W Greiner

QUANTUM MECHANICS
SPECIAL CHAPTERS

Springer-V erlag Berlin Heidelberg GmbH

Greiner Greiner
Quantum Mechanics Mechanics I
An Introduction 3rd Edition (in preparation)

Greiner Greiner
Quantum Mechanics Mechanics II
Special Chapters (in preparation)

Greiner· Milller Greiner

Quantum Mechanics Electrodynamics
Symmetries 2nd Edition (in preparation)

Greiner Greiner· Neise . StOcker

Relativistic Quantum Mechanics Thermodynamics
Wave Equations 2nd Edition and Statistical Mechanics

Greiner· Reinhardt
Field Quantization

Greiner· Reinhardt
Quantum Electrodynamics
2nd Edition

Greiner· Schafer
Quantum Chromodynamics

Greiner· Maruhn
Nuclear Models

Greiner· Milller
Gauge Theory of Weak Interactions
2nd Edition
 Walter Greiner

QUANTUM
MECHANICS
SPECIAL CHAPTERS

With a Foreword by
D. A. Bromley

With 120 Figures,

75 Worked Examples and Problems

Springer
Professor Dr. Walter Greiner
Institut fiir Theoretische Physik der
Johann Wolfgang Goethe-Universităt Frankfurt
Postfach Il 19 32
D-60054 Frankfurt am Main
Germany
Street address:
Robert-Mayer-Strasse 8-10
D-60325 Frankfurt am Main
Germany
email: [email protected]

Title of the original German edition: Theoretische Physik, Ein Lehr- und Obungsbuch,
Band 4a: Quantentheorie, Spezielle Kapitel, 3. Aufl., © Verlag Ham Deutsch, Thun 1989

1st Edition 1998, 2nd Printing 2001

ISBN 978-3-540-60073-2
Library of Congress Cataloging-in-Publication Data.
Greiner, Walter, 1935 - [Quantenmechanik, English] Quantum mechanics. Special ehapters / Walter Greiner; with a
foreword by D. A. Bromley, p. cm. Includes bibliographical referenees and index
ISBN 978-3-540-60073-2 ISBN 978-3-642-58847-1 (eBook)
DOI 10.1007/978-3-642-58847-1
1. Quantum theory, 2. Electrodynamies, 3. Quantum field theory, 4. Mathematical physics. 1. Greiner, Walter, 1935 -
Theoretische Physik, English, Band 4a. Il. Title. QCI74.12.G74513 1998 530.12-dc21 97-24126

This work is subject ta copyright. AII rights are reserved, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction an
microfilm Of in any other way, and storage in data banks. Duplication of this publicatian or parts thereof is pennitted
only underthe provisions of the German Copyright Law of September 9, 1965, in its current vers ion, and permission
for use must always be obtained from Springer-Verlag. Viol.tions are liable for prosecution under the German
Copyright Law.

© Springer-Verlag Berlin Heidelberg 1998

Originally published by Springer-Verlag Berlin Heidelberg New York in 1998

The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the
absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and
therefore free for general use.

Typesetting: Data conversion by A. Leinz, Karlsruhe

Cover design: Design Concept, Emil Smejkal, Heidelberg
Copy Editor: V. Wicks
Productian Editor: P. Treiber
SPIN 10850520 56/3111 - 5 4 3 2 1 O - Printed on acid-free paper
Foreword to Ear Her Series Editions

More than a generation of German-speaking students around the world have

worked their way to an understanding and appreciation of the power and
beauty of modern theoretical physics - with mathematics, the most funda-
mental of sciences - using Walter Greiner's textbooks as their guide.
The idea of developing a coherent, complete presentation of an entire field
of science in a series of closely related textbooks is not a new one. Many
older physicists remember with real pleasure their sense of adventure and
discovery as they worked their ways through the classic series by Sommerfeld,
by Planck and by Landau and Lifshitz. From the students' viewpoint, there
are a great many obvious advantages to be gained through use of consistent
notation, logical ordering of topics and coherence of presentation; beyond this,
the complete coverage of the science provides a unique opportunity for the
author to convey his personal enthusiasm and love for his subject.
The present five-volume set, Theoretical Physics, is in fact only that part
of the complete set of textbooks developed by Greiner and his students that
presents the quantum theory. I have long urged him to make the remaining vol-
umes on classical mechanics and dynamics, on electromagnetism, on nuclear
and particle physics, and on special topics available to an English-speaking
audience as well, and we can hope for these companion volumes covering all
of theoretical physics some time in the future.
What makes Greiner's volumes of particular value to the student and
professor alike is their completeness. Greiner avoids the all too common "it
follows that ... " which conceals several pages of mathematical manipulation
and confounds the student. He does not hesitate to include experimental data
to illuminate or illustrate a theoretical point and these data, like the theoret-
ical content, have been kept up to date and topical through frequent revision
and expansion of the lecture notes upon which these volumes are based.
Moreover, Greiner greatly increases the value of his presentation by includ-
ing something like one hundred completely worked examples in each volume.
Nothing is of greater importance to the student than seeing, in detail, how the
theoretical concepts and tools under study are applied to actual problems of
interest to a working physicist. And, finally, Greiner adds brief biographical
sketches to each chapter covering the people responsible for the development of
the theoretical ideas and/or the experimental data presented. It was Auguste
Comte (1798-1857) in his Positive Philosophy who noted, "To understand a
science it is necessary to know its history". This is all too often forgotten in
VI Foreword to Earlier Series Editions

modern physics teaching and the bridges that Greiner builds to the pioneering
figures of our science upon whose work we build are welcome ones.
Greiner's lectures, which underlie these volumes, are internationally noted
for their clarity, their completeness and for the effort that he has devoted to
making physics an integral whole; his enthusiasm for his science is contagious
and shines through almost every page.
These volumes represent only a part of a unique and Herculean effort
to make all of theoretical physics accessible to the interested student. Beyond
that, they are of enormous value to the professional physicist and to all others
working with quantum phenomena. Again and again the reader will find that,
after dipping into a particular volume to review a specific topic, he will end
up browsing, caught up by often fascinating new insights and developments
with which he had not previously been familiar.
Having used a number of Greiner's volumes in their original German in
my teaching and research at Yale, I welcome these new and revised English
translations and would recommend them enthusiastically to anyone searching
for a coherent overview of physics.

Yale University D. Allan Bromley

New Haven, CT, USA Henry Ford II Professor of Physics
1989
Preface

Theoretical physics has become a many-faceted science. For the young stu-
dent it is difficult enough to cope with the overwhelming amount of new
scientific material that has to be learned, let alone obtain an overview of the
entire field, which ranges from mechanics through electrodynamics, quantum
mechanics, field theory, nuclear and heavy-ion science, statistical mechanics,
thermodynamics, and solid-state theory to elementary-particle physics. And
this knowledge should be acquired in just 8-10 semesters, during which, in
addition, a Diploma (Masters) thesis has to be worked on and examinations
prepared for. All this can be achieved only if the academic teachers help to
introduce the student to the new disciplines as early on as possible, in order
to create interest and excitement that in turn set free essential new energy.
At the Johann Wolfgang Goethe University in Frankfurt am Main we
therefore confront the student with theoretical physics immediately, in the
first semester. Theoretical Mechanics I and II, Electrodynamics, and Quantum
Mechanics I - An Introduction are the basic courses during the first two years.
These lectures are supplemented with many mathematical explanations and
much support material. After the fourth semester of studies, graduate work
begins, and Quantum Mechanics II - Symmetries, Statistical Mechanics and
Thermodynamics, Relativistic Quantum Mechanics, Quantum Electrodynam-
ics, the Gauge Theory of Weak Interactions, and Quantum Chromo dynamics
are obligatory. Apart from these, a number of supplementary courses on spe-
cial topics are offered, such as Hydrodynamics, Classical Field Theory, Special
and General Relativity, Many-Body Theories, Nuclear Models, Models of El-
ementary Particles, and Solid-State Theory.
This volume of lectures provides an important supplement on the subject
of Quantum Mechanics. These Special Chapters are in the form of overviews
on various subjects in modern theoretical physics. The book is devised for
students in their fifth semester who are still trying to decide on an area of
research to follow, whether they would like to focus on experiments or on
theory later on.
The observation by Planck and Einstein that a classical field theory -
electrodynamics - had to be augmented by corpuscular and nondeterministic
aspects stood at the cradle of quantum theory. At around 1930 it was recog-
nized that not only the radiation field with photons but also matter fields,
e.g. electrons, can be described by the same procedure of second quantization.
VIII Preface

Within this formalism, matter is represented by operator-valued fields that

are subject to certain (anti-)commutation relations. In this way one arrives at
a theory describing systems of several particles (field quanta) which in partic-
ular provides a very natural way to formulate the creation and annihilation
of particles. Quantum field theory has become the language of modern the-
oretical physics. It is used in particle and high-energy physics, but also the
description of many-body systems encountered in solid-state, plasma, nuclear,
and atomic physics make use of the methods of quantum field theory.
We use second quantization (creation and annihilation operators for par-
ticles and modes) extensively. The lectures begin with the quantization of the
electromagnetic fields. As well as the state vectors with a well-defined (sharp)
number of photons, the coherent (Glauber) states are discussed, followed by
absorption and emission processes, the lifetime of exited states, the width
of spectral lines, the self-energy problem, photon scattering, and Cherenkov
radiation. In between it seemed fit to elucidate on the Aharanov-Bohm and
Casimir effects. Many applications are hidden in Exercises and Examples (e.g.
two-photon decay, the Compton effect, photon spectra of black bodies).
Fermi and Bose statistics and their relationship with the way of quantiza-
tion (commutators, anticommutators) are discussed in the third chapter. Here
also, tripple commutators leading to para-Bose and para-Fermi statistics are
reflected upon. After describing quantum fields with interaction (Chap. 4) we
address renormalization problems, not in full (as done in the lectures on quan-
tum electrodynamics and on field quantization), but in a rather elementary
way such that the student gets a feeling for the problems, their difficulties,
and their solution.
In Chaps.6 to 9 the methods of quantum field theory are applied to
topics in solid-state and plasma physics: quantum gases, superfluidity, pair
correlations (Hanbury-Brown-Twiss effect and Cooper pairs), plasmons and
phonons, and the quasiparticle concept give an impression of the flavor of
these fields. The following chapters are devoted to the structure of atoms
and molecules, containing many fascinating subjects (Hartree, Hartree-Fock,
Thomas-Fermi methods, the periodic system of elements, the Born-Oppen-
heimer approach, various types of elementary molecules, oriented orbitals,
hybridization, etc.).
Finally we present an elementary exhibition of Feynman path integrals.
The method of quantization using path integrals, which essentially is equiv-
alent to the canonical formalism, has gained increasing popularity over the
years. Apart from their elegance and formal appeal, path-integral quantiza-
tion and the related functional techniques are particulary well suited to the
implementation of conditions of constraint, which is necessary for the treat-
ment of gauge fields. Nowadays any student of physics should at least know
where and how the canonical and the path-integral formalisms are connected.
Like all other lectures, these special chapters are presented together with
the necessary mathematical tools. Many detailed examples and worked-out
problems are included in order to further illuminate the material.
It is clear from what we have said so far that these lectures are meant to
give an elementary (but not naive) overview of special subjects a student may
 Preface IX

hear about in colloquia and seminars. The lectures may help to furnish better
orientation in the vast field of interesting modern physics.
We have profitted a lot from excellent text books, such as

E.G. Harris: A Pedestrian Approach to Quantum Field Theory (Wiley,

New York 1972),
G. Baym: Lectures on Quantum Mechanics (W.A. Benjamin, Reading, MA
1974),
L.D. Landau, E.M. Lifshitz: Quantum Mechanics (Pergamon, Oxford
1977),

which have guided us to some extend in devising certain chapters, examples,

and exercises. We recommend them for additional reading. The biographi-
cal notes on outstanding physicists and mathematicians were taken from the
Brockhaus Lexikon.
This book is not intended to provide an exhaustive introduction to all
aspects of quantum mechanics. Our main goal has been to present an elemen-
tary introduction to the methods of field quantization and their applications in
many-body physics as well as to special aspects of atomic and nuclear physics.
We hope to attain this goal by presenting the subjects in considerable detail,
explaining the mathematical tools in a rather informal way, and by including
a large number of examples and worked exercises.
We would like to express our gratitude to Drs. J. Reinhardt, G. Plunien,
and S. Schramm for their help in preparing some exercises and examples and
in proofreading the German edition of the text. For the preparation of the
English edition we enjoyed the help of Priv. Doz. Dr. Martin Greiner. Once
again we are pleased to acknowledge the agreeable collaboration with Dr. H.J.
K6lsch and his team at Springer-Verlag, Heidelberg. The English manuscript
was copy edited by Dr. Victoria Wicks.

Frankfurt am Main, Walter Greiner

August 1997
Contents

1. Quantum Theory of Free Electromagnetic Fields . . . . . . . 1

1.1 Maxwell's Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Electromagnetic Plane Waves. . . . . . . . . . . . . . . . . . . . . . 3
1.3 Quantization of Free Electromagnetic Fields . . . . . . . . . . . 5
1.4 Eigenstates of Electromagnetic Fields. . . . . . . . . . . . . . .. 12
1.5 Coherent States (Glauber States) of Electromagnetic Fields 16
1.6 Biographical Notes .... . . . . . . . . . . . . . . . . . . . . . . . .. 29

2. Interaction of Electromagnetic Fields with Matter. . . . .. 31

2.1 Emission of Radiation from an Excited Atom .......... 33
2.2 Lifetime of an Excited State. . . . . . . . . . . . . . . . . . . . . .. 35
2.3 Absorption of Photons. . . . . . . . . . . . . . . . . . . . . . . . . .. 48
2.4 Photon Scattering from Free Electrons . . . . . . . . . . . . . .. 55
2.5 Calculation of the Total Photon Scattering Cross Section.. 57
2.6 Cherenkov Radiation of a Schrodinger Electron. . . . . . . .. 63
2.7 Natural Linewidth and Self-energy. . . . . . . . . . . . . . . . .. 74

3. Noninteracting Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1 Spin-Statistics Theorem. . . . . . . . . . . . . . . . . . . . . . . . .. 98
3.2 Relationship Between Second Quantization
and Elementary Quantum Mechanics .. . . . . . . . . . . . . .. 99

4. Quantum Fields with Interaction . . . . . . . . . . . . . . . . . . .. 109

5. Infinities in Quantum Electrodynamics:

Renormalization Problems . . . . . . . . . . . . . . . . . . . . . . . .. 133
5.1 Attraction of Parallel, Conducting Plates Due
to Field Quantum Fluctuations (Casimir Effect) ........ 133
5.2 Renormalization of the Electron Mass. . . . . . . . . . . . . . .. 143
5.3 The Splitting of the Hydrogen States 2S 1 / 2-2p3/2:
The Lamb Shift. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 149
5.4 Is There an Inconsistency in Bethe's Approach? . . . . . . . .. 156

6. Nonrelativistic Quantum Field Theory

of Interacting Particles and Its Applications. . . . . . . . . .. 161
6.1 Quantum Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.2 Nearly Ideal, Degenerate Bose-Einstein Gases. . . . . . . . .. 174
XII Contents

7. Superfluidity..................................... 193
7.1 Basics of a Microscopic Theory of Superfluidity . . . . . . . .. 194
7.2 Landau's Theory of Superfluidity . . . . . . . . . . . . . . . . . .. 205

8. Pair Correlations Among Fermions and Bosons ........ 213

8.1 Pair-Correlation Function for Fermions . . . . . . . . . . . . . .. 213
8.2 Pair-Correlation Function for Bosons ................ 218
8.3 The Hanbury-Brown and Twiss Effect. . . . . . . . . . . . . . .. 223
8.4 Cooper Pairs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 226

9. Quasiparticles in Plasmas and Metals: Selected Topics. .. 241

9.1 Plasmons and Phonons .... . . . . . . . . . . . . . . . . . . . . .. 246

10. Basics of Quantum Statistics . . . . . . . . . . . . . . . . . . . . . . . 255

10.1 Concept of Quantum Statistics and the Notion of Entropy. 255
10.2 Density Operator of a Many-Particle State . . . . . . . . . . .. 256
10.3 Dynamics of a Quantum-Statistical Ensemble . . . . . . . . .. 272
10.4 Ordered and Disordered Systems:
The Density Operator and Entropy. . . . . . . . . . . . . . . . .. 276
10.5 Stationary Ensembles . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

11. Structure of Atoms ... . . . . . . . . . . . . . . . . . . . . . . . . . . .. 285

11.1 Atoms with Two Electrons. . . . . . . . . . . . . . . . . . . . . . .. 285
11.2 The Hartree Method . . . . . . . . . . . . . . . . . . . . . . . . . . .. 292
11.3 Thomas-Fermi Method . . . . . . . . . . . . . . . . . . . . . . . . .. 293
11.4 The Hartree--Fock Method . . . . . . . . . . . . . . . . . . . . . . .. 297
11.5 On the Periodic System of the Elements . . . . . . . . . . . . .. 305
11.6 Splitting of Orbital Multiplets . . . . . . . . . . . . . . . . . . . .. 306
11.7 Spin-Orbit Interaction. . . . . . . . . . . . . . . . . . . . . . . . . .. 312
11.8 Treatment of the Spin-Orbit Splitting
in the Hartree-Fock Approach . . . . . . . . . . . . . . . . . . . .. 324
11.9 The Zeeman Effect ... . . . . . . . . . . . . . . . . . . . . . . . . .. 327
11.10 Biographical Notes ....... . . . . . . . . . . . . . . . . . . . . .. 332

12. Elementary Structure of Molecules .................. 335

12.1 Born-Oppenheimer Approximation. . . . . . . . . . . . . . . . .. 337
12.2 The Ht Ion as an Example . . . . . . . . . . . . . . . . . . . . . .. 339
12.3 The Hydrogen Molecule. . . . . . . . . . . . . . . . . . . . . . . . .. 346
12.4 Electron Pairing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 349
12.5 Spatially Oriented Orbits. . . . . . . . . . . . . . . . . . . . . . . .. 351
12.6 Hybridization................................. 353
12.7 Hydrocarbons................................. 356
12.8 Biographical Notes .,. . . . . . . . . . . . . . . . . . . . . . . . . .. 358
 Contents XIII

13. Feynman's Path Integral Formulation

of Schrodinger's Wave Mechanics. . . . . . . . . . . . . . . . . . .. 361
13.1 Action Functional in Classical Mechanics
and Schrodinger's Wave Mechanics . . . . . . . . . . . . . . . . .. 362
13.2 Transition Amplitude as a Path Integral. . . . . . . . . . . . .. 365
13.3 Path Integral Representation
of the Schrodinger Propagator . . . . . . . . . . . . . . . . . . . .. 370
13.4 Alternative Derivation of the Schrodinger Equation. . . . .. 374
13.5 Biographical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 376

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 377

Contents of Examples and Exercises

1.1 The Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2 Computation of the Magnetic Contributions to the Energy
of an Electromagnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Momentum Operator of Electromagnetic Fields. . . . . . . . . . . . . 13
1.4 Matrix Elements with Coherent States . . . . . . . . . . . . . . . . . .. 18
1.5 The Mean Quadratic Deviation of the Electric Field
Within the Coherent State. . . . . . . . . . . . . . . . . . . . . . . . . . .. 20
1.6 The Aharonov-Bohm Effect. . . . . . . . . . . . . . . . . . . . . . . . . .. 21
2.1 Selection Rules for Electric Dipole Transitions. . . . . . . . . . . . .. 38
2.2 Lifetime of the 2p State with m = 0 in the Hydrogen Atom
with Respect to Decay Into the Is State . . . . . . . . . . . . . . . . .. 40
2.3 Impossibility of the Decay of the 2s State of the Hydrogen Atom
via the p . A Interaction .... . . . . . . . . . . . . . . . . . . . . . . . .. 41
2.4 The Hamiltonian for Interaction Between the Electron Spin
and the Electromagnetic Field . . . . . . . . . . . . . . . . . . . . . . . .. 42
2.5 Lifetime of the Ground State of the Hydrogen Atom
with Hyperfine Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.6 One-Photon Decay of the 2s State in the Hydrogen Atom. . . . .. 46
2.7 Differential Cross Section dO'/drl for Photoelectric Emission
of an Electron in the Hydrogen Atom (Dipole Approximation) .. 50
2.8 Spectrum of Black-Body Radiation. . . . . . . . . . . . . . . . . . . . .. 53
2.9 The Compton Effect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 59
2.10 Two-Photon Decay of the 2s State of the Hydrogen Atom ..... 60
2.11 The Field Energy in Media with Dispersion. . . . . . . . . . . . . . .. 64
2.12 The Cherenkov Angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 73
2:13 Plemlj's Formula. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 78
3.1 Do the Commutators and Anticommutators
Fulfill the Poisson Bracket Algebra? . . . . . . . . . . . . . . . . . . . .. 85
3.2 Threefold Commutators from an Expansion of Paraoperators . .. 87
3.3 More on Paraoperators: Introduction of the Operator Gjk . . . . . 89
3.4 Occupation Numbers of Para-Fermi States ................ 91
3.5 On the Boson Commutation Relations ..... . . . . . . . . . . . . .. 95
3.6 Consistency of the Phase Choice for Fermi States
with the Fermion Commutation Relations . . . . . . . . . . . . . . . .. 97
3.7 Constancy of the Total Particle-Number Operator. . . . . . . . . .. 102
4.1 Nonrelativistic Bremsstrahlung. . . . . . . . . . . . . . . . . . . . . . . .. 112
4.2 Rutherford Scattering Cross Section. . . . . . . . . . . . . . . . . . . .. 121
XVI Contents of Examples and Exercises

4.3 Lifetime of the Hydrogen 2s State with Respect

to Two-Photon Decay (in Second Quantization) ............ 123
4.4 Second-Order Corrections
to Rutherford's Scattering Cross Section. . . . . . . . . . . . . . . . .. 128
5.1 Attraction of Parallel, Conducting Plates
Due to the Casimir Effect .. . . . . . . . . . . . . . . . . . . . . . . . . .. 137
5.2 Measurement of the Casimir Effect. . . . . . . . . . . . . . . . . . . . .. 140
5.3 Casimir's Approach Towards a Model for the Electron. . . . . . .. 142
5.4 Supplement: Historical Remark on the Electron Mass. . . . . . . .. 144
5.5 Lamb and Retherford's Experiment . . . . . . . . . . . . . . . . . . . .. 150
5.6 The Lamb Shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.1 The Field-Theoretical Many-Particle Problem. . . . . . . . . . . . .. 162
6.2 Equilibrium Solution
of the Quantum-Mechanical Boltzmann Equation ........... 167
6.3 Equilibrium Solution of the Classical Boltzmann Equation . . . .. 172
6.4 From the Entropy Formula for the Bose (Fermi) Gas
to the Classical Entropy Formula . . . . . . . . . . . . . . . . . . . . . .. 173
6.5 Proof of the H Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 174
6.6 Entropy of a Quantum Gas . . . . . . . . . . . . . . . . . . . . . . . . . .. 181
6.7 Distribution of N Particles over G States
(Number of Combinations) . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.8 Stirling's Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 186
6.9 Entropy and Information. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 187
6.10 Maxwell's Demon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 191
7.1 Choice of Coefficients for the Bogoliubov Transformation. . . . .. 199
7.2 An Analogy to Superftuidity in Hydrodynamics. . . . . . . . . . . .. 209
8.1 Pair-Correlation Function for a Beam of Bosons ............ 220
8.2 Boson Pair-Correlation Function as a Function
of the Quantization Volume . . . . . . . . . . . . . . . . . . . . . . . . . .. 222
8.3 The Debye Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 233
8.4 Correlation Length of a Cooper Pair. . . . . . . . . . . . . . . . . . . .. 236
8.5 Determination of the Coupling Strength of a Bound Cooper Pair 238
9.1 Electrostatic Potential of a Charge in a Plasma. . . . . . . . . . . .. 249
9.2 Classical Dielectric Function . . . . . . . . . . . . . . . . . . . . . . . . . . 250
9.3 Details of Calculating the Dielectric Function E(q,W) . . . . . . . .. 252
10.1 Density Operators in Second Quantization ................ 264
10.2 Transformation Equations for Field Operators. . . . . . . . . . . . .. 268
10.3 Commutation Relations for Fermion Field Operators. . . . . . . .. 270
10.4 Density Operator of a Mixture. . . . . . . . . . . . . . . . . . . . . . . .. 274
10.5 Construction of the Density Operator
for a System of Unpolarized Electrons. . . . . . . . . . . . . . . . . . .. 275
10.6 Systems of Noninteracting Fermions and Bosons . . . . . . . . . . .. 281
11.1 Calculation of Some Frequently used Integrals. . . . . . . . . . . . .. 288
11.2 Proof of (11.49) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
11.3 The Hartree-Fock Equation as a Nonlocal Schrodinger Equation. 301
11.4 An Approximation for the Hartree-Fock Exchange Term .. . . .. 304
11.5 Application of Hund's Rules. . . . . . . . . . . . . . . . . . . . . . . . . .. 311
11.6 The Wigner-Eckart Theorem. . . . . . . . . . . . . . . . . . . . . . . . .. 314
 Contents of Examples and Exercises XVII

11.7 Derivation of the Spin-Orbit Interaction. . . . . . . . . . . . . . . . .. 317

11.8 Transformation of the Spin-Orbit Interaction ..... . . . . . . . .. 323
11.9 The Stark Effect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 329
12.1 Calculation of an Overlap Integral and Some Matrix Elements
for the Ht Ion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 343
13.1 Momentum and Energy at the End Point
of a Classical Trajectory ................. . . . . . . . . . . .. 364
13.2 The Transition Amplitude for a Free Particle .............. 369
13.3 Trotter's Product Rule .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 372
1. Quantum Theory
of Free Electromagnetic Fields

From the lectures on classical electrodynamics we know Maxwell's equations

as the basic equations describing all classical electromagnetic phenomena. To
account for quantum effects the Maxwell equations have to be quantized.
We are thus led to quantum electrodynamics. Quantum electrodynamics also
deals with the quantization of the electron-positron field, the pion field, and
other fields and describes their interaction with the quantized electromagnetic
field (i.e. the quantized electromagnetic waves). To begin with, we briefly
recapitulate Maxwell's classical equations.

1.1 Maxwell's Equations

The Maxwell equations of motion for an electromagnetic field read
laB
VxE+-- 0,
c at
V·D 47rQ,
(1.1 )
laD 47r .
VxH--- -J,
c at c
V·B o·,
here we have used the ems system. Taking the divergence of the second equa-
tion and combining it with the time derivative of the third equation, we deduce
the continuity equation for the electric charge and current densities p and j:
'"" . aQ
v . J + at = O. (1.2)
The electric and magnetic field strenghts, E and B, are expressable in terms
of the vector potential A and the scalar potential 'P,
1 aA
E = -~8t-V'P, B = VxA. (1.3)
As an immediate consequence the first and last of Maxwell's equations (1.1)
are automatically fulfilled.
The potentials A and 'P are not unique; the modified potentials
1 aX (1.4)
A
I
= A + VX, 'P = 'P - ~8t
I
'

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
2 1. Quantum Theory of Free Electromagnetic Fields

where x(r, t) is an arbitrary function depending on position r and time t, yield

the same fields E and B. This modification of the potentials, which leaves
the fields strenghts E and B unchanged, is called a gauge transformation. In
our lectures about quantum mechanics it is proven that the wavefunction W
has to be replaced by

w' = wexp(~x) (1.5)

for a gauge transformation (1.4) once the electromagnetic potentials are in-
troduced via the minimal coupling
e
p ----+ p - - A (1.6)
c
into the Schrodinger equation. The form of the wave equation, which besides
the Schrodinger equation could also be the Pauli equation or the Dirac equa-
tion or any other, remains unaltered. Sometimes the transformation (1.5) is
called a gauge transformation of first degree, whereas the transformation (1.4)
is denoted as a gauge transformation of second degree.
Inserting (1.3) into the second and third of Maxwell's equations (1.1), we
arrive at
411' .
-),
c
(1.7)
-47rQ.

Once the vector A is described in an Euclidean coordinate system, the first

part of the left-hand side of the first equation can be rewritten as
v x (V x A) = V (V· A) - V2 A. (1.8)
The last term V2 A represents a vector with components L'lA i , where the
Laplace operator L'l acts on the components Ai of the vector potential sep-
arately. Making use of (1.8), we can further simplify the equations resulting
from (1.7), once the gauge transformation (1.4) is applied, which leads to new
potentials A' and cp'. We choose the Lorentz gauge fulfilling

V . A' + ~ acp'
= O. (1.9)
c at
Here the gauge function X is determined from

V2X_~~X
2
= _(V'A+~acp) (1.10)
2 c at c at
Equations (1. 7) finally read

1 a 2A' 411' .
V2A' --- --),
c2 at 2 c
(1.11)
1 a 2 cp'
V 2 cp' - -47rQ.
c2 at 2
 1.2 Electromagnetic Plane Waves 3

1.2 Electromagnetic Plane Waves

For a completely empty space j = 0 and (! = O. For this case we can always
find a gauge function X such that
V· A'(r, t) 0,
(1.12)
ip/(r, t) o
for all rand t; see Exercise 1.1. This gauge is called the Coulomb gauge.
As a consequence we find transverse plane waves as a solution for A' and
consequently for E and B. In the following we drop the primes attached to
A' and ip' for convenience and deduce from (1.9) and (1.11) that

V2A _ ~ fJ2A
c2 fJt2 0,
...."
v . A 0, (1.13)

ip o.
A typical plane-wave solution of these equations is charaterized by a real
vector potential A governed by the wave-number vector k, called the wave
vector for short, and the real polarization vector e:
A(r, t) 2elAoi cos (k . r - w t + 0:)
-- A oe i(k·r-wt) + A*oe -i(k·r-wt)
= Ao ei(k.r-wt) + C.c .. (1.14)
Here Ao = lAo Ie eia is the amplitude and "c.c." stands for the "complex
conjugate" of the first term. It is easy to see that the ansatz (1.14) is a
solution of the first equation of (1.13) if
w = kc = Iklc, (1.15)
and of the second equation of (1.13) if the polarization vector Ao is perpen-
dicular to the wave vector k,
Ao -L k. (1.16)
We say that Ao is transverse as it fulfills the last relation, i.e. Ao' k = O. The
corresponding electric and magnetic fields follow from (1.3) and (1.14):
E -2klAole sin (k . r - wt + 0:) ,
(1.17)
B -2lAolk x e sin (k· r - wt + 0:)
The Poynting vector is defined as
S = ~ExH (1.18)
47r
and, obviously, is parallel to k; remember that H = B for a charge-free and
current-free space. The time average of the Poynting vector, S, over one period
T = 27r /w of oscillation is given by
2
S = ~ IAol2 (1.19)
27rc
and represents the intensity of the electromagnetic plane wave.
4 1. Quantum Theory of Free Electromagnetic Fields

EXERCISE

1.1 The Coulomb Gauge

Problem. Show that the general solution of the Maxwell equations can be
expressed by potentials AI and <pI with V . AI = 0 and <pI = 0 if V . j = 0
and (} = O.

Solution. According to the gauge transformation (1.4) the two requirements

V . AI = 0 and <pI = 0 lead to
aX c<p(r,t) , (1)
at
V 2 x -V·A(r,t). (2)
Differentiating the second equation with respect to time and then inserting
the first equation yields
1 aA
2
V <p(r,t)+V. --a
c t
= o. (3)
But this equation is only valid if (} = 0; in fact it follows from Maxwell's
equation V . E = 47T (}:

V.E = V. (_~aA -V<P)

c at
= - (V. :~ + V2<p) = 47T (}. (4)

A consistency check with another one of Maxwell's equations then gives

V. (V x H - ~ aD) = V. (V x H) _ ~ a(v· D)
cat c at
= 47T (V. j) . (5)
c
Since the first term on the left-hand side is always equal to zero, i.e.
dive curl H) = 0, and the second term now vanishes because V D = 47T (} = 0,
we conclude that
V·j = o. (6)
The explicit solution for the gauge function X is obtained from a time inte-
gration of (1):

x(r, t) = c J<per, t) dt + const. (7)

Note that the two requirements V . AI = 0 and <pI = 0 automatically fulfill

the Lorentz gauge (1.9), so that the wave equations (1.11) still hold, but only
with (} = V . j = O.
 1.3 Quantization of Free Electromagnetic Fields 5

1.3 Quantization of Free Electromagnetic Fields

Once again we write down Maxwell's equations without source terms:

V·B 0,
V·E 0,
1 aB (1.20)
VxE --- ,
c at
1 aE
VxB = +--.
c at
The first three of these equations are automatically fulfilled if the Coulomb
gauge (1.12)

'P 0,
(1.21 )
V·A
°
is considered and (1.3) is used. The last of (1.20) then leads to the wave
equation

(1.22)

for the vector potential, which is already known from (1.13).

For the development of the quantum theory of an electromagnetic field it is
of advantage to express the field by a set of discrete variables. We confine the
electromagnetic field to a large box (cube) of volume fl = £3; in this respect
the normal modes of the electromagnetic field are discretized. The general A
field is the superposition of these normal modes A k ". The Fourier coefficients
ak,,(t) of this expansion will be treated as the actual field variables, which
describe the dynamics. We aspire to the quantization of these field variables. z
The boundary conditions of the field at the box's sides now come into the
problem. It is most convenient to require periodicity of A at the sides. This
means that the normal modes are determined such that full wavelengths fit
into the box. Mathematically, this translates into

A(£,y,z,t) A(O,y,z,t) ,
A(x,£,z,t) A (x, 0, z, t) , (1.23) Fig. 1.1. The electromag-
netic field is confined to a
A (x, y, £, t) A(x,y,O,t) . box; a normal wave mode
with two full wavelengths
Remembering that for normal modes all degrees of freedom vibrate with the along the x axis is indi-
same frequency, we are lead to the ansatz cated
A (x, y, z, t) = A (x, y, z) e iwt (1.24)
Inserting it into the wave equation (1.22) we arrive at

(~'; + ~:) A (x, y, z) = 0. (1.25)

6 1. Quantum Theory of Free Electromagnetic Fields

A normal solution is given by

Ak<1 = Nk E:k<1 eik ·x (0' 1, 2) , (1.26)
where

(1.27a)
or
(1.27b)
The two polarization vectors E:kl and E:k2 are perpendicular to the wave vector
k, which determines the direction of propagation:
E:k<1' k = o. (1.28)
This is a result of (1.21) and (1.26). We choose the two polarization vectors,
Fig. 1.2. The polarization
vectors cku are perpendic- which now both lie in the plane perpendicular to k, as two linearly indepen-
ular to the wave vector dent vectors orthogonal to each other (see Fig. 1.2):
k. This reflects the trans-
verse character of electro- (1.29)
magnetic waves The normalization factor Nk appearing in (1.26) will be determined later. In
order to fulfill the boundary conditions (1.23), the wave vector k has to take
on the form
211"
k = (kx, ky, kz) = L {nl' n2, n3} , nl, n2, n3 E Z, (1.30)

where ni are integer numbers. k is a vector in k space (see Fig. 1.3).

The most general solution for the A field is now a superposition of all
normal modes. Since the spatial part of the normal modes are pure imaginary
exponentials, see (1.26), this superposition is equal to a Fourier series:
~--------------"kx
Fig. 1.3. Illustration of k
A(x, t) = L L [ak<1 (t) Ak<1 (x) + a;w (t) Ak<1 (x)]
k <1= 1,2
space k.>O

L L Nk E:k<1 (ak<1 (t) eik-x + ak<1 (t) e- ik .x ) . (1.31)

k <1=1,2
k.>O

We have made sure that the vector potential, which determines the real
electromagnetic fields, stays real by adding the complex conjugate to each
term of the sum; in order to avoid double counting the summation has to
be restricted to those vectors k = {k x , ky, kz} with kz > O. Each term in
the sum (1.31) contains a term with eik . x and also one with e- ikx , so that
the summation has to be limited; see also Fig. 1.4. The factors eiwkt have
Fig. 1.4. The half space been absorbed into the time-dependent Fourier coefficients ak<1' Because of
kz > 0 is shown with a the normalization factor N k , the Fourier coefficients do not represent the
hatching. Every vector k
amplitudes of the A field alone, but nevertheless reflect a measure for the
of this half space can be
transformed by a reflec- latter. Later on aku and at]" will be interpreted as creation and annihilation
tion into -k belonging to operators for photons. Because aku and a ku will not commute anymore as
the lower half space operators, we have to pay attention to their precise ordering from now on.
 1.3 Quantization of Free Electromagnetic Fields 7

Inserting the expansion (1.31) into the wave equation (1.22) yields

"L "
L Nk eka [(_k2 aka (t)_~d2aka(t))
2 d2 e ik·x
k a = 1,2
c t

from which we deduce

o (1.32)

This differential equation has the solution

aka (t) = ak~ (0) e- iwkt + ak~ (0) eiwkt , (1.33)

so that the most general A field (1.31) becomes
A(x, t) = L Nk eka [ak~ (0) ei(kx-wktl + ak~ * (0) e-i(kx-wktl
k,a
kz>O

+ ak~ (0) ei(k,x+Wktl + ak~ * (0) e-i(k'X+Wktl] (1.34)

Somehow, the restriction kz > 0 of the sum over k is still disturbing.

Therefore we redefine the free constants ak~ (0) and ak~ * (0) appearing in
(1.33) as follows:

ak~ (0) = aka (0) }

for kz > O. (1.35)
ak~ (0) = -( -It a~ka (0)
Then (1.34) transforms into
A(x, t) = L Nkeka [aka (0) ei(k'x-Wkt) + aka (0) e-i(k,x-wktl
k,a
kz >0

- (-1)" (a~ka (0) ei(k'x+Wktl + a-ka (0) e-i(kX+Wktl)] k

(1.36)

Whereas the quantities Nk and Wk do not depend on the direction of k,

the polarization vector eka changes into e-ka = _(_1)17 eka with a = 1, 2 ..---
8-k2 /
as k turns its direction around; in this respect the right handedness of the
three vectors k, ek1 and ek2 is guaranteed (see Fig. 1.5). We deduce further:
/'
,/
,/
A(x, t) = L Nk eka [a kU (0) ei(k-x-wktl + aka (0) e-i(kX-wktl]
,/
)J -k
k,u
kz>O Fig. 1.5. Illustration of

+ L Nk e-ka [a~ka (0) eiwkteik.x + a-ku (0) ei(-kX-Wdl]

the relation ek" = - ( -1}"
x e-k", (T = 1,2
k,a
kz>O
8 1. Quantum Theory of Free Electromagnetic Fields

L Nk eka [aka (0) ei(k'x-wkt) + aka (0) e-i(k'X-Wk t )]

k,a
kz>O

+ L Nk eka [aka (0) e-iwkteik.x + aka (0) eiwkte-ik.X]

k,a
k%<O

L Nk eka [aka (t) e ik .x + aka (t) e- ik .x ] (1.37)

k,a
The second part of the summation in (1.36) is identical in form to the first
part, except that the summation now runs over the half space kz < O. In the
last step of (1.37) the relation
aka (t) = aka (0) e- iwkt (1.38)
has also been introduced, which is consistent with (1.33) and (1.35). From
this relation it follows that
daka(t) . ( )
dt = -lWk aka (t) , 1.39
which for all k and (J can be interpreted as the equation of motion for the
field. In fact, we will soon show that they can be derived from a Hamiltonian
that is identical to the total energy of the field.
Let us first calculate the energy. We know from the lectures on classical
electrodynamics,l that the energy of the electromagnetic field is given by

H = 1 { d3 (E2 + B2)
81T J£3 X

1 { d3 [~aA. aA* + (V x A) . (V x A*)] (1.40)

81T J£3 x c at
2 at
where we have used expression (1.3) and the Coulomb gauge (1.12) in the
last step. Now we insert the Fourier expansion (1.37). The first term in (1.40)
with use of (1.39) then gives
{ d3 x_1_ aA . aA* _
1£3 81TC2 at at -

(1.41)
We exploit the orthogonality relations
{d3xeik,xe-ik'.x = L 3 8kk , (1.42)
1£3
and

(1.43)
_(_1)17 81717 , ,

1 See W. Greiner: Classical Electrodynamics (Springer, New York 1996).

 1.3 Quantization of Free Electromagnetic Fields 9

so that (1.41) transforms into

rd3x
JL3
_l_aA . aA*
87fc 2 at at
=

+(-It[akaa_ka+akaa~ka]} . (1.44)
Analogously the computation of the second line of (1.40) becomes

rd3x~87f (V' x A)· (V' x A*)

J£3 = """ 8k:N't £3 {[aka aka + aka aka]
~
k,a "
- (-It [aka a-ka + aka a~ka]};
(1.45)
consult Exercise 1.2. This is identical to the expression (1.44) except for the
sign of the second term. Therefore the addition of both expressions, which is
equal to the relation (1.40), yields
2
H = L 4Wk 2 N't £3 (aka aka
7fC
+ aka aka) . (1.46)
k,a
This resembles very much the energy of a system of uncoupled harmonic
oscillators, which becomes even more apparent if we choose the normalization
constant of the normal modes (1.26) to be

Nk = V27ffic 2 . (1.47)
£3 Wk

Then the energy of the electromagnetic field becomes equal to a sum of ener-
gies corresponding to "harmonic field oscillators":

H = ~ Lliw k (akaaL +aLaka) . (1.48)

k,a
In other words, the normal modes of the electromagnetic field normalized with
(1.47) behave like classical oscillators with energy liw k .
Up to now, aka and aka have been treated as classical amplitudes. On this
level, we could rewrite (1.48) to give

H = L nwk aka aka· (1.49)

k,a
However, we prefer to stick to the form (1.48), where we have emphasized the
exact ordering of the amplitudes. In a moment we will interpret aka and aka
as noncommuting operators; as a consequence of which the two expressions
(1.48) and (1.49) will not be identical anymore.
10 1. Quantum Theory of Free Electromagnetic Fields

EXERCISE

1.2 Computation of the Magnetic Contributions to the Energy

of an Electromagnetic Field

Problem. Compute the expression (1.45), i.e. the integral

I = rd3x~(V
JL3 87r
x A)(V x A*),

using the Fourier expansion (1.37).

Solution. With
A(x, t) = L Nk ek17 [ak17 eik .x + ak17 e- ik .x ]
k,17
and (1.37) we calculate
V x A(x,t) = i LNkk x ek17 [ak17eik.x -ak17e-ik.X]
k,17
and, analogously,
VxA*(x,t) = i LNklk'xekl171 [akl17leikl'X-akl17le-ikl.X].
k', (7'

Altogether we get for the integral I

I = - 8~ L L NkNk' (k x ek(7)' (k' x ek ' 17 / )

k,u k',a'

x fL3
* ik X] [akl17 eik'·x - akl17,e
d3 x [ ak17e ik .x - ak17e-' ' * -ikl.X]

U sing the identity

(k x ek(7)' (k' x ekl(]"l) = (k· k') (ek17' ek ' 17 / ) - (ek17' k') (ek l17 1 . k)
and the orthogonality relation

JL3
rd3xei(k-k ' )'''' = L3 bkk'
we deduce further that

I = - 8~ L L L3 NkNk' [(k . k') (ek17 . ek'17 / ) - (ek17 . k') (ek l17 1 . k)]
k,a k',o'

X [(ak17akl17' + ak17ak'17') bk,-k l - (ak<Takl17' + ak17akl17/) bkk /] .

The relation e k17 . k' = 0 holds not only if k = k', but also if k = - k', because
both k and -k are perpendicular to ek17' In addition we have ek17' ek17 ' = b1717 1
as well as ek17 . e-k17' = -(-1)17 b1717 ,; consult Fig. 1.5. Moreover Wk = W-k
and Nk = N-k. Therefore we get

I = '" ~ k 2N'f L 3 [( ak17ak17 + ak17ak(7) - ( -1) (ak17a-ka + ak17a~k(7) ]

17
L87r
k,17
 1.3 Quantization of Free Electromagnetic Fields 11

With k 2 = wkj c2 we finally obtain Exercise 1.2.

1= L
k,CT
L~~ N'f£3 [(akCTakCT + akCTakCT) - (-It (akua-kCT + akCTa':.kCT)] ,

which is identical to the stated result (1.45).

Now we introduce the quantization. The expression (1.48) for the energy
of an electromagnetic field motivates us to proceed in an analogous manner
as with the quantization of phonons of the harmonic oscillator. 2 From now on
we designate aka as ala and interpret ala as the Hermitian adjoint of aka. If
the operators ala and ak' a' possess different quantum numbers, i.e. belong to
different oscillators, it is reasonable to assume that they commute with each
other. We require that

(1.50)

In addition

[akCT , ak'CT/L = [ala, aL,] _ = 0 (1.51 )

should hold. Then the Hamiltonian (1.48) should be interpreted as an operator

describing the electromagnetic field:

iI = 'L-2-
"' nwk (atka ak a + ak a at)
ka
k,a
(1.52)

Obviously the zero-point energy of the electromagnetic field

L~k
k,a
is infinite because of the existence of infinitely many field oscillators. However,
we will discover that this infinite contribution plays no role in most of the
physical problems. Only in special cases is it necessary to consider the zero-
point energy, for example, in the change in the zero-point energy resulting from
a change in the volume or the boundary conditions of the physical system. As
an explicit example we will discuss the Casimir effect.
Heisenberg's equations of motion for the operators aka read

or (1.53)

Formally they are identical to the classical equations of motion (1.39).

2 See W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1995).
12 1. Quantum Theory of Free Electromagnetic Fields

1.4 Eigenstates of Electromagnetic Fields

The Hamiltonian (1.52) splits into separate contributions resulting from the
various normal modes of the electromagnetic field. As a consequence the
eigenstate has to factorize with respect to different normal modes. The eigen-
state I ... , nkcr, ... , nk' cr', ... ) of the electromagnetic field is a direct product
of eigenstates Inkcr) belonging to separate field oscillators. We write
I .. ·,nkcr, ... ,nk'a""') = .. ·Inka) .. ·Ink'a') .... (1.54)
The relation
t
akcrakcr
A AInka )
nka Inka) (nka = 0,1,2, ... ) (1.55)
represents the eigenvalue equation for the number operator f.ha of the normal
mode kcr; in a moment we will realize which particles are to be counted here.
This and the two following relations have been adopted directly from our
introductory course in quantum mechanics and need no further specification.
As a direct consequence of the commutation relations (1.50) we note
aka I· .. , nka, ... ) = Vnka I· .. , nka - 1, ... ) (1.56)
and
at.,. I· .. , nka, ... ) = Vnka + 1 I· .. , nkcr + 1, ... ) . (1.57)
Using (1.54) and (1.55) together with (1.52), we observe that the states
I ... , nka, ... ) are indeed eigenstates of the Hamiltonian H:

HI .. ·,nka, ... ) = '2)iwk (at.,.aka+~)I ... ,nka, ... )

k,a

Lnwk (nka +~) I .. · ,nka, ... )

k,a
E I ... ,nka, ... ), (1.58)
where the energy E is given by

E = L nwk ( nkcr + ~) . (1.59)

k,a
I:k, cr nwk/2 represents the now-known infinitely large zero-point energy.
The momentum of the classical electromagnetic field is given by

p = r
iL3
d 3 x _l_E x B ;
41fc
(1.60)

compare with the lectures on classical electrodynamics. In Exercise 1.3 we will

transform this expression into
 1.4 Eigenstates of Electromagnetic Fields 13

P = Lnk ilk rr ilkrr ' (l.61)

k,rr
where (l.37), (1.3), and (1.12) will be used. This enables us to define the
momentum operator of the electromagnetic field as

(l.62)
k,rr k,rr
Because of (l.55), the states (l.54) are also eigenstates of the momentum
operator:
pI· .. , nkrr, ... ) L nknkrrl··· ,nkrr,"')
k,rr
L nk nkrr I ... , nkrr, ... ) (l.63)
k,rr
with eigenvalues
(1.64)

Based on relations (l.55) to (l.64) the electromagnetic field may be interpreted

as a state consisting of nkrr photons with energy ru.vk, momentum nk, and
polarization ekrr. The quanta of the field are the photons themselves. The
electromagnetic field as a whole consists of many of these photons; depending
on the occupation numbers nka the photon states describe various states of the
field. According to the relation (l.57) the operator ilka increases the number
of photons with quantum number ka by one; therefore this operator is called
a photon creation operator. On the other hand the operator ilka annihilates
a photon with quantum number ka, which follows from (l.56); it is called
an annihilation operator. The operator Nkrr of (l.55) counts the number of
photons belonging to the mode ka and is called a number operator.

EXERCISE

1.3 Momentum Operator of Electromagnetic Fields

Problem. Show that the momentum operator of the electromagnetic field

can be brought into the form

P = LnkilLilkrr .
k,rr
Start with the classical expression for p and use the Fourier expansion (l.37)
in the Coulomb gauge.

Solution. The expressions for the classical fields read E = -~ ~1- Vtp and
B = V x A. In the Coulomb gauge (without any sources) the potentials
become 'jJ = 0 and V . A = 0; the Fourier expansion of the vector potential
yields
14 1. Quantum Theory of Free Electromagnetic Fields

Exercise 1.3.

where Nk is given by (1.47). The momentum of the electromagnetic field is

p = r
lL3
d3xE x B
47rc
.

From Heisenberg's equations of motion (1.53) we get akl> = -iWkiikl>' so that

the electric field E becomes

E, 1 BA
= -~8t = i
~
L N kWkCkl> (akl>e
'k A l·r t
A

- akl>e
'k)
-1 ·r
. (1)
k,1>

Similiarly we get for the magnetic field B

B = L Nk'V x ck'I>,(iik'I>,eik'.r +iil'I>,e-ik'.r)
k',a'

(2)
k',er'

Observe that the electromagnetic fields have become operators. Then the mo-
mentum operator p results:

p= _1_
47rC 1£3
rd 3 xE x B

- 4:c2 L L N k N k ,Wk Ck!7 X (k' X Ck'I>')

k,a k',u'

We multiply the last two square brackets explicitly, use

lL3
rd 3xe i (k-k').r = L30kk' ,

transform
Ckl> X (k' X ck'I>') = k' (ckl>' ck'I>') - Ck'l>' (ckl>' k') ,
and arrive at

p = - 4~:2 L L NkNk,Wk [k' (ckl> . ck'I>') - Ck'l>' (ckl> . k')]

k,a k',er'

(3)

First we concentrate on the term rv 0kk' symmetric in k and k':

Ps = 4L32
7rC
LL NkNk'Wk [k' (ckl> . ck'I>') - ck'l>' (ckl> . k')]
k,cr k',u'
 1.4 Eigenstates of Electromagnetic Fields 15

The term rv E:ku ·k' vanishes for k = k' since E:ku ..l k. Furthermore E:ku 'E:ku' = Exercise 1.3.
buu" so that the double sum in Ps breaks down:

With

i.e.

The sum Ek, u ~k over all momenta is equal to zero, because positive and
negative contributions cancel pairwise. Thus
(4)

We still have to work out the antisymmetric term'" bk,-k' in (3):

Pas =
£3
-41rc2 L NkN-kWk[-k(E:ku·Cku,)+Cku(E:ku· k )]
k,cr,cr'

Again E:ka' k = 0 vanishes and E:ku' E:-ku' = -(-1)"" baa'; see Fig. 1.5. We
deduce

- £3 2
41rC
Lk
[NkN_kWk (-k) (o'klo'-kl + o'llo'~kl)

+NkN-kWk (+k) (o'k20'-k2 + o'l20'~k2)]

The two terms both include a factor k multiplied with other quantities, which
are invariant under the interchange k --t -k. Because the sum goes over all
positive and negative components of k, terms cancel pairwise again. Therefore

Pas = O.
Then the total momentum of the electromagnetic field finally reads
16 1. Quantum Theory of Free Electromagnetic Fields

1.5 Coherent States (Glauber States)

of Electromagnetic Fields

The electric field is determined by

(1.65)

where (1.3), (1.12), (1.37), and (1.39) have been used. It is tiresome to carry
along the sum over k and rJ all the time; therefore we will suppress this sum
in the following and only track one representative term out of (1.65):

E = +i j27flkv
----V-C [a (t)
·k e lX - a* (t) e- l·k .X] . (1.66)

This term describes one normal mode of the electromagnetic field. We are
tempted to change the classical electric field of E in (1.66) straight forwardly
into an operator E by writing

,
E(x, t) = +i j27flkv
----v-e [a(t) e ·k
l ·X - at (t) e- l·k .X] . (1.67)

However, some care has to be taken. We start again with the original classical
equation E = - ~ ~1 and define the operator for the electric field by the
operator equation

(1.68)

For the computation of the commutator we need to consider only the terms
in the expression for iI and A. with the same a and at; they characterize one
normal mode. The operators belonging to different normal modes commute
because of (1.50). We get

-- IV L3wk
c
27fhc2 " ..
--liWk ( i) [,
--
h
c ae ik.x+,t
a e -ik.x,t,]
a a
' -

_ .j27flkv
- I L3
k [, ik·x _ ,t
e ae a e
-ik.x]
, (1.69)

which is identical with the straightforward proposition (1.67). The last ap-
proach makes use of the definition of the vector potential as a quantum-
mechanical operator and of Heisenberg's equations of motion. It represents
the correct quantum-mechanical approach.
Generally the electromagnetic field contains many photons. It is charac-
terized by an N-photon state like the one of (1.54), which is a direct product
of states Ink,,) describing nk" photons of mode krJ. With In) we simply denote
the state corresponding to the mode characterized by the operators a, at. In
 1.5 Coherent States (Glauber States) of Electromagnetic Fields 17

order to determine the classical value of the electromagnetic field E, we de-

termine the expectation value of the E operator within the n-photon state
In). With (1.67) we find

(nIEln) = i J27rnw
A 'k
--v-c(nlae"x - ate-'
'k
,xln) = 0, (1. 70)

where
(1.71)
enters because of relations (1.56) and (1.57). On the other hand the expecta-
tion value of the square of the electric field, which apart from a factor of 87r
describes part of the energy density, is equal to

27rnw
- - c ' c (nl (ae''k ,X - ate-'
'k
'X)
87r£3
x (ate- ik .x - ae ik .x ) In)

-- naa +a an
27rnw (IAAt
87rL3
AtAI)

4n;3 (nI2a t a+1In) = 2n;3 (n+~) (1. 72)

This result is surprising. The expectation value of the E field in the n-photon
state vanishes, whereas, except for the zero-point energy, the energy density
is equal to the half of n photons times nw per volume L3. The other half of
the expected energy density results from the term i1 . i1 t/87r. This puzzle is
resolved once we realize that statistical phases are assigned to the n photons
in the field. After we average over the phases, the expectation value of E
vanishes. On the other hand the phases drop out for the energy density; the
energy density is a real positive number for every space-time point.
It was Glauber,3 who introduced a field state, for which the expectation
value of the E operator resembles the classical E field. It allows for an un-
certainty in the photon number in order to define the phase of the field more
precisely. The Glauber state Ie), often also called the coherent state, reads

L
00

Ie) bn In), (1. 73)

n=O
where
en e-!l c I2
bn (1.74)
yQ
and c represents a complex number. The probability of finding n photons in
the Glauber state is given by the square of the amplitude:
2 Icl 2n e- Ic12
Ibnl =
n.
I . (1.75)
The sum over all these probabilities has to be unity; indeed, we find
3 R.J. Glauber: Phys. Rev. Lett. 10 (1963) 84.
18 1. Quantum Theory of Free Electromagnetic Fields

f: Ibnl
n=O
2 = e- 1c12 f: Ic~~n =
n=O
e- 1c12 e+ 1c12 = 1, (1. 76)

Since the Glauber state has no definite photon number, the expectation value
of the annihilation operator is not zero; this is in contrast to the incoherent
state (1.54) with definite photon number. We calculate
00 00

m=On=O

m,n=O

m=O

c, (1. 77)

In the same manner we get

(1. 78)
With these two results we deduce the expectation value of the E field (1.67)
in the Glauber state:

(cIElc) =
A

+i
J21f1iW
----ve(clae1X -
'k t
a e- 1
'k
'Xlc)

+i J21fhw
----ve[ce ok oX_c*e- 1ok ,x]
1

+ iJ2~n; e 2 i Icl sin(k' x + 6c )

-21cl J21fhw
----v- e sin(k ' x + 6c ) , (1. 79)

This expression has the form of a classical electromagnetic wave; see (1.17),
Its amplitude is given by Icl and its phase is determined from the phase of
c = lei ei8c , It is worth mentioning that the second line of (1. 79) is formally
identical to the expression (1.67); the creation and annihilation operators a, at
have been substituted by the complex amplitudes c, c*,

EXERCISE

1.4 Matrix Elements with Coherent States

Problem. Show the validity of the following relations:

(a) (cia tale) lel 2 = (ii)
(b) (claatle) = lel 2 + 1 = (ii) +1
 1.5 Coherent States (Glauber States) of Electromagnetic Fields 19

(c) (clataatalc) = Ic1 4 +lc1 2 = (n 2 ) Exercise 1.4.

(d) (cla 2 Ie) = c2
(e) (cla t2 lc) = C*2

Solution. We already know the action of a on Ie):

a Ie) = c Ie) ; (1)
that means that Ie) is an eigenstate of a with eigenvalue e. The conjugate
equation reads
(ale))t = (ele))t = (ela t = (cle*,
or in short
(ela t = (ele*. (2)
We also remember the commutation relation
aa t - ata = 1. (3)
With the help of (1)-(3) we prove relations (a)-(e):
(a) (elatale) e*(elale) = e*e(ele) = lel 2 ,
because(ele) =1.
(ell + atale)
(ele) + (cia talc) = lel 2 + 1 ,
because of (a) and (3)
(ela t (l +ata)ale)
(elatale) + (elatataale)
lel 2 + e*e*ee = Icl 4 + lel 2
(d) (ela 2Ic) (elaale) = e(elajc) = e2
(e) (ela t2 Ic) (elatatle) = e*(elatle) = e*2.
The photon number operator n = ata has the expectation value (elnle) = lel 2
and (eln2Ie) = lel 4 + lel 2 for its square. This shows again that the coherent
state has no definite photon number, since c is an arbitrary complex number.

The Glauber state has no definite photon number n. We therefore ask for
the uncertainty ,1n of the photon number within the coherent state Ie), which
is defined as

6n = V(cl (n - (n))2Ie). (1.80)

With the results of Exercise 1.4 we determine
J(cln2Ie) - (elnle)2
Jlel 4 + lel 2 - lel 4
lei = J (elnle) . (1.81 )
20 1. Quantum Theory of Free Electromagnetic Fields

The so-called relative uncertainty

6n 6n
n (elnle)
1
(1.82)
J(elnlc)
becomes smaller the larger the average photon number (elnle) becomes in the
Glauber state. For coherent states with very many photons the E field still
behaves like a classical field [see (1.79)] and, in general the relative uncertainty
of the photon number becomes very small. Such states play an important
role in laser physics. Coherent states have minimum uncertainty, so that the
equality sign holds in Heisenberg's uncertainty relation.

EXERCISE

1.5 The Mean Quadratic Deviation of the Electric Field

Within the Coherent State

Problem. Determine the mean quadratic deviation of the electric field within
a coherent state and show that it vanishes in the classical limit.

Solution. The mean quadratic deviation of the electric field is defined by

(elE . E tie) - (eIEle)2. Using expression (1.67) for the E operator we find
A At 2nru.u ·k t·k
(eIE· E Ie) = ----v-c· c(el (iie1X - ii e- 1 .X)

x (iite- ik .x - aeik .x ) Ie)

2nru.u (e Iaa
----v- AAt + aAt aA- aae i2k·x - aAt aAt e -i2k.XI e) .
AA

With the results derived in Exercise 1.4 it follows further that

(eIE.Etle) = 2~n; (2IeI2+1_e2ei2k.x_e*2e-i2k.x). (1)

The expectation value of E within the Glauber state has already been calcu-
lated in (1. 79). We recapitulate:
(eIEle) (eIEle)*
2nru.u 2 2
----v- c . c (lei + le* I
-e2 ei2k-x - (*)2
e e -i2k.x) . (2)
Substraction of (2) from (1) yields the mean quadratic deviation
t 2 2n ru.u
(elE . E Ie) - I(eIEle) I = ----v-. (3)
A A A

In the classical limit, i.e. n ---> 0, the mean quadratic deviation goes to zero.
 1.5 Coherent States (Glauber States) of Electromagnetic Fields 21

EXAMPLE

1.6 The Aharonov-Bohm Effect

Here we will discuss an effect that gives interesting insight into the nature
of the electromagnetic field and its role for quantum mechanics. It is a typi-
cal interference phenomenon that can be understood at an elementary level;
strangely enough, its importance was recognized only 30 years after the de-
velopment of quantum mechanics. 4
Let us recapitulate how the interaction between charged particles is de-
scribed in the classical picture. Following the conception by Faraday and
Maxwell this interaction is not given via a remote action, but as a local
interaction with an electric field E(x, t) and a magnetic field B(x, t). It is
sufficient to know the field strengths E and B at each space-time point of
the particle's trajectory to describe the motion of a particle with charge e.
Newton's equation of motion holds with the Lorentz force

F= e ( E+ ~x B) . x (1)

In classical electrodynamics the Coulomb potential Ao(x, t) and the vector

potential A(x, t) are merely introduced as technical quantities to facilitate
formulations:
laA
E -Y'Ao - ~8t' (2)
B Y'xA.
These potentials were not expected to have a physical interpretation. Beyond
that, the potentials are not unique; gauge transformations

AIo 1 aX
Ao - ~7ii' (3)
AI A+ Y'x
with an arbitrary function X(x, t) leave the electromagnetic fields and, conse-
quently, also the Lorentz force invariant.
In quantum mechanics the notion of a force is not of much use. One starts
instead with a Hamiltonian. The Hamiltonian of a nonrelativistic charged
particle moving in an electromagnetic field contains the potentials Ao and A,
e)2
H = - 1 ( p - - A + eAo , (4)
2m c
and leads to the Schrodinger equation

iIL:ttJ!(x,t) = L~ (-ilLY' - ~Af +eAo] tJ!(x,t). (5)

The wavefunction tJ! now directly depends on the potentials Ao and A and only
indirectly on the fields E and B. Even when fields vanish, the wavefunction
tJ! is still influenced by the potentials.
4 Y. Aharonov, D. Bohm: Phys. Rev. 115 (1959) 485.
22 1. Quantum Theory of Free Electromagnetic Fields

Example 1.6. To begin with, we study the special case in which A leads to a stationary
and curlfree field B = V x A = o. If the wavefunction!lio represents a solution
of the Schrodinger equation (5) with A = 0 but otherwise arbitrary Ao(x, t),
then we can easily show that the wavefunction

!li(x, t) = !lio(x, t) exp (i ;~ l dx' . A (X')) (6)

solves the Schrodinger equation including the vector potential A. Here r ==

r (x) characterizes an arbitrary curve in three-dimensional space which ends
at point x. Let us investigate the momentum operator acting on a wavefunc-
tion of type (6). Now the gradient also acts on the phase factor: to be more
precise, it acts on the upper boundary of the integration. We get

V eXP(i:Cl dX'.A(X')) =i:CA(X)exP(i;cl dX'.A(X')) . (7)

Some care has to be taken when performing this step; this transformation
only holds when the curl of A is vanishing. (We convince ourselves with a
simple counter example. A homogeneous magnetic field in the z direction
B = B· e z = V x A is achieved A = ~Brecp; we choose the integration path
in the radial direction, so that e r . ecp = 0 and the path integral vanishes.
Then (7) cannot be fulfilled.) Only if the curl of the vector potential is equal
to zero is the path integral in depend end of the chosen path. With the help of
(7) we find

exp (i:C l dx' . A (x ') )

x [( -ifiV - ~A)!lio - ifi~A!lio] (Sa)

and

(-ifiV - ~Af!li exp (i;c l dx'· A (x'))

x [( -ifiV - ~A) 2!lio + ( -ifiV - ~A) ~A!lio
- iii ~: A (( -iliV - ~A) !lin + ~A!lio) ]
exp (i;c l dx'· A(X')) (-ifiV)2!lio, (8b)

so that (5) reduces to

iii :t!lio = (2~ (-ifiV)2 + eAo) !lio . (9)

According to (6) the presence of a vector potential leads to a change in the

phase of the wavefunction.
All physical observables depend only on the absolute square of the wave-
function !li. At first, the additional factor from (6) seems to have no signifi-
cance. But if two amplitudes superimpose coherently to a total wavefunction,
 1.5 Coherent States (Glauber States) of Electromagnetic Fields 23

detector Fig. 1.6. Interference pat-

tern behind two slits

phase factors are measurable. A diffraction experiment represents an ideal

example. If particles, such as electrons, are projected onto an inpenetrable
wall with two slits (see Fig. 1.6) an interference pattern arises for the prob-
ability distribution of the transmitted particles. The in-going beam can be
described by a plane wave, which has the same phase at both slits. Then,
strictly speaking, the Schrodinger equation should be solved with iJt = ce ip .x
for x < 0 with the boundary condition iJt = 0 at the wall. A simpler way to
explain the interference pattern is to apply Huyghen's principle. After passing
through the slits the two partial waves superimpose linearly and coherently
at the detection screen:
(10)
The phases depend on the travelled distances,
_ 0 P Si
'Pi = 'Pi = r/i = ;: , (11)

where X = lilp represents the de Broglie wavelength divided by 27r. The

maxima of the interference pattern appear for phase differences of ('P2 - /2 'PI)
in
liJtl 2 Icl 21ei<Pl + e i<P21 2
ICl21exp [i ('P2 ; 'PI)] + exp [-i ('P2; 'PI)] 12
41 12
C cos
2 'P2 -
2
'PI ' (12)

which are integer multiples of 7r. Assuming a large distance between the screen
and slits, L » d, we approximate the angle of diffraction by tan j3 = YI L ;::j
5sId (see Fig. 1.6), so that
58
0_
O'P = 'P2 - 'PI = "hp 5s = "hp LYd = X Yd
= XL ; (13)
compare with Fig. 1.6. For the interference maxima we then find
L
Yn = n27rX d . (14)
24 1. Quantum Theory of Free Electromagnetic Fields

Example 1.6. Because of the finite width of the slits a small destructive interference leads to
a decrease in intensity for the interference maxima; the interference pattern
is shown in Fig. 1.6.
What is going to change if a vector potential A (x) is introduced into
the diffraction experiment? According to (6) new phase factors come into
play. It seems plausible to identify the path from the source through the slits
to the point x on the screen as the integration path r (x). Different paths
n r
(x) and 2 (x) belong to the two slits and correspond to different partial
wavefunctions !Ii! and !Ii2 ; see Fig. 1.7. Then the new phases result in

ipi = ip? +~ lri dx'· A(x') , (15)

Fig. 1.7. Two paths are

possible for the wave prop-
agation from the source
through the two slits to ~_--....--.,.detector
the detector. An addi-
tional phase difference oc-
curs because of the pres-
ence of an A field; it cor-
responds to the path in-
tegral 1 A . dx along the
closed path r = n - r 2

so that the phase difference becomes

Dip = Dip° +~
fie
(J, -J, )
r2 r1
dx' . A (x') (16)

Now we realize that the difference between two path integrals can be written
as an integral over the closed path r. Furthermore, applying Stoke's theorem,
we get

i dx' . A (x') = J dF . (V x A)

J dF· B (x') = cJ>. (17)

This is just the magnetic flux going through the surface surrounded by the
curve r r r
= 2 - l . The phase difference

Dip = Dip° + ~ cJ> (18)

changes by an amount that is proportional to the magnetic flux inside the
two possible trajectories (slit 1 or slit 2) of the electron. Only if the curl of
A vanishes will there be no phase difference, Dip = Dip°, in every part of this
region and the vector potential will not influence the observed interference
pattern. Such an A field really has no physical significance. Nevertheless, we
could think of an experiment for which B = V x A =I- a does not vanish in
a tiny region behind the two slits. Everywhere else the magnetic field is zero;
see Fig. 1.8.
 1.5 Coherent States (Glauber States) of Electromagnetic Fields 25

Fig. 1.8. A magnetic field

B = V x A =F 0 is intro-
duced within a tiny region
(hatched) in between the
two beam paths. In this
wayan additional phase
difference between the two
beams occurs, which re-
sults in a displacement
of the interference pattern
(dashed)

This is easy to achieve. Think of a very long solenoid perpendicular to the

scattering plane. A homogenous magnetic field
47r
B = -vle z (19)
c
prevails in the interior of the solenoid carrying a current I and wound with
v turns per unit length; this follows directly from V x B = (47r/c)j. In the
exterior region of the solenoid the magnetic field is negligibly small as the field
lines close at infinity. The vector potential corresponding to an ideal solenoid
with radius R reads

~re<p for r::; R

A (r) = { 27rR (20)
it Ie
27r r <p
for r > R;
e<p is the azimuthal unit vector in polar coordinates. A<p = A· e<p is illustrated
in Fig. 1.9. It is straightforward to confirm, that V x A = 0 in the exterior
region and that the magnetic field takes on the value B = (¢ / 7r R2) in the
interior region, where cP = j B . dS = B7r R2 holds for a homogenous magnetic
field of the solenoid.
Because of (3) the vector potential is only determined up to a gauge trans- c~ A

formation; however, due to the presence of a magnetic flux, it is not possible

to find a gauge transformation such that A (x) becomes zero everywhere in
the exterior region. In the set-up illustrated in Fig. 1.8 the presence of the
magnetic flux tube leads to a displacement of the interference pattern. The
new maxima now lie at

(21)
A detailed analysis shows that the envelope of the interference pattern does
not change. This shows that on average the electron beam is not deflected
Fig. 1. 9. Illustration of
by the localized B field; only quantum-mechanical interference effects are the vector potential A =
influenced. Nevertheless the result seems to be stunning. Although the elec- {O, 0, A",} of an ideal sole-
tron beam encounters no magnetic field at any place, its presence results in a noid
26 1. Quantum Theory of Free Electromagnetic Fields

Example 1.6. displacement of the interference pattern. Referring to the concept of a local
interaction, which we have good reasons to believe in, we come to the con-
clusion that the vector potential A(x, t) plays a more fundamental role than
the magnetic field B(x, t). On the other side because, of the gauge freedom,
A(x, t) contains too much information. The only true and decisive quantity
remains the phase integral exp [i ;c Ir
dx l . A (Xl) ].
.Q; .QJ
I 2
The experiment sketched in Fig. 1.8 was performed in a slightly modi-
\ fied form soon after the proposal by Aharonov and Bohm. 5 Instead of two
\ I
\ I slits a biprism was used to achieve larger intensities. Such a biprism consists
\ I of a thin electrically charged wire between two grounded plates. It deflects
\ J
\ I the beam as if it came from two spatially separated sources Qi and Q~; see
Fig. 1.10. Chambers 6 added a thin ferromagnetic metal fragment, a so-called
whisker, behind the biprism and was then able to detect the change in the in-
terference pattern. Somewhat more elegant was the experiment of Mollenstedt
and Bayh. 7 With a clever electron optical device they succeeded in separat-
ing the two partial beams by a distance of about 60 11m. A freely suspended
air solenoid of 5 mm length and 7 11m radius (!) was introduced into the
Fig. 1.10. Electrostatic gap. Varying the current through the solenoid, they were able to displace the
biprism interference pattern continuously; the envelope remained unaltered. The dis-
placement was in full accordance with the expected result (21). Although the
Aharonov-Bohm effect follows naturally from quantum mechanics, its exis-
tence had been questioned repeatedly. New experiments 8 use small toroidal
magnets instead of solenoids; one of the two beams goes through their open-
ings. In this manner the magnetic scattering field can be eliminated. In addi-
tion the torus was covered with a superconductor to exclude any penetration
of the electrons into the magnetic region. None of these precautions changes
the experimental results.
The Aharonov-Bohm effect (the influence of the potential on the phase
of a wavefunction) does not only hold for magnetic fields. In fact, a relation
analogous to (6) can be derived for the case of electric fields. The presence
of a spatially constant but time-dependent electrostatic potential Ao(t) alters
the phase of a wavefunction according to

tlf(x,t) = tlfo(x,t)exp [-i(e/h) jt dtIAo(t l )] • (22)

Because E = - V Ao = 0, no electric field is present. Again, the appearance

of such a phase factor can be demonstrated with an interference experiment.
As sketched in Fig. 1.11 an electron beam is split, guided through two
metalic boxes, which serve as Faraday cages, and finally brought to interfere.
With a suitable device the beam is cut into separate wave pieces, which are

5 In fact the experiment was proposed 10 years earlier by W. Ehrenberg and R.E.
Siday, Proc. Roy. Soc. 62B (1949) 8, but at that time it did not raise much
interest. Independently, Aharonov and Bohm developed the idea and analyzed it
on more profound theoretical grounds.
6 R.G. Chambers: Phys. Rev. Lett. 5 (1960) 3.
7 G. Mollenstedt, W. Bayh: Naturwiss. 49 (1962) 81.
8 Osakabe et al.: Phys. Rev. A 43 (1986) 815.
 1.5 Coherent States (Glauber States) of Electromagnetic Fields 27

Fig. 1.11. Sketch of an

experiment that demon-
strates the dependence of
the phase of a wavefunc-
tion on a constant but
temporally changing elec-
tric potential Ao( t)

shorter than the length of the boxes. As soon as the wave is inside the box, a
voltage U is applied for a short time. The phases then become

'Pi = 'P? - ne lt2 dtAoi (t)

tl
, (23)

so that

(24)

is obtained for the relative phase. In order to have as much similarity as pos-
sible to (18) for the magnetic field case, the relation E = -V Ao has been
employed and an integral over a spatiotemporal region has been introduced.
The electric flux defined by the double integral of (24) is gauge invariant as in
the case for the magnetic flux of (17). The electric flux also leads to a trans-
lation (shift) of the interference pattern. This holds although the electrons
are inside the Faraday cages and thus do not encounter any E field as the
potential is switched on. On the other side the electron wavefunction is spread
all over space with, of course, different strengths.
From these considerations about the Aharanov-Bohm effect we conclude
that in quantum mechanics the interaction of a charged particle with an elec-
tromagnetic field results in a phase factor

Rl2 = exp [i~ /2 (~A' dx - Ao dt)] = exp (i;c /2 Ap dX P ) . (25)

This contains the potentials Ao and A as fundamental quantities and not the
field strengths E and B. The factor Rl2 specifies the probability amplitude
for a particle to propagate from space-time point (Xl, h) to point (X2' t2)'
The phase factor depends on the trajectory X (t). In principle, all possible
trajectories have to be considered, each of them with its own phase factor.
28 1. Quantum Theory of Free Electromagnetic Fields

Example 1.6. In fact, this conception provides the basis for an alternative, yet equiva-
lent, formulation of quantum mechanics; i.e. the "path integral" approach 9 to
quantum mechanics.
Finally we now show that the expression (25) also incorporates "macro-
scopic effects" of the electromagnetic field as the deflection of a charged par-
ticle due to the Lorentz force. Again we concentrate on the case of the simple
two-slits experiment and on small deflection angles. The considerations leading
to (19) and (21) are repeated with the additional assumption that the elec-
trons run through a sector of width D, which is penetrated by a homogenous
magnetic field B; see Fig. 1.12. The magnetic flux determines the deflection;
it is given by

<P = f dF . B ~ dD B . (26)

Fig. 1.12. In the region

of width D behind the
two slits a magnetic field
B acts on the electrons,
I·. ••

which not only leads to Q ~_......._ _d~I,..· .

I.-.::
a shift of the interfer-
ence pattern but also to a
shift of its envelope. This
is caused by the Lorentz
force, which acts on the
electrons as they move
through the B region
D
III
L

From (21) the shift of the interference pattern is obtained as

L1Y = !:x.e<P = DL eB . (27)

d lie pc
In a classical argumentation, the Lorentz force F = (elc)vB acts during the
short period T ~ D Iv; according to Newton's equation of motion this force
results for small deflection angles in a momentum Py = T(elc)vB in the y
direction. The deflection angle then is tan;3 = pylp ~ LlY I L. The deflection
due to the Lorentz force becomes

LlY = L Py = DL eB . (28)
P pc
This proves (in first order in (3) that the phase factor (25) includes the ac-
tion of the Lorentz force. A more detailed analysis would show that the total
probability distribution is shifted as sketched in the Fig. 1.12; this is in con-
trast to the Aharanov~Bohm effect, for which the envelope of the interference

9 See for example R.P. Feynman, A.R. Hibbs: Quantum Mechanics and Path Inte-
grals (MacGraw-Hill, NY 1965) and Chap. 13 of this book.
 1.6 Biographical Notes 29

patter remains unchanged, indicating that on average no action of a force has Example 1.6.
occured. 10

1.6 Biographical Notes

PAULI, Wolfgang, 'Vienna 25.4.1900, tZiirich 15.12.1958, professor at the

"Eidgenossische Technische Hochschule" (Federal Technical University) in
Zurich since 1928. Pauli was a student of Arnold Sommerfeld and Max Born.
In 1945 he was awarded the Nobel prize for the discovery of the exclusion prin-
ciple that carries his name. He also developed the first theory of the electron
spin, which led to the Pauli equation.

10 As a supplement to our presentation we recommend the short treatise by M.

Danos: Amer. Journ. of Physics Teachers 50 (1982) 64 on the Aharanov-Bohm
effect.
2. Interaction of Electromagnetic Fields
with Matter

In most cases we can think of matter as being build up of particles of mass

mi and charge ei. The interaction of these particles with each other can often
be described in terms of a potential V( ... , Xi,"" Xj, .. . ). For example, for
the case of Coulomb interaction this potential would be
1", e·e·
VCoulomb( ... ,Xi, ... ,Xj, ... ) = -~I t J I' (2.1)
2 .. Xi - Xj
',J
i"/;j
Such a modeling of many-particle systems is widely used in atomic, molecular,
and solid-state physics (atoms, molecules, and their interaction), in nuclear
physics (protons and neutrons and their interaction), and also in elementary
particle physics (quarks and their interaction). The Hamiltonian for such a
many-particle system for nonrelativistic particles reads

,
Hmp " Pi 2
= '~--+V. (2.2)
i 2mi
The index 'mp' stands for 'many particles'.
How does this many-particle system interact with electromagnetic fields?
From the lectures on quantum mechanics' we remember that it is necessary
for a quantum theory to remain gauge invariant. The interaction between ra-
diation and matter has to be in such a way that gauge transformations of
the form (1.4) do not change the observable quantities predicted from theory
(eigenvalues, expectation values, transition probabilities, and so on). This im-
portant requirement for gauge invariance is fulfilled with the minimal coupling
Pi
,
---> Pi - -ei A' ( Xi·
A ) ( 2.3 )
c
Together with the Hamiltonian for the radiation field, (1.40) and (1.48), the
Hamiltonian for the many-particle system (2.2) then becomes

H = L (Pi - 7 A(Xi)) (Pi - 7 A(Xi)) + V

2mi

J
i

E . E* + 11 . i1*
+ 8~ d3 x (2.4)

----
Hmp + Hrad + Hint.
I W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,
Heidelberg 1994).

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
32 2. Interaction of Electromagnetic Fields with Matter

This is the Hamiltonian for the total system "matter + radiation". The in-
teraction operator Hint for interaction between the many-particle system and
the electromagnetic field follows from calculating the square of the expression
inside the sum of (2.4),

,
Hint = """ [ei"
--Pi' A(Xi) + -e;,
- A2 (Xi) ] . (2.5)
L i m·c 'l,
2mc 2 1.

Here, the so-called Coulomb gauge V . A = 0 has been employed, which allows
us to write p . A + A . P as 2p . A. In the following, we suppress the sum and
the corresponding index i for simplicity, so that
, e,' e2 , 2
Hint = - mcP' A(X) + 2mc2A (X) (2.6)
represents the interaction of one particle with the radiation field. We will come
back to the explicit sum (2.5) when necessary. Now we separate the interaction
(2.6) into a term proportional to A and another term proportional to A2,
+ Hint
, , I ' "
Hint = Hint , (2.7)
and deduce further with (1.31) that

(2.8)

and
H' /I
(a) ~nhw' int
hw ~

(b)K
~ jk'

i (2.9)
n
hw The interaction Hint between the electromagnetic field and the many-particle
~

k hw system will remain small. Therefore we will treat it as a perturbation. Hence,

~
the unperturbed system consists of the many-particle system and the radiation
k'
i,f
field, Le.

Fig. 2.1a,b. Absorption flo = Hmp + Hrad (2.10)

and reemission of a pho- with state vectors
ton (light scattering): (a)
inelastic photon scatter- Imp + rad) = Imp)! ... nka ... )rad . (2.11)
ing (Raman scattering),
(b) elastic photon scatter- Here, mp represents the quantum numbers of the many-particle system. Later
ing. For the process in (a) we will replace Imp) by la), indicating that we consider atoms specifically.
both energy (tu..; ---+ tu..;') The interaction (2.7) will generate transitions between the states (2.11). We
and momentum (hk ---+ will calculate the transition amplitudes according to formulas known from
hk') change, whereas for
the process in (b) only quantum mechanics.
momentum (its direction) A few more words about the interactions (2.8) and (2.9) are in order.
changes Since H{nt contains only creation operators (it and annihilation operators (i
 2.1 Emission of Radiation from an Excited Atom 33

for photons linearly, HInt induces transitions for which one photon is either
created or annhilated. On the other side HI~t induces transitions where two
photons are involved. For example the processes (at a, aa t ) describe absorption
and reemission of one photon or vice versa (see Fig. 2.1). We will demonstrate
this point further with various examples.

2.1 Emission of Radiation from an Excited Atom

An atom is in the initial state lai) and decays into the final state laf) by
emitting a photon with wavevector k and polarization 0". Subscripts i and f
stand for "initial" and "final", respectively. The initial and final states of the
total system "atom + radiation field" are denoted by Ii) and If), respectively,
and, more explicitly, are given by

Ii) lai)I··· nko- ... ) ,

(2.12)
If) laf)I·· .nko- + 1. .. ).
They are graphically depicted in Fig. 2.2.

Fig. 2.2. Illustration of

lad I ... ,nko-, ... ) the initial and final state
energy of the total system "atom
of the
+ radiation field" . For
- - - ladl ... , nko- + 1, ... ) atom
the initial state the many-
particle system occupies
the state lai) and the ra-
diation field contains nko
photons characterized by
(k,G"). For the final state
In perturbation theory the transition probability per unit time reads the many-particle system
(Fermi's golden rule)2 occupies the energetically
lower state laf) and the ra-
( tran~. Prob.)
time
(2.13a) diation field contains one
additional photon with
quantum numbers (k,G")
Mfi

where I and II describe a complete set of states. Here, Mfi represents the
transition matrix element. In our case the contribution of first order to Mfi
results merely from HInt and we get with (2.8)

2 W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1994).
34 2. Interaction of Electromagnetic Fields with Matter

(2.14)

In the second step we have taken into account that only the term with the
creation operator at,
contributes within the infinite sum. Equation (1.57)
has been employed for the last step. The 8 function appearing in (2.13a)
guarantees energy conservation:
Ef - Ei = Ear + nwk - Eai Ear - Eai + nwk = O. (2.15)
~ '-v-'
=E f =Ei

The photon with energy nwk takes away the energy Eai - Ear from the atom.
With (2.13a) and (2.14) we obtain the transition probability per unit time
for the transition of the atom from lai) to laf) by emission of a photon in
first-order perturbation theory:

( trans.. prob.) = -.;-

27l'1
UIH{ntli) 12 8(Ef - E i )
A

tIme emission It

27l' ( e ) 2 (27l' he 2 )
Ii: me L3 Wk (nka + 1)
x l(aflp'E:kae-ik,xlai)12
x8(Eaf - Eai + nwk) . (2.16)
The factor nka + 1 in the radiation matrix element has deep and far-reaching
consequences. The probability for emission is not solely proportional to the
number of photons nka already contained in the initial state; this would be
stimulated emission alone. Emission also exists if initially no photon is present
at all (nka = 0); this is known as spontaneous emission. Once again, the in-
duced emission is proportional to the number of photons nka already present
in the radiation field. In some way they shake the excited state and stimulate
the transition. It is the zero-point oscillations of the radiation field that effect
the transition because of their "shaking". These zero-point oscillations are of
pure quantum-mechanical origin and are always present. The 1 appearing in
nka + 1 results from the commutation relations (1.50); this is for the same
reasons as the term ~ appears in (1.52) for the zero-point energy of the radia-
tion field. It is really remarkable and very satisfying indeed that the theory we
have developed so far automatically includes spontaneous emission. We will
 2.2 Lifetime of an Excited State 35

first concentrate on the spontaneous emission process. Later on we will come

back to the induced emission in connection with the absorption process.

2.2 Lifetime of an Excited State

An excited state may decay via spontaneous emission of light. If no photon
is present at the beginning, we have to set nka = 0 in (2.16) and then we
can calculate the transition probability per unit time for this process. We also
sum over all directions k and polarizations u of the photon. The lifetime T of
state lai) is defined as

n: L 1UIH[ntl
211" """"'
k,rr
' i ) 12 J(Eaf - E a, + Wk)

(2.17)
We investigate the limiting case L3 -+ (X) for this equation. The volume of the
box to which the electromagnetic field is confined goes to 00; the electromag-
netic field then occupies the whole space. For this limiting case we will show

f
that

Lk
->
L3
(211")3 d 3k (2.18)

becomes valid. For this we consider the three-dimensional lattice space with
axes nx,ny,n z ; ni E Z. Every lattice point {nx,ny,n z } represents a normal
mode of the electromagnetic field with a certain polarization.
Every vector n = {nx, ny, n z } represents such a normal mode; each vector
begins at the origin. Into a lattice volume t:mx 6.ny 6.n z , which reaches from
nx to nx +6.n x , ny to ny +6.ny and n z to n z +6.n z , fall exactly 6.n x 6.ny 6.n z
arrow heads (lattice points). Therefore, exactly 6.n x 6.ny 6.n z electromagnetic
normal modes exist in this volume. According to (1.30)
Lki = 211"ni (i = 1,2,3 for x,y,z) (2.19)
Fig. 2.3. All normal
and thus
modes n are counted as
L 6.k i 211" 6.ni , (2.20) the number of points in-
side the box with volume
from which llnx llny llnz
6.n x 6.ny 6.n z
L3
(211")3 6.k x 6.k y 6.k z (2.21 )

follows for the number of normal modes falling into the interval [ki' k i + 6.kiJ.
In the limiting case L -+ (X) and 6.k i -+ 0 (2.21) goes over to
L3
(211")3 d 3k (2.22)
from which (2.18) follows.
36 2. Interaction of Electromagnetic Fields with Matter

We return now to the determination of the lifetime (2.17) and choose the
two polarization vectors eka as depicted in Fig. 2.4. It follows that
L l(aflp'ekae-ik'''lai)12 I(af Ip. ek
A
2 e -ik'''1 ai )1 2
a=I,2

(2.23)
Light emitted from atoms typically has energies of Iiw ~ 10 eV. We conclude
that
Fig. 2.4. The polarization
vector ekl points into the 27l" Iiw
k·x ~ TaBohr lie aBohr
plane of the page; ek2 falls
into the plane given by p 10eV ~ x 1O- 8 cm
and k and is of course per-
pendicular to k. The kz 1.97 x 10- 5 eV cm
coordinate axis is chosen 2.7 x 10- 3 « 1 (2.24)
to be parallel to the vec-
tor (afljllai); hence, {) rep- and expand the exponential appearing in (2.23) into a Taylor series:
resents the polar angle in
k space
e-ik-x = 1 - ik . x - ~(k . X)2 + .... (2.25)
Because of (2.24) we restrict the expansion to the first term only; thus (2.23)
becomes
L l(aflp'ekae-ik-xlai)12 ~ !ek2·(aflpl a i)1 2 . (2.26)
0"=1,2

For reasons that will become clear in a moment this approximation is called
the dipole approximation. The next terms in expansion (2.25) would lead
to magnetic dipole radiation, electric quadrupole radiation, and so on. The
quantum-mechanical theory of multipole radiation represents an important
concept in nuclear physics; see the specialist literature. 3 Looking again at
Fig. 2.4 we can further simplify expression (2.26) into

Ilek2il(aflplai)lcos(900-19f = l(aflplai)12sin219. (2.27)

Using also (2.23) and (2.22), we find that the lifetime (2.17) now becomes

( ~)
T i
-;
f
= ~
27l"m
J d3 k ~ I(aflp lai)1 2 sin2 {) 8(Ear -
Wk
Eai + liw k ) .
(2.28)
We introduce spherical coordinates in k space and set the k z axis along the
direction of the vector (aflplai) (see Fig. 2.5). Then it follows that
d3 k k 2 dk sin {) d{) d<p

(2.29)

Fig. 2.5. Spherical coor- In the last step, the integration J d<p = 27l" has been performed because ex-
dinates in k space pression (2.28) does not dependent on <po Thus we deduce
3 J.M. Eisenberg, W. Greiner: Nuclear Theory, Vo!' 2. Excitation Mechanism of the
Nucleus, 3rd ed. (North-Holland, Amsterdam 1975).
 2.2 Lifetime of an Excited State 37

e2 27r
27rm 2 c3
J dWk Wk !(af!p !ai)!
"2

X o(Ea, - Ea. + !UJJk) 1'" sin 3 1} d1}

~ m~:3liWfi !(af!fJ !ai)!2 (2.30)

with Wfi = (Ea. - E a,) / Ii as the transition frequency. The matrix element
(af!fJ !ai) can be cast in another form, i.e.
dx
(af!fJ !ai) (af!mill!ai)
i A A

-fim(af!xHmp - Hmpx!ai)

-i; (Ea. - Ea,) (af!x!ai)

-im Wfi (af!x!ai) , (2.31)
so that (2.30) becomes

( ~)
T i ..... f
= 4e:~!i !(af!x!ai)!2
3nc
. (2.32)

The Hamiltonian Hmp from (2.2) represents the unperturbed many-particle

system and has been used in (2.31) for the operator differentiation
dx i
ill =
A

-fi[X' HmpJ- .
Applying the trick (2.31), we recognize that the momentum matrix element
e(af! fJ !ai) can be expressed as a matrix element of the dipole operator ex. By
now the designations dipole approximation and dipole radiation introduced
earlier should have become clear.
Yet another form can be derived for the expression (2.32) of the lifetime
once Heisenberg's equations of motion are applied twice,
d2 x dfJ
(af!m dt 2 !ai) (af!ill!ai)
i A A

-fi(af!fJHmp - HmpfJ!ai)

-k(Ea. - Ea,)(af!fJ!ai)
i dx
-fi(Ea• - Ea,)m(af! ill !ai)

(-k) (-k) (Ea. - Ea,)m(af!xHmp - Hmpx!ai)

- ; (Ea. - EaY(arlx!ai) ,

(2.33)
38 2. Interaction of Electromagnetic Fields with Matter

Finally this yields for (2.32)

( fiwfi)
7 .I-+f
= 4e21{afld2Xlai)12
3c3 dt 2
(2.34)

This relation can be regarded as the quantum-mechanical analogue to the

Larmor equation; it states that the energy radiated by an electron accelerated
nonrelativistically is given by

energy _ 2e21 d2 x 12
(2.35)
time - 3c3 dt 2
We illustrate the theory developed so far with a couple of examples.

EXERCISE

2.1 Selection Rules for Electric Dipole Transitions

Problem. Show that the selection rules for dipole transitions are given by
1:11 = ±l and I:1m = ±1,0, where I and m represent the angular momentum
quantum numbers of the electron.

Solution. With spherical coordinates the wavefunction of the electron takes

on the form

The matrix element describing electric dipole transitions is given by

M rv Jd3r1P~'l'm,(r)r1Pnlm(r).
To begin with we express the components of the position vector in terms of

If ·
spherical harmonics. We need

- - sin {) el'P

If .
Yl1 CI9, rp)
81T '

- sin {) e -1'1'
81T '

YlOCrJ, rp) f3 cos{).

V4;
Addition and subtraction yield

sin {) cos rp

sin {) sin rp
 2.2 Lifetime of an Excited State 39

cos'l9 = ~YlO('l9,cp) Exercise 2.1.

= azYlO ('l9, cp) .

Here ai, bi (i = x, y), and az are constants, which can be read off immediately,
i.e.

ax -b x -~fi
2 3'

ay by -;ifi,
az ~.
We deduce further that
x l' sin 'l9 cos cp r(a xYl1 + bxY1- 1 )
y l' sin 'l9 sin cp r(a Y Yl1 + byY1- 1 )
z l' cos 'l9
Referring to the matrix element M above, we multiply these spherical har-
monics further with Yim('l9, cp). The following integrals appear:

J dS? Yi;m' ('l9, cp )Y1,±1 ('l9, cp )Yim( 'l9, cp)

and

J dS? Yi;m' ('l9, cp )YlO ( 'l9, cp )Yim( 'l9, cp)

with
dS? sin'l9 d'l9 dcp .
Generally

J dS?Yi:m3('l9, CP)Yi2m2('l9, CP)Yi1ml ('l9, cp)

= ((2 h +( 1?(21 2
47r 23 +1
r1)) 1

2 (1112 13Imlm2m3)(1112131000)

holds. The Clebsch-Gordan coefficients (hI2blmlm2m3) vanish except for 4

1) ml+m2=m3,
i.e. m ± 1 = m' or Lim = m' - m = ±1,
for the x and y components
and m + 0 = m' or Lim = m' - m = 0,
for the z component.

4 See e.g. W. Greiner: Quantum Mechanics - Symmetries, 2nd ed. (Springer, Berlin,
Heidelberg 1994).
40 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.1. 2) The Clebsch-Gordan coefficient (111'1000) is only nonzero if I + I' + 1 is an

even number and if I and I' differ at most by 1. Therefore I - I' = ±1.

Hence, we get the following selection rules:

t:..1 ±1 } for the x and y components,

t:..m ±1
t:..l
t:..m
±l}
o for the z component.

For all other cases the angular integrals vanish, so that the matrix element is
zero.

EXERCISE

2.2 Lifetime of the 2p State with m = 0 in the Hydrogen Atom

with Respect to Decay Into the Is State

Problem. Calculate the lifetime T of the 2p state with m = 0 of the hydrogen

atom with respect to decay into the Is state. The wavefunctions are given by 5
1 e -ria
--
'IjJ(ls)
V7ra 3 '

'IjJ(2p, m = 0) __1_?:. e-r/2ay'2 cos 1'J

8V7ra3 a '
where a = 1i,2/ me 2 denotes the Bohr radius.

Solution. We remember that

1 4 e 2 Wl3s 2p
Ii, 1(2p,m
T 3 c3
where

nwl 2
s p
= EI -
s
E2
P
= me 4
21i,2
(1 _~)
4
= ~8 ea2
has been used. In spherical coordinates the unit vectors are given by e±1 =
=f ~(ex±iey) and eo = e z . As a consequence, for m = 0 only the z component
of the matrix element of r does not vanish. Thus, it follows that

1
T
4(3)3 :3 (2)4
3 8" ~ 1(2p,m = 01zlls)1 2
This matrix element becomes

5 W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1994).
 2.2 Lifetime of an Excited State 41

M (2plzlls) Excercise 2.2.

J d 3 r1j;*(2p)z1j;(ls)

_1_
87ra 3
J~ a
e- r / 2a v'2 cos '/J(r cos '/J)e- r / a r2 dr sin'/J d'/J dcp .

The integration over cp is simple:

M = v'21°O r dr e 171" cos2 '/J sin'/J d'/J

-4
4a 0
4 -3r /2a
0

2 V21°O
- - err d -3r/2a 4
34a4 0

4v'2 (~r a,
so that
1
T

or

T = (3)8
2
a (hc)4
c e 2
(3)8
2
a
ca 4
Inserting numbers yields
T ::::! 1.6 X 10- 9 s .

EXERCISE

2.3 Impossibility of the Decay of the 2s State

of the Hydrogen Atom via the p . A Interaction

Problem. Show that the 2s state of the hydrogen atom cannot decay via the
p . A interaction into the Is state by emitting a photon, i.e.

(2slcka . pe- ik .r lls) = o.

Solution. We choose k to be parallel to the z axis, so that

e -ik·r = e -ikz .

Then the matrix element becomes (jJ = -ihV):

(2slc·pe- ik .r lls) = ~ J d3r1j;28e-ikzc.V1j;18.

Since
_ 1 -ria
1j;18 -.J7ra 3 e ,

we deduce that
42 2. Interaction of Electromagnetic Fields with Matter

Excercise 2.3.

and derive
h
(2Slc· pe- 1'k 'Tlls) = -:-
la
J c· r e- 1'k ·z
d 3 r 1/J2s'l/Jis--
r
.

Since k points in the z direction and c is perpendicular to k, the vector c has

to lie in the x-y plane. Then, the scalar product c . r represents the projection
of r into the x-y plane and the integrand factorizes into a part that depends
only on x and y, and into another part depending only on z. Because 1/J2s
and 1/Jls are spherically symmetric, the integrations over x and y vanish. As a
consequence the matrix element vanishes.

EXERCISE

2.4 The Hamiltonian for Interaction

Between the Electron Spin and the Electromagnetic Field

Problem. Up to now we have neglected the electron spin. The magnetic

moment jJ, = (eh/2mc)ir of the electron means an additional spin-dependent
term has to be taken into account for the interaction of an electron with the
electromagnetic field, in addition to the usual terms H' '" p . A and H" '" A2
discussed up to now. Derive an expression for this additional term.

Solution. The interaction energy of a magnetic dipole JL with a magnetic

field B is equal to
Hili = - JL . B .
Now B = rotA = V x A and p = -ihV = hk hold, so that
k -iV or V = ik,
iJ ik x A,
Hili -- - (AkxAA)
ieh '0'.
2mc
Here plane-wave photons have been assumed, as usual. With

AA -_ "
~
NkCku [Aaku (t) e ik·r + aAtku (t) e -ik.r]
k,u

H"' becomes
H"' = - ieh "N
2mc ~ k
ir. (k xc
ku
) [a ku eik . r + atku e- ik .r ]
k,u

The operator k= -iV acts on the exponential function:

H"'
A ieh
= - 2mc
- - L NO'· (k xc) [a
k ku ku
e1'k · r - t e- 'k]
aku r1 ·

k,u

Here k is no longer an operator; the minus sign in the second term results
from differentiation of the exponential.
 2.2 Lifetime of an Excited State 43

EXERCISE

2.5 Lifetime of the Ground State of the Hydrogen Atom

with Hyperfine Splitting

Problem. The magnetic interaction between the spin of the electron and the
nuclear spin splits the ground state of the hydrogen atom into two levels: one
with total spin 1 and one with total spin O. The photon, which is emitted
from the transition between these two states, comes with a wavelength of
>. : : : 21cm. Use the results of Exercise 2.4 to calculate the lifetime of this
transition. For the calculation of the lifetime you may choose one particular
state of the degenerate spin-l triplet. Why?

Solution. The spin of the electron 1p'I)e and that of the proton 1P,2)p couple
to the total spin wavefunction

18p,) = l: (~~81P,1P,2P,) 1p'I)e 1P,2)p .

1-'1,1-'2

8 takes the values 8 = 1 (triplet) or 8 = a (singlet). (j122jlmlm2m) denote

the Clebsch-Gordan coefficients. 6 As the initial state we choose
Ii) = Ils)e I i)e Ii)nucleusl no photon)rad ;
a Is electron 11s)e comes with spin up I i)e, the nucleus has spin up Ii)nucleus,
and the radiation field is initially represented by the a-photon state.
This initial state has total spin S = 1 and its z projection is also p, = 1.
It is a member of the spin triplet. Since each of these degenerate spin-triplet
states should have the same width (decay time) - because of the rotational
symmetry of the problem - we may restrict ourselves to calculating the decay
of this particular member of the spin triplet. In the final state If) the electron
is still in the Is state, but the spin orientations of the electron and the nucleus
have changed. In addition, one photon exists in the state I· .. , lka, ... )rad.
1
If) = 11s)e V2 (I i)e Il)nucleus -Il)e Ii)nucleus) I···, lka,·· ')rad .

This final state carries angular momentum a for the electron-nucleus system.
The one photon carries the total angular momentum of the state, i.e. 1. The
lifetime is given by the expression
1
T
l: 2
2; iUIif'''li)i 8(DoE - hck).
final states
The 8 function guarantees energy conservation; here DoE = nwk = hcko, with
ko = 21l'j21 em-I, represents the energy difference between the two states,
which is equal to the energy nwk = hck of the emitted photon. Furthermore
(see Exercise 2.4),

6 W. Greiner, B. Muller: Quantum Mechanics - Symmetries, 2nd ed. (Springer,

Berlin, Heidelberg 1994).
44 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.5. HA'" -_ --2-

ien
me
L Nk U ·
(k X cka ) (Aaka eik·r - aAtk e -ik.r)
a
k,a
with N'f = 27rne
2 / PWk. The term including aka annihilates a photon; since
we only calculate the process of creating a photon, we discard this term. With
Wk = ek it follows that

(27rlie _1_ .(k x cka ) aka

1

H'" = ien
2me £
)"2 ""'
3 L
2
At e -ik·r .
rJ:U
k
,a
vck

With c5(ax) = I!I c5(x) this yields

~ (2~er C~~2) ~: L ~ lUlu. (k X cka)a!ae-ik.rli)12

k,a
x lie1 c5(ko - k) .
Again, we employ the dipole approximation
eikor ~ 1
and replace the summation over k for the final photon states with an integra-
tion

~ --+ (2~3)3 / d 3k.

Now k has become a continuous variable; compare again with (2.18). It follows
that

~ = 4~2:2 2~ L (/ d:k lUlu 0 (k x cka)aLlif c5(ko - k))

a

Furthermore
aLii) = 11s)el i)e IT)nucleusl···, lka,·· .)rad ,

so that we derive the matrix element as

M = \ ~((T Ie (1 Inucleus - (1 le(T Inucleus) (Isle (lkall

X 0-. (k X cka)llls)elT)elT)nucleusI1ka)).
The scalar product is
u· (k x cka) = &x{k X cka)x + &y{k X cka)y + &z{k X cka)z.
We use the Pauli matrices

(Jx =
(0 1)
1 0 '
A

(Jy =
(0 -i)
i 0 '
A

(Jz =
(1 0)
0 -1

and deduce
 2.2 Lifetime of an Excited State 45

Because of the orthogonality of If) and Ii) only those configurations contribute Exercise 2.5.
for which the electron switches from the initial state Ii)e to the state Il)e due
to 0-. Hence, only (Jx and (Jy contribute:
1 .
M = - V2 [(k x ck(J")x + l(k x ck(J")Y] .

Initially we have chosen the spin to be along the z axis. In such a system z
it is not convenient to evaluate the cross products appearing in M. Thus, we
rotate the coordinate system in such a way, that the z' axis coincides with
the direction of k (see Fig. 2.6):
x x' cos fJ - z' sin fJ ,
y y' ,
z x' sin fJ + z' cos fJ . x

Via these transformations, Fig. 2.6. Illustration of

the rotation of the coordi-
nate system
(J"

becomes

L ~ {[(k x ck(J")x' cosfJ - (k x ck(J")z' sinfJ]2

(J" (J"

+ [(k x Ck(J")y,]2}

in the rotated coordinate system. The latter has been chosen such that
k = ke z "
so that
(k x ck(J");'
(k x CkG" )z'
(k x Ck(J" )~,
We get

(J"

fJ represents the angle between the spin vector and k. Finally we obtain

~T = ~~
4m 2 c2 211"
(J d k ~k2(1
3
k 2
+ cos 2 fJ)c5(ko - k)) ,

and inserting d 3 k = k2 dk sin fJ dfJ dcp we have the final result:

1 --2J+
e 2 11,-2 -1
4m c 2
(1 cos 2 fJ)k3 dk sin fJ dfJ c5(ko - k)

~~~ J
T

k 3c5(ko - k) dk
4m 2 c2 2 3
~ e 2 11, k 3
3 m2c2 o'
46 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.5. Inserting the value ko = 21r /21 em -1, we calculate a lifetime of
T ~ 3.4 X 10 14 s ~ 10 7 years.

Now it has become obvious that these kinds of transitions are not observable
in a laboratory. However, such photons are well known in radioastronomy.

EXERCISE

2.6 One-Photon Decay of the 2s State in the Hydrogen Atom

Problem. Use the results of Exercise 2.5 to calculate the lifetime of the 2s
state of the hydrogen atom. The spins of the electron and proton couple as in
Exercise 2.5, i.e. to total spin 1. Assume that, because of the spin-dependent
interaction, the 2s state decays into the ground state by emitting one photon.
Later on we will demonstrate, in Exercise 2.10, that the two-photon decay
process is much faster.

Solution. In a manner analogous to Exercise 2.5 we arrive at

1
T

81r:~C2 / d:k k 2(1 + cos 2 19)8(ko - k) 1(lsi e- ik .r I2s) 12 .

Here we have analyzed the spin-dependent part of the matrix element in a

way similar to that in Exercise 2.5, except that the dipole approximation has
not been employed. The spatial part of the matrix element remains; in fact,
this part may be drawn out of the integral:

~ 81r:~c2 ( / k 3 dk sin 19 d19 dcp (1 + cos 2 19))

x 8(k o - k) l(lsle-ik.rI2s)12

~ e2 1i, k31(lsle-ik.rI2s)12
3 m2c2 0
We expand:
e -ik·r = l '- 1 1 .2(k ·r )2 + ...
k ·r+2"1

Since (lsI2s) = 0 the first term of the expansion does not contribute. The
second term leads to the matrix element
M = (lsi k . r 12s) .
The Is and 2s wavefunctions depend only on r. The scalar product yields
k . r = kr cos 19 as the z axis is chosen parallel to k. Thus the matrix element
is proportional to
 2.2 Lifetime of an Excited State 47

M rv i1f sin'!? cos '!? d'!? = 0; Exercise 2.6.

as a consequence, the second term of the expansion does not contribute. There-
fore we start all over with
(lsle- ik .r I2s) ~ -~(lsl(k.r)212s).
The corresponding wavefunctions are given by
1 e -ria
--
1/Jls
va 3 7r
and

1/J2s = 1 ( 2 - - e -ria , r)
4v27ra 3 a
so that the matrix element now becomes
M (lsi e- ik .r I2s)
~ -~Jr2dr-l-e-rla(k.r)2
2 va 3 7r
x ~
4 27ra 3
(2 - .c)
a
e- rla sin'!? d'!? d<p.

Performing the <p integration and using k . r = kr cos'!? gives

M = - -k
--
4V2a 3
2
10
00
r4e-2rla (2 - -)
r dr
a
l1f
0
sin'!? cos 2 '!?d'!?,

_~ roo r4e-2rla (2 _.c) dr,

6V2a 3 io a
1 2
rc;(ka) .
16v2
Because of energy conservation the energy of the emitted photon is
IE2s - Elsi = lieko· We thus arrive at

1 2
T
e2 h
~ -3 ~ko
m c
3(1 rc;(koa) 2) 2
16v2
or

From
3 e2
lieko = -- a
8 a ' lie
it follows that ko = ~ ~ and

~ ~ _1_ (~)7 o:lO~.

T 768 8 a
Finally, using 0: ~ 1/137 and a = 5.292 x 10- 9 cm we get
48 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.6. T ~ 3 X 10 8 S ~ 9.5 years.

The large lifetime underlines that the one-photon decay is strongly suppressed.
This is in contrast to the two-photon decay, which we will encounter in Exer-
cise 2.10.

2.3 Absorption of Photons

For the process of absorption of photons, the initial and final states are given
by
(2.36)
and
(2.37)
lar) respectively. For most cases of interest lai) represents the ground state of the
atom or the nucleus. The final state differs from the initial state in that the
many-particle system is in an excited state and the radiation field has lost
one photon characterized by kl7.
The relevant matrix element now reads

Fig. 2.7. As it absorbes a

photon the many-particle
system changes from lai)
to laf) x lai) I··· ,nku,"')

e
--
mc
J21f hc
-
L3 Wk
2
- ( afP'ekue
1 ik .XI)
A
ai

x (... , nku - 1, ... Iakul.·., nku",,)

- -e
mc
J21fhc2 ( af 1P . eku e ik·x 1ai ) Vr;;:;-:-
-L3
Wk
A
nku ; (2.38)

compare this with (2.14). Here we have assumed again that only H{nt con-
tributes to this process in first order and that (1.56) determines the photon-
matrix element. According to (2.13a) the transition probability per unit time
for the absorption process is given by

(
trans. Prob.) 21f (~)2 (21flic 2 ) nku
time absorPtion- Ii mc L3 Wk

x 1(aflfJ· eku eikxlai) 12 8(Ef - E i ) . (2.39)

Here
Ef Ear + (nku - l)nwku ,
Ei Ea; + nkunwku ,
 2.3 Absorption of Photons 49

so that the 8 function can be rewritten as

8(Eaf - E a, - Wku)
8(Ea, + Wku - E af ) . (2.40)
Inserting this into (2.39) and making use of the relation
(aflp·ekueik.xlai) = (a;!p·ekue-ik.xlaf)*
yields

( trans. Prob.)
time absorption

)2 (211"nc I( aip·eku
I' . af )1 2
2
-211" ( - e - -) nku e- ikxl
n mc L3 Wk
X 8(Ea, + Wku - E af ) . (2.41)

r
A comparison with (2.16) for the emission process shows that the transition
probability for stimulated emission, which is the term in (2.16) proportional to
nku, is identical to the absorption probability per unit time (2.41). Observe that
lai),laf) in (2.16) are the same states as laf) and lai) in (2.41), respectively.
The final state (ground state of the atom) for emission is identical to the
initial state (again the ground state of the atom) for absorption.
~
We will calculate the cross section ai ..... f(ka) for the absorption of a photon ~
with momentum nk and polarization a. It is defined as the transition prob- ~
atom
ability per unit time for the absorption of a photon divided by the incoming Jphoton
photon flux jphoton: Fig. 2.8. Illustration of
nku the absorption process
jphoton = V c. (2.42)

This yields

(2.43)
The incoming radiation will possess a certain frequency spectrum. If we in-
tegrate over this spectrum, the 8 function appearing in (2.43) will disappear
and, instead, the intensity of the absorbed photons from the radiation will
appear in (2.43). Soon we will recognize that the atomic levels (states of the
particle system) are not sharp either; so far we have assumed the opposite,
which expresses itself in terms of the 8 function appearing in (2.43). The
atomic levels have a naturallinewidth; see Sect. 2.7.
50 2. Interaction of Electromagnetic Fields with Matter

EXERCISE

2.7 Differential Cross Section dajdS2 for Photoelectric Emission

of an Electron in the Hydrogen Atom
(Dipole Approximation)

Problem. Consider the photoelectric emISSIOn of an electron out of the

ground state of a hydrogen atom. Assume that the incoming photon has a
very large energy, so that the wavefunction of the electron to be emitted can
be approximated as a plane wave. Furthermore, assume that the photon mo-
mentum is parallel to the x axis and that the polarization vector points in the
direction of the z axis. Make use of the dipole approximation and calculate
the differential cross section for the emission of an electron into the solid angle
element dS2. Neglect the electron spin, because it is irrelevant in this process.

Solution. According to (2.41) the transition probability per unit time is given
by

2; (~J 2 (~:~) nkcr 1(ali> . ckcr eik.rlb) 12 c5(Eb + 1Y..Jk ~ Ea) .

Here Ib) = lIs) and la) is described by a plane wave
eiq .r
la) = 'ljJq = (r Iq) = IT-i'
VL3

The energy of the latter is purely kinetic and is given by

fj}q2
Ea = -~,
2m
so that we arrive at

h27r e )2 (27rlie)
( me L3 nkcr
I(q IAP . ckcr eikrl 2 2m
. Is )1 u, (li2q2 ~ E Is " .. ) ,
~ flJ..Uk
Wk

with the Is wavefunction

'ljJls(r) = (r lIs) = ~ e- r / a .
V7ra 3
The cross section reads [compare with (2.43)]

a(ka) =

L
k
-+ (2:)3 J q2 dqdS2e ,

where d 3q = q2 dq dS2e represents the volume element in q space and dS2e =

sin {) d{) drp the solid angle element for the electron. Thus,
 2.3 Absorption of Photons 51

da Exercise 2. '1.
dD (k, a)

x8 (Ii;~ - E 1s - ~k) .

We have employed the dipole approximation, i.e. k . r « 1, and therefore

Choosing Cka = e z , so that the photon is linearly polarized in the z direction,

one obviously gets
Ii Ii a
-;-V ·e z
t i oz
and

Now
a a
oz cos 1J or '
e2
re
mc2 '
Wk ck,

!
and
£3 li2 1
- r e -k-2 q2 dq I(ql cos1J11s)1 2
21l' m a
li2q2 )
X 8 ( 2m - E 1s - ~ .

Since 8(ax) = (1/lal)8(x) we obtain

8 2;
21i": (q2 _ (E 1s +~))
2m 2 2
ti,28(q - qo)

for the 8 function, where q5 stands for q5 = (2m/li2) (Els + ~). Furthermore,
8(q 2 - qo)
2 1 [8(q - qo) + 8(q + qo) 1 .
= -2
qo
Because q, qo > 0, it is impossible for their sum, q + qo, to become zero; as a
consequence, 8(q + qo) = 0 and
li2q2 ) 2m 1
6 ( 2m - E 1s - ~ --6(q-qo) .
li2 2qo
It follows that
52 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.7. da L3 2
ds:? = 21rreka2
J 2
q dq l(qlcos'l911s)1 2q0c5(q-qo)
2 1

or
da L3 qo 2
d He
n = - re -k 21(qOlcos'l911s)1
21r a
We expand the plane waves into spherical harmonics 7
+1
L L
00

eiQo ·x = 41r i 1jl(qOr)Yim('I9,!P)Yim('I9',!P'),

1=0 m=-l
where 'I9,!p and 'I9',!P' represent the corresponding polar angles of qo and x
with respect to the z axis. We notice also that

cos'19, = ( 41r
3 )! Y (")
'19 , !P .
lO

Then, we deduce for the matrix element

(ql cos '191 Is)
J
M
e 4 +1
~ ~ m~/ _i)ljl(qOr)
DO

d3 x

Y('I9', !p')
1

X Yi::n('I9, !p)Yi::nW, !p') (4~) "2 lO e- r / a

J
or
41r 00 +1
M ~ r 2 e- r / a dr ~m~I(-i)ljl(qOr)Yi::n('I9,!p)

J )
1

X ds:?' Yi::n W, !p') (4~ "2 YlO W, !p') ,

respectively. Because of the orthonormality relation

Fig. 2.9. Illustration of
the geometry for the pho-
toelectric emission of a
J ds:?' Yi:n ('19', !p') Yi'm' ('19', !p') = c511' c5mm' ,
photon the summations break down:

M ~ Jr 2 e- r/a dr [(-i)ljl( qOr)Yi:n('I9,!P) (4~)!] c5l1c5mo .

Since

cos '19 [Yim (4~ ) !] c511 c5mo

Y(4~)!
lO

Y 1*o('I9,!p) (4~) "2

7 W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,
Heidelberg 1994).
 2.3 Absorption of Photons 53

1
we get Exercise 2. 7.

M = - 47ri cos {) r 2d r J1. (qor ) e -ria .

00
~
Y7ra 3 L3 0

Substituting x = ria yields

M -47ri cos {) V/oF

;V f (qoa)
with

f(qoa) 1 00
x 2 dxh(qoax)e- X
2qoa
2 .
(1 + q3a 2 )
With the abbreviations re = e2/mc 2 and a = fi? Ime 2 the scattering cross
section finally takes on the form
dCT L3 qo 167r 2 cos 2 {) a3 f2( )
dst e 27r r e ka2 7r L3 qoa

8a2 (~:) 2 cos2 {) ~ f2(qoa).

The scattering cross section is largest for {) = 0, i.e. in the direction of the
incoming photon. Most of the electrons are obviously kicked out directly by
the photon. Observe, however, that the solid angle dste = sin {) d{) d<p is zero
in this direction; as a consequence the maximum of the electron intensity is
shifted towards angles {) ; o.

EXAMPLE

2.8 Spectrum of Black-Body Radiation

Here we come back to the derivation of Planck's radiation formula, which we

considered at the very beginning of our lectures on quantum mechanics. 8 At
that time we pursued Einstein's derivation; right now we develop an advanced
point of view and explicitly work out some of the assumptions that were
employed during the former approach.
We consider a sample of atoms in thermal equilibrium, which means that
owing to thermal collisions the number of atoms that are excited is equal
to the number of atoms that are deexcited. The number of atoms in state
If) is denoted by Nf and for those in state Ii} the number is N i . Transitions
occur between the two states; photons may be absorbed or emitted from the
radiation field. Already with this background we understand the following two
differential equations:

8 W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1994).
54 2. Interaction of Electromagnetic Fields with Matter

Example 2.8. dNf -N (trans. Prob.) N (trans. prob.)

dt f t'lme abs. + 1 t'lme em.
dNi -N (trans. Prob.) R (trans. Prob.)
dt 1 t'lme em.
+ f t'lme abs.
They express the coupling between the accumulations of both states according
to the radiation. We assume that a coupling of atoms takes place only via the
radiation field and that collisions between atoms do not contribute to the
transitions. Equilibrium requires that
Nf Nj=O
and

Nf = exp ( - tf/r) = exp (_ (Ef - E j ») .

Ni exp ( _ ~ ) kBT

Here T represents temperature and kB is Boltzmann's constant. With the first

two equations it then follows that
Nf (trans. prob.jtime)em.
Nj (trans. prob.jtime)abs.
where we have used (2.41), (2.16), and E j - Ef = nwk. Solving for nka results
in
1
nka = -----,-----,---
exp (a)
-1 .
This is Planck's distribution, from which the radiation energy E rad in the
volume £3 immediately follows:
nwk £3 3
dErad = 2 £3 nka (27r)3 d k,

where k is the wavevector in d 3 k. Furthermore, a factor 2 for the polarization

has been inserted and (2.22) has been used. Now we can define a spectral func-
tion u(w), which describes the radiation energy per volume in the frequency
range between wand w + dw. Making use of (2.29) we obtain

u(w)dw = 1{),<p
1 w~dwk .
2nwknka- (
27r
)3--3-smfJdfJdcp
c
87rhw~ dWk
(27r)3cJ (exp (k:wT ) - 1)
hw 3 1
--z1
7rC
(exp (k~wT ) - 1) dWk . (1)

This agrees with an earlier result derived in a more elementary way. 9

9 W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1994).
 2.4 Photon Scattering from Free Electrons 55

2.4 Photon Scattering from Free Electrons

Free, noninteracting particles are described by the Hamiltonian

, P'2
H- - (2.44)
- 2m·
Its normalized eigenstates are denoted by Iq) and are confined to a volume
L3;

(2.45)

The momentum vectors nq obey periodic boundary conditions at the surface

of L 3 , which is in close analogy to the case of electromagnetic waves [compare
with (1.30)].
The normalization is given by

(2.46)

The energy eigenvalues belonging to (2.44) are

(1iq)2
Eq = --. (2.4 7)
2m
At first, we consider the process of a free electron absorbing a photon. The
initial and final states are given by

Ii) Iqi)I···,nk,,, ... ),

(2.48)
If) Iqf)I.··, nkO" - 1, ... ) .

Transitions between these two states can only be caused by H{nt from (2.8).
We deduce that

UIH{ntl i ) = (qrl(···, nkO" - 1,···1 ( - ~J p. €kO"J ~:~ Fig. 2.10. Absorption and
emission of a photon by
x (akO" ei b: + akO" e- ik .X) Iqi)I···, nkO", ... )
a free electron is prohib-
ited. The photon is indi-
e J21rfic(
-- I'
- - qfP·€kO"e ikxl) ~
. qi ynkO" cated by a wavy line, the
me L3 wk
electron by a straight line.
_~J21rfic
me L3 wk
(-qf· €kO"h/nkO"
The interaction point is
called vertex

X
JL3
e -iQf·X.tk-x eiqi · X d3
L3/2 e L3/2 x. (2.49)

The integral yields

JmL3
e-iqf·X ei(qi+k).X
--
m d 3 x = 6qf ,qi+k . (2.50)

This reflects momentum conservation, i.e.

(2.51 )
56 2. Interaction of Electromagnetic Fields with Matter

Furthermore, another restriction has to be considered for the transition

probability: According to (2.13a) for the transition probability per unit time
conservation of energy

hwk + (fiqi)2 (2.52)

2m
is a result of the 8 function.
It is straightforward to show that the two equations (2.51) and (2.52)
cannot be fulfilled simultaneously. The best way to prove this is to argue
relativistically. We remember that the norm of four-vectors is constant. Also
we simplify the situation with c = 1. For four-vectors it holds that

Pe +)
( P'"( = Pe' , ( Pe + P'"( ) 2 = 2
Pe' = 2
Pe .
Since p~ = m; and P; = 0 it follows that 0 + 2P'"(Pe + = m; m;, so that
P,"(Pe = O. In the rest frame of the electron the four-vectors become
Pe = (me,O),
Hence, it follows that
E'"( = 0;
which means that no photon is present. Therefore it is impossible for a free
electron to absorb a photon. We have to draw the conclusion that processes
of first order caused by H{nt do not exist. As a consequence we investigate
now the processes of first order with H{~t; see (2.9). This part of the interac-
tion contains terms of the form aLak' 0"" Obviously, they describe processes
in which a photon of the kind k' a' is absorbed and another photon kO' is
emitted. Such photon scattering is best illustrated with Feynman diagrams as
in Fig. 2.11.
The full lines describe the electron, whereas the wave-like lines represent
the photon. The graph is to be read from bottom to top in accordance with
the direction of time. The incoming photon kiO'i is absorbed (annihilated) by
the incoming electron. Another photon kWf is produced. In this annihilation-
creation process the electron scatters into the state qf. The initial and final
direction states for this process are
of time
Fig. 2.11. Sketch of a Ii)
(2.53)
process, where an electron If) Iqr) I· .. ,nkjO"j - 1, ... ,nkf(J"f + 1, ... ) .
in state Iqi) absorbs a pho-
ton kjO"j and emits another
photon kWf, so that the The transition from Ii) to If) is caused by the two terms with iL!Wfak;O"; in
electron scatters into state (2.9). According to (2.9) and (2.13a) the transition probability per unit time
Iqf) becomes
trans. prob.)
(
time
2)2 lekfO"f . ek;0";1 2
photon scattering

27r (~)2 (27rlic

Ii 2mc2 £3 Wk;Wkf

X I(qrl ( ... ,nkWj - 1, ... ,nkWf + 1, .. ·1

 2.5 Calculation of the Total Photon Scattering Cross Section 57

A At
x ( akiO"iakfO"fe
i(ki-kr)·x + akfO"fakiO"ie
At A i(ki-kr)'X)

x Iqi) I ... ,nkiO"jl' .. , nkfO"f' ... ) 12

X <5 (hwi + (hqi)2 _ hwf _ (h q f )2) . (2.54)
2m 2m
Owing to (1.56) and (1.57) the matrix element becomes

IMfd 2 = 1(qfl ei(ki-kr).X Iqi) 12

x (y'nkiO"i Jnk{O"f + 1+ y'nkiO"i JnkfO"{ + 1) 2
4nkiO".(nkfO"f + 1) l(qflei(ki-kr),xlqi)12 (2.55)
so that (2.54) results in

( trans. Prob.)
time
2
photon scattering

871" (~)2 (271"hc2 ) le"kfO"f' e"k,O",1 2

h 2mc 2 L3 WkiWkf

X 1(qfle i(ki-kr)·x
Iqi)
12 nkiO"i (nkfO"f + 1)
X <5 (hwk + (hqi)Z _ hwk _ (hqr)2) (2.56)
'2m {2m

.
(q le,(ki-kr),xlq') =
J
The matrix element between the electron states is calculated to be
e-i(qf+kr)'X ei(qi+k.).X
d3 x---==-------::=--
f ' JV JV
= <5qf +k{,ki+qi' (2.57)
It expresses momentum conservation:
h(qf + kr) = h(qi + ki) . (2.58)
Energy conservation is contained in the <5 function of (2.56):
hw . + (hqi)2 = hw + (hqr)2 . (2.59)
k, 2m k{ 2m
The occurence of the factor nkfO"f + 1 in (2.56) is interpreted as stimulated
emission of photons kWf in the final state; apparently the scattering is en-
hanced if photons kflJ"f of the final state are already present. Often this is not
the case, so we set nk{O"f = 0.

2.5 Calculation
of the Total Photon Scattering Cross Section
To calculate the total photon scattering cross section IJ"totai, we sum (2.56) over
all the final states of the photon as well as the final states of the electrons; this
result is then divided by the incoming photon flux nkiO"c/ L3. With nkfO"f = 0,
this yields
58 2. Interaction of Electromagnetic Fields with Matter

O"total

x8 (nw . + (liqi)2
k,2m
_ nw _ (liqr)2)
2m k, . (2.60)

If we employ the Kronecker delta function qr = qi + k i - kr, the sum over qr

breaks down. Furthermore, we make use of the relation

~ L~
""'3
-->oo
L3
(271")3
J d3kr ,

which is proven after (2.18), and get

O"total = li~4 L Jd kr ICk,O", . ckiO"i 12

3
m C 0", WkiWk,

X 8 (hc(k i _ kr) + (liqi)2 _ li2(qi + ki - kr)2) (2.61)

2m 2m
From the two conservation laws, (2.58) and (2.59), it follows that the wave-
length of the scattered photon differs from that of the incoming photon by
(see Exercise 2.9):

.6,A = Ar - Ai = (~C) (1 - cos'l9) ; (2.62)

here cos'l9 = ki . kc / kikr represents the cosine of the angle between the incom-
ing and the scattered photon. Evidently this small shift is of pure quantum-
mechanical origin; it vanishes for the limiting case li -+ O. Hence, we pursue
the classical approximation and set Ai = Ar. As a consequence the 8 function
in (2.61) becomes
1
8(lic(ki - kr ))) = hc8(ki - kr). (2.63)
Here k = k i - kr ;::::: 0 has been employed for the energy of the electron:
li 2 li 2 li 2 li2 li2
-(qi + k)2 = _q2 + _ k 2 + -qi' k ;::::: -q~. (2.64)
2m 2m 1 2m m 2m 1
With (2.63) the total photon scattering cross section (2.61) becomes

O"total = ( ne4
me2
) ""'
6
0",
J k f2 dkf d Jtk,
n ICk""i'
Cif
k 12 ~
2kckw, '(k f - k·)
/00 U
ftC
1

(~:2 ) 2 LJ drlk, ICkiO"i . ck,O", 12 . (2.65)

0",
The photon is scattered into the solid angle drlf (see Fig. 2.12).
ki The differential cross section for an incoming photon with, polarization
Fig. 2.12. Geometry of ckiO"i' to be scattered into the solid angle drl k, with polarization ck,O", is given
the photon scattering by
 2.5 Calculation of the Total Photon Scattering Cross Section 59

(2.66)

If the photons are unpolarized, expression (2.65) has to be averaged over the
initial polarizations and has to be summed over the final polarizations to
obtain the total cross section. We get (see Fig. 2.13)

"21""""
L L ICkiO'i' ckfO'fl 2 "21[ ICkil' ckf11 2 + ICkil' Ckf 2 12
€kf 2
+ ICki2 . ckfll2 + ICki2' Ckf212] Fig. 2.13. Wavenumber
1 and polarization vectors of
"2 [1 +0+0 + cos 2 '!9] the absorbed (i) and emit-
1 ted (f) photons
"2 (1 + cos2 '!9) (2.67)

and perform the integration over dstf = sin '!9 d'!9 drp in (2.65). The outcome
is

O'total = -811" ( -e-22 ) 811" 2

= -re ~ 0.66 b . (2.68)
3 mc 3
Here, re = e 2 Imc 2 ~ 2.8 fm represents the classical electron radius. Expression
(2.68) can also be derived classically and is known as Thomson's scattering
cross section.

EXERCISE

2.9 The Compton Effect

Problem. Show that for photon scattering a (small) shift in wavelength b.A =
Af - Ai = h(l- cos '!9) I mc occurs as a result of the nonrelativisitic conservation
laws for momentum and energy, (2.58) and (2.59). Assume that the electron
is initially at rest.

Solution. Since the electron is at rest initially we have qi = O. With lick = I1w
we derive from (2.58) and (2.59) that
n} 2
lick i lickf + ~ (1)
2m
and
ki = kf + qf· (2)
From the second equation we directly get
q; = (k i - kf)2 = k~ - 2ki . kf + k; .
Inserting this into the first equation gives

+ -1i( k i2 -
2k i k f cos '!9 + kf2) ,
2
lick i = lickf
2m
where '!9 represents the angle between k i and k f . With lik = hi A it follows
that
60 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.9. h
c-
A
Multiplication with AA' results in

eh(A' _ A)
'---v---"
~
2m
(A'A +~)
N
_hm cos{) '
2

=~A

~A = l::...-
mc
[! (A' +~) -
2 A N
cos{)]

If we assume that the shift in wavelength stays small, i.e. A' ::::: A, then
A' A A,2 + A2 2A2
'I + N = AN ::::: --:\2 = 2
and
h
- (1 - cos{)) .
me

EXERCISE

2.10 Two-Photon Decay of the 2s State of the Hydrogen Atom

Problem. Determine the lifetime of the 2s state of the hydrogen atom as-
suming a decay with the emission of two photons. It is necessary to combine
the matrix element of second order, iI' rv p. A, with the matrix element of
first order, iI" rv A2. Do not solve the problem exactly, only estimate the
order of magnitude.

Solution. The lifetime is given by the expression

1
T

where nek 1 and nek 2 represent the energies of the emitted photons. The sum
over the final states goes with a summation over the polarizations and an
integration over the momenta of the two photons,

L
final states
--+ (2:)3 L Jd3k1 (2£:)3 L Jd k
0"1 0"2
3 2

(£3 represents the normalization volume) and, consequently,

where (E2s - E 1s ) = hcko and 8(o:x) = 1/(lo:I)8(x) have been used.

The Hamiltonian entering the matrix element M = (lsliIrI2s) splits into
iIr = iI' + iI", where according to (2.8) and (2.9)
 2.5 Calculation of the Total Photon Scattering Cross Section 61

L (27r!ic ~ PA . E:k" (
e 2) A ..k·x At -.k·x Exercise 2.10.
fI' - mc L3 Wk ak" e + ak" e . )
k,"
and
fIll

X
A A
( ak"ak',,' e
i(k+k')·x + ak"ak,,,,e
A At i(k-k')·x

+ ak"ak'
At, i(-k+k')·x + At At i(-k-k').X)
,,' e ak"a k , ,,' e

represent the interaction terms '" p . A and '" A 2 , respectively.

We are interested only in photon creation, so we restrict our consideration
to

and
fIll

fIll creates two photons, so it is sufficient to consider the matrix element of

first order perturbation theory. On the other side fI' creates only one photon
in first order. Hence, the two photons have to be created successively, i.e. via
a process of second order in perturbation theory. We arrive at two different
matrix elements:
MN) (flfI"li) (first order) ,

Mr(.2) "
~
(f1iI'ln) (nIH'li)
Ei _ En
( d d )
secon or er .

In the latter term the (time) order of the creation of the two photons is
important:

11 s) lIs)

1
or
Fig. 2.14. Feynman dia-
12s) 12s) grams of second order

Therefore, we obtain two contributions of second order. Both amplitudes have

to be added coherently. They are illustrated in Fig. 2.14. With w = ck, terms
carrying dimensions can be explicitly factorized out:

M -
-
~
2mc 2
(27r!ic
P
2
) 1
CVklk2
M'
.
62 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.10. Thus the matrix element M' is dimensionless:

M' = el' e2(lsl e- i (k , +k 2 )·x 12s)
+ ~ "'" ((lS lp . e2 e- ik2 ,x ln) (nip· el e- ik"X 12s)
m L... E2s - En - lick l
n
+ (lslp, el e-ik1,xln)(nlp, e2e-ik2,xI2S))
E 2s - En - liek2

Here el and e2 are the polarization vectors of the first and second photons,
respectively. We then get

~7 = r;e "'" "'"

4(27r)3 L...L...
0'1 G2
J J d3 k l
kl
d3k2IM'125(k - k - k )
k2 0 1 2,

where re = e2 /me 2
represents the classical electron radius. Since we are inter-
ested only in an order of magnitude estimation, we approximate the dimen-
sionless matrix element M' as

As a consequence we obtain

~
7
= r;e
4(27r)3
J J d 3k l
kl
d 3 k 2 8(k o - kl - k 2) .
k2
Because of spherical symmetry [f(k) = f(lkl)J the relation

J d 3 k f(k) = J J k 2 dk dD f(lkl) = 47r J k 2 dk f(lkl)

holds, so that

1 =
-;. r;e
4(27rp167r 2 J J kl dk l k2 dk2 5 [(ko - kl ) - k2J .

Owing to the 8 function, the outcome of the integration over k2 is

1
7
2
r_e_
27r
e l0
ko
k1(ko-kddk 1

r;c (~k~ _ ~k~)

27r 2 3
2
Tee k3
127r 0'
where ko is the maximum energy that can be emitted. Since
3 Z2 e 2
liek o = E 2s - E ls = 8 -a-
where Z is the central charge of the nucleus (for hydrogen Z = 1), we deduce

ko = ~Z2~
8 a
with a = e2 / lie ~ l~7' It then follows that
 2.6 Cherenkov Radiation of a Schrodinger Electron 63

1 Exercise 2.10.
T

where the Bohr radius a = fj? /me 2 has been used. We obtain T = 0.114 Z-6 s.
Hence, we deduce for the hydrogen atom TH = 0.114s. Indeed, the two-photon
decay is much faster than the one-photon decay with T = 9.5 years (compare
with Exercise 2.6).
With an exact treatment of the matrix element M' we would derive
.!. = 8.226 Z6 s-1 .
T

We realize that the approximation we have used is well justified!

Historical Remark. Over many years two-photon decay has attracted much
interest. Qualitatively it is different from the one-photon transition and repre-
sents another confirmation of the quantum field theory of radiation. We refer
the interested reader to the publications by M. Goppert-Mayer: Ann. Phys. 9
(1931) 273 and J. Shapiro, G. Breit: Phys. Rev. 113 (1959) 179.

2.6 Cherenkov Radiation of a Schrodinger Electron

In the previous subsections we have seen that an electron moving freely in the
vacuum cannot radiate, i.e. cannot emit a photon. Also, such an electron can-
not absorb a photon, because - as for the photon emission - the conservation
laws of momentum and energy cannot be fulfilled simultaneously; see again
the discussion following (2.51) and (2.52). At least one additional particle has
to be involved to absorb the recoil.
Let us now turn to the motion of a charged particle (electron) in a dielectric
medium. From classical electrodynamics we know that the speed of light c' in
such a medium is given by
, c c c
c = -
n
= ---
,ftJi
~ --
~.
(2.69)

The variable c:(w) is known as the dielectric constant and depends on the
frequency w of the electromagnetic wave. Ii- ~ 1 is called the magnetic per-
meability of the medium. n(w) ~ y'c:(w) represents the refractive index. The
frequency (w) dependence of these quantities indicates the dispersion of elec-
tromagnetic waves in the medium. In general the speed of light c' in the
medium is smaller than the speed of light c in vacuum, i.e. c' < c. For this
reason a particle can move with a velocity v > c' in the medium, which is be-
yond the speed of light (in a medium). We will convince ourselves that under
these circumstances a charged particle is able to emit and absorb photons.
The medium acts as an additional particle that absorbs the recoil and in this
way permits the conservation laws to be fulfilled.
The energy of the electromagnetic field is given by expression (1.40):

H rad = - Ii,
81l' L3
d 3 x(E 2 +B 2 ). (2.70)
64 2. Interaction of Electromagnetic Fields with Matter

However, this is no longer the total energy of the system. A contribution

to the energy from the medium is missing that expresses the reaction of the
particles in the medium to the electromagnetic wave. First of all we illuminate
this problem with the following example.

EXAMPLE

2.11 The Field Energy in Media with Dispersion

The expression for the energy flux density (Poynting vector),

c
S = -ExH, (1)
47r
is valid for any arbitrary electromagnetic field. This also holds when dispersion
is present. We can see this in the following way. 10 Directly at an interface (for
example medium-vacuum) the tangential components of E and H are contin-
uous (see the lectures on classical electrodynamics). Therefore we decompose
them into tangential (t) and normal (n) components:
E(l) Eil) + E~l) , (2)
E(2) E(2) + E(2)
tn'
(3)
H(l) HP) + H~l) , (4)
H(2) Hi 2) + H~2) . (5)
Consequently, the Poynting vector at side 1 of the interface can also be de-
composed into a normal and a tangential component:
S(l) 4~ [(Ei l ) + E~l)) x (HP) + H~l))]
~ [E(l) x H(l)
47r t t
+ E(l)
t
x H(l) + E(l) x H(l)]
n n t

~
47r
[S n
+ S(1)]
t'
(6)
where we have distinguished between tangential and normal components, i.e.
S(l) E(l) x H(l) (7)
2 n t t'
St(l) E(l) x H(l)
t n
+ E(l)
n
x H(l)
t·
(8)

The term
E(l) x H(l)
n n
= 0 (9)
vanishes. Similarly we arrive at the result
Fig. 2.15. At the inter-
face H?) H?) and S(2) = 4~ [Sn + si 2)] (10)
E~l) = E?) hold, i.e. the
tangential components of for side 2 ofthe interface. Obviously, because of the continuity of the tangential
E and H are continuous components, we have
10 We follow here arguments given by L.D. Landau, E.M. Lifshitz: Electrodynamics
of Continuous Media (Pergamon, Oxford 1960).
 2.6 Cherenkov Radiation of a Schrodinger Electron 65

(11) Example 2.11.

the normal component of the energy flux is continuous at the interface. This is
what one expects. We would not have obtained this result if we had used Ex B
or D x H or D x B in (1). We realize that (1) represents the correct general
expression for the energy flux density; moreover, in vacuum it coincides with
the already known result (1.18).
The change of energy per second in a normalized volume of the medium
is described by div S. Applying Maxwell's equations in a medium,
laD
rotH - - -
47r .
-3,
at c c
laB
rotE + - - 0,
c at (12)
divB 0,
divD 47rp,

we obtain after a short calculation

divS = _~ (E. aD +H. aB) (13)
47r at at
For the case of a dielectric medium without dispersion, for which E and jl are
pure constants, (13) can be interpreted as the change in the electromagnetic
energy density,
1
u = -(EE2 + jlH2) , (14)
87r
with time, i.e.
au
at = -div S. (15)

Such a simple interpretation is not possible for media with dispersion. More-
over, dispersion present in a medium in general causes dissipation of energy; a
medium showing dispersion is absorbing. In order to calculate this dissipation
we consider a monochromatic electromagnetic wave and determine the aver-
age of expression (13) over time. Because (13) is quadratic in the real field
quantities E and H we have to be careful with a transition to complex-valued
fields and have to rewrite all expressions with the real forms. To avoid a mis-
understanding we now call the monochromatic fields E and H and denote
them in complex-valued form:

E ----+ ~(E + E*) ,

aD ----+
1
"2( -iwEE + iWE* E*) ,
at (16)
H ----+ ~(H + H*),
aB ----+
1
"2( -iwjlH + iWjl* H*) .
at
66 2. Interaction of Electromagnetic Fields with Matter

Example 2.11. Products like E . E and E* . E* vanish with time averaging, because they
contain factors e'F2iwt. We deduce that

~
47r
(E .
-
aD
at + -H . aB)
at = iw [(e* -
167r
e)E . E* + (11* -
r'
II.)H . H*]
r'

~
87r
(e"IEI 2 + Jl"IHI 2 )
Q. (17)
The quantity Q can be understood as a mean amount of heat produced per
normalized time and normalized volume; according to (13) it describes the
disappearing energy flux averaged over time and per unit time and volume
element. Furthermore, we can write Q also in the form

Q (18)

with

e" ~(e* - e) ,

Again, E and H stand for the real field quantities and the bar indicates a
time average over a period T = 27r / w.
Equation (18) is very important as it demonstrates that energy dissipation
(absorption) is governed by the imaginary parts of e and Jl. The two terms in
(18) are interpreted as electric and magnetic losses. The sign of Q must always
be positive because the second law of thermodynamics means the total entropy
of a closed system, which is a measure for the equipartitioning of energy, must
increase. The dissipation of energy results in production of energy. Therefore,
Q > 0 has to hold for all times. From this and from (18) we conclude that the
imaginary parts of e and Jl must always be positive:
e" > 0, Jl" > o. (19)
This is a very general statement. It is a result of the second law of thermody-
namics and is valid for all materials and frequencies. The real parts e' and Jl'
of e and Jl are given by
e +e*
e' =-- (20)
2
No restrictions exist for them. Depending on the particular physical circum-
stances, they could be either positive or negative.
For a real physical material, nonstationary processes are always thermo-
dynamically irreversible. Thus a variable electromagnetic field always suffers
from electric and magnetic losses in a real material. Often they remain small,
but they are always present. As a conclusion, the imaginary parts, e"(W) and
Jl"(w), do not become rigorously zero in any frequency domain. This finding
is of fundamental importance, as we will recognize in the following. Obviously
this does not exclude the existance of frequency domains for which e"(W) and
Jl"(w) take on very small numbers, so that only minor losses occur. Such fre-
quency domains are called domains of transparency of the material. For them
it makes sense to introduce the notion of the intrinsic energy U of a material
 2.6 Cherenkov Radiation of a Schrodinger Electron 67

in an electromagnetic field, presuming, of course, absorption to be neglected. Example 2.11.

It then has the same significance as for a constant field.
To determine U we have to consider a field formed by a bunch of monochro-
matic waves whose frequencies fall into a narrow interval around the mean
frequency woo Owing to its strict periodicity, a purely monochromatic field
would not allow the electromagnetic energy to be localized in time. This can
be achieved only with a wave packet. Hence, we introduce

E Eo(t) e- iwot ,
(21)
H Ho(t) e- iwot ,

where Eo(t) and Ho(t) are slowly varying functions in time compared to
e- iwot .
To determine the intrinsic energy we have to calculate the right-hand side
of (13) for transparent materials. We insert the real parts of (21) into (13)
and obtain

div S = _ ~ [(E + E*) . CD + n*) + (H + H*) . (B + 1J*)]. (22)

47r 2 2 2 2

Averaging this expression over time, i.e. with respect to the period T = 27r /wo,
. . *. .*
leads to the products E· D, E*· D ,H· B, and H* . B vanishing and results energy
in density

divS = __1_ [(E.n* +E*.n)+(H.B* +H*.B)] (23)

167r time
The derivative fJD(t)/fJt will be a function of E(t). We write
fJD fJB Fig. 2.16. A wave packet
at fE at
A A

= and analogously = gH, (24) consisting of electromag-

netic waves allows us to
where j and 9 have to be thought of as operators. If Eo in (21) were constant study the intrinsic energy
in time such that D(t) = c:(w)E = c:(w)Eoe- iwut , we would deduce that of a material because of
dispersion
fJD
- iWoc:(wo)E == jE = f(wo)E, (25)
fJt
and thus
f(w) = - iwc:(w) . (26)
We expand the function Eo(t) and, analogously, Ho(t) from (21) into a
Fourier integral:

Eo(t) = J Eo(a) e- iCtt da. (27)

Compared to e- iwot , the amplitude Eo(t) changes only slowly with time; there-
fore we expect only those components with
a« Wo (28)
to appear in the Fourier integral. Then we can write for each single component
68 2. Interaction of Electromagnetic Fields with Matter

Exampel 2.11. jEo(a) e-i(wo+<»t J(a + wo)Eo(a) e-i(wo+<»t

::::: J(wo)Eo(a) e-i(wo+<»t
+ df(wo) aEo(a) e-i(wo+<»t
dwo
(J(w o) + a d~~o)) Eo(a) e-i(wo+<»t . (29)

Inserting (27) and (29) into (21) we deduce that

jEo(t) e- iwot
1
jE(t)
j Eo(a) e-i(wo+<»t da

1 jEo(a) e-i(wo+<»t da

1 da (J(wo) + a d~~o)) Eo (a) e-i(wo+<»t

f(wo)E(t) + d~~o) (1 aEo(a) e-i<>t da) e- iwot

J(wo)E(t) + dJ(wo) i aEo(t) e- iwot . (30)

dwo at
Now we omit the index zero for Wo and with the help of (24) and (21) we are
led to the result
aD(t) = _ iWE(w)E(t) + d(wE) aEo(t) e- iwt . (31)
at dw at
From here we calculate the first term in (23). We assume transparency, neglect
the imaginary part of E (transparency of the material is assumed!), and finally
obtain

__1_ (E. b* + E*. b) = __1_ (iWE(W)E. E*(t) + d(wE) Eo' aE*

167f 167f dw at
_ iWE(w)E* . E(t) + d(wE) Eo' aE o )
dw at
= __1_ d(wE) (E* . aEo(t) E. aEo(t))
167f dw 0 at + 0 at
= __1_d(wE) ~(E. E*). (32)
167f dw dt
For the last step we have used Eo . Eo = E . E*, which follows from (21).
Analogously we deduce an expression for the second term in (23), the only
difference is that E has to be replaced by f.L and E by H. Hence, we write (23)
in the form

(33)
 2.6 Cherenkov Radiation of a Schri:idinger Electron 69

where Example 2.11.

U = _1_ (d(WE) E. E* + d(wJl) H. H*) (34)

16~ dw dw
represents the time average of the electromagnetic part of the intrinsic energy
per normalized volume in a transparent medium. Introducing again the real
fields E and H via
E+E* H+H*
E = 2 (35)
2
we have
E .E = i (E . E + 2E . E* + E* . E*) = ~ E . E* (36)
and an analogous result for the H term. This yields for (34)

'IT = ~ (d(WE) E. E + d(wJl) H· H) . (37)

8~ dw - - dw - -
For media without dispersion E und Jl do not depend on frequency and, as
expected, (37) becomes identical to the expression (14). However, although
small, absorbtion is always present. As a consequence the total energy density
'IT transforms into heat as the electromagnetic energy flux is turned off from
outside. According to the second law of thermodynamics this heat is produced
and is not transformed back into electromagnetic energy. Therefore,
'IT> 0,
from which
d(wE) > 0 and d(WJl) > 0 (38)
dw dw
follow in general.

We return to (2.70) for the energy of the free electromagnetic field. This
expression has to be modified for media with dispersion according to (37) of
the last exercise:

Hrad = J d 3 xu

J d3x ~
8~
(d (WE(W))
dw
E. E + d(WJl(w)) H· H)
dw
(2.71)

In comparison with (2.70) the additional terms of the form w(dE(w)/dw)E2

and w(dJl(w)/dw)H 2 describe the reaction of the medium to the energy con-
tained in the electromagnetic waves; to be more precise, they describe the
effect on the energy stored by the molecular currents induced by the electro-
magnetic waves. For most materials the permeability is Jl ~ 1, which indeed
is a very good approximation, and, as a consequence, H = B. Then, the field
energy (2.71) in the medium becomes
70 2. Interaction of Electromagnetic Fields with Matter

H rad -- Jd 3 X 81l'
1 (d(WE(W))E
dw . EBB)
+ . (2.72)

For the case of media with no frequency dependent dielectric constant E(W) =
E= const we deduce the expression
H rad = J 3
d x-
1
81l'
(EE. E + B· B) , (2.73)

which is well known from electrodynamics. We have set J1 = 1.

We use the results (1.44) and (1.45) obtained earlier to derive the quantum-
mechanical operator for the field energy corresponding to (2.73). The expres-
sion (1.41) for J£3
E2 d 3 x was given by

J -E
L381l'
1
. Ed 3x = "'""
~ 81l'c2 k
k,a
2
Wk N2L3 [(Aaka aka
At At aka
+ aka A )

Then the classical expression of (2.73),

J 3 1
d x-EE·E
81l' '

becomes

(2.74)
Similarily we deduce from (1.45) that

(2.75)

Hrad =

(2.76)

We realize that terms not conserving the photon number [second sum in (2.76)]
vanish for a constant dielectric constant. With the refractive index, n 2 = E,
the remaining term becomes

A "'"" W~ 2N 2L3 [A At At A ]
H rad = ~ 41l'c 2 n k aka aka + a ka akl7 (2.77)
k.17
 2.6 Cherenkov Radiation of a Schrodinger Electron 71

This result has to be compared with (1.46). To obtain a far analogous to

(1.48), we have to choose the normalization constant as

(2.78)

so that

(2.79)

This change in the normalization becomes noticeable for the expression for
the vector potential. From (1.37) we deduce with (2.78) that

A(x,t) = L (2.80)
k,a

Then, only a normalization constant is changing for the interaction Hamilto-

iI:
nian nt [see (2.8)J:

iI!mt -- -~ ~
me~
(2.81 )
k,a

With this result we now determine the transition amplitude per unit time for
a free electron with momentum liq emitting a photon with momentum lik.
After the emission of the photon the electron has momentum Iiq' = Ii(q - k).
We find

( tran~. Prob.) = 27rIMfi126 (li2q2 _ li2q'2 _ fu.J k ) , (2.82)

tIme q-->q' Ii 2m 2m
where the relevant matrix element is given with the help of (2.45) and (2.81)
as

,
x [ak'a' e ik'·x + ak'a'
,t e -ik'.X] I'f/;q ())
x I ... ,Oka,··· )

e
(2.83)
me
Obviously,
('f/;q'(X)lp·ckae-ik.xl'f/;q(x)) = liq·cka 6q',q-k, (2.84)
and we obtain the following preliminary result for (2.82):

( tran~. prob.) =
tIme q-->q-k

(2.85)
72 2. Interaction of Electromagnetic Fields with Matter

Since (q - k) 2 = q2 + k2 - 2kq cos rJ, it follows for the 8 function that

li2q2 li2(q - k)2 ) (2li 2qkCOSrJ li2k 2 )
8 ( -- - - fu,,;k 8 - - - - fu,,;k
2m 2m 2m 2m
m ( !ik fu,,;km)
li2qk 8 cos rJ - 2!iq - !iqlik

1 8( rJ nfu,,;k fu,,;k m )
vlik cos - 2mvc - mvli~n

V~k 8 (cosrJ - :v - ~n::::) .

Here we have used 8(ax) = (1/IaI)8(x) and liq = mv. The final result then
reads
41l' 2 e 2 li2 1q' . ekO'I 2
( tran~. Prob.) =
m 2 £3vlikwkn2
tIme q-->q'=q-k

x8 ( cos rJ - - c - -
nfu,,;k)
- (2.86)
nv 2mcv
The photon is emitted at an angle rJ with respect to the electron trajectory.
This Mach angle is given by

cos rJ = ~
nv
[1 + fu,,;n 22 ]
2mc
. (2.87)

If the photon energy is much smaller than the rest energy of the electron,
fu,,;« mc2 (mc 2 ~ O.5MeV), we obtain approximately

c c'
cosrJ = - = -. (2.88)
nv v
This is the classical Mach angle (see Fig. 2.17), which in this context is also
named after its Russian discoverer as the Cherenkov angle. Equation (2.88)
can be fulfilled only if the particle velocity v is larger than the wave velocity
c' = cln, i.e. v > cln. For a vacuum, n = 1 and v can never become larger
than c. Therefore a photon emission from a free particle is not allowed in
vacuum; see again the discussion about photon absorption after (2.52).
In the following we are interested in the loss of energy of the electron per
unit length along its trajectory (see Fig. 2.18). It is given by

Fig. 2.17. cos 19 = c'lv

determines the classical
Mach angle. Cherenkov
radiation can be inter-
preted as a Mach pheno-
menon: supersonic flight
of an electron through a
medium with c' < c
 2.6 Cherenkov Radiation of a Schrodinger Electron 73

dW = ~ dW = ~ L nwk (trans. Prob.) (2.89)

dx v dt v ka time q--->q-k

We calculate

a a

q2 sin 2 fJ
q2(1 - cos 2 fJ)
m 2v 2
Ji2 (1 - cos 2 fJ)

and use (2.18), which can be expressed in spherical coordinates in k space: Fig. 2.18. Illustration for

L
k
----> (£3)3
271'
roo k
io
2 dk 11
-1
d( cos fJ) r
io
27r
drp. (2.90)
the energy loss of a parti-
cle in a medium along the
trajectory x
We then derive
dW
dx

~/k
(2.91 ) ek2~ ~
\...ek 1
In the last step we have not been careful enough; as it is, the integral would
Fig. 2.19. Illustration of
diverge. But this is not the case because, as a result of (2.87), cos fJ can only the polarization vectors
range between -1 and 1 and thus limits the photon energy to relative to the particle mo-
nw < (nv/c - 1) 2mc 2 (2.92)
mentum hq = mv
- n2
The integration domain in (2.91) does not reach out to infinity, only to the
limiting energy (2.92).

EXERCISE

2.12 The Cherenkov Angle

Problem. Show that the Cherenkov angle is given by the expression

cosfJ = -c
nv
[nw
1 + __ (n 2 -1)
2mc
2
g2]
1--
c2

if the relativistic instead of the nonrelativistic expression for the particle en-
ergy is applied.
74 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.12. Solution. As E2 = p 2c 2 + m 2c4 , it follows from energy conservation that

vn2c2q2 + m2c4 = vn2c21q _ kl 2 + m 2c4 + likc .

n
Taking the square yields

,,2 21 kl2 =ltcq-

,,2 2 2 2Enkc 2 2k 2
Itcq- -n +n-nc 2-
where the electron energy is given by E = vn2c2q2 + m 2c4. From
Iq - kl 2 = q2 - 2qkcos'I'J + k2
we derive
E k 2
cos'I'J - + - ( n -1).
n1icq 2qn 2
Now
mv
E and 1iq = ---r=:===;;=~
VI - V2 /c 2
as well as ck = nw and it follows that

cos'I'J = -
vn
c[
1 + -n-w1-
2mc2
g2_(n 2 -1)
c2
1

2.7 Natural Linewidth and Self-energy

The photon emitted from a quantum system corresponds to a wave of finite

duration and of finite length. The spectral line emitted from the system has

r*IIEui
a (natural) linewidth because the initial state has only a finite lifetime as a
result of its decay. According to Heisenberg's uncertainty relation the energy
uncertainty flE is related to the lifetime T by flE ~ n/T. Up to now we
have considered only sharp spectral lines. This is reflected in the 6 function
6(Ea, - E a, -nwk) appearing in the expression for the transition probability
(2.16). Evidently something must be wrong in our considerations employed
E Uf so far. In fact the mistake comes from perturbation theory as we have taken
it over from the Introduction to Quantum Mechanics. l l Now we will modify
Fig. 2.20. A finite life- this formalism and include a finite linewidth. 12
time means a finite level Hence, we again consider the emission of a photon from an atom. Initially,
width r and a finite spec-
tralline width result from the atom is in state lai) and, to keep things simple, is only allowed to decay
the uncertainty relation into one final state laf). As previously no photon should be present in the
~t~E ~ n initial state and only one in the final state. We write

11 W. Greiner: Quantum Mechanics ~ An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1994).
12 This was first examined by V. Weisskopf, E. Wigner: Z. Phys. 63 (1930) 54 and
65 (1930) 18.
 2.7 Natural Linewidth and Self-energy 75

Ii) lai)I···, Oka, ... ) ,

(2.93)
If) laf)I···,lka, ... ).
With states I~) of the total system expanded in terms of unperturbed states
I¢n),

(2.94)

the time-dependent Schrodinger equation reads

(2.95)

Here I¢n) represent the solutions of the unperturbed system, which consists
of the atomic many particle system and the radiation field Ho = Hmp + Hrad
[compare with (2.4)]:
(2.96)
Inserting (2.94) into (2.95) and considering (2.96) leads to a system of coupled
differential equations:

dcm(t)
dt
_
-
_ ~" ()(r!. IH'. Ir!.)
h L..t Cn t 'I'm tnt 'l'n exp
(i(Em -h En)t) . (2.97)
n

We denote the amplitude of state Ii) with Cio and those of states If) with
Cfka. The indices indicate that Ii) has no photon and If) has one photon of
the kind ka. Then (2.97) read

d~;o _~ L(iOIHintlfka)exp (~(Ei - Ef -nwk)t) Cfka,

ka
dCjka
dt
i,
-li:(fkaIHintliO)exp ( -li:(E
i i - Ef -nwk)t ) CiO· (2.98)

Be aware that we consider only the two atomic states Ii) and If). On the
right-hand side of the first equation all states laf)lka) are considered, into
which the initial state lai)IO) can decay. On the right-hand side of the second
equation only the term proportional to Cio is taken into account; in principle
also matrix elements with two-photon states la f) I ... , 2ka, ... ) appear, but the
amplitudes of these states stay negligibly small. For the amplitudes CiO of the
decaying state lai) we make the ansatz

CiO(t) = exp (-~~Eit) . (2.99)

The energy shift ~Ei remains unknown for the moment and still has to be
determined. ~Ei can be a real, imaginary, or complex-valued quantity. We
insert (2.99) into the second equation of (2.98) and deduce

dClta = -~(fkaIHintliO)exp (-~(Ei + ~Ei - Ef -nwk)t)

After an integration over time from 0 to t we get
76 2. Interaction of Electromagnetic Fields with Matter

A • exp [-*(Ei + !:I.Ei - Er - nwk)t] - 1

Ctku(t) = (fkO"IHintlzO) Ei + !:I.Ei _ Er _ nwk . (2.100)

So far we have not used the first equation of (2.98) at all. Also, we did not de-
termine !:I.Ei . We choose !:I.Ei such that the first equation of (2.98) is fulfilled.
We insert (2.99) into the left side ofthe first equation of (2.98), substitute the
amplitudes (2.100) into the right side, and get

- ~!:I.g
It 1
exp (-~!:I.gt)
It 1

- ~ L(iOIHintlfkO") (fkO"IHintliO)
ku

exp [-*(Ei + !:I.Ei - Er - nwk)tJ - 1 (i )

x E!:I.E E nw exp ~(Ei - Er - nwk)t
i+ i- f- k n

i '"
- ~ L.J
I( fkO" IHint I'20 )1 21 -
A exp [*(Ei + !:I.Ei - Er - nwk)t]
E!:I.E E nw
n ku i + i - r- k

X exp ( -~(Ei + !:I.Ei - Er - nwk)t + ~(Ei + !:I.Ei - Er - nwk)t)

Division by (-illt) exp( -*!:I.Eit) on both sides results in

!:I.Ei = L 1(fkO"IHintl iO )12

ku

(2.101)

For large times t ---+ 00 it holds that

ixt
lim ( 1 - e ) = -1 - i7rS(x) . (2.102)
t-->oo X X

We prove this important relation in Exercise 2.13. Applied to (2.101) this

leads to
'" 1(fkO"IH[ntl iO )12
f: Ei + !:I.Ei - Er - nwk
- in L 1 (fkO"IH[ntl iO )1 2 S(Ei + !:I.Ei - Ef - nwk).
ku

Since we want to determine !:I.Ei only up to second order in e, i.e. in H[nt'

it is consistent to leave out !:I.Ei on the right-hand side of the last equation.
Consequently, we obtain

1(fkO" IH{nt liO) 12

!:I.Ei = L
k,u
Ei - Ef - nwk

- in L 1 (fkO"IH[ntl iO )1 2 S(Ei - Er - nwk). (2,103)

k,u
 2.7 Natural Linewidth and Self-energy 77

Hence, the energy shift !:lEi is complex valued. The real and imaginary parts
can be read off right away from (2.103):

" 1UkaIH{ntl iO ) 12
(2.104a)
~ Ei - Er - nwk '
k,IT
-1r L 1 UkalH{nt liO) 126(Ei - Er - nwk)' (2.104b)
k,IT

The sum L:kIT is over all photons with all possible momenta nk and polar-
izations a. It is straightforward to see that a consideration of several atomic
states lar) leads to an additional summation L:f for the right-hand side of
the first equation of (2.98); hence, the result (2.104) modifies to

" " 1UkaIH{ntliO) 12 (2.105a)

~~ Ei-Er-nwk '
k,IT f
8'(!:lEi) = -1r L L IUkaIH{ntliO)W6(Ei - Ef - nwk). (2.105b)
k,IT f
The sum over final states (L: f ) is not restricted in (2.105a); on the other
hand, because of the 6 function, the sum in the second equation is restricted
by Er < Ei·
We draw the conclusion that the processes of emission and absorption
of photons as described by (2.105) generally do not need to fulfill energy
conservation. Such processes are called virtual processes. We can think in
terms of the physical atom having dissociated for a short time into "atom + Fig. 2.21. Self-energy of
virtual photon". According to (2.104) and (2.105) it is these virtual photons a bound electron. Pho-
of all possible momenta and polarizations that cause the real energy shift. tons are emitted and re-
The diagrams in Fig. 2.21 illustrate this point again. absorbed. Such photons
Contrary to this the sums in (2.104b) and (2.105b) only cover photons are called virtual photons.
Also the electron in the
that have energy conservation: intermediate state f (be-
Ei = Ef + nw. tween emission and reab-
sorption of the photon) is
In contrast to the virtual photons discussed before these photons are real. virtual
Obviously the energy shift !:lEi possesses an imaginary part only if the state
lai) is allowed to decay into If) via spontaneous emission without violation of
energy conservation; see Fig. 2.22. This is expressed by the 6-function 6(Ei -
Ef - nw).
Equation (2.105b) can also be cast into the form

-~8'(!:lEi) = LL2;IUkaIH{ntliO)126(Ei-Ef-nw) , _.....z..._~r-----l~-!I

kIT f ___---"_---l~-h
where we recognize the expression (2.17) for the reciprocal lifetime l/Ti of the ______.L.._h
state lai) on the right-hand side. Therefore, we identify Fig. 2.22. The state lai)
1 can only decay into ener-
(2.106) getically lower states lar)
Ti
78 2. Interaction of Electromagnetic Fields with Matter

The physical interpretation of ~(~Ei) has now become clear. We go back to

(2.99) and combine it with the "usual" time dependence exp( -*Eit) of the
state lai); we deduce

¢a, = ua,(x)exp (-k(Ei + ~Ei)t)

ua,(x)exp (-k(Ei + ~(~Ei))t) exp (-~t) (2.107)

or

(2.108)

The state lai) decays with lifetime T. On the same footing the state obtains a
level shift ~(~E), which results from the emission and reabsorption of (vir-
tual) photons (see Fig. 2.21).
Another consequence from the decay of state lai) becomes clear by looking
at (2.100). Inserting
i
~Ei = ~(~Ei) - 2'ifi

yields

() _ ( 1 I I' )ex p (-*(Ei + ~(~Ei) - fiwk)t) exp( -,i t / 2) -1

jk(J Hint to
A

Clk" t - Ei + ~(~Ei) _ Ef - fiw k + (i/2)fi'i

The probability of finding a photon with frequency Wk, momentum fik, and
polarization (J in the radiation field after a long time t » 1/'i is given by

lim
t-+oo
ICfk,,12 = l(Jk(JIH{ntliO)12
(2.109)

Of course this reflects the intensity distribution of the emitted line: The spec-
tralline has a Breit- Wigner distribution with center fiw k = Ei + ~(~Ei) - E f
and width fi'ii see Fig. 2.23. Because of the self-energy of the electron, emis-
lIw
sion and reabsorption of photons results in the spectral line being shifted by
Fig. 2.23. Breit-Wigner ~(~Ei)' In Sect. 5 we will come back to the determination and the necessary
form of the spectral line renormalization of the self-energy in more detail.

EXERCISE

2.13 Plemlj's Formula

Problem. Prove the relation

ixt
1 --e - )
lim ( - = P ( -1 ) - i7r8(x) . (1)
t-HXJ X X
 2.7 Natural Linewidth and Self-energy 79

Solution. The relation Exercise 2.13.

iot e
ixt
lim (1 - e ) = lim (-i ixt ' dtl) (2)
t---+= X t---+=

holds because
t
i
-i 0 eixt'dt ' = -i - elxt
1 . , [t' =t
ix t'=O
1
= - - ~.
x
ixt
x
(3)

Now, let us introduce a small imaginary part and let c: go to 0+:

iot e = - lim lim (i

iot ei(X+io:)t' dtl)
- lim (i ixt ' dtl) ; (4)
t---+oo t---+oo 0:---+0+
uniform convergence in c: is assumed. We interchange the two limiting proce-
dures

lim lim (i
t---+CXJ c---+O+ iorot ei(x+io:)t' dt') = - lim lim (i
0: ..... 0+ t ..... oo
t ei(x+io:)t' dtl)
io
_ lim (i roo ei(x+io:)t' dtl) . (5)
0: ..... 0+ io
The last step is possible because the integral is defined as long as c: > 0 and
also remains finite for t ...... 00.
This integration can be performed:

r
io
00
ei(x+io)t'dt ' =
i(x
1
+ ic:)
ei(x+io:)t'
[tl=OO
t'=o
(6)

Since this integral also converges at the upper bound because of the conver-
gence factor exp( -c:t'), i.e. exp(i{x + ic:)t')lt'== = 0, we deduce that

- lim (i
0:---+0+ io
r= ei(x+io:)t' dt') =
o:~rg+ (x ~ ic: )
o:~rg+ ex + 7c:)(~C:- iC:))
o:~rg+ (x2: c: 2 - i x2 : c: 2) (7)
The imaginary part represents a well-known representation of the r5 function:

lim
0---+0+
(~)
x + c:
= 7l'r5(x). (8)

To prove this we check the defining properties of the r5 function:

lim (11 -
00
- -c:- d x)
x 2 + c: 2
(1 1
0---+0+ 7l' -00

r 00
d(x/c:) )
1 + (X/c:)2

(11
o:.:...rg+ ;;: -00

00
lim - -dy- )
0---+0+ 7l' -00 1 + y2
1, (9)
80 2. Interaction of Electromagnetic Fields with Matter

Exercise 2.13.
I: 6(x)f(x) dx = lim
0->0+
[~1°°
1r -00 x2
C
+ c2
x (f(O) + xl' (0) + ~2 !" (0) + ... ) dX]

f(O).1 + lim (1'(0) 1 ~cdX2

+
00
+ ...)
c

1 +
£ ...... 0+ 1r -(X) X

f(O) + lim (1'(0) c

00
Y dy )
£ ...... 0+ 1r -00 1 y2
f(O) . (10)
Thus, 6£(x) = c/1r(x 2 + c 2 ) approximates the 6 function for small and de-
creasing c; see Fig. 2.24. The real part of the last part of (7),

lim
£ ...... 0+
(- -+x)
-
c
:Z;2 2
- P
-
(1)
-
X
(11)

represents the so-called principal value of 1/ x, which is defined by

Fig. 2.24. Approximation

of the (j function 1 00
-00
p (~)
X
f(x) dx = lim+
£ ...... 0
(1-£ -00
f(x) dx
X
+ 1 £
00
f(x) dX)
X
(12)

Hence, we obtain

(13)
3. Noninteracting Fields

So far we have learnt that the classical radiation field takes on particle prop-
erties (quanta with energy flw and momentum lik) as soon as it has been
quantized. Thus, we might suspect that all wave fields show particle char-
acter as soon as they have been quantized and that, on the other side, all
particles appearing in nature could be understood as quanta of a field. With
certain particles, such as electrons, the question arises, what is the wave field
that contains these particles as quanta? We presume that it might be the
wavefunction 'lj;(x, t) and begin with the nonrelativistic Schrodinger equation
fJ'lj;(x t) 1i2
fJ ' = _ _ V2'lj;(X, t) + V(x)'lj;(x, t) = H'lj;(x, t) (3.1)
A

iii
t 2m
for a particle with mass m in a potential V(x). Equation (3.1) is the fun-
damental equation of elementary quantum mechanics. As with the Maxwell
equations representing the field equations for the electromagnetic fields (com-
bined in the vector potential A and Coulomb potential ¢), we consider the
Schrodinger equation to be a field equation for the wave field 'lj;(x, t). Now we
quantize Schrodinger's wave field 'lj;(x, t) in complete analogy to the quantiza-
tion (1.50) of the vector potential (and with that of the electric and magnetic
fields) in the el~ctromagnetic case. We assume {'lj;n(x)} to be a complete set
of solutions of H,

(3.2)
so that we can write the most general solution of (3.1) as

(3.3)
n

The bn can be understood as normal coordinates of the system; then 'lj;n(x)

are called normal states. Insertion of (3.3) into (3.1) yields

dbn(t) = -~E b (3.4)

dt Ii n n .

This represents the equations of motion for the normal coordinates. It is called
first (i.e. the usual) quantization. In contrast to that formulation we will now
learn how to introduce creation and annihilation operators for field quanta
(which may be particles, photons, and so on). The formulation of quantum
mechanics in terms of such creation and annihilation operators is called sec-
ond quantization. To achieve this, our goal is now to find a Hamiltonian depen-

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
82 3. Noninteracting Fields

ding only on the normal coordinates bn , introduced above, and yielding the
equations of motion (3.4). It could be that H is given by

H = ! d 3x1/;*(x, t) [- 2: V2 + V(X)] 1/;(x, t)

= ! d 3 x 1/;* (x, t)H1/;(x, t) , (3.5)

because it represents the expectation value of the energy. The quantity

1/;* [- : : V2 + V(X)] 1/;

represents the energy density. With the help of (3.2) and (3.3) and employing
the orthonormality (1/;nl1/;m) = bnm we straightforwardly calculate for (3.5)

H ! d3 x1/;*(x,t)H1jJ(x,t)

(3.6)
n
This resembles a Hamiltonian consisting of an infinite sum of harmonic oscil-
lators with energies En and frequencies Wn = En/n. Hence, we interpret
At
b~ -> bn ,

bn -> bn (3.7)
as creation and annhilation operators of these oscillator quanta and require
the commutation relations

(3.8)

Then the Hamiltonian of (3.6) becomes

A ""' At A (3.9)
H = LEnbnbn
n

and the equations of motion for bn read

(3.10a)
or
."dbn
In-
dt

n'

(3.10b)
 3. Noninteracting Fields 83

Formally they are identical to the "classical" equations of motion (3.4). As

in Chap. 1, where the quanta of the electromagnetic field (photons) appeared
explicitly in the theory, the theory developed here describes quanta that obey
Bose~Einstein statistics. The reason for this lies in the commutation relations
(3.8), which are the same as in (1.50). These commutation relations allow us
to construct states of the form
Nmtitnes

1~
- - b b ···b 10)
VNm! m m m
_l_(b t )Nm 10) (3.11)
VNm! m ,

in which N particles of the kind m appear with the same wavefunction 'lj;m(x).
The number operator fin for particles of the kind "n" is
A At A
N n = bnbn , (3.12)
and the general state vector
I ... , nn, ... , nn', ... ) (3.13)
of the 'Ij; field is, as previously [compare with (1.54)], a direct product of the
state vectors of the field oscillators Inn)
I···,nn,···,nn"···) = ···,lnn)···lnn')···. (3.14)
Here
Nnl···,nn,o .. ,nn"···) == nnl···,nn,···,nn"···) (3.15)
represents the eigenvalue equation for the particle-number operator.
The formalism developed so far holds for bosons, because, as we have
realized, it allows several particles to have identical wavefunctions. Now, we
attempt to modify the formalism, so that it can also describe fermions. We
keep

as the Hamiltonian. We also require that the equations of motion (3.1Ob)

again lead to the classical equations (3.4). After some trials we observe that
the result (3.10b) can be derived from (3.1Oa) also with commutation relations
different from (3.8). For example with the anticommutation relations l

(3.16)

instead of the commutation relations (3.8), where the anticommutation oper-

ation is given by

[A,BL = AB+BA.
1 Those go back to P. Jordan, E. Wigner: Z. Physik 47 (1928) 631.
84 3. Noninteracting Fields

We easily calculate with (3.9)

(3.17)
and recognize that we have derived the same equations of motions as previ-
ously in (3.1Ob). Notice, however, that the anticommutation relations (3.16)
do not possess the same algebraic properties as the commutation relations do.
The commutation relations (3.8) possess the same algebra as the classical
Poisson brackets

~(BABB _ BBBA)
A B
{ ,}q,p
=
8. Bqi BPi Bqi BPi .
(3.18)

We elaborate on this point in some more detail.

The canonical momenta Pj and the corresponding coordinates qj obviously
fulfill the equations
{qi,Pj}q,p
{qi, qj }q,p
Different canonical variables Qi and Pj, once introduced, also have to obey
{Qi, Pj}q,p
{Qi,Qj}q,p (3.19)
These are the fundamental Poisson bracket relations. They are independent
of the set of coordinates and momenta qi, Pj; therefore we can skip the index
at the curly brackets. We observe from (3.8) and (3.16) that the commutators
(3.8) as well as the anticommutators (3.16) formally obey the same rules.
However, the algebraic properties of the Poisson brackets, which we will list
in the following, resemble those of the commutators (3.8) only and not of the
anticommutators (3.16). As a result of the definition (3.18) we calculate
 3. Noninteracting Fields 85

{A,B} -{B,A}
{A,e} 0, if c is a number,
{(AI + A 2 ), B} (3.20)
{AIA2' B} + Al {A2' B} ,
{AI, B}A2
{A, {B, e}} + {B, {e, A}} + {e, {A, B}} = 0 (Jakobi identity) .
Here we have paid attention to the ordering of the various quantities to in-
clude possibly noncommuting quantities. We will demonstrate in Exercise 3.1
that the commutators (3.8) do indeed fulfill the equations analogous to (3.20),
whereas the anticommutators (3.16) do not fulfill these equations. Hence,
we conclude that the anticommutators do not have a classical analogy. This
should not lead to a misunderstanding: the number operator N = L:n Nn
and the Hamiltonian iI = L:n EnNn have classical limits because both are
bilinear in b~ and bn and commute with each other. This shows that the anti-
commutation relations represent something typically new, which appears only
in quantum mechanics. We will see soon that they incorporate the Pauli prin-
ciple, which does not exist on a classical level. These conclusions are supported
by the following physical observation. To become measurable classically a field
amplitude has to be very strong; it must be possible to have a large number
of particles in the same state so that their fields can add up coherently (see
Sect. 1.5 on coherent states). This implies that particles that give rise to clas-
sical fields have to obey Bose-Einstein statistics. We conclude, for example,
that light quanta have to be Bose particles because strong electromagnetic
fields can be produced and measured classically. For the case of electrons in
a metal, which obey Fermi~Dirac statistics, only quantities like the energy,
charge, or current density are measurable classically; these quantities are bilin-
ear combinations of the field amplitudes (operators b+ and b). The amplitude
of the electron field alone is linear only in b+ or b and cannot be measured
classically.

EXERCISE

3.1 Do the Commutators and Anticommutators

Fulfill the Poisson Bracket Algebra?

Problem. Show that the commutation relations (3.8) but not the anticom-
mutation relations (3.16) fulfill the algebra formulated in (3.20).

Solution. For the commutators [A, B] = AB - BA we calculate explicitly:

1) [A,B] = AB - BA = -(BA - AB) = - [B,A] ,
2) [AI + A 2, B] = (AI + A2)B - B(AI + A2)
= AlB - BAI + A2B - BA2
= [AI, B] + [A2' B] ,
86 3. Noninteracting Fields

Exercise 3.1. 3) [AI, B] A2 + Al [A2' B] = (AlB - BAI)A2 + A I (A 2B - BA 2)

= AIBA2 - BAIA2 + AIA2B - AIBA2
= AIA2B - BAIA2 = [AIA2' B] ,
4) [A, [B, C]] + [B, [C, All + [C, [A, Bll
= A[B,C]- [B,C1A+ B[C,Al- [C,A1B +C[A,Bl- [A,BJC
= A(BC - CB) - (BC - CB)A + B(CA - AC)
- (CA - AC)B + C(AB - BA) - (AB - BA)C
=ABC-ACB-BCA+CBA+BCA-BAC
-CAB +ACB + CAB - CBA - ABC + BAC = o.
Indeed, the commutators fulfill the algebra (3.20). Now we consider the anti-
commutators {A, B} = AB + BA:
1) {A,B} = AB + BA # -{B,A} = -(AB + BA),
this relation does not hold!
2) {AI + A2, B} = (AI + A2)B + B(AI + A2)
= AlB + BAI + A2B + BA2 = {AI, B} + {A2' B} ,
3) {AIA2' B} = AIA2B + BAIA2' whereas
{AI, B}A2 + Al {A2' B} = AIBA2 + BAIA2 + AIA2B + AIBA2
= {AIA2' B} + 2AIBA2 # {AIA2' B} ,
this relation does not hold either!
4) {A, {B,C}} + {B, {C, A}} + {C, {A, B}} # 0,
since the anticommutators yield only terms with positive sign,
which do not cancel.

(3.21 )
The term b~b~bnbn vanishes because (3.16) means it is antisymmetric with
respect to the exchange of the first two (or the last two) operators. From
(3.16) it follows that
~t ~t_ _ ~
bnbn - 0 - bnb n .
A

In shorthand notation, relation (3.21) can be written as N~ = Nn . With nn

we denote the eigenvalue of b~bn; then
A At A
Nnlnn ) = bnbnlnn ) = nnlnn) (3.22)
holds and due to (3.21) it follows that
2 ~t ~ ~t ~ 2
= bnbnbnbnlnn) = nnlnn) (3.23)
A

Nnln n )
 3. Noninteracting Fields 87

Therefore n; = nn, so that

or (3.24)
We realize that for the case of Fermi statistics [anticommutation relations
(3.16)] at most one particle can appear with the wavefunction 1/Jn(x). We
say: only one particle can occupy the state l1/Jn) for the case of Fermi-Dirac
statistics. This is the Pauli principle. It stems from the introduction of the
commutation relations (3.16) (for fermions). They represent the fundamental
element of the theory.
We add that generalizations of this statistic exist. They were introduced by
Green,2 Greenberg and Messiah,3 and Ohnuki and Kamefuchi 4 and are called
parastatistics. For example, it is possible to bring in s particles into one state;
we call this parastatistics of degree s. In its limiting cases, it can represent a
Bose as well as a Fermi statistic. The fundamental commutation relations are
not given by twofold commutators but by threefold commutators:

[iii, [iij, iikJ 'fL = 2Dij iik . (3.25)

A particle that obeys parastatistics can be thought of as a composite particle

with intrinsic degrees of freedom. For example, parastatistics has been applied
to atomic nuclei, which have been interpreted as par abo sons (with integer
spin) or parafermions (with half-integer spin), respectively.5 In this respect
the transition from even-even nuclei as pure bosons to complex particles with
intrinsic degrees of freedom can be described ("loss of the bosonic character" ).
Intrinsic degrees of freedom are not important for even-even nuclei as initial
products; they are "frozen". For the final products these intrinsic degrees
of freedom are "open". We shall further discuss this point in the following
exercises.

EXERCISE

3.2 Threefold Commutators from an Expansion

of Paraoperators

Problem. Often it is convenient to expand the paraoperators into usual op-

erators, i.e.
s
o'i = L:iii(a) ,
a=l

where s determines the degree of the parastatistics. Here the operators &i(a)
obey the standard commutation relations
2 H.S. Green: Phys. Rev. 90 (1953) 270.
3 O.W. Greenberg, A.M.L. Messiah: Phys. Rev. 138 (1965) B1155.
4 Y. Ohnuki, S. Kamefuchi: Phys. Rev. 170 (1968) 1279; Ann. Phys. NY 51 (1969)
337; Ann. Phys. NY 78 (1973) 64; Prog. Theor. Phys. 50 (1973) 258.
5 See for example H.J. Fink, W. Scheid, W. Greiner: J. Phys. G 3 (1977) 1119.
88 3. Noninteracting Fields

Exercise 3.2. [ai(a),aj(a)]

} for the para-Bose statistics
{ai(a), aj(j1)} o for ai-j1
or

{ai(a), a}(a)} 8ij , } for the para-Fermi statistics.

[ai(a), aj (j1)] o for a i- j1
Show that the threefold commutators for the paraoperators follow from these
commutation relations. For convenience we have denoted the anticommutator
by curly brackets: {,} = [,1+ .

tt
Solution. First we show for the para-Bose case that

[a" (a), il,}] ~ [~a,(n), (al(p), a,(o)) 1

[~a,(n), ~ (al{P), a,(p))1
(because {ai(a), a} (j1)} = 0 for a i- j1)
L [ai(a), a} (j1)ak(j1) + ak (j1)aj (j1)]
0.(3

L {[ai(a), aj(j1)ak(j1)] + [ai(a), ak(J1)iij(J1)]}

0,(3

L {a j (j1) '[iii (a)~A:k(j1)f + [iii (a), ii} (j1)] al (/1)

0,(3

+ii,(P) [a, (a), a) (P)] + ja,(a)::'(Pljaj(p)}

L {8ij 8 (3a k(j1) + 8ij 8 (3ak(j1)}

a 0

0,(3
s

28'j Lada)
a=1

and consequently

[iii,{iij,iik}] = 28 ij iik.
An analogous proof holds for the para-Fermi statistics
 3. Noninteracting Fields 89

Since Exercise 3.2.

[ai(a),a;U3)] = - [a;(3),ai (a)] = 0

for a i= (3 it follows that

[a" [aJ,a,]] ~ [t,a,(a),t, [iiJ(~),iik(m] 1

L [ai(a), a; ((3)ad(3) + ak((3)a;((3)]
a,,6=1

L {[ai(a), a; ((3)ak ((3)] + [ai(a), ak((3)a}(f3)]}

a,,6

L {a}((3)1ai(a)~":k((3)1+ [ai(a),a}((3)] ak((3)

a,,6

+<ik(~) [a,( 0), iiJ (~)] + jo,( o~:,(mlaJ (~) }

L {6ij 6 ,6ak((3) + ak ((3) 6ij 6k,6 }
a
a,,6
s
26ij L ak(a) ,
a=1
i.e.

EXERCISE

3.3 More on Paraoperators: Introduction of the Operator Cjk

Problem. In order to simplify the threefold commutators we introduce the

operator Cjk:

-Cjk + s6j k for para-Fermi statistics,

+Cjk + s6 jk for para-Bose statistics.

Here the anticommutator is denoted by curly brackets: {,} = J, 1+. What

do the threefold commutators (3.25) look like with the help of jk ? Use the e
expansion of Exercise 3.2 for ai = 2::=1 ai(a) to derive an explicit represen-
tation of Cjk and use this representation to discuss the action of Cjk on 10).
90 3. Noninteracting Fields

Exercise 3.3. Solution. From

[aj,al] = -ejk +SOjk
and (3.25) it follows for para-Fermi statistics that
=0
, ,~
[ai, Cjk - SOjk] = [ai, Cjk ] - S[ai, Ojk]
= [ai, ejk] = 20ikaj .
Analogously we deduce for para-Bose statistics that
{aj,an = e jk + SOjk,
[ai, ejk + SOjk] = [ai, ejk] = 20ikaj .
With the use of the operator ejk the commutators for the para-Bose and
para-Fermi statistics become identical.
In order to obtain a representation for jk we transform e
•
ejk = SOjk - [aj,al] = SOjk - L [a j (a),alU3)]
0<,(3 ' - - - v - - - '
=0 for 0<#(3

•
SOjk - L (aj(a)al(a) - al(a)aj(a)
0<=1

•
SOjk - L {{aj(a), al(a)} - 2al(a)aj(a)}
0<=1
• •
SOjk - L Ojk +2 L al(a)aj(a) ,
0<=1 0<=1
'-v---'
.liik

i.e.

L al(a)aj(a)
8

ejk 2
0<=1
for para-Fermi statistics.
The result for para-Bose statistics is the same. From
aj(a)IO) = 0
it follows that
ejklO) = O.
Furthermore,
(olal(a) (ak(a)IO))t
0,
 3. Noninteracting Fields 91

so that Exercise 3.3.

(OIG jk = O.
These properties become important for Exercise 3.4!

EXERCISE

3.4 Occupation Numbers of Para-Fermi States

Problem. With the results derived in Exercises 3.2 and 3.3 discuss the oc-
cupation numbers of the para-Fermi states:
XN = (al)NIO);
the degree of the para-Fermi statistics is given by s. Hint: Examine the norm
of the state!

Solution. We suppress the quantum number k because it does not turn out
to be necessary. The norm of the state XN is given by

i.e.
IlxN11 = (Ola N(at)NIO) .
First of all, we consider the simple cases with N = 1,2, .... Furthermore, it is
convenient to pull a factor out of the matrix element:

IlxN11 =
s
~ (Ola N(at)NIO) .
For N = 1 we get

IlxI11 = ~(OlaatIO).
s
We employ the operator G:
[a,a t ] = -G+s,
i.e.

IlxI11 = ~(Olata
s
- G + slO) .
Since 0,10) = GIO) = 0 and (010) = 1 it follows that
1
IlxI11 = -(OlsIO) = (010) = 1.
s
Hence, we deduce
IlxI11 = 1.
The N = 2 case yields

IIx211 ~(OlaaatatIO)
s s
12 (Ola(ata - G + s)atIO)
12 (OlaataatIO) - 12 (0 IaGa t 10) + ~ (OlaatIO)
s s s~
=1
92 3. Noninteracting Fields

Exercise 3.4. Because [a, Cj = 2a we can make the transformation

Ilx,,11 = 1+ 12 (Olaat(ata - C + s)IO) - ~(OI(Ca + 2a)a t I0)

8 s
= 1+ 12 (0Iaa tat aI0)_12 (0ICaatI0)_12(012aatI0)
8 '--v--' s ~ s
=0 =0
2
= -2--
s

For N = 3 we find
IIx311 13 (OlaaaatatatIO) = 13 (Olaa lata - C + s] atatlO)
8 8

~(OlaaataatatIO)
83
- ~(OlaaCatatIO)
S3

+ ~(OlaaatatIO)
2 8
~
= IIx211

13 (Olaaa t lata - C + 8] atlO)

8

- 13 (Ola [Ca + 2aj atatlO) + IIX211

8

~(OlaaatataaIO)
S3
- ~(OlaaatcatIO)
s3

+ 12 (OlaaatatIO) - 13 (OlaCaatatIO)
8 s
~
= IIx211
2
-3(0IaaatatI0) + IIx211
8

~(Olaaatat
3
lata - C + 8] 10) - ~(OlaaatCatIO)
3
8 8

+21Ix211- 813 (01 [Co' + 2aj aatatlO) - ~llx211

8

~(OlaaatatataIO)
83
- ~(OlaaatatcIO)
s3

+ 12 (OlaaatatIO) - 13 (OlaaatCatIO) + 211x211

s 8

- 13 (OICaaatatIO) - 23 (0 Iaaa tat 10) - ~llx211

8 8 8

= 311x211- 13 (OlaaatCatIO) - 2~IIX211.

8 s
Since CIO) = 0 it holds that (CIO)) t = (Olct = 0 = (OIC; therefore C = ct,
so that Cat = atC + 2a t . Hence, we derive
 3. Noninteracting Fields 93

311x211- 13 (Olaaa t
s
[a t6 + 2a t] 10) - 2~llx211
s
Exercise 3.4.

311x211- ~ (0Iaaa tat610) -~(OlaaatatIO) - 2~llx211

s 3 ' - . - - - ' s3 s
=0

311x211- 3~llx211 = 3 (1-~) Ilx211,

i.e.

From these three examples we discover the general formula

IlxN11 = N!l (l_~) ... (l_N;l).

This can be proven by induction, but this proof is very tedious!
The norm of a state has to be positive definite, i.e. IlxN II > O. We recognize
from the general relation for IlxN11 that

IlxN11 > 0 <==? (l_N;l) > OJ

thus,
N ::; s.
We conclude that a state of para-Fermi statistics of the degree s contains at
most N = s particles. For states with N > s the norm vanishes or becomes
negative!

We return to the Fermi-Dirac statistics and determine the matrix elements

of bt and bn . We begin with
t
Nnlnn) = bnbnlnn) = nnlnn) ,
A A A

(3.26)
where nn = 0 or 1 because of (3.24). We examine the vector
At
bnlnn)
and apply the operator Nn = btbn to it. This gives

(3.27)
This reveals that btlnn) represents an eigenvector of btbn with eigenvalue
1 - nn. Then we note
(3.28)
94 3. Noninteracting Fields

The constant Cn can be determined from the scalar product

A At
(nnl bnbnlnn )
At A
(nnl 1 - bnbnlnn )
(nnl (1 - nn)lnn)
1 - nn = ICn l2 (3.29)
and results in
Cn = e io "v1 - nn == 8 nV1 - nn . (3.30)
Here the phase factor 8 n = eion is of modulus 1. We now consider the state
bnln n ); in complete analogy to (3.27), it holds that
At A A
bnbnbnlnn )
't
(1 - bnbn)bn Inn)
A A

bnlnn ) - bnNnlnn )
(1 - nn)b n Inn) . (3.31)
Obviously bnln n ) is also an eigenvector of N n with eigenvalue 1 - nn, which
means
bnlnn ) = Dnl1 - nn) (3.32)
with

Thus
D n -- e io~ ynn
~ =
- 8'nynn,
~

where the phase factor a~ = an can be chosen as in (3.30). Therefore we are

allowed to write
Dn = eiD:ny'n;; = 8 n y'n;; (3.33)
and (3.32) becomes

bnln n ) = eio "y'n;;11 - nn) = 8 n y'n;;11 - nn) . (3.34)

Owing to (3.9), the most general fermion state I· .. ,nn,· .. ,nn',· .. ) factorizes
into one particle fermion states In); hence,
(3.35)
This relation is in complete analogy to (3.14) for the case of bosons.
For the boson case (3.14) the occupation numbers are nn = 0,1,2, ... ,
whereas for the fermion case (3.35) the occupation numbers remain as 0 or 1
because of (3.24), i.e. nn = 0,1; this is the principal difference between the
two cases. In order to distinguish the differences due to the different quanti-
zation more clearly, we list again the most important relations for bosons and
fermions:
 3. Noninteracting Fields 95

(a) bosons

bnl'" nn''') ~I···nn -1···),

(3.36)
b~I'" nn" .) ...jnn+ 11··· n n+ 1 ···).
(b) fermions

bnl' ··nn·") mi·· ·1- n n

8 ny'lJn ... ),
(3.37)
b~I" ·nn···)
For both cases bn represents an annhilation operator because bn reduces the
particle number nn by one. On the same footing b;t" represents a creation
operator for both cases. For the boson case the phase factor has been chosen
to equal 1; this is feasable because with this the boson commutation relations
At At At At
bnbn , - bn,bn 0,
bnbn , - bn,bn 0, (3.38)
A At At A
bnbn , - bn,bn 6nn ,

could be deduced from relations (3.36). In the following problem we again

explain this step.

EXERCISE

3.5 On the Boson Commutation Relations

Problem. Derive the boson commutation relation (3.8) from relations (3.36)
for the boson creation and annhilation operators.

Solution.
(a)

(bnb n , - bn,b n ) I··· nn'" nn' ... )

= (,;n;; yin;; - ,;n;; yIn;;) I ... nn-1 ... nn' -1 ... ) 0

for arbitrary I· .. nn ... nn' .. -), i.e.

(b)
At At At At
( bn bn' - bn' bn ) I· .. n n ... n n , ... )

for arbitrary I· .. nn ... nn' ... ), i.e.

96 3. Noninteracting Fields

Exercise 3.5. (c)

"t 't'
(bnb n, - bn,bn) I··· nn'" nn' ... )
(Jnn + 1Jnn + 1- FnFn) I .. ·nn·") = I .. ·nn·")
for n' = n,
(Jnn' + 1Fn - FnJnn' + 1) I .. · nn-l'" nn'+1"') = 0
for n' i=- n ,
i.e.

For the case of fermions the choice of phase factor is more complicated.
We select en = eia " in such a way that fermion commutation relations
't't , + bn,b
bnb 't 't 0,
n n
(3.39)
bnbn, + bn,bn o
follow from (3.37) with an arbitrary state I··· nn'" nn' ... ). Here it is an ex-
cellent excercise to convince ourselves that relations (3.39) do not follow from
(3.37) for an arbitrary but fixed choice of phase en = eia ". We immediately
calculate that
(bnbn, + bn,bn)I'" nn'" nn"")
=; enen, (Fnyfn;; + yfn;;Fn) I··· nn'" nn' ... )
i=- o. (3.40)
Here it was assumed that en depends only on nn and not on the remaining
occupation numbers. Evidently this strategy leads to a contradiction with
(3.39). In order to remove it we proceed in the following way: We order the
states Inn) of the system in an arbitrary but then fixed manner; for example,
(3.41 )
The operation of bk or bl
on these states is then given by (3.37), where we
choose the phase ek = + 1 or -1 depending on the number of preceding
occupied states to be even or odd. In other words: If an even number of states
nv = 1 is to be found within the states reaching from InI) to Ink-I), then we
set

bklnI···nk···) y'nklnl ... nk-I"') ,

(3.42a)
't J1 - nklnI ... nk+1 ... )
bklnI···nk···)

for even L:~:~ nv. For the case when the number of occupied states within
InI) to Ink) is odd, we fix the phase ek = -1; hence,
 3. Noninteracting Fields 97

bklnl ... nk"') -y'nklnl ... nk-l ... ) ,

(3.42b)
At
bklnl ... nk"') -vI - nklnl ... nk+l"')
for odd 2::~::~ nv. We can combine the two relations (3.42a) and (3.42b) to
give

bklnl' .. nk"')
(3.42)
At
bklnl ... nk"')

The difference between the phase choice (3.42) and the choice (3.40) attempted
before is that now the phase fh depends on the occupation numbers of the
preceding states Ink)' In (3.40) 8 k depended on the index k only. This choice of
phase builds on the construction of the state Inl ... nk ... ) out of the creation
operators bt:

The operation
bk Inl ... nk ... ) = bk (bitt (b~t2 ... (bttk ... 10)

makes the commutation of bk with all operations preceding (bl)n k necessary.

Because
bkbt = -btb k for 1J =f. k ,
a factor -1 comes up; it appears 2::~::~ nv times. This leads to

and explains naturally the choice made for the phase factors.

EXERCISE

3.6 Consistency of the Phase Choice for Fermi States

with the Fermion Commutation Relations

Problem. Show that the phase choice (3.42) is consistent with the commu-
tation relations (3.39) for an arbitrary state Inl ... nk ... ).

Solution. We compute
b1bklnl ... nk'" nl ... ) (1)
and
(2)
where Inl ... nk ... nl ... ) represents an arbitrary state. With no loss of gener-
ality we assume 1 > k for the ordering of the operators. So that the operations
(1) and (2) are not identical to zero from the very beginning, nk and nl must
98 3. Noninteracting Fields

Exercise 3.6. both be equal to 1 for the state Inl ... nk··· nl ... ). The operation (1) then
yields
hkhzinl ... nk··· nl···) = 8l8kfokfollnl··· nk - 1··· nl - 1···). (3)
It empties first state I and then state k. The operation
hlhklnl···nk···nl··-) = 8;8~fokfollnl···nk-1···nl-1···) (4)
empties first state k and then state I. Thus
~k-l
8~ = 8k = (-l)L..v nv

remains unaltered. The phase 8f = -8l differs exactly by the factor -1 from
8 l because state k has been emptied by hk before, so n~ nk - 1 = o.
Obviously the result is
(5)
or
(6)
This is exactly the second relation in (3.39). The first relation in (3.39) can
be deduced in a completely analogous manner. The same holds for
A At T
bkbl + blbk = 81k ·
Since the ket vectors Inl ... nk···) represent all possible states of the many-
particle problem, they form a complete set; consequently, the anticommuta-
tion relations (3.16) follow as operator equations from (3.42).

3.1 Spin-Statistics Theorem

We have become acquainted with two possibilities for the quantization, the
one with commutators (3.8) and the other with anticommutators (3.16). With
a Hamiltonian of the form (3.9) both possibilities lead to equations of motion
[(3.lOb) and (3.17)] for the operators, which are formally identical to the
"classical equations of motion" (3.4). This reflects a criterion to accept both
quantization procedures as possible quantization schemes. Immediately the
question has to be raised as to when to quantize with commutators and when
to quantize with anticommutators. Put in another way, when does Bose-
Einstein statistics and when does Fermi-Dirac statistics apply? When should
the Pauli principle hold and when not?
An answer exists in the general form of a theorem from Pauli: 6 Parti-
cles with integer spin have to obey Bose-Einstein statistics and particles with
half-integer spin have to obey Fermi-Dirac statistics. We presume that the
spin quantum number is already contained in the quantum number n of the
wavefunction 1/Jn for fermions.
6 W. Pauli: Phys. Rev. 58 (1940) 716.
 3.2 Relationship Between Second Quantization and Quantum Mechanics 99

Later in this series on theoretical physics we will discuss and explain this
theorem in much more detail. 7 For the moment we only indicate the deeper
roots: the energy spectra of Hamiltonians have to be bounded from below (i.e.
have to be essentially positive) and the validity of microcausality has to be
claimed.
Remark. We assume the existence of spin-1/2 particles to be fundamental;
all other particles with different spins (0,1,3/2, ... ) can be thought of as being
built up of spin-1/2 particles. Under this aspect it seems to be plausible that
spin-1/2 particles have to obey the "more fundamental" Fermi-Dirac quanti-
zation. Then all particles with integer spin have to be Bose particles because
they are build up of an even number of Fermi particles. Hence, for example
the exchange of two spin-1 particles would correspond to an exchange of two
pair's of spin-l/2 particles. If the Bose quantization held for the elementary
spin-1/2 particles themselves, then consequently all particles would be Bose
particles. Thus, we have to demonstrate that in any case spin-1/2 particles
have to obey Fermi-Dirac statistics.

3.2 Relationship Between Second Quantization

and Elementary Quantum Mechanics

The formulation of quantum mechanics (Hamiltonian, equation of motion, and

so on) in terms of creation and annihilation operators is called second quan-
tization. It was first presented in a famous paper by P. Jordan, O. Klein, and
E.P. Wigner. 8 Now, we could imagine that a quantum-mechanical equation
(for example, the Schrodinger equation) could describe more (or new and ad-
ditional) quantum-mechanical processes after second quantization has been
applied than before. Put into other words, the conventional many-particle
Schrodinger equation could yield different (more or less) results compared to
a second-quantized Schrodinger equation. This is not the case. We will show
that conventional (elementary) quantum mechanics follows from the second-
quantized theory; the former is contained in the latter. However, the second-
quantized theory is more general. It can be extended to describe processes
such as particle creation and annhilation as occuring, for example, in (3 decay,
quantum electrodynamics, and the theory of strong interactions.
We concentrate again on the expansion (3.3) of the wave field. We apply
the quantum relations (3.8) and (3.16) to the coefficients bn , so that the
coefficients bn turn into operators bn
bn --> bn ;

the field 1jJ(x, t) of (3.3) becomes a field operator ~(x, t). We write for the
field operator

7 W. Greiner, J. Reinhardt: Quantum Electrodynamics, 2nd ed. (Springer, Berlin,

Heidelberg 1994).
8 P. Jordan, O. Klein: Z. Phys. 45 (1927) 751; P. Jordan, E.P. Wigner: Z. Phys.
47 (1928) 631.
100 3. Noninteracting Fields

(3.43a)
n n
for the Hermitean conjugate field operator we get
'¢t(x,t) = L,)~7jJ~(x) = L,)~(nlx). (3.43b)
n n

Here also the bracket notation for the states 7jJn(x) = (xln), 7jJ~(x) = (nix)
has been indicated. We shall use it from time to time when it seems suitable.
Obviously, the field operator '¢(x, t) annihilates a particle at position x and
at time t; it is a linear combination of annihilation operators bn which de-
pends on position (x) and time (t). Similarily '¢ t (x, t) has to be interpreted
as an operator creating a particle at position x and time t. The commutation
relations for the field operators '¢( x, t) and '¢t (x, t) can be derived from those
A At
between bn and bn ,:

n n'

n
5(x - x'). (3.44)
Here (3.8) and (3.16) and, for the last step, the completeness relation of the
functions 7jJn(x) have been used. Analogously we find

['¢(x, t), '¢(X', t)] ± 0, (3.45a)

['¢t (x, t), '¢t (x', t)] ± o. (3.45b)

Relations (3.44) and (3.45) represent the well-known equal-time commutation

relations for field operators. "Equal-time" indicates that the field operators
are commuted at different positions (x, x') but at the same time t, t' = t. In
our lectures on field quantization 9 of this series we generalize the commutation
relations (3.44) and (3.45) with respect to t I t'. Equations (3.44) and (3.45)
can be interpreted in the following way: All over space the field operators
do not influence each other at the same time except at an equal space-time
point.
The Hamiltonian iI, (3.9), follows from the expectation value of the one-
particle Hamiltonian [- ;~ V 2 + V] [compare with (3.1)] with the field oper-
ators '¢:
iI = J x'¢t 2n:
d3 (x, t) [- V2 + V] '¢(x, t) (3.46a)

J ~ b~1P~(x)
d3 x [- : : V2 + V] ~ bn,7jJn' (x)
9 W. Greiner, J. Reinhardt: Field Quantization (Springer, Berlin, Heidelberg 1996).
 3.2 Relationship Between Second Quantization and Quantum Mechanics 101

L b~bn' Jd3x1jJ~(X) [- ; : V2 + V] 1jJn'(X)

n)n'

L Jd3x1jJ~(x)En,1jJn'(x)
b~bn'

L En,b~bn' Jd3x1jJ~(X)1jJn'(X)
n,n'

n,n'

(3.46b)
n,n' n

It is interesting that Heisenberg's equations of motion for the field operators,

- T at 1jJ(x, t) = ['
no, 1jJ(x, t), H'] _ ' (3.47)

with the Hamiltonian iI of (3.46) precisely result in the time-dependent

Schrodinger equation; this demonstates the consistency of the theory. We find
for the right-hand side of (3.47)

[~(x,t),iIL = J
[~(X,t), d3x'~t(x',t) [-2n: Vt2 + V(X')] ~(X"t)L
= J d3 x' {~(x, t)~t (x', t) [- 2n: V,2 + V(x')] ~(x', t)
-~t(x',t) [-;: V'2+V(X')] ~(X"t)~(x,t)}.
(3.48)
Employing the commutation relation (3.44),
~(x, t)~t (x', t) = o(x - x') =t= ~t (x', t)~(x, t) ,
we transform the second to last term in (3.48) into

Jd3X'~(X, t)~t(x', t) [- 2n: Vt2 + V(x')] ~(x', t)

J d3x'o(x - x') [-;: V,2 + V(x')] ~(x,t)
J ~t
=t= d 3 x' (x', t)~(x, t) [- 2n: V,2 + V(X')] ~(x', t)
[- ; : V2 + V(X)] ~(x, t)
+(=t=1) Jd3x~t(x',t) [-;: V,2 + V(x')] ~(x,t)'~(x',t)
[- ; : V2 + V(X)] ~(x, t)
+(=t=1)(=t=1) Jd3x'~t(X',t) [-;: V,2 + V(X')] ~(x',t)~(x,t).
102 3. Noninteracting Fields

We realize that in any case the second term is positive (and independent of the
statistics) and cancels the last term in (3.48). Thus, independent of the choice
of commutators or anticommutators for the quantization (Bose-Einstein or
Fermi-Dirac statistics), we get for the commutator

[~(x, t), iT] _ = [- : : V2 + V(X)] ~(x, t).

Equation (3.47) becomes
A 2
_~81/J~~,t) = [-:mV2+V(x)]~(X,t). (3.49)

Thus, Schrodinger's equation also holds for the field operators.

With the help of the field operators we are also able to define the particle-
number density operator. It seems reasonable to try the expression
n(x, t) = ~t (x, t)~(x, t) (3.50)
and to introduce the total particle-number operator

N(t) = f d3 x n(x, t) = f d3 x ~t (x, t)~(x, t) . (3.51 )

With (3.43) we immediately calculate

N(t) = f d3 x L:b~1/J~(x) L:bn'1/Jn'(X)

f
n n'

L: b~bn' d3x1/J~(x)1/Jn'(x)
n,n'
"" At A -_ "
~bnbn " A
~Nn. (3.52)
n n

This is in agreement with our previous result (3.12). Moreover,

dN
ill = -Iii [AN, HA] _ = 0; (3.53)

see the following exercise. The operator of the total particle number is a
constant in time; this is a very satisfying result.

EXERCISE

3.7 Constancy of the Total Particle-Number Operator

Problem. Show that dN /dt = o.

Solution. Since
dN
ill = -Iii [AN,HA] ,
it is sufficient to show that
 3.2 Relationship Between Second Quantization and Quantum Mechanics 103

[N,iT] = 0 Exercise 3. 7.

with

N = J d 3 x';P (x', t)1/J(x', t) ,

iT = J x;j;t
d3 (x, t) [- ; : V2 + v(X)] ;j;(x, t) .

It holds

NiT = J d3 x' d3 x ;j;t(x');j;(x');j;t(x) [- ; : V2 + v(x)] ;j;(x)

and

iT N = J d3 x' d3 x;j;t (x) [- ; : V2 + v(X)] ;j;(x);j;t (x');j;(x') ,

where we have left out the argument t. With the help of the commutation
relations
;j;(x');j;t(x) = o(x - x') ± ;j;t(x);j;(x')
NiI is transformed into

NiT = J d3 x' d3 x;j;t(x') [o(x' - x) ± ;j;t(x);j;(x')]

X
h2
[ - 2m V2 + V(x) ] ;j;(x)

J d3 x' d3 x;j;t(x') [- ; : V2 + V(X)] ;j;(x)o(x - x')

±J d3 x' d3 x;j;t(x');j;t(x);j;(x') [- ; : V2 + v(x)] ;j;(x).

We concentrate on the second term. Since

[;j;(x'),;j;(x)t = [;j;t(x'),;j;t(x)t = 0,

it follows that

J d3 x'd 3 x;j;t(x');j;t(x);j;(x') [-2h: V +V(X)] ;j;(x)

2

= J d3 x' d3 x;j;t(x);j;t(x');j;(x') [- ; : V2 + v(x)] ;j;(x)

= J d3 x' d3 x;j;t(x) [- ; : V2 + V(X)] ;j;t(x');j;(x');j;(x)

± J d3 x' d3 x;j;t (x) [- ; : V2 + v(x)] ;j;t (x');j;(x );j;(x')

J d3 x' d3 x;j;t(x) [- ; : V2 + v(x)]

x {±{;j;(x);j;t(x') - o(x' - x)}} ;j;(x')
104 3. Noninteracting Fields

Exercise 3.7. and we deduce

NfI = J d3x~t(x')
d 3 x' [- :~ V2 + V(X)] ~(x)8(x - x')

-J d3x~t(x)
d 3 x' [- :~ V2 + V(X)] ~(x')8(x - x')

+ J d3x~t(x)
d 3 x' [- :~ V2 + v(x)] ~(x)~t(x')~(x')
fIN,
i.e.

The ket vector 10) designates the vacuum state; it shall represent the state
without any real particles,lO so that

bnlO) = 0
for any bn . Because of (3.43) we immediately realize that

(3.54)
n
This result confirms that the vacuum does not contain particles. Thus, it
becomes clear that the vacuum expectation value of the field operator vanishes:
(Ol~(x, t)IO) = o. (3.55)
Hence, we expect that
(3.56)
describes a state where a particle stays at position x. To convince ourselves
we first calculate the action of the particle density operator n(x, t) of (3.50)
on this state and derive
n(x', t)~t (x, t)IO) ~t (x', t)~(x', t)~t (x, t) 10)
~t(x',t) [8(x-x')±~t(x,t)~(x',t)] 10)
8(x - x')1jJt (x, t)IO) . (3.57)
The last term of the second to last step vanishes because of (3.54). We realize
that ~t (x, t)IO) represents an eigenvector of the particle-number density oper-
ator with eigenvalue 8(x' - x). The eigenvalue is zero everywhere except at
x' = x. At this point (x' = x) the particle density becomes so large that an
integration over the vicinity of this position yields 1. Then the validity of the
following relation becomes clear:

10 Later on we will explain the concept of vacuum in more detail.

 3.2 Relationship Between Second Quantization and Quantum Mechanics 105

N~t(x,t)IO) = ~t(x,t)IO). (3.58)

It can be proven with the help of (3.57). We get

J d3x' n(x', t)~t (x, t)IO) = J d3x' c5(x' - x)~t (x, t)IO)
= ~t(x, t)IO) ,
which shows that also ~t (x, t) 10) is an eigenvector to the total particle-number
operator with eigenvalue 1. The interpretation of ~t(x, t)IO) as a one-particle
state has been justified! Similarily,
~t(XI' t)~t (X2' t)IO) (3.59)
represents a two-particle state with a particle at Xl and one at X2. Many-
particle states can be constructed in an analogous fashion. We come back to
the one-particle state (3.56) and construct a superposition of such states,

(3.60)

where the function Xl (x) is an ordinary function and should not be interpreted
as an operator! According to the rules of quantum mechanics we interpret
(3.61 )
as the probability of finding a particle within the volume d 3 x in state
~t(x, t)IO). Because of (3.57) this probability is identical to the probability
of finding a particle in d 3 X at all. Evidently IX I (x, t Wd 3 x plays the role of
the probability density, which in conventional quantum mechanics is given by
11jJ(x, tWo Now we will construct XI(X) in such a way that lXI, t) becomes an
eigenvector to fI with eigenvalue E. This leads to the eigenvalue equation
(3.62)
or

J d3x ~t (x, t) [- 211: V2 + V(X)] ~(x, t) J d3x' Xl (x')~t (x', t)IO)

= E J d 3 x' XI(X')~t(x', t)IO) . (3.63)

We transform the left-hand side into

Jd3x~t(x,t) [-211: V 2Jd3x'XI(X')~(x,t)~t(x',t)10)

+V(X)]

Jd3x~t(x,t) +[- : : V2 V(X)]

J x [c5(x-x')±~t(x',t)~(x,t)]
d3 x'XI(x') 10)

J J ~t(x,
d3 x d 3 x' +
t) [- : : V2 V(X)] 8(x - x')xI(x')IO)

Jd3x~t(x,t) +[-:: V2 V(x)] xI(x)IO). (3.64)

106 3. Noninteracting Fields

Using (3.64), we can summarize (3.63) as

f d3x~t (x, t) {[- : : V2 + V(X)] XI(X) - EXI(X)} 10) O.

Since this holds for any arbitrary ~t (x, t) it follows that

[- : : V2 + V(X)] XI(X) = EXI(X), (3.65)

This is the one-particle Schrodinger equation for XI(X), The interpretation

(3.61) and the validity of the Schrodinger equation (3.65) for Xl (x) mean the
function Xl (x) has to be identical with the one-particle wavefunction of ele-
mentary quantum mechanics. Hence, the formalism of the second quantization
yields the conventional one-particle Schrodinger equation.
In order to derive the many-particle Schrodinger equation out of the field-
theoretical formalism we introduce the many-particle state

IXn,t) = fd3XI···fd3.'rnXn(XI""'Xn)~t(XI,t) ... ~t(Xn,t)10) (3.66)

in analogy to (3.60). Then we interpret

IXn(XI,'" ,xn )1 2 d3xI ... d3xn (3.67)
as the probability of finding particle 1 in d3xI, particle 2 in d3x2, ... , particle
n in d3xn. In analogy to (3.62) we now require that the state IXn, t) obeys the
Schrodinger equation
(3.68)
where iI is given by (3.46a). This requirement leads to a condition for the
amplitude function Xn(XI, ... , xn). In order to determine it we write (3.68)
in full detail as

f d3x~t (x, t) [- 2h: V2 + V(X)] ~(x, t)

x f d3x~ ... d3x~Xn(X~, ... ,x~)~t(x~,t) ... ~t(x~,t)10)
= E f d3x~ ... d3x~Xn(X~,.··,x~)~t(x~,t) ... ~t(x~,t)10). (3.69)

The left-hand side becomes

f d3Xd3x~···d3x~~t(x,t) h2 V2+V(X) ]
[ -2m ~(x,t)
x Xn(X~,···,x~)~t(x~,t).··~t(x~,t)IO)

f d 3x d3x~ ... d3x~~t (x, t) [... J Xn(x~"", x~)

x ~(x, t)~t (Xl, t)··· ~t (X~, t)IO)

f d3Xd3x~ ···d3x~~t(x,t)[···JXn(X~'···'X~)
x [O(x - xD ± ~t (X~, t)~(x, t)] ~t (x;, t) ... ~t (X~, t)IO)
 3.2 Relationship Between Second Quantization and Quantum Mechanics 107

J d 3x d3x~ ... d3x~,(f;t (x, t) [... j Xn(x~, ... ,x~)

X {O(X - X~),(f;t (x;, t)· .. ,(f;t (X~, t)IO)

±,(f;t(X~, t) [O(X - x;) ± ,(f;t(X~, t),(f;(X, t)] ,(f;t(X~, t)··· ,(f;t(X~, t)IO)}
Jd3xd3x~ ... d3x~,(f;t(X, t) [... j Xn(X~"" ,X~)
X { l)±l)i-lO(X - xD,(f;t (X~, t)
,
X ... ,(f; t (x~_l , t),(f; t (X~+ 1, t) ... ,(f; t (X~, t) 10) }

Jd3X~ d 3x; ... d3X~ l)±l)i-l,(f;t(X~, t) [- : : V: 2 + V(X D]

,
X Xn(x~, ... ,x~),(f;t(x~,t) ... ,(f;t(x~,t)IO). (3.70)
Observe that within the last expression, all operators ,(f;t(x~, t) appear
on the far right except for the operator ,(f;t (x~, t) with argument x~. This
field operator appears left of [... J. All operators ,(f;t(x~, t) to the right of
[... J Xn(Xl ... Xn) can be shifted to the left of [... j without any restriction.
This gives

J d 3xi··· d3X~ l)±l)i-l,(f;t(x~, t),(f;t(x~, t)··· ,(f;t(X:_l' t),(f;t(X:+l' t)···

,
x ,(f;t(x~,t) [_Ii::? + V(x D] Xn(xi,'" ,x~)IO).
If we shift the operator ,(f; t (xD appearing on the left within this expression to
its "normal" place between ,(f; t (X~_l , t) and ,(f; t (x;+ l' t), we have to make i-I
permutations. This produces a factor (±l)i-l. Finally we get

Jd3x~ ... d3x~ 1)±1)i-l(±1r- 1 ,(f;t(xi, t)··· ,(f;t(X:_l' t),(f;t(x;, t)

,
x ,(f;t(x:+1,t) ... ,(f;t(x~,t) [ - \:i
2 2'
+V(x~) Xn(xi,···,x~)IO)
]

Jd3X~' .. d3x~,(f;t(xi,t) ... ,(f;t(x~,t)

XL, [ _\:i 2 2'

+V(X')Xn(x~,···,x~)IO).
]
(3.71)

It is remarkable that all distinctions have vanished as a result of the various

± signs. With the help of (3.71), we can write (3.69) in the form
J d 3 xi ... d3x~,(f;t(xi, t)··· ,(f;t(x~, t)

x { [~( - Ii::t) + V(x~) - E1Xn(x~,···, x~)} 10). (3.72)

108 3. Noninteracting Fields

This has to hold for arbitrary states ~t (x~, t) ... ~t (x n , t) 10); hence, we deduce

(~{ - :: v;Z + V(X~)} ) Xn(X~,···, x~) = EXn(Xl,···, xn) . (3.73)

This represents an n-particle Schrodinger equation for the amplitude

Xn(Xl, ... , xn). We realize that the formalism of second quantization also
incorporates the elementary quantum mechanics of an arbitrary number of
nonintemcting particles. Since the field-theoretical formulation, i.e. the second
quantization, clearly expresses the quantum nature of the processes (creation
and annhilation of particles) as well as the wave character (particles are an-
nihilated and created in specific states Xi) it should be acknowledged as the
true and complete formulation of quantum mechanics.
4. Quantum Fields with Interaction

We examine now the Hamiltonian of two (or more) free particle fields and
their mutual interaction. It appears to be obvious to simply add the Hamil-
tonians of the free fields and to introduce their interaction suitably. We have
already come across the interaction between charged particles and photons.
In the following we elaborate on how this well-known system is completely
described from the field-theoretical point of view, i.e. in the language of sec-
ond quantization. With the introduction of creation and annhilation operators
aL., aka for photons the radiation field itself has already been treated field-
theoretically. But we now also want to express the particle processes with
particle creation and annihilation operators b~ and bn . The noninteracting
particle field will be described with Hamiltonian (3.46) and the photon fields
through Hamiltonians (1.40) or (1.52). The interaction (2.6) between the two
fields is introduced via minimal coupling:
, ,e ,
p ---7 p - - A .
c
We insert this replacement directly into (3.46) and obtain the total Hamilto-
nian by addition of the mentioned terms:

H = J d3x~t(x,t) [2~ (fJ - ;A)2 + V(x)] ~(x,t)

+ J d 3x ~(E2
87r
+ 13 2)
Hmp + Hrad + Hint. (4.1)
Here Hmp represents the Hamiltonian of the particle field (many-particle
Hamiltonian) ,

Hmp = Jd3x~t(x,t)[-:~V2+V(X)]~(X,t), (4.2)

and Hrad represents the Hamiltonian of the radiation field,

3 8~ (E2 + 13 2) =
H rad = J d x L( Wk + ~) ata ka . (4.3)
k,cr
The Hamiltonian for the interaction between the particle field and the radia-

, J
tion field is given by

Hint = [en '

d 3x'ljJ't (x, t) --.-A· V
2
+ -e22A2
, ] ,
'ljJ(x, t). (4.4)
Imc mc

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
110 4. Quantum Fields with Interaction

The particle-field operators ¢, ¢t are given by (3.43) and are an expression of

the particle annihilation operators bn and creation operators b~. Analogously
to (2.8) and (2.9), (4.4) also splits into parts H{nt and H{~t, which are pro-
portional to A and A 2, respectively. Substituting into (4.4) the expansion
(1.37) for the vector potential field operator A in terms of photon creation
and annhilation operators,

and for the particle field operator the expansion (3.43) in terms of particle
creation and annihilation operators,
¢(x, t)
n

n
we obtain the result
Hint H'int + HI!int' (4.5)
where

and
H[~t L L L L b~ bn,€k 1 ,<71 • €k2,<72
k1,Ul k 1 ,a2 n n'

x [M(kl,k2,n,n')akl,<71ak2,<72

+ M(kr, -k2' n, n')ak1,<71 at,<72

+ M( -kl' k2' n, n')aL,<71 ak2,<72
+ M( -kl, -k2, n, n')aL,<71 at,<72] . (4.7)
The matrix elements M(k,n,n') and M(k 1 ,k 2,n,n') are given by

J27r3fic JdVV;~(X) (- ~mc

M(k,n,n') 2 .eli eik.Xv) V;n'(x) ( 4.8a)
£ Wk
and
27r 2 lic 2 1
£3 J Wk1Wk2
X J dV V;~(X) 2:c 2 ei (k 1+k2)'XV;n' (x) , (4.8b)

respectively. The unperturbed Hamiltonian

Ho = Hmp + Hrad (4.9)
has eigenvectors
I··· N n ·· ')1'" nk<7"') ( 4.10)
 4. Quantum Fields with Interaction 111

with eigenvalues

L EnNn
n
+Lnwk (nkCT +~)
k,cr
(4.11)

Hence, it holds that

Hoi'" Nn ·· ,)1", nkCT"')

= (LEnNn + L nwk (nkCT + ~)) I··· Nn ·· ,)1", nkCT" -). (4.12)

n k,a

The interaction (4.5), more explicitly (4.6) and (4.7), causes transitions be-
Fig. 4.1. Scattering of a
tween states (4.10). For example, the term proportional to b~bn,akCT annihi- particle from In') to In) by
lates a photon with momentum nk and polarization (1, annihilates a particle absorption of a photon kO"
in state In'), and at the same time creates a particle in state In). We say that
through the annihilation of a photon of the kind k(1, a particle is scattered
from state In') to state In). The matrix element EkCT' M(k,(1,n,n') in (4.6)
represents the amplitude for this process. Analogously, the term proportional
to b~bn,at describes the emission of photon k(1 with simultaneous scattering
of the particle from In') to In). In the same spirit the terms contained in H{~t
can be interpreted. We illustrate these processes in Figs. 4.1-4.5.
Often the interactions (4.6) and (4.7) are written in abbreviated form:
HA ,

LLLEkCT' [M(k,(1,n,n')b~bn,akCT +H.c.] (4.13)

int
k,a n n'

fI:~t L L L L Ek1,CTI . Ek2,(T2

k1 ,0'"}k2,CT2 n n'

x [M(k 1 , (11, k 2 , (12, n, n')b~bn,ak,,(T, ak 2,(T2 Fig. 4.2. Scattering of a

particle from In') to In) by
+ M(k 1 , (11, -k2, (12, n, n')b~bn,akl(Tl at(T2 + H.c.] (4.14) emission of a photon kO"

This notion is indeed correct. For instance we consider the Hermitean conju-
gated (H.c.) part of the first expression (4.13) for H{nt explicitly:

LLLEk(T' M*(k,(1,n,n')b~,bnat·
k,a n n'

Owing to (4.8a) we point out that

Ek(T . M*(k, n, n')

( partial )
integration

J
Fig. 4.3. Scattering of a
27rnc
L3 wk
2
dV 7f;*,(x)
n
(-.;!!-
Imc
e- ik .x . v) 7f;n(x)
particle from In') to In)
by absorption (annihila-
tion) of two photons k 2 0"2
Ek(T . M( -k, n', n) . and klO"I
112 4. Quantum Fields with Interaction

From this we deduce

LLL>:kO" M*(k,n,n')b~,bnaL
k,O' n n'

k,a n n'

L L L EkO' . M( -k, n, n')b~bn,aLO' ( 4.15)

k,O' n n'

We have exchanged the summation indices in the last step. Result (4.15)
is identical to the second term of (4.6). On the same footing it is easy to
Fig. 4.4. Scattering of a show the identity of (4.14) and (4.7). The field-theoretical formalism (second
particle from In') to In) quantization) presented here clearly expresses the quantum character of the
with simultaneous emis- processes. However, it does not modify the physical content of the theory; it is
sion of two photons k 2 0'2
and kjO'j only the particle feature that is explicitly illustrated by second quantization.

Fig. 4.5. Scattering of a

particle from In') to In)
via emission of a pho-
ton k 2 0'2 and simultane-
ous absorption of another
photon k j O'j , and vice
versa. Here we can think
of particle scattering as
being caused by photon
scattering

EXAMPLE

4.1 Nonrelativistic Bremsstrahlung

This example can be solved rigorously with the methods developed and pre-
sented during the course of the first three sections. We will now approach it
with the tools of second quantization in order to demonstrate the advantages
of this formalism.
We know from classical electrodynamics that an accelerated charged parti-
cle radiates. In particular such a Bremsstrahlung process happens in the colli-
sion of two charged particles; during the collision the particles first slow down
and then accelerate again or vice versa. Bremsstrahlung occurs in various colli-
sion processes: for example, heavy-ion scattering (see Fig. 4.6), proton~proton
scattering; and electron~electron scattering.
Here we discuss electron~nucleus scattering, i.e. the collision of an electron
and an atomic nucleus and their bremsstrahlung. Because the mass of the
atomic nucleus with A nucleons is much larger than the electron mass,
mnucleon A :::::i 2000 A ,
melectron
 4. Quantum Fields with Interaction 113

recoil effects can be neglected and the nuclei can be considered as being at rest.
This approximation is not valid for nucleus-nucleus bremsstrahlung. Without
any loss of generality we place the nucleus at the origin x = 0 of our coor-
dinate system and denote the position vector for the electron as x (compare
with Fig. 4.7). The interaction potential between the electron and nucleus is
described by the Coulomb potential
Ze Ze
-TxT
2 2
V(x) = = --;:- . (1)

We treat it as a perturbation; according to (4.2) we then write for the inter-

action operator Fig. 4.6. Bremsstrahlung

Hv = Jd3x~t(x)V(x)~(x). (2)
of atomic nuclei in heavy-
ion collisions. Bremsstrah-
lung represents an im-
The states In) of the free particle are now given by plane electron waves portant background for
quasimolecular radiation
1 . from molecules temporar-
1j;q(x) = (xlq) = ,fL3 e,q·", . (3) ily formed during the col-
lision. This process has
nq represents the electron momentum. According to (3.43) and (3) the field been calculated by Rein-
hardt et al. 1 and experi-
operator ~(x, t) becomes mentally detected by J.8.
Greenberg et al. 2
~(x, t)

and
~t(x,t) (4)
q

The Hamiltonian for the (free) particles reads [see (4.2)1 e

Hmp = J dV ~t (x, t) ( - n~:2) ~(x, t)

J dV L b~1j;~(x) (- n~:2) L bq,1j;q'(x)
J
q q'

L (nq')2
~b~bq'
A A

dV 1j;~(x)1j;q'(x)
q,q'

Fig. 4.7. For electron-

nucleus scattering the nu-
cleus can be assumed to
'" (nq? At A
L..- 2m bqb q . (5) rest at the origin of the co-
q
ordinate system

Here, m represents the electron mass. Expression (5) is plausible; the total
kinetic energy is given as a sum of one-particle energies belonging to the
various states 1j;q(x) multiplied by the number operator bkbq. With (4) the
(Coulomb) interaction (2) explicitly reads
1 J. Reinhardt, G. 8off, W. Greiner: Z. Phys. A 276 (1976) 285.
2 H.P. Trautvetter, J.8. Greenberg, P. Vincent: Phys. Rev. Lett. 37 (1967) 202.
114 4. Quantum Fields with Interaction

Example 4.1.
L b~bq' JdV e; ; V(X):;;
q,q'

L b~bq,v(q' - q) , (6)
q,q'

where

V(q' - q) = J dV ~3 ei(q'-q).xV(x) (7)

represents the Fourier transform of the interaction potential V(x). For the
case of Coulomb interaction this expression can be computed in closed form:

V(q) = J dV ~3 eiq .x ( - f:~) = - J

~~2 dVeli:~
- Ze 2
£3
(-~)
q2
JIxl
dV 6e iq .x = Ze 2
£3q2
J(6~)
Ixl
eiq,xdV

- 4~~~2 J 8(x) eiq .x dV

41fZe 2
- £3q2 . (8)

Use has been made of the relation

1
6 = -41f8(x) ,
1X1 (9)

which should be familiar from classical electrodynamics. Also, a partial in-

tegration has been employed twice. The explicit expressions of the radiation
matrix elements (4.8a), and eventually also (4.8b), are needed. They can be
easily calculated with the electron wavefunctions (3). We consider only the
dominating interaction with the radiation field H{nt (4.6); the relevant matrix
elements (4.8a) are

M(k, (J, q, q') =

J21fhc
L3 wk
2
(-~)
me q' . ' 8q,q'+k'
€Ok a (10)

Hence, the interaction operator (4.6) becomes

H(mt -_ (_ me en) "" ""~~

k,a q

The sum over q' breaks down because of the Kronecker delta in (10). From
our previous investigations into the emission and absorption of photons from
free particles (see Chap. 2, (2.48) and following) we already know that these
 4. Quantum Fields with Interaction 115

processes do not exist for free particles. They are only possible in a surround- Example 4.1.
ing medium with an index of refraction n > 1 (Cherenkov radiation, Chap. 2).
Therefore bremsstrahlung must occur first as a scattering of the charged par-
ticle at the nucleus from an initial to an intermediate state and a successive
photon emission with simultaneous scattering of the particle into the final
state. It is the presence of the atomic nucleus as a force center which makes
the emission of radiation (bremsstrahlung) feasible. In other words: brems-
strahlung is a second-order process, for which H{nt' (11), as well as H v , (6),
contribute as a perturbation. We list the transition matrix element up to
second order:

M· = (11 k + HI') + L UIH{nt + Hv II) (IIH{nt + Hvl i ) (12)

ft mt V Z
I
E i - E r + ZrJ
.
The initial and final states have to be as follows:
Ii) lone electron with ql)plno photon)R
Iql)pIO)R = btIO)pIO)R, (13)
If) lone electron with q3)plone photon krY)R
IQ3)paLT IO)R = bbalO)paL IO)R . (14)
Indices P and R indicate particle and radiation (photon) space, respectively.
It is immediately seen that the first-order term in (12) cannot contribute, i.e.

UIH{nt + Hvl i ) = UIH{ntl i ) = 0,

because it leads to the problem of photon emission from free particles, as
mentioned earlier, which does not exist because of the impossibility of energy
and momentum conservation. UIHvli) does not contribute because with Hv
alone no photon is created. Thus, only the transition matrix elements of second
order,

M. _ '"' UIH{nt + Hv II) (IIH{nt + Hv Ii)

fl - Y Ei - Er + irJ '
(15)

remain for the bremsstrahlung process. There are only two possibilities for
the intermediate states 11):
Ih) lone electron with Q2)P Ino photon)R
IQ2)pIO)R = bq2 10)pIO)R (16)
and
Ih) lone electron with q;)plone photon krY)R
Iq~)paLIO)R = b~2,IO)paLIO)R' (17)
We illustrate these two processes in Fig. 4.8.
From our previous investigations we know that momentum conservation
holds at a vertex (from the Latin expression "convertere", which means "to
turn around") of the form shown in Fig.4.9 [compare, for example, with
(2.51)]. Hence we can immediately write down momentum conservation for
the vertices illustrated in Fig. 4.8 for which a photon is emitted:
116 4. Quantum Fields with Interaction

Fig. 4.8a,b. Diagrams il-

lustrating bremsstrahlung
of an electron scattered at
a nucleus. For case (a),
first scattering at the nu-
cleus into the intermedi-
ate state q2 = q3 + k
takes place because of the
Coulomb potential, which
is indicated with - X -x; ku
then the emission process (b)
of photon k, u follows with
simultaneous scattering of
the particle into the final
state q3. For case (b), first
the photon ku is emitted fiq3 + fik,
(18)
with simultaneous scatter- fiql - fik.
ing of the particle into the
intermediate state q~ =
ql - k; then the particle is Since only the two intermediate states IIv) of (16) and (17) are possible, the
scattered again, this time matrix element (15) simplifies to
into the final state q3 be-
cause of the Coulomb po- Mfi = UIH"{ntlh)(hIHvl i ) + UI Hvlh)(I2IH{ntl i ) . (19)
tential V (x) Ei - Ell Ei - El2
The sum runs over only two terms, Ih) and Ih); as a consequence, the in-
finitesimal quantity 7J is not needed in the denominator, which usually ma-
ka nipulates poles occuring eventually in the infinite sum. Expression (19) exactly
reflects the contributions from the two diagrams in Fig. 4.8, which have to be
added coherently.
With (6) and (7) the relevant matrix elements are calculated straightfor-
wardly:

q]
(btIO)pIO)RI Lb~bqIV(q' - q) Ib~110)pI0)R)
q,q'
Fig. 4.9. Electron-photon
vertex

L V(q' - q) (01 bq2b~ (8q ql - b~}ql) 10)

1

q,q'

L V(q' - q)8q'ql (0Ibq2b~10)

q,q'

L V(q' - q)8q',ql (018q2 ,q - bq2b~10)

q,q'

q,q'

(20)
 4. Quantum Fields with Interaction 117

Similarly follows Example 4.1.

A
UIHvlh) = \jAt t
bq310)pakaIO)R I

x 2:V(q - q')bFJq' Ib:zIO)paLIO)R}

q,q'

(21)

where we have used (18). With (11) the radiation matrix elements are calcu-
lated as follows:

x { (h;, 10)p Ih:+>, h, Ih;, 10)p ) '( at.IO)R i:"., 10)R;

I
+ (hl, 10)pi h,b,+" &l,lO) p) '( al.lo~~·I~;::' 10)R),}
en
me

en (22)
me
Similarly it follows that
118 4. Quantum Fields with Interaction

Example 4.1.

(23)

As we have already realized earlier in the context of (18) the two 6 functions
appearing in (22) and (23) express momentum conservation at the photon
vertices. Obviously, it also follows naturally from the formalism of second
quantization. We still have to calculate the energy denominators of (19). Since
we deal with real particles, this is an easy task:
qr
/'i 2
2m '
(24)
/'i 2 q2
Ef = __ 3 + nwk .
2m
Energy conservation for the total process is deduced from the 6 function ap-
pearing in the expression for the transition probability per unit time,

( trans. Prob.) = 271'IM'1 2 6(E _E-) (25)

time /'i fl f 1,

which is a result of the general formalism of time-dependent perturbation

theory. For our case, energy conservation reads

(26)

With this result the energy denominators of (19) can be cast into the following
form:

Here the particle velocities

 4. Quantum Fields with Interaction 119

nqi nq3
Vi = and V3 = - (29) Example 4.1.
m m
have been introduced. With (20), (23), and (28) we can now write down the
bremsstrahlung matrix element, (19) explicitly; we get

Mfi = -41lZe 2
V(q3 + k - qi)2
(-en)
mc
J21f!ic 2
VWk

XEkcr· (nw k
(1-~-~) - nw (1-~+~))
kc 2mc k kc 2mc

X Ekcr . (1 _k.~3 kc
_ ~ -l-_--;-~-.:-l-+-"""'nk,-)
2mc
-
kc 2mc
(30)

All the time [see (24)] we have assumed nonrelativistic electrons, i.e. vlc« l.
Moreover, the photon momentum nk will remain small with respect to the
particle momentum in general, so that
nk« nq = mv.
Consequently, the denominator of (30) can be replaced by 1, and the denom-
inator of the first factor of (30) can be approximated by
(q3 +k - qd 2 ~ (q3 - qi)2 .
Then, (30) simplifies to

41fZe 2 n2
Mfi (31)
Vm 21Lv 12
where
Lv = V3 - Vi (32)
represents the difference between the electron velocities in the initial and final
state.
The total cross section IJ for bremsstrahlung is defined as the sum over
all transition probabilities per unit time for transition into all possible final
states and divided by the incoming electron current
(33)

(34)

A summation over electron states on the one hand and over photon states on
the other hand has been performed already; also replacement (2.18) has been
applied:
120 4. Quantum Fields with Interaction

Example 4.1.
L
£3--+ 00
------+
L3
(27r)3
j d3k,
k

L3 j 3
(35)
L
L 3 --->00
------+
(27r)3 d q3.
q

dDe is the solid angle element in which the electrons scatter and dDk the one
for photons (see Fig. 4.10). Inserting (31) into (34) yields

a = VI1 (47rZe 2fi2)2 27re 2 27r 1

m2 -fi-Ii: (27r)6
j q32 dq3 df2e j k 2 dkdDk
IEkcr . 6v 12m 2 2
x 16v 14w3 /i:28(q3 - ql) . (36)

k,(f Since
Fig. 4.10. Solid angle ele-
ments into which the elec-
tron and photon are scat-
tered
(only positive q3 have to be considered in the integral J ... dq3) and
. rJ
16vl = 2VI sm"2 (37)

(see Fig. 4.11) we arrive at

a = 2- (47r Z e2fi2 ) 2 27re2 27r _1_ ~ q2 2m _1_

VI m2 fi fi (27r)62ql I fi2 (2VI)4

x jdD jdDk k 2 dk IEkcr' 6v 12

e w3 sin4 rJ/2

Fig. 4.11. The 8(q3 - ql)

function guarantees q3 =
Z2 e4 ~
m2vtl67r2fi
j j
dD
e
dDk k 2 dk IEkcr . 6v 1
W3 sin 4 rJ/2
ql and thus also IV31 =
IVII. Since we have ne- Z2 e4 e2
m 2 vt 167r 2c 3 fi
j j dDe dD k dw IEk,cr . 6vl .
w sin4 rJ/2
(38)
glected the photon energy
of the final state this re-
The differential cross section for scattering of an electron into the solid angle
sult describes elastic scat-
tering of the electron. Di- element dDe with angle rJ and simultaneous emission of a photon with po-
rectly from the figure we larization Ekcr in the frequency domain between wand w + dw into the solid
deduce l.6vl = 2VI sin 11/2 angle element dDk can be read off conveniently:
d3 a Z2 e4 e21Ekcr . 6v 12
(39)
dDedf2kdw - m 2Vfsin 4 rJ/2 167r 2c3fiw
The first factor represents Rutherford's scattering cross section for elastic elec-
tron scattering at the nucleus (consult Exercise 4.2). Apparently the second
factor in (39) has to be the probability dW /dDk dw that a photon with fre-
quency w will be emitted into dDk. The probability for two events to occur at
the same time is equal to the product of the single probabilities. Hence, the
differential cross section (39) takes on a very plausible form:

d3 a ( da ) dW (40)
df2e dDk dw = df2e Rutherford df2k dw .
 4. Quantum Fields with Interaction 121

From (39) we unveil that the energy spectrum of the photon behaves like Example 4.1.
dw / w; as a consequence the probability of a photon being emitted with en-
ergy zero appears to increase to infinity. This behavior also shows up in the
relativistic theory of bremsstrahlung; it is known as the infrared catastrophe.
In order to better understand and remove this unreasonable result a careful
analysis of the actual experimental conditions is required for the observation
of bremsstrahlung. The crucial point is that every counting device possesses -----~ Ze
only a finite energy resolution. If it detects an electron scattered inelastically
into a finite energy interval around w = 0, then it will also detect an electron -----~ Ze
scattered elastically. Therefore we have to consider the elastic as well as the
inelastic scattering cross section for a comparison with experiment; both pro-
cesses, of course, up to the same order in e 2 . Since the bremsstrahlung part of
(39) is of second order in e 2 relative to elastic scattering the so-called radia-
tion corrections up to order e 2 have to be taken into account for the electron
scattering cross section

(d~J elastic .

Two different contributions exist. On the one hand, graphs of the form given
in Fig. 4.12 describe the second-order scattering of electrons in a Coulomb
potential (confer Exercises 4.2 and 4.4). Fig. 4.12. Second-order
On the other hand, a second-order interaction of the electron with itself Coulomb scattering
exists via the radiation field; the graphs depicted in Fig. 4.13 give an illus-
tration. They depict a virtual photon, which is first created at the electron
site and then falls back again to the electron like a boomerang. This is in
contrast to the graphs depicted earlier for which the second photon also runs
to the source (Ze); consult again the discussion about self-energy at the end
of Chap. 1. A systematic calculation of these graphs within relativistic theory
shows that the contributions from radiation corrections contain a divergent ---~Ze
term, too, which exactly cancels the divergence (39) of bremsstrahlung at
tu.J = O. Hence, the infrared catastrophe does not exist in reality!

EXERCISE

4.2 Rutherford Scattering Cross Section

Problem. Derive Rutherford's cross section for the scattering of an electron

at a fixed point nucleus with charge Ze.
---~Ze

Solution. In first quantization the Hamiltonian reads

p2 Ze 2
H = ---.
A

2m r
As discussed earlier, in second quantization the interaction takes on the form
Fig. 4.13. Radiation cor-
A _
Hv -
' " ' " At A
- ~~bqbq'
q q'
471' Ze 2
Plq_q'1 2
_
- ----v- ~~bq-kbqk2'
471' Ze 2
q k
'" '" At A 1 rections to Coulomb scat-
tering
122 4. Quantum Fields with Interaction

Exercise 4.2. We consider a transition from a particle state Ii) = bklO) to the final state
If) = bt,IO). tip and tip' denote the particle momenta before and after the
collision, respectively. The transition matrix element becomes

Mfi
A
= UIHvl z) -
. _
--v
47rZe 1 A At A At
2 ,",
L k (Olbp,bq_kbqbpIO)
q,k
2

--v
47rZe 1
2 ,", At A A At
L k2 (01 (5 p'q-k - bq_kbp' )bqbpIO)
q,k
41l'Ze 2 ,", 1
- -V L 5p 'q_k 5pq k 2
q,k
47rZe 2 1
--V(p_p,)2 .
To obtain the cross section one has to sum over all final momenta p' and

!
divide by the electron current np/mL3;
L3 m L3 12 I 27r 167r 2 (Ze 2)2 1
(J = np (27r )3 p dp dS?P'1i L6 (pI _ p)4

X 5 (ti 2p2 _ n2p/2 )

!
2m 2m
4mZ 2e4 2m p/2 dp' dS? , 5(p2 - p/2)

! !
ti2p ti2 P (pI _ p)4

4m 2 Z2 e4 p'2 5(p _ p') I

dS?, - dp
+ p'2 -
!
n4 P p2 (p2 2ppl cos 7'J)2
4m 2 Z 2 e4 1
d S? ,--,-----------:-:-;;-

2!
ti4 p (1 - cos 7'J)2

Z2 e dS?
4E2 4sin 4 7'J/2 .
Thus, we arrive at the well-known result
d(J) Z2 e4 1
(
dS? Rutherford = 16E2 sin4 7'J /2
for the differential cross section. Rutherford's scattering cross section diverges
for small scattering angles. Performing an integration over the solid angle we
realize that the total scattering cross section (J = f (d(J / dS?)dS? diverges, too.
This stems from the strong singularity'" 1jt'!4 at 7'J = 0; this singularity has
the same origin as the infrared catastrophe we have already experienced in a
calculation of bremsstrahlung.
 4. Quantum Fields with Interaction 123

EXERCISE

4.3 Lifetime of the Hydrogen 2s State with Respect

to Two-Photon Decay (in Second Quantization)

Problem. Calculate the lifetime of the 2s state of the hydrogen atom with
respect to two-photon decay (compare with Exercise 2.10) with the methods
of second quantization.

Solution. The relevant Hamiltonians already derived are given by

k,u n,n'
and (1)
iI"

with matrix elements

M(k, cr, n, n')

(2)
H' produces two photons in second order, whereas H" directly produces two
photons. No other term in (4.7) contributes to the two-photon process. In the
initial state the electron occupies the 2s state and no photon (radiation) is
present:
(3)
In the final state the electron occupies the Is state and two photons exist with
quantum numbers (k r , crr) and (k~, crD:

If) = btsiil f!
II
f
iikt ,
f'O"(
,10)eI0)rad' (4)
With the initial and final states given we can calculate the matrix elements
(2) of the Hamiltonians (1)

MN) UliI"li)
rad (01 e(Oliik;,II; iikf'"fblsH"b~s IO)e 10)rad
L L LM(-kl,crl,-k2,cr2,n,n')
124 4. Quantum Fields with Interaction

Exercise 4.3. Using the commutation relations

AAt AtA s:
aiaj - ajai Vij ,
A At AtA
bib j + bjbi 8ij
we obtain
A At A t t At
(Olak'f' U'akf
f '
ufblsbnbn,ak 1,vl _ b2s lO)
_ ak2,v2
A A AtAt A At A At
= (Olak;,u;akf,Ufakl.Ul a k2 ,(2 0 ) (0Iblsbnbn,b2sI0)
1

for the matrix element of operators. After commutation the photon part be-
comes:
(Olak'pUr,(atk},Ul ak f,Ue + 8k k 8O"l,O"f )
1, f at
k2,U2 10)

For the electron part it follows:

(Olblsb~ (8 n ' ,2s - b~sbn') 10)

8n , ,2s (Olblsb~ 10)
At A
8n',2s(018n,ls - bnb1sl0)

i.e. the electron has to occupy the 2s state as its initial state and the Is state
as its final state. Then we get

Mr\l) = L L LM(-k1,0'1,-k2,0'2,n,n')8n',2s 8n,ls

X + 8k k,8 ,8k k 8 ,at )

f 0"1,(7 f,
(8 k2, kf 80'"2,Ut 8k1, k,8 2, f a2 ,U f 1, f 0"1

M( -k;, 0';' -k[, 0'[, Is, 2s) + M( -k[, O"f, -k;, 0";, Is, 2s) . (5)
In order to calculate the matrix element of second order we need the interme-
diate states
(6)
for which only one photon is present (H' can produce only one photon in first
order). The matrix element reads

(7)
 4. Quantum Fields with Interaction 125

We evaluate the matrix elements separately: Exercise 4.3.

UIH'lz) (0Iak;,a;ak"a,b1sH' aLa}! 10)
LLM(-k,O",n,n')
k,a n,n'

k,u n,n'

Furthermore,

and

(Olak;,a;ak"a,a!,aaLa z 10)
(Olak;,a; (aLak"a, + c5k,kAj,a,) aLaz 10)
(olak;,a;aLak"a,aLaz 10)
+ c5k,k,c5a,a, (Olak;,a;aLaz 10)
(Olak;,a;aL (aLaz ak"a, + c5k.,k,c5a.,a,) 10)
c5k,k,c5a,a, (Olc5k.,k; c5a.,a; + aLaz ak;,a~ 10)
Z)
, 'atk,a 10) + 15k , kf 15
15 k kf 15 O'z,CTr (Ola kflat (T,fff
15 k k'c5
%1 f aZ)a r,

f a,a r' + at
15k kf 15O'z,O'r (Olc5 k, k'c5
Z) k,O' ak'f,u t,10)

+ c5k,k, c5a,a, c5k.,k; c5a.,a;

c5k.,kfc5a.,afc5k,k;c5a,a; + c5k,kfc5a,a,c5k.,k;c5a.,af ,
so that
L L M( -k, 0", n, n')c5n ,zc5n ,ls
k,o: n,n'

X (c5k.,kfc5a.,afc5k,k;c5a,a; + I5k,kfI5a,a,I5kZlk;I5a"a;)
LL M( -k, 0", n, n')c5n , ,zc5n ,ls
k,a n,n'

(8)
126 4. Quantum Fields with Interaction

Exercise 4.3. Analogously we derive

k',a'm,m'

k',(j' m,m'

It holds

(Oluk""zul',,,,IO) (Olbk"k,b""", + uL,Ak""z 10)

Ihzok' b" zo'"
and
A At A At A At
(0Ibzbmbm,b2sI0) = bm',2s(0Ibzbm I0) = bm',2s bm,z,
so that
(zIH'li) L L M( -k', a', m, m')bk zo k,b""",bm',2s bm,z . (9)
k' ,er' m,m'

Hence,

X [bkzok,b""a,bk,k;ba,a; + bk,k,ba,a,bk"k;b"",,;]}

x [L L
k',a'm,m'
M(-k',a',m,m')bk"k,baz,a,bm',2sbm,z]

L
z,kz ,(J z
E-
1
~E
Z
[M( -k;, aL Is, z)bkz,k,bazoa'

+M( -kr, ar, Is, Z)bkzok;b"z,a;] M( -kz, a z , z, 2s) . (10)

We write
1 1
Ei - E z E 2s - E z - nw z '
where E z represents the energy of the electron and nw z the photon energy of
the intermediate state z:

M2) = L [E _~ -nw M(-k;,a;,ls,z)M(-kr,ar,z,2s)

z 2s z f

+ E 2s _ ~z _ nw; M( -kr, af, Is, z)M( -kL aL z, 2S)] . (11)

 4. Quantum Fields with Interaction 127

Two terms arise as a result of the different time ordering of the photon emis-
sion as can be read from the arguments of the matrix elements; see Fig. 4.14. lIs)
Still the sum over electron states z remains in MA2). The total matrix
element Mfi = Mc~1) + MA2) becomes
Mfi = M( -kL O'L -kc, O'c, Is, 2s) + M( -kc, O'c, -k;, O'L Is, 2s) 12s)

+ ~ [E _ ~ or
-Iiw M( -kL O'L Is, z)M( -k f , O'c, z, 2s)
Z 28 Z f lIs)

+ E28_L_liwcM(-kc,O'f,IS,Z)M(-kr,O'Lz,2S)]. (12)

J
The expression for the lifetime is given by (compare with Exercise 2.10):
12s)
-:;1 L6 n2c
= (27l')6 27l' 'L,
" d 3k f d 3'1
kc Mfi 12-'( ')
u ko - kc - kc (13)
Fig. 4.14. Different time
CTf a f
ordering of two-photon de-
with ncko = E 2s - E 1s . From (2) we pull out those factors carrying dimensions: cay

and

M(k,O',n,n') = (27l'nc 2 )
L3 Wk
1/2 (_~)
lmc
Jd 3 x 1Pn* (eik'XEka' V) 1Pn"
Since M(k, 0', n, n') appears in the form M(k, 0', n, n')M(k', 0", m, m') within
the matrix element Mfi we can pull out of the total matrix element all quan-
tities carrying dimensions:

M _ ~ (27l'nc 2 ) __I_ M ,. (14)

r. - 2mc2 L3 cJkc k; r.,
where M!i results from Mfi by substitution of the matrix elements (13) with

and

M'(k, 0', n, n') = ~J d 3 x 1P~ (eik,xEka . p) 1Pn' .

These matrix elements and in particular the sum L:z in (12) are difficult to
calculate because not only bound states but also continuum states have to be
taken into account. Since Mfi in (14) takes on the same form as in Exercise
2.10 we can now use the same arguments as before. We set

The further calculation is identical to the one presented in Exercise 2.10!

128 4. Quantum Fields with Interaction

EXERCISE

4.4 Second-Order Corrections

to Rutherford's Scattering Cross Section

Problem. Determine the corrections of second order for Rutherford's scat-

tering cross section.

Solution. We will calculate the contributions of the following diagrams:

We have to deal with two terms for the determination of the transition matrix
element:

In the first sum we substitute p - kl = q and in the second sum p' + kl = q;

inserting Hv we derive

Fig. 4.15. Second-order 1

diagrams for Rutherford x "'"' Op' q'-k'Oq' qOq q"-k"Oq" p - -
L ' " , k'2k"2
scattering q',k'q",k"
2m 16n 2(Ze 2)2 1
Ii? L6 2 L (p2 _ q2)(pl _ q)2(p _ q)2
64n 2m(Ze 2)2 L3
q

J q2 dq dnq

J
h2 L6 (2n)3 (p2 _ q2)(pl _ q)2(p _ q)2
8m(Ze 2)2 q2 dqdnq
nh2L3 (p2 _ q2)(pI2 + q2 _ 2p" q)(p2 + q2 - 2p. q) .
The determination of this integral is far from easy. We follow a method pre-
sented by Dalitz. 3

3 R.H. Dalitz: Proc. Roy. Soc. A 206 (1950) 509.

 4. Quantum Fields with Interaction 129

Let us consider the integral Exercise 4.4.

It = J d3q
[(PI - q)2 + A2](p§ - q2)

lim 27T 1 +1
d(cos'!9)
100 q2 dq
(P1 2 +q2-2qPlcos'!9+A2)(p§_q2 +iE)
1 100-00
E->O+ -1 0

+1 q2 dq
lim 7T dt .
E->O+ -1 (PI 2 + q2 - 2qPIt + A2)(p§ - q2 + if)
1/ (p~ + ,\ 2 ) represents the Fourier transform of 1/ 47Tr exp( -r /'\), which is
the exponentially screened Coulomb potential. As we will realize later on, the
introduction of a finite ,\ is equivalent to giving the photon a finite mass ,\.
The integrand becomes singular at q = ±(p2 + if/2) and q = PIt ±
iJp~(1 - t 2 ) + A2. We evaluate the integral with the help of the residue theo-
rem as we close the integration path in the upper half of the complex q plane.
Here the two poles with positive imaginary part are enclosed by the integra-
tion path. Since the poles of the integrand are of first order the residues are
determined by
Res(J(q) , qi) = lim (q - qi)f(q) .
q-+qi

For the integral we derive

It =

where

r VPi(1-t 2 )+A2.

After some intricate transformations of the integrand we get

7T 2i
It = -
PI
1+
- 1
1
dt T(
1
p2t
1
- ip r
1 T)
P2 - PI t + 1
=
7T 2i
-
PI
1+
-1
1 d
d t In(p2 - PIt
dt- + iT)
2·
~ [In(p2 - PI + iA) -In(p2 + PI + iA)]
PI
Thus, we have

lim
E->O+
J [(PI - q)2 +
d 3q
A2](p§ - q2 + if)
= 27T 2 In P2 - PI
PI P2 + PI
+ iA .
+ iA
Via partial differentiation with respect to the parameter A we obtain

- ·
11m
{->O+
J ((q - Pl)2
2Ad 3q
+ A2)2(p§ - q2 + if)
i7T 2 ( i i)
PI P2 - PI + iA P2 + PI + iA
from which follows
130 4. Quantum Fields with Interaction

Exercise 4.4. lim

<->0+
J d3 q
[(q - Pl)2 + A2F(p~ - q2 + ic)
7T 2 1
A P2 - PI - A2 + 2p2Ai .
Our matrix element MA2) takes on the form

and with a small trick it can be transformed into the form of 12 : First, we
observe that a 0 function appears in the formula for the cross section, which
guarantees energy conservation. Hence, we only need to calculate the matrix
element for the case p~2 = p~ and we write P2 . p~ = p~ cos {). We make use of
the identity

~=
ab
1+-1
1 dz (a(l
2
+ z) + b(l -
2
Z))-2

and apply it to our matrix element:

1
[)..2 + (pz - q)2][)..2 + (P2 - q)2]

12 +1 dz [2
-1 [).. + (P2 -I 2 1+
q) ]-2-
z + [).. 2 + (P2 - 2 1-
q) ]-2-
z] -2

1 +1

-1
dz
2
2 2 -2
- [(PI - q) + A ]

with
1
PI 2[(1+z)P2+(1-z)p],
2212212
A ).. + 2(1 - z )P2 + 2(z - 1)p2 . P2
I

1
)..2 + (1 - z2)p 2 2 (1 - cos{J)

)..2 + p2 sin2 ~(1 _ Z2) .

2
Furthermore, PI = p~( cos 2 {) /2 + t 2 sin 2 {) /2) holds and our matrix element
reads
M(2) = lim 8m(Ze 2)21+ 1 dz 7T 2 1 .
fl )..->0 fi 27T£3 -1 2 A P2 2 - PI 2 - A2 + i2p2A
Inserting PI and A, performing the integration over z, and introducing the
abbreviation 'Y = })..4 + 4p~()..2 + p~ sin 2{J/2) we obtain
. -87Tm(Ze 2)2 1
Mf~2) = hm ---:-;~----'- -----,--;-
.\->0 2
1t £3 'YP2 sin {J /2
)..P2 sin {J/2 i I 'Y + p~ sin {J/2)
x ( arctan
'Y
+ -2 n 2 . {J/2
'Y - P2 sm
 4. Quantum Fields with Interaction 131

In the limit A ---+ 0 the first term inside the bracket vanishes, so we finally Exercise 4.4.
arrive at
M(2) ~ 41l'm(Ze 2 )2 i 1 2p~sint9/2 A-+O
fi ~ - fj,2L3 p3 sin2 t9/2 n A ----400.

The matrix element diverges as the introduced photon mass A goes to zero.
Here the long range of the Coulomb potential leads to a divergence which
questions the result of the previous problem: How can we trust the result
from first-order perturbation theory if the second order already diverges? On
the other hand the experiments are in very good agreement with Ruther-
ford's result. Later in this volume we will recognize how these and other
divergences can be avoided within the framework of renormalization theory,
which is widely discussed in relativistic quantum electrodynamics.
5. Infinities in Quantum Electrodynamics:
Renormalization Problems

A nice feature of quantum electrodynamics is the smallness of the interaction

between the charged particles and the radiation field; as a consequence, this
interaction can always be treated as a perturbation. But in all this brightness
a disturbing spot exists: With the use of the supposedly well-functioning per-
turbation theory some quantities that should remain small become infinitely
large. Discussing the self-energy in Chap. 2, we have had our first encounter
with this problem. Now we have to deal with these difficulties in detail in
order to overcome them.

5.1 Attraction of Parallel, Conducting Plates

Due to Field Quantum Fluctuations (Casimir Effect)

We begin with the infinite zero-point energy of the radiation field in the vac-
uum, which we have already encountered in (1.52):

(5.1)

In our previous considerations we have simply disregarded this contribu-

tion; we have concentrated on energy differences, so that (5.1) has cancelled.
Thereby we have followed the argument often loosely applied that the abso-
lute value of energy has no importance and that an arbitrary constant can be
added or subtracted from it. In general this statement will not hold for sure;
for example, within the general theory of relativity the absolute value of en-
ergy is physically relevant. It is the total energy that enters as a source for the
gravitational field in Einstein's gravitational equations and thus determines
the metric and the curvature of space.
For the quantization of the electromagnetic field as demonstrated in
Chap. 1 we started with the classical expressions and then translated them
into quantum mechanics. This procedure is somewhat questionable. We illus-
trate this point again with the classical expression (1.48) for the energy of the
radiation field

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
134 5. Infinities in Quantum Electrodynamics: Renormalization Problems

H ~L nwk (ak"a'k" + a'k"ak,,) (5.2a)

k,"
(5.2b)

On a classical basis both expressions are identical. From the viewpoint of

quantum mechanics they are not equivalent since (1.50) means the amplitudes
a and a* become operators. Hence, expressions (5.2a) and (5.2b) become

H' " .. (a' a't + a't a" )

=
2'1L" , Itwk k" k" k" k"
k,"

L nwkaLak" + ~ L nwk
k,u k,O'
(5.3a)

and

iI = L nwkal"ak" , (5.3b)
k,"
respectively. Evidently both Hamiltonians differ by the presence of the infinite
zero-point energy. Now we could argue that we have to pay careful attention
to the ordering of a, a* in the classical expression (before quantization). Then
(5.3a) would be the correct Hamiltonian for the free electromagnetic field.
We could also follow the reverse direction and ask what classical expression
for the radiation energy would result from (5.3b) if we take into account the
precise ordering. Indeed, a lengthy backward calculation l starting from (5.3b)
yields

H rad 2-
87l'
Jd 3 x
X {E2 + ii + y'~V2 [E. (V x B) - (V x B) . E]} , (5.4)

where the operator 1/ y' - V2 is defined by

- -1- eik.x =-e

1 ik.x
y'_V2 k
for plane waves. The electromagnetic field has to be decomposed into plane
wave Fourier modes in order to calculate the action of this operator properly.
Note the operator character of the field operators E and 13 in (5.4); if we
consider E and 13 as classical fields the term [E. (V x B) - (V x B) . EJ
vanishes and (5.4) is identical to the classical field energy (1.40). From this
discussion we realize that the classical field energy can be transformed into
a Hamiltonian of the wave field in many ways. We have to ask nature, i.e.
experiment, for the correct Hamiltonian! The same holds for the zero-point
energy of the electromagnetic field [(5.1) and (5.3a)].

1 E.G. Harris: A Pedestrian Approach to Quantum Theory (Wiley, New York 1972).
 5.1 Attraction of Parallel, Conducting Plates 135

Now the question arises whether somehow the zero-point energy (or parts
of it) can be observed in an experiment. Casimir2 pointed out such a possibility
and Lifschitz and Fierz 3 have both investigated his argument in more detail. 4
The line of thought is as follows. The zero-point energy of the electromag-
netic field is given by
1
L '2 fkvn ,
n
Em Elv

cJQr-1J
where n stands for k. The frequencies Wn of the electromagnetic field depend
on the geometry of the volume in which the field is confined. If the geometrical
form changes the frequency of the normal modes, the zero-point energy will
also change. We consider a rectangular box of lenght R and basal area A =
L2. This box, with conducting walls, represents our volume (quantization
~~.~
volume), which determines the frequencies of the electromagnetic field via its I. RI1J.I~ ~/d
geometry. We introduce a second plate at a distance d from the first one and
are interested in the energy of the system depending on the position of one Fig. 5.1. Modification of
a rectangular box leads to
of the plates. Hence we subtract the energy of a reference configuration for a modification of the zero-
which the mobile plate is fixed at a certain distance (for instance 1/rJ with point energy
1/7] = 1/2; see Fig. 5.1). The difference in the energy is
U(d,R,A) = (EI+En)-(Em+Elv), (5.5)
where Ei (i = I, II, III, IV) represents the zero-point energy of the free elec-
tromagnetic field in the corresponding region depicted in the figure. In the
following, we will move the walls of our quantization volume to infinity, i.e.
U(d, A) = lim U(d, R, A) .
R->oo

Every single term Ei is formally divergent because we consider an infinite

number of normal modes with increasing frequency. Therefore it is necessary
to find a physically reasonable cut-off procedure.
In our case, the wavelength could serve as a cut-off parameter. From classi-
cal electrodynamics we know that a good conductor turns into a bad conductor
at short wavelengths (X-rays!). Hence, we can cut off the energy exponentially:

Ei = L ~fkvn exp ( -A ~n) ; (5.6)

n
for the final result we will again let A ---> O.
In a rectangular waveguide with dimensions d x L x L the frequencies of
the classical and the quantum-mechanical normal modes are given by (see
Chap. 1)
Wlmn Cklmn(d, L, L)

2 H.B.G. Casimir: Proc. Netherlands Aka. Wetenschapen 51 (1948) 793.

3 E.M. Lifschitz: Soviet. Phys. JETP 2 (1956) 73; M. Fierz: He!. Phys. Acta 33
(1960) 855.
4 For a review of the Casimir effect in all its facets see G. Plunien, B. Miiller, W.
Greiner: Phys. Rep. 134 (1986) 87.
136 5. Infinities in Quantum Electrodynamics: Renormalization Problems

where I, m, and n are positive integers. Because there are two possible polar-
izations, the potential energy becomes

U(d, A) = lim lim

R-->oo '\-->0
!nc{
2
[2 """
L
[,m,n
klmn exp( -)..k1mn ) + (d ----> R - d)]

(5.7)

where, for example, the bracket (d ----> R/7]) is a shorthand notation of the first
expression with the substitution of Rh instead of d.
We will present the further calculation in Exercise 5.1. After performing
the limiting cases R --> 00 and ).. --> 0 we obtain

U(d A) = _~ ficA (5.8)

, 720 d3 '

This result is finite and does not depend on the cut-off procedure. From here
we calculate the force per unit area:

F = _~ (U(d,A)) = _ fiC7r 2 •
(5.9)
ad A 240d4 '

evidently this is not only a finite result again but, more astonishing, the force
depends only on the universal constants fi and c. Obviously it is indepen-
dent of e, the coupling between the electromagnetic field and matter. This
force depends only on the "zero-point pressure" of the zero-point oscillations
of the photon vacuum. Therefore we are allowed to say that this attracting
force between the conducting plates is of pure quantum-mechanical nature. It
vanishes for the limiting case fi --> O.
We have realized that two parallel plates alter the allowed normal frequen-
cies and thus the zero-point energy of the quantization volume (for the case
R ----> 00 of the "universe"). From this alteration results a force that pushes
the plates together.
An extension of this theory onto dielectric substances at finite temper-
atures was performed by Lifschitz and Fierz. An experimental test of the
theory started in 1957 with studies by Abrikosova and Deriagin,5 whereas the
first real evidence was provided by Sparnaay. 6 Details of this experiment are
presented in Exercise 5.2. From then on this experiment has been repeated
several times with different methods. The theoretical predictions have been
confirmed every time. 7

5 1.1. Abrikosova, B.V. Deriagin: Sov. Phys. JETP 3 (1957) 819; 4 (1957) 2.
6 M.J. Sparnaay: Physica 24 (1958) 751.
7 T.H. Boyer: Ann. of Physics 56 (1970) 474.
 5.1 Attraction of Parallel, Conducting Plates 137

EXERCISE

5.1 Attraction of Parallel, Conducting Plates

Due to the Casimir Effect

Problem. Under the assumption that the length L of the plates is large com-
pared to the distance d, derive the relation given in the text for the potential
energy of the system of plates depicted in Fig. 5.1.

Solution. To begin with we calculate the first expression in (5.7) for the
potential energy; afterwards we replace d with the corresponding other quan-
tities. We get
U he L klmn(d, L, L) exp[->'k1mn(d, L, L)] ,
l,m)n

where

If we assume A = L2 » d 2 then at finite energy many modes fit into the

volume; this means that we can choose m/ Land n/ L to be very large and
at the same time the energy would stay small. We are able to replace the
summations over n und m by integrations:

U = he ~
00 100 100
m=O dm n=O dnklmn(d,L,L)exp[->.klmn(d,L,L)]

or

U 00
he ~ 100 1
m=O
00

dm n=O dn (~r + (:y)2 + (n;f

x exp [-A (';)' + ("Z)' + C;:')' 1

We set
d
y a
7r

U = he 2
L2
L00 100 dxdy
7r 1=1 0

The substitution

Z = (la)2 (x 2 + y2) = (_al )2r2

leads to the surface element

rdr d¢ = ~ (~r dz = ~ (J f l2 dz .
138 5. Infinities in Quantum Electrodynamics: Renormalization Problems

r
Exercise 5.1. The integration over ¢ goes up to 7r /2 (only positive z)! We get

U = ~nc ~ ~ l n/2
d¢ 1 00
dz v'zTI (± exp ( -A±v'zTI)

with A = L2. Continuing,

u = ~hc 7r;~ f l31°O

1=1
dz v'zTI exp ( _ :l v'zTI) ,

where 0: = A7r. Now

d~3 exp ( -0:~v'zTI) = - (~r (z + 1)rz+l exp ( -o:~rz+l) ,

so that

U = -~ncA7r2 ~
4 L..t
1=1
ioroo ~
z+ 1 ~
d0: 3
{exp (_ o:l rz+l)} .
d

We assume uniform continuity; hence we are allowed to interchange the order

of the summation, integration, and differentiation:

1
U = --hcA7r
4
2 -d
3
d0: 3
10
00
-dz- Lexp
Z
00

+ 1 1=1
(O:l
--rz+l )
d

Since
~e--YI _ 1
L..t - 1- e--Y ,
1=0
it follows that

:l
~exp (- v'zTI) 1 - exp (-(o:/d)v'zTI) 1 ,
1
----;--.,-------;::==_ -

u = -~hcA7r2 d 3 roo ~ [ exp (-(o:/d)JZ+T) 1

4 d0: 3 io Z +1 1 - exp (-(o:/d)JZ+T)

We substitute u = JZ+T with dz = 2JZ+T du, so that

1 2 d 3 JOO du e--Yu
U = --hcA7r - - ---
2 d0: 3 1 U 1 - e--YU
with "( = o:/d, or

U = -~hcA7r2~ JOO du __1_ .

2 d0: 3 1 U e-Yu - 1

Since dido: = (l/d)(d/d"() we obtain

u=
 5.1 Attraction of Parallel, Conducting Plates 139

Now we substitute x = e"YU and arrive at Exercise 5.1.

U = ~ncA-7l"2~
2 do;2
[.!.1°O (x dx- 1)2 ]
0; e"Y

With z = x - I we get

u= ~ncA-7l"2
2
d [.!.1°O d;]
2
do;
2
0; e"Y-l Z

~ncA'71"2
2
d22
do;
[.!. (_~)OO ]
0; Z e"Y-l

1n A
"2 C
2 d2
'71" do;2
[1a 1 1]
eO'./d -

nc;;A d~2 [(~r eO'.~!~ 1]

The Taylor series expansion of y / (e Y - 1) reads:

-
y 00
- "'"'-y
Bn n
eY -1 - L n! '
n=O
where Bn represent the Bernoulli numbers, which are given by Bo = 1, Bl =
-1/2, B2 = 1/6, B3 = 0, B4 = -1/30, .... Hence 8

u= nc'71"2A ~ [(~)2 ~ Bn (~)n]

2d do;2 L n! d 0;
n=O

'71"2 d2 [00 Bn (o;)n-2] (1)

-ncA-
2d do;2 "'"'
L -n! - d .
n=O
The actual difference between the potential energies of the two configurations
of plates (see Fig. 5.1) is given by
U(d, R, A) = (EI + En) - (Em + E lv ) . (2)
We insert result (1) into (2) and let the distance between plates go to infinity,
R --t 00, as well as 0; --t 0:

Here (d --t R - d) means that d inside the first bracket has been replaced by

(R - d) and so forth.

8 See, for example: G. Artken, H.S. Weber: Mathematical Methods for Physicists
(Academic Press, New York 1995).
140 5. Infinities in Quantum Electrodynamics: Renormalization Problems

Exercise 5.1. We consider the first terms of the expansion:

U(d,A) =

As we sum over all four contributions, the terms with a- 2 and a- 1 cancel.
The next term vanishes once differentiated with respect to a. Therefore we
are left with

U(d, A) = lim lim 1[2 hcA~2 [_~ ( 1

R-->oo <>-->0 2 da 720 (R - d)3

+ :3 - (R;7])3 - (R - ~/7])3 ) + ...J

lim lim [-~hcA (~ + 1 ___
1_ _ 1 )
R-->oo <>-->0 720 d3 (R - d)3 (Rh)3 (R - Rh)3
+ terms, which contain powers of a]
and get

U(d,A)
Hence, the potential stays finite and does not depend on the cut-off parameter
.\ nor on the coupling constant a; it depends only on the dimensions of the
plates. The resulting force
[j 1[2 A
F = - 8d U (d,A) = - 240 nc d4
is attractive; both plates attract each other!

EXAMPLE

5.2 Measurement of the Casimir Effect

The first measurements of the Casimir effect were performed by Abrikosova9

and collaborators as well as by Sparnaay.10 Here we briefly describe the
method applied by Sparnaay (see Fig. 5.2).
The attraction prevailing between the two metal plates at very short dis-
tances (d) causes a rotation of the weigh beam (b). Then the capacity of
the condenser (c) is changed. Since the scale is supported at the point e, the
distance d between plates of the measurement condenser decreases, i.e. the
capacity increases. The measured capacity is a measure for the expansion of
the spring and thus for the force to be measured. With this method it is pos-
sible to achieve a precision of;::;:; 10- 3 dyn/cm 2 . The procedure is as follows. A
9 1.1. Abrikosova, B.V. Deriagin: SOy. Phys. JETP 3 (1957) 819; 4 (1957) 2.
10 M.J. Sparnaay: Physica 24 (1958) 751.
 5.1 Attraction of Parallel, Conducting Plates 141

Fig. 5.2. Sparnaay's ap-

paratus for the measure-
ment of the Casimir effect

e b

mmm

distance (d) between plates is chosen which is large enough so that the force
acting between the plates is smaller than the measurement resolution. With
the help of weight Wand spring S the scale is brought into equilibrium. Now
the distance between the plates is changed with the micrometer screws (m)
and the corresponding force is measured via the condenser (c).
In this way the dependence of the force on the distance d is obtained
which is depicted in Fig. 5.3. The width of the measured points goes back
to the micrometer screws. In the region F « 0.01 dynjcm 2 the quality of
the measurement of the force becomes uncertain. The broken line represents
the theoretical prediction as derived above. We see that the agreement be-
tween theory and experiment is remarkably good within the experimental
uncertainty. The experimental results obtained by Abrikosova show the same
agreement with the theoretical predictions.
The measured dependence of the force on the distance excludes expla-
nations (such as van der Waals forces) other than the Casimir effect. The

0.15 I---,.t-r--,,.-t---+---I---l

0.1
1\
\ theory
,Ill I
0.05 1----+--;\.-,--+1----1---11---1

0.01
,
x~x

-0.01 0.5 1.5 +- +

Fig. 5.3. Comparision be-
tween theoretical and ex-
perimental results for the
Casimir effect
142 5. Infinities in Quantum Electrodynamics: Renormalization Problems

Example 5.2. experiment described here has been presented in an oversimplified way. In
real life, precision work is necessary: The plates have to be entirely free of
dust, which is achieved only by very complicated procedures. Furthermore,
all electrostatic charge must be removed since it would lead to additional
forces. In addition, rigorous requirements have to be demanded for the plates
to be really flat. These are only some points that have to be taken care of. l l

EXAMPLE

5.3 Casimir's Approach Towards a Model for the Electron

The outcome of the zero-point energy for the plate condenser inspired
Casimir12 to develop a model for the electron. He assumed that the elec-
tron can be viewed as a spherical shell with homogenous surface charge e (see
Fig. 5.4). The electrostatic energy of a spherical shell with radius a is given
by
e2
Ee = -2a
The corresponding pressure is Pe = -aE/aV. With V = 47Ta 3/3 we obtain
p. _ _ _
1_ aEe
e - 47Ta2 aa
Fig. 5.4. Spherical shell
model of the electron i.e. the electrostatic pressure puffs up the spherical shell.
The presence of the spherical shell alters the normal modes of the electro-
magnetic field. The typical wave-number is k n rv l/a, so that

EN = L ~nwn = ~ne Lkn

n n
= -C (~:)

with constant C. The corresponding zero-point pressure reads

1
PN = - - - - - = -C--.
aEN ne
47Ta 2 aa S7Ta 4
The electron remains stable once the electrostatic pressure and the zero-point
pressure cancel each other:
e2 nc
Pe + PN = 0 = -7Ta
C S4
S7Ta-4' -

i.e. for the case that the constant,

e2
C = - = a
ne '
is equal to Sommerfeld's fine-structure constant. Since C follows from the geo-
metrical boundary conditions for the electromagnetic field inside the spherical

11 G. Plunien, B. Miiller, W. Greiner: Phys. Rep. 134 (1986) 87.

12 H.B.G. Casimir: Physica 19 (1953) 846.
 5.2 Renormalization of the Electron Mass 143

shell, it would be a first-time opportunity to calculate an elementary constant, Example 5.3.

i.e.
e2
0: = nc'
The simple result for the plate condenser seems to be encouraging: if we use
the spherical shell as a crude model and set A = 7ra 2 as well as d = a it follows

EN = - 720d3
7r 2 ncA
= -
7r 3
720a
nc 7r 3
= - 360
(ftc)
2a ;

hence, C = 7r 3 /360 ~ 0.036 whereas 0: ~ 1/137 ~ 0.0073.

However, the correct calculation for the spherical shell 13 leads to the failure
of this model: The value for C turns out to be C = -0.093 and has the wrong
sign. Thus, the resulting pressure is also positive and, like the electromagnetic
pressure, puffs up the spherical shell!

5.2 Renormalization of the Electron Mass

Now we examine in more detail the self-energy of the electron, which we

first encountered in Chap. 2 and which turned out to be an infinite quantity.
The problem of self-energy of a charged particle already exists in classical
electrodynamics. For example, a small conducting sphere of radius Rand
with charge e possesses an electric field
e r
E = ---
r2 r '
r;::: R;

see Fig. 5.5. For the limit of a point charge (R --+ 0) the electric field energy
of this sphere

(5.10)

becomes infinite.
This is the longitudinal (Coulomb) energy of the field. In quantum field
theory an additional energy exists as a result of the transverse electromagnetic
field. In the following, we will investigate this quantity within the framework
of nonrelativistic quantum electrodynamics, which we have pursued so far. Fig. 5.5. Electric field
In our lectures on quantum electrodynamics 14 we will become acquainted of a charged, conducting
with the relativistic theory, which is more exact and correct. For our purposes sphere

13 T.H. Boyer: Phys. Rev. 174 (1968) 1764.

14 W. Greiner, J. Reinhardt: Quantum Electrodynamics, 2nd ed. (Springer, Berlin,
Heidelberg 1994).
144 5. Infinities in Quantum Electrodynamics: Renormalization Problems

here it is sufficient to consider the nonrelativistic interaction operator of the

electron with the transverse radiation field:
e e2 2
Hint - mcP· A + 2mc2 A
A A

H;nt + H;~t ; (5.11)

it already allows us to present the essential line of thought in a simple and
clear fashion. As usual we treat Hint as a perturbation.
Iq) (5.12)
stands for a free electron state with momentum p = fiq. Its wavefunction is
given by
1 .
1/Jq (x) = (xlq) = y'[) e1qx . (5.13)

We designate the state with one free electron of momentum p and no photon
as
(5.14)
Then according to first-order perturbation calculation for the free electron we
calculate the contribution

(5.15)

EXAMPLE

5.4 Supplement: Historical Remark on the Electron Mass

According to classical electrodynamics the electron with its charge (-e) pro-
duces an electrostatic field where the energy is given by
e2
W = aR . (1)

The coefficient a depends on how the charge (-e) is distributed over the
small sphere with radius R. The order of magnitude for a is 1: for example,
a = 1/2 for a charge distribution confined to the spherical surface (compare
with result (5.10) for a conducting sphere) or a = 3/5 for a uniform charge
distribution over the whole sphere. The eigenmass of the electron consists of
two parts:

1) the mechanical mass m', which is not coupled to the field energy and
2) the electromagnetic mass mel, which is coupled to the field energy.

Hence, we have
m = m'+mel. (2)
 5.2 Renormalization of the Electron Mass 145

According to (5.10) the mass coupled to the field is determined by Example 5.4.
W e2
mel = 2 = a Rc2 . (3)
Strictly speaking, classical electrodynamics provides the following expression
for the mass coupled to the electrostatic field:
4W
mel = '3 c2 . (4)
The factor 4/3 instead of 1 results from the instability of the charge distri-
bution as identical charge elements inside the sphere repell each other. Thus,
additional forces of nonelectromagnetic origin have to be introduced in order
to produce equilibrium. If we consider nonlinear electrodynamics these diffi-
culties do not occur because they lead to stable charge distributions inside
the electron. For this case the correct relation (3) between the field-coupled
mass and the field energy is obtained. Lorentz succeeded in showing that the
electromagnetic mass depends on the velocity v of the charge according to the
relation

(5)
VI - v2 /c 2
This relation is easy to understand because Maxwell's equations have to be
Lorentz covariant. Hence, the electrostatic energy has to transform like the
fourth component of a four-vector, from which (5) follows (consult the lectures
on classical mechanics and classical electrodynamics of this series) .15
According to Lorentz the mechanical mass m' should be a constant; he
assumed the validity of Newtonian mechanics. However, experimental expe-
rience showed that the total electron mass changes according to the law (5).
It appeared that the total electron mass possesses field character. Nowadays
a value of R « 10- 16 cm for the electron radius, deduced from scattering
experiments, is used. Lorentz used a value R ~ 10- 13 cm and calculated with
(3) a reasonable mass of m ~ 10- 27 9 = melectron. Then came the theory
of relativity and required that the mechanical mass m' had to obey the law
(5), toO.15~16 The apparent consistency of these considerations achieved so
far broke down and the nature of the electron mass again faded away into
darkness.

In first order the term p.A from (5.11) does not contribute because it does
not contain terms of the form at a or aa t, which describe the emission and
A 2
reabsorption of photons. On the other hand, the term A describes such simul-
taneous creation and annihilation processes, as shown graphically in Fig. 5.6.
With the help of (2.9) and (5.14) we deduce immediately that

Fig. 5.6. Emission and

15 W. Greiner: Classical Mechanics I (Springer, New York) in preparation; W. absorption processes as
Greiner: Classical Electrodynamics (Springer, New York 1996). contained in the term pro-
16 W. Greiner, J. Rafelski: Special Relativity (Springer, New York) in preparation. portional to jp in (5.11)
146 5. Infinities in Quantum Electrodynamics: Renormalization Problems

(5.16)
Obviously this contribution to the energy of the free electron is infinite but
independent of the electron momentum p. It is equal for all electrons and, as
soon as energy differences are taken into account, it drops out. Therefore we
will not consider this contribution further.
Next we turn to the second-order contribution of the term p . A appear-
ing in the interaction (5.11), i.e. [(,t. This contribution will be much more
interesting; in second order it yields the contribution

E(2) = L (Pljj~nt II) (Iljj~nt Ip)

p J Ep - E J

~ ' " 1(Ilp·AlpW (5.17)

m 2 c2 LJ E p - EJ

to the energy. Since A is linear in the photon creation and annihilation op-
erators a, at the intermediate states II) have to contain one photon. Hence,
they are of the form
't t
II) bqIO)pak.,-IO)rad,
(5.18)
't
Ip) bplO)p ,
and
(hp)2
Ep ,
2m

EJ (hq)2 + fu.uk (5.19)

2m
denote the corresponding energies. With (2.8) we obtain immediately
 5.2 Renormalization of the Electron Mass 147

(5.20)

The Kronecker delta administers momentum conservation; hence we derive

for (5.17) the expression
E(2) _ ~,,(27rfic2) Inp· tkO"I 2 (5.21) Fig. 5.7. Diagram of the
p - m c ~
2 2 L3 Wk 1i2p2 _ ( 1i2(P-k)2 nkc)' self-energy. An incoming
k,O" 2m 2m + electron of momentum p
This is exactly the real part ~(.1Ei) of the self-energy obtained earlier in scatters into the interme-
diate state q by emitting
(2.105a). The processes involved are exemplified in the diagram in Fig. 5.7. a photon ku and then it
We continue with (5.21) and use scatters back into the old
state p by reabsorbing the
photon

(5.22)
Here {j represents the angle between the photon and electron (see Fig. 5.8).
Starting with

L tkO"tkO" = tkltkl + tk2 t k2 ,

a

we have introduced a third unit vector k/k to construct

, k k
I = tkltkl + tk2tk2 + k k'
Fig. 5.8. The geometry
Hence, we obtain
of the emission process of
a virtual photon: p repre-
(5.23) sents the electron momen-
tum and k represents the
photon momentum
which we have used in (5.22). Insertion of (5.22) into (5.21) leads to

E(2) = _e 2_ (27rnc
___ 2 ) _L3_
p m 2 c2 £3 (27r)3

X J d3 k -
1
Wk 1i2p2 _
n2p2(1 - _
_ _---:,.---=----'-_
(/i
cos 2 {j)
2p2 _ 1i2 p ·k
_-'---c,---_-.,-
(/ik)2
+
nkc)
+
J
2m 2m m 2m

~ nc2 k 2 dk d(cos {j) d4> (np)2(1 - COS 2 {j)

m 2 c2 (27r)2 kc (liP
-nkc 1 - me cos {j + 2me2
like)
148 5. Infinities in Quantum Electrodynamics: Renormalization Problems

--e 2- --27T
(np?
m 2 e2 (27T)2
1 1+ 00
dk
1 1 - cos 2 {)
d(cos{))---------;:,,--
1 - ~ cos{) + 2'::,~2

1+ 1
0 -1

2
(np)2 ( - - -e2 1 00
dk )
-2- - d(cos{))(1 - cos 2 {)) v hkc'
m me 7T -1 0 1 - C cos {) + 2mc 2

Obviously the last integral over k diverges logarithmically because for large k
it behaves like J dkjk. With other words: The self-energy E~2) is divergent.
If we combine this contribution with the energy of zeroth order E~O) the total
energy of the free particle becomes
Ep E(O)
p
+ E(2)
p
(np)2

1+ 1
2m
(np)2 ( 2 1
- -e2 - dk )
00
- - - d( cos {))(1 - cos 2 {)) v hkc
2m me 7T _ 1 0 1 - C cos {) + 2mc 2
'- v "
=D
(np)2 (1 _ D). (5.24)
2m
D stands for the divergent part. If the extreme nonrelativistic approximation
vje« 1, fuvjme 2 « 1 is applied to D, which, of course, cannot be fulfilled in
the total integration domain, then we arrive at the result

D = 41+1
me 7T -1
d(cos{))(I-cos 2 {))
io
roo dk
~ (~)
io
roo dk
1
me
2 7T 3

-4 - e2 00
dk
37T me2 0 '

which is often referred to in the literature but inconsistently. Within this

approximation the expression for D diverges even linearly and (5.24) becomes

E
p
= (np)2
2m
(1- ~~ ioroo 37T mc2
dk) . (5.25)

We will pursue the following point of view: The mass m in the energy of zeroth
order,

E(O) (np)2 (5.26)

p 2m'
represents the mass of the naked electron, which does not interact with the
electromagnetic field. It can only be fictitious because the electromagnetic in-
teraction cannot be switched off. Only under the presence of the electromag-
netic field is the mass of the electron experimentally observable. Therefore
1 1
-(1 - D) (5.27)
m
 5.3 The Splitting of the Hydrogen States 2S 1/ 2-2p3/2: The Lamb Shift 149

has to hold (in lowest order). In other words: The unknown fictitious mass m
of the electron has to be conditioned (divergent) such that
m
m exp = 1- D
always has a finite value, i.e. the value for the electron mass deduced from ex-
periment. This procedure of interpreting the divergent self-energy as a change
from the fictitious mass m of the electron to the real mass m exp is known as
the renormalization of the mass.

5.3 The Splitting of the Hydrogen States 2S 1 / 2-2p3/2:

The Lamb Shift

According to both Schrodinger's and Dirac's theories the 2S 1/ 2 and 2p1/2

states of hydrogenlike point nuclei are degenerate. This is true for pointlike
central nuclei, i.e. in the case of hydrogen for pointlike protons. It has been
one of the greatest achievements of atomic theory that the fine structure of
the atomic spectra (splitting of the states with j = l ± s) could be explained
quantitatively thanks to the consideration of spin-orbit coupling (which is
automatically contained in the Dirac theory). However, the precise measure-
ment of the hydrogen 2S 1/ 2 - 2p3/2 splitting led to some doubts with respect to
theory; small deviations from the theoretically predicted splitting have been
observed in experiment.
At first (1930-40) the accuracy of the measurements did not allow the ex-
act determination of this shift. Only the development of microwave techniques
made a precise investigation possible. In 1947 Lamb and Retherford 17 investi-
+
2p3/2
gated the energetic position of the 2S 1/ 2 state with the help of high-frequency
spectroscopy and discovered a shift of 9910 MHz
D.v ;::::; 1058 MHz (present value) . ! 2SI/2
This is depicted schematically in Fig. 5.9. 11058 MHz
2PI/2
This so-called Lamb shift is a result of the interaction of the electron with
its virtual radiation field. We have discussed the latter in the last section. experiment
Now we want to present the experiment, which has played a key role in the
development of quantum electrodynamics, and then we concentrate on the
f
2p3/2
corresponding theoretical calculations.
An atomic beam of hydrogen in the 1s 1/ 2 ground state can be produced
by dissociation of molecular hydrogen at high temperatures. An impinging 10950 MHz
electron current excites some of the atoms into the (n = 2) state. Via optical
transitions the two 2p1/2 and 2p3/2 levels soon decay into the Is ground
state. However, from the 2S 1/ 2 level only the transition into the 2P1/2 level is
! 2PI/2
2SI/2
possible, since the l = 0 transition into the 1s 1/ 2 ground state is prohibited; Dirac equation
the lowest electromagnetic multipole is the electric dipole with l = 1. Hence, Fig. 5.9. Energy levels
the 2S 1/ 2 state can be regarded as metastable. These metastable atoms are of the relevant hydrogen
then collected in a metal target. Contrary to the case for an atomic beam with states

17 W.E. Lamb, Jr., R.C. Retherford: Phys. Rev. 12 (1947) 241.

150 5. Infinities in Quantum Electrodynamics: Renormalization Problems

atoms in the ground state, an intensive emission of electrons is observed from

the metastable atoms; electrons can escape more easily from the metastable
states because of the lower binding energy (Coulomb excitation). The intensity
of the metastable beams can be altered as they have to run through a spatial
region where electromagnetic radiation of the excitation frequency 2S 1/ 2-2p3/2
prevails. Moreover several of the magnetic substates can be studied with the
help of the Zeemann effect. If a magnetic field is applied to the microwave
region, the excitation into the 2P3/2 state can be studied with three different
frequencies (see Fig. 5.10).
Lamb and Retherford's experiment is sketched in the next example.

m
Llv
+3/2

+1I2
2P3/2
-112
Fig. 5.10. Sketch of the
splitting of the hydro- -3/2
gen energy levels 2P3/2,
2Pl/2, 2S 1 / 2 in a magnetic +1/2
field and of the excitation 2S1/2
+1/2
frequencies of Lamb and 2Pl/2 -1/2
Retherford -1/2
IBI

Since the experimental result of the hydrogen 2P3/2-2s1/2 splitting de-

viates from the predictions of the Dirac equation by about 10%, this could
have easily been used against the theory. However, in 1947 a strong belief
in quantum electrodynamics already prevailed. The Lamb shift was thought
to be a consequence of the interaction with the electromagnetic field. When
the corresponding calculation for the shift was performed for the first time, it
turned out to be infinite. BethelS demonstrated that these difficulties can be
removed with a renormalization of mass. We will follow Bethe's thinking and
perform a nonrelativistic calculation.

EXAMPLE

5.5 Lamb and Retherford's Experiment

At first, H2 molecules are dissociated into hydrogen in an oven. Then the H

atoms are bombarded with electrons and are excited to transitions into states
with n = 2. Mainly metastable excitations into the 2S 1/ 2 states survive. Owing
to the low binding energy of their electrons these excited hydrogen atoms are
able to emit electrons intensively. The corresponding current is measured with
a collector and an attached galvanometer.

18 H.A. Bethe: Phys. Rev. 72 (1947) 339.

 5.3 The Splitting of the Hydrogen States 2S 1 / 2-2p3/2: The Lamb Shift 151

As the hydrogen atoms have to pass through a resonator which contains Example 5.5.
electromagnetic radiation with the transition frequency 2S 1 / 2 --> 2P3/2 the
current diminishes since the induced gamma emission causes the number
of metastable atoms to be reduced. The experimental set-up is sketched in
Fig.5.1l.

magnetic
fieldB
10- 8 of all atoms are now
in the metastable state

beam
collector with Fig. 5.11. Lamb and
resonator gal vanometer Retherford's experimental
2.4-18.5 cm set-up

The splitting due to the Lamb shift is of the order of 10- 6 eV, which lies
in the microwave region. Technically it is more convenient to observe this
energy difference directly with a high-frequency method instead of optically
investigating the splitting of the Ha line. In particular, the latter procedure
is difficult to perform because of the Doppler shift of the atoms moving with
respect to the source; the resulting frequency shift,
/:)./.1 = 7 X 10- 7 VTjA/.I,
is proportional to /.I (T in Kelvin; A atomic weight).
As soon as 2s --> 2p transitions occur in the field region, the current is
reduced because metastable atoms have been removed.

In)
The self-energy of a state In) of the hydrogen atom is represented by the
diagram shown in Fig. 5.12. This diagram is analogous to the one for a free
electron, sketched in Fig. 5.7. The only difference is that now the electron
states In) are bound and not free. The correction to the energy is given by

E(2) = ~ '" '" (27'i1ic 2

) I(n'lfJ . €k<7lnW (5.28)
n m2e2 L L
n' kG'
Pke E n -E, n
-Me'

in analogy to (5.17) and (5.21), where the dipole approximation

P . €k<7 e -ik·x = p. €k<7 , Fig. 5.12. Self-energy of
a state In) ofthe hydrogen
has been used. Employing (5.23) and (2.18) we simplify this expression to atom
152 5. Infinities in Quantum Electrodynamics: Renormalization Problems

(5.29)

Since

f ds?(l - cos 2 '19) -27l' /-1 d(cos'l9)(l- cos2'19)

-27l' (x _~3) [1 = 8;,

it follows that
E(2) = ~ e2 li L roo kdk l(n'lplnW (5.30)
n 37l' m 2 e
n'
loo En - En' - like'

This energy is as divergent as the self-energy for a free particle obtained pre-
viously in (5.25). At this point Bethe argued as follows: For a free electron the
matrix elements (n'lpln) yield only diagonal contributions and (5.30) becomes
the formerly derived second term of (5.25)

E(2)
p
= _~ ~
37l'me2
((liP )2)
2m
roo dk.
lo
(5.31)

However, according to the mass-renormalization scheme this contribution to

the kinetic energy of a free electron is interpreted as a contribution of the
electromagnetic mass to the total mass m exp of the electron. For a bound
electron, which has no sharp momentum, a term analogous to (5.31) is ob-
tained once the square of the momentum (lip)2 is replaced by the expectation
value of (nlv2In). With the completeness relation
(5.32)
n'

(5.31) becomes

E(2)
n(mass)

(5.33)

E~~;nass) represents the correction to the kinetic energy according to the

change of the electromagnetic electron mass in the state In). The energy
E~~;nass) has to be subtracted from (5.30) in order to obtain the real and
observable energy shift ~E~2):

~E(2) ~ e li ' " roo kdkl(n'lvln)12

2
=
n 37l' m 2 e ~ lo
n'

x [ 1 1- ]
+like (5.34)
En - En' - Me .
 5.3 The Splitting of the Hydrogen States 2S 1 / 2-2p3/2: The Lamb Shift 153

Because of
1 1 k--+oo
------+- --4
En - En' - like like like(En - En' - like)
(5.34) is logarithmically divergent. Here Bethe assumed that the integral anal-
ogous to (5.34) has to converge in a relativistic theory. This turned out to be
correct for the following reason. In a relativistic theory, states exist with pos-
itive and negative energy, which, according to the equation
(5.35)
+moc2
have to lie either above +moe2 or beneath -moc 2 . According to Dirac the --+---
lower states have to be occupied with electrons; otherwise an electron would
decay into even lower states and emit radiation. Thus lower states occupied
with electrons prevent a "radiation catastrophe" because the Pauli principle
prohibits such transitions.
If we attempt to construct a wave packet out of relativistic wavefunctions
that describes an electron with positive energy, states of negative energy do
not contribute as they are already occupied according to the Pauli princi-
Fig. 5.13. Typical energy
ple. Together the states with positive and negative energy are complete; but
spectrum of a relativistic
no subset is complete! In other words: With states of positive energy alone theory
we cannot construct a 83 (r) function. Hence, a wave packet consisting of
wavefunctions with positive energy cannot be compressed to arbitrary small
volumes; this can only be achieved up to the Compton wavelength of the
electron,
h
ACompton
me
In a way, the electron has acquired a spatial dimension; it is no longer localized
as a point. For this reason a relativistic theory automatically contains this
necessary restriction of the integration over k [compare with (5.34)]. This leads
to the convergence of the correponding integral. For a nonrelativistic theory
this convergence can be achieved by introducing a cut-off of the integral (5.34)
at the photon energy
me 1
like ~ me 2 ==} kCompton
ACompton .

This yields

1 ] kCompton
X
[ - lie In(En - En' - like) 0
154 5. Infinities in Quantum Electrodynamics: Renormalization Problems

2.~"I( 'IAI )12(En -E')l E 2

3 fj, 2 3 ~ n pn
7f men'
n nEn-
E n,-mc
E n - n'

~ 3~ fj,~:c3 ~ l(n'lpln)12(En - En~~2En I·

En') in 1 (5.36)

In the last step (En' - En) has been neglected with respect to mc 2 in the
nominator of the logarithm. In order to simplify the sum over n' even further,
(En' - En) is replaced within the argument by a suitable mean value
(En' - En) ---> (En' - En)av. (av. = average) . (5.37)
This approximation is not as crude as it appears. The argument of the loga-
rithm
mc2
~ 105
En' -En
is a large number and the logarithm function is a slowly varying function.
Even a deviation in (En' - En)av. by about a factor of 10 results in only a 20%
incorrect value for the expression (5.36). With the approximation (5.37) the
logarithm in (5.36) can be written in front of the summation; the remaining
summation can be calculated as follows:

n' n'

n'

(nip, (Ho - En)pln)

(nip, (Hop - pHo)ln) . (5.38)
Here

;m +
A2
Ho Vex)
represents the Hamiltonian of the hydrogen atom. The commutator results in
A A fj,
Hop-pHo = --:-VV(x). (5.39)
I

Since

Vex)
and
.6.V(x)
hold, we get for (5.38)
(nip, (Hop - pHo)ln)
(nl~v, (-~VV(x)) In)
fj,2 J'Ij.;~V . (VV)',pn d3 x
 5.3 The Splitting of the Hydrogen States 2S 1 / 2-2p3/2: The Lamb Shift 155

-n? JC'V1jJ~)· (VV)1jJn d3 x

n,2 J 1jJ~(V2V)1jJn d3 x + h2 J 1jJ~(VV) . V1jJn d3 x.

From the last equation it follows that

_h2 J (V1jJ~)1jJn . (V V) d3 x - h2 J 1jJ~(V1jJn) . VV d3x

= h2 J 1jJ~(V2V)1jJn d3x.
The hydrogen eigenfunctions are real and consequently the two expressions
on the left-hand side are identical, so that

h2 J 1jJ~V· [(VV)1jJnJ d3x = _~h2 J 1jJ~(V2V)1jJnd3x.

Thus,

(nip, (Hop - pHo)ln) _~h2 J 1jJ~1jJn(V2V) d3x

~2 J 1jJ~(X)1jJn(x)Ze2411'()(x) d3x
21fn,2 Ze 211jJn (0) 12 (5.40)
and the energy shift (5.36) becomes

f).E~2) = ~Z e:h311jJn(0)12In I mc
2
I (5.41 )
3 m c (En' - En)av.
We know from the theory of the hydrogen atom that states with nonzero
angular momentum (l i- 0) vanish at the origin:
1jJni (0) = 0 for l i- o.
Only s states have a nonvanishing probability within the nucleus:

(5.42)

Here n represents the principal quantum number and aB = h 2/(me 2) is the

Bohr radius. Still we need a value for (En' - En)av. ; Bethe calculated this
value numerically to be 17.8 Ry ~ 242.2 eV. Thus he arrived at

;:=2
f).E(2)
~ 1040 MHz. (5.43)

In view of the approximations employed, this result is in excellent agreement

with the experimental value for the Lamb shift of 1057 MHz. Kroll and LamV 9
performed the relativistic calculation for the self energy. They could confirm
Bethe's conjecture about the convergence of the relativistic generalisation

19 N.M. Kroll, W.E. Lamb, Jr.: Phys. Rev. 75 (1949) 388.

156 5. Infinities in Quantum Electrodynamics: Renormalization Problems

of expression (5.34) and could achieve an even better agreement with the
experimental result; see the volume on quantum electrodynamics. 2o
In the lectures on quantum electrodynamics we realize that no infinity (in-
cluding vacuum polarization, which we will not touch here) is observable; they
are included in the finite values of the observed mass and charge (mass and
charge renormalization). Not all field theories are renormalizable. However,
quantum electrodynamics is renormalizable and because of that it is possibly
the most successful theory in physics. Nevertheless, the removal of infinites
by renormalization is not satisfying aestheticly; we return to this point in our
lectures on QED.

5.4 Is There an Inconsistency in Bethe's Approach?

The largest radius of the electron charge distribution useful for our calcula-
tions is given by the Compton wavelength a = h/me. We write the interaction
with the radiation field as

HI = -~
me
JF(x - x')p· A(x') d 3 x' ,

F(x) where F(x) represents the charge distribution of the electron. Hence, the
electron interacts with only that part of the radiation field for which A < a,
i.e. lik < me, holds. The maximum photon momentum is given by k max =
me/Ii. On the other hand, the electron is localized within the Bohr radius
ao = li2 /me 2 . In the course of the calculation, the dipole approximation has
x been applied, i.e. exp(ikx) ~ 1; however,
me li2 lie
Fig. 5.14. Possible form-
factor for the electron
kx ~ kmaxao = Ii me2 e2

and

(~) -1 = ~ ~ 137,

so the applicability condition for the dipole approximation is no longer ful-

filled. Furthermore, the largest photon energy is assumed to be identical to
the electron rest mass me2 . Again, this assumption makes sense only if the
maximum momentum corresponds to k max = me/Ii, so that the electron is
smeared out over the Compton wavelength. However, all measurements of the
electron charge distribution clearly signal that the electron is pointlike to a
very good degree (re «0.1 fm). Hence, there is no physical justification for the
cut-off procedure employed by Bethe. Therefore his result has no significance;
but historically it is interesting. The correct (relativistic) calculation, which
does not need these dubious approximations was performed by Schwinger,

20 W. Greiner, J. Reinhardt: Quantum Electrodynamics, 2nd ed. (Springer, Berlin,

Heidelberg 1994).
 5.4 Is There an Inconsistency in Bethe's Approach? 157

Weisskopf, and Feynman 21 and by Kroll and Lamb. Also the so-called vac-
uum polarization is an important contribution to the Lamb shift. We discuss
it in the volume on quantum electrodynamics, where the correct relativistic
calculation due to Wichmann and Kro1l 22 is presented. In recent years it has
become possible to ionize heavy atoms completely and the investigation of, for
example, lead (82 protons) with only a single electron present allows the direct
measurement of the Lamb shift. This opens an interesting field of research,
particularly for testing quantum electrodynamics in strong central fields. 23

EXERCISE

5.6 The Lamb Shift

Problem. Solve the classical equations of motion for a particle in an oscil-

lating electromagnetic field and calculate the mean quadratic displacement of
the electron resulting from the action of a superposition of electromagnetic
waves where the energy ~ mu is associated to every mode of the radiation field.
Calculate the change in the potential energy for the electron resulting from
the displacement. After that, employ the additional term in the potential to
calculate the shift of the hydrogen energy level (Lamb shift) with the help of
perturbation theory.

Solution. At first we solve the classical equation of motion for the displace-
ment llx of a particle in an oscillating electric field E = Eo exp(iwt):
d2 = llx = __
~E ~Eoeiwt ==} e_ E - -e- E 0 eiwt
dt 211x m m mw 2 mw 2 '
so that

We now assume that E contains the modes of the (quantized) radiation field,
so that we can substitute E ---> 2.:ka Eka and obtain the expectation value

(lllxI2) -_ ~ " IEwkal

m 2 '~ 4
2
.
k,a k

Since

Ekrr =
. (2----v-
1
~
"~2)1/2
7rnc
wk
1/2

it follows (see Exercise 1.3) that

21 J. Schwinger, V.F. Weisskopf: Phys. Rev. 73 (1948) 1272;
R.P. Feynman: Phys. Rev. 76 (1948) 939.
22 E.H. Wichmann, N.M. Kroll: Phys. Rev. 101 (1956) 843.
23 See e.g. S.M. Schneider, W. Greiner, G. Soff: "Vacuum polarization contribution
to hyperfine-structure splitting of hydrogenlike atoms", Phys. Rev. A50 (1994)
118; Z. Phys. D31 (1994) 143; H. Persson, S.M. Schneider, W. Greiner, G. Soff,
J. Lindgren: "Self-energy correction to the hyperfine-structure splitting of hydro-
genlike atoms", Phys. Rev. Lett. 76 (1996) 1433.
158 5. Infinities in Quantum Electrodynamics: Renormalization Problems

Exercise 5.6.

and
27rn e2 "'" 1
£3 m 2 ~w3

J
k,a k

27rn e2 £3 1
£3 m2 (27r)3 2 d3 k(ek)3

-2- - ne-2
(27r)2 m 2 e3
1 J 00
k 2 dk
-
k3
- dD

2(~)2 roo
0

.: e dk ,
me 7r nc k io
where the factor 2 keeps track of the polarizations (J. This integral diverges
logarithmically; hence we introduce a cut-off for both integration boundaries:

(I~xl
2
) =
2 e2
- -
(
-
n ) 21krnax -dk .
7r nc me km'n k
The displacement ~x means the change in the potential given by (Taylor
series expansion):
V(x + ~x) = [1 + ~x· V + ~(~x· V)2 + ... J V(x).
We average the potential over all directions. Here we have to consider that
the spherical symmetric potential V(x) depends only on the modulus of x:

(V(x+~x)) = ~Jsin'l9d'l9(1+~X'V+~(~x'V)2+ ... )V(X).

The term '" ~x . V represents an odd function. Thus, the average vanishes:

(V(x + ~x)) = ~J sin 'l9 d'l9(a· V)(a· V)V(x) + V(x) ,

where a = ~x. Now

(a· V)(a· V)V(x) aiOiaiJjV(r) (sum convention!)
aiajOiOjV(r)
and
or oV

so that
X·oV
(a· V)(a· V)V(r) aao-.2-
, J ' r or

a.a.
'J
[(8rij _ Xi Xj ) oV
r3 or
+ XjXi
r2
o2V]
or2
2
( a _ (a.r)(a.r)) av +(a.:')202V
r r3 ar r or2
a2 ( 1 - cos 2 'l9 ) -oV + a 2 cos 2 'l9--
-
02V
r or ar2
 5.4 Is There an Inconsistency in Bethe's Approach? 159

and Exercise 5.6.

J sin 19 d19 (a· V)(a· V)V(r)

17r sin 13 d13 [:2 (1 - cos213) ~~ + a cos213 ~:~]2

a2 2 OV 21 02V
---+a ---
r 3 or 3 or2

a2 [~OV + 02V]
3 r or or2 .
The expression within the last square brackets is identical to V 2 V (because
of the spherical symmetry!); hence,
(v(x + ~x)) = (1 + i(lL\xI2)V2) V(x) .
With

v r
and

V2V = _e 2 V 2 (~)
it follows that

Vpert = ~(I~xI2)V2V

27re 2 (l~xI2)<53(x)
3

~ e4 (~)2In (kmax ) <53(x) .

3 lie me kmin
This is a perturbation potential. Thus, in lowest-order perturbation theory
the energy shift becomes

~E (1PlVpertl1P)

e (~)2 In (k max )
~3 tie
4
me kmin
3
(1P1<5 (x)I1P)

e4 (~)2In (k max )
~3 lie 11P(O)12,
me k min
where 1P is the unperturbed ground-state wavefunction.
This expression has exactly the form of (5.41), from which we have calcu-
lated the Lamb shift. However, the cut-off momenta tik min and tik max remain
to be determined. The infrared divergency is not physical because a bound
and localized electron cannot be put into oscillations by photons with long
wavelenghts. Thus, kmin ~ l/aB = a/A. should hold. Here aB = ti2/(me 2 )
denotes the Bohr radius and A. = ACompton/(27r) = h/(27rme) = ti/(me) the
Compton wavelength divided by 27r. As the upper boundary we choose again
160 5. Infinities in Quantum Electrodynamics: Renormalization Problems

Exercise 5.6. the Compton wavelength kmax = 1/A. In the framework of the classical theory
we cannot give a satisfactory derivation for the latter. With (5.42) we arrive
at

b.E = - 4 a5
-me
371" n 3
(1)
2 Z 3 In -
Za
Onl.

With this result the Lamb shift of the hydrogen atom (Z = 1) becomes b.E ~
660 MHz, which in view of the many crude approximations is in surprisingly
good agreement with the experimental result. This illustrative explanation
of the Lamb shift as a consequence of the vacuum fluctuations was given by
Welton in 1948. 24

24 T. Welton: Phys. Rev. 74 (1948) 1157.

6. N onrelativistic Quantum Field Theory
of Interacting Particles and Its Applications

From Chap. 3 we know how an arbitrary number of particles, which do not

interact with each other but which are subject to an external potential V (x),
are to be described within the framework of second quantization. Now we will
extend this formalism with respect to particles interacting with each other.
First, we construct an additional term in the Hamiltonian, which describes
the potential interaction energy V (x, x') between two particles at positions
x and x', respectively. As a generalisation of (3.63) we make the ansatz

fI = Jd3x~t(x,t)fI~~(x,t)
+~Jd3x Jd3x/~t(x,t)~t(xl,t)V(x,xl)~(x,t)~(xl,t), (6.1)

where
ll,2v 2
fI~ = - - - + V(x) (6.2)
2m
represents the unperturbed one-particle Hamiltonian for a particle in an ex-
ternal potential V (x). ~(x, t) represents the field operators (3.43). fI itself is
the Hamiltonian for the many-particle system in second quantization or, more
informatively, the field-theoretical Hamiltonian. We have to find its solutions,
i.e. its eigenstates. In order to see the connection between the field-theoretical
problem (6.1) and the "classical" many-particle problem of elementary quan-
tum mechanics we introduce the n-particle state

-J
IXn, t) - d 3 Xl .. ' J
d 3 XnXn (Xl, ... , Xn) tJi'At (Xl, t) ... tJi'At (Xn' t) 10)
(6.3)
in analogy to our previous considerations (3.66). Xn (Xl"'" Xn) represents
an ordinary function of the n-particle coordinates Xl,"" x n . According to
the conventional rules of quantum mechanics, we interpret
1Xn (Xl, ... , Xn) 12 d3Xl ... d3Xn (6.4)
as the probability of finding particle 1 in d3xl, particle 2 in d3x2,"" and
particle n in d3xn. We determine the amplitude function Xn (Xl, ... ,xn ) in
such a way that IXn, t) is an eigenstate of the field-theoretical Hamiltonian
(6.1):
(6.5)

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
162 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Indeed, it will be shown in Exercise 6.1 that (6.5) provides the following
defining equation for the amplitude function Xn (Xl, ... ,Xn):

[ t;n ( /i2
-2m V;+V(Xi)
) 1 n n ]
+2t;t;V(Xi,Xj) Xn(XI, ... ,Xn)

= EnXn(XI, ... ,X n ). (6.6)

This implies that Xn (Xl, ... , Xn) obeys the conventional many-particle
Schrodinger equation. Hence, the field-theoretical problem (6.5) is equivalent
to the conventional many-particle problem (6.6). This is the same result as
obtained in Chap. 3 [see (3.73)]; now this equivalence also holds for interacting
particles.

EXERCISE

6.1 The Field-Theoretical Many-Particle Problem

Problem. Show the equivalence between the field-theoretical many-particle

problem H IXn, t) = En IXn, t) and the conventional many-particle problem

[~( - :: V; + V (Xi») + ~ f.1V (Xi,Xj)] Xn (Xl, ... ,Xn )

where

Solution. We write HIXn, t) = EnlXn, t) as

(Ho + V) IXn, t) = En IXn, t) ,
where

Ho J d 3 x tirt (x, t)H~tir(x, t)

with

and

V = ~J d3 x J d 3 x'tir t (x,t)tir t (x',t)V(x,x')tir(x,t)tir(x',t).

We have already shown in (3.73) the equivalence between HoIXn, t)

EnlXn, t) and the "conventional" many-particle problem. Thus, it remains
to show that
VI Xn, t)
is equivalent to
 6. Nonrelativistic Quantum Field Theory of Interacting Particles 163

1
2'LV(x i ,Xj)Xn(X1, ... ,xn) . Exercise 6.1.
i,j

We write
V IXn, t)
~/ d 3x / d3x'~t(X,t)~t (XI,t)V(X,XI)~(X,t)~(X',t)
x/ d3x~ ... / d3x~Xn(x~, ... ,x~)~t(x~,t) ... ~t(x~,t)IO)
~/ d 3x d 3x' d3x~ ... d3x~ ~t (x, t)~t (x', t) V (x, x')
x ~(x, t)~(X', t)Xn(x~, ... ,x~)~t(x~, t) ... ~t(x~, t) 10).
We note the commutation relation
~(X)~t(X') = ±~t(X')~(X) + 6(x - x'),
V IXn, t)
~/ d 3x d 3x ' d3x~ ... d3x~ ~t (x, t)~t (x', t)V(x, x')
I I A A I At I At I
xXn(x1, .. ·,xn)!li(x,t)!li(x ,t)!li (x1,t) ... !li (xn,t) 10)
~/ d 3x d 3x ' d3x~ ... d3x~ ~t (x, t)~t (x', t)V(x, x')
x Xn(X~, ... , x~)~(x, t) [6(x' - x~) ± ~t(x~, t)~(X', t)]
x ~t (x~, t) ... ~t (x~, t) 10) .
The successive commutations are performed analogously to the method for
(3.69) and the equations following it:
Vlxn,t)
~ / d3xd3x~ ... d3x~ ~t(x, t) L v (x, x~) Xn (x~, ... , x~)
,
x ~(x, t)~t (x~, t) ... ~t (x~, t) 10) .
In the same manner the second annihilation operator goes through the com-
mutations; it produces a second sum

Vlxn,t) = ~/ d3xd3x~ ... d3x~~t(X',t) ... ~t(x~,t)2..:V(xj,x~)

',)
xXn (x~ ... x~) 10) .
Since this holds for arbitrary states
~t (x~, t)··· ~t (x~, t) 10) ,
it follows that

~ LV (x~,xj) Xn (x~, ... ,x~)

i,j
164 6. Nonrelativistic Quantum Field Theory of Interacting Particles

For the case when the external potential V(x) vanishes the n particles
move only as a result of their mutual interactions. Then the field operator -tP
can be expanded into plane waves:

-tP(x, t) = bk (t)L
k
ik·x
&. (6.7)

If we further assume that the particle-particle interaction depends only on

the relative distance between the particles, i.e.
v (x, x') = V (x - x') , (6.8)
(6.1) becomes

A '"' '"' At A
H = L...J L...J bk1 bk2 2m
!i2k~ JV
d3 x
exp [i (k2 - k 1 ) . x]
k, k2

+ '"'
L...J '"'
k,
L...J '"'
L...J '"'
k2 k3 k4
At bk2
L...J bk1 At bk3
A bk4
A d x
V J J 3 3 '
x
dL3

X exp [i (k3 - kd . x] exp [i (k 4 - k 2) . x'] V (x - x') (6.9)

We employ

J d3x ei(k 2-k,)·x

L3 = c5k2 ,k , (box normalization)

and use the Fourier transform of the interaction potential:

V(q) = J~:V(x)eiqX; (6.10)

then (6.9) simplifies to

A '"' !i2 k 2 At A
H = L...J 2m bkbk + '"' '"' '"' - At At A A
L...JL...JL...J V(q)bkl+qbk2_qbk2bkl· (6.11)
k k, k2 q

The last term becomes more plausible once we calculate in more detail:

= V(k3-kl)c5k3+k4,k,+k2 (6.12)
with the substitution z = x - x'. We set q = k3 - kl' so that
k3 + k4 = kl + k2
or

follows from the Kronecker delta symbol. Therefore the last term of (6.9)
becomes
 6.1 Quantum Gases 165

L L L bL bl,bk,+qbk2-q V (q)
k, k2 q
LLLbl~_qbl~+qbk~bk~ V(q)
k~ k~ q

(6.13)

which is identical to the last term in (6.11). The interaction term arising in
(6.11) is illustrated in Fig. 6.1.
The operators bk 1 and bk2 annhilate two particles with corresponding mo-
menta Ilki and Ilk2 at the vertex. In the same interaction (at the same vertex)
the operators bL +q and bl, _q create two new particles with momenta Il( ki +q)
and ll(k2 - q). For this process, momentum conservation is guaranteed since
Il(k i + q) + ll(k2 - q) Ilk I + Ilk2, (6.14)
(momentum after collision) (momentum before collision) . k2
Fig. 6.1. Momentum con-
This momentum conservation stems from the assumption (6.8) that the inter- servation at a vertex for
action between particles depends only on x - x', which reflects translational two- particle collisions
invariance. We can rephrase this result: During the scattering the two parti-
cles exchange the momentum Ilq (at the vertex); the corresponding amplitude
is determined by v (q).

6.1 Quantum Gases

A quantum gas is defined as a large number of particles that collide (inter-
act) with each other according to the interaction potential v (x - x'). The
particles arrive with a variety of different momenta. Within this ensemble we
denote the number of particles with momentum Ilk by N (k); N (k) will change
with time. Mutual interactions mean that particles with momentum Ilk are
scattered to another momentum; on the other hand particles with a different
momentum can adopt the momentum Ilk after the collision. The change in
N(k) is described symbolically by the following equation:

XX
k k' k'-q k+q

fJN(k)
(6.15)
fJt
k' q

k+q k'-q k' k

Evidently the X represents a binary scattering process. We sum over all

binary collisions that lead to a particle with momentum Ilk after the collision;
they increase the number N(k). We subtract all processes that annihilate a
166 6. Nonrelativistic Quantum Field Theory of Interacting Particles

particle with momentum lik; they decrease the number N(k). In order to
translate the symbolic relation (6.15) into a quantitative equation we have
to introduce the transition probability per unit time for each diagram. We
calculate the latter in first-order perturbation theory, where the interaction
term is treated as a perturbation potential. We use

bnl ... N n · .. ) vNnl ... N n - l ... ) ,

(6.16a)
b~1 ... N n · .. ) VNn + 11 ... N n + 1 ... )
for bosons and

bnl ... N n · .. ) envNnl ... I-Nn ... ) , (6.16b)

b~1 ... N n · .. ) e nVI-Nn l ... 1- N n ... )
for fermions, where the phase en has been explained previously in (3.42).
With this, the transition probability per unit time for the diagram depicted
in (6.15) becomes, according to Fermi's golden rule,

( trans. prob.) k
time X k
I = 211" 1~
h 12
V(q) N(k + q)N(k' - q)

k+q k'-q
x [1 ± N(k)][1 ± N(k')]

x 0 (:: (Ik + ql2 + Ik' - ql2 - k 2 - k,2) )

(6.17)
The plus or minus sign within the factors (1 ± N) has to be chosen for bosons
or fermions, respectively. Since the matrix element enters quadratically into
Fermi's golden rule the square-root factors appearing in (6.16) are carried into
Nand (N ± 1). The argument of the 0 function expresses energy conservation
for the scattering process; note that because of elastic collisions only kinetic
energy is considered. Analogously to the process for (6.17), the transition
probability per unit time for the second scattering process appearing in (6.15)
can be calculated. We then arrive at the following master equation for (6.15):
8N(k)
at
LL
k' q
2; IV (q) 120 ( : : (Ik + q 12 + Ik' - q 12 - k 2 - k'2))

x {N(k + q)N(k' - q) [1 ± N(k)] [1 ± N(k')]

- N(k)N(k') [1 ± N(k + q)] [1 ± N(k' - q)]} . (6.18)
We realize that because of factors 1 - N(k), transitions into already occu-
pied states (N (k) = 1) vanish for fermions; this reflects the Pauli principle.
For bosons on the other hand, transitions into already occupied states are
amplified; this is reflected in the factors of the type 1 + N(k). The master
equation (6.18) represents the quantum-mechanical generalization of the Boltz-
mann equation. The scattering probability therein is proportional to IV (q) 12;
 6.1 Quantum Gases 167

this goes back to the fact that outside the interaction region the particles
here move freely without interactions, so the matrix element (6.10) is calcu-
lated with plane waves (Born approximation). In order to find an equilibrium
solution of the Boltzmann equation (6.18) we have to require that
aN(k) = 0
at . (6.19)

This leads to the sufficient condition

N(k + q)N(k' - q) [1 ± N(k)] [1 ± N(k')]
= N(k)N(k') [1 ± N(k + q)] [1 ± N(k' - q)] . (6.20)
We will solve this equation in Exercise 6.2; the result is

N(k) = 1 (6.21)
Aexp [E(k)/kBT] 'f 1 .
Here, A is a normalization constant,

E (k) = h2 k 2 (6.22)
2m
is the kinetic energy, and kBT represents an energy related to temperature.
The plus and minus signs in (6.21) hold for fermions and bosons, respectively;
only with this assignment (6.21) is a solution of (6.20) possible.

EXERCISE

6.2 Equilibrium Solution

of the Quantum-Mechanical Boltzmann Equation

Problem. Find the equilibrium solution of the Boltzmann equation. Proceed

from the sufficient condition (6.20).

Solution. We start from

N(k + q)N(k' - q) [1 ± N(k)] [1 ± N(k')]
= N(k)N(k') [1 ± N(k + q)] [1 ± N(k' - q)]
and
N(k + q)N(k' - q) N(k)N(k')
[1 ± N(k + q)] [1 ± N(k' - q)] [1 ± N(k)] [1 ± N(k')]
or
I ( N (k + q) ) I ( N (k' - q) )
n 1 ± N (k + q) + n 1 ± N (k' _ q)
N(k)) (N(k'))
= In ( l±N(k) +In l±N(k') .
So In[N/(1 ±N)] is a collision invariant. Functions a(k) that fulfill this balance
for all possible collisions are given by
168 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Exercise 6.2. 1 (particle-number conservation) ,

E(k) (energy conservation)
owing to energy and particle-number conservation. Therefore every collision
invariant is a linear combination of these elementary invariants, i.e.

In C:~{k)) = ao + a(r, t) - (3(r, t)E(k)

or
N(k) = 1
Aexp [(3(r, t)E(k) - a(r, t)] =t= 1 '
where A = e- ao is a normalization constant. We define the local temperature
as
A 1
(3(r, t) = kBT or T(r, t) = kB(3(r, t) ,
kB being the Boltzmann constant, and the chemical potential as
a(r, t)
p, = - -
(3(r,t)'
Then we deduce
N(k) = [Aexp{(3(r,t)[E(k)-p,(r,t)]}=t=l]-l,

for - Bose-Einstein -----+ 1+N

for + Fermi-Dirac -----+ 1- N .
If the chemical potential vanishes we obtain (6.21).

The normalization constant A is determined from the total number N p of

particles in the system according to

N p = LN(k) =
k
L3 J d3 kN(k)

L Jd kAexp(h k /2mk
3 3 1
BT) ± 1 .2 2
(6.23)

The distribution functions calculated with (6.21) are called the Fermi-Dirac
(for the plus sign) and the Bose-Einstein (for the minus sign) distributions. l In
order to proceed further we need the notion of the entropy of a quantum gas,
which characterizes the disorder of the gas. We will explain some basics about
entropy in Exercise 6.6, where we will also derive the following expression:
S = ±kB L [(1 ± N(k)) In(l ± N(k)) =t= N(k) lnN(k)] . (6.24)
k
1 Historically it is interesting to know that Einstein was the referee of the manu-
script of the Indian physicist Bose, unknown at that time, which had been sub-
mitted to Zeitschrift fUr Physik. He realized its deep implications and improved
the paper by suggesting some valuable comments to Bose.
 6.1 Quantum Gases 169

ks represents the Boltzmann constant. Again, the upper sign holds for bosons
and the lower sign for fermions. We will now show that entropy will increase
permanently, i.e.
dS
-dt >
-
o. (6.25)

This follows from the Boltzmann equation (6.18). For the proof we restrict
ourselves to the case of bosons; for fermions the arguments are the same. We
start from (6.24), which we differentiate with respect to time; then we employ
again the Boltzmann equation (6.18). This gives
dS
dt
kB ~ {O~~k) {In[l + N(k)]-lnN(k)}
oN(k) 1 oN(k) 1 }
+ [1 + N(k)]----at [1 + N(k)] - N(k)----at N(k)

ks L O~~k) {In[l + N(k)]-lnN(k)}

k

ks L L L ~ IV(q)1
k k' q
2

x 8 [2n~ (Ik + q 12 + Ik - q 12 - k 2 - k'2)]

x {N(k + q)N(k' - q)[l + N(k)][l + N(k')]
- N(k)N(k')[l + N(k + q)][l + N(k' - q)]}
x {In[l + N(k)]-lnN(k)} . (6.26a)
From (6.26a) three different, but similar, equations can be derived. The first
one follows by making the replacements q -> -q, k -> k + q, k' -> k' - q.
With the fact that 1V( -q) 1= 1V(q) 1taken into account this equation reads
dS
dt kS LLL 2;IV(q) 12
k k' q

x 8 {:~ [Ik 12 + 1k'12 - (k + q)2 - (k' - q)2] }

x {N(k)N(k')[l + N(k + q)][l + N(k' - q)]

- N(k + q)N(k' - q)[l + N(k)][l + N(k')]}
x {In[l + N(k + q)]-lnN(k')} . (6.26b)
We obtain the third equation by making the replacements q -> -q and k -> k'
in (6.26a):
dS
dt
kB LLL 2;IV(q) 12
k k' q

x 8 [2n~ (Ik' - q 12 + Ik + q 12 - k'2 - k 2)]

X {N(k' - q)N(k + q)[l + N(k')][l + N(k)]
170 6. Nonrelativistic Quantum Field Theory of Interacting Particles

- N(k)N(k')[l + N(k' - q)][l + N(k + q)]}

x {In[l + N(k')]-lnN(k')} . (6.26c)
The fourth equation results from the substitutions q --+ -q and k --+ k' in
(6.26b):
dS
dt
kB LLL 2;IV(q) 12
k k' q

x 8 {:: [k'2 +k2- (k' _ q)2 - (k + q)2] }

x {N(k')N(k)[l + N(k' - q)][l + N(k + q)]
- N(k' - q)N(k + q)[l + N(k')][l + N(k)]}
x {In[l + N(k' - q)]-lnN(k')} . (6.26d)
Addition of all four equations (6.26a)-(6.26d) yields

4 dS
dt

x 8 [:: (Ik + q 12 + Ik' - q 12 - k'2 - k 2)]

x {N(k + q)N(k' - q')[l + N(k)][l + N(k')]

- N(k)N(k')[l + N(k + q)][l + N(k' - q)]}
x {ln {N(k + q)N(k' - q)[l + N(k)][l + N(k')]}
-In {N(k)N(k')[l + N(k + q)] [1 + N(k' - q)]}} . (6.27)
With the abbreviations
u= N(k + q)N(k' - q)[l + N(k)][l + N(k')] (6.28a)
and
v= N(k)N(k')[l + N(k + q)][l + N(k' - q)] , (6.28b)
the two last factors appearing in (6.27) can be written as
(u-v)(lnu-Inv) . (6.29)
This quantity is positive for u > v and also for u < v; it vanishes for u = v.
The condition u = v is just the equilibrium condition (6.20) or (6.19). Hence,
we conclude that
dS
- > 0 (6.30)
dt
always holds except for equilibrium. Entropy increases steadily and reaches its
maximum once the system is in equilibrium. This important conclusion con-
tains, on the one hand, the second law of thermodynamics (increase of entropy)
and, on the other hand, the so-called H theorem from Boltzmann (entropy
is at a maximum in equilibrium). We consider now the classical Boltzmann
equation, which we obtain from (6.18) once we perform the following limiting
cases:
 6.1 Quantum Gases 171

(a) h -+ 0 (classical limit) ,

(b) £3 -+ 00 (large volume) , (6.31)
(c) l±N(k) ~ 1 (gas is far from degeneracy) .
The first two conditions do not need further explanation. Condition (c) means
that every state k is occupied according to a distribution function N(k).
However, the latter should be close to zero so that the average occupation
numbers for a momentum interval are equal for the classical and quantum-
mechanical case.
We then say that the gas is far from degeneracy. Now we introduce further
the classical velocity distribution function f (v) via

LN(k)
k
===} £3 J d3 vf(v) (6.32)

and take notice of the substitutions

hk -+ mv and hq -+ mu. (6.33)
Thus, the quantum-mechanical momentum lik and momentum transfer liq
of the collision are identified with the classical particle momentum mv and

J
classical momentum transfer mu, respectively. Then

~
=? £3 J d3
_q_
(211")3
= £3 m 3
li3 (211")3
d3 u (6.34)

and (6.18) becomes

of (v)
at J J d3v' d3u £6 m 3 211" 1 (mu) 12
h3 (211")3 Ii Ii
v
xo ( -Iv+ul
m 2+-Iv
m, -ul 2 - -mv 2 - -mv '2)
2 2 2 2
x [f(v+u)f(v'-u)-f(v)f(v')]. (6.35)
Rewriting the Fourier transform (6.10) as

V(q) =
(211" )3/2
-£3
J 3 eiq . x
d xV(x)~

J
(211")
(211"1i/m)3/2 3 eimu,x/fi _ (211"h) 3/2 V (u)
d xV(x) - - --
£3 (211"1i/m)3/2 - m £3
(6.36)
and setting the transition probability as

2; V 1 (u) 12 = W (u) , (6.37)

we find that (6.35) becomes

8fa~v) J J d 3 v' d 3 uw (u)

m m, m 2 m 2]
xo [ 2(v+u) 2 2
+2(v -u) - 2 v - 2 u
x[f(v+u)f(v'-u)-f(v)f(v')]. (6.38)
172 6. Nonrelativistic Quantum Field Theory of Interacting Particles

This is the classical Boltzmann equation; it corresponds to the master equation

(6.15) and the quantum-mechanical Boltzmann equation (6.18). The quantity
n does not occur anymore. Note, however, that we tacitly made the desci-
sian in (6.37) to replace the quantum-mechanical transition probability by a
(classical) expression w (u) that is more or less independent of This makes n.
sense because for both the quantum-mechanical and for the classical case a
probability exists for the scattering of two particles with momentum transfer
mu. Each single term in (6.38) may be interpreted and understood according
to the master equation (6.15). The c5 function expresses energy conservation
in each single collision. We underline these points further with the following
exercises.

EXERCISE

6.3 Equilibrium Solution of the Classical Boltzmann Equation

Problem. Show that

f (v) = exp (-mv 2 /2kBT)
represents an equilibrium solution of the classical Boltzmann equation. Is this
solution unique?

Solution. We denote

h f (v + u) h f (v' - u)
(1)
f{ f (v) , and f~ f (v') .

The integrand of (6.38) vanishes if

hh = fU~ and lnh + lntz = lnff + lnf~· (2)
Functions a (v) are called collision invariants if they fulfill this balance for all
possible collisions. In fact, (2) states that the value of the quantity In h + In h
remains the same before and after the collision. As a result of energy and
momentum conservation
ai(v) mVi (i = 1,2,3) ,
a4 (v) mv 2
2
represent collision invariants. Conservation of particle number yields another
collision invariant ao (v) = 1. In addition to these five functions ai, only the
angular momentum may be thought of as a linear independent collision invari-
ant. Since usually "infinitely extended" gases are considered, the definition of
an angular momentum L = r x p makes no sense since no reference point
exists for r. Consequently, In f (v, r, t) has to be a linear combination of the
ai:
lnf(v,r,t) = a(r,t)+A(r,t)mv-{:J(r,t);v 2 .
 6.1 Quantum Gases 173

We define the temperature T = 1/ (kBfJ) , the chemical potential f.t(r, t) Exercise 6.3.
a(r, t)/fJ(r, t) and the average flow velocity u = )../fJ(r, t) and obtain

f(v,r,t) = ~C:Tf/2exp{fJ(r,t)(f.t(r,t)-;[v-u(r,t)l2)}, (3)

where the particle density p = N /V has been introduced so that

J d3 vf(v,r,t) = p.

If the average velocity u = 0 and f.t = 0, we obtain the simplified form

3/2
f (v, r, t) = p
(
~k ) exp (-mv 2 /2kBT) (4)
27r BT

EXERCISE

6.4 From the Entropy Formula for the Bose (Fermi) Gas
to the Classical Entropy Formula

Problem. Show that expression (6.24) for the entropy of a quantum gas
becomes the classical formula for the entropy

S = -kB L N(k) InN(k) ,

k

when the gas is far from degeneracy.

Solution. The expression for entropy is given by

S = -kB L {N(k) In N(k) =f [1 ± N(k)]ln[1 ± N(k)]} ,

k

where the upper sign holds for bosons and the lower sign for fermions. As we
have already stressed a classical ideal gas is characterized such that the num-
ber of possible states for the particles is large and that the average occupation
number N(k) of each single state remains small, i.e. the particles occupy a
large number of states independently. The effects of quantum statistics, which
for the case of fermions allows up to only one fermion per state but for bosons
allows the average occupation number to become arbitrary large, do not play
a role in the classical limit. We denote gases for which the average occupa-
tion number N(k) « 1 remains small as ''far from degeneracy". Since then
N (k) « 1 , we get 1 ± N (k) ~ 1 , and the expression for the entropy of the
Bose (Fermi) gas becomes

Selass. = -kB L N(k) InN(k) .

k
174 6. Nonrelativistic Quantum Field Theory of Interacting Particles

EXERCISE

6.5 Proof of the H Theorem

Problem. Use the classical expression for the entropy derived in the previous
problem to prove the H theorem (pronounced 'eta theorem', because H is
meant to represent the greak letter eta).

Solution. The classical expression for entropy is given by

S = -kB L N (k) In N (k) ,
k
i.e.
dS = -kB
dt
L aN(k)
k at
[lnN(k) + 1]

We insert aN(k)/at and use immediately the classical distribution function

(6.35); it follows that
dS
dt
x 8 (m Iv + ul 2 + mIv' _ ul 2 _ mv2 _ mV'2)
2 2 2 2
x [J (v + u) f (v' - u) - f (v) f (v')] [lnf (v) + 1]
The equilibrium solution of the Boltzmann equation reads
f(v) = Cexp(-mv 2 /2k B T) .
Inserting this solution into dS /dt, we immediately realize that the bracket
[J(v + u)f(v' - u) - f(v)f(v ' )]
vanishes, i.e. dS / dt = 0 in equilibrium.
On the other hand,
+ u)f(v' - u) - f(v)f(v ' ) = 0
f(v
follows from the requirement dS / dt = 0; hence, we conclude that the Maxwell-

)]
Boltzmann distribution

f (v) = C exp [- C::~

represents a unique solution.

6.2 Nearly Ideal, Degenerate Bose-Einstein Gases

We again focus on (6.23), which we will now study in more detail for the case
of a Bose--Einstein gas. Considering the Bose-Einstein distribution law we
realize that the number of molecules or atoms in the state with energy E = 0
is given by
 6.2 Nearly Ideal, Degenerate Bose~Einstein Gases 175

1 1
No = A-I (6.39a)
e XO - 1 '
where the normalization constant A has been written as
(6.39b)
Rearranging (6.39a) we get:

e
~xu
= -No
-. (6.40)
No + 1
Since No cannot become negative and since (6.40) approaches 1 for large No,
we deduce that
(6.41)
The value for the normalization constant xo is determined from the require-
ment to conserve the particle number:

N - '" 1 (6.42)
- ~ eo exp (Ek/kBT) - 1.
The ratio between the total number of particles and the number of particles
No in the state with E = 0 is given by
No 1 ( 1 ) (6.43)
N = N e XO -1 .
If e Xu > 1 it holds that No/N < 1; if, on the other hand, exa ~ 1 we get
(No/N) --7 1. Hence, the state with E = 0 is strongly populated only for
e xa ~ 1. Particles in such a state are called a condensate. In order to analyse
the expression for N further, we separate the state with E = 0 from (6.42)
and change the sum over k into an integral:

N -
- eo
1
---
-1
+ -1 (sm)3/2
4
-2
!i
7r £ 3
1
0
00
El/2
dE -----::-=-:-:---=:---
exoexp(E/kBT) -1 '
(6.44)

where we have also performed a substitution from momentum to energy E =

(1ik)2/2m. With a further substitution the integral can be written as
3/2
~ ( SmkBT ) £3 F' (6.45)
4 fi2 7r

with

F' (6.46)

and
x = E/kBT.
We expand the integral:

F' =

(6.47)
176 6. Nonrelativistic Quantum Field Theory of Interacting Particles

We integrate term by term and use

(6.48)

then

F'

(6.49)

Hence, we obtain

(6.50)
or
Nh3
e-xo + 2- 3 / 2 e-2xo + 3- 3 / 2 e- 3xo + ... = (6.51)
(27rmkBT)3/2 £3 '
if e Xo is not close to 1. We can rewrite (6.51) as

T_~[_l
- 27rmkB L3F
(N __
l-1 )]2/3 .
e XO
(6.52)

Owing to (6.41) the largest value of e- XO becomes ~ 1, so the largest value of

F' [in (6.50)] is
Fa = 1+T 3/ 2 + 3- 3 / 2 + . .. = ((3/2) ~ 2.612 , (6.53)
where ((x) = 2:;1 p-x represents Riemann's ( function. In the region where
e XO becomes so large that we can neglect 1/ (e Xo - 1) with respect to N we
derive with (6.53) that

(6.54)

In the second region, e Xo is close to 1, so we can approximate F by Fa.

However, e XO should still be sufficiently different from 1 so that 1/ (e Xo - 1)
can be neglected:
2
T xO "" 1 = 27rmk B
h (N
L3 Fa
)2/3
= Ta (6.55)

In the third region, F ~ Fa and eXO become so close to 1 that l/(e Xo - 1)

cannot be neglected anymore; instead, we get
1 1
(6.56)
eXO - 1 Xa
Thus,

N (6.57)
or
 6.2 Nearly Ideal, Degenerate Bose~Einstein Gases 177

1
(6.58)
Xo
Since Xo > 1 this region corresponds to T < To. We arrive at the result that
a critical temperature

To - - -
2
/i - - - - (P) 2/3
(6.59)
- 27rmkB 2.612
exists, which depends on the density p = N / L3 of the Bose~Einstein gas;
beneath this critical temperature nearly all particles are in the state with
energy E = O. This phenomenon is known as Bose~Einstein condensation. It
should be clear that for T = 0 all particles occupy the state with E = O.
Only one fluid occurs in nature for which this quantum effect becomes
observable macroscopically: liquid helium, 4He. We insert its density into re-
lation (6.59) and obtain the critical temperature To = 3.13 K. Below this
temperature 4He represents a macroscopic Bose-Einstein condensate. 2
After these introductory statistical considerations we now focus on the
quantum features of the Bose~Einstein gas. First we consider the noninter-
acting ground state of a Bose gas with N particles
l<po(N)) = IN, 0, 0, .. ·) , (6.60)
i.e. all N particles occupy the state with E = O. In the following, we will
consider a box of volume V = L3 and periodic boundary conditions. The
creation and annihilation operators for the state with E = k = 0 are denoted
as o'b and ao. The usual relations

0,0 I<Po (N)) N l / 21<po(N - 1)) ,

(6.61 )
o'bl<po(N)) (N + 1)1/21<po(N + 1))
hold. Neither ao nor o'b annihilate the ground state; this is in contrast to the
well-known property 0,010) = 0 of a vacuum state, in which no particles are
present. However, the Bose operators multiply the ground state with large
numbers of the order VN. Therefore it is convenient to switch to new opera-
tors bo and bb which we define as
bo = v~1/2ao and bb = v~1/2ab . (6.62)
Since
1,
we get
A At
lbo, bol V~l
, (6.63)
and
2 Recently Bose~Einstein condensates of various total spin F = 0 atoms have been
observed: see e.g. M.H. Anderson, J.R. Ensher, M.R. Matthews, C.E. Wiemann,
E.A. Cornell: Science 269 (1995) 198; K.B. Davis, M.O. Mewes, M.R. Andrews,
N.J. van Druten, D.S. Durfee, D.M. Kurn, W. Ketterle: Phys. Rev. Lett. 75
(1995) 3969; C.C. Bradley, C.A. Sackett, J.J. Tollett, R.G. Hulet: Phys. Rev.
Lett. 75 (1995) 1687.
178 6. Nonrelativistic Quantum Field Theory of Interacting Particles

(N)
V
1/2
Ipo(N -1)) ,
(6.64)
( ----v-
N+l)1/2
Ipo(N + 1)) .

We are interested in the behavior of a gas, i.e. in large particle numbers and
volumes. Hence, we consider the so-called thermodynamic limit: N --+ 00,
V --+ 00, but constant density p = N jV --+ const. Then bo and bb multiply
the ground state with a constant whereas the commutator becomes
A At
[b o, bol --+ 0 (6.65)
in the thermodynamic limit, i.e. we obtain the classical limit for which bo and
bb are ordinary C numbers. 3 For the case of photons (see Chap. 1) we have
already seen that a field for which the quanta are particles behaves classically
once a large number of particles occupy the same state.
So far we have considered a non interacting ground state; now we inves-
tigate a system of bosons at T = 0, which is slightly nonideal. We keep the
interaction term from (6.11) but, for simplicity, we replace v(q) by the factor
gj2V:
H = ~ to. At A 9 ~ At At A A ;:
~ItWkakak + 2V (6.66)
A

~ aklak2ak3ak4ukl+k2,k3+k4'
k klk2k3k4

9 is often called the pseudopotential. Owing to momentum conservation we

manipulate the wave-number vectors k l , ... , k4 appearing in (6.66). Further-
more, we then separate the contribution for k = 0 as an interaction term of
the Hamiltonian

+ ~'
~
(4 At A At A + AtAt
akakaoaO
AA
aka_kaOao
AtAt A A ) ]
+ aoaoaka-k ; (6.67)
k

here 2:' signals that terms with k = 0 have to be omitted from the summation.
Also, we have kept only terms up to second order in ak and and have al
neglected fourth-order terms; momentum conservation means that terms with
an odd number of these operators do not exist.
As a further approximation we replace ao and ab in (6.61) and (6.67) by
C numbers:
(6.68)
where No represents the number of particles in the state with k = O. It follows
that

~
9
2V (J\T2 +
1Vo
27\T
1'0 L;:
~'(AtA At A )
akak + a_ka_k

"T~'(AtAt
+ 1'0 AA))
~ aka_ k + aka-k . (6.69)
k
3 The phrase C numbers means "commuting" numbers, i.e. ordinary real or complex
numbers.
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 179

Analogously we get for the particle number operator

N' = " "t,
~ akak = 1Vo
1\T 1""(,t,
+2 ,t')
~ akak + a_ka_k . (6.70)
k
We set p = N/V, insert (6.70) into (6.69), and obtain

fI = ~Vgl+~L'[(EZ+pg)(atak+a~ka_k)
k

(6.71)

Here we have assumed that N - No « N and, consequently, that terms like

(I:k' at ak)2 can be neglected. This also means that most particles are in
the condensate. We look for a solution of our problem (6.71), Le. the energy
eigenvalues of the Hamiltonian (6.71). For this special case the problem can be
solved exactly with a canonical transformation (Bogoliubov transformation):
We introduce new operators 4

ak uk.;h + vkA~k' Ak Ukak - vka~k ,

(6.72)
akt uk A'tk + vkA_k,
' 't
Ak ' - Uk a_
Vkak ,t .
k
The coefficients should be real and spherically symmetric. The transformation
is canonical, once the new operators obey the canonical commutation rules:

(6.73)

By inserting (6.72) into (6.73) we realize that this is only fulfilled if

uk - v~ = 1 (6.74)
holds for every k. For reasons that will become clear later on we call the new
operators quasiparticle operators. We insert (6.72) into (6.71) and obtain

H, =
1
2Vgp 2
+"
~ (Ek + pg)Vk + pgukVk
,'[ 0 2 ]

+ ~ L ' {[(EZ + pg)(uk + V~) + 2UkVkpg] (AtAk + A~kA_k)}

k
1",,{[
+2 ~ pg(uk2 + Vk)
2
+ 2UkVdEk0
+ pg)
]

X (AtA~k + AkA_ k )} , (6.75)

where Uk and Vk are related via (6.74); however, the corresponding ratio may
be chosen freely. In order to make fI diagonal in At, Ak we use this freedom
to eliminate the last line in (6.75); we require that
pg(uk + v~) + 2UkVk(EZ + pg) = o. (6.76)

4 For the following, see also Sect. 7.1, where we follow the same steps in a slightly
different way.
180 6. Nonrelativistic Quantum Field Theory of Interacting Particles

We choose
Uk = cosh 'Pk , Vk = - sinh 'Pk (6.77)
such that (6.74) is fulfilled; then (6.76) becomes
2UkVk _ pg
(6.78)
u~ + v~ - EZ + pg
or
2 sinh 'Pk cosh 'Pk sinh 2'Pk pg
(6.79)
sinh2 'Pk + cosh 'Pk
2 cosh 2'Pk + pg . EZ
Use of the known trigonometric relations then yields
pg
tanh 2'Pk = EO . (6.80)
k + pg

Since -1 ::; tanh 2'Pk ::; 1 this last equation can be fulfilled for all k only once
9 > 0, i.e. the pseudopotential has to be repulsive! It follows that
v~ = u% - 1 = ~ [E;1(E2 + pg) - 1] , (6.81)
where we have defined
Ek = [(E2 + pg)2 + (pg)2] 1/2 . (6.82)
With the use of (6.74) to (6.82) we are able to write the Hamiltonian (6.75) in
its final form:

fl = ~Vgp2+~~'[Ek-(E2+pg)]
1~' At' At'
+"2 L Ek(Ak Ak + A_kA_k) . (6.83)
k

Here E2 stands for the free particle energies, which are obtained from (6.82)
if the interaction 9 = 0, Le. E2 = (hk)2/2m. The operator AlAk resembles
a particle-number operator and has eigenvalues 0,1,2, .... Hence, the ground
state of fl is determined by the requirement that
for all k i- o. (6.84)
Be aware that 10) represents a complicated combination of unperturbed eigen-
functions since neither ak nor al
annihilate the groundstate 10). Because of
(6.84) we obtain for the ground-state energy

E =
,
(OIHIO) = "2 Vgp
1 2
'7' (Ek -
+"21~' 0
Ek - pg). (6.85)

Furthermore, all excited states correspond to different numbers of noninter-

acting bosons where each boson possesses the excitation energy E k :
2 2 ]1/2
Ek = [ (h;~ + pg) _ (pg)2 (6.86a)

For k ---> 0 this becomes

Ek(k ---> 0) ( pg) 1/2

>';:j m hk = cshk . (6.86b)
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 181

We realize that (6.86a) and (6.86b) describe a system of quanta with mo-
mentum fik. The operators ih, At
annihilate and create those quanta we have
called quasiparticles. The quantity Cs in (6.86b) represents a velocity, which we
can interprete as the speed of sound in the nearly ideal degenerate Bose gas.
The long-wave excitations (quasiparticles) corresponding thereto are called
phonons. In the limit of large momenta (6.86a) becomes lik
fi 2 k 2
Ek ~-- (6.87) Fig. 6.2. Quasiparticle
2m ' energies. For low momen-
i.e. for large momenta the quasiparticles do not interact with each other. tum the energy grows lin-
Hence, we obtain the qualitative result presented in Fig. 6.2. early with k, for large
momentum quadratically.
In this section we have established that particles of an interacting gas
These two domains cor-
can be treated as a gas consisting of noninteracting quasiparticles. Phonons respond to the phonon
represent a first example for quasiparticles. Later on we will encounter further and particle sectors, re-
examples. spectively

EXAMPLE

6.6 Entropy of a Quantum Gas

In order to understand the physical significance of the quantity "entropy"

we consider a closed macroscopic system. If we separate a small part of the
system from the rest, this subsystem is no longer closed and can interact
with other parts of the system. We now investigate one of these subsystems;
its distribution function w(E) is a function of energy alone. For a better
understanding we imagine that our system is confined to a fixed box. We use a
coordinate system in which the box does not move. Under these circumstances
momentum and angular momentum no longer represent constants of motion.
Energy remains the only constant of motion that determines the distribution
function of the system. With r(E) we denote the number of quantum states
for which the energy is smaller than or equal to E. Thus, the number of states
of the subsystem with energies between E and E + dE is given by
dr(E) dE (1)
dE
and the probability distribution for the energies reads

W(E) = d~~) w(E) (2)

with the normalization condition

J W(E) dE = 1. (3)

We proceed with the assumption that W(E) possesses a sharp maximum at

E = E (average) and that it differs significantly from zero only in the vicinity
of E. Therefore we can introduce a width b.E such that
W(E) b.E = 1. (4)
182 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Example 6.6. The number of quantum states that correspond to the energy interval ilE is
given by

ilr = d~~) ilE. (5)

The normalization condition (4) can then be written as
w(E)ilr = 1. (6)
The interval ilE corresponds approximately to the average fluctuation of the
energy of the subsystem; hence, ilr represents the "order of smearing" of
the macroscopical state of the subsystem over its microscopical states. For a
classical system, w(E) is replaced by its classical distribution function p(E)
and ilr corresponds to the phase-space element ilp ilq in which the system
prevails for nearly all the time. In order to find the number of states we
consider the Bohr-Sommerfeld quantization rule

-1
2n1i
f pdx = n+-1
2
for a particle with one degree of freedom. The integral f p dx represents the
area confined by the closed classical phase trajectory of the particle, i.e. the
trajectory in the p-x plane (phase space) of the particle. By dividing this area
into cells of area 27rn each we obtain n cells in total. On the other hand, n
represents the number of quantum-mechanical states with energies that are
not larger than the corresponding value of the phase trajectory considered.
We realize that in the quasiclassical limit every quantum-mechanical state
corresponds to a phase-space cell with area 27rn. Put into other words, the
number of states in the volume element ilp ilx of phase space is given by
ilpilx
27rn .
If we consider s degrees of freedom this number turns out to be
ilPl ilXl ilP2 ilX2 ilps ilxs
27rn 27rn 27rn
i.e.
ilps ... ilps ilXl ... ilxs
(27rn)s
in total. We introduce the abbreviations ilp ilPl ... ilps and ilq
ilXl ... ilxs and obtain
ilpilq
(27rn)s .
This relation yields well known results; for example, for one particle
d3 p
(27rn)3
represents the number of states within the momentum region d3 p and the unit
volume of the configuration space.
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 183

Hence, we arrive at the "weight" of a phase-space cell: Example 6.6.

fj.T = fj.p fj.q . (7)
(27rh)s
Another approach to the derivation of the number of states in the phase-space
volume is as follows. The number N of plane waves in a box of length £ with
periodic boundary conditions for the walls is given by
£3
N = (27rP dkx dky dkz ;

see (2.22). For a small cube dV = d 3x we can write

dN = dV dkx dky dk x = dVdpx dpy dpz = d 3 xd 3 p
27r 27r 27r 27rh 27rh 27rh (27rhP'
In addition, if every particle in the plane-wave state possesses s spin degrees
of freedom 1,2, ... ,s we get
dN dNl dN2 ... dNs
d3Xl d 3Pl d3x2 d 3 p2 dqdp
(27rhP (27rh)3 (27rh)3s '
where dq = d3xl ... d 3 x s and dp = d 3pl ... d 3 ps. This is exactly the former
result (7); here we have counted only the three space and three momentum
degrees of freedom for each of the particles separately, whereas in (7) we
counted all degrees of freedom from 1 to 5. Note that we have here 2 x 35
degrees of freedom, whereas in (7) we have 2 x 5.
fj.T is called the statistical weight of the macroscopic state of the subsys-
tem, and its logarithm
S = Infj.T (8)
is called entropy. Hence, in the classical limit we have
fj.p fj.q
S = In (27rh)s . (9)

Since the number of states fj.T always has to be fj.T » 1 we conclude S ::::: O!
In statisitical mechanics it is customary to set
S = kB Infj.T (10)
instead of the dimensionless expression S = In fj.T. Since entropy has to fulfill
the thermodynamic relation T dS = dE the proportionality constant kB has
to have the dimension

[ te;;::!~ure] .
kB is called the Boltzmann constant. With this relation we are able to derive
the entropy for an ideal, a Bose-Einstein, and a Fermi-Dirac gas.
(a) Ideal Gas. We consider a group of N j particles as an independent system.
Its statistical weight is given by fj.Tj , and the statistical weight of all particle
groups is given by
184 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Example 6.6. 1l.T = II 1l.T j . (11)

j

Within the classical limit the average occupation numbers are small compared
to unity, i.e. the number of particles N j is small compared to the number Gj
of states (Nj « G j ); however, N j should still be a large number. Therefore
we assume that the particles are distributed completely independently of each
other over the various states. Once we assign to each of the N j particles one
of the G j states we arrive at (Gj)N; possible distributions. Some of them are
identical because N j ! permutations between the N j undistinguishable particles
are possible without changing the distribution. Hence, the statistical weight
is given by

(12)

and we get

S kB In 1l.T = kB L In 1l.Tj =?
j

S = kB L(Nj InGj -lnNj!) . (13)

j

Since N j is large we can use Stirling's formula (see Exercise 6.8) and write
InNj!::::O N j In Nj/e, i.e.

""'
S = kB~Njln eGo
NJ (14)
j J

or, introducing the average particle number per state n = Nj/G j ,

S = kBLGjnjln~ (15)
j nj
or
(16)
j j

The first part becomes

Sl = kB L Gjnj = kB LNjnj = kBN = const.

j j

This constant is often dropped. Then the entropy of the ideal gas reads

S1 = -kB L Gjnj In nj ; (17)

j

this corresponds to the expression derived earlier.

(b) Fermi Gas. Now N j ::::0 G j . The number of ways to distribute N j identical
particles over G j states in such a way that a state is occupied with one particle
at most results from the number of combinations of N j out of Gj elements.
This reflects a standard task of combinatorics.
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 185

Combinatorics can be illustrated with playing cards. For most of the hands Example 6.6.
of cards it does not matter in which order the cards are handed out since
afterwards it is possible to rearrange them arbitrarily. Instead of the possible
k-tupel only the subsets consisting of k elements of the given set are of interest.
They are called combinations of order k of n elements without repetition. For
an ideal gas we have seen that the number of permutations of n elements is
given by n!. If only k elements are picked from the n elements then n elements
are available to occupy the first place of the k-tupel; for the second place only
n - 1 elements are available and for place k only n - (k - 1) elements remain.
Hence, the number of variations of order k consisting of n elements without
repetition is given by
n!
n· (n - 1) . (n - 2) ..... [n - (k -1)] = (n _ k)! . (18)

(Example: For six people to take a seat from ten chairs 10!/4! = 151200 pos-
sibilities exist.)
If we further allow the possibility of rearranging k elements, i.e. k permu-
tations, then we obtain exactly

k!(nn~ k)! - (~) (19)

combinations. (Example: In a lottery, 6 numbers are picked out of 49 numbers.
The number of possibilities is (~9) = 13983816.)
After these considerations we obtain for the statistical weight of N j
fermions in G j states

(20)

We take the logarithm, make use again of Stirling's formula In N! ;:;:::

N In(NIe), and obtain

SF = kB LG j InGj - N j InNj - (G j - N j ) In(Gj - Nj ) . (21)

j

Again we introduce the average number of particles per state Tij = NjlG j
and it follows that
SF = kB LG j [Tij In Tij + (1 - Tij) In(1 - Tij)] . (22)
j

For the case that the G j is independent of j we can take them out of the
summation and arrive at the entropy relation for the Fermi gas already men-
tioned.
(c) Bose Gas. For this case an arbitrary number of particles can occupy
each quantum state. As a consequence the statistical weight 6..rj is given as
the number of ways to distribute N j particles over G j states; all particles are
allowed to occupy the same state. Using arguments analogously to the case
of fermions we obtain
(Gj + N j - I)! .
6..r, = (23)
< (G j - 1)!Nj ! '
186 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Example 6.6. see Exercise 6.7. We neglect the -1 because it is small compared to the G j
and N j , take the logarithm, and again use Stirling's formula. This results in

BB = kB l:)Gj + Nj ) In(Gj + Nj ) - N j InNj - G j InGj (24)

j

or with nj

BB = kBLGj [(l+nj)ln(l+nj)-nj lnnjJ (25)

j

Again, if we assume G j to be independent of j we arrive at the entropy

expression for the Bose gas given earlier.
In the limit Nj «: Gj , i.e. nj «: 1, the entropy for both quantum gases
becomes identical to the one for an ideal Maxwell-Boltzmann gas.

EXERCISE

6.7 Distribution of N Particles over G States

(N umber of Combinations)

Problem. Show that the number of ways of distributing, without restriction,

N particles over G states is given by
M = (G+N -I)!
(G - l)!N!

Solution. We imagine N balls as points lying on a line:

···11····1·1··
We numerate the G cells and separate them by G - 1 vertical lines. In the
sketch shown we have: 3 balls in the first cell, 0 in the second, 4 in the third,
1 in the fourth, and 2 in the fifth. The total number of sites that are either
occupied by points or lines is G + N - 1. Hence the number of possibilities is
identical to the number of ways of choosing G - 1 sites for the vertical lines,
i.e. the number of combinations of G - 1 elements out of N + G - 1; thus,
(N + G -I)! (N+G-1)!
M=
(G -I)! [N + G - 1 - (G - l)J! (G -l)!N! .

EXERCISE

6.8 Stirling's Formula

Problem. Prove Stirling's formula

InN! = Nln (~)

for large N.
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 187

Solution. It holds that

In N! In (1 x 2 x 3 x ... x N)
In 1 + In 2 + In 3 + ... + In N
N
L In(n) .
n=2
For large N we replace the sum by an integral:
2 345

jN In(x) dx = [x In(x) - xlf = N In(N) - N + 1 . Fig. 6.3. Approximation

leading to Stirling's for-
mula
Since N is very large we neglect the 1 and write N = N In e; we obtain

InN! ~ Nln (~)

for N » I! According to Fig. 6.3 we realize that the sum Z=~=2In(n) (illus-
trated by the step function), i.e. In N!, is approximated by the two integrals:

jN In(x) dx < InN! < jN In(l + x) dx;

hence, Stirling's formula represents a lower estimate for In N!. Since

j{' In(l + x) dx = (N + 1) In(N + 1) - N +1- 2ln 2

ft In(x) dx NlnN -N+1
--> 1
'
for N --+ 00, the relative error of Stirling's approximation vanishes as N --+ 00.

EXAMPLE

6.9 Entropy and Information

In this example we will study briefly the connection between entropy and the
information about a system.
First, we have to find a statistical definition of information. We consider
a situation for which Po different successive situations may arise, all with the
same probability. We denote the information as I. For the initial situation
10 = 0 since all processes are equally probable; we have no information about
the system.
We consider two independent processes, which come with probabilities POI
and P02 , respectively; the total probability is then given by
Po = POI P02 · (1)
The information should be additive, i.e.
I = It + 12 . (2)
This property leads to the following, plausible definition of information:
I = KlnP (3)
188 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Example 6.9. with a constant K. Hence, we obtain

I = Kln(POl P02 ) = h + h· (4)
Furthermore it is customary to think of the information I as a dimensionless
quantity, so the constant K has to be a number. In order to find a reasonable
unit for information we consider the following example.
A problem exists with n different and independent choices; each single
choice corresponds to either 0 or 1. The total number of possible choices is
then P = 2n; the information becomes I = KIn P = K n In 2. It is desirable to
identify I with n and to take into account the binary structure of the decision
between 0 and 1. Hence, we write
1
K = -1- = 10g2 e .
n2
This choice defines the unit of information: The bit (from "binary digit").
That is,
I = 10g2 P (5)
measures the information in bits.
Example: We choose one from 32 cards and the information we have is
(6)
If we choose one card each from two sets of cards, consisting each of 32 cards,
then the information is given by
I = 10g2(2 10 ) = 10 bits
since P = PI P2 with PI = P2 = 32 = 2 5 . The property of addition of the
information is realized.
So far we have considered events that all come with equal probability.
However, if we consider a set of G letters then our familiar formula would be
I = (G log2 27) bits
since 26 letters constitute our alphabet and the spacing between the words is
considered the 27th letter. The information per letter would then be

~ = 10g2 27:::::; 4.76 bits.

However, this result is not conclusive; we know that different letters do not
appear with the same frequency within a sentence. For example, the frequency
of the letter "E" is approximately p :::::; 0.105, whereas "Q" appears with only
p:::::; O.OOL
In order to find the correct relation we consider an "alphabet" consisting
of only two "letters": 0 and 1. In total we have G sites: No sites have a "0"
and NI sites have a "1".
It holds that No + NI = G and the probabilities that one of the G sites
contains a "0" or "1" are
No
Po = G and PI (7)
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 189

i.e. Example 6.9.

Po + PI = 1. (8)
N ow we can fill the G sites with numbers 0 and 1, one number for each site
only! This automatically leads to the number of ways to occupy No of the sites
with zeros since the remaining sites are then filled with ones. The number of
ways to fill No sites with zeros is equal to the number of combinations of G
objects from which No are picked, i.e.
G!
p = N o!N1 !. (9)
Hence, the information becomes
I = KlnP = K(lnG!-lnNo!-lnN1 !).
If the "sentence" we inspect is very long we can make use of Stirling's formula
InN! ~ N (inN -1), i.e. if G» 1, No» 1, and Nl »1, then
I ~ K{G[ln(G) -1]- No [In(No) -1]- Nl[ln(Nl) -I]}. (10)
Since G = No + Nl it follows that
I ~ K(GlnG - No In No - NdnNl) (11)
or

I~-KG
NoI nNo
(- NlI nNl)
-+- - .
G G G G
With (7) we obtain the information per "letter" of the message:
I
G ~ -K(po lnpo + pdnpI).
This generalized to M different symbols 0,1, ... is called Shannon's formula:
I = -KGl.:>jlnpj. (12)
j=1

If we consider the realistic frequencies of the various letters we obtain for

our example of a 27-letter alphabet information of I IG ~ 4.03 bit per letter,
i.e. smaller than calculated (4.76 bit). Our definition of the information (3) is
completely equivalent to the definition of entropy
(13)
where kB represents the Boltzmann constant. Hence, it makes sense to choose
K = kB for the constant K and to define a new, thermodynamic unit for
the information. In order to characterize the relationship between entropy
and information in more detail we differentiate between two different forms
of information:
(1) free information, which occurs if the possible cases can be viewed as
abstract and have no assigned physical meaning,
(2) bound information I g , which occurs if the possible cases are interpreted
as a manifestation of a physical system. (Planck has denoted these forms
as complexions).
190 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Example 6.9. It is this bound information which is connected to the physically relevant
situations and which we take to be related to entropy. We identify the possible
cases with the complexions; then it holds that

bound information number of entropy

complexions
initial state Igo = 0 Po 8 0 = kB In Po
final state Ig1 -=I- 0 PI < Po 8 1 = kB In PI

because Ig1 reduces the number of possible states from Po to Pl.

It is obvious that the system considered cannot be isolated, because when
information is obtained, the entropy decreases; the number of possibilities, i.e.
complexions, is reduced. This information must come from an outer system
for which entropy increases.
We obtain:
Io o and Po complexions,
It > 0 and Po complexions,
where It = K In(Po/ Pt) and thus (K = k B)
Ig1 = kB(lnPo -lnPt) = 80 - 81 (14)
or
81 = 80 - Ig1 , (15)
respectively. The bound information appears as a negative term in the total
entropy of the system:
bound information = entropy decrease = increase of negentropy N ,
where negentropy stands for the negative of entropy.
Let us consider a closed system. Then the law on the increase of entropy
gives
or (16)
i.e. an increase in entropy 8 1 can cause an increase in the entropy 8 0 as well
as a decrease in the information I g1 : As the entropy of a system increases the
information about this system decreases at the same time. An increase of ne-
gent ropy results in a gain of information, i.e. negentropy corresponds directly
to the information about a system. Let us now elucidate the significance of
these conceptions with several examples. At first we point out the difference
between free and bound information.

(1) A person possesses information (free).

(2) This person passes the information to a friend via acoustic waves (voice)
or electric waves (telephone). The information is realized in a physical
process and, hence, becomes bound information.
 6.2 Nearly Ideal, Degenerate Bose-Einstein Gases 191

(3) The friend is partly deaf and does not understand some words; thus, the Example 6.9.
bound information gets lost.
(4) After some time the original person forgets the information and loses the
free information since it was only present within thoughts (brain).

In the following we will consider a further example which clarifies the

relationship between information and entropy.

EXAMPLE

6.10 Maxwell's Demon

We consider a container divided into two parts A and B. A small hole is
present in the partitioning wall between A and B. Both volumes are filled
with the same gas under identical pressure. We now assume a living creature,
i.e. the demon, exists who is able to "see" every single molecule. He opens
and shuts the hole in such a way that the fast molecules from B go into A and
that the slow molecules from A pass into B. Hence, the temperature rises in
A without the input of work from outside the system. This is a contradiction
of the second law of thermodynamics .

..• ••A.·· • B.·

..•••...•..••••••
.....
..1. ••••• ~

• • • •• • Fig. 6.4. Typical set-up

for Maxwell's demon

In order to exorcize the demon we have to raise the question of whether

he can really see each single molecule?5 We have assumed the system to be
isolated; hence, the demon stays in a container with constant temperature,
which is filled with the corresponding black-body radiation. Here comes the
solution: It is impossible to see anything in the interior of a black body. The
demon can see only the thermal radiation but no molecules. Thus, he cannot
use the shutter and cannot violate the second law.
What other possibilities may come to his mind? He might get a lamp so
that he can see the molecules. However, the lamp represents a radiation source
which is not in equilibrium with the black body; it produces negative entropy
within the system. The demon is able to see the molecules, i.e. the negentropy
is transformed into his information. Now he can fulfill his demonic destiny:
he produces a higher temperature in A, i.e. negative entropy; he transforms
his information into negentropy. We arrive at the cycle
negentropy -+ information -+ negentropy.
5 P. Demers: Can. J. Research 22 (1944) 27; L. Brillouin: J. App!. Phys. 22 (1951)
334.
192 6. Nonrelativistic Quantum Field Theory of Interacting Particles

Example 6.10. Since the transformation process is at best complete, the demon will also
allow some "wrong" molecules to pass from time to time and, thus, it holds
again that
6..8 :::: 0, because 6..N ::; 0,
Le. negentropy N gets lost and entropy is produced; as a consequence the
second law of thermodynamics is not violated!
7. Superfiuidity

In the preceding chapter we introduced the concept of a "quasiparticle". We

now intend to elucidate this notion with respect to the interpretation of ex-
perimental facts and refer to the properties of liquid helium.
Two stable isotopes of the element Helium occur in nature: 3He (two pro-
tons and one neutron: a fermion) and 4He (two protons and two neutrons: a
boson). Both isotopes remain liquid down to temperatures T ~ 0 K at low
pressures; only at high pressures (of about 30 atmospheres) they do solidify.
This is in contrast to the remaining elements, which become solid easily. Since
these properties are related to quantum effects, liquid helium is denoted as a
quantum liquid. Fermi statistics state that every quantum state of 3He can be
occupied by only one particle. On the other hand 4He obeys Bose statistics;
thus it has the properties of a nearly ideal Bose gas with a weak interaction
between the particles. We will examine 4He in the following. If 4He, which

P/atm
50

30
He II
10 Fi~. 7.1. Phase diagram
of He at low temperature
o 6 8 9 10 TIK
To

is in equilibrium with its vapour, is cooled below To = 2.17 K, it develops

into a new phase denoted as Hell. This new phase has remarkable properties:
He II flows through tiny capillary tubes with no friction; evidently, its viscos-
ity is zero. On the other hand, direct measurements of the viscosity (with a
cylindrical viscosimeter) show that the viscosity of He II corresponds to the
viscosity of He I above To. This peculiar behavior has been explained with a
two-fluid model by Landau 1 and Tisza2 : He II consists of two fluids penetrat-
ing each other. One component is superfiuid and therefore shows no viscosity;
1 L.D. Landau: J. Phys. (Moscow) 5 (1941) 71; 11 (1947) 91.
2 L. Tisza: Nature 141 (1938) 913.

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
194 7. Superfluidity

Pn/P the flow is curl free, Le. curl Vs = 0, where Vs represents the velocity of the
superfluid component. Its density is Ps. The second component with density
Pn and velocity Vn is represented by a "normal" fluid, which has viscosity.
This picture from Landau has been confirmed experimentally. 3
The presence of two components in He II allows two different excitation
T modes, which differ in the relative orientation between Vs and Vn: Once Vs
and Vn are parallel a perturbation in He II leads to a change in density and
Fig. 7.2. The ratio of nor- pressure (for constant T and constant entropy); it is denoted as the first (or
mal fluid density pn to to- ordinary) sound.
tal density P = pn + p. If Vn and Vs point to opposite directions a wave is obtained, leading to
as a function of tempera- periodic changes in Psi Pn. This ratio depends on temperature and P = Ps + Pn
ture (qualitatively). P. in-
is constant (also the pressure is constant), so we obtain a temperature wave,
dicates the density of the
superfluid phase the so-called second sound. A detailed investigation of this problem shows that
the corresponding speed of sound is proportional to (PsIPn)1/2. The second
sound was observed for the first time by Peshkov. 4

7.1 Basics of a Microscopic Theory of Superfiuidity

If liquid helium (4He) is cooled down to almost absolute zero its viscosity van-
ishes (to the degree of measureability) and it flows without friction through
tiny capillaries and tiny openings. This was observed for the first time by
Kapitza. 5 Landau's explanation of this phenomenon uses the concept of a
"quantum liquid" (bosonic liquid), as we have discussed in the previous sec-
tion. Now we will investigate in detail the microscopic theory of superfluid
helium developed by Bogoliubov. 6 The method developed is of general inter-
est and often finds applications once perturbation theory becomes useless, for
example for the theory of superconductivity (see Chap. 8).
4He atoms represent neutral bosons (spin 0) and interact only weakly
with each other because of polarization forces (van der Waals forces). The
Hamiltonian for N 4He atoms can be cast into the form
N
H = L H(ri) + L V(lri - rjl) . (7.1)
i=l i<j

H(ri) represents the operator for the kinetic energy of the ith He atom and
V(lri - rjl) describes the interaction between the ith and jth atoms. For
a complete and orthogonal set of eigenfunctions we choose the solutions of
H(ri); these are given by plane waves

'Pk(r) = me'
1 'k
·r (7.2)

which are normalized with the conventional periodic boundary conditions for
a large volume £3 [compare for example with (1.23)]. Hence, it holds that
3 E.L. Andronikashvili: J. Phys. (Moscow) 10 (1946) 20l.
4 V.P. Peshkov: J. Phys. (Moscow) 10 (1946) 389.
5 P.L. Kapitza: Nature 141 (1938) 74.
6 N.N. Bogoliubov: J. Phys. (Moscow) 9 (1947) 23.
 7.1 Basics of a Microscopic Theory of Superfiuidity 195

(7.3)
2m
According to Chaps. 3 and 6 [(6.7) to (6.13) in particular]' the total Hamilto-
nian can be rewritten in second quantization:

with

V(q) J d 3 r V(lrl) e+ iq . r

-J drd(cos'!9) d<pr2 V(r)e+iqrcos19

J
-271' drr 2 V(r) 1- 1
dxe- iqrx

J
271' dr r2 V(r) eiqr ~ e- iqr
lqr

~J dr rV(r) sin(lqlr) . (7.6)

Here
(7.7)
represents the momentum transfer of particle 1 onto particle 2. The creation
and annihilation operators for the 4He atoms have been denoted as [;t and [;,
respectively. Conservation of total momentum is expressed by the Kronecker
delta symbol
(7.8)
appearing in (7.5). We obtain this symbol because of the box normalization.
In quantum electrodynamics, where we use a normalization with respect to
o-functions (infinitly large box), (7.8) becomes a O(kl +k2 - k~ - k;) function.
196 7. Superfluidity

The second term in (7.4) describes a two-particle interaction or, to be more

precise, a two-boson interaction. Every single term within this summation
describes the annihilation of a pair of 4He atoms with momenta lik~ and lik~
and the creation of another pair with momenta likl and Iik 2. According to
(7.7) and (7.8) conservation of the total momentum
lik~ + lik~ = 1ik 1 + Iik2 (7.9)
Fig. 7.3. Momentum con-
servation at an interaction holds for each such scattering process (see Fig. 7.3). This stems from the mo-
vertex mentum conserving potential V(lri -rjl), which only depends on the distance
of the interacting particles and is therefore translationally invariant.
Inserting (7.5) into (7.4) we get

if = L ckhlhk + 2~3 L (7.10)

k k, +k2=k~ +k~
where the summation runs over only the momentum-conserving combinations
of kl' k2, k~, k~.
For the ground state of a system of bosons all particles occupy the lowest
one-particle state. Contrary to fermions, bosons do not have to obey the Pauli
principle; this leads to the condensation of bosons.
The lowest one-particle state is characterized by k = 0, which is equivalent
to Ck=O = o. For the case when a small repulsion exists between the atoms,
most of the atoms will still remain in the lowest state with k = 0 and remain
in the condensate. In other words, even if the interaction V between the 4He
atoms stays small, no atoms will remain in the condensate, where no deviates
only slightly from the total number N of 4He atoms.
It should be clear that at the very end of our calculations the volume size
has to go to infinity, i.e. L3 -+ 00; in this way we remove the dependence on
the periodic boundary conditions for a normalization box with finite length
L. Then, of course, N -+ 00 increases, too. We require for this limit that the
density
N
= const. (7.11)
£3
remains constant. Now
and (7.12)
where no represents the number operator for particles in the ground state.
Because of (OlnoIO) = no;:;:; N it follows for their difference that
A At T
bobo _ bobo = ~ -+ o. (7.13)
£3 £3 £3
In the limit L3 -+ 00 we are allowed to forget the operator character of h6
and ho; within this limit they commute and can be replaced by numbers.
Bogoliubov suggested the introduction of new boson operators
 7.1 Basics of a Microscopic Theory of Superfluidity 197

IT
ak
no k '
yInObob
(7.14)
At 1 At A
ak yInObkb o
no
for k '" o. They represent boson-boson-hole pairs. Equation (7.10) then be-
comes
_1_N2V(0) + ",' (lik)2 ata
2£3 L...J 2m k k
k

+ 273 L' V(k)[a~a~k + aka-k + 2a~akl + iI'

k

iII + iI' . (7.15)

The prime (') attached to the summation sign indicates that values k = 0 are
excluded from the summation. Those terms that contain three or four boson
operators a, at are collected in iI' and can be neglected because the boson
operators (7.14) are of the order liN. This can be understood from a small
excercise:

k
+ 1 At A no + 1
' " At A no
L...Jbkbk-- - bobo- -
k no no
IV (no + 1_
no (no + 1))
no no
A) no
(N -no - -
A
+ 1
no
~ N - no -> 0 for L -> 00 , (7.16)
where (7.14) has been used. It follows that a~ak '" I1N 2 so
",'At A ",' 1 N 1
L...J akak '" L...J N2 '" N2 = N -> 0.
k k
Thus, in the limit N -> 00 we are allowed to neglect the term iI' in (7.15),
which contains four operators of a and at. We focus on the interesting part,
Le.

(7.17)
198 7. Superfiuidity

No doubt the interaction matrix elements V(k) are small, but perturbation
theory is not allowed to run the show. Infinitely many terms exist in the
interaction sum of (7.17) for pairs k and -k once Ikl ---. O. In other words,
scattering out of the condensate with one-particle states k = 0 into infinitely
many states k, -k might take place which are energetically all close to Ikl = 0;
a perturbation calculation then becomes impractical.
Bogoliubov suggested the following approach: Take the Hamiltonian fIo
and diagonalize it with a canonical transformation in the boson operators:

u(k)Ak + v(k)A~k'
(7.18)
u(k)At + v(k)A_k .
This is exactly the same as in (6.72). We will repeat now some of the steps
presented in Sect. 6 in more detail and with a different stress. u(k) and v(k)
represent real functions with the property

u( -k) u(k) ,
(7.19)
v( -k) v(k) .

This guarantees that no asymmetry occurs between the bosons at and a~k'
or lh and A~k' respectively. Also, (7.18) can now easily be transformed into

U(k)ak - V(k)a~k
u 2 (k) - v 2 (k)
(7.20)
A'tk --
u(k)at - v(k)Lk
u 2 (k) - v 2 (k)
This suggests the normalization
u 2 (k) - v 2 (k) = 1. (7.21)
The latter follows from the requirement that the Ak should again fulfill the
boson commutation relations
[Ak,A
, 't]k, _ 8kk , ,

[Ak' Ak,]_ 0, (7.22)

['t
Ak,A't]
k, _ O.

Of course, these relations must also hold for ak, at. This implies

[u(k)Ak + v(k)A~k,U(k')At, + v(k')A_k,L

u(k)u(k') [Ak' At,] _ + v(k)v(k') [A~k' A_k,]_
u2 (k)8kk' - v 2 (k)8 kk ,
(u 2 (k) - v 2 (k)) 8kk' (7.23)
 7.1 Basics of a Microscopic Theory of Superfluidity 199

and we arrive at the stated normalization relation (7.21). The normalization

(7.21) and the reality of u(k), v(k) are fulfilled with the ansatz
1
u(k)
VI ~ D~ ,
(7.24)
Dk
v(k) =
VI ~ D~ .

For the moment the values of the coefficients Dk remain unknown. Inserting
(7.24) into (7.18) we deduce

1 ' 't
~(Ak+DkA_k) ,
(7.25)
1 't '
~(Ak+DkA_k) .
1 ~ Dk

Using this result for ih in (7.17), we finally express iII in terms of the new
boson operators Ak , At. This calculation will be presented in detail in Exercise
7.1. Here, we state the outcome:

iIo + LE(k)AtAk, (7.26)

k

where

~: V(O) + ~ L k
[E(k) ~ C~~2 + ~~V(k))] (7.27)

and

E(k) (1ik)4 no (Iik )2 V(k)

(7.28a)
4m 2 + mL3 '
~ (E(k) ~ (1ik)2 ~ no V(k)) (7.28b)
no V(k) 2m £3
Note that no/ L3 = p, the density, and compare these results with (6.75)-
(6.86a). As to be expected, they are the same, except that the interaction
matrix element V(k) of (7.5) has in our former treatment been simplified to
a uniform constant g, i.e. V(k) ...... g.

EXERCISE

7.1 Choice of Coefficients for the Bogoliubov Transformation

Problem. Determine the coefficients Dk of the canonical transformation

(7.25) in such a way that the Hamiltonian iII diagonalizes in the boson op-
, 't
erators A k , A k .
200 7. Superfiuidity

Exercise 7.1. Solution. We have

N2 , (lik)2
2L3 V(O) + Lk -2- a!a k
m
no '"""V(k)
+ 2L3 ~
(At At
aka_k A A + 2At
+ aka_k A)
akak (1)
k

The prime (') at the sum indicates that the terms with k = 0 have to be
omitted from the summation. We introduce new boson operators rh, A! [see
(7.18) and (7.24)]:

u(k)Ak + v(k)A~k ,
(2)
aAtk u(k)A! + v(k)A_k ,
where
1
u(k) v(k) (3)
)1- D~ , )1- D~ .
Inserting (2) into (1) gives
N2
iI = 2L3 V(O)

+ ~' C~~2 + 2~V(k)) [u 2 (k)A!Ak +v2(k)A_kA~k

+ u(k)v(k)(A!A~k + A_kAk)]

+ L' 273 V(k) [u(k)v(k)(A!Ak + AkA! + A~kA_k + A-kA~k)

k

+ u2(k)(AlA~k + AkA_ k ) + v2(k)(A_kAk + A~kAl)] . (4)

With the help of the commutation relations (7.22) and (7.19) we obtain
N2
iI = 2L3 V(O)

+ ~' [C~~2 + 2~ V(k)) v 2 (k) + 2~ V(k)U(k)V(k)]

+ ~' {C~~2 + 2~ V(k)) [u 2(k) + v 2(k)]

+ 2~ V(k)u(k)v(k) }AlAk

+ ~' [C~~2 + 2~ V(k)) u(k)v(k)

+ 2~ V(k)(u(k)2 + V(k)2)] (A!A~k + AkA_k) (5)

To bring the Hamiltonian into diagonal form, the relation

 7.1 Basics of a Microscopic Theory of Superfluidity 201

C~~2 + ~~V(k)) u(k)v(k) + ~~V(k) [u 2(k) +v 2(k)] = 0 (6) Exercise 7.1.

has to hold. With (3) this one becomes

( (fik)2 nOV(k)) ~ no V(k)I+D~ 0, (7)

2m + £3 1 - D2k + 2£3 1 - D2k
and we deduce for the coefficients Dk

Dk = ~( (fik) 2 + (fik)2 no V(k) _ (fik)2 _ no V(k)) (8)

no V(k) 2m m £3 2m £3

Hence, the Hamiltonian (5) takes on the form

Ho + 2:' E(k)AkAk (9)

k

with
N2
+ 22:
l' ( (fik)2 no ) (10)
2£3 V(O) £(k) - 2m - £3 V(k)
k

and

E(k) (11)

Equation (9) describes a system of free bosons - quasiparticles - with one-

particle energy £(k). Ho describes the energy of the system in the ground
state 14>0); for the state 14>0), free of quasiparticles, it holds that
Akl4>o) = 0, for all k i- 0 (12)
and the energy becomes
(4)oIHI4>o) (4)oIHol4>o)
(4)014>0) (4)014>0)
= Ho· (13)

The coefficients Dk can also be determined by the requirement that the

energy of the ground state 14>0) of the system, (12), takes on a minimum.
According to (5) the ground-state energy is given by

Ho =
(14)

or
N2
Ho 2£3 V(O) (15)

D~,
+ "7:
"'"" [( (fik')2 no V(k')) no V(k') Dk' ]
2m + £3 1- D~, + £3 1 - D~,
202 7. Superfluidity

Exercise 7.1. From the requirement that the variation of Ho with respect to Dk should
vanish, i.e.
8Ho
0, (16)
8Dk
the condition for Dk is obtained:

o. (17)

This is the same result as stated in (7).

The operators A1 , jh from (7.26) and (7.22) describe new boson oper-
ators. According to (7.26) the excitations of the system can be described as
single-particle excitations
't (7.29)
Akl¢o) ,
where I¢o) characterizes the ground state with
jhl¢o) = o. (7.30)
These new one-particle states (7.29) are called quasiparticles. According to
(7.20) they consist of a linear combination of creation and annihilation op-
erators a~k and ak of the customary boson operators. The latter themselves
represent particle-hole pairs, as is clearly expressed in the defining equations
(7.14). As a result of the transition to quasiparticles, an important part of
the interaction between the original boson particles (described by bl,h) has
already been taken care of, i.e. "diagonalized away". Hence, quasi particles can
be viewed as freely moving structures consisting of a superposition of ordi-
nary particles and holes. An illustrative example of a quasiparticle is given,
for example, by the motion of a charged particle in a strong electrolyte, which
consists of positive and negative charge carriers.
The particle ED, which is here positively charged, attracts negative charge
carriers around it (Debye cloud) and is thus screened (see Fig. 7.4). Out of
the originally charged particle a neutral quasiparticle has developed: A part of
+ + the Coulomb interaction with the electrolytic medium has been "transformed
away" with the Debye cloud. It is evident that a quasiparticle, evolved in this
+ + =
~ + + + --

- --+ Q +=--
+-- + way, possesses a different mass as compared to the original particle, which
in general also depends on the velocity. This is in complete analology with
-_ + \::J + - the nontrivial k dependence of E(k) from (7.28a). We refer to and already
+ --=--+ + + --==-+ recommend Exercise 9.1, where the screening potential of the Debye cloud is
+ + calculated.
The part flo is constant and independent of the quasiparticle operators
Fig. 7.4. Illustration of a
quasiparticle by a Debye
Al,Ak [see (7.27)]. Obviously, it represents the energy of the ground state of
cloud around a charged the system; we have shown this already in Exercise 7.1.
particle ED in a strong elec- The operator P of the total momentum of the system can be expressed in
trolyte terms of quasiparticle operators, too. With (7.14) and (7.18) we obtain
 7.1 Basics of a Microscopic Theory of Superfluidity 203

P Lhkblbk
k

Lhkala k
k

Lhk[u(k)Al +v(k)A_k] [u(k)Ak +v(k)A~k]

k

L hk {[u 2 (k) - v2 (k)] AlAk

k

+ u(k)v(k) (AlA~k + A-kAk) } (7.31)

The second term inside the curly bracket vanishes as it is summed over k,
because positive and negative k contributions cancel pairwise. With the nor-
malization (7.21) we get

P = LhkAlAk. (7.32)
k

Thus, it now becomes clear that the low-lying excitations of the system con-
taining Helium atoms at low temperatures are given by elementary (quasipar-
ticle) excitations of the energy Ek (which we now write also E(k) to indicate
its k dependence more clearly) and the momentum (hk). For small excitations
no ~ N, E(k) from (7.28a) can be rewritten as

Ek = E(k) =
(hk)4 + N(nk)2 V(k) . (7.33)
4m 2 mL3
For small momenta this becomes

E(k) =
~
- 3 V(O) nlkl (1 + ... ) .
The veloci yrghhe quasiparticles is given by
(7.34)

Cs
= (fJE)
fJnk k=O
= J NV(O) =
mL3
J V(O)
Pm'
(7.35)

Here P = Nj L3 denotes the particle density. It represents the velocity with

which the elementary excitations are propagating, i.e. the sound velocity.
Hence, (7.34) can also be written as
(7.36)
The sound velocity (7.35) has to be positive and real; otherwise the ground
state would decay. This leads to the condition
V(O) > 0, (7.37)
which can be transformed into

V(O) = lim
Iql-.O
J d 3 r V(lrl) e iq .r

J d 3 r V(lrl) > 0, (7.38)

204 7. Superfluidity

where (7.6) has been used. This implies that the interaction energy V(lr I) has
to be essentially positive, i.e. repulsive.
Now we will look at the quasiparticle energies E(k) from (7.33) for large
momenta. For this purpose we write (lik)2/2m in front of the square root and
arrive at
N4m
E(k) 1 + (lik)2£3 V(k)

(lik)2 ( 1 N4m )
:;::j 2m 1 + "2 (lik)2£3 V(k) + ...
(lik)2 NV(k)
:;::j 2m +-0. (7.39)

As k increases we get
lim V(k) -+ 0 (7.40)
k->oo

according to (7.6). Then the quasiparticle energy (7.39) for larger momenta
becomes

lim E(k) = (lik)2 , (7.41)

k->oo 2m
which is just the kinetic energy of a free single atom. Intuitively this is correct;
think of the quasiparticle model of the Debye cloud around the particle. For
large velocities of the EB particle we expect the Debye cloud to be stripped off
so that we are dealing more or less with a free particle.
Hence, the momentum dependence of the quasiparticle energies (7.33) in
boson systems with weak and repelling interactions has the qualitative form
illustrated in Fig. 7.5.
Let us now understand how superfluidity comes about, at least qualita-
tively. Consider a small particle (which we here call a cluster, it could be one
or several dirt atoms or a piece of the surrounding wall, etc.) moving through a
quantum liquid with energy-momentum characteristics of the type discussed
here (see Fig. 7.5), i.e. its quasiparticle energies rise linearly with momentum
at low lik and quadratically at higher lik. The cluster can lose energy only by
causing excitations in the fluid. At temperature oF 0 there are already excita-
tions in the fluid at which the cluster may scatter and thus lose energy, but

E(hk)

E "'" cshk (phonon)

Fig. 7.5. The full curve

illustrates the qualitative
behavior of the quasiparti-
cle energies in a superfluid
medium. The free particle \ p;.;sumably long
energy is indicated as the whirlpool line
dashed curve
p ""'hk
 7.2 Landau's Theory of Superfluidity 205

at zero temperature this is not the case. The initial momentum of the cluster
shall be tiq, the momentum of the excitations tik. In a scattering event of the
cluster with the fluid, energy and momentum are conserved.

tiq' + tik,
(7.42)
ti2(q')2 + E(k) .
2m
Here tiq' denotes the momentum of the cluster after the scattering. The ele-
mentary excitations of the fluid (quasiparticles) E(k) are given by (7.39). The
equations (7.42) cannot be satisfied. To see this let us follow up on energy
conservation, which reads
ti 2q2 ti 2
2m = 2m (q - k)2 + E(k) (7.43)
or
ti2 ti 2
0= -mq·k+2mk+E(k), (7.44)
and it follows for the angle 0: between q and k that
coso: = Ek/tik + tik/2m = S: + tik/2m , (7.45)
v v v v
where v = tiq/m is the initial velocity of the cluster. We learned that the
low-energy excitations (quasi particles) of the fluid are phonons, for which
Ek/tik > es . Hence for the excitation (emission) of a phonon the cluster
velocity v must be larger than es , i.e. v > es . This clearly follows from (7.45)
for k -+ 0, otherwise the angle 0: becomes imaginary. We can state that a
cluster moving with velocity v < Vcrit (where Vcrit = es in this case) cannot
lose energy to the fluid; thus there is no friction and one has superfluidity.
It does not matter, of course, in which coordinate system the analysis is
carried out. Here we have choosen the fluid at rest and the cluster moving in
the fluid. The same holds if the cluster is at rest and the fluid moves along the
cluster. Our arguments also hold if the cluster is a dislocation of the surface
surrounding the fluid (rough surface).
For liquid Helium the critical velocity is vcr it « eg • The relation between
momentum and energy is believed to appear as indicated in Fig. 7.5, where the
critical velocity is given by the slope of the dashed line. Also indicated are the
rotons and curls of high momentum; the latter are believed to be responsible
for the critical velocity.

7.2 Landau's Theory of Superfiuidity

Let us investigate here in more detail the ability of He II to flow through a
capillary without viscosity below the critical temperature To. For that purpose
we consider the Helium at T = 0; the super fluid liquid is in its ground state
and is not excited.
We examine a liquid that flows with constant velocity v through a cap-
illary. If friction were present the liquid would lose kinetic energy and the
current would slow down.
206 7. Superfluidity

It is appropriate to consider the problem in the rest frame of the liquid,

i.e. the liquid is at rest and the walls are moving with velocity -v. If viscosity
were present then the resting Helium would start to move. This motion starts
locally and gradually because of elementary excitations. Once an elementary
excitation (quasiparticle) with momentum p = hk and energy E(k) = E(p)
is present the energy Eo of the liquid (in the system for which it was initially
at rest) is equal to E(k) and its momentum is Po = p. We transform back
into the rest frame of the capillary with the help of a Galilei transformation,
and we arrive at
Mv 2
[; = Eo + Po' v + -2- and P Po + Mv, (7.46)
where M represents the mass of the liquid.
If only one excitation is present, i.e. Eo = E(p), ko = p, it follows that
(7.47)

Mv 2 /2 is the kinetic energy of the liquid and E(p) + P . v represents the

change of energy due to the elementary excitation of the quasiparticle. Since
for the case of an intrinsic friction (viscosity) the energy of the fluid, which is
moving with velocity v, should decrease, it has to hold that E(p) + p. v < O.
If p is given, E(p) + P . v becomes minimal once p and v are anti parallel, i.e.
E(p) -Ipllvl < O! Thus,
E(k)
v > -k-' nk = p. (7.48)

Only if this condition is fulfilled can elementary excitations take place, i.e.
quasiparticles can be excited, and the liquid slows down. Then it possesses
viscosity because energy is lost for the quasiparticle excitations.
Condition (7.48) has to be fulfilled at least for some values of p. As already
outlined before, this implies that elementary excitations can occur only once
v is larger than the minimum of E(p)/p. We are thus led to a critical velocity:

v > (E(P))
p
. mm
= Vcrit (7.49)

for quasiparticles to be excited. If v < Vcrit then condition (7.48) cannot be

fulfilled: no quasiparticles can be excited and no energy is lost for intrinsic
degrees of freedom. Then the liquid is superfiuid.
The minimum value of E(p)/p corresponds to the point on the curve
E(p)(p) for which
dE(p) E(p)
-- -- (7.50)
dp p
Figure 7.6 sketches the situation.
Geometrically this corresponds to the point for which a straight line
through the origin is at the same time a tangent to E(p) [see (7.34)!]. For
all other values of p, which are not solutions of (7.50), E(p)/p is larger than
the minimal value described by (7.50). Superfiuidity can occur only if the ve-
locity of the liquid is smaller than the velocity of an elementary excitation;
 7.2 Landau's Theory of Superfl.uidity 207

Fig. 7.6. The quasiparti-

cle energy in K (kB rep-
resents Boltzmann's con-
20
Ll + (p - Po )2 stant) and its expansion
,,, / 2p, around the minimum

3 4
Po

otherwise superfluidity would be destroyed because the fluid would lose en-
ergy as a result of the excitation of quasiparticles. dE / dp is the velocity of an
elementary excitation. For every Bose liquid this required condition is fulfilled
for one point, i.e. for p = O. Near p = 0 the elementary excitations move with
the sound velocity. These are the phonons with E(p) = csp! The condition for
superfluidity is not fulfilled for sure once the velocity of the liquid is v > cs .
For the case of Helium, however, more than one point exists that fulfills this
condition. For small p the experimental spectrum E(p) (see Fig. 7.6) shows
the expected behavior for phonons, i.e. E(p) = CsP = csnk. However, another
minimum exists at Po. In thermal equilibrium most of the elementary exci-
tations have energies very close to the minimum of E(p), i.e. for p = 0 and
p =Po!
In such a case, we may expand around Po (approximation with a parabola)

,
and obtain

E(p) = kB.1 + (p ~:o? (7.51)

where .1 = E(po) and 1-£ are constants characterizing the minimum. Quasi-
particles that obey this dispersion relation are called Tatons. For 4He these
parameters are

L1 To;::, 8.5K,
ko = Po/n (7.52)

1-£ 0.16m(4He) .
The second point [i.e. second solution of (7.50)] we were looking for lies right
of the minimum for rotons at Po as can be identified from Fig. 7.6. With (7.51)
we calculate the critical velocity Vcrit from (7.49). It follows from condition
(7.50), which now reads
p - Po kBL1 + (p - po)2/21-£
1-£ L1
It has the solution

p = V2I-£kBL1 + P5
and, hence,
208 7. Superfluidity

vcrit = ~ ( VP5 + 2J-lkBL1- po)

Thus, superfluidity exists only if

v< Vcrit = ~ ( VP5 + 2J-lkBL1- po) (7.53)

or, once we consider the experimental values (7.52) for 4He (they show that
P5 » 2J-lL1),
kBL1
v < --.
Po
We can express this in another way: There is no superfluid flow in 4He if the
flow velocity is v > k BL1/po = kBTo/Po! Remember that L1 = To is measured
as a temperature - see (7.52). Consider now the Bose liquid for T « To but
T =f O. In this case excitations may be present. Think of a "quasiparticle gas"
that moves with v relative to the liquid. This quasiparticle gas can be at rest
relative to the wall, the latter moving with -v with respect to the liquid: it
crawls, so to speak, out of the vessel walls because of the temperature, which
causes the quasiparticle excitations. The distribution function of the gas at
rest is denoted as p(E); p(E - p. v) represents the distribution of the moving
gas. Hence, the total momentum of the system is given by

p = (2~)3 J d 3 ppp(E-p·v). (7.54)

For small velocities we expand with respect to p. v. The first term is linear
in p. Because of the isotropic orientation of p, this term vanishes with an
integration over all directions. The second term is

p = -(211")3
1 J 3
d PP(P,v)fJE'
fJp (7.55)

The integration over angles produces a factor 411"/3 and we arrive at

p = -~411" t
dpp 4dp (E) . JO
(7.56)
3 dE 10
For phonons the relation E = CsP holds. Via partial integration we obtain

p= -v 411" roo dp p4 dp(p)

10
1
3cs dp
1611"
v-
00

3C-s 0
dpp 3 p(p) . (7.57)

The integral

411" 1 00
dpcs pp(p)p2 = Jd 3 pE p(E)

gives the energy Eph of a unit volume of the phonon gas. We arrive at

P = v4Eph
-- = vm*. (7.58 )
3c~
 7.2 Landau's Theory of Superfiuidity 209

Obviously we can interpret this in the following way. The phonon gas has
an effective mass m *. Like all the liquid, this part of it also represents mass
transport. The excitations are allowed to collide with the walls and to ex-
change momentum; this leads to viscosity. Evidently, viscous flow can occur
in a Bose gas that contains elementary excitations and does not violate the
condition for the existence of superfluidity. The total transported mass m* is
not identical to the total mass of the liquid. We obtain the following picture.
For T = 0 no excitations exist, i.e. the motion is superfluid. For T 1= 0 ele-
mentary excitations exist in the system, which cause a part of the gas (mass
m*) to be viscous and converts it into a "normal" liquid. A superfluid rest
remains, so that

P = Pn + Ps·
This is the two-fluid picture we mentioned earlier.
The portions Pn and Ps depend on temperature: For T = 0 we have Pn = 0
and Ps = P, for 0 < T < To we expect Ps 1= 0 and Pn 1= 0, and for T > To
we expect Ps = 0 and Pn = P (compare again with Fig. 7.2). Notice that two
different kinds of elementary excitations "contribute" to the normal fluid:
phonons with E ~ p and rotons with E ~ p2. Their relevance depends on
temperature. For T ;::;:; 0.6 K the contributions from phonons and rotons are
comparable whereas for larger temperatures the rotons dominate. However,
except for the dispersion relation, no difference exists between phonons and
rotons.

EXERCISE

7.2 An Analogy to Superfluidity in Hydrodynamics

Problem. Explain the following Gedanken experiment. A basin is filled with

water. If a thin object, such as a rasor-blade, is moved through the water, a
laminar current is observed for small velocities of the object. Once the velocity
is increased above a critical value (approximately 23 cm/ s) the current changes
its character.

Solution. The critical velocity Vcrit (;::;:; 23 cm/ s) represents the velocity of
surface waves on the water. If v > Vcrit the object produces waves in such a
way that the latter run away from the object, interfere according to Huygen's
principle, and form a cone (with a fixed opening angle;::;:; 20°). In contrast
to the Mach cone phenomenon appearing at supersonic flight velocities, this
angle does not depend on the shape of the object, the velocity or any other
exterior factor. It is a consequence only of the dispersion relation for water
waves. In the case of deep water, where the depth of the water is much greater
than the wavelength>. of the surface wave, the dispersion relation of the wave
reads
9>' (1)
2n
210 7. Superfluidity

Exercise 7.2. c denotes the speed of propagation and 9 = 9.8 m/s 2, the gravitational accel-
eration. Thus, the angular velocity w = 27fc/ A is
w = g/c. (2)
Now, consider a radially symmetric surface wave centered around r = 0,

sin C~r -wt) , (3)

i.e. the phase ¢ at radial distance r and time t is

¢(r, t) =
27fr
T - wt 27fr
= T - t T'
)27f9 (4)

The characteristic wavelength is obtained from the condition d¢/dA = 0,

87fr 2
Ach = -2-' (5)
gt
In the case of an object moving with velocity v, waves centered around each
point ofits trajectory superimpose (cf. Fig. 7.7).The characteristic wavelength
of the waves centered around X2 is
87fr,2
Ach = gt 2 .
(6)

Fig. 7.7. If the object

moves faster than Vcrit, a
cone with fixed angle, in-
dependent of the object's
velocity, emerges

Using (5) and (6) we find

gt 2
¢ = - 4r' . (7)

Employing r' = Vr2 + v2t2 - 2rvt cos 1j;, the condition d¢/dt = 0 (stationary
phase) leads to

t = ~;(cos1j;±Vcos21j;-8/9). (8)

For cos 2 1j; < 8/9, i.e. 11j;1 > 19.5°, no real solution exists. In this case the phase
has no stationary point, i.e. all phases contribute equally, and destructive in-
terference leads to a vanishing amplitude outside this cone. If we consider this
problem quantum mechanically, we could say that the moving object emitts
quasiparticles (we could call them "hydrons" f and in this way could obtain
7 J.L. Synge: Science 138 (1962) 13.
 7.2 Landau's Theory of Superfluidity 211

momentum and energy as far as v > vcrit. We have the same situation as in the v < Vcrit

case of superfluidity: The laminar current corresponds to the superfluid state;

no excitations do occur! The current cone determined by the quasiparticles
corresponds to the normal, nonsuperfluid state.
Fig. 7.8. Laminar current
of a fluid around a thin ob-
ject. If the object moves
faster than Vcrit "hydrons"
are emitted and a Mach-
cone-type wave emerges
with well-defined cone an-
gle, independent of the ob-
ject's velocity
8. Pair Correlations Among Fermions
and Bosons

We will make use of the formalism of second quantization developed so far

in order to deepen our insights into the properties of quantum-mechanical
many-body systems.
So-called correlations coin very fundamental and characteristic many-body
effects. In general, correlations of a many-particle system (bosons or fermions)
describe the influence of the quantum numbers of a single particle on all
remaining particles of the system and their quantum numbers. We accentuate
in advance: Correlations are not only produced by special interactions (forces)
between particles but are also a consequence of the corresponding fundamental
statistics of fermions and bosons (commutation relations between fermionic
and bosonic operators). In the following, we will concentrate on correlations
that result from pure exchange effects.
The lowest correlations that can be studied are the so-called pair corre-
lations. In the following, we will focus on these correlations for fermions and
bosons.

8.1 Pair-Correlation Function for Fermions

We consider a gas of N free, noninteracting fermions with spin 1/2 in a volume

V. The ground state Ipo) of this N-particle system is characterized such that
all states with spin projection s = ±1/2 and momentum p are occupied up
to the Fermi momentum PF, i.e.

ns(p) = (¢olbl(p)bs(p)l¢o) = {01 lor

~or liP I ::::: PF ,
p > PF.
(8.1)

The validity of this relation becomes instantly clear once the representation
PF

I¢o) = II b!(p)IO) (8.2)

s,P

of the ground state is inserted. The Fermi momentum is fixed by the condition
that the average particle-number density n of the system is constant:
2
VLl. (8.3)
s,P Ipl~PF

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
214 8. Pair Correlations Among Fermions and Bosons

In the thermodynamic limit (limN->CXJ' limv->CXJ' NjV = const) this leads to

the relation

n = 2
j PF d3p
(27r)3 =
p~
37r 2 . (8.4)

Before we concentrate directly on the two-particle correlation function in a

Fermi gas, we first calculate the expectation value of the one-particle density
matrix operator and of the one-particle density operator.
The operator of the one-particle density matrix is defined via field opera-
tors:
(8.5a)
Its form is motivated by the well-known expression 1jJ*(r)1jJ(r) for the proba-
bility density. It fulfills the symmetric relation
A (r, r ') = Ps
Ps At( r I ,r ) . (8.5b)
The one-particle density operator for particles with spin projection s is the
diagonal element (r = r') of the one-particle density matrix:
(8.6a)
Hence, the total density operator is given by

p(r) = LPs(r). (8.6b)

In order to evaluate the ground-state expectation value of ps(r, r') we employ

the expansion of the field operators into free momentum states:

¢s(r) j (27r)3
d3p
e
ip.r jj ( ).
s P , (8.7)

we obtain
ps(r, r') (<I>olps(r, r')I<I>o)
j (27r)3 (27rp
d 3P d3 PI
e
-ip.r ip' .r'
e
At A I
(<I>olbs(p)bs(p )1<1>0) . (8.8)

The matrix element appearing in (8.8) is nonvanishing only if p = pi, i.e.

(8.9)
and if momenta p lie inside the Fermi sphere. This yields

ps(r, r') =

(8.10)

Here, the momentum integration has been performed in spherical coor-

dinates and the substitution ~ = PFlr - r'l and (8.4) have been used.
 8.1 Pair-Correlation Function for Fermions 215

j1 denotes a spherical Bessel function. It is related to the regular Bessel

functions In+l(X)
2
via the relation jn(x) = V7r/2XJn+l(X).
2
In particular,
j1(X) = (sinx/x 2 ) - (cos X/X). The one-particle density follows directly from
the one-particle density matrix (8.10):

(¢olp(r)l¢o) = 2 lim 3n j1~~) = n. (8.11)

1;--->0 2 <"

After these preliminary considerations we now calculate the pair-correlation

function for a free Fermi gas.
We are well aware of the Pauli principle for fermions from elementary
quantum mechanics. As an immediate consequence particles with an identical
spin projection tend to avoid each other. To be more precise: Two particles
with the same spin quantum number are not allowed at the same space point.
We will now inquire the probability amplitude of finding a particle with spin
projection s at place r inside a Fermi gas (s = 1/2) given another particle
with s' at r'. That part IIss' of this probability amplitude normalized to
unity and independent of the average particle-number density is denoted as
the pair-correlation function. For a free and homogenous Fermi gas it is exclu-
sively a function of the relative distance between the two particles. Hence, the
following properties of II,s' become plausible: Particles with opposite spin,
i.e. s of s', are allowed to approach each other closely; the correlation function
has to be independent of the relative distance, i.e.
for s of s' . (8.12a)
Such particles are uncorrelated. On the other hand, for particles with identical
spin it has to hold that

II '(Ir-rl) =
, {O for Ir - r'l --+ 0
(8.12b)
ss 1 lor
C Ir - r 'I --+ 00 ' s=s'.
This represents the so-called Pauli blocking for small distances; for large dis-
tances the correlations vanish, too. We will now determine the probability
amplitude of finding a particle at place r' given another one at place r at
the same time. A mathematical formulation of this problem is provided in the
following way: With the help of the field operators -0s(r) and -0,,(r') a particle
with spin s at r and another one with spin s' at r' are annihilated from the
(N-particle) ground state I¢o). The amplitude of transition into an (N-2)-
particle state 1¢;;-2) is given via (¢;;-21-0s,(r')-0s(r)I¢0). After a summation
over all possible final states we arrive at the total probability amplitude for
the annihilation of two particles:

(¢o 1-01 (r)-0!, (r')-0s' (r')-0s (r) I¢o) (~) 2 IIss' (Ir - r'l) ,
(8.13a)
216 8. Pair Correlations Among Fermions and Bosons

where use has been made of the completeness relation for two-particle states.
From the form of the probability amplitude (8.13a) it becomes obvious that
an equivalent formulation of the problem is provided by first annihilating a
particle with spin 8 at position r and then calculating the expectation value
with the remaining (N-1)-particle state I4>N-1) = -0.(r) 14>0) to find a particle
with spin 8' at position r':
(¢F -11-0!, (r')-0s' (r') I4>N -1)
= (4)01-0! (r)-0!,(r')-0s' (r')-0s(r)l4>o) . (8.13b)
For the further manipulation of the amplitude (8.13b) we insert an expansion
into momentum states (8.7) for each field operator and obtain

(~) 2 lIss'(lr _ r'l)

J d3p d 3q d 3p' d 3q' e-i(p-p').r e-i(q-q').r'
(2Jr)3 (2Jr)3 (2Jr)3 (2Jr)3
X (4)0IhUp)h!,(q)h s ,(q')h8 (p')14>0). (8.14)
In order to obtain nonvanishing values for the appearing matrix element, the
momentum integrations should run over only the Fermi sphere. Furthermore,
we use the commutation relations for the operators hl(p) and h8 ,(p'), which
follow directly from the corresponding relations for the field operators pre-
sented earlier:

[h 8 (p),h 8 '(P')L = 0,
(8.15)
(2Jr)3 888 , 8(p - p') .

The factor (2Jr)3 on the right-hand side of the last equation is a normaliza-
tion factor which corresponds to the dp3/ (2Jr)3 volume element in momentum
space. Indeed, taking the vacuum expectation value and integrating over mo-
menta yields

(8.16)

(8.17)

(8.18)

(8.19)

(8.20)

(8.21)

(8.22)
 8.1 Pair-Correlation Function for Fermions 217

We transform further:
(¢o Ih! (p )h!, (q )h8' (q')b 8(p') I¢o)
-(27r)3588 ,5(p' - q)(¢0Ih!(p)h 8,(q')1¢0)
+ (¢olh! (p )h8 (p')h!, (q)h8' (q') I¢o)
(27r)65(p - p')5(q - q')
x (¢olb! (p )b8(p') I¢o) (¢olh!, (q)b 8, (q') I¢o)
- (27r)65 88 ,5(p' - q)5(q' - p) (¢olh!(p)bs(p)l¢o) .
Inside the integration domain the remaining matrix elements are all equal to
unity. Inserting (S.14) provides
2 jPF d 3p jqF d3q
(
~) 1I88 ,(lr-r'l) = (27r)3 1 (27r)3 1

- 588' jPF (~:~3 e-ip.(r-r') jqF (~:~3 e-iq.(r-r') .

Taking into account (S.3) and (S.10), we deduce that

(~f lIss'(lr - r'l) = (~)2 5ss,[ps(r,r'W (S.23)

and we find the pair-correlation function

1I88 '(lr - r'l) = 1 - 588' (3jl(PFlr - r'I))2 (S.24)

PFlr - r'l
_ _ , {,3~[s_in~(p~F~lr_-
__ r~'I)~-~p~F~lr_-_r~'7Ic_o_s~(p~F~lr_-
__r~'I~)l~}2
- 1 U 88'-
(PFlr - r'I)6
This correlation function possesses the required properties discussed before;
it is sketched in Fig. S.l.

IIss
1

Fig. 8.1. Pair-correlation

function of noninteract-
ing spin-1/2 particles with
identical spin projection
I
rr/4pF Ir-r'l

We realize that the probability of finding two fermions with parallel spins
for relative distances smaller than the inverse Fermi momentum is drastically
reduced. The Pauli principle induces long-range correlations for the motion of
fermions with equal spin; they imply that these fermions cannot approach each
other too closely. This effective repulsion results explicitly from the exchange
symmetry of the many-particle wavefunction and not from the real, external
interactions. This observation is formally expressed in the following rewriting
of the amplitude (S.13b):
218 8. Pair Correlations Among Fermions and Bosons

(¢o I~! (r )~!, (r')~s' (r')~s( r) I¢o)

(¢ol~! (r )~s( r) I¢o) (¢ol~!, (r')~s' (r') I¢o)
- (¢o I~! (r )~s' (r') I¢o)(¢ol~!, (r')~s (r) I¢o)
(¢olps(r) I¢o) (¢o Ips' (r')I¢o)
-(¢olps(r, r') I¢o) (¢o Ips( r, r')I¢o)oss'
(i)2 - oss,(ps(r,r'))2. (8.25)

The first term represents the direct term, whereas the second one stands for
the exchange term. The relative minus sign reflects the antisymmetry.

8.2 Pair-Correlation Function for Bosons

Here we will calculate the pair-correlation function for a free, noninteracting

Bose gas. We will come back to the mathematical formulation of the correla-
tion function as presented in (8.13a); this equation is also valid for a system
of N bosons. However, changes do occur because bosons do not have to obey
Pauli's principle, Le. arbitrarily many bosons are allowed to occupy the same
momentum state.
We consider an N-particle state of the form

I¢) = ~ IT [at(p;)r(p;) 10), (8.26)

Pi

where N represents a normalization factor. Furthermore,

at(p)I¢) In(p) + 11¢[n(p)+11) ,
a(p)I¢) = In(p) I¢[n(p)-l)) .
The particle-number density is given via

(¢10 t (r)0(r)I¢) = J(~:~3 n(p) = n. (8.27)

Again, we introduce an expansion into free momentum states for the field
operators, this time for the boson operators:

'( )- J
'P r -
3
d p e ip·r'a ( p ) .
(27rp (8.28)

An analogous calculation of the expectation value

(¢10 t (r )0 t (r')0( r')0( r) I¢)
leads to the matrix element

The latter is only nonvanishing if p = p' and q = q', or if p = q' and q = p'.
Both cases are identical once p = q. Hence, we can write the following:
 8.2 Pair-Correlation Function for Bosons 219

(¢Ia t (p)a t (q )a( q')a(p') IqI)

[1 - (21l')35(p - q)](21l')66(p - p')6(q - q')
X (qlla t (p)a t (q)a(q)a(p)lqI)

+ [1 - (21l')36(p - q)](2'71y5(p - q')6(q - p')

x (qllat(p)at(q)a(p)a(q)lqI)
+ (21l')96(p - q)6(p - p')6(q - q')
x (¢Ia t (p)a t (p)a(p)a(p) IqI) . (8.29)
We determine the appearing matrix elements one after the other:
(qlla t (p)a t (q)a( q)a(p) I¢)
= n(p) (qI[n(p)-lj lat (q)a( q) I¢[n(p)-lj)
= n(p) n(q) ,
(qllat(p)at(q)a(p)a(q)lqI) = n(p)n(q),
(qlla t (p)a t (p )a(p )a(p) IqI)
= n(p) (qI[n(p)-lj la t (p )a(p) I¢[n(p)-l))
= n(p) (n(p) - 1) .
Thus, we obtain for (8.29)
(qlla t (p)a t (q )a( q')a(p') IqI)
= (21l')6[6(p - p')6(q - q') + 6(p - q')6(q - p')]n(p) n(q)
- (21l')96(p - q)6(p - p')6(q - q') n(p) (n(p) + 1). (8.30)
We deduce the probability amplitude of finding a particle at position rand
another one at r':

(8.31 )

n2 + If d3p
(21l')3
eip.(r-r') n(p)1
2
- f d3p n(p) (n(p)
(21l')3
+ 1) .
Let's inspect this result closer. In two ways it is different from the result
obtained for fermions. First, the sign of the second term is positive which is a
direct consequence of the exchange symmetry of bosons. Second, an additional
term shows up which is a result of the fact that many bosons are allowed
to occupy the same state. Furthermore we notice that this latter term is
independent of the positions and, according to dimensions, is smaller than
the first two terms by an order l/V; this will be rectified in Exercise 8.2.
For a homogenous Bose gas that is extended to infinity, the pair correlation
function II (r - r') is defined via
220 8. Pair Correlations Among Fermions and Bosons

(q'Jlcpt (r)cp t (r')cp( r')cp( r) 1q'J) n 2 II(r - r') (8.32)

n2 + If d 3p eip.(r-r') n(p)1
2
(27r)3
The first term appearing in this equation emerges from the second one for the
case r = r'. In other words: the exchange symmetry causes bosons to cluster.
According to (8.32) the probability for finding two particles in the same space
point at the same time doubles. 1 The following example provides a further
illustration of this point.

EXAMPLE

8.1 Pair-Correlation Function for a Beam of Bosons

We calculate the pair-correlation function for a particle beam of non-

interacting bosons, a photon beam for example. The average particle-number
density n(p) is characterized by a Gaussian momentum distribution
n(p) = ae-{3(p-po)2/ 2 , (1)
centered around momentum Po. At first we determine the Fourier integral of
the second term in (8.32) with (R = r - r'):

(2)

We substitute k = p - Po and calculate the k integral in polar coordinates;

this yields:

J(R) =

(3)
with
F(R) = LX> dk ke-{3k /2 sin (kR) .
2
(4)

The remaining parameter-dependent integral can be found, for example, in

the book by Gradshteyn and Ryzhik. 2 However, we prefer to determine it
explicitly with a favorite standard trick: The determination of the integral is
reduced to the solution of a simple, ordinary differential equation.
The integral F(R) can be expressed as a differentiation of another integral
with respect to the same parameter:
d
F(R) = - dRH(R) , (5)

1 Such effects are of some importance in e.g. high-energy heavy-ion collisions, where
in one encounter up to 10000 pions (bosons!) are produced.
2 1.M. Ryzhik, A. Seffrey, 1.S. Gradshteyn: Table of Integrals, Series, and Products,
5th ed. (Academic Press, New York 1994).
 8.2 Pair-Correlation Function for Bosons 221

On the other hand, a partial integration of (4) yields Example B.l.

F(R) = ~H(R). (6)

Hence, we arrive at an ordinary differential equation for the integral H(R):

dH R
dR = -jjH. (7)

With the initial condition

H(R = 0) = rooo dk e-{3k /2

V~ (8)
2

in = 2{3
we find the solution via a separation of variables:
rH(R)

iH(O)
dH'
H'
_~
(3
r dR' R' '
io
R

H(R) ~ -R 2/2{3
= V (9)
2{33 e .

An insertion into (6) yields F(R) and according to (3) we get the result for
the wanted Fourier integral:

J(R) _ ae ipo · R _R2/2{3

(10)
- (2rr{3)3/2 e
If we take the square of the modulus, the phase drops out and we get
2
IJ(R)12 = _ a _ e-R2/{3 (11)
8rr3{33
This was the method used for the second term of (8.32) in the case of the
photon beam. The first term, i.e. n 2 , immediately follows once we take the
limit R = 0 for the second term:
2 a2
n = 8rr3{33' (12)

In total we obtain
n 2 II(r - r') = n 2 (1 + e- lr - r '1 2 /(3) . (13)
A qualitative sketch of the pair-correlation function is illustrated in Fig. 8.2.

n
2

--- ~=t:------- Fig. 8.2. Pair-correlation

function of noninteracting
o~ _____________ ~

bosons
Ir- r'l
222 8. Pair Correlations Among Fermions and Bosons

Example S.l. Obviously, a significant increase can be observed for the probability of
finding two bosons at small relative distances at the same time. In particular,
the probability of finding a pair at the same position is twice as large as the
asymptotic value for large relative distances.

EXERCISE

8.2 Boson Pair-Correlation Function

as a Function of the Quantization Volume

Problem. For the derivation of (8.31) we have assumed, without mentioning

it, that the quantization volume is equal to the unit volume V = 1. Show that
the last term in (8.31) is of the order 1jV with respect to the first two terms.

Solution. We indicate again the derivation of (8.31), but this time we use an
expansion of the field operators into discrete momentum states, i.e.

L evV
ip·r
<j;( r) = ITT o'k . (1)
p

Then, we obtain for the matrix element in (8.31):

(¢I<j;t (r )<j;t (r')<j;( r')<j;( r) I¢)
1
V2 'L...
" e i(p-p')r e i(q-q')·r' ("'I At' t, , I"')
'+' apaq aq,ap' '+' . (2)
p,q,p',q'

The treatment of the matrix element appearing in (2) is straightforward and

completly analogous to the previous case. Instead of 0 functions for continuous
momenta we have to deal with the corresponding Kronecker symbols:
(¢IO,tO,~ o'q'o'p' I¢) = (opp,Oqq' + Opq,Oqp') np nq
- Opq Opp' Oqq' np (np + 1) . (3)
Inserting (3) into (2) and carrying out the summations yields
(¢I<j;t (r)<j; t (r')<j;( r')<j;( r) I¢)
~ '" n
V2 L... p
n
q
+~ '" e-ip.(r-r') eiq.(r-r') n p n q
V2 L...
p,q p,q
1
- V2 L np (np + 1)
p

n2 + 1~
V 'L...
" e-ip.(r-r')nP 12 - ~
V2 'L...
" n
p
(n
p
+ 1) . (4)
p p

Once the continuum limit is performed, i.e. replacement of 2:p with

V J d3pj(27r)3 etc., we realize that the third term in (4) is of the order IjV
with respect to the first term.
 8.3 The Hanbury-Brown and Twiss Effect 223

8.3 The Hanbury-Brown and Twiss Effect

Hanbury-Brown and Twiss 3 have carried out an experiment to measure the

pair-correlation function of noninteracting bosons. Here we will discuss only
the basic setup of the experiment; for most of the theoretical details we refer
to the original publications.
In this experiment, the probability amplitude is measured to detect two
photons in different space points at neighboring time; the two photons belong
to a coherent photon beam which is produced by a source Q. At first, the
primary beam is divided into two identical coherent beams via a semiper-
meable mirror S. In this way we avoid placing two detectors one behind the
other. Coming from the mirror, the two partial beams hit the two detectors
D1 and D2, which are at different distances from the mirror. In Fig. 8.3 the
experimental setup is sketched schematically.

Fig. 8.3. Schematic setup

of the Hanbury-Brown and
Twiss experiment

The light intensities, which are a measure for the average photon numbers,
are measured: J 1(t) in detector D1 at time t and J 2(t + T) in detector D2 at a
later time t + T. The difference in time T defines the relative distance r = cT
between two points in the primary beam. The product of the two intensities is
averaged with respect to t over the observation window T. This measurement
prescription is equivalent to the measurement of two photons at two different
points of the beam with a relative distance cT. For fixed T the observed,
correlated average value

shows the same characteristic trend as a function of T as has been derived for
the pair-correlation function II(r) in the previous example.
This experiment appears to reveal a typical quantum effect in Bose gases.
It is quite remarkable that it can also be explained completely within the
tools of classical electrodynamics. This should not be too surprising; we re-
member that a light beam (macroscopic electromagnetic field) contains very
many photons and, thus, also contains many photons with the same momen-
tum. An analysis of correlation functions in terms of Glauber states, which
we have briefly introduced in Chap. 1 in order to calculate, for example, the
classical value of the E field, would lead to classical quantities such as the

3 R. Hanbury-Brown, R.Q. Twiss: Nature, 177 (1956) 27; 178 (1956) 1447.
224 8. Pair Correlations Among Fermions and Bosons

intensity. As squares of electric and magnetic field amplitudes they represent

a measure for the number of photons. Hence, the Hanbury-Brown and Twiss
experiment teaches us that the bosonic character of photons is already con-
tained in the superposition principle for classical electromagnetic fields. The
following oversimplified consideration shall illucidate this aspect. More formal
and more detailed comments can be found, for example, in the textbook by
R.G. Newton. 4

Q
Fig. 8.4. Oversimplified
setup of the Hanbury- D2
Brown and Twiss experi- r'2
ment

For simplicity we assume that the source of the light beam consists of two
closely adjacent emitters Ql and Q2; see Fig. 8.4. Furthermore, Ql is required
to emit coherent light with a wave-number vector k1 and an amplitude a1; Q2
emits coherent light with k2 and a2. The light of the two emitters comes with
the same polarization; the relative phases of the amplitudes vary randomly
between 0 and 1r. From Ql, light with the amplitude a1 e ik1 ·rl reaches the
detector D 1; here r 1 represents the distance vector from Ql to D 1. In the
same manner the light originating from Q2 and falling into the detector D1
positioned at r~ has the amplitude a2 eik2·r~. According to the superposition
principle the total field amplitude reaching D1 is given by
(8.33a)
where we did not write down the polarization vector. The intensity becomes

J1 = IAl12 = lalI2+la212+R(aia2ei(k2.r~-kl.rtl). (8.33b)

Analogously, for the total amplitude and intensity at D2 we find:

(8.34a)

(8.34b)

R denotes the real part of the corresponding quantity. As stated earlier the
relative phases of al and a2 can vary randomly. A more suited measure
for the corresponding total intensities is given by an average over the rela-
tive phases of the partial amplitudes; this average is equivalent to the time
average performed in the Hanbury-Brown and Twiss experiment. Suppose
a1,2 = Q1,2 ei!h.2, then the averaging implies

4 R.G. Newton: Scattering Theory of Waves and Particles (Springer, Berlin, Hei-
delberg 1982).
 8.3 The Hanbury-Brown and Twiss Effect 225

1
(aiaz) = aiaz-
27r
l 0
Z7J"
d~e-l€ = 0,
•
(8.35)

As a consequence, terms containing products alaZ drop out after the average
is taken. Hence, the averaged intensities become
(8.36)
The product of the averaged intensities (h)(Jz ) is independent ofthe distance
between the detectors. This does not hold for the average of the product of
intensities; we obtain for the product hJz
h Jz = IA1Azlz = lar eik,.(r l +r2) + a~ eikdr; +r;)
(8.37)
Multiplying out and averaging over the relative phases eliminates terms of the
form alazlall z , alazlazl z , ... , and we obtain
+ lazl 4 + lallzlazlzlek,.r, eik2·r; + ek2·r; eikl.r212
lall 4
(J1)(JZ) + 21 a ll z lazl z
x cos [k2 . (T~ - T~) - kl . (Tl - T2)] . (8.38)
If both emitters Ql and Q2 are very close together, which should be the case
for a well-collimated light beam, we can replace Tl - TZ with T~ - T~ in the
above calculation.
Hence, we deduce for the correlated intensity as a function of the relative
distance between the detectors:
(8.39)
with p = k z - k 1. This function takes on a maximum for Tl = TZ. In a realistic
light beam, p has a distribution; a Gaussian, for example. Thus, an additional
averaging over different p yields again a function of the form (13) of Example
8.l.
The quantum-mechanical interpretation of the separate terms in (8.37)
would be as follows: The ai and the a~ terms represent the probability ampli-
tudes of observing a photon pair from Ql and Qz, respectively. They lead to
the position-independent terms of (8.39). The mixed term ala2 represents the
probability amplitude of measuring a photon from Ql and another one from
Q2' The two probabilities of registering one photon from Ql at Tl and the
other one from Q2 at T2 or vice versa cannot be distinguished. The interfer-
ence of the two amplitudes leads to the cosine term in (8.39). Therefore, the
tendency of photons to clump together can be understood as a consequence
of the superposition principle.
The HBT effect has been used in astrophysics for the analysis of the spacial
extension of cosmic light sources (stars, galaxies, and so on). It has also become
of importance for pions emitted from hot compressed nuclear matter regions
as they occur in high-energy heavy-ion encounters where nuclear shock waves
226 8. Pair Correlations Among Fermions and Bosons

cause the heating and compression of the nuclear matter. It serves there as a
tool to measure the spatial extend of the source. 5

8.4 Cooper Pairs

A key ingredient for our understanding of superconductivity is the pairing of

electrons during their motion in a crystal. At first, this is not to be expected
because two equally charged particles repel each other, so pairing implies that
they attract. Indeed, under certain conditions an attraction is accomplished by
the interaction of the electrons with the lattice. With a simple but instructive
model we want to find out the interaction mechanism leading to pairing.
For the moment we forget about all the structure effects in a crystal and
consider a metal with volume Vasa potential box (cube) filled with electrons.
Furthermore, the interaction between the electrons is neglected for the time
being, so that the normalized electron eigenfunctions with periodic boundary
conditions are given by
I
1f;n(r) = JV 'k
e' "T ; (S.40)

see Fig. 1.1. The electrons possess the energy

hk 2
Ek
n
----.!!.
- 2m (S.4la)

The possible k vectors inside a box of length L are [see (1.30)]

27r 27r
k n = Ln = L{n 1 ,n2 ,n3 } , nv E Z. (S.4lb)

nv are integers with -00 ::; nv ::; +00. Every single state (8.40) can be filled
with two electrons (spin degeneracy); as a consequence the state with the
lowest energy for the total system is characterized by

N= 2. (S.42)
nwithk,,<k F

N represents the total number of electrons within the box and kF is known
as the Fermi momentum. From (S.4la) the Fermi momentum is related to the
Fermi energy:

EF
h2 kF2
= __
. (8.43)
Fig. 8.5. The Fermi sea. 2m
All states with Ikl ::; kF In k space all states within the sphere of radius IkFI, i.e. Ikl ::;; IkFI, are
are occupied with elec- occupied; see Fig. 8.5.
trons. Two noninteracting
electrons with momenta These states constitute the Fermi sea. For the case that Nand L become
tikI and tik2 outside of the very large, the vectors k n are very close together according to (8.4lb) and the
Fermi sphere are indicated summation over n in (8.42) can be replaced:

5 D. Ardouin: Int. J. Mod. Phys. E (1997).

 8.4 Cooper Pairs 227

Ln, ---->
2~ Jdk x or

Ln ----> L3
(2'llV
J V
d 3k = (27r)3
J d 3k. (8.44)

Here, the normalization volume V = L3 has been introduced. The quantity

V
(8.45)
(27r )3
can be interpreted as the level density in k space. From (8.42) we get
V 47r 3
N = 2 (27r)3 3kF , (8.46)

from which

kF = (37r N)
V 2
1/3
(8.47)

follows. N IV = n represents the electron density.

Now we consider two electrons that are just outside the Fermi sphere (see
again Fig. 8.5) and interact with each other via a weak attractive force. The
interaction between the electrons inside the Fermi sphere and their interaction
with the electron pair outside the Fermi sphere are neglected for the time
being; however, for a realistic model of superconductivity these interactions
have to be taken into account. The two electrons outside the Fermi sphere are
assumed to have opposite spin: in accordance with the Pauli principle they
can simultanously occupy the same spatial wavefunction, Without interaction
the energy eigenfunction of the electron pair reads
eik,.r, eik2·r2 .
t) = _ _ _ _ e-(·/Ii)(c:.'+c: k 2 )t
ol'(T
'P 1,
T
2, JV JV (8.48)

The interaction leads to the electrons scattering from each other; as a conse-
quence (8.48) no longer represents an energy eigenstate, We have to take a
superposition of unperturbed (free) states (8.48) of the form

(8.49)

with
ak,k2(t) = e-iEt/liak,k2' (8.50)
Here E stands for the total energy of the electron pair. The correlated wave-
function (8.49) is a double Fourier series in T1 and T2, and ak,k2(t) represents
the probability amplitude of finding particle 1 with momentum fik1 and par-
ticle 2 with momentum fik 2 . Since states inside the Fermi sphere are occupied
it has to hold that
(8.51)
228 8. Pair Correlations Among Fermions and Bosons

because of Pauli's principle. We now study the behavior of the wavefunction in

time. If no interaction between the two electrons is present, i.e. E = ck, + ck 2,
it follows that
. a
Ih at ak,k2(t) = (ck,+ck,)ak,k 2(t) (8.52)

directly from (8.49) and the Schrodinger equation, and thus

ak,k 2(t) = exp [-i ( ck, ~ ck2 ) t] ak,k2(0) . (8.53)

This is a pure phase factor, which is already known from the uncorrelated
wavefunction (8.48). Once the particles interact they will constantly change
their momentum because of the ever recuring scattering from each other: In
one scattering event kl' k2 will become k~, k~. The Hamiltonian of the two-
electron pair with interaction V( TI, T2) reads
h2 2 h2 2
H =- - V I - -V2 + V(TI,T2) (8.54)
A

2m 2m
and we deduce from (8.49) and the time-dependent Schrodinger equation that
. a
Ihatak,k2(t) (ck, +Ck2)aT,T2(t)

(8.55)

Evidently, the matrix element (k i k21 V (T I, T2) Ik~ k~) describes the elementary
scattering process k I k2 - 4 k~ k~ mentioned above. In detail, it becomes

(kIk2IV(TI,T2)lk~k;) = :2 J d3TId3T2 eik'T' eik2·r2

(8.56)
In general, the Fourier transform of a matrix element (kIk2IVlk~k;) is given
by

x (kIk2IVlk~k;)e-k;.r, e-ik;.k;. (8.57)

This means that an arbitrary amplitude (kIk2IVlk~k~) generally implies
a nonlocal potential (TIT2IVIT~T~). The corresponding non local Schrodinger
equation reads

(8.58)

Hence, a change in the wavefunction P(TI' T2, t) at positions TI, T2 also de-
pends on the wavefunction at the distant positions T~, T~. Only if
(8.59)
 8.4 Cooper Pairs 229

does (8.58) turn again into a local Schrodinger equation. In this case

J
(klk2IV( Tl, T2, T~, T~) Ik~ k~)
2..
V2 d 3r 1 d 3r 2 d 3r'1 d 3r'2 eikl·Tl eik2·r2

V(rl,r2)J(rl - r~)J(r2 - T~)e-k~.rl e-ik;k;

J
X

d3rl d3r2 ei(kl -k~)·rl ei (k 2-k;).r 2V( rl, r2) , (8.60)

which can be simplified further if the potential V depends only on the relative
distance (for two-body forces this is always the case), i.e.
V(Tl' r2) = V(TI - r2) .
Then, it holds that
(klk2IV( T}, T2, r~, T~) Ik~ k~)

~2 J exp {i [(k 1 - k~) . (Tl - T2) + (k2 - k~ + kl - k~) . r2]}

x V(TI - T2) d3rl d3r2

J exp (iq· r) V(r) d 3r J exp [i(k 1 + k2 - k~ - k~) . r2] d3r2

V1 V ( q)J(kl + k2 - kl" - k2 ) , (8.61)

where the momentum transfer is

_ k
q-l-l-
k' _ kl - k~ - (k2 - k~)
2
k~ - k~ == k _ k' .
2
The total momentum K of the pair is given by K = kl + k2 = k~ + k~.
Indeed, the J function in (8.61) expresses that the total momentum of the
pair is conserved during the collision. Hence, we can also write down result
(8.61) in the form

(klk2IV(Tl,T2,r~,T;)lk~k;) = ~V(k - k')J(K - K'). (8.62)

Since the total momentum is conserved during the scattering process, the
matrix element of a nonlocal interaction in general will only connect states
with identical total momentum K = kl + k2 = k~ + k~. Thus,

(klk2IV(Tl,T2,r~,T;)lk~k;) = ~Vkk'(K)J(K - K') (8.63)

and the nonlocal potential takes on the following form in coordinate space:

(rlr2IVIT~r;) = J d3 K d3keiK·Reik.r

x Vkk'(K)8(k - k')e-iK'.R' e-ik'.r'

(rIV(R - R')lr') , (8.64)

where
230 8. Pair Correlations Among Fermions and Bosons

R R'
r~ + r~
2 (8.65)
r r' r~ - r~
represent the center-of-mass and relative coordinates, respectively. Let us con-
sider an eigenstate with energy E and sharp (conserved) total momentum. The
amplitude (8.50) can be transformed into
a k,k2 (t) = a k(K) e-iEtjli (8.66)
and (8.55) becomes

(E - ck, - Ck2)ak(K) = L Vkk,(K)adK). (8.67)

k'

Although the total momentum fiK only appears as a parameter, these equa-
tions are very difficult to solve with a general force. Electrons in a metal
experience two kinds of forces: electrons repel each other as a result of the
Coulomb interaction and they are attracted by the ions in the crystal lat-
tice. A moving electron pulls the ions a little bit in its direction and away
from their corresponding crystal lattice sites. Since the ions are very heavy,
the shifts remain very small. But the moving ions pull along other electrons
which in this manner effectively follow the first electrons. In other words, two
electrons that usually repel each other are able to attract each other in a
crystal lattice. Of course, this depends on the structure of the crystal lattice
and the kind of ions. The interaction resulting from these repulsions and at-
tractions are in some metals attractive for electrons near the Fermi surface.
This can be approximated as
Vo
Vkk,(K) = { 6' for kF < kr, k2' ki, k~ < ka ,
otherwise,
(8.68)

where Vo is positive and ka is slightly larger than kF (see Fig. 8.6).

Then, (8.67) becomes

(E - Ck, - ck 2)ak(K) = - Vo
V "'"
L.J ak,(K) . (8.69)
k'

The prime (') for the summation sign indicates that the k' summation only
runs over those k' that lie inside the shaded spherical shell sketched in Fig. 8.6:

kF < I~ ± k'i < k a . (8.70)

From the relation between k' and K given after (8.61) it is clear that
K
Fig. 8.6. Electrons in- 2+ k (8.71)
side a small spherical shell K -k.
kF < k < ka and at the 2
same time outside the
Fermi sphere kF attract Condition (8.70) corresponds exactly to the spherical shell for the involved
each other particle momenta noted in (8.68).
 8.4 Cooper Pairs 231

The sum 2:~, in (8.69) cannot vanish. This can be seen as follows: The
amplitude ak(K) has to be different from zero for some k; otherwise the
correlated wavefunction (8.49) would vanish identically. If 2:~, in (8.69) were
equal to zero then it would follow for the corresponding k on the left side of
(8.69) that

E _ _ (nk)2 (nK)2
- Ek, + Ek2 -
m
+ 4
m
. (8.72)

For a given K and E the last equation can only be fulfilled for one k; con-
sequently, only one ak (K) can be different from zero. As a consequence the
sum would not vanish, i.e. 2:~, ak' (K) f= 0, which contradicts the original
assumption. Thus, we have to deduce 2:~, ak,(K) f= O.
In order to solve (8.69), both sides are divided by (E - Ek, - Ek2) and are
then summed over k. This leads to
~' vo~' 1 ~' (8.73)
~ ak(K) = -V ~ (E-E -E ) ~ adK).
k k k, k2 k'
Since 2:~ak(K) = 2:~, ak,(K) we get
vo~' 1
(8.74)
1 = -V ~ (E-Ekl -Ek 2)'
Graphically this equation can be solved easily: Let us depict the function
l' 1
X( E) = - ~ -:-=-----;- (8.75)
V ~ (E - Ek, - Ek 2)
and look for intersections with the line -1 I vo, the latter being parallel to the
E axis (see Fig. 8.7).

X(E)

Fig. 8.7. Graphical solu-

tion of (8.74)

Evidently, the function X(E) possesses poles at all possible energy values
E = Ekl + Ek2 of a noninteracting electron pair outside the Fermi sphere.
These electron pairs need to have a total momentum nK and single mo-
menta nk 1 , nk2 in the spherical shell between nkF and nk a . Hence, the
lowest pole position has to lie on the Fermi surface itself, i.e. at energy
E = 2EF = 2(nkF)2 12m. The function X(E) has many intersections with
the straight line -l/vo above 2EF. Essentially these energies correspond to
the energies of unperturbed (noninteracting) pairs; qualitatively, the corre-
sponding states are similar to these uncorrelated pairs. Since Vo is positive
232 8. Pair Correlations Among Fermions and Bosons

[see (8.68)] an additional intersection point below 2EF exists:

Ebound = 2EF - 2.:1. (8.76)
This implies that the interaction leads to a bound state of the two electrons,
which is qualitatively different from the uncorrelated two-electron states out-
side the Fermi sphere.
In order to inspect the last result in more detail we consider the case
K = 0 of vanishing total momentum and look for its binding energy Ebound.
Then,
ki = k and k2 = -k
and the function X(E) in (8.75) can be easily calculated with the volume
element in k space:

X(E) = r kad3 k 1
JkF (27f)3 E - 2Ek
r kad3k 1

lea
JkF (27f)3E-(hk)2/m
m 1
(8.77)
27f 2 h2 eF k dEk E - 2Ek '
where Ea = (hk a)2/2m. Since ka is close to kF, the k in the integrand can be
replaced by kF. For E < 2EF it follows that

X(E) - N(O) In 1 2E a - E 1 (8.78)

2 2EF - E '
where

(8.79)

simply represents the density of states near the Fermi surface. The last
identity can easily be verified by using the well-known relation o(j(x)) =
(1/1df Idxl)o(x - a) for 0 functions, where a is the position x for which
f(x) = O. Once according to (8.76) E = Ebound = 2EF - 2.:1 is inserted
into (8.78), so that the group of equations
1
X(E)
Vo
_ N(O) In 12E a - 2EF + 2.:11 1
2 2.:1 Vo
2.:1 e- 2/ N (O)vo (8.80)
2(Ea - EF) + 2.:1
following from (8.74) is automatically solved. We get

e 2/ N (O)vo - 1.
(8.81 )

Now
(8.82)
 8.4 Cooper Pairs 233

is of the same order of magnitude as the Debye energy nwD. The Debye fre-
quency WD characterizes the maximum frequency of lattice vibrations in met-
als; see Exercise 8.3.

EXERCISE

8.3 The Debye Frequency

Problem. Determine the maximum frequency of the lattice vibrations in a

cubelike crystal of length L. The total number of atoms inside the cube is
N, so the number of degrees of freedom is 3N. Assume that the number of
lattice-vibration modes is 3N at most.

Solution. The attractive interaction, which leads to the bonding of electrons

into Cooper pairs, is a result of the exchange of phonons, i.e. the elementary
vibrational excitations of the crystal lattice. The effect of these phonons on the
electrons has been introduced approximately in (8.68) as effective interactions
between the electrons with momenta k, k':

for kF < Ikl, Ik'i < ka

(1)
otherwise.
kF represents the momentum of the Fermi surface and ka is a cutoff momentum
for the interaction. We can understand this maximum momentum in such a
way that for energy differences
/i 2 k a2
= __ /i 2 k 2
.6.E ~ Ea - EF ___
F (2)
2m 2m
above the Fermi surface, excitation modes of the lattice that can propagate
(transmit) the electron-electron interaction are no longer present. This max-
imum energy is called Debye energy ED,
ED = nwo; (3)
Wo is called the Debye frequency. In the context of the Debye model we can
quite easily give an approximate estimate for Wo.
Suppose we have a solid cube of length L. The excitations (vibrations) of
the lattice are given as standing waves, which can be constructed in a cube.
Then the waves 'P(k) are
'Pk(X) rv sin(kxx) sin(kyY) sin(kzz) ,
where boundary conditions are assumed in such a way that the excitations
vanish at the edge of the solid system, i.e. we require
sin(kixi)lxi=o = sin(kixi)lxi=L = 0, i = 1,2,3. (4)
This leads to a spectrum for the wavevectors
27rni
ki = (5)
L
234 8. Pair Correlations Among Fermions and Bosons

Exercise 8.3. where ni represent integer numbers. This implies that the level density in k
space for a volume L is given by
L3
Pk = (27r)3· (6)
The maximum value for k is determined in such a way that we consider N
particles in the solid-state system which have exactly 3N vibrational degrees
of freedom. We divide the solid-state excitations into longitudinal and trans-
verse oscillations. kmax is determined by requiring that the maximum number
of longitudinal oscillation modes is equal to Nand 2N for the transverse
oscillation modes because of two polarization degrees of freedom. Hence, for
longitudinal oscillations

l o
lkmax'
d 3 kpk = N

or

N. (7)
The maximum wave number
k max
3 6.".2N (8)
= "V
is essentially given by the number of lattice atoms per volume. Making use of
the dispersion relation
(9)
with the propagation velocity VI (speed of sound) of the longitudinal oscila-
tions in the solid and a similar calculation for the transverse waves, we obtain
for the Debye frequency

_ 3 67r2N/V.
WD = W max = VIVt
vr + vr ' (10)

Vtrepresents the corresponding propagation velocity of the transverse waves.

For metals, the Debye frequency is usually about 30 meV.

Typical values are nwD/cF ;: : ; 1/100 and N(O)vo ;::::; 1/4. Thus, with the
help of (8.82) the relation (8.81) can be simplified to
.:1 ;::::; nwD e- 2 / N (O)v o . (8.83)
This is the binding energy per electron in the bound (electron-pair) state. The
extremely nonlinear dependence of this pair energy .:1 on the force parameter
Vo is remarkable. This mechanism of pairing of electrons was discovered by
L.N. Cooper,6 hence the name Cooper pairs.
6 L.N. Cooper: Phys. Rev. 104 (1956) 1189. This idea was most essential for under-
standing superconductivity. Together with J.R. Schrieffer and J. Bardeen, L.N.
Cooper received in 1972 the Nobel Prize for physics.
 8.4 Cooper Pairs 235

Our analytical studies make use of the assumption that the total momen-
tum of the Cooper pair is fiK = O. If this is not the case, i.e. for fiK =1= 0, the
number of possible k values in (8.75) decreases rapidly. As already stated ear-
lier, the poles of x( E) characterize the energies of the noninteracting (or only
weakly interacting) pairs. Since now fewer k values are available (allowed), the
value of X(E) for E < 2cp (see Fig. 8.7) decreases. Hence Ebound = 2cp - 2.1
approaches closer to 2cp and the binding energy 2.1 of the Cooper pairs be-
comes smaller. This decrease is strong for an increasing total momentum fiK.
Fig. 8.8. Electrons that
It is important to note that the smaller the number of states is, which are
are in opposite directions
connected via the attractive interaction (klk2IVlk~k;), the smaller the bind- outside the Fermi sphere,
ing energy of the Cooper pairs becomes. This also follows from (8.83) for i.e. (k 1 = -k 2 ), have total
N(O) --+ O. We obtain the largest binding energy for electrons that possess the momentum fiK = 0 and,
total momentum fiK = 0; for them kl = -k2' Those single electron states lie hence, the largest level
density. This implies that
outside the Fermi sphere in opposite directions (see Fig. 8.8).
that their pair energy be-
In order to determine the wavefunction of a Cooper pair we use (8.69), comes largest
according to which
1
aK(K) = E const. (8.84)
- Ck, - ck2

This is so because the right-hand side of (8.69) is independent of k, i.e. it is

constant with respect to k. Inserting (8.84) into (8.66) and this into (8.49) and
using kl = K /2 + k and k2 = K /2 - k, results for a fixed total momentum
fiK, in

VJ (rl,r2) =
1
N 1eiK '(T,-T2)/2_
V
L ' Ee-
k
ik .(r,-r2)

Ck, - Ck 2
(8.85)

= rl
J'
Hence, with r - r2 the relative wavefunction of the Cooper pair reads
d3k e ik .r
tp(r) = N2 -- . (8.86)
(21T)3 E - Ck, - ck2

Nl and N2 are suitable normalization factors. For a pair with K = 0 this

results in

41Tm [cos kpr sin kpr] 2mL1

- N2 ----.::2 - - 2-
It ar
+ -2-3-
a r
, a = fi 2 k p ,
(8.87)

where (8.72) has been used; this will be calculated in detail in Exercise 8.4.
This relative wavefunction has its first extremum at

from which
236 8. Pair Correlations Among Fermions and Bosons

1 h,zk
d = rmax = - ~ ~ 1O-4 cm (8.88)
a 2mLl
follows. The correlation length d approximately characterizes the extension of
a Cooper pair. It is remarkable that cp(r) in (8.87) represents an s wavefunc-
tion; it does not depend on the angle and is thus proportional to Yoo CO, cp).
Hence, the angular momentum of a Cooper pair is l = O.
Superconductivity is a result of the simultaneous interaction of all elec-
trons. All electrons at the Fermi surface build up correlated pairs with the
same total momentum. Their correlation can be such that a current is pro-
duced, which is then called a super-current. In order to stop it, all electron
pairs have to be stopped at the same time. This is not the case for a con-
ventional conductor: electrons contributing to the current are stopped one
after the other. In a superconductor the correlated electron current has to
be stopped in total and instantly. Of course, this is very difficult; this is the
reason why the super-current flows for ever. 7

EXERCISE

8.4 Correlation Length of a Cooper Pair

Problem. Determine the relative wavefunction cp( r) of a Cooper pair for the
total momentum K = 0 [see (8.87)]. Estimate the extension of a Cooper pair.

Solution. According to (8.86), cp(r) is given by

cp(r) = N2 Jd3k
-- -----
e ik .r
(2np E - 10k, - ck 2
(1)

From (8.72), for K = 0 we have

(hk)2
lOki +ck2 = - - . (2)
m
With (8.76) we obtain

cp(r) = N2 J d3k eik .r

(2np 2cF _ 2.1 - (hk)2 jm

= N2
J d3k
(2np (h 2 jm) (k~
eik ' T
- k2 - 2mL1jh2)
(3)

We make use of a partial wave decomposition of eik ' T , i.e.

L inp'n + l)jn(kr)Pn(cosrJ).
00

eik ' T = (4)

n=O
The angle integration can then be performed. The jn(kr) are spherical Bessel
functions. Pn (cos rJ) represent Legendre polynomials for which the orthogo-
nality relation
7 This argument goes back to J.R. Schrieffer: Theory of Superconductivity (Ben-
jamin, New York 1964).
 8.4 Cooper Pairs 237

L" Pn (cos t9)Pm (cos '19) sin '19 &0 _2_0

2n + 1 nm
(5)
Exercise 8.4.

holds. Inserting (4) into (3) yields

<p(r) = N2 ~
J 00 k 2 dk (2n + l)jn(kr)
(21f)3 (h2 /2m) (k~ - k 2 - 2m6./h2)

x J J d<p dt9sin t9Pn(cost9) . (6)

The '19 integration can also be written as [Po(x) = 1]

J dt9sint9Pn(cost9)Po(cost9) = 20no, (7)

where (5) has been used. With jo(x) = sin x/x we obtain

<p(r) =
41fm
N2---,;:2
J k 2 dk sin kr
(21f)3 kr (k~ _ k2 - 2m6./h2) . (8)

Since the k integration should go only over a range 6.k = ka - kF « kF above

the Fermi surface we can write
k~-k2 ~ kF(kF-k) = -kFX. (9)
Hence,
sin kr = sin [(k - kF + kF )r] sin [(kF + x)r] (10)
and insertion into (8) yields

<p (r )
~ - N 241fm
~ --
1L1k d X
sin [(kF + x)r]
. (11)
h2r 0 x + 2m6./!i2kF
Mainly small x values contribute to the integral. Hence, we are allowed to
shift the upper integration bound to approximately 00; remember that a =
2m6./!i2k F. This gives

<p(r) ~
41fm ( sin kFr
- Nr--;:z
nr
10
00
dx -
x+a
xr
cos - + cos kFr 10
00
dx -
x+a
xr) (12)
sin -

~ N2 ~~ {sin kFr[+ci(ar) cos(ar) + si(ar) sin(ar)]

+ coskFr[-ci(ar) sin(ar) + si(ar) cos(ar)]} . (13)
The integral sine si(x) and the integral cosine ci(x) appear here; they are
defined via the integrals

si(x) jX sint t dt '

00
(14)

ci(x) j00
x cost dt.
t
(15)

The series expansion of these functions reveals their asymptotic behavior for
large r:
238 8. Pair Correlations Among Fermions and Bosons

Exercise 8.4. M-l (-I)m(2m)! ]

si(x) - cos x [ "~ x2m+l
+ O(lxl- 2M - 1 )
m=O

N-l (-I)m(2m - I)! ]

+sinx [ ~l x2m + O(lxl- 2N ) , (16)

N-l (-I)m(2m -I)! ]

ci(x) cosx [ ~l x2m + O(lxl- 2N )

+sinx [~ (-;2:~;n)! + O(IXI-2M-1)] (17)

with M, N = 1, .... For x --> 00 we obtain

cosx
si(x) : : : !
x (18)
sinx
ci(x) : : : !
x
Insertion into (13) yields, for large 1',

_ N2 47rm cos kFr _ N2 27rkF cos kFr

rp( 1') : : : ! ti,2 1'2 Ll 1'2 . (19)
In order to get an impression of the extension of a Cooper pair we take into
account the next terms from (17) for the expansion of the wavefunction (13)
and obtain
47rm (cos kFr sin kFr)
rp (l' ) : : : ! -N2 ---.::2 - - 2 -
It 0:1'
+ --2-3-
0: l'
. (20)

Determining the first extremum of this function rpmax(rm), we arrive at the

relation
o:rmax : : : ! 1. (21)
Thus, the extension of the Cooper pair is of the order of magnitude
1 ti,2 k F
d - rmax : : : ! -;; = 2mLl (22)
with numerical values for rm in the range::::::! 10- 4 cm.

EXERCISE

8.5 Determination of the Coupling Strength

of a Bound Cooper Pair

Problem. Determine the coupling strength Vo needed to obtain a bound

Cooper pair (E < 0) with total momentum K = 0 for the case of vanishing
Fermi momentum (kF = 0).

Solution. We know from (8.74)-(8.80) that the following relation has to be

fulfilled for a bound state (K = 0):
 8.4 Cooper Pairs 239

1 {ka d3 k 1 Exercise 8.5.

(1)
- Vo = ikF (27r)3 E-h2k 2/m'
With E = 2cF - 2..1 [see (8.76)] and kF = 0 we obtain
1 {ka d3k 1
Vo io (27r)32Ll+h 2k 2/m
m {k" k2
(2)
27r 2 h 2 io dk 2mLl/h2 + k2 .
The integral can be calculated in an elementary way:

J dX~2
l+x
=x - arctan x .

The result is

!.-
Vo
= ~
27r 2 h 2
(ka - J 2mLl arctan
h2
ka
J2mLl/h2
) (3)

We want to find out the conditions under which a bound pair is able to exist;
hence, we consider the limiting case ..1 -> 0 and obtain

(4)

as the minimum interaction strength. Contrary to the case for kF =I- 0 in

(8.81), a bound pair is obtained only for interaction strenghts that are larger
than a certain minimum value voin. This is because of suppression of the
pairing effect as the result the small phase-space level density d 3k/(27r)3 for
Ikl -> O. This becomes more clear once we look at the approximate relation
(8.81), which in principle holds only for kF > 0:
..1 = Ca - CF (5)
e 2 / N (O)vo_1'
Here, N(O) represents the density of states near the Fermi surface. Of course,
for kF -> 0 this density becomes N(O) -> 0; thus, within this approximation
the binding energy ..1 vanishes for all couplings Vo.
9. Quasiparticles in Plasmas and Metals:
Selected Topics

In the last chapters we saw that the concept of a quasiparticle in many-particle

physics has many advantages: A system of interacting particles can be treated
- in a suitable approximation - as a system of noninteracting quasiparticles.
Now we will make further use of this approach for an electrically neutral
system consisting of electrons and positively charged ions, i.e. a plasma. Such
plasmas appear manyfold in nature: the gas in a flame or in an electric dis-
charge, the matter in plasma reactors or in a star. In all these cases we are
dealing with some kind of a plasma. Here we will assume the plasma to be
isotropic and homogeneous; then, the physical properties are translationally
invariant.
It is convenient to simplify the problem by assuming a self-consistent field:
the particles interact with an electrostatic potential <p(x, t), which in return
has to be calculated from the average charge density p(x, t) of the plasma.
With the techniques developed in Chap. 6, the Hamiltonian of the system
becomes, in second quantization,

(9.1)

The index i distinguishes the particle species, i.e. electrons or ions. We expand
~t and ~ into plane waves,
, , e iq .",
Wi = Li
biq fTj
vV
(9.2)

(box normalization), where V = £3 represents the normalization volume. For

the Hamiltonian we now get

(9.3)

With the help of the Fourier transform of the potential <p,

0(q2 - qd = JVd3x .
e,(q2-qd''''<p(x) (9.4)

the interaction Hamiltonian can be written as [see (6.7) and the equations
following it]

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
242 9. Quasiparticles in Plasmas and Metals: Selected Topics

Hr = LLLeib!qjbiq2c,O(q2 - qt). (9.5)

qj q2
Since the potential <p(x, t) is a one-body potential, only one creation and one
annihilation operator appear; however, in nondiagonal form. Equations (9.3)
and (9.5) express the Hamiltonian in second-quantized form, which enables
us now to study such relevant physical quantities as, for example, the time
dependence of the particle-number operator. Mathematically speaking, in its
most general form we will study products of the form
A _ At A
Niq'q - biq,biq .
The time dependence of an arbitrary operator 6 is governed by Heisenberg's
equation of motion:
h aA A A
-:--a
1 t
0 = [H,OJ-,
which is in our case

(9.6)

Here [A, BJ- = AB - BA denotes the minus commutator. With the help of
(9.3) and (9.5) we can calculate the commutator. The result does not depend
on the choice of boson or fermion commutation relations for the operators biq .
We obtain
a At A
at biq,biq
(9.7)

The first part of (9.7) is a consequence of Ho: For the case q = q' the operator
b!q,biq represents the particle-number operator for particles of species i with
momentum q; for this case the first term in (9.7) vanishes, i.e. Ho does not
effect any change in the particle number. Only the potential changes the
particle number. The first term within the square brackets ~ b!q,b sp annhilates
a particle with momentum p and creates another one with momentum q';
it describes the increase in the particle number with momentum q' due to
scattering from other momentum states. The second term within the brackets
describes the decrease of the particle number with q' due to scattering to the
momentum p.
In order to find the relationship to classical quantities it is convenient to
introduce the distribution function of the particle species i:
(9.8)

where 10:) represent the states of the system and Pa is the probability of
finding the system in state 10:). Equation (9.7) becomes

:t Fi(q', q, t) = ~(Eiq, - Eiq)Fi(q', q, t)

 9. Quasiparticles in Plasmas and Metals: Selected Topics 243

+ i~i L [0(p - q)Fi(q',p, t) - 0(q' - p)Fi(p, q, t)] . (9.9)

p

The physical significance of the distribution function Fi(q', q, t) becomes clear

once we consider the particle-number operator N = ,(ft,(f. Its mean value with
respect to the states 10:) is obtained by averaging with Po: over the expectation
values (o:INlo:):
(Ni(X, t)) = L Po: (o:l,(fl (x, t),(fi(X, t)lo:) . (9.10)

We insert (9.2) and obtain

A
(Ni(x,t)) =
" '" '" At A eiq .x
'L.,..L.,..L.,..P",(o:lbipbiP+qlo:)V
q p '"
iq·x
' " ' " At Aip
L.,.. L.,.. (b ip b +q I
e
V
q P
iq·x
L L Fi(p,p + q, t) e V . (9.11)
q p

This result suggests the introduction of the distribution function Fi(x, p, t) in

phase space (coordinate and momentum space) according to
iq·x
Fi(x,p,t) LFi(P,p+q,t)eV; (9.12)
q
thus,
LFi(X,P,t) . (9.13)
p

On the other hand we derive for the momentum distribution for the particle
species i

(9.14)

Hence, Fi(x, p, t) is directly connected with the particle-number operator.

The definition (9.8), however, means that Fi represents not a classical, but
a quantum-mechanical distribution function;l frequently it is known as the
Wigner function.
For a given potential tp [see (9.4) and (9.5)] the quantum-mechanical dis-
tribution function is determined from (9.9). On the other hand the potential tp
depends on the Wigner function itself. In order to formulate this problem self-
consistently, we simply insert (N(x, t)) into the Poisson equation and obtain
for tp

\7 2 tp _- - 'L.,..L.,..L.,..
" '" '" 471' e i ( p,p+q,t ) e iq·x .
VFi (9.15)
. q p

1 If Fi(x, p, t) were a classical distribution function, Fi(x, p, t) would specify the

probable number of particles in the volume d 3 x and in the momentum range d 3 p.
The expressions (9.13) and (9.14) show that this is not the case for the quantum-
mechanical distribution function; however, the Wigner function is analogous to
the classical distribution function.
244 9. Quasiparticies in Plasmas and Metals: Selected Topics

From these two equations, (9.9) and (9.15), the distribution function Fi(p,p+
q, t) and the averaged potential cp have to be determined by successive it-
eration. Note, however, that (9.15) already represents a certain approxima-
tion because the true charge density has been replaced by the averaged one.
By starting from an initial distribution function Fi(O), an averaged potential
cp(O) can be calculated. With this potential cp(O) another new approximation
for Fi can be determined, and so forth. In atomic physics this approxima-
tion is known as the Hartree approximation. Equation (9.15) and (9.9) are
the quantum-mechanical analogy to the so-called Vlasov equations in classical
plasma physics.
To gain more insight into this system of equations let us consider the
equilibrium case for which the charge density and the potential cp vanish. In
equilibrium (cp = cp(O) = 0 --> <prO) = 0), (9.9) takes on the form

of;(O) (q', q, t) i ( ) (0)(,

ot = Ii E iq, - E iq Fi
)
q , q, t ,
and it follows that
Fi(O) (q', q, t) = F?)(q', q) e(i/h)(Eiq,-Eiq)t .
For cp = 0, the states 10:) in (9.8) have a fixed particle number; as a consequence
Fi(O)(q',q,t) rv 8q'q, which leads to

Fi(O)(q',q,t) = F?\q)8q'q.
We now consider the case in which Fi deviates only slightly from equilibrium
and oscillates around it. Hence, we set
Fi (q ' , q, t) = F(O)()J:
i q Uqq' + Fi ( l )q(,,q) e -iwt , (9.16a)

where FP) should stay small compared to Fi(O). Also for the potential cp we
set
cp = cp(O) + cp(l) e -iwt . (9.16b)
We insert this into the equation of motion (9.9) and deduce

iwFP) (q', q) e- iwt = iv(q', q)FP) (q', q) e- iwt

+ i~i L [<p(l)(p _ q) e- iwt (Fi(O) (q')8 q ,p + FP)(q',p) e- iwt )
p

_ <p(1)(q'_p)e- iwt (F}O) (q)8 pq + Fi(l)(p,q)e- iwt )] ,

where Vi(q', q) = (1/h)(Eiq , - Eiq). This equation can be rearranged as

i[w - Vi(q', q)]Fi(l) (q', q)
;:::; i~i (<p(l)(q' _ q)Fi(O) (q') _ <p(1)(q' _ q)Fi(O) (q)) ,

once the terms on the right-hand side proportional to <p(1) F(l) are neglected.
<p(l) and F(l) are treated as small quantities, so quadratically small terms,
such as the term <p(l) F(l) just mentioned, can be neglected. This yields
 9. Quasiparticles in Plasmas and Metals: Selected Topics 245

(9.17)

with the frequency

Vi(q', q) = ~ (Eiq, ~ Eiq) . (9.18)

Again we insert the solution (9.17) into the Poisson equation (9.15) and arrive
at

(9.19)

Owing to (9.4) it holds that

rp(x) = Lcp(1)(~q)eiq.X (9.20)
q

and we have
V2rp(X) = ~ Lq2cp(1)(~q)eiq.X. (9.21 )
q

Then, (9.19) can be written as

LE(q,w)q2cp(1)(~q) = 0, (9.22)
q

where

(9.23)

E( q, w) is called the dielectric function of the plasma. In addition to the trivial

case 1jJ(1)(~q) = 0, (9.22) has the solution
E(q,W) = O. (9.24)
The solutions w(q) of this equation yield the frequencies of the waves prop-
agating with momentum q in the plasma. For the solution of (9.24) it is
convenient to express the momentum distribution Fi(O) (p) in terms of the ve-
locity distribution f?)(v) with v = lip/m, to replace the summation over p
by an integration, and to let the volume V go to infinity:

Lp
----> V J d 3 v;

we get

E(q,W) = 1 + L 47re; Jd 3 v fi(O) (v) ~ f?)(v ~ hq/mi) . (9.25)

,. q21i w~q·v+hq2/2mi

In order to find the classical limit of (9.25), we expand the nominator of the
integrand into a Taylor series and let Ii ----> 0:
246 9. Quasiparticies in Plasmas and Metals: Selected Topics

(9.26)

An interpretation of E( q, w) in the present form is not possible because the

denominator of the integrand in (9.26) may have zeros; the integrand is sin-
gular. However, it is possible to regularize the integral: w becomes complex
valued and the conventional methods of complex variables can be applied.
With a suitable choice of the integration path around the singularity on the
real axis the integral becomes regular. With an analytical continuation the
desired result is finally obtained. Hence, we replace w with w + iT} and make
use of Plemelj's well-known formula (see Exercise 2.13)

. +1
hm --.- = P
1)--+0+ X IT}
(1).l7TO(x) ,
-
X
- (9.27)

where P denotes the principle value of the integral. We obtain

E(q,W) = EI(q,W) + iE2(q,w) (9.28)
with the real part

EI(q,W) = 1 + "'""'
2
47re i PJd3v f(O)(
i v -
) f(O)(
i
ti~/
V -'''1
+ n2q2/2mi
mi
)
(9.29a)
, miq2
~ w - q. v

:::1 J
and the imaginary part

E2(q,W) = - L, d3v [fi(O) (v) - fi(O) (v - :~)]

x0 (w _q. v + n2q2)
2mi
. (9.29b)

This prescription regularizes the integral at the zeros of the denominator.

Hence, in general the solutions of the eq uation E( q, w) = 0 are complex valued.
As a consequence the plasma waves either increase exponentially or decrease
exponentially. With more advanced methods it can be shown that the solu-
tions w have a negative imaginary part and are thus damped exponentially if
fi(O)(V) is a monotonous decreasing function in v = Ivl. This condition is al-
2
ways realized in thermal equilibrium because then f(v) '" e- V . If the plasma
is removed far from equilibrium, solutions with a positive imaginary part may
appear and the perturbations increase exponentially. In this way one obtains
plasma instabilities, which have also been observed experimentally.

9.1 Plasmons and Phonons

After these introductory remarks we will now study in more detail the impor-
tance of the quasiparticle concept for a plasma. We neglect quantum correc-
tions to the frequency and employ the classical dielectric function (9.26).
Furthermore, for the equilibrium distribution functions of the electrons
and ions we assume Fermi functions:
 9.1 Plasmons and Phonons 247

for v < VF\} _ _3_ _ _

- 471"1/ 8(v VF,),
3 (9.30)
f(v)

for V> VF i Fi
where vF; represents the Fermi velocity of the particle species i. It is deter-
mined from v
471" k 3 _ 471" P~i
P = 3 F, - 3"""iiJ Fig. 9.1. Fermi distribu-
tion function of the elec-
trons or the ions
= !!... (3P)1/3 ,
(9.31 )
mi 471"
where p represents the particle density, which is identical for electrons and
ions because of homogeneity. Since we can rewrite FlO)(v) according to (9.30),
li(O) = (3P/471"v~J8(v - VF,), where 8 denotes the step function. The deriva-

tive of I?) becomes

Oli(O)
q . - - = -q.- v ( -3p- ) 8(V-VFJ. (9.32)
ov V 471"V~i
For convenience let us introduce the abbreviations
W
q = Iql,

(47::;r/ 2
, (9.33)

{ I for z < 1 }
u(z) = = 8(z - 1) .
o for z> 1
Here Wpi (i = electron or ion) are characteristic frequencies, which, as we
shall see, characterize the plasma oscillations to zero order. This enables us
to perform the integration (9.26) and we obtain

f(q,W) = 1 +"".
~
471"e; Jd3v q. /o~
W- q .v +
Oli(O)
, miq2 17]

1 - L. 471"e; (~) Jd v q. v 8(v - VF,). 3 . (9.34)

miq2 471"VF3.
, v W- q .v +
, 17]

We use (9.27) and arrive at (compare with Exercise 9.3)

1 + Zi
~1
3 wpi
f(q,W) = 1+"" - 222
2 q v Fi ( I .
I
2-ziln - - +l7rZi U (Zi)
1- z·1 ) (9.35)

In order to search for the zeros we assume Zel » 1, i.e. W » qVFd for the
electrons; for the ions this holds all the more because VF iou = vFel(meI/mion),
i.e. VF iou «VFel'
Furthermore,
2 1 2 1
2_ZlnI1+ZI ;:::;
3 z2 5 Z4
for Z» 1 . (9.36)
1-z
248 9. Quasiparticles in Plasmas and Metals: Selected Topics

We insert this into (9.35) and obtain

E(q,W) ~ 1- W~cl (1-

w2
mel)
mion
_~w~clvL(1+m~l/mrOn)q2
5 w4
(9.37)

In order to find the eigenmodes of the plasma we have - according to (9.24)

- to set E( q, w) = 0 and solve approximately for w2 :

w2 ~ w2
Pel
(1 _ mel)
mion
+ ~v2
5 F 01
q2. (9.38)

These oscillations are known as plasma oscillations. The corresponding quanta

are called plasmons. Their frequency is approximately constant and equal to
the plasma frequency w pel • The recoil correction is negligible (mel/mion « 1).
The correction term proportional to q2 the frequency to be momentum de-
pendent and causes dispersion. The plasma oscillations (9.38) are not damped
because the imaginary part of (9.35) proportional to u(z) vanishes because
Z» l.
Yet another solution for E(q, w) = 0 exists with Zion» 1 but Zel « 1; this is
possible because Zel/ Zion = mel/mion. However, the solution has a completely
different physical interpretation. At first we use the approximations

2 _ Zion In I~ + Zion I ~ 2
- Zion 3z;on '
(9.39)
2 - Zel In 111 +- Zel
Zel I ~ 2

and deduce from (9.35) that

w2 w2 3 w2 W
E(q,W) = 1- Pion + 3~ + i...2::~ (9.40)
w2 q v Fe}
2 2 2 3 2 q v Fel .

We set E(q,W) = 0, solve it, taking into account the outlined approximations,
and get

(9.41)

where
W= (9.42)

For long waves (small q) we have

_ _ VFel
w -- (w- -
Pion )
q_
-VFcl
- (mel) 1/2
-- q
v'3 w pel v'3 mion

and hence
1/2
_ VFel ( mel
W~W=- - -) q. (9.43)
v'3 mion

This is the typical relation for sound waves (proportionality between frequency
and momentum), i.e.
 9.1 Plasmons and Phonons 249

where the speed of sound Cs is given by

1/2
VF el ( mel ) (9.44)
Cs = y'3 mien .

The experimental results for longitudinal sound waves in metals are in good
agreement with (9.44); this holds particularly for alkali metals where a 20%
agreement is found. The oscillations associated with these quanta are denoted
as phonons. For large wavelenghts we obtain from (9.41)

W ~ W (1 __ (~) i
3y'3 mien
1/2) , (9.45)

i.e. a small negative imaginary part. Here the phonons are weakly damped.
In general the weak damping of waves in plasmas is called Landau damping. 2

EXERCISE

9.1 Electrostatic Potential of a Charge in a Plasma

Problem. Use (9.35) to determine the electrostatic potential of a charge Q

in a plasma.

Solution. Without a plasma the potential would be Q/r and the correspond-
ing Fourier transform would then be
41l"Q
tp(q) = Vq2 (V = L3 = volume) .

In the presence of a dielectric medium the potential becomes

tp(q)
tp(q) -+ tp(q,w) = - ( - ) .
(' q,w
Since the charge Q is stationary we have w = 0, i.e.
41l"Q
tp(q) = Vq2('(q, 0) ,

where according to (9.35)

1
E(q,O) 1 + q2)..2

with

2 See e.g. A.L. Fetter, J.D. Walecka: Quantum Theory of Many Particle Systems
(McGraw Hill, New York 1971); U.C. Kittel: Quantum Theory of Solids (Wiley,
New York 1966).
250 9. Quasiparticles in Plasmas and Metals: Selected Topics

Exercise 9.1. We invert the Fourier transformation and get

<p(r) = ~ jd3q<p(q)eiq.x
(27r )3

(2~)3 10 00
q2 dq 107r sin 19 q d19 q 1027r d<pq<p(q) eiqrcos t9 q .

Since the integrand is independent of <pq the integration over <pq can be per-
formed:

<p(r) =
V
(27r)3 ioroo q2 dq ior sin19 q
.
d19q <p(q)elqrCost9q .

Substitution of v = qr cos 19 q and dv = qv( - sin 19 q) d19 q yields

<p(r) = 4~2 j l dq <p(q) j (-dv) :i;

-V j q dq<p(q)-.1
2 ·_·
_(e- 1qr e1qr )
47r 2 tqr

- ; j q2 dq <p(q) sinqr.
27r qr
Insertion of <p(q) gives

(r) = 2Q roo <i dq sin qr .

<p 7r io r q2 + (1/,\.)2
This integral can be solved analytically:
roo x sin bx dx ~ e-(ab)
io a2 + x2 2 '
to give

<p(r) = 2. e- r /). .
r
-=--
Within the distance A the charge is screened. If instead of the Fermi distri-
bution a Maxwell-Boltzmann distribution is used for the derivation of (9.35)
~~~ .Q_~~ we arrive at
-
A = (~)
47rpe 2 .

Fig. 9.2. Illustration of This quantity is denoted as the Debye screening length. It characterizes the
Debye screening of a charge screening of a charge by a plasma charge cloud which is built up around the
Q in a plasma charge Q (see Fig. 9.2 and Exercise 9.2).

EXERCISE

9.2 Classical Dielectric FUnction

Problem. Use (9.26) to determine the classical dielectric function for particles
obeying a Maxwell distribution.
 9.1 Plasmons and Phonons 251

Solution. The normalized Maxwell distribution is given by Exercise 9.2.

f(O)(v) = ~e-v2/a2
1l"3/20;3

0; = (2k BT/m)1/2 represents a thermal velocity. Furthermore, according to

(9.26)

Eelass -
-
1+~
~ 41l"e;
i
--2
miq
J d 3 q. (of?) /av)
V
w - q. v
.

We consider only one-particle species and choose q along the z direction:

q = qe z . Then the x-v integration can be easily performed.
Insertion of the Maxwell distribution yields, for s = 0,

E(q,W) = 1-
81l"e 2p
23/25
mq 1l" 0;
J 3 q·ve- v2 / a2
dv-=------
w-q·v
The choice of cylindrical coordinates leads us to

E(q,w)=l-
81l"e 2p
23/25
1271" 1
dv<p
00
vl!dvl!
mq1l" 0; 0

1
0
00 qvz e- V 2/ a 2 e- V •2/ a 2
%

X dv z - - - - - - - - - - - - - - -

1
-00 w - q. V z

1-
161l"e p
2 1/2 5
2
00
VI!
d
VI! e
/a21°O d
_v 2
• Vz
qvz e- Vz
-=--"-------
2/
a
2

mq 1l" 0; 0 -00 W - q . Vz
The elementary integral over vI! is of the type

1o
00
x
2n+l
c
_px 2
=--
2pn+l
n!
(p > 0) ,

1
and therefore
81l"e 2p 00 qvze-v~/a2
E(q,W) = 1 - / dv z .
2
mq 1l"1 20;3 -00 W - qv z

l°°
Adding a zero in the form of adding and subtracting an identical term yields

E(q,W) = 1 - 81l"e 2p dV (qvz-w+w) e _v z2/a 2

W - qv z

1
mq 2 1l"1/20;3 -00 z
2 2/ 2

1- 81l"e p 00
dv (
- e_ 2/ 2
Vz a
we-
+- - - a- )
V
%

mq21l"1/20;3 -00 z W - qv z
1_ 81l"e 2p
mq 2 1l"1/20;3
(1 00

-00
w e-v~/a2 dv z
W - qv z
_ 0;,;7r)

1 + A21q2 - A21q2 (q:) Z ( ; )

where

A = (m0;2 )1/2
81l"pe 2
( ~)1/2
41l"pe 2
252 9. Quasiparticles in Plasmas and Metals: Selected Topics

Exercise 9.2. and

Z(z) = 100

-00
dx--
z - x
e-x2

represents the so called Fried-Conte junction, 3 which can be tabulated nu-

merically.

EXERCISE

9.3 Details of Calculating the Dielectric Function c(q,w)

Problem. Use expression (9.34) to verify the result (9.35) for E(q,W).

Solution. According to (9.34) we have

E (q , w) -_ 1 - L. 47re-
- - - ) j d3v -
; ( -3p
miq2 47rv~
q·v O(V-VFJ
- ----'---...::..:....-
v w - q . v + i1]
~ t , v .f

=1

We now determine I:
I = r2"" dcp r sin {} d{) roo v2 dv qv cos {} __0.. :.(v_-_V_F..,:.i.:....)_
1"
io io io v w - qv cos {} + i1]
cos {}
2
27r sin {} d {} qVF {} . .
o 'w - qVFi cos + 11]
The substitution U = cos {}, du = - sin {} d{} yields
r+ 1 udu
I = 27rqv~i LI W - qVFiU + i1]
Making use of (9.27) for the solution of the integral leads to

I = 27rqv~i (p
i-I
r+ 1
udu
w - qVFiU
-i7rjUdUO(w- qVFiu ))

r+l w -udu
27rqv~., {p i-I qVFiU
- i7rjUduo [(u -~)
qVF,
qVFi]}

27rqv 2 (p
Fi
r+l
udu
i-I W - qVFiU
_ i7r_w__ )
(qvFJ 2
With the integral

j xdx
a + bx
= ~ _ ~ lnx
b b2
it follows that

3 B.D. Fried, C. Conte: The Plasma Dispersion Function (Academic Press, New
York 1961).
 9.1 Plasmons and Phonons 253

I
2
27rQVF i {[ (_
U

qVFi
) -
w
- (
qVF i
)2 In Iw - qVFi ul ]+1}-1
.
- l 7qVFi
r -)2
(
W Exercise 9.3.

= 27rqv~i ( - q:Fi - q:Fi In I: ~ :~:: 1- i7r (qv:J2 )

Using the substitutions

Zi,

mi

U(Zi) { I for Zi < I} = e(z _ 1)

o for Zi> 1
we obtain, for (9.34), the desired result:

E(q,W) = 1- l:. -32 q2v~

-w~
- ' ( -2-ziln 1 w - w/ Zi
W +W/Zi
1 +i7rziu(zi) )
t '

3 wp Zi +1 .
1 + " - --;- 2- Zi In 1- - 1 + l7rZiU(Zi) ) .
~i
2 q2 v2Fi ( z·-l
t
10. Basics of Quantum Statistics

In this chapter we present some basics on the quantum-theoretical description

of macroscopic many-particle systems. A detailed disclosure of thermodynam-
ics and quantum statistics is out of reach for this book. 1 Here we will first
introduce the concept of a quantum-theoretical ensemble and the so-called sta-
tistical operator, or density operator in short, of a many-particle system, and
discuss its properties. We will see that with this approach not only can we
calculate expectation values of observables, but we are also led to new no-
tions such as entropy. Looking back on the chapter on quantum gases we will
discuss in detail the density operator of a canonical ensemble.

10.1 Concept of Quantum Statistics

and the Notion of Entropy

In classical mechanics as well as in quantum theory the dynamics of a phys-

ical many-particle system is described by differential equations for the time
evolution of those quantities assigned to observables. Hamilton's equations
for position and momentum coordinates are only one example in classical me-
chanics. In quantum theory the time evolution of observables is governed by
Heisenberg's equations of motion for the assigned operators, or, equivalently,
by the Schrodinger equation of the state vector. Knowledge of all initial con-
ditions is crucial for the fixing of the dynamics of the system. In classical
mechanics, for example, 6N space and momentum coordinates have to be
known at an initial time for an N-particle system. The quantum-theoretical
state vector of a system is fixed only by a measurement of a complete set of
commutable observables at an initial time. Once all possible initial conditions
are given in principle, so that the information about a system is completely
known, the state vector of the system is unmistakably fixed. We say that the
system is in a pure state. However, in practice a pure state cannot be real-
ized experimentally, neither in the classical nor in the quantum-mechanical
case. The reading of initial positions and momenta of classical systems and the
measuring of all observables, which characterize the state vector of a quantum
system, always come with uncertainties. Errors that occur in the measurement
for the initial conditions cause a "wrong" state vector to be fixed. Thus, in
1 We refer to W. Greiner, L. Neise, H. Stocker: Thermodynamics and Statistical
Mechanics, 2nd ed. (Springer, Berlin, Heidelberg 1994).

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
256 10. Basics of Quantum Statistics

principle the information about the system is incomplete. As a consequence

of these uncertainties with respect to initial conditions we always have to deal
with a whole group of pure states; all of them represent possible candidates
for the description of the "true" state of the system. Anyhow, for a macro-
scopic system with N ~ 10 23 degrees of freedom it is impossible to fix a pure
state; in addition, all this information would be extremely useless. Therefore
we will attempt to describe the state of a many-particle system with a few
global observables such as the total energy, pressure, and so on.
If one pure state cannot be given for a system anyhow, so that only a
group of possible pure states are realized, we speak of a mixture. The multi-
tude of pure states of the system realized under defined conditions is denoted
as an ensemble. A quantum-mechanical description of a many-particle system
is given in the form of mixed states that give information about the probabil-
ities of single pure states realized in the ensemble. It is the task of quantum
statistics to record the properties of a quantum-theoretical mixture.

10.2 Density Operator of a Many-Particle State

We have to develop a mathematical formalism that takes into account the
incomplete information about the state vector of a quantum-mechanical many-
particle system.
We denote the energy eigenstates of a many-particle system as 11ji~(n»)'
which are the solution of the stationary Schrodinger equation
(10.1)
These many-body state vectors are enumerated by the index n, which ab-
breviates all required quantum numbers necessary to identify a special state,
the corresponding energy of which is E(n). In some cases n will also contain
continuous quantum numbers like the total momentum of the system etc.,
but for simplicity we will take n here as one discrete variable. Complications
due to the continuous character of some quantum numbers or due to the con-
tinuous part of the energy spectrum can then be easily taken care of later
on. The state vectors 11ji~(n») are assumed to form a complete orthonormal
basis of Hilbert space, characterized by the completeness and orthonormality
relations
(1ji~(n)IIji;(m») = bn,m and 2:= 11ji~(n»)(Iji~(n)1 = i, (10.2)
n

respectively. In the last relation the operator

PI>f/n) = 11ji~(n») (Iji~(n) I , (10.3)
denotes the projector on the particular state number n, which denotes the
property
PI>f/n)PI>f/n) PI>f/n) , (10.4)
since Iljin) (ljinlljin) (ljin I = Iljin) (ljin I according to (10.2). If we now assume that
the system is exactly in the state number n, we say the system is in a pure
 10.2 Density Operator of a Many-Particle State 257

state, and the projector Plpn) is called the density matrix of this pure state.
However, as stated above, one usually cannot determine exactly in which
state n the system is. For example, if we allow an uncertainty of b.E for the
measurement of the energy E of the system there will be very many state
vectors l!li~(n») which fulfill E < E(n) < E + b.E. The reason is that in a
macroscopic system with many particles (N ---+ 00, V ---+ 00, N/V = const) the
energy levels are very close together. We therefore do not have the information
about exactly which state vector the system is in. This lack of information
about the exact state vector can be accounted for if we allow each state vector
to be assumed with a certain probability pn. We write
(0) = (!li~(n)IOI!li~(n»)pn. (10.5)
The probabilities pn shall be normalized according to

(10.6)
n
This average can be rewritten, by inserting the completeness relation from
(10.2), as

(0) = L (!li~(n) l!li~(m»)(!li~(m) 101!li~(n»)pn

nm

In the last step we have rearranged the bra and ket vectors such that we can
furthermore write
(10.7)
where the density matrix P is defined as
(10.8)
n

i.e. as a sum of all possible projection operators weighted by the probability

pn with which the actual physical system can be in state number n. We then
say the system is in a mixed state. For example, if we know that the energy
of the system is in the range E, E + b.E, according to its preparation or due
to a measurement, then we know
pn 0 for E(n) < E or E(n) > E + b.E
pn t 0 for E < E(n) < E + b.E .
However, the actual values of the probabilities pn with E < E(n) < E + b.E
are still not known to us. Here we make an important assumption that cannot
be justified from first principles. We assume that all state vectors that are
compatible with our knowledge about the system (here, energy in the range
E, E + b.E) are equally probable. This corresponds exactly to the assumption
of equal probabilities for the six possible outcomes in an experiment of tossing
a dice. This assumption can of course be wrong, if the dice is not ideal, but
258 10. Basics of Quantum Statistics

it is the most probable assumption as long as we do not have any further

information. Then we find in our example

if E < E(n) < E + I:l.E ,

(10.9)
otherwise,
where n(E, I:l.E) is the number of energy eigenstates in the interval E, E+I:l.E,
corresponding to the six possible outcomes in a dice-throwing experiment. In
the case of a continuous energy spectrum one needs to know the so-called level
density or density of states g(E), i.e. the number of states per energy interval,
and then one can calculate n(E, I:l.E) ~ g(E)I:l.E, as long as I:l.E « E.
One calls the exact state vector of a physical system the microstate. On the
other hand our knowledge about the systems defines the so-called macrostate.
In our example the macrostate is given by the energy of the system and in
most cases by the particle number and volume. The number n(E, N, V),
which turns out to depend only weakly on I:l.E as long as I:l.E « E, is then
the number of microstates of the system, each of which is represented by a
single pure state IqF~(n))' which are compatible with the given macrostate.
Our fundamental assumption is that all microstates within the small energy
interval are equally probable. And this leads to (10.9), which is called the
micro canonical probability distribution. We denote this by the subscript "mc" .
If we prepare a large number of systems, say N, with the same Hamilto-
nian, energy interval, particle number, and volume, then an exact measure-
ment of all microstates would yield a certain number N n of systems that are
exactly on microstate (or pure state vector) number n. One generally calls
such a large number of identical systems with the same macrostate an en-
semble. Up to now we have dealt with a macrostate defined by the energy,
particle number, and volume of the system and the corresponding ensemble
is called the microcanonical ensemble. If the number of systems is very large,
N -> 00, we expect that
~ -> pn for N -> 00, (10.10)
Le. that the relative frequency of microstate n in the ensemble tends to the
probabilities pn. Therefore the averages (10.5) and (10.7) are also called the
ensemble averages of the observable O. Now mathematical statistics shows
that there exists a unique function of the probabilities which is a measure
for the predictability of a probability experiment: the so-called uncertainty
function. It is defined as
(10.11)

where Pi is the probability of the possible outcome i and the sum runs over
all of them. This function is exactly zero if and only if there is no uncertainty
about the outcome of the experiment, i.e. if one of the Pi is 1 and all others
are zero. This means the outcome of the experiment is then only the special
event i. Note that lIn 1 = 0 and limX->CXl x In x = o. On the other hand 1i
is constructed such that the case of equal probabilities has the maximum
uncertainty. For an event with n possible outcomes we obtain Pi = lin and
 10.2 Density Operator of a Many-Particle State 259

1i = - Ln
-n In n I
i=l
l
= -n n In n = In n .
J&
(10.12)

The maximum uncertainty is thus given by the logarithm of the number of

possible outcomes of the experiment.
We show now that the constant probability distribution (all Pi equal) is in-
deed the distribution with the highest unpredictability. We therefore calculate
the variation of 1i with a variation of the Pi, and for 1i to have a maximum
this must vanish:
(10.13)
i=l i=l

However, the Pi are not independent of each other and from the normalization
we have a constraint for the allowed variations 5Pi:
(10.14)
;=1 ;=1

If we multiply (10.14) with a Lagrange multiplier, Cl:, and add the resulting
equation to (10.13) we find

L 5p; (In Pi - 1 + Cl:) = o.

i=l

Now we can argue that the variations 5Pi may be assumed to be independent
of each other, such that each term of the sum must vanish itself if we account
later on for the normalization condition by a suitable choice of Cl:. Then we
find
In Pi = 1 - Cl: or Pi = e 1 -o<. (10.15)
The important fact is that the right-hand side of (10.15) is independent of
i so that all Pi are equal and the normalization then yields Pi = 1/ n for n
possible outcomes. We have shown in this way that our assumption of equal
probabilities in the micro canonical ensemble in a given small-energy shell
corresponds to the assumption of maximum uncertainty.
Now, if the macrostate of the system is not specified by the energy but
rather by the temperature, T, of the surroundings of the system (e.g. a heat
bath) the system can assume all possible energy values and not only a small
interval around E (we have assumed b.E « E!). The reason is that the
heat bath is assumed to be very large compared to the system of interest
(Ebath » E) and the exchange of energy between the heat bath and system
therefore leads to fluctuations of the energy of the system. Fluctuations are
generally deviations of an observable from its mean value. However, the mean
value of the energy, i.e. of the Hamiltonian, is still fixed:
U = (H) = L(1ji~(n)IHI1ji~(n))pn
n

(10.16)
n
An ensemble of systems in which the macrostate is specified by a fixed mean
value of the energy and fixed particle number and volume is called a canonical
260 10. Basics of Quantum Statistics

ensemble. The corresponding canonical probability distribution can now be

calculated by requiring maximum uncertainty (10.13),

(10.17)
n

where the normalization (10.14) restricts the variations of the pn:

(10.18)
n

Now we have the additional constraint (10.16), which leads to the variational
condition
(10.19)
n

If (10.18) is multiplied by a Lagrange multiplier Q and (10.19) by -(3, and

if these two equations are added to (10.17), we obtain (the minus sign is
arbitrary, but makes (3 positive)

Here we can again assume all variations opn to be independent of each other,
so that
pn = ce-{3E(n) ,

where c = exp( Q - 1) can be determined from the normalization condition

(10.14) which yields the canonical probability distribution
e-{3E(n)
Pen = 2:k e-{3E(k) (10.20)

Here the probabilities of the states drop exponentially with the energy. In
Thermodynamics and Statistical Mechanics 2 it is shown that from (10.16)
and the fact that the entropy of the system is given by
(10.21)
n

where kB is Boltzmann's constant, one can identify the up to now unknown

Lagrange parameter (3 to be
1
(10.22)
(3 = kBT'
where T is the temperature of the heat bath that fixes the mean energy of
the system.
It is very interesting to rewrite the probabilities pn as the matrix elements
of the operator p defined in (10.8). From that follows
p;: = (tJiihn)IPeltJiE(n))' (10.23)

2 W. Greiner, L. Neise, H. Stocker: Thermodynamics and Statistical Mechanics,

2nd ed. (Springer, New York 1997).
 10.2 Density Operator of a Many-Particle State 261

Now we see immediately that the pn of (10.20) can be obtained as the ex-
pectation value of an abstract operator
, e-{3H
Pc = Tr(e-{3H) , (10.24)

the canonical density operator. Here the trace can be rewritten in the basis
of the It[/~(n) as

(10.25)
n n

and in the last step we used the fact that the abstract operator e-{3H, which
is to be understood as

exp( -PH) = f (_~)k Hk ,

k=O

can ,directly be replaced by an eigenvalue, since the It[/~(n)) are eigenstates

to Hk with the eigenvalues [E(n)Jk. In the canonical ensemble we have as-
sumed that the macrostate of the system is defined by the temperature, vol-
ume, and particle number. However, in many systems one does not even know
exactly the particle number. Consider, for example, vapor in equilibrium with
its corresponding liquid phase. The actual particle number in the vapor will
then no longer be a fixed quantity, rather the actual number of particles in
the vapor will fluctuate around a mean value (N). The liquid phase acts here
as a particle reservoir for the vapor in the same way as the heat bath in our
canonical ensemble acts as an energy reservoir. If we are interested only in the
properties of the vapor, disregarding its direct interaction with the liquid, we
must take into account that we do not have knowledge about its actual par-
ticle number. We can only say that there will be an average particle number
which must be calculated from the probabilities pn,N that define the proba-
bility of finding the system with exactly N particles in the N-particle state
number n. We require the normalization
00

(10.26)

The mean particle number is then given as the average of the actual particle
number, N,
00

L LNpn,N = (N). (10.27)

N=O n

Furthermore, in most cases a particle reservoir acts also as a heat bath that
fixes a certain temperature, so our macroscopic knowledge about the system
is only the corresponding mean value, exactly as in the canonical ensemble:
00

L L E(n, N)pn,N = (H). (10.28)

N=O n
262 10. Basics of Quantum Statistics

Here the energy eigenvalues must be numbered by two indices: N enumer-

ates the different possibilities for no particle at all in the system (N = 0),
one particle, two particles, and so on, whereas n still enumerates all energy
eigenstates at given particle number. These eigenstates then also carry these
two indices: 11]i~(~,N))' If we now again ask for that probability distribution,
pn,N, that maximizes our uncertainty about the system, we have to vary the
uncertainty function (10.11), which now reads
<Xl

- L Lpn,N Inpn,N = H, (10.29)

N=O n
under the subsidiary conditions (10.26), (10.27), and (10.28), with respect to
the up to now unknown probabilities. Completely analogously to the method
for the canonical ensemble, this can be done by varying (10.26), (10.27), and
(10.28) and multiplying the resulting variations by three unknown Lagrange
multipliers, say a, -(3, and ,,/, and adding these equations to the variation of
(10.29). One obtains
<Xl

N=O n
Again we can argue that each of the brackets must vanish separately if we
fulfill the subsidiary conditions (10.26)-(10.28) later on by appropriate choice
of the Lagrange multipliers. We find
pn,N = exp [-(3(E(n, N) - IlN)] (10.30)
gc 2:.";'=0 2:.n' exp [-(3(E(n',N') -IlN')] ,
where we have determined a from (10.26) and introduced the chemical po-
tential 11 by the relation 'Y = (311· The physical meaning of 11 = "//(3 as the
chemical potential must be derived from condition (10.27), which determines
this parameter and we refer the reader again to the lecture on thermodynam-
ics and statistical mechanics for this calculation. 3 The meaning of (3 is the
same as in the canonical ensemble and is given by (10.22). A system whose
macrostate is determined by a mean energy or temperature, a certain volume,
and a mean particle number or chemical potential is called a grand canonical
system and (10.30) is the corresponding probability distribution of the grand
canonical ensemble. This means that in a set of N systems (N -> 00) with
common temperature, volume, and chemical potential (T, V,Il) the probabil-
ity of finding one of these N systems exactly with N particles and in the
special microstate number n is given by (10.30). Exactly as for (10.24) for
the canonical density operator, we can also find the representation of the free
density operator of the grand canonical ensemble:
, exp[-(3(H - IlN)]
(10.31 )
pgc = Tr {exp[-(3(H -Il N )]} ,
where the trace now contains not only a sum over all expectation values with
a complete set of basis vectors in the N-particle Hilbert space, but also a sum
3 W. Greiner, L. Neise, H. Stocker: Thermodynamics and Statistical Mechanics,
2nd ed. (Springer, New York 1997).
 10.2 Density Operator of a Many-Particle State 263

over all particle numbers [see (10.30)]:

DO

Tre-{3(H-J-LN) = '"'" ,"",(tj/n,N

L...J L...J E(n,N)
Ie-{3(H-J-LN)Itj/n,N )
E(n,N) , (10.32)
N=D n

where we used here the complete set of energy eigenfunctions but, as we know,
any other complete set can also be used to evaluate the trace even though the
actual calculation will be more complicated.
It is clear that for other definitions of the macrostate other probability
distributions and other ensembles will be needed, but for most situations the
canonical or grand canonical ensemble is sufficient. Before we proceed it is
useful to state some general properties of the density matrix, which follow
directly from the general definition (10.8) and hold in all ensembles:
Trp = 1, (10.33)
since the trace of the exponential operators in (10.24) or (10.31) exactly can-
cels the c number in the denominator, which is given explicitly by (10.25) or
(10.32), respectively.
pt = p, (10.34)
i.e. the density matrix is an Hermitean operator, which follows from the fact
that p is a sum of Hermitean projection operators, (10.8), where the weights,
pn, are real probabilities.
p2 = P <=> P = Itj/~(~,N») (tj/~(~,N) I , (10.35)
i.e. if p2 = jJ then jJ describes a pure state of a
system in an exactly known
microstate (here state number n, and the system has exactly N particles). In
this case p is a projector onto this microstate, otherwise jJ2 # p. Finally we
remember that averages of observables have to be calculated as
(0) = Tr(pO) = Tr(Op) , (10.36)
where the cyclic invariance under the trace may be used. Note that
this average contains a purely quantum-mechanical average such as
(tj/~(~.N) 101tj/~(~,N») and additionally a statistical average over all these quan-
tities weighted with the probabilities, pn,N, according to our starting point
(10.5).
We can now write the entropy according to (10.21) as the average of
-klnjJ, i.e. 4
s (-klnp) = -kTr(jJlnp)
-k L(tj/~(n)ljJlnpltj/~(n») , (10.37)
n

where the logarithm operator is to be understood as a series expansion. How-

ever, we can use

4 The expression originates from ideas of L.E. Boltzmann and M. Planck; the gen-
eral interpretation in terms of information theory has been formulated by C.
Shannon and E.T. Jaynes (see Example 6.9).
264 10. Basics of Quantum Statistics

(10.38)
i.e. the states I!IiE(n)) are eigenstates of the density operator and the proba-
bilities pn are its eigenvalues. Therefore the matrix element in (10.37) can be
evaluated to be the number pn In pn such that (10.37) is only a more com-
plicated formulation of (10.21). This formulation has the advantage that it is
representation free and holds for all kinds of ensembles. Special forms of the
density operator and the density matrix will be exemplified in the following
problems. However, here we note some of the results here. Expressed by field
operators tf,t (x) and tf,( x) the density operator takes on the form
jJ(x) = tf,t(x)tf,(x) . (10.39)
Analogously, the operator for the density matrix reads
jJ(x,x') = tf,t(x')tf,(x). (10.40)

EXAMPLE

10.1 Density Operators in Second Quantization

The starting point for our consideration is an element 1!Ii) of the N-particle
Hilbert space. In order to represent this rather abstract state vector we need
to have a complete basis of the Fock space. For this, arbitrary one-particle
states 1,8) can be used; we assume that they are eigenstates of an operator
B. ,8 stands for a corresponding set of quantum numbers. Out of states 1,8)
an arbitrary n-particle state can be constructed by taking direct products:
symmetric for bosons and antisymmetric for fermions. It is convenient to
know about the "unity operator" i in Fock space:
(1)

The various terms express the contributions from the ground state, the one-
particle, two-particle, ... states to the complete 1. They represent symmetric
or antisymmetric many-particle states for bosons or fermions, respectively.
An explicit distinction becomes necessary only once commutation relations
are derived; see Exercise 10.3.
We project the state 1!Ii) onto the Fock space i:
1 !Ii) i !Ii)
1

10)(01!Ii) +L 1,81) (,81 1!Ii) +L 1,81,82)(,81,821!Ii)

(3,

+".+ L 1,81·",8N)(,81,,·,8NI!Ii)· (2)

(31 · .. (3N
 10.2 Density Operator of a Many-Particle State 265

The quantum numbers 13 stand, for example, for momentum or position quan- Example 10.1.
tum numbers. The expansion (2) reduces to exactly one term for a pure n-
particle state: 1131··· 13n) = ItJi n). The quantities 1(131··· 13nltJi)j2 represent the
probabilities for the realization of a n-particle state with quantum numbers
131· . ·13n· Creation and annihilation operators b1 and b(3 are defined via their
action on a many-particle state, i.e.
b11 0 ) = 113), b(310) = 0, (3)

b1,113) = 113'13), b(3' 113') = 10), (4)

and so forth. The operator b1 transforms an (n-1)-particle state into an n-
particle state; the opposite holds for the adjoint operator b(3. This leads to
the interpretation in terms of creation and annihilation operators. Applying
b1 onto the Fock space i leads with the help of (3), (4), and (2) to the
decomposition
b(3 b(31
~t ~t A

= 113) (01 + L 113131) (1311 + L 113131132) (1311321 + ... ; (5)

for the adjunct operator b(3 we get

b(3 b(3 i
= 10)(131 + L 1131)(131131 + L 1131132)(131132131 + .... (6)
(3, (3, (32
A special class of creation and annihilation operators is represented by the
field operators ~t(x) and ~(x). They are already known from (3.43a,b). In
analogy to (5) and (6) they can be defined via an expansion with respect to
the complete Fock-space basis of position states:

~t(x) = Ix)(OI + J d3x1lxxl)(xll

+ J d3xl d3x2Ixxlx2)(xlx21··· (7)

~(x) 10)(xl + J d3xllxl)(xlxl

+ J d3xl d3x2Ixlx2)(xlx2xl··· . (8)

These operators create and annihilate a quantum at position x. They obey the
commutation relation (3.44), i.e. [~(x), ~t (x')]± = 8(x - x'). The operators
b1, b(3 and ~t (x), ~(x) are related via a unitary transformation of the form
~t(x) = Lb1(13lx) == L b1tJih(x), (9)
(3 (3

~(x) = L b(3(xl13) == L b(3tJi(3(x) ; (10)

(3 (3
see (3.43a,b) and Exercise 10.2.
266 10. Basics of Quantum Statistics

Example 10.1. It is possible to define the density operator with the help of the field
operators (9) and (10). With (7) and (8) we get
p(x) tPt(x)tP(x)

Ix)(xi + J d3xllxXl)(XlXI

+ Jd3xl d3x2IxXIX2)(X2XIXI + ... (11)

Again, the various contributions result from the one-particle, two-particle, ... ,
n-particle subspace of the Fock space. The one-particle density describes the
probability of finding a particle in an n-particle state Iwn) at position x and
is represented by the matrix element
(w n Ip( x) Iwn)

J d3x2 ... d 3x n (wn lxx2 ... Xn )(XX2 ... xnlwn)

nJ d3x2 ... d3xnwt (x, X2, ... ,xn)w(x, X2,· .. ,xn ) . (12)

Here, the density has been normalized to the particle number n.

Equation (12) can be written in another, equivalent form, namely

PllPn)(X) = J d3xl .. . d3 xnwt(XIX2·· ·xn) (~b(X - Xi))

x W(Xl' X2··· xn) . (13)

In this form (12) and (13) are valid for fermions and bosons. From (13) we
read off the local density operator in coordinate representation:
n
PllPn) (x) = Lb(X-Xi). (14)
i=l

In view of (12)-(14) it becomes clear that the operator p(x) in (11) acts on
n-particle Hilbert spaces. We also stress that the form of the density operator
(11) does not imply a space representation of p. The special appearance of
the density operator comes from its expression in terms of field operators.
The arguments x have to be viewed merely as parameters (position quan-
tum numbers) which just characterize this class of creation and annihilation
operators.
In order to make this point even clearer we discuss another form of the
density operator. According to the unitary transformation (9) and (10) this
leads to
p(X) = Lb~b{3(alx)(xl;3). (15)
a,{3
Let's take the expectation value in analogy to (12):

PllPn) (x) = L(wnlb~b{3lwn)wl(x)w{3(x) L na{3wl(x)w{3(x). (16)

a{3 ot{3
 10.2 Density Operator of a Many-Particle State 267

Due to the hermiticity of the matrix na(3 a unitary transformation (j can be Example 10.1.
found which diagonalizes (na(3):

The diagonal elements result to be

This suggests the introduction of "rotated" creation and annihilation opera-

tors, i.e.

1/

and (16) takes on diagonal form:

Pll[<n) (x) = L nalj/l(x )lj/a(x)

(xl L naIO')(O'lx) .
Then, the density operator becomes
(17)

A comparison with (12) yields the identity

(Ij/nlp(x)llj/n) = (xlnl[<n Ix) . (18)
The numbers na have a very simple meaning: they are the occupation num-
bers, or occupation probabilities, for the one-particle state 10') in the many-
particle state Ilj/n). As a distinction to the form (11) the density operator (17)
acts in the one-particle Hilbert space.
Now we focus on the density matrix. One of its forms is defined as
p(x, Xl) = Jjjt (XI)Jjj(X) . (19)
Forming the matrix element with respect to an n-particle state Ilj/n) we are
led to the density matrix normalized to n:
(20)

J d3x2 ... d3 x n (lj/n IXI X2 ... x n) (XX2 ... Xn Ilj/n)

nJ d3x2 ... d3 x nlj/t (Xl, X2,.··, xn)lj/(x, X2,···, x n) .

268 10. Basics of Quantum Statistics

Example 10.1. The one-particle density PI<pn) (x) in the n-particle state from (12) results as a
special case (diagonal element): PI<pn) (x) = PI<pn)(x,x). From (20) we deduce
the spatial representation for the nonlocal operator of the density matrix
[compare with (13)]:
n
P<p n (x, x') = L 8(x' ~ xi)8(x ~ Xi) . (21)
;=1
As in the case of the density operator the form

p(x, x') = L b~b/3(alx') (x 1;3) (22)

ex/3
can be taken as a basis. It follows after insertion of (9) and (10) into (19).
The relation
PI<pn)(x, x') = L(tJrnlb~b{3ltJrn)(alx')(xl;3)
ex/3
L (xl;3)nex/3(alx') (23)
ex/3
leads to
L 1;3)nex{3(al . (24)
ex/3
Obviously, the same one-particle operator is assigned to both the density and
the density matrix [compare (24) with (16)]. According to (23) for the density
matrix the nondiagonal matrix elements nex/3 of nl<pn) are essential:

PI<pn)(x,x') = (x'lnW,)lx) = Lnex/3 tJrl(x')tJr/3(x). (25)

ex/3

EXERCISE

10.2 Transformation Equations for Field Operators

Problem. Derive the equations for transformation between the operators

b1,b/3 and the field operators pt(x),P(x).

Solution. We begin with the transformation equations for the operators b1

and pt (x). Making use of the Fock space i,
i = 10)(01 + L 1;3)(;31 + L 1;3;31)(;31;31 + ... ,
/3 (3/3,

and the expansion (7) for the field operator pt(x) (see Example 10.1) we find
pt(x) Ipt(x)

= (10)(01 + L 1;3)(;31 + L 1;3;31)(;31;31 + ... )

/3 /3/3,
 10.2 Density Operator of a Many-Particle State 269

x (IX)(OI + J d3X1IxX1)(X11
Exercise 10.2.

+ J d3X1 d3X2IxX1X2)(X2X11 + ... ) .

This yields

~t(X) = L 1,6) (,6lx)(OI

{3

+L J d3x11,6,61)(,6lx)(,61Ix1)(X11

J
(3{3,

+ L d3x1 d3x2 1,6,61,62) (,61 X)

{3{3,{32

x (,61,62Ix1X2)(X1X21 + ... ;
all the other matrix elements like (,610) and (,6,61Ix) vanish. Only the matrix
elements between states having the same number of particles are different from
zero. Keeping this in mind, we easily understand that the equation above can
also be written as

~t(X) = L(,6lx) (1,6)(0 1

(3

+L 1,6,61)(,611 +L 1,6,61,62)(,62,611 + ... )

J
{3, {3,{32

x (10)(01 + d3X1Ix1)(X11

+ J d3x1 d3X2Ix1X2)(X2X11 + ... ) .

Remembering (5) of Example 10.1, we identify the expression within the first
bracket as the decomposition of the operator b1; the second term yields just
the i [see (1) in Example 10.1]. Thus, we obtain the wanted transformation
from b1 to ~t(x):

~t (x) = L b1(,6lx) . (1)

(3

For the inversion of (1) it is necessary to project with (xl,6):

Jd3x~t(x)(xl,6) J = d3 x L b1, (,61Ix)(xl,6)

{3,

L b1, (,611,6) = b1;

(3,

this leads to

b1 = Jd3x~t(x)(xl,6)· (2)
270 10. Basics of Quantum Statistics

Exercise 10.2. The corresponding transformation equations for the annihilation operators
are obtained by taking the adjoint of (1) and (2):
ir(x) = 2:)(3(xl(3), (3)
(3

b(3 = J d3 x ir(x) ((3lx) . (4)

EXERCISE

10.3 Commutation Relations for Fermion Field Operators

Problem. Derive the commutation relations between fermion creation and

annihilation operators.

Solution. For the case of fermions we have to take the Fock space of the
antisymmetric many-particle states as a basis. The action of the antisym-
metrization opemtor A on a one-particle product state 1(31(32 ... (3N) creates
a completely antisymmetric many-particle state normalized to the particle
number N:
VNlAI(31) 1(32) .. ·1(3N)
VNlAI(31(32' .. (3N) , (1)
where the index 'a' symbolizes antisymmetry. We find for the matrix element
between two arbitrary many-particle states:
a ((3~(3~ ... (3'tv, 1(31(32 ... (3N) a = 0 for N' =J N (2)
and

for N = N'. (3)

8(3',v(31 8f:J',vf:J2 8f:J',vf:JN

The unity operator in the Fock space of the antisymmetric states is given by

i = 10)(01 + L 1(31)((311 + L ~1(31(32)aa((31(321 +... (4)

(31 (31 (32
from which the expressions for the fermion creation and annihilation operators
follow. We get for the creation operator
b1 = b1 i = 1(3)(01 + L 1(3(3~)a ((3~1
(3;

+ L ~ 1(3(3~(3~)aa((3~(3;1 + ... (5)

(3;f:J~
 10.2 Density Operator of a Many-Particle State 271

and for the annihilation operator Exercise 10.3.

b13 = 10)(fJl + L lfil) a(fJfil 1
/31

+ L ~ Ifilfi2)aa(fJfilfi21 + .... (6)

fh/32
Here we have used the idea that the action of the operator b1 on an antisym-
metric N-particle state leads to an antisymmetric (N + I)-particle state, Le.

(7)
In order to derive commutation relations we determine the operator products
b!3'b1 and b1b13' according to (5) and (6). Let us have an explicit look on the
first terms:
(8)
131/31
+ 2;: _(2~)2Ifilfi2)a a(fJ' filfi2IfJfJ~fJ~)a a(fJ~fJ~1 + ... . (9)
13;13;131132
The appearing matrix elements have to be analyzed according to (3) and yield

a(fJ'fillfJfJ~)a = 813'13 8!3113; - 813'13;8!3113

and

a(fJ'filfi2IfJfJ~fJ~)a = 813'13(8/3113; 8!3213; - 8!3113; 8(3213)

- 813'13; (8!3113 8!3213; - 8/3,13; 8/3213)
+ 813'13; (8/3113 8/3213; - 8!3,13; 8(3213) .
Inserting this into (8) and performing the summation leads us to

b13,b1 813'131 0)(01 + 8!3'13 L IfJ~)(fJ~I-lfJ)(fJ'l

+ 8!3'13 L (2~)2 (lfJifJ~)a a(fJ~fJ~1 - IfJ~fJi)a a(fJ~fJ~l)

13; 13;

- ~ (2~)2 (lfJfJ~)aa(fJ'fJ~I-lfJ~fJ)aa(fJ'fJ~l)

+~ (2~)2 (lfJfJ~)a a(fJifJ'l - IfJ~fJ)a a(fJ~fJ'l) + ...

8~'13 (10)(01 + L IfJi)(fJ~1 + L ~lfJ~fJ~)a a(fJ~fJ~1 + ...)

13; 13; 13;
-lfJ)(fJ'l- L IfJfJ~)aa(fJ'fJ~I-'" .
13;
272 10. Basics of Quantum Statistics

Exercise 10.3. Calculating the next higher-order terms in (8) we convince ourselves that
those terms proportional to the factor bf3'f3 add up to the Fock space i [see
(4)].
We obtain
bf3 ,b1 = bf3f3,i -1,6)(,6'1- LI,6,6Daa(,6',6~1
f3;
- (more remaining terms) , (10)
where the remaining terms in the operator product b1bf3' originate with op-
posite sign. Furthermore,

b1bf3' = 1,6)(010)(,6'1 +L 1,6,6~)a (,6~1~1)a (,6',811 + ...

f3;fh
1,6) (,6'1 +L 1,6,6~) a a(,6' ,6~ I + (more remaining terms) . (11 )
f3;
Thus, with (10) and (11) we are led to the well-known commutation relations
for fermions:
(12)

10.3 Dynamics of a Quantum-Statistical Ensemble

The dynamics, i.e. the time evolution, of a quantum mixture is determined

by the Hamiltonian of the system. If no measurements are performed on the
system over an arbitrary time interval the information about the system,
which is reflected in the weights pn, does not change. This implies that the
pn remain constant over the corresponding time interval, i.e.

(10.41 )

The dynamics of each (pure) state ItPn) of the ensemble is described by the
Schrodinger equation (Schrodinger picture)
(10.42)
Here, we have set It = 1, which is the same as replacing II by II lit: i.e.
II -; II lit. As we ask for the evolution of the whole ensemble we have to find
the evolution equation for the density operator. Directly from the Schrodinger
equation (10.42) and its adjoint equation it follows that the density operator
of the pure state (10.3) is
(iOt ItPn) )(tPn I (1IItPn))(tPnl,
ItPn)( -iOt (tPn I) ItPn)((tPnllI) ,
so that
 10.3 Dynamics of a Quantum-Statistical Ensemble 273

Wt (I!p"n) (!p"n I)
i[(atl!p"n»)(!p"nl + I!p"n)(at(!p"nl)l
(HI!p"n»)(!p"nl-l!p"n)(H(!p"nl)
HI!p"n)(!p"nl-l!p"n)(!p"nIH
[HI!p"n) (!p"n Il-
and therefore
(10.43)
Since the weights pn are constants, the ensemble averaging (10.8) over (10.43)
can be performed directly:

LpnWtPI<lm) = Lpn[H,PI.pn)l·
nEE nEE
Finally we arrive with (10.8) at
WtPE = [H, PEl. (10.44)
Hence, we have derived an evolution equation for the density operator of
the quantum-theoretical mixture. It can be understood as the quantum-
mechanical analogue to Liouville's equation of classical many-particle mechan-
ics.
With the help of the evolution equation for the density operator we succeed
in deriving the equation for the time evolution of an arbitrary observable. 0
denotes the Schrodinger operator, which is assigned to the observable (O)E
and does not explicitly depend on time, i.e. dO / dt = aD/at = o. We consider
now the change with time of the ensemble average (10.7)
(O)E = Tr(pEO) ,
and indicate some of the intermediate steps. First,

mEE
+(!p"mlpEo(atl!p"m»)] .
Making use of the Schrodinger equation (10.42) and the evolution equation
(10.44) we deduce

at(O)E = L -i[ - (!p"ml[H,PEOll!p"m) + (!p"ml[H,PElOI!p"m)]

mEE
L -i(!p"mlpE[H, Oll!p"m) .
mEE
This leads us to the evolution equation for the averages:
at(O)E = -iTr(pE[O,HJ). (10.45)
274 10. Basics of Quantum Statistics

EXERCISE

10.4 Density Operator of a Mixture

Problem. Determine the form of the density operator of a mixture for the
case in which information is given about its pure states.

Solution. We assume that the ensemble consists of a finite number N of

pure states. If we have no information about the pure state of the system we
have to proceed in such a way that all possible state vectors l\]in) are realized
with the same probability. This implies that all weights have to be equal, i.e.
pn = p. This leads us to the following form of j):
N
fJ = p L I\]in) (\]in I ; (1)
n=l

it is the sum over projectors onto the pure states of the ensemble.
We can also provide another ansatz for the density operator, which is
equivalent to (1). For this we only have to take into account that on the one
hand the sum occuring in (1) represents a part of the complete unity operator
in Fock space and that on the other hand all expectation values always imply
ensemble averaging, i.e. taking the trace with the density operator. We write
(1) as
p = pi - pQ, (2)
where the operator Q simply represents the remaining sum over all projectors
onto pure states, which do not belong to the ensemble. Thus, it holds that
for nEE. (3)
Once the action of Q onto one l\]in) from the ensemble vanishes, this part in
the density operator does not contribute in all possible trace expressions that
contain p. In this respect, the form of the density operator equivalent to (1)
is simply given by
pi = pi. (4)
We still have to determine the weight p. It follows in a straightforward manner
from the general property Tr(fJ) = 1 of density operators. We arrive at the
result

pi i i (5)
Tr(i) N'
where N reflects the number of possible pure states of the ensemble.
With absolutely no information about the pure state of the system the
density operator can be set proportional to the unity operator.
 10.3 Dynamics of a Quantum-Statistical Ensemble 275

EXERCISE

10.5 Construction of the Density Operator for a System

of U npolarized Electrons

Problem. Construct the density operator for a system of unpolarized elec-

trons by using the spin eigenstates of (j z.

Solution. Only the spin part of a possible pure many-electron state is of im-
portance. The pure spin states of the system are the two possible eigenspinors
of the operator (j z. The two spinors

(1)

form the ensemble. For the construction of the density operator we need a
representation of the projectors I!lin) (!lin I onto the pure spinstates Xl±l in
2 2
first place. They result as dyadic products:

F± = 4±! 0X!±!, (2)

or more explicitly:

(1, 0) (~ )
(3)
(0, 1) (~)
Hence, we obtain the form
p = p+F+ + p_F_ (4)
for the density operator. Now we have to determine the weights p+ and p_,
i.e. the probabilities for the realization of a pure state with a z projection of
the spins of + ~ and - ~, respectively. In general it has to hold that

(5)
ms

In more detail this becomes

Tr(p) = (1, 0) [p+ (~ ~) + p_ (~ ~)] (~)

+ (0, 1) [p+ (~ ~) + p_ (~ ~)] (~) ,

which leads to the condition

(6)
However, in an unpolarized many-electron system no spin projection is fa-
vored, so the probabilities p+ and p_ are equal. As a result of (6), (4) takes
on the form
276 10. Basics of Quantum Statistics

Exercise 10.5. ,1, 1, 1(10 ) __ i (7)

P = "2 P+ + "2 P - ="20 1 Tr(i) ,

note that in our case Tr(i) = N = 2.

This result demonstrates that an unpolarized electron system represents
the physical realization of an ensemble for which we have no knowledge about
its pure state. p+ = p_ reflects the absence of information.

10.4 Ordered and Disordered Systems:

The Density Operator and Entropy

In the preceding subsections we have learned how the existing lack of knowl-
edge about the state vector Itli) is reflected in the density operator for the
description of a state of a physical many-particle system. We have also real-
ized that the uncertainty determined by Pis restrained from below and above
by two, so-to-say, extreme density operators. The lower bound is fixed by the
density operator of the pure state:
(10.46)
This implies that the state of the system is entirely known. On the other hand
we have experienced the density operator
, i i
(10.47)
pu = Tr(i) = N'
which has to be chosen when there is full ignorance about the state of the
system (see Exercise 10.4 and 10.5); the index 'U' stands for 'unawareness'.
Often it is important to possess a meaningful measure for the 'deviation'
of the density operator fJ from its pure state PI '/I). This measure can also
be understood as a measure for the missing information about the system. In
accordance with information theory this measure, which we will simply denote
as deviation in the following and which is assumed to be a positive number
a, is defined as a functional a[fJ] of the density operator in the form
alp] = -Tr[pln(fJ)]· (10.48)
The deviation alfJ] can also be understood as a measure for the disorder of the
system. Here we will not justify (10.48) from the information theory point of
view and refer to Example 6.9 instead, but we will straightforwardly prove the
most important properties of a[fJ]. We proceed with the eigen representation
of the density operator:
fJltlin) = pnltlin) with Lpn = 1 and 0 <::;; pn <::;; 1.
n

Within this representation (10.48) becomes:

(10.49)
n
 10.4 Ordered and Disordered Systems: The Density Operator and Entropy 277

compare with (12) of Example 6.9. Since 0 :::; pn :::; 1 holds for all weights, we
deduce at first positive definiteness:
o-[p] 2: o. (10.50)
Next we show that the measures 0" [Pu] and O"[PIIP)] corresponding to the density
operators Pu and PIIP) are indeed upper and lower bounds for every other O"[p],
so that
(10.51)
Once all weights pn except one (p' = p = 1) are set equal to zero we obtain
from the eigen representation (10.49) the density operator PIIP) of a pure state.
The limit limx_o x In(x) = 0 leads to
(10.52)
As it should do, the deviation 0" vanishes exactly for the density operator of
the pure state since it implies a complete knowledge about the state vector of
the system. In order to determine the upper bound for 0" we have to calculate
O"[pu]. The density operator pu originates from (10.49) in such a way that we
set all weights equal to pn = p = liN. This yields

o-[pu] = - ~
N 1In (1)
N N = - NIn N~ 1
lIN

1
-N(1n1-lnN)N = In(N). (10.53)

Hence, the claim (10.51) explicitly reads

o :::; O"[p] :::; In(N). (10.54)
This relation can be proven by rewriting the expression (10.49) for o-[p] in the
following form:

O"[p] = Lpn In (;n) = Lpn In (p~N)

n n

In(N) Lpn + Lpnln (P~N)

n n

< In(N)+Lpn(P~N-1) = In(N) ,

n
where we have used the relation In(x) :::; x -1. This is the proof of statement
(10.54).
In Sect. 10.3 we have learned that the time evolution of an isolated system
is only governed by its corresponding Hamiltonian iI and that for this case
the weights pn are constant [see (10.41)]. Hence, also the deviation 0" [see
(10.49)] remains constant:

:t O"[p] = O. (10.55)
We should mention another aspect here. The deviation 0" also represents a
measure for the disorder within a system. This becomes plausible once we
278 10. Basics of Quantum Statistics

remember, for instance, the example of the unpolarized electron system (Ex-
ercise 10.5). For this system the density operator was Pu = i/ Tr(i), which led
to the largest deviation IJ = In(2). Unpolarization means arbitrary orientation
of the single spins. With respect to spin orientation the system is completely
disordered. However, once a certain spin orientation is favored (magnetiza-
tion) this reflects a state of less disorder. Hence, the deviation IJ also becomes
smaller:
IJ = -[(1 - p) In(l - p) + pln(p)] :::: In(2), p :::: 1/2,
where p represents the weight of the less favored spin orientation. Once the
system is in a pure state, i.e. all spins are oriented in the same direction
(complete magnetization), IJ vanishes. Then the system is in a state of largest
order.
Another conclusion is at hand: (10.55) implies that for the case of an
isolated system, which is left to its own, the degree of disorder does not change.
Detailed studies show that the deviation IJ increases because of the influence
of external stochastic perturbations. The measure IJ behaves like an order
parameter. In fact, IJ is related to the thermodynamic entropy. The quantum-
mechanical definition of entropy is given by
S = kBIJ, (10.56)
where kB denotes Boltzmann's constant (see Examples 6.6 and 6.9). From
experience we can say that a system always tends to a state of maximum
entropy. In other words, the density operator p of a quantum-mechanical
mixture always maximizes the deviation dp]. In the next subsection we will
consider the extremal properties of IJ to pin down the density operators for
stationary ensembles. For further information on these subjects we refer the
reader to the lectures on thermodynamics and statistical mechanics. 5

10.5 Stationary Ensembles

According to (10.44) the vanishing of the commutator between the density

operator and the Hamiltonian,
[p,H]_ = 0, (10.57)
is a necessary and sufficient condition for the presence of a stationary ensem-
ble. For example, the density operator pu (10.47) of a completely disordered
system is stationary because the unity operator commutes with every ar-
bitrary Hamiltonian. A sufficient condition for stationarity is given once p
becomes a function of the Hamiltonian or a function of the operators 0 that
commute with H:
p = p(H,O), [O,H]_ = O. (10.58)

5 W. Greiner, L. Neise, H. Stocker: Thermodynamics and Statistical Mechanics,

2nd ed. (Springer, New York 1997).
 10.5 Stationary Ensembles 279

Of particular importance are stationary ensembles that maximize the devi-

ation a; examples are the operator pu and the generalized grand canonical
ensembles. They are determined from the variational principle
6a[p] = 6 Tr[pln(p)] = 0, (10.59a)
with the single restriction
Tr(p) = 1, (10.59b)
i.e.
6 {Tr[p In(p)] + A Tr(p)} = O. (1O.59c)
A represents a Lagrange multiplier. The consideration of fixed expectation
values for certain observables Oi (additional information about the system),
such as
(10.60)
leads to further constraints for the variation (1O.59a-c); with Lagrange mul-
tipliers Ai the variational equation then becomes

6 [Tr[p In(p) 1+ A Tr(p) +L Ai Tr(PO;)] = 0 (10.61)

t

and we arrive at various grand canonical ensembles. If we assume an eigen

representation of p, i.e.
(10.62)

the general variational equation (10.61) reads

L 6pnn (In(pnn) + 1 + A + L AiGin) = 0, (10.63)

n t

where Gin = (y;nIOily;n). Since the variations 6pnn fulfill only the condition

L6 pnn = 0,
n

and are otherwise arbitrary, the relation (10.63) can be fulfilled only once
every square bracket vanishes for its own. This immediately leads to

pnn = exp [- ( A +L AiOin)] = ~ exp ( - L AiOin) (10.64)

t t

and to the density operator

(10.65a)

Here we have formally set 1/ Z = e - A. Clearly A and therefore also Z are

normalization factors. The normalization factor Z is known as the partition
function and follows from Tr(p) = 1:
280 10. Basics of Quantum Statistics

(1O.65b)

The Lagrange parameters Ai are denoted as fugacities; their physical inter-

pretation is coupled to the corresponding observables Oi.
Finally, we consider two important special cases of (10.63). The canonical
ensemble maximizes a with the additional constraint that the average energy
(H)E = E of the ensemble possesses a given fixed value. It is determined by
80' = 8 Tr[pln(p)] = 0 (1O.66a)
and the conditions
Tr(p) = 1, Tr(pH) E. (10.66b)
According to (1O.65a-b) the canonical density operator becomes

.ok -1e -[3H (1O.67a)

Zk '
with
1
Zk Tr(e-[3H) , (J = kT'
(10.67b)

In addition to the mean energy the mean particle number (N)E = N is also
kept fixed for the so-called grand canonical ensemble. The corresponding vari-
ational equations
80'[.0] = 8 Tr[pln(p)] = 0 (10.68a)
together with the conditions
Tr(p) = 1, E, (1O.68b)
lead to the grand canonical density operator

PGk = 1 exp [-(J(H

-Z " - J.LN) ] (10.69a)
Gk
with the partition function
ZGk = Tr { exp [-(J(H - J.LN)]} . (1O.69b)
The parameter J.L is called the chemical potential. This should do for some
of the most important ensembles. In Chap. 6, about quantum gases, we have
already discussed one application of the grand canonical density operator.
For further applications of these methods we refer to Thermodynamics and
Statistical Mechanics. 6

6 W. Greiner, L. Neise, H. Stocker: Thermodynamics and Statistical Mechanics,

2nd ed. (Springer, New York 1997).
 10.5 Stationary Ensembles 281

EXAMPLE

10.6 Systems of Noninteracting Fermions and Bosons

We consider a canonical ensemble of free noninteracting fermions and bosons.

For each case we will determine the partition function and the mean occupa-
tion numbers. The latter will lead us to the Fermi-Dirac and Bose-Einstein
statistics.
We use the occupation-number representation to express the Hamiltonian
if and the particle-number operator N:
(1)

and
(2)

respectively. na represents the number operator for particles in the quantum

state a, where a stands for a characteristic set of quantum numbers such as
momentum k and spin projection 0'. Furthermore, ta denotes the correspond-
ing one-particle energy. Pauli's principle means that each one-particle state
for fermions can be occupied at most only once, i.e.
na = 0,1 (fermions) , (3)
whereas for bosons each occupation number can be a natural number, i.e.
na = 0,1,2, ... (bosons) . (4)
The grand canonical density operator takes the form

PGk = _1_ exp[ -t3(if - /-IN)] = Zl exp (-t3 :L(t a - /-l)n a ). (5)
ZGk Gk a

At first we calculate the grand canonical partition function:

ZGk = Tr { exp [- t3 (if - /-IN)] } . (6)
Ina! n a2 ... n,,; ... ) represents an arbitrary many-particle state; then, the
trace-operation implies:
(n cq n 0:2 ... n 0: 1 .. ·1 exp [- J J,-
r:I(if . . liN)]
. Ina1 n 0:2 ... n Qi ... )
{-"n u ," }

II exp [-t3(tct; - /-l)n".] , (7)

where the sum runs over all possible configurations of occupation numbers
n a ,. The ith one-particle state may have quantum numbers ai' For a general
many-particle state everyone-particle state contained in it appears with occu-
pation numbers according to (3) or (4). Hence, we are allowed to interchange
the summation with the product series: first we sum over possible individual
282 10. Basics of Quantum Statistics

Example 10.6. occupation numbers n and then we take the product over all allowed sets of
quantum numbers ai. We get:
(8)
<> n
For fermions we arrive at
zb = II {I + exp [-,8(10<> -IL)]} (9)

and for bosons we get

Z3k = II {I - exp [-,8(10<> -1L)]}-1 (10)

<>
Let us now calculate the mean particle number

n" = Tr(n"PGk) (11)

for a system of independent fermions and bosons in thermodynamic equilib-
rium in the state with energy E". The corresponding grand canonical ensemble
comes with a mean energy
E = Tr(H PGk)
and a mean particle number
N = Tr(NPGk) .
The trace operation (11) implies that

n" = Z
1
Gk
L n" exp [-,8(E" -1L)n,,]
{".nV···nO i "·}

x II exp [-,8(Eai -1L)n a.] (12)

aiel"
As with the determination of the partition function we interchange the prod-
uct series with the summation:

n" = Z~k [L nexp [-,8(10" - lL)n]] II L exp [-,8(Eai -1L)n]. (13)

n aiel" n
We insert the partition function according to (8); then all factors in this
equation containing an ai -I- 1/ drop out and
2::n nexp [-,8((" -1L)n] (14)
2::n exp [-,8(E" -1L)n] .
Here we have to distinguish between fermions and bosons. For fermionic sys-
tems, (14) takes on the form
1
(15)
exp [(,8" - IL)] +1 .
This is the so-called Fermi distribution.
 10.5 Stationary Ensembles 283

For the case of bosons we obtain from (14) Example 10.6.

-b L:~=o nexp [-,8((", - p,)n]
n", L:~=o exp [-,8((", - p,)n]

Summation of the geometric series yields the Bose~Einstein distribution:

-b 1
(16)
n", = exp [,8((", - p,)]-1 .
The temperature and chemical potential appearing in the two distributions
(15) and (16) are determined by the fixing of the average total energy

and the average particle number

'"
of the ensemble considered.
11. Structure of Atoms

In this chapter we discuss several approaches to the calculation of properties

of atoms. Exact calculations are extremely difficult. This is because of the re-
pulsive Coulomb interaction between electrons and the spin-orbit interaction
of the electrons, which results from the interaction of the electron spin with
the electric field of the atom.
First we discuss atoms with two electrons, which represent the simplest
system next to the hydrogen atom. Then we will consider systems with
many electrons and introduce different approaches (Hartree, Hartree-Fock,
and Thomas-Fermi).

11.1 Atoms with Two Electrons

For the case of light atoms, the spin-orbit coupling can be neglected to a
good degree. This leads to a considerable simplification. Typical atoms with
two elecrons are for example He, H- , Li+, and Be++. Nowadays it is possible
to completely strip heavy ions, e.g. uranium, of all electrons in heavy-ion
accelerators. Once uranium is shot through a thin foil with an energy larger
than about 500 MeV per nucleon, i.e. with a total energy of more than 235 x
500 MeV = 117.5 GeV, it loses most ofits electrons. In this way it is possible to
produce a heavy ion with no, one, two, or more electrons with a nonnegligible
probability. This offers the opportunity to study systems with few electrons
in the strong central fields of heavy ions experimentally and, in particular, to
check the effects of quantum electrodynamics in these systems (self-energy,
vacuum polarization). The Schrodinger-type Hamiltonian for a light atom
with two electrons is given by

H =
pi + -p~ - -Ze 2 - -Ze 2 +
-
e2
. (11.1)
2m 2m Tl T2 ITl - T21

The first two terms in (11.1) describe the kinetic energy of the two electrons
with mass m. The next two terms represent the attractive interaction between
the electrons and the nucleus. Z is the nuclear charge and Tl and T2 represent
the position vectors of the electrons from the nucleus (see Fig. 11.1). The last
term appears because of the repulsive Coulomb interaction between the two
electrons. The kinetic energy of the nucleus is neglected in (11.1), because
the nucleus is assumed to stay motionless in space as a result of its large
mass. The Hamiltonian H is supplemented by Ha , indicating the spin of the

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
286 11. Structure of Atoms

electrons. Because neither spin-orbit or spin-spin interactions are important

for light atoms, they will be neglected here. Nevertheless we always have
to remember that the total wavefunction for the two electron system also
contains a spin wavefunction. It is important for the exchange symmetry of
the total wavefunction, as we shall see soon.
Because of the electron-electron interaction in (11.1) it is impossible to
solve the problem of the atom with two electrons exactly. Therefore we have
Z to apply an approximation scheme, which will be discussed in the follow-
Fig. 11.1. An atom with ing. In the first place the electron-electron interaction is neglected. This ap-
two electrons. For simplic- proximation allows the problem to be solved. Afterwards the influence of the
ity the nucleus is assumed
electron-electron interaction [the last term in (11.1)] is treated in the frame-
to be large and at rest
work of perturbation theory. In zeroth order the electrons feel only the usual
Coulomb potential -Ze 2 /r. Then the lowest-energy state the two electrons
occupy is the Is state. The wavefunction then reads
(11.2)
<PIs describes the Is state and xtn is the spin wavefunction of the two electrons,
which are coupled to the spin j with projection m. The spatial wavefunction
<Pls(r) is given by
1
<PIs (r) = y7r
(Z)3/2
ao e- Zr / ao . (11.3)

ao = ;,,2/me 2 is Bohr's atomic radius. Because the spatial part of the wave-
function is symmetric under exchange of the electron indices, the spin part
has to be antisymmetric. This can be fulfilled only by the singlet state

x8 = ~ [xr (1) XL (2) - XL (1) Xr (2)] . (11.4)

The antisymmetry of the total wavefunction against electron exchange ensures

the Pauli principle is fulfilled. The energy of an electron in a Coulomb potential
is given by
(Ze)2
---
2ao 2
For two electrons it is

Eg = ---
Z2 2
e
(11.5)
ao
The superscript 0 refers to the zeroth order of perturbation theory. The sub-
script 0 reflects the ground state.
Now we will calculate the perturbative contribution of the electron-
electron interaction. In first order the change in the energy is
e2
D.E = ("pI I 11"p) , (11.6)
TI - T2

where"p is given by (11.2). Inserting (11.3) into (11.6) leads to

 11.1 Atoms with Two Electrons 287

/j,E J
d3rl d3T2 [_1_ ( Z) ~l2 e-Z(rl +r2)/ao e2
Vii ao
T
ITl - T21

X [;. (!) e-Z(c,+c,l/oo

J
d3rl d3T2 ~ (~)6 e2 e-2Z(r,+r2)/aO. (11.7)
7f2 ao ITl - T21
To solve the integrals, we use the Fourier representation of 1/1Tl - T21 (see
Exercise 11.1):

h -
1
T21
= J d3 k e ik .(r,-r 2 ) 47f
(27f)3 k2 .
(11.8)

For the energy shift we then obtain

/j,E = Jd3k~ (~)6IJd3Teik.r-2zr/aDI2 (11.9)

27f4k2 aD

Here the integrals over Tl and T2 have been exchanged with the integration
over k. The integral over Tl is complex conjugated to the integral over T2,
which explains the last factor in (11.9). Now the following relation holds (see
Exercise 11.1):

Jd 3re -2Zr/a Deik.r = 167fZ (11.10)

2'
ao [k2 + (2Z I ao)2]
Inserting this into (11.9) yields

/j,E = J d3k 128e 2 (Z)8 1 (11.11)

7f2 k 2 ao [k 2 + (2Z lao)2 ] 4 .

We integrate over the angular part of the volume element d 3 k = k 2 dk sin rJ

drJd<.p. Because the integrand depends only on k and not on the angle, the
integration over the angle is easily performed. With the substitution of x =
ao/2Zk we get (see Exercise 11.1)
4Ze 2 roo dx 5 Ze 2
(11.12)
/j,E = 7fao io [x2 + 1]4 - 8~ .
Thus the energy of the ground state in first-order perturbation theory is given
by

Eo = - Ze 2 (Z -~) . (11.13)
ao 8
The calculated energy of the ground state represents an upper value for the
real energy. This can be seen as follows. The wavefunction in first-order per-
turbation theory can be regarded as a test wavefunction. If the expectation
value of the corresponding Hamiltonian is generated, an upper limit is always
288 11. Structure of Atoms

obtained. This statement is nothing more than the variational principle from
Ritz. 1

EXERCISE

11.1 Calculation of Some Frequently used Integrals

Problem. Calculate the following integrals.

() Ja
d3k ik.r 47r
(27r)3 e k2 =
1
iTT·
(b)J d3
Te
-ar
e
ik·r _
-
4m:t
(P +( 2)2·

rOO dx 57r
(c)
io (1 +x 2 )4 32

Solution.
(a) We calculate the integral

I
1
= J d3k eik .r 47r .
(27r)3 k2
(1)

It is more comfortable to introduce spherical coordinates

kcos8coslP , x T cos 8' cos lP' ,

k cos 8sinlP, y T cos 8' sin lP' , (2)
k sin 8 , z Tsin 8' .
It follows that
h = 47r 3 roo
k 2 dk j+~ cos 8 d8 dlP k12 {27r
(27r) io -~ io
x exp {ikT [cos 8 cos 8' (cos lP cos lP' + sin lP sin lP') + sin 8 sin 8'l} . (3)
If we adjust the k-coordinate system in a way such that k falls onto the z
axis, (3) can be simplified to

4 1
h =~
00
k 2 dk j+~ cos8d8 127r dlP zeikrSinll.
1 (4)
(27r) 0 - ~ 0 k
After the intergration over lP is performed, the substitution x = sin 8 gives
h = ~ 27r {= dkj+l dxe ikrx
(27r)3 io -1

47r 3 27r roo dk ~ eikrXIX=+l . (5)

(27r) io lkT x=-l
----
I Consult, for example, W. Greiner: Quantum Mechanics - An Introduction, 2nd
ed. (Springer, Berlin, Heidelberg 1994).
 11.1 Atoms with Two Electrons 289

Because ei<p = cos 'P + i sin 'P it results in

1
Exercise 11. 1.

h =
47r
--27r2
00
dk - kr
sin -

1
(27r)3 0 kr
47r
--27r2-1 00 dTsin
--
T
(27r)3 roT
47r 1 7r 1
--27r2- - (6)
(27r)3 r 2 r
(b) For the calculation of the integral

h = Jd 3 Te-<>lr 1eik.r (7)

we use an analogous strategy to that for (a). We apply spherical coordinates

and perform the integration over the angle:

h = J r 2 dr cos BdBd'Pe-ar eikrsin9

47r 1 o
00 kr _
sin _
r 2 dre- ar _
kr k
47r 1 0
00
rdre-<>rsinkr. (8)

1
This results in

-47r 8 dre- ar sin k r

00
--
k 8a 0

47r 8 [ e-<>r ] 00
--- (-asinkr - kcoskr)
k 8a a + k
2 2 0

47r 8 k B (2 2)-1
-TBaa2 +k 2 = -47r 8a a +k
2 2 -2 87ra
-47r(-1)(a +k) 2a = 2.
(a 2 + P)
(c) We employ the analytic continuation of the integrand to the complex
plane and apply the residue theorem in the following way:

lim
R->oo
f
-R
+R R
f(x)dx = 27riLRes(f(z),z=zk) ,
k=1

where 8'(Zk) > 0 is valid for all Zk, i.e. we close the integration contour in the 1m
upper half-plane.
If Zk is a pole of order n, the residue is determined by

Res (f(z), Z = Zk) Re

-R +R
with Z2 =-i

g(Z) = (z - zkt fez) ,

Fig. 11.2. Poles of the
which can be easily derived by taking the Laurent series representation of functionJ(z) = 1/(1+z 2)4
fez). We have and integration contour
290 11. Structure of Atoms

Exercise 11.1. (XJ dx 1 J+oo dx

io (1 + x 2)4 = 2" -00 (1 + x 2)4 '
because the integrand is an even function. Once we investigate the complex
function J(z) = 1/(1 + z2)4 with respect to singularities, we find two poles of
order n = 4 at Zl = +i and Z2 = -i. For the residue in the upper half-plane
we get with

g(z) (z _ i 4 1
) (z - i)4(z + i)4
and
g(z) (-4)( -5)( -6)(z + i)-7
the result
5
Res (J(z); z = Zl) = ~(-4)( -5)( -6)(2i)-7 25 i .
3!
The integral thus has the value
roo dx 5
-Jr.
io (1 + x 2)4 32

A very interesting exercise is the determination of the ionization energy.

This is the minimum energy needed to remove an electron from an atom. A
hydrogen-like atom remains thereafter. Hence, the ionization energy is the
difference between the energies of the hydrogen-like atom and the atom with
two electrons:
(11.14)
with EJe = -Z2e2/2ao and E5 e as in (11.13). As an example we consider
He. The measured ionization energy amounts to 1.807 Rydberg (Ry). One
Rydberg is the ionization energy of a hydrogen atom, or 1 Ry = e2/2ao
me 4 /2ti 2 = 13.606eV. For EJe we know the exact result
Z2 e2 me4
E~e = - - - = _Z2_ = _Z2 Ry (11.15)
2ao 2ti2
with ao = ti 2 /me 2 . Thus for the energy E5 e of the ground state of the atom
with two electrons we have
Ege = EJe - Er:;~asured = -4 - 1.807 = -5.807 Ry (11.16a)
for Z = 2. Using the approximate result (11.13) we obtain
Ege = -5.5 Ry . (11.16b)
As expected this lies above the exact result. The value presented in (1l.16b)
in not very accurate with a deviation of about 5%. The reason for this is that
the electron does not feel the complete nuclear charge, since the other electron
shields it somewhat. In order to improve the result an effective nuclear charge
 11.1 Atoms with Two Electrons 291

Zeff hal:> to be introduced. In other words: our trial wavefunction must be

altered. One electron in a Is state now possesses the wavefunction

<PIs (r) = _1_ (Zeff)3/2 e-Zeffr/aO (11.17)

~ ao
instead of (11.3). Zeff is varied until the minimum energy value is reached.
The expectation value of if for the new trial wavefunction can be estimated
in the following way. The electron-electron interaction provides the energy
shift 5Zeffe2/8ao, i.e. the same value as in (11.12), but Z has to be replaced
by Zeff. The other two kinetic contributions each yield the energy +Z;ff e2/2ao,
i.e. the kinetic energy one electron has in a Coulomb potential Zeffe2/r. Care
has to be taken when dealing with for the expectation value of the Coulomb
potential in the Hamiltonian, since there the true Z appears. It is

_ Ze 2 _ ~ (_ Zeffe 2 ) (11.18)
r - Zeff r .
Now the effective charge appears in the expectation value. The energy contri-
bution of the potential, i.e. the expectation value, is

(\_ Ze~e2\) = _ z::e 2 .

We summarize all this and arrive at

E6€ (Z, Zeff) = ~ Zeff e2 + 2 (Z;ff e2 _ Z Zeff e2 ) (11.19)

8 ao 2ao ao
The number 2 in front of the second term states that the energy contribution
is the same for the two electrons. The result (11.19) can be rewritten as

E6 e (Z, Zeff) = (~Zeff - 2Z Zeff + Z;ff ) ::

[ Zeff _ (Z _~)]2 aoe2 _ (Z _~)2

16 16
e
ao
2
. (11.20)

Obviously Eo (Zeff ) has a minimum at

5
Zeff = Z - 16 (11.21)

We take this value and the energy of the ground state becomes

E2e(z) = - ( Z - -5)22
~ = -2 (Z5
- )
- 2 Ry. (11.22)
o 16 ao 16
For He we obtain
E6e (Z = 2) = 5.695 Ry. (11.23)
Compared to the preceding calculation (11.16b) the additional energy shift is
enormous. However, the accuracy is not sufficient, because the same method
leads to an incorrect result for the H- atom, for which we obtain
Eo (H-) = -0.945 Ry , (11.24a)
292 11. Structure of Atoms

whereas the H atom has a ground-state energy of -1 Ry. According to (11.14)

the ionization energy becomes
H-
E ion = -0.055 Ry . (11.24b)
This energy is negative and implies instability for H-; in other words the
energy of the ground state of the H- atom is higher than that of a H atom
plus an electron. The calculated instability of the H- atom is in contrast to
the experimental findings: the H- atom is stable. This means that the cal-
culational method has to be further improved. In practice a many-parameter
dependent trial wavefunction is constructed and minimized towards these pa-
rameters. Since this process is rather involved, we do not further discuss this
approach.2

11.2 The Hartree Method

The Hartree method is constructed in such a way that it works for large atoms.
It is a "mean-field" method. This means that the electrons create a central
potential as a mean field, in which they are moving. The mean potential is a
function of the electron wavefunctions. First we construct an ansatz for the
mean potential and solve the Schrodinger equation with it. Then a new mean
potential is reconstructed with the help of the wavefunctions obtained. With
the new mean potential the Schrodinger equation is solved again, to obtain
improved wavefunctions. These iteration steps are repeated several times until
convergence is obtained, i.e. until we reach consistency. We are now going to
explain this procedure in more detail.
To begin with we construct an ansatz for the wavefunction of an N-electron
system, which fulfills our physical expectations as well as possible. The sim-
plest ansatz is a product ansatz, which constructs the entire wavefunction
from None-particle wavefunctions:
(11.25)
ipi(k) represents the single-particle wavefunction of the kth electron. The
wavefunction (11.25) is not antisymmetric, so one has to be aware of unphys-
ical states. This can be taken into consideration: for example, each electron
is placed into another single-particle state. In this way the Pauli principle is
fulfilled.
Initially, the mean potential is constructed as follows. The mean charge
density of the electron in state ipj at the point r' is given by -elipj (r') 12.
Considering the Coulomb interaction between the electrons, we find that the
potential energy of another electron at position r is

J r'Ir ~2
d3 r'llipj(r)1 2 . (11.26)

2 See, e.g., H.A. Bethe, R. Jackiw: Intermediate Quantum Mechanics, 2nd ed. (Ben-
jamin, London 1968).
 11.3 Thomas-Fermi Method 293

Summing over all electrons yields

Vi' (r) = Jd r,_e_ 3

2
_ "'.I'P (r')1 2
Ir -r'l L J
_ Ze 2
r
J#'
Ze 2
Vi (r) - - (11.27)
r
for the potential for the electron in state 'Pi. The last term describes the
contribution of the interaction with the nucleus. The wavefunction of the ith
electon is determined by the Schrodinger equation
h2 Ze 2
[- 2m \7 2 - -r- + Vi (r) ] 'Pi (r) = ti'Pi (r) . (11.28)

Equations (11.27) and (11.28) represent a system of coupled equations, which

unfortunately can only be solved numerically. The solutions 'Pi(r) are in gen-
eral not orthogonal, so they have to be orthogonalized. The calculation can
be further simplified once Vi (r) is replaced by the potential averaged over all
angles

Vi'(r) = -1
47T
J . d'Psm19d19Vi(r). (11.29)

This method was introduced the first time by Hartree. It is therefore called
the H artree method.

11.3 Thomas-Fermi Method

To determine the mean behavior of the potentials Vi (r), Fermi and Thomas
developed a semi-classical approach, which can be applied to many-electron
atoms. First the mean potential (11.27) for the ith electron is approximately
given by

V (r) = J 2
d 3 r,_e_ _ '"
Ir -r'l L
J
I'P· (r')1 2
J
_ Ze 2
r
. (11.30)

Contrary to (11.27) the sum also runs over j = i: it is supposed that for
a many-electron atom the influence of one electron can be neglected. The
electron density is given by

L l'Pj (r')1 2
= p(r') . (11.31)
j

How does this density depend on the potential V (r)? This is a self-consistent
problem, since V(r) again is given by p(r'). According to the Thomas-Fermi
method, 'Pi(r) is replaced by a plane wave
'Pi (r) rv e(i/h)Pi(r).r. (11.32)

The justification for this is that the kinetic energy of the electrons in heavy
atoms is very large; the potential varies only slowly compared to the spatial
294 11. Structure of Atoms

change (oscillation) of the electrons. The total value of the linear momentum
Pi (r) depends on the energy ti of the ith electron via

ti = PT (r) + V (r) . (11.33)

2m
The above approximation is valid only if
ti » V (r) , (11.34)
so the kinetic energy dominates. Consequently, the potential can be neglected
and we are allowed to consider free electrons as in (11.32). Within this semi-
classical approach an electron in state 'Pi is more or less uniformly distributed
over the classically allowed space and has a small probability in the classically
forbidden region. This is illustrated in Fig. 11.3. Electrons occupy all states
up to a maximum energy CF, which is the Fermi energy. The distance from
the potential curve V (r) (we assume a spherical potential) to the energy (t;
or tF) of the electron provides the kinetic energy of the electron.

r
Fig. 11.3. Electron dis-
tribution in the ground
state of the Thomas-
Fermi model. The kinetic
energy of an electron at
the Fermi edge EF is
marked for various dis-
tances r from the center

For a given r, electron states exist for which the kinetic energies are be-
tween 0 :::; Ekin (r) :::; tF - V (r). The maximum momentum is given by the
kinetic energy of the electron with energy tF, i.e.
PF (r) = {2m [EF - V (r)]} 1/2 . (11.35)
Only electrons with energy ti 2: V (r) are considered. Energetically lower lying
electrons are hardly contributing to the density at position r. The electron
density at r is then given by the sum over all electrons allowed to stay at r,
i.e.

p (r)

P~ (11.36)
37r 2 h3 .
In correspondence with (1l.32) we have used J d 3n = J d 3 kL 3 /(27r)3 for
the number of plane waves in volume L3; consequently the density becomes
J d 3n/L3 = (2/(27r)3) J d 3k. The factor 2 takes into account that two elec-
trons (with spin projection +1/2 and -1/2) can occupy a plane wave state.
With (11.35) and (11.36) we obtain
 11.3 Thomas-Fermi Method 295

() _ {2m[EF-V(r)]}3/2 (11.37)
p r - 37r 2 n3
for the electron density. Insertion of (11.37) and (11.31) into (11.30) deter-
mines the mean potential V (r) of a many-electron system. The following in-
tegral equation results:

V(r) = JdVlr~2r'IP(r')
J d3r,e2 {2m [EF - V (r,)]}3/2
Ir - r'137r 2 n 3
(11.38)

There is also the possiblity to determine the potential V (r) with a differential
equation. To do this we transform (11.30) into a Poisson equation:
for r> 0, (11.39a)
where ,1 = V . V represents the Laplace operator. Here we have used the
relation ,1lr - r'I- 1 = -47r!5Ir - r'l. If we additionally require a rotationally
symmetric potential V(r), we obtain
1 a 2 a 4e 2 3/2
r2 orr or [-V (r)] = 37rn3 {2m [EF - V (r)]} (11.39b)

by using (11.37) and (11.39a). Notice that the angular part of the Laplace
operator disappears once applied to V(r). Equation (11.39b) represents the
determining equation for V (r). It is a nonlinear differential equation. To solve
the problem we use the asymptotic behavior of V(r) for r ~ 0 and r ~ 00.
In the first case we expect V(r) ~ -Ze 2/r, since for r ~ 0 the screening
of the Coulomb potential of the central nucleus by the electrons decreases.
If r is equal to the atomic radius R, we require EF - V(R) = 0 for neutral
atoms, i.e. from outside no charge is observed. In this case the electrons totally
screen the central charge. If the atom is k times ionized and possesses only
(Z - k) electrons, a Coulomb potential of the form (Z - k)e 2 /r has to show
up outside the atom. For simplicity we restrict ourselves to neutral atoms.
Since for neutral atoms we set V(R) = 0, EF = 0 is fixed, too. This will be of
use later. Let us make the ansatz
Ze 2
V (r) = --¢(x) (11.40a)
r
and introduce the new combination for the variable r and of the parameter
ao

x
(11.40b)
b =

With this we obtain for (11.38) and (11.39b) the differential equation

1/2 d2 ¢ (x) = ,,3/2 (x) . (11.41)

X ~ 'I'
296 11. Structure of Atoms

One boundary condition for ¢(x) is ¢(O) = 1, because for r ---+ 0 the potential
V(r) has to agree with the Coulomb potential. Unfortunately (11.41) can be
solved only numerically. The solution has the following form:
1 - 1.59x for x ---+ 0
¢(x) = { . (11.42)
144/x 3 for x ---+ 00

For x ---+ 00 this solution behaves unphysically. ¢(x) vanishes only for x ---+ 00
and not for x ;:: (R/b )Zl/3, i.e. for x larger than the radius R of the electron
distribution. For x (or r) ---+ 0 the potential has the form
Ze 2 e 2
V(r) ~ - - + 1.80Z4/ 3 _. (11.43)
r ao
The reason for the appearance of the last term in (11.43) originates from
the finite contribution of the electrons near the origin. The potential at the
origin increases because of the outer electron shell. There is another interest-
ing property predicted by the Thomas-Fermi model: Since ¢(x) in (11.40a)
does not depend on atomic parameters (radius, charge, etc.), the form of the
atom is always the same. Only the length scale is changed. Because the scale
parameter r is proportional to Z-1/3 in correspondence with (11.40b), the
extension of the atom decreases with Z. According to (11.43) the second term
dominates for large Z and r ~ ao, i.e. V(r) can be approximately set equal
to V(r) = 1.80(e 2 /ao)Z4/3 near r ~ ao. With this we obtain for the electron
density p(r) [see (11.37)] of neutral atoms (tF == 0) the behavior p(r) rv Z2.
Hence, according to (11.36) it follows a behavior rv Z2/3 for the mean elec-
tron momentum. A typical atomic density distribution, calculated with the
Thomas-Fermi model, is sketched in Fig. 11.4.

p(r)
140
120
100
80
Fig. 11.4. A typical den-
sity distribution of a large 60
atom (mercury) within
the Thomas-Fermi model. 40
r is given in atomic length
units 20
0 0.2 0.4 0.6 0.8 1.0 r

So far, we have not considered angular momentum. If one takes angular

momentum states into account, V(r) has to be modified. The simplest way
to do this is to add the angular momentum barrier (centrifugal potential) to
the Coulomb potential V(r). Such an effective potential then reads

V.elf () _ V () l (l + 1) !i,2
r - r + mr
2 (11.44)
 11.4 The Hartree-Fock Method 297

-
A bound state exists only if Veff(r) has only one minimum for 1= 0 states for
Z ::; 4; for 4 < Z ::; 19 it has a minimum for I = 0 and I = 1, and so forth.
This means that for Z ::; 4 only bound s states exist, but no p states, etc.
All this is in good agreement with experiments, e.g. the first bound p state
appears at Z = 5.
Finally it should be mentioned that for small and large distances from
the atomic nucleus the Thomas-Fermi approximation breaks down. For small
distances the solution of the potential varies too strongly and for large dis-
tances the behavior of the potential is unphysical (see remark given above).
A drawback of the Thomas-Fermi model is the neglection of the exchange
interaction of the electrons. 3

E
\
\
\ l(l +1),,2
\" --z;;;-;:2
\ Fig. 11.5. The sum of the
" ........ Coulomb potential V (r)
.......... _
-------
Vcrc(r)
and centrifugal potential

-- -------------
[1(1+ 1)n?l/2mr 2 yields the
.,..- effective potential Ve/f(r).
,/' If a minimum arises at
,I-- V(r)
E < 0, then stable atoms
I
with angular momentum

,
I
I exist. For very large an-
gular momenta this is not
the case; the atoms are un-
stable

11.4 The Hartree-Fock Method

We have not yet considered the exchange effect of the electrons; the Hartree-
Fock method will take care of that. For that purpose a test wavefunction
is constructed, which is already entirely antisymmetric. One uses the Slater
determinant of one-particle states:

1/;(1,2, ... ,N) (11.45)

All one-particle wave functions are already orthonormal, i.e. (<pil<pj) = Dij. The
expectation value of the Hamiltonian is calculated in the following way. The
kinetic energy is given by

3 Modern versions of the Thomas-Fermi model can be found in the book by R.M.
Dreizler, E.K.U. Gross: Density Functional Theory - An Approach to the Quan-
tum Many-Body Problem (Springer, Berlin, Heidelberg 1995).
298 11. Structure of Atoms

(11.46)

A partial integration has been used and Gauss' law was applied. The interac-
tion energy of the electrons with the nucleus is given by

~J d3 r l<pdr) I2 -:;:-.
- L..- Ze 2 (11.47)
.=1
In order to calculate the contribution of the electron-electron interaction it
is of advantage to employ second quantization, which we encountered in the
third chapter. There the electron-electron interaction takes on the following
form [see Chaps. 4 and 6, especially (6.1) and Exercise 6.1]:
1 "J
2:1 3 e3' At2
At ,A
/ ,A )
d rd r Ir_r'I\1/>s(r)1/>s' (r)1/>s' (r)1/>s(r) . (11.48)

The subscripts 8 and 8' attached to the field operators ¢!(r), ¢s(r) refer to
the spins. The expression inside the bracket "(- .. )" describes the annihilation
and subsequent creation of two electrons at points rand r' with spin 8 and
8', respectively. This interaction is weighted with the factor e2 /1r - r'l. The
bracket itself gives the expectation value. This expectation value provides the
so-called pair-correlation function:
Nl (r) ¢1, (r') ¢s' (r') ¢s (r»)
= N! (r) ¢s (r) )N!, (r') ¢8' (r') )
At A , At ' A
- bS8'(1/>s (r) 1/>8 (r ) )(1/;8 (r) 1/>s (r») . (11.49)
This will be proven in Exercise 11.2. The second term allows for the possible
exchange of electrons at positions rand r'. The density of the electrons at
point r is
p(r) = L Nl (r) ¢8 (r») (11.50)

and the "correlation density'

ps(r,r') = Nl(r)¢s(r'») (11.51 )
i
with spin s

EXERCISE

11.2 Proof of (11.49)

Problem. Prove relation (11.49), i.e.

(¢! (r )¢1, (r')¢s' (r')¢8( r»)
= N!(r)¢s(r»)(¢l,(r')¢s,(r'»)
-bss,(¢!(r)¢s(r'»)Nl(r')¢s(r») . (1)
The expectation value refers to a Slater determinant.
 11.4 The Hartree-Fock Method 299

Solution. If <pier) are the orbital wavefunctions and i includes all quantum Exercise 11.2.
numbers needed, then the field operators ~!(r) and ~s(r) are given by

(2)

and
~s(r) = L<Pj(r)hjs (3)
j

[compare with (3.43)]. The left-hand side of (1) can then be written as

(~! (r )~!, (r )~s' (r')~s( r))

(4)

In the following, we limit our consideration to the calculation of

(h!)js' hks,hls )' For this we have to consider the expectation value of the Slater
determinant. The corresponding state is given by
(5)
ps

All states, from the lowest up to the highest, are occupied. By anticommuta-
tion on both sides (bra and ket) we find that 11/1) can be written as
(6)
which are the corresponding creation operators belonging to the two annihi-
lation operators in (4). Since we apply the anticommutation on bra and ket,
no sign is changed. The combination (ks') and (Is) does not appear in 11/1'),
since otherwise two electrons would be in the same orbit. Applying hks' his on
11/1) gives

(7)
The same result is obtained once h!);;s' is applied to the left-hand side. There
is another simplification: hks,hl s annihilates states (ks') and (ls). In order to
have the same state in the expectation value on the left-hand side (otherwise
the expectation value is 0), it has to hold that (is) = (ls) and (is') = (ks'),
or (is) = (ks') and (is') = (ls). The latter gives an additional minus sign,
because one has to transform the expectation value into the form
At At A A)
(blsbks,bks,bls . (8)
This leads to
(bi/bjk - bjlbikbss' )(b1kbss' - 1)(b1kbss' - 1) (9)
Inserting (9) into (4) gives
300 11. Structure of Atoms

Exercise 11.2. (¢1 (r )¢!, (r')¢s' (r')¢s(r))

L 'P: (r)'Pj(r')'Pj (r')'Pi(r) - 6ss' L 'P: (r)'Pj (r')'Pi(r')'Pj (r)
ij ij

Furthermore,

ij

Here we have used the fact that the expectation value is only different from
zero if the same state is annihilated in bra and keto For r' = r in (11) the
squared modulus appears. Inserting this formula into (10) yields

(¢1 (r)¢!, (r')¢s' (r')¢s (r))

= (¢~(r)¢s(r))N!,(r')¢s,(r'))
At A , At, A
- 6ss ,(1fJs (r)1fJs(r ))(1fJs (r )1fJs(r)). (12)
Note that the subscripts i and j are connected with sand s', respectively.

If we summarize (11.46)-(11.51), we obtain for the expectation value of

the Hamiltonian

(if) = - L
,
j d r (h2I v :(rW+ Z;2 1'Pi(r)1 2)
3
2

+~2 "
~
j d3 rd 3 r' e2 l'Pi(rWI'P ·(r')1 2
Ir - r'l J
'J

-~ z= 6 j d rd r'Ir ~ r'l 'P:(r)'Pi(r')'Pj(r')'Pj(r).

'J
SiSj
3 3
2
(11.52)

Here 6SiSj means that the summation is only over such states i and j, that have
the same spin quantum number. In order to obtain the equation of motion for
'Pi(r), (11.52) has to be minimized. It needs to be guaranteed, that {'Pi(r)}
remains an orthonormal set. The variation with respect to 'Pt (r) immediately
provides the following nonlinear equation:

( - - -ze
-h2fJ.
2m r
2)
'Pi (r) + j 2
d3r, - -e - L"
Ir - r'l
'P* ( 'r)
. J
J
x ['Pj(r')'Pi(r) - 'Pj(r)'Pi(r')6 siS;l = Ei'Pi(r), i = 1, ... ,N. (11.53)
These are the Hartree-Fock equations. 4 They differ from the Hartree equations
(11.28) by the additional -'Pj(r)'Pi(r') part in (11.53). The antisymmetric
part
4 D.R. Hartree: Proc. Cambridge Phil. Soc. 24 (1928) 89, III; V. Fock: Z. Physik
61 (1930) 126; J.C. Slater: Phys. Rev. 35 (1930) 250.
 11.4 The Hartree-Fock Method 301

indicates the amplitude with which the electrons in states i and j can be
found at positions rand r', respectively. This term automatically fulfills the
Pauli priciple in (11.53).

EXERCISE

11.3 The Hartree-Fock Equation

as a N onlocal Schrodinger Equation

Problem. Transform the Hartree-Fock equation (11.53) into the form of a

nonlocal Schrodinger equation.

Solution. The Hartree-Fock equation (11.53) is

fWi(r) = (-~11-
2m
ze
r
2
) 'Pi(r) + jd 2
3 r,_e_ - " 'P*(r')
Ir - r'l L
)
)
(1)
Obviously

V(r) = e 2+
-Z-
r
jd 3,
r -e -
Ir -r'l
2 (Z *" )'P(r) )
"'P(r
L)
)=1
)
(2)

represents the local potential energy consisting of the central potential energy
-Ze 2 /r and of the mean energy of the electrons in the field of all electrons
(Hartree energy)

j d3r' Ir - r'l (~'P*(r')'P(r')).

e2
L)
)=1
)
(3)

The (last) exchange term in (1) can be written as

j U(r, r')'Pi(r') d3 r' , (4)

where
2 Z
U(r, r') = -Ir ~ r'l L 'Pj (r')'Pj (r)t5 siSj •
(5)
)=1

It characterizes the nonlocal interaction energy. The exchange energy is non-

local and the Hartree-Fock equation (1) can be rewritten in the form

Ci'Pi(r) = (- :: 11 + V(r)) 'Pi(r) + j U(r, r')'Pi(r') d3r' . (6)

Equation (5) guarantees hermitecity,

U(r, r') = U*(r', r) , (7)
302 11. Structure of Atoms

Exercise 11.3. which enforces the reality of the eigenvalues ti. In order to convince ourselves
we rewrite (6) for <'ok(r):

Ek<'ok(r) = (- :: L1 + V(r)) + j U(r, r')*<'ok(r') dV . (8)

After multiplication of (6) with <'ok(r) and (8) with <'oi(r), subsequent integra-
tion over r and subtraction of both equations yields

(Ei - 10k) j <'ok (r)<'oi(r) d3r

= j[<'ok(r)U(r,r')<'oi(r') - <'oi(r)U*(r,r')<'ok(r')]d 3 rdV. (9)

Because of (7) the right-hand side vanishes and it follows that

(10)
for i = k because I !<Pi(r)lZ d 3 r = 1.
In (5) the nonlocal density (also called the density matrix) appears:
z
p(r,r') = L<Pj(r)<pj(r'). (11)
j=l

If the sum runs over all states (Le. Z ----> 00) and not only over the occupied
states, it would result in
00

p(r,r') L<Pj(r)<pj(r') = c5(r - r') (12)

j=l

and
(13)

The nonlocal interaction in (5) or (6) would then be local. Equation (13)
remains approximately valid for the nonlocal density (11). Often it is practical
to consider the mean potential

U(r) = j U(r, r') d3 r' (14)

and the effective potential

Ueff(r)<Pi(r) = j U(r,r')<Pi(r')d 3 r'. (15)

These relations define U(r) and Ueff(r). In Example 11.4 the exchange poten-
tial will be considered in more detail and a handy expression will be derived
within a reasonable approximation.
 11.4 The Hartree-Fock Method 303

The energy eigenvalue Ei of (11.53) is about the same as the negative

value of the energy needed to transport an electron from the ith state out of
the atom, i.e. equal to the ionization energy of the ith electron. This can be
seen from the following. Multiplying (11.53) from the left side with ip;(r) and
integrating over r yields

Ei = - ! d3 r (IV~~W + Z;2 Iipi (r W)

+ "'!d3 rd 3 r,_e_ 2
-ip*(r)ip*(r')
~
J
Ir -r'l Jt

X [ipj(r')ipi(r) - ipj(r)ipi(r')] OSiSj • (11.54)

A comparison with (11.52) shows that this is exactly the amount by which
(if) is reduced once all terms refering to the fixed "i" are canceled. -Ei is the
energy needed to remove an electron from the ith state. Notice that in (11.52)
the sum is over all indices i, j. If the term refering to the ith state is removed,
then the fixed i appears twice in the double sum. Here we assume that the
wavefunctions of the other particles are not changed. This is of course only
approximately true; for this reason Ei can only be approximately identified
with the ionization energy. These facts are known as "Koopmans I theorem" .5
The Hartree-Fock state (ll.45) represents an improved test wavefunction
in comparison to the Hartree state. Therefore the Hartree-Fock state is lower
in energy than the analogous state calculated with the Hartree method. How-
ever, it turns out that the calculated energy is still about 1 e V per electron
pair too large. Corrections have to be introduced.
Instead of looking deeper into details, it is worth studying the form of
the general solution: As a first, good approximation to start with we assume
the mean potential experienced by the electron to be spherically symmetric.
The wavefunction of the electron can then be described by quantum numbers
n, l, ml, ms. Here n represents the radial quantum number and 1 the orbital
angular momentum with projection mi. Furthermore, ms is the magnetic
projection of the spin (±1/2) and n - 1 - 1 gives the number of nodes in the
radial wavefunction. The wavefunction then takes on the form
(11.55)
In the hydrogen atom all orbital angular momenta 1 belonging to the same n
are degenerate; this is a special property of the 1/1" potential. With several
electrons present, this is no longer valid. Identical orbital angular momenta
belonging to the same main quantum number are no longer degenerate.
We consider a fixed 1. For a given 1 there are (2s + 1)(2l + 1) = 2(2l + 1) =
4l + 2 states. These 4l + 2 states constitute a shell. The Hartree-Fock trial
wavefunction is then constructed in such a way that all one-particle states are
filled from below; we have to indicate how many electrons sit in an orbit. In
general the last shell is not entirely filled. As a result of the Pauli principle
only 2 electrons fit into an s shell (l = 0), only 6 into an p shell (l = 1), 10
electrons into a d shell (l = 2), and so on. For example, the nitrogen atom N
5 T.H. Koopmans: Physica 1 (1933) 104.
304 11. Structure of Atoms

has two electrons in the Is orbit, two in the 2s shell, and three in the 2p shell.
The electron configuration is then given by (ls)2(2s)2(2p)3. The number in
front of s or p indicates the main quantum number n.

EXAMPLE

11.4 An Approximation for the Hartree-Fock Exchange Term

We intend to determine an approximation for the exchange term of the ex-

pectation value (11.52) of iI in the Hartree-Fock approach. For that purpose
we have to rewrite this term:

X = -~ I)SiSj
t,J
J d 3rd 3r'Ir ~2 r'l <p:(r)<Pi(r')<pj(r')<pj(r)
-e2Jd3rld3r2 ~ p;(rl,r2) , (1)
r12
where rl2 = Irl - r21 and ps(rl' r2) is given by (11.51). A factor 2 results,
because the combination Si = Sj appears twice in the sum over i and j. We
now introduce relative (r = rl - r2) and center-of-mass coordinates [R =
(rl + r2)/2]:
(2)

For Ps(R, r) we want to find a suitable substitute within the Thomas-

Fermi approach. For that purpose we replace in the sum over the product
<p;(rd<pi(rj) of occupied states (11.51) by a sum over plane waves with their
wave numbers fulfilling Ikl < Ikpl:

-1 L e.1 ik·(rj-r 2)
V .
J

::::: _1_ Jeik.r d k 3

l
(27r)3
47r kF (R) sin kr 2
-()3
27r 0 r
-k-k dk
1 1
27r 2 r3 (sin kp(R)r - kp(R)r cos kp(R)r)

::::: ps(R,r). (3)

Inserting (3) into (2) and integrating over r leads to

12
X::::: -e 47r447r J r 2 dr d 3R ~1 [1 . kpr - kpr cos kpr)
r3 (sm ]2
-e2-1-Jd3Rk~(R) . (4)
47r
3

Using (11.36) with kp = npp we get

 11.5 On the Periodic System of the Elements 305

x ~ _e 2 ~ (~r/3 J3 d R p4/3(R) . (5)

Example 11.4.

If we now take into account that

p(r) = L rpi(r)rpi(r) ,
insert (5) into (11.52) and vary with respect to rpi(r), we obtain

-J 3r'Ir ~ r'l L
d
J
rpj(r')rpj(r)rpi(r')8 si ,sj ---> _e 2 (~p(r) r/3 (6)

instead of the nonlocal term in (11.53). The idea of this approximation goes
back to Slater,6 the correct implementation leading to (6) to Kohn and Sham. 7

11.5 On the Periodic System of the Elements

The periodic system of the elements is depicted in Fig. 11.8 on page 331. Each
element is assigned to a box. The charge number is given in the upper right
corner of each element symbol. Below the element symbol the last occupied
shells are indicated. In the lower part of each box the configuration of the
ground state is described. Here one should notice that the lower angular mo-
mentum in a many-particle atom is energetically lower than the others. The
explanation for this is that the probability of finding an electron with lower
angular momentum is larger in the vicinity of the atomic nucleus. For this
reason the electrons feel more the Coulomb part and less the screening due
to the other electrons.
With these arguments we get a better understanding of the periodic sys-
tem. In a hydrogen atom there is one electron in a Is shell, in a He atom the
Is shell is filled, in Li and Be the 2s shell fills up and since all electrons are
paired the ground state has zero angular momentum. In the next row, from
Na to Ar, first the 3s orbit and then the 3p orbit is filled. From the fourth row
on problems arise. First of all one expects, that the 3d orbit is filled after the
3p orbit. It turns out that the 4s orbit collapses in energy in such a way that
3d and 4s are almost energetically degenerate. This leads to a complicated
situation, in which for one time the 4s orbit and for the next time the 3d
orbit predominates. The fourth row ends with the 4p orbit entirely filled.
The 3d electrons do not reach out of the atom as far as the 4s or 4p
electrons do. This is due to the radial wavefunction Rnl(r). Therefore the
3d orbit hardly plays a role for chemical properties. However, a completely
filled 3d orbit is responsible for magnetic properties, e.g. in Fe or Ni. These
considerations can be repeated for the remaining rows.
The outer electrons determine the chemical properties of an atom. If there
is a similiar configuration in the outer shell of two atoms, then both atoms
6 J .C. Slater: Phys. Rev. 81 (1951) 386.
7 W. Kahn, L.V. Sham: Phys. Rev. 140 (1965) A1133.
306 11. Structure of Atoms

have similiar chemical properties. Hence, these atoms are arranged in the
same column of the periodic system. The last column on the right-hand side
indicates rare gases. For rare gases the outer shell is entirely filled, Le. all
electrons are paired. For this reason the chemical activity is low, or rather is
nonexistant. Alkali elements are presented in the first column of the periodic
system. They are characterized by a weakly bound s electron in the outer
shell. The alkali metals are for that reason chemically very active.
We draw particular attention to the recent discovery of elements with
Z = 104 and Z = 112.8 They are precursers of the predicted superheavy
elements,9, expected to be centered around Z = 114. Their chemical prop-
erties had also been predicted by utilizing Dirac-Hartree-Fock calculations.
The word "Dirac" in this context means that Hartree-Fock calculations as
described here are based on the relativistic Dirac-Hamiltonian. It will be in-
teresting to watch the discovery of further superheavy elements in the near
future. 10

11.6 Splitting of Orbital Multiplets

In order to obtain the state of an atom it is not sufficient to know the

electron configuration. For example, for Be the first excited configuration is
2.5 eV above the ground state (ls)2(2s)2. The excited configuration is given
by (ls)2(2s)(2p). Altogether there are 12 possible states, that is 2(2l + 1) =
4 x 1 + 2 = 6 for the 2p orbit multiplied with 2 = 4 x 0 + 2 for the 2s or-
bit. The energies of these 12 states differ only slightly. This is because for
different states of the same configuration the electron-electron interaction is
only slightly different, but it is different. In addition the spin-orbit interaction
has to be taken into account; it is small for light atoms and can be treated
pertubatively once the electron-Blectron interaction is calculated. In order to
see how the states of a configuration split, we first think about the quantum
numbers serving for the classification of the states. The total angular momen-
tum operator commutes with the Hamiltonian. As a result the total angular
momentum J is a good quantum number. If we were to neglect the spin-orbit
interaction, the total orbital angular momentum L = Li Ii and the total spin
S = Li Si would commute with the Hamiltonian, i.e. Land S would be nearly
good quantum numbers. Taking the projection of the angular momenta onto
the z axis into account, we can classify a state belonging to a fixed J by
(11.56)
Here, ML = Li mL(i) and Ms = Li ms(i). The spin projections ms(i)
of the various single particles are not good quantum numbers, because the

8 See e.g. S. Hofmann, G. Miinzenberg, et al.: Z. Phys. A 350 (1955) 277 and 288,
and further references therein.
9 W. Greiner: Int. J. Mod. Phys. E 5 (1996) 1.
10 The nuclear aspects of superheavy nuclei are discussed in W. Greiner, J.A.
Maruhn: Nuclear Models (Springer, Berlin, Heidelberg 1996); J. Eisenberg, W.
Greiner: Nuclear Theory, Vols.I-3 (North-Holland, Amsterdam 1985).
 11.6 Splitting of Orbital Multiplets 307

entire spin wavefunction is a linear combination of different projections; for

two electrons we have, for example,

XSMs = 2: (~~ S Ims(l) ms(2) MS) xt(I)(I)x!s(2)(2) .

ms(I),m,(2) (11.57)

Here (~~Slms(l)ms(2)Ms) represents a Clebsch-Gordan coefficient for the

SO(3) of the spin group. Applying B = BIz + B2z on (11.57), where BIz
acts only on the spin function of the first and B2z on the spin function of
the second electron, results in the eigenvalue Ms = ms(l) + ms(2). The entire
spin wavefunction XSM, in (11.57) has to be antisymmetric. For a two-electron
state we therefore have only one singlet:

(11.58)

The triplet spin function XIMs is symmetric under exchange of both particles.
This situation is depicted in Fig. 11.6. If the electron-electron interaction is
neglected, then all states of the same main quantum number would be degen-
erate. The electron-electron interaction causes states belonging to the same
configuration but with different L or S to differ somewhat in energy. If various
configurations differ by several eV, then the difference between the states be-
longing to one configuration is given by severallO- I eV. States with the same
Land S are degenerate. Once the spin-orbit interaction is added, this degen-
eration vanishes. The splitting only amounts to 10- 2 -10- 3 eV for light atoms.
For heavy atoms the spin-orbit coupling becomes important and cannot be
neglected or treated perturbatively.

E
,..-------4= .... Spin-orbit splitting
(order of magnitude "" IO- L IO-3 eV).
" Different Land S multiplets
-.,...-f------f:::=:: .... [see (11.56)] of the same
\------t---'"": -- configuration.
'--------1= / They lie "" 10- 1 e V apart.

'T'<~"1 'V 5
Different configurations.
Fig. 11.6. Configurations
and their splitting into
'pm
L - S multiplets. The lat-
ter split further as a result
of the spin-orbit coupling

In spectroscopy, states are described by 2S+1 LJ. J is the total angular

momentum, L the total orbital momentum, and S the total spin. This means,
that, for example, 3P2 corresponds to a state with L = 1, S = 1, and J = 2.
It is called "triplet P two". At the beginning of the periodic table one can
still clearly assign an L, S, and J. Thus a hydrogen atom (H) is described by
2S 1 / 2 , because a spin-l/2 electron is found in a Is orbit. The total angular
momentum is J = 1/2. For the case of a He atom a second electron is added
308 11. Structure of Atoms

to the Is orbit. The Pauli principle means it has to be antisymmetric and

therefore antiparallel with respect to the first one [see (11.57)]. Therefore the
total spin is S = 0, the total angular momentum is J = 0, and the state is
described by 1 So ("singlet S zero"), i.e. a singlet state. In general, filled shells
are described by ISO states.
For the next heavier elements the assignment of states becomes more com-
plicated. For example, boron (B) has the configuration (ls)2(2s)2(2p), i.e.
L = 1, S = 1/2. The total angular momentum possesses two possible values
J = 1/2 or 3/2. The ground state is given by either 2P 1/ 2 or 2P 3 / 2. The
degeneracy for J = 1/2 is twofold (2J + 1 = 1 + 1 = 2) and for J = 3/2
fourfold (2J + 1 = 2 x (3/2) + 1 = 4). For carbon (C) with the ground-state
configuration (ls)2(2s)2(2p)2 it is worse. One finds six possible states for the
first electron in the 2p orbit (notice the factor 2 for the spin) and only five
for the second electron as a result of the Pauli principle. Altogether there are
6 x 5/2 = 15 different states. The factor 1/2 comes from the nondistinction
of the electrons, i.e. we cannot say which one is the first and which one is the
second electron. What are the possible LSJ combinations? Let us begin with
the simplest one: Two electrons in a p orbit can be at most coupled to the
total orbital angular momentum L = 2. The orbital wavefunctions then read
R n =2 L=1 (rl)R21 (r2)(112Imlm2MdYlml (I)Y1m2 (2)
R 21 (rl)R 21 (r2)
Yl1 (D 1 )Yl1 (D2) for M2 =2
If [YlO (D 1 )Yl1 (D2) + Yl1 (DdYlO (D2)] for ML = 1

x
If [Yl-l(DIYll(D2) + 2Yoo(Dl)Yoo(D2)
+Yll(Dl)Yl-l (D2)] for M2 =0
If [YlO (f?I)Y1- 1(D 2) + Y1- 1(D 1)YlO (f?2)] for M2 =-1
for M2 = -2.
(11.59)
Here D 1 ,2 denote the polar angles of the electron 1 or 2, respectively. Obviously
the wavefuntions (11.59) are symmetric against exchange ofthe two electrons.
For the entire wavefunction to be antisymmetric the electrons have to be
coupled to spin 0:

XS=OM,=O =
, ~ (Xll(I)xLl(2)
v2 22 2 2
- XL1(I)Xll(2))
2 2 22
, (11.60)

since this spin wavefunction is antisymmetric. For this case we have J = L =

2, i.e. the state is labeled by ID 2. It includes five states (2J + 1 = 5). From
15 possible states ten remain. Next we couple the two electrons to L = 1 in
the spatial part of their wavefunction, i.e. in their orbital angular momenta.
Owing to the symmetry properties of the Clebsch-Gordan coefficients for
L = h = l2 = 1, the resulting orbital state is antisymmetric. Thus, the entire
spin wavefunction has to possess S = 1, i.e.
 11.6 Splitting of Orbital Multiplets 309

These spin wavefunctions are symmetric. With L = 8 = 1 it is possible

to generate three different total angular momenta because of the couplings
[L = 1 Q9 8 = ll[J] (J = 0,1,2). The corresponding labels are
3P o, 3P 1 , and 3P 2 ,

i.e. altogether there are nine [(2 x 0 + 1) + (2 xI + 1) + (2 x 2 + 1) = 9)]. With

them we have constructed 5 + 9 = 14 out of 15 possible states so far. The
missing state has L = O. The spatial wavefunction is symmetric, therefore the
spin part has to be antisymmecric (8 = 0). With this lSo state we finally
cover all 15 states.
Which energies do these states have? Which state is the lowest? The an-
swers to these questions are given by empirical rules, the so-called Hund's
rules:
1) The LS multiplet with largest spin 8 is lowest in energy.
2) If there are several multiplets with different orbital angular momenta L
but with identical spin 8, then the state with largest L has the lowest
energy.
3) The total angular momentum J is then fixed by the spin-orbit coupling
as follows. Once the last shell is filled to less than half, the total angular
momentum J of the ground state is given by J = IL - 81. Once the shell
is more than half full, it is J = L + 8.
There is no simple proof of these rules, except by explicit detailed calcula-
tions or by "experience" (i.e. experiments). But one can give some arguments
contributing to a better understanding. In order to understand the first rule,
one has to note that the larger 8 becomes, the more symmetric the spin wave-
function becomes. The maximum value of 8 is 8 = (lj2)ne, where ne is equal
to the number of the valence electrons. The corresponding spin wavefunction
reads
1 1 1
XHl)xi (2) ... XHne) .
2 2 2
(11.62)

The spin wavefunction is entirely symmetric. In order to guarantee antisym-

metry of the total wavefunction, the orbital part has to become "more an-
tisymmetric" as the spin part becomes "more symmetric". This implies that
the spatial distance between the electrons becomes increasingly larger. The
correlations due to the Pauli principle (here: antisymetric orbital wavefunc-
tions) drive electrons apart. This leads to an effective decrease in the electron-
electron repulsion. Thus it is understandable that the state with largest 8 is
lowest in energy. In order to understand the second rule, we look at a classical
310 11. Structure of Atoms

z analogy, illustrated in Fig.ll.7. Two balls are connected with two strings at
point O. The two balls are kicked and obtain an orbital angular momentum. It
o depends only on the velocity v, because the distance R between the balls and
the z axis is determined by the velocity v (interplay between graviational and
centrifugal force). If the orbital angular momentum increases, the distance
between the two balls becomes larger. The same situation holds in an atom.
The larger the orbital angular momentum is, the larger the distance between
the electrons becomes. The role of the gravitational force is adopted by the
central Coulomb force of the nucleus. The electrons behave similarly to the
balls of our classical example. An increased distance leads to a lower repulsion
between the electrons.
Before we now consider the spin-orbit interaction more closely, a technical
point has to be mentioned. For the construction of the Hartree-Fock state, the
Fig. 11.7. Classical me- wavefunction is built from a determinant involving the orbital wavefunctions.
chanical model for illus- However, often many angular momentum states exist for the same configura-
tration of Hund's rules: tion. The entire wavefunction cannot be described by a single determinant, but
The faster the balls rotate
(i.e. the larger the angular
rather it is described by a linear combination of several determinants. Gener-
momentum becomes) the ally the Hartree-Fock state therefore is a linear combination of determinants.
further they fly apart: 'P As an example we consider the L = S = 0 state of two p electrons [consult
and, consequently, R in- the preceding discussion, (11.59) and (11.60)]. Except for the normalization,
crease the wavefunction is given by
[Yll (I)Y1 - 1 (2) + Y1 -l(I)Yll (2) - Y lO (I)YlO (2)]
x [XI(I)X!l(2) - X!l(I)XI(2)] . (11.63)
2 2 2 2

The factors attached to each term are given by the Clebsch-Gordan coeffi-
cients [see (11.59) and (11.60)]. The wavefunction (11.63) can be expressed
by the following linear combination of determinants:

1
Yll(I)X~l (1)
2
1
Yll (2)x~ l (2)
2
1 1
YlO (I)XI (1) YlO(I)X~l (1)
2
1 1
2
(11.64)
Y lO (2)XI (2) YlO(2)X~l (2)
2 2

In practice, one first constructs a set of Hartree-Fock one-particle states and

forms a linear combination of determinants from them. Then the expectation
value of the Hamiltonian is calculated.
 11.6 Splitting of Orbital Multiplets 311

EXERCISE

11.5 Application of Hund's Rules

Problem. Apply Hund's rules to the atoms B, C, N, 0, F, and Ne.

Solution.
Boron (B): As can be deduced from the periodic table there is a p electron.
This can be coupled to J = 1/2 or 3/2 (L = 1, S = 1/2). Since Land S are
fixed, only the third of Hund's rules is important. It states that the minimum
value of J has to be taken because the shell is filled to less than half. As a
consequence 2p 1 represents the ground state.
2
Carbon (C): The following orbits are allowed:
ID 2, 3p O, 3p 1 , 3p 2, and ISO,
which altogether amounts to 15 states. According to the first of Hund's rules
the state with the largest S is lying lowest. This is the case for the states
3p O, 3p 1 , and 3P 2. The second of Hund's rules makes no difference since it is
L = 1 for all states. The third rule states that 3p o is the energetically lowest
state, because the 2p shell (six possible states) is filled only to one third (two
electrons in the p shell).
Nitrogen (N): Nitrogen contains three electrons in the p shell. Altogether
we have (6 x 5 x 4)/(1 x 2 x 3) = 20 possible states. This number can be
deduced as follows. In a p orbit (two spin multiplied by three orbital degrees of
freedom) there are six possible states. The first electron then has six states to
choose from. The second can find a place in only five states (Pauli's principle)
and the third in only four. The denominator takes into consideration that
the electrons are undistinguishable (3!). Three p electrons can be coupled
to L = 0,1,2. L = 3 is not possible, because the wavefunction is totally
symmetric with respect to exchange of coordinates and we are not able to
construct an antisymmetric state with three electrons (this can be done for
two electrons at most). For the total wavefunction to be antisymmetric, L = 1
and 2 can belong only to S = 1/2 and L = 0 only to S = 3/2. This can be
seen if the states are explicitly coupled to L = 0,1 (or S = 1/2, 3/2 in spin)
and a permutation is applied to the functions. L = 2, S = 1/2 splits into
2D3/2 and 2D5/2 as a result of the spin-orbit interaction. L = 1 and S = 1/2
splits into 2P 1/ 2 and 2P 3/ 2; L = 0, S = 3/2 leads to 4S 3/ 2 . In total we have
20 states. The 4S 3/ 2 state possesses the largest value and therefore it is lowest
in energy.
Oxygen (0): Oxygen has four electrons in a p orbit, which therefore is more
than half full. Always two electrons have to be paired. The property of the
ground state is then determined by the two remaining electrons. The problem
therefore is equivalent to the C atom once we consider the holes instead of
the electrons. There are exactly two. By applying the first of Hund's rules, we
find that the 3p O, 3p 1 , and 3P 2 states are lowest. But now the shell is more
than half full, so according to the third of Hund's rules the state 3P 2 is lowest.
312 11. Structure of Atoms

Exercise 11.5. Fluorine (F): F is equivalent to the B atom. There are five electrons in the p
orbit, i.e. only 1 hole exists. It is L = 1, 8 = 1/2, i.e. the possible states are
2P1/2 and 2P 3 / 2 . Since the shell is more than the half full, the 2P 3 / 2 state lies
energetically lowest.
Neon (Ne): In Ne the p shell is full, i.e. all electrons are paired. There is only
one state: ISO.

11. 7 Spin-Orbit Interaction

The spin-orbit interaction is due to the magnetic moment of the electron with
the magnetic field caused by the electric current of the central nucleus. In the
electron's rest frame the nucleus is orbiting and thus causing a magnetic field
at the position of the electron. One may also understand it by remembering
that a particle (electron) moving with the velocity Ve through a electric field
E (of the central nucleus) sees a magnetic field proportional to Ve x E. We
shall derive the spin-orbit interaction from the relativistic Dirac equation in
Example 11.7. The result is
(11.65)

Here "i" represents the particle index and Si the spin of the ith electron.
G(ri) has the form

G(ri) = _ _e_ Vi x Ei = _1_ (~dVh)) t. (11.66)

2moc2 2m~c2 ri dri
Here Vi is the velocity operator ofthe ith electron, Ei is the electric field acting
on the ith electron and V(r) is the corresponding potential with eE(r) =
-(dV /dr)(r /1') for the central symmetric case. L = r xp stands for the orbital
angular momentum and S = (fi/2)u for the spin. The form (11.65) of the spin-
orbit interaction is discussed in detail in Example 11.7. At the moment we
take the form of this force for granted and show the resulting consequences.
Exercise 11.8 presents the details of the transformations necessary for (11.66).
For light atoms the spin-orbit coupling (11.65) can be treated perturbatively
and we need only the matrix elements of
(11.67)
The orbital angular momentum L and the spin 8 are approximately good
quantum numbers. Q stands for the remaining quantum numbers further
classifying the state. For given Q, L, and 8 several values M L , Ms exist.
The corresponding states are degenerate.
Without spin-orbit coupling the Hamiltonian does not depend on the elec-
tron spin. The exact wavefunction is then given by a linear combination of
products between the spatial wavefunctions IJrb1,ML (1'1, ... , rk) of k electrons
and the spin wavefunction Xr~s (81 , ... , 8 k )· The wavefunction then takes on
the form
 11. 7 Spin~Orbit Interaction 313

(11.68)

with certain coefficients a". The matrix element (QLSMLMSIGi .

siIQLSMLM's) then becomes
L a~a",(aNLMLIGila'NLM£). (aSMslsila'SM's). (11.69)
aali

Here we have separated the spatial part from the spin part. The first matrix
element in (11.69) is the element of Gi between states I}/b'1ML
and 1}/2;~~;~.
The second matrix element is related to the spin part. Here it is useful to
apply the Wigner~Eckart theorem. It states, that
(11.70)
i.e. the matrix element of the vector operator Si is proportional to the matrix
element of the total spin S with respect to the same states. This is explained
in more detail in Example 11.6. The proportionality factor does not depend
on Ms and MS'. A similar relation holds for the spatial part:
(11.71)
With this, the matrix elements for the spin~orbit interaction of the ith electron
read
(NLSMLMsIGi · SiINLSM~LM's)
'" (LMLILILMD' (SMsISISM's) . (11.72)
Summing over all electrons we obtain
(QLSMLMsIHsoIQLSM'sM L )
~(QLS)(LMLILILMD . (SMsISISM's)
= ~(QLS)(QLSMLMsIL· SIQLSM~M's) . (11.73)
The proportionality factor ~(QLS) depends on the form of the operator G i ,
but it is the same for all states of the LS multiplet. We realize the usefulness
of the Wigner~Eckart theorem: From the relative complicated sum over one-
particle spin~orbit couplings (11.65), the structure L . S with total spin S
and total orbital angular momentum L is peeled off in (11.73); we would
not have expected such a remarkable relationship from the beginning. The
L . S operator is easy to diagonalize once the relationship to the total angular
momentum J2 = (L + S)2 is noted. We obtain
(J 2 _L2_S2)
L·S = (11.74)
2
The matrix element vanishes, if J' =1= J and/or M' =1= M. This is simply
a consequence of the fact that Hso is rotationally invariant, i.e. a scalar of
SU(2) [or 0(3)]' while J and M classify the states of the various multiplets and
those within these multiplets, respectively. The value of the matrix element
(11.73) is
314 11. Structure of Atoms

(QLSJMIHsoIQLSJM)

= ~(Q2LS) [J(J + 1) - L(L + 1) - S(S + 1)]. (11.75)

This equation describes the splitting of the energy levels resulting from the
spin-orbit interaction. The splitting of two neighboring levels J and J - 1
within the same LS multiplet (i.e. levels with identical Land S) is now easy
to calculate:

~(QLS) [J(J + 1) - J(J - 1)]

2
~(QLS)J . (11.76)
This relation is called Lande's interval rule. For ~ > 0 the mUltiplet is regu-
lar. For ~ < 0 the multiplet is inverted. A connection to the third of Hund's
rules exists: if the shell is less than half full with electrons, ~(QLS) has to
be positive, so the level with J - 1 lies lower than the one with J that the
minimum value (energetically lowest state) appears for the total angular mo-
mentum J = IL - SI. If the shell is more than half full, ~(QLS) has to be
negative and the energetically lowest state has the total angular momentum
J = L + S. The empirical rule therefore is traced back to the sign and the
precise structure of ~(QLS).

EXAMPLE

11.6 The Wigner-Eckart Theorem

The states
IJM), M = -J, -J + 1, ... , J (1)
represent the 2J + 1 eigenvectors of the squared angular momentum operator
)2 and Jz:
)2IJM) = n?J(J + l)IJM), (2)
JzIJM) = nMIJM). (3)
They form a multiplet of SU(2) [or 0(3)] and J can be considered as the
Casimir operator classifying multiplets. If the representation of IJ M) is given
in a fixed coordinate system with the z axis as the quantization axis, then in
a system rotated by the Euler angles a, (3, 'Y these vectors read II

IJM)' = LD!..tM,(a,(3,'Y)IJM'). (4)

M'
Here D!..tM,(a, (3, 'Y) is the expectation value of the rotation operator
D(a,(3,'Y) with respect to IJM) and IJM') in the unrotated system:
D!..tM,(a,(3,'Y) = (JMID(a,(3,'Y)IJM') (5)
11 J.M. Eisenberg, W. Greiner: Nuclear Theory Vol. 1: Nuclear Models, 3rd ed.
(North-Holland, Amsterdam 1987).
 11.7 Spin-Orbit Interaction 315

with Example 11.6.

D(a, (3, "Y) (6)
For a fixed J the vectors (1) span an invariant subspace 12 t J C t. This means
that the repeated application of shift operators [generators of SU(2)]
(7)
onto the vectors IJ M) of t J leads either to the zero vector or to the vectors
belonging to t J . 13 Equation (4) determines the transformation properties of
these vectors.
Sets of operators are now called irreducible tensor operators once they
transform under rotation according to
J
DyiP D- 1 = L D~M,yJj), (8)
M'=-J

where yJjl with M = -J, ... , J represents the (2J + 1) components of the
irreducible tensor operator of order J.
With
ITJM) (9)
we denote quantum-mechanical states that fulfill the eigenvalue equations (2)
and (3). The quantity T includes all remaining quantum numbers and the
following is valid:
(10)
The Wigner-Eckart theorem now states the following about matrix elements
of vectors (9) with irreducible tensor operators:

(TJMly(klIT'J'M') = 1 (rJlly(k11IT'J')(J'kJIM'qM)
q ..j2J + 1
q = -k, -k + 1, ... , k . (11)
The quantity
(T Jlly(k 1liT' J') (12)
is the reduced matrix element. It depends only on T, T', J, J', and k and no-
longer on q. Its brackets are indicated by a double bar, like (II II). (J' kJIM' qM)
denotes a Clebsch-Gordan coefficient. Obviously the Wigner-Eckart theo-
rem states that dependence of a matrix element on the quantum numbers
M qM is solely contained in a Clebsch-Gordan coefficient. For its proof see
the literature. 14
Equivalent to the transformation property (8) of the irreducible tensor
operators is the validity of the following commutation relations:
12 E is the entire space, which is built up from vectors IJ M) for all J and M E
{-J, -J + 1, ... , J}.
13 See also W. Greiner, B. Muller: Quantum Mechanics - Symmetries, 2nd ed.
(Springer, Berlin, Heidelberg 1994).
14 M.E. Rose: Theory of Angular Momentum (Wiley, New York 1957).
316 11. Structure of Atoms

Example 11.6. n[k(k + 1) - q(q ± 1)1~ T~!)l ,

nqT(k)
q , (13)

where Jo = Jz is the z component of the total angular momentum operator

and J± are the shift operators as given in definition (7). For these operators
the following relations also hold:
J±IT1M) (1M ± 111±11M)ITlM ± 1) ,
(T1MIJ± = (lMI1±11M =t= I)(T1M =t= 11 , (14)
J oI1M) = nMI1M).
We now go back to (11.70). We easily check that the spherical components of
the spin operator of the ith particle
, 1"
T±l = =t= y'2(Sx±iSY )i,

To = (Sz)i (15)
fulfill the commutation relations given in (13) with k = 1. Hence, (Sq)i,
q = -1,0,+1, represents the components of the irreducible vector operators
Si and S = L:~=l Si (n is the particle number) is the total spin operator.
Using the commutation relations (13) and the relations (14) we obtain

(16)
with q = -1,0, +1. The first matrix element on the right-hand side of (16)
is the reduced matrix element, which is independent of Ms, M~, and q. We
will now show the validity of (16). First we derive the following relations from
(13):

[Jo,T±l] ±nT±l,

[J±,T±l] 0, (17)

[J±,TCf1 ] nhTo.
In (17) we have used the abbreviations introduced in (15).
From (17) and (14) it follows that
(T1MIT±IT'lM') ±(M - M')(T1MIT±1IT' 1M')
= 8M ,M'±l (T1MIT±l IT'lM') .
Therefore this matrix element has to be proportional to (T1MIJ±IT'lM').
Furthermore, by using (14) again it follows from (17) that
(T1MI[l±,T±dIT'lM') = °
8M,M'±1((T1M' ± IIT±lIT' 1M')(lM' ± 21J±11M' ± 1)
- (T1M' ± 21T±llT'lM' ± 1)(JM' ± IIJ±11M')). (18)
This equation can be rewritten in another way:
 11. 7 Spin-Orbit Interaction 317

(TJM' ± 1 1'1\1 IT'JM') (TJM' ± 2IT±1IT'JM' ± 1) Example 11.6.

(19)
(JM' ± 1IJ±IJM') (JM' ± 2IJ±IJM' ± 1)
It immediately follows that

(TJMIT±1IT' JM') = =f ~C(T±l' J,T, T')(JM'IJ±IJM) . (20)

With (17), (14), and (20) we get

(TJMIToIT' JM')
l",,((JMIJ'fIJM ± l)(TJM ± 1IT±1IT' JM')
ltv 2
- (TJMIT±1IT'JM =f l)(JMIJ'fI JM' =f 1))
=fl A ,

"" C(T±1' J, T, r )6 MM ,
ltv 2
X ((JM'IJ'f IJM' ± l)(JM' ± 1IJ±IJM')
- (JM'IJ±IJM' =f l)(JM'I J'fI JM' =f 1))
C(T±l' J, T, r')6 MM ,(JMIJoIJM') . (21)
We realize that the coefficient C(T±1' J, T, T') has to be independent of the
component Tz . Therefore one can write
(TJMITqlr' JM') C(T±l' J, r,r')(JMIJ~IJM)
= (rJIITIIT' J)(JMIJ~IJM') ,

where the standard components of Jq have to be taken:

A 1 A A A

J~ = =f y'2J± , J~ = Jo .
This proves the validity of (16).

EXAMPLE

11.7 Derivation of the Spin-Orbit Interaction

Spin-1/2 particles moving in an external vector field experience a spin-orbit

interaction. We will derive this interaction from the Dirac equation, which
reads

(1)

with
H HD - e[a· A(r) - Ao(r)] , (2)
2
HD ea· p + moc (J . (3)
A A

HD is the free Dirac Hamiltonian and a and j; represent the well-known Dirac
matrices
318 11. Structure of Atoms

Example 11. 7. A

U=
(0iT iT)
0 '
A

(3 =
(i -10) '
0 (4)

where the 2 x 2 Pauli matrices are given by

iI in (2) includes the interaction with the external electromagnetic field,

where A represents the vector potential and Ao(r) the Coulomb potential.
The charge of the Dirac particle is designated as e. Finally, p = -inV corre-
sponds to the momentum operator.
The Dirac equation (1) includes the interaction with the external fields
A(r) and Ao(r), and in particular also the spin-orbit interaction. In order to
extract the latter we investigate the Dirac equation (1) in the nonrelativistic
limit. The spin-orbit coupling will then appear automatically.
Equation (1) includes only the coupling to the electromagnetic field. But
elementary particles experience forces that have no electromagnetic origin
(e.g. strong interaction, weak interaction, gravity). An immediate example
is given by nucleons, which interact with each other through 7r-meson and
p-meson fields and also show pronounced spin-orbit coupling. 15 In order to
take into account such additional interactions of the Dirac particle we rewrite
the Dirac equation (1) in the form

(i ~ iILplL + moe) ~ ~ ~ Qa~

= (5)

with i = -i/3&, ')'4 = (3.

The Qa operators on the right-hand side of (5) characterize the different
kinds of interactions. The following possibilities exist.
1) The interaction with a scalar field V (r). The interaction operator then is
simply given by
Qs = V(r). (6)
This kind of interaction acts like a space-dependent mass, because
V(r)e/e 2 can be directly added to the mass m in (5) in order to obtain
this coupling.
2) The interaction with a vector field BIL" The interaction operator reads

Qv = -i L ilL BIL , BIL = {Bo(r), -B(r)}. (7)

IL
A special vector field of this kind is given by the electromagnetic field; it
is described by the four-potential
AIL = {Ao(r), -A(r)}. (8)
A glimpse back at (2) and (5) reveals that in this case we have
BIL = eA IL · (9)
.,-::----
15 (See footnote 12 on page 315).
 11.7 Spin-Orbit Interaction 319

3) The interaction with a tensor field of the form Example 11. 7.

QT = L -yJ1.".tC/ w . (10)
J1.,11
An example is given by the interaction of a non-Dirac-like magnetic dipole
with the electromagnetic fields
F - 8AJ1. _ 8AlI (11)
/L II - 8x ll 8xJ1.·
In this case it is
(12)
where /-10 = eh/2moe, mo is the rest mass and 9 the coupling constant. For
electrons 9 = 0, but for nucleons 9 # o.

We now consider the case of a particle that simultaneously interacts with

a scalar and a vector field. Inserting (6) and (7) into (5) and multiplying from
the left with /J yields, after rearrangement,

ih~~ = {en. [p - ~B(r)] + /J [moc2 - V(r)] + Bo(r)} 1/1. (13)

Next we consider the nonrelativistic limit of this Dirac equation. It can be

done either very elegantly with the help of the Foldy-Wouthuysen transfor-
mation,16 or in a more elementary fashion, as will be described next. We make
the ansatz

1/I(r,t) = 1/I(r)exp (-i~t) (14)

and write the four-component spinor as

1/I(r) = (<p(r)) . (15)

x(r)
<p(r) and x(r) describe two-component (Pauli) spinors, i.e. column vectors
with two components. For simplicity we choose
B(r) =0, (16)
so that in (13) only the two terms V(r) and Bo(r) remain. We insert (14) and
(15) into (13), and arrive at a system of coupled differential equations:
{E - Bo(r) - [moc 2 - V(r)]} <p(r) c(u· p)x(r) , (17)
{E - Bo(r) + [moe 2 - V(r)]} x(r) e(u . p)<p(r) . (18)
We split the rest mass from the total energy, i.e. E = E' + me 2 , and get
[E' - Bo(r) + V(r)] <p(r) e(u· p)x(r) , (19)
[2moc2 + E' - Bo(r) - V(r)] x(r) e(u. p)<p(r) . (20)

16 See, for example, W. Greiner: Relativistic Quantum Mechanics - Wave Equations,

2nd ed. (Springer, Berlin, Heidelberg 1994).
320 11. Structure of Atoms

Example 11.7. These equations are still exact. Now we intend to investigate the nonrelativis-
tic motion of a spin-1/2 particle influenced by the external fields Bo(r) and
V (r). For this we take into consideration only terms up to order v 2 / c2 . An
expansion up to the first order in [E' - Bo(r) - V(r)]/2moc2 yields for x(r)
from (20)
iT· P ()
( 1 - E' - Bo (r) - V (r )) --<p
X (r ) = r . (21)
2moc2 2moc
Inserting this expression into (19), we obtain the differential equation
iT· P ( E' - Bo(r) - V(r))
[E' - Bo(r) + V(r)] <p(r) = -
2mo
1- 2
moc
2 iT· p<p(r)
(22)
for the two-component spinor function <p( r). Using the relations
(iT·A)(iT·B) = A·B+i(a-·[AxB]) (23)
and
(a- . p) f (r ) (a- . p) = f (r ) (a- . p) (iT . p) -
ili( a- . V f) (a- . p)
= f(r)p2-ili{(Vf·p)+i[a-·(Vfxp)]}) (24)
we can transform (22) into
H'<p(r) = E'<p(r) (25)
where

H' (1- E' - Bo(r) - V(r)) p2 + Bo(r) _ V(r)

2moc2 2mo
+ li{a-· [VBo(r) x p]} + li{a-· [VV(r) x p]}
4m5c2 4m5c2
-~ [VBo(r)· p]- ~ [VV(r)· p] . (26)
4m5c2 4m5c2
For the theory to be consistent up to the order v 2 / c2 we have to pay attention
to the normalization of <p( r). In Dirac theory the charge density
p = e(<pt<p+xtX) (27)
is normalized to

J pd 3 r = 1. (28)

In lowest order it follows from (21) that

a-.p
X ~ --<po (29)
2moc
With the help of (23) we then obtain for the charge density

p ~ e<pt (1 + 4!;C 2 ) <p (30)

and for the normalization integral

 11. 7 Spin-Orbit Interaction 321

(31) Example 11. 7.

which represents the proper normalization condition for 'P(r). On the other
hand, the normalization for a Schrodinger wavefunction 'lj; is given by

(32)

Hence, we have to introduce a proper rescaling of the wavefunction 'P in the

nonrelativistic range:
'lj; = g'P (33)
with the property

J'lj;t'lj;d 3 r = J 'Ptgtg'Pd 3 r = 1. (34)

9 is in general an operator. Comparing (34) with (31) we obtain

'2 )1/2 '2
9 = ( 1 + 4 P2 2 ~ 1 + S P2 2 . (35)
moc moc
The Hamiltonian belonging to 'lj; is obtained by multiplying (25) with 9 from
the left:
gHIg-1g'P = Elg'P. (36)
Thus,
H'lj; = E'lj;, (37)
where H is given by

fl = (1 + p2
Sm6c2
) iI' (1 p2)
- Sm6 c2
p2 + Bo(r) _ V(r) _ [EI - Bo(r)]2 - V 2(r)
2mo 2m6c2
+ h{u· [VBo(r) x p]} + h{u· [VV(r) x p]}
4m6c2 4m6c2
h2 h2
---2-2 \J 2 Bo(r) + --2-2 \J2V(r) (3S)
Smoc Smoc
up to order v 2 / c2 . We have used the relations
p 2V(r) - V(r)p2 = -h2\J2V(r) - 2ih(VV(r)· p)
and
EI - Bo(r) + V(r)) ,2 ~ ,2 [EI - Bo(r)]2 - V 2 (r)
( 1- 2 P ~ P - 2 . (39)
2moc c
The first three terms in (3S) correspond to the known nonrelativistic Hamil-
tonian. The additional terms are relativistic corrections of order v 2 / c2 . We
distinguish them as
(40)
322 11. Structure of Atoms

Example 11. 7. The first term

Ii? li 2
2 2 '\1 2Bo(r)
U1 = - - 8
moc
+ ~'\12V(r)
8m c
(41)
o
was introduced for the first time by Darwin. For the case of the Coulomb field
we have Bo = Ze 2 /r and with '\12(1/r) = -41T 5(r) it follows for U1 that

(42)

The term

(43)
2m6c2
represents a correction to the kinetic energy as a result of the velocity depen-
dence of the particle mass. The last term
U3 = h{u· [V Bo(r) x p]} + Ii{u· [VV(r) x p]} (44)
4m6c2 4m6c2
describes the spin-orbit interaction. In a spherical symmetric field we find
dBo(r) r
VBo(r) ----
dr r
(45)
dV(r) r
VV(r) ---
dr r
With the help of the known expressions for the angular momentum operator,
t = n1 = [f x p], and the spin operator, 8 = (1i/2)iT, we obtain, with an
insertion of (45) into (44),

U3 = dBo(r) ~ + dV(r) (8' l)

(46)
dr 2m5c2r dr 2m6c2r'
The first term describes the spin-orbit interaction with the electromagnetic
field; the second one represents the interaction with the nuclear field. Since the
nuclear interaction is stronger for smaller distances than the electromagnetic
one, the second term provides the main contribution. This is the reason for
the relativistic contribution to the nuclear spin-orbit coupling. We estimate
its order of magnitude with the help of the oscillator model. The potential
in this case is given by VCr) = (1/2)Mw 2r 2, so dV/dr = Mw 2r. When all
factors for the second term in (46) are taken into account it follows that

dV(r) (8' l) I ~ (1iw)2 ~ (6 MeV)2 ~ 0.02 MeV. ( 47)

I dr 2m6c2r 2moc2 2000 MeV

From this consideration alone we would deduce that the spin-orbit strength,
which is caused by relativistic effects of the central potential (which simulates
the mesonic exchange), provides a contribution that is too small to explain the
strong spin-orbit strenght in the shell model. The coupling has to be caused
by the mesonic field directly. At the moment the mesonic field theory is not
far enough developed to confirm or to exclude such additional terms with
 11.7 Spin-Orbit Interaction 323

certainty. The spin-orbit coupling in the nucleon-nucleon interaction leads to Example 11. 7.
a corresponding term in the potential; its influence increases with the number
of nucleons. This agrees with the empirically secured observation that the
spin-orbit coupling for nearly completely full shells is stronger than at the
beginning of a new shell. For a continued discussion of the nuclear interaction
see the book by Eisenberg and Greiner.1 7

EXERCISE

11.8 Transformation of the Spin-Orbit Interaction

Problem. Show that the spin-orbit coupling

Hso = L 2;c2 (Vi x E i ) . Si

i

can be rewritten in the form

, '"' 1 dV,
Hso = 6
i
2 d2 2
m c ri ri
L i · Si,

where Li and 8i are the orbital angular momenta and the spins of the elec-
trons.

Solution. We assume that the potential, which generates the mean electric
field in the atom, is a central potential V (r). Then the electric field experienced
by the ith electron becomes
1 dV
eEi = -ViV(r) = -Ti--· (1)
ri dri
For Hso we obtain

H' so = L 1
--8i·
' (Ti , ) 1 dV
X Vi - - . (2)
i 2mc2 r-1.. dr-1.
Here the order of Ti and Vi have been exchanged. The orbital angular mo-
mentum Li of the ith electron is L; = m(Ti x Vi). Therefore we obtain for
Hso
,
Hso = --
1 L 1 dV,
--(8;· Li).
'
(3)
2m 2 c2 i r-1. dr-1.
This is the desired expression.

17 M. Eisenberg, W. Greiner: Nuclear Theory Vol. I: Nuclear Models, 3rd ed. (North-
Holland, Amsterdam 1987).
324 11. Structure of Atoms

11.8 Treatment of the Spin-Orbit Splitting

in the Hartree-Fock Approach
We want to throw some light on the observation that ~ < 0 for a less
than half full shell and ~ > 0 otherwise, and calculate an explicit expres-
sion for Gi in (11.66) for a central potential V(ri). Then Ei = -Vi V(ri) =
-(dVdri)(ri/ri) and with (11.66) we have
e 1 dV
G(ri) =
A A

---2--Vi x ri
2mc ri dri
lei 1 dV(ri)
- - - - - - r· x p.
A

2m 2 c2 ri dri • •
lei 1 dV(ri) t (11. 77)
2m2c2G~ i·

ti represents the orbital angular momentum of the ith electron. The spin-
orbit coupling part of the Hamiltonian is then written as

(11. 78)

In order to determine ~(QLS) [see (11.73) and (11.76)] it is sufficient to take

special Ms and ML values, since ~(QLS) is independent of them. We choose
M£ = ML and Ms = Ms, and with (11.73) obtain
(QLSMLMsIHsBIQLSMLMs)
~(QLS)(QLSMLMslt. SIQLSMLMS)
~(QLS) (QLSMLMs ILzszIQLSMLMS)
~(QLS)MLMs . (11.79)
Only the diagonal matrix elements of both spin and angular momentum
operators contribute; only Sz and Lz are diagonal. In general the state
IQLSMLMs ) is a linear combination of Slater determinants, which we ex-
press in abbreviated form as L,6 a,6 IQf3) with L,6la,612 = 1. The symbol 13
runs over all Slater determinants denoted by IQj3). The many-particle func-
tions IQj3) are functions of the single-particle wavefunctions <Pnlm,XmSl where
<Pnlm, represents the angular momentum and radial part and Xms the spin
part. By construction IQj3) is an eigenstate of Lz and Sz with eigenvalues
ML and Ms. We get:
(QLSMLMSIHsoIQLSMLMS)

= L
,6,6'
a~a,6'(Qj3l L
i
2 ; 2
m c ri ri
~V li . silQj3') . (11.80)

Only for 13 = 13' do we obtain nonzero matrix elements. This can be seen as
follows. The matrix element goes with a sum of one-particle operators. If we
pick one term, for example, the ith term, then it acts on the ith electron and
does not modify the states of the other electrons. This means that in IQj3)
and IQj3') all electrons must have the same one-particle wavefunctions except
 11.8 Treatment of the Spin-Orbit Splitting in the Hartree-Fock Approach 325

for the ith electron. However, the ith electron also has to possess the same
wavefunction, since IQ,8) and IQ,8') are eigenstates of t z . If the ith electron
were not in the same state, ML would be different for IQ,8) and for IQ,8'), so
that the matrix element would vanish.
In order to calculate (Q,8IHsoIQ,8) we perform the sum over all occupied
states:
(Q,8I HsoIQ,8) (11.81)

= L
nlmlms
Jd3ripnlmt(r)2m;c2r ~~ Lipnlmt(r)· (mslslm s).
s has diagonal matrix elements (mslszlms) = lims only for sz. Therefore only
tz appears in (11.81) and we have
(Q,8I HsoIQ,8) (l1.82a)

with

'T/nl f 3 21 dV
d rIIPnlmt(r)1 ~dr' (l1.82b)

Since V is assumed to be spherical symmetric, 'T/nl is independent of mi. From

this we immediately conclude that for closed shells the spin-orbit coupling
disappears. Indeed, for closed shells we have
(11.83)

because the sum is performed over all ml and ms within a shell. In fact, for
a completely occupied shell for any given occupied state with ml, another
occupied state with -ml exists. Consequently (l1.82a) vanishes. Therefore
we only need to focus our attention on the last open shell. We denote it as nl
and we get

(Q,8I HsoIQ,8) (11.84)

where ML and Ms represent total projections of the orbital angular momen-
tum and the spin, respectively. As a result of Hund's rule the lowest lying
state is described by the maximum spin. Therefore we set Ms = S, where all
electrons are alligned parallel. First we consider a shell less than half full with
electrons. From (11.80) or (11.84) we then obtain

(11.85)

Here we have used L:i3lai312 = 1. Comparing this with (11.79) we deduce

~(QLS) =li2'T/nl . (11.86)

2m 2 c2
According to (l1.82b) 'T/nl includes (l/r)(dVdr). Since in general V(r) in-
creases with r, 'T/nl and, consequently, ~(QLS) are positive. This is in accor-
dance with the first part of the third of Hund's rules. For a shell that is more
than half full, not all spins can be parallel. It holds that
326 11. Structure of Atoms

o. (11.87)
occupied nonoccupied all states
states states of a shell

In other words, ML and Ms of the nonoccupied states are the same as the
negative values of the corresponding values for the occupied states. Hence,
we switch from a particle to a hole picture, for which all holes are coupled
parallel in spin. Consequently, we obtain

L mlm s = -MLMS· (11.88)

occ.states

In comparison with (11.86) we obtain

~(QLS) = _ h,2 TJnl . (11.89)

2m 2 c2
Now the third of Hund's rules is completely explained. A special situation is
given for a half-full shell. All electrons are coupled parallel to spin S. Hence,
in every state there is exactly one electron with ms = +1/2 without a partner
with ms = -1/2. This means that all orbital states have to be occupied. As
a consequence every occupied one-particle wavefunction with ml has as its
counterpart an occupied one-particle wavefunction with -mi. Therefore the
component ML vanishes, i.e. ML = O. This is the only possible value and,
consequently, there is only one L = o. The spin-orbit coupling disappears for
half-open shells.
In addition, the strength of the spin-orbit coupling can be estimated. The
factor TJnl behaves as

TJnl '" J rlrpntl

d3 2e 2 e2
r3 '" -(Z---=-!-ao-)-3 . (11.90)

Here the Bohr radius ao = h,2/(me 2 ) has been inserted (see Sect. 11.1). For ~
we then get

~ '" 2 Z2 2
aom c
(2e2ao ) = Z0:2Ry (11.91)

with the fine-structure constant 0: = e2 Inc ~ 1/137. The splitting strength is

proportional to 0: 2 ~ 5 X 10- 5 and increases linearly with Z. For small Z the
perturbative treatment of the spin-orbit coupling is justified. This approach
is known as LS or Russell-Saunders coupling. As Z increases the approach
becomes worse and for large Z the spin-orbit coupling has to be treated from
the beginning. Since (L· S) does not commute with Land S, Land S are no
longer good quantum numbers. Instead one uses to the so-called j j coupling,
where first an individual coupling to a total angular momentum ji = li + Si
is introduced and then to J = ~i j i·
 11.9 The Zeeman Effect 327

11.9 The Zeeman Effect

Although we have already discussed the Zeeman effect in another volume of

this lecture series,18 it is wise to bring up again the most important ingredients
in order to complete the discussions. The Zeeman effect describes the splitting
of the levels in an atom in the presence of a weak magnetic field B. Again we
apply perturbation theory in order to treat this problem. The weak magnetic
interaction reads
, e, e '
Hmag = - 2me L· B - meS. B , (11.92)
with the gyromagnetic factors 1 for orbital angular momentum and 2 for spin.
For a magnetic field in the z direction an energy shift results in first-order
perturbation theory:
(QLSJMIHmagIQLSJM)
_ eB (QLSJMI(L z + 2Sz)IQLSJM)
2me
_ eB (MIi+ (QLSJMISzIQLSJM)). (11.93)
2me
Here Jz = Sz + Lz = M was used. In order to determine the matrix element
of Sz we use the Wigner-Eckart theorem (see Example 11.6). It states that
(QLSJMISzIQLSJM) (JIJIMOM)(QLSJIISIIQLSJ)
(11.94)
(QLSJMIJzIQLSJM) (JIJIMOM) (QLSJIIJIIQLSJ)
The elements on the right-hand side are the reduced matrix elements. The
same Clebsch-Gordan coefficients (J1JIMOM) appear as factors in the nu-
merator and denominator and therefore cancel. The Wigner-Eckart theorem
states
(QLSJMIJ· SIQLSJM)
~ (QLSJIIJIIQLSJ) (QLSJII SIIQLSJ) (11.95)
and analogously
(QLSJMIJ 2IQLSJM) ~ (QLSJIIJIIQLSJ}2 (11.96)
with the same reduced matrix element (QLSJIIJIIQLSJ) as in (11.95). The
ratio between (11.95) and (11.96) is equal to (11.94), i.e.
(QLSJMISzIQLSJM) (QLSJIJ· SIQLSJ)
(11.97)
(QLSJMIJzIQLSJM) (QLSJIIJ 2I1QLSJ) ,
or with the replacements J2 ~ J(J + 1)li2 and Jz ~ +MIi

(QLSJMISzIQLSJM) = J(J~ 1)1i (QLSJMIJ· SIQLSJM) . (11.98)

18 W. Greiner: Quantum Mechanics - An Introduction, 3rd ed. (Springer, Berlin,

Heidelberg 1994).
328 11. Structure of Atoms

On the right-hand side we can substitute J . S by

" '2"
'2
(J - L
'2
+S
'2
)
(11.99)
J·S=S+L·S= 2 .

With this we obtain for (11.98)

(QLSJMISzIQLSJM)
Mh [J(J + 1) + S(S + 1) - L(L + 1)]
(11.100)
J(J+1) 2
The energy splitting (11.93) now becomes

LJ.EQLSJ(M) = - ehB gM (11.101)

2mc
with the gymmagnetic factor
J(J + 1) + S(S + 1) - L(L + 1)
(11.102)
9 = 1+ 2J(J + 1) .
The last equation is also called Lande's 9 factor of the LS coupling. The atom
behaves as if it had a magnetic moment geh/2mc. If we have S = 0 (L = J),
then the orbital part only contributes to the magnetic moment and we have
9 = 1. For L = 0 there is a contribution from the spin and with S = J
we obtain a gyromagnetic factor of 9 = 2. In general 9 is between 1 and 2.
The splitting of the energy levels caused by a weak magnetic field is called
Zeeman's effect.
A magnetic field that is stronger than the spin-orbit field is responsible
mainly for the splitting of the LS multiplet. The state IQLSMLMs) then has
the energy shift
, ehB
(QLSMLMSIHmagIQLSMLMs) = - 2mc (ML + 2Ms) . (11.103)

States with the same ML + 2Ms are degenerate in energy. Adding the spin-
orbit interaction results in an additional energy shift, which is
(11.104)
The total energy shift is then
eltB
LJ.EQLS(ML,Ms) = --(ML + 2Ms) + ~(NLS)MLMs· (11.105)
2mc
This combined splitting is called the Paschen-Back effect. If, however, the
magnetic field becomes comparable to the spin-orbit field, then Hmag + Hso
should be considered together. If the magnetic field becomes too large, A 2
terms (A for vector potential) have to be added and we are confronted with
a highly nonlinear problem.
 11.9 The Zeeman Effect 329

EXERCISE

11.9 The Stark Effect

Problem. Discuss the splitting of levels under the influence of a constant

electric field Eze z along the z axis (Stark effect). Apply perturbation theory.
How do the states split in dependence on E z ?

Solution. The Hamiltonian of an atom in an electric field E = Eze z can be

written as follows:
fl = flo - dEz (1)
with
N N
d e 2)ri' e z ) Ld i . (2)
i=l i=l

Here di is the dipole moment of the ith electron. All electrons feel the same
field E z .
The splitting of the nth energy level in first-order perturbation theory is
given by
(3)
However, the dipole operator d changes parity, so the matrix element in (3)
is different from zero only if Ina') has the opposite parity to Ina). Equation
(3) then contributes only if odd and even parities are mixed in the nth energy
level. For larger atoms the levels split because of the electron-electron inter-
action. For hydrogen the energy levels split rv E z . For larger atoms one has to
consider second-order perturbation theory. The energy splitting then is

t::.Ea = E; L l(na'ldlna)lZ (4)

Ena. - E na ,
a'oIa
The splitting behaves like the square of E z . The squared term also becomes
noticeable in hydrogen for increasing E z .
330 11. Structure of Atoms
 Periodic system of elements
.h
R:
fiiili! Jr
.&
= ir
I Jt
4~

=I:~k
ggg $~
!if!TIl4(
I::
0j; ~i ......
M ......
1~ <0
1 >-3
::r
1"1""'' ;;; ctl
&l N
&; it
S
§
tr:I
II ffi1
g.
=,
Fig. 11.8. Table of the periodic system of elements. Notice the incorporation of the recently discovered pre-superheavy elements with
Z = 104, ... , Z = 112. They were predicted by theory, as were their chemical properties.

w
w
.....
332 11. Structure of Atoms

11.10 Biographical Notes

HARTREE, Douglas Rayner, *Cambridge 27.3.1897, t12.2.1958, British

physicist and mathematician. 1929-37 professor for applied mathematics, then
for theoretical physics at the University of Manchester. 1946-58 professor in
Cambridge. 1932 he became a member of the Royal Society. Hartree's most
important achievement was his approximate method for the calculation of
quantum-mechanical wavefunctions of many-particle systems. Among other
things, Hartree has also considered problems about digital calculators, ballis-
tics, and physics of the atmosphere. Main works: numerical analysis (1952),
the calculation of atomic structures (1957).

FERMI, Enrico, *Rom 29.9.1901, tChicago 28.11.1954. Assistent professor

at the University of Florence between 1924 and 1926. Between 1927 and 1938
professor of theoretical physics at the university of Chicago. In 1938 he re-
ceived the Noble Prize of physics for the creation of new radioactive elements
by neutron bombardment and for the discovery of chain reactions, which were
caused by slow neutrons. He mainly worked in nuclear physics and developed
the Fermi statistics named after him. In 1942 he put the first atomic reactor
into operation.

THOMAS, Llewellyn Hilleth, *London 21.10.1903, American physicist,

British by birth. Between 1929 and 1946 professor in Columbus (Ohio), then at
Columbia University in New York. Between 1968 and 1976 he was in Raleigh
(NC). Most important works were among, about other things, quantum me-
chanics of atoms (especially about the spin-orbit interaction of electrons) and
about the theory of particle accelerators.

HUND, Friedrich, *Karlsruhe 4.2.1896, t31.03.1997, German physicist. From

1929 to 1946 professor of mathematical physics at the University of Leipzig.
Between 1946 and 1951 and between 1951 and 1956 he was professor for
theoretical physics at the universities in Jena and Frankfurt/M, respectively,
and since 1956 at the University of Gottingen. In 1943 he received the Max-
Planck medal.

LANDE, Alfred, -Elberfeld, t30.10.1976 (today belonging to Wuppertal)

13.12.1888, American physicist of German origin. 1922-31 professor in Tiibin-
gen, then at the Capital University in Columbus (Ohio). In 1921-23 he devel-
oped a theory of the mulitplet spectra and of the Zeeman effect (Lande vector
model), where he introduced the 9 factor, now called the Lande factor after
him.

PASCHEN, Friedrich Louis Carl Heinrich, ·Schwerin (Mecklenburg, Ger-

many) 22.1.1865, tPotsdam 25.2.1947, German physicist. Between 1901 and
1923 professor for physics at the University of Tiibingen and in 1924-33 at the
University of Berlin, where in addition he was president of the "Physikalisch-
Technische Reichsanstalt". His work was in quantum and spectral physics. He
discovered the first two lines of the spectral series of hydrogen named after
 11.10 Biographical Notes 333

him. The discovery of the splitting of the spectral lines in strong magnetic
fields is due to him and Back (Paschen-Back effect).

BACK, Ernst, *21.10.1881, t20.7.1959; German physicist, professor in Ho-

henheim and Tiibingen.

STARK, Johannes, German physicist, *15.4.1874 Schickenhof, in Thansiis,

district of Amberg, t21.6.1957 Traunstein. S. became a professor in Hannover;
in 1909 he went to Aachen, in 1917 to Greifswald, and in 1920 to Wiirzburg.
He founded the "Jahrbuch der Radioaktivitat und Elektronik" in 1904 and
discovered in 1905 the (optical) Doppler effect in so-called channel rays and
in 1913 the Stark effect. He was awarded the Nobel Prize in 1919. In 1933 he
became the president of the "Notgemeinschaft der Deutschen Wissenschaft".
He was a friend of P. Lenard, a supporter of "German physics", which dis-
missed quantum theory and the theory of relativity as the "product of Jewish
thinking" .
12. Elementary Structure of Molecules

For molecules a more difficult situation arises, because, in general, we are no

longer dealing with a spherical symmetric problem. Generally the electrons
do not experience a central potential. In order to find out more about this
problem some approximations have to be applied. The atomic nuclei are lo-
cated at classical equilibrium positions, around which they slowly oscillate
(large mass of the nuclei!). On the other hand the electrons are moving fast in
the Coulomb field of both nuclei (if we deal with a two-atom molecule). This
offers an adiabatic approximation for the motion of the atomic nuclei. This
only works because the masses of the atomic nuclei are much larger than the
ones of the electrons. The ratio between nucleon (M) and electron (m) mass
is
M
- ;::j 1836. (12.1)
m
For heavier atoms this ratio changes to typical values of about lO+4 - lO+5.
Because of this mass ratio the electrons move much faster than the atomic
nuclei. To a high degree the positions of the nuclei can be regarded as fixed.
Vibrations of the nuclei around their equilibrium positions can be treated
adiabatically. This shows up as a slow deformation of the molecular orbits.
In addition to electronic excitations, atomic nuclei can vibrate against each
other and rotate as a whole. We estimate the energies of the corresponding
excitations. With a being the typical extension of the molecule it follows that
/)"p ;::j Ii/a from Heisenberg's uncertainty principle for a typical electronic
excitation and therefore

(12.2)

As for atoms the orders of magnitude are some eV. For the relative vibration
of the nuclei we assume a harmonic potential Mw2(R - R O)2/2. Here R - Ro
corresponds to the distance from the equilibrium position Ro (see Fig. 12.1).
If the relative distance between the two nuclei is (R - Ro) ;::j a, then the cor-
responding excitations are approximately equal to the electronic excitations.
Hence, we can set
1 li2
_Mw 2 a 2 ;::j --
2 2ma 2
and obtain an estimate for w:

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
336 12. Elementary Structure of Molecules

Fig. 12.1. Typical molec- V(R)

ular potential for the nu-
clei in the molecule. Ro
characterizes the relative I ~ (1/2)Mw2(R - RO)2
equilibrium position of the
two nuclei. R represents I
the relative nuclear dis- I
tance I
R

fiw ~ (m/M)1/2 h22 . (12.3)

ma
The factor (m/M)1/2 appears in the vibrational energy and h2/ma 2 is the
electronic excitation. We deduce that
m
Evib ~ ( M
)1/2 Eel. (12.4)

It is easier to excite a molecule by rotation than by vibration. For a rotation

the nuclei do not have to stretch out of their equilibrium position. In zeroth
approximation the equilibrium distance remains the same during the rotation.
Only the rotation-vibration interaction (centrifugal force!) is able to change
the distance between the rotating nuclei. The rotational energy is given by
E _ h2 l(l + 1)
rot - 28 . (12.5)

Here l represents the angular momentum of the molecule and 8 its moment
of inertia (8 ~ Ma 2 ). Hence, we obtain

h 2 2l(l
Erot ~ ma + 1) (m)
M . (12.6)

Here the factor (m/M) appears; it lowers the rotational energy with respect
to the electronic excitation:
Erot ~ ( : ) Eel. (12.7)
The excitation of a molecule in general is given by the combination of all
excitations mentioned, i.e.
E = Eel + Evib + E rot . (12.8)
The ratio of the energies is
m)1/2 m
Eel: EVib : Erot = 1: ( M :M . (12.9)

Obviously the rotational energy is lowest: the excitation steps for a rotation
are smaller by a factor of Jm/M ~ J1/2000 ~ 1/40 than for those a vibra-
tion and these are again smaller by the same factor compared to those for the
electrons.
 12.1 Born-Oppenheimer Approximation 337

12.1 Born-Oppenheimer Approximation

To describe the various motions of the molecule we begin with the Schrodinger
equation. The Hamiltonian is given by
(12.10)
where

(12.11)

represents the kinetic energy of the electrons and

2 '2

TN = L2~1I
11=1
(12.12)

is the kinetic energy of the nuclei. VeN represents the attractive electron-nuclei
potential. Vee describes the repelling electron-electron interaction. VNN indi-
cates the repelling Coulomb interaction between the nuclei. Since the masses
of the nuclei are very large, TN can be neglected. This step is called the
Born-Oppenheimer approximation. In the following, we will explain the ap-
proximation in more detail.
If we neglect the kinetic energy TN of the nuclei (static approximation:
fixed distance R of the nuclei), the relative distance R between the nuclei
only occurs as a parameter. The Schrodinger equation becomes
[Te + Vee(r) + VeN(r, R)] 'Pn(r, R)
= [€n(R) - VNN(R)] 'Pn(r, R). (12.13)
Here r indicates the position of the electron. The solutions 'Pn(r, R) depend
parametricaly on the distance between the nuclei. The energy of this state
is given by the electronic energy €n(R) lowered by VNN(R). The solutions
'Pn (r, R) represent a complete set of functions. The true wavefunction 1jJ( r, R)
can be expanded within this set:
(12.14)
m

The coefficients cPm(R) are to be found and, in general, depend on R. 1jJ(r, R)

is the solution of the full Schrodinger equation, which takes into consideration
the kinetic energy TN of the atomic nuclei, i.e.
(Te + TN + Vee + VeN + VN N)1jJ(r, R) = E1jJ(r, R) . (12.15)
Inserting (12.14) into (12.15) and using (12.13), we obtain

L(€m(R) +TN)cPm(R)'Pm(r,R) = EL cPm(R)'Pm(r, R) . (12.16)

m m

Now we multiply from the left-hand side with 'P~ (r, R), integrate over the full
space, and get
338 12. Elementary Structure of Molecules

L f d3Tip~(r, R)TNrPm(R)ipm(r, R) + €n(R)rPn(R) ErPn(R) .

m
(12.17)
Here we have used the orthogonality of the functions 'Pn(r, R). TN is propor-
tional to the Laplace operator LlR, which acts on rPm'Pm. It holds that
(12.18)
The index R indicates the action of the operators in R space. The first term
in (12.18) is proportional to TNrPn. The rest is brought to the right-hand side
of (12.17). The result reads
(12.1ga)
m

with

CnmrPm(R) -n2 L 2~
'" '"
f d3T'P~(r,R) (12.1gb)

x [2V RorPm(R) . V Ro'Pm(r, R) + rPm (R)LlRoipm(r, R)J .

The sum over a comes from TN [see (12.12)J and V Ro acts only on the co-
ordinate R", of the nucleus a, which appears in R = J(R 2 - RI)2. Now,
the order of magnitude of C nm is (m/M)1/2 times smaller than the electronic
kinetic energy. This can be seen as follows. The order of magnitude of the
term'" n2LlRo 'Pm/2M", (the kinetic energy of the nucleus) is proportional to
-(m/ M",)n2(Llr'Pm/2m); we have simply replaced LlRo by Llr and introduced
the electronic kinetic energy _n2Llr'Pm/2m. The factor m/ M", indicates that
the contribution of LlRo to Cnm is smaller by this factor than the kinetic
energy of the electron.
The first term in (12.1gb) remains to be estimated. For this
we approximate rPm by a harmonic oscillator wavefunction: rPm ;:::;
exp(-(R-Flo)2Mw/2n), Ro being the equilibrium position of nucleus a.
We have

(12.20)

6R indicates the shift from the equilibrium position. The factor M is canceled
by l/M in (12.1gb) and the contribution is proportional to the vibrational
energy Iiw. As noted earlier, this goes like'" (m/M)1/2. As a summary, the
C nm term can be neglected or treated with the help of perturbation theory.
Without the C nm term, (12.1ga) reduces to
(12.21)
This equation has an interesting interpretation: the energy of the electron
states €n(R) acts like an effective potential in R. We imagine that the elec-
trons build a "medium" in which the atomic nuclei move. This medium acts
as an elastic band. If the nuclei try to leave the equilibrium position, they will
be drawn back. There is an equilibrium position where €(R) has a minimum
deep enough to generate binding. The elastic band behavior is then nothing
other than the expansion up to the order (R - Flo) 2 .
 12.2 The Ht Ion as an Example 339

The C nm produce a mixing between different states 'Pn and 'Pm. This
mixing between the 'Pn(R) states can be neglected in lowest order, because
the C nm are small [of order (m/M)1/2, as explained previously]. Accordingly
the wavefunction is approximately given by
(12.22)
Here 1/ stands for all quantum numbers of level n. Env indicates the energy
of the molecule, which is calculated from (12.21).
In order to describe vibrations and rotations of the molecule cn(R) is
expanded in coordinates describing vibration and rotation, respectively. The
expansion in (5R = IR - Rol up to the squared order leads to a harmonic
vibrational potential (see Fig. 12.1). cN(R) does not depend on the angles
(Euler angles). Hence the rotations of the molecule are free. An excitation
of the molecule is a combination of excitations of the harmonic vibrational
oscillator and of the rotations.
We summarize: in the Born-Oppenheimer approximation, first the energy
levels of the electrons are determined for fixed distances R of the nuclear
centers. The electron energy cn(R) plays the role of a potential, in which the
nuclei are moving. If this potential has one or several deep enough minima,
one or several bound states of the molecule can exist. If the minima are only
weak or do not exist at all, then the molecule is not bound.

12.2 The Ht Ion as an Example

The simplest molecule is represented by the Ht ion. We neglect the motion
of the nuclei and the spin-orbit coupling, so that the Hamiltonian takes on
the form

(12.23)

The first term is the kinetic energy of the electron, the second and third terms
describe the attractive Coulomb interactions between electron and nuclei 1
and 2 with position vectors Rl and R2, respectively. The last term represents
the repulsive interaction between the two protons (nuclei). Furthermore, r
describes the electron's position and Rl and R2 the positions of the two
protons (see Fig. 12.2). It is impossible to determine the eigenfunctions of
(12.23) exactly and analytically. Hence, test wavefunctions are constructed
and the energy is minimized by variation.
In case of the Ht ion we use a test wavefunction that is a linear combi-
nation of hydrogen wavefunctions referring to the first (Vh) and second ('l/J2)
proton:
(12.24)
This crude method is particularly instructive and transparent. For the lowest
state we take the Is wavefunction, i.e.

(12.25a)
340 12. Elementary Structure of Molecules

Fig. 12.2. Illustration of

the coordinates used: Ti
is the position of the ith
electron and Rex of the
ath nucleus. The distance
R = J(R2 - Rl)2 be-
tween the nuclei depends
on both Rl and R2

and, correspondingly,

1/J2 = 1 e-lr-R21/ao (12.25b)

J( 7ra5)
Here ao = !i2 / em = 0.529 A represents the Bohr radius. The problem of the
Ht ion is reflection symmetric with respect to a plane lying exactly between
the two protons. This implies that the wavefunction possesses the same sym-
metry and that the wavefunction can be classified by its parity. If the parity
is positive we have a = f3, and if it is negative a = -(3. So we end up with
(12.26)
where N± is an appropriate normalization factor. For the norm (1/J±11/J±1 we
find

1 = N~ [2 ± 2Jd 3r1/J!(r)1/J2(r)]
With the so-called overlap integral

U(R) = J d3r1/J!(r)1/J2(r)

= (1 + R+ R:)
ao 3a o
e-R/ao

(see Exercise 12.1), in which R = J(R 1 - R2)2 , we have

N _ 1 (12.27)
± - J2[1 ± U(R)]
Obviously the overlap integral gives the spatial overlap of the two Is wave-
functions.
 12.2 The Hi Ion as an Example 341

The expectation value of H from (12.23) is now given by

c±(R)
('IjIIIHI'IjIl) + ('IjI2IHI'IjI2) ± 2('IjIIIHI'IjI2) (12.28)
2(1 ± U)
We now use the abbreviation 11) and 12) for the states 'IjIl and '1j12, respectively.
Since by symmetry we have (1IHI1) = (2IHI2), it follows that

(R) = (1IHI1) ± (1IHI2)

(12.29)
c± (1 ± U) .

Now, (1IHI1) is given by

(1IHI1) = J d3 r'ljl[(r)H'IjIl(r)

2
Cl+ eR - J e
d3rl'ljll(rWlr_RI'
2
(12.30)

Cl is the energy of the Is state in the hydrogen atom, which is -1 Ry. This
contribution comes from the first two terms in (12.23). The Coulomb energy
e2/ R, with R = IRI - R21, results from the repulsive interaction between
both protons and the last term in (12.30) comes from the contribution of
the interaction between the electron and the second proton. Inserting the
wavefunction, we arrive at (see Exercise 12.1)

(1IHI1) = Cl + e2 (1 + R) e-2Rjao. (12.31 )

R ao
For the second term in (12.29) we get

(1IHI2) = J d3 r'ljl[(r)H'IjI2(r)

(Cl + ~) U - J d 3r'ljlt(r)'IjI2(r)lr ~2R21 . (12.32)

The last integral in (12.32) is the so-called exchange integral

A = J d3 r 'IjI[(r)'IjI2(r) Ir ~2R21
-
_ -
e
ao
2
(1 +- R) e - Rjao .
ao
(12.33)

This integral will be determined in Exercise 12.1. With

U = (1 + R
ao
+ R2) e-Rjao
3ao
we can calculate the energy c±(R) as a function of R. The behavior of c±(R) is
depicted in Fig. 12.3. The state with positive parity is bound, since c+(R) has
a minimum. On the other hand, the energy of the state with negative parity
has no minimum; this state is never bound. Experimentally one finds a binding
energy of -2.8 eV for 'IjI+ (r) when the distance between the two protons is
342 12. Elementary Structure of Molecules

Fig. 12.3. Energy for the s(R) \

,
-
bound [E+CR)] and un-
bound [E-CR)] orbit of the
" s_(R) Lunbound orbit
Hi ion as a function of the
distance R between the
two protons
'-
1.3A - - -____ _
I R
I
I

1.06 A. Our calculated energy is still to high, which is to be expected with the
crude test wavefunction (12.26).
The wavefunction of an electron, for example, 'l/J±(r), is called a molecular
orbit wavefunction. The form of the test wavefunction (12.24) is also called
a linear combination of atomic orbits (LCAO). 'l/J+ is called the bound and
'l/J- the unbound orbit. 'l/J- is not bound, which can be seen from the following
argument. Since 'l/J- has negative parity, 'l/J-(r) = -'l/J-( -r), the wavefunction
has a zero at r = O. However, the electron profits especially from the attractive
interaction with the protons, which is at a maximum between the protons.
For very large proton distances the approximated function 'l/J+ (r) becomes
more realistic. For R -> 0 the wavefunction (12.26) becomes the wavefunction
for a He+ atom, because for R « ao the electron practically sees a point nu-
cleus with charge number Z = 2; indeed 'l/J+(r) becomes the Is wavefunction
of the hydrogen atom. This means that for small proton distances the ap-
proximation (12.26) becomes inaccurate. This is the reason why the binding
energy calculated within this method is too small (see Fig. 12.4).

R
Fig. 12.4. Schematic com-
parision between calcu-
lated and realistic energy
E+CR) for the bound orbit realistic

This phenomenon is illustrated in Fig. 12.5. For large proton distances R

we have a H atom and an isolated proton. The calculated value approaches
the accurate one, which is -1 Ry. But for R -> 0 our crude calculation yields
-3 Ry, whereas the realistic value is -4 Ry (this is the ground-state energy
for a He+ atom).
 12.2 The Hi Ion as an Example 343

Fig. 12.5. Hi ion: illus-

'" /e_(R) - e2 / R (calculated) tration of the calculated
"'- and realistic energies (mi-
-_
.... --._---
~_--~==::.::::..:=II-- -1 Ry
R nus the Coulomb repul-
sion of the protons) for
bound [c+(R)] and un-
-3R
e+(R) - e2 / R (realistic) bound [c_(R)] orbits. For
-4R small R (R -> 0) the limit
e+(R) - e2 / R (calculated) of a He 2 + ion is ap-
proached

EXERCISE

12.1 Calculation of an Overlap Integral

and Some Matrix Elements for the Ht Ion

Problem. Calculate

(a) the overlap integral U(R) = (¢1(R)I¢2(R),

(b) the energy (¢lIHI¢l), and
(c) the interaction integral (¢1(R)IHI¢2(R))

for the Hamiltonian (12.23) and the wavefunction (12.25a,b).

Solution.
(a) The overlap integral is

U(R) = ~Jd3rexp(-lr-RII/ao-lr-R21/ao)
rra o
The introduction of relative R = Rl - R2 and center-of-mass coordinates
Rs = !(Rl + R 2 ) of the two nucleis gives

11 = Jd3rexp(-lr-~R-Rsl/ao-lr+~R-Rsl/ao)
J d 3 rexp ( -Ir - ~RI/ ao -IT + ~RI/ao)
because of the obvious translation invariance of 11. The azimuthal symmetry
of the integrand of h offers the use of prolate-elliptical coordinates: l
R
x "2 [(e - 1)(1 _1J2)]1/2 cos cp,

y ~ [(e - 1)(1 - 1J2)F/2 sin cp,

R
z "2c'1J
1 H. Stocker: Taschenbuch mathematischer Formeln und moderner Verfahren (Harri
Deutsch, Thun 1996).
344 12. Elementary Structure of Molecules

Exercise 12.1. with ~ E [1, (0), TJ E [-1,1]' cP E [0,21r) and R = IRI - R21. Using

Ir =f ~RI = ~(~ =f TJ) ,

it then follows for h:

h = (~) 3 1 2
1< dcp [II dTJ [JO d~ (~2 - TJ2)

X exp {-R [(~ - TJ) + (~ + TJ)] / 2ao}

21r (~) 3 [II dTJ /00 d~ (e _ TJ2) e-Rf,/ao

21r (~) 3 /00 d~ (2e _~) e-Re/ao

~R3 e-R/ao 2ao (~ _ 2ao + a6 )

4 R 3 R R2
Hence, we have

U(R) = (~+ R
2 ao
+3a6R2) e-R/ao .

(b) The expectation value of H with respect to 'I/JI(R) is

Hl1 ('1/JI(R)IHI1/J1 (R))
= ~
1ra
J d 3 rexp (-Ir - R11/ao) H exp (-Ir - R11/ao) .
o
The first two terms of H together give the Hamiltonian, which has the eigen-
function 1/J1 as the ground state; the coordinate system has been translated
by R 1 , i.e.

H01/J1 = (- 2~ L1r - Ir ~2Rll) 1/J1

= c11/J1
with C1 = -e2/2ao. The expectation value of 1/Jl, Ho = (1/J1IHol1/Jl) is then C1·
The fourth term of H is independent of r; hence it follows that

HI = ~Jd3rexp(-2Ir-R11/ao)
'Irao
IR e R I
1 -
2
2

e2
R
with R = IR1 - R21.
The expectation value of the third term of H is
-
H2 = -3
1
1rao
J 3
d rexp(-2Ir - Rli/ao) I
_e 2
R I.
r- 2
The introduction of prolate-elliptical coordinates [see part (a) of this exercise]
gives
 12.2 The Ht Ion as an Example 345

Exercise 12.1.

The sum of all terms leads to (12.31):

Hll = ('!hIHI7J!11
co + HI + H2
+ -e + -e e- 2R / ao
2 2
co
R ao
= co + e2
R
(1 + aoR e- 2R/ aO )

(c) The interaction integral is given by

H12 (7J!I(R)IHI7J!2(R»)
= ~Jd3rexp(-lr-Rll/ao)
1rao
Hexp(-lr-R21/ao).

With the help of parts (a) and (b) of this exercise, the contribution of H12
can be written down immediately; from the terms 1, 3, and 4 of H we get

H12 = coU(R)
2
+ eR U (R)+-31
1rao
J
d r
3
Ir
_e 2
-
R I
1

X exp (-Ir - Rll/ao -Ir - R21/ao)

(co + ~) U(R) - A(R).

Here, A(R) corresponds to the already mentioned exchange integral (12.33).
In prolate-elliptical coordinates one obtains
2
A= 2e3
ao
(!!:.) 211 d'l] Jroo1 de (e _'1]2)_1_
2 -1 e- 'I]
e-R~/ao

~ (R)211 d'l] roo de(e+'I])e-Rf,/ao

2ao ao -1 Jl
e2 (R) 2 roo de e e-R~/ao
ao ao Jl
-e
ao
2
( 1+ -
ao
R) e- R / ao .
346 12. Elementary Structure of Molecules

12.3 The Hydrogen Molecule

The next, more difficult molecule is the hydrogen molecule, with two electrons.
Again we choose a test wavefunction and calculate the expectation value of
the Hamiltonian. We construct a symmetric test wavefunction and make the
following ansatz:
1
tPs(I,2) = 2[1 + U(R)] [tPl(rl) + tP2(rl)][tPl(r2) + tP2(r2)] Xsingiet .
(12.34)
The index "S" stands for "singlet". The subscripts 1 and 2 on tP indicate
the atomic centers with respect to which the one-particle wavefunctions are
defined. The arguments rl and r2 in the wavefunctions describe the positions
of the two electrons. We choose again a symmetric wavefunction, since we
want to have a state that is energetically as low as possible. Consequently,
we set both electrons in the same molecular orbit tP+(r) = tPl(r) + tP2(r),
which is in accordance with our last section about the HI ion. Because both
electrons are now in the same spatial state, their spins have to be anti parallel
due to the Pauli principle: therefore we have a state Xsingiet with S = o. If we
want a triplett state, i.e. S = 1, we have to put one electron into the tP+ orbit
and the other one into the tP- orbit. Since tP- is energetically unfavored, the
triplett states are higher in energy.
The test wavefunction (12.34) has two disadvantages. First, for very small
distances between the two protons the results have to be identical to those for
the Helium atom. This is not the case here, because for R ----> 0 both electrons
are in a Is orbit of a H atom and not in the Is orbit of the He atom. We see
the second disadvantage once we multiply out the wavefunction (12.34):
tPs(I,2) '" [tPl(rdtPl(r2) + tP2(rdtP2(r2)]
+ [tPl(rl)tP2(r2) + tPl(r2)tP2(rd]· (12.35)
The expression in the first bracket [ ... ] describes two electrons both attached
to the vicinity of the first or second proton. The wavefunction of the second
bracket stands for one electron attached to the first proton and the other
electron to the second proton. There is some problem with the wavefunction
of the first bracket because it will be quite improbable that the two electrons
are bound to the same proton for large distances R between the atomic centers.
This is mainly because the energy of the H- ion is higher than that of the
two H atoms. For large separations R, the true two-electron wavefunction will
therefore be better described by the terms in the second bracket in (12.35).
An alternative method for determining the energy of the H2 molecule is
given by the valence binding or Heitler-London method. In (12.35) only the
term for which one electron can be found around each proton is used, i.e. for
the singlet state

(12.36)
 12.3 The Hydrogen Molecule 347

Analogously we have for the triplet state

(12.37)

Both functions are already normalized. U(R) is the overlap integral given after
(12.26).
We take the expectation value of the Hamiltonian in order to estimate the
energy:
(H)±
(12IHI12) + (21IHI21)
2(1 ± U2)
(12IHI21) + (21IHI12) (12.38)
± 2(1 ± U2) .
Furthermore we use
(12IHI12) (21IHI21)
and
(12IHI21) (21IHI12) .
Hence, (12.38) is simplified to
(12IHI12) (21IHI12)
e± = 1 ± U2 ± 1 ± U2 . (12.39)

The Hamiltonian has the form

H = - -vi - -v~ - .,.----...,.

ITI -
e2
2m 2m RII

(12.40)

The first two terms describe the kinetic energies of the first and second elec-
tron, respectively. The next four terms represent the attractive Coulomb in-
teraction between electrons and nuclei. The second from last term represents
the repulsive interaction between the nuclei, and the last term that between
the two electrons. It holds that
, e2
(12IHI12) = 2el + R + Vc(R), (12.41)
where R == IRI - R21 . Here we have taken advantage of

(11-:l-ITl~2RlI11) = el;

the same holds for (2IHI2). Furthermore, we have

Vc(R) = J d 3r l d3r21'lfI(Tl)121'lf2(T2W

x
(
h- e2 e2
T21 - IT2 - RII - ITI -
e2
R21
)
'
(12.42)
348 12. Elementary Structure of Molecules

and

(12IHI21) = U 2 (2EI + ~) + A(R) (12.43)

with

A(R) / d3TI d3T2 '0t (rl)'0~(r2)'0I(r2)'02(rl)

e2 e2 e2 )
( (12.44)
x Irl - r21 - Ir2 - RII - Irl - R21
A(R) is the contribution from the exchange interaction. The exchange integral
(12.44) is not easy to calculate. As a crude estimate it is proportional to the
square of the overlap intergral

U 2 = / d3TI d3r 2 '0i(rl)'0~(r2)'0I(r2)'02(rl)

( / d 3r l '0t(r l )'02(r l )) ( / d3T2'0~(r2)'0I(r2)) (12.45)

Inserting (12.41)-(12.44) into (12.39) we obtain

E±(R) = 2 e2 VeeR) ± A(R) (12.46)

EI + R + 1 ± U2
or equivalently

(12.47)

Now it turns out, that VeeR) + e2/ R is always positive and A(R) + U 2 e 2 / R
is mostly negative. For this reason E+(R) is lower in energy than c(R). If
the "binding energy" c (R) - 2El or E+ (R) - 2EI is a function of R, the first
expression shows no and the last one minimum (see Fig. 12.6).

Fig. 12.6. Expectation

value of the Hamilto-
nian for the symmetric R
[c:+(R)] and antisymmet-
ric [c (R)] spatial wave-
function (12.36) and -27.2eV
(12.37), for the H2 mole-
cule -30eV

From this result it is obvious that the hydrogen molecule is bound in the
state '0+, i.e. the ground state is a spin singlet state. Note that the binding is
due to the negative term A(R) + U 2 e 2 / R. It is thus the large overlap of the
wavefunctions that causes the binding. The strength of the binding is, as a
 12.4 Electron Pairing 349

rough estimate, proportional to the overlap of the two electron wavefunctions.

The triplet state is higher in energy because 1/J- disappears in the middle of the
two protons. This follows from the antisymmetry of 1/J- [see (12.37)]. Between
the protons is the place where the attractive interaction between electron and
proton can be used in an optimum way.
For larger distances R between the two protons we have two hydrogen
atoms and for R ---+ 0 the He atom has to appear. The energy of the ground
state has to be somewhere in between. The ground state of the He atom is
a singlet state. Also the H2 molecule should have such a state. However, the
test wavefunction produces an energy that is too high for R ---+ 0; the binding
energy is too low. This has to be expected from a test wavefunction.
A real molecule shows a 1/R6 behavior for large R, which is known as the
van der Waals interaction. It is attractive. We cannot obtain such an interac-
tion here since more accurate calculations have to be performed: for the two
hydrogen atoms, a second-order perturbation calculation with respect to the
electron-proton interaction and to the electron-electron interaction provides
the correct 1/R6 behavior. The van der Waals interaction makes up approx-
imately 1/1000 of the molecular binding calculated here. This demonstrates
that our wavefunction, which also has the wrong limit for R ---+ 0, cannot
describe such a small effect. For R » 1 the energy is described fairly well, but
the behavior of the wavefunction is not.

12.4 Electron Pairing

For the H2 molecule we have seen that a large overlap of the two electron
wavefunctions causes binding. This has been the case for the the spin singlet
state. In the triplet state the overlap is small, which leads to a repulsion.
Now we can imagine why there is no H-He molecule: the two electrons of
the He atom in the ground state are in a spin singlet state. Both electrons
are coupled antiparallel. If both nuclei are very close to each other, the three
electrons (one from H and two from He) see a Li nucleus. Since the Is orbit is
completely occupied, the third electron has to go into an orbit with one main
quantum number higher. In the Li- ion the last electron has an ionization
energy of only 0.4 Ry. As a consequence we expect that there is no binding in
the H-He molecule. If H and He are separated (see Fig. 12.7), the electron of
the H atom has the possibility of interacting both with the electron of the He
atom that is parallel to the electron of the H atom and with the electron that
is oriented antiparallel. The last case does not work, since the interaction is
spin independent and cannot change the spin in the exchange integral (see
Fig. 12.7). Therefore it can only interact with the electron of the He atom
that is aligned parallel. This is similar to the triplet state of He, where (by
definition) two electrons are aligned parallel and interact with each other. The
triplet state is not bound: just as it is not bound the H-He molecule. In other
words, there is no H-He atom.
The reason for this can also be seen directly from the wavefunctions. For
this we build the Slater determinant of the three-electron system:
350 12. Elementary Structure of Molecules

Fig. 12.7. Illustration of (a) exchange (interaction)

the impossible (a) and impossible
possible (b) interactions
of the electrons in a H-
He molecule. However, the
C&--®
H He
triplet interaction in case
(b) is repulsive. Therefore (b) exchange (interaction)
no H-He molecule exists possible

CB-®
H He

¢l(l)xi(l) ¢l (l)xl (1) ¢2(I)Xi(1)

¢(1, 2, 3) N '!f!l (2)Xi (2) ¢l (2)xl (2) '!f!2(2)Xi (2) (12.48)
'!f!l (3)Xi (3) ¢l (3)Xl (3) '!f!2(3)xi (3)
Here ¢l is the Is helium wavefunction and ¢2 the Is hydrogen wavefunction.
Xi and Xl are the spin wavefunctions with ms = ±1/2, respectively. We
calculate the scalar product of (12.48) and obtain, after a direct calculation,

N- 2 = 6 J
d3rl d3r2 d3r3¢i (1)¢i (2)¢~(3)

x [¢1(l)¢2(3) -¢2(1)¢1(3)]¢1(2)
6(1 - U 2 ) . (12.49)
N is the normalization constant and U denotes the overlap integral between
¢l and ¢2:

U = J d 3r¢i(r)¢2(r).

The factor 6 in (12.49) appears because we find, for example,

J d3rl¢i (l)¢2(1) = J d3r2¢i (2)¢2(2) .

In other words: a permutation was performed at some places in order to show

that two terms are equal. Furthermore, Xi and Xl are orthogonal to each
other.
A similar calculation leads to the energy:

c(R) = 1 _1 U2 J d3rl d3r2 d3r3¢i (1 )¢i (2)¢~ (3)

x H[¢1(l)¢2(3) -¢2(1)¢1(3)]¢1(2) . (12.50)

Here the normalized wavefunction has already been used with the norm of
(12.49). The factor 6 cancels, because it also occurs in the numerator of
(12.50). From (12.50) we see that only that part appears that describes elec-
trons with the same spin in H and He atoms, i.e.
 12.5 Spatially Oriented Orbits 351

Only electrons with the same spin direction can interact with each other
because of the spin-independent Coulomb interaction. Look again at the spin
wavefunctions in (12.48)! Two electrons in the same spatial state but with
opposite spin are called paired electrons. In the example of the H-He molecule
we observed that paired electrons interact repulsively with other electrons.
This explains why rare gases show no chemical activity. In rare gases the
shells are all closed, so all electrons are paired. The chemical activity arises as
a result of non paired electrons. They still have a "free" place, which can be
occupied by an electron belonging to another atom. The number of possible
bindings is determined by the number of valence electrons, i.e. the nonpaired
electrons in the outer shell. In addition, there is still the repulsive interaction
with the paired electrons belonging to deeper shells. But this can be neglected,
because the overlap of the outer-shell electrons with the inner-shell electrons
is small.
The chemically active electrons are mostly outer (i.e. from higher shells)
s or p electrons. The d and f electrons are often lying too close to the in-
dividual nucleus to bind with electrons of other atoms. The binding results
in all electrons being paired for a molecule. Hence, such molecules have total
spin zero. One exception is given by the molecules of the transition elements
(with unpaired d and f electrons). Since the unpaired d and f electrons hardly
overlap, they remain unpaired and cause little or no binding.

12.5 Spatially Oriented Orbits

The alkaline earth metals, which are found in the second main group of the
periodic system, have closed orbits in their ground states: for example, Be has
the occupation 1s22s2. In this state they are chemically inactive. However, it
does not cost much energy (some eV) to lift an s electron into the next p orbit.
The p electron now has three different possibilities for orientation according
to the projection onto the z axis (me = 0, ±1). Here, we have placed the z axis
along the line connecting the two atomic nuclei. Figure 12.8 shows the state
for m = O. For m = ±1 the wavefunction is perpendicular to the connecting
z axis. If the p electron binds with an another p electron (of the other atom),
very strong binding results as long as both p electrons are in a m = 0 state.
Then the two electrons can interact actively at a relatively large two-center
distance, and therefore keep the Coulomb repulsion of the nuclei relatively
low.
We call states orientated along the x, y, and z axis Ipx), Ipy), and IPz); P
stands for the p orbit. The spatial part of the wavefunction has an angular tnl='±!
dependence of the form
Fig. 12.8. Qualitative
picture of the angular dis-
{3 ~,
YlO =
v{3
4.;
cos(} =
V4.;r
(12.51a) tribution of a p elctron
with angular-momentum
1M2(Yl l
v,t,
- Y1-t) = {f;Y
--
47T r '
(12.51b)
projection ml = 0 and
ml =±1
352 12. Elementary Structure of Molecules

1 (3x
Ipx) = J2(Yn + Y1- 1 ) = Y4;-;::- . (12.51c)

Here the cartesian coordinates are expressed in terms of the spherical ones.
In order to see how the bindings come about in a p orbit, we look, for
example, at the C 2 molecule. In the ground state C has two unpaired 2p
electrons, i.e. the valence number is 2. The two 2s electrons can be excited
and also contribute to binding; we will not take this into account for the
present consideration. The question is: Which state (configuration) is lowest in
energy? We search for an arrangement with the largest overlap of the electron
wavefunctions. This is the case if both electrons are in the IPz) state (see
Fig. 12.9).

a bonding
Fig. 12.9. (j bond of two
C atoms: the IPz) orbits of
the two carbon nuclei have
a large overlap at a rela-
tively large distance R be-
tween the C nuclei

Once this binding has occured, there will be no place anymore for another
electron because the interaction with it would be repulsive. The binding de-
scribed above is also called u binding (u bond), since it possesses me = 0, in
analogy to the notation of the s orbit with I = O. Since the spherical symme-
try is violated for a molecule, the molecular orbits cannot be classified by the
angular momentum. On the other hand, the projection onto the z axis (L z )
still represents a good quantum number for two-atom molecules.
The next p electrons cannot occupy the IPz) state anymore as they have
to switch over to the Ipx) and Ipy) states. As can be seen in Fig. 12.10 these
electrons will have a weaker binding since the overlap is lower: With the
reduction in the central distance R the overlap increases but, in addition, the
Coulomb repulsion of the nuclei increases. The wavefunction of this binding
has the projection me = ±1 onto the z axis. Thus it is also called 7r binding
(7r bond). So the C 2 molecule has two bindings, a u and a 7r binding.

Fig. 12.10. Molecular or-

rrhonding
bits with m[ = ±1 only
have relatively small over-
lap for a given two-center
distance R. This results in
weak binding. Since me =
± 1 one speaks of 7r bind-
ing R
 12.6 Hybridization 353

Nitrogen has three 2p valence electrons. One takes a (J bond and the other
two have to switch over to the IPx) and Ipy) orbits. They experience a weaker
7r binding.
An interesting exception is the O 2 molecule. The 0 atom has four 2p va-
lence electrons. The first three take one (J and two 7r bonds as for the nitrogen
molecule. For the fourth electron there is now no binding orbit available. To-
gether with the fourth electron of the other atom it forms a state with an
overlap that is as low as possible: one electron in Ipx) and the other in Ipy).
Both electrons now form an antibound state (as for 1f;- in H2). A triplet state
has spin 1. The O 2 molecule represents an exception from the rule that states
that all two-atom molecules for which the atoms are identical and do not
belong to transition elements have spin 0 in the ground state.

12.6 Hybridization
One way to amplify the (J binding is to deform the orbit in such a way that
the overlap becomes larger (see Fig. 12.lla,b).

(a) deformed Iflr) orbit Fig. 12.11a,b. Hybridiza-

tion of a state: schematic
illustration of the ampli-
fication of the overlap ef-
fect. The CT bond is in-
creased by deformation
and, hence, an increased
(b) large overlap of lWO overlap of the Ipz) orbit.
Ipz) orbiL, (a) Ipz) orbit. (b) Overlap
of two Ipz) orbits

This only works if different angular momenta are mixed with each other.
It becomes possible because the rotational symmetry is broken for a two-atom
molecule. An example is the Li2 molecule. Li has one 2s electron in the outer
shell. But only little energy is needed to shift the 2s electron into a 2p orbit.
We denote the s orbit by Is) and the p orbit by Ip); then we are able to
construct a test wavefunction
(12.52)
A is the variation parameter. The s wavefunction is spherically symmetric (see
Fig. 12.12). The p wavefunction has positive sign in the range z > 0 and a
negative one for z < 0 [see Fig. 12.12 and (12.51a)]. We have put the p electron
into the IPz) state.
For z > 0 both wavefunctions amplify each other. For z < 0 a negative
interference (attenuation) appears. For this reason the shape appears on the
right-hand side of Fig. 12.12. The overlap with the electron of the other atom
is now larger and therefore the binding is stronger. The phenomenon of mixing
two atomic orbits described here is called hybridization.
354 12. Elementary Structure of Molecules

Fig. 12.12. Mixing of the

Is) and Ipz) orbits causes
a strongly deformed wave-
function. This illustrates --Jloo-
the principle of hybridiza-
tion

Next we consider the water molecule H 2 0. The atoms are not lined up,
which leads to a larger dipole moment. In the following, we will explain the
phenomenon and follow two opposite extremes: no hybridization and maxi-
mum hybridization. The truth lies somewhere in between.
Without hybridization we can assume the following: The 0 atom has four
p electrons; if no hybridization occurs, they remain P electrons. We place one
electron into the Ipx) and one into the Ipy) state. The last two are paired in
the IPz) state. The electrons in Ipx) and Ipy) bind with the electrons of the
H atoms. Since Ipx) and Ipy) are oriented perpendicular to each other, the
calculated angle of the binding is 90°. Experimentally one finds lOso. If one
takes into account that the two protons of the H atoms repel each other, then
the angle becomes larger. Thus, we can imagine how the lOso arises.
The other way of arguing is to assume that the 2s and 2p orbits are
energetically degenerate. (In reality the 2p orbit is somewhat higher.) Now
mixing can occur. The states with lowest energy, which are obtained by a
variation procedure, are
11) ~(Is) + Ipx) + Ipy) + IPz)) , (12.S3a)
12) ~(Is) + Ipx) - Ipy) - IPz)) , (12.S3b)
13) ~(Is) - Ipx) + Ipy) - IPz)) , (12.S3c)
14) ~(Is) - Ipx) - Ipy) + IPz)) . (12.53d)
They have maximum mixing.
Now we consider the spatial orientation of these states (12.S3a-d). In
(12.S3a) the p part has the form [see (12.S1a-c)]

Ipx) + Ipy) + IPz) '" x + y + z . (12.54)

r
The wavefunction is oriented in direction (1,1,1). For positive x, y, and z
the wavefunction is positive and for negative x, y, and z it is negative. If we
overlap the Is) state with this state, the wavefunctions add up in the direction
+(1,1,1) and substract in the direction -(1,1,1). One proceeds similarly
with the remaining states in (12.S3b-d). 12) is pointing into the (1, -1, -1)
direction, 13) into (-1,1, -1), and 14) into (-1, -1, -1). The four orbits form
a tetrahedron (see Fig. 12.13). The oxygen nucleus sits in the middle: from it
the four hybridized states 11),12),13),14) point starlike to the corners of the
tetrahedron. These legs form an angle of 109.6° to each other which follows
from the elementary geometrical relation sine B/2) = .j2[3.
The calculated angle is too high. But as already mentioned, complete
hybridization is an extreme case, because the sand p orbits in the oxygen atom
are not degenerate. The truth lies in between no and maximum hybridization.
 12.6 Hybridization 355

H 11) Fig. 12.13. Structure of

the water molecule H 2 0.
The fully hybridized sand
p orbits point from oxy-
gen into the direction of
the tetrahedron's corners
(states 11)·· ·14). The an-
gle between two neighbor-
ing legs is 109.6 0 , which
agrees well with the an-
gle 105 0 between both H
atoms obtained from ex-
periments
paired
p electrons
of 0

The ammonia molecule NH3 is another interesting example. The molecule N

forms a tetrahedron with 107 between the axes. The nitrogen nucleus sits in
0

one of the corners, for example, in the upper corner (see Fig. 12.14). If the 2p
electrons of the nitrogen atom were only in the Ipx), Ipy), and IPz) orbits, their H
,, H
angle would be 90 Via mixing with the 2s state, the angle becomes larger.
0 •

In the ammonia molecule the ground state splits into two energetically close ,, IH I
configurations. This can be explained by the fact that the N atom can be on , I I
either side of the pyramid (see Fig. 12.14). The wavefunctions representing the , I I
,-¥I
N atom on the upper or lower side may be added or subtracted. The antisym-
N'
metric wavefunction belongs to the ground state of NH 3. The symmetrically
Fig. 12.14. The two,
combined wavefunction slightly above lies. One speaks of a splitting of the
nearly degenerate ammo-
ground-state configurations as a result of tunneling of the N atom through nia configurations (NH3).
the plane spanned by the three hydrogens. N can be located either
We can understand some of the properties of water from the electronic above or below the plane
properties of the H2 0 molecule. The molecule has ten electrons altogether, spanned by the three hy-
drogens
two from the two H atoms and eight from the oxygen atom. The two Is
electrons in the 0 atom can be neglected in the following consideration. They
belong to a lower-lying closed shell and are, so to speak, inert and therefore
inactive. The remaining eight electrons fill the hybridized orbits (12.53a-d).
In addition the oxygen attracts the electrons, which implies polarization of
the water molecule.
Altogether the 0 atom has four charged legs, two in the directions of the
H atoms and two away from the 0 atom (see Fig. 12.15). If now a second H 2 0
molecule comes close to the first one, the H atoms, which now appear to be
more positively charged, can bind the oxygen atom with a negative leg (see
Fig. 12.15). This kind of binding is called a hydrogen bond. With only 0.2eV it
is quite weak. The mean distance between a H atom and an 0 atom of another
molecule is about 1.8 A, whereas the distance between Hand 0 within the
same molecule is about 1 A. This explains the low binding energy! Each water
molecule may obviously undergo hydrogen bonding to four neighbors. This
mechanism accounts for the accumulation of many water molecules of the
form H 2n O n = (H 2 0)n in water (cluster formation).
356 12. Elementary Structure of Molecules

Fig. 12.15. Principle of __ thr~u.gh polarization

the hydrogen bond of two negative posItive legs
H 2 0 molecules. It leads to
an accumulation (building legs "'" H
of clusters) of many water
molecules, which are con-
nected tetrahedronlike
H oH ~----__ ~

There are many more molecules that form hydrogen bonds. They belong
to the water-soluble substances because they untergo hydrogen bonding with
the H20 molecules of the water. Substances that do not form hydrogen bonds,
are not water soluble. Oils belong to this category. Heating causes the hydro-
gen bonds to break and the substance becomes more fluid, i.e. its viscosity
decreases.

12.7 Hydrocarbons

H The considerations of the last sections have shown that the geometry of a
CH4 slruc lure molecule depends on the structure of the electron orbits and the hybridiza-
tion. An excellent example of how electron configurations and the geometry
\09.6° of molecules are connected with each other can be found in hydrocarbons.
H In carbon the ground state is given by the configuration (ls)2(2s)2(2p)2. In
H this state there are two p valence electrons. But the C atom has the most
bonds via the (ls)2(2s)1(2p)3 configuration, i.e. one 2s electron is shifted into
Fig. 12.16. Structure of the 2p orbital. Now the C atom has four valence electrons, which can bind
the methane (CH 4 ) mole- with electrons of other atoms. We again use the completely hybridized states
cule: One s electron and (12.53a-d) and occupy each of the four states with one electron. These elec-
three p electrons of the C
are hybridized and form a
trons form (5 bonds with the electrons of the H atom and give rise to the
bonds with the electrons methane (CH 4 ) molecule. The angle between the bonds corresponds exactly
of the H to the calculated angle of 109.6 0 (see Fig. 12.16).
In the C 2 H4 molecule, ethylene, a planar structure occurs (see Fig. 12.17).
This is explained by the following hybridization of the orbits:
11) IPz) , (12.55a)

12) IIls) + tilPx) , (12.55b)

13 ) IIIS) - {flpx) + V1lpy) , (12.55c)

14) = IIIS) - {flpx) -lIlpy) . (12.55d)

Fig. 12.17. Structure of

the ethylene (C 2 H 4 ) mole- These states follow from a variational calculation. The first state points in
cule. The x, y, z coordi- ±z direction, whereas the last three states form an angle of 1200 in the x-y
nates are indicated plane. The spatial electron-orbit distribution is illustrated in Fig. 12.18. The
 12.7 Hydrocarbons 357

Fig. 12.18. The hybrid

orbits 12), 13), and 14) of
the C2 H4 molecule [see
(12.55b-d)] form an angle
of 120 0 to each other. The
z axis points out of the
plane

z axis (out of plane)

II)

electrons in states 13) and 14) form a bond with the electrons of the hydrogen
atoms.
The electrons in states 11) and 12) form a bond with the electrons of the
other C atom. The orbits of the second half (CH 2 ) of the C 2 H4 molecule are
mirror-reflected to the first half. The electrons in 12) form a (J bond and the
ones in 11) a 1': bond, since they are oriented perpendicular to the binding axis.
The reason for the planar structure of the C2H4 molecule is the 1': bond. This
bound becomes effective only if both CH 2 parts of the molecule are located
in the same plane. The orbits for the 1': bond are oriented perpendicular to it
(along the z axis in Fig. 12.18) Rotation around the x axis causes the 1': bond
to break up (see Fig. 12.19). Thus, the 1': bond enforces the coplanarity of the
two CH 2 complexes.

Fig. 12.19. The bond

H~ 12)u ~H types in ethylene (C 2 H4).
The 7r bond is mediated by
the 11) orbit, which origi-
C==C
a/ II)rr "- a nates in the respective car-
bon atom and points out
H/I4) I~H of the plane of the paper
(along the z direction).
(~----t~x
For a better understand-
z ing look also at Figs. 12.10
and 12.18
A linear structure is observed in the acetylene molecule C 2 H 2 . This
molecule is of the form H-C == C-H and is oriented along, for example, the x
axis. The lIs) state now only mixes with Ipx) in the form

(12.56)

Because of the positive and negative interference of the wavefunctions, we get

one state aligned in the +x direction and another one in -x direction. The
remaining Ipy) and IPz) states form 1': bonds. This is illustrated in Fig. 12.20.
The two 1': bond orbits are oriented in the first case in the y and in the second
case in the z direction. It is perhaps useful to have a look again at Fig. 12.10,
describing the 7r bond.
358 12. Elementary Structure of Molecules

Fig. 12.20. The linear y

structure of the acetylene
molecule (C 2 H 2 ) along the H _-..::u_ _ c _-..::u__ c __u;;;""_H
x axis. The two 7r bonds
between the carbon mole-
cules come from the Ipy) (!!fo----'l~x
and IPz) orbits, respec- z
tively, and are oriented
along the y axis and z axis,
respectively

H In the carbon compounds discussed so far all bonds were localized, i.e. we
know which bonds appear between which atoms. A particularly interesting
situation appears in the case of the "aromatic" compound benzene. It consists
of a ring with six carbon atoms, on each of which a hydrogen atom is bound
(see Fig. 12.21).
The bonds C-H are all hybridized a bonds. One bond between the C atoms
is also of a type. Hence, six out of eight valence electrons of two neighboring
carbon atoms are involved in a bonding, the other two will form a 11' bond.
c Therfore, between the six C atoms of the benzene-ring only three 11' bonds
U
H
will be possible. This is illustrated in Fig. 12.21 by thicker lines. The states of
the electrons causing the 11' bonds are oriented perpendicular to the a bonds,
Fig. 12.21. The structure
of the benzene molecule i.e. out of the plane of the benzene ring. Because of the symmetry it is not
CsHs. It is ring-shaped possible to say between which C atoms the 11' bond appears. In Fig. 12.22 two
and lies in a plane possibilities are depicted.
Indeed, the state with lowest energy is represented by a mixing of these
two possibilities. The electrons of the 11' bonds can be visualized as moving
~c
c ]'( c freely in the ring. Due to the larger extension available for these electrons
]'(1 their kinetic energy decreases (uncertainty principle!). With the lower kinetic
energy the state is additionally lowered and, hence, stronger bonds result.
c c
~C Summarizing, we can say that one can already reach good qualitative
understanding of the structure of molecules by simple arguments such as the
or overlap of the electrons in singlet states and hybridization or by increasing the
number of valence electrons by exciting an atom (e.g. carbon atom). But this
procedure is not sufficient to make accurate, quantitative predictions of the
structure of molecules. This can be accomplished only by large-scale computer
calculations.

Fig. 12.22. Two equal 12.8 Biographical Notes

possibilities for the 7r bond
between the C atoms in
the benzene ring. Because VAN DER WAAL, Johannes D., *Leiden 23.11.1837, t Amsterdam 9.3.1923,
of the symmetry of the Dutch physicist. He worked as a teacher, whose science was largely self-taught.
wavefunction one cannot He studied at the University of Leiden and became physics professor in 1877
distinguish between the at the University of Amsterdam. In 1910 v.d.W was awarded the Nobel Prize
two cases in physics for his work on the equation of state of gases and liquids.

BORN, Max, *Breslau 11.12.1882, tG6ttingen 5.1.1970, German physicist.

B. was the son of an anatomy professor at Breslau. He was educated at the
 12.8 Biographical Notes 359

universities of Breslau, Heidelberg, Zurich, and G6ttingen, where he reveived

his PhD in 1907. With the rise of Hitler he moved to Britain, returning in
1953. Born shared with W. Bothe the 1954 Nobel Prize for physics. On his
tombstone in G6ttingen on finds the fundamental equation of quantum me-
chanics: pq - qp = h/27ri.

OPPENHEIMER, J. Robert, *New York City 22.4.1904, tPrinceton

18.2.1967, American physicist. O. came from a wealthy family. He studied
at Harvard, Cambridge, and G6ttingen, where he obtained his PhD in 1927.
He worked on the development of the atom bomb in Los Alamos during World
War II. In 1963 he received the Fermi award.

LONDON, Fritz, *Breslau 7.3.1900, tDurham 30.3.1954, German physicist.

L. was the son of a mathematics professor at Bonn and the brother of the
physicist Heinz London. He studied in Frankfurt/Main, Munich, and Bonn
where he received his degree in philosophy in 1921. He mainly worked with
Heitler on superfiuidity, superconductivity, and wave mechanics.
13. Feynman's Path Integral Formulation
of Schrodinger's Wave Mechanics

In this final chapter we present an alternative formulation of quantum mechan-

ics operating with new mathematical methods: the so-called path integrals.
The first step towards such a description were made by Dirac, but the math-
ematical foundation and beauty was put forward by Feynman. It contributes
essentially to a fundamental understanding of quantum mechanics and allows
a derivation of exact equations in complex quantum field theory. It should be
noted from the beginning that up to now path integral formulations have not
played such an important role for solving certain field-theoretical problems; on
the one hand, analytical solutions are only possible in very simple cases, and
on the other hand, numerical calculations are extremely computer intense.
However, path integrals often allow an approximate solution for physical pro-
cesses, such as phase transitions, where perturbative methods fail.
In the following, we first remember the connection between classical me-
chanics and Schrodinger's wave mechanics in order to emphasize the impor-
tance of the action S. It turns out that the action represents the central
quantity that enters into the path integral description of transition ampli-
tudes (Greens functions). Its derivation implicitely assumes Born's probabil-
ity interpretation of Schrodinger's wave field tP' as "background knowledge".
After we exclusively derive the transition amplitudes with the help of the
path integration, we will derive the path integral representation from the
conventional description of Schrodinger's Greens function. This shows the
equivalence of both formulations. Taking for granted the connection between
wavefunction and transition amplitude, we will derive the Schrodinger equa-
tion in the framework of the path integral formulation. For further reading
see the literature. 1

1 R.P. Feynman, A.-R. Hibbs: Quantum Mechanics and Path Integrals (McGraw-
Hill, New York 1965); C. Itzykson, J.-B. Zuber: Quantum Field Theory (McGraw-
Hill, New York 1980); W. Greiner, J. Reinhardt: Field Quantization (Springer,
Berlin, Heidelberg 1996).

W. Greiner, Quantum Mechanics

© Springer-Verlag Berlin Heidelberg 1998
362 13. Feynman's Path Integral Formulation of Schriidinger's Wave Mechanics

13.1 Action Functional in Classical Mechanics

and Schrodinger's Wave Mechanics
We begin with a short review of the classical action S, which plays a central
role for the path integral formalism. An elegant access to the description of a
classical particle is offered by Hamilton's action principle. The starting point
is the action functional. L is the Lagrangian of the system and ta and tb are
two fixed times; then
(tb
S[b, a] = S[X(tb), x(ta)] = it" dt' L(x(t'), ±(t'), t') (13.1)

defines the so-called action functional, which assigns a defined value for any
path x(t) connecting the fixed end points Xa = x(t a ) and Xb = X(tb). Within
an infinite number of such paths the classical trajectory of a particle is rep-
resented by the special path x(t), for which the action (13.1) takes on an
extremum, more precisely a minimum. The necessary condition for this is
that for a small variation 8x of the classical path x the corresponding action
remaines unchanged in first order in 8x:
8S = S[x + 8x] - S[x] = o. (13.2)
As we know this leads to

8S = 8x aL
a±
l
t
t"
b + t
ita
b
dt' (aL _
ax
~ aL) 8x.
dt' a±
(13.3)

For fixed end points 8x(ta) = 8X(tb) = o. Hence the Euler Lagrange equation
for the classical trajectory follows:
aL d aL
ax - dt ax = O. (13.4)
In classical mechanics the form of the action functional is most interesting,
and not the value of the action itself. We consider the action along a classical
trajectory as a function of the upper bound

Sclass.[X(t)] = (t dt' L(x(t'),x(t'),t'). (13.5)

ita
From this, the energy and momentum at the end point can be calculated (see
Exercise 13.1):
aSelass .
p (13.6a)
ax
aselass.
E ---- (13.6b)
at
We now turn to quantum mechanics. Schrodinger's concept 2 for the derivation
of an equation for the wavefunction iJ! followed the analogy between quantum
mechanics and classical mechanics, on the one side, and to wave optics and
its borderline case, i.e. geometrical optics, on the other side. If we describe
the complex field iJ! in the form
2 E. Schriidinger: Ann. d. Physik 79 (4) (1926) 361; 79 (4) (1926) 489.
13.1 Action Functional in Classical Mechanics and SchrCidinger's Wave Mechanics 363

lli(x, t) = a(x, t) e iS [xl/1l (13.7)

the wave equation for lli has to be fixed in such a way that, first, la(x, tW ==
p(x, t) = Illilli* I fulfills the continuity equation, and that, second, the Hamil-
ton Jacobi equation ("eikonal equation") emerges in the limit Ii ----> 0 for S.
This classical limit of wave mechanics strengthens the importance of the ac-
tion. Nevertheless Schrodinger's equation, understood as a wave equation, has
only a partial connection to quantum mechanics. Only Born's interpretation
of the field lli as a guiding field for the particles, i.e. of lli being the probabil-
ity amplitude and Illil2 the probability density, established the fundamental
significance of the Schrodinger equation

(iii :t - iT) Illi(t)) = O. (13.8)

From this it becomes plausible that the action S(x) enters somehow into the
phase of probability amplitudes.
These preliminary remarks suffice and we now introduce the Schrodinger
propagator for later use. We define the Greens operator by

( ata,)
iii
, t')
- G(t, H =
~
iii 1 5(t - t') . (13.9)

The spatial Greens function

G(x, tj x', t') = (xIG(t, t')lx')
fulfills the equation

(iii! - iTx) G(x, tj x', t') = iM(x - x')5(t - t') . (13.10)

Once G is known, the time-dependent solution Illi(t)) can be deduced accord-

ing to
Illi(t)) = G(t, to)llli(to)) . (13.11)
For the case of a Hamiltonian iT that does not depend explicitly on time, a
formal solution of (13.9) can be given immediately:

G(t,t') = B(t-t')exp(-~iT(t-t')). (13.12)

The Greens function in space and time results from this as a matrix element

G(x, tj x', t') = B(t - t')(xl exp ( -~iT(t - t')) Ix') . (13.13)

From here we will derive later the path integral description for the Greens
function (13.13).
364 13. Feynman's Path Integral Formulation of Schrodinger's Wave Mechanics

EXERCISE

13.1 Momentum and Energy at the End Point

of a Classical Trajectory

Problem. Determine the momentum and energy at any end point of a clas-
sical trajectory.

Solution. The starting point is the classical action along the trajectory x(t)
as a function of the upper bound:

Sclass.[X(t),xoj = t dt' L[x(t'),x(t'),t'] .

lto
(1)

Generally, the variation of the action is then

8S = 8x fJL
fJi;
It
to
+ rt dt' 8x (fJLfJx _ ~
lto
fJL)
dt' fJi;
(2)

Along the classical trajectories the integral term vanishes. Assuming, further-
more, 8x(t o) = 0 and p = fJLlfJi; yields
8S=p8x. (3)
This equation has to be understood in the following way. It compares the
variation of the action with respect to paths that differ by the endpoints x(t)
taken at the same time t. Equation (3) tells us that the partial derivative of
the action with respect to the coordinate of the upper end point is equal to
the corresponding momentum, i.e.

(4)
Analogously, we can take the action as an explicit function of time as we
consider paths that end in the given location x at different times t.
The definition of the action (1) means the total time derivative of the
action along a trajectory is

dS~I;SS' = L[x(t),x(t),tj. (5)

On the other hand one has

-- + -
dS fJSclass. fJSclass. -'-
dt class. - --x (6)
fJt fJx
if we understand Sclass. in the above mentioned sense as a function of coordi-
nates and time. From this it follows that
fJS .
at = L - px = -E. (7)
 13.2 Transition Amplitude as a Path Integral 365

13.2 Transition Amplitude as a Path Integral

Consider, for example, the double-slit experiment in quantum mechanics. An

electron source Q is mounted in point a. At the screen S there appears an
interference pattern, which is described by Itli1 2 . We seek the probability am-
plitude K(b, a) with which a particle propagates from point a to point b. Now,
in quantum mechanics not only the classical trajectories with minimal action
contribute to the probability amplitude in b, but also all other possible paths
joining the points a and b. We are led to the following postulates.

1) All possible paths contribute equally, Le. formally in the same way as the
amplitude; but different paths contribute with different phases.
2) The phase of the contribution of a given path is determined by the action
S along this path (measured in Ii).

Fig. 13.1. Paths in the

double-slit experiment

The probability P(b, a) of reaching point Xb at time tb from point Xa at

time ta is defined by the absolute square P(b, a) = IK(b, a)12 of an amplitude
K(b, a). Futhermore the amplitude is given by a sum of contributions ¢[x(t)]
of all paths joining a and b, i.e.
K(b, a) = ¢[x(t)] . (13.14)
sum over all paths
from a to b

The contribution of each path has a phase proportional to the action S along
the path, i.e.

¢[x(t)] = cexp{ks[x(t)]}. (13.15)

Here C is a normalization constant that has yet to be determined. It is the

same for all paths, since all paths contribute with the same weight. Before
we go into detail with the mathematical tools and specifications, in particular
before we explain the meaning of what a "sum over all paths" is in (13.14),
we shall clarify why in the classical limit one particular path, i.e. the classical
trajectory, is most important: the classical limit means S » Ii. Now, every
path contributes in an equal manner, but for the classical path SIIi is very
large. The classical path x(t) is the one for which the phase in (13.15) becomes
366 13. Feynman's Path Integral Formulation of Schrodinger's Wave Mechanics

Fig. 13.2. The classical

path x(t) and neighboring tb
paths xv(t)

Xa Xb x

stationary. Hence, infinitesimal neighboring paths have the same phase in first
order. Compared to the classical one other paths lead to strongly oscillating
phases (see Fig. 13.2). The corresponding contributions in (13.14) interfere de-
structively and cancel each other. Therefore the amplitude (13.14) essentially
reduces to

K(b,a) ~ J(b,a)exp (~Sclass.) , (13.16)

where J(b, a) will be a weakly changing function.

We will now discuss the sum contained in (13.14) in more detail and we
will be guided by the analogy to Riemann's definition for an integral. For this
we discretize the time interval T = tb - ta into N equal partial intervals of
length E:

NE tb - ta , tb > ta ,

E ti+1 - ti , i=l, ... ,N-1,

(13.17)
to ta , tN = tb,
Xo Xa , XN = Xb·

Let X a , ta and Xb, tb be the two fixed end points. We now construct a certain
path by choosing special points Xi for all intermediate time points ti and
connect the selected points by straight lines (see Fig. 13.3). By refining this

t
~+---------------L

ti+l +-___________ ~

tj +-----------R-

Fig. 13.3. A discretized ta+---.....~

path from (Xa, ta) to
(Xb, tb)
Xa Xi Xb x
Xi+)
 13.2 Transition Amplitude as a Path Integral 367

lattice, we can approximate every path with any desired quality. Hence, it is
natural to define the sum over all paths as a multiple integral over all values
of Xi (i = 1, ... , N - 1; XQ, XN fixed):

K(b, a) rv J dXl ... dXN-l <f>[x(t)].

Let us now define the sum implied in (13.14) more precisely. In Riemann's
integral E can be made smaller and smaller. Here one cannot directly generate
a well-defined limiting value in this way. In order to force convergence, a
normalization factor A(E), depending on E, has to be introduced. Summation
leads to the path integral:

K(b,a) =
. 1
!~ A
JA'"
dXl dXN-l
- A - exp
{ i
/is[b,a]
}
, (13.18)
N--->oo

with the action

S[b, a] =
Ito
tb dt' L(x, x, t') .

S is a line integral along a route dXl ... dXN-l through the points Xi. The
above construction of the path integral is not unequivocal. Instead of connect-
ing the Xi by straight lines it is more useful in general to choose sections of
classical paths. Such a construction is possible for any Lagrange function.
We introduce the abbreviated notation:

K(b, a) = lb Dx(t) exp { ~S[b, a]} . (13.19)

For the amplitude (13.18) and (13.19), an important property can be derived.
Let us consider the path integral for two successive events (see Fig. 13.4) with
ta < tc < tb. The action fulfills the obvious property:
S[b, a] = S[b, c] + S[c, a] .

tc+----~.""

Fig. 13.4. Two succes-

sive events (xc, t c) and
a (Xb, tb) and the corre-
sponding paths
Xa Xc Xb X
368 13. Feynman's Path Integral Formulation of Schr6dinger's Wave Mechanics

Now we consider (13.18):

K(b, a) = lim ~ j dXI ... dXM-I dx c dXM+! ...
<--->0 A A A A A
N--->oo

dXN-I
x -A-exp (iIi(S[b,c] + S[c,a]) ) , (13.20)

with ta < tl'" tM-I < tM == tc < tM+!'" < tb·

The integrations can be performed in any order, in particular as follows.
First, for a fixed Xc, S[a, c] will be a constant with respect to the integrations
over XM+! up to XN-I' The same is valid for S[c, b] with respect to the
integrations over Xl up to XM-I. The integration over the point in between,
Xc, is performed at the end. Rewriting (13.20) accordingly yields

K(b, a) = j dx c l~ (~j d~l ... dX~_1 exp {~s[c,a]})

1 j -A--
x (A dXN-I exp { Iii S[b ,c]}) ,
dXM+! ... -A--

which can also be denoted as

K(b, a) = j dx c K(b, c)K(c, a) . (13.21)

Therefore the amplitude K(b, a) can be calculated in such a way that first
the path integration is performed from a to a fixed intermediate point c and
afterwards from c to b and then over all possible intermediate points c. Equa-
tion (13.21) expresses therefore, how amplitudes of successive time events are
multiplied. We can immediately generalize this to

K(b,a) = jdXI ... dXN_I ... dXi ... K(b,N-I)K(N-I,N-2) ...

xK(i+l,i) ... K(l,a). (13.22)
For an infinitesimal time interval f = ti+! - ti between points Xi+! and Xi the
following is valid up to first order in f:

K(i+l,i) = ~exp(~S[i+l,il)
~exp [~lti+l dt' L[X(t'),X(t'),t']]

~
1
A [ifL(XHI+Xi XHI-Xi
exp Ii 2' f ' 2
f)] (13.23)

Thus the integrand from (13.22), i.e. the amplitude for any complete path,
can be written as
N-I
¢[x(t)] = lim
<--->0
II K(i + 1, i) . (13.24)
N--->oo i=O

Let us specify (13.22) for the case of a particle's motion in a potential V(x).
According to the discretized form implied by (13.23) the corresponding La-
grangian reads
 13.2 Transition Amplitude as a Path Integral 369

L = ; (Xi+1C- Xi) 2 _ V(Xi) . (13.25)

Insertion into (13.22) yields according to (13.23)

J
r-
K(b,a) = lim dXl ... dXN-l
<--+0
N-->oo

X A1N exp [~~ ( ; (Xi+lf - Xi V(Xi)) 1; (13.26)

this is the probability amplitude of finding a particle at b coming from a. The

sum in the exponent of (13.26) has to be understood as a Riemann sum of
the action integral along a certain path. This will be further illuminated in
the following exercise.

EXERCISE

13.2 The Transition Amplitude for a Free Particle

Problem. Determine the transition amplitude K(b, a) for a free particle with
the help of (13.26).

Solution. The starting point is (13.26) with V = 0 (free particle):

K(b,a) =!~ J 1 [. N-l

dXl ... dxN-lANexp ~~ L(Xj+l-Xj)2 ,
1 (1)
N->oo J=O

with Xo = Xa, XN = Xb, and N E = (tb - ta). The various integrations appearing
in (1) are reducible to simple Gaussian integrals by means of the quadratic
supplement in the exponent. The successive performance of the N - 1 inte-
grations leads to a set of Gaussian integrations:

1 00

-00
dye a (x- y )2+.6(z-y)2 = (~) 1/2 exp [~(X
a. + j3 a. + j3
_z)2] (2)

1:
Let us begin with the integration over Xl (J-L = im/2nf):

dXl el'(X2- x ,)2+I'(XI-Xo)2 = ( ; ; ) 1/2 el'(X2- Xo)2 /2 .

1:
Integration over X2 gives:
dXl dX2 el'[(X3- X2)2+(X2 -x,)2+(Xl -xo)2]

( -7r) 1/2
2J-L
1
-00
00
dX2 el'(X2-Xo)2 /2+I'(X3- X2)2

( ; ; ) 1/2 (3:;2) 1/2 el'(X3- Xo)2 /3

( (-7r)2) 1/2 el'(X3-Xo)2 /3 .

3J-L2
370 13. Feynman's Path Integral Formulation of Schrodinger's Wave Mechanics

Exercise 13.2.

i:
Performing the N - 1 integrations one after the other yields the expression

dX1··· dXN-l exp {J.l [(XN - xN_d 2 + ... + (Xl - xo)2]}

= ~ (-J.l7r) (N-1)/2 e/1>(XN-X o )2/N . (3)

The amplitude (1) then takes the following form:

K(b,a) = ~~
. (1)
A(f)
N (27rill£) N /2
~
N ..... oo

x ( m )1/2 exp [im..:.(X-;-b_-_X_a-.;,)_2] (4)

27rili( tb - t a) 21i (tb - t a )
But now the limits f --+ 0, N --+ 00 should exist. The only way to guarantee
the convergence is to fix the normalization factor:

A(f) = C::lif) 1/2 (5)

We will derive this factor in another way in the next subsection.

Then the result is
K b m 1/2 [.1m (Xb - Xa )2]
( ) (6)
( ,a) = 27rili(tb _ ta) exp 21i (tb - ta) .
It is remarkable that this expression is identical with the spatial representation
of the Greens function of the one-dimensional free Schrodinger equation. 3

13.3 Path Integral Representation

of the Schrodinger Propagator

Now we will represent the one-dimensional Greens function (13.13) as a path

integral. It is given by

G(Xata;Xbtb) = (xblexp(-kHT)iXa) , T=tb-ta>O, (13.27)

with H= t + V. The following identity holds for the operator function:

, 'N
eO = (eO/ N ) . (13.28)
If we substitute A = i(tb - ta)/Ii into (13.27), the Greens function becomes
G(xata; Xbtb) = lim (xbl e-A(T+V)/N e-A(T+V)/N ... e-A(T+V)/Nlx a ) .
N ..... oo
We now apply the product formula (see Exercise 13.3):
3 See e.g. W. Greiner, J. Reinhardt: Quantum Electrodynamics, 2nd ed. (Springer,
Berlin, Heidelberg 1994).
 13.3 Path Integral Representation of the Schrodinger Propagator 371

lim [(e->-(T+V)/N)N _ (e->-T/N e->-V/Nt] = 0, (13.29)

N--->oo

and obtain:
G(Xata;Xbtb) = lim (xbl(e->-T/N e->-V/N)Nlxa). (13.30)
N--->oo

From here only a few more steps lead to the path integral. Inserting a complete
set of spatial states

i = J dXi IXi)(Xil , i=1, ... ,N-1,

yields

lim
N--->oo
J dXl ... dXN-l
N-l
X II (xi+11 e->-T/N e->-v/Nlxi) , (13.31)
i=O

where Xo = Xa and XN = Xb. Now the matrix elements appearing in (13.31)

have to be determined. Since the operator for the potential energy V is diag-
onal in space, we have

(13.32)
In order to calculate the spatial matrix element of the operator e->-T/N we
insert a complete set of momentum eigenstates:

(xi+11 e->-T/Nlxi) = J dp (xi+11 e->-T/Nlp)(plxi)

J dp (xi+1lp)(plxi) e->-p2 /2mN

_1_
21rh
/00 dpe->-p2/ 2mN eip
-00
(X;+l- X ;) .

1:
Such Gaussian integrals can be calculated, resulting in

dx e-<>x 2+,Bx = If; e,B2 /4<> .

Thus, we obtain

'/ __ (mN) 1/2 [-mN(Xi+l - Xi)2] (13.33)

(Xi+ll e->-T Nlxi) 21rAh exp 2Ah2

Inserting (13.32) into (13.31) results in

We insert E = (tb - ta)/N = hA/iN and sum the exponential expressions:

372 13. Feynman's Path Integral Formulation of Schrodinger's Wave Mechanics

lim
<->0
J dXl .. . dXN-l ( m )~2
--.-
27rlru:
N->CXJ

x exp {~~ [ ; (Xi+! - Xi)2 - V(Xi)] } (13.34)

With this we have deduced the path integral expression for the Schrodinger
propagator. It is identical in form with the expression for the transition ampli-
tude K(b, a) in (13.26). Now the undetermined normalization factor appearing

r
in (13.26) can be identified as

A(t:) = C%lit: /2
(13.35)

Note that we were led to the Greens function K(b, a) of the Schrodinger
equation exclusively by the path integration method.

EXERCISE

13.3 Trotter's Product Rule

Problem. Prove Trotter's product rule:

J~CXJ { (ex p [- ~ (1' + 11)]) N

_ (exp [_ ~ 1'] exp [_ ~ 11]) N } o. (1)

Solution. We first show that the two operator functions

A
and with a = -
N
differ only by commutation terms, which vanish in the limit N -; 00.
An operator function is defined by its Taylor series, e.g.

C(a) = ~ (a)n
~ n!
(dnC)
dan
I (2)
n=O a=O
In the following, a useful operator identity will be applied:
, aA' -aA ~ (a)n , ,
K(a) = e Be = ~ - , [A,B](n), (3)
n.
n=O
with [A,E](o) = 13, [A,E](1) = [A,E], [A, 13](2) [A, [A, 13]], .... For the
proof of (3) the coefficients (dnkjdan)la=o of the Taylor series have to be
calculated:

k(a) = ~ (a)n
~ n!
(d nk)
dan
I .
n=O a=O
 13.3 Path Integral Representation of the Schrodinger Propagator 373

Thus: Exercise 13.3.

n=O
K(O) B',
n=1
dK (4)
do:
= eaA[A, B] e- aA , (5)

dK I = [A, B](1) = [A, B] ; (6)

do: a=O

n=2

AeaA[A, B] e- aA - eaA[A, B]A e- aA (7)

= eaA[A, [A, Bll e- aA , (8)

d2 KI
d0: 2 a=O
= [A, [A, Bll = [A, B](2) . (9)

For any n one has

dn K I
-n
" [A, Ell
= [A,··, , (10)
do: a=O "---v---'
n tinles

Inserting (10) into the Taylor series for K(o:) yields the identity (3).
Turning to the operator function 0(0:) = e- aT e- av and calculating ex-
plicitly the first terms of its Taylor series, we obtain:
n=O
O(o:)la=o i·,
n=1
dO
(-)i'0(0:) + (_)e-aTVe- aV (11)
do:
(_ )i'0(0:) + (-) e-aTV eaT e- aT e- av (12)

, ,
(-)TG(o:) ~ -----:mr-[T,v](m)
+ (-) ('V + ~ (_o:)m " ) G(o:)
' , (13)

(-)(i' + V)O(o:) + (-) f

m=l
(-o:t [i', V](m)O(o:) ,
m.
(14)

(15)
374 13. Feynman's Path Integral Formulation of Schrodinger's Wave Mechanics

Exercise 13.3.
" ~ (-a)m " ) dC
( (-I)(T+ V) + (-1) ~l ~[T, VjCm) da (16)

00 ( )m-l
2 '" -a " ,
+ (-1) ~l (m _ I)! [T, VjCm)G(a) , (17)

(18)

In the way indicated, all higher derivatives can be determined. Then one gets

dnCI
dan 0=0 = (-I)n(T
' + v)n
, + {commutator terms}.

Inserting this into the Taylor expansion (2) and performing the summation
one obtains
2
C(a) = F(a) + ~ [T, Vj + 0(a 3 ) (19)
Hence we find
[F(a)t - (C(a))N = 0(a 2 ), (20)
i.e. the above difference is at least proportional to a 2 = ).,2/ N2. In the limit
N --> 00 the right-hand side of (20) vanishes, which proves the validity of
Trotter's formula (1).

13.4 Alternative Derivation of the Schrodinger Equation

In the preceding subsection we convinced ourselves that the path integral for
the transition amplitude is identical with the Greens function for the one-
dimensional Schrodinger equation

( - 2m a2 + V(x, t) ) P(x, t) = in ata P(x, t) .

n2 ax2 (13.36)

The Greens function K(b, a), or if we specify the arguments in more detail,
K(xt, x't'), represents the probability amplitude for the propagation of a par-
ticle from place x' at time t' to x at a later time t. The probability amplitude
for finding a particle at time t at position x is described by the wavefunc-
tion P(x, t). If we are not interested in the particle's past, we can define a
wavefunction by

P(x, t) = J
dx' K(xt, x't')P(x', t') . (13.37)

The integral core K(xt, x't') propagates the wavefunction from time t' to time
t. We now consider the special situation in which t and t' differ only by an
 13.4 Alternative Derivation of the Schrodinger Equation 375

infinitesimal time interval f. Applying the corresponding expression (13.23)

for the transition amplitude results in

ll/(x, t + f) =
['>0 dx , exp [if
Ii1 Loo li,L (x+x' x-x' f)]
- 2 - ' - f - ' t + "2 If/(x, t).
(13.38)
The Lagrangian is given by

L = ; j;2 - V (x, t) .

Hence, we obtain explicitly

ll/(x, t + f)
= 1 -00
00,1
dx - exp [im(
A
-
2lif
x - x ')2 - -if V (x+x'
Ii
--, t
2
+ -f)]
2
If/ (x' ,t ) .
(13.39)
The exponent includes all phases (x - X')2 / f. It leads to strong oscillations in
the case of large deviations x' - x = ~. Consequently, the main contribution
to the integral is expected for close points x and x'. After substitution of
x' = x +~,
ll/(x, t

= i:
+ f)
d~~expG:e)exp[-~v(x+~,t+~)]If/(x+~,t),
we expand the integrand and the left-hand side of the above equation, keeping
only terms up to linear in f and quadratic in ~. This leads to

i:
The integrals are given by

d~exp G:e) C%lif) 1/2 (13.40)

1 00
-00 d~ exp e~
em) 0, (13.41)

i:
2lif

d~exp G:e) e i~ C%lifr/2 (13.42)

Thus we obtain, up to 0(1'2),

If/(x, t)
a
+ I' at If/(x, t) (13.43)

= Ii1 [( --:;:;:;:-
27rilif) 1/2 (
If/(x,i
t) f a
ilif ax
- Ii, V(x, t)ll/(x, t) + 2m
2
2 1f/(x, t) ) 1.
376 13. Feynman's Path Integral Formulation of Schrodinger's Wave Mechanics

The normalization factor A(f) has to be choosen in such a way that (13.43) is
fulfilled in the limit f ---+ O. Therefore we have to choose A(f) = (27rnf/m)-1/2,
which we have seen already [see (13.35) and Exercise 13.2J. It follows that
8 k ~ 82
tIt + f 8t tIt tIt - fi VtIt - 2i m 8x 2 tIt .
In first order in f this equation is fulfilled only if
8 n2 82
in-tIt = - - - t I t + VtIt
at 2m 8x 2
is valid. This is the Schrodinger equation.

13.5 Biographical Notes

DIRAC, Paul Adrien Maurice, English physicist, 'Bristol 8.8.1902, tFlorida

20.10.1984. Since 1927 D. was a member of St. Johns College at the University
of Cambridge and from 1932 he was professor of mathematics and physics.
From 1953 he was professor of mathematics and physics at the University
of Oxford. Together with E. Schrodinger, Dirac obtained the Noble Prize for
physics in 1933. His main work was quantum mechanics and nuclear physics.
He set up the relativistic wave equation with spin 1/2 (Dirac equation).

FEYNMAN, Richard Philips, American physicist, 'New York 11.5.1918,

tLos Angeles 15.12.1988. Between 1950 and 1988 Feynman was professor at
the California Institute of Technology in Pasadena. He formulated quantum
electrodynamics and developed a graphical description for calculating com-
plicated field-theoretical processes (Feynman diagrams). In 1965 he received
the Noble Prize for physics.
Subject Index

Absorption 49 Dirac equation 317

Action 362 Dispersion 64
Aharonov-Bohm effect 21 Dispersion relation 207
Annihilation operator 13 Dissipation of energy 65
Anticommutation relation 83 Distribution function 242

Black body radiation 53 Electric dipole transitions 38

Bogoliubov transformation 179, 199 Electron pairing 349
Boltzmann equation 166 Emission 33
Born approximation 167 Ensemble 256
Born-Oppenheimer approximation Entropy 168,181,255,278
337 Equal-time commutation relations 100
Bose gas 218 Euler Lagrange equation 362
Bose-Einstein condensation 177
Bose-Einstein distribution 168,283 Fermi distribution 282
Bose-Einstein statistics 83,98,281 Fermi energy 226,294
Boson 83 Fermi gas 215
Breit-Wigner distribution 78 Fermi momentum 226
Bremsstrahlung 112 Fermi's golden rule 33,166
de Broglie wavelength 23 Fermi-Dirac distribution 168
Fermi-Dirac statistics 93,98, 281
Canonical ensemble 280 Fermion 83
Casimir effect 11,133 Feynman diagrams 56
Charge renormalization 156 Field operator 99
Cherenkov angle 73 Field oscillators 12
Cherenkov radiation 63,115 Fugacities 280
Clebsch-Gordan coefficients 43
Coherent state 16 Gauge transformation 2, 21
Collision invariants 172 Glauber state 16
Compton effect 59 Grand canonical ensemble 280
Compton wavelength 153 Greens function 363
Continuity equation 1 Gyromagnetic factor 328
Cooper pairs 226,234
Coulomb gauge 3,4,32 H theorem 170
Creation operator 13 Hamilton's action principle 362
Cross section 119 Hanbury-Brown and Twiss effect
223
Debye frequency 233 Hartree approximation 244
Debye screening 250 Hartree method 292
Density operator 255,256 Hartree-Fock exchange term 304
Dielectric function 245 Hartree-Fock method 297
Dielectric medium 63 Hund's rules 309
Diffraction 23 Hybridization 353
Dipole approximation 36, 44 Hyperfine structure 43
378 Subject Index

Infrared catastrophe 121 Quantum electrodynamics 1

Interacting particles 161 Quantum field 109
Interference experiment 26 Quantum field theory 161
Quantum gas 165
Lamb shift 149,157 Quantum liquid 193
Landau damping 249 Quantum statistics 255
Larmor equation 38 Quasiparticle operator 179
Liouville equation 273 Quasiparticles 193,202, 241
Lorentz force 21,28
Lorentz gauge 2 Radiation field 32, 109
Reduced matrix element 315
Mach angle 72 Renormalization 133,143
Magnetic flux 24 Rotational energy 336
Many-particle system 31 Rotons 207
Mass renormalization 149,156 Rutherford scattering cross section
Master equation 166 121
Maxwell equations 1, 5
Mean-field 292 cr bond 352
Minimal coupling 31 Schrodinger equation 99
Mixed states 256 Second law of thermodynamics 170
Molecular orbit 342 Second quantization 112
Momentum operator 13 Selection rules 38
Multipole radiation 36 Self-consistent field 241
Self-energy 74
n-particle Schrodinger equation 108 Shannon entropy 189
Naturallinewidth 49,74 Singlet state 346
Nonlocal Schrodinger equation 301 Spin-orbit interaction 312
Nonrelativistic particles 31 Spin-statistics theorem 98
Normal modes 5,9 Spontaneous emission 34
Number operator 12,13 Stark effect 329
Stimulated emission 34,49
One-particle density matrix 214 Stirling's formula 186
One-photon decay 46 Stoke's theorem 24
Superfluidity 193
7rbond 352
Pair correlations 213 Thomas-Fermi method 293
Para statistics 87 Thomson's scattering cross-section 59
Particle field 109 Transition matrix element 33
Partition function 279 Transition probability 33
Paschen-Back effect 328 Triplet state 347
Path integral 28,361 Two-fluid model 193
Pauli blocking 215 Two-particle correlation function 214
Pauli matrices 44 Two-photon decay 60
Pauli principle 87, 166
Periodic system 305 Vacuum state 104
Photoelectric emission 50 Variational principle of Ritz 288
Photon flux 49 Vertex 115
Photon scattering 55 Vibrational energy 336
Planck's distribution 54 Virtual photon 77
Plane wave 3 Vlasov equation 244
Plasma 241
Plasmons 246 Wigner function 243
Poisson brackets 84 Wigner-Eckart theorem 313,314
Polarization 58
Poynting vector 3 Zeeman effect 327
Principle value 80 Zero-point energy 11, 12,34,133
Pure state 255 Zero-point oscillations 34