Atomic Molecular Optical Physics by Hertel C Schulz, Volume 1

Graduate Texts in Physics
Ingolf V. Hertel
Claus-PeterSchulz
Atoms,
Molecules and
Optical Physics 1
Atoms and Spectroscopy
For further volumes:

www.springer.com/series/8431
Graduate Texts in Physics publishes core learning/teaching material for graduate- and advanced-
level undergraduate courses on topics of current and emerging fields within physics, both pure and
applied. These textbooks serve students at the MS- or PhD-level and their instructors as compre-
hensive sources of principles, definitions, derivations, experiments and applications (as relevant)
for their mastery and teaching, respectively. International in scope and relevance, the textbooks
correspond to course syllabi sufficiently to serve as required reading. Their didactic style, com-
prehensiveness and coverage of fundamental material also make them suitable as introductions or
references for scientists entering, or requiring timely knowledge of, a research field.
Series Editors
Professor Richard Needs
Cavendish Laboratory
JJ Thomson Avenue
Cambridge CB3 0HE, UK
rn11@cam.ac.uk
Professor William T. Rhodes

Department of Computer and Electrical Engineering and Computer Science
Imaging Science and Technology Center
Florida Atlantic University
777 Glades Road SE, Room 456
Boca Raton, FL 33431, USA
wrhodes@fau.edu
Professor Susan Scott

Department of Quantum Science
Australian National University
Science Road
Acton 0200, Australia
susan.scott@anu.edu.au
Professor H. Eugene Stanley

Center for Polymer Studies Department of Physics
Boston University
590 Commonwealth Avenue, Room 204B
Boston, MA 02215, USA
hes@bu.edu
Professor Martin Stutzmann

Walter Schottky Institut
TU Mnchen
85748 Garching, Germany
stutz@wsi.tu-muenchen.de
Ingolf V. Hertel r Claus-Peter Schulz
Atoms, Molecules and

Optical Physics 1
Atoms and Spectroscopy
Ingolf V. Hertel Claus-Peter Schulz
Max-Born-Institut fr Nichtlineare Optik Max-Born-Institut fr Nichtlineare Optik
und Kurzzeitspektroskopie und Kurzzeitspektroskopie
Berlin, Germany Berlin, Germany
ISSN 1868-4513 ISSN 1868-4521 (electronic)

ISBN 978-3-642-54321-0 ISBN 978-3-642-54322-7 (eBook)
DOI 10.1007/978-3-642-54322-7
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2014952813
Springer-Verlag Berlin Heidelberg 2015

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publishers location, in its current version, and permission for use must always be obtained from Springer.
Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations
are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, 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.
While the advice and information in this book are believed to be true and accurate at the date of pub-
lication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any
errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect
to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

To my Wife Erika
IVH
To my Wife Gudrun
CPS
Preface
Atomic, Molecular and Optical physics short AMO physics is one of the canon-
ical fields of physics, a profound knowledge of which is essential for understanding
almost any other area of modern physics. And while its roots reach back over a
century and are closely connected with the early days of modern physics, current
research in AMO physics is still highly productive in respect of both, cutting edge
applications and fundamental insights as several N OBEL prizes in recent years
have documented convincingly.
Looking back at the technical development of modern industrial society which
is closely connected with modern physics one may refer (COSE, 1998) to the
20th century as that of the electron while the 21st is the century of the photon. This
interesting particle, the essential ingredient of modern optics and quantum optics,
surprises humankind since N EWTON with its wave-particle dualism. It does not only
play a key role in todays information technology but is, from a general point of
view, also the primary carrier of any information which can be obtained about the
constituents of matter and materials. Even collisions of particles with mass under the
influence of the C OULOMB force may be viewed as exchange of virtual photons.
The textbooks presented here try to give a fairly comprehensive overview on the
whole field. They cover state of the art experimental methods, and combine this
with preparing the basis for a serious, theory based understanding of key aspects in
modern AMO research. The two volumes, originally written in German language
(H ERTEL and S CHULZ, 2008), are a genuine authors translation not just an En-
glish mirror image of the original. We have rewritten much of the text, extended
it wherever appropriate, and updated a number of aspects to catch up with recent
progress in the field.
On the one hand we address advanced students of physics, chemistry and other
neighbouring fields, typically at the end of their undergraduate studies, or during
their doctoral work. On the other hand we also wish to reach young postdocs or
even mature scientists, who feel it is time they connect freshly with the topics ad-
dressed here. We consider the basics of classical geometrical optics and wave optics
as well as electrodynamics to be well known by our readers. We also expect a cer-
vii
viii Preface
tain basic knowledge and understanding of atomistic concepts in physics, as well as

of elementary quantum mechanics.
We do, however, provide in Chaps. 1 and 2 of this Vol. 1 a brief repetition of these
topics essentially an extended list of keywords focussed on basic understanding
and knowledge. In the main part we cover the standard scope of atomic physics,
touch some modern aspects of spectroscopy, and try to lead the reader up to state-
of-the-art research in some main areas of the field wherever possible and as far as
space permits. The sequence of chapters follows essentially the logics of perturba-
tion theory. The strongest perturbation is treated first. Thus, after the introductory
chapters where pure C OULOMB interaction and the H atom have been discussed, in
Chap. 3 we allow for coarse deviations from the 1/r potential and focus on quasi-
one-electron systems. This, and some common sense, allows us already to introduce
the periodic system of elements. Next, in Chap. 4, we have to treat optically induced
and spontaneous transitions: they are a central theme in AMO physics. This requires
a brief introduction to time dependent perturbation theory, a topic which is indis-
pensible in AMO physics, but which is often neglected in undergaduate quantum
mechanics. To allow the reader a step by step approach towards the more demand-
ing topics, we implement at this point only the semiclassical approach by which
95 % of standard atomic physics may be treated (resorting occasionally to somewhat
hand waving arguments) and postpone field quantization to Vol. 2.
Chapter 5 further extends this knowledge, treating shapes and widths of spectral
lines and introducing multiphoton processes as well as transitions into the contin-
uum. We are now ready to understand in Chap. 6 a next step of complication, fine
structure (FS) interaction. In order to allow the reader to appreciate the experimen-
tal efforts, we also give a brief introduction to high resolution and precision laser
spectroscopy. This leads us automatically to the L AMB shift and calls for a short
side step into the basics of quantum electrodynamics (QED). In Chap. 7 two elec-
tron systems are treated, mainly the He atom and He like ions. Exchange interaction
may be smaller or larger than FS, depending on the system, but the step to multielec-
tron systems adds a new degree of complexity and sets the stage for a quantitative
treatment of the PAULI exclusion principle.
The next finer step in the hierarchy of perturbations is treated in Chap. 8, includ-
ing interactions between atomic electrons and external magnetic and electric fields,
leading to Z EEMAN and S TARK effect, respectively. At this point, a small detour
into the world of interaction between atoms and very intense laser fields is appro-
priate, as the theoretical formalism used is essentially an extension of the so called
dynamical S TARK effect. As a last refinement we include in Chap. 9 hyperfine in-
teractions between the atomic nucleus and electrons. These lead to very small but
highly significant splittings of atomic energy levels (HFS) and offer a wealth of prac-
tical applications. In the last Chap. 10 of Vol. 1 we are finally ready to treat genuine
multi-electron systems with a large number of electrons. We discuss the appropriate
theoretical tools (such as HF equations, CI methods, and DFT), and present some
relevant methods of X-ray spectroscopy and sources for generating X-ray radiation.
As a rule, we try to avoid extensive mathematical derivations. Rather, in the spirit
of these books we prefer to give the reader some general guidance on how to reach
Preface ix
the final, physically important results which we discuss and illustrate usually in
some detail. In addition, we provide several appendices for the reader interested
in more detail. We have e.g. collected a toolbox for angular momentum algebra in
atomic and molecular physics without any claim for full mathematical consistency,
but quite compact and possibly useful in practice.
Some words about formats, notation, units, typography appear in order:
Each chapter begins with a brief motto setting the tune of the chapter, followed
by short abstract guiding the reader through the text. At the end of each section a
short summary recalls what the readers should have learned from the preceding
text. All chapters build upon each other, but may be read by advanced readers
also individually: this is facilitated by intensive cross referencing of formulas and
figures, extended indices covering both volumes, a list of acronyms and important
terminology as well as references at the end of each chapter.
For clarity and homogeneity we do not reproduce original drawings or other
material from the literature. Rather, all published data have been redrawn (after
digitalization if necessary), are presented in a standard format, and all sources
used in the figures and text are properly quoted.
We consequently use the SI-System for all measurable quantities, and we empha-
size the pedagogical and practical value of a dimensional analysis for complex
physical formulas.1 On the other hand, atomic units (a.u.) facilitate the writing
of many relations in atomic and molecular physics dramatically. Hence, we use
them intensively considering, however, Eh , a0 and t0 etc. simply as abbrevia-
tions for quantities with dimensions. Phrasings such as we set , e, me , c equal
to unity are avoided, since they are highly misleading.
The finite number of letters in the Latin and Greek alphabets makes some incon-
sistencies or unusual designations unavoidable: we mention specifically, that in
order to allow the use of E for the electric field strength (an important quantity
in AMO) we use W (with appropriate indices) for energies of various types (with
the exception of the atomic unit of energy which is internationally defined as Eh ).
Occasionally we use the letter T for kinetic energy and try to avoid the neigh-
bourhood of time and temperature which are often also designated by T . Vectors
are written as r or k, unit vectors in these directions are er and ek , respectively.
We write operators as H , vector-operators as p and tensors of rank k as Ck . For
the unit operator and unit matrix we use 1. For integer numbers we mostly use
calligraphic letters such as N , while number densities are simply N to distin-
guish them from the index of refraction n which is also an often used quantity
throughout this text. Oscillations and other periodic processes are mostly char-
acterized by their angular frequencies (sometimes also by their frequencies )
and the corresponding energies are (or h).
1 We make, however, use of allowed prefixes (NIST, 2000a), such as cm1 as unit of wavenumbers
(which appears ineradicable in the literature). We also use accepted units outside the SI (NIST,
2000b), such as the enormously practical energy unit eV (electronvolt), or b (barn) as unit for cross
sections.
x Preface
Finally, we hope that these books will become a continuing source of reference
for the fastidious reader, working in or just needing to use AMO physics in her or
his special field. We ask all of you to kindly provide us with the necessary feedback.
We shall try to react to useful suggestion promptly. At the home page of the books,
http://www.mbi-berlin.de/AMO/book-homepage, we shall continuously report on
the status, list errata and possibly present additions. For additional reading and
cross referencing we have collected a few related textbooks and monographs in the
reference list below, just as typical examples without any claim for completeness.
Berlin Adlershof, Germany Ingolf V. Hertel
January 2014 Claus-Peter Schulz
Acronyms and Terminology
AMO: Atomic, molecular and optical, physics.

a.u.: atomic units, see Sect. 2.6.2.
CI: Configuration interaction, mixing of states with different electronic configu-
rations in atomic and molecular structure calculations, using linear superposition
of S LATER determinants (see Sect. 10.2.3).
DFT: Density functional theory, today one of the standard methods for computing
atomic and molecular electron densities and energies (see Sect. 10.3).
FS: Fine structure, splitting of atomic and molecular energy levels due to spin
orbit interaction and other relativistic effects (Chap. 6).
HF: H ARTREE -F OCK, method (approximation) for solving a multi-electron
S CHRDINGER equation, including exchange interaction.
HFS: Hyperfine structure, splitting of atomic and molecular energy levels due to
interactions of the active electron with the atomic nucleus (Chap. 9).
NIST: National institute of standards and technology, located at Gaithersburg
(MD) and Boulder (CO), USA. http://www.nist.gov/index.html.
QED: Quantum electrodynamics, combines quantum theory with classical elec-
trodynamics and special relativity. It gives a complete description of light-matter
interaction.
References
ATKINS, P. W. and R. S. F RIEDMAN: 2010. Molecular Quantum Mechanics. Oxford: Oxford Uni-
versity Press, 2nd edn.
B ERGMANN, L. and C. S CHAEFER: 1997. Constituents of Matter Atoms, Molecules, Nuclei and
Particles. Berlin, New York: Walter der Gruyter, 902 pages.
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics
64. Berlin, Heidelberg: Springer Verlag, 3rd edn., 343 pages.
B ORN, M. and E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
Preface xi
B RANSDEN, B. H. and C. J. J OACHAIN: 2003. The Physics of Atoms and Molecules. Prentice Hall
Professional.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
COSE (Committee Optical Science and Engineering): 1998. Harnessing Light: Optical Science
and Engineering for the 21st Century. Washington, D.C: National Academy Press, 360 pages.
D EMTRDER, W.: 2010. Atoms, Molecules and Photons. Berlin, Heidelberg, New York: Springer,
2nd edn.
D RAKE, G. W. F., ed.: 2006. Handbook of Atomic, Molecular and Optical Physics. Heidelberg,
New York: Springer.
E DMONDS, A. R.: 1996. Angular Momentum in Quantum Mechanics. Princeton, NJ, USA: Prince-
ton University Press, 154 pages.
H ERTEL, I. V. and C. P. S CHULZ: 2008. Atome, Molekle und optische Physik 1; Atomphysik
und Grundlagen der Spektroskopie. Springer-Lehrbuch. Berlin, Heidelberg: Springer-Verlag,
1st edn., 511 pages.
H ERTEL, I. V. and C. P. S CHULZ: 2010. Atome, Molekle und optische Physik 2; Molekle und
Photonen - Spektroskopie und Streuphysik, vol. 2 of Springer-Lehrbuch. Berlin, Heidelberg:
Springer-Verlag, 1st edn., 639 pages.
NIST: 2000a. Reference on constants, units, and uncertainties: SI prefixes, NIST. http://physics.
nist.gov/cuu/Units/prefixes.html, accessed: 8 Jan 2014.
NIST: 2000b. Reference on constants, units, and uncertainties: Units outside the SI, NIST.
http://physics.nist.gov/cuu/Units/outside.html, accessed: 8 Jan 2014.
S TEINFELD, J. I.: 2005. Molecules and Radiation, An Introduction to Modern Molecular Spec-
troscopy. Mineola, NY: Dover Edition, 2nd edn.
W EISSBLUTH, M.: 1978. Atoms and Molecules. Student Edition. New York, London, Toronto,
Syndey, San Francisco: Academic Press, 713 pages.
Acknowledgements
Over the past years, many colleagues have encouraged and stimulated us to move
forward with this work, and helped with many critical hints and suggestions. Most
importantly, we have received a lot of helpful material and state of the art data for
inclusion in these textbooks.
We would like to thank all those who have in one or the other way contributed
to close a certain gap in the standard textbook literature in this area that is at
least what we hope to have achieved. Specifically we mention Robert Bittl, Wolf-
gang Demtrder, Melanie Dornhaus, Kai Godehusen, Uwe Griebner, Hartmut Ho-
top, Marsha Lester, John P. Maier, Reinhardt Morgenstern, Hans-Hermann Ritze,
Horst Schmidt-Bcking, Ernst J. Schumacher, Gnter Steinmeyer, Joachim Ullrich,
Marc Vrakking und Roland Wester; their contributions are specifically noted in the
respective lists of references.
Of course, all other sources are also documented there which we have used for
information and which have provided the data used to generate the figures in these
books.
One of us (IVH) is particularly grateful to the Max-Born-Institute for provid-
ing the necessary resources (including computer facilities, library access, and office
space etc.) for continuing the work on this book after official retirement.
xiii
Contents of Volume 1
1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Overview, History and Magnitudes . . . . . . . . . . . . . . . . 1
1.1.1 Quantum Nature of Matter . . . . . . . . . . . . . . . . . 2
1.1.2 Orders of Magnitude . . . . . . . . . . . . . . . . . . . . 5
1.2 Special Theory of Relativity in a Nutshell . . . . . . . . . . . . . 10
1.2.1 Kinematics and Dynamics . . . . . . . . . . . . . . . . . 10
1.2.2 Time Dilation and LORENTZ Contraction . . . . . . . . . 13
1.3 Some Elementary Statistics and Applications . . . . . . . . . . . 14
1.3.1 Spontaneous Decay and Mean Lifetime . . . . . . . . . . 15
1.3.2 Absorption, LAMBERT-BEER Law . . . . . . . . . . . . 17
1.3.3 Kinetic Gas Theory . . . . . . . . . . . . . . . . . . . . 18
1.3.4 Classical and Quantum Statistics Fermions and Bosons 20
1.4 The Photon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.1 Photoelectric Effect and Quantization of Energy . . . . . 26
1.4.2 COMPTON Effect and Momentum of the Photon . . . . . 28
1.4.3 Pair Production . . . . . . . . . . . . . . . . . . . . . . 30
1.4.4 Angular Momentum and Mass of the Photon . . . . . . . 30
1.4.5 Electromagnetic Spectrum . . . . . . . . . . . . . . . . . 31
1.4.6 PLANCKs Radiation Law . . . . . . . . . . . . . . . . . 31
1.4.7 Solar Radiation on the Earth . . . . . . . . . . . . . . . . 34
1.4.8 Photometry Luminous Efficiency and Efficacy . . . . . 37
1.4.9 X-Ray Diffraction and Structural Analysis . . . . . . . . 40
1.5 The Four Fundamental Interactions . . . . . . . . . . . . . . . . 43
1.5.1 COULOMB and Gravitational Interaction . . . . . . . . . 44
1.5.2 The Standard Model of Fundamental Interaction . . . . . 46
1.5.3 Hadrons . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.5.4 The Electron . . . . . . . . . . . . . . . . . . . . . . . . 49
1.6 Particles in Electric and Magnetic Fields . . . . . . . . . . . . . 51
1.6.1 Charge in an Electric Field . . . . . . . . . . . . . . . . 52
1.6.2 Charge in a Magnetic Field . . . . . . . . . . . . . . . . 53
xv
xvi Contents of Volume 1
1.6.3 Cyclotron Frequency and ICR Spectrometers . . . . . . . 54

1.6.4 Other Mass Spectrometers . . . . . . . . . . . . . . . . . 54
1.6.5 Plasma Frequency . . . . . . . . . . . . . . . . . . . . . 56
1.7 Particles and Waves . . . . . . . . . . . . . . . . . . . . . . . . 57
1.7.1 DE BROGLIE Wavelength . . . . . . . . . . . . . . . . . 57
1.7.2 Experimental Evidence . . . . . . . . . . . . . . . . . . 58
1.7.3 Uncertainty Relation and Measurement . . . . . . . . . . 61
1.7.4 Stability of the Atomic Ground State . . . . . . . . . . . 63
1.8 BOHR Model of the Atom . . . . . . . . . . . . . . . . . . . . . 64
1.8.1 Basic Assumptions . . . . . . . . . . . . . . . . . . . . . 65
1.8.2 Radii and Energies . . . . . . . . . . . . . . . . . . . . . 67
1.8.3 Atomic Units (a.u.) . . . . . . . . . . . . . . . . . . . . 67
1.8.4 Energies of Hydrogen Like Ions . . . . . . . . . . . . . . 68
1.8.5 Correction for Finite Nuclear Mass . . . . . . . . . . . . 68
1.8.6 Spectra of Hydrogen and Hydrogen Like Ions . . . . . . 69
1.8.7 Limits of the BOHR Model . . . . . . . . . . . . . . . . 69
1.9 STERN-GERLACH Experiment and Space Quantization . . . . . 70
1.9.1 Magnetic Moment and Angular Momentum . . . . . . . 70
1.9.2 Magnetic Moment in a Magnetic Field . . . . . . . . . . 71
1.9.3 The Experiment . . . . . . . . . . . . . . . . . . . . . . 72
1.9.4 Interpretation of the S TERN-G ERLACH Experiment . . . 75
1.9.5 Consequences of the STERN-GERLACH Experiment . . . 77
1.10 Electron Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
1.10.1 Magnetic Moment of the Electron . . . . . . . . . . . . . 79
1.10.2 EINSTEIN-DE-HAAS Effect . . . . . . . . . . . . . . . . 79
Acronyms and Terminology . . . . . . . . . . . . . . . . . . . . . . . 81
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2 Elements of Quantum Mechanics and the H Atom . . . . . . . . . 87
2.1 Matter Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.1.1 Limits of Classical Theory . . . . . . . . . . . . . . . . . 87
2.1.2 Probability Amplitudes in Optics . . . . . . . . . . . . . 88
2.1.3 Probability Amplitudes and Matter Waves . . . . . . . . 89
2.2 SCHRDINGER Equation . . . . . . . . . . . . . . . . . . . . . 90
2.2.1 Stationary SCHRDINGER Equation . . . . . . . . . . . 91
2.2.2 HAMILTON and Momentum Operators . . . . . . . . . . 91
2.2.3 Time Dependent SCHRDINGER Equation . . . . . . . . 92
2.2.4 Freely Moving Particle The Most Simple Example . . . 94
2.3 Basics and Definitions of Quantum Mechanics . . . . . . . . . . 95
2.3.1 Axioms, Terminology and Rules . . . . . . . . . . . . . 95
2.3.2 Representations . . . . . . . . . . . . . . . . . . . . . . 99
2.3.3 Simultaneous Measurement of Two Observables . . . . . 100
2.3.4 Operators for Space, Momentum and Energy . . . . . . . 101
2.3.5 Eigenfunctions of the Momentum Operator p . . . . . . 102
2.4 Particles in a Box And the Free Electron Gas . . . . . . . . . . 103
2.4.1 One Dimensional Potential Box . . . . . . . . . . . . . . 103
Contents of Volume 1 xvii
2.4.2 Three Dimensional Potential Box . . . . . . . . . . . . . 104

2.4.3 The Free Electron Gas . . . . . . . . . . . . . . . . . . . 105
2.5 Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . 107
2.5.1 Polar Coordinates . . . . . . . . . . . . . . . . . . . . . 107
2.5.2 Definition of Orbital Angular Momentum . . . . . . . . . 109
2.5.3 Eigenvalues and Eigenfunctions . . . . . . . . . . . . . . 109
2.5.4 Electron Spin . . . . . . . . . . . . . . . . . . . . . . . 114
2.6 One Electron Systems and the Hydrogen Atom . . . . . . . . . . 117
2.6.1 Quantum Mechanics of the One Particle System . . . . . 117
2.6.2 Atomic Units . . . . . . . . . . . . . . . . . . . . . . . . 118
2.6.3 Centre of Mass Motion and Reduced Mass . . . . . . . . 119
2.6.4 Qualitative Considerations . . . . . . . . . . . . . . . . . 119
2.6.5 Exact Solution for the H Atom . . . . . . . . . . . . . . 121
2.6.6 Energy Levels . . . . . . . . . . . . . . . . . . . . . . . 122
2.6.7 Radial Functions . . . . . . . . . . . . . . . . . . . . . . 123
2.6.8 Density Plots . . . . . . . . . . . . . . . . . . . . . . . . 124
2.6.9 Spectra of the H Atom . . . . . . . . . . . . . . . . . . . 126
2.6.10 Expectation Values of r k . . . . . . . . . . . . . . . . . . 126
2.6.11 Comparison with the BOHR Model . . . . . . . . . . . . 127
2.7 Normal ZEEMAN Effect . . . . . . . . . . . . . . . . . . . . . . 128
2.7.1 Angular Momentum in an External B-Field . . . . . . . . 129
2.7.2 Removal of m Degeneracy . . . . . . . . . . . . . . . . 130
2.8 Dispersion Relations . . . . . . . . . . . . . . . . . . . . . . . . 131
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3 Periodic System and Removal of Degeneracy . . . . . . . . . . . 137
3.1 Shell Structure of Atoms and the Periodic System . . . . . . . . 137
3.1.1 Electron Configuration . . . . . . . . . . . . . . . . . . . 138
3.1.2 PAULI Principle . . . . . . . . . . . . . . . . . . . . . . 138
3.1.3 How the Shells are Filled . . . . . . . . . . . . . . . . . 139
3.1.4 The Periodic System of Elements . . . . . . . . . . . . . 140
3.1.5 Some Experimental Facts . . . . . . . . . . . . . . . . . 142
3.2 Quasi-One-Electron System . . . . . . . . . . . . . . . . . . . . 144
3.2.1 Spectroscopic Findings for the Alkali Atoms . . . . . . . 145
3.2.2 Quantum Defect . . . . . . . . . . . . . . . . . . . . . . 146
3.2.3 Screened COULOMB Potential . . . . . . . . . . . . . . . 148
3.2.4 Radial Wave Functions . . . . . . . . . . . . . . . . . . 149
3.2.5 Precise Calculations for Na as an Example . . . . . . . . 150
3.2.6 Quantum Defect Theory . . . . . . . . . . . . . . . . . . 152
3.2.7 MOSLEY Diagrams . . . . . . . . . . . . . . . . . . . . 159
3.3 Perturbation Theory for Stationary Problems . . . . . . . . . . . 161
3.3.1 Perturbation Ansatz for the Non-degenerate Case . . . . . 161
3.3.2 Perturbation Theory in 1st Order . . . . . . . . . . . . . 162
3.3.3 Perturbation Theory in 2nd Order . . . . . . . . . . . . . 163
xviii Contents of Volume 1
3.3.4 Perturbation Theory for Degenerate States . . . . . . . . 164

3.3.5 Application of Perturbation Theory to Alkali Atoms . . . 165
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
4 Non-stationary Problems: Dipole Excitation with One Photon . . . 169
4.1 Electromagnetic Waves: Electric Field, Intensity, Polarization
and Photon Spin . . . . . . . . . . . . . . . . . . . . . . . . . . 170
4.1.1 Electric Field and Intensity . . . . . . . . . . . . . . . . 170
4.1.2 Basis Vectors of Polarization . . . . . . . . . . . . . . . 171
4.1.3 Coordinate Systems . . . . . . . . . . . . . . . . . . . . 174
4.1.4 Angular Momentum of the Photon . . . . . . . . . . . . 175
4.2 Introduction to Absorption and Emission . . . . . . . . . . . . . 176
4.2.1 Stationary States . . . . . . . . . . . . . . . . . . . . . . 176
4.2.2 Optical Spectroscopy General Concepts . . . . . . . . 177
4.2.3 Induced Processes . . . . . . . . . . . . . . . . . . . . . 178
4.2.4 Spontaneous Emission Classical Interpretation . . . . . 181
4.2.5 The EINSTEIN A and B Coefficients . . . . . . . . . . . 184
4.3 Time Dependent Perturbation Theory . . . . . . . . . . . . . . . 186
4.3.1 General Approach . . . . . . . . . . . . . . . . . . . . . 186
4.3.2 Perturbation Ansatz for Transition Amplitudes . . . . . . 187
4.3.3 Transitions in a Monochromatic Plane Wave . . . . . . . 188
4.3.4 Dipole Approximation . . . . . . . . . . . . . . . . . . . 189
4.3.5 Absorption Probabilities . . . . . . . . . . . . . . . . . . 190
4.3.6 Absorption and Emission: A First Summary . . . . . . . 193
4.4 Selection Rules for Dipole Transitions . . . . . . . . . . . . . . 196
4.4.1 Angular Momentum and Selection Rules . . . . . . . . . 196
4.4.2 Transition Amplitudes in the Helicity Basis . . . . . . . . 198
4.4.3 Transition Matrix Elements and Selection Rules . . . . . 200
4.4.4 An Example for E1 Transitions: The H Atom . . . . . . . 201
4.5 Angular Dependence of Dipole Radiation . . . . . . . . . . . . . 203
4.5.1 Semiclassical Picture . . . . . . . . . . . . . . . . . . . 204
4.5.2 Angular Distributions from Quantum Mechanics . . . . . 206
4.6 Strength of Dipole Transitions . . . . . . . . . . . . . . . . . . . 212
4.6.1 Line Strength . . . . . . . . . . . . . . . . . . . . . . . 212
4.6.2 Spontaneous Transition Probabilities . . . . . . . . . . . 213
4.6.3 Induced Transitions . . . . . . . . . . . . . . . . . . . . 215
4.7 Superposition of States, Quantum Beats and Jumps . . . . . . . . 217
4.7.1 Coherent Population by Optical Transitions . . . . . . . . 217
4.7.2 Time Dependence of Optically Excited States Quantum
Beats . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4.7.3 Quantum Jumps . . . . . . . . . . . . . . . . . . . . . . 224
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
Contents of Volume 1 xix
5 Linewidths, Photoionization, and More . . . . . . . . . . . . . . . . 227

5.1 Line Broadening . . . . . . . . . . . . . . . . . . . . . . . . . . 227
5.1.1 Natural Linewidth . . . . . . . . . . . . . . . . . . . . . 227
5.1.2 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . 232
5.1.3 Collisional Line Broadening . . . . . . . . . . . . . . . . 233
5.1.4 DOPPLER Broadening . . . . . . . . . . . . . . . . . . . 234
5.1.5 VOIGT Profile . . . . . . . . . . . . . . . . . . . . . . . 236
5.2 Oscillator Strength and Cross Section . . . . . . . . . . . . . . . 238
5.2.1 Transition Rates Generalized . . . . . . . . . . . . . . . 238
5.2.2 Oscillator Strength . . . . . . . . . . . . . . . . . . . . . 238
5.2.3 Absorption Cross Section . . . . . . . . . . . . . . . . . 240
5.2.4 Different Notations Radiative-Transfer in Gases . . . . 242
5.3 Multi-photon Processes . . . . . . . . . . . . . . . . . . . . . . 244
5.3.1 Two-Photon Excitation . . . . . . . . . . . . . . . . . . 245
5.3.2 Two-Photon Emission . . . . . . . . . . . . . . . . . . . 248
5.4 Magnetic Dipole and Electric Quadrupole Transitions . . . . . . 250
5.5 Photoionization . . . . . . . . . . . . . . . . . . . . . . . . . . 254
5.5.1 Process and Cross Section . . . . . . . . . . . . . . . . . 255
5.5.2 BORN Approximation for Photoionization . . . . . . . . 256
5.5.3 Angular Distribution of Photoelectrons . . . . . . . . . . 260
5.5.4 Cross Sections in Theory and Experiment . . . . . . . . . 261
5.5.5 Multi-photon Ionization (MPI) . . . . . . . . . . . . . . 265
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
6 Fine Structure and L AMB Shift . . . . . . . . . . . . . . . . . . . . 273
6.1 Methods of High Resolution Spectroscopy . . . . . . . . . . . . 274
6.1.1 Grating Spectrometers . . . . . . . . . . . . . . . . . . . 274
6.1.2 Interferometers . . . . . . . . . . . . . . . . . . . . . . . 278
6.1.3 D OPPLER Free Spectroscopy in Atomic Beams . . . . . 282
6.1.4 Collinear Laser Spectroscopy in Ion Beams . . . . . . . . 283
6.1.5 Hole Burning . . . . . . . . . . . . . . . . . . . . . . . 284
6.1.6 D OPPLER Free Saturation Spectroscopy . . . . . . . . . 285
6.1.7 R AMSEY Fringes . . . . . . . . . . . . . . . . . . . . . 288
6.1.8 D OPPLER Free Two-Photon Spectroscopy . . . . . . . . 289
6.2 Spin-Orbit Interaction . . . . . . . . . . . . . . . . . . . . . . . 293
6.2.1 Experimental Findings . . . . . . . . . . . . . . . . . . . 293
6.2.2 Magnetic Moments in a Magnetic Field . . . . . . . . . . 294
6.2.3 General Considerations About LS Interaction . . . . . . 295
6.2.4 Magnitude of Spin-Orbit Interaction . . . . . . . . . . . 296
6.2.5 Angular Momentum Coupling . . . . . . . . . . . . . . . 297
6.2.6 Terminology for Atomic Structure . . . . . . . . . . . . 301
6.3 Quantitative Determination of Fine Structure . . . . . . . . . . . 303
6.3.1 FS Terms from D IRAC Theory . . . . . . . . . . . . . . 303
6.3.2 Fine Structure of the H Atom . . . . . . . . . . . . . . . 306
6.3.3 Fine Structure of Alkali and Other Atoms . . . . . . . . . 307
xx Contents of Volume 1
6.4 Selection Rules and Intensities of Transitions . . . . . . . . . . . 310

6.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 310
6.4.2 Transitions Between Sublevels vs. Overall Transition
Probabilities . . . . . . . . . . . . . . . . . . . . . . . . 310
6.4.3 Some Useful Relations for Spectroscopic Practice . . . . 313
6.5 L AMB Shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
6.5.1 Fine Structure and L AMB Shift for the H Line . . . . . . 316
6.5.2 Microwave and RF Transitions D OPPLER Free . . . . . 317
6.5.3 Experiment of L AMB and R ETHERFORD . . . . . . . . . 317
6.5.4 Precision Spectroscopy of the H Atom . . . . . . . . . . 319
6.5.5 LAMB Shift in Highly Charged Ions . . . . . . . . . . . 322
6.5.6 QED and F EYNMAN Diagrams . . . . . . . . . . . . . . 324
6.5.7 On the Theory of the L AMB Shift . . . . . . . . . . . . . 326
6.6 Electron Magnetic Moment Anomaly . . . . . . . . . . . . . . . 331
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
7 Helium and Other Two Electron Systems . . . . . . . . . . . . . . . 341
7.1 Introduction and Empirical Findings . . . . . . . . . . . . . . . 342
7.1.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
7.1.2 He I Term Scheme . . . . . . . . . . . . . . . . . . . . . 343
7.2 Some Quantum Mechanics of Two Electrons . . . . . . . . . . . 344
7.2.1 HAMILTON Operator for the Two-Electron System . . . . 344
7.2.2 Two Particle Wave Functions . . . . . . . . . . . . . . . 345
7.2.3 Zero Order Approximation: No e e Interaction . . . . . 346
7.2.4 The He Ground State Perturbation Theory . . . . . . . 348
7.2.5 Variational Theory and Present State-of-the-Art . . . . . 350
7.3 PAULI Principle and Excited States in He . . . . . . . . . . . . . 351
7.3.1 Exchange of Two Identical Particles . . . . . . . . . . . . 351
7.3.2 Symmetries of Spatial and Spin Wave Functions . . . . . 352
7.3.3 Perturbation Theory for (Singly) Excited States . . . . . 355
7.3.4 An Afterthought . . . . . . . . . . . . . . . . . . . . . . 358
7.4 Fine Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
7.5 Electric Dipole Transitions . . . . . . . . . . . . . . . . . . . . 362
7.6 Double Excitation and Autoionization . . . . . . . . . . . . . . . 365
7.6.1 Doubly Excited States . . . . . . . . . . . . . . . . . . . 365
7.6.2 Autoionization, FANO Profile . . . . . . . . . . . . . . . 366
7.6.3 Resonance Line Profiles . . . . . . . . . . . . . . . . . . 369
7.7 Quasi-two-Electron Systems . . . . . . . . . . . . . . . . . . . . 371
7.7.1 Alkaline Earth Elements . . . . . . . . . . . . . . . . . . 371
7.7.2 Mercury . . . . . . . . . . . . . . . . . . . . . . . . . . 372
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
Contents of Volume 1 xxi
8 Atoms in External Fields . . . . . . . . . . . . . . . . . . . . . . . . 377

8.1 Atoms in a Static Magnetic Field . . . . . . . . . . . . . . . . . 377
8.1.1 The General Case . . . . . . . . . . . . . . . . . . . . . 377
8.1.2 ZEEMAN Effect in Low Fields . . . . . . . . . . . . . . . 380
8.1.3 PASCHEN-BACK Effect . . . . . . . . . . . . . . . . . . 384
8.1.4 Do Angular Momenta Actually Precess? . . . . . . . . . 386
8.1.5 In Between Low and High Magnetic Field . . . . . . . . 388
8.1.6 Avoided Crossings . . . . . . . . . . . . . . . . . . . . . 392
8.1.7 Paramagnetism . . . . . . . . . . . . . . . . . . . . . . . 394
8.1.8 Diamagnetism . . . . . . . . . . . . . . . . . . . . . . . 396
8.2 Atoms in an Electric Field . . . . . . . . . . . . . . . . . . . . . 399
8.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 399
8.2.2 Significance . . . . . . . . . . . . . . . . . . . . . . . . 399
8.2.3 Atoms in a Static, Electric Field . . . . . . . . . . . . . . 400
8.2.4 Basic Considerations about Perturbation Theory . . . . . 401
8.2.5 Matrix Elements . . . . . . . . . . . . . . . . . . . . . . 402
8.2.6 Perturbation Series . . . . . . . . . . . . . . . . . . . . . 405
8.2.7 Quadratic STARK Effect . . . . . . . . . . . . . . . . . . 405
8.2.8 Linear STARK Effect . . . . . . . . . . . . . . . . . . . . 407
8.2.9 An example: RYDBERG States of Li . . . . . . . . . . . 409
8.2.10 Polarizability . . . . . . . . . . . . . . . . . . . . . . . . 411
8.2.11 Susceptibility . . . . . . . . . . . . . . . . . . . . . . . 413
8.3 Long Range Interaction Potentials . . . . . . . . . . . . . . . . . 414
8.4 Atoms in an Oscillating Electromagnetic Field . . . . . . . . . . 418
8.4.1 Dynamic STARK Effect . . . . . . . . . . . . . . . . . . 418
8.4.2 Index of Refraction . . . . . . . . . . . . . . . . . . . . 420
8.4.3 Resonances Dispersion and Absorption . . . . . . . . . 421
8.4.4 Fast and Slow Light . . . . . . . . . . . . . . . . . . . . 422
8.4.5 Elastic Scattering of Light . . . . . . . . . . . . . . . . . 427
8.5 Atoms in a High Laser Field . . . . . . . . . . . . . . . . . . . . 432
8.5.1 Ponderomotive Potential . . . . . . . . . . . . . . . . . . 432
8.5.2 KELDISH Parameter . . . . . . . . . . . . . . . . . . . . 434
8.5.3 From Multi-photon Ionization to Saturation . . . . . . . . 434
8.5.4 Tunnelling Ionization . . . . . . . . . . . . . . . . . . . 436
8.5.5 Recollision . . . . . . . . . . . . . . . . . . . . . . . . . 438
8.5.6 High Harmonic Generation (HHG) . . . . . . . . . . . . 439
8.5.7 Above-Threshold Ionization in High Laser Fields . . . . 441
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
9 Hyperfine Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 447
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
9.2 Magnetic Dipole Interaction . . . . . . . . . . . . . . . . . . . . 452
9.2.1 General Considerations and Examples . . . . . . . . . . 452
9.2.2 The Magnetic Field of the Electron Cloud . . . . . . . . 453
xxii Contents of Volume 1
9.2.3 Nonvanishing Orbital Angular Momenta . . . . . . . . . 457

9.2.4 The FERMI Contact Term . . . . . . . . . . . . . . . . . 458
9.2.5 Some Numbers . . . . . . . . . . . . . . . . . . . . . . . 459
9.2.6 Optical Transitions Between HFS Multiplets . . . . . . . 460
9.3 ZEEMAN Effect of Hyperfine Structure . . . . . . . . . . . . . . 461
9.3.1 Hyperfine Hamiltonian with Magnetic Field . . . . . . . 462
9.3.2 Low Magnetic Fields . . . . . . . . . . . . . . . . . . . 462
9.3.3 High and Very High Magnetic Fields . . . . . . . . . . . 464
9.3.4 Arbitrary Fields, BREIT-RABI Formula . . . . . . . . . . 467
9.4 Isotope Shift and Electrostatic Nuclear Interactions . . . . . . . . 471
9.4.1 Potential Expansion . . . . . . . . . . . . . . . . . . . . 471
9.4.2 Isotope Shift . . . . . . . . . . . . . . . . . . . . . . . . 473
9.4.3 Quadrupole Interaction Energy . . . . . . . . . . . . . . 477
9.4.4 HFS Level Splitting . . . . . . . . . . . . . . . . . . . . 480
9.5 Magnetic Resonance Spectroscopy . . . . . . . . . . . . . . . . 482
9.5.1 Molecular Beam Resonance Spectroscopy . . . . . . . . 482
9.5.2 EPR Spectroscopy . . . . . . . . . . . . . . . . . . . . . 484
9.5.3 NMR Spectroscopy . . . . . . . . . . . . . . . . . . . . 487
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
10 Multi-electron Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . 495
10.1 Central Field Approximation . . . . . . . . . . . . . . . . . . . 496
10.1.1 Hamiltonian for a Multi-electron System . . . . . . . . . 496
10.1.2 Centrally Symmetric Potential . . . . . . . . . . . . . . . 497
10.1.3 HARTREE Equations and SCF Method . . . . . . . . . . 498
10.1.4 HARTREE Method . . . . . . . . . . . . . . . . . . . . . 500
10.1.5 THOMAS-FERMI Potential . . . . . . . . . . . . . . . . . 501
10.2 HARTREE-FOCK Method . . . . . . . . . . . . . . . . . . . . . 503
10.2.1 PAULI Principle and SLATER Determinant . . . . . . . . 503
10.2.2 HARTREE-FOCK Equations . . . . . . . . . . . . . . . . 506
10.2.3 Configuration Interaction (CI) . . . . . . . . . . . . . . . 508
10.2.4 KOOPMANs Theorem . . . . . . . . . . . . . . . . . . . 509
10.3 Density Functional Theory . . . . . . . . . . . . . . . . . . . . 510
10.4 Complex Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . 512
10.4.1 Spin-Orbit Interaction and Coupling Schemes . . . . . . 512
10.4.2 Examples of Complex Spectra . . . . . . . . . . . . . . . 514
10.5 X-Ray Spectroscopy and Photoionization . . . . . . . . . . . . . 519
10.5.1 Absorption and Emission from Inner Shells . . . . . . . . 520
10.5.2 Characteristic X-Ray Spectra MOSLEYs Law . . . . . 522
10.5.3 Cross Sections for X-Ray Ionization . . . . . . . . . . . 524
10.5.4 Photoionization at Intermediate Energies . . . . . . . . . 527
10.6 Sources for X-Rays . . . . . . . . . . . . . . . . . . . . . . . . 530
10.6.1 X-Ray Tubes . . . . . . . . . . . . . . . . . . . . . . . . 530
10.6.2 Synchrotron Radiation, Introduction . . . . . . . . . . . 531
Contents of Volume 1 xxiii
10.6.3 Synchrotron Radiation, Quantitative Relations . . . . . . 536

10.6.4 Undulators and Wigglers . . . . . . . . . . . . . . . . . 540
10.6.5 Free Electron Laser (FEL) . . . . . . . . . . . . . . . . . 542
10.6.6 Relativistic THOMSON Scattering . . . . . . . . . . . . . 543
10.6.7 Laser Based X-Ray Sources . . . . . . . . . . . . . . . . 543
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546
Appendices
Appendix A Constants, Units and Conversions . . . . . . . . . . . . . 551
A.1 Fundamental Physical Constants and Units . . . . . . . . . . . . 551
A.2 SI and Atomic Units . . . . . . . . . . . . . . . . . . . . . . . . 553
A.3 SI and GAUSS Units . . . . . . . . . . . . . . . . . . . . . . . . 554
A.4 Radian and Steradian . . . . . . . . . . . . . . . . . . . . . . . 554
A.5 Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . 556
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
Appendix B Angular Momenta, 3j and 6j Symbols . . . . . . . . . . 559
B.1 Angular Momenta . . . . . . . . . . . . . . . . . . . . . . . . . 559
B.1.1 General Definitions . . . . . . . . . . . . . . . . . . . . 559
B.1.2 Orbital Angular Momenta Spherical Harmonics . . . . 562
B.2 Coupling of Two Angular Momenta: CLEBSCH-GORDAN
Coefficients and 3j Symbols . . . . . . . . . . . . . . . . . . . . 564
B.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 564
B.2.2 Orthogonality and Symmetries . . . . . . . . . . . . . . 565
B.2.3 General Formulae . . . . . . . . . . . . . . . . . . . . . 566
B.2.4 Special Cases . . . . . . . . . . . . . . . . . . . . . . . 567
B.3 RACAH Function and 6j Symbols . . . . . . . . . . . . . . . . . 568
B.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 568
B.3.2 Orthogonality and Symmetries . . . . . . . . . . . . . . 569
B.3.3 General Formulae . . . . . . . . . . . . . . . . . . . . . 570
B.3.4 Special Cases . . . . . . . . . . . . . . . . . . . . . . . 571
B.4 Four Angular Momenta and 9j Symbols . . . . . . . . . . . . . 571
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
Appendix C Matrix Elements . . . . . . . . . . . . . . . . . . . . . . . 575
C.1 Tensor Operators . . . . . . . . . . . . . . . . . . . . . . . . . . 575
C.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 575
C.1.2 WIGNER-ECKART Theorem . . . . . . . . . . . . . . . . 576
C.2 Products of Tensor Operators . . . . . . . . . . . . . . . . . . . 578
C.2.1 Products of Spherical Harmonics . . . . . . . . . . . . . 579
C.2.2 Matrix Elements of the Spherical Harmonics . . . . . . . 580
C.3 Reduction of Matrix Elements . . . . . . . . . . . . . . . . . . . 582
xxiv Contents of Volume 1
C.3.1 Matrix Elements of the Spherical Harmonics in LS

Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . 583
C.3.2 Scalar Products of Angular Momentum Operators . . . . 585
C.3.3 Components of Angular Momenta . . . . . . . . . . . . 586
C.4 Electromagnetically Induced Transitions . . . . . . . . . . . . . 587
C.4.1 Electric Dipole Transitions . . . . . . . . . . . . . . . . 588
C.4.2 Electric Quadrupole Transitions . . . . . . . . . . . . . . 588
C.4.3 Magnetic Dipole Transitions . . . . . . . . . . . . . . . 589
C.5 Radial Matrix Elements . . . . . . . . . . . . . . . . . . . . . . 590
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
Appendix D Parity and Reflection Symmetry . . . . . . . . . . . . . . 593
D.1 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
D.2 Multi-electron Systems . . . . . . . . . . . . . . . . . . . . . . 594
D.3 Reflection Symmetry of Orbitals Real and Complex Basis States 595
D.4 Reflection Symmetry in the General Case . . . . . . . . . . . . . 599
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603
Appendix E Coordinate Rotation . . . . . . . . . . . . . . . . . . . . . 605
E.1 EULER Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . 605
E.2 Rotation Matrices . . . . . . . . . . . . . . . . . . . . . . . . . 606
E.3 Entangled States . . . . . . . . . . . . . . . . . . . . . . . . . . 609
E.4 Real Rotation Matrices . . . . . . . . . . . . . . . . . . . . . . 610
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 611
Appendix F Multipole Expansions and Multipole Moments . . . . . . 613
F.1 Laplace Expansion . . . . . . . . . . . . . . . . . . . . . . . . . 613
F.2 Electrostatic Potential . . . . . . . . . . . . . . . . . . . . . . . 614
F.3 Multipole Tensor Operators . . . . . . . . . . . . . . . . . . . . 616
F.3.1 The Quadrupole Tensor . . . . . . . . . . . . . . . . . . 617
F.3.2 General Multipole Tensor Operators . . . . . . . . . . . 619
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
Appendix G Convolutions and Correlation Functions . . . . . . . . . 623
G.1 Definition and Motivation . . . . . . . . . . . . . . . . . . . . . 623
G.2 Correlation Functions and Degree of Coherence . . . . . . . . . 625
G.3 Gaussian Profile . . . . . . . . . . . . . . . . . . . . . . . . . . 626
G.4 Hyperbolic Secant . . . . . . . . . . . . . . . . . . . . . . . . . 627
G.5 LORENTZ Profile . . . . . . . . . . . . . . . . . . . . . . . . . . 628
G.6 VOIGT Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . 628
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 629
Contents of Volume 1 xxv
Appendix H Vector Potential, Dipole Approximation, Oscillator

Strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
H.1 Interaction of the Field of an Electromagnetic Wave with an
Electron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
H.1.1 Vector Potential . . . . . . . . . . . . . . . . . . . . . . 631
H.1.2 Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . 632
H.1.3 Static Magnetic Field . . . . . . . . . . . . . . . . . . . 633
H.1.4 Relation Between Matrix Elements of p and r . . . . . . 634
H.1.5 Ponderomotive Potential . . . . . . . . . . . . . . . . . . 634
H.1.6 Series Expansion of the Perturbation and the Dipole
Approximation . . . . . . . . . . . . . . . . . . . . . . . 635
H.2 Line Strength and Oscillator Strength . . . . . . . . . . . . . . . 636
H.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 636
H.2.2 THOMAS-REICHE-KUHN Sum Rule . . . . . . . . . . . 639
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
Appendix I FOURIER Transforms and Spectral Distributions of Light 643
I.1 Short Summary on FOURIER Transforms . . . . . . . . . . . . . 643
I.2 How Electromagnetic Fields are Written . . . . . . . . . . . . . 646
I.3 The Intensity Spectrum . . . . . . . . . . . . . . . . . . . . . . 648
I.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
I.4.1 Gaussian Distribution . . . . . . . . . . . . . . . . . . . 650
I.4.2 Hyperbolic Secant . . . . . . . . . . . . . . . . . . . . . 651
I.4.3 Rectangular Wave-Train . . . . . . . . . . . . . . . . . . 652
I.4.4 Rectangular Spectrum . . . . . . . . . . . . . . . . . . . 652
I.4.5 Exponential and LORENTZ Distributions . . . . . . . . . 653
I.5 Fourier Transform in Three Dimensions . . . . . . . . . . . . . 655
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
Appendix J Continuum . . . . . . . . . . . . . . . . . . . . . . . . . . 659
J.1 Normalization of Continuum Wave Functions . . . . . . . . . . 659
J.2 Plane Waves in 3D . . . . . . . . . . . . . . . . . . . . . . . . . 661
J.2.1 Expansion in Spherical Harmonics . . . . . . . . . . . . 661
J.2.2 Normalization in Momentum and Energy Scale . . . . . 662
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
Index of Volume 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665
Index of Volume 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 679
Contents of Volume 2
1 Lasers, Light Beams and Light Pulses . . . . . . . . . . . . . . . . 1

1.1 Lasers A Brief Introduction . . . . . . . . . . . . . . . . . . . 1
1.1.1 Basic Principle . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 FABRY-P ROT Resonator . . . . . . . . . . . . . . . . . 4
1.1.3 Stable, Transverse Modes and Diffraction Losses . . . . . 6
1.1.4 The Amplifying Medium . . . . . . . . . . . . . . . . . 9
1.1.5 Threshold Condition and Stationary State . . . . . . . . . 11
1.1.6 Laser Rate Equations . . . . . . . . . . . . . . . . . . . 12
1.1.7 Line Profiles and Hole Burning . . . . . . . . . . . . . . 14
1.2 Gaussian Beams . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.2.1 Diffraction Limited Profile of a Laser Beam . . . . . . . 17
1.2.2 FAUNHOFER Diffraction . . . . . . . . . . . . . . . . . . 23
1.2.3 Ray Transfer Matrices . . . . . . . . . . . . . . . . . . . 26
1.2.4 Focussing a Gaussian Beam . . . . . . . . . . . . . . . . 29
1.2.5 Measuring Beam Profiles with a Razor Blade . . . . . . . 34
1.2.6 The M 2 Factor . . . . . . . . . . . . . . . . . . . . . . . 34
1.3 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.3.1 Polarization and Time Dependent Intensity . . . . . . . . 36
1.3.2 Lambda-Quarter and Half-Wave Plates . . . . . . . . . . 38
1.3.3 S TOKES Parameters, Partially Polarized Light . . . . . . 41
1.4 Wave-Packets . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.4.1 Description of Laser Pulses . . . . . . . . . . . . . . . . 45
1.4.2 Spatial and Temporal Intensity Distribution . . . . . . . . 49
1.4.3 Frequency Combs . . . . . . . . . . . . . . . . . . . . . 49
1.5 Measuring Durations of Short Laser Pulses . . . . . . . . . . . . 52
1.5.1 Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 52
1.5.2 Correlation Functions . . . . . . . . . . . . . . . . . . . 53
1.5.3 Interferometric Measurement . . . . . . . . . . . . . . . 54
1.5.4 Experimental Examples . . . . . . . . . . . . . . . . . . 59
1.6 Nonlinear Processes in Gaussian Laser Beams . . . . . . . . . . 61
1.6.1 General Considerations . . . . . . . . . . . . . . . . . . 61
xxvii
xxviii Contents of Volume 2
1.6.2 Cylindrical Geometry (2D Geometry) . . . . . . . . . . . 63

1.6.3 Conical Geometry (3D Geometry) . . . . . . . . . . . . 65
1.6.4 Spatially Resolved Measurements . . . . . . . . . . . . . 66
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2 Coherence and Photons . . . . . . . . . . . . . . . . . . . . . . . . 71
2.1 Some Basics for Quantum Optics . . . . . . . . . . . . . . . . . 72
2.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 72
2.1.2 First-Order Degree of Coherence . . . . . . . . . . . . . 72
2.1.3 Quasi-Monochromatic Light . . . . . . . . . . . . . . . 75
2.1.4 Temporal or Longitudinal Coherence . . . . . . . . . . . 78
2.1.5 Higher-Order Degree of Coherence . . . . . . . . . . . . 82
2.1.6 Photon Bunching Experiments . . . . . . . . . . . . . 84
2.1.7 Spatial or Lateral Coherence . . . . . . . . . . . . . . . 86
2.1.8 Astronomical Interferometry . . . . . . . . . . . . . . . 91
2.1.9 H ANBURY B ROWN -T WISS Stellar Interferometer . . . . 95
2.1.10 Bunching and Anti-Bunching . . . . . . . . . . . . . . . 98
2.2 Photons, Photon States, and Radiation Modes . . . . . . . . . . 100
2.2.1 Towards Quantization of the Radiation Field . . . . . . . 100
2.2.2 Modes of the Radiation Field . . . . . . . . . . . . . . . 102
2.2.3 Density of States and Black Body Radiation . . . . . . . 105
2.2.4 Number of Photons per Mode . . . . . . . . . . . . . . . 106
2.2.5 The Multi-Mode Field and Energy . . . . . . . . . . . . 109
2.3 Field Quantization and Optical Transitions . . . . . . . . . . . . 110
2.3.1 Second Quantization and Photon Number States . . . . . 110
2.3.2 The Electric Field Operator . . . . . . . . . . . . . . . . 113
2.3.3 G LAUBER States . . . . . . . . . . . . . . . . . . . . . . 114
2.3.4 Addendum for Multi-Mode States . . . . . . . . . . . . . 117
2.3.5 Interaction Hamiltonian for Dipole Transitions . . . . . . 118
2.3.6 Perturbation Theory and Spontaneous Emission . . . . . 121
2.3.7 Spontaneous Emission in a Cavity . . . . . . . . . . . . 127
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3 Diatomic Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.1 Characteristic Energies . . . . . . . . . . . . . . . . . . . . . . 136
3.1.1 Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . 137
3.1.2 Electronic Energy . . . . . . . . . . . . . . . . . . . . . 138
3.1.3 Vibrational Energy . . . . . . . . . . . . . . . . . . . . . 138
3.1.4 Rotational Energy . . . . . . . . . . . . . . . . . . . . . 139
3.2 B ORN O PPENHEIMER Approximation . . . . . . . . . . . . . . 139
3.2.1 Molecular Potentials . . . . . . . . . . . . . . . . . . . . 139
3.2.2 General Form of Molecular Potentials . . . . . . . . . . 141
3.2.3 Nuclear Wave Functions . . . . . . . . . . . . . . . . . . 142
3.2.4 Harmonic Potential and Harmonic Oscillator . . . . . . . 143
Contents of Volume 2 xxix
3.2.5 M ORSE Potential . . . . . . . . . . . . . . . . . . . . . 145

3.2.6 VAN DER WAALS Molecules . . . . . . . . . . . . . . . 148
3.3 Nuclear Motion: Rotation and Vibration . . . . . . . . . . . . . 151
3.3.1 S CHRDINGER Equation . . . . . . . . . . . . . . . . . 151
3.3.2 Rigid Rotor . . . . . . . . . . . . . . . . . . . . . . . . 152
3.3.3 Population of Rotational Levels and Nuclear Spin . . . . 155
3.3.4 Specific Heat Capacity . . . . . . . . . . . . . . . . . . . 158
3.3.5 Vibration . . . . . . . . . . . . . . . . . . . . . . . . . . 160
3.3.6 Non-Rigid Rotor . . . . . . . . . . . . . . . . . . . . . . 163
3.3.7 D UNHAM Coefficients . . . . . . . . . . . . . . . . . . . 165
3.4 Dipole Transitions . . . . . . . . . . . . . . . . . . . . . . . . . 166
3.4.1 Rotational Transitions . . . . . . . . . . . . . . . . . . . 167
3.4.2 Centrifugal Distortion . . . . . . . . . . . . . . . . . . . 170
3.4.3 S TARK Effect: Polar Molecules in an Electric Field . . . 171
3.4.4 Vibrational Transitions . . . . . . . . . . . . . . . . . . 172
3.4.5 Vibration-Rotation Spectra . . . . . . . . . . . . . . . . 174
3.4.6 RYDBERG -K LEIN -R EES Method . . . . . . . . . . . . . 177
3.5 Molecular Orbitals . . . . . . . . . . . . . . . . . . . . . . . . . 179
3.5.1 Variational Method . . . . . . . . . . . . . . . . . . . . 179
3.5.2 Specialization for H+ 2 . . . . . . . . . . . . . . . . . . . 181
3.5.3 Charge Exchange in the H+ 2 System . . . . . . . . . . . . 184
3.5.4 MOs for Homonuclear Molecules . . . . . . . . . . . . . 190
3.6 Construction of Total Angular Momentum States . . . . . . . . . 197
3.6.1 Total Orbital Angular Momentum . . . . . . . . . . . . . 197
3.6.2 Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
3.6.3 Total Angular Momentum . . . . . . . . . . . . . . . . . 198
3.6.4 H UNDs Coupling Cases . . . . . . . . . . . . . . . . . . 199
3.6.5 Reflection Symmetry . . . . . . . . . . . . . . . . . . . 201
3.6.6 Lambda-Type Doubling . . . . . . . . . . . . . . . . . . 205
3.6.7 Example H2 MO Ansatz . . . . . . . . . . . . . . . . . 206
3.6.8 Valence Bond Theory . . . . . . . . . . . . . . . . . . . 210
3.6.9 Nitrogen and Oxygen Molecule . . . . . . . . . . . . . . 211
3.7 Heteronuclear Molecules . . . . . . . . . . . . . . . . . . . . . 215
3.7.1 Energy Terms . . . . . . . . . . . . . . . . . . . . . . . 215
3.7.2 Filling the Orbitals with Electrons . . . . . . . . . . . . . 216
3.7.3 Lithiumhydrid . . . . . . . . . . . . . . . . . . . . . . . 218
3.7.4 Alkali Halides: Ionic Bonding . . . . . . . . . . . . . . . 220
3.7.5 Nitrogen Monoxide, NO . . . . . . . . . . . . . . . . . . 224
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
4 Polyatomic Molecules . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.1 Rotation of Polyatomic Molecules . . . . . . . . . . . . . . . . 231
4.1.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.1.2 Spherical Rotor . . . . . . . . . . . . . . . . . . . . . . 234
xxx Contents of Volume 2
4.1.3 Symmetric Rigid Rotor . . . . . . . . . . . . . . . . . . 234

4.1.4 Asymmetric Rigid Rotor . . . . . . . . . . . . . . . . . . 236
4.2 Vibrational Modes of Polyatomic Molecules . . . . . . . . . . . 239
4.2.1 Normal Modes of Vibration . . . . . . . . . . . . . . . . 239
4.2.2 Energies and Transitions of Normal Modes . . . . . . . . 242
4.2.3 Linear, Triatomic Molecules AB2 . . . . . . . . . . . . . 243
4.2.4 Nonlinear Triatomic Molecules AB2 . . . . . . . . . . . 245
4.2.5 Inversion Vibration in Ammonia . . . . . . . . . . . . . 247
4.3 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
4.3.1 Symmetry Operations and Elements . . . . . . . . . . . 253
4.3.2 Point Groups . . . . . . . . . . . . . . . . . . . . . . . . 254
4.3.3 Eigenstates of Polyatomic Molecules . . . . . . . . . . . 257
4.3.4 JAHN -T ELLER Effect . . . . . . . . . . . . . . . . . . . 262
4.4 Electronic States of Some Polyatomic Molecules . . . . . . . . . 266
4.4.1 A First Example: H2 O . . . . . . . . . . . . . . . . . . . 266
4.4.2 Hybridization sp 3 Orbitals . . . . . . . . . . . . . . . 270
4.4.3 Electronic States of NH3 . . . . . . . . . . . . . . . . . 274
4.4.4 sp 2 Hybrid Orbitals Forming Double Bonds . . . . . . . 275
4.4.5 Triple Bonds . . . . . . . . . . . . . . . . . . . . . . . . 276
4.5 Conjugated Molecules and the H CKEL Method . . . . . . . . . 277
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
5 Molecular Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . 289
5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
5.2 Microwave Spectroscopy . . . . . . . . . . . . . . . . . . . . . 292
5.3 Infrared Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . 296
5.3.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
5.3.2 F OURIER Transform Infrared Spectroscopy . . . . . . . . 298
5.3.3 Infrared Action Spectroscopy . . . . . . . . . . . . . . . 302
5.4 Electronic Spectra . . . . . . . . . . . . . . . . . . . . . . . . . 305
5.4.1 F RANCK -C ONDON Factors . . . . . . . . . . . . . . . . 305
5.4.2 Selection Rules for Electronic Transitions . . . . . . . . 309
5.4.3 Radiationless Transitions . . . . . . . . . . . . . . . . . 311
5.4.4 Rotational Excitation in Electronic Transitions . . . . . . 312
5.4.5 Classical Emission and Absorption Spectroscopy . . . . . 314
5.5 Laser Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . 317
5.5.1 Laser Induced Fluorescence . . . . . . . . . . . . . . . . 317
5.5.2 REMPI for a Simple Triatomic Molecule . . . . . . . . 320
5.5.3 Cavity Ring Down Spectroscopy . . . . . . . . . . . . . 327
5.5.4 Spectroscopy of Small Free Biomolecules . . . . . . . . 328
5.5.5 Other Important Methods . . . . . . . . . . . . . . . . . 333
5.6 R AMAN Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . 334
5.6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 334
5.6.2 Classical Interpretation . . . . . . . . . . . . . . . . . . 337
Contents of Volume 2 xxxi
5.6.3 Quantum Mechanical Theory . . . . . . . . . . . . . . . 338

5.6.4 Experimental Aspects . . . . . . . . . . . . . . . . . . . 342
5.6.5 Examples of R AMAN Spectra . . . . . . . . . . . . . . . 343
5.6.6 Nuclear Spin Statistics . . . . . . . . . . . . . . . . . . . 345
5.7 Nonlinear Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 348
5.7.1 Some Basics . . . . . . . . . . . . . . . . . . . . . . . . 349
5.7.2 An Example . . . . . . . . . . . . . . . . . . . . . . . . 353
5.8 Photoelectron Spectroscopy . . . . . . . . . . . . . . . . . . . . 355
5.8.1 Experimental Basis and the Principle of PES . . . . . . . 355
5.8.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 358
5.8.3 TPES, PFI, ZEKE, KETOF, MATI . . . . . . . . . . . . 362
5.8.4 PES for Negative Ions . . . . . . . . . . . . . . . . . . . 364
5.8.5 PEPICO, TPEPICO and Variations . . . . . . . . . . . . 366
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
6 Basics of Atomic Collision Physics: Elastic Processes . . . . . . . . 383
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
6.1.1 Integral and Total Cross Sections . . . . . . . . . . . . . 385
6.1.2 Principle of Detailed Balance . . . . . . . . . . . . . . . 387
6.1.3 Integral Elastic Cross Sections . . . . . . . . . . . . . . 389
6.2 Differential Cross Sections and Kinematics . . . . . . . . . . . . 393
6.2.1 Experimental Considerations . . . . . . . . . . . . . . . 393
6.2.2 Collision Kinematics . . . . . . . . . . . . . . . . . . . 396
6.2.3 Mass Selection of Atomic Clusters . . . . . . . . . . . . 400
6.3 Elastic Scattering and Classical Theory . . . . . . . . . . . . . . 402
6.3.1 The Differential Cross Section . . . . . . . . . . . . . . 402
6.3.2 The Optical Rainbow . . . . . . . . . . . . . . . . . . . 403
6.3.3 The Classical Deflection Function . . . . . . . . . . . . . 405
6.3.4 Rainbows and Other Remarkable Oscillations . . . . . . 409
6.4 Quantum Theory of Elastic Scattering . . . . . . . . . . . . . . . 418
6.4.1 General Formalism . . . . . . . . . . . . . . . . . . . . 418
6.4.2 Angular Momentum and Impact Parameter . . . . . . . . 421
6.4.3 Partial Wave Expansion . . . . . . . . . . . . . . . . . . 422
6.4.4 Semiclassical Approximation . . . . . . . . . . . . . . . 425
6.4.5 Scattering Phase Shifts at Low Kinetic Energies . . . . . 428
6.4.6 Scattering Matrices for Pedestrians . . . . . . . . . . . . 432
6.5 Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
6.5.1 Types and Phenomena . . . . . . . . . . . . . . . . . . . 436
6.5.2 Formalism . . . . . . . . . . . . . . . . . . . . . . . . . 438
6.5.3 An Example: Electron Helium Scattering . . . . . . . . . 441
6.6 B ORN Approximation . . . . . . . . . . . . . . . . . . . . . . . 444
6.6.1 Scattering Amplitude and Cross Section in FBA . . . . . 445
6.6.2 RUTHERFORD Scattering . . . . . . . . . . . . . . . . . 446
6.6.3 B ORN Approximation for Phase Shifts . . . . . . . . . . 447
xxxii Contents of Volume 2

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
7 Inelastic Collisions A First Overview . . . . . . . . . . . . . . . . 453
7.1 Simple Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
7.1.1 Reactions Without Threshold Energy . . . . . . . . . . . 453
7.1.2 The Absorbing Sphere Model . . . . . . . . . . . . . . . 455
7.1.3 An Example: Charge Exchange . . . . . . . . . . . . . . 456
7.1.4 M ASSEY Criterium for Inelastic Collisions . . . . . . . . 457
7.2 Excitation Functions . . . . . . . . . . . . . . . . . . . . . . . . 460
7.2.1 Impact Excitation by Electrons and Protons . . . . . . . . 460
7.2.2 Electron Impact Excitation of He . . . . . . . . . . . . . 461
7.2.3 Finer Details in e + He Impact Excitation . . . . . . . . 464
7.2.4 Electron Collisions with Rare Gases . . . . . . . . . . . 465
7.2.5 Electron Impact at Atomic Mercury The
F RANCK -H ERTZ Experiment . . . . . . . . . . . . . . . 466
7.2.6 Molecular Excitation by Electron Impact . . . . . . . . . 468
7.2.7 Threshold Laws for Excitation and Ionization . . . . . . 470
7.3 Scattering Theory for the Multichannel Problem . . . . . . . . . 472
7.3.1 General Formulation of the Problem . . . . . . . . . . . 472
7.3.2 Potential Matrix and Coupling Elements . . . . . . . . . 477
7.3.3 The Adiabatic Representation . . . . . . . . . . . . . . . 478
7.3.4 The Diabatic Representation . . . . . . . . . . . . . . . . 480
7.4 Semiclassical Approximation . . . . . . . . . . . . . . . . . . . 484
7.4.1 Time Dependent S CHRDINGER Equation . . . . . . . . 484
7.4.2 Coupling Elements . . . . . . . . . . . . . . . . . . . . . 486
7.4.3 Solution of the Coupled Differential Equations . . . . . . 487
7.4.4 L ANDAU -Z ENER Formula . . . . . . . . . . . . . . . . . 489
7.4.5 A Simple Example: Na+ + Na(3p) . . . . . . . . . . . . 492
7.4.6 S TCKELBERG Oscillations . . . . . . . . . . . . . . . . 496
7.5 Collision Processes with Highly Charged Ions (HCI) . . . . . . . 499
7.5.1 Above-Barrier Model . . . . . . . . . . . . . . . . . . . 501
7.5.2 An Experiment on Electron Exchange . . . . . . . . . . 504
7.5.3 HCI Collisions and Ultrafast Dynamics . . . . . . . . . . 506
7.6 Surface Hopping, Conical Intersections and Reactions . . . . . . 506
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511
8 Electron Impact Excitation and Ionization . . . . . . . . . . . . . . 515
8.1 Formal Scattering Theory and Applications . . . . . . . . . . . . 515
8.1.1 Close-Coupling Equations . . . . . . . . . . . . . . . . . 516
8.1.2 Theoretical Methods and Experimental Examples . . . . 520
8.2 B ORN Approximation for Inelastic Collisions . . . . . . . . . . 525
8.2.1 FBA Scattering Amplitude . . . . . . . . . . . . . . . . 525
8.2.2 Cross Sections . . . . . . . . . . . . . . . . . . . . . . . 527
8.2.3 B ORN Approximation and RUTHERFORD Scattering . . . 528
Contents of Volume 2 xxxiii
8.2.4 An Example . . . . . . . . . . . . . . . . . . . . . . . . 529

8.3 Generalized Oscillator Strength . . . . . . . . . . . . . . . . . . 530
8.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 530
8.3.2 Expansion for Small Momentum Transfer . . . . . . . . 531
8.3.3 Explicit Evaluation of GOS for an Example . . . . . . . 533
8.3.4 Integral Inelastic Cross Sections . . . . . . . . . . . . . . 534
8.4 Electron Impact Ionization . . . . . . . . . . . . . . . . . . . . . 534
8.4.1 Integral Cross Sections and the L OTZ Formula . . . . . . 537
8.4.2 SDCS: Energy Partitioning Between the Electrons . . . . 539
8.4.3 Behaviour at the Ionization Threshold . . . . . . . . . . 540
8.4.4 DDCS: Double-Differential Cross Section
and the B ORN -B ETHE Approximation . . . . . . . . . . 544
8.4.5 TDCS: Triple-Differential Cross Sections . . . . . . . . . 549
8.4.6 Electron Momentum Spectroscopy (EMS) . . . . . . . . 558
8.5 Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
8.5.1 Direct and Dielectronic Recombination . . . . . . . . . . 563
8.5.2 The Merged-Beams Method . . . . . . . . . . . . . . . . 564
8.5.3 Some Results . . . . . . . . . . . . . . . . . . . . . . . 565
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
9 The Density Matrix A First Approach . . . . . . . . . . . . . . . 573
9.1 Some Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 575
9.1.1 Pure and Mixed States . . . . . . . . . . . . . . . . . . . 575
9.1.2 Density Operator and Density Matrix . . . . . . . . . . . 581
9.1.3 Matrix Representation for Selected Examples . . . . . . 582
9.1.4 Coherence and Degree of Polarization . . . . . . . . . . 585
9.2 Theory of Measurement . . . . . . . . . . . . . . . . . . . . . . 588
9.2.1 State Selector and Analyzer . . . . . . . . . . . . . . . . 588
9.2.2 Interaction Experiment with State Selection . . . . . . . 590
9.3 Selected Examples of the Density Matrix . . . . . . . . . . . . . 596
9.3.1 Polarization Matrix and S TOKES Parameters . . . . . . . 596
9.3.2 Atom in an Isolated 1 P State . . . . . . . . . . . . . . . . 602
9.4 Angular Distribution and Polarization of Radiation . . . . . . . . 611
9.4.1 Formulation of the Problem . . . . . . . . . . . . . . . . 611
9.4.2 General Discussion . . . . . . . . . . . . . . . . . . . . 616
9.4.3 Details of the Evaluation . . . . . . . . . . . . . . . . . . 619
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
10 Optical B LOCH Equations . . . . . . . . . . . . . . . . . . . . . . . 625
10.1 Open Questions . . . . . . . . . . . . . . . . . . . . . . . . . . 625
10.2 Two Level System in Quasi-Monochromatic Light . . . . . . . . 629
10.2.1 Dressed States . . . . . . . . . . . . . . . . . . . . . . . 629
10.2.2 R ABI Frequency . . . . . . . . . . . . . . . . . . . . . . 630
10.2.3 Rotating Wave Approximation . . . . . . . . . . . . . . 631
xxxiv Contents of Volume 2
10.2.4 The Coupled System . . . . . . . . . . . . . . . . . . . . 633

10.3 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
10.3.1 M OLLOW Triplet . . . . . . . . . . . . . . . . . . . . . 635
10.3.2 AUTLER -T OWNES Effect . . . . . . . . . . . . . . . . . 637
10.4 Quantum Systems in Strong Electromagnetic Fields . . . . . . . 639
10.4.1 Temporal Evolution of the Density Matrix . . . . . . . . 639
10.4.2 Optical B LOCH Equations for a Two State System . . . . 640
10.5 Excitation with Continuous Wave (cw) Light . . . . . . . . . . . 642
10.5.1 Relaxed Steady State . . . . . . . . . . . . . . . . . . . 643
10.5.2 Saturation Broadening . . . . . . . . . . . . . . . . . . . 643
10.5.3 Broad Band and Narrow Band Excitation . . . . . . . . . 645
10.5.4 Rate Equations . . . . . . . . . . . . . . . . . . . . . . . 646
10.5.5 Continuous Excitation Without Relaxation . . . . . . . . 647
10.5.6 Continuous Excitation with Relaxation . . . . . . . . . . 648
10.6 B LOCH Equations and Short Pulse Spectroscopy . . . . . . . . . 649
10.6.1 Excitation with Ultrafast Laser Pulses . . . . . . . . . . . 649
10.6.2 Ultrafast Spectroscopy . . . . . . . . . . . . . . . . . . . 652
10.6.3 Rate Equations and Optical B LOCH Equations . . . . . . 653
10.7 STIRAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
10.7.1 Three Level System in Two Laser Fields . . . . . . . . . 657
10.7.2 Energy Splitting and State Evolution . . . . . . . . . . . 659
10.7.3 Experimental Realization . . . . . . . . . . . . . . . . . 661
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666
Appendices
Appendix A First B ORN Approximation for e + Na(3s) e + Na(3p) 669
A.1 Evaluation of the Generalized Oscillator Strength . . . . . . . . 669
A.2 Integration of the Differential Cross Section . . . . . . . . . . . 672
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 672
Appendix B Guiding, Detecting and Energy Analysis of Electrons and
Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673
B.1 SEM, Channeltron, Microchannel Plate . . . . . . . . . . . . . . 673
B.2 Index of Refraction, Lenses and Directional Intensity . . . . . . 678
B.3 Hemispherical Energy Selector . . . . . . . . . . . . . . . . . . 680
B.4 Magnetic Bottle and Other Time of Flight Methods . . . . . . . . 683
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 686
Appendix C Statistical Tensor and State Multipoles . . . . . . . . . . 687
C.1 Multipole Expansion of the Density Matrix . . . . . . . . . . . . 687
C.2 State Multipoles and Expectation Values of Multipole Tensor
Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 690
C.3 Recoupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694
Contents of Volume 2 xxxv
Appendix D Optical Pumping . . . . . . . . . . . . . . . . . . . . . . . 695

D.1 A Standard Case: Na(3 2 S1/2 3 2 P3/2 ) . . . . . . . . . . . . . 695
D.2 Multipole Moments and Their Experimental Detection . . . . . . 698
D.3 Optical Pumping with Two Frequencies . . . . . . . . . . . . . . 700
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 702
Index of Volume 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 703
Index of Volume 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715
About the Authors
Ingolf V. Hertel was born in 1941 in Dresden, 1967

Diplom in Physics, Universitt Freiburg, Ph.D. thesis
at the University of Southampton UK, 1969 Dr. rer. nat.
Universitt Freiburg, Assistent University Mainz, 1970
Associate Professor University Kaiserslautern, 1978
Full Professor for Experimental Physics Freie Univer-
sitt Berlin, 1986 Full Professor Universitt Freiburg,
Extended Research Periods at JILA University of Col-
orado Boulder USA and Orsay France, 1992 to 2009
Director at Max Born Institute for Nonlinear Optics and
Short Pulse Spectroscopy in Berlin-Adlershof, 1993 to
2009 also Full Professor FU Berlin, since 2010 Wil-
helm und Else Heraeus Senior Professor for the En-
hancement of Teachers Education at Humboldt Univer-
sitt zu Berlin.
Claus-Peter Schulz was born in 1953 in Berlin, 1981

Diplom in Physics TU Berlin, 1987 Dr. rer. nat.
Freie Universitt Berlin, Postdoc at JILA University
of Colorado Boulder USA, 1988 Assistent Universitt
Freiburg, since 1993 Staff Scientist at Max Born In-
stitute for Nonlinear Optics and Short Pulse Spec-
troscopy in Berlin-Adlershof, Extended Research Peri-
ods at Universit Paris-Nord and Orsay France as well
as at JILA Boulder USA.
xxxvii
Basics
1
Here we provide a compact summary of the most important

concepts, experiments, observations, phenomena and models
that are essential in understanding matter on an atomic and
sub-atomic length-scale. These are taught typically to second
year undergraduates in physics. Readers who feel sufficiently
familiar with these basics of todays quantum view of matter
may safely skip the chapter.
Overview
Section 1.1 gives a brief survey of the canonical subject areas in physics, of
physics history and the quantum nature of atomic phenomena. An introduc-
tion to orders of magnitude of length, time and energy follows. Section 1.2
summarizes some essentials of special relativity. Section 1.3 introduces some
elements of statistical mechanics and thermodynamics. The photon, key par-
ticle in this text book, enters the scene in Sect. 1.4. Section 1.5 makes a very
short excursion into the nature of the four fundamental interactions and to
the standard model of elementary particles. Section 1.6 deals with the mun-
dane subject of how free, charged particles move under the influence of an
external electromagnetic field. Particles and waves (Sect. 1.7) and the B OHR
model of the H atom (Sect. 1.8) lead us to the foundations of modern physics.
Section 1.9 introduces one of the key concepts of quantum mechanics: space
quantization as discovered in the famous S TERN -G ERLACH experiment
which in turn led, more or less directly, to the discovery of the electron spin,
treated in Sect. 1.10.
1.1 Overview, History and Magnitudes
Table 1.1 gives a compact overview of the canonical subject areas of modern
physics and connects them to the content of these textbooks: marked in italics
are those themes that are at least partially treated here. Otherwise, the table is
more or less self explaining.
Springer-Verlag Berlin Heidelberg 2015 1

I.V. Hertel, C.-P. Schulz, Atoms, Molecules and Optical Physics 1,
Graduate Texts in Physics, DOI 10.1007/978-3-642-54322-7_1
2 1 Basics
Table 1.1 Standard subject areas of physics

Theory Canonical themes Applications and
of modern physics special subjects
Classical mechanics and Atomic physics Meteorology
special relativity Metrologya
Thermodynamics and statistics Molecular physics Chemical physics
Electrodynamics and optics Scattering physics
Quantum mechanics Quantum optics
Nonlinear optics
Quantum electrodynamics (QED) Lasers
Ultrafast physics
Cluster physics
Solid state physics Surface physics
(condensed matter) Semiconductor physics
Medical physics
Biophysics
Quantum field theory
Quantum chromodynamics Nuclear physics Reactor physics
General relativity Elementary particle physics
Grand unification Astrophysics Plasma physics
Quantum geometrodynamics Astro particle physics
a Scientific standards (units) and measuring techniques
At the beginning of the past century, the history of atomic, molecular and optical
(AMO) physics was almost identical to the history of physics at large. Without any
claim for completeness, we collect in Table 1.2 some milestones in the topical fields
of this textbook, and Table 1.3 highlights some key developments in theoretical
particle physics which we shall touch very briefly in Sect. 1.5. Of course, a host
of fascinating details during the development of modern physics cannot even be
mentioned here.
1.1.1 Quantum Nature of Matter
Based on experiments, many fundamental observations in physics can only be un-

derstood in terms of quantum mechanics. Some examples are
photoelectric effect (E INSTEIN 1905),

C OMPTON effect (C OMPTON 1922),
frequency distribution of blackbody radiation (P LANCK 1900),
diffraction and interference of particle beams i.e. the wave-particle duality (DE
B ROGLIE wavelength 1923),
specific heat capacity at low temperatures (E INSTEIN, D EBYE 1906),
line spectra of atoms (RYDBERG, B OHR 1913).
1.1 Overview, History and Magnitudes 3
Table 1.2 Highlights of science history from the idea of the atom to modern physics of atoms,
molecules and quantum optics (incomplete list)
400 BC D EMOCRITOS oo (indivisible)
1808 AD DALTON Multiple proportions
1811 AVOGADRO Molecular theory of gases
1814 F RAUNHOFER First useful spectrometer
1834 FARADAY Electrolysis (FARADAY constant)
1868 M ENDELEEV Periodic table of elements
1869 H ITTORF Cathode rays
1886 G OLDSTEIN Channel rays
1895 RNTGEN X-rays
1896 B ECQUEREL Radioactivity
1897 J.J. T HOMSON e/m for electrons
1898 Marie & Pierre C URIE Polonium, radium
1898 W IEN e/m for ions
1900 P LANCK E = h
1903 RUTHERFORD Atomic nuclei
1905 E INSTEIN E = mc2
1913 B OHR Atom model
1913 M ILLIKAN e-determination
19211922 S TERN & G ERLACH Space quantization
1925 Max B ORN (N OBEL prize 1954) Fundamental research in quantum
mechanics
1926 S CHRDINGER Wave equation
1927 H EISENBERG Uncertainty relation
1947 L AMB and R ETHERFORD L AMB shift for excited H
19581966 S CHAWLOW, T OWNES, Basov, P ROKHOROV, Maser, laser and spectroscopy
M AIMAN, JAVAN, K ASTLER
1986 N OBEL prize Dudley R. H ERSCHBACH, Dynamics of chemical elementary
Yuan T. L EE and John C. P OLANYI processesa
1989 N OBEL prize Norman G. R AMSEY, R AMSEY fringes, atomic clocksb , ion
Hans D EHMELT and Wolfgang PAUL trapsa
1996 N OBEL prize R. F. C URL Jr., H. K ROTO, Discovery of fullerenesa . . . C60 etc.
R. E. S MALLEY
1997 N OBEL prize S. C HU, Methods to cool and trap atoms with laser
C. C OHEN -TANNOUDJI, W. D. P HILLIPS light
1999 N OBEL prize Ahmed Z EWAIL Femto(second) chemistrya
2001 N OBEL prize Eric A. C ORNELL, Cold atoms and
Wolfgang K ETTERLE, Carl E. W IEMAN B OSE -E INSTEIN condensationa
2002 N OBEL prize John F ENN, Koichi TANAKA Electro spray, molecular beamsa , MALDI
mass spectroscopya
2005 N OBEL prize Roy G LAUBER, Theory of optical coherencec and
John H ALL and Theodor H NSCH laser precision spectroscopya
2007 N OBEL prize Gerhard E RTL Chemical processes at surfaces
a Work from several preceding years
b Work from the 1950s
c Work from the 1960s
4 1 Basics
Table 1.3 Theory on the way from electrodynamics to the standard model of the fundamental
interactions (the dates of N OBEL prizes given here refer to discoveries and developments which
typically happened much earlier)
ca. 1850 James Clerk M AXWELL Electrodynamics
1918 N OBEL prize Max P LANCK Energy quanta
1921 N OBEL prize Albert E INSTEIN Law of the photoelectric effect
1932 N OBEL prize Werner H EISENBERG Creation of quantum mechanics
1933 N OBEL prize Erwin S CHRDINGER, Paul D IRAC Wave equations for matter
1949 N OBEL prize Hideki Y UKAWA Prediction of mesons
1954 N OBEL prize Max B ORN Statistical interpretation
of quantum mechanics
1963 N OBEL prize W IGNER, G OEPERT-M AYER, J ENSEN Structure of the nucleus
1965 N OBEL prize T OMONAGA, S CHWINGER, F EYNMAN Quantum electrodynamics
1967 N OBEL prize B ETHE Theory of nuclear reactions
1969 N OBEL prize G ELL -M ANN Quark model
1979 N OBEL prize G LASHOW, W EINBERG, S ALAM Theory of weak interaction
1982 N OBEL prize W ILSON Re-normalization, critical
phenomena
1999 N OBEL prize T H OOFT, V ELTMAN Quantum structure
of electro-weak interaction
2004 N OBEL prize G ROSS, P OLITZER, W ILCZECK Asymptotic freedom of quarks
We encounter the phenomenon of quantization that will be a steady companion

during all volumes of this textbook. Of central importance is the relation between
the momentum p of a particle and its wavelength (or wave vector k) that quantifies
the so called wave-particle dualism by
p = h/ and p = k with k = 2/, (1.1)
with the fundamental P LANCK constant (see also Sect. 1.4.6):
h = 6.62606896(33) 1034 J s = 4.13566733(10) 1015 eV s. (1.2)
Closely related and supplementing the above we shall also have to discuss the quan-
tization of energy. As examples we mention the
energy of a photon (see Sect. 1.4.1):
Wph = h = (1.3)
energies of the H atom (see Sect. 1.8):
Wn = Eh /2n2 with n = 1, 2, 3, . . . . (1.4)

Equally important is the quantization of angular momentum and its direction in

space (so called space quantization) as we shall explicate later on various levels
of abstraction (see e.g. Sect. 1.9). At this point it may suffice to say that angular
momenta are measured in either integer or half integer multiples of . The following
equations summarize this:
absolute value of angular momentum:

J = j (j + 1)2 (1.5)
angular momentum quantum numbers

for bosons
j = 0, 1, 2, . . . (1.6)
for fermions
j = 1/2, 3/2, . . . (1.7)
(2j + 1) angular momentum projections:
Jz = mj (1.8)
with projection quantum numbers
mj = j, j + 1, . . . , j. (1.9)
Here J is the absolute value of the angular momentum, Jz its projection on a given
axis (here z) in space. One distinguishes orbital angular momentum (integer quan-
tum number) and intrinsic angular momentum or spin (integer for so called bosons,
half integer for so called fermions, see Sect. 1.5). To characterize these properties
of a given orbit or particle we shall refer to its angular momentum (or spin) as being
0, 1/2, 1, etc.
1.1.2 Orders of Magnitude
Before going into details, we want to give an orientation about orders of magnitude
for some relevant physical observables with which we shall have to deal. Orders
of magnitudes are always important in physics and the student is well advised to
roughly memorize some of these data. This will turn out to be very helpful when
trying to find her or his bearings later on when indulging into real, own measure-
ments. (A list of the most important fundamental constants of nature is provided in
Appendix A.)
Length Scales from Atomic Physics to Astrophysics

An overview about the whole length scale on which relevant physics occurs gives
Fig. 1.1. The smallest possible length unit one may construct from the known
fun-
damental constants is the so called P LANCK length P = /(mP c) = G/c3 =
6 1 Basics
see enlarged scale
distance to the moon

diameter of the earth
distance to the sun

molecules
e (Coulomb nano world
law)
diameter of
Planck length
our galaxy
bacteria
atomic nuclei
next star
universe
insects
human
atoms
virus
cells
1035 1015 1010 105 100 105 1010 1015 1020
dominant forces: length / m
strong interaction gravitation (among quasi neutral objects)
electroweak interaction
electromagnetic interaction
Fig. 1.1 Length scales in the universe and the range of dominance of the four fundamental forces:
strong, electro-weak, electromagnetic and gravitational interaction (see Sect. 1.5)
1.616252 1035 m, which may be viewed as describing the granularity of space

(with the P LANCK mass mP = c/G = 2.176437375 108 kg. Both, P and mP ,
are constructed from the reduced P LANCK constant , the gravitational constant G
and the velocity of light c (see Appendix A).
At the other end of the length scale are cosmic objects such as our galaxy (milky
way) with a diameter of ca. 100 000 light years (1 light year 9.46 1015 m) and
finally the known universe. Its age is currently estimated to be about 13.7 109 a
(see M ATHER and S MOOT 2006), from which one has to assume an upper limit of
its extension of 1.30 1026 m.
In contrast, atomic and molecular physics focus on objects in a length range
from 0.5 1010 m109 m. In this context it is useful to visualize typical dimen-
sions (radii) of the building blocks of matter as done very schematically in Fig. 1.2.
Elementary particles in the strict sense are only quarks and electrons, their exten-
sion being (if any) definitively below 1 am. While such length scales are in general
of little relevance to phenomena observed in atoms and molecules, it should be
Fig. 1.2 Blow up of the size size in am

length scale Fig. 1.1 for the
size range of building blocks 1010 m atom 100 000 000 am
of matter on the left side
given in m, right in attometers 1014 m
(1 am = 1018 m). The nucleus 10 000 am
images of elementary
proton, 1 000 am
particles are schematics only 1015 m neutron
(see also Sect. 1.5) (nucleons)
1018 m quark 1 am
electron
? ? (lepton)
homo sapiens
Planck time
universe
lifetime
0 meson
human life
one year
neutron
lifetime
earth
see enlarged scale
10-44 10-16 10-12 10-8 10-4 100 104 108 1012 1016
time / s
Fig. 1.3 Time scales in the universe. The range marked with the fat double arrow is shown in
Fig. 1.4 on an enlarged scale
pointed out that the finite extension of the atomic nuclei play a role (proton radius
ca. 0.881015 m) for ultra-high precision spectroscopy (see Chap. 9). On the other
end of the atomic length scale we have to be aware that the wavelength of electro-
magnetic radiation (spectral range of the visible, VIS, from 380 nm to 760 nm, see
Sect. 1.4.5) is of high relevance in all spectroscopic investigations.
Time Scales from Atomic Physics to Astrophysics

Typical time scales are communicated in Fig. 1.3. Again one may define a short-
est time (granularity of time), the P LANCK time as tP = P /c = G/c5 =
5.39124(27) 1044 s. For comparison, the mean lifetime of a 0 meson is
0.84 1016 s, atomic excited states live on the order of 109 s (see Chap. 4), and
the neutrons mean lifetime is 886 s (see Sect. 1.5.3).
On the other end of the scale we find the age of our earth: it exists now for
about 4.55 109 a. The universe that was created 13.7 109 a ago (according to
our present understanding, see M ATHER and S MOOT 2006). Out of these more than
60 decades, the time scale of interest for AMO physics as well as for technology,
chemistry, biology and medicine is illustrated in Fig. 1.4.
Energy Scales in Physics

Besides space and time, energy plays a key role in AMO physics. An orientation
gives Fig. 1.5 where the typical energy content of physical objects and excitation
energies of quantum systems are summarized. In our context mainly energies be-
tween some eV and some tens of keV are significant.
It is useful to characterize atomic and molecular phenomena according to their
characteristic interaction energies. All of them are in some way based on electro-
magnetic forces. In the following chapters we shall first treat the strongest interac-
tions and then refine our considerations step by step towards smaller perturbations
of the dominant effects. The quantitative treatment occurs then in the spirit of per-
turbation theory: one first formulates and solves the most simple problem and im-
proves the calculations by adding one perturbation after the other, ideally each some
orders of magnitude smaller than the former. The changes may then be calculated
8 1 Basics
electron motion molecular chemical reaction stop clock

in atoms and vibration / rotation (explosion)
molecules
primary
processes fast
of photo- digital electronics
synthesis life time of excited
atomic states fast camera shutter
ultrafast physics
1018 1015 1012 109 106 103 1
atto femto pico nano micro milli s
characteristic time scales / seconds (s)
Fig. 1.4 Enlarged time section from seconds to attoseconds (1018 s) from Fig. 1.3. A presently
very active area of research in AMO is ultrafast and attosecond physics
with excellent success by approximative methods. This hierarchy of perturbations is

summarized quantitatively in Table 1.4.
Corresponding to different spectroscopic precision one often uses different units
for characterizing the relevant interaction. Most often in spectroscopy the unit
wavenumber or cm1 is used (see Eq. (1.79)). Alternatively, energies in atoms and
molecules are measured in eV or in atomic units (a.u., see Sect. 1.8.3) of energy,
Eh = 27.211 eV. The upper end of the energy scale for AMO physics may be seen
as related to the rest mass energy of the electron me c2 0.511 MeV (with me being
the mass of the electron and c the speed of light in vacuum, see Sect. 1.2).
rest mass of e
electronic
W & Z boson
rest mass p+
excitation in atoms,
molecular molecules,
molecular vibrations solids
rotation
1eV 1meV 1eV 1 keV 1 MeV 1 GeV 1 TeV 1 PeV
1 nK 1 K 1 mK 1K 1000 K 1 MK 1000 MK
LHC lead
room
protons
Bose-Einstein CMBR tem- interior fusion

LHC
condensates perature of stars reactor
Fig. 1.5 Energy and equivalent temperature of quantum systems: the scale starts today at one
or several 100 pK, referring to the coldest B OSE -E INSTEIN condensates (BEC) and cosmic mi-
crowave background radiation (CMB(R)) at 2.725 K on the one end of the scale. On the other end
we find collisions artificially generated at the large hadron collider (LHC) with 14 TeV for protons
and more than 1000 TeV for lead nuclei. AMO is interested mainly in the red marked energy range
Table 1.4 Orders of magnitude of atomic interactions here for typical examples such as H, alkali
metal atoms and He; the structure of volume 1 of these textbooks essentially follows this energetic
scheme
Interaction Order of magnitude See
cm1 eV kHz K
Pure C OULOMB Z/r 30000 4 1015 43000 Chap. 2
Exchange (in He n = 2) 1000 to 6000 0.12 to 0.7 3 1010 1400 Chap. 7
to 1.81011 to 8600
C OULOMB screening 3000 0.4 1014 4300 Chap. 3
Fine structure (FS) 1 to 1000 104 to 0.1 3 1010 1.4 Chap. 6
to 3 1013 to 1400
External el. mag. fields 1 104 3 1010 1.4 Chap. 8
Hyperfine structure 103 to 1 107 to 104 3 107 1.4 103 Chap. 9
to 3 1010 to 1.4
The ratio of these two energies defines a dimensionless fundamental constant,

the so called fine structure constant

Eh e2 1
= = , (1.10)
me c2 40 c 137
which should be memorized. We shall often come across this very important con-
stant, typically indicating some influence of, or connection with special relativity.
The currently best value of , measured with very high accuracy, is reported in
Appendix A, based on 2010 CODATA (NIST 2010). There, one also finds precise
conversion factors between different units of energy.
We finally mention that one may, from fundamental constants, also construct a so
called P LANCK energy WP = c2 (c/G)1/2 = 1.221 1019 GeV. This is an energy
of cosmic magnitude which may be related to the first moments after the big bang.
There are indications for unification of the fundamental forces at such energy, i.e. at
least three coupling constants might become equal at this energy.
Section summary
The history of AMO physics is identical to the early history of physics at
large. We have recalled some basic observations documenting the quantum
nature of submicroscopic matter.
The essence of (any) physics is to obtain a quantitative description of nature.
Numbers and orders of magnitude are thus essential, and a good feeling for
the scales of lengths, time and energy is important for working in physics.
Section 1.1.2 gives a summary.
Specifically, we memorize P LANCKs constant h 6.63 1034 J s and the
dimensionless fine structure constant 1/137 which will be steady com-
panions on our voyage through AMO physics.
10 1 Basics
1.2 Special Theory of Relativity in a Nutshell
1.2.1 Kinematics and Dynamics
We cannot give here a serious introduction to the theory of relativity and assume
the reader to be somewhat familiar with E INSTEINs special theory of relativity.
Generally speaking, throughout this textbook relativistic effects will be treated as
a kind of afterthought where necessary (e.g. in Chap. 6). It is, however, useful to
recall here some formulas for later use.
We first remember that N EWTONs equations, specifically his second axiom, re-
main fully valid under relativistic conditions. Thus,
dp
=F (1.11)
dt
at high velocities v, has just to be applied to the relativistic momentum:

mv
p= = mv and p = mc = mc 2 1. (1.12)
1 2
Here m is the rest mass of a particle (its intrinsic mass in its rest frame, also called
invariant mass) and
1 v
= with = (1.13)
1 2 c
is the so called L ORENTZ factor. One may also rewrite (1.13) as
2 2 2 = 1 (1.14)
which is invariant under L ORENTZ transformation (it is a constant). For highly rel-
ativistic particles 1 one obtains

1 1/ 2 2 . (1.15)
Next we recall the equivalence of mass and energy. In the particles rest frame
the famous E INSTEIN relation reads
Wrest = mc2 . (1.16)

2 we obtain
If we multiply (1.14) on both sides by Wrest
2 m2 c4 2 2 m2 c4 = m2 c4 ,
and with (1.12) follows
2 m2 c4 p 2 c2 = m2 c4
W 2 p 2 c2 = m2 c4 . (1.17)
1.2 Special Theory of Relativity in a Nutshell 11
Here W is interpreted as total relativistic energy of the moving particle:1

W = mc = 2
m2 c4 + p 2 c2 or = W/mc2 . (1.19)
For the particle at rest (p = 0 and = 1) it becomes identical to the E INSTEIN

relation (1.16). Obviously with (1.13) the
fundamental relation = v/c < 1
must hold if divergence of energy is to be avoided: the velocity of a particle with

finite rest mass is always less than the speed of light. In contrast, particles without
rest mass, such as the photon, exist only moving at the speed of light.
We emphasize that for several interacting particles in a given frame of reference

relativistic energy conservation Wi = const
holds with energies Wi according to (1.19) to which potential energies may have
to be added due to internal or external fields. In addition

relativistic momentum conservation p i = const
holds, with momenta p i according to (1.12).

This may be summarized in most compact form by saying that the so called
momentum four-vector P is conserved in a closed system:

P= P i = const with P i = (Wi /c, p i ).
i
Note that
P 1 P 2 = W1 W2 /c2 p1 p 2
is the inner product of two four vectors P 1 and P 2 . The length of a four-vector is
invariant under L ORENTZ transformation (M INKOWSKI Norm):
1 One may abbreviate

m
mrel = = m (1.18)
1 2
and write the relativistic momentum and energy as
p = mrel v and W = mrel c2 ,
respectively. In modern theoretical literature the introduction of this relativistic mass is, however,
usually omitted to avoid confusion: it is the energy that changes with velocity, while the rest mass
m is L ORENTZ invariant.
12 1 Basics
W2
P2 = p2
c2
W 2 p 2 c2 = W 2 p 2 c2 = const. (1.20)
This expression describes a system of particles with total energy and total momen-
tum W , p in one frame, and W , p in an other frame of reference, respectively.
This is completely equivalent to (1.17), which was referring to a single particle and
its rest frame on the right hand of the equation. As an example, for two (or more)
particles the transformation from the laboratory system (lab) to the centre of mass
system (cm) would read
2
Wlab plab
2 2
c = Wcm
2
, (1.21)
where in the laboratory frame Wlab and p lab are the sums of relativistic energies and
momenta of all particles, respectively, while Wcm is the sum of all particle energies
in the centre of mass system in which per definition the sum of all momenta is
p cm = 0.
The kinetic energy T of a relativistic particle is defined by
T + mc2 = W = mc2 . (1.22)
The L ORENTZ factor may thus be written as

T
= 1 + wk with wk = . (1.23)
mc2
Often it is useful to know the velocity of a particle as a function of its kinetic energy.
With (1.23) and the definition (1.13) one finds

1 + wk /2 3
= 2wk 2wk 1 wk + . (1.24)
1 + wk 4
The approximate equality is an expansion for small kinetic energies. The relation
between the magnitude of the momentum p and the kinetic energy T is derived by
inserting (1.22) into (1.19) and squaring, so that
p 2 c2 = T 2 + 2T mc2 from which follows

T T
p = 2mT 1 + 2mT 1 + (1.25)
2mc2 4mc2

wk
p = mc 2wk 1 + wk /2 mc 2wk 1 + .
4
For small kinetic energies one recovers the classic relation T = p 2 /2m.
A nonrelativistic treatment is only possible if T mc2 . For an electron this is a
rather limited range of energies since

Wrest e = me c2 = 0.511 MeV. (1.26)
1.2 Special Theory of Relativity in a Nutshell 13
Fig. 1.6 Relativistic Doppler observer

shift in lab-system

moving source v
1.2.2 Time Dilation and LORENTZ Contraction
We want to transform times and positions that are known in one frame of reference,
say the rest frame of a particle with coordinates x , y , z , t , moving at a velocity
v in respect of another frame, say the laboratory system with coordinates x, y, z, t.
A distance x in the moving system is seen from the laboratory system at rest as
much shorter (L ORENTZ contraction):
x = x / . (1.27)
Conversely, the time difference t of two events measured in the moving system
will appear to be stretched in a measurement in the laboratory system (so called time
dilation):
t = t . (1.28)
We recall the so called twin paradox: the twin brother travelling in a spacecraft at
nearly the speed of light returns only a little bit older, while his twin who remained
on earth has already become an old man.
Closely related is the relativistic D OPPLER shift. Assume, as indicated in
Fig. 1.6, a source moving with velocity v to emit radiation of angular frequency
(wave vector k with k = /c, wavelength = c/ ). In the laboratory sys-
tem at rest one registers an angular frequency (wave vector k) at an angle with
respect to v:
k 1 1

= = = = = . (1.29)
k (1 cos ) 2 1 cos
For absorption one just has to replace by +. For perpendicular observation

( = /2) this leads in
both cases to the non-classical, so called quadratic D OPPLER
effect / = 1/ = 1 2 .
In forward and backward direction, = 0 and , respectively, (1.29) simplifies
(again with opposite signs for absorption):

k 1 1

= = = = (1 ) = = 2 1. (1.30)
k (1 ) 1
In the limit of highly relativistic energies, with 1, radiation is emitted essen-

tially into forward direction at frequencies
= 2 . (1.31)
14 1 Basics
We shall come back to this remarkable fact in the context of synchrotron radiation
in Sect. 10.6.2.
In the limit of very small velocities v one expands (1.29) in powers of = v/c
to recover the classical D OPPLER shift:
/ cos = (v/c) cos or (1.32)

v0
= k v. (1.33)
For later use we also give an expansion in terms of the kinetic energy. For wk =
T /(mc2 ) = 1 1 we obtain from (1.30) (at = 0 and ):

wk
= wk 2wk + wk 2wk 1
2 + . (1.34)
2
Section summary
Most of the relations from special relativity communicated here will be used
quite often throughout this textbook. The reader may want to memorize at
least (1.12), (1.13), (1.17), (1.20), (1.22), and (1.26)(1.28).
1.3 Some Elementary Statistics and Applications
Thermodynamics and statistics are together a big and important theme in physics
and physical chemistry. Many substantial textbooks exist on the subject and a vari-
ety of important aspects may be found well presented by online scripts in the Inter-
net. Here we present only a collection of topics and formulas from this wide field
with particular relevance to atomic and molecular physics. We start with some rather
elementary remarks about exponential probability distributions. They will be exem-
plified for spontaneous decay of unstable (excited) states of quantum systems and
for absorption of radiation. We then present a collection of formulas from kinetic
gas theory and end with probability distributions for classical particles, fermions
and bosons. But before going into specific examples we define some general termi-
nology.
A probability distribution w(x) describes the probability w(x)dx to find a ran-
dom variable between the values x and x + dx. Properly normalized the probability
to find the system with any value of x must be unity:

w(x)dx = 1. (1.35)
0
The average value of any observable f (x) which depends on x is

f = f (x)w(x)dx. (1.36)
0
1.3 Some Elementary Statistics and Applications 15
Specifically, the mean or expectation value of the variable is given by

x = xw(x)dx, (1.37)
0
and the so called variance is

2
2 = x x w(x)dx = x 2 2xx + x 2 w(x)dx
0 0

= x 2 2x 2 + x 2 = x 2 x 2 . (1.38)

The square root of the variance is called standard deviation, = x 2 x 2 , and
gives a measure of the width of the distribution.
1.3.1 Spontaneous Decay and Mean Lifetime
Exponential decay probabilities play an important role throughout quantum physics

in general, be it in AMO, nuclear or condensed matter physics. We have already
used notions like unstable particles or states, decay and half lifetime, tacitly
assuming that the reader is somewhat familiar with these concepts, and we shall
have to use them again quite often.
Quantum mechanics does not give us any certainty for finding quantum objects
at a given time, at a particular position in space or in a specific state: it only makes
predictions about probabilities. Thus, if a particle or a state of a quantum system is
not stable it decays into one or more other particles or states with a certain decay
constant A (also called decay rate) which is measured in units [A] = s1 . In many
cases it may be calculated by quantum mechanics (or possibly by QED or QCD).
We note explicitly two characteristic of such probabilities which hold for most of
the processes we are interested in:
1. The decay constant A does not depend on the number of particles that are in-
cidentally present in the experiment but only on the properties of the object(s)
studied.
2. It does also not depend on the time at which the particle or state decays: the decay
occurs at some arbitrary, a priori unknown time and we cannot predict what that
time will be; we only know the probability Adt that a decay may happen within
a time interval dt.
No predictions can be made about the destiny of a specific particle or state. How-
ever a quite accurate prediction can be made for a large ensemble of, say N , parti-
cles or objects in unstable states.2
Here and in the following N gives the number of particles in a volume V of interest while
2 Note:
N = N /V refers to the number density (or particle density) measured in units [N] = m3 .
16 1 Basics
In the process N decreases during the time dt by
dN = N Adt. (1.39)
The minus sign here indicates reduction. If we start at time t = 0 with N0 particles
in one particular initial state, the number N (t) of particles that at time t are still
found in their initial state is obtained by integration:
t
dN
= Adt ln N (t) ln N0 = At
N 0
N (t) = N0 eAt = N0 et/ = N0 et ln 2/1/2 (1.40)
where we have introduced the mean lifetime = 1/A and the so called half lifetime
1/2 = ln 2/A = ln 2 = 0.692 . The latter is the time during which half of the ini-
tial particles have decayed. Often one is interested in the number of decay processes
per unit time in the sample (at time t):3
dN N (t) ln 2 N0 t/
A= = AN (t) = = N (t) = e . (1.41)
dt 1/2
The exponential decay law is of fundamental nature and describes the statistical
(also called spontaneous or natural) decay of excited atomic or molecular states or
of electron hole pairs (excitons) in a solid as well as the decay of radioactive nuclei
(and thus the decrease of radioactive radiation with time) or of elementary particles
(barions, mesons, etc.) such as the decay of the neutron which we shall describe
in Sect. 1.5.3.
The exponential distribution or decay law (1.40) is displayed in Fig. 1.7. One
directly recognizes the significance of the half lifetime 1/2 . Alternatively to a linear
display shown in Fig. 1.7(a) one often uses a logarithmic scale for the probability.
As seen in Fig. 1.7(b) the exponential decay then gives a straight line. This is often
used to recognize exponential decays and to estimate the half lifetime.
Finally, we may recast (1.40) into a probability w(t)dt for any of the initial N0
particles to decay between a time t and t + dt:
1 dN 1
w(t) = = et/ . (1.42)
N0 dt
We note that this probability distribution is normalized according to (1.35): any
given particle will definitely decay at some time between t = 0 and . Conversely,
the probability that a particle has not yet decayed before a time t is exp(t/ ).
And with (1.37) we verify indeed that the mean lifetime (also mean or average
decay time) is t = = 1/A as introduced in (1.40). The variance (1.38) for the
exponential decay function is 2 .
3 In nuclear physics this is called activity which must not be confused with the decay constant (or
rate) A for which in nuclear physics often the letter is used.

(t) / 0 (t) / 0 in log. scaling
1.0 (a) 1.0 (b)
0.8 1/2
1/4
0.6
1/2 1/8
0.1
0.4
1/4
0.2
1/8
0.0 0.01
0 1 2 3 4 5 6 t / 1/2 0 1 2 3 4 5 6 t / 1/2
Fig. 1.7 Exponential decay law (a) in linear, (b) in logarithmic display. Note that the time axis is
scaled here in units of the half lifetime 1/2 so that for t = 1 and t = 2 and t = 31/2 the probability
decreases to 1/2, 1/4 and 1/8, respectively as indicated by the dashed lines
1.3.2 Absorption, LAMBERT-BEER Law
Exponential distribution functions of the kind just discussed do not only describe
probabilities as a function of time. Another important application is the absorption
of fast moving particles or electromagnetic radiation (i.e. photons, including visible
light as well as -rays) when passing through matter.
We discuss prototypically the absorption of visible light of intensity I (z) when
passing over a distance z through a medium. The intensity is defined as the total
energy transported in the light beam per unit time and area. It is measured in [I ] =
J s2 m2 = W m2 .
Alternatively, in the particle picture, we may describe the photon flux (z) =
I (z)/(h) with [] = particles s1 m2 as a function of position z (each photon
having an energy h). As indicated in Fig. 1.8, light with intensity I (z) at position
z is absorbed on the way through a medium. In analogy to (1.39) and (1.40), the
intensity dI absorbed over a short distances dz is proportional to that distance and
Fig. 1.8 Absorption of light

schematic for deriving of I(z) I(z) -dI
the L AMBERT-B EER law
dz
z
18 1 Basics
the incoming radiation dI = I (z)dz. In a homogeneous medium which extends

from zero to infinity, we find that the initial intensity I0 (at z = 0) is reduced to4
I (z) = I0 exp(z) (1.43)
at position z. Thus, the result is again an exponential decay law here it is the light
intensity which decays with the optical path z travelled.
The proportionality constant is called absorption coefficient and is measured
in units [] = m1 . It may be understood on an atomistic, statistical basis: the ab-
sorber medium may e.g. be an atomic gas in its electronic ground state, it may also
be glass or a liquid. In any case it consists of many absorbing particles (atoms,
molecules, defects in a solid), each of which has a characteristic absorption cross
section ([ ] = m2 ) for the photons which pass through the medium. The absorp-
tion coefficient is then the product of cross section and particle density (of the
absorbers), = N .
For a medium of finite thickness d the transmitted intensity is thus given by the
so called L AMBERT-B EER law
I (d) = I0 exp(d) = I0 exp( N d). (1.44)
and the decrease of the photon flux with z is given by
(z) = 0 exp(z) = 0 exp( N z). (1.45)
The probability w(z)dz that any of the photons arriving in the initial flux 0 at z = 0
is absorbed between z and z + dz is given by the probability distribution
1 d 1
w(z) = = ez/ l with l = 1/( N ). (1.46)
0 dz l
In analogy to the notation used in Sect. 1.3.1, here l is called mean free path length
of the photon. It is the distance a photon can on average move freely in the
medium without being absorbed. And exp(z/ l) is the probability that a photon
has survived up to distance z.
1.3.3 Kinetic Gas Theory
The statistical interpretation of the properties of ideal and real gases by the kinetic
theory of gases has played an important role in the history of atomic and molecular
physics. Here we just communicate some basic terminology and results without
derivation.
4 Inthe chemical literature one often writes = ln 10C 2.303C, with C being the con-
centration of the absorbent (e.g. in dilute liquids or gases), measured in [C] = mol L1 . The
molar absorption (or extinction) coefficient thus has the dimension [] = L mol1 cm1 =
1000 mol1 cm2 . The so called absorption (or extinction) is then log(I0 /I (x)) = Cx.
One mol of a gas (an SI unit) is defined as the amount of a substance that
contains as many elementary entities (e.g. atoms, molecules, ions, electrons) as there
are atoms in 12 g of the carbon isotope 12 C. The corresponding number of particles
is the AVOGADRO constant:
NA = 6.02214179(30) 1023 mol1 . (1.47)
The relative atomic (or molecular) mass Mr , formerly called standard atomic
(molecular) weight, of an isotope (substance) X is defined as Mr (X) = m(X)/
[m(12 C)/12] where m(X) is the mass of the corresponding atom and m(12 C) that
of 12 C. Note, that one typically finds Mr given in [Mr ] = g mol1 so that the mass
of one atom (molecule) of this substance is m = Mr /NA . Formally, however, Mr is
a dimensionless number and m = (Mr /NA ) g mol1 .
The classical equipartition theorem states that in a system at thermodynamic
equilibrium each degree of freedom that enters quadratically into the total energy
of the system contributes equally to it with kB T /2. The total internal energy of the
complete system is then given by5
kB T RT
U = f NA =f (1.48)
2 2
where f is the number of degrees of freedom per particle, T the absolute tempera-
ture of the gas, kB the B OLTZMANN constant (1.83) and R the molar gas constant
R = N A kB .
In an ideal (atomic) gas only the kinetic energy is of relevance and f = 3. Tem-
perature then corresponds to the average internal kinetic energy u of the particles
according to
1 3
u = mv 2 = kB T . (1.49)
2 2

These particles
move in the gas with an average velocity v 2 , their average momen-
2
tum being m v . From this, one may calculate the pressure p of an ideal gas as the
momentum transfer per unit time and area by elastic collisions and back reflections
with and from the walls containing the gas. Since 1/6 of all atoms move into one
direction we obtains with the particle density N
p = N mv 2 /3 = N kB T . (1.50)
For NA particles, i.e. for mol in a volume V we have N = NA /V and obtain

from (1.50) the well known ideal gas law for a macroscopic system:
pV = NA kB T = RT (1.51)
5 We use the traditional letter U for this energy/ mol and u for its average per particle.
20 1 Basics
In a real gas particles do collide not only with the walls but also with each
other. With the particle velocity v, the particle flux (i.e. the number of particles
passing through a unit area per unit time) is vN . If is the gas kinetic cross section,
measured in [ ] = m2 , the time tcol and the distance l a particle moves on average
freely between two collisions are
1
tcol = and l = v tcol , (1.52)
vN
respectively. The bracket indicates that one has to average over the velocity dis-
tribution N(v) of the gas particles. Somewhat more precisely, in the denominator v
is the relative velocity between the colliding particles.
If only one particle species
is of interest their average relative velocity is 2v . Thus, the so called mean free
path becomes
1
l= , (1.53)
2 N
where is an average gas kinetic cross section. For typical elastic collisions between
atoms and molecules it is on the order of 1019 m2 . Relations similar to (1.53) may
also be written for the absorption of ions and nucleons or of light, X- or -rays, see
(1.46).
Finally, we come back to the equipartition theorem. For molecules one has to
add 1 to the number of degrees of freedom f for each accessible rotation, and 2
for each vibrational mode (for kinetic and potential energy). Thus, in a solid f =
6 per atom, in a diatomic molecular gas f = 7 per molecule (three translational
coordinates, two rotational axes, one vibrational mode), while for linear and bent
triatomic molecules we have f = 9 and 10, respectively and so on. However, in
these cases one has to account for quantization of rotational and vibrational energies:
they become inaccessible (frozen) at very low temperature. Thus, the equipartition
theorem in its simple form (1.48) holds only at sufficiently high temperature.
1.3.4 Classical and Quantum Statistics Fermions and Bosons
When speaking of mean energy, velocity, lifetime or free pathway of particles (pho-
tons, atoms, molecules, ions, electrons, nuclei etc.) one implies that these quantities
are described by a statistical distribution. Classically, B OLTZMANN statistics pro-
vides the statistical distribution of energies in all fields of physics. It refers to the
probability for finding a certain energy u per particle in an ensemble. This energy
may be kinetic energy or internal excitation of the particles (e.g. electronic, vibra-
tional and rotational excitation in a molecule). Quantum physics requires certain
modifications, but for low enough particle densities and sufficiently high tempera-
tures these are very small, as we shall see in a moment.
In view of the possible quantization of energy we have to distinguish discrete
and continuous energy states. Energies of the system may be realized by different
quantum mechanical states. Different states i with identical energies ui are denoted
as degenerate and the number of possible realizations of one energy ui is called

degeneracy gi . In the case of a continuum of energies u a density of states g(u)
characterizes the number g(u)du of states (here per unit volume) in an energy inter-
val between u and u + du.
The B OLTZMANN distribution is derived from classical statistical mechanics.
The number density Ni (or dN , respectively) of particles with energy ui (or between
u and u + du, respectively) may be written as
Ni gi
= exp(ui /kB T ) (1.54)
N Z(T )
or dN g(u) exp(u/kB T )du (1.55)
with the total number density N of the particles and the so called partition function
Z(T ) = gi exp(ui /kB T ) which ensures that summation over all states i on the
right hand side of (1.54) gives 1.
Normalization in the continuum case needs some more detailed consideration.
We exemplify this by the distribution of velocities vx , vy , vz in an ideal gas. With
the particle mass m the kinetic energy is u = m(vx2 + vx2 + vx2 )/2 and g(vx , vy , vz ) =
const (since for < vx,y,z < no velocity vector is a priori more probable than
another) one may write the velocity distribution
3/2
dN m m(vx2 + vx2 + vx2 )

= exp dvx dvy dvz , (1.56)
N 2kB T 2kB T
which is normalized such that integration over all velocities gives 1. If, on the other
hand, one is interested in the probability of finding a particle with a certain mag-
nitude of velocity in a range v and v + dv one has to integrate over all angles so
that dvx dvy dvz = 4v 2 dv. This leads to the well known M AXWELL -B OLTZMANN
velocity distribution

dN 2 m 3/2 2 mv 2
= v exp dv, (1.57)
N kB T 2kB T
which is again properly normalized so thatthe integral over all velocities 0 v <
is 1. The most probable velocity is vm = 2kB T /m. One may rewrite (1.57) as an
energy distribution by substituting u = mv 2 /2:
3/2
dN 2 1 u
= u exp du. (1.58)
N kB T kB T
Comparing with (1.55)

we see that the density of states in the continuum of kinetic
energies is g(u) u. The mean energy is given by
3/2
2 1 u 3
u = u u exp du = kB T (1.59)
kB T 0 kB T 2
22 1 Basics
state 1 2 3 1 2 3 1 2 3
Boltzmann Bose-Einstein Fermi-Dirac
Fig. 1.9 How two particles may be distributed onto three states; this illustrates the key difference
between the statistics according to B OLTZMANN (distinguishable classical particles, here red and
black), B OSE -E INSTEIN (indistinguishable particles, no further limitations), and F ERMI -D IRAC
(indistinguishable particles, PAULI principle)
and thus fully recovers the classical equipartition theorem for three degrees of free-
dom in the
form (1.49). It also confirms the expression (1.49) for the averaged square

velocity v 2 = 3kB T /m, while v = 8T kB /m.
So much about classical statistics. When reconsidering this from a quantum me-
chanical point of view one has to account for phase space quantization as well as for
the indistinguishability of identical particles. The former aspect implies that even
the continuum is not completely continuous. Rather, the 6 dimensional phase space
(3 position and 3 momentum coordinates) has a finite cell size h3 . From this and
with the degeneracy ge = 2s + 1 due to the spin s of the particles under consid-
eration (see Eq. (1.8)) follows the density of states for a gas of non-interacting
particles in the continuum:

gs (2m)3/2 4 2m3/2
g(u) = u = gs u. (1.60)
4 2 3 h3
Its dimension is Enrg1 L3 . We shall give a derivation of g(u) for the model of a
free electron gas in Sect. 2.4.3 and discuss the quantization of electromagnetic radi-
ation in Sect. 2.2.2, Vol. 2. Here we refrain from elaborating on the derivation of the
statistical distributions, refer to the standard textbooks on statistical thermodynam-
ics and quantum statistics, and just summarize some key results.
In respect of indistinguishability, quantum mechanics knows two kinds of par-
ticles that behave differently: Bosons and fermions which we have introduced al-
ready in Sect. 1.1.1 as particles with integer and half integer spin s, respectively.
For fermions (e.g. e , e+ , p, 3 He, etc.) the PAULI exclusion principle (N OBEL prize
1945) holds so that each (discrete or continuum) quantum state can only be occu-
pied by one particle at most. In contrast, bosons (e.g. photons, 2 H = D, 4 He, 12 C,
etc.) are not restricted by the PAULI principle, i.e. each state may be occupied by
many particles. But again, identical bosons are indistinguishable in contrast to clas-
sical theory. Figure 1.9 illustrates the fundamental differences between the three
statistics for the most simple example: how can two particles be distributed onto
three states?
Detailed considerations on the probabilities for populating (many) energy levels
for many particles lead to the different statistics: the B OLTZMANN distribution for
classical particles, the B OSE -E INSTEIN distribution6 for bosons, and the F ERMI -
D IRAC distribution7 for fermions. To compare all three statistics (concentrating on
a quasi continuum of states) we also rewrite the B OLTZMANN distribution (1.55)
suitably:
1
B OLTZMANN dN = g(u)du (1.61)
exp[(u )/(kB T )]
1
F ERMI -D IRAC dN = g(u)du (1.62)
exp[(u )/(kB T )] + 1
1
B OSE -E INSTEIN dN = g(u)du. (1.63)
exp[(u )/(kB T )] 1
We have introduced the so called chemical-potential8 which here allows us to

properly normalize the distributions. Note that the dimensions of these equations are
m3 since with the density of states according to (1.60), g(u)du is a number per unit
of volume, and so is N . In the case of discrete states, one simply has to replace dN
by Ni /N and g(u) du by gi /Z(T ). The fractions displayed in the middle of these
equations are the B OLTZMANN factor, the F ERMI function and the B OSE -E INSTEIN
function, respectively (not distribution, as they are sometimes called colloquially).
Proper normalization requires that the integration over all energies 0 u <
is carried out using the whole right hand expression including the density of states.
The result must be N , the number density of the gas under consideration. In this way
one determines the chemical-potential . As we shall see in a moment, it depends
on T , N , m, and ge and on the type of statistics to be applied.
It is interesting to note that the three statistics differ only by an additive constant
0, 1 or 1 in the denominator and of course by the specific value of . As it
turns out, for sufficiently high temperature T and/or not too high density N they are
virtually indistinguishable. However, they differ substantially at low temperature: at
T = 0, bosons are all in the lowest state and the total energy of the system is zero,
while fermions populate a band of energies up to a maximum value F , called Fermi
energy (see Sect. 2.4.3).
6 B OSE first applied this to photons while E INSTEIN generalized it to any bosons.
7 F ERMI and D IRAC developed it independently from each other in 1926, F ERMI somewhat earlier
than D IRAC.
8 Inthermodynamics, the chemical potential is defined as the partial derivative G/ N of the free
enthalpy G (G IBBS potential) with respect to the particle number N at constant temperature and
pressure. Thus, gives the amount of energy that is necessary to change the number of particles
of a system (by 1) without disturbing the equilibrium of the system.
24 1 Basics
For a gas of free, noninteracting particles, we may insert the density of states
g(u) from (1.60) explicitly, introduce a quantity

4 2m3/2
A = gs (1.64)
N h3
which emphasizes the quantum nature of these statistics, and obtain:

dN udu
B OLTZMANN =A (1.65)
N exp( u
kB T )

dN udu
F ERMI -D IRAC =A u (1.66)
N exp( kB T ) + 1

dN udu
B OSE -E INSTEIN =A u . (1.67)
N exp( kB T ) 1
For normalization we have to integrate the right hand sides of (1.65)(1.67). By

substituting x = u/(kB T ) and = /(kB T ) the normalization condition becomes

xdx !
A(kB T )3/2
=1 (1.68)
0 exp(x ) +
with = 0 and 1 for the B OLTZMANN, F ERMI -D IRAC and B OSE -E INSTEIN
probability distributions, respectively.
For theB OLTZMANN distribution the integral can be evaluated in closed form
and gives /2 exp( ) so that one finds for the chemical-potential :
gs (2mkB T )2/3
exp( ) = (1.69)
N h3
gs (2mkB T )3/2
or = = ln . (1.70)
kB T N h3
At high temperatures and not too high densities the chemical-potential is thus neg-
ative. Just to obtain some feeling for typical values of this quantity we note that at
normal conditions N = NL = 2.687 1025 m3 and 273 K for He with m 4 u,
s = 0 and gs = 1 one finds exp( ) 252106 compared to which the additive
constant = 1 in the denominators of (1.66) and (1.67), respectively, is irrele-
vant. This is characteristic for gases under standard conditions where no difference
among the three statistics is observable, and 0.293 eV is virtually identical
for all three statistics.
However, at low temperatures and/or high particle densities significant differ-
ences exist. For F ERMI -D IRAC and B OSE -E INSTEIN statistics has to be deter-
mined from a numerical integration of (1.68) with = 1, respectively.
It is important to note, that for B OSE -E INSTEIN gases must not be positive to
avoid singularities in (1.67) (see, however, Sect. 2.2.2, Vol. 2). For = 0 the integral
on the left side of (1.68) reaches its maximum and can be evaluated in closed form.
(a) (b)
He 100 mbar at 2.5 K He 100 mbar at 1K

exp (u ) +
Tc = 0.18K Tc = 0.32K
u
0 1 2 3 4 5 6 0 1 2 3 4 5 6
u / k BT
Fig. 1.10 Comparison of the three statistical energy distributions for an atomic mass 4 u
at 100 mbar and two different temperatures (a) 2.5 K and (b) 1 K. The red lines refer to
B OSE -E INSTEIN statistics as applicable to He atoms, the grey line illustrates the classical
M AXWELL-B OLTZMANN distribution and the black line represents the energy distribution ob-
tained from F ERMI -D IRAC statistics
By comparing it to (1.64) one derives the so called critical temperature
2/3 2 2 2/3 2/3

Tc = 2N 2/3 (3/2)gs = 3.31 gs N (1.71)
mkB mkB
using the value (3/2) = 2.612 from the R IEMANN zeta function. At this critical
temperature the much celebrated B OSE -E INSTEIN condensation (BEC) occurs (the
pioneering work of C ORNELL, K ETTERLE, and W IEMAN was honoured with the
N OBEL prize in C ORNELL et al. (2011)).
In Fig. 1.10 we show two examples of the three statistical distributions with T
somewhat above the critical temperature Tc . The energy distributions shown have
been calculated for a gas of mass 4 u at a pressure of 100 mbar for two different tem-
peratures (a) T = 2.5 K and (b) 1 K. Clearly, in nature He gas would be described
by the red curve corresponding to B OSE -E INSTEIN statistics, while the other two
curves are just drawn for comparison. We recognize the differences for the tree
statistics, but it is also clear by extrapolating the trends seen in this figure that at
only several times the critical temperature, not the slightest difference will be rec-
ognizable.
In contrast to B OSE -E INSTEIN statistics where the chemical-potential has to be
0, in the F ERMI -D IRAC case can assume also positive values (no singularities
occur in Eq. (1.66)). One particularly important application is the model of a free
electron gas for electrons in a metal. There the number density of electrons is very
high and temperatures may be low. In that case, is called F ERMI energy, which can
assume rather high positive values (in units of kB T ). We shall discuss and illustrate
this in Sect. 2.4.3 and shall find that the energy distribution looks very different from
a M AXWELL -B OLTZMANN distribution.
26 1 Basics
Section summary
Statistical distributions are important in many areas of classical and quantum
physics. They describe the probability to find an observable at a certain posi-
tion in space or time or with a particular energy, frequency etc. Characteristic
are the mean value (1.37) and the variance (1.38).
Most common is the exponential distribution which we have introduced in
the context of spontaneous decay of excited states (1.40). It may be charac-
terized by the half-lifetime of the excited states (the time after which half of
the excited states have decayed) which is t1/2 = ln 2/A = ln 2, with the
mean lifetime and A the transition probability. Analogue relations hold for
the absorption of radiation though matter, described by the L AMBERT-B EER
absorption law (1.44).
We have reviewed a few basic concepts from kinetic gas theory: the average
free energy per atom and degree of freedom is kB T /2; the mean free path
between two collisions in a gas is l = 1/( 2 N ), with 1015 m2 the gas
kinetic cross section and N the particle density.
We have also discussed the three relevant statistics (1.61)(1.63) for energy
distribution in an ensemble of particles: the classical B OLTZMANN statistics,
F ERMI statistics (valid for fermions, i.e. particles with half integer spin) and
the B OSE -E INSTEIN statistics (valid for bosons, particles with integer spin
quantum number). At temperatures a few degrees above the critical tempera-
ture (1.71) for B OSE -E INSTEIN condensation all three distributions are almost
identical.
1.4 The Photon
From classical wave optics we know that light can be described as electromagnetic
waves: diffraction and interference are the experimental observations onto which
this viewpoint is based. In terms of geometrical optics, light propagation may even
be described simply by so called light rays or beams a notion which can be based
on wave optics as a special case (more in Chap. 1, Vol. 2).
However, light has also particle properties. The key observations documenting
this aspect are summarized in this section. Quantum mechanics, to be addressed in
some detail in the next chapter, reconciles both points of view or rather: it provides
a set of rules for a consistent interpretation of the experimental observations.
1.4.1 Photoelectric Effect and Quantization of Energy
One of the fundamental observations on the quantum nature of light is the photo-
electric effect. With light of a wavelength (frequency = c/) one illuminates a
metal surface from which electrons emerge. One measures the kinetic energy T of
1.4 The Photon 27
potential energy
e
T
h
WA
electron
sea x
Fig. 1.11 Simple potential well model to explain the photoelectric effect. The photon of energy
h rises an electron from the electron sea in the metal (bound) into the continuum (free). A hole
is left in the sea
these emitted electrons and makes some quite remarkable observations (astonishing
at least at the time early in the 20th century when this was discovered):
in contrast to the classical expectation the energy of the photoelectrons is inde-

pendent of the intensity of the irradiating light: the latter only determines the
number of emitted electrons
the observed kinetic energy T of the photoelectrons has a maximum value
T (max) = h WA , (1.72)
where h is the P LANCK constant (1.2) and WA the so called work function or
electron affinity of the metal surface from which electrons emerge (for experi-
ments in the gas phase WA has to be replaced by the ionization potential WI of
the atoms or molecules studied).
E INSTEIN (1905) in his annus mirabilis presented the interpretation for the photo-
electric effect one of the key steps in the early days of modern physics for which
he received the N OBEL prize in physics9 in 1921: Light energy exists only in well
defined energy packets of
Wph = h = . (1.73)
This energy packet is the elementary quantum of light, called photon. Light obvi-
ously has both: wave and particle character.
To get some numerical feeling, let us consider yellow light (from the sun or
from a sodium street lamp) at a wavelength of = 589 nm. With c = we have
= 5.09 1014 Hz. Thus, Wph = h = 3.37 1019 J = 2.10 eV is the energy of
the photon!
One may visualize the photoelectric effect by a very simple potential well model
for quasi free electrons in a metal. The energetic relations between T , WA and h
are illustrated in Fig. 1.11. If the photon has an energy h > WA it may eject an
electron from the electron sea (where it is bound but may freely move within the
metal) into the continuum (where it is unbound). If the electron originates from the
9 Note: not for his at least equally important theory of special relativity.
28 1 Basics
(a) scattered radiation detector (b) scattered photon p' = h/ '

W' = h'
collimating p'
apertures

incoming incoming free
-radiation photon electron p'e + p' = p
p=h/ pe = 0
metal target W = h We = m e c 2 p'e
electron
scattered electron We' = c2p'e2 + me2 c 4
Fig. 1.12 (a) Experimental scheme to study the C OMPTON effect. (b) Kinematics of the scattering
process. Prior to collision energy (momentum) are W (p) and We (pe ) for photon and electron,
respectively; the dashed quantities refer to the situation after the process
surface, it will have a kinetic energy T = T (max) as given by (1.72). An electron

hole is left in the sea.
The photoelectric effect is the basis for modern photoelectron spectroscopy
(PES) as we shall elaborate on in some detail in Sect. 5.8 in Vol. 2. One may eas-
ily visualize from Fig. 1.11 that a precise determination of the spectrum of kinetic
energies T of the emitted electrons may serve as a sensitive tool to determine the
electronic structure of the object studied. The potential well is of course only a very
rough approximation of reality, which more correctly would be described by the
band structure of a solid surface.
Note: The photoelectric effect as described above refers to low photon and elec-
tron fluxes (linear regime). If one uses very intense laser pulses which can readily
be generated with state-of-the-art laser techniques, the situation changes. With in-
creasing light intensity the process becomes no longer linear, i.e. more than one
photon will be involved in the emission process of one electron (see Sect. 8.5.1).
Then the phenomena observed may even approach the classical expectation, which
so surprisingly was not met in the original photoelectron emission experiments at
low intensity.
1.4.2 COMPTON Effect and Momentum of the Photon
The C OMPTON effect (N OBEL prize in 1927) may be observed with an experi-
mental setup as sketched schematically in Fig. 1.12(a). Highly energetic photons
( -radiation) are scatted from quasi free metal electrons. The momentum of the
photon is
p = k and p = h/ = /c. (1.74)
It enters into the kinematics of the experiment as illustrated in Fig. 1.12(b).

Both, momentum p = p + p e and energy W + We = W + We have to be con-
served during the scattering process, using the relativistic expression (1.19) for the
electron energy. With some algebra one finds that the wavelength of the scat-
1.4 The Photon 29
= 0 incoming radiation
(dashed) = 90

= 45 intensity of the
scattered = 135
- radiation

= 0.710 = 0.710 ' = 0.751
Fig. 1.13 Wavelengths of radiation after C OMPTON scattering
tered -radiation is shifted in respect of the incoming . The shift depends on the
scattering angle :
= C (1 cos ). (1.75)
Experimentally one observes the predicted shift of wavelength in the scattering

signal as a function of the scattering angle illustrated in Fig. 1.13. This shift is
a direct consequence of the momentum of the photon. Thus, along with the pho-
toelectric effectthe C OMPTON effect constitutes an important proof for the particle
properties of the photon. The parameter in (1.75) is the so called C OMPTON wave-
length of the electron (with the electron mass me ):10
h
C = = 2a0 = 2.4262 1012 m. (1.76)
me c
It determines the overall magnitude of the wavelength shift and is independent of

the irradiating wavelength . We compare several characteristic lengths to convey
some feeling for the relevant orders of magnitude. The C OMPTON wavelength of
a particle corresponds to the wavelength of a photon whose energy is equal to the
rest mass energy mc2 of that particle. For the electron it is in between atomic and
nuclear radius:
atomic radius (H atom 1s ) a0 = 0.529 1010 m

C OMPTON wavelength of e C = 2.4262 1012 m
proton radius Rp = 0.875 1015 m.
10 Often the reduced C OMPTON wavelength /me c = a0 = 3.8110 1012 m is used. In rel-
ativistic quantum mechanics one typically measures lengths in units of the reduced C OMPTON
wavelength /me c, and atomic energies in units of me c2 .
30 1 Basics
1.4.3 Pair Production
The photoelectric effect and the C OMPTON effect are two major mechanisms by
which high energy photons interact with matter. For completeness we also mention
pair production: in the vicinity of a nucleus a photon can be converted into an elec-
tron and a positron. This process is symbolically written as e +e+ and has the
energy balance
h = 2me c2 + Te + Te+ , (1.77)
so that this process becomes possible if and only if the energy of the photon
h > 2me c2 1.022 MeV, the rest mass of one electron and one positron. The ex-
cess energy is converted (essentially) into kinetic energy Te+ +Te of the two emerg-
ing particles. For momentum conservation (at threshold the electron and positron
momenta are very small) the process can only occur in the presence of a nucleus
which by C OULOMB interaction carries away the surplus momentum of the
photon. Pair production may also be viewed as excitation of an electron into the
world from the D IRAC sea in which a whole is created (the positron).
All three processes, photoelectric effect, C OMPTON effect and pair production,
are the key mechanisms for absorption of high energy photons in by atoms (specifi-
cally in the solid state). We shall come back to this in Sect. 10.5.3.
For completeness we mention that the exactly inverse process is extremely im-
probable due to phase space considerations. In contrast, the generation of two pho-
tons by positron-electron annihilation e +e+ 2 is a well known process, in
which energy and momentum conservation is easily realized. The two photons are
emitted in exactly opposite direction. This process is exploited in positron emission
tomography (PET), today a widely used medical technique for precise tumour imag-
ing. The positron in this case originates from an artificial isotope attached to a drug
which is accumulated specifically in tumour cells. Detecting the two photons after
e e+ annihilation in coincidence allows one to localize their origin in the human
body.
1.4.4 Angular Momentum and Mass of the Photon
For completeness we mention already here that the particle photon also has an
intrinsic angular momentum , called the photon spin with a quantum number s = 1.
We shall learn more about the experimental evidence in Sect. 4.1.4. The photon
spin will play an important role in various contexts, and a quantum mechanical
description of the photon will be presented in Sect. 2.2, Vol. 2.
We may even attribute a mass to the photon, the equivalent of its energy:
mP h = h/c2 . (1.78)
Note, however, that the rest mass of the photon is zero: it exists only as a particle
moving with the speed of light. We shall later on discuss that this has serious conse-
quences for space quantization of its angular momentum. Briefly, the massless parti-
1.4 The Photon 31
cle photon has only two substates with sz = jz = , while according to (1.9) a par-
ticle with rest mass and angular momentum j = s = 1 has three possible substates.
1.4.5 Electromagnetic Spectrum
Electromagnetic radiation is the key for most spectroscopic studies in atomic and
molecular physics. The relevant radiation ranges from radio frequency photons
whose energy is in the eV energy range up to the hard X-ray region with energies
up to MeV. Figure 1.14 gives an extended overview on the whole electromagnetic
spectrum of relevance. Note that slightly different definitions are used for the spec-
tral ranges shown in Fig. 1.14, depending on the field of application. We follow the
specifications of ISO 21348 (2007).11 Different units are used in different spectral
regions: Frequencies in the very low energy range, wavelengths in the infrared
(IR), in the visible (VIS), ultraviolet (UV) and vacuum-ultraviolet (VUV) spectral
range.12 For still shorter wavelengths, i.e. in the extreme ultraviolet (XUV), in the
soft and hard X-ray region as well as for -rays one uses energy units ( in eV,
keV, MeV). In spectroscopy a convenient measure is the reciprocal wavelength, the
so called wavenumber
= 1/ (1.79)
that is proportional to the photon energy
Wph = hc = = h = hc/ (1.80)

= 1.239841875(31) 104 eV cm.
The SI unit of wavenumbers is m1 but commonly one still uses [ ] = cm1 , often
literally called wavenumber. Up to date energy conversion factors are found e.g.
at NIST (2011).
1.4.6 PLANCKs Radiation Law
Quantum mechanics, one may say, was triggered by understanding the photoelec-
tric effect due to E INSTEIN (1905) (N OBEL prize 1921). But before that (1900),
P LANCKs law had already revolutionized the world of physics (N OBEL prize 1918)
giving an accurate interpretation of black body radiation for which the depen-
dence on wavelengths and absolute temperature T had been measured with very
high precision. The characteristic behaviour is illustrated for several temperatures
in Fig. 1.15. The interpretation of this fundamental distribution of radiation forced
11 Except for RF and MW where we follow the technical literature.

12 Called
vacuum-ultraviolet since absorption in air forces one to work in vacuum with these
wavelengths.
(a) 21cm H-line CMBR maximum of H2O window cosmic -radiation
32
solar radiation nuclear reactions

between
K-edge of Cu K-edge
H
THz C and O
Lyman-
(b) LF MF HF VHF UHF MIR NIR
Cs atomic clock
9192631770 Hz
FIR
soft
RF EUV X-ray
microwave IR visible UV hard X-ray -ray
XUV
A BC VUV
(c) 1 km 1m 1 mm 1 m 1 nm 1pm 1 fm
(d) 106 Hz 109 Hz 1012 Hz 1015 Hz 1018 Hz 1021 Hz

(e) 1 neV 1 H9 1 meV 1 eV 1 keV 1 MeV 1 GeV
microwave
oven
(f) cellular phones optical fibre EUV for medical

communication lithography diagnostics
radio, TV (13.5 nm)
radar
EPR light bulb HHG

NMR
synchrotron radiation X-ray tube
IR-FEL FEL XFEL

TiSa
(g) electronic oscillators and
antennas clystron, magnetron various lasers and NLO laser plasma nuclear reactions
Fig. 1.14 Spectrum of electromagnetic waves (for acronyms see text and p. 81ff.). (a) Specific sources and properties, (b) terminology, (c) wavelength scale, (d)
frequency scale, (e) energy scale, (f) examples for applications, (g) examples for methods of generation. Note the narrow range of visible radiation. Only a few
special sources of radiation are specifically emphasized such as CMB(R) and H Lyman-. The ultraviolet (UV) regions UVA, UVB, and UVC are abbreviated
1 Basics
here by A, B, and C
1.4 The Photon 33
1.0
1.0 5772 K (sun)
1073 K
3400 K 288 K (earth)
0.5 4000
10 106
u( ) / J m3 m-1
2856 K
0.5 10 0 5 10 15 m
1873 K
100
0.0
0 1 2 3 4 5
visible spectrum (390 to 750) nm wavelength / m
Fig. 1.15 P LANCKs law at different temperatures: 5772 K effective black body temperature of
our sun, 3400 K special purpose, short-lived incandescent bulbs, 2856 K CIE standard illumi-
nant A (about 100 W tungsten incandescent bulb), 1873 K blast furnace at the discharge aperture,
1073 K dark red glow (kitchen stove, grill fire), 288 K earth surface temperature (average)
Max P LANCK to introduce, at the beginning very reluctantly, a new fundamental

constant of action h (units [h] = J s). P LANCK was well aware that this manifested
the breakdown of classical physics! Today, the P LANCK constant is a fundamental
physical constant, known with very high precision (see Eq. (1.2)).
Here without derivation (it will follow in Sect. 2.2.3, Vol. 2) we communicate
P LANCKs law for the spectral energy density distribution of the black body radiator
as
8h 3 d

u()d = , (1.81)
c 3 exp(h/kB T ) 1
per unit of frequency with [u()]

= J m3 Hz1 , or as
3 d

u()d = (1.82)
2 c3 exp(/kB T ) 1
per unit of angular frequency, [u()]

= J m3 s, with the speed of light c and the
B OLTZMANNconstant
kB = 1.3806504(24) 1023 J K1 . (1.83)
Alternatively one plots as done in Fig. 1.15 the distribution as a function of the
wavelength so that [u()]
= J m4 = 103 mJ m3 m1 and
8hc d

u()d = . (1.84)
exp(hc/kB T ) 1
5
34 1 Basics
Equations (1.81) and (1.84) are the standard form of P LANCKs law for the black
body radiation density as originally published by P LANCK (1900). Often one is
also interested in the intensity at a given point, direction and wavelength per unit of
projected area, solid angle and frequency interval with [L ] = W m2 sr1 Hz1
(or [L ] . . . per wavelength interval), also called spectral radiance. Since black
body radiation is intrinsically isotropic, the corresponding expressions are obtained
by simply multiplying (1.81) and (1.84) with c/4 , i.e. the prefactor 8hc is re-
placed by 2hc2 . Typically measured is the spectral distribution of the intensity,13
i.e. of the radiation power emitted (per area) into the forward hemisphere, with
[I()] = W m2 nm1 . Integration over the cos angular distribution (projection
of the surface area onto the direction of emission) gives a factor of so that
2hc2 d
I()d = . (1.85)
5 exp(hc/kB T ) 1
The wavelength max at which this spectral distribution of radiation reaches its
maximum, decreases with temperature as seen in Fig. 1.15. Explicitly one finds
(from dI()/d = 0 for the maximum) the so called W IEN wavelength displace-
ment law:
max T = b with b = 2.8977721(26) 106 nm K. (1.86)
Finally, the total power emitted per unit area from the surface of a black body
is obtained by integrating (1.85) over all wavelengths. This gives the important
S TEFAN -B OLTZMANN law for the intensity of black body radiation:
2 5 kB4 4
I (T ) = I()d = T = B T 4 . (1.87)
0 15 h3 c2
Thus, the (spectrally integrated) intensity depends on the fourth (!) power of
the absolute temperature T . The proportionality factor B = 5.670373(21)
108 W m2 K4 is called S TEFAN -B OLTZMANN constant.
1.4.7 Solar Radiation on the Earth
At this point, a few words are in order on the radiation which we receive every day
from our sun. Some relevant parameters are summarized in Table 1.5. The spectral
distribution of the solar radiation at the top of the earth atmosphere is shown in
Fig. 1.16 (wiggly red line). Since about 2004 it is constantly monitored by satellite.
The solar constant S is the integral over this spectrum. The daily results can be
obtained online from SORCE (2012) (the data reported there are renormalized to
13 In radiometry one uses the terms irradiance or radiant flux (see also Sect. 1.4.8). For consistency
with the general custom in AMO physics we usually call this quantity intensity of the radiation,
measured in [I ] = W m2 .
1.4 The Photon 35
Table 1.5 Some properties of sun and earth in the context of solar radiation (data from SSE 2012,
unless otherwise specified)
Mean radius sun R 6.9551 105 km
Mean radius earth RE 6371.0 km
Mean distance sun-eartha RSE 149.60 106 km 1 uab
Solar constantc S 1360.8(5) W m2 (from KOPP and L EAN 2011)
Radiation power emitted by sun 384 109 PW (into 4 sr)
Radiation power received by earthd 173.5 PW (from the sun)
Effective temperature TS 5772 K (for the above value of S)
Sun surface temperature Tph 4400 K to 6600 K (top to bottom photosphere)
Albedo (B OND) a 0.306 (fraction of radiation reflected)
Earth temperaturee TE 254 K (effective black body)
TEa 288 K (average at surface)
a Dueto the ellipticity of the orbit the distance varies between parhelion (minimum) and aphelion
(maximum) by about 6.9 % between 4th of January and 4th of July
b Per definition the astronomical unit of length is 1 ua = 149597870700 m; 1 ua is almost identical
to the average distance RSE of the earth from the sun

c Defined as mean irradiance at 1 ua from the sun
d Above atmosphere
eT
E = [S (1 a)/(4B )]1/4 from (1.87); receiving surface RE
2 , emitting 4R 2
E
solar spectrum at the top of the earth atmosphere

2.0
RAYLEIGH scattering
O2
O2 PLANCK at 5772 K
I () / W m2 nm-1
1.0 0.05
H 2O
solar H2O 20
H2O
~
O3 spectrum
at sea H2O 20
H2O,
level CO2
0.0
0 1000 2000 3000 4000
UV vis IR wavelength / nm
Fig. 1.16 Spectral intensity distribution I() of the solar radiation: as measured by SORCE
(2012) above the earths atmosphere (for > 2400 nm from ASTM 2008); at sea level (AM1.5
global tilt spectrum from ASTM 2008, see text); black body radiator at 5772 K. Also indicated
are the main absorbing molecules in the earths atmosphere
36 1 Basics
Fig. 1.17 Definition of air mass coefficient AM for solar radiation standards. Table on the right:
solar radiation (integrated from 280 to 4000 nm) at normal incidence for different zenith angles
a the astronomical unit of length 1 ua, see Table 1.5). The spectral distribution is
remarkably stable over time, the average changes being less than the width of the
red line in Fig. 1.16. As illustrated, the distribution is approximated reasonably well
by a P LANCK distribution from a black body at 5772 K (smooth black line). The
latter is obtained from (1.85) by multiplying it with (R /1 ua)2 , where R is the
solar radius. No further scaling is needed to obtain this fit.
As we see, the fit is not perfect, but considering the gigantic nuclear fusion re-
actor which our sun actually is, and considering its complicated photosphere, the
relative similarity with a black body radiator is quite remarkable and so is the sta-
bility of the distribution. A variety of values for the sun temperature may be found
in the literature. From the recent, highly accurate measurement of the solar constant,
S = 1360.8 W m2 , by KOPP and L EAN (2011) (see also SORCE 2012) and with
appropriate scaling of the S TEFAN -B OLTZMANN law (1.87), we obtain the effective
black body temperature of the sun as TS = (1 ua /R )1/2 (S/B )1/4 = 5772 K.
Also shown in Fig. 1.16 is the spectral distribution of radiation which reaches the
ground level after partial absorption and scattering of the incoming solar radiation
by atmospheric gases. As sketched in Fig. 1.17, this obviously depends on the zenith
angle of the sun ( = 90 latitude angle). For not too large the optical path
length through the atmosphere is given by hx = h0 / cos , where h0 7.7 km is the
effective vertical height of the earths atmosphere (i.e. the height where the pressure
has dropped to 1/e of its value on ground). The ratio
hx 1
AM = (1.88)
h0 cos
is called air mass coefficient. An approximate empirical formula (see e.g. H ONS -
BERG and B OWDEN 2012) for the intensity reaching the earth surface is
0.678
I = 1.1 I0 0.7AM , (1.89)
1.4 The Photon 37
where I0 is the radiation arriving on top of the atmosphere (essentially the solar
constant S, given in Table 1.5). The prefactor 1.1 accounts for radiation scattered
from the air and retroreflected from the ground, the exponentials reflect somehow
the L AMBERT-B EER absorption law. With reference to Fig. 1.17 we also note that
the intensity depends in addition on the angle of incidence ( ) onto the receiving
surface, Isur = I cos( ), with (1.89) referring to normal incidence.
As a standard value, AM = 1.5 (short AM1.5) has been adopted, considered rep-
resentative for most industrialized countries at noon time. Based on measurements
and modelling, one also has defined two standards spectral distributions arriving on
the earth surface, both for AM1.5: (i) direct normal incidence of radiation and (ii)
hemispherical (global) incidence on a 37 degree tilted surface, which also includes
scattered and retroreflected radiation (ASTM 2008, G173-3).14 The latter spectrum
is shown in Fig. 1.16 and referenced the table with Fig. 1.17.
We note at this point that this spectrum has its maximum at ca. 500 nm, i.e. near
the maximum of the spectral sensitivity of the human eye at 555 nm. The evolution-
ary context behind this coincidence is evident.
In contrast, the surface of our earth with an average temperature of ca. 288 K
emits in the IR spectral range with a maximum at about 10 m where the green-
house gas CO2 absorbs (we shall come back to this in Sect. 5.3.1, Vol. 2) fortu-
nately we have to say: without it, the temperature on the earth surface would cor-
respond to the effective black body temperature of the earth TE = 254 K and life
would not be possible.
1.4.8 Photometry Luminous Efciency and Efcacy
Black body radiators are very inefficient when used to generate visible light. Even
the sun emits only a fraction of its total radiation energy into the visible (VIS) spec-
tral region: only 46 % of the total radiation power are emitted between 380 and
760 nm, as one finds from integrating (1.87). Incandescent light bulbs at typical
temperatures convert much less energy into the visible spectral range (e.g. 6.5 %
for a 100 W bulb with its tungsten wire at 2856 K). Worldwide it is understood that
intense efforts are needed to save energy. Primary energy resources must be used
in a most efficient way and energy efficient lighting is an import potential. As a
consequence, the good old light bulb, which humankind has gotten so used to since
over 100 years, is approaching its last days very rapidly.
The overall efficiency in creating luminosity as registered by the human eye is
even much worse, since u() has to be multiplied by the physiological sensitiv-
ity V () of the eye. This so called photopic luminous efficiency function peaks at
555 nm,15 where the human eye has its maximum sensitivity. It is standardized by
14 See also ISO 60904-3 (2008) or DIN EN 60904-3. Note that for AM1.5 normal incidence and
incidence on a 37 degree tilted surface differs only by 2 %.

15 Vision in bright light is called photopic, in contrast to scotopic at low light levels and maximum
sensitivity at ca. 498 nm.
38 1 Basics
Table 1.6 Relation between photometric and radiometric quantities after O HNO (2010)
Photometric Unit Relation to lm Radiometric Unit
Luminous flux lm (lumen) Radiant flux W (Watt)
Luminous intensity cd (candela) lm sr1 Radiant intensity W1 sr1
Illuminance lx (lux) lm m2 Irradiance W m2
Luminance cd m2 lm sr1 m2 Radiance W sr1 m2
Luminous exitance lm m2 Radiant exitance W m2
Luminous exposure lx s Radiant exposure J m2
Luminous energy lm s Radiant energy J (joule)
Color temperature K (Kelvin) Radiance temperature K
the Commission international de lclairage (CIE) and forms the basis for all lu-
minosity determinations. For back of an envelope calculation it may roughly be
approximated by a Gaussian, centred at 560 nm with a FWHM 104 nm (from a
least square fit). But for precise work tabulated values from the literature have to be
used, e.g. from CIE (or as plotted by D ICK LYON 2008, with detailed references).
For a brief and concise introduction into photometry we refer to O HNO (2010)
from whom we also have adopted Table 1.6. It compares (physiologically weighted)
photometric quantities with (directly energy related, physical) radiometric quanti-
ties.16 The relevant photometric SI unit is the candela, cd, defined as . . . the lumi-
nous intensity, in a given direction, of a source that emits monochromatic radiation
of frequency17 540 1012 Hz and that has a radiant intensity in that direction of
(1/683) W per sr.
However, the physically more fundamental unit is the lumen (lm), the unit for the
luminous flux. It measures the photometric equivalent of the total radiation power
emitted by a given light source. Thus, the hypothetical, ideal light source driven by
1 W electric power emits 683 lm at 555 nm. It is said to have an overall luminous
efficacy of 683 lm W1 and its luminous efficiency is defined as 100 % it emits the
maximum possible amount of visible light.
All other light sources have (much) lower luminous efficacy and efficiency. One
may compute the

V ()I()d
luminous efficiency = 0 (1.90)
I()d
0
from the spectral radiation intensity I() = 4u()/c

of the source and the photopic
luminosity function V (). For a black body radiator I() = cu()/4
is obtained
from (1.84) and the denominator in (1.90) is B T 4 , the total radiation power emit-
16 Note that this radiometric terminology somewhat confusingly differs from the standards used
elsewhere in optical and laser physics. A laser beam e.g. is typically characterized by its intensity
measured in units [I ] = W m2 , while here in radiometry the corresponding quantities in Table 1.6
are called irradiance or radiant exitance.
17 That is at a wavelength of 555 nm where the human eye has its maximum sensitivity.
1.4 The Photon 39
ted per unit area. This luminous efficiency is a function of T and has a maximum of
about 14.5 % at T = 7000 K. For our sun it is ca. 14 %, while for the above men-
tioned CIE standard Illuminant A the efficiency is only about 2.5 %, correspond-
ing to a luminous efficacy of ca. 17 lm W1 . A typical standard 100 W incandescent
bulb emits a total luminous flux of 1360 lm and thus its efficacy is 13.6 lm W1 . Of
all the electric energy used to heat the tungsten wire of the bulb, 98 % is lost as IR
radiation and eventually dissipated as heat! So called halogen lamps are somewhat
more efficient since a special chemical process reduces evaporation of the cathode
material and the temperature of the tungsten wire can be significantly higher. Still,
the general efficiency problem of incandescent lamps remains.
Thus, the challenge is, to exploit more efficient ways for converting electrical
energy into visible light. One direction that currently is followed uses miniaturized
fluorescent lamps. Typically, in a mercury gas discharge UV light is generated at
253.7 and 185 nm. It is then converted into visible light of different wavelengths
by fluorescing materials, so called phosphors. Typically three to five different phos-
phors with reasonable efficiency are used, whose fluorescence bands by additive
colour mixing appear more or less as white light. Much current development is
focused onto these materials. By judicious choice of the fluorescent bands one may
obtain a more or less continuous coverage of the visible range thus imitating the
visible part of the black body radiation without loosing energy into other spectral
regions.
The colour temperature of such a source is defined as the temperature of a black
body radiator that best matches that spectrum in the visible. Presently, the lumi-
nous efficacy of compact fluorescent lamps (commercial energy saving lamps) is
on the order of (50 to 60) lm W1 (7.3 % to 8.7 % efficiency) but up to more
than 100 lm W1 (14.6 % efficiency) can be achieved with long tubular fluorescent
lamps. For street lighting also high and low pressure sodium discharge lamps are
used, easily recognized by their yellow-orange light (around 590 nm). Low pres-
sure Na discharges have the highest luminous efficacy in state-of-the-art lighting
technology of up to 200 lm W1 (29 % efficiency).
Light emitting diodes (LED) could, in principle, be able to supersede these val-
ues: they convert electric current so to say more or less directly into light. Again,
additive colour mixing of at least three LEDs is required for generating white light
in practical devices. The luminous efficiency is presently of a similar order of magni-
tude as for fluorescent lamps, somewhere between 15 and 25 %. However, strategies
are discussed to achieve much higher efficiencies (e.g. B RETSCHNEIDER 2007), and
the price for such high-tech products typically decreases exponentially with years
of experience and mass production.
Occasionally the question comes up about a theoretical maximum of luminous
efficiency. It is difficult to answer: consider a (very hypothetical) light source con-
verting electric power to a 100 % into a spectrum, which in the visible (380 and
760 nm) corresponds exactly to the ideal white light of our sun, and is zero else-
where. According to (1.90) this would correspond to a luminous efficiency of 38 %.
Hence, conversion of 100 % electric energy into visible light is per definition
a completely unrealistic goal: it would mean totally green illumination. Thus, any
40 1 Basics
Fig. 1.18 B RAGG reflection l

sta e
at two lattice planes of a cry rfac
crystal with a distance d. G su
Note that the B RAGG angle
is defined complementary to X-r
ay
in k'
the angle of incidence
commonly used in reflection ut
optics k al o
n
sig
d
d lattice planes
s
d
claims for a luminous efficiency higher than 3438 % have to make massive com-
promises on the effective colour temperature. We shall certainly witness an exciting
development in the lighting industry over the coming years.
1.4.9 X-Ray Diffraction and Structural Analysis
Electromagnetic radiation in all spectral ranges light in the widest sense is today
one of the most important tools to reveal the structure and dynamics of matter. In
later chapters we shall learn a lot about the various spectroscopic techniques that are
used in this context.
At this point we want to mention, at least briefly, one of the key methods for struc-
tural analysis: X-ray diffraction, i.e. scattering and interference of electromagnetic
radiation of very short wavelength from crystalline matter. It is treated systemati-
cally in textbooks and monographs on solid state physics (see also Sect. 1.7.2).
The basis for a variety of such methods is multiple beam interference from the
crystal lattice structure studied. As sketched in Fig. 1.18 a multitude of parallel, so
called lattice planes may be thought to reflect the X-rays. According to Fig. 1.18
the optical path difference between two rays reflected from neighbouring planes is
2s = 2d sin , where d is the distance of two lattice planes in the crystal and the
so called B RAGG angle.
Thus, the reflected X-ray beams (shown in Fig. 1.18 are only two of them) inter-
fere constructively if and only if B RAGGs law (also called B RAGG condition)
2d sin = z with z = 0, 1, 2 . . . (1.91)
holds. Here is the wavelength of the scatted X-ray radiation and z is an integer.
For reference we also report some further quantitative relations relevant to X-ray
scattering. One defines a reciprocal lattice vector
G = hg 1 + kg 2 + lg 3 (1.92)
1.4 The Photon 41
constructed with the basis vectors of the unit cell g 1 , g 2 , g 3 in the reciprocal lattice,
using the so called M ILLER indices h, k, l that characterize the lattice planes in the
crystal. Without entering into details of crystal lattice theory we just note that these
basis vectors in the reciprocal lattice relate to the ordinary basis vectors a 1 , a 2 , a 3
in the crystal lattice by
g i a j = 2ij , (1.93)

and in Fig. 1.18 we identify d 2/|G|. With wave vectors k and k of incoming
and outgoing X-ray radiation B RAGGs law (1.91) is now written
k = k k = G. (1.94)
Normally one investigates only elastic scattering where |k| = |k | and one may
rewrite (1.94) as (k G)2 = k 2 or G2 2kG = 0. We may finally write the condi-
tion for diffraction:
2
G G
2kG = G 2
or k = . (1.95)
2 2
In the latter form the diffracted wave vectors describe a plane bisecting the recip-
rocal lattice vector G and being perpendicular to it. One may construct such planes
for all elementary reciprocal lattice vectors and combine them to a closed surface
in reciprocal lattice space. They form the so called B RILLOUIN zone (BZ) that rep-
resents all wave vectors of radiation that can be B RAGG reflected by the crystal.
BZs are a very important concept in solid state physics, specifically relevant for the
theory of band structure as we shall briefly discuss in Sect. 2.8. Since several dif-
ferent lattice vectors with different alignment in space exist (typically more than 3,
see Eq. (1.92)), BZs may be rather complex surfaces. Corresponding surfaces may
also be constructed from 2G, 3G etc. and one distinguishes the 1st, 2nd, 3rd, etc.
B RILLOUIN zone.
The intensity of the diffracted X-ray radiation depends on the (electron) charge
density distribution (r) in the unit cell of the crystal and is proportional to the
absolute square of the so called structure factor

F(hkl) = d3 r(r) exp(iG r) (1.96)
cell

= Fj (G) exp i2(xj h + yj k + zj l) .
j
The summation has to be carried out over all atoms in the unit cell of the crystal,
each of which is characterized by its atomic form factor18

Fj (q) = d3 rN (j ) (r) exp(iq r). (1.97)
atom
18 The correct quantum mechanical equivalent (8.21) will be discussed in Vol. 2.

42 1 Basics
(a) (b)
Fig. 1.19 (a) X-ray diffraction image from a large single crystal (80 100 50 m) of the human
enzyme prolidase, recorded with h = 13.05 keV at the beamline BL14.1 of the Free University
Berlin at BESSY.The maximum resolution of the diffraction image corresponds a lattice plane
distance of the crystal lattice of 0.25 nm. In the squares magnified sections of the image are shown.
(b) Secondary structure mode of the enzyme in dimer form with unbound Mn2+ (red spheres).
With kind permission from M UELLER et al. (2007) and private communications
These form factors have to be determined for each atom by integration over the par-
ticle density N (j ) (r) of all its electrons (the dimension of N being L3 ). Without
entering into the details we note here that the imaginary part of the atomic form fac-
tor is related to the photo-absorption cross section, while the real part characterizes
the elastic (coherent) photon scattering. For radially symmetric charge distributions
it is
(j )
N (r) sin(qr) 2
Re Fj (q) = 4 r dr, (1.98)
0 qr
with the momentum transfer q = 2k sin(/2) = 4 sin(/2)/ where = 2 is the

light scattering angle. Today these atomic form factors are well known and tabulated
(see e,g. C HANTLER et al. 2005).
We cannot go into details of different experimental methods for obtaining X-ray
diffraction data. But we mention that in addition to laboratory X-ray sources, syn-
chrotron radiation (SR, see Sect. 10.6.2) plays a key role in structural analysis, in
particular for large biological molecules. In order to illustrate the astonishing per-
formance of state-of-the-art SR diffraction techniques, we show in Fig. 1.19 one
particularly impressive example: an X-ray diffraction pattern from a human enzyme
which has been obtained with synchrotron radiation by rotational exposure. One
illuminates the object with rather monochromatic X-ray light (W/ W 5000 to
10 000) and rotates the crystal with a certain angular increment, in the present case
through 0.5 . During this rotation the multitude of reflexes shown in Fig. 1.19 be-
comes visible.
1.5 The Four Fundamental Interactions 43
Section summary
Photons have (i) a well defined energy W = h = (with = c/ and
= ck) as determined by the photoelectric effect, (ii) a momentum p = k
documented by the C OMPTON effect, and (iii) an angular momentum . They
have no rest mass and exist only moving with the speed of light.
The spectrum of electromagnetic radiation (we call it light in a general
sense) ranges from radio frequencies ( km, h 109 eV) to -rays
(105 nm, h 108 eV). The visible spectrum is only a very small part of
it (the wavelength range from 380 and 760 nm).
P LANCKs radiation law (1.81) was a corner stone in the development of quan-
tum physics. It describes the spectrum of a black body such as our sun or
incandescent light bulbs. The maximum of the spectrum shifts with tempera-
ture according to W IENs displacement law max T 2.9 106 nm K, the total
intensity (per unit area) follows the S TEFAN -B OLTZMANN law I (T ) = B T 4 .
A hypothetical, ideal light source with a luminous efficiency of 100 %, driven
by 1 W electric power emits 683 lm (lumen) at 555 nm. Incandescent light
bulbs have a luminous efficiency of only 23 %.
X-ray radiation is a very powerful tool for structural analysis. Constructive
interference occurs at angles of incidence (in respect of the lattice plane)
according to the B RAGG law 2d sin = z. Other formulations of the B RAGG
law make use of the reciprocal lattice vector (1.92) and M ILLER indices. The
structure factor (1.96) and the atomic form factor (1.97) describe the intensi-
ties in the diffraction pattern.
1.5 The Four Fundamental Interactions
We cannot give here an even brief introduction into particle physics and the un-
derlying theory. The so called standard model (SM) of quantum chromodynamics
(QCD) is a sophisticated theory which cannot be treated in passing. However, it is
important to know a few basics when discussing the interactions relevant in atomic,
molecular and optical physics. The four fundamental interactions
1. Gravitation
2. Electromagnetic interaction
3. Weak interaction
4. Strong interaction
may be seen as what holds the world together in its innermost folds as far as we
understand it today.
In our daily life we are confronted almost exclusively with the first two of these
forces in particular gravitation plays a key role in our everyday experience, while
electromagnetic interactions are perceived commonly in a more indirect manner:
through the action of various machines and equipment, via lighting gears (or occa-
44 1 Basics
Table 1.7 The four fundamental interactions, exchange bosons and coupling constants
Interaction Fermion Exchange Mass Couplingb Range Dependence
(e.g.) boson / GeV c2 /m on distance
Gravitation e , p, n Gravitona 0 5.9 1039 1/r 2
Electromagn. e , p Photon 0 7.30 103 1/r 2
Weak e , W -boson 80.4 105 1018 1/r 5
Z0 -boson 91.2 107 1/r 7
Strong p, n -meson 135.139 1 1015 1/r 7
Quarks Gluons 0 0.119
a Hypothetical, not yet found
b Values for gravitation and electromagnetic interaction refer to a pair of protons
sionally electrostatic discharges), magnetic attraction (e.g. by compass needles) or

even via natural phenomena such as lightning.
In contrast, the whole physics and chemistry of atoms, molecules, condensed
matter and quantum optics is essentially determined by electromagnetic interaction
while the extremely weak gravitational force is of relevance only in exceptional
cases and so are the weak and strong forces which can be neglected almost com-
pletely in AMO and condensed matter physics (we shall mention some exceptions
in later chapters).
Weak and strong interactions play, however, a crucial role when we are interested
in the structure of nuclei, in their stability and decay and of course if we are
interested in the genesis and development of the universe. The unified description
of electromagnetic and weak interaction is called electroweak interaction.
In Table 1.7 the most important characteristic of the four fundamental interac-
tions are summarized: The second column gives some examples of building blocks
of matter (here fermions, i.e. particles with spin quantum number s = 1/2) which
are subject to these forces. These fermions (and their composites) interact with each
other by exchanging characteristic virtual particles, the so called exchange bosons
or vector bosons (with spin quantum number s = 1, except for the graviton which
should have a spin s = 2 and is as yet unobserved). They are listed in the third col-
umn and their respective mass is given in column four of Table 1.7. An order of
magnitude of the strengths of these forces is estimated in column five of Table 1.7
a direct comparison being, however, impossible due to the different dependence on
distance and symmetry properties of these forces. Only electromagnetic interaction
(C OULOMBs law) and gravitation decrease universally with 1/r 2 (if r refers to the
distance of two particles), the range being given as in contrast to the weak and
strong forces that act only on close encounter.
1.5.1 COULOMB and Gravitational Interaction
The interaction of two particles depends in general in a complex manner from their
distance r. Only the well known force laws for gravitation
m1 m2 r
F g = G (1.99)
r2 r
and electromagnetic interaction, i.e. C OULOMBs law
1 q1 q2 r
Fe = (1.100)
40 r 2 r
can be written in this simple, closed form, with masses m1 , m2 and charges q1 , q2 of
the interacting particles, respectively. They may be derived from a scalar potential
V (r) 1/r by
F = grad V (r). (1.101)
While the gravitational constant G has to be measured as one of the fundamental

constants experimentally, the electric constant 0 is related to the speed of light
c = 0 0 which in SI units is a defined quantity as well as the magnetic constant 0 .
The potential energy of an electron at a distance a0 (first B OHR orbit, see atomic
units in Sect. 1.8) from a positive charge e is V (a0 ) = e2 /(40 a0 ) = 2 me c2 .
The fine structure constant according to (1.10) is thus the coupling constant of
C OULOMBs law in units of me c2 .19
Exact values for the coupling constants G and 1/(40 ), as well as , are docu-
mented in Appendix A. The electromagnetic force between two protons is 1.21036
stronger than gravitation among these particles. Thus, electromagnetic interaction
plays a dominant role on the atomic and molecular length scale (and is still relevant
at subatomic distances). For the structure and properties of atoms, molecules and
condensed matter it is more or less exclusively responsible. It determines e.g. the
orbits of electrons around atomic nuclei (or more precisely: the probability distribu-
tions of the electrons).
In contrast, macroscopic objects of our daily life as well as planets and stars
in the universe are essentially uncharged when seen from the outside: they con-
sist of an almost identical number of protons and electrons (quasi-neutrality). Thus,
macroscopic objects interact with each other only via gravitation that determines
their motion (in spite of the extremely small coupling constant G) as long as these
macroscopic objects are not driven by other mechanical, chemical or electrome-
chanical means. Even then gravitation usually plays a crucial role, if we think e.g.
of driving a car, flying an airplane, steering a rocket, or if we consider the tidal mo-
tions of the oceans. The whole complex structure of todays universe has developed
almost exclusively under the influence of graviton for now about 13.75 billion years
since about 1 million years after the big bang.
19 We note here that for very high energies the electromagnetic coupling constant em changes
(essentially ln W ); at energies 90 GeV (roughly corresponding to the mass of the Z boson)
experiments have determined 1/em 1/128.
46 1 Basics
1.5.2 The Standard Model of Fundamental Interaction
The other two forces, strong interaction and weak interaction, play an important role
on a subatomic length scale only: as indicated in Table 1.7, their range of interaction
is finite in complete contrast to C OULOMBs law and gravitation. Strong interac-
tion is mainly responsible for the cohesion of nuclear matter, i.e. it takes care of
the stability of nuclei which otherwise under the influence of electrostatic repulsion
would simply explode. Weak interaction on the other hand plays a decisive role in
-decay, i.e. when an electron is emitted from an atomic nucleus.
Electromagnetic, strong and weak interaction are described today consistently
and convincingly by the standard model of QCD only gravitation resists so far
a unified interpretation. At the core of the theory is a set of objects that may be
considered point like20 (<1018 m) real elementary particles without any internal
structure but with some well defined properties in terms of mass, charge and spin
(intrinsic angular momentum). They are summarized in Fig. 1.20.
The standard model distinguishes two varieties of elementary particles:
1. Matter particles
6 quarks are subject to strong interaction and carry one of three colour charges
red, green or blue; they also carry a charge +2/3e or 1/3e and are
subject to electromagnetic as well as to weak interaction.
6 leptons are all subject to weak but not to strong interaction i.e. they are
colourless. The electron, muon and tauon carry also an electric charge e
and are subject to electromagnetic interaction. In contrast, the corresponding
neutrinos do not have an electric charge. The have, however, a (very small)
mass.
All of these are fermions with spin 1/2. They are complemented (not shown
in Fig. 1.20) by the corresponding antiquarks (in the colours antired, antigreen,
antiblue) and antileptons, their charges are opposite.
2. Exchange particles (or gauge bosons) are considered to be exchanged between
the matter particles during interaction. They are all bosons with spin quantum
number s = 1 and may be seen as mediators of the forces. They are formed for a
very short time only, i.e. virtually, and are unobservable directly.21
20 For experts: e p collision experiments at HERA, with momentum transfer up to Q2 =

40 000 GeV2 , have not shown any deviations of the observed scattering signal from that expected
theoretically for a point-like electron and quark structure. This corresponds to a smallest resolved
size <103 fm, i.e. less than 1/1000 of the proton radius.
21 The virtual exchange of particles may be seen to support our phantasy in understanding the
interaction at a distance: a virtual particle of energy W or mass m = W/c2 can be created for a very
short time given by the limits of the energy-time uncertainty relation to be discussed in Sect. 1.7.3.
The three generations of Gauge or exchange

matter particles (fermions) particles (bosons)
Quarks: each may have

Mass1) 2.4 MeV 1.27 GeV 171.2 GeV

electro-
one of the colours
Charge 0
Spin
red green blue
2/3
1/2 u 2/3
1/2 c 2/3
1/2 t 1 magnetic
interaction
Name Up Charm Top Photon
4.8 MeV 104 MeV 4.2 GeV 0

strong
-1/3
1/2 d -1/3
1/2 s -1/3
1/2 b 0
1 g interaction
Down Strange Bottom Gluon2)
< 2.2 eV <0.17 MeV <15.5 MeV 91.2 GeV

e 0
Z0
Neutrinos
0 0 0 electroweak
colourless
1/2 1/2 1/2 1 interaction

El. neutrino Myon neutr. Tau neutr. Z Boson
0.511MeV 105.7MeV 1.777 GeV 80.4 GeV

-1
e -1
-1
1
W
electroweak
Leptons
1/2 1/2 1/2 1 interaction

Electron Myon Tauon W Boson
1) 1MeV/c 2 = 1.073 10-3 u = 1.957 a.u. (me)

HIGGS
2) each 1 colour + 1 anti colour (8 combinations) Boson*)
Fig. 1.20 Elementary particles according to the standard model: matter particles (fermions, spin
1/2) and exchange particles (gauge bosons, spin 1). The masses and charges are indicated:
masses are given in terms of their energy equivalent Wm = mc2 . *) According to CERN (2013)
On 4 July 2012, the ATLAS and CMS experiments at CERNs Large Hadron Collider announced
they had each observed a new particle in the mass region around 126 GeV /c2 . . . consistent with
the H IGGS boson but it will take further work . . . . This is based on theoretical work for which
E NGLERT and H IGGS (2013) received the N OBEL prize
We have discussed already in Sect. 1.4 the photon. It is the exchange particle
for electromagnetic interaction by which all charged particles interact (in addition
to weak or strong interaction and to gravitation). Strong interaction is characteristic
for quarks as well as for antiquarks (not shown in Fig. 1.20) with charge 2/3e
and +1/3e. They are never observed as free particles and occur only as compound
systems (confinement). On the other hand, at very close distance their interaction
tends to zero (so called asymptotic freedom).
In analogy to electric charge one attributes to the quarks a colour charge by
which they interact with each other. The colour comes in three varieties red, green
or blue (in our graphs coloured red, grey or black). In each composite particle the
colours have to add in such a manner that their sum leads to white (in analogy to
the colours observed from the visible spectrum of light). Antiquarks are coloured
with complementary colours antired, antigreen or antiblue. The exchange bosons
of the strong interaction are called gluons (reminding us of glue). They come in
8 different varieties encompassing one colour and one anticolour each (while the
48 1 Basics
Proton Mass
mp = 1836.15 me
u
Charge = 1e0 = 1.67261027 kg
u Spin = 1/2 = 938.27231 MeV/c 2
d stable = 1.007 276 47 u
Neutron
Charge = 0 mn = 1838.68 me
u Spin = 1/2 = 1.6749 1027 kg
d unstable = 939.5656 MeV/c 2
d t1/2 = 10 min 14s = 1.008664 92 u
Fig. 1.21 Structure of the two nucleons: proton and neutron. They are built from up and down
quarks (u and d, respectively). The special choice of colours is here without direct relevance and is
supposed to just illustrate the principle of colours in quantum chromodynamics: each of the three
fundamental colours (commonly called red, green, blue) has to be present exactly once. The wiggly
lines indicate exchange of gluons (strong interaction)
photon does not contain any charge). Both, gluons and the photon are massless
particles while the three exchange bosons of weak interaction Z0 , W+ and W are
again colourless but have each a (different) mass.
1.5.3 Hadrons
Barions and Mesons

All further particles observed in nature or generated by particle accelerators and
storage rings are composed of the elementary particles summarized in Fig. 1.20. Of
special importance are the two nucleons, proton p and neutron n. Their composition
by three quarks each is sketched in Fig. 1.21. Note that the charge of the proton
(q = 1e) and neutron (q = 0) is obtained by simply adding the charges of the quarks
involved (2/3, +2/3, 1/3) and (2/3, 1/3, 1/3), respectively. The three colour
charges red, green and blue of the three quarks add to white in each case.
The spin of proton and neutron has to be constructed from the three constituents,
each of which has spin 1/2. According to the rules of angular momentum algebra
a spin of 1/2 or 3/2 could in principle result. Apparently 1/2 for both, neutron and
proton is realized.
In contrast, the mass is not simply additive. While neutron and proton have
masses about 1840 times that of an electron, the mass of up and down quarks are
only somewhat larger than the mass me of the electron according to Fig. 1.20 a fac-
tor 5 and 10, respectively. Obviously the binding energy Wb = mc2 contributes
most to the masses of proton and neutron.
In addition to these two long-lived nucleons, which are special cases of baryons
(heavy particles), about a hundred others are known or predicted. They are con-
structed from three quarks in a similar manner. All of them are rather short-lived
and differ in their inner structure, energy (and thus mass), spin and charge.
u u
n d d p
d u
W _
e
Fig. 1.22 F EYNMAN diagram for the neutron decay by weak interaction (half lifetime
1/2 = 10 min +14 s): by emission of an electron e and an antielectron neutrino e a down quark
d (charge 1/3e) is changed into an up quark u (charge +2/3e). The interaction occurs via a
virtual W boson
In contrast to the barions the so called mesons (medium heavy particles) are built
from a quark and an antiquark with complementary colours (colour and anticolour),
which makes them also white. Barions and mesons together are called hadrons.
Hadrons are all subject to strong interaction, to weak interaction and as far as
electrically charged also subject to electromagnetic interaction.
Decay of the Neutron

One typical example of such an interaction is -decay of the neutron through weak
interaction. The isolated neutron is not stable outside the atomic nuclei, albeit rather
long-lived (half lifetime about 10 min 14 s, mean lifetime 885.7 0.8 s). Schemati-
cally this decay is illustrated in Fig. 1.22 by a so called F EYNMAN diagram (more
about these useful diagrams in Sect. 6.5.6).
The three constituents of the neutron move as indicated by the arrows through
time. A down quark emits a virtual W boson and thus the neutron becomes a
proton (see Fig. 1.21). The W boson finally decays into a newly generated lepton
pair e + e . From energy and momentum conservation it is clear that at least two
particles have to be created, since these two fundamental conservation laws could
otherwise not be fulfilled.
1.5.4 The Electron
The leptons electron, muon and tauon and their respective neutrinos are, in contrast
to the hadrons genuine elementary particles, at least as far as we know today. They
interact among each other as well as with quarks, protons and other barions and
atomic nuclei by electro-weak and (as far as charged) by electromagnetic interac-
tion.
In atomic, molecular and optical physics we are almost exclusively concerned
about electromagnetic interactions and by far the most important particle is the elec-
tron. Let us have a somewhat closer look at it. Its rest mass is
me = 9.10938291(40) 1031 kg = 5.4857990946(22) 104 u.

50 1 Basics
Fig. 1.23 Principle of the

d +
setup for determining the
elementary charge according U ~ 5000 V
to M ILLIKAN

As an elementary particle, the electron is considered as point like, more precisely:

the C OULOMB law is valid at least down to r 1018 m. Thus, the electron is at
least three orders of magnitudes smaller than the proton. The electron has a mass
me and an inherent angular momentum, the so called electron spin. The latter is
characterized by a spin quantum number s = 1/2 and has a measurable value of
angular momentum in respect of any given direction in space, say z, of sz = /2.
Connected with the spin is its magnetic moment ge B s, as we shall discuss in
more detail in Sect. 1.9.
Note: the electron has all these properties irrespective of its vanishing spatial
extension. From a classical viewpoint this is a very strange object! The so called
classical electron radius

re = e2 / 40 me c2 = 2 a0 = 2.82 106 nm (1.102)
is to be considered a purely numerical quantity equating the C OULOMB energy at

a distance re to the rest mass energy me c2 of the electron, with being the fine
structure constant (1.10). Historically, re was thought to be related to the radius of
a homogeneously charged sphere whose C OULOMB energy is equal to me c2 . To-
day we know that the electron is as point-like a particle as we can measure it (see
footnote 20). Nevertheless, re often appears as a useful abbreviation in physical rela-
tions. From Sect. 1.4.2 we recall in this context the reduced C OMPTON wavelength
of the electron C = C /2 = a0 another length scale related to the electron.
Attempts to measure the electron charge have already been undertaken in 1899
by J.J. T HOMSON, who determined e/me from the deflection of a moving electron
in a magnetic field (Nobel prize for T HOMSON 1906).
The first accurate determination of e goes back to M ILLIKAN (1913) who re-
ceived the N OBEL prize in physics 1923 for this work. As sketched in Fig. 1.23 he
determined the charge q of small oil droplets with mass m by letting them fall or rise
in an electric field. The velocity v of the particles is determined by the gravitational
force mg, the electric force qE and the frictional force 6rv = mg qE with
E = U/d. The charge q = N e is a (small) multiple of the elementary charge e
and can be determined by variation of the field if one knows the radius of the par-
ticles r and the viscosity of the carrier gas. The elementary charge is of course
a very important fundamental constant that may be determined today with great
accuracy:
e = 1.602176487(40) 1019 C . (1.103)
Its precise value is constant subject of improvement and the interested reader may
want to consult NIST (2010).
1.6 Particles in Electric and Magnetic Fields 51
Generally, free and isolated particles are only observed with an integer multiple
of e (positive, zero or negative). We point out again, however, that the bound quarks
and antiquarks have charges 2/3e or 1/3e.
Directly related to the elementary charge is the energy unit electronvolt
1 eV = 1.602176487(40) 1019 J [per particle]. (1.104)
It corresponds to the change in kinetic energy T = (mv 2 /2) = eU of a charge

e after passing a voltage difference U = 1 V. The unit eV is conveniently and com-
monly used in AMO physics.
Section summary
The four fundamental interaction responsible for keeping the world at its
innermost together are gravitation, electromagnetic interaction, weak inter-
action, and strong interaction. Gravitation and electromagnetic interaction en-
ergies are 1/r, i.e. have in principle infinite range, and are dominant in
macroscopic physics and atomic physics, respectively. Weak and strong inter-
action are of short range on the order 1018 and 1015 m, respectively, and
from an atomic physics point of view refer to forces inside the nuclei.
The standard model of elementary particle physics describes electromagnetic,
weak and strong interaction in a unified scheme. It comprises 6 quarks (matter
particles, fermions), 6 leptons (including the electron, all fermions, subject
to weak and electromagnetic interaction, if charged) and 4 types of exchange
bosons, including the photon which mediates electromagnetic interaction. The
quarks have charges 2e/3, e/3, spin 1/2 and come with three kinds of
colour charges. They are subject to weak, strong and electromagnetic forces.
They are building blocks of all other massive particles, including the nucleons
(neutron and proton) which in turn are constituents of atomic nuclei.
Leptons are a different kind of elementary particles, subject only to weak
and electromagnetic interaction (if charged). Of particular interest in atomic
and molecular physics is the electron. We memorize its electric charge e
1.6 1019 C. To move it through an electric field requires energy or sets
energy free. The atomic unit of energy is thus 1 eV 1.6 1019 J.
1.6 Particles in Electric and Magnetic Fields
We now have a look at characteristic motions and frequencies of charged particles

in electric (E) and magnetic fields (B) as useful background for later discussion.
One aspect is mass spectroscopyand energy analysis or energy selection for which
the properties of trajectories can be exploited. In the following we shall explore the
possibilities and limits of such methods.
52 1 Basics
A particle of momentum p, velocity v and charge q experiences in an electro-

magnetic field the L ORENTZ force:
dp
F= = q(E + v B). (1.105)
dt
In this form the equation is correct even in the relativistic limit. However, for sim-
plicity we shall restrict the following discussion to the nonrelativistic limit unless
pointed out differently.
1.6.1 Charge in an Electric Field
Consider a particle beam with velocity v moving in +x-direction. It enters a purely

electric field E = U/b at x = 0. We choose a geometry such that mvx = 0 and
mvy = qU/b and assume the field to be strictly limited to the lengths of the capacitor
as sketched in Fig. 1.24. This can actually be achieved by placing the aperture plates
(vertical, grey lines) at potential U/2 in a specific distance to the capacitor plates as
investigated in the early days of particle optics by H ERZOG (1935).
For the lateral deflection d of the beam we obtain from integrating N EWTONs
second law of motion
2
q U 2 q U qU 2
d= t = = . (1.106)
2m b 2m b v 4T b
Geometry (b and ) as well as the voltage U applied are known. The ratio (and only
the ratio) of kinetic energy T to charge q may be determined according to (1.106) by
the electric deflection method! This is valid in general, completely independent of
how complicated the setup may be. In practical applications one often uses special
capacitor designs e.g. cylindrical or hemispherical capacitors (the latter is described
in some detail in Appendix B.3, Vol. 2) as well as segments of these for determining
the particle energy if its charge is known. In addition to a measurement of the energy,
for extended, divergent particle beams such setups allow to focus the trajectories
from the entrance aperture onto the exit aperture of an energy analyzer.
Fig. 1.24 Trajectory (red + U

line) of a charged particle in
the electric field of a plane b x
capacitor (heavy black lines) q, v d
which is limited by entrance
apertures (heavy grey lines) 0

y
1.6.2 Charge in a Magnetic Field
To also determine the mass of a particle one needs an additional measurable quan-
tity. In a purely magnetic field B the movement of the charge q is determined by
dp
F= = qv B
dt
with the force F being perpendicular to v and B (see Fig. 1.25). The change in
energy
dW
= vF = qv v B 0
dt
is thus identical to zero, and v = const. This holds also relativistically so that = 0
and = const.
The relativistic equation of motion thus reads
dp dv
= m = qv B,
dt dt
with the rest mass m of the particle and p = mv. The particle moves on a circle
that is determined by balancing the centrifugal force Fc and the magnetic force
F = qvB:

dp 2 2
= Fc = mv = mv = qvB, (1.107)
dt r r
p
so that mv = p = qBr or = rB (1.108)
q
mv p
and r = = . (1.109)
qB qB
If one wants to use this orbital radius r for mass selection, again only one parameter
can be extracted: with a magnetic field the ratio of momentum and charge p/q can
be measured.
The expressions given above also hold in the nonrelativistic case where 1.
For highly relativistic energies as they are e.g. realized in electron storage rings built
for the generation of synchrotron radiation (see Sect. 10.6.2) one has more or less
Fig. 1.25 Moving charge

(red line)
in a magnetic field q, m B
B grey pointing into the
v
drawing plane
r

v

54 1 Basics
exactly v c and the relativistic energy (1.19) becomes mr v 2 mc2 = W . Thus,

(1.107) can be written

dp W c
or = (1.110)
dt r r
where is the radial acceleration a related to c. The latter relation follows from
(1.107) with a = Fc /( m) = v 2 /r c2 /r.
1.6.3 Cyclotron Frequency and ICR Spectrometers
According to (1.109) in a sufficiently extended magnetic field a charged particle

with rest mass m, charge e and kinetic energy T moves on a circular orbit. Its angular
frequency is called cyclotron frequency:
v eB 1 eB 1
c = = = . (1.111)
r m m 1 + T /(mc2 )
The turnaround time is Tc = 2/c = 2 m/Be. At low kinetic energies (T
mc2 ), cyclotron frequency and turnaround time are independent of radius and ve-
locity of the particle studied and simply depend on the ratio of mass to charge. At
high velocities a relativistic correction is required.
Ion cyclotron resonance (ICR) mass spectrometers are today among the most
important and accurate mass spectrometers for chemical and biological analysis.
One stores ions with different m/q in an ion trap and lets them interact with high
frequency electromagnetic radiation (RF from MF to HF, see Fig. 1.14). If the fre-
quency of this radiation is in resonance with c for a particular m/q in the trap,
these ions (and only these) are accelerated. The thus induced image currents can
readily be detected.
In practice one may either tune the RF field over the bandwidth of interest, or
alternatively use a very broad band (short pulse) and F OURIER analyze the result-
ing signal (FT-ICR). This kind of technique is also used for optical spectroscopy
and discussed in some detail in Sect. 5.3.2, Vol. 2. Todays commercial devices al-
low to analyze m/q up to several 1000 u (where q can be rather high so that even
large biomolecules may be studied) and achieve mass resolutions (FWHM) up to
m/ m 100 000.
1.6.4 Other Mass Spectrometers
The so called magnetic mass spectrometers are based in fact on a combination of

electric and magnetic fields. As we have seen
T
the electric field selects: = f (U, geometry)
q
p
the magnetic field selects: = rB.
q
If one combines the two types of fields judiciously one may achieve with T =
p 2 /2m a determination of m/q:
m 1 p 2 /q 2
= . (1.112)
q 2 T /q
In the classical, so called double focussing mass spectrometers, ions pass succes-
sively electric and magnetic fields. Such devices are still used today but have lost
much of their importance in relation to the FT-ICR mass spectrometers discussed in
Sect. 1.6.3.
Another type of mass spectrometers, today widely used, is the so called
quadrupole mass spectrometer (QMS). It exploits the mass dependence of the dy-
namic stability regions for ion trajectories in a rapidly oscillating electric quadrupole
field (see e.g. D EHMELT and PAUL 1989): for specific ratios of DC to AC fields only
ions with a specific mass to charge ratio m/q may pass on a stable trajectory through
the setup while all other ions undergo increasing oscillations and do not reach the
exit slit. The interesting mathematics follows M ATHIEUSs differential equation (see
e.g. W IKIPEDIA CONTRIBUTORS 2013).
In time of flight mass spectrometers one accelerates ion in electric fields to select
m/q through their time of flight to the detector. In state-of-the-art setups of this type
one exploits a variety of tricks to focus ions of the same mass but from different
starting positions at the ion source (W ILEY and M C L AREN 1955) as well as those
with different initial kinetic energies (M AMYRIN 1994) at the same time on the
detector. With present commercial solutions high mass to charge ratios m/q (up
to 4000 u) may be reached and high mass resolution (m/ m 60 000) can be
achieved.
For special applications two other, classical setups are still useful and deserve to
be mentioned. We already mentioned J.J. T HOMSONs method with parallel electric
(E) and magnetic (B) fields which nowadays is used to analyze particle emission
with high kinetic energies T . Assume the fields to point into +y-direction, the ions
enter the field perpendicular in +z-direction (velocity vz ) and pass through them on
a length l. In this case the L ORENTZ force (1.105) leads to acceleration both in y-
as well as in x-direction. After a time t1 = l/vz the particles have passed the fields
and acquired velocities
qE qEl qBvz qBl

vy = t1 = and vx = t1 = . (1.113)
m mvz m m
Behind the fields the ions are detected on a position sensitive detector: J.J. T HOM -
SON used a photographic plate, today sophisticated electronic devices are available
as we shall describe in Appendix B.1, Vol. 2. If the detector is positioned at z = s
far behind the field, the ions hit the detector at a time t2 = s/vz and their deflections
are
qEls qEls qBls qBls
y = v y t2 = 2
= and x = vx t2 = = . (1.114)
mvz 2Wkin mvz 2mWkin
56 1 Basics
Fig. 1.26 Charges in a solid, electrons

displaced by an external
++++++
electric field
++++++

++++++
x ions
Eliminating vz leads to the so called T HOMSON parabolas

m E 2
y= x . (1.115)
q B 2 ls
They describe the positions where the ions hit the detector and allow to determine
m/q and for known q. In addition, with (1.114) one may also determine the kinetic
energy.
In contrast to the T HOMSON setup, the so called W IEN filter uses crossed electric
and magnetic fields to select charged particles according to their velocity. For parti-
cles that hit the crossed E and B fields perpendicularly one may set the ratio of the
fields such that the resulting forces (1.105) for a given velocity v just compensate
each other, i.e. q(E vB) = 0. For
v = E/B
the charged particles just fly on straight line trajectories. If the mass is known one
may thus determine the momentum and hence the kinetic energy.
1.6.5 Plasma Frequency
Electrons in plasmas but also in clusters and condensed matter (in particular quasi
free electrons in metals and semiconductors) may carry out collective oscillations,
so called plasma oscillations with a characteristic frequency that plays an impor-
tant role in many areas of physics. The simplest model starts with is a quasi neutral
plasma of electrons e and ions with a density N of quasi free electrons, equal to
that of the ions. Assume we displace all electrons in respect of the ions by a distance
x as shown in Fig. 1.26.
This leads to a surplus of surface charge density = eN x that causes an elec-
tric field between electrons and ions E = /r 0 . The equation of motion for each
electron in this field is thus

me x = eE = e/r 0 = e2 N/r 0 x.
This differential equation is simply that of a harmonic oscillator with a force con-
stant kP = e2 N/r 0 . The angular frequency of the oscillations

kP N e2
p = = (1.116)
me me r 0
1.7 Particles and Waves 57
is called plasma frequency, sometimes also L ANGMUIR or D RUDE frequency. It

plays a key role when describing electrical and optical properties of electron gases
in metals, semiconductors or plasmas. For reference but without going into details
we mention that the plasma frequency is, however, not an oscillation frequency in
the standard sense, and that the dielectric function of the plasma is given by
p2
r () = 1 , (1.117)
2
which for < p becomes negative.
To interpret this result, we consider a plane wave exp[i(kx t)] propagat-
ing in +x-direction, with k = n/c being the magnitude of the wave vector in

the medium and n = r the index of refraction. According to (1.117), below the
plasma frequency n becomes imaginary and we write k = i. Correspondingly the
wave, exp(x) exp(it), is damped. Thus, the medium absorbs at all frequen-
cies < p and is transparent for > p .
We finally note that in isolated particles, the situation is somewhat more compli-
cated, e.g. so called plasmon resonance phenomena will be observed at p .
Section summary
In an electromagnetic field E, B particles with a charge q are subject to the
L ORENTZ force F = q(E + v B), where v is the particles velocity.
This can be used to manipulate and guide the charged particles (typically elec-
trons and ions) and to select their energy and/or mass/charge ratio.
We note two interesting frequencies: (1) In a homogeneous magnetic field B
a free electron moves on a circle, perpendicular to the field (or spirales) with
the cyclotron
frequency c = v/r = eB/( me ). (2) The plasma frequency
p = Ne 2 /(m e r 0 ) is characteristic for oscillations of quasi free electrons
in clusters and metals (particle density N ).
1.7 Particles and Waves
1.7.1 DE BROGLIE Wavelength
In 1923, Louis DE B ROGLIE (N OBEL prize in physics 1929), a French aristocrat,

argued in his PhD thesis that particles should also have wave properties, just as
electromagnetic waves have particle properties as demonstrated by the photoelec-
tric and the C OMPTON effect (see Sect. 1.4.1 and Sect. 1.4.2). His postulate, now
well proven, states that in full analogy to light (1.74) the momentum p of a particle
is related to its wave vector k with k = 2/dB and wavelength dB by
p = k and p = h/dB = /c, (1.118)

58 1 Basics
(a) (b) diffraction patterns

target
e-beam
screen
cathode +U
aperture
Wkin low Wkin high
Fig. 1.27 D EBYE -S CHERRER diffraction schematically: (a) setup and generation of the diffrac-
tion cones; (b) typical diffraction patterns for a poly-crystalline target: top view onto the screen (a)
as seen from the right
with dB commonly called DE B ROGLIE wavelength.

The postulate of DE B ROGLIE gave birth to the famous concept of wave-particle
duality. In its most simple form a matter wave may be written as
(r) = C exp(ik r), (1.119)
i.e. as a plane wave with C being a suitably chosen normalization constant.

In the nonrelativistic limit a slow electron with kinetic energy T has a wavelength
h 1.23 nm
dB = = . (1.120)
2me T T / eV
At a kinetic energy 100 eV the wavelength of an electron is thus on the order of the
size of an atom 0.1 nm. Note, however, that (1.118) also holds in the relativistic
case, with p given by (1.25). Specifically, for electrons that have been accelerated
by a voltage U one has to use

eU
p= 2me eU 1 + . (1.121)
2me c2
With the rest energy me c2 = 0.511 MeV for electrons this leads already at a moder-
ate kinetic energy of 50 keV to a reduction of the wavelength of 2.5 % in comparison
to (1.120). In a precision experiment this clearly has to be accounted for.
1.7.2 Experimental Evidence
DEBYE-SCHERRER Diffraction of Electrons

We remind the reader here of a some applications that illustrate the wave nature of
matter in a particularly obvious manner. The D EBYE -S CHERRER method, which is
also well established in X-ray diffraction, is one of the most prominent examples.
It is used for the structural analysis of poly-crystalline material which may e.g. be
deposited on a thin carbon film. The setup is sketched in Fig. 1.27(a).
(a) fluorescence screen electron energy (b)

selecting grids
electron gun
target
electron beam
elastically
diffracted electrons
Fig. 1.28 Low energy electron diffraction (LEED): (a) experimental setup; (b) typical diffraction
pattern from a single crystalline ordered surface
The diffraction structures are generated by many single reflexes from the
micro-crystals and appear according to (1.91) under the B RAGG angles 2 =
arcsin(z/2d) with z = 0, 1, 2 . . . The various distances d of the crystals lattice
planes contain information about the crystal structure. Since the crystals are aligned
at random into all directions of space, all electrons diffracted into an angle 2 will
be emitted into a cone around the incoming electron beam. For each m and each
d such a cone exists. For each cone a circle is formed at its intersection with the
plane of observation (photographic plate, or today more likely a CCD camera). In
Fig. 1.27(b) such diffraction patterns for electrons of low and higher energy are
sketched schematically.
Low Energy Electron Diffraction (LEED)

Electron diffraction is used in a variety of methods for structure determination of
matter. In contrast to the D EBYE -S CHERRER method, diffraction from single crys-
tals leads to point patterns. Low energy electrons may be used with advantage for
structure determination of surfaces in a setup illustrated in Fig. 1.28. The method is
known as LEED (low energy electron diffraction) and commercially available.
Diffraction of Neutrons, Atoms and Molecules from a Grating

Low energy neutrons have a DE B ROGLIE wavelengths on the order of atomic
diameters (1 eV = 0.029 nm). In contrast to charged particles that interact by the
C OULOMB force with the electron shells of the atoms, neutrons interact only at
short distance with the atomic nuclei through the strong force (see Sect. 1.5). At the
same time they penetrate very deep into the matter studied. Thus, thermal neutrons
are an excellent tool for structural analysis. The special form of atomic shell does
not play any role in neutron scattering which makes a detailed evaluation of diffrac-
tion patterns much more simple than those from charged particles. Such experiments
are today an indispensable method in structure determination from materials re-
search to biology and a number of dedicated nuclear reactors are available for this
purpose worldwide. In future, alternatively so called spallation sources are planned
or under construction, e.g. the European Spallation Source (ESS) is currently being
built in Lund, Sweden. First neutrons are expected for 2019.
Thermal and supra-thermal neutral He atoms are also used for structural inves-
tigations. Since they cannot enter into the material they are particularly well suited
for surface analysis. Figure 1.29 documents impressively that slow He atoms may
60 1 Basics
Fig. 1.29 Scattering of a 10 5

neutral helium atom beam 0.
through a transmission 1. order
diffraction grating with 10 4
100 nm slit distance as
intensity / particle s-1

3.
reported by S CHLLKOPF
and T OENNIES (1996) 10 3
2. 5.
4.
7.
10 2 6.
8. 9.
10
6 4 2 0 2 4 6
scattering angle / mrad
even be diffracted by quasi macroscopic optical elements, and they interfere. For
this nice experiment a transmission grating was manufactured by state-of-the-art
nano-technology. The diffraction pattern may be completely understood in terms
of K IRCHHOFFs diffraction theory which has been developed more than 150 years
ago.
One may of course ask how far one can push such kind of wave optics with
particles and particle beams. This is an interesting theme which may even be po-
tentially relevant for technical applications. Atom lithography and atom optics with
very cold atoms is subject to current research. We have already mentioned briefly in
Sect. 1.3.4 research on B OSE -E INSTEIN condensation (BEC). One particular fasci-
nating aspect is that the wavelength of the matter waves increasing with decreasing
temperature according to (1.118). It eventually reaches values comparable to the av-
erage distance between atoms (particle density N 3 ). With (1.118) and (1.50)
this happens when
h2
T N 2/3 . (1.122)
3mkB
Comparison with (1.71) shows that this estimate agrees with the critical temperature
for BEC apart from a numerical prefactor and the missing spin degeneracy. This
is a nice, plausible result.
One may also push the wave concepts into the opposite direction and ask about
the macroscopic limits of the wave-particle
duality: at constant kinetic energy the
DE B ROGLIE wavelength decreases m, i.e. with the square root of the mass.
Diffraction experiments with larger molecules are thus increasingly more difficult
and much more challenging.
Z EILINGER and his associates (see e.g. A RNDT et al. 1999) were able to show
that even such big objects as Fullerene molecules, C60 , are subject to wave mechan-
ics. In a single slit diffraction experiment one observes quite normal diffraction
patterns, as documented in Fig. 1.30 and as one actually would expect. The ul-
timate conundrum in this context is about really large objects as sketched in the
cartoon Fig. 1.31 the question is, what will happen? We leave this as a little mental
Fig. 1.30 Diffraction of C60 1200
signal pulses / (50 s)

molecules by a slit.
Experiment according to 1000
A RNDT et al. (1999). The 800
grey background indicates the
structure of C60 600
400
200
100 50 0 50 100
detector position / m
Fig. 1.31 Wave-particle

duality for macroscopic
objects? Cartoon after
Wolfram v. Oertzen
?
exercise for the reader who may want to estimate orders of magnitude for a pertinent
experiment.
1.7.3 Uncertainty Relation and Measurement
The H EISENBERG uncertainty relation
px x (1.123)
belongs today well-nigh to the canonical intellectual inventory of an educated mem-

ber of society (as far as she or he shows any interest for sciences at all). Expressed in
words the uncertainty relation (1.123) states that space and momentum or some-
what more general canonical conjugate quantities cannot be measured simultane-
ously with accuracy.22
The classical Gedankenexperiment devised by H EISENBERG and illustrated in
Fig. 1.32 analyzes the attempt to study an electron with an optical microscope. As
it turns out, all endeavours to localize the electron with precision, are limited by its
wave nature. If we want to observe the electron by light, its wavelength h has to
22 Note that measurements of microscopic observables are based on probability distributions. Thus,
the exact value on the right hand side of the uncertainty relation (, h/2, h etc.) depends on the na-
ture of that distribution and on the exact definition of uncertainty, e.g. FWHM, foot width, position
of the first minimum, etc.
62 1 Basics
Fig. 1.32 Classical Gedankenexperiment according to H EISENBERG why an electron may not
be localized by an optical microscope better than allowed by the uncertainty relation
Fig. 1.33 Trying to localize
x
an electron beam by an
aperture leads to diffraction
px
and hence to uncertainties in e-
the determination of the p

electrons lateral momentum
x
be small and the opening angle of the microscope has to be kept large, in order
to obtain high resolution. According to A BB the smallest structure x one may
resolve in an optical microscope is determined by its numerical aperture n sin and
the wavelength
h
x = , (1.124)
n sin
with n = 1 for the optical index of refraction (in vacuum). Small and large sin
imply, however, an intrusion into the experimental situation, since the photon will
thereby lose or change its momentum p by
h
px = p sin = sin ,
h
and transfer it to the electron. Together with the A BB relation (1.124) we obtain
px x h and thus have derived the uncertainty relation (1.123). Any other
method trying to localize the electron leads eventually to the same result.
Assume e.g. we try to localize an electron beam by a slit this leads to diffrac-
tion as indicated in Fig. 1.33. To obtain an estimate for the uncertainty in the x-
component of the electron momentum px = p sin we may use the angle min at
which the first minimum of the diffraction pattern occurs. The latter is well known
from the corresponding formula in wave optics:

sin min = .
x
Using this one reads from Fig. 1.33
h/p
px = p sin = p =p = px x = h,
x x
which again is equivalent to the uncertainty relation (1.123).
We note at this point that an analogues uncertainty relation between energy W

and time t also holds:
W t . (1.125)
1.7.4 Stability of the Atomic Ground State
The big dilemma of physics at the beginning of the 20th century was the inability
to understand stable electron orbits around the atomic nuclei. Electrons on circu-
lar orbits are accelerated charged particles which according to classical physics
should continuously radiate electromagnetic waves and hence loose energy. Thus,
they should get slower and slower with time and being negatively charged should
eventually fall into the positively charged nucleus!
Alas! Why do the electrons remain on stable orbits? As we shall see in Sect. 1.8,
B OHR simply postulated that stable electron orbits have angular momenta equal to
integer multiples of . Among theses stationary states according to B OHR there
are also excited ones which decay spontaneously by emitting radiation according to
an exponential decay law exp(At). But still, the ground state is completely stable
even in the B OHR model!
A quantitative solution of this problem is given by quantum mechanics. Inter-
estingly, the few basics about the wave nature of the electron which we have dis-
cussed so far, already allow us to answer this puzzle in principle. We simply use
the H EISENBERG uncertainty relation to obtain an estimate for the minimum en-
ergy without adopting a specific model for the atom: Let a be the mean radius of
the atom. With a high probability electrons will be found within that radius. This
immediately leads to an estimate of the uncertainty p in a determination of the
electrons momentum by writing (1.123) now
a p and p p /a. (1.126)
Consequently the kinetic energy of the electron becomes
p2 2
T= .
2me 2me a 2
Together with the potential energy
e2
V =
40 a
this gives a total energy W T + V . Assuming the equality to hold one obtains
2 e2
W= . (1.127)
2me a 2 40 a
64 1 Basics
W obviously depends on the atomic diameter a. So let us now find the lowest pos-
sible energy! We just apply the standard rule of calculus to find this minimum:
dW 1 42 0 e2 me a !
= = 0,
da 4 me a 3 0
from which we obtain for the radius of the atomic ground state:
42 0 0 h2
a0 = = . (1.128)
e2 me e2 me
Putting this back into (1.127) the total energy of the atomic ground state becomes
1 me e4 me e4
Wmin = 2 2
= 2 . (1.129)
32 2 0 80 h2
This is exactly the value of the B OHR model for the ground state of the hydrogen
atom! Big surprise with such a simple estimate? Well, the idea is obviously right.
One should, however, not overestimate the quantitative assertion: the correct numer-
ical result is a direct consequence of the specific form of (1.126) with on the right
hand side, rather than h/2 or something else (see footnote 22). And from this we
have derived the specific estimate for the minimum product of position and momen-
tum uncertainties.
Section summary
We have discussed here some elements of quantum mechanics. While the pho-
ton was recognized as the particle associated with electromagnetic waves,
we now associate matter waves with the well known particles. The DE
B ROGLIE wavelength of particles dB = h/p also holds for relativistic ve-
locities if the correct expression for p is used.
In the nonrelativistic limit, for
slow electrons, we memorize dB 1.2 nm / T / eV. Convincing experimen-
tal evidence comes from diffraction experiments with particles.
The uncertainty relation may be said to express the particle wave duality in
terms of minimal uncertainties in a measurement of two canonic conjugate
variable, e.g. for momentum and position p x or for energy and time
W t . The precise value of the lower limit (, h/2 or the like) depends
on the specific problem and on the accuracy which appears tolerable.
The stability of the atomic ground state can be viewed as a direct consequence
of the uncertainty relation. (Why do negative electrons not fall into the positive
nucleus?)
1.8 BOHR Model of the Atom
In 1913, during the big time of quantum mechanical discoveries, Niels B OHR
(N OBEL prize in physics 1922) worked as a young postdoc with Ernest RUTHER -
1.8 BOHR Model of the Atom 65
FORD in England. RUTHERFORD had developed an atomic model based on his scat-
tering experiments of alpha particles by atoms according to which . . . the atoms
consist of a positively charged nucleus surrounded by a system of electrons kept
together by attractive forces from the nucleus; the total negative charge of the elec-
trons is equal to the positive charge of the nucleus. Further, the nucleus is assumed
to be the seat of the essential part of the mass of the atom, and to have linear di-
mensions exceedingly small compared with the linear dimensions of the whole atom
. . . . J.J. T HOMSON expanded this by assuming that . . . the atom consists of a
sphere of uniform positive electrification, inside which the electrons move in circu-
lar orbits (both quotes from B OHR 1913).
As discussed above, the big challenge in physics was then to explain the highly
problematic fact that these atoms do not collapse: why do electrons circle on sta-
ble atomic orbits without loosing energy? Remember, this was 10 years before
DE B ROGLIE postulated his matter waves as discussed in Sect. 1.7.1! B OHR , so
to say, just cut the Gordian knot and developed the fundamental ideas for a the-
ory of atomic structure. He knew from BALMERs spectroscopic work, that the
energy levels of atoms follow the phenomenological equation W n2 with in-
teger numbers n. These numbers became B OHRs quantum numbers. Since accord-
ing to classical mechanics the energy of a particle on a circular orbit scales in-
versely to the square of its orbital angular momentum, B OHR suggested that the
angular momenta of stable orbits be directly proportional to these quantum num-
bers. And since the P LANCK constant h, then already known from P LANCKs ra-
diation law and E INSTEINs explanation of the photoelectric effect, has the unit of
an angular momentum, he postulated that angular momenta be quantized in units
of h/2 = . From a classical view point, this quantization could neither be proven
nor even be justified, but B OHR demonstrated that with these assumptions the ex-
perimentally observed spectrum of hydrogen could be explained with unique preci-
sion.
1.8.1 Basic Assumptions
Many versions and numbers of B OHRs postulates may be found in the literature
and in a multitude of Web pages on the subject. In his original paper B OHR (1913)
used the following two principle assumptions:
1. The dynamical equilibrium of the systems in the stationary states can be dis-
cussed by help of the ordinary mechanics, while the passing of the systems be-
tween different stationary states cannot be treated on that basis.
2. The latter process is followed by the emission of a homogeneous radiation, for
which the relation between the frequency and the amount of energy emitted is
the one given by P LANCKs theory.
B OHR, in principle, considered elliptical orbits but (wisely) focused on circular

ones for which he assumed a simple condition: that the angular momentum of the
66 1 Basics
electron round the nucleus in a stationary state of the system is equal to an entire
multiple of a universal value which he identified as h/2 , today written as . In
concise form, the angular momentum L = r p is quantized and its magnitude on
a stationary circular orbit is
h
L = rme ve = n = n, (1.130)
2
with the velocity ve , the electron mass me and the radius r of the orbit.
The second of B OHRs principle assumptions is about the photon energies emit-
ted in a transition from one stationary orbital n to another one n . In mathematical
form it reads:
h = Wn Wn . (1.131)
In classical mechanics, an electron (charge = e) may be on any circular orbit

around the nucleus (charge = +Ze) for which the centrifugal force is balanced by
the C OULOMB force:
me ve2 Ze2 Ze2

= = r= . (1.132)
r 40 r 2 40 me ve2
Thus, we have derived a classical expression for the radius, while the kinetic energy
may be written in terms of the potential V (r)
me ve2 Ze2 1
T= = = V (r).
2 80 r 2
(Note that the latter equality is directly given by the C OULOMB law but also corre-
sponds to the classical virial theorem.) Thus, the total energy of the system is
Ze2
W = T + V = T = < 0, (1.133)
80 r
with the minus sign characterizing bound states: one defines zero total energy W = 0
for an electron that is just no longer bound (no kinetic energy at r ).
Finally, the quantization condition (1.130) is introduced by rewriting the kinetic
energy, again using classical mechanics:
L2 (n)2
W = T = = . (1.134)
2me r 2 2me r 2
1.8.2 Radii and Energies
One may now obtain independent expressions r(n) = an and W (n) = Wn for any
given n = 1, 2, . . . < by solving the set of Eqs. (1.133) and (1.134) accordingly.
One finds B OHRs orbital radii
n2 0 h2 n2
an = = a0 (1.135)
Z e2 me Z
for the stationary states of hydrogen (Z = 1) and hydrogen like ions (Z > 1), with
n called principle quantum number and a0 B OHRs radius:
0 h2
a0 = = 5.2918 1011 m. (1.136)
e2 me
The latter is used as unit of length in atomic physics (we memorize a0 0.05 nm).
The total energy (or term energy) of the stationary states is
Z 2 me e4 Z2 Z2
Wn = = E h = R hc. (1.137)
n2 802 h2 2n2 n2
The quantity Eh is used as atomic energy unit23 and amounts to
me e4 e2
Eh = 2R hc = = = 4.359 1018 J =
27.211 eV, (1.138)
402 h2 40 a0
while the RYDBERG constant R = Eh /(2hc) = 10 973 731.568527(73) m1 is

used in spectroscopic context, often given in wavenumbers, cm1 . Today it is
the most precisely measured fundamental constant (see also Sect. 6.5.4).
1.8.3 Atomic Units (a.u.)
It is often convenient to measure observable properties of quantum systems in so

called atomic units (a.u.), i.e. write energies as W/Eh with the atomic unit of en-
ergy Eh according to (1.138), lengths in r/a0 with B OHRs Radius a0 according to
(1.136), and times in t/t0 . With the atomic velocity

v0 = Eh /me = Eh /me c2 c = /me a0 = c, (1.139)
(with the fine structure constant , Eq. (1.10)) one defines
a0 a2 2 2 h3
t0 = = 2me 0 = 0 4 = 2.4189 1017 s. (1.140)
v0 h me e
23 Notto be confused with the rest energy of the electron Wrest = me c2 , also called natural unit of
energy (see e.g. NIST 2010).
68 1 Basics
Fig. 1.34 Orbital radii rn = a0 n2 /Z and energies Wn = Eh Z 2 /(2n2 ) for H (Z = 1, red) and He+
(Z = 2, black) according to the B OHR model
The atomic unit of mass is the rest mass of the free electron:
1 a.u. 1me = 9.10938291(40) 1031 kg. (1.141)
Note, however, that this must not be confused with the unified atomic mass unit
1 u = 1.660538921(73) 1027 kg, (1.142)
which is defined as 1/12 of the mass of an unbound atom of the nuclide 12 C at rest
and in its ground state. It is a unit outside the SI but widely used and recommended
by the Comit International des Poids et Mesures.
1.8.4 Energies of Hydrogen Like Ions
Ions with only one electron, i.e. He+ , Li++ , . . . , U91+ are called hydrogen like. For
the examples H and He+ Fig. 1.34 illustrates the magnitudes of orbital radii rn =
n2 a0 /Z and energies Wn = (Z 2 /2n2 )Eh .
1.8.5 Correction for Finite Nuclear Mass
So far we have always assumed that the nuclear mass is at rest (mn = ). However,
correctly we should describe the electron motion in the centre of mass system.
For a two particle system this kinematic correction is done by simply replacing
the electron mass me with the reduced mass m e:
me mn
me m
e = . (1.143)
me + mn
B OHRs radius has thus to be replaced by

me
a0 a 0 = a0 , (1.144)
e
m
and the term energies become
Z2 Z2
Wn = E h = RH hc, (1.145)
2n2 n2
with Eh and the RYDBERG constant R (1.138) being replaced by
e
m e
m
E h = Eh and RH = R , respectively. (1.146)
me me
1.8.6 Spectra of Hydrogen and Hydrogen Like Ions
From the term energies (1.137) one derives with (1.131) the famous RYDBERG
(RYDBERG -R ITZ) formula for the spectra of hydrogen and hydrogen like ions H,
D, He+ , Li++ , Be+++ , . . . , U91+ :

1 1
Wn1 n2 = = h = Wn1 Wn2 = Z R 2 22
(1.147)
n1 n2

1 1 1
= = Z R 2 2
2
in wavenumbers (1.148)
n1 n2

1 1
= Z 2 R c 2 2 as transition frequency. (1.149)
n1 n2
State-of-the-art spectroscopy of the hydrogen atom contributes today decisively to
the most precise measurements of fundamental constants. We shall discuss their so-
phisticated methods and their amazing precision in Sect. 6.5.4. In order to compare
these experimental results with various theoretical predictions one has to correct
them of course according to (1.146) with the different kinematic correction factors
for each nuclear mass.
1.8.7 Limits of the BOHR Model
The B OHR model works surprisingly well for H and H like ions. B OHRs term
energies (1.137) in these cases are identical to those derived from nonrelativistic
quantum mechanics. Unfortunately, the model fails for all other atoms in spite of
some serious efforts made in the early days of quantum mechanics. It also fails in
the relativistic case when ve /c = (e2 /20 hc)(Z/n) = Z/n. Here the fine struc-
ture constant 1/137 according to (1.10) appears again. Relativistic effects ob-
viously become important for large Z and low n. We may neglect them as long as
Z/n 1.
70 1 Basics
So far we have also completely ignored the spin of the electron which is also of
relativistic origin. The electron spin leads to an extension of the concept of angu-
lar momentum and one of its practical consequences is fine structure splitting (see
Sect. 6), which the B OHR model of course does not consider.
We have to be aware that the concept of electron orbitals with well defined radii
will have to be reconsidered very critically: strictly speaking, from todays under-
standing of quantum mechanics, the B OHR model is simply not correct as we shall
discuss in detail in Sect. 2.6.11. Still, it serves in many cases as a helpful first step
towards formulating and understanding the correct quantum mechanical theory.
Section summary
The B OHR model of hydrogen (and H-like atoms), historically one of the cor-
nerstones in the development of modern physics, makes surprisingly correct
predictions for energy levels, Wn = Eh Z/(2n2 ) and transition frequencies
h = Wn Wn . . . though just based on heuristic postulates. Even the B OHR
radii rn = a0 n2 /Z allow a reasonable interpretation as we shall see in the next
chapter. Thus, the B OHR model remains a reference for quick cross checks
and estimates. Also, its educational value should not be underestimated.
For higher precision, some small correction for finite mass of the nucleus are
needed as summarized in Sect. 1.8.5.
Using the results of the B OHR model, we have introduced the system of
atomic units (a.u.) with energies measured in Eh (twice the binding energy
of the H atoms ground state), lengths in a0 (the radius of the first B OHR or-
bit), time in t0 (the time an electrons needs to circle once around that orbit),
and the mass in me (the rest mass of the electron).
1.9 STERN-GERLACH Experiment and Space Quantization
One of the key experiments for quantum mechanics has been performed in 1922 by
S TERN (N OBEL prize 1943) and G ERLACH. It showed with then unsurpassed clarity
that classical mechanics and electrodynamics are unable to explain the observations
on the level of atomic dimensions.
1.9.1 Magnetic Moment and Angular Momentum
Before trying to understand this experiment and its consequences, we have to recall
some basics from mechanics and electrodynamics. According to the B OHR model
the electron circles around the atomic nucleus, a motion that in a classical picture
may be viewed as a spinning top.
As indicated in Fig. 1.35 this orbit is connected with an electric current
e ev
I= = .
t 2r
1.9 S TERN -G ERLACH Experiment 71
Fig. 1.35 Magnetic moment

of an electron on a circular ML e0
v
orbit r
+I
of
This current, enclosing an area A (A = r 2 ), generates a magnetic moment M
the magnitude
ev evr eL
M = IA = r 2 = = .
2r 2 2me
It is proportional to the magnitude of L and oriented into the opposite direction. The
magnetic moment of an orbital angular momentum is thus
e L
M= L = B . (1.150)
2me
Note: the gyromagnetic ratio derived here, M/L = e/2me = ge B is indepen-
dent of the specific geometry of the motion. It does not depend on whether a point
charge or an extended charge distribution orbits around the nucleus. However, this
above classical derivation is valid only for orbital angular momenta. It will have
to be modified in view of the results of the S TERN -G ERLACH experiment.
In the B OHR model angular momenta (1.130) are quantized and appear only
in units of . Thus, L/ is an integer number in B OHRs theory, and the unit of
magnetic moment is the so called B OHR magneton
e 1
B = = 927.400915(23) 1026 J T . (1.151)
2me
1.9.2 Magnetic Moment in a Magnetic Field
In a magnetic field B as a consequence of the L ORENTZ force a magnetic mo-

ment M experiences a torque
T = M B. (1.152)

Its potential energy T d is thus given by
VB = M B = MB cos , (1.153)
where is the angle between M and B and zero energy has been assumed for
M B. Minimum potential energy corresponds to M parallel to B, and L an-
tiparallel to it.
72 1 Basics
Fig. 1.36 Precession of the B dL

angular momentum L in a L
magnetic field B T L
e0
ML
The force on the magnetic dipole M as whole in this potential is
F = grad V (r).
Since M is independent of r, one obtains

B B B
F = M B = Mx + My + Mz , (1.154)
x y z
(see e.g. S TERN 1921, rewritten in SI units).
Thus, in a homogeneous magnetic field no net force acts on a magnetic dipole,
while in an inhomogeneous field it experiences a force proportional to its magnetic
moment. However, the torque (1.152) acting on the dipole moment leads to a change
of the angular momentum of the spinning top. With dL = Ld the equation of
motion is
dL d
T= =L = LL . (1.155)
dt dt
In the special geometry shown in Fig. 1.36 with L B, this becomes
L
LL = MB = B B.

The torque T forces the angular momentum L to precess around B (gyroscopic
motion) with an angular frequency, called L ARMOR frequency:
M B e
L = B= B= B. (1.156)
L 2me
1.9.3 The Experiment
Otto S TERN had invented the so called molecular beam method24 and already ap-
plied it to determine the M AXWELL -B OLTZMANN velocity distribution in gases
(see Sect. 1.3.4). In his famous paper Ein Weg zur experimentellen Prfung der
24 In this and later context, an atomic or molecular beam source may be thought to just consists of
a reservoir with the species to be investigated (possibly heated, or cooled, possibly mixed with an
inert carrier gas). Through a small orifice in it, the atoms or molecules diffuse or stream into the
vacuum, and their divergence angle is collimated by one or more apertures along the beam axis.
z (a) z (b)
atom
oven magnet B on S
S B off
N x y
magnet B on N
slit photo plate profile
Fig. 1.37 Schematic of the S TERN -G ERLACH experiment. (a) The atomic beam from the oven
is collimated through a slit, may then be deflected in an inhomogeneous magnetic field, and will
finally hit a photographic plate. (b) Side view of magnet poles and atom beam
(a) z
y
z
(b) y
Fig. 1.38 Two wire field: it corresponds approximately to the field created by two wires through
which a current flows. The corresponding magnetic field lines are illustrated in (a). In an experi-
ment the field is realized by a pair of suitably shaped permanent magnets as indicated in (b). The
red line parallel to the y-direction indicates the position of the atomic beam entering the space be-
tween the magnets from the back. It will be deflected by the inhomogeneous field in z-direction
Richtungsquantelung im Magnetfeld he proposed (S TERN 1921) an experiment to

determine the magnetic moment of an atom, based on the deflection of an atomic
beam in an inhomogeneous magnetic field.
The Original Atomic Beam Experiment

The experiment as it was actually carried out in 1922 by S TERN and G ERLACH
with silver atoms (Ag) is illustrated schematically in Fig. 1.37. With (1.154) and
B = (0, 0, B) at y = 0 the symmetry of the setup leads there to Fx = 0 and Fy = 0.
Since the components Mx and My will precess around the z-axis, the force due to
these components averages out. Thus, we are left with the z-component of the force
(1.154):
z B
Fz = M . (1.157)
z
A small but important detail is the physical realization of an inhomogeneous

field. In the original S TERN -G ERLACH experiment a so called two wire field was
used which is illustrated in Fig. 1.38.
74 1 Basics
z
(a)
x
z
(b) x
(c) x
Fig. 1.39 Classical expectation about the S TERN -G ERLACH experiment: (a) View onto the
atomic beam coming out of the yz plane without magnetic field. (b) Alignment of the magnetic
moments prior to entering the magnet, distributed at random. (c) Atomic beam profile after deflec-
tion by the magnetic field. The distribution would arise from B/z being strongest for x = 0 and
decreasing for |x| > 0
What Do We Expect?
Let us assume the profile of the atomic beam without magnetic field arriving in the
detector plane is shaped as indicated in Fig. 1.39(a). Classically, the magnetic mo-
ments M are expected to be randomly distributed in all directions as indicated in
Fig. 1.39(b), their projections M z onto the +z-direction ranging from B L/ to
+B L/. The atomic beam enters the inhomogeneous magnetic field and its atoms
are deflected corresponding to their distribution of Mz . Behind the magnet the re-
sulting spatial distribution is registered by a suitable detector. With the magnetic
field switched on, the classical expectation is to observe a broadening of the beam
in z-direction proportional to the field and the average magnetic moments reflect-
ing the randomly distributed Mz components. The deflection on the z-axis (x = 0)
will be largest since there the field gradient B/z is highest.
And what Is Observed in the Experiment?

In their experiment with Ag atoms S TERN and G ERLACH used a photographic plate
as detector. The traces of the silver beam condensed on it. The highly remarkable
result of these efforts is a pattern schematically sketched in Fig. 1.40 in complete
contrast to the classical expectation: obviously there are only two directions into
which the atomic magnetic moments are oriented. The particular (elliptical) shape
is due to the vanishing field gradient at both sides of the beam.25
More clearly one recognizes the result by tracing the measured signal along the
z-axis in Fig. 1.40 and plotting the absorbance due to the silver deposited as a func-
tion of z. This is schematically shown in Fig. 1.41. The comparison with and without
25 A quaint note aside: the silver traces only became visible after Otto S TERN blew the smoke of
his (sulfur containing) cigar onto the photographic plate an early, unintended contribution to the
photochemistry and catalysis of the photographic development process.
Fig. 1.40 Schematic sketch z

of the pattern observed in the
S TERN -G ERLACH
experiment: the beam is split
y
into two components
Fig. 1.41 Result of the z

S TERN -G ERLACH classical
experiment (cut along the expectation
z-axis in Fig. 1.40): Splitting B on
in two discrete states when B=0
the magnet is switched on and
comparison with the classical B on
expectation
atom signal
magnetic field bears out the dramatic splitting into two components. This observa-
tion, which is also made for many other atoms, is a direct consequence of space
quantization as we shall discuss in a moment.
L ANGMUIR -TAYLOR Detector

We note here in passing that today many very efficient methods for particle de-
tection exist. Todays reproductions of the S TERN -G ERLACH experiment (but also
many modern atomic beam experiments) use detectors exploiting the so called
L ANGMUIR -TAYLOR effect which in itself is interesting. It is based on the quan-
tum tunnelling effect, to which the valence electron of an atom with low ionization
potential WI (e.g. K) is exposed when hitting a metal surface with high work func-
tion WA (e.g. W).
Figure 1.42 gives an essentially self explaining schematic of such a detector with
the relevant potential energy diagram (see also the potential well model introduced
in Sect. 1.4.1). The electrons tunnel into the metal, the K+ ions remain and can
be detected as ion current after field extraction and amplification, possibly with a
particle multiplier (see Appendix B.1, Vol. 2).
1.9.4 Interpretation of the S TERN-G ERLACH Experiment
How can the results of the S TERN -G ERLACH experiment be interpreted? We re-
member, the B OHR model postulates that angular momenta are quantized according
to (1.130): Thus, we expect L = with to be a positive integer number or zero.
Let us assume now, that a similar relation holds for the components of angular mo-
menta, say for Lz in z-direction. Then Lz would assume 2 + 1 integer multiples
of :
Lz = m with m = , + 1, . . . , . (1.158)
76 1 Basics
(a) tungsten surface (b) K atom

W vacuum vacuum W
x x
WA V(x ) WI
V(x )
electron
lake WA (tungsten) > WI (potassium)
(c) K atom on a tungsten surface (d) detector setup schematically

K electronics for ion
electron x atoms current amplificaton
tunnells
electron after e tunnelling + 50V

lake a K+ ion remains hot tungsten wire
on the surface in a metal cylinder
Fig. 1.42 Schematic of a L ANGMUIR -TAYLOR detector: (a) Potential for an electron at an isolated
tungsten surface, (b) potential and energy of the valence electron in a K atom, (c) potential felt by
a K atom when hitting a tungsten surface, (d) simple detector setup
This quantization is called space quantization (in German language somewhat more
precisely: Richtungsquantisierung). The number m is often called magnetic quan-
tum number since it was originally observed in a magnetic field. We prefer, how-
ever, the more general term projection quantum number.
In contrast to a classical statistical distribution of angular momenta, as illustrated
in Fig. 1.39(b), there are only 2 + 1 allowed projections of L = onto the z-
axis. The number 2 + 1 is called multiplicity of the state (in Sect. 1.3.4 we have
introduced it as degeneracy a term used in the context of energies). According to
(1.150) the magnetic moment is M = B L/, thus space quantization of L also
implies quantization of M. The component Mz may thus assume values from |M|
to |M| which correspond to the components |L| to |L| of the orbital angular
momentum.
A closer quantum mechanical analysis (see Chap. 2) shows that the conjectures
made above are nearly correct except for the fact that the magnitude of the angular
momentum is given by

L = ( + 1), (1.159)
which for large values of gives again L .

But even this heuristic introduction of space quantization, guided by B OHRs
postulates which even turns out to be quantum mechanically correct for orbital
angular momenta does not provide a full explanation of the S TERN -G ERLACH
experiment: the observed splitting in two components (see Fig. 1.41), i.e. the ob-
servation of a multiplicity 2, is obviously contrary to the multiplicity predicted for
the lowest nonvanishing orbital angular momentum ( = 1): one would expect a
multiplicity 2 + 1 = 3, i.e. triple splitting!
1.9.5 Consequences of the STERN-GERLACH Experiment
The S TERN -G ERLACH experiment revealed three dramatic non-classical facts:
1. Space quantization completely unexpected from a classical point of view but

plausible in the framework of B OHRs quantum theory
2. The observed multiplicity does not correspond to the expected whole numbered
angular momenta according to the B OHR model. The twofold splitting 2 =
2j + 1 only allows the conclusion that the atom studied (Ag) has an angular
momentum quantum number j = 1/2. This holds for silver atoms but also for
the alkali metal atoms such as Na, K, . . . . We thus have to extend the B OHR
model and assume that also half integer angular momenta exist. All experimental
observations of quantum systems confirm the hypothesis: Angular momenta J
only exist as integer or half integer multiples of . We have already anticipated
the relevant quantum mechanical relations in (1.5)(1.9).
3. Finally, a quantitative evaluation of the S TERN -G ERLACH experiment shows that
also the magnitude of the classically predicted magnetic moment (1.150) does
not lead to a correct estimate of the observed deflections. One has to generalize
(1.150) and defines the
J
magnetic moment for J : MJ = gJ B (1.160)

and its projection onto a given axis, say z, is given by
MJ z = gj B mj with mj = j, j + 1, . . . , j. (1.161)
The so called L AND g-factor is gL = 1 for pure orbital angular momenta and =2
for electron spin states. We shall derive general expressions gj in Chap. 8. A mul-
titude experiments confirms that the magnetic moment of atoms and molecules is
indeed proportional to the total angular momentum J and oriented opposite to it.
Equation (1.156) has also to be modified:
e
L ARMOR frequency for J : j = gj B = gj L . (1.162)
2me
Section summary
The S TERN -G ERLACH experiment has been another corner stones in the de-
velopment of quantum mechanics. It proofs what we call space quantization:
angular momenta J are oriented in space so that their projection on a given
axis is Jz = mj with mj = j, j + 1, . . . , j where j is the angular mo-
mentum quantum number.
Orbital angular momenta have integer quantum numbers, intrinsic quantum
numbers of particles may be half integer (fermions) or integer (bosons).
78 1 Basics
Fig. 1.43 Vector diagram for z

electron spin (heavy black
arrow): It precesses on a cone
Sz =+ -
around the z-axis with two S 2
possible orientations (dotted
circles with arrow), so that its 3
projection onto the z-axis S =-
2
(indicated by the two red, y
dashed lines) is either +/2
or /2
x Sz = - -
2
S
With each angular momentum J a magnetic moment M = gj B J / is

1
associated, with the B OHR magneton B = e/2me 927.4 1026 J T .
The L AND gj factor is 1 for orbital angular momenta and 2 for the electron
spin.
In an external magnetic field B magnetic moments have a potential energy
VB = M B = MB cos(M, B), they precess under the influence of
the torque T = M B with the L ARMOR frequency j = gj Be/2me . Only
in an inhomogeneous field an overall force (1.154) acts on them.
1.10 Electron Spin
A conclusive explanation for the ground-breaking observations of S TERN and G ER -

LACH was presented not until 1925 by G OUDSMIT and U HLENBECK in the context
of unusual atomic line splitting in a magnetic field (anomalous Z EEMAN effect, see
Sect. 8.1.2): The electron has an intrinsic magnetic moment MS which is associated
with an intrinsic angular momentum S, the electron spin. Identifying in (1.5)(1.9)
J with S and setting the spin quantum number s = 1/2 explains the multiplicity of
2 observed in the S TERN -G ERLACH experiment. Two possible orientations for the
spin exist with projection quantum numbers ms = 1/2, so that S = |S| = 3/2
and Sz = /2. One may visualize these relations in a so called vector diagram,
shown in Fig. 1.43.
In summary, the electron spin is characterized as follows:

3
magnitude |S| = s(s + 1) = 0.88
2
quantum number s = 1/2
(1.163)
multiplicity 2s + 1 = 2
z component Sz = ms with ms = 1/2.
1.10 Electron Spin 79
1.10.1 Magnetic Moment of the Electron
Quantitative evaluation of the deflection observed in the S TERN -G ERLACH exper-

iment for a range of atoms such as Ag, K, Na, . . . shows that the projection of
the magnetic moment onto a given axis is Mz = B , with the B OHR magneton
(1.151).
If we were to apply relation (1.150) M = (e/2me ) L = B L/, which is
valid for orbital angular momenta, to the z-component of the spin, Sz / = 1/2,
we would obtain Mz = B /2 in contrast to the experimentally observed value
Mz = B . Thus, we must apply the generalized expression (1.160) for the mag-
netic moment of the spin.
Specifically for an electron without angular momentum, i.e. for = 0 (as it is the
case with H, Na, K, Ag in the ground state) the magnetic moment of the atom is
determined exclusively by the intrinsic magnetic moment of the electron. From this
follow
S
the magnetic moment of the electron MS = ge B and (1.164)

the g factor of the electron ge 2 (1.165)
We emphasize that this value of the electron g factor cannot be explained by any
kind of classical charge distribution on an orbit, for which always (1.150) holds,
i.e. gL = 1.
In contrast, the relativistically correct D IRAC theory leads exactly to ge = 2 for
the electron spin. Thus, according to (1.161) the component of the magnetic moment
of the electron in respect of a given axis, say z, becomes indeed Mz = ge B /2 =
B as experimentally observed.
It is interesting to note that the L ARMOR frequency (1.162) for an electron with
spin s and ge = 2 becomes
e e
s = ge L = ge B B, (1.166)
2me me
and is, according to (1.111), (nearly) identical to the cyclotron frequency c =
(e/me )B of an electron in a magnetic field (in the nonrelativistic limit).
We point out, however, that high precision measurements show a small, but signif-
icant difference between L and c , i.e. document a deviation from (1.165), in fact
one finds ge 2.0023 . . . . A theoretical understanding of this so called anomalous
magnetic moment of the electron is provided by quantum electrodynamics (QED)
for which T OMONAGA, S CHWINGER and F EYNMAN received the N OBEL prize in
1965 as we shall explicate in Sect. 6.6.
1.10.2 EINSTEIN-DE-HAAS Effect
The so called E INSTEIN - DE -H AAS effect gives one further impressive piece of ev-
idence for the non-classical nature of the magnetic moment of the electron based
80 1 Basics
Fig. 1.44 Principle of the spring wire laser beam setup

E INSTEIN - DE -H AAS for measuring
experiment mirror the twisting
of the wire
C + solenoid
-
ferromagnetic iron core
on a macroscopic measurement. In this experiment one uses the fact (not to be dis-
cussed here in detail) that ferromagnetism is caused by many electron spins which
are oriented in parallel.
One determines the torque exerted onto a (soft) ferromagnetic iron core (cylin-
der) due to a change of magnetization: the magnetic field of the solenoid tends to
orient the magnetic moments of the electron spins in parallel to this magnetic field
and hence there will also be a preferential orientation of the electron spins. These
spins all together constitute a total inherent angular momentum of the cylinder. In
order to change it, a torque has to act on this cylinder which in the setup shown
in Fig. 1.44 arises from the external magnetic field. The torque is measured by the
twisting angle of a thin spring wire (silica).
In a quantitative experiment one starts with a completely demagnetized probe
and allows a well defined current pulse to flow through the solenoid. This leads to
a changing magnetic field Hsolenoid (t) as sketched in Fig. 1.45. During this pulse a
torque acts on the system which was originally at rest.
After the magnetic field has decayed, the remanent magnetization Mrem is as-
sumed to be N times the z-component of the magnetic moment Mz of the individ-
ual electrons. With (1.161) one obtains
e e
Mz = ge B ms = ge Sz Mrem = Brem = N ge Sz , (1.167)
2me 2me
where N is the number of electrons that after the pulse remain oriented (magnetized)
in the direction of the applied magnetic field (remanence). In turn, Brem may easily
Hsolenoid
Bmagnet
Brem
t
Fig. 1.45 Temporal evolution of the magnetic field generated by the solenoid in Hsolenoid after
starting to discharge the capacitor C along with the magnetic induction Bmagnet resulting from it
in the ferromagnetic cylinder leading to a remanent part Brem
Acronyms and Terminology 81
be determined e.g. by measuring a voltage induced by the remanent magnetic field

in a probe coil.
Due to angular momentum conservation we have
N Sz = Irod rod , (1.168)
with Irod = mR 2 /2, the moment of inertia of the ferromagnetic cylinder, m being its
mass and R its radius. The angular frequency of the cylinder rod after application
of the magnetic field may be determined from the maximum twist angle max of the
silica wire onto which the cylinder is mounted: the initial kinetic energy has then
been transformed into potential energy:
Irod 2 kr 2
= max . (1.169)
2 rod 2
From the ratio of Brem according to (1.167) and the angular momentum Irod rod
according to (1.168) we obtain
Brem N ge S z e e
= = ge .
Irod rod N Sz 2me 2me
This finally allows one to determine ge . A quantitative evaluation of such experi-

ments confirms ge 2.
Section summary
The spin related properties of the electron have again be summarized in
(1.163)(1.165).
The E INSTEIN - DE H AAS experiment demonstrates the electron spin on a
macroscopic level using in a ferromagnetic rod. When its magnetization is
reversed in external magnetic field, the magnetic moments of N the electrons
have to reversed from +/2 to /2. Thus, the total angular momentum of
all electrons in the rod changes by N . Since N is a large number, this in-
duces a macroscopic rotation of the rod, which can be measured, confirming
the microscopic measurements of ge .
AC: Alternating current, oscillating electric voltage and current.

BEC: B OSE -E INSTEIN condensation.
BESSY: Berlin Electron Strorage ring for Synchrotron Radiation, Germanys
third generation synchrotron radiation source in Berlin-Adlerhof.
BZ: B RILLOUIN zone, represents all wave vectors of radiation which can be
B RAGG-reflected by a crystal lattice. Important concept in solid state physics.
82 1 Basics
CCD: Charge coupled device, semiconductor device typically used for digital
imaging (e.g. in electronic cameras).
chemical-potential: In statistical thermodynamics defined as the amount of energy
or work that is necessary to change the number of particles of a system (by 1)
without disturbing the equilibrium of the system, see in Sect. 1.3.4.
CIE: Commission international de lclairage, International Commission on Illu-
mination, provides e.g. colorometric tables (http://files.cie.co.at/204.xls).
CMB(R): Cosmic microwave background, radiation at 2.725 K from the origin of
the universe.
DC: Direct current, unidirectional electric voltage and current.
ESS: European Spallation Source, large scale facility for generating neutrons for
structural research, see http://europeanspallationsource.se/.
EUV: Extreme ultraviolet, part of the UV spectral range. Wavelengths between
10 nm and 121 nm according to ISO 21348 (2007).
FIR: Far infrared, spectral range of electromagnetic radiation. Wavelengths be-
tween 3 m and 1 mm according to ISO 21348 (2007).
FT: F OURIER transform, see Appendix I.
FWHM: Full width at half maximum.
HERA: Hadron-Elektron-Ring-Anlage, for collision experiments between elec-
trons of 30 GeV and protons of 820 GeV operated at DESY very sucessfully
until 2007.
HF: High frequency, part of the RF spectrum. Wavelengths from 10 m to 100 m
or frequencies from 3 MHz to 30 MHz according to ISO 21348 (2007).
ICR: Ion cyclotron resonance, spectroscopy (specifically mass spectroscopy)
based on irradiation with a radio frequency in resonance with the cyclotron fre-
quency of an ion in magnetic field (see Sect. 1.6.3).
IR: Infrared, spectral range of electromagnetic radiation. Wavelengths between
760 nm and 1 mm according to ISO 21348 (2007).
LED: Light emitting diode.
LEED: Low energy electron diffraction, see Sect. 1.7.2.
LF: Low frequency, part of the RF spectrum from 30 kHz up to 300 kHz.
LHC: Large hadron collider (not to be confused with left hand circularly polarized
light), high energy physics facility at CERN (Geneva) providing particles with
collision energies up to 14 TeV for protons and up to 1 PeV for heavy ions.
MF: Medium frequency, part of the RF spectrum from 300 kHz up to 3 MHz.
MIR: Middle infrared, spectral range of electromagnetic radiation. Wavelengths
between 1.4 m and 3 m according to ISO 21348 (2007).
MW: Microwave, range of the electromagnetic spectrum. In spectroscopy MW
usually refers to wavelengths from 1 mm to 1 m corresponding to frequencies
between 0.3 GHz and 300 GHz; ISO 21348 (2007) defines it as the wavelength
range between 1 mm and 15 mm.
NIR: Near infrared, spectral range of electromagnetic radiation. Wavelengths be-
tween 760 nm and 1.4 m according to ISO 21348 (2007).
PES: Photoelectron spectroscopy, see Sect. 5.8, Vol. 2.

PET: Positron emission tomography, medical exploitation of positron-electron
annihilation (see Sect. 1.4.3).
QCD: Quantum chromodynamics, the theory of strong interaction (color force).
A fundamental force describing the interactions of quarks and gluons, the con-
stituents of all hadrons.
interaction.
QMS: Quadrupole mass spectrometer, a brief explanation is found in Sect. 1.6.4.
RF: Radio frequency, range of the electromagnetic spectrum. Technically, one
includes frequencies from 3 kHz up to 300 GHz or wavelengths from 100 km to
1 mm; ISO 21348 (2007) defines the RF wavelengths from 100 m to 0.1 mm; in
spectroscopy RF usually refers to 100 kHz up to some GHz.
SI: Systme international dunits, international system of units (m, kg, s, A,
K, mol, cd), for details see the website of the Bureau International des Poids
et Msure http://www.bipm.org/en/si/ or NIST http://physics.nist.gov/cuu/Units/
index.html.
SM: Standard model, of elementary particle physics. The basis of todays under-
standing of matter.
SR: Synchrotron radiation, electronmagnetic radiation in a broad range of wave-
lengths, generated by relativistic electrons on circular orbits.
THz: Tera-Hertz, spectral region of electromagnetic radiation. Wavelengths range
covering parts of MW and IR.
UHF: Ultra high frequency, part of the RF spectrum. Wavelengths from 10 cm to
1 m or frequencies from 3 GHz to 300 MHz according to ISO 21348 (2007).
UV: Ultraviolet, spectral range of electromagnetic radiation. Wavelengths be-
tween 100 nm and 400 nm according to ISO 21348 (2007).
UVA: Ultraviolet a, part of the UV spectral range. Wavelengths between 315 nm
and 400 nm according to ISO 21348 (2007).
UVB: Ultraviolet b, part of the UV spectral range. Wavelengths between 280 nm
UVC: Ultraviolet c, part of the UV spectral range. Wavelengths between 100 nm
VHF: Very high frequency, part of the RF spectrum. Wavelengths from 1 m to
10 m or frequencies from 300 MHz to 30 MHz according to ISO 21348 (2007).
VIS: Visible, spectral range of electromagnetic radiation. Wavelengths between
VUV: Vacuum ultraviolet, spectral range of electromagentic radiation. part of the
UV spectral range. Wavelengths between 10 nm and 200 nm according to ISO
21348 (2007).
XUV: Soft X-ray (sometimes also extreme UV), spectral wavelength range be-
tween 0.1 nm and 10 nm according to ISO 21348 (2007), sometimes up to
40 nm.
84 1 Basics
References
A RNDT , M., O. NAIRZ, J. VOS -A NDREAE, C. K ELLER, G. VAN DER Z OUW and A. Z EILINGER:
1999. Wave-particle duality of C60 molecules. Nature, 401, 680682.
ASTM: 2008. G173-03 Reference Spectra Derived from SMARTS v. 2.9.2, American Soci-
ety for Testing and Materials (ASTM). http://rredc.nrel.gov/solar/spectra/am1.5/ASTMG173/
ASTMG173.html, accessed: 7 Jan 2014.
B OHR , N.: 1913. On the constitution of atoms and molecules. Philos. Mag., 6, 26, 125.
B OHR , N. H. D.: 1922. The N OBEL prize in physics: for his services in the investigation of the
structure of atoms and of the radiation emanating from them, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/1922/.
B RETSCHNEIDER , E.: 2007. Efficacy limits for solid-state white light sources, Laurin Publishing
Co., Inc., Pittsfield MA, USA. http://www.photonics.com/Article.aspx?AID=28677, accessed:
7 Jan 2014.
DE B ROGLIE , L.: 1929. The N OBEL prize in physics: for his discovery of the wave nature of
electrons, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1929/.
CERN: 2013. The search for the H IGGS boson, Geneva. http://home.web.cern.ch/about/
physics/search-higgs-boson, accessed: 8 Jan 2013.
C HANTLER , C. T., K. O LSEN, R. A. D RAGOSET, J. C HANG, A. R. K ISHORE, S. A. KO -
TOCHIGOVA and D. S. Z UCKER : 2005. X-ray form factor, attenuation, and scattering tables
(version 2.1), NIST. http://physics.nist.gov/ffast, accessed: 7 Jan 2014.
C OMPTON , A. H.: 1927. The N OBEL prize in physics: for his discovery of the effect named after
him, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1927/.
C ORNELL , E. A., W. K ETTERLE and C. E. W IEMAN: 2001. The N OBEL prize in physics: for
the achievement of Bose-Einstein condensation in dilute gases of alkali atoms, and for early
fundamental studies of the properties of the condensates, Stockholm. http://nobelprize.org/
D EHMELT , H. G. and W. PAUL: 1989. The N OBEL prize in physics: for the development of the
ion trap technique, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1989/.
D ICK LYON: 2008. Luminosity function, Wikimedia Commons. http://commons.wikimedia.
org/wiki/File:Luminosity.png, accessed: 7 Jan 2014.
E INSTEIN , A.: 1905. ber einen die Erzeugung und Verwandlung des Lichtes betreffenden
heuristischen Gesichtspunkt. Ann. Phys., 17, 132.
E INSTEIN , A.: 1921. The N OBEL prize in physics: for his services to theoretical physics, and
especially for his discovery of the law of the photoelectric effect, Stockholm. http://nobelprize.
org/nobel_prizes/physics/laureates/1921/.
E NGLERT , F. and P. W. H IGGS: 2013. The N OBEL prize in physics: for the theoretical discovery
of a mechanism that contributes to our understanding of the origin of mass of subatomic par-
ticles, and which recently was confirmed through the discovery of the predicted fundamental
particle, by the ATLAS and CMS experiments at CERNs large hadron collider, Stockholm.
http://nobelprize.org/nobel_prizes/physics/laureates/2013/.
H ERZOG , R.: 1935. Berechnung des Streufeldes eines Kondensators, dessen Feld durch eine
Blende begrenzt ist. Arch. Elektrotech., 29, 790802.
H ONSBERG , C. and S. B OWDEN: 2012. PVCDROM Air Mass, UNSW and Solar Power
Labs at ASU, Australia. http://www.pveducation.org/pvcdrom/properties-of-sunlight/air-mass,
accessed: 7 Jan 2014.
ISO 21348: 2007. Space environment (natural and artificial) process for determining solar
irradiances. International Organization for Standardization, Geneva, Switzerland.
KOPP , G. and J. L. L EAN: 2011. A new, lower value of total solar irradiance: evidence and
climate significance. Geophys. Res. Lett., 38, L01706.
M AMYRIN , B. A.: 1994. Laser-assisted reflectron time-of-flight mass-spectrometry. Int. J. Mass
Spectrom. Ion Process., 131, 119.
References 85
M ATHER , J. C. and G. F. S MOOT: 2006. The N OBEL prize in physics: for their discovery of the
blackbody form and anisotropy of the cosmic microwave background radiation, Stockholm.
M ILLIKAN , R. A.: 1923. The N OBEL prize in physics: for his work on the elementary charge
of electricity and on the photoelectric effect, Stockholm. http://nobelprize.org/nobel_prizes/
physics/laureates/1923/.
M UELLER , U., F. H. N IESEN, Y. ROSKE, F. G OETZ, J. B EHLKE, K. B UESSOW and U. H EINE -
MANN : 2007. Crystal structure of human prolidase: the molecular basis of PD disease,
Hinxton, UK: PDB entry 2okn. The European Molecular Biology Laboratory (EMBL-EBI).
http://www.ebi.ac.uk/pdbe-srv/view/entry/2okn/summary.html, accessed: 7 Jan 2014.
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/cuu/
Constants/, accessed: 7 Jan 2014.
NIST: 2011. Conversion factors for energy equivalents, NIST. http://physics.nist.gov/cuu/
Constants/energy.html, accessed: 7 Jan 2014.
O HNO , Y.: 2010. Radiometry and photometry for vision optics. In: M. BASS, ed., Handbook of
Optics, vol. II, 37.1. New York: McGraw-Hill.
PAULI , W.: 1945. The N OBEL prize in physics: for the discovery of the exclusion princi-
ple, also called the Pauli principle, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1945/.
P LANCK , M.: 1900. Zur Theorie des Gesetzes der Energieverteilung im Normalenspektrum.
Verh. Dtsch. Phys. Ges., 2, 235245.
P LANCK , M. K. E. L.: 1918. The N OBEL prize in physics: in recognition of the services
he rendered to the advancement of physics by his discovery of energy quanta, Stockholm.
http://www.nobelprize.org/nobel_prizes/physics/laureates/1918/.
S CHLLKOPF , W. and J. P. T OENNIES: 1996. The nondestructive detection of the helium dimer
and trimer. J. Chem. Phys., 104, 11551158.
SORCE: 2012. SORCE Solar Spectral Irradiance, Boulder, Co.: Laboratory for Atmo-
spheric and Space Physics, University of Colorado and NASA. http://lasp.colorado.edu/
lisird/sorce/sorce_ssi/, accessed: 7 Jan 2014.
SSE: 2012. Solar System Exploration Our Solar System, NASA. http://solarsystem.nasa.gov/
planets/profile.cfm?Display=Facts&Object=Sun, accessed: 7 Jan 2014.
S TERN , O.: 1921. Ein Weg zur experimentellen Prfung der Richtungsquantelung im Magnet-
feld. Z. Phys., VII, 249253. Nachdruck: Z. Phys. D, Atoms, Molecules and Clusters 10, 111
116 (1988).
S TERN , O.: 1943. The N OBEL prize in physics: for his contribution to the development of the
molecular ray method and his discovery of the magnetic moment of the proton, Stockholm.
T HOMSON , J. J.: 1906. The N OBEL prize in physics: in recognition of the great merits of his the-
oretical and experimental investigations on the conduction of electricity by gases, Stockholm.
T OMONAGA , S.-I., J. S CHWINGER and R. P. F EYNMAN: 1965. The N OBEL prize in physics:
for fundamental work in quantum electrodynamics, with deep-ploughing consequences for
the physics of elementary particles, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1965/.
W IKIPEDIA CONTRIBUTORS: 2013. Mathieu function, Wikipedia, The Free Encyclopedia.
http://en.wikipedia.org/wiki/Mathieu_function, accessed: 7 Jan 2014.
W ILEY , W. C. and I. H. M C L AREN: 1955. Time-of-flight mass spectrometer with improved
resolution. Rev. Sci. Instrum., 26, 11501157.
Elements of Quantum Mechanics
and the H Atom 2
Quantum mechanics provides the tools for a quantitative

understanding of atoms and molecules. The reader is expected
to be familiar at least with the main concepts. Here we want to
repeat and refurbish the most important notions and methods so
that we can work with them directly in the following chapters.
Overview
If the reader is already familiar with quantum mechanics he may just want
to browse this chapter and return later if necessary. However, readers who
have experienced quantum mechanics up to now only as compulsory math-
ematical exercise may perhaps read this chapter with advantage and find it
helpful to approach the indispensable instruments without big formal hurdles.
In Sects. 2.12.3 we summarize a minimum of formalism. Section 2.4 treats
as a first example the well known particle in a box and the free electron gas
which in atomic and solid state physics is an important elementary model.
Section 2.5 gives an overview of how to treat angular momenta, needed in
all following text, specified in Sect. 2.5.4 for the electron spin s = 1/2. Sec-
tion 2.6 offers a crash course in nonrelativistic quantum mechanics of the
H atom essential knowledge for all the following chapters. We refrain here
from formal derivations in favour of a plausible, possibly somewhat hand wav-
ing introduction. Finally, Sect. 2.7 presents a first, elementary approach to in-
teractions of atomic electrons with external fields, which will be extended and
deepened in Chap. 8. Clearly, such a brief introduction into quantum mechan-
ics cannot substitute a profound study of the formal theory. It should, however,
facilitate the access to it and make the reader fit for the following chapters.
2.1 Matter Waves
2.1.1 Limits of Classical Theory
The classical picture of a well defined trajectory with known coordinate x(t) and
momentum p(t) looses its validity in quantum mechanics, as sketched in the phase

88 2 Elements of Quantum Mechanics and the H Atom
diagram Fig. 2.1: position and momentum cannot be measured simultaneously. They
may only be determined with an accuracy in accordance with the uncertainty rela-
tion pi xi h/2 . Quantum mechanics only makes statements about probability
amplitudes (r, t). These may assume the form of so called Wave-packets which
localize particles in space and time as well as the uncertainty relation allows it. One
finds a particle at position r and time t with the probability
2
w(r, t) = (r, t) . (2.1)
This is the key assumption of the statistical interpretation of quantum mechanics

as formulated by Max B ORN (1927) in the early days of quantum mechanics, for
which he received the N OBEL prize in physics in 1954. Quantum mechanics allows
to describe the evolution of a wave packet quantitatively. If formed at time t = 0
with widths in momentum and space, pi (0) and xi (0), respectively, one finds for
larger times t > 0 always pi (t) xi (t) > pi (0) xi (0): the wave packed diverges
as indicated in Fig. 2.1.
2.1.2 Probability Amplitudes in Optics
One may easily visualize the concept of a probability amplitude for the example
of photons in a YOUNGs double slit experiment. The probability to find a photon
at position r and time t is proportional to the intensity I (r, t) of the light, and the
latter is proportional to the square of the field amplitude. Let us consider only one
component of polarization, say Ex . The dependence of the intensity on position may
then be written as
2 2
I (r) Ex (r) = (r) = w(r).
The last two equalities can help us to get used to quantum mechanical terminology:
we call the quantity (r) now the (position dependent) probability amplitude or
wave function. In the case of light, it is simply represented by the electric field
component Ex . One determines it according to the laws of optics as solution of the
corresponding wave equation
(r) + k 2 (r) = 0 (2.2)
Fig. 2.1 Classical trajectory p

(red line) and quantum
mechanical probability (grey
shaded) in phase space. Note: ics
also at the beginning of a echan
mm
quantum mechanical ntu
trajectory, i.e. a qua
wave-packet, its position and
momentum are not precisely
defined corresponding to classical
the uncertainty relation x
2.1 Matter Waves 89
(with = 2 /x 2 + 2 /y 2 + 2 /z2 ) for given boundary conditions. For optical

applications a variety of approximations may be used such as H UYGENS -F RESNEL
principle or the diffractions theory of K IRCHHOFF. As the wave function is derived
from a linear partial differential equation (PDE), we may use the linear superposition
principle to describe the interference of waves. For diffraction by two slits we thus
have
(r) = 1 (r) + 2 (r), (2.3)
if 1 and 2 each describe the wave from one or the other slit. With these ampli-
tudes, the probability to find an atom at the detector positions r is given by:
2 2
w = (r) = 1 (r) + 2 (r) .
This expression contains interference terms 1 2 and is not simply a superposition

of probabilities but rather (we omit here the r):

w = |1 |2 + |2 |2 + 2 Re 1 2 . (2.4)
We may see this expression as a reinterpretation of the classical YOUNG double slit
experiment in terms of quantum mechanical probability amplitudes.
In an actual experiment one may now reduce the light intensity such that only
one single photon at a time is near the double slits and participates to the observed
interference pattern. One may easily verify such a setup with a particle counter
detecting single photons one may even hear the individual photons click.1 If a
sufficient number of such single photon events are recorded and added up, one finds
quite counter intuitively that the classical diffraction pattern known from optics
is recovered! Thus, the probability distribution of each individual photon behind the
double slit is determined by the wave amplitudes = 1 + 2 and it is impossible
to say through which slit the particle has passed (see Fig. 1.31). One also says that
one photon always interferes only with itself. We shall discuss this statement in some
detail in Chap. 2, Vol. 2 and quantify it statistically.
2.1.3 Probability Amplitudes and Matter Waves
From the perspective just discussed let us have a look at the matter waves intro-
duced in Sect. 1.7.1. We have already familiarized ourselves with the relation be-
tween momentum and wavelength as postulated by DE B ROGLIE. Matter waves are
1 For the experts: One may invest a lot of thought and substantial experimental efforts to make
absolutely sure that only isolated photons reach the double slit setup (for a recent review on single
photon sources see E ISAMAN et al. 2011). In the present context we shall be content if the average
time tav between each interference event (photon counted) is long compared to the coherence time
of the photon source c = 1/ tav , with being the bandwidth of the source (see Sect. 2.1.4,
Vol. 2).
also characterized by an amplitude (r), determining the probability for finding a

particle at position r in a volume element d3 r:
2
dw(r) = w(r)d3 r = (r) d3 r. (2.5)
And as in optics, diffraction and interference occurs, e.g. at a double slit for which
(2.4) describes the experimentally observable signal of particles.
However, in contrast to electromagnetic radiation, where was identified with
the electric (or magnetic) field, for matter waves (r) cannot be attributed to any
directly measurable physical quantity: the observable physics is described by the
probability w(r). So we call (r) simply the probability amplitude for finding a
particle at position r.
Otherwise, analogue considerations are valid for photons and particles of matter.
If e.g. we try to localize them on their pathways, we loose the interference patterns!
One important general rule holds: interference phenomena are observed if differ-
ent but indistinguishable pathways exist on which the particles may proceed toward
the detector. In contrast, no interference is observed if the two pathways may in
principle be distinguished even when no distinction is made in the actual experi-
ment.
Section summary
In classical mechanics well defined trajectories in position r(t) and mo-
mentum p(t) space describe the motion of particles. In contrast, quantum
mechanics describes probability amplitudes (r) and probabilities dw(r) =
|(r)|2 d3 r for finding a particle at a certain position r in a volume element
d3 r.
Diffraction and interference of matter waves can be described by concepts in
analogy to those used in wave optics. The big difference is that for photons
the probability amplitude is also a directly measurable quantity (electric or
magnetic field) while for matter waves (r) cannot be measured directly
only probabilities |(r)|2 refer to the real world.
2.2 SCHRDINGER Equation
Wave equations for matter waves were derived independently by Erwin

S CHRDINGER and PAUL D IRAC who in 1933 jointly received the N OBEL prize in
physics for their ground breaking work. The nonrelativistic S CHRDINGER equa-
tion is most commonly used in AMO physics and will briefly be introduced here.
As we shall see in Chap. 6, the relativistically correct D IRAC equation, valid for
fermions, is significantly more complex and describes inherently also the elec-
tron spin. Its key results may, however, also be introduced as perturbations into
the S CHRDINGER equation which then leads to sufficiently accurate results in the
non relativistic energy regime.
2.2 SCHRDINGER Equation 91
2.2.1 Stationary SCHRDINGER Equation
In contrast to photons, particles of matter may be exposed to external forces which

change their momentum and hence their wavelength. One may start from the wave
equation (2.2) and try to guess, how to modify it for matter waves. Consider-
ing a particle of mass m with a total energy W moving in a conservative poten-
tial V (r), we simply use the (nonrelativistic) energy conservation law of classi-
cal mechanics W = Wkin + V to determine its momentum from the kinetic en-
ergy Wkin :

p 2 = 2mWkin = 2m W V (r) .
With W being a constant of motion we derive from this and (1.118) the absolute
value of the wave vector k = p/ and insert it into (2.2):
p2 2m(W V (r))
(r) + (r) = (r) + (r) = 0.
2 2
Rewritten, this is already the stationary S CHRDINGER equation
2
(r) + V (r)(r) = W (r), (2.6)
2m
or somewhat more compact
(r) = W (r)
H (2.7)

with the eigenfunction (r) and the eigenenergy W . The H AMILTON operator H
represents the total energy (briefly just Hamiltonian):
= + V (r).
2
H (2.8)
2m
In the often encountered one-dimensional case the S CHRDINGER equation (2.6) is
further simplified:
2 d2 (x)
+ V (x)(x) = W (x). (2.9)
2m dx 2
2.2.2 HAMILTON and Momentum Operators
We may write the H AMILTON operator (2.8) even more suggestive:
= 2 + V (r) =
p2
2
H + V (r). (2.10)
2m 2m
Here we have introduced the momentum operator (a vector operator)

/x
p = i = i /y
(2.11)
/z
so that

2 2 2

p2 =
p
p = 2 2 = 2 = 2 + + . (2.12)
x 2 y 2 z2
Thus, (2.10) is the operator form of the classical energy conservation law
p2
W = Wkin + V = + V (r).
2m
2.2.3 Time Dependent SCHRDINGER Equation
So far we have discussed probability amplitudes only as a function of r, the par-

ticles position in space. For many problems in AMO physics such a description
of quantum systems by stationary states is fully sufficient. However, often time de-
pendence is equally important. In respect of photons we know, that electromagnetic
waves follow a time dependent wave equation, which is a 2nd order partial differ-
ential equation (PDE) in space and time derived from the M AXWELL equations. In
contrast, for matter waves the time dependent S CHRDINGER equation holds:
(r, t) = i (r, t)
H (2.13)
t
2 (r, t)
or more explicitly: (r, t) + V (r) (r, t) = i .
2m t
It cannot be derived, and we just communicate it here as it was found by Erwin
S CHRDINGER in the beginning of 1926 by the way: during a winter ski holiday
in the Swiss Alps. We point out some key aspects:
The S CHRDINGER equation is a linear PDE of 2nd order in space and 1st order
in time! As a consequence, time dependence of stationary states is truly complex.
The linear superposition principle may be applied to the solutions.
The statistical interpretation of quantum mechanics understands solutions
(r, t) of this PDE as probability amplitudes for finding a particle at a posi-
tion r in space at time t according to (2.1).
Although the S CHRDINGER equation cannot be derived, it is able to describe
a wealth of atomic phenomena and to predict observables quantitatively (under
nonrelativistic conditions) with excellent accuracy: we emphasize that is this very
fact which defines the validity of a physical theory.
Equivalently, formal (algebraic) quantum mechanics can be deduced from a con-
sistent set of axioms, which are, however, also heuristically assumed.
2.2 SCHRDINGER Equation 93
Consistent alternatives of wave equations for matter waves are the D IRAC equa-
tion for fermions, a multi component spinor equation and the K LEIN -G ORDON
equation, a single component PDE of 2nd order in time, which turns out to be
valid for bosons both are relativistically correct.
(r, t) = H
If the Hamiltonian itself does not depend on time, H (r), the time depen-
dence of the wave function may be factored by a product ansatz:
(r, t) = (r)(t) (2.14)
(r, t) = i (r, t)
H (r)(t) = i (r)(t)
H
t t

H (r) i (t)
= W.
(r) (t) t
The last identity (with the constant W which has to be determined) must hold so
that the former equality can be valid for all values of r and t. One thus has to
solve id(t)/dt = W (t) and H (r) = W (r). In this case, the time dependent
equation leads to the trivial solution

W
(t) exp i t . (2.15)

The position dependent part is nothing but the stationary S CHRDINGER equation
(2.6) and the parameter W introduced is the total energy of the system. The overall
wave function is given by

W
(r, t) = (r) exp i t . (2.16)

Note: The time dependence is truly complex and the imaginary unit i in the prefactor
= H
is needed for the solution! In the present case, however, with H (t) the time
dependence is trivial in the sense that only
2 2
w(r, t) = (r, t) = (r) (2.17)
can be measured, which contains only information about the stationary state.
In order to find these stationary atomic states for a given potential V (r)
which are the equivalent to B OHRs stationary orbits one has to solve the station-
ary S CHRDINGER equation (2.6) under suitably chosen boundary conditions. For
bound states this typically leads to a whole series of discrete total energies W < 0
with a set of quantum numbers characterizing the states. The corresponding series
of wave functions for individual electrons called orbitals describe the probability
of finding the particles at a given position in space.
2.2.4 Freely Moving Particle The Most Simple Example

As the most simple example we consider now a freely moving particle with mass m,
(kinetic) energy W and momentum p. The relevant stationary S CHRDINGER equa-
tion (2.6) reads
2
(r) = W (r) and has solutions (r) = C exp(ikr). (2.18)
2m
As one verifies by inserting the solution with wave vector k = p/ into the equation,
the total energy is given by W = 2 k 2 /(2m) = p 2 /(2m) = Wkin . With the time
dependence (2.16) the whole probability amplitude for this free particle is simply a
plane wave:
2
k
(r, t) = C exp i(t kr) = C exp i t kr (2.19)
2m

W pr
= C exp i t .

With this we obtain the important dispersion relation for free matter waves:
W p2 k 2 2 2
= = = = (k) or W = k . (2.20)
2m 2m 2m
Note: The probability to find this particle, w(r, t) = | (r, t)|2 = |C|2 , is indepen-
dent of position and time as expected for an infinitely extended plane wave. In
other words, a particle with well defined momentum may not be localized at all as
expected according to the uncertainty relation (1.123).
Section summary
The stationary S CHRDINGER equation (2.6) may be gleaned from the classi-
cal wave equation combined with a free interpretation of the DE B ROGLIE
wavelength in a conservative potential. It is validated by the excellent
agreement of its predictions with experimentally observed data in the sub-
microscopic world at nonrelativistic energies.
In its most compact form it reads H = W , where the H AMILTON operator
(2.10), also called Hamiltonian, is constructed in full analogy to its classical
counterpart, just replacing the momentum by the quantum mechanical mo-
mentum operator p = i.
The time dependence of the wave function is described by the time dependent
S CHRDINGER equation (2.13). In the case that the Hamiltonian itself is not
time dependent, the (stationary) solutions are given by the product of the solu-
tions (r) of the stationary equation (2.6) and a simple exponential function
exp(i(W/)t).
The most simple solution of the time dependent S CHRDINGER equation is a
plane wave exp[i((W/)t kr)], with the energy W = k 2 /(2m).
2.3 Basics and Denitions of Quantum Mechanics 95
2.3 Basics and Denitions of Quantum Mechanics
2.3.1 Axioms, Terminology and Rules
Here we summarize briefly the fundamental axioms of quantum mechanics and re-
call some terminology and rules which we shall use later on:
Quantum States and Wave Functions

States of quantum systems (in the world, in atomic physics . . . ) are described by
state vectors for which we introduce here the bra and ket notation of D IRAC,
| and | , respectively. Typically, states may be expressed in terms of basis states,
or basis vectors, say |f1 , |f2 , |f3 , . . . , |fn , . . . . We speak about a complete basis
set, if each state of a system can be written as ket vector

| = ci |fi , alternatively as bra vector | = ci fi |. (2.21)
i=1 i=1
Note: the sum includes all bound (discrete) and free (continuum) states.
One calls a basis orthonormal if
fk |fi = ki , (2.22)
where | is the scalar product of two state vectors | and | . With this, the
projection of the state | onto the basis vector |fk is

fk | = ci fi |fk = ck . (2.23)
i=1
Wave functions, the most commonly used representation of states, are formally
obtained by expanding | in a continuous position basis {|r } where r extends
over all points in 3D position space. We write (2.21) as ket

| = d r r r
3
or as bra | = d 3 r r r . (2.24)
Per definition we have, in analogy to (2.22),

r r = r r , (2.25)
and obtain from (2.24) the definition of a wave function in position space,

r| = |r = d3 r r r r = (r), (2.26)
and (r) = |r . In practice, wave functions are determined by solving the sta-
tionary S CHRDINGER equation. With (2.24) and (2.25) we derive (after one 3D
integration) the scalar product in terms of wave functions:2

| = (r)(r)d3 r = | . (2.27)
For the eigenstates |k of the S CHRDINGER equation the orthonormality relation

(2.22) thus reads:

i |k = i (r)k (r)d3 r = ik . (2.28)
We finally note that, equivalent to the state expansion in position space, one may
define wave functions (p) in momentum space by

| = d3 p p p and

p| = d3 p p pp = (p).
Operators
Linear operators play a key role in quantum mechanics: in general an operator
changes an object onto which it acts (e.g. a state vector, a wave function, another
and let it act on a ket
operator) into something different. Let us call the operator A
vector | :

A|
= | .
Linearity implies that for a superposition of states | = c1 |1 + c2 |2 +

c1 |1 + c2 |2 + = c1 A|
A 1 + c2 A|
2 + .
B
The product of two operators A is defined by

B)|
(A B|
=A . (2.29)
B
Operator multiplication is distributive, i.e. A C
= (A
B)
C = A(
B C)
but not neces-

sarily commutative. In general AB = B A and one defines a commutator
B]
[A, =A
B B
A (2.30)
2 We shall use the bra| and |ket symbols rather loosely. In particular, we shall often write wave
functions simply as | , |k , etc.

which only in special cases may vanish (see Sect. 2.3.3). With | and | being
two state vectors or wave functions one defines as matrix element of A between

states | and | (which may or may not be basis states of an operator A:

= (A)d
A = |A 3 r. (2.31)
Without going into details we define the so called adjoint (or Hermitian conjugate)
by
operator A

= |A
A or A = |A (2.32)

3
or d r = (A)d
A 3 r. (2.33)

Of particular importance are the so called Hermitian operators, let us call them O.
They are self-adjoint:
O
O. (2.34)
By this definition and with (2.31)(2.33) the matrix elements

O| = (O)
= |O d3 r = (O)d 3r (2.35)
= |O
= |O (2.36)
are all equivalent for Hermitian operators.3
Observables
Observables are all physical quantities which can in principle be measured (ob-
served). Every quantity which is observable in classical physics is represented quan-
tum mechanically by a linear Hermitian operator, let us call it O. A quantum sys-
tems can be characterized by a set of eigenstates (eigenvectors) |fk of an observable
O which it may in principle assume. From the eigenvalue equation
k = k |fk
O|f (2.37)
one determines the eigenvalues k of the observable O for the eigenvectors |fk . In
any individual physical measurement of the observable O only one of its eigenvalues
k can be observed.
From (2.37) and the orthonormality of the |fk basis one sees that the matrix
in the basis of its eigenstates between |fi and |fk are
elements of O
k = k ik ,
Oik = fi |Of (2.38)
3 To verify these relations one simply expands | and | in a basis of eigenvectors (eigenfunc-

tions) of O.
ik is diagonal and because of (2.36) the eigenvalues k must be real

i.e. the matrix O
as one has to demand for measurable quantities!
Superposition and Expectation Values

In general, a quantum systems to be investigated may be found in a state | which
Let the state | be a linear superposition of
is not an eigenstate of the operator O.

eigenstates |fi of the operator O with eigenvalues i according to (2.37):

| = ci |fi with the expansion coefficients ci = fi | . (2.39)
i
The latter relation follows directly with (2.22). If we measure now the observable
many times (as one does in a real experiment), the result of each individual mea-
O
The probability to detect this particular
surement is one of the eigenvalues i of O.
eigenvalue i is determined by the probability amplitude ci :

O|
=O ci |fi = ci i |fi . (2.40)
The average value measured for this observable, i.e. the result of many individ-
ual measurements applied to the same state | , is called expectation value of the
observable:

O
|ci |2 i = |O . (2.41)
The latter identity follows from

|O| = ci fi O ck f k = ci f i k
ck Of
i k

= k ci ck fi |fk = k ci ck ik = i |ci |2 .
i k i k i
()
Somewhat more general, if the state of the system is | = i ci |gi , i.e. it is
its expectation
given in an arbitrary basis {|gi } and |gi are not eigenstates of O,
value is
()
()
= |O|
O = ci
gi |O|
()
ck |gk = ci k c() . (2.42)
gi |O|g k
i k ik
Unit Operator
We note in passing a nice mathematical trick by rewriting (2.39) as

| = |fi ci = |fi fi | = |fi fi | | .
i i i
From this we deduce the often very useful fact that the quantum mechanical unit
operator4 (which does not change the operand) may be written as

1= |fi fi |, (2.43)
i
as long as the states |fi represent a complete orthogonal basis and the summation
goes over all states of this basis.
Quantization
Experimentally, when an observable O is determined in an individual measurement
one always finds one of its eigenvalues. With the measurement one prepares also
One
the corresponding eigenfunction (eigenstate, eigenvector) of the observable O.
may say that by a measurement one projects the eigenvector out of the state |
under investigation.
Example: HAMILTON Operator

The eigenvalues of the H AMILTON operator H are the eigenenergies Wn of a sys-
tem. Thus, the Hamiltonian is a particularly important example of an observable. Its
eigenvectors |n and eigenenergies are determined by solving the S CHRDINGER
equation (2.7). We may also write it in algebraic form:
|n = Wn |n .
H
Example: Spin Projection onto the z-Axis

As a further example we mention the projection of the spin of a particle onto a given
axis in 3D space. Typically, but not necessarily, one measures the
Sz component of
the spin. We have already discussed such a measurement in Sect. 1.10, the S TERN -
G ERLACH experiment. The eigenvalues are here ms and we write the eigenstates
quite formally as |sms . The eigenvalue equation is then

Sz |sms = ms |sms .
2.3.2 Representations
We have already made use of different ways to describe quantum systems and their
changes. The standard terminology speaks about representations of states and oper-
ators.
4 In matrix representation this corresponds to the identity matrix

1 0 0
0 1 0

1=. . . .
.. .. . . ...
0 0 1
SCHRDINGER Representation
In the S CHRDINGER representation (or picture) the operators are differential oper-
ators. The states are represented by wave functions. The scalar product is an integral
according to (2.27) and the orthogonality of basis states is defined by (2.28). The
matrix elements of an operator A in respect of an arbitrary basis set |fk , |fi , . . .
are
(in general not eigenfunctions of A)

k = fi (r)Af
Aik fi |Af k (r)d3 r = fk A
fi = A , (2.44)
ki
where we have used the definition of adjoint operators (2.32). If the operator is
Hermitian, i.e. represents and observable, this relation reads

Oik fi |Ofk = fi (r) Of k (r) d3 r (2.45)

= k (r) fi (r)d3 r
Of i = Oki
= fk |Of
. (2.46)
HEISENBERG Representation
In the H EISENBERG representation (or picture) the operators (we mention in this
context the N OBEL prize for H EISENBERG 1932) A are matrices, which are de-
termined by their matrix elements Aik . The states, say | or | , are vectors in
H ILBERT space5 which we write
| = = b1 f 1 + b2 f 2 + b3 f 3 +
| = = c1 f 1 + c2 f 2 + c3 f 3 +
with their components bi f i or ci f i , respectively. The scalar product is here

| = bi ci . (2.47)
Both representations are physically as well as mathematically fully equivalent.
2.3.3 Simultaneous Measurement of Two Observables
The following is a generalization of the H EISENBERG uncertainty relation which

states that two canonically conjugate coordinates, such as position and momentum,
cannot be measured simultaneously.
Two observables A and B can be measured simultaneously if and only if the
eigenstates of A are also eigenstates of B,
i.e. if the following holds:
i = i |i and B|
A| i = i |i .
5A H ILBERT space is an extension of the 3D vector space to an arbitrary or infinite number of

dimensions in quantum mechanics to an infinite dimensional function space.
Thus, also the following relations must be valid for simultaneous measurability of
and B:
the operators A
B|
A i = A
i |i = i A| A|
i = i i |i = B i .
B
Hence, the operators must commute, A =! B
A.
Equivalently:
and B
Simultaneous measurement of two observables A is possible if and only
if their commutator vanishes:
B
A B
A = [A,
B]
= 0. (2.48)
2.3.4 Operators for Space, Momentum and Energy
From the above follow some simple recipes of how to translate classical quantities
into quantum mechanical operators. One simply has to substitute:

r r and pi i =p
i
xi
(2.49)

or p i , i , i = i =
p.
x y z
All other substitutions are derived from these rules. In particular, the classical
Hamiltonian total energy
p2
Hclass = + V (r) = Wkin + V with p 2 = p p
2m
becomes the H AMILTON operator:
= 1 (i) (i) + V (r) = + V (r).

2
H
2m 2m
x are the prime
Position in space x and its canonically conjugate momentum p
example for non-commuting observables. As seen from comparing

x x(x) = i
p x(x) = i x (x) + (x)
x x

x (x) = ix
with x p (x)
x
the commuted operators p x x and x p
x generate completely different results and
hence, the observables x and px cannot be measured simultaneously. This is, so to
say, a formal endorsement of the H EISENBERG uncertainty relation.
2.3.5 Eigenfunctions of the Momentum Operator

p
We shall now find the eigenfunctions and eigenvalues of the momentum in the
S CHRDINGER picture, starting with the one dimensional case:
d(x)
x (x) = px (x)
p i = px (x).
dx
One easily verifies that (x) = exp(ipx x/) = exp(ikx x) are solutions of this eigen-
value problem. Each value of px (with < px < ) is an eigenvalue of the mo-
mentum operator p x in x-direction: the eigenfunction is thus a plane wave with a
continuum of eigenvalues.
This is easily extended into 3D space. The eigenvalue equation

p (r) = i(r) = k(r) is solved by
a plane wave (r) = C exp(ik r).
The unit directional vector for an arbitrary direction in space is e = ax ex +ay ey +

az ez with ax2 + ay2 + az2 = 1. If we are interested in the magnitude of the momentum
p in any given direction e, the operator providing this information is:

e = e
p p = ax p
x + ay p y + az p
z = i ax + ay + az . (2.50)
x y y
Any plane wave exp(ik r) in arbitrary direction k and with arbitrary magnitude k
e , with eigenvalue p cos , as one easily verifies
is an eigenfunction of p
e exp(ik r) = (ax kx + ay ky + az kz ) exp(ik r) = e k exp(ik r)

p
= k cos exp(ik r) = p cos exp(ik r), (2.51)
with p = k and the angle between e and k just as one would guess.
Section summary
In the S CHRDINGER picture, states (bra | and ket |) of quantum sys-
tems are represented by wave functions (r) and (r), respectively. The
H EISENBERG picture uses state vectors in H ILBERT space.
Quantum states may be expressed as a linear superposition of states (2.21)
from a complete, orthonormal basis with fi |fk = ik . The unit operator
may be written i |fi fi |.
The S CHRDINGER representation uses differential operators. Classical the-
ory is translated by replacing r r and p p = i.
Matrix elements of an operator A k = f (r)Af
are Aik fi |Af k (r)d3 r
i
in the S CHRDINGER picture. In the Heisenberg representation operators are
defined by the corresponding matrices.
The adjoint A of an operator A is defined by A
| = |A .
Hermitian

operators are self-adjoint A = A, and Aik = Aki holds.

2.4 Particles in a Box And the Free Electron Gas 103
Observables are represented by Hermitian operators. With O|f k = k |fk

their eigenvalues k are the only values of that observable which can be ob-
served experimentally.
Such an experiment projects the eigenstate |fk of the observable out of a
and B,
state | under investigation. If the operators for two observables, A
commute (commutator [A B]
= 0) they can be measured simultaneously.
The average value of an observable in a state | is called its expectation
= |O
value: O = (O)d 3 r.
Eigenstates of the momentum operator p are plane waves exp(ik r), with
p = k.
2.4 Particles in a Box And the Free Electron Gas

2.4.1 One Dimensional Potential Box
We first consider the one dimensional problem of a particle in a potential box, i.e.
between two infinitively high walls at a distance L. In between the walls, the 1D
S CHRDINGER equation (2.9) simply reads
2 d2 n (x)
= Wn n (x).
2m dx 2
Solutions can in principle be (x) = sin(kx) or cos(kx). However, since the wave
function cannot penetrate into the wall, it must vanish on the walls which at x = 0 is
only possible for the sin(kx) solution. For continuity on the other wall sin(kL) = 0
and thus k = n/L must hold, with n = 1, 2, 3, . . . . In summary, the solutions
(eigenfunctions) are standing waves with nodes on both walls of the box. The
eigenenergies assume discrete values Wn :

2 nx 2 k 2 h2 n2
n (x) = sin with Wn = = . (2.52)
L L 2m 8mL2
The expectation value of the momentum p x is derived from

x n = n (x)
px = n p px n (x)dx

2 nxL d sin nx
= i
sin L
dx
L0 L dx

i2n L nx nx
= sin cos dx 0.
L2 0 L L
This corresponds to the fact that the particle in the box moves with equal probability
back and forth. In contrast, the square of the momentum,

2 2 L nx 2 d2 sin nx
x = n p
p x2 n = n p x2 n dx = sin L
dx
L 0 L dx 2

2(n)2 L nx 1 h2 2
= sin2 dx = n ,
L3 0 L 4 L2
=
is not zero. With this and H px2 /2m we verify Wn in (2.52).
2.4.2 Three Dimensional Potential Box
A next step towards reality is the extension into the 3D space. The movement of a
particle is now restricted to a large but finite 3-dimensional box, with rigid walls
beyond which the probability to find a particle is zero. For simplicity we assume
the box to be a cube with edge length L as illustrated in Fig. 2.2(a). Inside the box
particles move freely. Stationary solutions are plane waves (2.18) which we write in
the box as real functions now as product in three dimensions:
(x, y, z) = sin(kx x) sin(ky y) sin(kz z). (2.53)
To be continuous, the wave function must vanish on the walls of the cube:

sin(kj L) = 0 kj = n j for j = x, y, z. (2.54)
L
With these boundary conditions, and in analogy to the 1D case (2.52),
the energy of the particle in a box becomes
2 2 2 k 2 2 2 2
W= kx + ky2 + kz2 = = n , (2.55)
2m 2m 2mL2
now with three integer quantum numbers nx , ny , nz and n2 = n2x + n2y + n2z .
One may view this in k or n space as indicated in Fig. 2.2(b). Equation (2.54) says
that precisely one solution exists for each lattice point with integer nx , ny and nz .
Thus, one reads from the figure that the total number of states with quantum num-
bers 1 to nx , 1 to ny and 1 to nz , i.e. the number of states with n n2x + n2y + n2z is
given by NZ (n) = 1/8 4/3 n3 . Expressing n by the energy W , the total number
of states with energies W is
1 (2mW )3/2 3
NZ (W ) = L .
6 2 3
Fig. 2.2 Boundary condition

(a) z (b) n z kz
for a particle in a potential
box (cube of edge length L); L
(a) the nodes of the wave n
L
function are on the walls of y
the box in 3D position space; L ny ky
(b) one counts the number of
states n in the n or k space x nx kx
2.4 Particles in a Box And the Free Electron Gas 105
If the particle has a spin s, we also have to account for the energy degenerate ge =
2s + 1 possible orientations in space. Dividing by the volume of the box L3 , we
obtain the total number of available states per unit volume:
ge (2mW )3/2
NZ (W ) = . (2.56)
6 2 3
From this we derive the number of states in an energy interval from W to W + dW ,
called density of states (DOS), here per unit volume:

dNZ (W ) ge (2m)3/2 4 2m3/2
g(W ) = = W = ge W. (2.57)
dW 4 2 3 h3
We note in passing, that assuming the phase space to be quantized in unit cells of
size h3 leads to exactly the same result.6 For later use we also give the density of
states in respect of a specific element of solid angle d and express the energy by k,
the magnitude of the wave vector:
d2 NZ (W ) d ge mk
dg = = d. (2.58)
dW 4 (2)3 2
2.4.3 The Free Electron Gas
An important application are electrons (mass me , s = 1/2, ge = 2). In the context

of the photoelectric effect (Sect. 1.4.1) we have already introduced the free electron
gas model, where one assumes the electrons in a metal to move freely in an electron
sea. Now, as a 1st order approximation, we describe this situation quantitatively by
electrons in a 3D box. It turns out that such a model serves well in many areas of
physics, as a first step e.g. when introducing electron bands in solid state physics,
but also in atomic physics for statistical models of electrons in a large atom (see
Sect. 10.1.5).
Due to the high particle density Ne (electrons per unit volume), F ERMI -D IRAC
statistics of electrons in the bulk is quite different from that in gases treated in
Sect. 1.3.4. We estimate Ne = NA /Mr g mol1 , with the number of valence
electrons per atom, NA the AVOGADRO number, the (mass) density of the mate-
rial and Mr the relative atomic mass. It is typically on the order of 1028 1029 m3 .
In contrast to the situation treated in gases, the chemical-potential is now pos-
itive and kB T . At absolute zero temperature, T = 0 (or rather at very low
values of T ), each available state (with any given spatial wave function according to
Eqs. (2.53) and (2.54)) will be filled by two electrons (with spin up and spin down,
ge = 2). The potential box will thus be filled up to an energy F . With (2.56) the
6 Thesize of phase space with momenta up to p is (4/3)p 3 L3 . Expressing p in terms of kinetic

energy W , and dividing by h3 and L3 gives the number of phase space cells per unit volume:
NZ = (4/3)(2mW )3/2 / h3 . Differentiation in respect of W yields (2.57).
w(W ) / eV -1 (a) W
w (W ) / eV -1 (b)
W
0.2 0.2
T = 0K
T = 50 K
0.1 0.1 T = 500K
Ne Ne
T = 293 K
0 0
0 2 4 6 8 6.8 6.9 7.0 7.1 7.2
F W / eV F
Fig. 2.3 F ERMI -D IRAC probability distribution according to (2.61) as a function of energy for
electrons in a metal with a F ERMI energy F = 7 eV at different temperatures T : (a) energy range
from 0 eV to 9 eV, (b) expanded scale around F ; at T = 0 the electron density Ne extends up to
F (grey shading up to the dashed vertical red line); the full red line indicates the density of states
W
total number of electrons per volume with kinetic energies between 0 and F is
thus

1 2me F 3/2
Ne = 2NZ (F ) = . (2.59)
3 2 2
We may invert this to obtain the so called F ERMI energy F . At absolute zero tem-
perature F is identical to the chemical-potential introduced in Sect. 1.3.4, and
corresponds to the maximum electron energy. One state after the other is filled, each
by two of electrons, up to the F ERMI energy:
2 2 2/3
F = 3 Ne . (2.60)
2me
Typical F ERMI energies for metals range from 1.6 eV (Cs) to 14.3 eV (Be).
Expressing the prefactor A given by (1.64) with ge = 2 in terms of the F ERMI
energy F one may write the probability distribution for finding a given energy W
in the electron gas in a box:

dNe 3 1 3/2 W
w(W )dw = = dW (2.61)
Ne 2 F exp[(W )/(kB T )] + 1
with F F as long as kB T F .
This is illustrated in Fig. 2.3 for a characteristic example with F =7 eV (about

the value for Cu). The full red line indicates the density of states W accord-
ing (2.57). The grey shaded area under this curve indicates for T = 0 K the filling
with electrons up to F (dashed vertical red line). For temperatures T > 0 K the
states will be occupied according to the F ERMI -D IRAC statistics (1.66). As clearly
seen on an enlarged energy scale in Fig. 2.3(b) the boundary between occupied and
2.5 Angular Momentum 107
unoccupied states broadens as temperature increases (width 2kB T ) the reduc-

tion of the probability for W < F is approximately compensated by the increase
at W > F , and the F ERMI distribution remains normalized, i.e. 0 w(W )dw 1.
For metals at room temperature one typically finds kB T /F < 1/100 and the nor-
malization condition holds very well.
However, at higher temperatures, kB T /F 0.1, or for different densities of
states (e.g. in semiconductors), the chemical-potential must be readjusted in or-
der to maintain the normalization of w(W ). In solid state physics it is often called
F ERMI level WF = , which, strictly speaking, is identical to the F ERMI energy F
only at T = 0.
Section summary
The particle in a box model allows a most simple description of electrons
moving freely in a metal. The wave functions (2.53) in the 3D box (vol-
ume L3 ) have nodes at theboundaries of the box. The energies are W =
2 2 n2 /(2mL2 ) with n = n2x + n2y + n2z and ni representing positive val-
ues in 3D integer number space.
From this the DOS (2.57) is derived, which according to (2.57) is
ge m3/2 W , where m is the particles mass and ge = 2s + 1 the degen-
eracy due to its spin.
In the case of electrons (Fermions) each state can be filled with up to 2 elec-
trons. Then, at T = 0 K the highest energy with occupied states is the F ERMI
energy F = 2 (3 2 Ne )2/3 /(2me ).
At temperatures T > 0 the boundary between occupied and unoccupied states
smears out according to (2.61). The width of the boundary layer is on the
order of kB T .
2.5 Angular Momentum
Angular momenta play a central role in atomic and molecular physics and Ap-
pendix B gives a summary on the essentials: their abstract definitions, properties,
combinations and the relevant algebra. Here we introduce the S CHRDINGER pic-
ture of orbital angular momenta as used in the quantum mechanics of the H atom.
The orbital picture emerging from this concept is directly accessible to physical
imagination and visualization. At the end of this section we shall generalize the
basic concepts, including electron spin.
2.5.1 Polar Coordinates
Quantum mechanical problems may be treated with advantage in polar rather than in
Cartesian coordinates if they have a symmetry centre, e.g. if the potential depends
only on the distance r from origin V (r) = V (r) as in the C OULOMB case. The
Fig. 2.4 Cartesian (x, y, z) z

and polar coordinates (r, , )
r

y

x
transformations between Cartesian (x, y, z) and polar coordinates (r, , ) are read
from Fig. 2.4,
x = r sin cos
y = r sin sin (2.62)
z = r cos ,
and the volume element transforms as
dxdydz r 2 sin d ddr. (2.63)
Conversion of the relevant observables such as the kinetic energy

p2 2 2 2 2 2 2
= = + +
2m 2m 2m x 2 y 2 z2
requires some in principle trivial partial differentiations and mathematical rear-
rangements which we shall not explicate here. The result is
2

p2 2 2 2 1 2 L
= = r + with (2.64)
2m 2m 2m r 2 r r 2mr 2

2 1 1 2
L = 2
sin + 2 . (2.65)
sin sin 2
In full analogy to classical mechanics, the form of (2.64) suggests two components
of kinetic energy:

2 1 2
radial energy Hr = r and (2.66)
2m r 2 r r
2
rotational energy rot = L .
H (2.67)
2mr 2
This suggestive way of writing the energy already implies that

2
L might be the
quantum mechanical equivalent to the classical angular momentum. In the follow-
ing, a more formal approach to derive an expression for the angular momentum will
be sketched.
2.5.2 Denition of Orbital Angular Momentum
In classical mechanics angular momentum is defined as L = r p. The quantum

mechanical equivalent is the orbital7 angular momentum operator
L which accord-
ing to the recipe (2.49) is obtained by substituting p
p:

L = r
p. (2.68)
This has to be expressed in polar coordinates. We just show this for one coordinate
by way of example:

z = i x y .
L (2.69)
y x
With (2.62) one transforms the expression in brackets into polar coordinates:
x y z
= + + = r sin sin + r sin cos +0
x y z x y

= y +x =x y .
x y y x
Thus, the operator for the z component of orbital angular momentum is
z = i .
L (2.70)

Obviously this is constructed in complete analogy to the linear momentum p x =

i/x, with the pair of canonically conjugate coordinates ( px , x) replaced by
z , ).
(L
The transformation of Lx and L y is slightly more complicated. In summary one
finds, here without proof, that the operator of the squared orbital angular momen-
tum 2x + L
2y + L
2z is indeed given by (2.65). Some details are explained in
2
L =L
Appendix B.
2.5.3 Eigenvalues and Eigenfunctions
We are now ready to discuss eigenvalues and eigenstates of angular momentum op-
erators. They will be used in more or less all of the following chapters. In the spirit
of this textbook, again we only sketch the basic concepts, present a collection of nec-
essary tools, and refer the reader to the standard textbooks on quantum mechanics
and angular momentum algebra for details.
7 Orbital to distinguish it e.g. from the spin angular momentum.

Component L z of the Orbital Angular Momentum

In polar coordinates the z-axis represents a preferred direction, since the polar angle
is defined in respect of it. With (2.70), the eigenvalue equation of L z is an ODE
of 1st order
z () = z ()
L
(2.71)
i = z ()

and may be integrated directly. The solution is

z
= C exp i

with a normalization constant C. We now have to apply some physics: which values
of z are really meaningful? Obviously () has to be unique:

! ! z
(0) = (2) or equivalently exp(0) = exp i 2 . (2.72)

This is only possible for integer values of z / = m with m = 0, 1, 2, . . . . Then

z
exp i 2 = exp(im2) = 1,

and m () = Cm exp(im) are the eigenfunctions solving (2.71). We call m the
projection quantum number (in the literature often also somewhat misleadingly
called magnetic quantum number). These wave functions are orthonormal:
2
!

mm = m |m = Cm Cm exp(im) exp im d (2.73)
0

with Cm = 1/ 2.
The standard phase convention uses real normalization constants Cm . In summary,

for the projection of the angular momentum onto the z-axis
the eigenvalue equation is z m = mm ,

L
1
with eigenfunctions m = exp(im), (2.74)
2
and eigenvalues m where m = 0, 1, 2, . . . .
Components in x- and y-Direction

For the Lx and L
y components the calculation is more complicated but in principle
trivial. We communicate without proof: Lx , L
y , and L
z cannot be measured simul-
taneously. Each pair of them does not commute. On the contrary, one may show that
the following commutation rules for angular momenta hold:
x , L
[L y ] = L
x L y L
y L z ,
x = iL
(2.75)
[L z ] = iL
y , L x z , L
and [L x ] = iL
y .
i are represented by different functions of

Corresponding to this all components L
and (not shown here).
Square of the Orbital Angular Momentum

With (2.65) we write the eigenvalue equation for
2
L as
2
L Y (, ) = L2 Y (, ) (2.76)
with the eigenvalue L2 and use a the product ansatz towards its solution:
Y (, ) = ( )(). (2.77)
z and substitute ( )m ()
As azimuthal part we try the eigenfunctions (2.74) of L
into (2.65). This leads to

1 ( ) 1 2
2
sin m () + 2 ( ) 2 m ()
sin sin
= L2 ( )m () = L2 Y (, )

1 m2
2 sin 2 = L2 . (2.78)
sin sin
Thus, only one ODE remains to be solved. Several procedures lead to the cor-
rect solutions. One may directly use the associated L EGENDRE polynomials, known
from the mathematics of ODEs, or (perhaps more elegantly) by exploiting the prop-
erties of the angular momentum operators to find appropriate recursion formulas.
In any case, one has to demand physically reasonable boundary conditions. In anal-
ogy to (2.72) for the z-component, the wave functions must be finite and unique for
0 . Without proof we communicate here that such physically meaningful
solutions exist for which the following relations hold (for completeness we include
again the z-component):
eigenvalue equation for

2 2
L L Y m (, ) = ( + 1)2 Y m (, ) (2.79)
eigenvalues of
2
L L2 = ( + 1)2 (2.80)
eigenvalue equation for
Lz z Y m (, ) = m Y m (, )
L (2.81)
quantum numbers = 0, 1, 2, . . . and m = 0, 1, . . . ,
(2.82)
degeneracy 2 + 1 (2.83)
Fig. 2.5 Vector diagram z

illustrating the 2 + 1 Lz /
possible orientations of the
2
angular momentum in space L
(shown for = 2 as an
example) 1
L
0
y
-1
x -2
z only acts
The validity of (2.81) follows directly from (2.74) and (2.77), since L
onto the component of Y m (, ). This implies
2
2
L L z
z=L L
2
or z = 0.
L ,L (2.84)
Equivalently: z can be measured simultaneously. This also holds for

2 2
L and L L
x as well as for
and L
2
L and Ly not, however, for the components L
x , L
y and L
z
which according to (2.75) do not commute with each other.
Vector Diagram
With (2.79) one may write the magnitude of the angular momentum as:

|
L| = ( + 1). (2.85)
For a given set of quantum numbers m the exact direction of the total angular
momentum is undefined. Precisely defined is only the magnitude and component
Lz = m in respect of the z-axis. One visualizes these relations with the help of
a vector diagram shown in Fig. 2.5. ofthe electron spin (see Fig. 1.43). Figure 2.5
illustrates the example = 2, |
L|/ = 6 2.45 with L z / = m = 2, 1, 0, 1, 2.
One may consider the vector arrows statistically
distributed around the z-axis, i.e.
on cones of height m with a side length ( + 1).
Spherical Harmonics
The eigenfunctions of
2
L and Lz are called spherical harmonics Y m (, ). General
formulas and properties are summarized in Appendix B.1.2, and specific expres-
sions up to = 3 are tabulated in Table B.1. A graphical illustration of the angular
dependence is shown in Fig. 2.6.
The Y m (, ) are orthonormalized:
2

d Y m (, )Y m (, ) sin d = m m . (2.86)
0 0
x y
|Y00|2 |Y10|2 |Y11|2 |Y20|2 |Y21|2 |Y22|2
Fig. 2.6 3D-plot of the s, p and d spherical harmonics; plotted are the squared moduli as a func-
tion of angles, the shading colours indicate the sign of Y m (, 0). For an alternative representation
in the real basis see Appendix D.3 and in particular Fig. D.1
The complex conjugate is given by

Y m (, ) = (1)m Y m (, ), (2.87)
and inversion at the origin (r r) leads to
Y m ( , + ) = (1) Y m (, ), (2.88)
which describes (see detailed discussion in Appendix D) so called positive or nega-

tive parity depending on whether is even or odd, respectively.
At this point, we introduce an important, commonly used notation:
orbital angular momenta = 0, 1, 2, 3, 4, . . . are labelled s, p, d, f, g, . . . .
In the following text we shall, for compactness, usually write the spherical harmon-
ics in bra and ket form, substituting

Y m (, ) | m and Y m (, ) m|. (2.89)
In this notion, the orthogonality relations (2.86) and the matrix elements of an oper-
ator A are written as

m m = mm m .
and A m, m = m|A (2.90)
Several useful relations are summarized in Appendixes B and C.

It has to be pointed out, that the complex form of the spherical harmonics
Y m (, ) illustrated in Fig. 2.6 is well adapted for many problems in atomic physics
but it is by no means the only possible representation of the angular part of atomic
orbitals. An alternative frequently used in molecular physics (see Chaps. 3 and 4,
Vol. 2) and always in chemistry, are real combinations of spherical harmonics. They
are described in some detail in Appendix D.
2.5.4 Electron Spin
As shown by the S TERN -G ERLACH experiment, the electron has in addition to mass
and charge one further property which we have identified in Sect. 1.10 as an intrinsic
angular momentum, called spin. The spin is characterized by the spin quantum num-
ber s = 1/2. Its magnitude is |S| = s(s + 1), and two orientations with angular
momentum components /2 and /2 are possible. Closely related to the spin, the
electron has also a magnetic moment with a g factor as defined by (1.162) close
to ge 2.
One simply transfers the formal rules which we have introduced in Sects. 2.5.2
2.5.3 onto the properties of the spin. Clearly, the spin may not be imaged in position
space. However, we may define quite formally in analogy to the orbital angular
momentum L a new vector operator S with a square magnitude operator S 2 and a
component Sz in z-direction for which the general commutation rules for angular
momenta (2.75) and (2.84) are valid:
2
[
Sx ,
Sy ] = i
Sz , [
Sy ,
Sz ] = i
Sx , [
Sz ,
Sx ] = i
S ,
Sy ,Sz = 0.
(2.91)
They imply, as in the general case, that the components of the spin cannot be mea-
sured simultaneously. However, it is possible to determine its magnitude together
with one of its components is (e.g. the z-component
Sz ).
We now introduce spin states |sms . In analogy to (2.79)(2.83) (for the orbital
angular momentum)
2 |sms = s(s + 1)2 |sms

S and
Sz |sms = ms |sms (2.92)
with ms = s, s + 1, . . . , s holds. Specifically, for the electron spin with s = 1/2

and ms = 1/2 only two basis states exist,
1 1 1 1
= | = | ,
2 2 = | = |+ and 2 2 (2.93)
for which the spin points into +z- and z-direction, respectively. We have intro-
duced here three equivalent notations which are commonly used for compactness.
One also finds in the literature the notation spin function and without bra or ket.
Alternatively, one speaks about spin up () and spin down () states. In any
case, the relations
2 | = 3 2 |
S 2 | = 3 2 |
S
4 4
(2.94)

Sz | = | Sz | = |
2 2
hold, together with the orthonormality relations
| = | = 0 and | = | = 1. (2.95)
From the commutation rules (2.91) one may derive how the other components of
S
act onto the basis (here without proof):
i i
Sx | = | Sx | = | Sy | = | Sy | = | . (2.96)
2 2 2 2
In this basis each arbitrary spin state of an electron may be expressed as
| = + | + | , (2.97)
with the probability amplitudes + and normalized to unity: | = |+ |2 +

| |2 = 1. The probabilities for finding in this state the and component (or spin
up and spin down ) is given by |+ |2 and | |2 , respectively. The expectation
values of the spin components Sk (with k = x, y or z) in the spin state (2.97) are
readily obtained from Sk = |Sk | using (2.94) and (2.96).
Note that although the three components of the spin cannot be measured simulta-
neously, magnitude and phase of the (complex) amplitudes + and determine the
orientation of the spin in three dimensional space uniquely. Assume we start with
a pure basis state say | prepared with an ideal S TERN -G ERLACH experiment,
pointing into an arbitrary direction of space. In a given coordinate system with x-,
y-, and z-direction (differing from that S TERN -G ERLACH coordinate frame) this
state | will again be a superposition of the type (2.97). The mathematics to do
such a coordinate rotation in general is summarized in Appendix E. Experimen-
tally one may determine the expectation values Sk in respect of the x-, y-, and
z-coordinates by using a second, rotatable S TERN -G ERLACH setup. With this one
may perform experiments in each of the corresponding directions and repeat them
many times to obtain expectation values. From these (one needs at least two of them)
the probability amplitudes + and in the new system may easily be derived.
It is convenient to write these amplitudes with (E.16) and = 0 as

+ = cos exp i and = sin exp i , (2.98)
2 2 2 2
which are normalized by definition. As an exercise, the reader may show with the
aid of (2.94) and (2.96), that the parameters and give the polar and azimuthal
angles, respectively, at which the so defined spin state is oriented in space.
The above provides a toolbox which is fully sufficient to describe the properties
of the spin states. Nevertheless one often writes perhaps for historical reasons or
better visualization operators in the form of matrices and eigenstates as vectors,
the so called spinors

+
= and = + (2.99)

1 0
with the basis = and = .
0 1
The matrix elements of the operators

Sx ,
Sy and
Sz are obtained from (2.96). They
may be summarized as spin operators in matrix form,

S= , (2.100)
2
with
being a vector operator, the so called PAULI vector, which is built from the
PAULI matrices:

0 1 0 i 1 0
x = y = and z = (2.101)
1 0 i 0 0 1

3 1 0
S = Sx2 +
Sy2 +
2
Sz2 = 2 . (2.102)
4 0 1
For later use we note here that the PAULI matrices anti-commutate
i j + j i = 2ij , (2.103)
and thus
x2 = y2 = z2 =
1 and x y = y x = iz . (2.104)
Another useful relation is obtained by applying (2.103):

3 2 3 2 2 1 0
= r 2
2
3(S r) = (x x + y y + z z) = r
2 2
S . (2.105)
4 4 0 1
Section summary
Orbital angular momentum operators may be derived from L=r p , with

p = i, or be constructed from the commutation rules (2.75).
From the spatial representations of the operators z according to
2
L and L
(2.65) and (2.70), respectively, one obtains eigenvalues and wave functions
as summarized in (2.79)(2.83), and schematically illustrated in the vector
diagram Fig. 2.5.
The shape of the orbitals with lowest angular momentum s, p and d shown
in Fig. 2.6 should be memorized. General formulas and properties are sum-
marized in Appendix B.1.2, specific expressions are tabulated in Table B.1.
The electron spin obeys the same commutation rules as orbital angular mo-
menta. Its intrinsic angular momentum, the spin, is however s = 1/2 and the
projection quantum number ms = 1/2. A frequently used representation of
the spin operators are the PAULI matrices (2.101) and (2.102) which act on the
so called spinors two component representations of the spin eigenfunctions.
2.6 One Electron Systems and the Hydrogen Atom 117
2.6 One Electron Systems and the Hydrogen Atom
We are now fully prepared to treat a particle in a centrosymmetric potential V (r),

specifically the one electron problem. Much of the following will be based on what
we have learned about angular momenta in Sect. 2.5. We keep the discussion gen-
eral as long as possible and specialize to the H atom, i.e. to the pure, attractive
C OULOMB potential V (r) < 0 in Sect. 2.6.5.
2.6.1 Quantum Mechanics of the One Particle System
We start by formulating the S CHRDINGER equation precisely. With (2.64)(2.67)

we may write the Hamiltonian (2.10) as:
2
=
H
p2
+ V (r) = Hr + L + V (r) (2.106)
2me 2me r 2

2 1 2
with Hr = r , (2.107)
2me r 2 r r
and obtain the one particle S CHRDINGER equation in polar coordinates:

2
r + L
H + V (r) n m (r, , ) = Wn m n m (r, , ). (2.108)
2me r 2
In Sect. 2.5.3 we have discussed in detail the eigenfunctions Y m (, ) and eigen-

values 2 ( + 1) of
2
L . To solve (2.108) we make the separation ansatz
n m (r, , ) = Rn (r)Y m (, ), so that (2.109)

2
r + L + V (r) Rn (r)Y m (, ) = W Rn (r)Y m (, ).

H
2me r 2
Since Hr and V (r) act only on the radial and 2

L only on the angular part we obtain
with (2.107) and (2.79):

2 1 d 2 d 2 ( + 1)
r + + V (r) Rn (r) = W Rn (r). (2.110)
2me r 2 dr dr 2me r 2
The sum of centrifugal potential 2 ( + 1)/(2me r 2 ) and C OULOMB potential

1/r is called effective potential:
2 ( + 1)
Veff (r) = V (r) + . (2.111)
2me r 2
With this and the substitution
Rn (r) = un (r)/r (2.112)
one obtains a relatively simple, one dimensional, ordinary differential equa-

tion (ODE) which may be integrated without problems:
2 d2 un
2
+ Wn Veff (r) un (r) = 0. (2.113)
2me dr
Note: The total energy does not depend on the projection quantum number m and is
thus be written W = Wn . Zero energy is usually set for the completely unbound
system, i.e. for electron and nucleus at infinite distance with no kinetic energy.
Bound electrons have negative energies Wn < 0, while free electrons have total
energies W > 0. Extending (2.89) to the full electron wave function one often ab-
breviates
Rn (r)Y m (, ) |n m . (2.114)
More specifically, one even writes these atomic orbitals shorthand as n = 1s, 2s,
2p, 3s, 3p, 3d, etc. in the notation introduced in Sect. 2.5.3.
2.6.2 Atomic Units
We recall here the concept of atomic units (a.u.), introduced in Sect. 1.8.3:
energy Eh = me e4 02 h2 /4

length a0 = 0 h2 e2 m1
e / = / me Eh (2.115)
time t0 = 202 h3 e4 m1
e /.
Numerical values of these quantities are given in Appendix A, and the most recent,
accurate updates can be found at NIST (2010). We use these definitions here to
rewrite the radial S CHRDINGER equation 2
(2.113). We multiply (2.113) by me /
and a0 , and apply the identity a0 = / me Eh to obtain in dimensionless form:
2
1 d2 un ( + 1) V (r/a0 )
+ W n /E h + un (r) = 0. (2.116)
2 d(r/a0 )2 2(r/a0 )2 Eh
For simplicity, one may just substitute r/a0 r and Wn / Eh Wn as well as

V (r)/Eh V (r). This just implies that all observables are measured in a.u. and
the radial S CHRDINGER equation (2.113) reads now
1 d2 un
2
+ Wn Veff (r) un (r) = 0 (2.117)
2 dr
( + 1)
with Veff (r) = V (r) + and for the H atom V (r) = Z/r.
2r 2
Following this scheme one may rewrite all atomic equations in a dimensionless,
rather clean looking form. Theoreticians, in particular, like this kind of equation very
much and even give the recipe to just set = e = me = 1 which really oversimpli-
fies what has to be done. The procedure has one serious disadvantage: a dimensional
analysis is no longer possible which is often highly commendable to check com-
plex calculations. Thus, we typically try to use equations in a form indicated by
(2.116), and carry the a.u. a0 , Eh , and t0 explicitly along. Sometimes one may even
be able to combine elementaryconstants to truly dimensionless quantities, such as
the fine structure constant = Eh /me c2 according to (1.10).
2.6.3 Centre of Mass Motion and Reduced Mass
Up to now we have treated the problem as if the electron would orbit around a space
fixed centre. As the nuclear mass is much larger than the electron mass me in the
case of the H atom with a proton as nucleus mp 1840me the centre of mass
is indeed very close to r = 0. For more demanding precision one has, however, to
correct for the difference. As in classical mechanics (see corrections to the B OHR
model described in Sect. 1.8.5), in quantum mechanics too, one transforms the two
body problem into an effective one particle problem by replacing the electron mass
me with the reduced mass m e (1.143) of the system. All a.u. have, in principle, to
be replaced correspondingly (kinematic correction):
me e
m e
m
a0 a 0 = a0 , Eh E h = Eh , and t0 t0 = t0 . (2.118)
e
m me me
For simplicity we shall, however, in the following text continue to use me and the
units a0 , Eh and t0 , and refer to the exact calculations if relevant.
2.6.4 Qualitative Considerations
While we know already the angular part of the hydrogenic wave functions, we
are still left with the task to find physically meaningful solutions to the radial
S CHRDINGER equation (2.117). They have to behave reasonably at r = 0 (see
below) and must not diverge for r . From this follows necessarily that only a
particular set of discrete total energies Wn < 0 leads to such solutions. To find these
is the task at hand.
Before applying mathematics we want to obtain a qualitative picture to support
our physical intuition. Figure 2.7 illustrates this for the = 0 case in a C OULOMB
potential V (r) = Z/r. We derive the kinetic energy of the electron at different
W,V(r )
0
Wn r
Wkin
rcl
V
un (r ) large classically for-
bidden region:
= h /p exponential
small decrease of
wavefunction r
Fig. 2.7 Bound state radial wave functions for s states, schematic. Top: C OULOMB potential
V (r) 1/r (red) and total energy Wn < 0 (black) determine the classical turning point rcl
(onset of the classically forbidden region). Bottom: The characteristic behaviour of radial wave
functions un is explained by changes of the kinetic energy Wkin in different regions of the poten-
tial (see text)
positions in the potential from Wkin = Wn V (r) and take the corresponding DE
B ROGLIE wavelength = h/p = h/(2me Wkin )1/2 as an indication for changes in
the radial wave function un (r). It obviously will change more rapidly for small
r (large Wkin ) than in the neighbourhood of the classical turning point rcl , where
Wkin = 0. In the classically forbidden area with r > rk (negative kinetic energies)
we expect exponential damping of the wave function, as illustrated in Fig. 2.7.
In a next step we explore the limits for very large and very small r. For the
limiting case r we may neglect the potential altogether and (2.117) becomes
a simple oscillator equation:
1 d2 un
+ Wn un (r) = 0.
2 dr 2

The classical solution is un (r) exp(i 2Wn r). Since for bound states Wn < 0,
we note for large r

lim un (r) r n exp 2|Wn |r . (2.119)
r
In the opposite limit r 0 the centrifugal term ( +1)/2r 2 dominates the potential
in (2.117) and
1 d2 un ( + 1)
un (r) = 0
2 dr 2 2r 2
r0
has the solution un (r) = Ar +1 as one easily verifies. Thus, we note for small r

lim Rn (r) = lim un (r)/r r . (2.120)
r0 r0
2.6.5 Exact Solution for the H Atom
We now specialize to the H atom. In the present chapter we neglect the size of the
atomic nucleus (positively charged with +Ze) since nuclear radii rnuc are much
smaller than atomic radii, typically ratom 105 rnuc . Thus, we are dealing with a
pure C OULOMB potential
1 Ze2
V (r) = , (2.121)
40 r
apart from small and very small effects which will be treated in Chaps. 6 and 9,
respectively. The general solution of the radial S CHRDINGER equation (2.110) is
found by using a power series of the type

...
Rn (r) = exp 2|Wn |r Ak r k ,
k=
which includes the limiting cases just discussed. Well known results from mathemat-
ics are used and we summarize here the results, again without proof. For hydrogen
like systems i.e. for one electron in the C OULOMB potentialof a Z fold charged
nucleus the radial function is
Rn (r) = unl (r)/r = An e/2 Ln+

2 +1
() (2.122)
3/2
2Z Z 2 (n 1)!
where = r/a0 and An =
n a0 n2 [(n + )!]3
with the so called associated L AGUERRE polynomials:

n 1
[(n + )!]2 k
2 +1
Ln+ () = (1)k+1 . (2.123)
(n 1 k)!(2 + 1 + k)! k!
k=0
With An these radial functions are orthonormalized:

Rn (r)Rn (r)r 2 dr = nn . (2.124)
0
We introduce here finally the often used term good quantum numbers: They char-
acterize the eigenvalues of such observables that may be measured simultaneously
with the H AMILTON operator, i.e. their operators commutate with the Hamiltonian.
We already know and m as typical examples: they are part of the set of quantum
numbers characterizing the total energy of the system; z are simultaneously
2
L and L

measurable with H .
Wn / eV n
=0 =1 =2 =3 =4
0
8
6
5s 5p 5d 5f 5g
- 0.85 4
4s 4p 4d 4f
- 1.51 3
3s 3p 3d
- 3.40 2
2s 2p Z2
Wn = E h
2n 2
independent of
-13.6 1
1s
Fig. 2.8 Term energies of the hydrogen atom (Z = 1) for different n and
Table 2.1 The lowest atomic levels, their energies in the H atom and the degeneracy of the states
(with Eh = 27.2 eV)
Shell Orbital n m Wn Degeneracy
Without spin With spin
sum in shell sum in shell
K 1s 1 0 0 Eh /2 1 1 2 2
L 2s 2 0 0 Eh /8 1 2
2p 2 1 0, 1 3 4 6 8
M 3s 3 0 0 Eh /18 1 2
3p 3 1 0, 1 3 6
3d 3 2 0, 1, 2 5 9 10 18
N 4s, p, d, f 4 Eh /32 16 32
2.6.6 Energy Levels
These solutions of the S CHRDINGER equation are the quantum mechanical equiv-
alent to B OHRs stationary orbits. Substituting unl (r) according to (2.122) into the
radial equation (2.113) one finds the eigenenergies Wn for the H atom. Remark-
ably, they are identical to the energies (1.137) from the B OHR model.8 These results
are summarized in Fig. 2.8 and Table 2.1.
We recall that the states are characterized by the principle quantum number
n = 1, 2, 3, . . . , the angular momentum quantum number (0 n 1) and
the projection quantum number m ( m ). Each set n m of quantum num-
bers refers to a different wave function (atomic orbital). In addition, we also have to
consider the spin of the electron, with projections ms = 1/2.
8 The corrections for finite mass of the atomic nucleus (Sect. 1.8.5) also apply.
Veff (r ) / E h
=1 =2
0.2
5 10 15 20
0
n=2 n=3
r /a 0
COULOMB potential
- 0.5 n =1 = Veff for = 0
Fig. 2.9 Illustration of degeneracy: Shown are the pure C OULOMB potential (red line) and the
effective potentials (black) for the H atom. The term energies Wn are indicated by horizontal
lines in their respective effective potentials: for = 0 (dotted red), for = 1 (dashed grey) and for
= 2 (heavy, full red). As indicated, the C OULOMB potential leads to a characteristic degeneracy
of states with equal n but different
However, for the H atom the energies Wn depend only on the principle quantum
number n. Thus, the energy levels are degenerate, the total degeneracy in a shell
n being gn = 2n2 . This is summarized in Table 2.1 for n = 1 to 4, also showing
the assignment of orbitals with equal principle quantum number n (i.e. with equal
energies and similar orbital radii) to specific shells (K, L, M, N . . . corresponding
to n = 1, 2, 3, 4, . . .).
It is important to note that degeneracy is a special property of the pure
C OULOMB potential, while m degeneracy occurs for all atoms if no external field
is present. Figure 2.9 illustrates degeneracy in the potential energy diagram, also
showing the effective potentials (2.111) for = 1 and = 2 (for = 0 effective and
C OULOMB potential are identical).
2.6.7 Radial Functions
The radial wave functions have very specific shapes, which may be understood from
the properties of the effective potentials. Figure 2.10 illustrates this schematically by
way of example for the n = 3 level and orbital angular momenta = 0 and 1. Shown
are C OULOMB potential, centrifugal potential and effective potential for = 1. The
classically forbidden areas (Wkin < 0) are grey shaded, the classical turning points.
While the radial functions for = 0 start with a finite value at r = 0, for = 1
the probability there is = 0 (since ( + 1)/(2r 2 ) ). Oscillations of the radial
functions are expected only in between the classically turning points.
Table 2.2 presents the radial wave functions Rn (r) for the six energetically low-
est states (n 3) of atomic hydrogen (Z = 1) and H-like atoms (Z > 1) in closed
form.
From the radial wave functions one derives the probability distributions
2
w(r)dr = Rn (r) r 2 dr (2.125)
Fig. 2.10 Schematic V, W =0 V, W =1

Veff
illustration of the relation
between potential (top) and
( +1)/ 2r 2
corresponding wave function 0 0
(bottom) for different , W30 r W r
31
exemplified for the n = 3 1/r
1/r
level with = 0 and 1. Grey
shaded is the classically Rn (r)
forbidden region, red-black
circles indicate classical
turning points 0 r 0 r
Table 2.2 Radial wave n Rn (r) with = 2Zr/(na 0 ) and a 0 = a0 me /m

e
functions for the energetically
lowest states of H and H-like 1 0 R10 (r) = 2( aZ0 )3/2 e/2
atoms
2 0 R20 (r) = 1
( Z )3/2 (2 )e/2
2 2 a 0
1 R21 (r) = 1
( Z )3/2 e/2
2 6 a 0
3 0 R30 (r) = 1
( Z )3/2 (6 6 + 2 )e/2
9 3 a 0
1 R31 (r) = 1
( Z )3/2 (4 )e/2
9 6 a 0
2 R32 (r) = 1 ( Z )3/2 2 e/2

9 30 a 0
for the electron to be found between r and r + dr. For comparison with the classical
picture of an orbiting electron one has to consult these probability distributions as
a function of distance from the nucleus. A graphical illustration of the radial wave
functions Rn (r) and the radial probability distributions w(r) is shown in Fig. 2.11.
Closer inspection of the probability distributions shows, that their maxima match ex-
actly the radii of the respective B OHR orbits for the largest = n 1 at any given n!
2.6.8 Density Plots
The complete solutions (atomic orbitals) n m (r, , ) of the S CHRDINGER equa-

tion (2.108) consist of radial and angular part. With the spherical harmonics
Sect. 2.5.3 and the just discussed radial functions the original ansatz (2.109)
n m (r, , ) = Rn (r)Y m (, ) is validated. With (2.86) and (2.124) these wave
functions are already orthonormalized:

n mn m =
n m n m r 2 dr sin d d = nn mm . (2.126)
The square of the wave function |n m (r, , )|2 represents the probability distribu-
tion for finding electrons per volume element, or equivalently the electron density
within the atom as a function of position in space. Since these atomic orbitals form
R n l (r) /a 0-3/2 r 2 [R n l (r) ] 2 /a 0-1
0.1 0.04 0.1 0.1 0.1

0.4
3s 3p 3d 3s 3p 3d
0
0.2 0.2 0 10 20
0.5 0.1 0.1 0.1
2s 2p 2s 2p
0 0
2.0 0 10 20 0 10 20
0.4
1s 1s
1.0 0.2 blow up
0.4
0.2 1s
0 10 20 r /a0 0 10 20
0 2 4
Fig. 2.11 Radial wave functions of the H atoms Rn (r) and probability distributions r 2 Rn 2 (r) for
the K, L and M shell. The dashed, vertical lines in the probability plots for the highest at a given
n indicate the corresponding B OHR radii. Note the drawing for the 1s orbital on a blown up r-scale
3
z 1s 0 2p 1 3s0 3p 1
5 10 10
0 0 0 0
-5 -10 -10
-3
-3 0 3 -5 0 5 -10 0 10 -10 0 10
3d 1 3d 2 20 4 s 0 40 5s 0
10 10
0 0 0 0
-10 -10
-20 -40
-10 0 10 -10 0 10 -20 0 20 -40 0 40
x
Fig. 2.12 Density plots for some characteristic H atom wave functions n m . Plotted are equiden-
sity lines (red high, grey low probability density). The distances are given in a0
a base for solving many key problems in atomic and molecular physics, we recom-
mend our readers to visualize and memorize their general shape intensively. One
finds a host of instructive Internet-sites, e.g. with Java applets to generate the H or-
bitals in a variety of displays. Thus, we present in Fig. 2.12 only a small selection of
cuts through the 3D density distributions. Plotted are the contour lines of the den-
sity |n m (x, y = 0, z)|2 = |Rn (r)Y m (, )|2 in the zx plane on a linear scale (in
contrast to many presentations in the WWW, where one finds the density plotted on
a log-scale). For clearness we have indicated the highest densities by red areas.
We finally note again, that only the ns orbitals have a finite density at the origin,
|ns (0)| > 0, while
n m (0, , ) 0 for > 0. Specifically, from (2.122) and with
Y00 (, ) = 1/ 4 we record for later use:

n00 (0)2 = Z 3 / a 3 n3 . (2.127)
0
2.6.9 Spectra of the H Atom
Just as in the B OHR model, the transition energies (wavenumbers, frequencies) in

atomic hydrogen are given by the RYDBERG formula (1.147)(1.149). Traditionally,
one characterizes spectral lines series for transitions n n

1 1 Eh m e
= h = Wn Wn = Z 2
2
2 (2.128)
n n 2 me
according to n (with n < n). Clearly, due to the m and (for the H atom) degener-
acy, the angular momentum quantum numbers do not influence the position of the
spectral lines as one also verifies by a glance at Fig. 2.8. This holds at least in 1st
order approximation. Finer effects will be discussed in Chap. 6.
A lot of scientific detective work of the early pioneers went into discovering the
connection between spectra and term energies, and the series are named after those
who originally discovered them. Most prominent are the LYMAN (n = 1 in the
VUV), BALMER (n = 2 in the VIS and near UV) and PASCHEN series (n = 3 in
the near IR). Within the series, the BALMER lines are historically referred to as H-
alpha, H-beta, H-gamma etc., more generally the lines of the different series
are designated as Ly-, and Ly-, Ly- , . . . , Ba-, Ba-, Ba- , . . . and so on, with
= n + 2, = n + 3, etc.
2.6.10 Expectation Values of r k
One often needs to know the expectation values of the electron distance from origin
r to a certain power k. In principle, they could be determined by a large number of
individual measurements of this value in an suitably designed experiment. Quantum
mechanics provides:

k
r = n |r k |n = Rn (r)r k Rn (r)r 2 dr = 2
Rn (r)r 2+k dr (2.129)
0 0

with 2
Rn (r)r 2 dr = 1.
0
These integrations are trivial but somewhat tedious. Using (2.122) and (2.123) one
obtains (for later reference):9
9 Here too, for high precision measurements one has to replace a0 a 0 = a0 me /m

e.

n2 1 ( + 1)
r n = a0 1+ 1
Z 2 n2
"
#
2 n4 3 ( + 1) 1/3
r n = a02 2 1 + 1
Z 2 n2

1 1 Z
= (2.130)
r n a0 n2

1 1 Z2
=
r 2 n a02 n3 ( + 1/2)

1 1 Z3
= .
r 3 n a03 n3 ( + 1/2)( + 1)
2.6.11 Comparison with the BOHR Model
One often hears the verdict that the B OHR model is basically wrong albeit predict-
ing the correct term energies Wn and spectra (1.131) for the H atom in agreement
with quantum mechanics and thus should be abolished when teaching modern
atomic physics. We do not adhere to such dogmatic view.
Of course, B OHRs orbits have to be replaced by atomic orbitals, i.e. by den-
sity distributions of the electrons within the atoms, and B OHRs quantization condi-
tion (1.130), L = n, has to be confronted with its quantum mechanical analogue,
z () = m(). Obviously, m n 1 is not equal to n as postulated in the
L
B OHR model but it is a good first guess.
On the other hand, the concept of angular momentum quantization, of stationary
states, and of radiation emitted upon transition between them with h = Wn was
a brilliant, daring and instrumental step on the way to understand the quantum nature
of microscopic phenomena even though the fact that states without orbital angular
momentum ( = 0) do exist, does not fit at all into the B OHR model. However, the
higher the angular momenta, the closer the atomic orbitals correspond to B OHRs
orbits. And in general, the correspondence principle holds:
quantum mechanical and classical values of observables approach each other

for very large quantum numbers.
A quantitative comparison of atomic orbitals with B OHRs orbits in hydrogen like

atoms shows that the maxima of the electron radial distribution (see Fig. 2.11) for
max = n 1 are indeed found at r = a0 n2 /Z, i.e. they correspond exactly to the
B OHR radii according to (1.135). And more specifically, electron orbitals with m =
may to some extend be associated with electrons moving on a circle a notion
which holds again particularly well for large . Such circular states are actually
an interesting subject studied in highly excited RYDBERG atoms.
However, the maxima of the probability distributions are of course not directly
observable; rather one may compare expectation values such as n |r|n = r . For
the largest angular momenta = n 1 one obtains from (2.130)

1 2 a0 a0
r = lim n + n = n2 , (2.131)
n 1 2 Z Z
which obviously agrees in the limit with B OHRs prediction (1.135). Conversely,
for the smallest values of = 0, the average radius r is distinctively larger than
predicted by the B OHR model, namely (3/2)n2 a0 /Z.
In summary, we do not feel that the B OHR model should be completely forgotten.
Apart from its outstanding historical impact, we shall come across a number of spe-
cific aspects in modern AMO physics for which useful, simple models or concepts
have been stimulated by images of electrons moving on classical trajectories com-
bined with appropriate quantization rules. We may think of B ORN -O PPENHEIMER
approximation for molecular physics and of semiclassical trajectory calculations in
atomic scattering theory, to mention just two important, and very successful exam-
ples or, as we shall see in the following section, the derivation of the magnetic
moment of an electron associated with its orbital angular momentum.
Section summary
The eigenenergies of the H atom, Wn = Z 2 /(2n2 )Eh , depend only on the
principle quantum number n . . . a specific consequence of the pure C OULOMB
potential. For precision measurements this has to be corrected by m e /me
where m e is the reduced mass of the electron.
Electron wave functions for the H atom can be expressed in analytical form
n m (r, , ) = Rn (r)Y m (, ) with the spherical harmonics Y m (, ) and
the radial function Rn (r) being proportional to the L AGUERRE polynomials.
The asymptotic behaviour of the wave functions should be memorized:
limr Rn (r) exp( 2|Wn |r) and limr0 Rn (r) r .
good quantum numbers characterize the eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator.
2.7 Normal ZEEMAN Effect
The so called normal Z EEMAN effect is actually not normal at all. It is observed
only in special cases. It concerns atoms in an external magnetic field. The word
normal simply alludes to the classical interpretation of such a situation ignoring
the electron spin. We shall treat atoms in external fields in great detail in Chap. 8.
Thus, the subject is touched here just briefly, and only since it confronts us for the
first time with the removal of a specific energy degeneracy, the m degeneracy in this
case.
2.7 Normal ZEEMAN Effect 129
Fig. 2.13 Electron circling B z

in a magnetic field B and its
orbital magnetic moment M Mz
M = - B L z /
e- v
2.7.1 Angular Momentum in an External B-Field
As discussed in Sect. 1.9, an orbital angular momentum has a magnetic moment

(1.150) which we rewrite in operator form

L
= e
M L = B , (2.132)
2me
with the B OHR magneton B . Its potential energy (1.153) in an external magnetic
field B is

B = M
V B = B L B. (2.133)

We choose z B so that VB = +B (L z /B), as indicated in the vector diagram
B ,
Fig. 2.13. The Hamiltonian contains now an additional term V

=H
H 0 + V 0 + B Lz B,
B = H (2.134)

assuming the unperturbed S CHRDINGER equation to be
0 n m = W (0) n m .
H
0 may e.g. be the H AMILTON operator (2.106) for the H atom. Thus, in the pres-
H
ence of an external field the S CHRDINGER equation to be solved reads

H0 + B (L z /)B |n m = W (0) + W |n m , (2.135)
with W being the change of total energy in respect of the unperturbed state. Here
and in the following we use state vectors instead of wave functions n m |n m
for compact writing.
The present problem is a particular simple case for the so called perturbation the-
ory to which we shall come later in some detail. At this point we just remember that
according to (2.79) and (2.81) the eigenfunctions n m (r, , ) = Rn (r)Yn (, )
of the H atom, or their eigenstates |n m , respectively, are also eigenfunctions
(eigenstates) of z with
2
L and L
z |n m = m|n m .
L (2.136)
Fig. 2.14 The normal Wm m=1

Z EEMAN effect
m=0
B
m =1
Thus, the eigenstates |n m of the unperturbed Hamiltonian H 0 are also eigenstates

so that we may insert (2.136) into (2.135) to obtain:
of the full H

(H0 + B mB)|n m = W (0) + Wm |n m
with Wm = B mB. (2.137)
2.7.2 Removal of m Degeneracy
The latter relation obviously implies that in a magnetic field the energy degeneracy
for different m is removed. The originally identical energies of the 2 + 1 states
|n m which correspond to a given value of n and , now split into 2 + 1 different
energy sub-levels. According to (2.137) the splitting is proportional to m and B.
The origin of this splitting is that the magnetic field breaks the spherical symmetry
characteristic for the unperturbed H atom.
We illustrate this for the example of an np state |npm with = 1 and sublevels
m = 1, 0, and 1. Figure 2.14 shows the energy changes Wm as a function of the
magnetic field B.
One may observe this splitting e.g. in optical emission spectra. Figure 2.15 com-
pares the emerging spectra for (a) a p s and (b) a d p transition. The individ-
ual transitions are indicated by black, downward arrows. The selection rules applied
in this plot are = 1 and m = 0, 1. They will be derived and discussed in
Sect. 4.4.
As the degeneracy and hence the number of split levels is 2 + 1, a manifold
of transitions may occur if states with > 1 are involved. This is illustrated in
Fig. 2.15(b) where the upper levels correspond to a d state. Since, however, for
the normal Z EEMAN effect the splitting between neighbouring levels is always
B B according to (2.137), independent of and m, and since for all transitions
m = 0, 1, one nevertheless sees only a line triplet in all cases.
As mentioned at the beginning of this section: in reality this kind of Z EEMAN
effect is seen only in special situations (see Sect. 8.1.2), since usually the spin of
electrons plays an important role and complicates the observations.
We note here an important message from a situation which one typically en-
counters in quantum systems with two or more degenerate states |1 , |2 , |3 : The
degeneracy is removed as soon as some additional, perturbing interaction V1 has to
be considered for which the matrix elements i|V1 |j between some of these states
do not vanish.
2.8 Dispersion Relations 131
m=+2
(a) (b) +1
+1
BB 0
p 0
d 1
1 BB
2
+1
BB
p 0
s m = 1 0 +1
1
m = 1 0 +1
Fig. 2.15 Normal Z EEMAN effect for (a) p s and (b) d p transitions. In case of equal
splitting in the excited and ground states one observes in each case a line triplet in spite of the
5-fold splitting in the d state
Section summary
A magnetic field B removes the central symmetry and hence the m degener-
acy. For the so called normal Z EEMAN effect, theory predicts a level splitting
Wm = B mB.
The selection rules for optical transitions are = 1 and m = 0, 1.
2.8 Dispersion Relations
We make a brief detour here, illuminating the borderline between atomic, molecular,
optical and solid state physics. Traditionally, dispersion relations are used in optics
and characterize an important material property: the dependence of the wavelength
(or wavenumber k = 2/) of electromagnetic radiation on its frequency . From
a quantum mechanical point of view one may generalize this to describe the relation
between the energy of a photon W = (or in fact the energy of any other particle)
and its wave vector k.
For the massless particle photon in vacuo, with c = = /k, the dispersion
relation
W = = c|k| (2.138)
is obviously a linear relation between energy and wave vector. In contrast, the en-
ergy of a freely moving, nonrelativistic electron (whose mass is me ) is
me 2 p2 2 k 2
W (k) = WP + v = WP + = WP + , (2.139)
2 2me 2me
where we have used the DE B ROGLIE relation (1.118). WP allows for arbitrary en-
ergy calibration and may e.g. account for a potential energy or the rest mass energy.
Thus, in this case the dispersion relation is quadratic. Both cases are illustrated
graphically in Fig. 2.16.
Fig. 2.16 Dispersion W W

relations (a) linear for (a) (b)
a massless particle (e.g.
a photon) and (b) quadratic
for a particle with mass (e.g.
an electron), moving freely in
3D space
k k
It should be noted that the quadratic relation (2.139) for particles with mass is
also in accord with the quantum mechanical calculation for a free particle in a 1D-
or 3D-box according to (2.52) or (2.55). Providing the box is large enough, the
energies may be considered continuous, i.e. they generally follow (2.139). However,
this is only the most simple model for the electronic band structure in a solid. The
particle in a box model does not account for the fact that the electrons in a solid
do not really move freely: rather, the electrons move in the lattice of atomic ions
and experience a periodic potential with strong attraction close to the ionic cores.
Elsewhere the potential is partially screened by bound and other free electrons in
the solid. We mention here two crucial consequences arising from this situation.
First, the dispersion relation (2.139) will have to be modified, which in principle
requires a serious band structure calculation. Many phenomena may, however, be
explained by the so called parabolic approximation. It parameterizes the band en-
ergy by introducing an effective mass me of the electron which may even depend on
the direction into which the electron moves:
2
kx ky2 kz2
W (k) = WP 2
+ + . (2.140)
2mx 2my 2mz
The sign allows to extend the concept also to electron holes (positive charges in
the latice) and both these quasi-particles may have different masses. This mass
enters into all further calculations on the dynamics and statistics of electrons and
holes in a solid, e.g. into the F ERMI -D IRAC statistics discussed in Sect. 2.4.3.
Second, one has to account for the periodicity of the motion in the lattice. The
solutions for this problem are so called B LOCH waves
(r) exp(ikr)uk (r) (2.141)
which are the product of a plane wave exp(ikr) and a periodic function uk (r). The
latter has to obey the periodic boundary condition
uk (r + T ) = uk (r), (2.142)
where T is any translation from one elementary cell of the lattice into another. Over-
all we still expect the dispersion relation (2.139) for electron energies to be more or
less valid. As detailed studies show, one may have to replace the electron mass me
by a (different) effective mass meff but the quadratic dependence of the energy W
2.8 Dispersion Relations 133
on the wave vector typically holds. The electron motion in general averages over
the periodic potential. However, the energies and eigenfunctions will be influenced
strongly by the lattice potential if the electrons are particularly close to the ions or
particularly far away from them. This is specifically relevant when the electron wave
vector k = k BZ is at (or close to) the B RILLOUIN zone (BZ), i.e. if the correspond-
ing wave functions are constructively interfering (see Sect. 1.4.9). Such electrons
feel the periodic potential strongly, all others experience only an average.
We cannot go into detail of these concepts which are fundamental for the the-
ory of band structure in solid states. We just emphasize some aspects which are
also of importance in molecular physics. Thus, let us discuss the particularly simple
situation of the 1D case, which e.g. describes a chain of atoms with a distance a
from each other (also, with slight modifications, a ring like molecule). According to
(1.95) and (1.93) the B RAGG condition for the ntn BZ then reads

k = kBZ = n . (2.143)
a
The corresponding wave functions are essentially exp(ikBZ x) and exp(ikBZ x), and
for free travelling electrons the energy in that situation would be 2 k 2BZ /2me in both
cases. We have two energetically degenerate solutions. However, the most general
solutions are linear superpositions of both, i.e. standing waves, with the two physi-
cally reasonable cases:
(x) exp(ikBZ x) exp(ikBZ x). (2.144)
These correspond to cos kBZ x and sin kBZ x. The probability |(x)|2 to find the elec-
tron close to the lattice ions is maximal in the first case, minimal in the latter. The
consequence of the perturbing periodic potential is removal of the degeneracy, sim-
ilar to Z EEMAN splitting discussed in the last section. And since the perturbing
potential is highly attractive, we expect the energy to be lowered in the first case,
and to be risen in the second case. In effect, at the boundary of the BZ we expect
the energies to split into two, and a gap to arise between two bands of the otherwise
continuous 2 k 2 /2me distribution.
This is illustrated in Fig. 2.17(a). Since the periodicity of the system does not
favour any particular origin in the reciprocal lattice, one projects all the possible
energies onto the first BZ as shown in Fig. 2.17(b). Electrons in the system may as-
sume all energies W in the grey shaded energy bands. The energy regions marked
band gap are energetically forbidden.
In summary, in the solid state case continuous energy bands, with gaps in be-
tween them, replace the discrete energy levels which we have discussed for the
atomic case (specifically for the H atom in the present chapter). One must, how-
ever, be aware of the fact that Fig. 2.17 shows only a particularly simple situation
(1D case, one valence electron only). In general, the band structure of solid states
is much more complicated, and depending on how many electrons are available to
fill the bands, it provides the basis for such different materials as metals, isolators or
semiconductors.
(a) W W (b)
3rd 2nd 1st 1st 2nd 3rd band gap
band gap
band gap
k k
- 3 a
2 - 2 3
a -
a a 0
a a a
Fig. 2.17 Emergence of energy bands in a periodic system (lattice constant a) and band gaps.
(a) Allowed energies ( ) with distortions at the boundaries of the 1st, 2nd, and 3rd B RILLOUIN
zone in comparison with the free particle ( ) as a function of k. (b) Projection onto the first
B RILLOUIN zone
Section summary
Dispersion relations describe how the energy W of the system depends on
the wave vector k. The most simple cases are (a) the photon (a massless
particle) W = ck, and (b) an electron (a particle with mass) moving freely
W = 2 k 2 /(2me ).
The influence of an average potential can be accounted for by an effective
mass, replacing me which even may depend on the direction of the electrons
momentum.
B LOCH waves, (r) exp(ikr)uk (r), are constructed to include the period-
icity of the lattice in the function uk (r) = uk (r + T ).

BZ: B RILLOUIN zone, represents all wave vectors of radiation which can be
B RAGG-reflected by a crystal lattice. Important concept in solid state physics.
chemical-potential: In statistical thermodynamics defined as the amount of energy
or work that is necessary to change the number of particles of a system (by 1)
without disturbing the equilibrium of the system, see in Sect. 1.3.4.
DOS: Density of states, number of states for a specified observable per unit of
that observable. Most frequently the observable is the energy of a particle in a
system. Typically it is given per unit volume of the system studied.
good quantum number: Quantum number for eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator (see Sect. 2.6.5).
References 135

ODE: Ordinary differential equation.
PDE: Partial differential equation.
21348 (2007).
References
B ORN , M.: 1927. Das Adiabatenprinzip in der Quantenmechanik. Z. Phys., 40, 167192.
B ORN , M.: 1954. The N OBEL prize in physics: for his fundamental research in quantum
mechanics, especially for his statistical interpretation of the wave function, Stockholm.
E ISAMAN , M. D., J. FAN, A. M IGDALL and S. V. P OLYAKOV: 2011. Invited review article:
single-photon sources and detectors. Rev. Sci. Instrum., 82, 071101.
H EISENBERG , W. K.: 1932. The N OBEL prize in physics: in recognition of the great merits
of his theoretical and experimental investigations on the conduction of electricity by gases,
Stockholm. http://www.nobelprize.org/nobel_prizes/physics/laureates/1932/.
ISO 21348: 2007. Space environment (natural and artificial) Process for determining solar
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/cuu/
Constants/, accessed: 7 Jan 2014.
S CHRDINGER , E. and P. A. M. D IRAC: 1933. The N OBEL prize in physics: for the discov-
ery of new productive forms of atomic theory, Stockholm. http://nobelprize.org/nobel_prizes/
Periodic System and Removal of Degeneracy
3
The S CHRDINGER equation for the hydrogen atom has been

solved in a fully analytic manner. This was possible due to the
special mathematical properties of the 1/r C OULOMB
potential. We introduce now step by step deviations from this
particular simple case, and aim at describing more and more
subtle effects and later on also more complex problems as they
are observed in spectroscopic and dynamic studies of atoms,
molecules and clusters.
Overview
This is a quite compact and important chapter. The attentive reader should
be able to work through it rather quickly after having refreshed her or his
basic knowledge in the previous two chapters. We shall make here the first
steps to generalize the methods successfully applied for the H atom, and
allow for an interaction potential which is no longer strictly proportional
to 1/r. This forms the basis for understanding the physics behind the peri-
odic system of the elements which is summarized in Sect. 3.1. The most sim-
ple multi-electron systems are the alkali atoms. Their energy levels are dis-
cussed in Sect. 3.2 phenomenologically and analyzed qualitatively and quan-
titatively, briefly explaining quantum defect theory. In Sect. 3.3 we introduce
time independent perturbation theory a tool that will be used later on quite
frequently and illustrate it by way of example for the energies of alkali
atoms.
3.1 Shell Structure of Atoms and the Periodic System
The periodic system provides the structural basis for our understanding of atoms
and molecules. Thus, albeit fairly well known, it appears worthwhile to summarize
here the underlying concepts, observations and definitions.

138 3 Periodic System and Removal of Degeneracy
3.1.1 Electron Conguration
The theory of the hydrogen atom as a prototype atom contains already all ingredi-
ents needed for understanding of how the more complex atoms are built up. The
periodic system of the elements follows naturally from the so called Aufbau princi-
ple (from German Aufbau = build-up): to a first order approximation one treats
the N electrons of an atom (with a nuclear charge Z) as independent of each other
and assumes that their respective wave functions look very similar to those of the
electron in a hydrogen atom. It turns out that this approach serves astonishingly well
for a first guess. Of course one has to modify the potential that each individual elec-
tron sees with all the other electrons around it clearly cannot simply be Z/r:
the nuclear charge will partially be screened off by (N 1) electrons.
We shall discuss this screening and its consequences in detail later. Here we
simply note, that each electron (numbered as i = 1, 2, . . . , N ) is characterized by a
set of quantum numbers:
Quantum numbers (ni i mi msi ) (3.1)

with the principle quantum number ni = 1, 2, . . . ,
angular momentum quantum number i = 0, 1, 2, . . . , ni 1
the projection quantum number mi = i , i + 1, . . . , i
and the spin quantum number msi = 1/2.
They correspond to the quantum numbers of the electron in an H atom. The entirety
of quantum numbers for all electrons of an atom in a particular state is called its
configuration {n1 1 m1 ms1 , n2 2 m2 ms2 , . . . , nN N mN ms N }, (3.2)
or somewhat more precisely the electron configuration of the atom.
3.1.2 PAULI Principle
In its best known form the PAULI exclusion principle (short PAULI principle,
N OBEL prize 1945) states that no two electrons in the same atom can have the
same four quantum numbers, briefly
(na a ma msa ) = (nb b mb msb ) if a = b. (3.3)
More generally, no two identical fermions (i.e. particles with half integer spin
s = 1/2, 3/2, etc.) can be in the same quantum state. This is an empirically con-
firmed property of fermions. In Sect. 7.3.1 we shall discuss and use the quantum
mechanical formulation: The total wave function of identical fermionsis antisym-
metric in respect of the exchange of two particles. We shall see there that both for-
mulations are completely equivalent.
3.1 Shell Structure of Atoms and the Periodic System 139
Table 3.1 Electron shells Shell States n Number of states

K 1s 1 2
L 2s, 2p 2 8
M 3s, 3p, 3d 3 18
N 4s, 4p, 4d, 4f 4 32
The PAULI principle has profound consequences on the structure of matter.

It accounts for the properties of atoms, for the order of the periodic system,
for the rules that determine how molecules are formed and condensed matter is
built up.
It is completely mind-boggling to think about it in philosophical terms, using the
concepts and experience of everyday life. One might ask: how do fermions com-
municate with each other, how do they know which quantum numbers their equals
occupy? Is the PAULI exclusion principle something like a force? A nonlocal force?
And so on. . . . The strict, formally correct answer is of course: these are simply the
wrong questions and we just have to accept it as a mathematical constraint. In any
case, the PAULI principle is certainly among the most profound concepts in physics
and part of the Miracle of Existence (M ARGENAU 1984).
The counterpart of the PAULI principle states for bosons, i.e. for particles with
zero or integer spin (e.g. for photons or 4 He atoms), that the total wave function of
identical bosons is symmetric in respect of the exchange of two particles which
as such is amazing enough. But there is no constraint with respect to filling theses
states: each respective quantum state may be filled with any number of identical
bosons.
3.1.3 How the Shells are Filled
One defines so called electron shells which are occupied by all electrons of an atom
with the same principle quantum number n. The shells are denoted by the letters K,
L, M, N, . . . , as summarized in Table 3.1. As already noted in Sect. 2.5.3, the letters
s, p, d, f, g stand for the angular momentum quantum numbers = 0, 1, 2, 3, 4. The
number of states in a shell with the principle quantum number n is 2n2 (including
spin states with ms = 1/2) all of which are degenerate in the H atom case.
According to the PAULI principle, each state may be occupied with no more than
one electron. The electron configuration of an atom is written in compact form as
illustrated in Table 3.2 for the ground state of some light atoms. For larger atoms
(e.g. Na) one summarizes the inner shells by the symbol of the next smaller rare gas
atom in brackets [ ].
The Aufbau principle of the periodic system assumes that these shells are sub-
sequently filled as the number of electrons increases. Figure 3.1 illustrates the shell
filling scheme up to neon graphically.
s p s p
L L
K H K He s p s p
L L L L
K Li K Be K B K C
L L L L
K N K O K F K Ne
Fig. 3.1 How the K and L shell of the periodic system of elements are filled with electrons. Arrows
indicate the spin projection (1/2) of the electrons
Table 3.2 Ground state Z Atom Ground state configuration Shell

electron configuration of
some light atoms 1 H 1s K
2 He 1s 2
3 Li 1s 2 2s L
4 Be 1s 2 2s 2
5 B 1s 2 2s 2 2p
... ... ...
10 Ne 1s 2 2s 2 2p 6
11 Na 1s 2 2s 2 2p 6 3s = [Ne]3s M
3.1.4 The Periodic System of Elements
Table 3.3 gives a complete overview of the periodic table of elements, indicating by
colour shadings how the electron shells are filled. The elements within one group
(vertical columns) have equivalent outer shell electrons (differing only by n) and
typically show corresponding similarities in their chemical and physical properties.
Each period (horizontal row) corresponds to the filling of one particular shell and
comprises elements with different electron configuration and usually different prop-
erties. Exceptions are the d block of transition elements, the Lanthanides and the
Actinides, for which deferred inner electron shell filling occurs. The radioactive el-
ements of the 7th period have been (except for Fr and Ra) generated artificially
at heavy ion storage rings (in very low concentration). They typically have mean
lifetimes of only seconds to minutes, and little is known about their electron config-
uration or even about their physical and chemical properties. Nevertheless, they are
interesting objects for fundamental studies.
We refer the reader to numerous, well linked presentations in the internet, such
as W IKIPEDIA CONTRIBUTORS (2014). The source par excellence for further infor-
mation is, however, NIST (2011) from where a host of tabulated properties of the
elements including all available spectroscopic data can be retrieved. Quite instruc-
tive is also the animation from the U NIVERSITY OF C OLORADO (2000).
Table 3.3 Periodic system of the elements. Left on the top of the elements the atomic number is given, below each element the configuration of the last
3.1
built in electron is shown. The filling of different shell is marked by colour shadings: s electrons and p electrons determine the main groups, adding
d electrons leads to the subgroups in the middle of the periodic table that contain all transition metals. The deferred addition of 4f and 5f electrons
occurs for Lanthanides and Actinides
Shell Structure of Atoms and the Periodic System
141
Be Mg Zn Cd Yb Hg
25
3d 104s 24p 6
[Kr]4d 105s 25p 6
[Xe]4f 145d 106s 2

3s 23p 6
[Ar]3d 104s 2
[Kr]4d 105s 2
[Xe]4f 146s 2
20
3s 2
WEA and WI / eV
15
10
0
Z= 0 20 40 60 80
He Ne Ar Kr Xe Rn
Fig. 3.2 Ionization potentials WI (red) and electron affinities WEA (grey) of the atoms as a func-
tion of nuclear charge Z. The full vertical lines indicate shell closures, the dashed lines the closure
of subshells, corresponding to the electron configuration given
The essential basis for the experimentally well confirmed shell structure of the
elements is (beyond the PAULI principle) the astonishingly good model of quasi-
independent electrons in the so called central field approximation. We shall explain
this in detail in Chap. 10 where a profound understanding of the periodic system
will emerge.
3.1.5 Some Experimental Facts
The ionization potentials (IP) WI for all natural elements of the periodic system
are presented in Fig. 3.2. The diagram illustrates impressively the shell nature of
atomic structure: The IPs assume very pronounced maxima for the rare gas atoms
(He, Ne, Ar, Kr, Xe, Rn) that are characterized by closed shell. They are chemically
particularly inert. Smaller maxima are also in between, whenever a subshell for one
particular is completely filled.
A complementary behaviour is seen in the electron affinities WEA (grey line),
i.e. the negative of the binding energies, which may be gained by the formation of
a negative ion (anions) energies. They correspond to the negative binding energies
of the electrons in the anion (WB = WEA ) and are particularly high, if the corre-
sponding neutral atom has just one vacancy in the outer shell for the electron to be
accepted which holds for the H atom and the halogens (F, Cl, . . .) and becomes
zero for the rare gases with their closed shells: no stable rare gas anions are known
to exist (see also Sect. 6.4.5, Vol. 2).
The atomic radii also support the shell structure of atoms very clearly. They are
plotted in Figs. 3.3 and 3.4 as functions of nuclear charge Z. Of course, the term
3.1 Shell Structure of Atoms and the Periodic System 143
Fig. 3.3 Atomic radii as a H Li Na K Rb Cs

function of nuclear charge 0.3
determined by different van der Waals Wigner-Seitz
r / nm
methods: W IGNER -S EITZ
radii (red) and VAN DER
WAALS radii (black) 0.2
0.1
He Ne Ar Kr Xe
Z
Fig. 3.4 Calculated atomic H Li Na K Rb Cs

radii (red), covalent radii 0.3
(black star) and averages
from various binding lengths
(black circle)
r / nm
0.2
0.1
0
He Ne Ar Kr Xe
Z
atomic radius is not very well defined as we have learned in Chap. 2 the size of
an atom is characterized by the probability to find its electrons in a certain distance
of the nucleus, and a limit cannot be defined uniquely. One may e.g. use the so called
W IGNER -S EITZ radius, rWS . That is the radius of a sphere of the same volume that
the atom occupies on average in the condensed phase if such a phase exists for that
particular atom. From particle density N or mass density , relative atomic mass Mr
and AVOGADRO number NA one finds

3 3 3Mr
rWS = = 3 . (3.4)
4N 4NA
A similar quantity is the VAN DER WAALS radius, that gives the distance up to
which chemically not bound atoms can approach each other (for a more precise def-
inition see Sect. 3.2.6, Vol. 2). Both quantities are shown in Fig. 3.3 for the elements
H to Ba.
Alternatively, Fig. 3.4 shows calculated atomic radii. They are extracted e.g. from
quantum mechanical expectation values, as discussed in Sect. 2.6.11 relying on
good computations of the respective atomic wave functions. An empirical deter-
mination is based on so called covalent radii that are derived from a set of well
known binding lengths of molecules, preferably diameters. By comparison with
other molecules one may improve these estimates without big efforts as also shown
in Fig. 3.4 (asterisks). In spite of all ambiguities in defining the term atomic radius
one recognizes very clearly the general trends as a function of atomic number. Rare
gases typically have the smallest radii, alkali metals the largest: electrons in closed
shells all see essentially the same, high charge, while in the outermost shell of an
alkali atom this charge is strongly screened for the one valence electron due to the
core electrons.
We shall discuss this in detail in the following sections of the present chapter.
Before doing so we mention one more little difference between Fig. 3.3 and Fig. 3.4:
while the minima of the atomic radii for rare gases are quite pronounced in the
latter case, such clear cut minima cannot be identified for the VAN DER WAALS and
W IGNER -S EITZ radii. Obviously, in addition to the electron density the influence
of polarizability plays here an important role.
Section summary
According to the PAULI principle each identical fermion in a quantum system
must differ by at least one quantum number.
The periodic system of the elements is based on the PAULI principle. The
PAULI principle prevents electrons from falling all into the energetically low-
est level and thus ensures the variety of the chemical elements. One may say
that it is behind the mystery of life.
The atomic shells are denominated by K, L, M, N, . . . according to their prin-
ciple quantum number n = 1, 2, 3, 4, . . . , respectively. Each shell may contain
up to 2n2 electrons. The Aufbau principle (building up principle) of the peri-
odic system says that electrons fill states with quantum numbers n mms one
after the other (essentially) in numerical sequence.
Chemical and physical properties of the elements depend crucially on the
number of electrons in the outermost shell. Completely filled shells corre-
spond to the rare gases; within each main period they have the highest ion-
ization potential and the smallest radius (as represented e.g. by the W IGNER -
S EIZ radius Eq. (3.4)).
3.2 Quasi-One-Electron System
The most simple cases of multi-electron systems are those where one or several
inner shells are completely filled with electrons (closed shells) and only one electron
is found in the outermost shell. A look at the periodic system in Table 3.3 identifies
these as the elements of the 1st group, the alkali atoms and the alkali like ions. Their
electronic ground state configuration is {[Rg]ns}, where [Rg] stands for a rare gas
configuration of the atomic core, e.g. Li: {[He]2s} Na: {[Ne]3s} K: {[Ar]4s}, etc.
In the following we shall focus on the active electron in the respective new shell,
filled only with this one electron of particular interest. It is called valence electron
(in the original German literature also Leuchtelektron = lighting electron).
3.2 Quasi-One-Electron System 145
Fig. 3.5 G ROTRIAN diagram

for the lithium atom: for the WI = + ion
ns, np, nd, and nf 5.392

configuration of the valence 5
electron the energy terms are 5p 5d 5f
5s
marked by horizontal lines. 4p 4d 4f
4s .7
69
term energy ( Wn - W2s ) / eV

Some optically allowed 4 18
497
transitions (double arrows) 3p 3d
.2
are indicated with
.2
wavelengths given in nm. The 3s
460
3
81
.3
diagram has been generated
2.6
610
from the NIST data bank. The
term energies are related to
2
2
the respective (negative)
323.
2p
binding energies Wn of the
valence electron by
0.8
Wn W2s , with WI = W2s 1
67
being the IP of the system
0
2s angular momentum
3.2.1 Spectroscopic Findings for the Alkali Atoms
The most detailed and precise information comes again from spectroscopy. The
spectra and hence the term schemes of the alkalis turn out to be similar to those
of the H atom with its 1s ground state electron with two important modifications:
(i) the active valence electron is now initially in an ns state with n > 1 and the inner
shells are filled (thus no VUV radiation is observed) and (ii) the degeneracy is
now removed (which makes the spectra richer, i.e. more complicated to interpret,
than that of the H atom).
As a characteristic example Fig. 3.5 shows the term diagram of Li(1s)2 n as
derived from a host of spectroscopic information collected by generations of spec-
troscopists. This type of term scheme indicating the possible transitions is usually
called G ROTRIAN diagram.
The general relation between the energies Wn of stationary states and spectral
lines observed (in wavenumbers) is again
1 1
n n = = (Wn Wn ), (3.5)
hc
now, however, with different energies for each .1
The basics of emission, absorption and fluorescence spectroscopy will be intro-
duced briefly in Sect. 4.2.2 and a variety of more refined modern methods will be
described later, e.g. in Sect. 6.1. Here we only give a survey over the collected re-
1 Strictly speaking, one should even introduce an additional quantum number j for the total angular
momentum, which becomes increasingly relevant as Z gets larger. We have already familiarized
ourselves with it in the context of the S TERN -G ERLACH experiment in Sect. 1.9.5 and shall come
back to it in detail in Sect. 6.2.5.
Wn / E h H Li Na K Rb
ionization continuum
0
n=8
- 0.020
- 0.031 n=4 6p 4f 5f
4s 4p 4d 4f
4s 4p
4d 4f 5s
5p 4d 4f 6s 4d 7s 7p 5d
- 0.056 n=3 4p 3d 5p 6p
3s 3p 3d 3p 3d 6s
3s 4s 5s
4p 5p
- 0.125 n=2 3p
2s 2p 2p
4s 5s
2s 3s
- 0.500 n=1
1s
Fig. 3.6 Overview on the term energies of the alkali atoms in comparison to the H atom. Charac-
teristic is the removal of degeneracy and the decrease of the energies in respect of the H atom for
small angular momenta
sults for the alkalis and refer the interested reader to the spectroscopic data bank of
K RAMIDA et al. (2013) which we have already mentioned.
A comparison of the energies of the H atom with those of all alkali atoms (va-
lence electron) shows Fig. 3.6. The characteristic removal of degeneracy lets the
term energies of the alkali valence electrons always lie below those of the H atom,
Wn < Eh /(2n2 ), owing to the higher nuclear charge Z. However, this lowering of
energy gets smaller as the orbital angular momentum gets larger. Figure 3.6 clearly
documents that Wns < Wnp < Wnd < Wnf . For the nf terms the energies of the al-
kali atoms are practically identical to those in atomic hydrogen. We shall understand
in a moment why that is so.
3.2.2 Quantum Defect
We start by summarizing the experimental findings in a compact form: the just

described similarity with the H atom suggests to write the energies of the alkali
atoms as
Eh
Wn = with n = n . (3.6)
2n2
The parameter is called quantum defect and n the effective quantum number. For
the moment this is simply a parameter, determined experimentally, which allows
one to order the spectroscopic data in a systematic manner.
A comparison of experimental data shows that strongly depends on the orbital
angular momentum and upon closer inspection also slightly on the principle
Fig. 3.7 Quantum defects

(n, ) of the Na atom as a 1.37
Na
(n, )
ns
function of binding energy 1.36
Wn for orbital angular
momenta = 0, 1, 2, 3 and 4. 1.35
Full black circles are 0.89
experimental data taken from np
0.87
K RAMIDA et al. (2013), open 0.85
circles are denoted there as - 4 - 2 - 0.8 - 0.4 0
determined by interpolation
or extrapolation . . . or by 0.014
semi-empirical
calculations. . . . Full lines nd
represent linear least square 0.012
fits to the measured data.
Note the break in energy 0.010
scale for the ns and np series, 0.0015
and the different scale for the nf
nd, nf and ng series 0.0010
0.0004 ng
0.0000
- 1.6 - 1.2 - 0.8 - 0.4 0
Wn / eV
quantum number n. Figure 3.7 illustrates this for Na as example. Obviously, for
large n the quantum defect approaches a constant value that depends only on . We
shall come back to the theoretical interpretation of in the framework of the so
called quantum defect theory (QDT) in Sect. 3.2.6. For a quantitative comparison
one fits the experimentally determined term energies, i.e. the excitation energies
in respect of the ground state, by an extended RYDBERG -R ITZ formula (see e.g.
W EBER and S ANSONETTI 1987):
Eh
term energy Wn = Wn0 0 (3.7)
2(n (n, ))2
with (n, ) = + B/(n )2 + C/(n )4 + (3.8)
or (n, ) + DWn . (3.9)
Here Wn0 0 is the ground state energy, which is related to the IP by WI = Wn0 0 .
The quality of present day spectroscopic data (K RAMIDA et al. 2013) allows a very
precise determination of the parameters. The thus derived quantum defects of all
alkali atoms for large n are summarized in Table 3.4 and Fig. 3.8.
One sees very clearly that the quantum defect increases strongly with the nuclear
charge Z and decreases with increasing orbital angular momentum. We also want
to emphasize that for large n (small |Wn |) decreases indeed linearly with Wn
(not shown here) according to (3.9) which allows for (i) a determination of WI with
spectroscopic accuracy and (ii) analytic continuation of (n, ) (W, ) into the
continuum where W > 0. In the following we shall try to understand these findings
Table 3.4 Quantum defect of the alkali atoms for large n

0 1 2 3 4
Valence electron ns np nd nf ng
Atom Z
H 1 0 0 0 0 0
Li 3 0.40 0.05 0.002 0.00 0.00
Na 11 1.3478 0.8551 0.01489 0.0016 0.00025
K 19 2.180 1.7115 0.2577 0.0013 0.0017
Rb 37 3.1223 2.6535 1.3355 0.017 0.0025
Cs 55 4.0494 3.5916 2.4663 0.03341 0.007
Fig. 3.8 Quantum defect of H Li Na K Rb Cs

the alkali atoms for high n as
a function of . Note the scale 4 ns
quantum defect
change in the ordinate at 0.05.

In particular the f orbitals 3 Z=3 11 19
np
and even more so the g
orbitals are so far away from 2
nd
the atomic core that the
1
quantum defect is close to
zero Z=37 55
0.04 nf
0.02 ng
0.00
1 10 20 30 40 50 Z
Fig. 3.9 Model of an alkali r

atom as quasi-one-electron 1e0
system +Ze0
(Z-1)e0
qualitatively and quantitatively by looking at the potentials and wave functions, and
finally we shall introduce QDT which explains that the quantum defect is not just
some kind of fudge parameter.
3.2.3 Screened COULOMB Potential
A quite reasonable model for the alkali atoms assumes that the Z 1 core elec-
trons (which fill the inner shells of the core ion completely) do influence the energy
terms of the one outermost electron (the valence electron) only by screening the
pure C OULOMB potential of the nucleus. As illustrated in Fig. 3.9, we consider an
N electron problem (nuclear charge Z = N ). At large distances, the valence elec-
2 4 6
0
-1/r VS (r) r / a0
-2
VS (r ) / E h
-4
-6 - Z/r
-8
-10
Fig. 3.10 Effective potential VS (r) in which the valence electron of an alkali atom moves (red),
compared to a fully screened C OULOMB potential 1/r and the unscreened potential Z/r that
the electron experiences close to the nuclear charge Z (here with Z = 11 representing the Na
atom). The grey shaded area indicates the radius of the ionic core
tron sees only the screened nuclear charge 1e and experiences something about
the nucleus only if it dives into the ionic core. One may treat this simplified problem
with nearly the same methods as the H atom except that we no longer have a pure
1/r potential.
Rather we have to solve the problem for a screened Z/r potential (in a.u.)

Z/r r 0
VS (r) = 1/r + VC (r) in between (3.10)

1/r r
where the ionic core is accounted for by a suitable, smooth potential, in the simplest
case e.g.
VC (r) = (Z 1) exp(r/rS )/r (3.11)
as illustrated schematically in Fig. 3.10.
3.2.4 Radial Wave Functions
Thus, the S CHRDINGER equation for this quasi-one-electron model is

2
p
H n m (r) = + VS (r) n m (r) = Wn n m (r). (3.12)
2me
Its solution depends on the spatial coordinates of one electron only. Tacitly we have
also assumed the core potential to be spherical symmetric. Consequently, we may
derive a radial equation in the same manner as done for the H atom, using the ansatz
n m (r) = Rn (r)Y m (, ).
Again we substitute un (r) = rRn (r) and obtain in complete analogy to (2.113) the
radial differential equation (in a.u.)

d2 un ( + 1)
+ 2 W n VS (r) + un (r) = 0 (3.13)
dr 2 2r 2
with the important difference that VS (r) is no longer the pure C OULOMB potential
but rather a suitably screened one as described by (3.10).
The task at hand now is to identify from the infinite manifold of solutions for
(3.13) those which are physically meaningful. They have to reproduce the correct
asymptotic behaviour both for large and small distances r. Since we have derived
the corresponding relations in the case of the H atoms without any reference to
the interaction potential, (2.119) and (2.120) must be valid in the present case too.
r the radial wave function has to be damped exponentially, un
Thus, for
exp( 2|Wn |r), and Wn is the sought-after energy of the stationary states of the
system. On the other hand, for r 0 the radial function has to follow un r +1 .
This very behaviour allows us to explain the experimentally observed depen-
dence of the quantum defect (Table 3.4): according to (2.125) the probability to find
an electron at a distance r from the nucleus is w(r) = 4r 2 Rn 2 (r) = 4u2 (r).
n
Specifically for small distances it is thus r 2 +2 . Hence, the higher its angular mo-
mentum, the less the electron is influenced by the ionic core. For very large the
wave function will be essentially hydrogen like, and the quantum defect correspond-
ingly small, just as documented in Sect. 3.2.2.
For a truly quantitative analysis of wave functions and eigenenergies we have
to integrate (3.13) explicitly. Even though the radial S CHRDINGER equation for
a screened potential of the type shown in Fig. 3.10 cannot be solved analytically,
numerical integration of (3.13) may be done today on any PC without problems.
A number of robust and simple procedures are available, e.g. the often used RUNGE -
K UTTA method. Typically one integrates for a given trial energy Wn from the inside
outwards as well as from the outside inwards, starting with the asymptotic forms just
discussed. One accepts only such solutions that can be matched continuously and
differentiable and by variation of Wn one thus obtains the discrete eigenstates and
eigenenergies of the system.
Of course, the result of such calculations can only be as good as the potential
VS (r) which so far we have only described qualitatively. There are a number of
useful approximations to estimate VS (r). In the most simple approach one guesses a
parametric form which is fitted such that the experimentally observed energies of a
few states are reproduced exactly. With the thus determined empirical core potential
one may derive the wave functions for any state of the system and compute other
quantities, such as the energy terms, optical transition probabilities, polarizabilities
and so on.
3.2.5 Precise Calculations for Na as an Example
Today, the quasi-one-electron model discussed above is, generally speaking, of

mainly pedagogical value and helps us to understand the physical origin of the ex-
(a) w(r ) (b)

1 K-shell
2p 0.3 L- shell of the ion core / 40
-1
0 3s
4 r 2Rn (r) / a 0
0.2
1 2s 3p
2
0
0.1
5 3d
1s
0 0.0
0 1 2 0 5 10 15 r / a0
Fig. 3.11 Radial electron probability distributions w(R) in Na for (a) the core electron orbitals
and (b) orbitals of the valence electron (alternatively in the 3s ground state and in the 3p and 3d
excited states). Also shown is the cumulated radial electron probability in the Na+ core (Z = 11,
closed [Ne] shell), down-scaled by a factor of 40. Grey shaded is the radius of the ion core (1.8a0 ).
The distributions shown by red lines have been calculated ab initio using a finite difference atomic
structure code, FDAlin, conveniently provided by S CHUMACHER (2011). The black dashed and
dash dotted lines are derived from QDT and are not expected to be valid inside the ionic core (see
Sect. 3.2.6). Note in particular the excellent agreement for 3s
perimental observations. With efficient, compact computing programmes and com-

puters available, one may calculate energy eigenvalues, wave functions, transition
probabilities and other properties of the alkali atoms with nearly unlimited preci-
sion.2 There is no need to restrict such calculations to the valence electron, and the
wave functions of all filled orbitals for the alkalis may be obtained readily. Some de-
tails of appropriate methods will be discussed in Chap. 10 in the context of genuine
multi-electron systems.
However, the general character of the orbitals for each individual electron re-
mains very similar to that of the corresponding electron orbital in atomic hydrogen.
Figure 3.11 shows by way of example the probability distributions for the electrons
in Na (electron ground state configuration 1s 2 2s 2 2p 6 3s). They have been computed
with a simple DFT programm that generates these data on any PC within seconds
(the principles of DFT will be introduced in Sect. 10.3). Shown in Fig. 3.11(a) are
the radial probability distributions (integrated over all angles) for the orbitals of the
core electrons (1s, 2s, 2p). The grey shaded areas indicates again the literature value
for the ion core radius (0.95 nm = 1.8a0 ).
Figure 3.11(b) shows the corresponding probability density for the valence elec-
tron in the 3s ground state (alternatively
3p or 3d). In addition, the cumulated radial
electron probability w(r) = core Nn 4r 2 Rn 2 (r) in units [w] = 1/a is given
0
for the ion core with Nn being the number of electrons in the respective n states,
i.e. 2 for the K shell (1s 2 ), and 2 + 6 for the L shell (2s 2 2p 6 ). This cumulated dis-
tribution for the ionic core clearly exhibits the different shells and underlines again
2 This holds at least for the smaller alkali atoms where relativistic effects play a minor role and
spin-orbit interaction can be treated as a small perturbation.

quantitatively that the valence electron stays predominantly outside the core. This
holds a fortiori for excited states, here very clearly seen for the 3d state the orbital:
its radial wave function (R3d (r) r 2 for small r) leads to very little probability for
finding the electron inside the core as Fig. 3.11 illustrates.
This explains again convincingly why the quantum defect decreases so rapidly
with as illustrated in Fig. 3.8. It also clarifies the overall striking similarity of the
alkali spectra with those of the hydrogen atom which we have seen in Fig. 3.6. The
nf electrons, for which at the origin R4f (r) r 3 and thus w(r) r 8 , practically
never come close to the nucleus and the quantum defect can essentially be neglected
as for all higher values of (see Fig. 3.8).
We recall again: degeneracy observed in H and H like ions is just a very specific
consequence of the pure C OULOMB potential. Deviations from the C OULOMB po-
tential lead to different energies for different , i.e. to a removal of this degeneracy.
The deviations are the larger the more the electron feels of the ionic core. For very
large the situation is practically identical to the pure 1/r potential in the H atom
case.
3.2.6 Quantum Defect Theory
Even though in principle the full quantum mechanical problem of bound states in
small alkali atoms may be treated numerically, it is useful to return once more to the
model of a single electron in an effective potential. So far, we have not yet under-
stood why the extended RYDBERG -R ITZ formula works so well with essentially
constant quantum defect (3.8) for large n as documented by Fig. 3.7. This remark-
able observation hints at some genuine physical background, even though the ef-
fective quantum number n and the quantum defect were introduced in (3.6) just
as useful empiric parameters. It turns out that understanding this physical origin of
the quantum defect allows one to extract useful concepts and further quantitative
information.
The roots of the so called quantum defect theory (QDT) go back to the early
days of quantum mechanics and H ARTREE (1928), while between 1950 and 1990
S EATON (1983), FANO (1986), J UNGEN (1996) and their students have developed
it into a powerful theoretical framework and extended it to multichannel problems
(MQDT). It has been applied successfully to calculate oscillator strengths for atomic
and molecular transitions including high lying RYDBERG states, to determine accu-
rate ionization potentials and photoionization cross sections (see also Sect. 5.5), to
understand autoionizing series and perturbations in multi-electron spectra, as well
as in electron ion scattering theory and is used even in solid state structure the-
ory.
The key to QDT is that for large r far outside the atomic core electronic wave
functions evolve in a pure C OULOMB potential and may be derived as analytic func-
tions also for non-integer quantum numbers with the correct damping at large r
(in a.u.)

un (r) r n exp 2|Wn |r = r n exp r/n , (3.14)
r
Wn being the binding energy and n the effective quantum number.3

Effective range theories for collisions (which will briefly be covered in Sect. 6.4.5,
Vol. 2) may be viewed as a generalization of QDT for the non-C OULOMB case. Re-
cently, QDT has attracted renewed interest in the context of collisions and reactions
with ultra-cold atoms and molecules (e.g. O SPELKAUS et al. 2010; I DZIASZEK and
J ULIENNE 2010, and references there).
As an introduction we take a quantitative look at the radial wave functions for
large principle quantum numbers n where (n, ) is essentially independent of n.
We solve (3.13) numerically to obtain un (r) = rRn (r), using the S CHRDINGER
applet from S CHMIDT and L EE (1998). As specific examples we choose the 18s
and 20s states and use the screened potential VS (r) according to (3.10) shown in
Fig. 3.10.4 The screening parameter was chosen rS = 0.4190a0 to yield the experi-
mentally determined quantum defect = 1.348. In Fig. 3.12(a) we show for com-
parison the exact radial wave functions for atomic hydrogen, in (b) the computed
wave functions of n = 18 and 20 in Na are plotted.
For r > 0 the wave functions u18s (r) and u20s (r) in both cases (H and Na) have
of course the same number of nodes (crossings through zero), n 1 = 17 and 19
for 18s and 20s, respectively. The probability distributions of these wave functions
lies almost entirely outside the ionic core (Na+ core radius is 0.095 nm = 1.8a0 ).
At intermediate r they show a pronounced oscillatory character
resembling effec-
tively sin(kr r) or cos(kr r) functions, with kr = 2/r T = |VS | |Wn | cor-
responding essentially to the DE B ROGLIE wavelength r for the local kinetic en-
ergy T .
We also see (in the intermediate r range) that a change from n n 2 corre-
sponds approximately to a shift by 1 oscillation (counting from the last maximum)
downward or a phase shift of 2 when extrapolated into the continuum. There,
for large r one might essentially expect a sin kr or cos kr like behaviour (see, how-
ever, the more precise discussion below). Most important in the present context:
the wave functions for Na are phase shifted towards smaller r in respect of the
strictly C OULOMB wave functions for H clearly by more than 1/2 of the oscil-
latory period as indicated by the red dotted lines and the red arrow. Extrapolated
into the continuum this translates into a phase shift of = as we shall see in a
moment.
The physical origin of this phase shift is recognized in the blow ups Fig. 3.12(c)
for H and (d) for Na: Even though the electron comes only very rarely close to the
atomic core (grey shaded region), due to the strong attraction of the nucleus in Na
(Z = 11) its wave function oscillates there much more rapidly than for the H atom
one may attribute this simply to the much shorter DE B ROGLIE wavelength r in the
3 The r n factor used here in contrast to (2.119) can improve the convergence.
4 This is a rather crude choice. It leads, however, to qualitatively correct wave functions. The radius
of the ionic core for Na+ is typically given in the literature as 0.095 nm = 1.8a0 . At this distance
this screened potential VS (r) is about 1.1/r.
(a) (c)
0.6 0.2
H 18s
r Rn / arb. un.
0 0
H 20s
-0.2
0 10 20
50 H 18s H 20s
-1.0
Na+ core
(b) (d) Na 18s
0.6 0.2
Na 20s
0 0
-0.2
0 10 20
50 Na 18s Na 20s
-1.0 r / a0
0 500 1000
Fig. 3.12 Illustration of the quantum defect as phase shift = for Na in the 18s and
20s states: (a) The radial wave functions rRn (r) for a pure C OULOMB potential (H atom) are
compared (d) to those in an Na+ pseudopotential VS (r) (see Fig. 3.10). All wave functions are
normalized to their minimum, which is set to 1. The blow ups of the r scale (c) and (d), respec-
tively, illustrate the strong influence of the Na+ ionic core (grey shaded area) onto the Na wave
functions. It leads to a shift of the rRn maxima in respect of the H atom; this shift is directly
related to the quantum defect (n, )
deeper attractive potential VS . While the Na 18s and 20s wave functions are nearly
identical in this small r range (apart from their magnitude owing to normalization),
they differ dramatically from their counterparts in the H atom. Outside the core the
potential is purely C OULOMBic for both Na and H hence the two pairs of wave
function differ essentially only by this phase shift .
QDT treats the outside wave functions as a linear superposition of two pure
C OULOMB functions. The mathematics involved is not completely trivial and
we can only outline here just a few basics, following the excellent review of
S EATON (1983) and his concise mathematical summary from 2002. The radial
S CHRDINGER equation (3.13) is solved in two parts:
"
FI (r) for 0 r < r0 with VS (r)
u(r) = (3.15)
FII (r) for r0 r with 1/r.
(If alkali like ions are discussed with a remaining charge ZC of the ionic core, the
1/r potential has to be replaced by ZC /r.)
Both functions must be joined at a suitable distance r0 with continuous loga-

rithmic derivatives. The first part FI (r) requires either some experimental input or
a numerical solution, but FI is insensitive to the total energy W as long as |W |
|VS (r)|; Fig. 3.12(d) may serve as illustration.
The second part of (3.15) can be solved fully analytically. One uses a scaled
energy and redefines a scaled radial distance r:
= 2(W/Eh )/ZC2 and r := ZC r/a0 . (3.16)
With this (3.13) is rewritten as

" #
d2 ( + 1) 2
+ + u = 0. (3.17)
dr 2 r2 r
Since this is an ODE of 2nd order it has two linearly independent sets of C OULOMB
functions as solutions for which one finds various forms in the literature. For QDT
S EATON (1983) defines s(, ; r) as the regular solution, and c(, ; r) as the irreg-
ular solution. When treating the H atom in Sect. 2.6.5 we have considered only the
regular solution un s(n , ; r) for n = 1/n2 (n being an integer); asymptoti-
cally they are r +1 for r 0 and r n exp(r/n) for r . It is important to
note that for all other values of the regular solution s(n , ; r) diverges at large r
hence the n give the eigenvalues of the H problem.
The irregular solutions c(, ; r) are r at the origin so that any radial
wave function R(r) rc(, ; r) would diverge hence they were considered non-
physical for the H problem. It is interesting to note, however, that for large r and
n = 1/(n + 1/2)2 (and only for these) they are damped r n exp(r/n).
Hence, it is plausible as shown already by H ARTREE (1928) that at arbitrary
energies = 1/n2 a physically meaningful solution of the problem for r > r0 is:

FII (r) = cos n s(, ; r) + sin n c(, ; r)
(3.18)
= (1)n cos()s(, ; r) + sin()c(, ; r) .
Here n is the effective quantum number and = n n is the quantum defect as

defined by (3.6) and more specifically by (3.7).
These superpositions of regular and irregular C OULOMB function for a given
quantum defect are used in QDT for r > r0 . In general they cover most of the
interesting part of the wave function as documented in Fig. 3.12. For not too low
n they are fully sufficient to calculate matrix elements of r and hence transition
probabilities, as we shall discuss in Chap. 4. Typically, in the Na case one would
choose r0 1.8a0 , corresponding to the Na+ ionic core radius.
Generally speaking, C OULOMB wave functions are special cases of confluent
hypergeometric functions which are derived from (3.17), possibly recast into various
other standard mathematical forms.
Bound States
Specifically for bound states = 1/(n )2 < 0, S EATON shows

sin(n ) n3 1/2
s(, ; r) = (1) cos n K
(2)1/2 K 2

(3.19)
n3 1/2
cos(n )
c(, ; r) = (1) + sin n K
(2)1/2 K 2
with the normalization factor

1/2
K n , = n2 n + + 1 n (3.20)
where (x) is the Gamma function (with (n 1) = n!).

2
n , ; r = Wn , +1/2 (3.21)
n
is the a so called W HITTAKER W function, while (n , ; r) is a linear combination

of two such functions. The asymptotic behaviour for large r is
n n
r r r r
exp and exp . (3.22)
r n n r n n
For integer values n = n the regular functions are orthonormalized:

s(n , ; r)s(n , ; r)dr = n3 /2 nn . (3.23)
0
Since n = (n+1 n ) 2/n3 in the limit of large n, the bound state functions
s(n , ; r) are said to be normalized per unit energy in this limit.
Since the component of the solutions (3.19) diverges for large r it is not suit-
able for the description of any realistic physical wave function. And indeed, when
inserting (3.19) with (3.21) into the general solution (3.18) the terms cancel and
one obtains for r > r0 simply:
3 1/2
n 2r
FII n , ; r = (1) K n , Wn , +1/2 . (3.24)
2 n
This is the sought-after general wave function in a pure C OULOMB potential outside
the atomic core for arbitrary energy! The W HITTAKER functions are included in
advanced mathematical programmes, e.g. in Mathematica, and we have computed
these functions for the 18s and 20s case of Na with n = n 1.3848. The result
is completely identical to the wave functions shown in Fig. 3.12(b) and (d) within
the limits one could see in the graphs except inside the ionic core where it is not
expected to be valid.
As a second example, illustrating the power as well as the limitations of QDT,

we have computed the 3s, 3p and 3d functions of Na in the same manner, using
the precise (experimentally determined) values (3s) = 1.37289, (3p) = 0.88283
and (3d) = 0.01023 derived from the K RAMIDA et al. (2013) data. The results for
the electron densities FII2 (r) are plotted in Fig. 3.11(b) as dashed and dash dotted
black lines and can be compared with the exact results from the DFT calculations.
The agreement is astonishingly good (and we cannot discuss here potential short-
comings of the DFT method). Clearly, the FII (r) wave functions are not expected
to be relevant inside the ionic core, where 3p and 3d are seen to diverge. But with
a suitable cutoff one may confidently use such type of wave functions (even for
low n), e.g. for calculating transition probabilities, polarizabilities, and other prop-
erties of the alkali atoms: QDT appears to lead to wave functions that represent most
of the significant r range very well.
One final brief and potentially useful remark (see e.g. F REEMAN and K LEPPNER
1976): A small change in the quantum defect (say by n ) changes the term
energies according to (3.7) by

1 1 1 1
Wn = 2 + = 1 3 . (3.25)
2n 2(n + )2 2n2 (1 /n )2 n
Conversely, any perturbation Wn known to scale (e.g. for the H atom) with the
principal quantum number as 1/n3 may be accounted for by a corresponding addi-
tion to the quantum defect. As we shall see in later chapters, a variety of important
perturbations that originate from interactions inside the ionic core do indeed scale
like 1/n3 , e.g. fine and hyperfine structure splitting, certain effects from interactions
with an electric fields, polarization and so on. They may all be treated as additive to
the quantum defect as long as n . Hence, also the corresponding changes of
the wave functions may be described by QDT as explained above.
Continuum States
We have already indicated that the behaviour of continuum wave functions may be
gleaned from analytic continuation of the quantum defect (n, ) (, ) for =
k 2 > 0. What we have vaguely addressed as phase shift between the oscillations
of the bound state wave functions will then become indeed a genuine phase shift
() = (, ) (3.26)
between a pure outgoing C OULOMB wave and the continuum wave function dis-
torted by the screened Z/r potential of the ionic core. Here too, (3.18) applies for
the outside part of the wave function, FII , and at a suitable r0 one has to join FI
and FII . Again various forms for the regular and irregular pure C OULOMB contin-
uum wave functions are used. According to S EATON (2002), Eqs. (86)(90) and
(113)(118)
' (
s(, ; r) Im exp i( /2) Wi/km +1/2 (2ikr)
' ( (3.27)
c(, ; r) Re exp i( /2) Wi/km +1/2 (2ikr)
0.4
s (, ;r)
0.0
c( , ;r)
-0.4
0 5 10
r
Fig. 3.13 C OULOMB continuum s waves ( = 0) at = 4 for attractive potential: regular solution
(red line, zero at origin), irregular (black line, finite at origin) adapted from S EATON (2002). The
asymptotic behaviour (3.30) (pink and grey dashed lines, respectively) gives a nearly perfect match
to the exact solutions already at r 3
where Wi/km +1/2 (2ikr) is again a W HITTAKER W function, this time to a com-
plex argument, and for the presently considered attractive C OULOMB potential we
have r > 0. The C OULOMB phase shift
= (k, ) = arg (1 + i/k) (3.28)
decreases rapidly with k and corresponds to the phase difference of the C OULOMB
in respect of a free spherical wave (not to be confused with the phase shift () due
to the ionic core potential VS being different from purely C OULOMB).
We have suppressed here a discussion on the important question of normalization
and simply note that

s(1 , ; r)s(2 , ; r)dr = (1 2 ), (3.29)
0
i.e. s(1 , ; r) is normalized to scale (see Appendix J). Again, the regular solution
goes to 0 at the origin while the irregular solution stays finite or diverges. For large r
these wave functions are essentially outgoing spherical waves apart from a slowly
varying logarithmic phase shift characteristic for the C OULOMB potential (see (J.8))
i.e. they behave essentially like sine and cosine functions:

2 2
lim s(, ; r) = sin and lim c(, ; r) = cos (3.30)
r k r k
1
with = kr + ln(2kr) /2 + (k, ). (3.31)
k
Figure 3.13 illustrates these continuum functions for = 0 at a moderate energy =
4 or k = 2 where 0 0.078 102 . The exact solutions (adapted from S EATON
(2002), Fig. 1) are apparently matched rather perfectly by their asymptotic form
(3.30) already at surprisingly small values of r 3.
We now come back to the general QDT theme of constructing a wave function
for r > r0 , this time in the continuum, using the analytic continuation () de-
rived from bound state quantum defects. Being interested in large r we insert the
asymptotic functions (3.31) into (3.18) and obtain for r > r0

2
FII (, ; r) sin + () (3.32)
r k
with the phase shift () according to (3.26) that just describes the shift of the
maxima between the regular continuum C OULOMB wave and the wave function of
the system with quantum defect very similar to the bound state situation illus-
trated in Fig. 3.11, except that now the phase shift is a genuine phase between the
sin[ ] in a clean 1/r potential and sin[ + ()] influenced by the screened ionic
core. We shall return to these types of outgoing spherical waves in the context of
photoionization in Sect. 5.5.4 and collision theory in Chaps. 6 and 7, Vol. 2.
3.2.7 MOSLEY Diagrams
It is often advantageous to use an alternative empirical description of the energy

terms that supports the comparison of different elements within a row of the periodic
system. Instead of using an effective quantum number n one introduces an effective
nuclear charge Z , thus accounting for the fact that electrons see only a fraction
of the nuclear charge. The quantity qs = Z Z is then called screening parameter.
The binding energies of the electrons thus become
Z 2 (Z qs )2
Wn = 2
Eh = Eh , (3.33)
2n 2n2
or in wavenumbers
(Z qs )2 (Z qs )2
n = E h /2 = R
hcn2 n2

2|Wn | n Z Z qs
= = = . (3.34)
Eh R n n
Plots of the square roots of the energies as a function Z are called M OSLEY di-
agram and should give according to (3.34) a straight
line. The screening parameter
qs is then obtained from the axis intercept of 2|Wn |/Eh where Z qs = 0. Full
screening of the C OULOMB potential would imply qs Z 1. By way of exam-
ple we discuss the energy terms for Na and Na-like ions with data taken from the
K RAMIDA et al. (2013) data bank. The series starts with Z = 11 (Na) and continues
with Mg, Al, Si, P, S, Cl, etc.
Figure 3.34 presents the data for the ns, np and nd terms with n = 3, 4 and 5.
The graph documents that indeed the square root of the energies follows to a very
good approximation a straight line as a function of the nuclear charge Z as predicted
by (3.34). Admittedly we still need two parameters, qs and n n to fit a whole
series well. As expected, for states with the highest orbital angular momentum the
4
3s
3p
Z-10.03
3d
3 2.91
4s
2Wn / E h 2 4p
Z-10.02 4d 5s
3.90 5p
1 5d
Z-10.02
4.97
0
10 11 12 13 14 15 16 17 18 19 20 Z
Na I Al III P V Cl VII Ca X
Fig. 3.14 M OSLEY diagram for Na like ions and different n states. Following spectroscopic
traditions we designate the degree of ionization with roman numbers: I for neutral atoms, II for
singly ionized, III for doubly ionized atoms, etc. Symbols correspond to the experimental data
according to K RAMIDA et al. (2013), the lines are fits following essentially (3.34). Nearly perfect
screening (qs = 10) is shown for the d states
energies agree best with the straight lines as validated in Fig. 3.14 particularly well
for the 3d, 4d and 5d series. The screening in all three cases is nearly perfect (qs =
10) and die slope of the lines is almost ideally 1/n. For the p and s states the
screening is obviously not perfect: in particular for the 3s states the screening is
only just above qs = 9, i.e. the electron still sees effectively nearly two charges of
the nucleus. In any case, the surprisingly good agreement of experimental data with
the prediction (3.34) supports very impressively the model of a quasi-one-electron
system over a whole isoelectronic series.
For higher Z the lines typically fall into the X-ray region. M OSLEY diagrams are
therefore also used successfully to characterize X-ray spectra from inner shells (see
Sect. 10.5.2).
Section summary
The spectra of alkali atoms resemble those of the H atom in the VIS except
that now degeneracy is removed and consequently several series of lines are
observed for each principle quantum number n.
Thus, a good concept to understand these spectra is a quasi-one-electron
model where all but one active electron (the valence electron) constitute the
closed shell of the ionic core. They (partially) screen the charge of the nucleus
but otherwise do not participate in standard spectra.
The energy terms of the alkali atoms (and alkali like ions) depend thus
on n and of the valence electron and may be described by Wn =
Eh /[2(n )2 ]. The quantum defect (n, ) depends strongly on and
weakly on n. For large n it becomes nearly independent it. The higher , the
smaller is , i.e. the closer the energy levels are to those of the H atom (and
H like ions, respectively).
3.3 Perturbation Theory for Stationary Problems 161
These findings are explained by the low probability of the valance electron to
be found close to the nucleus, being r 2 +2 for r 0.
Quantum defect theory (QDT) gives a very useful analytic foundation to the
observed behaviour of quantum defects: it turns out that the quantum defect
relates to the phase shift = between a pure C OULOMB wave function
(H atom) and the wave function for the alkalies.
M OSLEY diagrams summarize the spectral lines of a whole series of ions with
equal electron configuration as a function of the nuclear charge Z. Plotting

2|Wn |/Eh vs. Z leads to essentially straight lines.
3.3 Perturbation Theory for Stationary Problems

3.3.1 Perturbation Ansatz for the Non-degenerate Case
As we have seen, numerical integration of the S CHRDINGER equation is straight

forward at least for a quasi-one-electron problem. Nevertheless it is important not
to rely completely on the black box of powerful computer programs. As we have
just seen from QDT by way of example, alternative approaches may give profound
new insight into the underlying physics based on experimental observations.
An even more general approach, which in many cases helps to develop a genuine
understanding, is to view an additional interaction as a perturbation on top of a
problem that has already been solved. We thus use the opportunity to supplement
the above discussion by introducing an important quantum mechanical tool that will
be used frequently in the following chapters: time independent perturbation theory.
Instead of solving a problem by direct integration of the S CHRDINGER equa-
tion as described in Sect. 3.2.5, one may try to estimate the changes expected when
slightly changing the interactions in a given problem for which the solution is
known. The basic idea is to consider the already solved problem (in the present
example the H atom wave functions and energies) as 0th order solution for the mod-
ified problem (here the system with a potential changed from 1/r to a screened
Z/r potential). If that perturbation is not too large one may hope that the new so-
lution may not be too much different from the 0th order solution and can be derived
from the latter.
We summarize here the procedures in recipe like manner and refer the readers
to standard textbooks on quantum mechanics for a rigorous presentation of the un-
derlying perturbation theory. So, let the Hamiltonian for the original problem (here
the H atom) be H 0 , the eigenstates (0) (a complete basis), and the corresponding
k
(0)
eigenenergies Wk . The 0th order S CHRDINGER equation
0 (0) = W (0) (0)

H (3.35)
k k k
is assumed to be solved. Let the Hamiltonian for the new problem be given by
=H
H (r, p
0 + U ), (3.36)
with the perturbation operator U (in the most simple case just a scalar potential).
For perturbation theory to be applicable we have to assume that for the averaged per-
| |H
turbation |U 0 | holds. We thus write the S CHRDINGER equation (3.12)
in the form
0 + U
(H )k = Wk k , (3.37)
which has to be solved by the perturbation approach. One then expands the ener-
gies and wave functions into a series in terms of small quantities Wk and k
assuming that U is also small on the order of . The formally correct procedure is
to compare quantities of the same power in the smallness parameter and thus to
obtain correction terms of the order , 2 , etc. These are the corrections in 1st, 2nd,
etc. order perturbation theory. The key quantities characterizing the perturbation are
the matrix elements:

(0) (0)

Uj k = j U k = j(0) (r)U (r, p)k(0) (r)d3 r. (3.38)
3.3.2 Perturbation Theory in 1st Order
We abbreviate the procedure just outlined somewhat. Since the eigenfunctions i(0)
of the unperturbed Hamiltonian H 0 , derived from (3.35), form a complete, orthonor-
mal basis set, any function may be constructed as linear superposition from them
including the solution of (3.37). Thus, we use the perturbation ansatz
(0)
k = ai i with |ak | 1 and |ai | 1 for i = k. (3.39)
i
The conditions |ak | 1 and |ai | 1 are crucial; they establish the essence of the
perturbative approach.
Introducing (3.39) into the S CHRDINGER equation (3.37) we obtain

0 + U
[H Wk ] ai (0) = 0
i
i

0
H
(0) ai (0) Wk ai (0) = 0
ai i + U i i
i i i
(0) (0)
ai (0) = 0.
ai W i W k i + U (3.40)
i
i i
In the last step we have made use of 0th order solution (3.35). Next, (3.40) is multi-
(0) (0) (0)
plied from the left with k and integrated over all space. With k |i = ki
one obtains
(0) (0) (0)
ak Wk(0) Wk + ak k(0)
U k + ai k U i = 0.
i=k
Since |ai | 1 for i = k, and since the matrix elements of the perturbation operator
are small too, one can neglect to 1st order the whole sum over i = k in this
U
expression. With this ak may be factored out, and one arrives immediately at the 1st
order correction for the energy
(0) (0)
W = Wk Wk = Ukk = k U k ,
(0)
(3.41)
determined by the diagonal matrix element of the perturbation operator.

In order to also derive the correction for the wave functions, one multiplies (3.40)
(0)
from the left with j for j = k and integrates:
(0) (0) (0)
0 = aj W j W k + ai j U i .
i
Inserting the 1st order solution (3.41) for the energy correction this becomes
(0) (0) (0) (0) (0)
0 = aj Wj Wk k U k ai j U i .
(0)
+
i
If again one neglects all terms that are quadratically small, the third term can be
dropped and of the sum only the term with i = k remains (since ak 1), so that
(0) (0) (0) (0)
0 = aj Wj Wk + j U k ,
from which finally

(0)|
j |U
(0)
k
aj = (0) (0)
for j = k
Wk Wj
follows. Thus, the wave function in 1st order perturbation theory is
| (0)
j(0) |U Uj k
(0) k (0) (0) (0)
k = k + (0) (0)
j = k + (0) (0)
j . (3.42)
j =k Wk Wj j =k Wk Wj
3.3.3 Perturbation Theory in 2nd Order
For the next step we insert the results of 1st order perturbation theory again into
the S CHRDINGER equation (3.37) and repeat the whole procedure. The 2nd order
correction for the energy is thus derived as follows:
| (0)
j(0) |U
k(0) H0 (0) + k (0)
Wk = +U k (0) (0)
j
j =k Wk Wj
| (0)
j(0) |U
(0) (0) (0) (0)
k k
(0) k
= Wk + U k + (0) (0)
U j
j =k Wk Wj
(0)
|Uj k |2
Wk = Wk + Ukk + (0) (0)
. (3.43)
j =k Wk Wj
| (0) are the matrix elements of the perturbation. The pro-

Again, Uj k = j(0) |U k
cedure may in principle be repeated as often as necessary. A few points need to be
considered:
1. We see that many states (in principle an infinite number) may contribute to the
determination of wave functions as well as of energies. Practical considerations
will limit the efforts, and one has to choose judiciously which states are included
in a perturbation expansion.
2. The contribution of individual states j depends both on the magnitude of the
(0) (0)
matrix elements |Uj k | and on the resonance denominators 1/(Wk Wj ). The
closer a the energy of a perturbing state j to that of state k which is perturbed,
the stronger the modification.
(0) (0) (0) (0)
3. If several states 1 , 2 , 3 are degenerate, i.e. have the same energy W1 =
(0) (0)
W2 = W2 , one has to be careful with the perturbation approach because of
the resonance denominators. Only if the non-diagonal matrix elements of the
perturbation disappear, i.e. only if U12 = U13 = U23 0, one may follow the
perturbation procedure just described.
3.3.4 Perturbation Theory for Degenerate States

In the more general case with g fold degeneracy and nonvanishing off-diagonal
matrix elements, the problem has to be approached in a more general manner. For
clearness we write the Hamiltonian in matrix form with
H 0 + U
j k = j |H |k = W (0) j k + Uj k
k
(0)
W1 + U11 U12 U13 ... U1g
(0)
U 21 W + U 22 U23 ... U2g
=
2
H U31 U21
(0)
W3 + U33 ... U3g . (3.44)

... ... ... ... ...
(0)
Ug1 Ug2 Ug3 ... Wg + Ugg
k = Wk k thus becomes a matrix eigenvalue
The S CHRDINGER equation H
equation,
W
(H 1)| = 0, (3.45)
with the g g identity matrix
1, W an eigenvalue, and | is now an eigenstate
vector with g components (which may refer to the basis of the unperturbed system).
In other words, we have now to solve a set of g linear algebraic equations, which
can lead to a set of g eigenvalues W and state vectors. Generally speaking, one
has to diagonalize the Hamiltonian matrix by a unitary transformation. The matrix
elements of the diagonalized matrix are the eigenvalues of the system.
As well known, a solution of (3.45) only exists if the determinant of the matrix
vanishes. Thus, one first has to find the solutions of
W
det(H 1) = 0. (3.46)
In general this leads to a nonlinear equation of degree g in W with up to g solu-

tions Wk , the eigenenergies of the perturbed system. With these one can go back
into (3.45) and solve the equations for each value Wk to find the eigenstates of the
system.
We note that the procedure described here does not require the perturbation to be
small. It does not even need the 0th order states to be degenerate (which would imply
(0) (0) (0)
W1 = W2 = = Wg ). The procedure may ultimately be used universally
as long as the perturbation matrix Uij is sufficiently well known. The accuracy of
the solution finally depends only on the number and artful choice of the basis states
included in the computation. We shall encounter many applications in later chapters.
3.3.5 Application of Perturbation Theory to Alkali Atoms
To illustrate how to use 1st order perturbation theory, we briefly return to the alkali
atoms as a simple example. We emphasize the pedagogical aspect of this subsection
that is not meant to generate accurate numbers for energy levels. With the interaction
potential (3.10) discussed in Sect. 3.2.3 the Hamiltonian of the perturbed problem
(in a.u.) is written
= 1 2 + VS (r) = H
H 0 + VC (r) with VS (r) = 1/r + VC (r), (3.47)
2
and we recognize
0 = 2 /2 1/r
H
being the unperturbed Hamiltonian of the H atom.
The spherically symmetric perturbation potential U (r) = VC (r) originates from
the interaction of the valence electron with the full nuclear charge Z at small r and
screening at larger distances. For demonstration we use again the particularly simple
model potential VC introduced in (3.11). Figure 3.15 illustrates this potential for Na
(Z = 11). The screening radius has now been calibrated, as described below, to fit
the experimentally determined ground state binding energy W3s .
The matrix elements of the perturbation (3.11) are computed with the hydrogen
(0)
eigenfunctions Rn (r)Y m (, ) (0th order solutions), which have been treated in
detail in Sect. 2.6.1. Since the perturbation potential is spherically symmetric and
Fig. 3.15 Core screening in W / Eh

Na with rS = 0.98a0
0 1 2 3 4 5
W3s
r / a0
-2
-4
VC (r) =
-6 (Z 1) -r/rS
e
-8 r
Table 3.5 Binding energies Levels Binding energies / eV

for some levels of the s, p,
and d series in Na. exp. 1st order WH
Experimental data (exp.) are 3s 5.13907 5.139 1.5117
compared with 1st order 3p 3.0357 2.70 1.5117
perturbation theory (1st
order). WH are the 3d 1.5221 1.63 1.5117
corresponding energies for 4s 1.9477 1.87 0.8503
the H atom. For details see 4p 1.3864 1.30 0.8503
text
5s 1.0227 0.92 0.54422
5p 0.7951 0.74 0.54422
6s 0.6294 0.55 0.37794
acts only on the radial part of the wave function, the matrix elements simply are

(0) (0)
VC n m,n m = mm VC (r)Rn (r)Rn (r)r 2 dr = VC n ,n . (3.48)
0
In 1st order perturbation theory only the diagonal matrix elements are needed. We
(0)
make explicit use of the radial wave functions Rn (r) of the H atom as given in
(2.122), using the series expansion (2.123) for the L AGUERRE polynomials. The
numerical integration of (3.48) can be done with standard desktop computing pro-
grammes within seconds.
(0)
Inserting the binding energies Wn = 1/2n2 of the hydrogen atom into (3.41)
and noticing that with VC (r) < 0 the matrix elements are also negative, the energies
of the alkali atoms in 1st order perturbation theory finally are (in a.u.)
(0) 1
Wn = Wn + VC n ,n = |VC n ,n |. (3.49)
2n2
All energy terms are indeed lowered in respect of the H atom as experimentally
observed.
In Table 3.5 we communicate some numbers that we have derived in this man-
ner for Na and compare them with the experimental data from the NIST data bank
(K RAMIDA et al. 2013). For comparison, H atom binding energies are also shown.
The data illustrate quite clearly what may be achieved with such a simple approxi-
mation. Of course, the results can only be as good as the perturbation potential that
describes the additional interaction and one cannot expect miracles from (3.11) with
only one free parameter a rather crude guess of the perturbation potential. An ad-
ditional problem is, that the wave functions experience a strong phase shift in the
ionic core, as we have seen in Sect. 3.2.6. This is of course not taken into account
in 1st order perturbation theory where the energy is computed from 0th order wave
functions for the H atom.
In view of these difficulties, the results shown in Table 3.5 may be considered
reasonable. But it becomes also obvious that 1st order perturbation theory should
only be used for a first orientation in such problems with several competing influ-
ences. If precision is demanded one has to resort at least to the methods outlined in
Sect. 3.2.
Section summary
Perturbation theory can often give important qualitative and semi- quantitative
insights into the physics expected from exposing atoms or molecules to spe-
cific interactions as long as the averaged perturbation is small compared to
| |H
the Hamiltonian defining the basic interaction in the system, |U 0 |.
In 1st order perturbation theory the change of energy (of non-degenerated
(0) (0)
states) is W = Ukk = k |U |k , while the 1st order wave function is
given by (3.42).
Higher order approaches are needed if first order (diagonal matrix elements)
vanishes.
For degenerate states a different approach has to be taken. One expresses the
perturbed Hamiltonian in matrix form (3.44), using as many and as good basis
functions as appropriate. One then has to diagonalize this matrix, using the
standard procedures of linear algebra. The procedure can be extended to quite
general problems, providing suitable basis functions are available.

IP: Ionization potential, of free atoms or molecules (in solid state physics the
equivalent is called workfunction).
MQDT: Multichannel quantum defect theory, advanced form of QDT for the in-
terpretation of complex atomic and molecular spectra, especially of highly ex-
cited RYDBERG states (see Sect. 3.2.6).
QDT: Quantum defect theory, interprets experimental spectra by phase shifts in

the radial wave functions and makes predictions for scattering processes (see
Sect. 3.2.6).
VUV: Vacuum ultraviolet, spectral range of electromagentic radiation. Part of the
21348 (2007).
References
FANO , U. and A. R. P. R AU: 1986. Atomic Collisions and Spectra. Orlando: Academic Press Inc.,
409 pages.
F REEMAN , R. R. and D. K LEPPNER: 1976. Core polarization and quantum defects in high
angular-momentum states of alkali atoms. Phys. Rev. A, 14, 16141619.
H ARTREE , D. R.: 1928. The wave mechanics of an atom with a non-Coulomb central field. Part I.
Theory and methods. Proc. Camb. Phil. Soc., 24, 89110.
I DZIASZEK , Z. and P. S. J ULIENNE: 2010. Universal rate constants for reactive collisions of
ultracold molecules. Phys. Rev. Lett., 104, 113202.
J UNGEN , C.: 1996. Molecular Applications of Quantum Defect Theory. New York, London: Taylor
& Francis, 664 pages.
K RAMIDA , A. E., Y. R ALCHENKO, J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
Spectra Database (version 5.1), NIST. http://physics.nist.gov/asd, accessed: 7 Jan 2014.
M ARGENAU , H.: 1984. The Miracle of Existence. Woodbridge, CT, USA: Ox Bow Press.
NIST: 2011. NIST physics laboratory holdings by element, NIST. http://physics.nist.gov/
PhysRefData/Elements/per_noframes.html, accessed: 7 Jan 2014.
O SPELKAUS , S. et al.: 2010. Quantum-state controlled chemical reactions of ultracold potassium-
rubidium molecules. Science, 327, 853857.
PAULI , W.: 1945. The N OBEL prize in physics: for the discovery of the exclusion princi-
ple, also called the pauli principle, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1945/.
S CHMIDT , K. and M. A. L EE: 1998. Visual Schrdinger: A visualizer-solver. http://fermi.la.asu.
edu/Schroedinger/, accessed: 7 Jan 2014.
S CHUMACHER , E.: 2011. FDAlin programme, computation of atomic orbitals (Windows and
Linux), Chemsoft, Bern. http://www.chemsoft.ch/qc/fda.htm, accessed: 5 Jan 2014.
S EATON , M. J.: 1983. Quantum defect theory. Rep. Prog. Phys., 46, 167257.
S EATON , M. J.: 2002. Coulomb functions for attractive and repulsive potentials and for positive
and negative energies. Comput. Phys. Commun., 146, 225249.
U NIVERSITY OF C OLORADO: 2000. Davids wizzy periodic table, Physics 2000. http://www.
colorado.edu/physics/2000/applets/a2.html, accessed: 7 Jan 2014.
W EBER , K. H. and C. J. S ANSONETTI: 1987. Accurate energies of ns, np, nd, nf, and ng levels
of neutral cesium. Phys. Rev. A, 35, 46504660.
W IKIPEDIA CONTRIBUTORS: 2014. Periodic table, Wikipedia, The Free Encyclopedia. http://en.
wikipedia.org/wiki/Periodic_table, accessed: 7 Jan 2014.
Non-stationary Problems: Dipole Excitation
with One Photon 4
A quantum system, such as an atom, may only be observed by

changing it. Electromagnetic waves can induce transitions
between stationary states and are the basis of spectroscopy
one of the most important methods for studying quantum
systems in general. In this chapter we want to recapitulate the
quantum mechanical tools needed and then treat in detail the
rules and phenomena which underly light induced, electric
dipole (E1) transitions.
Overview
The present chapter concentrates on electric dipole transitions (E1), while
Sect. 5.4 will also treat electric quadrupole (E2) and magnetic dipole (M1)
transitions. After some basics and terminology on electromagnetic radiation,
polarization, and photon spin (Sect. 4.1), the essentials of spectroscopy are in-
troduced in Sect. 4.2, the E INSTEIN A and B coefficients are defined, and the
classical model of a radiating oscillator is reviewed. The advanced reader may
jump over this section and ignore also Sect. 4.3.14.3.5, where the elements
of time dependent perturbation theory are summarized. However, Sect. 4.3.6
with terminology and some key results as well Sect. 4.4 with essentials on
selection rules for dipole (E1) transitions are needed in the following sections
and should be read carefully. The same holds for Sect. 4.5 where the angular
characteristics of dipole radiation are presented. Section 4.6 may be used by
the expert reader just as a source of reference with details on the evaluation
of matrix elements and E INSTEIN coefficients. In Sect. 4.7 photoinduced lin-
ear combinations of states are discussed a theme of broad relevance. In this
context we also introduce quantum beats and indicate some spectroscopic per-
spectives. Finally, we ask the very fundamental, almost philosophical question
whether electrons may really jump from one stationary state into another
and present experiments illuminating this profoundly.

170 4 Non-stationary Problems: Dipole Excitation with One Photon
4.1 Electromagnetic Waves: Electric Field, Intensity,

Polarization and Photon Spin
Before discussing radiation induced transitions we define the terminology for de-
scribing electromagnetic waves used here and in most of the following text. For
brevity, we shall speak about electromagnetic waves, electromagnetic radiation
and light more or less synonymously, although the latter is often used more specif-
ically for the visible part of the spectrum.
4.1.1 Electric Field and Intensity
For the present purpose it is sufficient to consider only plane, monochromatic waves
extending over all space, and to concentrate on the electric field component of the
wave which is responsible for E1 transitions.1
The electric field vector E(r, t) is an observable in the real world depending on
position vector r and time t. We thus write it as2
i
E(r, t) = E0 eei(krt) e ei(krt) (4.1)
2
with the (real) field amplitude E0 , the unit polarization vector e, and the wave vector
k with |k| = 2/ = /c.
Even though it is sometimes convenient to use a complex representation (and take
its real part as observable after all calculations are done), we emphasize strongly that
it is important to write E(r, t) as a real quantity in order to be able to describe all
observed physical phenomena as done here neither of the two summands in (4.1)
can be ignored as we shall see shortly!
From classical electrodynamics we know that the field amplitude E0 is related to
the (time averaged) intensity I of the electromagnetic wave by

E0 = 2I /(0 c) = 2I Z0 = 27.45 I "1/2 (4.2)
with the electric constant 0 , the speed of light c and Z0 the characteristic vacuum
impedance. For practical purposes we communicate a handy numerical expression:

I
E0 = 2745 V m1 . (4.3)
W cm2
1 We find this approach (leading to correct results for E1 transitions) conceptually more accessible
than the general, rigorous treatment of all transition types based on the vector potential. The latter
is outlined in Appendix H, while in Chaps. 1 and 2, Vol. 2, we shall generalize (4.1) and learn how
to treat spatial distributions and quasi-monochromasy of real light beams adequately.
2 Theoverall phase angle 0 does not play a role in the present discussion and will generally be
ignored. However, we shall have to come back to it in Chaps. 1 and 2, Vol. 2.
4.1 Electromagnetic Waves: Electric Field, Intensity, Polarization 171
In reality, one usually has to deal with light of a certain bandwidth, i.e. with an
intensity distribution I() (intensity per unit angular frequency). Then one has to
replace I I()d and to integrate expressions derived for transition probabilities
over the whole available spectral range.
4.1.2 Basis Vectors of Polarization

In (4.1) the vector nature of the field is expressed by e, the unit polarization vector
which may be complex since it is always accompanied by its conjugate complex e .
Using a (real) Cartesian basis

1 0 0
ex = 0 , ey = 1 , ez = 0 , (4.4)
0 0 1
and assuming for simplicity that the light propagates in +z-direction (z k) each
arbitrary polarization vector may be expressed as a suitable linear combination of
the basis pair (ex , ey ). This is due to the fact that freely propagating light is trans-
versely polarized and the ez component has no relevance. The basis vectors are
clearly orthonormal ei ej = ij .
Alternatively, one may write e in a (complex) spherical basis, also called helicity
basis with basis vectors eq

1 0
1 1
e+1 = (ex + iey ) = i
= e1 , e0 = ez = 0

2 2 0 1
(4.5)
1
1 1
and e1 = (ex iey ) = i = e+1 .
2 2 0
In this basis, the polarization of light which propagates in the +z-direction (e0 ) may
again be expressed by only two basis vectors (e+1 , e1 ). One easily verifies again
the orthonormality
eq eq = qq with eq = (1)q eq . (4.6)
For atomic problems the helicity basis is often better adapted, since atoms too are
usually described in a spherical coordinate system. For reference, we express the
Cartesian basis in terms of the helicity basis:
1 i
ex = (e+1 e1 ) and ey = (e+1 + e1 ). (4.7)
2 2
For later use we also give the unit polarization vectors for linearly polarized light in
45 and 135 direction in respect of the x-axis:
1 1
e 45 = (ex + ey ) and e 135 = (ex ey ). (4.8)
2 2
In the spherical basis this is written as:

1
e 45 = (i 1)e+1 + (i + 1)e1
2
(4.9)
1
e 135 = (i + 1)e+1 + (i 1)e1 .
2
Again, for light propagating in +z-direction this pairs of polarization vectors con-
stitutes an orthonormal basis set.
We now give a few explicit examples for the wave fields. If e = ex , i.e. for light
polarized linearly in x direction, (4.1) becomes simply
E x (r, t) = E0 sin(kr t)ex , (4.10)
while linear polarization in y direction is described by
E y (r, t) = E0 sin(kr t)ey . (4.11)
The unit vector e = e+1 describes left hand circularly polarized light (LHC), also
called + light. Inserting (4.5) into (4.1) gives
1
E +1 (r, t) = E0 sin(kr t)ex + cos(kr t)ey , (4.12)
2
and e1 stands for right hand circularly polarized light (RHC) or light:
1
E 1 (r, t) = E0 sin(kr t)ex cos(kr t)ey . (4.13)
2
An illustration of + light gives Fig. 4.1. Shown is the E vector at a fixed time
t = /2 along the z-axis. As indicated E rotates at a fixed position clockwise
around the wave vector k of the light, i.e. with positive helicity. One verifies this
most directly in (4.12) with kr = 0, or correspondingly in Fig. 4.1, if one considers
the time propagating.3
The most general unit polarization vector for light propagating into an arbitrary
direction can be written in either Cartesian or spherical basis:

1
1
e = ax e x + ay e y + az e z = aq e q with |aq |2 = 1. (4.14)
q=1 q=1
In the spherical basis it may be specialized for any elliptically polarized light prop-
agating parallel to the +z axis (i.e. into e0 direction) by
eel = a+ e+1 + a e1 = ei cos e+1 ei sin e1 , (4.15)
3 Methods for generating and detecting polarized light will be discussed in Chap. 1, Vol. 2.
z
+ light (LHC)
k
+t
+t 0
k y
+t
Fig. 4.1 Schematic illustration of left hand circularly polarized light + (LHC). The grey arrows
indicate the direction of the electric field perpendicular to the z k axis at a fixed time, correspond-
ing to t = /2. For larger times the sense of the rotation is indicated. The somewhat surprising
definition for left hand circular polarized light is of historical origin: in times prior to the laser,
people used to look into the oncoming light beams, hence they considered + light as rotating
anti-clockwise, i.e. LHC polarized
with the component a0 = 0 in this coordinate system). The ellipticity angle de-
scribes the degree of ellipticity, the alignment4 angle gives the direction of the
ellipse in respect of ex . Converting (4.15) into Cartesian coordinates with (4.5) and
inserting into (4.1) gives an explicit, parameterized expression for the elliptic field
vector in terms of and :

E el (r, t) = (E0 / 2)
'
cos sin(kr t ) + sin sin(kr t + ) ex (4.16)
(
+ cos cos(kr t ) sin cos(kr t + ) ey .
This describes the standard form of an ellipse on which the E vector rotates if (for
fixed r) the phase kr t + varies between 0 and 2 . As sketched in Fig. 4.2,
this ellipse is inclined by an angle in respect of the x-axis, and for 0 /2
its major and minor axes a and b, respectively, are:

a = E0 (cos + sin )/ 2 = E0 sin( + /4)
(4.17)
b = E0 (cos sin )/ 2 = E0 cos( + /4).
4 Unfortunately the terms alignment and orientation are mixed up time and again in the liter-
ature: Alignment refers to the direction of a polar vector (e.g. the E vector in the case of linearly
polarized light). In contrast, orientation specifies the sense of rotation of an axial vector (e.g. the
angular momentum of left and right circularly polarized light).
Fig. 4.2 Polarization ellipse x

seen for right hand elliptically
polarized light, seen into the
+z-direction (k points into
the plane). As indicated, the a
light is right hand polarized
(negative helicity). The b y
parameters are 60 ,
64 or 0.34
Note that a 2 + b2 = E02 . In the literature instead of the ellipticity angle often a
(somewhat ambiguously defined) so called ellipticity is used:
= b/a = cot( + /4). (4.18)
Three special cases of are of particular significance according to (4.15):
1. Left hand circularly polarized light ( + ) corresponds to an ellipticity angle

= 0 , while
2. right hand circularly polarized light ( ) is characterized by = +/2.
3. Linearly polarized light corresponds to = /4, i.e. sin = cos = 1/ 2.
From (4.16) one finds that in this case:
E(r, t) = E0 e() sin(kr t)

(4.19)
= E0 (cos ex + sin ey ) sin(kr t).
4.1.3 Coordinate Systems
The main part of this chapter will be concerned with absorption and emission of
electromagnetic radiation from atoms. In general, one has to distinguish two dif-
ferent coordinate systems: one in which the atom is best described we call it the
atomic frame (at) while the photon may possibly be better described in another
frame we call it the photon frame (ph).
These coordinate systems may differ from each other. In the spirit of Sect. 4.1.2
a convenient choice for the photon frame is illustrated in Fig. 4.3. We shall use this
henceforth unless otherwise stated. The atomic frame may e.g. be defined by an ex-
ternal electric or magnetic field in respect of which the projection quantum numbers
m are defined. The photon frame is defined as shown in Fig. 4.3 by the direction of
light propagation (wave vector k), with the axes z(ph) k and y (ph) z(at) .
(at)
The unit vectors eq in the atomic frame (helicity basis) indicated in Fig. 4.3
by heavy red arrows may in a classical picture be seen to represent three different
classical dipole oscillators. This will be discussed in Sect. 4.5.1.
Fig. 4.3 Coordinate system z (at) z ( ph)

for the radiating atom (red,
(at)) and for the photon k
(black, ph). The three y (ph)
classical oscillators which
x ( ph)
correspond to the basis e 0(at ) k
vectors e(at) (at) (at)
1 , e0 and e1 are m= 0
indicated by thick, red arrows
e -1
(at)
y (at)
m= +1 e +1
(at)
m = -1 k
x (at)
4.1.4 Angular Momentum of the Photon
At this point we want to extend our knowledge about the properties of photons sum-
marized in Sect. 1.4. In addition to momentum and energy the photon also possesses
an intrinsic angular momentum, the so called photon spin with a quantum number
sph = 1. Projected onto the wave vector k LHC and RHC polarized light have angu-
lar momenta ms with ms = 1 and 1, respectively. Owing to the transverse nature
of electromagnetic waves the third component with ms = 0 does not exist.
In a particularly beautiful and fundamental experiment this has been observed for
the first time by B ETH (1936). The setup and the results are shown schematically
in Fig. 4.4. LHC photons are prepared by passing linearly polarized light through
a /4 plate. These photons then pass through a /2 plate where they are converted
into RHC. This implies an angular momentum change of 2 per photon which is
transferred to the /2 plate in order to conserve the total angular momentum of the
system. For more efficiency the light beam is retroreflected behind another /4 and
quartz fibre mirror and light beam for experimental result:

measuring the twist light torque M (measured
by the twist of the quartz
/4 plate fibre) as a function of
coated on top circular polarization
x y
defined by the angle
LHC LHC
5
M / 10-9 dyncm
/2 plate
RHC RHC 0
/4 plate
x y xy-polarization plane -5
of linear polarization
linearly polarized light in = -135 -90 -45 0 45
Fig. 4.4 Experiment of B ETH (1936) on mechanical detection and measurement of angular mo-
mentum of light
passes the /2 plate again. As a consequence a measurable torque M = dL/dt =

2 (2I A/) twists the plate which is suspended on a quartz fibre, I being the
intensity of the light beam, A its effective area and the energy of one photon.
The experiment confirms quantitatively that the photon has indeed a spin projec-
tion for LHC and RHC light, respectively. The photon is a boson! In contrast to
electrons (which are fermions) several photons may occupy identical states a pre-
requisite for a laser, as we shall see in Chap. 1, Vol. 2.
Section summary
We describe electromagnetic radiation by a monochromatic plane wave. As
a physical observable it is represented by a real valued function (4.1). Both
exponential terms will turn out to be relevant when describing the interaction
of the electromagnetic field with quantum systems.
The helicity basis (4.5) is convenient for describing orthonormal (4.6) unit ba-
sis vectors for problems in atomic physics. The most general unit polarization
vector for light propagating into +z-direction is given by
eel = ei cos e+1 ei sin e1 .
Ellipticity angles = 0 and /2 describe LHC ( + ) and RHC ( ), respec-

tively, = /4 corresponds to linear polarization ( ). The alignment angle
gives the direction of the major axis of the polarization ellipse (for linear
polarization the E vector) with respect to the x-axis.
The photon is a particle with angular momentum (spin) sph = 1, i.e. it is a
boson. As experimentally verified by B ETH (1936), the spin projection onto
the z-axis of propagation is for light, respectively.
4.2 Introduction to Absorption and Emission

4.2.1 Stationary States
Figure 4.5 schematically shows a typical system of stationary states of an atom

which converge towards the ionization limit. As we have seen for the H atom and
the alkali atoms the term energies are found by solving the S CHRDINGER equation
|n = Wj |n
H
at physically reasonable boundary conditions (in the following we shall often ab-
breviate all quantum numbers by one letter, e.g. n j = j, a or b). Energy zero is
defined such that for bound states Wj < 0 while the free electron has a continuous
spectrum of energies Wkin = W 0. As discussed in the previous two chapters, the
asymptotic behaviour of the radial wave function of the bound states is given by

lim Rj (r) exp 2|Wj |r
r
while the continuum states are described essentially by C OULOMB wave functions.
4.2 Introduction to Absorption and Emission 177
Fig. 4.5 Stationary states of W ionization

an atom with IP WI continuum
0
|j
bound
| 4 stationary
states
| 3
W2 | 2
WI
W1 | 1
4.2.2 Optical Spectroscopy General Concepts
According to B OHR, transitions |b |a between two stationary states are ac-

companied by emission (or are induced by absorption) of electromagnetic radiation
with a frequency ba given by:
hba = ba = hc ba = |Wb Wa |. (4.20)
Here ba = 2ba is the angular frequency of the radiation, ba = c/ba its vacuum
wavelength and ba = 1/ba = ba /c its wavenumber. Figure 4.6 illustrates three
basic types of setups for spectroscopic investigations. Specialties and refinements
will be discussed in detail later, but essentially all spectroscopic measurements fol-
low in principle one of these three concepts:
1. Emission spectroscopy: A hot gas or a plasma (e.g. a spectral lamp) emits by

spontaneous decay energy in the form of photons. The emitted light is analyzed
according to its wavelength by some type of spectrometer (in Fig. 4.6 symbolized
by a prism). As indicated in Fig. 4.6 (right) many excited states |b are involved,
each of them may decay into several states |a of lower energy. Thus, the light
emitted originates from a variety of different combinations of upper and lower
states. Hence, emission spectra are typically rich of lines and difficult to analyze.
2. Absorption spectroscopy: Only light of the characteristic transition frequencies
according to (4.20) is absorbed, all other light just passes through the target. Ide-
ally, only one initial (ground) state |a is populated in the cold target, and the
number of lines observed is significantly smaller as compared to emission spec-
troscopy. One typically uses white light (e.g. from a synchrotron radiation source
in the VIS, UV, VUV or X-ray spectral region, or from a glow bar in the IR).
After passage through the target the light is analyzed, today usually by optical
multichannel analyzers (OMA) which allow for simultaneous registration of the
signal over a whole spectral range. The alternative, to make the light monochro-
matic before sending it through the target and to scan its wavelength over the
region of interest is usually much less efficient.
3. Fluorescence spectroscopy: In a first step, the atoms or molecules to be studied
are excited very selectively into one specific state |b , typically by a laser (laser
hot gas,
plasma
emission spectroscopy
cold gas
white light absorption spectroscopy
lamp
fluorescence spectroscopy,
las
laser induced fluorescence

er
specifically excited atoms exciting laser line
Fig. 4.6 3 basic methods of spectroscopy very schematic. Left: experimental setups, middle: type
of observed spectra, right: term schemes and transitions
induced fluorescence, LIF). All observed emissions |a |b then start from

this one level, leading to very clear spectra. Since the exciting wavelength may
also be varied, the method is very powerful in identifying the nature of unknown
transitions. If the emission occurs delayed (on the ms to s time scale) the process
is called phosphorescence.
4.2.3 Induced Processes
Transitions which occur under the influence of an external electromagnetic radiation

field are called induced. Such transitions |b |a may occur in both directions:
we distinguish absorption and stimulated emission.
Absorption
As derived in Sect. 1.3.2 the L AMBERT-B EER absorption law (1.44)
dI
= Na I I (z) = I0 e Na z = I0 ez (4.21)
dz
describes the exponential reduction of the intensity I (z) of electromagnetic radia-
tion when passing through an absorbing medium of particle density Na with [Na ] =
number of absorbing particles m3 . Here Na is assumed to be constant over dis-
tance z and time t.
We now want to understand the process on a microscopic level and derive quan-
titative expressions for the absorption cross section ([ ] = m2 ) and the absorp-
tion coefficient = Na ([] = m1 ). Since the energy of a photon is we can
Fig. 4.7 Atomistic view on

the absorption of |b Nb
electromagnetic radiation Nph(0) Nph(t)
n (n-1)
|a Na
translate the intensity of the radiation field I ([I ] = W m2 ) into its photon density
Nph ([Nph ] = number of photons m3 ) or energy density u = Nph ([u] = J m3 )
by
I = cNph = cu. (4.22)
As schematically indicated in Fig. 4.7, on an atomistic level the absorption process
corresponds to a loss of photons from the field, i.e. a reduction of Nph by dNph
which is identical to a loss dNa of particle density Na in the initial state |a and gain
dNb of density Nb in the final excited state |b .
In contrast to the situation described by (4.21), we are now interested in the
change of N with time t at fixed position z for a constant intensity I . Since a change
over a distance dz corresponds to a change over a time dt = dz/c we may rewrite
(4.21) with (4.22) as
I dI dI dNph dNa dNb

Na = = = = = . (4.23)
dz cdt dt dt dt
With this one defines an absorption (or excitation) rate5
1 dNa I
R ba = = ba = ba (4.24)
Na dt
for the transition into the state |b from state |a . We have introduced here also the
photon flux = I /.
Note that the rate R ba has the dimension T1 ; it is the probability for absorption
of photons by an atom per unit of time. To obtain the total number of events per unit
time one has to multiply it by the number of atoms in the volume under observation.
In the above discussion we have tacitly assumed that the radiation to be absorbed
is strictly monochromatic and tuned into resonance with the absorption line. In real-
ity, however, the intensity has a spectral distribution, which we indicate by I() with
[I()] = W m2 s1 . Thus, only that fraction of intensity can be absorbed which is
in resonance with the angular frequency ba of the transition. Usually one refers the

transition probability to the spectral radiation density u(), i.e. the energy density
5 Forclarity, throughout this chapter we shall use R ba , B ba etc. for rates and coefficients averaged
over initial and summed over final substates, since in general we have to account for degenerate
energy levels. In contrast, Rab (e) refers to transitions between specific substates b and a with
polarization e.
Table 4.1 Definitions in the context of absorption and emission of electromagnetic radiation
Symbol Eq. Term Unit Remarks
absorption cross section m2 effective absorbing area
N particle density m3
=N absorption coefficient m1 I = I0 exp(x)
= amplification coefficient m1 if < 0
I = c Nph light intensity W m2
Nph = I /(c ) photon density m3
I = E02 /(2Z0 ) intensity E0 field amplitude
Z0 = 1/(c0 ) impendence in free space 376.73 also vacuum impedance
Nph = r 0 2
2 E0 energy density J m3 radiation field
Nph

u() = spectral radiation density J m3 Hz1
0 E02
= energy
Vol = I
c = 2 = I ()
2c W s2 m3
u() = 2u()
bandwidth (frequency) Hz
. . . (angular frequency) s1
du per spectral range d, which is related to the spectral intensity distribution I()
by
I() du d du u()

u() = = = = , (4.25)
c d d d 2
with the spectral radiation density given alternatively per unit angular frequency
= 2 or per unit frequency of the electromagnetic radiation. (We shall use the
former unless mentioned otherwise.)
Relevant for the transition probability is the intensity at the transition frequency.
Consequently, we now rewrite (4.24) as:
1 dNa I(ba )
R ba = = B ba = B ba u(
ba ). (4.26)
Na dt c
The thus defined constant B ba is called E INSTEIN coefficient for absorption with
[Bba ] = m3 s2 J1 . Note that rates, transition probabilities and matrix elements
from state |a to state |b are typically indexed from right to left.
Table 4.1 summarizes the terms and definitions most frequently used for the
quantitative description of radiation induced dipole transitions. Most of the discus-
sion to follow in this chapter will focus on the E INSTEIN coefficients. They contain
the essence of the transitions in quantum systems. Polarization and frequency de-
pendence play an important role.
Stimulated Emission
Up to now we have assumed that only one dominantly populated ground state level
contributes to the absorption process. And the absorption as such was thought to be
weak enough so that it does not change the population significantly. We now con-
Fig. 4.8 Stimulated emission

|b Nb
Nph(0) Nph(t)
n (n+1)
|a Na
sider the inverse process, the so called stimulated emission which we have schemat-
ically illustrated in Fig. 4.8. For this to happen, the excited state must be populated.
For simplicity we assume for the moment that all atoms are initially found in the
excited state |b as indicated in Fig. 4.8. In complete analogy to (4.24) and (4.26)
the corresponding emission rate is
1 dNb I(ba )
R ab = = B ab = B ab u(
ba ). (4.27)
Nb dt c
We thus find that more photons may come out of the system than going into it if
the excited state has a higher population than the ground state so that this process
dominates. B ab is called E INSTEIN coefficient for stimulated emission.
In general, both processes absorption and stimulated emission happen in a
quantum system. According to (1.54) the ratio of the population densities are given
by Nb /Na = (gb /ga ) exp[(Wb Wa )/(kB T )] in thermodynamic equilibrium. At
room temperature kB T 25 meV is small, while electronic excitation in atoms typ-
ically requires some eV; hence Wb Wa kB T and Nb Na . In contrast, if the
energy gaps are much smaller, as it is the case e.g. for molecular vibrational and ro-
tational states, one has to consider induced emission also in standard spectroscopic
absorption experiments.
A completely different situation arises when skillful schemes are applied to
achieve a significant population of some specific excited state |b such that Nb > Na
(for some state |a , not necessarily the ground state). Then, stimulated emission can
be stronger than absorption and laser action becomes possible. We shall discuss this
in some detail in Chap. 1, Vol. 2.
4.2.4 Spontaneous Emission Classical Interpretation
From experimental experience we know that excited atoms may also decay sponta-
neously. Quantum electrodynamics (QED) explains this by the interaction of quan-
tum systems with the vacuum field: In a fully quantized description the electromag-
netic field is represented by harmonic oscillators, their energy being (Nj + 1/2)j .
Here Nj is a (typically very large) integer number proportional to the intensity of
the radiation at the frequency j , and j /2 is the so called zero-point energy. The
latter represents the eigenenergy of the ground states of the field oscillators at zero
intensity of the radiation. This vacuum field is essentially a consequence of the
uncertainty relation. One may say, somewhat figuratively, that it is this (isotropic)
zero point oscillation of the quantized radiation field, which forces an excited quan-
tum system to decay spontaneously. We shall present a quantitative description of
spontaneous emission based on this concept in Chap. 2, Vol. 2.
Here we consider a heuristic, classical approach in which the atomic electron is
treated as an accelerated charge with a dipole moment D(t) = er(t). This clas-
sical picture does not allow to extract quantitative results. It leads, however, to an
intuitive understanding. According to classical electrodynamics such a dipole emits
electromagnetic waves. In the nonrelativistic limit the electric field at large distances
R from the source is given by

E(R, t) =
1 k

k
t = er (t ) ,
D (4.28)
2
40 c R k k 40 c R 2
with t = t R/c being the retarded time, and r representing the components of
the dipole acceleration perpendicular to the wave vector k. We identify this with
projecting r onto the x (ph) y (ph) plane of the photon frame introduced in Fig. 4.3.
For a harmonic dipole oscillator r(t) = r (at) exp(iba t) with amplitude r (at) in the
atomic frame and angular frequency ba the field amplitudes are given by
2
eba (ph)
Ex (R, t) = r (at) ex ei(kRba t) (4.29)
40 c2 R
2
eba (ph)
Ey (R, t) = r (at) ey ei(kRba t) . (4.30)
40 c2 R
In Sect. 4.5.1 these expressions will be used for a semiclassical derivation of angular
and polarization characteristic of atomic dipole radiation. Here we are just interested
in the overall power emitted. From (4.28) we obtain with (4.2) the intensity I =
0 c|E(R, t)|2 , and on average6 an energy
dW
|D(t)| 2
d = I R 2 d = sin2 k dk (4.31)
dt (4)2 0 c3
is emitted per unit time into a solid angle dk = sin k dk dk . Here k is the angle
between D and k. Integration of sin2 k over all solid angles gives a factor 8/3 so
that the total radiation power emitted is
2
dW 1 e2 dp
P= = | 2=
D| (4.32)
dt 60 c 3 60 me c dt
2 3
with [P ] = J s1 . For later use we have introduced also the momentum p = me r of

the oscillating electron.
6 The bar on top of quantities indicates temporal averaging.

2 = |2 D|2 = e2 |r (at) /2|2 4

For the harmonic dipole oscillator we have |D| ba ba
and hence the average energy radiated by one atom per unit time is
dW 1 e2 |r (at) /2|2 ba
4
P= = . (4.33)
dt 6 0 c3
Due to this radiation the excited state of the atom decays. In a classical picture
the oscillation amplitude (i.e. the orbital radius of the electron) continuously de-
creases contrary to the spectroscopic observation of discrete, stationary states:
the atom (or any other quantum system) is either in the excited or in the ground
state.
Thus, quantum mechanically (4.33) is interpreted as a probability statement:
The probability that one excited atom emits a photon of energy during the time
interval dt is
dW 1 e2 |r (at) /2|2 ba
3
R ab dt = A ab dt
(spont)
dt = = (4.34)
3 0 c 3
with the spontaneous decay rate R ab

(spont)
for transitions |a |b between two en-
ergy levels. The decay constant ([Aab ] = s1 ) derived from this classical model is
3 |er (at) /2|2

ba 4 3 (at) 2 32 3 c |r (at) /2|2
A ab = = r /2 = , (4.35)
30 c3 3c2 ba 3 3
with 1/137 being the fine structure constant (1.10). The exponential decay law
ensuing form (4.34) has already been discussed in Sect. 1.3.1. For the excited state
density Nb we obtain according to (1.39) and (1.40) in the absence of other pro-
cesses
1 dNb 1
= A ab = Nb (t) = Nb0 eAab t = Nb0 et/ab
Nb dt ab
with the average excited state lifetime ab = 1/A ab , the half lifetime 1/2 =
(ln 2)ab , and Nb0 the excited state density at time t = 0.
We note at this point, that (4.35) is almost identical to the exact expression
for the so called E INSTEIN A-coefficient for spontaneous emission, which will be
discussed in Sect. 4.6.2. The specific properties of the emitting atom, here repre-
sented by |r (at) /2|, require of course a stringent quantum mechanical interpreta-
tion. For a prominent case, the 3p 3s transition in sodium (the so called Na
D line at = 589 nm), we try an intelligent guess by setting r (at) 0.190 nm
(the atomic radius, see Fig. 3.4). From (4.35) we obtain A 3s3p 3.2 107 s1 or
= 1/A ab 31 ns which is at least on the same order of magnitude as the cor-
rect, experimentally determined value 16.2 ns. Of course, we do not expect exact
results from this classical model.
absorption stimulated spontaneous

emission emission
|b Nb
n (n -1) n (n+1)
|a Na
initial final initial final initial final
I I I
ab ab
characteristics
of spectra ab

Fig. 4.9 Absorption, induced and spontaneous emission schematically. In the top row the atom-
istic view point is indicated, below typical spectroscopic patterns are illustrated: the detected in-
tensity as a function of angular frequency of the radiation
Fig. 4.10 Two level system Nb |b

and E INSTEIN coefficients;
= = =
for deriving the principle of Bab u(ab) Aab Bba u(ab)
detailed balance
Na |a
4.2.5 The EINSTEIN A and B Coefcients
Figure 4.9 summarizes the above considerations and findings. In a real experiment
one has to consider all three processes simultaneously for describing an atomic sys-
tem in the presence of an electromagnetic field correctly. A special case is a system
of atoms, molecules or condensed matter oscillators in thermodynamic equilibrium
with its own radiation field.
E INSTEIN has taken this thermodynamic equilibrium as a starting point for a very
elegant and simple derivation of P LANCKs radiation law Fig. 1.15. It is based on
the so called principle of detailed balance: under stationary conditions each process
must be in equilibrium with its inverse. One treats the problem as a representative
two level system sketched in Fig. 4.10. The kinetics of this system is described by so
called rate equations. The population of excited and ground state may change due
to the three processes as follows:
dNb dNa
= A ab Nb Nb B ab u(
ba ) + Na B ba u(
ba ) = . (4.36)
dt dt
Stationarity means that the particle densities do not change:
dNb,a
= 0 = A ab Nb Nb B ab u(
ba ) + Na B ba u(
ba )
dt
A ab Nb A ab
=1 = u(
ba ).
(Na B ba Nb B ab )u(
ba ) B ba (Na /Nb ) B ab
On the other hand, according to (1.54) thermodynamic equilibrium implies

Nb gb W gb ba
= e kB T = e kB T (4.37)
Na ga ga
with the degeneracies gb and ga of ground |a and excited states |b , respectively.
Thus the spectral radiation density becomes
A ab A ab
ba ) =
u( = .

Bba (Na /Nb ) Bab Bba (ga /gb )eba /(kB T ) B ab

By comparing this with P LANCKs radiation law (1.81)
3 1

u() = /(k
,
2 3
c e BT ) 1
the following relations between the A and B coefficients emerge:
3
A ab = R ab = 2 3 B ab and gb B ab = ga B ba .
(spont)
(4.38)
c
Note the 3 factor interestingly, it agrees with the classical prediction (4.35).
We point out that the B coefficients used here refer to spectral radiation densities
u() per unit angular frequency.7
Section summary
We have introduced three prototype methods for obtaining information from
quantum systems by interaction with an electromagnetic field: emission, ab-
sorption and fluorescence spectroscopy.
Induced processes (absorption and induced emission) are introduced phe-
nomenological. The rates for both processes are proportional to the spectral
ba ) = I(ba )/c at the transition frequency studied.
radiation intensity u(
A classical interpretation of spontaneous emission encounters severe prob-
lems which have led to B OHRs second theorem and to the probabilistic
interpretation of the emission process. Nevertheless, it provides an informa-
tive first guess on the spontaneous emission probability and useful information
(4.32) for later use.
E INSTEINs has derived P LANCKs radiation law from rate equations for a
system of quantum oscillators in thermal equilibrium with the radiation field.
This leads immediately to quantitative relations between the E INSTEIN coef-
ficients for spontaneous and induced emission.
The readers should memorize the famous factor 3 (or 3 ) between them:
A ab 3 B ab = 3 (ga /gb )Bba it will play an important role on several oc-
casions.
7 In
the literature often u() = 2 u()
per unit frequency is used. Then one has to replace B ba
B ba
()
= B ba /2 in all relevant equations.
4.3 Time Dependent Perturbation Theory

After this phenomenological introduction into absorption and emission of photons
and equipped with a working description of electromagnetic waves, we are now
ready to develop a quantum mechanical understanding of dipole transitions. In order
to quantitatively describe the temporal change of atoms under the influence of elec-
tromagnetic radiation, a semiclassical, perturbative approach will be applied at this
point: the atom is treated fully quantum mechanically. The electromagnetic field,
however, is considered as a classical time dependent perturbation U (r, t), which
leaves the initial state of the atom essentially intact apart from the small transi-
tion amplitudes to be derived. Spontaneous emission will be treated as a kind of
afterthought by making use of the A/B ratio (4.38) for the E INSTEIN coefficients
just discussed.
As already mentioned, a rigorous derivation of spontaneous emission requires
QED and the quantization of the electromagnetic field. Since this concept is signif-
icantly less transparent than the semiclassical approach we defer its introduction to
Chap. 2 in Vol. 2. All spectroscopic relations derived in the following will, however,
remain valid as long as the intensity of the radiation field is high enough so that
its statistical properties are irrelevant, and on the other hand not too high so that
the problem may be treated with sufficient accuracy by perturbation theory. These
constraints will be released in Chaps. 2 and 10 in Vol. 2, respectively.
4.3.1 General Approach

We briefly recall from quantum mechanics how non-stationary states are treated
by solving the time dependent S CHRDINGER equation (2.13) approximatively. We
communicate the necessary, basic tools from time dependent perturbation theory for
deriving probabilities for transitions to state |b from state |a . The reader who is
sufficiently familiar with these tools may continue directly at Sect. 4.3.6.
One starts with the known 0th order solution for an unperturbed problem:
0 j (r) = Wj j (r).
H (4.39)
The time dependent S CHRDINGER equation (2.13) for H 0 = H
0 (t) has a trivial
time dependence (2.16) which we write here with j = Wj /

(0) Wj
j (r, t) = j (r) exp i t = j (r) exp(ij t).

(
Adding a time dependent perturbation U p , r, t) the Hamiltonian becomes
(t) = H
H 0 + U
(
p , r, t). (4.40)
Abbreviating (r, t) | (t) and j (r) |j we now make the ansatz

(t) = cj (t)eij t |j (4.41)
j =0
4.3 Time Dependent Perturbation Theory 187
where cj (t) is the time dependent probability amplitude of state j (r) and the prob-
ability to find a certain final state |j at time t is given by
2
wj (t) = cj (t) . (4.42)
If the summation (4.41) is carried out over sufficiently many states and the basis |j
is complete, it represents a solution as close to exact as desired.
Using this ansatz and the time dependent Hamiltonian (4.40), the time dependent
S CHRDINGER equation (2.13) is written as
| (t)
H0 + U p , r, t) (t) = i
( (4.43)
t
[cj (t)eij t ]|j
cj (t)eij t H0 + U
(p , r, t) |j = i .
t
j j
Inserting H0 |j according to (4.39) on the left hand side of this equation, and using
the rules of product differentiation on the right hand side leads to

(
cj (t)eij t Wj + U p , r, t) |j
j
dcj (t)

ij t ij t
=i cj (t)(ij )e + e |j .
dt
j
With i(ij ) = Wj the first terms in the sums on either side cancel. One multiplies
from the left with b|eib t and applies b|j = bj . With Wb = b and the time
dependent perturbation matrix elements

Uij (t) = i|U ( (
p , r, t)|j = i (r)U p , r, t)j (r)d3 r (4.44)
a system of linear differential equations is obtained for the probability amplitudes

cb (t) with the (angular) transition frequency bj = (Wb Wj )/:
dcb (t) i
= cj (t)Ubj (t)eibj t . (4.45)
dt
j
Note that this is a general and still exact formulation for atoms or molecules ex-
posed to a time dependent interaction. In particular, it is not restricted to the dipole
approximation nor even to monochromatic plane waves.
4.3.2 Perturbation Ansatz for Transition Amplitudes
In practice one can only sum over a finite number of terms. A perturbation approach
(
is possible if |U 0 |, i.e. if the averaged perturbation is small com-
p , r, t) | |H
pared to the averaged Hamiltonian of the unperturbed system (H atom, alkali atom
etc.). Then the initial conditions will change only slightly with time. To obtain the
probability amplitude cb (t) for the final state b in 1st order time dependent per-
turbation theory one starts in 0th order by setting on the right hand side of (4.45)
ca(0) (t) 1 (initial state a) and cj(0) (t) 0 for all j = a. From this one obtains for
b = a in 1st order:
dcb (t) i
= Uba (t)eiba t . (4.46)
dt
If the interaction is switched on at time t = 0, integration over time leads to the time
dependent transition amplitude in 1st order perturbation theory:

i t
cb (t) = Uba t eiba t dt . (4.47)
0
The full solution in 1st order is then given by

(t) |a eia t + cj (t)|j eij t . (4.48)
j =a
One may iterate the procedure by inserting (4.47) into (4.45) to obtain the 2nd order
solution and so on.
Often one is interested only in how the probability amplitude cb (t) evolves after
many oscillation of the field, i.e. for t 1/ba . In the limit t the 1st order
probability amplitude (4.47)

i
)ba (ba )
cb () = Uba t ei(ba )t dt = U (4.49)
0
is apart from numerical prefactors nothing but the F OURIER transform (I.2)
of the perturbation potential at the frequency ba of the transition. In conclusion
atomic or molecular transitions can only be excited by an interaction potential
which contains F OURIER components U )ba () which are in resonance with the tran-
sition frequency ba = .
4.3.3 Transitions in a Monochromatic Plane Wave
We now specialize to an electromagnetic monochromatic plane wave of frequency

and intensity I . Guided by the mathematical form of the electromagnetic field
(4.1), we introduce now still quite general a perturbation amplitude eE0 and a
transition operator
D=D(p , r):8
8A somewhat heuristic derivation of eE0 and p, r) for electric dipole transitions will be pre-
D(
sented below in Sect. 4.3.4, while a more rigorous, general derivation and specialization is found
in Appendix H.
i it
(
Uba (t) = b|U p , r, t)|a = b| eE0 De D eit |a
2 (4.50)
i
= eE0 Dba eit

Dba eit with
Dba = b|D|a .
2
The transition amplitude (4.47) now becomes

eE0 t
Dba ei(ba )t
Dba ei(ba +)t dt

cb (t) = (4.51)
2 0

eE0 Dba ei(ba )t D ei(ba +)t

= ba . (4.52)
2 i(ba ) i(ba + )
We note here that both terms, exp(it) and exp(+it), are relevant. And one
sees already in (4.51) that significant contributions are expected only for stationary
phases, i.e. for ba = 0, respectively. Otherwise the contributions to the integral
oscillate rapidly and disappear in the limit t . With the angular frequency of
the radiation > 0 (per definition), and a referring to the initial, b to the final state,
the first term in (4.52) describes absorption, the second induced emission:
absorption b a: ba > 0
D exp(it) term relevant
(4.53)
emission a b: ba < 0
D exp(+it) term relevant.
Obviously, both exponential functions originating from writing the field of an

electromagnetic wave (4.1) as a real quantity are indispensable.
4.3.4 Dipole Approximation
We now apply this general formalism to the so called electric dipole transitions
(E1 transitions) and present here a slightly heuristic introduction of the interaction
potential. It leads to the same result as the more stringent (but also less evident)
derivation given in Appendix H.1.6. In both cases the key simplification exploits the
fact that the wavelengths of interest are typically large compared to atomic dimen-
sions ( a0 ). Hence, the electromagnetic field E(r, t) described by (4.1) may be
expanded in powers of r/ or k r 1. To 1st order exp(ik r) reduces to 1, and
only the temporal change of the electric field E(t) needs to be considered.
The force on an electron from this oscillating field is eE(t), and for the
electron-nucleus pair the
electric dipole moment is D = er, (4.54)
the interaction energy in semiclassical dipole approximation is
i
U (r, t) = D E(t) = er E(t) = E0 er eeit e eit . (4.55)
2
It depends on the electron position r and on time t. The field amplitude E0 is related
to the radiation intensity I via (4.2), and e describes the polarization of the field.
Hence, by comparing (4.55) with (4.50) one finds that the dipole transition operator
for absorption is given by

D = r e = D e/e and eE0 = e 2I /(c0 ), (4.56)
and since r = r its adjoint, the operator for emission, is

D = r e . The relevant
dipole transition matrix elements are
for absorption for emission Dba = r ab e , (4.57)

Dab =

Dba = r ba e,
with r ab = r ba . Note that E0 is measured here in V m1 and is not a F OURIER

component (per unit ) of the field, so that U (r, t) is indeed an energy. Of course
it is not mandatory to treat only strictly monochromatic waves, and we shall extend
the description correspondingly in the next subsection.
The electric dipole approximation is a very reasonable, far reaching approxima-
tion. For many spectroscopic applications treating it in 1st order perturbation theory
is sufficient. However, only single photon transitions can be described in 1st or-
der. Multi-photon processes require higher order perturbation theory and will be
addressed in Sect. 5.3. They are very important in modern laser spectroscopy where
often intense radiation fields are used. For two level systems, one may even find
essentially exact solutions still within the framework of the electric dipole ap-
proximation. This will be discussed in Chap. 10, Vol. 2.
On the other hand, it is important to realize at this point, that the 1st order pertur-
bation formalism summarized in (4.46)(4.53) is not restricted to E1 transitions. It
may also be applied to higher order interactions arising from expanding exp(ik r)
in (4.1) beyond the first term as long as only one photon is involved. In order to
treat such processes accurately one has to use the quantum mechanically correct per-
turbation operator U (p , r, t) derived from the vector potential A(r, t) as elaborated
in Appendix H.1.6. In Sect. 5.4 we shall treat the most important cases, magnetic
dipole and electric quadrupole transitions (M1 and E2 transitions, respectively).
They may become important if E1 transitions between two levels of interest are
forbidden.
4.3.5 Absorption Probabilities
In the following we evaluate explicitly the probability amplitude for absorption, i.e.
we focus for the moment only on the first term in (4.52), with ba > 0 (Wa < Wb ).
If the electromagnetic field is switched on at t = 0, we have cb (t) 0 for t < 0
while for t > 0 the transition amplitude evolves as
eE0 ei(ba )t 1
cb (t) =
Dba . (4.58)
2 i(ba )
Fig. 4.11
g( ) / (t/2 ) 1
g() = sin2 [(ba )t/2]/(ba )2
as a function of ba in
units of 2/t . For t it
becomes proportional to the
delta function
-3 -2 -1 0 1 2 3
( ab - ) / (2 /t)
Hence, the probability to find state |b populated at time t > 0 is given by

i( )t
2 e2 E02
2 e
ba 1 2
(abs)
wba (t)
= cb (t) = 2 |Dba |
2i(ba )
2 ba
e2 E02 2 sin ( 2 t) D02 t
= |
D ba | = |
Dba |2 g(). (4.59)
2 (ba ) 2 2 2
Figure 4.11 illustrates the characteristic frequency dependence g(). For large times
t 1/(ba ) it becomes arbitrarily narrow, and at the same time arbitrarily high
(t/2 at = ba ). This line profile is normalized such that g()d = 1, and
is given by

2 sin2 ( ba2 t) t
g() = (ba ). (4.60)
t (ba )2
Thus, g() becomes a representation of the D IRAC delta function in the limit
t .
(abs)
According to (4.59) wba (t) grows linearly with time. The approximation is of
course only valid for not too large fields and not too large times so that the general
assumption of perturbation time |cb (t)|2 1 is fulfilled for all b = a. At the same
time we are interested in the stationary state, i.e. in times t 1/(ba ).9
(abs) (abs)
We now divide wba (t) by time t to obtain the transition rate Rba = wba (t)/t,
i.e. the transition probability per unit time which in this approximation obviously
becomes independent of time. This is indeed the quantity one determines in an ab-
sorption experiment.
Finally, we have to account for the fact that in practice strictly monochromatic
electromagnetic waves do not exist. Each radiation has a certain bandwidth =
2 . We introduce this frequency dependence by replacing the intensity I in a
spectral range from to + d by I()d, where I() = cu() is the spectral
9 We shall see in Chap. 5 that these somewhat contradictory requirements can already be overcome
by accounting for a finite lifetime of excited states, and a full treatment will be given in Chap. 10,
Vol. 2 in the framework of the optical B LOCH equations.
intensity distribution. Thus, in (4.59) the square of the perturbation amplitude e2 E02
according to (4.56) has to be replaced by
2
e2 E02 e2 I ()d. (4.61)
c0
(abs)
The absorption rate dRba = wba (t)d/t for radiation of angular frequency be-
tween and + d is then given by
e2 I()
dRba = I ()|
Dba |2 g()d = 4 2 |
Dba |2 g()d, (4.62)
0 c 2
which is identical to the excitation probability of state |b from |a per unit time,
with the fine structure constant 1/137 according to (1.10).
So far we have tacitly assumed that the absorption occurs between two isolated
states |a and |b with sharp energies Wb and Wa , i.e. at a well defined sharp line of
angular frequency ba = (Wb Wa )/. The total transition rate Rba is then obtained
by integrating (4.62) over all frequencies:

I() 4 2
Rba = dRba = 4 2 |
Dba |2 g()d = Dba |2 I(ba ). (4.63)
|

A brief dimensional analysis

[Rba ] = 1 r 2 [I] = Enrg1 T1 L2 Enrg L2 T1 1/T1 = T1
shows that Rba is indeed a rate, i.e. gives the probability for an absorption process
per atom per unit of time. In the present perturbative approach we have assumed
that Rba tobs 1 during the relevant observation time tobs .
We see now explicitly that the excitation probability is proportional to the
F OURIER component of the spectral radiation density u() = I()/c at the tran-
sition frequency ba . The integration (4.63) is of course based on the assumption
that I() is constant over the absorption line. For typical classical radiation sources
this is trivially a correct assumption. In laser spectroscopy, however, this is not nec-
essarily true as we shall explain in Sect. 5.2.3 and more profoundly in Chap. 10,
Vol. 2.
For the integration in (4.63) we have assumed that the electromagnetic wave
interacts for a sufficiently long time with the system, so that the limit 1/ba t
in (4.60) is well approximated. With typical periods of the light field on the order
of femtoseconds (fs) this approximation may even be applied to excitation by ns
pulses often used in laser spectroscopy. But (4.62) is still valid for much shorter
interaction times: according to (4.60) this does, however, imply a broadening of the
line. We shall come back to this aspect in Sect. 6.1.7.
4.3.6 Absorption and Emission: A First Summary
To derive (4.63) for the absorption process we have evaluated the first exponent
in (4.51). The second becomes only relevant if ba < 0, i.e. for induced emission
(Wa > Wb ). This leads in complete analogy to the rate Rab | Dab |2 = |r ab e |2 for
induced emission, with identical prefactors. Because of hermicity r ab = r ab holds
and thus Rba = Rab . The rates for induced emission and absorption for a specific
transition between the states |b |a are identical.
In contrast to the general discussion in Sects. 4.2.34.2.5 we refer here to indi-
vidual substates described by a set of quantum numbers e.g. j m and not to energy
levels which may consist of several such (degenerate) sublevels. The key quantity
which characterizes an electric dipole transition is the dipole transition operator, ac-
cording to (4.56) Dab = r ba e, i.e. essentially the scalar product between dipole
matrix element D ba = er ba and polarization vector e with the dipole transition
amplitude

r ba = b|r|a = b (r)ra (r)d3 r = a|r|b = r ab . (4.64)
Note that r ba is a vector, specific for each system and each transition,10 and the
transition probabilities also depend critically on the polarization e of the radiation.
Before evaluating this in detail, we briefly summarize the results for electric dipole
transitions obtained so far.
We rewrite the probability (4.63) for inducing E1 transitions per unit time be-
tween the substates |a = |ja ma and |b = |jb mb with polarization e:
4 2
Rba = I (ba )|r ba e|2 = Rab (4.65)

4 2 c
= |r ba e|2 u(
ba ) = B(jb mb ; ja ma ; e)u(
ba ).

In full analogy to (4.26) and (4.27) we have introduced the substate and polarization
specific E INSTEIN B coefficients:
4 2 c
B(jb mb ; ja ma ; e) = |r ba e|2 = B(ja ma ; jb mb ; e). (4.66)

Again, these B coefficients are defined in respect of a spectral intensity distribution
I(ba ) = cu(
ba ) per unit angular frequency at the transition frequency ba =
|Wb Wa |/.
10 More precisely, er
ba is called dipole length matrix element. According to (H.25) the correspond-
ing dipole velocity matrix element is
p|a = iba me eb|r|a .
eb|
Both formulations lead to identical results if exact wave functions are used. For approximate solu-
tions (i.e. quite generally) significant differences may occur.
At this point the wisdom of the semiclassical approximation ends. Spontaneous

emission cannot be understood in this manner. However, we may apply some hand
waving arguments based on the E INSTEIN relation (4.38) with A/B = 3 /( 2 c3 ).
It was derived from P LANCKs radiation law and detailed balance between two iso-
lated levels, assuming a completely isotropic, unpolarized radiation field.
In the present discussion, absorption and induced emission between the substates
|a ja ma and |b jb mb are the result of a plane wave with well defined direction
k and polarization e. In contrast, spontaneous radiation can in principle be emitted
into 4 with two orthogonal polarization vectors. Thus, we divide (4.38) by 4 2,
multiply it by the solid angle element, and insert (4.66) for B. Thus, we may glean
for the probability of spontaneous emission with well defined polarization per unit
time into a solid angle d:
(spont) 3 d
dRab = B(ja ma ; jb mb ; e) = C|r ba e|2 d
2 c3 8
(4.67)
2 3
ba e2 ba3
= C r ab e d with C = = .
2c2 8 2 0 c3
A discussion of the corresponding E INSTEIN A coefficients will be postponed un-
til we have derived the angular distributions of the radiation patterns in Sect. 4.5.
A clean derivation of (4.67) will be given in Chap. 2, Vol. 2.
In summary, absorption and emission of electromagnetic radiation due to E1
transitions is characterized by
(abs) (ind) (spont)

Rba = Rab dRab (4.68)
2 2
Dba |2 = |r ba e|2 =
| Dab = r ab e .
According to (4.57) the matrix elements for transitions |ja ma |jb mb are
Dba =

r ba e and Dab = r ab e = Dba for absorption and emission, respectively. They
determine
whether the transition may occur at all (so called selection rules),
its dependence on polarization and angular distribution,
and its overall strength.
In the following sections these aspects will be discussed in detail, in particular the
key quantity r ba e in (4.68) will be evaluated for different polarizations and ge-
ometries. As we shall see, this may be somewhat involved since in general, r ba and
e are described most conveniently in different coordinate frames, e.g. in the atomic
(at) and photon (ph) frame, respectively, as introduced in Sect. 4.1.3.
However, we do not want to end this section without treating at least one par-
ticularly simple example. We choose excitation (and induced emission) by linearly
polarized light and exploit the fact that r ba = (xba , yba , zba ) is a well defined vec-
tor which may be represented by its Cartesian components. Let (r) be the angle
between r ba and the polarization vector e of the absorbed electromagnetic radia-

tion not to be confused with the polar angle of e in respect of the z-axis or with
the azimuthal angle k of z(ph) introduced in Fig. 4.3. For a transition between well
defined initial and final states |ja ma |jb mb one may now write (4.65) sim-
ply as
I(ba )
Rba = Rab = 4 2 |r ba |2 cos2 (r) (4.69)

with |r ba |2 = (xba
2 + y 2 + z2 ). This expression may be somewhat misleading in as
ba ba
far as depending on the alignment of the polarization vector not all components of
r ba contribute equally; e.g. for the simplest geometries where the linear polarization
is chosen parallel to any of the axes (4.69) becomes

I( ) |xba |2 for e x
ba
Rba = Rab = 4
2
|yba |2 for e y (4.70)

|zba |2 for e z.
In a typical absorption experiment with a given polarization vector, say e z, one

sums over all final states b and averages over all initial states a. For an isotropically
populated target one may show that the absorption rate then becomes independent
of polarization (see e.g. Sect. 4.6.3 or Appendix H.2.1).
Section summary
The semiclassical approach describes radiation induced processes by a classi-
cal, time dependent electromagnetic field and treats the interacting atomic or
molecular system quantum mechanically.
Time dependent perturbation theory is usually a good approximation for low
enough radiation intensities. Time dependent transition amplitudes (4.47) are
computed from the matrix elements of the perturbation operator U (t).
In the limit t , the transition amplitude (4.49) between initial (a) and
final state (b) is essentially the F OURIER component of Uba (t) at the transition
frequency ba .
The dipole approximation describes so called E1 transitions and is usually an
excellent approximation for a quantitative description of transition probabili-
ties as long as the wavelength of the radiation is a0 . It neglects the change
of the field with position over atomic distances. In the dipole length formu-
lation the interaction operator is E0 D where E0 is the field amplitude and

Dba = r ba e, with e being the unit polarization vector of the radiation.
Both exponential terms of the real field amplitude (4.1) matter: the
D exp(it) term is responsible for absorption, the
D exp(it) term for in-
duced emission. While the semiclassical treatment gives quantitatively cor-
rect results for absorption and induced emission, spontaneous decay requires,
strictly speaking, the use of QED to be discussed in Chap. 2, Vol. 2.
4.4 Selection Rules for Dipole Transitions
4.4.1 Angular Momentum and Selection Rules
As we have learned in Sect. 4.1.4 the photon has a spin sph = 1. But only two
projections ms q = 1 in respect of the direction of propagation, |sph q = |1 +1
and |1 1 are realized, describing LHC and RHC polarized light, respectively. One
speaks of positive and negative helicity. The fact that the third component q = 0 is
missing, reflects the transverse nature of free electromagnetic waves. According to
relativistic quantum field theory this is a consequence of the photon being a massless
particle.
Quite down-to-earth, the angular momentum properties of the photon allow us
to derive selection rules for E1 transitions without any computational efforts: we
simply apply the classical conservation law for angular momenta. For the system
atom + photon the total angular momentum and its projection onto a particular
direction in space is conserved during absorption or emission of a photon. Thus,
transitions may only be observed if the total angular momentum of the whole system

J is conserved. Denoting the total angular momentum operators of the atom with Ja
and J b in the lower and the upper state, respectively, the following rules in vectorial
form hold:
for absorption
J =Ja +
S ph Jb
(4.71)
for emission
J = J b J a + S ph .
For general use we associate J a and

J b with the atomic angular momentum quan-
tum numbers ja and jb , the corresponding projection quantum numbers being ma
and mb . They may be integer or half integer as discussed already in Sect. 1.1.1 and
1.9.5. For the model of a spinless electron (which has been assumed so far) one
identifies j . In the following chapters, the general rules developed below will
be extended to examples of increasing complexity.
With a photon spin quantum number sph = 1 and conservation of the total angular
momentum of the complete system, jb can only assume three values: ja + 1, ja
and ja 1. The vector diagram Fig. 4.12 illustrates this graphically. However, if
ja = 0 the transition is only allowed if jb = 1 and vice versa if finally jb = 0
initially only ja = 1 is possible. No transitions are allowed between states with
ja = jb = 0. Finally, for ja = 1/2, a dipole transition may only lead to final states
with jb = 1/2 or 3/2. One summarizes these selection rules in compact form as so
Fig. 4.12 Triangular relation Sph

between angular momenta of
initial and final atomic states Jb Sph

J a and J b , respectively and Ja Sph
Ja Ja
the photon spin S ph
Jb
Jb
4.4 Selection Rules for Dipole Transitions 197
Fig. 4.13 Absorption (top, z (at) mb=ma +1 mb=ma - 1

z (at) |b
Dba = r ba e) and emission
absorption
(bottom, Dab = r ab e ) of

|a
circularly polarized light ma ma
(left: LHC + ) and y (at) y (at)
(right: RHC ). The k || z k || z
selection rules m = 1 are
summarized in (4.74). Here x (at) in x (at) in
the atomic at and photon (ph) +(LHC) light -(RHC) light
coordinate frames are
identical, with the photon k mb k mb
wave vector k z(at) |b
emission
|a
out ma = m b - 1 out ma=mb +1
y (at) y (at)
x (at) x (at)
called triangular relation:
j = 0, 1 with ja , jb 0, but 0 0
abbreviated: (ja jb 1) = 1, (4.72)
while (ja jb 1) = 0 indicates that the triangular relation does not hold.
Clearly, the sum of the projections of the angular momenta before and after the
process have to be conserved as well. Hence, for dipole transitions the corresponding
selection rule is
q = m = mb ma = 0, 1. (4.73)
Different q (photon spin projection quantum number) correspond to different polar-

izations of the light (and possibly also to different coordinate systems). In respect of
the z(at) axis of the atom we have:
type of radiation q= absorption emission

m ja ma jb mb jb mb ja ma
+ (LHC) z(at) k 1 mb = ma + 1 ma = mb 1 (4.74)
(RHC) z(at) k 1 mb = ma 1 ma = mb + 1
(lin. pol.) z(at) E(r, t) 0 mb = ma ma = mb
Schematically this is illustrated in Fig. 4.13 for absorption and emission of pure
circularly polarized + (LHC) and light (RHC), propagating in +z(at) direction.
Figure 4.14 shows the situation for linearly polarized light.
Note that due to the transverse nature of electromagnetic waves the selection rule
m = 0 refers to a coordinate system perpendicular to that in the two other cases:
for linearly polarized light it is convenient to choose the photon frame (ph) in respect
Fig. 4.14 Emission and mb= ma linearly

absorption of linearly z (at) |b polarized light
polarized light ( transitions) |a
with m = 0. The coordinate ma
k
system is chosen such that the y (at)
(linear) polarization vector
e 0 || z (at) absorption
e z(at) x (at)
mb
z (at) |b
|a
ma= mb
y (at)
emission
x (at) k
e 0 || z (at)
of the atomic coordinate frame (at) with the electric field vector E z(ph) z(at) and
the light travelling in the x (at) y (at) plane, as illustrated in Fig. 4.14.11
4.4.2 Transition Amplitudes in the Helicity Basis
The selection rules just outlined are necessary but not sufficient conditions for E1
transitions. To obtain quantitative expressions for the transition probabilities sum-
marized in (4.68) one has to evaluate | Dba |2 = |r ba e|2 explicitly. To do so, a spe-
cific coordinate system has to be chosen. For the unit polarization vector e we have
discussed suitable basis sets in Sect. 4.1.2. In principle, one may use Cartesian (xyz)
or polar coordinates (r ). Depending on the geometry of the experiment one or the
other choice may be advantageous. For linear polarization we have shown by way
of example that a Cartesian basis with e parallel to one of the axes leads directly
to the particularly simple result (4.70). However, since the angular part of atomic
wave functions is usually presented in polar coordinates, for a general description
the corresponding helicity basis (introduced in Sect. 4.1.2) is more appropriate for
expressing the polarization vector.
We introduce spherical components rq of the position vector r. With x =
r sin cos , y = r sin sin and Table B.1 one easily verifies
11 This is possible as long as no other preferential direction is enforced by a particular experiment.

More generally, linearly polarized light propagating in z(at) direction
may be written with (4.15)
as a linear combination of + and light (cos = sin = 1/ 2). Absorption or emission of
this light involves superpositions of m = 1 transitions (sometimes called ) and leads to the
generation of linear combinations of states as we shall explain in more detail in Sect. 4.5.2 and
Sect. 4.7.1.

4
r1 = (x iy)/ 2 = r Y11 (, ) = rC11 (, )
3
(4.75)
4
and r0 = z = r Y10 (, ) = rC10 (, ).
3
With (4.5)(4.7) the position operator r (a real valued vector) becomes:

1
1

r = r = r = xex + yey + zez =

rq eq =r C1q (, )eq (4.76)
q=1 q=1

1
1
1

r C1q (, )eq = r (1)q C1q (, )eq = (1)q rq eq = r .
q=1 q=1 q=1
We write the transition amplitudes between angular momentum substates

|b jb mb |a ja ma for a specific polarization q with mb |rq |ma as abbre-
viation if no misunderstanding is possible
b jb mb |rq |a ja ma = b |r|a jb mb |C1q |ja ma := mb |rq |ma

jb mb
jb 1 ja
= rba (1) 2jb + 1 jb C1 ja ,
mb q ma
(4.77)
where the radial transition matrix element is

b |r|a = rba = rab = Rb (r)Ra (r)r 3 dr. (4.78)
0
Explicit evaluation of these matrix elements is greatly facilitated by the W IGNER -

E CKART theorem (C.7)(C.9), which we have used in the second line of (4.77). It
allows us to factorize the overall transition probabilities, not only into radial and
angular part, but also into a contribution from geometrical effects (dependence on
projection quantum numbers m, q), and from total angular momenta.12
We may now evaluate the dipole transition matrix elements (4.57). For absorp-
tion of light with polarization e we find

1
r ba e = b|r|a e = b jb mb |r|a ja ma e = mb |rq |ma eq e
q=1

1
= b |r|a jb mb |C1q |ja ma eq e. (4.79)
q=1
12 More details for evaluating these matrix elements are made available in Appendix C.
Note that emission differs from absorption as described by (4.57). The polarization
vector e (creation of a photon) replaces e (annihilation of a photon):

1
r ab e = rab
ja ma |C1q |jb mb eq e = (r ba e)
q=1

1
= rab (1)q ja ma |C1q |jb mb eq e . (4.80)
q=1
However, since Dab = r ab e = (r ba e) =

Dba , and since transition rates depend
only on the absolute squares of the matrix elements, the probabilities for absorption
and emission between well defined substates |b jb mb |a ja ma are the same
if the photon polarization q is the same. We come back to this in Sect. 4.4.3.
The magnitude of the radial matrix element rba determines the overall strength
of a dipole allowed transition. Explicit analytical formulas are available only for
atomic hydrogen and H like ions (see Appendix C.5). In general rba has to be eval-
uated numerically from the radial wave functions Ra (r) and Rb (r). The latter are
usually obtained from numerical integration of the radial S CHRDINGER equation
as detailed in Sect. 3.2.
The angular part jb mb |C1q |ja ma of the transition amplitude (4.77) is of a more
general nature. It depends only on the respective angular momentum quantum num-
bers of the system studied and determines the selection rules for E1 transitions. In
Sect. 4.6 some general rules and expressions will be given for the transition prob-
abilities (A and B coefficients). In the following we shall explicitly recover and
quantify the selection rules (4.72) and (4.73), which were derived there from the
concept angular momentum conservation.
4.4.3 Transition Matrix Elements and Selection Rules
We now discuss the simplest case, assuming that the light polarization is given by a
basis vector in the atomic frame, e := eq . All transition rates (4.68) are now propor-
tional to |
Dba |2 = |r ba eq |2 . With (4.79) this implies
2 2
b |r|a jb mb |C1q |ja ma .
(abs) (ind) (spont)
Rba = Rab dRab (4.81)
The selection rules for any E1 transition are essentially contained in the angular
part of the matrix elements. With (4.77) the angular part of the transition probability
(4.81) is
2
jb 1 ja
jb mb |C1q |ja ma 2 = (2jb + 1) jb C1 ja 2 . (4.82)
mb q ma
Whether a dipole transition is possible at all between two levels with total angular
momenta ja and jb depends on reduced matrix element jb C1 ja . It does not
vanish if and only if the triangular relation (4.72) holds, as already gleaned from
angular momentum conservation.
Whether transitions can occur between specific sublevels depends on the 3j sym-
bol in (4.82). It can only be finite if the sum of its projection quantum numbers in
the second row vanishes (see Appendix B.2). Hence, (4.73) is confirmed as addi-
tional selection rule. The underlying conservation of total angular momentum has
been illustrated in Fig. 4.13.
Note: We point out again the differences between absorption and emission as a
consequence of replacing e in (4.79) by e to obtain (4.80) with r = r (real valued
vector). E.g., as spelled out in (4.74), + light absorption increases the projection
quantum number in the nal state by q = 1: mb = ma + 1; in contrast, emission
of + decreases of m in the nal state: ma = mb 1. But in both cases the se-
lection rule (4.73) holds per denition. This is a consequence of Dab = r ab e =

(r ba e) =
Dba : the transition probabilities depend only on the absolute squares of
the matrix elements. A corresponding consideration holds for light with q = 1.
For linearly polarized light, according to (4.74) and Fig. 4.14, the unit polar-
ization vector becomes e = e0 so that with (4.81) and (4.82) the transition rate is
simply proportional to
2
jb 1 ja
for E z(at) mb = ma . (4.83)
mb 0 ma
Thus, with such linearly polarized light one observes transitions with m = 0, often
called transitions. In contrast to (4.74) the E vector of the photon and not its wave
vector k points into the z(at) direction. The propagation angle within the x (at) y (at)
plane does not play a role in this case.
For reference we also note that

1

|r ba |2 = mb |rq |ma 2 = |xba |2 + |yba |2 + |zba |2 . (4.84)
q=1
The explicit evaluation of the matrix elements in (4.81) may become quite com-
plicated, depending on the angular momentum coupling schemes in which the states
|ja ma and |jb mb are described. In later chapters we shall have to discuss this in
more detail for various examples.
4.4.4 An Example for E1 Transitions: The H Atom
For the moment we just evaluate the most simple case for single electron transitions
between uncoupled, pure orbital angular momentum states, |na a ma and |nb b mb ,
ignoring the electron spin. Inserting (C.28) into (4.81) one obtains
Wn / eV
n =0 =1 =2 =3 =4
0
8
6
5s 5p 5d 5f 5g
-0.85 4 4s 4p 4d 4f
-1.51 3
3s 3p 3d
-3.40 2
2s 2p
-13.6 1 1s
Fig. 4.15 G ROTRIAN diagram for the H atom illustrating the selection rule = 1 in E1 tran-
sitions (shown are only emission lines, the electron spin is neglected)
2 2
Rba na |r|nb a ma |C1q | b mb
2
= na |r|nb (2 a + 1)(2 b + 1)
2 2
a 1 b a 1 b
ma mb +q ( a 1 b ) . (4.85)
ma q mb 0 0 0
The first 3j symbol confirms again the selection rules (4.72) and (4.73). The second
3j symbol, however, adds an additional selection rule: since its projection quantum
numbers all vanish, a + 1 + b must be even, lest it disappears according to (B.49).
Thus with the constraint ( a 1 b ) = 1 a new selection rule holds:
a = b 1. (4.86)
The physical origin of this rule is parity conservation which is discussed in more
detail in Appendix D. Briefly, the overall parity of the system has to be conserved,
i.e. its symmetry in respect of inversion at the origin has to be the same before and
after absorption or emission of the photon. The angular part of the wave functions
for such pure states is completely described by spherical harmonics, their parity
being (1) . The photon which is represented by C1 has negative parity. The parity
of the whole system before absorption of the photon is thus (1) a +1 while it is
(1) b after the photon has been absorbed. Thus, for parity conservation a + 1 + b
must indeed be even.
Up to now, we have used pure states for the hydrogen and alkali atoms. As an
example, Fig. 4.15 shows the term scheme of the H atom with some E1 transitions.
Specifically, for an s p transition three processes are allowed: |0ms |1mp with
ms = 0and mp = 0 or 1. According to (C.44) all three angular matrix elements
are 1/ 3, independent of the substates, so that
|ns |r|np |2
for an = 0 = 1 transition: Rab . (4.87)
3
4.5 Angular Dependence of Dipole Radiation 203
The electron spin, so fare neglected, will modify details of the term scheme
Fig. 4.15. The corresponding selection rules will be discussed in Chap. 6. In Ap-
pendix C one finds compact formulas for the evaluation of (4.82) in the general case.
In Chap. 7 we shall see, however, that transitions between states with angular mo-
mentum quantum numbers 0 and 1 are characteristic for closed shell multi-electron
systems. They occur e.g. among so called singlet states where the spins of several
electrons compensate each other. E1 transitions may then indeed be described by
the orbital angular momenta of the optically active electron, e.g. p s. In the
remainder of the present chapter we shall identify jb = 1 ja = 0 processes as
1 P 1 S transitions within a singlet system, indicated by the superscript, while
1 0
the subscript stands for j .
To actually calculate the transition probabilities, in addition to the angular part
of the matrix elements, which we have so far discussed in detail, one also has to
compute the radial transition matrix elements. For the H atom Appendix C.5 sum-
marizes the necessary formulas. A rather comprehensive collection of spontaneous
transition probabilities for many atoms of the periodic system is found at K RAMIDA
et al. (2013).
Section summary
Selection rules for light induced E1 transitions may directly be inferred from
For the total angular moments the triangular relation (ja jb 1) = 1 must hold,
the photon having a spin 1.
For the projection quantum numbers q = m = mb ma = 0, 1 reflects the
photon polarization q = 1 (with respect th z(at) k) and q = 0 (with respect
th z(at) E).
Quantitatively, the transition probabilities are proportional to
2
jb 1 ja
2
rab (2jb + 1) jb C1 ja 2 .
mb q ma
4.5 Angular Dependence of Dipole Radiation
We now turn to the angular characteristics for absorption and emission of electro-
magnetic radiation in dipole transitions. We make explicit use of the coordinate
frames introduced in Sect. 4.1.3: the atomic frame (at) for describing the atom and
the photon frame (ph) in which the light is best described. The latter is character-
ized by the polar (k ) and azimuthal angle (k ) of the wave vector k as shown in
Fig. 4.3. Section 4.5.1 presents a first somewhat heuristical semiclassical ap-
proach by interpreting the classical oscillator as dipole transition matrix element
(at)
r ab . In Sect. 4.5.2 we shall show that this may be cast into an exact quantum me-
chanical formulation with essentially identical results, now including all necessary
numerical prefactors.
4.5.1 Semiclassical Picture
To express the classical electron oscillator r(t) = r (at) exp(iba t) described in

Sect. 4.2.4 in quantum mechanical terms we assume its amplitude r (at) to be propor-
(at)
tional to the dipole transition matrix element r ab and reintroduce the time depen-
(at)
dence of the stationary initial and final states (2.16). Multiplying r ab from the left
with eia t and from the right with eib t , and using the representation (4.79) in the
helicity basis we obtain the quantum mechanical equivalent to a classical oscillator:
(at)
r(t) eia t r ab eib t = ma |r0 |mb e0 eiba t
(at)
(4.88)
(at) i t
+ ma |r1 |mb e1 e ba
(at)
+ ma |r1 |mb e+1 eiba t .
The three components directly reflect the three types of classical oscillators illus-
trated graphically in Fig. 4.3. They correspond to the three types of transitions which
we have already summarized in (4.74):
For q = 0 ( light) the electron oscillates linearly in z(at) direction with an am-
plitude ma |r0 |mb . From the classical picture we expect zero radiation in z(at)
direction and maximum intensity emitted into the x (at) y (at) plane with linear po-
larization z(at) .
The terms with q = +1 and q = 1 represent an electron on a circular orbit in the
x (at) y (at) plane. The emitted light may propagate into the whole solid angle 4 .
Maximum intensity and fully circularly polarized light is intuitively expected
for propagation into +z(at) direction, while light detected in the x (at) y (at) plane
will be linearly polarized z(at) axis.
More quantitatively, one may glean electric field amplitudes from (4.29) and (4.30)
(at)
by projecting r ab onto the x (ph) y (ph) plane. For light emitted from a linear oscil-
(at)
lator with amplitude ma |rq |mb e0 one reads from Fig. 4.3
(ph) (ph)
Ey = 0 and Ex ma |r0 |mb sin k (4.89)
independent of k . Hence, light is always linearly polarized, with e x (ph) . The

emitted intensity distribution
(ph) 2 (ph) 2 2
I (k ) Ey + Ex ma |r0 |mb sin2 k (4.90)
is of doughnut type as illustrated in Fig. 4.16.

(at)
By the same arguments, the circular dipoles ma |r1 |mb e1 oscillate in the
x (at) y (at) plane. With (4.7) and Fig. 4.3 the projections onto the x (ph) y (ph) plane are
(ph) ma |r1 |mb (ph) ma |r1 |mb

Ey i and Ex cos k . (4.91)
2 2
z (at) k
k= 90 k< 90
E,
k
x (at) y (at)
E, k
Fig. 4.16 Angular distribution of dipole radiation from a linear oscillator ( light)
I (k , ) sin2 k . For two different directions of the wave vector k the alignment of the polar-
ization vector is indicated (red, full and dashed vector arrows, respectively)
z (at) k
E, +
k
y (at )
k
x (at) E,
k
Fig. 4.17 Angular distribution of dipole radiation from a circular oscillator ( light)
I (k , ) (1 + cos2 k )/2. The polarization vector is indicated for two different directions of
observation: at 0 < k < /2 (full red ellipse with vector arrow), corresponding to elliptically
polarized light, and for k = /2 (dashed red arrow) corresponding to linear polarization
Hence, the angular distribution for light emitted in q = 1 transitions is
(ph) 2 (ph) 2 2 (1 + cos2 k )

I (k ) Ey + Ex ma |r1 |mb , (4.92)
2
again independent of k , as illustrated in Fig. 4.17. In contrast to light, the po-
larization of light changes with k . The field amplitudes (4.91) imply that full
circular polarization is only observed for k = 0 and , i.e. in z(at) direction.
(ph)
The cos k factor reduces the Ex component, and in general elliptically polarized
light is observed at angles in between these limits as indicated in Fig. 4.17. One
(ph)
special case is k = /2 (half maximum intensity) where Ex = 0 so that light
becomes linearly polarized with its e vector aligned in the x y (at) plane z(at) .
(at)
In Fig. 4.18 the standard geometry for observation of the different transition types
is shown, as e.g. realized for observing the so called normal Z EEMAN effect (see
Sect. 2.7). The three cases q = 0, 1 may in principle be observed independently
since the corresponding atomic lines are split in the magnetic field, with frequencies
0 and 1 = 0 L /2 for transitions with m = q = 0 and 1, respectively.
(a) (b) - +
(c)
-
z (at ) +
-1 0 1
+1 spectrum
1P
1 0
mb = -1
-1 0 1
Bex field (d)

1S mg = 0
0 spectrum
magnet
-1 0 1
Fig. 4.18 Normal Z EEMAN effect as realized in a 1 S0 1 P1 transition; (a) energy levels in an
external magnetic field B ex z(at) , (b) experimental setup, (c) spectrum from light emitted at
k = 0, (d) spectrum from and light emitted at k = /2
We emphasize again that the level scheme Fig. 4.18(a) should not be misinter-
preted: the arrows indicating transitions with 1 and 0 refer to different coordinate
systems. The external magnetic field defines here the direction of z(at) B ex ; emis-
sion parallel to that axis (i.e. k z(at) B ex ) implies transitions with m = q = 1
( light) only; in contrast, m = q = 0 transitions imply light, linearly po-
larized with the electric field E B ex , emitted into any direction except in z(at)
direction.
4.5.2 Angular Distributions from Quantum Mechanics
What has been deduced above from a more or less classical picture also follows from
a rigorous quantum mechanical treatment. The angular and polarization character-
istic of emission and absorption rates are obtained by inserting (4.80) and (4.79),
respectively, into the general expression (4.68). Typically, the matrix element r ba
and the polarization vector e are described in different coordinate frames, most con-
veniently in the atomic (at) and photon frame (ph) as illustrated in Fig. 4.3. The
(spont) (ind) (abs)
transition rates dRab , Rab , and Rba between two angular momentum basis
states |jb mb |ja ma are then proportional to
1 2
(at) (ph) 2 (at) (ph) 2

r e = r e = mb |rq |ma eq(at) e(ph)
ba ab
q=1
1 2

(ph)
= b |r|a jb mb |C1q |ja ma eq e .
(at)
(4.93)

q=1
We emphasize that the unit vector e(ph) represents the polarization of the
detected radiation for a spontaneous transition between |jb mb |ja ma ,

inducing radiation for an induced transition between |jb mb |ja ma ,
absorbed radiation for absorption between |jb mb |ja ma .
Note that irrespective of the direction of the above arrows (4.93) holds as writ-
ten (see also the note on p. 201). To obtain practical relations for the angular dis-
(at)
tributions, such as (4.90) and (4.92), we have to express eq and e(ph) in the same
coordinate system. Which of these frames is more convenient depends on the exper-
imental situation. For illustration we carry this out explicitly as a kind of exercise
on coordinate rotation, using the rules explained in Appendix E.
Angular Distribution of Spontaneous Emission

We first consider spontaneous emission. We are interested in the angular distribu-
tion and polarization of radiation emitted from a well defined transition in the (at)
(at)
system. Thus we have to express eq in the photon frame. We rotate the photon sys-
tem (old) through the E ULER angles into the atomic system (new). No initial
rotation around z(ph) is required as seen in Fig. 4.3, i.e. = 0. Rotation around y (ph)
through = k makes z z(at) . Finally, we have to rotate the frame around z(at)
through = k so that x is transferred into x (ph) . Thus, the atomic basis vectors
(ph)
are expressed in terms of basis vectors eq according to (E.14):
(ph)
q =e
iqk
e(at) dq1 q (k )eq . (4.94)
q
For the three standard transitions m = q = 0, 1 we insert the rotation matri-

ces dq1 q (k ) according to (E.12) into (4.94) and obtain explicit expression for the
atomic basis in the photon system according to Fig. 4.3 with z(ph) k:
m = q = 0 ( component)
(at) sin k (ph) (ph) (ph)

e0 = e1 e1 = sin k ex . (4.95)
2
m = q = 1 ( + component)

k (ph) k (ph)
e1(at) =e ik
cos2 e1 + sin2 e1 . (4.96)
2 2
m = q = 1 ( component)

k (ph) k (ph)
e1 = eik sin2 e1 + cos2 e1 .
(at)
(4.97)
2 2
Note that in the photon frame only the q = 1 basis vectors are considered, with
polarization perpendicular to the photon wave vector k. Only these are of physical
(at) (ph)
relevance in the scalar products (4.93) eq e(ph) . The e0 component is dropped
since it would indicate (un-physical) linear polarization in z(ph) direction. Hence,
(at)
eq is no longer a unit vector and its magnitude depends on k . We may, however,
express it in terms of an angle dependent normalization factor fq (k , k ) and the
(ph)
general unit vector (4.15) for elliptically polarized light in the photon frame eel :
(ph)
e(at)
q = fq (k , k )eel , so that (4.98)
(at)
(ph)
r ab e(ph) = ma |rq |mb fq (k , k )eel e(ph) . (4.99)
q
Assuming we can detect one specific transition q = m only, and register all light
(ph)
emitted from this transition without further discrimination, i.e. e(ph) := eel , the
emission rate becomes proportional to
(at) (ph) 2
r e = ma |rq |mb 2 fq (k , k )2 . (4.100)
ab el
Specifically, according to (4.95) for q = 0 ( component) the emitted light is

(ph) (ph)
always linearly polarized x (ph) so that eel := ex and f0 (k , k ) = sin k . Thus,
with (4.100) the rate (4.67) for spontaneous emission into a solid angle d due to a
transition |ja ma |jb mb is
3
ba
dR
(spont)
= mb |r0 |ma 2 sin2 k d. (4.101)
2c 2
One easily verifies that this probability has indeed the dimension T1 . The intensity
of radiation emitted per atom seen at a detector of an area A in a distance R is
obtained by multiplying (4.101) with the photon energy ba , dividing it by A,
and identifying the solid angle with d = A/R 2 :
4
(spont) ba mb |r0 |ma 2 sin2 k .
I = 2 2
(4.102)
2c R
This angular distribution corresponds exactly to the previously derived semi-
classical pattern (4.90) shown in Fig. 4.16. The angular part jb mb |C1q |ja ma of
the matrix element mb |r0 |ma according to (4.77) is typically on the order of 1,
while the radial matrix element b |r|a may vary over a wide range depending on
the size of the atomic orbitals and the overlap and symmetry of the wave functions.
We now turn to the circular components of the emitted radiation pattern
(q = 1). The atomic basis vectors projected onto the photon frame are

e1 = f1 eel = f1 (k , k ) ei cos e+1 ei sin e1 .
(at) (ph)
(4.103)
(ph)
The parameters and of the unit elliptic basis vector eel (4.15) are found from
a comparison of (4.103) with (4.96) and (4.97). We see immediately that = /2
(the main axis of the polarization ellipse is y (ph) ); with cos2 +sin2 = 1 and with
the trigonometric identity 2(cos4 (k /2) + sin4 (k /2)) = (1 + cos2 k ) we derive
(1 + cos2 k )1/2
f1 (k , k ) = ieik (4.104)
2
as normalizing prefactor.13 The ellipticity angle is given by:

" 2
2 cos (k /2) if q = 1
cos = (4.105)
(1 + cos2 k )1/2 sin2 (k /2) if q = 1.
For emission at k = 0, i.e. along the z(at) axis we have cos = 1 or = 0 and
sin = 0 or = 1 for q = +1 or 1, respectively. As discussed already, this implies
emission of pure circularly polarized +(LHC) or (RHC) light. In contrast,
at k = /2 one finds cos = sin = 1/ 2, corresponding to light emission with
linear polarization vector in the x (at) y (at) plane (see Fig. 4.3).
The angular distributions for m = q = +1 and m = q = 1 follow naturally
by inserting f1 (k , k ) from (4.104) into (4.100) and this in turn into (4.67). Again,
(ph)
the overall intensity of the emitted light is found by setting e(ph) := eel . We finally
obtain for q = 1 transitions
3
mb |r1 |ma 2 1 + cos k d,
ba 2
(spont)
dR = (4.106)
2c2 2
and the intensity per atom at a distance R is
4
(spont) ba mb |r1 |ma 2 1 + cos k .
2
I = (4.107)
2c2 R 2 2
This confirms again the angular characteristics (4.92) of the radiation found by the
semiclassical model as illustrated Fig. 4.17.
We may integrate the radiation characteristic (4.101) and (4.106) over the full
solid angle and thus derive the overall spontaneous emission probability for one
specific transition. Both, the sin2 k distribution (4.103) for the -components as
well as the (1 + cos2 k )/2 distribution (4.107) for the transitions lead to the same
factor 8/3:

8 1
sin k d = 2
2
sin k dk =
3
= 1 + cos2 k d. (4.108)
4 0 3 4 2
Hence, from (4.101) as well as from (4.106) the integral spontaneous emission prob-
ability for a specific transition with ma + q = mb becomes:
13 The overall phase factor i exp(ik ) is here of no significance for measurable quantities, since
these are proportional to the absolute squares of the matrix elements.
3
(spont) 4 ba mb |rq |ma 2
Rq = A(ja ma ; jb mb ; q) = 2
3 c
3
4 ba b |r|a 2 jb mb |C1q |ja ma 2 .
2
(4.109)
3 c
Equations (4.95)(4.109) describe spontaneous emission processes between well
defined states |ja ma |jb mb . This situation is realized experimentally if a spe-
cific initial state is prepared and only one specific final state is detected. Then the
angular distributions (4.102) and (4.107) are observed. One method (by far not the
only one) to prepare and detect such specific transitions is the Z EEMAN effect as
illustrated in Fig. 4.18.
We finally mention that a more general theory of radiation emitted from excited
atoms will be presented in Sect. 9.4, Vol. 2 as an interesting example for applications
of the density matrix.
Angular Dependence for Absorption and Induced Emission

In contrast to spontaneous emission where the polarization and angular distribution
was seen as the result of a specific transition, one may also be interested in excitation
(or induced emission) by a specific polarization e(ph) . In which case we have to
express e(ph) in the atomic frame. To do so, one rotates the atomic system (at) into
the photon system (ph). One reads from Fig. 4.3 that first the atomic frame has to be
rotated around its z(at) axis through = k and then around the new y axis through
= k , while a final rotation around the new z(at) axis is not necessary so that = 0.
Inserting these E ULER angles into (E.14) we obtain relations inverse to (4.94):

eiqk dq1 q (k )eq .
(ph) (at)
eq = (4.110)
q
Explicit, general expressions similar to (4.95)(4.97) will be presented in Sect. 4.7.

We shall derive there the general case in which absorption or induced emission from
a well defined initial substate, say |ja ma , leads to a coherent superposition of final
states |jb mb .
Presently we specialize to the angular dependence of induced transition rates
for absorption (and induced emission) between specific substates |jb mb |ja ma
(with q = mb ma ) by a well defined polarization q = 0 or 1 of the light. In this
case we have to evaluate
I(ba ) 2 (ph) 2
Rba = 4 2 mb |rq |ma eq(at) eq = Rab . (4.111)

Without going into details of the derivation we just report a few examples:
1. Excitation of a m = q = +1 transition with LHC or RHC light ( ) implies

(ph)
e(ph) := e1 . We obtain
" 4
I(ba ) 2 cos (k /2) for + light
Rba = 4 2 mb |r+1 |ma
sin4 (k /2) for light.
2. Similarly, the excitation probability for a m = q = 1 transition induced with

+ or light as a function of polar angle k is given by
" 4
I(ba ) 2 sin (k /2) for + light
Rba = 4 2 mb |r1 |ma
cos4 (k /2) for light.
3. And for excitation of a m = q = 0 transition with circularly polarized light the

angular dependence is
I(ba ) 2 sin2 (k )
Rba = 4 2 mb |r0 |ma .
2
4. Finally, if the same transition can also be induced by linearly polarized light
(ph)
x (ph) , i.e. if e(ph) := ex . We find
I(ba ) 2
Rba = 4 2 ma |r0 |mb sin2 (k ),

which obviously is a more efficient process. The pattern is the same as shown in
Fig. 4.16.
We must emphasize, however, that it is nontrivial to study such specific transitions:

the initial substate would have to be prepared and the final substate to be detected
selectively. While in principle possible, standard spectroscopic experiments average
over all initially populated substates and sum over all accessible final substates as
discussed in the following section.
Section summary
We have derived the angular distributions of radiation emitted for q = 0 (q =
mb ma ) transitions and for q = 1 from a semiclassical picture as well as
from quantum mechanical rates for E1 transitions.
We should memorize these distributions according to Figs. 4.16 and 4.17.
The polarization depends on the emission angle k with respect to the z(at)
axis. For q = 0 transitions, only linearly polarized light ( light) is emitted,
with its polarization vector parallel to z(at) for k = /2. For q = 1 transi-
tions, at k = 0 pure circularly polarized light ( light) is emitted, while at
k = /2 one finds linearly polarized light with polarization vector perpen-
dicular to z(at) . For other emission angles the light is elliptically polarized.
The three components of radiation may be separated experimentally, e.g. by
exploiting the Z EEMAN effect as illustrated for a 1 S0 1 P1 transition in
Fig. 4.18.
For induced transitions one may derive the angular dependence of rates for
transitions induced between specific sublevels by a specific polarization. The
general case is more complex however, as we shall see in a moment, it
may be greatly simplified for an initially isotropic population of all sub-
states.
4.6 Strength of Dipole Transitions
4.6.1 Line Strength
According to (4.68) the key parameter for all absorption and emission processes is
|r ba e|2 . It contains the necessary information on polarization and angular charac-
teristics for E1 transitions between specific substates. In contrast, to characterize the
overall strength of a dipole transition between levels ja and jb with several degen-
erate ma and mb substates one introduces (see also Appendix H.2) a symmetrically
defined line strength with the dimension L2 :

S(jb ja ) = |r ba |2
mb ma

= b jb mb |rq |a ja ma 2
mb ma q
2
= (2jb + 1)b |r|a jb C1 ja 2 (4.112)
2
= (2ja + 1)b |r|a ja C1 jb 2 S(ja jb ).
Here we have used (4.84) with (4.77), (4.82) and the 3j orthogonality relation
(B.41). The radial matrix element b |r|a is given by (4.78). We recall that b and
a represent all quantum numbers needed to characterize the radial wave function.
We also note that ja and jb may refer to various kinds of the angular momentum
quantum numbers, representative for the states under discussion in the simplest
case these may be orbital angular momenta or L, but could also be total angular
momenta as we shall discuss in Chaps. 6 and 9.
The line strength is a useful reference quantity. However, when describing spe-
cific cases one has to distinguish between types of radiation (polarized or unpolar-
ized, unidirectional or averaged over all solid angles) as well as between types of
transitions (selected substates, sums over final states, averages over initial states).
This will be discussed below in terms of E INSTEIN coefficients. Figure 4.19 defines
these for transitions between the substates |jb mb and |ja ma with polarizations
characterized by q = mb ma .
jb mb
B( ja ma ; jb mb ;q)
A( ja ma ; jb mb ;q)
B( jb mb ; ja ma ;q)
ja ma
Fig. 4.19 E INSTEIN coefficients for absorption, B(jb mb ; ja ma ; q), for induced emission,
B(ja ma ; jb mb ; q), and for spontaneous emission, A(ja ma ; jb mb ; q), between substates of an up-
per and a lower level characterized by the angular momentum quantum numbers jb and ja , respec-
tively
4.6 Strength of Dipole Transitions 213
4.6.2 Spontaneous Transition Probabilities
The spontaneous transition probability (4.109) for one specific transition |ja ma
|jb mb and polarization q is now rewritten with (4.82), (4.77) and (4.112)
A(ja ma ; jb mb ; q)
3 2
4ba 2 jb 1 ja
=
b |r|a (2jb + 1) jb C1 ja 2
3c2 mb q ma
3
4ba 3 2
= mb |rq |ma 2 = 4ba jb 1 ja
S(jb ja ). (4.113)
3c2 3c2 mb q ma
It has the dimension T1 . The 3j symbol accounts for all dependencies on angular
momenta and polarization while all specific properties of the atom (or molecule) are
included in S(jb ja ).
Total decay probability and spontaneous lifetime of the upper level are deter-
mined by summing over all final states and polarizations:
1
A ab = = A(ja ma ; jb mb ; q)
ja jb ma q
jb 2
3 3
(4.114)
4ba 1 ja 4ba S(jb ja )
= S(jb ja ) = .
3c2 mb q ma 3c2 (2jb + 1)
m q a
Here we have again used the orthogonality relation (B.41) for the 3j symbols. It is
important to note that the spontaneous lifetime ja jb is independent of the quantum
number mb of the sublevel in the excited state!
Explicitly, the prefactor in the above equations is given by
3
4ba 32 3 c 19 3 s
2 7.235 108 m
= = 1.083 10 ba 2 = . (4.115)
3c2 33ba m 3ba s
For practical use we insert the line strength (4.112) into (4.114) and introduce atomic
units a0 and Eh :
4 3 2
2
A ab = 2 3 Wba a0 b |r/a0 |a jb C1 ja 2
3c
4 5 me c2 2
= (Wba /Eh )3 b |r/a0 |a jb C1 ja 2 (4.116)
3
2.1420 1010 2
= (Wba /Eh )3 b |r/a0 |a jb C1 ja 2 .
s
The reduced matrix element jb C1 ja is typically on the order of 1. For the sim-
plest case, j = , it is given by (C.30), in the general case it may be evaluated as
explicated in Appendix C.3. The radial matrix element expresses the genuine atomic
physics: it is determined by the overlap between initial and final state wave func-
tions, weighted with the radius.
The dependence of A on the nuclear charge Z is particular interesting. For hy-
drogen like ions we have Wba Z 2 and r 1/Z, so that the spontaneous lifetime
ja jb = 1/A ab Z 4 decreases with the 4th power of the nuclear charge, a result
which is important for atomic X-ray physics.
With (4.114) one may derive the line strength directly from the experimentally
accessible natural lifetime of the excited state:
3c2 (2jb + 1) 1
S(jb ja ) = 3
. (4.117)
4 ba ja jb
On the other hand, with (4.113) the individual transition probabilities between indi-
vidual substates may be written as
2
jb 1 ja 1
A(ja ma ; jb mb ; q) = (2jb + 1) , (4.118)
mb q ma ja jb
and (4.77), the transition amplitude between substates, may be written

jb 1 ja
mb |rq |ma = (1)jb mb S(jb ja ). (4.119)
mb q ma
We also recall at this point, that even though according to (4.114) the excited state
lifetime is identical for all excited substates the angular radiation characteristic and
the polarization of the emitted light depends on mb and q as described by (4.101)
(4.109).
Experimentally one often detects the fluorescence of an atom without polariza-
tion and final state analysis. The respective transition probability is obtained by
summing over all polarizations q according to (4.101)(4.106) and over all final
states:
(spont)
dRja jb mb ( ) = d 3 S(jb ja ) (4.120)
2c2 ba
*+ 2 ,
jb 1 ja
2
jb 1 ja 1 + cos2 k
+
mb 1 ma mb 1 ma 2
ma
2 -
jb 1 ja
+ 2
sin k .
mb 0 ma
Usually, this still needs to be averaged over all populated initial states mb and de-
scribes in general a non-isotropic angular distribution. Only if all initial excited
states |jb mb are equally populated ( (2jb + 1)1 ) the 3j orthogonality relation
(B.41) may be applied and finally leads to an isotropic distribution:
(spont) 1 d
dRtot = . (4.121)
jb ja 4
4.6 Strength of Dipole Transitions 215
The general case of radiation from an anisotropically populated atom will be treated
in Sect. 9.4, Vol. 2.
4.6.3 Induced Transitions
We now consider induced processes, i.e. absorption and stimulated emission. For
transitions between specific substates induced by a polarization q (in respect of the
atomic frame) we rewrite (4.66) using (4.79):
4 2 c 2 2
B(jb mb ; ja ma ; q) = b |r|a jb mb |C1q |ja ma (4.122)

= B(ja ma ; jb mb ; q).
Again B refers to a spectral radiation intensity distribution I() = cu()

per unit
angular frequency.
The ratio of spontaneous (4.113) to induced transition probability (4.122) for a
specific set of two substates is given by
A(ja ma ; jb mb ; q) ba
3
4 h
= = . (4.123)
B(ja ma ; jb mb ; q) 3 2 c3 3 3ba
Note that this ratio differs by a factor 1/3 from (4.38) where isotropic radiation
and summation over all final states is assumed. In contrast, presently a transition
between specific substates is considered. The spontaneous rate has been integrated
over all emission angles to obtain A, while B has been determined for a specific
polarization and direction of the inducing light.
In full analogy to (4.113) we may write the coefficients for stimulated emission
and absorption between specific sublevels as
2
4 2 c jb 1 ja
B(ja ma ; jb mb ; q) = B(jb mb ; ja ma ; q) = S(jb ja )
mb q ma
2
3 2 c3 (2jb + 1) jb 1 ja
= (4.124)
ba3 jb ja m b q ma
2
33 jb 1 ja
= (2jb + 1) A ab ,
4h mb q ma
using the identities A ab = j1

b ja
and ab = 2c/ba .
In a typical absorption or induced emission experiment (even with a polarized
laser beam) one measures the absorption coefficient averaged over all initial and
summed over all final states. Assuming isotropic initial population we obtain for the
averaged E INSTEIN absorption coefficient (independent of q):
1
B ba = B(jb mb ; ja ma ; q)
2ja + 1 m m
b a
1 4 2 c jb 1 ja
2
= S(jb ja ) (4.125)
2ja + 1 mb q ma
mb ma
4 2 c S(jb ja ) 2 c3 2jb + 1 1 3ab 2jb + 1

= = = A ab .
3 (2ja + 1) ba 3 2j + 1
a ja jb 4h 2ja + 1
And equivalently the averaged induced emission coefficient is
4 2 c S(jb ja ) 2 c3 1 3ab
B ab = = = A ab . (4.126)
3 (2jb + 1) ba 3
ja jb 4h
The relations (4.125) and (4.126) just derived express a very important property of
absorption and induced emission without which quantitative spectroscopy would be
much more difficult (see also Appendix H.2):
The averaged B ba and B ab coefficients do not depend on the polarization q.
Finally, as a consequence of this independence of polarization of induced pro-

cesses, we recover the standard E INSTEIN relation for the averaged coefficients of
absorption, induced, and spontaneous emission by comparing (4.125), (4.126), and
(4.114). With g = 2j + 1 we obtain:
2 c3 3ab
ga B ba = gb B ab and B ab = A ab = A ab . (4.127)
ba3 4h
The agreement with (4.38) eventually justifies the hand waving introduction of nu-
merical factors for spontaneous emission in (4.67). However, we emphasize again,
that this is not a derivation of the spontaneous emission rate which we postpone to
Chap. 2 in Vol. 2.
Section summary
The line strength S(jb ja ) introduced in (4.112) characterizes the overall
strength of an atomic transition between energy levels jb and ja .
Various relations among the E INSTEIN A and B coefficients have been de-
rived which are useful for practical computations.
The spontaneous lifetime of substates in a given excited level jb is indepen-
dent of the initial projection quantum number mb .
Absorption as well as stimulated emission in an isotropically populated target
is independent of the polarization of the light used!
4.7 Superposition of States, Quantum Beats and Jumps 217
4.7 Superposition of States, Quantum Beats and Jumps
4.7.1 Coherent Population by Optical Transitions
So far we have discussed only transitions between basis states in the atomic frame.
But of course, initial and final states may also be linear superpositions of these.
Even if the initial state is a basis state, absorption as well as emission of light with
polarization e(ph) may create a linear superposition of states |(b) and |(a) ,
respectively:

(b) mb |rq |ma eq(at) e(ph) |jb mb (4.128)
mb q

(a) ma |rq |mb e(at) e(ph) |ja ma . (4.129)
q
ma q
For illustration we begin by discussing the absorption of elliptically polarized

(ph)
light characterized by eel as defined by (4.15). We assume it to propagate parallel
to the z(at) direction, i.e. k z(ph) = z(at) . The amplitudes of eel are ei cos and
(ph)
(ph) (ph)
ei sin for e+1 and e1 , respectively. With (4.128) absorption of this light by a
pure basis state |ja ma leads to an coherent superposition of excited substates:

(b) ma + 1|r1 |ma ei cos |jb ma + 1
(4.130)
ma 1|r1 |ma ei sin |jb ma 1 .
Often one cannot choose the atomic frame arbitrarily, since the experimental
setup already defines a preferential (at) frame, e.g. if an external magnetic or electric
field is applied, or if one wants to analyze a scattering experiment where a well
defined collision plane exists. The atom has to be described then in an appropriately
chosen coordinate system, which may differ from the (ph) frame which refers e.g.
to the position of a photon detector or to a laser beam for exciting the target. We
choose atomic (at) and photon system (ph) again according to Fig. 4.3 and shall
have to transform either e(at)
q into the photon frame or e
(ph) into the atomic frame as
discussed in Sect. 4.5.2.

There, our discussion was focussed on the characteristic of emission and absorp-
tion for well defined transitions between basis states. Here we want to resume the
discussion about transitions induced by absorption or stimulated emission of elec-
tromagnetic radiation with well defined polarization (characterized by e(ph) ). By
using (4.110) we now expand e(ph) explicitly into an atomic basis. Referring to the
angles defined in Fig. 4.3 we obtain:
k (at) sin k (at) k (at)
e e0 + eik sin2
(ph)
e1 = eik cos2 e (4.131)
2 1 2 2 +1
k (at) sin k (at) k (at)
e+1 = eik sin2 e1 + e0 + eik cos2
(ph)
e . (4.132)
2 2 2 +1
Fig. 4.20 Excitation of a

z ( at ) k, z ( ph)
coherent superposition of
substates m b with b = 1 k
form a = 0 by linearly
polarized light (1 P1 1 S1
E
transition). The electric field y ( ph )
vector E is orthogonal to the
z(at) axis
- ( at ) y ( at )
k k
x ( at ) k x (ph )
Elliptically polarized light propagating into an arbitrary direction is described

(ph) (ph)
by a superposition of e+1 and e1 according to (4.15). The resulting expressions
complicated in the general case. For linearly polarized light (cos =
are somewhat
sin = 1/ 2) which propagates into an arbitrary direction k , k with arbitrary
polarization angle we find
(ph) i sin cos k cos ik (at) (at)

e = e e1 + sin k cos e0
2
i sin cos k cos ik (at)
e e+1 , (4.133)
2
which for = 0 and /2, respectively, simplifies to
cos k (at) cos k (at)

= eik e1 + sin k e0 + eik e+1
(ph) (at)
ex (4.134)
2 2
i (at)
ey = eik e1 + eik e+1 , respectively.
(ph) (at)
(4.135)
2
A full description of the excited atomic state |(b) after absorption of such photons
is obtained by inserting (4.133) into (4.128). We exemplify this for two specific
situations.
Let us assume first, as sketched in Fig. 4.20, excitation from an initial state
|ja ma with linearly polarized light, its electric field vector E being aligned per-
pendicular ( = /2) to the z(at) axis ( light). Thus, the polarization vector is
(ph)
ey as described by (4.135), independent of the propagation angle k . Inserting
(4.135) into (4.129) leads to the excited state:

(b) i eik ma 1|r1 |ma |jb ma 1
2 (4.136)

+ eik ma + 1|r1 |ma |jb ma + 1 .
Specifically, for a transition jb = b = 1 ja = a = 0 (1 P1 1 S0 ). In this

case ma = 0, so that from (4.119) with the3j symbols (B.48) the matrix elements
ma 1|r1 |ma = ma + 1|r1 |ma = rba / 3 are identical. Hence, the angular part
of the excited state wave function simply becomes
i
() = (b) mb |r1 |ma eik Y11 (, ) + eik Y11 (, )
2

rba 1
= sin sin( k ). (4.137)
2
The angular dependence of the charge distribution may be expressed as |()|2 ,

with and referring to the atomic coordinate frame. In Fig. 4.20 this dumbbell
shaped charge cloud is imaged, with its main axis aligned along (at) = k /2
in the x (at) y (at) plane (parallel to the electric field vector E) and independent of the
propagation angle k as seen from (4.135).
Another, trivial case has already been discussed in the context of Fig. 4.14 where
the E vector points into the z(at) direction: in the present language we would de-
(ph) (at)
scribe this by k = /2 with ex = e0 according to (4.134). Thus we expect
only m = 0 transitions. The excited state dumbbell now points into the z(at) direc-
tion.
As a slightly more complex, but instructive example we discuss excitation by
circularly polarized light, which propagates in the x (at) y (at) plane. Inserting (4.132)
with k = /2 into (4.129), one obtains for the upper state, excited from the initial
|ja ma state:

(b) 1 eik mb |r1 |ma |jb ma 1 + eik mb |r1 |ma |jb ma + 1
2
1
+ mb |r0 |ma |jb ma . (4.138)
2
If we specialize again to an b = 1 a = 0 transition

(1 P1 1 S0 ), we
have mb |r1 |ma = mb |r1 |ma = mb |r0 |ma = rba / 3, and obtain for the angu-
lar part of the wave function
rba rba
() eik Y11 (, ) + eik Y11 (, ) + Y10 (, )
2 3 6

rba 1
= i sin sin( k ) + cos . (4.139)
2 2
Apart from the phase factor i the first term of (4.138) and (4.139) [in square brack-
ets] is obviously identical to the wave function generated by linearly polarized light
as described by (4.136) and (4.137) and imaged in Fig. 4.20. The second term cor-
responds to a dumbbell in z(at) direction.
Fig. 4.21 Illustration of a 1 P1 1 S0 transition, induced by circularly polarized light which prop-
agates in y (at) direction
We remember that LHC polarized light corresponds to a pure m = 1 transition

(ph)
in respect of the (ph) with z(ph) k, represented by a toroidal Y11 state. However,
in respect of the atomic frame it is described by a linear superposition of states as
illustrated in Fig. 4.21 for the special case k y (at) .
In the general case of elliptic polarization (not spelled out here mathematically)
the charge distribution will be sort of a squeezed doughnut instead of the dumbbell
shown in Fig. 4.20, lying in the x (ph) y (ph) plane, also aligned along (at) .
4.7.2 Time Dependence of Optically Excited States Quantum

Beats
An interesting aspect of coherent, linear superposition of states is the time depen-

dence which may be generated by optical excitation. While degenerate states within
one atomic level just exhibit the trivial exp(iba t) time dependence of wave func-
tions derived from the S CHRDINGER equation, this may dramatically change if
the atomic (or molecular) sublevels studied are no longer degenerate. Such situa-
tions arise e.g. in external magnetic fields where the energies of the |j m levels
are
Wj m = Wj 0 + mgj B B = Wj 0 + mj . (4.140)
Wj 0 is here the energy of the level without magnetic field, gj the L AND g factor
of the level, B the B OHR magneton, B the magnitude of the external field, and
j = gj L the L ARMOR frequency for J (1.162). If we neglect spontaneous decay
the temporal evolution of the magnetic substates will be given by

j m (t) = |j m ei(j 0 +mj )t , (4.141)
with j 0 + mj = Wj m /. The atomic frame is chosen such that z(at) B, and in all
relevant equations of the preceding section the basis states |j m have to be replaced
by (4.141). A common trivial time dependence exp(ij 0 t) may be pulled out of
the expressions and contributes to an overall non-measurable phase. However, the
frequency shift mgj L = mgj 8.79 1010 s1 T1 B is different for each magnetic
substate and leads to changing amplitudes.
We first have a look at the simplest case, excitation by linearly polarized light
described by (4.136) and (4.137). The E vector of the light is assumed to be per-
pendicular to the magnetic field. Inserting the states (4.141) into (4.136) instead of
|j m and identifying k = k (0) with the phase angle at time t = 0 we obtain the
angular part of the wave functions for a 1 P1 state:
irba
(, , t) ei10 t
6
i( (0)+g t) i(k (0)gj L t)

e k j L Y
11 (, ) + e Y11 (, ) (4.142)

rba 1 i10 t
= e sin sin k (t) (4.143)
2
with k (t) = k (0) + gj L t. (4.144)
The charge distribution |(, , t)|2 contains now a time dependent azimuthal an-
gle k (t). In order to observe this dynamics, the time origin has to be known with
sufficient accuracy in any case better than 1/(2gj L ). For a magnetic field on the
order of 1 T it should be around 0.1 ns. Assume then the atom is excited by a short
light pulse. The dumbbell shown in Fig. 4.20 will then rotate around the z(at) axis
(k (t) decreases with time) with an angular frequency gj L which may be observed
as quantum beat. In contrast, if the atom is excited by a continuous wave (CW) light
beam the L ARMOR precession will tend to generate a disc like charge cloud.
The other example discussed above, excitation by circularly polarized light prop-
agating perpendicular to the magnetic field according to (4.138), may also lead to
quantum beats. Since m = 0 states do not show an energy shift in the magnetic field
the corresponding component Y10 (, ) of the wave function remains unchanged.
In contrast, the Y11 components describe again a dumbbell rotating in the x (at) y (at)
plane as before. Thus, pulsed excitation leads to counter clockwise rotation around
the z(at) axis of the toroidal charge distribution |+1 (ph) shown in Fig. 4.21. For CW
excitation a completely isotropic distribution is expected.
No such dynamics will emerge if the exciting circularly polarized light is prop-
agating parallel to the B field (z(ph) B): the toroidal charge cloud will just rotate
around its symmetry axis without observable change of shape or alignment. The
same holds for linear polarization with e B. In summary, the dynamics in a lin-
ear superposition of non-degenerate states can only be observed if the substates are
populated non-isotropically and if a suitably geometry is chosen.
How can one actually measure such dynamics or wave-packets of the charge
cloud distribution? The so called quantum beat spectroscopy exploits these oscilla-
tions and is used today in several varieties as very efficient method to measure small
energy differences in excited states. Two (or more) neighbouring energy levels in
the excited state are needed as indicated in Fig. 4.22. The initial ground state |a
is excited coherently into the two states |b1 and |b2 by a short light pulse. For
|b 2
21
|b1
|f
|a
Fig. 4.22 Schematic of a typical four level scheme for the observation of quantum beats. Note
that the excited levels must be populated coherently. Quantum beats are generated by interference
of the emission lines from different excited states decaying into the same final state
the moment,14 the term coherent simply expresses a well defined phase relation
among the excited states as a consequence from a sufficiently short duration of the
exciting light pulse 1/21 (with 21 being the excited state level splitting).
Equivalently, this implies a F OURIER limited bandwidth, broad enough to simulta-
neously excite both levels b1 and b2 . At time t = 0 the excited state is then given
by

(0) = c1 |b1 + c2 |b2 , (4.145)
where cj are e.g. the coefficients used in (4.142). However, the following consider-
ations are valid much more generally, for two or more states which are excited co-
herently and have slightly different energies Wj . Quantum beats are not restricted to
j mj sublevels which are split in a magnetic field, as just discussed. Other examples
are highly excited RYDBERG states in atoms, molecular rotational and vibrational
states etc.
The excited states evolve in time according to cj |bj exp(iWj t/) with
Wj / = j . Let us now assume that these states decay by spontaneous emission
into a lower lying, final state |f . The lifetime of the excited states j = 1/Aj
is accounted for by a damping term exp(Aj t/2) for the amplitude, i.e. by
exp(Aj t) = exp(t/j ) for the probability:

(t) = c1 |b1 e(i1 +A1 /2)t + c2 |b2 e(i2 +A2 /2)t . (4.146)
The characteristic frequency for the dynamics clearly is 21 = 2 1 since the

trivial time dependence exp(i1 t) may be pulled out. The intensity of the emitted
radiation according to (4.93) will in this case be
2
If b |r f b ed |2 = f |r f b ed (t) , (4.147)
14 A detailed discussion of coherence will be given in Chap. 2, Vol. 2.

(a) (a) (b)

J=1 M
intensity
+1
0
Re (FT)
-1
intensity
0 2 4 6
t/ s
(b)
Re (FT)
0 4 8 12 0 1 2 3 4 0 10 20 30 40
t/ s / MHz / MHz
Fig. 4.23 Quantum beats in molecules according to C ARTER and H UBER (2000): (a) fluorescence
of the Z EEMAN levels for a R(0) line of the transition 17U in CS2 (magnetic field B 1.5 mT). The
laser polarization was perpendicular to B and enables coherence between the M = 1 sublevels;
(b) fluorescence after coherent excitation of the 000 101 rotational line in the vibrational band
(3,0)
231+
0 I(0,0) of the S1 S0 transition in acetone; quantum beats arise from a mixture between
singlet and triplet states; (a ) and (b ) give the real parts of the F OURIER transformed signals (a)
and (b), respectively
with ed being the polarization registered in the detector system. Obviously, the tran-
sition amplitudes for both processes |f |b1 and |f |b2 interfere. Inserting
(4.146) into (4.147) one obtains
Iab |r f b1 ed |2 |c1 |2 eA1 t + |r f b2 ed |2 |c2 |2 eA2 t (4.148)

(A +A )t/2
+ (r ab1 ed )(r ab2 ed )c1 c2 e 1 2 cos (21 + )t .
Without exploring the details of the polarization dependence we recall, however,

that the coefficients cj also depend on expressions of the type r bj a ee with ee being
the polarization vector of the exciting radiation. The phase between excitation and
emission amplitudes is summarized in . Clearly, a typical interference pattern is
recognized in (4.148): the coherent superposition of radiation from both states leads
to periodic increase and decrease of the observed intensity, the so called quantum
beats, with a period 2/21 and damped as exp[(A1 + A2 )t/2]. If both states have
the same lifetime 1/A and their excitation matrix elements are identical (4.148)
simplifies to
' (
Iab 1 + cos (21 + )t eAt . (4.149)
One may interpret the scheme Fig. 4.22 as a YOUNGs double slit experiment: since
a priori it is undetermined which path the light will take in the overall absorption-
emission process, i.e via state |b1 or |b2 in analogy to the two slits in YOUNGs
experiment, and since on the other hand the final state is well defined, the amplitudes
have to be added coherently and the typical interference pattern is generated.
Quantum beats have already been observed in the 60s of the past century. How-
ever, only after the introduction of tuneable, pulsed lasers it was possible to study
and use them in full beauty. Figure 4.23 presents two nice examples from molecu-
lar physics: (a) a rather clear spectrum is shown for a rotational transition of free
CS2 molecules in a magnetic field where exactly two Z EEMAN levels participate.
The F OURIER transform of the temporal evolution in (a ) shows essentially one dif-
ference frequency. A somewhat more complicated molecule is acetone for which
a quantum beat spectrum is reproduced in (b). It originates from contributions of
several, close lying electronic states. The temporal spectrum thus looks somewhat
intricate, and the F OURIER transform (b ) documents that more than four transitions
participate in the beats. The fact that the signals are not fully modulated from zero
to maximum is attributed to additional, incoherent contributions from other levels.
4.7.3 Quantum Jumps
We want to conclude this chapter with a short consideration of quantum jumps

in some sense the opposite of quantum beats. Studies of the type reported here are
of fundamental relevance for our basic understanding of stationary states and tran-
sitions between them. We remember: stationary states in a quantum system (atom,
molecule, etc.) have a discrete, well defined energy, notwithstanding that we may
create quantum systems in coherent superpositions of states, as just exemplified. As
we shall see in Chap. 10, Vol. 2, even superpositions of ground and excited states
may be generated. However, if we try to determine experimentally in which of the
states the system at a specific instant of time resides, we obtain a unique answer:
either in the excited or in the ground state. Quantum mechanics only allows us to
determine a probability amplitude for one or the other state. The system is never in
between just as the famous S CHRDINGERs cat, which is either dead or alive.
An experiment which documents this in a particularly clear manner has been
carried out by S AUTER et al. (1988) and is illustrated in Fig. 4.24. One single(!)
Ba+ ion is trapped in an ion trap (see e.g. D EHMELT and PAUL 1989) and is excited
continuously with a laser tuned into resonance between ground state |g and a short-
(a) fluorescence / 103 pulses s-1 (b)

e 10 -8 s
|e
m >> e 6
exc
fluor
|m
itati
3
esce
on
ng
0
nce
rki
0 100 200 300 400

pa
|g time / s
Fig. 4.24 Quantum jumps as observed by S AUTER et al. (1988). (a) Part of the term scheme for
Ba+ and the transitions involved. (b) One observes the fluorescence of the state |e as a function
of time. Grey shaded are times where the fluorescence signal vanishes while the ion is parked in
state |m
lived excited state |e . One observes the fluorescence signal emitted from state |e .
At the same time one also induces (with a much smaller probability) a transition
to the metastable state |m from where the ion can decay only with a very low
probability back into the ground state (the average lifetime |m is several seconds).
As documented in Fig. 4.24(b) the almost continuous fluorescence signal emitted
from |e is interrupted in irregular intervals (grey shaded) a clear indication that
the ion is not available for the excitation-fluorescence cycle. Remember, there is
always one but only one ion in the trap! We thus have to conclude that during the
dark periods the ion is parked in the excited, metastable state |m . Obviously, most
of the time one observes a fluorescence signal, i.e. the ion spends most of its time in
the ground state |g from where it can easily be excited for 108 s into the state |e
and thus be monitored. The intermissions by excitation into |m are rather infrequent
(intervals of up to more than 100 s) and have different duration times. A statistical
analysis of these intermissions (not shown here) simply reproduces the exponential
lifetime distribution of the state |m .
This experiment thus documents very explicitly that the ion exists always in a
very well defined state: except for very short times (on average for some 108 s)
in the fluorescing state |e the ion is found either in the ground state |g and may
be detected by excitation-fluorescence cycles, or it is found in the (dark) meta sta-
ble state |m and cannot be detected by fluorescence. There is nothing in between
and a transition from one into the other state occurs (arbitrarily fast) by a quantum
jump. It is only the probability for such a quantum jump which is described by the
exponential decay function for the state |m .
Section summary
Absorption and induced emission with fully polarized light (linear, circular
or elliptic polarization) leads in general to a coherent superposition of basis
states. We have derived closed expressions for these states as a function of
polarization.
If at least two states b1 and b2 with slightly different energies b1 and b2
are exited coherently and with an anisotropic overall population, one may
observe quantum beats if they radiate to the same final states. Coherent exci-
tation in this context implies short pulses with a duration t 1/ b2 b1 .
Single ion storage allows one to observe quantum jumps in an atomic system
which is excited with a certain, low probability: the atom is found either in
the excited or in the ground state. The observed change of population occurs
suddenly and in statistically determined time intervals.
CW: Continuous wave, (as opposed to pulsed) light beam, laser radiation etc.
E1: Electric dipole, transitions induced by the interaction of an electric dipole
with the electric field component of electromagnetic radiation.
E2: Electric quadrupole, transitions induced by the interaction of a quadrupolar

charge distribution with the electromagnetic radiation field.
IP: Ionization potential, of free atoms or molecules (in solid state physics the
equivalent is called workfunction).
LHC: Left hand cicularly, polarized light, also + light.
LIF: Laser induced fluorescence, radiation emitted from a quantum system after
excitation by laser radiation (see Sect. 5.5.1, Vol. 2).
M1: Magnetic dipole, transitions induced by the interaction of a magnetic dipole
with the magnetic field component of electromagnetic radiation.
OMA: Optical multichannel analyzer, spectrometer which allows simultaneous
registration of a whole spectrum.
interaction.
RHC: Right hand cicularly, polarized light, also light.
21348 (2007).
References
B ETH , R. A.: 1936. Mechanical detection and measurement of the angular momentum of light.
Phys. Rev., 50, 115125.
C ARTER , R. T. and J. R. H UBER: 2000. Quantum beat spectroscopy in chemistry. Chem. Soc.
Rev., 29, 305314.
D EHMELT , H. G. and W. PAUL: 1989. The N OBEL prize in physics: for the development of the
S AUTER , T., H. G ILHAUS, I. S IEMERS, R. B LATT, W. N EUHAUSER and P. E T OSCHEK: 1988.
On the photo-dynamics of single ions in a trap. Z. Phys. D, 10, 153163.
Linewidths, Photoionization, and More
5
What we have learned in the preceding chapter still requires

some consolidation, quantification and extension. Linewidth,
dispersion, oscillator strength and cross sections have to be
defined and understood. Multi-photon processes, M1 and E2
transitions will be introduced. Finally, we shall address in some
detail photoionization processes, i.e. various types of photo-
induced transitions from discrete, bound states into the
continuum of unbound states.
Overview
Section 5.1 introduces realistic finite linewidths into the description of optical
transitions. It should be easy to read and is of central importance for the entire
spectroscopy. In Sect. 5.2 closely related to this topic cross sections for
excitation involving lines of finite width are discussed. The concept of optical
oscillator strength is introduced (for which Appendix H.2 provides additional
background). Section 5.3 offers a brief introduction into multi-photon pro-
cesses without which much of modern spectroscopy would not be possible.
Likewise, E2 and in particular M1 processes are important in many areas of
spectroscopy; the reader may, nevertheless, consider the somewhat mathemat-
ical Sect. 5.4 as being mainly for later reference without loosing the present
context. This holds to some extend also for Sect. 5.5 where photoionization
is treated which plays a key role in many areas of modern physics. We shall
make use of it later, e.g. in Sect. 7.6.2 and Chap. 10.
5.1 Line Broadening
5.1.1 Natural Linewidth
Even if we are still missing some key tools for treating spontaneous emission it
is important to develop a pragmatic way to work with it correctly in the context of
spectroscopy. One essential consequence is the finite lifetime = t of stationary
states and the finite width = W of atomic levels, both connected through the
uncertainty relation (1.125). In Sect. 4.7.2 we have already introduced in passing a

228 5 Linewidths, Photoionization, and More
Fig. 5.1 Broadening of an

excited state by spontaneous
decay with L ORENTZ
distribution |b
Wba
|a
somewhat heuristical method to account quantum mechanically for finite lifetimes

of excited states. We now want to have a closer look at this approach without
claiming formal mathematical stringency. One describes the temporal evolution of
atomic states by time dependent probability amplitudes cj (t) as previously done in
the perturbation ansatz (4.41). Now we account, however, for the finite lifetime =
1/A of excited states |j which has been derived from 1st order perturbation theory
more precisely: from the E INSTEIN relation (4.38) between A and B coefficients.
An atom which is prepared in this excited state at time t = 0 will decay with time
t according to the probability distribution
2
wj (t) = cj (t) = eAj t = et/j .
In the spirit of perturbation theory the former 0th order approximation cj (t) = 1
now becomes cj (t) = exp(t/2j ). In the ansatz (4.41) we thus replace
j (r)eij t by
j (r)et/2j eij t = j (r)ei(j i/2j )t . (5.1)
One may read this equation as if excited states were now characterized by a complex
eigenfrequency or eigenenergy, substituting
i Aj
j j = j i or (5.2a)
2 2
j
Wj Wj i with j = Aj = . (5.2b)
2 j
Schematically this is illustrated in Fig. 5.1.

With (5.2a) the system of ODEs (4.45) for the time dependent probability ampli-
tudes becomes
dcb (t) i
= cj (t)Ubj (t)ei(bj tiAj t/2) , (5.3)
dt
j
5.1 Line Broadening 229
and instead of (4.58) the final result in 1st order perturbation theory is
eE0 ei((ba )tiAt/2) 1

cb (t) =
Dba (5.4)
2 i(ba iA/2)
(we drop here the index of Aj ) with E0 given by (4.56) and

Dba in dipole approxi-
mation by (4.57). For times t 1/A (steady state) the probability amplitude for |b
becomes independent of time, we call it cr :
i
cr = cb (t ) . (5.5)
ba iA/2
cr is a typical resonance amplitude, well known from classical physics where it de-
scribes a forced oscillator. It depends on the detuning ba and on the damping
constant A.
LORENTZ Prole
Somewhat more precisely, the decay constant A used here has to be identified with
A ab given in (4.114)(4.117) as long as only transitions between two isolated lev-
els are considered. From (5.5) follows the probability to find an atom in the excited
state if CW light of angular frequency is absorbed:
A2 /4
cb ()2 gL () = 2 . (5.6)
A ( ba )2 + A2 /4
This is the well known L ORENTZ profile (L ORENTZ distribution), describing an ab-
sorption line with a natural linewidth. The distribution is given here per unit angular
frequency, i.e. it has the dimension T. Its FWHM is nat = A A ab . In respect of
energy the L ORENTZ profile reads
2 ( /2)2 /2
gL (W ) = = . (5.7)
(W Wba ) + ( /2)
2 2 (W Wba )2 + ( /2)2
nat is related to the frequency bandwidth nat , the energy width of the state,
its natural lifetime nat and its spontaneous decay rate A:
1
nat = 2 nat = A = = . (5.8)
nat
Both profiles, gL () and gL (W ), are normalized to unity:

gL ()d = gL (W )dW = 1. (5.9)

Their maxima at = ba are 2/ nat and 2/ , respectively.

Frequency integrated excitation and absorption rates have been treated in the
preceding chapter. With (4.62) and I() (intensity distribution per unit angular fre-
quency, dimension Enrg T1 L2 T = MT2 ) the absorption rate in a frequency interval

to + d may be written
I()
dRba = BgL () d. (5.10)
c
B is the E INSTEIN coefficient according to (4.124) or (4.126), depending on exper-
imental conditions.1 The rate for spontaneous emission of a photon with angular
frequency is given by
(spont)
dRab = AgL ()d,
with the spontaneous decay rate A A ab , equal to the inverse lifetime of any of the
substates of level b in respect of decay into all substates of level a.
We may now introduce a frequency dependent absorption coefficient by multi-
plying B ba according to (4.126) with the normalized line shape (5.6):
gb 3ab A 2ab /4
B ba () = B ba gL () = . (5.12)
ga 2h ( ba )2 + A 2ab /4
B ba is the averaged absorption probability from the lower levels a to the upper levels
b for a given polarization.2
The attentive reader may find the above derivation of the frequency dependent
absorption coefficient somewhat heuristic. We shall come back to this in Sect. 10.5,
Vol. 2 where a more systematic deduction will be presented which turns out to
lead to identical results as the one above.
Population Density and Saturation Broadening

It is instructive at this point to learn something about the population density in the
excited state from (5.12). We recall the rate equation (4.36) and apply it for sta-
tionary conditions, now at nearly resonant excitation by quasi-monochromatic CW
radiation. We consider an isolated ensemble of atoms (no thermalization) in which
only the ground state (density Na ) and the excited state b (density Nb ) are popu-
lated. To keep things simple we assume a pure two level system,3 which may easily
be realized, e.g. if jb = ja + 1 by mb = jb , ma = ja and q = 1. In this case we
insert (5.12) instead of Bba into (4.36), we set gb /ga = 3, Bab () = Bba () and the
1 Ifinstead of individual transitions between substates, averaged probabilities are studied, we have
to replace in (4.68) the squared transition matrix element
1 1
|
Dab |2 = |r ab e|2 by |
Dab |2 = |r ab e|2 , (5.11)
ga m m ga m m
a b a b
2 If one is interested in the absorption profile for one specific transition between substates
jb mb ja ma , one would have to replace in (5.12) the prefactor (gb /ga ) 3 and the linewidth
A A(jb mb ja ma ; q) according to (4.123).
3 If several excited or lower state sublevels are involved one has, in addition, to account for optical
pumping (see Appendix D in Vol. 2).

decay constant A = Aab for this particular transition (see footnote 2). We replace

the spectral radiation density u() I /c with I0 being the intensity of the (very
narrow band) laser radiation. Then we obtain from (4.36)
B ba ()I /c
Nb /Na = ,
Aab + B ab ()I /c
from which we extract the relative population of the excited state
Nb (R /2)2
= (5.13)
Nb + Na (Aab /2)2 + R2 /2 + ( ba )2
with the abbreviation R2 = (33 Aab )I /(2hc).

At very low intensity (R Aab ) the L ORENTZ distribution is recovered with
a FWHM= Aab . The maximum excitation probability is proportional to the laser
intensity I . At high intensity (R > Aab ) the
line profile is still a L ORENTZ distri-

bution, however, the FWHM is now s = Aab + 2R and increases I we
2 2
speak of power broadening or saturation broadening: even for extremely high inten-
sity the maximum population (stationary limit) in the excited state at resonance
= ba is always Nb /(Nb + Na ) < 1/2, i.e. the transition becomes saturated. By
excitation with CW radiation one can never excite more than 50 % of the atoms!
We emphasize again, that the above derivation is to some extent guess
work: perturbation theory is strictly valid only for population densities Nb /(Nb +
Na ) 1/2. However, as we shall see in Sect. 10.5.2, Vol. 2, the non-perturbative
treatment confirms (5.13).
Homogeneous Line Broadening

Closely related to the above, we introduce here the concept of homogeneous line
broadening as opposed to inhomogeneous broadening which will be discussed
in Sect. 5.1.4. If the absorption (or emission) profile as a function of frequency is
identical for each individual quantum system (atom), the broadening is called ho-
mogeneous. In contrast, if each individual quantum absorber/emitter has a different
frequency profile, the broadening is called inhomogeneous.
The L ORENTZ profile (natural line profile) describes the prototype of homo-
geneous line broadening: absorption of a photon irrespective of its frequency
always excites the upper state as a whole. All atoms of a kind (say H atoms) are
identical in this respect and cannot be distinguished by their absorption frequency.
Conversely, if excited to a specific state they all emit the same spectrum with a prob-
ability described by gL (). The concept of homogeneous vs. inhomogeneous line
broadening plays an important role in many spectroscopic problems and methods
and is of particular relevance for understanding amplification in laser systems (see
Sect. 1.1 in Vol. 2).
Some Numbers
To develop some feeling for the orders of magnitude of natural linewidths we com-
municate two typical values:
For the LYMAN alpha line (1s 2p) in hydrogen like atoms we use (4.117)
with (C.56) and Table C.2. With these values4 we obtain A(1s2p) = 6.2658
108 Z 4 s1 . Specifically, for the 2p state of the H atom (H line at =
121.57 nm) the lifetime becomes nat = 1.596 ns. From this follows a linewidth
nat 99 MHz.
For other atoms one has to compute the radial wave functions and from these
the radial matrix elements numerically, as described by way of example for
alkali atoms in Chap. 3. One of the strongest atomic transitions is the Na D
line (3s 3p) at = 589 nm, with a lifetime 16.2 ns, nat = 1/nat =
6.15 107 s1 or nat 9.8 MHz.
These linewidths have to be compared to the transition frequencies ba = c/,

being 24.66 and 5.089 1014 Hz for H and Na, respectively. Natural linewidths are
thus extremely small, with nat /ba = / 4 108 and 2 108 , respec-
tively. For the Na D line the bandwidth in wavenumbers is nat /c = 0.00033 cm1
and in wavelengths units = c nat / 2 1.06794 103 nm. Hence, ex-
tremely good spectral resolution is required to record such line profiles experimen-
tally.
We note here that the natural linewidth is a lower boundary for the width of any
spectral line. It is directly related to the lifetime of the excited states by the H EISEN -
BERG uncertainty relation. A variety of influences and experimental conditions may
lead to further broadening of spectral lines as we shall see later.
5.1.2 Dispersion
For later reference we note that the complex resonance amplitude (5.5) with reso-
nance energy Wba and linewidth = / (FWHM) may be written:
cr (W ) = |cr | exp(i) with

2 /4
magnitude |cr | = (5.14a)
(Wba W )2 + 2 /4

/2
and phase = arctan + ( for W > Wab ). (5.14b)
Wba W
Here we have normalized |cr (W )| at resonance to |cr (Wba )| = 1.
Alternatively, we may express cb (W ) by its real and imaginary parts:
(Wba W ) /2
Re(cr ) = |cr | cos = (5.15a)
(Wba W )2 + 2 /4
2 /4
Im(cr ) = |cr | sin = . (5.15b)
(Wba W )2 + 2 /4
4 In the particular case of s p transitions A is even independent of the electron spin.

(a) 1 (b) 1.0

absorption
phase

0.5
dispersion
0.5
absorption
-8 -4 0 8
- 0.5
-8 -4 0 4 8
(W-Wba) /
Fig. 5.2 Alternative representations of the resonance amplitude (5.14a)(5.15b) as a function of

photon energy W = . (a) Magnitude and phase, (b) imaginary part (absorption) and real part
(dispersion)
We note that the imaginary part reproduces (apart from the normalization) the
L ORENTZ distribution (5.7) and thus corresponds to the absorption probability,
while the real part describes a dispersion type line shape. Both profiles are illus-
trated in Fig. 5.2. We shall come to these again in Sect. 7.6.2 where so called FANO
resonances are treated, as well as in Sect. 8.4.3 in the context of dispersion of the
refractive index as a function of wavelength.
5.1.3 Collisional Line Broadening
At higher pressures atoms and molecules in the gas phase collide frequently with
each other. One may rationalize the influence of such collisions on the linewidth by
assuming that the emission (or absorption) process is not completely disrupted, but
looses all its phase memory. One assumes that the interactions occur statistically
distributed (exponential probability distribution), with an average time interval tcol .
As will be discussed in detail in Sect. 2.1.3, Vol. 2 these considerations lead again
to a L ORENTZ distribution of the frequency spectrum with a bandwidth inversely
proportional to the collision time.
The average time tcol between two collisions is derived with (1.52) from the
particle density N of the perturbing gas, the collision cross section and the rel-
ative velocity v between radiating atoms and perturbing gas particles (being not
necessarily the same). Thus, collisional line broadening (or pressure broadening) is
characterized by a Lorentzian line shape with a FWHM

col = 1/tcol = N v = v p/kB T p 8/( Mk BT ) (5.16)

col p 8/( Mk B T )/2, (5.17)
where indicates averaging over the relative velocity distribution in the gas stud-
ied. The particle density is related to the pressure p by (1.50) at temperature T . The
Fig. 5.3 On the origin of the D OPPLER profile for an absorption line: A red or blue shift occurs,
depending on the direction of the atomic velocity (red and black marked atoms, respectively). No
shift occurs for atoms at rest or moving k (grey)

average (relative) velocity v 8kB T /( M) is derived from the M AXWELL -
B OLTZMANN distribution (1.57). The reduced mass M of the colliding pairs has to
be used here, since it is their relative velocity that counts. Typical gas kinetic cross
sections are 1015 cm2 (mostly elastic).
For our standard example, Na atoms, say in a cell with 100 mbar Ar buffer gas
(M = 14.6 u) at T = 554 K, the Na vapour pressure is 1 Pa = 0.01 mbar, sufficiently
high for nice spectroscopy. At these conditions the frequency bandwidth becomes
about col = 1.84 108 Hz, about an order of magnitude larger than the natural
linewidth. It increases linearly with pressure and may become very important for
dense gases. The collisional broadening too is a homogeneous line broadening since
it is of statistical nature and concerns all atoms in a gas in the same way.
If this additional broadening is of similar magnitude as the natural linewidth
one has to convolute both profiles. Specifically, the convolution of two Lorentzian
profiles (see Appendix G.5) leads again to a Lorentzian line profile with additive
linewidths (FWHM):
1 1
1/2 = col + A = + . (5.18)
tcol nat
5.1.4 DOPPLER Broadening
While collision broadening may often be avoided in spectroscopic studies of isolated

atoms and molecules, simply by reducing the target density, a much more cumber-
some and important broadening mechanism is due to the thermal motion of atoms
and molecules in the gas phase. Figure 5.3 illustrates the origin of this D OPPLER
broadening of spectral lines. The D OPPLER shift for a particle with a velocity com-
ponent vx in k direction is, in the non-relativistic limit, according to (1.32)
vx 2
ba = ba = vx , (5.19)
c ba
where ba = 2ba is the angular transition frequency in the rest frame of the atom,
ba the corresponding wavelength, and ba = 2 ba the shift of absorbed (or
emitted) angular frequency in respect of ba .
1/2
-4 -2 0 2 4
- b a / 1/2
Fig. 5.4 Comparison of a L ORENTZ profile (black line), typical for the natural line profile and
collision broadening, and a Gaussian distribution (red line), arising from thermal line broaden-
ing. Compared are here profiles of the same FWHM 1/2 , renormalized to equal maxima for
comparison
According to (1.56), the B OLTZMANN distribution taken for one velocity com-
ponent vx (in the direction of light propagation k) is
1 kB T
w(vx )dvx = exp vx2 / 2v02 dvx with v02 = v 2 = (5.20)
2v02 m
being the variance of the distribution, v0 = kB T /m its standard deviation, m the

atomic mass, T the temperature, and kB the B OLTZMANN constant. Thus, with
(5.19) each atom (molecule) absorbs at a different frequency according to its ve-
locity component vx .
Inserting vx from (5.19) into (5.20) leads to a Gaussian line profile

1 1 ba 2 kB T
gD ()d = exp d, with D = ba (5.21)
2D 2 D mc2
being the standard deviation of the Gaussian. The FWHM is

kB T 2 kB T
D = 8 ln 2D = ba 8 ln 2 = 8 ln 2 . (5.22)
mc2 ba m
Since d/D = d/D , the D OPPLER line profile in frequency space is ob-
tained from (5.21) by just replacing
, with =
D D /2 = ba kB T /mc
2
and a FWHM D = ba 8 ln2 kB T /mc 2 2

2.4ba kB T /mc . As usual, the
D OPPLER profile is normalized: gD ()d = gD ()d = 1.
Figure 5.4 illustrates that the wings of a G AUSS distribution are strongly sup-
pressed in comparison to a L ORENTZ distribution of the same FWHM.
We note here, that D OPPLER shift satisfies of course momentum and energy con-
servation. We prove this in the non-relativistic limit for emission of a photon with
energy into direction n. Energy conservation requires
1 1
+ a + mv 2a = b + mv 2b , (5.23)
2 2
where v b and v a are the atomic velocities, b and a the electronic energies in
the upper and lower state, respectively. Momentum conservation implies

mv b = mv a + n. (5.24)
c
Inserting v a from (5.24) into (5.23), abbreviating as usual ba = b a , and
neglecting terms /mc2 vx /c, one easily verifies

ba vx ba vx
= ba 1 + or ba = = ba .
1 vx /c c 2 c
Here vx = v b n is the projection of the initial atomic velocity v b onto the direction
of propagation of the photon. Thus we have recovered the non-relativistic D OPPLER
shift according to (5.19), q.e.d.
In practice, D OPPLER broadening in the gas phase is usually much larger than
the natural linewidth or collision broadening at not too high pressures. As an
example, we refer once again to the Na D lines. At 554 K with m = 23 u the
D OPPLER width becomes D 1.8 109 Hz at = 5.09 1014 Hz as com-
pared to nat = 9.79 106 Hz corresponding to a natural lifetime nat = 16.2 ns.
The D OPPLER width in this case is thus two orders of magnitude larger than the
natural linewidth and one order of magnitude larger than the collision broadened
line at 100 mbar. Spectroscopy of emission or absorption lines in the gas phase
the most important source of our knowledge about the structure of atoms and
molecules has to fight massively against D OPPLER broadening. In later chapters
we get to know a variety of interesting methods which deal artfully with this prob-
lem.
In view of Fig. 5.3 one easily recognizes that D OPPLER broadening in con-
trast to natural line broadening and collisional broadening is inhomogeneous:
each atom emits its specific wavelength according to its instantaneous velocity. In
an absorption process each particular frequency excites only a specific group of
particles in a velocity interval from vx to vx + vx related through (5.19) to the
natural linewidth by vx = ba nat .
5.1.5 VOIGT Prole
In the most general case (i.e. at high temperatures and high pressures) one has to
account for both effects, D OPPLER broadening and collision broadening. One has
to convolute then a Lorentzian with a Gaussian profile. This is not completely triv-
ial, but can be done as briefly sketched in Appendix G.6. The result, a so called
VOIGT profile, is illustrated in Fig. 5.5 for the specific example of a L ORENTZ and
D OPPLER profile with equal width. A rather accurate approximate formula for the
linewidth of a VOIGT profile is (G.25).
Such delicate studies and evaluations of spectral lines were done rather exten-
sively in the second half of the past century, partially in order to better understand
Fig. 5.5 Comparison of 1.0

G AUSS (black), L ORENTZ
(grey), and VOIGT (red) line
profiles the latter being a 0.5
D
convolution of the former
two. All three profiles are L
normalized to V

g()d = 1. The arrows
indicate the FWHM of the - 2.0 0 2.0
distributions - ba / 1/2
collision processes. Modern spectroscopy and collision physics have advanced over
the past decades substantially. Thus it may be said that collisional and D OPPLER
broadening have more or less been overcome by a toolbox of sophisticated spectro-
scopic methods.
Nevertheless, a detailed evaluation of line profiles still plays an important role
for remote spectroscopy of dense gases, high pressure gas discharges, and plas-
mas in general in laboratory studies as well as in astrophysics. In addition to the
effects mentioned here, several other mechanisms such as S TARK broadening in
plasmas contribute to the line shapes. In otherwise inaccessible media systematic
profile measurements of spectral lines often provide the only viable experimental
access to important physical properties such as particle densities and tempera-
tures.
Section summary
The natural line profile arises from the finite lifetime ab of the excited state. It
is described by a L ORENTZ distribution (5.12) with a FWHM nat = A ab =
1/ab .
Collisional line broadening is also represented by a L ORENTZ profile. It is
for studies in gases at high pressure p, its FWHM being col
relevant
p 8/( Mk B T ) at a temperature T .
D OPPLER broadening arises from transition frequencies shifted due to differ-
ent velocities of freely moving atoms or molecules in a gas. The B OLTZMANN
velocity distribution
leadsto a Gaussian line profile (5.21) with a FWHM of
D (2/ba ) 8 ln 2 kB T /m.
Typically nat col and nat D .
The VOIGT profile is a convolution of L ORENTZ and Gaussian profile and
describes a situation where both, D OPPLER broadening and collisional broad-
ening, are important, i.e. at high pressure and high temperature.
We distinguish homogeneous and inhomogeneous line broadening, if each
absorber or emitter has the same line profile or a different resonant frequency,
respectively. The natural line profile and collision broadening are typical for
the former, D OPPLER broadening for the latter.
5.2 Oscillator Strength and Cross Section
5.2.1 Transition Rates Generalized
In Sect. 5.1.1 we have started to extend the concept of transition rates between dis-
crete stationary states to levels broadened by spontaneous emission. We may con-
sider this as a first step to make the transition into the continuum. The considerations
presented there need not be restricted to the natural linewidth or to homogeneous
line broadening. They can also be applied to transitions broadened by collisions or
by the D OPPLER effect as just discussed. We simply have to replace gL () in (5.10)
by the corresponding profiles. In this context one usually normalizes to energy scale.
With W = and I() = dI /d = dI /dW the transition rate (4.62) is rewritten
as
dRba
= 4 2 I(W )|
Dba |2 g(W ), (5.25)
dW
with |Dba |2 according to (4.57) in dipole approximation (see also footnote 1). As
introduced in Sect. 1.3.4, g(W ) is now an appropriate density of states which de-
scribes the probability to encounter and excite the corresponding final states. The
area under this distribution function (whatever its overall width) must of course be
properly normalized according to (5.9). In this reading dRba /dW is a transition rate
(or a rate of absorption from the photon point of view) per unit energy interval (di-
mension T1 Enrg1 ), and the density of states g(W ) may in the most general case
describe an extended continuum or quasi-continuum of very closely spaced final
states: it specifies how many states exist per unit energy.
Replacing in (5.25) the specific interaction for electromagnetic radiation with
matter by the general transition operator U as introduced in Sect. 4.3.1, this expres-
sion turns out to be just a special application of the famous
dRba 2 2
F ERMIs golden rule: = |Uba | g(W ), (5.26)
dW
which in the literature is usually written without explicit reference to its differential
nature per unit energy interval.
5.2.2 Oscillator Strength
In Sect. 4.6.1 we have introduced the line strength S(jb ja ), which has the dimen-
sion L2 . For ease of comparing dipole transitions between different transitions and
atoms, one also defines a dimensionless quantity, called oscillator strength
2
(opt) 2Wba S(jb ja ) 2Wba
Dba
2Wba zba 2
fba = = = . (5.27)
Eh 3ga a02 E h a0 E h a0
5.2 Oscillator Strength and Cross Section 239
Briefly, the line strengths S(jb ja ) is the sum over the squares of all transition matrix
(opt)
elements between sublevels of a and b, while fba characterizes absorption (or
induced emission) from one substate |ja ma of the initial level a to the final level b
averaged over all initial substates and summed over all final states of that level.5
The degeneracy factor ga = 2ja + 1 compensates the summation over all initial
states in the definition of S(jb ja ) according to (4.112), the factor 1/3 compensates
(opt)
summation over all polarizations. We emphasize that fba is independent of the
polarization used!
With the transition energy Wba = ba being related as usual to the term energies
by Wba = (Wb Wa ) and with the definition (5.27) we have
(opt) (opt)
Wba = Wab and ga fba = gb fab .
(opt)
Thus fba > 0 for absorption and <0 for induced emission. As shown in Ap-
pendix H.2 the important
(opt)
T HOMAS -R EICHE -K UHN sum rule fba = N , (5.28)
b
strictly holds in dipole approximation, with N being the number of active electrons.
The summation (and possibly integration) has to be carried out over a complete
basis set of states describing the system, and potentially includes positive as well as
(opt)
negative values of fba .
The term oscillator strength is best understood in the context of the classical
(opt)
picture of one oscillating electron: an oscillator strength fba = 1 for a particular
transition implies that the whole ability of an active electron to absorb electromag-
netic radiation is concentrated in that one particular transition. A nearly perfect ex-
ample provides Na. As discussed in detail in Sect. 3.2 its one valence electron is per-
fectly described by a quasi-one-electron system. The oscillator strength for its yel-
(opt)
low NaD line (which we have already mentioned several times) is f3p3s 0.98,
i.e. it contains ca. 98 % of all available oscillator strength. In contrast, for the H

atom f2p1s = 0.416, f3p1s = 0.073, f4p1s = 0.029,
(opt) (opt) (opt) (opt)
n=5 fnp1s = 0.041. In
addition one has to consider ionization for which integration over all continuum
(opt)
states gives dfba = 0.435.
When applying the T HOMAS -R EICHE -K UHN sum rule to larger systems one has,
however, to be aware that the sum (5.28) is to be taken literally and may include tran-
sitions to fully occupied levels which are forbidden transitions as a consequence
of the PAULI principle.
(opt)
A very useful numerical relation between oscillator strength fba (for excitation
b a) and spontaneous transition probability Aab (for the decay process a b) is
derived from (5.27) and (4.114):
5 In the second and third equality of (5.27) we have omitted these summations for simplicity. Precise
details and alternative expressions are summarized in Appendix H.2. The last expression refers to
linear polarization, parallel to the z-axis.
(opt)
ga fba = Cgb Aab 2 (5.29)
0 me c
with C = = 1.4992 1014 nm2 s.
2e2
5.2.3 Absorption Cross Section
If one describes the absorption process so to say from the view point of the photon,
the question arises, which cross section does the photon see during its interaction
with an atom. With the photon energy W = we may rewrite (5.25) for the number
of photons absorbed per unit of time and energy interval as (see also footnote 1)
dRba I(W )
= 4 2 W |
Dba |2 g(W ). (5.30)
dW
Now, I(W )/() = is the number of photons per unit time per unit area and per
spectral energy range, i.e. photon flux per unit energy. Thus, generalizing (4.24) we
obtain the absorption cross section for a transition from |a to |b as a function of
the energy of the absorbed photons:
2
2W
Dba
g(W )Eh a 2 .
ba = 4 W |
2
Dba |2 g(W ) = 2 2 (5.31)
E h a0 0
On the right hand side we have rewritten this expression using atomic units a0 and
Eh , so that ba is given essentially as a product of dimensionless factors: Since
|
Dba |2 /a02 is dimensionless, Eh is an energy, and the density of states g(W ) has the
dimension Enrg1 , ba has indeed the dimension L2 of a cross section (here given
in units a02 ). If we refer here to the averaged cross section for transitions between
levels a and b (see footnote 1) the term in square brackets is the oscillator strength
as defined in (5.27) and (H.37). Thus we may write in compact form
(opt) (opt)
ba = 2 2 fba g(W )Eh a02 = 2 2 re cfba g(W ). (5.32)
Inserting numerical values for the atomic constants we obtain6
ba = 4.034 1018 cm2 Eh fba g(W ) = 1.098 1016 cm2 eV fba g(W ).
(opt) (opt)
The classical electron radius re = e2 /(40 me c2 ) in (5.32) recognizes the fact

that this equation can also be obtained from the classical forced oscillator model
except for the oscillator strength which can only be determined quantum mechani-
(opt)
cally (for a one electron system 0 fba 1).
6 Note that different authors use different units. Thus, one finds different prefactors in different
publications. The value used here corresponds to FANO and C OOPER (1968), while e.g. C OOPER
(1988) uses 8.067 1018 cm2 and measures energies in RYDBERG = Eh /2.
For excitation of a resonance level, which is determined only by the natural life-
time, the cross section (5.32) may be rewritten in an even more transparent manner.
(opt)
With (5.29) we express fba by Aab and identify g(W ) as the L ORENTZ distribution
(5.7) with a FWHM = = nat , normalized according to (5.9). By inserting the
atomic constants the photo-absorption cross section becomes
gb 2 ( /2)2
ba (W ) = . (5.33)
ga 2 (W Wba )2 + ( /2)2
Its maximum value A reached for W = Wba (corresponding to = ba ) is univer-

sally given by
gb 2ba
A = ba (Wba ) = , (5.34)
ga 2
and depends only on the wavelength ba at resonance and on the degeneracies. To
give a numerical value we refer again to our standard example, the Na D transition
at ba = 589.6 nm from which we obtain A = 1.6 109 cm2 (gb /ga = 3). This is
indeed an extremely large absorption cross section.
The relation between absorbed light intensity I , particle density Na,b and E IN -
STEIN coefficient Bba has already been discussed in Sect. 4.6.3. There I expressed
the total intensity absorbed over the distance x by the whole spectral line consid-
ered arbitrarily narrow. With ba (W ) we may write this now frequency or energy
dependent. The intensity I(W ) absorbed in an energy interval W is now simply
I(W ) = Na ba (W )I(W ) x, (5.35)

and the corresponding energy dependent absorption coefficient (in the chemical
literature called extinction coefficient) becomes () = Na ba ().
However, when deriving (5.34) from (5.33) we have tacitly assumed essentially
monochromatic light with light nat . This may be achieved but not in stan-
dard absorption experiments. Rather, one usually excites with light for which the
opposite is true, light nat . In that case the experiment averages over the
whole spectral line and one measures an averaged cross section

I (W Wdet )ba (W )dW
ba (Wc ) = . (5.36)

I (W )dW
Here we have accounted for some detuning Wdet of the light source in respect of
the center of the spectral line. This convolution (see Appendix G) thus allows one
to describe a realistic spectroscopic experiment where the laser is tuned over the
spectral lines.
We assume now for simplicity that the spectral distribution of the light source
(e.g. a tuneable dye laser) is also characterized by a Lorentzian profile, its FWHM
being light . Using (G.21) the convolution integral (5.36) can be evaluated, leading
again to a (properly normalized) L ORENTZ profile gL (; FWHM) with a FWHM=

nat + light . One finds that
2 gb
ba () = nat gL (; nat + light ) (5.37)
2 ga
with ba being the detuning of the light source from the maximum ba of
the spectral line. On resonance ( = ba ) the averaged absorption cross section
becomes
___ 2 gb nat
A = . (5.38)
2 ga nat + light
Thus, in practice the huge maximum absorption cross section (5.34) is reduced to an
effectively much smaller value if the bandwidth of the light source is large compared
to the natural linewidth. The total absorbed intensity (W cm1 ) is then given by
___
I = Na A I x.
For other line profiles similar considerations may be applied to derive averaged
cross sections. For instance, to describe a mainly collision broadened line, one
would have to replace nat = / by col according to (5.16) and reduce the
cross section (5.33) by nat / col .
Note, however, that the linear averaging described above is only applicable in
the linear absorption regime, i.e. as long as I(W )/I(W ) 1, and the relative
absorption may be computed directly from (5.35). Otherwise one has to apply its
integrated form, the L AMBERT-B EER law (1.45) at each incident frequency, and
finally average over the line shape of the radiation.
5.2.4 Different Notations Radiative-Transfer in Gases
In general, this is not a trivial task, in particular if molecular spectra with many lines
are considered. They may even arise from several different absorbing gas molecules
as it is the case in radiative-transfer of the incoming solar radiation through our
earths atmosphere. In the spectroscopic literature and in data banks (see e.g. ROTH -
MAN et al. 2009) one often rewrites (5.33) as a function of the wavenumber = 1/
gb 2 /
ba ( ) = = Sba L( , ba , ), (5.39)
ga 2 ( ba )2 + 2
where the L ORENTZ line shape L( , ba , ) is given as a function of wavenumbers,
and is the half width of the spectral line profile at half maximum (HWHM) in
wavenumbers [ ] = cm1 . Note that L( , ba , ) is measured in units [L] = cm
and normalized so that L( , ba , )d = 1.
The parameter Sba is usually also called line strength (not to be confused
with the line strength S(jb ja ) which we have used elsewhere, with the dimen-
sion L2 ). It is measured in units [S] = cm1 /(molecule cm2 ) to be identified
as wavenumber per column density. In radiative-transfer calculations one has also
to account for changes in the density Na (x) of absorbing molecules with position
x in space and defines the column density as CD = Na (x) x, measured in units
[CD ] = molecule cm2 . Thus, (5.35) is rewritten for a given spectral line, centred
at ba , as
dI( ) = I( )Sba L( , ba , )dCD . (5.40)
If only one single absorbing line is considered, the relative intensity transmitted
over a spectral range is determined by
ba + /2
I 1 /
= exp CD Sba d ,
I0 ba /2 (ba )2 + 2
where we have assumed that the incident spectral intensity is constant over and
centred at ba .
For the general multiline problem (also accounting for D OPPLER broadening)
one has to replace Sj L( , j , ) j Sj V ( ; j ) in (5.40), with VOIGT line pro-
files (Sect. 5.1.5) centred at each relevant spectral line j . By integrating (5.40) over
CD one derives the relative transmission I( )/I0 ( ) for the spectral intensity at each
of interest. To model a genuine molecular spectrum one also has to account for the
thermal population of the initial levels, and include induced emission in addition to
absorption. Finally, this spectrum I( ) thus obtained has to be convoluted with the
resolution of the spectrometer.
For atmospheric modelling, reemission of radiation and scattering from the at-
mosphere has also to be included, as well as absorption and scattering from aerosols
and small droplets. Thus, the problem gets rather involved. References to a number
of codes available for atmospheric radiative-transfer are reported by W IKIPEDIA
CONTRIBUTORS (2013). They all rely on information from molecular spectroscopy
as documented in databanks such as HITRAN, and allow to model solar spectra on
the surface of the earth as presented in Sect. 1.4.7.
Section summary
We have related the rate for optically induced E1 transitions to the famous
F ERMI golden rule (5.26).
(opt)
The optical oscillator strength fba is a useful, dimensionless quantity which
characterizes the strength of an E1 transition. For a (quasi)single electron sys-
tem, in the classical limit of a harmonic electron oscillator it would be equal
to 1, in quantum mechanical reality it obeys (5.28), the T HOMAS -R EICHE -
(opt)
K UHN sum rule, b fba = 1 and is always <1.
The photo-absorption cross section (5.32) for an infinitely narrow bandwidth
laser beam is very large, specifically A = 2 /(2)(gb /ga ) at the maximum
of a L ORENTZ distribution if only natural line broadening (FWHM nat =
1/ab ) is relevant. The measured cross section depends, however, strongly on
the line profile of the absorbed light (FWHM light ). If the latter is large
___
compared to the former, the photo-absorption cross section becomes A
A nat / light .
5.3 Multi-photon Processes
Up to now we have determined transition probabilities in 1st order perturbation

theory. Thus, we have included only processes where a single photon is absorbed or
emitted. This is in order as long as the light intensities used are low enough so that
the excitation probabilities are small and changes of the wave function of the target
are negligible. However, with todays powerful laser sources, this is by no means a
trivial assumption.
On the contrary: multi-photon processes are at the heart of todays laser spec-
troscopy on the one side they may be something to avoid carefully if one aims at
strictly linear spectroscopic data. On the other hand, they may be used artfully for
accessing a broad range of objects, states and transition types which would other-
wise be unaccessible. Not only in physics and physical chemistry are they of high
spectroscopic relevance, they also have become indispensable tools in biology and
medicine. A quick Internet search for two-photon will convince you that a host of
applications relates to confocal microscopy, fluorescence microscopy and biological
imaging.
Hence, we are well advised to approach this growing field of trans-disciplinary
research by trying to understand at least the basic physics involved applying tools
we have already acquired. In the following we shall define the terminology, derive
some quantitative expressions for the cross section of two-photon excitation, and
mention some experimental aspects (more examples will be met on many occa-
sions in later chapters). We shall then discuss two-photon decay processes for sim-
ple atoms with very long-lived excited states, which can only decay by two-photon
emission the dipole selection rules prevent any other decay channels. Multi-photon
ionization will be the subject in Sect. 5.5.5 and from the perspective of strong laser
fields in Sect. 8.5.
We consider the excitation of a target atom or molecule Tg into a state |b with
energy Wb from an initial state |a with energy Wa by several, say N photons of
angular frequency :
Tg(a) + N Tg(b). (5.41)
Of course, energy must be conserved in such processes, so that
N = Wb Wa = Wba = ba . (5.42)
While in (4.46) perturbation theory was applied only in 1st order, one now has
to carry it out consequently at least up to N th order to determine the transition
probability for this particular type of process. While in 1st order the transition rate
(4.63) was proportional to the intensity I of the laser field, in N th order it becomes
proportional to the N th power of it. Thus, the N photon transition rate will be
(N ) (N )
Rba = ba N I N with = I /, (5.43)
5.3 Multi-photon Processes 245
(N )
where ba is a so called generalized cross section. The photon flux is given in
units [] = Photons m2 s1 and the unit of the generalized cross section [ (N ) ] =
m2N sN 1 is thus a little bit unused but meaningful. In the following we shall talk
about processes which are induced by an intense laser field which is, however, not
yet strong enough to completely deform the subject of investigation or even to
destroy it completely.
5.3.1 Two-Photon Excitation
The question whether two photons may be absorbed simultaneously was first studied
theoretically by a student of Max B ORN, Maria G PPERT-M AYER (1931), who
later in 1963 won the N OBEL prize (together with W IGNER and J ENSEN) for her
work on the nuclear shell structure. In her PhD thesis ber Elementarakte mit
zwei Quantensprngen (On elementary acts with two quantum jumps) she laid the
foundations of this field which, however, only started to flourish with the advent
of powerful lasers, long after her pioneering work. We have already prepared in the
last chapter the tools for understanding what happens, so that now it will take only
a few steps to sketch the basic theory necessary for catching up with state-of-the-art
research in this field. In Sect. 5.5.5 we shall return to it.
We simply have to reintroduce the transition amplitude (4.52) (obtained in 1st or-
der perturbation theory) on the right hand side of the original ODE (4.45), use (4.50)
and to integrate. This leads to the probability amplitude in 2nd order perturbation
theory:

ie2 E02 t Db
D a ei(ba 2)t

D

D a eiba t
dt
b
cb (t) = + (5.44)
42 0
( a ) ( a )

+ emission terms .
The expression sums (in principle) over all basis states | of the system.7 Charac-
teristic for all absorption terms of this sum (first line) are the resonance denomina-
tors ( a ) where a = W Wa gives the energy difference between initial
state |a and an intermediate state | , each with exponentials in the nominator re-
sembling those in the 1st order approximation (4.51).
We see now clearly how two photon absorption comes into play: the first ex-
ponent i(ba 2)t in (5.44) common to all terms in the sum simply replaces
i(ba )t in (4.51): the double angular frequency 2 of the electromagnetic wave
assumes the role of . These terms, characteristic to 2nd order, oscillate rapidly and
average out in the integration unless 2 = ba in analogy to terms with ba
7 We have written out here only the absorption terms; the emission terms are identical, except that
is replaced by and
D by
D (and vice versa).
in 1st order. Terms with exp(iba t ) vanish in any case when the integration is car-
ried out.
Hence, for absorption only the first terms of the sum over in (5.44) contribute
to the integral, and exp[i(ba 2)t ] is pulled out of the sum. (The correspond-
ing emission terms in (5.44) contain exp[i(ba + 2)t ] and describe two-photon
emission in the case that ba < 0.) All further steps correspond to those in 1st or-
der: one carries out the integration from t = 0 to t and makes the transition from
probability amplitudes to transition probabilities in the limit t . Dividing that
by t gives a transition rate proportional to E04 (hence to I 2 ) and to the D IRAC delta
function this time of the argument ba 2.
Thus, the resonance frequency for a transition from |a to |b by simultaneous
absorption of two photons (both of energy) becomes 2 = ba as expected.
The necessary extension to levels of finite width or for transitions into the contin-
uum follows the scheme used in the one-photon case. Specifically, for a L ORENTZ
broadened line one replaces
(ba 2) gL (2), (5.45)
where the line profile (5.7) of the excited state |b is now a function of twice the
photon energy. From the thus derived rate one pulls out the square of the photon
flux (I /)2 to obtain the generalized cross section for two-photon excitation in
2nd order perturbation theory:
2
b| D| |
D|a
ba = (2)3 2 ()2 gL (2).
(2)
(5.46)
W a

As indicated in this expression, in principle one has to sum not only over infinitely
many discrete basis states | but also to integrate over the continuum. This can
lead to rather elaborate calculations. Fortunately, however, with differing more
and more from a, the increasing denominators as well as the decreasing magnitude
of the matrix elements D a = |
D|a usually leads to satisfactory results with only
a finite number of terms.
One verifies easily that this generalized two-photon cross section is derived in
units [ba ] = m4 s since the line profile is given in [g(2)] = eV1 and the ma-
(2)
(2)
trix elements are measured in units of length. Note that ba as defined by (5.43)
refers in principle to a strictly monochromatic electromagnetic wave (e.g. a highly
stabilized laser source) which is much narrower than the atomic linewidth given by
gL (2). For a radiation source of broader bandwidth only the fraction of the ra-
diation spectrum I() overlapping with the atomic linewidth is relevant just as
in the one-photon case. For a spectrum with finite bandwidth, in the present case of
two-photon absorption one has first to convolute I() with itself, i.e. with I( ),
and the result in turn with gL (2) to finally obtain an expression analogous to
(5.38).
Several extensions of the procedure described above are obvious. Excitation pro-
cesses with two photons of different frequency and/or polarization may be consid-
ered by adding to the interaction potential (4.55) a second electric field vector. The
expressions obtained by carrying this consequently through to 2nd order pertur-

bation theory are correspondingly more complicated and contain interesting cross
terms of the two frequencies. Also, one may include more than two photons into the
calculation by applying time and again the procedure described above to the original
ODE (4.45) as often as required, which leads to accordingly more complex final
expressions for the generalized cross section.
Clearly, (5.46) is also valid for multi-electron systems. One simply has to re-
place the matrix elements of D by a sum over all electrons according to (H.29) as
exemplified for the two electron case and single photon absorption in Sect. 7.5. In
detail the evaluation will not always be trivial to do.
As mentioned above, although in principle understood since 1931, multi-photon
transitions became experimentally accessible only after 1960 when lasers were in-
troduced as efficient light sources for spectroscopy: generalized multi-photon cross
sections are extremely small but according to (5.43) this may be compensated
by sufficiently high light intensity. To give one example: the two-photon ionization
cross section for Xe is (2) = 1.16 1049 cm4 s at a wavelength of = 193 nm
(L AMBROPOULUS 1985). From a typical classical light source one expects at such
wavelengths a few W cm2 at most, which would lead to an ionization rate of
ca. 1013 s1 per atom, i.e. on average one expects an atom to be ionized in 105
years! In contrast, a typical excimer laser at that wavelength (pulse duration of 10 ns,
pulse energy 100 mJ) may be focussed onto an area of 0.1 mm2 . This corresponds
to 1010 W cm2 and would lead to an ionization rate of about 107 s1 per atom. In
other words, during 10 ns, the duration of one laser pulse, there is already a 10 %
chance for such a process to happen, and thus the two-photon process can be de-
tected conveniently. Consequently, excitation and ionization by two or more photon
absorption plays an important role in todays laser spectroscopy.
One important advantage of multi-photon processes is that they allow to over-
come some of the limitations imposed by the rather rigid dipole selection rules for
single photon spectroscopy. Two-photon selection rules can be read directly from
(5.46). The product of the matrix elements b| D| |
D|a implies that selection
rules are multiplied so to say. Each of the two matrix elements corresponds to a
set of specific dipole selection rules. The product thus has the same effect as if the
two processes were carried out one after the other although the absorption of the
two photons is simultaneous. In place of the selection rules for one-photon transi-
tions summarized in Sect. 4.4.3, for two photon transitions on obtains for the total
angular momentum j = 0, 2, the parity selection rule becomes = 0, 2, and
for the projection quantum number m = 2q holds (if two photons of polarization
q = 0, 1 are absorbed). Using different polarizations for the two photons allows
for additional flexibility.
Important examples for two-photon absorption transitions between s states as il-
lustrated in Fig. 5.6 for the 1s 2 S 2s 2 S transition in the H atom. The intermediate
states | over which we have to sum here are all the bound |np states as well as
the corresponding states |p in the ionization continuum only to these the matrix
elements Db and D a do not vanish. We point out that these states are energeti-
cally not positioned in between initial and final state and thus, are far from being
Fig. 5.6 Two-photon W / Eh

transition 1s 2s in atomic s p
0.0
hydrogen (full red arrows). |
3s 3p
The red dashed arrows
indicate the coupling from - 0.125 2s 2p
initial state |1s to the
intermediate |p states and
from them to the final |2s
states by Dp1s and D2sp ,

respectively
- 0.500 1s
resonant. Nevertheless such processes can be studied with high efficiency and have
played a key role in the ultra high precision spectroscopy of atomic hydrogen which
led to the N OBEL prize for Ted H NSCH (2005). As we shall discuss in Sect. 6.1.8,
crucial for this kind of precision spectroscopy is the possibility to study two-photon
absorption in the gas phase completely free of D OPPLER broadening.
As a somewhat exotic but interesting field of atomic physics we finally mention
spectroscopic studies relating to parity violation. They rely on such high precision
spectroscopy. In this context, e.g. so called nuclear anapole moments have been
observed in 2 photon excitation of heavy alkali atoms (see e.g. B OUCHIAT 2007).
5.3.2 Two-Photon Emission
The inverse process, induced two-photon emission, is of course equally possible.

But even spontaneous two-photon emission has been observed in some cases where
one-photon transitions are strictly forbidden. Since in effect it is the vacuum field
which induces such transitions, these are extremely weak processes which accom-
pany the decay of very long-lived excited states. Particularly well studied exam-
ples are 2s 2 S states8 of hydrogen like ions (much easier to manipulate and to de-
tect than neutral species). First experiments for two-photon decay were reported by
L IPELES et al. (1965) and N OVICK (1972). Metastable, hydrogen like He+ ions in
the 2s 2 S state were investigated with at that time still rather elaborate photon-
photon-coincidence counting techniques. For He+ the 2s 2 S state the lifetime was
determined to be 2 ms. This is in very good agreement with theoretical predictions
of a two-photon decay rate R1s2s = 8.228 Z 6 s1 , i.e. for Z = 2 ca. 1.9 ms.
The fact that the decay rate increases Z 6 suggests to study such processes for
highly charged ions which are available in state-of-the-art ion storage rings. Fig-
ure 5.7 illustrates such an experiment with He like 58 Ni26+ performed by S CHF -
FER et al. (1999). Figure 5.7(a) shows the term scheme of the ion. In addition to
the 2 1 S0 state of interest in this experiment several other states are excited in this
experiment which may decay by one or the other mechanism into the ground state.
8 The
designations 2 S, 2 P used here and 1 S0 , 3 P0 , etc. used in Fig. 5.7 will be explained in detail in
Chap. 6.
W- W11S / keV (a) (b)
coincident
poton rate
0
1.65fs
7.80 21P 1
71ps
E1 2 3 P2
154ps 13fs 23P1
21S0 3
2.5ns 2 P0
1
10
7.75 E1E1
M2
8
2 3 S1
2E1 2.3ns
6
1
4
7.70 E1 10
/ ke
M1 8
2 6
2
V
4
58Ni26+ 2 2
/ keV

0
0 11S0 0
Fig. 5.7 Two-photon decay of the 1s2s 1 S0 state in He like 58 Ni26+ . (a) Term scheme with several
neighbouring levels and transitions. (b) Two-photon coincidences as a function of the emitted pho-
ton energies 1 and 2 . The red dotted line emphasizes the transition of interest 2 1 S0 1 1 S0 .
Figure adapted from S CHFFER et al. (1999)
Lifetimes and process types are noted in Fig. 5.7(a).9 Figure 5.7(b) shows the mea-
sured coincidence rate for detecting simultaneously two photons as a function of
their respective energies, 1 and 2 . Accidental coincidences with some of the
other processes make the results somewhat intransparent hence the red dotted
line indicates the 2E1 decay process of interest. Energy conservation requires that
1 + 2 = W2 1 S0 W1 1 S0 . One sees clearly that energy sharing among the two
photons leads to a broad distribution with a flat maximum for similar energies while
the probability that one of the photons carries all energy is zero (barely recognizable
in Fig. 5.7 due to accidental coincidences).
Precision measurements of such decay processes are of considerable interest
as an important testing ground for relativistic multi-electron quantum mechanical
computations. Interestingly, special two-photon transitions (3 P0 1 S0 otherwise
strictly forbidden for 2E1 processes) are discussed in the context of investigating
parity violation in atomic physics (B OUCHIAT and B OUCHIAT 1997).
Section summary
Multi-photon processes can be induced by intense electromagnetic radiation.
(N ) (N )
The rate for absorption of N photons is Rba = ba (I /)N (at not too
(N )
high intensities I ), the generalized cross section ba being measured in
m2N sN 1 . Resonance conditions are now N = ba .
9 We cannot enter into a discussion of the interesting details here, which require knowledge of the
underlying atomic physics still to be explored in later chapters. For the advanced reader it may
be particularly interesting that the high nuclear charge Z implies strong spin-orbit interaction and
leads to configuration mixing. Thus, the transition 2 3 P1 1 1 S0 is dominated by an E1 process,
while transitions 2 3 P2,0 1 1 S0 with J = 2 and 0, respectively, are strictly E1 forbidden due to
N photon cross sections may be computed in N th order perturbation theory.

This involves in principle summation (and integration) over all intermedi-
ate discrete (and continuum) states as exemplified for two-photon excitation
in (5.46). Corresponding expressions hold for induced multi-photon emission
processes.
In addition, (very weak) spontaneous two-photon (2E1) emission processes
are also observed. Their probability rises with the nuclear charge Z 6 and
has been studied for highly charged ions.
5.4 Magnetic Dipole and Electric Quadrupole Transitions

If electric dipole transitions (E1) are forbidden, higher order single photon pro-
cesses may become effective (in addition or alternatively to the multi-photon pro-
cesses just discussed). We have already noticed this in passing while discussing
two-photon decay in Sect. 5.3.2. The present section describes how magnetic dipole
(M1) and electric quadrupole transitions (E2) arise when treating the full pertur-
bation operator in 1st order perturbation theory. Albeit much weaker than E1 pro-
cesses, such transitions are observable in a variety of interesting cases. Of particular
spectroscopic and practical relevance are M1 transitions, being the basis for all mag-
netic resonance spectroscopy (see Sects. 9.5.2 and 9.5.3 for more details about EPR
and NMR, respectively).
So far, we have treated the interaction potential of the electromagnetic wave with
the atomic (or molecular) electrons only approximatively. The full expression (H.21)
it
(r, t) = eE0
U De D e+it (5.47)
2

2I
with (H.23) eE0 = eA0 = ecB0 = e (5.48)
c0
eikr ieikr
and (H.22)
D= e
p= e (5.49)
me me
contains the term exp(ik r) = 1 + ik r + , which in dipole approximation was
just replaced by 1. We shall now treat the second term of this expansion:
i

D(2) = (k r)(e
p ).
me
We first rewrite10 this by trivial addition and subtraction of an extra term:
i i

D(2) = [k r e
p e r k p] + [k r e
p +e r k
p ].
2me 2me
10 Above we have indicated once the precedence of vector multiplication by brackets. In all follow-
ing expressions we omit this notation: vector products always have precedence over scalar products
while these again have precedence over simple scalar multiplications.
5.4 Magnetic Dipole and Electric Quadrupole Transitions 251
We exploit now some rules for multiple products from vector algebra, using the fact
that k e B. The first bracket is identical to the scalar product of two vector
products (making sure not to permutate r and p):
i i

D(2) = k e r
p+ [k r e
p +e r k
p ].
2me 2me
Per definition r p= L is the angular momentum of the electron and k e B 0 =

(B 0 /2)(e it e ), see also (H.7). In addition, we note that with k e the projec-
it
tion of the position vector onto the polarization vector, e r, and the projection of the
momentum vector onto the wave vector, k p , represent perpendicular components
of the r and p operator, respectively. Hence, they commute and we may write
ik B 0 i

D(2) = L+ [k r e
p +k
p e r].
2me B0 2me
Next we apply the commutation rule (H.17), according to which ]

p=i(me /)[r, H
and obtain:
ik 1 ].

D(2) = B0
L [H k r e r k r e r H
2me B0 2
Multiplication with eE0 according to (5.48) and substituting in the first term k/ =
1/c finally gives:
e ecB0 ].
ecB0
D(2) = i L B0 [H k r e r k r e r H (5.50)
2me 2
We have thus managed to expand the operator (5.47) describing the interaction of
electromagnetic waves with matter beyond the simplest dipole approximation in
a suggestive manner. Obviously, (5.50) expresses the second most important term
as two different types of interactions due to the electromagnetic field: as we shall
see in a moment, the first one is responsible for magnetic dipole (M1), the second
one for electric quadrupole transitions (E2). Both have the same time dependence
as specified in (5.47) and may be treated completely analogous to electric dipole
transitions in 1st order perturbation theory. We thus simply need to determine the
transition matrix elements between |a and |b of ecB0 D(2) for the two components.
The magnetic dipole term may then be written as
1
L
(ecB0 B0
D)M1 = B B 0 = M (5.51)
i
where according to (1.150) we have inserted explicitly B OHRs magneton B to
express the characteristic nature of this interaction. The magnetic dipole interac-
tion energy M B(t) induces M1 transitions in complete analogy to the electric
dipole interaction energy D E(t) which is responsible for E1 transitions. Let us
now obtain an estimate for the relative order of magnitude for these two types of
transitions. Both interactions enter the transition probability quadratically. We as-

sume L/ 1 and D = erab ea0 /Z we obtain (exploiting E0 = cB0 ) as
an order of magnitude
|M B 0 |2 Z 2 2 (Z)2
= 1.33 105 Z 2 . (5.52)
|D E 0 |2 4m2e a02 c2 4
The probability to induce an M1 transition is thus (for light atoms) about five orders
of magnitude smaller than for an E1 transition at a given intensity of the radi-
ation. The same holds for spontaneous decay probabilities. Sure enough, for high
charge states M1 or even M2 may become appreciable, as we have already seen in
Sect. 5.3.2. The M1 transitions are crucial for all magnetic resonance spectroscopy
where transitions between different magnetic substates |j mb |j ma within the
same electronic level are studied (see Sects. 9.5.2 and 9.5.3) strictly forbidden for
electric dipole transitions.
For a quantitative treatment of M1 processes we also have to account for the elec-
tron spin. As mentioned already in Chap. 1, the magnetic moment associated with
the electron spin is
= 2B . This implies that the magnetic moment M L = B L/
of the orbital angular momentum has to be replaced by M = B ( L + 2S)/. Since
M1 transitions are single photon processes, selection rules and line strength for M1
transitions thus follow from the corresponding transition matrix elements. For the
determination of transition rates and cross sections one simply has to replace in all
relevant expressions of Sects. 4.35.2 the electric dipole transition matrix element
eE0 Dba by the corresponding one for the magnetic dipole interaction:
1 B B B B
(ecB0 b|L + 2
(M1)
D)ba = S|a = b|J + S|a . (5.53)
i ec B ec B
The components of total angular momentum and spin are thus projected onto the
magnetic field direction of the electromagnetic wave with B E k. Tools for ex-
plicit evaluation of the matrix elements are summarized in Appendix C.4.3. Details
will be discussed in several of the following chapters.
Let us come back now to the second part in (5.50), the electric quadrupole term,
from which the transition matrix element follows as
eE0
a|eE0
D|b (E2) = (Wb Wa )b|k r e r|a
2
eE0
= b|ek r e r|a . (5.54)
ba
With k = /c we have introduced here the unit vector ek = k/k in the direction of
the wave propagation and the transition wavelength ba . Alternatively with (5.47)
(5.49) and (4.1) the Hamiltonian for the electric quadrupole interaction may also be
written as

(r, t)(E2) = e ba E(t) r (k r).
U (5.55)
2
5.4 Magnetic Dipole and Electric Quadrupole Transitions 253
Since k and e are always perpendicular to each other, the two projections of r
onto these directions correspond to orthogonal components of r. Thus, depending
on the choice of the (at) and (ph)frame (see Fig. 4.3), the relevant matrix ele-
ments are of the type xy = Q22 / 3 (propagation in x-direction, polarization in
y-direction or vice versa), xz = Q21+ / 3 (propagation
in z-direction, polarization
in x-direction or vice versa) or yz = Q21 / 3. Light propagation under 45 in
the xy plane with a polarization angle vector also aligned within
the xy plane (az-
imuthal angle 45 ) is described by (x 2 y 2 )/2 = Q22+ / 3. The terminology
used here for the real components of the so called quadrupole tensor (an irreducible
tensor operator) is summarized in Appendix F.3. Any other geometry may be ex-
pressed as a linear combination of the components of this tensor operator, typically
by rotation of this operator with the help of the rotation matrix of rank 2 according
to Appendix E. Quantitatively the transition probabilities and cross sections for E2
transitions are obtained by replacing the electric dipole transition matrix elements
eE0Dba throughout Sect. 4.3 with
(1)q+1
b|eE0
D|a (E2) = b|Q2q |a . (5.56)
3ba
Here q characterizes the geometry of absorption or emission as just outlined. It
is advantageous that angular momentum algebra allows to derive the selection rules
in a relatively straight forward manner. For multi-electron atoms one has to account
for the respective angular momentum coupling schemes which will be discussed
in later chapters. A brief survey on selection rules for E2 transitions is given in
Appendix C.4.2.
We finally estimate the relative probability for such transitions. The E2 ma-
trix element (5.54) is eE0 xi xj /ba , xi and xj being two position coordinates,
while the E1 matrix element is eE0 xi . For a rough estimate we derive the
averaged position coordinates from (2.130) with xi xj r 2 n n4 a02 /Z 2 and
xi r n n2 a0 /Z. Thus, the ratio of the transition probabilities at identical
intensities is on the order of
2 2
|a|eE0
D|b (E2) |2 n a0
. (5.57)
|D E 0 |2 Zba
This ratio is thus determined by the square of the ratio of atomic dimension to
wavelength. For a typical transition in the VIS or UV spectral range for small nu-
clear charges Z and low principle quantum numbers n, one estimates a ratio of
(0.1 nm /300 nm)2 107 . For high Z, the probability for E2 processes even de-
creases with Z 2 in contrast to M1 transitions which increase with Z 2 as discussed
above. Thus, E2 processes have not played a major role in classical atomic, molec-
ular or solid state spectroscopy except in the X-ray region where they may be
considerable.
However, E2 transitions are of increasing interest as states with ever higher prin-
ciple quantum numbers n are studied, since orbital radii increase with n2 . A so
called RYDBERG atom with n = 100, say, has a diameter of ca. 500 nm, and thus
a size comparable to optical wavelengths. For a recent experimental example, also

illustrating the details of evaluating (5.54), see e.g. T ONG et al. (2009) who studied
RYDBERG states in ultra-cold rubidium for n = 2759. Quite generally speaking,
RYDBERG atoms, these nearly macroscopic quantum objects, often surprise with a
lot of interesting and rich physics.11
Section summary
By expanding the spatial dependence exp(ik r) of the electromagnetic
field beyond the first, constant term one obtains magnetic dipole and elec-
tric quadrupole interaction terms, proportional to B b| J + S|a and to
b|Q2q |a /ba , respectively.
The rates for the resulting M1 and E2 transition can be evaluated in 1st order
perturbation theory in full analogy to those for E1 transitions. In comparison
to the latter M1 transitions are typically less probable by a factor of 105 Z 2 ,
E2 transitions by about n4 (a0 /ba )2 Z 2 .
5.5 Photoionization
The discovery and interpretation of the photoelectric effect at the beginning of

the 20th century was one of the fundamental mile stones on the way to modern
physics as we have outlined in Sect. 1.4.1. Today one classifies the corresponding
processes as photoionization and addresses with this term a major part of the contin-
uum physics of atoms and molecules. While in the past chapters we have essentially
discussed transitions between discrete, bound states which are induced by electro-
magnetic fields, we now want to give at least a brief introduction into, so to say, this
other half of atomic and molecular physics. The fundamental concepts were already
explored in the 1930s, while in the 1960s and 1970s they have been brought into full
fruition, when extensive experimental studies on photoionization have been carried
out: The availability of synchrotron radiation (M ADDEN and C ODLING 1963) as
intensive light source in the VUV and X-ray region gave way to rich harvesting of
precise data and profound physical insight. Modern laser physics too has made an
important contribution to the development of photoelectron spectroscopy as a key
tool for studying the electronic and ro-vibronic structure of atoms, molecules and
solid states. Todays efforts are mainly concentrated on molecules and clusters. In
surface physics too these methods, original developed in atomic physics, are now
widely used. Photoelectron spectroscopy has matured to an indispensable tool in
many areas of physics, chemistry and materials research.
The following introduction outlines some basic theoretical considerations and
illustrates these by a few experimental examples. Characteristic findings for multi-
electron systems will be presented in Chap. 10, and a more detailed discussion of
11 This expression is often used like a synonym for up to now only poorly understood.
5.5 Photoionization 255
methods and results is given in the context of molecular photoelectron spectroscopy

in Sect. 5.8, Vol. 2. Readers interested in the historical development of this important
field are referred to the original literature and a number of good review articles (see
e.g. B ETHE and S ALPETER 1957; B URGESS and S EATON 1960; C OOPER 1962;
C OOPER and Z ARE 1968a; M ANSON and S TARACE 1982; S AHA 1989; S CHMIDT
1992, and references there).
5.5.1 Process and Cross Section
We want to obtain a quantitative description of the ionization process for an atomic

(or molecular) target, denoted in the following as Tg, by interaction with an elec-
tromagnetic field. We assume that the field induces a transition from a bound state
|a = |n into an unbound continuum state |b = | during which one photon of
energy is absorbed. For simplicity we ignore again the electron spin (as a rule
the spin does not change its orientation for low Z atoms) and also concentrate on
effective one electron systems.
While bound states are characterized by a principle quantum number n, an orbital
angular momentum quantum number , and a binding energy Wn < 0, continuum
states describe the free electron with an asymptotic kinetic energy > 0 and its
angular momentum . The continuum energy is related to the wavenumber ke of
the electron at infinite distance from the nucleus by
= 2 ke2 /(2me ). (5.58)
The photoionization process may be written schematically as

Tg(n ) + Tg+ + e with WI = , (5.59)
with the ionization potential WI which is related to the binding energy Wn =

WI < 0 of the initial target state Tg(n ).
A closely related process is the removal of an electron from an anion, the so
called photo-detachment:

Tg (n ) + Tg + e . (5.60)
It may be treated in a similar manner as photoionization, the ionization potential

WI being replaced by the electron affinity WEA (see Sect. 3.1.5). The main differ-
ence between the two processes is the rapid decrease of the interaction potential
with r between the remaining neutral atom and the emerging electron, in contrast to
C OULOMB interaction in the ionization case, which leads to a quite different asymp-
totic behaviour of the continuum wave functions. We shall discuss this in Sect. 5.5.4.
Just as for optical excitation, the photoionization cross section is given by (5.31).
However, the transition matrix element, Da = r a e in E1 approximation according
to (4.57), is now determined between the discrete initial state |a and the final con-
tinuum state | . Usually one normalizes the latter in energy scale (see Appendix J.1)
so that its dimension is L3/2 Enrg1/2 . In this normalization the density of states
(number of states per energy interval), g(W ) in (5.31), is already built into the wave
functions.12 The photoionization cross section thus becomes
a () = 4 2 |
Da |2 . (5.61)
Note that the squared matrix element has the dimension L2 Enrg1 so that a is
still a genuine cross section, with dimension L2 , measured e.g. in [a ] = b (see
Appendix A.2).
Extending now the definition (H.34) to the continuum, one introduces the optical
oscillator strength density (OOSD)
(opt)
dfa + WI
=2 2
|
Da |2 = 2 |za |2 . (5.62)
d E h a0 Eh a02
The photoionization cross section may thus be written as

(opt)
dfa
a () = 2 2 Eh a02 . (5.63)
d
The numerical prefactor is the same as for (5.32) while g() is now included in
the OOSD. It has the dimension Enrg1 in contrast to excitation of discrete states
(opt)
where the oscillator strength fba for a specific transition is a dimensionless num-
ber.
In the literature one also finds different notations of (5.63). They may, however,
be transferred into each other by (H.34)(H.44). For linear polarization, often the
field vector is assumed to point into z-direction, without loss of generality. With
(4.77) one finds (still neglecting the electron spin)

za = |z|a = r|n m C10 | m . (5.64)
Various other choices of geometry are possible in analogy to the treatment of

transitions between bound states which we have discussed earlier.
5.5.2 BORN Approximation for Photoionization
Following B ETHE and S ALPETER (1957), for high but not relativistically high pho-
ton energies
WI me c2 (5.65)
12 Alternativelyand completely equivalent one may employ continuum wave functions which are
determined as an orthogonal basis set for a large but finite volume L3 . In this case a density of
states corresponding to (5.31) has to be used in (5.61).
Fig. 5.8 Coordinates and z

angles for describing
photoionization
K
k ke

y
e
x
one may approximate the continuum wave function of a free electron by a plane
wave (1st order B ORN approximation, FBA):

me ke ike r
r|k e = e . (5.66)
(2)3 2
The normalization factor under the square root is the density of states in an element
of solid angle d per unit volume (dimension L3 ) and energy according to (2.58).
The geometry for describing the photoionization process is sketched in Fig. 5.8:
the ionizing light propagates into z-direction with a wave vector k. We assume it
to be linearly polarized, its polarization vector e pointing in x-direction. After ion-
ization the electron is emitted into the direction k e . At these high energies we can
no longer use the dipole approximation. Rather the full expression (H.22) for the
transition operator has to be employed:
1 ikr 1 ikr

D= e e
p= e p E . (5.67)
me me
We denote the component of e
p in the direction of the polarization vector e as p
as introduced in (2.50). The transition matrix element for photoionization of the
state |a is thus written
1 ikr 1 ikr
Da = k e |
D|a = k e | e p e |a =
pe k e | e |a . (5.68)
me me
e e k so that p
The latter identity holds since p e and eikr commute, and we can
apply (2.35). We recall now that according to (2.51) each plane wave is an eigen-
e with the eigenvalue p cos = ke cos , where is the angle between
vector of p
polarization vector e and the direction of electron emission k e . Thus, applying (5.66)
the matrix element is written in integral form:
3/2
ke
Da = k e |
D|a = cos (2)3/2 ei(kk e )r a (r)d3 r. (5.69)
me
With K = k e k we may abbreviate

3/2
a (K) = (2) eiKr a (r)d3 r (5.70)
which is just the F OURIER transform of the initial wave function a (r), i.e. its rep-
resentation in momentum space. The matrix element Da refers to electrons which
are emitted into a specific solid angle. Hence, inserting (5.69) into (5.61) leads to
the differential cross section for photoionization in FBA:
1 ke3 2
da () = 4 2 cos2 a (k e k) d. (5.71)
me
The function a (k e k) may be derived either by (5.70) from the wave function
in space or directly from the S CHRDINGER equation transformed into momentum
space (see e.g. B ETHE and S ALPETER 1957, Chap. 8).
Specialization to ns Initial States and H Like Atoms

The general form of the F OURIER transform is a (p) = Fn (p) Y m (, ). For ns
initial states with radial wave functions Rns (r) it may be written as:

1
a (K) = 2
Rns (r)r dr eiKr sin d. (5.72)
2 0 0
For the angular integration we choose z K so that eiKr = eiKr cos and

2
a (K) = drRns (r)r sin Kr (5.73)
K 2 0
which can be readily integrated. For large K = |k e k| and hydrogen like initial ns
states it may be approximated by
3/2
2 2Z 5/2 a0
ns (k e k) = (5.74)
n3/2 a04 |k e k|4
(B ETHE and S ALPETER 1957, Eq. (70.4), in a.u.). Introducing this into (5.71) leads
to a differential photoionization cross section for H like ns electrons:
1 2 ke3 8Z 5 a3
da () = 4 2 cos2 2 3 8 0 d
me n a0 |k e k|8
Z5 1 cos2
= 32 d. (5.75)
me n3 (a0 ke )5 [1 (v/c) cos ]4
The expression in brackets [ ] in the second row is obtained by exploiting (5.65)

and (5.59) = = ke /2me ; with the velocity of the ejected
so that with (5.58) 2 2
electron v = 2/me and the photon momentum k = /c one obtains 2k/ke =

v/c 1 and neglects (k/ke )2 . Consequently, further expansion in powers of v/c
and neglecting all nonlinear terms finally gives
Z 5 a02 cos2 [1 + 4(v/c) cos ]

da = 64 d. (5.76)
n3 (2/Eh )7/2
Angular distributions will be discussed in the next subsection. Here we just integrate
over the full solid angle. From Fig. 5.8 we read cos = sin cos and integration
4 2
over . . . d = 0 d 0 . . . sin d simply gives a factor 4/3. Thus, for H
like atoms in ns states the integral photoionization cross section in FBA is given
by:13
4
da 256 Z 5 a02
ns = d = 3 . (5.77)
0 d 3 n (2/Eh )7/2
Even if this result of the B ORN approximation is only valid at high (but not rel-
ativistically high) energies, it describes the general trend for H like orbitals reason-
ably well: a dramatic decrease of the cross section with photon energy ()7/2 ,
and strong dependence on nuclear charge Z 5 which leads to substantial photoion-
ization cross sections for the heavier atoms, as well as the decrease for higher or-
bitals n3 . As we shall see in Chap. 10 the inner shells of large atoms may very
well be described by H like orbitals so that the FBA leads to quite realistic first es-
timates for absorption of X- and -rays (see Sect. 10.5.3). A comprehensive data
source for such cross sections is C HANTLER et al. (2005).
Specifically for photoionization of the H 1s orbital (5.77) gives in numbers:
1s / cm2 = 1.609 1023 (/ keV)7/2 . (5.78)
Even though FBA is a high energy approximation, it is instructive to show it over

the whole energy range. One has to replace the approximation (5.74) by the exact
F OURIER transform of the wave function and obtains
1s / cm2 = 1.609 1023 ( WI )3/2 (/ keV)5 . (5.79)
Thus, FBA predicts the photoionization cross section to disappear at threshold =

WI = 13.6 eV, to rise quickly to a maximum at around 20 eV and then to decrease
()7/2 . As illustrated in Fig. 5.9 this is not even close to reality at threshold:
Generally one finds (also from theoretical arguments) that photoionization cross
sections are finite at threshold (see also Sect. 7.2.7 in Vol. 2). They decrease rapidly
with increasing energy but convergence of FBA and the exact cross section from
the NIST data base (C HANTLER et al. 2005) is reached only when is several
times WI . At = 10WI the cross section is already below 1 % a rather typical
behaviour. The NIST data shown in Fig. 5.9 are based on a careful selection of
state-of-the-art theory. The experimental data in Fig. 5.9 are from one of the very
few quantitative data sets for photoionization of atomic hydrogen. They have been
determined from a H plasma in a shock tube (PALENIUS et al. 1976). Considering
the experimental difficulties one may speak about satisfactory agreement between
theory and experiment.
We point out, however, that the extremely structureless behaviour seen in Fig. 5.9
is a specialty of the H atom. Already for He (Chap. 7) we shall see between the
13 Since the integral is independent of the same expression holds for unpolarized light.
experiment (a) 10-18 (b)

(shock tube)
10-20
1s / 10-18 cm2
4
exact calc. (NIST)
1s / cm2
10-22
10-24
2 10-26
BORN approximation
10-28
10-30
0 10-32
20 40 60 80 100 0.01 0.1 1 10 100
/ eV log( / keV )
Fig. 5.9 Cross section for photoionization of atomic hydrogen as a function of photon energy .
Exact calculations (red lines) according to C HANTLER et al. (2005) are compared with the B ORN
approximation (black lines) computed by (5.79), and with experimental data (grey squares) from
PALENIUS et al. (1976). (a) Linear plot for low , (b) logarithmic plot for energies from 13.7 eV
to 400 keV. The dashed line indicates the decrease W 7/2
first and the second ionization threshold (He+ and He++ , respectively) a rich spec-
trum of so called autoionizing lines, superposed on the general trend of the pho-
toionization cross section. For larger atoms, the photoionization cross section may
even pass at low energies through a so called C OOPER minimum as will be dis-
cussed in Chap. 10. Nevertheless, in the limit of very high photon energies one
always finds the dramatic decrease of the photoionization cross section discussed
here.
5.5.3 Angular Distribution of Photoelectrons
The angular distribution of the emitted photoelectrons for polarized light is charac-
terized according to (5.76) mainly by cos2 = sin2 cos2 . At high photon ener-
gies the term 4(v/c) cos leads to a shift of the angular distribution into the direction
of the ionizing light, but may in general be neglected for non-relativistic electrons.
Without derivation we note here the result for the general, low energy case. Strict
application of the dipole approximation (E1) leads to an angular dependence of the
differential photoionization cross section
da a
= 1 + P2 (cos ) where 1 2, (5.80)
d 4
which is characterized by the so called parameter (also anisotropy parameter),
with the L EGENDRE polynomial P2 (cos ) = (3 cos2 1)/2. For ionization of
pure ns states as in H(1s) case the just discussed, (5.76) is reproduced with = 2
neglecting the high energy term which is included in the B ORN approximation in
contrast to the simplified dipole approximation.
The maximum of the electron signal for ns ionization is thus found at = 0,
i.e. in the direction of the polarization vector e (see Fig. 5.8) while perpendicular
to it, i.e. parallel to the wave vector k of the ionizing light no electrons are emitted.
One may say, the angular distribution of the state generated by linear polarized light
corresponds to a |px orbital. Its angular probability distribution is represented by
the angular distribution of the photoionization cross section. For not too high photon
energies the continuum states generated by photoionization are thus equivalent to
those obtained in dipole excitation (see Sect. 4.7).
For unpolarized or circularly polarized light one simply has to average the
differential cross sections for polarization in x- and y-direction. According to
Fig. 5.8 this implies averaging over the corresponding polarization angles with
cos2 = sin2 cos2 and = sin2 sin2 , respectively and leads to

da a
= 1 P2 (cos ) . (5.81)
d 4 2
For = 2 this characterizes a torus shaped (doughnut like) distribution around the
z-axis.
Both expressions are generally valid for not too high in photoionization of
atoms and molecules initially prepared in an isotropic population of states. One
finds, however, that usually depends on photon energy. As we shall see, is an
important quantity which describes the contributions of different partial waves in
single photon ionization processes.
We finally note that the photoelectron yield emitted at the so called magic angle
mag = 54.736 for which P2 (cos mag ) = 0 does not depend on . Systematic
studies of the integral photoionization cross section a are preferably done at this
emission angle (or at = mag , respectively).
5.5.4 Cross Sections in Theory and Experiment
The B ORN approximation is a high energy approximation. Its advantage is its sim-
plicity and the fact that it includes all multipole moments of the incoming plane
wave up to infinite order. We have seen, however, that it does not give reliable re-
sults at energies below about 5 to 10 WI . Yet it is this very energy range that
contains most of the continuum oscillator strengths, and provides the background for
many interesting atomic and molecular phenomena. Fortunately, one may in princi-
ple apply here the electric dipole (E1) approximation which according to (5.62) and
(5.64) just requires an accurate computation of the relevant radial matrix elements.
Thus, one only has to determine sufficiently exact wave functions for the initial
bound and the final continuum states.
As discussed in detail in Chaps. 2 and 3, the initial state |a of the system is
described by a wave function with well defined quantum numbers n m:
un
|a = Rn (r)Y m (, ) = Y m (, ). (5.82)
r
The asymptotic behaviour of the radial functions for the bound states is according
to (2.119) and (2.120)

un (r) r +1 and un (r) exp 2|Wn |r . (5.83)
r0 r
In contrast, the continuum wave functions | for a given energy of the free electron
may in principle contain all orbital angular momenta :

u (r)
| = a Y m (, ). (5.84)
r
=0 m =
The expansion coefficients a characterize the respective boundary and ionization

conditions, while the radial functions u (r) = rR (r) have to be calculated from
the radial S CHRDINGER equation (2.117). There, the binding energy Wn has to be
replaced by the continuum energy = k 2 /2 (in atomic units) and the potential has
to be adapted. While stable solutions for bound states exist only for well defined,
discrete binding energies, to any positive energy a meaningful wave function may
be determined.
As already discussed in Sect. 3.2.6, wave functions in the continuum have to
be normalized on the energy scale as described in Appendix J. In the case of pho-
toelectron emission (5.84) has to represent a radial outgoing electron flux and the
functions u (r)/r have essentially the character of spherical waves. Their asymp-
totic behaviour (in atomic units) is given by
u (r) r +1 and (5.85)

r0

2 ZC
u (r) cos kr + ln(2kr) + + (5.86)
r k 2 k

with k = 2, the magnitude of the wave vector,14 and the remaining charge ZC of
the atom after the ionization process. For photoionization the outgoing wave in the
field of the remaining ion is characterized the standard phase shift of the wave in
respect of a free C OULOMB wave in a ZC /r potential, by , the C OULOMB phase
shift (3.28) which rapidly decreases with k and by its characteristic logarithmic
phase.15
photo-detachment (5.60) from anions (ZC = 0) the asymptotic ra-
In the case of
dial function is 2/k cos(kr /2 + ). In practice, it may often suffice to
determine the radial functions un (r) and u (r) by solving the simple radial, one
particle S CHRDINGER equation (2.117). However, depending on the complexity of
the system studied and on the desired quality of the computation it may become nec-
14 In
the literature the energy is often given in RYDBERG units so that the factor 2 under the root is
omitted.
15 Forthe ionization of H like atoms only a naked ion remains so that for the outgoing wave in
the pure C OULOMB potential = 0. For all other cases is related with the quantum defect
according to (3.26).
essary to use more sophisticated methods, such as multi configuration H ARTREE -

F OCK (MCHF) to be discussed in Chap. 10. Finally, for multi-electron systems one
has to sum over the ionization cross sections for all electrons in the atom or molecule
studied which can be ionized at a given photon energy.
With the thus characterized wave functions of the initial bound state (5.82) and
the final free continuum state (5.84) the integral cross section can be calculated.
Without loss of generality as in (5.64) we again assume the polarization vector
to point into z-direction, i.e. we set the dipole transition operator D = ie r =
irC10 (, ). Then (5.61) gives after averaging over all initial states and summation
over all accessible final states:
2
4 2
C10 | m .
a () = r|n m (5.87)
2 + 1
mm
While the radial matrix elements |r|n are determined by the specific prop-
erties of the atom and the photon energy, the matrix element m |C10 | m (see
Appendix C.2.2) may be evaluated in close analogy to Sect. 4.4 where we have de-
scribed optical excitation processes. This leads to corresponding dipole selection
rules for photoionization: again = 1 must hold, since a photon with angular
momentum is absorbed and the transition = 0 is forbidden as a consequence of
parity conservation. We note, however, that in the continuum both final states, with
= 1 may be superposed at each photon energy above the photoionization
threshold. Also the selection rule m = m holds again for light linearly polarized in
z-direction. Evaluation of the sum and the matrix elements in (5.87) finally leads to
4 2 2
a () = r, 1 + ( + 1)r, +1
2
(5.88)
3
with the corresponding radial matrix elements

r, 1 = ru 1 (r)un (r)dr. (5.89)
0
Hence, unless the initial state is an s state with = 0 the photoionization cross sec-
tion contains contributions from two matrix elements involving continuum functions
with = + 1 and 1.
With some additional effort the continuum function (5.84) may be determined in
such a way that it describes an electron emitted into the direction k e . Without going
into details this leads to (5.80) with an energy dependent anisotropy parameter (see
Eq. (2) in C OOPER and Z ARE 1968a,b):
1
() = (5.90)
(2 + 1)[ r, 1
2 + ( + 1)r, +1
2 ]
'
( 1)r, 1
2
+ ( + 1)( + 2)r, +1
2
(
6 ( + 1)r, +1 r, 1 cos +1 () 1 () .
Fig. 5.10 Angular 1
e- yield / arb. un.

distribution of electrons at the
photo-detachment from Cu
according to C OVINGTON et
al. (2007)
0
- 45o 0o 45o 90o 135o 180o
polarization angle
Obviously is determined by the ratio r, 1 /r, +1 of the two radial matrix
elements involved and by the phase differences +1 () 1 () of the two partial
waves. One easily verifies that for = 0 (and thus r, 1 = 0) = 2 holds as
already found by the B ORN approximation. One also confirms without difficulty
that in the general case 1 2.
Figure 5.10 illustrates this by way of example for an experimentally determined
angular distribution in the photo-detachment from an anion
Cu [Ar]3d 10 4s 2 1 S + Cu[Ar]3d 10 4s 2 S + e
studied by C OVINGTON et al. (2007). Since the electron configuration of the copper
anion may be considered He like (two s electrons on top of a closed 3d shell), and
since one s electron is detached, one expects = 2 and hence according to (5.80)
a clean cos2 angular distribution. Figure 5.10 largely confirms this expectation.
A small deviation cannot be excluded by the experiment (the minimum at 90 may
not be exactly zero). One would attribute this to the fact that the description of
the electron wave function by a pure product wave function of all orbitals (in this
case a 4s electron has to be detached) is not completely correct for such a large
atom. Rather one has to account for some configuration mixing as to be discussed
in Sect. 10.2.3.
Figure 5.11 shows the rather typical ion beam apparatus used in this experiment.
The setup is in principle uncomplicated and self explaining. It requires, however,
high experimental precision. The determination of the kinetic energy of the electrons
is achieved by a 160 spherical sector analyzer (for a hemispherical analyzer see
Appendix B.3 in Vol. 2) which is characterized by its high resolution and good
focussing properties. The angular distribution shown in Fig. 5.10 has been recorded
as usual in such experiments: one rotates the laser polarization which is much easier
than turning the whole electron analyzer-detector setup. For = 0 the polarization
direction is parallel to the direction of electron detection.
Quantitative computations of cross sections for such multi-electron atoms require
a major effort. First, the wave functions for bound and continuum states have to be
determined accurately. Then the one electron dipole operator has to be replaced by
(H.29) summing over all electron coordinates. For a realistic description of such an
atom one also has to account for the spin and angular momentum coupling. The
ensuing matrix elements have to be evaluated using the angular momentum algebra
outlined in Appendix C.
2.4
m
4m
insertable 90o selector insertable
FARADAY magnet FARADAY
insertable cup cup
FARADAY
polarization einzel cup
adjustable
rotor lens einzel beam slits
polarizer lens
b e am l a ser
aperture anion source
ion o bea electrostatic
45 e - m steering elements
FARADAY laser power meter
cup
spherical electron energy
analyzer (160o segment)
Fig. 5.11 Typical experimental setup for studying photo-detachment from anions, here according
to C OVINGTON et al. (2007). The ion beam is focussed with several lenses and guided by deflec-
tors. A 90 magnet selects one anion species which crosses in the interaction region a CW laser
beam. The polarization of the latter may be rotated
V(r ) / Eh
0.5
e-
N Wkin
- 15 15
r /a0
WI
- 0.5
- 1.0
Fig. 5.12 Illustration of multi-photon ionization for the example of an H atom. In the continuum
the electron has a kinetic energy Wkin corresponding to (5.91) depending on the number of
absorbed photons N (the schematic is drawn to scale in a.u. for photons with a wavelength
= 800 nm). The electrons kinetic energy Wkin is indicated by downward pointing red arrows,
which indicate here that 0 to 5 additional photons beyond those necessary for ionization have been
absorbed (ATI)
5.5.5 Multi-photon Ionization (MPI)
Processes of N th order (perturbation theory) correspond to the absorption of N

photons as already discussed in Sect. 5.3 for excitation processes (and correspond-
ingly for emission). Of course, ionization of a target atom or molecule (Tg) is also
possible by such processes, even if the individual photon energy is below the ioniza-
tion potential, < WI . This is sketched in the potential energy diagram Fig. 5.12.
Fig. 5.13 Photoelectron

emission spectra for Xe (a) I = 2.21012 Wcm-2
according P ETITE et al.
(1988), clearly showing ATI.
Nd-YAG laser pulses of a
electron signal / arb. un.

wavelength 1064 nm 1.165eV
(1.165 eV) at two different
laser intensities I are
compared. While (a) at (b) I = 1.11013 Wcm-2
I = 2.2 1012 W cm2 only
one or two extra photons
above threshold are absorbed,
(b) this number rises up to
about nine if the intensity is
increased by a factor of only
five 0 2 4 6 8 10 12 14
Wkin / eV
The process and its energy balance is now given by
2 ke2
Tg + N Tg+ + e (Wkin ) with WI + N = Wkin = . (5.91)
2me
Obviously the kinetic energy Wkin of the electron after ionization depends on the
number N of photons absorbed. The probability for a multi-photon ionization pro-
cess (MPI) depends on the photon flux = I /() just as in the case of multi-
photon excitation. In N th order perturbation theory one writes the transition rate
from a bound initial state |a into the continuum state | as
(N )
Rke a = (N ) N I N , (5.92)
where (N ) is a generalized ionization cross section as we already know it from the

excitation process described by (5.43).
As illustrated in Fig. 5.12 it is possible that more photons are absorbed than nec-
essary to reach the threshold for ionization so called above-threshold ionization
(ATI) process. A very clear experimental verification is shown by the photoelectron
emission spectra in Fig. 5.13. In this, by now classic, experiment of P ETITE et al.
(1988) Xe atoms were ionized with Nd-YAG laser pulses ( = 1064 nm) of 130 ps
duration. The photon energy is 1.165 eV and the spectra document that far more
photons are absorbed than the 11 necessary for reaching the ionization threshold
clear evidence for the ATI process: the additional energy is found as kinetic en-
ergy Wkin of the electrons according to the energy balance in (5.91). Comparison
of Fig. 5.13(a) and (b) shows that the number of additional photons absorbed rises
quickly with increasing laser intensity.
For MPI the angular distribution of the emitted electrons may become much more
complicated than for one-photon ionization, which is described by (5.81) implying
an angular momentum transfer of . If N photons are absorbed, angular momen-
tum transfer of up to N is possible and leads to a correspondingly expanded
summation over all accessible continuum states from (5.84) which will be reflected
in the differential cross section. Considering parity conservation, for a two-photon
ionization process the partial waves = and 2 contribute (as long as 0),
for a three photon process we have = 1 and 3 and so forth. When linearly
polarized light is used the selection rule m = 0 remains valid, but for circularly
polarized light also higher values of N m N are accessible. It is obvious
that the angular distribution of the emitted electrons is correspondingly more com-
plex.16
As in multi-photon excitation, discussed in Sect. 5.3, the transition amplitudes
are now to be treated at least in N th order perturbation theory. In the most simple
case, two-photon ionization of an initial single electron n state to a final state of
energy by linearly polarized light, the generalized differential ionization cross
section according to C OOPER and Z ARE (1969) is

dn (, ) m|r0 | m m|r0 |n m 2
(2)
, (5.93)
d
Wn W +

using the transition amplitudes in the helicity basis introduced in (4.79) and (4.77).
One has to average over all initial substates n m, to sum over all intermediate states
m (in principle over all discrete bound states as well as over continuum states),
and over all final continuum states m. One may guess that the evaluation of this
expression may become a formidable enterprize. For circular polarization, or differ-
ent polarization in each of the two steps, the expression becomes even more com-
plicated. One sees that this gets even more intricate if one of the intermediate states
may be excited resonantly or nearly resonantly ( W Wn ). Then the typ-
ical damping terms must be included, and one has to account for possible saturation
effects: in the limit one just populates (non-isotropically) an intermediate state from
which ionization occurs in the second step.
As a rule, (5.81) will not describe the angular distribution of the emitted elec-
trons after MPI. Rather, the (generalized) differential cross section for N photon
ionization will involve several parameters:
(N ) (N )
dn (, ) n 1 + 2 cos2 + 4 cos4 + + 2N cos2N
= . (5.94)
d 4 1 + 2 /3 + 4 /5 + + 2N /(2N + 1)
Owing to the continuous progress with intense, tuneable and reliable laser systems
the interest in studying such processes experimentally has increased tremendously
over the paste decades. Multi-photon ionization, in particular via resonant inter-
mediate states (so called resonantly enhanced multi-photon ionization, REMPI) is
today a universally employed tool for studying the structure and dynamics of atoms,
molecules, clusters and even biomolecules.
16 Mutatis mutandis this also holds for single-photon ionization of non-isotropically populated ini-
tial states which may e.g. be prepared by additional photons in an optical pumping process with
polarized light (see Appendix D in Vol. 2 or H ERTEL and S TOLL 1978).
7p 2P3/2 8p 2P3/2 8d 2D5/2

electron signal
1+1 REMPI 1+1 REMPI 2+1 REMPI
0 90 180 0 90 180 0 90 180

polarization angle / o
Fig. 5.14 Resonant multi-photon ionization of Cs atoms according to C OMPTON et al. (1984).
Measured data (points) and computed (red lines) angular distributions of the photoelectron yield
for the processes listed in (5.95)
We have to confine the discussion here on two examples. C OMPTON et al. (1984)
have studied REMPI transitions in Cs atoms. The experimental setup is similar to
that shown in Fig. 5.11, but much more simple since the target atoms are gener-
ated by a simple atomic beam oven. However, electron detection and polarization
rotation are practically the same. The light source was a pulsed dye laser, tuneable
to a number of wavelengths, with 10 ns pulse duration and an intensity of moder-
ate 108 W cm2 . Cs atoms have been excited with one or two photons being in
resonance with the intermediate state depending on selection rules. One further
photon of the same wavelength served for ionization. Observed were the following
processes:
1

1
Cs(7p 2 P3/2 ) Cs+ + e ( s and p)
1

Cs 6s 2 S1/2 1 1
Cs(8p 2 P3/2 ) Cs+ + e ( s and p)
2
(5.95)

23 1
Cs(8d 2 D5/2 ) Cs+ + e ( s and p).
Figure 5.14 shows the measured and calculated angular distributions of photoelec-
trons. Comparison with Fig. 5.10 emphasizes the rich structure of the photoelectron
signal as a function of polarization angle for the present MPI processes. This is
what we expect according to (5.94). We shall return to this in the context of intense
laser fields, e.g. in Sect. 8.5.3.
In recent years so called electron imaging spectrometers (EIS) or velocity map
imaging (VMI) devices have been developed. With these very elegant and efficient
methods one collects in one shot (e.g. for one laser pulse) a complete angular and
energy distribution of the photoelectrons such images have, of course, to be av-
eraged over many events. Several variants of these techniques are used, all of them
based on the fact thatthe photoelectrons expand around their point of origin on
spheres of radius r = t 2Wkin /me . At a well defined time t one projects this spher-
ical shell onto an extended position sensitive detector and determines the kinetic
energy Wkin and by determining x/t and y/t also the emission angles and in
respect of the polarization vector.
(a) phosphorescent (b)
electron signal
screen
/ arb. un.
+1600 V MCP
6 cm
0
10
el
0V 00 0 ix
3 cm
-1 0 /p
-5 y
pp / pix0 x
0 00
50
-160 V el
10 -1
Fig. 5.15 Imaging photoelectron spectroscopy for H according to R EICHLE et al. (2001).
(a) Optical EIS system with a special field shape (red dotted the equipotential lines) and two
trajectories for equal momenta p leading to the same detector position. (b) Photoelectron signal
from H in an infrared laser field with I = 1.7 1013 W cm2 . The polarization direction of the
laser points into y-direction.
The angular distribution of the photoelectrons is projected onto the
xy plane, the distance x 2 + y 2 reflects the electron kinetic energy
Here we show in Fig. 5.15(a) a typical setup used for a nice experiment with
atomic hydrogen anions by the group of Hanspeter H ELM, one of the pioneers of
these methods. With a judiciously shaped electric field one optimizes the collec-
tion of electrons which are amplified by a multi-channel plate electron multiplier
(MCP), detected by a phosphorescent screen, and finally optically registered and
digitized by a CCD camera. Figure 5.15 shows a measured electron distribution for
the photo-detachment from H with a typical ATI structure: The distance from the
centre represents Wkin , and each concentric angular distribution corresponds to one
particular value of N according to (5.91) each with its own nontrivial angular dis-
tribution corresponding to (5.94). We shall come back to these powerful techniques
at several occasions, e.g. in Chap. 5 and Appendix B.4, both in Vol. 2.
Section summary
Photoionization processes may be described as transitions between bound and
continuum states, induced by the electromagnetic field. In dipole approxima-
tion the cross section (5.61) is simply proportional to the absolute square of
the respective transition matrix element. It may be expressed in terms of the
OOSD (5.62).
FBA may be used to obtain an estimate of the cross sections at photon energies
> 5 to 10 WI . At threshold ( = WI ) the photoionization cross section
is finite (in contrast to the prediction of FBA).
The angular distribution of the ejected electron is an important characteristic
of the photoionization process, essentially imaging the continuum wave func-
tions accessed. For ionization with linearly polarized light one observes in the
general case a distribution [1 + P2 (cos )], with being the anisotropy
parameter (1 2) and the angle between polarization and ejected
electron.
Multi-photon ionization is a more complex process and must be treated at

least in N th order perturbation theory (if N photons are absorbed), allowing
for ATI where the excess energy is carried by the ejected electron as given by
(5.91). The angular distributions of ATI electrons are correspondingly more
intricate since angular momenta up to N may be exchanged between the
electromagnetic field and the ejected electron.
ATI: Above-threshold ionization, in multi-photon ionization, if more photons are

absorbed than necessary for ionization.
CW: Continuous wave (as opposed to pulsed) light beam, laser radiation, etc.
EIS: Electron imaging sprectrometer, specifically for photoelectron spectroscopy.
EPR: Electron paramagnetic resonance, spectroscopy, also called electron spin
resonance ESR (see Sect. 9.5.2).
FBA: First order B ORN approximation, approximation describing continuum
wave functions by plane waves; used in collision theory and photoionization (see
Sect. 6.6, Vol. 2 and Sect. 5.5.2, respectively).
HITRAN: High-resolution transmission molecular absorption database, http://
www.cfa.harvard.edu/hitran (ROTHMAN et al. 2009).
HWHM: Half width at half maximum.
M2: Magnetic quadrupole, transitions induced by the interaction of a magnetic
quadrupole with the magnetic field of electromagnetic radiation.
MCHF: Multi configuration H ARTREE -F OCK, method to determine wave func-
tions for multi-electron systems (see Sect. 10.5.4).
MCP: Multi channel plate, electron multiplier with many amplifying elements.
MPI: Multi-photon ionization, ionization of atoms or molecules by simultaneous
absorption of several photons.
NMR: Nuclear magnetic resonance, spectroscopy, a rather universal spectro-
scopic method for identifying molecules (see Sect. 9.5.3).
References 271
OOSD: Optical oscillator strength density, characterizes the strength of photoion-

ization per energy interval (see Sect. 5.5.1).
REMPI: Resonantly enhanced multi-photon ionization, ionization of atoms or
molecules by several photons with one resonant intermediate state.
VMI: Velocity map imaging, experimental method for registration (and visual-
ization) of particle velocities as a function of their angular distribution (see Ap-
pendix B, Vol. 2).
21348 (2007).
References
B ETHE , H. A. and E. E. S ALPETER: 1957. Quantum Mechanis of One- and Two-Electron Atoms.
Berlin, Gttingen, Heidelberg: Springer Verlag, 369 pages.
B OUCHIAT , M. A.: 2007. Linear stark shift in dressed atoms as a signal to measure a nuclear
anapole moment with a cold-atom fountain or interferometer. Phys. Rev. Lett., 98, 043003.
B OUCHIAT , M. A. and C. B OUCHIAT: 1997. Parity violation in atoms. Rep. Prog. Phys., 60,
13511396.
B URGESS , A. and M. J. S EATON: 1960. A general formula for the calculation of atomic photo-
ionization cross sections. Mon. Not. R. Astron. Soc., 120, 121151.
C OMPTON , R. N., J. A. D. S TOCKDALE, C. D. C OOPER, X. TANG and P. L AMBROPOU -
LOS : 1984. Photoelectron angular-distributions from multi-photon ionization of cesium atoms.
Phys. Rev. A, 30, 17661774.
C OOPER , J. and R. N. Z ARE: 1968a. Angular distribution of photoelectrons. J. Chem. Phys., 48,
942943.
C OOPER , J. and R. N. Z ARE: 1968b. Correction. J. Chem. Phys., 49, 4252.
C OOPER , J. and R. N. Z ARE: 1969. Photoelectron angular distributions. In: S. G ELTMAN et al.,
eds., Lectures in Theoretical Physics, vol. XI-C, 317337. New York: Gordon and Breach.
C OOPER , J. W.: 1962. Photoionization from outer atomic subshells. A model study. Phys. Rev.,
128, 681693.
C OOPER , J. W.: 1988. Near-threshold K-shell absorption cross-section of argon relaxation and
correlation-effects. Phys. Rev. A, 38, 34173424.
C OVINGTON , A. M. et al.: 2007. Measurements of partial cross sections and photoelectron angu-
lar distributions for the photodetachment of Fe and Cu at visible photon wavelengths. Phys.
Rev. A, 75, 022711.
FANO , U. and J. W. C OOPER: 1968. Spectral distribution of atomic oscillator strengths. Rev.
Mod. Phys., 40, 441507.
G PPERT-M AYER , M.: 1931. ber Elementarakte mit zwei Quantensprngen. Ann. Phys.
(Berlin), 9, 273294.
H NSCH , T. W.: 2005. N OBEL lecture: passion for precision, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/2005/hansch-lecture.html.
H ERTEL , I. V. and W. S TOLL: 1978. Collision experiments with laser excited atoms in crossed
beams. In: Adv. Atom. Mol. Phys., vol. 13, 113228. New York: Academic Press.
L AMBROPOULUS , P.: 1985. Mechanisms for multiple ionization of atoms by strong pulsed
lasers. Phys. Rev. Lett., 55, 21412144.
L IPELES , M., R. N OVICK and N. T OLK: 1965. Direct detection of 2-photon emission from
metastable state of singly ionized helium. Phys. Rev. Lett., 15, 690693.
M ADDEN , R. P. and K. C ODLING: 1963. New autoionizing atomic energy levels in He, Ne, and
Ar. Phys. Rev. Lett., 10, 516518.
M ANSON , S. T. and A. F. S TARACE: 1982. Photo-electron angular-distributions energy-
dependence for s subshells. Rev. Mod. Phys., 54, 389405.
N OVICK , R.: 1972. 2-Photon decay of metastable hydrogenic atoms. Science, 177, 367.
PALENIUS , H. P., J. L. KOHL and W. H. PARKINSON: 1976. Absolute measurement of photoion-
ization cross-section of atomic-hydrogen with a shock-tube for extreme ultraviolet. Phys. Rev.
A, 13, 18051816.
P ETITE , G., P. AGOSTINI and H. G. M ULLER: 1988. Intensity dependence of non-perturbative
above-threshold ionization spectra experimental-study. J. Phys. B, At. Mol. Phys., 21, 4097
4105.
R EICHLE , R., H. H ELM and I. Y. K IYAN: 2001. Photodetachment of H in a strong infrared
laser field. Phys. Rev. Lett., 87, 243001.
ROTHMAN , L. S. et al.: 2009. The HITRAN 2008 molecular spectroscopic database. J. Quant.
Spectrosc. Radiat. Transf., 110, 533572.
S AHA , H. P.: 1989. Threshold behavior of the M-shell photoionization of argon. Phys. Rev. A,
39, 24562460.
S CHFFER , H. W., R. W. D UNFORD, E. P. K ANTER, S. C HENG, L. J. C URTIS, A. E. L IV-
INGSTON and P. H. M OKLER : 1999. Measurement of the two-photon spectral distribution
from decay of the 1s2s S-1(0) level in heliumlike nickel. Phys. Rev. A, 59, 245250.
S CHMIDT , V.: 1992. Photoionization of atoms using synchrotron radiation. Rep. Prog. Phys., 55,
14831659.
T ONG , D., S. M. FAROOQI, E. G. M. VAN K EMPEN, Z. PAVLOVIC, J. S TANOJEVIC, R. C OTE
E. E. E YLER and P. L. G OULD: 2009. Observation of electric quadrupole transitions to Ryd-
berg nd states of ultracold rubidium atoms. Phys. Rev. A, 79, 052509.
W IKIPEDIA CONTRIBUTORS: 2013. Atmospheric radiative transfer codes, Wikipedia, The Free
Encyclopedia. http://en.wikipedia.org/wiki/Atmospheric_radiative_transfer_codes, accessed:
8 Jan 2014.
Fine Structure and L AMB Shift
6
We now refine our treatment of atoms with an essential step: we

include the electron spin and its interactions. In this context we
also meet the L AMB shift, very important in modern physics as
it is one of the major testing grounds for QED. To study such
fine details of atomic spectra one needs sophisticated
spectroscopic methods into which we give a first introduction.
Overview
This chapter focuses on some key topics of atomic physics. It should be stud-
ied carefully. To start, Sect. 6.1 introduces some state-of-the-art experimental
methods used today in high precision spectroscopy. Section 6.2 introduces
into the phenomenology and terminology of spin-orbit interaction and an-
gular momentum coupling, while Sect. 6.3 may be seen as the quantitative
counterpart perhaps somewhat strenuous, but important. The focus is on
the H atom, described by the D IRAC theory and the only system for which
fine structure (FS) splittings can be computed exactly. Alkali atoms serve as
further examples. Section 6.4 is concerned with selection rules and intensi-
ties for FS transitions, a somewhat dry but necessary subject which the reader
may perhaps skip on first reading and come back later. The L AMB shift is
presented in Sect. 6.5 with the classical experiments as well as with some
results from modern high precision spectroscopy for the H atom and highly
charged ions. Understanding the physics should be supported by Sect. 6.5.6
with a very brief introduction to F EYNMAN diagrams and a glimpse on QED.
Section 6.6 concludes the chapter with a short excursion into the physics of
the electron magnetic moment anomaly.

274 6 Fine Structure and L AMB Shift
6.1 Methods of High Resolution Spectroscopy
6.1.1 Grating Spectrometers
For precise measurements of atomic and molecular spectra correspondingly precise

instruments and special detection schemes are needed. In absorption or emission a
first approach will be the use of high resolution spectrometers. We take the oppor-
tunity to briefly recall some pieces from undergraduate optics, which are important
but often forgotten. The most straight forward instruments used are monochroma-
tors (designed to generate quasi-monochromatic light from a broad band source)
or spectrometers (to identify the wavelengths of spectral lines). Typically, an opti-
cal setup of focal length f images the entrance slit of a spectrometer onto its exit
slit after passing the light through suitable dispersive devices, based on refrac-
tion (prism spectrometers, see e.g. B ORN and W OLF 2006, Chap. 4.7) or diffraction
(grating spectrometers). We review here only some facts about the latter which are
most commonly used in modern spectroscopy (for more details see e.g. B ORN and
W OLF 2006, Chap. 8.6.1). A variety of geometries may be applied.
A particularly clear and efficient setup, illustrated in Fig. 6.1, uses a reflect-
ing, spherical grating of radius R which acts both as focussing optics (focal length
f = R/2) and dispersive element. While the ruling of spherical gratings is some-
what demanding, the simplicity of the setup is a great advantage and the reflecting
geometry avoids losses in optical lenses and can be used also in the VUV and XUV.
Grating, as well as entrance and exit slit are mounted on a so called ROWLAND
circle which has the diameter R/2. This geometry is based on the inscribed angle
theorem according to which the inscribed angles (here and ) belonging to the
same intercepted arcs on a circle (here AF and FB, respectively) do not depend on
the position of their apex on the circle. Without diffraction we would have =
and the entrance slit at A would be imaged by the grating onto B (0th order diffrac-
tion). Due to diffraction and interference (z = 1, 2, . . . ) the exit angle depends on
wavelength as to be discussed in a moment and the beam is focused onto the
Fig. 6.1 ROWLAND circle

F focus of grating
entrance
A slit
exit
slit B
spherical grating
6.1 Methods of High Resolution Spectroscopy 275
(a) GN (b) (c)

GN GN
ray 1 ray 2 FN
sin FN
sin
Fig. 6.2 (a) Standard geometry for a reflecting grating. Schematically indicated is the optical
delay for two light rays on the way in (fat, red line) and out (fat black line). Angles of incidence
() and diffraction () in respect of the grating normal (GN) are taken as positive to the left,
negative to the right of GN. (b) Blazed grating with a blazing angle B between the facette normal
(FN) and GN. (c) Echelle grating in a near L ITTROW arrangement; shown is an R3 echelle with a
blaze angle of 71.5
exit slit B. Typically, the grating is moved along the circle to vary the wavelengths
which reaches B.
The actual diffraction and interference geometry of a reflective grating is shown
in Fig. 6.2(a). We see that the optical delay between rays 2 and 1 from neighbour-
ing grooves of the grating at a distance d is given by s = d(sin + sin ). Note
that < 0 in the arrangement of Fig. 6.2(a). We expect constructive interference in
the outgoing wave front if the grating equation
c
p = z , or in frequency space = z (6.1)
d s
with p = (sin + sin ), s = pd and z = 0, 1, 2, . . .
holds, with z being the order of the interference and pd the optical delay. An im-
portant characteristic of any spectrometer is the change of the deflection angle with
wavelength. One derives the change of the interference maximum with wavelength
(at a fixed angle of incidence ) from (6.1), and finds for a grating the
d 1 z
angular dispersion = . (6.2)
d cos d
A detailed, rather straight forward calculation (B ORN and W OLF 2006,

Chap. 8.6.1) shows that the (normalized) total intensity reflected from a grating
with N coherently1 illuminated grooves (slits, facets) is

1 sin(N pkd/2) 2 2
T (p) = 2 T0 (p) with k = . (6.3)
N sin(pkd/2)
1 The concept of lateral coherence will be discussed in some detail in Chap. 2, Vol. 2. Essentially,
coherent illumination requires that s sin d holds for the width s of the spectrometer entrance
slit, where d is the full opening angle of the optics illuminating the grating.
=0 =1 =2
1
(a)
0 /d /d
2
1 (b)
Fig. 6.3 (a) Normalized interference function T (p)/T0 (p) of a diffraction grating with N coher-
ently illuminated grooves in 0th, 1st and 2nd order for two wavelengths differing by ; (b) both
traces added, illustrating that the two wavelengths are (just) distinguishable when the principle
maximum of one overlaps with the first side minimum of the other (here in 2nd order)
T0 (p) represents an overall normalized intensity function for a single groove of the
grating.2 It depends on the shape and reflectivity of an individual groove and varies
smoothly.
Figure 6.3(a) shows the interference function T (p)/T0 (p) due to multi-ray in-
terference for two, slightly different wavelengths and + , with maxima
at p = z/d for z = 0, 1 and 2 according to (6.1). According to the so called
R AYLEIGH criterium, illustrated in Fig. 6.3(b), the two wavelengths can just be
distinguished when the principle maximum of the latter coincides with the first side
minimum in the fringe pattern of the former (i.e. from N coherently interfering
rays). With (6.1) and (6.3) the principle maxima are positioned at
N pkd/2 = N pd/ = N z
while the optical delay (p + p)d for the first side minima corresponds to

N (p + p)d/ = (N z + 1) p = .
Nd
On the other hand, the principle maxima for a wavelength + correspond to
+
p + p = z p = z .
d d
By equating the two expressions for p we obtain the
resolving power / = N |z|, (6.4)
2 Inthe most simple case, a transmission grating consisting of slits with a width s, this would be
the single slit transmission function
T0 (p) = I0 sinc2 (ksp/2), with sinc(x) = sin(x)/x.
which holds for all interference based spectrometers. For grating spectrometers it
determines the spectral resolution usually given as difference of the wave-
length between two neighbouring spectral lines which can just be distinguished.
Modern gratings are fabricated by holographic methods or are ruled and may have
several thousands of grooves per mm and a length of up to a few 100 mm, thus
N may in principle be up to 100 000 or more, albeit difficult to illuminate coher-
ently. Typically they operate in 1st or 2nd order. Special shape and reflectivity of
the surface may help to channel all diffracted intensity into one interference order,
thus avoiding ambiguities in the observed spectra. Specifically, blazed gratings have
grooves with facets that are aligned such that the maximum diffraction angle in re-
spect of the facet normal (FN) is equal to the entrance angle in respect of FN as
indicated in Fig. 6.2(b). One reads from the figure that this blaze angle is
B = ( + )/2 (6.5)
where + is derived from (6.1) for a specific wavelength. The diffracted signal
then appears as if directly reflected from the facets. Even though such geometric
interpretation is somewhat questionable (typically the wavelengths are on the same
order of magnitude as the groove distance) detailed calculations show that one may
indeed maximize in this manner the light reflected for a specific order of interference
and wavelength B one says the grating is blazed for that wavelength.
At last we mention so called echelle gratings (from the French word chelle,
stairs or ladder) which have experienced a true renaissance in recent years. In con-
trast to standard gratings they have only a moderate number of grooves N , but very
large z 1 and large blaze angles. A typical setup is shown in Fig. 6.2(c). They
may be viewed as an intermediate between grating spectrometer and interferometer.
One fabricates them today by ruling or chemical etching. Characteristic for echelle
spectrometers is the combination of high dispersion according to (6.2) and high re-
solving power according to (6.4). A disadvantage is that due to the high value of z
the spectra may overlap in different orders. In frequency space we read from (6.1)
that frequencies and + FSR have a maximum of order z and z +1, respectively,
at the same optical path difference s if
c
FSR = = . (6.6)
s z
Thus, frequencies differing by more than FSR cannot be assigned unambiguously.
Obviously, the higher the order of interference z for which the system is designed,
the more serious is the problem. One may overcome it by crossing the echelle with
a suitable second dispersive element (grating or prism) which leads to an unambigu-
ous 2D pattern, ideally suited for state-of-the-art CCD devices for detection.
Finally, one should note that in practice the degree to which the theoretical re-
solving power is attained (by any grating spectrometer or monochromator) depends
not only on the angles and (and on z and N ) but also on the optical quality of
the grating surface, the uniformity of the groove spacing, the quality of the associ-
ated optics in the system, and (most importantly) on the width of the (entrance and
Fig. 6.4 M ICHELSON reflecting

interferometer mirror 2
s1
collimating s=s2 - s1
lens
s2
light
source beam
splitter reflecting
mirror 1
focussing lens
detector
exit) slits (or detector elements). Any departure of the diffracted wavefront greater
than /10 from a plane (for a plane grating) or from a sphere (for a spherical grat-
ing) will result in a loss of resolving power due to aberrations at the image plane.
The grating groove spacing must be kept constant to within about 1 % of the wave-
length at which theoretical performance is desired. (PALMER and L OEWEN 2005,
Chap. 2.4)
6.1.2 Interferometers
Expressions analogue to (6.4) also hold for all interference based spectrometric de-
vices, such as M ICHELSON, M ACH -Z EHNDER, FABRY-P ROT (FPI) etc. interfer-
ometers. In an genuine interferometer N may vary from 2 beams in a M ICHELSON
setup to several 100 in an FPI. On the other hand, the order of interference z is
very large in all interferometers: For an optical path difference s of the interfering
beams it is
z = s/ = s/c. (6.7)
In the visible (VIS) a moderately large optical path difference of s = 6 cm corre-
sponds already to z = 105 . One may thus in principle obtain very high resolution
with interferometers. Typically one observes fringes of equal inclination (slightly
different divergence angles lead to a modification of the optical path difference s).
Since in various contexts M ICHELSON interferometers will play a role later on
(e.g. in F OURIER transform spectrometers, see Sect. 5.3.2 in Vol. 2), we show such
a setup in Fig. 6.4 for reference without entering here into details (N OBEL pize for
M ICHELSON 1907). The key elements are a beam splitter and two reflecting mirrors
between which the light travels an optical distance s1 and s2 , respectively. The op-
tical path difference s = s2 s1 may be varied by moving one of the mirrors with
high precision. Unfortunately, spectra obtained with interferometers are not unam-
biguous (just as gratings used in higher order diffraction): assume that the central
Fig. 6.5 FABRY-P ROT highly reflecting
(fringes of equal inclination)

interferometer plane surfaces lens
interference pattern
r',t' r,t
3
C . 2
B P
2
L
A 1
possibly n >1
maximum for the wavelengths 1 (frequency 1 ) corresponds to order z. A some-

what shorter wavelength 2 will be observed at the same optical path difference s
in z + 1 order if
z = 1 s/c and z + 1 = 2 s/c = 1 = (2 1 ) s/c.
Completely equivalent to (6.6), we obtain for the frequency difference 2 1 of

two neighbouring maxima the so called
free spectral range FSR = c/ s. (6.8)
Hence, two lines can be distinguished in an interferogram only if their frequency

difference is < FSR . Since the resolution of a M ICHELSON interferometer
(N = 2) is on the same order of magnitude as its free spectral range it cannot di-
rectly be used as a spectrometer. Quite generally, for spectroscopic applications of
interferometers some type of additional selection has to be employed.
FABRY-P ROT interferometers (FPI) are used for a number of purposes. Not
least, each laser resonator is essentially an FPI as will be discussed in Sect. 1.1.2
in Vol. 2. But it is also an indispensable tool for high resolution spectroscopy. In
Fig. 6.5 a typical setup is shown schematically with the characteristic elements. It
consists of two plane parallel surfaces mounted in a distance of L with an amplitude
reflectivity r and r , respectively. For the light intensity this corresponds to reflection
coefficients R = |r|2 and R = |r |2 , respectively, and to transmission coefficients
T = |t|2 and T = |t |2 . If the losses are negligible R + T = 1. The space between
the mirrors may be empty or filled by a medium with an index of refraction n (this
may e.g. be realized by one plane, coated, parallel glass plate). Light incident at
an angle in respect of the surface normal is reflected many times at these highly
reflecting surfaces and finally brought to interfere on the detector P .
The optical path difference between neighbouring beams is given by the distance
ABC in Fig. 6.5 (red dotted line) as s = 2Ln cos (note that it decreases with the
tilt angle !). For perpendicular incidence the free spectral range (6.8) is thus
c
FSR = , (6.9)
2nL
or on the wavelengths scale

2
FSR . (6.10)
2nL
The order of interference becomes
z = 2nL/. (6.11)
and the phase difference between neighbouring, interfering beams is:

2 s cos
= = 2kLn cos = 2 . (6.12)
FSR
The full field amplitude transmitted through the FPI is found from adding all partial
amplitudes
2
Et = tt E0 + rr exp(i)tt E0 + rr exp(2i)tt E0 + , (6.13)
which is a geometrical series that can be written in closed form. Neglecting losses,
remembering that intensities are I |E|2 , and evaluating the geometrical series
emerging from (6.13) one obtains the
It 1
A IRY function = (6.14)
I0 1 + F sin2 ( 2 )
for the ratio of transmitted to incident intensity I0 as a function of frequency (using

(6.12)). Here F is the so called finesse coefficient

4 RR
F= . (6.15)
(1 RR )2
In the resonant case = (z + m)2 the full initial intensity is transmitted. Figure 6.6
shows some examples for the transmission through an FPI at different reflectivi-
ties, where R = R is assumed. The FWHM h of the transmission lines of the FPI
follows from (6.14) with It /I0 = 0.5 F sin2 (h /4) = 1 h 4/ F . Finally,
with (6.12) the FWHM ofthe frequency h follows: at perpendicular incidence
2 h / FSR = h = 4/ F holds and one defines for an FPI

FSR R
the finesse F = = F= (6.16)
h 2 1R
FSR
or on the wavelengths scale F =
h
With (6.10) the resolving power of an FPI is

R 2L
= = F z. (6.17)
h (1 R)
Fig. 6.6 Transmission of a I/Imax LORENTZ, F = 0.7

FABRY-P ROT interferometer 1.0
as a function of the phase
difference according to
(6.12) for different finesse F F = 0.7
(or reflection coefficients (R = 4.5%)
R = R ); R = 4.5 % is the h
reflectivity of uncoated glass 0.5
F=6
surfaces). The phase change (R = 60%)
is given in respect of one F = 30
(R = 90%)
transmission maximum;
= 2 corresponds to
= FSR
0 2 3 4

Hence, the finesse F of an FPI corresponds to the (effective) number N of inter-

fering beams in the general formula (6.4). In practice, F is limited by mechanical
imperfections of the mirrors. As a rule of thumb the surfaces must be ground to
at least /F precision over the whole surface. Typically, F may be up to a few
100. We also note, that the reflected intensity Ir is complementary to (6.14), so that
Ir () = I0 It (). While in transmission (e.g. for slightly divergent light) one ob-
serves bright rings on a dark background, in reflection one finds dark rings on a
bright background. Very important is the fact that in the interior of the resonator
the radiation intensity is significantly enhanced: since only a fraction T = 1 R of
the internal intensity leaves the resonator, within the resonator the intensity is given
by Iint = It /(1 R). In this sense, a FABRY-P ROT resonator acts as a storage
for light. This is exploited in the construction of lasers but also in spectroscopy: in
the interior of a FABRY-P ROT resonator one may conduct highly sensitive spec-
troscopy using relatively low intensity light sources. This is understood simply by
considering the light beam to pass a sample inside the resonator many times. Dur-
ing each passage the light beam is correspondingly attenuated. So called cavity ring
down spectrometers (CRD) use this cleverly by feeding a laser pulse into an FPI res-
onator and following its decay (see Sect. 5.5.3 in Vol. 2). A F OURIER transform of
this decay function allows a very sensitive determination of the samples absorption
spectrum.
In practice for lasers and spectroscopy one uses not only plane parallel mirrors.
Very often concave, highly reflecting surfaces are employed which lead to stabil-
ity (see Sect. 1.1.3 in Vol. 2) and an additional concentration of the available light
inside the resonator. A particularly useful setup consists of two concave mirrors
of radius R positioned in a distance R. A (nearly) plane wave (e.g. a laser beam)
entering such a device is focused in the centre of the resonator and may be used ef-
ficiently for studying nonlinear processes such as multi-photon excitation of atoms
and molecules: according to (5.43) such signals rise with some power of the inten-
sity while the detected volume only decreases linearly with the cross section in the
focus of the beam.
(a) collimation fluorescence detector term and (b)

aperture(s) excitation
skimmer scheme
for the Na D2 line
oven
inclined
3P 3/2
incidence
detector beam excitation emission
splitter
FABRY-PEROT CW dye laser beam 3S1/2
as marker tunable, perpendicular
Fig. 6.7 (a) Schematic of an atomic or molecular beam setup for high resolution spectroscopy.
(b) Partial term scheme of Na, with the two Na D2 HFS lines recorded
6.1.3 D OPPLER Free Spectroscopy in Atomic Beams
The precise observation of finer effects in atomic spectra is not only limited by the
resolving power of the spectrometers but very often also by the D OPPLER broad-
ening (5.22). By studying spectra in atomic (or molecular) beams this handicap
may be overcome. A typical experimental setup and detection scheme is sketched
in Fig. 6.7: The atomic beam emerges from an oven (preferably at high pressures
to allow for adiabatic cooling) and is collimated by a skimmer and some apertures
to a divergence angle . The velocity distribution has its maximum at vm and
a FWHM = v. Let the transition wavelength at resonance be ba and the angle
of light incidence . With (1.32) the resulting D OPPLER shift ba and D OPPLER
broadening D is then
vm vm v
ba = ba cos = cos and D = cos , (6.18)
c ba ba
respectively. When exciting (or detecting) perpendicular to the beam
cos cos(/2 ) = sin ,

0, and only the perpendicular ve-
hence the average shift vanishes, ba ba
locity components contribute to a small D OPPLER broadening:
v
D = 2 sin . (6.19)
ba
For illustration we show in Fig. 6.8 the hyperfine structure (HFS) of the NaD2
line, recorded with the atomic beam setup Fig. 6.7(a). The term scheme is shown in
Fig. 6.7(b); a detailed description of HFS will be given in Chap. 9. To demonstrate
the dependence of the D OPPLER profiles on the angle of light incidence in respect
of the atomic beam, two spectra for different angles of light incidence ( = 90 and
40 ) are recorded simultaneously. The atomic beam is slightly supersonic with
vm 1400 m / s and v 700 m / s. It is collimated to 2.5 . The exciting,
narrow band light (ba 589 nm) comes from a fine tuneable dye-laser.
inclined irradiation perpendicular irradiation

1800 1700
MHz MHz
D
900
MHz
4

marker
100 MHz
Fig. 6.8 Hyperfine splitting of the Na D2 line (32 P3/2 3 2 S1/2 transition) recorded with the
atomic beam apparatus sketched in Fig. 6.7(a). One clearly sees the two hyperfine components with
perpendicular (black) and with inclined (red) irradiation as function of the exciting frequency .
The full red line gives 4 the signal extracted from the combined signal (red dashed). The marker
shown at the bottom was simultaneously recorded with the FPI
One observes the whole fluorescence signal as a function of the incident fre-
quency and thus registers simultaneously the excitation spectrum for perpen-
dicular and inclined incidence of the light beam. Both contributions can easily be
separated as indicated in Fig. 6.8. While tuning the laser, a reference signal is simul-
taneously recorded from a small fraction of the laser beam which passes through an
FPI. This signal appears and vanishes in intervals corresponding to the free spectral
range of the FPI (here 100 MHz). It generates a convenient marker for calibrat-
ing the recorded spectrum. For perpendicular irradiation, the two HFS components
of the Na D2 line are clearly separated by ca. 1700 MHz. The observed linewidth
D 100 MHz corresponds to that predicted by (6.19). With some effort it can
be further reduced.
Inclined incidence (red line Fig. 6.8) leads to a D OPPLER shift ba ba
1800 MHz and a broadening of D 900 MHz much larger than for perpendic-
ular incidence again in full agreement with the predictions of (6.18) for the given
atomic beam conditions.
6.1.4 Collinear Laser Spectroscopy in Ion Beams
Laser spectroscopy on ions is often performed with energetic ion beams (the follow-
ing also holds for high energy neutral particle beams). In such experiments the laser
beam travels parallel or antiparallel to the fast ion or neutral beam. The method
has two advantages: On the one hand, one may tune the absorption frequency of the
ions by varying the energy of the beam and keeping the laser frequency fixed L at
some convenient laser line. On the other hand, D OPPLER broadening can be reduced
considerably (so called D OPPLER narrowing). The D OPPLER shift of the absorption
frequency ba as seen from the laser beam is according to (1.30) and (1.34)

ba = ba wk + 2wk + wk2 L 2wk (1 wk /2 + ), (6.20)
for ions with a rest mass m and a kinetic energy Wkin , with wk = Wkin /(mc2 ). The
+ and signs refer to irradiation head-on or parallel to the ion beam, respectively.
This shift may be considerable for ions and allows for calibrated scanning through
the finer structures of spectra by tuning the acceleration voltage U , with Wkin = qeU
for ions of charge qe.
Typically the ions are prepared in an ion source with a finite energy width Wk at
low kinetic energies. According to (6.20) this would lead to a D OPPLER broadening

2Wk
D ba . (6.21)
mc2
During acceleration of the ion beam Wk usually stays constant, and so does wk =
Wk /(mc2 ). Hence, for higher but not too high kinetic energies (Wk Wkin
mc2 ) the D OPPLER width is derived from (6.20):
d ba wk ba Wk
D wk ba (1 + 2wk + ) . (6.22)
dwk 2wk 2Wkin mc2

Comparison with (6.21) shows that the linewidth is reduced by a factor 4Wkin /Wk
in respect of a measurement in the ion source due to D OPPLER narrowing. Typ-
ical for ion beam spectroscopy would be a 100 keV beam and an initial width
Wk = 1 eV in the source (much above thermal energy but characteristic for ion
sources), amounting to a reduction factor of 630.
For illustration, in the case of He+ ions the thermal D OPPLER width at room
temperature according to (5.21) is D (293 K) = ba 6.1 106 . The linewidth in
the 100 keV ion beam (Wk = 1 eV) would be D = ba 3.66 108 a factor
of about 170 less than even the thermal D OPPLER width! This already allows a
convenient resolution (even if not a high precision measurement) of the L AMB shift
for the n = 2 states in He+ which is around 1.423 106 ba . The D OPPLER shift
according to (6.20) in this case is 0.007ba , which allows for sufficient tunability
(in the per mill range) by changing the acceleration voltage. The method has been
used (and is still used) extensively for the determination of HFS and FS structure as
well as of the L AMB shift for heavy ions, specifically so for their unstable isotopes.
6.1.5 Hole Burning
The basis of many modern methods of D OPPLER free spectroscopy is the so called
hole burning by which a certain part of a population of the sample atoms or
molecules is depleted by monochromatic excitation. This allows a nice illustration
of the terms homogeneous and inhomogeneous linewidth.
As we have seen in Sect. 5.1.4, in gas phase the D OPPLER width is typically much
larger than the natural linewidth: D nat . To record the absorption probability
as a function of the frequency for an ensemble of atoms or molecules we may excite
with a (quasi-monochromatic) narrow band tuneable laser in the neighbourhood of a
resonance frequency ba . As discussed in Sect. 5.1.4 and Sect. 6.1.3 the absorption
(a) (b)
BOLTZMAN N BOLTZMANN
with hole
LORENTZ
0 (velocity || of light) 0
Fig. 6.9 Velocity distribution of atoms or molecules in the gas phase parallel to the wave vector
k of the light. (a) M AXWELL -B OLTZMANN distribution prior to absorption and L ORENTZ profile
for a particular velocity class vtest (dashed). (b) Velocity distribution with hole, which has been
created by the radiation absorbed at vtest
frequency ba of an atom (in the thermally broadened distribution) moving with

a velocity v is shifted by the D OPPLER effect to given by (5.19). When tuning
the laser one observes an absorption signal which is proportional to the velocity
distribution w(vx ) according to (5.20). This is shown in Fig. 6.9(a) by the full red
line. We may turn the argument around and say that a given frequency is absorbed
by atoms or molecules belonging to a velocity class

vx () = v cos = c 1 = ba ( ba ) (6.23)
ba
where vx is the projection of the atomic velocity onto the wave vector k of the
light, with k x. More precisely, atoms with velocities v slightly smaller or larger
than vx () also absorb light of the frequency with a probability given by the
Lorentzian natural line profile on the velocity scale. This is illustrated in Fig. 6.9(a)
by the red dashed line.
Hence, when irradiating with a fixed test frequency test one only excites atoms
of a particular velocity class around vtest = vx (test ) of the initial velocity distribu-
tion. If the light intensity is sufficiently high this results in a noticeable depletion
of this velocity class in the initial state D OPPLER profile as shown in Fig. 6.9(b):
the remaining velocity distribution of atoms in the initial state has then typically a
hole at vtest . The width of the absorbing velocity class corresponds to the natural
linewidth nat = ba nat : all atoms within the homogeneous linewidth partici-
pate in the depletion correspondingly. The same argument hold for collision broad-
ened profiles of a width (5.17) if that is much larger than the natural linewidth.
6.1.6 D OPPLER Free Saturation Spectroscopy
One may exploit hole burning very elegantly for spectroscopy. This has first been
realized by H NSCH et al. (1971) who coined the term D OPPLER free saturation
spectroscopy. A typical experimental setup is shown in Fig. 6.10 and the principle
las beam
er splitter
phase sensitive
detection
differential
beam chopper amplifier
splitter
mirror
mirror
probe beam pumpbeam
photo
sample vapour (gas) in cell reference beam detectors
e.g. Na or H(2s,2p)
Fig. 6.10 Experimental setup for D OPPLER free saturation spectroscopy according to H NSCH
et al. (1971)
probe beam pump beam
(a)
(b)
(c)
Fig. 6.11 Principle of D OPPLER free saturation spectroscopy: circles with arrows indicate atoms
and their velocity vectors, full colours indicate resonance, grey out of resonance. The incident
frequency is (a) blue shifted in respect of ba , (b) exactly ba or (c) red shifted. Consequently
(a) the probe beam excites atoms moving to the left, the pump beam only those moving to the
right; (c) the probe beam excites atoms moving to the right, the pump beam only those moving to
the left; (b) pump and probe beam excite only (the same) atoms at rest
of the method is illustrated in Fig. 6.11. One splits the laser beam into an intense
pump beam and a weaker probe beam, and sends them counter-propagating through
the cell with the sample. The pump beam burns holes into the D OPPLER profile of
the target atoms in the ground state. Ideally the transition is then saturated for that
frequency, i.e. nearly one half of the atoms in that particular velocity class are found
in the excited state. All atoms not in resonance with the pump beam frequency are
unaffected. Hence, in most situations the probe beam does not notice anything of this
preparation since it comes from the opposite side and is in resonance with a different
velocity class. Only if both, pump and probe beam match exactly a resonance transi-
tion for atoms at rest the probe beam sees the same atoms as the pump beam. This
leads to a reduction of the absorption exactly when the laser frequency corresponds
to the atomic resonance frequency for atoms at rest. One calls this reduction L AMB
dip. The signal may be recorded in comparison to a reference signal without pump-
crossover
equivalent experiment:
DOPPLER profile saturation signal
probed in Na bulb
1700
MHz
100 MHz marker
Fig. 6.12 D OPPLER free saturation spectrum of the two main hyperfine lines in the
3 2 P3/2 2 S1/2 transition of sodium (Na). The spectrum has been recorded by a most unso-
phisticated setup with Na in a bulb to illustrate the robustness of the method. One recognizes
the two characteristic L AMB dips showing the HFS splitting of 1700 MHz and one (physically
insignificant) crossover signal in between
Fig. 6.13 Fine structure of 0 500 000 / GHz

the H BALMER (n = 2 3)
transition. (a) BALMER series BALMER series (a)
(n = 2 n ), overview.
(b) Blow up of the
(n = 2 3) transition: sticks DOPPLER 50 K
spectrum (theory) according broadened (b)
to K RAMIDA (2010);
300 K
D OPPLER broadened spectra
simulated for 300 K (dashed)
and 50 K (full line). (c) shift
D OPPLER free saturation DOPPLER free
saturation crossover
spectrum according to
spectrum (c)
H NSCH et al. (1974). More
details are explained in
Sect. 6.5.1
-5 0 5 10 / GHz
ing and typically is registered by phase sensitive detection which synchronizes the
detected signal with a chopper for the laser beam to minimize statistical fluctuations
(using a so called lock-in amplifier).
If several allowed transitions are accessible over the tuning range of the laser,
correspondingly several L AMB dips appear. This is illustrated in Fig. 6.12 for the
hyperfine structure of the Na atom which we have already encountered in Sect. 6.1.3,
Fig. 6.8. In contrast to the latter experiments, the setup for recording the saturation
spectrum Fig. 6.12 was particularly simple: just a bulb with Na vapour, no opti-
cal surfaces, and rather weak pump and probe laser beams. Consequently the two
L AMP dips in the Na D2 line (3 2 P3/2 3 2 S1/2 ) separated by ca. 1700 MHz HFS
splitting are small, but clearly visible on the D OPPLER profile, and illustrate the
robustness of the method. Unfortunately, the method is not completely unambigu-
ous: so called crossover signals may arise from two closely spaced lines exactly
halfway between them.
In any case, this D OPPLER free saturation spectroscopy is an excellent method
to determine resonance lines with an accuracy on the order of the natural linewidth
irrespective of being taken in a gas cell with much larger D OPPLER width. A par-
ticular nice and important example (H NSCH et al. 1972, 1974) is illustrated in
Fig. 6.13. In this pioneering work the fine structure of the H line has been deter-
mined for the first time with high accuracy. To emphasize the resolution Fig. 6.13(a)
gives an overview for the BALMER series of atomic hydrogen. For comparison
(b) shows two D OPPLER broadened spectra in the frequency range of interest, sim-
ulated for 300 K and 50 K, respectively, which obviously do not allow to resolve
the fine structure in detail. In contrast, saturation spectroscopy (c) gives a very clear
picture of the fully resolved fine structure. We shall come back to it later.
6.1.7 R AMSEY Fringes
Even if one may overcome the D OPPLER broadening and other pitfalls of spectro-
scopic observations by clever methods, one often encounters the situation that the
resolution is limited simply because not enough time is available to perform a mea-
surement since, alas, the fundamental conditions imposed by the uncertainty relation
(1.125) between time and energy provide apparently stringent boundaries:
W t .
If no other effect interferes, the accuracy W by which an atomic transition may be

determined increases with the interaction time between radiation field and sample
(in Sect. 4.3.5 we had simply assumed t ). Norman R AMSEY (N OBEL prize
1989) has in the context of magnetic resonance spectroscopy (1950) which will
be detailed in Sect. 9.5.1 invented and explored a particularly ingenious method
which allows one in principle to prolong the interaction time effectively and thus to
enhance the resolution.
The idea is amazingly simple: in practice it will never be possible to expose the
object of investigation to the radiation for an arbitrary long time. A finite interac-
tion time, on the other hand, leads to a line profile with finite width as indicated in
Fig. 4.11. However, if instead of one wave-train one uses two which are coherently
phase coupled, this problem can be overcome. The result can easily be understood
extending our earlier discussion using time dependent perturbation theory. If nec-
essary, the matrix elements Dba of the perturbation operator have to be replaced
depending on the type of interaction, but otherwise we start with (4.51). Integration
has now to be carried out over the two wave-trains, i.e. from 0 to and in addi-
tion from T to T + . The system thus interacts with two coherent alternating fields
of a duration each. One easily verifies that the excitation probability (4.59) (see
also Fig. 4.11) which vanishes at = 1 is now modulated by a rapidly oscillat-
ing amplitude of the frequency /T . These so called R AMSEY fringes allow for a
correspondingly sharper determination of the resonance frequency.
Fig. 6.14 R AMSEY fringes: 1.22/ T

Line profile as calculated by wab
R AMSEY (1950) for the
excitation with two spatially 0.6
separated oscillating fields.
The resolution is determined
now by the time T between = 0.64/ T
0.4
the two excitation processes
(full red line, resonant, fast
oscillations) while the overall =1.4/
linewidth corresponds to the 0.2
much shorter interaction time
with only one of these (0 - )T (resonant)
fields (black dashed line, not -1 1 3
0.0
resonant)
-1.0 - 0.5 0 0.5 1.0
(0 - ) (non resonant)
As a consequence one obtains an overall line profile as shown in Fig. 6.14. The
practical realization for transitions in the radio frequency and microwave region
caused no problem even in 1950 since there the phase coherence is a built in prop-
erty of these waves. With modern laser techniques the required fixed phase relation
between two different laser pulses is no problem either. We shall get to know a
few practical, state-of-the-art realizations in Fig. 6.16 and Fig. 6.28 as well as in
Sect. 9.5.1.
6.1.8 D OPPLER Free Two-Photon Spectroscopy
Another very elegant way to realize D OPPLER free spectroscopy is based on two-
photon excitation (introduced in Sect. 5.3.1).
The method has been applied experimentally for the first time in 1974 nearly
simultaneously by several groups. For a detailed review the interested reader is re-
ferred to G RYNBERG and C AGNAC (1977). One uses a standing wave, e.g. a cir-
cularly polarized wave field, created by a FABRY-P ROT interferometer advan-
tageously in a confocal resonator with spherical mirrors so that focussing the laser
beam provides sufficient intensity for efficient excitation of two-photon processes.
The gas cell with the sample atoms or molecules is positioned into the focus of the
mirrors possibly the mirrors may even act as walls of the gas cell.
The basic idea of the methods is the following. If v is the velocity of an atom or
molecule in the gas cell and k the wave vector of the exciting light (in a FABRY-
P ROT resonator defined by the beam axis), then according to (1.33) the D OPPLER
shift in 1st is k v (angular frequency). For the inverse direction of light propagation
(k k) the sign of the D OPPLER shift changes for this particular atom. Let us
assume that a two-photon process is possible in the target atom between levels of
energy Wa and Wb . An atom in a resonator with the two counter-propagating waves
thus interacts simultaneously with two waves of angular frequency k v and

+ k v, respectively. Hence, the resonance condition for two-photon absorption is
Wba = Wb Wa = ( k v) + ( + k v) = 2. (6.24)
This is a remarkable finding: Independent of the individual velocity of an atom (or

molecule) in the resonator all sample particles may be excited by absorbing two
photons of the right energy = (Wb Wa )/2. If both beams have the same polar-
ization there is also a D OPPLER broadened background: each atom may also absorb
two photons from one of the waves travelling in either direction (backwards and for-
wards) so that the terms k v do not cancel. This signal is, however, much weaker
and broadened so that it can be separated easily from the D OPPLER free signal.
Often, by clever choice of the polarization, thus exploiting the M selection rules
for two-photon transitions, the background may even be suppressed completely: for
example, in a two-photon ns n s transition M = q1 + q2 = 0 must hold with
q1 and q2 being the components of the angular momentum of photon 1 and 2, re-
spectively. If one uses RHC polarized light (q1 = 1) travelling in one direction, in
the other direction, however, LHC polarized light (q2 = 1), then such a transition is
possible. But an ns n s transition may not be induced by two photons from one
beam since in that case we should have M = 2.
These possibilities to record D OPPLER free spectra of atomic or molecular tran-
sitions in a simple gas cell continue to be of great practical value in todays spec-
troscopy and are used increasingly for high precision measurements and more com-
plex system. Initially, experiments have often been conducted for alkali atoms. Fig-
ure 6.15 illustrates as an example a high resolution measurements of the hyperfine
structure for the transition 3s 2 S1/2 5s 2 S1/2 in the Na atom for which (a) shows
the relevant parts of the term scheme. Figure 6.15(b) presents the results of a two-
photon absorption experiment. For understanding the method we do not need to
discuss details of HFS which will be subject to Chap. 9. We only note that F char-
acterizes the total angular momentum of the system including electron and nuclear
spin. For a two-photon process the total angular momentum may change by 0 or 2 so
that in the present case only the transitions F = 2 F = 2 and F = 1 F = 1
may be accessed in a two-photon process. The absorption process is detected by
registering the spontaneously emitted fluorescence 5s 2 S1/2 3p 2 P1/2 and 2 P3/2
at 615 nm and 616 nm which is emitted after the absorption process. A tuneable,
pulsed dye laser has been used as light source. Similar to the experiment discussed
in Sect. 6.1.3, for precise determination of the frequency differences a small part
of the laser light is recorded simultaneously after passing through FABRY-P ROT
reference interferometer. This produces calibration marks (c) at a distance of its free
spectral range (here 300 MHz). One recognizes that with this relatively simple setup
one can already resolve structures with an accuracy of a few MHz.
Of fundamental importance is the precise determination of the energetic distance
between ground state 1s 2 S1/2 and first excited s state 2s 2 S1/2 in atomic hydro-
gen. As already discussed in Sect. 5.3.2 the 2s 2 S1/2 state decays with a rate of
8.228Z 6 s1 , which for the H atom amounts to a frequency linewidth of only 1.3 Hz.
150 MHz
5 S1/2 (a)
(b)
4S1/2 4P 4P3/2
61 1/2
6n
laser wavelength
61
602.23 nm
5n
330 n cence 3P3/2
m
3P1/2 0 300 MHz dye laser frequency
re s
m
m
m
m
589 n
fluo
590 n
330 n
(c)
1772 MHz
3S1/2
0 600 MHz atomic transition
frequency
Fig. 6.15 Two-photon transition between HFS states of Na, exemplified for the 3s 2 S1/2
5s 2 S1/2 transition, according to G RYNBERG and C AGNAC (1977). (a) Part of the term scheme
with fluorescence lines (in nm) for detection; (b) measured two-photon spectrum; (c) calibration
with a FABRY-P ROT interferometer: the marker lines are spaced by one free spectral range of the
FPI of 300 MHz, corresponding to a change in absorption frequency of 600 MHz
This is a major challenge for precision spectroscopy. Two photons at a wavelength

of 243 nm are needed. They may be are generated by frequency doubling (SHG) of
a highly stabilized and calibrated CW dye laser at 486 nm. The detection of the ex-
citation is typically achieved by quenching the excited, metastable 2s 2 S1/2 atoms
(i.e. transferring them into the 2p 2 P states) and registering the emitted LYMAN-
radiation. We shall come to more details of this detection method in the context of
Sect. 6.5.3.
Figure 6.16 illustrates an interesting realization of this experiment by H NSCH
and coworkers (G ROSS et al. 1998) who employ a scheme with R AMSEY fringes
(which we have introduced in Sect. 6.1.7). In the present case the R AMSEY fringes
are generated by an AOM which chops the laser radiation in a sequence of pulses.
Comparison of Fig. 6.16(b) and (c) shows that a substantial improvement of resolu-
tion may be achieved by a judicious choice of pulse durations and distances. Clearly,
in addition to excellent resolution a precision measurement of also requires accurate
calibration of the frequency. We shall discuss this aspect in Sect. 6.5.4.
Finally, we want to point out that the two-photon process discussed here provides
also the basis for a frequently used state-of-the-art method for H atom (radical)
detection, e.g. in chemical gas phase reactions:in a so called 2 + 1 REMPI process
one uses the resonant 2 photon excitation of the 2s 2 S1/2 state (243 nm) and then
ionizes this state efficiently by a third photon, typically of the same wavelength.
Section summary
High resolution spectroscopy is the key to the investigation and understanding
of finer details in atomic structure, such as FS and HFS. A brief summary of
some essential tools has been given in this section.
vacuum chamber
cold finger
(a) gas discharge
H atoms 2S-
FARADAY
cage detector
dye laser SHG AOM
486 nm 243 nm Fabry-Perot

resonator
2S Signal / cps 2S Signal / cps photon counter
7500
800 Hz
2S signal / cps
(b) 100 @ 243 nm (c)

5000
50
2500
0 0
- 80 - 40 0 40 80 - 20 - 10 0 10 20
detuning / kHz @ 121 nm detuning / kHz @ 121 nm
Fig. 6.16 Two-photon excitation of the 1s 2 S1/2 2s 2 S1/2 transition in the H atom according to
G ROSS et al. (1998). (a) Scheme of the experiment. An acousto-optic modulator (AOM) generates
suitable pulse sequences at 243 nm. (b) Excitation probability as a function of detuning in respect
of the 2h resonance when irradiated by 50 s pulses at 10 kHz repetition rate; (c) excitation rate
3.2 kHz, otherwise as (b)
Interference based spectrometers such as grating spectrometers and interfer-

ometers have a resolving power / = N |z| where N is the number of
interfering rays and z the order of interference.
A particular important device is the FABRY-P EROT interferometer. The ef-
number of interfering rays is given by its finesse F = FSR / h =
fective
R/(1 R), with FSR = c/2L being the free spectral range between the
two plates in a distance L, h the FWHM of the transmitted line, and R the
reflectivity of the plates.
The most straight forward way to avoid D OPPLER broadening of optical tran-
sitions in atoms and molecules is provided by molecular beam spectroscopy
with perpendicular irradiation or emission. Collinear spectroscopy in high en-
ergy ion beams is another efficient method to dramatically reduce D OPPLER
linewidths.
In addition, several ingenuous methods have been devised for state-of-the-art
laser spectroscopy. Experimentally rather easy to realize are saturation spec-
troscopy (based on hole burning in intense radiation fields) and two-photon
excitation, both allowing to achieve sub-D OPPLER resolution even in gas
cells.
6.2 Spin-Orbit Interaction 293
A particularly powerful method to prolong the interaction time in spec-

troscopy is the exploitation of R AMSEY fringes which are an important in-
gredient of modern atomic clocks.
6.2 Spin-Orbit Interaction
6.2.1 Experimental Findings
In Chap. 3 we have described in detail how for alkali atoms the degeneracy of energy
levels with the same principle quantum number n is removed for different orbital
angular momentum quantum numbers . However, as we have already noted in the
preceding section, on closer inspection one finds that even the n levels are again
split into a finer substructure and the same holds for the absorption and emission
lines of the H atom.
Quantitatively one finds that this so called fine structure splitting (FS) is on the
order of magnitude
WFS Eh
Z2 = (Z)2 (6.25)
Wn me c2
with the atomic unit of energy Eh , the rest energy of the electron me c2 , and =
Eh /me c2 the fine structure constant (1.10) which we have encountered already on
several occasions. This order of magnitude already suggests that FS must be of
relativistic origin.
For light atoms the splitting is rather small and non-trivial to measure in the gas
phase. For the H atom in the first excited state FS splitting is of similar magnitude
as D OPPLER broadening (5.22) at room temperature,
FS 0.3 cm1 and D = D /(2c) (0.10.2) cm1
in wavenumbers. Hence, some effort is necessary to see the splitting as already

illustrated in Fig. 6.13. Much more convenient to observe is the FS splitting for
alkali atoms. Form example, the D lines of Na (at 17000 cm1 ) are a doublet at
wavelengths
D1 : 1 = 589.0 nm and D2 : 2 = 589.6 nm. (6.26)
Figure 6.17 illustrates the term scheme. The splitting is about 17 cm1 while
the D OPPLER width amounts to D 0.06 cm1 , so that spectroscopic separa-
tion is possible without problems. One needs an experimental resolution of about
/ = / 1000. That can easily be achieved with a grating monochromator.
All this observed fine structure is connected with the electron spin. As already
discussed in Sect. 1.10 the spin (quantum number S) is accompanied by a magnetic
moment which interacts with the orbital angular momentum (quantum number L)
of the electron. This is called spin-orbit (or LS) interaction. Since two possible
Fig. 6.17 Fine structure 3p 2P3/2

transitions in the first excited
state of Na 3p 2P1/2
3s 2S1/2
orientations of the electron spin, sz = 1/2, are possible in respect of a given axis
here defined by the angular momentum, the levels are split into two components.
For the example of a 3p electron shown in Fig. 6.17 the two states are characterized
by a total angular momentum j = 3/2 and 1/2. For the 3s state the orbital angular
momentum is zero, hence j = 1/2 and no splitting is observed. We shall discuss this
in detail below.
6.2.2 Magnetic Moments in a Magnetic Field
We recall briefly the experimental findings communicated in Sect. 1.91.10. With

the B OHR magneton B 14 GHz T1 according to (1.160), the magnetic moments
of electron are for the orbit

L = gL B L = e
M L, (6.27)
2me
and for the spin

S = ge B S = ge e
M S. (6.28)
2 me
The g factor of the orbit is gL 1, and for the spin it is exactly ge = 2 in D IRAC
theory. On closer inspection there is, however, a small deviation from this value so
that ge 2.0023 . . . , as will be elaborated in Sect. 6.6.
Now we analyze the interaction of the combined magnetic moments with an
external field B. For simplicity we choose B z. Each of them alone would have an
interaction energy
B = M
B = M
V z B (6.29)
depending on the projections of M L or M S onto B. With (6.28) and (6.27) the

interaction potentials for orbital and spin magnetic moments thus are:

L = gL B Lz B with m |V
V L | m = gL B Bm (6.30)

S = ge B Sz B with sms |V
V S |sms = ge B Bms . (6.31)

Fig. 6.18 Schematic z z

illustration of the origin of BL
spin-orbit interaction: the
orbital magnetic field B L is
L L
+ Ze y y
viewed as arising from the -e
nucleus circling around the
electron x -e x +Ze
If only the orbital magnetic moment or only the spin magnetic moment were acting,
these expressions would represent the level splittings in an external B field, and one
would observe
2 + 1 possible orientations in space for the angular momentum ( m )

or alternatively
2s + 1 = 2/2 + 1 = 2 for the spin (ms = 1/2).
6.2.3 General Considerations About LS Interaction
The situation is further complicated by the interaction between the magnetic mo-
ment M S of the electron spin and the magnetic field B L created by the orbiting
electron. To derive this interaction quantitatively one needs to know the effective
magnetic field at the position of the electron spin. Thus, one transforms the mo-
tion into a coordinate system where the atomic nucleus rotates around an electron
considered at rest as indicated in Fig. 6.18. Without going into details of the un-
derlying electrodynamics we note: The L ORENTZ force F = qv B (1.105) onto
a charge q moving with velocity v in a magnetic field B may be viewed as the re-
sult of an effective electric field E eff = v B. Conversely, a charge moving in an
electric field E generates a magnetic field
vE
B = . (6.32)
c2
In the potential V (r) of the ionic core of an atom (atomic nucleus + core elec-
trons) the electric field E experienced by an electron is derived from
r dV
eE = F = .
r dr
Hence, the electron moving in this field generates according to (6.32) a magnetic
field
v r dV 1 1 dV 1 1 dV
BL = = (r me v) = L (6.33)
ec2 r dr eme c2 r dr eme c2 r dr
which obviously is proportional to the orbital angular momentum L. We now con-

sider instead the positive nuclear charge circling around the electron, i.e. we sub-
stitute this expression with a positive sign into (6.29) and obtain with (6.28) and
ge = 2:

LS = MS B L = 1 ge 1 dV L S .
2
V (6.34)
4 me c r dr 2
2 2
We have introduced here without proof the so called T HOMAS factor 1/2 which
follows directly from D IRAC equation when re-transforming V LS into the nuclear
system at rest.
Introducing Eh , a0 and the fine structure constant = Eh /(me c2 ) the spin-
orbit (or LS) interaction is

LS = a 3 ge (r) L S Eh
2 40 dV
V with (r) = a03 . (6.35)
4 0
2 r dr
Note that for an attractive, pure C OULOMB potential (r) = Zr 3 . To simplify the
writing, in the remainder of this chapter all equations will be given in a.u. (unless
otherwise indicated).
6.2.4 Magnitude of Spin-Orbit Interaction
The diagonal matrix element of (6.35) may be written (with ge

= 2):
V L
LS = a S
2
with a = ge n |(r)|n (6.36)
4

2 1 dV 2 1 dV 2
= n | |n = 2
Rn (r) r dr.
2 r dr 2 0 r dr
Here we have introduced the spin-orbit coupling parameter a. Only for the H
atom and H atom like ions it can be evaluated in closed form. The results are, how-
ever, characteristic and may readily be extended to quasi-one-electron systems such
as the alkalis. For V (r) = Z/r and (r) = Zr 3 the coupling constant a can
be derived from the expectation value (2.130) of r 3 for hydrogenic wave func-
tions:

2 1 2 Z4
a = ge Z 3 = . (6.37)
4 r 2 n3 ( + 1/2)( + 1)
This disappears for = 0. For > 0 we obtain a good estimate by approximating
L
S s so that FS splitting is seen to be on the order of
2 Z4 VLS (Z)2
LS
V and ,
2 n3 ( + 1/2)( + 1) |Wn | n( + 1/2)( + 1)
where |Wn | = Z 2 /2n2 has been introduced to obtain the relative perturbation. At
small Z 11 it is very small, only about 106 . . . 103 for n 3. In general, the
FS splitting decreases 1/n3 (for any given , spectroscopically relevant being
4).
However, spin-orbit interaction becomes substantial for heavy atoms because of
its proportionality to Z 4 . Perturbation theory will break down altogether for Z 1.
The physics emerging from such a situation is subject to current research with atoms
of nuclear charges beyond the natural stability limit for 92 U. Such investigations are
carried out today at heavy ion accelerators and storage facilities where this limit can
easily be surpassed in highly energetic collisions.
6.2.5 Angular Momentum Coupling
Including spin-orbit interaction the Hamiltonian becomes

2 2
L
2
2
H 0 (r) + (r)
FS (r) = H L
p
S = r + 2 + V (r) + (r) L
S (6.38)
2 2 2r 2
r2 /2 and
2
with p L /(2r 2 ) being the radial and angular kinetic energy. To evaluate the
fine structure term 2 (r)L
S/2 we have to consider the coupling of spin and orbit
through their respective magnetic fields and derive the resulting eigenstates of the
Hamiltonian (6.38).
One formally introduces an operator of total angular momentum
L +
J = z +
S and its z component Jz = L Sz , (6.39)
for which the standard commutation rules hold in full analogy to those for orbital,
(2.84), (2.75), and spin angular momenta, (2.91):

J , Jz = 0
2
(6.40)

Jx , Jy = iJz , Jy , Jz = iJx , Jz , Jx = iJy . (6.41)
This implies that one can measure simultaneously the magnitude and one of the
components of the total angular momentum but not two components. Specifically
we shall use eigenstates |j mj for which
J |j mj = j (j + 1)2 |j mj and Jz |j mj = mj |j mj .

2
(6.42)
J has 2j + 1 possible orientations, i.e. Jz is characterized by

Correspondingly,
2
the quantum numbers

mj = j, j + 1, . . . , +j. (6.43)
L and
For the constituents S of this total angular momentum corresponding relations
hold. However, their quantization axis is now the direction of
J . One visualizes this
Fig. 6.19 Vector model for

the coupling of spin s = 1/2 J
Jz
and orbital angular S
momentum = 2 forming a
|j = 5/2 mj = 5/2 state (to L
S
scale)
Fig. 6.20 Vector model for Jz mj

the possible projections Jz of 3/2 3/2
a total angular momentum 1/2 1/2
J = 3/2
-1/2 -1/2
- 3/2 - 3/2
by a vector model as shown in Fig. 6.19. The following constraints must hold:

|
J | = j (j + 1), |
L| = ( + 1), |
S| = s(s + 1) (6.44)
with = 0, 1, 2, . . . , n 1,
together with the triangular relation ( sj ) = 1, or more explicitly:
| s|, | s| + 1, . . . j | + s| (6.45)
We have already encountered the triangular relation in Sect. 4.4.1 when discussing
selection rules for dipole transitions and the photon spin sph = 1. For an electron
spin s = 1/2 and > 0 it leads to j = 1/2, while for s = 1/2 and = 0 we
have j = 1/2. The possible projections mj are illustrated in Fig. 6.20 for the case
of j = 3/2.
Scalar Product of Angular Momentum Operators

With the definition (6.39) the square of the total angular momentum is
L+
J = (
S)2 = L
L + 2 S +
2 2 2
S (6.46)
and the scalar product of angular momentum and spin operators becomes
1 2 2 2
L
S= J L S . (6.47)
2
The whole H AMILTON operator (6.38) can then be rewritten:
2
2 2
FS (r) =
H
p 2r L 2
+ 2 + V (r) + (r) L
J
2
S . (6.48)
2 2r 4
Hence, we have to search for eigenstates |( s)j mj created by the coupling of
L
and
S which are at the same time eigenstates of J ,L and L
S and thus of
2 2 2
S.
For these states (6.42) must be valid simultaneously with

1 2 2
S ( s)j mj =
L J L S ( s)j mj
2
(6.49)
2
1
= j (j + 1) ( + 1) s(s + 1) ( s)j mj .
2
L and
During this evaluation we have exploited the fact that S commute, since they
operate on different spaces (position and spin space, respectively). Hence, also
2
L
and L as well Si
2 i
S commute. Since S Si =
2 2 2 2
L L i =L S for any component i,
and since
L
x
S=L y
Sx + L z
Sy + L Sz . (6.50)
one may see with (6.46) that L and S commute also with L S and
2 2 2
J . Hence,
L ,
S and FS according to (6.48) and j , and s are
2 2 2
J all commute with H
good quantum numbers (see Sect. 2.6.5).
L
Note that S does not commute with L z and not with
Sz as one sees from mul-
tiplying (6.50) with L x nor L
z since neither L y commute with L z , and analogical
for Sz . Thus, neither m nor ms are good quantum numbers. On the other hand,
z +
Jz = L Sz does commute with L S, since with (6.50)
1 2 2 2
L
[ S, Jz ] = J , Jz L , (Lz + Sz ) S , (Lz + Sz ) = 0.
2
Thus, mj is a good quantum number, and we note that because of Jz = L
z +
Sz
m + ms = mj (6.51)
must always hold for eigenstates of the Hamiltonian (6.38), while neither m nor ms
are good quantum numbers.
The eigenstates of the complete Hamiltonian H FS which we are looking for are
thus characterized by n, , s, j, mj :
FS |n sj mj = Wn j |n sj mj .
H
L
We note, that (6.42) holds for these states. They are also eigenstates of S and
according to (6.47) we have (in units of )
2
1
L
n sj mj | S|n sj mj = j (j + 1) ( + 1) s(s + 1) . (6.52)
2
Eigenstates of the Total Angular Momentum

The relations developed above allow to write the sought after states |n( s)j mj by
linear combinations of product states |n sm ms = |n m |sms in such a way that
m + ms = mj . This has to be done such that the Hamiltonian (and thus also L S)
becomes diagonal. For spin and orbital angular momentum this can be achieved by
using the definitions and commutation rules for angular momenta as sketched above.
We refer to the relevant extensive literature on angular momentum algebra and to
the brief extract of the most important formulas collected in Appendix B.2.3 The
angular part of the eigenstates for (6.48) is written

( s)j mj = | m |sms m sms ( s)j mj or
m ,ms

mj + s s j
= 2j + 1 | m sms (1) , (6.53)
m ms mj
m ,ms
using the so called C LEBSCH -G ORDAN coefficients sm ms | sj mj or 3j sym-

bols ( ), respectively.
The Most Simple Example

The most simple example for angular momentum coupling is offered by two elec-
trons with spin quantum numbers s1 = s2 = 1/2 and zero angular momentum. Their
total spin is

S =
S1 +
S2
so that the triangular relation (4.72) with (1/2 1/2 S) = 1 allows for (2 1 + 1) = 3
states with total spin S = 1 (so called triplet state) and one state for the total spin
S = 0 (singlet):

1/2
(s1 s2 )SM = s1 m1 s2 m1 |s1 s2 SM |s1 m1 s2 m2
m1 =1/2
meter2 =1/2

1/2
s1 s2 S
= 2S + 1 (1)M |s1 m1 s2 m2 .
m1 m2 M
m1 =1/2
meter2 =1/2
With help of the readily accessible 3j symbols (e.g. with a Java applet from S VEN
G ATO R EDSUN 2004) one verifies that the nonvanishing C LEBSCH -G ORDAN coef-
ficients are
1 1 1 1 1 1 1 1
1 1 = 1
2 2 2 2 11 = 1, 2 2 2 2
1 1 1 1 1 1 1 1
10 =
2 2 2 2 2 2 2 2 10 = 1/ 2
1 1 1 1
00 = 1 1 1 1 00 = 1/ 2.
2 2 2 2 2 2 2 2
3 Several slightly different definitions are used. We follow B RINK and S ATCHLER (1994).
Table 6.1 Triplet and singlet states with C LEBSCH -G ORDAN coefficients for the coupling of
two electrons with spin 1/2. The symmetry of the spin functions is given as well as symbolic
representations of the two spin 1/2 states involved
Multiplicity Symmetry |SM = | (1, 2) S M

|11 = |(1)(2) 1 1
equal
|(1)(2) + |(1)(2)
Triplet symmetric |10 = 1 0
2 phase
|11 = |(1)(2) 1 1
opposite
|(1)(2) |(1)(2)
Singlet antisymmetric |00 = 0 0 phase
2
The corresponding spin functions for triplet (symmetric) and singlet states (anti-
symmetric) are summarized and symbolized in Table 6.1, using the abbreviations
| = |1/2 1/2 and | = |1/2 1/2 as introduced in Sect. 2.5.4. The numbers in
brackets behind the spin designations and refer to electron number (1) or (2).
Note that the symmetry of a full wave function in spin and position is the product
of the symmetries in spin and position space. As we shall elaborate on in Sect. 7.3.1
and Sect. 10.2.1 the PAULI principle demands the full wave function to be antisym-
metric!
6.2.6 Terminology for Atomic Structure
To characterize atomic eigenstates in a compact and transparent manner an unam-

biguous terminology is used. It may readily be applied to atoms with a single active
electron as well as to multi-electron systems. In the case of the so called RUSSEL -
S AUNDERS coupling (somewhat misleadingly also called LS coupling) one first
composes
as vectorial sums over all active electrons a total angular momentum

L= Li (quantum
number L) from all individual orbital angular momenta and
a total spin S= S i (quantum number S) from all individual electron spins. As
just discussed for a single electron, both angular momenta are then coupled to the
overall total angular momentum J =S +L (quantum number J ).
Note that all relations for angular momenta presented in Sect. 6.2.5 refer there
to individual electrons with quantum numbers sm ms (uncoupled) and j mj (cou-
pled). They do, however, also hold for the quantum numbers of the combined total
angular momenta LSML MS and J MJ . In Sect. 7.3 we shall illustrate this in de-
tail for the example of the helium atom, and also introduce there the alternative jj
coupling scheme which is relevant for heavier atoms where spin-orbit interaction is
substantial.
Here we just communicate the general terminology of LS coupling. The number
of possible orientations of the total spin is called
multiplicity = 2S + 1 (6.54)
of the state. For a one electron system such as alkali atoms we have S = s = 1/2
and the multiplicity is always 2 1/2 + 1 = 2, one calls this a doublet. As we
have just seen in Sect. 6.2.5, atoms with two electrons have singlet and triplet states
(S = 0 and S = 1, respectively). In systems with N electrons in principle multi-
plicities up to 2N 1/2 + 1 could be constructed. Typically, quintuplets are the
highest multiplicities observed (constructed from 4 electrons, i.e. half the number
of electrons in a completely filled outer valence shell with two s and four p elec-
trons).
To distinguish the quantum number of a single electron from those of the full
atom on characterizes them with lower case and capital letters, respectively. Hence,
the total orbital angular momentum of an atomic state (quantum number L) is char-
acterized by upright capital letters S, P, D, F, G for L = 0, 1, 2, 3, 4 (in analogy to
s, p, d, f, g, . . . for the i quantum numbers of individual electrons). As superscript
on the left of the angular momentum one writes the multiplicity, as subscript on
the right the total angular momentum is given. In summary, an atomic state with
N electrons in a configuration {n1 1 . . . nN N } (see Sect. 3.1.1), total orbital an-
gular momentum L, total spin S and overall total angular momentum J is written
as
{n1 1 . . . nN N } 2S+1 LJ . (6.55)
Often the electron configuration is dropped for brevity. As additional information
for multi-electron spectra one often finds a superscript o for odd parity, especially
if the latter cannot be identified uniquely from the angular momentum quantum
number (more about parity is discussed in Appendix D).
We give a few examples:
For the H atomas well as for alkali atoms s states with = 0 and j = J =
| 1/2| = 1/2 are denoted as ns 2 S1/2 ,
and p states with = 1 and j = J = | 1/2| = 1/2, 3/2 are correspondingly
np 2 Po1/2 and np 2 Po3/2 , respectively,
while for a d state with = 2, j = | 1/2| = 3/2, 5/2 and hence nd 2 D3/2 or
nd 2 D5/2 holds.
The ground state of the He atom (two 1s electrons, total spin = 0) is characterized
by 1s 2 1 S0 ,
while the first excited states are in the singlet system 1s2s 1 S0 and 1s2p 1 Po1 ,
and the corresponding triplet states are 1s2p 3 Po2 , 1s2p 3 Po1 , 1s2p 3 Po0 and
1s2s 3 S1 . They will be discussed in detail later.
6.3 Quantitative Determination of Fine Structure 303
We mention, finally, that in the spectroscopic literature and relevant data collec-
tions the spectra of neutral atoms are marked with the roman number I, those of
singly ionized atoms with II and generally the spectra of the N fold ionized species
with roman N + 1. Thus, one speaks of the C I, O II, Fe III spectra and refers in
this way to the spectra of the neutral C atom, and of the ions O+ and Fe++ , respec-
tively.
Section summary
Spin-orbit (or LS) interaction (6.35) arises as magnetic dipole energy of the
electron spin oriented in the magnetic field due to the electron orbital angular
momentum. It is proportional to L S and to the gradient of the potential in
which the electron moves.
Quantitatively, the resulting fine structure (FS) splitting is on the order of
2 Z 4 and decreases 1/n3 with the principal quantum number (its relative
magnitude being (Z)2 /n).
With spin-orbit interaction the orbital and spin angular momenta couple to a
total angular momentum J = L+ S. With | m and |sms describing the
orbital and spin angular momentum, respectively, the states of the coupled
angular momenta |(ls)j mj are a sum of products of the former (involving so
called C LEBSCH -G ORDAN coefficients or 3j symbols).
good quantum numbers are n j mj with mj = m + ms , while m and ms are
not. The L
S interaction term is diagonal in this scheme, its eigenvalue being
( /2)[j (j + 1) ( + 1) s(s + 1)].
2
In multi-electron systems the angular momenta of several active electrons are

coupled. For low atomic numbers Z the FS can be treated as a small pertur-
bation, and RUSSEL -S AUNDERS coupling (also called LS coupling) gives a
good description of the atomic structure: first all electron orbital momenta i
couple to L (quantum number L) and all electron spin momenta s i couple to
S (quantum number S). Then L and S couple to a total angular momentum J
(quantum number J ).
A specific electron configuration it then characterized by 2S+1 LJ , with 2S + 1
giving the multiplicity of the spin states.
6.3 Quantitative Determination of Fine Structure
6.3.1 FS Terms from D IRAC Theory
As we have seen in Sect. 6.2.3 the spin-orbit splitting is on the order of 2 Z 2 =

Z 2 Eh /(me c2 ) hence the name fine structure constant for . The comparison
with me c2 already indicates that this interaction is of relativistic nature. Hence, an
exact treatment of the fine structure requires the solution of the D IRAC equation,
which combines the principles of both, quantum mechanics and special relativity
in contrast to the S CHRDINGER equation. It describes fermions (such as the
electron) correctly by so called spinors (with four complex values) rather than by
one complex wave function, and it provides automatically the correct electron spin
s = 1/2 with its magnetic moment and ge = 2, as well as the concept of antimatter.
Unfortunately, an exact solution and comparison with precision spectroscopy is
only possible for the H atom and does even in this case as we shall see in Sect. 6.5
not lead to a complete agreement with experiment. Interactions with the vacuum
field, to be treated with QED, enter into the term energies with small contributions
only about an order of magnitude smaller than those introduced by the (relativisti-
cally correct) D IRAC theory. Thus, we refrain here from a detailed treatment of the
D IRAC equation which is found in all major quantum mechanics textbooks.
Clearly, computations with the D IRAC equation (even approximate ones) are sub-
stantially more complex than those with the S CHRDINGER equation. Hence, one
tries to make the latter relativistic, i.e. to include electron spin and relativis-
tic effects by two component spin functions (as outlined in Sect. 6.2.5), originally
introduced by PAULI, and by adding appropriate interactions derived from D IRAC
theory.
As summarized below, there are three relativistic terms which have to be added
to the S CHRDINGER equation. All of them are on the order (Z)2 , i.e. of similar,
rather small magnitude. If they cannot be treated exactly, perturbation theory usu-
ally leads to very satisfactory results for most problems in AMO physics and quan-
tum chemistry. In combination with QED and higher order approximations, one can
even explain state-of-the-art high precision experiments extremely well for not too
large atomic nuclei, i.e. as long as (Z) 1 can be assumed in appropriate series
expansions.
Relativistic Correction Term to the Kinetic Energy

From the relativistic expression of the total energy (1.19) and with (1.22) one obtains
the relativistic contribution to the kinetic energy in SI units

p2
Wkin = m2e c4 + p 2 c2 me c = me c 1 + 2 2 me c2
2 2
(6.56)
me c

and approximates by expanding the root 1 + x 2 = 1 + x 2 /2 x 4 /8 +
p2 1 p4
Wkin = .
2me 8 m3e c2
Thus, the relativistic, quantum mechanical correction term is to lowest order
2 2 2
4
p 1 p 1 1 Eh p2
H1 = 3 2 = = Eh ,
8me c 2 2me me c2 2 me c2 2me Eh
or in a.u. (with the fine structure constant = Eh /me c2 ):

2 4
1 = p 2 2
0 (r) = 2
p
H = H0 V (r) with H + V (r). (6.57)
2 4 2 2
For H atoms this leads to

2 2 2 2 Z Z2
H1 = H0 V (r) = H0 + 2H0 + 2 .
2 2 r r
1 obviously depends only on r. In 1st order perturbation theory the energy shift is
H
2
1 |n sj mj =
2
0 Z + Z |n
02 + 2H
Vrel = n sj mj |H n |H 2
2 r r
2
1 1
= Wn2 + 2ZWn + Z2 2
2 r r
with Wn = Z 2 /(2n2 ). Inserting the matrix elements for the H atom according to
(2.130) one obtains in a.u.

(Z)2 3 1
Vrel = Wn . (6.58)
n 4n + 1/2
D ARWIN Term
The so called DARWIN term emerges from D IRAC theory as in a.u.
2
3 = Z (r).
H (6.59)
2
This perturbation is non-zero only at the origin r = 0 and we only need to evaluate
it for s states ( = 0) the only states which have a finite probability at the origin.
In 1st order perturbation theory we obtain
3 |n sj mj = Z 2
VD = n sj mj |H ns|(r)|ns
2

Z 2 2 Z 2 2 (Z)2
VD = ns (r) (r)d3 r = ns (0) = Wn , (6.60)
2 2 n
where the last identity is derived for the H atom from inserting Rn0 (0) according to
(2.122), using the explicit expressions for the L AGUERRE polynomials (2.123).
We note that the relativistic correction as well as the DARWIN term represent
( independent) shifts of the energy levels and not splittings. Only if the term posi-
tions can be computed exactly (i.e. essentially only for the H atom and H like ions)
one may hope to compare experiment and theory in respect of these shifts. Already
for the alkali atoms, where degeneracy is significantly removed, such an absolute
comparison of these very small effects is usually beyond the limits of the combined
accuracy of theory and experiment.
Spin-Orbit Term
In contrast to the two terms discussed above the spin-orbit interaction V LS (6.35)
leads to a splitting which can be calculated and measured with good accuracy and
L
is relevant for all atoms. In the S coupled scheme it is diagonal in respect of the
sj mj quantum numbers:
2
LS = n |(r)|n sj mj |
VLS = V L
S| sj mj . (6.61)
2
With (6.52) and with the spin-orbit interaction parameter a defined in (6.36) one
may write the energy change due to spin-orbit interaction as
a
Wn j VLS = j (j + 1) ( + 1) s(s + 1) . (6.62)
2
Obviously, the fine structure splitting depends on j and . States with = 0 and
j = s do not split, all other states do. Systems with one active electron and s = 1/2
form doublet energy levels with j = 1/2. If the potential V (r) is known, the fine
structure splitting is readily derived from the matrix element n |(r)|n .
6.3.2 Fine Structure of the H Atom
For the H atom and H like ions and > 0 (6.62) becomes with (6.37):
"
Z4 2 1 for j = + 1/2
VLS = (6.63)
2n3 2 ( + 1/2)( + 1) ( + 1) for j = 1/2.
Interestingly, the total energy with all FS terms included depends only on j . The
exact solution of the D IRAC equation (e.g. B JORKEN and D RELL 1964, Eq. 4.14) is
usually expanded in powers of (Z)2 and is

(Z)2 (Z)2 1 3
Wnj = 1+ + O(Z) me c2 ,
6
(6.64)
2n2 n j + 1/2 4n
in units of the electron rest energy, while in a.u. it reads

Z2 Z4 2 1 3
Wnj = + + Eh . (6.65)
2n2 2n3 j + 1/2 4n
In the following we stay with a.u. unless otherwise mentioned. The terms are split
into FS levels with j = 1/2, if and only if 1, by
Z4 2 1
Wn = Eh . (6.66)
2n3 ( + 1)
s1/2
0.730 p 3/2 0.091

0.122
2s, 2p p1/2 p 3/2 s1/2
0.213 0.243 p 3/2
0.456
p1/2 p3/2 p1/2
1.186 s1/2 p1/2
Wn s1/2 WNJ
BOHR -
SCHRDINGER Vrel relativistic VD DARWIN VLS spin orbit DIRAC
Fig. 6.21 Fine structure of the H atom, n = 2 levels. All energy shifts in cm1
Figure 6.21 summarizes the situation for the n = 2 levels of atomic hydrogen:
The non-relativistic S CHRDINGER equation (as well as the B OHR model) pre-
dicts identical energies Wn for the 2s and 2p levels (left). The different contri-
butions to Wnj from D IRAC theory are indicated as dotted lines. The final results
Wnj = Wn + Vrel + VD + VLS , are shown on the right. Note that D IRAC theory
predicts Wnj for s1/2 and p1/2 to be identical.
6.3.3 Fine Structure of Alkali and Other Atoms
The LS splitting decreases rapidly with n and as one would expect for an in-
teraction 1/r 3 . For not too large (typical in spectroscopic studies) we memo-
rize that fine structure splitting quite generally is Z 4 /n3 . Spin-orbit coupling is
particularly important for larger atoms. As we shall see in the following chapters,
RUSSEL -S AUNDERS (or LS) coupling works very well for light atoms where spin-
orbit coupling is small. It breaks down for large Z and other coupling schemes take
over, such as the so called jj coupling.
As already mentioned, only for H and hydrogen like ions He+ , Li++ , etc. (and
a few special exceptions, see Sect. 6.5.5) absolute calculations of term energies are
possible with a precision to verify or falsify the energetic level positions shown in
Fig. 6.21. However, the FS splittings due to L S interaction for individual n levels
may be computed with sufficient accuracy for comparison with experiment. Thus,
LS according to (6.35) to the
for practical applications it is usually sufficient to add V
standard S CHRDINGER equation, which for effective one electron systems simply
amounts to (6.62). For very heavy atoms it may become necessary to also account
for the relativistic kinetic energy shift.
For effective one electron systems such as alkali atoms and > 0 the energy
levels split in doublets according to (6.62). The spin-orbit coupling parameter a
(6.36) summarizes the specific FS properties of a given atom in a state characterized
by n . Fine structure splitting for alkali atoms is significantly larger than for the
Table 6.2 Fine structure splitting of the first excited p states of H and the alkali atoms
Atom H Li Na K Rb Cs
n 2p 2p 3p 4p 5p 6p
Wn j / cm1 0.365 0.335 17.196 57.71 237.595 554.039
H atom. Table 6.2 gives a survey for the first resonance lines (ns 2 S1/2 np 2 P1/2
and np 2 P3/2 ) of different alkali atoms in comparison to the H atom. Note that for
Cesium (Cs) FS splitting is already 5 % of the total transition energy not really a
small effect any more. We summarize a few general rules valid for H and the alkali
atoms but also beyond:
The higher n and the smaller the FS splitting. For example, in Na already the
5p level is split by only 2.47 cm1 and for the 3d level, where the electron hardly
ever comes close to the nucleus it is only 0.05 cm1 .
For term energies usually Wnj = +1/2 > Wnj = 1/2 holds this is called normal
ordering. For higher terms and heavier elements there are cases where the en-
ergies are inverted: the one electron picture is just an approximation which may
break down under extreme conditions.
Expression (6.62) for the FS splitting is (approximately) valid in many cases
also for more complex atoms. Then, of course, the overall total angular momen-
tum J , the total orbital angular momentum L and the total spin S > 1/2 replace
j , and s. The formula has proved to be valuable in particular for larger Z and n,
at least as long as LS coupling is still valid.
Using (6.62) one derives the distance of two neighbouring fine structure levels in
a multiplet, the so called L AND interval rule:
a
WFS = WJ WJ 1 = J (J + 1) (J 1)J = aJ. (6.67)
2
The energy differences between any two neighbouring FS levels with J and J 1
in an FS multiplet are proportional to J . For doublets, (6.67) gives the splitting.
Occasionally, equation (6.37) (valid for the H atom) is adapted by substituting Z

with an effective nuclear charge Zeff . However, a comparison with experimental
data for the alkali atoms shows that such a formula has only rather limited predictive
power for FS splittings even within one series of n at fixed .
A much more consistent parametrization of a given n 2 L series for all n is
obtained for the effective change of the quantum defect due to FS as derived from
QDT and documented in Fig. 6.22. Using (3.25) we obtain
j = n 3 Wn j (6.68)
which may be applied since spin-orbit interaction scales with 1/n3 .

In Fig. 6.22 we exemplify this for the FS splitting of the 2 P and 2 D series in
Na and Rb, summarizing (as a function of binding energy) all data reported by
NIST for the FS splitting Wn j in Na and Rb effectively by two parameters for
0.4
0.02223 0.3688
0.022 np 2P
np 2P
0.3
0.020 0.00444
0.04
Na (Z=11) Rb (Z=37)
nd 2D
0.00042 0.02
nd 2D
0.00023
nf 2F
0 0.00
-3.0 -2.0 -1.0 0.0 -3.0 -2.0 -1.0 0.0
/ eV
Fig. 6.22 Effective change j = n 3 Wn j of the quantum defect due to FS for Na and Rb.
The experimental data points have been derived from the NIST data (K RAMIDA et al. 2013). Note
that j follows very clearly the linear least squares fits (red lines) as a function of the binding
energy Wn typical for quantum defects (compare Fig. 3.7)
each n 2 L series. As shown in Fig. 6.22, the linear least squares fits represent the
experimentally observed data extremely well. We have to recall: j is the change
of the quantum defect due to FS splitting. It has to be compared to (n, ) as shown
in Fig. 3.8 (for Wnl 0, n 1). Obviously, for Na the FS splitting is a minor effect,
in particular for the P1/2 2 P3/2 series where (, 1) 0.8551 and j 0.022.
In contrast, for the same series in Rb FS splitting with j 0.369 is no longer
negligible in comparison to (, 1) 2.6535.
Section summary
An exact treatment of fine structure requires the solution of the D IRAC equa-
tion. However, very good approximations can be achieved by incorporating
electron spin and the three essential interaction terms into the S CHRDINGER
equation:
Relativistic correction to the kinetic energy, according to (6.57),
the DARWIN term, non-zero only at the origin and thus relevant only for s
electrons, and
spin-orbit interaction which can be diagonalized for one electron systems
by for |( s)j mj states, and is given by (6.61) and (6.62).
The first two terms represent shifts and can be compared with experiment
essentially only for H and H like ions. Spin-orbit interaction leads to splittings
which are readily observed and compared to theory for all atoms.
For atomic H (and H like ions) D IRAC theory predicts a splitting between
np3/2 and np1/2 states, but equal energies for the latter and ns1/2 .
For larger alkali atoms FS splitting is no longer a small effect and amounts
e.g. for the first resonance line in Cs already to 5 % of the transition en-
ergy.
6.4 Selection Rules and Intensities of Transitions

6.4.1 Introduction
For fine structure transitions all selection rules hold as derived and discussed for the
general case in Sect. 4.4. In particular, for dipole allowed (E1) transitions the trian-
gular rule (Ja Jb 1) = 1 holds for the total initial Ja and final Jb angular momenta.
Their projection quantum numbers may change according to Ma Mb = 0, 1
depending on the polarization of the emitted or absorbed light. We use here capital
letters, indicating that all selection rules also hold for systems with several electrons.
In the following we shall apply the general expressions for transition probabilities
derived in Sect. 4.6 to systems which ar well described by LS coupling. Specifically
we are interested in the relative intensities of several lines within one multiplet. This
requires some extension of the angular momentum algebra used so far. These tools
albeit they may appear somewhat uninspiring in the beginning will later on
prove very useful for describing various other, more complex situations. Details are
presented in Appendixes C and B.3.
To illustrate the present subject, Fig. 6.23 gives a schematic overview of the
dipole allowed transitions between fine structure levels in the H atom up to n = 3
(term energies are not to scale). The thickness of the arrows indicates the strengths
of the individual transitions within the different multiplets (only to be compared
within one multiplet). All transitions between the doublet levels of the H atom (as
well as of the alkalis) lead to double spectral lines or to triple lines one of which is
typically very weak (see the 3d 2 D 3p 2 P transition). We point out here that the
relative intensities within a multiplet in emission differs from those in absorption.
For example, the two components of the first resonance line in emission, 2p 2 P1/2
1s 2 S1/2 and 2p 2 P3/2 1s 2 S1/2 occur with of probability, while the ratio of the
absorption cross sections for the inverse processes is 1 : 2.
6.4.2 Transitions Between Sublevels vs. Overall Transition

Probabilities
We recall that LS coupling implies that orbital angular momentum L and spin an-
gular momentum S combine to a total angular momentum J of the system. The
following considerations are independent of whether L and S refer to a single ac-
tive electron or represent several electrons which, in turn, are composed of orbital
and spin angular momenta from several active electrons. Our first task is to calculate
the line strengths S(Jb Ja ) in the (LS)J coupled scheme for transitions between in-
dividual fine structure levels with quantum numbers Ja and Jb . According to (4.112)
6.4 Selection Rules and Intensities of Transitions 311
Fig. 6.23 Selection rules for n =0 =1 =2

fine structure transitions 3d5/2
term energies
(emission) in the H atom 3p3/2
3s1/2
not to scale
3 3d3/2
3p1/2
2p3/2
2 2p1/2
2s1/2
1
1s1/2
it is proportional to the radial transition matrix element and the corresponding re-
duced matrix element of the renormalized spherical harmonics tensor C1 :
2
S(Jb Ja ) = (2Jb + 1)b |r|a Lb SJb C1 La SJa 2 . (6.69)
The ratios of transition probabilities within individual multiplets (as indicated by the
red arrow in Fig. 6.24) may be derived by further reduction of the reduced matrix
element in (6.69). Since the dipole operator er responsible for E1 transitions does
not act on the electron spin, S remains constant during the transition. In (6.69) this
is formally born out by C1 , a spherical tensor of rank one which only acts onto the
angular momentum part of the coupled scheme |La SJa . Hence, using (C.46) and
the 6j symbols described in detail in Appendix B.3 we may pull out the spin:
2
S(Jb Ja ) = b |r|a (6.70)
" #2
La Lb 1
(2Jb + 1)(2Ja + 1)(2Lb + 1) Lb C1 La 2 .
Jb Ja S
From this we may derive the spontaneous lifetime(s) of the upper level(s). First,
with (4.114) we obtain the individual spontaneous transition probability A(Ja Jb )
from an upper multiplet substate |Jb Mb into all substates |Ja Ma of one specific
lower Ja level. This probability is independent of Mb and Ma but still depends on
Jb and Ja . We then sum over all final Ja levels to obtain the total decay probability
of the level Jb :
43 S(Jb Ja )
A(Ja Jb ) = ba
(6.71)
3c2 2Jb + 1
Ja Ja
3
4ba
b |r|a 2 Lb C1 La 2
3c2
" #2
La L b 1
(2Ja + 1)(2Lb + 1) .
Jb Ja S
Ja
Jb b L b S
A(Lb La)
A(Jb Ja)
Ja a L a S
Fig. 6.24 Transitions between multiplet levels b Lb S and a La S, explaining the terminology for
transition probabilities between individual J sublevels and overall L levels
The sign in the second line accounts for the radial transition matrix elements not
being strictly independent of Ja and Jb and the angular frequencies ba differing
slightly for different multiplet components. Significant deviations occur only for
heavy atoms.4
We now exploit the 6j orthogonality relation (B.67) and obtain:
4ba3
A(Ja Jb ) = b |r|a 2 Lb C1 La 2 = A(La Lb ) = 1 (6.72)
3c 2
Ja
3
4ba S(Lb La )
A(La Lb ) = . (6.73)
3c 2Lb + 1
2
Note that (6.72) is not only independent of Mb but also of the initial Jb level: we
thus have found the overall decay probability A(La Lb ) of the upper b Lb S level
into all substates of the lower a La S level. It only depends on La and Lb and on the
radial matrix element b |r|a . Thus A(La Lb ) is equal to the inverse spontaneous
lifetime of the upper level (more precisely: of each substate |Jb Mb ) as far as no
other, competing decay channels exist.
With (6.73) we have introduced also the overall line strength S(Lb La ) for the
multiplet transition, analogously to (6.69). Obviously this relation is completely
equivalent to (4.114) where an uncoupled scheme was assumed. Only the initial
and final orbital angular quantum numbers Lb and La and the radial matrix element
determine the excited state lifetime: La Lb = is the lifetime of the atomic tran-
sition. Take e.g. a sodium atom in the first excited 3p 2 P1/2 or 3p 2 P3/2 resonance
state: both states decay into the 3s 2 S1/2 ground state and each of the 6 different
excited substates has the same lifetime in respect of spontaneous decay.
All above equations still contain the reduced matrix element of C1 . It may be fur-
ther evaluated by angular momentum algebra. Specifically, for quasi-one-electron
systems such as alkalis with S = s = 1/2 we put La = a and Lb = b , respectively,
4 The lifetimes of the 3 2 P1/2 and 3 2 P3/2 states in Na are reported as 16.30 and 16.25 ns, respec-
tively, while for Cs 14 % difference is observed with 34.7 ns for the 6 2 P1/2 level and 30.4 ns for
the 6 2 P3/2 level (see e.g. S TECK 2010, and references there). This warrants an improved theoret-
ical treatment with more accurate wave functions for each individual j state. These could e.g. be
derived by QDT as outlined in Sect. 3.2.6 with the quantum defect differences among FS levels as
presented in the last paragraphs of Sect. 6.3.3.
the angular momentum quantum numbers of the electron. The reduced matrix ele-
ment Lb C1 La can be evaluated explicitly with (C.30):
" #2
2 l lb 1
S(nb b sjb na a sja ) = nb b |r|na a (2jb + 1)(2ja + 1) a
jb ja 1/2
2
b 1 a
(2 b + 1)(2 a + 1) . (6.74)
0 0 0
The 6j symbol implies the triangular relation (ja jb 1) = 1, or equivalently the se-
lection rule j = 0, 1 (however jb = 0 ja = 0). The 3j symbol demands in
addition that a + 1 + b is even, i.e. that b = a 1 ensuring parity conservation.
Finally, with (B.53) the last line in (6.74) is simply written as
( b + a + 1)/2. (6.75)
In the more general case, where L is already a coupled orbital angular momentum
as typical for more complex atoms, one has again to apply the reduction scheme
described above onto that coupling scheme, do this step by step and exploit (C.30)
where possible.
To compute the radial matrix element nb b |r|na a one has to know the wave
functions in detail. Useful programmes are now readily available. One may then
obtain the spontaneous lifetime = Lb La from (6.72).
Once is known (either from theory or from experiment) it is often useful to
express all quantities in terms of it. Inserting (6.72) into (6.70) one obtains for the
line strength
" #2
3c2 2Lb + 1 La Lb 1
S(Jb Ja ) = 3
(2J b + 1)(2J a + 1) (6.76)
4ba Jb Ja S
by which in turn all other relations maybe expressed: e.g. the state to state sponta-
neous transition probabilities A(Ja Ma ; Jb Mb ) according to (4.113) as well as the
corresponding coefficients B(Ja Ma ; Jb Mb ) = B(Jb Mb ; Ja Ma ) for induced transi-
tion probabilities (4.124).
6.4.3 Some Useful Relations for Spectroscopic Practice
We communicate probabilities for transitions between the components of two mul-

tiplets, inserting (6.76) into the respective expressions from Sect. 4.6.5 The sponta-
neous emission probability from level Jb of a multiplet b Lb S to a specific level Ja
of a (lower lying) multiplet a La S summed over all polarizations, emission angles
5 We recall here that the direction of a transition is always read from right to left, e.g. B(Jb Ja )
implies a transition to Jb Ja from Ja .
and final projection quantum numbers Ma is obtained from (4.114) (independent of

the initial Mb quantum number):
3 " #2
4ba S(Jb Ja ) La Ja S 2Lb + 1
A(Ja Jb ) = = (2Ja + 1) . (6.77)
3c2 2Jb + 1 J b Lb 1
For induced transitions, emission or absorption, the corresponding probability from
level Jb of a multiplet b Lb to level Ja of a multiplet a La is obtained from (4.125)
or (4.126), respectively. Averaged over an isotropic population of initial Mb sub-
states and summed over all final Ma states (independent of the polarization em-
ployed) the B coefficient is
" #2
4 2 c S(Jb Ja ) 2 c3 La Ja S 2Lb + 1
B(Ja Jb ) = = (2Ja + 1) (6.78)
3 2Jb + 1 ba3 J b Lb 1
2 c3
= A(Ja Jb ), (6.79)
ba
3
and for the inverse process (absorption or emission)

" #2
4 2 c S(Jb Ja ) 2 c3 La Ja S 2Lb + 1
B(Jb Ja ) = = (2Jb + 1)
3 2Ja + 1 ba
3 Jb Lb 1
2Jb + 1
= B(Ja Jb ). (6.80)
2Ja + 1
One sees that the relative intensities (branching ratios) to different final Ja or
Jb levels depend on the degeneracies of the latter and on the squared re-coupling
coefficients (6j symbols) but not on M or on the polarization.
We have already used the 6j orthogonality relation (B.67) to show that the spon-
taneous lifetime does not depend on the initial total angular momentum Jb . In the
same way one may prove a corresponding relation for the induced processes: from
(6.78) and (6.80) one finds that induced emission and absorption coefficients are
independent of the initial Ja or Jb , respectively, if the final states are not resolved:
2 c3 1 2 c3
B(La Lb ) = B(Ja Jb ) = = A(La Lb ) (6.81)
Ja
ba3 ba
3
2 c3 2Lb + 1 1
B(Lb La ) = B(Jb Ja ) = 3 2L + 1
(6.82)
Jb
ba a
2Lb + 1
= B(La Lb ). (6.83)
2La + 1
Obviously, the familiar E INSTEIN relations (4.127) are valid for the individual
fine structure components (6.77)(6.80) with j := J , as well as for the multiplet
averaged quantities (6.81)(6.83) where j := L.
Table 6.3 Branching ratios in 2 S 2 P and 2 P 2 D transitions calculated by (6.80) and (6.78).
The relative intensities can only be compared within one multiplet. The sum over all transitions
starting from one Ja or Jb are normalized to unity
to from to from
2S 2P 2P 2P 2P 2D 2D
1/2 1/2 3/2 1/2 3/2 3/2 5/2
2S
1/2 1 1 2P
1/2 5/6
2P
1/2 1/3 2P
3/2 1/6 1
2P
3/2 2/3 2D
3/2 1 1/10

1 1 1 2D
5/2 9/10

1 1 1 1
In the older literature, branching ratios within multiplets were often presented
in extended sets of tables or derived by equating different sums from (6.81) and
(6.83). Today these relative line strengths and transition probabilities are evaluated
readily using the 6j symbols (computed e.g. with S VEN G ATO R EDSUN 2004).
Hence, we only communicate by way of example in Table 6.3 the branching ratios
for 2 S 2 P and 2 P 2 D multiplets as relevant e.g. for H and the alkali atoms. It
is interesting to note that for a 2 S 2 P transition the absorption probability for the
1/2 P3/2 is twice that for the S1/2 2P1/2 transition. In contrast, the inverse
2S 2 2
processes P1/2 S1/2 and P3/2 2 S1/2 have identical transition probabilities,
2 2 2
as already mentioned in the context of Fig. 6.23.
Section summary
Selection rules for transitions between fine structure levels follow the general
concepts derived in Sect. 4.4 from angular momentum conservation now
applied to the total angular momentum quantum number J and its projection
M onto a given axis.
E1 transitions are allowed for J = 0, 1 (but 0 0). For linearly and cir-
cularly polarized light M = 0 and 1 must hold, with M referring to the
directions of polarization and light propagation, respectively. In addition, par-
ity of the whole system (photon plus atom) has to be conserved which for
single electron systems requires = 1.
Transition probabilities between individual J and J levels can be derived
from line strengths in the usual manner, involving the radial matrix element
between initial and final n and n states and the reduced matrix element
Jb C1 Ja in the coupled system. Using angular momentum algebra the lat-
ter may be reduced to expressions which refer to the initial and final orbital
angular momenta L and L .
Spontaneous emission probability from one specific initial substate |J M of
level nL to all substates |J M of the final level n L is independent of J
(a) (b)
(e) (a) (f) (d) (c) (b) (g)
crossover
LAMB shift (a) (b) (c) (d) (e) (f) (g)

1057.846 MHz 0 10 / GHz
Fig. 6.25 Fine structure and L AMB shift for the H atom. (a) Term scheme and transitions of the
H BALMER line between the n = 3 and n = 2 levels (energies not to scale). (b) Measurement
by D OPPLER free saturation spectroscopy and original assignment according to H NSCH et al.
(1974); positions and heights of the stick spectrum (vertical full red lines) give the most recent
theoretical data according to K RAMIDA (2010)
and M. It only depends on L and L and on the corresponding radial matrix

element. Thus, it is the inverse of the overall lifetime of the upper level.
Also, for induced emission and absorption, when averaging over all (equally
populated) initial M levels, the transition probabilities are independent of the
initial J and M and of the polarization.
6.5 L AMB Shift
6.5.1 Fine Structure and L AMB Shift for the H Line
In Sect. 6.1 we have already familiarized ourselves with optical methods for observ-
ing E1 transitions between the n = 2 and n = 3 levels in the H atom with high pre-
cision. Figure 6.25 shows (a) the relevant term scheme and reproduces in (b) again
the well resolved spectrum from Fig. 6.13. The assignment of the spectral lines to
transitions is indicated by lower case letters. For a quantitative comparison of line
intensities one would use (6.74) with the radial dipole matrix elements from Ap-
pendix C.5. However, one important feature shown in Fig. 6.25 cannot be explained
by the FS theory discussed so far: We remember that according to (6.65), the en-
ergy terms derived from D IRAC theory depend only on j but not on . This rule
is obviously broken for the n = 2 levels states in H as revealed by the experiment,
the 2p 2 S1/2 being slightly higher in energy than the 2p 2 P1/2 level: This important
observation is called L AMB shift.
6.5 L AMB Shift 317
6.5.2 Microwave and RF Transitions D OPPLER Free
The main enemy for optical precision spectroscopy of the H atom, with its small
mass M and high average thermal velocity v , is the D OPPLER effect which
according to (5.22) also depends on the transition frequency ba :
D ba v /c = ba v .
If one tries to determine the FS splitting or other small effects from the difference
of two optical transitions D OPPLER broadening enters proportionally to ba with
dramatic influence on the result. However, in some cases it is possible to access the
energetic difference between different |j mj states within one multiplet directly,
typically by M1 transitions as described in Sect. 5.4. The necessary alternating mag-
netic fields can be generated without problems in microwave resonators.
While the corresponding optical E1 transitions are observed in the VIS or UV re-
gion (at wavenumbers between (10 000 to 100 000) cm1 ) the FS splittings amount
to some cm1 or even below. D OPPLER broadening can be neglected completely
in this microwave regime (1 cm1 = 30 GHz) and one may determine transition fre-
quencies with orders of magnitude higher precision than by typical optical spec-
troscopy and even finer effects may be measured with radio frequency transitions
(RF) at frequencies from 1 MHz to 1 GHz, depending on the external magnetic field
which one usually applies in addition.
6.5.3 Experiment of L AMB and R ETHERFORD
The most prominent, and in its consequences far reaching experiment was the first
determination of the L AMB shift for which L AMB and K USCH (1955) obtained the
N OBEL prize.
In 1947, Willis E. L AMB and R. C. R ETHERFORD were looking for a transition
between the 2s 2 S1/2 and 2p 2 P1/2 level which according to D IRAC should be de-
generate. Thereby they exploited the fact that the 2p 2 P1/2 state decays directly into
the ground state ( = 1.6 ns) while the 2s 2 S1/2 state is metastable. It may only de-
cay by two-photon emission with a lifetime of 120 ms as we have learned in
Sect. 5.3.2.
The experimental scheme, sketched in Fig. 6.26, is interesting. A atomic beam
(see footnote 24, Chap. 1) of hydrogen emerges from a hot oven where H2 disso-
ciates. By electron impact a certain fraction of these H atoms is excited into the
2s 2 S1/2 and 2p 2 P levels. While the 2p 2 P1/2,3/2 atoms rapidly decay by sponta-
neous emission into the ground state, the metastable 2s 2 S1/2 atoms in the atomic
beam reach a microwave resonator. There, the microwaves if on resonance in-
duce the transition 2s 2 S1/2 2p 2 P1/2,3/2 . If one of these transitions occurs, the
2p 2 P1/2,3/2 states decay rapidly into the ground state. One detects the transition
by the loss of metastable atoms reaching the detector. The L ANGMUIR -TAYLOR
detector (see Sect. 1.9.3) is sensitive only to excited (2s 2 S1/2 ) atoms.
magnetic field B (variable) LANGMUIR

extrapolation to B = 0 TAYLOR
detector
electron beam (hot W metal sheet)
ground state excited
H atoms H atoms
oven
1s 2S1/2 2s 2S1/2
H+
Ly is emitted,
all atoms in
np states decay tuneable microwave
resonator
1...10 GHz
Fig. 6.26 Scheme of the experimental setup of L AMB and R ETHERFORD for measuring mi-
crowave induced transitions from the 2s 2 S1/2 state of the H atom
transition frequency / GHz
16
QED
DIRAC theory 2s 2 S1/2 (m j = 1/2) 2 p 2 P3/2 (m j = 3/2)
12
2s 2 S1/2 (m j = 1/2) 2 p P3/2 (m j = 1/2)

8
2s 2 S1/2 (m j = 1/2) 2 p P1/2 (m j = -1/2)
4 2s 2 S1/2 (m j = 1/2) 2 p P1/2 (m j = 1/2)

2s 2 S1/2 (m j = 1/2) 2 p P3/2 (m j = -1/2)
}
~ 1060 MHz
0 0.1 0.2 0.3 B / T
= W (2s 2 S1/2 ) - W (2p 2 P1/2 )
Fig. 6.27 Results of the original L AMB shift experiment for atomic hydrogen. Transition fre-
quencies between the 2s 2 S1/2 and the 2 p 2 P1/2 and 2p 2 P3/2 levels as a function of an external
magnetic field B. Dashed lines: D IRAC theory; full lines: fits to the experiment and QED. The
remaining energy difference between the 2s 2 S1/2 and 2p 2 P1/2 levels at B = 0 (ca. 1060 MHz) is
called L AMB shift
The experiment was carried out at several fixed microwave frequencies and the
splitting of the two states was registered by tuning this magnetic field so that res-
onant transitions occurred. In Fig. 6.27 the results are compared with D IRAC the-
ory and quantum electrodynamics (QED). In the limit of vanishing magnetic field
a finite value of 1060 MHz ( = 0.0353 cm1 ) for the resonance frequency of the
2 S1/2 2 P1/2 transition remains the L AMB shift, by which the 2 2 S1/2 level
2 2
lies higher than the 2 2 P1/2 level. In contrast, D IRAC theory predicts exact agree-
ment of both energies.
6.5 L AMB Shift 319
hydrogen
cryostat
vacuum chamber cooling Cesium
atoms
FARADAY cage
laser clock
chopper
2S
243 nm time resolved
detector
2F photon counting
microwave
F
dye laser 486 nm resonator
probe laser fluorescence
detector
4/7 F 1/2 F
cooling lasers
frequency comb
9.2 GHz
70 fs Ti:Sapphire
laser from Cs clock
cooling lasers
Fig. 6.28 Schematic of an advanced experiment for determining the 1S2S transition frequency
in the H atom by D OPPLER free two-photon spectroscopy according to N IERING et al. (2000). The
setup in the vacuum chamber corresponds essentially to Fig. 6.16. The frequency comb is the heart
of the experiment, providing a high precision frequency standard which is directly synchronized
with an atomic Cesium fountain clock, sketched in the spirit of J EFFERTS and M EEKHOF (2011)
6.5.4 Precision Spectroscopy of the H Atom
In spite of these apparent advantages of microwave spectroscopy, optical methods

have surpassed them by far essentially on the basis of D OPPLER free two-photon
spectroscopy as described in Sect. 6.1.8. Today, modern laser spectroscopy allows
one literally to count the frequency of optical transitions with a never dreamed of
precision. The pioneering work of Ted H NSCH and his collaborators who devoted
many years and a wealth of brilliant ideas on the spectroscopy of atomic hydro-
gen has led to an amazing accuracy. Jointly with John H ALL, Ted H NSCH (2005)
received the N OBEL prize in physics (shared with Roy G LAUBER) for the devel-
opment of laser-based precision spectroscopy, including the optical frequency comb
technique.
We shall discuss this fascinating experimental technique in some detail in
Sect. 1.4.3, Vol. 2. Briefly, one uses trains of extremely short, precisely clocked
light pulses. Their repetition frequency is directly compared or synchronized with a
time standard, e.g. with a Cs atomic clock (see e.g. N IERING et al. 2000). Instead of
measuring the wavelengths of optical transitions by essentially analogue techniques
as done by spectroscopists for the past two hundred years, one now compares the
frequencies of the electromagnetic waves inducing the optical transitions in the VIS
and UV spectral range directly to the comb frequencies. Thus, counting an essen-
tially digital technique is employed ensuring orders of magnitude higher accuracy.
Just to indicate the complexity and sophistication of such experiments, Fig. 6.28
reproduces roughly the scheme which was originally used by N IERING et al. (2000)
(for recent improvements see e.g. J ENTSCHURA et al. 2011; PARTHEY et al. 2011;
M ATVEEV et al. 2013, the latter being a simultaneous effort of two institutes via a
920 km fiber link). Part of it, the two-photon excitation of an atomic H beam in a
confocal resonator, has already been described in Sect. 6.1.8. A so called Cesium
fountain atomic clock is used, which is currently employed world wide as primary
time and frequency standard. It is sketched in Fig. 6.28 on the right. A compact
description is found e.g. on the NIST Web-pages (J EFFERTS and M EEKHOF 2011).
The special trick with this setup is the long effective interaction time T 1 s of
the atoms with the microwave field: one exploits the method of R AMSEY fringes as
described in Sect. 6.1.7.
To optimize this concept the Cs atoms are first cooled with 6 diode lasers. By a
brief detuning pulse applied to the bottom laser one kicks the atoms upwards,
pushes them so to say through the microwave resonator. The grey trajectory in
Fig. 6.28 shows schematically the path which the atoms take. Under earth gravity
they move freely up to a classical turning point and then fall back into the resonator
indeed some kind of an atom fountain. One induces the ground state hyperfine
transition in atomic Cs which is detected from a change of the fluorescence signal
induced by the probe laser.6 As explained in Sect. 6.1.7 the total interaction time
T in the frame work of the R AMSEY fringe method is given by the time of flight
of the atoms from the resonator to the turning point and back to the resonator. The
precision of the reference line is then 1/T , typically on the order of 1 Hz.
Based on this type of optical measurements the spectroscopy of the H and D
atom is today known with extreme precision. Specifically, the n = 2 L AMB shift
W (2s 2 S1/2 ) W (2p 2 P1/2 ) corresponds to
for the H atom = 1057.847(9) MHz and (6.84)

for the D atom = 1059.28(6) MHz.
Figure 6.29 summarizes the orders of magnitude of the different contributions to

the n = 2 L AMB shift in the H atom. We add here: predicted by QED calculations
but not seen in Fig. 6.29 is a tiny energy rise of the 22 P3/2 state by 0.00037 cm1
while the 22 P1/2 state is lowered by 0.000479 cm1 . The physical origin of the
L AMB shift will be discussed in Sect. 6.5.6 where we shall also understand that
these very small shifts are characteristic for P states.
A particular challenge, for experiment and theory alike, is the L AMB shift of the
1s S1/2 ground state of hydrogen. In principle one may extract it from a very pre-
cise measurement of the 1S1/2 2S1/2 transition frequency and D IRAC theory. One
problem is to avoid uncertainties in the fundamental constants needed to compute
an exact value of the D IRAC prediction. One of the first ultra high precision mea-
surements which circumvents this problem was reported by W EITZ et al. (1994).
Without going into the sophisticated details of such an experiment, we just men-
tion that the 1S1/2 2S1/2 frequency is compared with 4 that for the 2S 4S/4D
6 Werecall that 1 s is defined as the duration of 9 192 631 770 periods of the transition frequency
between the two hyperfine levels of the ground state in 133 Cs.
6.5 L AMB Shift 321
Fig. 6.29 Overview of the SCHRDINGER
wavenumber / cm-1
term positions of the n = 2 BOHR DIRAC QED
states for the H atom in 0
S CHRDINGER, D IRAC and =1 =1 2P
j = 3/2 3/2
QED approximation; term
distances are shown to scale 0.365 cm-1
and may be compared directly
=0 2S
1/2
j = 1/2 2P
0.5 = 0, 1 =1 1/2
0.035 cm-1
transitions thus deriving simultaneously an ultra precise value for the RYDBERG
constant: 7
R = 10 973 731.568539(55) m1 (6.85)

R c = 3.289841960364(17) 1015 Hz. (6.86)
Using this value of R , numerical values for the hydrogen 1S 2S transition fre-
quencies from B OHR /S CHRDINGER (1.149) or D IRAC theory (6.65) are obtained
with appropriate accuracy:
3R c
S CHRDINGER : (1S 2S) = = 2.467381470272 1015 Hz,
4
D IRAC : (1S1/2 2S1/2 ) = 2.467411580803 1015 Hz . (6.87)
The very recently confirmed (M ATVEEV et al. 2013) experimental value is
(1S1/2,F =1 2S1/2,F =1 ) = 2.466061413187018(11) 1015 Hz, (6.88)
measured with a relative accuracy of 4.5 1015 ! To compare this value measured
for atomic H with the prediction from D IRAC theory (6.87) one has to correct it by
the kinematic correction factor (1.146). With the present CODATA value for me /mp
one obtains
(1S1/2 2S1/2 ) me / = 2.46740447220939(56) 1015 Hz . (6.89)
A distinct difference remains on the one hand due to hyperfine interaction (in
(6.88) indicated by the subscripts F = 1), and to the 1S1/2 L AMB shift on the other.
A quantitative evaluation of HFS is relatively straight forward and experimentally
well under control by measurements of the corresponding level splittings (we shall
treat HFS in Chap. 9). In fact, the HFS splitting of the H 1sS1/2 ground state was for
many years one of the best determined transition frequencies until the (1S1/2
7 2010 CODATA values, M OHR et al. (2012), see also Appendix A.

2S1/2 ) transition acquired this merit. After all this is done one obtains the ground
state L AMB shift W (1s 2 S1/2 ) W (1s 2 S1/2 D IRAC)
for the H atom = 8172.840(22) MHz and (6.90)

for the D atom = 8183.970(22) MHz,
the presently agreed global average from several independent measurements. Note
that this is again an upward shift of the D IRAC term level towards a slightly smaller
binding energy just as for the 2s 2 S1/2 level.
We point out that the whole field of ultra high precision spectroscopy and the
interpretation of its results by QED is still rapidly evolving. Several not fully re-
solved issues and discrepancies remain. The interested reader is referred to the liter-
ature for details. Unless otherwise specified, numerical values given here reflect the
state-of-the-art according to some recent, extensive reviews by E IDES et al. (2001),
K ARSHENBOIM (2005), K RAMIDA (2010).
The critical reader may perhaps ask now, what this extreme precision might be
good for. Let us quote from the master of precision spectroscopy, Ted H NSCH et
al. (2005): Even beyond more precise measurements of fundamental constants and
searches for their possible time variations, there are many good practical reasons
for pushing the art of measuring time and frequency to the feasible limits. Advances
in time and frequency metrology will enable the synchronization of clocks over large
distances. Such synchronization is, e.g. needed in astronomy for very long baseline
interferometry. Better clocks will improve the performance of satellite navigation
systems. They are also crucial for the precise tracking of remote space probes. In
telecommunications, better atomic clocks will be needed for network synchroniza-
tion, as the need for bandwidth increases. Geologists can apply better clocks to
study the variability of earths rotation or to follow the drifts of continents with
millimeter precision. Astronomers can study irregularities in the periods of pulsars.
For physicists interested in fundamental science, better clocks will permit new strin-
gent tests of special and general relativity. Very likely, there will be new unexpected
discoveries, as the art of precision spectroscopy and optical frequency metrology
continues to advance.
And one may add in our present context, that progress in basic research has
always been nourished by quantum jumps in precision. The L AMB shift itself is an
excellent example: its discovery has paved the way for quantum electrodynamics
(QED), today one of the cornerstones of modern physics.
6.5.5 LAMB Shift in Highly Charged Ions
Atomic physics with highly charged ions (HCI) is a very active and productive
branch of modern physics. Advanced high energy ion storage rings and state-of-the-
art electron beam ion traps (EBIT), accessible as user facilities, offer challenging
perspectives for a broad range of interesting fundamental science (see also Sect. 7.5,
6.5 L AMB Shift 323
Vol. 2) as well as for many technological, biological and medical applications. Criti-
cal tests of QED and of atomic-nuclear interactions with high Z ions are one exciting
aspect of the spectroscopy of such systems (a compact review is given by B EIERS -
DORFER 2010). As we shall discuss in Sect. 6.5.6, the relevant coupling constant for
QED is also Z, and so called higher order loops (nth order perturbation theory)
enter into the term energies with (Z)n Thus QED may have a problem with highly
charged ions where Z is no longer really small, the fine structure constant being
1/137.
In the past, theory and experiment have mostly focussed on the 1S1/2 L AMB
shift of hydrogen like ions. In singly charged He+ the experimental value for the
2 2 S1/2 2 2 P1/2 transition is 14 041.13(17) MHz in excellent agreement with theory,
giving 14 041.18(13) MHz. The main contribution from QED are the one-loop ef-
fects, self-energy and vacuum polarization. Effects from the finite size of the nuclei
are included in comparisons of theory and experiment, and become equally impor-
tant for larger nuclei. We shall discuss volume effects of this kind in Sect. 9.4 in the
context of HFS.
In total, one expects for hydrogen like, (Z 1)-fold charged ions and ns levels a
L AMB shift (see e.g. J OHNSON and S OFF 1985)
(Z)4 Z42
WLamb = F (Z)me c 2
= F (Z)Eh , (6.91)
n3 n3
given in terms of the electron rest mass energy and a.u., respectively. F (Z) is a
slowly varying function of Z, but also depends slightly on n. For the neutral H
atom F (1) 10 as reflected in Fig. 6.29 for = 2. Inverting the experimental
results, (6.84) and (6.90), we obtain F (1) = 10.042037(27) and 10.39828(9) for
the 1S1/2 and 2S1/2 L AMB shift, respectively.
Note the proportionality WLamb Z 4 2 /n3 (in a.u.) which we have found to
be characteristic also for fine structure (6.65). The additional factors / F (Z)
make the L AMB shift about an order of magnitude smaller than FS.
The overall trend of the L AMB shift is illustrated over the whole nuclear charge
range from Z = 1 . . . 91 in Fig. 6.30(a). The dominating one-loop contributions and
finite nuclear size effects are plotted together with the total theoretical value (red
line) in units of (/)[(Z)4 /n3 ]me c2 on a logarithmic scale. The vacuum polar-
ization gives a negative contribution.
Experimentally, spectroscopy in the hard X-ray regime in storage rings has pro-
vided impressive results for multiply charged, hydrogen like ions up to U91+ . Ele-
gant methods have been employed in these studies, as sketched in Sect. 6.1.4. The
1S1/2 L AMB shift is typically determined by comparing the 1S1/2 2P1/2 and
1S1/2 2P3/2 transitions with D IRAC theory (making appropriate corrections for
HFS and kinematics). Figure 6.30(b) according to S THLKER (see G UMBERIDZE
et al. 2005) compares the theory presented in (a) with these results from several
sources on a linear scale. One may describe the agreement as quite satisfactory.
However, so far truly high precision spectroscopy for HCI is not yet in sight, at
least not one competing with the extreme accuracy of the 1s 2s transition in the
neutral H atom. Thus, the best test for two-loops QED contributions would presently
10
LAMB shift 10
self-energy
1 8
- vacuum
polarization
6
0.1
(a) 4
(b)
0.01 finite size
of the nucleus 2
0.001
0 20 40 60 80 100 0 20 40 60 80 100
nuclear charge Z nuclear charge Z
Fig. 6.30 1S1/2 L AMB shift for H like ions as a function of nuclear charge Z. (a) Main contri-
butions to the L AMB shift according to J OHNSON and S OFF (1985) from so called one-loop QED
effects (to be discussed in Sect. 6.5.6) and from the finite nuclear size. (b) Comparison of this
theory with experiments according to G UMBERIDZE et al. (2005)
still be the L AMB shift of the H 1S1/2 ground state (known to be ca. 4 parts in 105 )
were it is not hampered by the experimental uncertainties of the proton radius!
On the other hand, for 91 fold ionized Uranium, U91+ , the 1S1/2 L AMB shift
amounts to (460 4.5) eV by no means a tiny effect as in H atoms. And consid-
ering the difficulties of this experiment, the accuracy is impressive but with about
1 % by far not yet good enough for a test of two-loop contributions.
Recent theoretical and experimental advances appear to offer promising perspec-
tives for a direct comparison of the 2s 2 S1/2 2s 2 P1/2 L AMB shift in lithium like
highly charged ions (B EIERSDORFER 2010). A number of interesting questions also
remain to be clarified in respect of HFS in highly charged ions. Other fascinating
possibilities and experimentally demanding challenges arise in this context from
the advent of X-ray lasers, e.g. at XFEL P ROJECT (2011), for which theoretical
predictions are made e.g. for laser induced fluorescence in the strong field regime
(P OSTAVARU et al. 2011). In summary, the field remains open, highly active and
exciting.
6.5.6 QED and F EYNMAN Diagrams
What is the physical origin of the L AMB shift? Very generally speaking: the inter-
action of the electron with the vacuum field. The fundamental theory for describing
such processes, Quantum Electrodynamics QED and Quantum Field Theory, was
developed by T OMONAGA, S CHWINGER and F EYNMAN. In 1965 they received the
N OBEL prize for their ground breaking work, which is one of the cornerstones of
modern theoretical physics including the standard model i.e. far beyond its origin
and the applications in atomic physics which we discuss here.
However, even a brief introduction into QED would lead us far beyond the scope
of the present textbook. Thus we confine ourselves here to a heuristic presenta-
6.5 L AMB Shift 325
Table 6.4 Most simple Graph Description

F EYNMAN graphs for e , e+
and as well as for some 1 Free electron e
elementary electromagnetic 2 Free positron e+
interactions 3 Free photon
4 Electron emits photon
5 Electron absorbs photon
6 Positron emits photon
7 Positron absorbs photon
8 -annihilation by e e+ pair generation
9 e e+ -pair annihilation by generation
10 Self energy of the electron (one-loop)
11 Vacuum polarization (one-loop)
tion of F EYNMAN diagrams (or F EYNMAN graphs). They first appeared in Richard
F EYNMANs famous 1949 paper on Space-Time Approach to Quantum Electrody-
namics as a picturesque visualization representing the mathematical framework for
describing the interactions of electrons, photons and other subatomic particles. They
are a compact form of symbolizing perturbation integrals and will that is our hope
here help us to understand what causes the L AMB shift and other related phenom-
ena. QED expands interactions into a series of perturbation integrals accounting for
the influence of the quantized electromagnetic field. These perturbation series are
constructed for sequences of virtual or real events and warrant good book keeping
to account for all possible processes which might occur and this is done with the
help of the F EYNMAN graphs. Each node (or vertex) in such a graph characterizes
one individual interaction process, several subsequent nodes imply a higher order of
perturbation theory. One speaks about one-loop, two-loop processes and so on.
The movements of individual particles and their elementary interaction in space
and time are represented by graphical symbols which are summarized in Table 6.4.
The free electron (1) moves here from the left to the right, its antiparticle, the
positron, (2) moves from right to left. The exchange particle, for electromagnetic
interaction, the photon, (3) is characterized by a wiggly line. Graphs (49) are self
explaining, graph (10) represents the typical interaction of an electron with the elec-
tromagnetic magnetic radiation in vacuum: a free electron moving in space con-
stantly emits (virtual) photons which it quickly reabsorbs again. This is one of many
processes which contribute to the self energy of the electron. It plays a key role for
the so called mass re-normalization by which QED overcomes problematic diver-
gences. Another important process is vacuum polarization (11). In 1st order (one-
loop) it generates one (virtual) e e+ pair, which is then immediately annihilated
again. Graph (11) may obviously be constructed from (8) and (9).
We have somewhat loosely used the terms quickly and immediately which
may be quantified by the energy-time uncertainty relation (1.125): the larger the
energy of the emitted virtual photon, the shorter the time between emission and
reabsorption, and similarly for pair formation and annihilation.
From these basic diagrams one may construct more complex ones to describe
various interaction processes between electrons and photons. The following rules
have to be observed:
Energy and momentum conservation holds for each vertex.

Lines which leave the diagram
are real particles. For all of them together the
energy conservation law Wi = const must hold, with the particle energies in
relativistic form: Wi2 = pi2 c2 + m2i c4 .
Lines which connect vertices represent virtual particles. For them, energy con-
servation does only need to hold within the uncertainty relation. However, these
virtual particles cannot be observed!
Mathematically, vertices are represented by coupling constants, virtual particles
by so called propagators. The fine structure constant = Eh /(me c2 ), as de-
fined in (6.103), is the coupling constant for electromagnetic interaction, spec-
ifying the interaction between electron and photon. Theinteraction amplitude
between an electron and a Z-fold charged particle is Z .
In general, the propagator for bosons is f (q) 1/(|q|2 + m2boson ) with q being
the momentum of the boson. Specifically for a photon f (q) 1/|q|2 .
The total amplitude of a graph is the product of all coupling constants, and prop-
agators involved. Scattering cross sections are proportional to the square of the
amplitude.
In Table 6.5 F EYNMAN graphs are collected for several important and instructive
examples of interactions. The graphs are more or less self explaining. We refrain
here from a detailed description of the mathematical expressions represented by
these graphs. Only for C OULOMB scattering we have indicated, how one may glean
the underlying concepts: the well known RUTHERFORD scattering formula emerges
in this case without any effort.
6.5.7 On the Theory of the L AMB Shift
According to QED the electromagnetic field has a finite amplitude, even in its vac-
uum ground state, determined by the zero point energy of the corresponding quan-
tized harmonic oscillator. And the electron constantly emits and reabsorbs photons
as indicated by the self energy graph in Fig. 6.31(a). Consequently, a force pulls
6.5 L AMB Shift 327
Table 6.5 Lowest order F EYNMAN graphs for several important interaction processes in atomic
physics
Example Process Graph
1 C OULOMB scatter- e- e-
ing e + Ze Z
prop 1/q 2
aga
tor
+Ze
+Ze

The scattering amplitude at the two vertices is and Z , respectively.
The
virtual photon receives from a 2recoil momentum q = 2me W sin /2.
the electron
The total amplitude is thus Z /q . Hence, the scattering cross section is:
d Z2 2 Z 2 e4
4
. (6.92)
d" q W sin4 /2
2
This is the well known RUTHERFORDscattering formula.

2 e e+ scattering e- e-
e- e-
+
e+ e+ e+ e+
In addition to the C OULOMB term (as above), only two one-loop graphs are relevant:
the annihilation and re-generation of an e e+ pair by emission of a virtual photon.
3 C OMPTON scattering

+
e- e- e- e-
4 Pair production e+ e e-
e+
+Ze
The black cross symbolizes the atomic nucleus with which the electron interacts. The
following diagrams also need the interaction with an atomic nucleus to allow for en-
ergy and momentum conservation.
5 Bremsstrahlung:
e + Ze e + Ze + e-
e- + e-
+Ze e- +Ze
6 Electron bound in a C OULOMB potential. In the current literature the so called
D IRAC -C OULOMB propagator is summarized by just two parallel lines, compris-
ing several contributions
e + Ze = + + + ...
the electron back and forth into a Zitterbewegung (trembling motion), so that ef-
fectively the electron is smeared out over a (still very small) finite volume instead
of being point like. Due to this apparent charge distribution the electron sees
(a) (b)
self-energy
vacuum
polarization
Fig. 6.31 F EYNMAN graphs of (a) one-loop and (b) two-loop radiative corrections for the L AMB
shift. Graphs adapted from B EIERSDORFER (2010)
effectively a potential which close to the origin is slightly lifted up in respect

of the Z/R C OULOMB potential. And this in turn rises the energies of atomic
states with wave functions that have a finite probability at the origin. This is mainly
relevant for s electrons. And indeed the experimentally observed L AMB shift (6.84)
(illustrated in Fig. 6.29) and (6.90) is positive and significant only for S states. QED
can give reasonable estimates for the radius of this trembling motion and has pro-
found tools to calculate WLamb in terms of a perturbation series in principle
with arbitrary accuracy, if all parameters are known with sufficient accuracy and
computing power is not limited.
As we cannot enter here into the depths of QED we turn the question around and
try to glean an estimate for the Zitter radius rL from the measured L AMB shift.
To keep it simple, we assume the electron to see for r rL a constant potential
Z/rL , instead of Z/r (in a.u.). In 1st order perturbation theory we thus obtain
an estimate for the L AMB shift from the expectation value (3.41) of the perturbation
U (r) = 1/rL (1/r) which is the difference of the smeared out potential and
the original pure C OULOMB potential. Since rL is expected to be very small one
can pull out the radial wave function at the origin and obtains for a state n m (all
quantities in a.u.):
Z Z
WLamb = n | |n (6.93)
r rL
2
rL 1
1 2 2 2

= n m (0) 4Z r dr = n m (0) ZrL2 .
0 r rL 3
According to (2.127) for the H atom |n00 (0)|2 = Z 3 /(n3 ) so that

2 Z4 2
WLamb = r . (6.94)
3 n3 L
This clear result already explains the difference between ground state (n = 1) and
first excited state (n = 2) L AMB shifts in atomic hydrogen by about a factor of 8
(see (6.84) and (6.90), respectively). Comparing (6.94) for n = 2 with h from
(6.84) gives an estimate for rL 1.39 103 a0 3.0 102 C = 7.36 1014 m,
6.5 L AMB Shift 329
with the C OMPTON wavelength C of the electron (1.76). QED provides similar es-
timates (see e.g. E IDES et al. 2001, Eq. (28))
2 2 2 3
r = ln (Z)2 2C = ln (Z)2 a02 . (6.95)

This corresponds to r 2 3.4 102 C for the characteristic effective spread
of the electron which obviously is below the C OMPTON wavelength but signifi-
cantly above the proton radius (1015 m). With (6.95) one would obtain a L AMB
shift (in a.u.)
Z4 2 4
WLamb = 3
ln (Z)2 .
n 3
If we interpret 4/3 ln[(Z)2 ] as F (Z), one may compare this directly to the gen-
eral formula (6.91) with numerical values 13 and 10, respectively which is not too
bad for such a rough estimate.
As a historical remark we point out that Hans B ETHE (1947), in his first calcula-
tion of the H 2S1/2 L AMB shift, obtained a very similar expression when calculating
the self energy of the bound electron. Instead of 1/ 2 = me c2 /Eh as argument in the
logarithm he obtained me c2 /(8.9Eh ) which led him to 1040 MHz already pretty
close to the experimental value (6.84). Today, of course, with several orders of mag-
nitude better experimental accuracy and highly developed computational methods
in QED it is obvious that in addition to the self-energy also the vacuum polarization
has to be taken into account (see Fig. 6.31(a)) as well as higher order corrections.
The curious reader may admire the diagrams for the relevant two-loop corrections
in Fig. 6.31(b). Also a range of additional effects have to be considered.
For the hydrogen atom in the n = 1 and 2 levels the present state-of-the-art is
summarized in Table 6.6. We cannot enter into a deeper discussion which is ongo-
ing among the experts. However, the agreement between theory and experiment is
already now truly impressive.
At the moment, the largest uncertainty which limits stringent tests of bound state
QED still appears to be the proton radius. Its value rp = 0.8775(51) fm in 2010 CO-
DATA (NIST 2010) already strongly depends on optical precision spectroscopy. In-
dependent measurements from nuclear physics give different values and have much
larger error bars. This is a startling situation, the proton being one of the most im-
portant building blocks of matter and its size so poorly known!
The most recent and fascinating development in this context is a measurement
of the L AMB shift in muonic hydrogen by P OHL et al. (2010). The muon (see
Sect. 1.5.2) is the bigger brother of the electron, its mass being more than 200
times larger, with otherwise more or less identical properties. Hence, in muonic
hydrogen, p, all the physics of the H atom remains in principle unchanged ex-
cept for the magnitude of the measurable quantities. In particular, the B OHR radius
is 200 times smaller in p and thus the muons probability to be close to the pro-
ton is correspondingly higher. Consequently, the L AMB shift is influenced much
more dramatically by the protons structure as in the case of atomic hydrogen. The
Table 6.6 Main QED contributions to the L AMB for the 1S, 2S and 2P states of atomic hydrogen,
and comparison with the presently best experimental values
Contribution 1s 1/2 2s 1/2 2p 1/2 2p 1/2 2p 3/2
MHz MHz MHz MHz
Self energy (one-loop) 8383.339466(83)a 1072.958444f 12.84692(2)e 12.54795(2)e
Vacuum pol. (one-loop) 214.816607(15)a 26.852075(2)g 0.00035e 0.00008e
Two-loop corrections 0.7310(33)a 0.02598(7)e 0.01279(7)e
Proton sizea 1.253(50)a 0.1566(62)g 0 0
Recoil correctiona 2.401782(10)a 0.0145e 0.0177e
Radiative recoil corr.a 0.0123(7)a
Other corrections 0.002(1)a 0.0002e 0.0001e
Recent theory (total) 8172.894(51)a 1057.844(2.5)d 12.83599(8)e 12.51746(8)e
Experiment 8172.840(22)b 1057.847(9)c
a B IRABEN (2009)
b E IDES et al. (2001)
c K RAMIDA (2010)
d J ENTSCHURA et al. (2005)
e J ENTSCHURA and PACHUCKI (1996)
f J ENTSCHURA et al. (2001) (non-perturbative)
g order of magnitude: 1s1/2 scaled by 1/n3
p 2S1/2 2P1/2 L AMB shift which the P OHL et al. collaboration succeeded to
measure with high precision is 49881.88(76) GHz. On the basis of FS and HFS
calculation and QED they deduced from their measurement a value for the proton
radius rp = 0.84184(67) fm which claims an order of magnitude higher accuracy
and differs by 5 standard deviations from the above CODATA value!
Section summary
The L AMB shift manifests itself most clearly by a splitting of the 2s1/2 and
2p1/2 terms in atomic H for which D IRAC theory predicts identical energy.
We memorize (2s1/2 ) (2p1/2 ) 1060 MHz.
It was first discovered in 1947 by L AMB and R ETHERFORD using microwave
spectroscopy and triggered the development of QED. Today two-photon spec-
troscopy of the H and D atom is the method of choice. It has achieved an
amazing precision, not least due to the work of H NSCH and collaborators,
using frequency comb techniques and state-of-the-art atomic clocks.
Agreement with equally accurate QED computations is convincing. We have
encountered a couple of F EYNMAN graphs which are useful to understand the
origin of such radiative corrections for high precision atomic and molecular
spectroscopy. In these terms, the L AMB shift arises dominantly from self en-
ergy, with some correction from vacuum polarization both are single loop
terms.
6.6 Electron Magnetic Moment Anomaly 331
In a more picturesque language L AMB shift arises from the interaction of

the atomic electron with the vacuum field of electromagnetic radiation. This
causes a Zitterbewegung (trembling motion) of the electron which lets it
see an effectively smeared out C OULOMB potential. This, in turn, leads to
a rise of the energy term for electrons which come close to the nucleus, i.e.
essentially only for s electrons.
Heavy ion physics also provides a rich field for studies of the L AMB shift,
in particularly so as QED expands radiative corrections in series of powers
(Z)n .
6.6 Electron Magnetic Moment Anomaly
According to D IRAC theory the electron has a magnetic moment
S = ge B s e
M with s = 1/2, ge = 2 and B =
2me
as introduced in (6.28). We have already mentioned on several occasions that the
value ge = 2 differs slightly from reality.
One defines the electron magnetic moment anomaly as
ae = (ge 2)/2. (6.96)
After the preceding discussion we rightly suspect the origin of this deviation to be
again the interaction of the electron with the vacuum field. But let us first have a
look at the experimental determination of ae which in itself is fascinating. Again
we encounter an experiment of unbelievable accuracy. All experiments to determine
ge 2 are essentially based on a very precise comparison of the (non-relativistic)
cyclotron frequency of the electron in an external magnetic field B according to
(1.111)
e
c = B (6.97)
me
with the L ARMOR precession frequency of its magnetic moment (1.162):
e
s = ge L = ge B. (6.98)
2me
Knowing both frequencies allows one to derive the anomaly by (s c )/c =
(ge 2)/2. An older, but still instructive experiment is shown schematically in
Fig. 6.32. A 100 ns, 100 keV electron beam is partially polarized by scattering from
a gold foil. Electrons drift on helical trajectories into a magnetic field trap, their
polarization being perpendicular to both, the field and their velocity. After a well
defined time they are pulsed out of the trap and their polarization is measured by
scattering from a second gold foil. For a hypothetical value ge = 2 the spin orienta-
tion of the electrons after each full circle would be the same as at the entrance. Due
Fig. 6.32 Experimental unpolarized elektrons

setup for measuring g 2 polarization foil B exit cylinder
according to W ESLEY and
R ICH (1971). For details see
text
trap
region detector
injection cylinder B analyzing foil
to the anomaly, spin precession time and circling time differ, thus allowing a direct
determination of ae .
Todays most precise method goes back to D EHMELT and his collaborators (VAN
DYCK et al. 1986). It too uses the difference between s and c , however, in a highly
refined and compact manner. Thereby VAN DYCK et al. (1987) succeeded to store
a single electron in a so called P ENNING trap. With this method they were, for the
first time, able to measure ge 2 with high precision. D EHMELT and PAUL (1989)
received the N OBEL prize for the development and application of such ion traps
together with Norman R AMSEY, who was honoured for his method of separated
field oscillators (which we have discussed in Sect. 6.1.7). Both inventions were in-
strumental for the development of precision spectroscopy and time measurement,
today a standard in atomic clocks as well as for high resolution quadrupole mass
spectrometers and many other applications. As indicated in Fig. 6.33(a), a magnetic
field B of several Tesla (T), parallel to the axis of the trap, leads to cyclotron mo-
tion of the electron. In addition, an electric quadrupole field (equipotential surfaces
x 2 + y 2 2z2 = const) keeps the electrons in the centre of the trap.8
The classical motion of an electron in such a trap is characterized by three eigen-
frequencies: One axial frequency,

2eU0
z = with d02 = 02 + 2z02 ,
me d02
and two radial frequencies,

c c2 z2
c = + , the modified cyclotron frequency, (6.99)
2 4 2

c c2 z2
and m = , the magnetron frequency. (6.100)
2 4 2
8 The so called PAUL trap uses a similar design, however, instead of using a magnetic field the
electrons are stabilized by superposing an oscillatory field onto the static electric quadrupole field.
negative potential
on the upper cap
energy
electron
positive potential trajectories
on the ring
negative potential
on the lower cap (a) (b)
Fig. 6.33 (a) P ENNING trap for the determination of g 2 according to VAN DYCK et al. (1986).
(b) Simplified scheme for the eigenfrequencies of Geonium
Quantum mechanically, the L ARMOR frequency (6.98) of the electron spin has to
be added. For sufficiently low temperatures and high magnetic field it makes sense
to look for quantum mechanical solutions to this system (using S CHRDINGER or
D IRAC equation). The electron then assumes discrete energy levels Wnm which are
a combination of electron spin states and harmonic oscillator, the so called L ANDAU
levels:
2
ge 1 1 1
Wnm = c ms + n + c n + + ms . (6.101)
2 2 2 2
The first term describes the spin states with the projection quantum number ms ,
the second term contains the modified cyclotron resonances and the third term is
the leading relativistic correction term with /c = c /me c2 . The name Geonium
atom was coined for this system. Its lowest terms are shown in Fig. 6.33(b). By
cooling the system to very low temperatures one may indeed trap one single elec-
tron and bring it into its energetically lowest states. Transitions are detected by
induced currents. The experimental challenge is a most accurate determination of
the anomaly frequency a = gc /2 c between the states n = 0, ms = 1/2 and
n = 1, ms = 1/2 as indicated in Fig. 6.33(b). From this the electron anomaly ae is
obtained, if one knows c and the axial frequency z with sufficient accuracy.
Such a high precision warrants a fundamental understanding of the physics in
these P ENNING traps which has improved during the past two decades substantially.
Many methodological improvements have been introduced. We cannot go into the
details and have to refer the interested reader to the original literature, in particular to
the work of G ABRIELSE and collaborators (P EIL and G ABRIELSE 1999; H ANNEKE
et al. 2008) who were able to overcome several imperfections of the trap by a variety
of clever experimental tricks. They improved the accuracy of the measurement by a
factor of 10.
In the course of these improvements they also succeeded to really document the
population of the quantum states shown in Fig. 6.33(b) in the P ENNING trap. This
is illustrated in Fig. 6.34 which shows the measured probability of quantum jumps
between the individual n states. The number of quantum jumps observed depends
directly on the thermal population of the excited states. It reduces with temperature
4.2 K 1.0
0.0
cycltoron energy / c
3.2 K 1.0
probability, Pn
0.0
2.0 K 1.0
0.0
1.6 K 1.0
0.0
0.08 K 1.0
0.0
0 10 20 30 40 50 012
time / min n
Fig. 6.34 Quantum jumps between lowest states of the one electron cyclotron oscillator. A dra-
matic reduction with cavity temperature is seen. Based on this, on the right a population analysis
of the n states in the trap is shown. Adapted from P EIL and G ABRIELSE (1999)
as quantified by the population analysis on the right of Fig. 6.34. At the lowest
temperatures, indeed, only the ground state is still populated and no quantum jumps
are detected any more.
We recall that we have discussed quantum jumps in atoms already in Sect. 4.7.3.
However, in the present case we are discussing quantum jumps in a truly macro-
scopic quantum system. Thus, this experiment should be kept in mind in the context
of a presently very fashionable discussion about the limits (or extensions) of quan-
tum mechanics into the macroscopic world!
In the most recent high precision experiment (O DOM et al. 2006; H ANNEKE et
al. 2008) a one electron cyclotron oscillator (Geonium) was cooled down to below
100 mK and the magnetic field was 10.6 T. In this case s c 149 GHz while the
axial frequency was tuned to z 200 MHz. The magnetron frequency was m
134 kHz. The resulting presently most accurate value is given by 2010 CODATA
(NIST 2010) as9
ae = (ge 2)/2 = 1.15965218076(27) 103 . (6.102)
In this context, the reader should bear in mind that the determination of fundamen-
tal physical constants is today based on an intricate system of weighted averages of
a whole number of precision measurements for different observables (ae is one of
them). They are thus not independent of each other and any change in the system
9 We note for completeness that the anomaly for the muon is also known with high accuracy
(albeit to some orders of magnitude less well): a = 1.16592091(63) 103 .
1 2
g= 21 + + ... + ... ... + ....
2
DIRAC SCHWINGER + 5 more two-loop graphs
Fig. 6.35 F EYNMAN diagrams for ge 2 in 1st order, 2nd order (one-loop) and some examples
for 4th order (two-loop contributions). Note, the open photon propagator indicates interaction with
the external magnetic field
of constants entering this CODATA system changes other constants as well. Specif-
ically, the correlation coefficient between ae and , the fine structure constant, is
given as r = 0.9384 by the system. This means, the recommended numerical values
for both fundamental constants depend very strongly on each other and we have
to add: on a precise theoretical interpretation.
The theory of the electron magnetic moment anomaly has first been developed by
J. S CHWINGER. According to this work the first nonvanishing term for ae is of 2nd
order in QED and amounts to ae = /2 = 1.161 103 . This is already very close
to reality. Figure 6.35 shows the graphs up to 2nd order of the perturbation series
completely, as well as two further correction terms of 4th order (the additional 5
possible graphs or 4th order are also two-loop contributions and are not difficult to
guess).
The result of such QED computations is often written as
aeQED = A1 + A2 (me /m ) + A2 (me /m ) + A3 (me /m , me /m ),
where even the virtual generation of and Leptons are included. Each of these
individual terms is in turn expanded in a series:
2 3
(2) (4) (6)
Ai = Ai + Ai + Ai + .

Clearly, the A1 series contributes the most (in the A1 series the term of 2nd order
(2)
is the S CHWINGER term A1 = 0.5). When evaluating the perturbation integrals
behind these coefficients ab initio one major concern is good bookkeeping. To illus-
trate the efforts necessary: In 6th order there are already 72 graphs which contribute
to A(6)
1 and in 8th order all 891 F EYNMAN graphs have been included. On that basis,
according to G ABRIELSE ET al. (2006), the results for ae are inverted to determine
todays most precise value of the fine structure constant . Based on AOYAMA et
al. (2007) and H ANNEKE et al. (2008), who report 137.035999084(51),10 the 2010
10 More recently, AOYAMA et al. (2012) have performed even 10th order QED calculations (in-
volving 12672 diagrams) and report ae = 1.15965218178(77) 103 from which they derive
1 = 137.035999173(35).
CODATA data bank (M OHR et al. 2012) gives
e2
= = 1/137.035999074(44). (6.103)
40 c
Section summary
The deviation of the ge factor from 2 is given as the so called anomaly of the
magnetic moment of the electron, ae = (ge 2)/2 1.16 103 and can be
determined today with an amazing accuracy of about 3 1010 .
Experimentally it is measured by sophisticated techniques in electron traps
(P ENNING trap, first used by D EHMELT in this context), where quasi-
macroscopic states of the electron are prepared the so called Geonium.
Essentially, one measures the difference between the L ARMOR and the cy-
clotron frequency.
The precise determination of the magnetic moment of the electron provides
another stringent test for the potential of QED calculations. While the domi-
nant contribution arises from the one second order perturbation term (single
loop), presently the complete contributions up to 8th order have been com-
puted.

AOM: Acousto-optic modulator, device to modulate and shift the frequency of
light by diffraction in a B RAGG grating generated by sound waves (usually RF).
CRD: Cavity ring down spectrometer (see Sect. 5.5.3, Vol. 2).
CW: Continuous wave (as opposed to pulsed) light beam, laser radiation etc.
EBIT: Electron beam ion trap, source for highly charged ion beams see Sect. 7.5,
Vol. 2.
FPI: FABRY-P ROT interferometer, for high precision spectroscopy and laser res-
onators (see Sect. 6.1.2).
HCI: Highly charged ions, see Sect. 7.5, Vol. 2.
References 337
LHC: Left hand cicularly, polarized light, also + light.
QDT: Quantum defect theory, interprets experimental spectra by phase shifts in
the radial wave functions and makes predictions for scattering processes (see
Sect. 3.2.6).
interaction.
REMPI: Resonantly enhanced multi-photon ionization, ionization of atoms or
molecules by several photons with one resonant intermediate state.
RHC: Right hand cicularly, polarized light, also light.
SHG: Second harmonic generation, doubling of a fundamental frequency, for in-
frared or visible light typically by methods of nonlinear optics.
index.html.
21348 (2007).
tween 0.1 nm and 10 nm according to ISO 21348 (2007), sometimes up to 40 nm.
References
AOYAMA , T., M. H AYAKAWA , T. K INOSHITA AND M. N IO: 2007. Revised value of the eighth-
order contribution to the electron g-2. Phys. Rev. Lett., 99, 110 406.
AOYAMA , T., M. H AYAKAWA , T. K INOSHITA AND M. N IO: 2012. Tenth-order QED contribu-
tion to the electron g-2 and an improved value of the fine structure constant. Phys. Rev. Lett.,
109, 111 807.
B EIERSDORFER , P.: 2010. Testing QED and atomic-nuclear interactions with high-Z ions.
J. Phys. B: At. Mol. Phys., 43, 074 032.
B ETHE , H. A.: 1947. The electromagnetic shift of energy levels. Phys. Rev., 72, 339341.
B IRABEN , F.: 2009. Spectroscopy of atomic hydrogen how is the Rydberg constant determined?
Eur. Phys. J. ST, 172, 109119.
B JORKEN , J. D. AND S. D. D RELL: 1964. Relativistic Quantum Mechanics. New York: McGraw
Hill.
B ORN , M. AND E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
B RINK , D. M. AND G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University
Press, 3 edn., 182 pages.
D EHMELT , H. G. AND W. PAUL: 1989. The N OBEL prize in physics: for the development of the
VAN DYCK , R. S., P. B. S CHWINBERG AND H. G. D EHMELT : 1986. Electron magnetic-moment
from geonium spectra early experiments and background concepts. Phys. Rev. D, 34, 722
736.
VAN DYCK , R. S., P. B. S CHWINBERG AND H. G. D EHMELT : 1987. New high-precision com-
parison of electron and positron g-factors. Phys. Rev. Lett., 59, 2629.
E IDES , M. I., H. G ROTCH AND V. A. S HELYUTO: 2001. Theory of light hydrogenlike atoms.
Phys. Rep., 342, 63261.
F EYNMAN , R. P.: 1949. Space-time approach to quantum electrodynamics. Phys. Rev., 76, 769
789.
G ABRIELSE , G., D. H ANNEKE , T. K INOSHITA , M. N IO AND B. O DOM: 2006. New determi-
nation of the fine structure constant from the electron g value and QED. Phys. Rev. Lett., 97,
030 802.
G ROSS , B., A. H UBER , M. N IERING , M. W EITZ AND T. W. H NSCH: 1998. Optical Ramsey
spectroscopy of atomic hydrogen. Europhys. Lett., 44, 186191.
G RYNBERG , G. AND B. C AGNAC: 1977. Doppler-free multi-photonic spectroscopy. Rep. Prog.
Phys., 40, 791841.
G UMBERIDZE , A. et al.: 2005. Quantum electrodynamics in strong electric fields: The ground-
state lamb shift in hydrogenlike uranium. Phys. Rev. Lett., 94, 223 001 and personal commu-
nication from T. S THLKER.
H ANNEKE , D., S. F OGWELL AND G. G ABRIELSE: 2008. New measurement of the electron
magnetic moment and the fine structure constant. Phys. Rev. Lett., 100, 120 801.
H NSCH , T. W.: 2005. N OBEL lecture: Passion for precision, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/2005/hansch-lecture.html.
H NSCH , T. W., M. H. NAYFEH , S. A. L EE , S. M. C URRY AND I. S. S HAHIN: 1974. Precision-
measurement of Rydberg constant by laser saturation spectroscopy of Balmer alpha line in
hydrogen and deuterium. Phys. Rev. Lett., 32, 13361340.
H NSCH , T. W., I. S. S HAHIN AND A. L. S CHAWLOW: 1971. High-resolution saturation spec-
troscopy of sodium D lines with a pulsed tunable dye laser. Phys. Rev. Lett., 27, 707710.
H NSCH , T. W., I. S. S HAHIN AND A. L. S CHAWLOW: 1972. Optical resolution of Lamb shift
in atomic-hydrogen by laser saturation spectroscopy. Nat. Phys. Sci., 235, 63.
H NSCH , T. W. et al.: 2005. Precision spectroscopy of hydrogen and femtosecond laser frequency
combs. Philos. Trans. R. Soc. Lond., Ser. A, 363, 21552163.
J EFFERTS , S. AND D. M EEKHOF: 2011. NIST-F1 cesium fountain atomic clock, NIST. http://
www.nist.gov/physlab/div847/grp50/primary-frequency-standards.cfm, accessed: 8 Jan 2014.
J ENTSCHURA , U. AND K. PACHUCKI: 1996. Higher-order binding corrections to the Lamb shift
of 2p states. Phys. Rev. A, 54, 18531861.
J ENTSCHURA , U. D., S. KOTOCHIGOVA , E. O. L. B IGOT , P. J. M OHR AND B. N. TAY-
LOR : 2005. The energy levels of hydrogen and deuterium (version 2.1), NIST. http://
physics.nist.gov/PhysRefData/HDEL/, accessed: 8 Jan 2014.
J ENTSCHURA , U. D., P. J. M OHR AND G. S OFF: 2001. Electron self-energy for the K and L
shells at low nuclear charge. Phys. Rev. A, 63, 042 512.
References 339
J ENTSCHURA , U. D., A. M ATVEEV, C. G. PARTHEY, J. A LNIS, R. P OHL, T. U DEM, N. KO -

LACHEVSKY and T. W. H NSCH : 2011. Hydrogen-deuterium isotope shift: From the 1s2s-
transition frequency to the proton-deuteron charge-radius difference. Phys. Rev. A, 83, 042 505.
J OHNSON , W. R. AND G. S OFF: 1985. The Lamb shift in hydrogen-like atoms, 1 less-than-or-
equal-to Z less-than-or-equal-to 110. At. Data Nucl. Data Tables, 33, 405446.
K ARSHENBOIM , S. G.: 2005. Precision physics of simple atoms: QED tests, nuclear structure
and fundamental constants. Phys. Rep., 422, 163.
K RAMIDA , A. E.: 2010. A critical compilation of experimental data on spectral lines and energy
levels of hydrogen, deuterium, and tritium. At. Data Nucl. Data Tables, 96, 586644.
K RAMIDA , A. E., Y. R ALCHENKO , J. R EADER and NIST ASD T EAM: 2013. NIST Atomic
L AMB , W. E. AND P. K USCH: 1955. The N OBEL prize in physics: for his discoveries concerning
the fine structure of the hydrogen spectrum, and for his precision determination of the magnetic
moment of the electron, respectively, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1955/.
M ATVEEV , A. et al.: 2013. Precision measurement of the hydrogen 1s2s frequency via a 920-km
fiber link. Phys. Rev. Lett., 110, 230 801.
M ICHELSON , A. A.: 1907. The N OBEL prize in physics: for his optical precision instruments
and the spectroscopic and metrological investigations carried out with their aid, Stockholm.
M OHR , P. J., B. N. TAYLOR AND D. B. N EWELL: 2012. CODATA recommended values of the
fundamental physical constants: 2010. Rev. Mod. Phys., 2013, 15271605. http://physics.nist.
gov/constants, accessed: 8 Jan 2014.
N IERING , M. et al.: 2000. Measurement of the hydrogen 1s2s transition frequency by phase
coherent comparison with a microwave cesium fountain clock. Phys. Rev. Lett., 84, 5496
5499.
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/
cuu/Constants/, accessed: 7 Jan 2014.
O DOM , B., D. H ANNEKE , B. DU RSO AND G. G ABRIELSE: 2006. New measurement of
the electron magnetic moment using a one-electron quantum cyclotron. Phys. Rev. Lett., 97,
030 801.
PALMER , C. AND E. L OEWEN: 2005. Diffraction grating handbook, New York: Newport corpo-
ration. http://www.gratinglab.com/Information/Handbook/Cover.aspx, accessed: 8 Jan 2014.
PARTHEY , C. G. et al.: 2011. Improved measurement of the hydrogen 1s2s transition frequency.
Phys. Rev. Lett., 107, 203 001.
P EIL , S. AND G. G ABRIELSE: 1999. Observing the quantum limit of an electron cyclotron: QND
measurements of quantum jumps between fock states. Phys. Rev. Lett., 83, 12871290.
P OHL , R. et al.: 2010. The size of the proton. Nature, 466, 213216.
P OSTAVARU , O., Z. H ARMAN AND C. H. K EITEL: 2011. High-precision metrology of highly
charged ions via relativistic resonance fluorescence. Phys. Rev. Lett., 106, 033 001.
R AMSEY , N. F.: 1950. A molecular beam resonance method with separated oscillating fields.
Phys. Rev., 78, 695699.
R AMSEY , N. F.: 1989. The N OBEL prize in physics: for the invention of the separated oscillatory
fields method and its use in the hydrogen maser and other atomic clocks and the separated oscil-
latory fields method, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1989/.
S TECK , D.: 2010. Alkali D line data. http://steck.us/alkalidata/, accessed: 8 Jan 2014.
S VEN G ATO R EDSUN: 2004. 3j6j9j-symbol java calculator, Sven Gato Redsun. http://www.
svengato.com/, accessed: 8 Jan 2014.
T OMONAGA , S.-I., J. S CHWINGER AND R. P. F EYNMAN: 1965. The N OBEL prize in
physics: for fundamental work in quantum electrodynamics, with deep-ploughing conse-
quences for the physics of elementary particles, Stockholm. http://nobelprize.org/nobel_prizes/
W EITZ , M., A. H UBER , F. S CHMIDT-K ALER , D. L EIBFRIED AND T. W. H NSCH: 1994.

Precision-measurement of the hydrogen and deuterium 1s ground state Lamb shift. Phys. Rev.
Lett., 72, 328331.
W ESLEY , J. C. AND A. R ICH: 1971. High field electron g-2 measurement. Phys. Rev. A, 4,
1341.
XFEL P ROJECT: 2011. European XFEL project, technical information, Hamburg, Ger-
many: DESY. http://xfel.desy.de/technical_information/photon_beam_parameter/, accessed: 8
Jan 2014.
Helium and Other Two Electron Systems
7
So far our discussions were restricted to atomic systems with

effectively one active electron in an attractive C OULOMB or
screened C OULOMB field. For the majority of atoms this simple
model cannot be maintained, since the electrons repel each
other pairwise and have to obey the PAULI principle. The
helium atom (He) is the simplest and purest example for
introducing the key problems which one encounters in true
multi-electron systems, as well as the methods for describing
and understanding them.
Overview
After a general introduction and a survey of the experimental observations
(Sect. 7.1) the quantum mechanical basis for treating multi-electron systems
is reviewed in Sect. 7.2. The reader should know about these tools or get
used to them by reading this section. Section 7.3 expands this theme by intro-
ducing electron exchange and the characteristic excited state configurations.
Fine structure interaction in He and He like ions are addressed in Sect. 7.4
a consolidation of our knowledge acquired in Chap. 6 which will turn out
to be essential also for later chapters. In Sect. 7.5 the most important selec-
tion rules for E1 transitions in multi-electron system are treated. This directly
leads to double excitation in Sect. 7.6. It is of importance far beyond atomic
physics: resonances of states imbedded into a continuum are found in all fields
of physics. The ensuing interference structures, known as FANO resonances
and their origin are described in a practical, easy to access approach. Finally,
in Sect. 7.7 the alkaline earth atoms and the Hg atom are treated: they are re-
lated to He in an analogous manner as alkali atoms are related to the H atom.

342 7 Helium and Other Two Electron Systems
7.1 Introduction and Empirical Findings
7.1.1 Basics
Helium, with its atomic number (and nuclear charge) Z = 2, is the most simple of
all multi-electron systems. Nevertheless, it eludes an exact calculation of its wave
functions, energies and other properties as indeed any multi-electron or manybody
system in general, starting with the classical three body system.
Helium is a very rare element. It owes its presence on earth mainly to the radioac-
tive decay of heavy elements: the He atom is an particle which has captured two
electrons. Typically it is isolated from natural gas, where at special sources, it may
be found with concentrations up to 7 %. There are two natural isotopes: 42 He (relative
probability of wrel = 99.999866(3) %, atomic mass m(42 He) = 4.00260325415(6) u,
nuclear spin I = 0), and the very rare 32 He (wrel = 0.000134(3) %, m(32 He) =
3.0160293191(26) u, I = 1/2).
In the ground state the two electrons of He are described by 1s 2 1 S0 . The first ion-
ization potential for separating one of these electrons is 24.5873876 eV. For ioniza-
tion of the second electron (in the hydrogen like ion He+ , spectroscopically called
He II) an energy Z 2 Eh /2 = 54.4177630 eV is needed. The total binding energy of
both electrons thus amounts to 79.0051506 eV.
Correspondingly, the spectrum of He extends from the infrared spectral range
(IR) through the visible (VIS), the ultraviolet (UV) far into the vacuum-ultraviolet
spectral range (VUV). Figure 7.1 shows by way of example a typical spectrum in
the UV/VIS spectral range which has been constructed with data from K RAMIDA
et al. (2013). Marked in red are a few characteristic emission lines ending on the
1s2s 1,3 S and 1s2p 1,3 P states. This is, however, only a section of the full spectrum
of He I and one can imagine clearly that, in the early phase of atomic physics, the
interpretation of such spectra was by no means trivial. In particular since all optical
emission and absorption lines from neutral He appear grouped into two seemingly
independent sets of spectra a singlet and a triplet system. Each of them demands
an independent assignment of term energies. In the early time of spectroscopy one
associated these even with two types of helium: para an ortho He. Today this puzzle
is of course completely resolved: one observes two sets of spectra with different
total spin quantum numbers: He with its two electrons is thus the prototype element
for which all basic phenomena of multi-electron systems can be studied very clearly.
1s2s 1S - 1s 4p 1P 1s2p 3P - 1s5d 3D 1s2s 1S - 1s3p 1P 1s2p 3P - 1s 3d 3D

1s2s 3S - 1s5p 3P 1s2p 1P - 1s4d 1D 1s2p 3P - 1s 3s 3S
200 300 400 500 600 700 800 nm
Fig. 7.1 Spectrum of He I in the VIS and UV spectral range, according to K RAMIDA et al. (2013).
Only a few, particularly intense lines are designated here (red)
7.1 Introduction and Empirical Findings 343
Of particular importance (for all areas of modern physics) is the so called exchange
interaction.
7.1.2 He I Term Scheme
In Fig. 7.2 the energy levels of the neutral He atom below the first ionization thresh-
old are summarized. This diagram represents extensive material and detective work
from generations of spectroscopists. We recognize in Fig. 7.2 the two term systems
already mentioned. According to their fine structure splitting we designate them as
singlet and triplet system. The ground state of He belongs to the singlet system.
Optical dipole transitions (E1) are observed only within each of the two systems
(indicated by red arrows in Fig. 7.2). However, one finds that collision processes
may very well induce transitions between singlet and triplet system and vice versa
we shall discuss such processes in Sect. 7.2, Vol. 2.
Obviously, degeneracy is removed in He: a consequence of a (partial) screening
of the nuclear charge which each electron experiences by the other electron. As in
the case alkali atoms one may characterize the term energies by an effective quantum
number n with a quantum defect = n n or by an effective charge Z :
Z 2 Eh Z 2 Eh
Wn = 2
= . (7.1)
2n 2n2
singlet triplet H
+) / eV
He+ + e-
W W(He
0 1S 1P 1D 1F 3S 3P 3D 3F
W
n=5
W-
n=4
-1 4 3F
2,3,4
n=3
31P1 31 D 2 3 3P 3 3D 123
1,2,3
-2 31S0 0 1,2
0,1
3 3 S1
-3 2 3 P0
-3.62318 eV
n=2
21P1
-4 2 3P
0,1,2
21S0 23P1
-3.62331 eV
23S1 2 3 P2
-24.59 ground state (1s ) 2 1S0
Fig. 7.2 Energy levels of the neutral helium atom, He I (so called G ROTRIAN diagram). Note that
the position of the ground state (1s)2 1 S0 is not drawn to scale. The spectra split into a singlet
and a triplet system. The inset shows the (inverted) triplet splitting of the 1s2p 3 P0,1,2 states. For
comparison the corresponding term energies of the H atom are shown: clearly, the quantum defects
of the He terms are small and converge rapidly with n
Fig. 7.3 Coordinates for the e- z

two electron atom He r12
r2 e-
12
r1
y
Z+
For the ground state of neutral He the measured binding energy 24.5873876 eV
gives n 0.7439 (or = 0.256) and Z 1.344, respectively. From the view
point of one of the two electrons the C OULOMB potential of the nucleus is screened
already in the ground state quite strongly. For all singly excited states of He the
quantum defect lies significantly below 1 and converges for the 1sns 1 S0 levels
rapidly to 0.139 only. Similarly Z rapidly converges to 1 and already for n > 3
and > 0 the energies of singlet and triplet terms are already very close to those in
atomic hydrogen. If we assume the configuration of the He electrons to be 1sn this
is easy to understand: the 1s electron finds itself in a nearly H like orbit (with Z = 2,
however) and screens the charge of the nucleus very strongly. The second electron
thus sees essentially a nuclear charge Z = 1.
Nevertheless, the excited electron in He behaves different from the valence elec-
tron e.g. in alkali atoms since the interaction with the second electron is still very
direct. It is important to realize that both electrons can in principle not be distin-
guished! This will be a key issue in the quantum mechanical treatment. We shall
find that this indistinguishability is the key to the puzzle with the two term systems
of He.
Section summary
He is the prototype of two electron systems and indeed the basis for under-
standing any multi-electron system. Its first ionization potential is 24.6 eV,
the second electron requires 54.4 eV for ionization.
The spectra of He (ranging from the IR to the far VUV region) show two
characteristic types of term systems which in E1 transitions do not mix: a sin-
glet and a triplet system. Understanding this particular feature will be a key
to the understanding of electron-electron interaction and electron exchange:
both electrons are in principle indistinguishable.
7.2 Some Quantum Mechanics of Two Electrons
7.2.1 HAMILTON Operator for the Two-Electron System
A two electron system is characterized by two position vectors r 1 and r 2 as sketched

schematically in Fig. 7.3. Up to now we have not yet introduced any formalism
for treating multi-electron systems. We follow the general rules already used for
7.2 Some Quantum Mechanics of Two Electrons 345
deriving observables in the one electron case, specifically the Hamiltonian. We just
have to add the components: the energies for the two individual electrons 1 and 2
p 21 Ze2 p 22 Ze2
1 =
H 2 =
and H , (7.2)
2me 40 r1 2me 40 r2
respectively, and the repulsive C OULOMB energy between the two electrons
ee = e2
H . (7.3)
40 r12
The full H AMILTON operator for the two electron system is thus
p 21 Ze2
p2 Ze2 e2
=H
H 2 + H
1 + H ee = + 2 + . (7.4)
2me 40 r1 2me 40 r2 40 r12
ee depends on
Note that the interaction term H
r12 = |r 1 r 2 | = r12 + r22 2r1 r2 cos 12 , (7.5)
and thus breaks the original spherical symmetry an essential which we have ex-
ploited when solving the one electron case.
In the remaining chapter we shall (unless mentioned otherwise) again write the
H AMILTON operator in a.u.:
=H
H 2 + H
1 + H ee = Z Z + 1 . (7.6)
2 r1 2 r2 r12
7.2.2 Two Particle Wave Functions
With (7.6) we have now a Hamiltonian which depends on two position vectors.
We continue to use polar coordinates and the wave functions will now depend on
the coordinates {r1 , 1 , 1 , r2 , 2 , 2 }. The most simple ansatz for a wave function
would be
(r 1 , r 2 ) = 1 (r1 , 1 , 1 )2 (r2 , 2 , 2 ). (7.7)
This product ansatz is strictly valid only for the model of independent particles, i.e.
for two electrons that move completely uncorrelated and experience from each other
only an average potential. This turns out to be a good 1st order approximation for
the 1sn singly excited states of He. However, for the ground state and for doubly
excited states where the two electrons interact closely this approximation is only of
limited value.
By consequently extending the probability definition of the wave function in the
one particle case, as introduced with (2.5), the probability to find electron 1 at posi-
tion r 1 and simultaneously electron 2 at position r 2 in a volume element d3 r 1 d3 r 2
is given by
2
dw12 = (r 1 , r 2 ) d3 r 1 d3 r 2 (7.8)
2 2
= 1 (r1 , 1 , 1 ) 2 (r2 , 2 , 2 ) d3 r 1 d3 r 2 .
From this, the probability to find electron 1 at position r 1 and electron 2 anywhere
becomes

2 2
dw1 = d r 1 (r 1 , r 2 ) d3 r 2 = 1 (r1 , 1 , 1 ) d3 r 1 .
3
(7.9)
2
As usual, we assume the one electron wave functions (orbitals) 1 (r1 , 1 , 1 ) and
2 (r2 , 2 , 2 ) to be orthogonal and normalized.
The indices 1 and 2, more generally and , stand for the usual sets of quantum
numbers:
= {n m ms }, = {n m ms }. (7.10)
For completeness and later use we also include here the projection quantum number
ms = 1/2 of the electron spin. They will become important in Sect. 7.3.
In the compact bra-ket form, such two particle product states are written

(1, 2) = (r 1 ) (r 2 ) = |n m n m , (7.11)
where the indices 1 and 2 (given in round brackets) are an abbreviation for the
coordinates of the two electrons. Assuming that the single particle orbitals (r j )
are orthonormalized, this hold also for the two particle states:

(1, 2) (1, 2) = (1) (1) (2) (2) (7.12)
= n m |n m n m |n m
= .
It is important to note that the Hamilton operator (7.6) for the two electron system
in this form does not depend on electron spin! All influence of the relative orienta-
tion of the spins of the two electrons will enter through the anti-symmetrization rule
to be discussed in Sect. 7.3. For the moment we ignore the spin and just note that
the Hamiltonian does not change if we exchange the two electrons, i.e. their position
coordinates r 1 and r 2 .
7.2.3 Zero Order Approximation: No e e Interaction
In the spirit of perturbation theory we first define the unperturbed problem and
start with its solution. Thus, we neglect in (7.4) the repulsive screening term (7.3),
to be treated later as perturbation, and write the Hamiltonian in 0th order in a.u.:

p 2i Z
H 1 + H
0 = H 2 i =
with H (7.13)
2 ri
Hi describes the eigenvalue problem of the one particle system, i.e. the hydrogen
like He+ ion. Its energies Wni are given by
2
i |ni i mi = Wni |ni i mi = Z |ni i mi .
H (7.14)
2n2i
For the two particle system we thus have in 0th order

H0 | = H 1 |1 |2 + H 2 |2 |1

= H 1 |n1 1 m1 |n2 2 m2 + H 2 |n2 2 m2 |n1 1 m1
= (Wn1 + Wn2 )|n2 2 m2 |n1 1 m1 = (Wn1 + Wn2 )| ,
so that the energy of the He atoms in 0th order is approximation is

Z2 1 1
Wn1 n2 = Wn1 + Wn2 =
(0)
+ . (7.15)
2 n21 n22
Note that these energies are calibrated in respect of the fully separated He++ + e +
e system. Experimentally measured are spectral lines (i.e. energy differences).
Tabulated one typically finds term energies Wn1 1 n2 2 W1s 2 > 0 and ionization po-
tentials WI > 0 in respect of the He(1s 2 ) ground state (e.g. K RAMIDA et al. 2013).
(Alternatively, binding energies for the valence electron are reported in the literature
in respect of the first ionization threshold Wn1 1 n2 2 + WI (I).)
Figure 7.4 compares for He with Z = 2 the true, experimentally determined en-
ergies (left) with the 0th order approximation (right) for the neutral atom (He I)
as well as for the ionic system He+ + e (He II). The ionization energy of the hy-
drogen like He+ ion (Z = 2) can be computed from S CHRDINGER equation as
(0)
WI (II) = Z 2 Eh /2 = 2Eh = 54.422 . . . eV. This is more ore less exact: according
to K RAMIDA et al. (2013) at present the best measured value is1

WI (II) = WI He+ (1s) = 54.41776203 eV, (7.16)
while the value given for the ionization potential of neutral He is

WI (I) = WI He 1s 2 = 24.58738777 eV. (7.17)
Thus, the (true) total energy of the neutral He ground state is
W1s 2 = WI (I) WI (II) = 79.0051498 eV = 2.903387219Eh . (7.18)
In contrast, 0th order approximation (7.15) gives
(0) Z2
W1s 2 = Eh 2 = 4Eh = 108.8 eV (7.19)
2
1 The e /me for finite mass and

difference is due to the necessary kinematic corrections with m
further corrections such as the L AMB shift.
^
W - W1s 2 / eV experiment H0 without e- e- interaction Wnn''' / eV
WI (II) + WI (I) He++ + 2e-

=79.005 0
72.96
3sn 3s
65.40 2sn 3s 2
2s - 13.6 - 20
60 2s2 - 27.2
autoionization
40 - 40
He+ + e-
WI (I)= 24.59 - 54.42
20 -13.60 1s - 60
1s 2s 1S0 1s2s 3S1
20.62eV 19.82eV 1s2s,p
0 He - 80
1s 2 1S 0 - 54.42
0.0eV
- 100
- 108.84
1s1s
Fig. 7.4 Experimentally determined energies of He and He+ (left, black) and ansatz for perturba-
tion theory in 0th order (right, red)
(0) (0)
and WI (I) = WI (II). And similarly, the 0th order binding energy of one electron
in an n state is Z 2 ((1 + 1/n2 ) + 1)/2Eh = 2Eh /n2 , specifically Eh /2 =
13.6 eV in the 1s2s or 1s2p states (the excitation energy in 0th order would be
Eh /2 (2Eh ) = 1.5Eh ).
As documented in Fig. 7.4 the real energies lie obviously much higher due to
screening of the attractive C OULOMB potential by the repulsive 1/r12 term. Obvi-
ously, theory faces a major challenge if one aims at an accuracy matching that of the
experiment!
Even more complex is the situation when considering two excited electrons
in excited n states. In 0th order a 2 2 configuration would lead to an excita-
tion energy 3Eh which is higher than the first ionization potential in 0th order.
Hence, it is already embedded into the continuum of the singly ionized helium and
has an excitation energy Eh in respect of the He+ 1s 2 S0 ground state as indi-
cated in Fig. 7.4. Such states really exist and we shall explore this kind of dou-
ble excitation and the ensuing phenomenon of FANO resonances in some detail in
Sect. 7.6.
7.2.4 The He Ground State Perturbation Theory
Thus, the screening term (7.3) has substantial influence on the total energy of the
He ground state. In a first step one may try to approach the problem by perturbation
ee > 0 we expect indeed to come closer to the experimental value
theory. Since H
if we account for it. For the He(1s 2 ) ground state in 0th order both electrons are
Table 7.1 One electron Exp. 0th order 1st order

binding energies / eV (or
ionization potentials) of the H 0.76 13.6 003.4
ground state for He and He He 24.58741 54.4 20.4
like ions: experimental data Li+ 75.64018 122.4 71.4
according to K RAMIDA et al.
(2013) are compared with 0th Be++ 153.8945 217.7 149.6
and 1st order perturbation B3+ 259.3752 340.1 255.1
theory C4+ 392.0872 489.9 387.7
described by 1s hydrogen like wave functions (with Z = 2) and the wave function
(7.11) is written as

1s 2 (r 1 , r 2 ) = 1s (r 1 )1s (r 2 ) or 1s 2 = |100 |100 . (7.20)
In 1st order perturbation theory the energy correction term according to (3.41) is:
(0) ee |1s 2 = 1s 2 | 1 |1s 2

W = W1s 2 W1s 2 = 1s 2 |H (7.21)
r12

1 1 2 2
= 100|100| |100 |100 = d3 r 1 d3 r 2 100 (r 1 ) 100 (r 2 ) .
r12 r12
To evaluate this integral we insert the 1s wave function from Table 2.2:
100 (r) = Y00 (, )R1s (r) = (Z)3/2 1/2 eZr .
With this the integral is written explicitly as

2Zr2
(Z)6 e
W = 2 e2Zr1 d3 r 1 d3 r 2 ,
r12
and the inverse distance between the two electrons 1/r12 may be expanded into a
power series according to (F.2) involving L EGENDRE polynomials. Leaving the
details as an exercise to the interested reader we simple convey that integration
leaves only the first term from this series and the final result (with Z = 2) is simply
5
W = ZEh = 34.01 eV.
8
Thus, in 1st order perturbation theory the total energy of the He ground state is
(1) 5
W1s 2 = Eh Z 2 + ZEh = 74.79 eV. (7.22)
8
Considering the crudeness of the ansatz this results is astonishingly close (about
5 %) to the experimental result (7.18).
It is instructive, to compare the experimental binding energies (or ionization po-
tentials) for several members of the He like isoelectronic sequence to WI (I) from
(7.18) with 0th and 1st order perturbation theory according to (7.19) and (7.22).
This is summarized in Table 7.1.
Obviously, the relative accuracy of 1st order perturbation theory improves with
increasing nuclear charge Z. While for the H anion perturbation theory completely
fails, it is quite satisfactory for the four fold ionized carbon where an agreement
between experiment and perturbation theory to about 1 % is reached.
We note here in passing, that the He like 1s 2 1 S0 state of H is the only existing
and experimentally documented, stable state of the hydrogen anion.
7.2.5 Variational Theory and Present State-of-the-Art
Variational methods are very important in quantum mechanics for the determination
of energies, in particular for ground states. Helium is a particular clear example for
its efficiency, and we shall use variational procedures in many other instances as
well. The variational principle (also called the R ITZ or R AYLEIGH -R ITZ method
after the inventors) makes use of an important theorem according to which the mini-
mal value of energy W which is calculated for a given class of functions is always
the best:

|
|H H d3 r
W = min = min 3 (7.23)
| d r

with (r) = c (r 1 , r 2 ).

Here r refers to the full configuration space of both electrons, and the coefficients
c have to be varied until W reaches its minimum.
In detail, one chooses a test wave function for the ground state which is in prin-
ciple able to describes the properties of the system under study as well as possible.
With this test wave function one calculates the energy as expectation value of the
H AMILTON operator so that W is derived as a function of the parameters ck defin-
ing the wave function and one minimizes W according to the standard procedures
of analysis. Specifically for He the so called H YLLERAAS wave functions

N
(s, t, u) = exp(ks) c ,2m,n s t 2m un (7.24)
mn
with s = r1 + r2 , t = r1 r2 and u = r12
have proved to be very efficient. Thus, the energy is computed from

W = (s, t, u)H (s, t, u) / (s, t, u) (s, t, u) (7.25)
by variation of k and the coefficients c ,2m,n to obtain the minimum energy W . The
resulting wave function obviously does not describe two independent electrons
for which the wave function would be given by the product of the individual or-
bitals. Rather, the electrons in a H YLLERAAS type wave function are highly corre-
lated: they influence each other more directly as could be described by an averaged
7.3 PAULI Principle and Excited States in He 351
potential. With this ansatz and typically 5 parameters a very high accuracy can be
achieved. H YLLERAAS already obtained W1s1s = 79.001 eV, to be compared with
(7.18). Modern state-of-the-art calculations include all relevant corrections such
as the finite He mass, L AMB shift, nuclear size and the recalculation to the present
value of R . The agreement between experiment and theory is on the order of 107
(D RAKE and M ARTIN 1998). The data presently reported in
Section summary
The Hamiltonian (7.6) of a two electron system is given by adding up all
energies of the two single particle systems plus the mutual repulsion of the
two electrons.
The most simple ansatz for two particle wave functions is a product of sin-
gle particle wave functions for electron 1 and 2. Such a simple independent
particle model is often a reasonable 0th and 1st order approximation.
Neglecting the 1/r12 screening potential it leads to energies for He which are
substantially too low, but predicts interesting doubly excited states.
Including, however, the 1/r12 term in 1st order perturbation theory, a hy-
drogenic (1s)2 product wave functions already gives a total energy for the
He(1s 2 ) ground state which agrees to within 5 % with experiment.
Excellent agreement is obtained by variational theory (7.23), also called
R AYLEIGH -R ITZ method. For the He ground state, so called H YLLERAAS
wave functions (7.24) have proved very useful. They imply strong correlation
between the two electrons.
7.3 PAULI Principle and Excited States in He
7.3.1 Exchange of Two Identical Particles
Up to now we have not yet explicitly accounted for the fact that electrons are
fermions. Before clarifying this aspect specifically for He, we take the opportunity
to discuss multiparticle wave functions in a more general context.
It is straight forward to generalize the two particle wave function introduced in
Sect. 7.2.2 to a system with N identical particles. To include the spin properties of
the particle we introduce so called spin-orbitals which are characterized by the three
quantum numbers n m referring to position space and the spin quantum number
ms = 1/2 (orientation of the spin in respect of the z-axis):
ms
n mms (q i ) = n m
s
(ri , i , i )1/2 (si ). (7.26)
Here we have introduced a coordinate q i for each particle, representing the position
vector r i and a symbolic spin variable si . The total wave function for the N particle
quantum state is then written (q 1 , q 2 . . . q j . . . q N ). In Sect. 3.1.2 we have already
mentioned the symmetry properties:
In respect of exchange of two particles, the state vectors (wave functions) of a

quantum system of identical particles are symmetric for bosons, and antisym-
metric for fermions.
ij which exchanges particle i and j one may write

Defining an exchange operator P
this in mathematical form
for bosons
Pij (q 1 ..q i ..q j ..q N ) = (q 1 ..q j ..q i ..q N ) = (q 1 ..q i ..q j ..q N )
(7.27)
and for fermions
ij (q 1 ..q i ..q j ..q N ) = (q 1 ..q j ..q i ..q N ) = (q 1 ..q i ..q j ..q N ).
P
(7.28)
The latter relation is completely equivalent to the PAULI exclusion principle: assume
two identical particles to be in the same quantum state, say particle i and particle j .
In that case exchanging the two obviously leads to
(q 1 ..q i ..q j ..q N ) = (q 1 ..q j ..q i ..q N ). (7.29)
For fermions this contradicts (7.28) unless 0. Thus, two identical fermions
cannot exist in the same spin-orbital or in the classical formulation by PAULI J R .
(1925):
There are never two or more equivalent electrons in an atom for which the
values of all quantum numbers are identical.
In contrast, for bosons no conflict arises between (7.29) and (7.27) so that any num-
ber of identical bosons may be in the same spin-orbital, i.e. have the same set of
quantum numbers.
7.3.2 Symmetries of Spatial and Spin Wave Functions
For a two electron system such as He and He like ions it is convenient to express the
total wave function as a product2
(1, 2) (q 1 , q 2 ) = s (r 1 , r 2 )(s1 , s2 ) s (1, 2)(1, 2) (7.30)
of a spatial part of the wave function, s (r 1 , r 2 ), which we have treated so far

exclusively, and a spin function for the two electrons (1, 2).
2 If no ambiguities can arise, we abbreviate here and in the following r j and sj as well as qj by j .
We may apply the exchange operator onto spatial and spin function separately.
With
12 s (r 1 , r 2 ) = s (r 2 , r 1 ),
P (7.31)
and since the two electrons are indistinguishable, the states s (r 1 , r 2 ) and
s (r 2 , r 1 ) belong to the same, non-degenerate eigenvalue. They may differ only
by a scalar factor:
12 s (r 1 , r 2 ) = s (r 2 , r 1 ) = s (r 1 , r 2 ).
P
Applying this operation twice must reproduce the original state, hence 2 = 1 and
= 1, i.e.
s (r 2 , r 1 ) = s (r 1 , r 2 ). (7.32)
The spatial part of the wave function, may thus be either symmetric in respect of
exchange of the two electrons or antisymmetric. We shall call these wave functions
+s (r 1 , r 2 ) and s (r 1 , r 2 ), respectively.
Since electrons are fermions (spin 1/2), for their total wave function in He
(1, 2) = (2, 1) (7.33)
must hold, specializing (7.28). Thus, with (7.30) obviously follows that
the spatial part s (r 1 , r 2 ) of the wave function, must be symmetric if the spin
function (1, 2) is antisymmetric
vice versa, the spatial part s (r 1 , r 2 ) of the wave function, must be antisymmet-
ric if the spin function (1, 2) is symmetric.
For the spin function we have already developed the necessary tools in Sect. 6.2.5:
two electrons with spin s = 1/2 couple to a total spin
S1 +
S2 =
S (7.34)
with S = 1 or S = 0. The corresponding spin combinations and state vectors

|(1, 2) = |SMS are illustrated in Table 6.1. There, we have identified two dif-
ferent types of spin functions: the symmetric triplet functions with S = 1 and
MS = 1, 0, +1
1 1
= (1)(2) , = (1)(2) and
1 1
0 |(1)(2) + |(1)(2) (7.35)
=
1
2
and the antisymmetric singlet function with S = 0 and MS = 0
0 |(1)(2) |(1)(2)
= . (7.36)
0
2
Again, the numbers (1) and (2) indicate which of two electrons is represented by
the spin function or , with

(1)(2) = 1 1 , (1)(2) = 1 1 and so on
2 2 2 2
being compact writing for |ms1 ms2 = |s1 ms1 |s2 ms2 .
Note the orthonormality of the total spin functions:

M
S S SMS = S S MS MS . (7.37)
The total angular momentum of the system is also constructed from the individual
angular momenta of the two electrons:
L1 +
L2 =
L. (7.38)
For the ground state we have so far only discussed the spatial part of the wave
function (7.20), which is symmetric with respect to exchange of r 1 by r 2 . Thus,
the corresponding spin function must be antisymmetric. Due to the PAULI principle
(total anti-symmetry, at least one quantum number different) no triplet configuration
is possible for the ground state and the complete ground state wave function is
s |(1)(2) |(1)(2)
= 1s
1s 2 (1, 2) 2 (r 1 , r 2 ) . (7.39)
2
In this case inclusion of the spin coordinates is just a formality which does not
change the energy, the H AMILTON operator (7.6) as such being independent of the
spins.3 In summary, the ground state of He with 1 = 2 = 0, a total spin S = 0 and
a total orbital angular momentum L = 0 is a singlet 1s 2 1 S0 state.
We use here the terminology introduced in Sect. 6.2.6, i.e. we characterize the
total spin and the total orbital angular momentum with capital letters S and L and
designate the states of He by n1 1 n2 2 2S+1 LJ . As already mentioned in Sect. 6.2.6
this coupling scheme is called RUSSEL -S AUNDERS coupling (or LS coupling). An
alternative would be the jj coupling, where first the spin s and the orbital angular
momentum of each individual electron couple to a single electron total angular
momentum j ; and then the resulting j i would couple to a total angular momentum
J of both electrons. In Sect. 7.3.4 we shall further analyze why He chooses to be
LS and not jj coupled.
In general, the total wave functions for He in LS coupling are
singlet states S (1, 2) = +s (1, 2)00 (1, 2) or (7.40)

M
triplet states T (1, 2) = s (1, 2)1 S (1, 2) with MS = 1, 0, 1. (7.41)
3 We mention, however, a very remarkable property of this description of singlet states: the wave
function (state) is not separable, i.e. we cannot write it as a simple direct product of states from the
two separated electrons. One calls such states entangled, see also Appendix E.3.
We specialize this now for the excited states of He and consider electron con-
figurations {n1 1 n2 2 } for which n1 = n2 . In the following we shall approximate
the corresponding spatial part of the wave functions by products of orbitals of the
type ns1 1 (r 1 )ns (r 2 ). Since both electrons are in principle indistinguishable the
2 2
spatial orbitals have to be combined to symmetric or antisymmetric linear superpo-
sitions:
1
+s (1, 2) = ns1 1 m1 (r 1 )ns2 2 m2 (r 2 ) + ns1 1 m1 (r 2 )ns2 2 m2 (r 1 ) (7.42)
2
1
s (1, 2) = ns1 1 m1 (r 1 )ns2 2 m2 (r 2 ) ns1 1 m1 (r 2 )ns2 2 m2 (r 1 ) . (7.43)
2
We want to emphasize that this kind of ansatz is already an approximation, the so

called independent particle model is used here. For the ground state of He it is not
a very good approximation: as we have seen in Sect. 7.2.4 and Sect. 7.2.5 the two
electrons are directly correlated. However, for singly excited states the independent
particle model turns out to describe the system rather well.
7.3.3 Perturbation Theory for (Singly) Excited States
We now focus on such excited states in which one of the electrons remains in its 1s
ground state configuration |100 , the other one, however, is raised into an excited
state |n m . With the Hamiltonian (7.6) we obtain in 1st order perturbation theory
for singlet (7.40) and triplet states (7.41) a total energy
ee (r12 )|S = W1 + W2 + Hee +

WS = W (0) + S |H
(1)
and
(7.44)
ee (r12 )|T = W1 + W2 + Hee
WT = W (0) + T |H
(1)
,
(1)
respectively. We have to compute the diagonal matrix element Hee which deter-
mines the energy shift in respect of the undisturbed system (W (0) = W1 + W2 ).
Since the H AMILTON operator does not depend explicitly on the spin and since the
spin functions (7.37) are orthonormal, they can be pulled outs of the integral:

(1) ee |S,T = s (1, 2) 1 s (1, 2) .
Hee = S,T |H
r12
Inserting (7.42) and (7.43) leads to
1 s
(1)
Hee =
s
100 (r 1 )n m (r 2 ) 100
s s
(r 2 )n m (r 1 ) (7.45)
2
1 s
(r 1 )n ms
(r 2 ) 100
s s
(r 2 )n m (r 1 ) .
r12 100
1L
Kn
Kn
Jn 3L
1sn
Fig. 7.5 Singly excited He(1sn ) state with total angular momentum quantum number L: the
term is raised by C OULOMB screening Jn and split into one singlet 1 L and three degenerate
triplet states 3 L by exchange interaction Kn
One easily verifies that this perturbation term can be written as

(1)
for singlet terms Hee + = Jn + Kn and (7.46)
(1)
for triplet terms Hee = Jn Kn , (7.47)
with the C OULOMB integral

s
Jn = (r 1 )2 1 s (r 2 )2 d3 r 1 d3 r 2 (7.48)
100 n m
r12
and the exchange integral

1 s
Kn = s
100 s
(r 1 )n m (r 2 ) 100 s
(r 2 )n m (r 1 )d3 r 1 d3 r 2 . (7.49)
r12
The influence of Jn and Kn on the total energy is illustrated in Fig. 7.5.
The C OULOMB integral can be interpreted in a rather evident manner. For clarity
we briefly switch to SI units and introduce
2
the charge density of electron 1 at position r 1 (r 1 ) = e s (r 1 ) and
2
the probability to find electron 2 at position r 2 w(r 2 ) = s (r 2 ) .
With these we rewrite (7.48) as

" #
100 (r 1 )
Jn = d3 r 2 wn m (r 2 ) e d3 r 1 . (7.50)
40 r12
One may thus read the C OULOMB integral as follows: {} is the repulsive interaction
energy of electron 2 in position r 2 with the charge density of the other electron in
state |100 integrated over the whole charge distribution of electron 1. The exter-
nal integration just averages this repulsion over the probability to find 2 anywhere in
space. In total, Jn is the overall electrostatic repulsion energy of the two electrons.
It leads to a raise of the terms as sketched in Fig. 7.5. In contrast, the exchange inte-
gral Kn is not open to an intuitive interpretation. Albeit also of electrostatic nature
(1/r12 term) it is generated by a typical quantum mechanical effect, the exchange
of the two electrons as a consequence of symmetrizing or anti-symmetrizing the
spatial part of the wave function. This becomes particularly evident if one formally
rewrites the perturbation (7.46) and (7.47) as
1
ee
H (1)
= Jn (1 + 4
S 1
S 2 )Kn , (7.51)
2
so that the total Hamiltonian becomes
=H
H 2 + Jn 1 (1 + 4
1 + H S 1
S 2 )Kn . (7.52)
2
One easily verifies the identity of this expression with (7.6) using

1 2 2 2 1 2 3 3 1 2
S1S2 = S S1 S2 = S = 2 S 3
2 2 4 4 4
and the eigenvalue equation for the total spin

S =
S1 +
S 2 (in a.u. =
)
2
S |SMS = S(S + 1)|SMS .
One finds that the expectation value of (7.51) is indeed identical with (7.46) and
(7.47), respectively, depending on whether S = 0 or 1.
One speaks about exchange interaction almost like its own kind of force which
results from the anti-symmetrization rule for wave functions of fermions. It is this
exchange interaction which enforces the splitting of singlet and triplet terms. The
magnitude of this interaction energy is given by the exchange integral Kn that is
by a purely electrostatic interaction. The states which diagonalize the H AMILTON
operator obviously have to be such that also S 1
S 2 is diagonal, which implies diag-

onalization of S . Our initial ansatz to first couple the spins through S = S1 +
2
S2

thus finds its justification by the operator 2 (1 + 4S 1 S 2 )Kn which enters the Hamil-
1
tonian. We shall expand this somewhat more precisely in the next section.
We emphasize again that, according to the scheme sketched in Fig. 7.5, triplet
states generally have lower energies than singlet states (for otherwise identical quan-
tum numbers). A second look at Fig. 7.2 confirms this statement. The physical origin
is also quite clear after the preceding discussion: in triplet states the spin function
is symmetric, the spatial part of the wave function is antisymmetric. However, this
implies that the probability for both electrons to come close to each other is smaller
than for a symmetric spatial function. In a triplet state, the two electrons are never
found at the same position simultaneously. Consequently, the average repulsive en-
ergy is smaller in triplets than in singlets, and hence triplet states lie lower than
singlet states. Very generally, H UNDs rules hold:
For a given electron configuration the states with the highest multiplicity
2S + 1 have the lowest energy.
Among these, the states with highest L have the lowest energy.
We point out that the special form (7.52) of writing the H AMILTON operator ex-
plicitly with exchange interaction is valid also for multi-electron systems where one
has to sum the S i

S j term over all pairs of electrons. This Hamiltonian plays a key
role in the H EISENBERG model of solid state ferromagnetism: it is the magnitude
of the exchange interaction which determines whether it is energetically favourable
for the spins to orient parallel to each other and hence whether a material can be
ferromagnetic or not.
Finally, we mention a recent systematic and quantitative comparison between
experiment and the above 1st order approximation for the 1sn series in singly ex-
cited He. H UTEM and B OONCHUI (2012) explicitly evaluated the C OULOMB and
exchange integrals, using hydrogen orbitals and spherical harmonics series. They
found e.g. for the first excited 1s2s 1 S and 3 S states a discrepancy of 5.3 % and
2.4 %, respectively, for the 1s2p 1 P and 3 P states 7.2 % and 4.1 % (as expected
with the 1st order results higher than the experiment). For larger n the deviations
get significantly smaller, but are still more than 1 % at the n = 4 level. Obviously, at
the present level of precision in modern spectroscopy, a significantly better approxi-
mation is required. We shall come back to state-of-the-art computations in Chap. 10.
7.3.4 An Afterthought
What kind of force acts on the spins? It is definitively not magnetic, since, so far,
we have not included any magnetic moments of the spin, spin orbit interaction and
the like. Still, the spins prefer obviously to be oriented parallel or antiparallel; the
following considerations should help to understand why.
Turning this question around, one is tempted to search for a total wave function
differing from (7.30) which, nevertheless, is consistent with the anti-symmetrization
rule (7.33). One possible way to realize such states from a linear combination of
orbitals are so called S LATER determinants. They will be discussed and used in
some depth in Chap. 10. For the present case of a singly excited He atom in the
configuration {1s, n } the following expressions are possible:
s (1)(1)
1 s (1)(1) n
1 (1, 2) = 1s
2 1s (2)(2) n (2)(2)
s s
s s (1)(1)
1 1s (1)(1) n
2 (1, 2) = s s
2 1s (2)(2) n (2)(2)
s s (1)(1)
1 1s (1)(1) n
3 (1, 2) = s (7.53)
2 1s (2)(2) n (2)(2)
s
s s (1)(1)
1 1s (1)(1) n .
4 (1, 2) = s
2 1s (2)(2) n (2)(2)
s
The numbers in brackets () refer again to the coordinates of electron 1 and 2. Obvi-
ously one may pull out (1)(2) and (1)(2) from 1 (1, 2) and 4 (1, 2), respec-
tively. These linear combinations of atomic orbitals are thus identical to the triplet
states T (1, 2) = s (1, 2)11 (1, 2) with an antisymmetric spatial wave functions
according to (7.41) just as used in the previous description of the He excited states.
However, the states 2 (1, 2) and 3 (1, 2) cannot be identified with any of the states
used so far albeit antisymmetric beyond any doubt: exchange of electron 1 and
2 implies interchanging two rows in the determinants, which by definition changes
the sign of the determinant.
To understand the significance of these state functions 1 (1, 2) to 4 (1, 2) let
us use them to express the H AMILTON operator (7.6) in matrix representation. By
simply multiplying this operator from the left with j (1, 2)| and from the right
with |i (1, 2) and making use of the orthonormality relation (7.37) for the spin
functions we obtain the 4 4 matrix

W1s + Wn + Jn Kn 0 0 0
0 0 0 0
H=
0 0 0 0
0 0 0 W1s + Wn + Jn Kn

0 0 0 0
0 W1s + Wn + Jn Kn 0
+0
. (7.54)
Kn W1s + Wn + Jn 0
0 0 0 0
We have written it in terms of a diagonal and a non-diagonal part. One readily ver-
ifies that Jn and Kn are indeed the C OULOMB and exchange integrals, (7.48) and
(7.49), while W1s and Wn are the single particle energies in the 1s and n state, re-
spectively. Obviously, the S LATER determinants (7.54) diagonalize the Hamiltonian
only partially, they are thus not a complete set of eigenfunctions. We may, however,
try to diagonalize the non-diagonal term, i.e. the 2 2 block in the middle of the
second matrix in (7.54). Using the standard procedure we write the characteristic
equation
W) = 0
det(H
from which we obtain two solutions for W , the two missing eigenenergies W =
W1s + Wn + Jn Kn . And we happily note that this result is exactly the same 1st
order result (7.44) with (7.46) and (7.47) that we have obtained with the correct
state functions (7.40) and (7.41), respectively. The latter ones, i.e. +s (1, 2)00 (1, 2)
and s (1, 2)10 (1, 2), are of course also the result of the diagonalization procedure
which we have sketched here: they simply are linear combinations of 2 (1, 2) and
3 (1, 2) to the eigenvalues W+ and W while 1 (1, 2) and 4 (1, 2) are already part
of the triplet system.
Summarizing, it is the exchange interaction, recognized in (7.54) as the off-
diagonal term Knl , which enforces the parallel or antiparallel orientation of the
electron spins! Combination of spin-orbitals which do not diagonalize the spin and
orbital angular momentum simultaneously are no eigenvalues of the H AMILTON
operator under discussion.
Of course this statement is fully based on the fact that exchange interaction
is by far the dominant perturbation as assumed in (7.54). For high Z this will
change significantly with spin-orbit (or LS) interaction increasing. As we shall see
in Sect. 10.4.1 LS interaction may indeed become of the same order of magnitude
or even larger than exchange. Then it will be reasonable to diagonalize the Hamil-
tonian first in respect of the dominant LS interaction and later on add exchange
interaction a minor perturbation. This explains without effort the transition from
RUSSEL -S AUNDERS (or LS coupling) to jj coupling for atoms with high nuclear
charge Z.
Section summary
The two electrons in the He atom are in principle indistinguishable, hence
proper antisymmetrization of the He wave functions according to the PAULI
principle is mandatory. This leads to singlet and triplet formation, with an-
tiparallel and parallel electrons spins, respectively.
Consequently, the He 1s 2 ground state can only be a singlet state as the spatial
quantum numbers of the two electrons are identical.
In contrast, for each singly excited 1sn configuration a singlet and a triplet
state exists. To a good 1st order approximation their spatial part is described
by the independent electron model as product wave functions of the respective
single electron orbitals, symmetrized and antisymmetrized according to (7.40)
for singlet and triplet states, respectively.
A reasonable estimate for the energy of these states is obtained in 1st order
as W1sn = W1s + Wn + Jn Kn with the C OULOMB integral, Jn , ac-
counting for the repulsion of the two electrons, and the exchange integral,
Kn , characterizing the electron exchange energy.
We recognize the exchange energy Kn to be responsible for the electrons to
orient antiparallel or parallel as singlet or triplet system, respectively.
7.4 Fine Structure
Closer inspection shows of course that He levels with S, L > 0 may also have a fine
structure (FS) due to spin-orbit (LS) interaction quite analogous to that discussed
in Chap. 6. One finds that only the He triplet levels (S = 1) are split while the singlet
states (S = 0) remain un-split (as the name indicates). The order of magnitude of FS
splitting in He is similar to that in atomic hydrogen. The largest splitting is found
for the He 1s2p 3 P state (overall ca. 1.06 cm1 or 0.00013 eV, see inset in Fig. 7.2).
This has to be compared with the singlet-triplet splitting between the 21 P and 23 P
state (2048 cm1 or 0.25 eV). FS interaction is indeed a very small perturbation in
comparison to exchange interaction and the description of the He levels in terms of
singlet and triplet states is an excellent approximation.
Thus, for a description of the FS states we have to start now with the total spin

S = S1 + S 2 and the total orbital angular momentum L= L1 + L2 for the two elec-
trons which are already coupled by exchange interaction as just discussed. Under
the influence of LS coupling a total angular momentum J = L + S of the system
7.4 Fine Structure 361
Fig. 7.6 Fine structure 2

1s2 p 3 P oJ J J
splitting of the 1s2p 3 PJ state 0 0
in He and He like ions. Note
0.9879 3.13 950
the inversion of the triplet
states for small Z. All 2
energies in cm1 , different 1 1
2 2.27 0
scale for each of the three 0.0764 1 150
systems
He I Z = 2 Li II Z = 3 F VIII Z = 9
is formed for each configuration {n1 1 n2 2 LS}. The rules which have been used in
Sect. 6.2.5 for the angular momenta of individual electrons apply with appropriate
modifications. The corresponding eigenstates of the two electron system which di-
agonalize J and Jz are constructed in full analogy to (6.53) and may be written
2
explicitly

(LS)J MJ = ( 1 2 )LML (s1 s2 )SMS LML SMS (LS)J MJ . (7.55)
m ,ms
This combined coupling scheme for the total L and S angular momenta is called
RUSSEL -S AUNDERS coupling (or LS coupling), as already mentioned. Again, as in
the single electron case, the multiplets split into 2S + 1 components (if S L).
However, the magnitude of the splitting can no longer be derived simply from
(r)LS as in the single electron case (at least not for low Z). The Hamiltonian
must include all magnetic interactions between individual electrons and will contain
terms like

HLS = Li
i (r i ) Si (7.56)
but also all terms of the type L2

S1, L1 S 2 (spin other orbit) as well as spin-spin
interaction. The latter is not simply S 1S 2 but has to account for the correct dipole-
dipole interaction. In Chap. 9 we shall discuss such kind of problems in some detail
in the context of hyperfine interactions between electron spin and nuclear spin. The
ensuing computations are rather complicated and cannot be described any longer by
the simple interval rule (6.67). We do not want to go into the details here, but note
that the deviations are particularly strong for small Z, specifically for He. Figure 7.6
(left) shows this for the case of the first excited triplet state 2 3 P with the highest FS
in He. The term positions are completely inverted in this case, i.e. the highest J has
the lowest energy. For comparison the corresponding triplet state splittings for the
He like ions Li+ and F7+ are also shown. In the case of the fluorine ion at least
the ordering of the states is again normal. Nevertheless, the interval rule (6.67),
according to which the distances between the FS levels should be proportional to
the higher J is not really fulfilled.
For larger Z (and singly excited states) the spin-own-orbit interaction of the
L2
2 (r 2 ) S 2 type dominates (2 refers to the excited electron) and the FS splitting
can be expressed again approximately by
a
L2
VLS = 2 (r 2 ) S 2 = J (J + 1) L(L + 1) S(S + 1) (7.57)
2
Fig. 7.7 Fine structure Be I 1s 2 2s2p 3 P oJ Mg I [Ne] 3s3p 3 P oJ

splitting of the alkaline earth
J=2 J=2
metals beryllium (Z = 4) and
magnesium (Z = 12). The
terms are normally ordered, 2.345 40.614
i.e. for higher J the energy is
higher. Note that for Mg the
1
energy splittings are about
1 20.059
WFS J , i.e. the L AND
interval rule holds 0 0.645 0
in analogy to the one electron systems. We illustrate this in Fig. 7.7 for the example
of alkaline earth metals. Their role in respect of He corresponds to that of the alkali
atoms in respect of the hydrogen atom: they have two valence electrons on top of a
closed shell core (see also Sect. 7.7.1). One sees that for Mg the interval rule holds
again very well.
We recall, however, that fine structure interaction grows according to (6.66) with
Z 4 /n3 . For large Z and not too large n it may become substantial. We shall come
2
back to this again in Chap. 10 and find that RUSSEL -S AUNDERS coupling breaks
down for large Z.
Section summary
For He (as well as for other light atoms) spin-orbit interaction (7.57) is much
smaller than exchange interaction (7.49). Hence, for light atom spin-orbit
coupling can be considered a small perturbation which does not change the
coupling of the spins.
A quantitative interpretation of the fine structure of two electron systems thus
starts with total spin S =S1 + S 2 and total orbital angular momenta L=

L1 + L2 which couple in turn to the overall angular momentum of the system

J = L + S.
The overall magnitude for FS splitting in the case of He is on the same order
of magnitude as for H. Unfortunately, the simple formulas used there for the
splitting as well as the L AND interval rule cannot be applied in general (for
light atoms): spin-spin interaction and spin-other-obit interactions have to be
accounted for.
7.5 Electric Dipole Transitions
At the beginning of this chapter we have already introduced selection rules for E1
transitions in He. We want to explore some basic questions about photon absorp-
tion and emission in multi-electron systems, using the two electron He system as a
(hopefully) still transparent example. We now have to expand our definition of the
dipole operator (4.54). For one electron at position r i the dipole moment was just
7.5 Electric Dipole Transitions 363
D i = er i . For N active electrons the perturbation energies (4.55) for all potentially
active electrons have to be added:
/N 0
ieE0
U (r, t) = D i E(t) = r i eeit + e eit . (7.58)
2
i=1
Here E(t) is again the electric field component of the wave and E0 represents the
amplitude. The dipole transition operator (4.56) for a multi-electron system is cor-
N (i)
respondingly D = i=1 D with D(i) = (D i e)/e = r i e. Time dependent per-
turbation theory has to be applied in full analogy to the one electron problem. Only
the dipole transition matrix element is somewhat more complicated.
The following considerations are based on the independent particle model. For
He dipole transition matrix element has the form:

Dba = b (1, 2)r 1 + r 2 a (1, 2) e.
(7.59)
Explicitly with the wave functions according to (7.40) and (7.41) we obtain
s s
(1, 2)Sb b (1, 2)r 1 + r 2 a
MS

Dba = b
M
(1, 2)Sa Sa (1, 2) e,
abbreviating again (r 1 ) = (1) and (r 2 ) = (2). We first note that neither r 1 nor r 2
act directly on the spin component of the wave function. We may thus pull the latter
out of the integral and apply the orthonormality relations for the spin functions:
s s MS M

Dba = b (1, 2)r 1 + r 2 a (1, 2) Sb b (1, 2)Sa Sa (1, 2) e (7.60)
s s
= b (1, 2)r 1 + r 2 a (1, 2) eSb Sa MSb MSa .
This is a very important result: the spin state remains unchanged by E1 tran-
sitions in He. In particular, there are no E1 transitions between singlet and triplet
system (these so called intercombination lines are forbidden). Correspondingly, the
+ and symmetry of the spatial wave functions remain also unchanged in E1 tran-
sitions (this holds even for E transitions of all orders since the spin is not involved).
This selection rule thus explains the experimental observation of the two quasi iso-
lated systems of lines historically para and ortho helium, i.e. transitions occur
only within the singlet or the triplet system.
For He this rule holds rather strictly. For He like ions with large Z one observes
more and more also (weak) intercombination lines: as Z increases, spin-orbit in-
teraction (7.57) also increases and has to be accounted for. This leads to a mixing
of the pure singlet or triplet character of states in the independent particle model
so that triplet and singlet components are added to the respective other multiplicity
(so called configuration interaction, CI). For this very reason intercombination lines
also become possible.
Beyond the forbidden intercombination lines the model of independent particles
supports another important selection rule: allowed are only pure one electron transi-
tions even within the singlet or triplet system i.e. only one of the quantum numbers
is allowed to change in a one-photon E1 transition. The proof is quite instructive:
In the independent particle model the spatial part of the wave function is writ-
ten as symmetric or antisymmetric products of one electron orbitals according to
(7.42) or (7.43), respectively. Let the initial configuration be {a} = {1a, 2a} =
{n1a 1a m1a , n2a 2a m2a }, the configuration after the transition {b} = {1b, 2b} =
{n1b 1b m1b n2b 2b m2b }. The dipole transition matrix element between these is
s s

Dba = b (1, 2)r 1 + r 2 a (1, 2) e (7.61)
1 s
= 1b (1)2b s
(2) 1b
s s
(2)2b (1)r 1
2

+ r 2 s (1) s (2) s (2) s (1) e.
1a 2a 1a 2a
We note that r 1 contributes to the matrix element only through that part of the wave
function which depends on r 1 while r 2 is only relevant for the wave function de-
pending on r 2 . Hence the components of (7.61) factorize. This results in expressions
of the type
1 s s s s

Dba = 1b (1)1a (1) 2b (2)r 2 e2a (2)
2
1 s s s s
1b (1)2a (1) 2b (2)r 2 e1a (2)
2
1 s s s s
2b (1)1a (1) 1b (2)r 2 e2a (2)
2
1 s s s s
+ 2b (1)2a (1) 1b (2)r 2 e1a (2) +
2
and corresponding expressions with r 2 replaced by r 1 and vice versa, leading to
identical results (one has to integrate over the whole space). Recalling that the one
electron wave functions are orthonormal this implies
s s

Dba = 1b1a 2b (r)r e2a (r) (7.62)
s s
1b2a 2b (r)r e1a (r)
s s
2b1a 1b (r)r e2a (r)
s
+ 2b2a 1b (r)r e|1a
s
(r) ,
i.e. one of the quantum numbers has to remain the same before and after the tran-
sition. What remains of the double integrals over r 1 and r 2 are just single elec-
tron transition matrix elements: they are determined by the standard selection rules
which we have developed in Sect. 4.4 for one electron systems. Thus, the model of
independent particles allows only single electron transitions, i.e. only such transi-
tions are allowed where one electron changes its quantum number while all others
remains in its initial configuration.
Specifically, if one electron is originally in the ground state 100, only transition
{100n m} {100n m } or {100n m} {n m n m} may occur for which
the usual selection rules = 1 and m = 1, 0 hold, together with Sb = Sa
7.6 Double Excitation and Autoionization 365
the prohibition of intercombination transitions. Note however, that the selection

rule (7.60) for the spin projection quantum numbers MS has to be replaced for fine
structure transitions within the triplet system by the J and MJ rules which we
have outlined in Sect. 6.4.
Section summary
The dipole transition operator for an N electron system is given by D=
N (i)

i=1 D , with D = r i e.
(i)
Exploiting the structure of the wave functions one finds that E1 transitions
between singlet and triplet system (intercombination lines) are strictly forbid-
den.
The independent electron model allows only E1 transitions in which the quan-
tum numbers of only one electron change.
The usual = 1 and m = 1, 0 selection rules hold. In addition, for fine
structure transitions J = 0, 1 with 0 0 must hold.
7.6 Double Excitation and Autoionization
7.6.1 Doubly Excited States
In the previous discussion of excited He states at least one of the electrons was as-
sumed to be in its 1s ground state, the other one in an excited n state. However,
one may very well envisage configurations where two electrons are excited irre-
spective of the question how such doubly excited states might be populated. The
independent particle model predicts in 0th order perturbation theory energies given
by (7.15). For the example of a series with the configuration 2 n (n 2) this
would be (in a.u.)

(0) Z2 1 1 Z2
W2 n = + 2 .
2 4 n 4
For He (Z = 2) these doubly excited states are thus expected at energies 1Eh =
27.2 eV, while the binding energy of the He+ ion is 2Eh = 54.4 eV. Hence,
the doubly excited states {2 n } lie in the ionization continuum of He+ + e and
converge to the excited He+ (2s) state at 2Eh /4 = 13.6 eV. The {3 n } se-
ries has energies 13Eh /18 = 19.6 eV and converges to the He+ (3s) state at
2Eh /9 = 6.0 eV.
C OULOMB and exchange interaction raise the lower limits of these series, but
the general feature remains: these states are embedded in the ionization continuum.
The expected limits for these series are schematically indicated in Fig. 7.4 right (red
horizontal lines, marked 2s and 3s ). Series of absorption lines are indeed
observed experimentally in an energy range indicated by the grey shaded areas in
Fig. 7.4 on the left, marked autoionization (upper limits shown as black lines).
7.6.2 Autoionization, FANO Prole

Autoionization and related resonance phenomena are an important and fascinating
topic not only in atomic physics. It is thus important to recognize the characteristic
features and to understand the physics behind the experimental observations.
At first sight one wonders how such states might be populated at all, since dou-
ble excitation (7.62) is forbidden for E1 transition in the independent particle model.
However, in reality the concept of independent particles is obviously far from being
perfect. We have seen this already for the ground state. In the case of doubly ex-
cited states the two electrons may get rather close to each other. Thus, it is no longer
sufficient to account for the other electron by an averaged screening term. Modern
theoretical calculations include the electron correlation by different techniques, e.g.
by CI, i.e. by linear combination of several configurations, or very elegantly
by a judiciously chosen, non-Cartesian coordinate system, where the distance be-
tween the two electrons is one of the free coordinates. Using such correlated wave
functions one may indeed describe two electron E1 transitions of the type

h + He 1s 2 He n n . (7.63)
Beautiful examples for such spectra are obtained in absorption experiments with
high resolution synchrotron radiation. A few samples, the 2 n and 3 n series,
are shown in Fig. 7.8. Note that the characterization of the individual lines em-
phasizes already the complexity of the correlated electronic states which cannot be
fully described by the configuration {2 n } and {3 n }. We cannot enter into the
computational details here.
Figure 7.8 immediately raises the question about the origin of the peculiar line
shapes observed. The experiment directly records the He+ ion current as a function
of the photon energy from a highly monochromatized synchrotron radiation beam
which is passed through a He filled cell.4 We first note that the observed lines, i.e.
the He(n n ) = He states,5 are energetically indeed embedded in the ionization
continuum of helium, He+ (1s) + e . The two series 2 n and 3 n of doubly
excited states shown here converge for n towards the corresponding excited
ionic states He+ (2 ) and He+ (3 ), respectively.
These doubly excited states may in principle decay through optical transitions to
a lower lying state by emission of a photon. Alternatively and indeed much more
efficiently they can decay simply by ionization since they lie energetically within
the ionization continuum:
He He+ (1s) + e . (7.64)
Such process in which a state embedded into an ionization continuum decays by
electron emission is called autoionization. The second electron returns here to the
ground state, now of the He+ (1s) ion. The photon energy initially deposited
4 Note that the signals shown are not generated by differentiation as often done in spectroscopy for
better assignment of line centres the genuine line shapes are shown here.
5 Two asterisks, short for double excitation.
n = 2+ He (sp, 2 n ) 1 P o WI (n = 2)
65.405
3- 3+
4+ 5+
x 15 6+
4- 5- 7+
photoionization yield
60 61 62 63 64 65
He (3,1n )
WI (n =3)
4 5 72.963
n=3 6 14
15
q =1.45 8 16
17
19
3,-13
= 0.21
Wr = 69.88
70 71 72 72.88 72.90 72.92 72.94
excitation energy / eV
Fig. 7.8 He autoionization spectra recorded with high resolution synchrotron radiation; top:
(sp, 2n) series (D OMKE et al. 1991), bottom left: (3, 1n ) series, inset bottom right: enlarged
detail (S CHULZ et al. 1996). Red vertical lines indicate the limits He+ (n = 2) and He+ (n = 3)
towards which the two series converge. For the state (3, 13 ) the red line shows by way of example
a FANO profile fit (7.70) with the parameters q, and Wr
for exciting the He state is partially used for ionization (creating He+ (1s) from
He(1s 2 )) while the excess energy, WI [He(1s 2 )], is carried as kinetic energy
Wkin by the ejected electron.
The peculiar shape of the autoionizing lines documented very neatly in Fig. 7.8
is explained as a special kind of interference: the final state He+ + e (Wkin ) may
in principle be reached from the ground state through two different channels:
either one induces the ionization process directly

+ He 1s 2 He+ (1s) + e (Wkin ), (7.65)
or one excites one of these doubly excited states which subsequently decays into
the continuum:

+ He 1s 2 He He+ + e (Wkin ). (7.66)
In both cases the total energy balance is given by:

= Wkin + WI He 1s 2 .
The experimentally observed line shape results from the fact that both channels
(direct ionization and ionization via the resonance He ) can by principle not be
distinguished in the experiment a situation which is characteristic for any kind
of interference experiment, we may e.g. think of YOUNGs double slit experiment.
Let us call the probability amplitudes for the two processes cd and cr , respectively,
written as

cd = Aei and cr = cr (W )ei(W ) . (7.67)
According to the general principles of quantum mechanics these amplitudes for the
two indistinguishable processes have to be added coherently. The probability for
observing an absorption signal is then given by
2
S(W ) = |cd + cr |2 = Aei + |cr |ei (7.68)
= A2 + |cr |2 + 2A|cr | cos( ).
Close to resonance the phase angle changes rapidly and one expects a character-
istic interference pattern as indeed observed in the autoionization spectra Fig. 7.8.
A quantitative treatment of this phenomenon has been described for the first time
by FANO (1961). Briefly, one considers amplitude A and phase for the direct pho-
toionization process to be essentially constant and independent of the photon en-
ergy W = over an energy range relevant for a single resonance. For cr (W ) one
assumes the typical behaviour of a (complex) resonance amplitude expected when
W passes through the resonance, as e.g. for a harmonic oscillator. We have already
discussed this in the context of ordinary optical resonance absorption in Sect. 5.1.2.
As illustrated in Fig. 5.2(a), the magnitude |cr (W )| rises and decreases rapidly close
to a resonance at energy Wr , its linewidth being , while the phase changes from
zero to . Due to the superposition of cr (W ) with cd one obtains indeed very pro-
nounced line shapes as observed in the experiment. The famous FANO line profile
(more precisely the B EUTLER -FANO profile, B EUTLER being the spectroscopist
who first observed such line shapes) is given by6
(q + )2 W Wr
with = (7.70)
1 + 2 /2
with the linewidth , the reduced energy , and the FANO line shape parameter q.
The latter includes the relative phase and coupling strength between the direct and
the resonant process. We have fitted the He(3, 13 ) line shown in the experimental
data Fig. 7.8 by way of example with this profile and obtain a nearly perfect fit! In
the Sect. 7.6.3 we investigate in a systematic manner different forms of such FANO
resonance profiles and try to understand them in some more detail.
6 For fitting a measured signal the original formula has to be slightly generalized, e.g. to

C (q + )2 C
S() = + A2 2 , (7.69)
q2 +1 1+ 2 q +1
accounting for the background signal A2 and the strength C of the resonance.
Fig. 7.9 Resonance Im

amplitude cr () in the = 0 (resonance)
complex plane, B = 1 1.0
=1 0.5 = -1
>>1 << - 1
Re
- 0.5 0 0.5
7.6.3 Resonance Line Proles
Resonances of the type discussed here are observed in many areas of physics. One
finds them in optical spectroscopy as discussed here, in atomic and molecular col-
lision physics with electrons, ions and atoms, but also in nuclear physics or high
energy physics, as well as in solid state physics. They always occur when a final
state may be reached on different pathways, one of which involves a resonant, quasi
stable state. To understand the behaviour described by the FANO profile (7.70) we
visualize the excitation amplitudes involved in the complex plane.
With the normalized energy according to (7.70) one writes magnitude |cr ()|
and phase () of the resonance amplitude cr or alternatively real and imaginary
part according to (5.14a)(5.15b), scaled with a resonance strength B:
1
cr () = B and tan () = or (7.71)
2 + 1
|cr ()|2 |cr ()|2
Re(cr ) = and Im(cr ) = . (7.72)
B B
We have already illustrated the resonance amplitude in Fig. 5.2. Figure 7.9 shows
an alternative visualization in the complex plane: with (7.72) one easily verifies:

2
2 B 2 B
Re(cr ) + Im(cr ) = . (7.73)
2 2
Thus, when the reduced energy is tuned through resonance ( = 0), cr rotates in
the complex plane on a circle with radius B/2 around a centre at (0, iB/2).
We have now to add this resonance amplitude cr () to the (constant) direct am-
plitude cd as suggested by (7.68). With (7.71) the profile of the resonance line (i.e.
the absorption or ionization cross section) becomes:

i i() 2 i B iB 2

(W ) Ae + cr () e
= Ae + 2 + . (7.74)
+ 1 2 + 1
Autoionization resonances are observed close to resonance 0 (W = Wr )
on a background from direct ionization. Depending on the phase and the relative
magnitude A/B of the direct amplitude vector addition of cd and cr () in the com-
(a) = 90 o (b) = - 26 o (c) =206 o (d) = - 90 o

A = 0.25 A = 0.56 A = 0.56 A = 1.00
Im ii
ii ii ii 0.5
-0.5
0.5 0.5
iii 0.5 i iii i iii i
iii 0.5 i
0
1 -1 0
-0.5 0.5 Re
S( ) 1.5 ii
1.25 1.25 1.25
i iii i ii ii iii
i iii
iii i ii
-10 0 10 -10 0 10 -10 0 10 -10 0 10
Fig. 7.10 Top row: absorption amplitudes (full red arrows) in the complex plane as a sum of
direct ionization amplitude cd = A exp(i) (dashed black arrow) and resonant amplitude cr ()
(dashed red circle) according to (7.71) with B = 1. Bottom row: Corresponding absorption profiles
(ionization signals) S(). Depending on the parameters very different FANO line shapes emerge as
a function of reduced energy = (W Wr )/( /2). In each case the complex amplitudes and the
profiles are marked for different energies i < ii < iii
plex plane generates different line profiles as illustrated in Fig. 7.10.7 The shapes
range from (a) a pure absorption profile for a phase angle = 90 (or whenever
the direct amplitude vanishes), via dispersion like profiles (b) and (c) to completely
destructive interference (d), i.e. to a transparent window in the absorption spectrum
cd and cr () are of equal magnitude and opposite phase at resonance.
Section summary
Doubly excited states in He, embedded in the ionization continuum of He+ +
e may be excited by E1 due to a break down of the independent electron
model.
Pronounced interference structures are observed due to coherent superposi-
tion of the amplitudes for direct ionization and autoionization. These can be
described by the FANO profile (7.70).
It may be rationalized by addition of a direct and a resonance amplitude. While
the former is constant, the magnitude of the latter changes rapidly over the
resonance according to (7.71).
7 One may derive from (7.74) with some algebra the modified FANO formula (7.69) with

2
4B B C/(AB) + B/A + 2 sin
C = AB 4 + sin + and q = for cos = 0
A A 2 cos
q = 0 for = /2 and q for = /2; C and thus q are in principle double valued.
7.7 Quasi-two-Electron Systems 371
7.7 Quasi-two-Electron Systems
7.7.1 Alkaline Earth Elements
The alkaline earth metal atoms are in respect of the He atom in the same role as the
alkali atoms in respect of atomic hydrogen: their two potentially active valence elec-
trons are screened from the nuclei by Z 2 electrons in the completed filled ionic
core levels. In full analogy to the alkali metal atoms (Sect. 3.2.3) with one electron,
the Hamiltonian for two valence electrons of alkaline earth atoms is essentially that
of He, except that in (7.6) the pure nuclear C OULOMB potential has to be replaced
by a screened one which approaches 2/r1 (and 2/r2 ) for sufficiently large distance
from the nucleus. Correspondingly, the term schemes and spectra are very similar to
those of He, but in detail even richer, including also several series of doubly excited
states. They are very well documented by K RAMIDA et al. (2013): energy levels,
wavelength tables with transition probabilities, G ROTRIAN diagrams and simulated
spectra can be obtained within seconds. Thus, here we show in Fig. 7.11 only the
G ROTRIAN diagram of Be as one typical example.
Be+ ( 2p 2P3/2,1/2) displaced terms displaced terms

2pns 2pnp 2pnd 2pns 2pnp 2pnd
binding energy / cm -1
1Po 1 3P o 3P 3D
WI = 13.282 1 P1 0 0
13
5 5
Be+ ( 2S 4 4
1/2)
3
11 3 3
2sns 2snp 2snd 2p 2 2sns 2snp 2snd
10 1S 1P o 1D 1S 3S 3 Po 3D
WI = 0 1 2 0 0
n 0
9.32270 5
5 4 5 5
5 4 4
4 4 4 - 10000
8 3 2p 2 3
3 1D 2p2
2 3 3P
7 3 0 -20000
3
6
- 30000
2
5
- 40000
3 Be I - 50000
singlet 2
2 - 60000
Be I
1 triplet
- 70000
2 1s 22s 2 1S0 Be ground state - 75192.63
0
Fig. 7.11 Term scheme of the beryllium atom derived from K RAMIDA et al. (2013) (G ROTRIAN
diagram for the strongest transitions, Aab > 5 106 s1 ). Singlet terms are marked by red, triplet
terms by black lines
The ground state configuration of Be I is 1s 2 2s 2 , its first ionization potential

is 9.32 eV, the ground state of the Be+ ion (Be II) being 1s 2 2s. To reach the first
excited state of Be+ (2s 2 2p 2 P3/2,1/2 ) additional 3.96 eV, for full removal of the
second valence electron 18.21 eV are needed. The doubly excited states of the al-
kaline earth atoms (also called displaced terms) partially start already below the
first ionization potential as documented in Fig. 7.11 for the Be** 2pnp 1,3 P series.
Doubly excited states above the ionization limit show the typical autoionization
with partially overlapping line structures. Singlet-triplet intercombination lines are
forbidden as in the He case, although weak transitions are observed already in Mg
and the prohibition is more and more released as Z increases: spin-orbit coupling
induces the characteristic configuration interaction between singlet and triplet terms
for larger Z so that the initial assumptions for the electron spin conservation rule
(7.60) in E1 transitions is no longer valid. Hence the rule can be violated.
The fine structure splitting in the first excited 3 PoJ state is for the Be atom still
very weak, 0.645 cm1 (J = 0 J = 1) and 2.345 cm1 (J = 1 J = 2). For
Mg, however, it is already 20.059 cm1 and 40.714 cm1 , respectively in both
cases thus normally ordered (contrary to He) and in the latter case in good agree-
ment with L ANDs interval rule (6.67). For the heavier alkaline earth metals spin-
orbit interaction becomes rapidly more significant, with larger FS splitting and an
enhancement of intercombination lines.
7.7.2 Mercury
As a last example for quasi-two-electron atoms we discuss mercury, Hg. Its ground
state configuration is [Xe]4f 14 5d 10 6s 2 , i.e. the K to N shells are completely filled,
and all subshells of the O shell are filled except for the 5f electrons. The building-
up of the P shell starts with the two 6s 2 electrons.
Figure 7.12 shows part of the Hg emission spectrum, constructed from the
NIST data assuming here a somewhat unrealistically high electron temperature
(Te = 15 000 K) which allows us to see also several transitions between higher lying
states. In a standard low temperature gas discharges typically the 6s 2 1 S0 6s6p 3 Po2
is by far the dominant line due to a combination of a reasonable transition probabil-
ity (Aab = 8 106 s1 ) and low enough excitation energy Wexc = 4.89 eV. Hence,
the population density of the excited 63 Po2 state exp(Wexc /kB Te ) is high enough.
Note that this line is an intercombination line which in strict RUSSEL -S AUNDERS
coupling (LS coupling) is completely forbidden. Obviously, with its nuclear charge
Z = 80 and a nucleon number from A = 196204 Hg belongs to the heavy elements
of the periodic system and LS coupling starts to break down as spin-orbit interac-
tion increases Z 4 /n3 (see Sect. 6.3.2). Nevertheless, as shown in Fig. 7.13, one
continues to classify the states as singlet or triplet system.
The term scheme shows a remarkably large FS splitting of the 6 3 PJ states. With
1767 cm1 (J = 0 J = 1) and 4631 cm1 (J = 1 J = 2) the terms are nor-
mally ordered and in fair agreement with L ANDs interval rule. Its order of magni-
tude is getting close to the splitting between singlet and triplet system, which in the
7.7 Quasi-two-Electron Systems 373
line intensity / arb. un
6s 6p 3P o1 - 6s7d 3D2
6s 2 1S- 6s6p 3Po
6s6p 3Po - 6s 7d 3D
6s6p 3P o2 - 6s6d 3D3
6s6p 3P o0 - 6s 7s 3S1
6s 6p 3P o1 - 6s7s 3S1
6s6p 3P o2 - 6s7s 3S1

6s 6p 1P o1 - 6s6d 3D2
200 300 400 500 600 700 800
wavelength / nm
Fig. 7.12 Sticks spectrum of Hg in the UV and VIS after K RAMIDA et al. (2013) (S AHA-LTE
spectrum, electron temperature 1.23 eV)
6snp 6snp 6snd 6snf 6snp 6snp 6snd 6snd 5d 9(2D5/2) cm-1
WI = 1S 1P o 1D 1F o 3S 3Po 3D 3Fo 2
10.4375 0 1 2 3 0 2,1,0 3,2,1 4,3,2 6s 6p 0
10 10 9 8 6 10 9 8 6
9 8 7 5 9 8 7 5 - 10000
9 8 7 6 8 6
7
8 7 complex - 20000
7 terms
7 - 30000
6
6
3 Po - 40000
2
5
6 3 Po
1
3 Po - 50000
4 0
singlet triplet
3 - 60000
nm
3.7
2 25 - 70000
1
6 [Xe] 4f 14 5d 10 6s 2 1S0 Hg I ground state - 84184 - 80000
0
Fig. 7.13 Term scheme of the mercury atom derived from K RAMIDA et al. (2013) (G ROTRIAN
diagram for the strongest transitions in Hg, Aab > 1 106 s1 ). Singlet terms are marked by red,
triplet terms by black lines. Intercombination lines are dashed black
case of 6 1 Po1 and 6 3 Po2 is about 10 026 cm1 . Still, the line with the highest transi-
tion probability is 6s6p 3 Po2 6s6d 3 D3 (Aab = 1.3 108 s1 ) a clean LS allowed
intra-triplet transition.
Section summary
The two valence electrons in alkaline earth metals and Hg essentially behave
like those in He, with the nuclear charge effectively shielded by the filled inner
shells.
Even for Hg the characterization of valence states by RUSSEL -S AUNDERS

(LS) coupling is still approximately correct. However, the FS splitting is large
and a number of rather strong intercombination lines is observed.

rations in atomic and molecular structure calculations, using linear superpositon
index.html.
VUV: Vacuum ultraviolet, spectral range of electromagnetic radiation. Part of the
21348 (2007).
References
D OMKE , M. et al.: 1991. Extensive double-excitation states in atomic helium. Phys. Rev. Lett.,
66, 13061309.
D RAKE , G. W. F. and W. C. M ARTIN: 1998. Ionization energies and quantum electrodynamic
effects in the lower 1sns and 1snp levels of neutral helium (4 He I). Can. J. Phys., 76, 679698.
FANO , U.: 1961. Effects of configuration interaction on intensities and phase shifts. Phys. Rev.,
124, 18661878.
H UTEM , A. and S. B OONCHUI: 2012. Evaluation of Coulomb and exchange integrals for higher
excited states of helium atom by using spherical harmonics series. J. Math. Chem., 50, 2086
2102.
References 375
PAULI J R ., W.: 1925. ber den Zusammenhang des Abschlusses der Elektronengruppen im Atom
mit der Komplexstruktur der Spektren. Z. Phys., 31, 765783.
S CHULZ , K., G. K AINDL, M. D OMKE, J. D. B OZEK, P. A. H EIMANN, A. S. S CHLACHTER
and J. M. ROST: 1996. Observation of new Rydberg series and resonances in doubly excited
helium at ultrahigh resolution. Phys. Rev. Lett., 77, 30863089.
Atoms in External Fields
8
The interaction of atoms with external magnetic fields

(Z EEMAN effect) has already been introduced in Chaps. 1
and 2, while radiation induced transitions where treated in
Chaps. 4 and 5. Here we generalize and deepen what is already
known, and develop the tools for a quantitative description of
atoms and molecules in external magnetic and electric fields.
Thus, this chapter provides the essential basis for understanding
this type of interaction also in a more complex environment and
gives first examples of how to approach macroscopic properties
of matter, such as magnetism and dielectric polarization.
Overview
Sections 8.18.4 present the essentials on these topics. The reader should be-
come well acquainted with these key tools in modern atomic physics even
though parts of it may seem somewhat strenuous at first sight. These concepts
are essential also for molecular physics, as well as for a fundamental under-
standing of many properties of condensed matter and plasmas. This will be
illustrated in Sects. 8.1.6, 8.1.7, 8.1.8, 8.2.10, 8.3 and 8.4.2 for a number of
selected examples. In these presentations the reader will also find various ref-
erences to modern developments in atomic physics which will brighten the
study of these classical themes with currently hot topics as e.g. fast and
slow light in Sect. 8.4.4 or, in Sect. 8.5, with a first approach towards the
rapidly developing modern research field of matter in intense and ultra-intense
laser fields.
8.1 Atoms in a Static Magnetic Field
8.1.1 The General Case
Pioneering work on atoms in magnetic fields was already performed at the end of
the 19th century (the second N OBEL prize was awarded to L ORENTZ and Z EEMAN
1902). Here we recall our previous discussion on magnetic moments in magnetic

378 8 Atoms in External Fields
fields and on spin-orbit interaction. According to (6.35), spin-orbit interaction with-

out external magnetic fields may be written in a.u.
2
LS = (r)
V L L
S = a S (8.1)
2
and with (6.36) and (6.49) its expectation value (i.e. the FS splitting) is
LS = a
VLS = V J (J + 1) L(L + 1) S(S + 1) (8.2)
2
with the fine structure constant 1/137 and the spin-orbit coupling parameter
a = 2 (r) /2. We recall that (r) = (dV /dr)/r. In the following, we shall use for
clarity SI units and rewrite (8.1):
V L
LS = (a0 )2 (r) L
S/22 = a S/2 . (8.3)
The quantum numbers for orbital angular momentum, spin and total angular mo-
mentum are L, S and J , the corresponding projection quantum numbers ML , MS
and MJ (also called magnetic quantum numbers).
B depends on the mag-
In an external magnetic field B the interaction energy V

netic moments ML and MS of orbital angular momentum L and electron spin
S.

They define the total magnetic moment MJ = ML + MS . Assuming B z we
have to add (6.30) and (6.31) with the g factors gL = 1 (orbit) and ge = 2 (spin),
and obtain (B = e/2me is the B OHR magneton)
B = M
V J B = (M L + M S ) B = B (gL Lz + ge
Sz )B (8.4)

z + 2
L Sz Jz +
Sz
= B B = B B = M J z B. (8.5)

z + 2
In the last steps we have used Jz = L Sz and defined the z-component MJ z =
B (Jz + J .
Sz )/ of M
The quantum numbers for orbital angular momentum, electron spin and total
angular momentum are again L, S and J , their projections onto the z-axis ML , MS
and MJ , respectively (magnetic quantum numbers). We consider the problem for
the completely unperturbed system as solved:
0 |nLML SMS = WnLS |nLML SMS

H
H0 comprises all electrostatic interactions (including exchange) but no spin-orbit

interaction: spin and orbit are still uncoupled in this 0th order solution, with n,
L, S, ML and MS being good quantum numbers. We restrict our considerations to
such cases where the perturbation by the magnetic field and/or spin-orbit interaction
is small compared to the energy difference between neighbouring states n L S or
n L S , so that only the substates of one nLS multiplet contribute to energetic shifts
and level splittings.
8.1 Atoms in a Static Magnetic Field 379
Table 8.1 Limiting cases for the Z EEMAN effect

B field Perturbation Optimal basis Effect
low B a/B VB VLS |LSJ MJ anomalous Z EEMAN
high B a/B VLS VB |LML SMS PASCHEN -BACK
Using the binomial formula for L

S in (8.3) and with (8.5), the full Hamiltonian
of an atom in a magnetic field becomes:
= H
H 0 LS
+V +VB
J
L Jz +
2 2 2
0 S Sz (8.6)
=H +a + B B
2 2
unperturbed spin-orbit interaction magnetic interaction.
It is important to realize at this point that neither LML SMS nor LSJ MJ is a set of
good quantum numbers for this combined Hamiltonian: as discussed in Sect. 6.2.5
the states |LML SMS do not diagonalize spin-orbit interaction. On the other hand,
the |LSJ MJ states do not diagonalize the magnetic interaction in (8.6): Sz and
2
J
do not commute since the latter contains also the components Sx and Sy . In contrast,
per definition Jz commutes with J as well as with L and S (since Jz = L z +
2 2 2
Sz )
, Jz ] = 0.
so that [H
Thus, in the general case where magnetic and spin-orbit interactions are equally
important, MJ = ML + MS is a good quantum number of the system, while J is not
a good quantum number.
If we want to treat the problem in perturbation theory we have to choose in 0th
order the most appropriate coupling scheme, |LML SMS or |LSJ MJ depending
on which approximately describes the states best. This depends on the magnitude
of the magnetic field as summarized in Table 8.1. The two limiting cases for very
small and very large external magnetic field are called anomalous Z EEMAN effect
and PASCHEN -BACK effect, respectively.
Before entering into the details it is commendable to obtain an estimate for the
order of magnitude of the magnetic interaction. According to (8.5) it will be in the
range of B B. Without much effort one may obtain in the laboratory magnetic fields
up to 1 T to 2 T, using soft iron electromagnets or permanent magnets. With state-
of-the-art super-conducting magnets one reaches 10 T to 30 T that are e.g. required
for high resolution NMR and EPR spectroscopy. With B = 5.788 105 eV T1 ,
B < 30 T and L z + 2
Sz / 1 one thus expects splittings up to a maximum of
VB < 2 103 eV =
14 cm1 , (8.7)
corresponding to 103 104 of typical electronic excitation energies for valence

electrons. According to Table 6.2 this upper limit is much larger than the fine struc-
ture splitting of the 2p 2 P state in H or Li (0.37 and 0.36 cm1 , respectively), and for
He and Be we have seen an FS splitting of about 1 cm1 . However, for the heavier
alkalis, from Na upward, the splitting that can be achieved with laboratory magnetic
fields typically remains below FS splitting in the first excited states. To fully treat
lower and higher Z elements and states with low as well as high principle quantum
numbers n we have to consider both cases characterized in Table 8.1.
8.1.2 ZEEMAN Effect in Low Fields
In practice the case of low magnetic fields, B a/B , is particularly impor-

tant. In 0th order we start now with a Hamiltonian H 0 + VLS whose eigenstates
|J MJ |LSJ MJ are spin-orbit coupled as described in Sect. 6.2.5. The 2S + 1
levels of a multiplet 2S+1 LJ are characterized by their total angular momenta J ,
each consisting of 2J + 1 degenerate magnetic substates |LSJ MJ . In an external
magnetic field we have to add VB according to (8.5), which we treat in 1st order as
a small perturbation. The additional magnetic energy thus becomes
B = B J MJ |Jz +
VB = V Sz |J MJ B (8.8)

B
= J MJ |Jz |J MJ + J MJ |
Sz |J MJ B. (8.9)

We now exploit that J MJ |Jz |J MJ = MJ and obtain

J MJ |
Sz |J MJ
VB = 1 + B MJ B.
MJ
In the next step we use some angular momentum algebra derived in Appendix C.1.2.
By applying the projection theorem in the form (C.18) onto the matrix element of

Sz the magnetic energy becomes

J MJ |
S
J |J MJ
VB = 1 + B MJ B. (8.10)
J (J + 1)2
Using the binomial formula this can be further evaluated as 2 SJ =J +
2 2
S
J
( J +
S)2 = S
2 2 2
L . Inserting this into the above matrix element the eigen-
values J (J + 1) etc. emerge and one finds for the numerical factor in the square
2
bracket

J MJ |
S
J |J MJ 3J (J + 1) + S(S + 1) L(L + 1)
gJ = 1 + = . (8.11)
J (J + 1)2 2J (J + 1)
This is nothing but the L AND gJ factor phenomenologically introduced in (1.161).

The final result for the magnetic energy of the total angular moment may now be
written
VB = gJ B MJ B = MJ z B, (8.12)
with MJ z being the expectation value for the z-component of the total magnetic
moment of the electron, defined in (8.5). Thus, in a (not to high) magnetic field each
J level in a multiplet splits into 2J + 1 equally separated components.
We point out here, that the above derivation also holds for multi-electron systems
in RUSSEL -S AUNDERS (LS) coupling in low magnetic fields. The key assumption
for deriving the gJ factor from the Hamiltonian (8.6) is that the magnetic moments
of the electron orbits and spins are strictly proportional to the respective angular
momenta, with gL = 1 and ge = 2. This implies that the total
orbital angular mo-
mentum and the total spin are exactly L = i Li and S = i S i (where i refers to
the electrons involved); only after that coupling, the total overall angular momentum
J = L + S is composed. Note that assumptions also define the RUSSEL -S AUNDERS
coupling scheme.
The Vector Model

In Fig. 8.1 this somewhat abstract derivation is illustrated by the vector model as
introduced in Sect. 6.2.5. We point out, however, that the logic of this essentially
classic model is not completely compelling, i.e. one has to know somehow the cor-
rect quantum mechanical result in order to present it graphically. But it helps to
make the relation between angular momentum addition and the resulting magnetic
moments plausible. Again, the z-axis is taken parallel to B, and one starts with the
definition of the total angular momentum:
J = L + S.
Since LS coupling is assumed to be much stronger than the magnetic interaction,

L and S precess fast around J . Consequently, the magnetic moment of J is an
averaged quantity MJ = ML + MS . The averaging corresponds to a projec-
tion of ML and MS onto the J direction. According to (8.4) the effective inter-
action with the magnetic field B is derived from this averaged magnetic moment
MJ
VB = ML + MS B = MJ B = MJ z B. (8.13)
In contrast, J precesses at a much slower frequency around B and the mag-
netic moments are projected onto the unit vector J /|J |. The magnitude of
Fig. 8.1 Vector model for z,B slow

the anomalous Z EEMAN
effect Jz J
S
fast
L
ML
MJz
MJ
MS
the averaged moments are simply given by the respective scalar products
M J /|J |. In summary, with ML = B L/ and MS = 2B S/ one writes
(8.13)

ML J J MS J J
VB = MJ B = + B
|J | |J | |J | |J |

B L J + 2S J (J + S) J Jz
= J B = B B = gJ B MJ B.
J 2 J 2
Obviously, the ratio of scalar products to J 2 in the squared bracket is the gJ factor.
Comparing this semiclassical relation to (8.11) and (8.12) one finds that they are
equivalent: one just has to apply again the binomial formula for the scalar product
S J , and interpret J 2 , S 2 and L2 by the respective quantum mechanical expectation
values J (J + 1)2 , etc.
Some Examples
Singlet system: We remember the so called normal Z EEMAN effect which we
have introduced already in Sect. 2.7. The standard experiment is illustrated in
Fig. 4.18. We are now able to identify the special situation in which this clas-
sically expected splitting into three components is actually observed: the total
spin has to be S = 0 so that no fine structure splitting is observed in this case. A
typical example is He, where the spins of both electrons s1 = 1/2 and s2 = 1/2
may be composed to a triplet (S = 1) or a singlet (S = 0) system as discussed
in detail in Sect. 7.3. In the latter case J = L which according to (8.11) is then
characterized by a g factor gJ = 1. Hence, independent of J (or L) all J MJ sub-
states in the singlet system have the same splitting B B between neighbouring
magnetic levels MJ and MJ + 1. In Fig. 8.2 this is illustrated for the example of
MJ = +2 "normal"
(a) +1
(b)
gJ = 1
ZEEMAN spectrum
0
equal splitting
1D : J=2 -1 observed in xy direction

2
-2
1P : J=1 +1
gJ = 1
1
0
observed in z direction
-1 - +
MJ = +1 0 -1
- +
linearly polarized
(no emission in z direction right and left
in z direction) circularly polarized
Fig. 8.2 So called normal Z EEMAN effect exemplified by a 1 P 1 D transition (e.g. in the He
singlet system). (a) Term scheme. (b) Spectrum (schematic); since gJ is the same for the upper and
lower levels only the classical triplet of lines is observed (red: components, pink: components)
Fig. 8.3 Triplet term scheme MJ = +2

with allowed E1 transitions 3P
2 +1
and the expected spectrum J=2 0
(schematic) below gJ = 3/2 -1
-2
3S +1
1
J=1 0
gJ = 2 -1

a 1 P1 1 D2 transition. Since gJ = 1 for all J one observes only three compo-

nents for all transitions within a singlet system.
Triplet system: This is the counterpart to the singlet case just discussed. Fig-
ure 8.3 shows as an example a 3 S1 3 P2 triplet transition. According to (8.11)
the L AND g factors are different in the lower and upper state, g(3 S1 ) = 2 and
g(3 P2 ) = 3/2, respectively, so that in the magnetic field all together nine lines
of different frequency are observed. This is indicated below the term scheme in
Fig. 8.3.
Doublet system: For S = 1/2 there are 2S + 1 = 2 fine structure levels with total
angular momenta J = |L 1/2|. The g factor (8.11) is
J + 1/2
gJ = .
L + 1/2
The most important 2 S1/2 2 P3/2,1/2 transitions (e.g. the Na D line doublet)
have g factors g(2 S1/2 ) = 2, g(2 P1/2 ) = 2/3 or g(2 P3/2 ) = 4/3. This leads to
a splitting scheme in the magnetic field as shown in Fig. 8.4(a). Sketched in
2P MJ = (a) doublet spectra (b)

3/2 gJ = 4/3
+3/2
+1/2 MJ = 2S 2P
2P
1/2 gJ = 2/3 1/2 - 3/2
-1/2 +1/2
J = 3/2 - 3/2 -1/2 2S 2P
J = 1/2 1/2 - 1/2
2P 2D
3/2 - 5/2
2P 2D
2S
1/2 3/2 - 3/2
+1/2
gJ = 2
J = 1/2 -1/2 2P
1/2 -
2D
3/2
+ - + -
6 lines 4 lines
Fig. 8.4 Alkali doublets (a) splitting of the 2 S1/2 and 2 P1/2,3/2 states in a magnetic field with
allowed E1 transitions (e.g. for the Na D1 and D2 lines). (b) corresponding doublet spectra
(schematic) and more examples
Fig. 8.5 Stick spectrum (calculated) of transitions between magnetically split sublevels in a septet
7 S 7 P . This could e.g. be in Cr I between the configurations 3d 5 ( 6 S)4s 3d 5 ( 6 S)4p) at a
3 4
wavelength 425.4 nm. Pink the , red the components detected in the xy plane
Fig. 8.4(b) are the corresponding spectra. They may turn out to be rather com-
plex. For reference we summarize the whole fine structure splitting with (8.8) and
(8.2) explicitly for the example 2 P3/2,1/2 . In a low magnetic field with g3/2 = 4/3
and g1/2 = 2/3 we obtain by combining (8.8) and (6.62):
*
4
B BMJ + a2 for J = 32 , MJ = 32 , 12 , 12 , 32
VB + VLS = 32 (8.14)
3 B BMJ a for J = 12 , MJ = 12 .
Septet: For heavier atoms one also observes lines with high multiplicity. In
Fig. 8.5 we just show as one arbitrary example the characteristic splitting of a
line in the Cr I spectrum. (This calculation just demonstrates how complex such
spectra can be, without going into detail for chromium which has a rather com-
plicated electronic structure).
We finally note, that the magnetic level splitting as complicated as it may appear
from the above examples often serves as an important tool for revealing term
configurations in the spectra of complex atoms and molecules.
Line Strengths
For a quantitative evaluation of spectra one needs to calculate the transition probabil-
ities between individual magnetic sublevels J MJ J MJ splitted in the magnetic
field. These are easily obtained from (4.118) and (4.124) for spontaneous emission
and absorption, respectively. One simply has to evaluate the corresponding 3j sym-
bols for MJ = q.
The relative strengths of different lines J J within one multiplet nLS
n L S require evaluation of (6.70), i.e. of the corresponding 6j symbols. With the

tools communicated in Appendixes B.2 and B.3 this can be done without problems.
Several Java applets for computing 3j , 6j and 9j symbols are available in the In-
ternet (e.g. S VEN G ATO R EDSUN 2004).
8.1.3 PASCHEN-BACK Effect
The case of very high magnetic fields B a/B can be of relevance for light atoms
such as H, He or Li, as well as for all highly excited states since FS interaction de-
creases with the principle quantum number as 1/n3 (see (6.66)). Also in the extreme
magnetic fields of some stellar atmospheres this case has to be considered.
Fig. 8.6 Vector model for z,B

Z EEMAN effect in high
magnetic fields Sz
S
Lz
L
MS
ML
In 0th order we completely neglect now the spin-orbit coupling and start with the
Hamiltonian:
H 0 + V
P B = H B
=H0 + B (L
z + 2
Sz )B. (8.15)

Then L and S are decoupled, and this situation is called the PASCHEN -BACK effect.
In this approximation L z and
Sz as well as Jz are well determined and commute

with H . Thus LSML MS are good quantum numbers as well as MJ = ML + MS .
In the vector model one may visualize this as illustrated in Fig. 8.6: both angu-
lar momenta precess around the external magnetic field B. We may compare this
directly with Fig. 8.1, where the corresponding situation for low magnetic field is
displayed (in both figures L = 1, S = 1, MJ = 3/2 is assumed and presented to
scale). We thus choose the uncoupled states, short |ML MS , as basis and obtain
with (8.15)
B |ML MS = B B(L
V z + 2
Sz )|ML MS

= B B(ML + 2MS )|ML MS ,
and the splitting in the magnetic field is simply given by
VB = B B(ML + 2MS ). (8.16)
This is an exact solution of the problem if spin-orbit interaction can be com-

pletely neglected in comparison to the magnetic splitting. Schematically this is
summarized in Fig. 8.7 for 2 P3/2, 1/2 and 2 S1/2 states. In addition to the quan-
tum numbers ML , MS and MJ = ML + MJ also the z-component = Mz /B =
ML + 2MS of the magnetic moment is displayed (in units of the B OHR magneton
B ) which determines the splitting. Note that the level characterized with = 0
is doubly degenerate since it may be composed in two different ways from ML
and MS . Thus, the number of states remain the same as in the low field case,
i.e. (2 (3/2) + 1) + (2 (1/2) + 1) = 6 for the 2 P3/2, 1/2 states. Conversely,
MJ = 1/2 refers to two levels with zero splitting as well as to splittings : albeit
Jz is a good quantum number, it does not uniquely characterize the states.
MJ ML MS
+3/2 +1 +1/2 2
2P
+1/2 0 +1/2 1
3/2
-1/2 -1 +1/2 0
+1/2 +1 -1/2
2P
1/2 -1/2 0 -1/2 -1
- 3/2 -1 -1/2 -2
2S 0 +1/2 1
1/2 +1/2
without -1/2 0 -1/2 -1

B field + -
Fig. 8.7 PASCHEN -BACK effect: splitting and E1 transitions in a high magnetic field, illustrated
for the example 2 P3/2, 1/2 2 S1/2 (cf. weak B field Fig. 8.4). One sees the classical line triplet
as in the normal Z EEMAN effect. Shown are the magnetic quantum numbers MJ = ML + MS
as well as the projection of the magnetic moment onto the B or z-axis, = ML + 2MS , that
determines the splitting. Different energies may belong to the same MJ , but one specific energy
may also be realized for different MJ
In order to identify allowed optical transitions we have to account for the fact
L and
that S are now completely decoupled. And since the E1 dipole operator acts
only on the orbital component L of the states and not onto the spin
S, the selection
rules for E1 transitions in a high magnetic field are
ML = 0, 1, MS = 0 and L = 1 (8.17)
as indicated in Fig. 8.7. Since in the upper and lower level the splitting equally
depends only on = ML + 2MS , six different transitions are possible but only
three different frequencies are observed: in high magnetic fields one thus observes
the classical normal Z EEMAN triplet.
8.1.4 Do Angular Momenta Actually Precess?
Before generalizing the above considerations we want to address the conceptually

important question whether the magnetic moments (and hence the angular mo-
menta) really precess in a magnetic field, be it in an external field or in the in-
ternal magnetic field that causes spin-orbit coupling. In other words, we want to
understand whether the suggestive images used in the vector model in Fig. 8.1 and
Fig. 8.6 represent reality or, alternatively, only reflect the probability to find the
angular momenta pointing into a specific direction.
To answer the question let us consider just the spin of an electron in an isolated
1/2 atom (L = ML = 0), and let us switch on an external field B z at time t = 0.
2S
Let the spin orientation in space at this moment be given by its polar and azimuthal
angles, 0 0 . This may e.g. be achieved by a passing the atom through a S TERN -
G ERLACH magnet whose axes are aligned correspondingly in respect of B. This
particular spin state prepared at t = 0, is described according to (E.16) by

0 0 1 1 0 0 1 1
| = cos exp i + sin exp i , (8.18)
2 2 2 2 2 2 2 2
in the basis of the spin states |SMS = | 12 12 . In the magnetic field both spin com-
ponents will change with time according to (2.16) as exp(VB t/) corresponding
to their energies VB in the magnetic field. With (8.16) VB = 2B BMS and we ob-
tain the spin state as a function of time:

(t) = cos 0 exp i 0 exp i B B t 1 1
2 2 2 2

0 0 B B 1 1
+ sin exp i exp i t .
2 2 2 2
This may also be written as

(t) = cos 0 exp i (t) 1 1 + sin 0 exp i (t) 1 1 ,
2 2 2 2 2 2 2 2
2B B
where (t) = 0 + t = 0 + j t (8.19)

with the (generalized) L ARMOR frequency j for the electron spin as defined in
(1.162). This clearly shows that the azimuthal angle of the spin grows with j t
while the polar angle remains unchanged: the spin thus precesses indeed with the
L ARMOR frequency around the axis of the magnetic field in fact just in the direc-
tion indicated in Fig. 8.1 and Fig. 8.6, the angular precession frequency j being
independent of the polar angle . This result is in accordance with the classical
model (Sect. 1.9.2, see in particular Fig. 1.36) of a spinning top with a magnetic
moment onto which the external magnetic field exerts a torque apart from the
factor g
= 2, which is needed to describe the intrinsic magnetic moment of the elec-
tron. We also recall at this point that it is this very precession with the L ARMOR
frequency that is used for the ultra precise determination of the electron magnetic
moment anomaly. As explained in Sect. 6.6 one essentially exploits the difference
between L ARMOR frequency and cyclotron frequency to determine the very small
deviation of the electron L AND factor from g = 2.
To develop some feeling for the relevant time scales of this precession we con-
sider a reasonably high field B = 0.5 T. In that case the L ARMOR precession time is
h/2B B = 71.4 ps. The inner atomic field which the electron spin sees in the 2p
state of atomic hydrogen due to the angular momentum is of the same order of mag-
nitude (by dividing the spin-orbit splitting according to (6.66) by B one estimates
0.26 T). However, as the splitting is proportional to Z 4 /n3 , for high Z atoms the
magnetic field acting on the electron spin at the position of the nucleus may easily
become 30 T or even more comparable to the highest fields which can be generated
by supra conducting magnets. The precession times are reduced correspondingly.
In the theory of magnetism one occasionally uses the term exchange fields, in
particular in the context of ferromagnetism. To derive such (hypothetical) exchange
fields one simply divides the exchange interaction which is responsible for ferro-
magnetism by B , as we have done in the present context. The numerics of this
procedure may lead to fields that are orders of magnitude larger than those dis-
cussed above. Of course they are just a mathematical construct! As we have learned
in (7.49) in the context of the helium atom, exchange interaction is of purely elec-
trostatic nature, enforced by anti-symmetrization of the electronic wave function,
and definitively not generated by magnetic fields. The example of He illustrates very
clearly that exchange interaction may be on the order of magnitude of an eV, in
comparison to which spin-orbit splitting (Fig. 7.6) is minute.
8.1.5 In Between Low and High Magnetic Field
In the preceding sections we have approached the splitting of energy levels in an

external magnetic field from two different limits: for spin-orbit coupling VLS being
large compared to the interaction of the atoms magnetic moment with the external
magnetic field, VB , we started with LS coupled states |n(LS)J MJ as zero order
approximation. In contrast, if VB is large compared to VLS it was appropriate to start
from the uncoupled |nLML SMS basis and add VLS later on as a small perturbation.
In between these two limits, i.e. if both interactions are of similar magnitude we
cannot avoid a full diagonalization of Hamiltonian (8.6). We write
=H
H 0 + H
BLS with
a B
H B + V
BLS = V LS = 2 L
S+ (Lz + 2
Sz )B. (8.20)

Considering the unperturbed system as solved in the uncoupled basis |nLML SMS
0 |nLML SMS = WnLS |nLML SMS ,
H
and assuming that only one principle quantum number n contributes, a linear com-
bination of uncoupled states

|LSMJ = cML MS |LML SMS (8.21)
ML MS
is the obvious ansatz for the perturbed system. The eigenvalues of H BLS are the
sought-after splittings V,MJ (B) of the energy levels in the magnetic field:
BLS |LSMJ = V,MJ (B)|LSMJ .
H (8.22)
Note that |LSMJ are also eigenstates of Jz = L

z +
Sz :
Jz |LSMJ = MJ |LSMJ with MJ = ML + MS . (8.23)

Fig. 8.8 Scheme of the =ML+2MS ML MS

transition from low to high B
field. The level splittings are 2 +1 +1/2
not drawn to scale. The
dashed horizontal line
indicates the energetic
position of a hypothetical nL gJ = 4/3 MJ 1 0 +1/2
level without spin-orbit 2P 3/2
3/2 1/2
interaction and without nP - 1/2 a/2
a/2
external magnetic field - 3/2 0 - 1 +1/2
a
1/2 0 +1 - 1/2
2P - 1/2
1/2
gJ = 2/3
unper- -1 0 - 1/2
turbed - 2 - 1 - 1/2
anomalous
no field ZEEMAN effect PASCHEN-BACK effect
The magnetic quantum number MJ by itself does not yet fully specify an eigenstate
of the total system. Thus, we use the additional parameter which is chosen such
that in the limit of high fields = ML + 2MS = MJ + MS . Hence it corresponds
to the projection = M z /B of the magnetic moment onto the z-axis. This has
already been exemplified for a 2 P3/2 state in Fig. 8.7 (see labels on the right).
Note that there may be several solutions for each value MJ , as well as several MJ
values for one specific . In detail Fig. 8.8 illustrates schematically the connection
between quantum numbers and splittings in the transition from low to high B field:
the states that have already been identified in Fig. 8.7 for the high field case are
connected with the low field situation such that MJ = ML + MS . This is, however,
not yet an unambiguous rule for the change from low to high field. In addition, we
have applied the so called non-crossing rule according to which states with equal
MJ must not cross!
The latter rule emerges from the full treatment of the problem: we have to di-
agonalize the combined Hamiltonian (8.20) for magnetic and spin-orbit interaction
by the expansion coefficients cML MS for the states |LSMJ according to (8.21).
Using (8.16) we write H BLS in matrix form:

LML SMS HBLS |LML SMS (8.24)
a
= B (ML + 2MS )B ML ML MS MS + 2 LML SMS L
S|LML SMS

The matrix elements of L
S are derived in Appendix C.3.2. Specifically we use
(C.59)(C.62) and the important result that the off-diagonal matrix elements are
non-zero only for MJ = ML + MS = ML + MS = MJ :

LML SMS HBLS |LML SMS = (M +M )(M +M ) .
L S L S
(8.25)
BLS matrix is thus very simple. Only states for

The structure emerging for the H
which MJ = MJ interact.
BLS for a 2 P state in the uncoupled |LML SMS basis

Table 8.2 Matrix elements of H
ML MS |1 12 |0 12 |1 12 |1 12 |0 12 |1 12
2 1 0 0 1 2
MJ 3/2 1/2 1/2 1/2 1/2 3/2
1 12 | 2B B + a/2 0 0 0 0 0

0 12 | 0 B B a/ 2 0 0 0

1 12 | 0 a/ 2 a/2 0 0 0

1 12 | 0 0 0 a/2 a/ 2 0

0 12 | 0 0 0 a/ 2 B B 0
1 12 | 0 0 0 0 0 2B B + a/2
This H AMILTON matrix may easily be diagonalized. The ansatz (8.21) leads to
the following system of linear equations

cL,S
' ( cL+1,S
LML SMS HBLS W 1|LML SMS

= 0,
(8.26)
...
cL,S
BLS W
with {. . . } representing the matrix corresponding to H 1. The eigenenergies
W in the general case are then obtained from solving the secular equation
' (
det ML , MS HBLS W1|ML , MS = 0. (8.27)
We want to discuss this in some more detail for the important case of a 2 P3/2,1/2
doublet, for which Table 8.2 represents (8.24). It has been obtained using Table C.1.
Obviously, the secular equation (8.27) factorizes in this case into four rather conve-
nient expressions:
2B B + a/2 W = 0

B B W a/ 2
=0
a/ 2 a/2 W

a/2 W a/ 2
=0
a/ 2 B B W
2B B + a/2 W = 0.
These have six solutions Wi = V,MJ (B) that are the sought-after energy splittings
in the B field. With x = B B/a they may be written:
1
V2,3/2 /a = 2x + (8.28)
2

2
1 1 3 4 2
V1,1/2 /a = x + 1+ x+ x (8.29)
2 4 4 9 3
V, M / a
J
MJ = 3/2 =ML+2MS
2
1
5 1/2
2P
3/2 5 10
0 -1/2 0
2P
1/2 1/2 0 B B / a
-5 -1/2
-1
- 3/2 -2
Fig. 8.9 Transition from a low to a high magnetic field B for the example of a 2 P3/2,1/2 system.
Plotted as full red lines are the splittings VMJ of the levels as a function of the field strength B.
The parameters MJ = ML + MS and = ML + 2MS characterize the 6 states (see text). Dashed
black (red) lines show the extrapolation in the case of a low (high) field. Two avoided crossings are
marked by pink circles. The energy splitting VMJ is given in units of the fine structure parame-
ters a, the field strength B in units of a/B

2
1 1 3 4 2
V0,1/2 /a = x 1+ x+ x (8.30)
2 4 4 9 3

2
1 1 3 4 2
V0,1/2 /a = x + 1 x + x (8.31)
2 4 4 9 3

2
1 1 3 4 2
V1,1/2 /a = x 1 x + x (8.32)
2 4 4 9 3
1
V2,3/2 /a = 2x + . (8.33)
2
The result of this diagonalization of H BLS is plotted to scale in Fig. 8.9. In a high
field the splitting is proportional to B = (ML + 2MS )B, in a low field to gJ MJ B.
An important consequence of the fact that states with equal MJ couple (as ex-
pressed by nonvanishing matrix elements in the Hamiltonian Table 8.2) is clearly
seen in Fig. 8.9:
There are no curve crossings for states with equal MJ = ML + MS .
This implies a repulsion of the corresponding terms (here MJ = 1/2). One sees
that the linear splitting V,MJ corresponding to g3/2 = 4/3 and g1/2 = 2/3 in the
case of low fields is only valid over a rather small range of the magnetic field
Table 8.3 Limiting values for the energy splitting V,MJ (B) of a 2 P3/2,1/2 doublet in very weak
(left) and very high (right) magnetic field B, respectively. The levels are determined by the quantum
numbers J and MJ in the weak field, and by = MJ + 2MS and MJ in the high field limit
Weak field Strong field
J MJ BB a BB a MJ
3/2 3/2 2B B + a/2 2B B + a/2 2 3/2
3/2 1/2 (2/3)BB + a/2 BB 1 1/2
1/2 1/2 (1/3)BB a a/2 0 1/2
1/2 1/2 (1/3)BB a a/2 0 1/2
3/2 1/2 (2/3)BB + a/2 BB 1 1/2
3/2 3/2 2B B + a/2 2B B + a/2 2 3/2
since it would lead to two crossings (marked with pink circles) which are avoided.
While theses states with MJ = 1/2 belonging to the 2 P1/2 and 2 P3/2 doublet
levels strongly interact, the MJ = 3/2 states of the 2 P3/2 term are not at all in-
fluenced by the 2 P1/2 level the latter simply has no MJ = 3/2 component.
Thus the splitting of these terms is at all field strengths always linear in B:
V2,3/2 = 2B B + a/2.
With the eigenvalues (8.28)(8.33) one may now solve the system of equations
(8.26) for the coefficients cML MS which according to (8.21) determine the respective
states |LSMJ . In the case of small B fields these will approach the corresponding
C LEBSCH -G ORDAN coefficients, for large fields, however, one of the |LML SMS
states will dominate the linear superposition, in fact the state with ML = 2MJ
and MS = MJ . We refrain from entering into the details of this somewhat
tedious but trivial computation.
However, it is instructive to verify that the general solutions (8.28)(8.33) con-
verge to the limiting cases for small and large fields as previously derived in
Sect. 8.1.2 and 8.1.3, respectively. Table 8.3 summarizes the results of expand-
ing the roots for BB a and BB a into a power series of x = BB /a and
1/x = a/BB ; the states are characterized in these limits by the quantum numbers
J , MJ and , MJ , respectively.
Comparing the three left columns in Table 8.3 with (8.14) we see that indeed the
dependence of the term energies on the magnetic field is exactly reproduced. The
three right most columns agree fully with (8.16) for the PASCHEN -BACK effect if
one neglects the fine structure term for very high fields. The remaining small shifts
of the terms energies are indicated schematically in Fig. 8.8 and Fig. 8.9 most
evident for the |LSMJ = |1 1/2 0 1/2 and |1 1/2 0 1/2 states.
8.1.6 Avoided Crossings
Problems of the type just discussed how do degenerate or nearly degenerate en-
ergy levels of quantum systems change under the influence of external or internal
fields (magnetic or electric) is of quite general importance. It was first treated by
B REIT and R ABI (1931) for the splitting of hyperfine levels in magnetic fields (so
called B REIT-R ABI formula, Eq. (9.53)). Prominent examples are found in mag-
netic resonance spectroscopy, i.e. in electron spin resonance (EPR) and nuclear spin
resonance (NMR) which will be treated in some detail in Chap. 9. However, similar
problems occur also in different context e.g. in molecular bond formation or in
atomic collision processes as we shall see in Vol. 2. Whenever two or more states,
originally separated in energy, are exposed to an interaction that may become larger
than their initial separation, one may in principle expect crossings of the energies
as a function of the perturbation. Such a perturbation induced energy degeneracy
warrants detailed inspection.
We treat the problem for the general case, choosing, however, the most simple
situation of only two relevant states involved, |1 and |2 . Be H 0 the unperturbed
Hamiltonian and V (q) the perturbation as a function of a characteristic experimental
parameter q (e.g. the strength of an electric field, the distance of two atoms and so
on). Assume that in 0th order

H0 1(0) = W (0) 1(0) and H 0 2(0) = W (0) 2(0) .
1 2
We seek for eigenstates |a and eigenenergies Wa of the perturbed system:

H0 + V (q) |a = Wa |a where a = 1 or 2

or H 0 + V (q) Wa |a = Wa(0) + V (q) Wa |a = 0.
By multiplication from the left with b| (where b = 1 or 2) one obtains a system of

linear equations
/ 0
(0)
W1 W + V11 V12 c1
=0 (8.34)
V
(0)
12 W W + V22
2
c2
for the expansion coefficients cab of the eigenstates:

|1 = c11 1(0) + c12 2(0) and

|2 = c21 1(0) + c22 2(0) .
Solutions exist if and only if the secular equation holds:

(0)
W1 W + V11 V12
=0 (8.35)
V12 W2 W + V22
(0)
(0)
W1 W + V11 W2(0) W + V22 |V12 |2 = 0.
The corresponding energies W1 and W2 thus are:

(0) (0)
W2 + W1 V11 + V22
W1,2 = + (8.36)
2 2

1 (0) (0) 2
W1 W2 + V11 V22 + 4|V12 |2 .
2
W2 W = 2V12
W1
qx
perturbation parameter q
Fig. 8.10 Avoided crossing. As a function of a perturbation parameter q the energies W1 and W2
of two states are plotted: with (full lines) and without (dashed) interaction, respectively, i.e. with
finite or negligible off-diagonal matrix element V12 (q). The perturbation parameter q may be an
external magnetic or electric field, or e.g. the distance between two atoms
= 0, the re-
If the off-diagonal terms of the perturbation matrix vanish, V12 = V21
sults correspond to those obtained from perturbation theory in 1st order (for two
non-degenerate states):
(1) (0) (1) (0)
W1 = W1 + V11 and W2 = W2 + V22 .
Thus in this situation (V12 = 0) and depending on the behaviour of V11 (q) and
V22 (q) as a function of a perturbation parameter q, curve crossings are in principle
possible, i.e. W (1) (q) = W1(0) W2(0) + V11 V22 = 0 may occur at a certain
value of q = qx . If, however, the two states interact and V12 (qx ) = 0, the curve
crossing is avoided as illustrated in Fig. 8.10. According to (8.36) the splitting at the
approximate crossing point qx is
W = W1 W2 = 2|V12 |. (8.37)
We summarize: The potential energy curves of two states |a and |b do not cross
as a function of a characteristic perturbation parameter q, if the interaction potential
at the approximate crossing point is finite, i.e. if Vab = 0. Such a situation is called
avoided crossing.
Specifically, for the change of atomic energies in a magnetic field we have seen
in Sect. 8.1.5 that the matrix elements (8.24) of the perturbation HBLS do not vanish
if MJ = ML + MS = ML + MS = MJ : states with equal ML + MS = MJ couple.
Hence, such states cannot cross as illustrated in Fig. 8.9 (non-crossing rule).
8.1.7 Paramagnetism
The magnetic properties of matter are determined by the magnetic dipole moments
of its constituents (atoms and molecules). We give here a brief outline of how to
relate macroscopic quantities (volume magnetic susceptibility) to the microscopic

magnetic properties which we have treated above in the present chapter.
According to the basics of electrodynamics, the magnetic B field within a macro-
scopic material (e.g. an ensemble of atoms) is given by
B = 0 (H + M) = 0 H + M 0 H = r 0 H . (8.38)
H is the magnetic field strength in the material and M the so called magnetization,
i.e. the magnetic moment per unit volume of this material (both measured in A m1 )
with = r 0 being the magnetic permeability (0 = 4 107 N A2 ), r the
relative magnetic permeability and M = r 1 the magnetic susceptibility (both
dimensionless). Occasionally one also finds Mmol given per mol (so called molar
susceptibility).
Thus, the present task is to express the magnetization
M = (r 1)H = M H = M B/0 (8.39)
in microscopic terms. As long as we are interested in paramagnetism we may exploit

the fact that M is typically a very small quantity, so that to 0th order B = 0 H .
For of a single atom in an |J MJ state, MJ z = gJ B MJ is the effective mag-
netic moment, projected onto the direction of the external field according to (8.12)
valid for not too high magnetic fields. For the whole ensemble one has to average
MJ z over the population of the magnetic substates that are populated in a thermal
equilibrium (essentially) according to the B OLTZMANN distribution (1.54). With the
magnetic energy VB (MJ ) of an atom in an |J MJ state we obtain for the average:

MJ MJ z exp(VB (MJ )/kB T )
MJ z = . (8.40)
exp(VB (MJ )/kB T )
The partition function in the denominator ensures correct normalization. At not too
low temperatures T the energy (8.12) of an individual magnet VB (MJ ) is very small
compared to the average thermal energy kB T at room temperature we typically
have VB (MJ ) 105 eV kB T 0.010 eV. Thus the partition function is 1
and the term in the nominator of (8.40) may be expanded as

VB (MJ )
MJ z exp VB (MJ )/kB T gJ B MJ 1
kB T
B
= gJ B MJ + (B gJ MJ )2 .
kB T
We insert this into (8.40) and see that positive and negative values of MJ compensate
each other in 0th order while the 1st order terms in the lead to a finite positive
contribution B/T . We finally obtain for the magnetization
C
M = NMJ z = B,
T
where N is the particle density of the material (dimension L3 ). With (8.39) the well
known
C
C URIE law M = (8.41)
T
for paramagnetism follows, M being the magnetic susceptibility.
For not too dense matter the C URIE constant C may be easily evaluated for
specific cases of L, S and J . If one extends the parameter range to very low values
of the temperature T or to very high magnetic fields B the full sum (8.40) has to
be used, the so called L ANGEVIN function, with the variable B/T . It leads to a
saturation of the magnetization at large B/T . In the general case one has, however,
also to account for the fact that a more rigorous treatment gives M = M/H . As
a simple exercise the reader may evaluate in detail the case L = 0, S = J = 1/2.
Paramagnetism requires a nonvanishing total angular momentum J of the ma-
terial under investigation. In the electronic ground state in which we normally find
matter the angular momentum is often zero and the paramagnetism of the atoms or
molecules studied is determined by their electron spin. As it turns out, not many
materials in nature happen to have unsaturated electron spins. Typical examples are
the alkali metals and a few other metals (not the alkaline earth metals, however).
Among the molecules are only a very few: one important example for molecular
paramagnetism is oxygen, O2 in its 3 1 ground state, as we shall see in Sect. 3.6.9,
Vol. 2.
8.1.8 Diamagnetism
Diamagnetism is a general property of all matter. It is relevant mainly if the atoms

or molecules under study do not have a permanent dipole moment. To understand
diamagnetism we have to use the H AMILTON operator for the interaction with the
electromagnetic field in its exact form expressed by the vector potential A of the
field as explained in Appendix H.1. In summary, one replaces p p + eA so that
=
H
p2 LS
+ V (r) + V
2me
in an external field becomes
= p2 aL
S eA
p e2 2
H + V (r) + + + A
2me 1 23 4 2 me 2me
1 23 4 potential, 1 23 4 1 23 4 1 23 4
kin. energy, spin-orbit, paramagn., diamagn. (8.42)
etc. etc.
which correspond to >0 < 0.
As explicated in Sect. 4, Sect. 5.4 and Appendix H, the term eA p /me leads
to electric and magnetic dipole interaction, er E and gJ B B J , respectively,
which are responsible for electric polarizability (Sect. 8.2.10) and paramagnetism
(Sect. 8.1.7). It is also the cause for electric quadrupole interaction (Sect. 5.4). In
contrast, the quadratic term in (8.42) leads to the so called ponderomotive potential
in intense fields (to be discussed in Sect. 8.5.1), as well as to diamagnetism as we
shall show now.
In comparison to paramagnetism (as far as present in a material) diamagnetism
is a very small effect, counteracting the external magnetic field according to L ENZs
law. From a classical point of view it describes the deceleration of the electron orbits
around the nucleus as a consequence of the external B field: the magnetic moment
of one atom induced by the latter is
MJ z =
)M H =
)M B/0 (8.43)
with )M being the magnetic susceptibility per particle (dimension L3 ). Since MJ z

itself depends on B we have to read here the magnetic energy (6.29) as a differential
equation, i.e. it builds up as the magnetic field increases:
)M B

dVdmag = MJ z dB = dB so that
0
1)M 2 1 dVdmag
Vdmag = B or )M = 0
. (8.44)
2 0 B dB
Let us now derive )M from the quadratic term in (8.42). For a static B field
according to (H.12) the vector potential is A = (B r)/2. Thus, the diamagnetic
component of the Hamiltonian is
2 2
dmag = e 1 (B r) (B r) = e B 2 r 2 sin2 ,
V
2me 4 8me
with being the angle between B and r. For a spherically symmetric atom one aver-
ages over all solid angles using sin2 = 2/3 and obtains in 1st order perturbation
theory the diamagnetic energy
dmag = e2 2 2 2 e2
V B r sin = B2 r2 (8.45)
8me 12me
with r 2 being the expectation value of the squared electron distance from the
atomic centre. The diamagnetic energy (8.45) thus obtained has to be compared
to (8.44). We finally obtain with 0 = 1/(0 c2 ), the fine structure constant and
the atomic unit of length a0 :
2 2
e2 2 2 r r
)M = 0
r = a03 2 2 = 1.12 104 a03 2 . (8.46)
6me 3 a0 a0
The diamagnetic susceptibility is thus negative and extremely small as expected.

For atoms in their ground state the expectation value of r 2 is on the order of mag-
nitude of a02 . (In the case of atomic hydrogen we may use the closed expressions
according to Sect. 2.6.10.)
With the particle density N or alternatively the AVOGADRO constant NA the

macroscopic quantities per volume or per mol are
M =
)M N and Mmol =
)M NA , (8.47)
respectively. With this and (8.47) even for solid materials in the most dense close-
packed arrangement of spheres the macroscopic diamagnetic susceptibility is only
on the order of M 0.83 104 . We mention in passing that while the value
of )M increases with r 2 the particle density decreases with 1/r 3 so that the
macroscopic M decreases with the size of the atoms.
We finally note that in standard atomic spectroscopy the diamagnetic energy is
minute and does not play a significant role. However, this is different when high
lying RYDBERG states are studied. Since the radius of these atoms increases with
n2 and r 2 a02 n4 the ratio of the diamagnetic term (8.45) to the normal magnetic
field splitting (8.8) is
dmag
V e B
Ba 2 n4 106 . . . 107 n4 . (8.48)
VB 6gJ 0 T
For n = 34 at only 1 T it is already 1. Consequently, spectroscopy of RYDBERG

atoms in magnetic fields cannot neglect the diamagnetic term.
Section summary
The interaction energy of an electron with an external magnetic field is de-
scribed by VB = B (Lz + 2Sz )B/.
In low magnetic fields B (spin orbit interaction V L
LS = a S/2 V B ) one
uses the |(LS)J MJ basis and considers V B as a small perturbation. The en-
ergy is then gJ B MJ B with the L AND factor gJ in (8.11).
In the limit of high magnetic fields, i.e. for very low Z atoms, extremely large
B fields, or high principle quantum numbers n, on treats V B in the uncoupled
|LML SMS scheme, while V LS is a small perturbation. In this case, L, S, ML
and MS are good quantum numbers as well as MJ = ML + MS . In addition,
= ML + 2MS has to be introduced.
Vector models are useful to visualize the projection of magnetic moments
onto each other and onto the external field. They are based on the precession
of magnetic moments at L ARMOR frequency, which can be rationalized in
quantum mechanical terms.
By diagonalizing the full Hamiltonian (8.6), its matrix form being given in
Table 8.2, one may also treat the intermediate situation between low and
high fields exactly. The energies are given by B REIT-R ABI type formulas, see
(8.28)(8.33). Characteristic are interactions between states with equal MJ
which lead to the non-crossing rule for such states.
Avoided crossings are important also in a broader context: they occur if de-
generate or quasi-degenerate states are coupled by a non-diagonal interaction
which changes as a function of a characteristic parameter.
8.2 Atoms in an Electric Field 399
From the microscopic energy of atomic dipole moments in a magnetic field

one derives the macroscopic magnetic susceptibility M of matter. Specifi-
cally C URIEs law for the temperature dependence of M = C/T in param-
agnetic materials, and the diamagnetic susceptibility (M < 0) has been ob-
tained. The latter is very small, but rises n4 .
8.2 Atoms in an Electric Field

8.2.1 Introduction
The splitting of energy terms in static electric fields, the so called S TARK effect, has
been discovered by Johannes S TARK in 1913 (N OBEL prize 1919). In traditional
spectroscopy it plays only a minor role since static electric fields available in the
laboratory (see Table 8.4) are typically very small compared to the inner atomic
fields. The latter are on the order of magnitude
Ze V
Eatom = 2
= 5.14 1011 Z . (8.49)
40 a0 m
This may be compared to the electric field strength for electric breakdown in air
which is about 10 kV cm1 = 1 106 V m1 which is thus far below the electric
field strength needed to modify the atomic energies significantly.
8.2.2 Signicance
Nevertheless, for quite a number of reasons it is important to discuss this theme

thoroughly:
1. Electric fields play a key role in the structure and composition of matter: be
it in molecular bonding, the forces in crystal lattices, soft matter or plasmas.
The electric fields by which neighbouring atoms and ions interact are essential
ingredients for our understanding of the properties of matter.
Table 8.4 Typical electric field strengths

Example |E| / V m1
In an electric power line 102
Close to a charged up plastic hair comb 103
Surface of the drum of a photocopier or laser printer 105
Electric breakdown in air 106
In an H atom on the first B OHR orbit 5 1011
In focused short pulse laser at an intensity of I = 1020 W cm2 3 1013
On the surface of a Uranium nucleus 3 1022
2. As an example, the polarizability of atoms approaching each other is a typical

effect to be explained by electric fields. The electron clouds of two neighbouring
atoms repel each other, thus forming two interacting dipoles. As we shall see in
Sect. 8.3 this allows us to describe the long range potentials by which the atoms
attract each other. Corresponding considerations are basic for understanding the
formation of molecular bonds.
3. The polarizability of atoms and molecules is of fundamental importance in many
areas of physics, specifically so in alternating electromagnetic fields. In the field
of an electromagnetic wave polarizability leads to the refraction of light. The
dependence of the index of refraction on the frequency (dispersion) can be ex-
plained on an atomistic basis as will be shown in Sect. 8.4.2.
4. In Chap. 6 we have documented by way of example that spectral lines may today
be measured with a relative accuracy of up to 1013 or 1014 . In such precision
measurements electric stray fields of only some V m1 may influence the results
and need to be considered seriously!
5. In modern atomic and molecular physics highly excited RYDBERG states play an
important role. Since the radius of the excited states increases with n2 the inner
atomic field decreases with n4 . For n = 100 the inner atomic field is on the
order of only some kV cm1 . With corresponding external fields that may easily
be generated in the laboratory the electronic energy levels of RYDBERG atoms
and molecules can be influenced significantly.
6. An important field of current research is concerned with oscillating electromag-
netic fields at very high intensity, as e.g. generated in the focus of intense ultra-
short laser pulses. The amplitude of the electric field E0 is related to the laser
intensity I by (4.3):

E0 I
1
= 2745 . (8.50)
Vm W cm2
Present technology allows for intensities of up to at least 1020 W cm2 . Thus,
electric fields today available in the laboratory are far beyond inner atomic fields
according to (8.49). And presently worldwide several laser systems are built up,
to push the limits even further. The vision is to study matter in the laboratory
under extreme conditions that otherwise only exist in the interior of stars. Buzz
words are highly relativistic plasma dynamics (ion acceleration, nuclear fusion)
or new approaches for particle physics (extreme energy densities can lead to
particle generation).
8.2.3 Atoms in a Static, Electric Field
A static electric field which is a polar vector field in contrast to the axial magnetic
field breaks the symmetry of the H AMILTON operator. The perturbation by the
electric field may be written in 1st order approximation
VE (r) = D E = er E (8.51)
and hence H (r) = H

(r). For convenient calculation of the matrix elements we
rewrite (8.51)
for Ez as VE (r) = ezE = er cos = eErC10 (, ) (8.52)
making use of (4.75) and of the renormalized spherical harmonics Ckq according to
(B.29). Note that the squared orbital angular momentum operator
2
L (which implies
differentiation with respect to ) does not commute with this perturbation. Hence
2
L does no longer commute with H and L is not any more a good quantum number
in contrast to the situation for an atom in a magnetic field where according to (8.5)
the relevant interaction is proportional to L z + 2Sz , which commutes with
2
L but
no longer with
2
J .
However, even with (8.52) included in the Hamiltonian it still commutes with L z

since C10 (, ) = C10 ( ) and VE (r) does not depend on , while Lz = i/ acts
only on , and Sz does not act on the spatial coordinates at all. Hence, ML and MS
remain good quantum numbers in the electric field. As we shall see, the interaction
matrix elements depend only on |ML |. Hence, it is reasonable to characterize the
eigenstates by L2z and |ML |. In the extreme case we expect a superposition of many
orbitals with different L but a constant value of |ML |: a so called hybridization
occurs a phenomenon crucial for the understanding of chemical bonding.
In the following considerations we shall concentrate for simplicity on systems
with only one active electron (the valence electron). In the general case one has to
replace in (8.51) the vector r by the sum r i over all electron coordinates. Typi-
cally, for coupled systems the reduction formulas for evaluating the matrix elements
are accordingly more complicated.
8.2.4 Basic Considerations about Perturbation Theory
A comparison between the presently discussed perturbation (S TARK effect) with

those previously treated is shown in Table 8.5, along with the most important effects
related to these interactions.
An estimate for the order of magnitude of the S TARK effect is obtained from
(8.52) by

VE (r) ea0 E,
assuming that the angular dependent matrix elements are on the order of 1 and
the radial matrix element with r a0 . For the breakdown field strength in air
Emax = 106 V m1 we thus find

VE (r) < 5 105 eV =
0.4 cm1 .
The effect is thus expected to be indeed very small, by comparison to (8.7) even
significantly smaller than the Z EEMAN effect under standard laboratory conditions.
We distinguish two limiting cases depending on the non-C OULOMB term VnC in
the perturbation hierarchy Table 8.5:
Table 8.5 Perturbation hierarchy for effective one electron atoms: Summary of interactions and
consequences (gqn stands here for good quantum number)
H like Alkali like FS splitting Z EEMAN effect S TARK effect
=H
H C (r) +VnC (r) +VLS +VB +VE (r)
pure C OULOMB electrostatic spin-orbit ext. B field ext. E field
Tkin + C/r not 1/r L S B (Lz + 2Sz )B er E
L degeneracy L degeneracy M degeneracy |M| degeneracy
removed removed removed
(LS)J MJ ,
[H Sz ] = 0 ,
[H
2
L ] = 0
coupling J no longer gqn L no longer gqn
VnC (r) VE (r) : In this case the electric field removes for the first time L
degeneracy. This has to be treated by perturbation theory for degenerate states
(see Sect. 3.3.4). One finds the so called linear S TARK effect.
VnC (r) VE (r) : L degeneracy is already removed when the E field is
switched on. Due to the symmetry of the perturbation potential (z has odd par-
ity) all diagonal matrix elements disappear. Hence, the energy does not change
in 1st order perturbation theory and one has to resort to 2nd order for determin-
ing the energy shift. Consequently the so called quadratic S TARK effect E 2 is
observed!
8.2.5 Matrix Elements
The perturbation VE by the electric field (8.51) does not act on the spin in contrast
to the B field (8.6). The spin S and its projection MS thus remain conserved. This
facilitates the evaluation of the matrix elements

J M VE (r)| J M (8.53)

= eE J M rC10 ( ) J M = eE r| J M C10 ( )|J M
significantly. For later use we derive these now in detail. All quantum numbers re-
lated to the radial wave functions are summarized by while J M stands here for all
angular momentum quantum numbers. Depending on the strength of the interaction
as compared to FS interaction the states are again described most suitably in the
uncoupled |LML SMS or in the coupled scheme |(LS)J MJ , respectively. Here L,
S and J stand again for the total orbital angular momentum, the totals spin and the
total angular momentum of the system, respectively, while the E field is assumed
to act only on one active electron with the angular momentum . For the present
derivations we shall assume for simplicity L = . However, the following may eas-
ily be extended to coupled orbital angular momenta. Obviously, the matrix elements
(8.53) are essentially the same which we have already encountered in Chap. 4 in
the context of describing E1 transitions induced by linearly polarized light. We thus
obtain the same selection rules as in Sect. 4.4.
Strong Electric Field

We first consider the case of a high electric field in the uncoupled scheme:

S Ms L M VE | SMs LM

= eErn n S Ms L M C10 |SMs LM

= eErn n L M C10 |LM S S Ms Ms (8.54)

with rn n = n r|n = Rn (r)Rn (r)r 3 dr (8.55)
0
and the radial wave function Rn (r). Using (C.29) and (C.30) we obtain

+M L 1 L

L M VE | LM = eErn n (1) 2L+L
2L + 1
M 0 M
5
L 5C1 L (8.56)

L M VE | LM = eErn n M M L L1 (1)M 2L + 1 (2L + 1)

L 1 L L 1 L
, (8.57)
M 0 M 0 0 0
assuming for (8.57) a pure one electron system L = and L = and making
use of the 3j symbols symmetries (B.37) and (B.49): Thus, the matrix element
is only non-zero for L = L 1. Hence, as already mentioned, L is no longer
a good quantum number, while the projection quantum number M = M is con-
served. In contrast to the magnetic field where the angular momentum (an axial
vector) explicitly is included in the perturbation, in the case of an electric field the
quantity z (derived from the polar vector r) does not act on the projection of the
angular momentum. Using (B.53) for the relevant 3j symbols we finally obtain:

(L + 1)2 M 2
(L + 1)M VE | LM = eErn n (8.58)
(2L + 1)(2L + 3)

L2 M 2
(L 1)M VE | LM = eErn n . (8.59)
(2L 1)(2L + 1)
The matrix elements only depend on M 2 and thus on the absolute value |M| of
the projection quantum number. This is a consequence of the fact that on inversion
(thereby changing +M to M) the sign of VE changes, while at the same time one
(and only one) of the spherical harmonics of the wave functions changes since they
have different parity due to L = L 1.
According to (8.58) and (8.59) the influence of the electric field decreases with
increasing |M| L. Thus, as we shall elaborate in Sect. 8.2.7, states with the highest
projection quantum number show the weakest S TARK effect.
Again, the matrix elements show that

2
L does not commute with z, hence neither
with H. On the other hand, M is more specific as observable in an electric field.
Thus, as detailed in Appendix D, it is convenient in such cases to use the real rather
than the usual complex basis for angular momentum eigenstates.
Weak Electric Field

We now compute the matrix elements for the S TARK effect in the coupled scheme
|(LS)J M , i.e. for the case when the electric interaction is small compared to spin-
orbit interaction, VE VLS . Again the interaction operator VE does not act on the
spin of the electron. The expression equivalent to (8.56) for the matrix element is
obtained by making use of the tools developed in Sect. C.3.1. From (C.53) one
obtains

L SJ M VE | LSJ M

= eErn n 2J + 1 (2J + 1) 2L + 1 (2L + 1)
" #
J J 1 L L 1 L 1 L
M M L L1 (1)MS
.
M M 0 J J S 0 0 0
(8.60)
Specifically with (C.55) for a one electron system with S = 1/2

L SJ M VE | LSJ M
= eErn n M M L L1 (1)M3/2 (8.61)

J J 1 J J 1
2J + 1 (2J + 1)
M M 0 1/2 1/2 0
holds. Explicitly from (B.53) and (B.55) this can simply be written as

L SJ M VE | LSJ M = eErn n M M L L1 (8.62)

(J +1)2 M 2
for J = J + 1,

(2J +2)
(1)2M1 (2J +1)
(1) for J = J, and
2J
2J (J +1)(2J +2) M

J 2 M 2

2J for J = J 1.
Here too the projection quantum number is conserved (in this case the projection
of J ), and only the matrix elements with L = L 1 are non-zero. Even the rule that
the states with highest projection quantum number |M| are least disturbed remains
valid, since a quantitative evaluation of (8.62) shows that the opposite trend for
J = J contributes only very little.
Fig. 8.11 Typical energy W, n

level positions in atoms; we
consider the S TARK shift for
the state marked with a black W n +1
dot
Wn
W n 1
8.2.6 Perturbation Series

In the following we specify the S TARK effect in more detail for quasi-one-electron
systems. Depending on whether the states under investigation are still essentially1
degenerate (for the H atom, H like atoms, high RYDBERG states) or not (e.g. for
the alkali atoms) we have to apply perturbation theory for degenerate states or use
the standard perturbation series, respectively. In the latter case energy and wave
function of a state |a are changed as
a|VE |b 2
Wa = Wa(0) + a|VE |a + and
Wa Wb
b=a
a|VE |b (0)
a = a(0) + .
Wa Wb b
b=a
According to (8.57) and (8.60) quite generally a|VE |b 1 , i.e. the diagonal
term vanishes. With (8.52) we have
|zab |2 r 2 |a|C10 |b |2
Wa Wa(0) = |eE|2 = |eE|2 ab
, (8.63)
Wa Wb Wa Wb
b=a b=a
and the change of energy depends on the square of the electric field: if the degener-
acy is already removed, a quadratic S TARK effect is observed. For effective one elec-
tron systems, |a and |b represent |n m or |n sJ M , depending on the coupling
case. Only states with equal m or M, respectively and with = 1 interact (mix).
Only the absolute values |m | or |M| of the projection quantum numbers matter.
8.2.7 Quadratic STARK Effect

Considering the typical term positions according to Fig. 8.11, equation (8.63) allows
us to make already a few qualitative statements:
(a) the S TARK effect will always lower the terms since the atomic level spacing
typically decreases with increasing principle quantum number n as sketched in
1 Note that this depends on the precision of the measurement: for very weak electric fields and very
high precision even the H atom levels of equal n are already split due to FS interaction.
Fig. 8.12 Estimating the z

relative magnitude of the y
matrix elements 2pz |C10 |2s
(top) and 2px |C10 |2s x
(bottom). It is obvious that the
|M|= 0 |M|= 0
components of the integral in
the upper case overlap much |pz |s
more than in the lower case.
The absolute value of the C10
matrix element with the lower
projection quantum number |M|= 1 |M|= 0
|m| = 0, i.e. 2pz |C10 |2s , |px |s
thus has to be significantly
larger
Fig. 8.13 Visualizing why E z E z

an M = 0 state is more
readily polarized than an + y + y
|M| = 1 state
x x
|M|= 0 |pz |M|= 1 |px
Fig. 8.11; thus the series in (8.63) contains always many more closely spaced
levels for which Wa Wb < 0 as compared to those levels where the reverse is
true; this is particularly pronounced for the ground state;
(b) higher lying levels are stronger influenced by the S TARK effect since Wa Wb
decreases with increasing principle quantum number n;
(c) within one level states with larger |m | are lowered less as already mentioned.
In the uncoupled case the latter statement is read directly from the matrix ele-
ments (8.58) and (8.59). It is valid, however, also in respect of |M| in the coupled
case as explicit evaluation of the matrix elements (8.62) shows. One may visual-
ize this with the help of Fig. 8.12 where the components of the matrix elements
2pq |C10 |2s are illustrated. In physical terms positive and negative charges may be
displaced by the E field more easily for the |2pz state along the z-axis than for the
|2px or |2py state. In the latter case the positive charge would have to be more or
less extracted from the negative charge cloud as illustrated in Fig. 8.13.
Thus, for the quadratic S TARK effect one expects a characteristic dependence of
the energies of different |M| states as a function of the applied electric field E as
sketched in Fig. 8.14 for the example of a 2 P3/2,1/2 doublet.
Corresponding to the splitting with |M| in emission or absorption spectra one
finds in an electric field only two polarization components and : these are lin-
early polarized (perpendicular to each other) since the electric field does not enforce
an orientation onto the system again in contrast to the situation in a magnetic field
where + (LHC) and (RHC) light is emitted at two different frequencies.
Fig. 8.14 Quadratic S TARK

W W (0)
effect for the example of the
Na 3 2 P1/2,3/2 states 32P 3/2 |m|=3/2
3P
|m|=1/2
32P1/2
|m|=1/2
without field with field E
8.2.8 Linear STARK Effect
The situation is completely different if the external E field is the dominant interac-
tion which removes degeneracy. Thus, let us assume states of different parity are
degenerate without the external E field. This is the case for the H atom and H like
ions (as long as FS can be neglected) as well as for extremely high field strengths
where the splitting is negligible (e.g. in molecular bond formation between sev-
eral atoms) but also for highly excited RYDBERG states and high as we shall see
in Sect. 8.2.9.
For the 1s 2 S ground state of the H atom all matrix elements of VE vanish ac-
cording to (8.57) or (8.62) because of L L1 . The H atom ground state has no linear
S TARK effect. However, for the first excited state (with FS neglected) there are four
degenerate states, written in the real basis,
|2s0 , |2pz , |2px , |2py . (8.64)
The states |2s0 , |2pz are characterized by M = 0, while |2px and |2py have a
projection quantum number |M| = 1. According to (8.57) all diagonal matrix ele-
ments disappear, as well as those to different M and M . With (8.58) or (8.59) only
two matrix elements are non-zero:
1
2pz |VE |2s0 = 2s0|VE |2pz = eEr2s2p . (8.65)
3
The radial matrix element (8.55) between the states |2s and |2p may be de-
rivedfrom the radial wave functions of the H atoms (Table 2.2) and is r2s2p =
(3 3/Z)a0 .
The H AMILTON matrix is thus
2s0 2pz 2px 2py

W
(0) 3eEa0 0 0 2s0

0 2pz
0 + VE = 3eEa0
H
W (0) 0 , (8.66)

0 0 W (0) 0 2px

0 0 0 W (0) 2py
0 | = W (0) | . The S CHRDINGER equation

with H
0 + VE W )| = 0
(H
may be solved algebraically. Since only two states couple, this amounts to solving a
set of the linear equations for the coefficients cn ,m
(0)
W W 3eEa0 c2s0
= 0, (8.67)
3eEa0 W (0) W c2pz
for which the secular equation is

(0) 2
W W (3eEa0 )2 = 0.
It has two possible solution for the energy values W1 and W2 :
W (0) W1,2 = 3eEa0 or

W1 = W (0) + 3eEa0 and (8.68)
W2 = W (0) 3eEa0
With (8.67) the coefficients c2s0 and c2pz are derived (properly normalized)

(1) c2s0 = c2pz = 1/ 2 and (2) c2s0 = c2pz = 1/ 2
for W1 and W2 , respectively. In summary, the eigenenergies and eigenstates of the

H(n = 2) states, perturbed by an external electric field, are:

(1) W (0) + 3eEa0 : |2 = |2s0 |2pz / 2 (8.69a)

(2) W (0) 3eEa0 : |2+ = |2s0 + |2pz / 2 (8.69b)
(3) W (0) : |2px (8.69c)
(4) W (0) : |2py . (8.69d)
These states correspond to the electronic charge distributions illustrated schemat-

ically in Fig. 8.15. The energy Wa in the electric field (including the quadratic
S TARK effect which has not been evaluated here explicitly) depends on the field
strength E as sketched in Fig. 8.16.
We note that the asymmetric charge distribution of the states |2 and |2+ re-
sults in a finite dipole moment D at of these special, excited states in the electric
field. We may thus explain the raising or lowering of the energies as due to the in-
teraction of this dipole moment with the electric field. From (8.68) one reads the
magnitude of the dipole moment directly from the energies:
VE = Wa Wa(0) = 3eEa0 = D at E. (8.70)
Hence, the dipole moment of the H atom in the state |2 and |2+ states is 3ea0
and 3ea0 , respectively.
Fig. 8.15 Dipole states (so

called S TARK states |2 of
the excited H atom in an
electric field as the sum and z =
difference of |2pz and |2s ,
respectively
+ =
(|2pz |2s) / 2 = |2
Fig. 8.16 S TARK effect for

W W (0) E, z |2
the H(2s, 2p) states as a
function of the field strength. x
Linear S TARK effect of
degenerate states (black E, z |2px
|2
dashed lines) and transition to 2s, 2p
the quadratic S TARK effect x
(full red lines). On the right |2px,y
the corresponding orbitals are E, z
indicated |2+
|2+ x
8.2.9 An example: RYDBERG States of Li
As an experimental example for the S TARK effect we shall discuss here highly ex-
cited atomic states (so called RYDBERG atoms). This is a wide field and still subject
of present research. We present just one particularly impressive pioneering exper-
iment of Z IMMERMAN et al. (1979). It is considered a benchmark experiment for
pertinent theoretical investigations (see e.g. M ENENDEZ et al. 2005) and demon-
strates a number of important aspects of the S TARK effect. The experimental setup
is relatively simple, as illustrated schematically in Fig. 8.17. A Li atomic beam (see
footnote 24 in Chap. 1) is excited by a resonant multi-photon process with three
laser frequencies: the steps are 2s 2p (671 nm), 2p 3s (813 nm), and finally
3s 15p (626 nm). At a fixed DC S TARK field, the latter laser is tuned through
ca. 100 cm1 . Excitation of the RYDBERG states is detected by field ionization in a
pulsed electric field (HV pulser), applied shortly after the laser pulse.
The spectra, observed for a number of electric field strengths, are reproduced in
Fig. 8.18 (black, vertical traces with horizontal excitation lines). The red lines in
Fig. 8.18 give the theoretically expected energy dependence of the relevant states
as a function of the electric field strength. The observed absorption lines follow the
theoretical understanding very impressively. There are two sets of states, for M = 0
and |M| = 1, which are excited by light polarized parallel and perpendicular to the
E field ( and light), respectively. In the previously discussed case of the 2s 2p
delayed trigger
HV pulser
DC
atomic beam
laser beams
collecting
ion detector
electronics
Fig. 8.17 Experimental setup according to Z IMMERMAN et al. (1979) for RYDBERG spectroscopy
in an electric field. Two laser systems tuned to the transitions 2s 2p and 2p 3s, respectively,
excite the Li atoms into the 3s state. This is then further excited by a very narrow band, tuneable
laser into the n = 15 region. Shortly after the excitation, in addition to the S TARK field a pulsed,
high electric field (HV power supply) is applied which ionizes all RYDBERG atoms. Detected is
the ion signal
binding Lithium I (a) M=0 (b) M =1

energy / cm-1
440
16s
470
n =15 15
15p
500 15p
15s
530
0 2500 5000 0 2500 5000

electric field / Vm-1
Fig. 8.18 RYDBERG levels n 15 in an electric field as a function of the field strength, for atomic
Li from Z IMMERMAN et al. (1979). (a) |M| = 0, (b) |M| = 1. Black spectra represent the experi-
mental excitation probabilities, red lines give the calculated term positions
states this would correspond to excitation of the |2 , |2+ states (with light) and
the |2px and |2py states (with light).
In the present case, for n = 15, practically all levels with 2 14 = n 1
are degenerate without electric field (a total of 13 levels). Thus, all these levels split
with increasing electric field by linear S TARK effect. For the 15p state and very
weak electric field (300 V m1 ) one recognizes that the degeneracy is already
removed and the energy decreases initially in a quadratic fashion with increasing
field. As the electric field gets larger VE supersedes the initial splitting and the
linear S TARK effect takes over also for the 15p state. The 15s state behaves analo-
gously, however, due to the larger initial splitting the linear region is reached only
at about 2000 V m1 . There, however, already numerous avoided crossings with the
neighbouring levels n = 14 are encountered, as evident from Fig. 8.18 left (M = 0).
A similar behaviour is seen at higher energies for the 16s levels.
It is also interesting to compare the |M| = 0 and |M| = 1 levels. Even though
the general pattern is very similar in both cases, one recognizes very clearly that the
interaction and hence the repulsion in the vicinity of avoided crossings for terms
of equal symmetry is much larger for |M| = 0 in comparison with |M| = 1: this
clearly reflects the magnitude of the corresponding matrix elements according to
(8.58) and (8.59), respectively.
8.2.10 Polarizability
We have seen that the S TARK effect is quadratic when the degeneracy is already
removed among interacting states of different parity. This is the case for almost all
atoms in their ground and lower excited states at moderate electric fields. It does not
only change the energy but also the wave functions, and thus the electronic charge
distribution around the nucleus changes. This is called polarization. Quantitatively,
the electric field induces an electric dipole
D el = E E (8.71)
in each atom, with the (microscopic) polarizability E being a characteristic prop-

erty of the atom. Changing the electric field by dE changes the interaction energy
by
dW = D el dE = E EdE. (8.72)
Thus, the total energy of the induced dipole is

E
W W (0) = E EdE = E 2 . (8.73)
2
Conversely, one may derive the polarizability E from the energy change with
changing field strength. With (8.72)
W
Del =
E
holds, and using (8.63) the polarizability (here of an atom in state |a ) becomes
Del 1 Wa e2 |E|2 |zab |2

E = = =
E E E E E Wa Wb
b=a
|zab
|2 2 |a|C |b |2
rab 10
= 2e2 = 2e2 . (8.74)
Wb Wa Wb Wa
b=a b=a
After averaging over all initial substates states with projection quantum number ma ,
inserting the transition frequencies ba = Wba / with Wba = Wb Wa , and using
(opt)
the oscillator strength fba as defined in (5.27), one may write the polarizability E
in compact form as:2
e2 fba
(opt)
E = 2
. (8.75)
me ba
b=a
We note in passing that the above discussion refers to the so called induced po-
larization. The situation is different, if the medium consists of particles with a per-
manent dipole moment (e.g. water molecules). In that case the permanent electric
dipole moments will tend to orient along the electric field. Similar to the discussion
for magnetic moments in the context of paramagnetism, one has to consider the en-
ergetics and statistics involved in this orientation processes to obtain the orientation
polarization.
It is instructive to compare this quantum mechanically exact calculation with the
classical model (J.J. T HOMSON) considering the atom as a harmonic oscillator of
eigenfrequency 0 . The driving mechanical force Fm = me 02 z must be compen-
sated by the electric force Fel = eE in the field E:
Fm = me 02 z = Fel = eE.
The displacement z corresponds to a dipole moment
e2
D = ez = E,
me 02
from which the classical expression for the polarizability follows:
e2 1
E = . (8.76)
me 02
Comparison with (8.75) shows that the classical formula corresponds to an atom
with only one single transition frequency ba = 0 . In contrast, the quantum me-
chanical result distributes the ability of the electrons to oscillate over all transition
2 Note that the SI unit of the polarization is [E ] = A2 s4 kg1 = C m2 V1 . In a.u. the polarization
(au)
isE = E /(40 a03 ). Often the esu system is still used in this context (see also Appendix A.3)
with
(esu)
(esu) E E 106
E = or 3
= E ,
40 cm 40 m3
(esu)
indicating that E is usually given in cm3 . This scaling allows a direct comparison of the polar-
izability with the volume of the atoms that are polarized.
frequencies. The T HOMAS -R EICHE -K UHN sum rule (5.28), which for a one elec-
(opt)
tron system reads fba = 1, may be seen as the mathematical manifestation of
this picture.
The classical formula (8.76) allows us to obtain a rough estimate for E in the
ground state of the H atom. Setting 0 = Eh /2 leads to
E (40 )4a03 . (8.77)
Generalizing this with the aid of (8.75) one finds the (static) polarizability to be
proportional to the third power of the extension a of the polarized object. This fully
corresponds to the findings of classical electrodynamics.
8.2.11 Susceptibility
In analogy to the magnetic properties Sect. 8.1.7 one derives the macroscopic di-
electric susceptibility from the microscopic quantity polarizability E . Electric
displacement D and electric field E are related by
D = 0 E + 0 E = r 0 E = 0 E + P, (8.78)
where = r 0 is the macroscopic dielectric permittivity, with the electric con-

stant 0 = 1/0 c2 and the relative dielectric permittivity r (formerly dielectric
constant). With (8.78) the (electric) polarization of a medium is3
P = (r 1)0 E = 0 E, (8.79)
measured in [P] = C m2 . It relates to the polarizability by
P = N E E, (8.80)
with N being again the particle density (number of atoms per volume). Thus, the
dielectric susceptibility is
N E E
= (r 1) = or per atom a = . (8.81)
0 0
To obtain some feeling for the order of magnitude of we make an estimate for
our favourite Na atoms where nearly all oscillator strength is concentrated in the
main transition 3s 3p ( 3p3s = 1.696 104 cm1 ). Setting f3p3s 1 we obtain
with E from (8.75) the static dielectric susceptibility per atom
e2
a = = 3.12 1028 m3 . (8.82)
0 me (2c 3p3s )2
3 Note that this differs slightly from the scheme in (8.39) for the magnetization M.
This may be compared to the atomic volume. With rNa 3a0 it is Vat 1.7
1029 m3 or roughly a /4 . This is also what classical theory predicts.
Section summary
The interaction of atoms with a static electric field E is described by the dipole
energy er E = erC10 ( )E, assuming the field to be parallel to z. This op-
erator breaks the symmetry of the Hamiltonian, so that the orbital angular
momentum L is no longer a good quantum number, in contrast to M. Inter-
action matrix elements are only finite between states with L and L 1. They
are independent of the electron spin.
Consequently, the change of the energy of LM levels in an electric field
(called S TARK effect) depends on M 2 . If L degeneracy for the energy lev-
els is already removed, the S TARK effect is quadratic in E and negative. For
L > 0 the levels split into L + 1 sublevels while no S TARK effect occurs for
isolated s states.
If L degeneracy is not yet removed, or the splitting between different L levels
is small compared to the dipole energy, the dipole operator mixes different L
states, but M is conserved. The splitting of the new states is then linear in E
(for not too high fields).
Typically, the S TARK splitting is a small effect in standard spectroscopy. How-
ever, since it depends quadratically on the radial matrix elements, it increases
with principle quantum number n4 and becomes substantial for high RYD -
BERG states.
The polarizability of atoms is following the quadratic S TARK effect also
derived from the same interaction. According to (8.75) it is obtained as sum
over all relevant oscillator strengths divided by the respective squared transi-
tion frequencies. It is related to the relative dielectric permittivity r and to
the dielectric susceptibility by (8.81).
8.3 Long Range Interaction Potentials
We shall now address the question how two atoms or molecules (or their ions) inter-
act as they approach each other. At this point we are interested in distances R large
enough so that no chemical bonds can be formed yet, while on the other hand the
respective charge distributions of the particles begin to influence each other. This
range of distances is so to say the precursor for the formation of macroscopic
matter and plays a key role in plasma physics, scattering physics, kinetic gas theory
etc. A detailed derivation for the relevant interaction potentials has been given by
B UCKINGHAM (1967). We summarize here the key results and make some plausi-
bility considerations.
8.3 Long Range Interaction Potentials 415
Monopole Monopole: R 1 In the most elementary case one ion of charge q1 e

interacts with another one of charge q2 e by the well known C OULOMB law:
q1 q2 e 2
V (R) = R 1 . (8.83)
40 R
Monopole Permanent Dipole: R 2 Following the rules of electrostatics, the

potential of a charge qe in the field E of a dipole (Fig. 8.19), e.g. a diatomic het-
eronuclear molecule, is given by:
qeD eR qeD
V (R) = D E = 2
= cos
40 R 40 R 2
R 2 with eR = R/R. (8.84)
Fig. 8.19 Geometry

q R D
monopole permanent dipole
+
For later use we also note without proof the corresponding electric field that is ob-
tained as gradient of (8.84):
1
E= 3
D 3eR (D eR ) . (8.85)
40 R
Permanent Dipole Permanent Dipole: R 3 Two heteronuclear, diatomic

molecules with permanent dipole moments present a prototypical example. As in-
dicated in Fig. 8.20 both dipoles may be oriented differently in respect of R. Com-
bining (8.84) and (8.85) gives:
1
V (R) = D 1 E(D 2 ) = 3
D 1 D 2 3(D 1 eR )(D 2 eR )
40 R
D1 D2
= [cos 12 3 cos 1 cos 2 ] R 3 . (8.86)
40 R 3
Fig. 8.20 Geometry D1 D2

permanent dipole
1 2
permanent dipole
R
Monopole Quadrupole: R 3 Without detailed reasoning we communicate that

the same dependence on distance holds also for the interaction of a point charge
with a quadrupolar charge distribution, the latter being e.g. a neutral atom in a p
state or a homonuclear molecule:
V (R) R 3 . (8.87)
Monopole Induced Dipole: R 4 This is a particular important case, often en-

countered, e.g. when an electron or an ion of charge qe interacts with a neutral atom
and polarizes its electron charge cloud as indicated in Fig. 8.21. With the polariz-
ability E of the electron shell according to (8.75) we have:

E 2
V (R) = D ind dE = E E dE = E(R) (8.88)
2
2 (au)
E qe E (qe)2 1 E q 2
= = or = in a.u.,
2 40 R 2 32 2 02 R 4 2R 4
Fig. 8.21 Geometry q R -

monopole induced dipole
+ - - +
-
(au) (esu)
with E = E /(40 a03 ) = E /a03 (see footnote 2). Note that this so called
polarization potential is always attractive. For an estimate of the order of magni-
tude we consider an H atom in the field of a singly charged ion and use (8.77)
as an approximation for the polarizability. In this case V (R)/Eh = 2(R/a0 )4 ,
so that in a distance of one B OHR radius the polarization potential would be
just 2 a.u. Of course the formula holds strictly only for significantly larger dis-
tances.
Quadrupole Quadrupole: R 5 Important examples for this case are e.g. the
interaction of two homonuclear, diatomic molecules as well as the interaction of
one excited atom in a px state interacting with such a molecule. The geometry is
shown in Fig. 8.22. Without entering into further details, we record the dependence
on the distance:
F (1 , 2 )
V (R) = R 5 . (8.89)
R5
Fig. 8.22 Geometry

quadrupole quadrupole + 1 + 2
+ R +
Permanent Dipole Induced Dipole: R 6 An example for this case is the in-
teraction of a heterogeneous molecule with a neutral atom as indicated in Fig. 8.23.
This case too is easily derived since the induced dipole is always aligned parallel to
the field of the permanent one. In analogy to the case monopole induced dipole
one obtains the potential as

E 2
V (R) = D ind dE = E E dE = E(R)
2
8.3 Long Range Interaction Potentials 417

E 1 2
= D 3e R (D e R ) (8.90)
2 40 R 3
E D 2
= 2 6
1 + 3 cos2 E R 6 ,
2(40 ) R
Fig. 8.23 Geometry D

permanent dipole induced
dipole
R +
where the second line follows with (8.85). In the limiting cases, D R and D R,
the angular dependence in the bracket gives a factor of 4 and 1, respectively. In any
case one finds an attractive, non-isotropic potential proportional to R 6 .
Induced Dipole Induced Dipole: R 6 The famous VAN DER WAALS potential
occurs in the interaction of all neutral atoms or molecules depending on their polar-
izability. It always ensures an attractive interaction at large distances as far as not
another of the above discussed cases dominates.
Fig. 8.24 Geometry induced A - R - B

dipole induced dipole
-+- + + - -
- -
One may visualize this interaction to arise from spontaneous charge fluctuations
that lead to dipole formation in one of the atoms. This initial dipole in turn induces
by polarization a dipole in the other atom and so on, until a stable situation as de-
picted in Fig. 8.24 is reached.
The resulting dipole dipole interaction will lead to an expression similar to
(8.90) except for the fact that there is no preferred direction and one has to average
over all alignment angles . In the quantum mechanical calculation one expands the
electrostatic interaction of all charges involved (valence electrons and ion core) for
large R into a 1/R N series. Summing over all electron coordinates r A (at atom A)
and r B (at atom B) one encounters typical dipole terms H di (er A )(er B )/R 3 that
correspond to (8.86).
One has to treat these in 2nd order perturbation theory (in 1st order the dipole
terms of a neutral charge distributions disappear). The interaction energy in 2nd
order is the sought-after polarization potential:
|a|H
di |b |2
V (R) = . (8.91)
Wa Wb
b=a
Here a specifies again the initial state (usually the ground state) and b are all inter-
mediate states. Here too, as discussed in Sect. 8.2.7, for the vast majority of relevant
terms Wa Wb < 0 holds, and the sum is negative. Comparing this with (8.74) one
sees that this essentially amounts to compute the polarizability of both atoms. Since
di R 3 and V (R) according to (8.91) depends quadratic on the matrix elements
H
di , the overall interaction potential is attractive and proportional to R 6 :
of H
C
VAN DER WAALS potential V (R) = . (8.92)
R6
One may evaluate (8.91) approximatively (see B UCKINGHAM 1967), and finds
3 WA WB A B
V (R) = . (8.93)
2(40 )2 WA + WB R 6
The key parameters are the polarizabilities, A and B , together with average bind-
ing energies, WA and WB , of particle A and B, respectively one typically uses the
respective ionization energies.
The force derived from VAN DER WAALS interaction is often called dispersion
force as the polarizability E () depends on the frequency of the external electric
field.
Section summary
The long range interaction potentials between ions, atoms and molecules play
an important role in spectroscopy as well as in scattering physics. They are
determined by the structure of the interacting particles and their plausibility.
For large R the interaction potential is 1/R for monopole-monopole inter-
action, 1/R 2 for monopole-permanent dipole, 1/R 3 for permanent
dipole-permanent dipole, 1/R 4 for monopole-induced dipole, 1/R 5
for quadrupole-quadrupole, and 1/R 6 for dipole-induced dipole as well
as induced dipole-induced dipole.
8.4 Atoms in an Oscillating Electromagnetic Field
8.4.1 Dynamic STARK Effect
How does an atom behave in an oscillating electromagnetic field? As discussed in

detail in Chap. 4, optical transitions are induced only by resonant irradiation. How-
ever, even for non-resonant irradiation one may observe (much weaker) transitions
of 2nd order. We shall discuss this so called R AMAN scattering in detail in Chap. 5,
Vol. 2.
In contrast, at present we want to ask whether also the eigenenergies of the atomic
states do change in an oscillating electromagnetic field. The answer is of course yes,
because the quadratic S TARK effect does not distinguish between positive and neg-
ative field. A more detailed consideration in the framework of QED shows that the
treatment in the static case according to (8.63) has to be amended by introducing
the photon energy in the resonance denominators. One speaks of dressed states.
Even when considering only one atomic level, we have to allow for the possibility
8.4 Atoms in an Oscillating Electromagnetic Field 419
that for a very short time t a virtual photon is emitted or absorbed, as we have
discussed and visualized by F EYNMAN diagrams in Sect. 6.5.6. Such transient ex-
citation is possible even for non-resonant irradiation since the energy W of states is
defined only within the limits of the uncertainty relation, so that W > / t holds.
On a very short time scale the levels thus get completely blurred and may be quasi
excited. One often talks in this context somewhat misguiding about virtual
intermediate levels.
Thus, we expect a shift of the atomic levels by a modified quadratic S TARK ef-
fect. A clean derivation of the dynamic polarizability warrants some quite serious ef-
forts. We reduce these here to a heuristic consideration by replacing the total energy
of the system Wa prior to absorption or emission of a photon (angular frequency )
by
Wa Wa + n
and the energies Wb of the intermediate states by
Wb Wb + (n 1).
The signs refer to absorption or emission, respectively. Introducing these energies

into the expression (8.74) for the static polarizability, summing over emission and
absorption terms, and averaging over all initial states, we obtain in the dynamic case:

1 |zba |2 |zba |2
E () = e 2
+ .
ga Wb Wa Wb Wa +
b=a,ma
The first term stands for absorption, the second for emission. As a last step we
replace again Wb Wa = ba :

1 |zba |2 / |zba |2 /
E () = e2 + (8.94)
ga ba ba +
b=a,ma
2e2 1 ba |zba |2 e2 fba

(opt)
E () = = . (8.95)
ga 2 2
ba me 2 2
b=a,m a b=a ba
(opt)
We have again introduced the optical oscillator strength fba according to (5.27).
In the static limit 0, expression (8.75) for the static polarizability is recovered.
Conversely, in the limit of very high frequencies we make use of the T HOMAS -
(opt)
R EICHE -K UHN sum rule (5.28) for fba (considering only one active electron) and
obtain
e2
E () for ba , (8.96)
me 2
where the minus sign indicates that the induced dipoles have opposite direction as
the polarizing electromagnetic field.
Finally, the shift of the atomic energy levels is derived from (8.73), now using
the dynamic polarizability (8.94):
E () 2 E () 2 E () I
Wa Wa(0) = E = E0 =
2 4 2 0 c
e2 I fba
(opt)
Wa Wa(0) = . (8.97)
20 cme 2 2
b=a ba
We have averaged here over the square E(t) 2 = E0 cos(t) 2 of the electric field
strength and introduced the intensity I of the electromagnetic radiation, using (4.2).
8.4.2 Index of Refraction
From elementary optics we know that the index of refraction n and the relative
dielectric permittivity relate as

n = r (8.98)
(assuming for the relative magnetic permeability = 1 in very good approxima-
tion). Hence, with the dynamic polarizability (8.95) and (8.81), the index of refrac-
tion n is obtained from
N e2 fba
(opt)
N E ()
n2 1 = r 1 = = = 2 2
. (8.99)
0 0 me
b=a ba
The resonance frequencies for typical optical materials are usually far away from
the visible region and dispersion dn/d < 0 (so called normal dispersion). One may
express (8.99) in terms of the wavelength, = 2c/ and write it as so called
S ELLMEIER equation, which is often used in optics:
Bi 2
n2 = 1 + . (8.100)
2 Ci
i
Usually one treats this as an empirical relation with typically three pairs of S ELL -
MEIER coefficients Bi , Ci obtained from experiment (a comprehensive set of data
for optical materials is given by P OLYANSKIY 2012).
For thin media (gases) n 1, so that n2 1 = (n 1)(n + 1) 2(n 1) and we
obtain as a good approximation
N e2 fba
(opt)
n1+ . (8.101)
20 me 2 2
b=a ba
We finally mention that in condensed matter the particle density is N 1/Vat ,

so that = a N becomes on the order of 1 quite different from magnetism
where the magnetic susceptibility was a very small quantity. Thus one has to make
an important modification: in dense media the atoms experience an electric field, the
so called L ORENTZ field, which is already modified as compared to the field in the
surrounding vacuum. According to C LAUSIUS -M OSSOTI this leads to a modified
formula:
N e2 fba
(opt)
n2 1
= . (8.102)
n2 + 2 30 me 2 2
b=a ba
In the limit of n 1, of course, (8.101) is recovered.
8.4.3 Resonances Dispersion and Absorption
So far we have tacitly assumed that the frequency of the radiation is far away from
resonances, ba = . In order to include also frequency regions close to resonance
we have to consider damping, i.e. we must account for the finite lifetimes b =
1/b of excited states. As in Sect. 5.1.1, we simply introduce a complex energy (or
transition frequency) ba ba ib /2 so that
1 1
.
ba ba ib /2
Inserting this into (8.94) and (8.99) leads to a complex index of refraction
nc = n + i. (8.103)
For thin media, (8.101) is now replaced by real and imaginary parts:
(opt)
N e2 fba ba ba +
n=1+ + (8.104)
40 me ba ( )2 + b2 ( + )2 + b
2
b=a ba 4 ba 4

N e2 fba
(opt)
b b
= + . (8.105)
40 me 2ba ( )2 + b2 ( + ) 2 + b
2
b=a ba 4 ba 4
Far from all resonances, i.e. for negligible b , one recovers from (8.104) the original
relation (8.101). In the general case the wave vector is now also complex and one
has to replace k (2/0 )nc ek = (2/0 )(n + i)ek , with ek being the unit vector
in the direction of propagation. Thus, the electric field (4.1) in a medium becomes
i
E(r, t) = E0 eei(2nz/0 t) + e ei(2nz/0 t) e2z/0 (8.106)
2
where we have assumed the light to propagate along the z-axis. The vacuum wave-
length 0 is shortened to = 0 /n, while implies exponential damping of the
wave (in physics texts is often called extinction coefficient). The intensity of the
radiation I |E(r, t)|2 decreases with distance as
I = I0 exp(4z/0 ). (8.107)
n normal dispersion (a)

anomalous
dispersion
1
absorption lines (b)
Fig. 8.25 Schematic of the complex index of refraction nc as a function of the incident radiation
frequency ; (a) real part, n, and (b) imaginary part, . The absorption lines shown in the
extinction coefficient clearly correspond to regions of anomalous dispersion in the index of
refraction n
Comparing this to the L AMBERT-B EER law (4.21) we see that we have found a
microscopic interpretation of the absorption coefficient introduced there.
In the vicinity of an isolated resonance, ba , only one term in the sums
(8.104) and (8.105) dominates and the real index of refraction is
(opt)
N e2 fba ( ba )/b
n=1+ , (8.108)
40 me ba b [( ba )/b ]2 + 1/4
showing close to resonance the characteristic anomalous dispersion region with

dn/d > 0 (or dn/d < 0). For the absorption coefficient we find from (4.21),
(8.107) and (8.105) the familiar L ORENTZ profile:
(opt)
4 N e2 fba
= = . (8.109)
ba 40 cme [( ba )/b ]2 + 1/4
As illustrated in Fig. 8.25, the real part n (8.104) of the index of refraction
may have several regions of normal and anomalous dispersion, while the imagi-
nary part (8.105) represents absorption, with several L ORENTZ profiles of dif-
ferent strength. Clearly, these absorption lines correspond to regions of anomalous
dispersion in the real part n().
8.4.4 Fast and Slow Light
We do not want to end this discussion on the interaction of electromagnetic radiation

with a medium without at least mentioning one fascinating aspect of modern optics
that presently enjoys much attention (e.g. B OYD et al. 2010, and further publications
in that special issue of J. Opt.), and may be of practical interest in the future. The
essential ingredients that we briefly recall here, are already taught in undergraduate
physics.
We consider the plane wave (8.106) propagating in +z-direction. For simplicity,

we assume linear polarization, and use only the second exponential (with positive
exponent) which we write4
E + (t, z) = E0 ei(tk()z)()z (8.110)

where k() = 2n/0 = 2/ and () = 2/0 ,
with 0 and = 0 /n being again the wavelengths in vacuo and in the medium,
respectively, while k() is the real (propagation) and () the imaginary part (ab-
sorption) of the wave vector in the medium, with n and according to (8.104)
and (8.105). The positions of constant phase = t kz are given by z =
(/k)t /k and propagate with the

dz c
phase velocity vp () = = = , (8.111)
dt =const k() n()
where we have used the familiar relations vp = and c = 0 with = 2.

To transport information, one has to imprint some recognizable temporal struc-
ture onto the wave. The amplitude E0 must depend on time and position, i.e. it
becomes a field envelope. In the spirit of the so called SVE approximation (see
footnote 1 in Appendix H) we consider a pulse of a quasi-monochromatic wave,
with a carrier frequency c , and a pulse duration t 1/c . At a given position in
space, say at z = 0, we thus write (8.110) as
E + (t) = E0 (t)eic t . (8.112)
As derived in Appendix I.2 we may write this as F OURIER transform (I.22),

+ 1 ) c )eit d,
E (t) = E( (8.113)
2
where

)
E() = E0 (t)eit dt, (8.114)

is the F OURIER transform of the field envelope

1 )
E0 (t) = E()e it
d. (8.115)
2
When describing also the spatial dependence, we have to account for dispersion:
as we have seen in the previous section, in a medium neither n nor vp are constant
4 Weuse here the notation of Appendix I.2 where the relations to the full description are given
which in the present case would just be space consuming without leading to further insight.
with , and the wave vector k() is not simply proportional to . However, as only
a small interval of frequencies around c contributes, we may expand k around

c dk c
kc = : k() = kc + = + , (8.116)
vp (c ) d c vp (c ) c vg
with vg = d/dk. For an arbitrary value of z we replace the time t by t k()z/,

as in (8.110). We use the above expansion for k() and abbreviate = c to
obtain from (8.113)
E + (t) E0+ (t kz/)

ez = )
E( )ei(c + )ti(kc + /vg )z d( )
2 =
+
1 ) i(t vzg )
= ei(c tkc z)z E( )e d( ) .
2
In the last step we have exploited the fact that d( ) = d. Comparing the above
expression in square brackets with (8.115) we realize that it is just the field ampli-
tude shifted in time by z/vg . Hence, we may write the electric field of the wave at
t kz/:

z i(kc zc t)z
E0+ (t kz/) = E0 t e . (8.117)
vg
We clearly recognize that the carrier wave with carrier frequency c propagates with
the wave vector kc just as (8.110), while the E0 (t z/vg ) describes the form of the
wave-packet. If E0 (t) has its maximum at t = 0, as determined by an appropriate
)
choice of E(), at a different position z in space one finds this maximum E0 (0)
obviously for z = vg t. Thus, characteristic structures of the wave-packet travel with
the so called
d c c
group velocity vg = = = . (8.118)
dk n + d
dn ng
Here we have used d/dk = 1/(dk/d) and differentiated k = n/c from (8.111)
in respect of . In analogy to n = c/vp (which may be called phase index), one
introduces a
c dn
group index ng = =n+ . (8.119)
vg d
In the range of visible light for transparent materials as a general rule one finds
n > 1 and dn/d > 0 (normal dispersion), so that typically vg < vp < c: light pulses
propagate at velocities smaller than in vacuum. So far the canonical discussion on
phase and group velocity.
The discussion about slow and fast light starts with the fact that under cer-
tain conditions dispersion may be negative as already illustrated in Fig. 8.25 for
the standard dispersion behaviour of conventional materials close to a resonance.5
A look at the formula for the group index (8.119) shows, that one may then indeed
find n to be smaller than 1 or even negative which in principle implies the possi-
bility of group velocities larger that the speed of light in vacuum vg > c. One speaks
about superluminal propagation phenomena. And if n < 0, and hence vg < 0, this
would even lead to light travelling backward before it has arrived. In contrast, in
other regions of the spectrum vg may be extremely small so that light crawls so to
say, or is being stopped. And indeed, such phenomena have been observed during
the past years in fancy experimental setups (see e.g. B OYD and G AUTHIER 2002).
To caution the reader quite in the beginning: E INSTEINs paradigm according to
which information can never be communicated faster that the speed of light in vac-
uum is by no means violated as a consequence of group velocities vg > c.
We want to exemplify these phenomena quantitatively on the basis of a simple
case study. Let us consider Na vapour as medium through which light is supposed
to propagate. For this consideration Na represents to a good approximation a two
level system if we allow only light frequencies in the vicinity of the 3 2 S 3 2 P
transition (Na D-line). Let us assume a target density of N = 2 1013 cm3 which
may be achieved in a gas cell without problems. The other relevant parameters are
the oscillator strength for the transition, fNaD = 0.98, and the decay probability for
the excited state A = = 6.15 107 s1 (corresponding to a lifetime = 16.2 ns),
0 589 nm (in vacuo) and 0 3.2 1015 s1 . Hence, the prefactor in (8.108) is
Ne2 fNaD /(40 me 0 ) = 0.08 and 0 / = 5.2 107 .
In Fig. 8.26 the situation close to the resonance line is depicted to scale as a
function of 0 in units of the linewidth. Figure 8.26(a) shows the real index of
refraction (phase index), and Fig. 8.26(b) the absorption coefficient. One recognizes
that n undergoes the typical change between normal and anomalous dispersion, and
finally back to normal with values close to 1 for which (8.108) is valid. Absorption,
on the other hand, is quite substantial contrasting the very moderate changes of the
index of refraction. Figure 8.26(c) reports the group index ng according to (8.119). It
shows dramatic changes in the resonance region. Correspondingly, the group veloc-
ity (8.108) also varies very fast across the resonance as shown in Fig. 8.26(d). When
approaching a resonance vg becomes very small, increases again, passes through
zero for 0 = /2 accompanied by two singularities in each case, and as-
sumes a very small negative value exactly on resonance, 0 = 0.
Generally speaking, experiments trying to observe the expected phenomena are
extremely difficult, the main problem being that the most interesting effects, relying
on |vg | c or |vg | > c, or even vg < 0, are expected where absorption is particularly
high. Thus, typically instead of passive two level systems one investigates three and
more levels schemes which also act as amplifiers when pumped with a suitable laser.
Beyond atomic gases, today specially designed solid state materials are employed,
5 Today, in addition a variety of artificial, specially designed smart, meta and nano materials
exist, as well as photonic fibres, with extended regions of quite unusual optical properties.
(a) n (c) 106 ( 0) /

1.08
1.04 -4 -2 2 4
( 0) / - 106
1.0 ng - 2106
-4 -2 2 4
- 3106
0.96
(d) - 4106
0.92 vg /c = 1/ng
210- 5 1.0
(b) / 1000 cm-1
15
110- 5
10 0.5
5 ( 0) / ( 0) /
0
-4 -2
- 410- 6
-4 -2 0 2 4 0 2000 4000
Fig. 8.26 (a) Absorption and (b) index of refraction (showing changing dispersion) in the
vicinity of a resonance line. The data is shown to scale for the example of the Na D line at
N = 2 1013 cm3 . For comparison (c) the group index and (d) the group velocity are shown;
note the different scales in the left and right half of (d), the overall profile being symmetric around
0 = 0
such as doped light transmitting fibres, photonic crystals, and assemblies of quantum
dots.
Propagation of light pulses faster than light as well as slowing down of such
pulses has been observed. A detailed analysis allows to understand all observations
reported in the literature from the basic principles of optics even if they may ap-
pear as mysterious superluminal effects: information can be transported by such
schemes, but not faster than the speed of light in vacuum. Genuine discontinuities
within a wave always propagate with velocities less than c, while precursors of the
wave may reach the observer much earlier. Without proof we mention that energy
transport in an electromagnetic wave is always characterized by a propagation ve-
locity
2n
c,cf = (8.120)
n2 + 1
which is definitively less than c, the speed of light in vacuum.
While such superluminal effects may just be a fascinating theme of fundamental
research, the inverse case, light propagating at extremely low velocities, vg c, al-
beit perhaps less spectacular, could possibly be of much more practical relevance in
data transmission technology: One may envisage special delay lines, data switches
or optical storage devices that profit from these phenomena. One could e.g. slow
down a data stream temporarily by a control laser if the traffic somewhere down
the transmission lines is jammed and later on release it again on demand. Fig-
ure 8.26(d) shows in principle, how low the group velocity can be for close to a
resonance line.
Fig. 8.27 Light extremely

slowed down according to T = 450 nK
H AU et al. (1999). One sees 25
= 7.05 0.05 s
light signal / arb. un.

the not delayed reference 20 L = 229 3 m
pulse (open circles) and the vg = 32.5 0.5 ms-1
pulse delayed by ca. 7 s in 15
the B OSE -E INSTEIN
condensate (full circles). The 10
delay corresponds to a speed 5
of light of ca. 32 m s1
0
-2 0 2 4 6 8 10 12
/ s
An important breakthrough was achieved by H AU et al. (1999), who were able

to demonstrate for the first time in a very cold atom gas (to be precise: in a B OSE -
E INSTEIN condensate) group velocities down to 17 m s1 . Figure 8.27 shows an
example from this work. One sees two light pulses in comparison: the undelayed
reference pulse and a signal pulse delayed by ca. 7 s! Since then, much progress
has been made in this field and many ingenious concepts have been explored. These
effects are now studied in a variety of media and technological applications appear
feasible (we refer the interested reader to review and the recent special issue allready
mentioned: B OYD and G AUTHIER 2002; B OYD et al. 2010, and to references there).
8.4.5 Elastic Scattering of Light
With the tools developed in the previous sections we are now also able to treat the
important subject of elastic scattering of light. Light cannot only be absorbed, excit-
ing atoms or molecules in the absorbing medium as presented in detail in Chap. 4.
It can also be scattered elastically.
We first note that the very intense light scattering one observes from dust and
smoke particles or from small water droplets in vapour or fog (e.g. in the head-
lights of cars, in discos and laser shows) will not be subject to this section. This
kind of elastic light scattering is called M IE scattering. It dominates if the scattering
particles have dimensions on the order of the wavelength or larger. The angular dis-
tribution of M IE scattering depends on the size and form of the scattering particles
and has to be calculated directly from M AXWELLs equations with corresponding
boundary conditions. Such computations may become quite complicated. For the
identification and analysis of nano and micro particles M IE scattering is an often
used important tool. Its theoretical treatment leads, however, beyond the scope of
this textbook (a very detailed treatment is e.g. found in B ORN and W OLF 2006).
Here we shall discuss light scattering from atoms (and molecules) which es-
sentially arises from atoms being polarized by the electromagnetic radiation. More
precisely, the oscillating electric field E(t) of the wave induces a time dependent
dipole moment D(t) = E E(t) in the atom, with E being the polarizability of the
atoms discussed above. These dipoles in turn radiate at the frequency of the irradi-
ating field. One may calculate the emission classically from (4.33). The power P
emitted per solid angle is
dP
|D(t)| 2 E2 |E|2 4
= sin2
= sin2 (8.121)
d (4)2 0 c3 (4)2 0 c3
and depends on the angle between the polarization of the incident radiation and
the direction of the detected radiation. The latter is often replaced by the angle
between the polarizations of incident and scattered radiation, with sin2 = cos2 .
The differential cross section for this so called R AYLEIGH scattering of polarized
light is obtained by dividing dP /d by the incident light intensity I = c0 |E|2 =
c0 E02 /2:
2 4
E
dR 3
= 2
cos2 = R cos2 . (8.122)
d 2
16 0 c 4 8
Integration over all azimuthal and polar angles gives a factor 8/3, so that the inte-
gral R AYLEIGH cross section is given by
2 4
E 2
8 3 E
R = = . (8.123)
602 c4 302 4
Equation (8.122) represents again the typical doughnut like radiation characteristic
which we have already encountered with resonant fluorescence in Sect. 4.5. One
may observe R AYLEIGH scattering quite conveniently from a laser beam propagat-
ing in air. Depending on the laser intensity it can be seen in a darkened laboratory
even in completely dry and dust free air, and one easily verifies that the emission
parallel to the direction of linear polarization of the scattered light vanishes com-
pletely (i.e. perpendicular to the laser beam in one particular azimuthal direction).
For unpolarized light one has to average (8.122) over the two directions of po-
larization and obtains
dR 3
= R 1 + cos2 , (8.124)
d 16
where is now the angle between incident and scattered light direction.
For low frequencies , in the IR and often even in the VIS spectral region, E
is essentially independent of incident frequency and corresponds to the static value
(8.75). Thus, if we define a mean oscillator frequency 0 , characteristic for each
atom, by
1 f (opt)
2
= ba
2
, (8.125)
0 b=a ba
the total elastic scattering cross section for 0 is given by
4
R = e . (8.126)
04
This is formally identical to the classical relation already found by Lord R AYLEIGH,
who considered 0 to be the eigenfrequency of the atomic electron.
The proportionality R 4 in (8.123) gives a very clear answer to the often
posed question why is the sky blue?: blue light is scattered much more efficiently
than red light essentially by molecular oxygen and nitrogen being the main con-
stituents of air. Their absorption bands are in the VUV region and do not influence
the visible spectrum. The sky, which we register as elastically scattered light from
the sun, thus appears blue. Conversely, this also explains the red colour of the sun
rising and setting.
Note, however, that for higher photon energies, in contrast to the classical low
energy limit (8.126), the elastic scattering cross section (8.123) shows pronounced
structures over a wide range of wavelengths as expected from the polarizability
(8.94). In the vicinity of resonance frequencies particularly intensive scattering phe-
nomena are observed even where absorption is still negligible.
For high enough photon energies, W = 0 , the polarizability simply be-
comes E = e2 /(me 2 ) according to (8.96), with 0 being representative for the
binding energy of any relevant electron in the atom. Then the integral cross section
in (8.122)(8.124) approaches the so called T HOMSON cross section (measured in
units of barn, see Appendix A.2)
e4 8 4 2 8 2
R e = = a0 = r = 0.665 b, (8.127)
ab 2 2
60 me c 4 3 3 e
de
so that for polarized light = re2 sin2 . (8.128)
d
The electrons thus behave quasi like free electrons for which this cross section was
first derived by J.J. T HOMSON on a fully classical basis. Here is the fine structure
constant, a0 the atomic units of length, and re the classical electron radius (see
Appendix A). Note that (8.127) is a rather suggestive relation, but we recall that re
has nothing to do with a radius of the electron the latter being as point-like as we
can measure today.
So far, in all above discussion we have neither considered the particle properties
of the photon nor used special relativity for treating its motion. Both will become im-
portant at high photon energies W , i.e. when = W/me c2 is no longer very small.
For really free electrons energy and momentum conservation must be satisfied si-
multaneously, so that the electron will accept and the photon will loose some energy:
C OMPTON scattering occurs, which we have introduced briefly in Sect. 1.4.2. We
rewrite (1.75) as
W 1
= (8.129)
W 1 + (1 cos )
where W = and W = are the incident and scattered photon energy, re-
spectively, while is the photon scattering angle in respect of the incoming light
beam. Without derivation we communicate here the (relativistically correct) K LEIN -
N ISHINA formula for the differential C OMPTON scattering cross section:

dC re2 W 2 W W
= + 2 + 4 cos .
2
(8.130)
d 4 W W W
Note, that in the low energy, non-relativistic limit = W/me c2 1 the C OMPTON
shift vanishes according (8.129), i.e. W /W 1, and the K LEIN -N ISHINA formula
(8.130) reproduces the differential T HOMSON cross section (8.128) for elastic pho-
ton scattering from a free electron.
Still, the treatment of elastic light scattering from atoms (or molecules), i.e. from
bound electrons is not yet complete. First, up to now we have treated only one single
active electron. In the long wavelengths range and even in the region of atomic
resonances this can (in principle) be cured by simply summing (8.125) or (8.95)
over a sufficient number of absorption frequencies and thereby accounting for all
Ne electrons of a system. Nevertheless, the scattering process will be dominated
by only a few resonance transitions that are closest to the frequency of the incident
radiation.
At higher energies, W = ba , specifically for the scattering of X- and
gamma-rays where the wavelength becomes comparable or even smaller that atomic
dimensions, and energy and momentum conservation has to be considered, the sit-
uation is much more complicated (a detailed discussion is found e.g. in K ANE et al.
1986). Most commonly, the form factor approximation is used: one sums explicitly
over all electrons, thereby accounting for their probability density, and superposes
the scattered radiation appropriately. Two cases are distinguished (see e.g. H ANSON
1986, who also describes explicitely how the polarization angle is converted into
scattering angle):
1. The equivalent to R AYLEIGH scattering (elastic) at these energies is called co-

herent scattering. One argues that the recoil momentum q = k k from the
scatted photon can be accepted by the atom as a whole without significant en-
ergy exchange (with q = 2k sin(/2) = 4 sin(/2)/). Thus, the photon energy
is conserved and the radiation from all electrons corresponding to their density
is superposed coherently. From (8.127) one then derives the differential cross
section for coherent scattering of unpolarized radiation
dR r2
= e F 2 (q, Z) 1 + cos2 , (8.131)
d 2
where F(q, Z) is the atomic form factor as defined in (1.97), here emphasizing
its dependence on atomic number Z.
2. C OMPTON scattering from atoms, called incoherent scattering, is described in a
analogue manner, applying the K LEIN -N ISHINA formula (8.130) but now adding
cross sections for photons of different final energy incoherently. As detailed e.g.
by H UBBELL et al. (1975) one generalizes the definition of atomic form factors
to include excited states, with |a describing the ground and |b any excited state:

Z
F (b) (q, Z) = b| exp(iq r i )|a so that F (a) (q, Z) = F(q, Z). (8.132)
i=1
One defines an incoherent scattering function, essentially adding cross sections:

S(q, Z) = F (b) (q, Z)2 . (8.133)
b=a
The differential cross section for incoherent scattering of unpolarized radiation

is then:
2
daC r2 W W W
= e S(q, Z) + sin 2
. (8.134)
d 2 W W W
State-of-the-art numerical values of form factors for all atoms are found e.g. at
C HANTLER et al. (2005) and B ERGER et al. (2010), including scattering angle in-
tegrated coherent and incoherent scattering cross sections, photoionization data etc.
One finds that the integrated cross sections decrease with increasing photon energy
for both, coherent and incoherent light scattering.
We finally mention T HOMSON scattering from a relativistic electron beam as an
interesting topic with potential applications for the generation of short pulsed X-ray
radiation. We shall briefly discuss this in Sect. 10.6.2.
Section summary
In an oscillating electromagnetic field (angular frequency ) the energy of
atomic energy levels also changes (so called dynamic S TARK effect). Due to
the quadratic nature of the interaction the effect is obtained by simply replac-
2 with 2 2 ( denotes the relevant excitation frequencies). The
ing ba ba ba
atomic polarizability E () in the field is obtained by the same substitution
from its static limit E .

The index of refraction n = r , in a medium of density N , is obtained from
the atomic polarizability by n2 = 1 + N E ()/0 .
In spectral regions with dn/d > 0 we speak about normal dispersion, while
dn/d < 0 corresponds to anomalous dispersion. The latter situation is as-
sociated with absorption. Introduction of the respective damping constants
(linewidths of the transitions) leads to a complex index of refraction nc =
n + i.
We distinguish phase and group velocity, vp = /k = c/n and vg = d/dk =
c/(n + dn/d), respectively. Typically vg < vp < c. Near a resonance vg is
very small. However, in the region of anomalous dispersion it rapidly changes
and we can find vg > c or even vg < 0. Still, information transport can only
occur with velocities c.
8.5 Atoms in a High Laser Field
Modern developments in ultrafast laser science open new dimensions of light matter
interaction and are the basis for an active, cutting edge field of research (a glimpse
of the kind of processes studied today may be obtained e.g. from H ICKSTEIN et al.
2012). The radiation intensity I = W/(A t) in a focussed laser pulse scales with
the total energy W of the pulse, its focal area A, and pulse duration t. Thus, the
intensity of a 10 fs laser pulse is e.g. 106 times higher than that of a 10 ns pulse
which contains the same energy W . As discussed in Sect. 8.2.2 and exemplified in
Table 8.4, gigantic electric field strengths can be generated in this manner, presently
surpassing the field that an electron experiences on the first B OHR orbit in the H
atom by two orders of magnitude. And the experimental limits are being pushed
further.
Atoms and molecules exposed to such extreme conditions, react with a wealth of
astonishing new phenomena warranting also theoretical approaches quite different
from that outlined in Sect. 8.2.
8.5.1 Ponderomotive Potential
As discussed previously, an electromagnetic field interacting with a free electron

cannot just transfer photon energy directly onto the electron (C OMPTON scatter-
ing, as discussed in Sect. 8.4.5, is only relevant at very high photon energies ).
If, however, a third body (e.g. an atomic core) is present, energy and momentum
conservation may in principle be achieved with the help of this third particle. Our
discussion thus starts with a very conventional approach: we solve the classical,
non-relativistic equation of motion of an electron oscillating in an electric field of
amplitude E0 and frequency ,
dv
me = eE0 cos t,
dt
and obtain velocity and kinetic energy of the electron in the stationary case:
eE0 1 e2 E02
v(t) = sin t me v 2 = sin2 t. (8.135)
me 2 2me 2
The deflection of the electron from its average position is given by
eE0
x= cos(t) = x0 cos(t). (8.136)
2 me
For the oscillation amplitude one calculates with (4.2)

eE0 e 2I e2 2I
x0 = 2 = 2 = 2 2
, (8.137)
me me 0 c 4 c me 0 c
8.5 Atoms in a High Laser Field 433
UP / eV x 0 / nm
IH 1800 nm 800 nm
400 nm 1800 nm 800 nm
mec 2 200 nm 400 nm
106
50nm (b) 200 nm
10 50nm
104 (a) 13nm
1 13nm
102
1 0.1
10-2 0.01
1012 1014 1016 1018 1020 1012 1014 1016 1018 1020
I / Wcm-2 I / Wcm-2
Fig. 8.28 Ponderomotive potential (a) and maximum amplitude of an electron (b) in the field of
a short laser pulse of intensity I and wavelength ; the full red line corresponds to the wavelength
= 800 nm of the Ti:Sapph laser
which in convenient units reads:

x0 / nm = 1.3607 107 [/ nm]2 I / 1012 W cm2 . (8.138)
The average energy Up inherent to this quiver motion is called ponderomotive

potential:
1 e2 E02 1
Up = me v 2 = = me 2 x02 .
2 4me 2 4
Inserting (4.2) one obtains
e2 I e 2 I 2
Up = 2
= I 2 , or (8.139)
20 cme 8 2 0 c3 me

Up / eV = 9.3375 108 [/ nm]2 I / 1012 W cm2 . (8.140)
We point out that this expression is completely identical to (H.20), formally derived
from the term quadratic in the vector potential A in the exact (semiclassical) Hamil-
tonian (H.1) for an atom in an electromagnetic field.
The order of magnitude of Up and x0 is illustrated by Fig. 8.28 for a number
of wavelengths as a function of laser intensity I . The full red lines refer to the
fundamental of the Titanium-sapphire laser (short Ti:Sapph) at = 800 nm the
work horse of ultrafast laser science. As an example, an electron in the focus of
a laser beam at an intensity of 1014 W cm2 experiences a ponderomotive potential
of Up = 5.976 eV, and according to (8.138) the corresponding excitation amplitude
is x0 = 0.87 nm a huge electron motion as compared to typical atomic radii of
0.1 nm to 0.25 nm (see Sect. 3.1.5).
Clearly, electrons bound to an atom or molecule exposed to such field strengths
will experience dramatic changes of their wave functions and term energies. We thus
have to compare the ponderomotive potential (8.139) to the binding energies of the
electrons in atoms. In the low intensity limit, we expect energy shifts as we have
derived them for the dynamic S TARK effect in (8.97). And indeed, a comparison
(opt)
with (8.139) shows that in the limit of high frequencies ba , with fba = 1
both expressions become identical.
However, for really intense laser fields, in particular at longer wavelengths as
characterized in Fig. 8.28, a complete break down of the bound state description
developed so far in atomic physics is expected. Two specific limits are indicated:
on the one hand, the system becomes highly relativistic if Up > me c2 . The intensity
necessary to reach this limit decreases according to (8.139) with the square of the
wavelength. On the other hand, for intensities above IH the electric field in the laser
focus is larger than the atomic field EH that an electron experiences in the H atom
at a distance a0 from the nucleus. This intensity is independent of wavelength:
2
0 c 2 0 c e
IH = E = = 3.51 1016 W cm2 . (8.141)
2 H 2 40 a02
8.5.2 KELDISH Parameter
There are other aspects for considering a laser field to be high. One of these is
derived from the ratio of ionization potential WI to ponderomotive potential Up .
For reasons to be discussed in Sect. 8.5.4 one defines the so called

WI 0 cme 2 WI
K ELDYSH parameter = = (8.142)
2Up e2 I

WI / eV
or in numerical terms = 2.31 10 3
I /1012 W cm2 / nm2
introduced in the pioneering work of K ELDYSH (1965). This parameter character-

izes the transition from atom with laser field > 1 to a situation that one may
describe as laser field with atom < 1. For the above example, an H atom with
WI = 13.6 eV in a radiation field with I = 1014 W cm2 at = 800 nm we find
1. At this intensity the atomic energy is thus comparable with the energy im-
posed onto the electron by the laser field. For an H atom, this may be called an
intense laser field. We also emphasize at this point that the K ELDYSH parameter de-
pends on the wavelength: the longer the wavelength, the more efficient is the laser
field!
8.5.3 From Multi-photon Ionization to Saturation
Multi-photon ionization (MPI) has already been subject to our discussion in

Sect. 5.5.5. There we have used perturbation theory: up to N th order for N pho-
ton absorption. As described there, the cross section for MPI depends on the laser
Fig. 8.29 Multi-photon

ionization signal from Xe at
800 nm as a function of laser
intensity according to
L AROCHELLE et al. (1998).
The slope in the loglog
display gives with (5.92) an
indication of the number of
photons N involved in the
process. Sketched in red are
the slopes for processes with
9 and 5, respectively.
For the direct MPI of Xe at
least 9 photons are needed
observed for the lowest
intensities. At an intensity
I = 1014 W cm2 the process
is saturated. The ion yield is
compared with different
theories
intensity as I N . However, if the laser field becomes comparable to inner-atomic

fields this approach is bound to break down. Atoms and molecules often behave
quite surprisingly in high laser fields: one even finds that the processes become
more and more classical as intensities increase. For instance, at very high inten-
sities atomic energy levels are shifted substantially and electrons can escape from
the atoms by tunnelling or above-barrier processes, as we shall see in a moment.
Their kinetic energies then increase as laser intensity increases a phenomenon
in direct contrast to the canonical observations with the photoelectric effect at low
intensities.
The transition between perturbative, tunnelling, and above-barrier region is, how-
ever, seamless. A benchmark type of experiment is shown in Fig. 8.29. It has been
reported by L AROCHELLE et al. (1998), who very carefully measured the MPI yield
from Xe with femtosecond laser pulses at 800 nm. The ionization potential of Xe
is WI = 13.44 eV, and with WI / = 8.67 the number of photons required for
ionization is 9. With an ion yield I N one would, in a double logarithmic plot,
expect a slope of N = 9. As illustrated in Fig. 8.29 this is indeed the case for the
lowest intensities, while at intermediate intensities the experimental data appear to
follow a power law I 5 , as indicated in the graph. Xe is a quite complex atom with
dense series of states above the first excited state, the latter requiring ca. 5 photons
to be excited. Obviously at such high intensities the resonance condition is washed
out due to the dynamic S TARK effect, and after this first step has been reached the
subsequent ionization of this state occurs readily.
As also seen in Fig. 8.29 the rise of the ion yield above ca. 1014 W cm2 con-
tinues to decrease dramatically. One may conclude that at these intensities almost
all atoms are already ionized in the centre of the laser focus one speaks of satu-
ration. The continuing rise with intensity is essentially due to a geometrical effect:
enhanced ionization now also occurs of the peripheral zone of the laser beam (hav-
ing a Gaussian profile). Thus, the volume in which saturation intensity is reached
increases, and with it the total ion yield. The upper scale in Fig. 8.29 gives the
K ELDYSH parameter (8.142) for comparison. Saturation intensity obviously corre-
sponds to 1, that is to say saturation sets in where the moderate field becomes
a very high one.
8.5.4 Tunnelling Ionization
At very high intensities the internal atomic field will be modified substantially by the
external (oscillating) electric field. Lets assume the atomic potential to be Coulom-
bic with charge Ze. When adding a linearly polarized laser field, an atomic electron
sees a time dependent, overall potential
Ze2
V (r, t) = eE(t)z with z = r cos , (8.143)
40 r
as illustrated in Fig. 8.30 for the time of maximum field (amplitude E0 ).

A bound electron can thus tunnel out of the atom as indicated in Fig. 8.30(a)
or may even leave the atom above-barrier (b), if the latter is lowered sufficiently.
This happens at a critical intensity Icr when at the saddle point V (rs ) = WI . From
dV (r)/dr|rs = 0 one finds
2 c03
Icr = (WI )4 (8.144)
2Z 2 e6

Icr 4.0 109 WI 4
.
W cm2 Z2 eV
For an H atom (Z = 1, WI = Eh /2), the critical intensity is Ic = 1.37

1014 W cm2 which is easily reached in a focused femtosecond laser pulse.
In this picture the K ELDYSH parameter may be visualized in an alternative in-
terpretation: since the laser field oscillates, the crucial question is, whether the elec-
tron can escape the atom fast enough before the field reverses its sign. Considering
Fig. 8.30(a), one estimates the distance tu through which the electron has to tun-
nel for simplicity from a so called zero range potential (red dashed line). From
Fig. 8.30(a) one reads:
tu = WI /(eE0 ) (8.145)
(a) V(r) / E h (b) - e E0 z V(r) / E h

- eE0 z 0.5 0.5
tu
- 10 15 - 10 10
z / a0 z / a0
WI WI
e- e-
- 0.5 - 0.5
- 1.0 saddle point - 1.0
Fig. 8.30 Model to understand atomic ionization in a high electric field, in particular in an intense
laser field: (a) tunnelling (b) electron emission above-barrier. Sketched are cuts through the
potential parallel to the direction of the E field at the time of maximum field in z-direction
the electron leaves the atom its kinetic energy is Wkin6 = WI , its velocity v =
As
2WI /me , and consequently the tunnelling time becomes

tu me WI 0 cme WI
ttu = = = . (8.146)
v 2eE0 2 e2 I
In order to allow the electrons to leave that atom for good, the tunnelling time must
be distinctively smaller than one half of the oscillation period, say ttu < 1/(2). One
then defines the K ELDYSH parameter as

0 cme 2 WI
= 2ttu = , (8.147)
e2 I
in agreement with (8.142). Saturation of the ion signal, observed in Fig. 8.29 for
high laser intensities ( 1), is thus found to happen at intensities and frequencies
for which the electron has sufficient time to escape when the oscillation field reaches
its maximum value E0 . With this visualization of the ionization process it is obvious
that ionization is more probable when the field oscillates less frequently, i.e. at larger
wavelengths.
The essentially classical ADK theory (A MMOSOV et al. 1986), often used with
astonishingly good results for atoms, neglects the frequency dependence of the pro-
cess completely. It considers saturation to be reached when the field is high enough
for direct above-barrier ionization as sketched in Fig. 8.30(b). To give a typical
value: in the H atom at = 800 nm, = 0.9 when this critical field intensity (8.144)
is reached. A detailed understanding of the relevant processes is subject to current
research. Some important concepts and consequences will be discussed in the fol-
lowing.
6 One should take this with a grain of salt: Tunnelling is a quantum mechanical process, while in a
classical picture the electron can only leave the atom above-barrier.
Fig. 8.31 Non-sequential

double ionization of He by
multi-photon processes with 105
160 fs pulses at 780 nm and
comparison with different
theories according to
103
WATSON et al. (1997). The
ion signal / arb. un.

He+ (+) and He2+ () ion He+ He 2+
yield has been measured by
WALKER et al. (1994). The 101
seq
dashed red lines (seq)
represent a theory with one
eq
active electron only, the full
10-1
ns
seq
black line (nseq) is a model
calculation for non-sequential
ionization
10-3
1014 1015 1016
intensity / Wcm-2
8.5.5 Recollision
If the timing between laser field and electron ejection is favourable, the electron may
even return to the atom. This so called rescattering of electrons was first discussed
in a pioneering paper by C ORKUM (1993). In a high, oscillating electric field the
trajectory of an electron depends of course on the exact point in time when it starts.
A simple classical calculation shows that the electron returns indeed to its starting
point, if is has not yet travelled too far when the sign of the electric field is reversed.
C ORKUM found, that (at the origin) the rescattered electron can acquire a kinetic
energy of up to
(el)
3.17 Up Wkin . (8.148)
This happens for an electron ejected at time t = 0, if the phase angle of the field
E(t) cos(t + ) is 17 . The physics of these backscattered electrons is very
interesting and continues to be a hot topic in current research.
One phenomenon associated with back scattered electrons is the ejection of a
second electron, leading to the so called non-sequential double ionization. It can be
recognized by a very special behaviour of the MPI cross section as a function of
laser intensity, illustrated for He as an example in Fig. 8.31.
Generally speaking, one expects processes of the following type for a multi-
electron system A:
A + N1 A+ + e
A+ + N2 A2+ + e
...
Aq+ + Nq+1 A(q+1)+ + e .
Fig. 8.32 Visualization for V(r ) / E h e-

the generation of high
harmonics (HHG). If the
electron is emitted - e E0 r HHG
above-barrier at the right 3.17Up+WI 3.17Up
time it may be back scattered
by the inverting field with a
kinetic energy up to r <0 0 r>0
Wkin 3.17Up . This energy r / a0
+ ionization potential is then
WI
available in principle for
HHG
e-
If these processes occur consecutively one speaks of stepwise or sequential ion-

ization. But in the case of strongly correlated systems one may also consider non-
sequential ionization, i.e. the simultaneous emission of several electrons in one gen-
uine multi-electron process. Alternatively, two electrons may be ejected by a recol-
liding electron as discussed in summary
(el)
A + N1 A+ + e Wkin 3.17Up
A2+ + 2e A+ + e &,
which is also a non-sequential ionization process. Characteristic for non-sequential

ionization is the kink in the double logarithmic plot of the ion yield as a function of
intensity very clearly seen for He++ in Fig. 8.31.
8.5.6 High Harmonic Generation (HHG)
Rescattered electrons cannot only eject a second electron. They may also be recap-
tured by the ion and emit electromagnetic radiation during this process: this leads
to the generation of electromagnetic waves with frequencies that are multiples of
the original laser frequency (fundamental). This process is called high harmonic
generation (HHG) and has attracted worldwide considerable interest during the past
years.
The HHG mechanism is illustrated schematically in Fig. 8.32. The recolliding
electron has a potentially high excess energy that may be emitted during the capture
process as radiation. According to (8.148), the energy of the recolliding electron can
be as high as 3.17Up . Thus, photon energies up to HHG 3.17Up + WI may be
emitted upon capture of the electron.
This HHG process is used in current research very successfully to generate short
pulses in the soft X-ray region (XUV). An intense femtosecond laser pulse is fo-
cussed into a dense gas target (e.g. a gas jet, a gas filled cell or capillary). One
(a) (b)
photons per band width

cutoff
plateau (3.17Up+WI)/ 8
/ 107 nm-1
intensity
H31 H43
2 H37
4 H49
H55
0
26 24 22 20 18 16 14
HHG frequency wavelength / nm
Fig. 8.33 (a) Schematic HHG spectrum with plateau and cutoff at 3.17Up + WI . The frequency
distances are 2. (b) Example of an experimentally observed HHG spectrum from BALCOU et al.
(2002). 30 fs pulses at ca. 800 nm were focussed into a Ne gas jet. Different focussing conditions
(full and dotted lines) lead to quite different efficiencies
obtains the XUV radiation in forward direction. Typically it contains a broad spec-
trum of harmonics HHG = (2N + 1) of the fundamental as shown schemati-
cally in Fig. 8.33(a). For symmetry reasons, usually only odd harmonics are emit-
ted.
The scheme indicates the particularly high efficiency for low harmonics, fol-
lowed by a long plateau with frequencies at a distance of 2 up to the so
called cutoff at 3.17Up + WI , which is easily understood in view of Fig. 8.32. In
Fig. 8.33(b) gives as a typical experimental example the spectrum of Ne. As shown
in the figure one may modify the emitted output by judicious choice of focussing
conditions. This is a consequence of the highly nonlinear process. Optimization of
HHG generation for practical application is currently a hot topic in AMO research.
Special temporal and spatial pulse shaping may be used to improve the conversion
efficiency substantially. HHG is currently being used as a convenient, table top, time
resolved short pulse radiation source in the near X-ray region. It has considerable
application potential for X-ray spectroscopy. The shortest wavelengths achievable
depend on the target, on the pump laser intensity, as well as on its frequency (since
UP 2 ).
During the past decade the generation of attosecond laser pulses (1 as = 1018 s)
by superposition of several harmonics has developed very successfully (see e.g. the
reviews by K RAUSZ and I VANOV 2009; S ANSONE et al. 2011). As it turns out, the
harmonics generated are coherent; superposing them artfully (see e.g. T ZALLAS et
al. 2003) and filtering the resulting radiation suitably corresponds to interference in
a F OURIER series, leading to a sequence of pulses with an individual pulse duration
below 1 fs. As always, a new method that improves earlier techniques by one or
two orders of magnitude opens new perspectives with an unforeseeable potential in
basic research and applications. Certainly we shall witness exciting developments
of attosecond science in the years to come (as illustrated e.g. by S ANSONE et al.
2010; B OGUSLAVSKIY et al. 2012; V RAKKING and E LSAESSER 2012).
Fig. 8.34 ATI spectra of Ar

according to PAULUS et al.
(1994), obtained with 40 fs, 107
630 nm laser pulses at 3.17Up
intensities of
count rate / arb. un.

(a) 6 1013 W cm2 , 105 (d)
(b) 1.2 1014 W cm2 ,
(c) 2.4 1014 W cm2 and =0.7
(d) 4.4 1014 W cm2 (the
103
traces are vertically slightly (b) (c)
displaced for better
visibility); the black arrows (a) = 1.33 =0.94
indicate the maximum 101 = 1.88
classical back scattering 0 20 40 60 80
energy of 3.17 Up electron energy Wkin / eV
8.5.7 Above-Threshold Ionization in High Laser Fields
Before ending this chapter we return briefly to ATI processes which we have intro-
duced already in Sect. 5.5.5. The question here is, how these processes change as
the laser intensity is increasing from MPI through the tunnelling regime and to
above-barrier ionization?
As a particularly suggestive example we show in Fig. 8.34 the spectra for Ar that
were studied by PAULUS et al. (1994) with beautifully resolved ATI peaks. Argon
has an ionization potential of 15.4 eV, according to (8.147) the laser intensities
used here thus correspond to K ELDYSH parameters of (a) 1.88, (b) 1.33, (c) 0.94
and (d) 0.7 a range covering the critical transition from moderate to above-barrier
behaviour. This is reflected quite evidently in the electron spectra: while at the low-
est intensity (a) an unspectacular ATI spectrum is observed, quite comparable to that
shown in Fig. 5.13 for Xe, the higher intensities promote very pronounced structures
as a function of electron energy, that remind us of the plateau seen in HHG which we
have discussed in the last section. There, the cutoff was identified as corresponding
to the maximum energy of rescattered electrons.
Thus it appears self-evident to attribute the plateaus or beats in the ATI spectra
also to recollision: obviously recolliding electrons too may absorb further photons.
Without going into the finer details of these observations we indicate by arrows
in Fig. 8.34 the maximum kinetic energy 3.17 Up of the rescattered electrons.
Obviously, there is even more structure and one should not over-stress the simple
rescattering model for such a highly complex process. We just mention that seri-
ous quantum mechanical model calculations achieve very good agreement with the
experimental data.
It is interesting to note that ATI may also be observed when ionizing quite large
molecules by (moderately) high intensity lasers. This is exemplified in Fig. 8.35 for
C60 according to C AMPBELL et al. (2000). The ionization potential is in this case ca.
7.6 eV, much lower than for argon. The intensities are thus equivalent to those used
in Fig. 8.34 as documented by the corresponding K ELDYSH parameters . Here too
one may recognize, albeit weakly, something like a prolonged plateau for the higher
Fig. 8.35 ATI spectra of C60

according to C AMPBELL et
1.56eV (795nm) I / Wcm-2
al. (2000); the laser intensities 9.01013 ( = 0.85)
105 7.51013 ( = 0.92)
for the four traces is given in
count rate / arb. un.

the legend. The vertical, grey 4.71013 ( = 1.17)
lines spaced at a distance of 3.71013 ( = 1.31)
the photon energy (for 3.17Up
795 nm) allow identification 103
of the ATI peaks
101
0 5 10 15 20 25 30
electron energy Wkin / eV
intensities. Clearly the decrease of the electron signal beyond the 3.17 Up limit
is significantly slower for higher intensities. It is also evident that in this case the
laser intensity must not get too high as the clear ATI peaks in the electron spectrum
smear out: this large, finite system has many active electrons (there are 240 valence
electrons in C60 ). Interaction among them thermalizes the electron motion at the
highest laser intensities. Similar trends may be recognized for Ar: in that sense, C60
with its high symmetry may be seen as a kind of super atom.
Section summary
Todays short pulse lasers allow one to generate extremely high intensities I
of electromagnetic radiation. The corresponding electric field strengths can
easily surpass the inner atomic electric fields, even by orders of magnitude.
Correspondingly, in the Hamiltonian the term quadratic to the vector potential
of the field must be considered. It gives rise to the ponderomotive potential
(8.140), Up I 2 .
The K ELDYSH parameter = WI /2Up characterizes the field strengths:
it is considered to be high for 1.
Recollision of the electron ejected by the field and forced to return to the
atom by the field provides a useful concept for understanding non-sequential
ionization, HHG and ATI in high intensity fields. A recolliding electron may
acquire a kinetic energy up to 3.17 Up .
HHG by intense femtosecond laser pulses offers excellent perspectives for
time resolved spectroscopy with short X-ray pulses. It is also the basis at-
tosecond pulses.
ADK: A MMOSOV, D ELONE, and K RAINOV, (1986) theory for strong field ion-
ization (see e.g. Sect. 8.30).

ATI: Above-threshold ionization, in multi-photon ionization, if more photons are
absorbed than necessary for ionization.
esu: electrostatic units, old system of unities, equivalent to the G AUSS system for
electric quantities (see Appendix A.3).
HHG: High harmonic generation, in intense laser fields.
HV: High voltage, electric voltages typically higher than 1000 V.
LHC: Left hand circularly, polarized light, also + light.
MPI: Multi-photon ionization, ionization of atoms or molecules by simultaneous
absorption of several photons.
interaction.
RHC: Right hand circularly, polarized light, also light.
index.html.
SVE: Slowly varying envelope, approximation for electromagnetic waves (see
Sect. 1.2.1, specifically Eq. (1.38), Vol. 2).
Ti:Sapph: Titanium-sapphire laser, the workhorse of ultra fast laser science.
VUV: Vacuum ultraviolet, spectral range of electromagnetic radiation. part of the
21348 (2007).
40 nm.
References
A MMOSOV , M. V., N. B. D ELONE and V. P. K RAINOV: 1986. Tunnel ionization of complex
atoms and of atomic ions in an alternating electromagnetic field. Sov. Phys. JETP, 64, 1191
1194.
BALCOU , P. et al.: 2002. High-order-harmonic generation: towards laser-induced phase-matching
control and relativistic effects. Appl. Phys. B, 74, 509515.
B ERGER , M. J., J. H. H UBBELL, S. M. S ELTZER, J. C HANG, J. S. C OURSEY, R. S UKUMAR,
D. S. Z UCKER and K. O LSEN: 2010. XCOM: Photon cross sections database (version 1.5),
NIST. http://physics.nist.gov/xcom, accessed: 8 Jan 2014.
B OGUSLAVSKIY , A. E., A. E. B OGUSLAVSKIY, J. M IKOSCH, A. G IJSBERTSEN, M. S PANNER,
S. PATCHKOVSKII, N. G ADOR, M. J. J. V RAKKING and A. S TOLOW: 2012. The multi-
electron ionization dynamics underlying attosecond strong-field spectroscopies. Science, 335,
13361340.
B ORN , M. and E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
B OYD , R., O. H ESS, C. D ENZ and E. PASPALKALIS: 2010. Slow light. J. Opt., 12, 100301.
B OYD , R. W. and D. J. G AUTHIER: 2002. Slow and fast light. In: Progress in Optics,
vol. 43, 497530. Amsterdam: Elsevier.
B REIT , G. and I. I. R ABI: 1931. Measurement of nuclear spin. Phys. Rev., 38, 20822083.
B UCKINGHAM , A. D.: 1967. Permanent and induced molecular moments and long-range inter-
molecular forces. Adv. Chem. Phys., 12, 107.
C AMPBELL , E. E. B., K. H ANSEN, K. H OFFMANN, G. KORN, M. T CHAPLYGUINE
M. W ITTMANN and I. V. H ERTEL: 2000. From above threshold ionization to statistical elec-
tron emission: the laser pulse-duration dependence of C60 photoelectron spectra. Phys. Rev.
Lett., 84, 21282131.
C ORKUM , P. B.: 1993. Plasma perspective on strong-field multi-photon ionization. Phys. Rev.
Lett., 71, 19941997.
H ANSON , A. L.: 1986. The calculation of scattering cross-sections for polarized X-rays. Nucl.
Instrum. Methods A, 243, 583598.
H AU , L. V., S. E. H ARRIS, Z. D UTTON and C. H. B EHROOZI: 1999. Light speed reduction to
17 meters per second in an ultracold atomic gas. Nature, 397, 594598.
H ICKSTEIN , D. D. et al.: 2012. Direct visualization of laser-driven electron multiple scattering
and tunneling distance in strong-field ionization. Phys. Rev. Lett., 109, 073004.
H UBBELL , J. H., W. J. V EIGELE, E. A. B RIGGS, R. T. B ROWN, D. T. C ROMER and R. J. H OW-
ERTON : 1975. Atomic form factors, incoherent scattering functions, and photon scattering cross
sections. J. Phys. Chem. Ref. Data, 4, 471538.
K ANE , P. P., L. K ISSEL, R. H. P RATT and S. C. ROY: 1986. Elastic-scattering of gamma-rays
and X-rays by atoms. Phys. Rep., 140, 75159.
K ELDYSH , L. V.: 1965. Ionization in the field of a strong electromagnetic wave. Sov. Phys. JETP,
20, 1307.
K RAUSZ , F. and M. I VANOV: 2009. Attosecond physics. Rev. Mod. Phys., 81, 163234.
References 445
L AROCHELLE , S., A. TALEBPOUR and S. L. C HIN: 1998. Non-sequential multiple ionization of

rare gas atoms in a ti:sapphire laser field. J. Phys. B, At. Mol. Opt. Phys., 31, 12011214.
L ORENTZ , H. A. and P. Z EEMAN: 1902. The N OBEL prize in physics: in recognition of the
extraordinary service they rendered by their researches into the influence of magnetism upon
radiation phenomena, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1902/.
M ENENDEZ , J. M., I. M ARTIN and A. M. V ELASCO: 2005. The stark effect in atomic Rydberg
states through a quantum defect approach. Int. J. Quant. Chem., 102, 956960.
PAULUS , G. G., W. N ICKLICH, H. L. X U, P. L AMBROPOULOS and H. WALTHER: 1994. Plateau
in above-threshold ionization spectra. Phys. Rev. Lett., 72, 28512854.
P OLYANSKIY , M.: 2012. RefractiveIndex.Info, MediaWiki. http://refractiveindex.info, accessed:
10 Jan 2014.
S ANSONE , G., L. P OLETTO and M. N ISOLI: 2011. High-energy attosecond light sources. Nat.
Photonics, 5, 656664.
S ANSONE , G. et al.: 2010. Electron localization following attosecond molecular photoionization.
Nature, 465, 763767.
S TARK , J.: 1919. The N OBEL prize in physics: for his discovery of the Doppler effect in canal
rays and the splitting of spectral lines in electric fields, Stockholm. http://nobelprize.org/
S VEN G ATO R EDSUN: 2004. 3j 6j 9j -symbol java calculator, Sven Gato Redsun. http://www.
T ZALLAS , P., D. C HARALAMBIDIS, N. A. PAPADOGIANNIS, K. W ITTE and G. D. T SAKIRIS:
2003. Direct observation of attosecond light bunching. Nature, 426, 267271.
V RAKKING , M. J. J. and T. E LSAESSER: 2012. X-ray photonics: X-rays inspire electron movies.
Nat. Photonics, 6, 645647.
WALKER , B., B. S HEEHY, L. F. D IMAURO, P. AGOSTINI, K. J. S CHAFER and K. C. K ULAN -
DER : 1994. Precision-measurement of strong-field double-ionization of helium. Phys. Rev.
Lett., 73, 12271230.
WATSON , J. B., A. S ANPERA, D. G. L APPAS, P. L. K NIGHT and K. B URNETT: 1997. Nonse-
quential double ionization of helium. Phys. Rev. Lett., 78, 18841887.
Z IMMERMAN , M. L., M. G. L ITTMAN, M. M. K ASH and D. K LEPPNER: 1979. Stark structure
of the Rydberg states of alkali-metal atoms. Phys. Rev. A, 20, 22512275.
Hyperne Structure
9
We shall now discuss a further step of refinement in our

understanding of atomic spectra. Hyperfine structure (HFS)
arises from the interaction of the electrons in the atomic shell
with the atomic nucleus. Aside from its general spectroscopic
relevance in atomic, molecular and nuclear physics in recent
years also in ultra-cold boson and fermion gases HFS is the
basis for NMR and in detail also for EPR. Both are today
among the most important methods for structural analysis in
molecular physics, chemistry, biology, medicine and materials
research.
Overview
This is not a particularly easy chapter to study. Nevertheless, the reader
will earlier or later have to deal with this important and methodolog-
ically fundamental topic. After an introduction into the underlying interac-
tions in Sects. 9.1 and 9.2 we shall discuss in Sect. 9.3 the Z EEMAN effect,
again for low, high and arbitrary magnetic fields, following the discussion in
Sects. 8.1.28.1.6 there applied to the electron spin, here to the nuclear spin.
Electric interactions (Sect. 9.4) and isotope shift (Sect. 9.4.2) are specialties in
the interaction of atomic nuclei with atomic electron and require some math-
ematical effort making use of formulas derived in the appendices. Finally,
Sect. 9.5 introduces some state-of-the-art experimental techniques: molecular
beam spectroscopy, EPR and finally NMR spectroscopy. All sections build
on the previous ones, but the text may be read without having mastered each
preceding detail.
9.1 Introduction
In Table 9.1 we review the perturbation hierarchy already discussed in previous

chapters, now including hyperfine interaction between the atomic electron cloud
and the atomic nucleus. The latter gives rise to a hyperfine structure (HFS) of atomic

448 9 Hyperfine Structure
Table 9.1 Hierarchy and relevance of interactions in atomic physics (ordered according to relative
strength)
Interaction Characteristic Remarks
1. C OULOMB degeneracy only for H atom and H like
Ions
2. deviation from Z/r removal of degeneracy for alkali atoms and all others
potential
3. spin-orbit fine structure splitting (2S + 1) FS levels (S electron
J =L + S for low Z spin), each (2J + 1)-fold
J = J i for high Z degenerate
4a. external field magnetic (2J + 1)-fold degeneracy
removed
4b. external field electric (2J + 1)-fold degeneracy
removed but states with
M = |M| remain degenerate
Depending on field strength 3. and 4. may change sequence
5. radiative corrections L AMB shift sensitive test for QED
6. atomic nucleus electron hyperfine structure 2I + 1 states (I nuclear spin),
cloud (in analogy to 3. but F =J +I each (2F + 1)-fold degenerate
much weaker)
a volume effects form and mass of atomic isotope shift of spectral lines
nucleus
b magnetic dipole B j field hyperfine splitting
of electron cloud
c electric quadrupole moment additional shift
E field of electron cloud
7a. external field magnetic (2F + 1)-fold degeneracy
removed
7b. external field electric analogue to 4b
6. with 7.: Important spectroscopic tool for nuclear physics, also for chemistry (atomic nuclei
probe their chemical environment, NMR)
and molecular energy levels, subject of the present chapter. The magnetic moment
of an atomic nucleus is are related to its spin I and the nuclear g factor as
I = gI N
M I / z = gI N Iz /
and has the projection M (9.1)
with the nuclear magneton

e me
N = = B (9.2)
2mp mp
= 5.051 1027 J T1 = 3.152 108 eV T1 =
7.623 MHz T1 .
9.1 Introduction 449
Note the positive sign of the nuclear magnetic moment,1 in contrast to the magnetic
moment of the electron:
S = gB
M S/ (9.3)
with the B OHR magneton
e mp
B = = N (9.4)
2me me
= 927.4 1026 J T1 = 5.788 105 eV T1 =
14.00 GHz T1 .
Precise values N and B are found in Appendix A. We emphasize that the nuclear
magneton (9.1) is very small, N B /1836, and hence HFS is small too.
As a quantitative basis for the following discussion, Table 9.2 presents a range
of typical examples for nuclear moments. We remember that protons, neutrons and
atomic nuclei are not elementary particles in contrast to the electron. Hence, the
nuclear g factors, gI = MI /I N are not even approximately integer numbers, as
documented in Table 9.2.
The usual eigenvalue equations for angular momenta hold also for the nuclear
spin. For the magnitude we have
2
I |I MI = 2 I (I + 1)|I MI , (9.5)
and 2I + 1 nuclear spin orientations are possible in respect of a given z-axis:
Iz |I MI = MI |I MI with MI = I, I + 1, . . . I. (9.6)
If no external magnetic field is present, hyperfine interaction leads to orientation

of the nuclear spin
I in respect of the total angular momentum J of the electron
charge cloud. Thus, I and J couple and form the total angular momentum of the
whole atom (electron cloud and nucleus)
=
F I +
J. (9.7)
In full analogy to FS coupling (Chap. 6) we just need to replace:
L
S
J
(9.8)

J
IF
Correspondingly, in the coupled scheme (J I )F we have:
1 Some authors use a different definition for gI : gI N (see e.g. the review of A RIMONDO
et al. 1977).
Table 9.2 Properties of some hadrons and atomic nuclei (S TONE 2005). The notation A Z X refers
to an atomic nucleus X with Z protons (atomic number) and a total of A nucleons (atomic mass
number). The unit of area, 1 b (see Appendix A.2) corresponds to the area of an average size atomic
nucleus
Nucleon or Spin Land factor Magnetic moment Quadrupole NMRb
atomic nucleus I gI = MI /(I N ) MI /N momenta
Q / eb
Proton p 1/2 5.58569471(5) 2.79284736(2) 0 +
Neutron n 1/2 3.8260854(10) 1.9130427(5) 0
Deuteron 21 D 1 0.8574382284 0.8574382284 0.0286(2)
3 He
2 1/2 4.25499544(6) 2.12749772(3) 0
4 He
2 0 0 0
6 Li
3 1 0.8220473(6) 0.8220473(6) 0.00083(8)
7 Li
3 3/2 2.1709513(13) 3.256427(2) 0.0406
12 C
6 0 0 0
13 C 1/2 +1.4048236(28) +0.7024118(14) 0 +
6
14 N
7 1 0.40376100(6) +0.40376100(6) +0.02001(10)
15 N
7 1/2 0.56637768(10) 0.28318884(5) 0 +
16 O
8 0 0 0
19 F
9 1/2 +5.257736(16) +2.628868(8) 0 +
23 Na
11 3/2 1.478348(2) +2.217522(2) +0.109(3)
29 Si
14 1/2 1.11058(6) 0.55529(3) 0 +
31 P
15 1/2 +2.2632(6) +1.13160(3) 0 +
39 K
19 3/2 0.26098(2) +0.39147(3) +0.049(4)
67 Zn
30 5/2 0.3501916(4) +0.875479(9) +0.150(15)
85 Rb
37 5/2 0.541192(4) +1.35298(10) +0.23(4)
129 Xe
54 1/2 1.555952(16) 0.777976(8) 0
133 Cs
55 7/2 0.7377214(9) +2.582025(3) 0.00371(14)
199 Hg
80 1/2 1.0117710(18) +0.5058855(9) 0
201 Hg
80 3/2 0.3734838(9) 0.5602257(14) +0.38(4)
235
92 U 7/2 0.108(10) 0.38(3) 4.936(6)
Comparison ge = |MS /(SB )| MS /N
Electron e 1/2 2.002319 . . . 1838.2819709(8)
Muon 1/2 2.002331 . . . 8.8905971(2)
Note: for I = 0 or I = 1/2 the quadrupole moment is always Q 0
aA precise definition for Q gives (9.71); see also Appendix A.2
b Isotopes marked with + are particularly useful for NMR
|J I F MF = 2 F (F + 1)|J I F MF
F with (9.9)
F = J I, J I + 1, . . . , J + I for I < J and
F = I J, I J + 1, . . . , J + I for I > J.
9.1 Introduction 451
(a) z (b) z
Fz MF
Fz 3 3
F I F
2
1 1
0
- 1 -1
J S -2
- 3 -3
L
Fig. 9.1 (a) Vector model for coupling L and S to J , and of J and I to F , (b) F has 2F + 1
possibilities of orientation in space
The vector model in Fig. 9.1(a) visualizes these relations for the example of a 2 P3/2
level: first the orbital angular momentum L (here L = 1) couples with the spin S
(here S = 1/2) forming a total angular momentum J (here J = 3/2) of the electron
charge cloud. Orbital angular momentum and spin precess around J , which finally
couples with the nuclear spin I (here I = 3/2) forming a total angular momentum
F (here F = 3) of the atom. J and I precess around F , whose absolute value is

|F | = F (F + 1) (here = 3.46). In turn, F has 2F + 1 possible orientations in
space, MF = F, F + 1, . . . F , as illustrated in Fig. 9.1(b).
Including spin-orbit (LS) and hyperfine interaction terms (J I ), as well as (if
appropriate) magnetic interactions, we write the H AMILTON operator:
=H
H LS + V
0 + V MD + V
B + V
vol + V
Q . (9.10)
In the following we consider, one by one, the different contributions to HFS: the
magnetic dipole interaction (V MD ), the HFS-Z EEMAN effect (V
B ), volume shifts

(Vvol ) and finally the electric quadrupole interaction (VQ ).
Section summary
Hyperfine structure is caused by the interaction between electron cloud and
atomic nucleus. The nuclear magnetic properties are summarized by the nu-
clear magneton N and the gI factor. While N is about three orders of mag-
nitude smaller than B for the electron, the g factors are of similar magnitude.
Angular momentum coupling between the nuclear spin I and the total angular
momentum J of the electron charge cloud to an overall angular momentum
F of the atom occurs in complete analogy the coupling of electron spin S and
orbital angular momentum L to
J.
In addition to magnetic dipole interaction and interaction with external fields,
HFS is also determined by volume shifts and electric quadrupole interac-
tion.
9.2 Magnetic Dipole Interaction
9.2.1 General Considerations and Examples
The main contribution to HFS is due to the nuclear analogue of spin-orbit coupling
in FS. It is treated in a similar spirit as the latter in Sect. 6.2.3. We write this magnetic
dipole hyperfine interaction between the magnetic field of the atomic electron cloud
B J and the magnetic dipole moment of the atomic nucleus MI as

I
J
V J = gI N
I B
MD = M I BJ = A 2 with (9.11)

the magnetic dipole HFS coupling constant A = gI N J . (9.12)
J is the operator for the average magnetic field of the electron cloud, projected
B
onto J . It is caused by the electron orbital motion and their spin magnetic moments.
Its relation with J is written as
J = J
B J /. (9.13)
MD may be diagonalized by coupling

In analogy to the treatment of fine structure, V
nuclear spin I and total angular momentum J of the electron cloud according to
(9.7), so that in analogy to (6.47)
1 2 2 2
I
J= F I J . (9.14)
2
The change of the eigenenergy is then derived from (9.11), and we obtain
A
WMD = F (F + 1) I (I + 1) J (J + 1) , (9.15)
2
corresponding to (6.62). This expression gives already a good phenomenological de-
scription of the empirically observed hyperfine structure splitting. The computation
of J and thus A will be subject to the next sections.
As for fine structure splitting, from (9.15) follows the L AND interval rule:
WMD = WMD (F ) WMD (F 1) = AF. (9.16)
The energetic distance WMD of two neighbouring HFS levels, F and F 1, in an

HFS multiplet is proportional to F .
We present some examples: For the ground and the first excited states of the
H atom the hyperfine structure is depicted in Fig. 9.2. With I = 1/2 (proton) we
obtain for the 1s 2 S1/2 ground state (J = 1/2) an HFS splitting into a doublet F = 0
and 1 (numerical values in the figure). The same holds for the excited 2s 2 S1/2 and
2p 2 P1/2 states. The 2p 2 P3/2 state (J = 3/2) also forms an HFS doublet, however,
with F = 1 and 2.
9.2 Magnetic Dipole Interaction 453
n =2
2701 2p 2P3/2 F =2
23.65
2p 3/2 9911.201 F =1 F =1
2s 2S1/2 excited
13 670
1057.847(9) 177.56 F =1 states
2s1/2 , 2p1/2 F =0 59.17
2p 2P1/2
243 nm F =0
2 466 061 102. 474 851(34)
n =1
2 466 061 413.187 074(34) 243 nm
43 770
1s 2S1/2 F =1 ground
1420.41 state
1s1/2 8173
F =0
BOHR DIRAC LAMB shift hyperfine structure
Fig. 9.2 Fine and hyperfine structure of the H atom (compare to Fig. 6.29). All energy splittings
are given in MHz as taken from the compilation of K RAMIDA (2010). The nuclear spin of the
proton (H+ ) is I = 1/2. The ground to excited state distance is not to scale. The scale for the
splittings in more and more magnified towards the right.
The deuteron, d = pn, has a nuclear spin I = 1. As shown in Fig. 9.3, the HFS
structure of the hydrogenic isotope deuterium (D) is thus quite different from atomic
H. The 1s 2 S1/2 ground state and the excited states with J = 1/2 (e.g. 2s 2 S1/2 and
2p 2 P1/2 ) split also into doublets, however, with F = 1/2 and 3/2. In contrast, the
2p 2 P3/2 state forms a triplet with F = {J I, J I + 1, J + I } = {1/2, 3/2, 5/2}.
Somewhat more complicated is the situation e.g. for sodium (Na), as shown in
Fig. 9.4. With a nuclear spin I = 3/2 the maximum number of levels is (2I + 1) = 4.
The 3s 2 S1/2 ground state is again an HFS doublet (since J = 1/2) with F = I
J = {1, 2}. The same holds for the lower level of the excited FS doublet 3p 2 P1/2 .
However, the 3p 2 P3/2 , J = 3/2 level forms a quartet with the components F =
{J I, J I + 1, J I + 2, J + I } = {0, 1, 2, 3}.
9.2.2 The Magnetic Field of the Electron Cloud
Atomic hyperfine structure has been studied in the past in great detail, and the rel-
evant parameters are known for most atomic systems rather well. Today HFS is
used as a very sensitive probe in molecular and condensed matter spectroscopy. For
a quantitative comparison with theory, i.e. for the computation of A, one has to
evaluate (9.13), i.e. to determine the magnetic field B J of the electron cloud at the
position of the nucleus.
n =2 2p 3/2 2p 2P3/2 F = 5/2

F = 3/2 4.55
2.70
9912.61 F = 1/2
2s 2S F = 3/2
1/2
40.92 excited
1059.28
F = 1/2 F = 3/2 states
2p 2P1/2
2s1/2 , 2p1/2 13.63
2 466 732 407.521 71(15)
F = 1/2
1s 2S1/2 F = 3/2
n =1
327.38
F = 1/2
8184 ground
1s1/2 state
BOHR DIRAC LAMB shift hyperfine structure
Fig. 9.3 Fine and hyperfine structure of the deuterium atom. The nuclear spin is I = 1. Otherwise
as in Fig. 9.2
Fig. 9.4 Hyperfine structure F=3

of the sodium atom (Na) in 3 2P3/2 62.5
the 32 S1/2 ground and 3P
F=2
32 P1/2 ,3/2 excited states. The F=1
515 521.3 36.1 F=0
level splittings shown are not
to scale, numerical values are 16.4
given in MHz, with HFS
F=2
splittings taken from 192
A RIMONDO et al. (1977), 3 2P1/2
F=1
while ground to excited state
508 333 195.7
and FS splittings are derived
from K RAMIDA et al. (2013)
3S 3 2S1/2 F=2
1771.6
F=1
screened
COULOMB hyperfine
potential fine structure structure
In addition to the magnetic field from the electron orbits (similar to the situation
for spin-orbit interaction treated in Sect. 6.2.3) we have to account for the magnetic
field from the dipole moments of the electron spins. That makes it somewhat more
complicated. We restrict the discussion to a single active electron. Fortunately, the
magnetic fields of orbit and spin are additive and we treat these contribution, B L and
B S , now one after the other.
For B L the situation is even somewhat less complicated than in the FS case:
According to classical electrodynamics the magnetic field of a rotating shell with a
magnetic moment M and a radius r is inside the shell
0 M
B= 2 .
4 r 3
With (1.150), M =M L = B L/, the field of the orbiting electron at the nu-
cleus is thus represented by the operator

L = 0 2B L =
B
e 1
3L, (9.17)
4 r 3 40 me c 2 r
with the B OHR magneton B = e/2me and 0 = 1/0 c2 . The respective contribu-
tion to the hyperfine dipole energy is
a3
N = 0 2B gI N 1 L I = gI me 0 2L I Eh .
2
LI = B
V L M (9.18)
4 r 3 2 4 mp r 3 2
We may compare this to (6.35) for the spin-orbit interaction in the FS case.2
We point out that the above derivation of (9.18) has been obtained as energy of the
nuclear dipole M N in the magnetic field B L of the electron orbit. It is interesting to
realized that the same result is obtained if one computes the energy of the electronic
dipole moment M L in the field B I of the localized nuclear dipole moment.
Let us now turn to the electron spin. Its magnetic moment M s of is associated
with a magnetic dipole field (e.g. JACKSON 1999, Eq. (5.56))
s = 0 1 (3er M
B s er M
s ), (9.19)
4 r 3
with er = r/r, being the unit vector in r direction. The resulting interaction does
not only depend on the angle between S and I and on the distance r of the electron
from the nucleus. It also depends on the angle between r and S. By inserting the
magnetic moment M s = ge B S/ of the electron spin S with ge = 2.0023, and
adding the orbital field B L (9.17), one obtains the total field B J of the electron
cloud for r > 0. The singularities at r = 0 (vanishing distance between nucleus and
electron) require special attention. We suppress here the details of the analysis and
just communicate the result for the total magnetic field caused by the electron at
arbitrary distance r:

0 B L ge 4ge
BJ = + (3er S er S) + S (r) . (9.20)
2 r 3 2r 3 3
2 The main difference is the ratio of electron to proton mass, me /mp 1/1836, determining the
order of magnitude for the HFS splittings. The electron spin S has been replaced by
I and corre-
spondingly the ge factor by gI . We also note the factor 2, since no T HOMAS factor has to be applied
presently (compensation of the coordinate transformation is not needed). And finally, a03 /r 3 has
now replaced the more complicated (r) a consequence of the localized nuclear dipole.
The overall magnetic dipole hyperfine interaction (9.11) is then:

MD = 0 B gI N 2
V
ge
I er
L + 3 (3
I I
S er S)
4 2 r 3 r

8ge
+ (r)I S . (9.21)
3
The (r) term, only relevant at the nucleus, is called F ERMI contact term. Note that
this expression is strictly valid only for effective one electron, one nucleus systems.
In the multi-electron case one has to add the fields from all individual electron or-
bits and spins, projected onto L z and Sz , respectively. In addition, for molecules
and solid state materials one has to sum over all relevant nuclei at their respective
positions.
We note in passing, that for L = 0 and apart from the F ERMI contact term, (9.21)
is completely analogue to the electric dipole-dipole interaction, where (8.86) de-
scribes the interaction of two electric dipoles at a distance r.
For vanishing angular momentum, L = 0 (a situation typical for many atoms and
most, non-radical, organic molecules), one may write (9.21)
MD =
V S A
I, (9.22)
with the so called hyperfine coupling tensor

0 ge B gI N 1 8
A= (3e r e r 1) + (r) . (9.23)
4 2 r3 3
This tensor plays a key role in the theory of all magnetic resonance methods
(EPR, NMR, etc.). It contains only operators derived from the components of the
position vector r. In order to determine the hyperfine splitting in 1st order, one just
has to determine the diagonal matrix elements of (9.22), i.e. one has to average over
all angular momenta and all position coordinates. Averaging over the latter needs
only to be done for the A tensor. As we shall see in a moment, for atoms which are
described in spherical coordinates this implies averaging 1/r 3 , quite similar to the
fine structure problem. In the case of complex molecules or solid state materials this
averaging becomes, however, much more elaborate. Apart from the F ERMI contact
term containing (r) to be treated below, one obtains an average hyperfine tensor
2 2
3x rr 3xy 3xz

0 ge B gI N
5 r 5 r 5
3yx 2 r 2
A = 3y
3yz
. (9.24)
4 2 r 5 r 5 r 5

3z rr
2 2
3zx
r5
3zy
r5
5
Averaging over the electron cloud is indicated by . . . . As evident from com-

paring with Table D.2 the elements of A are expressions of the quadrupole type
that characterize the charge distribution of the electron density around the nucleus.
It is this local environment that determines the HFS. The nuclear spin is thus a sen-
sitive probe for such charge distributions and hence an important tool for structural
analysis of complex systems.
9.2.3 Nonvanishing Orbital Angular Momenta
Let us return to atoms and focus our attention on r > 0. This is reasonable if the or-
bital angular momentum is larger than zero. Electrons with > 0 almost never come
close to the nucleus, and the terms containing (r) in (9.20) and (9.21) are obsolete
when determining the magnetic field and the HFS dipole interaction, respectively.
To compute the matrix elements in the coupled (effectively one electron) scheme
|n(( S)J I )F MF some serious angular momentum algebra has to be applied, with
6j or even 9j symbols (as sketched in Appendix B.4, the latter allow the recoupling
of schemes with four angular momenta, SJ I to F , in a similar manner as the 6j
symbols for the coupling of three angular momenta discussed in the context of fine
structure). Here we just outline the crucial steps of such an analysis.
First we derive an estimate for the average magnetic field of the electron cloud
B J by recalling the projection theorem (C.17) used already successfully in the
context of spin-orbit interaction. We apply it to the tensor operator B J ,
J M|BJ
J | J M
J M BJ q | J M = J M Jq | J M ,
J (J + 1)2
and write this symbolically for all components q in vector form

BJ J J
BJ = . (9.25)
J (J + 1)
Comparing this expression with the original ansatz (9.13), we obviously have found
an expression for

BJ J
J = . (9.26)
J (J + 1)
Inserting (9.20) for r > 0 and
J = L +S, we have to evaluate:

0 B
J = 2L + g e (3S e r e r
S) (
L +
S) . (9.27)
4 r 3 J (J + 1)2
Since p and hence r

L = r L = 0, this leads to

0 B 1 2 2
J = 2L + ge 3( S er )2
S , (9.28)
4 J (J + 1) r
2 3
where the averaging over the radial dependence and angular components can be
carried out independently. For s = 1/2 we find with (2.105), that
+3(
S er )2
2
S = 0, (9.29)
and finally obtain with |

2
L | = 2 ( + 1)
0 2 ( + 1)
J = B n |r 3 |n , (9.30)
4 J (J + 1)
and the magnetic dipole HFS coupling constant (9.12) for effective one electron
systems with > 0 becomes
0 2 ( + 1)
A= B N gI n |r 3 |n , or (9.31)
4 J (J + 1)
2 me 2 ( + 1) 1
= gI n | 3 |n in a.u.
4 mp J (J + 1) r
In hindsight, this remarkable result justifies our somewhat relaxed dealings with
averaging processes: since is a good quantum number in any coupling scheme, we
have to consider only diagonal matrix elements of the HFS dipole interaction (as
long as no electric field is active). For comparison with FS, and specifically with the
spin-orbit coupling constant a according to (6.36), see footnote 2. The additional
factor ( + 1)/J (J + 1) originates from the projection of the magnetic field B L
onto J .
Specifically, for H and H like ions we find with the expectation value 1/r 3
according to (2.130)
0 1 Z3
A= gI N B 3 , or (9.32)
J (J + 1)(2 + 1) a0 n3
me 1 Z3
= 2 gI in a.u.
mp J (J + 1)(2 + 1) n3
As we shall see in a moment, this special formula holds even for = 0.
9.2.4 The FERMI Contact Term
We return once more to the Hamiltonian (9.21), and discuss now the term with the
delta function (r), the F ERMI contact term. It plays a role only for = 0, since
according to (2.120) the radial wave function at small r is Rn (r) r . Only s
states have a finite probability |n 0 (0)|2 at the origin. As just outlined, for the one
electron case all other terms in (9.23) disappear for = 0. Hence, the contact term
remains the only contribution to HFS, and (9.21) becomes

MD = 0 ge B gI N 2 (r) S I ,
V for = 0. (9.33)
3 2
The diagonal matrix element is easily evaluated. In the |(LS)J I F MF coupling
scheme we have as usual
1

S
I /2 = F (F + 1) S(S + 1) I (I + 1) and
2

2 2
(r) = n00 (r) (r)d3 r = n00 (0) in general, (9.34)
Z3
and = specifically (9.35)
a03 n3
for H and H like ions according to (2.127). Thus, we finally obtain for effective one
electron systems the HFS splitting for = 0 in 1st order perturbation theory:
20
WMD = ge B gI N (r) I /2
S (9.36)
3
20 2
= ge B gI N n00 (0) F (F + 1) S(S + 1) I (I + 1) .
3
With ge 2 and (9.15) the magnetic dipole HFS coupling constant becomes
16 0 2 4 2 me 2
A= B gI N n00 (0) = gI n00 (0) a03 Eh . (9.37)
3 4 3 mp
Specifically for hydrogen and hydrogen like ions, |n00 (0)|2 is given by (2.127).
Interestingly, one obtains for the present case with = 0 (i.e. for J = S = 1/2)
exactly the same expression (9.32) for A already found for > 0. For precision
measurements one has, of course, to replace again a0 a0 me /m e.
9.2.5 Some Numbers
We compute a few numbers explicitly and first note:

0 N B
= 1.9082 108 Hz.
a03 h
For the H atom in its ground state (F = 0 or 1 and J (J + 1) = 3/4) we obtain with
gI 5.586 the hyperfine coupling constant A/ h 1421 MHz. With the interval
rule (9.16) this value is identical to the splitting of the HFS doublet. The HFS split-
ting in the ground state of the H atom is thus nearly one order of magnitude smaller
than the L AMB shift (6.90). Its experimental value is today one of the best known
spectroscopic quantities.3 It can be determined directly as microwave M1 transition,
and the first high precision measurement was carried out by R AMSEY and collabora-
tors (C RAMPTON et al. 1963), using a hydrogen maser. The presently recommended
best value (K ARSHENBOIM 2005) for the ground state HFS splitting of H is
HFS (1s1/2 ) = 1420.405751768(1) MHz, i.e. = c/ 21.1 cm.
3 Presently best known is the frequency of the 1s 2 S1/2 2s 2 S1/2 transition, see (6.88).
This famous 21 cm line of atomic hydrogen is very important in state-of-the-art ra-

dio astronomy. For instance, the abundance of atomic hydrogen in the universe has
been mapped out by recording this transition.
The HFS splitting in the excited states is much smaller. With (9.32) one calculates
for H in the 2s 2 S1/2 state
HFS (2s1/2 ) = HFS (1s1/2 )/8

= 178 MHz
and thus it is also about an order of magnitude smaller than the respective L AMB
shift. For the 2p 2 P1/2 and 2p 2 P3/2 states the splitting is again smaller, by a factor
3 and 7.5, respectively.
The HFS splitting of the deuterium atom in its ground state (F = 1 and 2) is sig-
nificantly smaller than for hydrogen, due to the smaller gI = 0.8574382284 factor.
Its value is also known with high precision:
HFS (1s1/2 ) = 327.38435230(25)MHz. (9.38)
In alkali atoms the HFS is comparatively large, in spite of the larger principle
quantum number n. This is attributed to the much higher effective nuclear charge
Zeff , which according to (9.32) enters as A (Zeff /n)3 . Also, the higher values of F
lead to larger splitting. According to the interval rule (9.16) by a factor of 2 for 23
11 Na
in the ground state (I = 3/2 F = 2) in comparison to 1 H (I = 1/2 F = 1).
1
On the other hand, the radius of the 3s electron in Na is larger by a factor of 7 as

compared to the 1s radius in the H atom (see Fig. 3.4 in Sect. 3.2), reducing 1/r 3
and |n00 (0)|2 correspondingly. Finally, the g factor of the Na nucleus is only 1/4
of that for the proton. The experimental value
HFS (3s1/2 ) 1772 MHz,
already mentioned above, is plausible: from (9.32) we derive with this value an
effective charge of Zeff 4.6 to be compared with Z = 11 for Na.
9.2.6 Optical Transitions Between HFS Multiplets
Selection rules and probabilities for optical (E1) dipole transition between two HFS
states |a and |b are derived in full analogy to those for FS transitions, as detailed
in Sect. 6.4. One just has to replace the quantum numbers according to the scheme
(9.8). In particular, the triangular relation (Fa Fb 1) = 1 holds as well as the selec-
tion rule for projection quantum numbers MF = 0, 1 with the corresponding po-
larization of emitted or absorbed light. As an example, Fig. 9.5 shows the optically
allowed HFS transitions for the 2 P1/2 , 2 P3/2 2 S1/2 doublets in Na(I = 3/2). Part
of an experimental spectrum has already been presented in Sect. 6.1.3.
Line intensities are expressed again by reduced matrix elements and 6j symbols
from which intensity ratios are derived in the same manner as for FS multiplets.
If one wants to determine the matrix elements quantitatively, one has to apply the
9.3 ZEEMAN Effect of Hyperfine Structure 461
Fig. 9.5 E1 transition F=3

between hyperfine 2P
3/2 F=2
components of a Na D F=1
doublet 2 P 2 S with 2P F=2 F=0
I = 3/2 (schematic, not to 1/2
F=1
scale)

F=2
2S F=1
1/2
reduction formalism twice in order to resolve the dependence on hyperfine and fine
structure coupling. The final expressions can be evaluated with the aid of 6j and
3j symbols except for the radial matrix elements that are characteristic for each
individual atom and transition. They are responsible for the overall strength of a line.
Of course, parity conservation also holds. In the case of one single active electron
this implies b = a 1.
Section summary
Magnetic hyperfine dipole interaction V MD = M I BJ is treated in the
same spirit as spin-orbit interaction for FS, with B J = J J / being the
effective magnetic field of the electron cloud. Consequently, the level splitting
is given by (9.15), in complete analogy to (6.62) for FS.
The main difference to FS interaction is the contribution of the electron spin
J . This leads to somewhat more complex expressions for the interaction
to B
energy (9.21)(9.23).
We distinguish the case of nonvanishing orbital angular momentum, in which
expectation values 1/r 3 determine the splitting, while for zero orbital angu-
lar momentum the F ERMI contact term is relevant, i.e. the electron density at
the origin. Interestingly, the final result (9.32) for HFS splitting in the case of
H and H like ions is identical.
HFS splitting is nearly an order of magnitude less than the L AMB shift.
For optical transitions between HFS multiplets standard selection rules hold,
i.e. (Fa Fb 1) = 1 and MF = 0, 1.
9.3 ZEEMAN Effect of Hyperne Structure
The problem is solved very similar as for the Z EEMAN effect of fine structure. How-
ever, some specialties and the fact that it forms the basis for NMR spectroscopy,
make a detailed discussion advisable.
9.3.1 Hyperne Hamiltonian with Magnetic Field
We have to add the interaction (6.29) between the external magnetic field B and
the magnetic moments of electron cloud M J and nucleus M I to the H AMILTON
operator (9.11) in the field free case:
MDB = V
H MD + V J B M
B = MI B J M I B

I
J Jz Iz
= A 2 + gJ B gI N B. (9.39)

In the second line, we have assumed B z as usual. The first term has just been
treated in detail. And for simplicity sufficient in many cases one often ne-
glects the third term, since the gyromagnetic ratios for electrons and nuclei differ
by three orders of magnitude, gJ B gI N , and also A gI N B for not too
high B.
Note, however, that this is not acceptable for high field NMR and EPR, using
super-conducting magnets (B some tesla) today standard methods in analytical
organic chemistry. Also, the objects of NMR are often characterized by J = 0, so
that the Z EEMAN splitting is determined exclusively by the interaction of the nu-
clear moment with the external field, which in turn is modified by the local field
anisotropy due to a molecular environment. Thus, we shall slightly different from
the usual treatment in atomic physics text books distinguish between weak, high,
and very high external field,
gJ B B A, gI N B A gJ B B, and A gI N B,
respectively. In the latter case we have to account fully for the term Iz B.
For all cases, we note that Jz and Iz do not commute with
I J . Strictly speaking,
neither F nor MF nor MS or MI are good quantum numbers.
9.3.2 Low Magnetic Fields
Nevertheless, for very low fields the coupling scheme

J + may be considered
I =F
a good 0th order approximation with states |[(LS)J I ]F MF , and the term I
J

is diagonalized as usual. Neglecting the term B Iz , the HFS Hamiltonian (9.39)
becomes in analogy to (8.6):
2 2 Jz
MDB A I J + gJ B J B = A F
H
J
2
I + gJ B B . (9.40)
2 2 2
The squared angular momenta have the well known eigenvalues 2 F (F + 1), etc.,
and gJ B B Jz / is treated as a perturbation.
In spite of the formal similarity with fine structure splitting in a magnetic field,
there is a significant quantitative difference. To see this, we rewrite quite formally
MD term in (9.39) in two ways:
the V

I J
I
A = g I N B J or as (9.41a)
2

I J
J
A 2 = gJ B B I . (9.41b)

In our above derivation we used view point (a): the nuclear spin I orients itself in the
magnetic field B J of the electron cloud. The alternative view point (b) considers the
splitting of the substates |J MJ of the electron cloud in the field B I of the nuclear
moment.
(a) From (9.41a) we get an estimate for the order of magnitude of the field BJ of the
electron cloud at the nucleus. Specifically, for the H atom in its 1s 2 S1/2 ground
state with A = 5.87 106 eV and gJ 2 we obtain:
BJ A/(gI N ) 33.3 T. (9.42)
We recall: a soft iron magnet at saturation can typically generate a 2 T field,

simple super-conducting magnets several T, and only with state-of-the-art high
field magnets 30 T and more may be reached. Hence, the magnetic field of the
charge cloud in an atom is by orders of magnitude larger than magnetic fields
that can be generated conveniently in the laboratory.
(b) Alternatively, from (9.41b) we estimate the average magnetic field of the nu-
cleus at the position of the electron(s), and obtain a direct comparison between
the first and second terms in (9.40):
BI A/(gJ B ) 0.05 T. (9.43)
Thus, the transition from the low (B BI ) to high field case (B BI ) in HFS
Zeeman splitting occurs at magnetic fields that can be achieved conveniently in
the laboratory in contrast to the PASCHEN -BACK effect for FS.
For low magnetic fields in this sense, B BI according to (9.43), J , I , F and MF

are approximately good quantum numbers, and, in analogy to FS, the term B in
(9.40) can be treated as perturbation in the coupled scheme |J I F MF :
A
WMD = F (F + 1) J (J + 1) I (I + 1)
2
Jz
+ gJ B J I F MF | |J I F MF B. (9.44)

Evaluating the diagonal matrix element of Jz / in analogy to (8.10) gives
Jz F MF | |F MF
J F
gJ J I F MF | |J I F MF = gJ MF = g F MF (9.45)
F (F + 1)2
(a) z B (b) z B
Fz F Jz
I BJz J
J
S
L Iz
I
Fig. 9.6 (a) Vector model for the HFS Z EEMAN effect in a low magnetic B field, (b) Vector model
for the HFS Z EEMAN effect in a high B field
with the HFS g factor
F (F + 1) + J (J + 1) I (I + 1)
gF = gJ . (9.46)
2F (F + 1)
As shown in Fig. 9.6(a), one may visualize this again by a vector diagram. The
magnetic moment M J = gJ B J / of the electron cloud is averaged due to the
precession of J and
I around F . The contribution of M I may be neglected here

in contrast to MS when treating the fine structure splitting: hence the small but
important difference between the g factors for FS according to (8.11) and for HFS
according to (9.46). The hyperfine splitting (9.44) in a low, external magnetic field
B finally becomes:
A
WMDB = F (F + 1) J (J + 1) I (I + 1) + gF B BMF . (9.47)
2
As a typical example Fig. 9.7 shows the HFS splitting of the Na D lines.
9.3.3 High and Very High Magnetic Fields
Let us now consider the opposite case: high external fields B BI , for which the
coupling of I and is broken. However, the field is assumed to be not yet high
J to F
enough for breaking the spin-orbit coupling (LS)J . This implies B a/B , a be-
ing the spin-orbit coupling parameter (6.36). We thus use the H AMILTON operator
in its original form (9.39), and have to evaluate it somewhat differently: strongest is
now the interaction of the electron clouds magnetic moment MJ with the external
field B. As visualized by the vector model in Fig. 9.6(b), this leads to a fast preces-
sion of J around B, which is assumed parallel to the z-axis, as usual. This involves
the formation of 2J + 1 (anomalous) Z EEMAN sublevels. At not too high external
magnetic fields B < BJ the next smaller interaction is that between the magnetic
WHFS MF = 3
F= 2
1
3
0 0.94 MHz / 10 - 4 T
-1 MF
-2 F= 2
3 2P
3/2
62.5 MHz 2 -3 1
2 0
2 1
0 3 2P1/2 -1
-1 192 MHz -2
36.1 MHz -2
1 0.23 MHz / 10 - 4 T
1
0 1
16.4 MHz
nm
-1 1
589
0
0
9.6
0 -1
.0 n
58
m
F= MF = 2
2 1
0 0.7 MHz / 10-4 T
3 2S1/2 -1
-2
1772 MHz
1 1
0
-1
Fig. 9.7 Hyperfine structure of the Na D lines with splitting in a low magnetic field for the 3 2 S1/2
ground and the 3 2 P1/2 and 3 2 P3/2 excited states (schematic, not to scale)
moment MI of the nucleus and the field B J = J J / of the electron cloud. We

recall that BJ is on the order of 30 T. One may view B J as consisting of a static
component B J z and one component which rotates in the xy plane. Since the pre-
cession of J and hence of the B J field of the electron cloud is much faster than
that of the weakly interacting I , the components B J x and B J y average out. Thus,
I precesses in the B J z field, i.e. also around the z-axis.
Using the definition (9.11), (9.12) for the HFS coupling constant (A = gI N J )
we thus rewrite the H AMILTON operator (9.39) with (B.8) in the spherical basis:

MDB = A Jz Iz J+ I J I+ + gJ B Jz gI N Iz B.
H (9.48)
2
The appropriate basis states are now of course |J MJ I MI for uncoupled J and I .
In 1st order perturbation theory the HFS energies are given by the diagonal matrix
elements of HMDB which in this basis are
WMDB = AMJ MI + (gJ B MJ gI N MI )B, (9.49)
since the terms J+ I and J I+ have, according to (B.17), no diagonal compo-
nents. As we have seen in the last section, they cannot be neglected for small fields.
However, for high B fields their contribution averages out (the vanishing diago-
nal matrix elements are the mathematical equivalent for the fast rotation in the xy
plane); hence, the term AMJ MI describes the remaining interaction of the electron
3/2
WHFS
MJ + 3/2 1/2 MI
- 1/2
- 3/2
2B B 3/2
F MF + 1/2 1/2
- 1/2
3
2P 3 -3 2B B/3 - 3/2
3/2 3 A
2
2 -2 2B B/3
I = 3/2 2 A
1 1 -3/2
A 0 -1 -1/2
0
-1/2 1/2
2B B 3/2
- 3/2
- 3/2 - 1/2
1/2
zero field low field high field 3/2
Fig. 9.8 Hyperfine splitting in low and high magnetic field, schematic for a 2 P3/2 state with
I = 3/2 (e.g. Na)
cloud with the nuclear spin. We recognize (8.12) in the term gJ B BMJ , which re-
flect spin-orbit with the external magnetic field B, while the nuclear spin interaction
with B is given by gI N MI B relevant only for very high external fields.
Figure 9.8 shows schematically the transition from low to (medium) high
field gI N B A gJ B B for the already rather complex example of the Na
32 P3/2 state. According to (8.11) we have gJ = 4/3, from Table 9.2 I = 3/2 and
gI = 1.478, and the experimental value of the HFS constant is A 20 MHz. For
B 1.8 T this leads to gI N B = A, while for B 103 T one finds gJ B B = A.
The right part in Fig. 9.8, where we have neglected gI N MI B completely, is thus
valid for B = 102 to 101 T. Each FS term is spit here into (2I + 1) = 4 sub-
levels. The distances between the terms are 3A/2 and A/2, for |MJ | = 3/2 and 1/2,
respectively. For the transition from low to high field the non-crossing rule holds
just as for FS splitting in an external magnetic field: States with equal MJ + MI
do not cross (corresponding to ML + MS in Sect. 8.1.5). This leads to the transitions
from MF MJ + MI as illustrated schematically in Fig. 9.8.
For yet higher fields (above a few tesla), today accessible with superconduction
magnets, one has to account also for the term gI N MI B. However, even the total
electronic angular momentum J is then no longer a good quantum number, since
at B 10 T the spin-orbit coupling constant becomes a = gJ B B. This makes
the whole evaluation again more complicated. One obtains a picture similar to the
PASCHEN -BACK effect in Fig. 8.8 on the right. However, each of the 6 sublevels
shown there is now split again into 4 sub-sublevels. In each case the highest belongs
to MI = 3/2.
9.3.4 Arbitrary Fields, BREIT-RABI Formula
As the splitting patterns in the case of high fields tends to become rather complex,
we shall now focus on two significantly simpler model cases, and follow their term
energies as a function of the magnetic field in detail. We have already noted that the
magnetic fields needed to decouple I and J are not particularly high. The general
case, where the field may neither be considered low nor high, is thus most often
encountered when studying or applying HFS in contrast to FS. One has to diago-
nalize the full Hamiltonian (9.39) in analogy to the procedure for FS described in
Sect. 8.1.5. And one may simultaneously also solve the case of very high fields. As
a basis, either |(J I )F MF or |J MJ I MI states may be used.
Using the |(J I )F MF basis, I J is diagonal, and one has to evaluate the

matrix elements of Jz and Iz . This is done in a manner quite similar to that in
Sect. 8.1.5. We sketch this here for the most simple case, nuclear spin I = 1/2 and
charge cloud angular momentum J = 1/2 with gJ 2. This is e.g. the case for the
ground state of atomic H with = 0 and J = S = 1/2.
As no complications from coupling electron spin and orbit arise, the splitting
pattern becomes rather clear. The matrix elements Jz / obtained from (C.66) and
(C.67) are summarized as follows:

F MF 11 1 0 1 1 00

F MF
J I F MF |Jz |J I F MF
12 0
= 1 1 0 0 .
1 0 0 0 0 1
2
1 1 0 0 1
0
2
1
0 0 0 2 0 0
Since in this case J and I have the same value, the corresponding matrix for Iz
is almost identical: only the off-diagonal matrix elements have inverse signs cor-
responding to (C.46) and (C.47). The field free term in (9.40) is diagonal in the
(J I )F MF coupling scheme:
MD |J I F MF
J I F MF |V A
= MF MF F (F + 1) I (I + 1) J (J + 1) .
2
For J = I = 1/2 we have F = 1 and 0, and this becomes A/4 and 3A/4, respec-
tively. With the abbreviation

gJ gI
= B N , (9.50)
2 gJ
the full Hamiltonian (9.40) becomes


F MF 11 10 1 1 00

1 1 A/4 B 0 0 0
MDB
H =1 0 0 A/4 0 + B .

1 1 0 0 A/4 + B 0
0 0 0 + B 0 3A/4
Obviously only states with MF = 0 mix. Thus, we have to diagonalize

HFS W )| = 0,
(H
by solving the secular equation

HFS W ) = 0.
det(H
The solutions for MF = 1 (with MJ = MI = 1/2 or MJ = MI = 1/2, respec-

tively) are:
A
B.
W1 = (9.51)
4
For the mixed terms MF = 0 (MJ = 1/2 and MI = 1/2) one finds:

A A
W0 = 1 + (2+ B/A)2 . (9.52)
4 2
One verifies by insertion, that for low and intermediate fields the thus found four
eigenenergies may be written in the form of the more general B REIT-R ABI formula
(1931), valid for J = 1/2:4

A A 2MF gJ B B gJ B B 2
W = 1+ + . (9.53)
2(2I + 1) 2 I + 1/2 A A
We recognize the similarity with the splitting of the fine structure in a magnetic field
for a J = 3/2, 1/2 doublet according to (8.29)(8.32).
From (9.51) and expansion of the root in (9.52) for very low fields B B A,
one finds
WHFS WMD + gF B BMF
B0
with WMD given by (9.15), while gI N B is again neglected. We thus have retrieved
the low field expression (9.47) with gF = gJ /2 according to (9.46) for I = J =
1/2. In the opposite limit, B A/B , expansion of the root in (9.53) recovers
(9.49).
Figure 9.9 illustrates (9.51) and (9.52) as a function of the external magnetic field
B (with gI N B being neglected). In analogy to fine structure splitting in a magnetic
field, we see the linear behaviour for very small fields, and repulsion of the terms
4 Note that the in front of the root has to be applied judiciously: only (2J + 1)(2I + 1) different
energies correspond to eigenstates.

WHFS / A MJ MI A
+1/2 +1/2 MJ MI
2
} A /4
} - A /4
J = 1/2 MF = 1 1/2 -1/2
1
I = 1/2 2 BB
F=1 0
A /4 {
- 3A /4
F=0
{ -1
1 2 B / AB-1
-1
-1/2 -1/2
0 } A /4
} - A /4
-2
-1/2 +1/2
Fig. 9.9 Z EEMAN effect for HFS: transition from low to high magnetic field for I = J = 1/2. Left:
energies according to the B REIT-R ABI formula (9.53). Red and black, full lines refer to MI = +1/2
and 1/2, respectively. Dashed red lines show the energies one would expect without HFS. Right
(not to scale): extrapolation to (moderately) high B fields according to (9.49) however, without
the small gI N MI B term
WHFS / A MJ MI
+1/2 -1/2
NMR } gI N B - A / 2
+1/2 +1/2
164
B / AB-1 EPR
164
-1/2 -1/2
NMR } gI B + A / 2
N
-1/2 +1/2
Fig. 9.10 Z EEMAN effect for HFS at very high B fields. Left: extension of Fig. 9.9 to the (very)
high field situation for the example of the H atom according to (9.51)(9.52). For sufficiently high
field strength B (9.49) holds, now including the gI N MI B term (not to scale). Right: limiting
case A gI N B. Shown are also the potential M1 transitions in the microwave (EPR) and radio
frequency range (NMR)
with equal total projection quantum number, here MF = 0, in the transition to high
fields.
While the term gI N B is indeed negligible for low and moderately strong fields,
it becomes important for extremely high fields today exploited e.g. in EPR spec-
troscopy. This transition from moderately high to very high field is illustrated in
Fig. 9.10. For the limit of very high fields the splitting is indicated on the right
(not to scale): the sign of the nuclear spin orientation for the level with highest
energy has changed as compared to lower magnetic fields! According to (9.49),
Fig. 9.11 Hyperfine splitting WHFS / A MJ MI

of 6 Li a fermion with I = 1 1/2 1
and half integer J in its 4 1/2 0
electronic ground state F MF 1/2 - 1
2s 2 S1/2 for low and 3/2
intermediate magnetic fields. 2
1/2
The energies are given in - 1/2
units of the hyperfine 3/2
coupling constant - 3/2 0
20 40 60 B / mT
(A h222 MHz) 1/2 1/2
- 1/2
-2
- 1/2 - 1
-4 - 1/2 0
- 1/2 1
the crossover occurs when AMJ MI gI N MI B = 0, for atomic hydrogen at

B 164A1 B 3 T.
For later discussion, we have indicated in Fig. 9.10 transitions which are rele-
vant for EPR and NMR spectroscopy, in the microwave and radio frequency region,
respectively. All of them are M1 transitions.
In all the above discussion we have chosen the coupled |J I F MF basis as a
starting point. Of course one may also carry out the diagonalization in the uncou-
pled basis |J MJ I MI and obtain the same result. In this case the second and third
term of the H AMILTON operator (9.39) are already diagonal, and one has to evaluate
the matrix elements of I
J . In Fig. 9.11 we illustrate this for the HFS of 6 Li as a
6
second example. Li has recently obtained some interest in the physics of ultracold
gases as a model fermion, since it offers the possibility to form a molecule through
F ESHBACH resonances in a magnetic field. Its HFS is still relatively uncomplicated.
The Hamiltonian, derived from (9.48) using (B.14) and (B.15), is displayed as ma-
trix in Table 9.3.
Again, one may diagonalize this Hamiltonian without problems, and finds a so-
lution described by the B REIT-R ABI formula. Figure 9.11 shows the term energies
as a function of the magnetic field applied. The ordering of the terms shown at the
right in Fig. 9.11 is again only valid for intermediate fields. For very high B fields
above some tesla the terms with positive energy shift, WHFS , invert their ordering,
similar to Fig. 9.10.
Section summary
The hyperfine Hamiltonian (9.39) in an external magnetic field B consists of
three terms:
I J interaction ( A), interaction of B with the electron spin
( Jz B) and interaction of B with the nuclear spin ( Iz B).

In weak magnetic fields, B BI , the latter may be neglected, the coupled
(I J )F MF HFS-levels split linearly with B (with BI A/(gJ B ) being the
magnetic field the electron feels from the nucleus).
9.4 Isotope Shift and Electrostatic Nuclear Interactions 471
Table 9.3 H HFS for 6 Li in the electronic ground state 2s 2 S1/2 , written in the uncoupled
|J MJ I MI basis. The nuclear spin is I = 1, the HFS coupling constant A = h 221.864(64) MHz
(WALLS et al. 2003), and with (9.50) and Table 9.2 we have + = 1.0016B , and = 1.0007B
MJ MI 1
2 1
1
2 0 12 1 1
2 1 12 0 12 1
2 + B
1 A
1 0 0 0 0 0
2
1 A

0 0 B B 2 0 0 0
2
2
12 1 0 A
2 2 A2 + B 0 0 0

2 1 A2 + + B
1 A
0 0 0 2 0
2
12 0 0 0 0 A
2 2 B B 0
12 1 0 0 0 0 0 A
2 B
At very high fields B BJ the (I J )F coupling breaks and the uncoupled

J MJ I MI levels split linearly with B (with BJ A/(gI N ) being the mag-
netic field the nucleus feels from the electron).
In between these limits typical B REIT-R ABI formulas (9.53) describe the
splitting. Levels with equal MF = MJ + MI do not cross.
9.4 Isotope Shift and Electrostatic Nuclear Interactions
Aside from magnetic dipole interactions V MD and external fields, atomic mass A
(for equal atomic number Z) and shape of the nuclei may also influence the energy
levels of atoms. The so called isotope shifts (IS) and quadrupole interaction do
not change the number of hyperfine levels, but modify their overall position and
splittings. Most of the following considerations are presented again for effective one
electron systems. They may, however, be generalized for multi-electron systems by
summing over all electron coordinates.
9.4.1 Potential Expansion
The interaction energy between an atomic electron at position r and the charge
distribution n (R) of the nucleus is read from Fig. 9.12:

Ven (r) = V (R, r)n (R)d3 R, (9.54)
e
with V (R, r) = (9.55)
40 |r R|

and the nuclear charge Ze = n (R)d3 R. (9.56)
Fig. 9.12 Atomic nucleus electron

z
and electron position with R
nuclear and electron r- r
coordinates R and r,
respectively, describing the
interaction potential
(9.54)(9.56)
n (R) R
x y
nucleus
Let us discuss the potential created by the electron close to the nucleus. Since the
extension of the nuclear charge cloud is extremely small compared to the electron
cloud, we can safely expand the potential around the origin,

e
3
V 1 2V
3

V (R, r) = + X + X X + , (9.57)
x 0

40 r 2 x x 0
=1
where x1 = x, x2 = y and x3 = z (also for X ). Thus, (9.54) becomes

e2 Z 3
V
Ven (r) = + n (R)X d3 R (9.58)
40 r x 0
=1

1 2 V
+ n (R)X X d3 R + . (9.59)
2 x x 0

With the first term in (9.58) we have recovered the C OULOMB energy of a point
charge Ze, which is already included in the unperturbed atomic Hamiltonian (and
close to the nucleus is correct also for multi-electron systems). The second term
would be proportional to the electric dipole moment of the atomic nucleus which
is zero for symmetry reasons. The first nonvanishing terms, relevant in the present
context, are the second order terms (9.59) obviously tensor components. We can
always find a coordinate system in which the axes are aligned such that this tensor
is diagonal:

1
V n (R)X X d3 R
2

1
= V n (R)X2 d3 R (9.60)
2

1 2
= V n (R)R d R + n (R) 3X R d R
2 3 2 3
(9.61)
6
= Vvol (r) + VQ (r) (9.62)

A = 234 236
233 235 238
0 1 2 W / cm-1
Fig. 9.13 Example for isotope shift (IS) of spectral lines in the VIS range, here for uranium where
HFS IS. The measured isotope shift of a spectral line at 424.44 nm according to S MITH et al.
(1951) is shown for different isotopes with nucleon number A indicated by vertical black lines.
For comparison a scaling A2/3 is shown by red vertical lines below (see explanation on p. 476)
The judiciously rewritten second line consists of two terms. The first is the so called
volume term, the second is a sum over the components of a traceless tensor, the so
called quadrupole tensor. The corresponding energy shifts are obtained in 1st order
perturbation theory by multiplying Vvol (r) and VQ (r) with the electron probability
distribution |el (r)|2 and integrating them over all r space. This will be explicated
in the next two subsections.
9.4.2 Isotope Shift
Isotope effects arise for atoms with equal nuclear charge Z and different atomic
mass numbers A. We distinguish two effects of spectroscopic relevance: the so
called mass effect and the volume shift, which in turn are composed of different
contributions. While the former is of simple kinematic nature, the volume shift iden-
tified by Vvol (r) in (9.61)(9.62) is due to the finite size of the nucleus and reflects
small deviations from a pure Z/r potential. It is responsible for the major part of IS
for atoms with large Z.
Valuable information about the atomic nuclei can be gleaned from the IS. A typ-
ical example for an isotope specific optical spectrum is shown in Fig. 9.13. Isotope
shifts for atomic nuclei of different mass and shape have been determined since
the 1950ies quite intensively, and have led to a wealth of spectroscopic data which
contributed substantially to a systematic investigation of nuclear structure (see e.g.
A NGELI 2004). Recent investigations focus on a comprehensive study of artificial
isotopes over a large range of atomic numbers and mass numbers, allowing e.g. to
determine the radii of such species. On the other hand, one performs high precision
measurements for light atoms, in particular for the isotopes of He+ and Li+ . As
already mentioned in Chap. 6, these quantities allow very sensitive tests of theory,
including relativistic and radiative contributions (QED) to atomic energies. Below
we present only a few details.
Mass Effect
As explained already in Sect. 1.8.5 and Sect. 2.6.1, for a precise calculation of
eigenenergies of states in one electron systems one has to replace the electron mass
me in the S CHRDINGER equation by the reduced mass m e = me /(1 + me /M),

with M being the nuclear mass. This allows one to separate the motion of centre
of gravity, and the problem may be treated as a one particle problem. As the mass
enters linearly into the H AMILTON operator p 2 /2me and me /M 1 one obtains, to
a very good approximation, the resulting energy change (kinematic correction) by

W me
= . (9.63)
W M
Thus, to compare the optical transition frequencies of different isotopes one does
not need to know the absolute values of the term energies. This fact is of particular
importance for the precise interpretation of the spectra of small atomic numbers.
The shift is largest of course for atomic hydrogen between 1 H and 1 D with

W 1 1 1
= 2.72 104 .
W 1836 1 2
For the LYMAN line this amounts to 22.4 cm1 , in excellent agreement with ex-
periment. A precise comparison requires of course a detailed knowledge of HFS and
L AMB shift for both nuclei, as detailed in Sect. 9.2.5 and Sect. 6.5.4 (for a review
on recent precision measurments see J ENTSCHURA et al. 2011).
Quite generally, a good 1st approximation for the mass effect is obtained by
rescaling a 0 = a0 me /m e and E h = Eh m e /me . It is largest for small atoms. How-
ever, strictly speaking, the above kinematic correction is only exact for one elec-
tron atoms. For multi-electron systems the separation of the nuclear motion is no
longer trivial and one has to consider the motion of the nucleus in respect of all
electrons (see e.g. D RAKE et al. 2005). A detailed analysis shows e.g. for the
He atom that separation of the centre of mass motion leads to an additional term
(m e /M) r 1 r 2 (in a.u.), the so called mass polarization. Its expectation value is
not trivial to compute and depends on the electronic state of the system. Fortunately,
it leads to significant contributions only for small A. Beyond the middle of the pe-
riodic system of elements one may apply (9.63) to very good approximation. (We
note in passing that the mass effect for the uranium case, displayed schematically
in Fig. 9.13, does amount to just 0.0012 cm1 between 233 U and 238 U as com-
pared to the experimental value of 2.20 cm1 , i.e. it is without any spectroscopic
relevance).
On the other hand, only for the small atoms one may hope at all to achieve a
quantitative comparison of two isotopes with high precision. Recent work concen-
trates on the isotopes of He and Li where a precision below 1 MHz can be reached
today. However, as it turns out, even for these seemingly simple atoms a theoret-
ical calculation on such level of precision becomes a formidable task. In addition
to a precise computation of mass polarization it requires inclusion of higher order
radiative corrections (QED), a complete analysis of all HFS contributions as we
have discussed them for the H atom, and finally of the volume terms to be treated
next.
Fig. 9.14 Schematic V(r) Ri Rk r

illustration of modified
C OULOMB potentials Vi (red)
and Vk (pink) due to the finite
size of the nuclei in two Vk
different isotopes i and k of
an atom with effective radii
Ri and Rk . Black dashed line Vi
indicates the corresponding
C OULOMB potential
Volume Shift
We now discuss the volume term in (9.61) which describes the interaction of the
nuclear charge distribution with the field of the electron close to the origin. It is only
relevant if the electron has a finite probability at the origin, i.e. for s states.5 We
write the expectation value of R 2 for the atomic nucleus

n (R)R 2 d3 R = Ze R 2 ,
and use the P OISSON equation for the electron charge cloud at r = 0,

el (r) e 2
V |0 V |0 = = el (0) .

0 r=0 0
Thus, the volume shift of the energy levels becomes

1 Ze2 2
Wvol = V n (R)R d R =
2 3
el (0) R 2 . (9.64)
6 60
The volume term complements the 0th order pure C OULOMB term of the inter-
action in (9.58), and contributes to the isotope shift: Due the small but finite ex-
tension of the atomic nuclei, s electrons experience close to the origin a potential
which deviates from Z/r. For larger nuclei this becomes significant. The expecta-
tion value R 2 is directly related to the charge distribution, and thus to the number
of nucleons in the atomic nucleus. It should be noted that (except for the simplest
atoms) experimental measurements can only be compared with theory as difference
Wvol between different isotopes with the same nuclear charge. However, in pre-
cision measurements even higher terms Wvol = C1 R 2 + C2 R 4 + can be
recorded (F RICKE et al. 1995).
Figure 9.14 illustrates very schematically the potentials for two different
isotopes A = i and A = k at equal Z, in and near to the nucleus resulting from this
finite extension of the nuclei. One may estimate Wvol from the difference R 2
5A relativistic treatment shows that, in addition to s1/2 electrons, a small contribution also arises
for p1/2 electrons.
Table 9.4 Isotope shift for the example of 3 He and 4 He according to M ORTON et al. (2006).
Measured and calculated (Wvol ) values and determination of the average nuclear radius Rc for
3 He, using the known value R (4 He) = 1.673(1) fm
c
Transition Measurement Wvol Rc ( 3 He)

/ MHz / MHz / fm
1 3/2 2 P0 1/2)
3 He(2 3 S 3 45394.413 (137) 42184.368 (166) 1.985 (41)
4 He(2 3 S1 2 3 P0 )
1 3/2 2 P0 1/2)
3 He(2 3 S 3 1480.573 (30) 33668.062 (30) 1.963(6)
4 He(2 3 S1 2 3 P1 )
1 3/2 2 P0 1/2)
3 He(2 3 S 3 810.599 (3) 33668.066 (3) 1.9643 (11)
He(2 S1 2 3 P2 )
4 3
3 He(2 3 S 1/2 3 3 P 1/2) 45 394.413 (137) 42 184.368 (166) 1.985 (41)

1 0
4 He(2 3 S1 3 3 P2 )
Electron nucleon scattering 1.959 (30)
Nuclear theory 1.96 (1)
of the expectation values for the squared nuclear radii. Summing (9.64) over all s
electrons in the electron charge cloud (j ) leads to
Ze2 2 2
Wvol = R j (0) .
60
The volume shift according to (9.64) should e.g. explain the isotope shift for
uranium, shown in Fig. 9.13. The nuclear radius Rc may be estimated to a rough
approximation by the liquid drop model for atomic nuclei with
Rc = RLD A1/3 , (9.65)
which according to A NGELI (2004) reproduces the whole periodic system rela-
tively well6 with RLD = 0.9542 fm (for comparison, the proton radius is Rp =
0.8775(51) fm). Hence, we expect the volume shift to be

Wvol R 2 RLD2
A2/3 . (9.66)
As documented in Fig. 9.13, the isotope shifts for the 233 U to 238 U follow this
scaling rather well.
If one wants to derive reliable estimates for the squares of the nuclear radii R 2
from a measurement of the isotope shift, the electron density at the nucleus has to
be known very precisely and one has to know of course all other parameters which
determine the measured value for a particular transition. Todays experimental ac-
curacy warrants correspondingly elaborate, non-perturbative computational efforts
for some low Z atoms. Table 9.4 illustrates the state-of-the-art. Precision measure-
ments for several 2 3 S1 2, 3 3 P transitions in 3 He and 4 He are shown as derived
6 However,including terms proportional to A1/3 and to A1 in (9.65) leads to better fits for the
experimental data.
by M ORTON et al. (2006). To extract from these data the nuclear radius Rc for the
helium isotope 3 He, one has to account for the hyperfine structure of 3 He (transi-
tions F = 3/2 F = 1/2 and F = 1/2 F = 1/2), and to use the known
value Rc for 4 He.
9.4.3 Quadrupole Interaction Energy
While for electronic s states and more generally for s1/2 and p1/2 states the
electrostatic HFS interaction between the electron and the nucleus is completely
characterized by the volume shift (9.64), quadrupole interactions originate from a
not completely spherical shape of the nucleus which may be prolate or oblate. If
the potential of the electron cloud has also a non-spherical component at the position
of the nucleus (which requires j > 1/2), a quadrupole shift of the energies arises.
(The quadrupole term disappears for electronic states with j = 0 or j = 1/2 which
have no quadruple moment).
The reformulation of (9.60) into (9.61) corresponds to writing it in irreducible
form as outlined in Appendix F. It is thus advantageous to replace (9.57) by ex-
panding the potential (9.55) in terms of reduced spherical harmonics according to
(F.2). Outside the nucleus (i.e. for r > R), the quadrupole term VQ (r) in (9.61)
may be written as a scalar product of two irreducible tensor operators of rank 2,
as defined by (C.20). Explicitly, the energy shift due to quadrupole interaction is
extracted from (F.5) and becomes in 1st order perturbation theory

(e) 1 (r) 2 (R) Ze2
WQ = 3
C 2 ZeR C 2 = U2 Q2 , (9.67)
40 r 40
where the brackets . . . indicate averaging, i.e. multiplication with the probability
densities n (R)/(Ze) |el (r)|2 and integration over electronic and nuclear coor-
(r) (R)
dinates, r and R, respectively. The rank 2 tensor operators, C2 and C2 , are the
renormalized spherical harmonics (B.29) in respect of the electronic and nuclear
coordinates. We also have defined two rank 2 tensor operators, Q2 and U2 , which
allow a concise description of the nuclear and electronic charge clouds and their
interaction. Their components are7,8
for the nuclear quadrupole tensor Q2q = R 2 C2q (, ) (9.68)

1
and for the electron cloud tensor U2m = C2m (, ). (9.69)
r3
components Q2q with their expectation values Q2q ,

7 Note that we distinguish quadrupole tensor
and the nuclear quadrupole moment Q whose definition (9.71) differs by a factor 2eZ from the
expectation value Q20 .

8 While r k C (, ) = r k 4/(2k + 1)Y (, ) = C (r) are the rescaled (regular) solid har-
kq kq kq

monics, r k1 Ckq (, ) = r k1 4/(2k + 1)Ykq (, ) = Ikq (r) are called (rescaled) irregular
solid harmonics.
Fig. 9.15 Atomic nuclei (a) z Iz (b) z Iz

with quadrupole moment:
(a) Q < 0 describes an oblate
shape, (b) Q > 0 corresponds
to a prolate shape
x y x y
The former definition is identical to (F.14) for = 2, while the latter, albeit similar to
the quadrupole tensor of the electron charge cloud, differs from it by the weighting
factor 1/r 3 instead of r 2 . Since electronic and nuclear coordinates, r and R, are
independent of each other, the averaging . . . over the respective wave functions is
done independently.
Let us first consider the nucleus: Adopting the direction of the nuclear spin as
axis of reference, and assuming rotational symmetry, the nuclear quadrupole tensor
has only one nonvanishing component:
3X32 R 2
Q20 = R 2 C20 (cos ) = (9.70)
2
Its expectation value in a well defined state |I M is Q20 = I M|Q20 |I M . In
principle, this has to be averaged over the population probabilities w(M) unless a
pure |I M state is prepared. One defines the so called

nuclear quadrupole moment Q = n (R) 3X32 R 2 d3 R (9.71)
= 2eZQ20 ,
with the normalization (9.56). Numerical values are found e.g. in Table 9.2.9
Q may be computed if one knows the structure of the nucleus, i.e. the nuclear
charge distribution n (R). However, in practice one rather uses the HFS to deter-
mine Q and to compare it with suitable nuclear structure models.
We distinguish oblate nuclear shape (pancake like), shown in Fig. 9.15(a),
which corresponds to Q < 0, and prolate (cigar like) shape with Q > 0, as illus-
trated in Fig. 9.15(b).10 The nuclear spin I may be considered responsible for such
deviations of the shape from a pure sphere (see Table 9.2). With (F.25) the matrix
9 Usually,e is not mentioned in published data. As the expectation values Q2q of the quadrupole
tensor are correctly measured in units of b, to avoid misunderstandings we use for the quadrupole
moment Q the (non-SI) unit eb (see Appendix A.2).
10 It
is important to note that this definition of prolate and oblate only holds for the quadrupole
moment (9.70), defined by the shape of the charge cloud. The sign changes when general tensor
operators are used, constructed from angular momenta (Appendix F.3.2).
elements of Q20 can be reduced to
3M 2 I (I + 1)
I M|Q20 |I M = I Q2 I , (9.72)
(2I + 3)(I + 1)I (2I 1)
with M being the projection quantum number of the nuclear spin I onto a given
laboratory axis (defined e.g. by the z-axis of an oriented atom). The maximum value
of I M|Q20 |I M is obtained for M = I . Multiplied with 2eZ it is called the spec-
troscopic (nuclear) quadrupole moment:

I (2I 1)
Qs = 2eZI I |Q20 |I I = 2Ze I Q2 I . (9.73)
(2I + 3)(I + 1)
Thus, the reduced matrix element11 of the nuclear quadrupole tensor is

2 Qs (2I + 3)(I + 1)
I Q2 I = R I C2 I = . (9.74)
2Ze I (2I 1)
With this we may rewrite (9.72) for the nuclear quadrupole moment in a state |I M :
3M 2 I (I + 1)
Q(M) = 2ZeI M|Q20 |I M = Qs .
I (2I 1)
For further details we refer to Appendix F.3.1 and classical textbooks (e.g. B LATT
and W EISSKOPF 1952).
Summarizing, the spectroscopic nuclear quadrupole moment Qs is the expecta-
tion value of 2ZeQ20 in a nuclear state |I M = I with respect to the laboratory
frame. It has to be distinguished from the intrinsic nuclear quadrupole moment Qi
with respect to the nuclear symmetry axis which is not identical to the nuclear spin.
Typically, the nuclear spin I does not point into the direction of the and the nuclear
symmetry axis. If the nucleus is assumed to be rotationally symmetric (which is of-
ten the case but not necessarily so) spectroscopic and intrinsic nuclear quadrupole
moments are related by (see e.g. N EYENS 2003)
3MK2 I (I + 1)
Qs = Qi , (9.75)
(2I + 3)(I + 1)
where MK is the projection quantum number of I onto the nuclear symmetry axis.
Even for the maximum value of MK = I we note that Qs < Qi . This is a conse-
quenceof the nuclear axis being statistically oriented on a cone around the nuclear
spin ( I (I + 1) > I ). Only in the limit of large I = MK both quantities become
equal.
11 Note that in the nuclear physics literature several types of normalization of the reduced matrix
elements may be found (see footnote 2 in Appendix C).

We emphasize again that the atomic nucleus (as well as the electron cloud) can
have a quadrupole moment only if its spin is I > 1/2 (or its total angular momentum
J > 1/2, respectively): the reduced matrix element obeys the triangular relation
(I 2I ) = 1, i.e. I 1 must hold.
(r)
Let us now turn to the atomic electron charge cloud: The tensor U2 defined
by (9.69), characterizes its electric field. Its zero component may be written
3z2 r 2 1
U20 = 5
= 3 C20 ( ), (9.76)
2r r
and its matrix elements are
2[3MJ2 J (J + 1)]
J MJ |U20 |J MJ = J U2 J . (9.77)
(2J + 3)(2J + 2)2J (2J 1)
The reduced matrix element

1 (r)
J U2 J = 3 J C2 J (9.78)
r
may be computed for one electron systems in a straight forward manner. Again,
the expectation value of 1/r 3 has to be evaluated. In the s J coupled scheme
(r)
s J C2 s J is given by (C.51) and (C.54).
9.4.4 HFS Level Splitting
We can now evaluate the overall HFS level splitting due to magnetic dipole and elec-
tric quadrupole interaction. For the quadrupole interaction energy (9.67) we have to
evaluate the matrix elements

I J F MF |Q2 U2 I J F MF
of the of scalar product U2 Q2 . Fortunately, according to (C.48), they are diagonal

in F and MF and do not depend at all on MF . Thus, with (C.49) we can rewrite
(9.67) as
Ze2
WQ = (2I + 1) (2J + 1)
40
" #
I I 2
(1)F +J +I J U2 J I Q2 I . (9.79)
J J F
This is evaluated by inserting (9.78), (9.74) and the explicit expressions for the 6j
symbol (B.72). WQ is additive to the magnetic dipole interaction (9.15) as long
as WQ WMD . Except for really heavy atomic nuclei this is usually a reasonable
assumption. Summation of the two terms gives
Fig. 9.16 Example for the W HFS /A

influence of quadrupole F =3
2
interaction on HFS as a
function of the ratio of - 0.4 - 0.2 0.2 0.4
0
quadrupole coupling constant B/A
2
B and magnetic dipole -2
hyperfine constant A; 1
I = J = 3/2 is assumed 0 -4
1
WHFS = WMD + WQ = AK
2
3
2 K(K + 1) 2I (I + 1)J (J + 1)
+B , (9.80)
2I (2I 1)2J (2J 1)
with K = F (F + 1) I (I + 1) J (J + 1).
A is the magnetic dipole hyperfine constant, given by (9.31) and (9.37) for > 0
and = 0, respectively, while for the so called quadrupole coupling constant one
obtains

e 2J 1 1 2J 1 a03 Qs
B= Qs = Eh . (9.81)
40 2J + 2 r 3 2J + 2 r 3 ea02
Sometimes (9.80) is cast into the more general form
6( J )2 + 3
I I
J 2
I
2 2
J
WHFS = AI J + B , (9.82)
2I (2I 1)2J (2J 1)
where the averaging process is again interpreted in the spirit of the vector model,
and (9.80) is recovered by inserting
I = I (I + 1),
J = J (J + 1) and I
2 2
J =
K/2.
A typical example for the influence of quadrupole interaction on HFS is sketched
in Fig. 9.16 as a function of B/A. The coupling corresponds to the first excited states
in the main isotopes of the smaller alkali atoms: for the 3P3/2 state in 23 Na one finds
B/A = 0.155, for the 4P3/2 state in 39 K, B/A = 0.472 (A RIMONDO et al. 1977).
We emphasize again that (9.80) is only valid for I > 1/2, as the quadrupole mo-
ment disappears for I = 0 and 1/2. Correspondingly, (9.78) describing the electron
cloud disappears if J = 0 or = 1/2, as well as for = 0.
Section summary
Mass, size and shape of the atomic nucleus contribute to shifts of hyperfine
energy levels.
Only for small atoms, the simple kinematic mass effect due to the finite nu-
clear mass is of spectroscopic relevance.
To identify the influence of size and shape of the nucleus one expands the
interaction potential between nucleus and electron in a multipole series and
finds that quadratic terms give the first nonvanishing contributions.
The finite size of the nucleus, decreases the overall binding energy of s elec-
trons by the so called volume shift, Wvol R 2 which is on the order of
Wvol /Wvol 104 for the largest atoms. For high Z it is the largest contri-
bution to the isotope shift observed for atoms with equal Z and different A.
The quadrupole shift reflects the non-spherical shape of atomic nuclei with
finite spin I > 1/2, represented by the nuclear quadrupole moment (9.71).
It changes the energy of individual hyperfine levels F if the electron charge
cloud has also a quadrupole moment (J > 1/2 and > 0).
The overall HFS splitting is given by (9.80), with the quadrupole and magnetic
dipole HFS coupling constants, A (9.31) or (9.37) and B (9.81), respectively.
9.5 Magnetic Resonance Spectroscopy
In addition to high resolution, D OPPLER free optical spectroscopy, as discussed

in Chap. 6, and related methods of laser induced fluorescence, a number of so-
phisticated radio frequency (RF) and microwave (MW) resonance methods have
been devised to obtain information on atomic nuclei and their interaction of with
the surrounding electron charge cloud (the pioneering experiments of R ABI 1944,
were honoured with the N OBEL prize). They play an key role in modern molecular
physics and chemical analytics. Since D OPPLER shift and broadening are propor-
tional to the absorbed frequency, one may suppress them completely when determin-
ing FS and HFS by inducing the transitions between the respective levels directly.
We have already discussed the key ideas in the context of the L AMB shift exper-
iment in Sects. 6.5.2 and 6.5.3. Here we introduce molecular beam spectroscopy
with RF and MW, EPR and finally NMR spectroscopy. A variety of other elaborate
methods, including infrared and R AMAN spectroscopy, will be treated in Chap. 5,
Vol. 2.
9.5.1 Molecular Beam Resonance Spectroscopy
R AMSEY (1989) and his collaborators have performed since 1946 ground breaking
work in microwave and radio frequency resonance spectroscopy on magnetically or
electrically prepared and guided atomic and molecular beams. Typical experimental
equipment and methods which are used still today even for quite complicated
molecules are still based essentially on the original design from the 1970ies (see
e.g. G ALLAGHER et al. 1972). Of course, detection schemes, data acquisition and
storage procedures have been improved by orders of magnitude in the mean time.
Figure 9.17 illustrates the principle of an electrically focussing molecular resonance
apparatus for studying magnetic dipole transitions (M1) in the RF and MW domain.
First, one has to create a molecular beam of the atoms or molecules to be stud-
ied (see footnote 24 in Chap. 1). They are then state selected in electric quadrupole
fields and focused onto the detector. Without going into details, we note that this
9.5 Magnetic Resonance Spectroscopy 483
A z C B
quadrupole DC and HF field quadrupole
beam detector
source
beam axis
DC mixer
power supply detection
and data
HF acquisiton
ramp for tuning synthesizer
Fig. 9.17 Scheme of a R AMSEY molecular beam resonance apparatus for the observation of M1
transitions in molecules according to G ALLAGHER et al. (1972); for details see text
procedure is essentially an electrostatic equivalent to the S TERN -G ERLACH experi-

ment (Sect. 1.9) where magnetic dipole moments are deflected in an inhomogeneous
magnetic field. A quadrupole arrangement consists of four long metal rods, mounted
symmetric and parallel to the atomic beam axis. Connected to static electric poten-
tials (+ +), these rods generate an inhomogeneous electric field, close to the
beam axis with a strength E 2 , being the radial distance from the axis. This
field deflects the atoms/molecules corresponding to their electric dipole moment
and orientation in space. It may be configured such that only one orientation of the
dipole axis in respect of the z-axis is focused towards the beam axis. This mecha-
nism is effective even if as typical with atoms the dipole moment has first to be
generated by the S TARK effect (see Sect. 8.2.8).
In a typical R AMSEY apparatus a specific dipole orientation is focused onto the
detector by the combined action of the two quadrupole fields, A and B. In the region
C in between A and B a parallel plate capacitor provides a homogeneous electric
DC and an RF field, perpendicular to the z-axis (alternatively a microwave resonator
may be placed at this position). Inducing an M1 transition in the atoms or molecules
changes the component of the dipole moment in respect of the z-axis, so that they
are no longer focused onto the detector. This leads to a corresponding reduction of
the detected signal. Instead of tuning the RF, it is usually more convenient to change
the DC field in region C, thus tuning the transition by means of the S TARK effect
into resonance with the fixed RF.
To illustrate the excellent resolution of this method, Fig. 9.18 shows the HFS
spectrum of the highly polar, diatomic molecule lithium-iodide (LiI) showing sur-
prisingly rich structure for a molecule which at first sight appears to be a relatively
simple system. We note the impressive agreement between the computed and mea-
sured spectra.
Minimizing all other experimental inaccuracies, the limits of the resolution in
such a measurement are determined by the uncertainty relation: the interaction re-
gion C has finite length , say for example = 1 m. At a typical beam velocity of
1000 m s1 the interaction time is 1 ms. The frequency resolution will then
at best be about 1 kHz. In terms of modern precision measurements this warrants im-
provement. It is, however, not trivial to increase the length of the interaction region
C since various perturbations can easily obliterate the improvement. In particular,
signal
0
41705 41707 41709
transition frequency / kHz
Fig. 9.18 Experimentally observed and fitted radio frequency spectrum (red dots and lines) of nine
S TARK split HFS transitions (F1 = 3/2 7/2) in LiI according to C EDERBERG et al. (2005). The
LiI molecules are in the ground vibrational state (v = 0), slightly rotationally excited (J = 3). The
black lines represent a computed stick spectrum
it is difficult to keep the RF or even an MW frequency or the DC field sufficiently

constant over a longer distance. Thus, R AMSEY (1950) developed his ingenious
fringe method, which effectively extends the interaction time without acquiring new
uncertainties. We have discussed this scheme already in Sect. 6.1.7 and illustrated
it for a modern atomic clock. In the setup sketched in Fig. 9.17 one simply splits
region C into two short interaction regions which may be stabilized and controlled
conveniently. In between the atoms or molecules travel freely over a time T . The
resolution is thus improved as / T .
9.5.2 EPR Spectroscopy
Electron paramagnetic resonance (EPR), also known as electron spin resonance

(ESR), is the technical term for M1 transitions induced in a quantum system with
one or more unpaired electrons in an external magnetic field. As a rule, the orbital
angular momentum is = 0. The electron spin S orients itself with MS = 1/2 par-
allel or antiparallel to the external B field. Without interaction of the electron and its
molecular environment the energy of the system depends only on this orientation of
the electron spin in respect of B. As described in Sect. 6.2.2, the interaction energy
of the electron spin without orbital angular momentum in an external field B z is
S B/ h = ge B MS B/ h 2 14 GHz 1 B/ T,
W S / h = M (9.83)
2
where holds, depending on the spin orientation being parallel or antiparallel to
the direction of the field. Transitions may be induced between both states by a
microwave field. In this case, the electromagnetic field acts with its magnetic field
component (see Appendix H.1.1) onto the magnetic moment of the spin and induces
a magnetic dipole transition (M1) as described in Sect. 5.4. The selection rules do
NMR with 1H EPR

Wp / h MHz MI = - 1/2 Ws / h GHz
750 MS = + 1/2 500
'state of the art'
X band high field EPR
500
(3cm) recent 300
development
250
2007 ff. 100
10 20 30
0 0
'state of the art'
NMR up to 950 MHz B/T - 100
- 250
MS = - 1/2 - 300
- 500
MI = +1/2
- 750 - 500
Fig. 9.19 Typical (M1) transition frequencies in NMR (left scale, black) and EPR (right scale,
red) as a function of the external magnetic field B. In EPR the electron spin is flipped, in NMR the
nuclear spin (specifically, this graph is for 1 H)
not require a change of , so that they typically occur within one n level (also pos-
sible for = 0). Of course, again the total angular momentum, including the photon
angular momentum Jph = 1 is conserved: The triangular relation (Ja 1Jb ) = 1 must
hold. The most simple case is Ja = Jb = S = 1/2, where for the spin projection we
have
MS = 1. (9.84)
The microwave frequency needed to induce a typical EPR transition is thus
WS / h = 2WS / h = 2ge B MS B/ h 28 GHz B/ T, (9.85)
corresponding to the electrons L ARMOR frequency in the field B.

Figure 9.19 displays schematically the energy splitting for the magnetic moment
of a free electron in an external B field, and compares it to nuclear magnetic reso-
nance (NMR) with protons (1 H). The latter will be discussed in Sect. 9.5.3. For both
methods Fig. 9.19 indicates typical MW and RF frequency bands, used in present
research together with the respective magnetic fields. For chemical analytics the X
band is still used as a standard method. However, with increasing complexity of the
molecules investigated, very high magnetic fields are used (high field EPR), which
can only be reached with state-of-the-art super-conducting magnets. As the split-
ting is proportional to the external magnetic B field, one expects the resolution to
increase accordingly.
EPR spectroscopy lives of the fact that the magnetic moment of the electron is
a very sensitive probe for magnetic fields in its molecular environment. In a com-
WS (b)
MW source 1 circulator 3 detector fixed frequency
B
2
irradiation reflected reflected microwave
from source signal power
(c)
MW cavity B
(a) with probe differentiated signal (- 1)
Fig. 9.20 (a) Schematic of the microwave bridge in a CW-EPR spectrometer. The microwave
enters at point 1 the so called circulator which guides it a point 2 completely onto the probe. In
contrast, the reflected signal, entering at point 2, is routed towards point 3. (b) Spectra are recorded
at fixed frequency, the magnetic field B is tuned so that the resonance energy WS changes (straight
red lines). (c) The resulting absorption signal (full black line) as a function of B; the signal is
differentiated so that a dispersion like line shape (red, dashed) emerges
frequency counter and reference arm with attenuator

power meter detector diode
MW source PSD output signal
circulator
MW cavity with probe modulation unit

modulation coils
main magnet
Fig. 9.21 Scheme of the whole CW-EPR spectrometer with the microwave bridge Fig. 9.20. The
magnetic field is provided by the main magnet and the modulation coils. The probe is positioned
in the resonator cavity and the signal is detected by a phase sensitive detector (PSD)
plex molecule the unpaired electrons interact e.g. with the nuclear spins of different
atoms, leading to characteristic spectroscopic patterns, from which one may glean
very detailed information about the molecular structure. In the standard, continu-
ous EPR method (CW-EPR) a spectrum is obtained at fixed frequency of the mi-
crowave radiation by tuning the magnetic field. A characteristic setup is shown in
Figs. 9.209.21. At each resonance, losses occur in the microwave resonator which
can be detected with high sensitivity. In contrast to optical spectroscopy, one typ-
ically observes here differentiated signals (dispersion like) which often allow for
better resolution of the structures. Figure 9.22 shows an example for an experimen-
tally obtained EPR spectrum. We cannot go into the details here.
As the resolution increases with B (requiring correspondingly higher resonance
frequencies for the same transitions), modern EPR spectroscopy uses frequencies
Fig. 9.22 CW-EPR

spectrum for the ethanol CH3 CH2 O- H
(differentiated)
anion as an example obtained
absorption
in the X band, after SDBS
(2013). The different HFS
coupling constants a for the
various atoms lead to a
pronounced structure of the
spectra 0 20 40 60 80
B / mT
Fig. 9.23 Comparison of (a)
EPR signal / arb. units

Azz (a) (b)
an X band CW-EPR spectrum gy
and (b) a pulsed high field
EPR spectrum exemplified by gz
a complex molecule
according to P RISNER et al. gx gz
(2007); note that the signal gy gx
(a) is differentiated, however, Azz
(b) is not
332 336 340 6420 6440
B / mT
up to 700 GHz and super-conducting magnets, today up to 30 T (high field EPR).

This allows one to investigate structural details even in large molecules, specifically
in radicals. This is illustrated in Fig. 9.23 for the example of a stable nitroxyl radi-
cals (TEMPO in polystyrene). The spectrum recorded with conventional CW-EPR
spectroscopy in the X band (a) and the pulsed high field EPR spectrum (b) show a
similar magnitude of hyperfine coupling Azz as expected from (9.39) (more pre-
cisely, Azz is the z-component of the HFS coupling tensor Eq. (9.22)). In contrast,
the splitting of the lines, characterized by the g factor, scales with the magnetic field
B. The gJ factor (which describes the superposition of the electron cloud and the
electron spin) becomes itself a tensor in the field of a non-isotropic molecule. While
the anisotropy of this g tensor is barely indicated in Fig. 9.23(a) it becomes evident
and clearly measurable in the high field EPR spectrum (b).
9.5.3 NMR Spectroscopy
The frequencies for nuclear magnetic resonance (NMR) transitions with protons,
1 H, are also indicated in Fig. 9.19. For the energies of the nuclear spin states |I M
I
in an external magnetic field and the transitions between them, rules hold in com-
plete analogy to those for the electron spin states. Specifically, for an uncoupled
proton spin (e.g. in an atomic 1 S0 state), the energy splits into two levels in an ex-
ternal B field with
1
Wp / h = gp N MI B/ h = 42.56 MHz B/ T,
2
RF oscillator RF receiver
receiver coil
magnet magnet signal

pole pole processing
and control
probe
tuning coils
tuning
Fig. 9.24 Simplified schematic of an NMR spectrometer. A (magnetic) RF field at fixed frequency
is applied perpendicularly to the main magnetic field B. The latter is tuned through the resonances
which induce a current in the receiver coil due to the spin flip upon RF absorption. This signal is
amplified, filtered and recorded in a PC
with gp = 5.58569471(5) being the g factor of the proton and MI = 1/2 its spin
projection quantum number (magnetic quantum number). Again, M1 transitions
may be induced by interaction of the nuclear magnetic moment with magnetic field
component of an electromagnetic wave or with a local, oscillating magnetic field.
The transition frequencies for these nuclear magnetic resonances lie in the RF range
up to several 100 MHz:
Wp / h = gp N B/ h = 42.56 MHz B/ T. (9.86)
The experimental methods of NMR are in principle similar to those for EPR
except that microwave radiation and techniques are replaced by RF. Schematically,
a typical NMR spectrometer is shown in Fig. 9.24.
NMR spectroscopy has been developed to one of the most powerful tools for
structural analysis of complex molecules, up to proteins, since the resonance fre-
quency of each observable nucleus is highly sensitive to its local environment. Char-
acteristic is a diamagnetic screening, the so called chemical shift , typically given
in ppm in respect of the standard reference material tetramethylsilane (TMS), so
that the modified NMR transition frequencies are given by:12
Wp / h = gp N (1 )B/ h. (9.87)
Characteristic features in the spectra reflect the magnetic fields of neighbouring nu-
clear moments.
Figure 9.25 shows a typical experimental 1 H-NMR spectrum for ethanol as ex-
ample. It is characterized by the influence of neighbouring protons: the N = 3
protons of the methyl radical form a quartet (Itot = N 1/2 = 3/2, degeneracy
2 Itot + 1 = N + 1 = 4; so called N + 1 rule). They change the NMR signal
12 If the electron charge cloud involves unsaturated p orbitals, the chemical shift may also be para-
magnetic.
Fig. 9.25 1 H-NMR CH3 CH2 OH CH3 sees

spectrum of ethanol recorded CH2
absorption
at 89.56 MHz in CDCl3 , CH2 sees
according to SDBS (2013) CH3
OH
4 3 2 1 0
NMR shift / ppm
of the protons in the methylene correspondingly. Conversely, the N = 2 protons

of the methylene form a triplet and determine the NMR signal of the methyl radi-
cal.
A certain problem of NMR spectroscopy is the small difference of population in
the two states with MI = 1/2. It is given by

N+1/2 N1/2 Wp Wp
= 1 exp . (9.88)
N+1/2 kB T kB T
A numerical value is ca. 6.9 106 for protons in a magnetic field of 1 T at room
temperature. This means, the relative difference in population density is less than
105 . The expected absorption signal is then
S = h(B12 N+1/2 B21 N1/2 )() (9.89)
with the E INSTEIN Bba absorption coefficient and the spectral radiation density
() (see Chap. 4). The number of absorption and emission processes is propor-
tional to the occupation probability of the states. Hence, both processes occur nearly
equally often. The net absorption rate which may be detected is thus very small. The
fact that NMR is nevertheless very sensitive is due to the large number of absorbing
molecules in the solid or liquid probe (order of magnitude 1022 1023 ). In contrast,
in atomic beam experiments one always deals with just a few particles, which may,
however, be detected with a 100 % efficiency, due to the high population differ-
ence. With (9.88) and (9.89) one immediately sees that for a good NMR signal high
magnetic fields and low temperatures and crucial.
A significant enhancement of the detection probability for NMR (as well as for
EPR) is achieved by pulsed excitation with an RF pulse of significant bandwidth.
One detects the time dependent response of the system to this excitation and av-
erages over many such measurements. The F OURIER transform of this signal gen-
erates the true NMR spectrum, providing simultaneously information on the relax-
ation properties of the system. Significant further improvements have been intro-
duced by so called 2D or even multi dimensional procedures which employ several
RF pulses of different frequencies in judiciously chosen time sequences. In this way,
one may follow the systems response to an excitation of a specific nuclear spin, and
identify correlations among different, active nuclei. Spatial anisotropy in molecu-
lar crystals may significantly blur the resolution, particularly so in solid state NMR
spectroscopy. This problem can be overcome today by the so called magic-angle-

spinning (MAS): similar to photoelectron spectroscopy (see Sect. 5.5.3) the influ-
ence of such anisotropy is described by the L EGENDRE polynomial P2 (cos ), with
being here the orientation of the crystal axis. If one spins the probe sufficiently
fast around an axis pointing into the direction of the magic angle mag = 54.736
at which P2 disappears, such anisotropy averages out and one observes sharp
lines.
Apart from proton NMR also 13 C NMR plays an important role in molecular
structure analysis, in particular so for large biomolecules. The carbon isotope 13 C
occurs in nature with an abundance of 1.1 % and may also be enriched if necessary.
It has a nuclear spin I = 1/2, however, with gI = 1.4048236(28) its gyromagnetic
ratio is only 1/4 of that for the proton, gp = 5.58. Thus, corresponding to (9.86),
the splitting of the levels is only Wp / h = gI N B = 10.71 MHz B/ T. On the one
hand, according to (9.88) this leads to a lower sensitivity. On the other hand, this
also implies a weaker interaction with the environment and brings the advantage
that in contrast to 1 H NMR sharp lines are observed even in large, complex
molecular structures, while the small splitting may be overcome by superconduc-
tive high field magnets. Thus, a rich and flexible NMR toolbox is available for
structural investigations. The combination of specific, site selective probes (isotope
labelling), multidimensional methods and MAS allows today an extremely precise
determination of the spatial structure of even quite large biomolecules in a solid
state environment.
EPR and NMR spectroscopy are today highly developed, powerful methods for
structural analysis, available in many different varieties. In addition to the methods
briefly indicated above, we have to mention sophisticated double resonance methods
with simultaneous or sequential optical transitions, as well as the combination of
electron spin and nuclear spin flips, and various time resolved and/or imaging proce-
dures. In summary, we refer to a wide range of research with a host of applications in
physics, chemistry, biology, medicine and materials sciences. Excellent monographs
and reviews provide profound introduction into and in-depth information about the
field, and a series of N OBEL prices e.g. to E RNST (1991), W THERICH (2002), and
L AUTERBUR and M ANSFIELD (2003) illustrates important steps of progress and the
far reaching significance of magnetic resonance spectroscopies.
Section summary
Magnetic resonance spectroscopy exploits M1 transitions between different
magnetic sublevels of one n level in the RF and MW region (thus avoiding
D OPPLER broadening). Today, magnetic fields up to 30 T are used. The
transition frequencies for electron spin and proton spin flip are summarized in
Fig. 9.19.
Classical molecular beam spectroscopy, still used today, employs electric or
magnetic deflection and focussing to detect atoms and molecules sensitively
in well defined magnetic substates. These are depleted in a magnetic field by
the transitions studied.
In EPR spectroscopy one studies electron spin flip transitions in the MW spec-
tral region. It is applicable to molecules with unpaired electrons in the liquid
or solid state. The g tensor of the electron acts as a highly selective probe for
structural analysis of its chemical environment, e.g. in large molecular radi-
cals.
NMR spectroscopy is today one of the most powerful methods for chemical
analysis. Mostly 1 H spin flips are used, but other nuclei such as 13 C are also
serve as probes in large molecules. The observed transition frequencies are
very sensitive to the local environment of the nuclei.

CW: Continuous wave (as opposed to pulsed) light beam, laser radiation etc.
IS: Isotope shift, of spectral lines due to different atomic number (part of the
HFS).
MAS: Magic angle spinning, fast rotation of a solid state probe in NMR to over-
come spatial inhomogenities.
usually refers to wavelengths from 1 mm to 1 m corresponding to frequencies
between 0.3 GHz and 300 GHz; ISO 21348 (2007) defines it as the wavelength
range between 1 mm to 15 mm.
interaction.

References
A NGELI , I.: 2004. A consistent set of nuclear RMS charge radii: properties of the radius surface
r(N, Z). At. Data Nucl. Data Tables, 87, 185206.
A RIMONDO , E., M. I NGUSCIO and P. V IOLINO: 1977. Experimental determination of hyperfine-
structure in alkali atoms. Rev. Mod. Phys., 49, 3175.
B LATT , J. M. and V. F. W EISSKOPF: 1952. Theoretical Nuclear Physics. New York: John Wiley
& Sons, Inc., 1979, Springer-Verlag New York edn.
B REIT , G. and I. I. R ABI: 1931. Measurement of nuclear spin. Phys. Rev., 38, 20822083.
C EDERBERG , J. et al.: 2005. An anomaly in the isotopomer shift of the hyperfine spectrum of
LiI. J. Chem. Phys., 123, 134321.
C RAMPTON , S. B., N. F. R AMSEY and D. K LEPPNER: 1963. Hyperfine separation of ground-
state atomic hydrogen. Phys. Rev. Lett., 11, 338.
D RAKE , O. W. F., W. N RTERSHUSER and Z.-C. YAM: 2005. Isotope shifts and nuclear radius
measurements for helium and lithium. Can. J. Phys., 83, 311325.
E RNST , R. R.: 1991. The N OBEL prize in chemistry: for his contributions to the development of
the methodology of high resolution nuclear magnetic resonance (NMR) spectroscopy, Stock-
holm. http://nobelprize.org/nobel_prizes/chemistry/laureates/1991/.
F RICKE , G., C. B ERNHARDT, K. H EILIG, L. A. S CHALLER, L. S CHELLENBERG, E. B. S HERA
and C. W. D EJAGER: 1995. Nuclear ground state charge radii from electromagnetic interac-
tions. At. Data Nucl. Data Tables, 60, 177285.
G ALLAGHER , T. F., R. C. H ILBORN and N. F. R AMSEY: 1972. Hyperfine spectra of 7 Li35 Cl
and 7 Li37 Cl. J. Chem. Phys., 56, 59725979.
JACKSON , J. D.: 1999. Classical Electrodynamics. New York: John Wiley & sons, 3rd. edn., 808
pages.
J ENTSCHURA , U. D., A. M ATVEEV, C. G. PARTHEY, J. A LNIS, R. P OHL, T. U DEM, N. KO -
LACHEVSKY and T. W. H NSCH : 2011. Hydrogen-deuterium isotope shift: from the 1s2s-
transition frequency to the proton-deuteron charge-radius difference. Phys. Rev. A, 83, 042505.
K ARSHENBOIM , S. G.: 2005. Precision physics of simple atoms: QED tests, nuclear structure
and fundamental constants. Phys. Rep., 422, 163.
K RAMIDA , A. E.: 2010. A critical compilation of experimental data on spectral lines and energy
levels of hydrogen, deuterium, and tritium. At. Data Nucl. Data Tables, 96, 586644.
L AUTERBUR , P. C. and S. P. M ANSFIELD: 2003. The N OBEL prize in physiology or medicine:
for their discoveries concerning magnetic resonance imaging, Stockholm. http://nobelprize.
org/nobel_prizes/medicine/laureates/2003/.
M ORTON , D. C., Q. W U and G. W. F. D RAKE: 2006. Nuclear charge radius for 3 He. Phys.
Rev. A, 73, 034502.
N EYENS , G.: 2003. Nuclear magnetic and quadrupole moments for nuclear structure research on
exotic nuclei. Rep. Prog. Phys., 66, 633689.
References 493
P RISNER , T. F., M. B ENNATI and M. M. H ERTEL: 2007. Experimental example of an EPR

spectrum in the X band and high field EPR, private communication.
R ABI , I. I.: 1944. The N OBEL prize in physics: for his resonance method for record-
ing the magnetic properties of atomic nuclei, Stockholm. http://nobelprize.org/nobel_prizes/
R AMSEY , N. F.: 1950. A molecular beam resonance method with separated oscillating fields.
Phys. Rev., 78, 695699.
R AMSEY , N. F.: 1989. The N OBEL prize in physics: for the invention of the separated oscillatory
fields method and its use in the hydrogen maser and other atomic clocks and the separated oscil-
latory fields method, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1989/.
SDBS: 2013. Spectral database for organic compounds, National Institute of Advanced Industrial
Science and Technology (AIST), Japan. http://sdbs.db.aist.go.jp, accessed: 8 Jan 2014.
S MITH , D. D., G. L. S TUKENBROEKER and J. R. M C NALLY J R .: 1951. New data on isotope
shifts in uranium spectra: U236 and U234 . Phys. Rev., 84, 383384.
S TONE , N. J.: 2005. Table of nuclear magnetic dipole and electric quadrupole moments. At. Data
Nucl. Data Tables, 90, 75176.
WALLS , J., R. A SHBY, J. J. C LARKE, B. L U and W. A. VAN W IJNGAARDEN: 2003. Measure-
ment of isotope shifts, fine and hyperfine structure splittings of the lithium D lines. Eur. Phys.
J. D, 22, 159162.
W THERICH , K.: 2002. The N OBEL prize in chemistry: for his development of nuclear
magnetic resonance spectroscopy for determining the three-dimensional structure of bio-
logical macromolecules in solution, Stockholm. http://nobelprize.org/nobel_prizes/chemistry/
laureates/2002/.
Multi-electron Atoms
10
The last chapter in this volume is dedicated to eigenstates and

energies of atoms with many electrons. For these heavy, complex
atoms, in principle all electrons have to be treated equally. The
repulsion of the electrons among each other is of the same order
of magnitude as the C OULOMB attraction of the nucleus. So,
this problem can no longer be treated by perturbation theory.
Overview
In this final chapter of Vol. 1 much comes together what has been introduced
in the preceding text. In Sects. 10.110.2 we give an overview on the classi-
cal methods for computing multi-electron wave functions, applying what we
have learned in Chap. 3 and Chap. 7. In Sect. 10.3 a short excursion into den-
sity functional theory is added. With increasing atomic number Z not only
the number of electrons grows and hence the complexity of the problem.
Also the significance of different types of interaction changes: for light atoms
exchange interaction was dominant, and LS coupling gave a good descrip-
tion as described in Chap. 6. With increasing spin-orbit interaction, LS cou-
pling is no longer adequate, as we shall illustrate in Sect. 10.4. The energy
scale as a whole changes roughly Z 2 . Transition energies for quantum
jumps within the outermost electron shells are still in the VIS or UV spectral
range; the reader may familiarize him- or herself by way of a few characteris-
tic examples with the zoo of energy levels and coupling schemes in complex
atoms. Changes in the inner shells are associated with emission or absorption
of X-ray radiation, which treated in Sect. 10.5 complementing Chap. 4. In
addition, our understanding of photoionization (Chap. 5) is deepened by some
examples of multi-electron systems. Finally, Sect. 10.6 introduces the reader
to modern sources for the generation of X-rays.

496 10 Multi-electron Atoms
10.1 Central Field Approximation
The most generally used approach for solving the multi-electron problem remains
the independent particle model as in essentially all the preceding chapters: Each
of the N electrons moves in the C OULOMB potential of the nucleus and in the av-
eraged potential of all other (N 1) electrons (also called mean field). Any direct
correlation between the electrons due to their momentary, individual coordinates is
neglected this is the key assumption. The motion of each electron thus depends
on those of the other electrons only in as far as they contribute to the average re-
pulsion potential. The character of radial wave functions (electron orbitals) in such
a screened C OULOMB potential will be very similar to those in the hydrogen atom.
This will be manifested e.g. in the asymptotic behaviour at small and large r, and in
the number of zero-crossings. Nonetheless, pure hydrogen like orbitals are typically
no longer a good 0th order approximation for larger N and Z. One thus has to solve
the S CHRDINGER equation numerically, and then to refine the independent parti-
cle model step by step. The most prominent observations may be described already
in this manner without further effort.
10.1.1 Hamiltonian for a Multi-electron System
We consider an atom with nuclear charge Z and N electrons and start again by
just including electrostatic forces. For the moment, we neglect are all magnetic in-
teractions as well as relativistic and QED effects. Following the approach taken in
Chap. 7 for the two electron system He, the total energy is obtained as the sum over
all energies of each electron: kinetic energy, C OULOMB attraction by the nucleus,
and C OULOMB repulsion from all other electrons. Extending (7.4), the H AMILTON
operator for the multi-electron system becomes:
N 2
N N
=
p j Ze2 1 e2
H + (10.1)
2me 40 rj 2 40 rj k
j =1 j k=j
with rj k = |r j r k |. (10.2)
The factor 1/2 in front of the double sum compensates the double appearance of the
repulsive terms 1/rkj and 1/rj k , thus allowing to write the equation symmetric in j
and k. In atomic units (10.1) reads
N
N N
= 1 1
H hej + (10.3)
2 |r j r k |
j =1 j k=j
1 Z
with
hej = j , (10.4)
2 rj
10.1 Central Field Approximation 497
the latter representing the unscreened single electron H AMILTON operator in the
field of the nucleus. The full S CHRDINGER equation to be solved is
(q 1 , q 2 , . . . , q N ) = W (q 1 , q 2 , . . . , q N ),
H (10.5)
q j referring to position (rj , j , j ) and spin (sj ) coordinates of electron j .
In contrast to the simple models for He in singly excited states, for multi-electron
systems in general, the repulsion terms in (10.3) can no longer be regarded as a
(small) perturbation especially when summed over all electrons. Strictly speaking,
the terms |r j r k |1 make the potential non-isotropic and different for each
electron.
In addition, we have to recall that the N electrons are indistinguishable and oc-
cupy N different sets of quantum numbers, each of which is characterized by a
specific set of quantum numbers, different from all others (PAULI exclusion princi-
ple). The total wave function must be antisymmetrized with respect to exchange of
any two electrons.
This all leads to substantial complications, and any attempt to solve the
S CHRDINGER equation (10.5) in a direct way even in numerical form will be
prohibitively elaborate. One thus has to resort to intelligent approximation strate-
gies, the two most important of which will be sketched in Sect. 10.2 and Sect. 10.3.
10.1.2 Centrally Symmetric Potential

However, we first introduce some of the underlying essential concepts ignoring for
the moment the complications arising from the requirement of antisymmetrization.
The key idea of the central field approximation is, to represent the repulsion
among electrons by an averaged, centrally symmetric potential. This centrally sym-
metric potential allows one at least in essence to separate the problem, and to
solve the S CHRDINGER equation (10.5) by a product of single particle wave func-
tions (orbitals) for each electron:
(r 1 , r 2 , . . . , r j , . . . , r N ) = 1 (r 1 )2 (r 2 ) (r j ) N (r N ). (10.6)
Basically, this is the mathematical formulation of the independent particle model.
The indices 1, 2, . . . , , . . . , N of the one-particle wave functions represent the
usual sets (7.10) of quantum numbers
= {n m ms } (10.7)
for the N occupied orbitals of the N atomic electrons. Together, they charac-
terize the total wave function1 (1, 2, . . . , , . . . , N ) and the total energy W =
W {1, 2, . . . , , . . . , N }. The individual orbitals are assumed to be orthonormal
(r) (r) = .
1 Hereand in the following the index stands for a set of quantum numbers {n m } (later also
for {n m ms }. If no ambiguities can arise, we abbreviate q j , as well as r j and sj by j as
we also have done in Chap. 7.
For electron j the centrally symmetric potential may be written as

N
1
Vj (rj ) = | | , (10.8)
|r j r k |
k=j
where | | stands for integration over all other electrons k = j . For very small
as well as for very large rj this potential will behave as discussed for the active
electron in alkali atoms (see (3.10) and Fig. 3.10).
With this potential the full Hamiltonian (10.3) may be rewritten as composed of
a centrally symmetric part
N

c = 1
H hj + Vj (rj )
e
(10.9)
2
j =1
and a (hopefully) small perturbation term which is the difference between (10.3)
and (10.9):
+N
N
,
1 1
nc = H
H H c = Vj (rj ) (10.10)
2 rj k
j k=j
Hnc contains all non-spherical parts of the interaction. Its diagonal term disappears
in 1st order perturbation theory and is neglected in the central field approximation.
However, one may considered H nc as a perturbation in higher order. Typically it
is small if one has already used reasonable orbitals for determining Vj (rj ). The
so called M LLER -P LESSET perturbation theory accounts for just this remaining
difference, by computing 2nd (MP2), 3rd (MP3) and 4th order (MP4) corrections.
10.1.3 HARTREE Equations and SCF Method
Obviously one has to find the one-particle wave functions (orbitals) 1 (r), 2 (r)
to N (r) for all N electrons ignoring for the moment the spin dependence. In
order to derive the corresponding one-particle S CHRDINGER equations, one fol-
lows a variational approach, similar to that introduced in Sect. 7.2.5. Starting from
the product ansatz (10.6) and the Hamiltonian (10.9) one minimizes the total energy
|H| of the system. We do not go into details, but the general idea outlined here
leads to plausible results.
The orbitals (r) are found as eigenfunction of N S CHRDINGER equations
e

h + V (r) (r) = Wn (r), (10.11)
differing by the repulsive potential V (r) which according to (10.8) is averaged

over all electrons, except electron . Equation (10.11) is solved by the usual type of
spatial wave functions
n m (r) = Rn (r)Y m (, ). (10.12)

Of course, the radial functions Rn (r) are here not simply hydrogenic wave func-
tions. Rather, one has to compute
Rn (r) = un (r)/r
numerically as already practiced for quasi one electron systems in Sect. 3.2.4 from
the radial S CHRDINGER equation

d2 un Z ( + 1)
+ 2 Wn + V (r) un (r) = 0. (10.13)
dr 2 r r2
So far, this all looks straight forward. However, the determination of the averaged
centrally symmetric repulsive potentials according to (10.8) is by no means trivial:
in principle one already needs to know the orbitals of all electrons for the averaging
procedure!
Assume for a moment we know the orbitals. Then, averaging over the repulsion
from all other electrons k = j implies to insert (10.6) into (10.8) from left and
right and to integrate over all other coordinates. As the orbitals are orthonormal, the
relatively simple expressions
N
N
1 2
3 | (r )|
V (r) = r
r = d r (10.14)
|r r | |r r |
= =
emerge. While refers to the quantum numbers of the orbital for electron j , the
summation has to be carried out over all other N 1 occupied orbitals, from = 1
to N .
Now, one may formally insert this expression back into (10.11) and obtains for
every occupied orbital = 1 to N :
+ N ,
)|2
3 | (r

he (r) + d r (r) = E h (r). (10.15)
|r r |
=
These coupled integro-differential equations for (r) are called H ARTREE equa-
tions. Each of them contains the solutions (r ) of all others! Its a major task to
solve them. In practice this is done iteratively, as we shall discuss in the next sec-
tion. We point out again that the potentials V (r) may be different for each electron
orbital.
After all this is done, in 1st order (H ARTREE approximation) the total wave func-
tion defined by (10.6) is obtained from the N orbitals . The corresponding
expectation value of the total energy WHartree {1, 2, . . . , N } of the system is derived
from (10.9), using (10.14) for the mean repulsive potentials:
N
N N
c | = 1
WHartree = |H + J (10.16)
2
=1 =1 =
Fig. 10.1 Scheme for finding the self-consistent field potential for a multi-electron system (SCF
method) illustrated for the H ARTREE approximation

e
with = (r)
h (r) = d3 r (r)
he (r) (10.17)
1
and J = (r) r r r (r) r (10.18)

| (r)|2 | (r )|2
= d3 rd3 r .
|r r |
Here is the energy a single electron would have in the unscreened C OULOMB
potential of the nucleus, and J the averaged repulsions between two electrons
which we have already introduced as C OULOMB integral in Sect. 7.3.3.
Note that WHartree is not equal to the sum over all orbital energies E h from
(10.15) due to the factor 1/2. Without this factor the repulsive energies would
be counted twice.
10.1.4 HARTREE Method
The search for the best set of one-particle wave functions for all N occupied or-
bitals rests upon an iteration procedure first introduced and explored by H ARTREE.
It was later on justified by S LATER, based on the variational method. It is sketched
schematically in Fig. 10.1. One first guesses a plausible 0th order approximation
for the central field potentials V (r) for all electrons of the configuration under in-
vestigation.
A possible ansatz for this potential is the T HOMAS -F ERMI model, to be dis-
cussed in the next section. With this potential one computes all electron orbitals in
the configuration, and uses these in turn to derive improved potentials V according
to (10.14). With these potentials one repeats the procedure. At the end of each in-
tegration one checks how well the wave functions obtained in the calculation agree
with those at the input. As long as the agreement is poor, one repeats the computa-
tion until it converges, i.e. until the potential reproduces itself within a preset limit.
The whole procedure is called self-consistent field method (SCF).
10.1.5 THOMAS-FERMI Potential
The classical textbook approach for the interaction potential in 0th order is the
so called T HOMAS -F ERMI model. One assumes that the (many) electrons inside
the atom follow essentially the rules for a F ERMI electron gas, as introduced in
Sect. 2.4.3.
However, in the present case the electrons are not contained in a box. Rather they
are distributed in a potential V (r) < 0 which is determined by the C OULOMB poten-
tial of the nucleus and the screening from all electrons, just as outlined in Sect. 3.2.3.
Thus, the maximum kinetic energy, F , electrons may have inside the atom is as-
sumed to be position dependent, and given by F (r) = V (r). The electron density
(2.59) becomes also position dependent:

1 2me 3/2 3/2
.
Ne (r) = V (r) (10.19)
3 2 2
On the other hand, one assumes the electrostatic potential V (r)/e to be related to
the charge density eNe (r) by the P OISSON equation:
e2
V (r) = Ne (r). (10.20)
0
We may eliminate Ne (r) and obtain the differential equation

e2 8 2m3e 3/2
V (r) = V (r) or (10.21)
40 3 3

8 2 3/2
V (r) = V (r) in atomic units, (10.22)
3
from which V (r) can be derived. Since the potential is considered spherically sym-
metric, it depends only on r and = (1/r)d2 /dr 2 . The boundary conditions at
very small and very large r are those for the screened C OULOMB potential given by
(3.10). Thus, one scales this potential as
Z
V (r) = (r/b), (10.23)
r
Fig. 10.2 T HOMAS -F ERMI 0 0.2 0.4 0.6 0.8 1.0

potential for the atomic core 0
of neon (Z = 10). The full r /a 0
red line gives -20 -1/r
(9(r/b) + 1)/r: that is the
averaged potential which a THOMAS-FERMI model
V(r) / Eh
valence electron sees. For -40
comparison also shown is the
C OULOMB potential of the -60
neon nucleus 10/r (limit
for r 0) and the -80 -Z/r
completely screened potential
1/r (limit for r )
-100
with b being a Z dependent parameter:

3 2/3 1
b= 1/3
= 0.8853Z 1/3 . (10.24)
8 2 Z
With this, and x = r/b one finally obtains from (10.22) the so called
d2 3/2
T HOMAS -F ERMI equation = . (10.25)
dx 2 x 1/2
The limiting values of (x) are 1 and 0 for x 0 and x , respectively. The
T HOMAS -F ERMI equation is treated in the literature on differential equations (see
e.g. Z WILLINGER 1997) and describes with (10.23) a universal potential which
scales with Z. Its relation to the electron density (in a.u.) is
3/2
Z
Ne (x) = for x 0. (10.26)
4b3 x
In the literature, one finds numerical data for the T HOMAS -F ERMI function (x)
as well as approximate analytical formulas, e.g. in L ATTER (1955):

(x) 1 + 0.02747x 1/2 + 1.243x 0.1486x 3/2 (10.27)
1
+ 0.2302x 2 + 0.007298x 5/2 + 0.006944x 3 .
Note that this (x) with (10.23) gives the full atomic model potential with
the nuclear charge Z, screened by Z electrons as experienced by an infinitely
small probe charge. For one of the atomic electrons one might e.g. use the poten-
tial [(Z 1)(r/b) 1]/r as a initial input into the SCF calculation of atomic
orbitals. Figure 10.2 shows this model potential for the example of Ne (Z = 10), in
comparison to the pure C OULOMB limits for very small and very large distances.
Section summary
The Hamiltonian (10.1) for a multi-electron system with nuclear charge Z and
N electrons contains in addition to a sum over all single electron Hamiltoni-
10.2 HARTREE-FOCK Method 503
ans the repulsive potentials among the N electrons which make it the overall
interaction potential non-isotropic.
A surprisingly good key approximation is the independent particle model, ac-
cording to which the total wave function may be written as product (10.6) of
all single electron orbitals. It is based on the central field approximation which
represents the electron repulsion by a centrally symmetric screening potential.
Mathematically this may be cast into a set of N coupled integro-differential
equations. Without antisymmetrization (i.e. neglecting electron exchange)
these are the H ARTREE equations (10.15).
Typically, they are solved by an iterative procedure in which at each iteration
step the N orbitals are obtained from numerical integration of the radial equa-
tion; the screening potentials are obtained from averaging over the orbitals of
the previous iteration step. This is repeated until the orbitals are reproduced
in each step, i.e. until they are self consistent (SCF approximation).
An initial estimate for the repulsive screening potential may be obtained from
the T HOMAS -F ERMI model, which is based on a statistical electron distribu-
tion in the C OULOMB potential Z fold charged nucleus. A closed expression
may be obtained from (10.23) with (10.27).
10.2 HARTREE-FOCK Method
The above discussion of the H ARTREE SCF method was meant as a first introduc-
tion on how to calculate wave functions and energies for multi-electron systems.
So far, we have completely neglected the PAULI principle. It demands for some
major additions to the procedure, first introduced by F OCK and independently by
S LATER the so called H ARTREE -F OCK (HF) method (also HF approximation or
HF theory). It provides the basis for essentially all advanced methods of atomic and
molecular structure computation and modern quantum chemistry. In detail it is a
rather sophisticated theory and we cannot give here a tutorial for those who actu-
ally want to use it (we refer to the numerous textbooks on quantum chemistry, e.g.
S ZABO and O STLUND 1996). In the following we just sketch some of its essential
ingredients.
10.2.1 PAULI Principle and SLATER Determinant
To account for the fact that the N electrons in an atom are indistinguishable
fermions, we have to anti-symmetrize the total wave function in the S CHRDINGER
equation (10.5). The H AMILTON operator is still represented by (10.3) it is not
explicitly spin dependent and we start with product wave functions similar to
(10.6).
However, HF theory introduces so called spin-orbitals (q ) = s (r j ) (j )
(as already applied for He in Sect. 7.3.1) and uses products of these:
(q 1 , . . . , q N )
= 1 (q 1 )2 (q 2 ) . . . (q 2 ) . . . N (q N )
= 1s (r 1 )1 (1)2s (r 2 )2 (2) . . . s (r j ) (j ) . . . N
s
(r N )N (N ).
The indices = {n m ms } represent again the standard sets of four quantum

numbers for each electron. Antisymmetrization is now realized most conveniently
by the so called S LATER determinant (of an N N matrix) which generates an
antisymmetric sum of product wave functions:
(q 1 , q 2 , . . . , q N )

1 (q 1 ) 2 (q 1 ) ... (q 1 ) . . . N (q 1 )

1 (q 2 ) 2 (q 2 ) ... (q 2 ) . . . N (q 2 )

1 . . . ... ... ... ... . . .
= . (10.28)
N ! 1 (q j ) 2 (q j ) ... (q j ) . . . N (q j )
... ... ... ... ... . . .

1 (q ) 2 (q N ) ... (q N ) . . . N (q N )
N
Exchange of two particles corresponds to exchanging two rows of the determinant.

This changes the sign of the total wave function. Thus, (q 1 , q 2 , . . . , q N ) is indeed
antisymmetric in respect of exchanging two electrons.
One verifies again the PAULI principle (see Sect. 7.3.1): if any two sets of quan-
tum numbers, e.g. 1 and 2, are equal, two whole columns are equal, and the deter-
minant becomes 0.
The explicit form of the S LATER determinant (10.28) is very clear, but somewhat
lengthy to write and read. Thus, one usually abbreviates it. We write it as a state
vector:
1
| = det 1s (1)1 (1) . . . s (j )j (j ) . . . N
s
(N )N (N ) . (10.29)
N!
Examples Restricted vs. Unrestricted HF

Let us discuss again He as the most simple example, using this terminology. We
know already the results from Chap. 7 to compare with. The general (approximate)
H ARTREE -F OCK wave function of He reads:

(1, 2) = 1 det s (1)1 (1) s (2)2 (2) . (10.30)
1 2
2!
In general, the spatial orbitals for different electrons will be different. If one fully
allows for this possibility, one speaks of unrestricted H ARTREE -F OCK (UHF).
For closed shells, one may assume that each spatial orbital is occupied twice,
once by an electron with spin and once by an electron with spin. This approxi-
mation is called (closed shell) restricted H ARTREE -F OCK (RHF). It is obviously a
reasonable approximation for the He ground state, since the two electrons are abso-
lutely identical, both being in a 1s orbital:
1 s
1s 2 (1, 2) = det 1s (1)(1)1s s
(2)(2)
2!
(1)(2) (1)(2)
= 1s
s s
(r 1 )1s (r 2 ) .
2
This is exactly the result (7.39) for the He ground state: a singlet state with total spin
S = 0, MS = 0. Things are more complicated in the excited state apart from the
trivial cases (1) = (2) and (1) = (2). The S LATER determinant in UHF reads
now
1 s
1sn (1, 2) = det 1s s
(1)(1)n (2)(2)
2!
s (r ) s (r )(1)(2) s (r ) s (r )(1)(2)
1s 1 n 2 1 1s 2
= n .
2
Note that this is neither a singlet nor a triplet state. Thus, as we know from Sect. 7.3,
these are not eigenstates of the Hamiltonian (10.9) for the excited He system. We
have already discussed in Sect. 7.3.4 how to amend this fact by constructing singlet
and triplet states from linear combinations of S LATER determinants.
Now let us have a brief look at Li. Its ground state configuration being 1s 2 2s, the
appropriate S LATER determinant would read
1 s
1s 2 2s (1, 2, 3) = det 1s s
(1)(1)1s s
(2)(2)2s (3)(3)
3!
1 s
or alternatively = det 1s (1)(1)1s s s
(2)(2)2s (3)(3) .
3!
In the first case electron 1 and 3 have identical spin, in the second case electron 1
and 3 have opposite, 2 and 3 identical spin. Obviously the two electrons in the 1s
orbital are no longer fully equivalent (as indicated by 1s ). Thus it makes sense to
apply UHF. We shall see in a moment how this works out when actually determining
the spin-orbitals in a H ARTREE -F OCK calculation. Clearly, the UHF procedure will
be more expensive than RHF, and hence one often restricts calculations for large
system with many electrons to RHF.
We finally mention, that again orthonormality is required for all sets of orbitals.
However, orthogonality between orbitals with opposite spin is automatically en-
sured since = = 0. This eventually leads to one set of orthogonal spatial
orbitals for spin and one orthogonal set for spin which are not mutually or-
thogonal.
Simultaneous Measurement of Observables

A few words on the simultaneous measurement of observables are now required:
for electron j the angular momentum operators Lj and
2
Lj commute according to
(B.2). Since
2
Lj is part of the j operator and commutes with any function of r j ,

also Lj commutes with the one particle H AMILTON operator. Thus, also L= Lj
and hence z commute with the N particle Hamiltonian. And since the
2
L and L
latter is completely independent of the spins, it also commutes with
S= S j as
well as with
S and with
2
Sz .
One thus may generate fully antisymmetric states | LSML MS , which are eigen-
states of the H AMILTON operator and simultaneously eigenstates of the total an-
gular momentum and of the total spin.
We have already constructed such states for atomic He. For the ground state of
He the above S LATER determinant provides such a state. For the excited states,
however, one needs for the diagonalization of H c linear combinations of several
S LATER determinants in order to obtain RUSSEL -S AUNDERS (LS) coupled eigen-
states | LSML MS . In Sect. 7.3.4 we have discussed this already in detail for the
example of He atoms in their first excited states.
10.2.2 HARTREE-FOCK Equations
In the H ARTREE -F OCK approximation (HF) one determines the S LATER determi-
nant (10.29) in such a way that it minimizes the expectation value of the H AMILTON
operator (10.3) which (to 1st order) is then the total energy of the system, replacing
(10.16):2
N
N N
| = 1
WHF = |H + [J K ]. (10.31)
2
=1 =1 =1
As in the H ARTREE approximation, is the energy (10.17) of a single electron in

the unscreened C OULOMB field of the nucleus, while J is the C OULOMB integral
(10.18). The difference between (10.31) and (10.16) is due to the characteristic
exchange integrals
1
K = (q) q r r (q) q (10.32)
s s s s
(r) (r ) (r) (r )
= | 2 d3 rd3 r .
|r r |
2 Onemay wonder why the sum does not exclude = . Note, however, that in this case
C OULOMB and exchange integral just cancel each other.
We have met these already in Sect. 7.3.3 for the He case. They are so to say
representatives of exchange interaction. The scalar product in front of the integral is
=1 if the orbitals have the same spin, it is =0 if they have opposite spin.
We cannot go into details of the variational procedure. The main outcome is again
a set of N coupled integro-differential equations, this time for the spin-orbitals.
These H ARTREE -F OCK equations may be written as
+ N , + N ,

he (r) + Vd (r) (r) Vex (r) (r) = E h (r). (10.33)
=1 =1
The summations are over all occupied spin-orbitals (see footnote 2).
The functions Vd (q j ) and Vex (q j ) are called direct potentials (or C OULOMB
terms) and exchange potentials, respectively. They are, however, not potentials in the
standard sense. Rather, they are defined through the spin-orbitals (q j ), with q =
r, s, which in turn are to be determined from the HF equations. The direct potential
corresponds essentially to the screening term in the central potential (10.14) of the
H ARTREE method:
(r ) (r )
1
Vd r = q r r q = d3 r . (10.34)
|r r |
In principle, this matrix element of 1/|r r | implies integration over the spatial
coordinates r and the scalar product of spin functions. However, since bra and ket
vector in Vd are the same, the spin product always gives 1. In contrast, the exchange
potential is defined as a so called non-local potential
1
Vex (r) = q r r q (10.35)

(q ) (q ) s s
(r ) (r )
3
= dq = |
2
d r ,
|r r | |r r |
with the characteristic exchange of the sets of quantum numbers vs. . It involves
the orbital to be determined by (10.33) (and not just the average over all other
orbitals as the direct potential).
Thus, (10.33) is not even an eigenvalue equation for the spin-orbitals. We men-
tion in passing that, very formally, one defines a so called F OCK operator which
allows one to rewrite (10.33) so that it looks like an eigenvalue equation. Since from
our perspective no deeper insights into atomic and molecular structure are gained
by this concept, we suppress further details.
The solution of the H ARTREE -F OCK equations for the spin-orbitals may in prin-
ciple be done by iteration following closely the scheme of the H ARTREE method
described in Sect. 10.1.4. Alternatively, one introduces predefined basis sets (e.g.
special Gaussian type functions) from which the orbitals are constructed as linear
combinations. The variational procedure is then applied to the coefficients, and thus
an algebraization of the problem is achieved.
In any case, one finally obtains, in addition to the binding energies W of all
electrons, the symmetries of the states, their orbitals, and the effective self consistent
potentials. We recall again, that the total energy is then obtained from (10.31) and
not (!) by simple summation over all orbital energies E h (which would account for
the repulsive terms twice).
The S LATER determinants (HF states) are, however, not yet necessarily a correct
description of the atom. Rather, they provide, a good starting point. As mentioned
above, in the next step one has to diagonalize the total Hamiltonian. As long as spin-
orbit interaction does not play a major role (i.e. for light atoms where the L
S terms
can be neglected or may be treated as small perturbations), one may diagonalize
L and
2 2
S by linear combinations of several HF states within one configuration,
invoking some angular momentum algebra with C LEBSCH -G ORDAN coefficients.
We have explained this in detail in Sect. 7.3.4 for the He atom.
The states thus obtained describe the system as good as possible in the framework
of the independent particle model and Russel-Saunders coupling (LS coupling).
10.2.3 Conguration Interaction (CI)
Irrespective of this trivial LS diagonalization, the model of independent electrons

which simply move in a screened central field, is an approximation which often
does not meet demands for state-of-the-art accuracy: correlations between individ-
ual electrons are not included in simple product wave functions from one single
configuration. Various post HF approaches are available, such as the M LLER -
P LESSET perturbation theory discussed in Sect. 10.1.2. In respect of exchange in-
teraction, the non-locality of the potential is a key property, which makes a sim-
ple perturbation approach difficult or impossible. One often tries nevertheless with
clever local approximations, to estimate a quasi-potential for exchange (see e.g.
Sect. 10.3).
Some serious consequences from neglecting correlations have already been dis-
cussed in the context of autoionizing doubly excited states of He (see Sect. 7.6.1).
One may try to include such correlations by a linear combination of orbitals from
several configurations. This approach is called configuration interaction (CI). One
optimizes the additions from other configurations e.g. by variational methods.
The difference between the HF result (if necessary including the spin diagonal-
ization) and the exact energy is defined as correlation energy:
Wcor = Wexact WHF . (10.36)
For He in its ground state it amounts to Wcor = 0.114 eV (1.4 %), for neon already
10.3 eV (already 3 % of the total energy Wexact 3507 eV). Clearly, for atoms
of the size of Ne and larger, efforts beyond the HF method are required. For even
larger atoms the additional problem of spin-orbit coupling and relativistic effects
have to be accounted for. They have to be included in a CI calculation. Yet another,

serious problem arises from the fact that in spectroscopy one always determines
transition energies, which are the difference between typically large total binding
energies for all electrons. This may lead to rather big errors in the calculated energy
difference, which is to be compared with the directly measured value even if the
relative precision of the total calculated energy appears astonishingly high.
10.2.4 KOOPMANs Theorem
Helpful in this context is the so called KOOPMANs theorem: It states that the ener-
gies E h calculated for each orbital with the HF equations (10.33) for each electron
(more precisely for each set of quantum numbers) is approximately equal to the
ionization energy of that electron. Consequently
the excitation energy of an atom is approximately equal to the difference of

the respective orbital energies.
This is, however, just a 1st order approximation, since changes in the atomic
structure due the excitation or ionization process are not accounted for.
Today, the H ARTREE -F OCK method is almost exclusively used as an initial basis
for often very elaborate CI procedures. These are by no means limited to atoms.
Similar and much extended methods are used in molecular physics and quantum
chemistry, as well as in solid state physics. A variety of powerful programmes are
even commercially available (see also Sect. 3.6.8, Vol. 2). A detailed description
would, however, reach far beyond the scope of the present book.
Section summary
The H ARTREE -F OCK (HF) method employs electron wave functions cor-
rectly antisymmetrized as S LATER determinants (10.28) constructed from
spin orbitals (q ) = s (r j ) (j ).
The latter are derived by solving the H ARTREE -F OCK equations (10.33), a set
of coupled integro-differential equations which include a non-local exchange
potential.
The total HF energy is obtained from (10.31) as a sum of single (unscreened)
electron energies, C OULOMB screening energies and exchange integrals.
However, in the general case one has to diagonalize the Hamiltonian (in LS
coupling together with L and
2 2
S ) by a linear superposition of S LATER deter-
minants for a given electron configuration.
These HF energies are still limited by the model of independent electrons
moving in a central field. To account for electron correlations one has to go
further, in the simplest case to M LLER -P LESSET perturbation theory. A rig-
orous treatment requires CI, i.e. a linear superposition of several electron con-
figurations.
KOOPMANs theorem gives a very useful 1st order approximation to ioniza-

tion potentials and excited state energies for a single active electron.
10.3 Density Functional Theory
Walter KOHN (1998) received the N OBEL prize for the development of density func-
tional theory (DFT). This method is today the method of choice for a range of ap-
plications in atomic, molecular and solid state physics (in their ground states) for
energy and structural determination of complex systems. Different from the proce-
dures discussed so far, DFT does not focus on the wave functions of a system for
physical interpretations these are anyhow often not of primary relevance. Rather,
DFT as its name already implies, is a method to determine electron densities.
The key goal is the characterization of the ground state properties for (interact-
ing) many-particle systems, based on their single electron densities. The starting
point of DFT is the fact (as proven by H OHENBERG and KOHN), that knowing the
electron density (r) of the ground state for an electronic system (with or without
interaction) characterizes the system completely.
In the model of independent particles, which interact only via an averaged po-
tential with each other and with an external field or the field of many atoms (in a
molecule or solid state material), the electron density of a system with N electrons
is given by
N

(r) = j (r)2 . (10.37)
j =1
The so called KOHN -S HAM orbitals j are obtained from the eigenvalue equations
(in a.u.)

1
+ vKS (r) j (r) = j j (r). (10.38)
2
The density dependent (effective) KOHN -S HAM potential

(r )d3 r
vKS (r) = v0 (r) +
+ vxc (r) (10.39)
|r r |
is composed by an external potential v0 (for a single atom of nuclear charge Z
this is simply Z/r), the repulsion of the electrons among each other (the second
term) and a non-classical exchange potential vxc ((r)). The latter is the functional
derivative of the exchange energy function Wxc ((r)) in respect of (r):
Wxc ((r))
vxc (r) = . (10.40)
(r)
The nature of exchange interaction has been discussed already in Chap. 7.
10.3 Density Functional Theory 511
Equations (10.37)(10.39) are called the KOHN -S HAM equations. Without ex-
change they are equivalent to the H ARTREE method. The key problem, and the main
efforts of DFT thus is, to find an adequate treatment of the exchange potential. In the
most simple variety, the so called local density approximation (LDA), one chooses
a local potential derived from the local density functional which similar as in the
T HOMAS -F ERMI model (Sect. 10.1.5) is directly related to the electron density:

Wxc (r) = xc (r) (r)d3 r. (10.41)
Here xc () is the exchange correlation energy per electron at constant electron
density .
In any case, the total state of the system is described by the corresponding
S LATER determinant (r) constructed with KOHN -S HAM orbitals according to
(10.28). The procedure then follows a variational approach: the electron density
(r), which ultimately defines the KOHN -S HAM orbitals, is optimized until the ex-
pectation value of the H AMILTON operator (10.3)
|(r)
W = (r) |H (10.42)
reaches a minimum. In mathematical form one writes

|(r) = 0.
W (r) |H (10.43)
Then (10.42) gives the best ground state energy within the framework of DFT, and
the corresponding (r) describes the best electron distribution for the ground state.
Unfortunately we cannot go into more detail of this method which becomes
increasingly important in modern structure theory of atoms, molecules and solid
states. Figure 10.3 shows some typical results for atomic electron densities ob-
tained with a simple PC programme (S CHUMACHER 2011) found on the Inter-
net. According to the authors, the programme uses the Xalpha functional and
The model is not spin polarized and non-relativistic. Overall, it has an accu-
racy between LDA and LSDA. Plotted in Fig. 10.3 is the radial electron density
4r 2 Ne (r) = 4 nj j r 2 Rn2j j (r), summed over all occupied orbitals nj j so that

4 0 r 2 Ne (r)dr = Z. The maxima of this distribution show the different, occu-
pied shells. For comparison the completely smooth T HOMAS -F ERMI electron den-
sity (10.26) is also plotted clearly it may only be used as a 0th order approxima-
tion.
A standard textbook on DFT is PARR and YANG (1989). In the mean time,
this powerful method has developed in a variety of brands. Particularly important
progress has been made in recent years with time dependent density functional the-
ory (TDDFT) which has become a key tool for physical and quantum chemical
structure calculations and simulations of dynamic processes in complex systems.
With certain tricks it is now possible to study excited states and, very important, op-
tical transitions. The review of M ARQUES and G ROSS (2004) provides an excellent
introduction, while more recent advances are summarized by C ASIDA (2009).
2 1s 2 2s 2 2p 6
1s 2 6C
12 1s 11Na 20 18Ar
6 10 2s 2 2p 6 10 Ne
8 10 15
1s 2
4r 2 Ne(r) / a0-1
2He
4
6 10 2s 2 2p 6
2s 2 2p 2
2 4 3s 2 3p 6
10 5
2 3s only 2
1s
0 0 0
0 2 4 0 2 4 0 2 4
r/ a0
Fig. 10.3 Computed radial electron density for several atoms (full lines) in atomic units. For
comparison the densities obtained from the T HOMAS -F ERMI potential (dashed lines). For Na,
with its isolated valence electron in a 3s orbital, its radial density is multiplied by a factor of 10
(black, dash dotted line)
Section summary
The density functional theory (DFT) provides a very efficient, self consistent,
widely used method for electronic structure calculations (mainly for electronic
ground states) in atoms, molecules and condensed matter.
It concentrates on electron densities, rather than on wave functions, and ap-
proximates exchange by different types of judiciously chosen functionals, in
the simplest case by a local potential (LDA).
10.4 Complex Spectra
10.4.1 Spin-Orbit Interaction and Coupling Schemes
Up to now we have restricted our description of multi-electron systems completely

onto electrostatic interactions. As already discussed for the He atom, exchange in-
teraction (which is of such electrostatic origin) leads to coupling of spin and orbital
angular momenta of the N individual electrons, such that they combine to a total
spin and a total angular momentum of the atom:
N
N

S=
Si and
L=
Li .
i=1 i=1
We recall that this kind of coupling by exchange interaction dominates for small
nuclear charge Z and is called RUSSEL -S AUNDERS coupling or (somewhat am-
biguously) LS coupling.
In contrast, spin-orbit interaction is of magnetic origin. As long as it is small (e.g.
in the case of He) it may be added as a perturbation, typically being proportional
L
to S and to Z 4 (see Sect. 6.2.4). Under the influence of this perturbation, total
10.4 Complex Spectra 513
Fig. 10.4 Doublet splitting 8000 Tl
fine structure splitting / cm-1

in the 3rd main group of the ... ns 2(1S0) np 2Po1/2, 3/2
periodic system as a function
of the nuclear charge Z. Red 6000
for the ground state
ns 2 (1 S0 )np 2 P01/2,3/2 and grey 4000
for the ns 2 (1 S0 )(n + 1)p In
2 P0
1/2, 3/2 excited state ... ns 2 (1S0)
2000 Ga (n+1)p 2Po1/2,3/2
B Al
0
0 20 40 60 80
Z
spin and total orbital angular momentum finally couple to an overall total angular
momentum of the active electrons:
L +
J = S.
Except for very small Z one typically considers only the interaction of spins with
their own orbit and neglects spin-spin as well as spin-other-orbit interactions. This
leads to fine structure splittings which may be described very similar to those of
quasi-one-electron systems. This also holds for the Z EEMAN splitting observed in a
magnetic field.
However, for large Z spin-orbit interaction terms become comparable or even
larger than exchange interaction so that RUSSEL -S AUNDERS (LS) coupling does
no longer describe the physical reality appropriately. Strictly speaking, L and S are
then no longer good quantum numbers even if they are often still used to charac-
terize the atomic states. We have already reported deviations from the standard (LS)
selection rules for optical transitions and we shall see more of it below, e.g. in the
case of the Hg atom, where strong intercombination lines are observed even though
in the term scheme one still distinguishes a singlet and a triplet system (Fig. 7.13).
In the extreme case for very high Z one typically finds spin-own-orbit interac-
tion to dominate and the coupling scheme changes completely. Most appropriately
it is described by jj coupling of the N atomic electrons: First all individual orbital
angular momenta Li (quantum number j ) couple with their respective spin Sj
(sj = 1/2) to form an individual total angular momentum J i (quantum number ji ).
Only then the individual J i couple together (under the influence of exchange inter-
action which may now be considered to be a perturbation) and form the total angular
momentum J of the atom:
N

Ji =
Li +
Si
J=
J i. (10.44)
i=1
There is a variety of cases in between LS and jj coupling which we shall

get to know by way of example in the following. First, we illustrate in Fig. 10.4
the increasing fine structure splitting for the elements in the 3rd main group of the
Fig. 10.5 Exchange Sn
W (2S+1PJ) - W ( 3P0 ) / cm-1

configuration
interaction and fine structure
... np(2P) (n+1)s 1P
splitting as a function of Z 4000 1
spin-orbit interaction
exchange interaction
for the example of the terms
np(n + 1)s in the 4th main Ge 3P
group of the periodic system. 2
Plotted is the difference of the 2000 C
term energies in respect of the Si
3 P state 1P
0 1
3P 3P
2 1
0 3P
0
0 10 20 30 40 50
Z
periodic system (Table 3.3). While FS is essentially negligible for the element boron
(B), in the case of thallium (Tl) it reaches an order of magnitude of eV.
A direct comparison of spin-orbit and exchange interaction becomes possible in
the 4th main group: in Fig. 10.5 the excited ns 2 np(n + 1)s configuration for sin-
glet and triplet states with their respective fine structure splitting are shown. For the
carbon atom (small Z) exchange interaction clearly dominates, separating the 1 P1
and 3 PJ states by nearly 0.2 eV, while the fine structure splitting of the 3 PJ states is
nearly negligible. In contrast, for large Z fine structure splitting (i.e. spin-orbit in-
teraction) becomes comparable to the singlet-triplet splitting (i.e. to exchange inter-
action) or even larger in particular so for tin (Sn). Thus, RUSSEL -S AUNDERS cou-
pling hardly describes reality any longer correctly. The next element in this group,
lead (Pb), is not shown here. Event though spectroscopic data are still denominated
as singlets and triplets (K RAMIDA et al. 2013), missing terms in this scheme clearly
document the transition from LS to jj coupling.
10.4.2 Examples of Complex Spectra
The following Figs. 10.610.10 represent some characteristic examples for com-
plex atomic spectra. We shall discuss these briefly. The presentation used here to
show the term positions together with observed transitions are called G ROTRIAN
diagrams. However, to avoid confusion we communicate only a few of the most im-
portant transitions and by no means all known energy terms. The diagrams contain
in any case a lot of information as a result of many years of spectroscopic work
of many research teams. The full data sets are collected by K RAMIDA et al. (2013),
who have analyzed and evaluated the literature sources critically and presented them
in a user friendly manner.
We start with a typical and particularly important example, carbon (C), an el-
ement of the 4th group, for which a G ROTRIAN diagram is shown in Fig. 10.6.
According to H UNDs rules, the ground state is a triplet with the configuration
1s 2 2s 2 2p 2 3 P0 , where the two spins of the 2p electrons are parallel (symmetric),
the orbital wave function is, however, antisymmetric (PAULI principle). Since car-
bon is a light element (small Z = 6), spin-orbit interaction is weak, and hence the
C+ ionic excited state: 2 2 2

120
(2 3P) / 1000 cm-1
2P
110
100
2S
90
80
2D 70
60
2 2 3
4P
50
22 2P 40
2 1/2,3/2 C+ ionic ground state 1Po
120 2 2 2 30
2 22 2 22 2 22 3So 20
110 5P
3
100 1 o
P 3P
o 1
S 1 P 1 D 3 S 3P 3D 1 o o
P 1D 1Fo 3P 3D 3F
o o o
10
I 5 0
5 4
4 3Po
4 3
-1
1189
I / 1000 cm
6
70 3Do
11 .5
143
3 940 5 94 3
14 131
60 17 13
6
156
14
50
6
24
40 displaced
8
30 terms
5So
19
16
3
20 2
10 2
0 2 [He] 2 2 2 2 3P C ground state
o
Fig. 10.6 G ROTRIAN diagram for carbon, 6 C. WI ( 2 P1/2 ) = 90820.42 cm1 . Energies are
given in 1000 cm1 : left excitation energies of the neutral atom (C I), right the excitation ener-
gies of the ion (C II); wavelengths marked on the transitions are given in nm (rounded to integers)
splitting within the 3 PJ multiplet is relatively small: W (J = 2) W (J = 0) =

43.40 cm1 . On the left side of the diagram the terms are shown for which one
of the two 2p electrons is excited. The configurations are 2s 2 2p ns (left most),
2s 2 2p np (middle) and 2s 2 2p nd (right). The spectra resemble somewhat those of
the simple spectra of the alkali and alkaline earth atoms. The terms converge to the
lowest ionization limit at which the C+ ion in its ground state 2s 2 2p 2 P is formed.
The energetically next higher terms belong to the configuration 2s 2p 3 and
2s 2p 2 ns. Here a 2s electron is excited one speaks about displaced terms. They
converge to an ionization limit which belongs to excited states of the C+ ion as in-
dicated on the right side of Fig. 10.6. Typically, only the lowest displaced terms are
well known from experiments.
The term diagrams for the heavier elements of the 4th group of the periodic sys-
tem (Si, Ge, Sn and Pb) are similar except for the increasing spin-orbit interaction
as discussed already summarized in Fig. 10.5 for the ns 2 np (n + 1)s configurations
as examples.
Atomic nitrogen (N), for which the G ROTRIAN diagram is shown in Fig. 10.7, is
an example for elements of the 5th group of the periodic system. The ground state is
a quartet term (4 S0 ) with a configuration [He]2s 2 2p 3 . The excited states in which
2s 2 p 3
50
160 5 So
[He] 2s 2 2p 2 3P 0,1,2 N+ ionic ground state 40
150
30
2s 22p 2(3P) 2s 22p 2(1D) 2s2p 4
ns np nd ns np 20
4 o 4 o 2 o
2 P 4P 2So D 4Po S 2Do P 2P 4F 2F 4P 4D 2D 2D 2Fo 2 10
2Do 2Po D
120
WI 0
5 5
110 5 4
4 3 displaced
4 410
W - WI / 1000 cm-1
100 1558
terms
3 528 4P
90 859
123
132
7
4
6
3 1487
12
95 10
13
80
14
9
70
11
W - W(2 4S) / 1000 cm-1
60
13
3
11
50 4
11
8
17
40 5
2
12
30
0
20 2
10
2 [He] 2s 2 2 p 3 4S0 N ground state
0
Fig. 10.7 G ROTRIAN diagram for nitrogen, 7 N. WI ( 3 P0 ) = 117 225.7 cm1 , WI ( 1 D) =

132 541.9 cm1 . Otherwise as Fig. 10.6
one of the three 2p electrons is excited may be divided into three groups, related to
the N+ ionic state towards which they converge: 3 P, 1 D or 1 S. On the left of the dia-
gram the configurations 2s 2 2p 2 ns, 2s 2 2p 2 np and 2s 2 2p 2 nd with their respective
multiplets are given, converging to the 2s 2 2p 2 3 P configuration of the N+ ion. On
the right of these the less well known configurations 2s 2 2p 2 ns and 2s 2 2p 2 np are
displayed. They converge towards the N+ (2s 2 2p 2 1 D) state. Terms converging to
the 2s 2 2p 2 1 S state of the N+ ion are not known for nitrogen. However, such terms
are observed for the heavier elements of the 5th group (e.g. for As). The 4 P and 2 D
states on the right side of the diagram belong to the configuration 2s 2p 4 , where
one of the stronger bound 2s electrons is excited. As for carbon these are called
displaced states.
The G ROTRIAN diagram of oxygen (O), typical for the 6th group in the periodic
system, is shown in Fig. 10.8. The ground state configuration 1s 2 2s 2 2p 4 leads to
the same three states 3 P, 1 D and 1 S as in the case of the 4th group. The two extra
electrons have opposite spin and fill two orthogonal p orbitals. The ordering of
the 3 P multiplet is now inverted: for oxygen the 3 P2 term is energetically lowest,
while for carbon the terms are normally ordered. Since the O+ ion is isoelectronic
with neutral N, three terms (4 S, 2 D and 2 P) are possible in the O+ ionic ground
state configuration 2s 2 2p 3 . Thus, there are three ionization limits towards which
the excited state configurations of the neutral atom 2s 2 2p 3 nl may converge.
2 22 3(2Po)
O+ ion ground state 60
o 2 22 3(2Do)
2Po) [He] 2 2 2 3 4S
I ( 1Po o 1S 1,3D 1,3 o o
3P 1 P 1,3D
= 150.3 P
1,3D 1,3Po 1,3Fo 40
2Do) 1Do 3Do
I ( 1P 1,3F 1,3 o 1,3
S D
o 1,3Go
2 22 3(4S)
5
= 136.6 5 4 2 2 20
4 3 3Po
120 o o
3P 5 P
3 S 5S 3D 5Do
o
I 0
5 3
100 395
-1
4
I / 1000 cm
88
77
80
94
80
80
11 5
3 8 79
77
( 2 3P) / 1000 cm-1
99
95
103
60 displaced terms
98
1 30
40
2s2 2p4 1S
92
20
2s2 2p4 1D
0 [He] 2s2 2p4 3P O ground state
o
Fig. 10.8 G ROTRIAN diagram for oxygen, 8 O. WI ( 4 S3/2 ) = 109 837.02 cm1 . Otherwise as
Fig. 10.6
Next we discuss the term scheme of neon (Ne) as a typical representative of the
rare gases. All rare gases are characterized by a closed rare gas outer shell (ns 2 np 6 ).
The neutral ground state is thus always a 1 S0 state. The G ROTRIAN diagram of neon
is shown in Fig. 10.9. Its neutral ground state configuration is 1s 2 2s 2 2p 6 . Excita-
tion of one electron from the 2p shell leads to terms which converge to the Ne+
ion ground state 2s 2 2p 5 2 Pj with j = 3/2 and 1/2. The coupling scheme cannot be
expressed in the form of the typical RUSSEL -S AUNDERS coupling, rather one uses
a j scheme: In this case the ionic ground state configuration 2s 2 2p 5 2 Pj cou-
ples with the excited n electron to form an angular momentum K. Each [K] state
splits under the influence of spin-orbit coupling (K couples with the electron spin
of the excited electron) into a doublet with J = K 12 . One uses the terminology
(2S+1 Lj )nl 2S+1 [K]J , e.g. the first excited state of neon is (2 P3/2 )3s 2 [ 32 ]2 . In all
rare gases some of the first excited states, those with J = 0, 2 or 3, are metastable:
they cannot decay to the ground 1 S0 state due to the selection rule J = 0 1, but
0 0.
The famous red line of the helium-neon laser is a transition between two higher
lying excited states in neon: (2 P1/2 )3p 2 [ 32 ]2 (2 P1/2 )5s 2 [ 12 ]1 , marked red in
Fig. 10.9. The population of the upper level is achieved by collisions with metastable
He in the 1s2s 1 S0 excited state. Depopulation of the lower laser level is supported
by collisions with the glass wall of the laser cell.
Ne+ ionic excited state: 2s2p 6 2S 220

390
np 5 210
380 1 Po 4
370 200
3
360 190
230
[He] 2s 2 2p 5 2Po1/2, 3/2 Ne+ ionic ground states 50
220
2s 2 2p 5 (2Po 3/2) 2s 2 2p 5 (2Po 1/2) 40
210 ns np nd nf ns np nd nf
30
200 2[7/2]o 2
[9/2]o 27
2[5/2]o 2[5/2]o 2[7/2]o 20
190
2[3/2]o 2[3/2]o 2[5/2]o 2[3/2]o 2[5/2]o 2[7/2]o
2[3/2]o 2
10
180 [1/2]o 2[1/2]o 2[3/2]o 2[1/2]o 2[1/2]o 2[3/2]o 2[5/2]o
WI 0
170 6 6
5 5 4 4
4 3 1838
W - WI / 1000 cm-1
160 4
W - W(2 1 S0) / 1000 cm-1
150 3 627
140
3 displaced
130 terms
63 62
120
110 74
[He] 2s 2 2p 6 1S0
Ne ground state
0
o o
Fig. 10.9 G ROTRIAN diagram of neon, 10 Ne. WI ( 2 P3/2 ) = 173 929.75 cm1 , WI ( 2 P1/2 ) =
174 710.17 cm1 . The so called j - coupling scheme holds. Otherwise as Fig. 10.6
In neon one also observes some weak absorption lines in the XUV range. These
transitions lead from the ground state to highly excited states with the configuration
2s 2p 6 np. These are again displaced terms where one of the 2s electrons is excited
to an np level. Such terms are also found for heavier rare gases.
As a last example we consider aluminum (Al). Its G ROTRIAN diagram is shown
in Fig. 10.10. The ground state configuration [Ne]3s 2 3p contains only a single elec-
tron in the 3p shell. The excited states where this 3p electron is excited thus form
a simple, alkali like term scheme as indicated on the left in the diagram Fig. 10.10.
All these excited terms converge towards the Al+ ionic ground state 3s 2 1 S0 . In ad-
dition, there are series of complex terms, all except one embedded in the ionization
continuum, as shown on the right hand side of the diagram. A 3s electron is ex-
cited in this case. There are three terms with the electron configuration 3s 3p 2 , the
4 P multiplet having the lowest energy. Terms with higher excited configurations
J
3s 3p n finally converge all towards the 3s 3p 3 P excited state of the Al+ ion.
10.5 X-Ray Spectroscopy and Photoionization 519
o
100 Al+ ionic ground state: 3s 3p(3P ) 50
[Ne] 3s 2 1S 3p ns np nd
0
WI ( 3P )
o
Al+ ionic ground state 2 S 2P 4
P 4 Po 2S 2 2
P D 2D o4
2F o4
P D 40
= 85.7 5 30
3s 2(1S0) 5 4
5 4
70 ns np nd nf 20
3
2S 2 2 3 4
1/2 P3/2,1/2 D5/2,1/2 F7/2,5/2
W - W(3 2 P1/2) / 1000 cm-1
60 10
WI 0
W - WI / 1000 cm-1
6 4 5
40 5 5 4 08
5 3
30
4
5
3 112 3
4 131 displaced terms
20 4
34
39
4
10
[Ne] 3s 2 3p 2P1/2 Al ground state
3
0
Fig. 10.10 G ROTRIAN diagram for aluminium, 13 Al. WI ( 1 S0 ) = 48 278.37 cm1 . Otherwise
as Fig. 10.6
Section summary
With increasing atomic number spin-orbit interaction increases Z 4 and
RUSSEL -S AUNDERS (LS) coupling gives a less and less appropriate descrip-
tion.
In the limit if very high Z the orbital angular momentum and spin of each
electron couple with each other to j under the influence of spin-orbit inter-
action, and only at the end the different j couple by the weaker exchange
interaction in a jj coupling scheme.
We have discussed several examples of term schemes (as G ROTRIAN dia-
grams) for more complex atoms with rich structure and increasing complexity.
Excited states of noble gases are described by a j coupling scheme.
We memorize that C, N and O each have 3 lowest multiplet levels with the
same electron configuration 2s 2 2p 2 , 2s 2 2p 3 , and 2s 2 2p 4 , respectively. These
levels are spaced by about 1000 to 2000 cm1 and follow H UNDs rules. In
the case of O the lowest level is 3 P followed by 1 D and 1 S.
10.5 X-Ray Spectroscopy and Photoionization

Today, X-ray spectroscopy is not only an important tool in modern research for de-
tailed structural analysis be it in the context of biomolecules, nano-materials, soft
or crystalline matter it is also indispensable for analytical purposes in physics,
chemistry, medicine and technology. Analytical chemistry and physics makes use
of X-ray spectroscopy in many areas of application, e.g. in archaeological science
to determine the origin of historical findings, or to reveal whether some precious ob-
jects of art are original or fake to mention just two not so mundane examples. One
important asset for todays manifold applications of X-ray spectroscopy is the abun-
dant accessibility to state-of-the-art electron storage rings which have been built
exclusively for the generation and applications of synchrotron radiation, in partic-
ular in the VUV, XUV and X-ray spectral range. Since the methods for generating
and detecting this kind of electromagnetic radiation may typically be considered
part of optical physics (in its broadest sense) and since atoms with high Z play an
important role in this context, this subject has to be discussed here. We can, however,
only touch some aspects of this important area of research and collect some of the
most important basics. In the present section we focus on terminology, methods, and
key observations of spectroscopy. In Sect. 10.6 we shall then elaborate on some of
the pertinent sources for X-ray radiation as prerequisite for spectroscopy. For more
details the reader is referred to specialized literature (e.g. the rather comprehensive
book of ATTWOOD 2007).
10.5.1 Absorption and Emission from Inner Shells
Up to now we have been concerned almost exclusively with the spectra of valence
electrons, i.e. from electrons in the outermost atomic shell. We now turn to the inner
shells, in particular to those of larger atoms. Due to the characteristic dependence of
atomic energies on the square of the nuclear charge Z, we expect the corresponding
spectra at very short wavelengths. In uranium, e.g., an electron in the K shell has a
binding energy of more than 110 keV. Interestingly, in certain respects these spectra
are much easier to understand than those of the complex outer shells of intermediate
size atoms which we have discussed in the last section with the help of G ROTRIAN
diagrams: since all inner shells are filled, transitions between these levels can only
occur when a hole is created in one of the shells typically by collisions or by
photoionization.
We discuss these processes on the basis of a very schematic term scheme for
an element Tg (target) with high atomic number Z, as sketched in Fig. 10.11.
Orbitals and total angular momentum for each electron are again characterized by
the quantum numbers n j , the designation of the shells as K, L, M, N, . . . and
their respective subshells, following the usual schematic which we have introduced
in Sect. 3.1.3. Let us assume that all these inner shells are filled. Three types of
processes may be observed:
Absorption of a photon from one of the inner shell levels can (essentially) oc-
cur only into the continuum, since all inner shells are occupied. This leads to
photoionization

Tg(n ) + Tg+ (n )1 + e
(10.45)
with the energy balance WI = ,
where WI = WK , WL , WM etc. is the ionization potential from the respective
shell. The kinetic energy of the electron ejected is . The initial electron configu-
ration is described by (n ), and the symbol (n )1 indicates that a corresponding
(a) absorption (continua) (b) emission (lines)
n j shells
OI - O...
OI ...
3 3 75/2 NVII
3 3 5/2 NVI
3 2 3/2 NV
3 2 3/2 NIV
3 1 1/2 NIII
3 1 1/2 NII
3 0 /2 NI
NI ...
WM 2 2 5/2 MV
2 2 3/2 MIV
2 1 3/2 MIII
2 1 1/2 MII
2 0 1/2 MI
MI MII ...
WL edges
2 1 3/2 LIII
2 1 1/2 LII
2 0 1/2 LI
LI LII LIII 2 LI LII LIII
1
WK edges 2 L series
K1
1 0 1/2
K edge K series
Fig. 10.11 Absorption and emission of X-ray radiation: (a) absorption is only possible into the
ionization continuum, since the intermediate levels are fully occupied; (b) emission can only occur
after generation of holes in the K, L, M, . . . shells. Lines with red arrows correspond to photon
energies, black lines characterize the energetics of the system
hole has been created in the atomic shell. Since the energy balance has to be
maintained by the ejected electron, one observes as a function of the photon en-
ergy so called absorption edges: for energies below WI there is no absorption.
As soon as > WI , the corresponding shell can be ionized and the cross section
for photoionization jumps to a finite value. For still larger it decays again, as
we have already discussed in Sect. 5.5. We shall come back to this in a moment.
But first, we discuss the other two processes.
If such a whole exists in an inner shell it may be refilled by spontaneous de-
cay with an electron from a higher shell n . This process generates so called
characteristic emission lines
' ( ' 1 (
Tg+ (n )1 . . . n Tg+ n n +
(10.46)
with the energy balance Wn Wn = .
In contrast to the absorption edges just described, one now observes a spectrum
of discrete emission lines. Due to the well known 3 factor (4.38) this process
is particular efficient in the X-ray region and many orders of magnitude more
Fig. 10.12 AUGER electron

emission. Three processes are o o
relevant: (a) generation of a
whole in a lower shell by
photo-absorption (dashed red Wn''
upward arrow), (b) electron
Wn''''
transition from a higher level
refilling this hole (red
downward), (c) exchange of a
virtual photon (wiggly grey
line) and emission of an Wn (a) (b) (c)
electron (red upward arrow),
called an AUGER electron
probable than the absorption of a second photon. The corresponding line spectra
may be read in principle from Fig. 10.11. They are essentially one electron spec-
tra which may be understood more or less as directly as the spectra of hydrogen
like ions or better, alkali like ions if we want to account for the screening of
the nuclear Z/r potential. One may also discuss these spectra in terms of holes
being excited instead of electron transitions. We come back to this concept
quantitatively in a moment.
Finally a third process may occur. The hole is refilled by an electron from a
higher shell, but instead as in reaction (10.46) where a photon is emitted the
excess energy is used to eject a second electron. This so called AUGER electron
emission process may be visualized as
' ( ' 1 (
Tg+ (n )1 . . . n Tg+ n . . . n + ()
' 1 1 ( (10.47)
Tg++ n . . . n n + e ,
where () indicates a virtual photon which is exchanged between the two
electrons, giving the latter sufficient energy to leave that atom. As one reads
from Fig. 10.12 the energy balance is given by
= Wn + Wn Wn . (10.48)
10.5.2 Characteristic X-Ray Spectra MOSLEYs Law
In comparison to the rich material obtained from spectroscopy of outer shells, far
less detailed data exist for the absorption and emission of X-ray radiation from inner
shells for obvious reasons: only during the past decades dedicated synchrotron
radiation sources have provided intense, tuneable X-ray sources of excellent quality.
Fortunately, the available material is well documented and excellently backed up by
theoretical computations. We use here the NIST X-ray data-bank (C HANTLER et
al. 2005), which offers characteristic emission lines in tabulated form, as well as
10 K
5 data )
.56
(0.924 , 7 LI
fit parameters LII
(n*, qs) K
4 LIII
)
5.21
,1
9)
3 1
1.4
.95
keV) 1/2
2, (1
.06
MI
(1
) )
2 .82 .98 MV
, 21 99 , 24
(Wedge
6 0
5) (3.5 (4.1
, 4.9 ) NI
1 5 98 4 , 29.26 NIV
(2. 8) (6.4 5
11.7 NVII
13 ,
(5.1 ) 6 5 .89)
40.0 3 9 ,
0 (7.305 , (5.01
0 10 20 30 40 50 60 70 80 90
Z
Fig. 10.13 M OSLEY diagram for the lowest X-ray absorption edges for all elements (compiled
from the tables in C HANTLER et al. 2005). Grey and black lines give the results of fits with (10.50).
The numbers in brackets are the parameters (n , qs ). One sees that the inner shells cannot be fitted
optimally over the whole periodic system with one set of parameters
absorption edges and photoionization cross sections over a wide range of energies,
for all elements. The data are based mostly on state-of-the-art ab initio computations
which have been tested extensively with the available experimental material.
Figure 10.13 shows the X-ray absorption edges of the K, L, and some higher
shells of all natural elements of the periodic system. This so called M OSLEY dia-
gram is based on the assumption that the transition energies are obtained in analogy
to the RYDBERG -R ITZ formula (1.149) for an effective one electron system with an
effective nuclear charge Z = Z qs :
(Z qs )2 Eh
W = . (10.49)
n2 2
Note that this also implies KOOPMANs theorem (see Sect. 10.2.4) assuming that
holes behave just complementary to electrons. The idea is thus to interpret the em-
pirically determined ionization energies for the different shells (i.e. the position of
the absorption edges) grosso modo by a combination of a screening parameter qs
and the quantum defect n n . One plots the square root of the energies as a func-
tion of Z, as suggested by (10.49):

E h Z Z qs E h
|Wn | = = . (10.50)
2 n n 2
We have already used this so called M OSLEY formula (3.34) for the alkali metal
atoms. Ideally one would expect straight lines, as roughly confirmed by Fig. 10.13.
However, the K and L edges cannot be fitted optimally with one set of parame-
ters over the whole periodic system. For example, at low Z the effective quantum
number of the K shell is nicely n = 1.062 and the screening parameter qs = 1.49
is moderate. In contrast, for larger Z the best fit is obtained with n = 0.924 and
substantial screening qs = 7.56.
For Z > 13 one finds a somewhat crude, but rather useful rule of thumb for
obtaining the K energy:
WK (Z)/ eV 14 (Z 3)2 . (10.51)
For aluminum (Al) with Z = 13 this gives 1400 eV (true value 1560 eV, soft X-
ray) and for tungsten (W) with Z = 74 the formula leads to 70 574 eV (true value
71 676 eV, hard X-ray).
10.5.3 Cross Sections for X-Ray Ionization
We come now to a somewhat more quantitative discussion of photoionization for

multi-electron atoms. It is by far the most important cause for the absorption of
X-ray radiation by matter for photon energies up to some 100 keV. In this energy
range photoionization cross sections, a , are typically given in barn, [a ] = 1 b
(see Appendix A.2). Often one also finds the absorption coefficient (dimension
L1 ) tabulated. We have introduced it in Sect. 1.3.2 with the L AMBERT-B EER law
I = I0 exp(d) for the reduction of intensity I by passage through matter of
thickness d, with = Na a , where Na is the particle density of the absorbers.
The so called mass absorption coefficient / (its dimension being L2 M1 ) refers
to the absorbing mass per unit area instead to d. With the density = Na ma
(ma = atomic mass) this gives / = a /ma .
In Sect. 5.5 we have already treated photoionization in some detail. For high en-
ergies, the first B ORN approximation should give a first, rough guess, even though
we do not expect exact predictions. According to (5.77) the cross section for pho-
toionization in FBA is
Z5
ns (10.52)
n3 (2)7/2
for one electron in an H atom like s orbital, at photon energies |Wn |. Clearly,
the cross section increases strongly with atomic number Z, and decreases with
the principle quantum number n (and correspondingly with the respective electron
shell). Modified formulas will have to be applied for higher orbital angular momenta
(p, d etc.). Also, for the overall cross section one has to multiply (10.52) with the
number Nn of active electrons in each shell, so that a = Nn n for a given shell
n . The B ORN approximation is naturally only a first, very rough approximation
indicating a trend. One finds various approximation formulas in the literature. How-
ever, since excellent, theory and experiment based data banks exist in the literature
BORN L
L Al
106 BORN K
K
photo ionization
104
100 elastic
(coherent)
sum
a / b
photo
1 COMPTON ionization
pair
production
0.01 1 100 104
photon energy / keV
Fig. 10.14 Photo-absorption cross section (in barn) for aluminum as a function of photon energy.
The main contribution is due to photoionization. At higher energies pair production dominates
(according to C HANTLER et al. 2005). Black dashed lines give the photoionization cross section in
B ORN approximation
one should not loose time with such approximation when attacking practical prob-
lems. We use he NIST X-ray data bank of C HANTLER et al. (2005) and B ERGER
et al. (2010) and show in Fig. 10.14 the total photo-absorption cross section a for
aluminum (Al) over a broad range of photon energies. Al is an example for a light
atom (Z = 13). In contrast, Fig. 10.15 presents a for lead (Pb), a heavy element
(Z = 82) well known as a good X-ray absorber and frequently used for protection
from X- and -ray radiation.
For the L edge of lead at 16 keV one reads e.g. in Fig. 10.15 a 52 000 b
(roughly 200 that for Al at the same photon energy). With ma = 207 u this corre-
sponds to a mass absorption coefficient of ca. / = 15.0 m2 / kg. This implies e.g.
that a protection west with just 67 g lead per m2 the irradiation with a 16 keV X-ray
source is already reduced to 1/e 37 %.
As indicated Fig. 10.14 and Fig. 10.15 by dashed lines, the trend and the order
of magnitude of the B ORN approximation. The real cross section differs quantita-
tively somewhat and is of course much more structured. With increasing energy the
cross section drops typically ()5/2 2.5 (and not as predicted by B ORN ap-
proximation ()7/2 ). It jumps, however, at each absorption edge significantly:
i.e. the cross section assumes a finite value whenever the energy of the photon is
high enough to photoionize the respective shell as evident from Fig. 10.11. We re-
member: considering the energetics of individual electrons independently from the
reaction of the whole atom is the very content of KOOPMANs theorem according to
Sect. 10.2.4. The edge energy is then simply the HF orbital energy Wn according
M N
106 BORN L Pb
photo ionization
L
BORN K
104
elastic K
(coherent)
100
sum
pair
production
a / b
1 COMPTON photo
ionization
0.01 1 100 104

photon energy / keV
Fig. 10.15 Photo-absorption cross section (in barn) for lead as a function of photon energy. Oth-
erwise as Fig. 10.14
to (10.33). One recognizes the K and L edges very clearly for Al, for Pb in addition
also the N, M and O edges.
Figure 10.14 and Fig. 10.15 illustrate that energetic photons are not only attenu-
ated by photoionization. Without going into details we summarize the most impor-
tant mechanisms for absorption and scattering of X-ray photons:
1. Photoionization (photoelectric effect)
Tg + Tg+ + e Eq. (10.45)
2. C OMPTON scattering
e + h e + h see Sect. 8.4.5
3. Pair production
e + e+ for 2me c2
4. T HOMSON scattering (elastic)
; k k see Sect. 8.4.5.
Obviously, photoionization dominates the absorption cross section at photon en-

ergies below some 100 keV. Above 2me c2 = 1.022 MeV an electron positron pair
can be generated (pair production). Since both, energy and momentum conservation
must be fulfilled simultaneously, pair production can only occur in the presence of
a third particle, preferentially an atomic nucleus with high Z. Correspondingly the
absorption cross section increases again for hard radiation, as seen in particular
for Pb. For energies around 1 MeV incoherent photon scattering (C OMPTON effect)
plays a central role, as described by the K LEIN -N ISHINA formula (8.130).
In addition to these three processes elastic (also coherent or T HOMSON) scat-
tering of photons plays a (minor) role. As discussed in Sect. 8.4.5, at energies sig-
nificantly below 2me c2 = 1.022 MeV, but far above typical atomic resonances one
expects cross sections el Z 2 e . With the T HOMSON cross section e = 0.665 b
one verifies this for Fig. 10.14 (Z = 13, Al) and Fig. 10.15 (Z = 82, Pb) at the lowest
energies. As energy increases into the relativistic regime (L ORENTZ factor 1)
the elastic cross section drops dramatically (as expected: oscillation amplitudes of
relativistic electrons are much smaller than at classical energies). One roughly finds
el ()2 1/ 2 .
10.5.4 Photoionization at Intermediate Energies
In Sect. 5.5 we have discussed in some detail how (in principle) to compute pho-
toionization cross section specifically for the particularly simple example of the
H atom using the B ORN approximation. In the previous section we have presented
the photo-absorption cross sections for two selected metal atoms and focussed on an
overview from VUV to hard X-ray photons. We now present in some more detail,
by way of example, a non-trivial atom of intermediate nuclear charge. We take the
rare gas atom argon (Ar) with Z = 18 in the interesting range of photon energies
from threshold up to some keV.
Figure 10.16 shows the dependence of the photoionization cross section for Ar
in various ranges of energy. Experimental results and theoretical calculations from
a number of different groups are compared. They supplement each other obviously
quite well, even though a perfect agreement cannot be expected for such a rather
complex element. We do not enter into a discussion of the many details, both in ex-
perimental data as well as in theory. In recent years, high quality data are obtained
almost exclusively at synchrotron radiation electron storage rings (see Sect. 10.6.2),
and in Fig. 10.16 these data are recognized by their small error bars and the con-
sistent trends reported. Notwithstanding this progress, many data have still been
obtained with selected lines from spectral lamps or X-ray tubes. Figure 10.16(a)
gives a nice example. Shown there is the photoionization process

Ar [Ne]3s 2 3p 6 1 S + Ar+ [Ne]3s(3s)1 3p 6 + e (p) (10.53)
from a single subshell, the MI shell (3s 2 ). One distinguishes this process from oth-
ers by measuring the kinetic energy of the outgoing, free electron. The final state
is here a 1 P0 . The term scheme of Ar is quite similar to that of Ne (see Fig. 10.9).
In addition to the full K and L shells, [Ne] = 1s 2 2s 2 2p 6 , in Ar the M shell is filled
with 3s 2 and 3p 6 electrons. The first ionization potential, related to the configura-
tion [Ne]3s 2 3p 5 , is WI = 15.76 eV. To ionize a 3s electron, additional 13.48 eV are
1.5
1.0 (a) dipole-length appr. (b)
0.8 dipole-velocity
Ar 3s
0.6 Houlgate et al. 1.0
0.4 Samson and
Gardner
0.2
0.5
0.0
30 50 70 90 50 100 150 200
6 6
(d)
photionizaton cross section / Mb
2p (c')
3 1.0
4 2p
0
245 250 0.5
2
2s
(c) 0.0
0
300 500 600 800 1000
J.H. McCrary
(e') G.V.Marr
0.20 1s np 0.10 A.J.Bearden
(f) CXRO
0.08
0.15 ZCZ06
3205 3210 (e) 0.06 R.H.Millar
0.10 FFAST
0.04 XCOM
0.05 K edge K+M
double excitation 0.02
0.00 0.00
3210 3230 3250 2400 3600 4800 6000
100
(g)
1.00
0.01
1E-4
from FFAST (NIST)
1E-6 photon energy
10 100 1000 104 105 / eV
Fig. 10.16 Photoionization cross section of argon (Ar) in different energy regimes compari-
son of experiment and theory adapted from several authors. S AHA (1989): (a) partial cross sec-
tion for the 3s ionization above-threshold, showing a typical C OOPER minimum with MCHF
theory (full and dashed lines). S UZUKI and S AITO (2005): (b) total cross section up to 200 eV
(experiment only), (c) experiments from L edge up to 500 eV, (c ) blow up close to L edge,
(d) 500 eV to 1150 eV. Z HENG et al. (2006): (e) near K edge with high and (e ) very high res-
olution, (f) K edge over a broader energy range. Derived from the FFAST data bank (C HANTLER
et al. 2005): (g) overview for whole energy range
necessary. The data shown in Fig. 10.16(a) show the energy dependence of the pro-
cess (10.53) directly above its energetic threshold. Note that this particular process
contributes only very little to the total photoionization cross section.
However, this particular partial cross section shows in an impressive manner how
the photoionization cross section starts with the usual jump to a finite value a thresh-
old, drops down rapidly to nearly zero just above 40 eV photon energy (ca. 12 eV
electron energy), and finally rises again. This so called C OOPER minimum is not un-
usual in the photoionization of complex atoms above ionization threshold. It can be
understood easily from the theoretical background: The photoionization cross sec-
tion (5.88) contains two matrix elements (5.89) that essentially describe the overlap
between bound state and continuum wave functions (weighted by the distance r).
Depending on the positions of the nodes in the wave functions they may be positive
ore negative what can change with the energy of the continuum electron. Thus,
it may happen, that the two matrix elements in (5.88) just compensate for one par-
ticular energy. Alternatively one dominant matrix element may change from a pos-
itive to a negative value, passing necessarily zero at some energy which then leads
to zero contribution from this matrix element. In the present case things are even
a little bit more complex since these wave functions are, strictly speaking, multi-
electron wave functions comprising more than the ideal configuration according to
(10.53). Here the dominant matrix element is . . . 3s . . . 1 S|r|p 1 P0 where the con-
tribution from the 3s 2 3p 6 configuration is partially compensated by strong CI with
the 3s 2 3p 5 3d configuration. Correspondingly expensive is the multi configuration
H ARTREE -F OCK (MCHF) calculation shown in Fig. 10.16(a), which interprets the
experimental results rather well not differing much in its dipole lengths and veloc-
ity form which is a characteristic for any high quality calculation (see footnote 10
in Chap. 4).
Figure 10.16(b) shows the total photoionization cross section for the energy
range between 40 and 100 eV, which is dominated by ionization of the six 3p elec-
trons. Here too one recognizes a C OOPER minimum, this time at 50 eV, followed
by a maximum at ca. 80 eV so to say its counter part. In Fig. 10.16(c) and (c ) the
details in the vicinity of the L edge are illustrated: the ionization thresholds (termi-
nology see Fig. 10.11) for LIII , LII and LI at 248.4, 250.6 and 326.2 eV, respectively.
The contribution of the 2s electrons (LI ) is very small similar as in the M shell.
Below the LIII and LII thresholds one sees in (c ) indications of autoionizing reso-
nances, which we got to know in Sect. 7.6.2.
Figure 10.16(d) shows the rather boring decrease above the L edge, but still sig-
nificantly below the K edge. The latter is finally reached at 3205.9 eV as shown in
detail in Fig. 10.16(e), (e ) and (f). In (e) one recognizes the double excitation of K
and M shell 1s3p 4p 2 (again an autoionizing state). The most pronounced struc-
ture in this energy range is doubtless the excitation of RYDBERG states 1s np for
(n 4), shortly below the K edge, shown on an expanded scale in (e ).
Figure 10.16(g), finally, gives a full survey over the whole energy range from
threshold up to 100 keV in a loglog plot, extracted from the B ERGER et al. (2010)
data bank. These data do, of course, only reproduce the rough tendencies.
We cannot end this section without mentioning that in addition many measure-
ments of the anisotropy parameter have been reported (see Sect. 5.5.3), containing
additional information we refrain, however, from reproducing such data here.
Section summary
Absorption and emission of X-ray radiation from (and to) inner atomic shells
is governed by the PAULI principle which forbids transitions into completely
filled electron shells.
Absorption (at not too high photon energy, ) is mostly due to photoioniza-
tion. The absorption cross section as a function of shows a typical edge
structure displaying the ionization potentials of different shells. Beyond these,
the cross section decreases rapidly (Figs. 10.14 and 10.15).
At intermediate photon energies, photoionization cross sections of complex
atoms show a rich structure. E.g. so called C OOPER minima may occur,
shortly above an ionization threshold. They are attributed to cancellation of
two radial matrix elements for bound-free transitions, or to zero passage of
one dominant matrix element.
The absorption
edges WI are reasonably well described by M OSLEYs law,
predicting WI Z qs .
Based on this, a very rough estimate for the energy of K radiation is obtained
from (10.51).
At > 100 keV, the cross section starts to be dominated by C OMPTON scat-
tering, while for > 1 MeV pair production sets in.
Transitions of electrons from higher to lower levels can only occur into elec-
tron holes. This leads to characteristic X-ray emission spectra, similar to those
from quasi-one-electron spectra (alkali atoms).
Alternatively, the filling of inner shell holes may be accompanied by emission
of AUGER electrons which carry the excess energy.
10.6 Sources for X-Rays
10.6.1 X-Ray Tubes
The classical device for generating X-rays is the X-ray tube (originally RNTGEN-
Rhre, named after its inventor RNTGEN 1901, who received the N OBEL prize for
the discovery of X-rays). Even today this very simple method is still widely used.
A typical example of a state-of-the-art, small X-ray tube for analytical purposes is
illustrated by a photo of the device as well as by a schematic in Fig. 10.17. An elec-
tron beam generated by a hot cathode wire an a negatively biased W EHNELT cylin-
der hits a metal anode. Inside the metal it is strongly decelerated in the electric field
of atomic nuclei (preferably with high Z). This leads to X-ray bremsstrahlung
(the German verb for decelerating is bremsen). At the same time, a fraction of
the atoms in the anode is ionized by electron impact, which leads to hole formation
in the inner shells of the atoms. As a consequence emission of characteristic X-ray
radiation becomes possible.
Figure 10.18 shows characteristic spectra emitted from a Rhenium anode (Rh) at
different electron energies. The measured X-ray spectrum presented here has been
10.6 Sources for X-Rays 531
Fig. 10.17 Example of a

modern X-ray tube (photo
and schematic have been
kindly provided by
H ASCHKE and L ANGHOFF
2007). Top: Photo of a mini
X-ray tube with side exit of
the radiation. Bottom:
construction of this tube
schematic oil insulation glass body of the tube
cathode focal area anode
e-
filament
e- HV
WEHNELT cylinder
Be window screening
X-ray radiation
recorded at 90 detection angle to reduce the signal for the spectral analysis with a
crystal monochromator. One detects elastically as well as inelastically (C OMPTON-
Effekt) scattered X-ray radiation. Indicated in Fig. 10.18 are the literature values
for the K , K and L, emission lines for Rh as well as the energies expected
according to (1.75) for inelastically scattered photons. The weak signal at 8.02 keV
(CuK line) originates from a small impurity in the anode.
The spectrum clearly documents that the X-ray bremsspectrum cannot have a
higher energy than the generating electron beam at the anode, with Wkin = eU cor-
responding to the voltage U (HV in Fig. 10.18) applied to the anode. This implies
that the shortest wavelength emitted in the X-ray bremsspectrum follows the so
called rule of D UANE -H UNT:
hc
min = . (10.54)
eU
10.6.2 Synchrotron Radiation, Introduction
Today, for the spectroscopy with VUV, XUV and X-ray radiation electron storage
rings are available and intensively used at many places around the world, dedicated
to generate synchrotron radiation. In so called 3rd generation storage rings one uses
extremely well focussed, highly relativistic electrons. Typically, the electrons are
stored in a large, overall circular or elliptic structure in which they are guided by
deflection magnets (magnetic dipole fields). They are arranged with long straight
sections in between, where the electrons are focussed with different magnetic struc-
tures, mainly with quadrupole magnets, but other structures such as sextupole or
octupole magnets are also needed to control the chromaticity, i.e. the variation of
Rh K (inel.)
X-ray emiss. / arb. un.

Cu - K(inel.)
K (elast.)
40 keV
K (inelast.)
20 keV
L L K (elast.)
10 keV
0 5 10 15 20 25 30
/ keV
Fig. 10.18 X-ray bremsstrahlung and characteristic radiation (top) from the tube shown in
Fig. 10.17 with a rhenium (Rh) anode at three different electron energies (the spectra have been
kindly provided by H ASCHKE and L ANGHOFF 2007). The bremsstrahlungs continuum as well as
characteristic radiation (K , K and L, ) can be clearly recognized
focussing properties with the spread in electron momentum.3 Electromagnetic ra-

diation in a wide spectral range, so called synchrotron radiation (SR) is generated
where the electrons are accelerated, e.g. as they are deflected by the dipole magnets.
The overall energy of the electron beam is kept constant, compensating the radia-
tive losses by iterative re-acceleration in special microwave cavities somewhere in
the beam path. Typical data presented in the following refer (as far as not otherwise
mentioned) by way of example to BESSY II, the Berlin electron storage ring for
synchrotron radiation in Berlin-Adlershof.
In Fig. 10.19 the topography of BESSY II is sketched schematically. The ring
has a circumference of 240 m. In it electrons of a nominal energy of 1.7 GeV are
stored with a current of 100 mA to 400 mA. 32 dipole magnets keep the beam on
its ring trajectory. In between 16 about 4 m long straight sections are arranged into
which undulators or wigglers may be inserted. The orbital period of the electrons
amounts to ca. 0.8 s, and the electrons are kept in bunches at a temporal distance of
ca. 2 ns, each having a pulse duration typically below 20 ps. Under normal operating
conditions about 350 such bunches circle in the ring, each followed by a ca. 100 ns
gap. They are initially prepared and accelerated in a smaller synchrotron which is
indicated in Fig. 10.19 inside the ring. After injection into the storage ring electrons
are stored for several hours. The most recent mode of operation features continu-
ous refilling of the ring to compensate for lost electrons. Typically, some 50 user
stations, so called beamlines are installed for general use.
Key advantages of such synchrotron sources are their high brilliance and the
broad spectrum generated, typically from the terahertz region at the long wavelength
end down to soft or even hard X-ray radiation. By using suitable monochromators
3 The quality of the focussing is characterized by the so called emittance. That is the spatial exten-
sion of the electron beam multiplied by its divergence angle. As an example, BESSY II is charac-
terized by an emittance of (3 to 6) nm rad in horizontal and <0.1 nm rad in vertical direction.
target hall high frequency

frequ i jecction
inje
inj i target hall
electron gun
resona
nator
ators
rs
deflection
synchrotron magnets
circumference
ce
96m
undulatorr
deflection
magnets storage
g ringg
circumference 240 m
be 32 deflection magnets
am
line 16 straight sections
s
beamlines
nes
s
ne
mli
a mli bea
target hall be target hall
Fig. 10.19 Schematic of the topography in BESSY II. The electrons are pre-accelerated in a
microtron, further accelerated in a small synchrotron ring to the nominal energy, and then injected
into the storage ring. The radiation energy emitted is compensated by high frequency generators.
In between the deflection magnets, marked in red, are straight sections are arranged for insertion
devices and multipole focussing magnets. The beam lines exit tubes of radiation for the users
are shown here for clarity only in the lower half of the figure. Only one undulator is indicated
as an example for an insertion device (We thank Prof. JAESCHKE et al. 2007, for providing details
and sketches)
one may thus obtain intense, conveniently tunable, quasi-monochromatic radiation

of low divergence ideally suited for spectroscopic experiments of various kinds
as well as for X-ray diffraction experiments (see e.g. Fig. 1.19). Other, increas-
ingly important applications of synchrotron radiation are microscopy, holography
and lithography with X-ray radiation.
Brilliance is here not just a buzzword but a measurable characteristic of SR.
Eventually one is interested in the number of photons N emitted per unit time t and
area A (which is called photon flux , given in SI units [] = s1 m2 ) and per
solid angle ([] = sr, see Sect. A.4). This leads to the standard definition for the
radiance (see Table 1.6)
d3 N d
R= = , (10.55)
dtdAd d
with = W = hc/ = hc being the photon energy. The spectral radiance
)W = d ,
2
R (10.56)
ddW
i.e. it refers to radiance per unit energy interval (or per frequency, wavenumber or
wavelength). The synchrotron radiation community uses a special quantity called
spectral brilliance
3
B= , (10.57)
W/W
which refers to the relative bandwidth dW/W = d/ = d/. Usually, brilliance
is measured in units
1photon/ s
= 1 Sch. (10.58)
mm2 mrad2 0.1 % bandwidth
We abbreviate this somewhat complicated unit4 by Sch in honour of Julian

S CHWINGER (N OBEL prize in physics 1965 together with T OMONAGA and F EYN -
MAN ). As we shall see, S CHWINGER has provided the theoretical background for
this kind of radiation.
Figure 10.20(a) gives a survey of spectral brilliance for a variety of synchrotron
radiation sources of the 3rd and 4th generation, including insertion devices and un-
dulators as well as for the new Free Electron Laser sources (FEL) some already
operating, some under construction (dashed lines). For comparison the brilliance
achievable at dipole deflection magnets is shown at the bottom in Fig. 10.20(a).
Even this radiation is with ca. 3 1014 Sch still by orders of magnitude higher
than e.g. bright sun light on earth (maximum in the visible spectral range ca.
3 1011 Sch), standard X-ray tubes with 107 to 1010 Sch and other conventional
laboratory sources with the exception of some laser systems.
Brilliance does not only describe how many photons are available per solid an-
gle, area and wavelength interval. It also defines how well they can be focussed:
the less extended a light source is, and the smaller its emittance angle, the bet-
ter it may be focussed onto a target. As detailed in Appendix B.2, Vol. 2, ac-
cording to the H ELMHOLTZ -L AGRANGE relation (for paraxial beams equivalent
to the A BBE sine law) the product of n2 S is a constant when imaging light,
electron or ion beams. With the index of refraction n 1 being a constant in
4 This unit may be somewhat confusing. What it implies is simply that B gives the number of
photons emitted per second into a solid angle = 1 mrad2 from a source area S = 1 mm2
into an energy interval W = 103 W . Thus, into an arbitrary solid angle and energy interval the
source emits
photons S W
=B 3 .
s mm2 mrad2 10 W
Expressed in terms of standard spectral radiance (10.56) one would write
photons R )W
= S W,
s
so that with (10.56), (10.57) and W =
d2 )W .
B = 103 W mrad2 mm2 = 103 R
ddW
wavelength / nm
100 1 0.01 1000 100 10 1 0.1
FLASH DESY
Y XFEL
fs-Ti:Sapph DESY 1012
1033 (seeded) XFEL
LCSL
1031 LCSL
109
peak brillance / Sch (see text)
DESY
1029 DESY FLASH
FLASH
peak power / W
1027 6
plasma lasers 10
PETRA III HHG
1025 20 m ID SPring-8
X-ray Und.
1023 XUV Und. dye APS
exzimer
103
BESSY laser
Rubin
U49
U-49 U125
1021 U-125
ALS BESSY
1019 U5.0 100
1017 (a) (b)

BESSY
1015 dipole deflecton magnet
1013
101 102 103 104 105 106 1E-3 0.01 0.1 1 10
photon enenergy / eV
Fig. 10.20 (a) Brilliance for various synchrotron radiation sources of the third generation (in-
cluding insertion devices ID and specifically undulators U) and for Free Electron Lasers (FEL) as
a function of photon energy. (b) comparison of the peak power in these sources with a variety of
laser systems
the present case (and assuming that no photons are lost), B is thus a constant
too. Synchrotron radiation from third generation sources typically emerges from
an area of only some 102 mm2 and is extremely well collimated. Compared to
conventional light sources, SR sources are superior by many orders of magni-
tude.
Note, however, that the angular divergence of laser sources is usually even better
than SR. Also, from an experimental point of view the peak power of a source is
often more relevant than its brilliance for obtaining an experimental signal. Fig-
ure 10.20(b) thus compares typical laser systems with representative SR sources
from this point of view. Peak power determines the total number of photons avail-
able for an experiment per unit time. Also the total number of photons, i.e. the
available measuring time may be crucial, so that even X-ray tubes are still in use
and allow long time experiments, in particular for such experiments where SR
peak power cannot be exploited e.g. due to damaging side effects on delicate
targets.
(a) p z' (b) p z

k
k
. .
p p
= kx / kz = 1 / 2
x' LORENTZ transformation x
kz = 2 kz'
kz'
kx =
(c) kx' = kz' (d) kx' = kz'
k'
A
(e)

2
} time window for observation
2 t = 2sin( )/c
p
2 2 te = 2 /v
Fig. 10.21 For understanding the generation of synchrotron radiation (SR): polar plots for the
angular distribution of (a) the classical radiation characteristic from an electron on a circular orbit,
observed in its rest frame and (b) its L ORENTZ transform; (c) components of the wave vector
in a frame moving with the electron and (d) the same in the laboratory system; (e) observation
window for the radiation from this electron
Nevertheless, in addition to brilliance, excellent tunability, broad wavelength

range, as well as relatively good accessibility and user friendliness of the publicly
funded facilities make SR to an unsurpassed radiation source for a wide range of
applications. This is specifically true for the VUV, XUV and soft as well as hard
X-ray region, where SR is clearly superior even to laser sources.
10.6.3 Synchrotron Radiation, Quantitative Relations
Starting point for the strictly relativistic derivation of the intensity and angular dis-
tribution of SR is (4.32) which gives the radiation characteristic in the moving rest
frame of the electron. This has to be L ORENTZ transformed into the laboratory sys-
tem. Consequently a highly collimated radiation characteristic tangential to the orbit
of the electrons is emitted. The not completely trivial theory has been developed for
the first time in a famous paper by S CHWINGER (1949). The formula derived there
for the emission characteristic is so precise that synchrotron radiation is used to-
day as a calibration standard in radiation metrology. Without going into details we
just present the key ingredients by some hand waving arguments, summarized in
Fig. 10.21.
We apply the terminology introduced in Sect. 1.2. Dashed quantities refer to

the electron rest frame, un-dashed quantities to the laboratory system. Let the to-
tal energy of the electron be We with = We /(me c2 ) = (1 2 )1/2 , the energy
of the photons emitted is and (corresponding to wavenumbers k and k),
respectively. In the field B of a deflection magnet the highly relativistic electrons
( = 1) are accelerated onto a circular orbit with an instantaneous radius of cur-
vature me c/(eB) according to (1.109). In its moving rest frame x y z (one
chooses z p) the electron emits, as indicated in Fig. 10.21(a), the characteristic
radiation of a H ERTZ dipole, with an angular characteristics cos2 according to
(4.32).
If one transforms this radiation pattern into the laboratory system, its frequency
in z-direction and its wave vector component kz are dramatically magnified, while
the x-component kx = kx remains constant. Thus, the relativistic D OPPLER shift
(1.29) leads to a collimation of the radiation as indicated in Fig. 10.21(b) into the
direction of the electron trajectory a key characteristic of synchrotron radiation.
For 1 one obtains kz = 2 kz according to (1.31), which corresponds to a very
small (full) divergence angle
kx 1
2 2 = (10.59)
kz
for the radiation emitted, as one reads from Fig. 10.21(c) and (d). To illustrate this
by numbers: at BESSY II with its electron energy of 1.7 GeV one has 3300
and the full divergence angle of the radiation at a deflection magnet is 2 1 (in
words: one arc minute!).5
From (10.59) one may even obtain a simple estimate of the bandwidth. One has
to realize that an observer at large distance sees the radiation from a single electron
only for a very short time. As sketched in Fig. 10.21(e) this time represents the time
the electron needs to pass an arc segment 2. The electron needs a time te =
2/v, the light the time 2t = sin(2 )/c 2/c. An observer, looking
at the electron tangentially from a far distance, sees the light emitted from that
electron smeared out over a time

1 1 1 1 1
t 2 = = 3 ,
v c v c c 2 c
where we have used (1.15) for 1.6 Applying the uncertainty relation (1.125)
for time and energy one estimates from this the characteristic bandwidth of SR to
5 Of course, the emission angle effective for the experiment also depends crucially on the experi-
mental setup, which includes a certain length of the electron orbit. The angle given here thus
only holds for the vertical divergence perpendicular to the plane of the ring. In horizontal direction
the experimental apertures limit the SR beam.
6 Clearly,this time refers to the light emitted from a single electron and has nothing to do with
the actually measurable duration of the SR light pulses from bunches of electrons. Typical pulse
duration of modern synchrotrons are several ps.
be W /t = 2 3 c/. The exact calculation confirms the order of magnitude

from this crude estimate: A so called critical energy
3c 3 3 2 eB
Wc = c = = , (10.60)
2 2me
enters as key parameter into the S CHWINGER formula for the intensity distribution.
The corresponding critical wavelength is
4 4me c
c = = . (10.61)
3 3 3 2 eB
To characterize the angular distribution of the radiation, instead of referring to
polar ( ) and azimuthal angle () one uses for practical purpose two orthogonal
angles (in the ring plane) and (normal to it). In the literature the original for-
mula (Eq. (II.32) in the paper of S CHWINGER 1949) for the emitted photon flux is
rewritten in a variety of different ways.
A rather concise form is the following:7

d3 3 I /e 2 2 2
= 2
1 + x2 (10.62)
ddd/ 4 S c

x2
K2/32 [ ] + K 2
[ ] .
1 + x 2 1/3
Here I is the electron current in the ring and S the emitting area. The unit Sch
(S CHWINGER), introduced above, is obviously compatible with this formula writ-
ten essentially dimensionless (apart from cm2 s1 ). Kn is the generalized second-
order B ESSEL function, implemented in standard mathematical programs, and the
variables x and are defined by
3/2
x = and = 1 + x2 . (10.63)
2Wc
_ _2 _ _2
As it turns out, the two components in (10.62), K2/3 (/2) and x 2 K1/3 (/2)
(1 + x ), respectively, represent distributions of radiation intensity for two different
2
polarizations: the first one describes light linearly polarized parallel to the plane of
the ring, while the second term corresponds to a polarization normal to the ring
which is nonvanishing only outside the ring plane and has a phase shift of /4
in respect of that in the plane. Superposition of both components thus makes SR
7 For readers who want to follow these reformulations: S CHWINGER uses the esu system, and
his formula (II.32) gives the energy which an electron emits along the orbit over = 2 . For
the determination of the brilliance it has to be divided by 2 and by the photon energy. Now,
2/c = te is the period time for an electron on a circular orbit and, on the other hand, Ne =
2I /(ec) gives the number of electrons in the ring at a current I . With an emitting area S and
the fine structure constant = e2 /(40 c) one obtains (10.62).
light emitted above or below the ring plane elliptically polarized: it is a unique
potential of SR to provide short wavelength, tuneable radiation with elliptic or even
circular polarization. Typically such radiation is offered at dedicated beam lines. For
compact description of the radiation characteristic one usually presents the radiation
with polarization parallel to the ring plane as a function of the photon energy ,
i.e. one sets = 0 and obtains as a convenient numerical relation:
d3 I/A
= 1.326 1013 (We / GeV)2 Shor ( ) Sch. (10.64)
ddd/ S/ mm2
Alternatively one integrates over the angle:
d2 I/A
= 2.4577 1013 (We / GeV)Sint ( ) mrad Sch. (10.65)
dd/ S/ mm2
While from (10.62) one obtains Shor ( ) directly for x = = 0 and = W/2Wc ,
one finds Sint ( ) after some reformulation by integration:
Shor ( ) = 4 2 K2/3
2 ( )

(10.66)
Sint ( ) = 2 K5/3 (z)dz.

Both functions are illustrated in Fig. 10.22(a). One sees that the maximum bril-
liance of the SR is reached just below the critical energy, at = W/2Wc 1. For
BESSY II e.g., Wc = 2.5 keV, and the useful spectral range at a dipole deflection
magnet extends up to about 10 keV.
Finally, by integrating (10.62) over all photon energies , one obtains the over-
all angular distribution in respect of (angle normal to the ring plane) which is
plotted in Fig. 10.22(b):

d
2 5/2 5 x2
1+x 1+ . (10.67)
d 7 1 + x2
The FWHM angle is 1.3/ , in fair agreement with the rough estimate 2 = 1/
according to (10.59).
The total radiation emitted may after appropriate rewriting be obtained from
(4.32) or by integration of (10.65) over all frequencies:

e2 2 dp 2 e2 c 4 e4 We2 B 2
P= = = .
60 m2e c3 dt 60 2 60 m4e c3
This expression is obviously the relativistic analogue to (4.32), where in the last
steps (1.110) and = We /(me c2 ) has been inserted. Note the strong, inverse pro-
portionality on the 4th power of the radiating particles mass, here the electron.
Thus, the electron is the particle of choice, if one wants to generate synchrotron
radiation.
(a) (b)
S int ( ) 2.146 d
1/3
2.910 2/3
1.0
S hor ( ) d
1 0.8
0.1 0.6
1.339 1/2 e- F W H M ~_ 1.3/
0.4
0.01
0.2
0.001
0.0001 0.01 1.00 -2 0 2
= /2 Wc
Fig. 10.22 Characteristic energy and angular distribution of SR generated at a deflection magnet.
(a) Brilliance as a function of photon energy in the plane of the ring, Shor corresponding to (10.64)
and integrated over angle, Sint corresponding to (10.65). (b) Energy integrated intensity (10.67)
as a function of the angle above and below the plane of the ring, showing a characteristic angular
FWHM 1.3/
10.6.4 Undulators and Wigglers
An even more dramatic enhancement of brilliance and extension of the wavelength

range can be achieved by using so called insertion devices, specifically undula-
tors and wigglers. These are a sequence of periodically positioned dipole deflection
magnets, which are poled alternately in north-south and south-north direction. They
are mounted into the straight sections of the storage ring and force electrons onto
rapidly oscillating orbits which again leads to emission of radiation. Schematically
this is shown in Fig. 10.23(a) and illustrated in Fig. 10.23(b) by a photograph of an
undulator, type U49, implemented at BESSY II. One computes the radiation emitted
from the electron passing the undulator magnets quite analogously to deflection
magnets by L ORENTZ transformation from the electron rest frame into the labora-
tory. This again leads to a strongly forward peaked radiation characteristic. A rough
estimate for the wavelength is obtained as follows: the electron sees the periodic
distance u of the dipole magnets, due to L ORENTZ contraction (1.27) shortened
to = u / . In the emission process the relativistic D OPPLER shift (1.29) again
reduces this wavelength to
u u
= (1 cos ) u 1 + 2 /2 2 1 + 2 2 ,
2
where cos has been expanded for small emission angles , and (1.15) was inserted
for
= 1.
The exact calculation leads to the undulator equationfor the so called resonance
wavelength:

u K2
r = 2 1 + + 2 2 . (10.68)
2 2
(a) u 1
c =
N S N S N S u
e-
N S N S N S
1/ u (c)
intensity

(b)
Fig. 10.23 (a) Scheme of an undulator with Nu (here =3) periods of aligned, alternately poled
magnetic fields. They force the electrononto an oscillatory trajectory. The opening angle of the
central cone is narrowed by a factor 1/ Nu as compared to a deflection magnet. (b) Photograph
of a U49 undulator for X-ray spectroscopy at BESSY II. (c) Typical spectrum from an undulator
in the oscillation plane with odd harmonics. (We thank Eberhard JAESCHKE et al. 2007, for the
photograph)
Here K is the dimensionless magnetic deflection or undulator parameter
eBu B u
K= = 0.93373 (10.69)
2me c T cm
with B being the magnetic field in the undulator magnets. The resonance wavelength
r thus depends on the undulator wavelength u , the beam energy , the undulator
parameter K, and, remarkably, it also depends on the emission angle . A detailed
analysis shows that one has to distinguish two quite different situations: undulators
with K 1 and wigglers for which K 1.
Essential for an undulator (K 1) is, that the oscillation amplitudes of the elec-
trons remain small, so that the amplitudes of the radiation emitted at each of the Nu
magnet pairs can interfere constructively. This induces partially coherent radiation
with a
substantially reduced bandwidth / = / 1/Nu
around the resonance frequency r = c/(2r ) similarly to multi-beam interfer-

ence at a grating. This is accompanied by a corresponding enhancement of the in-
tensity and a

reduction of the angular spread c 1/ Nu .
At the same time, higher harmonics n = nr of the resonance frequency may be

emitted, as indicated in Fig. 10.23(c). The computation shows that in forward di-
rection only odd harmonics n = 2j + 1 are emitted, with the fundamental j = 0
dominating. At synchrotron radiation sources of the third or forth generation, undu-
lators are very popular for intense short wavelength applications due to their superior
performance (see Fig. 10.20).
In contrast, for K 1 the emission cones from the individual undulations be-
come wider and do no longer superpose coherently. Such a situation is then called a
wiggler. A key feature is the generation of very high harmonics with broader band-
width. With increasing K they dominate the overall wavelengths spectrum which
then becomes a broad continuum. In principle, this spectrum does not differ much
from that at a bending magnet except for the fact that it is much more intense and
is shifted to much higher photon energies than achievable at bending magnets with
a given electron energy. Wigglers thus provide convenient sources for soft and hard
X-rays.
10.6.5 Free Electron Laser (FEL)
In a long undulator with many periods of magnets and suitable design the coherently
superposed radiation from all the magnets may become so intense that it acts back
onto the electron beam which has generated it: a bunch structure emerges, and the
electrons ride so to say on the electromagnetic wave. This implies the feedback and
amplification of the radiation which is needed to achieve complete coherence and
which is characteristic for a laser. Such devices are called free electron laser (FEL).
The setup is very similar to that for an undulator shown in Fig. 10.23, however, with
many magnets and typically a highly relativistic electron beam designed and ded-
icated to this purpose. Such a facility is, in principle, a single user facility, even if
as rule precautions are taken that several users may work quasi in parallel at such
a system (i.e. intermittently). The first systems which have been working very suc-
cessfully for many years by now had been designed for the infrared spectral range.
Presently, worldwide a number of FELs are under construction or already in opera-
tion. They generate intense soft or even hard X-ray short pulses (some fs). A variety
of techniques are applied, one of these being the SASE principle (Self-Amplified
Spontaneous Emission), which was used e.g. at DESY for the VUV laser FLASH
(Free Electron LASer in Hamburg) and will also be implemented at the XFEL (X-
ray Free Electron Laser) there. More efficient, but also more delicate, are concepts
where the FEL action is seeded initially by an already coherent, albeit weak, laser
pulse which may be generated from high harmonics (HHG) of a Titanium-sapphire
laser, as discussed in Sect. 8.5.6. Such concepts are planned, under construction or
even already in operation at all FEL facilities worldwide.
A look at Fig. 10.20 reveals that with these new sources at methodological quan-
tum jump for the interaction of light with matter of various kinds is under way. Not
only is the brilliance at these facilities many orders of magnitude higher than any-
thing one may have dreamed of some years ago. Perhaps even more important is the
temporal structure of such sources, which are designed for coherent femtosecond
pulses in the soft and hard X-ray radiation. They will allow for a new generation
of experiments which we may look forward to with great expectations. To mention
just one out of a host of dream experiments: it should become possible to generate a
complete X-ray diffraction image from a single, large biomolecule during one sin-
gle laser pulse and thus one would circumvent many efforts and imponderabilities
which arise from the present necessity to crystallize such objects.
10.6.6 Relativistic THOMSON Scattering
Without entering into great details we just mention this interesting and timely sub-
ject, following the introduction to coherent, elastic light scattering in Sect. 8.4.5. The
basic idea is quite simple: if one collides a laser beam (photon energy L ) with a
highly relativistic electron beam ( = We /me c2 1) the radiation is exposed to the
relativistic D OPPLER shift twice on the way in as well as upon reemission. Thus,
for a head on interaction of laser and electron beam and observation in the direction
of the electron beam, the emitted radiation has, according to (1.31), an energy
X 4L 2 . (10.70)
One easily verifies that even at moderate electron energies multi keV radiation may
be obtained in such a setup, e.g. from a 50 MeV electron beam, Titanium-sapphire
laser radiation (800 nm) is scattered as 60 keV X-ray photons. The problem with
this concept as a short pulsed X-ray source is of course the low magnitude of the
T HOMSON cross section. However, if intense laser fields are used, nonlinear effects
come into play (nonlinear, relativistic T HOMSON scattering), essentially related to
the magnetic field component of the radiation. The topic is subject to state-of-the-art
research (see e.g. C HUNG et al. 2011, and references there).
10.6.7 Laser Based X-Ray Sources
As already indicated these big, new facilities will provide only limited amount of
measuring time, and many experiments stand already in line waiting for their turn.
Thus, it is worthwhile to search for alternatives, which do not necessarily offer the
ultimate specifications of the new XFELs, but are more readily accessible. One class
of such sources are laser based X-ray sources. We specifically mention such setups
where a strongly focussed, very short and intense Titanium-sapphire laser pulse
generates a hot plasma of elements with high Z. The electrons thus generated may
induce bremsstrahlung or characteristic X-ray radiation just as in an X-ray tube
with the essential difference that in this cases an X-ray pulse is generated with a
duration on the order of some 100 fs. This allows one to study dynamical processes.
One example of such a source is shown in Fig. 10.24(a). Its essential ingredients are
a Cu target which is locally ionized by an intense laser pulse (5 mJ, 50 fs, 800 nm),
focussed onto a few m2 . Since the target is vaporized during the laser pulse, one has
to use a moving band-target. A second, this time transparent, rapidly moving tape
protects the laser focussing optics from all debris particles which emerge from the
Cu target. As illustrated by the measured X-ray spectrum, shown in Fig. 10.24(b),
(a) titanium sapphire laser (b)

1 kHz, 5 W, 45 fs @ 800 nm
K
count rate / arb. units

K
tranparent
protection Cu band / keV
against debris target 2 4 6 8 10 12
Fig. 10.24 Laser based X-ray source for short pulses according to Z HAVORONKOV et al. (2005).
(a) Scheme of an experimental setup with a moving Cu tape source, and moving transparent tape
to protect the optics for focussing the Titanium-sapphire laser pump beam. (b) Measured charac-
teristic X-ray radiation spectrum
and demonstrated by time resolved X-ray diffraction experiments, one obtains a

sufficiently intense X-ray pulse to produce a measurable signal for studying the
ultrafast dynamics of non-equilibrium states of matter. The further development will
certainly provide interesting new applications to which one may look forward
especially in the context of the competition from the emerging new XFEL facilities.

BESSY: Berlin Electron Strorage ring for Synchrotron Radiation, Germanys
third generation synchrotron radiation source in Berlin-Adlerhof.
rations in atomic and molecular structure calculations, using linear superposition
Sect. 6.6, Vol. 2 and Sect. 5.5.2, respectively).
FEL: free electron laser, laser like radiation source using amplification in a spa-
tially oscillating, relativistic electron beam, (see Sect. 10.6.5).
FLASH: Free Electron LASer in Hamburg.

HF: H ARTREE -F OCK, method (approximation) for solving a multi-electron
S CHRDINGER equation, including exchange interaction.
HHG: High harmonic generation, in intense laser fields.
LDA: Local density approximation, simplest version of density functional theory.
LS: L, for orbital and S for spin angular momenta, unfortunately this is used in
two opposite contexts: (i) LS interaction characterizes the spin-orbit interaction
energy, while (ii) LS coupling denotes an angular momentum coupling scheme
for multi-electron systems where all orbital angular momenta and all spin an-
gular momenta of all electrons are first separately coupled together to L and S,
respectively, and finally L and S are coupled to J .
LSDA: Local spin density approximation, similar to LDA but for electrons with
one spin orientation only.
MCHF: Multi configuration H ARTREE -F OCK, method to determine wave func-
tions for multi-electron systems (see Sect. 10.5.4).
MP2: M LLER -P LESSET correction of 2nd order, perturbative approach to cor-
rect HF energies for contributions from non-spherical repulsive potentials.
MP3: M LLER -P LESSET correction of 3rd order, perturbative approach to cor-
MP4: M LLER -P LESSET correction of 4th order, perturbative approach to cor-
interaction.
RHF: Restricted H ARTREE -F OCK, assuming all spatial wave functions in a given
closed shell to be equal when computing atomic wave functions.
SASE: Self-Amplified Spontaneous Emission, scheme for FEL operation.
SCF: Self-consistent field, method for solving coupled integro-differential equa-
tions iteratively.
index.html.
SR: Synchrotron radiation, electronmagnetic radiation in a broad range of wave-
lengths, generated by relativistic electrons on circular orbits.
TDDFT: Time dependent density functional theory, a modern variety of DFT,
allowing also for excited state calculations.
UHF: Unrestricted H ARTREE -F OCK, allowing different spatial wave functions
for each orbital when computing atomic orbitals (specifically in closed shells).

VUV: Vacuum ultraviolet, spectral range of electromagnetic radiation. part of the
21348 (2007).
XFEL: X-ray free electron laser, as FEL but specifically designed to generate
coherent, short pulse X-ray laser like radiation, for details see Sect. 10.6.5.
40 nm.
References
ATTWOOD , D.: 2007. Soft X-rays and Extreme Ultraviolet Radiation, Principles and Applications.
Cambridge, UK: Cambridge University Press.
B ERGER , M. J., J. H. H UBBELL, S. M. S ELTZER, J.C HANG, J. S. C OURSEY, R. S UKUMAR,
D. S. Z UCKER and K. O LSEN: 2010. XCOM: photon cross sections database (version 1.5),
NIST. http://physics.nist.gov/xcom, accessed: 8 Jan 2014.
C ASIDA , M. E.: 2009. Time-dependent density-functional theory for molecules and molecular
solids. J. Mol. Struct., Theochem, 914, 318.
C HUNG , S. Y., H. J. L EE, K. L EE and D. E. K IM: 2011. Generation of a few femtosecond keV
X-ray pulse via interaction of a tightly focused laser copropagating with a relativistic electron
bunch. Phys. Rev. Spec. Top., Accel. Beams, 14, 060705.
H ASCHKE , M. and N. L ANGHOFF: 2007. Mini power X-ray tube and spectrum private
communication, see also: IfG Institute for Scientific Instruments GmbH. http://www.ifg-
adlershof.de/, accessed: 8 Jan 2014.
JAESCHKE , E., W. E BERHARD and M. S AUERBORN: 2007. Technical data, sketches, and pho-
tographs about BESSY II. We acknowledge grateful the material provided.
KOHN , W.: 1998. The N OBEL prize in chemistry: for his development of the density-functional
theory, Stockholm. http://nobelprize.org/nobel_prizes/chemistry/laureates/1998/.
L ATTER , R.: 1955. Atomic energy levels for the Thomas-Fermi and Thomas-Fermi-Dirac poten-
tial. Phys. Rev., 99, 510519.
M ARQUES , M. A. L. and E. K. U. G ROSS: 2004. Time-dependent density functional theory.
Annu. Rev. Phys. Chem., 55, 427455.
PARR , R. G. and W. YANG: 1989. Density Functional Theory of Atoms and Molecules. Interna-
tional Series of Monographs on Chemistry. New York, Oxford: Oxford University Press, 333
pages.
RNTGEN , W. C.: 1901. The N OBEL prize in physics: in recognition of the extraordinary ser-
vices He has rendered by the discovery of the remarkable rays subsequently named after him,
Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1901/.
S AHA , H. P.: 1989. Threshold behavior of the M-shell photoionization of argon. Phys. Rev. A,
39, 24562460.
References 547
S CHUMACHER , E.: 2011. FDAlin programme, computation of atomic orbitals (Windows and
Linux), Chemsoft, Bern. http://www.chemsoft.ch/qc/fda.htm, accessed: 5 Jan 2014.
S CHWINGER , J.: 1949. On the classical radiation of accelerated electrons. Phys. Rev., 75, 1912
1925.
S UZUKI , I. H. and N. S AITO: 2005. Total photoabsorption cross-section of Ar in the sub-keV
energy region. Radiat. Phys. Chem., 73, 16.
S ZABO , A. and N. S. O STLUND: 1996. Modern Quantum Chemistry. Dover, first revised edn.
Z HAVORONKOV , N., Y. G RITSAI, M. BARGHEER, M. W RNER, T. E LSSSER, F. Z AMPONI,
I. U SCHMANN and E. F RSTER: 2005. Microfocus Cu K source for femtosecond X-ray
science. Opt. Lett., 30, 17371739.
Z HENG , L., M. Q. C UI, Y. D. Z HAO, J. Z HAO and K. C HEN: 2006. Total photoionization cross-
sections of Ar and Xe in the energy range of 2.16.0 keV. J. Electron Spectrosc., 152, 143147.
Z WILLINGER , D.: 1997. Handbook of Differential Equations. Boston, MA: Academic Press, 3rd.
edn.
Appendices
Overview
These appendices offer a collection of concepts and relations which belong
to the standard mathematical tools in atomic, molecular and optical physics.
On many occasions we refer to these in the main text. We do not attempt here
to derive them in a closed, axiomatic manner rather a compact summary of
useful and important formulas and equations is presented. Usually they are
accompanied by some hints about their origins and connections to make them
plausible for the reader without requesting the substantial efforts needed to
go into the depth of derivations or proofs. Literature references are given for
further information.
Appendix A offers a tabulated overview of current values for the most important
physical constants used in atomic, molecular and optical physics and in these
textbooks. Some remarks about systems of units and dimensional analysis are also
included.
Appendix B defines the angular momentum operators and introduces some ba-
sics about angular momentum algebra with 3j symbols (respectively C LEBSCH -
G ORDAN coefficients), 6j symbols (respectively R ACAH functions), and 9j sym-
bols.
Appendix C is concerned with the rules of evaluating and transforming matrix
elements of irreducible tensor operators. This includes the very important W IGNER -
E CKART theorem, matrix elements for electromagnetically induced transitions and
the reduction of coupling schemes as well as some useful geometrical relations.
Finally we give some directions for the evaluation of radial matrix elements.
Appendix D formulates the concept of parity and introduces as an alternative
to the usual complex spherical harmonics real basis states for angular momenta,
typically used in quantum chemistry. The concept is also extended to half integer
angular momenta.
Appendix E offers a recipe type of access to the important theme of coordinate
rotation for angular momenta and irreducible tensor operators.
550 Appendices
Appendix F introduces the concept of multipole expansions and gives the most
important relations derived from it. Two alternatives to define multipole tensor op-
erators are discussed: based on spatial coordinates and on combinations of angular
momentum operators.
Appendix G describes the concept of convolution and illustrates it for the most
important distribution functions.
Appendix I presents some essentials about F OURIER transformations with spe-
cial emphasis on electromagnetic fields.
Appendix H provides the formal basis for treating electromagnetically induced
transitions are presented in the main text somewhat heuristically. It also derives
the important T HOMAS -R EICHE -K UHN sum rule.
In Appendix J, finally, we discuss the behaviour and the proper normalization of
continuum wave functions. Both are important for a correct treatment of ionization
and scattering processes.
Constants, Units and Conversions
A
A.1 Fundamental Physical Constants and Units
Modern physics relies on the knowledge of a relatively large number of so called

fundamental constants which are measured with high precision in an ongoing,
cooperative international effort. Table A.1 gives a survey on the most important
ones, used throughout these books.
Table A.1 2010 CODATA recommended values of physical constants extracted from NIST
(2010), published by M OHR et al. (2012)
Constant Value Equation
Universal constants
speed of light in vacuum c 2.99792458 108 m s1 defined
magnetic constant 0 4 107 N A2 defined
107
electric constant 0 8.854187817 . . . 1012 F m1 1
0 c2
= 4 c2
characteristic vacuum Z0 376.730313461 . . . " c0 = 1
impedance 0 c
= 0 /0
Newtonian constant of G 6.67384(80) 1011 m3 kg1 s2
gravitation G/c 6.70837(80) 1039 (GeV /c2 )2
P LANCK constant h 6.62606957(29) 1034 J s
4.135667516(91) 1015 eV s
angular momentum quant 1.054571726(47) 1034 J s h/2
c 197.3269631(49) MeV fm

P LANCK mass mP 2.17651(13) 108 kg c/G
P LANCK length P 1.616199(97) 1035 m /(mP c)
P LANCK temperature TP 1.416833(85) 1032 K mP c2 /kB
P LANCK time tP 5.39106(32) 1044 s P /c
Electromagnetic constants
elementary charge e 1.602176565(35) 1019 C

Graduate Texts in Physics, DOI 10.1007/978-3-642-54322-7
552 A Constants, Units and Conversions
Table A.1 (Continued)

B OHR magneton B 927.400968(20) 1026 J T1 e/(2me )
5.7883818066(38) 105 eV T1
B / h 13.99624555(31) GHz T1
nuclear magneton N 5.05078353(11) 1027 J T1 e/(2mp )
3.1524512605(22) 108 eV T1
N / h
7.62259357(17) MHz T1
J OSEPHSON constant KJ 483597.870(11) 109 Hz V1 2e/ h
VON K LITZING constant RK 25812.8074434(84) h/e2
Atomic and nuclear constants
electron mass me 9.10938291(40) 1031 kg
me 5.4857990946(22) 104 u
electron g factora ge 2.00231930436153(53) e B
g factor anomaly ae 1.15965218073(28) 103 g/2 1
proton mass mp 1.672621777(74) 1027 kg
proton-electron mass ratio me /mp 1836.15267245(75)
neutron mass mn 1.674927351(74) 1027 kg
mass of He nucleus m 6.64465675(29) 1027 kg
unified atomic mass unit 1u 1.660538921(73) 1027 kg m(12 C)/12
atomic unit of energy Eh 27.21138505(60) eV e4 me /(40 )2
atomic unit of length a0 0.52917721092(17) 1010 m 40 2 /me e2
(B OHR radius) Eh a02 = 2 /me
RYDBERG constant R 10 973 731.568539(55) m1 Eh /(2hc)
atomic unit of time t0 2.418884326502(12) 1017 s /Eh
fine structure 7.2973525698(24) 103 e2 /(4
0 c)
constant 1/137.035999074(44) = Eh /me c2
new datab
1/137.035999037(91)
classical electron radiusc re 2.8179403267(27) 1015 m 2 a0
C OMPTON wavelength C 2.4263102389(16) 1012 m h/me c = 2a0
T HOMSON cross section e
0.6652458734(13) 1028 m2 (8/3) 4 a02
= (8/3)re2
Proton radius Rp 0.8775(51) 1015 m
B OLTZMANN constant kB 1.3806488(13) 1023 J K1
5 4
S TEFAN-B OLTZMANN B 5.670373(21) 108 W m2 K4 2 kB
15 h3 c2
constant
AVOGADRO constant NA 6.02214129(27) 1023 mol1
molar volumed Vm 22.413953(21) 103 m3 mol1
L OSCHMIDT constantd NL 2.6516462(24) 1025 m3 NA /Vm
a.u. of dipole moment ea0 8.47835326(19) 1030 C m
A.2 SI and Atomic Units 553
Table A.1 (Continued)

energy equivalents 1 eV 1.602176565(35) 1019 J e 1V
1J 6.24150934(14) 1018 eV 1 J /1 eV
1u 0.931494061(21) GeV 1 u c2 /1 eV
me 0.510998928(11) MeV me c2 /1 eV
1 cm1 1.239841930(27) 104 eV hc/1 eV
eV in a.u. 1 eV 3.674932379(81) 102 Eh 1 eV /2R hc
eV in chem. units 1 eV
96.4853364(48) kJ mol1 1 eV NA
1
23.061 kcal mol
eV in wavenumbers 1 eV 8 065.54429(19) cm1 1 eV / hc
eV in degree K 1 eV 1.1604519(11) 104 K 1 eV /kB
a As usual in the literature, g is defined positive, see (1.164), in contrast to NIST
b By B OUCHENDIRA et al. (2011)
c This is just a constructed quantity: at this distance the C OULOMB energy of two electrons is equal
to the rest energy me c2 of an electron
d In ideal gas at 273.15 K and 100 kPa
A.2 SI and Atomic Units
In general, throughout these text books we consequently use the SI units, meter
(m), kilogram (kg), second (s), ampere (A), kelvin (K), mole (mol) and candela
(cd) the international standard of units.
In addition to these 7 base units (NIST, 2000a, Table 1), we employ
standard prefixes (NIST, 2000b), e.g. cm, mg, fs,

SI derived units (NIST, 2000a, Tables 2 and 3) when useful or common cus-
tom, such as joule (J), volt (V) or degree Celsius ( C),
several of the accepted units outside the SI (NIST, 2000c), such as electron-
Volt, 1 eV = e V = 1.602 1019 J, or barn, 1 b = 100 fm2 = 1028 m2 . . . ,
and make one exception for the occasionally used (non-SI) unit electron-barn,
1 eb = e 1 b = 1.60221047 A s m2 .
We point out here that all values of constants and physical quantities given in the
table above and elsewhere in these text books refer to the currently valid definitions
of the SI. These definitions may possibly change in 2014 and be replaced by a new
set of definitions, completely based on fundamental constants (see Table A.1) as has
been decided by the CGPM (2011) (see also W IKIPEDIA CONTRIBUTORS, 2013).
Alternatively, we often use atomic units (a.u.) which are commonly used in
atomic and molecular physics. We consider these, however, to be just simplifying
abbreviations: one reformulates equations such, that for lengths, times and energies
only the combinations
r/a0 , t/t0 , and W/Eh , respectively, (A.1)
occur in the final expressions which then become otherwise free of units and
include only some dimensionless parameters, such as the fine structure constant .
It is often helpful to leave the equations in this form which simultaneously includes
a dimensional analysis. However, for simplicity of writing we occasionally omit the
respective units a0 , t0 , and Eh (e.g. in a S CHRDINGER equation), if no ambiguities
can arise. The respective quantities are then measured in a0 , t0 , and Eh . For more
details on a.u. see Sect. 1.8.3.
We emphasize, however, that it is rather misleading, to talk about setting =
e = me = 1 as occasionally formulated in (mostly theoretical) literature. The
same holds for relativistic expressions where the confusing notion that one sets
c = 1 is often used even in general texts, specifically on particle physics.
A.3 SI and GAUSS Units
The attentive reader will find a number of useful references to standard (mostly
older) and sometimes important literature, where G AUSS, electrostatic (esu) or elec-
tromagnetic (emu) units are still used (G AUSS units are equivalent to esu and emu
for electrostatic and electromagnetic units, respectively). Even in current literature,
specifically in theoretical papers, these systems are still used. Obviously, the seem-
ing advantage of writing a few equations in a simpler way than SI demands, ap-
pears so attractive to some authors that they renounce general compatibility. Thus,
Table A.2 communicates some conversion rules between the SI and the G AUSS
systems, following essentially the Appendix on units and dimensions in JACKSON
(1999).
A.4 Radian and Steradian
Radian (rad) and steradian (sr) are two somewhat conspicuous units which are fre-
quently used in AMO physics. They are, however, genuine SI derived units (NIST,
2000a). A brief clarification may be useful.
Radian and steradian may be visualized as shown in Fig. A.1. The radian (rad) is
defined in two dimensions for the plane angle by the length of an arc on a circle
of radius r, divided by this radius r:

= . (A.2)
r
With the full circumference of the circle being 2r, the full angle thus becomes
= 2 rad = 360 , so that 1 rad = 360 /(2) = 57.296 .
A.4 Radian and Steradian 555
Table A.2 Relations and conversions between the SI and the Gaussian system
Quantity/relation SI G AUSS

velocity of light c= 1/0 0 c
dielectric permittivity r 0 r
magnetic permeability r 0 r
electric field E, displacement D, D = 0 E +P = r 0 E D = E + 4P = r E
polarization P unit E esu = E SI 40 E SI = E esu / 40
magnetic H and B field, B = 0 (H + M) B = H + 4M = r H
magnetization M unit = r 0 H B SI = B emu 0 /4
B emu = B SI 4/0
M AXWELL equations D = D = 4
E = B
t E = 1c B
t
B = 0 B = 0
H =J + D
t H = 1c (J + D
t )
L ORENTZ force F = q(E + v B) F = q(E + B) v
c

charge, current, etc. q esu = q SI / 40 q SI =
40 q esu
unit 1C = 1As 1esu =1 dyn cm

=
2.9979 109 esu 40 109 N m
=
= 3.3356 1010 C
q1 q2 q1 q2
C OULOMB potential V (r) = 4 0 r V (r) = r
unit electric dipole moment Cm esu cm
D EBYE 1D = 3.33564 1030 C m 1D = 1018 esu cm
Polarizability esu = SI /(4 )
E SI = (4 ) esu
E
E 0 0 E
B OHR magneton B = e/(2me ) e/(2me c)
(a) (b)
rad =
m
Fig. A.1 On the definition of (a) radian and (b) steradian
In three dimensions the steradian (sr) is defined as the solid angle which char-
acterizes any surface area of the size S on a sphere of radius r. In other words, the
solid angle is given as the surface S on a sphere of radius r divided by the square
of this radius:
S
= . (A.3)
r2
For a cone with a full opening angle 2 as shown in Fig. A.1 we have

= d = 2 sin d = 2(1 cos ). (A.4)
0
With the surface of a sphere being 4r 2 the full solid angle is 4 sr, and 1 sr cor-
responds to an angle = arccos(1 1/(2)) = 0.57195 rad = 32.77 or equiva-
lently to a cone with a full angle = 2 = 65.54 .
Note that expressed in SI base units, both rad = m m1 and sr = m2 m2 have
the unit 1, i.e. are dimensionless. The symbols rad and sr may thus be used where
advantageous to distinguish the respective observable quantities from others . . . or
be omitted wherever the expressions are unambiguous.
For small angles (A.4) may be expanded as

lim = 2 + O 4 .
0
The divergence of well collimated beams (light beams, electron or ion beams
etc.) is often given in mrad2 , i.e. in terms of the full divergence angle = 2 mea-
sured in mrad. In the case of axially symmetry, i.e. for paraxial rays in a conical
beam, the solid angle of divergence is thus
2

= 106 sr.
4 mrad
Thus, a (full) divergence angle of 1 mrad corresponds to a solid angle of =
(/4)106 sr. Note, however, that for a rectangular beam with a small diver-
gence 2 , measured in mrad2 , the solid angle is just = 106 (/ mrad)2 sr.
A.5 Dimensional Analysis

In physics and technology dimensional analysis is often a useful tool to
make a first check on the validity of a physical expression,

guess, what the relation between several physical quantities might be,
obtain orders of magnitude estimates from a given relation.
Throughout these textbooks we frequently make use of these possibilities. Thus, we

summarize here the terminology used in this context:
1. Physical quantities, such as the electron mass typically have
a symbolic name, such as me ,

a numerical value, here 9.1094 1031 , and
a unit, here kg.

Together we write for this quantity: me = 9.1094 1031 kg.
2. In dimensional analysis we try to avoid specific units for mass, length, and so on,
since physical laws must be independent of the specific choice of units. Rather
we use a basic set of dimensions: mass (M), length (L), time (T), charge (Q),
and temperature (') all written in block letters without serifs. For most of our
purposes all other dimensions of physical quantities can be expressed as a prod-
uct of powers of these basic dimensions, e.g. the dimension of force is MLT2 ,
of energy ML2 T2 , of electric current QT1 , or with the thermal energy being
Wtherm = kB T , the B OLTZMANN constant kB has the dimension ML2 T2 '1
and so on.
3. However, for convenience or for clarity we shall sometimes use more letters or
abbreviations for the dimensions of physical quantities, e.g. for volume V = L3 ,
for energy Enrg = ML2 T2 or R = L3/2 for the dimension of a radial wave func-
tion.
4. The units of a physical quantity A or of a dimension A are represented by square
brackets, [A] and [A], respectively. For instance the unit of a potential V (r)
would be [V (r)] = eV and more generally we have [Enrg] = eV or [Enrg] = J,
both are acceptable.

emu: electromagnetic units, old system of unities, equivalent to the G AUSS sys-
tem for magnetic quantities (see Appendix A.3).
index.html.
References
B OUCHENDIRA, R., P. C LADE, S. G UELLATI -K HELIFA, F. N EZ and F. B IRABEN: 2011. New
determination of the fine structure constant and test of the quantum electrodynamics. Phys.
Rev. Lett., 106, 080801.
CGPM: 2011. Resolution 1 of the General Conference on Weights and Measures (2011) CGPM:
On the possible future revision of the International System of Units, the SI, Bureau Interna-
tional des Poid et Messure (BIPM). http://www.bipm.org/en/si/new_si/, accessed: 8 Jan 2014.
JACKSON, J. D.: 1999. Classical Electrodynamics. New York: John Wiley & sons, 3rd edn., 808
pages.
M OHR, P. J., B. N. TAYLOR and D. B. N EWELL: 2012. CODATA recommended values of the fun-
damental physical constants: 2010. Rev. Mod. Phys., 2013, 15271605. http://physics.nist.gov/
constants, accessed: 8 Jan 2014.
NIST: 2000a. Reference on constants, units, and uncertainties: SI base and derived units, NIST.
http://physics.nist.gov/cuu/Units/units.html, accessed: 8 Jan 2014.
NIST: 2000b. Reference on constants, units, and uncertainties: SI prefixes, NIST. http://physics.
nist.gov/cuu/Units/prefixes.html, accessed: 8 Jan 2014.
NIST: 2000c. Reference on constants, units, and uncertainties: Units outside the SI, NIST.
http://physics.nist.gov/cuu/Units/outside.html, accessed: 8 Jan 2014.
NIST: 2010. Reference on constants, units, and uncertainties, NIST. http://physics.nist.gov/
cuu/Constants/, accessed: 7 Jan 2014.
W IKIPEDIA CONTRIBUTORS: 2013. Proposed redefinition of SI base units, Wikipedia, The Free
Encyclopedia. http://en.wikipedia.org/wiki/Proposed_redefinition_of_SI_base_units, accessed:
8 Jan 2014.
Angular Momenta, 3j and 6j Symbols
B
We note here the most important relations for angular momenta in compact form
for reference and use throughout these books. Closely related are the 3j , 6j and 9j
symbols relevant for the coupling of angular momenta. They play an important role
in atomic and molecular physics. We follow here with one important exception
concerning the normalization of the spherical components of angular momenta
strictly the definitions of B RINK and S ATCHLER (1994) and E DMONDS (1996). Ex-
plicit expressions for the 3j , 6j and 9j symbols are somewhat complicated but well
documented in both references as well as in many textbooks on quantum mechan-
ics. For practical evaluation purposes one may use today readily accessible, compact
3j , 6j and 9j calculators (see e.g. S VEN G ATO R EDSUN, 2004). Also, they are of-
ten integrated in standard mathematical programmes. Thus, we refrain here from
extensive tabulations and restrict ourselves to definitions, symmetry properties, or-
thogonality relations, a few matrix elements and some important special cases (as
collected from several sources: E DMONDS, 1996; B RINK and S ATCHLER, 1994;
W EISSTEIN, 2004a,b).
B.1 Angular Momenta
B.1.1 General Denitions
The generalized angular momentum operator J is defined as vector operator with

three real, Hermitian components Jx , Jy and Jz with
Ji = Ji for i = x, y, z, (B.1)
which satisfy the commutation relations

2
[Jx , Jy ] = iJz , [Jz , Jx ] = iJy , [Jy , Jz ] = iJx and
J , Ji = 0,
or in compact form:

560 B Angular Momenta, 3j and 6j Symbols
2

J
J = i
J and J ,
J = 0. (B.2)
Alternatively one may construct J from three spherical components J and J0 (so
called helicity basis) which are an irreducible representation of the rotation group
of rank 1. They are derived from the real components:1
1 1
J+ = (Jx + iJy ), J = + (Jx iJy ) and J0 Jz . (B.3)
2 2
The inverse relations are:
1 i
Jx = (J+ J ), Jy = (J+ + J ). (B.4)
2 2
We note that these operators are constructed in complete analogy to the spherical
components (4.75) of the position operator r in the helicity basis which we have
introduced in the context of dipole transitions in Sect. 4.4.
The spherical components J are not Hermitian, with
J+ = J and J = J+ , (B.5)
and their commutation relations are derived from (B.2):
[J0 , J ] = J and [J+ , J ] = J0 , (B.6)

2 2

J , J = 0 and J , J0 = 0. (B.7)
The scalar product of two angular momenta

J 1 and
J 2 is given by

J1
J 2 = J1x J2x + J1y J2y + J1z J2z = J1+ J2 + J10 J20 J1 J2+ (B.8)
in accordance with the general definition of scalar products of tensor operators

(C.20). Hence,
2
J is expressed alternatively in the real or in the helicity basis:
+1

J = Jx2 + Jy2 + Jz2 = J+ J + J02 J J+ =
2
(1)q Jq Jq . (B.9)
q=1
The operators J+ and J play a key role in the construction of angular mo-
mentum states. Using their commutation relations (B.6)(B.7) one may construct
angular momentum states without using spherical harmonics (2.79)(2.83) which
1 Very often one finds in textbooks (even in B RINK and S ATCHLER, 1994) these ladder operators
defined as J = Jx iJy . These are, however, not spherical components of the
J vector operator
and lead to asymmetric commutation relations. With the present definition which is also used
e.g. by B LUM (2012), W EISSBLUTH (1978) J fulfill both functions: as ladder (or stepping)
operators and spherical components of J.
B.1 Angular Momenta 561
exclusively represent orbital angular momenta with integer quantum numbers . All
relevant algebra for angular momentum states may be derived in that way. Specifi-
cally, if |J M are eigenstates of Jz = J0 with eigenvalues M, it follows from (B.6)
that J+ |J M and J |J M are also eigenstates of J0 with eigenvalues (M + 1) and
(M 1), respectively:
J0 J |J M = J J0 |J M J |J M

(B.10)
= (M 1)J |J M .
With (B.7) the new states J |J M are also eigenfunctions of

2
J . The operators
J+ and J are thus raising and lowering operators (or ladder operators) for the
projection quantum number M. According to (B.10) M increases or decreases in
steps of one. One may show (here without proof) that M has a maximum and a
minimum value
J M J, (B.11)
and clearly, this maximum and minimum value M = J can only be connected by
M = J, J + 1, . . . , J 1, J if both M and J are either integer or half integer
numbers in complete accordance with the experimental observation: integer values
arise from orbital angular momenta, and half integer values reflect the involvement
of spin 1/2 particles. The states thus generated obey the standard eigenvalue equa-
tions and are of orthonormalized:
and J0 |J M = M|J M
2
J |J M = 2 J (J + 1)|J M (B.12)

J M J M = J J M M . (B.13)
With some algebra one may show that the only nonvanishing matrix elements of the
spherical angular momentum components J0 , J+ and J are
J M|J0 |J M = M and (B.14)

J M 1|J |J M = a(M) (B.15)
2

with a(M) = J (J + 1) M(M + 1). (B.16)
Equivalently one may write:

J |J M = a(M)|J M 1 and J0 |J M = M|J M . (B.17)
2
We finally note, that all relations communicated in this section for the general
angular momentum J are completely equivalent to those for the orbital angular mo-
mentum L which we have derived in Sect. 2.5 (one may of course also introduce
spherical components L ). In contrast to L,
the general angular momentum is not
explicitly referring to position space and angles, and is not restricted to integer val-
ues of J and M.
B.1.2 Orbital Angular Momenta Spherical Harmonics
One special kind of angular momenta are the orbital angular momenta introduced in
Sect. 2.5. As described there, their respective operators may be represented in terms
of spatial coordinates more precisely, the operator of the squared orbital angular
momentum L and the components of
2
L can be expressed as differential operators
with respect to the azimuthal and polar angles in polar coordinates. Of particular
importance are

2 1 1 2
L = Lx + Ly + Lz =
2 2 2 2
sin + 2 (B.18)
sin sin 2
z = i
and L (B.19)

with the spherical harmonics Y m (, ) as joint eigenfunctions. These operators
have the eigenvalues 2 ( + 1) and m, respectively, corresponding to (B.12). Here
= 0, 1, 2, . . . and m is any integer value between m .
Without proof we communicate here the spherical harmonics,

m 2 + 1 ( m)! m
Y m (, ) = (1) P (cos ) exp(im) (B.20)
4 ( + m)!
2

with d Y m (, )Y m (, ) sin d = m m . (B.21)
0 0
P m (x) are the associated L EGENDRE polynomials

m dm P (x) d +m 2
1 2 m
P m (x) = 1 x2 m
=
1 x +m
x 1 , (B.22)
dx 2 ! dx
which in turn are derived from the ordinary L EGENDRE polynomials
1 d 2
P (x) =
x 1 . (B.23)
2 ! dx
As a function of cos the first L EGENDRE polynomials are:
P0 (cos ) = 1
P1 (cos ) = cos
3 1
P2 (cos ) = cos2 (B.24)
2 2
5 3
P3 (cos ) = cos3 cos
2 2
1
P4 (cos ) = 35 cos4 30 cos2 + 3 .
8
B.1 Angular Momenta 563
We also note the orthogonality relation

4
P (cos )P (cos )d = .
4 2 + 1
With P0 (cos ) = 1 follows for all L EGENDRE polynomials with > 0

P (cos )d = 0. (B.25)
4
The definition (B.20) of the spherical harmonics Y m (, ) implies the so called

standard phase convention according to C ONDON and S HORTLEY (1951) with the
factor (1)m . One finds that the c.c. is

Y m (, ) = (1)m Y m (, ), (B.26)
and inversion at the origin (r r) leads to
Y m ( , + ) = (1) Y m (, ). (B.27)
The spherical harmonics are said to have positive or negative parity depending on
whether is even or odd, respectively (see detailed discussion in Appendix D). In
compact form one obtains well suited for computational use

(1) +m (2 + 1)( m)! d +m (sin )2
Y m (, ) = (sin )m exp(im)

2 ! 4( + m)! d(cos ) +m

(1) (2 + 1)( + m)! d m (sin )2
= (sin )m exp(im).
2 ! 4( m)! d(cos ) m
(B.28)
For simplicity of writing one often also uses renormalized spherical harmonics

4
C m (, ) = Y m (, ). (B.29)
2 + 1
We note the special values

Y m (0, ) = (2 + 1)/(4)m,0 and C m (0, ) = m,0 . (B.30)
Explicit expressions for the renormalized spherical harmonics up to = 3 are given

in Table B.1. A graphical illustration of their angular dependence is shown in
Fig. 2.6.
Table B.1 Spherical harmonics according to (B.28) and (B.29)

m Symbol 2 Renormalized
L Lz spherical harmonics
C k (, ) = 4/(2 + 1)Y m (, )
0 0 s 0 0 C00 = 1
1 0 p0 22 0 C10 = cos

1 p1 22 C11 = 12 sin ei
2 0 d0 62 0 C20 = 12 (3 cos2 1)

1 d1 62 C21 = 32 sin cos ei

2 d2 62 2 C22 = 38 sin2 ei2
3 0 f0 122 0 C30 = 12 (5 cos3 3 cos )

1 f1 122 C31 = 14 3ei sin (5 cos 2 1)

2 122 2 C32 = 15 2 2i
f2 8 cos sin e

3 f3 122 3 C33 = 4
5
sin3 e3i
B.2 Coupling of Two Angular Momenta: CLEBSCH-GORDAN

Coefcients and 3j Symbols
B.2.1 Denitions
When coupling two angular momenta

J 1 and
J 2 to generate a third one

J1 +
J2 =
J
the eigenstates |J M of the latter are constructed from the uncoupled states |J1 M1
and |J2 M2 as

|J1 J2 J M = |J1 M1 J2 M2 J1 M1 J2 M2 |J1 J2 J M . (B.31)
M1 M2

Formally, one may read the sum |J1 M1 J2 M2 J1 M1 J2 M2 | 1 simply as quan-
tum mechanical unit operator (see Sect. 2.3.1) written in front of |J1 J2 J M . J1 , J2
and J are the corresponding (integer or half integer) angular momentum quantum
numbers, M1 , M2 and M the corresponding projection quantum numbers in respect
of the z-axis. The inverse transformation is

|J1 M1 J2 M2 = |J1 J2 J M J1 J2 J M|J1 M1 J2 M2 , (B.32)
JM
and the coefficients J1 J2 J M|J1 M1 J2 M2 are the conjugate complex ones of
J1 M1 J2 M2 |J1 J2 J M . According to the general convention one chooses the phase
such that the coefficients are real:
J1 J2 J
J1 J2 J M|J1 M1 J2 M2 = J1 M1 J2 M2 |J1 J2 J M = CM 1 M2 M
. (B.33)
B.2 Coupling of Two Angular Momenta 565
They are called C LEBSCH -G ORDAN coefficients. They are non-zero if and only
two conditions are fulfilled:
1. for the Ji quantum numbers (J1 J2 J ) = 1 must be fulfilled, with the triangular
relation defined by:
"
1 if |J2 J1 |, |J2 J1 | + 1, . . . J |J2 + J1 |
(J1 J2 J ) = (B.34)
0 else
2. for the projection quantum numbers Mi = M1 , M2 and M
Ji , Ji + 1, . . . Mi Ji and (B.35)
M = M1 + M 2
must hold.
The C LEBSCH -G ORDAN coefficients disappear if the triangular relation is not sat-
isfied, i.e. if (J1 J2 J ) = 0, or if M = M1 + M2 .
Alternative to the C LEBSCH -G ORDAN coefficients the W IGNER 3j symbols

J1 J2 J
M1 M2 M
are used. They are defined completely symmetric in respect of the three angular
momenta, and are related to the C LEBSCH -G ORDAN coefficients by

J1 J2 +M J1 J2 J
J1 M1 J2 M2 |J M = (1) (2J + 1)1/2
. (B.36)
M1 M2 M
Note the minus sign in front of M in the last column. The 3j symbols disappear
unless
(J1 J2 J ) = 1 and M1 + M2 = M. (B.37)
B.2.2 Orthogonality and Symmetries
The following orthonormality relations hold:

J1 M1 J2 M2 J M J1 M1 J2 M2 |J M = M1 M M2 M (J1 J2 J ) (B.38)
1 2
JM

J1 M1 J2 M2 J M J1 M1 J2 M2 |J M = MM J J (J1 J2 J ). (B.39)
M1 M2
Correspondingly, the 3j symbols satisfy the orthogonality relations:

J1 J2 J J1 J2 J
(2J + 1) = M1 M M2 M (J1 J2 J )
M1 M2 M M1 M2 M 1 2
JM
(B.40)
J1 J2 J

J1 J2 J

(2J + 1) = J J MM (J1 J2 J ) (B.41)
M1 M2 M M1 M2 M
M1 M2
J1 J2 J
2
(J1 J2 J )
and thus also = . (B.42)
M1 M2 M (2J + 1)
M1 M2
The 3j symbols are symmetric in respect of cyclic exchange of their columns.

Their sign changes according to (1)J1 +J2 +J when two columns are exchanged or
the sign of all projection quantum numbers is changed:

J1 J2 J J2 J J1 J J1 J2
= = (B.43)
M1 M2 M M2 M M1 M M1 M2

J2 J1 J
= (1)J1 +J2 +J etc. (B.44)
M2 M1 M

J1 +J2 +J J1 J2 J
= (1) . (B.45)
M1 M2 M
B.2.3 General Formulae
The 3j symbols can be computed either by suitable recursion formulae or from the
R ACAH formula:

J1 J2 J
(B.46)
M1 M2 M

= (1)J1 J2 M (M1 +M2 )M (J1 J2 J )

(J1 + M1 )!(J1 M1 )!(J2 + M2 )!(J2 M2 )!(J + M)!(J M)!
'
(1)t t! (J J2 + t + M1 )!(J J1 + t M2 )!
t
(1
(J1 + J2 J t)!(J1 t M1 )!(J2 t + M2 )!
with the so called triangular function
(a + b c)!(a b + c)!(a + b + c)!

(abc) = . (B.47)
(a + b + c + 1)!
B.2 Coupling of Two Angular Momenta 567
One has to sum over all values of t which lead to non-negative terms for the faculty
expressions in (B.47). The 3j symbols can be obtained very fast from simple com-
puting programmes, as already mentioned (e.g. S VEN G ATO R EDSUN, 2004, Java
script, also for 6j and 9j symbols).
B.2.4 Special Cases
Closed expressions for some frequently used cases are often useful. The 3j symbol
becomes specially simple if one of the angular momenta is zero:

J1 J2 0 (1)J1 M1 J1 J2 M1 M2
= . (B.48)
M1 M2 0 2J1 + 1
From (B.45) one finds the rule:

J1 J 2 J
= 0 if S = J1 + J2 + J is odd. (B.49)
0 0 0
A rather compact expression is also obtained if S = J1 + J2 + J is even and all

projection quantum numbers are zero:

1/2
J1 J2 J (S 2J1 )!(S 2J2 )!(S 2J )!
= (1)S/2 (B.50)
0 0 0 (S + 1)!
(S/2)!
.
(S/2 J1 )!(S/2 J2 )!(S/2 J )!
Frequently used relations are:

/ 0
J+ 1
J 1
J M1/2 J M + 1/2
2 2 = (1) (B.51)
M M 1
2
1
2
(2J + 1)(2J + 2)

J +1 J 1 (J M)(J M + 1)
= (1)J M1 (B.52)
M M 1 1 (2J + 1)(2J + 2)(2J + 3)

+1 J 1 2(J M + 1)(J + M + 1)
= (1)J M1 (B.53)
M M 0 (2J + 1)(2J + 2)(2J + 3)

J J 1 (J M)(J + M + 1)
= (1)J M (B.54)
M M 1 1 J (2J + 1)(2J + 2)

J J 1 M
= (1)J M (B.55)
M M 0 J (J + 1)(2J + 1)

J J 2
(B.56)
MM 2 2

6(J M 1)(J M)(J + M + 1)(J + M + 2)
= (1)J M
(2J + 3)(2J + 2)(2J + 1)2J (2J 1)

J J 2
(B.57)
M M 1 1

6(J + M + 1)(J M)
= (1)J M (1 + 2M)
(2J + 3)(2J + 2)(2J + 1)2J (2J 1)

J J 2 2[3M 2 J (J + 1)]
= (1)J M . (B.58)
M M 0 (2J + 3)(2J + 2)(2J + 1)2J (2J 1)
Note: Formula (B.54) is slightly erroneous in B RINK and S ATCHLER (1994).
B.3 RACAH Function and 6j Symbols

B.3.1 Denition
Often one has to couple three angular momenta. A typical case is the hyperfine
structure (HFS) of atoms in one electron systems, where orbital angular momentum,
electron spin and nuclear spin have to be considered. In the case of two electrons,
one may have no orbital angular momentum (ns orbital), the other electron may
be in a n configuration, e.g. Hg(6sn ). Here the spins of the two electrons cou-
ple to a total spin and the latter couples with the orbital angular momentum of the
n electron. In molecular physics too, a host of problems involves similar coupling
requirements, e.g. the molecular rotation may couple with the electrons orbital an-
gular momentum and its spin. And even in optical transitions between spin-orbit
coupled atomic states the photon adds an extra angular momentum from which the
optical selection rules are derived.
Generally, in order to determine the states of three angular momenta one starts
with the product states
|j1 m1 |j2 m2 |j3 m3 = |j1 m1 j2 m2 j3 m3
in an uncoupled representation. Formally, the total angular momentum vector oper-

ator is given by
J1 +
J = J2 +
J 3,
and one has to find eigenstates of J and Jz with the eigenvalues J (J + 1)2
2
and M. This may be done in two alternative schemes, based on the coupling of
two angular momenta described in Appendix B.2. One may first couple J 1 and
J2
to
J 12 and then combine J 12 with
J 3 to
J . According to (B.31) from the product
states |j1 m1 |j2 m2 one obtains
B.3 RACAH Function and 6j Symbols 569

(j1 j2 )J12 M12 = |j1 m1 |j2 m2 j1 m1 j2 m2 |J12 M12 ,
m1 m2
from which with |j3 m3

(j1 j2 )J12 j3 ; J M = (j1 j2 )J12 M12 |j3 m3 J12 M12 j3 m3 |J M (B.59)
M12 m3
follows. This state is simultaneously eigenstate of J , Jz and

2
J 12 with J12 being an
additional quantum number of the system. Alternatively one couples first J 2 with

J 3 to give J 23 and combines then J 23 with J 1 to
J . The eigenstate for the total
angular momentum J and its z-component Jz is then
2

j1 (j2 j3 )J23 ; J M = |j1 m1 (j2 j3 )J23 M23 j1 m1 J23 M23 |J M . (B.60)
m1 M23
This state is also eigenstate of

2
J 23 , and in this case J23 is an additional quantum
number. These two representations are of course not linearly independent from each
other they describe the same subspace of angular momentum states. Thus, one
may express each state in one of the two coupling schemes by a linear combination
of the states in the other scheme:

(j1 j2 )J12 j3 ; J M (B.61)

= j1 (j2 j3 )J23 ; J M j1 (j2 j3 )J23 ; J M (j1 j2 )J12 j3 ; J M .
J23
The expansion coefficients are scalar quantities and independent of M. To write

them in a symmetric manner one uses the so called W -function after R ACAH (1942)
or the 6j symbol:

j1 (j2 j3 )J23 ; J M (j1 j2 )J12 j3 ; J M (B.62)
= (2J12 + 1)1/2 (2J23 + 1)1/2 W (j1 j2 Jj3 ; J12 J23 )
" #
j1 j2 j3 J j1 j2 J12
with W (j1 j2 Jj3 ; J12 J23 ) = (1) . (B.63)
j3 J J23
B.3.2 Orthogonality and Symmetries
For the 6j symbols not to disappear, the 6 angular momenta must be coupled accord-
ing to the above definitions. This requires triangular relations between the angular
momenta. Schematically the following triangular relations must be satisfied:
" # " # " # " #
* * * * * *
, , , . (B.64)
* * * * * *
The 6j symbols are invariant against exchanging any two columns or any two rows:
" # " # " # " #
j1 j2 J12 j1 J12 j2 j2 J12 j1 J J23 j1
= = = , etc.
j3 J J23 j3 J23 J J J23 j3 j2 J12 j3
(B.65)
Finally, a number of sum rules hold for the 6j symbols. We note
" #
j1 j2 k
(1) (2k + 1)
2k
= 1, (B.66)
j2 j 1 J
k
and the orthogonality relation

" #" #
j1 j2 J j1 j2 J
(2J + 1) 2J + 1 = J J . (B.67)
j3 j4 J j3 j4 J
J
B.3.3 General Formulae
One may develop explicit expressions for the 6j symbols from (B.59)(B.61). One
finds according to R ACAH:
" #
j1 j2 J12
(B.68)
j3 J J23
(1)t (t + 1)!
= (j1 j2 J12 ) (j1 J J23 ) (j3 j2 J23 ) (j3 J J12 )
t
f (t)
with f (t) = (t j1 j2 J12 )!(t j1 J J23 )!(t j3 j2 J23 )!

(t j3 J J12 )!(j1 + j2 + j3 + J t)!
(j2 + J12 + J + J23 t)!(J12 + j1 + J23 + j3 t)!
and the triangular function (abc) according to (B.47). The summation is again
over all nonvanishing terms. Again it is commendable to use one of the available
Java-applets for practical evaluations.
The 6j symbols may also be expanded as sums over products of 3j symbols. An
important relation is the contraction formula:

j1 +j2 +j3 +1 +2 +3 1 j2 j3
(1)
m1 2 3
1 2 3

j1 2 j3 j1 j2 3

1 m2 3 1 2 m3
" #
1 2 3 1 2 3
= . (B.69)
m1 m2 m3 j1 j2 j3
B.4 Four Angular Momenta and 9j Symbols 571
For the special case that 1 + 2 + 3 is even and j3 = 1/2, a particularly simple
relation emerges:

j1 j2 3
= (1) 1 2 j2 j1 (2 1 + 1)(2 2 + 1) (B.70)
1/2 1/2 0
" #
1 2 3 1 2 3
.
j2 j1 1/2 0 0 0
B.3.4 Special Cases
Again, some closed expressions for special cases are occasionally useful. With s =
a + b + c and X = b(b + 1) + c(c + 1) a(a + 1):
" #
a b c (1)s (abc)be cd
= (B.71)
0 d e (2b + 1)(2c + 1)
" #
a b c 2(1)s+1 X
= (B.72)
1 c b 2b(2b + 1)(2b + 2)2c(2c + 1)(2c + 2)
" #
a b c 2(1)s
= (B.73)
2 c b (2b 1)2b(2b + 1)(2b + 2)(2b + 3)
[3X(X 1) 4b(b + 1)c(c + 1)]

(2c 1)2c(2c + 1)(2c + 2)(2c + 3)
* -
a+ 1
a 1
(1)s+1 (s 2c + 1)(2 + s)
2 2 = (B.74)
b+ 1
2 b c + 12 2 (2a + 1)(a + 1)(2b + 1)(b + 1)
* -
a+ 1
2 a 1
2 = (1) s+1 (s 2c + 1)(2 + s)
. (B.75)
b b+ 2 c
1 2 (2a + 1)(a + 1)(2b + 1)(b + 1)
B.4 Four Angular Momenta and 9j Symbols
Very briefly, we mention the coupling of four angular momenta, j1 , j2 , j3 and

j4 , which again may be accomplished in different sequences. E.g. one may cou-
ple |(j1 j2 )j12 , (j3 j4 )j34 ; j m or alternatively |(j1 j3 )j13 , (j2 j4 )j24 ; j m . The two
schemes relate to each other by a unitary transformation:

(j1 j3 )j13 , (j2 j4 )j24 ; j m (B.76)

= (j1 j2 )j12 , (j3 j4 )j34 ; j m (j1 j2 )j12 , (j3 j4 )j34 ; j (j1 j3 )j13 , (j2 j4 )j24 ; j .
j12 ,j34
As a short notation for the recoupling coefficients

(j1 j2 )j12 , (j3 j4 )j34 ; j (j1 j3 )j13 , (j2 j4 )j24 ; j (B.77)

1/2 j1 j2 j12
= (2j12 + 1)(2j34 + 1)(2j13 + 1)(2j24 + 1) j3 j4 j34

j13 j24 j
the so called W IGNER 9j symbols are used {given in curly brackets}.

Each column and each row has to satisfy a triangular relation, and an odd number

of permutations of rows or columns only changes the sign according to (1) all j s .
The orthonormality relation reads

(2j12 + 1)(2j34 + 1)(2j13 + 1)(2j24 + 1)
j13 ,j24

j1 j2 j12 j1 j2 j12
j3 j4 j34 j3 j4 j34 = j13 j j24 ,j .
13 24
j13 j24 j j13 j24 j
For later use, we also communicate here one useful formula. In the case that one of
the j s is zero, the 9j symbol reduces to a 6j symbol:

a b c " #
cf gh (1)b+d+c+g a b c
d e f = . (B.78)
[(2c + 1)(2g + 1)]1/2 e d g
g h 0
c.c.: complex conjugate.

References
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics,
vol. 64. Berlin, Heidelberg: Springer Verlag, 3rd edn., 343 pages.
C ONDON, E. U. and G. S HORTLEY: 1951. The Theory of Atomic Spectra. Cambridge, England:
Cambridge University Press, 441 pages.
R ACAH, G.: 1942. Theory of complex spectra II. Phys. Rev., 62, 438462.
S VEN G ATO R EDSUN: 2004. 3j 6j 9j -symbol java calculator, Sven Gato Redsun. http://www.
References 573
W EISSBLUTH, M.: 1978. Atoms and Molecules. Student Edition. New York, London, Toronto,
Sydney, San Francisco: Academic Press, 713 pages.
W EISSTEIN, E. W.: 2004a. Wigner 3j -symbol, Wolfram Research, Inc., Champaign, IL, USA.
http://mathworld.wolfram.com/Wigner3j-Symbol.html, accessed: 8 Jan 2014.
W EISSTEIN, E. W.: 2004b. Wigner 6j -symbol, Wolfram Research, Inc., Champaign, IL, USA.
http://mathworld.wolfram.com/Wigner6j-Symbol.html, accessed: 8 Jan 2014.
Matrix Elements
C
In this appendix some tools are collected which are useful for evaluating matrix ele-
ments of irreducible tensor operators. These matrix elements are needed to compute
transition probabilities for optical excitation of atomic and molecular states as well
as for the analysis of collision induced processes.
C.1 Tensor Operators
C.1.1 Denition
According to R ACAH (1942) one defines as irreducible tensor operator Tk of the

rank k each operator whose 2k + 1 components Tkq (q = k, k + 1, . . . , k) sat-
isfy the same commutation rule in respect of
J as the spherical harmonic operators
Ykq (, ). Following R ACAH this implies in our notation (B.3)1 for the spherical
components of the angular momentum operator J:
[Jz , Tkq ] = q Tkq and (C.1)

1
[J , Tkq ] = (k q)(k q + 1)Tkq1 . (C.2)
2
B RINK and S ATCHLER (1994) define equivalently a spherical tensor (operator)
Tk of rank k as a quantity defined by 2k + 1 components, which transforms upon co-
ordinate rotation in the same way as the irreducible representation Dk of the rotation
group (see Appendix E):

Tkq

= Tks Dksq ( ). (C.3)
s
1 We refer to J = 1 (Jx iJy ) while R ACAH gives the relation for (Jx iJy ). See also
2
footnote 1 in Appendix B.

576 C Matrix Elements
Here Tkq is the component of the tensor in the new coordinate system which arises
by rotating the old, un-dashed system through the E ULER angles . It is rep-
resented in term of the components Tks of the tensor in the old system. Important
examples of tensor operators are the spherical harmonics Y m ( ) (rank ) them-
selves, or somewhat more general the eigenstates |j m of the angular momentum
operator (rank j ).
The adjoint (or Hermitian conjugate) operator Tkq
of a tensor operator Tkq is
defined by (see Sect. 2.3.1)

J M Tkq |J M = Tkq J M J M = J M|Tkq JM , (C.4)
and with the C ONDON and S HORTLEY (1951) phase convention
Tkq

= (1)q Tkq . (C.5)
If Tkq is simply a complex function this is equivalent to Tkq

= Tkq
, as e.g. in the
case of the spherical harmonics (B.26).

A tensor is called Hermitian or self adjoint if Tkq

= Tkq . Obviously, with (C.4)
the diagonal matrix elements (and hence the eigenvalues) are then real: Hermitian
operators describe measurable, real quantities.
Tensors of rank 1 are called vector operators, e.g. the position vector r, the unit
vector of polarization e, the dipole moment D, as well as all angular momentum
operators, such as S,
L,
J, . The components of a vector operator V
I or F in the
spherical basis are
1 = 1 [V
V x iV
y ]; 0 = V
V z . (C.6)
2
C.1.2 WIGNER-ECKART Theorem
When evaluating matrix elements for a tensor operator, the W IGNER -E CKART theo-
rem is very useful. It says, that for any irreducible tensor operators of a given rank k,
the matrix elements between states |J M and |J M depend in identical form on
the projection quantum numbers M, M , and q:

J M Tkq | J M
5
= J Mkq J M J 5Tk J (C.7)

J J k 5
= (1)J k+M 2J + 1 J 5Tk J (C.8)
M M q

J M
J k J 5
= (1)
2J + 1 J 5Tk J . (C.9)
M q M
C.1 Tensor Operators 577
The thus defined reduced matrix element2 J Tk J is independent of M , M

and q. Due to the properties of the 3j symbol the matrix elements are non-zero only
if M = M q and if the triangular relation (J kJ ) = 1 holds!
In practice, to determine the reduced matrix element one has to evaluate just
one nonvanishing matrix element (preferably a particularly simple one) and to di-
vide it by the corresponding C LEBSCH -G ORDAN coefficient, e.g. J Tk J =
J 0|Tk0 |J 0 /J 0|J k00 if J + J + k is even.
An important example is the reduced matrix element of an angular momentum.
With

J M Jz |J M = J J M M M (C.10)

J 1 J 5
= (1)J M 2J + 1 J 5
J J
M 0 M
and (B.55) for the 3j symbol, one obtains:

5
J 5J J = J J J (J + 1). (C.11)
The W IGNER -E CKART theorem allows one to relate two matrix elements of dif-
ferent tensor operators, Sk and Tk of the same rank k with each other in a simple
manner. The ratio R of the reduced matrix elements of two tensor operators of the
same rank k may be computed from any pair of (nonvanishing) matrix elements of
their components:
J Tk J J M |Tkq | J M
R J , J, T, S = = . (C.12)
J Sk J J M | Skq | J M
Specifically for vector operators V with (C.7) we write

J M V q | J M = R J M J q | J M , (C.13)
which is independent of M and M and holds for all components V q .

The diagonal matrix elements of the scalar product V J may be written

J M|VJ | J M = J M|V J M J M J | J M , (C.14)
M
where we have simply inserted the quantum mechanical unit operator

1 = M | J M J M |. With (C.13), this expression becomes
Our definition of reduced matrix elements follows B RINK and S ATCHLER (1994), as-
suming, however, integer rank only, so that (1)2k = 1. Note that these reduced
matrix elements are not symmetric in J and J , rather (2J + 1) J T J =
k
(2J + 1) J Tk J . R ACAH (1942), E DMONDS (1996) and others use a symmetric
definition: J Tk J Racah = 2J + 1 J Tk J .

J M|V J | J M = R J J M J M
J M| J | J M (C.15)
M
= R J M|
2
J | J M = RJ (J + 1)2 .
In the second line the eigenvalue (B.12) of

2
J has been used, so that finally
J M|VJ | J M
R= . (C.16)
J (J + 1)2
Inserting this back into (C.13) gives the very useful projection theorem
J M|VJ | J M
J M Vq | J M = J M Jq | J M . (C.17)
J (J + 1) 2
realizes that (V J )J /J (J + 1) is the projection of V onto the unit vector

If one 2
J / J (J + 1), its physical relevance becomes evident. The projection theorem is

used e.g. in a coupled scheme of angular momenta, to express the matrix elements
of individual components in terms of the total angular momentum J. Specifically
z one obtains with J M |Jz | J M = MM M
for q = 0 =
J M|VJ | J M
z | J M =
J M|V M. (C.18)
J (J + 1)
C.2 Products of Tensor Operators

Direct products of matrices are an important concept in matrix algebra (see e.g.
B LUM, 2012, Appendix A). If A is N dimensional and B is n dimensional, then the
direct product A B is N n dimensional.
Similarly, from tensor operators Rk1 and Uk2 of rank k1 and k2 , respectively, one
may construct (2k1 + 1)(2k2 + 1) components of new tensor operators. These can
be expressed in irreducible form again as direct product Rk2 Uk2 by tensors TK of
rank K with |k1 k2 | K |k1 + k2 |:

TKQ = [Rk2 Uk2 ]KQ = k1 q1 U
R k2 q2 k1 q1 k2 q2 |KQ
q1 q2
(C.19)
= k1 q U
R k2 Qq k1 qk2 Q q|KQ .
q
Here k1 k2 q1 q2 |KQ is a C LEBSCH -G ORDAN coefficient as introduced by (B.31)

in the context of angular momentum coupling.
Specifically, with (B.36) and (B.48) the scalar product (rank K = 0) of two ten-
sor operators of the same rank k becomes simply

Rk Uk = kq U
(1)q R kq . (C.20)
C.2 Products of Tensor Operators 579
This is the generalization of the standard scalar product of two vectors R and S (see
e.g. (B.8)). The result is a single scalar quantity.
C.2.1 Products of Spherical Harmonics
Application ofthe relations given above to the renormalized spherical harmonics

(B.29), Ckq = 4/(2k + 1)Ykq , leads to a number of useful expansion formulae.
(1) (1)
With (C.20) the scalar product of the two tensor operators C = C (1 , 1 ) and
(2) (2)
C = C (2 , 2 ) at different angles becomes:
(1) (2)

C C = (1)m C m (1 , 1 )C m (2 , 2 ). (C.21)
m=
If the coordinate system is such that 2 2 defines the z-axis, we have C m (2 , 2 ) =

m0 , and the sum reduces to C 0 ( 0) = P (cos ), with being the angle between
the two direction in space characterized by (1 , 1 ) and (2 , 2 ). Rewriting

(1)m C m = C m
we obtain the addition theorem for the renormalized spherical harmonics:

P (cos ) = C m (1 , 1 )C m (2 , 2 ) (C.22)
m=

= C mp (1 , 1 )C mp (2 , 2 ). (C.23)
m=0,p=1
The second line is easily verified with the definition of real tensor operators (D.5)
applied to the (renormalized) real spherical harmonics C mp (, ) (D.7). One also
derives from (C.19) a useful expansion formula for the product of two spherical
harmonics at the same angle

kq m m C m (, )C m (, ) = k0 00 Ckq (, ), (C.24)
m m
where the normalization constant k0| 00 has been derived from the definition of
spherical harmonics at = = 0. Using the orthogonality relations (B.38) for the
C LEBSCH -G ORDAN coefficients one also obtains the inversion of this relation:

C m (, )C m (, ) = Ckq (, ) kq m m k0 00 . (C.25)
kq
C.2.2 Matrix Elements of the Spherical Harmonics
The often used matrix elements of the spherical harmonics in an | m basis, nor-
malized as

(2 + 1)(2 + 1)
m Ckq | m = C m (, )Ckq (, )C m (, )d,
4
(C.26)
may be evaluated by replacing the latter two functions with the expansion (C.25):

(2 + 1)(2 + 1)
m Ckq | m =
4

C m (, )Ck q (, )d k q k qm k 0k 00
k q

2 + 1
=
k q m k q k qm k 0k 00 .
2 + 1
kq
Thus, we finally obtain:

2 + 1

m Ckq | m =
k qm m k 00 0 (C.27)
2 + 1

m Ckq | m = (1)m 2 + 1 (2 + 1) m m+q k (C.28)

k k

0 0 0 m q m

m
k 5

m Ckq | m = (1)
2 + 1 5Ck . (C.29)
m q m
The last equation is simply the W IGNER -E CKART theorem for the matrix element.
Comparing the latter two equalities gives the reduced matrix element of Ck :

5 k
5
Ck = (1) (2 + 1) . (C.30)
0 0 0
We recall that this expression is zero if + k + is odd. We also note that

5 5
(2 + 1) 5
5
Ck = (1) Ck . (C.31)
(2 + 1)
For k = 0 this implies

5
5C0 = , (C.32)
C.2 Products of Tensor Operators 581
and more generally for = one obtains with (B.50)

(1)k/2 k!( + k/2)! (2 + 1)(2 k)!
Ck = . (C.33)
(k/2)!2 ( k/2)! (2 + k + 1)!
Note that this expression only holds for even k. The reduced matrix element vanishes
for odd k. For the special case C2 one obtains:

( + 1)
C2 = . (C.34)
(2 1)(2 + 3)
Note the minus sign!
Of special importance in respect of the selection rules for electric dipole transi-
tions is according to (4.79) the case k = 1. Because of (B.34) and (B.49)
parity conservation holds: = 1 (C.35)
(see also Appendix D). With (B.50) the only nonvanishing matrix elements are

5 +1
for = + 1
5C1 =
2 +3
(C.36)

(2 1) for = 1,

and because of (B.43)(B.2.3)

m C1q | m = (1)q m|C1q m (C.37)

= m C1q | m = (1)q m|C1q m .
Hence, in dipole transitions the selection rule for the projection quantum number
becomes m = q. The matrix elements vanish unless m = m 1 or m = m. A de-
tailed evaluation of (C.29), using (B.52)(B.55) leads to:

( m + 1)( + m + 1)
( + 1)mC10 | m = (C.38)
(2 + 1)(2 + 3)

( + m + 1)( + m + 2)
( + 1)(m + 1)C11 | m = (C.39)
2(2 + 1)(2 + 3)

( m + 1)( m + 2)
( + 1)(m 1)C11 | m = (C.40)
2(2 + 1)(2 + 3)

( m)( + m)
( 1)mC10 | m = (C.41)
(2 1)(2 + 1)

( 1 m)( m)
( 1)(m + 1)C11 | m = (C.42)
2(2 1)(2 + 1)

( 1 + m)( + m)
( 1)(m 1)C11 | m = . (C.43)
2(2 1)(2 + 1)
For the particularly simple case of s p transitions one finds that all allowed tran-
sitions have the same amplitudes:
10|C10 |00 = 00|C10 |10 = 11|C11 |00 (C.44)

= 1 1|C11 |00 = 00|C11 |1 1 = 00|C11 |11 = 1/3.
C.3 Reduction of Matrix Elements
Often matrix elements of tensor operators TKQ have to be evaluated in a scheme

of coupled angular momenta, such as |(LS)J M . Typically, these operators may be
composed according to (C.19) from two tensor operators, R k1 q1 (1) which acts only

on the first component of the eigenstate, and Uk2 q2 (2) which acts only on the second
component.
With the W IGNER -E CKART theorem (C.9), the matrix element of TKQ between
state | LSJ M and | L SJ M is:

L S J M TKQ |LSJ M

J K+M
J J K 5
= (1)
2J + 1 L S J 5TK LSJ . (C.45)
M M Q
For evaluating the reduced matrix elements of product tensors angular momentum
algebra provides a useful toolbox. In simple cases, with some algebra and exploiting
6j symbols (B.61), the reduced matrix elements in the coupled scheme |(LS)J M
may be recast into the uncoupled scheme |LML |SMS . We communicate here only
(1)
some results. If Rk acts only on the first angular momentum of the coupled scheme
(2)
and Uk 1, we have:
5 (1)

L S J 5Rk LSJ = S S (1)k+L +S+J (2J + 1) 2L + 1
" #
L L k 5
L 5Rk L . (C.46)
J J S
Conversely, if Rk
(1)
1 and U acts only onto the second angular momentum in
the coupled scheme, we obtain:
5 (2)

L S J 5Uk LSJ = L L (1)k+L+S+J (2J + 1) 2S + 1
" #
S S k 5
S 5Uk S . (C.47)
J J L
C.3 Reduction of Matrix Elements 583
Note: due to the well defined phases of the C LEBSCH -G ORDAN coefficients the se-
quence of the angular momenta is not arbitrary. For the scalar product (rank K = 0)
of two tensor operators of rank k the reduction becomes particularly clear. We first
notice that according to (C.7)
5
j1 j2 J M (Rk Sk )00 |j1 j2 J M = M M J J j1 j2 J 5Rk Sk j1 j2 J , (C.48)
where we have used the explicit expression (B.48) for the 3j symbol. Explicitly
(C.48) may be evaluated by inserting on the left side into the scalar product Rk Sk
the quantum mechanical unit operator. Then each part may be reduced according
to (C.46). By applying the orthogonality relations according to R ACAH (1942) (if
one has only to average over the angular momentum part) one finds for the reduced
matrix element of the scalar product:
5
j1 j2 J 5Rk Sk j1 j2 J = 2j1 + 1) 2j2 + 1 (C.49)
" #
j j k 5 5
(1)J +j1 +j2 1 1 j1 5Rk j1 j2 5Sk j2 .
j2 j 2 J
A somewhat different situation is encountered when a tensor TK is the product of

two tensor operators Rk1 and Sk2 which are constructed from the same coordinates,
momenta etc., and thus act on the same system. From simple angular momentum
recoupling, E DMONDS (1996) obtains in this case (Eq. (7.1.1)):
5
j 5(Rk1 Sk2 )K j = 2K + 1(1)K+j +j (C.50)
" k1 k2 K # 5
j 5Rk j jSk j .
j j j
j
C.3.1 Matrix Elements of the Spherical Harmonics in LS Coupling
Most often required are the matrix elements of the spherical harmonics Yk (or of the
renormalized Ck = 4/(2k + 1)Yk ). We have described these in Appendix C.2.2
already in a basis of pure n states. For LS coupled orbital angular momenta with
orbital quantum numbers and the reduced matrix elements Ck are given
by (C.28).3 In the coupled scheme with a total electron spin S the reduced matrix
element (C.46) is:
3 The following expressions are also valid in multi-electron systems as long as there is only one
active electron (while further electrons have zero orbital angular momentum as it is the case for
alkali atoms, or for He below the first ionization limit). Here and is identical to the total orbital
angular momenta L and L, respectively, while the total spin S 1/2 of the system can in principle
be composed from several particles.
5
S J 5Ck SJ = S S (2J + 1) 2 + 1 (2 + 1) (C.51)
" #
k k
(1)S J .
J J S 0 0 0
In this case, inverting the reduced matrix element leads to

5
J J (2J + 1)
5
5
SJ Ck SJ = (1) SJ 5Ck SJ . (C.52)
(2J + 1)
This expression is only non-zero if + + k is even. Thus, (C.45) reads now

S J M Ckq | SJ M

= S S 2J + 1 (2J + 1) 2 + 1 (2 + 1) (C.53)
" #
J J k k k
(1)M S .
M M q J J S 0 0 0
Special Case: Spin 1/2
For the case S = s = 1/2 (e.g. for an effective one particle system with spin 1/2
i.e. electron in an atom or a single active proton in a nucleus), using (B.70) reduces
(C.51) to:
5 5
5 5
1 5 5 1 j 1/2+k
j j k
j 5Ck 5 j = (1) (2j + 1) . (C.54)
2 2 1/2 1/2 0
This expression is even independent of and (of course + + k must be even)

and one obtains for (C.53):

1
1
j m Ckq j m = (1)j m +j 1/2+k 2j + 1 (2j + 1)
2 2

j j k j j k
. (C.55)
m m q 1/2 1/2 0
For the evaluation of line strengths in E1 transitions, we report here some values for
the reduced matrix element of C1 in the case of a one electron systems:
j j
5 5 1/2 3/2 5/2

5 5 1
1 5 1/2 1/3
j 5C1 5
1/3 0
2 5 2 :
j
3/2 1/3

1/15

2/5
(C.56)

5/2 0 2/5 1/35
C.3 Reduction of Matrix Elements 585
C.3.2 Scalar Products of Angular Momentum Operators
Scalar products of coupled angular momenta are most conveniently evaluated by

using binomial formulas. We exemplify this for the product of angular momentum
L and spin
S which is the crucial operator in spin-orbit interaction. As discussed in
Chap. 6, the operator L J and the projection Jz of the
S commutes with the square
2
total angular momentum J = L+

S. In the scheme of coupled angular momenta, i.e.
when the system is described by eigenstates |LSJ MJ , the good quantum numbers
are L, S, J and MJ . From the first binomial formula one obtains:

LSJ MJ L
S|LSJ MJ
1 2
LSJ MJ L
J
2 2
= S |LSJ MJ
2
2
= J (J + 1) L(L + 1) S(S + 1) J J MJ MJ . (C.57)
2
However, one often needs to express the matrix elements in the uncoupled basis of
eigenstates |LML SMS , e.g. in the case of strong magnetic fields as described in
Sect. 8.1.3. Thus, for uncoupled L and
S one returns to the definition of the scalar
product according to (B.8)

LML SMS L
S|LML SMS

= LML SMS L+
S |LML SMS (C.58)

LML SMS S+ |LML SMS + LML SMS
L Lz
Sz |LML SMS ,
and uses the matrix elements (B.14)(B.17) for the angular momenta. Since the
q act only onto the LML component of the state vector and
operators L Sq only onto
the SMS component one sees that the matrix elements are only nonvanishing for
MJ = ML + MS = ML + MS = MJ . (C.59)
Hence, MJ = ML + MS is a good quantum number, albeit J is not. Three types of

nonzero matrix elements exist in the uncoupled system:

L(ML + 1)S(MS 1) L
S|LML SMS

= L(ML + 1)S(MS 1) L+
S |LML SMS

= 2 L(L + 1) ML (ML + 1) S(S + 1) MS (MS 1) /4 (C.60)

L(ML 1)S(MS + 1) L
S|LML SMS

= LS ML 1 MS + 1|L S+ |LML SMS

= 2 L(L + 1) ML (ML 1) S(S + 1) MS (MS + 1) /4 (C.61)
L
Table C.1 Matrix elements of S/2 for a 2 P state in the uncoupled basis |LML SMS , with
= ML + 2MS .
ML MS |1 12 |0 12 |1 12 |1 12 |0 12 |1 12
2 1 0 0 1 2
MJ 3/2 1/2 1/2 1/2 1/2 3/2
1 12 | 1/2 0 0 0 0 0

0 12 | 0 0 1/ 2 0 0 0

1 12 | 0 1/ 2 1/2 0 0 0

1 12 | 0 0 0 1/2 1/ 2 0

0 12 | 0 0 0 1/ 2 0 0
1 12 | 0 0 0 0 0 1/2
L
LML SMS | S|LML SMS
z
= LML SMS |L Sz |LML SMS = 2 ML MS . (C.62)
Table C.1 provides as an example all LS matrix elements for a 2 P state (L = 1,

S = 1/2).
C.3.3 Components of Angular Momenta
As a further example we compute the z-component of an angular momentum which

acts only on one part of the coupled system. Typical applications are the components
Sz or Lz in a spin-orbit coupled scheme |(SL)J MJ , or Iz and Jz in the hyperfine
coupled scheme |(I J )F MF . We use here the latter quantum numbers, which for
other application just have to be replaced appropriately.
We first derive the reduced matrix element of J according to (C.46):
5
I J F 5J I J F (C.63)
" #
J J 1 5
= (1)1+I +J +F I I (2F + 1) 2J + 1 J 5J J
F F I
" #
J J 1
= (1)1+I +J +F I I J J (2F + 1)J (J + 1) 2J + 1 ,
F F I
where in the last step (C.11) was used. The nonvanishing matrix elements of the
components of J may now be written with the W IGNER -E CKART theorem (C.7),
e.g.

J I F MF Jz |J I F MF (C.64)

F F 1 5
= (1)F 1+MF MF MF 2F + 1 J I F 5J J I F .
MF MF 0
Inserting (C.63) yields:

C.4 Electromagnetically Induced Transitions 587

J I F MF Jz |J I F MF
= (1)2F +MF +I +J MF MF (C.65)
" #
F F 1 J J 1
2F + 1 (2F + 1)(2J + 1)(J + 1)J .
MF MF 0 F F I
Thus, these matrix elements are diagonal in MF , I and J . We may finally insert the
explicit expressions of the 3j symbols from (B.55) and (B.54). We have to distin-
guish two cases:
(a) F = F
J I F MF |Jz |J I F MF

" #
3F +I +J MF (2F + 1)(2J + 1)(J + 1)J J J 1
= (1)
F (F + 1) F F I
(1)2I +2J +1 F (F + 1) + J (J + 1) I (I + 1)
= MF (C.66)
2 F (F + 1)
except for F = 0 where J I F MF |Jz |J I F MF / = 0.
(b) F = F + 1
J I (F + 1)MF |Jz |J I F MF

" #
J J 1
= (1)3F +I +J +1
F F +1 I

2(F + MF + 1)(F MF + 1)(2J + 1)(J + 1)J

(2F + 2)

(1)4F +2I +2J +2 MF 2
= 1 (C.67)
2 F +1

(J + I F )(F + J + I + 2)((F + 1)2 + (J I )2 )
.
(2F + 1)(2F + 3)
In the last step we also have inserted the explicit expression for the 6j symbols
(B.72) to obtain these useful, compact relations.
C.4 Electromagnetically Induced Transitions

Probably the most important application for the tools described here are transition
matrix elements for transitions induced by electromagnetic radiation. We shall dis-
cuss these for electric dipole (E1) and quadrupole (E2), as well as for magnetic
dipole (M1) transitions.
C.4.1 Electric Dipole Transitions
Let us begin with emission or absorption of a photon in an E1 transition. As

discussed in detail in Chap. 4, the probability of such a transition between state
|a = | J M and state |b = | J M is proportional to the absolute square of
matrix elements of the type
1 2
2 2

J M r e| J M = r| J M C1q |J M eq . (C.68)

q=1
Since C1 may be seen as a tensor of rank 1, this expression contains in principle the
(2 1 + 1)(2 1 + 1) components of the product tensor. One may express these
in irreducible form by tensors of rank 0 (basically the total intensity), rank 1 (so
called orientation) and rank 2 (so called alignment). We shall come back to this in
Appendix C, Vol. 2.
For low Z atoms, LS coupling is characteristic according to the scheme
| LSJ M and | L S J M . In this case the selection rules are determined by
the matrix elements S L J M |C1q |SLJ M . Equation (C.53) factors this matrix
element into its important components. In an electromagnetically induced transi-
tion, the total electron spin does not change, i.e. S = 0. For E1 transitions we have
k = 1 and with q = 0, 1 one reads in (C.53) from the first 3j symbol the selection
rules for J and M: M = q and (J 1J ) = 1. Thus M = M 1 or M = M and
J = J 1 or J = J however, 0 0 is strictly forbidden. Finally, the last 3j
symbol recovers the parity selection rule L = L 1.
C.4.2 Electric Quadrupole Transitions
According to (5.56) the crucial matrix element for E2 transitions is

1
a|T0 T|b (E2) b|Q2q+ |a = b|r 2 |a (1)q b|C2q |a + b|C2q |a
2
(C.69)
1
and b|Q2q |a = b|r 2 |a (1)q b|C2q |a b|C2q |a ,
2i
where the definitions (D.5) have been inserted. Since k = 2, we note that q = 0,
1, or 2. In LS coupling we use again (C.53) for evaluation. Again S = 0,
and considering the triangular rule (J J 2) = 1, we read from the 3j symbols that
the total angular momentum J may now change by 0, by 1 or 2 units of ,
now with the restriction 0 0 and 1/2 1/2. With M = q follows in this case,
that transitions are possible with M = M, as well as with M 1 and even with
M 2. This leads to correspondingly structured quadrupole absorption and emission
patterns. In addition, for quadrupole transitions, parity conservation now implies
L = L 2.
C.4 Electromagnetically Induced Transitions 589
C.4.3 Magnetic Dipole Transitions
Selection rules for M1 transitions are derived from the matrix element (5.53)
a|eE0 L + 2
D|b (M1) b| J +
S|a B = b| S|a B , (C.70)
where the index B indicates that the coordinate frame for orbital angular momentum
L and spin
S is chosen here with its z-axis parallel to the magnetic RF field inducing
the transition. This particular polarization dependence can be expressed again by the
W IGNER -E CKART theorem (C.9). Since we discuss a dipole transition with k = 1
we obtain:

L + 2
b| S|a B = (1)J 1+M 2J + 1 (C.71)

J J 1 5 5
J 5 LJ + 2 J 5
SJ .
M M q
With | we describe the overlap integral of radial part of the wave function
between initial and final state. The 3j symbol determines the selection rule for the
magnetic quantum numbers M, with q referring in this case to the direction of the
magnetic field vector (rather than the electric field in the E1 case). In spectroscopic
experiments, typically, radio frequency (RF) or microwave radiation (MW) is used
in external, static magnetic field B st , defining the z-axis. If the radiation field vector
B points into this z-direction (i.e. if the polarization vector e and the wave vector k
lie in the xy plane), then q = 0 and transitions with M = M occur. To induce M =
1 transition (q = 1), the B vector of the inducing field has to be perpendicular
to static field B st .
The two reduced matrix elements in (C.71) determine the selection rules for the
quantum numbers J , L and S. They have to be evaluated (for not too high static
magnetic fields) according to (C.46) and (C.47) in the coupled system. With (C.11)
we have L LL = L L L(L + 1) and S SS = S S S(S + 1). Thus we
obtain
5

L S J 5 LLSJ = S S L L (1)k+L +S +J (2J + 1) 2L + 1
" #
L L 1
L(L + 1) and (C.72)
J J S
5

L S J 5 SLSJ = S S L L (1)k+L+S+J (2J + 1) 2S + 1
" #
S S 1
S(S + 1). (C.73)
J J L
Hence, the total spin and the total orbital angular momentum do not change during
a M1 transition in LS coupling! In contrast, transitions with J = J 1 or J = J
are possible, provided 0 0. We recall that the radial wave functions for different
principal quantum numbers and otherwise identical configuration are orthogonal.
Hence, only within one electronic state | = 0, so that M1 transitions do not

occur between different electronic states. Of course, this strictly holds only for pure
LS coupling. In the case of configuration mixing or jj coupling one has examine
the situation in detail. We finally note, that all above considerations may be applied,
mutatis mutandis, to a hyperfine coupled scheme as well typically prior to evalu-
ating the LS coupling.
C.5 Radial Matrix Elements
To obtain quantitative values for transition probabilities it is necessary to also com-

pute the radial parts of the matrix elements, as indicated by . . . |Tkq | . . . above.
In general, the radial part may completely be pulled out of the matrix element, and
can be computed independently. For single electron wave functions the radial matrix
elements are

r| = r 3 drRn (r)Rn (r). (C.74)
0
They are needed when evaluating probabilities for E1 transitions as described in

Chap. 4 but also for a quantitative description of the S TARK effect discussed in
Chap. 8. The radial eigenfunctions have to be known and these depend on even
though the orbital angular momentum L has dropped out of the angular part of the
matrix element (C.54).
In closed form, these matrix elements may only be derived for the hydrogen atom
and H like ions. Using the known radial functions (2.122) one finds

4 (n 1)! (n 1)!
n r|n = 2 2 (C.75)
n n [(n + )!]3 [(n + )!]3

+1 2 +1 /2 /2 2
L2
n + Ln+ ()re e r dr,
0
with = 2Zr/(na0 ) and = 2Zr/(n a0 ) and the associated L AGUERRE polyno-

mials

n 1
[(n + )!]2 k
2 +1
Ln+ () = (1)k+1 .
(n 1 k)!(2 + 1 + k)! k!
k=0
For not too large values of n and one may evaluate (C.75) in closed form,
using standard analytical programs like Maple, Mupad4 or Mathematica on a PC.
We assemble in Table C.2 some characteristic results for n |r|n in (atomic,
Z scaled) units a0 /Z.
4 Available as part of the Symbolic Math Toolbox add-on for MATLAB.

C.5 Radial Matrix Elements 591
Table C.2 Radial transition n1 1 n2 2 n |r|n

matrix elements for the a0 /Z
hydrogen atom an H-like
atoms 1 0 2 1 128 6/243 1.290

1 0 3 1 27 6/128 0.516

1 0 4 1 6144 15/78 125 0.305

1 0 5 1 250 30/6561 0.209

2 0 3 1 13 824 12/15 625 3.065

2 0 4 1 512 30/2187 1.282

2 0 5 1 576 000 60/5764 801 0.774

2 1 3 0 3456 18/15 625 0.938

2 1 3 2 27 648 180/78 125 4.748
If very high values for n and are required (RYDBERG atoms), there are suitable
alternatives. Typically, it is sufficient to compute n , 1|r|n, for n = n. Based
on earlier work a closed form has been given e.g. by T OWLE et al. (1996) (again in
units of a0 /Z):

()n (n + 1)!(n + )! 1/2

n , 1r|n, =
4Z (n )!(n 1)!

(4n n) +1 (4n n) (n n )N 1 N !
(C.76)
(n + n)n +n (2 1 + )!(N )!
2 F1 (, 2 + 1; N + 1; x)

(4n n) (n n )N 1 N !

(n + n )2 (2 1 + )!(N )!

2 F1 , 2 + 1 ; N + 1; x .
Here 2 F1 (, ; ; x) is the hypergeometric series

()t ()t
2 F1 (, ; ; x) = xt (C.77)
t
( )t t!
with ()t = ( + 1) ( + t 1)
2
x = n n /4n n

= min nr , nr N = max nr , nr ,

= min nr , nr + 2 N = max nr , nr + 2 ,
and nr = n and nr = n 1 are radial quantum numbers. The hypergeometric

series always terminates since and are negative numbers. Extended tables on
line strengths S(n sj n sj ) or intensities (2j + 1)A(jj ) (with the E INSTEIN
coefficient A(jj ) corresponding to (4.114)) are found in T OWLE et al. (1996) for H
like RYDBERG atoms. Since for sufficiently high n and all atoms become hydrogen
like, these results are of general relevance.

usually refers to wavelengths from 1 mm to 1 m corresponding to frequencies be-
tween 0.3 GHz to 300 GHz; ISO 21348 (2007) defines it as the wavelength range
between 1 mm to 15 mm.
References
R ACAH, G.: 1942. Theory of complex spectra II. Phys. Rev., 62, 438462.
T OWLE, J. P., P. A. F ELDMAN and J. K. G. WATSON: 1996. A catalog of recombination lines
from 100 GHz to 10 microns. Astrophys. J. Suppl. Ser., 107, 747760.
Parity and Reection Symmetry
D
The concept of parity conservation plays a key role in modern physics and is used
throughout this book. We present here a short introduction to parity and reflection
symmetry from an atomic and molecular physics perspective.
D.1 Parity
refers to spatial symmetry of states and is defined by:
The parity operator P

P(r) = (r). (D.1)
We have described the central force problem in polar coordinates. With r r

they change as (r, , ) (r, , + ). As long as the interaction is spatially
isotropic (e.g. the C OULOMB potential) inversion has no influence on the Hamilto-
nian, hence P and H commute:
H
[P, ] = 0.
Thus, the eigenfunctions of the Hamiltonian are also eigenfunctions of the parity
operator. We distinguish states with positive and negative parity (alternatively also
has eigenvalues Pat = 1:
called odd and even), i.e. P

P(r) = (r) = (r).
The product of two functions with unequal parity has negative parity, while the
product of two functions with the same parity has positive parity. We note that the
integral over all space disappears for a function F (odd) (r) with odd parity:
2
F (odd) (r)d 0. (D.2)
0 0
For atomic orbitals we write explicitly

n m (r) = PR
P n (r)Y m (, ) = Rn (r)Y m ( , + ).

594 D Parity and Reection Symmetry
With the definition of the spherical harmonics (B.28) one finds
Y m ( , + ) = (1) Y m (, ),
so that the parity of the states n m (r) is Pat = (1) , hence, spherical harmonics
which differ by = 1 have opposite parity.
Specifically, the relation
= r
Pr (D.3)
holds, and we may say that the dipole operator D = er has odd parity. As trivial
as that statement may seem, it has significant consequences in the context of selec-
tion rules for optical transitions. As discussed in detail in Chap. 4, for an optical
2
transition to occur, the integral 0 0 Y b mb (, )rY a ma (, )d must not disap-
pear. Thus, with (D.2) and (D.3) the product Y b mb (, )Y a ma (, ) must have odd
parity and transitions are thus only allowed if the parity of initial | a ma and final
state | b mb is different. A dipole transition thus connects only states with opposite
parity. The selection rule for single electron transitions is = 1.
Somewhat more general expressing the particle point of view the photon is
said to have odd parity, Pph = 1. The selection rule (4.86) for dipole transitions
may thus be inferred directly and elegantly from a general parity conservation rule
for the whole system including atom and photon. Since the photon with its negative
parity disappears during absorption and appears in emission, absorption or emission
of a photon must be accompanied by a change of parity in the atomic system, hence
the selection rule = 1.
D.2 Multi-electron Systems
The above discussion tacitly assumed a single electron, characterized by | m , ig-

noring its spin. The parity concept may, however, be extended naturally to multi-
electron systems, assuming that the spatial part of the total wave function |LM can
be written as a product of single electron functions (or as a linear combination of
products such as a Slater determinant). The total parity of an N electron system
is then the product of the parities of all its electron orbits. Thus, a configuration
{n1 1 n2 2 . . . nk k . . . nN N } has even or odd parity depending on whether
N
(1) 1 i
= (1) = 1
is positive or negative, respectively (or alternatively whether = 0 or 1). In principle

one has to sum over all electrons of an atom; however, in practice it usually suffices
to sum just over all electrons in open shells.
D.3 Reection Symmetry of Orbitals Real and Complex Basis States 595
D.3 Reection Symmetry of Orbitals Real and Complex Basis

States
At this point some general remarks about the choice of basis states or wave functions
and their symmetry properties are in order. For simplicity we first discuss again a
single electron and ignore its spin.
To study different atomic problems some choices of basis states may be more
suitable than others. The one most commonly used in atomic physics for the angu-
lar part of the wave functions (states) is the representation by (complex) spherical
harmonics.
There are, however, a variety of cases where another symmetries may be more
adaptable to a given problem, such as the interaction with an external electric field
(see Sect. 8.2). In that case the eigenstates are linear combinations of states with
M and |M| remains a good quantum number, while the sign of M is indefinite.
Similarly, in optical transitions the polarization of the exciting light may lead to
particular linear superpositions of different basis states as discussed in Chap. 4. All
these linear combinations are per definition also pure states, and in principle, one
can find an infinite number of such states which, properly orthonormalized, could
also serve as a basis set. In standard quantum mechanical textbooks the | m (or
|LM ) states are somewhat overemphasized and one is led to believe that these are
the only suitable basis states for atomic systems. That is only true if a system is
characterized by the magnitude of angular momentum z .
2
L and its z-component L
In addition, the measurement to be described must select states which diagonalize
both operators simultaneously, so that
2
L |LM = L(L + 1)2 |LM
(D.4)
z |LM = M|LM .
L
And indeed, in the presence of a magnetic field in z-direction a typical atomic

Hamiltonian remains diagonal in this basis as discussed in Sects. 2.7 and 8.1. Such
properties of the complex angular momentum basis, together with the fact that the
corresponding matrices for coupling and rotating angular momenta are extensively
tabulated, makes them very attractive. The price to be paid is that linear combi-
nations have to be used already for rather simple applications, such as atoms in
external electric fields, dipole excitation by linearly polarized light, as well as for
most of molecular physics and chemistry.
Thus, one may construct real tensor operators Tkqp (observables, spherical har-
monics etc.) as linear combinations from the complex ones
with Tkq = (1)q Tkq

for 0 < q k and p = 1:
i(p1)/2 i(p1)/2
Tkqp = (1)q Tkq + p Tkq = (1)q Tkq + p Tkq

(D.5)
2 2
while for q = 0: Tk0+ = Tk0 and Tk0 = 0.
Table D.1 Renormalized real spherical harmonics Ckq for k = 1 to 3

k p = +1 p = 1
0 C00+ = 1 C00 = 0
1 C10+ = cos C10 = 0
1 C11+ = sin cos C11 = sin sin
2 C20+ = (3 cos2 1)/2 C20 = 0

2 C21+ = 3 sin cos cos C21 = 3 sin cos sin

2 C22+ = ( 3/2) sin2 cos 2 C22 = ( 3/2) sin2 sin 2
3 C30+ = (5 cos3 3 cos )/2 C30 = 0

3 C31+ = ( 6/4) sin (5 cos2 1) cos C31 = ( 6/4)(5 cos2 1) sin

3 C32+ = ( 15/2) cos sin2 cos 2 C32 = ( 15/2) cos sin2 sin 2

3 C33+ = ( 10/4) sin3 cos 3 C33 = ( 10/4) sin3 sin 3
In complete analogy the real angular momentum states are defined by
i(p1)/2
|J Mp = (1)q |J M + p|J M for 0 < M J (D.6)
2
|J 0+ = |J 0 and |J 0 = 0 for M = 0.
Note that the relations between real and complex unit polarization vectors (4.7) are
also constructed according to that scheme. And by inserting (B.20) into (D.5) we
obtain the real spherical harmonics

2k + 1
Ykqp (, ) = Ckqp (, ) where
4

2(k q)! q
Ckq+ (, ) = P (cos ) cos(q) for 0 < q k (D.7)
(k + q)! k

2(k q)! q
Ckq (, ) = P (cos ) sin(q) for 0 < q k
(k + q)! k
Ck0+ (, ) = Pk0 (cos ) and Ck0 (, ) 0 for q = 0,

q
with the associated L EGENDRE polynomials Pk which can be derived from (B.22).
Some explicit expressions for the renormalized real spherical harmonics Ckqp (, )
are given in Table D.1. Their orthogonality relations are
2
4
d Ckqp (, )Ck q p (, ) sin d = kk qq pp . (D.8)
0 0 2k + 1
Rotation of the real tensor operators, real states and real spherical harmonics is
achieved by (E.25) using the real rotation matrices (E.26)(E.28).
D.3 Reection Symmetry of Orbitals Real and Complex Basis States 597
While the real spherical harmonics are eigenfunctions of 2z ,

2
L and L
2
L Ykqp = k(k + 1)2 Ykqp
2z Ykqp = q 2 2 Ykqp
L with q = 0, 1, . . . , k, (D.9)
they are neither eigenfunctions of Lz nor of L

x or L
y . However, the real spherical
harmonics and the corresponding real states (electron orbitals) are characterized by
a well defined reflection symmetry not only in respect of the xy plane (which is
also a plane of reflection symmetry for the complex basis) but also in respect of the
xz and yz plane. Reflection of a wave function at the xz plane e.g. is equivalent
to replacing by . In case of the complex basis this leads to Ykq (, ) =
(, ) = (1)q Y
Ykq kq (, ). However, by applying this relation to the definition
of the real spherical harmonics (D.5)
one finds for a reflection at the xz plane

v (xz)Ykqp = pYkqp with p = 1, (D.10)
where we have introduced the reflection operator

v (xz).
Often one wants to express (D.7) in Cartesian coordinates. We apply a very use-
ful relation communicated by C AOLA (1978). In our notation the so called solid
harmonics Ykq (x, y, z) are (renormalized):
Ckq (x, y, z)

4
= Ykq (x, y, z) = r k Ckq (, ) (D.11)
2k + 1

(kq)/2
(1)j (x + iy)j +q (x iy)j zkq2j
= (1)q (k q)!(k + q)! .
22j +q (j + q)!j !(k q 2j )!
j =max(q,0)
Inserting this into (D.5) leads to explicit expressions for the renormalized real solid
harmonics Ckqp . Table D.2 summarizes the most important cases.
Figure D.1 illustrates graphically the angular dependence of the three real basis
orbitals for k = = 1: |px , |py , and |pz (short: px , py and pz orbital, respec-
tively) and compares them to the usual complex basis orbitals |1 1 and |10 . Ob-
viously, the real spherical harmonics are represented by dumbbell like states parallel
to the x-, y-, and z-axes of the atom, respectively. In contrast the complex basis con-
sists of two doughnut like orbitals (|1 1 with opposite sense of orientation) while
the |10 orbital is also dumbbell like and identical to |pz .
One may think of preparing these basis states by optical dipole excitation.
However, since electrons have also a spin, and m are typically not
good quantum numbers. Thus, a complete optical preparation of the orbitals shown
in Fig. D.1 is only possible in very special cases. If e.g. one can make use of a
Table D.2 Renormalized k p = +1 p = 1

real solid harmonics Ckq for
k = 1 to 3 0 C00+ = 1 C00 = 0
1 C10+ = z C10 = 0
1 C11+ = x C11 = y
2 C20+ = (3z2 r 2 )/2 C20 = 0

2 C21+ = 3zx C21 = 3zy
2
2 C22+ = 3(x y 2 )/2 C22 = 3xy
3 C30+ = z(5z2 3r 2 )/2 C30 = 0

3 C31+ = 6x(5z2 r 2 )/4 C31 = 6y(5z2 r 2 )/4

3 C32+ = 15z(x 2 y 2 )/2 C32 = 15zxy

3 C33+ = 10x(x 2 3y 2 )/4 C33 = 10y(3x 2 y 2 )/4
z z
Fig. D.1 Angular part of
atomic p orbitals (spherical p =+1
harmonics) in the complex z
(top row) and real (bottom |+1 |0 |-1
row) basis. Plotted are the
squared moduli |Y1q |2 and
|Y1q |2 , respectively. The
colour shading indicates x y x y x y
regions of different signs of
the functions. The and
$ signs indicate positive and
negative reflection symmetry
in respect of the xy plane,
z p =+1
z
p =+1 z p = 1
| px | pz | py
while reflection symmetry in
respect of the xz plane
according to (D.10) is
characterized by red p = +1
and p = 1 x y x y x y
1S 1 P1 transition, it is indeed possible to prepare all five different basis states

0
individually by optical dipole excitation. In such a case, for example in a two elec-
tron system like He, the electron spin is compensated by another electron in the
system, but only one electron is excited:
(a) The complex basis set requires circularly polarized light for its preparation: left
and right hand circularly polarized, propagating along the z-axis to excite the
|1 +1 (equivalent to Y1+1 ) and the |1 1 (equivalent to Y11 ) states, respec-
tively. Linearly polarized light propagating perpendicularly to the z-axis and
with its E vector aligned parallel to the z-axis excites the |10 state (equivalent
to Y10 ).
(b) The real basis set requires linearly polarized light for its preparation: the |pz
state (identical to |10 ) is excited as just described, while the preparation of |px
D.4 Reection Symmetry in the General Case 599
Fig. D.2 Schematic z z z

illustration of a reflection
^
v (xz) in the xz plane being
constructed from an inversion y
P y y
at the origin (P ) and a x x x
rotation through about the
^
y-axis v (xz)
and |py requires linearly polarized light with its E vector parallel to x and y,
respectively.
Generally speaking, the real basis functions are the basis of choice when the sign
of the angular momentum is irrelevant, i.e. whenever there is no specific preferred
orientation. Thus, they are relevant if external electric fields are present. They play
a particularly important role in molecular physics and quantum chemistry as dis-
cussed in detail in Chaps. 34 of Vol. 2, but also in collision problems. There, the
electrostatic field between different, neighbouring atoms defines a suitable z-axis
and is typically very strong so that spin quantum numbers can be treated as a mere
afterthought. Thus, the orbital angular momentum quantum numbers and the abso-
lute value of their projection onto this axis become good quantum numbers. Real
atomic orbitals (AOs), as depicted in the bottom row of Fig. D.1, and linear com-
binations of them attached to different atoms are the basis for a quantitative
treatment of electronic wave functions in molecules.
Reflection symmetry is also an important property in the context of collision
problems. As we shall discuss in Chaps. 68, Vol. 2, reflection symmetry in respect
of the scattering plane is conserved during a binary collision for the combined wave
function of the interacting particles. Thus, either reflection symmetry of both parti-
cles remains constant during the collision, or both particles change their reflection
symmetry in respect of the collision plane.
D.4 Reection Symmetry in the General Case
Since reflection symmetry is such an important property of atomic and molecular

states, we now generalize the concept. So far, the discussion was restricted to the
orbital part of the state vectors, i.e. to the integer values of the total angular momen-
tum J . In these cases reflection symmetry is an intuitively obvious property of the
wave functions. However, it is possible to extend this concept to half integer angular
momenta as well.
We start by noting that a reflection v in respect of a given plane may be viewed
as an inversion at the origin with subsequent rotation through an angle around
the axis perpendicular to this plane. This is illustrated in Fig. D.2. Inversion at the
origin implies, that the parity operator P multiplies the state vectors |J M with
(1) , where is an odd or even integer depending on whether the parity of the
whole system is odd or even (typically determined by the orbital angular momentum
of the wave function). The rotation through is achieved with the help of rotation
matrices described in Appendix E. Thus, the state |J M after reflection in the xy

plane becomes

J

v (xy)|J M = (1) DJM M (, 0, 0)J M = eiM (1) |J M
M
= (1)+M |J M . (D.12)
We note here in passing that with this relation the |+1 and |1 p states in Fig. D.1
have both positive reflection symmetry in respect of the xy plane, while the |0 state
has negative reflection symmetry as intuitively obvious.
Reflection at the xz plane leads to

v (xz)|J M = (1) DJM M (0, , 0)J M
M
= (1) dMM
J
()|J M = (1)+M+J |J M . (D.13)
Here we have used the symmetry relation (E.9) for the reduced rotational matrix.
From (D.12) and (D.13) we see that the standard complex |J M representation are
eigenstates of the xy plane reflection operator but not of the operator for reflection
in the xz plane. In analogy to (D.5) one may, however, construct also for the general
case (J integer or half integer) states of specified reflection symmetry in respect of
the xz plane:

|J Mp = v(p, M) (1)M |J M + (1)+J p|J M with (D.14)

1/2 for M = 0
v(p, M) = 1/ 2 for 0 < M J and p = 1 or +i (D.15)

i/ 2 for 0 < M J and p = 1 or i.
Here p = (1)J is the eigenvalue of the reflection operator v (xz) in respect of

the xz plane. Applying (D.13) to (D.14) one easily verifies that |J Mp are orthonor-
malized eigenstates of
v (xz):
v (xz)|J Mp = p|J Mp .
(D.16)
For integer values of J = k, M = q and = J the previous relations (D.5) are

recovered, and p = 1. For half integer values of J the eigenvalues of the v (xz)
reflection are obviously p = i. The same eigenvalues are found for reflection in
the xy plane. However, half integer states with M belong to different reflection
symmetry, while for integer values of J (spherical harmonics) reflection symmetry
depends on |M|.
In general, the |J Mp states are not eigenstates of any of the components of
the angular momentum operator. The matrix elements of Jx , Jy , and Jz may be
derived from the definition (D.14) with (B.4) and (B.14)(B.17). In general one has
to distinguish different cases such as J, M = 0, 1/2, and 1 as well as p = 1 or i.
Thus, we note here only a few special aspects.
D.4 Reection Symmetry in the General Case 601
1. The states defined by (D.14) are eigenstates of Jz2 . One easily derives
"
+1 for p = +1 or +i

Jz |J Mp = iM J M(p)
1 for p = 1 or i
so that Jz2 |J Mp = M 2 2 |J Mp . (D.17)
2. Jz changes the reflection symmetry from +(1)J to (1)J and vice versa, and
so does Jx :

Jx |J Mp = a(M)J (M + 1)(p) + a(M)J (M 1)(p) (D.18)
as one verifies for J 1 and M > 0 easily with some algebra. In contrast Jy
does not change reflection symmetry:

Jy |J Mp = a(M)J (M + 1)p a(M)J (M 1)p (D.19)
where a(M) is typically1 given by

a(M) = J (J + 1) M(M + 1).
2
3. As specific examples, we have already discussed the p orbitals |px = |11+ ,
|py = |11 and |pz = |10+ depicted in the lower part of Fig. D.1. Without
proof we note here in passing the relations:
Jy |px = i|pz

Jy |pz = +1|px (D.20)
Jy |py = 0.
4. The simplest, and somewhat exceptional case with half integer J is a pure spin
state with = 0, J = 1/2, describing e.g. an electron in a 1 S1/2 state. The basis
states in the |J M representation are |1/2 1/2 and |1/2 1/2 , with spin point-
ing in the +z- and z-direction, respectively. Reflection on the xy plane gives
according to (D.12)

1 1 1 1 1 1 1 1

v (xy)
= i and
v (xy) = i . (D.21)
22 22 2 2 2 2
We see that the up and down spin states have opposite reflection symmetry in
respect of the xy plane. The corresponding states with defined xz reflection sym-
metry are obtained from (D.14) ( = even) with p = i:
1 These formulas hold for all J 1 and M 1. The other cases are easily derived individually
from the definition (D.14) with (B.4) and (B.14)(B.17).


1 1 1 1 1

|+i = i (D.22)
2 22 2 2

1 11 1 1
|i = i
+ . (D.23)
i 2 22 2 2
These states are identical to (E.17) and (E.18), respectively, obtained by coor-
dinate rotation for a spin state pointing into y direction. Reflection at the xz
plane gives with (D.13):
v (xz)|+i = i|+i and

v (xz)|i = i|i . (D.24)
In this particular case (and only in this one), these xy reflection symmetric states
are also eigenvectors of Jy . With (B.4) and (B.14)(B.17) one easily verifies that

Jy |i = |i . (D.25)
2
We finally apply the concept of reflection symmetry to optical transitions treated
in detail in Chap. 4. In Sect. 4.1.2 we have introduced linear (ex , ey ) and circular
(e+1 , e1 ) polarization vectors. We may write |ex , |ey and |e+1 , |e1 as the real
and complex representation of the corresponding photon states, respectively. Then
the definitions (4.4)(4.7) are in complete agreement with (D.14). Since for the pho-
ton J = = 1, M = 1, we have p = +1 for |ex and p = 1 for |ey , i.e. |ex
and |ey have positive and negative reflection symmetry in respect of the xz plane.
Since the (fully quantized) Hamiltonian for the system photon + atom is invariant
under reflection at any plane through the centre of the atom, the reflection symmetry
of the state vectors describing the combined system must conserve reflection symme-
try. Depending on whether we use light which is polarized in the x- or y-direction,
the atom changes or conserves its reflection symmetry in an optical transition as
indicated by the following scheme:
|J Mp; ex |J Mp; . . .

|J Mp; ey J M(p); . . . .
For example, the prototype 1 S0 1 P1 transition induced by light linearly polarized

in the x-direction induces no change in the reflection symmetry: The 1 S0 state has +
reflection symmetry in respect of the xz plane as well as the 1 P1 dumbbell pointing
into x-direction (see Fig. D.1). The y-polarized light excites the dumbbell pointing
into the y-direction, having negative reflection symmetry.
A somewhat less trivial example is depicted in Fig. D.3. In a 2 S1/2 2 P1/2
transition the initial state |Ja = |+i corresponds to (D.22), with its spin pointing
into the +y-direction it has p = +i reflection symmetry in respect of the xz plane.
Since after the excitation to a P state, the parity of the atom changes ( = 0 1),
the final state |Jb with angular momentum pointing into +y-direction has now
p = i symmetry. When exciting with ey light, the combined system |Ja |ey prior
Fig. D.3 Illustration of z z z z

reflection symmetry changing
and non-changing processes Ja Jb Ja Jb
in a 2 S1/2 2 P1/2 transition y y e y y
x
induced by light linearly ey
polarized in x- and x x x x
y-directions, respectively k k
2S
1/2
2P1/2 2S
1/2
2P1/2
to the transition has +i (1) = i reflection symmetry, thus the final state must
also have i symmetry, which corresponds to |Jb pointing into +y-direction. In
contrast, for excitation with ex polarized light, the combined initial system |Ja |ex
has +i (+1) = i symmetry, which must be conserved and corresponds to reversing
the orientation of the spin into |Jb .
Circularly polarized photons |e+1 or |e1 propagating into z-direction have
both positive reflection symmetry in respect of the xy plane (in analogy to the |+1
and |1 p states in Fig. D.1) and cannot change this reflection symmetry of the
atomic states. Formally one sees this from (D.12), since in a dipole transition both
and M change by 1 in respect of the photon frame where z k. Thus, (1)+M
remains constant during the excitation process and reflection symmetry in respect
of the xy plane does not change.
References
C AOLA, M. J.: 1978. Solid harmonics and their addition theorems. J. Phys. A, Math. Gen., 11,
L23L25.
Coordinate Rotation
E
We compile here the most important formulas for rotation of coordinate systems
and their consequences on angular momentum states and tensor operators without
going into details of the derivation.
E.1 EULER Angles
One specifies a coordinate rotation by three E ULER angles (, , ). As shown in

Fig. E.1 the rotation is performed in three steps:

(xyz) x y z x y z x # y # z# .
First, one rotates the system around its original z-axis (z = z) through an angle .
The following rotation through an angle around the new y axis (y = y ) brings
z into the direction z and x into direction x . The final rotation through an angle
is around the thus tilted z = z# axis (x x # and y y # ).
Fig. E.1 Definition of the z, z'

Euler angles (, , ) for
rotation of a coordinate z'', z# y#
system xyz
y'
y''
y

x x#
x' x''

606 E Coordinate Rotation
E.2 Rotation Matrices
The operator achieving this rotation is derived from the angular momentum opera-
tor
J:

D( ) = exp(i Jz ) exp(i Jy ) exp(i Jz ).
The corresponding rotation matrix elements, more precisely the matrix elements
J of the rotation group, may be derived with
of the irreducible representation D
some angular momentum algebra (following the notation of B RINK and S ATCHLER,
1994),1 and are written as

DJMN ( ) = J M|D( )|J N = exp(iM)dMN
J
() exp(i N ), (E.1)
with the reduced rotation matrix:
[(J + M)!(J M)!(J + N )!(J N )!]1/2

J
dMN () = (1)t
t
(J + M t)!(J N t)t (t + N M)! (E.2)
(cos /2)2J +MN 2t (sin /2)2t+N M .
Special cases:

4
DJM0 ( ) = CJ M (, ) = Y (, ) (E.3)
2k + 1 J M

(J M)! M
J
dM0 () = (1)M P (cos ) if M > 0 (E.4)
(J + M)! J
J
and d00 () = PJ (cos ). (E.5)
We note some important symmetry properties
DJMN ( ) = (1)MN DJMN ( ) = DJN M ( ), (E.6)

J
dMN () = (1)MN dNJ M () = dN
J
M () = dN M ()
J
(E.7)

= (1)J M dMN
J
( ) = (1)J +M dMN
J
( + ). (E.8)
J
dMN () = (1)J +M MN and (E.9)
J
dMN (0) = MN = (1)2J dMN
J
(2). (E.10)
1 Here the coordinate system is rotated. An alternative convention, also found in the literature, is to
rotate the states.

E.2 Rotation Matrices 607
Explicitly we communicate formulas for the frequently used cases J = 1/2, J = 1,

and J = 2 according to B RINK and S ATCHLER (1994):

1 1 1 1
d 12 1 = d 2 1 1 = cos d 2 1 1 = d 12 1 = sin (E.11)
2 2
2 2 2 2 2 2 2 2

2
d11 = d1 1 = cos
1 1 2 d11 = d11 = sin
1 1
2 2 (E.12)
sin
d01 = d10 = d01 = d10 =
1 1 1 1 d00 = cos .
1
2
2
d22 = d22
2
= cos4 (/2) (E.13)
2 = d2 = d2
20 = d02
2 = d 2
12 0 = d21 = d12
d20 2 2 2
02 d21

= ( 3/8) sin2 = sin (1 + cos )/2
2
d21 = d12
2 0 = d 2
21 = d12
2
2
d22 = d22
2 = sin4 (/2)
= sin (1 + cos )/2
2
d11 = 2
d11 2
d11 = d11
2
= (2 cos 1)(1 + cos )/2 = (2 cos + 1)(1 cos )/2

2 = d2
01 = d01 = d10
d10 2 2
2 = (3 cos2 1)/2
d00
= ( 3/2) sin cos .
An angular momentum state |J N # in the new system may be expressed by

eigenstates |J M in the original system as:

J |J N =
|J N # = D
|J M J M|D( )|J N = |J M DJMN ( )
M M

= exp(i N ) |J M exp(iM)dMN
J
(). (E.14)
M
The components Tkq ( # , # ) of tensor operators of rank k (e.g. Ykq or a vector with
k = 1) can be expressed by tensor components in the old system

Tkq # , # = Tkq (, )Dkqq ( )
q

= exp i q k
Tkq (, ) exp(iq)dqq () (E.15)
q
with , and # , # being the angles in the old and new system, respectively.
We illustrate the procedure by a particularly simple and important example,
a spin s = 1/2 state |1/2 1/2 # pointing into the = , = direction in the orig-
1/2
inal coordinate frame xyz. Inserting for this case dMN () from (E.11) into (E.14)
we obtain:
(a)
(b)
Fig. E.2 Rotation to describe a spin s pointing into (a) +y-direction, (b) +x-direction
#
1 i i 1 i 1
= exp exp cos + exp sin .
2 2 2 2 2 2 2 2
(E.16)
(For simplicity of writing we have dropped here J = 1/2.) The third rotation angle
around the final z# axis is somewhat arbitrary and just contributes a free phase
factor.2 Figure E.2(a) indicates the rotation angles for one specific case where the
spin s points into the +y-direction of the (xyz) coordinate system. Thus, the oblig-
atory rotation angles are = = /2 and = = /2. We choose = 3/2,
make use of (E.11)(E.13), and thus obtain from (E.16) for this particular spin state

1 #(y) 1 1 1
+ = i+ . (E.17)
2 2 2 2
If the spin s points into the y-direction, we make exactly the same rotation and
find from (E.14)

1 #(y) 1 1 1
= i + + (E.18)
2 i 2 2 2 .
Somewhat more compact, we designate the spin states parallel and antiparallel to
the y-axis with by | (y) and | (y) , respectively, and those which refer to the z-axis
with | and | . Equations (E.17) and (E.18) may then be written:
1 1
| (y) = i| | and | (y) = i| + | . (E.19)
2 i 2
2 Note, however, that for spin

1/2 this rotation angle is defined modulo 4 rather than modulo 2
(one more turn around the z axis changes the sign). For example, the angle = 3/2 in Fig. E.2
has not the same effect as = /2.
E.3 Entangled States 609
For completeness we also describe a state where the spin s points into the x-
direction. As illustrated in Fig. E.2(b) the E ULER angles in this case are (0, /2, )
and we obtain

1 #(x) 1
= | | . (E.20)
2 2
All these states are of course eigenstates of the squared spin angular momentum

S 2 , with spin quantum number s = 1/2. The eigenvalue of S 2 are 2 s(s + 1) =

3 /4. And specifically | and | are also eigenstates of Sz , the spin projection
2
onto the z-axis, with eigenvalues /2 and /2, respectively. What about the states
|1/2 (x,y) ?
Here is a little exercise for the active reader, which can be solved with a few
lines of algebra (for complex quantities): Using (B.4), (B.16) and (B.17), show that
the spin states | (y) and | (y) are indeed eigenstates of the spin component in y
direction with eigenvalues /2:

Sy | (y) = | (y) and
Sy | (y) = | (y) . (E.21)
2 2
And the same holds of course for the corresponding spin states pointing in x-
direction.
E.3 Entangled States
For many-particle systems we have to construct spin functions which account for
the fact that identical particles are indistinguishable. We have discussed this in the
main text on several occasions, see e.g. Table 6.1 in the context of fine structure and
Sect. 7.3 where specifically the He atom is treated. Let us take the latter example,
i.e. the two electrons in the He ground state (1s)2 1 S0 , or any singlet state of two
electrons in the continuum after ionization (with a symmetric spatial wave function).
The two spins of the electrons are anti-parallel and since we cannot distinguish
which of these two fermions has which spin we have to construct an antisymmetric
description of their joint state, according to (7.36)
0
= 1 (1)(2) (1)(2) , (E.22)
0
2
where the numbers in brackets refer to electron (1) and electron (2), respectively.
Obviously, this state of the two electrons is inseparable, i.e. it cannot be written
as product of two wave functions of the individual electrons. One calls such states
entangled.
Equation (E.22) refers to a particular coordinate frame and | and | reflect
the spin components with respect to its z-axis. The total (spin) angular momentum
operator is
S = S1 +
S 2 , and the singlet state (E.22) is constructed such that the
respective eigenvalues of S and
2
Sz are zero:3
2
Sz 00 = (
S2z )00 = 0 and
S1z + S 00 = 0. (E.23)
One may now be curious what happens if we define this entangled, antisymmetric
spin state with respect to another axis, say the y-axis. This may be done in complete
analogy to (E.22) from spin states with angular momentum parallel and antiparallel
to the y-axis:
0
= 1 (1)(2) (y) (1)(2) (y) . (E.24)
y
2
In full analogy to (E.23) we have now Sy |y0 = 0, i.e. the projection of the total spin
onto the y-axis is also zero. The relation (E.19) between the | (y) and | (y) states
and the | and | states parallel to the z-axis may now be inserted into (E.24):
with just a few lines of algebra (in which many terms cancel) one recovers again
(E.22), i.e. the singlet state with respect to the z-axis. Thus, the definition of this
entangled state does not depend on the coordinate axis and its angular projection is
zero in all coordinate systems.
So much about coordinate rotation for entangled states. Entanglement is a very
active field of current research with far reaching consequences, not only for funda-
mental conceptual and philosophical reasons (dating back to the famous E INSTEIN ,
ROSEN , P ODOLSKY 1935 paper, quoted by now over 5000 times), but also in view
of a number of potential future applications such as quantum cryptography or quan-
tum computation. We cannot enter here in a detailed discussion of this wide field
but refer the reader to some relatively recent reviews and references quoted there
(T ICHY et al., 2011; H ORODECKI et al., 2009; R AIMOND et al., 2001), see also the
respective chapters in B LUM (2012).
E.4 Real Rotation Matrices
For reference we also communicate the rotation procedure (FANO, 1960) for the
real harmonics, real states and real tensor operators as defined in Appendix D. In
this case, the transformation is achieved by

# #
Tkq p , = Tkqp (, )Dkqp,q p ( ) (E.25)
qp
3 The first relation is obvious, since each of the operators S1z and
S2z acts only onto one of the
states which have the eigenvalues /2, i.e. they compensate each other. The latter relation may
easily be verified as an exercise with the help of (B.8) and (B.17).
References 611
with p, p = 1. The real rotation matrices are obtained by applying the complex
rotation (E.15) onto the defining equations (D.5) and finally back-transformation to
the real basis. One obtains:4
For k q, q > 0:
(E.26)
Dkqp,q p ( )
k
k
(1)q dqq () cos(q + q ) + p dqq () cos(q q )

if p = p
= (1)q

(1)q p dqq
k () sin(q + q ) + d k

qq () sin(q q )

if p = p
For q or q = 0: (E.27)
"
k cos(q + q ) if p = p
Dkqp,q p ( ) = (1)q+q 2dqq ()
sin(q q ) if p = p
specifically Dkqp,0+ ( ) = Ckqp (, ) (see Eq. (D.7))
For q = q = 0: Dk0+,0+ ( ) = d00

k
(). (E.28)
References
E INSTEIN, A., B. P ODOLSKY and N. ROSEN: 1935. Can quantum-mechanical description of
physical reality be considered complete? Phys. Rev., 47, 777780.
FANO, U.: 1960. Real representations of coordinate rotations. J. Math. Phys., 1, 417423.
H ERTEL, I. V. and W. S TOLL: 1978. Collision experiments with laser excited atoms in crossed
beams. In: Adv. Atom. Mol. Phys., vol. 13, 113228. New York: Academic Press.
H ORODECKI, R., P. H ORODECKI, M. H ORODECKI and K. H ORODECKI: 2009. Quantum entan-
glement. Rev. Mod. Phys., 81, 865942.
R AIMOND, J. M., M. B RUNE and S. H AROCHE: 2001. Colloquium: Manipulating quantum en-
tanglement with atoms and photons in a cavity. Rev. Mod. Phys., 73, 565582.
T ICHY, M. C., F. M INTERT and A. B UCHLEITNER: 2011. Essential entanglement for atomic and
molecular physics. J. Phys. B, At. Mol. Phys., 44, 192001.
4 See also H ERTEL and S TOLL (1978), Eq. (15) (some small typos there have been corrected here).
Multipole Expansions and Multipole Moments
F
One often needs suitable approximations for potentials which at large distances are
nearly symmetric around a centre, but not completely. For scalar as well as for vector
potentials, multipole expansions may be derived. We treat here the important case
of an electrostatic potential arising from a charge distribution localized in a small
region around the origin as e.g. encountered when describing the interaction of
atomic nuclei with the electron charge cloud of an atom (see Fig. 9.12 in Chap. 9).
In the final subsection we shall also introduce a more general definition of multipole
tensor operators and multipole moments.1
F.1 Laplace Expansion
We first note that the inverse distance between two points r and r in space may be
expanded for r r into a power series of x = r /r:
1 1 1/2
= 1 2x cos + x 2 (F.1)
|r r | r

1 3 1 5 3
= 1 + x(cos ) + x 2 cos2 + x3 cos3 cos
r 2 2 2 2
where is the angle between r and r . With (B.24) we identify the angle dependent
terms in brackets as L EGENDRE polynomials. With this result in mind, we commu-
nicate the L APLACE expansion which holds for r < r:

r k
1
= Pk (cos )
|r r| r k+1
k=0
1 The term multipole moments is used in the literature for a variety of slightly different defini-
tions. As far as necessary for distinction, we shall add some specification, indicated in (brackets).

614 F Multipole Expansions and Multipole Moments

r k r k (r) (r )
k
= C kq (, )C kq , = C Ck (F.2)
r k+1 r k+1 k
k=0 q=k k=0

r k
k

= Ckqp (, )Ckqp , .
r k+1
k=0 q=0,p=1
In the second line of (F.2) we have made use of the addition theorem (C.22) for
the renormalized spherical harmonics, with , and , being the angles of the
position vectors r and r , respectively. In the following equality we have made use
(r) (r )
of (C.21) and the scalar product notation (C.20) for the two tensors Ck and Ck .
The third line of (F.2) introduces the real (renormalized) spherical harmonics as
defined in Appendix D.3.
F.2 Electrostatic Potential
With these relations we write the electrostatic interaction energy dV12 (r, r ) of a
charge dq, located at a position r , with a point charge e located at r

1 edq 1 e (r) k (r )
dV12 r, r =
= Ck r dq Ck . (F.3)
40 |r r| 40 r k+1
k=0
For an extended electric charge with density (r ), localized around the origin, the
charge element is dq = (r )d3 r , and the total interaction energy with the point
charge e is obtained from

e (r ) 3
V (r) = dV12 r, r = d r. (F.4)
40 |r r|
With (F.3) we may write this as a multipole expansion

e 1
k
V (r) = Qkq Ckq (, ) or (F.5)
40 r k+1
k=0 q=k

e 1
k
in real form = Qkqp Ckqp (, ), (F.6)
40 r k+1
k=0 q=0,p=1
with (charge) multipole moments

Qkq = Ckq , r k r d3 r , or (F.7)

Qkqp = Ckqp , r k r d3 r = Ckqp r r d3 r , (F.8)
F.2 Electrostatic Potential 615
where we have used in the second row for abbreviation the renormalized real solid
harmonics Ckqp as introduced in Appendix D.3.
These multipole expansions, derived from (F.3), are most appropriate if r r.
Thus, it is suitable e.g. for describing the field of an atomic nucleus as seen by an
electron far away from the nucleus (i.e. for 1), or for the far field of an atom in
a well defined polarization state (with 1).
The total charge is given by

q = Q00 = r d3 r . (F.9)
For an electronic charge cloud of an atom with N electrons it is q = N e, and for

an atomic nucleus with atomic number Z (protons in the nucleus) we have q = Ze.
If the charge distribution can be described by an electronic or nuclear wave func-
tion as (r ) = q|n m (r )|2 = q|Rn (r )C m ( , )|2 the angular part of the ma-
trix elements in (F.7) may be evaluated as described in Appendix C.2.2 if nec-
essary applying the recoupling rules from Appendix C.3. In this case, we find with
(C.30) that only even multipole moments are nonvanishing: k = 0 (monopole), k = 2
(quadrupole), k = 4 (octupole), etc.
More generally, for an arbitrary charge distributions, convenient expressions for
the real multipole moments (F.8) are obtained with the renormalized real solid har-
monics (D.11). The components of the dipole moment are:

Dz = Q10+ = z r d3 r Q10 = 0
(F.10)

Dx = Q11+ = x r d3 r

Dy = Q11 = y r d3 r .

If only positive or only negative charges are involved, the dipole moments are made
to disappear by suitable choice of the origin, thus defining a charge centre. They
also disappear if both charges have the same centre (characteristic for atoms, unless
they are placed in an electric field, see Sect. 8.2.8). In contrast, if positive and neg-
ative charges with different centres are to be described (typical for polyatomic and
heteronuclear diatomic molecules) the dipole moment is finite.
With (F.8) and Table D.2 the real quadrupole moments are:

1 2
Q20+ = 3z r 2 r d3 r Q20 = 0
2
3 3
Q21+ = 3 z x r d r Q21 = 3 zy r d r (F.11)

3 2 3
Q22+ = x y 2 r d3 r Q22 = x y r d3 r .
2 2
Note that a variety of slightly different terminologies is used in the literature.2 In

nuclear physics one defines (assuming symmetry around the nuclear z-axis) the
nuclear quadruple moment

Q= 3z2 r 2 r d3 r = 2Q20+ . (F.12)
Explicitly, in Cartesian coordinates the potential expansion (F.6) reads

e q D r 1 C2qp (x, y, z)
V (r) = + 3 + Q2qp + . (F.13)
40 r r 2 qp r5
The components (F.10) of the electric dipole moment D are zero for charges cen-
tred at the origin, while the quadrupole moments Q2qp according to (F.11) are fi-
nite for an anisotropic charge distribution. The renormalized real solid harmonics
C2qp (x, y, z) are found in Table D.2.
F.3 Multipole Tensor Operators
Closely related to the (charge) multipole moments just discussed, but without any
reference to the charge of the particles, one defines irreducible spatial multipole
tensor operators of rank k. They are constructed from the position vector and have
the components
Qkq (r) = r k Ckq (, ) = Ckq (x, y, z) (F.14)

for k q k. (F.15)
using again the (renormalized) solid harmonics (D.11). They may also be written as
real quantities according to (D.5):
Qkqp (r) = r k Ckqp (x, y, z) = Ckqp (x, y, z) (F.16)

for 0 q k and p = 1. (F.17)
For explicit expressions we refer to Table D.2. General rules for tensor operators
have been laid out in Appendix C.
2 E.g.
following JACKSON (1999), in classical electrodynamics one usually expands in terms of the
standards spherical harmonics, multiplying the terms in the sum (F.7) by 4/(2 + 1), so that the
(complex) multipole moments are given as

q m = Y m (, )R (R)d3 R.
F.3 Multipole Tensor Operators 617
The expectation values Qkq , i.e. the matrix elements J M|Q2q | J M aver-
aged over the population probabilities w(M) of | J M states, are called (spatial)
multipole moments:

Qkq = r 2 w(M)J M|Ckq ( )|J M . (F.18)
M
They characterize the particle distribution (e.g. for the protons in an atomic nucleus,
for the electrons in an atom, or for electrons and nuclei in a molecule).
For a given angular momentum J of a quantum state the above definitions (F.14)
and (F.16) allow to construct multipole tensor operators and multipole moments of
rank 0 k 2J . They all have a well defined parity
kq (r) = (1)k Qkq (r)
PQ (F.19)
but also a reflection symmetry in respect of the xy plane (see also Appendix D.4):
v (xy)Qkq = (1)k+q Qkq .

(F.20)
Here v (xy) is the reflection operator in respect of the xy plane.

The real tensor operators (F.16) and the corresponding multipole moments often
allow a more flexible and self evident interpretation of physical situations. They are
constructed such that, in addition to parity (F.19) and reflection symmetry in respect
of the xy plane (F.20), they also have a well defined reflection symmetry p = 1 in
respect of the xz plane:

v (xz)Qkqp = pQkqp . (F.21)
They are appropriate tools to express symmetries e.g. in molecules and interaction
processes such as excitation by photons or collisions. For rotations of the real mul-
tipole moments in space (E.25)(E.28) may be used.
F.3.1 The Quadrupole Tensor
Most commonly used is the quadrupole tensor Q2 with its components Q2q =
C2q (r, , ) according to Table D.2. Higher moments become increasingly com-
plicated. If the quantum state |J M of an object is known exactly, the components
of the quadrupole moment are obtained from the respective matrix elements of the
quadrupole tensor Q2 . These matrix elements can be derived from the W IGNER -
E CKART theorem (C.9):

J M
J J 2
J M Q2q | J M = (1) 2J + 1 J Qk J (F.22)
M M q
with J Q2 J = J C2 J J |r 2 | J . (F.23)
Clearly, systems with angular momenta j = 0 or 1/2 do not have a quadrupole

moment, since the 3j symbols must satisfy the triangular relation (j j 2) = 1.
Specifically, the Q20 component of the quadrupole tensor is

3z2 r 2 3 cos2 1
Q20 = C20 ( ) = = r2 . (F.24)
2 2
With (B.58) its diagonal matrix element is
3M 2 J (J + 1)
J M|Q20 | J M = J C2 J r 2 , (F.25)
(2J + 3)(J + 1)J (2J 1)
and if the radial wave functions Rj (r) are known, the radial matrix element r 2
is obtained from
2
r = J |r | J = r 4 |Rj |2 dr.
2
The extrema of (F.25) for M = J and M = 0 are:

J (2J 1)
J J |Q20 | J J = J C2 J r 2 and (F.26)
(2J + 3)(J + 1)

J (J + 1)
J 0|Q20 | J 0 = J C2 J r 2 . (F.27)
(2J + 3)(2J 1)
The reduced matrix element J C2 J still depends on the coupling scheme of the
states under discussion. For LS coupling detailed expressions for the renormalized
spherical harmonics have been derived in Sect. C.3.1.
Specifically, for a quasi-one-fermion system with spin s = 1/2 (one active atomic
electron or nuclear proton) we use (C.55) to obtain the diagonal matrix elements:

1 1
j mQ20 j m
2 2

1 1
= j mC20 j m j |r 2 |j
2 2

2j m1/2 j j 2 j j 2 2
= (1) (2j + 1) r
m m 0 1/2 1/2 0
1 3m2 j (j + 1) 2
= r . (F.28)
4 j (j + 1)
For the last equality we have used (B.58) for the 3j symbols and the fact that j is
half integer, so that (1)2j = 1. The highest (positive) value is found for m = 0,
the lowest (negative) value for m = j :

1 1

j 0Q20 j 0 = r 2 /4 (F.29)
2 2

1 1 1 2j 1 2
jj Q20 jj = r . (F.30)
2 2 4 j +1
F.3 Multipole Tensor Operators 619
The expectation values of the quadrupole tensor, i.e. the (spatial) quadrupole
moments according to (F.18), may be related to the (charge) quadrupole moments
defined in Sect. F.2 by

Q2q = qQ2q = r 2 J C2 J w(M)J M20|J M . (F.31)
For a prolate (cigar shaped) charge distribution, dominated by low M states, (F.31)
leads to Q > 0, while an oblate (pancake like) distribution is characterized by large
|M| and Q < 0 (see e.g. Fig. 9.15).
In nuclear physics, the quadrupole moment of a nucleus (assuming cylindrical
symmetry around the z-axis) is
traditionally defined by Q = 2eZQ2q .
Note the factor 2eZ (overall charge eZ) as already introduced in (F.12): averaging
J M|Q20 | J M over state populations w(M) is equivalent to integration over
the charge distribution (r) = eZ|(r)|2 . The probability density |(r)|2 refers
to the total wave function for the charged particles of interest, while the population
of the states w(M) depends on the specific experiment.3 For a single active spin 1/2
particle system (i.e. a proton in the nucleus) in a pure |jj state, instead of (F.30)
the nuclear quadrupole moment thus reads (see also B OHR and M OTTELSON, 1998,
Eq. (3-27))

1 1 2j 1
Qsp = 2 jj Q20 jj = eZ j |r 2 |j .
2 2 2j + 2
F.3.2 General Multipole Tensor Operators
Often it is desirable to use a more general form of irreducible tensor operators in

multipole expansions. Rather than constructing them from position operators one
may generate such operators from angular momenta. One straight forward pro-
cedure is to polarize the solid harmonics as introduced by FALKOFF and U H -
LENBECK (1950) for applications in nuclear physics. Using the normalization of
M ACEK and H ERTEL (1974) for the solid harmonics (FANO, 1960), these tensor
operators are obtained from

3
Tkq = (Ji1 Ji2 . . . Jik )i1 i2 . . . ik Ckq (r), (F.32)
i1 i2 ...ik =1
where ij = 1 to 3 stands for x, y and z. The angular momenta are given in a.u., i.e. in
units of . Applying this operation to the explicit expression for the (renormalized)
solid harmonics Table D.2 we obtain Table F.1.
3 For a general formalism see Chap. 9 in Vol. 2.

Table F.1 Multipole tensor operators constructed from angular momenta for k = 1 to 3 (only the
zero component is given for the octupole moment k = 3); for reference we also communicate the
relations with the FANO and M ACEK (1973) orientation (O1 ) and alignment parameters (A1+
and A2+ )
k p = +1 p = 1
0 T00+ = 1 T00 = 0
1 T10+ = Jz T10 = 0
1 T11+ = Jx T11 = Jy = J (J + 1)O1
T20+ = (3Jz2 T20 = 0
2
2 J )

2 T21+ = 3(Jz Jx + Jx Jz ) = J (J + 1) 3A1+ T21 = 3(Jz Jy + Jy Jz )

2 T22+ = 3(Jx2 Jy2 ) = J (J + 1) 3A2+ T22 = 3(Jx Jy + Jy Jx )
T30+ = (15Jz3 9
J Jz + 3Jz )
2
3 etc.
The Tk0 (= Tk0+ ) components are obviously diagonal in J and M. Explicitly, the
matrix elements for k = 1 are
M
J M|T10 |J M = M = J T1 J (F.33)
J (J + 1)
and describe the orientation of the state (nonvanishing angular momentum). The
latter equality follows from the W IGNER -E CKART theorem (C.9) and (B.55). The
matrix elements of T20 characterize the alignment (quadrupole moment or the
anisotropy) of the charge density distribution in a given state:
J M|T20 |J M = 3M 2 J (J + 1) (F.34)
2[3M 2 J (J + 1)]
= J T2 J .
(2J + 3)(2J + 2)2J (2J 1)
The latter follows again from the W IGNER -E CKART theorem, this time using
(B.58).
The reduced matrix elements for rank k = 1 and k = 2 follow directly from (F.33)
and (F.33), respectively:

J T1 J = J (J + 1) and (F.35)

J T2 J = (2J + 3)(J + 1)J (2J 1). (F.36)
For an arbitrary rank k one finds (see e.g. M ACEK and H ERTEL, 1974, and refer-
ences given there; note, however, that the factor (2J + 1)1/2 is specific for the
B RINK and S ATCHLER notation of the reduced matrix elements):

k! (2J + k + 1)!
J Tk J = k . (F.37)
2 (2J + 1)(2J k)!
We point out that these (general) multipole tensor operators and their expectation
values, the (general) multipole moments, may be finite also for odd rank in contrast
to the usual (charge or spatial) multipole moments as discussed above. This is a

direct consequence of their construction by angular moments, rather than by position
vectors.
The W IGNER -E CKART theorem provides a one to one relation between the ma-
trix elements of any irreducible tensor operator of rank k for a set of basis states
|j m say for the spatial multipole tensor operators (F.14) Qkq constructed from
position coordinates to any other irreducible tensor of the same rank in that basis
say the Tkq multipole tensor operator. With (C.12) this relation is simply given by
J Tk J
J M Tkq |J M = J M Qkq |J M , (F.38)
J Qk J
with J Qk J defined by (F.23). We emphasize that (F.38) also holds for the re-
spective expectation values (the multipole moments).
Often one is only interested in the expectation values (multipole moments) of
these tensors in an orbital angular momentum basis with sharp = (see e.g. Ap-
pendix D.2 in Vol. 2). We may then use the explicit expression (C.33) for the reduced
matrix elements of the renormalized spherical harmonics. These are nonvanishing
only for even k and we note that their sign is always negative in contrast to (F.37)!
Thus, the multipole moments Tkq (constructed from angular momenta) have the
opposite sign from Qkq (which are built from spatial coordinates)!
Expectation values of these multipole tensor operators are widely used to char-
acterize anisotropy and orientation of atomic or molecular systems in particular
if these cannot be described by pure states. Strictly speaking, they are measured in
units of k but one usually drops the unit. We shall refer to them as multipole mo-
ments as long as no confusion can arise with the (spatial) multipole moments con-
structed from position vectors.4 More details will be discussed in the context of the
density matrix in Chap. 9 and Appendix C, Vol. 2. In addition, we shall introduce
there yet another kind of irreducible tensor operators to characterize anisotropic
state populations, the so called statistical tensor operators. Their expectation values
are the state multipoles which again can be related by simple numerical factors with
any other irreducible representation of tensor operators, in analogy to (F.38).
is important to keep in mind that the sign of Tk0 is opposite to that of Qk0 . When speaking
4 It
about shapes of charge distributions we have to note that oblate implies Q20 < 0 and T20 > 0,
while a prolate shape is encountered if Q20 > 0 and T20 < 0.
References
B OHR, A. and B. R. M OTTELSON: 1998. Nuclear Structure, vol. 1: Single-Particle Motion. Sin-
gapore: World Scientific, reprint from 1969 edn., 471 pages.
FALKOFF, D. L. and G. E. U HLENBECK: 1950. On the directional correlation of successive nu-
clear radiations. Phys. Rev., 79, 323333.
FANO, U.: 1960. Real representations of coordinate rotations. J. Math. Phys., 1, 417423.
FANO, U. and J. H. M ACEK: 1973. Impact excitation and polarization of emitted light. Rev. Mod.
Phys., 45, 553573.
JACKSON, J. D.: 1999. Classical Electrodynamics. New York: John Wiley & Sons, 3rd edn., 808
pages.
M ACEK, J. and I. V. H ERTEL: 1974. Theory of electron-scattering from laser-excited atoms.
J. Phys. B, At. Mol. Phys., 7, 21732188.
Convolutions and Correlation Functions
G
In this appendix all distribution functions f (x) are normalized such that
+
f (x)dx = 1. (G.1)

In other chapters of these textbooks a different type of normalization may be more

convenient and are used whenever appropriate, e.g. normalization to the maximum
max[f (x)] = 1.
G.1 Denition and Motivation
The mathematical operation called convolution quantifies the overlap between two
mathematical functions f1 (x) and f2 (x) as one is reversed and shifted in respect of
the other. A convolution is defined as

(f1 f2 )(x) := f1 ( )f2 (x )d, (G.2)

which describes an important mathematical procedure, often required and used in

experimental physics and measuring techniques.
The cross-correlation of two (possibly complex) functions is defined by

G(x) = (f1 * f2 )(x) := f1 ( )f2 ( + x)d f1 (x) f2 (x). (G.3)

Thus, both procedures are identical if one of the functions is even, f (x) = f (x),
which is the case in most of the physical applications.1 If f1 f2 one calls G(x) =
(f * f )(x) an autocorrelation function.
1 An alternative method to solve (G.3) is given by F (f1 * f2 ) = 2 F (f1 )F (f2 ) where F (f ) de-
scribes the F OURIER transform of f (see Appendix I).

624 G Convolutions and Correlation Functions
S1
S2
x,
Fig. G.1 Convolution scheme of two rectangular profiles as manifested by two slits S1 and S2 of
unequal size. Their respective transmission profiles are shown as full black and red dashed lines.
As slit S2 is shifted over slit S1 in x, direction, the overlap (grey area) changes. The resulting
total transmission is shown as heavy red line (convolution)
Convolution and cross-correlation are well documented in mathematical tables as

well as in the web (often with illustrative animations, e.g. W EISSTEIN, 2011). Thus
we keep this brief recollection focused on the applications most often encountered
in AMO physics. Convolutions occur e.g. if a physical observable is analyzed and
detected that may assume a range of values described by a probability distribution
f1 ( ) as the variable is changed. The measurement device too, will never be sensi-
tive at one and only one value of . Rather the signal will be detected with a varying
detection probability f2 ( ). Hence, the detector is tuned over the region of interest
by shifting its detection efficiency through x. The resulting signal is described by
(G.3).
Two rectangular profiles are the simplest case e.g. manifested as transmission
of light through two slits. This is shown for illustration of the principle in Fig. G.1,
starting with the situation in the leftmost cartoon with both slits separated from
each other where transmission is zero. The signal will increase (linearly) when S2
is shifted into +x-direction as soon as the areas of the slits begin to overlap. The
maximum is reached when the overlap is complete and stays constant as long as
the slits fully overlap. Further shift will reduce the signal again linearly. The overall
result is a capped triangular shape of the transmission profile (heavy red line in
Fig. G.1). It will be triangular if the two slits have equal width.
Typical applications are found in any kind of spectroscopy on atoms, molecules,
solids or in nuclear and elementary particle physics: the system absorbs, emits, re-
flects photons, electrons, atoms or elementary particles as a function of energy,
frequency or wavelength , characteristic for the physics studied. A suitable detector
registers this signal with a specific detection probability which depends on . This
may e.g. be controlled by the slit position and width of a spectrograph, the transmis-
sion curve of a frequency filter or particle energy analyzer. Recording a spectrum
implies changing x and thus shifting the transmission maximum of the
spectrograph knowing that on the left and right of that maximum still some signal
is transmitted. The detector finally sums over all of this transmitted signal.
Convolution is also at work in nature when different processes may occur si-
multaneously in one system and overlay each other. A standard example is the
broadening of spectral lines emitted or absorbed by atoms and molecules. They
may e.g. move with a velocity described by the B OLTZMANN distribution and at the
same time undergo collisions. While the first effect leads to a significant, inhomo-
geneous spectral broadening due to the D OPPLER effect, collisions may also disturb
G.2 Correlation Functions and Degree of Coherence 625
the emission or absorption process and add to the broadening. The combination of
both effects is described by a convolution of the respective line profiles.
The physics involved in all these processes and examples will be discussed in
some detail in Sect. 5.1.1, and Chaps. 1, 2 in Vol. 2, and in other parts of these vol-
umes. In the present appendix we summarize a few useful mathematical expressions
which are needed to describe such measurements quantitatively. At the end one aims
of course at an intelligent de-convolution of the signal detected as a function of x,
i.e. one tries to extract the original profile f1 ( ) form the convoluted signal. This is
no trivial task unless the shapes of the two profiles involved are both well known.
Often one is content with just measuring the convolution and gleaning the width of
the profile f1 ( ) from known convolution formulas.
G.2 Correlation Functions and Degree of Coherence
In physics, the cross-correlation (G.3) also called (first-order) correlation func-

tion is often used to correlate an observable at different points in space r and/or
time t. Typical observables are the electric field, or the intensity of an electromag-
netic wave. They may represent pulses (wave-packets), random sequences of pulses,
stationary signals (e.g. continuous light beams) or even more or less statistical noise.
Correlation functions describe the coherence of these observables in space and/or
time in more colloquial terms one might say: the correlation function measures
traces of similarities over a distance in space and/or time.
The propagation of electromagnetic waves is described by its wave vector k and
angular frequency in the combination kr t. Thus, with EA (t) = E(r, t) and
EB (t ) = E(r , t ) representing the field at two positions in space2 and time, one
may keep r constant, focus on the temporal coherence, and write the correlation
function in dependence of the time delay = t t

1 Tav /2
G () =
(1)
EA (t)EB (t + ) = EA (t)EB (t + )dt. (G.4)
Tav Tav /2
The angle brackets . . . indicate averaging, the second equality giving one possible
recipe how to accomplish this averaging for the stationary case (e.g. for a CW light
beam or for continuous noise). Of course, integration has to be over a sufficiently
long time Tav so that all statistical short term fluctuations are smoothed out.
If EA and EB are square integrable functions, instead to average one has to inte-
grate over all times as in (G.3) and the 1/Tav factor becomes redundant. In normal-
ized form, the correlation function is called first-order degree of coherence:
EA (t)EB (t + )
g (1) () = . (G.5)
[|EA (t)|2 |EB (t)|2 ]1/2
2 The further treatment is independent of whether EA and EB relate actually to two different fields
or represent the same field at different positions in space.
Correlation functions are e.g. used in state-of-the-art ultrafast science for the
determination of shape and duration of short laser pulses as to be detailed in Chap. 1,
Vol. 2. In such a measurement, the electric field amplitude EA (t) (or the intensity
|EA (t)|2 ) of the pulse to be characterized is convoluted with a second light pulse
of well known profile EB (t) (or intensity |EB (t)|2 ). This is achieved by delaying
one pulse for a variable time in respect of the other. Both signals are multiplied
and integrated over a long time Tav . The signal is then recorded as a function of the
delay time and the result may finally be de-convoluted.
If EA (t) and EB (t) are identical (or just displaced in time) the first-order degree
of coherence (G.5) simply becomes the normalized autocorrelation function
E (t)E(t + ) E (t )E(t)
g (1) () = = = g (1) (). (G.6)
|E(t)|2 |E(t)|2
The important symmetry with respect zero follows directly from the definition (G.4)
for A = B. If the shape of a pulse is already well known, it allows one to determine
directly the pulse duration (see below for special examples).
Note that here the 1/Tav factor cancels out in (G.5) and (G.6): these definitions
hold for CW fields as well as for a single pulses or a pulse sequence. In practice, the
averaging may not be that trivial, especially if E(t) is not given in closed analytical
form. For CW fields the average . . . can be performed over times at a given po-
sition in space or alternatively over the whole ensemble in space at a given fixed
time. One calls a system ergodic if the temporal average is equal to the average over
a representative ensemble which is generally true for reasonable physical systems.
Usually the temporal average is more convenient to perform while determining the
ensemble average (averaging over the whole space) would be a rather difficult task.
According to the ergodicity theorem,both averages are identical.
G.3 Gaussian Prole

In reality, experimental profiles are not just rectangular. To be specific, we start by
discussing the Gaussian profile characteristic for a number of physical phenomena
and often also a good approximation for analyzer profiles. We write it in the form
most often used in atomic and laser physics:

2 1
fG (x; w, x0 ) = exp 2(x x0 )2 /w 2 (G.7)
w

with a FWHM of x = 2 ln 2w = 1.177w. (G.8)
We note that at x x0 = w gives the distance at which fG has decreased to 1/e2 .

One easily verifies that fG (x; w, x0 ) is normalized according to (G.1). The mean
value x and the variance 2 = (x x )2 are given by

2 1 + 2(xx0 )2 /w2
x = xe dx = x0 , and (G.9)
w
G.4 Hyperbolic Secant 627

2 2 1 + w2
(x x0 )2 e2(xx0 )
2 /w 2
2 = x x = dx = . (G.10)
w 4

The standard deviation is thus = w/2 and the FWHM is 8 ln 2 1.2w.
An important case is the B OLTZMANN distribution for velocities in a gas. The
1D distribution in respect of the component vx , say, is given by

m mvx2
w(vx )dvx = exp dvx (G.11)
2kB T 2kB T
with the particle mass m, the absolute temperature T , and the B OLTZMANN con-
stant kB . The average velocity is vx = 0, and the variance vx2 = kB T /m.
D OPPLER broadening (Sect. 5.1.4) is a direct consequence of this Gaussian dis-
tribution. We recall from (5.21) its variance D 2 = 2 k T /(mc2 ) with the angular
ba B
transition frequency ba and the speed of light c.
Interestingly, the convolution (here identical to the cross-correlation) of two
Gaussian profiles with 1/e2 at x = w1 and w2 , respectively, is again a Gaussian
profile, with w 2 = w12 + w22 . One verifies easily (using e.g. SWP 5.5, 2005)
(fG * fG )(x; w2 , x2 )

2 1
= exp 2( x1 )2 /w12 exp 2( x2 x)2 /w22 d
w1 w2

2 (x (x1 x2 ))2
= exp 2 , (G.12)
(w12 + w22 ) w22 + w12
the convoluted FWHM thus being the geometric mean of both widths

x = 1.177w = 1.177 w12 + w22 = ( x1 )2 + ( x2 )2 . (G.13)
Thus, the autocorrelation function or the first-order degree of coherence (G.6) of a

Gaussian temporal profile with a FWHM t has a

FWHM: t auto = 2 t. (G.14)
G.4 Hyperbolic Secant
The hyperbolic secant squared is another, often used line profile, in particular for
laser pulse intensities. Normalized according to (G.1) it is
sech2 [(x x0 )/w] 1
fH (x; w, x0 ) = = (G.15)
2w 2w cosh [(x x0 )/w]
2
2
1 2
= ,
2w e(xx0 )/w + e(xx0 )/w

with a FWHM of x = 2 ln( 2 1) w = 1.763w. (G.16)
The convolution is a not completely trivial integral. For the autocorrelation function
one finds (properly normalized)
1 (x/w) cosh(x/w) sinh(x/w)
(fH * fH )(x; w) = (G.17)
w [sinh(x/w)]3
3 sech4 [x/(2.24445w)] sech2 [x/(1.5429w)]

=
4 2.24445w 2 1.5429w
all with a FWHM of x = 2.720w. (G.18)
The first equality is exact, the sech4 gives an excellent approximation, and the sech2
is still a good approximation for small |x| 1.7; in the far wings, however, it is
somewhat too high.
G.5 LORENTZ Prole
Another profile often encountered is the L ORENTZ profile, describing e.g. the natu-
ral line profile or collision broadening (see Sects. 5.1.1 and 5.1.3, respectively). We
write the L ORENTZ profile
2 2 /4
fL (x; , x0 ) = (G.19)
(x x0 )2 + 2 /4
with a FWHM x = , (G.20)
again properly normalized according to (G.1). The L ORENTZ profile is a partic-

ularly wide profile (see Fig. 5.5) and its variance x 2 obviously diverges since
fL (x; , x0 ) 1/x 2 for large x, while its mean value is x = x0 .
The convolution of one L ORENTZ profile with another leads again to a L ORENTZ
profile (for a proof see e.g. Appendix I.4.5):
2 2 /4
(fL * fL )(x; 2 , x2 ) = (G.21)
(x (x1 x2 ))2 + 2 /4
with the FWHM of the convolution x = = 1 + 2 .
G.6 VOIGT Prole
Convolution of a L ORENTZ with a G AUSS profile is required e.g. for optical line
shapes of atoms or molecules in the gas phase if D OPPLER broadening and collision
broadening are on the same order of magnitude (or at low temperatures D OPPLER
and natural linewidth). This so called VOIGT profile
+
e2 /w
2 2
2 1
fV (x; , ) = (fG * fL )(x; , 0) = d, (G.22)
w 2 ( x)2 + 2 /4
is an integral which cannot be evaluated by standard integration formulas. With

some advanced functions this may, however, be achieved. Extended literature exists
on the subject. We essentially follow NIST-DLMF (2013) and W IKIPEDIA CON -
TRIBUTORS (2014). Properly normalized according to (G.1) and with the definitions
for w and used above, the VOIGT function can be expressed as

2 Re(exp(z2 ) erfc(iz))
fV (x; w, ) = (G.23)
w

2
with z = (x + i /2) (G.24)
w
where exp(z2 ) erfc(iz) is the complex error function (also FADEEVA function).
The complementary error function erfc(x) = 1 erf(x) is usually implemented in
modern symbolic mathematics programmes (we use SWP 5.5, 2005). Even an ap-
proximate linewidth (FWHM) for the VOIGT profile is found in the literature as

V = 0.5346 L + 0.2166( L )2 + ( D )2 , (G.25)
supposed to be accurate to within 0.02 %, where L and D are the FWHM

of L ORENTZ and D OPPLER profile, respectively. One example of this convoluted
D OPPLER and collision profile for L = D is shown in Fig. 5.5. In this case
the formula gives V = 1.638 L .

References
NIST-DLMF: 2013. Digital library of mathematical functions: 7.19 Voigt functions, NIST.
http://dlmf.nist.gov/7.19, accessed: 9 Jan 2014.
SWP 5.5: 2005. Scientific work place, Poulsbo, WA 98370-7370, USA: MacKichan Software,
Inc. http://www.mackichan.com/, accessed: 9 Jan 2014.
W EISSTEIN, E. W.: 2011. Convolution, Wolfram Research, Inc., Champaign, IL, USA. http://
mathworld.wolfram.com/Convolution.html, accessed: 9 Jan 2014.
W IKIPEDIA CONTRIBUTORS: 2014. Voigt profile, Wikipedia, The Free Encyclopedia. http://en.
wikipedia.org/wiki/Voigt_profile, accessed: 9 Jan 2014.
Vector Potential, Dipole Approximation,
Oscillator Strength H
H.1 Interaction of the Field of an Electromagnetic Wave with

an Electron
In view of its clearness and compactness we use in this book apart from a few ex-
ceptions in general the dipole length approximation for describing the interaction
of electromagnetic waves with atoms and molecules. It is thus appropriate to justify
this approach in some detail. In addition, we shall provide the basis for some useful
concepts related to the dipole approximation.
H.1.1 Vector Potential
In the quantum mechanically correct formulation for the interaction of an electro-

magnetic field with charged particles one replaces the momentum operator p=
i of the particle (charge q) by p field =
p qA, where A = A(r, t) is the vector
potential of the field. The Hamiltonian of an electron in the field thus becomes

p 2field 1
=
H + V (r) = p + eA)2 + V (r)
( (H.1)
2me 2me
p2 pA
e e2 2
= + V (r) + + A (H.2)
2me me 2me
where the so called C OULOMB gauge of the field has been used:
A = 0. (H.3)
For simplicity we have restricted the discussion to a single active electron. One
easily generalizes (H.2) to multi-electron systems by replacing momentum p and
position vector r with p i and r i , respectively, for each electron i, and summing
over all electrons.

632 H Vector Potential, Dipole Approximation, Oscillator Strength
The vector potential A relates to the electric and magnetic field vectors

E(r, t) = A(r, t) + Vext and B(r, t) = A(r, t) (H.4)
t
with Vext (r, t) describing an additional, external potential. If such a field is present,
one has to add Vext to V (r) in (H.1) and (H.2). Not included in this description are
of course all interactions related to the electron spin which follow from the D IRAC
equation, including spin-orbit interaction, as briefly discussed in Chap. 6.
In vacuum, we write the vector potential, electric field, and magnetic field for a
plane electromagnetic wave as:
A0 i(krt)
A(r, t) = ee + e ei(krt) = A0 e cos(kr t) (H.5)
2
iE0 i(krt)
E(r, t) = ee e ei(krt) = E0 e sin(kr t) (H.6)
2
iB0 k i(krt) k
B(r, t) = ee e ei(krt) = B0 e sin(kr t). (H.7)
2 k k
The relations among the amplitudes are derived from (H.4)

E0 = A0 , B0 = kA0 = A0 , and B0 = E0 /c. (H.8)
c
H.1.2 Intensity
The P OINTING vector

EB
S =EH =
0
gives the energy flux. With E B and 0 0 = 1/c2 we obtain its absolute value,
called intensity:
0 c[E0 (r, t)]2 it 2
If (r, t) = |S| = 0 c|E|2 = ee e eit (H.9)
4
0 c[E0 (r, t)]
2
= 1 + sin(2) cos(2t) . (H.10)
2
Here we have inserted for e the general unit polarization vector (4.15) with the
ellipticity angle . Obviously, depending on , the intensity If (r, t) may oscillate
fast (with 2) in position space and time, as emphasized by the index f. However,
independent of one obtains the cycle averaged intensity
2 2
I (r, t) = If (r, t) = 0 c E(r, t) = 0 c2 A(r, t)
0 c 2 0 c2 2 c 2 (H.11)
= E0 (r, t) = A0 (r, t) = B0 (r, t) .
2 2 20
H.1 Interaction of the Field of an Electromagnetic Wave with an Electron 633
In principle, the amplitude E0 (r, t) (and A0 and B0 ) as well as the averaged intensity
I (r, t) may still depend slowly on time t and position r. This is called SVE.1 More
about cycle averaged intensities is discussed in Sect. 1.3.1, Vol. 2. The electric field
is measured in units [E0 ] = V m1 , the intensity in [I ] = W m2 .
Note that all field quantities used here represent quasi-monochromatic waves:
their bandwidth is assumed to be much narrower than any atomic or molecular ab-
sorption line studied. The transition to a continuous spectrum is made by replacing
I I()d, where the spectral intensity distribution I() (or the spectral energy

density u() = I()/c) refer to the unit
of angular frequency interval. Finally, one
has to integrate over all frequencies: . . . d.
H.1.3 Static Magnetic Field
For a static, homogeneous external magnetic field B one derives the vector potential
by inverting (H.4):
1 1
A = r B = B r. (H.12)
2 2
This may be verified by inserting this expression into (H.4).2
For such a constant B field the Hamilton operator (H.2) becomes
=
H
p2
+ V (r) +
e
LB +
e 2 A2
, (H.13)
2me 2me 2me
where for the third term we have used the identity (B r) p = B (r

p ) for
the scalar triple product and the definition of angular momentum L=r p . This
term corresponds exactly to (6.29) with (6.27), i.e. we have derived here correctly
the interaction potential of an external magnetic field with the magnetic moment of
the orbital angular momentum. A heuristic derivation is given in Chaps. 1 and 6.
The final term in (H.13) represents a (usually) small correction
e 2 A2 e2 2 2
= r B sin , (H.14)
2me 8me
1 The SVE approximation demands that |E

0 /t| c E0 as well as |E0 /z| E0 /c etc. Then,
second order derivatives may be neglected in the general wave equation.
2 The triple vector product is expanded as

1 1
A = r B = [B r r B + r B B r].
2 2
The first two terms are directional gradients. B r = B and for a homogeneous B field
r B = 0. The third term also disappears since B = div B 0. With div r = 3 the forth
term = 3B. Finally, the whole right hand side gives (1/2)(B 3B) = B, q.e.d.
where is the angle between the external B field and the position vector r of the
electron in the atom. We estimate for, say, 30 T and r = a0 a maximum value of ca.
5 108 eV. Thus, in spectroscopy this term only plays a role if extreme accuracy
is asked for or very high magnetic fields and very large orbital radii are involved
(i.e. high lying RYDBERG states). On the other hand, it is this very term which is
responsible for all diamagnetism of matter.
H.1.4 Relation Between Matrix Elements of p and r
For the following considerations we derive an important relation between the matrix
elements of momentum and position. For electrons i and j the standard commuta-
tion rules for canonically conjugated observables (momentum and position) hold:
[xi , p
yj ] = 0 and [xi , p
xj ] = iij , etc. (H.15)
With these and the identity

a,
a,
b2 = [ b]
b + a,
b[ b] (H.16)
we can write
] = i p
[xi , H xi and [r, H ] = i p, (H.17)
me me

where r = r i and p= p i . Thus, for the matrix elements of
p between eigen-
states |a and |b of the Hamiltonian we derive
me ]|a = me (Wa Wb )b|r|a .
b|
p |a = b|[r, H (H.18)
i i
With (Wa Wb )/ = Wba / = ba the sought-after relation between the matrix
elements of
p and r is:
b|
p |a = ime ba b|r|a . (H.19)
H.1.5 Ponderomotive Potential
Before actually evaluating the matrix elements of the interaction in the Hamiltonian
(H.2) in detail, we have a second look at the term proportional to A2 , now for an
atom or molecule in an electromagnetic wave. Clearly, this term is a time dependent
additional energy in the Hamiltonian. By averaging A2 over one period we obtain
with (H.11) an estimate for its dependence on intensity I :
e2 A2 e2 I
Up = = . (H.20)
2me 20 cme 2
H.1 Interaction of the Field of an Electromagnetic Wave with an Electron 635
This is identical to the expression (8.139) discussed in Sect. 8.5.1 for the pondero-
motive potential Up . There, a completely classical picture of an electron oscillating
in the electric field of the wave is used to derive it.
In respect of standard laser spectroscopy, the ponderomotive potential is usually
negligible. As described in Sect. 8.5.1, however, it leads to very interesting phe-
nomena if the objects studied are exposed to very intense electromagnetic radiation
which can be generated without difficulties by todays short pulse lasers.
H.1.6 Series Expansion of the Perturbation and the Dipole

Approximation
now discuss the term proportional to A

p in (H.2) which is responsible for electro-
magnetically induced transitions. With (H.5) it is
e A0
(
U p , r, t) = p eei(krt) + e ei(krt)

me 2
eE0 it
=
De D e+it . (H.21)
2
We abbreviate the transition operator
D
eikr ieikr

D=
pe= e (H.22)
me me
(with the dimension L) and the amplitude

eE0 = eA0 = ecB0 = e 2I /(c0 ) (H.23)
(with the dimension of a force MLT2 ). For electromagnetic waves in the IR, VIS,
UV, and VUV spectral range the wavelength is usually very large compared to the
objects studied. We thus can assume k r 1 and expand the exponential function
in (H.22):
1

D= (1 + ik r + )
p e. (H.24)
me
In the electric dipole approximation (short: dipole approximation) only the first
term is taken into account and one speaks of E1 transitions. Thus, the dipole transi-
tion matrix elements of D between two eigenstates |a and |b is:
1

Dab = a|
p |b e = i a||b e = ia|r|b e. (H.25)
ba me ba me
For the last equality we have made use of (H.19). Strictly mathematical, both forms
of
Dab are completely equivalent if the wave functions used are exact solutions of
the unperturbed Hamiltonian. However, since we usually know the wave functions
only approximatively (except for some special cases such as the H atom) both for-
mulas (H.26) and (H.27) lead to slightly different results. Both variants are used in
the literature: they are called dipole velocity approximation (since
v = p /me ), and
the dipole length approximation, respectively.
The matrix elements of the interaction potential (H.21) are thus

i e 2I ()
ab (t) = a|U
U (r, t)|b = a||b eeit + e e+it (H.26)
2 me c0
or alternatively in dipole length approximation

U (r, t)|b = i E0 a|er|b eeit e e+it .
ab (t) = a|U (H.27)
2
The latter form is identical to that obtained from the heuristic considerations pre-
sented in Sect. 4.3.4.
All expressions above refer to a single active electron which is excited or de-
excited in an external electromagnetic field. For larger atoms and molecules more
than one electron (in principle all) can participate in the interaction even simul-
taneously. Such events typically lead to interesting phenomena (see e.g. autoioniza-
tion, Sect. 7.6). Thus, in general one has to sum the respective interaction potentials
for all electrons. The transition operator (H.22) will then read
N N
1 ikr i i ikr i

D= e
pi e = e i e. (H.28)
me me
i=1 i=1
Specifically, in dipole approximation for a multi-electron system the transition ma-

trix element (H.25) has to be replaced by
N N

Dab = i a| i |b e = ia| r i |b e. (H.29)
ba me
i=1 i=1
The first and second equality refer to the dipole velocity and dipole length form
of the transition operator, respectively. They are identical if (and only if) the wave
functions are exact. The sum has to be carried out in principle over all electrons i,
with N being the total number of electrons in the system.
H.2 Line Strength and Oscillator Strength

H.2.1 Denitions
Several, slightly differing, quantities are used in the literature to characterize the
overall strength of a dipole transition between levels a and b with angular mo-
mentum quantum numbers ja ma and jb mb , and degeneracies ga = 2ja + 1 and
H.2 Line Strength and Oscillator Strength 637
gb = 2jb + 1, respectively. In principle, j may refer to the orbital (L) or to the to-
tal angular momentum (J ), whichever is appropriate. Note that we abbreviate the
quantum numbers a ja ma := a, but we also use a and b for designating the respec-
tive energy levels as long as no confusion can arise.
We essentially3 follow C ONDON and S HORTLEY (1951) and define the line
strength symmetric in respect of initial and final energy levels and as a sum over
all polarizations:

S(jb ja ) S(ja jb ) := |r ba |2 b jb mb |r|a ja ma 2
mb ma mb ma
(H.30)
= b|x|a 2 + b|y|a 2 + b|z|a 2 .
mb ma
The dimension of the line strength is L2 .

In spherical coordinates (see 4.75) this may be written even more explicitly:

S(jb ja ) = b jb mb |rq |a ja ma 2 S(ja jb )
mb ma q
2 (H.31)
= b |r|a jb mb |C1q |ja ma 2 .
q mb ma
Using the W IGNER -E CKART theorem in the form (4.82) and the orthogonality re-
lations of the 3j symbols (B.42) we obtain the compact relations:4
2
S(jb ja ) = (2jb + 1)b |r|a jb C1 ja 2
2 (H.32)
= (2ja + 1)a |r|b ja C1 jb 2 S(ja jb ).
If LS coupling is appropriate, one may recouple the reduced matrix elements

Ja La SC1 Jb Lb S 2 using (C.46). Exploiting the orthogonality relation (B.67) for
the 6j symbols one obtains the relation
S(Jb Ja ) S(Lb La )
, (H.33)
2Jb + 1 2Lb + 1
Ja
where the equality sign holds in so far as the radial matrix element does not depend
on the individual fine structure levels J but only on the orbital quantum numbers L
and n which is a good approximation for light atoms.
3 The line strength S CS (jb ja ) used by C ONDON and S HORTLEY (1951) is related by S CS (jb ja ) =
eS(jb ja ) to the quantity used here which leads to more compact formulas for the A and B
(opt)
coefficients and for the oscillator strength fba .
4 Notethat a factor 3 arising here from the sum over all polarizations q is compensated by a factor
1/(2 1 + 1) = 1/3 from the orthogonality relations of the C LEBSCH -G ORDAN coefficients when
summing over ma and mb .
Complementary to the line strength, which is symmetric and has a dimension L2 ,

the so called oscillator strength f(opt) (e) is defined asymmetrically and dimension-
less for a given polarization with unit vector e. We first define it for a specific tran-
sition from one initial substate |a = |ja ma to one final substate |b = |jb mb :

2Wba 2Wba r ba 2
(opt)
fjb mb ja ma (e) = | |2
= e (H.34)
Eh a02
D ba
E h a0 .
Here Dba is the dipole transition matrix element (H.29). In the second equality we
use explicitly the dipole length form for a single electron system, with r ba as de-
fined by (4.79). The oscillator strength is thus proportional to the transition proba-
bility (4.65). And since it is also proportional to the absorbed (or emitted) photon
energy Wb Wa = Wba = ba , its value is positive for absorption and negative for
emission.
For the particular case of linear polarization with elin z this gives a simple ex-
pression for the oscillator strength:

(opt) Wba zba 2 me ba
fjb mb ja ma =2 =2 |zba |2 . (H.35)
E h a0
It may be further evaluated with (4.75) and the W IGNER -E CKART theorem in the
form (4.82):
2
|zba |2 = b jb mb |r0 |a ja ma
2 (H.36)
2 ja 1 jb
= (2jb + 1)b |r|a jb C1 ja 2 .
ma 0 mb
Note that this becomes zero for mb = ma .

In an experiment one typically averages over all initial substates |ja ma and sums
over all final substates |jb mb . If the initial level is populated
isotropically one ob-
tains for an arbitrary unit polarization vector e = q aq eq for the transition be-
tween from level jb to level ja with (H.34) and (4.93)

(opt) 1 (opt) 2 Wba r ba 2
fjb ja = f (e) = e (H.37)
ga m m jb mb ja ma ga E h m m a 0
a b a b
1 2
2 Wba
1

= b jb mb |rq |a ja ma eq aq e q
ga E h m m

a b q=1 q =1
2 Wba 2 2
1
= |aq |2 b |r|a jb mb |C1q |ja ma . (H.38)
ga E h m m
a b q=1
H.2 Line Strength and Oscillator Strength 639

Using (4.82), the orthogonality relation (B.42) for the 3j symbols, and q |aq |2 = 1
(unit vector e) the sums can be carried out:
2
Wba gb r
1
1
(opt)
fjb ja =2 | | jb C1 ja 2 |aq |2 (H.39)
E h ga
b a
a0 3
q=1
2
(opt) 2 Wba gb r

fjb ja = b | | a jb C1 ja .
2
(H.40)
3 E h ga a0
Finally by comparing this with (H.32) we obtain the expression
(opt) 2 Wba S(jb ja ) 2me ba S(jb ja )

fjb ja = = (H.41)
3ga Eh a02 3 ga
as used in Sect. 5.2.2. We note here one very important result:
The averaged oscillator strength for an isotropically populated initial level

and hence the probability for absorption or induced emission is independent
of the polarization.
Thus, we also may simply average (H.35)

(opt) 2 Wba zba 2 2me ba 1
fjb ja = = |zba |2 (H.42)
ga E h m a 0 ga m
a a
where we have exploited that here q = 0, and thus only contributions from ma = mb
arise. For an initial singly degenerate ns state (H.42) is reduced to (H.35). Equivalent
expressions are derived from (H.41) with (H.30) and the dipole moment D = er:
2me ba 2
b jb mb |r|a ja ma
(opt)
fjb ja = (H.43)
3ga m m
2
a b
2
2 Wba r
8 2 me Wba
= j m
b b b | | a a a =
j m |D ba |2 .
3ga Eh m m a0 3ga h2 e2 m m
a b a b
(H.44)
Note the factor 3 in the denominator in comparison to (H.42), compensating for the
equal contributions from the components of |r|2 = |x|2 + |y|2 + |z|2 .
H.2.2 THOMAS-REICHE-KUHN Sum Rule
We sum the oscillator strength according to (H.35) over all existing final states |b =
|b jb mb and rewrite the expression suitably:
(opt)
2me ba 2me ba
fjb mb ja ma = |zba |2 = a|z|b b|z|a (H.45)

b b b
1
= ime ba a|z|b b|z|a ime ab a|z|b b|z|a .
i
b
Now we apply (H.19) and obtain

(opt) 1
fjb mb ja ma = a|z|b b|pz |a a|pz |b b|z|a
i
b b
(H.46)
1 a|zpz pz z|a
= a|zpz |a a|pz z|a = = 1,
i i

where we have exploited completeness 1 = b |b b|. The last step realizes the
commutation rules (H.15) and normalization a|a = 1.
We note that this derivation does not depend on the specific initial substate |a =
|ja ma since summation is always over all final states |b . Hence, (H.46) is also valid
(opt)
for the average oscillator strength fjb ja . Thus, we have derived the very important
T HOMAS -R EICHE -K UHN sum rule:
(opt)
fjb ja = 1. (H.47)
b
The oscillator strength is often used in atomic and molecular physics to characterize
dipole transitions. It allows one to compare the strength of different transitions in
(opt)
different atoms or molecules. For systems with one active electron fjb ja 1 strictly
holds, the classical reference being an oscillating electron which has an oscillator
strength = 1.
For systems with Ne active electrons one has to replace in (H.45) the single
e (i)
electron coordinate z by N i=1 z and obtains the sum rule:
(opt)
fjb ja = Ne . (H.48)
b
Finally, we point out that the summation just discussed must include the ion-
ization continuum if the basis is to be complete. In this context, continuum states
are usually normalized in the energy scale (see Appendix J). Correspondingly one
defines the optical oscillator strength density (OOSD), df(opt) /d, with the dimen-
sion Enrg1 , where is the energy of the emitted electron in the continuum. Thus,
(H.48) includes summation over discrete states and integration for energies beyond
the ionization potential WI :

discrete (opt)
dfja
(opt) (opt)
fjb ja fjb ja + d. (H.49)
WI d
b b
Obviously for high energies, limW (df(opt) /d) = 0 must hold.


OOSD: Optical oscillator strength density, characterizes the strength of photoion-
ization per energy interval (see Sect. 5.5.1).
21348 (2007).
References
FOURIER Transforms and Spectral Distributions
of Light I
I.1 Short Summary on FOURIER Transforms

The F OURIER transform (FT) is a mathematical tool widely used in physics and
engineering. We collect here the most important definitions and relationships and
give a few practical examples, focussing on the description of short light pulses.
F OURIER transforms are based on the complex version of the F OURIER integral
(here without proof):

1
X(t) = eit d X( )ei d (I.1)
2

or = e 2it
d X( )e2i d.

Various, slightly different notations can be found in the literature. We use the stan-
dard notations of modern physics with the F OURIER transform1

)
X() = F X(t) = X(t)eit dt (I.2)

of the time dependent function X(t). Conversely, X(t) is recovered by the inverse
F OURIER transform

1
)
X(t) = F 1 X() = )
X()e it
d. (I.3)
2
The complex conjugate of the F OURIER transform of X(t) is

)
X () = X (t)eit dt, (I.4)

1 In

other notations the factor 1/2 is applied symmetrically as 1/2 , or completely avoided by
using frequency instead of angular frequency = 2.

644 I FOURIER Transforms and Spectral Distributions of Light
and correspondingly

1 ) ()eit dt.
X (t) = X (I.5)
2
By rewriting (I.1) as

1
X(t) = X( ) e i(t )
d d
2
one identifies the expression in square brackets as the D IRAC delta function:

1
(t ) = F 1 ei = ei(t ) d (I.6)
2

1 ic t 1
and equivalently (c ) = F e = ei(c )t dt. (I.7)
2 2
With this, one verifies the important relationship

2
X(t)2 dt = 1 )
X() d, (I.8)
2
called P LANCHERELs theorem (sometimes PARCEVALs theorem which actually

refers to F OURIER series or R AYLEIGHs theory). To proof it, one just has to insert
(I.3) and (I.2), rearrange the integrations and recover the D IRAC delta function. We
leave it to the reader to work out the details. As we shall see below, for electromag-
netic radiation P LANCHERELs theorem essentially states energy conservation.
In the context of describing short pulses of light, we note a very useful relation
for the F OURIER transform of an oscillation around a carrier frequency c with an
envelope function X(t): With X()) = F[X(t)] we can write

1 ) 1 ic t ) c )ei(c )t d
X(t)e ic t
=e ic t
X()e it
d e X(
2 2

1
= X( ) c ) .
) c )eit d = F 1 X( (I.9)
2
Thus, the carrier frequency just shifts the F OURIER transform F[X(t) exp(ic t)]
towards positive frequencies in respect of F[X(t)]. This formally also resolves the
problem one might envisage with negative frequencies in the spectral distribution
)
X(): since for any physically relevant distribution with a FWHM= one expects
c , there will be no contributions from negative frequencies in (I.2). An ana-
logue relation applies to a shift in time.
Another set of useful relations concerns F OURIER transforms of convolutions
(see Appendix G). The proof is similar to that for P LANCHERELs theorem.
I.1 Short Summary on FOURIER Transforms 645
1. The convolution theorem states that the F OURIER transform of a convolution of

)
two functions X(t) and Y (t) is the product of their F OURIER transforms X()
)
and Y ():

F (X Y )(t) = F X(t) F Y (t) = X() ) Y )(). (I.10)
This is proven as follows:

it

F (X Y )(t) = X t Y t t dt e dt

it
= X t e dt Y t t ei(tt ) dt

= X t eit dt Y t eit dt .

2. The F OURIER transform of a product of two functions is the convolution of their

F OURIER transforms:
1 1 ) )
F X(t)Y (t) = F X(t) F Y (t) = (X Y )(). (I.11)
2 2
And the proof for this relation is only slightly more complex, involving the
D IRAC delta function:

F X(t)Y (t)

1 i t 1 i t it
= )
X e d )
Y e d e dt
2 2

1 i t
= )
X e d
2

1
) ei( )t d eit dt
Y
2

1 1
= ) Y
X ) d d ei( )t dt
2 2

1
= X ) Y ) d d
2

1
= ) Y
X ) d .
2
The inverse relations of course also hold:

F 1 (X )Y))() = 2X(t)Y (t) and (I.12)

) Y
F 1 X() )() = (X Y )(t). (I.13)
3. In analogy to the convolution theorems, the cross-correlation theorems

F (X * Y )(t) = F X(t) F Y (t) = X ) ()Y
)() and (I.14)
1 1 ) )
F X (t)Y (t) = F X(t) * F Y (t) = (X * Y )() (I.15)
2 2
hold, as well as their inverse relations.
From these expressions follows directly a very important relationship between

F OURIER transform and autocorrelation function, the W IENER -K HINCHIN theo-
rem:

1 )
2
F
X() = X ( )X( + t)d or (I.16)

2
)X() = F X ( )X( + t)d . (I.17)

I.2 How Electromagnetic Fields are Written

With these tools at our command we want to describe wave-packets, i.e. short, co-
herent light pulses. At this point, we emphasize again that real electromagnetic fields
are real observables, i.e. measurable quantities! One often finds in the literature that
electromagnetic wave fields are for simplicity written in complex form. However,
albeit the greater mathematical simplicity of the complex form is indeed attractive,
one may easily overlook some important aspects by a too naive use of the complex
form as we have documented in Chap. 4, when describing absorption and induced
emission of light in quantum mechanical terms.
Thus, following (4.1) we write the electric field vector E(r, t) of a wave-packet
as a real quantity, explicitly using its two conjugate complex terms. For simplicity
of writing we ignore here the r dependence2 in E0 (r, t) and keep the position in
space constant kr = 0:
1
E(t) = E0 (t) eei(c tc ) + e ei(c tc ) (I.18)
2
1
eE (t) + e E + (t) with E + (t) = E0 (t)ei(c tc ) = E (t) .
2
In the spirit of the so called SVE approximation (see footnote 1 in Appendix H), we
have introduced here E0 (t), the time dependent envelope of the field now a real
function depending on time. The phase c may also depend on time and position (see
e.g. (1.46) in Vol. 2). It allows to specify the relative phase of the carrier oscillation
c t k c r c t .
2 Where necessary, we shall simply reintroduce the wave vector by replacing
Similarly, we keep the r dependence of the envelope in the back of our mind and set E (t)
E (r, t) when required.
I.2 How Electromagnetic Fields are Written 647
in respect of the envelope E0 (t) as illustrated in Fig. 1.24 in Vol. 2. In nonlinear

optics and ultrafast science it plays a significant role for a variety of processes. To
make things not too complicated here, we assume in the following c to be constant
with time, and we shall focus on the E + (r, t) term, so that we can still exploit the
advantages of complex arithmetic.
Within the SVE approximation we are interested in the temporal dependence of
the intensity of the electromagnetic field. According to (H.11), ignoring the position
dependence, the average intensity is
0 c 2 0 c
I (t) = If (t) = E0 (t) = E (t)E + (t). (I.19)
2 2
Here If (t) indicates cycle averaging over fast oscillations. For reference we also
communicate the cycle average energy density of the field
If (t) 0
uf (t) = = E (t)E + (t). (I.20)
c 2
The electric field of a light pulse may be expressed as inverse F OURIER trans-
form, i.e. as a linear superposition of oscillators with frequencies around the car-
rier frequency c in a bandwidth c :

1 )
E0 (t) = E()e it
d. (I.21)
2
To explicitly show the carrier frequency c and phase c , we apply (I.9):

E + (t) = E0 (t)ei(c tc ) = E (t)

eic ) 1 )+ ()eit d.
= E( c )eit d = E (I.22)
2 2
Conversely, the F OURIER transform of the field envelope is

)
E() = E0 (t)eit dt, (I.23)

or alternatively one may write3

)+ () = E(
E ) c )eic = E ) ()

= E + (t)eit dt = eic E0 (t)ei(c )t dt. (I.24)

3 Even )+ ()) = E
though (E + (t)) = E (t), note that (E ) (), since

+ + it
) () =
E E + (t)eit dt = E (t) e dt

= ) ().
E (t)eit dt = E

)
Note that in this terminology E0 (t) is a real quantity. E() will also be real if and
only if E0 (t) is symmetric around its maximum, while E )+ () is usually complex.
However, with these definitions E )+ () typically has a pronounced resonance at
> 0 so that contributions from negative frequencies to the inverse FT (I.22) can
safely be neglected. In other words: artificial truncations of the limits of integrals
are not needed, since anyhow only positive frequencies contribute when working
with E)+ (). These relationships will be illustrated below for specific examples.
I.3 The Intensity Spectrum
If one inserts E (t) according to (I.22) into expression (I.19) for the cycle averaged
intensity and integrates over all times, one recovers P LANCHERELs theorem (I.8)

0 c 1 + 2
I (t)dt =
E () d = I()d = F, (I.25)
2 2
leading to the fluence F (i.e. the total pulse energy Wtot per unit area). With the
second equality we have defined the spectral intensity distribution:
0 c )+ 2 0 c ) 2
I() = E () = E( c ) . (I.26)
4 4
The assumption that this indeed describes the spectrum of the radiation is by
no means as trivial as commonly assumed. Note, that this expression is not the
F OURIER transform of the intensity I (t)!
One may, however, rationalize this definition by remembering how optical (or
other) spectra of electromagnetic radiation are measured: Assume e.g. an interfero-
metric measurement (FABRY-P ROT interferometer, diffraction grating etc.), where
the signal S is generated by superposition of two (or many) amplitudes. A typical
interference pattern emerges from, say two wave fields E + (t) and E + (t + ), which
have traversed different optical path length c (corresponding to a delay time ). The
signal is then

S E (t) + E (t + ) E + (t) + E + (t + ) (I.27)

= 2I0 + 2 Re E (t + )E + (t) ,
and only the interference term [in square brackets] is relevant for the measurement
of the spectrum. If such a measurement is made with short pulses, one obviously
integrates over all times (or sufficiently long times), thus an autocorrelation function
of the field is recorded. To obtain the spectrum one has to F OURIER transform this
signal which according to (I.17) leads to
1 )+ 2
F E (t) * E + (t) = E () I ().
2
I.4 Examples 649
If a prism spectrograph is used a similar argument can be based on the wave optical
interpretation of the index of refraction.
Hence, the definition of the spectrum according (I.26) is justified. Proper normal-
ization is provided by the P LANCHEREL theorem in the form (I.25), which simply
states energy conservation: the fluence F is independent of whether we integrate in-
tensity over all times or spectral intensity distribution over all angular frequencies.
The units are [F ] = J m2 , [I (t)] = W m2 , and [I()] = J s m2 , respectively.
It is important to note that the above use of P LANCHERELs theorem is only pos-
sible for square integrable functions, i.e. for pulses of electromagnetic radiation with
finite duration. If we want to describe CW light sources the integration over all times
has to be replaced by an average over a sufficiently long time Tav . P LANCHERELs
theorem (I.25) must then be replaced by
Tav /2
1 0 c Tav /2 + 2
I (t)dt = E (t) dt = I()d = I, (I.28)
Tav Tav /2 2Tav Tav /2
where I() is now the spectral intensity of the ensemble, measured in units [I()] =
W s m2 , while I is the average intensity with [I ] = W m2 . More about station-
ary, quasi-monochromatic or chaotic light sources, their spectra and their coherence
properties will be found in Sect. 2.1, Vol. 2.
I.4 Examples
In the following we shall present the most commonly used temporal and spectral
profiles of light pulses. The F OURIER transforms collected here can be derived
by analytic integration, where necessary with the help of a suitable computer pro-
gramme (e.g. SWP 5.5, 2005). They are also found in standard textbooks or at
the web (e.g. W EISSTEIN, 2012; W IKIPEDIA CONTRIBUTORS, 2014). We write the
field envelope as

E0 (t) = E0 h t/|t0 | with h(0) = 1 and (I.29)
0 c 2
I0 = E (I.30)
2 0
where E0 is the maximum field amplitude, |t0 | a characteristic width, and I0 the
cycle averaged maximum intensity.
We point out that the spectral intensity distribution as defined in (I.26) refers to a
single pulse. It has the dimension [I()] = J s m2 . Integrated over all frequencies
it gives the fluence F per pulse, i.e. the total pulse energy per unit area:

F= I (t)dt = I0 h2 (t/t0 )dt. (I.31)

If one wants to extend the notion of the spectral intensity distribution to a CW

beam, one has to modify the normalization. As detailed in Sect. 2.1, Vol. 2, one
may assume that the beam consists of a statistical ensemble of single pulses which
are characterized by individual time constants i and their individual spectra Ii ().
One then has to average these spectra and to renormalize in respect of the average
duration c of the pulses. One obtains an overall spectrum:

1
I() = Ii () = w(i )Ii ()di . (I.32)
c 0
This spectrum has now the dimension W s m2 and the integration over all frequen-
cies gives the average, stationary intensity I of the beam.
I.4.1 Gaussian Distribution
Gaussian pulse shapes are probably the most popular ones, mainly due to their con-
venient mathematics. Starting from (I.18) we write the field envelope
2
t
E0 (t) = E0 h(t) = E0 exp 2 (I.33)
G

with a FWHM= 2 ln 2G = 1.665G . The corresponding cycle averaged intensity
(I.19) is now

t2
I (t) = I0 exp 2 2
G

with a FWHM t1/2 = 2 ln 2G = 1.177G . (I.34)
We also note the overall fluence

t2 G
F = I0 exp 2 2 dt = 0 cE0 =
2
G I0 . (I.35)
G 2 2 2
The F OURIER transform of the field envelope is again a Gaussian:

1 2 2
)
E() = G exp G and (I.36)
4

E ) c )ei = G exp 1 G2 ( c )2 i .
)+ () = E( (I.37)
4
Thus, with (I.26) the (experimentally measurable) intensity spectrum becomes
2

G2 I0 c 2
I() = I0 exp G ( c )2 = 2 exp , (I.38)
2 2 G G

with G = 2/G and a FWHM

1/2 = 2 ln 2G = 2 2 ln 2/G = 2.355/G . (I.39)
I.4 Examples 651

One easily verifies that I()d = F as in (I.35). In frequency space 1/2 =
2 ln 2/(G ) and the so called time-bandwidth product is
2 ln 2
t1/2 1/2 = = 0.441. (I.40)

I.4.2 Hyperbolic Secant
Another, quite popular distribution function for describing the time dependence of
short light pulses is the hyperbolic secant
t E0 2E0
E0 (t) = E0 sech = = (I.41)
s cosh(t/s ) et/s + et/s

with a FWHM of 2 ln(2 3)s = 2.634s . The corresponding intensity distribu-
tion is
2
2
I (t) = I0 sech (t/s ) = I0 t/
2
, (I.42)
e s + et/s

with a FWHM t1/2 = 2 ln(1 + 2)s = 1.763s . (I.43)
This normalization leads to a fluence

F = I0 sech2 (t/s )dt = 2s I0 . (I.44)

The F OURIER transform of the field envelope is again a hyperbolic secant
) s 2E0 s
E() = E0 s sech = /2 or (I.45)
2 e s + es /2
)+ () = E0 s eic sech s ( c ) .
E (I.46)
2
The spectral intensity profile for a pulse with the carrier frequency c is

0 c 2 s
I () = 2 2
E0 s sech ( c ) . (I.47)
4 2
In angular frequency space it has a FWHM

1/2 = 4 arcsech(1/ 2)/(s ) = 1.122/s (I.48)
and thus 1/2 = 0.179/s , so that the time-bandwidth product is
t1/2 1/2 = 0.315. (I.49)

I.4.3 Rectangular Wave-Train
For a number of situations a rectangular wave-train is a useful model of a light

pulse. Assume an (angular) carrier frequency c = 2/Tc , let the pulse begin at
time t = ti and end at ti + i ; during this time interval the amplitude is constant and
zero elsewhere, its relative phase i may be chosen freely. In our standard format
(I.18) the electric field of the wave is explicitly
"
1 E0 ei[c (tti )i ] for ti < t < ti + i
E i (t) = e + c.c. (I.50)
2 0 else
The intensity distribution has the same temporal shape with I0 = 0 cE02 /2. It is
convenient to directly evaluate the F OURIER transform (I.24) of Ei+ (t):
t0 +r

)+ () = E0
E exp i(c t c ti i t) dt
i
t0
exp[i(c )i ] 1
= E0 r exp i(ti + i ) (I.51)
i(c )i
)i () = E0 i exp[iti ] 1 exp[ii ] .
E (I.52)
ii
The spectral intensity distribution (I.26) is thus

0 c )+ 2 I0 i2 i ( c )
Ii () = Ei () = sinc2 , (I.53)
4 2 2
with the sinc x = (sin x)/x function. The overall fluence is

F= I (t)dt = I()d = I0 i .

The FWHM of Ii () is given by 1/2 = 5.566/i or in frequency space 1/2 =

0.886/i so that the time-bandwidth product is
t1/2 1/2 = 0.886, (I.54)
which is much broader than for the Gaussian or sech2 distributions.

Finally, we note in passing that all above expressions apply to a simple rectangu-
lar pulse (without oscillations) when one just sets c = 0 and i = 0.
I.4.4 Rectangular Spectrum
Another important case is a rectangular spectrum. Such a situation may arise, e.g.
if one spectrally filters a broad band light CW source with a sharp band-pass filter
(e.g. in stellar interferometry). The spectrum would then be
I.4 Examples 653
*
I / for c 12 < < c + 12
I() = (I.55)
0 else,
where I is the average stationary intensity of the beam. The integral (I.3) can easily
be carried out in this case and one obtains the inverse F OURIER transform:

I ic t t sin x
F 1 I() = e sinc with sinc(x) = . (I.56)
2 2 x
This formula turns out to be useful e.g. for understanding stellar interference spec-
troscopy (see Sect. 2.1.4, Vol. 2).
I.4.5 Exponential and LORENTZ Distributions
The One-Sided Exponential Distribution
*
0 if x < 0
h(x) = (I.57)
exp(x) if x 0
is used to describes an exponentially decaying field (oscillating at a carrier fre-
quency c ) and its cycle averaged intensity, respectively:

t
E(t) = E0 h and
2e

t
I (t) = I0 h with FWHM t1/2 = e ln 2 = 0.693e . (I.58)
e
The overall fluence is simply F = e I0 .

The F OURIER transform of the field envelope is complex in this case (asymmetric
distribution):
) 2e E0
E() = , and
1 + 2ie
)+ () = ei 2e E0
E .
1 + 2ie ( c )
The intensity spectrum becomes a L ORENTZ distribution
1 I0 e2 I0 2 /4
I() = = , (I.59)
1 + [2e ( c )]2 2 2 /4 + ( c )2
with a FWHM 1/2 = = 1/e . (I.60)
In frequency space this gives 1/2 = 1/(2e ) = 0.159/e , so that the time-
bandwidth product becomes
t1/2 1/2 = 0.110. (I.61)
This is an extraordinarily small time-bandwidth product in comparison with the

Gaussian or sech2 case. Note, however, that the temporal shape of this pulse
is
characterized by an extremely wide wing at positive times. Normalization
I()d = F is verified easily with dx/(1 + x 2 ) = .
The Two Sided Exponential Distribution

E0 (t) = E0 exp |t|/(2ee ) (I.62)
is a somewhat pathological pulse envelope, with

E + (t) = E0 exp i(c t ) |t|/(2ee ) . (I.63)
It has an intensity profile

I (t) = I0 exp |t|/ee (I.64)
with a fluence F = 2I0 ee , (I.65)
and a FWHM t1/2 = 2 ln 2ee = 1.386ee .

Die F OURIER transform of the field envelope is now a (real) L ORENTZ distribu-
tion
) 4ee E0
E() = , or alternatively (I.66)
1 + (2ee )2
)+ () = E0 ei
E , (I.67)
2 /4 + ( c )2
with an FWHM = 1/ee . The intensity spectrum is given by
2
I0
I () = (I.68)
2 2 /4 + ( c )2

with a FWHM 1/2 = 2 1 = 0.644 = 0.643/ee , (I.69)
so that the time-bandwidth product becomes in this case

tL 1/2 = 2 1(ln 2)/ = 0.142. (I.70)
We point out that (I.68) is not a L ORENTZ profile, and may, e.g. not easily be convo-
luted. We emphasize that in any case this is not a typical profile for characterizing a
real light pulse: it has a cusp at time zero and extremely wide tails, both in the time
and in the frequency domain.
I.5 Fourier Transform in Three Dimensions 655
For reference we also write (I.67) in normalized form
1
L(, )= (I.71)
2 /4 + ( c )2
2

with = 1. The inverse FT is given by (I.63) as
L(, )d

1 1 |t|/2+ic t

F 1 L(, ) =
L(, )eit d = e (I.72)
2 2
with = 1/ee . We have set here = 0.
We finally can now supplement the proof for the additivity of linewidths when
convoluting two L ORENTZ profiles, as introduced by (G.21). According to (I.12) the
inverse FT of a convolution of two L ORENTZ profiles is proportional to the products
of their inverse FT:

F 1 L 1 (, 1 ) L 1 (, 2 ) (I.73)
1 1 |t|/2i1 t 1 1 |t|/2+i2 t 1 (1 +2 )|t|/2+i(2 1 )t
= 2 e e = e .
2 2 2
Now we just transform this back, compare with (I.72), and obtain:
1 (1 +2 )|t|/2+i(2 1 )t
L1 (x, 1 ) L1 (x, 2 ) = F e (I.74)
2
2 (1 + 2 )2 /4
= .
(1 + 2 ) (x (2 1 ))2 + (1 + 2 )2 /4
Thus, the convolution of one L ORENTZ profile with another one leads again to a
L ORENTZ profile. The FWHM of the two profiles are just added:
= 1 + 2 . (I.75)
I.5 Fourier Transform in Three Dimensions
F OURIER transforms can be extended into n dimensions. Three dimensional FTs

will be employed when 3D wave-packets are introduced (Eq. (1.106) in Sect. 1.4,
Vol. 2). There, in principle, one has to integrate over plane waves exp(ikr) in full
momentum space k. The same holds for FBA (see Sects. 6.6, 8.2, and 8.4.4, Vol. 2).
Practical evaluation of these 3D FTs and their inverse is significantly more complex
than in one dimension. Complete operational tool-sets for the 2D and 3D cases have
recently been developed by BADDOUR (2009) and (2010), respectively. We briefly
summarize some key results for the 3D case.
The definition of the 3D F OURIER transform in analogy to (I.2) reads

) )
X(k) = X(kx , ky , kz ) = X(r)eikr d3 r (I.76)
R3
and its inverse in analogy to (I.3) is

1 )
X(r) = X(x, y, z) = X(k)e ikr 3
d k. (I.77)
(2)3 R3
We note in passing that the D IRAC delta function in three dimensions is

1
kk = ei(kk )r d3 r. (I.78)
(2)3 R3
Its characteristic property is expressed by

f (x) = f x x x d3 x
R3

2
= d sin d f x x x x 2 dx , (I.79)
0 0 0

specifically this ensures x x d3 x = 1. (I.80)
R3
In view of (I.79) one may also express the 3D delta function as
(x x )
x x = 2 . (I.81)
x sin
For problems with some kind of spherical symmetry (e.g. in atomic physics)
it is useful to rewrite (I.76) in spherical polar coordinates, with standard position
coordinates (r, r , r ) and coordinates in wave vector space (k, k , k ):
2
) k , k ) =
X(k, X(r, r , r )eikr r 2 sin r drdr dr . (I.82)
0 0 0
One expands X(r, r , r ) into a series of spherical harmonics

X(r) = X(r, r , r ) = X m (r)Y m (r , r ), where (I.83)
=0 m=
2

X m (r) = X(r, r , r )Y m (r , r ) sin r dr dr . (I.84)
0 0
By expanding also the plane wave (I.82) according to (J.13) one obtains the 3D
F OURIER transform in k space

+
) = X(k,
X(k) ) k , k ) = 4 (i) X m (k)Y m (k , k ), (I.85)
=0 m=
with the th order spherical H ANKEL transform of X m (r)

+ ' (
X m (k) = S X m (r) = X m (r)j (kr)r 2 dr. (I.86)
0
The inverse 3D F OURIER transform in analogy to (I.3) is defined as

1 )
X(r) = X(k)e ikr 3
d k, (I.87)
(2)3
which complementary to (I.82) may again be expressed in spherical coordinates.
)
Explicitly, one expands X(k) into a series of spherical harmonics

) = X(k, k , k ) =
X(k) ) m (k)Y m (k , k ),
X where (I.88)
=0 m=
2
) m (k) =
X ) k , k )Y m
X(k,
(k , k ) sin k dk dk . (I.89)
0 0
Using now the inverse spherical H ANKEL transform of X) m (k)

+
) 2 )
X m (r) = X(k)j (kr)k 2 dk, (I.90)
0
one finally obtains the inverse 3D F OURIER transform in spherical coordinates:

1 +

X(r) = X(r, r , r ) = ) m (k)Y m (k , k ).
(i) X (I.91)
4
=0 m=
Comparing (I.85) and (I.88) one also finds that

+ ' (
) m (k) = 4(i) X m (k) = 4(i) S X m (r)
X
from which the F OURIER expansion coefficients may be obtained.

Sect. 6.6 in Vol. 2 and Sect. 5.5.2, respectively).
FT: F OURIER transform, see Appendix I.
References
BADDOUR, N.: 2009. Operational and convolution properties of two-dimensional Fourier trans-
forms in polar coordinates. J. Opt. Soc. Am. A, 26, 17671777.
BADDOUR, N.: 2010. Operational and convolution properties of three-dimensional Fourier trans-
forms in spherical polar coordinates. J. Opt. Soc. Am. A, 27, 21442155.
SWP 5.5: 2005. Scientific work place, Poulsbo, WA 98370-7370, USA: MacKichan Software,
Inc. http://www.mackichan.com/, accessed: 9 Jan 2014.
W EISSTEIN, E. W.: 2012. Fourier transform, Wolfram Research, Inc., Champaign, IL, USA.
http://mathworld.wolfram.com/FourierTransform.html, accessed: 9 Jan 2014.
W IKIPEDIA CONTRIBUTORS: 2014. Fourier transform, Wikipedia, The Free Encyclopedia.
http://en.wikipedia.org/wiki/Fourier_transform, accessed: 9 Jan 2014.
Continuum
J
J.1 Normalization of Continuum Wave Functions
Many physical problems require the inclusion of continuum states in addition to

discrete bound states e.g. scattering processes, photoionization, electron impact
ionization an the like. In all these cases the normalization of the continuum states
poses a specific problem. We briefly indicate here how this problem may be solved
as one way among several possible approaches.
For the discrete spectrum the radial wave functions Rn (r) are normalized in the
usual fashion:

Rn (r)Rn (r)r 2 dr = un (r)un (r)dr = nn . (J.1)
In the following we use the radial functions in the form u(r) = rR(r). In the con-
tinuous spectrum, following B ETHE and S ALPETER (1957) Eq. (4.11ff), they are
normalized in T scale by demanding
T + T
druT (r) uT (r)dT = 1. (J.2)
0 T T
Here T (k) is any function of the electron wavenumber k in the continuum, possibly
k itself. In the theoretical treatment of photoionization normalization in scale is
frequently used, with being the energy of the electron in the continuum. The in-
tegration over dT needs only to be over a small interval 2 T around T , since all
other contributions cancel.1
With such normalized continuum functions one may represent any wave function
including the continuum as partial wave expansion (for simplicity written here only
for one electron systems):
1 Summation of (J.1) over all n states in the discrete spectrum would be the equivalent to this second
integration in (J.2). Since only one state contributes, the result is also =1.

660 J Continuum
(r, , ) (J.3)
+ ,
1
= Y m (, ) an m un (r) + dT (k)a m (T )uT (r) .
r k=0
m n= +1
The relation between normalization in T scale and in k scale is given by

1/2 1/2
dT dk
uT (r) = uk (r) = uk (r) . (J.4)
dk dT
Specifically, for normalization in energy scale, with

2 k 2 me a02 k 2 M 1 2
= = Eh and k = (J.5)
2M M 2 me a0 Eh
we obtain
(for a system with reduced mass M)

d me dk M 1 M 1
= Eh a02 k and = = . (J.6)
dk M d me Eh a02 k me a0 (2Eh )1/2
And for electrons with M me and = k 2 /2 we obtain most simply in a.u.
d/dk = k. (J.7)
To illustrate this by way of example, we discuss a general radial wave function in

the continuum. In a.u. it is asymptotically (see e.g. B URKE, 2006)

Z
u (r) = b sin kr /2 + ln(2kr) + + . (J.8)
k
Here Ze is the charge of a C OULOMB field (if present) which enters into =
arg ( + 1 iZ/k) and the slowly varying logarithmic C OULOMB phase. The
standard phase shift reflects the influence of further, non-Coulombic interactions.
For a free particle (Z/k) ln(2kr) as well as and vanish. To obtain the normal-
ization factor b in k scale normalization, we neglect the logarithmic phase, write
= /2 + + , and evaluate
k+ k k+ k
b
dk u (r) = i dk exp i k r + c.c.
k k 2 k k
sin kr
= 2b sin(kr + ) .
r
Inserting this into (J.2), and substituting the rapidly oscillating term sin2 (kr + )
by its average value 1/2, one may integrate over all r

sin kr
2b 2
dr sin(kr + ) sin(kr + ) = b2 ,
0 r 2
J.2 Plane Waves in 3D 661

where the identity 0 sin(|a|r)
r dr = 2 has been applied, so that b = 2/ . Hence,
normalized in k scale the radial wave function becomes

2 Z
u (r) = sin kr /2 + ln(2kr) + + . (J.9)
k
To determine the dimension of uk (r), we write the units in the expression under
the integral in (J.2) [udrdk] = 1, or as dimensional equation u2 L1 L1 = 1 so that
u = 1, i.e. uk (r) in k scale normalization is dimensionless. For clarity it might be
useful to explicitly write out the a.u., i.e. to replace r r/a0 and Z/k Z/(ka0 )
in (J.9), so that the expression becomes independent of the units used.
For normalization in scale we use (J.4)(J.7) and obtain in a.u.:

2 dk Z
u (r) = sin kr /2 + ln(2kr) + +
d k

2 Z
= sin kr /2 + ln(2kr) + + or explicitly (J.10)
k k

2 me Eh /2 1/4 Z
= sin kr /2 + ln(2kr) + + .
Eh a0 M ka0
(J.11)
The latter expression is independent of the units used. The prefactor (Eh a0 )1/2
is again obtained from a dimensional analysis: normalized in scale from (J.2)
we have [u2 drd] = 1, so that u2 = Enrg1 L1 and u = Enrg1/2 L1/2 . Note that
for electron scattering the mass factor me /M 1 disappears. Different authors use
slightly different notations and energy units (often without mentioning which).2
J.2 Plane Waves in 3D
J.2.1 Expansion in Spherical Harmonics
Without proof we note that a plane wave in 3D can be written as

(k) (r) = eikr = 4 i j (kr) Y m (k , k )Y m (r , r ), (J.13)
=0 m=
2 For example, expression (J.11) was introduced by C OOPER (1962) to photoionization as

u (r) 1/2 1/4 sin 1/2 r /2 + Z 1/2 ln 2 1/2 r + . (J.12)
Here, = 2/Eh , i.e. energies are measured in RYDBERG units = Eh /2, lengths in a0 .
662 J Continuum
with the polar coordinates (r, r , r ) representing r, while (k, k , k ) gives the mag-
nitude and direction of the wave vector k, and j (kr) = u (kr)/(kr) are spherical
B ESSEL functions. They are solutions of the radial S CHRDINGER equation (2.110)
for vanishing potential:
2
1 d2 u k ( + 1)
+ u (r) = 0. (J.14)
2 dr 2 2 2r 2
As one easily verifies, the most simple ones are:
sin x sin x cos x
j0 (x) = sinc x = and j1 (x) = . (J.15)
x x2 x
All others may in principle be derived from the recursion formula:
2 + 1
j +1 (x) = j (x) j 1 (x). (J.16)
x
Asymptotically the following relations hold:
" #
x /[(2 + 1)(2 1)(2 3) ] for x
j (x) = . (J.17)
sin(x /2)/x for x
With the addition theorem (C.22) and the angle between k and r one may also
write (J.13) as:

(k) (r) = eikr = (2 + 1)i j (kr)P (cos ). (J.18)
=0
J.2.2 Normalization in Momentum and Energy Scale
The probability density | (k) |2 of the plane waves (J.13) or (J.18) is given per
wavenumber to the third power (dimension L3 ) and per volume (L3 ), so they are
overall dimensionless. They may be normalized in k scale (see e.g. OVCHINNIKOV
et al., 2004):
1
)(k) (r) =
eikr = r|k . (J.19)
(2)3/2
To verify this normalization we first integrate over position space using (I.78):

k 2 |k 1 k = )(k) (r)
)(k) (r)d3 r a03 ka0 k a0 .
2 1
This is the continuum equivalent to the usual orthogonality a|b = ab of bound

state wave functions. Since ka0 is dimensionless, the expression is given per
wavenumber interval cubed (dimension L3 ) and thus reflects the built in density
of states. Finally, with (I.80), integration over k spaces confirms the correct normal-
ization of (J.19):

3
k k k d k = a03 ka0 k a0 d3 k 1.
To obtain 3D normalization in the scale, we note that

dk
d3 k = k 2 dkd = k 2 dd.
d
Hence (J.4) becomes in 3D
1/2
dk
)(k) (r) = k
)(k) (r).
(J.20)
d
Specifically for conversion into energy scale with (J.6)

1
d3 k = kdd, (J.21)
Eh a02
holds, and the 3D plane wave (J.19) normalized in energy scale becomes

ka0 ikr
) (r) = (2)
(k) 3/2
e . (J.22)
a03 Eh
)(k) (r)|2 is
Its dimension is now Enrg1/2 L3/2 so that the probability density |
correctly given per energy and volume.

References
B ETHE, H. A. and E. E. S ALPETER: 1957. Quantum Mechanis of One- and Two-Electron Atoms.
Berlin, Gttingen, Heidelberg: Springer Verlag, 369 pages.
B URKE, P.: 2006. Electron-atom, electron-ion and electron-molecule collisions. In: G. W. F.
D RAKE, ed., Handbook of Atomic, Molecular and Optical Physics, 705729. Heidelberg, New
York: Springer.
C OOPER, J. W.: 1962. Photoionization from outer atomic subshells. A model study. Phys. Rev.,
128, 681693.
OVCHINNIKOV, S. Y., G. N. O GURTSOV, J. H. M ACEK and Y. S. G ORDEEV: 2004. Dynamics of
ionization in atomic collisions. Phys. Rep., 389, 119159.
Index of Volume 1
Symbols Adjoint operator, 97, 576

1, identity matrix, 99 ADK theory, 437
3D F OURIER transform, 655657 Air mass coefficient (AM), 36
3j symbols, 300, 564568 A IRY function, 280
orthogonality, 565, 566 Al atom, G ROTRIAN diagram, 518
special cases, 567, 568 Alignment, 173, 588, 620
symmetries, 565, 566 angle, 173
6j symbols, 311, 568572 parameter, 620
orthogonality, 569, 570 Alkali atoms, 144, 146, 151, 165
special cases, 571, 572 comparison with H atom, 146
symmetries, 569, 570 overview about term energies, 146
9j symbols, 572 quantum defect, 146, 147
spectroscopy, 145, 146
A Alkaline earth metals
Above-barrier ionization, 436, 437 fine structure, 361, 362
Above-threshold ionization (ATI), 266, 441 G ROTRIAN diagram, 371
of Ar, 441 Angular dispersion, 275
of C60 , 441, 442 Angular momentum, 5, 65, 66, 72, 107117
Absorption, 193196 algebra, 575592
coefficient, 18, 241, 422 commutation rules, 559
molar, 18 conservation, 81
cross section, 18, 240242 E1-transitions, 196
broad band light, 242 coupling, 297, 298
monochromatic light, 241 definition, 109, 559573
definitions, units, 180 eigenstates, 561
edge, 521, 525 intrinsic, 78
E INSTEIN coefficient, 180 matrix elements, 561
inner shells, 520524 operators, 109, 111
introduction, 17, 18, 178180 scalar product, 298, 299, 585, 586
probability, 190192 projection theorem, 578
rate, 179, 191 real and helicity basis, 560
X-ray, 520525, 530 Anomalous magnetic moment of the electron,
C OMPTON scattering, 526 331
pair production, 526 Anomalous Z EEMAN effect, 380
photoionization, 526 Anti-symmetrization, 346, 357, 358, 388
T HOMSON scattering, 526 Areas of physics, 2
Addition theorem for Ckq , 579 Atom model, 63

666 Index of Volume 1
Atom radius, 121 B OSE -E INSTEIN condensate, 8

Atomic beam, 7375, 282, 317 B OSE -E INSTEIN statistics, 23
Atomic core, 148 Boson, 20, 22, 139
Atomic form factor, 41, 430, 431 Boundary condition, 104
Atomic hydrogen B RAGG reflection, 40, 41
eigenfunctions, 123126 B REIT-R ABI formula, 393, 468
Atomic orbitals, 118, 122, 125, 126 Z EEMAN effect in hyperfine structure,
Atomic radius, 29, 68, 143, 144 467471
Atomic size, 64 Bremsstrahlung (X-ray), 530
Atomic units, 67, 68, 118, 119 B RILLOUIN zone, 41, 133
Atoms in a magnetic field, see Z EEMAN effect
Atoms in an electric field, see S TARK effect C
Atoms in intense laser fields, 432, 442 C atom, 22
Attometer, 6 G ROTRIAN diagram, 514
Attosecond, 8 Candela, 38
Attosecond laser pulses, 440 CCD camera, 59, 269, 277
Aufbau principle, 138, 139 Central field approximation, 496
AUGER electron, 522 Centre of mass system, 119
Autocorrelation function, 623, 626, 628, 646, Centrifugal potential, 117, 120, 123
648 Centrosymmetric problems, 108
Autoionization, 260, 366370 Cesium fountain atomic clock, 320
Avoided crossing, 392394 Characteristic X-ray radiation, 530
S TARK effect in RYDBERG states, 411 Chemical shift
B NMR spectroscopy, 488
Band structure, 28 Chemical-potential, 23, 24, 105107
electronic, in a solid, 132 Classical forced oscillator
Barions, 48, 49 model for photon absorption cross section,
Barn, 524 240
Basis spin function, 115 C LAUSIUS -M OSSOTI formula, 421
Basis states: real and complex, 595599 C LEBSCH -G ORDAN coefficients, 300,
Basis vectors, 95 564568
G ROTRIAN diagram, 372 orthogonality, 565, 566
BESSY, 42, 532 symmetries, 565, 566
(anisotropy) parameter, 260, 263, 264, 269, CODATA, 9, 321, 329, 330, 334336, see also
529 Fundamental physical constants
B EUTLER -FANO profile, 368 data bank, 551
Black body radiation, 3134 Coherently excited states, 217220
Blazed grating, blaze angle, 277 Collisional line broadening, 233, 234
B LOCH wave function, 132 Colour temperature, 39
B OHR, 2, 64 Commutation rules, 114, 115, 634
magneton, 378, 449 Commutator, 96, 101
model of the atom, 6470 Complex spectra, 512519
comparison with quantum mechanics, C OMPTON effect, 2, 28, 29
127, 128 C OMPTON wavelength, 29
limits, 69, 70 Configuration interaction (CI), 363366, 508,
orbital radius, 67, 124 509
B OLTZMANN Continuum, 659663
constant, 19, 33 normalization, 659661
distribution, 20, 24, 25, 235, 624, 627 of eigenvalues, 102
factor, 23 wave function, 262
statistics, 2123 Convolution, 623629
B ORN approximation Gaussian profile, 626, 627
first, applied to photoionization, 257 Hyperbolic secant, 627, 628
first, for X-ray photoionization, 524 L ORENTZ profile, 628, 654
Index of Volume 1 667
theorem for F OURIER transform, 645 Density of states, 2124, 238, 240
VOIGT profile, 628, 629 particle in a box, 105, 106
C OOPER minimum, 260, 529 Detailed balance
Coordinate rotation, 575, 605611 principle of, 184
exercise, 207 Detuning, 229
Coordinate system Diamagnetism, 396398
atomic vs. photon, 174 Dielectric function, 57
cartesic and polar, 108 Diffraction
electron vs. nucleus, 295 D EBYE-S CHERRER, 58, 59
Core electron, 148, 151 experiment, 60
Core potential, 149 He scattering, 60
alkali atoms, 165 image, 42
Correlation function LEED, 59
1st order, 625 Dimensional analysis, 119, 192, 554, 556, 661
Correspondence principle, 127 Dipol vector, 183
C OULOMB gauge, 631 Dipole approximation, 189, 631640
C OULOMB integral, 356, 359, 365, 506 electric, 635
C OULOMB law, 50, 59 magnetic, 250254
C OULOMB potential, 119, 121, 123 Dipole excitation, linear combination of states,
H atom, 121 217225
screened, 148, 149 Dipole matrix element, 193
C OULOMB wave, 262 length approximation, 193, 636
Coupling
velocity approximation, 193, 636
jj , 301, 354, 360
Dipole operator, 594
LS, 301, 307, 354, 360, 361
magnetic field, 386
LS vs. jj , 512514
multi-electron system, 264, 363
RUSSEL -S AUNDERS, 301, 354, 360, 361,
Dipole oscillator, classical, 182
508
Dipole radiation
break down, 362
angular characteristic, 203212
spin-orbit, see Spin-orbit
Dipole transition
Crystal lattice, 40
C URIE constant, 396 amplitude, 193
C URIEs law, 396 E1 transitions in the H atom, 201203
Cyclotron frequency, 54, 332 in He, 362365
electron, 79 length approximation, 636
Cylindrical capacitor, 52 matrix element, 190, 635
operator, 190, 193, 263
D selection rules for E1 transitions,
Damping constant, 229 196203
DARWIN term, 305 selection rules for E1-transition, 202
D E B ROGLIE, 57, 89 velocity approximation, 636
wavelength, 5759 D IRAC delta function, 644
Decay constant, 15 D IRAC equation, 79, 93, 296, 303, 333
Decay rate, 15 Direct product, 578
Degeneracy, 21, 22 Dispersion, 421
, 123 anomalous, 422
m, 111 close to resonance, 232, 233
, removal of, 137146 normal, 420, 422
m, removal of, 130, 131 Dispersion relation, 131134
Degree of coherence matter waves, 94
1st order, 625 Displaced terms
Degrees of freedom, 19 alkaline earth term schemes, 372
Density distribution, 125 C atom, 515
Density functional theory, 510512 Ne atom, 518
D OPPLER free spectroscopy, see Spectroscopy, Electron, 4951, 114

D OPPLER free angular momentum, 50
D OPPLER broadening, 234, 236, 285 classical electron radius, 50
D OPPLER effect, 285, 317 C OMPTON wavelength, 29
classical, 14 elementary charge, 50, 51
quadratic, 13 g factor, 79
relativistic, 13 mass, 68
D OPPLER narrowing, 283 M ILLIKAN experiment, 50
D OPPLER profile, 287 orbital magnetic moment, 129
Double slit experiment, YOUNGs, 88, 89 point like, 49
Doubly excited states, 348, 365367 spin, 50, 70, 78, 112, 114, 128
Drehimpuls, 81 Electron bunches, 532
Dressed states, 418 Electron configuration, 138, 142, 144, 151,
D RUDE frequency, 57 302, 500, 508, 509
Duality Electron diffraction
wave-particle, 2, 4 D EBYE-S CHERRER, 58, 59
Duality, wave-particle, 58 LEED, 59
Electron hole, 521
E Electron magnetic moment, 50, 79, 81, 294,
E1-transitions, 635 387, 449
Echelle spectrometer, 277 anomaly, 331, 336
Effective mass of an electron, 132 Electron shell, 139, 140
Effective nuclear charge, 159 Electron spin, 114116
Effective potential, 117 resonance spectroscopy (ESR), see
Eigenenergy, 91 Electron paramagnetic resonance
Eigenfunction, 91, 99 (EPR)
nodes of the, 103 Electron storage ring, 53
of momentum, 102 synchrotron radiation, 531
Eigenstate, 98, 99 Electronvolt, 51
Eigenstates of angular momentum, 109113 Electrostatic potential, expansion, 614616
Eigenvalue, 97 Electroweak interaction, 44
Eigenvalue equation, 97, 99 Ellipticity angle, 172, 174
energy, 129 Emission, 193196
z , 110
of L inner shells, 520524
of momentum, 102 spectrum, 130
Eigenvalue problem, 102 Emittance, 532
Eigenvector, 97, 99 Energy analyzer, 52, see also Energy selector
E INSTEIN Energy balance, 120
E = mc2 , 3, 10 Energy conservation
photoelectric effect, 31 operator form, 91
E INSTEIN A and B coefficients, 184, 185, 193, relativistic, 11, 28
194, 212, 215, 216, 228, 230, 241, Energy levels
314, 489 fine structure splitting, 294
E INSTEINs paradigm on speed of light, 425 Energy packet, 27
E INSTEIN - DE -H AAS effect, 79, 81 Energy quantization, 26
Electric dipole (E1) transitions, 588 Energy scales, 7
Electric dipole moment Energy terms
of the electron-nucleus pair, 189 H atom, 122
Electric quadrupole (E2) transition, 250254, Energy zero
588 H atom, 118
Electric quadrupole moment Entanglement, 354, 609, 610
atomic nuclei, 449 EPR spectroscopy, 484487
Electromagnetic spectrum, 31 high B field, 487
Electromagnetic waves, 170176 X band, 486
Equivalence of mass and energy, 10 relativistic correction, 304, 305

Ergodicity, 626 spin-orbit term, 306
Error function H atom, 293, 296, 297, 306, 307
complementary, 629 He and He like ions, 360, 361, 362
complex, 629 interaction, see Spin-orbit interaction
E ULER angles, 605 interval rule (L AND), 308, 361, 362, 372
Exchange Hg atom, 372
boson, 44 Na D doublet, 293
integral, 356, 357, 359, 365 normal ordering of terms, 308
interaction, 343 quantum defect, 308, 309
spin orientation, 358, 360 splitting, 293, 308
operator, 351353 theory and experiment, 303310
Expectation value, 98, 99, 126 transitions
observable, 98 branching ratios, 315
r k for the H atom, 126, 127 multiplet, 312, 315
spin component, 115 transitions and selection rules, 310316
the momentum in a 1D box, 103 Fine structure constant, 9, 69, 293, 296
Experiment of B ETH, 176 electromagnetic coupling, 326
Extinction coefficient, 241, 421 high precision measurement, 335
molar, 18 Finesse, see FABRY-P ROT interferometer
Four vector (momentum), 11
F F OURIER transform, 643658
FABRY-P ROT interferometer, 279281 analysis, 54
finesse, 280 exponential distribution, 653655
finesse coefficient, 280 Gaussian, 650, 651
FADEEVA function, 629 inverse, 643
FANO lineshape(, 366 L ORENTZ, 653655
FANO lineshape, 369 rectangle, 652
Fast light, 422427 sech, 651, 652
F ERMI contact term, 456, 458, 459 spectroscopy, 643
F ERMI energy, 25, 106 Free electron gas, 105107
F ERMI level, 107 Free electron laser (FEL), 542, 543
F ERMIs golden rule, 238 Free spectral range, 279
F ERMI -D IRAC statistics, 23, 106 FT-ICR, 54, 55
Fermion, 20, 22, 44, 138 Fundamental interactions
Ferromagnetism, 80, 358 the four, 4351
F EYNMAN diagrams, 324326 Fundamental physical constants, 67, 551, 553
energy conservation, 326
for ge 2, 335 G
L AMB shift, 326 g 2, see Electron magnetic moment, anomaly
neutron decay, 49 Galaxy, 6
pair annihilation, 325 Gas kinetic cross section, 20
pair generation, 325 Gaussian profile, 235
propagator, 326 convolution, 626, 627
self-energy of the electron, 325 Generalized cross section for multi-photon
vacuum polarization, 325 processes, 245
vertices, 326 Geonium atom, 333335
Fine structure, 293316 gJ factor
alkali atoms, 307, 309 definition, 77
alkaline earth metals, 362, 372 quantum mechanical derivation, 380, 381
and electron spin, 293 vector model, 381, 382
BALMER H line, 288 Gratings, 274279
D IRAC theory, 303 G ROTRIAN diagram, 514518, 520
DARWIN term, 305 Al atom, 518
alkaline earth atoms, 371 ground state, 348

Be atom, 372 H AMILTON operator, 345
C atom, 514 He like ions
H atom, 202 ground state, 349
He atom, 343 H EISENBERG representation, 100
Hg atom, 372 H EISENBERG uncertainty relation, 100, 101
Li atom, 145 Helicity, 196
N atom, 515 Helicity basis, 171, 172, 174, 267, 560
Ne atom, 517 angular momentum, 560
O atom, 516 transition amplitudes, 198200
Ground state Hemispherical capacitor, 52
H atom, 64 Hermitian operator and conjugate, 97, 576
Group in periodic system of elements, 140 HFS, see Hyperfine structure
Group index, 424 Hg atom, 372374
Group velocity, 422424 G ROTRIAN diagram, 372
Gyromagnetic ratio, 71 High harmonic generation, 439, 440
plateau, 440
H H ILBERT space, 100
H atom, 69, 117128, 296, 302, 305, 306, History of physics, 2, 3
316 Hole burning, 284, 285
1S2S transition, 290, 291 H UNDs rules, 357
BALMER series, 126 H UYGENS -F RESNEL principle, 89
density plots, 124126 Hydrogen anion, 350
energy levels, 122 Hydrogen like ions, 68
expectation values of r k , 126, 127 Hydrogen maser, 459
fine structure, 287 H YLLERAAS wave function, 350
fine structure transition, 310 Hyperbolic secant
H AMILTON operator, 117 convolution, 627, 628
in a magnetic field, 129 Hyperfine structure, 287, 447493
L AMB shift, 317 coupling constant, 452
LYMAN series, 126 coupling tensor, 456, 457, 480
orbitals, 125 deuterium, 454, 460
PASCHEN series, 126 E1 multiplet transitions, 460
radial wave function, 120, 121, 123, 124 H atom, 453, 459
spectrum, 65, 69, 126 intervall rule, 481
wave functions (2D plot), 125 isotope shift and electrostatic interaction,
Hadrons, 49 471482
H AMILTON operator, 9194, 99, 101, 117, L ANDs interval rule, 452
121, 130 magnetic dipole and quadrupole, 481
He atom, 345 magnetic dipole interaction, 452
magnetic fields, 129 mass effect, 473, 474
H ANKEL transform Na atom, 282, 287, 290, 454, 460
spherical, 657 nuclear quadrupole moment, 477481
H ARTREE equations, 498500 quadupole interaction, 481
H ARTREE -F OCK, 503510 total angular momentum, 449
equations, 506, 508 vector diagram, 451
restricted, 504 volume shift, 475, 477
unrestricted, 504 I
He atom, 22, 59, 341375 Identity matrix,
1, 99
0th order approximation, 346348 Independent particle model, 345, 355, 496, 497
diffraction, 59 Index of refraction, 57, 62, 400, 420422
electron exchange, 351355 Induced transitions
excited states, 351360 dipole approximation, 189, 190
G ROTRIAN diagram, 343 probability, 215, 216
Insertion device, 540 L AMBERT-B EER law, 18, 178, 524

Intensity, 34, 632, 633 L AND g factor, see gJ factor
cycle averaged, 632 L ANDs interval rule
Intensity spectrum, 648, 649 fine structure, 308
Intercombination lines hyperfine structure, 452
forbidden in He, 363 L ANGEVIN function, 396
Interference, 89, 90 L ANGMUIR -TAYLOR detector, 75, 317
Interferometer, 278281 L APLACE expansion, 613
FABRY-P ROT, 279281 L ARMOR frequency, 72, 77, 79, 331, 485
free spectral range, 279 Laser based X-ray sources, 543, 544
opitcal path difference, 279 Lattice plane, 40, 42, 59
optical path difference, 278 L EGENDRE polynomial, 563
resolving power, 278 associated, 111
Interval rule, see Fine structure Leuchtelektron (valence electron), 144
Invariant mass, 10 Level splitting, 130
Ion beam spectroscopy, see Spectroscopy Li atom, 146
Ion cyclotron resonance G ROTRIAN diagram, 145
Spectrometer, 54 Light quantum, 27
Ionization Light scattering, 427432
above-barrier, 436 coherent, 430
non-sequential, double, 438 C OMPTON, 430, 431
Ionization potential, 27, 75, 142, 146 from relativistic electrons, 431
of alkali atoms, 147 incoherent, 431
IR spectral range, 37 M IE, 427
Irradiance, 34 R AYLEIGH, 428, 429
Irreducible representation T HOMSON, 429
rotation group, 560, 575, 606 Light storage, 281
tensor operator, 560 Light year, 6
Limits of classical physics, 87
J Line broadening, 227238
jj coupling, see Coupling, 513 by finite measuring time, 288
K homogeneous, natural, 231, 232
K shell, 139, 151 inhomogeneous, 236
K ELDYSH parameter, 434, 436 Lorentzian linewidth, 234
Kinematic correction, 68, 119, 321, 323, 474 Line spectra, 2
Kinetic gas theory, 1820 Line strength, 212, 636, 637
K IRCHHOFFs diffraction theory, 60, 89 Line triplet, 130
K LEIN -G ORDON equation, 93 normal Z EEMAN effect, 131
KOHN -S HAM Liquid drop model for nuclear radius, 476
equations, potential, orbitals, 510 Long range potentials, 414
KOOPMANs theorem, 509 induced dipole induced dipole, 417, 418
monopole induced dipole, 416
L monopole monopole, 414
L shell, 139, 151 monopole permanent dipole, 415
L AGUERRE polynomials, 121 monopole quadrupole, 415
L AMB dip, 286 permanent dipole induced dipole, 416
L AMB shift, 316 permanent dipole permanent dipole, 415
1st order perturbation theory, 328 quadrupole quadrupole, 416
BALMER H, 316 L ORENTZ factor, 10
experiment of L AMB and R ETHERFORD, L ORENTZ force, 55, 295
317, 318 L ORENTZ profile, 229
highly charged ions, 322, 324 convolution, 628
optical precision spectroscopy, 319, 322 numerical examples, 231, 232
theory, 326, 331 LS coupling, see also Coupling, 512
LS interaction, see Spin-orbit interaction Mean free path length, 18

Lumen, 38 Mesons, 49
Luminous efficacy, 38 Metastable states of rare gases, 517
Luminous efficiency, 38 M ICHELSON interferometer, 278
photopic, 37 Microscope resolution according to Abb, 62
Luminous flux, 38 M IE scattering, 427
Luminous intensity, 38 M ILLER indices, 41
Molar susceptibility, 395
M Molecular beam, 72, 282
Magic angle, see Photoionization resonance spectroscopy, 482484
Magic angle spinning (MAS), see NMR M LLER -P LESSET perturbation theory, 498,
Magnet poles, 73 508
Magnetic dipole (M1) transition, 250254, Momentum conservation, 429, 430
484, 488, 589, 590 F EYMAN graphs, 326
Magnetic field of the electron cloud, 453, 457 relativistic, 11, 30
Magnetic moment, 251 Momentum eigenfunction, 102
and angular momentum, 70, 71 Momentum operator, 91, 92
atomic nuclei, 447, 450 M OSLEY diagram, 160,161
in a magnetic field, 71, 72, 294, 295 Na-like ions, 160
of the electron, 331 X-ray absorption edges, 523
precessing in a magnetic field, 386, 388 Multi-electron atom, 344, 495547
Magnetic resonance spectroscopy, 482491 H AMILTON operator, 496498
Magnetic susceptibility, 395
H ARTREE method, 500
diamagnetism, 397
self-consistent field method, 500, 501
paramagnetism, 396
with one valence electron, 144
Magnetization, 395
Multi-electron photoionization, 524530
Magneton, 71
Multi-photon ionization, 244, 265269
B OHR, 79, 129, 294
angular distribution of electrons, 266269
Magnetron frequency, 332
kinetic energy of the electrons, 265
Main group
saturation, 434436
periodic system of elements, 141
Mass absorption coefficient, 525 Multi-photon processes, 244250
X-ray, 524 Multiple beam interference, 40
Mass correction Multiplicity, 76, 300, 302, 357, 363
relativistic, 54 Multipole expansion, 613622
Mass polarization, 474 Multipole moment, 613622
Mass selection, 53 Multipole tensor operator, 616621
Mass spectrometer, 5456 general, 619621
double focussing, 55
quadrupole, 55 N
time of flight, 55 N atom, G ROTRIAN diagram, 515
Matrix eigenvalue equation, 164 Na atom, 149
Matrix element, 100, 116, 575, 592 electron density distribution, 150152
angular momentum components, 586 radial electron density, 151
operator, 97 radial wave function, 149
reduction, 582587 Natural lifetime, 229
spherical harmonics Natural linewidth, 227229
LS-coupling, 583, 585 Natural unit of energy, 67
Matrix representation, 116 Ne atom, G ROTRIAN diagram, 517
Matter wave, 8794 Neon shell, 151
plane, 58 Neutron, 48
M AXWELLs equations, 92 Neutron diffraction, 59
M AXWELL -B OLTZMANN NIST data bank, 50, 118, 140, 146, 147, 157,
velocity distribution, 21, 72 159, 160, 166, 203, 259, 260, 309,
319, 342, 347, 349, 351, 371, 373, Nuclear quadrupole moment, 447, 449, 616
431, 514, 522, 525, 529, 551, 629 oblate or prolate, 478
NMR spectroscopy, 487491 Nuclear radius, 29, 121
apparatus, 488 liquid drop model, 476
CW spectrum of ethanol, 488 Nuclear spin, 449
magic angle spinning (MAS), 490 eigenvalue equations, 449
occupation probability of levels, 489 Nuclear spin resonance, see NMR
N OBEL prize in chemistry Nucleons, 48
Richard R. E RNST (1991), 490
KOHN and P OPLE (1998), 510 O
F ENN, TANAKA, W THRICH (2002), 490 O atom, G ROTRIAN diagram, 516
N OBEL prize in physics Oblate, 477, 478, 619
Wilhelm C. RNTGEN (1901), 530 nuclear shape, 478
L ORENTZ and Z EEMAN (1902), 377 Observable, 9799, 118
Joseph J. T HOMSON (1906), 50 commuting, 101
Albert A. M ICHELSON (1907), 278 non-commuting, 101
Max K. E. L. P LANCK (1918), 31 simultaneous measurement, 100, 101
Johannes S TARK (1919), 399 One electron cyclotron oscillator, 334
Albert E INSTEIN (1921), 27, 31 One particle problem, 117134
Niels H. D. B OHR (1922), 64 One sided exponential distribution, 653
C ORNELL, K ETTERLE, W IEMAN (1925), One-loop QED effects, 323, 324
25 Operator, 96, 97, 100
Arthur H. C OMPTON (1927), 28 energy, 101
Louis DE B ROGLIE (1929), 57 momentum, 101
Werner K. H EISENBERG (1932), 100 position in space, 101
S CHRDINGER and D IRAC (1933), 90 simultaneous measurement, 101
Otto S TERN (1943), 70 Optical path difference, 40
Isidor I. R ABI (1944), 482 Orbital angular momentum, 71, 295, 299,
Wolfgang PAULI (1945), 22, 138 302
Max B ORN (1954), 88, 89 components, 110, 111
L AMB and K USCH (1955), 317 eigenfunctions, 109113
T OMONAGA, S CHWINGER, F EYNMAN square, 111, 112
(1965), 79, 324, 534 Orbital energies, 509
R AMSEY, D EHMELT, PAUL (1989), 55, Orders of magnitude, 59
225, 288, 332, 482 energy scales, 7
G LAUBER, H ALL, H NSCH (2005), 248, length scales, 5
319 time scales, 7
M ATHER, S MOOT (2006), 6, 7 Orthonormality relation, 96
E NGLERT and H IGGS (2013), 46 Oscillator strength, 238, 239, 636640
N OBEL prize in physiology or medicine density, 256
L AUTERBUR, M ANSFIELD (2003), 490 sum rule, 239, 639, 640
Non-crossing rule, 391, 394
Non-local potential, 507 P
Non-stationary problems Pair production, 526
dipole excitation (E1), 169225 Paramagnetism, 394396
light matter interaction, 227270 Parity, 593, 594
Non-stationary states, 186, 187 conservation in E1 transitions, 202
normal Z EEMAN effect, 128131, 382386 multi-electron systems, 594603
Nuclear gN factor, 447, 449 Parity violation, 249
Nuclear magnetic moment, 447, 449, 485, 487, Particle detection, 75
488 Particle diffraction, 5861
Nuclear magnetic resonance, see NMR C60 , 60, 61
Nuclear mass, 119 He atoms, 59, 60
energy correction, 68, 69 Particles and waves, 57, 64
Partition function, 21 P LANCHERELs theorem, 644

PASCHEN -BACK effect, see Z EEMAN effect, P LANCK constant, 4, 33
high field P LANCK energy, 9
PAUL trap, 332 P LANCK length, 5
PAULI spin matrices, 116 P LANCK time, 7
PAULI principle, 22, 138, 139, 351355, P LANCKs radiation law, 3134
503506 E INSTEINs derivation, 185
P ENNING trap, 332 Plane wave, 94
Periodic system, 137144, 168 partial wave expansion, 661663
table of elements, 140144 Plasma frequency, 56, 57
Perturbation hierarchy Plasma oscillations, 56
with electric field, 402 Plasmon resonances, 57
Perturbation theory, 129 Pointing vector, 632
1st order, 162, 163 Polar coordinates, 107110
2nd order, 163, 164 H atom, 117
alkali atoms, 165 Polarizability, 144, 150, 411413
degenerate states, 164, 165 Polarization
stationary, 161167 circular, 172, 173
time dependent, 186196 dielectric, 411413
1st order, 190 induced, 412
Phase diagram, 88 orientation, 412
Phase difference elliptical, 172174
FPI interferometer, 280
linear, 171, 173, 174
Phase index, see Index of refraction
vector, 170176
Phase shift
basis, 171174
in QDT, 159
Polarization ellipse, 174
Phase space, 88
Polarization potential, 416
Phase velocity, 422424
Ponderomotive potential, 432434, 634,
Photo-absorption cross section, 42
635
aluminum, 525
Positron emission tomography (PET), 30
lead, 525
Photo-detachment, 255 Potential box
angular distribution of electrons, 264 one dimensional, 103, 104
Photoelectric effect, 2628 three dimensional, 104107
Photoelectron spectroscopy (PES), 28, 254, Potential well model, 27
268, 269 Power broadening, 230
imaging spectrometers (EIS), 268 Precession of angular momentum in a
Photographic plate, 59, 73, 74 magnetic field, 72
Photoionization, 254269 Principle quantum number, 67
angular dependence, 260, 261 Probability amplitude, 8790
anisotropy parameter, 260, 264 dependence on time, 92
Ar atom, 527 matter waves, 89, 90
B ORN approximation, 256260 photon, 88, 89
cross section, 255258 time dependent, 187
energy dependence, 259, 527530 Probability distribution, 61, 89, 151
magic angle, 261 energy, 20
theory and experiment, 261264 position, 123
with X-ray, 520, 524, 530 Probability interpretation, 89
Photometry, 3740 Product ansatz, 93, 345
Photon, 4, 2643, 62, 8890, 92 Projection theorem for angular momenta, 578
angular momentum, 30, 175198 Prolate, 477, 478, 619
elastic scattering, 527 nuclear shape, 478
flux, 179, 245 Proton, 48
momentum, 29 Proton radius, 7, 29, 324, 329
Q Rare gas, 142, 517

Quadrupole coupling constant, 481 no anions, 142
Quadrupole field, 55, 332 radii, 144
Quadrupole moment, 617619 Rare gas configuration, 144
electric, 253 Rare gas shell, 139
intrinsic, 479 Rate equations, 184
spectroscopic, 479 R AYLEIGH criterium, 276
Quadrupole tensor R AYLEIGH SCATTERING, 428, 429
atomic nucleus, 478 Real solid harmonics
electric, 253 renormalized, 597
Quantization, 25 Real spherical harmonics, 596
Quantization of the electromagnetic field, 325 renormalized, 596
Quantum beats, 220224 Reciprocal lattice vector, 40
Quantum defect, 146148 Recollision, 438, 439
fine structure, 308, 309 Reduced mass, 119
He atom, 343 Reduced matrix element, 577
theory, 152159 Reflection operator, 597, 617
Quantum electrodynamics (QED), 181, Reflection symmetry, 595603
324326 Relativity, see Special theory of relativity
Quantum jumps, 224, 225 Removal of degeneracy, 137144
Quantum mechanics Removal of m degeneracy, 130, 131
axioms, 9599 REMPI
definitions, 95104 H atom, 1S2S, 291
introduction, 87134 Resolving power
representations, 99, 100 FABRY-P ROT interferometer, 280
Quantum number, 127 interferometer, 277
angular momentum, 77, 112, 138 Resonance denominator, 164
good, 121, 299, 595, 597, 599 Rest mass, 10
in a box, 104 Rotation group, 575
principle, 122, 123, 138, 139, 147 irreducible representation, 560, 606
projection, 76, 138 Rotation matrix, 606
spin, 78, 114, 138 Rule of D UANE -H UNT, 531
spin projection, 78 RUNGE -K UTTA method, 150
Quantum state, 95, 96 RUSSEL -S AUNDERS coupling, 512
Quasi-one-electron system, 144161 RUTHERFORD, 65, 326
Quasi-two-electron system, 371374 RYDBERG, 2
Quiver motion atoms, 253
high, oscillating field, 433 atoms and diamagnetic interaction, 398
atoms in electric fields, 409411
R constant, 67, 69
Radial electron density constant, precision measurement, 321
computed with DFT, 511 states, 529
Radial matrix elements, 590, 592 RYDBERG -R ITZ formula, 69
Radial wave function, 166
Radian, 556 S
Radiance, 533 Saturation broadening, 230
spectral, 533 Saturation spectroscopy, see Spectroscopy,
Radiant flux, 34 D OPPLER free
Radiation Scalar product
spectral density, 33 of states, 95
spectral distribution, 31, 34 of tensor operators, 578
Radio frequency spectrum Scattering cross section, 326
lithium iodide, 483 S CHRDINGER equation, 9092
R AMSEY fringes, 288, 289, 320, 484 alkali atoms, 149
H atom, 117 general concepts, 177, 178

stationary, 91 high resolution, 274293
time dependent, 9294 Spectrum
S CHRDINGER representation, 100, 107 He atom, 342
Screening Hg atom, 372
He atom, 343 visible, 34
of nuclear charge, 138 Spherical harmonics, 112, 113
Screening parameter, 153 matrixelements, 580582
alkali like atoms, 159 products, 579
Na atom, 159 Spin
sech2 function, 651, 652 angular momentum, 99
convolution, 627 components, 115
Selection rules, 194 function, 301
Self adjoint operator, 97, see also Hermitian projection, 99, 140
operator Spin orientation and exchange interaction, 358
Self consistent field method, 498 Spin-orbit
Self-energy, 323 coupling, 568
S ELLMEIER equation, 420 coupling parameter, 296, 307, 378
Separation ansatz, 117 interaction, 293303, 513
Shell closure, 142 splitting
Shell structure of atoms, 137144 D IRAC theory, 306
Sinc function, 652 Spin-orbital, 503508
Singlet function, 353 Spinor equation, 93
Singlet states, 301 Splitting of energy levels
Singlet system, 342 due to magnetic field, 130
Singlet transitions, 203 Spontaneous decay rate, 229
S LATER determinant, 358, 503508 Spontaneous emission
Slow light, 422427 and QED, 186
Solid harmonics, 616 E INSTEIN A coefficient, 183
Space quantization, 7378 frequency dependence, 185
Special theory of relativity, 1014 introduction, 181183
and fine structure, 69 Spontaneous transition probability, 213, 214
L ORENTZ contraction, 13 Standard deviation, 15, 627
rest energy, 58 B OLTZMANN distribution, 235
rest frame, 13 Standard model, 4
time dilation, 13 of elementary particle physics, 4648
twin paradox, 13 Standard phase convention, 563
Spectral brilliance, 534 S TARK states, 408
of various X-ray sources, 534 S TARK effect, 399411
Spectral intensity distribution, 180 dipole states, 408
Spectral radiance, 34 dynamic, 418, 420
Spectral radiation density, 180 H(2s, 2p) states, 408
Spectrometer interaction potential, 400, 401
echelle, 277 linear, 407, 408
Spectroscopy matrix elements, 402
absorption, 177 high field, 403, 404
D OPPLER free low field, 404
ion beams, 283, 284 perturbation series, 405
microwave and RF transitions, 317 quadratic, 405, 406
molecular beams, 282, 283 RYDBERG atoms, 409411
saturation, 285288 significance, 399, 400
two-photon, 289291 State of a quantum system, 95, 96
emission, 177 State vector, 95, 129
fluorescence spectroscopy, 177 States of a quantum system, 95
Stationary states, 176 Trajectory

Statistics classical, 87
classical, 2026 Transition amplitude
elementary, 1426 perturbation ansatz, 187, 189
quantum, 2026 spherical basis, 198200
S TEFAN -B OLTZMANN Transition matrix element
constant, 34 radial, 312
law, 34 Transition operator
Steradian, 554, 556 dipole approximation, 190, 635, 636
S TERN -G ERLACH experiment, 7078, 99, Transition rate
114, 115 absorption, 191
interpretation, 75, 76 Transition rates in the continuum, 238
setup, 72, 73 Transmission grating, 60
Stimulated emission Trembling motion of an electron (so called
introduction, 180, 181 Zitterbewegung), 327
Strength of dipole transitions, 212217 Triangular relation, 197, 201, 298, 565, 577
Strong force, 59 Triplet functions, 353
Structure analysis, 4043 Triplet states, 301
Subgroup Triplet system, 342
periodic system of elements, 141 Tunnel ionization, 436, 437
Subshell, 142 Tunnelling effect, 75
Superluminal light propagation, 424427 21 cm line, 460
Superposition principle, 89, 92 Two electron system, 341375
Susceptibility, 413, 414, 420422 Hamiltonian, 344, 345
Synchrotron radiation, 42, 53, 531539 probabilities, 345
angle and energy dependence, 539 quantum mechanics, 344351
critical wavelength, 538 Two level system
generation schematically, 537 thermodynamic equilibrium, 184
Two particle
T wave function, 345, 346
Tensor operator, 575578 Two particle problem, 119
irreducible representation, 560 Two sided exponential distribution, 654
products, 578582 Two wire field, 73
real, 595 Two-photon emission, 248, 249, 317
Term levels, see G ROTRIAN diagram Two-photon excitation, 245248, 289, 320
Term scheme, see G ROTRIAN diagram H atom, 291
Terminology
of atomic structure, 301 U
Theory of special relativity Ultrafast physics, 8
rest energy, 67 Uncertainty relation
time dilation, 13 Gedanken-experiment, 61
T HOMAS -F ERMI equation, 502 H EISENBERG, 61, 63
T HOMAS -F ERMI potential, 501, 502 Undulator, 532
T HOMSON cross section, 527 Undulator and wiggler, 540542
T HOMSON parabolas, 56 Unit operator
T HOMSON scattering, 429 quantum mechanics, 98, 99, 577, 583
nonlinear, relativistic, 543 Unit vector of polarization, 170
Time dependent density functional theory,
511 V
Time-bandwidth product, 651 Vacuum field, 181
Total angular momentum interaction of an electron with the,
eigenstates, 299301 324
of the atomic charge could, 449 Vacuum polarization, 323
Total wave function, 138, 139 Vacuum-ultraviolet, 31
Valence electron, 75, 144, 146, 148 bremsstrahlung, 530

Na atom, 151 characteristic emission line, 521
VAN DER WAALS diffraction, 4043
potential, 417, 418 sources, 530544
radius, 142, 143 spectroscopy, 519524
Variance, 15, 235, 626 AUGER electron, 522
Variational method, 350, 351 characteristic lines, 522524
Vector boson, 44 M OSLEY formula, 522524
Vector diagram, 78, 112, 129 X-ray tube, 530, 531
Vector model, 298, 381, 385
Vector operator, 91, 114, 576, 577 Y
Vector potential, 396, 631634 YOUNGs double slit experiment, 88
Velocity distribution, 285
Virial theorem, 66 Z
VIS, visible spectral range, 31 Z EEMAN effect, 9
VOIGT profile, 236, 237
normal, 128131
convolution, 628, 629
anomalous, 78
Volume term in HFS, 473
fine structure, 377399
normal, 382386
W
anomalous, 379
Wave function, 8896, 100, 104
symmetric and antisymmetric, 353 avoided crossings, 391
two particles, 345, 346 classical triplet in a high field, 386
Wave nature of matter, 5863 examples, 382384
Wave vector, 4, 13, 57, 94 high B field, 384386
two-photon excitation, 289 interaction Hamiltonian, 377380
Wave-packets, 88 intermediate field, 388392
Wave-particle duality, 58, 61 limiting cases, 379, 392
Wavenumber, 31, 67 line strengths, 384
SI units, 31 low B field, 380384
W EHNELT cylinder, 530 selection rule for transitions, 386
W IEN filter, 56 hyperfine structure, 461471
W IENs displacement law, 34 B REIT-R ABI formula, 467471
Wiggler, 532, 541, 542 ground state of 6 Li, 470
W IGNER -E CKART theorem, 576578, 580, high B field, 464466
582, 586, 617, 620, 621, 637 low B field, 462, 464
W IGNER -S EITZ radius, 143, 144 Na D lines, 464
transition to very high fields, 469
X Zero point energy, 181
X-ray Zero range potential, 436
absorption edge, 520, 521, 523 Zitterbewegung (trembling motion), 327
Index of Volume 2
Symbols Benzene
12, 6 potential, 149 D6h point group, 262
H CKEL orbitals, 277285
A (anisotropy) parameter, 356, 360, 551, 557
A BBE sine law, 679 B ETHE formula, 528530, 534, 537, 538, 546
Adiabatic representation, 478480 ionization, 537
A IRY diffraction pattern, 25 B ETHE integral, 526
A IRY disc, 26 B ETHE ridge, 546, 548, 549, 553
Alignment, 596, 604, 616, 619, 698700 B ETHE surface, 545
angle, 36, 44, 605, 610, 611 Binary peak in (e, 2e) process, 553
parameter, 622, 623, 698 Birefringence, 38, 40, 41
Alkali halide potentials, 220224 B IRGE -S PONER plot, 177
Ammonia maser, 1, 252 Black body radiation, 105
Amplification profile, 15 B OLTZMANN
Amplified spontaneous emission, 11 distribution, 155, 173, 335, 389, 591
Anharmonicity constant, 162 statistics, 158
Anti-S TOKES lines, 334 Bond order, 195
Anti-symmetrization, 521 Bonding orbital, 183
Antibonding orbital, 183 B ORN approximation
Ar2 TPES spectrum, 369 first order, elastic, 444448
Arn clusters, 400 generalized oscillator strength (GOS),
Atomic form factor 530534
inelastic, 527 inelastic collisions, 460, 461, 522, 525530
AUGER electron, 533 integral inelastic cross sections, 534
AUTLER -T OWNES effect, 637639 B ORN phase shift, 447
Autocorrelation function, 5360 B ORN series, 553
Avoided crossing, 220224, 263 B ORN -O PPENHEIMER approximation,
139151
B collision processes, 458, 474
Baseline in interferometry, 87, 91 BPP, see Beam parameter product
Beam divergence, 21, 22 Branching ratio, 653
Beam parameter, 1721
product, 26, 34, 35 C
Beam radius, 1735, 650 C6 H +
4 absorption spectra, 328
Beam waist, 1721 C2 DFWN spectrum, 354
and coherence volume, 87 Carrier envelope, 46
and lateral coherence radius, 91 phase, 46, 55, 188

Carrier frequency, 45 Collision channels, 474

C ASIMIR operator, 541 Collision frame, 421, 422, 473
C AUCHY-S CHWARZ inequality, 83 Collision process
Cavity quantum electrodynamics, 127130 elastic, 383451
Cavity ring down spectroscopy, 327, 328 highly charged ions, 499506
CCD camera, 315, 508, 676, 701 and ultrafast dynamics, 506
Centre of mass system, 389, 395, 396 inelastic, 453513
Ceratron, 675 introduction, 383393
CH4 , electronic states, 273 kinematics, 396400
Channeltron, 674 COLTRIMS, 414, 501, 504, 685
Character tables, 257262 Complex beam parameter
C2v , 258, 259 Gaussian beams, 19
Cs , 259 Confocal parameter in Gaussian beam, 21, 30
D2h , 260 Conical intersection, 263, 312, 323, 332, 507
D3h , 322 Conical skimmer, 321
D6h , 278, 279 Convergent close-coupling, 460, 464
Oh , 260262 Correlation diagram, 192194, 217
Td , 273 Correlation function, 5260
Characteristic equation, 180 1st order, 72, 74
Charge exchange higher order, 5558
H+2 system, 184190 interferometric measurement, 5458
H+ + H, 185188 examples, see also Convolution
highly charged ions, 504, 505 C OULOMB integral, 182
over-the-barrier model, 501503 Coupling elements
Chemical shift, 358, 360
non-adiabatic, 486, 487
Chemical-potential, 106
Cross section
Classical rainbow, 403
differential, 393402
Classical trajectory, 402
integral, 394
Classical turning point, 406
momentum transfer, 395
Close-coupling
total, 385387
convergent calculations (CCC), 522, 524,
Cusp in the e + He excitation, 464
529, 538, 539, 547, 548, 551, 553
Cytosine
theory, 460, 461, 516519, 521, 523
Clusters photoelectron spectroscopy, 360
mass selection, 400402
supersonic molecular beams, 321 D
CO D E B ROGLIE
nuclear spin statistics, 156, 157 wavelength, 402
CO2 Decay of coherence (T2 ), 641
infrared spectrum, 245 Decay of excitation (T1 ), 641
laser, 245 Degenerate four wave mixing (DFWM), 351
normal modes, 244, 245 Degree of coherence, 601, 602
Coherence, 72100 1st order, 72
and incoherence, 580 2nd order, 84
angle, 90 N th order, 82
area, 91 Degree of polarization, 601, 602
length, 76, 7882 coherence, 585588
spatial, 8691 linear, 43
time, 43, 76, 77, 79, 82, 84, 598 Degree of temporal coherence
volume, 91 2nd order, 82
Coherent anti-S TOKES R AMAN spectroscopy Delay line, 52
(CARS), 351 Density functional theory, 211
Coherent S TOKES R AMAN spectroscopy Density matrix, 573624
(CSRS), 351 1 P state, 602611
optical excitation, 609623 of Hg, 466468

transformation, 585 of molecules, 468, 469
Density of states of rare gases, 465, 466
photons, 105 Electron impact ionization, 534562
Density operator, 581, 582 at threshold, 540544
Depletion spectroscopy, 321 double-differential cross section, 544549
Detailed balance, 387389 GOSD, 545
Diabatic representation, 480483 integral cross section, 537, 538
Diatomic molecules, 135228, see also L OTZ formula, 537
Molecules, diatomic, 229 single-differential cross section, 539, 540
Dielectronic recombination, 563, 564 triple-differential cross section, 549558
Differential cross section, 393402 WANNIER threshold law, 542544
Diffraction Electron jump
F RAUNHOFER, 2326, 89, 413, 414 in ion molecule reactions, 221
F RESNEL factor, 25 Electron momentum spectroscopy, EMS,
F RESNEL number, 8, 20, 25 558561
H UYGEN -F RESNEL principle, 23 Electron photo-detachment, 364
Dipole moment Electron scattering theory, 515525
of a diatomic molecule, 167 Electronic spectra of molecules, 305317
Dipole transition, 166178 classical spectroscopy, 314317
matrix element, 305, 480 laser induced fluorescence (LIF), 317320
diatomic molecule, 167 laser spectroscopy
polyatomic molecules, 242 biomolecules, 328333
operator, 119, 612 REMPI laser spectroscopy, 320327
rotational transitions, 312, 313
Direct excitation process, 482
selection rules, 309311
Directional intensity, 680
Electronic states
Dissociative ionization, 363, 366, 558
NH3 , 274, 275
Distorted wave approximation, 547, 553
of conjugated organic molecules, 277285
Double resonance spectroscopy, 295, 330
triatomic molecules, 266277
Dressed states, 629, 630
Ellipsoid of inertia, 232
coupled systems, 633635
Ellipticity angle, 36, 44, 597
three, 659
Emission
D UNHAM coefficients, 165, 166 in a narrow band laser field, 635639
DYSON orbital, 559 Energy defect
charge exchange, 501
E Energy imaging, see Velocity map imaging
e rare gases Energy selector
integral elastic cross section, 392 hemispherical, 680683
(e, 2e) process, 534 magnetic bottle, 683685
Effective potential, 152, 163, 406, 413, time of flight method, 683685
453455 Entanglement, 575, 608
Effective range expansion, 430 Ergodicity, 75, 598
Eikonal approximation, 425 ESCA, 355383
Elastic scattering, 383451 Ethylenfluoroacetat, 361, 362
classical theory, 402418 Exchange
Electric field and photon number, 116, 117 amplitude, 519
Electron and ion optics, 673685 cross section, 519
Electron configuration, 141, 195, 205, 217, integral, 182
226, 259, 261, 263, 264, 267 symmetry, 156
Electron impact excitation electrons, 203
GOS (for e Na), 533, 534 Excitation
of He, 461463 continuous, with relaxation, 648, 649
fine details, 464 continuous, without relaxation, 647, 648
narrow band vs. broad band, 645, 646 R AYLEIGH length, 19

with short laser pulses, 649652 Gaussian profile, 20, 59, 83, 644, 651
Excitation function, 459, 460472 Generalized oscillator strength
density (GOSD), 545549
F (GOS), 530534
FABRY-P ROT resonator, 46, 50 Genetic algorithm, 332
FANO lineshape, 441 G LAUBER approximation, 553
FANO -M ACEK theory, 611623 G LAUBER states, 114117, 596
Fast and slow axis, 38 Glory oscillations, 391, 412
Femtochemistry, 507 Glow bar, 297
Femtosecond spectroscopy, 224 G OUY phase, 20
F ERMI resonance, 244
F ESHBACH resonance, 437 H
Field envelope H + He, integral elastic cross section, 390
phase of, see Carrier envelope phase H2
Field quantization and transitions, 110131 MO ansatz, 206210
Fine structure nuclear spin statistics, 156, 157
in H UNDs case (a), 200 ortho and para hydrogen, 156
Fluence, 49 potentials, 208, 209
FORT trap, 626 reflection symmetry, 203, 204
F ORTRAT diagram, 314 valence bond theory, 210, 211
F OURIER transform H+2 MOs, 181184
limited pulses, 46, 47 H2 O
spectroscopy, 170, 293 absorption spectrum, 269
F OURIER transform spectroscopy, 298302, C2v group, 258
343 electronic states, 266270
of H2 O in the visible, 247 orbitals (EMS), 561
Fragment spectroscopy photoelectron spectroscopy, 358360
KETOF, 364 rotational levels, 237, 238
MATI, 364 structure, 237
F RANCK -C ONDON vibrational spectrum, 245247
factor, 305312 H2 , predissociation, 207
principle, 306309 Half wave plate, 40
F RANCK -H ERTZ experiment, 467 H AMILTON operator
F RAUNHOFER diffraction, see Diffraction diatomic molecules, 137, 138
Frequency comb, 4952 heavy particle collision, 396
F RESNEL rhomb, 40 relative motion, 397
Fringe spatial frequency Hanbury B ROWN -T WISS, 72
stellar interferometer, 91 experiment, 8486
Frozen-core approximation, 560 Harmonic oscillator, 143145
FTIR, see F OURIER transform spectroscopy He-He potential, 149
Helicity basis, 36, 37, 42, 596, 597, 602
G polarization vectors, 101
Gain narrowing, 11 H ELMHOLTZ -L AGRANGE relation, 679
Gas kinetic cross section, 389 Hemispherical capacitor, 680683
G AUSS radius, 22 H ERMITE functions, 144
Gaussian beam, 1735 H ERMITE polynomials, 144
beam waist, 1932 Hessian matrix, 240
complex beam parameter, 19 Hindered pseudorotation, 324
divergence angle, 22 Hole burning, 1416
intensity and power, 2123 with IR and UV, 330
nonlinear processes in, 6167 HOMO, 195
radius of curvature of the wave front, 19, HOOO
21, 30, 31 IR action spectrum, 302
microwave spectrum, 295 J EFFREYS -B ORN phase shift, 426, 448

H CKEL method, 279285 JWKB phase shift, 426, 427, 494
H UNDs coupling cases, 199201
Hybrid orbitals, 219 K
double bonding, 275, 276 K type doubling, 237
LiH, 219 K-matrix, 475, 521, 522
bonding, 273, 274
sp 3 orbitals, 270272 L
triple bonding, 276, 277 Laboratory frame, 396
Hyperspherical coordinates, 540542 Lambda-doubling, 205, 206
Lambda-half plate, 40
I Lambda-quarter plate, 3840
I2 studied by LIF, 318, 319 L AMBERT-B EER law, 386
Imaging methods, 357, 365, 367, 369, 370 L ANDAU -Z ENER formula, 489492
Impact parameter, 402, 404, 405, 408412, L ANGEVIN cross section, 454457
421, 422 Laser, 117
Incoherence by collisional excitation, 606609 amplifier medium, 911
Index of refraction basic principle, 3, 4
particle beams, 678680 diffraction losses, 8
Inertia tensor, 232 history, 1, 2
Infrared spectroscopy, 296305 longitudinal modes, 5, 16
action spectroscopy (IAS), 302304 population inversion, 1416
Integral cross section, 408, 409 rate equations, 1214
Intensity correlation function, 626 stability diagram, 7
Intensity interferometry, 98 threshold condition, 11, 12
Interaction experiment transverse modes, 68
state selective, 590596 Laser beam
Intercombination lines, 309 diameter, 22
Interference experiment, 78, 79, 87 M2 factor, 34, 35
spatial, 87 profile measurement, 34
YOUNGs, 78 Laser pulse, 4548
Interferometer frequency spectrum, 4548
M ACH -Z EHNDER, 52 Gaussian temporal profile, 46
M ICHELSON, 52, 59, 78 highest intensities, 3
stellar, M ICHELSON, 9195 measurement of ultrashort, 5260
Internal conversion, 312 mode coupled, 50
Inversion frequency, 250 spatial and temporal profile, 49
Inversion symmetry, 183, 202, 203 time dependence of sech2 , 46
Inversion vibration ultrafast, 59
in NH3 , 247252 Laser spectroscopy, 317334
p toluidine, 294 LCAO, 179, 180
Ion imaging, 66, 67 L ENNARD -J ONES potential, 149
Ion velocity map, see Velocity map imaging L EVINSON theorem, 429, 432
Irreducible representation Li3
density matrix, 614, 687, 693 high resolution spectroscopy, 321327
dipole operator, 311 potential surface, 323
of point groups, 257 Lifetime
excited molecular states, 306, 307, 312
J H2 O, 270
JABLONSKY diagram, 335 internal conversion, 656
JAHN -T ELLER natural, 631, 640, 644
effect, 237, 262265, 321323 photons in a resonator, 46, 8, 12, 14, 327,
theorem, 262 328
vibronic coupling, 265 rate equations, 653
resonance scattering, 523 diatomic, heteronuclear, 215218

RYDBERG state, 563, 564 diatomic, homonuclear, 179197
Light beam, 17 Molecular potential, 140
LiH potentials, 218220 Molecular spectroscopy, 289381
Line broadening Molecules
saturation, 643645 diatomic, 135229
Linewidth electron spin, 197, 198
homogeneous and inhomogeneous, 1416 electronic energy, 138
natural, 9, 127130 equilibrium distance, 136
L IOUVILLE equation, 639, 640 heteronuclear, 215226
L IPPMANN -S CHWINGER equation, 418420 rotational energy, 139
Lone pair (of electrons), 223, 267 total angular momentum states,
ammonia, 274 197214
Longitudinal coherence, 80 total energy diagram, 291
L ORENTZ profile, 4, 911, 76, 77, 81, 91, 107, vibration, 160163
127, 300, 301, 643, 645 vibrational and rotational constants, 154
resonance scattering, 439 polyatomic, 231288
L OTZ formula RYDBERG states, 224
electron impact ionization, 537, 538 valence states, 224
LUMO, 195 M OLLOW triplet, 635637
Momentum conservation, 554, 558
M Momentum imaging methods, see Velocity
Magic angle, 357 map imaging
Main axis of symmetry, 254 M ORSE potential, 145147, 163
MALDI, 330 MOs, 179, 180, see Molecular orbitals
M ALUSs law, 44 MOTRIMS, 508, 685
Maser Multi-mode states, 117, 118
microwave amplification by stimulated Multi-photon ionization, 67
emission of radiation, 1 Multiplicity, 197, 205, 214
NH3 , 252 Multipole moment, 690692, 698700
M ASSEY criterium, 457, 459, 479, 481, 488, Multipole tensor operator, 690692
491
modifed, 461, 491 N
M ASSEY parameter, see M ASSEY criterium N2
Matrix isolation spectroscopy, 333 nuclear spin statistics for bosons, 347
Mean free path length, 386 potentials, 211, 212
Measurement R AMAN spectrum, 343, 344
state analyzer, 588590 N2 + e shape resonance, 524
state selector, 588590 Na atom
theory of, 588596 hyperfine transition, 695697
Merged-beams experiment, 564, 565 Na+ + Na(3p), inelastic and super-elastic
M ICHELSON interferometer, 298 processes, 492496
Micro reversibility, 388 Na + Hg
Microchannel plate, 675 integral elastic cross section, 391
Microwave spectroscopy, 292296 Na+2 potentials, 493
Mode density, 105 Na3 REMPI spectroscopy, 321
Mode locking, 50 NaCl potentials, 221, 222
passive, 328 NaI potentials, 222224
Mode synchronization, 50 Natural lifetime, 127
Modes, see Laser N EWTON diagram, 397, 398, 400, 401
Modes of the radiation field, 103 NH3 , umbrella mode, 248252
Molecular beam (NH3 )n , FEICO spectra, 371
seeded, 321 NIST data bank, 147, 162, 168170
Molecular orbitals NO potentials, 224226
N OBEL prize in physics Optical-optical double resonance

F RANCK and H ERTZ (1925), 467 in Li3 , 325
Chandrasekhara V. R AMAN (1930), 334 Orientation, 616, 619, 698700
Isidor I. R ABI (1944), 630 Orientation parameter, 698
Robert H OFSTDTER (1961), 385 Oscillations
B LOEMBERGEN, S HAWLOW, S IEGBAHN glory, 391
(1981), 2, 317, 355 rainbow, 411
R AMSEY, D EHMELT, PAUL (1989), 695 shadow scattering, 412415
G LAUBER, H ALL, H NSCH (2005), 49, S TCKELBERG, 494, 496498
52, 72 symmetry, 415417
H AROCHE and W INELAND (2012), 128 Oscillator strength
N OBEL prize in chemistry density, 545
Robert S. M ULLIKAN (1966), 258 Overlap integral, 181
Gerhard H ERZBERG (1971), 314
H ERSCHBACH, L EE, P OLANYI (1986), P
221, 507 P toluidine, microwave spectrum, 294
C URL, K ROTO, S MALLEY (1996), 242 Partial wave analysis, 391, 392, 418, 427, 428,
Ahmed H. Z EWAIL (1999), 224, 507 433, 434, 475, 476, 484, 516, 517,
F ENN, TANAKA, W THRICH (2002), 330 523, 592
Non-adiabatic coupling, 477 e He and e Ne, 431, 432
Non-crossing rule, 194 Partial wave expansion, 422425, see Partial
Nonlinear spectroscopy, 348355 wave analysis
basics, 349353 Partial waves, 423
BOXCARS, 353 Partition function, 156, 173
four wave mixing processes, 352 PAUL trap, 333
Normal modes, 239243 P ENNING trap, 333
asymmetric stretch, 243 P ERCIVAL -S EATON hypothesis, 617
bending vibrations, 244 Periodic boundary conditions, 103
symmetric stretch, 243 Periodic system
transitions between, 242, 243 diatomic molecules, 194197
triatomic molecule, 245247 Phase matching, 352
Nuclear spin statistics, 157, 343348 Phase shift
population of rotational levels, 155157 scattering, 423
Nuclear wave function, 142, 143, 151161 Phosphorescence, 311
Number operator, 111 Photo-dissociation
of H+2 , 188190
O Photo-fragment spectroscopy, 333
O2 Photoelectron spectroscopy (PES), 355372
H ERZBERG bands, 315, 316 anions, clusters, 364366
nuclear statistics (bosons), 348 basics, 355358
paramagnetism, 214 PEPICO, TPEPICO, 366371
potentials, 211213 PFI, 363
R AMAN spectrum, 344, 345 TPES, 363
reflection symmetry, 204, 205 ZEKE, 363
O5+ + e Photoionization
dielectronic recombination, 563 anisotropy parameter, 356, 360
Oblate, 234236, 699 magic angle, 357
Odd and even molecular orbitals, 183 Photomultiplier, 674
Optical B LOCH equations, 625665 Photon
and short pulse spectroscopy, 649657 introduction, 100102
Optical multi channel analyzers, 316 modes of the radiation field, 102105
Optical pumping, 626, 695702 number per mode, 106108
with two frequencies, 700702 photon states, 100108, 110
Optical theorem, 425 Photon annihilation operator, 113
Photon bunching, 84 Pulse train, 50

Photon creation operator, 113 Pure state, 577
Photon number states, 110114 Pyrazine
Photon spin, 101 ZEKE and MATI spectra, 365
pulse, 648
P LANCKs radiation law, 106 Q
Plane wave impulse approximation, 559 Q factor of a resonator, 4
PN emission spectrum, 314 QED
P OCKELS cell, 41 cavity, 127130
P OISSON distribution, 116 Quality factor of a resonator, 4
Polarizability, 337 Quantum beats, 617
Polarization, 3545, 611 Quantum optics, 72134, 626
analyzer for linear, 44 Quantum system
circular, 38, 40, 44, 101, 155, 596, 597, 609 in electromagnetic field, 639642
degree of, 42, 43, 356, 580 temporal evolution, 639
density matrix for, 596602 Quarter wave plate, 3840
elliptical, 36 Quasi-monochromatic light, 43, 45, 7577
field induced, 349
incomplete, 4144 R
lambda-half plate, 40 R-matrix theory, 443, 464, 472, 522, 529
lambda-quarter plate, 3840 R ABI frequency, 630, 631
linear, 36, 38, 102, 154, 155, 189, 356, 495, non-resonant, 633
609, 695 resonant, 631
linear, elliptic, circular, 37 R ABI oscillation, 627, 648
measuring the degree of, 43, 44 Radial coupling
nonlinear, 350 collision induced transitions, 480
rotating the plane of linearly polarized Radiationless transitions, 311, 312
light, 40 Radiative corrections, 128
selection rule in electron impact excitation, Radiative recombination, 563
532 Rainbow
S TOKES parameter, 600 heavy particle collisions, 409417
S TOKES vector, 43, 600 optical, 403405
time dependence of intensity by, 3638 rapid oscillations, 411
Polyatomic molecules supernumerary, 410
vibration, 239253 R AMAN active transitions, 338, 341
Population inversion, 9, 14, 16, 252 R AMAN scattering
Potential differential cross section, 341
anharmonic, 147 graph, 339
diatomic molecules, 141 R AMAN spectroscopy, 334348
hypersurface, see Potential surface classic interpretation, 337, 338
L ENNARD -J ONES, 149 experimental aspects, 342, 343
surface, 140, 141, 262, 263, 265, 269, principle, 334337
506509 quantum mechanical theory, 338342
H atom as three body problem|(, 541, R AMSAUER minimum, 391, 431
542 Rate constant, 386
Mexican hat, 323 Rate equations, 386, 646, 647, 653657
VAN DER WAALS , 141, 148150 Ray tracing, 26
Potential hypersurface, see Potential surface Ray transfer matrix, 2629
Predissociation of H2 , 207 R AYLEIGH criterium
Principle moments of inertia, 232 resolution of optical instruments, 89
Prolate, 234236, 699 R AYLEIGH length, 19, 29, 30, 32, 63, 65
Pseudo-states, 522 R AYLEIGH line, 334
Pseudopotential, 437 Rb atom, hyperfine transition, 626
Pseudorotation, 321327 Reaction coordinate, 508, 509
Reaction microscope, 414, 508, 685 asymmetric (non-rigid), 295, 303

Reactions asymmetric (rigid), 236239
absorbing sphere, 455 spherical (rigid), 234
without threshold, 453, 454 symmetric top (rigid), 234236
Recoil peak in (e, 2e) process, 553 Ruby laser, 2
Recombination, 563566 RUTHERFORD
Reduced cross section, 407, 408 scattering, 407, 446, 447, 528, 529, 539,
Reduced scattering angle, 407, 408 559
Reflection symmetry, 203 RYDBERG
Refraction states in molecules, 224
of particle beams, 678 RYDBERG -K LEIN -R EES method, 177, 178
Relative velocity, 187, 386, 387, 390, 397, 398,
400, 409, 509 S
Resolving power S-matrix, 434, 475, 487, 494, 521, 522
FABRY-P ROT interferometer, 5 S-triazine, R AMAN spectrum, 345
R AYLEIGH criterium, 26 Saturation broadening, see Line width
Resonances, 436444 Saturation intensity, 643
autoionization, 436 Scattering, elastic, 383451
electron scattering Scattering amplitude, 410, 418420, 424, 426,
by molecules, 523525 433, 434, 488, 494, 576, 592, 606,
He , 441443 669
F ESHBACH, 412 direct, 519
formalism, 438443 first B ORN approximation, 525527
in electron scattering inelastic, 475, 476, 517, 518
H2 , 207 semiclassical, inelastic, 487, 488
N2 , 211 Scattering cross section
O2 , 214
differential
orbiting, 412 beam-gas experiment, 394
predissociation, 412, 436 crossed beam experiment, 395
shape, 412 elastic, 385
types, 436, 437 elastic, integral, 389392
Resonant capture, see Resonances, orbiting inelastic, 385
Resonator mode, 16 ionizing, 385
Resonator Q factor, see FABRY-P ROT reactive, 385
resonator total
Resonator quality factor, see FABRY-P ROT absorption experiment, 386
resonator Scattering kinematics, 396400
Resonator turnaround time, 6 Scattering length, 429431
Richtstrahlwert, 680 Scattering matrix, 432435
Room temperature, 148 Scattering phase, 418
Rotating wave approximation, 123, 631633 Scattering theory
Rotational constant, 155 classical, 405409
diatomic molecules, 153 multi channel problem, 472484
Rotational coupling phase shifts, 428435
collision induced transitions, 480 quantum mechanical, 418436
Rotational spectrum semiclassical, elastic, 425428
CO, 168 semiclassical, inelastic, 487489
Rotational temperature, 155 semiclassical approximation, 484499
Rotational transitions, 167170 S CHRDINGER equation
Rotor, diatomic molecules, 137
non-rigid, 163165 sech2 function, 47
rigid, 152157, 160162, 200 Secondary electron multiplier, 673678
Rotor, polyatomic, 231 Seeded molecular beam, 321
SF6 Sudden approximation, 560

Oh group, 261 Surface hopping, 312, 506509
Shape resonance, 437 SVE approximation, see Slowly varying
Short pulse generation, 4952 envelope approximation
light, see Polarization, circular Symmetry
Single atom in a MOT trap, 627 character tables, see Character tables
Single electron capture, 501 cylinder, 190
Slowly varying envelope approximation, 18, 42 g, u, 191
S OLEIL -BABINET compensator, 40 molecular physics, 253265
sp 2 orbital point groups, 254257
double bonding, 275, 276 reflection, 201205
Spatial coherence
degree of, 89 T
Spatial filter, 33 T-matrix, 420, 433, 434, 475, 476, 487, 517,
Specific heat capacity, 158, 160 519, 521, 522, 592, 594
Spectrum inelastic scattering, 517
electromagnetic radiation for molecular Telescope systems
spectroscopy, 290292 K EPLER and G ALILEI, 32, 33
Spontaneous line broadening, see Line Temporal coherence, 80
broadening, natural Tensor operator
Stability diagram statistical, 688
laser oscillation, 7 Tetrahedral angle, 272
Standard deviation Three body problem, 535, 541, 542, 555, 557
P OISSON distribution, 116 Threshold amplification, 14
S TARK effect, 171, 172 Threshold inversion, 14, 15
State multipole, 614, 687694 Threshold laws, 470, 472
State of a quantum system Tight binding method, 285
coherent and incoherent, 579 Time-bandwidth product, 48
pure and mixed, 575581 Toroidal energy analyzer, 556
State selection, 577 Transition
Stationary phase, 426428 field quantization, 110131
Statistical tensor, 687694 L ANDAU -Z ENER, 489492
operator, 688 non-adiabatic, 477, 478
Statistics perpendicular, 245
B OLTZMANN, 389 Triatomic molecules
exponential distribution, 76 linear, 243245
of coincidence methods, 560 nonlinear, 245247
Tunnelling
of measurement, 368
in predissociation, 207
quantum, 72
level splitting in NH3 , 247252
Stellar interferometer
Two level system, 119, 629635
Hanbury B ROWN -T WISS, 9598
excitation with CW light, 642649
Stimulated emission pumping, 658
STIRAP, 657665 U
energy splitting and evolution of states, Ultrafast laser pulses
659, 660 measurement, interferometric, 60
experiments, 661665 Ultrashort light pulse, see also Correlation
three level system, two laser fields, function
657659 Unimolecular dissociation, 321
S TOKES lines, 334
S TOKES parameters, 41, 42, 596, 598600, V
602 Valence states
experimental determination, 600, 601 in molecules, 224
S TOKES shift, 308 Van C ITTERT-Z ERNICKE theorem, 89
S TOKES vector, 42, 600 VAN DER WAALS
contact distance, 149 polyatomic molecules, 253

equation, 148 Vibronic coupling, 262, 265, 321327
potential, 148150, 409 Visibility, 80
radius, 150 VMI, see Velocity map imaging
Variational method, 179, 180
R ITZ, 208 W
Velocity WANNIER ridge, 542, 556
of electrons and atomic nuclei, 136 WANNIER threshold law, 470, 542544
Velocity map imaging (VMI), 66, 67, 188, 357, Wave equation, 17
384, 507, 508, 676685 general case, 350
Vertical binding energy, 358 Wave-packets, 4552, 82
Vertical ionization potential, 358 Whole burning spectroscopy, 330
Vibration-rotation spectra, 174177 W IGNER -E CKART theorem, 615, 671, 687,
CO, 174 688, 691, 699
PQR branches, 175 WKB phase shift, 426, 448
Vibrational quantum number, 152
Vibrational transitions Z
diatomic molecules, 172, 173 Zero point energy, 112, 145
in polyatomic molecules, 242 C60 , 242

Atomic Molecular Optical Physics by Hertel C Schulz, Volume 1

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Atomic Molecular Optical Physics by Hertel C Schulz, Volume 1

Caricato da

Copyright:

Formati disponibili

Graduate Texts in Physics

For further volumes:

Professor William T. Rhodes

Professor Susan Scott

Professor H. Eugene Stanley

Professor Martin Stutzmann

Atoms, Molecules and

ISSN 1868-4513 ISSN 1868-4521 (electronic)

Library of Congress Control Number: 2014952813

Springer-Verlag Berlin Heidelberg 2015

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

tain basic knowledge and understanding of atomistic concepts in physics, as well as

Acronyms and Terminology

AMO: Atomic, molecular and optical, physics.

1.6.3 Cyclotron Frequency and ICR Spectrometers . . . . . . . 54

2.4.2 Three Dimensional Potential Box . . . . . . . . . . . . . 104

3.3.4 Perturbation Theory for Degenerate States . . . . . . . . 164

5 Linewidths, Photoionization, and More . . . . . . . . . . . . . . . . 227

6.4 Selection Rules and Intensities of Transitions . . . . . . . . . . . 310

8 Atoms in External Fields . . . . . . . . . . . . . . . . . . . . . . . . 377

9.2.3 Nonvanishing Orbital Angular Momenta . . . . . . . . . 457

10.6.3 Synchrotron Radiation, Quantitative Relations . . . . . . 536

C.3.1 Matrix Elements of the Spherical Harmonics in LS

Appendix H Vector Potential, Dipole Approximation, Oscillator

1 Lasers, Light Beams and Light Pulses . . . . . . . . . . . . . . . . 1

1.6.2 Cylindrical Geometry (2D Geometry) . . . . . . . . . . . 63

3.2.5 M ORSE Potential . . . . . . . . . . . . . . . . . . . . . 145

4.1.3 Symmetric Rigid Rotor . . . . . . . . . . . . . . . . . . 234

5.6.3 Quantum Mechanical Theory . . . . . . . . . . . . . . . 338

Acronyms and Terminology . . . . . . . . . . . . . . . . . . . . . . . 448

8.2.4 An Example . . . . . . . . . . . . . . . . . . . . . . . . 529

10.2.4 The Coupled System . . . . . . . . . . . . . . . . . . . . 633

Appendix D Optical Pumping . . . . . . . . . . . . . . . . . . . . . . . 695

Ingolf V. Hertel was born in 1941 in Dresden, 1967

Claus-Peter Schulz was born in 1953 in Berlin, 1981

Here we provide a compact summary of the most important

1.1 Overview, History and Magnitudes

Springer-Verlag Berlin Heidelberg 2015 1

Table 1.1 Standard subject areas of physics

1.1.1 Quantum Nature of Matter

Based on experiments, many fundamental observations in physics can only be un-

photoelectric effect (E INSTEIN 1905),

We encounter the phenomenon of quantization that will be a steady companion

p = h/ and p = k with k = 2/, (1.1)

with the fundamental P LANCK constant (see also Sect. 1.4.6):

h = 6.62606896(33) 1034 J s = 4.13566733(10) 1015 eV s. (1.2)

energy of a photon (see Sect. 1.4.1):

energies of the H atom (see Sect. 1.8):

Wn = Eh /2n2 with n = 1, 2, 3, . . . . (1.4)

Equally important is the quantization of angular momentum and its direction in

absolute value of angular momentum:

angular momentum quantum numbers

with projection quantum numbers

1.1.2 Orders of Magnitude

Length Scales from Atomic Physics to Astrophysics

see enlarged scale

distance to the moon

distance to the sun

1.616252 1035 m, which may be viewed as describing the granularity of space

Fig. 1.2 Blow up of the size size in am

Time Scales from Atomic Physics to Astrophysics

Energy Scales in Physics

2/3 2 2 2/3 2/3