals, and we shall do the
same here. It is convention to take the principal symmetry axis of the molecule
as the z-axis.
Let us consider the simplest case, that of a heteronucIear diatomic, e.g., CO.
The CO molecule has C oov symmetry and from the appropriate character table
listed in Table 3.1 we can find that the L representation transforms like z and
therefore characterizes the symmetry properties of the 2pz orbital. The spher-
ically symmetric 2s atomic orbital is represented by the function x 2 + y2 + Z2,
which also transforms like L. In contrast, according to Table 3.1 the 2px and 2py
atomic orbitals associated with the functions x and y transform like the n
representation. The important point in constructing MOs is that only orbitals of
the same symmetry, i.e., belonging to the same irreducible representation, can mix
[3.8, 10, 11, 14]. For CO the 2pz and 2s atomic orbitals have L symmetry and
can therefore mix while the 2px and 2py orbitals have n symmetry and cannot
mix with either the 2pz or 2s orbitals.
The N2 molecule has Dooh symmetry and with respect to the inversion center
of the molecule we can symmetrically combine two 2pz atomic orbitals on each
N to obtain a symmetry adapted bonding orbital which transforms like Z2 (no
sign change upon inversion) and which according to Table 3.1 corresponds to
Lg • Similarly, two 2s atomic orbitals on each N can be combined such that the
symmetry adapted combination transforms like Z2 and therefore corresponds to
Lg , as well. These two bonding combinations are allowed to mix. The corres-
ponding 2s and 2pz anti bonding combinations each transform like z, i.e., switch
sign upon inversion, and according to Table 3.1 they each belong to the
representation Lu and therefore are allowed to mix. In contrast, the 2px and 2py
orbitals and all symmetry adapted combinations belong to the representations
nu and ng and cannot mix with the 2s orbitals.
Similar arguments to those for N2 can also be applied to the CO 2 molecule
which also has Dooh symmetry. Because the atomic orbitals of three atoms are
involved, a larger number of MOs results, in particular, we obtain two empty
antibonding MOs of 0" symmetry, the 50"; and 40"~ orbitals, and one empty n*,
52 3. Symmetry and Molecular Orbitals

Table 3.1. Selected abbreviated character tables

C oov E 2Ct, 00 (Jv

L Z, X 2 + y2, Z2
II 2 2cos¢ 0 (x, y), (xz, yz)

Dooh E 2Ct, 00 (Jv 2St, 00 C2

L. 1 x2 + y2, Z2

II. 2 2cos¢ 0 2 -2cos¢ 0 (xy, yz)

Lu 1 1 -1 -1 -1 z
IIu 2 2cos¢ 0 -2 2cos¢ 0 (x, y)

C2v E C2 (Jv (J~

At 1 Z, x 2 , y2, Z2
A2 -1 -1 R" xy
Bt -1 1 -1 x, R y, xz
B2 -1 -1 y, R" yz

C 3v E 2C 3 3(Jv

At Z, x 2 + y2, Z2
A2 1 -1 Rz
E 2 -1 0 (x, y), (R" Ry), (x 2 - y2, xy), (xz, yz)

Td E 8C 3 3C 2 6S 4 6(Jd

At 1 1 1 1 1 (x2 + y2 + Z2)
A2 1 1 1 -1 -1
E 2 -\ 2 0 0 (x2 _ y2), (2Z2 _ x2 _ y2)
Tt 3 0 -1 1 -1
T2 3 0 -\ -1 1 (x, y, z), (xy, xz, yZ)

the 2n:, orbital. The highest occupied molecular orbital (HOMO) and lowest
unoccupied (LUMO) molecular orbitals of CO 2 are shown in Fig. 3.2. The Is
orbital of oxygen is the nearly degenerate 1au and lag pair, and the Is orbital of
carbon is the 2ag orbital. Note that the Is orbital of carbon is nondegenerate and
gerade since the carbon atom is located in the inversion center of the molecule.
We shall see that the inversion symmetry has important consequences and leads
to different oxygen and carbon K-shell spectra.
 3.2 Some Molecular Orbitals and Irreducible Representations 53

Frontier Orbitals of CO 2 Fig. 3.2. Plot of the highest occupied

(HOMO) and lowest unoccupied (LUMO)
molecular orbitals for CO 2 • labeled by their
symmetry. taken from [3.8]

5u'g

3.2.2 lJydrogen Fluoride, Water, Ammonia, and Methane

Let us now consider bonds between hydrogen and second row atoms. The
simplest examples are the molecules hydrogen fluoride (HF), water (H 2 0),
ammonia (NH 3 ), and methane (CH 4 ) which are isoelectronic with neon. For all
four molecules MOs are constructed from the 2s and 2p orbitals of the second
row atoms and the hydrogen Is orbitals, under the symmetry constraints
dictated by the point groups of the molecules. From the Is orbitals of the off-
center hydrogens we can form symmetry adapted combinations corresponding
to a particular irreducible representation.
Figure 3.3 shows the MOs of HF, H 2 0, NH 3 , and CH 4 , labeled by their
symmetry, as a function of the orbital energies calculated by Snyder and Basch
[3.18]. We also show schematically the chosen molecular coordinate system and
the atomic orbital composition of the MOs. The figure is intended as a guide for
the following discussion. HF has the same symmetry as CO (Ca)V ) and according
to Table 3.1 the fluorine 2s and 2pz orbitals are allowed to mix and interact with
the hydrogen Is orbital. This gives rise to three a orbitals, a pair of bonding
orbitals, 2a and 3a, and an empty 4a* anti bonding orbital. The 2a orbital has
larger 2s and the 3a orbital more 2pz character. The 4a* orbital contains both 2s
54 3. Symmetry and Molecular Orbitals

HF Coov H2 O C 2v NH3 C 3v CH 4 Td

~' rlL ~ @
-_ ... _-
x

10 ~ 2p z
25
~2bi ~
4a* 1
__1_* =~~ ~2Lr~
4a'
f-
-
212 ~.®

-~~~---
~ @$
5
0 &- - - - - - - - - - ----------

$! ~ 1t2>-~
_ 2py
- 10 ~ ~
>
~ 2px 2py ~~
- 15
...>- 17r -r~
~1* ~)-r ~
. ~
0)

II>
c: -20 ~~ ~
"""~~ 2px 2py """

2.,
'
w
2
~
:0 -25
2PZ
25
Px
25 --
~ '®
0 2PZ @j
-30 f-
"""~ ~"""
~

25
2P Z
-35

- 40
25
2p z
~
~
T ~ f-

~
~

}o:.
-45

-W
15 '
t~
1a _.
-----@ ~ ~ ~ 1
-- ...
Fig. 3.3. Correlation between orbital energies for the isoelectronic molecules hydrogen fluoride
(HF), water (H 2 0), ammonia (NH3)' and methane (CH 4 ) . The orbital energies are taken from the
self-consistent field calculation of Snyder and Basch [3.18] and the MOs are labeled by their
irreducible representation in the indicated point groups of the molecules. Note that for H 2 0 the
irreducible presentations b. and b2 depend on the choice of the x- and y-axes. For each MO the
atomic orbital composition is shown. For HF, H 2 0 , and NH3 we have qualitatively followed the
MO composition calculated by Snyder and Basch, while the MOs for CH 4 are in the coordinate
system shown in the figure, which is also used for the canonical MOs tabulated in Table 3.2

and 2pz character [3.18]. The fluorine 2px and 2py orbitals form a pair of 11:
orbitals which are non bonding since there is no 11: orbital on the hydrogen. The
localized fluorine 1s orbital is given the 10' molecular label.
H 2 0 has C 2v symmetry and we choose the z-axis along the C 2 axis and the y-
axis perpendicular to the molecular plane. From table 3.1 the oxygen 1s and 2s
orbitals and the 2pz orbital can be seen to transform like the A 1, the 2px orbital
 3.3 Molecular Orbitals, Equivalent Orbitals and Hybrid Orbitals 55

like the BI and the 2py orbital like the B2 irreducible representations. The
localized oxygen Is orbital is given the la l label. Mixing of the 2s and 2pz
orbitals with the appropriate symmetry adapted pairs of hydrogen orbitals
results in three bonding, 2a l , Ib l and 3a l , and two antibonding, 4a! and 2b!,
MOs. The oxygen 2py orbital forms a non bonding n, the Ib 2 orbital. Another
way of constructing MOs for a molecule like H 2 0, using simple (1 and n
symmetry considerations with respect to symmetry planes in the molecule, is
discussed by Jorgensen and Salem [3.8] for methylene (CH 2 ).
NH3 has C3v symmetry and we choose the z-axis along the C 3 axis. Because
of the threefold symmetry in the x-y plane, the 2px and 2py orbitals are
degenerate and according to the character table they correspond to the irreduc-
ible representation E. Similar to H 2 0 the 2pz and all s orbitals of nitrogen
correspond to the Al representation, and we denote the nitrogen Is orbital as
la l . Mixing of the 2s and 2pz with appropriate symmetry adapted combinations
of the three hydrogen Is orbitals leads to the bonding 2a l , Ie and 3a l , and
antibonding 4a! and 2e* MOs shown in Fig. 3.3. Inspection of the wave-
functions [3.18] reveals that the 3a l MO has little hydrogen character and is,
essentially, the nitrogen lone pair orbital. Again, the MOs can also be derived
from considerations of (1 and n symmetries with respect to certain molecular
planes as discussed by Jorgensen and Salem [3.8] for the methyl (CH 3) radical.
Thus the 4a! MO in NH3 can be called a N-H (1* and the 2e* a N-H n* orbital.
For CH 4 the symmetry is that of the tetrahedral group Td and the cubic
symmetry requires all 2p orbitals to be degenerate, corresponding to the
representation T2 • Again all carbon s orbitals belong to the symmetric repres-
entation Al with the carbon Is or la l orbital lowest in energy. Note that the
mixing of the 2s and 2pz orbitals is symmetry forbidden. The MOs resulting
from combinations of the carbon 2px, 2py, and 2pz orbitals and symmetry
adapted combinations of the four hydrogen Is orbitals, labeled hI' h2' h3' and
h4' depend on the choice of the coordinate system. The schematic pictures for
the bonding 2al and It2 and anti bonding 3a! and 2t! MOs in Fig. 3.3 refer to
the coordinate system shown [3.10] and differ from those given by Jorgensen
and Salem [3.8]. Note in particular that the 3a! LUMO does not contain any p-
orbital component, which as we shall see is important in conjunction with the
dipole selection rule.

3.3 Molecular Orbitals, Equivalent Orbitals and Hybrid Orbitals

3.3.1 Molecular Orbital Versus Valence Bond Theory

For the description of molecules we have, so far, used MO theory, which is based
on the construction of MOs from the orbitals of the individual atoms, under
consideration of symmetry. The success of this LCAO scheme is based on its
mathematical convenience for the calculation of ionization potentials, which via
Koopmans' theorem can be associated with the orbital energies of the canonical
56 3. Symmetry and Molecular Orbitals

MOs. On the other hand, organic chemists have long used another model, based
on the sharing of pairs of electrons, to explain the similarity and additivity of
important "bond characteristics", such as bond energies and geometries, in
certain classes of molecules [3.19]. This model, originated by Lewis, was
developed by Pauling and others in the 1930s into the quantum mechanical
Valence Bond Theory [3.6]. Here we shall make use of some of the concepts of
the valence-based theories because they are very useful, if not fundamental, for
the classification of molecules and/or discussion of the properties of large
molecules. In particular, we shall consider an important subgroup of molecules,
the hydrocarbons, and develop simple concepts to explain many of their
properties.
Our earlier MO description of methane (CH 4 ) showed well the correspond-
ence between atomic orbitals, MOs and one-electron binding energies. In
particular, because of symmetry, the bonding between carbon and the four
hydrogens resulted in four bonding MOs, three degenerate 1t2 and a single 2a 1
orbital, shown in Fig. 3.3. However, in the MO model there is no indication that,
in fact, all four carbon-hydrogen bonds are chemically equivalent, e.g., have
identical bond lengths (~ 1.1 A), bond energies ( ~ 100 kcaljmol), and stretching
force constants (~500 Nm - 1). This equivalence of the carbon-hydrogen bonds
naturally emerges from the valence-based theorices, either the octet electron rule
in Lewis' theory or the covalent electron-pair bond in Valence Bond Theory. In
both valence-based theories each C-H bond consists of a shared electron pair.
One important concept born out of the electron-pair bond model is that of
hybridization. This mathematical concept consists of constructing equivalent
hybrid orbitals from combinations of atomic orbitals. For carbon the lowest
electronic configuration is (ls)2(2s)2(2Px)(2py) with two unpaired electrons. The
basis of the hybrid orbitals is the excited configuration (ls)2(2s)(2Px)(2py) (2pz)
with four unpaired electrons. Although this configuration lies about 100kcalj
mol or 4.3 eV above the ground state, this energy can be retrieved by the ability
of the excited carbon atom to form more than two bonds. For example, for
methane four hybrids are formed from a linear combination of the four atomic
2px, 2py, 2pz, and 2s functions, and because each hybrid consists of one sand
three p orbitals the hybrid is called an Sp3 hybrid. Similarly, we can form Sp2
hybrids from two p and one s orbital, and an Spl or simply sp hybrid from one p
and one s orbital (the "unused" p orbitals form 1t bonds, see later). Wave-
functions for the three types of hybridization are given in Table 3.2.
Hybridization forms a useful intellectual bridge between a mathematical
technique and simple structural ideas. For example, Sp3 hybrids form tetra-
hedral bonds (bond angle ~ 110°), Sp2 hybrids trigonal bonds (bond angle
~ 120°), and sp hybrids digonal bonds (bond angle 180°). Using the hybrid
concept, large molecules may be assembled by means of "localized" hybrid
bonds and it is possible to calculate many molecular properties such as heats of
formation, geometries, and dipole moments by assuming that the contributions
from the individual "localized" bonds are additive [3.22].
On first inspection, the molecular orbital and hybrid orbital concepts appear
to have little or no connection. However, in a series of papers beginning in 1949
 3.3 Molecular Orbitals, Equivalent Orbitals and Hybrid Orbitals 57

Table 3.2. CH 4 molecular and equivalent orbitals and carbon hybrid orbitals

Canonical CH 4 MOs for the tetrahedral group T., [3.20, 21]8

tP, (Ad = 0.5812s) + 0.19(lh,) + Ih 2) + Ih3) + Ih4 »)
tP 2 (T2 ) = 0.5512px) + 0.32(lh,) -lh 2) + Ih3) - Ih4 »)
tP 3 (T2 ) = 0.5512py) + 0.32(lh, ) + Ih 2) -l h3) - Ih4 »)

tP4 (T2 ) = 0.5512pz) + 0.32(lh,) - Ih 2) - Ih3) + Ih 4 »)

Equivalent CH 4 orbitals [3.20, 21]8

tP', = !(tPdA,) + tP 2 (12) + tP 3 (12) + tP4 (12))

= 0.2912s) + 0.28(12px) + 12p,) + 12pz») + 0.58Ih l ) - 0.07(1h2) + Ih3) + Ih4 »)

tP~ = !(tPdAd - tP 2 (12) + tP 3 (12) - tP4 (12))

= 0.2912s) + 0.28( - 12px) + 12py) - 12pz») + 0.58Ih2) - 0.07(1h,) + Ih3) + Ih4 »)

tP~ = !(tPdA,) + tP2 (12) - tP 3 (12) - tP4 (12))

= 0.2912s) + 0.28(12px) -12py) -12pz») + 0.58Ih3) - 0.07(lh l ) + Ih 2) + Ih4 »)

tP~ = !(tP, (Ad - tP 2 (12) - tP 3 (12) + tP4 (12))
= 0.2912s) + 0.28( - 12px) - 12py) + 12pz») + 0.58Ih 4 ) - 0.07(1 hi ) + Ih 2) + Ih3»)

Sp3 hybrids [3.21]

rPl = H12s) + 12px) + 12py) + 12pz»)

rP2 = !(12s) -12px) + 12py) -12pz»)
rP3 = !(12s) + 12px) - 12py) - 12pz»)
rP4 = !(l2s) -12px) - 12py) + 12pz»)
Sp2 hybrids [3.21]
I
1/1,= J3[12s)+J21 2Pz)]

1/12= ~[12s)+AI2PY)- ~12Pz)]

1/13 =_1 [12S) -
J3
fi 12P ) _J2
"';2 Y
_I 12PZ)]

sp hybrids [3.21]
I
XI = J2 [l2s) + 12pz)]

I
X2= J2[12S)-12Pz )]

8 The 2s orbital is orthogonalized to the Is orbital and the small contribution of the Is orbital in
tP I (A I) has been neglected
58 3. Symmetry and Molecular Orbitals

Lennard-Jones et al. [3.23, 24] showed that equivalent or transferrable bond

properties could be understood using MO theory, the key being the determin-
antal description of many electron wavefunctions. We have already encountered
the basic concept underlying this idea in conjunction with the canonical
Hartree-Fock equations (2.28). These were obtained from the general HF
equations (2.24) by a unitary transformation which diagonalized the Fock
operator. The canonical MOs cP i for methane obtained with a minimal basis
set of Slater orbitals by the Hartree-F ock method are listed in Table 3.2 [3.25],
labeled by their irreducible representations. These orbitals may be transformed
into the set of equivalent orbitals or valence orbitals cP;, also listed in Table 3.2, by
a suitable unitary transformation, i.e., the linear combinations indicated in the
table.
From the atomic composition of the equivalent orbitals it is apparent that
they consist of two contributions, "localized' partly on the carbon atom and
partly on one of the four hydrogen atoms. At this point it is important to realize
that the "localization" of electrons implied by the composition of equivalent-
orbital wavefunctions is no real physical but rather a mathematical phenom-
enon. The connection between equivalent orbitals and hybrid orbitals is estab-
lished by the fact that, except for an overall normalization factor of 2, the carbon
contribution to the equivalent orbitals is nearly identical to the carbon Sp3
hybrid orbitals listed in Table 3.2.

3.3.2 Ionization Potentials in Methane

Since valence or equivalent orbitals are associated with the non-diagonal HF
equation (2.24) they do not directly yield ionization potentials but, according to
(2.24) and (2.28), IPs may be obtained by solving the eigenvalue equations [3.26]
N
L
k=l
AikcPk = EcP i , (3.1)

i.e., the secular determinant

(3.2)

Here the matrix elements Aik correspond to those of the nondiagonal Fock
operator. As discussed by Hall [3.26] for methane the matrix Aik for the four
equivalent valence orbitals is given by

a b b b)
b a b b
(
Aij= b b a b . (3.3)

b b b a

The diagonal matrix elements a are the self-energies of the equivalent

orbitals, i.e., the C-H bonds, and the off-diagonal terms represent the interaction
 3.3 Molecular Orbitals, Equivalent Orbitals and Hybrid Orbitals 59

energies. Because of the tetrahedral symmetry each equivalent orbital can

interact with the other three, and all interactions are the same. The above matrix
can be transformed into the following diagonal form by means of a unitary
transformation:

a + 3b 0 o
( o a-b o oo ).
o 0 a-b o (3.4)
o 0 o a-b
The roots of the diagonal matrix correspond to IPs, similar to the case of the
diagonal HF equations. The single root (a + 3b) represents the binding energy of
the 2a 1 orbital, and the triple roots (a - b) are the binding energies of the triply
degenerate It2 MO.
Using experimental binding energies of 14.4 eV for the It2 and 22.9 eV for the
2a 1 MOs of methane [3.13] we can determine the parameters a = -16.5eV and
b = -2.1 eV. In the equivalent orbital scheme one can therefore view methane
as having four equivalent C-H bonds with a self-energy of 16.5 eV. The fourfold
degeneracy of this valence state is lifted by interaction of the equivalent orbitals
such that the energies of the resulting two states are the lowest two IPs of the
molecule, i.e., those of the MOs derived from the atomic 2p and 2s orbitals of
carbon.

3.3.3 Bonding in Ethane, Ethylene, and Acetylene

The usefulness of hybrid orbitals is elegantly demonstrated by considering the
bonding in the di-carbon molecules ethane (C 2 H 6 ), ethylene (C 2 H 4 ), and
acetylene (C 2 H 2 ) (Fig. 3.4). The main difference in bonding between the three
molecules occurs in their C-C bonds, and one may compose the associated
molecular orbitals in a particularly simple fashion from carbon hybrid orbitals.
Although the number of hydrogen atoms and the C-H bond geometries change,
the C-H bond energies [3.27] and bond lengths [3.28] are very similar in all
three molecules. The C-H bonds in all hydrocarbons lead to characteristic
resonance structures in K-shell excitation spectra which are very similar and
distinct from those due to C-C bonds. Therefore, it is a good approximation for
our purpose to separate the C-H from the C-C bonds and focus our discussion
on the bonding between the two carbon atoms.
For ethane, each of the four carbon Sp3 hybrid orbitals forms a separate
bond, either to another carbon atom or one of the three hydrogen atoms, and
the carbon bonds are said to be "saturated". If we choose the z-axis of our
coordinate system along the C-C bond, we can form C-C bonds by hybridiza-
tion of the atomic 2s and 2pz orbitals on each carbon atom as shown in Fig. 3.4,
while the C-H bonds are formed by mixing of symmetry adapted combinations
of hydrogen Is orbitals with carbon 2s-2pz or 2Px-2py combinations. All bonds
are composed of carbon Sp3 hybrids which can be constructed in our coordinate
60 3. Symmetry and Molecular Orbitals

Single Bond Ethane

u~c (3a 2u )

Double Bond Ethylene

"cc (1b 3g )

"cc
-+t-

Triple Bond Acetylene

"cc (1" g)
"~c

Fig. 3.4. Formation of C- C bonding and antibonding (denoted by an asterisk) orbitals from atomic
hybrid orbitals for ethane (C 2 H 6 ), ethylene (C 2 H4)' and acetylene (C 2 H 2 ). For a C atom four hybrid
orbitals can be formed by superposition of atomic 2s and 2px, 2py, and 2p, wavefunctions. For
ethane three hybrids on each C atom are used for bonding to hydrogen atoms while the fourth forms
the C- C bond. If we choose the C-C internuclear axis as the z-axis, a C-C (J bond is formed from
hybids of 2s and 2p, atomic character, as indicated. Two (J orbitals result, a bonding orbital (Jcc with
in-phase and an antibonding orbital (J~c with out-of-phase orbital amplitudes on adjacent atoms,
respectively. The C-C bond in ethane is said to be a single, saturated bond of Sp3 character. For
ethylene, the C-C (J bond is stabilized and shortened by an additionaln bond, originating from 2px
atomic orbitals. As for the (J bond, bonding nee and anti bonding n~e orbitals exist. Ethylene is
therefore double bonded and the C=C bond has Sp 2 character. Acetylene has a second n bond
originating from the atomic 2py orbital. The bonding and anti bonding n states are energetically
degenerate, respectively. The resulting C",C bond is called a sp hybridized triple bond. The shapes
of all orbitals on the left side are just schematic and indicate the position of nodes and the signs of the
wavefunctions. Calculated wavefunctions of the anti bonding orbitals including their group theoret-
icallabels, taken from [3.8], are shown on the right side. For the (J~e orbitals we have selected the
ones with the most C-C character from the various orbitals [3.8]
 3.4 Interactions Between Localized Orbitals: Conjugation 61

system by a linear combination of the Sp3 hybrids listed in Table 3.2. The angle
between any two bonds is ~ 110°. The symmetry adapted combinations of the
atomic 2s and 2p. orbitals lead to two C-C bond orbitals consisting of a filled
bonding ace orbital and an empty antibonding a~c orbital as illustrated in
Fig. 3.4. A three-dimensional picture of the anti bonding a~c 3a 2u orbital for
ethane, taken from [3.8], is also shown.
In the planar molecule ethylene we can explain all in-plane bonding by use of
Sp2 hybrids. Again we choose the z-axis of our coordinate system along the C-C
bond and use a coordinate-adjusted combination of the three Sp2 hybrids, listed
in Table 3.2, to account for the three in-plane C-H and C-C bonds. Note that
the bonds are at angles of ~ 120° with respect to one another. The carbon 2px
atomic orbitals, which are perpendicular to the molecular plane have not yet
been used and we can make a C-C n bond out of them. The so-formed n system,
composed of a bonding and an anti bonding MO, shown in Fig. 3.4, is ortho-
gonal to the in-plane a C-C bonds. This orthogonality is the foundation for the
separate treatment of the a and n bonds in hydrocarbons, underlying Ruckel
theory [3.10]. Illustrations of the antibonding n~c 1b 3g and a~c 4au orbitals for
ethylene [3.8] shown in Fig. 3.4 reveal their symmetry.
Finally, triple bonded acetylene has two orthogonal, energetically degener-
ate n orbitals which are orthogonal to the C-H and C-C a system composed of

n:
carbon 2s and 2p. atomic orbitals, i.e., the sp hybrids listed in Table 3.2. The two
anti bonding 1 orbitals, together with the 4a~ antibonding orbitals [3.8] are
plotted in the bottom right corner of Fig. 3.4.

3.4 Interactions Between Localized Orbitals: Conjugation

To facilitate the understanding of the bonding in the simple pseudo-diatomic

hydrocarbons (the prefix "pseudo" refers to the neglect of the C-H bonds), we
have in the last section introduced the "equivalent orbital" and "hybrid orbital"
concepts. For even larger molecules, e.g., hydrocarbon chains or rings, we can go
one step further. We use equivalent or hybrid orbitals to describe "localized"
orbitals or bonds and treat interactions between such orbitals or bonds as a
perturbation. Thus we account for delocalization or conjugation effects in a
perturbative fashion [3.22]. This view is supported by the fact that, in general,
we can write any MO t/li as a linear combination of equivalent or hybrid orbitals
according to

(3.5)

since "localized" bond orbitals C/Jk are themselves constructed from atomic
orbitals. Thus in our approach we form MOs t/li by a perturbative coupling of
individual C/Jk's. We now introduce a Hamiltonian .Yf such that the diagonal
62 3. Symmetry and Molecular Orbitals

matrix elements

(3.6)

correspond to characteristic "energies", the self-energies, of the "localized"

orbitals tPk • This is the same concept used for methane in the previous section.
The off-diagonal matrix elements

(3.7)

represent the orbital-orbital interaction energies. Furthermore, we assume that

the orbitals tP j are orthonormal,

(3.8)

The zero overlap assumption expressed by (3.8) simplifies the corresponding

secular equations for the energies, similar to the conventional Hiickel treatment
[3.10].

3.4.1 First and Second Order Perturbation Treatment

Let us consider the simple case of two interacting orbitals, represented by the
interaction matrix

( Al B12), (3.9)
B21 A2

whereB12 = B21 = B. If the two orbitals are energetically degenerate, Al = A2

= A, we have the well known case of first order perturbation and we obtain two
split levels with energies

El =A - B, (3.10)

and

E2 = A + B, (3.11)

as shown in Fig. 3.5. Here we have defined the interaction energy B as a negative
quantity. For the corresponding wavefunctions we obtain

(3.12)

and

(3.13)
 3.4 Interactions Between Localized Orbitals: Conjugation 63

First Order Interaction Fig. 3.5. Schematic energy level

diagram for the interaction of two
<1>,-<1>2 orbitals. The strong first order in-
A- S ... .. ....... .... . . .
/ 1'J:K)9C)\ teraction between energetically
degenerate orbitals q" and q,2 with
A ..... .-!!...../ \~ ...... A energy A leads to a higher energy
I'J:K) \ / ~ antibonding orbital q" - q,2 and
\ /
A + 8 .• . . . .• . . .... . . . . .\ <1> , +<1>2 I
a lower energy bonding orbital
q" + q,2. which are separated by
~ an energy 12BI, where the interac-
Second Order Interaction tion energy B is negative. The
weaker second order interaction
8 between energetically nondegener-
'1" +-- '~2
A,-A2 S2 ate orbitals with energies A, and A2
<1>, / i ............. A, + A -A
A, .... .. _ _ _ / ®oC>@) , 2 results in a repulsion between the
®oC> \ \ energy levels. The energies and
\ \
\ \ wavefunctions of the perturbed sys-
\ \ tem are given in the lower diagram
\ \
\ \ ~'2
82 \ ~ / - - _ •.•. • . A2
A 2 - - - · ·· · ···· · ·
A, - A2 B
/ 0
<I'2 - --<r"
A, - A2

If the two interacting states are nondegenerate, say Al > A 2 , and their separa-
tion is large compared to the perturbation, IBI ~ Al - A 2 , we have the other
familiar case of second order perturbation. In this case the interaction pushes the
two states apart according to
B2
EI = Al + Al - A2 (3.14)

and

(3.15)

as shown in Fig. 3.5 with wavefunctions

(3.16)

and

(3.17)

Here c is a normalization constant. It can be seen from (3.16) and (3.17) that the
wavefunctions of the composite system are predominantly those of the two
64 3. Symmetry and Molecular Orbitals

respective "localized" components since IBI ~ A1 - A 2 . Note, however, that the

change in the energies given by (3.14) and (3.15) may still be sizeable since the
perturbative correction is quadratic in B.

3.4.2 Interactions in Chain-Like Hydrocarbons

In order to apply the above concept, let us consider the example of interacting
"localized" p-orbitals in the polyenes and polyines, i.e., hydrocarbon chains with
alternating C-C double and single and triple and single bonds, respectively. For
the polyene series the starting point is ethylene (H 2 C=CH 2 ), characterized by
"localized" p-orbitals on each C atom, which interact to form a 1t bond. The next
larger molecule is butadiene (H 2 C=CH-CH=CH 2 ) with four interacting p-
orbitals. Note that in allene (H 2 C = C = CH 2 ), with two adjacent double bonds,
the adjacent 1t systems are orthogonal, resulting from interaction of two pairs of
orthogonal p-orbitals, and therefore the 1t system is not delocalized. The third
member of the polyene series with six interacting p-orbitals is hexatriene
(H 2 C = CH-CH = CH-CH = CH 2 ). If we replace the double bonds in the above
molecules by triple bonds we obtain the polyines acetylene (HC == CH), di-
acetylene (HC == C-C == CH), triacetylene HC == C-C == C-C == CH), etc., which
can be treated similarly to the polyenes with two orthogonal p-orbitals on each
C atom.
The general treatment of the polyenes and polyines requires consideration of
the interaction between the various atomic p-orbitals. This is done in Huckel
theory by starting from identical p-orbitals on the individual C atoms, charac-
terized by a self-energy IX, and considering nearest neighbor interactions {3
between the atomic p-orbitals [3.10]. For ethylene this yields the orbital
interaction matrix

(3.18)

and the eigenvalues are 81,2 = IX ± {3 such that the resulting filled bonding and
empty anti bonding 1t orbitals are separated by 2{3. Considering only nearest
neighbor interactions between p-orbitals, the matrix for butadiene is

IX {3 0 0)
{3 IX {3 0
( (3.19)
O{3IX{3 ,
o 0 {3 IX
and the bonding orbitals are at energies 8 1 = IX + 1.62{3 and 8 2 = IX + O.62{3
(note that {3 is negative in the conventional Huckel treatment) and the anti-
bonding orbitals are at 8 3 = IX - O.62{3 and 8 4 = IX - 1.62{3. Figure 3.6 illustrates
the interaction and resulting MOs and energy levels for the cases of ethylene and
butadiene, discussed above.
 3.4 Interactions Between Localized Orbitals: Conjugation 65

C atom Ethylene Butadiene

Fig. 3.6. Interaction of atomic p-orbitals on carbon atoms to form the 1t systems of ethylene
(H 2 C=CH 2 ) and butadiene (H 2 C=CH-CH=CH 2 ). For ethylene the interaction results in an
antibonding 1t: and bonding 1t. orbital. For butadiene (H 2 C=CH-CH=CH 2 ) there is an inter-
action between two (identical) 1t systems, 1t. and 1tb' and a splitting within the antibonding and
bonding system occurs, as shown. In first order perturbation theory the splitting within the
antibonding and bonding systems is treated independently. In second order, the interaction between
the two manifolds leads to an increased overall splitting but does not effect the internal splitting
within the bonding and anti bonding manifolds

From the examples it is apparent how the Hiickel scheme can be used to
predict the energies of the bonding and antibonding orbitals of a polyene or
polyine chain of any length. In fact, it is possible to obtain a general expression
for the Hiickel energies for the interaction of n orbitals in a chain of n atoms
[3.10]

em = IX + 2{3COS(~).
n+l
(3.20)

The eigenvalues are labeled by the index m = 1, 2, ... , n. For a polyene or

polyine n is an even number, and it is twice the number of double bonds. From
(3.20) we see that in the limit of a long chain, n -+ 00, the energies become a
continuous band with peaks at the extreme values IX + 2{3 and IX - 2{3. The
splitting in the anti-bonding and bonding manifolds for the polyenes, measured
by means of electron transmission spectroscopy and photoemission, respectively
[3.29], is shown in Fig. 3.7.
Although Hiickel theory is typically used for the description of 1t electron
systems, one of the most beautiful examples of the level splitting predicted by the
theory is that of a (J system, i.e., of the carbon 2s level in the paraffins [3.13,
30--32] shown in Fig. 3.8. In going from methane (CH 4 ) to ethane (H 3 C-CH 3 )
the 2a 1 , or carbon 2s derived orbital, splits into a bonding and anti bonding pair
due to interaction between the two carbon orbitals. For the larger paraffins the
splitting pattern follows that predicted by (3.20), e.g., a triplet for propane
66 3. Symmetry and Molecular Orbitals

Binding Energy reI. to Vacuum Level (eV)

30 20 10 30 25 20

IJi~
(c)
Theory
2. 2. Band

Splitting of1r System in Polyenes 2p l CH 4

~ I I C2 HS

M;
:;
~
~ -4 -4 f I C3 H S

~ C2 HS
i I II n-C4 H lO

&.
~
:{
.~

I, ,
-2 "~ f I I ,'J n-C s H 12
guc: -2
~
~ :e I, n-C s H 14

lYG,'"
W ~

'" 0 o ~
"
c:
.. ii'
I ,j n-C 7H 1S
.!l
ill, I ,i
I,,,
:; -= n-C s H 1S

II
~ +8 +8

~c: I n-C g H 20
., n;20
~+10 +10

.,
c:
.o
:!l
'g +12 +12

"
, Binding Energy reI. to Fermi Level (eV)

Fig.3.7. Splitting of the n system in the polyenes: ethylene, butadiene, cis- and trans-hexatriene, and
all-trans-l,3,5,7-octatetraene [3.29]. The electron affinities of the unoccupied n* leyels were meas-
ured by electron transmission spectroscopy and the ionization potentials of the occupied n system
by photoemission

Fig. 3.8. (a) Splitting of the carbon 2s atomic level in the gas-phase paraffins methane (CH 4 )
through nonane (C 9 Hzo) observed in X-ray photoemission spectroscopy (XPS) [3.13]. Note that at
XPS energies the cross section of the 2s level is enhanced relative to the 2p level. The binding energy
is referenced to the vacuum level. (b) XPS spectrum of the valence band of polyethylene showing the
splitting of the 2s level into two peaks [3.33]. Here the binding energy is referred to the Fermi level
and differs from that in (a) by a work function and extramolecular relaxation shift. (c) Splitting of
the 2s level with increasing paraffin size as calculated by an ab initio molecular orbital calculation
[3.13]. (d) Results of an ab initio calcula-tion of the valence band density of states for polyethylene
where the binding energy scale has been corrected as discussed by Andre et al. [3.34]. In order to
correlate the splitting ofthe 2s band in the paraffins with that in polyethylene (limit of infinitely long
chain) we have aligned the two peaks in the polyethylene density of states with the limits of the 2s
splitting in (c)

(H 3 C-CH 2-CH 3 ), a quartet for butane (H 3 C-CH 2 -CH 2 -CH 3 ), etc. In the long-
chain limit, i.e., for polyethylene ([-CH 2-CH 2 - ]n), two peaks separated by
about twice the splitting for ethane are found [3.30] as predicted by (3.20). Note,
though, that the splitting of the mainly 2p derived It 2 level of methane through
the paraffin series is more complicated than that ofthe 2s derived 2a 1 level, both
because of mixing with hydrogen orbitals, and because some of the anti bonding
 3.5 Splitting of Antibonding Orbitals Due to Bond-Bond Interactions 67

orbitals are unoccupied. In fact, it is transitions to these unoccupied anti bonding

orbitals which give rise to K-shell excitation resonances.

3.5 Splitting of Antibonding Orbitals

Due to Bond-Bond Interactions

3.5.1 The Linear Combination of Bond Orbitals Method

Often, either the bonding or the antibonding orbitals, but not both, are observed
by a given spectroscopic technique. As illustrated in Fig. 3.7, photoemission
[3.35] probes filled bonding orbitals and inverse photoemission [3.36] or
electron transmission spectroscopy [3.29, 37] probes empty antibonding or-
bitals. X-ray absorption spectroscopy also probes unfilled antibonding orbitals
which are the final states of the K-shell transitions. We can therefore simplify
our treatment by considering only the splitting of the anti bonding n* and (1*
orbitals. Hence, rather than starting with atoms we begin with an atomic pair,
i.e., a bond, and consider the energy splitting of the anti bonding orbitals of a pair
caused by first order bond-bond interactions between the pair of atoms, the so-
called Linear Combination of Bond Orbitals (LeBO) model. We neglect second
order interactions between the bonding and anti-bonding manifolds, which are
included in the Huckel treatment given in Sect. 3.4.2. As illustrated by Fig. 3.6,
second order interactions mainly lead to a repulsion between the two manifolds,
not significantly perturbing the internal level splitting within the bonding or
antibonding groups. Note that second order perturbations are, however, im-
portant in the consideration of collective molecular properties, like the total
energy, as discussed by Dewar and Dougherty [3.22].
The LeBO scheme is particularly useful if there is a characteristic "energy"
associated with a bond. We shall see below that this is, in fact, the case in K-shell
excitation spectra, where different molecular bonds are characterized by differ-
ent resonances which occur at specific resonance energies. Interactions between
adjacent bonds can then be treated by first or second order perturbation (Sect.
3.4.1) and lead to a splitting of K-shell resonances or resonance shifts as
schematically shown in Fig. 3.5. The LeBO formulation reduces the dimen-
sionality of the bond matrices by a factor of 2, as illustrated below for the case of
n*-n* interactions in the polyenes and (1*-(1* interactions in the paraffins.

3.5.2 Application to (1 and 1r Bonds in Hydrocarbons

If we define the self-energy of the n* anti bonding orbital of ethylene as A we can
account for the n* splitting in butadiene by an interaction energy B between the
two n bonds (Fig. 3.9). Hence the bond matrix for butadiene is

(3.21)
68 3. Symmetry and Molecular Orbitals

Butadiene Fig. 3.9. Illustration of bond- bond interactions

(conjugation) for the 1[ system in the polyenes and
the (J system in the paraffins, using the correspon-

~:::~~-!(:t ding cases, butadiene and propane, as examples.

For the polyenes the interaction B n - n is between
H--- C ~
. I.. C --- H 1[ bonds belonging to pairs of double bonded

II········· · ~ · · · · ··~\
atoms. For the paraffins the interaction b._. is
between the single bonds of a C-C pair. Overall,
Brr- rr there are always the same number of 1[ and spJ
hybrid orbitals on corresponding members of the
series, e.g., four 1[ orbitals in butadiene and four
spJ hybrid orbitals in propane, six 1[ orbitals in
Propane hexatriene and six spJ hybrid orbitals in butane,
etc.

and the eigenvalues are 81,2 = A ± B. Similarly for hexatriene the bond matrix
and eigenvalues are

U~ n·
and 8 1 = A + fiB , 8 2 = A, and 8 3 = A - fi B. Again, the extension to a
(3.22)

larger chain is obvious and the eigenvalues follow from (3.20). Another inter-
esting example is benzene. In this case the bond matrix is related to that of
hexatriene with an additional off-diagonal matrix element due to the ring
closure, i.e.,

(3.23)

The eigenvalues are 8 1 = A + 2B, and 8 2 , 3 = A - B, such that there are two 11:*
antibonding orbitals, one of which is a degenerate pair. An extension of the
above concepts leads to a general analytic expression for the energy levels of the
11: system in unsaturated cyclic molecules and to Hiickel's rule on aromatic
reactivity [3.10].
As a second example for the LCBO model, let us consider the C-C cr bonds
of a paraffin chain. The starting point of our treatment is the "self-energy"
associated with the C-C anti bonding cr* orbital in ethane (H 3 C-CH 3 ), which is
 3.6 Orbital Orientation, Symmetry, and the Dipole Selection Rule 69

split by an increasing number of bond-bond interactions as we increase the

chain length to propane (H 3C-CH r -CH 3) with two interacting C-C single
bonds, butane (H 3C-CH 2 -CH 2 -CH 3) with three interacting C-C single bonds,
etc. For propane the bond matrix is given by

(3.24)

and the eigenvalues are 61,2 = a ± b. If we only consider nearest neighbor bond
interactions we obtain the following bond matrix for butane:

a
( b
b
a
0)
b , (3.25)
o b a

with eigenvalues 61 = a + fib, 6 2 = a, and 6 3 = a - fib. Comparison be-

tween the 11: interactions in the polyenes and the (T interactions in the paraffins
shows a one-to-one correspondence of the bond matrices and eigenvalues of
propane and butadiene, butane and hexatriene, etc. This has previously been
discussed by Dewar [3.38] in terms of a comparison of 11: conjugation in the
polyenes and (T conjugation in the paraffins. While in the polyenes the
bond-bond interaction B is between 11: systems belonging to pairs of double
bonded atoms, the bond-bond interaction b in the paraffins is between C-C
single bonds, as shown clearly in Fig. 3.9.
This figure also shows that, although there are only three carbon atoms in
propane, there are four carbon hybrid orbitals involved in the C-C bonding,
similar to the four 11: orbitals in butadiene. As pointed out by Dewar [3.38] the
above scheme leads to an equivalent description of the 11: system in benzene and
the (T system in cyclopropane (C3H6)' In cyclopropane there are interactions
between two Sp3 hybrids on each of the carbon atoms and the bond matrix is
given by

a
( b
b b)
a b , (3.26)
b b a

which is identical in structure to that of the 11: system in benzene (3.23).

3.6 Orbital Orientation, Symmetry, and the Dipole Selection Rule

3.6.1 Orbital Orientation and Angular Dependence of the Dipole Matrix Element
One important feature of molecular orbitals is that they have strong directional
character and there is a one-to-one correlation between the spatial orientation
of the orbitals and the molecular geometry. Thus for oriented molecules and
70 3. Symmetry and Molecular Orbitals

polarized X-rays the intensities of resonances associated with, for example, a*

and n* final states should exhibit a dramatic and different angular dependence.
The polarization dependence of the intensity of a particular resonance
associated with a specific molecular orbital can be derived quite generally from
the expression for the X-ray absorption cross section given by (2.8). The angular
dependence is contained in the dipole matrix element and according to (2.8, 11)
we can write for the transition intensity Ii!'

(3.27)

assuming the X-rays are linearly polarized in the direction of the unit vector e.
Let us first evaluate the dipole matrix element <flrl i) in (3.27). We shall restrict
ourselves to K-shell excitation, firstly because of its importance for low-Z atoms
and secondly because it exhibits the strongest polarization-dependent effects.
The initial Is state Ii) = R1s(r) is spherically symmetric and it is to a very
good approximation represented by the atomic Is wavefunction of the excited
atom in the molecule. In a one-electron model, the upper state of a bound-state
transition can be represented by a LCAO wavefunction (2.34) of the correspond-
ing molecular orbital. Because of the localization of the Is initial state, the
atomic valence components of the excited atom in the LCAO wavefunction
dominate in the evaluation of the dipole matrix element. For the second row
atoms, one may therefore consider only these terms and write the final state
wavefunction as a linear combination of atomic 2s and 2p states on the excited
atom, i.e.,

= aR2s(r) + R 2p (r)( b sin fJ cos ¢ + c sin fJ sin ¢ + d cos fJ) , (3.28)

where the coefficients a, b, c, and d give the weight of the atomic orbitals in the
LCAO expansion, and R2s(r) and R 2p (r) are the radial atomic wavefunctions.
Equation (3.28) shows that the maximum amplitude of the final state orbital is in
the direction 0 determined by the superposition of the three p orbitals, namely,
0= bex + cey + dez , where the unit vectors ei are along the axes of our
coordinate system.
We express the position vector in spherical coordinates, r = r(sin fJ cos ¢ ex
e
+ sin fJ sin ¢ ey + cos ez ), the vector matrix element in (3.27) can be evaluated
by integration and we obtain

(3.29)

where

(3.30)

IS the radial dipole matrix element. Thus, according to (3.29), for K-shell
 3.6 Orbital Orientation, Symmetry, and the Dipole Selection Rule 71

excitation, the vector matrix element points in the same direction as the p-
component in the final state orbital on the excited atom and thus the polarization
dependence of the total matrix element can be expressed as a function of the
angle lJ between the direction e of the electric field vector and the direction 0 of
largest amplitude of the final state orbital, namely

(3.31)

As an example, let us consider the 0"* and n* orbitals of a triple bonded

diatomic, e.g., N2 or CO. We orient our coordinate system so that the 0"* orbital,
and therefore the internuclear axis, lies along the z-axis, and the two n* orbitals
lie along the x- and y-axes, and specify the orientation of the unit electric field
vector e by spherical angles (J' and 4J'. The polarization dependence of the
resonance intensity associated with the 0"* molecular orbital final state is then
simply given by

(3.32)

Therefore the 0"* resonance intensity is greatest for E along the internuclear
axis and vanishes when E is perpendicular to it. The polarization dependence of
a resonance associated with the two energetically degenerate n* orbitals can be
derived from (3.31) by integration over all E vector orientations in the x-y plane
and is obtained as the sum of the individual intensities of n* orbitals along ex
and along ey,

Ii/(n) oc le·ex l2+ le·eyl2

oc sin 2 (J' cos 2 4J' + sin 2 (J' sin 2 4J' oc sin 2 (J' • (3.33)

Resonances associated with 0"* and n* orbital final states, therefore, have a
strong and opposite polarization dependence. Equations (3.32) and (3.33) are the
foundation for the determination of molecular orientations on surfaces by
means of NEXAFS and detailed angular effects will be discussed in Chap. 9.
We have so far assumed that the final state has a p orbital component. For a
Is initial state this is the prerequisite for an allowed transition, dictated by the
dipole selection rule. Because the K-shell excitation occurs on a particular atom
due to spatial localization of the Is initial state, it is the local symmetry of the
MO final state on the excited atom which determines whether the transition is
allowed. The importance of spatial localization in K-shell excitation is revealed
by the radial part of the dipole matrix element given by (3.30). Clearly it is the
overlap of the Is and 2p functions near the atomic core which determines the
matrix element. This is the origin of the often used one-center approximation in
the calculation of K-shell X-ray emission spectra [3.39,40]. One simple method
of checking whether a transition from the K -shell to a particular final state is
allowed is to inspect a plot of the MO final state and look for a node, indicating
local p-character. Note that for second row atoms, d and f orbital final states
which also have nodes do not contribute to the MOs. Inspection of Fig. 3.2
72 3. Symmetry and Molecular Orbitals

reveals that in CO 2 transitions from the carbon K-shell are allowed to the 40":
(node) but not to the 50": (no node) MO, while both transitions are allowed for
excitation of the oxygen K-shell.

3.6.2 Group Theory and the Dipole Selection Rule

In many cases plots of the empty MOs are not available and another, more
general, method must be used to determine whether specific transitions are
allowed. Again, group theory is employed. Since the irreducible representations
contain all the symmetry information for a given point group, we can determine
whether the transition matrix element <flrli) is zero or not simply by inspec-
tion of the representations for the initial core state and the MO final state. The
group theoretical method is not only elegant but also of general validity and can
thus be applied to determine the selection rules for transitions between any two
states.
Before we consider this approach, we must discuss different ways of consid-
ering the effect of the core hole. The simplest approach is to use the broken
symmetry of the final state. The broken symmetry is obtained from the ground
state symmetry by assuming that the excited atom is changed by its core hole
and as such is different from other atoms in the molecule with the same atomic
number. In this case the Is initial state transforms like the totally symmetric
representation of the broken-symmetry point group and the selection rules for
the allowed transitions are worked out in the broken symmetry. The second
method treats the core hole as if it were delocalized (which of course in practice it
is not) and in this case the Is initial state corresponds to those irreducible
representations which describe the possible delocalized hole combinations in the
full symmetry of the molecular ground state, and the selection rules are worked
out in this symmetry. The third, and rigorously correct approach, is to consider
the whole set of broken symmetry configurations and form linear combinations
thereof such that the transition can take place within the higher symmetry of the
molecular ground state [3.41, 42].
In practice, the three procedures normally arrive at the same result. For a
quick check of the allowed transitions, it is often convenient to use the second
method since MO calculations are typically carried out in the ground state
molecular symmetry and the MOs published in the literature are labeled with
the irreducible representations of that symmetry.
For a given symmetry and initial and final state representations, the selection
rules are worked out as follows [3.14]. Since the dipole operator r is a vector
with components (x, y, z), the dipole transition will be allowed if the direct product
of the irreducible representations of the initial and final states is or contains the
irreducible representation to which x, y, or z belongs. Furthermore, if we find that
the direct product transforms like x, y, or z, we know from (3.31) that the
transition will be observed strongest if the electric field vector E is pointing in
that direction. The direct product is easily calculated from the respective
character table. First, we calculate the characters of the direct product which are
 3.6 Orbital Orientation, Symmetry, and the Dipole Selection Rule 73

equal to the product of the respective pairs of characters of the representations.

The direct product is then the sum of those irreducible representations whose
characters sum to the calculated characters of the direct product [3.14]. Tables
of direct products are given by Herzberg [3.43].

3.6.3 Applications of Group Theoretical Selection Rules

By way of illustration we will now discuss the expected K-shell spectra for NH3
and CH 4 . In both cases, the symmetry of the molecules in the ground state is
identical to that of the excited state with a hole on the N or C atom, respectively.
Thus we simply consider the selection rules for the molecules in the ground state.
For NH3 the nitrogen Is state transforms like AI' and the empty MOs
transform like Al and E, respectively. According to Table 3.1, for the C 3v point
group the direct product Al x Al = Al transforms like z. Hence the la l -+ 4a I
transition (Fig. 3.3) is allowed and it is strongest for E parallel to z, the symmetry
axis of the NH3 molecule. The second transition from the la l to the 2e orbital is
also allowed since Al x E = E, which transforms like (x, y). This transition is
maximized for E in the x-y plane, perpendicular to the symmetry axis of the
molecule. For CH 4 the la l -+ 3a I transition is forbidden since for the Td point
group the direct product Al x Al = Al transforms like X Z + yZ + zZ. The reason
is, of course, that the 3a I orbital has no p-character, as mentioned earlier. The
transition la l -+ 2t z is allowed since Al x 12 = 12 transforms like (x, y, z). There-
fore, the intensity of this transition has no angular dependence, as would be
expected for cubic symmetry.
COz in its ground state has the symmetry Dooh with the corresponding
character table given in Table 3.1. Let us first consider selection rules for this
symmetry. The allowed transitions follow from a general rule, which forbids
transitions between two states of the same inversion symmetry. Hence g -+ g and
u -+ u transitions are forbidden. The reason is that the dipole operator is un-
gerade with respect to inversion and hence the direct product of initial and final
states must be ungerade, too. For the oxygen K-shell in COz, the initial state is a
nearly degenerate lUg and luu pair and the transitions lUg -+ 2n:, luu -+ 5u:,
and lUg -+ 4u: are allowed (Fig. 3.2). In contrast, for the carbon K-shell excita-
tion, the initial2ug MO is of gerade symmetry only, and therefore the 2ug -+ 5u:
transition is forbidden.
Consideration of the core hole does not change the symmetry for the carbon
K -shell excitation, and therefore the above arguments also apply for this case.
For the oxygen K-shell excitation, however, the effective symmetry created by
the localized hole on one of the oxygen atoms is reduced to C oov . We now need
to work out how the irreducible representations in Doob map into those of C oov .
This is easily done by comparison of the characters for the symmetry operations
which are common to both symmetries. If the characters correlate, the two
respective representations map into each other. Simpler yet, correlation tables
between the irreducible representations of different point groups can be found in
Ibach and Mill's [3.44] and Herzberg's [3.43] books. Comparison of the
74 3. Symmetry and Molecular Orbitals

character tables in Table 3.1 readily shows that I'g and I'u in D oo h symmetry both
correspond to I' in C oo v symmetry, and the same applies for the n representa-
tions. Thus for the effective core-hole-induced C oo v symmetry we have two
empty (1* MOs which can both be reached from the 0 Is state, which is also of (1
symmetry. We therefore obtain the same result, independent of core hole
localization, for both the carbon and oxygen K-edge excitations.
Finally, we shall consider the selection rules for a bent triatomic molecule of
the general form XYX, such as N0 2 • This case will be of importance in later
chapters for the chemisorbed CO 2 and HCOO- species. The evolution of MOs
from a linear triatomic with symmetry Doo h (e.g., CO 2 ) to a bent triatomic
molecule with symmetry C 2v , can be conveniently treated in the form of a Walsh
diagram [3.8, 43] (Fig. 3.10). For the bent molecule we have chosen the z-axis
along the C 2 axis and the x-axis in the molecular plane. We find the upper three
states to be of b l , ai' and b2 symmetry and therefore need to work out the

Walsh Diagram for YX 2

A 5b,
___ l~Y
~x

A 7a,

m
---~50g 0<>G<:>0

A 2b 2I----------:::;:::=----i 21Tu 000

X 4b,
la 2

*X
6a, 11T9 He--o
5a,
l b2 11Tu m~
3b,
30u f><O>OO

A 4a,
40g CKO:>o

X 2b,
3a,
20u
30g
0--0
~

90· 180·
~XYX

Fig.3.10. Walsh diagram for the evolution of molecular orbitals from a linear triatomic molecule of
general form X YX, e.g., CO 2 , to those of a bent triatomic molecule, e.g., N0 2 , as a function of the
bending angle. The MOs are labeled by their irreducible representations. For the bent molecule we
have chosen the z-axis along the twofold symmetry axis and the x-axis in the molecular plane, as
shown by the inset. Note that our choice of the x- and y-axes determines which MOs correspond to
the bl and b2 representations. The labeling would be reversed if the y-axis were chosen in the
molecular plane
 3.7 Spin-Dependent Excitations 75

Table 3.3. Symmetry and polarization of resonances in a bent triatomic molecule X YX'

Core state Transition Polarization

(E vector orientation)

X Is la l -+ 6a l (uT) z
la l -+ 2b 2 (It*) Y
la l -+ 7a l (un z
la l -+ 5b l (un x

X Is Ib l -+ 6al (uT) x
Ib l -+ 2b 2 (It*) Forbidden
Ib l -+ 7a l {un x
Ib l -+ 5b l {un z

Y Is 2a l -+ 6a l {un z
2a l -+ 2b 2 (It*) Y
2a l -+ 7a l {un z
2a l -+ 5b l {un x

• C2v symmetry; the z-axis is in the C2 axis and the x-axis in the molecular plane, and assuming that
the Is ionization potential of atom X is larger than for atom Y

polarization of the transitions from the X 1s and Y 1s levels to these MOs. Let us
first consider the transitions in the full ground state symmetry. The 1s orbital of
the central atom Y (i.e., the 20" MO if we assume that the 1s binding energy of
atom X is larger than of atom Y) clearly transforms as the symmetric repres-
entation Al in the character table (Table 3.1). The 10" MO of the outer atoms X
can be constructed from a symmetric or antisymmetric linear combination of the
two 1s wavefunctions. The symmetric combination transforms like A l , the
antisymmetric combination like B l • Using Table 3.1 we can now work out the
polarization of the various transitions; the results are given in Table 3.3 [3.45].
In the Table we have also listed the selection rules for transitions to the 6al MO,
which is the LUMO for many triatomics, e.g., N0 2 with a O-N-O bond angle
of 134.0°.
For the localized-core-hole case, excitation of the central atom does not
reduce the symmetry, while excitation of the outer atom X results in a symmetry
reduction to C•. The Al and Bl representations in C2v symmetry correspond to
the A' representation in C. and the B2 representation to the A" representation
[3.43]. The X 10" MO transforms like A', and A' - A' transitions are allowed for
E in the molecular x-z plane (0"*), and A' - A" transitions are allowed for
Ell y (n*). Hence we obtain the same polarization dependence as for the full
ground state symmetry.

3.7 Spin-Dependent Excitations

Thus far we have tacitly assumed that it does not matter whether we excite the
spin-down or the spin-up electron in the K-shell. This is in fact true for alI closed
76 3. Symmetry and Molecular Orbitals

shell atomic and molecular systems. The reason is that for electric dipole
excitations there is no difference in exchange interaction in closed shell systems
between the different possible final states attainable by K-shell excitation. In
short, the dipole selection rule states that the total spin is conserved during
excitation because the dipole operator does not act on spin. For ionization, the
photoelectron therefore always has opposite spin to that of the electron left
behind in the K-shell and the two possible final states are energetically degener-
ate, and therefore there is only one IP. For bound state excitations the situation
is similar, with the electron in the K-shell and the excited electron in a Rydberg
or molecular orbital having opposite spins. Again, the two possible spin config-
urations for a given two-level system are energetically equivalent and there is
one transition energy.
Let us discuss three examples to illustrate spin-dependent excitations,
namely the K-shell spectra of the molecules N z, NO and Oz. The energy level
scheme for these molecules in the ground state is shown in Fig. 3.1. For N z and
Oz the 1au and lag levels are degenerate in the ground state and all levels up to
the 3aglevel are filled. The doubly degenerate 1ng (or 1nt) orbitals correspond to
the lowest unfilled levels, being empty in N z and half full (two electrons) for O 2 ,
For NO the equivalent of the 1nt orbitals are called the 2n* orbitals and one of
them is singly occupied. Hence N z is a closed shell molecule and both NO and
Oz are open shell molecules and therefore paramagnetic.
The energy levels relevant to the following discussion of the K-shell spectra
of the molecules are shown in Fig.3.11a. For N z we only consider the n*
excitation. According to the dipole selection rule excitation of a spin-down
(spin-up) electron from the 1au level leads to a spin-down (spin-up) electron in
the In: level and both processes are energetically equivalent, leading to only one
n* resonance. The spin-flip transition shown in Fig. 3.11a is dipole forbidden
since the total spin of the 3 n final state is different to that of the 11; initial ground
state. This transition can in fact be observed in ISEELS experiments carried out
under nondipole conditions [3.46] and the energy separation between the two
configurations, arising from a difference in exchange interaction, was found to
be d 1 = 0.82 ± 0.01 eV.
If a spin-up occupation of one of the 2n* orbitals in NO is assumed, only
excitation of a spin-down electron from the 1a (0 Is) or 2a (N Is) orbital to the
half occupied orbital is allowed by the dipole selection rule and the Pauli
principle. Excitation of either a spin-up or spin-down electron is allowed into the
second, empty 2n* orbital. The exchange splitting between the possible three
final states has not been observed and was therefore inferred to be < 1 eV
[3.49]. Ionization of a K-shell electron may lead to two different ionic states
resulting in two different IPs. As shown in Fig. 3.11 b, the two ionic final states
labeled 1 nand 3 n are energetically nondegenerate because of the different spin
coupling and hence exchange interaction in the two states. The exchange
splitting d z has been determined by photoemission to be 1.42 eV for N Is and
0.55 eV for 0 Is excitation [3.47]. Note that for ionization it is customary to
label the final state, by ignoring the photoelectron, as that of the ion left behind.
 3.7 Spin-Dependent Excitations 77

(a) Nitrogen (b) Nitric Oxide

Vacuum __. -__________. -__

---------/j.-:--,- - - - - - - Level

-fL=+=
! I ==<IH== 2,,'
l"g f

1,"
dipole
forbidden

-+t---
I
lau

(c) Oxygen

r
Vacuum Level

'"i

forbidden

I
lau 4+--
Fig. 3.11. (a) Energy level diagram illustrating the origin of the n* resonance in the nitrogen
molecule. The spin-dependent dipole selection rule connects the lu u (N Is) core level and the empty
In: level with conservation of the total spin. Hence from the 1;[ molecular ground state the spin-flip
transition to the 3 [J final state is forbidden but the spin-conserving transition to the I [J final state is
allowed. The energy separation ~I = 0.82 eV has been measured under nondipole conditions [3.46].
(b) In the open shell molecule nitric oxide the occupation of the 2n* orbital by a single electron in
the ground state leads to two different K-shell ionization potentials depending on whether a spin-up
or spin-down electron is excited from the 10' (0 Is) or 20' (N Is) orbital. The exchange splitting ~2 has
been measured by photoemission to be 1.42 eV for N Is and 0.55 eV for 0 Is excitation [3.47]. (c) K-
shell transitions in the paramagnetic O 2 molecule which in its 3;[ ground state has an unpaired
electron in each of the two degenerate In: orbitals. According to Hund's rule, the two spins are
parallel and we shall assume that they are both up. Then excitation of a spin-up electron from the K-
shell to the In:
orbitals is forbidden but excitation of a spin-down electron is allowed, corresponding
to a 3[J final state. As in NO there are two ionization potentials, labeled 2;[ and 4;[, as discussed in
the text. Also, in O 2 there are two 0'* resonances, corresponding to excitations of either a spin-up or
spin-down electron from the K-shell to the 30': orbital.

Relative to the 2II ground state of NO there is thus one electron less in the ionic
final state and this accounts for the change in spin by ± 1/2 between the 2 II
ground state and the 1 II and 3 II excited states. As required by the dipole
selection rule the spin of the total system (ion plus photoelectron) is preserved.
7'ii 3. Symmetry and Molecular Orbitals

Finally, we discuss the particularly interesting case of the O 2 molecule

(Fig.3.11c). In the 3}; ground state the In: orbitals are occupied by two
electrons with parallel spins according to Hund's rule, and for the following
discussion we shall assume that their spins are up. Because of the dipole
selection rule and the Pauli principle only a spin-down electron from the 100u
n:
orbital can therefore be excited into either 1 orbital, giving rise to the 3 II
molecular state. Similar to NO, there are two Is IPs corresponding to excitation
of a spin-up e}; ionic final state) and a spin-down (4}; ionic final state) electron
from the 100u or 10"g orbitals. The corresponding exchange splitting Ll4 has been
determined to be 1.1 eV by photoemission [3.16]. The K -shell spectrum of
oxygen is different from those of N2 and NO in that the transition to the next
higher 30"~ orbital (Fig. 3.1) has a transition energy less than the Is IP [3.50].
For the 10"g --> 30"~ excitation the dipole selection rule only allows the two 3};
final state configurations shown in Fig. 3.l1c, with a difference Ll3 in binding
energy of the 30"~ orbital. Thus the difference in 100g --> 30"~ excitation energy for
the two final state configurations, given by Ll3 - Ll4' corresponds to the splitting
of the two corresponding 0"* resonances in the NEXAFS spectrum. Experiment
shows this splitting to be remarkably large, ~ 3 eV [3.48], as discussed in
Sect. 4.2.8.
In the next chapter we shall give an overview of experimental techniques and
results for K-shell spectra of free molecules and apply the theoretical concepts
discussed above to obtain an understanding of the spectra.
4. Experimental and Calculated K-Shell Spectra
of Simple Free Molecules

Here an overview is given of the spectral features that are encountered in K-

shell excitation spectra of simple gas phase molecules. Experimental results ob-
tained by the inner shell electron energy loss, or ISEELS, technique - which
is discussed first - are compared to those obtained by different theoretical
techniques.

4.1 Experimental Methods: The ISEELS Techniques

Although historically X-ray absorption spectroscopy has been the technique of

choice for the study of inner shell excitation processes in matter, our detailed
understanding of the K-shell spectra of low-Z molecules is largely due to a
different technique, namely inner shell electron energy loss spectroscopy, or
ISEELS [4.1, 2]. For example, this technique which closely approximates X-ray
absorption spectroscopy has been used to study the carbon K-shell spectra of
more than one hundred molecules [4.2] while only a few such studies have been
carried out by means of X-ray absorption [4.3-13]. A summary of the early X-
ray absorption and some of the early ISEELS work has been given in a review
by Koch and Sonntag [4.14].
One reason for the dominance of IS EELS in investigating molecules lies in
the long-time unavailability of highly monochromatic, intense tunable photon
radiation in the 250-1000eV range. Even with the utilization of synchrotron
radiation in the early 1970s, problems with beam line optics impeded extensive
work in this spectral range [4.15]. Another important reason is the experimental
convenience of ISEELS measurements. ISEELS spectrometers are relatively
inexpensive and home based, which is why they have been called the "poor
man's synchrotron". Other factors are their extremely high and energy-inde-
pendent energy resolution ( ~ 50 me V) and the fact that ISEELS measurements
do not suffer from saturation effects in the recorded spectra which plagued early
X-ray absorption studies, e.g., the pioneering N2 K-shell work by Nakamura et
al. [4.16] as discussed by Connerade et al. [4.17]. The resolution capabilities of
ISEELS at the C K-edge of CO and the N K-edge of N2 are demonstrated in
Fig. 4.1, which shows vibrational fine structure in the Is -+ n* transitions of the
molecules [4.18-20]. Similar resolution has only very recently been achieved
80 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

Vibrational Structure In fT* Resonance Fig. 4.1. Vibrational structure of the

Is -+ n* transition recorded with high
1.0
N2 resolution ISEELS for N 2 [4.19] and CO
Eo = 2.5 keY (C K-shell) [4.20] gas. Eo denotes the
6E = 90 meV primary electron beam energy and I'lE
0.8 the instrumental energy resolution. The
N2 spectrum is least-squares fitted with
six Lorentzians
0.8

~ 0.4&.....1!~c...
c:

.e'" 400 401 402

~ ;-~-r~~~~--~-.--~~~~
~
'iii
~1.4
..:- CO
.s Eo = 1.5 keY
6E = 55 meV
1.0

0.6

0.2 ) \ . ' \.....

o I:....:---::::;;:....I.--.J..~::;:::::::::::::::.J
287 288
Energy Loss (eV)

with X-ray absorption spectroscopy [4.21, 22], ten years after the IS EELS
spectra in Fig. 4.1 were recorded. We will briefly describe the ISEELS technique.
The energy loss of electrons passing through gases was studied as early as
1914 by Franck and Hertz [4.23]. These experiments are often used in physics
textbooks to explain the different nature of electron versus photon interactions
with atoms. In collisions, energetic electrons (a few keY) can lose any fraction of
their energy through electronic excitations of atoms or molecules while the
interaction of photons with matter (except for energies above about 100keV
when incoherent scattering becomes important) is characterized by complete
loss of energy, i.e., absorption. The resonant character of X-ray absorption is the
reason for saturation effects in the recorded K-shell structures at high molecular
densities [4.17] while such effects are nonexistent in the nonresonant ISEELS
measurements. Ironically, Franck and Hertz's experiments, which demonstrated
the difference between X-ray and electron excitations of atoms, are also the basis
for application of electron energy loss spectroscopy in a manner equivalent to
photon absorption.
In 1930 Bethe [4.24] showed that under certain conditions the electron
energy-loss cross section can be directly related to the X-ray or optical absorp-
tion cross section, where the energy loss E) of the electron during the collision
 4.1 Experimental Methods: The ISEELS Techniques 81

event is equivalent to the photon energy. The latter equivalence is easy to

understand since the energy required to excite an atom from state Ii) to state
<II is furnished by energy loss in one case and photon absorption in the other. If
the energy loss of the electron during the collision event is small compared to its
kinetic energy, the interaction between the electron and the atom can be treated
in lowest order and the first Born approximation can be used. In this case the
number of electrons that have lost an energy E) in a collision with an atom and
are scattered in a direction characterized by the momentum qf = qi - Q. where
Q is the momentum transfer, is proportional to the double-differential cross
section [4.24, 25]

(4.1)

where I(Q, E) is a Q- and E)-dependent function [4.25, 26] and the summation
is over all electrons with coordinates rj in the target atom. In the limit Q' r ~ 1,
where r is the radius of the initial core state, the exponential factor in (4.1) can be
simplified by expansion in a power series

(4.2)

Because the states Ii) and <II are orthogonal, the matrix element ofthe first
term in the series vanishes. The second term yields the familiar dipole matrix
element, the third term the quadrupole matrix element and so on. Note that the
dipole approximation in X-ray absorption corresponds to approximating the
exponential term in (2.7) to 1, the first term of the power series, while in IS EELS
the exponential term in (4.2) is approximated by the first two terms 1 + i Q' r. If
we choose our experimental conditions such that the momentum transfer Q is
small, the dipole term will dominate over the higher multipole terms. Further-
more, if we define r = Ljrj and integrate over the finite angular acceptance of
the spectrometer up to a maximum angle Om from the incident beam direction,
we obtain [4.25-27]

(4.3)

where Eo is the incident beam energy and eQ is a unit vector in the direction of
the momentum transfer Q. It is interesting to compare the above expression to
that governing X-ray absorption which, according to (2.8. 11), can be written as
[4.26,28],

(4.4)

where q is the wave vector magnitude of the photoelectron and eE is a unit vector
in the direction of the electric field vector E of the X-rays. Thus in the limit of
small momentum transfer, ISEELS spectra are dominated by dipole transitions
where the Q vector takes the role of the E vector in X-ray absorption. The
"polarization" dependence of ISEELS has in fact been utilized to study the
82 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

anisotropy of solids [4.26,29-31], and it has also been used to probe structural
and dynamic anisotropies of surfaces [4.32], and the orientation of chemisorbed
molecules [4.33, 34].
Experimentally the ISEELS cross section dUe/dE) (units: cm 2 /eV) and X-ray
absorption cross section U x (units: cm 2 = 10 24 barn) are measured by moni-
toring the transmission of an electron or X-ray beam through a gas character-
ized by its volume density nv (atoms/cm 3 ) and the X-ray path length t through
the gas. If we denote by 10 , the incident photon or electron flux density, [e.g.,
photons/(scm 2 )] and by Ao the area of the sample exposed to the beam (cm 2 ),
the transmitted number of photons N ph is given by the familiar expression

(4.5)

The product nvux = J.i.x is called the linear X-ray absorption coefficient. In
ISEELS the incident beam has an energy Eo and the number of electrons
dNe/dE) having lost an energy E) through core excitation in the gas in given by

(4.6)

Here we have assumed that the sample density is kept sufficiently low that
multiple scattering events are negligible [4.35]. As discussed in detail by Inokuti
[4.25] who elucidated the physical concepts of the Bethe theory, the ISEELS
cross section per unit energy loss in the case of forward scattering (observation
along the incident beam direction with vanishing angular spectrometer accept-
ance) is even quantitatively related to the X-ray absorption cross section Ux' In
this case, corresponding to Q = 0, one obtains [4.25]

dUe -3
-d = BEo(E) uX (4.7)
E)
,

where the constant B is given by B = 8mc 2 (e 2 /hc). Note that the quantity in
parentheses is the dimensionless fine structure constant. Because of the finite
angular acceptance of ISEELS spectrometers, in practice, a range of Q vectors is
measured and the cross section dUe/dE) and X-ray absorption cross section U x
differ by a factor (E)-n with 2 ::; n ::; 3. Procedures for converting ISEELS
spectra recorded at finite spectrometer acceptance angles into absolute X-ray
absorption cross sections ha~e been discussed in the literature [4.26, 27] and
used to derive absolute transition intensities for free molecules [4.36, 37].
Equation (4.7) tells us that ISEELS is best suited to study excitations of
relatively low energy since its cross section drops off with the third power of E).
In practice, this limits ISEELS to the study of inner shells with binding energies
in the range ~ 1000eV.
Modern ISEELS spectroscopy of atoms and molecules was pioneered in the
early 1970s by Van der Wiel [4.38] and Brion [4.39] who showed that it was
possible to carry out measurements close to the optical limit by using sufficiently
 4.2 Characteristic Resonances in K-Shell Spectra 83

high electron beam energies and a forward scattering geometry. It is interesting

to note that ISEELS can also be applied to solids using thin film samples and
beam energies of l00keV or more in order to avoid multiple inelastic scattering
events in the sample [4.40,41]. A comprehensive review ofISEELS of solids has
been given by Egerton [4.41]. Surface ISEELS studies, which are typically
carried out in reflection geometry and with primary beam energies of a few ke V,
have been reviewed by De Crescenzi [4.42]. Such studies have investigated clean
surfaces [4.43] and metallic overlayers [4.44], as well as atomic [4.45] and
molecular [4.33, 34] adsorbates.
The instrumentation for ISEELS of free atoms and molecules has been
discussed by several authors [4.1, 20, 46-48]. Spectra are typically recorded with
electron energies Eo ~ 3 keY and a ±0.5° angular acceptance of the electron
energy loss analyzer around the direction of the incident beam. This corres-
ponds to a maximum momentum transfer of Q = 2.4 A- 1 at the oxygen K -edge
(530eV). Using the oxygen (Z = 8) K-shell diameter r ~ 2ao/Z = 0.13 A, where
ao = 0.529 A is the Bohr radius, the product Qr = 0.32 A indicates that the
dipole term in (4.2) is only a factor 6 larger than the quadrupole term.
Experimental studies, however, show excellent agreement with X-ray absorp-
tion, indicating that the dipole approximation holds better than expected. A
possible explanation is the relative weakness of the quadrupole matrix elements
[4.49]. In a few cases quadrupole transitions have been observed, the most
dramatic case being for PCl 3 [4.50].
We shall go on to discuss K-shell spectra offree molecules obtained with the
ISEELS technique and in a few cases with X-ray absorption and compare them
to spectra calculated with the XIX-MS and/or ST-MO techniques. The purpose
of this chapter is the understanding of the major characteristics of molecular K-
shell spectra. The basis for this understanding is provided by the spectra of
simple diatomics, pseudo-diatomics or other small molecules and we shall
therefore restrict our discussion here to those systems. Spectra of larger mole-
cules are discussed in Chap. 6 in conjunction with those of macromolecules and
polymers.

4.2 Characteristic Resonances in K-Shell Spectra

4.2.1 Overview
The K -shell spectra of atoms and molecules contain a variety of pronounced
resonances whic,h correspond to electronic transitions of a K -shell electron to
states near the vacuum level and therefore fall at an excitation energy close to
the IP. The purpose of this section is to survey the different kind of resonances
typically encountered in molecular K-shell spectra. To simplify matters we only
consider "simple" molecules, either diatomics or pseudo-diatomics (ignoring
hydrogens), or larger molecules where there is little interaction between adjacent
84 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

bonds. Thus we ignore all molecules where first order bond-bond interactions
(Sects. 3.4 and 3.5) between second row atoms are important, as in systems with
n and/or a conjugation. Also, second order bond-bond interactions must be
weak. Therefore we do not consider cases where the antibonding orbitals
associated with different neighbor bonds are similar in energy, e.g., C-C and
C-O a* or C = C and C = 0 n* orbitals, respectively. This selection allows us to
lay the foundation for the understanding of the spectra of complex free mole-
cules, as well as chemisorbed or polymeric molecules, which are discussed in
terms of the spectra of diatomic building blocks in Chap. 6.
The effective molecular potential and different types of final states for an
atom and a diatomic molecule respectively are depicted in Fig. 4.2. In construct-
ing this figure we have drawn from the theory discussed in Chap. 2. The plotted
potential for the diatomic molecule differs in shape from the muffin-tin potential
shown in Fig. 2.7 in that it assumes positive values at the periphery of the
molecule, i.e., contains a potential barrier. The barrier is due to an additional
centrifugal term 1(1 + l)h 2/(2mr2) in the molecular (or atomic) potential which
arises when the Schrodinger equation is written in spherical coordinates. Thus
the radial part of the wave equation written in the center-of-mass coordinate
system of the molecule contains an "effective potential" which is the sum of the
Coulomb and exchange potentials (2.58) and a centrifugal potential. The poten-
tial barrier term depends on angular momentum 1and for K-shell excitations of
atoms is small because the excited photoelectron has angular momentum 1 = 1.
For K-shell excitation of molecules, however, the final state wavefunction is only
of 1 = 1 nature in the coordinate system of the excited atom. In the center-of-
mass molecular coordinate system all I-values are allowed (Sect. 2.7.4 and
Appendix A) and therefore the potential barrier may be appreciable.
For a closed shell atom in its ground state, say a noble gas atom, the
Schrodinger equation predicts empty Rydberg states just below the vacuum
level (Ev) and a continuum of empty states above it. The same picture applies for
the core excited atom; the energies of the electronic states are however shifted.
Hence we expect the K-shell excitation spectrum to resemble that shown at the
top left of Fig. 4.2. Here we have simply assumed that transitions of the Is core
electron to all empty states are allowed and ignored all multi-electron excita-
tions. For a diatomic molecule we also have Rydberg states below, and con-
tinuum states above, Ev' In addition, there are unfilled or empty molecular
orbitals. As discussed in Sect. 3.1, it is convenient to simply label the MOs in
terms of a or n symmetry and denote unfilled MOs- by an asterisk. As illustrated
by Fig. 2.5, the lowest unoccupied MO (LUMO) for a diatomic is usually a n*
orbital with a a* orbital at higher energy. For the neutral molecule these states
typically lie above the vacuum level but the n* state is pulled below Ev by the
electron-hole Coulomb interaction (Fig. 2.4). This yields the schematic effective
potential and energy level diagram shown at the bottom right of Fig. 4.2 and the
K-shell spectrum shown above it. Let us now compare these simple concepts to
experimentally observed spectra.
 4.2 Characteristic Resonances in K-Shell Spectra 85

Rydberg tr-

IP Photon Energy IP Photon Energy

Rydberg
States

y X
Fig. 4.2. Schematic potentials (bottom) and K-shell spectra (top) of atoms and diatomic molecules.
The effective potentials include the centrifugal term [ oc 1(1 + 1)/r 2 ] , which is small for K-shell
excitation of an atom (I = I) but may be large for K-shell excitation ofa molecule because relative to
the center of the molecule large I-values may dominate in the wave function of the excited electron.
For the diatomic, the centrifugal term gives rise to the indicated barrier, which separates an inner
well near the atomic positions from a shallower outer well. The inner well contains all core and
valence atomic or molecular orbitals. The wavefunctions of the lowest energy Rydberg orbitals may
also be largely confined to the inner well but lie almost entirely in the outer well for the higher
members of the series, i.e., those close to the vacuum level. Unfilled MOs are indicated by an asterisk
label (e.g. 0'*). Resonances in K-shell spectra arise from electronic transitions from a Is initial state to
Rydberg or unfilled-MO final states. At the IP, corresponding to the threshold for transitions to
continuum states, a step-like increase in X-ray absorption is expected. These effects lead to the
characteristic features of atomic and molecular K-shell spectra schematically shown in the upper
part of the figure. In addition to these "one-electron" features other structures arising from "multi-
electron" transitions may be observed

We have already encountered the K-shell spectrum of the simplest case,

namely a noble gas atom, in Fig. 2.1. The spectra are characterized by a few
sharp resonances approximately 2-6 eV below the IP followed by a step-like
increase in absorption around 2 eV below the IP. To illustrate how this "atomic"
spectrum develops into a "molecular" one we show in Fig. 4.3 the K-shell
86 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

:r···"3f·······,····· ··~;···l
;~t '''2 NH, 1
i~I N2
H
.j
~1~1 -10 o 10 20
Energy from IP (eV)

Fig. 4.3. ISEELS K-shell spectra of neon (Ne) [4.51. 52]. ammonia (NH3) [4.53]. hydrazine (N 2H 4)
[4.51]. and nitrogen (N 2) [4.51. 54]. All spectra are aligned at the Is IPs [4.55]. The figure illustrates
the increasing number of resonances introduced by additional bonds: The Ne spectrum exhibits a
pronounced Rydberg resonance", 3eV below the IP. NH3 exhibits two pronounced bound state
resonances which are stronger relative to the continuum step than in Ne because of N-H valence
orbital admixtures to the Rydberg states. Note that Ne is the united atom ofNH 3. N2H4 exhibits an
additional N-N single bond (1* resonance close to the IP. Triple bonded N2 shows a pronounced n*
bound state resonance and a (1* shape resonance in the continuum around geV above the IP. The
structure around 5eV above the IP in N2 is a multi-electron resonance

spectrum of Ne [4.51, 52] and in comparison that of isoelectronic NH3 [4.51,

53]. The spectra look surprisingly similar, the main difference being the greater
intensity of the discrete transitions. This can be attributed to the fact that the
wavefunctions of the final states in NH3 are no longer pure Rydberg orbitals but
contain an admixture of hydrogen-derived valence orbitals [4.56, 57].
In contrast, the K-shell spectrum of the hydrazine (N 2H 4) molecule [4.51]
(Fig. 4.3) exhibits a new resonance around 1 eV below the IP. In hydrazine two
NH2 groups are bonded together by a N-N single bond and the additional
resonance arises from a transition to the N-N antibonding u* orbital, which is
unoccupied. Also note that the vertical scale in Fig. 4.3 is in absolute oscillator
strength units and the overall absorption step or edge jump is a factor of 2 larger
in N2H4 than in NH 3. This is the result ofthe fsum rule (2.15), which links the
oscillator strength to the number of electrons in the atom or molecule. More
precisely, it arises from the approximate fsum rule for the K-subshell as
discussed in Sect. 2.1.2. It also appears that the N-N u* resonance takes away
oscillator strength from the N-H valence/Rydberg resonances since these are
 4.2 Characteristic Resonances in K-Shell Spectra 87

less pronounced relative to the continuum cross-section than for NH 3 . This,

again, is a consequence of the approximate fsum rule for the K-shell, stating
that the integrated oscillator strength is a constant.
Finally, at the bottom of Fig. 4.3 we show the spectrum of the nitrogen
molecule [4.51, 54] which is linked by two N- N n bonds and one a bond, i.e., a
triple bond [4.58]. Indeed, this bonding situation is directly reflected by its K-
shell spectrum. The lowest energy resonance around 9 eV below the IP arises
from a transition of the is electron to the N-N anti bonding in: orbital, and the
highest energy resonance around 9 eV above the IP is due to a transition to the
N-N a:
antibonding orbital [4.54]. The weak structures below the IP are
Rydberg resonances which, in general, can be identified by their characteristic
term values, i.e., their separation from the IP [4.56]. The structure near 5eV
above the IP is a multi-electron excitation feature [4.59- 63]. Note that without
this structure the N 2 spectrum closely resembles the schematic spectrum in
Fig. 4.2.
The origin of the molecular resonances shown in Fig. 4.3 is confirmed by the
results of Xa-MS calculations by Wurth and Stohr, shown in Fig. 4.4 [4.64].
These calculations were performed using the transition state method, choosing
the atomic sphere radii to match calculated and experimental IPs. In Fig. 4.4 we
have plotted both experiment and theory in absolute cross section units of Mb
(= 10- 18 cm 2). The overall agreement between theory and experiment is re-
markably good. Except for the multi-electron peak in the N2 spectrum at 5 eV

Experiment Theory

4 (a) 5 2e' (b)

N- H' 4
3
2 3
_ 2
1 NH3 .c
D NH3
~ 1

~~I
j~ N, H,
1
!1!
U 0

i
4
3 o·
2
xV. I
1
0
- 10 0 10 20 - 10 0 10 20
Energy from IP (eV) Energy from IP (eV)

'''9
Fig. 4.4. Comparison of the molecular K-shell spectra of NH 3 , N 2H 4 , and N2 taken from Fig. 4.3
to those calculated by means of the Xa-MS technique [4.64]. Note that the experimental intensities
of Fig. 4.3 have been converted to cross section units of Mb, according to (2.12). Between the spectra
we have depicted those molecular orbitals, taken from [4.58], which are associated with specific
resonances in the spectra. The orbitals and corresponding resonances in the calculated spectra are
labeled by the irreducible representations of the MOs
88 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

above the IP, all experimental features are accounted for. This indicates that
such one-electron calculations can be used to assign all major resonances and
that features not reproduced by the calculations are most likely of multi-electron
character. To illustrate the origin of specific resonances we have also shown in
Fig. 4.4 orbital contours of unfilled MOs associated with the resonances. The
pictures are from [4.58] and are labeled with their irreducible representations.
The same labels are used for the resonances in the calculated spectra.
Before we discuss the main resonances encountered in K -shell excitation
spectra in more detail, it should be pointed out that similar resonance features
are also observed in the excitation spectra of the valence shells [4.65, 66].
However, the particularly simple symmetry and the localized nature of the Is
initial state greatly favors the study of the K -shell. As pointed out by Dehmer et
al. [4.67], valence shell spectra are complicated by the delocalized nature of the
valence-hole state leading to more complicated electron correlation and screen-
ing effects and strong particle-hole interactions, by a greater energy dependence
of the dipole matrix element, and by strong continuum--continuum coupling
between nearly degenerate ionization channels. These complications obscure the
intensities and energy positions of the resonances. In contrast, K-shell spectra
are less complicated and their interpretation can be guided by simple, general
rules, to be discussed.

4.2.2 7l'* Resonances

In the spectrum ofN 2 (Fig. 4.3) several sharp peaks are observed below the Is IP
(Fig. 4.2). The most pronounced resonance by far is the lowest energy peak,
which corresponds to a transition to a n* anti bonding orbital of the molecule. In
the following we shall refer to such commonly observed Is ~ n* transitions as
n* resonances. In all known K-shell spectra of low-Z molecules the n* reson-
ance, if present, is the lowest energy structure and its energy position falls below
the Is IP. In the ground state the final state valence orbital associated with such
transitions can be empty (e.g., CO) or partially filled (e.g., NO). For a molecule in
its ground state empty states lie above, and partly filled states below, the vacuum
level. In the core excited molecule there is an upward (smaller binding energy)
relaxation shift of the Is orbital and a downward shift of the outer "optical"
orbitals because of the increased attractive Coulomb potential created by the
core hole [4.68]. The Coulombic shift of the outer orbital results in a transition
energy of the n* resonance which is less than the Is IP. In fact, CO with a C Is
core hole and N2 with a N Is hole and an electron in the n* orbital resemble NO
in the equivalent core approximation and it is therefore not surprizing that the
n* resonances for N 2 and CO lie below the IP by nearly the same amount
(~9 eV) as the partially filled n* orbital for neutral NO lies below the vacuum
level, as discussed in Sect. 2.4.2.
A n* resonance is only observed for molecules with n bonding. For example,
experimental and calculated [4.64] spectra for molecules with C-O bonds
shown in Fig. 4.5 exhibit a n* [Link] for triple and double C-O bonds as in
 4.2 Characteristic Resonances in K-Shell Spectra 89

Experiment

J J
C K-edge OK-edge

(a) (c)
u· u·
CO
x- x3

1
~
c:
'0 I
::l IP
IP
€
.!!!.
~
on
c: [Link]
~

[Link]

310 320 530 540 550 560

Energy Loss (eV)

Theory
C K·edge o K·edge

J .~
(b) >T" (d)
0"

i
0"

xV. )(V2 :
CO , I

iii
.'c:" /
I
I
I
>T*
I - I

::l >T*

P
I
~
c:
I
!!
I [Link]
.!:

[Link]
280 290 300 310 320 530 540 550 560
Excitation Energy (eV)

Fig. 4.5. Top: Experimental ISEELS spectra above the C and OK-edges for carbon monoxide
(C=O) [4.51, 54], formaldehyde (H 2 C=O) [4.51, 69], and methanol (H3C-OH) [4.37, 51, 53].
Bottom: Corresponding K -shell spectra calculated by means of the Xa-MS technique [4.64] for the
same molecules. The Is IPs determined by photoemission (top) [4.55] and calculated (bottom) are
indicated by vertical lines. Note that no 71:* resonance exists for single-bonded CH 30H. The strong
energy dependence of the (1* resonance can be correlated with the change in C-O bond length:
1.128..\ for CO, 1.209..\ for H 2 CO, and 1.425..\ for CH 30H [4.70]

C=O [4.51, 54] and H 2 C=O (formaldehyde) [4.51, 69] but not for single
bonded H3C-OH (methanol) [4.37, 51, 53]. The natural width (in the absence of
instrumental broadening) of the resonances is determined by the lifetime of the
excited state, see (2.16), and the vibrational motion of the molecule. In the n*
resonance, the final state lifetime of the free molecule is determined by the decay
of the core hole, which predominantly proceeds via autoionization, i.e., deexcit-
ation of the excited electron in the n* state to the K-shell and simultaneous
90 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

emission of a valence electron [4.71]. The n* resonance lifetime width for N2

and for the C K -edge in CO is extremely small (::: 0.1 eV) and this allows
observation of the vibrational energy levels of the electronic final state with
intensities determined by the appropriate Franck-Condon factors (Fig. 4.1)
[4.72]. Within a series of molecules containing the same atoms, the intensity of
the n* resonance changes with the bond order between the atoms. This is
illustrated by Fig. 4.5, where the intensity of the n* resonance relative to the
absorption continuum changes between double bonded H 2C = 0 and triple
bonded C == o. Quantitative measurements [4.27, 73] show a decrease in the n*
intensity by a factor of 2.3 at the C K-edge and an increase by a factor of 1.3 at
the 0 K-edge between CO and H 2CO, respectively. This indicates that at the C
K -edge the n* resonance loses oscillator strength to the resonances associated
with C-H bonds (see below).

4.2.3 Rydberg and Mixed Valence/Rydberg Resonances

Between the n* resonance and the IP several sharp but weak resonances can be
observed (Figs. 4.3, 5) which correspond to excitations to Rydberg orbitals or, in
the presence of bonds to hydrogen atoms, to a mixture of Rydberg and
hydrogen-derived antibonding orbitals of the same symmetry [4.56, 74]. The
resonances merge into a continuous step-like feature about 2 eV below the IP as
expected from a Rydberg series of states. Figure 4.5 shows that the intensity of
these sharp resonances relative to the continuum absorption at the highest
energy is weakest for CO and increases with the addition of hydrogen atoms in
H 2C = 0 and H3C-OH. Furthermore, the intensity of these resonances is
stronger for the C K-edge excitation than for the 0 K-edge excitation. Rydberg
features are generally weaker in the higher energy edges because of the decreased
size of the Is orbital and hence the smaller spatial overlap with the Rydberg
orbital. In H 2CO, another reason for the reduction of the corresponding
resonance in the 0 K -edge relative to the C K -edge spectra stems from the
presence of C-H valence orbitals and the absence of O-H valence orbitals. The
weak intensity of the peaks for CO is associated with the large spatial extent of
Rydberg orbitals, which have most of their orbital density at the periphery of the
molecule, in the outer well (Fig. 4.2). Mixing of the Rydberg orbitals with
hydrogen-derived valence orbitals increases the intensity of the corresponding
resonance because of the more compact nature of the valence orbitals [4.75, 76].
The correlation between Rydberg and hydrogen-derived molecular orbitals
can be well illustrated for the simple molecules methane (CH 4 ), ammonia (NH 3),
water (H 20) and hydrogen fluoride (HF) whose K-shell excitation spectra [4.53,
77] are shown in Fig. 4.6. The spectra are quite similar and exhibit structures
primarily below the respective Is IPs, as expected. 1 We have denoted the two

1 Previously, weak structures in the continuum, in the range 10-20eV above the IP, have been
assigned in [4.79] to u* shape resonances associated with hydrogen bonds. We now believe that this
assignment is probably incorrect and that the structures more likely arise from shake-up continua.
 4.2 Characteristic Resonances in K-Shell Spectra 91

40"*

-10 0 10
Energy from IP (eV)
Fig. 4.6. ISEELS spectra of the isoelectronic molecules methane (CH 4 ), ammonia (NH 3 ), water
(H 2 0) [4.53] and hydrogen fluoride (HF) [4.77). In all cases the IP has been subtracted from the
measured electron loss energy, in order to compare the spectra on the same energy scale. Resonances
A and B correspond to transitions to the hydrogen-derived anti bonding molecular orbitals shown
on the right taken from [4.58). Peaks A correspond to transitions to a!, and for HF to the 417*, final
states, and peaks B are associated with (from top to bottom) the 2t1, 2e*, and 2b! MOs. These
orbitals are strongly mixed with Rydberg orbitals of the same symmetry but for K -shell excitations
the valence character of the final state largely determines the resonance intensity [4.78]

most prominent resonances as A and B. Originally these resonances were

attributed by Wight and Brion mainly to Rydberg states associated with 38 and
3p atomic final states [4.53], although it was pointed out that the 3p state for
NH3 was known to have appreciable anti bonding character. Later, in their
analysis of the HF spectrum, Hitchcock and Brion [4.77] pointed out that the
mixing of Rydberg and hydrogen-derived valence-type orbitals is a more general
phenomenon, and specifically associated the lowest energy resonance in HF
with the 4u* H-F anti bonding orbital (Fig. 4.6) [4.58]. In a direct comparison of
92 4. Experimental and Calculated K-SheJl Spectra of Simple Free Molecules

K-shell spectra for CH 4 , NH 3, and H 2 0 with that of the united atom Ne

(Fig. 4.3), Akimov et al. [4.57] attributed the shift of resonances A and B in
Fig. 4.6 to valence (electronegativity) effects and concluded that the states must
be of mixed valence/Rydberg type. This was later supported by ab initio
calculations by Chang et al. [4.80].
Recently it has become apparent from the detailed inspection of K-shell
spectra of cyclic [4.78] and linear [4.81, 82] hydrocarbons that many "Rydberg"
resonances observed over the years in hydrocarbons and molecules with C-H
bonds are in fact better described as mixed Rydberg/valence resonances, since
for K-shell excitations the peak intensities are often determined by the valence-
like rather than the Rydberg-like nature of the final state. Such findings lead to
the correspondence between resonances A and B in Fig. 4.6 with the hydrogen-
derived anti bonding orbitals discussed in Sect. 3.2.2 (compare Fig. 3.3) and
shown on the right side of the figure. Resonances A correspond to the valence
orbitals with a 1 symmetry mixed with the 3s a 1 Rydberg orbitals, or in the case
of HF to the 4<1* orbital. For CH 4 the Is (lad -+ 3ar transition is symmetry
forbidden and becomes allowed only through Herzberg-Teller vibronic coup-
ling [4.83]. Resonances B are associated with the 2t 2, 2e, and 2b 1 antibonding
orbitals in CH 4 , NH3 and H 2 0, which are mixed with the 3pt2' 3pe, and 3pb 1
Rydberg states, respectively. Note that for the core excited final state the empty
MOs are pulled below Ev by the effective Z + 1 Coulomb potential of the
equivalent core and can therefore mix more extensively with the Rydberg
orbitals. Thus, to a large degree the mixing is a final state effect. In addition, the
character of the unfilled orbital sampled is dependent on the spectroscopic
technique. For K-shell excitation, it is the small-R region of the Is core which
contributes the most. As noted by Mulliken [4.84], a MO is more valence
at small R and more Rydberg at large R, an effect which has been called
"Rydbergization".
More evidence for the valence nature of the final states associated with the
resonances in Fig. 4.6 comes from H + yield spectra of condensed and chemi-
sorbed water [4.85]. Here the enhanced H + yield at the first resonance position
(peak A) and its angular dependence were taken as direct evidence for the
valence character of the associated state. The correspondence between the
resonances and hydrogen-derived valence orbitals in Fig. 4.6 has important
consequences, since for chemisorbed molecules resonances A and B should
exhibit a different polarization dependence (with the exception of CH 4 ) and may
therefore be used to determine the orientation of the molecules on the surface 2 •
The survival of the Rydberg/valence resonances upon chemisorption may
yield interesting information regarding the strength of the chemisorption bond
for chemisorbed molecules. For strongly chemisorbed molecules one would

2 NEXAFS spectra of chemisorbed NH3 on Ni(J 10), see [Ref. 4.86, Fig. 3], reveal resonances which
were attributed to Rydberg transitions. The observed resonances are better described as transitions
to 4a l u* and the 2e n* N-H valence orbitals
 4.2 Characteristic Resonances in K-Shell Spectra 93

expect pure Rydberg resonances to be quenched upon chemisorption. In con-

trast, hydrogen-derived resonances are expected to survive the chemisorption
process because of their more localized nature, similar to n* and 0'* resonances.
Finally, C-H bond related resonances play an important role in the K-shell
spectra of hydrocarbons (Sect. 4.2.7) and especially in the spectra of polymeric
hydrocarbon chains as discussed in Chap. 6.

4.2.4 0'* Shape Resonances

Above the Is IP, the K-shell spectrum ofN2 and the C K-edge in CO exhibit two
resonances of which only the broader, higher energy one is observed for the 0
K -edge for CO (Fig. 4.5). The lower energy structure has been assigned to multi-
electron excitations associated with the prominent Is -+ n* transition [4.60-63,
87], while the highest energy peak is a so-called 0'* shape resonance [4.59, 88].
Shape resonances are one of the intriguing phenomena associated with molecu-
lar valence and core excitation spectra and the name arises from the idea [4.89,
90] that the excited state is stabilized against immediate decay by a hump in the
potential shape as indicated in Fig. 4.2. In all documented cases the potential
barrier arises from the centrifugal part of the potential and its shape and height
depend on the angular momentum of the final electron state [4.91]. Therefore
shape resonances are associated with particular I components of the electronic
wavefunction and these components are trapped only in certain directions, e.g.,
along the internuclear axis for the 0'* shape resonance.
Recently Natoli [4.92] has questioned the necessity of a potential barrier for
the existence of a continuum resonance and suggested that they should be
viewed more generally as arising from multiple scattering processes of the
photoelectron by the atomic centers. In the potential barrier model, the n*
resonance might be viewed as a "shape resonance below threshold" where the
wavefunction is trapped by an impermeable barrier perpendicular to the inter-
nuclear axis. However, this terminology has been criticized by Schwarz et al.
[4.93], who argue that resonances below and above the IP are of different nature
and suggest the name Feshbach resonances for the structures below the IP. This
terminology is based on that traditionally used in the area oflow energy electron
scattering, as reviewed by Allan [4.66].
Rather than using a scattering picture for the description of the highest
energy resonance in the N 2 spectrum in Fig. 4.3, the resonance can also be
explained in a molecular orbital picture. As discussed in detail by Sheehy et al.
[4.94], the molecular orbital picture naturally yields states of enhanced ampli-
tude on the molecule which lie above the vacuum level without specifically
employing a potential barrier concept. In this picture the 0'* shape resonance
can be described as a two step one-electron process where the K-shell electron is
first excited to a virtual molecular orbital of 0' symmetry, followed by the
emission of a photoelectron [4.95-98]. Examples of such M 0 final states are the
30': and 60'* MOs for N2 and CO shown in Fig. 2.5 or the 60'* MO for NO
pictured in Fig. 2.8. Because of the insignificance of the potential barrier concept
94 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

and therefore the "shape" of the potential for the existence of above-threshold
resonances in the MO description, we shall below refer to u* shape resonances
also simply as u* resonances.
The energy width of the u* resonance is related to the lifetime of the quasi-
bound electronic state. Because of the increasing decay probability of the
electron to continuum states, 0'* resonances become broader the higher they lie
in the continuum. Their asymmetric line shape with a tailing toward higher
energy is readily understood in the potential barrier picture emerging from XIX-
MS calculations. Since the probability for tunnelling through the barrier in-
creases with the kinetic energy of the photoelectron, the resonance shape is more
peaked at lower energy with a high energy tail. In another picture the asym-
metry can be understood in analogy to the "giant resonance" line shape [4.99],
which arises quite generally from the interaction of an electron with a potential
well. In particular, the giant resonance line shape is not only asymmetric but
shows dispersion on the high energy side such that it resembles an asymmetric
Gaussian profile with an underlying step function. The line shapes of the near
edge resonances are discussed in more detail in Chap. 7.
Furthermore, 0'* resonances are asymmetrically broadened by the vibra-
tional motion of atoms in the molecule [4.100]. This is shown in Fig. 4.7 for the
0'* resonance of the N2 molecule observed in the excitation of the 30'g inner
valence shell. For variation in intranuclear distance spanning the range of
ground state vibrational amplitude from 1.23 A to 0.965 A, the resonance
position shifts from about 8 eV to 24 eV. Since the electronic excitation is fast
compared to the nuclear motion the X-ray absorption spectrum is a weighted

20 ::_ R = 1.230A
:0-
:::?!
~15
b
c:
o
~ 10
Q)
(J)

~ 5
e
()
t"'~iriiIii~;;:

o~----~----~----~----~
o 10 20 30 40
hv-IP (eV)
Fig. 4.7. Cross section for photoionization of the 30'g valence level ofN2 (binding energy - 16eV)
in the excitation region of the 0'. shape resonance [4.100]. Calculations were carried out as a
function of the internuclear molecular distance R over a range spanning the ground state vibrational
amplitude. The position, strength and width are found to be strong functions of R. The solid curve
corresponds to a summation over the vibrational motions in the ground state and its shape exhibits
an asymmetry which distinguishes it from the resonance shape calculated for R fixed at the
eqUilibrium value 1.095 A (central dashed curve)
 4.2 Characteristic Resonances in K-Shell Spectra 95

average over the distances allowed in the ground state of the molecule. The R
dependence of the resonance position arises from the fact that 0'* orbitals are
directed along the internuclear axis between two atoms, as discussed below, and
it is therefore absent for n* resonances. The distance dependence of shape
resonances has been elegantly measured by vibrationally resolved partial photo-
ionization cross sections of the final ion. Examples are discussed in the reviews
by Dehmer et al. [4.67] and Nenner and Beswick [4.101].
Figure 4.7 shows that the energy position of the 0'* resonance is very
sensitive to the internuclear distance, a fact which has been discussed extensively
[4.79,92,94, 102, 103]; see Fig. 4.5. Systematic inspection of the 0'* resonance
position for different bonds reveals that it is a function not just of the bond
length but also of the detailed molecular potential. This dependence can be
approximated by consideration of the atomic numbers Z of the bonded atoms
[4.79]. For most low-Z molecules the 0'* resonance lies in the continuum above
the IP, but if the sum of atomic numbers of the two bonded atoms exceeds
Z = 15, it falls below the IP such as in O 2 [4.79, 104]. Measurement of ion
angular distributions is an elegant way to obtain final state symmetry informa-
tion [4.105] and this technique has recently been used to confirm the assignment
of the position of the 0'* resonance in O 2 [4.106]. The important correlation
between the 0'* resonance position and bond length is discussed in detail in
Chap. 8. The intensity of the 0'* resonance for a given bond length increases with
Z, being weakest for C-C bonds and strongest for F-F bonds.

4.2.5 Multi-Electron Features

Multi-electron effects have already been discussed in Sect. 2.5, and we saw that,
in principle, all electronic excitations are multi-electron in nature. The reason for
using a one-electron model, at all, stems from its simplicity and success in
explaining photoemission spectra that are dominated by peaks originating from
the "active" electron. However, closer inspection of photoemission as well as
absorption spectra reveal satellite peaks which can be associated with the
indirect excitation of "passive" electrons. The sudden creation of the core hole
potential induced by the absorption of an X-ray photon by the "active" electron
may knock one or more of the "passive" electrons into excited states.
In a simple picture, two types of effects can be distinguished. When the
excited state is a bound state, the process is called a shake-up, and when it is a
continuum state the process is referred to as a shake-off. In both processes the
total excitation energy is equal to the photon energy and in photoemission
spectra the satellites are therefore observed at lower kinetic energy than the
main active-electron peak. In X-ray absorption or ISEELS, multi-electron
features related to a specific primary transition are observed at higher photon
energy than the primary resonance. In general, shake-up structures are peak-like
while shake-off structures are step-like or just a smooth background.
Probably the best known multi-electron effects in K -shell spectra are those
in argon, first observed in 1963 [4.107] and extensively studied ever since by
96 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

Fig. 4.8- K-shell X-ray absorption

Argon spectrum of argon near threshold
K-edge (compare Fig. 2.1), recorded by
4 Lytle et al. [4.111]. The energy re-
(jj" gion containing various shake-up
.~
c: structures is shown enlarged under-
::l
neath and is plotted on an energy
-e 3 scale relative to the IP. Most of the
~ satellite peaks can be attributed to
1:)
m
;;::
two-electron excitations of Is and
3p electrons [4.109]
c: 2
e
uQ)

iii

3200 3220 3240 3260

Photon Energy (eV)

means of X-ray absorption [4.108, 109] and photoemission [4.110], spectro-

scopies. The Ar K-shell spectrum recorded by Lytle [4.111] is shown in Fig. 4.8
and exhibits large multielectron structures in the range 15-50eV above thres-
hold. Similar structures are also observed in the K-shell spectrum of neon
[4.112]. These shake-up structures arise from two-electron transitions involving
the 1s and 3p electrons with binding energies of [Link] and 15.8 eV, respect-
ively. As expected, the satellites are separated from the 1s IP by more than the 3p
binding energy because of the additional Coulomb and exchange terms associ-
ated with the two-hole final state [4.113] and correlation effects. In fact, the
main satellite peak is separated from the 1s IP by approximately the 3p binding
energy (18.3 eV) of the equivalent core atom, potassium. The various shake-up
peaks are associated with different electronic final state configurations, of which
the 1s3p 53d2 and 1s3p54p2 configurations give rise to the largest transition
strengths [4.109]. It has been pointed out that, in particular, the main two-
electron resonance at 3225 eV is associated with the 1s3p 53d 2 configuration and
that it can be reached via 1s-3p double excitation from a correlated ground state,
consisting of the pure 1S23p6 ground state mixed with double-excited 1S23p4
nln'l configurations.
Examples of shake-up structures in molecular K -shell spectra can be found in
the N 2 K -shell spectrum (Fig. 4.3), and the C K -shell spectrum of CO (Fig. 4.5).
In both cases a pronounced peak-like structure exists about 5 eV above the IP,
just below the u* shape resonance. High resolution spectra (Fig. 7.1) reveal
considerable fine structure in this resonance [4.21,54, 114], owing to its origin
as a double excitation feature. For example, in CO the double excitation
involves the final state configurations (1s) -1 (br) - 1(2n*)1 (Ryd.)1 and
(1s) -1 (5u) -1 (2n*)1 (Ryd.)1 resulting from transitions of the "active" 1s electron
 4.2 Characteristic Resonances in K-Shell Spectra 97

and a In or 5u "passive" valence electron to the 2n* and Rydberg orbitals,

respectively [4.21, 60-63, 87]. Careful inspection of the 0 K-edge spectra of CO
and H 2 CO in Fig. 4.5 reveals other weak structures in the continuum above the
u* shape resonances which also appear to be multi-electron in character.
Multielectron excitations involving double-core-vacancy excited states have
also been observed in molecular gas phase spectra [4.113, 115] and solids
[4.116]. Such features may lie several hundred eV above the IP and therefore
interfere with the EXAFS structure. Shake-up peaks can also be observed in the
near edge spectra of solids, as discussed by Stern [4.117].
To develop a better understanding of the origin of various resonances we
present below a comparison of selected experimental and theoretical results for
K-shell excitation spectra of simple molecules.

4.2.6 Correlation of Multiple Scattering and Molecular Orbital Calculations:

The N 2 Molecule
The N 2 molecule has served a special role over the years. It was not only the first
test case for the quantitative calculation of molecular K-shell spectra (for a
review see [4.67]) but it has also been thoroughly explored experimentally and
theoretically, deepening our understanding of the various phenomena associ-
ated with photoionization [4.67,91]. As a consequence, its K-shell spectrum has
been thoroughly studied experimentally [4.8, 16, 17,21,54, 71, 113, 118] and
theoretically [4.59, 60, 88, 95, 97, 100, 119-121] and provides a suitable example
as to the present extent and limitations of our understanding.
The experimental X-ray absorption cross section for N2 measured by
Bianconi et al. [4.118] is compared in Fig. 4.9 to the results calculated by Dehmer
and Dill with the Xoc-MS method [4.88] and by Rescigno and Langhoffwith the
MO St'ieltjes-Tchebycheff [4.95] technique. The MS and MO calculations give!
numbers for the n* resonance of 0.23 and 0.26, respectively. Since the total
(discrete and continuum) oscillator strength for the K-shell of the low-Z atoms is
approximately equal to 2, (2.15), the number of K-shell electrons [4.122], the n*
resonance intensity is about 10-15% of the total K-shell absorption intensity.
The corresponding n* resonance cross sections shown in Fig. 4.9, obtained by
means of (2.14) with a peak width l/(lb = 1 eV, are in good agreement with
experiment. The energy position of the n* resonance is 4 eV lower than experi-
ment for the MO calculation and 2eV higher than the measured value for the
MS theory.
The continuum cross sections obtained with the two techniques differ in
magnitude and the u* resonance position. The MO theory yields the continuum
cross section in excellent agreement with theory and the u* resonance is within
5% of the experimental intensity. The MS calculation overestimates the u*
resonance intensity by a factor of 1.6. At first sight it therefore appears that the
MO-ST calculation is superior to the Xoc-MS calculation in reproducing the
absolute transition intensity. However, we need to remember that both calcu-
lations are one-electron calculations and completely ignore multi-electron
98 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

Photon Energy hv (eV) Fig.4.9. (a) X-ray absorption cross section

400 410 420 430 440 450 for N2 measured by Bianconi et al. [4.118].
4 Superimposed on the spectrum as a dash-
A
N2 Experiment (a) dotted step function is twice the calculated
atomic N cross section [4.88]. (b) Calculated
3 xO.2 N2 K-shell excitation spectra by means
B of X<x-MS theory [4.59, 88] and
2 Stieltjes-Tchebycheff MO theory [4.95]. Note
that, although the calculated energy positions
for the n* and (1* resonances are about 3 eV
off the measured positions, the n*--O'* splitting
:c is reproduced within 1 eV by both calcu-
~
)( 0 lations. Also shown is the atomic step function
b
~A
of (a)
c: N2 Theory (b)
.2 5
""
UQ) """
en ""
rn
rn 4 " xO.2 B
0
0
3

2
""
rJ'-I
I
I
I
I '

-.-2 ___
""
"
o ""
-10 o 10 20 30 40
hv - IP (eV)

effects. Because such effects lower resonance intensities, one-electron calcu-

lations should overestimate the continuum resonance intensities, as discussed in
Sect. 2.5. In addition, continuum resonances are expected to be broadened by
vibrational effects as illustrated in Fig. 4.7. With respect to the 0'* position, both
calculations are off by 3 eV; the MO calculation places the resonance too low in
energy, while the MS calculation places it too high by the same amount.
Although the absolute 1t* and 0'* energy positions are incorrect by 2-4 eV, the
n*-O'* splitting is accurately predicted within 1 eV by both calculations.
The dependence of the 0'* resonance position and intensity for N 2 on the
molecular potential, calculated by means of the X/X-MS method, is shown in
Fig. 4.10. As our reference spectrum, we have recalculated the original results by
Dehmer and Dill [4.88], shown in Fig. 4.9. Our results for the continuum cross
section and potential are shown as solid lines in Fig.4.10d and Fig.4.10a,
respectively. The potential was constructed self-consistently with muffin-tin
radii of 0.55 A around the atomic centers, and an outer sphere radius around the
molecular center of 1.10 A, and is identical to that previously shown in Fig. 2.7.
The potentials are plotted along the internuclear axis as a function of the
distance Irl from the center of the molecule. Our reference potential and the
 4.2 Characteristic Resonances in K-Shell Spectra 99

Xa-MS Potentials and Cross Sections for N2

O~~-r __ ~~-r--~~~--r-~

leI "
-20 _ -20
>
~
-40 j:: -40
>
-60 ~c: -60
.,
o
;; -80 <>.
-80 Core
~ Hole
E
>
-100
-2 -1 o 2
~c: Distance from Molecular Center (AI Distance from Molecular Center (AI
0
5
<>. Ibl , ,-
Idl - - - - Touching spheres
/ 1\ tranSition state
-20 \ I
1 II _ . - OverlapPing spheres
II tranSition state
II
-40 II - - - Touchmg spheres
II self-consistent
I I T~uchlng spheres
I I
-60 I \ Non-self-conSistent
I \
I \
I \ r" '-..
-/
I\./
=.........._-...j
/ ' -...~-':.:~'---=-,",.

425 430 435

Distance from Molecular Center (AI Excitation Energy (eV)

Fig. 4.10.-4. Molecular potentials for N2 as a function of the distance Irl from the center of the
molecule, calculated by the Xcx-MS technique and the corresponding continuum cross sections. As a
reference we have chosen the self-consistent touching-sphere potential of Fig. 2.7, which was also
used by Dehmer and Dill [4.88] for their cross section calculation shown in Fig. 4.9. This potential
and the corresponding continuum cross section are shown as solid lines. Note that we have omitted a
vertical line in the solid-line potential at the molecular center (Fig. 2.7), which would indicate the
value ofthe "inner potential", in order to more clearly show it in the other potentials, calculated with
different assumptions. (a) Comparison with a potential (dotted line) calculated non-self-consistently
with the same sphere radii as the solid-line potential, by simple superposition of the atomic XIX
charge densities for nitrogen atoms. (b) Comparison with a potential (dashed line) calculated self-
consistently with the same parameters but using the transition state approximation where half an
electron was removed from the Is shell of one of the nitrogen atoms. (c) Comparison with a potential
(dashed-dotted line) calculated for atomic spheres of radius 0.63 A (overlapping spheres) and an outer
sphere of radius 1.18 A, using the transition state approximation. In this case the sphere radii were
determined by demanding that the calculated ls IP agreed with the experimental value of [Link].
(d) Comparison of the continuum cross section calculated with the potentials in (a-c). The line
patterns of the curves correspond to the ones chosen for the potentials

corresponding continuum cross section are compared to three potentials and

cross sections calculated with different assumptions. The potential shown as a
dotted line in Fig. 4.10a and the corresponding cross section in Fig. 4.l0d were
calculated non-self-consistently with the same sphere radii (touching spheres) as
the solid-line reference potential by simple superposition of the atomic Xa
charge densities for nitrogen atoms. The potential shown as a dashed line in
Fig. 4. lOb and the corresponding cross section were calculated self-consistently
tOO 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

with the same parameters but using the transition state approximation where
half an electron is removed from the Is shell of one of the nitrogen atoms.
Finally, the dashed-dotted potential in Fig.4.10c and the dash-dotted cross
section were calculated for atomic spheres ofradius 0.63 A (overlapping spheres)
and an outer sphere of radius 1.18 A, using the transition state approximation.
In this case the sphere radii were determined by setting the calculated Is IP to
the experimental value of [Link].
Ofthe various potentials shown in Fig. 4.10, the overlapping-spheres trans-
ition state calculation (dash-dotted) yields the a* resonance position (418.5 eV)
in best agreement with the experimental value of 41geV. As discussed by Wurth
and Stohr [4.64], the success of this approach is based on the semi-empirical
choice of the sphere radii to fit the experimental ionization potential. The a*
positions for the other potentials fall at 413.5 eV (dashed), 421.5 eV (solid), and
424eV (dotted), with the results for the touching-spheres transition state (dashed
line) showing the largest discrepancy with experiment.
The theoretical XIX-MS and MO-ST methods discussed in Chap. 2 provide
us with a complementary view as to the origin of the observed K-shell reson-
ances. For example, it is easy to understand in the MS picture why a correlation
exists between the a* resonance position and the bond length. In a scattering
picture the excited photoelectron gets trapped by the molecular potential in
the direction along the internuclear axis. It scatters back and forth between the
absorbing atom and its neighbor and in a simple EXAFS-like picture the
resonance condition is expected to depend on the inter-nuclear distance Rand
the photoelectron wavevector k, according to Rk = const., or Ek = const./R 2 ,
where Ek is the kinetic energy of the photoelectron, i.e., the energy above
threshold in the K-shell spectrum [4.92]. On the other hand, the MO picture is
very helpful for visualizing the origin of the resonances in terms of molecular
orbital pictures of the final states, as plotted for example, by Jorgensen and
Salem [4.58]. The question arises as to a link between the two seemingly
different MS and MO approaches. Such a link can indeed be established, and we
shall discuss it here for N 2 • In particular, we shall focus on the calculations for
the continuum a* shape resonance which is more intriguing than the n* bound-
state resonance. As mentioned above, calculations for bound states can be
carried out with great accuracy by MO theory (for N2 see [4.121]), while it is
more difficult to obtain reliable results for continuum states.
The MS picture views the a* shape resonance as a final ionic state in which
the photoelectron is trapped by the molecular potential in the direction of the
internuclear axis [4.59, 88]. The large oscillator strength of the transition arises
from the large spatial overlap of the core initial state with the compact final
state, trapped in the inner potential well (Fig. 4.2). The effective potential barrier
that limits the motion of the excited photoelectron is a centrifugal barrier [ oc I
(I + 1)/r2] for the 1 = 3 orfwave component of the au ionization channel. This
can be understood as follows. In an atomic picture (region I in Fig. 2.6) the
dipole excitation of a Is initial state leads to a p-like (l = 1) final state which
upon escaping to infinity is scattered by the anisotropic molecular potential into
 4.2 Characteristic Resonances in K-Shell Spectra 101

the entire range of angular momentum states contributing to the a ionization

channel. When the au wavefunction is expanded about the center of the molecule
(coordinate r in Fig. 2.6) it is the 1= 3 component which is resonantly trapped at
a particular kinetic energy of the photoelectron.
A seemingly different explanation for the a* resonance arises from MO
calculations. As first discussed by Rescigno and Langhoff [4.95], and later

(a) N2 3a~ Antibonding Orbital

(b) N2 Resonant au 2 =3 E igenchannel Wavefu net ion

2
'Ita
o
-2

-1
o

(e) N2 au 2=3 Wavefunction on and off Resonance

12.2 eV

16.3 eV

20.4 eV

Fig. 4.11. (8) Picture of the 317: anti bonding molecular orbital in N2 taken from [4.58], which in a
molecular orbital picture is the final state of the u· resonance excitation. (b) Eigenchannel contour
plot of the resonant 1= 3 uu Xa-MS wavefunction taken from [4.120]. The N2 molecule lies along
the z-axis with its center at y = z = O. Note the close correspondence of the central contours in the
y-z plane with the shape of the orbital in (a). (c) Radial plot from the molecular center of the I = 3 uu
Xa-MS wavefunction at different final state energies, below (12.2 eV), at (16.3 eV), and above 20.4eV)
resonance. The nuclear position of N is also marked. Note that the shape resonance wavefunction is
strongly localized on the molecule
102 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

clarified by Hermann and Langhoff[4.96] and Thiel [4.97], the (1* resonance can
be understood in a MO picture as arising from a transition to the 3(1~
unoccupied virtual orbital. This orbital, plotted in Fig. 4.11a, is spatially com-
pact, giving rise to the large (1* resonance intensity, and has p-like lobes along
the internuclear axis, centered on each N atom.
The link between the MS and MO concepts is established most elegantly by
viewing the MS wavefunction (Sect. 2.7.4) as similar to the contour maps used
for picturing MOs. This is done in Fig. 4.11, where the orbital contour of the 3(1~
molecular orbital is compared with the resonant I = 3 eigenchannel wave-
function for N2 [4.120]. In Fig. 4.11b, the N2 molecule lies along the z-axis with
the center of the molecule at y = z = O. At large distances the wavefunction has
the three nodal planes characteristic off orbitals. However, the four innermost
"orbital" contours, best visualized as the projection in the y-z plane, show a
striking resemblance with those of the 3(1~ molecular orbital in Fig. 4.11a. The
local p-like symmetry of the final state MS wavefunction at the site of each of the
two N atoms can again clearly be seen in the y-z plane projection. The
enormous enhancement of the wavefunction at resonance (16.3 eV) and its
compact spatial extent, with an intensity peaking at the N atom position, is
shown in Fig. 4.11c. Note that the (1* resonance energy in Fig. 4.11c is higher
than that in Fig. 4.9 because a different rx value was used in the exchange
potential (2.56).
The close correspondence between MS shape resonances and transitions to
antibonding molecular orbitals demonstrated for N2 holds quite generally and
in the following we shall discuss this useful correlation for a particularly
important class of molecules, the hydrocarbons.

4.2.7 Molecular Orbitals and Resonances of Simple Hydrocarbons

Because of their importance in catalytic processes and as the backbone of
polymers, hydrocarbons warrant special attention. The basic building blocks of
such systems are C-C bonds of different hybridization and additional C-H
bonds and we therefore expect their K-shell spectra to exhibit all the various
features discussed in the preceding sections.
Experimental ISEELS spectra for the simple hydrocarbons acetylene
(HC=CH), ethylene (H 2C=CH 2), and ethane (H 3 C-CH 3 ), shown (Fig. 4.12a)
exhibit rather complex structures. In particular, for C 2H 2 and C 2H 4 there are
many resonances above the IP, and at first sight it is difficult to make an
unambiguous assignment of the (1* shape resonances. It is for this reason that
the assignment of Sette et al. [4.79, 103], which was based on systematics of (1*
resonance positions in a variety of molecules was criticized by Piancastelli et al.
[4.123, 124].
Compared to the experimental spectra, the spectra calculated by means of
Xrx-MS theory [4.64] shown in Fig.4.12b are much simpler and exhibit the
various resonances expected from our previous discussions of other molecular
K-shell spectra. In particular, the Xrx-MS calculation for C 2 H 2 is in good accord
 4.2 Characteristic Resonances in K-Shell Spectra 103

Experiment Theory

rr* (a) rr* (bl

0*
0*
x5

x3

280 290 300 310 320 280 290 300 310 320
Excitation Energy (eV)

Fig. 4.12. Comparison of experimental ISEELS K-shell excitation spectra of acetylene (HC=CHl
[4.51], ethylene (H 2 C=CH 2 l [4.27], and ethane (H3C-CH3l [4.81] with XIX-MS calculations of the
same molecules [4.64]. Comparison of experimental and theoretical spectra indicates that all
resonances shown shaded originate from multi-electron transitions

with that carried out using the MO-ST method [4.125]. The good agreement of
experiment and theory shown in Figs. 4.4 and 4.5 indicates that many of the
peaks above the IP in the C 2 H 2 and C 2 H 4 spectra are of multi-electron
character, which the calculation, of course, does not reproduce. We have shaded
those structures in Fig. 4.l2a. In particular, in agreement with our classification,
the prominent peaks at 295eV in the C 2 H 2 spectrum and at 293eV for
C 2 H 4 have been attributed to two-electron excitations of the kind
(1s}-1(1n) - 1(2n*)1(RydV [4.63, 126, 127]. The involvement of the n system in
the double excitation process is also indicated by the absence of such resonances
in the spectrum of C 2 H 6 , shown at the bottom of Fig. 4.12.
Because the spectra of unsaturated hydrocarbons contain discrete n*, C- H*
and Rydberg features, it is interesting to compare the density of the associated
orbitals. This is done in Fig. 4.13 for the ethylene molecule. Again the Xcx-MS
method was used [4.64] and the core hole was located in the upper C atom,
indicated C*, in Fig. 4.13a. The labeling corresponds to the C 2v symmetry of the
excited molecule. The orbital contours show that the 2b 1 orbital, associated with
the n:.* resonance in the spectra in Fig. 4.12, is the most compact, in agreement
with the large resonance intensity. In comparison, the density of the 3b 2 C-H*
orbital and the 3b 1 Rydberg orbital which both contribute to the peak near
288 eV in the C 2 H 4 spectra in Fig. 4.12 is smaller by about a factor of 5. Note
that the C-H* orbital clearly exhibits density at the positions of the H atoms.
This orbital corresponds to the 2b 3u MO (D 2h symmetry) plotted by Jorgensen
104 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

(a) Geometry (b) 3b 2 Orbital (e-H')

x5
H

H H
y-z plane

(e) 2b 1 Orbital (71') (d) 3b 1 Orbital (Rydberg)

~ . .

• x-z plane
J x-z plane
l
Fig. 4.13a- d. Geometry and orbital contour plots of several obitals for ethylene. The irreducible
presentations correspond to the C 2Y broken symmetry in the presence of a core hole, placed on the
upper C atom labeled C*. The 3b 2 orbital shown in (b) is ofC-H* valence nature and gives rise to the
resonance at 288eV in the C 2 H4 spectra in Fig. 4.12. The 2b , orbital is the n* orbital associated with
the C = C bond and with the n* resonance near 285 eV. The 3b l orbital is of pure Rydberg nature
and, like the 2b , orbital, it is perpendicular to the molecular plane containing the hydrogens. It also
contributes to the resonance at 288 eV in Fig. 4.12. Note the reduced density of the C-H* and
Rydberg orbitals relative to the n* orbital, yet the similar density of the Rydberg orbital and the
C-H* valence orbital

and Salem [4.58]. The similar density ofthe 3b 1 Rydberg orbital is in contrast to
the conventional picture of a diffuse nature of Rydberg orbitals. Our results
therefore underscore the fact known from elementary quantum mechanics
[4.28] that this picture only holds for the higher members of the Rydberg series
while the first members, e.g., the 3b 1 orbital, are rather compact.

4.2.8 Exchange Splitting in the Oxygen Molecule

The open shell oxygen molecule is expected to exhibit a magnetic exchange
splitting of the (1* resonance. Such a splitting is indeed observed. In accordance
with the simple MO diagram shown in Fig. 3.1 Ie, the spectrum calculated by
means of a self-consistent spin-dependent XIX-MS calculation and the gas phase
ISEELS [4.51] spectrum exhibit three peaks, as shown in Fig. 4.14. The calcu-
lation utilized Slater's transition state method and a semiempirical choice of the
muffin-tin spheres, i.e., adjustment of their radii to yield the spin-dependent
experimental Is ionization potentials [4.128] 543.1 eV (41' ionic final state) and
544.2 eVe l' ionic final state) for the O 2 molecule.
 4.2 Characteristic Resonances in K-Shell Spectra 105

Fig. 4.14. (a) Spin- and polarization-

(a) Theory resolved NEXAFS spectrum calculated
-----" 9
.. " ¢ by means of self-consistent XCl multiple
scattering theory. All calculated oscil-
--u9 lator strengths were converted into
---u9 cross sections by convolution with a
Gaussian of [Link] FWHM. The calcu-
lated ionization potentials (same as
measured) corresponding to 4}; and 2 J;
ionic final states are also indicated. (b)
Experimental electron energy loss
spectrum of gas phase O 2 [4.51]
(b) Experiment

530 532 534 536 538 540 542 544 546

Excitation Energy (eV)

The labeling of the resonances in Fig. 4.14 is chosen to agree with that used
earlier for Fig. 3.11c. Resonance A at ~ 531 eV corresponds to the lau -+ In:
transition. According to Hund's rule, in the ground state, the two degenerate
n:
1 orbitals of O 2 are each singly occupied by electrons with parallel spins, and
we shall assume that their spins are up, as shown in Fig. 3.11c. The Pauli
principle and the dipole selection rule which forbids any spin-flip transition then
demand that the 1i* resonance corresponds to excitation of a spin-down electron
from the Is shell (lau orbital).
Resonances Band C in Fig. 4.14b primarily correspond to a* resonances,
i.e., spin-up and spin-down lag -+ 3a: transitions, respectively, the splitting
arising from two different final state spin configurations, see Fig. 3.11c. The
exchange splitting of the a* resonance in the gas phase spectrum in Fig. 4.14b is
remarkably large ( ~ 2.5 e V), and exceeds the calculated splitting (1.2 eV) in the
a* system shown in Fig. 4.l4a. This quantitative difference is due to the limited
accuracy of the calculated transition energies 3 but it does not affect the qualitat-
ive reliability of the calculated spectra.

3 In the XCl-MS transition state calculation we removed one half electron from the Is shell. This
allowed us to calculate the whole spectrum with the same potential. In order to calculate transition
energies between two orbitals more accurately one should also add half an electron to the upper
orbital. In our case this procedure would have resulted in an "orbital energy" for the 3u! orbital
above the vacuum level and was therefore not used.
106 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

Closer inspection of all transitions calculated by the Xo:-MS method reveals

that the schematic molecular orbital picture given in Fig. 3.l1c is oversimplified
in that it neglects the influence of Rydberg orbitals. The calculation predicts
several transitions to Rydberg orbitals of n character which mostly fall in the
region of resonance C. These transitions are also shown in Fig. 4.14a and have
been broadened by the same 1.0eV Gaussian function as all other calculated
transitions. Furthermore, resonance C itself is composed of two roughly equal
components of spin-down character which are not resolved in Fig. 4.14. We
believe this splitting to arise from a mixing of Rydberg orbitals with the 3a!
valence orbital. For O 2 the 3a! valence component dominates, as discussed
earlier in Sect. 4.2.3 in conjunction with Rydberg and hydrogen-derived reson-
ances. This valence-like nature of resonances Band C has the important
consequence that they are expected to remain largely unaffected by extra-
molecular bonds, in contrast to pure Rydberg resonances. As shown in Chap. 10,
this is supported by results for solid O 2 and physisorbed O 2 ,

4.3 Systematics of Resonance Positions

In our overview of the various characteristic resonances encountered in K-shell

spectra of low-Z molecules, we have already indicated a hierarchy of energy
positions. Typically n* resonances are lowest in energy, followed by Rydberg
and/or hydrogen-derived resonances and, at the highest excitation energies, a*
resonances. Here we are interested in the energetics of a characteristic resonance
within differently bonded molecules, e.g., the energy position of the n* resonance
in molecules with C=C, C=C, C=N, or C=O bonds. We shall restrict our
discussion here to "simple" molecules as defined in Sect. 4.2.1. This insures that
the resonances are characteristic of well-defined local bonds and not disturbed
by bond-bond interaction effects.
In our attempt to understand systematics among characteristic resonances,
it is advantageous to compare relative rather than absolute resonance positions.
If we ignore Rydberg transitions, a K-shell resonance corresponds to a trans-
ition from a localized Is to a molecular orbital final state which is delocalized
over the bonded pair of atoms. Therefore, transitions to the same MO can be
observed by Is excitation of either atom participating in the bond. For example,
the n resonance in CO is observed at the C as well as the 0 K-edge (Fig. 4.5). It is
to be expected that, although the absolute excitation energies for a characteristic
resonance are substantially different at different K-edges, the resonance posi-
tions relative to the respective Is IPs should be rather similar. This is because the
resonance position relative to the IP is a direct measure of the orbital energy of
the excited state MO (the "optical orbital") relative to the vacuum level, Ev , as
discussed in Sect. 2.6.1 and illustrated in Figs. 2.4, 4.2. By referencing resonance
positions to the corresponding IPs we can therefore significantly increase our
data base. Other reasons for this procedure will become apparent later.
 4.3 Systematics of Resonance Positions lO7

Table 4.1. n* resonance positions and Is ionization potentials in nonconjugated simple molecules

Is ionization n* resonance
potential' position ~. = E. - IP
z· Molecule b IP [eV] E.[eV] [eV] Reference

12 HC=CH 291.1 285.9 - 5.2 4.129

H 2 C=CH 2 290.8 284.7 - 6.1 4.27
F 2 C=CF 2 296.5 290.1 - 6.4 4.27
13 HC*=N 293.4 286.4 -7.0 4.130
HC=N* 406.8 399.7 - 7.1 4.130
14 N=N 409.9 401.0 - 8.9 4.54
C*=O 296.2 287.3 - 8.9 4.54
C=O* 542.5 534.1 - 8.4 4.54
H 2 C*=0 294.5 286.0 - 8.5 4.73
H 2 C=0* 539.4 530.8 - 8.6 4.73
F 2 C*=0 299.6 290.9 - 8.7 4.73
F 2 C=0* 540.8 532.7 - 8.1 4.73
15 N*=O 4lO.7 d 399.7 - 11.0 4.131
N=O* 543.2d 532.7 -lO.5 4.131
CF3N*=0 410.5 399.1 - 11.4 4.132
CF 3N=0* 542.3 530.8 -11.5 4.132
16 0=0 543.7 d 530.8 - 12.9 4.54

• Sum of atomic numbers of bonded pair

b The excited atom is indicated by an asterisk

, For a tabulation of IPs, see [4.55]

d Weighted average of doublet

Let us start with the lowest energy resonance observed for all unsaturated
bonds, the 11:* resonance. We have summarized in Table 4.1 the 11:* resonance
position E", the corresponding IP, and the difference ~" = E" - IP in a variety
of "simple" molecules. Note that -~" is usually called the term value in the
literature. We prefer our (opposite) sign definition of the energy difference
because it directly reflects whether the optical orbital lies below (negative
energy) or above (positive energy) E v , in accordance with the definition of
calculated orbital energies. We have grouped the molecules in Table 4.1 accord-
ing to Z, the sum of atomic numbers of the bonded atomic pair, such that a C-C
bond is characterized by Z = 12, etc. Within each group we have listed mole-
cules with different bond orders (triple and double bonds) and different 1s IPs,
arising either from excitation of different atoms or from chemical shifts for a
given atom [4.55].
Within each group, the 11: resonance position relative to the IP is remarkably
constant, to within about 1 eV, independent of the hybridization, the atom
whose K-shell is excited, and chemical shifts of the 1s IP. The independence of
~" to hybridization, and therefore bond length, can be understood by the
orthogonality of the 11: system to the internuclear axis and the resultant small
change in overlap of the p functions on adjacent bonded atoms as a function of
bond length. As expected from the delocalized nature of the MO final state, the
108 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

C K-edge Fig. 4.15. Carbon K-shell excitation spectra of

some fluorinated aldehydes [4.73]. In these mole-
15 cules the carbon Is IP exhibits a sizeable chemical
shift arising from charge transfer from carbon to
10 fluorine. When the spectra are plotted relative to
~
I the IP, it is seen that the Jr* and u* resonances
>Q) remain fixed in energy
C\I
I
5
0
10
.c 0
C,
c:
Q)
5
~
0 6
~ 4 o
·0
rn
0
2

-10 o 10
Energy from IP (eV)

energy ~" of the opticaln* orbital for heteroatomic bonds does not depend on
which atom gets excited. More surprizing is the insensitivity of ~" to chemical Is
binding energy shifts. This is illustrated in Fig. 4.15 for the carbon K-shell
excitation of fluorinated aldehydes, H 2C = 0, FHC = 0, and F 2C = 0, and in
Fig. 4.16 for the n* resonance in fluoroethylenes. For H 2C = CH 2, HFC = CFH,
and F 2C = CF2 (Fig. 4. 16a), the two carbon atoms in the molecule are equivalent
and the C Is IP shifts by as much as 5 eV upon fluorination, as for the aldehydes.
The n* resonance position follows this shift, leaving ~" unchanged. Further-
more, two distinctly shifted C = C n resonances are observed when the two
carbons are inequivalently bonded to hydrogen and fluorine as in H 2C = CHF,
H 2 C = CF 2, and HFC = CF 2 (Fig. 4.16b). The two n* resonances follow the IP's
for the respective carbon atoms.
Table 4.1 reveals that ~" is, however, dependent on Z, and the n* resonance
shifts away from the IP with increasing Z, corresponding to a lower-energy or
stronger-bound n* optical orbital. While the resonance is only about 5.5 eV
below the IP for C-C bonds, it is about 13 eV below the IP for the O 2 molecule.
This simply reflects the larger electron-nuclear Coulomb interaction, which
according to (2.19) increases with Z. To summarize, the n* resonance position
relative to the IP is mostly a function of Z, the sum of atomic numbers of the
bonded pair, and for a given Z is largely independent of hybridization, the
excited atom in a heteroatomic bond, and Is initial state chemical shifts. On an
absolute transition energy scale the n* resonance within a given Z group
therefore shifts with the Is IP.
The observations made for the n* resonances also hold for the Rydberg/C-H
resonances. The Rydberg/C-H resonances associated with the methyl group
 4.3 Systematics of Resonance Positions 109

10
10 8
,, ...... .C=C......... ,, 6
5 CH,
4

i 2
> 0
0
'1''" 295 8 H.... C..... ,
~ 10 = .... ,
H..... C
.r; 6
c;, ,H..... . .C=C.........H,
.."'" 5 4

..
Vi
0
2
:§ 0 0
'u., 8 5
0
6
H.... C C..... H
4
H..... = ....
H

Excitation Energy (eV)

Fig. 4.16a, b. K -shell excitation spectra near the 7[* resonances (shaded) for fluoroethylenes [4.27].
In (a) both carbon atoms are equivalent and their 7[* resonance position follows the IP, as for the
aldehydes shown in Fig. 4.15. For molecules with inequivalent carbon atoms, as in (b), two IPs and
two 7[* resonances are observed. Again, the 7[* positions move with the respective IPs

(CH 3-) are shown in Fig. 4.17 as a function of attached atoms or groups of
varying electronegativity. Similar to NH3 (discussed in Sect. 3.2.2), the CH 3
group is characterized by two empty Mas, a a*-like and a n*-like state [4.58].
Resonances A and B correspond to transitions to these final states which, as
discussed in Sect. 4.2.3, are mixed with Rydberg orbitals of the same symmetry.
The more intense n*-like resonance B occurs at 288.0eV in methane (CH3-H)
[4.53], [Link] in ethane (CH 3-CH 3) [4.81, 129], 288.5eV in methylamine
(CH 3-NH 2 ) [4.53], 289.4eV in methanol (CH 3-OH) [4.37, 53], and 290.5eV in
fluoromethane (CH3-F) [4.47, 51]. Its energy shift and that of the weaker
resonance A closely follows the chemical shift of the C Is IPs listed in the figure,
caused by the electronegative oxygen or fluorine ligands [4.55]. Thus, as for the
n* resonance of a specific atomic pair characterized by Z, the positions of the
Rydberg/C-H resonances are fixed relative to the IP.
In contrast to the rather simple behavior of the n* and Rydberg/C-H
resonance positions, the a* resonances exhibit a more complex energy depend-
ence. The representative data for molecules with "isolated", well-defined bonds
are summarized in Table 4.2, and their separation from the IP, .1a , is plotted in
Fig. 4.18 as a function of Z. In comparison we also show in this figure the data
110 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

Fig. 4.17. ISEELS spectra of molecules containing

B
the CH 3 group and ligands with different electro-
negativities. Peak positions A and B marked by
solid lines correspond to RydbergjC-H resonances
and follow the listed C I s IPs marked by a dashed
line [4.55]. The shown spectra are taken from the
following references: methane [4.53], ethane
[4.81], monomethylamine [4.53], methanol [4.37],
and ftuoromethane [4.51]. The shaded resonances
are associated with transitions to (top to bottom)
C-C, C-N, C- O and C-F anti bonding (1* orbitals
(shape resonances). Note that for CH3F the C-F (1*
resonance by chance falls at the same energy as the
lowest RydbergjC- H resonance A [4.133]

290 300
Electron Energy Loss (eVj

15

c:
:2
·iii
o
Q.
Q)
()
c:
'"
g -5
"'
Q)
Fig. 4.18. Energy positions of n* and (1*
a: resonances relative to the K -shell IPs as a
-10 function of Z, the sum of atomic numbers
rr* of bonded atomic pairs (see Tables 4.1 and
4.2). Note the large scatter of data points
-15 for the (1* relative to the n* positions
which is caused by the strong dependence
12 13 14 15 16 17 18
of the (1* position on hybridization and
Sum of Atomic Numbers Z
bond length
 4.3 Systematics of Resonance Positions 111

Table 4.2. (1* resonance positions and Is ionization potentials in nonconjugated simple molecules

Is ionization (1* resonance ~a = Bond

potential e position Ea - IP length d
z· Molecule b IP [eV] Ea [eV] [eV] Ref. [A]

12 HC;=CH 291.1 310 19 4.129 1.203(1)

H 2C=CH 2 290.8 301 10 4.129 1.337(2)
H 3C-CH 3 290.7 291.2 0.5 4.129 1.533(2)
13 HC*=N 293.4 307.9 14.5 4.130 1.158(3)
HC=N* 406.8 420.8 14.0 4.130
H3C*-NH2 291.6 291.5 - 0.1 4.53 1.465(2)
H3C-N*H2 405.1 404.6 - 0.5 4.53
F3C-N*0 410.5 407.5 - 3.0 4.132 1.546(8)
14 N=N 409.9 418.9 9.0 4.54 1.095(5)
C*=O 296.2 304.0 7.8 4.54 1.128(5)
C=O* 542.5 550.0 7.5 4.54
H 2C*=0 294.5 300.5 6.0 4.73 1.209(3)
H2C=0* 539.4 544.0 4.6 4.73
B*-F3 202.8 205.1 2.3 4.134 1.313(1)
B-F! 694.8 699.0 4.2 4.135
H3C*-OH 292.3 292.0 -0.3 4.53 1.425(2)
H 3C-O*H 538.9 537.4 - 1.5 4.53
H2N-NH2 406.1 405.1 - 1.0 4.132 1.449(4)
15 N*=O 410.7" 414.5 3.8 4.131 1.150(5)
N=O* 543.2" 546.3 3.1 4.131
F 3CN*=0 410.5 413.2 2.7 4.132 1.197(5)
H3C*-F 293.5 289.1 - 4.4 4.5,47 1.382(5)
H3C-F* 692.4 691.0 - 1.4 4.136
16 0=0 54P" 540.5 - 3.2 [4.54, 104] 1.207(5)
N*-F 3 414.4 407.1 -7.3 4.137 1.365(2)
N-F! 693.2 687.4 - 5.8 4.137
HO-OH 541.8 533.0 - 8.8 4.138 1.475(4)
17 F 2-O* 545.3 534.6 -10.7 4.139 1.405(1)
F1-0 695.1 683.8 - 11.3 4.139
18 F-F 696.7 682.2 -14.5 4.140 1.417(5)

• Sum of atomic numbers of bonded pair

b The excited atom is indicated by an asterisk

e For a tabulation ofIPs, see [4.55]

d For a tabulation of bond lengths, see [4.70, 141]

• Weighted average of doublet

from Table 4.1 for ~'" the n* resonance positions relative to the IP. Compared
to the n* positions, the 0"* energies exhibit large scatter. At first sight, the only
obvious trend is that all data lie within upper and lower bounds shown as
dashed lines. The upper bound is determined by the molecules with the highest
bond order (triple and double bonds) for a given Z, and the lower by those with
the lowest bond order (single bonds). This is strong evidence that ~a, in addition
to Z, also depends on the bond length R, which, of course, changes with
hybridization. This dependence is also intuitively expected because the molecu-
112 4. Experimental and Calculated K-Shell Spectra of Simple Free Molecules

lar a orbitals lie along the internuclear axis, in contrast to the 11: system. We can
make the following two general observations: (1) /l." decreases with increasing Z,
similar to but stronger than /l.". (2) The spread in /l." decreases with increasing Z,
and the upper and lower bounds appear to converge near Z = 18 (the F 2
molecule).
Let us try to qualitatively understand the trends in /l." with Z as shown in
Fig. 4.18. This is best accomplished by first ignoring the complications intro-
duced by different bond hybridizations, i.e., bond lengths. Since the data points
appear to converge near F2 (Z = 18) let us take its bond length R =.1.417 A as
our reference. From Table 4.2 we see that the molecules F 2-0 (Z = 17) with
R = 1.405 A and H3C-OH (Z = 14) with R = 1.425 A have similar bond lengths
and their /l." values all lie approximately on the solid line marked
"R = 1.42 A" in Fig. 4.18. Hence to a good approximation this line represents
the Z-dependence of /l." at constant interatomic distance. Its slope is approxim-
ately a factor of 2 larger (3.5 eV for /l.Z = 1) th~m that ofthe corresponding line
for /l." (1.8eV for /l.Z = 1) also shown in the figure. The larger Z-dependence of
the a* relative to the 11:* resonance position can be understood from Fig. 3.1. As
discussed in Sect. 3.2, the 11: MOs originate from the 2px and 2py orbitals of
adjacent bonded atoms, and, while the energy of the two atomic p orbitals
decreases with Z, the 11:-11:* splitting is approximately constant. Therefore the Z-
dependence of the 11:* resonance follows that of the atomic p orbitals and is
dominated by the electron-nuclear Coulomb interaction. The a MOs originate
from the 2pz and the 2s atomic orbitals and therefore their energy positions
depend on both orbitals, in particular, their energy position and mixing. Owing
to the increasing energy separation of the 2p and 2s orbitals with increasing Z,
the 2pz-2s mixing is reduced substantially. From Fig. 3.3, we see that the 2pz-2s
splitting is about 10eV for carbon but about 25eV for fluorine. This greatly
affects the asymmetry of the positions of the outermost a and a* pair relative to
the atomic 2p level as shown for the 3ag and 3au states in Fig. 3.1. Hence with
increasing Z, not only is the energy position of the atomic 2p level lowered but
also the increasing separation from the 2s level, which drops considerably more
in energy with Z, causes the asymmetric a-a* splitting of the outermost pair to
diminish. The overall effect is a larger downwards energy shift of the highest a*
level than for the 11:* level.
The scatter of the data points around the solid R = 1.42 Aline in Fig. 4.18 is
caused by the strong dependence of the a* resonance position on hybridization
and bond length. Again it is easy to visualize the effect of a distance change
in a molecular orbital representation similar to Fig. 3.1. The a-(J* splitting is
strongly dependent on the internuclear distance because the overlap of the 2pz
orbitals on adjacent bonded atoms, which are directed along the internuclear
axis, changes greatly with distance. Hence, at short bond length the splitting is
larger than at longer bond length and the a* level is higher for small rather than
large R. Furthermore, with increasing Z the 11:* orbital becomes filled, being
empty at Z = 14 (N 2), half full at Z = 16 (0 2) and filled at Z = 18 (F2)'
Therefore the range of bond hybridization and bond length decreases with
 4.3 Systematics of Resonance Positions 113

increasing Z. While N 2 is triple bonded, O 2 is double and F 2 single bonded. The

attractive interaction caused by the filled 1t bonding orbital is compensated by
the repulsive interaction arising from filling of the anti bonding 1t orbital. This
leads to the convergence of the upper and lower bounds near Z = 18.
Because of its important applications in the study of chemisorbed molecules,
whose bond lengths are largely unknown, the detailed dependence of the u*
resonance position on bond length will be discussed in more detail in Chap. 8 for
free, polymeric and chemisorbed molecules.
5. Principles, Techniques, and Instrumentation
ofNEXAFS

In this chapter a basic question is addressed: How can the X-ray absorption
signal from a single molecular layer on the surface of a bulk material be
measured? In particular, electron yield and fluorescence yield detectors and
experimental techniques are discussed and specific attention is given to the
problems of normalization and background correction of experimental data.

5.1 Achieving Adsorbate Sensitivity

The link between K -shell spectra and intramolecular structure outlined in the
last chapter for free molecules points to the potential of near edge spectroscopy
for investigating the internal structure of chemisorbed molecules, as well. Here
we shall discuss how such spectra can be obtained experimentally. Compared to
gas phase spectroscopy there are two major complications associated with the
recording of spectra for molecules bonded to surfaces. First, the molecular
density on the surface is small. Typical surfaces have an atomic density of abou~
10 15 atoms/cm 2 and since molecules usually cover a surface in a single layer this
number also characterizes the molecular density. In contrast, for gas phase
studies the electron or X-ray beam traverses a volume of gas and the molecular
density can be adjusted at will. Typically it is chosen to be in the 10 17_10 18
atoms/cm 2 range. Secondly, the signal from the molecular layer is superimposed
upon an unwanted background signal from the bulk substrate. In particular, it is
the combination of the two problems, the low concentration of adsorbate atoms
and the high concentration of substrate atoms, which leads to experimental and
instrumental challenges.
For the study of chemisorbed molecules on surfaces the ISEELS technique is
not well suited. The use of high energy electron beams [5.1] in conjunction with
a transmission geometry through a thin film substrate 1, especially in the form of
single crystals, is undesirable because of the awkwardness and general unavail-
ability of such samples. Even if such substrates were readily available ex-
periments would be impeded by unfavorable signal-to-background ratios for
realistic thicknesses ('" 500 A) of free standing films. Radiation damage of the

1 High energy ISEELS measured in transmission through thin films is discussed in [5.2].
 5.1 Achieving Adsorbate Sensitivity 115

molecular adsorbate layer resulting from the high power density of the electron
beam [5.3] may also be a problem. For the same reason, reflection geometry
experiments [5.4, 5] are problematic, too, and they have only recently been
applied to the study of chemisorbed molecules [5.6, 7]. In addition, "polariza-
tion" dependent studies are difficult or impossible, which, as we shall see below,
make X-ray absorption studies so very powerful.
The above considerations indicate that for our purposes X-ray absorption
spectroscopy is indeed the technique of choice. NEXAFS studies, however, are
not without experimental and instrumental challenges. The small concentration
of surface species necessitates the use of high photon intensities, as demon-
strated by a quick estimate. Using the K-shell X-ray absorption cross section
1 Mb/molecule = 10- 18 cm 2/molecule for N2 from Fig. 4.9 and assuming a N2
monolayer of 10 15 molecules/cm 2 on the surface we find that it takes 103
photons to create one absorption event. Also, let us assume that we can monitor
each absorption event by collecting the photoelectron created in the event and
energy analyze and detect the electron with a total efficiency (solid angle,
transmission and detection) of 10- 3. Then, to obtain a count rate of 10 3
counts/s, needed for a spectrum with reasonable statistics, a photon flux of
1 x 10 9 photons/s is required. The low concentration of the molecules on the
surface therefore necessitates the use of powerful soft X-ray sources.
The additional requirement of high spectral resolution, which can only be
achieved in the soft X-ray region by monochromator optics with a small angular
acceptance [5.8-11], eliminates all but synchrotron radiation sources. Only for
highly collimated synchrotron radiation does the match of the emission charac-
teristics of the source and the small angular acceptance of the monochromator
provide for X-rays of sufficient intensity and energy resolution. In the following
we shall simply assume that monochromatic soft X-rays of sufficient intensity
are available. Although this statement is not rigorously true at the time of
writing it will certainly be true in the near future with the construction of new
powerful synchrotron radiation sources around the world.
The essence of the remaining problem is to find an experimental detection
method for the core excitation event which is specifically sensitive to the signal
from the molecule and insensitive to the background signal from the substrate.
For practicability we shall require that conventional substrates of macroscopic
dimensions (e.g., 10 x 10 x 2mm 3 ) can be used. This leads to a desired geometry
where the X-ray beam is incident on the surface of a sample at a specific but
variable angle and the characteristic signal from the adsorbate layer is measured
by use of a suitable detector which views the surface.
We want to measure a signal which is proportional to the X-ray absorption
cross-section crx(hv) defined by (2.8). In the limit of small adsorbate concentra-
tions, of interest here, the number of photons absorbed in the molecular layer is
(5.1)
where 10 [photons/(scm 2)] is the incident photon flux density, Ao [cm 2] is the
area exposed to the beam, which depends on the X-ray incidence angle,
116 5. Principles, Techniques, and Instrumentation of NEXAFS

Auger Fig. 5.1. Schematic diagram of a photon absorption

Electron process resulting in a photoelectron and a core hole.
Photo-
electron The hole is filled by an electron from a higher shell,
either radiatively by emission of a fluorescent photon,
or nonradiatively by emission of an Auger electron
-----+------+---Ev
----~--~~~-B

hv FI uorescent
Photon

----~--~-----A

p [atoms/cm 2 ] is the atomic area density and O"x(hv) is given in units of

Mb/atom = 10 - 18 cm 2 /atom. The number of absorbed photons is therefore
directly proportional to the X-ray absorption cross section and, according to
Fig. 5.1, so is the number of created core holes and photoelectrons. Figure 5.1
suggests that the most direct method of measuring the X-ray absorption cross
section of chemisorbed molecules is to monitor the intensity of the photo-
electrons which constitute the primary excitation channel. However, from
inspection of the energy scale of Fig. 4.2, it becomes clear that this is not a viable
method. All structures below the IP corresponding to bound state excitations
would be lost since only free photoelectrons are measured in photoemission.
This leads to the measurement of Auger electrons or fluorescent photons
associated with the secondary process of core hole annihilation. Both channels
are a direct measure of the probability of the existence of a core hole created by
X-ray absorption.
The secondary deexcitation process occurs either radiatively, by emission of
a fluorescent photon, or nonradiatively, by Auger electron emission. The
fractions of the nonradiative and radiative decay rates relative to the total decay
rate are called Auger yield Wa and fluorescence yield wJ' respectively, and satisfy
the sum rule Wa + wJ = 1 [5.12]. The relative yields are a strong function of
atomic number Z as shown for the K- and L-shell fluorescence yields of the
elements in Fig. 5.2a. For the K-shell excitation of low-Z atoms and for the L-
shell excitation of all atoms with Z < 90 the Auger decay is faster and hence
dominates [5.12]. In particular, for C, N, and 0 atoms the Auger electron yield
is favored over the fluorescence yield by more than two orders of magnitude, see
Fig. 5.2b. Nevertheless, as detailed below, both channels prove to be useful in
recording NEXAFS spectra of chemisorbed molecules.
Before discussing the principles and experimental techniques of electron
yield and fluorescence yield detection we should note the possibility of using a
third approach, namely ion yield detection, as a measure of the molecular cross
section. In the time evolution of events, molecular fragmentation and ion
desorption from the surface follow the primary excitation ( - 10- 18 s) and the
secondary Auger decay (_10- 15 s) as a tertiary (_10- 13 s) event. In a model
first suggested by Carlson [5.14] for free molecules and later by Knotek and
 5.1 Achieving Adsorbate Sensitivity 117

Fig. 5.2. (a) Overview of the fluorescence

yields following the excitation of K, L 3 , L 2 ,
and L, shells as a function of atomic number
Z [5.12, 13]. (b) Fluorescence yields for the K-
shell excitation of low Z atoms. We have used
0.20 the values listed in Table 3 of the review by
Krause [5.12]
0.10

0.05

0.02
"0
~
>-
'"uc
Atomic Number Z
'"~'"
u
(b)
0 10- 1 P~
Si

.--
.2
LL Mg AI

10-2
;./Ne Na

F=N
f-C
B'
10-3
o 500 1000 1500 2000 2500
Photon Energy (eV)

Feibelman [5.15] for ionic surfaces, the ion yield following core level excitation is
directly related to the Auger decay process. The repulsive forces between holes
created in the valence shell by an Auger decay may result in a "Coulomb
explosion" between two bonded atoms.
The question arises whether ion yield detection, in the form of individual
fragments or the sum of all ions, can be used as a measure of the X-ray
absorption cross section. Unfortunately, this is not the case for chemisorbed
molecules as first shown by Jaeger et al. [5.16] and displayed in Fig. 5.3 using
the data of Treichler et al. [5.17] for CO chemisorbed on Ru(001). The ion yield
spectra of various fragments are found to be significantly different from the
Auger electron yield spectrum shown at the bottom. The Auger yield with its
linear response to the core hole population mirrors the average of all decay
channels and is proportional to the X-ray absorption cross section. The detailed
Auger spectra of chemisorbed molecules furthermore reveal that the vast
majority of electronically excited states, in particular the multielectron ("shake-
up") states observed in photoemission, have boiled down to a well-screened
adiabatic state before Auger decay takes place [5.18-21]. In contrast, ion
fragmentation and desorption selectively follow the minority of cases where the
primary multielectron excitations survive long enough to influence fragmenta-
tion and desorption, aided by the additional holes created through Auger decay.
118 5. Principles, Techniques, and Instrumentation of NEXAFS

Fig. 5.3. Ion yields of CO +, 0 +, 0 2 +,

1«4<.~
..~' ' . ,. .
500
~ .~... ~,.~.<!("~~
.
and C+ following 0 K-shell excitation of
CO on Ru(OOI) in comparison with the
V~ co+ Auger electron yield spectrum obtained
100 .J by detecting 0 K VV electrons of 510 eV
kinetic energy [5.17]. Note that the ion
4000
~~~J'

__r
yield does not follow the Auger yield,
which is representative of the 0 K-shell
X-ray absorption cross section
« / 0+
E
0
0 1000
.......
~

en
.......
en
.....
C
::J
500 ~.r
2
0

100
./~'~~:'
"a; 200 :ti
>- ~Mt'-:.\.",~
100 '.AI":;'~ C+
en I J~
.'=
c
::J Vi\~
...
.ri
CtI J:I
o (KVV) Auger Yield
530540 550560 570 580590 600
Photon Energy (eV)

Therefore, the ion yield does not statistically average over all Auger decay
events. This explains the small desorption rates and steps in the ion yield spectra
in Fig. 5.3 that occur at energies associated with specific multielectron-
multi hole excitation thresholds. Ion desorption therefore demonstrates that in
order to measure the absorption cross section a detection technique has to be
used which statistically averages over all Auger or fluorescence channels.
In the following we shall discuss the principles and techniques of electron
and fluorescence yield detection.

5.2 Electron Yield Detection

5.2.1 Principles
The principles of electron yield detection require an understanding of the basic
photoemission events which follow the absorption of an X-ray by a solid. Note
that photoabsorption leading to emission of a photoelectron dominates by
orders of magnitude over all other interactions of soft X-rays (hv < 2 keY) with
matter such as coherent and incoherent scattering (Figs. 5.10, 11). A schematic
 5.2 Electron Yield Detection 119

EFl EV
Binding Energy (eV) 0 Kinetic Energy (eV)

-hv,-- - - - _ Photoemission

A B VB I (a)

(e)
Auger
/ VB
"'z:tz:=p,zt;J/1 :
Y1ield Window s:~~:~ I ' I

Total ~77777177m77==a0=3
Ep Ea

Fig.S.4a-c. Energy level diagram and schematic photoemission spectra at different photon energies
for a sample containing atoms with two core levels A and B and a valence band VB. The energy zero
is chosen at the Fermi level EF , which lies below the vacuum level Ey by the work function rjJ. (a) hv,
below the excitation threshold of core level A, (b) hV2 just above the absorption threshold of shell A
but below its photoemission threshold, (c) hV3 far above threshold of shell A. At each energy the
various photoemission and Auger peaks and their inelastic tails are indicated. At the bottom,
window settings for different electron yield detection modes are shown. In the Auger yield mode the
detector window is set around the energy Ea of the Auger peak. For partial yield detection only
electrons with a kinetic energy in access of Ep are detected while all electrons are collected in the total
electron yield mode

photoemission spectrum of a solid sample is shown in Fig. 5.4. We shall assume

that the sample consists of two kinds of atoms, adsorbate atoms which have a
core level A and a valence level which overlaps with the valence band (VB) of the
substrate, and substrate atoms with a shallow core level B and the VB. The
electron binding energies (Eb) are measured relative to the Fermi level (E F ) of
the substrate, as indicated. The Fermi level also defines the zero of the photo-
electron kinetic energy and is separated from the vacuum level (Ev) by the work
function cp.
As the photon energy is increased from below [hVl < Eb(A)] to just above
the absorption threshold [hV2 > Eb(A)] of shell A the creation of the core hole
results in the appearance of a characteristic Auger peak in the photoemission
120 5. Principles, Techniques, and Instrumentation of NEXAFS

spectrum at a kinetic energy Ea while the photoemission peaks move with the
photon energy. The Auger peak is characteristic of the adsorbate atoms and
originates from the decay of an adsorbate valence electron into the core hole in
shell A with transfer of the decay energy to another adsorbate valence electron.
At the shown energy hV2 the photoelectron excited from the adsorbate shell A is
not free to leave the sample because its kinetic energy is not high enough to
overcome the work function. As the photon energy is increased even more (hv3)
a photoemission peak corresponding to level A is observed while the Auger peak
remains at a fixed kinetic energy. The intensity of the Auger peak, however, will
change with photon energy and follow the X-ray absorption cross section of
shell A. Therefore, by use of an electron energy analyzer we can select a window
and center it at the fixed energy Ea of the Auger peak. The recorded intensity will
directly give the X-ray absorption cross section of the shell A of the adsorbate
atom.
The Auger electron yield (AEY) detection technique is illustrated for CO
adsorbed on Ni(lOO) in Fig. 5.5, taken from the early work of Stohr and Jaeger
[5.22]. Figure 5.5a shows the photoemission spectrum of the sample taken at a
photon energy of 550eV, about 15 eV above the 0 K-edge threshold. It directly

(a)
~ CO on Ni(100)
'c::l
Ni 3p hv=550eV
.c
~ 5.0
iU
c:
Ol
Ui OKVV
Ni VB
c: Auger
e
ti
Q)
2.5
c;;
o
(5
""a. ia

470 490 510

Electron Kinetic Energy (eV)

(b)

3 - - C O on Ni(100) Ni 3p
---Clean Ni (100)
>Q) I
0
U;
Ni VB
(40
w"
"0
c;;
2
I T .... Fig. 5.5. (a) Photoemission spectrum at hv
= 550eV for CO on Ni(I00) showing the 0
:;: KVV Auger peak of CO and the Ni 3p peak and
c: A B

ti
e J
valence band. (b) Auger electron yield NEX-
1 AFS spectrum recorded at grazing X-ray in-
Q)
iii
I
cidence for the CO covered and clean Ni(I00)
----- /
surface. The Auger yield was monitored by set-
ting the electron energy analyzer window at Ea
0
500 520 540 560 580 = 510eV as shown in (a). The spectra are taken
Photon Energy (eV) from [5.22]
 5.2 Electron Yield Detection 121

corresponds to Fig. 5.4b with the Ni 3p and VB representing the two substrate
levels B and VB, and the 0 Is level for level A. For Fig. 5.5b the window of the
cylindrical mirror electron energy analyzer (Fig. 5.7 below) was set around
510eV, the energy of the 0 KVV (K-shell valence-valence) Auger transition,
and its intensity was recorded as a function of photon energy near the 0 K-edge.
Figure 5.4 shows that the Ni 3p and VB photoemission peaks move with photon
energy and at a certain energy will therefore sweep through the detector
window, set at the fixed 0 KVV energy. This effect results in the two prominent
peaks at the lowest and highest energies in Fig. 5.5b. The two peaks in-between,
labelled "A" and "B", are characteristic of the 0 K-shell absorption. This is
proven by their disappearance for the clean Ni(IOO) surface. In fact, the two
peaks are the n* and 0'* resonances familiar from the 0 K -edge spectrum for free
CO shown in Fig. 4.5.
The interference of photoemission peaks with the NEXAFS structure shown
in Fig. 5.5b may be overcome by simply spoiling the energy resolution of the
detector. This approach was used by Wang et al. [5.23] in their NEXAFS study
of ethylene on Pd(III). By applying a triangular modulation of 50 Vat I kHz on
the dc voltage that defined the electron pass energy through the analyzer,
possible photoelectron peaks in the recorded spectrum were averaged out. In
practice, other electron yield detection methods are often used, and these will be
discussed in the following.
A fraction of the Auger electrons from the adsorbate suffer an energy loss
and emerge from the sample with a kinetic energy less than Ea. Because the
primary Auger kinetic energy is independent of hv, so is the energy distribution
of the inelastic Auger electrons shown as a hatched area in Fig. 5.4. Therefore
the inelastic Auger intensity will follow the elastic one. This fact is utilized in the
partial electron yield (PEY) detection variant where only electrons of kinetic
energy larger than a threshold energy Ep are detected (Fig. 5.4). The elastic and
part of the inelastic Auger intensity serve as the signal. By suitable choice of E p ,
one can avoid new photoemission peaks of the substrate entering the kinetic
energy window of the detector over the NEXAFS energy range. Some photo-
emission peaks may already fall into the window at the lowest photon energy of
the NEXAFS scan but this simply increases the background. It is more import-
ant to avoid new peaks entering the window because this leads to NEXAFS
unrelated structures in the spectrum as for the AEY. Because of the flexibility in
choosing E p , the PEY allows one to avoid the interference problems encoun-
tered in AEY detection. PEY measurements are best carried out with a retarding
grid detector, as discussed in detail in Sect. 5.2.4, where a negative grid
potential - E p, is used to prevent electrons with kinetic energy less than Ep from
entering the detector. It is apparent that for the same solid detection angle, the
PEY mode offers a higher count-rate than the AEY mode. However, the signal-
to-background ratio is reduced.
The third, and simplest, detection technique consists of collecting electrons
of all energies from the sample and is referred to as total electron yield (TEY)
detection. The TEY signal is dominated by low energy electrons with kinetic
122 5. Principles, Techniques, and Instrumentation of NEXAFS

energy below about 20eV (Fig. 5.4), the so-called "inelastic tail". A fraction of
the electrons comprising the inelastic tail are inelastically scattered Auger
electrons from the adsorbate and it is this fraction which is responsible for the
desired NEXAFS signal. Although the total count-rate is very large for the TEY
mode, the signal-to-background ratio is typically very small. As a rule of thumb,
for an adsorbate monolayer on a metal substrate, about 1% of the TEY signal
originates from the adsorbate. Figure 5.4 shows that the Is photoemission peak
(peak A) of the adsorbate contributes to the TEY when the photon energy
exceeds the K-shell ionization potential (i.e., the Is binding energy relative to the
vacuum level). In principle, this would lead to a step-like increase in the TEY
about 5 eV above the lowest K -shell absorption structures arising from trans-
itions to molecular bound states or substrate states near the Fermi level. In
practice, no such step is observed because the measured elastic Is photoemission
intensity increases gradually at threshold to a maximum about 20eV above the
IP [5.24] and it is small compared to the total (elastic and inelastic) Auger
electron intensity. Because of the higher kinetic energy of the elastic Auger than
the elastic photoelectrons, the K-shell TEY signal from the adsorbate is always
dominated by the inelastic Auger channel [5.25]. Thus the AEY, PEY and TEY,
except for giving different signal-to-background and signal-to-noise ratios, yield
almost identical NEXAFS spectra.

5.2.2 Quantitative Description of Electron Yield

In practice, the key issue in electron yield NEXAFS spectroscopy is to find
means of suppressing the unwanted background without jeopardizing the signal
from the adsorbate layer, i.e., optimizing the signal-to-background and signal-
to-noise ratios. Figure 5.6a illustrates the problem. Photons as shown in the
figure penetrate into the substrate composed of atoms B to a depth which is
characterized by the photon mean free path and is given by the inverse of the
X-ray absorption coefficient /lx(hv) of the material, also called the linear
X-ray absorption coefficient. This coefficient /lx(hv) [cm -1] is related to the
absorption cross section (Tx(hv) [cm 2 /atom] by the atomic volume density of the
sample nv [atoms/cm 3 ] according to

(5.2)

The volume density ~ is calculated from the mass density nm [g/cm 3 ] by the
relation ~ = nmNo/A, where No = 6.022 X 1023 [atoms/mole] is Avogadro's
number and A [g/mole] is the atomic weight. In the soft X-ray region values for
(Tx(hv) and /lx(hv) have been tabulated by Veigele [5.28] and Henke et al. [5.29].
From these data the photon mean free path 1/ Jlx(hv) in most materials at
hv = 1000 eV is of the order of 1000 A.
Absorption of X-rays leads to the creation of photoelectrons and Auger
electrons. On their way to the surface these electrons are scattered inelastically
by electron--electron and electron-plasmon interactions and quasi-elastically by
 5.2 Electron Yield Detection 123

hv hv (a)

8
t
~=Z¢za.z=~'!lL~~M1~7frz.m- Adsorbate
Atoms A

Electron~
Escape
"
"
Depth
L - 50A. ,
_L ___________ _ Substrate
Atoms B
Photon
Penetration
Depth
sin 81p.- 500 A.

J __________ _
~ 100
;S
'"
"- 50
'"
e
u.
c
'"'"
:2
c
~ 10
u
~
W
5

5 10 50 100 500 1000

Electron Energy above Fermi Level (eV)

Fig. 5.6. (a) Photo absorption and electron production in a sample consisting of substrate atoms B
and an adsorbate layer A. Only electrons created within a depth L from the surface contribute to the
measured electron yield signal. Electrons originating from layer A constitute the NEXAFS signal;
those from layer B give rise to unwanted background. (b) Electron mean free path in solids as a
function of the electron kinetic energy above the Fermi level. The shaded area represents the
distribution typically found for different materials [5.26, 27]

electron-phonon interactions. The relative importance of the scattering mech-

anisms depends on whether the material is a metal, a semiconductor or an
insulator. Nevertheless, the electron scattering length or mean free path as a
function of kinetic energy follows a "universal curve" [5.26, 27, 30] shown in
Fig. 5.6b, where the dashed region indicates the variation found in different
materials. In particular, absorption of X-rays in the range 250eV ~ hv ~ 600eV
containing the C, N, and 0 K-edges leads to primary photoelectrons and Auger
electrons with mean free paths typically less than 10 A. The inelastic scattering of
these primary electrons results in an electron cascade as shown in Fig. 5.6a.
When the cascade reaches the surface only those electrons will escape into
vacuum which have sufficient energy to overcome the surface potential barrier.
This limits the origin of the TEY signal to an effective escape depth L since
electrons generated deeper inside the sample will have insufficient energy to
escape.
124 5. Principles, Techniques, and Instrumentation of NEXAFS

The value of L has been estimated by Gudat [5.31] in the 50-150eV spectral
range to be less than 50 A for metals and semiconductors and by Jones and
Woodruff [5.32] to be 65 A for aluminum metal and 130 A for aluminum oxide
at hv ~ 1600eV. The two latter numbers illustrate the fact that L is typically
longer in insulators (e.g., A1 2 0 3 ) than in semiconductors and metals (e.g., AI)
owing to the reduced electron-electron scattering mechanism at low kinetic
energy in insulators. To understand how the surface signal can be enhanced and
the background signal reduced let us consider their dependence on physical
parameters.
The AEY signal from the adsorbate layer with atomic area density pA,
K-shell cross section (1~, and nonradiative Auger yield w~, emitted into a solid
angle Q, is given by [see (5.1)]

I~(hv) = ~ IoAo(1~(hv)pAw~ . (5.3)

Here we have again assumed that the fraction of X-rays absorbed in the
adsorbate layer is small, i.e., (1~ pA ~ 1. We have explicitly indicated that the
NEXAFS structure arises from photon energy dependent modulations in
(1~(hv ).
Rigorous calculation of the electron yield from the substrate is difficult
because of the complexity of the electronic scattering and cascading processes.
In principle, it would be desirable to write down a general expression for the
electron yield as a function of the photon energy, the electron kinetic energy, the
X-ray incidence angle, and the electron emission angle. The TEY, PEY and AEY
would then be obtained from this expression by integration over the appropriate
kinetic energy range and detector acceptance. Unfortunately, such a general
theory does not exist at present and we shall therefore discuss an approximate
theory for the TEY and use it to explain the surface sensitivity enhancement in
the PEY and AEY detection modes.
For the TEY, we use a simple model that averages over the various
complicated electron scattering processes and describes them by effective para-
meters with physical meaning. Our model follows Fig. 5.6a and describes the
interaction of the X-rays with the bulk sample by an absorption coefficient
J.l~(e, hv) where e is the X-ray incidence angle on the sample measured from the
surface. The so-defined absorption coefficient includes the effects of X-ray
refraction at the surface and is in general obtained from the optical constants
[5.29, 33, 34]. However, in the soft X-ray region at sufficiently large incidence
angles (> 10°) we can neglect reflection and refraction [5.35] and express
J.l~(e, hv) in terms of the tabulated [5.28, 29] absorption coefficient J.l~(hv), see
(5.2), according to

B J.l~(hv)
[Link](e, hv) = -.-e-· (5.4)
sm

Equation (5.4) expresses the fact that absorption is geometrically enhanced, and
 5.2 Electron Yield Detection 125

the beam penetration into the sample shortened, at nonnormal X-ray incidence
angles (Fig. 5.6a). The number of photoelectrons created in the sample at a depth
z, within an increment dz, is given by

N~ = IoAoJl~(lJ, hv)e -1'~(8,hv)z dz . (5.5)

We have neglected the reflection of the X-ray beam at the surface which would
contribute a factor 1 - R(lJ), but at the photon energies and X-ray incidence
angles of interest here, the reflectivity R(lJ) is negligibly small [5.29]. The
primary electrons created propagate to the surface by inelastic scattering
processes and the creation of low-energy secondary electrons which, according
to Fig.5.6b, have the longest mean free paths. We describe the various
energy-dependent electron scattering lengths simply by an effective electron
scattering length L e , which is largely determined by the mean free paths 1; of the
low-energy secondaries according to I/Le = 1:(1/1;) and is, to first order, inde-
pendent of the photon energy. The multiplication and loss of electrons, their
transport to the surface and escape into vacuum is described by a single electron
gain function Ge(hv).
Experiment and theory [5.36-39] both show that for a given material the
shape of low-energy electron distribution ("inelastic tail") is independent of the
primary electron energy once it is higher than about 20eV. The distribution is
well described by the function Ek/(Ek + cjJ )4, where Ek is the kinetic energy
relative to the vacuum level and cjJ is the work function [5.38, 39]. On the other
hand, the number of low-energy electrons increases linearly with the primary
energy, i.e., with the photon energy. We can therefore write for the total gain
Ge(hv) = hvM, where M is a material constant describing the conversion effici-
ency into low-energy electrons. In analogy to the attenuation of X-rays de-
scribed by the exponential factor in (5.5), one can interpret the quantity 1/Le as a
linear electron-attenuation coefficient and mathematically describe the electron
scattering processes as the attenuation of a single electron multiplied by a gain
factor hvM. The primary electron generated in the increment dz inside the
sample then contributes a fraction dI~ to the TEY, emitted into a solid angle D,
given by

D
dI~ = 41t N~ e- z /L hvM . (5.6)

Here we have defined the effective electron escape depth L as the projection of Le
along the surface normal. Integration of(5.6) over the sample thickness gives the
TEY signal generated in the sample. If we assume that the sample is much
thicker than both the X-ray penetration depth sin lJ / Jl:Ahv) and the effective
electron escape depth L, we obtain for the TEY signal of the substrate [5.31]

B D Jl~(lJ, hv)
It (lJ, hv) = 41t IoAo Jl~(lJ, hv) + I/L hvM . (5.7)
126 5. Principles, Techniques, and Instrumentation of NEXAFS

The dependence of the TEY on the X-ray incidence angle (J is contained in the
ratio factor in (5.7).
For the following discussion we shall define for brevity Jl8 =
Jl~«(J, hv) = Jl~(hv)/sin (J. At more grazing X-ray incidence angles (J, the effective
X-ray absorption coefficient Jl8 increases and the ratio in (5.7), which can be
rewritten as Jl8L/(Jl8L + 1), approaches 1 in the limit LJl8 ~ 1. Thus at grazing
incidence angles the TEY is strongly enhanced and it is no longer proportional
to the X-ray absorption coefficient. This has been observed experimentally by
Gudat [5.31] and Martens et al. [5.40]. The physical reason for this effect is that
all X-rays are absorbed in a surface layer with thickness 1/ Jl8 ~ L and the
electron signal saturates. The only energy-dependent modulation of the electron
yield arises from the term Jl~«(J, hv). In the limit Jl8L ~ 1, which is fulfilled at
larger X-ray incidence angles, we obtain Jl8L/(Jl8L + 1) ~ Jl8L and the TEY is
proportional to the X-ray absorption coefficient. This was first pointed out by
Lukirskii [5.41] and Gudat and Kunz [5.42] and has been extensively utilized in
recording EXAFS spectra of bulk materials [5.25,43-45].
For a given substrate, values for Land M can be obtained from measure-
ment of the quantum yield yg, defined as the number of electrons emitted
(0 = 2n) per incident photon. Typically yg is measured at normal X-ray
incidence, (J = 90°, such that Jl8L ~ 1, and with this approximation we obtain
from (5.7)

B l~(hv) 1 B
Yo(hv)=--=-2 Jlx (hv)LhvM. (5.8)
loAo

Henke [5.46] has demonstrated the usefulness of (5.8) for the description of the
photon energy dependence of the quantum yield for several materials. From his
plots we obtain the quantum efficiencies of Au at hv = 300eV to be 6.5 x 10- 2
and 4.5 x 10- 2 at hv = 1500eV. Using tabulated values for Jl~(hv) [5.28] we can
calculate the product LhvM from (5.8) and obtain the values 42A (300eV) and
210A (1500eV). The ratio between the numbers is exactly that of the photon
energies, indicating that the product LM (=0.14A eV- 1 ) is independent of
photon energy, as assumed in the derivation of (5.7). The most direct way to
obtain the conversion efficiency M, and therefore L, is to measure the electron
gain Ge(hv) = hvM of the substrate directly. This is simply done by simultan-
eous measurement of the TEY with a pulse counting electron multiplier and a
current amplifier. Since all electrons from the same primary absorption process
reach the electron multiplier within a time interval ( < 10- 13 s) that is short
relative to the detector response time they will contribute to the same mUltiplier
pulse. The ratio of electron number yield obtained from the current measure-
ment to the pulse number yield is then equal to the number of electrons
produced inside the sample per primary electron, i.e., equal to the gain hv M.
Using the data for Au obtained by Eliseenko et al. [5.47] with AI-K.. (1487 eV)
radiation, Henke et al. [5.38] have obtained the value hvM = 4.3. Thus at
hv = 1487 eV each primary electron produces 4.3 electrons on the average. At
 5.2 Electron Yield Detection 127

hv = 300eV, we obtain hvM = 0.87 and there is no gain but rather a small loss
in the transport and emission process. Combination of these results with the
values for Lhv M obtained earlier gives the electron escape depth in Au as
L = 50 A, independent of hv in the 300-1500eV region. This value is in excellent
agreement with the values found for metals by Gudat [5.31] and Jones and
Woodruff [5.32].
An expression for the PEY signal I:(lJ, hv) from the substrate can be
obtained by suitable modification of(5.7). Because in PEY detection low-energy
electrons are eliminated, the electron gain hvM inside the substrate is equal to 1.
Furthermore, the effective electron escape depth L is greatly reduced according
to Fig. 5.6b such that the approximation f.l8L ~ 1 is valid for typical X-ray
incidence angles (lJ ~ 10 0) and we obtain

(5.9)

It is difficult to write down an expression for the electron signal from the
substrate when the AEY from the adsorbate is monitored. This would require
calculation of the inelastic photoelectron signal of the substrate at the energy of
the adsorbate Auger peak (Fig. 5.5a). As a rule of thumb, one would expect it to
be less than half of the PEY signal because the elastic photoemission peaks of
the substrate do not directly contribute (Fig. 5.5).

5.2.3 Adsorbate Versus Substrate Signal

We are now in a position to estimate the signal-to-background ratio SB obtained
in an electron yield measurement of a chemisorbed molecular layer on a
substrate. We assume that the adsorbate signal is given by the AEY (5.3). This
neglects any secondary electrons created by scattering of elastic Auger electrons
(Fig. 5.4) and therefore underestimates the adsorbate signal. We could compen-
sate for this loss by introducing a gain factor for the adsorbate signal which,
similar to that of the substrate, would be proportional to hv. It is simpler,
however, to account for this effect by eliminating the gain factor hvM from the
expression for the substrate signal, instead. Furthermore, we assume that
f.l8L ~ 1, which is valid for measurements at larger X-ray incidence angles
(Fig.5.6a). Both PEY and TEY signals are then described by (5.9), but with
different escape depths L, and the signal-to-background ratio is given by

S _ (1~(hV)pAQ)~
(5.10)
B - f.l~(lJ, hv)L

With the values appropriate for a monolayer of C atoms at hv = 300eV,

(1~ = 9 x 10- 19 cm 2 /atom [5.28], pA = 10 15 atoms/cm 2 , and Q)~ = 1 and those
for a Au substrate at the same energy f.l~ = 3 X 10- 3 A -1 [5.28], and L = 50A
we obtain SB = 6 X 10- 3 for TEY detection. For PEY detection L is shortened
128 5. Principles, Techniques, and Instrumentation of NEXAFS

to about S A according to Fig. S.6b and therefore SB is increased by an order

of magnitude.
The above estimates are for a Au substrate, which has a large absorption
coefficient, and therefore electron yield, in the 300-1000eV spectral range. For
example, relative to Au the absorption coefficients for other substrates at 300eV
are lower by factors 4.5 (Si), 3.S (Ag) and 2.2 (Cu) [S.28, 29] and therefore larger
signal-to-background values would be observed. We therefore expect typical SB
values of 1-2% for TEY, 10-20% for PEY, and > 20% for AEY measurements
for a chemisorbed monolayer. These numbers are in good accord with experi-
mental observations of typical signal-to-background ratios or edge jump ratios
in SEXAFS spectra of chemisorbed atoms [S.4S].
From (S.10) the dependence of SB on photon energy is determined by the
ratio (J~(hv)/ Jl~((}, hv). Since both quantities decrease with photon energy one
would expect to measure approximately the same SB values for carbon (K -edge:
28SeV) and sulfur (K-edge: 2470eV) monolayers on the same substrate. This is
in fact the case [S.4S] and, in retrospect,justifies the neglect ofthe gain factor in
(S.10). If we had not eliminated it, the increase in hvM from C (0.87) to S (7.5)
would have resulted in an unrealistically small SB value for a chemisorbed S
layer. This indicates that in reality both the substrate and the adsorbate TEY
signals increase with hv because of the contributions from their low-energy tails.
Another method which allows further reduction of the bulk background is
based on geometry and consists of detecting electrons at grazing emission
directions from the surface. Full appreciation of this effect requires an under-
standing of electron refraction at the surface. To escape into vacuum, an electron
has to overcome a surface potential barrier. The electron travelling in a free
electron band state crosses the surface conserving the parallel wave vector
component while the component normal to the surface alters because of the
kinetic energy required to overcome the barrier [S.48]. The refraction at the
surface can be expressed in terms of the surface barrier Vo, i.e., the energy
difference between the vacuum level and the bottom of the free-electron-like
conduction band, and the kinetic energy Ek of the electron measured from the
vacuum level. Note that Vo is not the work function, but the work function plus
the separation of the Fermi level from the bottom of the free-electron-like bands.
Electrons at the Fermi level already have kinetic energy and the work function is
simply the minimum energy needed to excite these electrons into vacuum.
Electrons approaching the surface at an angle IX from the inside surface normal
are refracted into vacuum at an angle (j from the outside surface normal
according to [S.49]

sin (j = (
E
k
+ V.
Ek °)1/2 sin IX • (S.11)

At a given kinetic energy only electrons which approach the surface inside
the crystal within a momentum cone (often called a "primary Mahan cone")
[S.48] centered around the surface normal can escape; those outside the cone do
 5.2 Electron Yield Detection 129

not have enough perpendicular momentum and are reflected back into the solid.
The maximum cone angle IXmax> corresponds to sin!5 = 1 in (5.11), i.e., when the
escaping electron emerges from the surface at !5 = 90°. The largest refraction
angle is 45° and occurs for c5 = 90° and Ek = Vo. For example, Williams et al.
[5.49] showed that for Cu Vo = 14eV, such that at Ek = 100eV the maximum
cone angle is IXmax = 70 ° and therefore refraction angles are as large as 20°, even
at this high kinetic energy.
Applying (5.11), we can now express the effective electron escape depth
Le cos IX inside the crystal as a function of the measured angle c5 outside the
crystal and obtain

L = Le ( Ek cos 2 !5 + Vo )1/2 (5.12)

Ek + Vo Ek + Vo
For PEY and AEY detection, only high energy electrons are detected and with
Ek ~ Vo refraction at the surface can be neglected. In this case the effective
escape depth L is reduced by a factor cos IX ~ cos!5 and therefore SB (5.10), is
increased by a factor 1/cos!5. This surface enhancement by grazing incidence
electron detection has in fact been extensively utilized in X-ray photoelectron
spectroscopy [5.50].
Finally, the background from the substrate can be reduced by proper choice
of the detector position relative to the electric field vector E of the polarized
X-ray beam. As discussed in detail in Chap. 9, synchrotron radiation is approxi-
mately linearly polarized with the E vector in the horizontal orbit plane of the
storage ring. By Yang's theorem [5.51], dipole excitation of a photoelectron in a
randomly oriented system leads to an angular distribution of the form

da(Ek' Y)
dQ
= a(Ek)
4n
(1 + peEk) (3
2
2
cos y
_ 1)) , (5.13)

where y is the angle between E and the electron emission direction, and Pis the
photoelectron asymmetry. Measurements and calculations of the energy de-
pendence of the photoemission cross section a(Ek) and peEk) are fairly abund-
ant for gases [5.52] and, depending on the atomic level and energy, peEk) varies
between -1 and 2. It has been recognized by Davis et al. [5.53] that photo-
electron asymmetry effects also exist for solid samples and, most importantly,
that in all studied cases the measured angular distributions showed a tendency
toward high asymmetry, characterized by P= 2. If we accept this result, (5.13)
becomes

(5.14)

and the photoemission intensity is peaked in the direction of E. This has

important consequences since by choosing a detector acceptance geometry 90 °
from the E vector, most of the photoemission intensity is eliminated. Since the
130 5. Principles, Techniques, and Instrumentation of NEXAFS

photoelectron signal constitutes a large fraction of the total electron intensity

from the substrate the background can be significantly reduced using the
photoelectron asymmetry effect.

5.2.4 Experimental Details and Detectors

A typical experimental arrangement for electron yield NEXAFS studies is
shown in Fig. 5.7. The X-ray beam from the monochromator is first collimated
by a set of moveable blades and then falls onto a metal grid with about 100
wires/cm and a transmission around 80%. The TEY signal from this grid,
amplified by a channeltron or spiraltron [5.54, 55] electron multiplier, serves as
a dynamic reference monitor of the X-ray intensity. The grid can be coated in
situ by evaporation of a suitable metal which does not have any absorption
edges in the energy range of interest. When the signal from the grid is only used
to normalize out fluctuations in the X-ray intensity emitted by the storage ring
(and not the energy dependent transmission function of the beam line optics, see
Sect. 5.5) it is often convenient and advantageous to simply use a "dirty" gold
grid. Because of its noble metal character, gold is typically covered with only a
monolayer of carbon and oxygen adsorbates whose K-edges are barely visible in
TEY detection. The stability of such a grid with respect to time coupled with its
high and rather smooth quantum efficiency make it an ideal reference monitor in
the soft X-ray region.

.....-t:.,
X-Ral/s......

~EII
~ Electron Multipl ier
II fEvaporator
Partial Yield
Detector

Fig. 5.7. Experimental arrangement for electron yield NEXAFS studies. The elliptically polarized
X-ray beam from the monochromator, with the major electric field vector component Ell in the
horizontal plane, is collimated and then transverses an in situ coatable metal grid which, in
conjunction with an electron multiplier, serves as a dynamic reference monitor. Electrons from the
sample are detected either by an electron energy analyzer such as the shown cylindrical mirror
analyzer (eMA) or a partial yield detector. This detector is shown in more detail in Fig. 5.9 and can
also be used for total electron yield detection by proper bias of its two metal grids. A phosphor
coated screen in the beam path serves for alignment purposes
 S.2 Electron Yield Detection 131

...
f
]:.~;:.~~~
Metal
To
Scaler

Grid

HV Power Supply
Fig. 5.8. Measurement scheme for total yield collection from a metal grid reference monitor.
The electron signal is amplified by a channeltron electron multiplier which is operated at voltages
VI"" + SOV and V2 "" + 2000 V, respectively. The amplified electron output current is collected by
a collector plate kept at V3 > V 2 by means of a high voltage battery box (2 kV ::::; V3 ::::; 3 kV). The
negative side of the floating battery box is connected to a current amplifier capable of monitoring
currents in the 10- 1°-10- 8 A range. Avoltage (0-10 V) signal proportional to the input current is
then fed into a voltage-to-frequency converter which is read by a computer via a data scalar

The electronics associated with the measurement of the TEY current from
the grid is shown in Fig. 5.8. Electrons from the grid, kept at ground potential,
are pulled into the channeltron cone by a small positive voltage (V 1 ~ 50 V).
Assuming a channeltron gain factor of 106 , one incident electron per second
results in a channeltron output of 106 e1ectrons/s or a current of 10- 13 A. Thus
current measurement techniques can conveniently be used. This is accomplished
most easily by use of a floating battery box which supplies a low-noise positive
bias potential (V 3) in the 2-3 kV range to the collector of the channeltron. The
high potential is generated by a series of ~ 300 V batteries which are suitably
insulated from each other and the battery box housing and soldered together to
eliminate noise. The achievable noise level is limited by leakage currents and is
about 10- 11 A. The negative side of the battery string is connected to the input
of a current amplifier (and therefore is close to ground potential) which gener-
ates an output voltage proportional to the input current. The signal (voltage) is
then converted to a frequency and fed into the scaler of a computer.
The signal from the sample is detected either by an electron energy analyzer
such as the often used and commercially available cylindrical mirror analyzer
(CMA) [5.56] or by a simple partial yield detector (PYD). Other electron
analyzers have been reviewed by Smith and Kevan [5.57]. As shown in detail in
Fig. 5.9, a PYD is easily built and assembled using two high transmission metal
grids for retardation and a double channel plate assembly for electron multipli-
cation [5.54]. Typically the first grid is operated at ground potential and the
second grid at a retardation voltage - Ep where Ep is the PEY cutoff energy
defined in Fig. 5.4. Suitable cutoff energies for PEY studies of low-Z atoms are
~ 180-230eV for carbon, ~ 290-340eV for nitrogen, and =::: 430-480eV for
oxygen, such that the corresponding Auger peaks around 260eV (C), 370eV (N),
132 5. Principles, Techniques, and Instrumentation of NEXAFS

Portia I [telCtron Yield [Link] Fig. 5.9. Assembly drawing of a double-channel plate

©
partial electron yield detector. All components are as-
Coverplote
sembled in the depicted order in the ceramic housing

g Compression
Spring
shown at the bottom. The electrons entering the detector
first traverse a metal grid typically kept at ground poten-

© Ceramic
Spacer
tial (V, = 0), and low energy electrons are then elimina-
ted by a second grid kept at a negative retarding voltage
v,
~ ht Metal Grid (V 2 ) . Typical retarding voltages are -2OOV for carbon,

©
- 320 V for nitrogen, and -450V for oxygen K-shell
CClromlc
Space,. measurements. The electron signal is amplified by a dual
channel plate arrangement. The channel plate voltages are
v,
~ 2nd [Link] Grid supplied by metallic ring contacts. Typical values are

©
V 3 ", + 30V, V 4 ", + 1000 V, and V s '" +2000V. The dual

--
Ceramic
Spoc.,.. channel plate arrangement operates at a total gain of
v,
©' Ring Contact

1sf Channel Piol.

10 7_ 108 and the electron output is collected by a collec-
tor which is connected for current measurement to a
floating battery box as shown in Fig. 5.8. The collector
voltage V6 is kept somewhat ('" + 100 V) larger than VS
V.
©b Ring Contoct

2nd Channel Plat.

~
V. RIng Contact

© Ceramic
Spacer

V.
£::) [Link]

ffi Ceramic
HaUling

and 510eV (0) fall in the detection window. By using a slightly positive bias
potential on the two grids ( ~ 10-20 V), the PYD can also be used for total
electron yield measurements.
The acceptance angle of PYDs can be made quite large, of the order of
10-20% of 4n sr, either by positioning it close to the sample or by using large
channel plates. In comparison, the acceptance of the CMA is about 7% of 4n sr.
The detector efficiency, defined as the ratio of the number of detected to incident
electrons, is determined by the electron optics only, since typical electron
multipliers have unit detection efficiency for electrons with kinetic energies in
the hundred eV range. PYDs typically have ~ 80% grid transmission and
therefore detection efficiency. For a double pass CMA the overall efficiency is
about 5% [5.56]. Typically the output signal from the CMA or the PYD is large
enough that current measurement techniques can be employed (Fig. 5.8).
The phosphor covered screen (Fig. 5.7) is a very useful device for alignment
of the X-ray beam in the chamber and for alignment of the sample in the beam.
In particular, at grazing X-ray incidence a slight misalignment of the beam will
cause it to generate a signal from the edge of the sample, which in a polarization
 5.3 Fluorescence Yield Detection 133

dependent study may lead to an incorrect determination of the molecular

geometry on the surface. Since the front surfaces of typical single crystal
substrates are polished, they reflect the soft X-rays at small incidence angles and
the reflected beam can be monitored on the phosphor screen for alignment
purposes. Also one may simply watch the shadow image of the sample in the
beam spot on the phosphor screen.

5.3 Fluorescence Yield Detection

In considering the detection of the fluorescent X-ray signal from a thin adsorb-
ate layer on a bulk substrate it is important to understand all interactions of the
incident X-rays with the sample that result in "secondary" photons. These are (1)
X-ray absorption followed by fluorescence decay, (2) coherent and incoherent
scattering of X-rays from the bulk substrate, (3) specular reflection of X-rays,
and (4) diffuse scattering by surface irregularities. In a theoretical treatment it is
convenient to discuss separately bulk-dominated effects such as X-ray absorp-
tion and scattering versus "surface" effects such as reflection and diffuse
scattering. This is done below.

5.3.1 Absorption and Scattering of Soft X-Rays

The relative cross sections of X-ray absorption and scattering, taken from the
work of Hubbell et al. [5.58], are shown in Fig. 5.10 for a copper (Z = 29) target.
The dominant interaction of soft X-rays with matter is by means of photoelectric
excitation, i.e., X-ray absorption, whose cross section U x is orders of magnitude
larger than the coherent scattering cross section ues • In a coherent scattering

Cu (Z=29)

Fig. S.10. X-ray cross sections for differ-

----- ......,
Des

,, ent interactions with a copper (Z = 29)

atom [5.58]. (f x is the photoelectric or X-
ray absorption cross section, (f" is the
-~.
coherent (Rayleigh) scattering cross sec-
,/ \~
Oil '\ tion, (fi. is the incoherent (Compton)
scattering cross section, (f ""ir is the cross
/ section for pair production, and (fn",,' is
the nuclear photoabsorption cross sec-
tion. The total cross section (ftot is the
Photon Energy (eV) sum of all others
134 5. Principles, Techniques, and Instrumentation of NEXAFS

event, also called Rayleigh or elastic scattering, a photon is scattered by an atom

without change of its wavelength. Incoherent scattering, also called inelastic or
Compton scattering, is characterized by a change in X-ray wavelength and is
very weak in the soft X-ray region with a cross section a i • that is orders of
magnitude smaller than a c•• Therefore we can neglect it altogether. We note,
however, that at high X-ray energies inelastic scattering becomes relatively large
(Fig. 5.10), and "photon energy loss" or "X-ray Raman" spectroscopy has in fact
been used to record the K-shell excitation spectrum of diamond [5.59].
The maximum K-shell X-ray absorption cross sections of the low-Z elements
in the 200-2500eV range and the energy-dependent X-ray absorption cross
sections of copper and gold, taken from the tables compiled by Veigele [5.28],
are plotted in Fig. 5.11. For the low-Z elements important data for K-shell
excitations are summarized in Table 5.1, such as the X-ray absorption cross-
section above (a::,aX) and below (a::'in) the K-edges, the edge jump ratio
J R = a::,ax/a::,in and the fluorescence yield wf'
Veigele [5.28] also lists coherent scattering cross sections in the soft X-ray
range but they have been derived by extrapolation and are not very accurate. In
particular, they neglect the strong dispersion effects (anomalous scattering)
occurring near the various absorption edges in the soft X-ray region. The
coherent scattering cross sections can be calculated, however, from the real part
11 and imaginary part 12 of the complex atomic scattering lac tors 1=11 + iI2
tabulated by Henke et al. [5.29] for the elements in the 100-2000 eV region.

Fig. 5.11. Soft X-ray photoelectric

(absorption) and coherent scattering cross
sections for selected elements. The photo-
electric cross sections for the K -shells of
the low-Z atoms boron (Z = 5) through
sulfur (Z = 16) are taken from the tables of
Veigele [5.28]. For the low-Z atoms the
maximum cross sections just above the K-
edge are plotted. For carbon, fluorine,
magnesium and phosphorus the jumps
/ / - - - - - - - - - - . . . . Au through the K-edge are also shown. The
coherent scattering (Rayleigh) cross sec-
I
tions for carbon, copper and gold were
~r-CU calculated by means of (5.16) using the
atomic scattering factors tabulated by
Henke [5.29]. Note that the so-obtained

"' ir'-'-'-'-'-'-'-
K-edge I cross section for copper agrees in magni-
C tude with that plotted in Fig. 5.10 but in
addition exhibits an anomalous dispersion
1011::!-.L-I-.I...I:-!+.J...L-!-::,:~~,*:!~~~L.J....~
1000 1500 2000 2500 effect near the L3 (-930eV) and L2
Photon Energy (eV) (-950eV) edges
 5.3 Fluorescence Yield Detection 135

Table 5.1. Cross section, edge jump ratio, and fluorescence yield for K-shell excitation of low-Z
atoms

Atom 2 max"
(5,
min"
(5, JR JR WI Wft1~ax

[10 5 b] [10 5 b] theory' exp.b [x 10- 2 ], [Wb]

B 5 14.7 0.66 22.3 28.5 0.17 2.5

C 6 9.8 0.51 19.2 24.3 0.28 2.7
N 7 6.9 0.39 17.7 21.3 0.52 3.6
0 8 5.1 0.33 15.5 19.1 0.83 4.2
F 9 3.9 0.28 13.9 17.4 1.3 5.1
Ne 10 3.2 0.24 13.3 16.0 1.8 5.8
Na 11 2.9 0.20 14.5 14.9 2.3 6.7
Mg 12 2.4 0.17 14.1 13.9 3.0 7.2
Al 13 2.0 0.15 13.3 13.1 3.9 7.8
Si 14 1.7 0.14 12.1 12.4 5.0 8.5
P 15 1.5 0.13 11.5 11.8 6.3 9.5
S 16 1.3 0.12 10.8 11.3 7.8 10.1

• [5.28]
b Calculated by use of the empirical formula JR = (125/2) + 3.5, given in [5.28]
, [Ref. 5.12, Table 3]

Using the relationship A = 2nhc/(hv) we obtain the relation

A A _ 12398.52
[ ] - hv[eV] , (5.15)

linking X-ray energy hv (units: eV) with X-ray wavelength A (units: A). Therefore
in the soft X-ray region the wavelength is much larger than the atomic diameter.
In this long wavelength limit, the cross section for coherent scattering from an
atom with scattering factors II and 12 is given by

(5.16)

where re = e 2 /mc 2 IS the classical electron radius and 8nr; /3 =

0.665 x 1O-24 cm 2 is the Thomson cross section. Equation (5.16) has been
derived from the basic theoretical expressions given by Warren [5.60] and is
valid for linearly polarized as well as unpolarized light, see (5.30). We have
plotted G"cs for carbon, copper and gold in Fig. 5.11 to demonstrate its depend-
ence on Z and for comparison with the respective X-ray absorption cross
sections. The elastic scattering cross sections are only plotted up to 2000eV
since Henke's tables do not list values for II at higher energies. Also, at higher
energies the long wavelength approximation assumed in the derivation of (5.16)
is no longer valid [5.29, 60, 61].
As discussed above, the coherent scattering cross section calculated from
(5.16) includes dispersion effects and therefore differs in its energy dependence
136 5. Principles, Techniques, and Instrumentation of NEXAFS

from that typically found in the tables [5.28]. This can be seen by comparison of
the copper cross sections (leo plotted in Figs. 5.10 and 5.11. Note, however, that
there is good agreement of the magnitudes in nonresonant energy regions. The
coherent scattering cross section varies significantly with Z and is about two
orders of magnitude higher for gold than for carbon. The maximum K -shell
photoelectric cross section decreases by about an order of magnitude between
boron (Z = 5) and sulfur (Z = 16) but it is still almost two orders of magnitude
higher for sulfur than the largest coherent scattering cross section value, that of
gold. It should be noted that the X-ray absorption cross section (Ix is related to
the imaginary part of the atomic scattering factor according to

(5.17)

where C = 2rehc = 7.0 x 10- 17 eV cm 2 . The (Ix values obtained with the f2
values of Henke et al. [5.29] are in good agreement with those tabulated by
Veigele [5.28]. Using the cross sections plotted in Fig. 5.11 we can now calculate
the fluorescent and scattered X-ray signals from the substrate atoms.
The theoretical description of the fluorescence yield from the bulk substrate
is somewhat simpler than that of the electron yield because X-rays interact less
strongly with matter than electrons. Hence the complicated inelastic scattering
processes of the photoelectrons on their way out of the substrate are nonexistent
for the fluorescent photons, or at least so small that they can be completely
neglected. We only need to take absorption of the fluorescent radiation into
account. In analogy to (5.6), we can write for the fluorescence signal dl' created
in the substrate at a depth z, within an increment dz,

d/B(hv)
f
= !!..- I A IIB«() hV)WB e -I'~(/J,hv)z e -
4n 0 0 r x ' f
zID(Ej) dz
'
(5.18)

where we have assumed that the fluorescence yield from the substrate can be
characterized by its strongest component w"
which occurs at an energy Ef'
D(Ef ) is the effective X-ray escape depth at the fluorescent energy, measured
along the surface normal. If the fluorescent radiation is detected at an angle h
from the surface normal, the X-ray escape depth is related to the X-ray
absorption coefficient according to D(Ef ) = cos h/ flx(Ef)' By integration over the
sample thickness, and assuming that the sample is much thicker than both
1/ fl~«(), hv) and D(Ef)' we obtain

B Q fl~«();hv)w'
If«(),hv) = 4n loAo fl~«(),hv) + l/D(Ef ) . (5.19)

The coherent scattering contribution from the substrate is derived in a

similar fashion to the fluorescence contribution, considering an X-ray absorp-
tion coefficient fl~«(), hv) en route to the scattering center inside the sample, an
effective X-ray escape depth D(hv), and integrating over an infinitely thick
 5.3 Fluorescence Yield Detection 137

sample. We obtain

B Q CT~.n~
Ic.«(J,hv) = 4n 10Ao Ji~«(J,hv) + I/D(hv) , (S.20)

where n~ (atoms/cm 3 ) is the volume density of the substrate and CT~. is given by
(S.16).
It is important to note that in the soft X-ray region Bragg scattering typically
does not exist. From the Bragg equation A = 2d sin (J, we know that the longest
Bragg-scattered wavelength is given by Amax = 2d, and since almost all natural
crystals have d-spacings ofless than 2 Awe see that Amax ~ 4 A, corresponding to
a minimum photon energy for Bragg scattering of about 3000eV.

5.3.2 X-Ray Reflection and Diffuse Scattering

In detecting photons from the sample one also needs to consider the intensity of
the reflected X-rays. Although the specular reflectivity is very low for incidence
angles larger than 10 typically used in NEXAFS studies, the high incident flux,
0

of the order of 10 10 photons/s, leads to a significant reflected signal which, when

directed into the detector, may dominate the fluorescence signal. Below we
consider the size of the angle-dependent X-ray reflectivity of a metal substrate in
the 2S0-2000eV spectral range.
The X-ray reflectivity of a material is calculated by means of the Fresnel
equations. We shall use here a formalism developed by Stern [S.62] and Heavens
[S.63] for the reflectivity at the interface between two homogeneous layers. By
defining (J as the grazing angle of incidence from the surface, ko, the component
of the incident wavevector perpendicular to the surface (in vacuum) is given by

2n. (J
ko =ysm (S.21a)
,

and k1 (the tilde denotes complex numbers), the wave vector component perpen-
dicular to the surface in the substrate, can be calculated according to

k-1 = Y2n J-8 1 - cos 2(J . (S.21b)

Here Ais the wavelength of the incident radiation in vacuum and 81 the dielectric
constant of the substrate, which can be calculated from the atomic scattering
factors 11 and 12 [S.29] according to

8-1 -_1_
re A2nV(r
)1 - 1·f)
2 . (S.22)
n
where re = 2.82 x 10- 13 cm is the classical electron radius and nv is the number
of atoms per unit volume in the substrate with atomic scattering factors 11
and 12.
138 5. Principles, Techniques, and Instrumentation of NEXAFS

We next consider the specular reflectivity. We denote as rm the reflection

coefficient at the interface between vacuum and substrate. The polarization
direction of the electric field vector E is characterized by the label m. For s-
polarized light, m = s, and E is perpendicular to the plane of incidence, i.e.,
parallel to the surface of the specimen, while for p-polarized light m = p, and E
lies in the plane of incidence. The Fresnel equations then define the polarization
dependent reflection coefficients at the vacuum-substrate interface in terms of
the wave vector components ko and k1 and the dielectric constant 81 according to

(5.23)

and
_
=
81 ko - k1
_ .
rp (5.24)
61 ko + k1
The measured specular reflectivity is obtained as
(5.25)

where Ro((J) is the Fresnel reflectivity of the perfectly smooth surface and the
asterisk denotes the complex conjugate. The exponential term accounts for a
finite surface roughness characterized by a root mean square (RMS) roughness ()
perpendicular to the surface [5.64].
Calculated reflectivities Ro((J) for gold and copper at 277 eV(C K Il ) and
2042 eV(Zr LI%) for an s-polarization geometry are shown in Fig. 5.12. For p-
polarization, the reflectivity near the extreme incidence angles (lJ = 0° or 90°) is
identical to that in s-polarization but exhibits a pronounced minimum at the
Brewster angle near lJ = 45 ° [5.65]. In particular, for linearly polarized radi-
ation the Fresnel equations show that the p-reflectivity is equal to the square of
the s-reflectivity at lJ = 45° and therefore extremely small. Figure 5.12 also
illustrates the dramatic effect of surface roughness by showing a reflectivity
curve calculated for a gold surface with () = 20 A vertical roughness at 277 eV.
This curve drops by 7 orders of magnitude as the X-ray incidence angle from the
surface is increased to 30°.
Elastic scattering processes occur not only in the bulk but also at the surface.
A real surface can be described by a distribution of structures with aRMS
roughness () perpendicular to the surface and a lateral correlation length A
characterizing the average separation of the structures parallel to the surface
[5.64]. This surface roughness influences the reflectivity when the phase differ-
ence of beams reflected at different points of the surface becomes large. In this
case the specularly reflected fraction R(lJ) is reduced according to (5.25) and an
"incoherent" fraction, called diffuse scattering. is created. The scattered radiation
is spatially incoherent but temporally coherent, i.e., it is of the same wavelength
as the incident radiation. If we neglect absorption by the surface, the diffuse
 5.3 Fluorescence Yield Detection 139

Fig. 5.12. Calculated reflectivities Ro(lI)

z for perfectly smooth copper (dotted curves)
and gold (solid curves) surfaces at 277 eV
(carbon K.) and 2042eV (zirconium L.).
The assumed geometry (s-polariz-
y
ation) is shown as an inset. Also shown
is a curve calculated with (5.25) for
hv = 277 eV assuming a gold surface with
a RMS vertical roughness () = 20 A. Here
K = 41t sin 11/). is the momentum transfer
upon reflection

--Au
.......... Cu

o 10 20 30 40 50 60 70 80 90
Incidence Angle 8 (deg)

scattering intensity integrated over all angles, called total integrated scatter, is
simply obtained as the difference of the intensity reflected by the perfectly
smooth and the rough surface, and is given by [5.66]

(5.26)

5.3.3 Adsorbate Fluorescent Signal and Substrate Background

Because of the low scattering cross section for low-Z atoms we can neglect the
elastically scattered radiation from the adsorbate layer. The X-ray fluorescence
signal INhv) from the adsorbate is given by an expression similar to (5.3) for the
AEY signal,

It(hv) = ~ IoAoO'~(hv)pAOJt . (5.27)

Again we have assumed the small concentration limit, i.e., O'~ pA ~ 1.

The background signal I back arises from the fluorescent signal of the sub-
strate and from the reflected and scattered X-rays according to

(5.28)

Here the reflectivity R(lJ) is zero except at the specular angle, Ids is the diffuse
scattering intensity (5.26), I~s is the elastically scattered intensity from the bulk
(5.20), and Ir is the fluorescent intensity from the bulk (5.19).
140 5. Principles, Techniques, and Instrumentation of NEXAFS

In order to understand the relative size of the various background signals in

comparison to the fluorescence signal from the adsorbate let us estimate them
assuming that all photons from the sample are detected without energy or
angular discrimination. Also, since it is necessary in practice to measure angle-
dependent NEXAFS spectra we shall carry out a worst case analysis and
consider the X-ray incidence angle that gives rise to the largest background
signal. We assume a collection angle Q = 2n and an incident photon flux
IoAo = 2 x 10 10 photons/s and consider the case of a monolayer
(pA = 10 15 atoms/cm 2) of carbon atoms on a copper surface. Below we evaluate
in tum all relevant intensities in (5.27) and (5.28) at a photon energy of
'" 300eV.
(1) It(hv): Using the X-ray absorption cross section O'~ = 10- 18 cm 2/atom
from Fig. 5.11, and wt = 3 x 10- 3 from Fig.5.2b, we obtain It = 3 x 104
photons/s for the fluorescence signal from the adsorbate.
(2) IoAoR(O): The specularly reflected background signal from the sample
is largest at grazing X-ray incidence and at the smallest X-ray incidence angle
'" 10 ° typically used in NEXAFS studies the reflectivity of the perfectly smooth
copper surface is Ro(100) ~ 10- 1 according to Fig. 5.12. Assuming aRMS
surface roughness of () = 20 A we can calculate the intensity in the reflected
beam with (5.25) to be IoAoR(100) ~ 6 x 108 photons/so
(3) Ids: Using the same parameters, we can estimate the diffuse scattering
intensity from (5.26) as Ids ~ 7 X 108 photons/so
(4) I~s: The coherently scattered intensity is obtained from (5.20). We
approximate the X-ray escape depth as l/D(hv) = {lAhv)/cos () ~ 2{lx(hv) and
with (5.2) and 0 = 90° we obtain I~s = (0'~s/30'~) X 10 10 photons/so Using the
cross sections for copper plotted in Fig. 5.11 yields I~. ~ 5 X 105 photons/so
(5) Ir: Using the same approximation for D(er) and approximating the
photon energy dependence of the absorption coefficient according to
{lx(6 c) ~ (hv/er)3 11x (hv) we obtain from (5.19) Ir ~ {wd[l + 2(hv/er)3
sin OJ} x 10 10 photons/so Around hv = 300eV only the outer shells of typical
substrate atoms are excited and fluoresce, and their fluorescence yields are very
small ('" 10- 5 ) [5.12, 13Y For copper the largest fluorescence yield ofthe outer
shells is that of the M 3 shell at 75 eV binding energy2, and with Wc = 5 x 10- 5 we
obtain the background intensity at 0 = 10 ° to be Ir ~ 2 x 104 photons/so
The above comparison predicts a signal-to-background ratio which, taken at
face value, would render fluorescence detection impractical if not unfeasible. In
particular, the specularly reflected and diffusely scattered intensities are about 4
orders of magnitude larger than the fluorescence signal from the adsorbate layer.
Fortunately, there are experimental remedies which reduce the background
signal by taking advantage of its angular and/or energy dependence.

2 Calculated fluorescence yields for the M and N shells of copper and gold are [5.67]: Copper: M 1
(123eV): 1.2xlO- s , M2 (77eV): 4.1 x 1O-S, and M3 (75eV): 5.1xl0- s . Gold: N4 (353eV): 1.5
x 10- 4 , Ns (335eV): 1.7 x 10- 4 , N6 (88eV): 3.0 x 10-5, and N7 (84eV): 3.5 x lO- s .
 5.3 Fluorescence Yield Detection 141

The conceptually simplest scheme, successfully applied in the hard X-ray

region [5.68, 69], is energy discrimination. The fluorescence radiation from the
adsorbate is emitted below the K-shell threshold, at an energy Be. For example,
for carbon Be ~ 277 eV, compared to a typical K-shell threshold of 285 eV. For
nitrogen (Be ~ 392) eV and oxygen (Be ~ 525) the fluorescence energy is also
about 8 eV below the respective K -shell thresholds of 400 eV and 533 eV. The
fluorescent radiation can therefore be isolated by means of a suitable mono-
chromator from the fluorescence radiation of the substrate atoms and from the
reflected and scattered radiation which has the same energy hv as the incident
radiation. In practice, this elegant scheme is impeded by low signal rates arising
from the restricted acceptance and low throughput of the analyzing mono-
chromator. For this reason this scheme has not yet been utilized for NEXAFS
studies of chemisorbed molecules.

5.3.4 Practical Scheme for Suppression of Background Signal

For practical purposes a useful detection scheme for fluorescence NEXAFS
studies consists of a high quantum efficiency detector which offers sufficient
energy resolution (-100-3OOeV) to eliminate unwanted fluorescence radiation
from the substrate and which is suitably positioned relative to the incident beam
and the sample in order to minimize the background signal from the substrate.
Wire proportional counters or doped semiconductor [e.g., Si(Li)] detectors
fulfill these requirements as discussed in more detail in Sect. 5.3.5.
Assuming that such detectors are available, the most important question is
how the extremely large background signals arising from the reflected and
diffusely scattered intensities can be reduced. Elimination of the reflected signal
is simply accomplished by placing the detector so that it does not see the
specularly reflected beam, e.g., underneath the sample (see Fig. 5.7). This
detector position also eliminates most of the diffusely scattered intensity. It is
important to understand this effect in more detail.
The diffuse scattering intensity Ids is peaked at the specular angle but extends
over the entire angular range. Unfortunately, very few experimental investiga-
tions of the angular dependence of the diffuse scattering intensity exist in the soft
X-ray re~ion [5.70-73]. An illustrative example at the long wavelength end,
A. = 100 A, is shown in Fig. 5.13, taken from Hogrefe'S thesis [5.70]. Here the
specularly reflected coherent intensity and the broad underlying diffuse intensity
was measured [5.70] for a 500 Athick gold film on a glass substrate which was
roughened prior to the film deposition by a polish with a 1 Jl.m diamond paste.
The experimental geometry is shown in Fig. 5.13a. The experimental spectrum,
shown as dots in Fig.5.13b, was measured in an s-polarization geometry
(4J = 0°) at an incidence angle (Jl = 10 ° from the surface. The spectrum was
normalized to the incident intensity and normalized to the Fresnel reflectivity
Ro. Therefore the reduction of the measured peak reflectivity at (J2 = 10° below
1 is entirely due to the finite surface roughness. The solid line in Fig. 5.13b is the
result of a calculation of the diffuse scattering intensity [5.70, 71] by means of a
142 5. Principles, Techniques, and Instrumentation of NEXAFS

Z <8>

0
a::
_0
500A Au/Glass
(b) ci> A = 100A Diffuse Intensity (e)
10 = 10'
CI)

c: A = 100A
~ 8,
~ -e.
;::; 10-2
1J
8, = 10' CI)

...
>
'in
.... Experiment C>
c:
c(
0
c:
$ Cl
c:
c: 10-4 .;:

~
1J 10-3
-10
..
~
(J
~ 10-6 rJl
eX 40 30 20 10 0 40 30 20 10 0
Scattering Angle 112 (deg.) Scattering Angle 112 (deg.)

Fig. 5.13. (a) Geometry for X-ray reflectivity and surface diffuse scattering measurements by
Hogrefe [5.70]. (b) Measured angular distribution of the diffuse scattering intensity from a 500 A
gold film on a glass substrate. The substrate was roughened before film deposition by polishing with
a 1 JIm diamond paste [5.70]. The data shown as dots were taken at A. = 100 A at an incidence angle
(Jl = 10° in an s-polarization geometry (</I = 0°). The experimental data have been normalized to the
incident photon flux 10 and the specular reflectivity Ro of the smooth gold surface such that the
reduction ofthe intensity at the specular angle below 1 is solely due to the finite vertical RMS surface
roughness lJ. The solid line is a fit with a scalar theory assuming lJ = 27 A, and a lateral correlation
length A = 2500 A. (e) Angular dependence of the diffuse scattering intensity as a function of (J2 and
</I. The lines represent equi-intensity profiles, normalized to the diffuse intensity at the specular angle.
The intensity variation with (J2 corresponds to the solid line in (b). Note that the scattered intensity
falls off much more rapidly with </I than with (J2' All data and theoretical results are taken from
[5.70]

so-called "scalar theory", which neglects polarization effects [S.64, 70, 71].
Because of the rather lengthy mathematical expressions for Ids in both the
"scalar theory" [S.64] or a perturbative "vector theory" [S.74] we shall here
discuss only its qualitative dependence on various parameters. We also note that
the scalar and vector theories give quantitatively different results at larger
scattering angles [S.70, 71] while both agree in their qualitative predictions. In
general Ids can be expressed as

(S.29)

where the angles have been defined in Fig. S.13.

According to (S.26) the total integrated scatter, i.e., the area under the solid
curve in Fig. S.13b, increases with increasing vertical surface roughness b at the
expense of the speculady reflected intensity. To first order, b does not affect the
 5.3 Fluorescence Yield Detection 143

angular distribution of the diffuse intensity. In contrast, the lateral correlation

length A, and A and (Jl strongly affect the angular distribution, which becomes
more peaked at the specular angle and diminished at larger scattering angles
with increasing A and (Jl and decreasing .Ie [5.70, 71, 75].
An important general result is revealed by Fig.5.13c, namely, that the
scattered intensity falls off much more rapidly in a direction perpendicular to the
plane of incidence (change of ¢, (J2 fixed), than in the plane of incidence (change
of (J2 , ¢ = 0°). The angular dependence shown by Fig. 5.13c was calculated by
means of the scalar theory [5.70, 71J but similar results are obtained with the
vector theory [5.72]. The result shown by Fig. 5.l3c is the key to eliminating
most of the diffusely scattered intensity by placing the detector out of the plane
of incidence. For a vertical sample mount, as shown in Fig. 5.7, the detector is
best mounted underneath the sample. By detecting only X-rays from the sample
within a 20° halfcone from the sample surface, centered about a vertical axis, this
detector geometry offers a reasonable solid acceptance angle (~5% of 4n sr) and
places the acceptance cone more than 70° away from the reflected beam, at any
X-ray incidence angle on the sample. Therefore, the diffusely scattered intensity
will be reduced by many orders of magnitude. Figure 5.l3c suggests that the
reduction factor will be at least 5 orders of magnitude, which would decrease the
diffuse scattering intensity to a value below the fluorescence signal from the
adsorbate.
A detector placed underneath the sample which views the sample in a
grazing geometry also reduces the coherently scattered and fluorescence back-
ground intensity. The reduction arises, similar to the case of electron yield
detection, from the angular dependence of the X-ray escape depth term
D = cosb/px in (5.19) and (5.20). At grazing emission angles (b> 70°) the
effective X-ray escape depth in the substrate and therefore I~ and I~s are greatly
reduced. At this point a comment is required about the angular dependence of
the elastically scattered intensity. For linearly polarized X-rays the scattering
cross section is not isotropic but, using the scattering geometry illustrated by
Fig. 5.14, is given by

(5.30)

Fig. 5.14. Coordinate system for a coherent X-ray

scattering process from an atom for linearly polarized X-
rays. k i is the incident wave vector along the y-axis, ks is the
scattered wave vector. The E vector lies along the x-axis. If
E lies in the scattering plane defined by k i and k" i.e., X
x = 90°, we speak ofp-polarization. The geometry", = 0° is
referred to as s-polarization
144 5. Principles, Techniques, and Instrumentation of NEXAFS

Note that the polarization factor in (5.30), 1 - sin 2 '" sin 2 X, differs from that for
unpolarized radiation, (1 + COS 2 6)/2, typically found in the X-ray literature,
where 6 is the angle between the incident and scattered beams. Integration of
(5.30) over all angles yields (5.16). For linearly polarized radiation the scattered
intensity vanishes for'" = X = 90°, i.e., when the wavevector kg of the scattered
radiation lies along the electric field vector E. For E perpendicular to the
scattering plane (s-polarization, '" = 0°) there is no angular dependence. Equa-
tion (5.30) indicates that the detector geometry dictated by the diffuse scattering
intensity is not optimal for suppression of the coherent scattering intensity,
which would favor a detector orientation perpendicular to the incident X-ray
beam, along the electric field vector. However, the suppression of the diffuse
intensity is more important, judged by its larger overall intensity.
The above general considerations are confirmed by recent measurements
carried out under conditions that resemble those of a NEXAFS study of a
carbon-containing adsorbate on a single-crystal sample [5.73]. Fischer et al.
measured the angular dependence of the scattered soft X-ray radiation from a
Pt(111) crystal at an incident photon energy of 275eV. The scattered photons
were detected with a proportional counter (Sect. 5.3.5) whose energy window
was centered around the incident photon energy. Since this energy coincides
with the carbon Ka. fluorescence energy, the experiment directly probed the
scattered X-ray background that would be present in a NEXAFS experiment of

Rotation
Axis
I (a) (b)

___
~
II

~--------/
/
/
/
/
/
/
/
/
II
Surface )n-Plane Out-of-Plane
Normal Detector

Out-of-Plane
o 25 50 75
Detector Angle of Incidence. e (deg)
Fig. [Link]. (a) Experimental arrangement used by Fischer et al. [5.73] to measure the scattered light
intensity off a polished Pt(lll) single crystal at hv = 275 eV. The (nearly) linearly polarized soft X-
rays with the electric field vector E in the horizontal plane were incident on the vertical sample at an
angle ewith respect to the surface (p-polarization). The sample could be rotated about a vertical axis
allowing measurements from grazing (e "" 0°) to normal (e = 90°) incidence angles. The scattered
light was measured with a proportional counter with a 2 x 2cm 2 detection area, located 13 em from
the sample. Two detector positions were explored, both at right angles with respect to the incident
beam direction: underneath the sample labeled "out-of-plane", and along the electric field vector
direction labeled "in-plane". (b) Measured scattered light intensity for in-plane and out-of-plane
e
detector positions [5.73]. For in-plane orientation the large peak around = 45° is due to the
specularly reflected beam. Note that at all angles the out-of-plane detection intensity is lower than
the in-plane detection intensity
 5.3 Fluorescence Yield Detection 145

a carbon-containing species on the same Pt(111) surface. The detector was

positioned such that its angular acceptance was about 10°. The experimental
geometry, pictured in Fig. 5.15a, is seen to closely reflect that of a typical
NEXAFS experiment shown in Fig. 5.7. The soft X-rays, with the (major)
electric field vector component E in the horizontal plane, are incident at an angle
e relative to the surface of the vertically oriented sample. By rotation about the
vertical axis the angle of incidence could be varied in the range 0° ~ e ~ 90° (p-
polarization geometry). The detector was positioned either underneath the
sample, perpendicular to the plane of incidence, labeled "out-of-plane", or along
the electric field vector direction, in the plane of incidence, labeled "in-plane".
The measured scattered intensity for the two detector orientations is shown
in Fig. 5.15b. Clearly the "out-of-plane" intensity is much smaller than the "in-
plane" intensity at all angles e. The large peak around e = 45° is due to the
specularly reflected X-rays which enter the detector at this angle for the "in-
plane" geometry. The angular dependence indicates that the scattered radiation
is due to diffuse surface scattering rather than coherent bulk scattering. The
results unambiguously demonstrate the advantage of "out-of-plane" detection
for X-ray fluorescence NEXAFS studies.

5.3.5 Experimental Details and Detectors

Fluorescence yield measurements are carried out using an experimental ar-
rangement similar to that shown in Fig. 5.7 with the partial electron yield
detector replaced by a suitable fluorescence yield detector. For a general review
of photon detectors in the soft-X-ray region the reader is referred to the classic
book by Samson [5.76] and the recent review by Timothy and Madden [5.55].
Two detectors, i.e., a gas proportional counter and a semiconductor Si(Li) diode,
suitable for soft X-ray fluorescence studies, are shown in Fig. 5.16. Both de-
tectors offer energy discrimination capabilities and reasonable detection efficien-
cies and have been utilized for NEXAFS [5.77-83].
The first fluorescence NEXAFS studies of chemisorbed low-Z molecules
[5.79] were carried out on the sulfur K-edge (2470eV) by use of a gas propor-
tional counter with a 127 jLm thick beryllium window of 5 cm diameter [5.79].
The detector shown in Fig. 5.16 used by Fischer et al. [5.77] and Arvanitis et al.
[5.80] for fluorescence studies at the carbon K-edge is modelled on the original
detector but utilizes two 1 jLm thick polypropylene windows of 1.75 cm diameter.
The inner window is supported by a stainless steel wire mesh of 60% transmis-
sion. The transmission of the polypropylene window as a function of photon
energy [5.84] is displayed in Fig. 5.17 and is seen to be about 83% for C K~
radiation. Other suitable windows for soft X-rays have been discussed by Henke
and co-workers [5.84, 85].
In order to improve the energy resolution, Fischer et al. [5.81, 82] have more
recently used a cylindrical detector geometry indicated in Fig. 5.18 and obtained
a carbon K~ peak whose FWHM was about 240eV or 85% of the photon
energy, close to the theoretical limit [5.85]. The detection efficiency of this
146 5. Principles, Techniques, and Instrumentation of NEXAFS

Gas Proport ional Detector

Removable Detector
Vacuum Detector Gas Head Assembly
(- 10-3 Torr) (- 200 Torr) with 0 Ring Sea l

X-rays High Voltage

Connector- Wiper
Assembly

Po lypropylene
Window
'pm Th ick
Alum inized (200A)
Polypropylene Window
(1pm) on Grid Support

SI (U ) "Quantum" Detector by Kevex Corporation

X- rays

"Quantum"/,
Window

Copper
Co ld Finger

Fig. 5.16a, b. Schematics of two energy resolving detectors for soft X-ray fluorescence detection. The
gas proportional counter (a) is the URY compatible design of Fischer et al. [5.77]. The URY
environment of the surface science chamber is decoupled from the ionization region filled with a
suitable detector gas (e.g., propane) by two 1 /lm thick circular polypropylene windows (each ~83%
transmitting for C K. radiation) of 1.75 cm diameter. The region between the two windows is
differentially pumped. The inner window is supported by a 60% transmitting stainless steel grid and
is capable of withstanding atmospheric pressure. It is aluminized (200 A) on the side facing the anode
wire in order to improve the electric field distribution. (b) The Si(Li) "Quantum" detector com-
mercially available from Kevex Corporation (San Carlos, CA, USA), which uses a special window (of
proprietary composition) to protect the high vacuum environment of the Si(Li) detector crystal. The
window offers a 10 mm 2 circular aperture and can withstand atmospheric pressure. The total
detector efficiency is shown in Fig. 5.17

detector is about 60%, as defined by the transmission of the 2 windows (83%

each) and the stainless steel grid (90%) and the unit quantum efficiency of the
detector gas. Its solid angle of acceptance is about 5 % of 4n sr. The arrangement
shown in Fig. 5.18 was designed with the goal of studying samples in the
presence of a gas, i.e., close to the reaction conditions used in real catalytic
reactions, and studies of this kind have been performed by Zaera et al. [5.82].
The efficiency of the commercially available "Quantum" Si(Li) detector
shown in Fig. 5.16 is plotted as a function of photon energy in Fig. 5.17. For
 5.3 Fluorescence Yield Detection 147

/
/'
./
---- "Quantum" detector
80
~
OK / c:
0'./ 0
'iii
/ 60 <I)

'E
L .:.... , Beryllium window (-8pm) <I)
c:
• detector <U

40 ~
~
0
"0
c:
20 ~
'--~--1pm Polypropylene
Transmission
.. '

o 1.0