Relativistic Many-Body Theory

Springer Series on
Atomic, Optical, and Plasma Physics
Volume 63
Series Editors
W.E. Baylis
Uwe Becker
Philip George Burke
Robert N. Compton
M.R. Flannery
Charles J. Joachain
Brian R. Judd
Kate Kirby
Peter Lambropoulos
Gerd Leuchs
Pierre Meystre
For further volumes:

http://www.springer.com/series/411
Ingvar Lindgren
Relativistic
Many-Body Theory
A New Field-Theoretical Approach
With 97 Figures
ABC
Ingvar Lindgren
Department of Physics
University of Gothenburg
SE-412 96 Göteborg
Sweden
ingvar.lindgren@physics.gu.se
ISSN 1615-5653
ISBN 978-1-4419-8308-4 e-ISBN 978-1-4419-8309-1
DOI 10.1007/978-1-4419-8309-1
Springer New York Dordrecht Heidelberg London
Library of Congress Control Number: 2011921721
c Springer Science+Business Media, LLC 2011

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are
not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject
to proprietary rights.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

To Eva
Preface
It is now almost 30 years since the first edition of my book together with John
Morrison, Atomic Many-Body Theory [6], appeared, and the second edition ap-
peared some years later. It has been out of print for quite some time, but fortunately
has recently been made available again through a reprint by Springer Verlag.
During the time that has followed, there has been a tremendous development
in the treatment of many-body systems, conceptually as well as computationally.
Particularly the relativistic treatment has expanded considerably, a treatment that has
been extensively reviewed recently by Ian Grant in the book Relativistic Quantum
Theory of Atoms and Molecules [2].
Also, the treatment of quantum-electrodynamical (QED) effects in atomic
systems has developed considerably in the last few decades, and several review
articles have appeared in the field [7, 11, 13] besides the book by Labzowsky et al.,
Relativistic Effects in Spectra of Atomic Systems [5].
An impressive development has taken place in the field of many-electron systems
by means of various coupled-cluster approaches, with applications particularly on
molecular systems. The development during the last 50 years has been summarized
in the book Recent Progress in Coupled Cluster Methods, edited by Čársky, Paldus,
and Pittner [14].
The present book is aimed at combining atomic many-body theory with quantum-
electrodynamics, which is a long-sought goal in quantum physics. The main prob-
lem in this effort has been that the methods for QED calculations, such as the
S-matrix formulation, and the methods for many-body perturbation theory (MBPT)
have completely different structures. With the development of the new method for
QED calculations, the covariant evolution operator formalism by the Gothenburg
atomic theory group [7], the situation has changed, and quite new possibilities
appeared to formulate a unified theory.
The new formalism is based on field theory, and in its full extent the unification
process represents a formidable problem, and we can in this book describe only
how some steps toward this goal can be taken. This book is largely based upon
the previous book on Atomic Many Body Theory [6], and it is assumed that the
reader has absorbed most of that book, particularly Part II. In addition, the reader
is expected to have basic knowledge in quantum field theory that is explained in
vii
viii Preface
books such as Quantum Theory of Many-Particle Systems by Fetter and Walecka [1]
(mainly parts I and II), An introduction to Quantum Field Theory by Peskin and
Schroeder [12], and Quantum Field Theory by Mandl and Shaw [10].
The material of this book is largely based upon lecture notes and recent publi-
cations by the Gothenburg Atomic-Theory Group [3, 4, 7–9], and I want to express
my sincere gratitude particularly to my previous coauthor John Morrison and to
my present coworkers, Sten Salomonson and Daniel Hedendahl, as well as to the
previous collaborators Ann-Marie Pendrill, Jean-Louis Heully, Eva Lindroth, Björn
Åsén, Hans Persson, Per Sunnergren, Martin Gustavsson, and Håkan Warston for
valuable collaboration.
In addition, I want to thank the late pioneers of the field, Per-Olov Löwdin, who
taught me the foundations of perturbation theory some 40 years ago, and Hugh
Kelly, who introduced the diagrammatic representation into atomic physics – two
corner stones of the later developments. Furthermore, I have benefitted greatly
from communications with many other national and international colleagues and
friends (in alphabetic order), Rod Bartlett, Erkki Brändas, Gordon Drake, Ephraim
Eliav, Stephen Fritzsche, Gerald Gabrielse, Walter Greiner, Paul Indelicato, Karol
Jankowski, Jürgen Kluge, Leonti Labzowsky, Peter Mohr, Debashis Mukherjee,
Marcel Nooijen, Joe Paldus, Vladimir Shabaev, Thomas Stöhlker, Gerhard Soff ,
Joe Sucher, Peter Surjan, and many others.
The outline of the book is the following. The main text is divided into three parts.
Part I gives some basic formalism and the basic many-body theory that will serve as
a foundation for the following. In Part II, three numerical procedures for calculation
of QED effects on bound electronic states are described, the S-matrix formulation,
the Green’s-function, and the covariant-evolution-operator methods. A procedure
toward combining QED with MBPT is developed in Part III. Part IV contains a
number of Appendices, where basic concepts are summarized. Certain sections of
the text that can be omitted at first reading are marked with an asterisk (*).
References
1. Fetter, A.L., Walecka, J.D.: The Quantum Mechanics of Many-Body Systems. McGraw-Hill,
N.Y. (1971)
2. Grant, I.P.: Relativistic Quantum Theory of Atoms and Molecules. Springer, Heidelberg (2007)
3. Hedendahl, D.: Towards a Relativistic Covariant Many-Body Perturbation Theory. Ph.D.
thesis, University of Gothenburg, Gothenburg, Sweden (2010)
4. Hedendahl, D., Salomonson, S., Lindgren, I.: . Phys. Rev. p. (to be published) (2011)
5. Labzowski, L.N., Klimchitskaya, G., Dmitriev, Y.: Relativistic Effects in Spactra of Atomic
Systems. IOP Publ., Bristol (1993)
6. Lindgren, I., Morrison, J.: Atomic Many-Body Theory. Second edition, Springer-Verlag, Berlin
(1986, reprinted 2009)
7. Lindgren, I., Salomonson, S., Åsén, B.: The covariant-evolution-operator method in bound-
state QED. Physics Reports 389, 161–261 (2004)
8. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body-QED perturbation theory: Connec-
tion to the two-electron Bethe-Salpeter equation. Einstein centennial review paper. Can. J.
Phys. 83, 183–218 (2005)
Preface ix
9. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body procedure for energy-dependent
perturbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502
(2006)
10. Mandl, F., Shaw, G.: Quantum Field Theory. John Wiley and Sons, New York (1986)
11. Mohr, P.J., Plunien, G., Soff, G.: QED corrections in heavy atoms. Physics Reports 293,
227–372 (1998)
12. Peskin, M.E., Schroeder, D.V.: An introduction to Quantun Field Theory. Addison-Wesley
Publ. Co., Reading, Mass. (1995)
13. Shabaev, V.M.: Two-times Green’s function method in quantum electrodynamics of high-Z
few-electron atoms. Physics Reports 356, 119–228 (2002)
14. Čársky, P., Paldus, J., Pittner, J. (eds.): Recent Progress in Coupled Cluster Methods: Theory
and Applications. Springer (2009)
Gothenburg, Sweden Ingvar Lindgren

Contents
1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 1
1.1 Standard Many-Body Perturbation Theory . . . . . . . . . .. . . . . . . . . . . . . . . . . 1
1.2 Quantum-Electrodynamics.. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 2
1.3 Bethe–Salpeter Equation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 3
1.4 Helium Atom: Analytical Approach . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 4
1.5 Field-Theoretical Approach to Many-Body Perturbation Theory.. . . 5
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 7
Part I Basics: Standard Many-Body Perturbation Theory
2 Time-Independent Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 13
2.1 First Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 13
2.1.1 De Broglie’s Relations . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 13
2.1.2 The Schrödinger Equation . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 14
2.2 Second Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 16
2.2.1 Schrödinger Equation in Second Quantization .. . . . . . . . . . . . . 16
2.2.2 Particle–Hole Formalism: Normal Order and
Contraction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 18
2.2.3 Wick’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 19
2.3 Time-Independent Many-Body Perturbation Theory .. . . . . . . . . . . . . . . . 20
2.3.1 Bloch Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 20
2.3.2 Partitioning of the Hamiltonian . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 21
2.4 Graphical Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 25
2.4.1 Goldstone Diagrams .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 25
2.4.2 Linked-Diagram Expansion .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 28
2.5 All-Order Methods: Coupled-Cluster Approach .. . . .. . . . . . . . . . . . . . . . . 30
2.5.1 Pair Correlation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 30
2.5.2 Exponential Ansatz .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 33
2.5.3 Various Models for Coupled-Cluster
Calculations: Intruder-State Problem .. . . . . .. . . . . . . . . . . . . . . . . 36
2.6 Relativistic MBPT: No-Virtual-Pair Approximation . . . . . . . . . . . . . . . . . 37
2.6.1 QED Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 39
xi
xii Contents
2.7 Some Numerical Results of Standard MBPT and CC

Calculations, Applied to Atoms . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 40
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 43
3 Time-Dependent Formalism .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 47
3.1 Evolution Operator .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 47
3.2 Adiabatic Damping: Gell-Mann–Low Theorem .. . . .. . . . . . . . . . . . . . . . . 51
3.2.1 Gell-Mann–Low Theorem . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 52
3.3 Extended Model Space: The Generalized
Gell-Mann–Low Relation .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 52
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 56
Part II Quantum-Electrodynamics: One-Photon

and Two-Photon Exchange
4 S-Matrix . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 59
4.1 Definition of the S-Matrix: Feynman Diagrams . . . . .. . . . . . . . . . . . . . . . . 60
4.2 Electron Propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 61
4.3 Photon Propagator .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 65
4.3.1 Feynman Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 66
4.3.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 68
4.4 Single-Photon Exchange .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 69
4.4.1 Covariant Gauge .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 69
4.4.2 Noncovariant Coulomb Gauge .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 72
4.4.3 Single-Particle Potential .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 74
4.5 Two-Photon Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 75
4.5.1 Two-Photon Ladder.. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 75
4.5.2 Two-Photon Cross . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 78
4.6 QED Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 79
4.6.1 Bound-Electron Self-Energy .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 79
4.6.2 Vertex Correction .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 82
4.6.3 Vacuum Polarization.. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 84
4.6.4 Photon Self-Energy .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 87
4.7 Feynman Diagrams for the S-Matrix: Feynman Amplitude .. . . . . . . . . 87
4.7.1 Feynman Diagrams .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 87
4.7.2 Feynman Amplitude .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 88
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 89
5 Green’s Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 91
5.1 Classical Green’s Function.. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 91
5.2 Field-Theoretical Green’s Function: Closed-Shell Case . . . . . . . . . . . . . 92
5.2.1 Definition of the Field-Theoretical Green’s Function.. . . . . . 92
5.2.2 Single-Photon Exchange . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 95
5.2.3 Fourier Transform of the Green’s Function . . . . . . . . . . . . . . . . . 96
Contents xiii
5.3 Graphical Representation of the Green’s Function . .. . . . . . . . . . . . . . . . .100

5.3.1 Single-Particle Green’s Function . . . . . . . . . . .. . . . . . . . . . . . . . . . .100
5.3.2 Many-Particle Green’s Function . . . . . . . . . . . .. . . . . . . . . . . . . . . . .105
5.3.3 Self-Energy: Dyson Equation .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .108
5.3.4 Numerical Illustration . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .110
5.4 Field-Theoretical Green’s Function: Open-Shell Case . . . . . . . . . . . . . . .110
5.4.1 Definition of the Open-Shell Green’s Function . . . . . . . . . . . . .110
5.4.2 Two-Times Green’s Function of Shabaev . .. . . . . . . . . . . . . . . . .111
5.4.3 Single-Photon Exchange . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .114
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .117
6 Covariant Evolution Operator and Green’s Operator . .. . . . . . . . . . . . . . . . .119

6.1 Definition of the Covariant Evolution Operator . . . . .. . . . . . . . . . . . . . . . .119
6.2 Single-Photon Exchange in the
Covariant-Evolution-Operator Formalism . . . . . . . . . . .. . . . . . . . . . . . . . . . .122
6.2.1 Single-Photon Ladder . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .126
6.3 Multiphoton Exchange .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .127
6.3.1 General .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .127
6.3.2 Irreducible Two-Photon Exchange.. . . . . . . . .. . . . . . . . . . . . . . . . .128
6.3.3 Potential with Radiative Parts . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .130
6.4 Relativistic Form of the Gell-Mann–Low Theorem .. . . . . . . . . . . . . . . . .131
6.5 Field-Theoretical Many-Body Hamiltonian.. . . . . . . . .. . . . . . . . . . . . . . . . .132
6.6 Green’s Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .134
6.6.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .134
6.6.2 Relation Between the Green’s Operator
and Many-Body Perturbation Procedures . .. . . . . . . . . . . . . . . . .136
6.7 Model-Space Contribution .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .138
6.7.1 Lowest Orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .139
6.7.2 All Orders .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .142
6.8 Bloch Equation for Green’s Operator . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .146
6.9 Time Dependence of the Green’s Operator. Connection
to the Bethe–Salpeter Equation .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .151
6.9.1 Single-Reference Model Space . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .151
6.9.2 Multireference Model Space . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .154
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .156
7 Numerical Illustrations to Part II. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .157

7.1 S-Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .157
7.1.1 Electron Self-Energy of Hydrogen-Like Ions . . . . . . . . . . . . . . .157
7.1.2 Lamb Shift of Hydrogen-Like Uranium .. . .. . . . . . . . . . . . . . . . .158
7.1.3 Lamb Shift of Lithium-Like Uranium . . . . . .. . . . . . . . . . . . . . . . .159
7.1.4 Two-Photon Nonradiative Exchange
in Helium-Like Ions . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .160
7.1.5 Electron Correlation and QED Calculations
on Ground States of Helium-Like Ions . . . . .. . . . . . . . . . . . . . . . .163
xiv Contents
7.1.6g-Factor of Hydrogen-Like Ions:

Mass of the Free Electron . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .164
7.2 Green’s-Function and Covariant-Evolution-Operator Methods .. . . . .165
7.2.1 Fine-Structure of Helium-Like Ions . . . . . . . .. . . . . . . . . . . . . . . . .165
7.2.2 Energy Calculations of 1s2s Levels
of Helium-Like Ions .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .167
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .168
Part III Quantum-Electrodynamics Beyond Two-Photon Exchange:

Field-Theoretical Approach to Many-Body Perturbation
Theory
8 Covariant Evolution Combined with Electron Correlation.. . . . . . . . . . . . .173

8.1 General Single-Photon Exchange . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .173
8.1.1 Transverse Part.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .174
8.1.2 Coulomb Interaction .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .181
8.2 General QED Potential .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .181
8.2.1 Single Photon with Crossed Coulomb Interaction . . . . . . . . . .181
8.2.2 Electron Self-Energy and Vertex Correction . . . . . . . . . . . . . . . .186
8.2.3 Vertex Correction with Further Coulomb Iterations . . . . . . . .192
8.2.4 General Two-Body Potential . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .193
8.3 Unification of the MBPT and QED Procedures:
Connection to Bethe–Salpeter Equation . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .193
8.3.1 MBPT–QED Procedure . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .193
8.4 Coupled-Cluster-QED Expansion . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .196
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .198
9 The Bethe–Salpeter Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .199

9.1 The Original Derivations by the Bethe–Salpeter Equation .. . . . . . . . . .199
9.1.1 Derivation by Salpeter and Bethe . . . . . . . . . . .. . . . . . . . . . . . . . . . .199
9.1.2 Derivation by Gell-Mann and Low . . . . . . . . .. . . . . . . . . . . . . . . . .202
9.1.3 Analysis of the Derivations of the
Bethe–Salpeter Equation . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .203
9.2 Quasi-Potential and Effective-Potential
Approximations: Single-Reference Case. . . . . . . . . . . . .. . . . . . . . . . . . . . . . .205
9.3 Bethe–Salpeter–Bloch Equation: Multireference Case . . . . . . . . . . . . . . .206
9.4 Problems with the Bethe–Salpeter Equation .. . . . . . . .. . . . . . . . . . . . . . . . .208
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .209
10 Implementation of the MBPT–QED Procedure

with Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .211
10.1 The Fock-Space Bloch Equation . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .211
10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs . . . . . . .213
10.3 Single-Photon Exchange: Virtual Pairs. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .216
10.3.1 Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .216
Contents xv
10.3.2 Full Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .219

10.3.3 Higher Orders .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221
10.4 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221
10.4.1 Two-Photon Exchange . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221
10.4.2 Beyond Two Photons .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .221
10.4.3 Outlook.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .224
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .224
11 Analytical Treatment of the Bethe–Salpeter Equation . .. . . . . . . . . . . . . . . . .225

11.1 Helium Fine Structure.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .225
11.2 The Approach of Sucher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .226
11.3 Perturbation Expansion of the BS Equation . . . . . . . . .. . . . . . . . . . . . . . . . .231
11.4 Diagrammatic Representation . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .233
11.5 Comparison with the Numerical Approach . . . . . . . . . .. . . . . . . . . . . . . . . . .235
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .235
12 Regularization and Renormalization .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237

12.1 The Free-Electron QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237
12.1.1 The Free-Electron Propagator .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237
12.1.2 The Free-Electron Self-Energy . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .239
12.1.3 The Free-Electron Vertex Correction . . . . . . .. . . . . . . . . . . . . . . . .240
12.2 Renormalization Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .242
12.2.1 Mass Renormalization .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .242
12.2.2 Charge Renormalization.. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .244
12.3 Bound-State Renormalization: Cutoff Procedures .. .. . . . . . . . . . . . . . . . .248
12.3.1 Mass Renormalization .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .248
12.3.2 Evaluation of the Mass Term. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .249
12.3.3 Bethe’s Nonrelativistic Treatment . . . . . . . . . .. . . . . . . . . . . . . . . . .251
12.3.4 Brown–Langer–Schaefer Regularization . . .. . . . . . . . . . . . . . . . .252
12.3.5 Partial-Wave Regularization . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .255
12.4 Dimensional Regularization in Feynman Gauge.. . . .. . . . . . . . . . . . . . . . .257
12.4.1 Evaluation of the Renormalized Free-Electron
Self-Energy in Feynman Gauge.. . . . . . . . . . . .. . . . . . . . . . . . . . . . .258
12.4.2 Free-Electron Vertex Correction in Feynman Gauge . . . . . . .261
12.5 Dimensional Regularization in Coulomb Gauge . . . .. . . . . . . . . . . . . . . . .263
12.5.1 Free-Electron Self-Energy in the Coulomb Gauge.. . . . . . . . .263
12.6 Direct Numerical Regularization of the Bound-State
Self-Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .267
12.6.1 Feynman Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .268
12.6.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .268
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .269
13 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .271

xvi Contents
Appendices
A Notations and Definitions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .273

A.1 Four-Component Vector Notations . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .273
A.2 Vector Spaces .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .275
A.2.1 Notations .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .275
A.2.2 Basic Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .275
A.2.3 Special Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .277
A.3 Special Functions.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .277
A.3.1 Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .277
A.3.2 Integrals over Functions . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .279
A.3.3 The Heaviside Step Function . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .282
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .282
B Second Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .283

B.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .283
B.2 Heisenberg and Interaction Pictures .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .286
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .287
C Representations of States and Operators . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .289

C.1 Vector Representation of States. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .289
C.2 Matrix Representation of Operators .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .291
C.3 Coordinate Representations .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .292
C.3.1 Representation of Vectors . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .292
C.3.2 Closure Property .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .292
C.3.3 Representation of Operators . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .293
D Dirac Equation and the Momentum Representation . . . .. . . . . . . . . . . . . . . . .295

D.1 Dirac Equation.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .295
D.1.1 Free Particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .295
D.1.2 Dirac Equation in an Electromagnetic Field . . . . . . . . . . . . . . . .300
D.2 Momentum Representation . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .300
D.2.1 Representation of States . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .300
D.2.2 Representation of Operators . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .301
D.2.3 Closure Property for Momentum Functions .. . . . . . . . . . . . . . . .302
D.3 Relations for the Alpha and Gamma Matrices . . . . . . .. . . . . . . . . . . . . . . . .302
Reference .. . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .303
E Lagrangian Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .305

E.1 Classical Mechanics .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .305
E.1.1 Electron in External Field . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .307
E.2 Classical Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .308
E.3 Dirac Equation in Lagrangian Formalism .. . . . . . . . . . .. . . . . . . . . . . . . . . . .309
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .310
Contents xvii
F Semiclassical Theory of Radiation .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .311

F.1 Classical Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .311
F.1.1 Maxwell’s Equations in Covariant Form . . .. . . . . . . . . . . . . . . . .311
F.1.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .315
F.2 Quantized Radiation Field . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .317
F.2.1 Transverse Radiation Field . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .317
F.2.2 Breit Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .318
F.2.3 Transverse Photon Propagator . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .321
F.2.4 Comparison with the Covariant Treatment .. . . . . . . . . . . . . . . . .322
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .324
G Covariant Theory of Quantum ElectroDynamics .. . . . . . .. . . . . . . . . . . . . . . . .325

G.1 Covariant Quantization: Gupta–Bleuler Formalism .. . . . . . . . . . . . . . . . .325
G.2 Gauge Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .327
G.2.1 General .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .327
G.2.2 Covariant Gauges .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .328
G.2.3 Noncovariant Gauge .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .329
G.3 Gamma Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .332
G.3.1 z D 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .333
G.3.2 z D 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .333
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .334
H Feynman Diagrams and Feynman Amplitude. . . . . . . . . . . .. . . . . . . . . . . . . . . . .335

H.1 Feynman Diagrams .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .335
H.1.1 S-Matrix .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .335
H.1.2 Green’s Function.. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .336
H.1.3 Covariant Evolution Operator .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .336
H.2 Feynman Amplitude.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .337
I Evaluation Rules for Time-Ordered Diagrams. . . . . . . . . . .. . . . . . . . . . . . . . . . .341

I.1 Single-Photon Exchange .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .342
I.2 Two-Photon Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .343
I.2.1 No Virtual Pair .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .344
I.2.2 Single Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .345
I.2.3 Double Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .346
I.3 General Evaluation Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .347
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .348
J Some Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .349

J.1 Feynman Integrals .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .349
R d 3 k eikr12
J.2 Evaluation of the Integral .2/ 3 q 2 k 2 Ci . . . . . . . . .. . . . . . . . . . . . . . . . .351
R d 3k ikr
J.3 Evaluation of the Integral .2/3 ˛1 kO ˛2 kO q 2ek 212Ci . . . . . .352
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .354
xviii Contents
K Unit Systems and Dimensional Analysis . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355

K.1 Unit Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355
K.1.1 SI System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355
K.1.2 Relativistic or “Natural” Unit System . . . . . .. . . . . . . . . . . . . . . . .355
K.1.3 Hartree Atomic Unit System .. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .356
K.1.4 cgs Unit Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .357
K.2 Dimensional Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .357
Abbreviations .. . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .361
Index . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .363
Chapter 1
Introduction
1.1 Standard Many-Body Perturbation Theory
The quantum-mechanical treatment of many-electron systems, based on the

Schrödinger equation and the Coulomb interaction between the electrons, was
developed shortly after the advent of quantum mechanics, particularly by John
Slater in the late 1920s and early 1930s [58]. Self-consistent-field (SCF) schemes
were early developed by Slater, Hartree, Fock, and others.1 Perturbative schemes for
quantum-mechanical system, based on the Rayleigh–Schrödinger and Brillouin–
Wigner schemes, were developed in the 1930s and 1940s, leading to the important
linked-diagram expansion, introduced by Brueckner [13] and Goldstone [28] in
the 1950s, primarily for nuclear applications. That scheme was in the 1960s and
1970s also applied to electronic systems [31] and extended to degenerate and
quasi-degenerate energy levels [10, 35]. The next step in this development was the
introduction of “all-order methods” of coupled-cluster type, where certain effects
are taken to all orders of the perturbation expansion. This represents the last –
and probably final – major step of the development of a nonrelativistic many-body
perturbation theory (MBPT).2 The first step toward a relativistic treatment of many-
electron systems was taken in the early 1930s by Breit [11], extending works made
somewhat earlier by Gaunt [25]. Physically, the Gaunt interaction represents the
magnetic interaction between the electrons, which is a purely relativistic effect.
Breit augmented this treatment by including the leading retardation effect, due to
the fact that the Coulomb interaction is not instantaneous, which is an effect of the
same order.
A proper relativistic theory should be Lorentz covariant, like the Dirac single-
electron theory.3 The Dirac equation for the individual electrons together with the
1
For a review of the SCF methods, the reader is referred to the book by Froese-Fischer [24].
2
By MBPT, we understand here perturbative methods based upon the Rayleigh–Schrödinger per-
turbation scheme and the linked-diagram expansion. To that group, we also include nonperturbative
schemes, such as the coupled-cluster approach (CCA), which are based upon the same formalism.
3
A physical quantity (scalar, vector, tensor) is said to be Lorentz covariant, if it transforms accord-
ing to a representation of the Lorentz group. (Only a scalar is invariant under that transformation.)
I. Lindgren, Relativistic Many-Body Theory: A New Field-Theoretical 1

Approach, Springer Series on Atomic, Optical, and Plasma Physics 63,
DOI 10.1007/978-1-4419-8309-1 1, c Springer Science+Business Media, LLC 2011
2 1 Introduction
instantaneous Coulomb and Breit interactions between the electrons represents for a
many-electron system all effects up to order ˛ 2 H(artree atomic units) or ˛ 4 me c 2 .4
This procedure, however, is NOT Lorentz covariant, and the Breit interaction can
only be treated to first order in perturbation theory, unless projection operators are
introduced to prevent the intermediate states from falling into the “Dirac sea” of
negative-energy states, as discussed early by Brown and Ravenhall [12] and later by
Sucher [62]. The latter approach has been successfully employed for a long time in
relativistic many-body calculations and is known as the no-virtual-pair approxima-
tion (NVPA).
A fully covariant relativistic many-body theory requires a field-theoretical ap-
proach, i.e., the use of quantum-electrodynamics (QED). In principle, there is no
sharp distinction between relativity and QED, but conventionally we shall refer
to effects beyond the NVPA as QED effects. This includes effects of retardation,
virtual pairs, and radiative effects (self-energy, vacuum polarization, vertex correc-
tion). The systematic treatment of these effects requires a covariant approach, where
the QED effects are included in the wave function. The main purpose of this book
is to formulate the foundations of such a procedure.
1.2 Quantum-Electrodynamics
Already in the 1930s deviations were observed between the results of precision
spectroscopy and the Dirac theory for simple atomic systems, primarily the hydro-
gen atom. Originally, this deviation was expected to be due to vacuum polarization,
i.e., spontaneous creation of electron–positron pairs in the vacuum, but this effect
turned out to be too small and even of the wrong sign. An alternative explanation
was the electron self-energy, i.e., the emission and absorption of a virtual photon
on the same electron – another effect that is not included in the Dirac theory. Early
attempts to calculate this effect, however, were unsuccessful, due to singularities
(infinities) in the mathematical expressions.
The first experimental observation of a clear-cut deviation from the Dirac theory
was the detection in 1947 by Lamb and Retherford of the so-called Lamb shift [34],
namely the shift between the 2s and 2p1=2 levels in atomic hydrogen, levels that
are exactly degenerate in the Dirac theory [17, 18]. In the same year, Bethe was
able to explain the shift by a nonrelativistic calculation, eliminating the singularity
of the self-energy by means of a renormalization process [5]. At the same time,
Kusch and Foley observed that the magnetic g-factor of the free electron deviates
slightly but significantly from the Dirac value 2 [32, 33]. These observations led
An equation or a theory, like the theory of relativity or Maxwell’s theory of electromagnetism, is

said to be Lorentz covariant, if it can be expressed entirely in terms of covariant quantities (see, for
instance, the books of Bjorken and Drell [7, 8]).
4
˛ is the fine-structure constant 1=137 and me c 2 is the electron rest energy (see Appendix K).
1.3 Bethe–Salpeter Equation 3
to the development of the modern form of the quantum-electrodynamic theory by

Feynman, Schwinger, Dyson, Tomanaga, and others by which the deviations from
the Dirac theory could be explained with good accuracy [20, 22, 23, 56, 64].5
The original theory of QED was applied to free electrons. During the last four
decades, several methods have been developed for numerical calculation of QED
effects in bound electronic states. The scattering-matrix or S-matrix formulation,
originally developed for dealing with the scattering of free particles, was made
applicable also to bound states by Sucher [60], and the numerical procedure was
refined in the 1970s particularly by Mohr [38]. During the last two decades, the
method has been extensively used in studies of highly charged ions to test the QED
theory under extreme conditions, works that have been pioneered by Mohr and Soff
(for a review, see [39]).
The Green’s function is one of the most important tools in mathematical physics
with applications in essentially all branches of physics.6 In 1990s, the method
was adopted to bound-state QED problems by Shabaev et al. [57]. This procedure
is referred to as the Two-times Green’s function and has recently been extensively
applied to highly charged ions by the St Petersburg group.
During the first decade of this century, another procedure for numerical QED
calculations was developed by the Gothenburg atomic theory group, termed the
Covariant-evolution-operator (CEO) method [36], which has been applied to the
fine structure and other energy-level separations of helium-like ions.
1.3 Bethe–Salpeter Equation
The first completely covariant treatment of a bound-state problem was presented

in 1951 by Salpeter and Bethe [6, 52] and by Gell-Mann and Low [26]. The
Bethe–Salpeter (BS) equation contains in principle the complete relativistic and in-
terelectronic interaction, i.e., all kinds of electron correlation and QED effects.
The BS equation is associated with several fundamental problems, which were
discussed in the early days, particularly by Dyson [21], Goldstein [27], Wick [65],
and Cutkosky [16]. Dyson found that the question of relativistic quantum mechanics
is “full of obscurities and unsolved problems” and that “the physical meaning of
the four-dimensional wave function is quite unclear.” It seems that some of these
problems still remain.
The BS equation is based upon field theory, and there is no direct connection
to the Hamiltonian approach of relativistic quantum mechanics. The solution of
the field-theoretical BS equation leads to a four-dimensional wave function with
individual times for the two particles. This is not in accordance with the standard
5
For the history of the development of the QED theory, the reader is referred to the authoritative
review by Schweber [55].
6
For a comprehensive account of the applications, particularly in condensed-matter physics, the
reader is referred to the book by Mahan [37].
4 1 Introduction
quantum-mechanical picture, which has a single time variable also for many-particle
systems. The additional time variable leads sometimes to “abnormal solutions” with
no counterparts in nonrelativistic quantum mechanics, as discussed particularly by
Nakanishi [40] and Namyslowski [41].
Much efforts have been devoted to simplifying the BS equation by reducing it
to a three-dimensional equation, in analogy with the standard quantum-mechanical
equations (for reviews, see [9, 15]). Salpeter [51] derived early an “instantaneous”
approximation, neglecting retardation, which led to a relativistically exact three-
dimensional equation, similar to – but not exactly equal to – the Breit equation.
More sophisticated is the so-called quasi-potential approximation, introduced by
Todorov [63], frequently used in scattering problems. Here, a three-dimensional
Schrödinger-type equation is derived with an energy-dependent potential, deduced
from scattering theory. Sazdjian [53, 54] was able to separate the BS equation into
a three-dimensional equation of Schrödinger type and one equation for the relative
time of the two particles, serving as a perturbation – an approach that is claimed
to be exactly equivalent to the original BS equation. This approach establishes a
definitive link between the Hamiltonian relativistic quantum mechanics and field
theory. Connell [15] further developed the quasi-potential approximation of Todorov
by introducing series of corrections, a procedure that also is claimed to be formally
equivalent to the original BS equation.
Caswell and Lepage [14] applied the quasi-potential method to evaluate the hy-
perfine structure of muonium and positronium to the order ˛ 6 me c 2 by combining
analytical and perturbative approaches. Grotch and Yennie [9, 30] have applied the
method to evaluate higher-order nuclear corrections to the energy levels of the hy-
drogen atom, and Adkins and Fell [1, 2] have applied it to positronium.
The procedure we shall develop in the following is to combine the covariant-
evolution-operator method with electron correlation, which will constitute a step
toward a fully covariant treatment of many-electron systems. This will form another
approximation of the full Bethe–Salpeter equation that seems feasible for electronic
systems.
A vast literature on the Bethe–Salpeter equation, its fundamental problems and its
applications, has been gathered over the years since the original equation appeared.
Most applications are performed in the strong-coupling case (QCD), where the fun-
damental problems of the equation are more pronounced. The interested reader is
here referred to some reviews of the field, where numerous references to original
works can be found [29, 40–42, 54].
1.4 Helium Atom: Analytical Approach
An approach to solve the BS equation, known as the external-potential approach,

was first developed by Sucher [59, 61] to evaluate the lowest-order QED contribu-
tions to the ground-state energy of the helium atom, and equivalent results were at
the same time also derived by Araki [3]. The electrons are here assumed to move
1.5 Field-Theoretical Approach to Many-Body Perturbation Theory 5
in the field of the (infinitely heavy) atomic nucleus. The relative time of the two
electrons is eliminated by integrating over the corresponding energy of the Fourier
transform, which leads to a Schrödinger-like equation, as in the quasi-potential
method. The solution of this equation is expanded in terms of a Brillouin–Wigner
perturbation series. This work has been further developed and applied by Douglas
and Kroll [19] and by Zhang and Drake [69, 70] by considering higher-order terms
in the ˛ and Z˛ expansions. This approach, which is reviewed in Chap. 11, can be
used for light systems, such as light helium-like ions, where the power expansions
are sufficiently convergent. The QED effects are here evaluated by means of highly
correlated wave functions of Hylleraas type, which implies that QED and electron-
correlation effects are highly mixed. A related technique, referred to as the effective
Hamiltonian approach, has been developed and applied to helium-like systems by
Pachucki and Sapirstein [43–45].
A problem that has been controversial for quite some time is the fine structure of
the lowest P state of the neutral helium atom. The very accurate analytical results
of Drake et al. and by Pachucki et al. give results close to the experimental results
obtained by Gabrielse and others [68], but there have for quite some time been
significant deviations – well outside the estimated limits of error. Very recently,
Pachucki and Yerokhin have by means of improved calculations shown that the
controversy has been resolved [46, 47, 66, 67].
1.5 Field-Theoretical Approach to Many-Body

Perturbation Theory
The methods previously mentioned for numerical QED calculations can for compu-
tational reasons be applied only to one- and two-photon exchange, which implies
that the electron correlation is treated at most to second order. This might be suffi-
ciently accurate for highly charged systems, where the QED effects dominate over
the electron correlation, but is usually quite insufficient for lighter systems, where
the situation is reversed. To remedy the situation to some extent, higher-order many-
body contributions can be added to the two-photon energy, a technique applied by
the Gothenburg and St Petersburg groups [4, 48].
In the numerical procedures for standard (relativistic) MBPT, the electron corre-
lation can be evaluated effectively to essentially all orders by technique of coupled-
cluster type. QED effects can here be included only as first-order energy corrections,
a technique applied particularly by the Notre-Dame group [49]. To treat electron
correlation, relativity and QED in a unified manner would require a field-theoretical
approach.
The above-mentioned methods for QED calculations are all based upon field-
theory. Of these methods, the covariant-evolution method has the advantage that
it has a structure that is quite akin to that of standard MBPT, which has the con-
sequence that it can serve as a basis for a unified field-theoretical many-body
approach. The QED effects can here be included in the wave function, which will
6 1 Introduction
make it possible to treat the QED and correlation effects in a more unified way.
To solve this problem completely is a formidable task, but it will be a main theme
of this book to describe how some steps can be taken in this direction, along
the line that is presently being pursued by the Gothenburg atomic theory group.
The covariant evolution operator, which describes the time evolution of the rela-
tivistic state vector, is the key tool in this treatment. This operator is closely related
to the field-theoretical Green’s function. It should be mentioned that a related idea
was proposed by Leonard Rosenberg already 20 years ago [50], namely of including
Coulomb interactions in the QED Hamiltonian, and this is essentially the procedure
we are pursuing in this book.
The covariant evolution operator is singular, as is the standard evolution operator
of nonrelativistic quantum mechanics, but the singularities can be eliminated in a
similar way as the corresponding singularities of the Green’s function. The regu-
lar part of the covariant evolution operator is referred to as the Green’s operator,
which can be regarded as an extension of the Green’s-function concept and shown
to serve as a link between field theory and standard many-body perturbation the-
ory. The perturbation used in this procedure represents the interaction between the
electromagnetic field and the individual electrons. This implies that the equations
operate in an extended photonic Fock space with variable number of photons.
The strategy in dealing with the combined QED and correlation problem is first to
construct a field-theoretical “QED potential” with a single retarded photon, contain-
ing all first-order QED effects (retardation, virtual pairs, radiative effects), which –
after proper regularization and renormalization – can be included in a perturbative
expansion of MBPT or coupled-cluster type. In this way, the QED effects can – for
the first time – be built into the wave function and treated together with the elec-
tron correlation in a coherent manner. For practical reasons, only a single retarded
photon (together with arbitrary number of Coulomb interactions) can be included in
this procedure at present time, but due to the fact that these effects are included
in the wave function, this corresponds to higher-order effects in the energy. When
extended to interactions of multiphoton type, this leads for two-particle systems to
the Bethe–Salpeter equation, and in the multireference case to an extension of this
equation, referred to as the Bethe–Salpeter–Bloch equation.
In combining QED with electron correlation, it is necessary to work in the
Coulomb gauge, in order to take advantage of the development in standard MBPT.
Although this gauge is noncovariant in contrast to, for instance, the simpler Feyn-
man gauge, it can be argued that the deviation from a fully covariant treatment will
have negligible effect in practical applications when handled properly. This makes
it possible to mix a larger number of Coulomb interactions with the retarded-photon
interactions, which is expected to lead to the same ultimate result as a fully co-
variant approach but with faster convergence rate due to the dominating role of the
Coulomb interaction.
The procedure can also be extended to systems with more than two electrons,
and due to the complete compatibility between the standard and the extended pro-
cedures, the QED effects need only be included where they are expected to be most
significant.
References 7
In principle, also the procedure outlined here leads to individual times for the
particles involved, consistent with the full Bethe–Salpeter equation but not with
the standard quantum-mechanical picture. We shall mainly work in the equal-time
approximation here, and we shall not analyze effects beyond this approximation in
any detail. It is expected that – if existing – any such effect would be extremely
small for electronic systems.
References
1. Adkins, G.S., Fell, R.N.: Bound-state formalism for positronium. Phys. Rev. A 60, 4461–75
(1999)
2. Adkins, G.S., Fell, R.N., Mitrikov, P.M.: Calculation of the positronium hyperfine interval
using the Bethe-Salpeter formalism. Phys. Rev. A 65, 042,103 (2002)
3. Araki, H.: Quantum-Electrodynamical Corrections to Energy-levels of Helium. Prog. Theor.
Phys. (Japan) 17, 619–42 (1957)
4. Artemyev, A.N., Shabaev, V.M., Yerokhin, V.A., Plunien, G., Soff, G.: QED calculations of
the n D 1 and n D 2 energy levels in He-like ions. Phys. Rev. A 71, 062,104 (2005)
5. Bethe, H.A.: The Electromagnetic Shift of Energy Levels. Phys. Rev. 72, 339–41 (1947)
6. Bethe, H.A., Salpeter, E.E.: An Introduction to Relativistic Quantum Field Theory. Quantum
Mechanics of Two-Electron Atoms. Springer, Berlin (1957)
7. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Fields. Mc-Graw-Hill Pbl. Co, N.Y. (196)
8. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Mechanics. Mc-Graw-Hill Pbl. Co, N.Y.
(1964)
9. Boldwin, G.T., Yennie, D.R., Gregorio, M.A.: Recoil effects in the hyperfine structure of QED
bound states. Rev. Mod. Phys. 57, 723–82 (1985)
10. Brandow, B.H.: Linked-Cluster Expansions for the Nuclear Many-Body Problem. Rev. Mod.
Phys. 39, 771–828 (1967)
11. Breit, G.: Dirac’s equation and the spin-spin interaction of two electrons. Phys. Rev. 39,
616–24 (1932)
12. Brown, G.E., Ravenhall, D.G.: On the Interaction of Two electrons. Proc. R. Soc. London,
Ser. A 208, 552–59 (1951)
13. Brueckner, K.A.: Many-Body Problems for Strongly Interacting Particles. II. Linked Cluster
Expansion. Phys. Rev. 100, 36–45 (1955)
14. Caswell, W.E., Lepage, G.P.: Reduction of the Bethe-Salpeter equation to an equivalent
Schrödinger equation, with applications. Phys. Rev. A 18, 810–19 (1978)
15. Connell, J.H.: QED test of a Bethe-Salpeter solution method. Phys. Rev. D 43, 1393–1402
(1991)
16. Cutkosky, R.E.: Solutions of the Bethe-Salpeter equation. Phys. Rev. 96, 1135–41 (1954)
17. Dirac, P.A.M.: . Roy. Soc. (London) 117, 610 (1928)
18. Dirac, P.A.M.: The Principles of Quantum Mechanics. Oxford Univ. Press, Oxford (1930,
1933, 1947, 1958)
19. Douglas, M.H., Kroll, N.M.: Quantum Electrodynamical Corrections to the Fine Structure of
Helium. Ann. Phys. (N.Y.) 82, 89–155 (1974)
20. Dyson, F.J.: The radiation Theories of Tomonaga, Schwinger, and Feynman. Phys. Rev. 75,
486–502 (1949)
21. Dyson, F.J.: The Wave Function of a Relativistic System. Phys. Rev. 91, 1543–50 (1953)
22. Feynman, R.P.: Space-Time Approach to Quantum Electrodynamics. Phys. Rev. 76, 769–88
(1949)
23. Feynman, R.P.: The Theory of Positrons. Phys. Rev. 76, 749–59 (1949)
24. Froese-Fischer, C.: The Hartree-Fock method for atoms. John Wiley and Sons, New York,
London, Sidney, Toronto (1977)
8 1 Introduction
25. Gaunt, J.A.: The Triplets of Helium. Proc. R. Soc. London, Ser. A 122, 513–32 (1929)
26. Gell-Mann, M., Low, F.: Bound States in Quantum Field Theory. Phys. Rev. 84, 350–54
(1951)
27. Goldstein, J.S.: Properties of the Salpeter-Bethe Two-Nucleon Equation. Phys. Rev. 91,
1516–24 (1953)
28. Goldstone, J.: Derivation of the Brueckner many-body theory. Proc. R. Soc. London, Ser. A
239, 267–279 (1957)
29. Grotch, H., Owen, D.A.: Bound states in Quantum Electrodynamics: Theory and Applications.
Fundamentals of Physics 32, 1419–57 (2002)
30. Grotch, H., Yennie, D.R.: Effective Potential Model for Calculating Nuclear Corrections to
the Eenergy Levels of Hydrogen. Rev. Mod. Phys. 41, 350–74 (1969)
31. Kelly, H.P.: Application of many-body diagram techniques in atomic physics. Adv. Chem.
Phys. 14, 129–190 (1969)
32. Kusch, P., Foley, H.M.: Precision Measurement of the Ratio of the Atomic ’g Values’ in the
2
P3=2 and 2 P1=2 States of Gallium. Phys. Rev. 72, 1256–57 (1947)
33. Kusch, P., Foley, H.M.: On the Intrinsic Moment of the Electron. Phys. Rev. 73, 412 (1948)
34. Lamb, W.W., Retherford, R.C.: Fine structure of the hydrogen atom by microwave method.
Phys. Rev. 72, 241–43 (1947)
35. Lindgren, I.: The Rayleigh-Schrödinger perturbation and the linked-diagram theorem for a
multi-configurational model space. J. Phys. B 7, 2441–70 (1974)
37. Mahan, G.D.: Many-particle Physics, second edition. Springer Verlag, Heidelberg (1990)
38. Mohr, P.J.: Numerical Evaluation of the 1s1=2 -State Radiative Level Shift. Ann. Phys. (N.Y.)
88, 52–87 (1974)
227–372 (1998)
40. Nakanishi, N.: Normalization condition and normal and abnormal solutions of Bethe-Salpeter
equation. Phys. Rev. 138, B1182 (1965)
41. Namyslowski, J.M.: The Relativistic Bound State Wave Function. in Light-Front Quantization
and Non-Perturbative QCD, J.P. Vary and F. Wolz, eds. (International Institute of Theoretical
and Applied Physics, Ames) (1997)
42. Onida, G., Reining, L., Rubio, A.: Electronic excitations: density-functional versus many-body
Green’s-function approaches. Rev. Mod. Phys. 74, 601–59 (2002)
43. Pachucki, K.: Quantum electrodynamics effects on helium fine structure. J. Phys. B 32, 137–52
(1999)
44. Pachucki, K.: Improved Theory of Helium Fine Structure. Phys. Rev. Lett. 97, 013,002 (2006)
45. Pachucki, K., Sapirstein, J.: Contributions to helium fine structure of order m˛ 7 . J. Phys. B
33, 5297–5305 (2000)
46. Pachucky, K., Yerokhin, V.A.: Reexamination of the helium fine structure (vol 79, 062516,
2009). Phys. Rev. Lett. 80, 19,902 (2009)
47. Pachucky, K., Yerokhin, V.A.: Reexamination of the helium fine structure (vol 79, 062516,
2009). Phys. Rev. A 81, 39,903 (2010)
48. Persson, H., Salomonson, S., Sunnergren, P., Lindgren, I.: Two-electron Lamb-Shift Calcula-
tions on Heliumlike Ions. Phys. Rev. Lett. 76, 204–07 (1996)
49. Plante, D.R., Johnson, W.R., Sapirstein, J.: Relativistic all-order many-body calculations of
the n D 1 and n D 2 states of heliumlike ions. Phys. Rev. A 49, 3519–30 (1994)
50. Rosenberg, L.: Virtual-pair effects in atomic structure theory. Phys. Rev. A 39, 4377–86
(1989)
51. Salpeter, E.E.: Mass Correction to the Fine Structure of Hydrogen-like Atoms. Phys. Rev. 87,
328–43 (1952)
52. Salpeter, E.E., Bethe, H.A.: A Relativistic Equation for Bound-State Problems. Phys. Rev. 84,
1232–42 (1951)
53. Sazdjian, H.: Relativistiv wave equations for the dynamics of two interacting particles. Phys.
Rev. D 33, 3401–24 (1987)
References 9
54. Sazdjian, H.: The connection of two-particle relativistic quantum mechanics with the Bethe-
Salpeter equation. J. Math. Phys. 28, 2618–38 (1987)
55. Schweber, S.S.: The men who madt it: Dyson, Feynman, Schwinger and Tomonaga. Princeton
University Press, Princeton (1994)
56. Schwinger, J.: Quantum electrodynamics I. A covariant formulation. Phys. Rev. 74, 1439
(1948)
58. Slater, J.: Quantum Theory of Atomic Spectra. McGraw-Hill, N.Y. (1960)
59. Sucher, J.: Energy Levels of the Two-Electron Atom to Order ˛ 3 Ry; Ionization Energy of
Helium. Phys. Rev. 109, 1010–11 (1957)
60. Sucher, J.: S-Matrix Formalism for Level-Shift Calculations. Phys. Rev. 107, 1448–54 (1957)
61. Sucher, J.: Ph.D. thesis, Columbia University (1958). Univ. Microfilm Internat., Ann Arbor,
Michigan
62. Sucher, J.: Foundations of the Relativistic Theory of Many Electron Atoms. Phys. Rev. A 22,
348–62 (1980)
63. Todorov, I.T.: Quasipotential Equation Corresponding to the relativistic Eikonal Approxima-
tion. Phys. Rev. D 3, 2351–56 (1971)
64. Tomanaga, S.: On Infinite Field Reactions in Quantum Field Theory. Phys. Rev. 74, 224–25
(1948)
65. Wick, G.C.: Properties of Bethe-Salpeter Wave Functions. Phys. Rev. 96, 1124–34 (1954)
66. Yerokhin, K.P.V.A.: Reexamination of the helium fine structure. Phys. Rev. Lett. 79, 62,616
(2009)
67. Yerokhin, K.P.V.A.: Fine Structure of Heliumlike Ions and Determination of the Fine Structure
Constant. Phys. Rev. Lett. 104, 70,403 (2010)
68. Zelevinsky, T., Farkas, D., Gabrielse, G.: Precision Measurement of the Three 23 PJ Helium
Fine Structure Intervals. Phys. Rev. Lett. 95, 203,001 (2005)
69. Zhang, T.: Corrections to O.˛ 7 .ln ˛/mc 2 / fine-structure splittings and O.˛ 6 .ln ˛/mc 2 / en-
ergy levels in helium. Phys. Rev. A 54, 1252–1312 (1996)
70. Zhang, T., Drake, G.W.F.: Corrections to O.˛ 7 mc 2 / fine-structure splitting in helium. Phys.
Rev. A 54, 4882–4922 (1996)
Part I
Basics: Standard Many-Body
Perturbation Theory
Chapter 2
Time-Independent Formalism
In this first part of the book, we shall review some basics of quantum mechanics and
the many-body theory for bound electronic systems that will form the foundations
for the following treatment. This material can also be found in several standard text
books. The time-independent formalism is summarized in the present chapter1 and
the time-dependent formalism in the following one.
2.1 First Quantization
First quantization is the term for the elementary treatment of quantized systems,
where the particles of the system are treated quantum-mechanically, for instance,
in terms of Schrödinger wave functions, while the surrounding fields are treated
classically.
2.1.1 De Broglie’s Relations
As an introduction to the quantum mechanics, we shall derive the Schrödinger equa-

tion from the classical relations of de Broglie.
According to Planck–Einstein’s quantum theory, the electromagnetic radiation is
associated with particle-like photons with the energy (E) and momentum (p) given
by the relations

E D h D !„
; (2.1)
p D h= D „k
1
This chapter is essentially a short summary of the second part of the book Atomic Many-Body
Theory by Lindgren and Morrison, and the reader who is not well familiar with the subject is
recommended to consult that book.

14 2 Time-Independent Formalism
where „ D h=2, h being Planck’s constant (see further Appendix K), the cyclic
frequency of the radiation (cycles/s) and ! D 2 the angular frequency (radians/s).
D c= (c being the velocity of light in vacuum) is the wavelength of the radiation
and k D 2= the wave number.
De Broglie assumed that the relations (2.1) for photons would hold also for ma-
terial particles, like electrons. Nonrelativistically, we have for a free electron in one
dimension
p2 „2 k 2
ED or „! D ; (2.2)
2me 2me
where me is the mass of the electron.
De Broglie assumed that a particle could be represented by a wave packet
Z
.t; x/ D dk a.k/ ei.kx!t / : (2.3)
The relation (2.2) then leads to the one-dimensional wave equation for a free
electron
@.t; x/ „2 @2 .t; x/
i„ D ; (2.4)
@t 2me @x 2
which is the Schrödinger equation for a free particle. This can be obtained from the
first of the relations (2.2) by means of the substitutions
@ @
E ! i„ p ! i„ : (2.5)
@t @x
2.1.2 The Schrödinger Equation
We can generalize the treatment above to an electron in three dimensions in an

external field, vext .x/, for which the energy Hamiltonian is
p2
EDH D C vext .x/: (2.6)
2me
Generalizing the substitutions above to2
p ! pO D i„r and x ! xO D x; (2.7)
where r is the vector gradient operator (see Appendix A.1) leads to the Hamilton
operator
pO 2 „2 2
HO D C vext .x/ D r C vext .x/ (2.8)
2me 2me
2
Initially, we shall use the ‘hat’ symbol to indicate an operator, but later we shall use this symbol
only when the operator character needs to be emphasized.
2.1 First Quantization 15
and to the Schrödinger equation for a single electron

@ O „2 2
i„ .t; x/ D H .t; x/ D r C vext .x/ .t; x/: (2.9)
@t 2me
For an N -electron system, the Schrödinger equation becomes correspondingly3
@
i„ .tI x 1 ; x 2 ; x N / D HO .tI x 1 ; x 2 ; x N /; (2.10)
@t
where we assume the Hamiltonian to be of the form HO D HO 1 C HO 2 [see Appendix

(B.19)]4
XN XN
„2 2
HO 1 D rn C vext .x n / DW hO 1 .n/;
nD1
2me nD1
X
N
e2 XN
HO 2 D DW hO 2 .m; n/: (2.11)
m<n
40 rmn m<n
Here, rmn is the interelectronic distance, rmn D jx m x n j and vext represents the
external (essentially nuclear) energy potential.
Generally, the quantum-mechanical operators A; O BO that represent the corre-
sponding classical quantities A; B in the Hamilton formulation (see Appendix E)
should satisfy the quantization condition
O B
ŒA; O D AOBO BO AO D i„fA; Bg; (2.12)
where the square bracket (with a comma) represents the commutator and the curly
bracket the Poisson bracket (E.10). For conjugate momenta, like the coordinate
vector x and the momentum vector p, the Poisson bracket equals unity, and the
quantization conditions for the corresponding operators become
O pOx D Œy;
Œx; O pOy D ŒOz; pOz D i„; (2.13)
which is consistent with the substitutions (2.7).

We shall be mainly concerned with stationary, bound states of electronic sys-
tems, for which the wave function can be separated into a time function and a space
function
.tI x 1 ; x N / D F .t/ .x 1 ; x 2 ; x N /:
3
Note that according to the quantum-mechanical picture the wave function has a single time also
for a many-electron system. This question will be discussed further below.
4
The symbol “D:” indicates that this is a definition.
As shown in standard text books, this leads to a separation into two equations, one
for the time part and one for the space part. The time equation becomes
@
i„ F .t/ D E F .t/;
@t
with the solution
F .t/ / eiE t =„ ;
and the space part is the standard time-independent Schrödinger equation
HO .x 1 ; x N / D E .x 1 ; x N /: (2.14)
Thus, for stationary states the time-dependent wave function is of the form
.tI x 1 ; x N / D eiE t =„ .x 1 ; x N /: (2.15)
The separation constant E is interpreted as the energy of the state.
2.2 Second Quantization
2.2.1 Schrödinger Equation in Second Quantization*
In the following, we shall consistently base our treatment upon second quantiza-
tion, which implies that also the particles and fields are quantized and expressed in
terms of (creation and absorption) field operators (see Appendices B and C). Here,
we shall first derive the second-quantized form of the time-dependent Schrödinger
equation (SE) (2.9), which reads
@
i„ j.t/i D H j.t/i: (2.16)
@t
With the partitioning (2.11), the operator becomes in second quantization (B.12)
1
HO D ci hi jh1 jj i cj C ci cj hij jh2 jkl i cl ck

(2.17)
2
and the state is expressed as a vector (C.4). The equation (2.16) is by no means
obvious, and we shall here indicate the proof. (The proof follows largely that given
by Fetter and Walecka [19, Chap. 1].)
For the sake of concretization, we consider a two-electron system. With the co-
ordinate representation (C.19) of the state vector
.x1 ; x2 / D hx1 ; x 2j.t/i ; (2.18)

2.2 Second Quantization 17
the SE (2.16) becomes
@
i„ hx 1 ; x 2 j.t/i D hx; x 2 jH j.t/i: (2.19)
@t
We consider first the effect of the one-body part of the Hamiltonian (2.17) op-
erating on the wave function (2.18), and we shall show that this is equivalent to
operating with the second-quantized form of the operator (B.19)
HO D ci hi jh1jj i cj

(2.20)
on the state vectorj.t/i.

We start by expanding the state vector in terms of straight products of single-
electron state vectors (t1 D t2 D t)
j.t/i D akl .t/ jkijli ; (2.21)
(akl D alk ). The coordinate representation of this relation is
.x1 ; x2 / D hx1 ; x 2j.t/i D akl .t/ hx 1jki hx 2jli : (2.22)
We now operate with the single-particle operator (2.20) on the state vector ex-
pansion (2.21)
HO 1j.t/i D ci hi jh1jj i cj akl .t/ jkijli : (2.23)
For j D k, the electron in position 1 is annihilated in the state k and replaced by an
electron in the state i , yielding
hi jh1jki akl .t/ ji ijli :
The coordinate representation of this relation becomes
hx1ji i hi jh1jki akl .t/ hx 2jli D hx 1 jh1jki akl .t/ hx 2jli
using the resolution of the identity (C.12). The right-hand side of (2.23) can also be
expressed
h1 .x 1 / k .x 1 / l .x 2 / akl .t/ D h1 .x 1 /.x1 ; x2 /:
Together with the case j D l this leads to
hx 1 ; x 2 jH1j.t/i D .h1 .x 1 / C h1 .x 2 // .x1 ; x2 / D H1 .x1 ; x2 /:
Thus, we have shown the important relation
hx 1 ; x 2 jH1j.t/i D H1 .x1 ; x2 /: (2.24)

A similar relation can be derived for the two-body part of the Hamiltonian, which
implies that
hx 1 ; x 2 jHj.t/i D H.x1 ; x2 / (2.25)
and from the relation (2.19)
@
i„ hx 1 ; x 2j.t/i D hx 1 ; x 2 jHj.t/i : (2.26)
@t
This is the coordinate representation of the Schrödinger equation (2.16), which is
thus verified. It should be observed that (2.16) does not contain any space coordi-
nates. The treatment is here performed for the two-electron case, but it can easily be
extended to the general case.
2.2.2 Particle–Hole Formalism: Normal Order and Contraction
In the particle–hole formalism, we separate the single-particle states into particle

and hole states, a division that is to some extent arbitrary. Normally, core states
(closed-shell states) are treated as hole states and virtual and valence states as parti-
cle states, but sometimes it might be advantageous to treat some closed-shell states
as valence states or some valence states as hole states.
If time increases from right to left, the creation/annihilation operators are said
to be time ordered. Time ordering can be achieved by using the Wick time-ordering
operator, which for fermions reads

A.t1 /B.t2 / .t1 > t2 /
T ŒA.t1 /B.t2 / D : (2.27)
B.t2 /A.t1 / .t1 < t2 /
The case t1 D t2 will be discussed later.

The creation/annihilation operators are said to be in normal order, if the
particle-creation and hole-annihilation operators appear to the left of the particle-
annihilation and hole-creation operators

cp cp ch ch ; (2.28)
where p,h stand for particle/hole states.

A contraction of two operators is defined as the difference between the time-
ordered and the normal-ordered products,
x y D T Œxy N Œxy: (2.29)

2.2 Second Quantization 19
In the following, we shall use curly brackets to denote the normal product [38]
N Œxy fx yg: (2.30)
From these definitions, it follows that the nonvanishing contractions of the electron-
field operators (B.28) are
(
i"p .t1 t2 /=„
O C .x1 / O C OC p .x 1 / p .x 2 / e t1 > t2

.x2 / D
.x2 / O C .x1 / D ;
0 t1 < t2

O .x1 / O .x2 / D O .x2 / O .x1 / D 0 t1 > t2
i"h .t1 t2 /=„ :
h .x 1 / h .x 2 / e t1 < t2
(2.31)
Here, O ˙ represents the positive-/negative-energy part of the spectrum, respectively,
and p and h denote particle (positive-energy) and hole (negative-energy) states,
respectively.
The results can be summarized as
O .x1 / O .x2 / D O .x2 / O .x1 / D

j .x 1 / j .x 2 / e
i"j .t1 t2 /=„
; (2.32)
if t1 > t2 for particles and t1 < t2 for holes with all other contractions vanishing.
2.2.3 Wick’s Theorem
The handling of operators in second quantization is greatly simplified by Wick’s

theorem [80] (for an introduction, see, for instance, Fetter and Walecka [19, Sect. 8]
or Lindgren and Morrison [40, Chap. 11]), which states that a product of creation
and annihilation operators AO can be written as the normal product plus all sin-
gle, double . . . contractions with the uncontracted operators in normal form, or
symbolically
AO D fAg
O C fAg:
O (2.33)
A particularly useful form of Wick’s theorem is the following. If AO and BO are op-
erators in normal form, then the product is equal to the normal product plus all
normal-ordered contractions between AO and B,
O or formally
AO BO D fAO Bg
O C fAO Bg:
O (2.34)
With this formulation, there are no further contractions within the operators to be
multiplied. This forms the basic rule for the graphical representation of the operators
and operator relations to be discussed below.
2.3 Time-Independent Many-Body Perturbation Theory
2.3.1 Bloch Equation
Here, we shall summarize the most important concepts of standard time-

independent many-body perturbation theory (MBPT) as a background for the
further treatment. (For more details, the reader is referred to designated books, such
as Lindgren–Morrison, Atomic Many-Body Theory [40].)
We are considering a number of stationary electronic states,j ˛ i .˛ D 1 d /,
termed target states, that satisfy the Schrödinger equation
˛
Hj i D E ˛j ˛
i .˛ D 1 d /: (2.35)
˛
For each target state, there exists an “approximate” or model state, j 0 .˛ D ˛
1 d /, which is more easily accessible and which forms the starting point for the
perturbative treatment. We assume that the model states are linearly independent
and that they span a model space. The projection operator for the model space is
denoted P and that for the complementary or orthogonal space by Q, which to-
gether form the identity operator
P C Q D I: (2.36)
A wave operator is introduced – also known as the Møller operator [53] – which
transforms the model states back to the exact states,
˛ ˛
j i D ˝j 0i .˛ D 1 d / (2.37)
and this operator is the same for all states under consideration.
We define an effective Hamiltonian with the property that operating on a model
function it generates the corresponding exact energy
Heffj ˛
0i D E ˛j ˛
0i .˛ D 1 d /; (2.38)
with the eigenvectors representing the model states. Operating on this equation with
˝ from the left, using the definition (2.37), yields
˝Heffj ˛
0i D E ˛j ˛
i; (2.39)
which we compare with the Schrödinger equation (2.35)
H ˝j ˛
0i D E ˛j ˛
i: (2.40)
Since this relation holds for each state of the model space, we have the important
operator relation
˝Heff P D H ˝P; (2.41)
which as known as the generalized Bloch equation.
2.3 Time-Independent Many-Body Perturbation Theory 21
The form above of the Bloch equation is valid independently on the choice
of normalization. In the following, we shall mainly work with the intermediate
normalization (IN), which implies
h 0j
˛ ˛
i D 1; (2.42a)
j ˛
0i D Pj ˛
i .˛ D 1 d /: (2.42b)
Then we have after projecting the Schrödinger equation onto the model space
PH ˝j ˛
0i D E ˛j ˛
0i (2.43)
and we find that the effective Hamiltonian (2.38) becomes in IN
Heff D PH ˝P: (2.44)
Normally, the multidimensional or multireference model space is applied in con-

nection with valence universality, implying that the same operators are used for
different stages of ionization (see further Sect. 2.5).
2.3.2 Partitioning of the Hamiltonian
For electrons moving in an external (nuclear) potential, vext , the single-electron

(Schrödinger) Hamiltonian (2.8) is
„2 2
hS D r C vext : (2.45)
2me
The corresponding Schrödinger equation
hS i .x/ D "i i .x/ (2.46)
generates a complete spectrum of functions, which can form the basis for numerical
calculations. This is known to as the Furry picture. These single-electron functions
are normally referred to as (single-electron) orbitals – or spin-orbitals, if a spin
eigenfunction is adhered. Degenerate orbitals (with the same eigenvalue) form an
electron shell.
The Hamiltonian for a many-electron system (2.11) is
N
X XN
„2 2 e2
H D r C vext C ; (2.47)
n
2me n n<m
40 rnm
where the last term represents the interelectronic interaction. For the perturbation
treatment, we separate the many-electron Hamiltonian into
H D H0 C V; (2.48)
where H0 a model Hamiltonian that is a sum of single-electron Hamiltonians
N
X X
N
„2 2
H0 D r C vext C u DW h0 .n/ (2.49)
n
2me n n
and V is a perturbation
X
N X
N
e2
V D un C : (2.50)
n n<m
40 rnm
The potential u is optional and used primarily to improve the convergence properties
of the perturbation expansion.
The antisymmetrized N -electron eigenfunctions of H0 can be expressed as de-
terminantal products of single-electron orbitals (see Appendix B)
H0 ˚A .x 1 ; x 2 x N / D E0A ˚A .x 1 ; x 2 x N /;
p
˚A .x 1 ; x 2 x N / D 1= N Š Af 1 .x 1 / 2 .x 2 / N .x N /g; (2.51)
where A is an antisymmetrizing operator. The determinants are referred to as Slater

determinants and constitute our basis functions. The eigenvalues are given by
X
N
E0 D "n (2.52)
nD1
summed over the spin-orbitals of the determinant.

Degenerate determinants form a configuration. The model space is supposed to
be formed by one or several configurations that can have different energies. We
distinguish between three kinds of orbitals
Core orbitals, present in all determinants of the model space
Valence orbitals, present in some determinants of the model space
Virtual orbitals, not present in any determinants of the model space
The model space is said to be complete, if it contains all configurations that can
be formed by distributing the valence electrons among the valence orbitals in all
possible ways. In the following, we shall normally assume this to be the case.
With the partitioning (2.48), the Bloch equation above can be expressed
.˝Heff H0 ˝/ P D V ˝P: (2.53)

2.3 Time-Independent Many-Body Perturbation Theory 23
With H0 of the form (2.49), it commutes with the projection operator P . Then, we
find that
Heff D PH0 P C P V ˝P (2.54)
and we shall refer to the second term as the effective interaction
Veff D P V ˝P: (2.55)
The partitioning leads to the commonly used form of the generalized Bloch
equation [36, 37, 40]
Œ˝; H0 P D Q .V ˝ ˝Veff / P; (2.56)
which is frequently used as the basis for many-body perturbation theory (MBPT).
The last term appears only for open-shell systems with unfilled valence shell(s)
and is graphically represented by so-called folded or backwards diagrams, first
introduced by Brandow in nuclear physics [7], (see further below).
If the model space is completely degenerate with a single energy E0 , the general
Bloch equation reduces to its original form, derived in the late 1950s by Claude
Bloch [4, 5],
.E0 H0 / ˝P D V ˝P ˝Veff : (2.57)
This equation can be used to generate the standard Rayleigh–Schrödinger perturba-

tion expansion, found in many text books.
The generalized Bloch equation (2.56) is valid for a general model space, which
can contain different zeroth-order energy levels. Using such an extended model
space represents usually a convenient way of treating very closely spaced or quasi-
degenerate unperturbed energy levels, a phenomenon that otherwise can lead to
serious convergence problems. This can be illustrated by the relativistic calculation
of the fine structure of helium-like ions, where a one-dimensional model space leads
to convergence problems for light elements, a problem that can normally be reme-
died in a straightforward way by means of the extended model space [50, 62]. But
the extended model space can also lead to problems, due to so-called intruder states,
as will be further discussed below.
With an extended model space, we can separate the projection operator into the
corresponding energy components5
X
P D PE I H0 PE D EPE : (2.58)
E
5
In the case of an extended model space, we shall normally use the symbol E for the different
energies of the model space.
Operating with the general Bloch equation (2.56) on a particular component, then
yields
.E H0 / ˝PE D Q .V ˝ ˝Veff / PE : (2.59)
Expanding the wave operator order-by-order
˝ D 1 C ˝ .1/ C ˝ .2/ C (2.60)
leads to the recursive formula

.E H0 / ˝ .n/ PE D Q V ˝ .n1/ .˝Veff /.n/ PE (2.61)
or
˝ .n/ PE D Q .E/ V ˝ .n1/ .˝Veff /.n/ PE ; (2.62)
where
.k/
Veff D P V ˝ .k1/ P: (2.63)
Here,
1
.E/ D (2.64)
E H0
and
Q .E/ D Q .E/ (2.65)
are known as the resolvent and the reduced resolvent, respectively [45].
The recursive formula (2.62) can generate a generalized form of the Rayleigh–
Schrödinger perturbation expansion (see [40, Chap. 9]), valid also for a quasi-
degenerate model space. We see from the form of the resolvent that in each new
order of the perturbation expansion there is a denominator equal to the energy dif-
ference between the initial and final states. This leads to the Goldstone rules in the
evaluation of the time-ordered diagrams to be considered in the following section.
Even if the perturbation is energy independent, we see that the wave operator
and effective interaction will still generally be energy dependent, due to the energy
dependence of the resolvent. In first order, we have
˝ .1/ PE D Q .E/VPE (2.66)
and in second order

˝ .2/ .E/PE D Q .E/ V ˝ .1/ .E/ ˝ .1/ .E 0 /PE 0 Veff
.1/
PE ; (2.67)
.1/
where Veff D P VP . Note that the wave operator in the last term operates on the
projection operator PE 0 and therefore depends on the corresponding energy E 0 . We
now have
0
ı˝ .1/ .E/ ı Q .E/ Q .E / Q .E/ 0
D V D V D Q .E/ Q .E /V
ıE ıE E0 E
D Q .E/˝ .1/ .E 0 / (2.68)
2.4 Graphical Representation 25
and we note that the last folded term in (2.67) has a double denominator. We can
express the second-order Bloch equation as
ı˝ .1/ .E/ .1/

˝ .2/ .E/PE D Q .E/V ˝
.1/
.E/PE C Veff .E/PE : (2.69)
ıE
In the limit of complete degeneracy space, the difference ratio, of course, goes
over into the partial derivative. We shall show in later chapters that the second-order
expression above holds also when the perturbation is energy dependent (6.77).
2.4 Graphical Representation
In this section, we shall briefly describe a way of representing the perturbation

expansion graphically. (For further details, the reader is referred to the book by
Lindgren and Morrison [40].)
2.4.1 Goldstone Diagrams
The Rayleigh–Schrödinger perturbation expansion can be conveniently represented

in terms of diagrams by means of second quantization (see above and Appendix B).
The perturbation (2.50) becomes in second quantization
1
VO D ci cj hi jf jj i C ci cj cl ck hij jgjkli ; (2.70)
2
where f is the negative potential f D u and g is the Coulomb interaction between

the electrons. When some of the states above are hole states, the expression (2.70) is
not in normal order. By normal ordering the expression, zero-, one-, and two-body
operators will appear [40, Eq. 11.39]
V D V0 C V 1 C V 2 ; (2.71)
where
X
hole
1X
hole
V0 D hi jf ji i C Œhij jgjkli hj i jgjkli ;
2
i ij

V1 D fci cj g hi jveffjj i ;
1
V2 D fc c cl ck g hij jgjkli : (2.72)
2 i j
In the one- and two-body parts, the summation is performed over all orbitals.
Here,
X
hole
hi jveffjj i D hi jf jj i C Œhi kjgjj ki hki jgjj ki (2.73)
k
is known as the effective potential interaction and can be represented graphically as

shown in Fig. 2.3. The summation term represents the Hartree–Fock potential
X
hole
hi jvHFjj i D Œhi kjgjj ki hki jgjj ki ; (2.74)
k
where the first term is a “direct” integral and the second term an “exchange” in-
tegral. In the Hartree–Fock model, we have u D vHF , and the effective potential
vanishes [40].
We can now represent the perturbation (2.72) by the normal-ordered diagrams
in Fig. 2.1. The zero- and one-body parts are shown in more detail in Figs. 2.2 and
2.3. In our diagrams, the dotted line with the cross represents the potential interac-
tion, f D u, and the dotted line between the electrons the Coulomb interaction,
g D e 2 =40 r12 . We use here a simplified version of Goldstone diagrams. Each
free vertical line at the top (bottom) represents an electron creation (absorption) op-
erator but normally we do not distinguish between the different kinds of orbitals
(core, valence, and virtual) as done traditionally. There is a summation of internal
lines over all orbitals of the same category. We use here heavy lines to indicate
j i j
-u
V D V0 C +
i k l
Fig. 2.1 Graphical representation the effective-potential interaction (2.72). The heavy lines rep-
resent the orbitals in the Furry picture. The dotted line with the cross represents the potential u
and the dotted, horizontal lines the Coulomb interaction. The zero-body and one-body parts of the
interaction are depicted in Figs. 2.2 and 2.3, respectively
i k l k

u
V0 D 6 + 6 ? + -
l
Fig. 2.2 Graphical representation of the zero-body part of the effective-potential interaction
(2.72). The orbitals are summed over all core/hole states
i i i i
u
= + ?+
j
j j j
Fig. 2.3 Graphical representation of the effective-potential interaction (2.73). For the closed or-
bital lines (with no free end), there is a summation over the core/hole states. The last two diagrams
represent the “Hartree–Fock” potential, and the entire effective-potential interaction vanishes when
HF orbitals are used
that the orbitals are generated in an external (nuclear) potential, i.e., the bound-state
representation or Furry picture.
By means of Wick’s theorem, we can now normal order the right-hand side
(r.h.s.) of the perturbation expansion of the Bloch equation (2.62), and
Each resulting normal-ordered term will be represented by a diagram.
The first-order wave operator (2.66)
˝ .1/ PE D Q .E/ VPE D Q .E/ .V1 C V2 /PE (2.75)
becomes in second quantization (2.72)

n o hi jv jj i
eff 1n o hij jgjkli
˝ PE D Q
.1/
ci cj C c c cl ck PE :
"j "i 2 i j "k C "l "i "j
(2.76)
This can be represented in the same way as the open part (V1 C V2 ) of the pertur-
bation (2.70) (Fig. 2.1), if we include the extra energy denominator according to the
Goldstone rules, summarized below.
In second order, we have from (2.67), using Wick’s theorem (2.34),
n o n o n o n o
˝ .2/ PE D Q .E/V ˝E.1/ C V ˝E.1/ ˝E.1/ PE 0 Veff .1/
˝E.1/
0 P 0 V
E eff
.1/
PE ;
(2.77)
where the hook represents a contraction. The first uncontracted term is represented
by combinations of the diagrams in Fig. 2.1, such as
(2.78)
considered as a single diagram. This diagram can be of two types.


? ? ? ?

Fig. 2.4 Examples of second-order wave-operator diagrams, excluding folded diagrams
If, on the one hand, both disconnected parts are open, the diagram is referred
to as linked.6 If, on the other hand, at least one of them is closed, the diagram
is referred to as unlinked.
.1/
In the unlinked part of the second term in (2.77), the closed part represents Veff , and
since the order of the operators in the normal product is immaterial, this unlinked di-
agram appears also in the third term and is therefore eliminated. The last contracted
term survives and represents the “folded” term. Here, the wave operator depends on
the energy (E 0 ) of the intermediate state, which might differ from the energy of the
initial state (E). We can then express the second-order wave operator by

˝ .2/ PE D Q .E/ V ˝E.1/ ˝E.1/ .1/
0 PE 0 Veff PE ; (2.79)
linked
where only linked diagrams are maintained (see Fig. 2.4).

The diagrams in Fig. 2.4 are second-order time-ordered Goldstone diagrams. In
these diagrams, time is supposed to run from the bottom (although the formalism
is here time independent). The diagrams are evaluated by the standard Goldstone
rules with a denominator after each interaction equal to the energy difference be-
tween the (model-space) state at the bottom and that directly after the interaction
(see Appendix I and [40, Sect. 12.4]). (In later chapters, we shall mainly use Feyn-
man diagrams, which contain all possible time orderings between the interactions.)
2.4.2 Linked-Diagram Expansion
2.4.2.1 Complete Model Space
Written more explicitly, the second-order wave operator (2.79) becomes

0
˝ .2/ PE D Q .E/ V Q .E/ V Q .E/ Q .E / VPE 0 V linked
PE : (2.80)
6
A closed diagram has the initial as well as the final state in the model space. Such a diagram can
– in the case of complete model space – have no other free lines than valence lines. A diagram that
is not closed is said to be open.
Here, the second term has a double resolvent (double denominator, which might
contain different model-space energies), and it is traditionally drawn in a “folded”
way, as shown in the left diagram below (see, for instance, [40, Sect. 13.3])
r s r 6 6s
c d
) c 6P 0 6d
E
a b
PE 0 PE a 6 6b
PE (2.81)
The reason for drawing the diagram folded in this way is that the two pieces –
before and after the fold – should be evaluated with their denominators indepen-
dently. In the general case, by considering all possible time-orderings between the
two pieces, together with the Goldstone evaluation rules, it can be shown that the
denominators do factorize. In a relativistic treatment, which we shall employ for the
rest of this book, the treatment is most conveniently based upon Feynman diagrams,
which automatically contain all possible time-orderings, and then it is more natural
to draw the diagram straight, as shown in the second diagram above. Factorization
then follows directly. The double bar indicates that the diagram is “folded.” In such
a diagram, the upper part has double denominators – one denominator with the en-
ergy of the initial state and one with that of the intermediate model-space state. The
second-order wave operator can then be illustrated as shown in Fig. 2.5. Note that
there is a minus sign associated with the folded diagram.
The general ladder diagram (Fig. 2.5) may contain a (quasi)singularity, when the
intermediate state lies in the model space and is (quasi)degenerate with the initial
state. This singularity is automatically eliminated in the Bloch equation and leads to
the folded term. Later, in Sect. 6.6 we shall discuss this kind of singularity in more
detail in connection with energy-dependent interactions, and then we shall refer to
the finite remainder as the model-space contribution (MSC).
We have seen that the so-called unlinked diagrams are eliminated in the second-
order wave operator (2.79). When the model space is “complete” (see definition
above), it can be shown that unlinked diagrams disappear in all orders of pertur-
bation theory. This is the linked-cluster or linked-diagram theorem (LDE), first
r 6 6s r 6 6s r 6 6s
c 6
P CQ 6d ) c 6Q 6d + c 6
P 6d
a 6 6b a 6 6b a 6 6b
P P PE
Fig. 2.5 Removing the singularity from a ladder diagram leads to finite remainder, represented by
a “folded” diagram (last). The double bar represents a double denominator (with a factor of 1)
demonstrated in 1950s by Brueckner [9] and Goldstone [21] for a degenerate model
space. It holds also for a complete quasi-degenerate model space, as was first shown
by Brandow [7], using a double perturbation expansion. This was demonstrated
more directly by Lindgren [37] by means of the generalized Bloch equation (2.56),
and the result can then be formulated7
Œ˝; H0 P D .V ˝P ˝Veff /linked P: (2.82)
This equation is a convenient basis for many-body perturbation theory, as developed,

for instance, in [40]. It will also constitute a fundament of the theory developed in
this book.
2.4.2.2 Incomplete Model Spaces
When the model space is incomplete, i.e., does not contain all configurations that
can be formed by the valence orbitals, the expansion is not necessarily completely
linked. As first shown by Mukherjee [41, 56], the linked-diagram theorem can still
be shown to hold, if the normalization condition (2.42a) is abandoned. As will be
discussed later, a complete model space often has the disadvantage of so-called
intruder states, which destroy the convergence. Then also other means of circum-
venting this problem will be briefly discussed.
2.5 All-Order Methods: Coupled-Cluster Approach
2.5.1 Pair Correlation
Instead of solving the Bloch equation order-by-order, it is often more efficient to

solve it iteratively. By separating the second-quantized wave operator into normal-
ordered zero-, one-, two-,. . . body parts
˝ D ˝0 C ˝1 C ˝2 C (2.83)
7
The Rayleigh–Schrödinger and the linked-diagram expansions have the advantage compared to,
for instance, the Brillouin–Wigner expansion, that they are size-extensive, which implies that the
energy of a system increases linearly with the size of the system. This idea was actually behind the
discovery of the linked-diagram theorem by Brueckner [9], who found that the so-called unlinked
diagrams have a nonphysical nonlinear energy dependence and therefore must be eliminated in
the complete expansion. The concept of size extensivity should not be confused with the term
size consistency, introduced by Pople [64, 65], which implies that the wave function separates
correctly when a molecule dissociates. The Rayleigh–Schrödinger or linked-diagram expansions
are generally not size consistent. The coupled-cluster approach (to be discussed below) does have
this property in addition to the property of size extensivity.
2.5 All-Order Methods: Coupled-Cluster Approach 31
with
8 n o
ˆ
ˆ ˝ D c

c xji
ˆ
< 1 i j
1n o
ij (2.84)
ˆ ˝2 D ci cj cl ck xkl ;
ˆ 2
:̂
etc:
the Bloch equation can be separated into the following coupled n-particle equations
Œ˝1 ; H0 P D .V ˝ ˝W /linked; 1 P;
Œ˝2 ; H0 P D .V ˝ ˝W /linked; 2 P; (2.85)
etc., where
W D Veff D P V ˝P (2.86)
is the effective interaction.
Usually, the two-body operator dominates heavily, since it contains the important
pair correlation between the electrons. Therefore, a good approximation for many
cases is
˝ 1 C ˝1 C ˝2 ; (2.87)
which yields
Œ˝1 ; H0 P D .V1 C V ˝1 C V ˝2 ˝1 W1 /linked; 1 P;

Œ˝2 ; H0 P D .V2 C V ˝1 C V ˝2 ˝1 W2 ˝2 W1 ˝2 W2 /linked; 2 P; (2.88)
where
W1 D .V1 C V1 ˝1 /closed; 1 ;
W2 D .V2 C V ˝1 C V ˝2 /closed; 2 : (2.89)
We see here that the equations are coupled, so that ˝1 appears in the equation of ˝2
and vice versa. This approach is known as the pair-correlation approach. Solving
these coupled equations self-consistently is equivalent to a perturbation expansion
– including one- and two-body effects – to essentially all orders. It should be noted,
though, that each iteration does not correspond to a certain order of the perturbative
expansion.
As a simple illustration, we consider the simplified pair-correlation approach
˝ D ˝2 (2.90)
omitting single excitations. (This would be exact for a two-electron system using
hydrogenic basis functions, in which case there are no core orbitals, but is a good
approximation also in other cases.) The equation for ˝2 is
Œ˝2 ; H0 P D .V C V ˝2 ˝2 W2 /linked; 2 P: (2.91)

6 6
6 6 6 6 6 6
6 6
= + 6 6+ + + folded
6 6 6 6 6 6
6 6
6 6
Fig. 2.6 Graphical representation of the pair function (2.96)
Operating on an initial two-electron state of energy E, the solution can be expressed
˝2 PE D Q .E/ .V C V ˝2 ˝2 W2 /linked PE : (2.92)
Solving this iteratively leads to
˝2.1/ PE D Q .E/VPE ; (2.93)

˝2.2/ PE D Q .E/ V ˝2.1/ ˝2.1/ PE 0 W2.1/ PE
0
D Q .E/V Q .E/VPE Q .E/ Q .E /VPE 0 VPE ;
etc:; (2.94)
where all terms are assumed to be linked. This leads to the “ladder sequence,”
illustrated in Fig. 2.6. Note that in the expression above, all energies of the first
term depend on the initial state, while in the folded term the wave operator depends
on the energy of the intermediate state (E 0 ) (c.f., the “dot product,” introduced in
Sect. 6.6).
Operating with ˝2 in (2.84) on the initial statejabi leads to the pair function
˝2jabi D xab
rs
jrsi D ab .x1 ; x2 /; (2.95)
which inserted in (2.88) leads to the pair equation
."a C "b h0 .1/ h0 .2// ab .x1 ; x2 / D .jrsi hrsjV jabi Cjrsi hrsjV j ab i
j cd i hcd jW2jabi/linked : (2.96)
(For simplicity, we work with straight product functions – not antisymmetrized –

in which case we sum over all combinations of r; s (without the factor of 1/2) with
rs
xab D xab
sr
.)
We can also express the pair function as
j ab i D Q .E/ I
Pair
jabi ; (2.97)
r s r 6 6s
r s r s
= + Q + c 6P 0
E 6d
W2
a b a b
PE PE a b a b
PE PE
Fig. 2.7 Graphical representation of the self-consistent pair equation (2.99). The last diagram
represents the “folded” term ˝2 W2 . The double line represents the double denominator (double
resolvent)
where Q .E/ is the reduced resolvent (2.65) and E is the energy of the initial state
jabi. I Pair represents the ladder sequence of Coulomb interactions (including folded
terms), corresponding to the heavy line in Fig. 2.6, and including the resolvent (final
denominator) leads to the pair function j ab i. The effective interaction W2 can be
expressed as
W2 D PE 0 I Pair PE ; (2.98)
which can be represented by the same diagrams as in Fig. 2.6 (with no final denom-
inator) if the final state (with energy E 0 ) lies in the model space. The pair function
(2.92) can now be expressed
0
Q .E/I
Pair
PE D Q .E/ V CV Q .E/I
Pair
Q .E /I Pair PE 0 I Pair PE PE : (2.99)
This relation can be represented graphically as shown in Fig. 2.7.
2.5.2 Exponential Ansatz
A particulary effective form of the all-order approach is the Exponential Ansatz or

Coupled-Cluster Approach (CCA), first developed in nuclear physics by Hubbard,
Coster, and Kümmel [12, 13, 23, 33, 34]. It was introduced into quantum chemistry
by Čižek [11] and has been extensively used during the last decades for more de-
tails. (The reader is referred to a recent book “Recent Progress in Coupled Cluster
Methods” [79], which reviews the development of the methods since the start.) The
CCA is a nonlinear approach, and the linear all-order approach (2.85), discussed
above, is sometimes inadvertently referred to as “linear CCA”(!) – a term we shall
not use here. In the exponential Ansatz, the wave operator is expressed in the form
of an exponential
1 1
˝ D eS D 1 C S C S 2 C S 3 C ; (2.100)
2 3Š
S is the cluster operator (in chemical literature normally denoted by T ). It can then
be shown that for a degenerate model space the cluster operator is represented by
connected diagrams only.8 This implies that the linked but disconnected diagrams
of the wave operator are here represented by the higher powers in the expansion of
the exponential.
For open-shell systems (with unfilled valence shell), it is convenient to represent
the Ansatz in the normal-ordered form, introduced by Lindgren [38, 40],
n o 1˚ 2 1 ˚ 3
˝ D eS D 1 C S C S C S C : (2.101)
2 3Š
This form has the advantage that unwanted contractions between the cluster oper-
ators are avoided. The cluster operator is completely connected also in this case, if
the model space is complete [41], which can be formulated by means of the Bloch
equation
ŒS; H0 P D Q .V ˝P ˝Veff /conn : (2.102)
Expanding the cluster operator in analogy with the wave-operator expansion (2.83)
in terms on one-, two-,. . . body operators,
S D S1 C S2 C S3 C (2.103)
yields
n o 1˚ 1˚ 1˚ 1 ˚ 3
˝ D eS D 1CS1 CS2 C S12 CfS1 S2 gC S22 C S12 S2 C S C :
2 2 2 3Š 1
(2.104)
With the approximation
S D S1 C S2 ; (2.105)
the cluster operators satisfy the coupled Bloch equations
ŒS1 ; H0 P D .V ˝ ˝W /conn; 1 P
ŒS2 ; H0 P D .V ˝ ˝W /conn; 2 P (2.106)
illustrated in analogy with Fig. 2.7 in Fig. 2.8. These equations lead to one- and two-
particle equations, analogous to the pair equation given above (2.96). Also these
equations have to be solved iteratively, and we observe that they are coupled, as are
the corresponding equations (2.88) for the full wave operator.
8
The distinction between linked and connected diagrams should be noted. A linked diagram can
be disconnected, if all parts are open, as defined in Sect. 2.4.
6 6
6 6
S1 W = + 6 + 6 +
6 6
6 6
6 6 6
+ + + 6
6 6
W1
6 6 6
6 6 6 6 6 6 6 6 6 6

S2 W = + 6 6+ 6 6 + 6 6 +
6 6 6 6 6 6 6 6 6 6
6 6
6 6 6 6
6 6 + + + +
6 6 6 6 6 6
6 6 6 6 6 6
6 6 + 6 6 + 6 6
W2 W1 W2
Fig. 2.8 Diagrammatic representation of the equations for the cluster operators S1 and S2 (2.106).
The circle with a cross represents the “effective potential” in Fig. 2.3. The second diagram in the
second row and the diagrams in the fourth row are examples of coupled-cluster diagrams. The
last diagram in the second row and the three diagrams in the last row represent folded terms (c.f.
Fig. 2.7)
The normal-ordered scheme is usually combined with a complete model space –

or complete active space (CAS) – and the valence universality. This might lead to
problems due to intruder states to be discussed further below.
For atomic systems with essentially spherical symmetry, on the one hand, the
cluster equations can be separated into angular and radial parts, where the former
can be treated analytically and only the radial part has to have solved numerically
(see, for instance, [40, Chap. 15]). For molecular systems, on the other hand,
analytical basis-set functions of Slater or Gaussian types are normally used to solve
the coupled-cluster equations, as described in numerous articles in the field.
As mentioned, the advantage of the normal ordering of the exponential Ansatz is
that a number of unwanted contractions between open-shell operators is avoided.
More recently, Mukherjee has shown that certain valence-shell contractions are
actually desired, particularly when valence holes and strong relaxation are in-
volved [26]. He then introduced a modified normal ordering
˝ D ffexp.S /gg; (2.107)
where contractions involving passive (spectator) valence lines are reintroduced com-
pared to the original normal ordering.
2.5.3 Various Models for Coupled-Cluster Calculations:

Intruder-State Problem
The early forms of coupled-cluster models were of single-reference type (SRCC)

with a one-dimensional (closed-shell) model space. In the last few decades, vari-
ous versions of multireference (MRCC) models with multidimensional model space
have appeared (for reviews, see e.g., [2, 41, 58]). These are essentially two ma-
jor types, known as valence-universal multireference (VU-MRCC) [38, 57] and
state-universal multireference (SU-MRCC) [27, 28] methods, respectively. In the
valence-universal methods, the same cluster operators are being used for different
ionization states and therefore particularly useful for calculating ionization ener-
gies and affinities. In the state-universal methods, specific operators are used for a
particular ionization stage and particularly used when different states of the same
ionization are considered or in the molecular case for studying potential energy sur-
faces (PES).
A serious problem that can appear in MBPT with a multireference model space
is what is known as the “intruder-state-problem.” This appears when a state out-
side the model space – of the same symmetry as the state under consideration –
has a perturbed energy between those of the same symmetry originating from the
model space. This will destroy the convergence of the perturbation expansion. This
problem was first observed in nuclear physics [74], but it was early observed also in
atomic physics for the beryllium atom [69]. Here, the ground state is 1s 2 2s 2 1 S , and
the excited state 1s 2 2p 2 1 S has a low unperturbed energy, while the true state lies
close to the 2s ionization limit. This implies that when the perturbation is gradually
turned on, a large number of “outside” states, 1s 2 2s ns, will cross the energy of the
1s 2 2p 2 1 S state, and there will be no convergence beyond the crossing point.
The convergence problem due to intruders is particularly serious in perturbation
theory, when the states are expanded order-by-order from the unperturbed ones.
In the coupled-cluster approach, which in principle is nonperturbative, it might be
2.6 Relativistic MBPT: No-Virtual-Pair Approximation 37
possible to find a self-consistent solution of the coupled equations without reference

to any perturbative expansion. It was first shown by Jankowski and Malinowski [24,
25, 48] that it was in fact possible to find a solution to the beryllium problem with a
complete model space. Lindroth and Mårtensson [44] solved the same problem by
means of complex rotation.
Several other methods have been developed to reduce the intruder-state prob-
lem. One way is to reduce the model space and make it incomplete. It was shown
by Mukherjee [56] that by abandoning the intermediate normalization (2.42a), the
linked character of the diagram expansion could still be maintained. The criteria for
the connectivity of the coupled-cluster expansion have been analyzed by Lindgren
and Mukherjee [41].
Another approach to avoid or reduce the intruder-state problem is to apply an
intermediate-effective Hamiltonian, a procedure developed by the Toulouse group
(Malrieu, Durand et al.) in the mid 1980s [15]. Here, only a limited number of roots
of the secular equation are being looked for. A modified approach of the method
has been developed by Meissner and Malinowski [51] and applied to the above-
mentioned beryllium case.
A third approach to the problem is the state-specific multireference (SS-MRCC)
approach, where a multireference is used but only a single state is considered [47].
This approach can be regarded as an extreme of the intermediate-Hamiltonian ap-
proach and is frequently used particularly for studying potential-energy surfaces.
All the coupled-cluster approaches can also be applied in the relativistic formal-
ism, although applications are here still quite limited. We shall return briefly to this
problem in Chap. 8.
2.6 Relativistic MBPT: No-Virtual-Pair Approximation
In setting up a Hamiltonian for relativistic quantum mechanics, it may be tempting

to replace the single-electron Schrödinger Hamiltonian in the many-body Hamilto-
nian (2.11) by the Dirac Hamiltonian (see Appendix D)
hD D c˛ pO C ˇmc 2 C vext ; (2.108)
which with the Coulomb interaction between the electrons
X
N
e2
VC D (2.109)
40 rij
i <j
yields the Dirac–Coulomb Hamiltonian
X
N
HDC D hD .i / C VC : (2.110)
i D1
This Hamiltonian, however, has several serious shortcomings. First, it is not bound
from below, because nothing prevents the electrons from falling into the “Dirac
sea” of negative-energy electron states. A many-electron state with a mixture of
negative-energy and positive-energy electron states can then be accidentally degen-
erate with a state with only positive-energy states – a phenomenon known as the
Brown–Ravenhall disease [8]. In Chap. 6, we shall derive a field-theoretical many-
body Hamiltonian that will be used in the further development. In this model, there is
no “Dirac sea,” but the negative-energy states correspond to the creation of positron
states, which are highly excited. Then there can be no Brown–Ravenhall effect.
Within the conventional many-body treatment, the Brown–Ravenhall effect can
be circumvented by means of projection operators [78], which exclude negative-
energy states, leading to the projected Dirac–Coulomb Hamiltonian
" #
X
N
HDCproj D C hD .i / C VC C: (2.111)
i D1
Including also the instantaneous Breit interaction (see Appendix F)

" #
e 2 X ˛i ˛j .˛i r ij /.˛j r ij /
VB D C ; (2.112)
80 r ij
i <1
rij2
where ˛i is the Dirac alpha matrix vector for particle i (see Appendix D), leads to
the projected Dirac–Coulomb–Breit Hamiltonian
" #
X
N
HNVPA D C hD .i / C VC C VB C; (2.113)
i D1
which is known as the No-Virtual-Pair Approximation (NVPA).

With the partitioning (2.48)
H D H0 C V; (2.114)
we choose the model Hamiltonian to be
X
N X
N
H0 D .hD C u/i DW h0 .i / (2.115)
i i
and the perturbation

X
N
V D u.i / C VC C VB : (2.116)
i
The Dirac–Coulomb and Dirac–Coulomb–Breit Hamiltonians, which are valid

only in the Coulomb gauge (see Appendix G.2), have been extensively used in
2.6 Relativistic MBPT: No-Virtual-Pair Approximation 39
relativistic MBPT calculations and particularly in self-consistent-field calculations

of Dirac–Fock type. In the latter type of calculations, the projection operators can
often be left out, since the boundary conditions usually excludes negative-energy
solutions (see the book by I.P. Grant for a modern review [22]).
NVPA is a good approximation for many purposes, and it includes all effects to
order ˛ 2 H, but it is not Lorentz covariant (see definition in the Introduction). In
later chapters, we shall consider a more rigorous many-body Hamiltonian, based
upon field theory.
2.6.1 QED Effects
As mentioned, we shall refer to effects beyond the NVPA as QED effects, although
this separation is to some extent arbitrary. These effects are of two kinds
Nonradiative effects, representing effects due to negative-energy states and to
retardation of the Breit interaction, shown in the upper line of Fig. 2.9. These
effects are also referred to as the Araki–Sucher effects [1, 76, 77].
Radiative effects, represented by the lower line of Fig. 2.9, which are “true”
quantum-electrodynamical effects due to the electron self-energy (first diagram),
vacuum polarization (next two diagrams), or vertex correction (last diagram) (see
further Chap. 4).
-
- -
- 6
-
-
- -
?6 6 6
6
Fig. 2.9 Nonradiative (upper line) and radiative (lower line) “QED effects.” These diagrams are
Feynman diagrams, where the orbital lines can represent particle as well as hole or antiparticle
states (see further Chap. 4). The second diagram in the first row is reducible (there is an interme-
diate time with no photon), while the remaining ones are irreducible
The QED effects can also be separated into reducible and irreducible effects, where
a reducible effect is represented by a diagram that can be separated into two le-
gitimate diagrams by a horizontal cut, such as the second nonradiative diagram in
Fig. 2.9.9 Remaining diagrams are irreducible.
2.7 Some Numerical Results of Standard MBPT

and CC Calculations, Applied to Atoms
In the book Atomic Many-Body Theory [18, Sect. 15.5] a brief summary is given of
the situation in the late 1970s concerning the numerical application of many-body
perturbation theories. Most effective at that time to handle the electron-correlation
problem were various pair-correlation approaches, based on works of Kelly [32],
Meyer [52], Sinanŏglu [75], Nesbet [59], Kutzelnigg [35], and others. Coupled-
cluster methods were available at that time but still relatively undeveloped. Also
methods of treating open shells and the quasi-degenerate problem, using the ex-
tended model space [37] (2.56), were available but not particularly well known.
In the three decades that have followed, a dramatic development regarding nu-
merical implementations has taken place. All-order methods, in particular, coupled-
cluster methods, have been developed to a stage of “almost perfection.” Also, the
open-shell techniques have been further developed and are now routinely used.
Here, two main lines have emerged, based upon multireference or single-reference
states. The latter technique has been developed mainly to circumvent the intruder
problem, although there are methods of dealing with this problem also in the mul-
tireference case, as was briefly mentioned above. We shall in no way try to review
this immense field here but limit ourselves to some comments concerning develop-
ments that are most relevant for the theme of this book. (We refer to the previously
mentioned book, edited by Čársky et al. [79], for more details.) We also call atten-
tion to a comprehensive review of all-order relativistic atomic calculations that has
recently been published by Safronova and Johnson [67].
The coupled-cluster approach was early applied to various molecular systems,
particularly by Čižek, Paldus, and coworkers in the Waterloo group [60, 61]. Ex-
tensions of the method and extensive calculations have been performed by Bartlett
and his collaborators at Gainesville [3, 66]. The paper by Purvis and Bartlett [66],
together with the simultaneous publication by Pople et al. [63], represent the first ap-
plications of CCA with both single and double excitations (CCSDs). Bartlett et al.
have later extended the technique to include part of triples, CCSD(T), and quadru-
ples, CCSD(TQ), techniques that are now widely spread.
9
Unfortunately, different definitions of reducible and irreducible diagrams occur in the literature.
We use in this book the original interpretation of the concepts, due to the pioneers Feynman, Dyson,
Bethe, Salpeter, and others [16, 73].
2.7 Some Numerical Results of Standard MBPT and CC Calculations, Applied to Atoms 41
In molecular calculations, functional basis sets of Slater or Gaussian-type are

normally used. For atomic systems, it is normally preferable to use numerical inte-
gration of the radial coordinates. Such techniques have been developed and applied
particularly by the groups at Notre Dame, Gothenburg, and Tel Aviv.
The Notre-Dame group has for a long time performed relativistic many-body cal-
culations on atomic systems by applying and further developed the spline technique
with piece-wise polynomial fitting [29]. This was first used for calculations to sec-
ond order (third order in energy) of the helium atom, and the sodium isoelectronic
sequence [30]. The method was then extended by Blundell et al. [6] to an all-order
technique (linear with singles and doubles) and applied to the Li atom and the BeC
ion and by Plante et al. [62] to a sequence of helium-like ions. In Table 2.1, we repro-
duce from the latter work the contributions to the ground-state ionization energies
due to (a) all-order Coulomb interactions, (b) same with one instantaneous Breit in-
teraction, (c) same with TWO instantaneous Breit interactions, (d) first-order QED
contribution (from [14]), (e) total ionization energy. Later, the Notre–Dame group,
partly together with Safronova, has extended the technique to full relativistic CCSD,
including also some triples, CCSD(T), and applied it extensively to various atomic
and ionic systems [46, 67] (see Tables 2.2 and 2.3).
The Gothenburg group developed numerical nonrelativistic all-order and
coupled-cluster approaches in the late 1970s and early 1980s. Mårtensson (Pen-
drill) [49] developed an all-order pair program (LD) – linear with doubles without
coupled clusters – based upon the first-order pair program developed by Mor-
rison [20, 54], and first applied it to the helium atom. This technique was later
Table 2.1 Contributions to the ground-state ionization energies of helium-

like ions. From [62] (in Hartrees)
Z Coulomb Breit Double Breit QED Total
10 43.962 0.010708 0.000048 0:004610 43.946
20 188.636 0.096696 0.000433 0:054905 168.485
40 792.126 0.83482 0.00409 0:57860 790.717
60 1,855.119 2.97236 0.01528 2:22984 1,849.832
80 3,472.330 7.51789 0.03914 5:89519 3,458.965
100 5,841.499 16.0999 0.0836 12:9704 5,812.513
Table 2.2 Binding energies of the two lowest states of

the lithium atom (in H)
Lithium atom
2 2S 2 2P References
Expt’l 198 159 130 246
Hartree–Fock 196 304 128 637
Difference 1 854 1 609
LSD 1 855 1 582 Blundell et al. [6]
CCSD 1 850 1 584 Lindgren [39]
CCSD 1 835 1 534 Eliav et al. [17]
Table 2.3 Correlation energy of some low-lying states of the sodium

atom (in H) (from [68])
Sodium atom
3 2 S 3 2 P1=2 3 2 P3=2 4 2 S References
Expt’l 6 825 2121 2110 1415
LSD 6 835 2118 2108 1418 Safronova et al. [68]
CCSD 6 458 Salomonson-Ynnerman [72]
CCSD 6 385 Eliav et al. [17]
CCSD(T) 6 840 Salomonson-Ynnerman [72]
Table 2.4 Correlation energy of the ground state of

the beryllium atom and the negative lithium ion (in
H) (from [10])
Beryllium atom and negative lithium ion
Be Li References
CCD 92:960 71:148 Bukowski et al. [10]
CCD 92:961 71.266 Salomonson-öster [70]
CCSD 93:665 72.015 Bukowski et al. [10]
CCSD 93:667 72.142 Salomonson-öster [70]
converted into a coupled-cluster program with doubles (CCD) by Salomonson [42]

and applied to various atomic systems. It was also applied to open-shell sys-
tems [55, 69] – in the second paper (concerning the beryllium atom) the famous
intruder problem, mentioned above, was probably observed for the first time in an
atomic system. The procedure of the Gothenburg group was extended to the full
CCSD procedure and applied by Lindgren [39] (see Table 2.2) and Salomonson
et al. [70, 72] (see Tables 2.3 and 2.4).
A relativistic version of the linear all-order pair program (LD) was developed
by Eva Lindroth [43], and applied to the helium atom. This was extended to a
relativistic coupled-cluster program by Salomonson and öster, who also devel-
oped a new numerical, highly accurate technique, referred to as the discretization
technique [71]. This technique was early applied relativistically as well as nonrela-
tivistically to a number of atomic systems [70, 72] and is used also in all later works
of the group.
In Tables 2.2–2.4, we have compared some all-order calculations for the lithium,
sodium, and beryllium atoms as well as for the Li ion. The calculations on Be and
Li demonstrate clearly the importance of single excitations in this case. The results
for sodium show the importance of triple excitations in this case. (The results by
Safronova et al. are probably fortuitous, indicating that effects of nonlinear coupled-
cluster terms and triples accidentally cancel.) The accurate results from numerical
integrations by Salomonson et al. are sometimes used as benchmarks for testing
calculations with finite basis sets [10].
References 43
The Tel-Aviv group has applied the relativistic coupled-cluster technique with
singles and doubles (CCSD) particularly to very heavy atoms and simple molecules
(see, for instance, the review article by Kaldor and Eliav [31], as well as Tables 2.2
and 2.3).
References

Phys. (Japan) 17, 619–42 (1957)
2. Bartlett, R.J.: Coupled-cluster approach to molecular structure and spectra: A step toward
predictive quantum chemistry. J. Phys. Chem. 93, 1697–1708 (1989)
3. Bartlett, R.J., Purvis, G.D.: Many-Body Perturbation Theory, Coupled-Pair Many-Electron
Theory, and the Importance of Quadruple Excitations for the Correlation Problem. Int. J.
Quantum Chem. 14, 561–81 (1978)
4. Bloch, C.: Sur la détermination de l’etat fondamental d’un système de particules. Nucl. Phys.
7, 451–58 (1958)
5. Bloch, C.: Sur la théorie des perurbations des etats liés. Nucl. Phys. 6, 329–47 (1958)
6. Blundell, S.A., Johnson, W.R., Liu, Z.W., Sapirstein, J.: Relativistic all-order calculations
of energies and matrix elements for Li and BeC . Phys. Rev. A 40, 2233–46 (1989)
7. Brandow, B.H.: Linked-Cluster Expansions for the Nuclear Many-Body Problem. Rev. Mod.
Phys. 39, 771–828 (1967)
8. Brown, G.E., Ravenhall, D.G.: On the Interaction of Two electrons. Proc. R. Soc. London,
Ser. A 208, 552–59 (1951)
9. Brueckner, K.A.: Many-Body Problems for Strongly Interacting Particles. II. Linked Cluster
Expansion. Phys. Rev. 100, 36–45 (1955)
10. Bukowski, R., Jeziorski, B., Szalewicz, K.: Gaussian geminals in explicitly correlated coupled
cluster theory including single and double excitations. J. Chem. Phys. 110, 4165–83 (1999)
11. Čižek, J.: On the Correlation Problem in Atomic and Molecular Systems. Calculations
of Wavefunction Components in the Ursell-Type Expansion Using Quantum-Field Theoreti-
cal Methods. J. Chem. Phys. 45, 4256–66 (1966)
12. Coster, F.: Bound states of a many-particle system. Nucl. Phys. 7, 421–24 (1958)
13. Coster, F., Kümmel, H.: Short-range correlations in nuclear wave functions. Nucl. Phys. 17,
477–85 (1960)
14. Drake, G.W.F.: Theoretical energies for the n D 1 and 2 states of the helium isoelectronic
sequence up to Z=100. Can. J. Phys. 66, 586–611 (1988)
15. Durand, P., Malrieu, J.P.: Multiconfiguration Dirac-Fock studies of two-electron ions: II. Ra-
diative corrections and comparison with experiment. Adv. Chem. Phys. 67, 321 (1987)
16. Dyson, F.J.: The radiation Theories of Tomonaga, Schwinger, and Feynman. Phys. Rev. 75,
486–502 (1949)
17. Eliav, E., Kaldor, U., Ishikawa, Y.: Ionization potentials and excitation energies of the alkali-
metal atoms by the relativistic coupled-cluster method. Phys. Rev. A 50, 1121–28 (1994)
18. E.Lindroth, Mårtensson-Pendrill, A.M.: Isotope Shifts and Energies of the 1s2p States in He-
lium. Z. Phys. A 316, 265–273 (1984)
N.Y. (1971)
20. Garpman, S., Lindgren, I., Lindgren, J., Morrison, J.: Calculation of the hyperfine interaction
using an effective-operator form of many-body theory. Phys. Rev. A 11, 758–81 (1975)
21. Goldstone, J.: Derivation of the Brueckner many-body theory. Proc. R. Soc. London, Ser. A
239, 267–279 (1957)
22. Grant, I.P.: Relativistic Quantum Theory of Atoms and Molecules. Springer, Heidelberg (2007)
23. Hubbard, J.: The description of collective motions in terms of many-body perturbation theory.
3. The extension of the theory to non-uniform gas. Proc. R. Soc. London, Ser. A 243, 336–52
(1958)
24. Jankowski, K., Malinowski, P.: A valence-universal coupled-cluster single- and double-
excitations method for atoms: II. Application to Be. J. Phys. B 27, 829–42 (1994)
25. Jankowski, K., Malinowski, P.: A valence-universal coupled-cluster single- and double-
excitations method for atoms: III. Solvability problems in the presence of intruder states. J.
Phys. B 27, 1287–98 (1994)
26. Jena, D., Datta, D., Mukherjee, D.: A novel VU-MRCC formalism for the simultaneous treat-
ment of strong relaxation and correlation effects with applications to electron affinity of
neutral radicals. Chem. Phys. 329, 290–306 (2006)
27. Jeziorski, B., Monkhorst, H.J.: ACoupled-cluster method for multideterminantal reference
states. Phys. Rev. A 24, 1668–81 (1981)
28. Jeziorski, B., Paldus, J.: Spin-adapted multireference coupled-cluster approach: Linear ap-
proximation for two closed-shell-type reference configuration. J. Chem. Phys. 88, 5673–87
(1988)
29. Johnson, W.R., Blundell, S.A., Sapirstein, J.: Finite basis sets for the Dirac equation con-
structed from B splines. Phys. Rev. A 37, 307–15 (1988)
30. Johnson, W.R., Blundell, S.A., Sapirstein, J.: Many-body perturbation-theory calculations
of energy-levels along the sodium isoelectronic sequence. Phys. Rev. A 38, 2699–2706 (1988)
31. Kaldor, U., Eliav, E.: High-Accuracy Calculations for Heavy and Super-Heavy Elements. Adv.
Quantum Chem. 31, 313–36 (1999)
32. Kelly, H.P.: Application of many-body diagram techniques in atomic physics. Adv. Chem.
Phys. 14, 129–190 (1969)
33. Kümmel, H.: In: E.R. Caianiello (ed.) Lecture Notes on Many-Body Problems. Academic
Press, N.Y. (1962)
34. Kümmel, H., Lührman, K.H., Zabolitsky, J.G.: Many-fermion theory in exp S (or coupled
cluster) form. Phys. Rep. 36, 1–135 (1978)
35. Kutzelnigg, W.: Methods in Electronic Structure Theory. Plenum, New York (1977)
36. Kvasnička, V.: Construction of Model Hamiltonians in framework of Rayleigh- Schrodinger
perturbation-theory. Czech. J. Phys. B 24, 605 (1974)
37. Lindgren, I.: The Rayleigh-Schrödinger perturbation and the linked-diagram theorem for a
multi-configurational model space. J. Phys. B 7, 2441–70 (1974)
38. Lindgren, I.: A Coupled-Cluster Approach to the Many-Body Perturbation Theory for Open-
Shell Systems. Int. J. Quantum Chem. S12, 33–58 (1978)
39. Lindgren, I.: Accurate many-body calculations on the lowest 2 S and 2 P states of the lithium
atom. Phys. Rev. A 31, 1273–86 (1985)
41. Lindgren, I., Mukherjee, D.: On the connectivity criteria in the open-shell coupled-cluster
theory for the general model spaces. Phys. Rep. 151, 93–127 (1987)
42. Lindgren, I., Salomonson, S.: A Numerical Coupled-Cluster Procedure Applied to the Closed-
Shell Atoms Be and Ne. Presented at the Nobel Symposium on Many-Body Effects in Atoms
and Solids, Lerum, 1979. Physica Scripta 21, 335–42 (1980)
43. Lindroth, E.: Numerical soultion of the relativistic pair equation. Phys. Rev. A 37, 316–28
(1988)
44. Lindroth, E., Mårtensson-Pendrill, A.M.: 2p 2 1 S state of Beryllium. Phys. Rev. A 53, 3151–56
(1996)
45. Löwdin, P.O.: Studies in Perturbation Theory. X. Lower Bounds to Eigenvalues in
Perturbation-Theory Ground State. Phys. Rev. 139, A357–72 (1965)
46. M. S. Safronova, W.R.J., Derevianko, A.: Relativistic many-body calculations of energy levels,
hyperfine constants, electric-dipole matrix elements, and static polarizabilities for alkali-
metal atoms. Phys. Rev. A 560, 4476–87 (1999)
References 45
47. Mahapatra, U.S., Datta, B., Mukherjee, D.: Size-consistent state-specific multireference
couloed cluster theory: Formal developmnents and molecular applications. J. Chem. Phys.
110, 6171–88 (1999)
48. Malinowski, P., Jankowski, K.: A valence-universal coupled-cluster single- and double-
excitation method for atoms. I. Theory and application to the C2C ion. J. Phys. B 26, 3035–56
(1993)
49. Mårtensson, A.M.: An Iterative, Numerical Procedure to obtain Pair Functions Applied to
Two-electron Systems. J. Phys. B 12, 3995–4012 (1980)
50. Mårtensson-Pendrill, A.M., Lindgren, I., Lindroth, E., Salomonson, S., Staudte, D.S.: Con-
vergence of relativistic perturbation theory for the 1s2p states in low-Z heliumlike systems.
Phys. Rev. A 51, 3630–35 (1995)
51. Meissner, L., Malinowski, P.: Intermediate Hamiltonian formulation of the valence-universal
coupled-cluster method for atoms. Phys. Rev. A 61, 062,510 (2000)
52. Meyer, W.: PNO-CI studies of electron correlation effects. 1. Configuration expansion
by means of nonorthogonal orbitals, and application to ground-state and ionized states
of methane. J. Chem. Phys. 58, 1017–35 (1973)
53. Møller, C., Plesset, M.S.: Note on an Approximation Treatment for Many-Electron Systems.
Phys. Rev. 46, 618–22 (1934)
54. Morrison, J.: Many-Body calculations for the heavy atoms. III. Pair correlations. J. Phys. B
6, 2205–12 (1973)
55. Morrison, J., Salomonson, S.: Many-Body Perturbation Theory of the Effective Electron-
Electron Interaction for Open-shell Atoms. Presented at the Nobel Symposium.., Lerum, 1979.
Physica Scripta 21, 343–50 (1980)
56. Mukherjee, D.: Linked-Cluster Theorem in the Open-Shell Coupled-Cluster Theory for In-
complete Model Spaces. Chem. Phys. Lett. 125, 207–12 (1986)
57. Mukherjee, D., Moitra, R.K., Mukhopadhyay, A.: Application of non-perturbative many-body
formalism to open-shell atomic and molecular problems - calculation of ground and lowest
singlet and triplet energies and first ionization potential of trans-butadine. Mol. Phys
33, 955–969 (1977)
58. Mukherjee, D., Pal, J.: . Adv. Chem. Phys. 20, 292 (1989)
59. Nesbet, R.K.: Electronic correlation in atoms and molecules. Adv. Chem. Phys. 14, 1 (1969)
60. Paldus, J., Čižek, J.: Time-Independent Diagrammatic Approach to Perturbation Theory
of Fermion Systems. Adv. Quantum Chem. 9, 105–197 (1975)
61. Paldus, J., Čižek, J., Saute, M., Laforge, A.: Correlation problems in atomic and molecular
systems. IV. Coupled-cluster approach to open shells. Phys. Rev. A 17, 805–15 (1978)
the n=1 and n=2 states of heliumlike ions. Phys. Rev. A 49, 3519–30 (1994)
63. Pople, J.A., Krishnan, R., Schlegel, H.B., Binkley, J.S.: Electron Correlation Theories and
Their Application to the Study of Simple Reaction Potential Surfaces. Int. J. Quantum Chem.
14, 545–60 (1978)
64. Pople, J.A., Santry, D.P., Segal, G.A.: Approximate self-consistent molecular orbital theory.I.
Invariant procedures. J. Chem. Phys. 43, S129 (1965)
65. Pople, J.A., Segal, G.A.: Approximate self-consistent molecular orbital theory. 3. CNDO re-
sults for AB2 and AB3 systems. J. Chem. Phys. 44, 3289 (1966)
66. Purvis, G.D., Bartlett, R.J.: A full coupled-cluster singles and doubles model: The inclusion of
disconnected triples. J. Chem. Phys. 76, 1910–18 (1982)
67. Safronova, M.S., Johnson, W.R.: All-Order Methods for Relativistic Atomic Structure Calcu-
lations, Advances in Atomic Molecular and Optical Physics series, vol. 55 (2007)
68. Safronova, M.S., Sapirstein, J., Johnson, W.R.: Relativistic many-body calculations of energy
levels, hypeerfine constants, and transition rates for sodiumlike ions. Phys. Rev. A 58, 1016–
28 (1998)
69. Salomonson, S., Lindgren, I., Mårtensson, A.M.: Numerical Many-Body Perturbation Calcu-
lations on Be-like Systems Using a Multi-Configurational Model Space. Presented at the Nobel
Symposium on Many-Body Effects in Atoms and Solids, Lerum, 1979. Physica Scripta 21,
335–42 (1980)
70. Salomonson, S., Öster, P.: Numerical solution of the coupled-cluster single- and double-
excitation equations with application to Be and Li . Phys. Rev. A 41, 4670–81 (1989)
71. Salomonson, S., Öster, P.: Relativistic all-order pair functions from a discretized single-
particle Dirac Hamiltonian. Phys. Rev. A 40, 5548, 5559 (1989)
72. Salomonson, S., Ynnerman, A.: Coupled-cluster calculations of matrix elements and ioniza-
tion energies of low-lying states of sodium. Phys. Rev. A 43, 88–94 (1991)
1232–42 (1951)
74. Schäfer, L., Weidenmüller, H.A.: Self-consistency in application of Brueckner’s method to
doubly-closed-shell nuclei. Nucl. Phys. A 174, 1 (1971)
75. Sinanŏglu, O.: Electronic correlation in atoms and molecules. Adv. Chem. Phys. 14, 237–81
(1969)
76. Sucher, J.: Energy Levels of the Two-Electron Atom to Order ˛ 3 Ry; Ionization Energy of He-
lium. Phys. Rev. 109, 1010–11 (1957)
Michigan
78. Sucher, J.: Foundations of the Relativistic Theory of Many Electron Atoms. Phys. Rev. A 22,
348–62 (1980)
79. Čársky, P., Paldus, J., Pittner, J. (eds.): Recent Progress in Coupled Cluster Methods: Theory
and Applications. Springer (2009)
80. Wick, C.G.: The Evaluation of the Collision Matrix. Phys. Rev. 80, 268–272 (1950)
Chapter 3
Time-Dependent Formalism
In this chapter, we shall summarize the fundamentals of time-dependent perturbation

theory. Although we shall be only concerned with stationary problems in this book,
it will be advantageous to apply time-dependent methods. We restrict ourselves in
this chapter to the nonrelativistic formalism and return to the relativistic one in later
chapters.
3.1 Evolution Operator
It follows from the second-quantized Schrödinger equation (2.16) that the state vec-
tor evolves in time according to
jS .t/i D eiH.t t0 /=„ jS .t0 /i : (3.1)
This is known as the Schrödinger picture (SP), indicated by the subscript “S.” In an-
other representation, known as the interaction picture (IP) [see Appendix B, (B.23)],
the Hamiltonian is partitioned according to (2.48), H D H0 C V , and the state vec-
tors and the operators are transformed according to
jI .t/i D eiH0 t =„ jS .t/i I OI .t/ D eiH0 t =„ OS eiH0 t =„ : (3.2)
This implies that the state vectors are normally much more slowly varying with
time, and most of the time dependence is instead transferred to the operators that
are normally time independent in SP.
The Schrödinger equation is in IP transformed to
@
i„ jI .t/i D VI .t/ jI .t/i; (3.3)
@t
with the solution

Z t
i
jI .t/i DjI .t0 /i dt1 VI .t1 / jI .t1 /i : (3.4)
„ t0

48 3 Time-Dependent Formalism
VI .t/ is the perturbation in the interaction picture, which is assumed to be time

independent in the Schrödinger picture.
For a stationary state of energy E, the time dependence (2.15) is eiE t =„ . It then
follows that the state in the IP is of the form
jI .t/i D eit .E H0 /=„ jI .t D 0/i : (3.5)
The time-evolution operator in IP, UI .t; t0 /, is defined by the relation
jI .t/i D UI .t; t0 / jI .t0 /i .t > t0 /: (3.6)
Evidently, we then have
UI .t; t/ D 1; (3.7)
UI .t; t1 / UI .t1 ; t2 / D UI .t; t2 /: (3.8)
From the relation (3.1), it follows that the corresponding evolution operator in SP is
US .t; t0 / D eiH.t t0 /=„ : (3.9)
Transforming (3.2) to IP then yields1
UI .t; t0 / D eiH0 t =„ eiH.t t0 /=„ eiH0 t0 =„ : (3.10)
This evolution operator satisfies the differential equation
@
i„ UI .t; t0 / D VI .t/ UI .t; t0 /; (3.11)
@t
which leads to the expansion2

Z t
i
U.t; t0 / D 1 dt1 V .t1 / U.t1 ; t0 /
„ t0
Z t i 2 Z t Z t1
i
D 1 dt1 V .t1 / C dt1 V .t1 / dt2 V .t2 / U.t2 ; t0 /;
„ t0 „ t0 t0
(3.12)
1
It should be noted that generally eiH0 t=„ eiH t=„ ¤ eiV t=„ , since the operators do not necessarily
commute.
2
Unless specified otherwise, we shall in the following assume that the evolution operators always
are expressed in IP and leave out the subscript I .
3.1 Evolution Operator 49
etc. By extending the second integration from t0 to t, this can be expressed [2,
Fig. 6.1]
Z Z Z
i t
1 i 2 t t
U.t; t0 / D 1 dt1 V .t1 / C dt1 dt2 T V .t1 /V .t2 / U.t2 ; t0 /;
„ t0 2 „ t0 t0
(3.13)
where T is the time-ordering operator, which orders the operators after decreasing
time (without any sign change). This leads to the expansion
Z
i t
U1 .t; t0 / D dt1 V .t1 /;
„ t0
Z Z t
1 i 2 t
U2 .t; t0 / D dt1 dt2 T V .t1 / V .t2 / ; (3.14)
2 „ t0 t0
etc., which can be generalized to [2, Eq. 6.23], [4, Eq. 4–56]
Z Z t
1 i n t
X1
U.t; t0 / D dt1 : : : dtn T V .t1 / : : : V .tn / : (3.15)
nD0
nŠ „ t0 t0
(We have here included the term n D 0 to replace the unity.)

We introduce the Hamiltonian density H.x/ by
Z
V .t/ D d3 x H.t; x/: (3.16)
We do not have to specify the perturbation at this point, but we shall later assume
that it is given by the interaction between the electrons (of charge e) and the elec-
tromagnetic radiation field (see Appendix E.3)
H.x/ D O .x/ec˛ A .x/ O .x/: (3.17)
Here, ˛ is the Dirac operator (see Appendix D) and A is the covariant radiation
field [Appendix (G.2)]
s
„ X
A .x/ D
"r akr ei kx C akr ei kx : (3.18)
20 !V kr
The evolution operator (3.15) can now be expressed
Z Z t
1 i n t 4
X1
U.t; t0 / D dx1 : : : dxn4 T H.x1 / : : : H.xn / : (3.19)
nD0
nŠ c„ t 0 t 0
The factor of c in the denominator is due to the fact that we now use the integration
variable x0 D ct. The integrations are performed over all space and over time as
indicated. Alternatively, this can be expressed
" #
i Z t
U.t; t0 / D T exp d x H.x/
4
: (3.20)
c„ t0
The evolution operator can be represented graphically be means of Goldstone

diagrams in the same way as the wave operator, discussed previously. As a simple
example, we consider the first-order interaction with a time-independent potential
interaction v.x/. In second quantization, the evolution operator becomes
Z t
i
U .1/ .t; t0 / D dt cr hrjv.x 1 /jai ca (3.21)
„ t0
or after summing over the states

Z t Z
i
U .1/ .t; t0 / D dt1 d3 x 1 O .x1 / v.x 1 / O .x1 /; (3.22)
„ t0
which is illustrated in Fig. 3.1 (left).

The two-body interaction can be given by a contraction of two perturbations
(3.17), corresponding to the exchange of one virtual photon, v.x1 ; x2 /, as will be
further discussed in Chap. 4. The corresponding, second-order evolution operator
then becomes (Fig. 3.1, right)
ZZ t
1 i 2
U .2/ .t; t0 / D dt1 dt2
2 „ t0
ZZ
d3 x 1 d3 x 2 O .x1 / O .x2 / v.x1 ; x2 / O .x2 / O .x1 /: (3.23)
In higher orders, the operator can have connected as well as disconnected parts
and can be separated into zero-, one-, two-,. . . body parts. The connected one- and
t t
O r O r s O
x2
x1 -
x1
O a O a b O
t0 t0
Fig. 3.1 Graphical representation of the evolution operator for first-order potential interaction and
single-photon exchange
3.2 Adiabatic Damping: Gell-Mann–Low Theorem 51
t t
O r O r s O
O a O a b O
t0 t0
Fig. 3.2 Schematic graphical representation of the connected one- and two-body parts of the evo-
lution operator
two-body pieces are schematically illustrated in Fig. 3.2. Expressions with uncon-
tracted photons fall in an extended photonic Fock space, as will be further discussed
in later chapters.
3.2 Adiabatic Damping: Gell-Mann–Low Theorem
For the mathematical treatment, we shall find it convenient to apply an “adiabatic

damping factor” to the perturbation,
V .t/ ! V .t/ e jt j ; (3.24)
where is a small, positive number, which implies that
H ! H0 as t ! 1: (3.25)
The expansion (3.19) then becomes
1 i n
X1
U .t; t0 / D
nD0
nŠ c„
Z t Z t
dx14 : : : dxn4 T H.x1 / : : : H.xn / e.jt1 jCjt2 j:::Cjtn j/ :
t0 t0
(3.26)
The damping is adiabatically ‘switched off’ at the end of the calculation. The evo-
lution operator satisfies now the equation (3.11)
@
i„ U .t; t0 / D V .t/ i U .t; t0 /; (3.27)
@t
where the upper sign is valid for t > 0.
3.2.1 Gell-Mann–Low Theorem
The damped perturbation (3.24) vanishes, when t ! ˙1, and the perturbed
(target) state vector approaches in these limits an eigenstate of H0 ,
ˇ ˛ ˇ ˛
ˇI .t/ ) ˇ˚ ; (3.28)
which we call the parent state. Gell-Mann and Low have shown that for t D 0 and
in the limit ! 0, the state vector
˛ U .0; 1/j˚i
lim jI .0/ D lim DWj i (3.29)
!0 !0 h˚jU .0; 1/j˚i
is a solution of the time-independent Schrödinger equation
.H0 C V /j i D Ej i ; (3.30)
where H0 is the model Hamiltonian (2.49) without the interaction. Here,
@
h˚j @ U .0; 1/j˚i
E D E0 C i„ : (3.31)
h˚jU .0; 1/j˚i
This is the famous Gell-Mann–Low theorem (GML) [3], [2, p. 61], [10, p. 336],
which represents one of the fundamentals of the theory presented here. The pertur-
bation, V , must in the limit ! 0 be time independent in the Schrödinger picture,
which is the case with the interaction (3.17).
3.3 Extended Model Space: The Generalized Gell-Mann–Low

Relation
The original Gell-Mann–Low theorem (3.29) is valid only in the single-reference

case (one-dimensional model space). The time-dependent MBPT was in 1960s and
1970s further developed by several groups [1,5,6,8,9,11], mainly in connection with
nuclear calculations. We shall extend this treatment here and prove a generalization
of the Gell-Mann–Low theorem for an arbitrary model space. This treatment follows
mainly that performed in [7] (see also [2, Sect. 6]).
We choose the parent states to be the (normalized) limits of the target states for
finite as t ! 1, as introduced by Tolmachev [11],
ˇ ˛˛ ˇ ˛
ˇ˚ D C ˛ lim ˇ˛ .˛ D 1; 2 d /; (3.32)
t !1
3.3 Extended Model Space: The Generalized Gell-Mann–Low Relation 53
where C ˛ is a normalization constant. The parent functions are then eigenfunctions

of H0 ,
ˇ ˛ ˇ ˛
H0 ˇ˚ ˛ D E0˛ ˇ˚ ˛ ; (3.33)
but generally we do not know which eigenvalue a specific target state will con-
verge to.
In analogy with the single-reference case (3.29), we construct the state
ˇ ˛
ˇ ˛ U .0; 1/ ˇ˚ ˛
ˇ ˛
D˝ ; (3.34)

0 jU .0; 1/j˚ i
˛ ˛
which is normalized in the intermediate normalization, h 0˛ j ˚ ˛ i.

We shall now demonstrate that this state is in the limit ! 0 an eigenstate of
the time-independent Hamiltonian of the system for all values of ˛,
ˇ ˛ ˇ ˛
H0 C V ˇ ˛
D E ˛ˇ ˛
.˛ D 1; 2; d /: (3.35)
This is a generalization of the original Gell-Mann–Low relation (3.29), and it

holds also for a quasi-degenerate model space with several energy levels [7].
In order to prove the theorem, we consider one term in the expansion (3.26)
Z
1 i n t
U.n/ .t; 1/ D dtn
nŠ „ 1
Z t
dtn1 T V .tn /V .tn1 / e.t1 Ct2 :::Ctn / : (3.36)
1
(As long as t does not approach C1, we can leave out the absolute signs in the
damping factor.) Using the identity
ŒH0 ; ABC D ŒH0 ; A BC C AŒH0 ; B C C ;
we obtain, noting that in IP VI .t/ D eitH0 =„ VS eitH0 =„ and that V is assumed to be

time independent in the SP,
@V12 .t/
D i ŒH0 ; V12 .t/ (3.37)
@t
and
@ @
H0 ; V .tn /V .tn1 / D i„ C C V .tn /V .tn1 / : (3.38)
@tn @tn1
This gives
Z Z t
1 i .n1/ t
H0 ; U.n/ .t; 1/ D dtn dtn1
nŠ „ 1 1
h @ @ i
T C C V .tn /V .tn1 /
@tn @tn1
e.t1 Ct2 :::Ctn / :
When integrating by parts, each term gives the same contribution, yielding
H0 ; U.n/ .t; 1/ D V .t/ U.n1/ .t; 1/ C i„ n U.n/ .t; 1/; (3.39)
where the last term originates from derivating the damping term. Introducing an
order parameter, ,
H D H0 C V .t/; (3.40)
the result can be expressed
@
H0 ; U .t; 1/ D V .t/ U .t; 1/ C i„ U .t; 1/: (3.41)
@
By operating with the commutator on the parent function (3.32), utilizing the fact
that the parent state ˚ ˛ is an eigenstate of H0 , we obtain for t D 0
ˇ ˛ @ ˇ ˛
H0 E0˛ C V U .0; 1/ ˇ˚ ˛ D i„ U .0; 1/ ˇ˚ ˛ ; (3.42)
@
where V D V .0/ or with the state (3.34)
ˇ ˛ @
U .0; 1/j˚ ˛ i
H0 C V E0˛ ˇ ˛
D i„ ˝ @˛ : (3.43)
0 jU .0; 1/j˚ i
˛
(Note that at t D 0 the Schrödinger and interaction pictures are identical.) We note
from the relation (3.34), that
ˇ ˛ ˇ ˛
@ ˛ @ U .0; 1/ ˇ˚ ˛ @
U .0; 1/ ˇ˚ ˛
j ˛
D ˝ ˇ ˇ ˛ D ˝ ˛ˇ
@ ˇ ˛
@
@ 0˛ ˇU .0; 1/ˇ ˚ ˛ ˇ
0 U .0; 1/ ˚
ˇ ˛
D ˇ ˇ E
˛ˇ @ ˇ ˛ ˇ ˛
ˇ U
0 @ .0; 1/ ˇ ˚ U .0; 1/ ˇ 0˛
˝ ˛ˇ ˇ ˛ ˝ ˛ˇ ˇ ˛: (3.44)
ˇ ˇ ˛ ˇ ˇ ˛
0 U .0; 1/ ˚ 0 U .0; 1/ ˚
Therefore, the r.h.s. of (3.43) can be expressed
@
@ U .0; 1/j˚ i
˛ ˇ ˛ @ ˇˇ ˛
i„ ˝ ˛ˇ
ˇ ˇ ˛ D E˛ ˇ ˛
C i ˛
;
ˇ ˛
0 U .0; 1/ ˚ @
3.3 Extended Model Space: The Generalized Gell-Mann–Low Relation 55
where D ˇ ˇ E
˛ˇ @ ˇ ˛
0 ˇ @ U .0; 1/ˇ ˚
E˛ D i„ ˝ ˇ ˇ ˛
˛ ˇU .0; 1/ˇ ˚ ˛
0
and this yields

ˇ ˛ @ ˇˇ ˛
H0 C V E0˛ E˛ ˇ ˛
D i„ ˛
: (3.45)
@
˛
˛
Provided that the perturbation expansion ofj converges, the r.h.s. will vanish as
! 0. Then
The generalized Gell-Mann–Low (GML) relation reads
ˇ ˛
ˇ ˛ ˇ ˛ U .0; 1/ˇ˚ ˛
ˇ ˛
D lim ˇ ˛
D lim ˝ ˛ ˇ ˇ ˛: (3.46)
!0

!0 ˇU .0; 1/ ˇ ˚ ˛
0
This state vector will satisfy the time-independent Schrödinger equation

ˇ ˛ ˇ ˛
H0 C V .0/ ˇ ˛
D E˛ ˇ ˛
; (3.47)
where H0 is the model or independent-particle Hamiltonian 2.49 and V is the

perturbation (3.16).
This relation is derived in the interaction picture with t D 0, which implies that
it holds also in the Schrödinger picture (SP). The perturbation must be time-
independent in SP, apart from possible damping, as is the case with the perturba-
tion (3.17).
The energy eigenvalue corresponding the Gell-Mann–Low state (3.46) becomes
2 D ˇ ˇ E
˛ˇ @ ˇ ˛ 3
0 ˇ @ U .0; 1/ˇ ˚
E ˛ D lim 4E0˛ C i„ ˝ ˛ ˇ ˇ ˛ 5: (3.48)
ˇ ˇ ˛
0 U .0; 1/ ˚
!0
This expression is not very useful for evaluating the energy, since the eigenvalue
E0˛ of the parent state is generally not known. The procedure is here used mainly
to demonstrate that the functions satisfy the Schrödinger equation. Instead, we shall
derive an expression for the effective Hamiltonian (2.54), which is the natural tool
for a multilevel model space.3
In the one-dimensional model space, singularities appear in U for unlinked
terms. In the general multidimensional case, singularities can appear also for
3
It should be noted that a necessary condition for the proof of the theorem given here is that the
parent state (3.33) is an eigenstate of the model Hamiltonian H0 [see (3.42)]. This is in conflict
with the statement of Kuo et al. [6], who claim that it is sufficient that this state has a nonzero
overlap with the corresponding target state.
linked diagrams that have an intermediate state in the model space. The remaining
diagrams are regular. In addition, so-called quasi-singularities can appear – i.e., very
large, but finite, contributions – when an intermediate state is quasi-degeneracy with
the initial state. All singularities and quasi-singularities are eliminated in the ratio
(3.46) – in analogy with the original Gell-Mann–Low theorem, although in the
general case there is a finite remainder, so-called model-space contribution (MSC).
The elimination of these quasi-singularities represents the major advantage of the
procedure using an extended model space. We shall see how this procedure can also
be applied in quantum-electrodynamical calculations in the following sections.
References
1. Brown, G.E., Kuo, T.T.S.: Structure of finite nuclei and the nucleon-nucleon interaction. Nucl.
Phys. A 92, 481–94 (1967)
N.Y. (1971)
(1951)
4. Itzykson, C., Zuber, J.B.: Quantum Field Theory. McGraw-Hill (1980)
5. Jones, R.W., Mohling, F.: Perturbation theory of a many-fermion system. Nucl. Phys. A 151,
420–48 (1970)
6. Kuo, T.T.S., Lee, S.Y., Ratcliff, K.F.: A folded-diagram expansion of the model-space effective
Hamiltonian. Nucl. Phys. A 176, 65–88 (1971)
8. Morita, T.: Perturbation Theory for Degenerate Problems of Many-Fermion Systems. Progr.
Phys. (Japan) 29, 351–69 (1963)
9. Oberlechner, G., Owono-N’-Guema, F., Richert, J.: Perturbation Theory for the Degenerate
Case in the Many-Body Problem. Nouvo Cimento B 68, 23–43 (1970)
10. Schweber, S.S.: An Introduction to Relativistic Quantum Field Theory. Harper and Row, N.Y.
(1961)
11. Tolmachev, V.V.: The field-theoretic form of the perturbation theory for many-electron atoms.
I. Abstract theory. Adv. Chem. Phys. 14, 421, 471 (1969)
Part II
Quantum-Electrodynamics:
One-Photon and Two-Photon Exchange
Chapter 4
S-Matrix
In Part I, we have considered methods for treating atomic many-body systems within
the standard relativistic MBPT and coupled-cluster schemes, in what is known as the
no-virtual-pair approximation (NVPA). In the second part, we shall include effects
beyond this approximation, which we shall refer to as quantum-electrodynamical
(QED) effects. We shall describe three methods for numerical calculations of QED
effects on bound states, developed in the last few decades, which are all based upon
field theory.1
In this chapter, we present the most frequently applied scheme for bound-state
QED calculations, namely the S-matrix formulation. In this chapter, we also come
into contact with the important question of the choice of gauge. The Maxwell
equations are invariant under a certain class of gauge transformations, as shown
in Appendix G. So far, practically all QED calculations have been performed us-
ing what is known as covariant gauges, particularly the Feynman gauge, where the
expressions involved are particularly simple. However, for bound-state problems,
where the Coulomb interaction often dominates, it would be more advantageous to
use the Coulomb gauge. It has been demonstrated by several authors [1,14] that it is
perfectly legitimate to use the Coulomb gauge also in QED calculations and that this
leads to results that are renormalizable and completely equivalent to those obtained
using covariant gauges.
In the next chapter, we consider the Green’s-function method, which is frequently
used in various fields of physics. In Chap. 6, we shall present the recently introduced
covariant-evolution operator method, which will form the basis for the unified ap-
proach we are developing in the following chapters.
1
From now on we shall for simplicity set „ D 1 but maintain the remaining fundamental constants.
In this way, our results will be valid in the relativistic or natural unit system as well as in the Hartree
atomic unit system. They will also be valid in the cgs unit system, as long as we stay consistently to
either the electrostatic or the magnetic version, but they will NOT be valid in the Gaussian system
that is a mixture of the two. With our choice, it will still be possible to perform a meaningful
dimensional analysis (see further Appendix K).

60 4 S-Matrix
4.1 Definition of the S-Matrix: Feynman Diagrams
The scattering matrix or S-matrix was introduced by Wheeler [18] and Werner
Heisenberg in 1930s, particularly for studying the scattering processes between ele-
mentary particles. The formalism is not particularly suited for bound-state problems
but has in the last few decades been applied also to such problems in connection
with QED calculations (see, for instance, the review article by Mohr, Plunien, and
Soff [11] for a modern update).
The S-matrix relates the initial state of a particle or system of particles, ˚i D
˚.t D 1/, before the interaction has taken place, to the final state after the inter-
action is completed, ˚f D ˚.t D C1/,
˚.t D C1/ D S ˚.t D 1/: (4.1)
We know that the time evolution of the state vector in the interaction picture is
governed by the evolution operator (3.6), which leads to the connection
S D U.1; 1/: (4.2)
This is assumed to hold also relativistically (see, for instance, Bjorken and Drell [2]).
With the expansion (3.26), this becomes
Z Z
1 i n
X1
SD dx1 : : : dxn4 T H.x1 / : : : H.xn / e.jt1 jCjt2 j:::jtn j/ :
4
nD0
nŠ c
(4.3)
Here, x is the four-dimensional coordinate vector x D .ct; x/, which explains the
factor of c in the denominator. The S-matrix is – in contrast to the evolution oper-
ator for finite times – Lorentz covariant (see footnote in the Introduction), which
is manifestly demonstrated by its form given here. We shall normally assume that
the perturbation density is given by the interaction between the electrons and the
electromagnetic radiation field (3.17)
H.x/ D O .x/ec˛ A .x/ O .x/: (4.4)
The S-matrix can conveniently be represented by so-called Feynman diagrams.

Feynman has in his famous papers from 1949 [5, 6] developed a set of rules for
evaluating the S-matrix for various elementary processes (see Appendix H), and
this has formed the basis for much of the developments that followed in quantum-
electrodynamics and field theory in general (see, for instance, the books by Mandl
and Shaw [10, Chap. 7] and Peskin and Schroeder [13]). This has also formed the ba-
sis for the diagrammatic representation of many-body perturbation theory (MBPT),
discussed earlier [8].
4.2 Electron Propagator 61
In order to represent the S-matrix by means of Feynman diagrams, this has to be

transformed into normal order, which can be performed by means of Wick’s theo-
rem (see Sect. 2.2). This leads to all possible (zero, single, double . . . ) contractions
between the perturbations H and to diagrams of the type shown in Fig. 2.9. (Details
of this process are found in standard text books, e.g., Fetter and Walecka [4]
or Lindgren-Morrison [8].) Below we shall illustrate this by a few simple
examples.
Even if the S-matrix formulation was initially set up for scattering problems, we
shall here be mainly concerned with applications to bound-state problems. Since
the final time of the scattering process is t D C 1, we cannot directly apply the
Gell-Mann–Low theorem [(3.31) and (3.46)]. Sucher [16] has, however, modified
the Gell-Mann–Low energy formula so that it can be applied also to the S-matrix.
With the S-matrix expanded in a perturbation series
X
SD S .n/ ; (4.5)
n
the energy shift can be expressed
P
i n h˚jS .n/j˚i
E D lim : (4.6)
!0 2 h˚jSj˚i
This energy formula can also be applied to a degenerate multistate model space –
but not in the case of quasi-degeneracy, when there are several distinct energy levels
within the model space. Furthermore, in the S-matrix formulation no information
can be derived for the corresponding change of the state vector or wave function.
For these reasons, the S-matrix formulation is not suited as a basis for a unification
with many-body perturbation theory that is our main concern in this book. We shall
return to this problem in later chapters.
Before we consider some physical processes, we shall define two very impor-
tant concepts, namely the Feynman electron and photon propagators that will be
frequently used in the following.
4.2 Electron Propagator
The contraction between two electron-field operators is defined as the difference

between the time and normal orderings (see Sect. 2.2)
h i h i
O .x1 / O .x2 / D T O .x1 / O .x2 / N O .x1 / O .x2 / : (4.7)
62 4 S-Matrix
! 6 i SF .x1 ; x2 /
Fig. 4.1 Graphical representation of the (bound-state) electron propagator. As before, we shall let
thick vertical lines represent electron propagators in the bound-state representation (Furry picture)
and thin lines in the free-electron representation
Since the vacuum expectation value vanishes for every normal-ordered product, it
follows that the contraction is equal to the vacuum expectation of the time-ordered
product2
˝ ˇ ˇ ˛
O .x1 / O .x2 / D 0ˇT O .x1 / O .x2 / ˇ0
˝ ˇ ˇ ˛
D 0ˇ.t1 t2 / O .x1 / O .x2 / .t2 t1 / O .x2 / O .x1 /ˇ0 (4.8)
considering that the electron fields operators are fermions that anticommute. is
the Heaviside step function [Appendix A, (A.29)].
The Feynman electron propagator is defined (see Fig. 4.1)3
˝ ˇ ˇ ˛
O .x1 / O .x2 / D 0ˇT O .x1 / O .x2 / ˇ0 DW i SF .x1 ; x2 /: (4.9)
Separating the field operators into particle (p) and hole (h) parts, O D O C C O ,
above and below the Fermi surface, respectively, it follows that the expression (4.8)
is identical to
˝ ˇ ˇ ˛
0ˇ.t1 t2 / O C .x1 / O C .x2 / .t2 t1 / O .x2 / O .x1 /ˇ0

"p .t1 t2 / "h .t1 t2 /

D .t1 t2 /
p .x 1 / p .x 2 / e .t2 t1 / h .x 2 / h .x 1 / e
2
In field theory, the vacuum state is normally the “true” vacuum with no (positive-energy) particles
or photons present. In the Dirac picture, this implies that the negative-energy states or “hole” states
of the “Dirac sea” are filled. In many-body applications without reference to field theory, the “vac-
uum” is normally a closed-shell state related to the system (finite or infinite) under study, obtained
for instance by removing the valence or open-shell single-electron states. Single-electron states
present in this vacuum state are referred to as hole states and those not present as virtual or particle
states. In our unified approach, we shall let hole states include negative-energy (antiparticle) states
as well as core states.
3
Note that we define here the electron propagator, using O rather than NO D O ˇ, which is more
frequently used. We find the present definition more convenient in working with the combination
of QED and MBPT.
4.2 Electron Propagator 63
x x x x
!
x x x x
Fig. 4.2 Complex integration of the electron propagator (4.10)
using the time dependence of the field operators in IP in Appendix B (B.28). As will
be demonstrated below,
The electron propagator can be expressed as a complex integral
Z
d! j .x 1 / j .x 2 /
SF .x1 ; x2 / D ei!.t1 t2 / ; (4.10)
2 ! "j C i sgn."j /
where is a small, positive number.

In order to verify the integral formula (4.10), we first consider the case t1 > t2 .
Here, the integrand vanishes exponentially as ! ! i 1, and we then integrate over
the negative half-plane, as illustrated in Fig. 4.2. Here, the poles appear at ! D "j
when this is positive. The contribution to the integral from this pole is 2i times
the pole value – with the minus sign due to the negative (clockwise) integration – or

i j .x 1 / j .x 2 / ei"j .t1 t2 / . Similarly, when t1 < t2 , we integrate over the pos-
itive half plane with the result Ci j .x 1 / j .x 2 / ei"j .t1 t2 / , when "j is negative.
It then follows that iSF , as defined by the integral, is identical to the time-ordered
vacuum expectation (4.8).
The Fourier transform of the electron propagator with respect to time is

j .x 1 / j .x 2 /
SF .!I x 1 ; x 2 / D ; (4.11)
! "j C i sgn."j /
which can be regarded as the coordinate representation (see Appendix C)
hx 1 jj i hj jx 2 i
SF .!I x 1 ; x 2 / D hx 1 jSOF .!/jx 2 i D (4.12)
! "j C i sgn."j /
64 4 S-Matrix
of the operator4
jj i hj j
SOF .!/ D : (4.13)
! "j .1 i/
Using the relation in Appendix (D.49), this can also be expressed
1
SOF .!/ D ; (4.14)
O
! hD .1 i/
where hOD is the Dirac hamiltonian operator (see Appendix D).

The contraction has so far been defined only for t1 ¤ t2 . For the bound-state
problem, it is necessary to consider also equal-time contractions. We then define
the time-ordering for equal time as
1
T .x1 /
.x2 / D .x1 /
.x2 /
.x2 / .x1 / .t1 D t2 /: (4.15)
2
In this case we have
.x1 /
.x2 / D h0j T .x1 /
.x2 / j0i
1
D h0j .x1 / .x2 / .x2 / .x1 / j0i
2
1X 1X
D p .x 1 / p .x 2 / h .x 1 / h .x 2 /
2 p 2 h
1X
D sgn."j / j .x 2 / j .x 1 /;
2
j
where j as before runs over particles and holes. This can still be expressed by the
integral above, as can be seen from the relation
1 "j z i sgn."j /
D C
"j z i sgn."j / ."j z/ C
2 2 ."j z/2 C 2
1
DP C i sgn."j / ı."j z/; (4.16)
"j z
P stands for the principal-value integration, which does not contribute here. There-
fore, the electron-propagator expression (4.10) is valid also for equal times.
4
As stated before, we use the ‘hat’ symbol to emphasize that the quantity is an operator. In cases
where this is obvious, the hat will normally be omitted.
4.3 Photon Propagator 65
4.3 Photon Propagator
The exchange of a single photon between the electrons corresponds to a contraction

(2.29) of two photon-field operators (3.18), defined as in the electron-field
case (4.7),
˝ ˇ ˇ ˛
A .x1 /A .x2 / D 0ˇT A .x1 /A .x2 / ˇ0
˝ ˇ ˇ ˛
D 0ˇ.t1 t2 /A .x1 /A .x2 / C .t2 t1 /A .x2 /A .x1 /ˇ0 ;
(4.17)
(the photon-field operators commute in contrast to the electron-field operators), and
in analogy with the electron propagator we have
The Feynman photon propagator is defined (see Fig. 4.3)
˝ ˇ ˇ ˛
A .x1 /A .x2 / D 0ˇT A .x1 /A .x2 / ˇ0 DW i DF .x1 ; x2 /: (4.18)
We shall also sometimes for convenience use the short-hand notation
DF .x1 ; x2 / D ˛ ˛ DF .x1 ; x2 / (4.19)
using the summation convention.

With A D AC
C A we see that (4.17) is identical to
˝ ˇ ˇ ˛
0 ˇ.t1 t2 / AC C ˇ
.x1 /; A .x2 / C .t2 t1 / A .x2 /; A .x1 / 0 ;
where the square bracket with a comma between the operators represents the com-
mutator (2.12) and noting that the photon-field operators do commute.
Before evaluating the photon propagator, we have to make a choice of gauge
(see Appendix G.2). In so-called covariant gauges, the field components are re-
lated by a Lorentz transformation. Most commonly used of the covariant gauges is
the Feynman gauge, because of its simplicity. In our work with combined QED and
electron correlation, however, it will be necessary to use the noncovariant Coulomb
gauge in order to take advantage of the development in standard many-body pertur-
bation theory. We shall demonstrate that this is quite feasible, although not always
straightforward.
z
-
1, ,2 i DF .x2 ; x1 /
Fig. 4.3 Graphical representation of the photon propagator
66 4 S-Matrix
4.3.1 Feynman Gauge
In the Feynman gauge we have, using the commutation rule in Appendix G (G.11),
1 0
AC
.x1 /; A .x2 / D r r 0 Œakr ; ak0 r 0 ei.kx1 k x2 /
20 !V
1
D g ık;k0 ır;r 0 eik.x1 x2 / :
20 !V
With kx1 D k0 x10 k x 1 and k 0 x2 D k00 x20 k0 x 2 (x0 D ct; ! D ck0 ), this
yields for the vacuum expectation in (4.18)
˝ ˇ ˇ ˛
0ˇT A .x1 /A .x2 / ˇ0
1 h i
D g .t1 t2 / eik.x1 x2 / C .t2 t1 / eik.x1 x2 /
20 ck0 V
X 1
D g eikr 12 .t1 t2 / eik0 .x10 x20 /
k
20 ck0 V
C .t2 t1 / eik0 .x10 x20 / ; (4.20)
with r 12 D x 1 x 2 . The sign of the exponent k r 12 is immaterial.

The expression in the square brackets of (4.20) can as in analogy with (4.10) be
written as a complex integral
Z 1
dq eiq.x10 x20 /
.t1 t2 /eik0 .x10 x20 / C .t2 t1 /eik0 .x10 x20 / D 2ik0 :
1 2 q 2 k02 C i
(4.21)
Thus (see Appendix, Sect. D.2),

Z
˝ ˇ ˇ ˛ 1 X ikr 12 1 dq eiq.x10 x20 /
0ˇT A .x1 /A .x2 / ˇ0 D ig e
1 2 q k C i
0 cV k 2 2
Z Z
1 d3 k ikr 12 1 dq eiq.x10 x20 /
! ig e 2
;
.2/3 1 2 q k C i
c0 2
(4.22)
with k0 D jkj, and the photon propagator (4.18) becomes in the Feynman gauge [c.f
Appendix (F.62)]
Z Z
g d3 k ikr 12 1 dq eiq.x10 x20 /
F
DF .x1 ; x2 / D e
1 2 q k C i
c0 .2/3 2 2
Z Z
g d3 k ikr 12 1 dz eiz.t1 t2 /
D e ; (4.23)
1 2 z c k C i
0 .2/3 2 2 2
where z D cq is the energy parameter. It then follows that

4.3 Photon Propagator 67
The Fourier transform of the photon propagator with respect to x0 D ct

becomes in Feynman gauge
Z
g d3 k eikr 12
F
DF .qI x 1 ; x 2 / D (4.24)
0 .2/3 q 2 k2 C i
and the inverse transformation becomes

Z
dq F
F
DF .x1 ; x2 / D D .qI x 1 ; x 2 / eiq.x10 x20 / : (4.25)
2 F
After integration over the angular part (see Appendix J), this becomes
Z 1
g 2 d sin r12
F
DF .qI x 1 ; x 2 / D 2 ; (4.26)
4 c0 r12 0 q 2 2 C i
where D jkj, and q D k0 is now decoupled from jkj.5 Fourier transforming

(4.25) with respect to space yields
g 1
F
DF .qI k/ D (4.27)
c0 q 2 C i
2
or in covariant notation
g 1
F
DF .k/ D ; (4.28)
c0 k C i
2
where k is the four-dimensional momentum vector, k 2 D k02 k2 .

The Fourier transforms with respect to time are similarly
Z
g d3 k eikr 12
F
DF .zI x 1 ; x 2 / D
0 .2/3 q 2 k2 C i
Z 1
g 2 d sin r12
D 2 ; (4.29)
4 0 r12 0 z2 c 2 2 C i
g 1
F
DF .zI k/ D ; (4.30)
0 z2 c 2 2 C i
5
In some literature jkj is denoted by k, but here we introduce a new notation (), reserving k for
the four-dimensional vector, in order to avoid confusion.
68 4 S-Matrix
which differ from the previous transforms with respect to momentum (4.26) and
(4.27) by a factor of c (see Appendix K.2). z D cq is the energy parameter. The
inverse transformation is here
Z
dz F
F
DF .x1 ; x2 / D D .zI x 1 ; x 2 / eiz.t1 t2 / : (4.31)
2 F
4.3.2 Coulomb Gauge
Above we have found an expression for the photon propagator in the Feynman
gauge, and by means of the formulas for gauge transformation in Appendix G.2,
we can derive the corresponding expressions in other gauges.
In the Coulomb gauge (G.19), the scalar part ( D 00) of the photon propaga-
tor is
1
C
DF00 .k/ D : (4.32)
c0 k2
Transforming back to four-dimensional space yields according to (4.23)
Z Z
1 d3 k eikr 12 dk0 ik0 .x01 x02 /
C
DF00 .x1 ; x2 / D e
c0 .2/3 k2 2
Z 1 Z
1 2 d sin r12 dk0 ik0 .x01 x02 /
D e
4 2 c0 r12 0 2 2
using the relation (J.17). With x0 D ct and z D ck0 , this can be expressed
Z
VC dz iz.t1 t2 /
C
DF00 .x1 ; x2 / D e ; (4.33)
e2c2 2
where VC is the Coulomb interaction (2.109). With the damping factor, the integral
tends to a delta function (A.15)
VC
C
DF00 .x1 ; x2 / ) ı.t1 t2 /; (4.34)
e2c2
but we shall normally use the more explicit expression (4.33).
From the relation (4.33), we find that the Fourier transform with respect to time
becomes
Z 1
1 2 d sin r12 VC
C
DF00 .zI x 1 ; x 2 / D D 2 2: (4.35)
4 c 2 0 r12
2
0 2 e c
4.4 Single-Photon Exchange 69
The vector part of the propagator is according to (G.19) (q D k0 )
1 ki kj
C
DFij .k/ D gij C (4.36)
c0 .k 2 C i/ k2
and transforming back to three-dimensional space yields
Z
1 d3 k eikr 12 ki kj
DFij .qI x 1 ; x 2 / D
C
gij C
c0 .2/3 q 2 k2 C i k2
Z 1
1 2 d sin r12 ki kj
D gij C
c0 0 q 2 2 C i k2
D cDFij
C
.zI x 1 ; x 2 / .z D cq/: (4.37)
4.4 Single-Photon Exchange
We consider now the exchange of a single photon between the electrons, represented
by the Feynman diagram in Fig. 4.4 (left). We start with a general covariant gauge,
such as the Feynman gauge, and consider then the noncovariant Coulomb gauge.
4.4.1 Covariant Gauge
The second-order S-matrix (4.3) is given by

ZZ
1 i 2
S .2/ D d4 x2 d4 x1 T H.x2 / H.x1 / e.jt1 jCjt2 j/ : (4.38)
2 c
With the interaction density (4.4), this becomes
ZZ h i
.ie/2
S .2/ D d4 x2 d4 x1 T
.x/˛ A .x/ .x/ 2

.x/˛ A .x/ .x/ 1
2
.jt1 jCjt2 j/
e ; (4.39)
c d
z D cq
1 - 2 = - + -
a b
Fig. 4.4 The Feynman representation of the exchange of a single, virtual photon between two
electrons. This contains two time-orderings
70 4 S-Matrix
where the contraction between the radiation-field operators yields the photon
propagator, iDF (4.18), or with the short-hand notation (4.19),
ZZ
.ie/2
S .2/ D d4 x2 d4 x1 .x1 / .x2 /
2
iDF .x2 ; x1 / .x2 / .x1 / e.jt1 jCjt2 j/ : (4.40)
Identification with the second-quantized form (see Appendix B)
˝ ˇ ˇ ˛
S .2/ D 1
2 cc cd cd ˇS .2/ˇab cb ca (4.41)
yields a particular matrix element of the S .2/ matrix

ZZ
˝ ˇ ˇ ˛
cd ˇS .2/ˇab D d4 x2 d4 x1 2
c .x1 / d .x2 / ie DF .x2 ; x1 /
b .x2 / a .x1 / e.jt1 jCjt2 j/

Z
dz ˝ ˇˇ ˇ ˛
D cd ie 2 DF .z; x 2 ; x 1 /ˇab
2
ZZ
c 2 dt1 dt2 eit1 ."a "c z/ eit2 ."b "d Cz/ e.jt1 jCjt2 j/
(4.42)
using the Fourier transform (4.31). After performing the time integrations (A.14)
Z
˝ ˇ .2/ˇ ˛ dz ˝ ˇˇ ˇ ˛
cd ˇS ˇab D cd ie 2 c 2 DF .z; x 2 ; x 1 /ˇab
2
2 ."a z "c / 2 ."b C z "d /: (4.43)
We introduce the single-photon interaction

I.x1 ; x2 / D Vsp .x1 ; x2 / D e 2 c 2 ˛1 ˛2 DF .x1 ; x2 / D e 2 c 2 DF .x1 ; x2 /
(4.44)
with the Fourier transform with respect to time
I.zI x 1 ; x 2 / D e 2 c 2 ˛1 ˛2 DF .zI x 1 ; x 2 / D e 2 c 2 DF .zI x 1 ; x 2 /; (4.45)
which has the form of an energy potential. We shall generally express the Fourier
transform of the interaction with respect to time as
Z
2c 2 d f .; x 1 ; x 2 /
I.zI x 1 ; x 2 / D e 2 c 2 ˛1 ˛2 DF .zI x 1 ; x 2 / D ;
z2 c 2 2 C i
(4.46)
where f .; x 1 ; x 2 / is a gauge-dependent function. This transform, as well as the

function f .; x 1 ; x 2 /, has the dimension of energy (or s 1 with our convention
with „ D 1).6
With the notation above the S-matrix element (4.43) becomes
Z
˝ ˇ .2/ˇ ˛ ˝ ˇ ˇ ˛
cd S ab D d z cd ˇ iI.z/ˇab 2 ."a z "c / 2 ."b C z "d /:
ˇ ˇ
(4.47)
In Appendix A.3.2, it is shown that

Z
dz 1
2 .a z/ 2 .b z/ 2
2 z c 2 2 C i
1
D 22 .a b/ 2 ; (4.48)
z c 2 2 C i
where we observe that the infinitesimally small quantity , appearing in the prop-
agators to indicate the position of the poles, is replaced by the adiabatic damping
parameter, , which is a finite quantity (that eventually tends to zero). This gives
˝ ˇ ˇ ˛
cd ˇS .2/ˇab D 22 ."a C "b "c "d / hcd jiI.z/j abi (4.49)
with z D cq D "a "c . This can also be expressed

˝ ˇ .2/ˇ ˛ ˝ ˇ ˇ ˛
cd ˇS ˇab ) 22 .Ein Eout / cd ˇ iI.z/ˇab ; (4.50)
where Ein and Eout are the incoming and outgoing energies, respectively. Using the
Sucher energy formula (4.6) and the relation (A.17)
lim 22 .x/ D ıx;0 ; (4.51)

!0
the corresponding energy shift becomes

˝ ˇ ˇ ˛
E .1/ D ıEin ;Eout cd ˇI.z/ˇab : (4.52)
Assuming that ˚out D ˚in D ˚ is the antisymmetrized state

1
j˚i Djfabgi D p jabi jbai ;
2
the first-order energy shift becomes
E D h˚jI.z/j˚i D habjI.z/jabi hbajI.z/jabi ; (4.53)
6
The constants of the expressions can be conveniently checked by dimensional analysis (see
Appendix K.2).
72 4 S-Matrix
which is consistent with the interpretation of the interaction I.z/ as an equivalent

energy-dependent perturbing potential.
We have seen here that the time integration – in the limit ! 0 – leads to
Energy conservation at each vertex with the propagator energy parameters
treated as energies.
Due to the energy conservation of the scattering process, only diagonal (“on-
the-energy-shell”) matrix elements are obtained from the analysis of the S-matrix.
Therefore, the technique cannot be used for studying quasi-degenerate states by
means of the extended-model-space technique (see Sect. 2.3). Off-diagonal elements
needed for this approach can be evaluated using the covariant-evolution operator
technique, demonstrated in Chap. 6.
4.4.1.1 Feynman Gauge
With the expression (4.29) of the photon propagator in Feynman gauge, the
corresponding interaction (4.45) becomes (z D cq)
Z
e2 2 d sin r12
I .zI x 1 ; x 2 / D 2
F
˛ ˛2 : (4.54)
4 0 r12 1 q 2 2 C i
The corresponding f function in (4.46) then becomes
e2 sin r12 e2 sin r12

f F .; x 1 ; x 2 / D 2
˛1 ˛2 D 2
.1 ˛1 ˛2 / :
4 0 r12 4 0 r12
(4.55)
Evaluating the integral in (4.54), using the result in Appendix J, we obtain
e2
I F .zI x 1 ; x 2 / D .1 ˛1 ˛2 / eijzjr12 =c ; (4.56)
40 r12
which agrees with the semiclassical potential (Appendix F.73).
4.4.2 Noncovariant Coulomb Gauge
In the Coulomb gauge, we separate the interaction into the instantaneous Coulomb
part and the time-dependent transverse (Breit) part,
I C D ICC C ITC : (4.57)
The transverse part of the interaction can be treated in analogy with the covariant
gauges. According to (4.44) we have
j
ITC .x1 ; x2 / D e 2 c 2 ˛1i ˛2 DFij
C
.x1 ; x2 /; (4.58)
which with (4.37) yields

Z
e2 d3 k .˛1 k/ .˛2 k/ eikr 12
ITC .zI x 1 ; x 2 / D ˛1 ˛2 2
0 .2/3 k q 2 k2 C i
Z
e2 d3 k .˛1 r 1 / .˛2 r 2 / eikr 12
D ˛1 ˛2
0 .2/3 k2 q 2 k2 C i
Z
e2 2 d sin r12 .˛1 r 1 / .˛2 r 2 /
D ˛1 ˛2
4 2 0 r12 q 2 2 C i 2
(4.59)
and the corresponding f function becomes (4.46)

e 2 sin.r12 / .˛1 r 1 / .˛2 r 2
fTC .; x 1 ; x 2 / D ˛1 ˛2 : (4.60)
4 2 0 r12 2
Performing the integration in (4.59), using the integrals in Appendix J, yields for
the transverse (Breit) part

e2 eijqjr12 eijqjr12 1
ITC .zI x 1 ; x 2 / D ˛1 ˛2 .˛1 r 1 /.˛2 r 2 / :
40 r12 q 2 r12
(4.61)
This agrees with the semiclassical result obtained in Appendix F.2 (F.54).
The instantaneous Breit interaction is obtained by letting q ) 0, (Fig. 4.5)

e2 1 .˛1 r 12 /.˛1 r 12 /
I Breit
D Inst
B12 D ˛1 ˛2 C ; (4.62)
40 r12 2 2r12
which is the interaction in the Dirac–Coulomb–Breit approximation (NVPA)

(2.112) and agrees with the expression derived in Appendix F (F.55).
c d c d
V12 B12
Fig. 4.5 Instantaneous 1 2 1 2
Coulomb and Breit
interactions between the
a b a b
electrons
74 4 S-Matrix
The (instantaneous) Coulomb part of the interaction becomes, using the relations
(4.33) and (4.35),
Z Z
e2 2 d sin r12 dz iz.t1 t2 /
ICC .x1 ; x2 / D 2 2
e
4 0 r12 2
Z
dz iz.t1 t2 /
D VC e ; (4.63a)
2
Z
e2 2 d sin r12
ICC .zI x 1 ; x 2 / D 2
D VC : (4.63b)
4 0 r12 2
This leads to, using (4.43),
Z
˝ ˇ .2/ˇ ˛ dz ˝ ˇˇ ˇ ˛
cd ˇS ˇab D cd iVCˇab 2 ."a z "c / 2 ."b C z "d /
2
and in analogy with (4.49)

˝ ˇ .2/ˇ ˛ ˝ ˇ ˇ ˛
cd ˇS ˇab D cd ˇ iVCˇab 2 ."a C "b "c "d /:
The Sucher energy formula (4.6) then gives the expected result for the first-order
energy shift
E .1/ D ıEin ;Eout hcd jVCjabi ; (4.64)
where, as before, Ein D "a C "b is the initial and Eout D "c C "d is the final energy.
Again this demonstrates that the interaction (4.45) represents an equivalent interac-
tion potential and that the energy is conserved for the S-matrix.
4.4.3 Single-Particle Potential
Finally, we consider in this subsection the simple case of an interaction between a

single electron and a time-independent external field, A .x/ (Fig. 4.6). Here, the
scattering amplitude becomes from (4.3) with the interaction density (4.4)
Z
S .1/ D ie d4 x O .x/ ˛ A .x/ O .x/ e jt j (4.65)
Fig. 4.6 Diagrammatic

representation of the b
interaction between an
electron and an external field. x
The heavy lines represent a
electronic states in the
bound-interaction picture
4.5 Two-Photon Exchange 75
with A D . =c; A/ according to (F.6) in Appendix F. In analogy with the

previous cases, this yields (dx0 D c dt)
hbjS .1/jai D iec hbj˛ Ajai 2 ."a "b /: (4.66)
We consider a scalar energy potential, V .x/ D e .x/ and A0 D V =ec, which

with is given by and the S-matrix element then becomes
hbjS .1/jai D 2 ."a "b / hbj iV jai : (4.67)
The Sucher energy formula (4.6) then yields the expected result
E .1/ D ı"a ;"b hbjV jai : (4.68)
4.5 Two-Photon Exchange
4.5.1 Two-Photon Ladder
We consider next the exchange of two uncrossed photons in a covariant gauge,

such as the Feynman gauge, illustrated in Fig. 4.7 (left). Again, this is a Feynman
diagram, which contains all relative time orderings of the times involved, still with
the photons uncrossed.
As before, we consider first this problem using a general covariant gauge, such
as the Feynman gauge, and then we shall consider the Coulomb gauge, in particular.
In analogy with the single-photon exchange, the S-matrix becomes
ZZZZ
.ie/4
S .4/ D d4 x1 d4 x2 d4 x3 d4 x4 O .x3 / O .x4 / iDF .x4 ; x3 /
4Š
iSF .x3 ; x1 / iSF .x4 ; x2 / iDF .x2 ; x1 / O .x2 / O .x1 / e.jt1 jCjt2 jCjt3 jCjt4 j/ ;
(4.69)
where DF is defined in (4.19).
c d c d
-z0 VC
3 4 3 4
t !1 z !2 u t !1 z !2 u
1 - 2 1 - 2
a b a b
E0 E0
Fig. 4.7 The Feynman representation of the two-photon exchange. The left diagram represents a
Coulomb and a transverse photon interaction in Coulomb gauge
76 4 S-Matrix
The vertices can here be permuted in 4Š ways, and this leads to pairwise identical
diagrams, related only by a reflection in a vertical plane. The 12 pairs represent
equivalent but distinct terms in the expansion, and by considering only one of them,
we have
ZZZZ
.ie/4
S .4/ D d4 x1 d4 x2 d4 x3 d4 x4 O .x3 / O .x4 / iDF .x4 ; x3 /
2
iSF .x3 ; x1 / iSF .x4 ; x2 / iDF .x2 ; x1 / O .x2 / O .x1 / e.jt1 jCjt2 jCjt3 jCjt4 j/ :
(4.70)
Identifying with the second-quantized expression and performing the time integra-
tions as in the single-photon case (4.43), using the interaction (4.44), the matrix
elements becomes7
ZZ ZZ
dz dz0 d!1 d!2 ˝ ˇˇ
hcd jS .4/jabi D cd .i/I.z0 I x 4 ; x 3 / iSF .!1 I x 3 ; x 1 /
2 2 2 2
ˇ ˛
iSF .!2 I x 4 ; x 2 / .i/I.zI x 2 ; x 1 /ˇab 2 ."a z !1 /
2 ."b C z !2 /2 .!1 z0 "c /2 .!2 C z0 "d /:
Integrations over !1 ; !2 then yield

ZZ
dz dz0 ˝ ˇˇ
hcd jS .4/
jabi D cd .i/I.z0 I x 4 ; x 3 / iSF ."a zI x 3 ; x 1 /
2 2
ˇ ˛
iSF ."b C zI x 4 ; x 2 / .i/I.zI x 2 ; x 1 /ˇab
22 ."a "c z z0 / 22 ."b "d C z C z0 /: (4.71)
(As shown before, the parameter in the electron propagators should here be
replaced by the adiabatic damping parameter .) After integration over z0 , we have
Z
dz ˝ ˇˇ
hcd jS .4/jabi D cd .i/I."a "c zI x 3 ; x 4 /iSF ."a zI x 3 ; x 1 /
2
ˇ ˛
iSF ."b C zI x 4 ; x 2 /.i/I.zI x 2 ; x 1 /ˇab
24 ."a C "b "c "d /: (4.72)
To evaluate this integral is straightforward but rather tedious, and we shall not
perform this here (see, for instance, [9]).
Next, we shall consider the special case, where we have one instantaneous
Coulomb interaction and one transverse-photon interaction (Fig. 4.7, right), using
the Coulomb gauge.
7
We have here an illustration of the general rules for setting up the S-matrix, given in Appendix
H, that there is (1) factor iSF for each electron propagator, (2) a factor iI for each single-photon
exchange and (3) a factor for each vertex.
4.5 Two-Photon Exchange 77
Separating the interaction according to (4.57), we now have

Z
dz ˝ ˇˇ
hcd jS jabi D
.4/
cd .i/ICC .zI x 4 ; x 3 / iSF ."a zI x 3 ; x 1 /
2
ˇ ˛
iSF ."b C zI x 4 ; x 2 / .i/ITC .zI x 2 ; x 1 /ˇab
22 ."a "c z/ 24 ."a C "b "c "d /: (4.73)
Inserting the expressions for the electron propagators (4.10) and the interaction
(4.46), this yields
D ˇ Z
ˇ dz jti htj jui huj
hcd jS .4/jabi D cd ˇ VC
2 "a z "t C it "b C z "u C iu
Z
2 c 2 d fTC ./ ˇˇ E
ˇab 24 ."a C "b "c "d /; (4.74)
z2 c 2 2 C i
where VC is the Coulomb interaction fTC (4.63b) and fTC is given by (4.60). The
products of the propagators can be expressed
1 1
"a z "t C it "b C z "u C iu

1 1 1
D C (4.75)
E0 "t "u "a z "t C it "b C z "u C iu
with E0 D "a C "b . The poles are here at z D "a "t C it , z D "u "b iu and
z D ˙.c i/. Integrating the first term over the negative half plane (z D c i)
and the second term over the positive half plane (z D c C i), yields8
D ˇ jtui htuj ˇ E
ˇ ˇ
hcd jS .4/jabi D i cd ˇ VC VT ˇab 24 .E0 Eout /; (4.76)
E0 " t " u
where
D ˇZ
ˇ
htujVTjabi D tuˇ c d fTC ./
ˇ E
1 1 ˇ
C ˇab (4.77)
"a "t .c i /t "b "u .c i /u
is the transverse-photon potential. The corresponding energy shift becomes in anal-

ogy with the single-photon case (4.53)
D ˇ jtui htuj ˇ E
ˇ ˇ
E D ˚ ˇ VC VT ˇ˚ : (4.78)
E0 " t " u
This holds when particle as well as hole states are involved.
8
This is an illustration of the rule given in Appendix H that there is a factor of i for each “non-
trivial” integration, not involving a factor.
78 4 S-Matrix
In principle, the adiabatic damping has to be switched off simultaneously at all

vertices. If the intermediate state is not degenerate with the initial state, the damping
can be switched off at each vertex independently, which leads to energy conservation
at each vertex, using the orbital energies of the free lines and the energy parame-
ters of the propagators. The degenerate case, which leads to what is referred to as
the reference-state contribution, is more complicated to handle [3, 9], and we shall
not consider that further here. This kind of contribution is easier to evaluate in the
covariant-evolution-operator formalism that we shall consider in Chap. 6.
4.5.2 Two-Photon Cross
For two crossed photons (Fig. 4.8), the S-matrix becomes

ZZ ZZ
dz dz0 d!1 d!2 ˝ ˇˇ
hcd jS .4/
jabi D cd .i/I.z0 I x 4 ; x 3 /iSF .!1 I x 4 ; x 1 /
2 2 2 2
ˇ ˛
iSF .!2 I x 2 ; x 3 / .i/I.zI x 2 ; x 1 /ˇab 2 ."a z !1 /
2 ."b z0 !2 /2 .!1 C z0 "c /2 .!2 C z "d /:
Integrations over !1 ; !2 yield

ZZ
dz dz0 ˝ ˇˇ
.4/
hcd jS jabi D cd .i/I.z0 I x 4 ; x 3 / iSF ."a zI x 4 ; x 1 /
2 2
ˇ ˛
iSF ."b z0 I x 2 ; x 3 / .i/I.zI x 2 ; x 1 /ˇab
22 ."a "c z C z0 / 22 ."b "d C z z0 /: (4.79)
Again, we consider the simpler case with one Coulomb and one transverse in-
teraction (Fig. 4.8, right), using the Coulomb gauge. Then the diagonal element
becomes
D ˇ Z
ˇ dz jti htj jui huj
habjS .4/jabi D ab ˇ VC
2 "a z "t C it "d z "u C iu
Z
2c d fTC ./ ˇˇ E
2
ˇab 24 .0/ (4.80)
z2 c 2 2 C i
c d c d
4 2 2
z u !2
!1 t u !2 3 4
z0 6 !1 t z6
Fig. 4.8 The Feynman 1 3 1
representation of the a b a b
two-photon cross E0 E0
4.6 QED Corrections 79
using the fact that "a C "b D "c C "d . Integration over z leads in analogy with
(4.77) to
D ˇ jtui htuj ˇ E
ˇ ˇ
habjS .4/jabi D i ab ˇ VC VTX ˇab 24 .0/; (4.81)
"a "d "t C "u
where VTX is the potential

D ˇZ
ˇ
htujVTX jabi D tuˇ c d fTC ./
ˇ E
1 1 ˇ
ˇab : (4.82)
"a "t .c i /t "d "u .c i /u
If "t and "u have the same sign, the denominators in the expression (4.81) can be
expressed
1 1
;
"a "t .c i /t "d "u .c i /u
which is in agreement with the evaluation rules for time-ordered diagrams, derived
in Appendix I.
The two-photon ladder and the two-photon cross have been studied in great detail
by means of the S-matrix technique for the ground-state of helium-like systems by
Blundell et al. [3] and by Lindgren et al. [9]. Some numerical results are given in
Chap. 7.
4.6 QED Corrections
In this section, we shall consider how various first-order QED corrections – be-
yond the no-virtual-pair approximation (see Sect. 2.6) – can be evaluated using the
S-matrix formulation. With this formulation, only corrections to the energy can be
evaluated. In Chap. 6, we shall demonstrate a way of including these effects directly
into the wave functions, which makes it possible to incorporate them into the many-
body procedure in a more systematic way. Some QED effects contain singularities
(divergences), which can be handled by means of regularization and renormaliza-
tion, as will be discussed in Chap. 12.
4.6.1 Bound-Electron Self-Energy
When the photon is emitted from and absorbed on the same electron, we have an
effect of the electron self-energy, illustrated in Fig. 4.9. This forms the major part
of the Lamb shift, discovered experimentally by Lamb and Retherford in 1947 [7].
80 4 S-Matrix
b
2 b
t
t ! z 2 1
1 a
a
Fig. 4.9 Diagram representing the first-order bound-electron self-energy. The second diagram
represents the Coulomb part of the self-energy in Coulomb gauge
This was the starting point for the development of modern QED (see the book by
Schweber [15]). The second most important part of the Lamb shift is the vacuum
polarization, to be treated below.
We treat first the self-energy and start with a covariant gauge and then consider
the noncovariant Coulomb gauge.
4.6.1.1 Covariant Gauge
For the electron self-energy (Fig. 4.9), we can set up the expression for the S-matrix
in analogy with the single-photon exchange (4.40),
ZZ
.ie/2
SSE D d4 x2 d4 x1 .x2 / iSF .x2 ; x1 / iDF .x2 ; x1 / .x1 / e.jt1 jCjt2 j/ :
2
(4.83)
Considering the equivalent case with 1 $ 2, the matrix element becomes

ZZ
dz d! D ˇˇ ˇ E
ˇ
hbjSSEjai D b ˇ iSF .!I x 2 ; x 1 / .i/I.zI x 2 ; x 1 / ˇa
2 2
2 ."a z !/ 2 .! C z "b / (4.84)
and after integration over !
hbjSSEjai D 22 ."a "b / hbj i˙."a /jai
with Z
dz
˙."a / D i SF ."a zI x 2 ; x 1 / I.zI x 2 ; x 1 / (4.85)
2
being the self-energy function.
The Sucher energy formula (4.6) yields the corresponding energy shift
ESE D lim i hbjSSEjai D ı"a ;"b hbj ˙."a /jai (4.86)

!0
using the relation between the Dirac delta function and the Kroenecker delta factor
in Appendix (A.17).
With the expressions for the electron propagator (4.10), the bound-state self-
energy becomes
D ˇ Z dz ˇ E
ˇ 1 ˇ
haj˙."a /jai D i at ˇ I.zI x 2 ; x 1 / ˇta
2 "a "t z C it
D ˇ Z dz Z
2c 2 d f ./ ˇˇ E
ˇ 1
D i at ˇ ˇta
2 "a "t z C it z2 c 2 2 C i
(4.87)
using the f function defined in (4.46).

In the Feynman gauge, we have
D ˇ Z dz Z
2c 2 d f F ./ ˇˇ E
ˇ 1
haj˙."a /jai D i at ˇ ˇta ;
2 "a "t z C it z2 c 2 2 C i
(4.88)
where f F is given by (4.55). Performing the z integration yields

D ˇZ c df F ./ ˇ E
ˇ ˇ
haj˙."a /jai D at ˇ ˇta (4.89)
"a "t .c i/t
and
Z
e 2 D ˇˇ ˛1 ˛2 c d sin r12 ˇ E
ˇ
haj˙."a /jai D at ˇ ˇta : (4.90)
4 2 0 r12 "a "t .c i/t
4.6.1.2 Coulomb Gauge
In the Coulomb gauge, the transverse part can be treated in analogy with the co-
variant gauge (4.89)
D ˇZ c dfTC ./ ˇ E
ˇ ˇ
haj˙."a /jaiTrans D at ˇ ˇta (4.91)
"a "t .c i/t
or with (4.60)
Z
e 2 D ˇˇ 1 c d sin r12
haj˙."a /jaiTrans D at ˇ
4 2 0 r12 "a "t .c i/t
h .˛1 r 1 / .˛2 r 2 i ˇˇ E
˛1 ˛2 ˇta : (4.92)
2
82 4 S-Matrix
For the Coulomb part, we insert the expression for ICC in (4.63b) into (4.85),
yielding
Z Z
ie 2 dz 2 d sin r12
˙."a /Coul D 2
S ."
F a zI x 2 ; x 1 /
4 0 r12 2 2
Z
dz
Di SF ."a zI x 2 ; x 1 / VC (4.93)
2
and
D ˇ Z dz ˇ E
ˇ 1 ˇ
haj˙."a /jaiCoul D i at ˇ VC ˇta : (4.94)
2 "a z "t C it
The integral can be evaluated as a principal integral (which vanishes) and half a
pole, yielding the result i sgn."t /=2. The self-energy then becomes
D ˇ ˇ E
1 ˇ ˇ
haj˙."a /jaiCoul D sgn."t / at ˇ VC ˇta : (4.95)
2
The electron self-energy is divergent and has to be renormalized, as will be dis-
cussed in Chap. 12. Some numerical results, using the Feynman gauge, are given in
Chap. 7.
4.6.2 Vertex Correction
The vertex correction, shown in Fig. 4.10, is a correction to the single-potential in-
teraction in Fig. 4.6, and the S-matrix becomes in analogy with the self-energy
ZZ
dz d!
hbjSVCjai D
2 2
D ˇ
ˇ
buˇ iSF .!I x 2 ; x 3 / iec ˛ A .x 3 /
ˇ E
ˇ
iSF .!I x 3 ; x 1 / .i/I.zI x 2 ; x 1 / ˇta
2 ."a z !/ 2 .! C z "b / (4.96)
b
3,
2
u ! A x b
A x 3, z t u
t ! 1 2
1
Fig. 4.10 Diagram a
representing the first-order a
vertex correction
and after integration over !
hbujSVCjtai D 22 ."a "b / hbuj iec ."a ; "a /A .x 3 / jtai ; (4.97)
where
Z
dz
."a ; "a / D i ˛ SF ."a zI x 2 ; x 3 / SF ."a zI x 3 ; x 1 / I.zI x2 ; x 1 /
2
(4.98)
is the first-order vertex-correction function.
4.6.2.1 Covariant Gauge
With the expression for the electron and the photon propagators in a covariant gauge,
we have
Z Z
dz 1 1 2c 2 d f ./
."a ; "a / D i˛
2 "a "u z C iu "a "t z C it z2 c 2 2 C i
Z
c d f ./
D ˛ (4.99)
."a "u c C i/."a "t c C i/
assuming positive intermediate states. The corresponding expressions of the partic-
ular gauge is obtained by inserting the expression for f .k/ in that gauge.
Comparing with the self-energy above, we find for the diagonal part, t D u, what
is known as
The Ward identity (see also Chap. 12).
@
˙."a / D 0 ."a ; "a /: (4.100)
@"a
Also, the vertex correction is singular and has to renormalized, as will be dis-
cussed in Chap. 12.
The transverse part in Coulomb gauge is analogous the expression in the covariant
gauge, using the corresponding f function. For the Coulomb part, we insert the
Coulomb interaction (4.63b) in expression (4.98), yielding
D ˇ Z ˇE
ˇ dz 1 1 ˇ
."a ; "a / D i uˇ ˛ VC ˇt
2 "a "u z C iu "a "t z C it
D ˇ VC ˇˇ E
ˇ
D uˇ sgn."t / ˛ ˇt (4.101)
"t "u
84 4 S-Matrix
provided "t and "u have different sign. If "t D "u this vanishes, which is consistent
with the Ward identity, since the corresponding self-energy contribution is energy
independent.
4.6.3 Vacuum Polarization
The field near the atomic nucleus can give rise to a “polarization effect” in the
form of the creation of electron–positron pairs, an effect referred to as the vacuum
polarization. The first-order effect, illustrated in Fig. 4.11, forms together with the
first-order self-energy (Fig. 4.9) the leading contributions to the Lamb shift.
In order to set up the S-matrix for the leading vacuum polarization (first diagram
in Fig. 4.11), we go back to the relation (4.39) for single-photon exchange
ZZ h i
d4 x2 d4 x1 T
.x/e˛ A .x/ .x/ 2

.x/e˛ A .x/ .x/ 1
.jt1 jCjt2 j/
e
leaving out the factor of 1/2, since we can interchange 1 and 2, and inserting the
contraction between the creation and absorption electron-field operators at vertex 2
to represent the closed orbital loop. Explicitly writing out the spinor components,
we have at this vertex
h i

.x2 /e˛ A .x2 / .x2 / D Tr
.x2 /e˛ A .x2 / .x2 / ;
b
-
1 - 2 ?= - ?+ - + - ?+

a

-
= - + - ?

Fig. 4.11 Diagram representing the first-order vacuum polarization according to (4.107). The
closed loop contains summation over all orbitals (particles and holes). The first and third dia-
grams on the r.h.s. of the first row vanish due to Furry’s theorem (see text). The first diagram in
the second row represents the Uehling part and the final diagrams the Wickmann–Kroll part. The
heavy lines represent the bound-state propagator and the thin lines the free-electron propagator
where “Tr” stands for the trace of the matrix, i.e., the sum of the diagonal elements.
The contraction leads here to iSF .x2 ; x2 /, according to the definition (4.9). We
then have the S-matrix element
ZZ
e 2 d4 x2 d4 x1 i˛1 DF .x2 ; x1 / Tr ˛2 .i/SF .x2 ; x2 / e.jt1 jCjt2 j/ :
With the Fourier transforms SF .!I x 2 ; x 2 / and DF .zI x 2 ; x 1 /, the time depen-
dence is
eit1 ."b "a z/ jt1 j eit2 .!!Cz/ jt2 j
and this leads after time integrations to the S-matrix element
.2/
hbjSSE jai D 2 ."a "b z/ 2 .! ! C z/
ZZ
d! dz D ˇˇ ˇ E
ˇ
e2 b ˇ ˛1 DF .zI x 2 ; x 1 / Tr ˛2 SF .!I x 2 ; x 2 / ˇa ;
2 2
(4.102)
and in the limit ! 0, using the relation (A.28) in Appendix A,
.2/
hbjSSE jai D 2 2 ."a "b /
Z
d! D ˇˇ ˇ E
ˇ
e 2
b ˇ ˛1 DF .0I x 2 ; x 1 / Tr ˛2 SF .!I x 2 ; x 2 / ˇa :
2
(4.103)
According to Sucher’s energy formula (4.6), we have in second order
E D lim i h˚jS .2/j˚i (4.104)

!0
and using the relation (A.17)
E D iı."a ; "b /
Z
d! D ˇˇ ˇ E
ˇ
e2 b ˇ ˛1 DF .0I x 2 ; x 1 / Tr ˛2 SF .!I x 2 ; x 2 / ˇa : (4.105)
2
It can furthermore be shown that only D 0, i.e., ˛ D 1 will contribute here [12].
The vacuum polarization contribution is divergent and has to renormalized, which
in this case turns out to be not too difficult (see below).
The bound-state electron propagator, SF .!/, is in operator form (4.14)
1
SOF .!/ D : (4.106)
O
! hbau .1 i/
86 4 S-Matrix
Expressing the Dirac Hamiltonian for an electron in an external (nuclear) potential

vext as hO bau D hO free Cvext , where hO free is the free-electron Hamiltonian, the propagator
(4.102) can be expanded as
1 1 1 1
D C V
O
z hbau .1 i/ O
z hfree .1 i/ O O
z hfree .1 i/ z hfree .1 i/
1 1 1
C V V C
O O
z hfree .1 i/ z hfree .1 i/ O
zhfree .1 i/
1 1 1
D C V
z hOfree .1 i/ z hO free .1 i/ z hO free .1 i/
1 1 1
C V V ;
z hO free .1 i/ z hO bau .1 i/ z hO free .1 i/
(4.107)
which leads to the expansion is illustrated in Fig. 4.11.

The first and third diagrams on the r.h.s. in the first row in Fig. 4.11 vanish due to
“Furry’s theorem”. According to this theorem, a diagram will vanish if it contains a
free-electron loop with an odd number of vertices [10, Sect. 9.1]. The first diagram
in the second row represents the Uehling part [17], and the second part is the so-
called Wickmann–Kroll [19] part. The Uehling part is divergent, but Uehling was
already in 1934 able to handle this divergence and derive an analytic expression
for the renormalized potential. The Wickmann–Kroll part is finite and has to be
evaluated numerically.
Both the Uehling and the Wickmann–Kroll effects can be expressed in terms of
single-particle potentials that can be added to the external potential, used to gen-
erate the single particle spectrum, and in this way the effects can be automatically
included in the calculations to arbitrary order (see, for instance, Persson et al. [12]).
In Table 4.1, we show the result of some accurate vacuum-polarization calcula-
tions. The diagrams above for the vacuum-polarization and the self-energy can
be compared with the corresponding many-body diagrams, discussed in Sect. 2.4
(Fig. 2.3). In the MBPT case, the internal lines represent core orbitals only, while in
the present case they can represent all orbitals – particle as well as hole states.
Table 4.1 Vacuum- 1:35682

36 Kr Uehling
polarization effects in the Wickmann-Kroll 0.01550
ground state of some
54 Xe Uehling 7:3250
hydrogen-like systems (in
eV) (from Persson et al. [12]) Wickmann-Kroll 0.1695
70 Yb Uehling 23:4016
Wickmann-Kroll 0.8283
92 U Uehling 93:5868
Wickmann-Kroll 4.9863
4.7 Feynman Diagrams for the S-Matrix: Feynman Amplitude 87
-
1 3 4 2
Fig. 4.12 Diagram representing the first-order photon self-energy
4.6.4 Photon Self-Energy
The interaction between the photon and the electron–positron fields can give rise
to another form of vacuum polarization, illustrated in Fig. 4.12. The S-matrix
for this process can be obtained from that of single-photon exchange (4.40) by

replacing ie 2 ˛1 DF .x1 ; x2 / ˛2 by
ZZ
d4 x3 d4 x4 .ie 2 / ˛1 DF .x1 ; x3 / i˘ .x3 ; x4 / .ie 2 / ˛2 DF .x4 ; x2 /;
(4.108)
where
i˘ .x3 ; x4 / D ˛3 O .x3 / O .x3 / O .x4 / O .x4 /˛4
D ˛3 O .x3 / O .x4 / O .x3 / O .x4 /˛4

D Tr ˛3 iSF .x3 ; x4 / iSF .x4 ; x3 / ˛4 (4.109)
is the first-order polarization tensor [10, Eqs. (7.22) and (9.5)]. The contractions
lead here to the trace as in the previous case, and there is also here a minus sign due
to the closed loop.
The photon self-energy is (charge) divergent and requires a renormalization, as
is discussed further in Chap. 12, and after the renormalization there is a finite re-
mainder.
4.7 Feynman Diagrams for the S-Matrix: Feynman Amplitude
4.7.1 Feynman Diagrams
We have in this chapter constructed S-matrix expressions for a number of Feynman

diagrams, and we summarize here the rules that can be deduced for this con-
struction. We also introduce the so-called Feynman amplitude, introduced by
Richard Feynman in his original works on quantum-field theory, which we shall
88 4 S-Matrix
find convenient to use also in other procedures to be discussed later. These rules are
also summarized in Appendix H.
The S-matrix is given by (3.26) and (4.3)
1 Z Z
X i n 1
SD dx14 : : : dxn4 T H.x1 / : : : H.xn / e.jt1 jCjt2 j:::jtn j/
nD0
c nŠ
with the interaction density (4.4)
H.x/ D O .x/ec˛ A .x/ O .x/:
This leads to the following rules: There is

An electron-field creation/absorption operator, O = O , for each outgoing/-
incoming electron orbital.
An electron propagator (4.10) (times imaginary unit) for each internal orbital
line
Z
O .x1 / O .x2 / D i SF .x1 ; x2 / D i d!
SF .!I x 1 ; x 2 / ei!.t1 t2 / :
2
A single-photon interaction [(4.44) and (4.45)] (times negative imaginary unit)

for each single internal photon line
Z
dz
I.x1 ; x2 / D .i/I.zI x 1 ; x 2 / eiz.t1 t2 / ;
2
where the interaction is given by (4.46)

Z Z
2c 2 d f .; x 1 ; x 2 / 2 d f .; x 1 ; x 2 /
I.zI x 1 ; x 2 / D D
z2 c 2 2 C i q 2 2 C i
( D jkj; z D ck0 D cq).

At each vertex a space integration and a time integral 2 .arg/, where the
argument is equal to incoming minus outgoing energy parameters.
A factor of 1 and a trace symbol for each closed orbital loop.
4.7.2 Feynman Amplitude
The Feynman amplitude, M, is

For the S-matrix defined by the relation
hcd jSjabi D 2ı.Ein Eout / hcd jMjabi ; (4.110)
where Ein ; Eout are the incoming and outgoing energies, respectively.
References 89
The first-order energy shift is given by
E D ıEin ;Eout hcd jiMjabi : (4.111)
References
1. Adkins, G.: One-loop renormalization of Coulomb-gauge QED. Phys. Rev. D 27, 1814–20
(1983)
(1964)
3. Blundell, S., Mohr, P.J., Johnson, W.R., Sapirstein, J.: Evaluation of two-photon exchange
graphs for highly charged heliumlike ions. Phys. Rev. A 48, 2615–26 (1993)
N.Y. (1971)
(1949)
7. Lamb, W.W., Retherford, R.C.: Fine structure of the hydrogen atom by microwave method.
Phys. Rev. 72, 241–43 (1947)
9. Lindgren, I., Persson, H., Salomonson, S., Labzowsky, L.: Full QED calculations of two-
photon exchange for heliumlike systems: Analysis in the Coulomb and Feynman gauges. Phys.
Rev. A 51, 1167–1195 (1995)
227–372 (1998)
12. Persson, H., Lindgren, I., Salomonson, S., Sunnergren, P.: Accurate vacuum-polarization cal-
culations. Phys. Rev. A 48, 2772–78 (1993)
(1989)
15. Schweber, S.S.: The men who madt it: Dyson, Feynman, Schwinger and Tomonaga. Princeton
University Press, Princeton (1994)
16. Sucher, J.: S-Matrix Formalism for Level-Shift Calculations. Phys. Rev. 107, 1448–54 (1957)
17. Uehling, E.A.: Polarization Effects in the Positron Theory. Phys. Rev. 48, 55–63 (1935)
18. Wheeler, J.A.: On the Mathematical Description of Light Nuclei by the Method of Resonating
Group Structure. Phys. Rev. 52, 1107–22 (1937)
19. Wichmann, E.H., Kroll, N.M.: Vacuum Polarization in a Strong Coulomb Field. Phys. Rev.
101, 843–59 (1956)
Chapter 5
Green’s Functions
The Green’s function is an important tool with applications in classical as well

as quantum physics (for an introduction, see particularly the book by Fetter and
Walecka [5, Chap. 3], see also the book by Mahan [8]). More recently, it has been
applied also to quantum-electrodynamics by Shabaev et al. [12]. As a background,
we shall first consider the classical Green’s function.
5.1 Classical Green’s Function
The classical Green’s function, G.x; x0 /, can be defined so that it describes the
propagation of a wave from one space-time point x0 D .t0 ; x 0 / to another space-
time point x D .t; x/, known as the Huygens’ principle (see, for instance, the book
by Bjorken and Dell [4, Sect. 6.2])
Z
.x/ D d3 x 0 G.x; x0 / .x0 /: (5.1)
The retarded Green’s function is defined as the part of the functions G.x; x0 / for
which t > t0
GC .x; x0 / D .t t0 / G.x; x0 /; (5.2)
where .t/ is the Heaviside step function (Appendix A.29), which implies
Z
.t t0 / .x/ D d3 x 0 GC .x; x0 / .x0 /: (5.3)
We assume now that the function .x/ satisfies a differential equation of

Schrödinger type

@
i H.x/ .x/ D 0: (5.4)
@t

92 5 Green’s Functions
Operating with the bracket on (5.3) yields

Z
@
iı.t t0 / .x/ D d x 0 i H.x/ GC .x; x0 / .x0 /
3
@t
using the relation (A.31), which implies that

the retarded Green’s function satisfies the differential equation

@
i H.x/ GC .x; x0 / D iı 4 .x x0 /: (5.5)
@t
– A relation often taken as the definition of the (mathematical) Green’s function.

The Green’s function can be used for solving inhomogeneous differential equa-
tions. If we have

@
i H.x/ .x/ D f .x/; (5.6)
@t
then the solution can be expressed
Z
.x/ D dx 0 GC .x 0 ; x/ f .x 0 /: (5.7)
5.2 Field-Theoretical Green’s Function: Closed-Shell Case
5.2.1 Definition of the Field-Theoretical Green’s Function
In the closed-shell case, the field-theoretical single-particle Green’s function can

be defined [5]1
D ˇ h iˇ E
ˇ ˇ
0H ˇT O H .x/ O H .x0 / ˇ 0H
G.x; x0 / D ; (5.8)
h0H j 0H i
where T is the Wick time-ordering operator (2.27) and O H ; O H are the electron-

field operators in the Heisenberg representation (HP) (B.27). The statej0H i is the
“vacuum in the Heisenberg representation,” i.e., the state in the Heisenberg repre-
sentation with no particles or holes. In a “closed-shell state,” the single reference
or model state is identical to the vacuum state (see Sect. 2.3).
1
Different definitions of the field-theoretical Green’s function are used in the literature. The def-
inition used here agrees with that of Itzykson and Zuber [7], while that of Fetter and Walecka [5]
differs by a factor of i.
5.2 Field-Theoretical Green’s Function: Closed-Shell Case 93
The Heisenberg vacuum is time independent and equal to the corresponding vac-
uum state in the interaction picture at t D 0, i.e.,
j0H i D U.0; 1/j0i ; (5.9)
where U.t; t0 / is the evolution operator (3.6) andj0i is the unperturbed vacuum or
the IP vacuum as t ! 1 [c.f. (3.28)].
Using the relation between the electron-field operators in HP and IP (B.25)
O H .x/ D U.0; t/ O .x/ U.t; 0/; (5.10)
we can transform the Green’s function (5.8) to the interaction picture

D ˇ h i ˇ E
ˇ ˇ
0 ˇU.1; 0/T U.0; t / O .x/U.t; 0/U.0; t0 / O .x0 /U.t0 ; 0/ U.0; 1/ˇ 0
G.x; x0 / D :
h0jU.1; 1/j0i
(5.11)
For t > t0 , the numerator becomes
D ˇ ˇ E
ˇ ˇ
0 ˇU.1; t/ O .x/U.t; t0 / O .x0 /U.t0 ; 1/ˇ 0 (5.12a)
and for t < t0

D ˇ ˇ E
ˇ ˇ
0 ˇU.1; t0 / O .x0 /U.t0 ; t/ O .x/U.t; 1/ˇ 0 (5.12b)
using the relation (3.8). From the expansion (3.15), we obtain the identity
X1 Z Z t
.i/ t
U.t; t0 / D dt1 : : : dt T V .t1 / : : : V .t / e.jt1 jCjt2 j /
D0
Š t 0 t 0
X1 Z Z t
.i/n t
D dt1 : : : dtn T V .t1 / : : : V .tn / e.jt1 jCjt2 j /
nD0
nŠ t1 t1
X1 m Z t1 Z t1
.i/
dt1 : : : dtm T V .t1 / : : : V .tm / e.jt1 jCjt2 j / ;
mD0
mŠ t 0 t 0
(5.13)
where we have included the unity as the zeroth-order term in the summation. If we
concentrate on the :th term of the first sum, we have the identity (leaving out the
damping factor)
Z t Z t
1
dt1 : : : dt T V .t1 / : : : V .t /
Š t0 t0
X Z t Z t Z t1 Z t1
1
D dt1 : : : dtn T dt1 : : : dtm T : (5.14)
mCnD
mŠ nŠ t1 t1 t0 t0
We can now apply this identity to the first part of the numerator (5.12a),
U.1; t/ O .x/U.t; t0 /. The interaction times of U.1; t/, O .x/ and U.t; t0 / are
time ordered, and hence the result can be expressed
Z Z 1 h i
1 1
dt1 : : : dt T V .t1 / : : : V .t / O .x/ : (5.15)
Š t0 t0
The same procedure can be applied to the rest of the expression (5.12a) as well as to
the other time ordering (5.12b). With the perturbation (3.16), the numerator of the
single-particle Green’s function (5.11) then becomes [5, Eq. 8.9]
D ˇ h iˇ E X Z Z
1 i n
1
ˇ ˇ
0H ˇT O H .x/ O H .x0 / ˇ 0H D

d4 x1 d4 xn
nD0
nŠ c
D ˇ h iˇ E
ˇ ˇ
0 ˇT O .x/ H.x1 / H.xn / O .x0 / ˇ 0 e.jt1 jCjt2 j / (5.16)
with integrations over all internal times. In transforming the time-ordering to normal
ordering by means of Wick’s theorem, only fully connected terms remain, since the
vacuum expectation of any normal-ordered expression vanishes (see Sect. 4.41).
The denominator in (5.8) becomes, using the relation (5.9),
h0H j 0H i D h0jU.1; 1/j0i D h0jS j0i;
where S is the S-matrix (4.2). Then

the Green’s function can be expressed
D ˇ h iˇ E
ˇ ˇ
0H ˇT O H .x/ O H .x0 / ˇ 0H
G.x; x0 / D : (5.17)
h0jS j0i
We see that this expansion is very similar to that of the S-matrix (4.3), the main
difference being the two additional electron-field operators. Therefore,
the Green’s function can also be expressed as
D ˇ h iˇ E
ˇ ˇ
0 ˇT O .x/U.1; 1/ O .x0 / ˇ 0
G.x; x0 / D ; (5.18)
h0jS j0i
where the time-ordered product is connected to form a one-body operator. This

leads to
Z Z
1 i n
X1
1
G.x; x0 / D d4 x1 d4 xn
h0jSj0i nD0 nŠ c
D ˇ h iˇ E
ˇ ˇ
0 ˇT O .x/ H.x1 / H.xn / O .x0 / ˇ 0 e.jt1 jCjt2 j / : (5.19)
The Green’s function is like the S-matrix Lorentz covariant.

The two-particle Green’s function is defined in an analogous way

˝ ˇ h iˇ ˛
0ˇT O H .x/ O H .x 0 / O H .x00 / O H .x0 / ˇ0
G.x; x 0 I x0 ; x00 / D (5.20)
h0jS j0i
and transforming to the interaction picture leads similarly to
Z Z
1 i n
X1
1
G.x; x 0 I x0 ; x00 / D d4 x1 d4 xn
h0jSj0i nD0 nŠ c
h0jT O .x/ O .x 0 / H.x1 / H.xn / O .x00 / O .x0 / j0i e.jt1 jCjt2 j / (5.21)
and analogously in the general many-particle case. Note that the coordinates are
here four-dimensional space-time coordinates, which implies that the particles have
individual initial and final times. This is in contrast to the quantum-mechanical wave
function or state vector, which has the same time for all particles. We shall discuss
this question further below.
We can transform the time-ordered products above to normal-ordered ones by
means of Wick’s theorem (see Sect. 2.2.3). Since normal-ordered products do not
contribute to the vacuum expectation value, it follows that only fully contracted
contribute to the Green’s function. The contractions between the electron-field op-
erators and the interaction operators lead to electron propagators (SF ) (4.9) on the
in- and outgoing lines as well as all internal lines (see Fig. 5.1).2 This allows time to
run in both directions, and both particle and hole states can be involved.
5.2.2 Single-Photon Exchange
The Green’s function for single-photon exchange in Fig. 5.2 can be constructed in
close analogy to that of the corresponding S-matrix in Sect. 4.4,
ZZ
G.x; x 0 ; x0 ; x00 / D d4 x2 d4 x1 iSF .x; x1 / iSF .x 0 ; x2 /
.i/e 2 DF .x2 ; x1 / iSF .x1 ; x0 / iSF .x2 ; x00 / e.jt1 jCjt2 j/ : (5.22)
x x x0
SF SF SF
SF SF SF
x0 x0 x00
Fig. 5.1 Graphical representation of the one- and two-particle Green’s function. The orbital lines
between dots represent electron propagators
2
In our notations, an orbital line between heavy dots always represents an electron propagator.
Fig. 5.2 Green’s function x x0

for single-photon exchange
SF !3 !4 SF
z
1 - 2
SF !1 !2 SF
x0 x00
E0
With the transforms (4.10) and (4.31), this becomes after integrating over the inter-
nal limes [using the relation (A.17)]
ZZ ZZ
0 0 d!3 d!4
G.x; x 0 ; x0 ; x00 / D eit !3 eit !4 eit0 !1 eit0 !2 d3 x 1 d3 x 2
2 2
ZZ Z
d!1 d!2 dz
iSF .!3 I x; x 1 /
2 2 2
iSF .!4 I x 0 ; x 2 / .i/e 2 DF .zI x 2 ; x 1 / iSF .!1 I x 1 ; x 0 /
iSF .!2 I x 2 ; x 00 /2 .!1 z !3 / 2 .!2 C z !4 /:
(5.23)
In the equal-time approximation, where the particles have the same ini-
tial and final times (t D t 0 and t0 D t00 ), the external time dependence becomes
eit .!3 C!4 / eit0 .!1 C!2 / . In the limit ! 0, we have after z-integration !1 C !2 D
!3 C !4 , and if we consider the diagram as a part of a ladder, this is equal to the
initial energy E0 .
We define the Feynman amplitude for the Green’s function as the function with
the external time dependence removed. This gives
G.x; x 0 ; x0 ; x00 / D Msp .x; x 0 I x 0 ; x 00 / ei.t t0 /E0 (5.24)
and
ZZ ZZ ZZ
d!3 d!4 d!1 d!2
Msp .x; x 0 I x 0 ; x 00 / D d3 x 1 d3 x 2
2 2 2 2
iSF .!4 I x 0 ; x 2 /.i/I.!1 !3 I x 2 ; x 1 / iSF .!1 I x 1 ; x 0 /
iSF .!2 I x 2 ; x 00 /22 .!1 C !2 !3 !4 / (5.25)
using the definition (4.44).
5.2.3 Fourier Transform of the Green’s Function
5.2.3.1 Single-Particle Green’s Function
Assuming the Heisenberg vacuum state j0H i to be normalized, the single-particle

Green’s function (5.8) becomes
˝ ˇ ˇ ˛
G.x; x0 / D 0H ˇT O H .x/ O H .x0 / ˇ0H

˝ ˇ ˇ ˛ ˝ ˇ ˇ ˛
D .t t0 / 0H ˇ O H .x/ O H .x0 /ˇ0H .t0 t/ 0H ˇ O H .x0 / O H .x/ˇ0H :
(5.26)
The retarded part (5.2) is then, using the relation (B.27) in Appendix B,
˝ ˇ ˇ ˛
GC .x; x0 / D 0H ˇ O H .x/ O H .x0 /ˇ0H

˝ ˇ ˇ ˛
D 0H ˇ eiHt O S .x/ eiHt eiHt0 O S .x 0 / eiHt0 ˇ0H : (5.27)
Inserting between the field operators a complete set of positive-energy eigenstates of

the second-quantized Hamiltonian H (2.17), corresponding to the .N C 1/-particle
system
H jni D En jni (5.28)
yields the Lehmann representation
X˝ ˇ ˇ ˛ ˝ ˇ ˇ ˛
GC .x; x0 / D 0H ˇeiHt O S .x/ˇn eiEn .t t0 / nˇ O S .x0 / eiHt ˇ0H (5.29)
n
summed over the intermediate states of the .N C 1/ system. The ground state as
well as the inserted intermediate states are eigenstates of the Hamiltonian H , and
setting the energy of the former to zero, this yields
X˝ ˇ ˇ ˛ ˝ ˇ ˇ ˛
GC .x; x0 / D 0H ˇ O S .x/ˇn eiEn .t t0 / nˇ O S .x0 /ˇ0H : (5.30)
n
Performing a Fourier transform of the Green’s function, including the adiabatic

damping e (see Sect. 3.2), yields ( D t t0 > 0)
Z ˝ ˇ ˇ ˛˝ ˇ ˇ ˛
1 0H ˇ O S .x/ˇn nˇ O S .x 0 /ˇ0H
GC .EI x; x 0 / D d e iE
GC .; x; x 0 / D i
0 E En C i
(5.31)
using
Z 1
i
dt ei˛t e t D : (5.32)
0 ˛ C i
Analogous results are obtained for the advanced part (t < t0 ) of the Green’s func-
tion, corresponding toˇ a holeˇ in the initial system.
˝ ˛
The expression nˇ O S .x/ˇ0H represents a state n .x/ of the .N C 1/ system in
the Schrödinger picture, and an equivalent expression of
the Fourier transform of the Green’s function becomes
X
n .x/ n .x 0 /
GC .EI x; x 0 / D i : (5.33)
n
E En C i
This implies that

the poles of the retarded/advanced single-particle Green’s function represent
the true energies of the vacuum plus/minus one particle, relative to the vacuum
state.
In order to show that the definition (5.8) of the Green’s function is compatible
with the classical definition [(5.1) and (5.5)], we form the reverse transformation
Z
dE iE X n .x/ n .x 0 /
GC .x; x0 / D GC .; x; x 0 / D e i : (5.34)
2 n
E En C i
We then find that

Z
@ dE iE X E En
i H.x/ GC .x; x0 / D e i n .x/ n .x 0 /:
@t 2 n
E En C i
(5.35)
Letting ! 0 and using the closure property (C.27)
X

n .x/ n .x 0 / D ı 3 .x x 0 /
n
and the integral

Z
dE iE
e D ı./ D ı.t t0 /;
2
we confirm that the retarded part of the Green’s function (5.8) satisfies the
relation (5.5)

@
i H.x/ GC .x; x0 / D iı 4 .x x0 /: (5.36)
@t
5.2.3.2 Electron Propagator
We consider now the zeroth-order single-particle Green’s function (5.19)

D ˇ ˇ E
ˇ ˇ
G0 .x; x0 / D 0 ˇT Œ O .x/ O .x0 /ˇ 0 ; (5.37)
where the vacuum and the field operators are expressed in the interaction picture.
Then we find that
the single-particle Green’s function is identical to the Feynman electron prop-
agator (4.9) times the imaginary unit i
G0 .x; x0 / iSF .x; x0 /: (5.38)

The retarded operator can be transformed in analogy with the Lehmann

representation above
X D ˇˇ ˇ
ˇ
ˇ
ˇ
ˇ E
ˇ
0H ˇ O S .x/ˇ n0 eiEn .t t0 / n0 ˇ O S .x0 /ˇ 0H ;
0
G0C .x; x0 / D (5.39)
n
˛
wherejn0 are eigenstates of the zeroth-order Hamiltonian for the .N C 1/-particle
system (B.22)
˛ ˛
H0jn0 D En0jn0
and En0 are the energies relative the vacuum. Performing the time integration yields
the Fourier transform
˛˝
X hxjn0 n0 jx 0 i
G0C .x; x 0 ; E/ D i : (5.40)
n
E En0 C i
The corresponding advanced function becomes

˛˝
X hxjn0 n0 jx 0 i
G0 .x; x 0 ; E/ D i : (5.41)
n
E En0 i
Both of these results can be expressed by means of a complex integral

Z ˛˝
dE hxjn0 n0 jx 0 i iE.t t0 /
G0 .x; x0 / D iSF .x; x0 / D i e ; (5.42)
2 E En0 C in
where n has the same sign as En0 , i.e., positive for particle states and negative for
hole or antiparticle states.
The zeroth-order Green’s function or electron propagator can also be expressed
in operator form as
i
GO 0 .E/ D iSOF .E/ D : (5.43)
E H0 ˙ i
5.2.3.3 Two-Particle Green’s Function in the Equal-Time Approximation
Setting the initial and final times equal for the two particles, t D t 0 and t0 D t00 , the
retarded two-particle Green’s function (5.20) becomes
˝ ˇ ˇ ˛
GC .x; x 0 I x0 ; x00 / D 0H ˇ O H .x/ O H .x 0 / O H .x00 / O H .x0 /ˇ0H
˝ ˇ
D 0H ˇ eiHt O S .x/ O S .x 0 /ei Ht
ˇ ˛
eiHt0 O S .x 00 / O S .x 0 /eiHt0 ˇ0H : (5.44)
We introduce a complete set of two-particle states (5.28), which leads to the

Lehmann representation
X D ˇˇ ˇ E
ˇ
D ˇ
ˇ
ˇ E
ˇ
GC .x; x 0 I x0 ; x00 / D 0H ˇ O S .x/ O S .x 0 /ˇ n eiEn .t t0 / n ˇ O S .x00 / O S .x0 /ˇ 0H
n
(5.45)
with the Fourier transform
˝ ˇ ˇ ˛˝ ˇ ˇ ˛
X 0H ˇ O S .x/ O S .x 0 /ˇn nˇ O .x 00 ; O .x 0 //ˇ0H
0
GC .EI x; x I x 0 ; x 00 /
D S S
n
E En ˙ i
(5.46)
˝ ˇ
with the upper (lower) sign for the retarded (advanced) function. Here, nˇ O S .x 0 /
ˇ ˛
O .x 00 /ˇ0H represents a two-particle state n .x; x 0 / in the Schrödinger picture,
S
which yields the Fourier transform
X 0
/ n .x 0 ; x 00 /
n .x; x
GC .EI x; x 0 I x 0 ; x 00 / D i : (5.47)
n
E En ˙ i
This implies that also in this case the poles of the Green’s function represent the
exact eigenvalues of the system, relative to the vacuum. Note that this holds in the
many-particle case only in the equal-time approximation, where there is only a sin-
gle time coordinate D t t0 .
5.3 Graphical Representation of the Green’s Function
We shall now demonstrate how the expansions of the Green’s-functions [(5.19) and
(5.21)] can be conveniently represented by means of Feynman diagrams [6], dis-
cussed in the previous chapter, and we start with the single-particle case.
5.3.1 Single-Particle Green’s Function
The zeroth-order Green’s function is (with our definition) identical to the Feynman
electron propagator times the imaginary unit (5.38) or equal to the contraction (4.9)
D ˇ h iˇ E
ˇ ˇ
G0 .x; x0 / D 0 ˇT O .x/ O .x0 / ˇ 0 D O .x/ O .x0 /; (5.48)
which we represent graphically as in Fig. 4.1.

x
G0 .x; x0 / D 6j
x0 (5.49)
5.3 Graphical Representation of the Green’s Function 101
This contains both time orderings, i.e., j represents both particle and hole/
antiparticle states.
In next order, the numerator of the Green’s function (5.16) has the form
ZZ
1 D ˇˇ ˇ E
ˇ
2 0ˇ d4 x1 d4 x2 T O .x/ H.x1 /H.x2 / O .x0 / ˇ0 : (5.50)
2c
The photon fields have to be contracted, which leads to a two-particle interaction, in

analogy with the single-photon interaction Vsp (4.44),
H.x1 /H.x2 / D v.x1 ; x2 / (5.51)
with the Fourier transform with respect to time v.zI x 1 ; x 2 /, which we represent
graphically as
1 2
(5.52)
The vacuum expectation (5.50) can then be illustrated by the following picture
1 2
h0jTŒ ]j0i
x0
(5.53)
where the vertical lines represent the electron-field operators. The procedure is now
to transform the time ordering to normal ordering (see Sect. 2.2), which we can do
by means of Wick’s theorem (2.34). This leads to a normal-ordered totally uncon-
tracted and all possible normal-ordered single, doubly, . . . contracted terms. In the
vacuum expectation only fully contracted terms will survive.
We can here distinguish between two cases: either the electron-field operators
are connected to each other and disconnected from the interaction or all parts are
connected to a single piece. The former case leads to the diagrams

- -
6 6 6 6
-
(5.54)
where the disconnected, closed parts represent the closed first-order S-matrix
diagrams

- -
Scl.1/ D 6 6C :- (5.55)
The diagrams in (5.54) can then be expressed G0 Scl.1/ .

Connecting all parts of the expression (5.53) leads to the diagrams
(5.56)
These diagrams are quite analogous to the S-matrix diagrams for vacuum polar-
ization and self-energy, discussed in Sect. 4.6, the only difference being that the
Green’s-function diagrams contain in- and outgoing electron propagators. We note
that all internal lines do represent electron propagators, containing particle as well
as hole states.
We can now see that the disconnected parts of the diagrams (5.54) are eliminated
by the denominator in the definition of the Green’s function (5.8). Therefore, we
can then represent the Green’s function up to first order by connected diagrams
only (Fig. 5.57).
6 + + -
(5.57)
We shall now indicate that this holds also in higher orders.

Next, we consider a Green’s-function term with two two-particle interactions.
h0jT [ ]j0i
(5.58)
We can here distinguish different cases.

We consider first the case where both interactions are disconnected from the
electron-field operators. Leaving out the latter, we then have
h0jT[ j0i
(5.59)
This corresponds to the vacuum expectation of the second-order S-matrix and leads
to connected diagrams
- -

- -
6 - 6 6 ? 6 ? 6 6
- -
(5.60)
and to the disconnected diagrams

- - - - - -
6 66 6 6 6- - -
(5.61)
.2/
We denote these diagrams by Scl D h0jS .2/j0i. In addition, we have the free
electron-field operators, which combine to the zeroth-order Green’s function G0 .
Therefore, we can express the corresponding GF diagrams as G0 Scl.2/ .
Next, we consider the case where one of the interactions in (5.59) is closed by
itself, while the remaining part is connected. This leads to disconnected diagrams,
where the disconnected part is the closed first order (5.55) and the connected part
is identical to the connected first-order diagrams in Fig. 5.56, which we can express
the disconnected diagram as GC.1/ Scl.1/ .
Finally, we have the case where all diagram parts are completely connected,
shown in Fig. 5.3, which we denote by GC.2/ .
Going to third order, we find similarly that we can have G0 D G .0/ combined
with the closed diagrams Scl.3/ , GC.1/ combined with Scl.2/ , GC.2/ combined with Scl.1/
.3/
and finally completely connected GC diagrams. This leads to the sequence
8 .0/
ˆ
ˆ G
ˆ
ˆ
ˆ .1/
< GC C G Scl
.0/ .1/
.2/ .1/ .1/ .2/
GC C GC Scl C G .0/ Scl ;
ˆ
ˆ
ˆ
ˆ CG .3/
C G .2/
S .1/
C G .1/
S .2/
C G .0/ .3/
S
:̂ C cl C cl cl
etc:;
which summarizes to
.G .0/ C GC.1/ C GC.2/ C /.1 C Scl.1/ C Scl.2/ C / D .G0 C GC /.1 C Scl/; (5.62)
-
6
- - -
- 6 6 ? 6
-
Fig. 5.3 Second-order connected diagrams of the one-body Green’s function, assuming a two-
body interaction
where GC represents all connected diagrams of the numerator of the GF expression

(5.21) and Scl represents all closed S diagrams. But the last factor is the vacuum
expectation of the S matrix to all orders
h0jSj0i D 1 C Scl ; (5.63)
which implies that this is canceled by the denominator in the definition (5.8). Hence,
the single-particle Green’s function can in the close-shell case be represented
by completely connected diagrams
" 1
X 1 1 n Z Z
iG.x; x0 / D 2
d4 x1 d4 x2n
nD0
nŠ c
#
D E
0T O .x/ H.x1 ; x2 / H.x2n1 ; x2n / O .x0 / 0 :
conn
(5.64)
This can also be expressed
˝ ˇ ˇ ˛
G.x; x0 / D 0H ˇT Œ O H .x/ O H .x0 /ˇ0H conn : (5.65)
The connectedness of the Green’s function can also be shown in a somewhat differ-
ent way. If we remove the two electron-fields operators and the denominator from
the Green’s function expansion (5.19), then we retrieve the vacuum expectation of
the S-matrix (4.3) h0jS j0i. Therefore, if the field operators are connected to each
other and the interactions among themselves, the result (after including the denom-
inator) is simply the zeroth-order Green’s function iG .0/ . If the field operators are
connected to one of the interactions, they form the connected first-order Green’s
.1/
function iGconn and the remaining interactions again form h0jS j0i. Continuing the
process leads to
G D G .0/ C Gconn
.1/
C Gconn
.2/
C ; (5.66)
which proves that the single-particle Green’s function is entirely connected.
5.3.1.1 One-Body Interaction
We shall now consider the case when we, in addition to the two-body interaction,
have a one-body interaction of potential type
(5.67)
The graphical representation can then be constructed in the same way as before,
and we then find in first order the additional diagrams
66
-
(5.68)
The first diagram is unconnected, and the closed part is a part of h0jS j0i and hence
this diagram is eliminated by the denominator of (5.19), as before. It is not diffi-
cult to show that the single-particle Green’s function is represented by connected
diagrams only, when we have a mixture of one- and two-body interactions. The
additional connected diagrams in second order are shown in Fig. 5.4.
5.3.2 Many-Particle Green’s Function
We now turn to the two-particle Green’s function (5.21). The zeroth-order Green’s
function is in analogy with the one-particle function (5.49) represented by
x x0
G0 .x; x 0 I x0 ; x00 / D 0 0
6 6 D iSF .x; x0 / iSF .x ; x0 /
x0 x00
(5.69)
or a product of two Feynman electron propagators.

-
6
-
-
-
6

6

Fig. 5.4 Additional second-order diagrams of the single-particle Green’s function – in addition to
those in Fig. 5.3 – with a combination of one- and two-body interactions
As mentioned earlier, the (initial and final) times of the two particles in principle
can be different, although we shall in most applications assume that they are equal,
as will be further discussed in the following.
In first order, we have in analogy with the single-particle case (5.53)
h0jT Œ ]j0i
(5.70)
This can lead to disconnected diagrams, composed of the zeroth-order function

(5.69) and the closed first-order diagrams (5.55). Another type of disconnected
diagrams is the combination of zeroth-order single-particle GF and the connected
first-order GF
-
6 6
(5.71)
It should be noted that both parts are here consider as open (not closed).3 Finally,
we can have an open two-particle diagram
3
Generally, a diagram is considered closed if it has no free lines/propagators, like the diagrams in
(5.60) and (5.61), while an open diagram has at least one pair of free lines, like those in Fig. 5.3.
An operator or a function represented by a closed/open diagram is said to be closed/open.
(5.72)
In second order, we can have the zeroth-order two-particle Green’s function,

combined with second-order closed diagrams, Scl.2/ , and connected first-order di-
agrams combined with first-order closed diagrams, Scl.1/ . In addition, we can have
disconnected diagrams with two open first-order single-particle diagrams (5.56).
Continuing this process leads formally to the same result as in the single-particle
case (5.62) – the diagrams with a disconnected closed part are eliminated by the
denominator. Formally, the diagrams can still be disconnected, like (5.71), since
there is a disconnected zeroth-order Green’s function part. We shall refer to such
diagrams as linked in analogy with the situation in MBPT (Sect. 2.4). The result is
then expressed
" Z Z
X1
1 1 n D ˇ h
ˇ
G.x; x 0
I x0 ; x00 / D 2
d 4
x1 d 4
x2n 0 ˇT O .x/ O .x 0 /
nD0
nŠ c
#
i ˇE
ˇ
v.x1 ; x2 / v.x2n1 ; x2n / O .x00 / O .x0 / 0ˇ ; (5.73)
linked
a result that can easily be extended to the general many-particle case.

The two-body interactions used here correspond to two contracted interactions
of the type (4.4). Uncontracted interactions of this kind cannot contribute to the
Green’s function, since this is a vacuum expectation. Therefore, the results above
can in the single-particle case also be expressed
Z Z
1 i n
X1
G.x; x0 / D d x1 d4 xn
4
nD0
nŠ c
D ˇ ˇ E
ˇ ˇ
0 ˇT O .x/ H.x1 / H.xn / O .x0 / ˇ 0conn (5.74)
including even- as well as odd order terms, and similarly in the many-particle case.
This can also be expressed
˝ ˇ ˇ ˛
G.x; x0 / D 0H ˇT Œ O H .x/ O H .x0 /ˇ0H conn

(5.75)
and in the two-particle case
˝ ˇ ˇ ˛
G.x; x 0 I x0 ; x00 / D 0H ˇT Œ O H .x/ O H .x 0 / O H .x00 / O H .x0 /ˇ0H linked : (5.76)
The linked character of the Green’s function can also in the two-particle case be
shown as we did at the end on the single-particle section. If all interactions of the ex-
pansion (5.21) are connected among themselves, they form the vacuum expectation
value of the S-matrix, canceling the denominator, and the electron-field operators
.0/
form the two-body zeroth-order Green’s function G2 . If one pair of field operators
are internally connected, then the remaining part is identical to the single-particle
Green’s function G1 , which has been shown to be connected. The result G1.0/ G1 is
disconnected but since both parts are open, this is linked with the convention we
use. If one pair of field operators are connected to some of the interactions and the
other pair to the remaining ones, the result is G1 G1 , which is also disconnected
but linked. Finally, all field operators can be connected to the interactions, which
leads to the connected two-particle Green’s function G2;conn . The remaining inter-
actions form h0jSj0i, canceling the denominator, and the result becomes G2;conn . In
summary, the two-particle Green’s function becomes
G2 D G2.0/ C G1;conn G1;conn C G2;conn ; (5.77)
which can be disconnected but linked. This argument can easily be generalized,
implying that
the many-particle Green’s function in the closed-shell case is linked.
5.3.3 Self-Energy: Dyson Equation
All diagrams of the one-particle Green’s function can be expressed in the form
ZZ
G.x; x0 / D G.x; x0 / C d4 dx1 d4 dx2 G0 .x; x1 / .i/˙.x1 ; x2 / G0 .x2 ; x0 /;
(5.78)
where ˙.x2 ; x1 / represents the self-energy. This can be represented as shown in
Fig. 5.5, i.e., as the zeroth-order Green’s function plus all self-energy diagrams.
Some of the second-order self-energy diagrams in Figs. 5.3 and 5.4 have the form
of two first-order diagrams, connected by a zeroth-order GF. All diagrams of that
kind can be represented as a sequence of proper self-energy diagrams, ˙ , which
have the property that they cannot be separated into lower-order diagrams by cutting
a single line. This leads to the expansion of the total self-energy shown in Fig. 5.6,
where the crossed box represents the proper self-energy. The single-particle Green’s
function can then be represented as shown in Fig. 5.7, which corresponds to the
Dyson equation for the single-particle Green’s function.
ZZ
G.x; x0 / D G0 .x; x0 / C d4 x1 d4 x2 G0 .x; x2 / .i/˙ .x2 ; x1 / G.x1 ; x0 /:
(5.79)
Fig. 5.5 The single-particle

Green’s function expressed in
terms of the self-energy
6
6
6 = 6+
6
Fig. 5.6 Expansion of the

6
total self-energy in terms of 6 6
proper self-energies. The
crossed box represents the = + 6 +
proper self-energy ˙
6 6
6
Fig. 5.7 Graphical

representation of the Dyson 6
equation for the
6
6 = 6+
single-particle Green’s
function (5.79), using the 6
6
proper self-energy ˙
x x0
x x0 x x0
x2 x20
˙ .x2 ; x20 I x1 ; x10 /
x1 x10
= 6 6 +
G.x1 ; x10 I x0 ; x00 /
x0 x00 x0 x00
x0 x00
Fig. 5.8 Graphical representation of the Dyson equation for the two-particle Green’s function
(5.80). The crossed box represents the proper two-particle self-energy
Similarly, the Dyson equation for two-particle Green’s function becomes

ZZZZ
0
G.x; x I x0 ; x00 / D G0 .x; x 0
I x0 ; x00 / C d4 x1 d4 x2 d4 x10 d4 x20
G0 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 /

G.x1 ; x10 I x0 ; x00 /: (5.80)
This equation is illustrated in Fig. 5.8, where the crossed box represents the proper
two-particle self-energy.
Table 5.1 Electron affinity of Ca atom (in meV)

4p1=2 4p3=2 Reference
Theory 19 13 Salomonson [11]
Theory 22 18 Avgoustoglou [3]
Theory 49 18 Dzuba [3]
Expt’l 24.55 19:73 Petrunin [10]
Expt’l 18.4 Walter [13]
Expt’l 17.5 Nadeau [9]
5.3.4 Numerical Illustration
Here, we shall illustrate the application of the Green’s-function technique for many-
body calculation by the electron affinity of the calcium atom (Table 5.1). The
negative calcium ion is a very delicate system, with a very feeble binding energy, and
it has been quite difficult to determine this quantity experimentally as well as theo-
retically. It is only recently that it has been possible to obtain reasonable agreement.
The calculation of Salomonson et al. is performed by means of the Green’s-
function method, that of Dzuba et al. by many-body perturbation theory and that of
Avgoustoglou by all-order pair-correlation method.
5.4 Field-Theoretical Green’s Function: Open-Shell Case
In this section, we indicate how the Green’s-function concept could be extended

to the open-shell case, when the model states are separated from the vacuum state.
It is recommended that Chap. 6 is first studied, where the treatment is more akin
to the normal situation in MBPT, discussed in Sect. 2.3. We leave out most details
here and refer to the treatment of the covariant evolution operator and the Green’s
operator, which is quite equivalent. In this section, we +in particular look into the
special approach due to Shabaev [12].
5.4.1 Definition of the Open-Shell Green’s Function
In the general open-shell case, singularities of the Green’s function can appear also
for connected diagrams, as in the covariant-evolution operator (see below). If we
consider a sequence of ladder diagrams of single-photon exchange, V , as discussed
in the next chapter (Fig. 6.3), considering only particle states (no-pair), the Feynman
amplitude for the Green’s function is the same as for the covariant evolution operator
(6.20) with no model-space states,
MD1C Q .E0 / V .E0 / C Q .E0 / V .E0 / Q .E0 / V .E0 / C ; (5.81)

5.4 Field-Theoretical Green’s Function: Open-Shell Case 111
where
Q jrsi hrsj
Q .E0 / D D
E0 H0 C i E0 "r "s C i
is the reduced resolvent (2.65) and E0 is the energy parameter (of the Fourier
transform) of the Green’s function. The GF becomes singular, when there is an
intermediate statejrsi of energy E0 . Including the residuals after removing the sin-
gularities (model-space contributions) leads as shown below (6.117) to a shift of the
energy parameter, E0 ! E D E0 C E,
M D 1C Q .E/ V .E/ C Q .E/ V .E/ Q .E/ V .E/ C : (5.82)
This is a Brillouin–Wigner perturbation expansion, and it can be summed to
1 jni hnj
MD D (5.83)
E H C i E En C i
with H D H0 CV .E/ andjni represents the exact eigenstates of the system with the
energy En . This agrees with the Fourier transform of the GF derived above (5.47),
demonstrating that the transform has poles at the exact energies. Consequently, this
holds also in the open-shell case.
The Green’s-function technique yields information only about the energy of the
system. This is in contrast to the Green’s-operator formalism, to be treated in the
next chapter, which can give information also about the wave function or state vector
of the system under study.
5.4.2 Two-Times Green’s Function of Shabaev
The use of the Green’s-function technique for atomic calculations has been further
developed by Shabaev et al. [12] under the name of the “Two-times Green’s func-
tion” (which is equivalent to the equal-time approximation, discussed above). This
technique is also applicable to degenerate and quasi-degenerate energy states, and
we shall outline its principles here.
We return to the extended-model concept, discussed in Sect. 2.3. Given are a
number of eigenstates (target states) of the many-body Hamiltonian
˛
Hj i D E ˛j ˛
i .˛ D 1 d /: (5.84)
The corresponding model states are in intermediate normalization the projections

on the model space
j ˛
0i D Pj ˛
i .˛ D 1 d /: (5.85)
The model states are generally nonorthogonal, and following Shabaev we

˚ˇ ˇ ˛
introduce a “dual set” ˇe0 , defined by
ˇ E ˇ ED ˇ
ˇeˇ ˇ ˇ e˛0 ˇˇ D ı˛;ˇ :
0 jD
˛
ˇ 0 h ˇ 0 (5.86)
Then the standard projection operator becomes

X ˇˇ ˇ E D ˇˇ
ˇ X ˇˇ ED ˇ
P D ˇe0 D ˇ eˇ ˇˇ
0 ˇ ˇ 0 0 (5.87)
ˇ 2D ˇ 2D
with the summation performed over the model space D. We also define an alterna-
tive projection operator as
X ˇˇ ED
ˇˇ
ˇ X ˇˇ ˇ E D ˇ ˇˇ
P 1 D ˇe 0 e0 ˇ:
ˇ
PD ˇ 0 0 ˇ (5.88)
ˇ 2D ˇ
Then
ˇ E ˇ E ˇ E ˇ E
ˇ ˇ ˇ ˇ
P ˇ e˛0 D ˇ ˛
0 and P 1 ˇ ˛
0 D ˇe ˛0 : (5.89)
The Fourier transform of the retarded Green’s function is generally (5.33)

X hxj n i h n jx 0 i
GC .EI x; x 0 / D i ; (5.90)
n
E En C i
where we let x; x 0 represent the space coordinates of all outgoing/incoming parti-

cles. It then follows that
I
dE GC .EI x; x 0 / D 2 hxj n i h n jx 0 i (5.91)
n
and
I
E dE GC .EI x; x 0 / D 2 hxj n i En h n jx 0 i; (5.92)
n
where n is a closed contour, encircled in the positive direction and containing the
single target energy En and no other pole. (This holds if all poles are distinct. In the
case of degeneracy, we can assume that an artificial interaction is introduced that
lifts the degeneracy, an interaction that finally is adiabatically switched off.) This
yields the relation [12, Eq. (44)]
H
E dE GC .EI x; x 0 /
En D H n : (5.93)
n
dE GC .EI x; x 0 /
Following Shabaev, we also introduce a “projected” Green’s function by

D ˇ ED ˇ E
ˇ ˇ ˇˇ
X xˇ 0 0 ˇx 0
gC .EI x; x 0 / D i ; (5.94)
E E ˇ C i
ˇ 2D
which is the coordinate representation (see Appendix C.3) of the corresponding

operator
ˇ ˛˝ ˇˇ ˇ ˛˝ ˇ
X ˇ 0ˇ 0
ˇ X P ˇ ˇ ˇ ˇP
gO C .E/ D i Di (5.95)
E E ˇ C i E E ˇ C i
ˇ 2D ˇ 2D
operating only within the model space.

The effective Hamiltonian (2.54) is defined by
Heffj ˛
0i D E ˛j ˛
0i
and we can then express this operator as

X ˇˇ E D ˇ
ˇ
Heff D ˇ
ˇ
0 E ˇ eˇ0 ˇ D Heff P 1 ; (5.96)
ˇ 2D
where
X ˇ E D ˇ
ˇ ˇ ˇˇ
Heff D Dˇ 0 Eˇ 0 ˇ: (5.97)
ˇ 2D
From the definition (5.94), it follows that

I
1
Heff D E dE g.E/; (5.98)
2
where the integration contour contains the energies all target states. As before, we
assume that the poles are distinct.
Expanding the effective Hamiltonian (5.96) order-by-order leads to
.0/ .1/ .0/ .1/

Heff D Heff C Heff Heff P C : (5.99)
The first-order operator Heff becomes

I
.1/ 1 .0/
Heff D E dE gC .E/; (5.100)
2
where D ˇ ED ˇ E
ˇ ˇ ˇˇ
.0/
X xˇ 0 0 ˇx 0
gC .EI x; x 0 / D i : (5.101)
E E ˇ C i
ˇ 2D
The effective Hamiltonian above is nonhermitian, as in the MBPT treatment in

Sect. 2.3. It can also be given a hermitian form [12], but we shall maintain the non-
hermitian form here, since it makes the formalism simpler and the analogy with the
later treatments more transparent.
5.4.3 Single-Photon Exchange
We shall now apply the two-times Greens function above to the case of single-
photon exchange between the electrons, discussed above (Fig. 5.2). We shall eval-
uate the contribution to the effective Hamiltonian in the general quasi-degenerate
case. In the equal-time approximation, the (first-order) Green’s function (5.25) is
given by
G .1/ .x; x 0 ; x0 ; x00 / D M.1/ 0 0

sp .x; x I x 0 ; x 0 / e
it .!3 C!4 / it0 .!1 C!2 /
e (5.102)
and the first-order Feynman amplitude is given by (5.23)

ZZ ZZ
0 0 d!3 d!4 d!1 d!2
M.1/
sp .x; x I x 0 ; x 0 / D i SF .!3 I x; x 1 /
2 2 2 2
SF .!4 I x 0 ; x 2 / I.!1 !3 I x 2 ; x 1 / SF .!1 I x 1 ; x 0 /
SF .!2 I x 2 ; x 00 / 22 .!1 C !2 !3 !4 /; (5.103)
after integrations over z.

The Fourier transform of the Green’s function with respect to t and t0 is
ZZ
0 dt dt0 iE 0 t iE t0 .1/
G .1/
.E ; E/ D e e G .x; x 0 ; x0 ; x00 /
2 2
D .E 0 !3 !4 / .E!1 !2 / M.1/ 0 0
sp .x; x I x 0 ; x 0 / (5.104)
or
ZZ
0 d!3 d!1
G .1/
.E ; E/ D i SF .!3 I x; x 1 / SF .E 0 !3 I x 0 ; x 2 /
2 2
I.!1 !3 I x 2 ; x 1 / SF .!1 I x 1 ; x 0 / SF .E !1 I x 2 ; x 00 /
22 .E 0 E/; (5.105)
after integrations over !2 ; !4 . With the expression for the electron propagator
(4.12), the matrix element of the Green’s function becomes
ˇZ Z

˝ ˇ .1/ 0 ˇ ˛ ˇ d!3 d!1 1 1
rs ˇG .E ; E/ˇtu D rs ˇˇ
2 2 !3 "r C iu E 0 !3 "s C is
ˇ
1 1 ˇ
I.!1 !3 / ˇ tu
!1 "t C it E !1 "u C iu ˇ
22 .E 0 E/: (5.106)
With jrsi and jtui in the model space, this is the same as the matrix element of
the projected Green’s function (5.94), considering only poles corresponding to the
relevant target states. We define the single-energy Fourier transform by
Z
dE 0
G.E/ D G.E 0 ; E/; (5.107)
2
which yields
ˇZ Z
˝ ˇ .1/ ˇ ˛ ˇ d!3 d!1
rs G .E/ tu D i rs ˇˇ
ˇ ˇ I.!1 !3 /
2 2

1 1 1
C
E "r "s !3 "r C ir E !3 "s C is
ˇ
1 1 1 ˇ
C ˇ tu :
E "t "u !1 "t C it E !1 "u C iu ˇ
(5.108)
We assume here that the initial and final states lie in the model space with all
single-particle states involved being particle states. The relevant poles are here
E D "t C "u D Ein and E D "r C "s D Eout :
The contribution of the first pole is

ˇZ Z
i ˇ d!3 d!1
rs ˇˇ I.!1 !3 /
2i 2 2

Ein 1 1
C
Ein Eout !3 "r C i Ein !3 "s C i
ˇ
1 1 ˇ
C ˇ tu : (5.109)
!1 "t C i Ein !1 "u C i ˇ
The last bracket yields 2i .!1 "t /, and integration over !1 yields
ˇZ ˇ
ˇ Ein 1 1 ˇ
i rs ˇˇ !3 I."t !3 / C ˇ tu :
Ein Eout !3 "r C i Ein !3 "s C i ˇ
(5.110)
Similarly, the other pole yields

ˇZ
ˇ d!1
i rs ˇˇ I.!1 "r /
2
ˇ
Eout 1 1 ˇ
C ˇ tu : (5.111)
Ein Eout !1 "t C i Eout !1 "u C i ˇ
The matrix element of P .1/ is similar with Ein and Eout in the numerator re-
.0/
moved. The matrix element of Heff P .1/ is obtained by multiplying by Eout , and the
first-order contribution then becomes
D ˇ ˇ E ˇZ
ˇ .1/ ˇ ˇ d!3
rs ˇHeff ˇ tu D i rs ˇˇ I."t !3 /
2
ˇ
1 1 ˇ
C ˇ tu : (5.112)
!3 "r C i Ein !3 "s C i ˇ
The photon interaction is in the Feynman gauge given by (4.46)
Z
2c 2 d f F .I x 1 ; x 2 /
I.qI x 1 ; x 2 / D (5.113)
q 2 c 2 2 C i
with f F given by (4.55). This gives

Z
2c 2 d f F .I x 1 ; x 2 /
I."t !3 / D (5.114)
."t !3 /2 c 2 2 C i
with the poles at !3 D "t ˙ .i/. Integrating the relation (5.112) over !3 then
yields
ˇZ ˇ
ˇ 1 1 ˇ
.1/
hrsjHeff jtui ˇ
D rs ˇ cd f F
C ˇ tu :
"t "r . i/ "u "s . i/ ˇ
(5.115)
This agrees with the result obtained with the covariant evolution operator (CEO)
method in the next chapter (6.16). The CEO result is more general, since it is valid
also when the initial and/or final states lie in the complementary Q space, in which
case the result contributes to the wave function or wave operator.
In contrast to the S-matrix formulation, the Green’s-function method is appli-
cable also when the initial and final states have different energies, which makes
it possible to evaluate the effective Hamiltonian in the case of an extended model
space and to handle the quasi-degenerate case.
The two-times Green’s function has in recent years been successfully applied
to numerous highly charged ionic systems by Shabaev, Artemyev et al. of the St.
Petersburg group for calculating two-photon radiative effects, fine structure separa-
tions, and g-factors of hydrogenic systems [1, 2, 14, 15]. Some numerical results are
given in Chap. 7.
References 117
References
1. Artemyev, A.N., Beier, T., Plunien, G., Shabaev, V.M., Soff, G., Yerokhin, V.A.: Vacuum-
polarization screening corrections to the energy levels of heliumlike ions. Phys. Rev. A 62,
022,116.1–8 (2000)
the n=1 and n=2 energy levels in He-like ions. Phys. Rev. A 71, 062,104 (2005)
3. Avgoustoglou, E.N., Beck, D.R.: All-order relativistic many-body calculations for the electron
affinities of Ca , Sr , Ba , and Yb negative ions. Phys. Rev. A 55, 4143–49 (1997)
(1964)
N.Y. (1971)
8. Mahan, G.D.: Many-particle Physics, second edition. Springer Verlag, Heidelberg (1990)
9. Nadeau, M.J., Zhao, X.L., Garvin, M.A., Litherland, A.E.: Ca negative-ion binding energy.
Phys. Rev. A 46, R3588–90 (1992)
10. Petrunin, V.V., Andersen, H.H., Balling, P., Andersen, T.: Structural Properties of the Negative
Calcium Ion: Binding Energies and Fine-Structure Splitting. Phys. Rev. Lett. 76, 744 (1996)
11. Salomonson, S., Warston, H., Lindgren, I.: Many-Body Calculations of the Electron Affinity
for Ca and Sr. Phys. Rev. Lett. 76, 3092–95 (1996)
13. Walter, C., Peterson, J.: Shape resonance in Ca photodetachment and the electron affinity of
Ca(1 S). Phys. Rev. Lett. 68, 2281–84 (1992)
14. Yerokhin, V.A., Artemyev, A.N., Beier, T., Plunien, G., Shabaev, V.M., Soff, G.: Two-electron
self-energy corrections to the 2p1=2 2s transition energy in Li-like ions. Phys. Rev. A 60,
3522–40 (1999)
15. Yerokhin, V.A., Artemyev, A.N., Shabaev, V.M., Sysak, M.M., Zherebtsov, O.M., Soff, G.:
Evaluation of the two-photon exchange graphs for the 2p1=2 2s transition in Li-like Ions.
Phys. Rev. A 64, 032,109.1–15 (2001)
Chapter 6
Covariant Evolution Operator
and Green’s Operator
The third method we shall consider for numerical QED calculation on bound states
is the covariant-evolution-operator (CEO) method, developed during the last decade
by the Gothenburg group [9]. This procedure is based upon the nonrelativistic time-
evolution operator, discussed in Chap. 3, but it is made covariant in order to be
applicable in relativistic calculations. Later, we shall demonstrate that this proce-
dure forms a convenient basis for a covariant relativistic many-body perturbation
procedure, including QED as well as correlational effects, which for two-electron
systems is fully compatible with the Bethe–Salpeter equation. This question will be
the main topic of the rest of the book.
6.1 Definition of the Covariant Evolution Operator
In the standard time-evolution operator (3.6), U.t; t0 /, time is assumed to evolve

only forwards in the positive direction, which implies that t t0 . Internally,
time may run also backwards in the negative direction, which in the Feyn-
man/Stückelberg interpretation [6, 15] represents the propagation of hole or
antiparticle states with negative energy. However, all internal times (ti ) are limited
to the interval ti 2 Œt; t0 .
In the S-matrix (4.2), the initial and final times are t0 D 1 and t D C1, re-
spectively, which implies that the internal integrations do run over all times, making
the concept Lorentz covariant.1
In order to make the time-evolution operator covariant also for finite times, it has
to be modified. This leads to what is referred to as the covariant evolution operator
(CEO), introduced by Lindgren, Salomonson, and coworkers in early 2000s [7–11].
The CEO is, as well as the S-matrix and the Green’s function, field-theoretical
concepts, and the perturbative expansions of these objects are quite similar. The
integrations are performed over all times, and therefore, these objects are normally
represented by Feynman diagrams instead of time-ordered Goldstone diagrams, dis-
cussed earlier (Sect. 2.4).
1
See footnote in the Introduction.

120 6 Covariant Evolution Operator and Green’s Operator
The evolution operator contains generally (quasi)singularities, when it is un-

linked or when an intermediate state lies in the model space. Later in this chapter,
we see how these singularities can be removed for the CEO, leading to what we
refer to as the Green’s operator, since it is quite analogous to the Green’s function,
which is also free of singularities.
As mentioned earlier, the covariant perturbation expansion we shall formulate
here leads for two-particle systems ultimately to the full Bethe–Salpeter (BS) equa-
tion [10]. In principle, the BS equation has separate time variables for the individual
particles, which makes it manifestly covariant. This is also the case for the CEO as
well as for the Green’s function. In most applications, however, times are equal-
ized, so that the objects depend only on a single time, which is known as the
equal-time approximation. This makes the procedure in line with the standard
quantum-mechanical picture, where the wave function has a single time variable,
.t; x 1 ; x 2 /, but the covariance is then partly lost. Here, we shall mainly work
with this approximation in order to be able to combine the procedure with the stan-
dard many-body perturbation theory.
As a first illustration, we consider the single-photon exchange with the stan-
dard evolution operator (Fig. 6.1, left), the Green’s function (middle), and the CEO
(right). In the standard evolution operator only particle states (positive-energy
states) are involved in the lines in and out. Therefore, this operator is NOT Lorentz
covariant. In the Green’s function, there are electron propagators on the free lines,
involving particle as well as hole states (positive- and negative-energy states), and
the internal times can flow in both directions between 1 and C1, which makes
the concept covariant. In the CEO electron propagators are inserted on the free lines
of the standard evolution operator with integration over the space coordinates, mak-
ing it covariant. This implies that we attach a density operator [3] to the free lines
O.x/ D O .x/ O .x/ (6.1)

with integration over the space coordinates. We can also see the CEO as the Green’s
function, with electron-field operators attached to the free ends.
r 6O ˙ O

˙ 6s
t t x x0
- r OC

OC

s - r O˙ O˙ s - r O˙ O˙ s
z Part. z Part. z
Particles 1 - 2 1 - 2 1 - 2
- Holes
- Holes
-
t OC OC u t O˙ O˙ u t O˙ O˙ u
t0 t0 x0 x00
t 6O ˙ O ˙ 6u
Fig. 6.1 Comparison between the standard evolution operator, the Green’s function, and the co-
variant evolution operator for single-photon exchange in the equal-time approximation
6.1 Definition of the Covariant Evolution Operator 121
We generally define the Covariant Evolution Operator (CEO) in the single-

particle case by the one-body operator2
ZZ
˝ ˇ ˇ ˛
1
UCov .t; t0 / D d3 x d3 x 0 O .x/ 0H ˇT Œ O H .x/ O H .x0 /ˇ0H O .x0 /: (6.2)
We use here the same vacuum expectation in the Heisenberg representation as

in the definition of the Green’s function (5.8) with two additional electron-field
operators, O .x/ and O .x0 /, with space integrations over x; x 0 . In contrast to
the Green’s function, we shall assume here that the number of photons does not
need to be conserved. When this number is conserved, the vacuum expectation is
a number and represents the corresponding Green’s function. The space integra-
tion makes the electron-field operators attached to this function, as illustrated in
the figure below (c.f. Fig. 5.1, left).
O
x
SF
SF
x0
O
In analogy with the expression (5.18) for the Green’s function, we can also ex-
press the covariant evolution operator as
ZZ
1
UCov .t; t0 / D d3 x d3 x 0 O.x/ U 1 .1; 1/ O.x0 /; (6.3)
where the density operators are connected to the standard (one-body) evolution
operator or S-matrix.
In expanding the S-matrix [see (5.16)], we obtain
X1 ZZ i n Z Z
1
1
UCov .t; t0 / D d3 x d3 x 0 d4 x1 d4 xn
nD0
nŠ c
h i
T O.x/ H.x1 / H.xn / O.x0 / e.jt1 jCjt2 j / ; (6.4)
1
where the operators are connected to form a one-body operator.
2
An “n-body operator” is an operator with n pairs of creation/absorption operators (for particles),
while an “m-particle” function or operator is an object of m particles outside our vacuum. In prin-
ciple, n can take any value n m, although we shall normally assume that n D m.
Similarly, the two-particle CEO becomes – in analogy with the corresponding

Green’s function (5.21) and Fig. 5.1 (right)
X1 ZZZZ i n
1
2
UCov .t; t 0 I t0 ; t00 /
D d3 x d3 x 0 d3 x 0 d3 x 00
nD0
nŠ c
Z Z h
d4 x1 d4 xn T O.x/ O.x 0 / H.x1 / H.xn /
i
O.x00 / O.x0 / e.jt1 jCjt2 j / : (6.5)
2
O O
x x0
SF SF
SF SF
x0 x0
O O 0
6.2 Single-Photon Exchange

in the Covariant-Evolution-Operator Formalism
We shall now consider the exchange of a single photon between the electrons in
the covariant-evolution-operator formalism. We consider here a general covariant
gauge (see Sect. 4.3), such as the Feynman gauge, and we shall later consider the
noncovariant Coulomb gauge.
We assume here that the initial state is unperturbed and return to the more gen-
eral situation in Chap. 8.
The CEO for the exchange of a single photon (Fig. 6.1, rightmost) is in the gen-
eral case given by
ZZ ZZ ( ZZ
0 O O 1
Usp .t; t I t0 ; t00 / D 3
d xd x 3 0
d 3
x 0 d3 x 00 .x/ 0
.x / d4 x1 d4 x2
2
iSF .x; x1 / iSF .x 0 ; x2 / .i/ e 2 DF .x2 ; x1 / iSF .x1 ; x0 /

)
iSF .x2 ; x00 / e.jt1 jCjt2 j/ O .x00 / O .x0 / (6.6)
6.2 Single-Photon Exchange in the Covariant-Evolution-Operator Formalism 123
[with DF defined in (4.19)] in analogy with the corresponding S-matrix and

Green’s-function expressions. The expression in the curly brackets is the corre-
sponding Green’s function (5.22) (the denominator does not contribute in first
order). The CEO contains additional electron creation/annihilation operators and
integration over the space coordinates at the initial and final times. This makes the
CEO into an operator, while the Green’s function is a function.
When the initial state is unperturbed, it implies with the adiabatic damping that
the initial time (t0 ; t00 : : :) is 1. From the definition of the electron propagator
(4.8), it can be shown that, as t0 ! 1,
Z
d3 x 0 iSF .x; x0 / O .x0 / ) O .x/; (6.7)
when the incoming state is a particle state. Therefore, we can leave out the propaga-
tors on the incoming lines, as illustrated in the first diagram of Fig. 6.2, correspond-
ing to the expression
ZZ ( ZZ
0 3 0 O O 0 1
Usp .t; t I 1/ D 3
d xd x .x/ .x / d4 x1 d4 x2 i SF .x; x1 /
2
)
0
i SF .x ; x2 / .i/e DF .x2 ; x1 / e
2 .jt1 jCjt2 j/ O .x2 / O .x1 /:
(6.8)
Identification with the expression for the second quantization (Appendix B) leads to
the matrix element
ZZ
˝ ˇ ˝
0
hrsjUsp .t; t I 1/jabi D e i.t "r Ct 0 "s /
dt1 dt2 rs ˇx; x 0 i x; x 0 jiSF .x; x1 /
ˇ ˛
iSF .x 0 ; x2 / .i/e 2 DF .x2 ; x1 /jx 1 ; x 2 i hx 1 ; x 2 jˇab
ei.t1 "a Ct2 "b / e.jt1 jCjt2 j/ ;
where we for clarity have indicated the integration variables (see Appendix C.3).
r 6O O
6s r 6 6s r 6 6s
x x0 x x0 x x0
s r s
r !1 !2 s r
z
1 - 2
- -
a O O b a b a b
E0 E0 E0
Fig. 6.2 The evolution-operator diagram for single-photon exchange
0
The external-time dependence is here eit .!1 "r / eit .!2 "r / , which in the
equal-time approximation (t D t 0 ) becomes eit .!1 C!2 "r "s / . Since in the limit
! 0 !1 C !2 D "a C "b D E0 is the initial energy and "r C "s is the final
energy, we have in this limit
hrsjUsp .t; 1/jabi D eit .E0 "r "s / hrsjMspjabi (6.9)

or
Usp .t; 1/jabi D eit .E0 H0 / jrsi hrsjMspjabi ; (6.10)
where Msp represents the Feynman amplitude. This is defined as the operator with-
out the external time dependence, in analogy with the Green’s function (5.23) (see
also Appendix H.2). This yields
ˇZ Z Z
˝ ˛ ˇ d!1 d!2 dz
rsjMsp jab D rs ˇˇ iSF .!1 I x; x 1 / iSF .!2 I x 0 ; x 2 /
2 2 2
.i/I.zI x 2 ; x 1 /2 ."a z !1 /
ˇ
ˇ
2 ."b C z !2 /ˇˇab (6.11)
and after integration over !1 ; !2 in the limit ! 0

D ˇ Z dz
ˇ
hrsjMspjabi D rs ˇ iSF ."a zI x; x 1 / iSF ."b C zI x 0 ; x 2 /
2
ˇ E
ˇ
.i/I.zI x 2 ; x 1 / 22 ."b C "b !1 !2 / ˇab : (6.12)
Inserting the expressions for the propagator (4.10) and the interaction (4.46) then
yields
ˇ Z
ˇ dz 1 1
hrsjMspjabi D rs ˇˇ i
2 "a z "r C ir "b C z "s C is
Z ˇ
2c 2 d f ./ ˇˇ
ab : (6.13)
z2 c 2 2 C i ˇ
With the identity (4.75), this can be expressed
1
hrsjMspjabi D hrsjVspjabi (6.14)
E0 " r " s
or
1
Msp .x; x 0 /jabi D Vspjabi ; (6.15)
E0 H0
6.2 Single-Photon Exchange in the Covariant-Evolution-Operator Formalism 125
where Vsp is the potential for single-photon exchange (4.77),

ˇZ 1
ˇ
hrsjVspjabi D rs ˇˇ c d f ./
0
ˇ
1 1 ˇ
C ˇab :
"a "r .c i /r "b "s .c i /s ˇ
(6.16)
The evolution operator (6.10) then becomes
ˇ ˛ eit .E0 H0 /

Usp .t; 1/ˇab D Vspjabi : (6.17)
E0 H0
The results above hold in any covariant gauge, like the Feynman gauge. They
do hold also for the transverse part in the Coulomb gauge by using the transverse
part of the f function (4.60).
The result (6.16) is identical to the Green’s-function result (5.115), when the final
state,jrsi, lies in the model space. In the CEO case, the final state can also lie in the
complementary Q space, in which case the evolution operator contributes to the
wave function/operator.
The CEO result can be represented by means of two time-ordered Feynman dia-
grams, as shown in Fig. 6.2. We then see that the denominators are given essentially
by the Goldstone rules of standard many-body perturbation theory [5, Sect. 12.4],
i.e., the unperturbed energy minus the energies of the orbital lines cut by a horizon-
tal line, in the present case including also k for cutting the photon line.3
When the initial and final states have the same energy, the potential (6.16) above
becomes
ˇZ 1 ˇ
ˇ 2 d f ./ ˇˇ
hcd jVspjabi D cd ˇˇ ab ; (6.18)
0 q 2 2 C i ˇ
where cq D "a "c D "d "b , which is the energy-conservative S-matrix result
(4.46) and (4.52).
We have seen here that the covariant evolution operator for single-photon
exchange has the time dependence eit .E0 H0 / , which differs from that of the
nonrelativistic evolution operator (3.11). We shall return to this question at the end
of this chapter.
3
It should be observed that a Goldstone diagram is generally distinct from a “time-ordered Feyn-
man diagram,” as is further analyzed in Appendix I.
6.2.1 Single-Photon Ladder
We can now construct the covariant evolution operator for some ladder-type interac-
tions provided no hole states are appearing in the intermediate states (see Chap. 8).
The diagram in Fig. 6.3 represents two reducible single-photon interactions with an
intermediate time (t 0 ) that separates the interactions. The Feynman amplitude is then
obtained by combining two single-photon interactions (6.16) with corresponding
resolvents,
MPE D .E/ Vsp .E/ .E/ Vsp .E/PE : (6.19)
where (see 8.11)
Z
˝ ˇ ˇ ˛ D ˇˇ 1 h 1 1 iˇ E
ˇ
rs ˇVsp .E/ˇtu D rs ˇ c d f ./ C ˇtu
0 E "t "s c E "r "u c
PE is the projection operator of the part of the model space with energy E, and .E/
is the resolvent (2.64).
This procedure can be repeated to a general single-photon ladder
MLadd PE D .E/ Vsp .E/ .E/ Vsp .E/ .E/ Vsp .E/PE : (6.20)
The corresponding part of the evolution operator is according to (6.10)
U0 .t; 1/Ladd PE D eit .EH0 / MLadd PE ; (6.21)
where subscript “0” is used to indicate that there are no intermediate model-space
states (see further below). This evolution operator can be singular due to inter-
mediate and/or final model-space states, which can be eliminated by means of
counterterms, leading to “folds” (model-space contributions, MSC), as we shall
demonstrate below.
It should be observed that
In the equal-time approximation the interactions and the resolvents as well as
the time factor of the ladder without folds all depend on the energy of the ini-
tial, unperturbed state.
The folds will affect the time dependence, as will be discussed in Sect. 6.9. In Part
III, we shall treat the ladder in the presence of virtual pairs and higher-order inter-
actions and see how the procedure can be fitted into a many-body procedure.
6 6
t
6- 6
6 6
t0
6 6
Fig. 6.3 Feynman diagram -
representing second-order
6 6
ladder diagram (6.19) E
6.3 Multiphoton Exchange 127
6.3 Multiphoton Exchange
6.3.1 General
We shall now briefly consider the general case of multiphoton exchange. We can
describe this by means of a general many-body potential, which we can separate
into one-, two-, . . . body parts,
V D V1 C V2 C V3 C (6.22)
and which contains all irreducible interactions.4 By iterating such a potential, all re-
ducible interactions will be generated. In Figs. 6.4 and 6.6, we illustrate the one- and
two-body parts of this potential, including radiative effects – vacuum polarization,
self-energy, vertex correction (see Sect. 2.6) – which, of course, have to be properly
renormalized (see Chap. 12).
The one-body potential contains an effective-potential interaction (Fig. 6.5) in
analogy to that in ordinary MBPT (2.73). In the effective potential here, however,
the internal lines can be hole lines as well as particle lines. This implies that the
second diagram on the r.h.s. in Fig. 6.5 contains the direct Hartree–Fock potential as
well as the radiative effect of vacuum polarization and the last diagram the exchange
V1
x = + + +
Fig. 6.4 Graphical representation of the one-body part of the effective potential (6.22), containing
the one-body potential in Fig. 6.5 as well as irreducible one-body potential diagrams, including
radiative effects
= + ?+
Fig. 6.5 Graphical representation of the “extended” effective potential interaction. This is analo-
gous to the effective potential in Fig. 2.3, but the internal lines represent here all orbitals (particles
as well as holes). This implies that the last two diagrams include the (renormalized) vacuum polar-
ization and self-energy
4
Concerning the definition of the concepts “reducible” and “irreducible,” see Sect. 2.6.
V2
=
-

Fig. 6.6 The two-body part of the effective potential (6.22) contains all irreducible two-body
potential diagrams
part of the HF potential as well as the and electron self-energy (both radiative effects
properly renormalized). All heavy lines here represent orbitals in the external (nu-
clear) potential, which implies that the vacuum polarization contains the Uehling
potential [16] (see Sect. 4.6) as well as the Wickmann–Kroll [17] correction, dis-
cussed earlier in Sect. 4.6.
6.3.2 Irreducible Two-Photon Exchange
We consider next the general two-photon exchange, illustrated in Fig. 6.7, still as-
suming the equal-time approximation and unperturbed initial state.
6.3.2.1 Uncrossing Photons
Generalizing the result for single-photon exchange (6.6), we find that the kernel of
the first (ladder)diagram becomes
iSF .x; x3 / iSF .x 0 ; x4 / .i/e 2 DF .x4 ; x3 / iSF .x3 ; x1 /

iSF .x4 ; x2 / .i/e 2 DF .x2 ; x1 /: (6.23)
This leads to the Feynman amplitude in analogy with (6.11)

ZZZZ ZZ
d!1 d!2 d!3 d!4 dz dz0
Msp .x; x 0 I x 0 ; x 00 / D iSF .!3 I x; x 3 /
2 2 2 2 2 2
iSF .!4 I x 0 ; x 4 / .i/I.z0 I x 4 ; x 3 / iSF .!1 I x 3 ; x 1 /
iSF .!2 I x 4 ; x 2 / .i/I.zI x 2 ; x 1 / 2 ."a !1 z/
6.3 Multiphoton Exchange 129
r 6 6s r 6 6s
x x0 x x0
r !3 !4 s r !3 !4 s
0
z
- 4 2
3 4
z0
t !1 !2 u t !1 !2 u
z z6
1 - 2
1 3
a b a b
E E
Fig. 6.7 Covariant-evolution-operator diagrams for two-photon ladder and “cross”
2 ."b !2 C z/ 2 .!1 z0 !3 /

2 .!2 C z0 !4 /: (6.24)
Integration over !1 ; !2 leads to

ZZ ZZ
d!3 d!4 dz dz0
Msp .x; x 0 / D iSF .!3 I x; x 3 / iSF .!4 I x 0 ; x 4 /
2 2 2 2
.i/I.z0 I x 4 ; x 3 / iSF ."a zI x 3 ; x 1 / iSF ."b C zI x 4 ; x 2 /
.i/I.zI x 2 ; x 1 / 22 ."a z z0 !3 /
22 ."b C z C z0 !4 / (6.25)
and over !3 ; !4
ZZ
dz dz0
Msp .x; x 0 / D iSF ."a z z0 I x; x 3 / iSF ."b C z C z0 I x 0 ; x 4 /
2 2
.i/I.z0 I x 4 ; x 3 / iSF ."a zI x 3 ; x 1 /
iSF ."b C zI x 4 ; x 2 / .i/I.zI x 2 ; x 1 /: (6.26)
Integration over z0 leads to the denominators

1 1 1
C
E "r "s "a "r z .c 0 i /r "b "s C z .c 0 i /s
and the remaining part of the integrand is

1 1 1 1
C :
E "t "u "a "t z C it "b C z C iu z c 2 C i
2 2
6.3.2.2 Crossing Photons
For the crossed-photon exchange in Fig. 6.7 (right), the corresponding result is
ZZZZ ZZ
0 d!1 d!2 d!3 d!4 dz dz0
Msp .x; x I x 0 ; x 00 / D iSF .!3 I x; x 4 /
2 2 2 2 2 2
iSF .!4 I x 0 ; x 2 / .i/I.z0 I x 4 ; x 3 / iSF .!1 I x 4 ; x 1 /
iSF .!2 I x 2 ; x 3 / .i/I.zI x 2 ; x 1 / 2 ."a !1 z/
2 ."b !2 z0 / 2 .!1 z0 !3 /
2 .!2 C z !4 /: (6.27)
Integration over the omegas yields

ZZ
dz dz0
Msp .x; x 0 / D iSF ."a C z0 zI x; x 4 / iSF ."b C z z0 I x 0 ; x 2 /
2 2
.i/I.z0 I x 4 ; x 3 / iSF ."a zI x 4 ; x 1 /
iSF ."b zI x 2 ; x 3 / .i/I.zI x 2 ; x 1 /: (6.28)
Integration over z0 leads to the denominators

1 1 1
C
E "r "s "a "r z .c 0 i /r "b "s C z .c 0 i /s
and the remaining part of the integrand is
1 1 1
:
"a "t z C it "b "u z C iu z c 2 C i
2 2
To evaluate the integrals above is quite complicated, but they are considered in
detail in [9, Appendix B] and in the thesis of Björn Åsén [2]. The two-photon ef-
fects have been evaluated for helium-like ions, and some results are shown in the
following chapter.
6.3.3 Potential with Radiative Parts
Two-photon potentials with self-energy and vacuum-polarization insertions can also

be evaluated in the covariant-evolution-operator formalism, as discussed in [9]. We
shall not consider this any further here, but return to these effects in connection with
the MBPT–QED procedure in Chap. 8.
6.4 Relativistic Form of the Gell-Mann–Low Theorem 131
6.4 Relativistic Form of the Gell-Mann–Low Theorem
We have in Chap. 3 considered the nonrelativistic form of the Gell-Mann–Low the-

orem, and we shall now extend this to the relativistic formalism. This theorem plays
a fundamental role in the formalism we shall develop here.
We shall start with
a conjecture that the time evolution of the relativistic state vector is governed by
the CEO in the equal-time approximation (in the interaction picture), in analogy
with the situation in the nonrelativistic case (3.6) (c.f. [4, Sect. 6.4]),
ˇ ˛ ˛ ˇ ˛
ˇ .t/ D UCov .t; t0 /ˇ˛ .t0 / : (6.29)
Rel Rel
We shall later demonstrate that this conjecture is consistent with the standard
quantum-mechanical picture (6.120) [see also (9.13)]. It should be noted that the
evolution operator does not generally preserve the (intermediate) normalization.
It can now be shown as in the nonrelativistic case in Sect. 3.3 that the conjecture
above leads to a
relativistic form of the Gell-Mann–Low theorem for a general quasi-dege-
nerate model space
ˇ ˛˛
ˇ ˛ ˛ ˇ ˛ UCov .0; 1/ˇ˚Rel
ˇ .0/ D ˇ ˛
D lim ˝ ˛; (6.30)
0 Rel j UCov .0; 1/ j˚Rel
Rel Rel
!0 ˛ ˛
˛
˛
which is quite analogous to the nonrelativistic theorem (3.46). Here,j˚Rel is, as
before, the parent state (3.32), i.e., the limit of the corresponding target state, as
the perturbation is adiabatically turned off,
ˇ ˛˛ ˇ ˛
ˇ˚ D C ˛ lim ˇ˛ .t/ ; (6.31)
Rel t !1
Rel
˛ ˛
(C ˛ is a normalization constant) andj 0 Rel ˛
D P j Rel
˛
is the (normalized) model
state. ˛
The state vector j Rel
˛
satisfies a relativistic eigenvalue equation, analogous to
the nonrelativistic (Schrödinger-like) Gell-Mann–Low equation (3.35),
ˇ ˛ ˇ ˛
H0 C VF ˇ ˛
Rel
D E ˛ˇ ˛
Rel
; (6.32)
where VF is the perturbation, used in generating the evolution operator (6.5).

In proving the relativistic form of the GML theorem, we observe that the
covariant evolution operator differs from the corresponding nonrelativistic operator
particularly by the replacement of the electron-field operators by the corresponding
density operators (6.1). It then follows that the commutator of H0 with the covari-
ant operator is the same as with the nonrelativistic operator, which implies that the
proof in Sect. 3.3 can be used also in the covariant case.
A condition for the GML theorem to hold is as in the nonrelativistic case that the
perturbation is time-independent in the Schrödinger picture (apart from damping),
which is the case for the perturbation we shall use here (see further below).
6.5 Field-Theoretical Many-Body Hamiltonian
In the unified MBPT–QED procedure, we shall apply the Coulomb gauge in order to
be able to utilize the developments of the MBPT procedure. In this gauge, we sepa-
rate the interaction between the electrons in the instantaneous Coulomb interaction
and the transverse interaction, with the Coulomb part being (2.109)
X
N
e2
VC D : (6.33)
40 rij
i <j
The exchange of a virtual transverse photon is represented by TWO perturbations

of the one-body perturbation
Z
vT .t/ D d3 x H.t; x/; (6.34)
where the perturbation density given by (4.4)
H.x/ D H.t; x/ D O .x/ ec˛ A .x/ O .x/ (6.35)
with A being the quantized, transverse radiation field (see Appendix F.2). The total
perturbation is then
VF D VC C v T : (6.36)
The perturbation (6.35) represents the emission/absorption of a photon. There-

fore, with this perturbation the GML equation works in a photonic Fock space,5
where the number of photons is not preserved. (The perturbation above is not time
independent in the Schrödinger picture as required by the GML relation, but it can
be transformed into equivalent, time-independent interactions, as will be demon-
strated in Sect. 10.1).
The model many-body Hamiltonian we shall apply is primarily a sum of Dirac
single-electron Hamiltonians in an external (nuclear) field (Furry picture) (2.108)
hD D c˛ pO C ˇmc 2 C vext : (6.37)
5
Also, the Fock space is a form of Hilbert space, and therefore we shall refer to the Hilbert space
with a constant number of photons as the restricted (Hilbert) space and the space with a variable
number of photons as the (extended) photonic Fock space (see Appendix A.2).
6.5 Field-Theoretical Many-Body Hamiltonian 133
[As before, we may include an optional potential, u, in the model Hamiltonian (2.49)
– and subtract the same quantity in the perturbation – in order to improve the con-
vergence rate for many-electron systems.]
However, since the number of photons is no longer constant in the space we work
in, we have to include in the model Hamiltonian also the radiation field, HRad [see
Appendix (G.12) and (B.20)], yielding
X
H0 D hD C HRad : (6.38)
The full field-theoretical many-body Hamiltonian will then be
H D H0 C VF D H0 C VC C vT (6.39)
sometimes also referred to as the many-body Dirac Hamiltonian. This leads with
the GML relation (6.32) to the corresponding Fock-space many-body equation6
H DE : (6.40)
In comparing our many-body Dirac Hamiltonian with the Coulomb–Dirac–Breit

Hamiltonian of standard MBPT (2.113), we see that we have included the radiation
field, HRad , and replaced the instantaneous Breit interaction with the transverse field
interaction, vT , in addition to removing the projection operators.
Using second quantization (see Appendices B and E),
the field-theoretical many-body Hamiltonian (6.39) becomes
Z
H D d3 x O .x/ c ˛ pO C ˇmc 2 C vext .x/ ec˛ A .x/ O .x/
ZZ
1 e2
CHRad C d3 x 1 d3 x 2 O .x1 / O .x2 / O .x2 / O .x1 /;
2 40 r12
(6.41)
where vext .x/ is the external (nuclear) field of the electrons (Furry picture).
We have here assumed that the Coulomb gauge is employed, and therefore the
operator A .x/ represents only the transverse part of the radiation field. (As
mentioned previously, it is quite possible to use the Coulomb gauge in QED
calculation, as demonstrated by Adkins [1], Rosenberg [13], and others.)
6
This equation is not completely covariant, because it has a single time, in accordance with the es-
tablished quantum-mechanical picture. This is the equal-time approximation, mentioned above and
further discussed later. In addition, a complete covariant treatment would require that also the inter-
action between the electrons and the nucleus is treated in a covariant way by means of the exchange
of virtual photons (see, for instance, [14]).
By treating the Coulomb and the transverse photon interactions separately,

a formal departure is made from a fully covariant treatment. However, this
procedure is, when performed properly, in practice equivalent to the use of a
covariant gauge.
We define the wave-operator in analogy with the nonrelativistic case (2.37)7
˛ ˛
j i D ˝j 0i .˛ D 1 d / (6.42)
but now acting in the extended photonic Fock space.

The effective Hamiltonian has the same definition as before (2.38), which leads to
Heff D PH ˝P (6.43)
and the effective interaction is defined by
Veff D Heff PH0 P D P .H H0 /˝P (6.44)
or using the Hamiltonian (6.39)
Veff D P VF ˝P D P VC C vT ˝P: (6.45)
This is a Fock-space relation, and the corresponding relation in the restricted space
without uncontracted photons is given by (6.123).
By solving the many-body equation (6.40) iteratively, all possible perturbations
will be produced. This is the basic principle of the covariant relativistic many-body
perturbation procedure we shall develop in this book. How this can be accomplished
will be discussed in the following. First, we shall treat the simple case of single-
photon exchange.
6.6 Green’s Operator
6.6.1 Definition
The vacuum expectation used to define the Green’s function (5.8) contains singu-
larities in the form of unlinked diagrams, where the disconnected parts represent
the vacuum expectation of the S-matrix. This is a number, and it then follows that
the singularities could be eliminated by dividing by this number. For the covariant
7
In the following, we shall leave out the subscript “Rel.”
6.6 Green’s Operator 135
evolution operator (CEO) (6.2), the situation is more complex, since this in an
operator, and the disconnected parts will also in general be operators. Therefore,
we shall here proceed in a somewhat different manner.
As mentioned,
we shall refer to the regular part of the CEO as the Green’s operator – in the
single-particle case denoted G.t; t0 / – due to its great similarity with the Green’s
function. We define the single-particle Green’s operator by the relation8
U.t; t0 /P D G.t; t0 / P U.0; t0 /P; (6.46)
where P is the projection operator for the model space, and analogously in the
many-particle case. Below we shall demonstrate that the Green’s operator is
regular.
The definition of the Green’s operator contains the important concept of a heavy
dot, which is defined in the following way.
If the operators are disconnected, there is no difference between the dot product
and an ordinary (normal-ordered) product. If the operators are connected to a dia-
gram of ladder type in Fig. 6.3 (6.20), then we have seen that all interactions in an
ordinary product depend on the energy of the initial state. We now introduce the con-
vention that in a dot product the operators do not operate beyond the heavy dot.9
The definitions above imply that the interactions and the resolvents to the left
of the dot depend on the energy of the unperturbed state at the position of the dot.
If we operate to the right on the part of the model space PE of energy E and the
intermediate model-space state lies in the part PE 0 of energy E 0 , we can express the
two kinds of products as
8
< APE 0 BPE D A.E/ PE 0 B.E/PE
ˆ
(6.47)
:̂ 0
A P BPE D A.E / P B.E/PE
E0 E0
with the energy parameter of A equal to E in the first case and to E 0 in the second
case. By the hooks we indicate that the operators must be connected by at least one
contraction. We shall soon see the implication of this definition.
8
The Green’s operator is closely related – but not quite identical – to the reduced covariant evolu-
tion operator, previously introduced by the Gothenburg group [9].
9
This can be compared with the situation in the MBPT Bloch equation (2.56), where – using the
heavy dot – the folded term could be expressed ˝ P Veff P , indicating that the energy parameters
of the wave operator depend on the intermediate model-space state.
6.6.2 Relation Between the Green’s Operator and Many-Body

Perturbation Procedures
From the conjecture (6.29) and the definition (6.46), we have in the limit of vanish-
ing damping
j˛ .t/i D N˛ U.t; 1/j˚ ˛ i D N˛ G.t; 1/ P U.0; 1/P j˚ ˛ i ; (6.48)
where N˛ is the normalization constant
1
N˛ D ˝ (6.49)
0 jU.0; 1j˚ i
˛ ˛
making the state vector intermediately normalized for t D 0. Here,j˚ ˛ i is the parent
state (6.31), andj ˛ i D N˛ U.0; 1/j˚ ˛ i is the target state (for t D 0). The model
state is
j 0˛ i D P j ˛ i D N˛ P U.0; 1/j˚ ˛ i :
This leads directly to
the relation
ˇ ˛ ˛ ˇ ˛
ˇ .t/ D G.t; 1/ˇ ˛
; (6.50)
0
which implies that the time dependence of the relativistic state vector is gov-
erned by the Green’s operator.
Therefore, the Green’s operator can be regarded as a time-dependent wave op-
erator – but it is NOT an evolution operator in the sense, discussed in Sect. 3.1.
For the time t D 0, we have the covariant analogue of the standard wave oper-
ator of MBPT (2.37)
j˛ .0/i Dj ˛
i D ˝Covj ˛
i (6.51)
with
˝Cov D G.0; 1/: (6.52)
It follows directly from the definition (6.46) that
P G.0; 1/P D P (6.53)
and the relation above can also be expressed
˝Cov D 1 C QG.0; 1/: (6.54)

6.6 Green’s Operator 137
We note here that it is important that the Green’s operator is defined with the dot
product (6.46). The definition of the wave operator (2.37) can be expressed
j ˛
i D ˝Cov P j ˛
i D ˝Cov P U.0; 1/j˚ ˛ i (6.55)
indicating that the energy parameter of the wave operator depends on the interme-
diate model-space state.
We shall also define a covariant effective interaction, analogous to the operator
of MBPT (2.55). The time dependence of the relativistic state vector is formally the
same as that of the nonrelativistic one (2.15) [which is verified below (6.120)] , i.e.,
in interaction picture
ˇ ˛ ˛ ˇ ˛ ˇ ˛
ˇ .t/ D eit .E ˛ H0 /ˇ˛ .0/ D eit .E ˛ H0 /ˇ ˛
(6.56)
or
@ ˇˇ ˛ ˛
.t/ D .E ˛ H0 /j˛ .t/i :
i (6.57)
@t
With the relation (6.50), this yields for the time t D 0

@ ˇˇ ˛ ˛ @
i .t/ j ˛i D i G.t; 1/ j 0˛ i D .E ˛ H0 /j ˛
i : (6.58)
@t t D0 @t t D0
Here, the r.h.s. becomes, using the GML relation (6.32) and the wave-operator rela-
tion (6.51),
.H H0 /j ˛ i D VFj ˛ i D VF ˝Covj 0˛ i : (6.59)
These relations hold for all model states, which leads us to the important operator
relation for the entire model space

@
i G.t; 1/ P D VF ˝Cov P; (6.60)
@t t D0
which we refer to as the reaction operator. Projecting this onto the model space,
yields according to the definition (6.44) to
the covariant relativistic effective interaction
@
Cov
Veff D P VF ˝Cov P D P i G.t; 1/ P: (6.61)
@t t D0
This is a relation in the photonic Fock space, closely analogous to the corre-
sponding relation of standard MBPT (2.55) [c.f. the relation (6.123)].
Our procedure here is based upon quantum-field theory, and the Green’s operator
can be regarded as a field-theoretical extension of the traditional wave-operator
concept of MBPT, and it serves as a connection between field theory and MBPT.
6.7 Model-Space Contribution
We shall now demonstrate how the singularities of the covariant evolution operator
can be eliminated in the general multireference case. We assume that the initial time
is t0 D 1. We also work in the equal-time approximation, where all final times
are the same.
We work in the restricted Hilbert space with no uncontracted photons and
consider a ladder of complete single-photon interactions (6.21), transverse and
Coulomb parts (see Fig. 6.3). (We shall later expand this to more general, irreducible
interactions.)
We start by expanding the relation (6.46) order-by-order, using the fact that
U .0/ .0/P D P ,
U .0/ .t/P D G .0/ .t/ P U .0/ .0/P D G .0/ .t/P

U .1/ .t/P D G .1/ .t/P C G .0/ .t/ P U .1/ .0/P
U .2/ .t/P D G .2/ .t/P C G .1/ .t/ P U .1/ .0/P C G .0/ .t/ P U .2/ .0/P
U .3/ .t/P D G .3/ .t/P C G .2/ .t/ P U .1/ .0/P C G .1/ .t/ P U .2/ .0/P
C G .0/ .t/ P U .3/ .0/P; (6.62)
etc.
It follows from (6.21) that the time dependence of the ladder is given by
eit .Ein Eout / , where Ein and Eout represent the incoming and outgoing energies.
When operating on the part of the model space of energy E, the operator can be
expressed as
U.t/PE D eit .EH0 / U.0/PE : (6.63)
Solving the equations (6.62) for the Green’s operator, we then have
G .0/ .t/P D U .0/ .t/P

G .1/ .t/P D U .1/ .t/P G .0/ .t/ P U .1/ .0/P
G .2/ .t/P D U .2/ .t/P G .0/ .t/ P U .2/ .0/P G .1/ .t/ P U .1/ .0/P
G .3/ .t/P D U .3/ .t/P G .0/ .t/ P U .3/ .0/P G .1/ .t/ P U .2/ .0/P
G .2/ .t/ P U .1/ .0/P; (6.64)
etc. We shall demonstrate that the negative terms above, referred to as counterterms,
will remove the singularities of the evolution operator.
It follows directly from the definition of the dot product above that the singulari-
ties due to disconnected parts are exactly eliminated by the counterterms. Therefore,
we need only consider the connected (ladder) part, and we consider a fully con-
tracted two-body diagram as an illustration (Fig. 6.3). It is sufficient for our present
purpose to consider only positive intermediate states, as in (6.20).
6.7 Model-Space Contribution 139
6.7.1 Lowest Orders
From the above, it follows that the zeroth-order Green’s operator is
G .0/ .t; E/PE D U .0/ .t; E/PE D eit .EH0 / PE D PE : (6.65)
(For clarity, we insert the energy parameter in the operator symbol.)

In first order, we have from (6.64)
G .1/ .t; E/PE D U .1/ .t; E/PE G .0/ .t; E 0 /PE 0 U .1/ .0; E/PE ; (6.66)
where we observe that the Green’s operator in the counterterm has the energy
parameter E 0 , due to the heavy dot in the expression (6.64). The first term is (quasi)-
singular, when the final state lies in the model space, and we shall show that this
singularity is eliminated by the counterterm.
From (6.63), we have
PE 0 U .1/ .t; E/PE D PE 0 G .0/ .t; E/ U .1/ .0; E/PE
(with the energy parameter E in the Green’s operator) and hence the corresponding
part of the Green’s operator (6.66) becomes

PE 0 G .1/ .t; E/PE D G .0/ .t; E/ G .0/ .t; E 0 / PE 0 U .1/ .0; E/PE (6.67)
(P commutes with H0 ). According to (6.20)
1 Vsp .E/
PE 0 U .1/ .0; E/PE D PE 0 .E/ Vsp .E/PE D PE 0 Vsp .E/PE D PE 0 PE
E H0 E E0
(6.68)
and hence we can express the first-order Green’s operator (6.66) as
ıG .0/ .t; E 0 ; E/
G .1/ .t; E/PE D QU .1/ .t; E/PE C PE 0 Vsp PE ; (6.69)
ıE
where QU .1/ .t; E/ D G .0/ .t; E/ Q .E/V .E/PE . We assume here that there is a
summation over E 0 , so that the entire intermediate model-space is covered. The dif-
ference ratio above is defined
ıG .0/ .t; E 0 ; E/ G .0/ .t; E/ G .0/ .t; E 0 / @G .0/ .t; E/

D ) ; (6.70)
ıE E E0 @E
which turns into a derivative at complete degeneracy. Furthermore,
.1/
PE 0 Vsp PE D PE 0 Veff PE (6.71)
G .0/ .E / 6 6 G .0/ .E 0 / 6 6 6 6
U .1/ .E / 6 PE 0
-
6 - U .1/ .E / 6 PE 0
-
6 ) 6 PE 0
-
6
6 6 6 6 6 6
PE PE PE
Fig. 6.8 Illustration of the elimination of singularity of the first-order evolution operator, due to
a final model-space state. The double bar represents the difference ratio/derivative of the zeroth-
order Green’s operator (c.f. Fig. 6.9)
is the first-order effective interaction, which is in accordance with the Fock-space

expression (6.45) [see also (6.123)].
The first-order elimination process is illustrated in Fig. 6.8.
In second order, we have from (6.21)
.2/ .1/ .1/

U0 .t; E/Ladd PE D U0 .t; E/ U0 .0; E/ PE : (6.72)
This can be (quasi)singular, if the final or intermediate state lies in the model space.
If there is a model-space state only at the final state, the counterterm will lead –
in complete analogy with the previous case (6.69) – to the contribution
ıG .0/ .t; E 00 ; E/
PE 00 W0.2/ PE 0 ; (6.73)
ıE
where
.2/
PE 00 W0 PE 0 D PE 00 Vsp .E/ Q .E/Vsp .E/PE (6.74)
is the second-order effective interaction without any intermediate model-space state.
If there is an intermediate model-space state in the second-order evolution oper-
ator (6.72), we have
PE 0
U0.1/ .t; E/PE 0 U0.1/ .0; E/ PE D U0 .t; E/ Vsp .E/ PE : (6.75)
E E0
The singularity will here be eliminated in a similar way by the corresponding
counterterm (6.64). If also the final state lies in the model space, there is an addi-
tional singularity, which is eliminated by replacing U0.1/ .t; E/ by the corresponding
Green’s operator (6.69), yielding for the entire second-order Green’s operator
ıG .0/ .t; E 00 ; E/ .2/

G .2/ .t; E/PE D G .0/ .t; E/ Q .E/V E/ Q .E/V E/PE C PE 00 W0 PE
ıE
ıG .1/ .t; E 0 ; E/
C PE 0 Vsp PE : (6.76)
ıE
6 6 6 6 6 6
G .1/ .E / 6 Q 6 G .1/ .E 0 / 6 Q 6 6 Q 6
- - -
) -
U .1/ .E / 6 PE 0 6 U .1/ .E / 6 PE 0 6 6 PE 0 6
- - -
6 6 6 6 6 6
PE PE PE
Fig. 6.9 Elimination of singularity of the second-order evolution operator, due to an intermediate
model-space state. This leads to a residual contribution that corresponds to the folded diagram in
standard many-body perturbation theory (Fig. 2.5). In addition, there can be a singularity at the
final state, as in first order (see Fig. 6.8)
6 6 6 6 6 6
G .1/ .E / 6P C Q 6 6P C Q 6 6P C Q 6
-
) -
+ -
U .1/ .E / 6P C Q 6 6 Q 6 6 P 6
- - -
6 6 6 6 6 6
Fig. 6.10 Elimination of the singularity of the second-order evolution operator due to an interme-
diate model-space state
The second-order elimination process, due to intermediate model-space state,

is illustrated in Fig. 6.9, and the corresponding part of the Green’s operator is il-
lustrated in Fig. 6.10. This process is quite analogous to the appearance of folded
diagram, discussed in connection with standard MBPT (2.81). Since we are here
dealing with Feynman diagrams, it is more logical to draw the “folded” part straight,
indicating the position of the “fold” by a double bar from which the denominators
of the upper part are to be evaluated. (The elimination process in first order has no
analogy in standard MBPT, since there final model-space states do not appear in the
wave function.)
For t D 0, we have from (6.76)
ıG .1/ .t; E 0 ; E/ .1/

QG .2/ .0; E/PE D Q .E/Vsp .E/ Q .E/Vsp .E/ PE CQ PE 0 Veff PE ;
ıE
(6.77)
which is quite analogous to the corresponding second-order wave operator in ordi-

nary time-independent perturbation theory (2.69). The only difference is here that
the derivative of the first-order Green’s operator leads in addition to the standard
folded term to a term with the energy-derivative of the interaction. The latter term is
sometimes referred to as the reference-state contribution [12], but here we shall re-
fer to both terms as the model-space contribution (MSC), which is more appropriate
in the general multireference case.
We have assumed so far that in the ladder the interactions are identical. If the
interactions are different, some precaution is required. We see in the second-order
expression that the differential/derivative in the last term should refer to the SEC-
OND interaction, while if we treat this in an order-by-order fashion, we would get
the differential of the FIRST interaction. If the interactions are in order V1 and V2 ,
then last term above becomes
ı. Q V2 /
PE 0 V1 PE (6.78)
ıE
(leaving out the arguments). This issue will be further discussed below.
6.7.2 All Orders
The procedure performed above can be generalized to all orders of perturbation

theory. We still consider a two-particle system in the ladder approximation. The
treatment here follows mainly those of [10, 11] but is more general.
We consider an evolution operator in the form of a ladder (6.21) with a general
interaction, V .E/, and with all intermediate model-space states removed, including
also the zeroth-order term,
U0 .t; E/PE D G .0/ .t; E/ 1 C .E/V .E/ C .E/V .E/ Q .E/V .E/ C PE ;
(6.79)
which may have a final model-space state. The corresponding Green’s operator is
according to (6.64)
U0 .t; E/PE G .0/ .t; E/ PE 0 .U0 .0; E/ 1/ PE ; (6.80)
which can be expressed
QU0 .t; E/PE C ıG .0/ .t; E 0 ; E/ PE 0 .U0 .0; E/ 1/ PE : (6.81)
But in analogy with (6.68), we have
PE 0 W0 .E/PE
PE 0 .U0 .0; E/ 1/ PE D ; (6.82)
E E0
where W0 is the effective interaction without intermediate model-space states or

folds, in analogy with (6.74),
W0 .E/PE D V .E/ C V .E/ Q .E/V .E/ C PE : (6.83)
Then the relation (6.81) becomes
ıG .0/ .t; E 0 ; E/
QU0 .t; E/PE C PE 0 W0 PE : (6.84)
ıE
The second term eliminates the singularity due to the final model-space state, and
we shall refer also to this as a folded contribution, in analogy with those eliminating
intermediate model-space singularities.
The Green’s operator with no folds (intermediate or final) is
G0 .t; E/PE D G .0/ .t; E/PE C QU0 .t; E/PE

D G .0/ .t; E/ 1 C Q .E/V .E/ C Q .E/V .E/ Q .E/V .E/ C PE :
(6.85)
The evolution operator with exactly one intermediate model-space state can be
expressed (G .0/ .0/ D 1)
QU0 .t; E/PE 0 .U0 .0; E/ 1/PE (6.86)
and the folded part of (6.84) provides a final fold, yielding the Green’s operator with
one intermediate or final fold,
ıG0 .t; E; E 0 /
G1 .t; E/ D PE 0 W0 PE : (6.87)
ıE
The evolution operator with two intermediate folds can be expressed in analogy
with (6.86)
U2 .t; E/PE D QU0 .t; E/PE 0 .U0 .0; E/ 1/PE 0 .U0 .0; E/ 1/PE : (6.88)
The two leftmost factors represent the Green’s operator (6.86) with one intermediate
fold, and including also a final fold we can replace this by the operator (6.87)
G1 .t; E 0 /PE 00 .U0 .0; E/ 1/PE :
This represents the operator with exactly two intermediate or final model-space state
with the singularity due to the leftmost one being eliminated. Eliminating also the
second fold leads to the Green’s operator with two folds
ıG1 .t; E 0 ; E/
G2 .t; E/PE D PE 0 W0 .E/PE : (6.89)
ıE
Continuing this process leads to (with somewhat simplified notations)

G.t; E/PE D G0 .t; E/ C G1 .t; E/ C G2 .t; E/ C PE
" #
ıG0 .t; E/ ıG1 .t; E/
D G0 .t; E/ C C C W0 PE :
ıE ıE
This yields
ıG.t; E/
G.t; E/PE D G0 .t; E/PE C W0 PE : (6.90)
ıE
Here, the second term represents all intermediate/final folds (model-space contri-
butions). This relation is valid for the entire model space and it is consistent with
[11, Eq. (6.54)] but more general. The expressions given here are valid for all times
and for the final state in P as well as Q spaces. The corresponding wave-operator
relation is obtained by setting t D 0.
We can find an alternative expression for the folded term in (6.90) by considering
G D G0 C G1 C G2 C :
From the expressions above, we find

ıG0
G1 D W0
ıE

ıG1 ı ıG0 ı 2 G0 2 ıG0 ıW0
G2 D W0 D W0 W0 D W C W0
ıE ıE ıE ıE 2 0 ıE ıE
ı 2 G0 2 ıG0
D W C W1 (6.91)
ıE 2 0 ıE
with
ıWn1
Wn D W0 (6.92)
ıE
being the effective interaction with exactly n folds. Similarly,10
10
ı ıG
ıG
ıG GE GE 0 V ııG
ıE E E
V 0
E E0 E
D I V D
ıE E E0 ıE ıE E E0
ıG
VE ııG
E E0 E
V C ııG V ııG
E E0 E
V 0
E E0 E ı2 G ı G ıV
D ıE E D V C
E E 0 ıE 2 ıE ıE
ı ı VE VE 0 ıV
V D
2
VE VE D VE
00 00 DV :
ıE ıE E E0 ıE
This can be generalized to
ı n .AB/ Xn
ı m A ı nm B
D
ıE n
mD0
ı E m ı E nm
(see further [11, Appendix B]).
ıG2
G3 D W0
ıE
ı 3 G0 3 ı 2 G0 ıW0 2 ı 2 G0 ıG0 ıW1
D W C W0 C W1 W0 C W0
ıE 3 0 ıE 2 ıE ıE 2 ıE ıE
or
ı 3 G0 3 ı 2 G0 ıG0
G3 D W C 2W1 W0 C W2 :
ıE 3 0 ıE 2 ıE
Summing this sequence leads to
ıG0
G D G0 C .W0 C W1 C W2 C /
ıE
ı G0
2
ı 3 G0
C 2 .W02 C 2W0 W1 C / C .W03 C / C : (6.93)
ıE ıE 3
It can be shown by induction [10] that this leads to
X1
ı n G0 n
G D G0 C W0 C W1 C W2 C : (6.94)
nD1
ıE n
Here,
Veff D W0 C W1 C W2 C (6.95)
is the total effective interaction, which leads to
X ı n G0 .t; E/ n
G.t; E/PE D G0 .t; E/PE C n
Veff PE : (6.96)
nD1
ıE
This relation is consistent with the results in [10, Eq. (100)] and [11, Eq. (61)], where
more details of the derivations are given. As the previous relation (6.90), it is valid
for all times and with the final state in Q as well as P space. In case the interactions
are different, the derivatives should be taken of the latest interactions.
We can generalize the treatment here and replace the single-photon potential by
the two-body part of the complete irreducible multiphoton exchange potential (6.22)
in Fig. 6.6, V ) V2 D V.
It follows from the treatment here that the counterterms eliminate all singularities
so that the Green’s operator is completely regular at all times.
6.7.2.1 Linkedness of the Green’s Operator
All parts of the expansions above are linked, so this demonstrates that
The Green’s operator is completely linked also in the multireference case.
The linkedness of the single-particle Green’s operator can be expressed, us-
ing (6.4),
" 1 ZZ n Z Z
X 1 i
G .t; t0 / D
1
d3 x d3 x 0 d4 x1 d4 xn
nD0
nŠ c
#
˝ ˇ ˇ ˛
0ˇT O.x/ H.x1 / H.xn / O.x0 / ˇ0 e.jt1 jCjt2 j/
linkedCfolded
(6.97)
and similarly in the many-particle case.

This represents a field-theoretical extension of the linked-diagram theorem of
standard many-body perturbation theory (2.82).
6.8 Bloch Equation for Green’s Operator*
We now want to transform the general expression above for the Green’s operator into
a general Bloch-type of equation (2.56) that, in principle, can be solved iteratively
(self-consistently). Iterations can be performed, only if the in- and outgoing states
contain only particle states of positive energy (no holes). Therefore, we assume this
to be the case. If we have an interaction with hole states in or out, we can apply a
Coulomb interaction, so that all in- and outgoing states are particle states, as will be
discussed further in later chapters.
We still work in the restricted Hilbert space with complete single-photon (or
multiphoton) interactions.
We want to have an equation of the form
G .n/ ; H0 P D V G .n1/ P C folded (6.98)
or
G .n/ PE D G .0/ PE C Q V G .n1/ C folded PE ; (6.99)
where V is the last interaction.

We start from the relation (6.90),
ıG
G D G0 C W0 ; (6.100)
ıE
where
G D G0 C G1 C G2 C
6.8 Bloch Equation for Green’s Operator 147
and Gm is the operator with exactly m intermediate/final folds,
ıGm1
Gm D W0 :
ıE
Furthermore, the total effective interaction is (6.95)
Veff D W0 C W1 C W2 C ; (6.101)
where Wm is the effective interaction (6.92) with m folds and

ıWm1
Wm D W0 : (6.102)
ıE
The folded contribution of order n > 0 is according to (6.99)
G .n/ QV G .n1/ G .0/ D G0.n/ QV G0.n1/ G .0/ C G1.n/ QV G1.n1/

CG2.n/ .n1/
Q V G2 C :
We then see that in the case of no folds we have (6.85)
G0.n/ .n1/
Q V G0 G .0/ D 0: (6.103)
In the case of a single fold, we have

ıG .n/ ıG .n1/
0 0
1 D G1.n/ QV G1.n1/ D W0 QV W0 :
ıE ıE
Here, all terms cancel except those where the last factor of Q V is being differen-
tiated in the first part of 1 and, in addition, terms with a fold in the final state.
Obviously, those terms do not appear in the second part of the difference. This
yields11
ı G .n/
0
1 D W0 ; (6.104)
ıE
11
Distinguishing the various interactions, we can write
G0 D G .0/ 1 C Q V1 C Q V1 Q V2 C
h ıG ı G0 i h ı G .0/ ı. Q V1 / i
0
1 D Q V1 W0 D C G0 1 C Q V2 C W0
ıE ıE ıE ıE
ı G1 h ı 2 G0 ı 2 G0 i
DW Q V 1 W0
ıE ıE 2 ıE 2
"
ı 2 G .0/ ı G .0/ ı. Q V1 / ı 2 . Q V1 /
D C 1 C Q V2 C C G .0/ 1 C Q V2 C
ıE 2 ıE ıE ıE 2
#
ı. Q V1 / ı. Q V2 / ı G1
C G .0/ 1 C Q V3 C C W0 DW :
ıE ıE ıE
where we have introduced the notation ı , with the asterisk indicating that the
differentiation applies only to the last interaction, including the associated resol-
vent, Q V ,
ı. Q Va Q Vb / ı. Q Va /
D Q Vb (6.105)
ıE ıE
and, in addition, differentiation of G .0/ in case there is no QV factor.
In the case of two folds, we have
.n/ .n1/
ıG1 ıG1
2 D G2.n/ QV G2.n1/ D W0 QV W0
ıE ıE
.n/
ı ıG0
D W0 W0
ıE ıE
.n1/
ı ıG0
QV W0 W0
ıE ıE
2 .n/ 2 .n1/
ı G0 2 ı G0 2
D .W0 / QV .W0 /
ıE 2 ıE 2
.n/ .n1/
ıG0 ıG0
C W1 QV W1 :
ıE ıE
With the convention above, we can express the folds
.n/ .n/
ı G0 ı G1
2 D W1 C W0 :
ıE ıE
Continuing this process leads to the total folded contribution

ı G0 ı G1 ıG
C CC .W0 C W1 C / D Veff
ıE ıE ıE
with differentiation with respect to the last factor of QV and to G .0/ , when no
factor of Q V appears.
We then have the generalized Bloch equation for an arbitrary energy-dependent
interaction (V)
ıG
G D G .0/ C QV GC Veff ; (6.106)
ıE
where veff is given by (6.101).

6.8 Bloch Equation for Green’s Operator 149
This equation is valid also when the interactions are different, and then it can
be expressed more explicitly as (n > 0)
X
n1
ı G .m/ .nm/
G .n/ D Q Vn G
.n1/
C Veff ; (6.107)
mD0
ıE
where Vn is the last interaction and the operator G .m/ is formed by the m last
interactions.
We can check the formula (6.106) by considering the first few orders,
G .0/ D eit .EH0 /

ıG .0/ .1/
G .1/ D QV G
.0/
C W0
ıE
ıG .0/ .2/ ı G .1/ .1/
G .2/ D Q V G .1/ C V C W0 ; (6.108)
ıE eff ıE
where the last term becomes
ı G .1/ .1/ ı. Q V / .0/ .1/ ı 2 G .0/ .1/ 2
W0 D G W0 C W0 : (6.109)
ıE ıE ıE 2
This can easily be shown to reproduce the expansions (6.90) and (6.96).
We can also illustrate the validity of the generalized Bloch equation (6.106)
by considering the third-order case with different interactions, V1 ; V2 ; V3 , (see
Fig. 6.11). For simplicity, we assume t D 0 and therefore make the replacement
G ! ˝.
If there is no model-space state directly after the first interaction (Fig. 6.11a), the
contribution becomes

ı. Q V3 /
Q V1 P D Q V3 Q V2 C
.2/
˝ P V2 Q V1 P
ıE
using the second-order expression (6.78) with the last two interactions (V2 ; V3 ).
a b
6 6
V3 - V3 -
ı˝ .2/
- ˝ .2/ - ıE
V2 V2
Q P
V1 - V1 - .1/
Veff
P P
Fig. 6.11 Third-order Green’s operator with different interactions
With a model-space state directly after the first interaction (Fig. 6.11b), the con-
tribution is

ı˝ .2/ .1/ ı ı. Q V3 /
Veff D Q V3 Q V2 C P V2 P V1 P
ıE ıE ıE

ı Q V3 ı Q V2 ı 2 Q V3
D Q V2 C Q V3 C P V2
ıE ıE ıE 2

ı Q V3 ıV2
C P P P V1 P:
ıE ıE
We can now identify the terms above with the Bloch equation (6.106), where the
differentiation should apply to the last interaction. Then we have
ı. Q V2 /
Q V3 ˝
.2/
D Q V3 Q V2 Q V1 P C Q V3 P V1 P
ıE
ı ˝ .2/ .1/ ı Q V3 ı 2 Q V3
Veff D Q V2 P V1 P C P V2 P V1 P
ıE ıE ıE 2
ı ˝ .1/ .2/ ı Q V3 .2/
Veff D W0 C W1.2/
ıE ıE
ı Q V3 ıV2
D P V2 Q V1 P C P P V1 P ;
ıE ıE
the sum of which is identical to the sum of the two previous expressions.
In most applications, we want to have an expression for the wave operator in the
form of a Bloch equation, where we start from a wave operator ˝I and then add an
interaction V that might be different from those involved in ˝I . The Bloch equation
is then of the form
˝P D .˝I C Q .E/V ˝I C folded/P;
where we want to find the form of the folded part. We then make the replacement
˝0 ) Q V ˝I0 in the expression (6.96), where ˝I0 is the wave operator without
folds, yielding
X ın. Q V ˝I0 / n
˝D Veff : (6.110)
nD0
ıE n
The sum can be reformulated as, noting the modified differentiating rules given
above,
X ın. Q V ˝I0 / n
XX n
ım Q V m ı
nm
˝I0 nm
˝D Veff D Veff Veff
nD0
ıE n nD0 mD0
ıE m ıE nm
X ım Q V m
X ı nm ˝I0 nm
X ım Q V m
D Veff V eff D ˝I Veff ;
mD0
ıE m nDm
ıE nm
mD0
ıE m
(6.111)
6.9 Time Dependence of the Green’s Operator. Connection to the Bethe–Salpeter Equation 151
and this leads to the relation
X ın. Q V / n
˝D QV ˝I C n
˝I Veff : (6.112)
nD1
ıE
Since the full wave operator appears only on the left-hand side, this equation does
not have to be solved self-consistently.
We can understand the appearance of the sequence of difference ratios above in
the following way. Each model-space contribution (MSC) should contain a differen-
tiation of all the following interactions. In ˝I , the last interaction, V, is not involved,
and therefore a differentiation of Q V for each interaction in ˝I is required.
We can illustrate the formula above with the third-order case considered pre-
viously (Fig. 6.11), now assuming that we have two Coulomb interactions (VC ),
followed by an energy-dependent potential (V ). Then, we have instead
ı QV
˝ .3/ P D QV Q VC Q VC P C P VC Q VC P
ıE
ı QV ı2 Q V
C Q VC P VC P QV
2
Q VC P VC P C P VC P VC P;
ıE ıE 2
which can be expressed
ı
QV .1/ .2/
˝ .3/ P D QV ˝I .2/ P C ˝I .1/ Veff C ˝I .0/ Veff
ıE
ı2 Q V 2
.1/
C ˝ I
.0/
V eff ;
ıE 2
where
.1/
˝I .1/ D Q VC Veff D P VC P
.2/
˝I .2/ D Q VC Q VC 2
Q VC P VC P Veff D P VC Q VC P:
This is in agreement with the general formula (6.112).
6.9 Time Dependence of the Green’s Operator. Connection

to the Bethe–Salpeter Equation
6.9.1 Single-Reference Model Space
Operating with the relation (6.96) on a model function, 0 , of energy E0 , yields

" 1
#
ˇ ˛ X ı n
G .t; E/ ˇ ˛
G.t; E0 /ˇ 0 D G0 .t; E/ C ˇ 0 :
0
n
E/n (6.113)
nD1
ıE
EDE0
We have here used the fact (6.44) that
Veffj 0i D .E E0 /j 0i D Ej 0i : (6.114)
The expansion (6.113) is a Taylor series, and the result can be expressed
G.t; E0 /j 0i D G0 .t; E/j 0i ; (6.115)
where G0 is the Green’s operator without model-space states (6.85). This implies
that the sum in (6.113), representing
the model-space contributions (MSC) to all orders, has the effect of shifting
the energy parameter from the model energy E0 to the target energy E .
From the relations (6.21) and (6.64), we have the Green’s operator for the ladder
without MSC in the present case, including also the zeroth order and the time factor,
h
G0 .t; E0 /j 0i D eit .E0 H0 / 1 C Q .E0 / V .E0 /
i
C Q .E0 / V .E0 / Q .E0 / V .E0 / C j 0i : (6.116)
The result (6.115) then implies that the Green’s operator with model-space contri-
butions (MSC) becomes
G.t; E0 /j 0i D G0 .t; E/j 0 i D eit .E H0 /

h iˇ ˛
1 C Q .E/ V .E/ C Q .E/ V .E/ Q .E/ V .E/ C ˇ 0 :
(6.117)
shifting also the energy parameter of the time dependence.12
From this, it follows that
@ ˇ ˛ ˇ ˛
i G.t; E0 /ˇ 0 D .E H0 / G.t; E0 /ˇ 0 (6.118)
@t
and, using (6.52),

!
@ ˇ ˛ ˇ ˛
i G.t; E0 / ˇ 0 D .E H0 / ˝ˇ 0 : (6.119)
@t
t D0
12
We observe here that also the zeroth-order term has changed its time dependence, which is
a consequence of the fact that the zeroth-order Green’s operator, G .0/ , is being modified by the
expansion (6.96).
According to (6.50), the Green’s operator has the same time-dependence as the
state vector, in the interaction picture
j.t/i D eit .E H0 / j i (6.120)
(withj i D j.0/i), which implies that the result above – which is a consequence
of the initial conjecture (6.29) – is in accordance with the elementary quantum-
mechanical result (2.15) and (3.2).
Setting the time t D 0 yields with the identity (6.52), ˝j 0 i D G.0; E0 /j 0 i,
the corresponding relation for the wave operator
ˇ ˛
j i D ˝j 0 i D 1 C Q .E/ V .E/ C Q .E/ V .E/ Q .E/ V .E/ C ˇ 0 ;
(6.121)
which is the Brillouin–Wigner expansion of the wave function.
From the relation (6.83), we have that the effective interaction without folds is
ˇ ˛ ˇ ˛
W0 .E0 /ˇ 0 D P V .E0 / C V .E0 / Q .E0 /V .E0 / C ˇ 0 : (6.122)
It can be shown in the same way as for the wave function that inclusion of the folds
(MSC) leads to the replacement E0 ! E (see [10]) and to the expression for the
full effective interaction (6.95)
ˇ ˛ ˇ ˛ ˇ ˛
Veffˇ 0 D W0 .E/ˇ 0 D P V .E/ C V .E/ Q .E0 /V .E/ C ˇ 0 :
But according to the definition (6.44), Veff D P .H H0 /˝P , which gives

ˇ ˛ ˇ ˛
Veffˇ 0 D P .H H0 /˝P D P V .E/˝ˇ 0 : (6.123)
This is an expression for the effective interaction in the restricted Hilbert space
with no uncontracted photons, equivalent to the photonic-Fock-space relation (6.45).
This is analogous to the MBPT result (2.55), but now the perturbation is energy
dependent.
We can generalize this treatment by replacing the single-photon potential V by
the irreducible multiphoton potential in Fig. 6.6, V ) V2 D V. Then we have from
(6.117)
G.t; E0 /j 0i D G0 .t; E/j 0 i D eit .E H0 /

ˇ ˛
1 C Q .E/ V.E/ C Q .E/ V.E/ Q .E/ V.E/ C ˇ 0
(6.124)
and
ˇ ˛ ˇ ˛ ˇ ˛
Veffˇ 0 D P .E H0 / ˝ˇ 0 D P V.E/˝ˇ 0 : (6.125)
From the relation, (6.118) we have

@ ˇ ˛ ˇ ˛
Q i G.t; E0 / ˇ 0 D Q.E H0 / ˝ˇ 0
@t t D0
ˇ ˛
D Q V.E/ C V.E/ Q .E0 / V.E/ C ˇ 0
ˇ ˛
D QV.E/ ˝ˇ 0 :
Combining this with (6.125) leads to the Schrödinger-like equation in the restricted
space
.H0 C V.E//j i D Ej i (6.126)
and an energy-dependent Hamilton operator
H D H0 C V.E/: (6.127)
These relation can be compared with the corresponding GML relations (6.32) and
(6.39) in the photonic Fock space. The equation (6.126) is identical to the effective-
potential form of the Bethe–Salpeter equation (9.20).
6.9.2 Multireference Model Space
We shall now investigate the time dependence of the Green’s operator in a general,
quasi-degenerate model space. We can express the relation (6.96), using the general
perturbation, as
X1
ı n G0 .t; E/ n
G.t; E/PE D G0 .t; E/PE C n
Veff PE ; (6.128)
nD1
ıE
valid in the general multireference (quasi-degenerate) case, where PE is the part of

the model space with energy E and Veff is given by (6.138). This can be formally
expressed as an operator relation
X1
ı n G0 .t; H0 / n
G.t; H0 /P D G0 .t; H0 /P C Veff P; (6.129)
nD1
ı.H0 /n
valid in the entire model space. We have here introduced the symbol A , which im-
plies that the operator A operates directly on the model-space˛ state to the right.
˛
Thus, H0˛BPE D EBPE D BH0 PE . Similarly, Heff
Bj 0˛ D E ˛ Bj 0˛ D

BHeff j 0˛ .
In analogy with (6.45), we have
Veffj ˛
0i D P .E ˛ H0 /˝j ˛
0i D P V.E ˛ /˝j ˛
0i (6.130)
or in operator form

Veff P D P .Heff H0 /˝ P D P V.Heff /˝P: (6.131)
The relation (6.129) is a Taylor expansion in analogy with (6.115), yielding
G.t; H0 / P D G0 .t; Heff

/P (6.132)
using the fact (6.44) that

Heff D PH0 P C Veff

: (6.133)
From (6.116), it follows
G0 .t; E/PE D eit .EH0 / 1 C Q .E/ V.E/ C Q .E/ V.E/ Q .E/ V.E/ C PE
(6.134)
or in operator form

G0 .t; H0 /P D eit .H0 H0 / 1 C
Q .H0 / V.H0 /

C Q .H0 / V.H0 / Q .H0 / V.H0 / C P: (6.135)
This leads in analogy with (6.117), using the relation (6.115), to

G.t; H0 /P D G0 .t; Heff

/P D eit .Heff H0 / 1 C Q .Heff

/ V.Heff /

C Q .Heff / V.Heff / Q .Heff / V.Heff / C P: (6.136)
From this, we conclude that

the general time dependence of the Green’s operator is given by
@
i G.t; H0 /P D .Heff

H0 / G.t; H0 /P: (6.137)
@t
This gives with (6.133)

@
P i G.t; H0 /
P D Veff P; (6.138)
@t t D0
which is the expected result.

In analogy with the single-reference case, the effective interaction becomes

Veff D P V.Heff /˝ P (6.139)
and the Schrödinger-like equation [10, Eq. (113)]
.H0 C V.E ˛ //j ˛

i D E ˛ /j ˛
i: (6.140)
This agrees with the equation derived in [10, Eq. (133)] and it is equivalent to the
Bethe–Salpeter–Bloch equation, discussed in Chap. 9 (9.30).
References
(1983)
2. Åsén, B.: QED effects in excited states of helium-like ions. Ph.D. thesis, Department of
Physics, Chalmers University of Technology and University of Gothenburg, Gothenburg,
Sweden (2002)
3. Bjorken, J.D., Drell, S.D.: Relativistic Quantum Fields. Mc-Graw-Hill Pbl. Co, N.Y. (196)
(1964)
5. E.Lindroth, Mårtensson-Pendrill, A.M.: Isotope Shifts and Energies of the 1s2p States in
Helium. Z. Phys. A 316, 265–273 (1984)
7. Lindgren, I.: Can MBPT and QED be merged in a systematic way? Mol. Phys. 98, 1159–1174
(2000)
8. Lindgren, I., Åsén, B., Salomonson, S., Mårtensson-Pendrill, A.M.: QED procedure applied
to the quasidegenerate fine-structure levels of He-like ions. Phys. Rev. A 64, 062,505 (2001)
Phys. 83, 183–218 (2005)
11. Lindgren, I., Salomonson, S., Hedendahl, D.: Many-body procedure for energy-dependent per-
turbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502
(2006)
227–372 (1998)
(1989)
14. Shabaev, V.M., Artemyev, A.N., Beier, T., Plunien, G., Yerokhin, V.A., Soff, G.: Recoil cor-
rection to the ground-state energy of hydrogenlike atoms. Phys. Rev. A 57, 4235–39 (1998)
15. Stuckelberg, E.C.G.: . Helv.Phys.Acta 15, 23 (1942)
16. Uehling, E.A.: Polarization Effects in the Positron Theory. Phys. Rev. 48, 55–63 (1935)
17. Wichmann, E.H., Kroll, N.M.: Vacuum Polarization in a Strong Coulomb Field. Phys. Rev.
101, 843–59 (1956)
Chapter 7
Numerical Illustrations to Part II
In this chapter, we shall give some numerical illustrations of the three QED methods
described in Part II, the S-matrix, the two-times Green’s function, and the covariant-
evolution-operator methods.
7.1 S-Matrix
7.1.1 Electron Self-Energy of Hydrogen-Like Ions
In the early days of quantum-electrodynamics, the effects were calculated analyti-

cally, applying a double expansion in ˛ and Z˛. For high nuclear charge, Z, such an
expansion does not work well, and it is preferable to perform the evaluation numer-
ically to all orders of Z˛. The first numerical evaluations of the electron self-energy
on heavy, many-electron atoms were performed by Brown et al. in late 1950s [14]
and by Desiderio and Johnson in 1971 [18], applying a scheme devised by Brown,
Langer, and Schaefer [13] (see Sect. 12.3).
An improved method for self-energy calculations, applicable also for lighter sys-
tems, was developed and successfully applied to hydrogen-like ions by Mohr [34,
39–42]. The energy shift due to the first-order electron self-energy is conventionally
expressed as:
˛ .Z˛/4
E D F .Z˛/ mc 2 (7.1)
n3
where n is the main quantum number. The function F .Z˛/ is evaluated numerically,
and some results are given in Table 7.1.
Performing accurate self-energy calculations for low Z is complicated due to
slow convergence. Mohr has estimated the first-order Lamb shift (self-energy C
vacuum polarization) by means of elaborate extrapolation from heavier elements
and obtained the value 1057.864(14) MHz for the 2s 2p1=2 shift in neutral hy-
drogen [40], in excellent agreement with the best experimental value at the time,
1057.893(20) MHz. More recently, Jentschura et al. [26] have extended the method

158 7 Numerical Illustrations to Part II
Table 7.1 The F .Z˛/ Reference F .Z˛/

function for the ground state
Desiderio and Johnson [18] 1.48
of hydrogen-like mercury
Mohr [38] 1.5032(6)
Blundell and Snyderman [9] 1.5031(3)
Mohr [34] 1.5027775(4)
Table 7.2 Ground-state Correction Value Reference

Lamb shift of hydrogen-like
uranium (in eV, mainly from Nuclear size 198:82
[43]) First-order self-energy 355:05 [34, 45]
Vacuum polarization 88:59 [49]
Second-order effects 1:57
Nuclear recoil 0:46
Nuclear polarization 0:20
Total theory 463:95
Experimental 460:2.4:6/
of Mohr to calculate directly the self-energy of light elements down to hydrogen

with extremely high accuracy. Accurate calculations have also been performed for
highly excited states [27].
The original method of Mohr was limited to point-like nuclei but was extended to
finite nuclei in a work with Gerhard Soff [45]. An alternative method also applicable
to finite nuclei has been devised by Blundell and Snyderman [9, 10].
7.1.2 Lamb Shift of Hydrogen-Like Uranium
In high-energy accelerators, like that at GSI in Darmstadt, Germany, highly charged

ions up to hydrogen-like uranium can be produced. For such systems the QED ef-
fects are quite large, and accurate comparison between experimental and theoretical
results can here serve as an important test of the QED theory in extremely strong
electromagnetic fields – a test that has never been performed before.
The first experimental determination of the Lamb-shift in hydrogen-like uranium
was made by the GSI group (Stöhlker, Mokler et al.) in 1993 [56]. The result was
429(23) eV, a result that has gradually been improved by the group, and the most
recent value is 460.2(4.6) eV [55]. The shift is here defined as the experimental
binding energy compared to the Dirac theory for a point nucleus, implying that it
includes also the effect of the finite nuclear size. In Table 7.2, we show the various
contributions to the theoretical value. The self-energy contribution was evaluated by
Mohr [34] and the finite-nuclear-size effect by Mohr and Soff [45]. The vacuum po-
larization, including the Wickmann–Kroll correction (see Sect. 4.6.3), was evaluated
by Persson et al. [49]. The second-order QED effects, represented by the diagrams
in Fig. 7.1, have also been evaluated. Most of the reducible part was evaluated by
Persson et al. [48]. The last two irreducible too-loop diagrams are much more elabo-
rate to calculate and have only recently been fully evaluated by Yerokhin et al. [60].1
1
See Sect. 2.6.
7.1 S-Matrix 159
-

- -
- - 6
6 6
6
6 - 6 ?
1
6
6- 6
6 6
6 6 6
6 6 6
6
6 6 66
6 6 6
6 6 6
Fig. 7.1 Second-order contributions to the Lamb shift of hydrogen-like ions (c.f. Fig. 5.3)
The main uncertainty of the theoretical calculation on hydrogen-like uranium

stems from the finite-nuclear-size effect, which represents almost half of the entire
shift from the Dirac point-nuclear value. Even if the experimental accuracy would
be significantly improved, it will hardly be possible to test with any reasonable ac-
curacy the second-order QED effects, which are only about 1% of the nuclear-size
effect. For that reason, other systems, like lithium-like ions, seem more promising
for testing such effects.
7.1.3 Lamb Shift of Lithium-Like Uranium
The 2s 2p1=2 Lamb shift of lithium-like uranium was measured at the Berkeley
HILAC accelerator by Schweppe et al. in 1991 [54]. The first theoretical evalu-
ations of the self-energy was performed by Cheng et al. [15] and the complete
first-order shift, including vacuum polarization, by Blundell [7], Lindgren et al. [31],
Table 7.3 2s 2p1=2 Lamb shift of lithium-like uranium (in eV)

Correction Ref. [7] Ref. [48] Ref. [58]
Relativistic MBPT 322.41 322.32 322.10
First-order self-energy 53:94 54:32
First-order vacuum polarization (12.56) 12.56
First-order self-energy C vac. pol. 41:38 41:76 41:77
Second-order self-energy C vac. pol. 0.03 0.17
Nuclear recoil (0.10) (0:08) 0:07
Nuclear polarization (0.10) (0.03) 0:07
Total theory 280.83(10) 280.54(15) 280.48(20)
Experimental 280.59(9)
Fig. 7.2 Feynman diagrams

representing the two-photon 6 6 6 6
exchange (ladder and cross)
-
for helium-like ions 6 6 6 6 6
-
6 6 6 6
and Persson et al. [48], the latter calculation including also some reducible second-
order QED effects. Later, more complete calculations were performed by Yerokhin
et al. [58]. The results are summarized in Table 7.3.
In lithium-like systems, the nuclear-size effect is considerably smaller than in
the corresponding hydrogen-like system and can be more easily accounted for. The
second-order QED effects in Li-like uranium are of the same order as the present
uncertainties in theory and experiment, and with some improvement these effects
can be tested. Therefore, systems of this kind seem to have the potential for the
most accurate test of high-field QED at the moment.
7.1.4 Two-Photon Nonradiative Exchange in Helium-Like Ions
Accurate S-matrix calculations of the nonradiative two-photon exchange for

helium-like ions (ladder and cross), corresponding to the Feynman diagrams in
Fig. 7.2, have been performed by Blundell et al. [8] and Lindgren et al. [30]. The
results are illustrated in Fig. 7.3 (taken from [30]). In the figure, the contributions
are displayed versus the nuclear charge, relative to the zeroth-order nonrelativistic
ionization energy, Z 2 =2 (in atomic Hartree units). The vertical scale is logarithmic,
so that 1 corresponds to ˛, 2 to ˛ 2 , etc.
As comparison, we show in the top picture of Fig. 7.3 the energy contribution due
to first-order Coulomb and Breit interactions as well as the first-order Lamb shift,
corresponding to Feynman diagrams shown in the top line of Fig. 7.4.
For low Z, the first-order Coulomb interaction is proportional to Z, the first-
order Breit interaction to Z 3 ˛ 2 , and the first-order Lamb shift to Z 4 ˛ 3 . For high
7.1 S-Matrix 161
Fig. 7.3 Various contributions to the ground-state energy of He-like ions. The top picture repre-
sents the first-order contributions, the middle picture the second-order contributions in the NVPA
as well as the screened Lamb shift, and the bottom picture contributions due to retardation and
virtual pairs (see Fig. 7.4). The values are normalized to the nonrelativistic ionization energy, and
the scale is logarithmic (powers of the fine-structure constant ˛)
6
6 6 6 6
6
6 6 6 6 6
6
6 6 6 6 6 6
6
6 6 6 6 6 6 6
6
6 6 6 6 6 6 6
6 6 6 6 6 6 6 6 6
6 6 6
6 6 6 6 - 6
6- 6 6- 6 6 6 6 -
6 6 6
6 6 6 6 6
6 6 6 6 6 6 6
Fig. 7.4 Feynman diagrams representing the one- and two-photon exchange, separated into
Coulomb, instantaneous Breit and retarded Breit interactions
Z, we can replace Z˛ by unity, and then after dividing by Z 2 , all first-order effects
tend to ˛ as Z increases, as is clearly seen in the top picture of Fig. 7.3 (see also
Fig. 10.9 and Table 10.1).
An additional Coulomb interaction reduces the effect for small Z by a fac-
tor of Z. Therefore, the Coulomb–Coulomb interaction, i.e., the leading electron
correlation, is in first order independent of Z and the Coulomb–Breit interaction
proportional to Z 2 ˛ 2 . The screened Lamb shift is proportional to Z 3 ˛ 3 and the
second-order Breit interaction (in the no-pair approximation) to Z 4 ˛ 4 . After di-
vision with Z 2 , we see (second picture of Fig. 7.3) that all second-order effects
tend to ˛ 2 . The corresponding Feynman diagrams are shown in the second row of
Fig. 7.4.
The third picture in Fig. 7.3 shows the effect of the retarded Coulomb–Breit and
Breit–Breit interactions without and with virtual pairs, corresponding to diagrams
in the bottom row of Fig. 7.4. For low Z, these effects are one order of ˛ smaller
than the corresponding unretarded interactions with no virtual pairs, while for high
Z they tend – rather slowly – to the same ˛ 2 limit. It is notable that for the
Coulomb–Breit interactions the retardation and virtual pairs have nearly the same
effect but with opposite sign. For the Breit–Breit interactions, the effects of sin-
gle and double pairs have opposite sign and the total effect changes its sign around
Z D 40.
7.1 S-Matrix 163
Table 7.4 Two-photon Z 2 3 S1 2 3 P0 2 3 P2

effects on some excited states
30 MBPT 49 541 88 752 75 352
of helium-like ions (in
Hartree, from [44]) QED 8:7 145 77.6
50 MBPT 53 762 123 159 79 949
QED 64 1340 767
80 MBPT 66 954 251 982 93
QED 966 9586 5482
Table 7.5 Two-electron effects on the ground-state energy of helium-like

ions (in eV, from [50])
Z Plante et al. Indelicato et al. Drake Persson et al. Expt’l
32 652.0 562.1 562.1 562.0 562.5˙1:5
54 1,028.4 1,028.2 1,028.8 1,028.2 1,027.2˙3:5
66 1,372.2 1,336.5 1,338.2 1,336.6 1,341.6˙4:3
74 1,574.8 1,573.6 1,576.6 1,573.9 1,568˙15
83 1,880.8 1,886.3 1,881.5 1,876˙14
More recently, Mohr and Sapirstein have performed S-matrix calculations also
on the excited states of helium-like ions and compared with second-order MBPT
calculations to determine the effect of nonradiative QED, retardation, and virtual
pairs [44], and some results are shown in Table 7.4.
7.1.5 Electron Correlation and QED Calculations on Ground

States of Helium-Like Ions
The two-electron effect on the ground-state energy of some helium-like ions has
been measured by Marrs et al. at Livermore Nat. Lab. by comparing the ionization
energies of the corresponding helium-like and hydrogen-like ions [33]. (The larger
effect due to single-electron Lamb shift is eliminated in this type of experiment.)
Persson et al. [50] have calculated the two-electron contribution by adding to the
all-order MBPT result the effect of two-photon QED, using dimensional regulariza-
tion (see Chap. 12). The results are compared with the experimental results as well
as with other theoretical estimates in Table 7.5. The results of Drake were obtained
by expanding relativistic and QED effects in powers of ˛ and Z˛, using Hyller-
aas type of wave functions [20]. The calculations of Plante et al. were made by
means of relativistic MBPT and adding first-order QED corrections taken from the
work of Drake [52], and the calculations of Indelicato et al. were made by means of
multiconfigurational Dirac–Fock with an estimate of the Lamb shift [25]. The agree-
ment between experiments and theory is quite good, although the experimental
accuracy is not good enough to test the QED parts, which lie in the range 1–5 eV.
The agreement between the various theoretical results is very good – only the results
of Drake are somewhat off for the heaviest elements, which is due to the shortcom-
ing of the power expansion.
7.1.6 g-Factor of Hydrogen-Like Ions: Mass of the Free Electron
The Zeeman splitting of hydrogen-like ions in a magnetic field is another good test
of QED effects in highly charged ions. The lowest-order contributions to this effect
are represented by the Feynman diagrams in Fig. 7.5.
The bound-electron g-factor can be expanded as [51]:

1h p i ˛ 1 .Z˛/2
gJ D 2 1 C 2 1 .Z˛/2 C C C ; (7.2)
3 2 12
where Z is the nuclear charge. The first term represents the relativistic value with
a correction from the Dirac value of order ˛ 2 . The second term, proportional to ˛,
is the leading QED correction, known as the Schwinger correction, and the follow-
ing term, proportional to ˛ 3 , is the next-order QED correction, first evaluated by
Grotsch [23].
Numerical calculations to all orders in Z˛ have been performed by Blundell
et al. [11] [only self-energy part, (b,c) in Fig. 7.5] and by the Gothenburg
group [6, 51] [incl. the vacuum polarization (d,e)]. The results are displayed in
Fig. 7.6, showing the comparison between the Grotsch term (the leading QED cor-
rection beyond the Schwinger correction) and the numerical result. (The common
factor of 2˛= has been left out.) More accurate calculations have later been per-
formed by the St Petersburg group, including also two-loop corrections and the
nuclear recoil [59, 61].
The g-factors of hydrogen-like ions have been measured with high accuracy by
the Mainz group, using an ion trap of Penning type [4, 24]. The accuracy of the
experimental and theoretical determinations is so high that the main uncertainty
is due to the experimental mass of the electron. Some accurate dates for H-like
carbon are shown in Table 7.6. By fitting the theoretical and experimental values,
a value of the electron mass (in atomic mass units) me D 0:0005485799093.3/
is deduced from the carbon experiment and the value me D 0:0005485799092.5/
from a similar experiment on oxygen [4]. These results are four times more ac-
curate than the previously accepted value, me D 0:0005485799110.12/ [37]. The
new value is now included in the latest adjustments of the fundamental constants
[35, 36].
a b c d e
6 6 6
6 - ? 6 -
- 6
- 66 - -
6 6
6 6 -
-
6 6
6 6 6
Fig. 7.5 Feynman diagrams representing the lowest-order contributions to the Zeeman effect of
hydrogen-like ions. Diagrams (b) and (c) represent the leading self-energy correction to the first-
order effect (a, d and e) the leading vacuum-polarization correction
7.2 Green’s-Function and Covariant-Evolution-Operator Methods 165
Fig. 7.6 The first-order, numerically evaluated, QED correction to the gj value of hydrogen-like
ions, compared with the leading analytical (Grotsch) term (7.2). Both results are first-order in ˛
but the numerical result is all order in Z˛, while the Grotsch result contains only the leading term
(from [51]). A common factor 2˛= is left out
Table 7.6 Theoretical Correction Value

contributions to the g-factor
Dirac theory 1.998 721 3544
of hydrogen-like carbon
(mainly from [5]) Finite nuclear-size corr. C0.000 000 0004
Nuclear recoil C0.000 000 0876
Free-electron QED, first order C0.002 322 8195
Free-electron QED, higher orders C0.000 003 5151
Bound-electron QED, first order C0.000 000 8442
Bound-electron QED, higher orders 0.000 000 0011
Total theory 2.001 041 5899
7.2 Green’s-Function and Covariant-Evolution-Operator

Methods
7.2.1 Fine-Structure of Helium-Like Ions
The two-times Green’s function and the covariant-evolution-operator methods have

the important advantage over the S-matrix formulation that they can be applied also
to quasi-degenerate energy levels. As an illustration, we consider here the evalu-
ation of some fine-structure separations of the lowest P state of helium-like ions
(see Table 7.7). The calculations of Plante et al. [52] are relativistic many-body
calculations in the NVPA scheme (see Sect. 2.6) with first-order QED-energy cor-
rections, taken from the work of Drake [20]. The calculations by Åsén et al. [2, 29],
Table 7.7 The 1s2p 3 P fine structure of He-like ions. (Values for Z D 2, 3 given in MHz and the
remaining ones in Hartree)
Z 3 P1 3 P0 3
P2 3 P0 3
P2 3 P1 Ref. Expt’l Ref. Theory
2 29616.95166(70) 2291.17759(51) Gabrielse et al. [62]
29616.9527(10) Giusfredi et al. [22]
29616.9509(9) Hessels et al. [21]
2291.17753(35) Hessels et al. [12]
29616.9523(17) 2291.1789(17) Pachucki et al. [57]
3 155704.27(66) 62678:41.65/ Riis et al. [53]
62678:46.98/ Clarke et al. [16]
155703.4(1,5) 62679:4.5/ Drake et al. [20]
9 701(10) 5064(8) 4364.517(6) Myers et al. [46]
680 5050 4362(5) Drake et al. [20]
681 5045 4364 Plante et al. [52]
690 5050 4364 Åsén et al. [2, 32]
10 1371(7) 8458(2) 7087(8) Curdt et al. [17]
1361(6) 8455(6) 7094(8) Drake et al. [20]
1370 8469 7099 Plante et al. [52]
1370 8460 7090 Åsén et al. [29]
12 3789(26) 20069(9) 16280(27) Curdt et al. [17]
3796(7) Myers et al. [47]
3778(10) 20046(10) 16268(13) Drake et al. [20]
3796 20072 16276 Plante et al. [52]
3800,1 20071 Artemyev et al. [1]
14 40707(9) Curdt et al. [17]
8108(23) 40708(23) 32601(33) Drake et al. [20]
8094 40707 32613 Plante et al. [52]
40712 Artemyev et al. [1]
18 124960(30) Kukla et al. [28]
124810(60) Drake et al. [20]
23692 124942 101250 Plante et al. [52]
23790 124940 101150 Åsén et al. [29]
124945(3) Artemyev et al. [1]
using the recently developed covariant-evolution-operator method, was the first nu-
merical evaluation of QED effects (nonradiative) on quasi-degenerate energy levels.
It can be noted that the energy of the 1s2p 3 P1 state, which a linear combination
of the closely spaced states 1s2p1=2 and 1s2p1=3 , could not be evaluated by the
S-matrix formulation (see, for instance, the above-mentioned work of Mohr and
Sapirstein [44]). Later, calculations have also been performed on these systems by
the St Petersburg group, using the two-times-Green’s-function method [1], where
also the radiative parts are evaluated numerically.
The accuracy of the experimental and theoretical fine-structure results is not
sufficient to distinguish between the first-order energy QED corrections and the
7.2 Green’s-Function and Covariant-Evolution-Operator Methods 167
Table 7.8 The transition Reference

1s2s 1 S0 1s2p 3 P1 for
Expt’l 7230.585(6) Myers et al. [19]
He-like Si (in cm1 )
Theory 7231.1 Plante et al. [52]
7229(2) Artemyev et al. [1]
Table 7.9 Two-photon

Z 2 3 S0 2 3 S1 2 3 S1
calculations on the
1s2s 1 S; 3 S states of 10 MBPT 116 005 47 638
helium-like ions (in Hartree, QED 6.2 1:2
first two columns from Åsén 18 MBPT 119 381 48 158
et al. [3], last column from QED 3.8 4.6
Mohr and Sapirstein [44]) 30 MBPT 128 349 49 542 49 541
QED 93 6.9 8.7
60 MBPT 177 732 57 025 57 023
QED 2358 216 224
numerical evaluation of Åsen and Artemyev. On the other hand, the experimen-
tal accuracy of the separation of Fluorine (Z D 9) seems to be sufficient to test even
higher-order QED effects. Here, present theory cannot match the experimental ac-
curacy, but this might be a good testing case for the new combined QED-correlation
procedure, discussed in the following.
As a second illustration, we consider the transition 1s2s 1 S0 1s2p 3 P1 for
He-like silicon, which has recently been very accurately measured by Myers
et al. [19] (see Table 7.8). Corresponding calculations have been performed by
Plante et al. [52], using relativistic MBPT with first-order QED correction, and by
Artemyev et al. [1], using the two-times Green’s function. Here, it can be seen that
the experiment is at least two orders of magnitude more accurate than the theoreti-
cal estimates. Also here the combined MBPT–QED corrections are expected to be
significant.
7.2.2 Energy Calculations of 1s2 s Levels

of Helium-Like Ions
The covariant-evolution-operator method has also been applied by Åén et al. [2, 3]
to evaluate the two-photon diagrams in Fig. 7.2 for the first excited S states of
some helium-like ions. The results are compared with relativistic MBPT results,
to determine the nonradiative QED effects, as in Table 7.4 above. The results are
shown in Table 7.9, where comparison is also made with some results of Mohr and
Sapirstein [44].
References
the nD1 and nD2 energy levels in He-like ions. Phys. Rev. A 71, 062,104 (2005)
2. Åsén, B.: QED effects in excited states of helium-like ions. Ph.D. thesis, Department of
Physics, Chalmers University of Technology and University of Gothenburg, Gothenburg,
Sweden (2002)
3. Åsén, B., Salomonson, S., Lindgren, I.: Two-photon exchange QED effects in the 1s2s 1 S and
3
S states of heliumlike ions. Phys. Rev. A 65, 032,516 (2002)
4. Beier, T., Djekic, S., Häffner, H., Indelicato, P., Kluge, H.J., Quint, W., Shabaev, V.S.,
Verdu, J., Valenzuela, T., Werth, G., Yerokhin, V.A.: Determination of electron’s mass from
g-factor experiments on 12 C C5 and 16 O C7 . Nucl. Instrum. Methods 205, 15–19 (2003)
5. Beier, T., Häffner, H., Hermanspahn, N., Karshenboim, S.G., Kluge, H.J.: New Determination
of the Electron’s Mass. Phys. Rev. Lett. 88, 011,603.1–4 (2002)
6. Beier, T., Lindgren, I., Persson, H., Salomonson, S., Sunnergren, P., Häffner, H.,
Hermanspahn, N.: gj factor of an electron bound in a hydrogenlike ion. Phys. Rev. A 62,
032,510 (2000)
7. Blundell, S.: Calculations of the screened self-energy and vacuum polarization in Li-like, Na-
like, and Cu-like ions. Phys. Rev. A 47, 1790–1803 (1993)
8. Blundell, S., Mohr, P.J., Johnson, W.R., Sapirstein, J.: Evaluation of two-photon exchange
graphs for highly charged heliumlike ions. Phys. Rev. A 48, 2615–26 (1993)
9. Blundell, S., Snyderman, N.J.: Basis-set approach to calculating the radiative self-energy in
highly ionized atoms. Phys. Rev. A 44, R1427–30 (1991)
10. Blundell, S.A.: Accurate screened QED calculations in high-Z many-electron ions. Phys. Rev.
A 46, 3762–75 (1992)
11. Blundell, S.A., Cheng, K.T., Sapirstein, J.: Radiative corrections in atomic physics in the
presence of perturbing potentials. Phys. Rev. A 55, 1857–65 (1997)
12. Borbely, J.S., George, M.C., Lombardi, L.D., Weel, M., Fitzakerley, D.W., Hessels, E.A.: .
Phys. Rev. A 79, 060,503 (2009)
13. Brown, G.E., Langer, J.S., Schaeffer, G.W.: Lamb shift of a tightly bound electron. I. Method.
Proc. R. Soc. London, Ser. A 251, 92–104 (1959)
14. Brown, G.E., Mayers, D.F.: Lamb shift of a tightly bound electron. II. Calculation for the
K-electron in Hg. Proc. R. Soc. London, Ser. A 251, 105–109 (1959)
15. Cheng, T.K., Johnson, W.R., Sapirstein, J.: Screend Lamb-shift calculations for Lithiumlike
Uranium, Sodiumlike Platimun, and Copperlike Gold. Phys. Rev. Lett. 23, 2960–63 (1991)
16. Clarke, J.J., van Wijngaaren, W.A.: . Phys. Rev. A 67, 12,506 (2003)
17. Curdt, W., Landi, E., Wilhelm, K., Feldman, U.: Wavelength measurements of heliumlike 1s2s
3
S1 1s2p 2 P0;2 transitions in Ne 8C , Na9C , M g 10C , and Si 12C emitted by solar flare
plasmas. Phys. Rev. A 62, 022,502.1–7 (2000)
18. Desiderio, A.M., Johnson, W.R.: Lamb Shift and Binding Energies of K Electrons in Heavy
Atoms. Phys. Rev. A 3, 1267–75 (1971)
19. DeVore, T.R., Crosby, D.N., Myers, E.G.: Improved Measurement of the 1s2s 1 S0 1s2p 3 P1
Interval in Heliumlike Silicon. Phys. Rev. Lett. 100, 243,001 (2008)
20. Drake, G.W.F.: Theoretical energies for the n D 1 and 2 states of the helium isoelectronic
sequence up to Z D 100. Can. J. Phys. 66, 586–611 (1988)
21. George, M.C., Lombardi, L.D., Hessels, E.A.: Precision Microwave Measurement of the
2 3 P1 2 3 P0 Interval in Atomic Helium: A determination of the Fine-Structure Constant.
Phys. Rev. Lett. 87, 173,002 (2001)
22. Giusfredi, G., Pastor, P.C., Natale, P.D., Mazzotti, D., deMauro, C., Fallani, L., Hagel, G.,
Krachmalnicoff, V., Ingusio, M.: Present status of the fine structure of the 2 3 P helium level.
Can. J. Phys. 83, 301–10 (2005)
23. Grotsch, H.: . Phys. Rev. Lett. 24, 39 (1970)
24. Häffner, H., Beier, T., Hermanspahn, N., Kluge, H.J., Quint, W., Stahl, S., Verdú, J., Werth,
G.: High-Acuuracy Measurements of the Magnetic Anomaly of the Electron Bound in Hydro-
genlike Carbon. Phys. Rev. Lett. 85, 5308–11 (2000)
References 169
25. Indelicato, P., Gorciex, O., Desclaux, J.P.: Multiconfiguration Dirac-Fock studies of two-
electron ions: II. Radiative corrections and comparison with experiment. J. Phys. B 20,
651–63 (1987)
26. Jentschura, U.D., Mohr, P.J., Soff, G.: Calculation of the Electron Self-Energy for Low Nuclear
Charge. Phys. Rev. Lett. 61, 53–56 (1999)
27. Jentschura, U.D., Mohr, P.J., Tan, J.N., Wundt, B.J.: Fundamental Constants and Tests of
Theory in Rydberg States of Hydrogenlike Ions. Phys. Rev. Lett. 100, 160,404 (2008)
28. Kukla, K.W., Livingston, A.E., Suleiman, J., Berry, H.G., Dunford, R.W., Gemmel, D.S.,
Kantor, E.P., Cheng, S., Curtis, L.J.: Fine-structure energies for the 1s2s 3 S 1s2p 3 P tran-
sition in heliumlike ArC . Phys. Rev. A 51, 1905–17 (1995)
29. Lindgren, I., Åsén, B., Salomonson, S., Mårtensson-Pendrill, A.M.: QED procedure applied
to the quasidegenerate fine-structure levels of He-like ions. Phys. Rev. A 64, 062,505 (2001)
30. Lindgren, I., Persson, H., Salomonson, S., Labzowsky, L.: Full QED calculations of two-
photon exchange for heliumlike systems: Analysis in the Coulomb and Feynman gauges. Phys.
Rev. A 51, 1167–1195 (1995)
31. Lindgren, I., Persson, H., Salomonson, S., Ynnerman, A.: Bound-state self-energy calculation
using partial-wave renormalization. Phys. Rev. A 47, R4555–58 (1993)
33. Marrs, R.E., Elliott, S.R., Stöhkler, T.: Measurement of two-electron contributions to the
ground-state energy of heliumlike ions. Phys. Rev. A 52, 3577–85 (1995)
34. Mohr, P.: Self-energy correction to one-electron energy levels in a strong Coulomb field. Phys.
Rev. A 46, 4421–24 (1992)
35. Mohr, P., Taylor, B.: CODATA recommended values of the fundamental constants 2002. Rev.
Mod. Phys. 77, 1–107 (2005)
36. Mohr, P., Taylor, B.: CODATA recommended values of the fundamental constants 2002. Rev.
Mod. Phys. (2008)
37. Mohr, P., Taylor, B.N.: CODATA recommended values of the fundamental physical constants:
1998. Rev. Mod. Phys. 72, 351–495 (2000)
88, 52–87 (1974)
39. Mohr, P.J.: Self-Energy Radiative Corrections. Ann. Phys. (N.Y.) 88, 26,52 (1974)
40. Mohr, P.J.: Lamb Shift in a Strong Coulomb Field. Phys. Rev. Lett. 34, 1050–52 (1975)
41. Mohr, P.J.: Self-energy of the n D 2 states in a strong Coulomb field. Phys. Rev. A 26, 2338–
54 (1982)
42. Mohr, P.J.: Quantum electrodynamics of high-Z few-electron atoms. Phys. Rev. A 32, 1949–57
(1985)
227–372 (1998)
44. Mohr, P.J., Sapirstein, J.: Evaluation of two-photon exchange graphs for excited states of
highly charged heliumlike ions. Phys. Rev. A 62, 052,501.1–12 (2000)
45. Mohr, P.J., Soff, G.: Nuclear Size Correction to the Electron Self Energy. Phys. Rev. Lett. 70,
158–161 (1993)
46. Myers, E.G., Margolis, H.S., Thompson, J.K., Farmer, M.A., Silver, J.D., Tarbutt, M.R.: Pre-
cision Measurement of the 1s2p 3 P2 3 P1 Fine Structure Interval in Heliumlike Fluorine.
Phys. Rev. Lett. 82, 4200–03 (1999)
47. Myers, E.G., Tarbutt, M.R.: Measurement of the 1s2p 3 P0 3 P1 fine structure interval in
heliumlike magnesium. Phys. Rev. A 61, 010,501R (1999)
48. Persson, H., Labzowsky, I.L.L.N., Plunien, G., Beier, T., Soff, G.: Second-order self-energy–
vacuum-polarization contributions to the Lamb shift in highly charged few-electron ions. Phys.
Rev. A 54, 2805–13 (1996)
49. Persson, H., Lindgren, I., Salomonson, S., Sunnergren, P.: Accurate vacuum-polarization cal-
culations. Phys. Rev. A 48, 2772–78 (1993)
50. Persson, H., Salomonson, S., Sunnergren, P., Lindgren, I.: Two-electron Lamb-Shift Calcula-
tions on Heliumlike Ions. Phys. Rev. Lett. 76, 204–07 (1996)
51. Persson, H., Salomonson, S., Sunnergren, P., Lindgren, I.: Radiative corrections to the electron
g-factor in H-like ions. Phys. Rev. A 56, R2499–2502 (1997)
the n D 1 and n D 2 states of heliumlike ions. Phys. Rev. A 49, 3519–30 (1994)
53. Riis, E., Sinclair, A.G., Poulsen, O., Drake, G.W.F., Rowley, W.R.C., Levick, A.P.: Lamb shifts
and hyperfine structure in 6 LiC and 7 LiC : Theory and experiment. Phys. Rev. A 49, 207–20
(1994)
54. Schweppe, J., Belkacem, A., Blumenfield, L., Clayton, N., Feinberg, B., Gould, H., Kostroun,
V.E., Levy, L., Misawa, S., Mowat, J.R., Prior, M.H.: Measurement of the Lamb Shift in Lithi-
umlike Uranium (U C89 /. Phys. Rev. Lett. 66, 1434–37 (1991)
55. Sthöhlker, T., Beyer, H.F., Gumberidze, A., Kumar, A., Liesen, D., Reuschl, R., Spillmann, U.,
Trassinelli, M.: Ground state Lamb-shift of heavy hydrogen-like ions: status and prspectives.
Hyperfine Interactions 172, 135–40 (2006)
56. Stöhlker, T., Mokler, P.H., et al., K.B.: Ground-state Lamb Shift for Hydrogenlike Uranium
Measured at the ESR Storage Ring. Phys. Rev. Lett. 14, 2184–87 (1993)
57. Yerokhin, K.P.V.A.: Fine Structure of Heliumlike Ions and Determination of the Fine Structure
Constant. Phys. Rev. Lett. 104, 70,403 (2010)
58. Yerokhin, V.A., Artemyev, A.N., Shabaev, V.M., Sysak, M.M., Zherebtsov, O.M., Soff, G.:
Evaluation of the two-photon exchange graphs for the 2p1=2 2s transition in Li-like Ions.
Phys. Rev. A 64, 032,109.1–15 (2001)
59. Yerokhin, V.A., Indelicato, P., Shabaev, V.M.: Self-energy Corrections to the Bound-Electron
g Factor in H-like Ions. Phys. Rev. Lett. 89, 134,001 (2002)
60. Yerokhin, V.A., Indelicato, P., Shabaev, V.M.: Two-Loop Self-Energy Correction in High-Z
Hydrogenlike Ions. Phys. Rev. Lett. 91, 073,001 (2003)
61. Yerokhin, V.A., Indelicato, P., Shabaev, V.M.: Evaluation of the self-energy correction to the
g Factor of S states in H-like Ions. Phys. Rev. A 69, 052,203 (2004)
62. Zelevinsky, T., Farkas, D., Gabrielse, G.: Precision Measurement of the Three 23 PJ Helium
Fine Structure Intervals. Phys. Rev. Lett. 95, 203,001 (2005)
Part III
Quantum-Electrodynamics Beyond
Two-Photon Exchange: Field-Theoretical
Approach to Many-Body Perturbation
Theory
Chapter 8
Covariant Evolution Combined
with Electron Correlation
In Part I, we have considered some standard methods for many-body calculations on

atomic systems. These methods are well developed and can treat certain electron-
correlation effects to essentially all orders of perturbation theory. In Part II, we have
considered three different methods for numerical QED calculations on bound sys-
tems, which have been successfully applied to various problems. All these methods
are, however, in practice limited to one- and two-photon exchange, implying that
electron correlation can only be treated in quite a restricted way. For many systems
the electron correlation is of great importance, and to evaluate the QED effects ac-
curately, it may be necessary to take into account also the combination of QED and
correlational effects, which has not been done previously.
In Part III, we demonstrate that one of the methods presented in the previ-
ous part, the covariant-evolution-operator method, forms a suitable basis for a
combined QED–MBPT procedure.1 This leads to a perturbative procedure that is ul-
timately equivalent to an extension of the relativistically covariant Bethe–Salpeter
equation, valid also in the multireference case and referred to as as the Bethe–
Salpeter–Bloch equation. In this work, we normally use the Coulomb gauge, and
we shall apply the equal-time approximation, discussed in Chap. 6. In Chap. 10, we
illustrate how this procedure can be implemented and give some numerical results.
8.1 General Single-Photon Exchange
In the treatment of single-photon exchange in Chap. 6, the incoming state was

assumed to be unperturbed. We shall now generalize this treatment and allow the in-
coming state to be perturbed, involving particle as well as hole states. As mentioned,
we shall deal particularly with the Coulomb gauge, where the total interaction is ac-
cording to (4.57) separated into a Coulomb and a transverse part (see Fig. 8.1).
I C D ICC C ITC : (8.1)
1
The treatment in Part III is largely based on [3, 6, 7] and the thesis of Daniel Hedendahl [2].

174 8 Covariant Evolution Combined with Electron Correlation
Fig. 8.1 In the Coulomb

gauge, the single-photon 6 6 6 6
x x0 x x0
exchange is separated into
!3
r 6 !4 6s r 6!3 !4 6s
a Coulomb and a transverse z
(Breit) part 1 2 1 - 2
!1
t 6 !2 6u !1 !2 6u
t 6
x0 x00 x0 x00
6 6 6 6
E E
The corresponding single-photon potential is similarly separated into:
Vsp D VC C VT : (8.2)
We start with the transverse part and consider the Coulomb part later.
8.1.1 Transverse Part
The kernel of the transverse part of the single-photon exchange in Coulomb gauge
according to (6.6) is given by:
i SF .x; x1 / i SF .x 0 ; x2 / .i/ITC .x2 ; x1 / i SF .x1 ; x0 / i SF .x2 ; x00 / e.jt1 jCjt2 j/ : (8.3)
The external time dependence is (with the notations in Figs. 8.1 and 8.2) in the
equal-time approximation in analogy with the previous case (6.9):
eit .!3 C!4 "r "s / eit0 .!1 C!2 "t "u / :
As before, we can argue that in the limit ! 0 !1 C !2 D !3 C !4 D E, i.e.,
equal to the initial energy, and the dependence becomes:
eit .E"r "s / eit0 .E"t "u / :
We then have the relation:
UT .t; t0 / D eit .EH0 / MT eit0 .EH0 / ; (8.4)
where MT is the corresponding Feynman amplitude, defined as before (6.11). This

yields:
ZZZZ Z
1 d!1 d!2 d!3 d!4 dz
MT .x; x 0 I x 0 ; x 00 / D iSF .!3 I x; x 1 /
2 2 2 2 2 2
iSF .!4 I x 0 ; x 2 / .i/ITC .zI x 2 ; x 1 / iSF .!1 I x 1 ; x 0 /
iSF .!2 I x 2 ; x 00 /2 .!1 z !3 / 2 .!2 C z !4 /
(8.5)
8.1 General Single-Photon Exchange 175
Fig. 8.2 Time-ordered

r 6O ˙ O ˙6s r 6O ˙ O ˙6s

evolution-operator diagrams 0
x x x x0
for single-photon exchange, !4 6s r 6
transverse part r 6!3 z 2 1 6s
- -z
1 !2 6u t 6 2
t 6!1 6u
x0 x00 x0 x00
t 6O ˙ O ˙6u t 6O ˙ O ˙6u
E E
leaving out the internal space integrations. (The factor of 1/2 is, as before, eliminated
when a specific matrix element is considered.)
After integrations over !2 ; !3 ; !4 , the amplitude becomes:
ZZ
0 1 d!1 dz
MT .x; x ; x 0 ; x 00 / D iSF .!1 zI x; x 1 / iSF .E !1 zI x 0 ; x 2 /
2 2 2
.i/ITC .zI x 2 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .E !1 I x 2 ; x 00 /:
(8.6)
Inserting the expressions for the electron propagator (4.10) and the interaction
(4.46), a specific matrix element becomes:
* ˇ Z Z
ˇ
ˇ d!1 dz 1 1
hrsjMTjabi D rs ˇi
ˇ 2 2 !1 z "r C ir E !1 C z "s C is
Z ˇ +
1 1 2c 2 d fTC ./ ˇˇ
ˇab ;
!1 "t C it E !1 "u C iu z2 c 2 2 C i ˇ
(8.7)
where fTC is the transverse part of the f function in Coulomb gauge (4.60). Integra-
tion over z now yields – in analogy with the treatment in Chap. 6
Z
MT D .i/2 c d fTC ./
times the propagator expressions

1 1 1
C
E "r "s !1 "r .c i /r E !1 "s .c i /s
and

1 1 1
C :
E "t "u !1 "t C it E !1 "u C iu
We have now four combinations that contribute depending on the sign of the
orbital energies (after integration over !1 ):
sgn."t /
sgn."r / ¤ sgn."t / W
"t "r .c i /r
sgn."t /
sgn."s / D sgn."t / W
E "t "s .c i /s
sgn."u /
sgn."u / D sgn."r / W
E "r "u .c i /r
sgn."u /
sgn."u / ¤ sgn."s / W (8.8)
"u "s .c i /s
times .i/.
The Feynman amplitude for the transverse part of the single-photon exchange
now becomes:
MT D .E/ iVT .E/ .E/; (8.9)
where .E/ is the resolvent (2.64). This yields for the present process:
˝ ˇ ˇ ˛ D ˇ ˇ E
i ˇ ˇ 1
rs ˇMT .E/ˇtu D rs ˇ VT .E/ ˇtu ; (8.10)
E "r "s E "t "u
where VT .E/ is now the generalized transverse-photon potential:
ˇZ
˝ ˇ ˇ ˛ ˇ t˙ r
ˇ ˇ ˇ
rs VT .E/ tu D rs ˇ c d fT ./ ˙
C
"t "r ˙ c
ˇ
t˙ s˙ u˙ r˙ u˙ s ˇ
˙ ˙ ˙ ˇ tu : (8.11)
E "t "s c E "r "u c "u "s ˙ c ˇ
Here, t˙ etc. represent projection operators for particle/hole states. The upper
or lower sign should be used consistently in each term, inclusive of the sign in the
front, but all combinations of upper and lower signs in the four term should be used,
corresponding to the 16 time-ordered combinations, shown in Fig. 8.3.
It should be noted that the expression above is valid also for the entire interaction
in any covariant gauge, using the appropriate f function.
We shall now illustrate the potential (8.11) by giving explicit expressions in a
few cases.
No pairs r<0 s<0 r; s < 0

r s -
r s
- 6 6 r s
t u
t <0 u<0 t; s < 0 t; s < 0

s
s
6 t 6
6 6 u
t t
u; r < 0 u; r < 0 t; u < 0 t; u; s < 0

r
r ?u
6 6 u t
?s
u t - u 6
t; u; r < 0 u; r; s < 0 t; r; s < 0 t; u; r; s < 0
t ? t ? ?u
6 ?u t ? 6 ?s -
u? r r ?
?s r ? r ? ?s
6
Fig. 8.3 All 16 time-ordered diagrams corresponding to the transverse single-photon exchange
given by (8.11)
No virtual pairs
r s
-
t 6 6u
E
The potential becomes here:

ˇZ
˝ ˇ ˇ ˛ ˇ
ˇ ˇ
rs VT .E/ tu D rs ˇˇ c d fTC ./
ˇ
1 1 ˇ
C ˇ tu (8.12)
E "r "u c E "t "s c ˇ
and the Feynman amplitude agrees with the previous result (6.16). This agrees with
the result of the evaluation of the corresponding time-ordered diagram according to
the rules of Appendix I.
Single hole in (t)
r s
u
6
t
E
The potential becomes here:
ˇZ ˇ
˝ ˇ ˇ ˛ ˇ 1 1 ˇ
rs ˇVT .E/ˇtu D rs ˇˇ c d fTC ./ C ˇ tu ;
"t "r c E "r "u c ˇ
(8.13)
which can also be expressed
ˇZ
˝ ˇ ˇ ˛ ˇ
rs ˇVT .E/ˇtu D .E "t "u / rs ˇˇ c d fTC ./
ˇ
1 1 ˇ
ˇ tu (8.14)
"t "r c E "r "u c ˇ
and the denominators of the Feynman amplitude become:
1 1 1
: (8.15)
E "r "s "t "r c E "r "u c
This agrees with the evaluation rules of Appendix I. We see here that one of the
resolvents in (8.9) can be singular (“Brown–Ravenhall effect”), which is eliminated
by the potential.
Single hole out (r)

r
6 s
t
u
E
The potential (8.11) becomes:
ˇZ
˝ ˇ ˇ ˛ ˇ
ˇ ˇ
rs VT .E/ tu D rs ˇˇ c d fTC ./
ˇ
1 1 ˇ
C ˇ tu : (8.16)
"t "r C ck E "t "s c ˇ
The denominators can here be expressed as:
1 1
.E "r "s / (8.17)
"t "r C c E "t "s c
and the denominators of the Feynman amplitude becomes:
1 1 1
; (8.18)
"t "r C c E "t "s c E "t "u c
which agrees with the evaluation rules of Appendix I.
Double hole in t,u
r s
t - u
E
The potential (8.11) is here:
ˇZ ˇ
˝ ˇ ˇ ˛ ˇ 1 1 ˇ
rs ˇVT .E/ˇtu D rs ˇˇ c d fTC ./ C ˇ tu (8.19)
"t "r c "u "s c ˇ
and the denominators of the Feynman amplitude become:

1 1 1 1
C : (8.20)
E "r "s "t "r c "u "s c E "t "u
We shall demonstrate explicitly here that this agrees with the evaluation rules of
Appendix I. With one time-ordering t34 > t2 > t > 1 and 1 > t34 > t2 , the
time integrations yield:
Z 1 Z 1 Z 1
id2 t2 id34 t34
.i/ 3
dt2 e dt34 e dt1 eid1 t1 : (8.21)
t34 t2 t2
Together with the alternative time ordering 1 $ 2, this becomes:

1 1 1
C (8.22)
d1234 d34 d1 d2
with the notations of Appendix I, which is identical to the result (8.20). Note that
this is NOT in agreement with the standard Goldstone rules of MBPT [4].
Single hole in and out (t,s)
6 s 6
u s
r 6 u 6
t
t r
E E
The potential (8.11) yields:

ˇZ
˝ ˇ ˇ ˛ ˇ 1 1
rs ˇVT .E/ˇtu D rs ˇˇ c d fTC ./ C
"t "r c E "t "s C c
ˇ
1 1 ˇ
C C ˇ tu :
E "u "r c "u "s C c ˇ
(8.23)
Using the notations of Appendix I:
d1 D "t "r cI d2 D "u "s C cI d3 D "a "t cI
d4 D "b "u C cI d34 D E "t "u I
d134 D E "r "u cI d234 D E "t "s CcI d1234 D E "r "s ;
the bracket above becomes:

1 1 1 1 d1234 d34 1 1
C C D (8.24)
d1 d234 d2 d134 d134 d234 d2 d1
and the denominators of the Feynman amplitude (8.10):

1 1 1 1 1 1
D ; (8.25)
d134 d234 d2 d1 d1 d2 d134 d234
which agrees with the rules of Appendix I.

8.2 General QED Potential 181
No pairs r; s < 0 t; u < 0 t; u; r; s < 0
r s r s t ? ?u
r s
t u
t u t u r ? ?s
Fig. 8.4 Same as Fig. 8.3 for the Coulomb interaction
8.1.2 Coulomb Interaction
The Coulomb part of the interaction is obtained in a similar way (see Fig. 8.4). In
analogy with (8.5), we now have:
ZZ Z
1 d!1 d!3 dz
MC D iSF .!1 / iSF .E0 !1 /
2 2 2 2
.i/ICC iSF .!3 / iSF .E0 !3 /
leaving out the space coordinates. After z integration, using (4.63b), and with the
explicit form of the propagators this leads to:
ˇ ZZ
ˇ d!1 d!3 1 1
hrsjMCjabi D rs ˇˇ i
2 2 !1 "r C ir E0 !1 "s C is
ˇ
1 1 ˇ
VC ˇ ab
!3 "t C it E0 !3 "u C iu ˇ
ˇ ˇ
ˇ i 1 ˇ
D rs ˇˇ˙ VC ˇ ab ; (8.26)
E0 " r " s E0 " t " u ˇ
where VC is the Coulomb interaction (2.109). Here, the plus sign is used if
sgn."t / D sgn."u / D sgn."r / D sgn."s / and the minus sign if sgn."t / D
sgn."u / ¤ sgn."r / D sgn."s /.
8.2 General QED Potential
We shall now see how the potential above can be extended to include also crossing
Coulomb interactions as well as various radiative effects.
8.2.1 Single Photon with Crossed Coulomb Interaction
We start by considering a transverse photon with a crossing Coulomb interaction

(Fig. 8.5), using the Coulomb gauge.
Fig. 8.5 Feynman diagram

representing the exchange of 6 6 6 6
x x0 x x0
a retarded covariant photon r !3 !4 s
with crossing Coulomb 4 !4 s r !3 3
interaction !1 v z w !2
1 ? 2
w !2
1
!1 v z6
2
t !5 3 4 !6 u
!6 u t !5
x0 x00 x0 x00
6 E 6 6 E 6
The Feynman amplitude becomes for the left diagram:

ZZ ZZ ZZ Z
1 d!1 d!2 d!3 d!4 d!5 d!6 dz
M.x; x 0 I x 0 ; x 00 / D
2 2 2 2 2 2 2 2
iSF .!3 I x; x 4 / iSF .!4 I x 0 ; x 2 / iSF .!1 I x 4 ; x 1 /
iSF .!2 I x 2 ; x 3 / iSF .!5 I x 1 ; x 0 / iSF .!6 I x 3 ; x 00 /
.i/ITC .zI x 4 ; x 3 / .i/VC .x 2 ; x 1 /
2 .!1 z !3 / 2 .!6 C z !2 /
2 .E !5 !6 / 2 .!1 C!4 !5 !2 /: (8.27)
Integrations over !3 ; !4 ; !5 ; !6 lead in the adiabatic limit to
!3 D !1 z; !4 D E !1 C z; !5 D E !2 C z; !6 D !2 z
and to
ZZ
1 d!1 dz
M.x; x 0 I x 0 ; x 00 / D iSF .!1 zI x; x 4 / iSF .E !1 C zI x 0 ; x 2 /
2 2 2
iSF .!1 I x 4 ; x 1 /iSF .!2 I x 2 ; x 3 /iSF .E !2 C zI x 1 ; x 0 /
iSF .!2 zI x 2 ; x 00 / .i/ITC .zI x 2 ; x 3 / .i/VC .; x 4 ; x 3 /:
(8.28)
More explicitly the electron propagators become:

1 1 1 1
C
.!1 "v C iv / .E "r "s / .!1 z "r C ir / .E !1 C z "s C is /

1 1 1 1
C :
.!2 w C iw / .E "t "u / .E !2 C z "t C it / .!2 z "u C iu /
(8.29)
The integrations over !1 ; !2 lead in analogy with (8.8) to
v˙ r v˙ r
sgn."v / ¤ sgn."r / W ˙ DW ˙
"v z "r i a
v˙ s˙ v˙ s˙
sgn."v / D sgn."s / W ˙ DW ˙
E "v C z "s ˙ i b˙
w˙ t˙ w˙ t˙
sgn."w / D sgn."w / W ˙ DW ˙
E "w C z "t ˙ i c˙
w˙ u w˙ u
sgn."w / ¤ sgn."u / W ˙ DW ˙ (8.30)
"w z "u i d
times .i/2 . Here, the first two terms should be combined with the last two, and the
propagators (8.28) reduce to:

1 1 v˙ r v˙ s˙
˙ ˙
E "r "r E "t "u a b˙

w˙ t˙ w t w˙ u w u˙
˙ ˙ : (8.31)
c˙ c d d˙
Upper or lower sign should be used consistently in all four operators in each
product. We use here the notations:
8 8
ˆ A˙ D "v "r c
ˆ
< ˆ
ˆ a˙ D "v z "r ˙ i
B˙ D E "v "s c <
b˙ D E "v C z "s ˙ i
ˆ C˙ D E "w "t c (8.32)
:̂ ˆ c˙ D E "w C z "u ˙ i
D˙ D "w "u c :̂
d˙ D "w z "u ˙ i:
The photon interaction has one pole in each half-plane, and for the combinations
where the electron propagator poles are in the same half-plane the z integration
leads directly to the replacement a˙ ! A˙ etc. This part then becomes 1=.E "r
"r / 1=.E "t "u / times

v˙ r v˙ s˙ w˙ t˙ w˙ u
VC1 D C C : (8.33)
A B˙ C˙ D
Again, upper or lower sign should be used consistently in all four operators in
each product. Expressing the Feynman amplitude in analogy with (8.10),
i ˇ ˇ 1
rs ˇM.E/ˇtu D rs ˇ VTC .E/ ˇtu ; (8.34)
E H0 E H0
the corresponding part of the potential becomes:

Z
hrsjVTC /1jtui D c d fTC ./ hvsjVCjtwi hrwjVC1jvui : (8.35)
When the electron propagators have one pole in each half-plane, we have to sep-
arate the propagators as before. For instance, the product
v˙ r w u v˙ r w t
D
a c "v z "r i E "w C z "t i
is rewritten as

v˙ r w t 1 1 v˙ r w t 1 1
C ) C ;
aCc a c aCc A C
after z integration, or

v˙ r w t 1 1
C :
E "w "t C "v "r "v "r ˙ c E "w "t ˙ c
Similarly,
v˙ r w u˙ v˙ r w u˙
D
a d˙ "v z "r i "w z "u ˙ i
is rewritten as

v˙ r w u ˙ 1 1 v˙ r w u ˙ 1 1
) :
ad a d˙ ad A D˙
But A D˙ D a d ; so this can also be written as:
v˙ r w u ˙
:
A D˙
All similar combinations lead to

v˙ r w t 1 1 v˙ r w u˙ v˙ s˙ w t
VC 2 D C
aCc A C A D˙ B˙ C

v˙ s˙ w u˙ 1 1
C ; (8.36)
bCd B˙ D˙
where upper and lower sings are used consistently in all four operators in each term.
The notations are defined in (8.32). This complete expression is quite complicated,
particularly due to the denominator a C c.
Equations (8.33) and (8.36) represent the complete potential for all 64 time-
ordered diagrams, corresponding to the Feynman diagram in Fig. 8.5:
Z
hrsjVTCjtui D c d fTC ./ hvsjVCjtwi hrwjVC1 C VC 2jvui : (8.37)
We can simplify the results above by assuming that the incoming orbitals t; u are
particle states. Then (8.33) reduces to

vC r vC sC wC tC v r C v s w uC
C C C (8.38)
A BC CC AC B DC
and (8.36) to

v rC wC tC 1 1 vC sC w uC 1 1
C C : (8.39)
aCc AC CC bCd BC DC
With no pairs, we then have:
vC sC wC tC vC sC wC tC
D
BC CC .E "v "s c/.E "w "t c/
with single pair (v)

v rC wC tC 1 1 v rC wC tC
C D
aCc AC CC E "w "t C "v "r

1 1
C
"v "r c E "w "t c
and double pairs (v; w)

v rC w uC v rC w uC
D :
AC DC ."v "r c/."w "u c/
This corresponds to the time-ordered diagrams shown below, and the results are
in agreement with the evaluation rules for time-ordered diagrams. Also here some
diagrams are complicated to evaluate, due to the denominator a C c.
r r s r s r s
s
v w v w
v w
? t
-
t
- 6
w v
t u t u
u u
E E E E
Another, probably more reasonable approximation is to assume that the inter-

mediate states v; w are particle states. Then only the simpler term (8.33) survives,
yielding

vC r vC sC wC tC wC u
VC1C D C C (8.40)
A BC CC D
and the potential

Z
C
hrsjVTC jtui D c d fTC ./ hvsjVCjtwi hrwjVC1Cjvui : (8.41)
This part of the potential can be generated by iterating the pair equation, as discussed
in Chap. 10. This is true also for repeated Coulomb crossings.
8.2.2 Electron Self-Energy and Vertex Correction
Next, we shall see how some radiative effects, previously treated in the S-matrix
formalism (see Sect. 4.6), can be included in the QED potential. Since we are using
the Coulomb gauge, we have to treat the Coulomb and transverse parts separately
as in Sect. 4.6.
We start with the transverse part, illustrated in Fig. 8.6 (left). The kernel is here
(c.f. 4.84):
i SF .x; x2 / i SF .x2 ; x1 / .i/ITC .x2 ; x1 /; (8.42)
where ITC is the transverse part of the interaction (4.59). The Feynman amplitude
becomes in analogy with previous cases (8.4)
ZZ
d!1 d!2 dz
MSE .x/ D iSF .!2 I x; x 2 / iSF .!1 I x 2 ; x 1 / .i/ITC .zI x 2 ; x 1 /
2 2 2
2 ."a !1 z/ 2 .!1 !2 C z/ (8.43)
integrated over the internal space coordinates and with the energy parameters given
in the figure. After integration over the omegas, this becomes:
Z
dz
MSE .x/ iSF ."a I x; x 2 / iSF ."a zI x 2 ; x 1 / .i/ITC .zI x 2 ; x 1 /: (8.44)
2
The matrix element of the evolution operator is:
hrjUSE .t/jai D eit ."a "r / hrtjMSE .x/jtai ; (8.45)
which we can express
˝ ˇ ˇ ˛ eit ."a "r / ˝ ˇˇ ˇ ˛

r ˇUSE .t/ˇa D r i˙."a /ˇa ; (8.46)
"a "r
where ˙."a / is the self-energy operator (4.85)

ˇZ ˇ
ˇ dz ˇ
hrj˙."a /jai D rt ˇˇ iSF ."a zI x 2 ; x 1 / ITC .zI x 2 ; x 1 /ˇˇ ta : (8.47)
2
x x
r !2
Fig. 8.6 Diagram
2 r
representing the transverse t
and Coulomb parts of the t !1 z 2 1
first-order self-energy of a
bound electron in the 1 a
covariant-evolution-operator a
formalism (c.f. Fig. 4.9)
Inserting the explicit expressions for the propagators, yields

ˇZ ˇ
ˇ c d fTC ./ ˇ
hrj˙."a /jaiTrans D rt ˇˇ ˇ ta (8.48)
"a "t .c i/t ˇ
consistent with the diagonal S-matrix (4.89).

It should be noted that the self-energy is diagonal in energy in the S-matrix for-
mulation, due to the energy conservation, while also nondiagonal parts will appear
in the covariant-evolution formulation. As we shall see, only the diagonal part is
divergent and has to be renormalized, as will be discussed in Chap. 12.
The Coulomb part of the self-energy (Fig. 8.6 right) becomes in analogy with the
S-matrix result (4.95)
ˇ Z ˇ
1 ˇ e2 2 d sin r12 ˇˇ
hrj˙."a /jaiCoul D ˇ
rt ˇ 2
2 4 0 r12 2 ˇ ta
1
D sgn."t / hrt jVC j tai (8.49)
2
with summation over positive- as well as negative-energy states.
8.2.2.1 General Two-Electron Self-Energy
We consider now a general self-energy operator (transverse part) in analogy with the
general single-photon exchange in Sect. 8.1, illustrated in Fig. 8.7. (The Coulomb
part can be treated similarly.) The kernel is now
i SF .x; x2 / i SF .x2 ; x1 / .i/ITC .x2 ; x1 / i SF .x1 ; x0 / i SF .x 0 ; x00 / e.jt1 jCjt2 j/

(8.50)
x x0
r 6!3
2
v 6 z !2 6u
Fig. 8.7 General 1
two-electron self-energy !1
t 6
with incoming and outgoing x0 x00
electron propagators E
and the Feynman amplitude

ZZZZ Z
0 d!1 d!2 d!3 d!4 dz
MSE .x; x I x 0 ; x 00 / D iSF .!3 I x; x 2 /
2 2 2 2 2
iSF .!4 I x 2 ; x 1 / iSF .!1 I x 1 ; x 0 / iSF .!2 I x 0 ; x 00 /
.i/ITC .zI x 2 ; x 1 / 2 .!1 z !4 /
2 .!4 C z !3 / 2 .E !1 !2 /: (8.51)
After integrations over !2 ; !3 ; !4 , this becomes

ZZ
0 d!1 dz
MSE .x; x ; x 0 ; x 00 / D iSF .!1 I x; x 2 / iSF .!1 zI x 2 ; x 1 /
2 2
.i/ITC .zI x 2 ; x 1 /iSF .!1 I x 1 ; x 0 /iSF .E !1 I x 0 ; x 00 /
(8.52)
as before, leaving out the internal integrations. Integration over z now yields
Z
MSE D .i/ 2
c d fTC ./
times the propagator expressions

1 1 1

"r "v .i /v !1 "r C ir !1 "v .c i /v
and

1 1 1
C : (8.53)
E "t "u !1 "t C it E !1 "u C iu
The integration over !1 yields another factor of i, and this leads in analogy with
(8.10) to:
˝ ˇ ˇ ˛ i 1
ruˇMSE .E/ˇtu D hru jVSE .E/ j tui ; (8.54)
"r "v .c i /v E "t "u
where VSE .E/ is the potential
ˇZ
˝ ˇ ˇ ˛ ˇ t˙ r r˙ u ˙
rs VSE .E/ tu D rs ˇˇ c d fTC ./ ˙
ˇ ˇ ˙
"t "r E "r "u
ˇ
t˙ v v˙ u ˙ ˇ
ˇ tu : (8.55)
"t "v ˙ c E "u "v c ˇ
If all states are particle states, we find that the bracket above becomes:
1 1 "r "v c
D
E "r "u E "u "v c .E "r "u /.E "u "v c/
and the Feynman amplitude
i 1
; (8.56)
.E "r "u /.E "u "v c/ E "t "u
in agreement with the evaluation rules for time-ordered diagrams, derived in

Appendix I.
Next, we consider some specific cases with virtual holes, and, as before, we apply
the potential to a Coulomb interaction.
Hole out (r) (remaining ones particle states)
r
6 r v
v
6u
t 6 t 6 6u
E E
The Feynman amplitude (8.54) becomes:

i 1 1 1
: (8.57)
"r "v c "t "r E "u "v ck E "t "u
This corresponds to the two time-ordered diagrams in the marginal.
Hole in (t)
r
6u r 6 6u
v
6
t
t v
6
E
E
The Feynman amplitude becomes:

i 1 1 1
C C
"r "v c "t "r E "r "u "t "v c

1 1
; (8.58)
E "u "v ck E "t "u
which can also be expressed as

1 1 1
i C :
."r "t /."t "v c/ .E "r "u /.E "u "v c/ E "t "u
(8.59)
After some additional algebra, this can be shown to be identical to

i 1 1
C (8.60)
."t "v c/.E "r "u / E "u "v c "t "r
corresponding to the time-ordered diagrams in the marginal.
8.2.2.2 General Vertex Correction
The general vertex correction (transverse part), illustrated in Fig. 8.8, leads to
ZZ Z
d!1 d!1 dz
MVx .x; x 0 ; x 0 ; x 00 / D iSF .!3 I x; x 2 / iSF .!3 zI x 2 ; x 3 /
2 2 2
iSF .E !3 I x 0 ; x 4 /.i/ITC .zI x 2 ; x 1 /
.i/VC .x 4 ; x 3 / iSF .!1 zI x 3 ; x 1 /
iSF .!1 I x 1 ; x 0 / iSF .E !1 I x 4 ; x 00 /: (8.61)
x x0
r 6!3 !4 6s
2
w 6
3 4
v 6 !2 6u
1
Fig. 8.8 General vertex t 6!1
correction with incoming and x0 x00
outgoing electron propagators E
More explicitly, the electron propagators become:

1 1 1 1
C
E "r "s !3 "w z iw !3 "r C ir E !3 "s C is
times

1 1 1 1
C : (8.62)
E "t "u !1 "v z iv !1 "t C it E !1 "u C iu
If the energies of the orbitals v and w have the same sign, then the integration over
z leads to: Z
MVx D i c d fTC ./
times

1 1 1 1
C
E "r "s !3 "w .c i /w !3 "r C ir E !3 "s C is
and

1 1 1 1
C (8.63)
E "t "u !1 "v .c i /v !1 "t C it E !1 "u C iu
and to the Feynman amplitude, in analogy with (8.54):
i ˇ ˇ 1
rs ˇMVx .E/ˇtu D rs ˇ VVx .E/ ˇtu ; (8.64)
E "r "s E "t "u
where VVx .E/ is the potential

ˇ Z
˝ ˇ ˇ ˛ ˇ ˇ ˛˝ ˇ
ˇ ˇ ˇ ˇ ˇ
rs VVx .E/ tu D ws ˇVC vu rv c d fTC ./

r˙ w s˙ w˙
˙ ˙
"r "w ˙ ck E "s "w c
ˇ
t˙ v v˙ u ˙ ˇ
˙ ˙ ˇ wt : (8.65)
"t "v ˙ c E "u "v c ˇ
If the orbitals v and w are of different kind (particle or hole), the evaluation becomes
more complicated. This case is expected to be less important.
If all states are particle states, we find that the brackets above become:
1 1
;
E "s "w c E "u "v c
in agreement with the evaluation rules for time-ordered diagrams. If r is a hole state
and the others particle states, we have instead

1 1 1
C
"r "w c E "s "w c E "u "v c
E "r "s 1
D :
."r "w c/.E "s "w c/ E "u "v c
This leads to the denominators of the Feynman amplitude (8.64)

1

."r "w c/.E "s "w c/.E "u "v c/.E "t "u /
and corresponds to the time-ordered diagram.
r
w6
6s
v 6
t 6 6u
8.2.3 Vertex Correction with Further Coulomb Iterations
The Coulomb interactions of the vertex correction can be iterated before the
photon interaction is closed, in the same way as for the retarded photon with
crossed Coulomb, treated above, leading to diagrams of the type shown in Fig. 8.9.
Assuming that the intermediate states v; w, as well as the states between the
Coulomb interactions are particle states, the corresponding analytical expression is
obtained from (8.64) by replacing the matrix element hwsjVCjvui by:
jxyi hxyj
hwsjVC VCjvui :
E "x "y c
r 6
2
w 6
6s
x 6 6y
v 6
1 6u
Fig. 8.9 Vertex correction t 6
with double Coulomb
interactions E
8.3 Unification of the MBPT and QED Procedures 193
6 6 6
6 6 6 Vsp 6 6 6
V QED 6 6
D - C C
6 6 6C
6 6
6 6 6 6 6 6
6 6 6
Fig. 8.10 Feynman diagram representing the “QED potential,” V QED , in (8.66). The first diagram
on the r.h.s. includes the Coulomb potential
8.2.4 General Two-Body Potential
We can now form a general “two-body QED potential” by adding the contributions
derived above:
V QED D Vsp C VTC C VSE C VVx ; (8.66)
where Vsp , as before, represents the combined Coulomb and transverse-photon ex-
change (8.2). This is illustrated in Fig. 8.10. Here, particle as well as holes are
allowed in and out.
8.3 Unification of the MBPT and QED Procedures: Connection

to Bethe–Salpeter Equation
We shall now see how the general QED potentials, derived above by means of
the field-theoretical Green’s operator, can be combined with the standard MBPT
procedure, leading to a unified MBPT–QED procedure. The procedure is valid for
an arbitrary (quasi-degenerate) model space and equivalent to an extension of the
standard Bethe–Salpeter equation, referred to as the Bethe–Salpeter–Bloch equa-
tion, briefly mentioned in Sect. 6.9 and further discussed in the next chapter (see
also [5]). The procedure is also applicable to systems with more than two electrons,
as will be briefly discussed at the end of this chapter.
8.3.1 MBPT–QED Procedure
The general potentials derived above – with possible hole states on the in- and
outgoing lines – cannot be used iteratively in the way discussed in Chap. 6. There-
fore, it cannot be used directly in a Bloch equation, like that in (6.106). For that
purpose, we shall insert one extra Coulomb interaction, when holes are present,
leading to the replacements
V QED ) V QED C V QED Q .E/VC C VC Q .E/V

QED
C VC Q .E/V
QED
Q .E/VC
(8.67)
6 6
6 6 6 6 6 6
V QED 6 6 V QED
V QED 6VP 6
6 6) + 6VP 6 + 6VP 6 +
VC 6VP 6
6 6
6 6 6 6 6 6
6 6
Fig. 8.11 Illustration of the modified potential (8.67), which can be iterated. It has only positive-
energy states in and out and is free from the Brown–Ravenhall effect
6 6 6 6
6 6 V QED ı ˝ QED
˝ QED ıE
D 6 6C 6Q 6 C 6 P 6
QED
˝ QED Veff
6 6
6 6 6 6
Fig. 8.12 Graphical representation of the single-photon Bloch equation (8.68). The last diagram
represents the “folded” term, i.e., the last term of the equation. This equation can be compared with
the Bethe–Salpeter equation in Fig. 9.4, valid only in the single-reference case, where there is no
folded contribution. The order-by-order expansion of this equation is illustrated in Fig. 8.13
illustrated in Fig. 8.11. This potential has only particle states (positive energy) in
and out, and can therefore be used iteratively in a Bloch equation.
When we have a single negative-energy state in the output, we can have a
vanishing denominator of the final resolvent, which leads to a singularity of the
Brown–Ravenhall type [1]. As demonstrated, though, at the beginning of this chap-
ter, such singularities cancel when combined with the general potential. But then it
is of vital importance that the potential and resolvent appear in “matching pairs.”
This will always be the case when the modified potential (8.67) is applied.
Inserting the modified potential (8.67) into the Bloch equation (6.106) leads to
ı ˝ QED QED
˝ QED D 1 C QV
QED
˝ QED C Veff ; (8.68)
ıE
QED
where Veff D P V QED .Heff /˝ QED P is the corresponding effective interaction
(6.139). In the last folded term only the last interaction, with the corresponding re-
solvent, is differentiated [see (6.105)]. The modified potential (8.67) is here regarded
as a single unit. This equation is illustrated graphically by the Dyson type of equa-
tion in Fig. 8.12. The iterative expansion of the equation is displayed in Fig. 8.13.
Solving the equation iteratively is equivalent to solving the corresponding version
of the Bethe–Salpeter–Bloch equation [see (6.140) and (9.30)].
The potential discussed above represents the dominating part of the QED ef-
fects. In order to get further, also irreducible combinations of transverse interactions
8.3 Unification of the MBPT and QED Procedures 195
6 6 6
˝ QED 6 6
V QED 6
D 6 6C C 6 6C + folded
6 6 6 6 6 6
Fig. 8.13 Graphical representation of the order-by-order expansion of the Bloch equation in
Fig. 8.12
6 6
6 6 6 6 6 6
6
V QED 6
6
˝ QED 6 6˝I 6 6 6 6 6
D C6 6+ 6 6+ 6 +..+ 6 P 6
6 6 6 6 6
6 6 6 6 6 6 6
6
Fig. 8.14 Graphical representation of the Bloch equation (8.70), where a standard pair function
(˝I ) is combined with a QED potential
should be included (see Fig. 6.6). Formally, we can express the corresponding Bloch
equation:
ı ˝ QED QED
˝ QED D 1 C QV
QED
˝ QED C Veff ; (8.69)
ıE
where V QED is the QED potential, based on the generalized multiphoton poten-
QED
tial, used previously (Fig. 6.6), and Veff is the corresponding effective interaction
(6.139). This corresponds to the full Bethe–Salpeter–Bloch equation (without sin-
gles). For the time being, though, it does not seem feasible to go beyond a single
transverse photon. However, the two-photon exchange can be approximated by in-
cluding one retarded and one instantaneous transverse (Breit) interaction.
The potential (8.67) can also be combined with standard pair functions without
virtual pairs (Fig. 2.6). This leads to the Bloch equation
ı ˝ QED QED
˝ QED D ˝I C QV
QED
˝ QED C Veff (8.70)
ıE
illustrated in Fig. 8.14 (and analogously in the generalized case). This implies that
the Coulomb interaction is iterated to much higher order than the transverse in-
teraction. But since the Coulomb interaction normally dominated heavily over the
transverse interaction, this procedure usually represents a much faster way of gen-
erating a perturbative scheme than that represented by (8.68) and Fig. 8.12.
In the next section, we shall describe how the QED potential (8.67) can be used
in a coupled-cluster expansion, in analogy with the standard procedure of MBPT,
described in Sect. 2.5. Then also single-particle effects can be included in a system-
atic way, and the procedure would, in principle, be fully equivalent to the complete
Bethe–Salpeter–Bloch equation with singles, applicable also to open-shell sys-

tems. This approach will also make it possible to apply the procedure to more than
two electrons.
In the next chapter, we analyze the Bethe–Salpeter and the Bethe–Salpeter–Bloch
equations further. In Chap. 10, we discuss how the iterative procedure discussed here
with a single transverse photon can be implemented and give some in numerical
illustrations. The renormalization procedure is discussed in Chap. 12.
8.4 Coupled-Cluster-QED Expansion
With the interactions derived above, we can construct an effective QED-Coupled-

Cluster procedure in analogy with that used in standard MBPT, described in
Sect. 2.5 (see [7]). Considering the singles-and-doubles approximation (2.105)
S D S1 C S2 ; (8.71)
the MBPT/CC equations are illustrated in Fig. 2.8. In order to obtain the corre-
sponding equations with the covariant potential (6.6), we make the replacements
illustrated in Fig. 8.15, which leads to the equations illustrated in Fig. 8.16.
The CC-QED procedure can also be applied to systems with more than two elec-
trons. For instance, if we consider the simple approximation (2.101)
1
˝ D 1 C S2 C fS22 g;
2
then we will have in addition to the pair function also the coupled-cluster term,
illustrated in Fig. 8.17 (left). Here, one or both of the pair functions can be replaced
by the QED pair function in Fig. 8.12 (right) to insert QED effects on this level. In
addition, of course, single-particle clusters can be included, as in the two-particle
case discussed above (Fig. 8.15).
6 6 6
6 6 6 6
) x = + + 6
? + 6 + 6 +
6
6 6 6 6
6 6 6
6 6 6 6 6 6 6 6
) )
6 6 6 6 6 6 6 6
Fig. 8.15 Replacements to be made in the CC equations in Fig. 2.8 to generate the corresponding
CC-QED equations (c.f. Figs. 6.4 and 6.5). The wavy line in the second row represents the modified
potential (8.67) with only particle states in and out
8.4 Coupled-Cluster-QED Expansion 197
6 6 6
S1 W = 6
+
x
+
x
+
6
x 6 6
6
6 6
6 6 6
+ + + 6
6 6
W1
6 6 6
6 6 6 6 6 6
6 6 x
S2 W = 6 6+ 6 6 + 6 6+ 6 6 +
6 6 6 6 6 6 6 6
6 6 6 6
6 6 6 6
6 6 + 6 6 + + + +
6 6 6 6 6 6 6 6
6 6 6 6 6 6
6 6 + 6 6 + 6 6
W2 W1 W2
Fig. 8.16 Diagrammatic representation of the QED-coupled-cluster equations for the operators S1
and S2 . The second diagram in the second row and the diagrams in the fourth row are examples
of coupled-cluster diagrams. The last diagram in the second row and the three diagrams in the last
row represent folded terms (c.f. the corresponding standard CC equations in Fig. 2.8)
6 6 6 6 6 6 6 6 6 6 6 6
6 6 6 6 6 6 6 6 6 6 6 6
Fig. 8.17 Diagrammatic representation of the QED-coupled-cluster term 12 S22 with standard pair
functions (left) and one and two inserted QED pair function, defined in Fig. 8.12, (right)
We can summarize the results obtained here in the following way:

When all one- and two-particle effects are included, the MBPT–QED proce-
dure is fully compatible with the two-particle Bethe–Salpeter(–Bloch) equation
– including singles.
The advantage of the MBPT–QED procedure is – thanks to the complete com-
patibility with the standard MBPT procedure – that the QED potentials need
to be included only in cases where the effect is expected to be sufficiently
important.
The procedure described here is based on the use of the Coulomb gauge (6.41),
and therefore not strictly covariant. As mentioned, however, in practice, it is equiv-
alent to a fully covariant procedure, and, furthermore, it seems to be the only
feasible way for the time being to treat effects beyond two-photon exchange in a
systematic fashion.
References
1. Brown, G.E., Ravenhall, D.G.: On the Interaction of Two electrons. Proc. R. Soc. London, Ser.
A 208, 552–59 (1951)
2. Hedendahl, D.: Towards a Relativistic Covariant Many-Body Perturbation Theory. Ph.D. the-
sis, University of Gothenburg, Gothenburg, Sweden (2010)
Phys. 83, 183–218 (2005)
turbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502 (2006)
7. Lindgren, I., Salomonson, S., Hedendahl, D.: Coupled clusters and quantum electrodynamics.
In: P. Čársky, J. Paldus, J. Pittner (eds.) Recent Progress in Coupled Cluster Methods: Theory
and Applications, pp. 357–374. Springer Verlag (2010)
Chapter 9
The Bethe–Salpeter Equation
In this chapter, we discuss the Bethe–Salpeter equation and its relation to the
procedure we have developed so far. We start by summarizing the original deriva-
tions of the equation by Bethe and Salpeter and by Gell-Mann and Low, which
represented the first rigorous covariant treatments of the bound-state problem. We
demonstrate that this field-theoretical treatment is completely compatible with the
presentation made here. The treatments of Bethe and Salpeter and of Gell-Mann and
Low concern the single-reference situation, while our procedure is more general.
We, later in this chapter, extend the Bethe–Salpeter equation to the multireference
case, which will lead to what we refer to as the Bethe–Salpeter–Bloch equation in
analogy with corresponding equation in MBPT.
9.1 The Original Derivations by the Bethe–Salpeter Equation
The original derivations of the Bethe–Salpeter equation by Salpeter and Bethe [15]
and by Gell-Mann and Low [9] were based on procedures developed in late 1940s
for the relativistic treatment of the scattering of two or more particles by Feyn-
man [7, 8], Schwinger [18, 19], Tomanaga [23] and others, and we here summarize
their derivations.
9.1.1 Derivation by Salpeter and Bethe
Salpeter and Bethe [15] start their derivation from the Feynman formalism of the
scattering problem [7, 8], illustrated in terms of Feynman graphs. A Feynman di-
agram represents in Feynman’s terminology the “amplitude function” or “kernel”
for the scattering process, which in the case of two-particle scattering, denoted
K.3; 4I 1; 2/, is the probability amplitude for one particle propagating from one
space-time point x1 to another x3 and the other particle from space-time x2 to x4 .
For the process involving one irreducible graph G .n/ , i.e., a graph that cannot be

200 9 The Bethe–Salpeter Equation
3 4
3 4 6
5
.n/
5 6
G 8
G .n/ 7
7 8
6 G .m/
1 2
1 2
Fig. 9.1 Examples of Feynman graphs representing scattering amplitudes in Eqs. (9.1) and (9.2)
of the Salpeter–Bethe paper [15]. The first diagram is irreducible, while the second is reducible,
since it can be separated into two allowed diagrams by a horizontal cut
3 4
G
K.3; 4I 1; 2/ = + + +
1 2
Fig. 9.2 Graphical representation of the expansion of the Feynman kernel in terms of irreducible
graphs
separated into two simpler graphs, as illustrated in Fig. 9.1 (left part), the kernel is
given by (in Feynman’s notations)
ZZZZ
K .3; 4I 1; 2/ D i
.n/
d5 d8 KCa .3; 5/KCb .4; 6/
G .n/ .5; 6I 7; 8/ KCa .7; 1/KCb .8; 2/; (9.1)
where KCa ; KCb represent free-particle propagators (positive-energy part). For a

process involving two irreducible graphs, the kernel illustrated in the right part of
the figure becomes:
ZZZZ
K .n;m/ .3; 4I 1; 2/ D i d5 d8 KCa .3; 5/KCb .4; 6/
G .n/ .5; 6I 7; 8/ K .m/.7; 8I 1; 2/: (9.2)
This leads to the sequence illustrated in Fig. 9.2, where G represents the sum of all
irreducible two-particle self-energy graphs. From this, Salpeter and Bethe arrived
at an integral equation for the total kernel
ZZZZ
K.3; 4I 1; 2/ D KCa .3; 1/KCb .4; 2/ i d5 d8 KCa .3; 5/KCb .4; 6/
G .5; 6I 7; 8/ K.7; 8I 1; 2/: (9.3)

9.1 The Original Derivations by the Bethe–Salpeter Equation 201
x x0
x x0 x x0
x2 x20
˙ .x2 ; x20 I x1 ; x10 /
x1 x10
G.x1 ; x10 I x0 ; x00 / = + G.x1 ; x1 I x0 ; x00 /
x0 x00 x0 x00
x0 x00
Fig. 9.3 Graphical representation of the integral equation (9.3) for the Feynman kernel of Salpeter
and Bethe – identical to the Dyson equation for the two-particle Green’s function (Fig. 5.8)
This is the equation for the two-particle Greens function (5.80) in the form of a
Dyson equation, in our notations written as:
ZZZZ
G.x; x 0 I x0 ; x00 / D G0 .x; x 0 I x0 ; x00 / C d4 x1 d4 x2 d4 x10 d4 x20
G0 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / G.x1 ; x10 I x0 ; x00 /

(9.4)
and depicted in Fig. 9.3 (see also Fig. 5.8). Note that the two-particle kernel K in the
terminology of Feynman and Salpeter–Bethe corresponds to our Green’s function
G, and the irreducible interaction G corresponds to our proper self-energy ˙ .
The proper (or irreducible) self-energy is identical to the irreducible two-particle
potential in Fig. 6.6. Furthermore, the electron propagators are in the Feynman–
Salpeter–Bethe treatment free-particle propagators. Note that the intermediate lines
in Fig. 9.3 represent a Green’s function, where the singularities are eliminated.
Salpeter and Bethe then argued that a similar equation could be set up for the
bound-state wave function. Since the free lines of the diagrams in the Feynman
formulation represent free particles, they concluded that the first (inhomogeneous)
term on the right-hand side could not contribute, as the bound-state wave function
cannot contain any free-particle component. This leads in their notations to the ho-
mogeneous equation:
ZZZZ
.3; 4/ D i d5 d8 KCa .3; 5/KCb .4; 6/ G .5; 6I 7; 8/ .7; 8/:
(9.5)
This is the famous Bethe–Salpeter equation. In the Furry picture we use here,
where the basis single-electron states are generated in an external (nuclear) po-
tential, the inhomogeneous term does survive, and the equation becomes in our
notations
ZZZZ
0 0
.x; x / D ˚.x; x / C d4 x1 d4 x2 d4 x10 d4 x20
G0 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / .x1 ; x10 /: (9.6)

x x0 x x0 x x0
x2 x20
˙ .x2 ; x20 I x1 ; x10 /
= x1 x10
+
˚
Fig. 9.4 Graphical representation of the inhomogeneous Bethe–Salpeter equation (9.6). ˙ repre-
sents the proper self-energy, which contains all irreducible interaction graphs and is identical to the
irreducible two-particle potential in Fig. 6.6. This equation can be compared with that represented
in Fig. 8.12, valid also in the multireference case
This is the inhomogeneous Bethe–Salpeter equation we shall use, and it is

graphically depicted in Fig. 9.4.
9.1.2 Derivation by Gell-Mann and Low
The derivation of Gell-Mann and Low [9] starts from the “Feynman two-body ker-
nel,” used in the definition of the Green’s function (5.20) (in their slightly modified
notations):
˝ ˇ ˇ ˛
K.x1 ; x2 I x3 ; x4 / D 0
ˇT Œ O H .x1 / O H .x2 / O .x4 / O .x3 /ˇ 0 : (9.7)
H H
T is the time-ordering operator (2.27) and O H ; O H are the particle-field operators

in the Heisenberg representation. 0 is the vacuum (ground state) of the interacting

system in the Heisenberg picture,j0H i.
In an Appendix of the same paper, Gell-Mann and Low derive a relation between
the interacting ( 0 ) and the noninteracting (˚0 ) vacuum states (both in the interac-
tion picture):
U.0; 1/˚0
c 0 D ; (9.8)
h˚0 jU.0; 1/j˚0 i
which is the famous Gell-Mann–Low theorem (3.29), discussed previously. Here, c

is a normalization constant (equal to unity in the intermediate normalization that we
use). This can be eliminated by considering
h˚0 jU.1; 1/j˚0 i

1Dh 0j 0 i D : (9.9)
c2 h˚0 jU.1; 0/j˚0 i h˚0 jU.0; 1/j˚0 i
9.1 The Original Derivations by the Bethe–Salpeter Equation 203
Inserting the expression (9.8) into the kernel (9.7), using the relation (9.9), yields:
h˚0 jU .1; 0/T Œ O H .x1 / O H .x2 / O H .x4 / O H .x3 /U.0; 1/j˚0 i

K.x1; x2 I x3 ; x4/ D ;
h˚0 jU.1; 1/j˚0 i
(9.10)
which is equivalent to the field-theoretical definition of the Green’s function
G.x1 ; x2 I x3 ; x4 / in (5.20).
Gell-Mann and Low then conclude that expanding the expression above in a
perturbation series leads to the two-body kernel of Feynman in terms of Feynman
diagrams, as we have performed in Chap. 5. This is identical to the expansion given
by Salpeter and Bethe, and hence leads also to the integral equation (9.3). Gell-
Mann and Low then use the same arguments as Salpeter and Bethe to set up the
Bethe–Salpeter equation (9.5) for the wave function. In addition, they argue that
single-particle self-energy parts can easily be included by modifying the single-
particle propagators.
The derivation of Gell-Mann and Low, which starts from the field-theoretical
definition of the Green’s function, has a firm field-theoretical basis. This is true, in
principle, also of the derivation of Salpeter and Bethe, which is based on Feynman
diagrams for scattering of field-theoretical origin.
In Sect. 9.1.3, we shall see how the Bethe–Salpeter equation can be motivated
from the graphical form of the Dyson equation in Fig. 9.3.
9.1.3 Analysis of the Derivations of the Bethe–Salpeter Equation
We can understand the Bethe–Salpeter equation graphically, if we let the Dyson

equation in Fig. 9.3 act on the zeroth-order state, ˚.x0 ; x00 /, which we represent
by two vertical lines without interaction. (The treatment can easily be extended to
the situation, where the model function is a linear combination of straight prod-
ucts.) From the relation (6.7), we see that the electron propagator acting on an
electron-field operator (with space integration) shifts the coordinates of the oper-
ator. Therefore, acting with the zeroth-order Green’s function on the model function
shifts the coordinates of the function according to:
ZZ
˚.x; x 0 / D d3 x 0 d3 x 00 G0 .x; x 0 I x0 ; x00 / ˚.x0 ; x00 /: (9.11)
This is illustrated in Fig. 9.5 (left) and corresponds to the first diagram on the right-
hand side of Fig. 9.4. Similarly, operating with the full Green’s function in Fig. 9.5
on the model function leads to
ZZ
.x; x 0 / D d3 x 0 d3 x 00 G.x; x 0 I x0 ; x00 / ˚.x0 ; x00 / (9.12)
illustrated in Fig. 9.5 (right). Then the entire equation (9.6), illustrated in Fig. 9.4, is
reproduced.
x x0
0 0
x x x x x x0
G0
G
x0 x00 ) x0 x00 )
x0 x00 x0 x00
˚
˚ ˚
Fig. 9.5 Graphical illustration of Eqs. (9.11) and (9.12)
The equation (9.12) is consistent with the definition of the classical Green’s func-
tion (5.1), which propagates a wave function from one space-time point to another
– in our case from one pair of space-time point to another. This equation can also
be expressed as an operator equation
˛ ˛
j .t; t 0 / D G.t; t 0 I t0 ; t00 / j .t0 ; t00 / ; (9.13)
where G is the Green’s operator, introduced in Sect. 6.6. The coordinate represen-
tation of this equation
˛ ˝ ˛ ˛
hx; x 0j .t; t 0 / D x; x 0 jG.t; t 0 I t0 ; t00 /jx 0 ; x 00 hx 0 ; x 00j .t0 ; t00 / (9.14)
is identical to (9.12).
This implies that:
The Green’s function is the coordinate representation of the Green’s operator.
The four-times Green’s operator represents the time propagation of the two-
particle Bethe–Salpeter state vector.
In the equal-time approximation, this is consistent with our previous result (6.50)
and with our conjecture (6.29).
It is of interest to compare the Bethe–Salpeter equation (9.6), depicted in Fig. 9.4,
with the Dyson equation for the combined QED-electron correlation effects in
Fig. 8.12. If in the latter more and more effects are included in the QED potential,
then the Coulomb interactions, represented by the standard pair function, become
insignificant. Then, this equation is identical to the Bethe–Salpeter equation. Solv-
ing the original BS equation iteratively, however, is extremely tedious and often
very slowly converging, due to the dominating Coulomb interaction. As mentioned
in the previous chapter, the QED-correlation equation is expected to be a faster
road to reach the same goal. One- and two-photon exchange in the QED potential
will very likely yield extremely good results, while such effects in the BS equa-
tion will often be quite insufficient (c.f. the discussion about the QED methods in
Part II).
9.2 Quasi-Potential and Effective-Potential Approximations: Single-Reference Case 205
9.2 Quasi-Potential and Effective-Potential Approximations:

Single-Reference Case
In the equal-time approximation, where we equalize the times of the two particles in
the Bethe–Salpeter equation (9.6), we can make a Fourier transformation of it with
a single energy parameter as in the treatment of the single-particle Green’s function
in Sect. 5.2.3. The Q part, falling outside the model space, then leads to:
Q .E/ D Q G0 .E/ .i/˙ .E/ .E/ (9.15)
leaving out the space coordinates and integrations.

Replacing the zeroth-order Green’s function with the resolvent (5.43)
i
G0 .E/ D ; (9.16)
E H0
we obtain
Q.E H0 / .E/ D Q˙ .E/ .E/: (9.17)
If we identify the proper self-energy with the generalized potential (8.11)
V.E/ D ˙ .E/; (9.18)
the equation above leads together with the relation (6.125)
P .H H0 /˝ .E/ D P V.E/ .E/ (9.19)
to
The effective-potential form of the Bethe–Salpeter equation
.E H0 /j i D V.E/j i (9.20)
frequently used in various applications. This equation was also derived above, using
the Green’s operator only (6.126).
The equation (9.20) can also be expressed
Q
j i Dj 0i C V.E/j i ; (9.21)
E H0
where 0 is the model state 0 D P . This is equivalent to the Lippmann–

Schwinger equation [12], frequently used in scattering theory. Formally, the equa-
tion (9.20) can also be expressed in the form of the time-independent Schrödinger
equation
H DE ; (9.22)
where H is the energy-dependent Hamiltonian
H.E/ D H0 C V.E/: (9.23)
Equation (9.20) operates entirely in the restricted Hilbert space with constant
number of photons. This can be related to the equivalent equation (6.32), derived by
means of the Gell-Mann–Low theorem, which operates in the photonic Fock space.
We can then regard the equation above as the projection of the Fock-space equation
onto the restricted space.
9.3 Bethe–Salpeter–Bloch Equation: Multireference Case
We can extend the treatment above to the general multireference case. From the
expression (6.117), using the fact that the Green’s operator at time t D 0 is identical
to the wave operator (6.52), we have in the single-reference case (one-dimensional)
model space
h iˇ ˛
j i D ˝j 0i D 1C Q .E/ V.E/ C Q .E/ V.E/ Q .E/ V .E/ C ˇ 0 ;
(9.24)
wherej 0i is the model state,j 0i D P j i, and
Q
Q .E/ D
E H0
is the reduced resolvent (2.65).

Operating on (9.24) from the left with Q.E H0 / now yields
Q.E H0 /j i D QV.E/j i ; (9.25)
which is identical to the equation (9.17) with the identification (9.18).

For a general multidimensional (quasi-degenerate) model space, we have
similarly
Q.E ˛ H0 /j ˛ i D QV.E ˛ /j ˛ i (9.26)
and
P .E ˛ H0 /j ˛
i D P V.E ˛ /j ˛
i: (9.27)
This leads to
.E ˛ H0 /j ˛
i D V.E ˛ /j ˛
i (9.28)
or in operator form

.Heff H0 /˝P D V.Heff /˝P (9.29)
9.3 Bethe–Salpeter–Bloch Equation: Multireference Case 207
using the notations introduced in Sect. 6.9. But

Heff ˝P D ˝Heff P D ˝H0 P C ˝Veff P;
which yields the commutator relation

˝; H0 P D V.Heff /˝P ˝P Veff P; (9.30)

where according to (6.139) Veff P D P V.Heff /˝P . Here, the energy parameter of

V.Heff / is given by the model-space state to the far right, while the energy param-
eter of ˝ of the folded term depends on the intermediate model-space state (see
footnote in Sect. 6.6). This equation is valid in the general multireference (quasi-
degenerate) situation and represents an extension of the effective-potential form
(9.20) of the Bethe–Salpeter equation. Due to its close resemblance with the stan-
dard Bloch equation of MBPT (2.56), we refer to it as the Bethe–Salpeter–Bloch
equation. This is equivalent to the generalized Bethe–Salpeter equation, derived in
Chap. 6 (6.140).
In analogy with the MBPT treatment in Sect. 2.5, we can separate the BS-Bloch
equation into

˝1 ; H0 P D V.Heff /˝P ˝P Veff .H0 /P ;
linked;1

˝2 ; H0 P D V.Heff /˝P ˝P Veff .H0 /P ; (9.31)
linked;2

etc. It should be noted that the potential operator V.Heff / is an operator or matrix
where each element is an operator/matrix. In the first iteration, we set Heff D H0 and
.1/
in the next iteration Heff D H0 C Veff etc. Continued iterations correspond to the
sum term in the expression (6.96), representing the model-space contributions. The
two-particle BS-Bloch equation above is an extension of the ordinary pair equation,
discussed in Sect. 2.5 (Fig. 2.6).
The Bethe–Salpeter–Bloch equation leads to a perturbation expansion of
Rayleigh–Schrödinger or linked-diagram type, analogous to that of standard MBPT
expansions. It differs from the standard Bloch equation by the fact that the Coulomb
interaction is replaced by all irreducible multiphoton interactions.
Solving the BS-Bloch equation (9.30) is NOT equivalent to solving the single-
state equation for a number of states. The Bloch equation (9.30) leads to a
Rayleigh/Schrödinger/linked-diagram expansion with folded terms that is size
extensive. The single-state equation (9.20), on the other hand, leads to a Brillouin–
Wigner expansion (see footnote in Sect. 2.4), that is not size extensive.
Due to the very complicate form of the potential of the Bethe–Salpeter–Bloch
equation, it is very difficult to handle this equation in its full extent. In the previous
chapters, we have considered a simpler way of achieving essentially the same goal.
9.4 Problems with the Bethe–Salpeter Equation
There are several fundamental problems with the Bethe–Salpeter equation and with
relativistic quantum mechanics in general, as briefly mentioned in the Introduction.
Dyson says in his 1953 paper [6] that this is a subject “full of obscurities and un-
solved problems.” The question concerns the relation between the three-dimensional
and the four-dimensional wave functions. In standard quantum mechanics, the
three-dimensional wave function describes the system at a particular time, while the
four-dimensional two-particle wave function describes the probability amplitude for
finding particle one at a certain position at a certain time and particle two at another
position at another time, etc. The latter view is that of the Bethe–Salpeter equation,
and Dyson establishes a connection between the two views. The main problem is
here the individual times associated with the particles involved, the physical mean-
ing of which is not completely understood. This problem was further analyzed by
Wick [24] and Cutkoski [4] and others. The relative time of the particles leads to
a number of anomalous or spurious states – states that do not have nonrelativistic
counterparts. This problem was analyzed in detail in 1965 by Nakanishi [13],
and the situation was summarized in 1997 in a comprehensive paper by
Namyslowski [14].
The Bethe–Salpeter equation was originally set up for the bound-state problem
involving nucleons, such as the ground state of the deuteron. The equation has
lately been extensively used for scattering problems in quantum chromodynamics,
quark–quark/antiquark scattering. The equation has also been used for a long time
in high-accuracy works on simple atomic systems, such as positronium, muonium,
hydrogen, and helium-like ions. The problems with the BS equation, associated with
the relative time, are most pronounced at strong coupling and assumed to be neg-
ligible in atomic physics, due to the very weak coupling. One important question
is, of course, whether this is true also in the very high accuracy that is achieved in
recent time.
To attack the BS equation directly is very complicated, and for that reason vari-
ous approximations and alternative schemes have been developed. The most obvious
approximation is to eliminate the relative time of the particles, the equal-time ap-
proximation or external-potential approach. The first application of this technique
seems to be have been made in the thesis of Sucher in the late 1950s [20, 21] for
the evaluation of the leading QED corrections to the energy levels of the helium
atom. This work has been extended by Douglas and Kroll [5] and by Drake, Zhang,
and coworkers [25, 26], as will be further discussed in the Chap. 11. Another early
application of an effective-potential approach was that of Grotch and Yennie [11] to
obtain high-order effects of the nuclear recoil on the energy levels of atomic hydro-
gen. They derived an “effective potential” from scattering theory and applied that in
a Schrödinger-like equation. A similar approach was applied to strongly interacting
nucleons in the same year by Gross [10], assuming that one of the particles was “on
the mass shell.” Related techniques have been applied to bound-state QED prob-
lems among others by Caswell and Lepage [2] and by Bodwin et al. [1]. A more
References 209
formal derivation of a “quasipotential” method for scattering as well as bound-state

problems was made by Todorov [22], starting from the Lippmann–Schwinger scat-
tering theory [12].
Several attempts have been made to correct for the equal-time approximation.
Sazdjian [16, 17] has converted the BSE into two equations, one for the relative
time and one eigenvalue equation of Schrödinger type. Connell [3] has developed
a series of approximations, which ultimately are claimed to lead to the exact BSE.
The approaches were primarily intended for strong interactions, but Connell tested
the method on QED problems.
References
1. Boldwin, G.T., Yennie, D.R., Gregorio, M.A.: Recoil effects in the hyperfine structure of QED
bound states. Rev. Mod. Phys. 57, 723–82 (1985)
2. Caswell, W.E., Lepage, G.P.: Reduction of the Bethe-Salpeter equation to an equivalent
Schrödinger equation, with applications. Phys. Rev. A 18, 810–19 (1978)
3. Connell, J.H.: QED test of a Bethe-Salpeter solution method. Phys. Rev. D 43, 1393–1402
(1991)
4. Cutkosky, R.E.: Solutions of the Bethe-Salpeter equation. Phys. Rev. 96, 1135–41 (1954)
6. Dyson, F.J.: The Wave Function of a Relativistic System. Phys. Rev. 91, 1543–50 (1953)
(1949)
(1951)
10. Gross, F.: Three-dimensionsl Covariant Integral Equations for Low-Energy Systems. Phys.
Rev. 186, 1448–62 (1969)
11. Grotch, H., Yennie, D.R.: Effective Potential Model for Calculating Nuclear Corrections to
the Eenergy Levels of Hydrogen. Rev. Mod. Phys. 41, 350–74 (1969)
12. Lippmann, B.A., Schwinger, J.: Variational Principles for Scattering Processes. I. Phys. Rev.
79, 469–80 (1950)
13. Nakanishi, N.: Normalization condition and normal and abnormal solutions of Bethe-Salpeter
equation. Phys. Rev. 138, B1182 (1965)
14. Namyslowski, J.M.: The Relativistic Bound State Wave Function. in Light-Front Quantization
and Non-Perturbative QCD , J.P. Vary and F. Wolz, eds. (International Institute of Theoretical
and Applied Physics, Ames) (1997)
1232–42 (1951)
16. Sazdjian, H.: Relativistiv wave equations for the dynamics of two interacting particles. Phys.
Rev. D 33, 3401–24 (1987)
17. Sazdjian, H.: The connection of two-particle relativistic quantum mechanics with the Bethe-
Salpeter equation. J. Math. Phys. 28, 2618–38 (1987)
18. Schwinger, J.: On quantum electrodynamics and the magnetic moment of the electron. Phys.
Rev. 73, 416 (1948)
19. Schwinger, J.: Quantum electrodynamics I. A covariant formulation. Phys. Rev. 74, 1439
(1948)
Helium. Phys. Rev. 109, 1010–11 (1957)
Michigan
22. Todorov, I.T.: Quasipotential Equation Corresponding to the relativistic Eikonal Approxima-
tion. Phys. Rev. D 3, 2351–56 (1971)
23. Tomanaga, S.: On Infinite Field Reactions in Quantum Field Theory. Phys. Rev. 74, 224–25
(1948)
24. Wick, G.C.: Properties of Bethe-Salpeter Wave Functions. Phys. Rev. 96, 1124–34 (1954)
26. Zhang, T., Drake, G.W.F.: A rigorous treatment of O.˛ 6 mc 2 / QED corrections to the fine
structure splitting of helium. J. Phys. B 27, L311–16 (1994)
Chapter 10
Implementation of the MBPT–QED Procedure
with Numerical Results
In this chapter, we shall see how the combined covariant-evolution-QED approach

developed in the previous chapters can be implemented numerically. In principle,
this is equivalent to solving the complete Bethe–Salpeter equation perturbatively,
but in practice, of course, approximations have to be made. We shall consistently
work in the Coulomb gauge.1
We shall restrict ourselves here to the exchange of a single transverse photon
together with a number of Coulomb interactions. We shall first apply the procedure
in the no-pair case, and later a different procedure in the presence of virtual pairs
will be applied. We work in the photonic Fock space, and initially we shall derive
some relations for that space.
10.1 The Fock-Space Bloch Equation
We have seen earlier that with the perturbation (6.35)
H.x/ D H.t; x/ D O .x/ ec˛ A .x/ O .x/; (10.1)
the wave function partly lies in an extended photonic Fock space, where the number
of photons is no longer constant. According to the Gell-Mann–Low theorem, we
have a Schrödinger-like equation (6.32) in that space
.H0 C VF /j ˛
i D E ˛j ˛
i; (10.2)
where VF is the perturbation (6.36) with the Coulomb and the transverse parts,
VF D VC C vT . We shall demonstrate below that, for single-photon exchange, this
leads to a perturbation that is time-independent in the Schrödinger picture, which is
a requirement for the GML theorem. Furthermore, in working in the extended space
1
This chapter is mainly based on [3, 5] and, in particular, on the thesis of Daniel Hedendahl [1].

212 10 Implementation of the MBPT–QED Procedure with Numerical Results
with uncontracted perturbations, it is necessary to include in the model Hamiltonian

(H0 ) also the energy operator of the photon field (6.38)
H0 D H0 C c ai .k/ ai .k/; (10.3)
where D jkj.
The wave operator is, as before, given by the Green’s operator at t D 0 (6.52),
which may now contain uncontracted photon terms
j ˛
i D ˝j ˛
0 i: (10.4)
˛
j 0˛ D P j ˛ i is the corresponding model state, which lies entirely within the re-
stricted space with no uncontracted photons.
From the GML equation (10.2), it can be shown in the same way as for the
restricted space that the standard Bloch equation (2.56) is still valid also in the ex-
tended space
˝; H0 P D VF ˝ ˝Veff P: (10.5)
The effective interaction is here given by Veff D P VF ˝P (6.61). The equation is

formally the same as in the standard MBPT (2.56), but the operators involved now
have somewhat different interpretation.
The expressions for single transverse-photon exchange are given by (8.11). In
the Coulomb gauge, these expressions involve the functions fTC , given by (4.60)

e2 sin.r12 / sin.r12 /
fTC ./ D ˛1 ˛2 .˛ 1 r 1 / .˛2 r 2 / : (10.6)
4 2 0 r12 2 r12
By means of the expansion theorem
X 1
sin r12
D .2l C 1/jl .r1 /jl .r2 / C l .1/ C l .2/; (10.7)
r12
lD0
where jl .r/ are radial Bessel functions and C l vector spherical harmonics [4],
we can express the function fTC as a sum of products of single-electron operators
[5, Appendix B]
1 h
X i
fTC ./ D VGl .r1 / VGl .r2 / C Vsrl .r1 / Vsrl .r2 / ; (10.8)
lD0
10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs 213
where
e p
VGl .r/ D p .2l C 1/ jl .r/ ˛C l ; (10.9a)
2 0
r hp
e
Vsrl .r/ D p .l C 1/.2l C 3/ jlC1 .r/ f˛C lC1 gl
2 0 2l C 1
p i
C l.2l 1/ jl1 .r/ f˛C l1 gl : (10.9b)
In (10.6), the first term represents the Gaunt interaction and the second term the
scalar retardation, which together form the Breit interaction (see Appendix F.2.2).
Each term in the expansion – which are all time independent in the Schrödinger
picture – will together with the Coulomb interaction (VC D e 2 =4 r12 ) form the
(time-independent) perturbation
VF D VC C VGl .r/ C Vsrl .r/: (10.10)
10.2 Single-Photon Potential in Coulomb Gauge:

No Virtual Pairs
We consider first the case where no virtual pairs are present. Inserting the perturba-
tion (10.10) into the Bloch equation (10.5) yields
˝; H0 P D VC C V l ˝P ˝ Veff ; (10.11)
where we use V l as a short-hand notation for the Gaunt and scalar-retardation parts
in (10.10). We consider first a number of instantaneous Coulomb interaction, form-
ing a standard pair function (2.97), including also the zeroth order,
˝I PE D 1 C Q .E/I
Pair
PE : (10.12)
This includes also the folds and is represented by the first diagram in Fig. 10.1. Then
we can perturb this by one of the V l terms, representing part of the transverse Breit
interaction, leading to the equation
˝ l ; H0 PE D V l ˝I PE ˝ l PE 0 I Pair PE (10.13)
or
E E
E h0 .1/ h0 .2/ c j˝ab
l
D V l j˝Iab i j˝cd
l
hcd jI Pairjabi ; (10.14)
r s
r
r s
r -
u u - 6u
t 6 t 6
t u t u
c 6 E0 t c 6 E0
6d 6d
a 6 6b a 6 6b a 6 6b a 6 6b a 6 6b
E E E E E
Fig. 10.1 Expansion of the Bloch equation (10.11) in the Fock space, no virtual pairs, leading to
single-photon exchange, including folded diagrams
(E; E 0 are here the energies of the unperturbed states jabi and jcd i, respectively).
This equation has the solution
E D ˇ Vl ˇ E
ˇ ˇ
hruj˝ab
l
D ruˇ ˇ˝Iab
E "r "u c
D ˇ Vl ˇ E˝ ˇ ˇ ˛
ˇ ˇ
ruˇ 0
ˇ˝ Icd cd ˇI Pairˇab ; (10.15)
.E "r "u c/.E "r "u c/
where ˝Ipq represents a pair function (10.12) starting from the statejpqi. Here, the
first term is represented by the second diagrams in the Fig. 10.1 and the last term by
the third “folded” diagram. The double bar indicates here the double denominator,
which yields the first-order derivative (difference ratio) of the potential.
By adding a second perturbation V l , we can complete the single-photon ex-
change between the electrons, which corresponds to solving the pair equation
E h0 .1/ h0 .2/ ˝sp PE D V l ˝ l PE ˝sp PE 0 I Pair PE : (10.16)
This yields
hrsjV l jrui hrujV l j˝Iab i

hrsj˝spjabi D
.E "r "s /.E "r "u c/
hrsjV l jrui hrujV l j˝Icd i

.E "r "s /.E "r "u c/E 0 "r "u c/
˝ ˇ ˇ ˛ hrsj˝spjcd i hcd jI Pairjabi
cd ˇI Pairˇab :
E "r "s
(10.17)
This is illustrated by the last two diagrams of Fig. 10.1 (except for the final folded
contribution).
10.2 Single-Photon Potential in Coulomb Gauge: No Virtual Pairs 215
Summing over and l, including the Gaunt as well as the scalar-retardation

parts and considering photon emission from both electrons, we see that the result is
in agreement with the Bloch equation (8.68) to first order
Q Vsp /
ı.
˝sp PE D Q .E/Vsp ˝I PE C PE 0 I Pair PE
ıE
ıVsp
D Q .E/Vsp ˝I PE C Q .E/ PE 0 I Pair PE
ıE
Q .E/ Q .E 0 /Vsp PE 0 I Pair PE (10.18)
with (the first part of) the potential (6.16),

ˇZ 1
ˇ
hrsjVsp .E0 /jtui D rs ˇˇ c d fTC ./
0
ˇ
1 1 ˇ
C ˇ tu (10.19)
E0 "r "u c E0 "t "s c ˇ
(the second part of the potential is generated by emitting a photon from the second
electron.)
We see that
the energy dependence of the potential is generated by an energy denominator
and the energy derivative (difference ratio) by a folded contribution (double
denominator), when operating in the extended space (10.15).
After the first interaction V1l , it is possible to add one or more Coulomb inter-
actions, before closing the photon, corresponding to the first diagram in Fig. 10.2.
This can be achieved by another iteration of the pair equation (10.14)
E h0 .1/ h0 .2/ c ˝ l PE D VC ˝ l PE ˝ l PE 0 I Pair PE : (10.20)
Instead of closing the uncontracted perturbation on the other electron, it can be
closed on the same electron as the emission occurred from. This leads to a self-
energy interaction, represented by the first diagram in Fig. 10.3. This is a radiative
effect (see Sect. 2.6), which is infinite and has to be renormalized, as discussed in
Chap. 12.
6 6
r0 6 6u0 6
r 6 -
t 6 6u
6
a 6 6b
E 6 6
Fig. 10.2 Crossing Coulomb interactions before closing the retarded interaction (left). Continu-
ing the process leads to the single transverse-photon exchange combined with high-order electron
correlation, including crossing Coulomb interactions (right)
r
k’ k’
u r r s
v k k
v
u
t t t u
a 6 6b a b a b
E E E
Fig. 10.3 Second-order contribution to the wave operator in the extended Fock space
A second perturbation V l can also be applied without contracting the first one,
leading to diagrams indicated by the second and third diagrams in Fig. 10.3. Closing
these photons lead to irreducible two-photon interactions, discussed in Sect. 8.2.
10.3 Single-Photon Exchange: Virtual Pairs
10.3.1 Illustration
The iterative procedure of the previous section works well in the no-pair situation,
when the repeated single-photon exchange leads to reducible diagrams of ladder
type, which means that they can in time-ordered form be separated into legitimate
diagrams by horizontal cuts.
In the presence of virtual pairs, we have to use a different procedure. As we
have seen above (Sect. 8.3), we have to combine the general potential with Coulomb
interactions (8.67) to be able to treat the potential in an iterative process. This will at
the same time eliminate the so-called Brown–Ravenhall effect of vanishing energy
denominators.
This potential (8.67) can be used directly in the Bloch equation (8.68). In prin-
ciple, we can use the corresponding full potential with QED effects (8.67), but for
simplicity we shall consider only the pure single-photon part.
We use pair functions (10.12) as input, and perturbing this with the potential
(8.67) leads in next order to
QED
ı. QV /
˝sp P D QV
QED
˝I C ˝I PI Pair P: (10.21)
ıE
For the evaluation we use, as before, the expansion (10.8)

1
˝ ˇ C ˇ ˛ X
rs ˇfT ./ˇtu D hsjVGl .r2 /jui hrjVGl .r1 /jti
lD0
C hsjVsrl .r2 /jui hrjVsrl .r1 /jti : (10.22)

10.3 Single-Photon Exchange: Virtual Pairs 217
r 6 6s
-
f t 6VP
Vsp
Usp 6u
VC
6 6
˝Iab
a 6 6b
As an illustration we consider the second term in Fig. 8.11 (shown above), when
there is a single hole (t) [c.f. (8.13)]. Then (10.21) becomes

ı. Q Usp /
˝sp P D Q Usp ˝I C ˝I PI Pair
P; (10.23)
ıE
where Usp D Vsp Q VC represents the transverse photon (Vsp ) with a Coulomb inter-
action. Here, only the first and the third terms of the potential Vsp (8.11) are relevant,
yielding for the first term
" #
jrC sC i hsC jV l juC i hrC jV l jt i hsC jV l juC i hrC jV l jt i
C
E "r "s "t "r c E "r "u c
ht uC jVCj˝Iab i
; (10.24)
E "t "u
where again V l represents the Gaunt and the scalar-retardation potentials for the
two electrons.
In order to evaluate the expression above, we first perform an additional iteration
of the pair equation (10.12)
˙
˛
E h0 .1/ h0 .2/ j˝ab D VCj˝Iab i Q VCj˝Icd i hcd jI
Pair
jabi (10.25)
yielding a new pair function with a single hole output. The solution can be expressed
in analogy with (10.15)
˛ ht uC jVCj˝Iab i ht uC jVCj˝Icd i ˝ ˇ Pairˇ ˛

˙
ht uCj˝ab D cd ˇI ˇab :
E "t "u .E "t "u /.E 0 "t "u /
(10.26)
The first part of the solution, illustrated in Fig. 10.4 (a), represents the last factor in
the expression (10.24). The second folded part will be used later in constructing the
complete folded contribution.
In evaluating the first term within the square brackets of (10.24), we first multiply
the pair function (10.26) without folded contribution by
hrC jV1l jt i

"t "r c
a b c s
6 d s
6
u u r
6 r 6 r u u
6
t 6 6
Vl
t - t t
6 6 6 6 6 6 6 6
a6 6
b a6 6
b a6 6
b a6 6
b
Fig. 10.4 (a–c): Generating a pair function with a hole output (10.25), combined with a single-
particle perturbation V1 , and closed with a perturbation V2 . The last diagram contains a crossing
Coulomb interaction, as discussed at the end of the section
yielding
˙
l hrC jV1l jt iht uC j˝ab i hrC jV1l jt i ht uC jVCj˝Iab i
hrC uC j˝ab iD D
"t "r c "t "r c E "t "u
and represented by the diagram (b) in Fig. 10.4. Then we close the photon by multi-
plying with hsC jV2l juC i and including the final denominator, yielding
hsC jV2l juC i hrC jV1l jt i ht uC jVCj˝Iab i

hrC sC j˝spjabi D ; (10.27)
E "r "s "t "r c E "t "u
which agrees with the corresponding part of (10.24) (Fig. 10.4c).
The second term in the brackets of (10.24) is evaluated in a similar way with a
different denominator.
The folded contribution is in lowest order from (10.21)
ı. Q Usp / ı. Q Vsp Q VC /
j˝Icd i hcd jI Pairjabi D j˝Icd i hcd jI Pairjabi :
ıE ıE
(10.28)
Here,
ı. Q Vsp Q VC / ı Q ıVsp ı. Q VC /
D Vsp Q VC C Q Q VC C Q Vsp : (10.29)
ıE ıE ıE ıE
The difference ratio ı. Q VC /=ıE is obtained from the folded part of (10.26) (VC
is energy independent). Similarly, the difference ratio of the relevant part of Vsp is
obtained by including an extra factor
1
:
E0 "r "u c
Finally, the difference ratio of Q is
ı Q 1
D 0 :
ıE .E "r "s /.E "r "s /
10.3 Single-Photon Exchange: Virtual Pairs 219
It should be noted that the last term in (10.29) is combined with the pair function
(10.26) with folded contribution, while the other terms are combined with that
function without that contribution. This is important to avoid the singularities of
Brown–Ravenhall type, mentioned above.
10.3.2 Full Treatment
We shall now generalize the treatment above and consider all 16 combinations of
the single-photon exchange (8.11) (see Fig. 8.3) essentially in one single step.
To begin with, we leave out the folded contribution. Then, the expression to eval-
uate is
hrjV l jti hsjV l jui htujVCj˝Iab i

E "r "s

t˙ r t˙ s˙ u˙ r˙ u˙ s
˙ ˙ ˙ ˙ :
"t "r ˙ c E "t "s c E "r "u c "u "s ˙ c
(10.30)
If we in this expression make the substitution t r $ us, we get an identical result

but with a and b interchanged. Therefore, we can replace the expression above by
the much simpler expression
hrjV l jti hsjV l jui htujVCj˝Iab C ˝Iba i

E "r "s

t˙ r u ˙ r˙
˙ ˙ : (10.31)
"t "r ˙ c E "r "u c
The last expression can be evaluated in the following way. We first evaluate the
matrix element htujVCj˝Iab i by iterating the pair equation (10.25) once, allowing
negative-energy states as output. We can separate the solutions into four block, de-
pending on the signs of the outgoing orbital energies, as illustrated in the matrix in
Fig. 10.5.
Next, we evaluate the matrix elements hrjV1jti for each value of and l, and
separate them in a similar way, shown in Fig. 10.6.
We now multiply the matrices in Figs. 10.5 and 10.6 (in that order), leading to
the matrix in Fig. 10.7. Here, we include the two denominator terms in the brackets
of (10.31) and sum over all t, particle as well as hole states.
Finally, we multiply the result by hsjV2jui and sum over and l, corresponding
to closing the photon (c.f. Fig. 10.4c), and apply the final denominator in (10.31).
If the input orbitals a; b are different, the procedure is repeated with a $ b.
The folded contribution in (10.21) is evaluated in a similar way [c.f. (10.28)].
Fig. 10.5 Representation 0 1

of the pair function in (10.25) B C
iterated one extra time and B
B
6t 6
u 6t C
C
separated into four blocks, B C
B C
B u C
depending on the signs of the B 6 6 6 6 C
outgoing orbital energies B C
B C
B a 6 C
B
B
6
b a 6 6
b C
C
B C
B C
B C
B C
B C
B 6
u C
B C
B C
B C
B t t u C
B 6 6 6 6 C
B C
B C
@ a 6 A
6
b a 6 6
b
Fig. 10.6 The matrix 0 1

elements hrjV1jt i, separated B C
in analogy with Fig. 10.5 B C
B r C
B
B
6V t
6r C
C
1
B C
B C
B C
B t 6 C
B C
B C
B C
B C
B C
B C
B C
B C
B C
B t ? C
B C
B C
B C
B C
Br C
B t r ? C
@ A
Fig. 10.7 Result of 0 1

multiplying the matrices in B C
B r 6 r 6 C
Figs. 10.5 and 10.6. The t line B C
represents particle as well as B
B t˙ 6 6u t˙ 6 C
C
B C
hole states B C
B
B 6 6 6 6u C
C
B a 6 C
B
B
6
b a6 6
b C
C
B C
B C
B C
B C
B C
B C
B r
6u r C
B t˙
6 t˙
6 C
B C
B C
B C
B 6 6 6 6u C
B C
@ a 6 6
b a6 6
b A
10.4 Numerical Results 221
6 6
6 6
-
6 6
6 6
Fig. 10.8 The Feynman diagrams representing a single transverse photon exchange combined
with high-order electron correlation (heavy horizontal line). The internal vertical lines represent
electron propagators with particle and holes. The numerical evaluation of this diagram is given
below
10.3.3 Higher Orders
In most of the cases treated above, it is possible also to insert Coulomb interaction
before the photon interaction is completed, as discussed in the no-pair case. This
is the case when the orbitals u; r or t; s are of the same kind (particle or hole), as
indicated in Fig. 10.4d. This corresponds to including another part of the potential
in (8.66).
After the completion of the single-photon exchange, the iteration process can be
continued with further Coulomb interactions, leading to the complete single-photon
exchange with electron correlation, including all combinations of particles and
holes, as illustrated in Fig. 10.8.
10.4 Numerical Results
10.4.1 Two-Photon Exchange
In Chap. 7 (Fig. 7.3), we showed the results of two-photon exchange for the ground-
state of helium-like ions, calculated using the S-matrix formulation. In Table 10.1,
we compare these results with those obtained by Hedendahl et al. [1, 3], in testing
the new covariant method described in the present chapter. The agreement, which is
found to be very good, is also displayed in Fig. 10.9, where the solid lines represent
the old S-matrix results and the squares the new covariant method. As before, the
scale is logarithmic and the norm is the nonrelativistic ionization energy.
10.4.2 Beyond Two Photons
Calculations have also been performed of the effect of electron correlation beyond
two-photon exchange for the ground state of helium-like ions by Hedendahl et al.
[1–3]. Some results are shown in Table 10.2 and also displayed in Fig. 10.10.
Table 10.1 Comparisons between two-photon effects for He-like ions ground
states, evaluated with the S-matrix and the covariant-evolution-operator methods (in
Hartree) (see Fig. 7.3)
Coul.–Breit Coul.–Breit. Coul.–Breit
Z Method NVPA retard. VP uncrossed.
6 S-matrix 1; 054:2 31:4 10:1
6 CEO 1; 054:9 31:5 10:0
10 S-matrix 2; 070:4 122:3 45:9
10 CEO 2; 071:0 122:4 45:9
14 S-matrix 5; 515 292:8 121:5
14 CEO 5; 517 292:8 121:2
18 S-matrix 8; 947 553:1 247:3
18 CEO 8; 950 553:3 248:2
30 CEO 23; 632 1; 909:9 1; 010
Fig. 10.9 Comparison of some two-photon exchange contributions (Coulomb–Breit NVPA,

Coulomb–Breit retardation, and Coulomb–Breit virtual pairs, no correlation) for the ground-state
of some helium-like ions obtained by S-matrix calculations (see Fig. 7.3) (heavy lines) and by
means of the covariant-evolution procedure (squares), described in this chapter (see Table 10.1,
c.f. Fig. 7.3 in Chap. 7) (from [1, 3])
The top line of the figure, representing the Coulomb–Breit interaction with corre-
lation without virtual pairs, contains the instantaneous as well as the retarded Breit
interaction. The former part lies within the no-virtual-pair approximation (NVPA)
and is therefore NOT a QED effect with the definition we have previously made. In
order to obtain the pure QED effect, the instantaneous part is subtracted, yielding the
retarded part, represented by the second line of the figure. The next line represents
the same effect with Coulomb crossings, and the bottom line represents the effect
of electron correlation on the Coulomb–Breit interaction with virtual pairs and no
10.4 Numerical Results 223
Table 10.2 Contributions Beyond two-photon Coul.–Breit

due to electron correlation
Z Unretarded Retarded Virt.pairs
beyond two-photon exchange
for the ground state of some 6 137 17 2.7
helium-like ions. This can be 10 223 40 7.3
compared with the 14 301 68 13
corresponding two-photon 18 372 100 21
exchange in Table 10.1 30 553 210 46
(in H) 42 688 322 71
Fig. 10.10 The effect of electron correlation beyond two-photon exchange – Coulomb–Breit
NVPA, Coulomb–Breit retardation with and without Coulomb crossings, and Coulomb–Breit vir-
tual pairs, all WITH electron correlation, for the ground-state of helium-like ions (c.f. Fig. 10.9)
(from [1, 3]). For comparison, the effect of pure retarded two-photon exchange without additional
correlation is also indicated
crossing Coulomb interactions. The corresponding Feynman diagrams are shown at

the bottom of the figure. This represents the first numerical bound-state calculation
beyond two-photon exchange.
In the figure, we have for comparison also indicated the effect due to doubly
retarded two-photon interactions (thin black line), estimated from the S-matrix
results. This comparison demonstrates the important result that – starting from
single-photon exchange – for light and medium-heavy elements the effect of
electron correlation is much more important than a second retarded photon in-
teraction.
10.4.3 Outlook
The results presented here are incomplete and represent only the nonradiative part
of the QED effect in combination with electron correlation. The corresponding ra-
diative effects with a single transverse photon are also possible to evaluate. Such
calculations are underway by the Gothenburg group. The effect due to double trans-
verse photons is presently beyond reach, but the effect can be estimated by replacing
the second transverse photon by an instantaneous Breit interaction.
The calculations performed so far with the procedure described here concern the
ground states of helium-like ions [3]. By extending the calculations to excited states,
it will be possible to make detailed comparison with experimental data. For instance,
very accurate data exist for some helium-like ions, as shown in Tables 7.7 and 7.8.
In some of these cases, the experimental results are at least two orders of magnitude
more accurate than the best theoretical estimates made so far. Furthermore, it seems
that standard procedures applied until now cannot be significantly improved in this
respect, so – to be able to make significant progress – there might be a need for
a new, improved procedure, like the MBPT–QED procedure presented here. Then,
it might be possible for the first time to observe the combined effect of QED and
electron correlation.
References
1. Hedendahl, D.: Towards a Relativistic Covariant Many-Body Perturbation Theory. Ph.D. thesis,
University of Gothenburg, Gothenburg, Sweden (2010)
2. Hedendahl, D., Lindgren, I., Salomonson, S.: Towards numerical implementation of the rela-
tivistically covariant many-body perturbation theory. Can. J. Phys. 87, 817–24 (2008)
turbation: Merging many-body perturbation theory with QED. Phys. Rev. A 73, 062,502 (2006)
Chapter 11
Analytical Treatment of the Bethe–Salpeter
Equation
11.1 Helium Fine Structure
The leading contributions to the helium fine structure beyond the first-order
relativistic contribution (NVPA, see, Sect. 2.6) were first derived in 1957 by
Araki [1] and Sucher [6, 7], starting from the Bethe–Salpeter (BS) equation [5]
and including the nonrelativistic as well as the relativistic momentum regions.
Following the approach of Sucher, Douglas and Kroll [2] have derived all terms of
order ˛ 4 H(artree)1, where no contributions in the relativistic region were found.
The same approach was later used by Zhang [8, 12] to derive corrections of order
˛ 5 log ˛ H and of order ˛ 5 H in the nonrelativistic region and recoil corrections
to order ˛ 4 m=M H (see also [10]). Later some additional effects of order ˛ 5 H
due to relativistic momenta were found by Zhang and Drake [11]. The radiative
parts are treated more rigorously by Zhang in a separate paper [9]. Using a different
approach, Pachucki and Sapirstein [4] have derived all contributions of order ˛ 5 H
and reported some disagreement with the early results of Zhang [8].2
We shall here follow the approach of Sucher in his thesis [7]. This is based di-
rectly on the BS equation, which makes it possible to identify the contributions in
terms of Feynman diagrams and therefore to compare them with the results obtained
in the previous chapters. This approach of Sucher is closely followed by Douglas
and Kroll [2] and by Zhang [8], and we shall in our presentation make frequent ref-
erences to the corresponding equations of Sucher (S), Douglas and Kroll (DK), and
Zhang (Z).
1
H(artree) is the energy unit of the Hartree atomic unit system (see Appendix K.1). In the rela-
tivistic unit system, the energy unit is mc 2 D ˛ 2 H .
2
This chapter is largely based on the paper [3].

226 11 Analytical Treatment of the Bethe–Salpeter Equation
11.2 The Approach of Sucher
The treatment of Sucher starts from the Bethe–Salpeter equation (9.5), which in our
notations (9.6) reads, leaving out the inhomogeneous term (S 1.1, DK 2.5),
ZZZZ
.x; x 0 / D d4 x1 d4 x2 d4 x10 d4 x20
G00 .x; x 0 I x2 ; x20 / .i/˙ .x2 ; x20 I x1 ; x10 / .x1 ; x10 /: (11.1)
G00 is the zeroth-order two-particle Green’s function, dressed with all kinds of
single-particle self-energies. ˙ is identical to the irreducible potential V (Fig. 6.6).
The undressed zeroth-order Green’s function is, using the relation (5.38),
G0 .x; x 0 I x2 ; x20 / D G0 .x; x2 / G0 .x 0 ; x20 / D iSF .x; x2 / iSF .x 0 ; x20 / (11.2)
and the corresponding dressed function is then
G00 .x; x 0 I x2 ; x20 / D G.x; x2 / G.x 0 ; x20 / D iSF0 .x; x2 / iSF0 .x 0 ; x20 /; (11.3)
where G is the full single-particle Green’s function, generated in the field of the
nucleus (Furry representation) (see Fig. 5.1), and SF0 the correspondingly dressed
electron propagator. The Green’s functions satisfy the relation (5.36) (S 1.5)

@
i h1 G.x; x0 / D iı 4 .x x0 /; (11.4)
@t
which leads to (S 1.6, DK 2.19)

ZZZZ
@ @
i h1 i 0 h2 .x; x 0 / D i d4 x1 d4 x2 d4 x10 d4 x20 ı 4 .x x2 /
@t @t
ı 4 .x 0 x20 / ˙ .x2 ; x20 I x1 ; x10 / .x1 ; x10 /
ZZ
Di d4 x1 d4 x10 ˙ .x; x 0 I x1 ; x10 / .x1 ; x10 /; (11.5)
where h1;2 are the Dirac single-electron Hamiltonians for electron 1 and 2.
We assume that the wave function is of the form
.x; x 0 / D .T; ; x; x 0 / D eiE T .; x; x 0 /; (11.6)
where T D .t C t 0 /=2 is the average time and D t t 0 is the relative time. Then

@ 0 @
i .x; x / D E=2 C i .x; x 0 /;
@t @

@ 0 @
i 0 .x; x / D E=2 i .x; x 0 /
@t @
11.2 The Approach of Sucher 227
leading to (S 1.9, DK 2.23)

@ @
E=2 C i h1 E=2 i h2 .; x; x 0 /
@ @
Z ZZ
D i d1 d3 x 1 d3 x 01 ˙ .; x; x 0 I 1 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /(11.7)
leaving out the average time.

Sucher then transfers to the momentum representation, but we shall here still
work in the coordinate representation with a Fourier transform only of the time
variables.
We define the Fourier transform with respect to time
Z
F ./ D d ei F ./ (11.8)
and the inverse transformation

Z
d i
F ./ D e F ./: (11.9)
2
Fourier transforming (11.7) with respect to yields
.E=2 C h1 / .E=2 h2 / .; x 1 ; x 01 /

Z ZZ
D i d1 d3 x 1 d3 x 01 ˙ .; x; x 0 I 1 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /: (11.10)
Performing the Fourier transform of the right-hand side with respect to 1 yields
Z ZZ
d10 d1 i 0 1 i 1 1
d1 e 1 e ˙ .; x; x 0 I 10 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /
2 2
ZZ
d10 d1
D 2ı.1 C10 / ˙ .; x; x 0 I 10 ; x 1 ; x 01 / .1 ; x 1 ; x 01 / (11.11)
2 2
or (S 1.16)
.E=2 C h1 / .E=2 h2 / .; x 1 ; x 01 /

Z ZZ
d1
Di d3 x 1 d3 x 01 ˙ .; x; x 0 I 1 ; x 1 ; x 01 / .1 ; x 1 ; x 01 /: (11.12)
2
Following Sucher, we express the relation (11.10) in operator form
FO j i D gO j i : (11.13)
The operator FO has the (diagonal) coordinate representation

˝ ˛
; x; x 0 jFj; x; x 0 D .E=2 C h1 / .E=2 h2 / (11.14)
and the operator gO has the (nondiagonal) representation
˝ ˛ i ˝ ˛
; x; x 0 jgj
O 1 ; x 1 ; x 01 D ; x; x 0 j˙Oj1 ; x 1 ; x 01 : (11.15)
2
We expand the interaction into
gO D gO c C gO ; (11.16)
where gO c represents the Columbic part of gO
i O
gO c D Ic (11.17)
2
and IOc is the Coulomb interaction with the (diagonal) coordinate representation
˝ ˛ e2
; x; x 0 jIOc j; x; x 0 D ; (11.18)
4jx x1 j
gO represents the remaining part of gO
gO D gO T C gT c C gO T c 2 C gO T T C C gO rad ; (11.19)
where gO T represents a single transverse photon, gO T c and gO T c 2 a transverse pho-

ton with one and two crossing Coulomb interactions, respectively, gO T T with two
irreducible transverse photons, and finally gO rad all radiative corrections. This corre-
sponds to the diagrams are shown in Fig. 11.1.
With the decomposition (11.16), the relation (11.13) becomes (S 1.30, DK 3.6)
1
j i D FO gO gO c j i (11.20)
with the coordinate representation
˝ 1 ˛
h; x; x 0j i D ; x; x 0 j FO gO j2 ; x 2 ; x 02
˝ ˛
2 ; x 2 ; x 02 jgO c j1 ; x 1 ; x 01 h1 ; x 1 ; x 01j i (11.21)
+ + + + + +
Fig. 11.1 Diagrammatic representation of the approximation in (11.19), used by Sucher

11.2 The Approach of Sucher 229
or noting that the representation of gO c is diagonal
˝ 1 ˛
h; x; x 0j i D ; x; x 0 j FO gO j1 ; x 1 ; x 01 gO c h1 ; x 1 ; x 01j i : (11.22)
Sucher defines the equal-time wave function (S 1.32, DK 3.8)

Z
˚.x; x 0 / D d .; x; x 0 / (11.23)
or in operator form
j˚i Dji hj i ; (11.24)
which gives with (11.22)
˝ 1 ˛
h; x; x 0j i D ; x; x 0 j FO gO jx 1 ; x 01 gO c hx 1 ; x 01j˚i : (11.25)
Summing over with the replacement (11.17), this can be expressed as (S 1.34)
Z
d O 1
j˚i D i F gO IOc j˚i : (11.26)
2
Using the identity (S 1.35, DK 3.11)
.A B/1 A1 C A1 B.A B/1 ; (11.27)
the BSE (11.26) becomes (DK 3.12)

Z
d h O 1 i
j˚i D i F C FO 1 gO .FO gO /1 IOc j˚i : (11.28)
2
The inverse of the operator OF is
1 1
FO 1 D ; (11.29)
E=2 C h1 E=2 hO 2
O
which is a product of electron propagators in operator form (4.14)
FO 1 D SOF .E=2 C / SOF .E=2 /: (11.30)
In the coordinate representation (4.12),
hxjj i hj jx 0 i hxjj i hj jx 0 i hxjj i hj jx 0 i

SF .!I x; x 0 / D D CC :
! "j C i sgn."j / ! "j C i ! "j i
(11.31)
Integration over then yields (S 1.44, DK 3.24)

Z
d 1 hx; x 0jrsi hrsjx0 ; x00 i
F D i . CC / ; (11.32)
2 E "r "s
which is also the negative of the Fourier transform of the zeroth-order Green’s func-
tion G0 .EI x; x 0 I x 0 ; x 00 /, or in operator form
Z
d O 1 i
F D G0 .E/ D . CC / : (11.33)
2 E hO 1 hO 2
Equation (11.28) then becomes (S 1.47, DK 3.26)3
Z
d 1
h1 C h2 C CC Ic C D i F g .F g /1 Ic ˚ D E ˚;
2
(11.34)
where
D D E h1 h2 : (11.35)
This is the starting point for the further analysis.
The operator on the left-hand side can be written in the form Hc C H , where
Hc D h1 C h2 C CC Ic CC (11.36)
is the Hamiltonian of the no-(virtual-) pair Dirac-Coulomb equation (Z 16)
Hc c D Ec c (11.37)
and
Z
d 1
H D CC Ic .1 CC / Ic CDi F g .F g /1 Ic
2
DH 1 CH 2 (11.38)
is the remaining “QED part” (S 2.3, DK 3.29, Z 17). The first part H 1 represents
virtual pairs due to the Coulomb interaction and the second part effects of transverse
photons (Breit interaction).
In order to include electron self-energy and vacuum polarizations, the electron
propagators (5.37) are replaced by propagators with self-energy insertions ˙./,
properly renormalized (DK 2.10),
jri hrj
S 0 ./ D : (11.39)
"r C ˇ˙./ C ir
Also renormalized photon self-energies have to be inserted into the photon lines.
3
In the following, we leave out the hat symbol on the operators.
11.3 Perturbation Expansion of the BS Equation 231
11.3 Perturbation Expansion of the BS Equation
The effect of the QED Hamiltonian (11.38) can be expanded perturbatively, using
the Brillouin–Wigner perturbation theory,
E D E Ec D h c jV
CV V CV V V C j ci
V
D h cj j ci ; (11.40)
1 V
where is the reduced resolvent (2.65)
Q 1 j c i h c j 1 j c i h cj
D Q .E/ D D D (11.41)
E Hc E Hc Dc
with
Dc D E Hc : (11.42)
The unperturbed wave function is in our case one solution of the no-pair Dirac–
Coulomb equation (11.37), c , and we can assume that the perturbation is expanded
in other eigenfunctions of Hc . Q is the projection operator that excludes the state
c (assuming no degeneracy). This leads to the expansion (S 2 19–21, DK 3.43,
Z 28)
E .1/ D h c jH j ci ; (11.43a)
E .2/ D h c jH H j ci ; (11.43b)
E .3/ D h c jH H H j ci ; (11.43c)
etc.
Since CCj c i D j c i and j c i D 0, it follows that h c jH 1j ci 0,
and the first-order correction becomes (DK 3.44)
Z
d 1
E D h c jH 2j c i D h c jD i
.1/
F J F 1 Icj c i (11.44)
2
and (DK 3.45)
J D g .1 F 1 g /1 : (11.45)
The second-order corrections are (DK 3.46)4
Ea.2/ D h c jH 1 H 1j ci D h c jIc Ic j c i ; (11.46a)
4
Note that the two Ic in (11.46a) are missing from [2, Eq. 3.46]. Equation (11.46b) agrees with
[8, Eq.3.30] but not with [2], where the factor Ic LCC should be removed.
Z
d 1
Eb.2/ D h cH 1 H 2i c D c Ic D i F J F 1 Ic c ;
2
(11.46b)
Z
d 1
Ec.2/ D h cH 2 H 1i c D cD i F J F 1 Ic Ic c ;
2
(11.46c)
Ed.2/ D h cH 2 H 2i c
Z Z
d 1 d 1
D cD i F J F 1 Ic Di F J F 1 Ic c : (11.46d)
2 2
These formulas can be simplified, noting that
Q
DD .E h1 h2 /; (11.47)
E Hc
which, using the relation (11.42), becomes (DK 3.41)

CC Ic CC
DD 1 C D : (11.48)
E Hc
According to DK Ea.2/ ; Ec.2/ and E .3/ do not contribute to the fs in order

˛ (Hartree). This holds also in the next order according to Zhang, but E .3/ will
4
contribute to the singlet energy in that order. In the relativistic momentum region,
the second-order part Ea.2/ contributes to the energy already in order ˛ 3 H and to
the fine structure in order ˛ 5 H [8, p. 1256].
Using the relation (11.42), we have Ec Hc D Dc CC Dc CC , and the
no-pair equation (11.37) can be written (DK 3.51)
.Dc CC Ic / c D 0: (11.49)
.2/
Then the second-order correction Eb (11.46b) can be expressed as
Z
d 1
F J F 1 Ic j
.2/
Eb D h c j.Ic Dc / i ci : (11.50)
2
This can be combined with the first-order correction E .1/ (11.44), yielding
Z
d 1
h c j.Ic C E/ i F J F 1 Ic j ci (11.51)
2
11.4 Diagrammatic Representation 233
with
E D E Ec D D Dc : (11.52)
Here, the E term differs in sign from (DK 3.54) and (Z 37).
The reason for the discrepancy between our result here and those of DK and Z
seems to be that the latter make the replacement (DK 3.48)
1 S1 C S2
F 1 D S1 S2 .S1 C S2 / S11 C S21 D D D 1 .S1 C S2 / ;
E h1 h2
(11.53)
which follows from (11.30), and then approximate D with Dc in the second-order
expression.
11.4 Diagrammatic Representation
To continue, we make the expansion (DK 3.45, Z 32)
J D g .1 F 1 g /1 D g C g F 1 g C ; (11.54)
where the first term represents irreducible terms and the remaining ones are
reducible. Furthermore, we make the separation (DK 3.53, Z 12)
g D gT C g; (11.55)
where gT represents the interaction of a single transverse photon and g the

irreducible multiphoton exchange of (11.19). The first-order expression (11.44) be-
comes
Z
d 1
E D h c jD i
.1/
F gT C gT F 1 gT C g C F 1 Ic j c i (11.56)
2
and the leading terms are illustrated in Fig. 11.2. The first term can be expanded in
no-pair and virtual-pair terms (a–c)
Z
d 1
E .1/
Dh c jD i F gT F 1 . CC C C C C C /Ic j c i : (11.57)
2
The second term in (11.56) represents in lowest order two reducible transverse
photons (d) and the third term irreducible (inclusive radiative) multiphoton
part, (e–h).
a b c d e f
g h

Fig. 11.2 Diagrammatic representation of the first-order expression (11.56)
a b c d
Fig. 11.3 Diagrammatic representation of the second-order expressions (11.58a)–(11.58d)
Similarly, the second-order expressions above become
Ea.2/ D h c jIc Ic j c i ; (11.58a)
Z
d 1
gT C gT F 1 gT C g C F 1 Ic
.2/
Eb D c Ic i F c ;
2
(11.58b)
Z
d 1
Ec.2/ D ci F gT C gT F 1 gT C g C F 1 Ic Ic c ;
2
(11.58c)
Z
d 1
Ed.2/ D cD F ŒgT C F 1 Ic
i
2
Z
d 1
Di F ŒgT C F 1 Ic c : (11.58d)
2
This is illustrated in Fig. 11.3. The first second-order contribution (11.58a)

represents two Coulomb interactions with double pair (Fig. 11.3a) and the next
References 235
contribution (11.58b) in lowest order a transverse photon and a Coulomb interaction

with double pair (b). The third contribution represents in lowest order one transverse
photon and two Coulomb interactions with a double pair (c). The last term repre-
sents two reducible transverse photons with at least one Coulomb interaction (d).
11.5 Comparison with the Numerical Approach
In the previous chapter, we have described an approach that is presently being

developed by the Gothenburg group of treating the Bethe–Salpeter equation numer-
ically. This is based on the covariant-evolution approach and the Green’s-operator
technique, described previously, and to a large extent upon the numerical techniques
developed by the group and applied to numerous atomic systems (see Sect. 2.7).
This new technique has the advantage over the analytical approach that all relativis-
tic effects are automatically included in the procedure. This simplifies the handling
appreciably, and it corresponds to the treatment of the entire Sect. 4 of Douglas and
Kroll [2] or to Sect. VII in the paper of Zhang [8].
The numerical technique of solving the Bethe–Salpeter equation, described in the
previous chapter, is presently only partly developed, but the effect of one transverse
photon with arbitrary number of crossing Coulomb interactions can presently be
handled as well as virtual pairs. This corresponds to most of the terms gT CgT c C
gT c 2 C of the expansion in (11.19) and to the numerous formulas of Sect. 5 of
Douglas–Kroll and of Sect. IV of Zhang.
Also part of the multiphoton effect can be treated numerically by iterating re-
ducible interactions with a single transverse photon, corresponding to the operator
gT F gT in the formulas above with crossing Coulomb interactions. These ef-
fects are treated in Sect. 6 of Douglas and Kroll. The irreducible interaction with
several transverse photons cannot be teated at present with the numerical tech-
nique, but this can be approximated with one retarded and one or several unretarded
photons (instantaneous Breit). Also radiative effects can be handled with the same
approximation.
References

Phys. (Japan) 17, 619–42 (1957)
3. Lindgren, I.: The helium fine-structure controversy. arXiv;quant-ph p. 0810.0823v (2008)
4. Pachucki, K., Sapirstein, J.: Contributions to helium fine structure of order m˛ 7 . J. Phys. B
33, 5297–5305 (2000)
1232–42 (1951)
Helium. Phys. Rev. 109, 1010–11 (1957)
Michigan
9. Zhang, T.: QED corrections to O.˛ 7 mc 2 / fine-structure splitting in helium. Phys. Rev. A 53,
3896–3914 (1996)
10. Zhang, T.: Three-body corrections to O.˛ 6 / fine-structure in helium. Phys. Rev. A 56, 270–77
(1997)
11. Zhang, T., Drake, G.W.F.: Corrections to O.˛ 7 mc 2 / fine-structure splitting in helium. Phys.
Rev. A 54, 4882–4922 (1996)
12. Zhang, T., Yan, Z.C., Drake, G.W.F.: QED Corrections of O.mc 2 ˛ 7 ln ˛/ to the Fine Structure
Splitting of Helium and He-like Ions. Phys. Rev. Lett. 77, 1715–18 (1996)
Chapter 12
Regularization and Renormalization
(See, for instance, Mandl and Shaw [17, Chap. 9] and Peskin and Schroeder [23,
Chap. 7].)
In previous chapters, we have evaluated some radiative effects in the S-matrix
(Chap. 4) and covariant-evolution operator formulations (Chap. 8). In this chapter,
we discuss the important processes of renormalization and regularization in some
detail.
Many integrals appearing in QED are divergent, and these divergences can be re-
moved by replacing the bare electron mass and charge by the corresponding physical
quantities. Since infinities are involved, this process of renormalization is a delicate
matter. In order to do this properly, the integrals first have to be regularized, which
implies that the integrals are modified so that they become finite. This has to be done
so that the process is gauge-independent. After renormalization, the regularization
modification is removed. Several regularization schemes have been developed, and
we shall consider some of them in this chapter. If the procedure is performed prop-
erly, the way of regularization should have no effect on the final result.
12.1 The Free-Electron QED
12.1.1 The Free-Electron Propagator
The wave functions for free electrons are given by (D.29) in Appendix D
3=2
pC .x/ D .2/ uC .p/ eip x eiEp t
3=2 ; (12.1)
p .x/ D .2/ u .p/ eip x eiEp t
where p is the momentum vector and pC represents positive-energy

p states (r D 1; 2)
and p negative-energy states (r D 3; 4). Ep D cp0 D c 2 p2 C m2 c 4 . The coordi-
nate representation of the free-electron propagator (4.10) then becomes
Z
d! X p;r .x 1 / p;r .x 2 / i!.t1 t2 /

hx1 jSOFfreejx2 i D e ; (12.2)
2 p;r ! "free
p .1 i/

238 12 Regularization and Renormalization
p is the energy eigenvalue of the free-electron function (Ep D j"p j). The
where "free free
Fourier transform with respect to time then becomes

X
p;r .x 1 / p;r .x 2 /
hx1 jSOFfreejx2 i D )
p;r
! "free
p .1 i/
Z
d3 p X eip.x1 x2 /
D u r .p/ u
r .p/
.2/3 r ! "free
p .1 i/
Z 3
d p 1
D uC .p/ uC .p/
.2/ 3 ! Ep .1 i/

1
C u .p/ u .p/ eip.x1 x2 / :
! C Ep .1 i/
The square bracket above is the Fourier transform of the propagator, and using the
relations (D.35, D.36), this becomes1

1 1 1
SFfree .!; p/ D C
2 ! Ep .1 i/ ! C Ep .1 i/
2
c ˛ p C ˇmc 1 1
C
2p0 ! Ep .1 i/ ! C Ep .1 i/
! C c ˛ p C ˇmc 2 ! C c˛ p C ˇmc 2
D D
! 2 Ep2 C i ! 2 .c 2 p2 m2 c 4 / .1 i/
1
D (12.3)
! .c ˛ p C ˇmc 2 / .1 i/
with Ep2 D c 2 p02 D c 2 p2 C m2 c 4 and ˛ˇ D ˇ˛. This can also be expressed
1
SFfree .!; p/ D ; (12.4)
! hfree
D .p/ .1 i/
where hfree
D .p/ is the momentum representation of the free-electron Dirac Hamilto-
nian operator (D.21), hO free O
D .p/.
covariant four-component form with ! D cp0
Formally, we can write (12.3) inp
with cp0 disconnected from Ep D c 2 p2 C m2 c 4 – known as off the mass-shell.
Then we have2
1
SFfree .p/ D cSFfree .!; p/ D ˇ (12.5)
6 p mc C i
1
In the following, we shall for simplicity denote the electron physical mass by m instead of me .
2
The factor of ˇ appears here because we define the electron propagator (4.9) by means of O
instead of the more conventionally used NO D O ˇ.
12.1 The Free-Electron QED 239
with 6 p D p D ˇ˛ p D ˇ.p0 ˛p/ D .p0 C˛p/ ˇ (see Appendix D). Note

that the two transforms differ by a factor of c (c.f. Sect. 4.3, see also Appendix K).
12.1.2 The Free-Electron Self-Energy
The S-matrix for the first-order free-electron self-energy (Fig. 12.1) is obtained from
(4.84) and (4.44) with the momentum functions (12.1) after time integrations
Z ZZ
dz 0 0
S .2/ .!I p0 ; r 0 ; p; r/ D e 2 c 2 d3 x d3 x 0 ur0 .p0 / eip x
2
˛ iSFfree .! zI x 0 ; x/ ˛ ur .p/ eipx iDF .z; x 0 x/:
(12.6)
The relation between the momentum and coordinate representations is
Z
d3 q free 0
SFfree .!I x 0 ; x/ D 3
SF .!; q/ eiq.x x/ ; (12.7)
.2/
Z
d3 k 0
DF .zI x 0 ; x/ D 3
free
DF .z; k/ eik.x x/ : (12.8)
.2/
Integration over the space coordinates then yields

Z Z
0 0 d3 q d3 k 3
S .!I p ; r ; p; r/ D e c
.2/ 2 2
ı .p q k/ ı 3 .p0 q k/
.2/3 .2/3
Z
dz free
ur0 .p0 / ˛ SF .! z; k/ ˛ DF
free
.z; k/ ur .p/
2
(12.9)
and integration over q
S .2/ .!I p0 ; r 0 ; p; r/ D ı 3 .p0 p/ ur0 .p0 / .i/˙ free .!; p/ ur .p/; (12.10)
where
Z Z
dz d3 k free
˙ free
.!; p/ D ie c 2 2
˛ SF .! z; k/ ˛ DF .z; k/:
2 .2/3
p0 r 0 6
!

qs
6 zk
Fig. 12.1 Diagram
representing the first-order
pr !
free-electron self-energy 6
In covariant notation we have, using z D ck0 ,

Z
d4 k free
˙ free .p/ D ie 2 c 2 ˛ SF .p k/ ˛ DF .k/; (12.11)
.2/4
which is the free-electron self-energy function. With the expression (12.5) for the
free-electron propagator, this becomes expressed in terms of gamma matrices
Z
d4 k 6 p 6 k C mc
ˇ˙ free
.p/ D ie c
2 2
DF .k/ (12.12)
.2/4 .p k/2 m2 c 2 C i
or
Z
d4 k 1
ˇ˙ free .p/ D ie 2 c 2 DF .k/: (12.13)
.2/ 4 6 p 6 k mc C i
With the commutation rules in Appendix (D.58), this becomes
Z
d4 k 6 p 6 k 2mc
ˇ˙ free
.p/ D 2ie c 2 2
DF .k/ (12.14)
.2/4 .p k/2 m2 c 2 C i
and with the photon propagator (4.28) we have in the Feynman gauge
Z
2ie 2 c d4 k 6 p 6 k 2mc 1
ˇ˙ free .p/ D : (12.15)
0 .2/ .p k/ m c C i k C i
4 2 2 2 2
fAs mentioned above, the factor of ˇ is due to our definition of the electron propa-
gator [c.f. (12.5)]g.
12.1.3 The Free-Electron Vertex Correction
We consider first the single interaction with an external energy potential

(Appendix D.41) e˛ A (Fig. 12.2 left). The S-matrix is given by
Z
0
S .1/ .! 0 ; !I p0 r 0 ; pr; q/ D iec d3 x ur0 .p0 / eip x ˛ A .x/ ur .p/ eipx (12.16)
or
S .1/ .! 0 ; !I p0 r 0 ; pr; q/ D iec ı 3 .p p0 / ur 0 .p0 / ˛ A .p p0 / ur .p/;

(12.17)
where A .q/ is the Fourier transform of A .x/.

12.1 The Free-Electron QED 241
p0 r 0 6!
p0 r 0 6

6q0
A .q/ A .q00 / zk
6q

pr
6
pr !
6
Fig. 12.2 Diagram representing the first-order free-electron vertex correction
The vertex-modified free-electron self-energy diagram in Fig. 12.2 (right) be-

comes similarly
Z ZZZ
dz 0 0
S .3/ .! 0 ; !I p0 r 0 ; pr/ D .ie/3 c 2 d3 x 1 d3 x 2 d3 x 3 ur 0 .p0 / eip x

2
˛ iSFfree .! 0 z; x 0 ; x 00 /˛ A .x 00 / ˛
iSFfree .! z; x 00 ; x/ ur .p/ eipx iDF .z; x 0 x/:
(12.18)
In analogy with (12.9), this becomes
Z Z Z 3 00 Z
d3 q d3 q0 d q d3 k 0
S .3/ .! 0 ; !I p0 r 0 ; pr/ D e 3 c 2 u 0 .p /
.2/ 3 .2/ 3 .2/ 3 .2/3 r
ı 3 .p q k/ ı 3 .p0 q0 k/ ı 3 .q q0 C q00 /
Z
dz free 0
˛ SF .! z; q0 /
2
˛ A .q00 / ˛ SFfree .! z; q/ ur .p/ DF .z; k/
(12.19)
0 00
and after integrations over q, q , and q
S .3/ .! 0 ; !I p0 r 0 ; pr/ D ie ı 3 .p p0 / ur 0 .p0 / .! 0 ; !I p0 ; p/ A .p p0 / ur .p/;

(12.20)
where
Z Z
0 0 dz d3 k free 0
.! ; !I p ; p/ D ie c 2 2
˛ SF .! z; p0 k/
2 .2/3
˛ SFfree .! z; p k/ ˛ DF .z; k/ (12.21)
is the vertex correction function. In covariant notations, this becomes in analogy

with (12.10)
Z
0 d4 k free 0
.p ; p/ D ie c
2
˛ SF .p k/ ˛ SFfree .p k/ ˛ DF .k/ (12.22)
.2/4
In the Feynman gauge, this becomes

Z
ie 2 d4 k 1
.p 0 ; p/ D 0
0 .2/ 4 6 p 6 k mc C i
1 1
2 : (12.23)
6 p 6 k mc C i k C i
Comparing with the self-energy function (12.15), we find the Ward identity
(4.100) [17, Eq. 9.60]
@
˙.p/ D .p; p/: (12.24)
@cp
Obviously, this relation holds independently of the gauge.
12.2 Renormalization Process
We shall here derive expressions for the mass and charge renormalization in terms
of counterterms that can be applied in evaluating the QED effects on bound states.
The process of regularization will be treated in the next section.
12.2.1 Mass Renormalization
We consider now a bare electron with the mass m0 . The corresponding free-electron
propagator (12.5) is then
1
SF bare .!; p/ D ˇ (12.25)
6 p c m0 c 2 C i
with ! D cp0 .
We now “dress” the bare-electron propagator with all kinds of self-energy in-
sertions in the same way as for the bound-electron propagator in Fig. 5.7. This
corresponds to the S-matrix in operator form3
iSF .!; p/
iSF .!; p/ C iSF .!; p/.i/˙.!; p/ iSF .!; p/ C D ;
1 ˙.!; p/ SF .!; p/
(12.26)
3
Note that ˙.!; p/ has the dimension of energy and that the product ˙.!; p/ SF .!; p/ is dimen-
sionless (see Appendix K).
12.2 Renormalization Process 243
D C C C D C
Fig. 12.3 Dyson equation for the dressed bare-mass electron propagator
˙ free .p/= D C 6 C 6 C
Fig. 12.4 Expansion of the proper self-energy operator for a bare electron
which leads to
1
SF bare;dressed .!; p/ D ˇ (12.27)
6 p c m0 c ˇ˙bare
2 .!; p/ C i
illustrated in Fig. 12.3. Here, the box represents the irreducible or proper self-energy

insertions, ˙bare .!; p/, illustrated in Fig. 12.4. We shall in the following refer to this
as the free-electron self-energy, ˙ free .!; p/,

˙bare .!; p/ D ˙ free .!; p/: (12.28)
To lowest order, the free-electron self-energy is in analogy with (4.85)

Z
d! bare
˙ free
.!; p/ D i S .!; p/ I bare .!I p/; (12.29)
2 F
where I bare is the interaction (4.44) in the momentum representation with the elec-
tronic charge replaced by the bare charge, e0 .
The bare-electron propagator itself is also associated with a bare-electron charge
(e0 ) at each vertex. The dressing of the electron propagator leads to a modification
of the electron mass as well as of the electron charge. One part of the free-electron
self-energy is indistinguishable from the mass term in the electron propagator, and
another part is indistinguishable from the electronic charge, and these parts give
rise to the mass renormalization and the charge renormalization, respectively. The
modification of the electron charge is here compensated by a corresponding mod-
ification of the vertex (to be discussed below), so that there is no net effect on the
electron charge in connection with the electron self-energy. On the other hand, there
is a real modification of the electron charge in connection with the modification of
the photon propagator, as we shall discuss later.
Instead of working with the bare-electron mass and charge with self-energy in-
sertions, we can use the physical mass and charge and introduce corresponding
counterterms (see, for instance, [13, p. 332]). The free-electron propagator with the
physical electron mass, m, is
1
SF free .!; p/ D ˇ (12.30)
6 p c mc 2 C i
and it has its poles “on the mass shell,” 6 p D mc [see Appendix (D.19)]. The dressed
propagator (12.27) should have the same pole positions, which leads with
m D m0 C ım (12.31)
to
ˇ
ımc 2 D ˇ˙ free .!; p/ˇ6p Dmc : (12.32)
This is the mass-counterterm. We can now in the dressed operator (12.27) replace
m0 c 2 by
ˇ
mc 2 ˇ˙ free .!; p/ˇ6p Dmc ;
which leads to
1
SF free;ren .!; p/ D ˇ ; (12.33)
6 p c mc ˇ˙ren
2 free .!; p/ C i
where
ˇ
free
˙ren .!; p/ D ˙ free .!; p/ ˙ free .!; p/ˇ6p Dmc : (12.34)
This represents the mass-renormalization. Both the free-electron self-energy and

the mass counterterms are divergent, while the renormalized self-energy is finite.
12.2.2 Charge Renormalization
12.2.2.1 Electron Self-Energy
The pole values (residues) of the dressed bare electron propagator should also be
the same as for the physical propagator, including the associated electronic charges.
The physical propagator (12.30) with the electronic charge
e2
e 2 SFfree .!; p/ D ˇ
6 p c mc 2 C i
has the pole value ˇe 2 =c. The dressed propagator (12.27) with the bare electron
charge is
e02 e02
ˇ D ˇ
6 p c m0 c 2 ˇ˙ free .!; p/ C i 6 p c mc 2 ˇ˙renfree .!; p/ C i
and its pole value at the pole6 p D mc is
ˇ e02 .6 p mc/ ˇ e02

lim D ˇ
6p !mc c 6 p c mc 2 ˇ˙ren free .!; p/ C i c 1 ˇ @c6p
@ free .!; p/ˇ
˙ren C i
6p Dmc
using l’Hospital’s rule. This gives us the relation
e02
e2 D ˇ (12.35)
1 ˇ @c6p
@ free .!; p/ˇ
˙ren 6p Dmc
or
ˇ
@ ˇ
e 2 D e02 1 C ˇ free
˙ren .!; p/ ˇ : (12.36)
@c 6 p 6p Dmc
Here, the second term, which is divergent, represents the first-order charge renor-
malization.
It is convenient to express the free-electron self-energy as
˙ free .!; p/ D A C B.6 p c mc 2 / C C.6 p c mc 2 /2 : (12.37)
It then follows that the constant A is associated with the mass renormalization,
ˇ
A D ˙ free .!; p/ˇ6p Dmc D ˇımc 2 (12.38)
and B with the charge renormalization,
@ ˇ
BD ˙ free .!; p/ˇ6p Dmc : (12.39)
@c 6 p
From (12.36), it follows that for the charge renormalization due to the dressing of
the electron propagator becomes
e D e0 .1 C B=2 C /: (12.40)
The constant C represents the renormalized free-electron self-energy that is

finite.
12.2.2.2 Vertex Correction
The modification of the vertex function shown in Fig. 12.2 can be represented by
ie0 .p; p 0 / D ie0 ˛ ie0 ˇ .p; p 0 /; (12.41)
where e0 is the “bare” electron charge. The vertex correction is divergent and can
be separated into a divergent part and a renormalized, finite part
.p; p 0 / D L˛ C ren
.p; p 0 /: (12.42)
The divergent vertex part corresponds to a charge renormalization, in first order

being
e D e0 .1 ˇL/: (12.43)
But this should be combined with the charge renormalization due to the dressing of
the electron propagators (12.40), which yields
e D e0 .1 ˇL C ˇB/; (12.44)
since there are two propagators associated with each vertex. Due to the Ward identity
(12.24), it then follows that the charge renormalization due to the electron self-
energy and the vertex correction exactly cancel. This holds also in higher orders.
12.2.2.3 Photon Self-Energy
We first transform the first-order photon self-energy (4.108) to the momentum rep-
resentation, using
Z
d4 k
DF .x1 ; x3 / D eik.x1 x3 / DF .k/;
.2/4
Z
d4 k 0
DF .x4 ; x2 / D eik .x4 x2 / DF .k 0 /;
.2/4
Z
d4 q
SF .x3 ; x4 / D eiq.x3 x4 / SF .q/;
.2/4
Z
d4 q 0
SF .x4 ; x3 / D eiq .x4 x3 / SF .q 0 /: (12.45)
.2/4
The space integrations over x3 and x3 give rise to the delta functions ı 4 .k q C q 0 /
and ı 4 .k 0 q C q 0 /, yielding with the bare electron charge e02 ,
Z
d4 k
ie 2 ˛ DF .k/ i˘3;4 .k/ ie02 ˛2 DF .k/
.2/4 0 1
Z
d4 q
i˘3;4 .k/ D Tr i˛3 SF .q/ i ˛4 SF .q k/ : (12.46)
.2/4
-
1 2 1 2 3 2
+ +

Fig. 12.5 Diagram representing the first-order vacuum polarization of the single photon (first-
order photon self-energy)
The photon self-energy represents a modification of the single-photon exchange,

illustrated in Fig. 12.5,
ie02 DF .k/ ) ie02 DF .k/ C ie02 DF .k/ i˘ .k/ ie02 DF .k/ C : (12.47)
With the form (4.28) of the photon propagator in the Feynman gauge, this becomes
ie02 g ie02 g ie02 g ie02 g

) C i˘ .k/ :
c0 k C i
2 c0 k C i
2 c0 k 2 C i c0 k 2 C i
(12.48)
From the Lorentz covariance, it follows that the polarization tensor must have
the form
˘ .k/ D g A.k 2 / C k k B.k 2 / (12.49)
and it can be shown that in this case only the second term can contribute [4,
p. 155], [17, p. 184]. This reduces the expression above to

ie02 g ie02 g e02 A.k 2 /
) 1
c0 k 2 C i c0 k 2 C i c0 k 2 C i
ie02 g
: (12.50)
c0 e02
k2 C c 0 A.k 2 / C i
The expression above represents the modification of the photon propagator due
to the photon self-energy. It is infinite and can be interpreted as a change of the
electronic charge – or charge renormalization – in analogy with the mass renor-
malization treated above.
The photon propagator has a pole at k 2 D 0, corresponding to the zero photon
mass [c.f. the free-electron propagator in (12.5)], and the pole value is proportional
to the electron charge squared, e0 2. If
A.k 2 D 0/ D 0; (12.51)
then also the modified propagator has a pole at the same place with a pole value
proportional to
e02
: (12.52)
2 2 ˇ
1 C 0 dA.k2 / ˇ 2
e
c 0 dk k D0
This cannot be distinguished from the bare charge and represents the physical elec-
tron charge,

2 e02 2 e02 dA.k 2 / ˇˇ
e D e0 1 ; (12.53)
2 ˇ c0 d k 2 k D0
2
e2
1 C 0 dA.k2 / ˇ 2
c 0 dk k D0
which is the charge renormalization.

The polarization tensor may have a finite part that vanishes at k 2 D 0, ˘ren ,
which is the renormalized photon self-energy. This is physically observable.
12.2.2.4 Higher-Order Renormalization
The procedure described above for the first-order renormalization can be extended
to higher orders. A second-order procedure has been described by Labzowsky and
Mitrushenkov [14] and by Lindgren et al. [15], but we shall not be concerned with
that further here.
12.3 Bound-State Renormalization: Cutoff Procedures
Before applying the renormalization procedure, the divergent integrals have to mod-
ified so that they become finite, which is the regularization procedure. Details
of this process depend strongly on the gauge used. Essentially, all QED calcula-
tions performed so far have been carried out in the so-called covariant gauges (see
Appendix G), preferably the Feynman gauge. In the remaining sections of this chap-
ter, we shall review some of the procedures used in that gauge, and also consider the
question of regularization in the Coulomb gauge.
Several regularization procedures have been developed, and the conceptually
simplest ones are the cutoff procedures. The most well known of these procedure
is that of Pauli–Willars and another is the so-called partial-wave regularization. A
more general and more sophisticated procedure is the dimensional regularization,
which has definite advantages and is frequently used today. We shall consider some
of these processes in the following sections.
12.3.1 Mass Renormalization
When we express the Dirac Hamiltonian (2.108) with the physical mass
„D D c˛ pO C ˇmc 2 C vext ; (12.54)

12.3 Bound-State Renormalization: Cutoff Procedures 249
Fig. 12.6 Diagram

r
representing the 2 r 6
renormalization of the
first-order self-energy of a
bound electron
t -
1 a 6
a
we have to include the mass counterterm (12.32) in the perturbation density (6.35)
H.x/ D ec O .x/ ˛ A .x/ O .x/ ımc 2 O .x/ˇ O .x/: (12.55)
The bound-electron self-energy operator is given by (8.47)

ˇZ ˇ
ˇ dz bou ˇ
hrj˙ bou ."a /jai D rt ˇˇ iSF ."a zI x 2 ; x 1 / I.zI x2 ; x 1 /ˇˇ ta (12.56)
2
and subtracting the corresponding mass-counterterm yields the renormalized self-

energy operator
˝ ˇ ˇ ˛
hrj˙ren
bou
."a /jai D r ˇ˙ bou ."a / ˇımc 2 ˇ a (12.57)
illustrated in Fig. 12.6. Here, both terms contain singularities, which have to be elim-
inated, which is the regularization process.
In the regularization process due to Pauli and Villars [21], [17, Eq. 9.21], the
following replacement is made in the photon propagator
1 1 1
) 2 ; (12.58)
k 2 C i k 2 C i k 2 2 C i
which cuts off the ultraviolet and possible infrared divergence.
12.3.2 Evaluation of the Mass Term
(See Mandl and Shaw [17, Sect. 10.2].)

6 D mc, the free-electron self-energy (12.15) becomes [17,
On the mass shell, p
Eq. 10.16]
Z
ˇ
2 2e 2 d4 k 6 k C mc 1
ım c D ˙ free .p/ j6pDmc D i : (12.59)
c 0 .2/ k 2pk C i k C i
4 2 2
In order to evaluate this integral, we apply the Pauli–Villars regularization scheme,

which we can express as
Z 2
1 1 1 dt
) 2 D : (12.60)
k 2 C i k 2 C i k 2 2 C i 2 .k 2 t C i/2
By means of the identity (J.4) in Appendix J with a D k 2 t and b D k 2 2pk

we can express the mass term
Z Z 2 Z
4ie 2 d4 k 1
.6 k C mc/x
ım c 2 D dt dx : (12.61)
0 .2/4 2 0 Œk 2 2pk.1 x/ tx3
With the substitutions q D p.1 x/ and s D tx, the k integral becomes, using
the integral (J.8) and (J.9) and 6 p D mc,
Z
d4 k .6 k C mc/x i mc x.2 x/
D (12.62)
.2/ Œk C 2qp C s
4 2 3 32 m c 2 .1 x/2 C tx
2 2
yielding
Z
e 2 mc 1
x C m2 c 2 .1 x/2
2
ımc 2 D dx .2 x/ ln : (12.63)
8 2 0 0 2 x C m2 c 2 .1 x/2
This is logarithmically divergent as ! 1 with the leading term being

Z 1
e 2 mc
2
2
x
ımc D dx .2 x/ ln C ln : (12.64)
8 2 0 0 m2 .1 x/2
To evaluate the second part of the integral, we need the following formulas
Z Z
x 2 ln x x 2
dx ln x D x ln x x dx x ln x D ; (12.65)
2 4
which leads to Z 1
x 3
I D dx .2 x/ ln
D : (12.66)
0 .1 x/ 2 4
ı
In all unit systems with „ D 1, the factor e 2 40 D c ˛, where ˛ is the fine-
structure constant (see Appendix K), and the mass term (12.59) becomes

3˛ mc 2 1
ım. / c D 2
ln C : (12.67)
2 mc 4
12.3.3 Bethe’s Nonrelativistic Treatment
Bethe’s original nonrelativistic treatment of the Lamb shift [3] is of great historical
interest, and it also gives some valuable insight into the physical process. Therefore,
we shall briefly summarize it here.
From the relation (4.90), we have the bound-state self-energy, using the Feynman
gauge (4.55),
Z 1
e2c d sin r12
hx 2 j˙ bou
."a /jx 1 i D hx 2 j˛jti htj˛ jx 1 i;
40 r12 0 "a "t c sgn"t
(12.68)
where r12 D jx1 x 2 j. For small k values and positive intermediate states, this
reduces to
Z 1
e2c d
˙ ."a / D 2 ˛jti
bou
htj˛ : (12.69)
4 0 0 " a "t c
The scalar part of ˛ ˛ cancels in the renormalization, leaving only the vector part
to be considered,
Z 1
e2c d
˙ bou ."a / D ˛jti htj˛: (12.70)
4 2 0 0 "a "t c
The corresponding operator for a free electron in the state pC is

Z 1
e2c ˛ d ˝
˙ free .pC / D ˛jqC q j˛ (12.71)
2
4 0 0 "pC "qC c C
restricting the intermediate states to positive energies. In the momentum representa-

tion, this becomes
Z
˝ ˇ ˇ ˛ e2 c ˝ 0 ˛ 1
d ˝ ˛
p0C ˇ˙ free .pC /ˇpC D pC j˛jqC qC j˛jpC :
2
4 0 0 "pC "qC c
(12.72)
But since ˛ is diagonal with respect to the momentum, we must have q D p D p0 .
Thus,
Z
˝ ˇ ˇ ˛ e 2 ˇˇ˝ ˛ˇ2 1
p0C ˇ˙ free .pC /ˇpC D ıp30 ;p 2
pC j˛jpC ˇ d: (12.73)
4 0 0
Obviously, this quantity is infinite. Inserting a set of complete states, this becomes
Z
˝ ˇ ˇ ˛ e2 ˝ ˛ 1
p0C ˇ˙ free .pC /ˇpC D ıp30 ;p p j˛jti htj˛jpC d: (12.74)
4 2 0 C 0
The free-electron self-energy operator can then be expressed as

Z 1
e2
˙ free .pC / D ıp30 ;p ˛jti d htj˛; (12.75)
4 2 0 0
which should be subtracted from the bound-electron self-energy operator (12.70).

We can assume the intermediate states ftg to be identical to those in the bound case.
This gives the renormalized self-energy operator
Z 1
e2 "a "t
bou
˙ren ."a / D ˛jti d htj˛ : (12.76)
2
4 0 0 "a "t c
The expectation value of this operator in a bound statejai yields the renormalized
bound-electron self-energy in this approximation, i.e., the corresponding contribu-
tion to the physical Lamb shift,
Z 1
e2 "a "t
haj ˙ren
bou
."a / jai D haj˛jti htj˛jai d : (12.77)
4 2 0 r12 0 " a "t c
This result is derived in a covariant Feynman gauge, where the quantized radia-
tion has transverse as well as longitudinal components. In the Coulomb gauge, only
the former are quantized. Since all three vector components above yield the same
contribution, we will get the result in the Coulomb gauge by multiplying by 2/3.
Furthermore, in the nonrelativistic limit, we have ˛ ! p=c, which leads to
Z 1
e2 "a "t
haj ˙ren
bou
."a / jai D hajpjti htjpjai d ; (12.78)
6 c 2 0 r12
2
0 "a "t c
which is essentially the result of Bethe.

Numerically, Bethe obtained the value 1,040 MHz for the shift in the first excited
state of the hydrogen atom, which is very close to the value 1,000 MHz obtained
experimentally by Lamb and Retherford. Later, the experimental shift has been
determined to be about 1,057 MHz. Bethe’s results was, of course, partly fortuitous,
considering the approximations made. However, it was the first success perfor-
mance of a renormalization procedure and represented a breakthrough in the theory
of QED.
We can note that the nonrelativistic treatment leads to a linear divergence of
the self-energy, while the relativistic treatment above gives only a logarithmic
divergence.
12.3.4 Brown–Langer–Schaefer Regularization
The bound-state electron propagator can be expanded into a zero-potential term, a

one-potential term, and a many-potential term
a a 6 a 6 a 6
2 2 2 2
6 x 6
4
t = t6 + x 3 + x 3
6
1 1 1 16
a a 6 a 6 a 6
Fig. 12.7 Expanding the bound-state self-energy in free-electron states according to (4.107)
SFbou .!; x 2 ; x 1 / D SFfree .!; x 2 ; x 1 /

Z
C d3 x 3 SFfree .!; x 2 ; x 3 / v.x 3 /SFfree .!; x 3 ; x 1 /
Z
C d3 x 3 d3 x 4 SFfree .!; x 2 ; x 4 / v.x 4 /
SFbou .!; x 4 ; x 3 / v.x 3 /SFfree .!; x 3 ; x 1 /; (12.79)
which leads to the expansion of the bound-electron self-energy, as illustrated in

Fig. 12.7,
D ˇ Z dz ˇ E
ˇ ˇ
haj˙ bou ."a /jai D at ˇ SFfree ."a zI x 2 ; x 1 / I.zI x2 ; x 1 / ˇta
2
D ˇZ Z
ˇ dz free
C at ˇ d3 x 3 S ."a zI x 2 ; x 3 / v.x 3 /
2 F
ˇ ED ˇ Z
ˇ ˇ
CSFfree ."a zI x 3 ; x 1 / I.zI x2 ; x 1 / ˇta at ˇ d3 x 3 d3 x 4
Z Z
dz
d3 x 3 d3 x 4 SFfree .!; x 2 ; x 4 / v.x4 /
2
ˇ E
ˇ
SFbou .!; x 4 ; x 3 / v.x3 /SFfree .!; x 3 ; x 1 / I.zI x 2 ; x 1 / ˇta ;
(12.80)
where I.zI x 2 ; x 1 / represents the single-photon interaction (4.45). We can then ex-
press this as
haj˙ bou ."a /jai D haj˙ free ."a /jai hajecA free
."a /jai C haj˙mpjai :
(12.81)
Here, the first term on the right-hand side is the average of the free-electron
self-energy in the bound state jai, the second term a vertex correction (4.98) with
v.x/ D e˛ A , and the last term the “many-potential” term.
We can now use the expansion (12.37) of the free-electron self-energy in (12.81),
where the first term (A) will be eliminated by the mass-counterterm in (12.57). We
are then left with the average of the mass-renormalized free-electron self-energy
(12.34), which is still charge divergent. If we separate the vertex operator in a di-
vergent and a renormalized part according to (12.42), it follows from (12.44) that
the charge-divergent parts cancel, and we are left with three finite contributions,
the mass-renormalized free-electron self-energy (12.34), the many-potential term
(12.79), and the finite part of the vertex correction (12.42)
hrj˙ren
bou
."a /jai D hrj˙ren
free
."a /jai hrjeA free;ren
."a /jai C hrj˙mpjai :
(12.82)
This is the method of Brown et al. [6], introduced already in 1959. It was first applied
by Brown and Mayers [7] and later by Desidero and Johnson [10], Cheng et al. [8,9],
and others. The problem in applying this expression lies in the many-potential term,
but Blundell and Snyderman [5] have devised a method of evaluating this terms
numerically with high accuracy (and the remaining terms analytically).
We can also express the renormalized, bound self-energy (12.57) as
hrj˙ren
bou
."a /jai D hrj˙ bou ."a /jai hrj˙ free ."a /jai
C hrj˙ free ."a /jai hrjˇımc 2jai ; (12.83)
where the second term is the renormalized free-electron self-energy (12.34), eval-
uated between bound states. This is illustrated in Fig. 12.8. The mass term can be
evaluated by expanding the bound states in momentum representation
˝
hrjˇımc 2jai D hrjp0 ; r 0 i p0 ; r 0 j˙ free ."p /jp; ri hp; rjai (12.84)
as illustrated in Fig. 12.9. The relation (12.83) can then be written as

˝ ˇ ˇ ˛ ˝ ˇ ˝
hrj˙ren
bou
."a /jai D r ˇ˙ bou ."a / ˙ free ."a /ˇa C r ˇp0 ; r 0 i p0 ; r 0 j˙ free ."a /
ˇ ˛
˙ free ."p /jp; ri hp; r ˇa ; (12.85)
where we note that in the last term the energy parameter of the self-energy operator
is equal to the energy of the free particle.
r r 6 r 6
2 2 2 r 6
( t - t 6 ) + ( t 6 - )
1 1 1 a 6
a a 6 a 6
Fig. 12.8 Illustration of the method of Brown et al.

p0 ; r 0 6
r 6 2
= hrjp0 ; r 0 i q; s
6 z; k hp; rjai
a 6 1
p; r
6
Fig. 12.9 Expansion of the mass term in momentum space
In this way, the leading mass-divergence term is eliminated, while the parts are
still charge-divergent, but this divergence is cancelled between the parts. The elimi-
nation of the mass-renormalization improves the numerical accuracy.
Mohr has developed the method further and included also the one-potential
part of the expansion in the two parts, thereby eliminating also the charge diver-
gence. In this way, very accurate self-energies have been evaluated for hydrogenic
systems [18–20].
12.3.5 Partial-Wave Regularization
An alternative scheme for regularizing the electron self-energy is known as the

partial-wave regularization (PWR), introduced independently by the Gothenburg
and Oxford groups [16, 24].
12.3.5.1 Feynman Gauge
Using the expansion (10.7)

X 1
sin r12
D .2l C 1/jl .r1 /jl .r2 / C l .1/ C l .2/; (12.86)
r12
lD0
the expression (4.90) for the bound-state self-energy in the Feynman gauge can be
expressed as
1 Z 1
e2 X ˛ jl .r/C l jti htj˛ jl .r/C l
˙ ."a / D 2
bou
.2l C 1/ c d
4 0 0 "a "t c sgn."t /
lD0
(12.87)
with a summation over the intermediate bound state jti. Similarly, for the free
electron
1 Z 1
e2 X ˛ jl .r/C ljq; si hq; sj˛ jl .r/C l
˙ .!/ D 2
free
.2l C 1/ c d
4 0 0 ! "q c sgn."q /
lD0
(12.88)
summed over free-electron statesjq; ri. Here, ! is the free-running energy parameter
and "q represents
p the energy of the free-electron state jq; si. On the mass shell,
! D "p D c p C m2 c 4 , this becomes
2 2
1 Z 1
e2 X l
˛ jl .r/C jq; si hq; sj˛ jl .r/C l
˙ free ."p / D .2l C 1/ c d :
2
4 0 0 "p "q c sgn."q /
lD0
(12.89)
The PWR can be combined with the Brown–Langer–Schaefer method discussed

above by expanding the remaining terms in (12.83) in a similar way.
The free-electron self-energy is diagonal with respect to the momentum p, when
all partial waves are included, but this is NOT the case for a truncated sum. Further-
more, it has nondiagonal elements with respect to the spinor index r.
The partial-wave regularization has not yet been applied in the Coulomb gauge,
but to be able to include the self-energy in a many-body calculation this will be
necessary.
In analogy with the Feynman-gauge result (12.87), the transverse part of the
self-energy in Coulomb gauge becomes
1 Z 1
e2 X
˙ bou;trans ."a / D 2 .2l C 1/ c d
4 0 0
lD0
˛jl .r/C ljti htj˛jl .r/C l .˛ r /jl .r/C ljti htj.˛ r /jl .r/C l = 2

"a "t c sgn."t /
(12.90)
using the expression (4.92). The corresponding mass term becomes in analogy with
(12.89)
1 Z 1
e2 X
˙ free;trans ."p / D .2l C 1/ c d
4 2 0 0
lD0
˛jl .r/C ljq; si hq; sj˛jl .r/C l .˛ r /jl .r/C ljq; si hq; sj.˛ r /jl .r/C l = 2
:
"p "q;s c sgn."q /
(12.91)
The Coulomb part in Coulomb gauge is obtained similarly from (4.93)
X1
e2
˙."a /bou;Coul D sgn." t / .2l C 1/
8 2 0 r12
lD0
Z 1
2 d jl .r/ C l jti htjjl .r/C l (12.92)
0
12.4 Dimensional Regularization in Feynman Gauge 257
using the value i sgn."t /=2 for the integral, and the corresponding mass term
X1
e2
˙."a /free;Coul D sgn." t / .2l C 1/
8 2 0 r12
lD0
Z 1
2 d jl .r/ C l jq; si hq; sjjl .r/C l : (12.93)
0
The main advantage of the PWR is that the bound- and free-electron self-energies
are calculated in exactly the same way, which improves the numerical accuracy,
compared to the standard procedure, where the mass term is evaluated analytically
(12.67). Since all terms are here finite, no further regularization is needed. The
maximum L value, Lmax , is increased until sufficient convergence is achieved. This
scheme has been successfully applied in a number of cases [16, 24].
It has been shown by Persson et al. [22] that the method of PWR gives the correct
result in lowest order with an arbitrary number of Coulomb interactions, while a
correction term is needed when there is more than one magnetic interaction. This is
due to the double summation over partial waves and photon momenta, which is not
unique due to the infinities involved. This problem might be remedied by combining
this method with dimensional regularization, as will be briefly discussed at the end
of the chapter.
12.4 Dimensional Regularization in Feynman Gauge
The most versatile regularization procedure developed so far is the dimensional reg-
ularization, which is nowadays frequently used in various branches of field theory.
In treating the number of dimensions (D) as a continuous variable, it can be shown
that the integrals of the radiative effects are singular only when D is an integer. Then
by choosing the dimensionality to be 4 , where is a small, positive quantity, the
integrals involved will be well defined and finite. After renormalization, one lets the
parameter ! 0. This method has been found to preserve the gauge invariance and
the validity of the Ward identity to all orders. The method was developed mainly
by t’Hooft and Veltman in the 1970s [27] (see, for instance, Mandl and Shaw [17,
Chap. 10], Peskin and Schroeder [23, Chap. 7] and Snyderman [25]).
Most applications are made in so-called covariant gauges, where the procedure is
now well developed. For our purpose, however, it is necessary to apply the Coulomb
gauge, and here the procedure is less developed. Important contributions have been
made more recently, though, by Adkins [1, 2], Heckarthon [11], and others.
Here, we shall first illustrate the method by evaluating the renormalized free-
electron self-energy, using the Feynman gauge. The problem with the Coulomb
gauge will be discussed in the next section.
12.4.1 Evaluation of the Renormalized Free-Electron

Self-Energy in Feynman Gauge
We start now from the form (12.12) of the free-electron self-energy in the Feynman
gauge and the photon propagator in momentum space (4.28)
Z
d4 k 6 p 6 k C mc
ˇ˙ free .p/ D ie 2 c 2 DF .k/
.2/4 .p k/2 m2 c 2 C i
Z
ie 2 c d4 k 6 p 6 k C mc 1
D 2 :
0 .2/ 4 .p k/ m c C i
2 2 2 k C i
(12.94)
Using the Feynman integral (J.2) (second version) with a D k 2 and b D .p k/2
m2 c 2 , this can be expressed as
Z Z
ie 2 c 1
d4 k .6 p 6 k C mc/
ˇ˙ free .p/ D dx : (12.95)
0 0 .2/ Œk 2 C .p 2 2pk m2 c 2 /x2
4
We shall now evaluate this integral using nonintegral dimension D D 4 ,

Z Z
ie 2 c 1
dD k .6 p 6 k C mc/
ˇ˙ free
.p/ D dx
0 0 .2/D Œk 2 C .p 2 2pk m2 c 2 /x2
Z Z
2ie 2 c 1
dD k .1 =2/.6 p 6 k/ .2 =2/ mc
D dx ;
0 0 .2/D Œk 2 C .p 2 2pk m2 c 2 /x2
(12.96)
after applying the anticommutation rules for the gamma matrices in Appendix D.59.
With the substitutions q D px and s D .p 2 m2 c 2 /x, this becomes
Z Z
2ie 2 c 1
dD k .1 =2/.6 p 6 k/ .2 =2/ mc
ˇ˙ free
.p/ D dx ;
0 0 .2/D Œk 2 C 2kq C s2
(12.97)
which is of the form of (G.23) and (G.24). This leads to

Z
2e 2 c.mc/ 1
.=2/
ˇ˙ free
.p/ D dx
0 .4/D=2 0 .2/
=2
m2 c 2
Œ.1 =2/.6 p 6 p x/ .2 =2/ mc
w
(12.98)
with w D q 2 s D m2 c 2 p 2 .1 x/ x. The Gamma function can be expanded

as (see Appendix G.3)
2
.=2/ D E C

with E D 0:5722 : : : being Euler’s constant, and furthermore
1 1
D 1 C ln 4 C ;
.4/D=2 .4/2 2
2 2 =2
m c w
D 1 ln C :
w 2 m2 c 2
This yields
=2
.=2/ m2 c 2 1
D .2= E C /
.4/D=2 w .4/2

1 C ln 4 C 1 ln w=m2 c 2 C
2 2
1 h w i
D ln C ; (12.99)
.4/2 m2 c 2
where
2
D E C ln 4 C : (12.100)

This leads to
Z 1 h w i
ˇ˙ free .p/ D 2K dx .6 p 6 p x 2mc/ ln 2 2
C
0 m c
Z 1
dx .6 p 6 p x C mc/
0
(12.101)
with
e2c c2 ˛
KD 2
D :
0 .4/ 4
We write the denominator in (12.98) as
p2
w D m2 c 2 xX I X D1 .1 x/ D Œ C .1 /x
m2 c 2
with
p 2 m2 c 2
D : (12.102)
m2 c 2
We then express the integral (12.101) as 2K.A C B C C / with

Z 1 Z 1
AD dx .6 p 6 p x 2mc/ C dx .6 p 6 p x mc/ ;
0 0
Z 1
BD dx .6 p 6 p x 2mc/ ln x;
0
Z 1
C D dx .6 p 6 p x 2mc/ ln Œ C .1 /x :
0
To evaluate this integral, we can use the formulas (12.65), yielding
Z 1 Z 1
dx ln.1 x/ D 1 dx x ln.1 x/ D 3=4;
0 0
Z 1
ln
dx ln Œ C .1 /x D 1 ;
0 1
Z 1
ln 1
dx x ln Œ C .1 /x D 1C 1C2 2
ln 2
;
0 .1 / 1 4.1 /2
which gives
A D .6 p =2 2mc/ C .6 p =2 mc/
B D 3 6 p =4 C 2mc

6p ln 6p
C D 1C C 1 C 2 2 ln 2
.1 / 1 4.1 /2

ln
.6 p 2m/ 1 C
1

1 ln 2
ln 1C
D6p 1C C C
.1 / 1 2.1 /2 4.1 /

ln
C2mc 1 C :
1
Subtracting the on-the-mass-shell value (6 p D m; D 0) yields for the A and B

terms

6 p mc 1
.A C B/ren D C :
2 2
For the C term, the on-shell value is 5mc=4, yielding

.2 / ln 3 2 ln 3
C ren
D 6p C C C mc C :
2.1 / 2 .1 / 4 1 4
The total on-shell value (mass-counter term) becomes
mc 2 ˛
ımc 2 D .3 C 4 C / : (12.103)
4
Collecting all parts, we obtain the following expression for the mass-
renormalized free-electron self-energy

c2 ˛ .2 / ln
ˇ˙ free .p/ D .6 p mc/ C 2 C C
4 1 .1 /2

mc 23
C 1 ln
1 1
(12.104)
with D .6 p 2 m2 c 2 /=m2 c 2 . This agrees with the result of Snyderman [25, 26].
12.4.2 Free-Electron Vertex Correction in Feynman Gauge
Next, we consider the free-electron vertex correction (12.23)

Z
ie 2 d4 k 6 p 0 6 k C mc
.p 0 ; p/ D
0 .2/4 .p 0 k/2 m2 c 2 C i
6 p 6 k C mc 1
2 : (12.105)
.p k/ m c C i
2 2 2 k C i
The Feynman parametrization (J.4), similar to the self-energy case, a D k 2 , b D

.k p/2 m2 c 2 , and c D .k p 0 /2 m2 c 2 , yields
Z 1 Z 1x Z
2ie 2 d4 k
.p 0 ; p/ D dx dy
0 0 0 .2/4
.6 p 0 6 k C mc/ .6 p 6 k C mc/
:
Œk 2 C .p 2 2pk m2 c 2 /x C .p 02 2p 0 k m2 c 2 /y3
With q D .px C p 0 y/ and s D p 2 x C p 02 y m2 c 2 .x C y/, the denominator is

of the form k 2 C 2kq C s
Z Z 1x Z
2ie 2 1 dD k .6 p 0 6 k C mc/ .6 p 6 k C mc/
.p 0 ; p/ D dx dy
0 0 0 .2/D .k 2 C 2kq C s/3
2 Z 1 Z 1x
2ie
D dx dy ŒC0 C C1 C C2 ;
c0 0 0
where the index indicates the power of 6 k involved,

Z
dD k .6 p 0 C mc/ .6 p C mc/
C0 D ;
.2/D .k 2 C 2kq C s/3
Z
dD k . 6 k/ .6 p C mc/ C .6 p 0 C mc/ . 6 k/
C1 D ;
.2/D .k 2 C 2kq C s/3
Z
dD k 6 k 6 k
C2 D :
.2/ .k 2 C 2kq C s/3
D
The coefficients C0 and C1 are convergent and we can let ! 0. With the
formula (G.23) (n D 3) and the contraction formulas (D.59), we then have
i .6 p C mc/ .6 p 0 C mc/
C0 D
.4/2 w
with
w D s q 2 D s .px C p 0 y/2 :
Similarly, we have for the numerator in C1
. 6 k/ .6 p Cmc/ C .6 p 0 Cmc/ . 6 k/ D 2.6 p 6 kC 6 k 6 p 0 /8mck
and with (G.24)
i 6 p 6 qC 6 q 6 p 0 4mcq
C1 D :
.4/2 w
The C2 coefficient is divergent and has to be evaluated with more care. Then the
situation is analogous to that of the self-energy (12.98). The numerator becomes
6 k 6 k D .2 / 6 k 6 k e e
6 ke 6k
and
Z
dD k .2 / 6 k 6 k C e e
6 ke 6k
C2 D ;
.2/ D .k C 2kq C s/
2 3
which can be evaluated with (G.25).

The evaluation of the integrals above is straightforward but rather tedious. The
complete result is found in [25].
12.5 Dimensional Regularization in Coulomb Gauge 263
12.5 Dimensional Regularization in Coulomb Gauge
12.5.1 Free-Electron Self-Energy in the Coulomb Gauge
For our main purpose of combining MBPT and QED, it is necessary to apply the
Coulomb gauge to take advantage of the developments in standard MBPT.
We shall first follow Adkins [1] in regularizing the free-electron self-energy in
the Coulomb gauge by expressing the bound states in terms of free-electron states.
We then start from the expression (12.12)
Z
d4 k 6 p 6 k C mc
ˇ˙ free .p/ D ie 2 c 2 DF .k/: (12.106)
.2/4 .p k/2 m2 c 2 C i
For the photon propagator, we use the expressions (4.32) and (4.36)

1 ı;0 ı;0 ki kj 1
C
DF .kI k/ D ı;i ı;j gij C 2 : (12.107)
c0 k2 k k 2 C i
The three terms in the propagator correspond to the Coulomb, Gaunt, and scalar-
retardation parts, respectively, of the interaction (4.59).
The Coulomb part of the self-energy becomes
Z
ie 2 c d4 k 0 .6 p 6 k C mc/ 0 1
; (12.108)
0 .2/ .p k/ m c C i k C i
4 2 2 2 2
Z
ie 2 c d4 k p e
e k C mc 1
D (12.109)
0 .2/ .p k/2 m2 c 2 C i k2 C i
4
using the commutation rules in Appendix (D.58). With q D p and s D p 2 m2 c 2 ,

the denominator is of the form k 2 C 2kq C s and we can apply the formulas (G.26)
and (G.27) without any further substitution (n D 1). This gives with k 0 ! q 0 D
p 0 , ki ! qi y D pi y, k D i ki ! py, and w D p 2 y 2 C .1 y/yp02
.p 2 m2 c 2 /y
Z
ie 2 c .mc/ dD k p e
e k C mc 1
0 .2/ k C 2kq C s C i k C i
D 2 2
2 Z 1
e c dy .=2/
D p Œp.1 y/ C mc :
0 .4/D=2 0 y .w=m2 c 2 / =2
Using (12.99), this yields the Coulomb contribution

Z 1
dy
K p .p .1 y/ C mc/ . ln.yX //
0 y
with K D e 2 c=.0 .4/2 / and w D m2 c 2 y X , X D 1 C .p2 =m2 c 2 /.1 y/. This

leads to
Z 1
dy
K p .p .1 y/ C mc ln y ln X
0 y
and the Coulomb part becomes (times K)
Z 1
4 32 dy
p C 2mc C p C 4mc p ..p .1 y/ C mc/ ln X:
3 9 0 y
(12.110)
The Gaunt term becomes, using (12.106) and the second term of (12.107),
Z
ie 2 c d4 k i .6 p 6 k C mc/ i 1
: (12.111)
0 .2/ .p k/2 m2 c 2 C i k 2 C i
4
The denominator is here the same as in the treatment of the self-energy in the
Feynman gauge, and we can use much of the results obtained there.4
In analogy with (12.95), we then have
Z Z
ie 2 c 1
d4 k i .6 p 6 k C mc/ i
dx
0 0 .2/ Œk 2 C .p 2 2pk m2 c 2 /x2
4
Z Z
ie 2 c 1 p Ce
d4 k .3 /mc .2 /.6 p 6 k/ e k
D dx 4 2
; (12.112)
0 0 .2/ Œk C .p 2pk m c /x
2 2 2 2
after inserting the Feynman integral (J.2) and applying the commutation rules in
Appendix (D.59).
With the substitutions k ! q D px and s D .p 2 m2 c 2 /x, the equation
above leads after applying (G.23) and (G.24) in analogy with (12.98) to
Z Z
ie 2 c 1
dD k .3 /mc .2 /.6 p 6 k/ ep Ce k
dx D 2
0 0 .2/ Œk 2 C 2kq C s/
Z 1
e2c .=2/
D dx Œ.3 /mc .2 / 6 p .1 x/ e
p .1 x/
0 .4/D=2 0 .w=/ =2
Z 1
e2c
D dx .1 x/ 3 0 p0 p C 3mc
0 .4/D=2 0
.=2/
C ..1 x/ 6 p mc/ ;
.w=/ =2
4
We use the convention that ; ; : : : represent all four components (0,1,2,3), while i; j; : : : rep-
resent the vector part (1,2,3).
12.5 Dimensional Regularization in Coulomb Gauge 265
where w D q 2 s D p 2 x 2 .p 2 m2 c 2 /x D m2 c 2 xY . This yields (times K)

Z 1 ˚
dx .1 x/ 3 0 p0 p 3mc Œ ln.xY /
0
2 ..1 x/ 6 p mc/g
using the relation (12.99) and the fact that ! 2 as ! 0. Then the Gaunt
part becomes

1 5 1
3 p0 p 3mc 0 p0 p C mc
0
2 4 4
Z 1
C dx .1 x/ 3 0 p0 p 3mc ln Y: (12.113)
0
Finally, the scalar-retardation part becomes similarly, using the third term of
(12.107) and the commutation rules (D.57),
Z
ie 2 c d4 k i ki .6 p 6 k C mc/ j kj 1 1

0 .2/ 4 .p k/ m c C i k k C i
2 2 2 2 2
Z
ie 2 c d4 k i ki j kj .6 p 6 kmc/ 2 i ki .k j pj k j kj / 1 1
D
0 .2/ 4 .pk/ m c C i
2 2 2
k k C i
2 2
Z
ie 2 c d4 k 6 p e
k mc C 2 i ki k j pj =k2 1
D
0 .2/4 .p k/2 m2 c 2 C i k 2 C i
with i ki j kj D k2 D ki ki . With the same substitutions as in the Gaunt case,
this becomes
Z Z
ie 2 c 1
dD k 6 p e
k mc C 2 i ki k j pj =k2
dx : (12.114)
0 0 .2/ Œk 2 2pkx C .p 2 m2 c 2 /x2
D
With the substitutions k ! q D px and s D .p 2 m2 /x the first part is of the

form (G.23) and (G.24) and becomes
Z 1
e2c .=2/
dx Œ6 p e
p x mc (12.115)
0 .4/D=2 0 w =2
and with (12.99)

Z 1
K dx Œ6 p e
p x mc . ln.xY // (12.116)
0
with K D e 2 c=.0 .4/2 / and w being the same as in the Gaunt case, w D q 2 s D
p 2 x 2 p 2 x C m2 c 2 x D m2 c 2 xY .
The second part of (12.114) is of the form (G.28) and becomes (ki k j ! qi q j
y D pi p j x 2 y 2 in first term, ! 12 gji D 12 ıij in second)
2
Z Z
1 1
p .=2/ .1 C =2/
K dx dy y 2 i pi p i pj p j
; j pj
0 0 w =2 w1C =2
Z 1 Z 1 i
p 2 pi p j pj xy
K dx dy y j
pj . ln.xyZ// ;
0 0 m2 c 2 Z
Z 1 Z 1
p 2p p2 xy
DK dx dy y C p . ln.xyZ//
0 0 m2 c 2 Z
with w D xy p2 xy C p02 x p 2 C m2 c 2 D xyŒp2 .1 xy/ p02 .1 x/ C

m2 c 2 D m2 c 2 Zxy.
Integration by parts of the first term yields (times K), noting that dZ=dy D
p2 x,
Z Z Z
1
p 1
1 1
p
dx y y 2p ln Z 0
C3 dx dy y p ln Z:
0 0 0
The total scalar-retardation part then becomes (with Z.y D 1/ D Y )
Z 1 Z 1
dx Œ6 p e
p x mc . ln.xY // dx 2p ln Y
0 0
Z 1 Z 1 Z 1 Z 1
p p
C3 dx dy y p ln Z C dx dy y p . ln.xyZ//
0 0 0 0
or
Z Z
1
0
1
p
dx p0 .1 x/ p.1 C x/ mc . ln x/ C dy y p
0 0
Z 1
dx 0 p0 .1 x/ p.1 x/ mc ln Y
0
Z 1 Z 1 Z 1 Z 1
p p
dx dy y p ln.xy/ C 2 dx dy y p ln.xy/
0 0 0 0
Z 1 Z 1
p
3 dx dy y p ln Z;
0 0
12.6 Direct Numerical Regularization of the Bound-State Self-Energy 267
which gives5

1 0 5 3 5
p0 p mc C 0 p0 p mc
2 6 4 36
Z 1
dx 0 p0 .1 x/ p.1 x/ mc ln Y
0
Z 1 Z 1
p
3 dx dy y p ln Z:
0 0
Summarizing all contributions yields the mass-renormalized free-electron self-

energy in Coulomb gauge
" Z 1
e2c 1 0 19 dy
.6 p mc/ p 0 C p p .p .1 y/ C mc/ ln X
0 .4/2 2 6 0 y
Z 1 Z 1 Z 1 #
p
C2 dx Œ.1 x/ 6 p mc ln Y C dx dy y 2p ln Z ;
0 0 0
(12.117)
where we have subtracted the on-shell (6 p D mc) value, mc.3 C 4/. (The expres-
sions for X; Y; Z are given in the text.) This is in agreement with the the result of
Adkins [1].
The treatment of the vertex correction is more complex and will not be repro-
duced here. Interested readers are referred to the papers by Adkins.6
12.6 Direct Numerical Regularization of the Bound-State

Self-Energy
As an alternative to the regularization procedure discussed above, we shall consider

a new more direct procedure, where the regularization is performed directly in the
bound state, without any transformation to free-electron states. This is presently
being tested by the Gothenburg group [12].
5
Z Z Z 1 Z 1
1 1
p 10 p 1
dx dy y ln.xy/ D I dx x dy y ln.xy/ D I
0 0 9 0 0 18
Z 1 Z 1
p 9
dx x dy y y ln.xy/ D :
0 0 50
6
A complete treatment of the vertex correction is being published separately in arXiv:quant-ph.
12.6.1 Feynman Gauge
The bound-state self-energy in the Feynman gauge is from (8.48)

ˇZ ˇ
ˇ c d f F ./ ˇ
hrj˙."a /jai D rt ˇˇ ˇ ta ; (12.118)
"a "t .c i/t ˇ
where D jkj and the function f F is given by (4.55). The integral over is con-
vergent, while the summation over t is (logarithmically) divergent.
With 3 dimensions of the k-vector space, we make the substitution
Z Z Z
d3 k 2 d
) d˝
.2/3 .2/3
and the self-energy becomes

ˇ Z ˇ
e 2 .2/ ˇ ˛1 ˛2 c d sin r12 ˇˇ
haj˙ ."a /jai D ˇ
at ˇ ta; (12.119)
4 2 0 r12 "a "t .c i/t ˇ
which would make the expression convergent for ¤ 0. In a similar way, the the
free-electron self-energy can be expressed. In analogy with the expression (12.85),
this leads to the renormalized bound-state self-energy
˝ ˇ bou ˇ ˛ ˝ ˇ ˇ ˛
r ˇ˙ren ."a /ˇa D lim r ˇ˙ bou ."a / ˙ free ."a /ˇa
!0
˝ ˇ ˝ ˇ ˇ ˛ ˇ ˛
C r ˇp0 ; r 0 i p0 ; r 0 ˇ˙ free ."a / ˙ free ."p /ˇp; r hp; r ˇa :
(12.120)
12.6.2 Coulomb Gauge
The transverse part of the self-energy expression in Coulomb gauge is in analogy

with the Feynman expression (12.118)
ˇZ ˇ
ˇ c d fTC ./ ˇ
hrj˙."a /jaiTrans D rt ˇˇ ˇ ta ;
ˇ (12.121)
"a "t .c i/t
where fTC is given by (4.60). The Coulomb part is given by (8.49)

ˇ Z ˇ
1 ˇ e2 2 d sin r12 ˇˇ
hbj˙."a /jaiCoul D sgn."t / bt ˇˇ 2 ˇ ta (12.122)
2 4 0 r12 2
and the renormalization can be performed as in the Feynman gauge.

The procedure of direct numerical regularization outlined here is presently being
tested by the Gothenburg group.
References 269
References
(1983)
2. Adkins, G.: One-loop vertex function in Coulomb-gauge QED. Phys. Rev. D 34, 2489–92
(1986)
3. Bethe, H.A.: The Electromagnetic Shift of Energy Levels. Phys. Rev. 72, 339–41 (1947)
(1964)
5. Blundell, S., Snyderman, N.J.: Basis-set approach to calculating the radiative self-energy in
highly ionized atoms. Phys. Rev. A 44, R1427–30 (1991)
6. Brown, G.E., Langer, J.S., Schaeffer, G.W.: Lamb shift of a tightly bound electron. I. Method.
Proc. R. Soc. London, Ser. A 251, 92–104 (1959)
7. Brown, G.E., Mayers, D.F.: Lamb shift of a tightly bound electron. II. Calculation for the
K-electron in Hg. Proc. R. Soc. London, Ser. A 251, 105–109 (1959)
8. Cheng, K.T., Johnson, W.R.: Self-energy corrections to the K-electron binding energy in heavy
and superheavy atoms. Phys. Rev. A 1, 1943–48 (1976)
9. Cheng, T.K., Johnson, W.R., Sapirstein, J.: Screend Lamb-shift calculations for Lithiumlike
Uranium, Sodiumlike Platimun, and Copperlike Gold. Phys. Rev. Lett. 23, 2960–63 (1991)
10. Desiderio, A.M., Johnson, W.R.: Lamb Shift and Binding Energies of K Electrons in Heavy
Atoms. Phys. Rev. A 3, 1267–75 (1971)
11. Heckarthon, D.: Dimensional regularization and renormalization of Coulomb gauge quantum
electrodynamics. Nucl. Phys. B 156, 328–46 (1978)
12. Hedendahl, D.: p. (private communication)
14. Labzowsky, L.N., Mitrushenkov, A.O.: Renormalization of the second-order electron self-
energy for a tightly bound atomic electron. Phys. Lett. A 198, 333–40 (1995)
15. Lindgren, I., Persson, H., Salomonson, S., Sunnergren, P.: Analysis of the electron self-energy
for tightly bound electrons. Phys. Rev. A 58, 1001–15 (1998)
16. Lindgren, I., Persson, H., Salomonson, S., Ynnerman, A.: Bound-state self-energy calculation
using partial-wave renormalization. Phys. Rev. A 47, R4555–58 (1993)
88, 52–87 (1974)
19. Mohr, P.J.: Self-Energy Radiative Corrections. Ann. Phys. (N.Y.) 88, 26–51 (1974)
20. Mohr, P.J.: Self-energy of the n D 2 states in a strong Coulomb field. Phys. Rev. A 26,
2338–54 (1982)
21. Pauli, W., Willars, F.: On the Invariant Regularization in Relativistic Quantum Theory. Rev.
Mod. Phys. 21, 434–44 (1949)
22. Persson, H., Salomonson, S., Sunnergren, P.: Regularization Corrections to the Partial-Wave
Renormalization Procedure. Presented at the conference Modern Trends in Atomic Physics,
Hindås, Sweden, May, 1996. Adv. Quantum Chem. 30, 379–92 (1998)
24. Quiney, H.M., Grant, I.P.: Atomic self-energy calculations using partial-wave mass renormal-
ization. J. Phys. B 49, L299–304 (1994)
25. Snyderman, N.J.: Electron Radiative Self-Energy of Highly Stripped Heavy-Atoms. Ann. Phys.
(N.Y.) 211, 43–86 (1991)
26. Sunnergren, P.: Complete One-Loop QED CAlculations for Few-Eleectron Ions. Ph.D. thesis,
Department of Physics, Chalmers University of Technology and University of Gothenburg,
Gothenburg, Sweden (1998)
27. t’Hooft, G., Veltman, M.: Regularization and renormalization of gauge fields. Nucl. Phys. B
44, 189–213 (1972)
Chapter 13
Summary and Conclusions
The all-order forms of many-body perturbation theory (MBPT), like the

coupled-cluster approach (CCA), have been extremely successful in calculations on
atomic and in particular on molecular systems. Here, the dominating parts of the
electron correlation can be evaluated to essentially all orders of perturbation theory.
A shortcoming, however, of the standard MBPT/CCA procedures is that
quantum-electrodynamics (QED) can only be included in a very limited fashion
(first-order energy). Particularly for highly charged systems, QED effects can be
quite important. Certain experimental data on such systems are now several orders
of magnitude more accurate than the best available theoretical calculation. It is
believed that this shortcoming is due to the omitted combination of many-body and
QED effects in presently available theoretical procedures.
The procedure presented in this book, which is based on quantum-field theory,
describes – for the first time – a road toward a rigorous unification of QED and
MBPT. The procedure is based on the covariant evolution operator, which describes
the time evolution of the relativistic wave function or state vector. The procedure is
for two-electron systems fully compatible with the relativistically covariant Bethe–
Salpeter equation, but it is more versatile.
The procedure is – in contrast to the standard Bethe–Salpeter equation – appli-
cable to a general multi-reference (quasi-degenerate) model space. It can also be
combined with the coupled-cluster approach and is, in principle, applicable to sys-
tems with more than two electrons.
The covariant evolution operator contains generally singularities that can be
eliminated. The regular part is referred to as the Green’s operator, which is a gen-
eralization of the Green’s-function concept.
In principle, the Green’s operator – as well as the Bethe–Salpeter equation – has
individual times for the particles involved. Most applications, though, use the equal-
time approximation, where the times are equalized, to make the procedure consistent
with the quantum-mechanical picture.
The Green’s operator for time t D 0 corresponds to the wave operator used in
standard MBPT, and the time derivative at t D 0, operating within the model space,
to the many-body effective interaction. This connects the field-theoretical procedure
with the standard MBPT.

272 13 Summary and Conclusions
The formalism presented here has been partially tested numerically by the
Gothenburg atomic theory group, and in cases where comparison can be made with
the more restricted S-matrix formulation, very good agreement is reported.
A big challenge is the renormalization of the radiative effects, which generally
has to be performed using the Coulomb gauge, to take advantage of the develop-
ments in MBPT/CCA. Schemes have been developed for this process but so far not
been implemented in a QED-MBPT procedure.
When the procedure is more developed, critical tests can be performed to find
out to what extent the new effects will improve the agreement between theory and
accurate experimental data.
Appendix A
Notations and Definitions
A.1 Four-Component Vector Notations
A four-dimensional contravariant vector is defined as1
x D x D .x 0 ; x 1 ; x 2 ; x 3 / D .x 0 ; x/ D .ct; x/; (A.1)
where D0; 1; 2; 3 and x is the three-dimensional coordinate vector xD.x 1 ; x 2 ; x 3 /

.x; y; z/. The four-dimensional differential is
d 4 x D cdt d3 x and d3 x D dx dy dz:
A corresponding covariant vector is defined as
x D .x0 ; x1 ; x2 ; x3 / D g x D .x0 ; x/ D .ct; x/; (A.2)
which implies that
x 0 D x0 x i D xi .i D 1; 2; 3/; (A.3)
g is a metric tensor, which can raise the so-called Lorentz indices of the vector.
Similarly, an analogous tensor can lower the indices
x D g x : (A.4)
These relations hold generally for four-dimensional vectors.

There are various possible choices of the metric tensors, but we shall use the
following
0 1
1 0 0 0
B 0 1 0 0 C
g D g D B @ 0 0 1 0 A :
C (A.5)
0 0 0 1
1
In all appendices, we display complete formulas with all fundamental constants. As before, we
use the Einstein summation rule with summation over repeated indices.

DOI 10.1007/978-1-4419-8309-1, c Springer Science+Business Media, LLC 2011
274 A Notations and Definitions
The four-dimensional scalar product is defined as the product of a contravariant

and a covariant vector:
ab D a b D a0 b0 a b; (A.6)
where a b is the three-dimensional scalar product
a b D ax bx C ay by C az bz :
The covariant gradient operator is defined as the gradient with respect to a con-
travariant coordinate vector:

@ 1 @
@ D D ;r (A.7)
@x c @t
and the contravariant gradient operator analogously

@ 1 @
@ D

D ; r ; (A.8)
@x c @t
r is the three-dimensional gradient operator
@ @ @
rD eO x C eO y C eO z ;
@x @y @z
where .eO x ; eO y ; eO z / are unit vectors in the coordinate directions.

The four-dimensional divergence is defined as
1 @A0
@ A D C r A D rA; (A.9)
c @t
where r A is the three-dimensional divergence
@Ax @Ay @Az

r A D C C :
@x @y @z
The d’Alembertian operator is defined as
1 @2
D @ @ D r 2 D r 2; (A.10)
c 2 @t 2
where
@2 @2 @2
r2 D D C C
@x 2 @y 2 @z2
is the Laplacian operator.
A.2 Vector Spaces 275
A.2 Vector Spaces
A.2.1 Notations
X; Y; : : : are sets with elements x; y; : : :.

x 2 X means that x is an element in the set X .
N is the set of nonnegative integers. R is the set of real numbers. C is the set of
complex numbers.
R n is the set of real n-dimensional vectors. Cn is the set of complex n-dimensional
vectors.
A X means that A is a subset of X .
A [ B is the union of A and B. A \ B is the intersection of A and B.
A D fx 2 X W P g means that A is the set of all elements x in X that satisfy the
condition P .
f W X ! Y represents a function or operator, which means that f maps uniquely
the elements of X onto elements of Y .
A functional is a unique mapping f W X ! R .C/ of a function space on the space
of real (complex) numbers.
The set of arguments x 2 A for which the function f W A ! B is defined is the
domain, and the set of results y 2 B which can be produced is the range.
.a; b/ is the open interval fx 2 R W a < x < bg. Œa; b is the closed interval
fx 2 R W a x bg.
sup represents the supremum, the least upper bound of a set
inf represents the infimum, the largest lower bound of a set.
A.2.2 Basic Definitions
A real (complex) vector space or function space X is an infinite set of elements, x,

referred to as points or vectors, which is closed under addition, x C y D z 2 X , and
under multiplication by a real (complex) number c, cx D y 2 X . The continu-
ous functions f .x/ on the interval x 2 Œa; b form a vector space, also with some
boundary conditions, like f .a/ D f .b/ D 0.
A subset of X is a subspace of X if it fulfills the criteria for a vector space.
A norm of a vector space X is a function p W X ! Œ0; 1 with the properties
(1) p.x/ D jjp.x/
(2) p.x C y/ p.x/ C p.y/ for all real ( 2 R) and all x; y 2 X
(3) that p.x/ D 0 always implies x D 0
The norm is written as p.x/ D jjxjj. We then have jjxjj D jj jjxjj , jjx C yjj
jjxjj C jjyjj , and jjxjj D 0 ) x D 0. If the last condition is not fulfilled, it is a
seminorm.
A vector space with a norm for all its elements is a normed space, denoted by
.X; jjjj/. The continuous functions, f .x/, on the interval Œa; b form a normed
Rb
space by defining a norm, for instance, jjf jj D Œ a dt jf .t/j2 1=2 . By means of
the Cauchy–Schwartz inequality, it can be shown that this satisfies the criteria for a
norm [4, p. 93].
If f is a function f W A ! Y and A X , then f is defined in the
neighborhood of x0 2 X , if there is an > 0 such that the entire sphere
fx 2 X W jjx x0 jj < g belongs to A [4, p. 309].
A function/operator f W X ! Y is bounded, if there exists a number C such
that
jjf xjj
sup D C < 1:
0¤x2X jjxjj
Then C D jjf jj is the norm of f . Thus, jjf xjj jjf jj jjxjj .
A function f is continuous at the point x0 2 X , if for every ı > 0 there exists
an > 0 such that for every member of the set x W jjx x0 jj < we have
jjf x f x0 jj ı [4, p. 139]. This can also be expressed so that f is continuous at
the point x0 , if and only if f x ! f x0 whenever xn ! x0 , fxn g being a sequence
in X , meaning that f xn converges to f x0 , if x converges to x0 [5, p. 70].
A linear function/operator is continuous if and only if it is bounded [4, p. 197,
213], [2, p. 22].
A functional f W X ! R is convex if
f .tx C .t 1/y/ tf .x/ C .t 1/f .y/;
for all x; y 2 X and t 2 .0; 1/.

A subset A X is open, if for every x 2 A there exists an > 0 such that
the entire ball Br .x/ D fy 2 X j jjy xjj < g belongs to A, i.e., Br .x/ A [1,
p. 363], [4, p. 98], [5, p. 57].
A sequence fxn g , where n is an integer (n 2 N), is an infinite numbered list
of elements in a set or a space. A subsequence is a sequence, which is a part of a
sequence.
A sequence fxn 2 Ag is (strongly) convergent toward x 2 A, if and only if for
every > 0 there exists an N such that jjxn xjj < for all n > N [4, p. 95,
348].
A sequence is called a Cauchy sequence if and only if for every > 0 there
exists an N such that jjxn xm jj < for all m; n > N . If a sequence fxn g is
convergent, then it follows that for n; m > N
jjxm xn jj D jj.xn x/ C .x xm /jj jjxn xjj C jjxm xjj < 2;
which means that a convergent sequence is always a Cauchy sequence. The opposite
is not necessarily true, since the point of convergence need not be an element of
X [3, p. 44].
A.3 Special Functions 277
A subset A of a normed space is termed compact, if every infinite sequence of

elements in A has a subsequence, which converges to an element in A. The closed
interval [0,1] is an example of a compact set, while the open interval (0,1) is non-
compact, since the sequence 1, 1/2, 1/3. . . and all of its subsequences converge to
0, which lies outside the set [5, p. 149]. This sequence satisfies the Cauchy conver-
gence criteria but not the (strong) convergence criteria.
A dual space or adjoint space of a vector space X , denoted as X , is the space
of all functions on X .
An inner or scalar product in a vector space X is a function h; i W X X ! R
with the properties (1)
hx; 1 y1 C 2 y2 i D 1 hx; y1 i C 2 hx; y2 i ; hx; yi D hy; xi;
for all x; y; y1 ; y2 2 X and all j 2 R , and (2) hx; xi D 0 only if x D 0.
A.2.3 Special Spaces
A.2.3.1 Banach Space
A Banach space is a normed space in which every Cauchy sequence converges to a

point in the space.
A.2.3.2 Hilbert Space

p
A Banach space with the norm jjxjj D C hx; xi is called a Hilbert space [1,
p. 364].
A.2.3.3 Fock Space
A Fock space is a Hilbert space, where the number of particles is variable or un-
known.
A.3 Special Functions
A.3.1 Dirac Delta Function
We consider the integral

Z L=2
dx eikx : (A.11)
L=2
Assuming periodic boundary conditions, eiLx=2 D eiLx=2 , limits the possible k

values to k D kn D 2 n=L. Then
Z L=2
1
dx eikn x D ıkn ;0 D ı.kn ; 0/; (A.12)
L L=2
where ın;m is the Kronecker delta factor

1 if m D n
ın;m D : (A.13)
0 if m ¤ n
If we let L ! 1, then we have to add a “damping factor” e jxj , where is a

small positive number, to make the integral meaningful,
Z 1
2
dx eikx e jxj D D 2 .k/: (A.14)
1 k2 C 2
In the limit ! 0, we have

Z 1
1
lim .k/ D lim dx eikx e jxj D ı.k/; (A.15)
!0 2 !0 1
which can be regarded as a definition of the Dirac delta function, ı.k/. Formally,
we write this relation as
Z 1
dx ikx
e D ı.k/: (A.16)
1 2
The function also has the following properties
lim .x/ D ıx;0 ;

!0
Z 1
dx .x a/ .x b/ D C .a b/: (A.17)
1
In three dimensions, (A.12) goes over into
Z
1
d3 x eikn x D ı 3 .kn ; 0/ D ı.knx ; 0/ ı.kny ; 0/ ı.knz ; 0/: (A.18)
V V
In the limit where the integration is extended over the entire three-dimensional
space, we have in analogy with (A.16)
Z
d3 x ikx
e D ı 3 .k/: (A.19)
.2/3
A.3.2 Integrals over Functions
We consider the integral

Z 1 Z 1
dx ı.x a/ f .x/ D lim dx .x a/ f .x/
1 !0 1
Z 1
1 2
D lim dx f .x/: (A.20)
2 !0 1 .x a/2 C 2
The integral can be evaluated using residue calculus and leads to

Z 1
dx ı.x a/ f .x/ D f .a/ (A.21a)
1
provided the function f .x/ has no poles. In three dimensions, we have similarly
Z
d3 x ı.x x 0 / f .x/ D f .x 0 / (A.21b)
integrated over all space.

The relations above are often taken as the definition of the Dirac delta function,
but the procedure applied here is more rigorous.
Next, we consider the integral over two functions
Z Z
1 2 2
dx .x a/ .x b/ D dx
.2/2 .x a/2 C 2 .x b/2 C 2
Z
1 1 1
D dx
4 2 i x a i x a C i
2

.x b C i/.x b i/

1 1 1
D
2i b a i. C / b a C i. C /
1 2. C /
D (A.22)
2 .a b/2 C . C /2
after integrating the first term over the negative and the second term over the positive
half plane. Thus,
Z
dx .x a/ .x b/ D C .a b/ (A.23)
and we see that here the widths of the functions are added.
Now, we consider some integrals with the functions in combination with

electron and photon propagators that are frequently used in the main text.
First, we consider the integral with one function and an electron propagator
(4.10)
Z Z
1 d! 1 2
d! ."a !/ D
! "j C i 2 ! "j C i ."a !/2 C 2
Z
d! 1 2
D :
2 ! "j C i ."a ! C i /."a ! i /
The pole of the propagator yields the contribution ."a "j /, which vanishes in
the limit ! 0, if "a 6D "j . Nevertheless, we shall see that this pole has a significant
effect on the result.
Integrating above over the positive half plane, with the single pole "a C i , yields
1
"a "j C i C i
and integrating over the negative half plane, with the two poles "j i; "a i ,
yields
2i 1
C
."a "j C i C i/."a "j i C i/ "a "j i C i
1
D ;
"a "j C i C i
which is identical to the previous result. We observe here that the pole of the prop-
agator, which has a vanishing contribution in the limit ! 0, has the effect of
reversing the sign of the i term.
The parameter originates from the adiabatic damping and is small but finite,
while the parameter is infinitely small and only determines the position of the pole
of the propagator. Therefore, if they appear together, the term dominates, and the
term can be omitted. This yields
Z
1 1
d! ."a !/ D (A.24)
! "j C i "a "j C i
noting that the parameter of the propagator is replaced by the damping

parameter .
Second, we consider the integral with the photon propagator (4.31)
Z Z
1 1 d! 1 1
d! 2 ." a !/ D
! 2 C i 2 2 ! C i ! C i
2

."a ! C i /."a ! i /

1 1 1
D
2 "a C i C i "a i C i
1
D (A.25)
"2a . i i/2
or, neglecting the term,
Z
1 1
d! ."a !/ D 2 (A.26)
!2 C i
2 "a 2 C i
noting that 0.
Finally, we consider the integrals of two functions and the propagators. With
the electron propagator, we have
Z
1
d! ."a !/ ."b !/
! "j C i
Z
1 1 1 1
D d!
.2i/2 ! "j C i "a ! i "a ! C i

1 1
:
"b ! i "b ! C i
Here, three of the combinations with poles on both sides of the real axis contribute,
which yields

1 1 1
C
2i ."b "j C i /."a "b 2i / ."a "j C i /."b "a 2i /

1
:
."a "j C i /."b "j C i /
The last two terms become

1 1 1 1

"a "j C i "b "a 2i "b "j C i ."b "a 2i /."b "j C i /
neglecting an imaginary term in the numerator. This leads to
Z
1 1
d! ."a !/ ."b !/ 2 ."a "b /:
! "j C i "a "j C i
(A.27)
Similarly, we find for the photon propagator

Z
1 1
d! ."a !/ ."b !/ 2 ."a "b /:
! 2 2 C i "a 2 C i
(A.28)
Formally, we can obtain the integral with propagators by replacing the func-
tion by the corresponding Dirac delta function, noting that we then have to replace
the imaginary parameter in the denominator by the damping factor .
A.3.3 The Heaviside Step Function
The Heaviside step function is defined as
.t/ D 1 t0 > t
D0 t 0 < t: (A.29)
The step function can also be given the integral representation

Z 1
d! ei!t
.t/ D lim i (A.30)
!0 1 2 ! C i
from which we obtain the derivative of the step function

Z 1
d.t/ d! !
D lim ei!t D ı.t/ (A.31)
dt !0 1 2 ! C i
where ı.t/ is the Dirac delta function.
References
1. Blanchard, P., Brüning, E.: Variational Methods in Mathematical Physics. A Unified Approach.
Springer-Verlag, Berlin (1992)
2. Debnath, L., Mikusiński, P.: Introduction to Hilbert Spaces with Applications., second edn.
Academic Press, New York (1999)
3. Delves, L.M., Welsh, J.: Numerical Solution of Integral Equations. Clarendon Press, Oxford
(1974)
4. Griffel, D.H.: Applied Functional Analysis. Wiley, N.Y. (1981)
5. Taylor, A.E.: Introduction to Functional Analysis. Wiley and Sons (1957)
Appendix B
Second Quantization
B.1 Definitions
(See, for instance [2, Chap. 5], [1, Chap. 11].) In second quantization – also known
as the number representation – a state is represented by a vector (see Appendix C.1)
jn1 ; n2 ; i, where the numbers represent the number of particles in the particular
basis state (which for fermions can be equal only to one or zero).
Second quantization is based on annihilation/creation operators cj /cj , which
annihilate and create, respectively, a single particle. If we denote byj0i the vacuum
state with no particle, then
cjj0i Djj i (B.1)
represents a single-particle state. In the coordinate representation (C.19), this corre-
sponds to the wave function
j .x/ D hxjj i (B.2)
satisfying the single-electron Schrödinger or Dirac equation. Obviously, we have
cj j0i D 0: (B.3)
For fermions, the operators satisfy the anticommutation relations

fci ; cj g D ci cj C cj ci D 0;
fci ; cj g D ci cj C cj ci D 0;
fci ; cj g D ci cj C cj ci D ıij ; (B.4)
where ıij is the Kronecker delta factor (A.13). It then follows that
ci cjj0i D cj cij0i ; (B.5)
283
284 B Second Quantization

which means that ci ci j0i represents an antisymmetric two-particle state, which we
denote in the following way1
ci cjj0i Djfi; j gi : (B.6)
The antisymmetric form is required for fermions by the quantum-mechanical rules.

A corresponding bra state is
h0jcl ck D hfk; lgj (B.7)
and it then follows that the states are orthonormal.

In the coordinate representation, the state above becomes
1
hx 1 x 2jfi; j gi D p i .x 1 / j .x 2 / j .x 1 / i .x 2 / : (B.8)
2
Generalizing this to a general many-particle system leads to an antisymmetric prod-

uct, known as the Slater determinant,
1
hx 1 ; x 2 ; x N jca cb cN j0i D p Detfa; b; N g
NŠ
ˇ ˇ
ˇ 1 .x 1 / 1 .x 2 / 1 .x N / ˇ
ˇ ˇ
1 ˇˇ 2 .x 1 / 2 .x 2 / ˇ
2 .x N / ˇ
D p ˇ ˇ : (B.9)
NŠ ˇ ˇ
ˇ .x / .x / ˇ
N 1 N .x 2 / N N
For an N -particle system, we define one- and two-particle operators by
X
n
F D fn ; (B.10)
nD1
Xn
GD gmn ; (B.11)
m<nD1
respectively, where the fn and the gmn operators are identical, differing only in the
particles they operate on. In second quantization these operators can be expressed
(see, for instance [1, Sect. 11.1])2
1
We shall follow the convention of letting the notation ji; j i denote a straight product function
ji; j i D i .x 1 / j .x 2 /, whilejfi; j gi represents an antisymmetric function.
2
Occasionally, we use a “hat” on the operators to emphasize their second-quantized form. We also
use the Einstein summation rule with summation over all indices that appear twice. Note the order
between the annihilation operators.
B.1 Definitions 285
FO D ci hi jf jj i cj ;

1
GO D ci cj hij jgjkli cl ck ; (B.12)
2
etc. (note order between the operators in the two-particle case). Here,
Z
d3 x 1 i .x1 / f j .x1 /;
hi jf jj i D
ZZ
hij jgjkli D d3 x 1 d3 x 2 i .x1 / j .x2 / g k .x1 / l .x2 /: (B.13)
We can check the formulas above by evaluating

˝ ˇ1 ˇ ˛
O
hfcd gjGjfabgi D fcd gˇ ci cj hij jgjkli cl ck ˇfabg
2
1 ˝ ˇˇ ˇ ˛
D 0 cd cc ci cj hij jgjkli cl ck ca cbˇ0 : (B.14)
2
Normal-ordering the operators yields
cl ck ca cbj0i D ık;a ıl;b ıl;a ık;b
and similarly

h0jcd cc ci cj D ıi;d ıj;c ıj;c ıi;d :
Then we have
O
hfcd gjGjfabgi D hcd jgjabi hdcjgjabi ;
which agrees with the results using determinantal wave functions (see, for instance
[1, Eq. 5.19])
1 ˝ ˇ
O
hfcd gjGjfabgi D cd dc ˇGjab
O bai : (B.15)
2
We define the electron-field operators in the Schrödinger representation (3.1) by
O S .x/ D cj j .x/I
O .x/ D c
j .x/: (B.16)
S j
Then, the second-quantized one-body operator can be expressed as

Z Z
FO D d3 x ci
i .x/ f cj j .x/ D d3 x O S .x/ f O S .x/ (B.17)
and similarly
ZZ
1
GO D d3 x 1 d3 x 2 O S .x1 /S .x2 / g O S .x2 / O S .x1 /: (B.18)
2
286 B Second Quantization
The nonrelativistic Hamiltonian for an N -electron system (2.11) consists of a

single-particle and a two-particle operator
XN X N
„2 2
H1 D rn C vext .x n / D h1 .n/;
nD1
2m nD1
X
N
e2 XN
H2 D D h2 .m; n/; (B.19)
m<n
40 rmn m<n
and in second quantization this can be expressed as

Z
HO D d3 x 1 O S .x 1 / h1 O S .x 1 /
ZZ
1
C d3 x 1 d3 x 2 O S .x 1 / O S .x 2 / h2 O S .x 2 / O S .x 1 /: (B.20)
2
B.2 Heisenberg and Interaction Pictures
In an alternative to the Schrödinger picture, the Heisenberg picture (HP), the states
are time independent and the time dependence is transferred to the operators,
O O O
j Hi Dj S .t D 0/i D eiH t =„j S .t/i I OO H D eiH t =„ OO S eiH t =„ : (B.21)
In perturbation theory, the Hamiltonian is normally partitioned into a zeroth-

order Hamiltonian H0 and a perturbation V (2.48), which using second-quantized
notations becomes
HO D HO 0 C VO : (B.22)
We can then define an intermediate picture, known as the interaction picture (IP),
where the operators and state vectors are related to those in the Schrödinger picture
by
O O O
j I .t/i D eiH0 t =„ j S .t/i I OO I .t/ D eiH0 t =„ OO S eiH0 t =„ : (B.23)
3
The relation between the Heisenberg and the interaction pictures is
O O O O O O
j Hi D eiH t =„ eiH0 t =„j I .t/i I OO H .t/ D eiH t =„ eiH0 t =„ OO I eiH0 t =„ eiH t =„ :
(B.24)
Using the relation (3.9), we then have
j Hi D U.0; t/j I .t/i I OO H .t/ D U.0; t/ OO I U.t; 0/: (B.25)
3
Note that HO and HO 0 generally do not commute, so that in general eiHO t=„ eiHO0 t=„ ¤ eiVO t=„ .
References 287
The state vector of time-independent perturbation theory corresponds in all pictures

considered here to the time-dependent state vectors with t D 0,
j i Dj Hi Dj S .0i/ Dj I .0/i : (B.26)
In the Heisenberg picture (B.21), the electron-field operators (B.16) become
O H .x/ D eiHO t =„ O S .x/ eiHO t =„ I O H .x/ D eiHO t =„ O .x/ eiHO t =„ (B.27)

S
and in the interaction picture (IP) (B.23)
O I .x/ D eiHO 0 t =„ O S .x/ eiHO 0 t =„ D eiHO 0 t =„ cj j .x/ e

iHO 0 t =„
i"j t =„
D cj j .x/ e D cj j .x/
O I .x/ D cj
j .x/ e
i"j t =„
D cj
j .x/; (B.28)
where j .x/ is an eigenfunction of H0 . We also introduce the time-dependent cre-

ation/annihilations operators in the IP by
cj .t/ D cj ei"j t =„ I cj .t=„/ D cj ei"j t =„ ; (B.29)
which gives
O I .x/ D cj .t/ j .x/ I O I .x/ D c .t/

j .x/: (B.30)
j
From the definition (B.23), we have
@ O @ h iHO 0 t =„ O iHO 0 t =„ i h i
OI .t/ D e OS e D i H0 ; OO I .t/ : (B.31)
@t @t
References
2. Schweber, S.S.: An Introduction to Relativistic Quantum Field Theory. Harper and Row, N.Y.
(1961)
Appendix C
Representations of States and Operators
C.1 Vector Representation of States
A state of a system can be represented by the wave function or Schrödinger func-

tion .x/, where x stands for all (space) coordinates. If we have a complete basis
set available in the same Hilbert space (see Appendix A.2), f j .x/g, then we can
expand the function as
.x/ D aj j .x/ (C.1)
with summation over j according to the Einstein summation rule. If the basis set is
orthonormal, implying that the scalar or inner product satisfies the relation
Z

hijj i D dx i j .x/ D ıi;j ; (C.2)
then the expansion coefficients are given by the scalar product

Z

aj D dx j .x/ D hj j i : (C.3)
These numbers form a vector, which is the vector representation of the state or
the state vector, 0 1
h1j i
B h2j i C
ˇ ˛ B C
ˇ DB B C:
C
(C.4)
B C
@ A
hN j i
Note that this is just a set of numbers – no coordinates are involved. N is here
the number of basis states, which may be finite or infinite. [The basis set need not
be numerable and can form a continuum in which case the sum over the states is
replaced by an integral.] The basis states are represented by unit vectors jj i
289
290 C Representations of States and Operators
0 1 0 1
1 0
B0C B1C
ˇ ˛ B C ˇ ˛ B C
ˇ1 D B
B0C
C ˇ2 D B
B0C
C
etc: (C.5)
B C B C
@A @A

The basis vectors are time independent, and for time-dependent states the time
dependence is contained in the coefficients
j .t/i D aj .t/jj i ; (C.6)
j i is a ket vector, and for each ket vector there is a corresponding bra vector
˝ ˇ
ˇ D a ; a ; ; (C.7)
1 2
where the asterisk represents complex conjugate. It follows from (C.1) that
aj D h jj i: (C.8)
The scalar product of two general vectors with expansion coefficients aj and bj ,
respectively, becomes
h j˚i D aj bj (C.9)
with the basis vectors being orthonormal. This is identical to the scalar product of
the corresponding vector representations
0 1
b1
B b2 C
h j˚i D a1 ; a2 ; B C
@ A: (C.10)

The ket vector (C.4) can be expanded as
j i Djj i hj j i: (C.11)
But this holds for any vector in the Hilbert space, and therefore we have the formal
relation in that space
jj i hj j I ; (C.12)
where I is the identity operator. This is known as the resolution of the identity. Us-
ing the expression for the coefficients, the scalar product (C.9) can also be expressed
as
h j˚i D h jj i hj j˚i; (C.13)
which becomes obvious, considering the expression for the identity operator.
C.2 Matrix Representation of Operators 291
C.2 Matrix Representation of Operators
The operators we are dealing with have the property that when acting on a function
in our Hilbert space, they generate another (or the same) function in that space,
OO .x/ D ˚.x/ (C.14)
or with vector notations

O i Dj˚i :
Oj (C.15)
Expanding the vectors on the left-hand side according to the above yields
O i hj j i Dj˚i :
ji i hi jOjj (C.16)
Obviously, we have the identity
OO ji i hi jOjj
O i hj: (C.17)
O i are matrix elements

The numbers hi jOjj
Z
O iD
hi jOjj dx O
i .x/ O j .x/ (C.18)
and they form the matrix representation of the operator

0 1
O
h1jOj1i O
h1jOj2i
OO ) @ h2jOj1i
O O
h2jOj2i A :

Standard matrix multiplication rules are used in operations with vector and matrix
representations, for instance,
1 0
0 1
h1j i h1j˚i
0 1
O
h1jOj1i O
h1jOj2i B C B
B h2j i C B h2j˚i C
C
O i Dj˚i ) @ h2jOj1i B C B C
Oj O O
h2jOj2i A B C D B C ;
B C B C
@ A @ A
hN j i hN j˚i
where
O i hj j˚i
hkj˚i D hkjOjj
summed over the index j .
292 C Representations of States and Operators
C.3 Coordinate Representations
C.3.1 Representation of Vectors
The coordinate representation of the ket vectorj i (C.4) is denoted as hxj i, and
this is identical to the corresponding state or (Schrödinger) wave function

hxj i .x/ I h jxi .x/: (C.19)
This can be regarded as a generalization of the expansion for the expansion coef-
ficients (C.1), where the space coordinates correspond to a continuous set of basis
functions.
The basis functions j .x/ have the coordinate representation hxjj i, and the co-
ordinate representation (C.1) becomes
hxj i D aj j .x/ D aj hxjj i : (C.20)
The scalar product between the functions .x/ and ˚.x/ is

Z
h j˚i D dx .x/ ˚.x/; (C.21)
which we can express as

Z
h j˚i D dx h jxihxj˚i : (C.22)
We shall assume that an integration is always understood, when Dirac notations of

the kind above are used, i.e.,
h j˚i D h jxi hxj˚i (C.23)
in analogy with the summation rule for discrete basis sets. This leads to the formal
identity
jxi hxj I; (C.24)
which is consistent with the corresponding relation (C.12) with a numerable
basis set.
C.3.2 Closure Property
From the expansion (C.1), we have

Z
.x/ D dx 0 0
j .x / .x 0 / j .x/: (C.25)
C.3 Coordinate Representations 293
This can be compared with the integration over the Dirac delta function
Z
.x/ D dx 0 ı.x x 0 / .x 0 /; (C.26)
which leads to the relation known as the closure property
0
j .x / j .x/ D ı.x x 0 / (C.27)
(with summation over j ). In Dirac notations, this becomes
hxjj i hj jx 0 i D ı.x x 0 /
or ˛
hxjIjx 0 D ı.x x 0 /; (C.28)
which implies that the delta function is the coordinate representation of the identity
operator (C.12). Note that there is no integration over the space coordinates here.
C.3.3 Representation of Operators
The coordinate representation of an operator is expressed in analogy with that of a

state vector
˛
OO ) hxjOjx
O 0 D O.x;O x 0 /: (C.29)
which is a function of x and x 0 . An operator OO acting on a state vectorj i is repre-

sented by Z
˛
O i ) hxjOjx
hx Oj O 0 hx 0j i D dx 0 O.x; O x 0 / .x 0 /; (C.30)
which is a function of x.
Appendix D
Dirac Equation and the Momentum
Representation
D.1 Dirac Equation
D.1.1 Free Particles
The standard quantum-mechanical operator representation

@
E ! EO D i„ I p ! pO D i„r I x ! xO D x; (D.1)
@t
where E; p; x represent the energy, momentum, and coordinate vectors, and
O p;
E; O xO the corresponding quantum-mechanical operators, was previously used
to obtain the nonrelativistic Schrödinger equation (2.9). If we apply the same
procedure to the relativistic energy relation
E 2 D c 2 p2 C m2e c 4 ; (D.2)
where c is the velocity of light in vacuum and me the mass of the electron, this
would lead to
@2 .x/
2 2
„2 D c pO C m 2 4
e c .x/; (D.3)
@t 2
which is the Schrödinger relativistic wave equation. It is also known as the
Klein–Gordon equation. In covariant notations (see Appendix, Sect. A.1), it can
be expressed as
„2 C m2e c 2 .x/ D 0: (D.4)
In contrast to the nonrelativistic Schrödinger equation (2.9), the Klein–Gordon
equation is nonlinear, and therefore the superposition principle of the solutions can-
not be applied. In order to obtain a linear equation that is consistent with the energy
relation (D.2) and the quantum-mechanical substitutions (D.1), Dirac proposed the
form for a free electron
@ .x/
i„ D c˛ pO C ˇme c 2 .x/; (D.5)
@t
295
296 D Dirac Equation and the Momentum Representation
where ˛ and ˇ are constants (but not necessarily pure numbers). This equation is
the famous Dirac equation for a relativistic particle in free space.
The equivalence with the equation (D.3) requires
.c˛ pO C ˇme c 2 /.c˛ pO C ˇme c 2 / c 2 pO 2 C m2e c 4 ;
which leads to
˛x2 D ˛y2 D ˛z2 D ˇ 2 D 1;

˛x ˛y C ˛y ˛x D 0 .cyclic/;
˛ˇ C ˇ˛ D 0; (D.6)
where “cyclic” implies that the relation holds for x ! y ! z ! x.

The solution proposed by Dirac is the so-called Dirac matrices

0 1 0
˛D I ˇD ; (D.7)
0 0 1
where D .x ; y ; z / are the Pauli spin matrices

01 0 i 1 0
x D I y D I z D : (D.8)
10 i 0 0 1
The Dirac matrices anticommute
˛ˇ C ˇ˛ D 0: (D.9)
With the covariant four-dimensional momentum vector (A.2) p D .p0 ; p/,

and the corresponding vector operator

i„ @
pO D .pO0 ; p/
O D ; i„r ; (D.10)
c @t
the Dirac equation (D.5) becomes
c pO0 c˛ pO ˇme c 2 .x/ D 0 (D.11)
or with
˛ D .1; ˛/ (D.12)
and ˛ pO D pO0 ˛ pO we obtain the covariant form of the Dirac Hamiltonian for

a free particle
HD D c˛ pO C ˇme c 2 (D.13)

D.1 Dirac Equation 297
and the corresponding Dirac equation
c˛ pO ˇme c 2 .x/ D 0: (D.14)
With the Dirac gamma matrices
D ˇ˛ ; (D.15)
this can also be expressed as (ˇ 2 D 1)

pO me c O me c
.x/ D 6 p .x/ D 0; (D.16)
O is the “p-slash” operator

where 6 p
O D pO D ˇ˛ pO D ˇ .pO0 ˛ p/

6p O : (D.17)
It should be observed that in the covariant notation pO0 is normally disconnected

from the energy, i.e., q
pO0 ¤ pO 2 C m2e c 2 : (D.18)
This is known as off the mass shell. When the equality above holds, it is referred to
as on the mass shell, which can also be expressed as
pO 2 D pO02 pO 2 D m2e c 2 O D me c:
or 6 p (D.19)
In separating the wave function into space and time parts,

i"p t =„
.x/ D p .x/ e ; (D.20)
the time-independent part of the Dirac equation (D.5) becomes
hO free
D .p/O p .x/ D "p p .x/; (D.21)
where
hO free O C ˇme c 2
D D c˛ p (D.22)
is the free-electron Dirac Hamiltonian. The Dirac equation can also be expressed
as "
p
ˇ ˇ˛ pO me c p .x/ D 0: (D.23)
c
Here, p .x/ is a four-component wave function, which can be represented by
1
p .x/ D p ur .p/ eip x I pO p .x/ Dp p .x/ : (D.24)
V
(Note the difference between the momentum vector p and the momentum operator
O eip x represents a plane wave, and ur .p/ is a four-component vector function of
p).
the momentum p. For each p, there are four independent solutions (r D 1; 2; 3; 4).
The parameter p in our notations p and "p represents p and r or, more explicitly,
p .x/ D p;r .x/ I "p D "p;r :
With the wave function (12.1), the Dirac equation (D.23) leads to the following
equation for the ur .p/ functions
"
p
ˇ ˇ˛ pO me c ur .p/ D 0
c
or
"p =c me c pO
ur .p/ D 0 ; (D.25)
pO "p =c me c
where each element is a 2 2 matrix. This equation has two solutions for each
momentum vector p:

"p =c C me c p
uC .p/ D NC I u .p/ D N (D.26)
p "p =c C me c
corresponding to positive (r D 1; 2) and negative (r D 3; 4) eigenvalues, respec-

tively. Defining the momentum component p0 – to be distinguished from the
corresponding operator component pO0 (D.10) – by
q
j"p j D Ep D cp0 I p0 D p2 C m2e c 2 (D.27)
gives

p0 C m e c p
uC .p/ D NC I u .p/ D N : (D.28)
p p0 C m e c
The corresponding eigenfunctions (12.1) are

1 1
pC .x/ D p uC .p/ eip x eiEp t =„ D p u .p/ eip x eiEp t =„
p .x/
V V
(D.29)
including the time dependence according to (D.20) and (D.27).
The vectors

p
u.p/ D uC .p/ and v.p/ D u .p/ D N (D.30)
p0 C m e c
D.1 Dirac Equation 299
satisfy the equations
.6 p me c/ u.p/ D 0 and .6 p C me c/ v.p/ D 0; (D.31)
where p0 is given by (D.27). Note that the negative energy solution corresponds
here to the momentum p for the electron (or Cp for the hole/positron).
Normalization
Several different schemes for the normalization of the u matrices have been used
(see, for instance, Mandl and Shaw [1, Chap. 4]). Here, we shall use
ur0 .p/ ur .p/ D ır 0 ;r ; (D.32)
which leads to

p0 C m e c
uC .p/ D jNC j .p0 C me c; p/
2
p
D jNC j2 .p0 C me c/2 C . p/2 D jNC j2 2p0 .p0 C me c/ (D.33)
using . p/2 D p2 D p02 m2e c 2 . This gives
1
NC D p (D.34)
2p0 .p0 C me c/
and the same for N .

With the normalization above, we have

2 p0 C m e c
uC .p/ uC .p/ D jNC j .p0 C me c; p/
p

1 p0 C m e c p p0 C ˛ p C ˇme c
D D (D.35)
2p0 p p0 m e c 2p0
and similarly
p0 .˛ p C ˇme c/
u .p/ u .p/ D ; (D.36)
2p0
which gives
uC .p/ uC .p/ C u .p/ u .p/ D I: (D.37)

D.1.2 Dirac Equation in an Electromagnetic Field
Classically, the interaction of an electron with electromagnetic fields is given by the

“minimal substitution” (E.15), which in covariant notations can be expressed as1
p ! pO C eA (D.38)
with the four-dimensional potential being

.x/
A .x/ D ; A.x/ : (D.39)
c
This implies that the Dirac Hamiltonian (D.13) becomes
HD D c˛ .pO C eA / C ˇme c 2 (D.40)
and that the interaction with the fields is given by the term
Hint D ec˛ A : (D.41)
D.2 Momentum Representation
D.2.1 Representation of States
Above in Sect. C.3 we have considered the coordinate representation of a state vec-
tor, a .x/ D hxjai. An alternative is the momentum representation, where the state
vector is expanded in momentum eigenfunctions. A statejai is then represented by
a .pr/ D hprjai, which are the expansion coefficients of the state in momentum
eigenfunctions
hxjai D hxjpri hprjai (D.42)
with summations over p and r. The expansion coefficients become
Z r Z
1
hprjai D d x hprjxi hxjai D
3
d3 x eipx ur .p/ a .x/: (D.43)
V
In the limit of continuous momenta, the sum over p is replaced by an integral and
V replaced by .2/3 .
1
In many text books, the minimal substitution is expressed as p ! pO C ec A , because a mixed
unit system, like the cgs system, is used. In the SI system – or any other consistent unit system – the
substitution has the form given in the text. The correctness of this expression can be checked by
means of dimensional analysis (see Appendix K).
D.2 Momentum Representation 301
Note that the momentum representation is distinct from the Fourier transform.
The latter is defined as
r Z
1
hpjai D ur .p/ hprjai D d3 x eipx a .x/
V
Z
! .2/3=2 d3 x eipx a .x/ (D.44)
using the identity (D.37).

In analogy with (C.23), we have
hajbi D hajp; ri hp; rjbi; (D.45)
which yields
jp; ri hp; rj I (D.46)
with implicit summation/integration over p and summation over r.
D.2.2 Representation of Operators
O O.x 2 ; x 1 / D hx 2 jOjx
Coordinate representation of an operator O: O 1 i.
Momentum representation of an operator O: O 1 r1 i.
O O.p2 r2 ; p1 r1 / D hp2 r2 jOjp
Transformation between the representations
ZZ
O 1 r1 i D
hp2 r2 jOjp O 1 i hx 1 jp1 r1 i:
d3 x 2 d3 x 1 hp2 r2 jx 2 i hx 2 jOjx (D.47)
The corresponding Fourier transform is according to (D.44)
O 1 r1 i ur .p1 /:
ur2 .p2 / hp2 r2 jOjp (D.48)
1
Any operator with a complete set of eigenstates can be expanded as
OO Djj i "j hj j where Ojj

O i D "j jj i : (D.49)
This gives the coordinate and momentum representations
O 1 i D hx 2 jj i "j hj jx 1 i;
hx 2 jOjx (D.50a)
O 1 r1 i D hp2 ; r2 jj i "j hj jp1 r1 i:
hp2 r2 jOjp (D.50b)
D.2.3 Closure Property for Momentum Functions
In three dimensions, we have the closure property (C.27)

0
j .x/ j .x / D ı 3 .x x 0 / (D.51)
and for a continuous set of momentum eigenfunctions this becomes

Z

d3 p pr .x/ pr .x 0 / D ı 3 .x x 0 / (D.52)
with summation over r. This can also be expressed as
hxjpri hprjxi D ı 3 .x x 0 / (D.53)
also with integration over p. From the closure property (D.51), we have
0 0
j .p; r/ j .p ; r / D ır;r 0 ı 3 .p p0 /; (D.54)
which leads to
hp; rjj i hj jp0 ; r 0 i D ır;r 0 ı 3 .p p0 /: (D.55)
D.3 Relations for the Alpha and Gamma Matrices
From the definition of the alpha matrices and the definitions in Appendix A, we find
the following useful relations:
˛ ˛ D 1 ˛2 D 2;
˛ ˛˛ D ˛˛ ˛ D 2˛;
˛ ˇ˛ D ˇ ˛ˇ˛ D 4ˇ;
˛ ˇ˛ D ˇ C ˛ˇ˛ D 2ˇ;
˛ A
6 ˛ D ˛ ˇ˛ A ˛ D 4 A
6 ; (D.56)
where 6 A is defined in (D.17). The gamma matrices satisfy the following anticom-
mutation rule:
C D 2g ;
6 B
A 6 CB
6 A
6 D 2AB: (D.57)
This leads to
D 2 ;
A
6 D 2 A
6 ;
Reference 303
D 4;
D 2 ;
A
6 D 2 A
6 ;
0 D D 1;
0 0 0
0 D 0e
;
0e
6 D A;
A 0
0 0 D e
;
0 A6 0 D e
A;
0 0 D e
e ;
0
A 0 e e
6 6 B D AB;
0 ˇ 0 D e
ˇe
e
;
0
A 6 6B 6C D e
0
ABeC
e; (D.58)
where eA D 0 A0 i Ai D 0 A0 C A
With the number of dimensions being equal to 4 , to be used in dimensional
regularization (see Chap. 12), the relations become
D 4 ;
D .2 / ;
A 6 D .2 / A
6 ;
D 4g ;
A 6 B
6 D 4AB A
6 B
6 ;
ˇ D 2 ˇ C ˇ ;
i i D 3 ;
i i D .2 / e ;
i A6 D .2 / A
i e
6 A;
i i D 4g e
e ;
6 B
i A i e e
6 D 4AB AB A6 B
6 ;
i ˇ i D 2 ˇ e
ˇe
e
C ˇ ;
i A6 6 B 6 C D 2 C
i
6 B 6 e
6 A AB eC
eCA6 B6 C 6 : (D.59)
Reference
Appendix E
Lagrangian Field Theory
Concerning notations, see Appendix A.
E.1 Classical Mechanics
In classical mechanics, the Lagrangian function for a system is defined as
L D T V; (E.1)
where T is the kinetic energy and V the potential energy of the system. Generally,
this depends on the coordinates qi , the corresponding velocities qPi D @q@t
i
, and
possible explicitly on time (see, for instance [2, Sect. 23])
L.tI q1 ; q2 I qP1 ; qP2 /: (E.2)
The action is defined as

Z
I D dt L.tI q1 ; q2 I qP1 ; qP2 /: (E.3)
The principle of least action implies that
ıI.q1 ; q2 I qP1 ; qP2 / D 0; (E.4)
which leads to the Lagrange equations

d @L @L
D 0: (E.5)
dt @qPi @qi
The Hamilton function can be defined as
H D pi qPi L; (E.6)
305
306 E Lagrangian Field Theory
where pi is the canonically conjugate momentum to the coordinate qi
@L
pi D : (E.7)
@qPi
It then follows that

@H @qi
D qPi D : (E.8a)
@pi @t
Furthermore, from the definitions above and the Lagrange equations, we have
@H
D pPi : (E.8b)
@qi
These are Hamilton’s canonical equations of motion.

We consider a general function of time and the coordinates and canonical mo-
menta f .tI pi ; qi /. Then the total derivative with respect to time becomes
df @f @f @qi @f @pi @f @f @H @f @H
D C C D C : (E.9)
dt @t @qi @t @pi @t @t @qi @pi @pi @qi
With the Poisson bracket of two functions A and B, defined by
@A @B @B @A
fA; Bg D ; (E.10)
@qi @pi @qi @pi
the derivative can be expressed as
df @f
D C ff; H g: (E.11)
dt @t
For a single-particle system in one dimension (x), the kinetic energy is T D
p 2 =2m, where m is the mass of the particle, which yields
p2 mv2
LD V D V;
2m 2
where v D xP is the velocity of the particle. Furthermore, pi qP i D p xP D p 2 =m D

mv2 , yielding with (E.6)
p2 mv2
H D CV D C V;
2m 2
E.1 Classical Mechanics 307
which is the classical energy expression. The canonically conjugate momentum

(E.7) is then
@L @L
pD D D mv;
@xP @v
which is the classical momentum.
E.1.1 Electron in External Field
The Lagrangian for an electron (charge e) in an external field, A D . .x/=c; A/,
is [1, p. 25]
1
P D mxP 2 eA xP C e .x/;
L.x; x/ (E.12)
2
where the last two terms represent (the negative of) a velocity-dependent potential.
The conjugate momentum corresponding to the variable x is then according to (E.7)
pi ! p D mxP eA: (E.13)
Using the relation (E.6), we get the corresponding Hamilton function
1 1
H D p xP L D mxP 2 e .x/ D .p C eA/2 e .x/: (E.14)
2 2m
We see that the interaction with the fields ( ; A) is obtained by means of the substi-
tutions
H ! H e .x/ p ! p C eA (E.15)
known as the minimal substitutions.

The corresponding equations of motion can be obtained from either the La-
grange’s or Hamilton’s equation of motion. We then have

d @L d
! .mxP eA/
dt @qPi dt
and
@L
! er .A x/
P C er .x/:
@qi
The same equations are obtained from the Hamilton’s equations of motion (E.8b).
The total time derivative can in analogy with (E.9) be expressed as
d @ dx @ @
D C C D C xP r
dt @t dt @x @t
giving
d @A
.mxP eA/ D mxR e e.xP r /A:
dt @t
From the identity
xP .r A/ D r .A x/
P .xP r /A;
we then obtain the equations of the motion
@A
mxR D er .x/ C e e xP .r A/ D e .E C v B/ (E.16)
@t
with v D xP being the velocity of the electron. This is the classical equations of
motion for an electron of charge e in an electromagnetic field. The right-hand side
is the so-called Lorentz force on an electron in a combined electric and magnetic
field. This verifies the Lagrangian (E.12).
E.2 Classical Field Theory
In classical field theory, we consider a Lagrangian density of the type
L D L. r ; @ r /; (E.17)
where r D r .x/ represents different fields and
@ r
@ r D : (E.18)
@x
The requirement that the action integral
Z
I D d4 x L. r ; @ r / (E.19)
be stationary over a certain volume leads to the Euler–Lagrange equations
@L @L
@ D 0: (E.20)
@ r @.@ r /
The field conjugate to r .x/ is
@L
r .x/ D ; (E.21)
@ Pr
E.3 Dirac Equation in Lagrangian Formalism 309
where the “dot” represents the time derivative. The Lagrangian function is defined
as Z
L.t/ D d3 x L. r ; @ r /: (E.22)
The Hamiltonian density is defined as
H.x/ D r .x/ Pr .x/ L. r ; @ r /: (E.23)
In quantized Lagrangian field theory, the fields are replaced by operators, satis-
fying the Heisenberg commutation rules at equal times [1, Eq. 2.31]
h i
O .x/; O .x 0 / D i„ ı; ı 3 .x x 0 / (E.24)
with the remaining commutations vanishing. In our applications, the quantized field
will normally be the electron field in the interaction picture (B.28) or the electro-
magnetic field (G.2).
E.3 Dirac Equation in Lagrangian Formalism
From the Dirac equation for a free electron (D.14), we can deduce the corresponding
Lagrangian density
L.x/ D O .x/ i„c ˛ @ ˇme c 2 O .x/: (E.25)
Using the relation (B.17), the space integral over this density yields the correspond-
ing operator
Z
L D d3 x L.x/ D i„c ˛ @ ˇme c 2 D c ˛ p ˇme c 2 (E.26)
(with pO D i„ @ ) and the corresponding Hamilton operator (E.6)
H D L D c ˛ p C ˇme c 2 ; (E.27)
since the fields are time independent. This leads to the Dirac equation for a free
electron (D.14).
We can also apply the Euler–Lagrange equations (E.20) on the Lagrangian
(E.25), which leads to
@L
@ D @ O .x/ i„ c ˛ ;
@.@ O /
@L
D O .x/ ˇme c 2 ;
@O
and
@ i„ c ˛ O .x/ C ˇme c 2 O .x/ D 0
with the hermitian adjoint
i„ c ˛ @ C ˇme c 2 O .x/ D 0; (E.28)
which is consistent with the Dirac equation for the free electron.
In the presence of an electromagnetic field, we make the minimal substitution
(D.38)
ie
p ! p C eA .x/ or @ ! @ A .x/; (E.29)
„
which leads to the Lagrangian density in the presence of an electromagnetic field
L.x/ D O .x/ c ˛ p ˇme c 2 C ec˛ A .x/ O .x/: (E.30)
This gives the the corresponding Hamiltonian density
H.x/ D O .x/ c ˛ p C ˇme c 2 ec˛ A .x/ O .x/; (E.31)
where the last term represents the interaction density
Hint .x/ D O .x/ ec˛ A .x/ O .x/: (E.32)
The corresponding Hamilton operator can then be expressed as

Z
HO D d3 x 1 O .x1 / c ˛ p C ˇme c 2 ec˛ A .x/ O .x1 /: (E.33)
References
2. Schiff, L.I.: Quantum Mechanics. McGraw-Hill Book Co, N.Y. (1955)
Appendix F
Semiclassical Theory of Radiation
F.1 Classical Electrodynamics
F.1.1 Maxwell’s Equations in Covariant Form
The Maxwell equations in vector form are1
r E D =0 ; (F.1a)
1 @E
r B D C 0 j ; (F.1b)
c 2 @t
r B D 0; (F.1c)
@B
r E C D 0; (F.1d)
@t
where is the electric charge density and j the electric current density. Equation
(F.1c) gives
B D r A; (F.2)
where A is the vector potential. From (F.1d), it follows that the electric field is of
the form
@A
E D r ; (F.3)
@t
where is the scalar potential. The (F.1a) and (F.1b) give together with (F.3) and
(F.2)
@
r 2 r A D =0 D c0 j 0
@t

1 @2 A 1 @
r A 2 2
2
r r AC 2 D 0 j (F.4)
c @ t c @t
1
As in the previous Appendices, the formulas are here given in a complete form and valid in any
consistent unit system, like the SI system (see Appendix K).
311
312 F Semiclassical Theory of Radiation
using the vector identity r .r A/ D r .r A/ r 2 A. Here, j 0 D c (with

0 0 D c 2 ) is the scalar or “time-like” part of the four-dimensional current den-
sity
j D j D .c ; j /; (F.5)
where the vector part is the three-dimensional current density j . Similarly, the four-
dimensional vector potential
A D . =c; A/ A D . =c; A/ (F.6)
has the scalar part =c and the vector part A. With the d’Alambertian operator
(A.10), these equations can be expressed as2
@
.rA/ D c0 j 0 ; (F.7)
@t
A C r .rA/ D 0 j ; (F.8)
which leads to Maxwell’s equations in covariant form
A r.rA/ D 0 j (F.9)
or
@ @ A @ .@ A / D 0 j : (F.10)
F.1.1.1 Electromagnetic-Field Lagrangian
We introduce the field tensor [2, Eq. 5.5]
F D @ A @ A : (F.11)
Then we find for instance

@ =c @Ax
F 01 D @1 A0 @0 A1 D D Ex ;
@x @ct
@Ax @Ay
F 12 D @2 A1 @1 A2 D D Bz ;
@y @x
etc., leading to the matrix
0 1
0 Ex =c Ey =c Ey =c
B Ex =c 0 Bx Bz C
F DB
@ Ey =c Bz 0
C: (F.12)
Bx A
Ez Bx Bx 0
2
Concerning covariant notations, see Appendix A.1.
F.1 Classical Electrodynamics 313
The Maxwell equations (F.10) can now be expressed as [2, Eq. 5.2]
@ F D 0 j (F.13)
using the identity

@ @ A @ @ A :
With r D A , the Euler–Lagrange equations (E.20) become
@L @L
@ D 0: (F.14)
@A @.@ A /
Using the field tensor (F.11) and the form of the metric tensor (A.5), we have
F F D @ A @ A .@ A @ A /
D @ A @ A .g @ g A g @ g A / : (F.15)
Here, and are running indices that are summed over, and we can replace them
with 0 and 0 , respectively. The derivative with respect to fixed and then gives
@ 0 0 0 0 0 0
F0 0 F D F F C F0 0 g g F 0 0 g g D 4F :
@.@ A /
(F.16)
We then find that with the Lagrangian
1
LD F F j A ; (F.17)
40
the Euler–Lagrange equations (F.14) lead to the Maxwell equations (F.13).

With the same Lagrangian, the conjugate fields (E.21) are
@L 1 @L
.x/ D D ; (F.18)
@AP c @.@0 A /
where the dot represents the time derivative and @0 D @

@x0 D 1 @
c @t D @0 . Using the
relation (F.16), this yields
1
.x/ D F 0 .x/: (F.19)
0 c
The Hamiltonian is given in terms of the Lagrangian by [2, 5.31]
Z
H D d3 x sN f .x/AP .x/ Lg; (F.20)
where N fg represents normal order [1, Chap. 11] (see Sect. 2.2).
F.1.1.2 Lorenz Condition
The Lorenz condition is3
1 @
rA D @ A D r A C D0 (F.21)
c 2 @t
and with this condition the Maxwell equations get the simple form
A D 0 j: (F.22)
Then also the electromagnetic fields have particularly simple form, given in (G.2).
F.1.1.3 Continuity Equation
Operating on Maxwell’s equations (F.9) with r yields:
r .A/ r r.rA/ D 0 rj:
Since D r 2 and r commute, this leads to the continuity equation
rj D @ j D 0: (F.23)
F.1.1.4 Gauge Invariance
A general gauge transformation is represented by
A)ACr ; (F.24)
where is an arbitrary scalar.
Inserted into the Maxwell equations (F.9), this yields:
.r / r.rr / D .r / r. / D 0;
which shows that the Maxwell equations are gauge invariant.
3
This condition is named after the Danish physicist Ludvig Lorenz, not to be confused with the
more well-known Dutch physicist Hendrik Lorentz.
F.1 Classical Electrodynamics 315
F.1.2 Coulomb Gauge
F.1.2.1 Transverse and Longitudinal Field Components
The vector part of the electromagnetic field can be separated into transverse
(divergence-free) and longitudinal (rotation-free) components
A D A? C Ak I r A? D 0 I r A k D 0: (F.25)
The electric field can be similarly separated

@A ? @A k
E D E? C Ek I E? D I Ek D r ;
@t @t
while the magnetic field has only transverse components due to the relation (F.2).
The separated field equations (F.4) then become
@
r2 C r A k D =0 ; (F.26a)
@t

2 1 @2 A k 1 @
r Ak 2 2 r r Ak C 2 D 0 j k ; (F.26b)
c @ t c @t

1 @2
r 2 2 2 A ? D 0 j T : (F.26c)
c @ t
The longitudinal and the scalar or “time-like” components (A k ; ) represent the
instantaneous Coulomb interaction and the transverse components (A ? ) represent
retardation of this interaction and all magnetic interactions, as well as the electro-
magnetic radiation field (see Sect. F.2.4).
The energy of the electromagnetic field is given by
Z
1 1 ˇˇ ˇˇ2 ˇ ˇ2
Erad D d3 x B C 0 ˇEˇ
2 0
Z Z
1 1 ˇˇ ˇˇ2 ˇ ˇ2 1 ˇ ˇ2
D d3 x B C 0 ˇE ? ˇ C d3 x 0 ˇE k ˇ : (F.27)
2 0 2
The last term represents the energy of the instantaneous Coulomb field, which is
normally already included in the hamiltonian of the system. The first term represents
the radiation energy.
Semiclassically, only the transverse part of the field is quantized, while the lon-
gitudinal part is treated classically [3, Chaps. 2 and 3]. It should be noted that the
separation into transverse and longitudinal components is not Lorentz covariant and
therefore, strictly speaking, not physically justified, when relativity is taken into ac-
count. It can be argued, though, that the separation (as made in the Coulomb gauge)
should ultimately lead to the same result as a covariant gauge, when treated properly.
In a fully covariant treatment also the longitudinal component is quantized. The

field is then represented by virtual photons with four directions of polarizations. A
real photon can only have transverse polarizations.
The Coulomb gauge is defined by the condition
r A.x/ D 0: (F.28)
Using the Fourier transform

Z
A.x/ D d4 k A.k/ eikx ; (F.29)
this condition leads to

Z
@Ai
D d4 k Ai .k/.i/ki eikx D 0
@x i
or
A.k/ k D 0: (F.30)
This is also known as the transversally condition and implies that there is no longi-
tudinal component of A. Maxwell’s equations then reduce to
r 2 D =0 : (F.31)
This has the solution

Z
1 .x 0 /
.x/ D d3 x 0 ; (F.32)
40 jx x 0 j
which is the instantaneous Coulomb interaction.

In free space, the scalar potential can be eliminated by a gauge transformation.
Then the Lorenz condition (F.21) is automatically fulfilled in the Coulomb gauge.
The field equation (F.4) then becomes
1 @2 A
r 2A D 0: (F.33)
c 2 @2 t
The relativistic interaction with an atomic electron (D.41) is then in the Coulomb
gauge given by
Hint D ec ˛ A? (F.34)
and in second quantization (see Appendix B)
X
HO int D

cj hi jec ˛ A?jj i cj ; (F.35)
ij
F.2 Quantized Radiation Field 317
where c ; c represent creation/annihilation operators for electrons. In the interaction

picture, this becomes
X
HO int;I .t/ D ci hi jec ˛ A ?jj i cj ei."i "j /t =„ :

(F.36)
ij
F.2 Quantized Radiation Field
F.2.1 Transverse Radiation Field
Classically the transverse components of the radiation field can be represented by

the vector potential [3, Eq. 2.14]
2 h
XX i

A.x; t/ D ckp "p ei.kx!t / C ckp "p ei.kx!t / ; (F.37)
k pD1

where k is the wave vector, ! D cjkj the frequency, and ckp =ckp represents the
amplitude of the wave with a certain k vector and a certain polarization "p . The
energy of this radiation can be shown to be equal to [3, p.22]
X X

Erad D 20 ! 2 ckp ckp D 0 ! 2 ckp ckp C ckr ckp : (F.38)
kp kp
By making the substitution

s s
„ „
ckp ! akp and ckp ! a ;
20 !V 20 !V kp

where akp ; =akp are photon creation/annihilation operators, the radiation energy
goes over into the hamiltonian of a collection of harmonic oscillators
1X
Hharm:osc D „! .akp akp C akp akp /:
2 kp
Therefore, we can motivate that the quantized transverse radiation field can be rep-
resented by the operator [3, Eq. 2.60]
s
X X2 h i
„
A ? .x; t/D
akp "p ei.kx!t / C akp "p ei.kx!t / : (F.39)
k
20 !V pD1
F.2.2 Breit Interaction
The exchange of a single transverse photon between two electrons is illustrated by

the time-ordered diagram (left) in Fig. F.1, where one photon is emitted at the time
t1 and absorbed at a later time t2 . The second-order evolution operator for this
process, using the interaction picture, is given by (see Sect. 3.1)
2 Z Z
i 0 0
U.2/ .0; 1/ D dt2 HO int;I .t2 / dt1 HO int;I .t1 / e.t1 Ct2 / ; (F.40)
„ 1 1
where is the parameter for the adiabatic damping of the perturbation. The in-
teraction Hamiltonians are in the Coulomb gauge given by (F.36) with the vector
potential (F.39)
s
X „ X
2
˝ ˇ ˇ ˛
HO int;I .t1 / D cr r ˇ akp ec ˛ "kp eikx ˇa
k1
20 !1 V p1 D1
1
ca eit1 ."a "p „!1 /=„ ;

s
X „ X2
˝ ˇ ˇ ˛
HO int;I .t2 / D cs r ˇ akp ec ˛ "kp eikx 2ˇa
k
20 !2 V p D1
2 2
it2 ."b "s C„!2 /=„
cb e ; (F.41)
which leads to the evolution operator

X
e2c2
U.2/ .0; 1/ D cr ca cs cb p
2„ 0 V !1 !2
k
X ˝ ˇ ˇ ˛
rs ˇ akp ˛ "p eikx 2 akp ˛ "p eikx ˇab I;

1
p1 p2
(F.42)
where I is the time integral. The contraction between the creation and annihilation
operators (G.10) yields . ! D !1 D !2 /
s
2 r s
r B12
6 2 1
b
1 a b
a
Fig. F.1 Diagrammatic representation of the exchange of a single, transverse photon between two
electrons (left). This is equivalent to a potential (Breit) interaction (right)
X ˝ ˇ ˇ ˛
rs ˇ akp ˛ "p eikx 2

akp ˛ "p eikx ˇab
1
p1 p2
X
2
˝ ˇ
D rs ˇ.˛ "p /2 .˛ "p /1 eikr 12 .r 12 D x 1 x 2 /: (F.43)
pD1
The time integral in (F.42) is

Z 0 Z t2
I D dt2 eit2 ."b "s C„!Ci /=„ dt1 eit1 ."a "r „!Ci /=„
1 1
1
D (F.44)
.cq C cq 0 C 2i /.cq ! C i /
with cq D ."a "r /=„ and cq 0 D ."b "s /=„.

The result of the opposite time ordering t1 > t2 is obtained by the exchange
1 $ 2 .r 12 $ r 12 /, a $ b, and r $ s, and the total evolution operator,
including both time-orderings, can be expressed as
e 2 c 2 X X ˝ ˇˇ ˇ ˛
2
U.2/ .0; 1/ D cr ca cs cb rs .˛ "p /1 .˛ "p /2 Mˇab
2„0 !V k pD1
(F.45)
with
eikr 12 eikr 12
MD C :
.cq C cq 0 C 2i /.cq ! C i / .cq C cq C 2i /.cq 0 ! C i /
0
(F.46)
This can be compared with the evolution operator corresponding to a potential in-
teraction B12 between the electrons, as illustrated in the right diagram of Fig. F.1,
Z
i 0
U.2/ .0; 1/ D cr ca cs cb hrsjB12jabi dt eit ."a C"b "r "s Ci/=„
„ 1

cr ca cs cb hrsjB12jabi
D : (F.47)
„ cq C cq 0 C i
Identification then leads to

e2c2 X eikr 12 eikr 12
B12 D .˛ "p /1 .˛ "p /2 C 0 : (F.48)
20 !V kp cq ! C i cq ! C i
We assume now that energy is conserved by the interaction, i.e.,
"a "r D "s "b or q C q 0 D 0: (F.49)

It is found that the sign of the imaginary part of the exponent is immaterial (see
Appendix J.2), and the equivalent interaction then becomes
e2 X ei kr 12
B12 D .˛ "p /1 .˛ "p /2 2 (F.50)
0 V q k 2 C i
kp
with ! D ck.
The "p vectors are orthogonal unit vectors, which leads to [3, Eq. 4.312]
X
3
.˛ "p /1 .˛ "p /2 D ˛1 ˛2 : (F.51)
pD1
This gives
X
2
O
.˛ "p /1 .˛ "p /2 D ˛1 ˛2 .˛1 k/.˛ O
2 k/ (F.52)
pD1
assuming "3 D kO to be the unit vector in the k direction. The interaction (F.50) then
becomes in the limit of continuous momenta (Appendix D)
Z
e2 d3 k h O O
i ei kr 12
B12 D ˛1 ˛2 .˛ 1 k/.˛2 k/ : (F.53)
0 .2/3 q 2 k 2 C i
With the Fourier transforms in Appendix J, this yields the retarded Breit interac-
tion
" #
e2 eijqjr12 eijqjr12 1
Ret
B12 D ˛1 ˛2 .˛1 r 1 /.˛2 r 2 / : (F.54)
40 r12 q 2 r12
Setting q D 0, we obtain the instantaneous Breit interaction (real part)

e2 ˛1 ˛2 1
Inst
B12 D C .˛1 r 1 /.˛2 r 2 / r12
40 r12 2
or using
˛1 ˛2 .˛1 r 12 /.˛1 r 12 /
.˛1 r 1 /.˛2 r 2 / r12 D C 2
;
r12 r12
we arrive at

e2 1 .˛1 r 12 /.˛1 r 12 /
Inst
B12 D ˛1 ˛2 C ; (F.55)
40 r12 2 2r12
which is the standard form of the instantaneous Breit interaction.

F.2.3 Transverse Photon Propagator
s r
r 2 1 s r s
+ = 1 - 2
6 6
1 b a 2 a b
a b
Fig. F.2 The two time-orderings of a single-photon exchange can be represented by a single
Feynman diagram
We shall now consider both time-orderings of the interaction represented in Fig. F.2
simultaneously. The evolution operator can then be expressed as
2 Z Z h i
i 0 0
U.2/ .0; 1/ D dt2 dt1 T HO int;I .t2 / HO int;I .t1 / e.jt1 jCjt2 j/ ;
„ 1 1
(F.56)
where
8
h i < HO int;I .t2 / HO int;I .t1 / t2 > t1
T HO int;I .t2 / HO int;I .t1 / D : (F.57)
: O
Hint;I .t1 / HO int;I .t2 / t1 > t2
In the Coulomb gauge, the interaction is given by (F.36) and the vector potential is
given by (F.39). The evolution operator for the combined interactions will then be
Z 0 Z 0
e2c2
U.2/ .0; 1/ D cr ca cs cb dt1 T
dt2 Œ.˛ A ? /1 .˛ A? /2
„2 1 1
it1 ."a "r Ci /=„ it2 ."b "s Ci /=„
e e : (F.58)
Here X „
T Œ.˛ A ? /1 .˛ A ? /2 D .˛ "p /1 .˛ "p /2
kp
2!0 V
8 i.k x !t / i.k x !t /
< e 1 1 1 e 2 2 2 t2 > t1

:
ei.k2 x2 !t2 / ei.k1 x1 !t1 / t1 > t2
or with r 12 D x 1 x 2 and t12 D t1 t2
„ X 1 Xe
2 i.kr 12 !t12 /
T Œ.˛ A ? /1 .˛ A ? /2 D .˛ "p /1 .˛ "p /2 ;
0 pD1 V k 2!
(F.59)
where the upper sign is valid for t2 > t1 . This yields

Z 0 Z 0 X
2
e2c2
U.2/ .0; 1/ D cr ca cs cb dt2 dt1 .˛ "p /1 .˛ "p /2
0 „2 1 1 pD1
1 Xe i.kr 12 !t12 /
eicq t12 e.t1 Ct2 / (F.60)
V k 2!
using the energy conservation (F.49).

The boxed part of the equation above is essentially the photon propagator (4.23)
Z
1 Xe i.kr 12 !t12 /
d3 k e i.kr 12 !t12 /
DF .1; 2/ D ) : (F.61)
V 2! .2/3 2!
k
This can be represented by a complex integral

Z Z
d3 k dz eizt12
DF .1; 2/ D i eikr 12 ; (F.62)
.2/ 3 2 z ! 2 C i
2
where is a small, positive quantity. As before, the sign of the exponent ik r 12 is

immaterial. The integrand has poles at z D ˙.! i/, assuming ! to be positive.
1 i! t12 ikr 12
For t2 > t1 , integration over the negative half plane yields 2! e e and for
1 i! t12 ikr 12
t1 > t2 integration over the positive half plane yields 2! e e , which is
identical to (F.61). The evolution operator (F.60) can then be expressed as
Z Z 0
e2c2 0
U.2/ .0; 1/ D cr ca cs cb dt2 dt1
0 „ 1 1
X2
.˛ "p /1 .˛ "p /2 DF .1; 2/ eicq t12 =„ e.t1 Ct2 /„ :
pD1
(F.63)
F.2.4 Comparison with the Covariant Treatment
It is illuminating to compare the quantization of the transverse photons with the

fully covariant treatment, to be discussed in the next chapter. Then we simply have
to replace the sum in (F.42) by the corresponding covariant expression
X
2

X
3

.akp ˛ "p /1 .akp ˛ " p /2 ) .akp ˛ "p /1 .akp ˛ "p /2 :
p1 p2 D1 p1 p2 D0
(F.64)
The commutation relation (G.10) yields
X
3

.akp ˛ "p /1 .akp ˛ "p /2 D ˛1 ˛2 1: (F.65)
p1 p2 D0
We then find that the equivalent potential interaction (F.50) under energy conserva-
tion is replaced by
Z
e2 d3 k ei kr 12
V12 D .1 ˛1 ˛2 / (F.66)
0 .2/3 q 2 k 2 C i
and with the Fourier transform given in Appendix J.2
e2
V12 D .1 ˛1 ˛2 / eijqjr12 : (F.67)
40 r12
We shall now compare this with the exchange of transverse photons, treated
above. We then make the decomposition
8
O
< 1 .˛1 k/.˛ O
2 k/
1 ˛1 ˛2 D : (F.68)
: O O
˛1 ˛2 C .˛1 k/.˛2 k/
The last part, which represents the exchange of transverse photons, is identical to
(F.52), which led to the Breit interaction. The first part, which represents the ex-
change of longitudinal and scalar photons, corresponds to the interaction
Z
e2 d3 k h O O
i ei kr 12
VC D 1 .˛1 k/.˛2 k/ : (F.69)
0 .2/3 q 2 k 2 C i
This Fourier transform is evaluated in Appendix J.3 and yields

Z Z
e2 d3 k q2 ei kr 12 e2 d3 k ei kr 12
VC D 1 2 D (F.70)
0 .2/3 k q k C i
2 2 0 .2/3 k 2 i
provided that the orbitals are generated in a local potential. Using the transform in
Appendix J.2, this becomes
e2
VCoul D : (F.71)
40 r12
Thus, we see that the exchange of longitudinal and scalar photons corresponds to
the instantaneous Coulomb interaction, while the exchange of the transverse pho-
tons corresponds to the Breit interaction. Note that this is true only if the orbitals
are generated in a local potential.
If instead of the separation (F.68), we would separate the photons into the scalar
part .p D 0/ and the vector part .p D 1; 2; 3/,
8
<1
1 ˛1 ˛2 D ; (F.72)
:
˛1 ˛2
then the result would be
e2
Ret
VCoul D eijqjr12 ;
4ı0 r12
e2
Ret
VGaunt D ˛1 ˛2 eijqjr12 ; (F.73)
40 r12
which represents the retarded Coulomb and the retarded magnetic (Gaunt) inter-
action. This implies that the longitudinal photon represents the retardation of the
Coulomb interaction, which is included in the Breit interaction (F.54).
If we would set q D 0, then we would from (F.73) retrieve the instantaneous
Coulomb interaction (F.71) and
e2
˛1 ˛2 ; (F.74)
40
which is known as the Gaunt interaction. The Breit interaction will then turn into the
instantaneous interaction (F.55). This will still have some effect of the retardation
of the Coulomb interaction, although it is instantaneous.
We shall see later that the interactions (F.73) correspond to the interactions in the
Feynman gauge (4.56), while the instantaneous Coulomb and Breit incinerations
correspond to the Coulomb gauge.
References
3. Sakurai, J.J.: Advanced Quantum Mechanics. Addison-Wesley Publ. Co., Reading, Mass.
(1967)
Appendix G
Covariant Theory of Quantum ElectroDynamics
G.1 Covariant Quantization: Gupta–Bleuler Formalism
With the Lorenz condition (F.21) @ A D 0, the Maxwell equations have a

particularly simple form (F.22)
A D 0 j: (G.1)
In this case, the covariant electromagnetic radiation field can be expressed in anal-
ogy with the semiclassical expression (F.39) and represented by the four-component
vector potential [4, Eq. 5.16]
s
„ X h i
C
A .x/ D A .x/ C A .x/ D "r akr ei kx C akr
ei kx :
2!0 V
kr
(G.2)
However, different equivalent choices can be made, as further discussed in Sect. G.2.
Here, we use the covariant notations
k D k D .k0 ; k/ I k0 D !=c D jkj I kx D !t k x
defined in Appendix A.1, and r D .0; 1; 2; 3/ represents the four polarization states.
Normally, the polarization vector for r D 3 is defined to be along the k vector –
longitudinal component – and for r D 1; 2 to be perpendicular – transverse compo-
nents. The component r D 0 is referred to as the time-like or scalar component (see
Sect. F.1.2).
The electromagnetic fields components are Heisenberg operators and should sat-
isfy the canonical commutation (quantization) rules (E.24) at equal times
A .t; x/; .t; x 0 / D i„ı; ı 3 .x x 0 /; (G.3)
where .x/ is the conjugate field (E.21). With the Lagrangian (F.17) the field 0
vanishes according to the relation (F.19), which is inconsistent with the quantization
325
326 G Covariant Theory of Quantum ElectroDynamics
rule (G.3). In order to remedy the situation, we add a term 2

0
.@ A /2 to the
Lagrangian (F.17), where is an arbitrary constant [3, Eq. 1-49, Eq. 3-98]
1
LD F F .@ A /2 j A : (G.4)
40 20
We can rewrite the extra term as

.@ g A / .@ A / :
20
Then the conjugate field (F.19) becomes [3, Eq. 3-100]
@L 1 0
.x/ D 0
D F 0 g @ A (G.5)
@.@ A / c0 0
and 0 ¤ 0 for ¤ 0.
The extra term in the Euler–Lagrange equations (F.14) leads to
@
@ .@ g A / .@ A / D @ g .@ A /
20 @.@ A / 0

D @ g .@ A / D @ .@ A / :
0
The Maxwell equations (F.10) then take the modified form [3, Eq. 3-99]
@ @ A .1 / @ .@ A / D 0 j : (G.6)
Setting D 1, we retrieve the same simple form of Maxwell’s equations as with

the Lorenz condition (G.1) – without introducing this condition explicitly. This is
usually referred to as the Feynman gauge.
The Lagrangian (G.4) is incompatible with the Lorenz condition, and to resolve
the dilemma this condition is replaced by its expectation value
h j@ Aj i D 0; (G.7)
which is known as the Gupta–Bleuler proposal [4, 5.35].

In the Feynman gauge, the commutation relations (G.3) become [4, 5.23]
A .t; x/; AP .t; x 0 / D ic 2 0 „g ı; ı 3 .x x 0 /: (G.8)
To satisfy this relation, we can assume that the polarization vectors fulfill the or-
thogonality/completeness relations [4, Eq. 5.18,19]
"r "r 0 D grr 0 ;

X
grr "r "r D g ; (G.9)
r
G.2 Gauge Transformation 327
and the photon creation and absorption operators the commutation relation [4,
Eq. 5.28]
h i

akr ; ak0 r 0 D ık;k0 grr 0 : (G.10)
Considering that the g-matric (A.5) used is diagonal, this leads to

Xh i X h i
"r akr ; "r 0 ak0 r 0 D "r "r 0 akr ; ak0 r 0 D g; ık;k0 (G.11)
rr 0 rr 0
and then it follows that the field operators (G.2) satisfy the commutation relation
(G.8).
With the Lagrangian (G.4) and the conjugate fields (G.5), the Hamiltonian of the
free field (F.20) becomes in the Feynman gauge ( D 1) [4, Eq. 5.32]
X
HRad D „! grr akr akr : (G.12)
k;r
G.2 Gauge Transformation
G.2.1 General
The previous treatment is valid in the Feynman gauge, where the Maxwell equations
have the form (G.1), and we shall here investigate how the results will appear in
other gauges.
The interaction between an electron and the electromagnetic field is given by the
Hamiltonian interaction density (D.41)
Hint D j A ; (G.13)
where j is the current density. The Maxwell equations are invariant for a gauge
transformation (F.24) A ) A C r , which transforms this interaction to

@
Hint D j A ) A C

j :
@x
Integration over space leads after partial integration to

Z Z
@ @j
3
d x j D d3 x D 0:
@x @x
Since is arbitrary, it follows that
@j
D ı j D rj D 0;
@x
which is the continuity equation (F.23). In analogy with (F.30), the corresponding
relation in the k space is
j .k/ k D 0: (G.14)
The single-photon exchange is represented by the interaction (4.44)
I.x2 ; x1 / D e 2 c 2 ˛1 ˛2 DF .x2 ; x1 /; (G.15)
which corresponds to the interaction density j DF j . In view of the relation

(G.14), it follows that the transformation
DF .k/ ) DF .k/ C k f .k/ C k f .k/;
where f .k/ and f .k/ are arbitrary functions of k, will leave the interaction un-
changed.
G.2.2 Covariant Gauges
In a covariant gauge, the components of the electro-magnetic field are expressed in

a covariant way. We shall consider three gauges of this kind.
G.2.2.1 Feynman Gauge
The photon propagator in the Feynman gauge is given by the expression (4.28)
g 1
DF .k/ D : (G.16)
c0 k 2 C i
G.2.2.2 Landau Gauge
With
1 k
f .k/ D ;
c0 .k 2 C i/2
the propagator (G.16) becomes

1 1 k k
DF .k/ D g 2 : (G.17)
c0 k 2 C i k C i
This leads to k DF D 0, which is consistent with the Lorenz condition (F.21)
rA D @ A D 0:
G.2.2.3 Fried–Yennie Gauge
With
1 k
f .k/ D .1 / 2 ;
2c0 .k C i/2
the propagator (G.16) becomes

1 1 k k
DF .k/ D g .1 / : (G.18)
c0 k 2 C i k 2 C i
With D 1, this yields the Feynman gauge, and with D 0 we retrieve the Lan-
dau gauge. The value D 3 yields the Fried–Yennie gauge [2], which has some
improved properties, compared to the Feynman gauge, in the infrared region.
G.2.3 Noncovariant Gauge
We consider only one example of a noncovariant gauge, the Coulomb gauge, which
is of vital importance in treating the combined QED-correlation problem. Here, the
Coulomb interaction is treated differently from the transverse part.
G.2.3.1 Coulomb Gauge
With
1 1 k0 1 1 ki
f0 D I fi D .i D 1; 2; 3/;
2c0 k 2 C i k2 2c0 k 2 C i k2
the propagator (G.16) can be expressed as
1
!
1 0
DF00 .k/ D k2 ; (G.19)
ki kj
c0 0 1
k 2 Ci
ıi;j k2
where the first row/column corresponds to the component D 0 and the second
row/column to D 1; 2; 3.
This leads to k i DFij D 0, which is consistent with the Coulomb condition (F.30)
r A D @i Ai D 0 .i D 1; 2; 3/:
The formulas above can be generalized to be used in dimensional regularization

(see Sect. 12.4), where the number of dimensions is noninteger (mainly from
Adikins [1], see also t’Hooft and Veltman [6]).
Following the book by Peskin and Schroeder [5], we can by means of Wick rota-
tion evaluate the integral
Z Z
dD l 1 dD l 1
D i.1/m
.2/D .l 2 /m .2/D .lE2
C /m
Z Z 1 D1
d˝D lE
D i.1/m 0
dlE :
4
.2/ 0 .lE C /m
2
We have here made the replacements l 0 D ilE 0

and l D lE and rotated the inte-
o
gration contour of lE 90 , which with the positions of the poles should give the
same result. The integration over dD lE is separated into an integration over the
0
D-dimensional sphere ˝D and the linear integration over the component lE . This
corresponds in three dimensions to the integration over the two-dimensional angular
coordinates and the radial coordinate (see below).
Z
dD k 1 i.1/n .n D=2/ 1
D ; (G.20)
.2/ .k C s C i/
D 2 n .4/D=2 .n/ s nD=2
Z
k
d4 k D 0; (G.21)
.k 2 C s C i/n
Z
dD k kk i.1/n .n D=2 1/ 1
D : (G.22)
.2/D .k 2 C s C i/n .4/D=2 .n/ s nD=21
G.2.3.2 Covariant Gauge
Compared to Adkins [1] (A1a), (A3), and (A5a):
p ! qI M 2 ! sI ! ! D=2I ˛ ! nI ! nI Q D p ! qI A ! g I
! w D q 2 s;
Z
dD k 1 i.1/n 1 .n D=2/
D ; (G.23)
.2/D .k 2 C 2kq C s C i/n .4/D=2 .n/ wnD=2
Z
dD k k i.1/n 1 .n D=2/
D q ; (G.24)
.2/D .k 2 C 2kq C s C i/n .4/D=2 .n/ wnD=2
Z
dD k k k
.2/D .k 2 C 2kq C s C i/n

i.1/n 1 .n D=2/ g .n 1 D=2/
D q
q : (G.25)
.4/D=2 .n/ wnD=2 2 wn1D=2
G.2.3.3 Noncovariant Gauge
Compared to Adkins [1] (A1b), (A4), and (A5b):
p ! qI M 2 ! sI ! ! D=2I ˛ ! nI ˇ ! 1I ! n C 1I k2 ! k2 I

Q D py ! qyI
1y
A ! g C ı;0 ı;0 I .AQ/ ! q y ı0 .1 y/q0 ;
y
! w D q 2 y 2 C .1 y/yq02 sy C 2 .1 y/
D q2 y 2 C yq02 sy C 2 .1 y/;
Z
dD k 1 1
.2/ .k C 2kq C s C i/ k 2
D 2 n 2
Z 1
i.1/n 1 .n C 1 D=2/
D dy y n11=2 ; (G.26)
.4/D=2 .n C 1/ 0 wnC1D=2
Z
dD k k 1
.2/ .k C 2kq C s C i/ k 2
D 2 n 2
Z 1
i.1/n 1 .nC1D=2/
D dy y n11=2 q y C ı;0 q0 .1 y/ ;
.4/D=2 .n/ 0 wnC1D=2
(G.27)
6 k ! 6 q y 0 q0 .1 y/ D q y 0 q0 ;
Z
dD k kk 1
.2/ .k C 2kq C s C i/ k 2
D 2 n 2
Z 1
i.1/n 1 ˚
D D=2
dy y n11=2
q y C ı;0 q0 .1 y/
.4/ .n/ 0
.n C 1 D=2/
Œq y C ı;0 q0 .1 y/g
wnC1D=2
1 ˚ .n D=2/
g C ı;0 ı;0 .1 y/=y (G.28)
2 wnD=2
1
6 k ! q y 0 q0 in first part and 6 k 6 k ! C .1 y/=y in second.
2
Z
dD k ki k kj 1
.2/ .k C 2kq C s C i/ k 2
D 2 n 2
Z 1
i.1/n 1 ˚ i j 3
D D=2
dy y n11=2
q q q y C q i q q j ı0 .1 y/y 2
.4/ .n/ 0
.n C 1 D=2/ 1˚
nC1D=2
C y g i q j C g j q i C g j i q
w 2
.n D=2/
ı0 g q0 .1 y/
ij
(G.29)
wnD=2
G.3 Gamma Function
The Gamma function can be defined by means of Euler’s integral

Z 1
.z/ D dt t z1 et : (G.30)
1
For integral values, we have the relation
.n/ D .n 1/Š (G.31)
and generally
.z/ D .z 1/ .z 1/: (G.32)
The Gamma function can also be expressed by means of
1
1 Y z z=n
D zeE z 1C e ; (G.33)
.z/ nD1
n
where E is Euler’s constant, E D 0:5772 : : :

The Gamma function is singular, when z is zero or equal to a negative integer.
Close to zero, the function is equal to
1
./ D E C O./; (G.34)

which follows directly from the expansion above. We shall now derive the
corresponding expression close to negative integers.
G.3 Gamma Function 333
G.3.1 z D 1
Y1
1 E .1C / 1C
D .1 C / e 1 e.1C /=n
(G.35)
.1 / nD1
n
Q
The first few factors of the product are (to orders linear in )
e1C D e1 .1 C /;

1C 1
1 e.1C /=2 D .1 / e1=2 .1 C =2/;
2 2

1C 2
1 e.1C /=3 D .1 =2/ e1=3 .1 C =3/;
3 3

1C 3
1 e.1C /=4 D .1 =3/ e1=4 .1 C =4/;
4 4
which in the limit becomes
eE 1 C Œ1 1=2 1=.2 3/ 1=.3 4/ eE
using the expansion
1 C 1=2 C 1=3 C 1=4 C C 1=M ) ln M C E : (G.36)
This gives
1
.1 / D C E 1 C O./: (G.37)

This can also be obtained from (G.32).
G.3.2 z D 2
Y1
1 E .2C / 2C
D .2 C / e 1 e.2C /=n
: (G.38)
.2 / nD1
n
Q
The first few factors of the product are (to orders linear in )
.1 C / e2C D .1 C / e2 .1 C /;
=2 e.2C /=2

D =2 e1 .1 C =2/;

2C 1
1 e.2C /=3 D .1 / e2=3 .1 C =3/;
3 3

2C 2
1 e.2C /=4 D .1 =2/ e2=4 .1 C =4/;
4 4

2C 3
1 e.2C /=5 D .1 =3/ e2=5 .1 C =5/;
5 5
which in the limit becomes
e2E .1 C Œ5=2 2=.1 3/ 2=.2 4/ 2=.3 5/ / e2E .1 C /:
This gives

1 1
.2 / D C E 1 1=2 C O./ : (G.39)
2
This is consistent with the formula
.2z/ D .2/1=2 22z1=2 .z/ .z C 1=2/: (G.40)
The step-down formula (G.32) yields

1 1
.3 / D C E 1 1=2 1=3 ; (G.41)
23
which can be generalized to
" #
.1/N 1 1 XN
1
.N / D C E : (G.42)
NŠ nD1
n
References
(1983)
2. Fried, H.M., Yennie, D.M.: New techniques in the Lamb shift calculation. Phys. Rev. 112,
1391–1404 (1958)
6. t’Hooft, G., Veltman, M.: Regularization and renormalization of gauge fields. Nucl. Phys. B
44, 189–213 (1972)
Appendix H
Feynman Diagrams and Feynman Amplitude
In this appendix, we shall summarize the rules for evaluatiig Feynman diagrams of
the different schemes, discussed in this book. These rules are based on the rules
formulated by Feynman for the so-called Feynman amplitude, a concept we shall
also use here.
H.1 Feynman Diagrams
H.1.1 S-Matrix
The Feynman diagrams for the S-matrix have an outgoing orbital line for each
electron-field creation operator
O .x/
6
6
and an incoming orbital line for each electron-field absorption operator
O .x/
6
6
and a vertex diagram
A iec˛ A
for each interaction point.
335
336 H Feynman Diagrams and Feynman Amplitude
This leads to an electron propagator for each contracted pair of electron-field

operators:
2
R
! 6 O .x1 / O .x2 / D i SF .x2 ; x1 / D i d!
SF .!I x 2 ; x 1 /
6 2
1
and a photon propagator for each contracted pair of photon-field operators:
z R
1, - ,2 A .x1 /A .x2 / D i DF .x2 ; x1 / D i dz
2 DF .zI x 2 ; x 1 /.
Thus, there is a photon interaction (4.45), including the vertices,
z R R
; 1 - ; 2
dz
2
.i/I.zI x 2 ; x 1 / D d z .i/ e 2 c 2 ˛1 ˛2 DF .zI x 2 ; x 1 /
for each photon exchange, and a corresponding diagram
2
1 2 iVC D i 4 e0 r12
for each Coulomb interaction, VC , between the electrons, and a potential diagram
iV
for each energy potential, V (Fig. 4.6).
H.1.2 Green’s Function
The Feynman diagrams of the Green’s function are identical to those of the S-matrix
with the exception that all outgoing and incoming lines represent electron orbitals.
H.1.3 Covariant Evolution Operator
The Feynman diagrams of the Covariant evolution operator are identical to those of
the Green’s function with the exception that there are creation/absorption operator
lines attached to all outgoing/incoming orbital lines.
H.2 Feynman Amplitude 337
H.2 Feynman Amplitude
The Feynman amplitude, M, contains

An electron propagator (4.10)
Z
d!
i SF .!I x 2 ; x 1 /
2
for each internal orbital line.

A photon interaction
Z Z
dz dz
.i/I.zI x 2 ; x 1 / D .i/ e 2 c 2 ˛1 ˛2 DF .zI x 2 ; x 1 /
2 2
for each photon exchange, including the vertices.

At each vertex space integrations and a time integral 2 .arg/, where the
argument is equal to incoming minus outgoing energy parameters.
A factor of 1 and a trace symbol for each closed orbital loop.
The integration over the energy parameters leads to a factor of i for each “non-
trivial.” (The integral is considered to be “trivial,” when it contains a function
from the time integration.)
The S-matrix is related to the Feynman amplitude by (4.110)
S D 2 ı.Ein Eout / M; (H.1)
which gives the energy contribution (4.111)
E D ıEin;Eout h jiMj i: (H.2)
The Green’s function is in the equal-time approximation related to the Feynman

amplitude, M, by (5.25)
G.x; x 0 I x0 ; x00 / D eit .E0 H0 / M.x; x 0 I x 0 ; x 00 / eit0 .E0 H0 / (H.3)
and analogously for the covariant-evolution operator (6.10)
U.t; 1/jabi D eit .E0 H0 / jrsi hrsjiMjabi eit0 .E0 H0 / : (H.4)
In this formalism, the contribution to the effective interaction is
h jVeff j i D h jMj i: (H.5)

338 H Feynman Diagrams and Feynman Amplitude
Illustrations
6 6
6 6
iSF 6 6iSF iSF 6 6iSF
- i Vsp iSF 6 6iSF - i Vsp
- i Vsp
6 6 iSF 6 6iSF
6 6
6 6
A:i B:1 C:i
6 6
6 6 6 6 Pair 6 6 Pair
iI iI
iSF 6 6iSF iSF 6 6iSF iSF 6 6iSF
- i Vsp - i Vsp - i Vsp
iSF 6 6iSF iSF 6 6iSF iSF 6 6iSF
iI Pair iI Pair iI Pair
6 6 6 6 6 6
D:1 E:i F:1
Diagram A is a first-order S-matrix (Sect. 4.4), and the Feynman amplitude is
M D iVsp .E0 /:
Diagram B is a first-order covariant evolution-operator diagram with the unper-

turbed state as input and with outgoing electron propagators (Sect. 6.2). Here, there
are three energy parameters and two subsidiary conditions. This leaves one nontriv-
ial integration, giving a factor of i. This gives a factor of i2 .i/2 D 1 and
MD .E0 / Vsp .E0 /:
Diagram C is a first-order covariant evolution-operator diagram with incom-

ing and outgoing electron propagators (Sect. 8.1). Here, there are five parameters
and three conditions, leaving two nontrivial integrations. This gives the factor
i4 .i/3 D i and (8.9)
MD .E0 / iVsp .E0 / .E0 /:
Diagram D is a first-order covariant evolution-operator diagram with incoming

pair function (Sect. 6.2.1). This gives i4 .i/4 D 1
M D I Pair .E0 / Vsp .E0 / .E0 / I Pair :

H.2 Feynman Amplitude 339
Diagram E is an S-matrix diagram with incoming and outgoing pair functions

(Sect. 6.2.1). This gives i4 .i/5 D i and
M D iI Pair .E0 / Vsp .E0 / .E0 / I Pair :
Diagram F is a first-order covariant evolution-operator diagram with incoming

and outgoing pair functions (Sect. 6.2.1). Here, there are seven parameters and four
subsidiary conditions, yielding i6 .i/6 D 1 and
MD .E0 / I Pair .E0 / Vsp .E0 / .E0 / I Pair :

Appendix I
Evaluation Rules for Time-Ordered Diagrams
In nonrelativistic (MBPT) formalism, all interaction times are restricted to the

interval .t; 1/, and the Goldstone diagrams are used for the graphical representa-
tion. In the relativistic (QED) formalism, on the other hand, times are allowed in the
entire interval .1; 1/, and then Feynman diagrams, which contain all possible
time orderings, are the relevant ones to use.
For computational as well as illustrative purpose, it is sometimes useful also in
the relativistic case to work with time-ordered diagrams. It should be observed,
though, that time-ordered Feynman diagrams are distinct from Goldstone diagrams,
as we shall demonstrate here.1
When only particles states (above the Fermi level) are involved, time runs in the
positive direction, and the time-evolution operator can be expressed as (3.12)
Z t Z t Z t1
U.t; 1/ D 1 i dt1 V .t1 / C .i/2 dt1 V .t1 / dt2 V .t2 / C ;
1 1 1
(I.1)
where V .t/ is the perturbation in the interaction picture (3.16)

Z
V .t/ D d3 x O .x/ec˛ A .x/ O .x/: (I.2)
Core states and negative energy states are regarded as hole states below the Fermi
level with time running in the negative direction. Then, the corresponding time in-
tegration should be performed in the negative direction.
1
The treatment here is partly based on that in [2, Appdices C and D].
341
342 I Evaluation Rules for Time-Ordered Diagrams
I.1 Single-Photon Exchange

t
r s
z
- 2
1
a b
E
We consider first the time-ordered diagram for single-photon exchange with only
particle states involved. The time restrictions are here
t > t2 > t1 > 1;
which corresponds to the evolution operator (I.1)

Z t Z t2
it2 d2
.i/ 2
dt2 V .t2 / e dt1 V .t1 / eit1 d1 : (I.3)
1 1
The contraction of the radiation-field operators gives rise to a photon propagator

(4.18)
A .x2 /A .x1 / D i DF .x2 ; x1 /
and this leads to the interaction (4.46)

Z Z
dz iz.t2 t1 / 2c 2 k dk f .kI x 1 ; x 2 /
I.x2 ; x1 / D e 2 c 2 ˛1 ˛2 DF .x2 ; x1 / D e :
2 z2 c 2 k 2 C i
The time dependence at vertex 1 then becomes eit1 d1 , where
d1 D "a "r z C i:
This parameter is referred to as the vertex value and given by the incoming minus
the outgoing orbital energies/energy parameters at the vertex. Similarly, we define
d2 D "b "s C z C i
d12 D d1 C d2 D E "r "s (I.4)
with E D "a C "b . This leads to the time integrals

Z Z
t
it2 d2
t2
eit d12 1
.i/ 2
dt2 e dt1 eit1 d1 D : (I.5)
1 1 d12 d1
I.2 Two-Photon Exchange 343
Together with the opposite time ordering, t > t1 > t2 > 1, the denominators
become

1 1 1 1 1 1
C D C : (I.6)
d12 d1 d2 E "r "s "a "r z C i "b "s C z C i
This leads to the Feynman amplitude (6.11)

Z
dz 1 1 1
Msp D i C
2 E "r "s "a "r z C i "b "s C z C i
Z
2c 2 k dk f .k/
(I.7)
z2 c 2 k 2 C i
or
1
hrsjMspjabi D hrsjVsp .E/jabi (I.8)
E "r "s
with
Z
1 1
Vsp .E/ D c dk f .k/ C : (I.9)
"a "r ck C i "b "s ck C i
If the interaction is instantaneous, then the time integral becomes

Z t
eit d12 eit .E"r "s /
i dt12 eit12 d12 D D ; (I.10)
1 d12 E "r "s
which for t D 0 is the standard MBPT result.
I.2 Two-Photon Exchange
Next, we consider the diagrams in Fig. I.1.

We extend the definitions of the vertex values:
d1 D " a " t z I d2 D " b " u C z I d3 D "t "r z0 d4 D "u "s C z0 ;
d12 D d1 C d2 D E "t "u I d13 D "a "r z z0 I d24 D "b "s C z C z0 ;
d123 D E "r "u z0 I d124 D E "t "s C z0 I d1234 D E "r "s ;
i.e., given by the incoming minus the outgoing energies of the vertex. There is a
damping term ˙i for integration going to 1.
r s
r s r s t r s
t t z 4 t
r s r s r
- r s
0 4 1 1 2
-
z -z
3 u t 4 t u z
u
t z 2 3 u 3 1
-
1 t 0 4
a b a 2 a 2 -
z
b b a 3 b
E D "a C "b E E E
Fig. I.1 Time-ordered Feynman diagrams for the two-photon ladder with only particle states (left)
and with one and two intermediate hole states (right)
I.2.1 No Virtual Pair
We consider now the first diagram above, where only particle states are involved.
We assume that it is reducible, implying that the two photons do not overlap in time.
Then the time ordering is
t > t4 > t3 > t2 > t1 > 1:
This leads to the time integrations

Z t Z t4 Z t3 Z t2
.i/4 dt4 eit4 d4 dt3 eit3 d3 dt2 eit2 d2 dt1 eit1 d1
1 1 1 1
eit d1234
D : (I.11)
d1234 d123 d12 d1
Changing the order between t1 and t2 and between t3 and t4 leads to the
denominators

1 1 1 1 1 1
C C : (I.12)
d1234 d123 d124 d12 d1 d2
Here, all integrations are being performed upward, which implies that all denomi-
nators are evaluated from below.
If the interaction 1–2 is instantaneous, then the integrations become
Z t Z t4 Z t3
it4 d4 it3 d3 eit d1234
.i/ 3
dt4 e dt3 e dt12 eit12 d12 D
1 1 1 d1234 d123 d12
(I.13)
I.2 Two-Photon Exchange 345
and together with the other time ordering

eit d1234 1 1 1
C : (I.14)
d1234 d123 d124 d12
If both interactions are instantaneous, we have
Z Z
t t34
eit d1234
.i/2 dt34 eit34 d34 dt12 eit12 d12 D (I.15)
1 1 d1234 d12
consistent with the MBPT result [1, Sect. 12.2].
I.2.2 Single Hole
Next, we consider the two-photon exchange with a single hole, represented by the
second diagram above. We still assume that the diagram is reducible, implying that
the two photons do not overlap in time. The time ordering is now
t > t4 > t3 > t2 > 1 and 1 > t1 > t4 ;
but the order between t1 and t is not given.

If this is considered as a Goldstone diagram, all times (including t1 ) are restricted
to tn < t, which leads to
Z 1 Z t1 Z t4 Z t3
it1 d1 it4 d4 it3 d3
dt1 e dt4 e dt3 e dt2 eit2 d2
t 1 1 1
eit d1234
D : (I.16)
d1234 d234 d23 d2
Note that the last integration is being performed in the negative direction, due to the
core hole. This is illustrated in Fig. I.2 (left).
Considered as a Feynman diagram, the time t1 can run to C1, which leads to
Z t4 Z t Z t4 Z t3
1 1 1 1
eit d1234
D : (I.17)
d1234 d1 d23 d2
Here, the last integration is still performed in the negative direction, this time from
C1 to t4 , and this leads to a result different from the previous one. In the Goldstone
case all denominators are evaluated from below, while in the Feynman case one of
them is evaluated from above (see Fig. I.2, right). For diagrams diagonal in energy
we have d1234 D 0, and hence d1 D d234 , which implies that in this case the two
results are identical.
t t
r s d1234 r s d1234
? ?
1
-z d234 1
-z d1 6
t 4 d23 ? t 4 d23
3 d2
? 3 d2
?
? ?
a 2 a 2
b b
Goldstone Feynman
Fig. I.2 Time-ordered Goldstone and Feynman diagrams, respectively, for two-photon exchange
with one virtual pair. In the latter case, one denominator (at vertex 1) is evaluated from above
Let us next consider the third diagram in Fig. I.1, where the time ordering is
t > t4 > t1 > t3 > t2 > 1:
Here, all times are limited from above in the Goldstone as well as the Feynman
interpretation, and this leads in both cases to
Z t Z 1 Z t1 Z t3
1 t4 1 1
eit d1234
D : (I.18)
d1234 d123 d23 d2
I.2.3 Double Holes
The last diagram in Fig. I.1, also reproduced in Fig. I.3, represents double virtual
pair. Considered as a Goldstone diagram, the time ordering is
t > t2 > t1 > t4 > t3 > 1;
which yields
Z t Z t2 Z t1 Z t4
1 1 1 1
eit d1234
D : (I.19)
d1234 d134 d34 d3
I.3 General Evaluation Rules 347
t t
r s d1234 r s d1234
2 ? 2 ?
d134 d2 6
z z
u ? u
1 d34 1 d12 6
t z0
- 4 d3 ? t z0
- 4 d3
3 ? 3 ?
a b a b
Goldstone Feynman
Fig. I.3 Time-ordered Goldstone and Feynman diagrams, respectively, for two-photon exchange
with two virtual pairs. In the latter case, two denominators (at vertices 1 and 2) are evaluated from
above
This is illustrated in Fig. I.3 (left).

Considered as a Feynman diagram, we have instead 1 > t2 > t1 , which leads to
Z t Z t4 Z t4 Z t1
1 1 1 1
eit d1234
D ; (I.20)
d1234 d3 d12 d2
where two integrations are performed in the negative direction.
I.3 General Evaluation Rules
We can now formulate evaluation rules for the two types of diagrams considered
here. For (nonrelativistic) Goldstone diagrams, the rules are equivalent to the stan-
dard Goldstone rules [1, Sect. 12.2]
There is a matrix element for each interaction.
For each vertex, there is a denominator equal to the vertex sum (sum of vertex
values): incoming minus outgoing orbital energies and z C i for crossing photon
line (leading to ck after integration) below a line immediately above the vertex.
For particle/hole lines, the integration is performed in the positive/negative
direction.
For the relativistic Feynman diagrams, the same rules hold, with the exception
that
For a vertex where time can run to C1, the denominator should be evaluated
from above with the denominator equal to the vertex sum above a line immedi-
ately below the vertex (with z i for crossing photon line, leading to Cck).
References
Appendix J
Some Integrals
J.1 Feynman Integrals
In this section, we shall derive some integrals, which simplify many QED calcula-
tions considerably (see the books of Mandl and Shaw [1, Chap. 10] and Sakurai [2,
Appendix D]), and we shall start by deriving some formulas due to Feynman.
We start with the identity
Z b
1 1 dt
D : (J.1)
ab ba a t2
With the substitution t D b C .a b/x, this becomes

Z 1 Z 1
1 dx dx
D D : (J.2)
ab 0 Œb C .a b/x2 0 Œa C .b a/x2
Differentiation with respect to a yields

Z 1
1 xdx
D2 : (J.3)
a2 b 0 Œb C .a b/x3
Similarly, we have
Z 1 Z x
1 1
D2 dx dy
abc 0 0 Œa C .b a/x C .c b/y3
Z d Z 1x
1
D2 x dy : (J.4)
0 0 Œa C .b a/x C .c a/y3
Next we consider the integral

Z Z Z 1
1 dk0
4
d k 2 D 4 jkj djkj
2
:
.k C s C i/3 1 .k 2 C s C i/3
349
350 J Some Integrals
The second integral can be evaluated by starting with

Z 1
dk0 i
Dp
1 k0 jkj C s C i
2 2
jkj2 s
evaluated by residue calculus, and differentiating twice with respect to s. The inte-
gral then becomes
Z Z
1 3i 2 jkj2 djkj i 2
4
d k 2 D D : (J.5)
.k C s C i/3 2 .jkj2 C s/5=2 2s
The second integral can be evaluated from the identity
x2 1 s
D 2 2
.x C s/
2 5=2 .x C s/ 3=2 .x C s/5=2
and differentiating the integral

Z p
dx
p D ln x C x 2 C s
x2 C s
yielding Z
x2 1
D :
.x 2 C s/5=2 3s
For symmetry reason, we find
Z
k
d4 k D 0: (J.6)
.k 2 C s C i/3
Differentiating this relation with respect to k leads to

Z Z
kk g 1 i 2 g
d4 k D d4 k D (J.7)
.k C s C i/
2 4 3 .k 2 C s C i/ 3 6s
using the relation (A.4).

By making the replacements
k )kCq qnd s ) s q 2 ;
the integrals (J.5) and (J.6) lead to

Z
1 i 2
d4 k 2 D ; (J.8)
.k C 2kq C s C i/3 2.s q 2 /
Z Z
4 k 4 q i 2 q
d k 2 D d k D :
.k C 2kq C s C i/3 .k 2 C 2kq C s C i/3 2.s q 2 /
(J.9)
R d3k eikr12
J.2 Evaluation of the Integral .2/3 q 2 k 2 Ci
351
Differentiating the last relation with respect to q leads to

Z
kk i 2 g 2q q
d4 k 2 D C ‹‹: (J.10)
.k C 2kq C s C i/4 12 s q 2 .s q 2 /2
Differentiating the relation (J.8) with respect to s yields

Z
1 i 2
d4 k 2 D ; (J.11)
.k C 2kq C s C i/ 4 6.s q 2 /2
which can be generalized to arbitrary integer powers 3
Z
1 .n 3/Š 1
d4 k 2 D i 2 : (J.12)
.k C 2kq C s C i/ n .n 1/Š .s q 2 /n2
This can also be extended to nonintegral powers
Z
1 .n 2/ 1
d4 k 2 D i 2 (J.13)
.k C 2kq C s C i/n .n/ .s q 2 /n2
and similarly
Z
k .n 2/ q
d4 k D i 2 ; (J.14)
.k 2 C 2kq C s C i/ n .n/ .s q 2 /n2
Z
kk 2 .n 3/ .2n 3/ q q:

g
4
d k 2 D i C ;
.k C 2kq C s C i/n 2 .n/ .s q 2 /n2 .s q 2 /n3
(J.15)
Z
d 3k eikr12
J.2 Evaluation of the Integral
.2/3 q 2 k2 C i
Using spherical coordinates k D .; ; /, . D jkj/, we have with d3 k D

2 d sin d d˚ and r12 D jx 1 x 2 j
Z Z 1 Z
d3 k eik.x1 x2 / 2 d
D .2/2 d sin eir12 cos
.2/3 q 2 k2 C i 0 q 2 C i
2
0
Z
i d eir12 eir12
1
D
4 2 r12 0 q 2 2 C i
Z 1
i d eir12 eir12
D ; (J.16)
8 2 r12 1 q 2 2 C i
where we have in the last step used the fact that the integrand is an eve n function
of . The poles appear at D ˙q.1 C i=2q/. eir12 is integrated over the posi ti ve
and eir12 over the negati ve half-plane, which yields e˙iqr12 =.4 r12 / with the
upper sign for q > 0. The same result is obtained if we change the sign of the
exponent in the numerator of the original integrand. Thus, we have the result
Z Z 1
d3 k e˙ik.x1 x2 / 1 2 d sin.r12 / eijqjr12
D D : (J.17)
.2/3 q 2 k2 C i 4 2 r12 0 q 2 2 C i 4 r12
The imaginary part of the integrand, which is an odd function, does not contribute
to the integral.
J.3 Evaluation of the Integral

Z
d 3k
O ˛2 kO
eikr12
˛ 1 k
.2/3 q 2 k2 C i
The integral appearing in the derivation of the Breit interaction (F.53) is

Z
d3 k eikr 12
I2 D .˛ 1 k/.˛ 2 k/
.2/3 q 2 k2 C i
Z
d3 k eikr 12
D .˛1 r 1 /.˛2 r 2 / : (J.18)
.2/ k .q 2 k2 C i/
3 2
Using (J.16), we then have
Z
i 1 d eir12 eir12
I2 D .˛1 r 1 /.˛2 r 2 /
8 2 r12 1 .q 2 2 C i/
Z 1
1 2 d sin.kr12 /
D .˛1 r 1 /.˛2 r 2 / : (J.19)
4 2 r12 0 2 .q 2 2 C i/
The poles appear at D 0 and D ˙.q C i=2q/. The pole at D 0 can be treated
with half the pole value in each half plane. For q > 0, the result becomes
1 eiqr12 1

4 r12 q 2
and for q > 0 the same result with q in the exponent. The final result then becomes
R
J.3 Evaluation of the Integral d3k
.2/3
˛1 kO ˛2 kO eikr12
q 2 k 2 Ci
353
Z
d3 k eikr 12 1 eijqjr12 1
.˛1 k/.˛2 k/ D .˛1 r 1 /.˛2 r 2 /
.2/3 q 2 k2 C i 4 r12 q2
1
D .˛1 r 1 /.˛2 r 2 /
4 2 r12
Z 1
2 d sin.r12 /
: (J.20)
0 2 .q 2 2 C i/
Assuming that our basis functions are eigenfunctions of the Dirac hamiltonian
hO D , we can process this integral further. Then the commutator with an arbitrary
function of the space coordinates is
h i
hO D ; f .x/ D c ˛ pO f .x/ C ŒU; f .x/ : (J.21)
The last term vanishes if the potential U is a local function, yielding

h i
hO D ; f .x/ D c ˛ pO f .x/ D ic ˛ r f .x/ : (J.22)
In particular
h i
hO D ; eikx D ic ˛ r eikx D c ˛ kO eikx : (J.23)
We then find that
1
.˛ r /1 .˛ r /2 eikx D hD ; eikx 1
hD ; eikx 2
(J.24)
c2
with the matrix element
hrsj .˛ r /1 .˛ r /2 eikx jabi D q 2 eikx (J.25)
using the notation in (F.49). The integral (J.18) then becomes
Z Z
d3 k O O eikr 12 d3 k q 2 eikr 12
I2 D .˛ 1 k/.˛2 k/ D
.2/3 q 2 k2 C i .2/3 k2 q 2 k2 C i
(J.26)
provided that the orbitals are generated by a hamiltonian with a local potential.
References
2. Sakurai, J.J.: Advanced Quantum Mechanics. Addison-Wesley Publ. Co., Reading, Mass.
(1967)
Appendix K
Unit Systems and Dimensional Analysis
K.1 Unit Systems
K.1.1 SI System
The standard unit system internationally agreed upon is the SI system or System
Internationale.1 The basis units in this system are given in the following table:
Quantity SI unit Symbol

Length Meter m
Mass Kilogram kg
Time Second s
Electric current Ampere A
Thermodynamic temperature Kelvin K
Amount of substance Mole mol
Luminous intensity Candela cd
For the definition of these units, the reader is referred to the NIST WEB page (see
footnote). From the basis units – particularly the first four – the units for most other
physical quantities can be derived.
K.1.2 Relativistic or “Natural” Unit System
In scientific literature, some simplified unit system is frequently used for conve-
nience. In relativistic field theory the relativistic unit system is mostly used, where
the first four units of the SI system are replaced by
1
For further details, see The NIST Reference on Constants, Units, and Uncertainty
(http://physics.nist.gov/cuu/Units/index.html).
355
356 K Unit Systems and Dimensional Analysis
Quantity Relativistic unit Symbol Dimension

Mass Rest mass of the electron m kg
Velocity Light velocity in vacuum c ms1
Action Planck’s constant divided by 2 „ kg m2 s1
Dielectricity Dielectricity constant of vacuum 0 A2 s4 kg1 m3
In the table, the dimension of the relativistic units in SI units are also shown. From
these four units, all units that depend only on the four SI units kg, s, m, A can be
derived. For instance, energy that has the dimension kg m2 m2 has the relativistic
unit me c 2 , which is the rest energy of the electron (511 keV). The unit for length is
„
D =2 0; 386 1012 m;
me c
where is Compton wavelength and the unit for time is 2c= 7; 77 104 s).
K.1.3 Hartree Atomic Unit System
In atomic physics, the Hartree atomic unit system is frequently used, based on the
following four units
Quantity Atomic unit Symbol Dimension

Mass Rest mass of the electron m kg
Electric charge Absolute charge of the electron e As
Action Planck’s constant divided by 2 „ kg m2 s1
Dielectricity Dielectricity constant of vacuum times 40 A2 s4 kg1 m3
4
Here, the unit for energy becomes
me 4
1H D ;
.40 /2 „3
which is known as the Hartree unit and equals twice the ionization energy of the
hydrogen atom in its ground state (27.2 eV). The atomic unit for length is
40 „2
a0 D
me 2
known as the Bohr radius or the radius of the first electron orbit of the Bohr hydro-
gen model (0:529 1010 m). The atomic unit of velocity is ˛c, where
e2
˛D (K.1)
40 c„
K.2 Dimensional Analysis 357
is the dimensionless fine-structure constant (1/137,036). Many units in these two

systems are related by the fine-structure constant. For instance, the relativistic length
unit is ˛a0 .
K.1.4 cgs Unit Systems
In older scientific literature, a unit system, known as the cgs system, was frequently
used. This is based on the following three units:
Quantity cgs Unit Symbol

Length Centimeter cm
Mass Gram g
Time Second s
In addition to the three units, it is necessary to define a fourth unit to be able to de-
rive most of the physical units. Here, two conventions are used. In the electrostatic
version (ecgs), the proportionality constant of Coulombs law, 40 , is set equal
to unity, and in the magnetic version (mcgs) the corresponding magnetic constant,
0 =4, equals unity. Since these constants have dimension, the systems cannot be
used for dimensional analysis (see below).
The most frequently used unit system of cgs type is the so-called Gaussian unit
system, where electric units are measured in ecgs and magnetic ones in mcgs. This
implies that certain formulas will look differently in this system, compared to a sys-
tem with consistent units. For instance, the Bohr magneton, which in any consistent
unit system will have the expression
e„
B D
2m
will in the mixed Gaussian system have the expression

e„
B D
2mc
which does not have the correct dimension. Obviously, such a unit system can easily
lead to misunderstandings and should be avoided.
K.2 Dimensional Analysis
It is often useful to check physical formulas by means of dimensional analysis,

which, of course, requires that a consistent unit system, like the SI system, is being
used. Below we list a number of physical quantities and their dimension, expressed
in SI units, which could be helpful in performing such an analysis.
In most parts of the book, we have set „ D 1, which simplifies the formulas.
This also simplifies the dimensional analysis, and in the last column below we have
(after the sign )) listed the dimensions in that case.
kg m 1
Œforce D N D 2
) ;
s ms
kg m2 1
Œenergy D J D Nm D ) ;
s2 s
kg m2
Œaction; „ D Js D ) 1;
s
J kg m2 1
Œelectric potential D V D D ) ;
As As3 As2
kg m 1
Œelectric field; E D V=m D ) ;
As3 Ams2
kg 1
Œmagnetic field; B D Vs=m2 D 2
) ;
As Am2 s
kg m 1
Œvector potential; A D Vs=m D 2
) ;
As Ams
kg m 1
Œmomentum; p D ) ;
s m
As
Œcharge density; D 3 ;
m
A
Œcurrent density; j D 2 ;
m
kg m 1
Œ0 D N=A2 D 2 2 ) 2 ;
A s A ms
A2 s4 A2 s3
Œ0 D Œ1=0 c 2 D 3
) :
kg m m
Fourier transforms
Z
dz
DF .x1 ; x2 / D DF .zI x 1 ; x 2 / eiz.t2 t1 / ;
2
ŒA.!; x/ D sŒA.x/;
ŒA.!; k/ D sm3 ŒA.x/;
ŒA.k/ D m4 ŒA.x/:
K.2 Dimensional Analysis 359
Photon propagator
1
ŒDF .x; x/ D ;
A2 m2 s2
s
Œ0 DF .x; x/ D ;
m3
Œ0 DF .k/ D sm;
s
Œ0 DF .k0 ; x/ D 2 ;
m
1
Œ0 DF .t; x/ D 2 ;
m
s2
Œ0 DF .z; x/ D z D ck0 ;
m3
Œ0 DF .z; k/ D s2 ;
1
Œe 2 c 2 DF .z; x/ D I.z; x/ D ;
s

e2 m
D :
0 s
Electron propagator
SOF .x; x/ ) s;
1
SF .x; x/ ) ;
m3
s
SF .z; x/ ) 3 ;
m
SF .z; k/ ) s;
SF .k/ ) m:
S-matrix
SO .x; x/ ) 1;
S.z; x/ ) s;
S.z; k/ ) m4 :
Self energy
1
˙O .z/ ) ;
s
1
˙.z; x/ ) ;
s m3
1
˙.z; k/ ) ;
s
m
˙.k/ ) 2 :
s
Vertex
.z; k/ ) 1;
m
.p; p 0 / ) :
s
Abbreviations
CCA Coupled-cluster approach

CEO Covariant evolution operator
GML Gell-Mann–Low relation
HP Heisenberg picture
IP Interaction picture
LDE Linked-diagram expansion
MBPT Many-body perturbation theory
MSC Model-space contribution
NVPA No-virtual-pair approximation
PWR Partial-wave regularization
QED Quantum electrodynamics
SCF Self-consistent field
SI International unit system
SP Schrödinger picture
361
Index
A covariant evolution operator, 3, 119

Adiabatic damping, 51 covariant vector, 273
all-order method, 30 creation operator, 283
annihilation operator, 283 cut-off procedure, 248
antisymmetry, 22
B D
Banach space, 277 d’Alembertian operator, 274
Bethe–Salpeter equation, 3, 119, 193, 199, 225 de Broglie’s relations, 13
effective potential form, 205 density operator, 120
Bethe–Salpeter–Bloch equation, 6, 156, 173, difference ratio, 139
193, 199, 206 dimensional analysis, 355
Bloch equation dimensional regularization, 163
for Green’s operator, 146 Dirac delta function, 277
generalized, 20, 23 Dirac equation, 1, 295
bra vector, 290 Dirac matrices, 296
Breit interaction, 38, 73, 318 Dirac sea, 38
Brillouin–Wigner expansion, 153 Dirac-Coulomb Hamiltonian, 37
Brown–Ravenhall effect, 2, 38, 178, 194, 216, discretization technique, 42
219 dot product, 135
Dyson equation, 108, 201, 243
C
Cauchy sequence, 276
closure property, 292 E
complex rotation, 37 effective Hamiltonian, 5, 20
configuration, 22 intermediate, 37
conjugate momentum, 15, 306 effective interaction, 23, 31
connectivity, 37 Einstein summation rule, 289
continuity equation, 314 electron field operators, 285
contraction, 18 electron propagator, 61, 98
contravariant vector, 273 equal-time approximation, 96, 120, 124, 173,
coordinate representation, 63, 292 205, 208
counterterms, 138 Euler-Lagrange equations, 308
coupled-cluster approach, 30 evolution operator, 47
normal-ordered, 34 exponential Ansatz, 33
coupled-cluster-QED expansion, 196 normal-ordered, 34
covariance, 1 external-potential approach, 4, 208
363
364 Index
F intermediate normalization, 21
Feynman amplitude, 88, 96, 124, 174, 335 intruder state, 23, 36
Feynman diagram, 28, 60, 87, 125, 335 irreducible diagram, 40, 127
fine structure, 165
first quantization, 13
Fock space, 133, 277 K
photonic, 6, 51, 132, 211 ket vector, 290
fold, 126 Klein-Gordon equation, 295
folded diagram, 23 Kronecker delta factor, 278
Fourier transform, 63
functional, 275
Furry picture, 21, 27 L
Furry’s theorem, 86 Lagrange equations, 305
Lagrangian function, 309
Lamb shift, 2, 79, 84, 157
G Laplacian operator, 274
g-factor, 164 Lehmann representation, 97
gamma function, 332 linked diagram, 28, 107
gauge Lippmann–Schwinger equation, 205
Coulomb, 59, 68, 76, 81, 315, 324, 329 Lorentz covariance, 1, 39, 60, 94, 119
covariant, 59, 65, 122, 125, 328 Lorentz force, 308
Feynman, 59, 65, 66, 81, 324, 326, 328 Lorentz transformation, 1
Fried-Yennie, 329 Lorenz condition, 314
Landau, 328
non-covariant, 329
gauge invariance, 314 M
gauge transformation, 327 many-body Dirac Hamiltonian, 133
Gaunt interaction, 213, 217 matrix elements, 291
Gell-Mann–Low theorem, 51 matrix representation, 291
relativistic, 131 Maxwell’s equations, 311, 312
Goldstone diagram, 25 metric tensor, 273
Goldstone rules, 24, 28, 125 minimal substitutions, 307
Green’s function, 3, 91 model space, 20
projected, 113 complete, 22
Green’s operator, 119, 134 extended, 23
Grotsch term, 164 model state, 20
Gupta-Bleuler formalism, 325 model-space contribution, 29, 56, 126, 138,
142
momentum representation, 300
H MSC, 56, 126
Hamiltonian density, 309 multi-photon exchange, 127
Hartree–Fock model, 26
Heaviside step function, 282
Heisenberg picture, 202, 286 N
Heisenberg representation, 92 no-virtual-pair approximation, 2, 37
helium fine structure, 225 non-radiative effects, 39
Hilbert space, 277 norm, 275
hole state, 119 normal order, 18
Hylleraas function, 5, 163
P
I pair correlation, 30
instantaneous approximation, 4 parent state, 52, 131
interaction picture, 47 partitioning, 21
Index 365
Pauli spin matrices, 296 photon, 87, 246

perturbation proper, 108
Brillouin–Wigner, 111 sequence, 276
Rayleigh–Schrödinger, 1, 24 set, 275
photon propagator, 65 size consistency, 30
Poisson bracket, 15, 306 size extensive, 207
polarization tensor, 87 size extensivity, 30, 207
principal-value integration, 64 Slater determinant, 22, 284
spin-orbital, 22
spline, 41
Q state-specific approach, 37
QED effects, 39, 59, 79 state-universality, 36
QED potential, 181 subset, 275
quantization condition, 15 Sucher energy formula, 61
quasi-degeneracy, 23, 56
quasi-potential approximation, 4
quasi-singularity, 56 T
target states, 20
time-ordering, 49
Wick, 18
R
trace, 85
radiative effects, 39
transverse-photon, 77
reducible diagram, 40, 127
two-times Green’s function, 3, 111
reference-state contribution, 78, 142
regularization, 79, 237
Brown–Langer–Schaefer, 252 U
dimensional, 257 Uehling potential, 86, 128
partial-wave, 255 union, 275
Pauli–Willars, 248 unit system, 355
renormalization, 79, 237 cgs, 357
charge, 244 Hartree, 356
mass, 242 mixed, 300
resolution of the identity, 290 natural, 355
resolvent, 24 relativistic, 59, 355
reduced, 24 SI , 355
unlinked diagram, 28
S
S-matrix, 3, 60, 157 V
scalar potential, 311 vacuum polarization, 2, 39, 84, 87
scalar product, 274 valence universality, 21, 36
scalar retardation, 213, 217 vector potential, 311
scattering matrix, 60 vector space, 275
Schrödinger equation, 15 vertex correction, 39, 82, 186, 240, 246
Schrödinger picture, 47
Schwinger correction, 164
second quantization, 16, 283 W
self energy Ward identity, 83, 242
electron, 39 wave operator, 20
self energy, 108 Wick’s theorem, 19
electron, 79, 157, 186, 239, 244 Wickmann-Kroll potential, 86

Relativistic Many-Body Theory

Uploaded by

Copyright:

Available Formats

Relativistic Many-Body Theory

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Relativistic Many-Body Theory

Uploaded by

Copyright:

Available Formats

Springer Series on

Atomic, Optical, and Plasma Physics

For further volumes:

Library of Congress Control Number: 2011921721

c Springer Science+Business Media, LLC 2011

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Gothenburg, Sweden Ingvar Lindgren

Part I Basics: Standard Many-Body Perturbation Theory

2.7 Some Numerical Results of Standard MBPT and CC

Part II Quantum-Electrodynamics: One-Photon

5.3 Graphical Representation of the Green’s Function . .. . . . . . . . . . . . . . . . .100

6 Covariant Evolution Operator and Green’s Operator . .. . . . . . . . . . . . . . . . .119

7 Numerical Illustrations to Part II. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .157

7.1.6g-Factor of Hydrogen-Like Ions:

Part III Quantum-Electrodynamics Beyond Two-Photon Exchange:

8 Covariant Evolution Combined with Electron Correlation.. . . . . . . . . . . . .173

9 The Bethe–Salpeter Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .199

10 Implementation of the MBPT–QED Procedure

10.3.2 Full Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .219

11 Analytical Treatment of the Bethe–Salpeter Equation . .. . . . . . . . . . . . . . . . .225

12 Regularization and Renormalization .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .237

13 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .271

A Notations and Definitions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .273

B Second Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .283

C Representations of States and Operators . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .289

D Dirac Equation and the Momentum Representation . . . .. . . . . . . . . . . . . . . . .295

E Lagrangian Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .305

F Semiclassical Theory of Radiation .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .311

G Covariant Theory of Quantum ElectroDynamics .. . . . . . .. . . . . . . . . . . . . . . . .325

H Feynman Diagrams and Feynman Amplitude. . . . . . . . . . . .. . . . . . . . . . . . . . . . .335

I Evaluation Rules for Time-Ordered Diagrams. . . . . . . . . . .. . . . . . . . . . . . . . . . .341

J Some Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .349

K Unit Systems and Dimensional Analysis . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . .355

1.1 Standard Many-Body Perturbation Theory

The quantum-mechanical treatment of many-electron systems, based on the

I. Lindgren, Relativistic Many-Body Theory: A New Field-Theoretical 1

An equation or a theory, like the theory of relativity or Maxwell’s theory of electromagnetism, is

to the development of the modern form of the quantum-electrodynamic theory by

1.3 Bethe–Salpeter Equation

The first completely covariant treatment of a bound-state problem was presented

1.4 Helium Atom: Analytical Approach

An approach to solve the BS equation, known as the external-potential approach,

1.5 Field-Theoretical Approach to Many-Body

2.1 First Quantization

2.1.1 De Broglie’s Relations

As an introduction to the quantum mechanics, we shall derive the Schrödinger equa-

I. Lindgren, Relativistic Many-Body Theory: A New Field-Theoretical 13

2.1.2 The Schrödinger Equation

We can generalize the treatment above to an electron in three dimensions in an

p ! pO D i„r and x ! xO D x; (2.7)

and to the Schrödinger equation for a single electron

For an N -electron system, the Schrödinger equation becomes correspondingly3

where we assume the Hamiltonian to be of the form HO D HO 1 C HO 2 [see Appendix

which is consistent with the substitutions (2.7).

.tI x 1 ; x N / D eiE t =„ .x 1 ; x N /: (2.15)

The separation constant E is interpreted as the energy of the state.

2.2 Second Quantization

2.2.1 Schrödinger Equation in Second Quantization*

.x1 ; x2 / D hx1 ; x 2j.t/i ; (2.18)

.tI x 1 ; x N / D eiE t =„ .x 1 ; x N /: (2.15)

.x1 ; x2 / D hx1 ; x 2j.t/i ; (2.18)

on the state vectorj.t/i.

j.t/i D akl .t/ jkijli ; (2.21)

.x1 ; x2 / D hx1 ; x 2j.t/i D akl .t/ hx 1jki hx 2jli : (2.22)

hx 1 ; x 2 jH1j.t/i D .h1 .x 1 / C h1 .x 2 // .x1 ; x2 / D H1 .x1 ; x2 /:

hx 1 ; x 2 jH1j.t/i D H1 .x1 ; x2 /: (2.24)

x y D T Œxy N Œxy: (2.29)

N Œxy fx yg: (2.30)

O .x1 / O .x2 / D O .x2 / O .x1 / D

Œ˝; H0 P D Q .V ˝ ˝Veff / P; (2.56)

Œ˝; H0 P D .V ˝P ˝Veff /linked P: (2.82)

Œ˝1 ; H0 P D .V1 C V ˝1 C V ˝2 ˝1 W1 /linked; 1 P;

Œ˝2 ; H0 P D .V C V ˝2 ˝2 W2 /linked; 2 P: (2.91)