Quantum Field Theory I

Graduate Texts in Physics
Edouard B. Manoukian
Quantum
Field Theory I
Foundations and Abelian and
Non-Abelian Gauge Theories
Series editors
Kurt H. Becker, Polytechnic School of Engineering, Brooklyn, USA
Sadri Hassani, Illinois State University, Normal, USA
Bill Munro, NTT Basic Research Laboratories, Atsugi, Japan
Richard Needs, University of Cambridge, Cambridge, UK
Jean-Marc Di Meglio, Université Paris Diderot, Paris, France
William T. Rhodes, Florida Atlantic University, Boca Raton, USA
Susan Scott, Australian National University, Acton, Australia
H. Eugene Stanley, Boston University, Boston, USA
Martin Stutzmann, TU München, Garching, Germany
Andreas Wipf, Friedrich-Schiller-Univ Jena, Jena, Germany
Graduate Texts in Physics publishes core learning/teaching material for graduate-
and advanced-level undergraduate courses on topics of current and emerging fields
within physics, both pure and applied. These textbooks serve students at the
MS- or PhD-level and their instructors as comprehensive sources of principles,
definitions, derivations, experiments and applications (as relevant) for their mastery
and teaching, respectively. International in scope and relevance, the textbooks
correspond to course syllabi sufficiently to serve as required reading. Their didactic
style, comprehensiveness and coverage of fundamental material also make them
suitable as introductions or references for scientists entering, or requiring timely
knowledge of, a research field.
More information about this series at http://www.springer.com/series/8431

Quantum Field Theory I

Foundations and Abelian and Non-Abelian
Gauge Theories
123
The Institute for Fundamental Study
Naresuan University
Phitsanulok, Thailand
ISSN 1868-4513 ISSN 1868-4521 (electronic)

ISBN 978-3-319-30938-5 ISBN 978-3-319-30939-2 (eBook)
DOI 10.1007/978-3-319-30939-2
Library of Congress Control Number: 2016935720
© Springer International Publishing Switzerland 2016

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG Switzerland
Preface to Volume I
This textbook is based on lectures given in quantum field theory (QFT) over the
years to graduate students in theoretical and experimental physics. The writing of
the book spread over three continents: North America (Canada), Europe (Ireland),
and Asia (Thailand). QFT was born about 90 years ago, when quantum mechanics
met relativity, and is still going strong. The book covers, pedagogically, the wide
spectrum of developments in QFT emphasizing, however, those parts which are
reasonably well understood and for which satisfactory theoretical descriptions have
been given.
The legendary Richard Feynman in his 1958 Cornell, 1959–1960 Cal Tech
lectures on QFT of fundamental processes, the first statement he makes, the very
first one, is that the lectures cover all of physics.1 One quickly understands what
Feynman meant by covering all of physics. The role of fundamental physics is to
describe the basic interactions of Nature and QFT, par excellence, is supposed to
do just that. Feynman’s statement is obviously more relevant today than it was
then, since the recent common goal is to provide a unified description of all the
fundamental interactions in nature.
The book requires as background a good knowledge of quantum mechanics,
including rudiments of the Dirac equation, as well as elements of the Klein-Gordon
equation, and the reader would benefit much by reading relevant sections of my
earlier book : Quantum Theory: A Wide Spectrum (2006), Springer in this respect.
This book differs from QFT books that have appeared in recent years 2 in several
respects and, in particular, it offers something new in its approach to the subject, and
the reader has plenty of opportunity to be exposed to many topics not covered, or
1
R. P. Feynman, The Theory of Fundamental Processes, The Benjamin/Cummings Publishing Co.,
Menlo Park, California. 6th Printing (1982), page 1.
2
Some of the fine books that I am familiar with are: L. H. Ryder, Quantum Field Theory;
S. Weinberg, The Quantum Theory of Fields I (1995) & II (1996), Cambridge: Cambridge
University Press; M. Peskin and D. V. Schroeder, An Introduction to Quantum Field Theory, New
York: Westview Press (1995); B. DeWitt, The Global Approach to Quantum Field Theory, Oxford:
Oxford University Press (2014).
v
vi Preface to Volume I
just touched upon, in standard references. Some notable differences are seen, partly,
from unique features in the following material included in ours:
• The very elegant functional differential approach of Schwinger, referred to

as the quantum dynamical (action) principle, and its underlying theory are
used systematically in generating the so-called vacuum-to-vacuum transition
amplitude of both abelian and non-abelian gauge theories, in addition to the well-
known functional integral approach of Feynman, referred to as the path-integral
approach, which are simply related by functional Fourier transforms and delta
functionals.
• Transition amplitudes are readily extracted by a direct expansion of the vacuum-
to-vacuum transition amplitude in terms of a unitarity sum, which is most closely
related to actual experimental setups with particles emitted and detected prior and
after a given process and thus represent the underlying physics in the clearest
possible way.
• Particular emphasis is put on the concept of a quantum field and its particle
content, both physically and technically, as providing an appropriate description
of physical processes at sufficiently high energies, for which relativity becomes
the indispensable language to do physics and explains the exchange that takes
place between energy and matter, allowing the creation of an unlimited number
of particles such that the number of particles need not be conserved, and for
which a variable number of particles may be created or destroyed. Moreover,
quantum mechanics implies that a wavefunction renormalization arises in QFT
field independent of any perturbation theory – a point not sufficiently emphasized
in the literature.
• The rationale of the stationary action principle and emergence of field equations,
via field variations of transformation functions and generators of field variations.
The introduction of such generators lead, self consistently, to the field equations.
Such questions are addressed as: “Why is the variation of the action, within the
boundaries of transformation functions, set equal to zero which eventually leads
to the Euler-Lagrange equations?”, “How does the Lagrangian density appear in
the formalism?” “What is the significance in commuting/anti-commuting field
components within the interaction Lagrangian density in a theory involving field
operators?” These are some of the questions many students seem to worry about.
• A panorama of all the fields encountered in present high-energy physics, together
with the details of the underlying derivations are given.
• Schwinger’s point splitting method of currents is developed systematically in
studying abelian and non-abelian gauge theories anomalies. Moreover, an explicit
experimental test of the presence of an anomaly is shown by an example.
• Derivation of the Spin & Statistics connection and CPT symmetry, emphasizing
for the latter that the invariance of the action under CPT transformation is not
sufficient for CPT symmetry, but one has also to consider the roles of incoming
and outgoing particles.
Preface to Volume I vii
• The fine-structure effective coupling ˛ ' 1=128 at high energy corresponding

to the mass of the neutral Z 0 vector boson based on all the charged leptons and
all those contributing quarks of the three generations.
• Emphasis is put on renormalization theory, including its underlying general
subtractions scheme, often neglected in treatments of QFT.
• Elementary derivation of Faddeev-Popov factors directly from the functional
differential formalism, with constraints, and their modifications, and how they
may even arise in some abelian gauge theories.
• A fairly detailed presentation is given of “deep inelastic” experiments as a
fundamental application of quantum chromodynamics.
• Schwinger line integrals, origin of Wilson loops, lattices, and quark confinement.
• Neutrino oscillations,3 neutrino masses, neutrino mass differences, and the
“seesaw mechanism.”
• QCD jets and parton splitting, including gluon splitting to gluons.
• Equal importance is put on both abelian and non-abelian gauge theories,
witnessing the wealth of information also stored in the abelian case.4
• A most important, fairly detailed, and semi-technical introductory chapter is
given which traces the development of QFT since its birth in 1926 without
tears, in abelian and non-abelian gauge theories, including aspects of quantum
gravity, as well as examining the impact of supersymmetry, string theory, and the
development of the theory of renormalization, as a pedagogical strategy for the
reader to be able to master the basic ideas of the subject at the outset before they
are encountered in glorious technical details later.
• Solutions to all the problems are given right at the end of the book.
With the mathematical rigor that renormalization has met over the years and
the reasonable agreement between gauge theories and experiments, the underlying
theories are in pretty good shape. This volume is organized as follows. The first
introductory chapter traces the subject of QFT since its birth, elaborating on
many of its important developments which are conveniently described in a fairly
simple language and will be quite useful for understanding the underlying technical
details of the theory covered in later chapters including those in Volume II. A
preliminary chapter follows which includes the study of symmetry transformations
in the quantum world, as well as of intricacies of functional differentiation and
functional integration which are of great importance in field theory. Chapter 3
deals with quantum field theory methods of spin 1/2 culminating in the study of
anomalies in the quantum world. The latter refers to the fact that a conservation law
3
It is rather interesting to point out that the theory of neutrino oscillations was written up in this
book much earlier than the 2015 Nobel Prize in Physics was announced on neutrino oscillations.
4
With the development of non-abelian gauge theories, unfortunately, it seems that some students
are not even exposed to such derivations as of the “Lamb shift” and of the “anomalous magnetic
moment of the electron” in QED.
viii Preface to Volume I
in classical physics does not necessarily hold in the quantum world. Chapter 4, a
critical one, deals with the concept of a quantum field, the Poincaré algebra, and
particle states. Particular attention is given to the stationary action principle as
well as in developing the solutions of QFT via the quantum dynamical principle.
This chapter includes the two celebrated theorems dealing with CPT symmetry
and of the Spin & Statistics connection. A detailed section is involved with the
basic quantum fields one encounters in present day high-energy/elementary-particle
physics and should provide a useful reference source for the reader. Chapter 5 treats
abelian gauge theories (QED, scalar boson electrodynamics) in quite details and
includes, in particular, the derivations of two of the celebrated results of QED
which are the anomalous magnetic moment of the electron and the Lamb shift.
Chapter 6 is involved with non-abelian gauge theories (electroweak, QCD, Grand
unification). 5 Such important topics are included as “asymptotic freedom,” “deep
inelastic” scattering, QCD jets, parton splittings, neutrino oscillations, the “seesaw
mechanism” and neutrino masses, Schwinger-line integrals, Wilson loops, lattices,
and quark confinement. Unification of coupling parameters of the electroweak
theory and of QCD are also studied, as well as of spontaneous symmetry breaking
in both abelian and non-abelian gauge theories, and of renormalizability aspects of
both gauge theories, emphasizing the so-called BRS transformations for the latter.
We make it a point, pedagogically, to derive things in detail, and some of such
details are relegated to appendices at the end of the respective chapters with the
main results given in the sections in question. Five general appendices, at the end
of this volume, cover some additional important topics and/or technical details.
In particular, I have included an appendix covering some aspects of the general
theory of renormalization and its underlying subtractions scheme itself which is
often neglected in books on QFT. Fortunately, my earlier book, with proofs not just
words, devoted completely to renormalization theory – Renormalization (1983),
Academic Press – may be consulted for more details. The problems given at the
end of the chapters form an integral part of the book, and many developments in
the text depend on the problems and may include, in turn, additional material. They
should be attempted by every serious student. Solutions to all the problems are given
right at the end of the book for the convenience of the reader. The introductory
chapter together with the introductions to each chapter provide the motivation and
the pedagogical means to handle the technicalities that follow them in the texts.
I hope this book will be useful for a wide range of readers. In particular, I
hope that physics graduate students, not only in quantum field theory and high-
energy physics, but also in other areas of specializations will also benefit from it
as, according to my experience, they seem to have been left out of this fundamental
area of physics, as well as instructors and researchers in theoretical physics. The
content of this volume may be covered in one-year (two semesters) quantum field
theory courses.
5
QED and QCD stand, respectively, for quantum electrodynamics and quantum chromodynamics.
Preface to Volume I ix
In Volume II, the reader is introduced to quantum gravity, supersymmetry, and

string theory, 6 which although may, to some extent, be independently read by a
reader with a good background in field theory, the present volume sets up the
language, the notation, provides additional background for introducing these topics,
and will certainly make it much easier for the reader to follow. In this two-volume
set, aiming for completeness in covering the basics of the subject, I have included
topics from the so-called conventional field theory (the classics) to ones from the
modern or the new physics which I believe that every serious graduate student
studying quantum field theory should be exposed to.
Without further ado, and with all due respect to the legendary song writer Cole
Porter, let us find out “what is this thing called QFT?”
6
Entitled: Quantum Field Theory II: Introductions to Quantum Gravity, Supersymmetry, and String
Theory” (2016), Springer.
Acknowledgements
In the beginning of it all, I was introduced to the theoretical aspects of quantum

field theory by Theodore Morris and Harry C. S. Lam, both from McGill and to
its mathematical intricacies by Eduard Prugovečki from the University of Toronto.
I am eternally grateful to them. Over the years, I was fortunate enough to attend
a few lectures by Julian Schwinger and benefited much from his writings as well.
Attending a lecture by Schwinger was quite an event. His unique elegant, incisive,
physically clear approach and, to top it off, short derivations were impressive. When
I was a graduate student, I would constantly hear that Schwinger “does no mistakes.”
It took me years and years to understand what that meant. My understanding of this
is because he had developed such a powerful formalism to do field theory that,
unlike some other approaches, everything in the theory came out automatically and
readily without the need to worry about multiplicative factors in computations, such
as 2’s and other numerical factors, and, on top of this, is relatively easy to apply.
Needless to say this has much influenced my own approach to the subject. He had
one of the greatest minds in theoretical physics of our time.
I want to take this opportunity as well to thank Steven Weinberg, the late Abdus
Salam, Raymond Streater, and Eberhard Zeidler for the keen interest they have
shown in my work on renormalization theory.
I acknowledge with thanks the support I received from several colleagues, while
visiting their research establishments for extended periods, for doing my own thing
and writing up the initial notes on this project. These include Yasushi Takahashi
and Anton Z. Capri from the University of Alberta, Lochlainn O’Raifeartaigh
and John Lewis from the Dublin Institute for Advanced Studies, and Jiri Patera
and Pavel Winternitz from the University of Montreal. For the final developments
of the project, I would like to thank Sujin Jinahyon, the President of Nare-
suan University, Burin Gumjudpai, Seckson Sukhasena, Suchittra Sa-Nguansin,
and Jiraphorn Chomdaeng of the university’s Institute for Fundamental Study
for encouragement, as well as Ahpisit Ungkitchanukit and Chai-Hok Eab from
Chulalongkorn University.
I am also indebted to many of my former graduate students, who are now
established physicists in their own rights, particularly to Chaiyapoj Muthaporn,
xi
xii Acknowledgements
Nattapong Yongram, Siri Sirininlakul, Tukkamon Vijaktanawudhi (aka Kanchana

Limboonsong), Prasopchai Viriyasrisuwattana, and Seckson Sukhasena, who
through their many questions, several discussions, and collaborations have been
very helpful in my way of analyzing this subject.
Although I have typed the entire manuscripts myself, and drew the figures as
well, Chaiyapoj Muthaporn prepared the LATEX input files. Without his constant
help in LATEX, this work would never have been completed. I applaud him, thank
him and will always remember how helpful he was. My special thanks also go
to Nattapong Yongram for downloading the endless number of papers I needed to
complete the project.
I was fortunate and proud to have been associated with the wonderful editorial
team of Maria Bellantone and Mieke van der Fluit of Springer. I would like to
express my deepest gratitude to them for their excellent guidance, caring, patience,
and hard work in making this project possible and move forward toward its
completion. I have exchanged more emails with Mieke than with anybody else on
the globe. This has led to such an enjoyable association that I will always cherish.
This project would not have been possible without the patience, encouragement,
and understanding of my wife Tuenjai. To my parents, who are both gone, this work
is affectionately dedicated.
Contents
1 Introduction
Donkey Electron, Bare Electron, Electroweak Frog, God
Particle, “Colored” Quarks and Gluons, Asymptotic
Freedom, Beyond Resonances into the Deep Inelastic
Region, Partons, QCD Jets, Confined Quarks, Bekenstein
– Hawking Entropy of a Black Hole, Sparticles, Strings,
Branes, Various Dimensions and even Quanta of Geometry,
AdS/CFT Correspondence and Holographic Principle,
CPT, and Spin & Statistics .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 33
Recommended Reading .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 42
2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 45
2.1 Wigner’s Symmetry Transformations in the Quantum World . . . . . . . 46
2.1.1 Wigner’s Symmetry Transformations .. .. . . . . . . . . . . . . . . . . . . . 47
2.2 Minkowski Spacetime: Common Arena of Elementary Particles . . . 50
2.3 Representations of the Dirac Gamma Matrices;
Majorana Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 54
2.4 Differentiation and Integration with Respect
to Grassmann Variables . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 56
2.5 Fourier Transforms Involving Grassmann Variables .. . . . . . . . . . . . . . . . 60
2.6 Functional Differentiation and Integration; Functional
Fourier Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
2.7 Delta Functionals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 68
Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 70
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
3 Quantum Field Theory Methods of Spin 1=2 . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 73
3.1 Dirac Quantum Field, Propagator
and Energy-Momentum Transfer: Schwinger-Feynman
Boundary Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 74
xiii
xiv Contents
3.2 The Dirac Quantum Field Concept, Particle Content,

and C, P, T Tansformations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 78
3.2.1 Charge Conjugation (C), Parity
Transformation (T), and Time Reversal
(T) of the Dirac Quantum Field . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 83
3.3 Re-Discovering the Positron and Eventual Discovery
of Anti-Matter.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85
3.3.1 h 0C j 0 i for the Dirac Equation . . . . . . .. . . . . . . . . . . . . . . . . . . . 89
3.4 Coulomb Scattering of Relativistic Electrons . . . .. . . . . . . . . . . . . . . . . . . . 93
3.5 Spin & Statistics and the Dirac Quantum Field;
Anti-Commutativity Properties Derived .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 96
3.6 Electromagnetic Current, Gauge Invariance
and h 0C j 0 i with External Electromagnetic Field . . . . . . . . . . . . . . . . . 98
3.7 h 0C j 0 i.e/ in the Presence of a Constant F Field
and Effective Action .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 103
3.8 Pair Creation by a Constant Electric Field . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
3.9 Fermions and Anomalies in Field Theory: Abelian Case . . . . . . . . . . . . 111
3.9.1 Derivation of the Anomaly . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 112
3.9.2 Experimental Verification of the Anomaly:
0 ! Decay.. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 117
3.10 Fermions and Anomalies in Field Theory: Non-Abelian Case . . . . . . 120
Appendix A: Evaluation of L1 2 . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 125
Appendix B: Infinitesimal Variation of the Exponential of a Matrix . . . . . . . 128
Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 129
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 131
4 Fundamental Aspects of Quantum Field Theory .. . . .. . . . . . . . . . . . . . . . . . . . 133
4.1 The Field Concept, Particle Aspect and Wavefunction
Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 135
4.2 PoincarKe Algebra and Particle States . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
4.3 Principle of Stationary Action of Quantum Field
Theory: The Rationale Of . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 146
4.3.1 A Priori Imposed Variations of Dynamical
Variables and Generators of Field Variations:
Field Equations.. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 148
4.3.2 Commutation/Anti-commutation Relations . . . . . . . . . . . . . . . . 155
4.3.3 Generators for Quantum Responses to Field
Variations: Internal Symmetry Groups ... . . . . . . . . . . . . . . . . . . . 156
4.3.4 Variations of Boundary Surfaces . . . . . . . .. . . . . . . . . . . . . . . . . . . . 157
4.4 Inhomogeneous Lorentz Transformations
and Energy-Momentum Tensor . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 159
4.5 Spin and Statistics Connection .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 162
4.5.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 166
4.5.2 The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 166
4.5.3 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 167
Contents xv
4.6 Quantum Dynamical Principle (QDP) of Field Theory .. . . . . . . . . . . . . 168

4.6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 175
4.7 A Panorama of Fields. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 176
4.7.1 Summary of Salient Features of Some Basic Fields. . . . . . . . 177
4.7.2 Spin 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 184
4.7.3 Spin 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 188
4.7.4 Spin 3=2 .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 192
4.7.5 Spin 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 199
4.8 Further Illustrations and Applications of the QDP . . . . . . . . . . . . . . . . . . . 204
4.9 Time-Ordered Products, How to Write Down
Lagrangians and Setting Up the Solution
of Field Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209
4.10 CPT .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 213
Appendix A: Basic Equalities Involving the CPT Operator .. . . . . . . . . . . . . . . . 217
Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 219
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 221
5 Abelian Gauge Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223
5.1 Spin One and the General Vector Field . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 224
5.2 Polarization States of Photons . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 226
5.3 Covariant Formulation of the Propagator .. . . . . . . .. . . . . . . . . . . . . . . . . . . . 228
5.4 Casimir Effect.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 231
5.5 Emission and Detection of Photons .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 235
5.6 Photons in a Medium .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 238
5.7 Quantum Electrodynamics, Covariant Gauges: Setting
Up the Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 241
5.7.1 The Differential Formalism (QDP) and
Solution of QED in Covariant Gauges ... . . . . . . . . . . . . . . . . . . . 241
5.7.2 From the Differential Formalism to the Path
Integral Expression for h 0C j 0 i . . . . . . .. . . . . . . . . . . . . . . . . . . . 246
5.8 Low Order Contributions to lnh 0C j 0 i . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 248
5.9 Basic Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 253
5.9.1 e e ! e e , eC e ! eC e . . . . . . .. . . . . . . . . . . . . . . . . . . . 257
5.9.2 e ! e ; eC e ! and
Polarizations Correlations . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 264
5.9.3 e ! e . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 269
5.10 Modified Propagators .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 271
5.10.1 Electron Self-Energy and Its Interpretation .. . . . . . . . . . . . . . . . 273
5.10.2 Photon Self-Energy and Its Interpretation;
Coulomb Potential . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 278
5.11 Vertex Part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 283
5.11.1 Charge renormalization and External lines . . . . . . . . . . . . . . . . . 287
5.11.2 Anomalous Magnetic Moment of the Electron .. . . . . . . . . . . . 290
xvi Contents
5.12 Radiative Correction to Coulomb Scattering and Soft

Photon Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 290
5.13 Lamb Shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 293
5.14 Coulomb Gauge Formulation . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 304
5.14.1 h 0C j 0 i in the Coulomb Gauge in the
Functional Differential Form . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 304
5.14.2 From the QDP to the Path Integral of
h 0C j 0 i in Coulomb Gauge.. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 308
5.15 Gauge Transformations of the Full Theory .. . . . . .. . . . . . . . . . . . . . . . . . . . 310
5.16 Vertex Function and Ward-Takahashi Identity; Full Propagators . . . 316
5.16.1 Ward-Takahashi Identity .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 316
5.16.2 Equations for the Electron Propagator and the
Vertex Function . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 318
5.16.3 Spectral Representation of the Photon
Propagator, Charge Renormalization, Coulomb
Potential in Full QED, Unordered Products of Currents . . . 323
5.16.4 Integral Equation for the Vacuum Polarization Tensor .. . . . 333
5.17 The Full Renormalized Theory . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 335
5.18 Finiteness of the Renormalized Theory; Renormalized
Vertex Function and Renormalized Propagators .. . . . . . . . . . . . . . . . . . . . 338
5.18.1 Finiteness of the Renormalized Theory .. . . . . . . . . . . . . . . . . . . . 339
5.18.2 The Renormalized Vertex and the
Renormalized Electron Propagator .. . . . .. . . . . . . . . . . . . . . . . . . . 342
5.18.3 The Renormalized Photon Propagator . .. . . . . . . . . . . . . . . . . . . . 343
5.19 Effective Charge and the Renormalization Group . . . . . . . . . . . . . . . . . . . 347
5.19.1 Renormalization Group Analysis . . . . . . .. . . . . . . . . . . . . . . . . . . . 347
5.19.2 The Fine-Structure Effective Coupling at High
Energy Corresponding to the Mass of the
Neutral Z 0 Vector Boson . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 354
5.20 Scalar Boson Electrodynamics, Effective Action
and Spontaneous Symmetry Breaking .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 355
5.20.1 Change of Real Field Variables of Integration
in a Path Integral . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 356
5.20.2 Goldstone Bosons and Spontaneous Symmetry
Breaking .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 357
Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 361
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 364
6 Non-Abelian Gauge Theories . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 369
6.1 Concept of Gauge Fields and Internal Degrees
of Freedom: From Geometry to Dynamics .. . . . . .. . . . . . . . . . . . . . . . . . . . 371
6.1.1 Generators of SU(N) . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 381
6.2 Quantization of Non-Abelian Gauge Fields
in the Coulomb Gauge.. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 382
Contents xvii
6.3 Functional Fourier Transform and Transition

to Covariant Gauges; BRS Transformations
and Renormalization of Gauge Theories . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 388
6.3.1 Functional Fourier Transform and The
Coulomb Gauge .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 388
6.3.2 Trasformation Law from the Coulomb Gauge
to Covariant Gauges in Non-abelian Gauge
Field Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 395
6.3.3 BRS Trasformations and Renormalization
of Non-abelian Gauge Field Theories .. .. . . . . . . . . . . . . . . . . . . . 395
6.4 Quantum Chromodynamics . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 403
6.5 eC e Annihilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 411
6.6 Self-Energies and Vertex Functions in QCD . . . . .. . . . . . . . . . . . . . . . . . . . 413
6.6.1 Fermion Inverse Propagator . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 414
6.6.2 Inverse Gluon Propagator.. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 415
6.6.3 Fermion-Gluon vertex . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 417
6.7 Renormalization Constants, Effective Coupling,
Asymptotic Freedom, and What is Responsible for the Latter? .. . . . 419
6.7.1 What Part of the Dynamics is Responsible for
Asymptotic Freedom? . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 423
6.8 Renormalization Group and QCD Corrections to eC e
Annihilation .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 423
6.9 Deep Inelastic Scattering: Differential Cross Section
and Structure Functions . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 429
6.10 Deep Inelastic Scattering, The Parton Model, Parton
Splitting; QCD Jets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 434
6.10.1 The Parton Model .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 434
6.10.2 Parton Splitting .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 437
6.10.3 QCD Jets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 444
6.11 Deep Inelastic Scattering: QCD Corrections . . . . .. . . . . . . . . . . . . . . . . . . . 447
6.12 From the Schwinger Line-Integral to the Wilson Loop
and How the latter Emerges .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 457
6.13 Lattices and Quark Confinement .. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 463
6.14 The Electroweak Theory I .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 469
6.14.1 Development of the Theory: From the Fermi
Theory to the Electroweak Theory . . . . . .. . . . . . . . . . . . . . . . . . . . 470
6.14.2 Experimental Determination of sin2 W . . . . . . . . . . . . . . . . . . . 481
6.14.3 Masses of the Neutrinos and the “Seesaw
Mechanism” . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 482
6.14.4 Neutrino Oscillations: An Interlude .. . . .. . . . . . . . . . . . . . . . . . . . 485
6.15 Electroweak Theory II: Incorporation of Quarks;
Anomalies and Renormalizability . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 488
6.15.1 Quarks and the Electroweak Theory . . . .. . . . . . . . . . . . . . . . . . . . 489
6.15.2 Anomalies and Renormalizability . . . . . .. . . . . . . . . . . . . . . . . . . . 491
6.16 Grand Unification .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 495
xviii Contents
Problems .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 504
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 506
General Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 513
Appendix I: The Dirac Formalism . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 515
Appendix II: Doing Integrals in Field Theory . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 521
Appendix III: Analytic Continuation in Spacetime Dimension

and Dimensional Regularization . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 529
Appendix IV: Schwinger’s Point Splitting Method of Currents:

Arbitrary Orders.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 533
Appendix V: Renormalization and the Underlying Subtractions . . . . . . . . . 541

Solutions to the Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 551
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 583
Notation and Data
ı Latin indices i; j; k; : : : are generally taken to run over 1,2,3, while the Greek
indices ; ; : : : over 0; 1; 2; 3 in 4D. Variations do occur when there are many
different types of indices to be used, and the meanings should be evident from
the presentations.
ı The Minkowski metric is defined by Œ D diagŒ1; 1; 1; 1 D Œ in
4D.
ı Unless otherwise stated, the fundamental constants „; c are set equal to one.
ı The gamma matrices satisfy the anti-commutation relations f ; g D 2 .
ı The Dirac, the Majorana, and the chiral representations of the matrices are
defined in Appendix I at the end of the book.
ı The charge conjugation matrix is defined by C D i 2 0 .
ı D 0 , u D u 0 , v D v 0 . A Hermitian conjugate of a matrix M is
denoted by M , while its complex conjugate is denoted by M .
ı The step function is denoted by ™.x/ which is equal to 1 for x > 0, and 0 for
x < 0.
ı The symbol " is used in dimensional regularization (see Appendix III).
is
used in defining the boundary condition in the denominator of a propagator
.Q2 C m2 i
/ and should not be confused with " used in dimensional
regularization. We may also use either one when dealing with an infinitesimal
quantity, in general, with
more frequently, and this should be self-evident from
the underlying context.
ı For units and experimental data, see the compilation of the “Particle Data
Group”: Beringer et al. [1] and Olive et al. [2]. The following (some obviously
approximate) numerical values should, however, be noted:
1 MeV D 106 eV
1 GeV D 103 MeV
103 GeV D 1 TeV
1 erg D 107 J
xix
xx Notation and Data
1J D 6:242 109 GeV

c D 2:99792458 1010 cm/s (exact)
„ D 1:055 1034 J s
„c D 197:33 MeV fm
1 fm D 1013 cm
(Masses) Mp D 938:3 MeV=c2 , Mn D 939:6 MeV=c2 ,

MW D 80:4 GeV=c2 , MZ D 91:2 GeV=c2 ,
me D 0:511 MeV=c2 , m D 105:66 MeV=c2 , m D 1777 MeV=c2 .
Mass of e < 2 eV=c2 , Mass of < 0:19 MeV=c2 , Mass of < 18:2 MeV=c2 ,
Mass of the neutral Higgs H 0 125:5 GeV=c2 .
For approximate mass values of some of the quarks taken, see Table 5.1 in
Sect. 5.19.2. For more precise range of values, see Olive et al. [2].
(Newton’s gravitational constant) GN D 6:709 1039 „ c5 / GeV2 .
(Fermi weak p interaction constant) GF D 1:666 105 „3 c3 / GeV2 .
Planck mass p„ c=GN 1:221 1019 GeV=c2 ,
Planck length „ GN =c3 1:616 1033 cm.
Fine structure constant ˛ D 1=137:04 at Q2 D 0, and 1=128 at Q2 MZ2 .
For the weak-mixing angle W , sin2 W 0:232, at Q2 MZ2 .
˛= sin2 W 0:034, at Q2 MZ2 .
Strong coupling constant ˛s 0:119, at Q2 MZ2 .
References
1. Beringer, J., et al. (2012). Particle data group. Physical Review D, 86, 010001.
2. Olive, K. A., et al. (2014). Particle data group. Chinese Physics C, 38, 090001.
Chapter 1
Introduction
Donkey Electron, Bare Electron, Electroweak Frog, God Particle, “Colored” Quarks
and Gluons, Asymptotic Freedom, Beyond Resonances into the Deep Inelastic
Region, Partons, QCD Jets, Confined Quarks, Bekenstein – Hawking Entropy
of a Black Hole, Sparticles, Strings, Branes, Various Dimensions and even Quanta
of Geometry, AdS/CFT Correspondence and Holographic Principle, CPT, and
Spin & Statistics
The major theme of quantum field theory is the development of a unified theory
that may be used to describe nature from microscopic to cosmological distances.
Quantum field theory was born 90 years ago, when quantum theory met relativity,
and has captured the hearts of the brightest theoretical physicists in the world. It is
still going strong. It has gone through various stages, met various obstacles on the
way, and has been struggling to provide us with a coherent description of nature
in spite of the “patchwork” of seemingly different approaches that have appeared
during the last 40 years or so, but still all, with the common goal of unification.
As mentioned in our Preface, Feynman, in his 1958 Cornell, 1959–1960 Cal
Tech, lectures on the quantum field theory of fundamental processes, the first
statement he makes, the very first one, is that the lectures will cover all of physics
[76, p. 1]. One quickly understands what Feynman meant by covering all of physics.
After all, the role of fundamental physics is to describe the basic interactions we
have in nature and quantum field theory is supposed to do just that. Feynman’s
statement is obviously more relevant today than it was then, since the recent
common goal is to provide a unified description of all the fundamental forces in
nature. With this in mind, let us trace the development of this very rich subject from
the past to the present, and see what the theory has been telling us all these years.
When the energy and momentum of a quantum particle are large enough, one is
confronted with the requirement of developing a formalism, as imposed by nature,
which extends quantum theory to the relativistic regime. A relativistic theory, as
a result of the exchange that takes place between energy and matter, allows the
creation of an unlimited number of particles and the number of particles in a given
physical process need not be conserved. An appropriate description of such physical
processes for which a variable number of particles may be created or destroyed, in
the quantum world, is provided by the very rich concept of a quantum field. For
example, photon emissions and absorptions, in a given process, are explained by the
introduction of the electromagnetic quantum field. The theory which emerges from
extending quantum physics to the relativistic regime is called “Relativistic Quantum
© Springer International Publishing Switzerland 2016 1

E.B. Manoukian, Quantum Field Theory I, Graduate Texts in Physics,
DOI 10.1007/978-3-319-30939-2_1
2 1 Introduction
Field Theory” or just “Quantum Field Theory”. Quantum Electrodynamics is an

example of a quantum field theory and is the most precise theory devised by man
when confronted with experiments. The essence of special relativity is that all
inertial frames are completely equivalent in explaining a physical theory as one
inertial frame cannot be distinguished from another. This invariance property of
physical theories in all inertial frames, as required by special relativity, as well as by
the many symmetries one may impose on such theories, are readily implemented in
the theory of quantum fields. The implementation of symmetries and describing
their roles in the explanation of observed phenomena has played a key role in
elementary particle physics.
Of course it took years before the appropriate language of quantum field theory,
described concisely above, by marrying quantum theory and relativity, was spelled
out and applied consistently to physical processes in the quantum world in the
relativistic regime. An appropriate place to start in history is when Dirac [47–49]
developed his relativistic equation of spin 1/ 2, from which one learns quite a bit
about the subsequent development of the subject as a multi-particle theory. We will
then step back a year or two, and then move again forward in time to connect
the dots between the various stages of the underlying exciting developments. His
relativistic equation, which incorporated the spin of the electron, predicted the
existence of negative energy states with negative mass, with energies going down
to 1, implying the instability of the corresponding systems. For example, an
electron in the ground-state energy of an atom would spontaneously decay to such
lower and lower negative energy states emitting radiation of arbitrary large energies
leading eventually to the collapse of the atom with the release of an infinite amount
of energy. Historically, a relativistic equation for spin 0, was developed earlier by
Klein and Gordon in 1926,1 referred to as the Klein-Gordon equation, which also
shared this problem, but unlike Dirac’s theory it led to negative probabilities as well.
Dirac being aware of the negative probabilities encountered in the theory of the latter
authors, was able to remedy this problem in his equation. To resolve the dilemma
of negative energies, Dirac, in 1930,2 assumed that a priori all the negative energy
states are filled with electrons in accord to the Pauli exclusion principle, giving rise
to the so-called Dirac sea or the Dirac vacuum, so that no transitions to such states
are possible, thus ensuring the stability of the atom.
The consequences of the assumption made by Dirac above were many. A negative
energy electron in the Dirac sea, may absorb radiation of sufficient energy so as
to overcome an energy gap arising from the level mc2 to Cmc2 , where m is
the mass of an electron, thus making such a negative energy electron jump to a
positive energy state, leaving behind a surplus of positive energy and a surplus of
1
Klein [128] and Gordon [101].
2
Dirac [50, 51].
1 Introduction 3
positive charge Cjej relative to the Dirac sea. This has led Dirac eventually,3 in
1931 [52], to interpret the “hole” left behind by the transition of the negative energy
electron to a positive energy state, as a particle that has the same mass as the electron
but of opposite charge. It is interesting to note that George Gamow referred4 to
Dirac’s predicted particle as a “donkey electron”, because it would move in the
opposite direction of an appropriate applied force. The physics community found
it difficult to accept Dirac’s prediction until Anderson5 discovered this particle (the
positron eC ), who apparently was not aware of Dirac’s prediction at the time of
the discovery.6 With the positron now identified, the above argument just given has
provided an explanation of the so-called pair production ! eC e by a photon
( in the vicinity of a nucleus).7 Conversely, if a “hole” is created in the vacuum, then
an electron may make a transition to such a state releasing radiation giving rise to the
phenomenon of pair annihilation. A Pair created, as described above, in the vicinity
of a positively charged nucleus, would lead to a partial screening of the charge of
the nucleus as the electron within the pair would be attracted by the nucleus and
the positively charged one would be repelled. Accordingly, an electron, in the atom,
at sufficiently large distances from the nucleus would then see a smaller charge on
the nucleus than an electron nearby (such as one in an s-state). This leads to the
concept of vacuum polarization, and also to the concept of charge renormalization
as a result of the partial charge screening mentioned above.
The Dirac equation is Lorentz covariant, that is, it has the same form in every
inertial frame with its variables being simply relabeled reflecting the variables used
in the new inertial frame. It predicted, approximately, the gyromagnetic ratio g D
2 of the electron, the fine-structure of the atom, and eventually anti-matter was
discovered such as antiprotons.8 It was thus tremendously successful. Apparently,9
Dirac himself remarked in one of his talks that his equation was more intelligent
than its author.10
Thus the synthesis of relativity and quantum physics, led to the discovery of the
antiparticle. The Dirac equation which was initially considered to describe a single
particle necessarily led to a multi-particle theory, and a single particle description
in the relativistic regime turned out to be not complete. A formalism which would
naturally describe creation and annihilation of particles and take into account this
3
Dirac [50, 51] assumed that the particle is the proton as the positron was not discovered yet at that
time. Apart from the large mass difference between the proton and the electron, there were other
inconsistencies with such an assumption.
4
Weisskopf [242].
5
Anderson [5, 6].
6
Weisskopf [242].
7
The presence of the nucleus is to conserve energy and momentum.
8
Chamberlain et al. [30].
9
Weisskopf [242].
10
For a systematic treatment of the intricacies of Dirac’s theory and of the quantum description of
relativistic particles, in general, see Manoukian [151], Chapter 16.
4 1 Introduction
multi-particle aspect became necessary. The so-called “hole” theory although it gave
insight into the nature of fundamental processes involving quantum particles in the
relativistic regime, and concepts such as vacuum polarization, turned out to be also
not complete. For example, in the “hole” theory, the number of electrons minus the
number of positrons, created is conserved by the simultaneous creation of a “hole”
for every electron ejected from the Dirac Sea. In nature, there are processes, where
just an electron or just a positron is created while conserving charge of course.
Examples of such processes are ˇ decay: n ! p C e C Q e , muon decay:
! e C Q e C , and ˇ C decay: p ! n C eC C e , for a bound proton
in a nucleus for the latter process. Finally, Dirac’s argument of a sea of negatively
charged bosons did not work with the Klein-Gordon equation because of the very
nature of the Bose statistics of the particles. A new description to meet all of the
above challenges including the creation and annihilation of particles, mentioned
above, was necessary.
After the conceptual framework of quantum mechanics was developed, Born,
Heisenberg, and Jordan in 1926 [26], applied quantum mechanical methods to
the electromagnetic field, now, giving rise to a system with an infinite degrees of
freedom, and described as a set of independent harmonic oscillators of various
frequencies. Then Dirac in 1927 [46], prior to the development of his relativistic spin
1/2 equation, also extended quantum mechanical methods to the electromagnetic
field now with the latter field treated as an operator, and provided a theoretical
description of how photons emerge in the quantization of the electromagnetic field.
This paper is considered to mark the birthdate of “Quantum Electrodynamics”, a
name coined by Dirac himself, and provided a prototype for the introduction of
field operators for other particles with spin, such as for spin 1/2, where in the latter
case commutators in the theory are replaced by anti-commutators [125, 126] for the
fermion field.
The first comprehensive treatment of a general quantum field theory, involving
Lagrangians, as in modern treatments, was given by Heisenberg and Pauli in 1929,
1930 [116, 117], where canonical quantization procedures were applied directly
to the fields themselves. A classic review of the state of affairs of quantum
electrodynamics in 1932 [68] was given by Fermi. The problem of negative
energy solutions was resolved and its equivalence to the Dirac “hole” theory was
demonstrated by Fock in 1933 [83], and Furry and Oppenheimer in 1934 [90], where
the (Dirac) field operator and its adjoint were expanded in terms of appropriate
creation and annihilation operators for the electron and positron, thus providing
a unified description for the particle and its antiparticle. The method had a direct
generalization to bosons. The old “hole” theory became unnecessary and obsolete.11
The problem of negative energy solutions was also resolved for spin 0 bosons by
11
As a young post-doctoral fellow, I remember attending Schwinger’s lecture tracing the Develop-
ment of Quantum Electrodynamics in “The Physicist’s Conception of Nature” [202], making the
statement, regarding the “hole” theory, that it is now best regarded as an historical curiosity, and
forgotten.
1 Introduction 5
similar methods by Pauli and Weisskopf in 1934 [170]. The fields thus introduced
from these endeavors have become operators for creation and annihilation of
particles and antiparticles, rather than probability amplitudes.12
The explanation that interactions are generated by the exchange of quanta was
clear in the classic work of Bethe and Fermi in 1932 [18]. For example, charged
particles, as sources of the electromagnetic field, influence other charged particles
via these electromagnetic fields. Fields as operators of creation and destruction of
particles, and the association of particles with forces is a natural consequence of field
theory. The same idea was used by Yukawa in 1935 [249], to infer that a massive
scalar particle is exchanged in describing the strong interaction (as understood in
those days), with the particle necessarily being massive to account for the short
range nature of the strong force unlike the electromagnetic one which is involved
with the massless photon describing an interaction of infinite range. The mass
of the particle may be estimated from the expression „=Rc, obtained formally
from the uncertainty principle, where R denotes the size of the proton, i.e., R D
1 fm D 1013 cm. In natural units, i.e., for „ D 1, c D 1, 1 fm 1=.200 MeV/.
This gives 200 MeV. Such a particle (the pion) was subsequently discovered
by the C. F. Powell group in 1947 [136].
As early as 1930s, infinities appeared in explicit computations in quantum
electrodynamics by Oppenheimer [168], working within an atom, by Waller
[233, 234], and by Weisskopf [239]. The nature of these divergences, arising
in these computations, came from integrations that one had to carry out over
energies of photons exchanged in describing the interaction of the combined
system of electrons and the electromagnetic field to arbitrary high-energies. By
formally restricting the energies of photons exchanged, as just described, to be
less than, say, , Weisskopf, in his calculations, has shown [239, 240], within the
full quantum electrodynamics, that the divergences encountered in the self-energy
acquired by the electron from its interaction with the electromagnetic field is of
the logarithmic13 type ln.=mc2 /, improving the preliminary calculations done
earlier, particularly, by Waller, mentioned above. That such divergences, referred
to as “ultraviolet divergence”,14 are encountered in quantum field theory should
12
It is important to note, however, that the matrix elements of these field operators between particle
states and the vacuum naturally lead to amplitudes of particles creation by the fields and to the
concept of wavefunction renormalization (see Sect. 4.1) independently of any perturbation theories.
13
The corresponding expression occurs with higher powers of the logarithm for higher orders in
the fine-structure constant e2 =4„c.
14
That is, divergences arising from the high-energy behavior of a theory. Another type of
divergence, of different nature occurring in the low energy region, referred to as the “infrared
catastrophe”, was encountered in the evaluation of the probability that a photon be emitted in a
collision of a charged particle. In computations of the scattering of charged particles, due to the
zero mass nature of photons, their simultaneous emissions in arbitrary, actually infinite, in number
must necessary be taken into account for a complete treatment. By doing so finite expressions for
the probabilities in question were obtained [22].
6 1 Introduction
(a) γ (b)
e+
γ γ
e− e− e−
e−
Fig. 1.1 Processes leading to an electron self-energy correction, and vacuum polarization, respec-
tively
be of no surprise as one is assuming that our theories are valid up to infinite

energies!15
The 2S1=2 , 2P1=2 states of the Hydrogen atom are degenerate in Dirac’s theory.
In 1947 [134], Lamb and Retherford, however, were able to measure the energy
difference between these states, referred to as the “Lamb Shift”, using then
newly developed microwave methods with great accuracy. Bethe [17] then made
a successful attempt to compute this energy difference by setting an upper limit for
the energy of photon exchanged in describing the electromagnetic interaction of the
order of the rest energy of the electron m c2 , above which relativistic effects take
place, relying on the assumption that the electron in the atom is non-relativistic,
and, in the process, took into consideration of the mass shift16 of the electron.
He obtained a shift of the order of 1000 megacycles which was in pretty good
agreement with the Lamb-Retherford experiment.
Very accurate computations were then made, within the full relativistic quantum
electrodynamics, and positron theory. Notably, Schwinger17 in 1948 [192], com-
puted the magnetic moment of the electron modifying the gyromagnetic ratio, g D 2
in the Dirac theory, to 2 .1 C ˛=2/, to lowest order in the fine-structure constant.
The computation of the Lamb-Shift was also carried out in a precise manner by Kroll
and Lamb [133], and, for example, by French and Weisskopf [85], and Fukuda et al.
[88, 89].
State of affairs changed quite a bit. It became clear that an electron is accompa-
nied by an electromagnetic field which in turn tends to alter the nature of the electron
that one was initially aiming to describe. The electron e , being a charged particle,
produces an electromagnetic field . /. This field, in turn, interacts back with the
electron as shown below in Fig. 1.1a. Similarly, the electromagnetic field ( ) may
lead to the creation of an electron-positron pair eC e , which in turn annihilate
each other re-producing an electromagnetic field, a process referred to as vacuum-
polarization, shown in part (b). Because of these processes, the parameters initially
15
See also the discussion in Sect. 5.19.
16
See also the important contribution to this by Kramers [130]. This reference also includes
contributions of his earlier work.
17
See also Appendix B of Schwinger [193].
1 Introduction 7
e− e−
E= p 2 + m2 , p E = p 2 + m2 , p
Fig. 1.2 As a result of the self-energy correction in Fig. 1.1a, where an electron emits and re-
absorbs a photon, the mass parameter m0 , one initially starts with, does not represent the physical
mass of the electron determined in the lab. Here this is emphasized by the energy dependence on
the physical mass m of an electron in a scattering process. The dashed lines represent additional
particles participating in the process
appearing in the theory, such as mass, say, m0 , vis-à-vis Fig. 1.1a, and the electron
charge, say, e0 , vis-à-vis Fig. 1.1b, that were associated with the electron one starts
with, are not the parameters actually measured in the lab. For example, the energy
of a scattered
q electron of momentum p, in a collision process, turned up to be not
p
2
equal to p C m0 but rather to p C m2 , self-consistently,18 with m identified
2 2
with the actual, i.e., tabulated, mass of the electron, and m ¤ m0 , with a scattering
process shown in Fig. 1.2, where the dashed lines represent other particles (such as
; e ; eC ), where the total charge as well as the total energy and momentum are
conserved in the scattering process.
Similarly, the potential energy between two widely separated electrons, by a
distance r, turned up to be not e20 =4r but rather e2 =4r, with e2 ¤ e20 , where
e is identified with the charge, i.e., the tabulated charge, of the electron. As we
will see later, the physical parameters are related to the initial ones by scaling
factors, referred to as mass and charge renormalization constants, respectively. An
electron parametrized by the couple .m0 ; e0 /, is referred to as a bare electron
as it corresponds to measurements of its properties by going down to “zero”
distances all the way into the “core” of the electron – a process that is unattainable
experimentally. On the other hand, the physical parameters .m; e/, correspond to
measurements made on the electron from sufficiently large distances.
One thus, in turn, may generate parameters, corresponding to a wide spectrum of
scales running from the very small to the very large. Here one already notices that in
quantum field theory, one encounters so-called effective parameters which are func-
tions of different scales (or energies). Functions of these effective parameters turn
out to satisfy invariance properties under scale transformations, thus introducing a
concept referred to as the renormalization group. Clearly, due to the screening effect
via vacuum polarization of eC e pairs creation, as discussed earlier, the magnitude
of the physical charge is smaller than the magnitude of the bare charge.
18
An arbitrary number of photons of vanishingly small energies are understood to be attached
to the external electron lines, as discussed in Footnote 14 when dealing with infrared divergence
problems.
8 1 Introduction
A process was, in turn, then carried out, referred to as “renormalization”,

to eliminate the initial parameters in the theory in favor of physically observed
ones. This procedure related the theory at very small distances to the theory at
sufficiently large distances at which particles emerge on their way to detectors as
it happens in actual experiments. All the difficulties associated with the ultraviolet
divergences in quantum electrodynamics were isolated in renormalization constants,
such as the ones discussed above, and one was then able to eliminate them in
carrying out physical applications giving rise to completely finite results. This
basic idea of the renormalization procedure was clearly spelled out in the work
of Schwinger, Feynman, and Tomonaga.19 The renormalization group,20 mentioned
above, describes the connection of renormalization to scale transformations, and
relates, in general, the underlying physics at different energy scales.
In classic papers, Dyson [59, 60] has shown not only the equivalence of the
Schwinger, Feynman, and Tomonaga approaches,21 and the finiteness of the so-
called renormalized quantum electrodynamics, but also developed a formalism for
computations that may be readily applied to other interacting quantum field theories.
Theories that are consistently finite when all the different parameters appearing
initially in the theory are eliminated in favor of the physically observed ones, which
are finite in number, are said to be renormalizable. Dyson’s work, had set up:
renormalizability as a condition for generating field theory interactions.
In units of „ D 1, c D 1, ŒMass D ŒLength1 . Roughly speaking, in a

renormalizable theory, no coupling constants can have the dimensions of negative
powers of mass. (Because of dimensional reasons, we note, in particular, that one
cannot have too many derivatives of the fields, describing interactions, as every
derivative necessitates involving a coupling of dimensionality reduced by one in
units of mass.)
The photon as the agent for transmitting the interaction between charged parti-
cles, is described by a vector – the vector potential. In quantum electrodynamics,
as a theory of the interaction of photons and electrons, for example, the photon is
coupled locally to the electromagnetic current. The latter is also a vector, and the
interaction is described by their scalar product (in Minkowski space) ensuring the
relativistic invariance of the underlying theory. To lowest order in the charge e of
the electron e, this coupling may be represented by the diagram Fig. 1.3a. On the
other hand, for a spin 0 charged boson ', say, one encounters two such diagrams,
each shown to lowest order in the charge e in Fig. 1.3b, where we note that in the
second diagram in the latter part, two photons emerge locally from the same point.
19
This is well described in their Nobel lectures: Schwinger [201], Feynman [75], Tomonaga [225],
as well as in the collection of papers in Schwinger [198, 201].
20
Stueckelberg and Peterman [209], Gell-Mann and Low [93], Bogoliubov and Shirkov [25],
Ovsyannikov [169], Callan [28, 29], Symanzik [213–215], Weinberg [237], and ’t Hooft [218].
21
The best sources for these approaches are their Nobel Lectures: Schwinger [201], Feynman [75],
Tomonaga [225], as well as Schwinger [198].
1 Introduction 9
(a) (b)
ϕ ϕ
e
γ
γ γ
e e e2
γ
e ϕ ϕ
Fig. 1.3 Local couplings for photon emission by an electron, and by a spin 0 charged particle
described by the field ', respectively
Quantum Electrodynamics, was not only the theory of interest. There was also
the weak interaction. The preliminary theory of weak interaction dates back to Fermi
[69, 70]. Based on weak processes such as ˇ decay: n ! p C e C Q e ,
he postulated that the weak interactions may be described by local four-point
interactions involving a universal coupling parameter GF . The four particles of the
process just mentioned, interact locally at a point with a zero range interaction.
The Fermi theory was in good agreement in predicting the energy distribution of
the electron. For dimensional reasons, however, the dimensions of the coupling
constant GF involved in the theory has the dimensions of ŒMass2 , giving rise to a
non-renormalizable theory.22 In analogy to quantum electrodynamics, the situation
with this type of interaction may be somehow improved by introducing, in the
process, a vector Boson23 W which mediates an interaction24 between the two
pairs (so-called currents), .n; p/ and .e ; Q e /, with both necessarily described by
entities carrying (Lorentz) vector indices, to ensure the invariance of the underlying
description. Moreover, in units of „ D 1, c D 1, a dimensionless coupling g is
introduced. The Fermi interaction and its modification are shown, respectively in
parts Fig. 1.4a, b.
In order that the process in diagram given in part Fig. 1.4b, be consistent with
the “short-range” nature of the Fermi interaction, described by the diagram on the
left, the vector particle W must not only be massive but its mass, MW must be
quite large. This is because the propagator of a massive vector particle of mass
MW , which mediates an interaction between two spacetime points x and x 0 , as we
will discuss below, behaves like ı .4/ .x x 0 /=M 2W for a large mass, signifying
22
It is interesting to point out as one goes to higher and higher orders in the Fermi coupling
constant GF , the divergences increase (Sect. 6.14) without any bound and the theory becomes
uncontrollable.
23
A quantum relativistic treatment of a problem, implies that a theory involving the W particle,
must also include its antiparticle W C , having the same mass as of W .
24
Such a suggestion was made, e.g., by Klein [129].
10 1 Introduction
(a) p
(b) p
n GF n g
(x) (x)
(x )
e−
W− g
e−
ν̃e
ν̃e
Fig. 1.4 (a) The old Fermi theory with a coupling GF is replaced by one in (b) where the
interaction is mediated by a vector boson with a dimensionless coupling g
necessarily a vanishingly small range of the interaction.25 Upon comparison of both

diagrams, one may then infer that
g2
GF : (1.1)
M 2W
Evidently, the Fourier transform of the propagator in the energy-momentum descrip-

tion, at energies much less than MW is, due to the ı .4/ .xx 0 / function given above,
simply =M 2W , and (1.1) may be obtained from a low-energy limit.
With some minimal effort, the reader will understand better the above two limits
and some of the difficulties encountered with a massive vector boson, in general, if,
at this stage, we write down explicitly its propagator between two spacetime points
x, x 0 in describing an interaction carried by the exchange of such a particle which
is denoted by26 :
Z
0 .dk/ i k .x x 0 /
4C .x x / D e 4C .k/; .dk/ D dk0 dk1 dk1 dk3 ; (1.2)
.2/4

1 k k
4C .k/ D 2 C 2 ; (1.3)
.k C M 2W i 0/ MW
25
Here is the Minkowski metric.
26
This expression will be derived in Sect. 4.7. For a so-called virtual particle k2 D k2 .k0 /2 ¤
M 2W . The i 0 in the denominator in (1.3) just specifies the boundary condition on how the k0
integration is to be carried out. These things will be discussed in detail later on and are not needed
here.
1 Introduction 11
where k0 is its energy, and k D .k1 ; k2 ; k3 / its momentum. Formally for M 2W !

1, 4C .k/ ! =M 2W , leading from (1.2) to
Z
0 .dk/ ik .x x 0 / 4
4C .x x / ! e D ı .x x 0 /; (1.4)
M 2W .2/4 M 2W
signalling, in a limiting sense, a short range interaction for a heavy-mass particle.

On the other hand, for jk j MW for each component, one has

4C .k/ ; (1.5)
M 2W
in the energy-momentum description.

Although the introduction of the intermediate boson W improves somehow the
divergence problem, it is still problematic. The reason is not difficult to understand.
In the energy-momentum description, the propagator of a massive vector particle,
as given in (1.3), has the following behavior at high energies and momenta
1 k k
4C .k/ ! ; (1.6)
k2 M 2W
providing no damping in such a limit. Moreover, as one goes to higher orders in

perturbation theory the number of integration variables, over energy and momenta
arising in the theory, increase, and the divergences in turn increase without bound
and the theory becomes uncontrollable.27 On the other hand, an inherited property
of quantum electrodynamics is gauge symmetry due to the masslessness of the
photon. In the present context of ultraviolet divergences, the photon propagator has
a very welcome vanishing property at high energies. This gauge symmetry as well
as the related massless aspect of the photon, which are key ingredients in the self
consistency of quantum electrodynamics, turned out to provide a guiding principle
in developing the so-called electroweak theory.
In 1956 [138], an important observation was made by Lee and Yang that parity
P is violated in the weak interactions. Here we recall that, given a process, its parity
transformed (mirror) version, is obtained by reversing the directions of the space
variables.28 This has led Gershtein and Zel’dovich [95], Feynman and Gell-Mann
[77], Sudarshan and Marshak [210], and Sakurai [178], to express the currents
27
The damping provided by the propagators of a massless vector particle, a spin 1/2 particle, and
a spin 0 particle, for example, in the ultraviolet region vanish like 1/energy2 , 1/energy, 1/energy2 ,
respectively.
28
It was later observed that the product of charge conjugation, where a particle is replaced by its
antiparticle, and parity transformation “CP”, is also not conserved in a decay mode of K mesons
at a small level [38, 39, 82]. As the product “CPT”, of charge conjugation, parity transformation,
and time reversal “T”, is believed to be conserved, the violation of time reversal also follows. For
a test of such a violation see CPLEAR/Collaboration [36].
12 1 Introduction
constructed out of the pairs of fields: .n; p/, .e ; Q e /; : : : in the Fermi theory to
reflect, in particular, this property dictated by nature. The various currents were
eventually expressed and conveniently parametrized in such a way that the theory
was described by the universal coupling parameter GF . The construction of such
fundamental currents together with idea of intermediate vector bosons exchanges to
describe the weak interaction led eventually to its modern version.
Quantum Electrodynamics may be considered to arise from local gauge invari-
ance in which the electron field is subjected to a local phase transformation ei#.x/ .
The underlying group of transformations is denoted by U.1/ involving simply
the identity as the single generator of transformations with which the photon is
associated as the single gauge field. In 1954 [247], Yang and Mills, and Shaw
in 1955 [203], generalized the just mentioned abelian gauge group of phase
transformations, encountered in quantum electrodynamics, to a non-abelian29 gauge
theory, described by the group SU.2/,30 and turned out to be a key ingredient in the
development of the modern theory of weak interactions. This necessarily required
the introduction, in addition to the charged bosons W ˙ , a neutral one. What
distinguishes a non-abelian gauge theory from an abelian one, is that in the former
theory, direct interactions occur between gauge fields, carrying specific quantum
numbers, unlike in the latter, as the gauge field – the photon – being uncharged.
As early as 1956, Schwinger believed that the weak and electromagnetic
interactions should be combined into a gauge theory [97, 159, 199]. Here we
may pose to note that both in electrodynamics and in the modified Fermi theory,
interactions are mediated by vector particles. They are both described by universal
dimensionless coupling constants e, and by g (see (1.1)) in the intermediate vector
boson description, respectively. In a unified description of electromagnetism and the
weak interaction, one expects these couplings to be comparable, i.e.,
1
g2 e2 D 4 ˛; where ˛ ; GF 1:166 105=.GeV/2 : (1.7)
137
in units „ D 1; c D 1. From (1.1), we may then estimate the mass of the W bosons
to be
s
4 ˛
MW 90 GeV=c2 ; (1.8)
GF
29
Non-abelian refers to the fact that the generators do not commute. In contrast a U.1/ gauge
theory, such as quantum electrodynamics, is an abelian one.
30
SU.2/ consists of 2 2 unitary matrices of determinant one. (The letter S in the group stands
for the special property of determinant one.) It involves three generators, with which are associated
three gauge fields. This will be studied in detail in Sect. 6.1.
1 Introduction 13
re-inserting the constant c for convenience, in good agreement with the observed
mass. We may also estimate the range of the weak interaction to be
„c
RW 2:2 1016 cm: (1.9)
MW c2
Glashow, a former graduate student of Schwinger, eventually realized [96]31 the

important fact that the larger group SU.2/ U.1/, is needed to include also
electrodynamics within the context of a Yang-Mills-Shaw theory. A major problem
remained: the local gauge symmetry required that the gauge fields associated with
the group must, a priori, be massless in the initial formulation of the theory.
The problem of the masslessness of the vector bosons was solved by Weinberg
[236, 238] and Salam [182, 183],32 by making use of a process,33 referred to as
spontaneous symmetry breaking, where a scalar field interacting with the vector
bosons, whose expectation value in the vacuum state is non zero, leads to the
generation of masses to them.34 This is referred to as the Higgs35 mechanism, in
which the group SU.2/ U.1/ is spontaneously broken to the group U.1/ with
the latter associated with the photon, and, in the process, the other bosons, called
W ˙ ; Z 0 , acquiring masses, thanks to the Higgs boson, and renormalizability may
be achieved. The latter particle has been also called the “God Particle”.36 The mere
existence of a neutral vector boson Z 0 implies the existence of a weak interaction
component in the theory without a charge transfer, the so-called neutral currents. A
typical process involving the neutral Z 0 boson exchange is in Q C e ! Q C e
shown in Fig. 1.5 not involving the muon itself. Neutral currents37 have been
observed,38 and all the vector bosons have been observed39 as well. It turned out
that the theory with spontaneous symmetry did not spoil the renormalizability of
31
See also Salam and Ward [189].
32
See also Salam and Ward [187–189] and Salam [181].
33
Some key papers showing how spontaneous symmetry breaking using spin 0 field may generate
masses for vector bosons are: Englert and Brout [63], Englert et al. [64], Guralnik et al. [110], and
Kibble [127].
34
Apparently the Legendary Victor Weisskopf was not impressed by this way of generating masses.
In his CERN publication [241], on page 7, 11th line from below, he says that this is an awkward
way to explain masses and that he believes that Nature should be more inventive, but experiments
may prove him wrong.
35
Higgs [119–121]. This work followed earlier work of Schwinger [200], where he shows, by
the exactly solvable quantum electrodynamics in two dimensions, that gauge invariance does not
prevent the gauge field to acquire mass dynamically, as well as of the subsequent work of Anderson
[7] in condensed matter physics.
36
This name was given by Lederman and Teresi [137].
37
Neutral current couplings also appear in Bludman’s [23] pioneering work on an SU.2/ gauge
theory of weak interactions but did not include electromagnetic interactions.
38
Hasert et al. [112, 113] and Benvenuti et al. [15].
39
See, e.g., C. Rubbia’s Nobel Lecture [176].
14 1 Introduction
Fig. 1.5 A process involving

the exchange of the neutral ν̃μ ν̃μ
vector boson Z 0
Z0
e− e−
the resulting theory with massive vector bosons. Proofs of renormalizability were
given by ’t Hooft [216, 217].40 It seems that Sydney Coleman used to say that
’t Hooft’s proof has turned the Weinberg-Salam frog into an enchanted prince.41
The “Electroweak Theory” turned up to be quite a successful theory.42
Another interaction which was also developed in the “image” of quantum
electrodynamics was quantum chromodynamics, as a theory of strong interactions
based, however, on the non-abelian gauge symmetry group SU.3/. Here one notes
that a typical way to probe the internal structure of the proton is through electron-
proton scattering. The composite nature of the proton, as having an underlying
structure, becomes evident when one compares the differential cross sections for
elastic electron-proton scattering with the proton described as having a finite
extension to the one described as a point-like particle. With a one photon exchange
description, the form factors in the differential cross section are seen to vanish
rapidly for large momentum transfer (squared) Q2 of the photon imparted to the
proton. As Q2 is increased further one reaches the so-called resonance region,43
beyond which, one moves into a deep inelastic region, where experimentally the
reaction changes “character”, and the corresponding structure functions of the
differential cross section have approximate scaling properties (Sect. 6.9), instead of
the vanishing properties encountered with elastic form factors, the process of which
is depicted in Fig. 1.6. Such properties indicate the presence of approximately free
point-like structures within the proton referred as partons, which consist of quarks,
gluons together with those emitted44 from their scattering reactions. This led to the
development of the so-called parton model,45 as a first approximation, in which
these point-like particles within the proton are free and the virtual photon interacts
40
See also ’t Hooft and Veltman [221], Lee and Zinn-Justin [139–142], and Becchi et al. [13].
41
See Salam [183], p. 529.
42
The basic idea of the renormalizability of the theory rests on the fact that renormalizability may
be established for the theory with completely massless vector bosons, as in QED, one may then
invoke gauge symmetry to infer that the theory is also renormalizable for massive vector bosons
via spontaneous symmetry breaking.
43
A typical resonance is C , of mass 1.232 GeV, consisting of a proton p and a 0 meson.
44
See, e.g., Fig. 6.7c.
45
Feynman [73, 74] and Bjorken and Pachos [21].
1 Introduction 15
} Anythin
g
nucleon
lepton
lepton
Fig. 1.6 In the process, “Anything” denotes anything that may be created in the process consistent
with the underlying conservation laws. The wavy line denotes a neutral particle (, Z 0 ,. . . ) of large
momentum transfer
(a) (b)
Fig. 1.7 (a) If interactions between quarks may be represented, as an analogy, by people holding
hands, then pulling one person would drag everybody else along. In the parton model, the situation
is represented as in part (b) rather than in part (a)
with each of its charged constituents independently,46 instead of interacting with the
proton as a whole.
The non-abelian gauge symmetry group SU.3/, is needed to accommodate
quarks and gluons, involving eight generators with which the gluons are associated.
Here, in particular, a quantum number referred to as “color” (three of them)47 is
assigned to the quarks. One of the many reasons for this is that the spin 3/2 particle
CC , which is described in terms of three identical quarks (the so-called u quarks)
as a low lying state with no orbital angular momentum between the quarks, behaves
as a symmetric state under the exchange of two of its quarks and would violate the
Spin & Statistics connection without this additional quantum number. The color
degrees of freedom are not observed in the hadronic states themselves and the latter
behave as scalars, that is they are color singlets, under SU.3/ transformations.
As the group SU.3/ involves “color” transformations within each quark flavor, it
may be denoted by SU.3/color or just by SU.3/C . The gluons also carry “color”
and direct gluon-gluon interactions then necessarily occur, unlike the situation with
46
For an analogy to this, see part (b) of Fig. 1.7b.
47
Greenberg [104], Han and Nambu [111], Nambu [163], Greenberg and Zwanziger [105], Gell-
Mann [92], and Fritzsch and Gell-Mann [87].
16 1 Introduction
photons in quantum electrodynamics since photons do not carry a charge. These

gluon-gluon interactions turn out to have an anti-screening effect on a source field
which dominate over the screening effect of quark/antiquark interactions leading to
the interesting fact that the effective coupling of quark interactions becomes smaller
at high-energies, and eventually vanish48 – a phenomenon referred to as asymptotic
freedom. This has far reaching consequences as it allows one to develop perturbation
theory at high energies, in the effective coupling, and carry out various applications
which were not possible before the development of the theory, and is consistent
with deep-inelastic experiments of leptons with nucleons, with the latter described
by point-like objects which, at high energies, scatter almost like free particles,49 as
mentioned above, the process of which is shown Fig. 1.6.
A particular experiment which indirectly supports the idea of quarks having
spin 1/2 stems from e eC annihilation to a quark-antiquark, in the center of mass
system. One would naïvely expect that the quark and the antiquark will emerge from
the process moving in opposite directions50 on their ways to detectors and will be
observed. This is not, however, what happens and instead two narrow jets of hadrons
emerge, moving back-yo-back, with the net jet-axis angular distribution consistent
with a spin 1/ 2 character of the quark/antiquark parents sources.
The electroweak theory and quantum chromodynamics together constitute the so-
called standard model51 with underlying gauge symmetry group SU.3/ SU.2/
U.1/.
The effective coupling in quantum chromodynamics is expected to become larger
at large distances increasing with no bound providing a strong confining force
of quarks and gluons restricting them within hadrons – a phenomenon that is
sometimes referred to as infrared slavery.52
From our discussion of quantum electrodynamics, we recall that the effective
coupling of a U.1/ gauge theory, as an abelian gauge theory, increases with
energies. On the other hand, the asymptotic free nature of non-abelian gauge
theories, imply that the effective couplings associated with the groups SU.3/,
SU.2/, decrease with energy. Due to the smallness of the U.1/ coupling in
comparison to the other two at the present low energies, this gives one the hope that
at sufficient high energies these three couplings merge together and the underlying
theory would be described by one single force. A theory which attempts to unify
the electroweak and the strong interactions is called a grand unified theory. Such
theories have been developed53 and the couplings seem to run to merge roughly
48
This was discovered by Gross and Wilczek [106] and Politzer [172]. See also Vanyashin and
Terentyev [229] for preliminary work on vector bosons.
49
Chromodynamics means Colordynamics, and the name Quantum Chromodynamics is attributed
to Gell-Mann, see, e.g., Marciano and Pagels [158].
50
See, e.g., Fig. 6.7d, and Sect. 6.10.3
51
The name “Standard Model” is usually attributed to Weinberg.
52
Unfortunately, no complete proof of this is available.
53
See, e.g., Georgi et al. [94] for pioneering work. See also Beringer et al. [16] and Olive et al.
[167].
1 Introduction 17
somewhere around 1015 –1016 GeV. This, in turn, gives the hope of the development
of a more fundamental theory in which gravitation, p which should be effective at
energy scale of the order of Planck energy scale „c=GN ' 1019 GeV,54 or
even at a lower energy scale, where GN denotes Newton’s gravitational constant,
is unified with the electroweak and strong forces. If the standard model is the
low energy of such a fundamental theory, then the basic question arises as to
what amounts for the enormous difference between the energy scale of such
a fundamental theory (1016 –1019 GeV) and the defining energy scale of the
standard model (300 GeV)? This has been termed as the hierarchy problem which
will be discussed again later. We will see in Vol. II, in particular, that the above
mentioned couplings seem to be unified at a higher energy scale of the order
1016 GeV, when supersymmetry is taken into account, getting it closer to the energy
scale at which gravitation may play an important role.
One may generalize the symmetry group of the standard model, and consider
transformations which include transformations between quarks and leptons, leading
to a larger group such as, for example, to the SU.5/ group, or a larger group, which
include SU.3/ SU.2/ U.1/. The advantage of having one larger group is
that one would have only one coupling parameter and the standard model would
be recovered by spontaneous symmetry breaking at lower energies. This opens the
way to the realization of processes in which baryon number is not conserved, with a
baryon, for example, decaying into leptons and bosons. The experimental55 bound
on lifetime of proton decay seems to be >1033 years and is much larger than the
age of the universe which is about 13.8 billion years.56 Such a rare event even if it
occurs once will give some support of such grand unified theories.
We have covered quite a large territory and before continuing this presentation,
we pose for a moment, at this appropriate stage, to discuss three aspects of
importance that are generally expected in order to carry out reliable computations
in perturbative quantum field theory. These are:
1. The development of a powerful and simple formalism for doing this.
2. To show how the renormalization process is to be carried, and establish that the
resulting expressions are finite to any order of perturbation theory.
3. The physical interpretation will be completed if through the process of renor-
malization, the initial experimentally unattainable parameters in the theory are
eliminated in favor of physically observed ones, which are finite in number, and
are generally determined experimentally as discussed earlier in a self consistent
manner.
54
The Planck energy (mass) will be introduced in detail later.
55
See, e.g., Olive et al. [167].
56
A decay of the proton may have a disastrous effect in the stability of matter over anti-matter itself
in the universe. See, however, the discussion given later on the dominance of matter in the visible
universe.
18 1 Introduction
Perturbatively renormalizable theories are distinguished from the non-renormaliz-

able ones by involving only a finite number of parameters in the theory that are
determined experimentally.
We discuss each of these in turn.
1. A powerful formalism is the Path Integral one, pioneered by Feynman,57 defining
a generating functional for so-called Green functions from R which physical
amplitudes may be extracted, and has the general structure dŒ¦ eiAction . Here
dŒ¦ defines a measure of integration over classical fields as the counterparts
of the quantum fields of the theory.58 “Action” denotes the classical action. In
the simplest case, the measure dŒ¦ takes the form ˘x d.x/ as a product of
all spacetime points. In gauge theories, due to constraints, the determination of
the measure of integrations requires special techniques59 and takes on a much
more complicated expression and was successfully carried out by Faddeev and
Popov in 1967.60 The path integral expression as it stands, involves continual
integrations to be carried out.
An equally powerful and quite an elegant formalism is due to Schwinger,61
referred to as the Action Principle or the Quantum Dynamical Principle. For
quantum field theory computations, the latter gives the variation of the so-called
vacuum-to-vacuum transition amplitude (a generating functional): •h 0C j 0 i as
any of the parameters of the theory are made to vary. The latter is then expressed
as a differential operator acting on a simple generating functional expressed
in closed form. This formalism involves only functional differentiations to be
carried out, no functional integrations are necessary, and hence is relatively
easier to apply than the path integral. We will learn later, for example, that the
path integral may be simply obtained from the quantum dynamical principle
by a functional Fourier transform thus involving functional integrals. Again the
application of the quantum dynamical principle to the quantization of gauge
theories with underlying constraints require special techniques and it was carried
out in Manoukian [150].
All the fundamental interactions in nature are presently described theoretically
by gauge theories, involving constraints, and the two approaches of their
quantization discussed above will be both treated in this book for pedagogical
reasons and are developed 62
57
See Feynman and Hibbs [78] for the standard pedagogical treatment. See also Feynman [75].
58
We use a general notation .x/ for the fields as functions of spacetime variable and suppress all
indices that they may carry to simplify the notation. These fields may include so-called Grassmann
fields.
59
Feynman [72], DeWitt [43, 44], and Faddeev and Popov [66].
60
Op. cit.
61
Schwinger [194–197, 201].
62
There is also the canonical formalism, see, e.g., Mohapatra [161, 162] and Utiyama and
Sakamoto [228].
1 Introduction 19
via the
Path Integral [66],
or via the
Action Principle (Quantum Dynamical Principle) [150].
2. Historically, Abdus Salam, was the first “architect” of a general theory of renor-
malization. In 1951, he carried out a systematic study63 of renormalization [180],
introduced and sketched a subtraction scheme in a general form. Surprisingly,
this classic paper was not carefully reexamined until much later. This was
eventually done in 1976 [147] by Manoukian, and inspired by Salam’s work,
a subtraction scheme was developed and brought to a mathematically consistent
form, and the finiteness of the subtracted, i.e., renormalized, theory was proved
by the author64 to any order of perturbation theory.65 by using, in the process, a
power counting theorem established by Weinberg [235] for integrals of a special
class of functions, thus completing the Dyson-Salam program. The subtraction
was carried out directly in momentum space and no cut-offs were introduced.
Shortly after the appearance of Salam’s work, two other “architects” of a
general theory of renormalization theory, Bogoliubov and Parasiuk, in a classic
paper in 1957 [24], also developed a subtraction scheme. In 1966 [118], Hepp
gave a complete proof of the finiteness of the Bogoliubov-Parasiuk to any order
of perturbation theory, by using in the intermediate stages ultraviolet cut-offs,
and in 1969 [251], Zimmermann formulated their scheme in momentum space,
without cut-offs, and provided a complete proof of finiteness as well, thus
completing the Bogoliubov-Parasiuk program. This scheme is popularly known
as the BPHZ scheme.
The equivalence of the Bogoliubov-Parasiuk scheme, in the Zimmermann
form, and our scheme was then proved by Manoukian,66 after some systematic
cancelations in the subtractions. This equivalence theorem67 unifies the two
monumental approaches of renormalization.68
63
Salam [180], see also Salam [179].
64
Manoukian [149].
65
For a pedagogical treatment of all these studies, see my book “Renormalization” [149]. This also
includes references to several of my earlier papers on the subject as well as many results related to
renormalization theory.
66
See Manoukian [149] op. cit.
67
This result has been also referred to as “Manoukian’s Equivalence Principle”, Zeidler [250],
p. 972. See also Streater [207].
68
I was pleased to see that our equivalence theorem has been also considered, by completely
different methods, by Figueroa and Gracia-Bondia [80]. For other earlier, and recent, but different,
approaches to renormalization theory, see, e.g., Epstein and Glaser [65], Kreimer [131, 132],
Connes and Kreimer [34, 35], and Figueroa and Gracia-Bondia [79, 81]. See also Landsman [135]
and Aschenbrenner [11].
20 1 Introduction
DS : Dyson Salam Weinberg Author

(program) (completion)
Author
(equivalence)
BP : Bogoliubov - Parasiuk Hepp - Zimmermann

(program) (completion)
Fig. 1.8 Developments of the general theory of renormalization from the DS and BP programs.
The intricacies of this layout also appear in Zeidler [250], pp. 972–975. Regarding the author’s
work shown in the above layout and of his completion of the renormalization program stemming
out of Salam’s, Streater [207] writes: “It is the end of a long chapter in the history of physics”
The development of the general theory of renormalization from the DS and

BP programs may be then summarized as given in Fig. 1.869 .
3. The physical interpretation of the theory is completed by showing that the
subtractions of renormalization are implemented by counterterms in the theory
which have the same structures as the original terms in the theory (i.e., in
the Lagrangian density),70 thus establishing the self-consistency involved in the
elimination of the initial parameters in the theory in favor of physically observed
ones. As mentioned above, for a theory to be renormalizable, i.e., involving
only a finite number of parameters determined, in general, experimentally, the
counterterms of the theory must be finite in number as well.
All particles due to their energy content experience the gravitational attraction.
Einstein’s theory of gravitation, also referred to as general relativity (GR), is
described by a second rank tensor with the energy-momentum tensor of matter
as its source from which the energy density of matter may be defined. It may
not be described just by a scalar or just by a vector field as they are inconsistent
with experiment. It is easy to see that due to the fact that masses have the same
signs (positive) a theory based on a vector field alone will lead to a repulsive
rather an attractive gravitational force.71 GR theory predictions are well supported
experimentally in our solar system.
The key observation, referred to as the principle of equivalence, of Einstein is that
at any given point in space and any given time, one may consider a frame in which
gravity is wiped out at the point in question. For example, in simple Newtonian
gravitational physics, a test particle placed at a given point inside a freely falling
elevator on its way to the Earth, remains at rest, inside the elevator, for a very
short time, depending on the accuracy being sought, and, depending on its position
relative to the center of the Earth, eventually moves, in general, from its original
position in a given instant. Einstein’s principle of equivalence applies only locally
69
The layout in Fig. 1.8 is based on Manoukian [149], and it also appears in Zeidler [250], p. 974.
See also Streater [207] and Figueroa and Gracia-Bondia [80].
70
For a detailed study of this see, Manoukian ([148]; Appendix, p. 183 in [149]).
71
Attempts have been made to include such fields as well for generalizations of Einstein’s theory,
but we will not go into it here.
1 Introduction 21
at a given point and at a given time. At the point in question, in the particular
frame in consideration, gravity is wiped out and special relativity survives. The
reconciliation between special relativity and Newton’s theory of gravitation, then
readily leads to GR, where gravity is accounted for by the curvature of spacetime
and its departure from the flat spacetime of special relativity one has started out
with upon application of the principle of equivalence. By doing this, one is able to
enmesh non-gravitational laws with gravity via this principle.
Quantum gravity (QG) is needed in early cosmology, black hole physics, and,
in general, to deal with singularities that arise in a classical treatment. QG must
also address the problem of the background geometry. A common interest in
fundamental physics is to provide a unified description of nature which is applicable
from microscopic to cosmological distances. A fundamental constant of unit of
length that is expected to be relevant to this end is the Planck length as well as the
Planck mass. Out of the fundamental constants of quantum physics „, of relativity
c, and the Newtonian gravitational one GN , we may define a unit of length and
mass, the Planck length and Planck mass, respectively, relevant in quantum gravity,
through the following
r s
„GN „c
`P D ' 1:616 1033 cm; mP D ' 1:221 1019 GeV=c2 :
c3 GN
(1.10)
In units „ D 1, c D 1, dimensions of physical quantities may be then expressed
in powers of mass (ŒEnergy D ŒMass; ŒLength D ŒMass1 D ŒTime; : : :), and, as
gravitation has a universal coupling to all forms of energy, one may hope that it
may be implemented within a unified theory of the four fundamental interactions,
with the Planck mass providing a universal mass scale. Unfortunately, it is difficult
experimentally to investigate the quantum properties of spacetime as one would be
working at very small distances.
GR predicts the existence of Black holes. Here it is worth recalling of the
detection (“Observational waves from a binary black hole merger”, Phys. Rev. Lett.
116, 061102 (1–16) (2016)) by B. P. Abbott et al. of gravitational waves from the
merger of two black holes 1.3 billion light-years from the Earth. Recall that a black
hole (BH) is a region of space into which matter has collapsed and out of which
light may not escape. It partitions space into an inner region which is bounded by
a surface, referred to as the event horizon which acts as a one way surface for light
going in but not coming out. The sun’s radius is much larger than the critical radius
of a BH which is about 2.5 km to be a black hole for the sun. We will see that
for a spherically symmetric BH of mass M, the radius of the horizon is given by
RBH D 2GN M=c2 .72
72
This may be roughly inferred from Newton’s theory of gravitation from which the escape speed
of a particle in the gravitational field of a spherically symmetric massive body of mass M, at a
distance r, is obtained from the inequality v 2 =2 GN M=r < 0, and by formally replacing v by
22 1 Introduction
One may argue that the Planck length may set a lower limit spatial cut-off. The
following formal and rough estimates are interesting. Suppose that by means of a
high energetic particle of energy E, hE 2 i hp 2 ic2 , with hp 2 i very large, one is
interested in measuring a field within an interval of size ı around a given p point in
space. Such form of energy acts as an effective gravitational mass M hE 2 i=c4
which, in turn distorts space around it. The radius of the event horizon of such a
gravitation mass M is given by r BH D 2GN M=c2 . Clearly we must have ı > r BH ,
otherwise the region of size ı that we wanted to locate the point in question will be
hidden beyond a BH horizon, and localization fails. Also:
˝ 2 ˛
hp 2 i p hpi „2=4ı 2 . Hence M „=2cı,
2GN M „GN
ı> 2
3 ;
c cı
p
which gives ı > rBH D „GN =c3 D `P .
Interesting investigations by Hawking73 have shown that a BH is not really a
black body, it is a thermodynamic object, it radiates and has a temperature associated
with it.74 In Chapter 7 in Vol. II , we will see, considering a spherically symmetric
BH, that its temperature is given by75
„c3
TBH D : (1.11)
8 GN M k B
where kB is the Boltzmann constant. Note that a very massive black hole is cold.
Recall that entropy S represents a measure of the amount of disorder with
information encoded in it, and invoking the thermodynamic interpretation of a BH,
we may write
@S 1
2
D ; (1.12)
@.Mc / T
which upon integration with boundary condition that for M ! 0, S ! 0, gives the
celebrated result

c3 k B A 2GN M 2
SBH D A D kB 2 ; A D 4 (1.13)
4„GN 4 `P c2
the ultimate speed c to obtain for the critical radius Rcritical D 2GN M=c2 such that for r < Rcritical
a particle cannot escape.
73
Hawking [114, 115].
74
Particle emission from a BH is formally explained through virtual pairs of particles created near
the horizon with one particle falling into the BH while the other becoming free outside the horizon.
75
A pedestrian approach in determining the temperature is the following. By comparing the
expression of energy expressed in terms of the wavelength of radiation : E D h c=, with
the expression E D kB T, gives T D hc=kB . On dimensional grounds 2 GN M=c2 , which
gives T „c3 =GN M kB . This is the expression given for the temperature up to a proportionality
constant.
1 Introduction 23
referred to as the Bekenstein-Hawking Entropy formula76 of a BH. This relation

relates quantum gravity to information theory. This result is expected to hold in any
consistent formulation of quantum gravity, and shows that a BH has entropy unlike
what would naïvely expect from a BH with the horizon as a one way classical surface
through which information is lost to an external observer. The proportionality of the
entropy to the area rather than to the volume of a BH horizon should be noted. It
also encompasses Hawking’s theorem of increase of the area with time with increase
of entropy. We will discuss the Bekenstein-Hawking Entropy formula below in
conjunction with loop quantum gravity and string theory.
Now we turn back to the geometrical description of gravitation given earlier, and
introduce a gravitational field to account for the departure of the curved spacetime
metric from that of the Minkowski one to make contact with the approaches of
conventional field theories, dealing now with a field permeating an interaction
between all dynamical fields. The quantum particle associated with the gravitational
field, the so-called graviton, emerges by considering the small fluctuation of the
metric, associated with curved spacetime of GR about the Minkowski metric, as
the limit of the full metric, where the gravitational field becomes weaker and the
particle becomes identified. This allows us to determine the graviton propagator
in the same way one obtains, for example, the photon propagator in QED, and
eventually carry out a perturbation theory as a first attempt to develop a quantum
theory of gravitation.
In units of „ D 1; c D 1, Newtons gravitational constant GN , in 4 dimensional
spacetime, has the dimenionality Œ GN D Œ Mass 2 , which is a dead give away of
the non-renormalizabilty of a quantum theory of gravitation based on GR. The non-
renormalizability of the theory is easier to understand by noting that the divergences,
in general, tend to increase as we go to higher orders in the gravitational coupling
constant without a bound, implying the need of an infinite of parameters need to
be fixed experimentally77 and hence is not of any practical value. Some theories
which are generalizations of GR, involving higher order derivatives, turn out to be
renormalizable78 but violate, in a perturbative setting, the very sacred principle of
positivity condition of quantum theory. Unfortunately, such a theory involves ghosts
in a perturbative treatment, due to the rapid damping of the propagator at high
energies faster than 1=energy2 , and gives rise, in turn, to negative probabilities.79
One is led to believe that Einstein’s general relativity is a low energy effective
theory as the low energy limit of a more complicated theory, and as such it provides
a reliable description of gravitation at low energies. Moreover, one may argue that
the non-renormalizability of a quantum theory based on GR is due to the fact that
one is trying to use it at energies which are far beyond its range of validity. As a
76
Bekenstein [14].
77
Manoukian [149] and Anselmi [9].
78
Stelle [204].
79
Unitarity (positivity) of such theories in a non-perturbative setting has been elaborated upon by
Tomboulis [224].
24 1 Introduction
matter of fact the derivatives occurring in the action, in a momentum description via
Fourier transforms, may be considered to be small at sufficiently low energies. In
view of applications in the low energy regime, one then tries to separate low energy
effects from high energy ones even if the theory has unfavorable ultraviolet behavior
such as in quantum gravity.80 Applications of such an approach have been carried
out in the literature as just cited, and, for example, the modification of Newton’s
gravitational potential at long distances has been determined to have the structure
GN m1 m2 h GN .m1 C m2 / GN „ i
U.r/ D 1C˛ C ˇ ; (1.14)
r c2 r c3 r 2
for the interaction of two spin 0 particles of masses m1 and m2 . Here ˛; ˇ; are
dimensionless constants,81 and the third term represents a quantum correction being
proportional to „.
Conventional quantum field theory is usually formulated in a fixed, i.e., in, a
priori, given background geometry such as the Minkowski one. This is unlike the
formalism of “Loop Quantum Gravity” (LQG) also called “Quantum Field Theory
of Geometry”. The situation that we will encounter in this approach is of a quantum
field theory in three dimensional space, which is a non-perturbative background
independent formulation of quantum gravity. The latter means that no specific
assumption is made about the underlying geometric structure and, interestingly
enough, the latter rather emerges from the theory. Here by setting up an eigenvalue
equation of, say, an area operator, in a quantum setting, one will encounter a
granular structure of three-dimensional space yielding a discrete spectrum for area
measurements with the smallest possible having a non-zero value given to be of the
order of the Planck length squared: „GN =c3 1066 cm2 .82 The emergence of
space in terms of “quanta of geometry”, providing a granular structure of space, is
a major and beautiful prediction of the theory. The 3 dimensional space is generated
by a so-called time slicing procedure of spacetime carried out by Arnowitt, Deser
and Misner.83 The basic field variables in the theory is a gravitational “electric” field,
which determines the geometry of such a 3 dimensional space and naturally emerges
from the definition of the area of a surface in such a space, and its canonical conju-
gate variable is referred to as the connection. By imposing equal time commutation
relation of these two canonically conjugate field variables, the quantum version of
the theory arises, and the fundamental problem of the quantization of geometry
follows. The basic idea goes to Penrose [171] whose interest was to construct the
concept of space from combining angular momenta. It is also interesting that the
proportionality of entropy and the surface area of the BH horizon in the Bekenstein-
Hawking Entropy formula has been derived in loop quantum gravity.84
80
Donoghue [55–57] and Bjerrum-Bohr et al. [19], Bjerrum-Bohr et al. [20].
81
Recent recorded values are ˛ D 3, and ˇ D 41=10 Bjerrum-Bohr et al. [19].
82
Rovelli and Smolin [174], Ashtekar and Lewandoski [12], and Rovelli and Vidotto [175].
83
Arnowitt et al. [10].
84
See, e.g., Meissner [160] and Ansari [8].
1 Introduction 25
Supersymmetry is now over 40 years old. Supersymmetry provides a symmetry

between fermions and bosons. Borrowing a statement made by Dirac, speaking
of theories, in general, it is a theory with mathematical beauty.85 The name
“Supersymmetry” for this symmetry is attributed to Salam and Strathdee as it
seemed to have first appeared in the title of one of their papers [184]. An abbreviated
name for it is SUSY, as some refer to this symmetry. The latter is not only
beautiful but is also full of thought-provoking surprises. To every degree of freedom
associated with a particle in the standard model, in a supersymmetric version,
there corresponds a degree of freedom associated with a partner, referred to as a
sparticle, with the same mass and with opposite statistics to the particle.86 Unlike
other discoveries, supersymmetry was not, a priori, invented under pressure set by
experiments and was a highly intellectual achievement. Theoretically, however, it
quickly turned out to be quite important in further developments of quantum field
theory. For one thing, in a supersymmetric extention of the standard model, the
electroweak and strong effective couplings do merge at energies about 1016 GeV,
signalling the possibility that these interactions are different manifestations of a
single force in support of a grand unified theory of the fundamental interactions.
Also gravitational effects are expected to be important at the quantum level at the
Planck energy of the order 1019 GeV, or possibly at even lower energies, giving the
hope of having a unified theory of the four fundamental interactions at high energies.
Supersymmetry leads to the unification of coupling constants. SUSY also tends to
“soften” divergences of a theory in the sense that divergent contributions originating
from fermions tend to cancel those divergent contributions originating from bosons
due to their different statistics.
One of the important roles that supersymmetry may play in a supersymmetric
extension of the standard problem is in the so-called hierarchy problem. The basic
idea of a facet of this is the following. A fundamental energy scale arises in the
standard model from the vacuum expectation value of the Higgs boson which sets
the scale for the masses in the theory, such as for the masses of the vector bosons.
It turns out that the masses imparted to the initially massless vector bosons, for
example, via the Higgs mechanism, using the parameters in the Lagrangian density
are in very good agreement with experimental results. On the other hand, if one
introduces a large energy scale cut off 1015 GeV, of the order of a grand unified
energy scale, or the Planck energy scale 1019 GeV, at which gravitation may play a
significant role, to compute the shift of the squared-mass of the Higgs boson, as a
scalar particle, due to the dynamics (referred to as radiative corrections), it turns out
to be quadratic87 in , which is quite large for such a large cut-off. This requires
that the bare mass squared of the Higgs boson to be correspondingly large to cancel
85
Here we recall the well known statement of Dirac, that a theory with mathematical beauty is
more likely to be correct than an ugly one that fits some experimental data [53].
86
This is such that the total number of fermion degrees of freedom is equal to the total bosonic
ones.
87
See, e.g., Veltman [230].
26 1 Introduction
such a quadratic dependence on and obtain a physical mass of the Higgs boson
of the order of magnitude of the minute energy88 . 10 2 GeV/, in comparison,
characterizing the standard model, and this seems quite unnatural for the cancelation
of such huge quantities.89 This unnatural cancelation of enormously large numbers
has been termed a facet of the hierarchy problem. Supersymmetric theories have,
in general, the tendency to cancel out such quadratic divergences, up to possibly of
divergences of logarithmic type which are tolerable, thus protecting a scalar particle
from acquiring such a large bare mass. So supersymmetry may have an important
role to play here.
SUSY relates fermions to bosons, and vice versa, and hence a generator is
required which is of fermionic type, that is, it carries a spinor index as in the
Dirac field90 to carry out a transformation fermion $ boson. Since the spins of
fermions and bosons are different, this necessarily means that such a generator does
not commute with the angular (spin) momentum operator as supersymmetry unites
particles of the same mass and different spins into multiplets. Bosons and fermions
have, in general, different masses, which means that SUSY is to be spontaneously
broken if such a symmetry is to have anything to do with nature. If supersymmetry
breaking sets at such an energy scale as 1 TeV or so, then some of the lowest mass
superpartners may be hopefully discovered.91
Of particular interest was also the development of the superspace concept as
an extension of the Minkowski one, where one includes an additional degree of
freedom usually denoted by92 D .a / to the space coordinates .t; x/, which
turns out to be quite convenient in defining and setting up SUSY invariant integrals
such as the action of a dynamical system.93 To describe dynamics this, in turn,
necessitates to introduce superfields of different types94 as functions of these
variables.
The extension of the algebra of the Poincaré group to a superalgebra was first
carried out by Gol’fand and Likhtman in [100] to construct supersymmetric field
theory models, and with the implementation of spontaneous symmetry breaking
88
Aad et al. [1] and Chatrchyan et al. [31].
89
As mentioned earlier, the question, in turn, arises as to what amounts for the enormous difference
between the energy scale of grand unification and the energy scale that characterizes the standard
model.
90
This point is of importance because an earlier attempt by Coleman and Mandula [33] to enlarge
the Poincaré group did not work. They considered only so-called “Bose” generators (that is tensors,
and not spinors) in their analysis.
91
Perhaps an optimist would argue that since antiparticles corresponding to given particles were
discovered, the discovery of superpartners associated with given particles would not be out of
the question either. The underlying symmetries involved in these two cases are, however, quite of
different nature.
92
This is called a Grassmann variable.
93
See Salam and Strathdee [185, 186].
94
Details on superfields will be given in Vol. II. The explicit expression of the pure vector superfield
has been recently obtained in Manoukian [155].
1 Introduction 27
by Volkov and Akulov in [232]. In [243], Wess and Zumino also,95 independently,
developed supersymmetric models in 4 dimensions, and this work has led to an
avalanche of papers on the subject and to a rapid development of the theory. In
particular, supersymmetric extensions of the standard model were developed,96
supergravity, as a supersymmetric extension of gravitational theory, was also
developed.97 Unfortunately, things do not seem to be much better for supergravity,
as far as its renormalizability is concerned.98
Now we come to String Theory. String Theory is a theory which attempts to
provide a unified description of all the fundamental interactions in Nature and,
in particular, give rise to a consistent theory of quantum gravity. A string is a
fundamental one dimensional extended object, and if it has p to do with quantum
gravity, it is, say, of the order of the Plank length `P D GN „=c3 1033 cm,
involving the three fundamental constants: Newton’s gravitational constant GN , the
quantum unit of action „, and the speed of light c. Since no experiments can probe
distances of the order of the Planck length, such a string in present day experiments
is considered to be point-like. When a string, whether closed or open, moves in
spacetime, it sweeps out a two dimensional surface referred to as a worldsheet.
String Theory is a quantum field theory which operates on such a two dimensional
worldsheet. This, as we will see, has remarkable consequences in spacetime itself,
albeit in higher dimensions. Particles are identified as vibrational modes of strings,
and a single vibrating string may describe several particles depending on its
vibrational modes. Strings describing bosonic particles are referred to as a bosonic
strings, while those involving fermionic ones as well are referred to as superstrings.
The remarkable thing is that the particles needed to describe the dynamics of
elementary particles arise naturally in the mass spectra of oscillating strings, and
are not, a priori, assumed to exist or put in by hand in the underlying theories.
The dimensionality of the spacetime in which the strings live are predicted by the
underlying theory as well and are necessarily of higher dimensions than four for
consistency with Lorentz invariance of spacetime at the quantum level, consisting
of a dimensionality of 26 for the bosonic strings and a spacetime dimensionality of
10 for the superstrings. The extra dimensions are expected to curl up into a space
that is too small to be detectable with present available energies. For example the
surface of a hollow extended cylinder with circular base is two dimensional, with
one dimension along the cylinder, and another one encountered as one moves on
its circumference. If the radius of the base of the cylinder is relatively small, the
cylinder will appear as one dimensional when viewed from a large distance (low
energies). Accordingly, the extra dimensions in string theory are expected to be
95
These basic papers, together with other key ones, are conveniently collected in Ferrara [71].
96
See Fayet [67], Dimopoulos and Georgi [45].
97
See Freedman, van Nieuwenhuizen and Ferrara [84], Deser and Zumino [42].
98
See, e.g., Deser et al. [41], Deser [40], Stelle [205, 206], and Howe and Stelle [124].
28 1 Introduction
small and methods, referred to as compactifications,99 have been developed to deal

with them thus ensuring that the “observable” dimensionality of spacetime is four.
Superstring theories involve fermions and are thus relevant to the real world,
but there are, however, several superstring theories. Also unlike the loop quantum
gravity, which provides a background independent formulation of spacetime with
the latter emerging from the theory itself, as discussed earlier, the strings in
string theories are assumed to move in a pre-determined spacetime, and thus
spacetime plays a passive role in them.100 A theory, referred to as M-Theory,101
based on non-perturbative methods, is envisaged to unify the existing superstrings
theories into one single theory, instead of several ones, and be related to them by
various limiting and/or transformation rules, referred to as dualities,102 and is of 11
dimensional spacetime. M-Theory is believed to be approximated by 11 dimensional
supergravity,103 and the spacetime structure is envisaged to emerge from the theory
as well. Bosonic strings involve tachyonic states. This is unlike the situation in
superstring theories in which supersymmetry plays a key role in their definitions,
and a process referred to as a GSO projection method, ensuring the equality of the
degrees of freedom of bosonic and fermionic states, as required by supersymmetry,
in turn implies that no tachyonic states appear in the theory.104
String theory was accidentally discovered through work carried out by Veneziano
in 1968 when he attempted to write down consistent explicit expressions of meson-
meson scattering amplitudes in strong interactions physics.105 This was, of course
before the discovery of QCD. With the many excited states of mesons and baryons
(resonances), it was observed experimentally that there exists a linear relationship
between spin J and the mass M squared of a resonance given by a linear relationship
dJ
with a universal slope W D ˛ 0; ˛ 0 Š 1 GeV2 ; (1.15)
dM 2
defining so-called Regge trajectories. Veneziano postulated and wrote down a
scattering amplitude of meson – meson scattering: p1 .m1 / C p2 .m2 / ! p3 .m3 / C
p4 .m4 /, which, in particular, showed that the amplitude involves the exchange of an
infinite number of particles (corresponding to arbitrary integer spins). This is unlike
the situation in conventional field theory as QED or the standard model, where they
involve the exchange of a finite number of particles to any given order. String theory
shares this property of the Veneziano amplitude. As a matter of fact the Veneziano
99
An idea used by Kaluza and Klein in their attempt to unify gravity and electromagnetism in a 5
dimensional generalization of general relativity.
100
See also Horowitz [122].
101
Townsend [227], Witten [244], and Duff [58].
102
Duff [58] and Schwarz [191].
103
Cremmer et al. [37].
104
The GSO method of projection was proposed in Gliozzi et al. [98, 99].
105
Veneziano [231], see also Lovelace and Squires [145] and Di Vecchia [54].
1 Introduction 29
amplitude may be derived from string theory. Nambu [164], Nielsen [166] and
Susskind [211] have shown that the famous expression of the amplitude postulated
by Veneziano may be interpreted as a quantum theory of scattering of relativistic
strings. Although, a priori, this was assumed to describe a strong interaction process,
Yoneya [248], and Scherk and Schwarz [190] made use of the fact that string theory
(involving closed strings) contains a spin 2 massless state, which was identified
with the elusive graviton, in addition to a whole spectrum of other excitation modes,
to propose that string theory provides a framework for the unification of general
relativity and quantum mechanics. As early as 1971, Neveu and Schwarz [165],
and Raymond [173] included fermions in their analyses, which eventually led to
the notion of superstrings, and during a short period of time, several types 106 of
superstrings were introduced in the literature.
Due to the assumed non-zero extensions of strings, it is hoped that they provide,
naturally, an ultraviolet cut-off .`P /1 and render all processes involving
strings ultraviolet finite. This is unlike conventional quantum field theory inter-
actions where all the quantum fields are multiplied locally at the same spacetime
points, like multiplying distributions at the same point, and are, in this sense, quite
troublesome.
In string theory, two strings with given vibrational modes, identifying two
given particles, may combine forming one string with an arbitrary number of
different vibrational modes associated with a myriad number of particles, defining
generalized 3-vertices. The combined string may again split into two strings with
associated vibrational modes, identified appropriately with two more particles,
describing a scattering process of 2 particles ! 2 particles. Thus interactions
involve string worldsheets of various topologies arise.
Other extended objects are also encountered in string theory called branes which,
in general, are of higher spatial dimensions than one, with the string defined as a one
dimensional brane. For example, an open string, satisfying a particular boundary
condition, referred to as a Dirichlet boundary condition, specifies a hypersurface,
referred to as a D brane, on which the end points of the open string reside. On the
other hand, the graviton corresponds to a vibrational mode of closed strings, and
since the latter, having no ends, may not be restricted to a brane and moves away
from it. This might explain the weakness of the gravitational field, if our universe
is a 3 dimensional brane embedded in a higher dimensional spacetime. Massless
particles encountered in string theory are really the physically relevant ones because
of the large unit of mass .`P /1 1019 GeV in attributing masses to the spectrum
of massive particles.107 As we will see a massless particle may acquire mass if,
106
Green and Schwarz [102, 103] and Gross et al. [107, 108].
107
A systematic analysis of all the massless field excitations encountered in both bosonic and
superstrings are investigated in Manoukian [152–154], in their respective higher dimensional
spacetimes, and include the determinations of the degrees of freedom associated with them. Note
that in four dimensional spacetime the number of degrees of freedom (spin states) of non-scalar
fields is always two. This is not true in higher dimensional spacetime. For example, the degrees of
freedom associated with a massless vector particle is 8 in 10 dimensions, while for the graviton is
30 1 Introduction
for example, the end points of the open string are attached to two different branes,
instead of a single brane.
We will learn the remarkable facts that Einstein’s general relativity as well as
Yang-Mills field theory may be obtained from string theory.
Interesting high energy scattering amplitudes have been computed in string
theory over the years,108 which provide a hint that space may not be probed
beyond the Planck length – a result shared with “loop quantum gravity”. It is worth
mentioning that the Bekenstein-Hawking Entropy relation has been also derived in
string theory.109
In recent years much work has been done, which is worth mentioning here
but rather briefly, indicating that general relationships may exist between field
theories and string theories, and consequently considerable attention was given
trying to make such a statement more and more precise, with the ultimate hope
of providing, in turn, a consistent and acceptable quantum theory of gravitation
relevant to our world but much work still remains to be done. In particular,
much study has been made to study the equivalence relation between certain four
dimensional gauge theories and superstring theories, referred to as the AdS/CFT
correspondence, where AdS space stands for anti-de-Sitter space, and CFT stands
for conformal field theory.110 Such correspondences have been also referred to
as Gauge/Gravity duality, as well as Maldacena duality, a duality which was
first proposed by Maldacena.111 Without going into technical details, the aim
of this work is to show, for example, the existence of an equivalence relation
between a certain supersymmetric SU.N/ Yang-Mills field theory in 4 dimensional
Minkowski spacetime, and a superstring theory in a 5 dimensional AdS space,
having one additional dimension to the Minkowski one, and with the 5 dimensions
of the AdS space supplemented by 5 extra dimensions defined by a five-sphere,
making up the 10 dimensions of superstrings mentioned earlier. The interest in this
work is that it deals with a connection between string theory (involving gravity) and
35, as shown later in Chapter 3 of Vol. II. In 4 dimensions, their degrees of freedom are, of course,
two.
108
See, e.g., Amati et al. [3, 4] and ’t Hooft [219].
109
See, e.g., Strominger and Fava [208] and Horowitz et al. [123].
110
AdS space and CFT symmetry may be introduced as follows. AdS space, in D dimensions,
may be defined in terms of coordinates z D .z 0 ; z 1 ; : : : ; z D1 ; z D / satisfying a quadratic equation
PD1 k 2 0 2 D 2 2
kD1 .z / .z / .z / D R , P for a given constant R 2 , embedded in a .DC1/ dimensional
space with interval squared ds 2 D jD1 dz j 2 dz 0 2 dz D 2 . On the other hand a DSphere
D1
PD C1
is defined in terms of coordinates y1 ; : : : ; y DC1 satisfying a quadratic equation jD1 .y j /2 D 2
for a given constant . The conformal group, as applied in 4 dimensional Minkowski spacetime, is
defined by a scale transformation x ! x , and a so-called special (conformal) transformation
x 0 x
0 2
D 2 C a ;
x x
for a constant 4-vector a , in addition to the Poincaré ones.
111
Maldacena [146]. See also Witten [245], Gubser et al. [109], and Aharoni et al. [2].
1 Introduction 31
supersymmetric gauge theories. This brings us into contact with the holographic
principle, in analogy to holography in capturing 3 dimensional images of objects
on a 2 dimensional (holographic) plate,112 showing that an equivalence relation
exists between the 3 and the 2 dimensional set-ups. The 4 dimensional quantum
field theory is like a hologram capturing information about the higher dimensional
quantum gravity theory. In this case the SU(N) theory provides a holographic
description of gravitational field. This is in analogy to black hole entropy with
its encoded information being proportional to the area rather than to the volume
of the region enclosed by the horizon. Perhaps holography is a basic property of
string theory and one expects that much has to be done before developing a realistic
quantum gravity, and in turn provide a background independent formulation for
string theory. The holographic principle was first proposed by ’t Hooft.113
We close this chapter by commenting on two symmetries which seem to be
observed in Nature, that is of the CPT symmetry and of the Spin & Statistics
connection and of their relevance to our own existence. We will see how these
symmetries arise from quantum field theory in Sect. 4.10 and Sect. 4.5, respectively.
CPT taken in any order, seems to be an observed symmetry in Nature, where C
stands for charge conjugation with which particles are replaced by their antiparticles
and vice versa, P represents space reflection, while T denotes time reversal.
Local Lorentz invariant quantum field theories preserve (Sect. 4.10) the CPT
symmetry. Experimentally, symmetry violations are well known to occur when
one restricts to one or to the product of two transformations in CPT in dealing
with some fundamental processes. For example the violation of parity was already
established in 1957114 as well as the violation of charge symmetry.115 Later, in
1964 CP violation, and hence also of T, was observed in neutral Kaon decays.116
The CP transformation and C, provide the fundamental relations between matter
and anti-matter. The question then arises as to why we observe, apart in accelerator
experiments, only one form (matter) than the other form in the visible universe –
a key criterion for our own existence. If an equal amount of matter and anti-matter
was produced at some stage then why, our visible universe is matter dominated.
Sakharov in 1967 [177] proposed that a key reason for this is CP violation. In
more details to explain this asymmetry, he proposed that (1) baryon number is not
conserved. (This is supported by recent grand unified field theories,) (2) CP and C
are violated, (3) the universe has gone through a phase of extremely rapid expansion
to avoid the pairing of matter and anti-matter. The violation of such symmetries, at
112
Recall that the two dimensional holographic plate which registers the interference of reflected
light off an object and an unperturbed Laser beam stores information of the shape of the three
dimensional object. As one shines a Laser beam on it an image of the three dimensional object
emerges.
113
’t Hooft [220], see also especially Thorn [223], as well as the analysis with further interpretations
by Susskind [212]. See also Bousso [27].
114
Wu et al. [246], Garwin et al. [91], and Friedman and Telegdi [86].
115
Garwin et al. [91].
116
Christenson et al. [32].
32 1 Introduction
the microscopic level, and their consequences on a macroscopic scale is certainly

intriguing.
Clearly, the “Spin & Statistics” connection, of which the Pauli exclusion
principle is a special case applicable to spin 1/ 2 particles, is important not only
in physics but in all of the sciences, and is relevant to our own existence. For
one thing, the periodic table of elements in chemistry is based on the exclusion
principle. In simplest terms, the upshot of this is that half-odd-integer spin fields
are quantized by anti-commutators, while integer spins fields are quantized by
commutators. This result is of utmost significance for our existence. As a matter of
fact the Pauli exclusion principle is not only sufficient for the stability of matter117
in our world but it is also necessary.118 In the problem of stability of neutral
matter, with a finite number of electrons per atom, but involving several nuclei, and
correspondingly a large number of electrons N, the stability of matter, based on the
Pauli exclusion principle, or instability of so-called “bosonic matter”, in which the
exclusion principle is abolished, rests rather on the following. For “bosonic matter”,
the ground state energy EN N ˛ , with ˛ > 1,119 where .N C N/ denotes
the number of the negatively charged particles plus an equal number of positively
charged particles. This behavior for “bosonic matter” is unlike that of matter, with
the exclusion principle, for which ˛ D 1.120 A power law behavior with ˛ > 1
implies instability as the formation of a single system consisting of .2N C 2N/
particles is favored over two separate systems brought together each consisting
of .N C N/ particles, and the energy released upon collapse of the two systems
into one, being proportional to Œ.2N/˛ 2.N/˛ , will be overwhelmingly large
for realistic large N, e.g., N 1023 . Dyson [61], has estimated that without the
exclusion principle, the assembly of two macroscopic objects would release energy
comparable to that of an atomic bomb, and such “matter” in bulk would collapse
into a condensed high-density phase and our world will cease to exist.121 Ordinary
matter, due to the exclusion principle, occupies a very large volume. This point
was emphasized by Ehrenfest in a discussion with Pauli in 1931122 on the occasion
of the Lorentz medal to this effect: “We take a piece of metal, or a stone. When
we think about it, we are astonished that this quantity of matter should occupy so
large a volume”. He went on by stating that the exclusion principle is the reason:
“Answer: only the Pauli principle, no two electrons in the same state”. In this regard,
117
For a pedagogical treatment of the problem of “stability of matter” and related intricacies, see
Manoukian [151], Chapter 14.
118
Lieb and Thirring [144] and Thirring [222].
119
Dyson [61], Lieb [143], and Manoukian and Muthaporn [156].
120
Lieb and Thirring [144] and Thirring [222].
121
In the Preface of Tomonaga’s book on spin [226], one reads: “The existence of spin, and the
statistics associated with it, is the most subtle and ingenious design of Nature – without it the
whole universe would collapse”.
122
See Ehrenfest [62].
References 33
a rigorous treatment123 shows that the extension of matter radially grows not any
slower than N 1=3 for large N. No wonder why matter occupies so large a volume.
The importance of the “Spin & Statistics” connection and the role it plays in our
world cannot be overemphasized. Needless to say, no quantum field theory treatment
is complete without the CPT Theorem and the Spin & Statistics Connection.
The present volume deals with the foundations of quantum field theory and with
the intricacies of abelian and non-abelian gauge theories. Volume II deals with
quantum gravity, supersymmetry, and string theory.
References
1. Aad, G. et al. (2012). Observation of a new particle in the search for the Standard Model
Higgs Boson with the ATLAS detector at the LHC. Physics Letters, B716, 1–29.
2. Aharoni, O. et al. (2008). N D 6 superconformal Chern-Simons matter theories, M2-branes
and their gravity duals. JHEP, 0810, 091.
3. Amati, D., Ciafaloni, M., & Veneziano, G. (1987). Superstring collisions at planckian
energies. Physics Letters, B197, 81–88.
4. Amati, D., Ciafaloni, M., & Veneziano, G. (1988). Classical and Quantum effects from
Planckian energy superstring collisions. International Journal of Modern Physics, 3, 1615–
1661.
5. Anderson, C. D. (1932). The apparent existence of easily deflectable positives. Science, 76,
238–239.
6. Anderson, C. D. (1933). The positive electron. Physical Review, 43, 491–494.
7. Anderson, P. W. (1963). Plasmons, gauge invariance, and mass. Physical Review, 130, 439–
442.
8. Ansari, M. H. (2008). Generic degeneracy and entropy in loop quantum gravity. Nuclear
Physics, B795, 635–644.
9. Anselmi, D. (2003). Absence of higher derivatives in the renormalization of propagators in
quantum field theory with infinitely many couplings. Class. Quantum Grav., 20, 2344–2378.
10. Arnowitt, R. S., Deser, S., & Misner, W. (2008). The dynamics of general relativity. General
Relativity and Gravitation, 40, 1997–2027. Reprinted from Gravitation: An Introduction to
current research (Chap.7), Edited by L. Witten. John Wiley & Sons Inc., New York/London,
1962.
11. Aschenbrenner, M. (1996). A decoupling theorem for the BPHZL-scheme. Annals of Physics,
250, 320–351.
12. Ashtekar, A., & Lewandoski, J. (1997). Quantum theory of gravity I: Area operators. Classical
Quantum Gravity, 14, A55–A81.
13. Becchi, C., Rouet, A., & Stora, R. (1976). Renormalization of gauge theories. Annals of
Physics, 98, 287–321.
14. Bekenstein, J. D. (1973). Black holes and entropy. Physical Review, D7, 2333–2346.
15. Benvenuti, A. et al. (1974). Observation of Muonless neutrino-induced inelastic interactions.
Physical Review Letters, 32, 800–803.
16. Beringer, J. et al. (2012). Particle data group. Physical Review D, 86, 010001.
17. Bethe, H. (1947). The Electromagnetic shift of energy levels. Physical Review, 72, 339–341.
18. Bethe, H., & Fermi, E. (1932). Über die Wechselwirkung von Zwei Elektronen. Zeitschrift
fur Physik, 77, 296–306.
123
Manoukian and Sirininlakul [157].
34 1 Introduction
19. Bjerrum-Bohr, N. E. J., Donoghue, J. F., & Holstein, B. R. (2003a). Quantum gravitational
corrections to the non-relativistic scattering potential of two mesons. Physical Review, D 67,
084033. Erratum: ibid., D71, 069903 (2005).
20. Bjerrum-Bohr, N. E. J., Donoghue, J. F., & Holstein, B. R. (2003b). Quantum corrections to
the schwarzchild and kerr metrics. Physical Review, 68, 084005–084021.
21. Bjorken, J. D., & Pachos, E. A. (1969). Inelastic electron-proton and y-proton scattering and
the structure of the nucleon. Physical Review, 185, 1975–1982.
22. Bloch, F., & Nordsieck, A. (1937). Notes on the radiation field of the electron. Physical
Review, 52, 54–59.
23. Bludman, S. (1958). On the universal fermi interaction. Nuovo Cimento, 9, 433–445.
24. Bogoliubov, N. N., & Parasiuk, O. S. (1957). On the multiplication of propagators in quantum
field theory. Acta Physics Mathematics, 97, 227–266. Original German Title: Über die
Multiplikation der Kausalfunctionen in der Quantentheorie der Felder.
25. Bogoliubov, N. N., & Shirkov, D. V. (1959). Introduction to the theory of quantized fields.
Interscience, New York.
26. Born, M., Heisenberg, W., & Jordan, P. (1926). Zur Quantenmechanik III. Zeitschrift fur
Physik, 35, 557–615. Reprinted in Sources of quantum mechanics, (ed. B. L. vander Waerden),
Dover Publications, New York (1968).
27. Bousso, R. (2002). The holographic principle. Reviews of Modern Physics, 74, 825–874.
28. Callan, C. G. (1970). Broken scale invariance in scalar field theory. Physical Review, D2,
1541–1547.
29. Callan, C. G. (1972). Broken scale invariance and asymptotic behavior. Physical Review, D5,
3202–3210.
30. Chamberlain, O., Segrè, E., Wiegand, C. and Ypsilantis, T. (1955). Observation of antipro-
tons. Physical Review, 100, 947–950.
31. Chatrchyan, S. et al. (2012). Observation of a New Boson at Mass 125 GeV with the CMS
experiment at LHC. Physics Letters, B716, 30–61.
32. Christenson, J. H. et al. (1964). Evidence for the 2 decay of the K02 meson. Physical Review
Letters, 13, 138–140.
33. Coleman, S., & Mandula, J. (1967). All possible symmetries of the S matrix. Physical Review,
150, 1251–1256.
34. Connes, A., & Kreimer, D. (1998). Hopf algebras, renormalization and noncommutative
geometry. Communications in Mathematical Physics, 119, 203–242.
35. Connes, A., & Kreimer, D. (2000). Renormalization in quantum field theory and the Riemann-
Hilbert problem I: The Hopf algebra structure and the main theorem. Communications in
Mathematical Physics, 210, 249–273.
36. CPLEAR/Collaboration (2000). T violation and CPT tests at CPLEAR, symmetries in
subatomic physics. In: 3rd International Symposium. AIP Conference Proceedings (Vol. 539,
pp. 187–196), Adelaide (Australia).
37. Cremmer, E., Julia, B., & Scherk, J. (1978). Supergravity theory in eleven-dimensions.
Physics Letters, B76, 409–412.
38. Cronin, J. W. (1981). CP symmetry violation: The search of its origin. Reviews of Modern
Physics, 53, 373–383.
39. Cronin, J. W., & Fitch, V. L. (1964). Evidence for the 2 decay of the K2 0 meson. Physical
Review Letters, 13, 138–140.
40. Deser, S. (2000). Infinities in quantum gravities. Annalen der Physik, 9, 299–306.
41. Deser, S., Kay, J. H., & Stelle, K. S. (1977). Renormalizability properties of supergravity.
42. Deser, S., & Zumino, B. (1976). Consistent supergravity. Physics Letters, 62B, 335–337.
43. DeWitt, B. (1964). Theory for radiative corrections for non-Abelian gauge fields. Physical
44. DeWitt, B. (1967a). Quantum theory of gravity. II. The manifestly covariant theory. Physical
Review, 162, 1195–1239.
References 35
45. Dimopoulos, S., & Georgi, H. (1981). Softly broken supersymmetry and SU(5). Nuclear
Physics, B193, 150–162.
46. Dirac, P. A. M. (1927). The quantum theory of the emssion and absorption of radiation.
Proceedings of the Royal Society of London, A114, 243–265.
47. Dirac, P. A. M. (1928a). The quantum theory of the electron, I. Proceedings of the Royal
Society of London, A114, 610–624.
48. Dirac, P. A. M. (1928b). The quantum theory of the electron, II. Proceedings of the Royal
Society of London, A117, 610–624.
49. Dirac, P. A. M. (1928c). Über die Quantentheorie des Elektrons. Physikalishce Zeitschrift, 29,
561–563.
50. Dirac, P. A. M. (1930a). A theory of electrons and protons. Proceedings of the Royal Society
of London, A126, 360–365.
51. Dirac, P. A. M. (1930b). On the annihilation of electrons and protons. Proc. Cambridge Phil.
Soc., 26, 361–375.
52. Dirac, P. A. M. (1931). Quantized singularities in the electromagnetic field. Proceedings of
the Royal Society of London, A133, 60–72.
53. Dirac, P. A. M. (1970). Can equations of motion be used in high-energy physics? Physics
Today, 23, 29. April, issue (4).
54. Di Vecchia, P. (2008). The birth of string theory. In M. Gasperini & J. Maharana (Eds.), String
theory and fundamental interactions: Gabriele Veneziano and theoretical physics: Historical
and contemporary perspectives. (Lecture notes in physics, vol. 737, pp. 59–118). Berlin/New
York: Springer.
55. Donoghue, J. F. (1994a). Leading correction to the newtonian potential. Physical Review
Letters, 72, 2996.
56. Donoghue, J. F. (1994b). General relativity as an effective field theory, the leading corrections.
Physical Review, D50, 3874–3888. (gr-qg/9405057).
57. Donoghue, J. F. (1997). In Fernando, C., & Herrero, M.-J. (Eds.), Advanced school on
effective theories. World Scientific, Spain. UMHEP - 424, gr-qc/9512024.
58. Duff, M. J. (1996). M-theory (The theory formerly known as superstrings). International
Journal of Modern Physics, A11, 5623–5642, hep–th/9608117.
59. Dyson, F. J. (1949a). The radiation theories of Tomonaga, Schwinger and Feynman. Physical
Review, 75, 486–502. Reprinted in Schwinger (1958a).
60. Dyson, F. J. (1949b). The S-Matrix in quantum electrodynamics. Physical Review, 75, 1736–
1755. Reprinted in Schwinger (1958a).
61. Dyson, F. J. (1967). Ground-state energy of a finite system of charged particles. Journal of
Mathematics and Physics, 8(8), 1538–1545.
62. Ehrenfest, P. (1959). Ansprache zur Verleihung der Lorentzmedaille an Professor Wolfgang
Pauli am 31 Oktober 1931. (Address on award of Lorentz medal to Professor Wolfgang Pauli
on 31 October 1931). In M. J. Klein (Ed.), Paul Ehrenfest: Collected scientific papers (p. 617).
Amsterdam: North-Holland. [The address appeared originally in P. Ehrenfest (1931). Versl.
Akad. Amsterdam, 40, 121–126.].
63. Englert, F., & Brout, R. (1964). Broken symmetry and the mass of gauge vector bosons.
64. Englert, F., Brout, R., & Thiry, M. F. (1966). Vector mesons in presence of broken symmetry.
Nuovo Cimento, A43, 244–257.
65. Epstein, H., & Glaser, V. (1973). The role of locality in perturbation theory. Annales de
l’Institute Henri Poincaré, A19, 211–295.
66. Faddeev, L. D., & Popov, V. N. (1967). Feynman diagrams for the Yang-Mills field. Physics
Letters, B25, 29–30.
67. Fayet, P. (1977). Spontaneously broken supersymmetric theories of weak, electromagnetic,
and strong interactions. Physics Letters, 69B, 489–494.
68. Fermi, E. (1932). Quantum theory of radiation. Reviews of Modern Physics, 4, 87–132.
69. Fermi, E. (1934a). Tentativo di una teoria dei raggi ˇ. Nuovo Cimento, 11, 1–19.
70. Fermi, E. (1934b). Versuch einer Theorie der ˇ - Strahlen. Zeitschrift fur Physik, 88, 161–171.
36 1 Introduction
71. Ferrara, S. (Ed.). (1987). Supersymmetry (Vol. 1 & 2). New York: Elsevier.
72. Feynman, R. P. (1963). Quantum theory of gravitation. Acta Phys. Polon., 24, 697–722.
73. Feynman, R. P. (1969a). The behavior of hadron collisions at extreme energies. In Proceed-
ings of the 3rd topical conference on high energy collisions, Stony Brook. New York: Gordon
& Breach.
74. Feynman, R. P. (1969b). Very high-energy collisions of hadrons. Physical Review Letters, 23,
1415–1417.
75. Feynman, R. P. (1972). The development of the space-time view of quantum electrodynamics.
In Nobel Lectures, Physics 1963–1970, 11 Dec 1965. Amsterdam: Elsevier.
76. Feynman, R. P. (1982). The theory of fundamental processes, 6th printing. The Menlo Park:
Benjamin/Cummings.
77. Feynman, R. P., & Gell-Mann, M. (1958). Theory of fermi interaction. Physical Review, 109,
193–198.
78. Feynman, R. P., & Hibbs, A. R. (1965). Quantum mechanics and path integrals. New York:
McGraw-Hill.
79. Figueroa, H., & Gracia-Bondia, J. M. (2001). On the antipode of Kreimer’s Hopf algebra.
Modern Physics Letters, A16, 1427–1434. hep–th/9912170v2.
80. Figueroa, H., & Gracia-Bondia, J. M. (2004). The uses of Connes and Kreimer’s algebraic
formulation of renormalization. International Journal of Modern Physics, A19, 2739–2754.
hep–th/0301015v2.
81. Figueroa, H., & Gracia-Bondia, J. M. (2005). Combinatorial Hopf algebras in quantum field
theory. I. Reviews in Mathematical Physics, 17, 881–961. hep–th/0408145v2.
82. Fitch, V. L. (1981). The discovery of charge conjugation parity asymmetry. Reviews of
Modern Physics, 53, 367–371.
83. Fock, V. (1933). C. R. Leningrad (N.S.) no. 6, pp. 267–271.
84. Freedman, D. Z., van Nieuwenhuizen, P., & Ferrara, S. (1976). Progress toward a theory of
supergravity. Physical Review, B13, 3214–3218.
85. French, J. B., & Weisskopf, V. F. (1949). The electromagnetic shift of energy levels. Physical
Review, 75, 1240–1248.
86. Friedman, J. I., & Telegdi, V. L. (1957). Nuclear emulsion evidence for parity nonconservation
in the decay chain C ! C ! eC . Physical Review, 105, 1681–1682.
87. Fritzsch, H., & Gell-Mann, M. (1972). Quatks and what else? In J. D. Jackson, & A. Roberts
(Eds.), Proceedings of the XVI International Conference on High Energy Physics (Vol. 2).
Chicago: Chicago University Press.
88. Fukuda, H., Miyamoto, Y., & Tomonaga, S. (1949a). A self consistent method in the quantum
field theory. II-1. Progress of Theoretical Physics, 4, 47–59.
89. Fukuda, H., Miyamoto, Y., & Tomonaga, S. (1949b). A self consistent method in the quantum
field theory. II-2. Progress of Theoretical Physics, 4, 121–129.
90. Furry, W. H., & Oppenheimer, J. R. (1934). On the theory of the electron and positive.
Physical Review, 45, 245–262.
91. Garwin, R. L. et al. (1957). Observations of the failure of conservation of parity and charge
conjugation in meson decays: The magnetic moment of the free muon. Physical Review, 105,
1415–1417.
92. Gell-Mann, M. (1972). Quarks. Acta Physica Austriaca Supplement IX, 9, 733–761.
93. Gell-Mann, M., & Low, F. E. (1954). Quantum electrodynamics at small distances. Physical
Review, 95, 1300–1312.
94. Georgi, H., Quinn, H. R., & Weinberg, S. (1974). Hierarchy of interactions in unified gauge
theories. Physical Review Lett, 33, 451–454.
95. Gershtein, S. S., & Zel’dovich, Y. B. (1956). On corrections from mesons to the theory of
ˇ-decay. Soviet Physics JETP, 2, 576. Original Russian version: Zhurnal Experimental’noi i
Teoreticheskoi Fiziki, 29, 698 (1955)
96. Glashow, S. L. (1961). Partial symmetries of weak interactions. Nuclear Physics, 22, 579–
588.
References 37
97. Glashow, S. L. (1980). Towards a unified theory: Threads in a tapestry. Reviews of Modern
Physics, 52, 539–543.
98. Gliozzi, F., Scherk, J., & Olive, D. (1976). Supergravity and the spinor dual model. Physics
Letters, 65B, 282–286.
99. Gliozzi, F., Scherk, J., & Olive, D. (1977). Supersymmetry, supergravity theories and the dual
Spinorl model. Nuclear Physics, B22, 253–290.
100. Gol’fand, A., & Likhtman, E. P. (1971). Extension of the Poincaré Group Generators and
Violation of P Invariance. JETP Letters, 13, 323–326. Reprinted in Ferrara (1987).
101. Gordon, W. (1926). Der Compton Effect nach der Schrödingerschen Theorie. Zeitschrift fur
Physik, 40, 117–133.
102. Green, M. B., & Schwarz, J. H. (1981). Supersymmetrical dual string theory. Nuclear Physics,
B181, 502–530.
103. Green, M. B., & Schwarz, J. H. (1982). Supersymmetrical string theories. Physics Letters,
109B, 444–448.
104. Greenberg, O. W. (1964). Spin and unitary spin independence in a paraquark model of baryons
and mesons. Physical Review Letters, 13, 598–602.
105. Greenberg, O. W., & Zwanziger, D. (1966). Saturation in triplet models of hadrons. Physical
Review, 150, 1177–1180.
106. Gross, D., & Wilczek, F. (1973). Ultraviolet behavior of non-Abelian gauge theories. Physical
107. Gross, D. J., Harvey, J. A., Martinec, E. J., & Rhom, R. (1985a). Heterotic string theory (I).
The free hetrotic string. Nuclear Physics, B256, 253–284.
108. Gross, D. J., Harvey, J. A., Martinec, E. J., Rohm, R. (1985b). The heterotic string. Physical
109. Gubser, S. S., Klebanov, I. R., & Polyakov, A. M. (1998). Gauge theory correlations from
non-critical string theory. Physics Letters, B428, 105–114. (hep-th/9802150).
110. Guralnik, G. S., Hagen, C. R., & Kibble, T. W. B. (1964). Global conservation laws and
massless particles. Physical Review Letters, 13, 585–587.
111. Han, M. Y., & Nambu, Y. (1965). Three-triplet model with double SU(3) symmetry. Physical
Review, 139, B1006–B1010.
112. Hasert, F. J. et al. (1973). Observation of neutrino-like interactions without muon-electron in
the Gargamellr neutrino experiment. Physics Letters, B 46, 138–140.
113. Hasert, F. J. et al. (1973). Search for elastic muon-neutrino electron scattering. Physics
Letters, B 46, 121–124.
114. Hawking, S. W. (1974). Black hole explosions? Nature, 248, 230–231.
115. Hawking, S. W. (1975). Particle creation by Black Holes. Communications in Mathematical
Physics , 43, 199–220.
116. Heisenberg, W., & Pauli, W. (1929). Zur Quantenelektrodynamic der Wellenfelder, I.
Zeitschrift fur Physik, 56, 1–61.
117. Heisenberg, W., & Pauli, W. (1930). Zur Quantenelektrodynamic der Wellenfelder, II.
Zeitschrift fur Physik, 59, 168–190.
118. Hepp, K. (1966). Proof of the Bogoliubov-Parasiuk theorem of renormalization. Communi-
cations in Mathematical Physics, 2, 301–326.
119. Higgs, P. W. (1964a). Broken symmetries, massles particles and gauge fields. Physics Letters,
12, 132–133.
120. Higgs, P. W. (1964b). Broken symmetries and the masses of Gauge Bosons. Physical Review
Letters, 13, 508–509.
121. Higgs, P. W. (1966). Spontaneous symmetry breaking without massless particles. Physical
Review, 145, 1156–1163.
122. Horowitz, G. T. (2005). Spacetime in string theory. New Journal of Physics, 7, 201 (1–13).
123. Horowitz, G., Lowe, D. A., & Maldacena, J. (1996). Statistical entropy of non-extremal four
dimensional black holes and U-duality. Physical Review Letters, 77, 430–433.
124. Howe, P. S., & Stelle, K. S. (2003). Supersymmetry counterterms revisited. Physics Letters,
B554, 190–196. hep–th/0211279v1.
38 1 Introduction
125. Jordan, P., & Pauli, W. (1928). Zur Quantenelektrodynamic Ladungfreier Felder. Zeitschrift
fur Physik, 47, 151–173.
126. Jordan, P., & Wigner, E. (1928). Über das Paulische Äquivalenzverbot. Zeitschrift fur Physik,
47, 631–651. Reprinted in Schwinger (1958a).
127. Kibble, T. W. B. (1968). Symmetry breaking in non-abelian gauge theories. Physical Review,
155, 1554–1561.
128. Klein, O. (1926). Quantentheorie und Fünfdimensionale Relativitätstheorie. Zeitschrift fur
Physik, 37, 895–906.
129. Klein, O. (1948). Mesons and nuclei. Nature, 161, 897–899.
130. Kramers, H. A. (1948). Non-relativistic quantum-electrodynamics and correspondence prin-
ciple. In Rapport et Discussions du 8e Conseil de Physique Solvay 1948 (pp. 241–265). R.
Stoop, Bruxelles, 1950.
131. Kreimer, D. (1999). On the Hopf algebra structure of perturbative quantum field theory.
Advances in Theoretical and Mathematical Physics, 2, 303–334.
132. Kreimer, D. (2003). New mathematical structures in renormalizable quantum field theories.
Annals of Physics, 303, 179–202.
133. Kroll, N. M., & Lamb, W. E. (1949). On the self-energy of a bound electron. Physical Review,
75, 388–398. Reprinted in Schwinger (1958a).
134. Lamb, W. E., Jr., & Retherford, R. C. (1947). Fine structure of the hydrogen atom by a
microwave method. Zeitschrift fur Physik, 72, 241–243. Reprinted in Schwinger (1958a).
135. Landsman, N. P. (1989). Large-mass and high-temperature behaviour in perturbative quantum
field theory. Communications in Mathematical Physics, 125, 643–660.
136. Lattes, C. M. G. et al. (1947). Processes involving charged mesons. Nature, 159, 694–697.
137. Lederman, L., & Teresi, D. (2006). The god particle: If the universe is the answer, what is the
question? New York: Mariner Books.
138. Lee, T. D., & Yang, C. N. (1956). Question of parity conservation in weak interactions.
Physical Review, 104, 254–258. See also ibid. 106, 1671 (1957).
139. Lee, B., & Zinn-Justin, J. (1972a). Spontaneously Broken Gauge symmetries. I. Preliminaries.
Physical Review, D5, 3121–3137.
140. Lee, B., & Zinn-Justin, J. (1972b). Spontaneously Broken Gauge symmetries. II. Perturbation
theory and renormalization. Physical Review, D5, 3137–3155.
141. Lee, B., & Zinn-Justin, J. (1972c). Spontaneously Broken Gauge symmetries. III. Equiva-
lence. Physical Review, D5, 3155–3160.
142. Lee, B., & Zinn-Justin, J. (1973). Spontaneously Broken Gauge symmetries. IV. general
Gauge formulation. Physical Review, D7, 1049–1056.
143. Lieb, E. H. (1979). The N 5=3 law for bosons. Physics Letters, A70(2), 71–73. Reprinted in
Thirring (2005).
144. Lieb, E. H., & Thirring, W. E. (1975). Bound for the kinetic energy of fermions which proves
the stability of matter. Physical Review Letters, 35(16), 687–689. [Erratum ibid., 35(16), 1116
(1975).] Reprinted in Thirring (2005).
145. Lovelace, C., & Squires, E. (1970). Veneziano theory. Proceedings of the Royal Society of
London, A318, 321–353.
146. Maldacena, J. (1998). The large N limit of superconformal theories and gravitation. Advances
in Theoretical and Mathematical Physics, 2, 231–252. (hep-th/9711200).
147. Manoukian, E. B. (1976). Generalization and improvement of the Dyson-Salam renormaliza-
tion scheme and equivalence with other schemes. Physical Review, D14, 966–971. ibid., 2202
(E).
148. Manoukian, E. B. (1979). Subtractions vs counterterms. Nuovo Cimento, 53A, 345–358.
149. Manoukian, E. B. (1983a). Renormalization. New York/London/Paris: Academic.
150. Manoukian, E. B. (1986a). Action principle and quantization of gauge fields. Physical Review,
D34, 3739–3749.
151. Manoukian, E. B. (2006). Quantum theory: A wide spectrum. Dordrecht: Springer.
152. Manoukian, E. B. (2012a). All the fundamental massless bosonic fields in bosonic string
theory. Fortschritte der Physik, 60, 329–336.
References 39
153. Manoukian, E. B. (2012b). All the fundamental bosonic massless fields in superstring theory.
Fortschritte der Physik, 60, 337–344.
154. Manoukian, E. B. (2012c). All the fundamental massless fermion fields in supersring theory:
A rigorous analysis. Journal of Modern Physics, 3, 1027–1030.
155. Manoukian, E. B. (2012d). The explicit pure vector superfield in Gauge theories. Journal of
Modern Physics, 3, 682–685.
156. Manoukian, E. B., & Muthaporn, C. (2003). N 5=3 Law for bosons for arbitrary large N.
Progress of Theoretical Physics, 110(2), 385–391.
157. Manoukian, E. B., & Sirininlakul, S. (2005). High-density limit and inflation of matter.
Physical Review Letters, 95, 190402: 1–3.
158. Marciano, W., & Pagels, H. (1978). Quantum chromodynamics. Physics Reports, C36, 137–
276.
159. Martin, P. C., & Glashow, S. L. (2008). Julian Schwinger 1918–1994: A biographical memoir.
National Academy of Sciences, Washington, DC, Copyright 2008.
160. Meissner, K. (2004). Black-Hole entropy in loop quantum gravity. Classical Quantum
Gravity, 21, 5245–5251.
161. Mohapatra, R. N. (1971). Feynman rules for the Yang-Mills field: A canonical quantization
approach. I, II. Physical Review, D4, 378–392, 1007–1017.
162. Mohapatra, R. N. (1972). Feynman rules for the Yang-Mills field: A canonical quantization
approach. III. Physical Review, D4, 2215–2220.
163. Nambu, Y. (1966). A systematic of hadrons in subnuclear physics. In A. de Shalit, H.
Feshback, & L. van Hove (Eds.), Preludes in theoretical physics in Honor of V. F. Weisskopf
(p. 133). Amsterdam: North-Holland.
164. Nambu, Y. (1969). In: Proceedings of the Internatinal Conference on Symmetries and Quark
Models (p. 269), Wayne State University. New York: Gordon and Breach, 1970.
165. Neveu, A., & Schwarz, J. H. (1971a). Factorizable dual model of pions. Nuclear Physics,
B31, 86–112.
166. Nielsen, H. (1970). Internatinal Conference on High Energy Physics, Kiev Conference, Kiev.
167. Olive, K. A. et al. (2014). Particle data group. Chinese Physics C, 38, 090001.
168. Oppenheimer, J. R. (1930). Note on the theory of the interaction of field and matter. Physical
Review, 35, 461–477.
169. Ovsyannikov, L. V. (1956). General solution to renormalization group equations. Doklady
Akademii Nauk SSSR, 109, 1112–1115.
170. Pauli, W., & Weisskopf, V. (1934). Über die Quantisierung der Skalaren Relativistischen
Wellengleichung. Helvetica Physica Acta, 7, 709–731.
171. Penrose, R. (1971). In T. Bastin (Ed.), Quantum theory and beyond (pp. 151–180).
Cambridge: Cambridge University Press.
172. Politzer, H. D. (1973). Reliable perturbative results for strong interactions. Physical Review
Letters, 30, 1346–1349.
173. Raymond, P. (1971). Dual theory for free fermions. Physical Review, D3, 2415–2418.
174. Rovelli, C., & Smolin, L. (1995). Discreteness of area and volume in quantum gravity.
Nuclear Physics, B442, 593–622.
175. Rovelli, C., & Vidotto, F. (2015). Covariant loop quantum gravity. Cambridge: Cambridge
University Press.
176. Rubbia, C. (1984). Experimental observation of the intermediate vecor bosons, W C , W and
Z 0 . Nobel Lecture, 8 December (pp. 240–287).
177. Sakharov, A. D. (1967). Violation of CP invariance, C asymmetry, and Baryon asymmetry of
the universe. Soviet Physics – JETP Letters, 5, 24–27.
178. Sakurai, J. J. (1958). Mass reversal and weak interactions. Nuovo Cimento, 7, 649–660.
179. Salam, A. (1951a). Overlapping divergences and the S-matrix. Physical Review, 82, 217–227.
180. Salam, A. (1951b). Divergent integrals in renormalizable field theories. Physical Review, 84,
426–431.
181. Salam, A. (1962). Renormalizability of gauge theories. Phys. Rev, 127, 331–334.
40 1 Introduction
182. Salam, A. (1968). Weak and electromagnetic interactions. In N. Svartholm (Ed.), Elementary
Particle Theory, Proceedings of the 8th Nobel Symposium, Almqvist and Wiksell, Stockholm.
183. Salam, A. (1980). Grand unification and fundamental forces. Reviews of Modern Physics, 52,
525–538.
184. Salam, A., & Strathdee, J. (1974a). Supersymmetry and non-abelian gauges. Physics Letters,
51B, 353–355. Reprinted in Ferrara (1987).
185. Salam, A., & Strathdee, J. (1974b). Supergauge Transformations. Nuclear Physics, B76, 477–
482.
186. Salam, A., & Strathdee, S. (1975b). Feynman rules for superfields. Nuclear Physics, B86,
142–152.
187. Salam, A., & Ward, J. (1959). Weak and electromagnetic interactions. Nuovo Cimento, 11,
568–577.
188. Salam, A., & Ward, J. (1961). On a gauge theory of elementary interactions. Nuovo Cimento,
19, 165–170.
189. Salam, A., & Ward, J. (1964). Electromagnetic and weak interactions. Physics Letters, 13,
168–170.
190. Scherk, J., & Schwarz, J. H. (1974). Dual models for non-hadrons. Nuclear Physics, B81,
118–144.
191. Schwarz, J. H. (1997). Lectures on superstrings and m-theory. Nuclear Physics Supp., B55,
1–32, hep–th/9607201.
192. Schwinger, J. (1948). On Quantum-electrodynamics and the magnetic moment of the electron.
Physical Review, 73, 416.
193. Schwinger, J. (1951). On gauge invariance and vacuum polarization. Physical Review, 82,
664–679.
194. Schwinger, J. (1951a). On the Green’s functions of quantized fields. I. Proceedings of the
National academy of Sciences of the United States of America, 37, 452–455.
195. Schwinger, J. (1951b). The theory of quantized fields. I. Physical Review, 82, 914–927.
196. Schwinger, J. (1953). The theory of quantized fields. II, III. Physical Review, 91, 713–728,
728–740.
197. Schwinger, J. (1954). The theory of quantized fields. V. Physical Review, 93, 615–628.
198. Schwinger, J. (Editor) (1958a). Selected Papers on Quantum Electrodynamics. New York:
Dover.
199. Schwinger, J. (1958b). A theory of fundamental interactions. Annals of Physics, 2, 404–434.
200. Schwinger, J. (1962a). Gauge invariance and mass. II. Physical Review, 128, 2425–2429.
201. Schwinger, J. (1972). Relativistic quantum field theory. In Nobel Lectures, Physics 1963–
1970, 11 Dec 1965. Amsterdam: Elsevier.
202. Schwinger, J. (1973a). A report on quantum electrodynamics. In J. Mehra (Ed.), The
physicist’s conception of nature. Dordrecht-Holland: D. Reidel Publishing Company.
203. Shaw, R. (1955). The problem of particle types and other contributions to the theory of
elementary particles. Ph.D. Thesis, Cambridge University.
204. Stelle, K. S. (1977). Renormalization of higher-derivative quantum gravity. Physical Review,
D16, 953–969.
205. Stelle, K. S. (2001). Revisiting supergravity and super Yang-Mills renormalization. In J.
Lukierski & J. Rembielinski (Eds.), Proceedings of the 37th Karpacz Winter School of
Theoretical Physics, Feb 2001. hep-th/0203015v1.
206. Stelle, K. S. (2012). String theory, unification and quantum gravity. In 6th Aegean Summer
School, “Quantum Gravity and Quantum Cosmology”, 12–17 Sept 2011, Chora, Naxos
Island. hep-th/1203.4689v1.
207. Streater, R. F. (1985). Review of renormalization by E. B. Manoukian. Bulletin of London
Mathematical Society, 17, 509–510.
208. Strominger, A., & Fava, G. (1996). Microscopic origin of the ‘Bekenstein-Hawking Entropy’.
209. Stueckelberg, E. C. G., & Peterman, A. (1953). La Normalisation des Constantes dans la
Théorie des Quanta. Helvetica Physica Acta, 26, 499–520.
References 41
210. Sudarshan, E. C. G., & Marshak, R. (1958). Chirality invariance and the universal fermi
interaction. Physical Review, 109, 1860–1862.
211. Susskind, L. (1970). Dual symmetric theory of hadrons. I. Nuovo Cimento, 69A, 457–496.
212. Susskind, L. (1995). The world as a hologram. Journal of Mathematics and Physics, 36,
6377–6396.
213. Symanzik, K. (1970). Small distance behaviour in field theory and power counting. Commu-
nications in Mathematical Physics, 18, 227–246.
214. Symanzik, K. (1971). Small distance behavior in field theory. In G. Höhler (Ed.), Springer
tracts in modern physics (Vol. 57). New York: Springer.
215. symanzik, k. (1971). small distance behaviour analysis in field theory and Wilson expansions.
Communications in Mathematical Physics, 23, 49–86.
216. ’t Hooft, G. (1971a). Renormalizable of massless Yang-Mills fields. Nuclear Physics, B33,
173–199.
217. ’t Hooft, G. (1971b). Renormalizable Lagrangians for massive Yang-Mills fields. Nuclear
Physics, B35, 167–188.
218. ’t Hooft, G. (1973). Dimensional regularization and the renormalization group. Nuclear
Physics, B61, 455–468.
219. ’t Hooft, G. (1987). Can spacetime be probed below string size? Physics Letters, B198, 61–63.
220. ’t Hooft, G. (1995). Black holes and the dimensional reduction of spacetime. In Lindström,
U. (Ed.), The Oskar Klein Centenary, 19–21 Sept 1994, Stockholm, (1994), World Scientific.
See also: “Dimensional Reduction in Quantum gravity”, Utrecht preprint THU-93/26 (gr-
qg/9310026).
221. ’t Hooft, G., & Veltman, M. J. G. (1972). Regularization and Renormalization of gauge fields.
222. Thirring, W. E. (Ed.). (2005). The stability of matter: From atoms to stars: Selecta of Elliott
H. Lieb (4th ed.). Berlin: Springer.
223. Thorn, C. B. (1991). Reformulating string theory with the 1=N expansion. In International A.
D. Sakharov Conference on Physics, Moscow (pp. 447–454). (hep-th/9405069).
224. Tomboulis, E. T. (1984). Unitarity in higher-derivative quantum gravity. Physical Review
Letters, 52, 1173–1176.
225. Tomonaga, S. (1972). Development of quantum electrodynamics: Personal recollections. In
Nobel Lectures, Physics 1963–1970, 6 May 1966, Amsterdam: Elsevier Publishing Company.
226. Tomonaga, S.-I. (1997). The story of spin. Chicago: University of Chicago Press. Translated
by T. Oka.
227. Townsend, P. K. (1995). The eleven-dimensional supermembrane revisited. Physics Letters,
B350, 184–187, hep–th/9501068.
228. Utiyama, R., & Sakamoto, J. (1976). Canonical quantization of non-abelian gauge fields.
Progress of Theoretical Physics, 55, 1631–1648.
229. Vanyashin, V. S., & Terentyev, M. V. (1965). Soviet Physics JETP, 21, 375–380. Original
Russian Version: Zhurnal Experimental’noi i Teoreticheskoi Fiziki, 48, 565 (1965)
230. Veltman, M. J. G. (1981). The infrared-ultraviolet connection. Acta Physica Polonica, B12,
437.
231. Veneziano, G. (1968). Construction of a crossing-symmetric regge-behaved amplitude for
linearly rising trajectories. Nuovo Cimento, 57A, 190–197.
232. Volkov, D. V., & Akulov, V. P. (1973). Is the neutrino a goldstone particle. Physics Letters,
46B, 109–110. Reprinted in Ferrara (1987).
233. Waller, I. (1930a). Die Streuung von Strahlung Durch Gebundene und Freie Elektronen Nach
der Diracshen Relativistischen Mechanik. Zeitschrift fur Physik, 61, 837–851.
234. Waller, I. (1930b). Bemerkung über die Rolle der Eigenenergie des Elektrons in der
Quantentheorie der Strahlung. Zeitschrift fur Physik, 62, 673–676.
235. Weinberg, S. (1960). High-energy behavior in quantum field theory. Physical Review, 118,
838–849.
236. Weinberg, S. (1967). A model of leptons. Physical Review Letters, 19, 1264–1266.
42 1 Introduction
237. Weinberg, S. (1973). New approach to renormalization group. Physical Review, D8, 3497–
3509.
238. Weinberg, S. (1980). Conceptual foundations of the unified theory of weak and electromag-
netic interactions. Reviews of Modern Physics, 52, 515–523.
239. Weisskopf, V. F. (1934). Über die Selbstenergie des Elektrons. Zeitschrift fur Physik, 89, 27–
39.
240. Weisskopf, V. F. (1939). On the self-energy of the electromagnetic field of the electron.
Physical Review, 56, 72–85.
241. Weisskopf, V. F. (1979). Personal impressions of recent trends in particle physics. CERN
Ref.Th. 2732, CERN, Geneva, Aug 1979. [Also presented at the 17 International School
of Subnuclear Physics “Ettore Majorana”, Erice, 31 July-11 August, 1979. pp.1–9, (ed. A.
Zichichi), Plenum, New York,1982].
242. Weisskopf, V. F. (1980). Growing up with field theory, and recent trends in particle physics.
“The 1979 Bernard Gregory Lectures at CERN”, 29 pages, CERN, Geneva.
243. Wess, J., & Zumino, B. (1974a). Supergauge transformations in four dimensions. Nuclear
Physics, B70, 39–50. Reprinted in Ferrara (1987).
244. Witten, E. (1995). String theory dynamics in various dimensions. Nuclear Physics, B443,
85–126, hep–th/9503124.
245. Witten, E. (1998). Anti-de-Sitter space and holography. Advances in Theoretical and Mathe-
matical Physics, 2, 253–291. (hep-th/9802150).
246. Wu, C. S. et al. (1957). Experimental tests of parity conservation in beta decay. Physical
Review, 105, 1413–1415.
247. Yang, C. N., & Mills, R. L. (1954). Conservation of isotopic spin and isotopic gauge
invariance. Physical Review, 96, 191–195.
248. Yoneya, T. (1974). Connection of dual models to electrodynamics and gravidynamics.
Progress of Theoretical Physics, 51, 1907–1920.
249. Yukawa, H. (1935). On the interaction of elementary particles. I. Proceedings of the Physical
Mathematical Society Japan, 17, 48–57. Reprinted in “Foundations of Nuclear Physics”, (ed.
R. T. Beyer), Dover Publications, Inc. New York (1949).
250. Zeidler, E. (2009). Quantum field theory II: Quantum electrodynamics. Berlin: Springer.
251. Zimmermann, W. (1969). Convergence of Bogoliubov’s method of renormalization in
momentum space. Communications in Mathematical Physics, 15, 208–234.
Recommended Reading
1. Becker, K, Becker, M., & Schwarz, J. H. (2006). String theory and M-theory: A modern
introduction. Cambridge: Cambridge University Press.
2. Davies, P. (Ed.). (1989). The new physics. Cambridge: Cambridge University Press.
3. DeWitt, B. (2014). The global approach to quantum field theory. Oxford: Oxford University
Press.
4. Manoukian, E. B. (1983). Renormalization. New York/London/Paris: Academic Press.
6. Manoukian, E. B. (2016). Quantum field theory II: Introductions to quantum gravity, super-
symmetry, and string theory. Dordrecht: Springer.
7. Martin, P. C., & Glashow, S. L. (2008). Julian Schwinger 1918–1994”: A Bio-graphical
memoir. National Academy of Sciences, Washington, DC, Copyright 2008.
8. Oriti, D. (Ed.), (2009). Approaches to quantum gravity. Cambridge: Cambridge University
Press.
9. Schweber, S. S. (2008). Quantum field theory: From QED to the standard model In M. Jo
Nye (Ed.), The Cambridge history of science. The modern physical and mathematical sciences
(Vol. 5, pp. 375–393). Cambridge: Cambridge University Press.
References 43
10. Streater, R. F. (1985). Review of Renormalization by E. B. Manoukian. Bulletin of London

Mathematical Society, 17, 509–510.
11. Weinberg. S. (2000). The quantum theory of fields. III: supersymmetry. Cambridge: Cambridge
University Press.
12. Weisskopf, V. F. (1980). Growing up with field theory, and recent trends in particle physics. The
1979 Bernard Gregory Lectures at CERN, 29 pages. “Personal Impressions of Recent Trends
in Particle Physics”. CERN Ref. Th. 2732 (1979). CERN, Geneva.
pp. 972–975.
Chapter 2
Preliminaries
This preliminary chapter deals with basic tools needed for dealing with quantum
field theory. It begins with Wigner’s theory of symmetry transformations by showing
how symmetry is implemented in the quantum world via unitary or anti-unitary
operators. This subject matter will be also important in deriving commutation
(anti-commutation) rules between symmetry generators such as in developing the
Poincaré algebra of spacetime transformations including those in supersymmetric
field theories which also involve fermion-boson exchanges. Some properties of
the Dirac equation and related aspects are summarized in Appendix I, at the end
of the book, for the convenience of the reader. Various representations of the
gamma matrices are spelled out, however, in the present chapter which turn out
to be important in modern field theory, and the concept of a Majorana spinor is
also introduced. Special emphasis is put in the remaining sections on functional
differentiations and functional integrations involving Grassmann variables as well.
Functional Fourier transforms are introduced which clearly show the intimate
connection that there exists between the functional differential formalism pioneered
by Julian Schwinger and the functional integral formalism pioneered by Richard
Feynman. The last section on delta functionals, makes this connection even more
transparent and shows the simplicity of the functional formalism in general, whether
it is in differential or integral forms. It is, however, relatively easier to functionally
differentiate than to deal with continual functional integrals. Both formalisms are
used in this book for greater flexibility and for a better understanding of quantum
field theory.

DOI 10.1007/978-3-319-30939-2_2
46 2 Preliminaries
2.1 Wigner’s Symmetry Transformations in the Quantum

World
Invariance of physical laws under some given transformations leads to conservation

laws and the underlying transformations are referred to as symmetry transforma-
tions. For example, invariance under time translation (e.g., by setting one’s clocks
back by a certain amount) or under space translation (by shifting the origin of
one’s coordinate system) lead, respectively, to energy and momentum conservations.
Invoking invariance properties in developing a dynamical theory, conveniently
narrows down one’s choices in providing the final stages of a theory. Obviously, not
all transformations of a given physical system are symmetry transformations. But
invoking the invariance of a system under a set of given transformations may provide
the starting point in describing a dynamical theory, and one may then consistently
modify the theory to take into account any symmetry breaking. One may also have
symmetry breaking spontaneously which will be discussed later.
A celebrated analysis of Eugene Wigner in the thirties1 originating on symmetry
under rotations in space, spells out the nature of the transformations implemented on
elements of a Hilbert space. This result is of central importance for doing quantum
physics2 and quantum field theory.
The physical situation that presents itself is the following. One prepares a system
in a state j i. The question then arises as to what is the probability of finding the
system in some state ji if j i is what one initially has?.3 The latter is given by
jh j ij2 .
If j 0 i, j 0 i denote the states j i, ji, resulting from a symmetry transforma-
tion, then invariance of the above probability under such a transformation means
that
˝ ˇ ˛
j 0ˇ 0 2
j D jh j ij2 ; (2.1.1)
and fj 0 i ; j 0 ig provide an equivalent description as fj i ; jig. One may scale

j 0 i ; j 0 i ; j i ; ji by arbitrary phase factors without changing the relevant
probabilities given in (2.1.1). Although such overall phase factors are not important,
the relative phases occurring when considering addition of such states are physically
relevant with far reaching consequences.4
1
See, e.g., his book Wigner [3]. See also Wigner [4].
2
For intricacies of Wigner’s Symmetry Transformations in quantum mechanics, see, e.g.,
Manoukian [1], pp. 55–65.
3
More generally one may also have initially a mixture described by a density operator.
4
One should also consider unit rays in this discussion and in the subsequent definitions but we will
not go into these points here.
2.1 Wigner’s Symmetry Transformations in the Quantum World 47
To understand Wigner’s Theorem, one needs to define the following operators:

1. An operator L is called linear or anti-linear, if for any j i; ji; Œ a j i C b ji ,
where a; b are, in general, complex numbers, then
L Œ a j i C b ji D a L j i C b L ji ; (2.1.2)
or
L Œ a j i C b ji D a L j i C b L ji ; (2.1.3)
respectively, with * denoting complex conjugation.

2. A linear or anti-linear operator U is called unitary or anti-unitary, if
hU j U i D h j i ; (2.1.4)
or
hU j U i D h j i D h j i ; (2.1.5)
respectively.
2.1.1 Wigner’s Symmetry Transformations
Under a symmetry transformation, there exists a unitary or anti-unitary operator U

such that (2.1.2)/(2.1.4) or (2.1.3)/(2.1.5) hold true, respectively, with
ˇ ˛ ˇ ˛
ˇ 0
D U j i ; ˇ 0 D U ji ; (2.1.6)
ˇ ˛
ˇŒ a C b 0 D U Œa j i C b ji: (2.1.7)
Continuous transformations, for which one may also consider infinitesimal ones
close to the identity operation, such as space or time translations and rotations, are
implemented by unitary operators since the identity operator I itself is trivially a
unitary one. An infinitesimal change of a parameter •, under consideration, is
implemented by a given operator G, referred to as the generator of symmetry
transformation, and the corresponding unitary operator, for such an infinitesimal
transformation, would take the form
U D I C i •G: (2.1.8)
48 2 Preliminaries
The unitarity condition U U D I D UU implies that • G is a self-adjoint

operator5:
.• G/ D • G: (2.1.9)
For example, for an infinitesimal space translation •x, via the momentum operator
P associated with the system in consideration, the corresponding unitary operator
is given by
Ujspace D I C i •x P: (2.1.10)
For infinitesimal time translation •, via the Hamiltonian H of the system, one has
Ujtime D I i •H: (2.1.11)
Moreover for a rotation of a coordinate system by an infinitesimal angle • about a

unit vector n, via the angular operator J,
Ujrotation D I C i • n J: (2.1.12)
The latter may be more conveniently rewritten by introducing parameters •! ij , and

rewrite J i as
1 ij k j k
Ji D " J ; •! ij D " ij k n k •; i; j; k D 1; 2; 3; (2.1.13)
2
where a summation over repeated indices is assumed, and " ij k is totally anti-
symmetric, with "123 D C1. We may then rewrite (2.1.12) as
i
Ujrotation D I C •! ij J ij : (2.1.14)
2
The generators H, P, and J of the transformations are self-adjoint.
In a relativistic setting, in which this book is based, these spacetime transfor-
mations may be combined by introducing, in the process, additional operators to
this set of operators, denoted by J 0i , i D 1; 2; 3, which impart a frame with
an infinitesimal velocity change •v. The unitary operator corresponding to these
infinitesimal spacetime transformations then takes the elegant form
1
U D 1 C i •b P C •! J ; •! D •! ; ; D 0; 1; 2; 3;
2
(2.1.15)
5
One may conveniently, in general, absorb • in G.
2.1 Wigner’s Symmetry Transformations in the Quantum World 49
with6 P 0 D H,
•b D .•; •x/; •! ij D " ij k nk • ; •! 0 i D •v i ; •! 00 D 0: (2.1.16)
First we have to learn how the labels attached to an event in different, so-
called, inertial frames are related. This will be taken up in the next section. The
transformation rules relating this different labeling, are referred to as Lorentz
transformations in Minkowski spacetime. On the other hand, the genertors P ; J
satisfy basic commutations relations and form an algebra in the sense that the
commutator of two generators is equal to a linear combination of the generators in
question. The underlying algebra is called the Poincaré algebra and will be studied
in Sect. 4.2.
In incorporating supersymmetry into physics, with an inherited symmetry exist-
ing between fermions and bosons, one introduces a generator7 of fermionic type,
that is, it carries a spinor index, referred to as a supercharge operator, and the
Poncaré algebra is extended to a larger one including anti-commutation relations
as well. This algebra is referred to as Super-Poincaré algebra.8
Typical discrete transformations, are provided by space P or time T reflections
(parity and time reversal) or charge conjugation C, for which a particle is replaced
by its antiparticle. A key criterion for finding out if T, for example, should be
unitary or anti-unitary is to eliminate the choice which would lead to the inconsistent
result that a Hamiltonian is unbounded from below. This is inferred as follows.
Under an infinitesimal time translation •, via the Hamiltonian H of the system
under consideration, followed by a time-reversal applied to it is equivalent to a time
reversal followed by a time translation, i.e., we have the equality
Œ1 i •H T D T Œ1 C i •.H C '/; (2.1.17)
up to a phase factor '. If T were unitary, this gives H T D T .H C '/. For a

state ji with positive and arbitrary large energy E, there would correspond a state
T ji with energy .E C'/ and the Hamiltonian would be unbounded from below.
That is, T is to be implemented by an anti-unitary operator which, in the process
of applying time reversal, it would complex conjugate the i factor on the right-
hand side of (2.1.17). A similar analysis applied to P leads one to infer that it is to
be implemented by a unitary operator. On the other hand, C, unlike P, and T, is
not involved with space and time reflections, and may be implemented by a unitary
operator. Accordingly, the product CPT, in turn, is implemented by an anti-unitary
6
With the Minkowski metric adopted in this book Œ D diagŒ1; 1; 1; 1, bi D bi ; b0 D b0 .
7
One may also introduce more than one such operator.
8
This will be studied in the accompanying book Manoukian [2]: Quantum Field Theory II:
Introductions to Quantum Gravity, Supersymmetry, and string Theory. Minkowski spacetime is,
in turn, extended to what has been called superspace.
50 2 Preliminaries
operator. The corresponding symmetry, embodied in the so-called CPT Theorem,

will be dealt with in Chap. 4.9
2.2 Minkowski Spacetime: Common Arena of Elementary

Particles
A fundamental principle which goes to the heart of special relativity is the

equivalence of inertial frames in describing physical laws. By this it is meant, in
particular, that dynamical equations take the same form in such frames up to a mere
relabeling of the variables of the underlying theories. Because of such relabeling
of the variables, these equations are said to transform covariantly as one goes from
one inertial frame to another and the corresponding rules of transformations are, in
general, called Lorentz transformations.
An event labeled by x D .x 0 ; x/, where x 0 D t (time), in one frame, say, F will
be labeled by x 0 D .x 00 ; x 0 / in some other frame, say, F 0 . The transformation rules
which connect this different labeling of the same event are the so-called Lorentz
transformations. One inertial frame F 0 may be moving with a uniform velocity
with respect to another frame F, as determined in F, with a possible orientation of
the cartesian space coordinate axes of F 0 as also determined in F at some initial
time x 0 D 0. If an x D 0 reading in F corresponds to an x 0 D 0 reading in
F 0 , then the Lorentz transformations are called homogeneous ones. Otherwise, they
are called inhomogeneous. In the former case, the origins of the space coordinate
axes set up in F and F 0 coincide at time readings x 0 D 0, x 0 0 D 0 by observers
located at the corresponding origins of these respective coordinate systems.
We use the notation .x / D .x 0 ; x/ x, D 0; 1; 2; 3, and our Minkowski
metric is defined by Œ D diag Œ 1; 1; 1; 1 , Œ D diag Œ 1; 1; 1; 1 .
Authors who have also used this signature for the metric include, Julian Schwinger,
Steven Weinberg,. . . . Others may use the minus of ours. Surprisingly, Paul Dirac
has used both signatures at different times of his career. It doesn’t matter which
one to use. It is easy to keep track of the relative minus sign. In our notation,
x 0 D x 0 , x i D x i for i D 1; 2; 3, x D x , x D x , Œ D
Œ • D diag Œ1; 1; 1; 1.
Under a homogeneous Lorentz transformation,
x 0 D x ; (2.2.1)
9
The reader may wish to consult Manoukian [1], pp. 55–65, 112–115, where additional details,
and proofs, are spelled out in quantum theory.
2.2 Minkowski Spacetime: Common Arena of Elementary Particles 51
where is independent of x, x 0 . Before we spell out the structure of the matrix

. /, we note that
@x 0
D ; (2.2.2)
@x
and hence from the chain rule
@
@ D @ 0 ; @ : (2.2.3)
@x
Quite generally, we have the Lorentz invariant property
.x 0 y 0 / .x 0 y 0 / D .x y / .x y /; (2.2.4)
from which one may infer that
D ; .1 / D ; (2.2.5)

@ 0 D @ ; D 0; @ @ : (2.2.6)
The Poincaré (inhomogeneous Lorentz) transformations in Minkowski spacetime

x ! x 0 are defined by
x 0 D x b : (2.2.7)
Consistency of such a transformation requires that its structure remains the same
under subsequent transformations leading to group properties spelled out below. For
a subsequent transformation x 0 ! x 00 , we may use (2.2.7) to write
x 00 D 0 x 0 b 0 ; (2.2.8)
or

x 00 D 0 x 0 b C b 0 : (2.2.9)
Thus a Poincaré transformation may be specified by a pair .; b/, with D

. /, b D .b /, satisfying the group properties:
1. Group multiplication: .0 ; b 0 /.; b/ D .0 ; 0 b C b 0 /.
2. Identity .I; 0/: .I; 0/.; b/ D .; b/.
3. Inverse .; b/1 D .1 ; 1 b/: .; b/1 .; b/ D .I; 0/.
4. Associativity Rule: .3 ; b3 /Œ.2 ; b2 /.1 ; b1 / D Œ.3 ; b3 /.2 ; b2 /.1 ; b1 /.
To spell out the general structure of the matrices , we first consider spatial
rotations of coordinate systems. A point x specified in a given 3D coordinate system
52 2 Preliminaries
will be read as x0 in a coordinate system obtained from the first by a c.c.w rotation
by an angle about a unit vector n,
x 0 i D R ij x j ; R ij R i k D • j k ; i; j; k D 1; 2; 3; (2.2.10)
with a summation over repeated indices understood. Here R i k are the matrix
elements of the rotation matrix with

R i k D •i k " ij k nj sin C •i k ni nk /.cos 1 ; (2.2.11)
" ij k is totally anti-symmetric with " 123 D C 1.

If for a given unit vector n, we use the notation x 0 D x./, then
from (2.2.10), (2.2.11), it is worth noting that
d
x./ D n x./; x.0/ D x: (2.2.12)
d
The general structure of the matrices , involving such rotations and so-
called Lorentz boosts is given by
ˇ k ˇj
i j D ij D R ij C .% 1/ R i k ; (2.2.13)
ˇ2
1=2
0 0 D 0 0 D % 1 ˇ 2 ; (2.2.14)
0 i D 0 i D % ˇ i ; (2.2.15)
i 0 D i 0 D % R ij ˇ j ; (2.2.16)
Using rather a standard notation for ˇ, its physical interpretation is as follows.

From the second identity in (2.2.5) it follows from the transformation law x 0 D
x in (2.2.1) that x D x 0 . The origin of the spacial coordinate system of
the frame F 0 is specified by x 0j D 0; j D 1; 2; 3. From this and (2.2.14), (2.2.15),
we have x 0 D % x 0 0 , x i D % ˇ i x 0 0 . That is, ˇ denotes the velocity with
which the origin of the spatial coordinate system of F 0 moves relative to the
corresponding origin of F. Now put a (massive m ¤ 0) particle at the origin
of the spatial coordinate system of F, the so-called rest frame of the particle, i.e.,
with energy-momentum p D .m; 0/. Then if no initial relative rotation of the
spatial coordinate systems of the frames is involved, the velocity of the particle in F 0
(the “laboratory” frame) would be simply ˇ, i.e., of momentum p 0 D m % ˇ.
With a relative rotation of the coordinate systems, the relation p 0j D j 0 m gives
p 0j D Rj k . m % ˇ k / showing simply a c.w. rotation of the earlier expression.
Equally important are the transformations (2.2.7), (2.2.13), (2.2.14), (2.2.15),
(2.2.16) for infinitesimal changes •b, •, •ˇ. To first order,
•x D x x 0 D •b •! x ; (2.2.17)
2.2 Minkowski Spacetime: Common Arena of Elementary Particles 53
where
D • C •! ; •! D •! ; (2.2.18)
•! i j D " ij k n k •; (2.2.19)

•! i 0 D •ˇ i ; (2.2.20)
•! 0 i D •ˇ i ; (2.2.21)
•! 0 0 D 0: (2.2.22)
and i; j D 1; 2; 3; ; D 0; 1; 2; 3.
When the parameters •b, •, •ˇ are led continuously to go to zero, the induced
transformations in the underlying vector space of particle states go over to the
identity. The latter being unitary, we may refer to Wigner’s Theorem (Sect. 2.1)
to infer that an induced transformation is represented by a unitary rather than by
an anti-unitary operator. Such transformations, in turn, generate an algebra of the
corresponding generators referred to as the Poincaré algebra. The corresponding
analysis will be carried out in Sect. 4.2.
Before closing this section, we note that the general structure in (2.2.5), rewritten
in matrix form reads
> D ; (2.2.23)
and since det D 1, we may infer that for the transformations in (2.2.1) det D
C1. It is not 1 as this transformation includes neither time nor space reflections.
From this and (2.2.2) we learn that the Jacobian of the transformation x 0 ! x is
given by det D 1, establishing the Lorentz invariance of the volume element in
Minkowski spacetime:
.dx/ D .dx 0 /; where .dx/ dx 0 dx 1 dx 2 dx 3 : (2.2.24)
This together with the definition of a Lorentz scalar ˚.x/ by the condition
˚ 0 .x 0 / D ˚.x/; (2.2.25)
under the above Lorentz transformations, guarantees the invariance of integrals of

the form
Z
A D .dx/ ˚.x/; (2.2.26)
such as the action integral, and lead to the development of Lorentz invariant theories.
54 2 Preliminaries
2.3 Representations of the Dirac Gamma Matrices;

Majorana Spinors
The Dirac formalism for the description of the relativistic electron in a quantum
setting via the Dirac equation . @ =i C m/ D 0, is summarized in Appendix I
at the end of the book. The Dirac representation of the gamma matrices, in
particular, are defined by

I 0 0 0I
0D ; D ; 5 i 0 1 2 3 D ; (2.3.1)
0 I 0 I 0
satisfying the anti-commutation relations: f ; g D 2 . Various representa-

tions of the gamma matrices, satisfying the same anti-commutation relations, may
be similarly defined. We here introduce two other representations which turn up to
be of quite importance in modern field theory. They arise as follows.
The unitary matrix

1 I I 1 1 I I
GD p ; G DG D p ; (2.3.2)
2 I I 2 I I
introduces via the transformation G G 1 the following representation of gamma

matrices, known as the chiral representation, in which 5 is diagonal

0 I 0 I 0
0D ; D ; 5D ; (2.3.3)
I 0 0 0 I
where for simplicity we have used the same notation for the resulting matrices.
Note that coincides with the one in the Dirac representation. The chiral
representation is important in dealing with massless particles and in the study of
supersymmetry. While the unitary matrix

1 I 2
G D G 1 D G D p ; (2.3.4)
2 2 I
leads to the so-called Majorana representation

0 0 2 5 2 0
D ; D ; (2.3.5)
2 0 0 2

i 3 0 0 2 i 1 0
1D ; 2
D ; 3
D ; (2.3.6)
0 i 3 2 0 0 i 1
2.3 Representations of the Dirac Gamma Matrices; Majorana Spinors 55

in which =i D =i , making the Dirac operator @ =i C m real. This
representation is particularly convenient in analyzing general results in field theory
such as the Spin & Statistics Connection .
If satisfies the Dirac equation in an external electromagnetic field e A .x/
>
of charge e then C D C , where D 0 , and C is the charge
2 0
conjugation matrix defined by C D i , satisfies the Dirac equation with
e A .x/ replaced by e A .x/ i.e., with the sign of the charge e reversed. (See
Eq. (I.3) in Appendix I at the end of the book).
In the Dirac and
chiral representations
the charge conjugation matrix are given by . 2 /> D 2

0 i 2 i 2 0
CDirac D ; Cchiral D : (2.3.7)
i 2 0 0 i 2
We note that 5 C ; 5 C , together with C , are anti-symmetric matrices.

Some properties of the charge conjugation matrix C D i 2 0 , in general, are:
(] stands for any of the operations .:/ ; .:/1 ; .:/> )
>
C \ D C ; C 1 C D . />; ŒC ; 5 D 0; C 1
5 C D 5 ;
(2.3.8)
>
C 1
Œ ; C D .Œ ; /> ; C 1 5
Œ ; C D 5 Œ ; :
(2.3.9)
Once a charge matrix C has been defined, one may define a Majorana spinor
, not to be confused with the Majorana representation of the gamma matrices, by
the condition D C , that is
>
DC ; D > C 1 D .C /> : (2.3.10)
From these definitions, one may infer the general structure of a Majorana spinor,
for example, in the chiral representation, to be of the form:
0 1 0 1
1 4
B 2 C B C
DB C B 3C
@ 3 A D @ 3 A : (2.3.11)
4 4
56 2 Preliminaries
2.4 Differentiation and Integration with Respect

to Grassmann Variables
Consider n real Grassmann variables 1 ; : : : n , i.e., variables that satisfy fi ; k g D

0, pairwise, which, in particular, implies that for any k: k2 D 0.
The left-hand derivative with respect to a variable j of any product i1 : : : ik is
defined by
@
i : : : ik D .1/ıj i1 : : :b
j : : : ik ; (2.4.1)
@j 1
where ıj denotes the position of j from the left 1, in the product i1 : : : ik , and
j is omitted on the right-hand side of the equation as indicated by the “hat” sign on
it. One may also define a right-hand derivative, taking a derivative from the right,
with a corresponding rule.
In particular note that,
@
i j D •i k j •j k i ; (2.4.2)
@k
and

@
; k D •j k : (2.4.3)
@j
Pn
One may define the differential operator d D kD1 dk @=@k satisfying the rule
d.i j / D di j C i dj : (2.4.4)
On the other hand by using the rule in (2.4.2) in applying the differential operator
d gives
d.i j / D di j dj i ; (2.4.5)
which upon comparison with (2.4.4) gives
dj i D i dj : (2.4.6)
To extend the above rules for differentiations with respect to complex variables,
one may proceed as follows. Given two real Grassmann variables R ; I satisfying
fR ; I g D 0, one may define a complex Grassmann variable D R C i I . Using
the notation for the complex conjugate of , the following anti-commutation
rules emerge
f ; g D 0; f ; g D 0; f ; g D 0: (2.4.7)
2.4 Differentiation and Integration with Respect to Grassmann Variables 57
Imposing a reality restriction on the product
D i I R ; (2.4.8)
i.e., . / D . /, leads to the following definition
.I R / D I R or .I R / D R I ; (2.4.9)
showing that complex conjugation of the product of two real Grassmann variables
reverses their order. This in turn implies that for complex (or real) Grassmann
variables 1 ; : : : ; n .
.1 : : : n / D n : : : 1 : (2.4.10)
One may also define two sets of derivatives with respect to variables i ; i , as
before, and note that
( )
@ @ @ @
; D 0; ; D 0: (2.4.11)
@i @j @i @j
We now consider integrations with respect to real Grassmann variables. Due to

the property 2i D 0, one has to investigate only the meanings of the following two
integrals, for a given i,
Z Z
di ; di i : (2.4.12)
Imposing translational invariance of these integrals under i ! i C ˛ i D i0 ,

with ˛ i anti-commuting with i , simply means that
Z Z Z Z
d 0i D di ; d 0i 0i D di i : (2.4.13)
In particular, translation invariance gives

Z Z Z
d 0i 0i D di i C di ˛i ; (2.4.14)
for arbitrary ˛ i . We may thus infer that

Z
di D 0: (2.4.15)
58 2 Preliminaries
The second integral in (2.4.12) may be normalized as

Z
di i D 1; (2.4.16)
for any i, a normalization condition that will be adopted in this book.

For any function f .i / D ˛ 0 C c1 i , where c1 is a c-number, and ˛ 0 anti-
commutes with i , one clearly has
Z
di .i ˇ i / f .i / D f .ˇ i /: (2.4.17)
This allows one to introduce a Dirac delta given by
•.i ˇ i / D .i ˇ i /; (2.4.18)
being a Grassmann variable itself. That is di •.i ˇ i / commutes with Grassmann
variables.
Before introducing integrations over complex Grassmann variables, consider
making a change of integration variables
i D Cij j ; (2.4.19)
in an integral, where a summation over j understood, and the Cij are c-numbers,
i; j D 1; : : : ; n. We are thus led to consider, in general, the integral
Z Z
d1 : : : dn 1 : : : n D d1 : : : d n J C 1i1 i1 : : : C n i n i n ; (2.4.20)
where J is the Jacobian of the transformation to be determined. Using the anti-

commutativity property of the j , we may write
i1 : : : in D " i1 ::: in 1 : : : n ; (2.4.21)
where " i1 ::: in is the Levi-Civita symbol totally anti-symmetric with " 1::: n D C1.
Using the definition of a determinant of the matrix Œ Cij
C1i1 : : : Cnin " i1 ::: in D det C; (2.4.22)
we immediately obtain
J D 1=det C; (2.4.23)
being the inverse of det C.

2.4 Differentiation and Integration with Respect to Grassmann Variables 59
As before, out of two real Grassmann variables R ; I , we may define a complex

one D R C i I . Now consider the transformation R ; I ! ; defined by

R 1 1 1
D ; (2.4.24)
I 2 i i
with the Jacobian of the transformation given by

1
1 1 1
J D det D 2 i: (2.4.25)
2 i i
Using the fact that I R D =2i, we may infer that

Z Z
1D dR dI I R D d d .2i/ . =2i/; (2.4.26)
or
Z
d d D 1: (2.4.27)
Hence for consistency with the above integral we may set

Z Z
d D 1; d D 1: (2.4.28)
For integrations over n complex Grassmann variables 1 ; : : : ; n , consider the

integral
Z
d1 d1 d dn
ID n exp Œ ii Aij j (2.4.29)
i i
where the Aij are c-numbers. The commutativity of i Ai1 1 , for example, with all
the Grassmann variables, and so on, allows us to rewrite the integral I as (since
.1 /2 D 0; : : : ; .n /2 D 0)
Z

ID d1 1 A1j1 j1 d1 dn n Anjn jn dn
Z
D d1 j1 dn j n A1j1 Anjn
Z
D d1 1 dn n "j1 :::j n A1j1 Anjn ; (2.4.30)
60 2 Preliminaries
where we have used (2.4.21). From (2.4.28) and the definition of a determi-
nant (2.4.22), the following expression for the integral emerges
Z
d1 d1 d dn
n exp Œ i i Aij j D det A: (2.4.31)
i i
For det A ¤ 0, the above integral may be used to give another useful integral. To
this end, we note the following. We may call the variables i in (2.4.31) anything
we like and use any notation for them. In the language of calculus they are dummy
variables. We may go even further by noting that we are free to make completely
different change of variables for the j and the former variables. Taking these two
points into account, we make the substitutions j ! j .A1 /j and i ! i
.A1 / i to obtain directly from (2.4.31)
Z
d1 d1 d dn
n expŒ i .i Aij j C i i C i i / D .det A/ exp Œ i i A1
ij j :
i i
(2.4.32)
This integral is to be compared with the corresponding one for integrations over
commuting complex variables
Z
dz1 dz1 dzn dz n 1
expŒ i . zi Aij z j C zi Ki C Ki z i / D exp i Ki A1
ij Kj :
2=i 2=i det A
(2.4.33)
For integrations over real commuting variables one also has

Z
dy1 dyn h 1 i 1 hi i
p p exp i yi Aij yj C Ki yi D p exp Ki A1 Kj :
2=i 2=i 2 det A 2 ij
(2.4.34)
2.5 Fourier Transforms Involving Grassmann Variables
The study of Fourier transform theory of Grassmann variables rests on the following
two integrals. The first one is
Z
d1 d 1 : : : dn dn exp i Œ . ˛/ C . ˛/
D .1 ˛1 /.1 ˛ 1 / : : : .n ˛n /.n ˛ n /; (2.5.1)
which, by now, should be straightforward to verify (see Problem

P 2.6). In the above
equation, we have used the standard notation .˛/ D j j .j ˛j /. The second
2.5 Fourier Transforms Involving Grassmann Variables 61
one is
Z
d1 d1 : : : dn dn .1 /1 .1 /"1 : : : .n /n .n /"n .1 ˛1 /
.1 ˛ 1 / : : : .n ˛n /.n ˛ n / D .˛ 1 /1 .˛1 /"1 : : : .˛ n /n .˛n /"n ; (2.5.2)
where the j ; "j are either 0 or 1, and a factor such as .0/0 , (e.g., when ˛j D
0; "j D 0), is to be replaced by 1 on the right-hand side of the equation.
The second one seems more difficult to verify but easily follows by first writing
./ D ı 0 C ı1 ; ./" D ı"0 C ı"1 : (2.5.3)
The integral on the left-hand side of (2.5.2) then becomes

Z
d1 .ı1 0 1 ı1 1 / .1 ˛1 / .1 ˛ 1 / d1 .ı"1 0 C ı"1 1 /
dn .ın 0 n ın 1 /.n ˛n /.n ˛ n / dn .ı"n 0 C ı"n 1 /: (2.5.4)
Therefore it remains to face the following integral which works out as follows:
Z
d .ı 0 ı1 /. ˛/. ˛/ d .ı"0 C ı"1 /
Z
D d .ı 0 ı1 /. ˛/ d . ˛/.ı"0 C ı"1 /
Z
D d .ı 0 C ˛ ı1 / d .ı"0 C ˛ ı"1 / D .˛/ .˛/" ; (2.5.5)
which leads from (2.5.4) to the expression on the right-hand side of (2.5.2). Now we
are ready to introduce Fourier transforms involving Grassmann variables.
Given a function FŒ; which is a linear combination of terms as
Œ.1 /1 .1 /"1 : : : .n /n .n /"n , we define its Fourier transform: .; / ! .; /
by
Z
Q ; D d1 d1 d dn
FŒ ::: n FŒ ; exp Œ i . C / : (2.5.6)
i i
To find the inverse Fourier transform: .; / ! .; /, we multiply the above
integral by
i d1 d1 : : : i dn dn exp Œ i. ˛ C ˛ / ;

62 2 Preliminaries
which commutes with everything, and integrate using (2.5.1) and then (2.5.2). The
right-hand becomes10
Z Z
d1 d1 : : : dn dn FŒ ; d1 d1 : : : dn dn exp i Œ . ˛/ C . ˛/
Z
D d1 d1 : : : dn dn FŒ ; .1 ˛1 /. 1 ˛ 1 / : : : .n ˛ n /.n ˛ n / D FŒ ˛; ˛ :
(2.5.7)
By equating this expression with the resulting one on the left-hand side of the
equation in question, the following expression emerges for the inverse Fourier
transform
Z
Q ; exp Œ i. C / ;
FŒ ; D i d1 d1 : : : i dn dn FŒ (2.5.8)
written in terms of variables .; /.

For integration over commuting complex variables, with z D .z 1 ; : : : ; z n /> , we
define a Fourier transform by
Z Y
n
dzj dzj
Q K ; K D
FŒ FŒ z ; z exp Œ i .z K C K z/ ; (2.5.9)
j
2=i
then for the inverse Fourier transform, we have

Z Y
n
dK dKj
FŒ z ; z D
j Q K ; K exp Œ i .z K C K z/ :
FŒ (2.5.10)
j
2 i
For integrations over commuting real variables, we define a Fourier transform as

Z Y
dj
n
Q K D
FŒ p FŒ exp Œ i K > ; (2.5.11)
j 2=i
and the inverse Fourier transform is then given by

Z Y
dKj Q
n
FŒ D p FŒ K exp Œ i K > : (2.5.12)
j 2 i
10
Note that the factors .j ˛j /. j ˛ j / commute with everything.
2.6 Functional Differentiation and Integration; Functional Fourier Transforms 63
2.6 Functional Differentiation and Integration; Functional

Fourier Transforms
We begin by extending the concept of differentiation with respect to Grassmann

variables 1 ; : : : ; n ; 1 ; : : : ; n to a continuum limit as n ! 1. Such an extension
for commuting variables will then become straightforward to carry out. By a
continuum limit as n ! 1, it is meant that one introduces Grassmann type
functions a .x/; a .x/ of a continuous variable x, such as specifying spacetime
points, moreover they may also depend, in general, on discrete parameters, such as
a spinor index. By Grassmann type functions, one also means that these functions
all anti-commute for all x and all a.
The simple rules of differentiations taken from the left with respect to a discrete
set of Grassmann variables, defined earlier in Sect. 2.4, such as
@ @ @ @
j D •j i ; D •j i ; j D 0; D 0; (2.6.1)
@i @i j @i @i j
@
j k m s D ıj i k m s j k m ıs i ; (2.6.2)
@i
become replaced, in this limit, by11
• •
b .x 0 / D ıba ı.x 0 x/; .x 0 / D ıba ı.x 0 x/; (2.6.3)
•a .x/ •a .x/ b
•
b .x 0 / c .y 0 / d .y 00 / e .x 00 / D ıba ı.x 0 x/ c .y 0 / d .y 00 / e .x 00 /
•a .x/
b .x 0 / c .y 0 / d .y 00 / ıea ı.x 00 x/; (2.6.4)
involving Dirac deltas instead of just Kronecker deltas. One uses the notation
•=•a .x/ etc, for the functional derivatives , i.e., for derivatives with respect to
functions.
These functions a .x/; a .x/ together with the functional derivatives with
respect to them satisfy the anti-commutation rules:
fa .x/; b .x 0 /g D 0; fa .x/; b .x 0 /g D 0; fa .x/; b .x 0 /g D 0; (2.6.5)

• • • • • •
; D 0; ; D 0; ; D 0;
•a .x/ •b .x 0 / •a .x/ •b .x 0 / •a .x/ •b .x 0 /
(2.6.6)
11
We interchangeably use the notations ı.x x 0 / and ı .D / .x x 0 / where D is the dimensionality
of spacetime.
64 2 Preliminaries

• 0 0 •
; b .x / D ıba ı.x x/; ; .x / D ıba ı.x 0 x/:
0
(2.6.7)
•a .x/ •a .x/ b

• •
; .x 0 / D 0; ; b .x 0 / D 0: (2.6.8)
•a .x/ b •a .x/
Important functionals, i.e., functions of functions, that occur repeatedly in field

theory are the exponentials of linear and/or bilinear forms such as:
XZ
expŒ i . C /; where .dx/ a .x/ a .x/; (2.6.9)
a
a .x/; a .x/ anti-commute with b .x 0 /; b .x 0 / as well as with the functional

derivatives with respect to the latter, and
XZ
exp Œ i A; where A .dx/ .dx 0 / a .x/Aab .x; x 0 /b .x 0 /; (2.6.10)
a;b
a .x/; a .x/ commute with Abc .x 0 ; x 00 /. Note that the above two functionals being
even in anti-commuting functions commute with everything.
Typical functional derivatives of such functionals are
•
exp Œ i . C / D i a .x/ exp Œ i . C /; (2.6.11)
•a .x/
•
exp Œ i . C / D C i a .x/ exp Œ i . C /; (2.6.12)
•a .x/
note the minus sign on the right-hand side of the first equation,
• XZ
exp Œ i A D i .dx 0 /Aab .x; x 0 /b .x 0 / exp Œ i A ; (2.6.13)
•a .x/ b
• XZ
exp Œ i A D i .dx 0 /c .x 0 /Aca .x 0 ; x/ exp Œ i A ; (2.6.14)
•a .x/ c
• •
exp Œ i A D i A.x; y/ac exp Œ i A C exp Œi A
•c .y/ •a .x/
! !
XZ XZ
00 00 00 0 0 0
.dx /d .x /Adc .x ; y/ .dx /Aab .x; x /b .x / ; (2.6.15)
d b
• • • •
exp Œ i A D exp Œ i A : (2.6.16)
•a .x/ •b .x 0 / •b .x 0 / •a .x/
A point which is perhaps not sufficiently emphasized is that a functional

derivative may be also applied in a simple manner to a derivative such as shown
below
Z Z
@ a .x/ D @ .dx 0 / ı.x x 0 /a .x 0 / D .dx 0 / a .x 0 /@ ı.x x 0 /;
Z
•
@ a .x/ D .dx 0 / ıab ı.x 0 x 00 / @ ı.x x 0 / D ıab @ ı.x x 00 /:
•b .x 00 /
(2.6.17)
We may Q take over the integral in (2.4.32) in a continuum limit, and use the
notation x a .da .x/ da .x/=i/ .D D/, which is even in anti-commuting
variables, to obtain for a matrix M = ŒMab .x; x 0 / :
Z
1 1
expŒ i M D .det M/ .D D/ expŒ i . M C C /; (2.6.18)
where we note that

Z ˇ
ˇ
det M D .D D/ expŒ i . M C C /ˇ : (2.6.19)
D0;D0
Here one may use a matrix notation for Mab .x; x 0 / h x j Mab j x 0 i, h x j Iab j x 0 i
D ıab ı.x x 0 /, and note that
XZ
1
h x j Mab j x 0 i.dx 0 /h x 0 j Mbc j x 00 i D ıac ı.x x 00 /: (2.6.20)
b
To see how these integrals work, consider, for example, the Dirac operator in the
presence of an external electromagnetic field A .x/ (see Eq. (I.2) in Appendix I at
the end of the book)
h @ i

M D e A C m M.e/; (2.6.21)
i
h @ i

h x j M.e/ j x 0 i D e A .x/ C m ı.x x 0 /; (2.6.22)
i
corresponding to the interaction of an electron with an external electromagnetic
potential. The following notation is often used for the matrix
@

M.0/ D 1
C m SC : (2.6.23)
i
66 2 Preliminaries
Subject to an appropriate boundary condition, we will learn in Sect. 3.1 how to invert
1
the matrix SC , thus defining the matrix SC . We may conveniently write M.e/ D
1
.SC e A/, and note that
1
det M.e/ D exp Œ Tr ln M.e/ D exp Œ Tr ln .SC e A/ : (2.6.24)
In particular,
det M.e/
D exp Œ Tr ln .I e SC A/ ; (2.6.25)
det M.0/
and the expression in the exponential, on the right-hand side may be formally
expanded as follows
X .e/n Z
Tr ln .I e SC A/ D .dx1 / : : : .dx n /A 1 .x1 / : : : A n .x n /
n>1
n
h i
Tr SC .x1 ; x 2 / 2 SC .x 2 ; x3 / 3 : : : SC .x n ; x1 / 1 ; .dx/ dx 0 dx1 dx 2 dx 3 ;
(2.6.26)
where the trace operation is over the gamma matrices. Such expressions will be
studied in detail later and applications will be given.
A 1
Using the notation M.e/ Œ SC , for the matrix depending on the elec-
tromagnetic potential, and taking into account the expression in (2.6.19), we may
equivalently rewrite (2.6.18) as
ZŒ ;
exp Œ i SC
A
D ; (2.6.27)
ZŒ 0; 0
Z h @ i

ZŒ ; D D D exp i eA C m C C :
i
(2.6.28)
We note that this defines a functional Fourier transform of

h @ i

exp i eA C m ; (2.6.29)
i
giving the Fourier transform from .; / to .; / variables We will recognize
the expression in the exponential in (2.6.29), multiplying the i factor, as the
Dirac Lagrangian in the presence of an external electromagnetic field, with .; /
in (2.6.28) representing external sources coupled to the Dirac field. ZŒ ; =ZŒ 0; 0
acts as a generating functional for determining basic components of an underlying

theory as we will see this on several occasions later on in the book. In particular
by functional differentiating (2.6.27) with respect to and then by , and setting
them equal to zero, gives12 i SC
A A
, with SC interpreted as the propagator of a charged
particle of charge e in the presence of an external electromagnetic field A .x/. SC A
may be then analyzed depending on how complicated the external field is. This will
be dealt with in the next chapter.
Q
By using the notation x d .x/ d.x/=.2=i/ .D D/, for complex

functions .x/; .x/, of a real variable, the integral (2.4.33) in the continuum limit
may be written for a matrix M
Z
1
expŒ i K M 1 K D .D D/ expŒ i . M C K C K /:
det M
(2.6.30)
On the other-hand for a real function .x/ of aQreal variable,
p the integral
in (2.4.34) in the continuum limit, with the notation x d.x/= 2=i .D/,
takes the form
hi i Z h 1 i
1 1
p exp K M K D .D/ exp i M C K : (2.6.31)
det M 2 2
Functionals FŒ •=•K; •=•K and FŒ •=•K may be applied, respectively,

to (2.6.30) and (2.6.31) to generate more complicated functional integrals.
The Functional Fourier transforms in the continuum case may be taken over
from (2.5.6), (2.5.8), (2.5.9), (2.5.10), (2.5.11), (2.5.12). For anti-commuting fields
we define
Z
QFŒ ; D .D D/ FŒ ; exp Œ i . C /; (2.6.32)
Q is then given by
and the inverse Functional Fourier transform of FŒ;
Z
FŒ ; D Q D/
.D Q FŒ
Q ; expŒ i . C /; (2.6.33)
Q
using the notation i da .x/ da .x// Q D/.
D .D Q
xa
For complex fields, the Functional Fourier transform reads (see (2.5.9), (2.5.10))
Z
Q ; K D
FŒK Q ; expŒ i . K C K /;
.D D/ FŒ (2.6.34)
12
See Problem 2.8.
68 2 Preliminaries
Q
where x .d .x/ d.x//=.2=i/ = .D D/, and for the inverse Fourier trans-
form one has
Z
FŒ ; D .DK Q FŒ
Q DK/ Q K ; K expŒ i .K C K/; (2.6.35)
Q
where now x .dK .x/ dK.x/.=.2 i/ = .DK Q DK/.
Q
For real fields, the Functional Fourier transform reads (see (2.5.11), (2.5.12))
Z
Q K D .D/ FŒ exp Œ iK;
FŒ (2.6.36)
Q p
where Q
x d.x/= .2=i/ = .D/, and the Fourier transform of FŒ K is given
by
Z
FŒ D Q FŒ K exp Œ i K;
.DK/ (2.6.37)
Q p
where dK.x/= .2 i/ Q
= .DK/.
x
2.7 Delta Functionals
Consider a real function K.x/ and define the functional

Z
1 • •
GD .dx 0 /.dx 00 / .i/ 0
M.x 0; x 00 / .i/ ; (2.7.1)
2 •K.x / •K.x 00 /
where M.x 0; x 00 / D M.x 00; x 0 /. Then we have the commutation relation

Z
•
i Œ G; K.x/ D i .dx 0 / M.x; x 0 / : (2.7.2)
•K.x 0 /
By using the relation MM 1 D I, we obtain

hi i
i Œ G; K.x/ C K.x/ exp KM 1K D 0: (2.7.3)
2
Since ŒG; ŒG; K.x/ D 0, we may use the identity
eA B eA D Œ A; B C B; for Œ A; Œ A; B D 0; (2.7.4)

2.7 Delta Functionals 69
with B D K, A D i G, to infer, upon multiplying (2.7.3), in the process, by

exp.iG/, that
hi i
K.x/ exp .i G / exp KM 1K D 0; (2.7.5)
2
for all x.
Therefore for every point x we may introduce a delta function ı.K.x// and
introduce the delta functional
Y
ı.K/ D ı K.x/ ; (2.7.6)
x
and use the fact that xf .x/ D 0 implies that f .x/ is a delta function ı.x/, up to
a multiplicative constant, to infer that the expression between the round brackets
in (2.7.5) is equal to ı.K/, up to a proportionality constant, since it is valid for all
x, i.e.,
hi i
exp KM 1K D C exp. i G ı.K/; (2.7.7)
2
where C is a K-independent multiplicative factor, and G is defined in (2.7.1).
Equivalence of the above formulation with the Functional Fourier transform
method developed in the previous section may be established by writing
YZ 1
d.x/
ı.K/ D exp Œ i K.x/.x/; (2.7.8)
x 1 2
which leads from (2.7.7) to

hi i YZ 1 h 1 i
d.x/
exp KM 1K D C exp i M C K.x/.x/ :
2 x 1 2 2
(2.7.9)
Since the left-hand side is unity for K D 0, we obtain for 1=C the functional
integral
YZ 1 h i i
1 d.x/
D exp M : (2.7.10)
C x 1 2 2
Obviously (2.7.9) coincides with that given in (2.6.31) since the .2/ factors
in (2.7.9) cancel out anyway in determining exp Œ i K M 1K=2/. The interest in the
expression in (2.7.7) is that one may also apply to it an arbitrary functional FŒı=ıK
which, as we will see later, essentially provides the solution of field theory.
70 2 Preliminaries
For complex functions K.x/; K .x/ of a real variable x, the same analysis as
above involving both functions, with
Z
• •
G D .dx 0 /.dx 00 / .i/ M.x 0 ; x 00 / .i/ ; (2.7.11)
•K.x/ •K .x/
gives

expŒ i K M 1 K D C exp i G ı.K ; K/; (2.7.12)
up to a (K ; K)-independent
Q multiplicative factor C, and ı.K ; K/ may be taken as
the product x ı.KR .x// ı.KI .x// or equivalently as
Y Z d .x/ d.x/

ı.K ; K/ D 2
exp i K .x/.x/ C .x/K.x/ : (2.7.13)
x
i .2/
For anti-commuting ; , paying special attention to this property, the same

analysis as above gives
expŒ i M 1 D C exp. i G / ı.; /; (2.7.14)
up to a (; )-independent multiplicative factor C, with

Z
• •
G D .dx 0 /.dx 00 / .i/ Mab .x 0 ; x 00 / .i/ ; (2.7.15)
•a .x 0 / •b .x 00 /
and ı.; /, absorbing any factors in C, may be taken as

YZ
ı.; / D d.x/ d.x/ exp i Œ .x/.x/ C .x/.x/ ; (2.7.16)
x
with the functional differentiations on the right-hand side of (2.7.14) to be carried

out before carrying out the integrations in (2.7.16).
Problems
2.1 Consider the matrix G D I cos C n sin , G D G 1 D I cos n sin ,

where n D a=jaj, for some non-zero dimensionless 3-vector a, and
hp ı p i1=2
cos D a2 C 1 C 1 2 a2 C 1 :
p
Show that G 0 . a C 1/ G 1 D 0 a2 C 1.
Recommended Reading 71
2.2 The Dirac Hamiltonian is given by H D 0 Œ p C m in the momentum

description. Use the result in Problem 2.1 to diagonalize the Hamiltonian in the
Dirac representation.
2.3 Evaluate the integral over the Grassmann variable j in (2.4.17).
2.4 As a curious property of integration over a Grassmann variable, consider again,
as in the previous
R problem, the function: f ./ D ˛ 0 C c1 of a single variable.
Show that df ./ D .@=@/f ./ implying an equivalence relationship between
integration and differentiation.
2.5 Use the property of complex conjugation of the product dR R of a real
Grassmann variable R to infer that
(i) .dR / D dR . From this conclude that for a complex Grassmann variable
, .d/ D d .
R R
(ii) d D d . This justifies of assignining the same real numerical
R R
values for both integrals: d , and d over complex Grassmann
variables.
2.6 Verify the validity of the integral in (2.5.1).
2.7 Verify the functional differentiations carried out in (2.6.15).
2.8 Evaluate the functional integral in (2.6.27) to the leading order in the external
A
potential as a functional of SC . Then use this expression to determine SC to this
leading order.
References

2. Manoukian, E. B. (2016). Quantum field theory II: Introductions to quantum gravity, supersym-
metry, and string theory. Dordrecht: Springer.
3. Wigner, E. P. (1959). Group theory, and its applications to the quantum mechanics of atomic
spectra. New York: Academic.
4. Wigner, E. P. (1963). Invariant quantum mechanical equations of motion. In: Theoretical physics
(p. 64). Vienna: International Atomic Energy Agency.
Recommended Reading

Chapter 3
Quantum Field Theory Methods of Spin 1=2
With the many-particle aspect encountered by emerging relativity and quantum

physics and the necessity of describing their creation and annihilation in arbitrary
numbers, one introduces the very rich concept of a quantum field. As a first
encounter with this concept, we investigate, in this chapter, the very significance
of a quantum field for the very simple system described by the celebrated and well
known Dirac equation as given in Eq. (I.1) in Appendix I, as well as of the Dirac
equation in the presence of a general external electromagnetic potential in Eq. (I.2).
The quantum field inevitably leads to the concept of a propagator as describing
the transfer of energy and momentum between emission sources and particle
detectors or between particles, in general. In this particular situation involved with
the Dirac field, this necessarily leads one to investigate and understand how to invert
the Dirac operator . @=i C m/, based on physical grounds, and define carefully the
boundary conditions involved in doing this (Sect. 3.1).
The experimental set-up where a particle may be emitted or detected before or
after having participated in some physical process, respectively, as well as of the
underlying physical process itself are quite appropriately described by introducing
the so-called vacuum-to-vacuum transition amplitude (Sects. 3.1, 3.2, 3.3, 3.4, 3.5,
3.6, 3.7, and 3.8). The latter amplitude leads quite naturally to the determination of
transition amplitudes of physical processes. In particular, the positron is readily re-
discovered by extracting the amplitude by such a method for the system described
by the Dirac equation in an external electromagnetic potential probe (Sect. 3.3).
Various applications will be given including to the Coulomb scattering of
relativistic electrons, derivation of the Euler-Heisenberg effective Lagrangian, as
an example of a modification of Maxwell’s equations, as well of the decay of the
vacuum and of the underlying Schwinger effect of pair productions by a constant
electric field (Sect. 3.8). Particular emphasis is put on deriving the spin & statistics
connection of the Dirac quantum field, and of the generation of gauge invariant
currents (Sects. 3.6, 3.9, and 3.10) based on Schwinger’s elegant point splitting
method. Finally it is noted that a conservation law that may hold classically does

DOI 10.1007/978-3-319-30939-2_3
74 3 Quantum Field Theory Methods of Spin 1=2
not necessarily hold quantum mechanically and lead to modifications of the theory
in the quantum world. The breakdown of such conservation laws lead to calculable
anomalies and are treated in Sect. 3.9 and 3.10, and we provide as well of a concrete
physical example where such an anomaly is actually verified experimentally.
3.1 Dirac Quantum Field, Propagator

and Energy-Momentum Transfer: Schwinger-Feynman
Boundary Condition
In this section, we investigate the very significance of a quantum field for the
very simple system described by the Dirac equation in (I.1). To this end, one may
consider the matrix element of the Dirac field .x/, treated as a quantum field,
between the no-particle state (vacuum) and a single particle state of a particle of
momentum-spin (p; ) as follows h vac j .x/ j p i.
The particle may have been produced by some external emission source, and a
much more transparent treatment for understanding the meaning of such a matrix
element is to introduce a source term .x/ on the right-hand side of the Dirac
equation to reflect this fact:

@
Cm .x/ D .x/: (3.1.1)
i
The presence of the external source .x/ makes the physical interpretation of
the problem at hand clearer, as it plays the role of an emitter and detector of
particles. The source is assumed to be of compact support in time, i.e., it vanishes
except within a given interval or given sets of intervals in time. Physically it means
outside the interval(s) the source ceases to operate, i.e., it is switched off, and is set
equal to zero.1 The explicit structure of .x/ is unimportant. It does not appear in
the expressions of transition amplitudes of quantum particles. The introduction of
such sources facilitate tremendously the computation of transition amplitudes, and
hence their inclusion in the analysis of the underlying physical problems is of quite
practical value. .x/, by definition of an external source, is not a quantum field. It
is not, however, a classical one either as such, but a Grassmann variable. This latter
point does not concern us yet at this stage and we will come back to it later.
Let j 0 i designates the vacuum state before the source .x 0 / is switched
on. After the source is eventually switched off, the latter may have created some
particles. This means that if j 0C i designates the vacuum state after the source is
switched-off, then quantum mechanics says that h 0C j 0 i is not necessarily one
or just a phase factor as the system may be in some other state involving some
1
It is important to emphasize, that the source need not vanish abruptly at any point and may be
taken to be continuous at a point where it vanishes (see Problem 3.1).
3.1 Dirac Quantum Field, Propagator and Energy-Momentum. . . 75
particles. This amplitude corresponds to the case where we begin with the vacuum
before the source .x 0 / is switched on, and end up with the vacuum after the source
is switched off. Here we have conveniently used the argument x 0 for the source.
We consider the matrix element h0C j .x/ j 0 i, and solve (3.1.1) for the latter.
We will then investigate the meaning of this matrix element later.
By carrying out a 4D Fourier transform of (3.1.1) and inverting, in the process,
the matrix . @=i C m/, we obtain for the matrix element in question
Z Z
h 0C j .x/ j 0 i .dp/ i p .xx 0 / . p C m/
D .dx 0 / e .x 0 /; (3.1.2)
h 0C j 0 i .2/4 . p 2 C m2 /
.dx/ D dx1 dx 2 dx 3 dx4 ; p.x x 0 / D p.x x 0 / p0 .x0 x 0 /; (3.1.3)
where we have used the relation . p C m/. p C m/ D . p2 C m2 /.

Regarding (3.1.2), we are faced with the interpretation of the pole that may arise
from the vanishing of the denominator . p2 C m2 /, whose physical significance and
the evaluation of the expression on the right-hand side of (3.1.2) provides the first
step for carrying out computations in quantum field theory.
The denominator in (3.1.2) may be rewritten as
p
. p2 C m2 / D . p 0 E/. p 0 C E/; E D p2 C m2 : (3.1.4)
If x 0 is such that x0 D x 02 > x 0 0 , as shown in Fig. 3.1, for all x 0 0 contributing to

the x 0 0 -integral in (3.1.2), that is, if the source is switched off before the time x 0 ,
we have the physical situation of a causal arrangement (x0 > x 0 0 ), where energy is
being transferred from the source to the point x2 , with the source playing the role of
an emitter. Thus we have to pick up the pole p0 D C E in (3.1.2) corresponding to
an energy gain at the point in question. To do this, we make a transition to a complex
p0 -plane. Let Im p0 denote the imaginary part of p 0 . For the case considered above,
x0 x0 0 T > 0, and the real part of .ip 0 T/ is given by Re .ip 0 T/ D .Im p 0 /T
(Fig. 3.2).
SOURCE
x1 0 x 0 x2 0
Fig. 3.1 x1 0 ; x2 0 are, respectively, two arbitrary points before the source is switched on and after
it is switched off at which h 0C j .x/ j 0 i is examined
p 0−plane
−E+i
x0 < x 0
x0 > x 0
E−i
Fig. 3.2 Figure showing the direction of the contour that is to be taken in the complex p0 plane
when applying the residue theorem, taken c.w. below the horizontal axis for x0 > x0 0 , and c.c.w.
above it for x0 < x0 0 , generating, in turn, two semi-circles of infinite radii, one below and one
above the horizontal axis, respectively
To pick up the pole at p 0 D C E we displace E by i

,
! 0, in (3.1.4), i.e.,
we make the replacement
. p2 C m2 / ! Œ p 0 .E i
/Œ p 0 C .E i
/ D . p2 C m2 i
/: (3.1.5)
We close the p 0 contour c.w. from below the horizontal axis, for which Im p 0 <
0, ensuring that the infinite semi-circle part of the contour in the lower plane gives
no contribution. The latter is as a consequence of the fact that expŒ.Im p 0 /T ! 0
for .Im p 0 /T ! 1, T > 0, and, in the process pick up the pole p 0 D C E, by
the application of the Cauchy Theorem. We note that for any real number a > 0,
a
is equivalent to
in complex integration for
! C 0. An application of the
residue theorem gives
Z Z
h 0C j .x/ j 0 i 0 d3 p 0
D i .dx / e i p .xx / . p C m/ .x 0 /; (3.1.6)
h 0C j 0 i .2/3 2p 0
p
where x0 > x 0 0 , and now p 0 D C p2 C m2 . Here we recall the source .x 0 /
acts as an emitter.
On the other hand for x0 D x1 0 in Fig. 3.1, i.e., x0 x 0 0 T 0 < 0, meaning
that the source is switched on after the time x 0 in question, energy is absorbed by
the source, now acting as a detector, with energy loss from point x1 . That is, in this
case, we have to pick up the pole at p 0 D E. We may then close the p 0 contour
c.c.w. from above the horizontal axis in the complex p0 plane by noting, in the
process, that expŒ.Im p 0 /T 0 ! 0, for Im p 0 ! C 1, T 0 < 0, to obtain by the
residue theorem
Z Z
h 0C j .x/ j 0 i 0 d3 p 0
D i .dx / ei p .xx / . p C m/ .x 0 /; (3.1.7)
h 0C j 0 i .2/3 2p 0
3.1 Dirac Quantum Field, Propagator and Energy-Momentum. . . 77
p
where x0 < x 0 0 , p0 D C p2 C m2 (not minus), and we made a change of
variables of integration p ! p to finally write down the above integral.
Accordingly, and quite generally, from (3.1.5), (3.1.6), (3.1.7) one may rewrite
the expression in (3.1.2) in a unified manner as
Z
h 0C j .x/ j 0 i
D .dx 0 / SC .x x 0 / .x 0 /; (3.1.8)
h 0C j 0 i
where SC .x x 0 / is referred to as the Dirac (fermion) propagator given by

Z
0 .dp/ i p .xx 0 / . p C m/
SC .x x / D e ; (3.1.9)
.2/4 . p 2 C m 2 i
/
responsible for the transfer of energy (and momentum) as discussed above. The i
term, with
! C 0, in the denominator, dictated by the physics of the situation,
specifies the (Schwinger-Feynman) boundary pcondition for the proper integration in
the complex p 0 plane. Also with p 0 D C p2 C m2 ,
Z
0 d3 p 0
SC .x x / D i e i p .xx / . p C m/; for x 0 > x 0 0 ; (3.1.10)
.2/3 2p 0
Z
0 d3 p 0
SC .x x / D i 3 0
e i p .xx / . p C m/; for x 0 < x 0 0 ; (3.1.11)
.2/ 2p

@
Cm SC .x x 0 / D ı .4 / .x x 0 /; (3.1.12)
i
as follows from (3.1.9). The propagator also satisfies

!
0 @
SC .x x/ C m D ı .4 / .x 0 x/: (3.1.13)
i

The field equation for .x/ .x/ 0 , in the presence of the external source,
is given by
!
@
.x/ C m D .x/: (3.1.14)
i
The vacuum expectation value of .x/ is then easily worked out to be

Z
h 0C j .x/ j 0 i
D .dx 0 / .x 0 / SC .x 0 x/; (3.1.15)
h 0C j 0 i
and from (3.1.10), (3.1.11),

Z Z
h 0C j .x/ j 0 i d3 p 0
D i .dx 0 / e i p .x x / .x 0 /. p C m/; (3.1.16)
h 0C j 0 i .2/3 2p 0
if x 0 < x 0 0 , i.e., if the source .x 0 / is switched on after the time x 0 ,

Z Z
h 0C j .x/ j 0 i 0 d3 p 0
D i .dx / 3 0
e i p .x x / .x 0 /. p C m/; (3.1.17)
h 0C j 0 i .2/ 2p
if x 0 > x 0 0 , i.e., if the source .x 0 / is switched off before the time x 0 .
Thus the Schwinger-Feynman boundary condition, based on physical grounds,
amounts in replacing the mass m by m i
in the denominator in defining the
propagator. Needless to say for m D 0 the i
should survive.
We note that due to the equality
d3 p 0 p
.dp/ ı. p2 C m2 / ™. p 0 / D 0
dp ı. p 0 p2 C m2 /; (3.1.18)
2p
where ™. p 0 / D 1 for p 0 > 0, ™. p 0 / D 0 for p 0 < 0, and due to the fact that a
Lorentz transformation does not change the sign of p 0 , we may infer the Lorentz
invariance of the measure on the right-hand side of (3.1.18) as well.
We are now ready to investigate, in the next section, the role of .x/ as a
quantum field and the particle content of the theory.
3.2 The Dirac Quantum Field Concept, Particle Content,

and C, P, T Tansformations
For the interpretation of the role played by the quantum field .x/ for the simple
system described by the Dirac equation, we consider the expression for the matrix
element h 0C j .x/ j 0 i=h 0C j 0 i in (3.1.6), that is for x 0 > x 0 0 in reference
to Fig. 3.1, where the source acts as an emitter.
With the functional time dependence of .x 0 / being of compact support in time
explicitly absorbed in .x 0 /, we may carry out the x 0 -integral in (3.1.6) over all x 0
to obtain
Z
h 0C j .x/ j 0 i d3 p
D e ip x . p C m/ i . p/; (3.2.1)
h0C j 0 i .2/3 2p 0
Z
0
. p/ D .dx 0 / e ip x .x 0 /: (3.2.2)
3.2 The Dirac Quantum Field Concept, Particle Content, and C, P, T. . . 79
Using the expression for the projection operator in Eq. (I.21) in Appendix I over
spin states, we rewrite (3.2.1) as
Z X
h 0C j .x/ j 0 i m d3 p ip x
D e u.p; / h p j 0 i: (3.2.3)
h 0C j 0 i
p 0 .2/3
where
h p j 0 i Œ i u.p; / . p/ ; (3.2.4)
and as it will be evident below, and as shown in the next section, h p j 0 i

represents an amplitude that the source, as an emitter, emits a particle of momentum-
spin (p; ) which persists after the source is switched off.
Since we have the vacuum state on the left-hand side of (3.2.3), we reach the
inevitable conclusion that .x/ has annihilated the particle in question to end up in
the vacuum.
At this stage, it is instructive to introduce the normalization condition of a particle
state,2 as well as of the resolution of the identity of the momenta and spin of such a
particle defined, in turn, by
Z X
p0 m d3 p
h p 0 0 j p i D .2/3 ı 0 ı 3 .p 0 p/; I D j p ih p j;
m
p 0 .2/3
(3.2.5)
to rewrite (3.2.3) as
Z X
h 0C j .x/ j 0 i h 0C j .x/ j p i m d3 p
D h p j 0 i: (3.2.6)
h 0C j 0 i
h 0C j 0 i p 0 .2/3
and infer from (3.2.3) that
h 0C j .x/ j p i
D e ip x u.p; /: (3.2.7)
h 0C j 0 i
A similar analysis as above dealing with h 0C j .x/ j 0 i in (3.1.16) for the

causal arrangement, now, with x 0 < x 0 0 , relative to the source .x 0 /, in reference
to Fig. 3.1, leads to the identification
Œ i . p/ u.p; / h 0C j p i; (3.2.8)
2
Particle states and the Poincaré Algebra will be considered in Sect. 4.2.
with the latter representing an amplitude that the source , as a detector, absorbs a
particle, and finally obtain
h p j .x/ j 0 i
D e ip x u.p; /: (3.2.9)
h 0C j 0 i
Thus we reach the inevitable conclusion that .x/ has created a particle since we
began with the vacuum and ended up having a particle detected by the source.
It remains to consider the situation for h 0C j .x/ j 0 i with x 0 < x 0 0 relative
to .x 0 /, and h 0C j .x/ j 0 i with x 0 > x 0 0 , relative to .x0 /. This will be
done in the next section.
In the remaining part of this section and in view of applications in the following
sections, we consider the Dirac equation in an external electromagnetic field A in
the presence of an external source . This is given by

@
e A .x/ Cm .x/ D .x/; (3.2.10)
i
with e denoting the charge. We define it to carry its own sign, i.e., e D jej for an
electron.
In analogy to (3.1.12), we introduce the propagator SC A
.x; x 0 /, now in the
presence of an external electromagnetic field, satisfying

@
e A .x/ C m S AC .x; x 0 / D ı .4 / .x x 0 /: (3.2.11)
i
The reason as to why we have written the functional dependence of SCA

as .x; x 0 /
0
rather than .x x / is that for A .x/ ¤ 0 translational invariance is generally
broken. One may convert (3.2.11) to the following integral equation
Z
SAC .x; x 0 / D SC .xx 0 / C e .dx 00 / SC .xx 00 / A .x 00 /SAC .x 00 ; x 0 /: (3.2.12)
This is easily verified by applying the uperator . @=i C m/ to (3.2.12), use (3.1.12)
and finally integrate over a delta function to see that (3.2.11) is indeed satisfied.
The expression for h 0C j .x/ j 0 iA =h 0C j 0 iA , now in the presence of the
external electromagnetic field A .x/, follows from (3.2.10) and takes the form
Z
h 0C j .x/ j 0 iA
h .x/iA D .dx 0 / SC
A
.x; x 0 / .x 0 /: (3.2.13)
h 0C j 0 iA
We consider the following physical situation where .x 0 / is so chosen that x 0 0

lies sufficiently in the remote past and x 0 lies sufficiently forward in the future,
with A .x 00 / in (3.2.12) being effective only in the region, x 0 0 < x 00 0 < x 0 . In
x 0
x 0
x0
Fig. 3.3 .x 0 / is so chosen that x 0 0 lies sufficiently in the remote past and x 0 lies sufficiently
forward in the future, with A .x 00 / being effective only in the region x 0 0 < x 00 0 < x 0
practice, one may choose x 0 0 x 00 0 x0 , if necessary, depending on A .x 00 /,

in a limiting sense for the time variables x 0 0 ; x0 (Fig. 3.3).
For a sufficiently weak external field A , (3.2.12) becomes
Z
SAC .x; x 0/ ' SC .xx 0/ C e .dx 00 / SC .xx 00 / A .x 00 /SC .x 00 x 0/: (3.2.14)
ı
The condition of a weak external field will be relaxed in Sect. 3.6 Appendix IV and
an all order treatment will be given.
We may carry out the integrations in (3.2.13), (3.2.14) to obtain3
Z X
m d3 p m d3 p 0 ip 0x
h .x/iA ' 0 .2/3 p 0 0 .2/3
e u.p 0 ; 0 /h p 0 0 j p iA Œ i u.p; /. p/ ;
p
0
(3.2.15)
where
h p0 i
hp 0 0 j p iA D .2/3 ı 0 ı 3 .p 0 p/ C i e u.p 0 ; 0 / A . p 0 p/ u.p; / ;
m
(3.2.16)
which coincides with (3.2.5) for e D 0. In the process of deriving (3.2.15), (3.2.16),
we have, by using (3.1.10), (I.21), conveniently rewritten the first term SC .x x 0 /
on the right-hand side of (3.2.14) as
Z X
0 m d3 p m d3 p 0 i. p 0 xp x 0 /
SC .x x / D i e
0
p 0 .2/3 p 0 0 .2/3

p0
u.p 0 ; 0 / Œ .2/3 ı 0 ı 3 .p 0 p/ u.p; /; (3.2.17)
m
3
See Problem 3.3.
for x 0 > x 0 0 , and defined the Fourier transform

Z
.dQ/ iQ x
A .x/ D e A .Q/: (3.2.18)
.2/4
We note that (3.2.16) is also valid for a time-independent external field A .x/ as
long as x 0 ; x 0 0 are taken large in a limiting sense. In this case A . p 0 p/
in (3.2.16) will be proportional to ı. p 0 0 p 0 /.
In the causal arrangement discussed above, the quantum field .x/, in a source
free region, destructs the particle after it has scattered off the external field A ,
with the transition amplitude for the process corresponding to p ! p 0 0 , say,
for p ¤ p 0 , given from (3.2.16) to be
h p 0 0 j piA D Œ i e u.p 0 ; 0 / A . p 0 p/u.p; / : (3.2.19)
The external electromagnetic field A .x/ may in turn create particles. This will
be taken up in Sect. 3.8.
The structure given in (3.2.16) is interesting as, in particular, it provides us with
the value of the magnetic moment of the charged particle (electron). To this end,
we may use the equation . p C m/u.p; / D 0 (see (I.10)), to write u.p; / D
. p=m / u.p; /, multiply the latter from the left by to derive the identity
p Œ ;
u.p; / D u.p; / p u.p; /; (3.2.20)
m 2m
and similarly
p 0 Œ ;
u.p 0 ; 0 / D u.p 0 ; 0 / C u.p 0 ; 0 / p 0 : (3.2.21)
m 2m
By multiplying (3.2.20) from the left by u.p 0 ; 0 / and (3.2.21) from the right by
u.p; /, and adding the resulting two equations lead to the following decomposition,
referred to as the Gordon decomposition,
e Œ u.p 0 ; 0 / u.p; / A .Q/

h . p C p 0 / Œ ; i
D e u.p 0 ; 0 / C Q u.p; /A .Q/
2m 4m
h . p C p 0 / ie i i
D u.p 0 ; 0 / e A .Q/ C Q A .Q/ Q A .Q/ Œ ; u.p; /;
2m 2m 4
(3.2.22)
and Q D p 0 p is the (four) momentum transfer. In reference to the second

term in the second equality on the right-hand side of (3.2.22), note that the specific
contribution over the sum of the indices ; , for D i, D j, is given by
i e ij k

" Q i Aj .Q/ Qj A i .Q/ u.p 0 ; 0 / S k u.p; / ; i; j; k D 1; 2; 3;
2m
(3.2.23)
with a sum over repeated indices understood, where S k D i " s `k Œ s ; ` =8 is the
spin matrix in (I.15), and in writing (3.2.23), we have used, in the process, the
identity
" ij k " s ` k D ı i s ı j ` ı i ` ı j s : (3.2.24)

We note that i " ij k Qi Aj .q/ Qj Ai .Q/ =2 denotes the Fourier transform of the
magnetic field
i ij k
B k .Q/ D " Q i Aj .Q/ Qj A i .Q/ (3.2.25)
2
and we may rewrite (3.2.23) as
ge
Œ u.p 0 ; 0 / u.p; / B.Q/; D S; (3.2.26)
2m
where is the magnetic dipole moment of the charged particle (electron) with
the g-factor equal to 2. The modification of this value due to so-called radiative
corrections in QED will be considered later in Sect. 5.11.2.
3.2.1 Charge Conjugation (C), Parity Transformation (T),

and Time Reversal (T) of the Dirac Quantum Field
In the remaining part of this section, we investigate the transformation rules of the
Dirac quantum field under charge conjugation .C/, parity transformation .P/, and
time reversal .T/.
In the presence of an external electromagnetic field, the Dirac equation reads
from (3.2.10), with D 0,

@
e A .x/ Cm .x/ D 0; (3.2.27)
i
> C
while from Eq. (I.3), C satisfies the same equation with sign of the
charge e reversed, i.e.,
h i
@ C
C eA .x/ Cm .x/ D 0; (3.2.28)
i
where C D i 2 0 is the charge conjugation matrix defined in Eq. (I.3). Hence we

may define the charge conjugation of a Dirac quantum field by
>
C .x/ C1 D C .x/; (3.2.29)
up to a phase factor.
On the other hand, according to Eq. (I.18), 0 .x 0 /, with x D .x 0 ; x/,
satisfies the same Dirac equation . @ =i C m/ .x/ D 0, i.e.,
@

C m 0 .x 0 / D 0; x 0 D .x 0 ; x/: (3.2.30)
i
Hence we may infer that the parity transformation of the Dirac quantum field may
be defined by
P .x/ P1 D 0 .x 0 /; (3.2.31)
For time reversal, we note that by setting x 00 D .x 0 ; x/, that we may rewrite the
Dirac equation . @ =i C m/ .x/ D 0 as
h 0 @0 00 @ i
C C m .x/ D 0: (3.2.32)
i i
Upon multiplying the latter equation by 5 C , making the substitution x 00 $ x,

and using the facts that
Œ 5 C ; 0 D 0; Œ 5 C ; 2 D 0; f 5 C ; 1 g D 0; f 5 C ; 3 g D 0;
(3.2.33)
as is easily verified, we may rewrite (3.2.32) as
h 2@ 2 0@ 0 1@ 1 3@ 3 i
C C m 5C .x 00 / D 0: (3.2.34)
i i i i
Finally we make use of the property of T, as an anti-unitary operator, that it complex
conjugates, and note that, e.g., in the Dirac representation, . 2 / D 2 , and
0 ; 1 ; 3 are real, to obtain upon multiplying (3.2.34) from the left by T1 , and
from the right by T, the equation
@

C m T1 5 C .x 00 /T D 0; (3.2.35)
i
3.3 Re-Discovering the Positron and Eventual Discovery of Anti-Matter 85
which by comparing it with the Dirac equation, gives the following transformation
rule for time reversal of the Dirac quantum field
T .x/ T1 D 5 C .x 00 /; x 00 D .x 0 ; x/; (3.2.36)
3.3 Re-Discovering the Positron and Eventual Discovery

of Anti-Matter
We have seen in Sect. 3.2, (3.2.1),(3.2.6), that the vacuum expectation value of
the Dirac quantum field .x/, in the presence of an external source .x/, i.e.,
satisfying (3.1.1), is given by
Z Z
h 0C j .x/ j 0 i 0 d3 p 0
D i .dx / 3 0
eip .xx / . p C m/ .x 0 /
h 0C j 0 i .2/ 2p
Z X
m d3 p h 0C j .x/ j pi
D hp j 0 i; (3.3.1)

p 0 .2/3 h 0C j 0 i
when the source .x 0 / is in operation, i.e., non-zero, and then switched off
sometime before the time x 0 , i.e., x 0 > x 0 0 in reference to Fig. 3.1, for all x 0 0
contributing to the above integral. Here h p j 0 i denotes an amplitude p that
the source emits a particle of momentum-spin (p; ), energy p 0 D C p2 C m2
(see (3.2.4) and later on in this section),
h 0C j .x/ j pi p
D eip x u.p; /; p0 D C p2 C m2; (3.3.2)
h 0C j 0 i
leading to the inevitable conclusion that .x/ annihilates such a particle to end up
with the vacuum. The so-called wave-function of the particle is given by (3.3.2).
Similarly, when the source .x 0 / is switched on after the time x 0 , i.e., x 0 0 > x 0
(see (3.1.16), (3.2.8), (3.2.9)),
Z Z
h 0C j .x/ j 0 i d3 p 0
D i .dx 0 / .x 0 / . p C m/ eip .x x/
h 0C j 0 i .2/3 2p 0
Z X
m d3 p h p j .x/ j 0 i
D 0 3
h 0C j pi ; (3.3.3)

p .2/ h 0C j 0 i
with
h p j .x/ j 0 i p
D e i p x u.p; /; p0 D C p2 C m2; (3.3.4)
h 0C j 0 i
h 0C j p i D Œ i . p/ u.p; / ; (3.3.5)
representing an amplitude that the source .x 0 / has absorbed a particle of

momentum-spin (p; ), thus playing the role of a detector, and leading to the
inevitable conclusion that .x/ creates such a particle which eventually goes to
the detector.
We have also considered the Dirac equation in an external electromagnetic field
in (3.2.10) under the same condition as in (3.3.1), with the source switched off
sometime before the time x 0 , and we have identified in (3.2.16), (3.2.10), the charge
of the particle to be e.
What is the role of .x/ if the source is nonzero and is finally switched off
sometime before the time x 0 , i.e., x 0 > x 0 0 instead of the situation encountered
in (3.3.3).
In this case, we may use the expression in (3.1.17) and the orthogonal projection
in Eq. (I.22), to write
Z Z
h 0C j .x/ j 0 i d3 p 0
D i .dx 0 / e ip .x x/ .x 0 /. p C m/
h 0C j 0 i .2/3 2p 0
Z X
m d3 p ip x
D e v.p; / Œ i .p/ v.p; /: (3.3.6)

p 0 .2/3
Upon comparing this equation with (3.3.1), (3.3.2), for the same causal arrange-
ment where now the source .x 0 / is also switched off before the time x 0 , we may
infer that .x/ annihilates a particle with wave-function
h 0C j .x/ j pi p
D eipx v.p; /; p0 D C p2 C m2 ; (3.3.7)
h 0C j 0 i
and
Œ i .p/ v.p; / h p j 0 i; (3.3.8)
represents an amplitude that the source .x 0 / creates a particle of momentum-spin

(p; ).
What is the nature of this particle? To answer this question, we introduce an
external electromagnetic potential probe ıA .x/ off which a particle, after being
created, may, for example, be scattered. Before doing this, we rewrite (3.3.6) as
ZX
h 0C j .x/ j 0 i m d3 p m d3 p 0
D
h 0C j 0 i 0
p 0 .2/3 p 0 0 .2/3
;
h 0C j .x/ j p 0 0 i
h p 0 0j p ih p j 0 i: (3.3.9)
h 0C j 0 i
where h p 0 0 j pi is defined in (3.2.5).

In the presence of the external probe, the Dirac equation for .x/ reads
" ! #
@
.x/ C e ıA .x/ C m D .x/: (3.3.10)
i
This simply amounts in the modification of the matrix element of .x/ between the
vacuum states, in the absence of the probe from
Z
h 0C j .x/ j 0 i
D .dx 0 / .x 0 / SC .x 0 ; x/; (3.3.11)
h 0C j 0 i
where

@
SC .x 0 ; x/ C m D ı .4 / .x 0 x/; (3.3.12)
i
to
Z
h 0C j .x/ j 0 iıA
' .dx 0 / .x 0 /
h 0C j 0 iıA
h Z i
SC .x 0 x/ C e .dx 00 / SC .x 0 x 00 / ıA .x 00 / SC .x 00 x/ ; (3.3.13)
in the presence of the external probe as a perturbation of the system.

Repeating the same analysis as the one leading to (3.2.15), (3.2.16), as before,
gives4
Z X
h 0C j .x/ j 0 iıA m d3 p m d3 p ip 0x
D 0 .2/3 p 0 .2/3
e v.p 0 ; 0 /
h 0C j 0 iıA 0
p

0 0
hp j p iıA Œ i .p/ v.p; / ; (3.3.14)

p0
h p 0 0 j p iıA D .2/3 ı 0 ı 3 .p 0 p/ i e v.p; / ı A . p 0 p/ v.p 0 ; 0 / :
m
(3.3.15)
4
See Problem 3.5.
Upon comparing this equation with (3.2.16), and taking into account (3.3.6),
(3.3.7), (3.3.8), we may infer that .x/ annihilates a particle of mass m and charge
e, i.e., of pcharge Cjej. For such a particle of momentum p, its energy is given
by p 0 D C p 2 C m 2 , and its wave-function is given in (3.3.7). Also h p j 0 i;
as defined in (3.3.8), represents the amplitude that the source .x/ creates such a
particle. This is the anti-electron, the positron.
As mentioned in the Introductory Chapter of the book, prior to its experimental
discovery of the positron, it was referred to as the donkey electron by George
Gamow because it would move in the opposite direction to that of an electron in
an applied field.
It is important to note that v.p; /, corresponding to the positron in its initial
state, is to the left of , and v.p 0 ; 0 /, corresponding to the its final state, is on
its right-hand side.
Under space reflection, we have from below Eq. (I.18), that
u.p; / ! 0 u.p; / D C u.p; /;

v.p; / ! v.p; / 0 D v .p; / D v.p; /;
and hence the electron and positron have opposite intrinsic parities.
Finally we note that for the source switched on after x 0 , i.e., x 0 0 > x 0 ,
Z Z
h 0C j .x/ j 0 i d3 p 0
D i .dx 0 / eip .xx / . p C m/ .x 0 /;
h 0C j 0 i .2/3 2p 0
Z X
m d3 p
D 0 .2/3
Œ i v.p; / .p/ e ip x v.p; /; (3.3.16)

p
where we have used Eq. (II.21). Now
h p j .x/ j 0 i p
D e ip x v.p; /; p0 D C p2 C m2 ; (3.3.17)
h 0C j 0 i
and .x/ creates a positron,
Œ i v.p; / .p/ D h 0C j p i; (3.3.18)
denotes an amplitude that the source .x/ absorbs a positron with momentum-spin
(p; ) and hence it plays the role of a detector.
Clearly to distinguish between the electron and positron we may rewrite the states
as j p; "i, with " D
for the electron/positron, respectively, with the conventions
p0
h p 0 0 ; " 0 j p ; "i D .2/3 ı 0 ı" 0 " ı 3 .p 0 p/; (3.3.19)
m
and
h 0C j .x/ j p ; i
D e ip x u.p; /; (3.3.20)
h 0C j 0 i
h p ; C j .x/ j 0 i
D e ip x v.p; /; (3.3.21)
h 0C j 0 i
with .x/ annihilating/creating an electron/positron, respectively,
h p ; j .x/ j 0 i
D e ip x u.p; /; (3.3.22)
h 0C j 0 i
h 0C j .x/ j p ; Ci
D eip x v.p; /; (3.3.23)
h 0C j 0 i
with .x/ creating/annihilating an electron/positron, respectively.

Clearly, this was the beginning of the discovery of anti-matter, in general,
thanks to the Dirac equation. It is interesting to remember (see the Introductory
Chapter of the book) that, in one of his talks, Dirac remarked that his equation was
more intelligent than its author.
With the above analysis now carried out, it will be quite evident in the sequel,
that the computations of transition amplitudes of processes in quantum field theory
become more manageable.
The amplitudes for detection of a particle/antiparticle and emission of parti-
cle/antiparicle are spelled out below using the sign of the charge " D
.5 To do
this we obtain, in the process, the expression for the vacuum-to-vacuum transition
amplitude h 0C j 0 i for the Dirac equation in the presence of an external source.
3.3.1 h 0C j 0 i for the Dirac Equation
For the Dirac equation (3.1.1) in the presence of an external source,

@
Cm .x/ D .x/; (3.3.24)
i
5
When it is evident we suppress the parameter " in amplitudes.
we have seen in (3.1.8) that

Z
h 0C j .x/ j 0 i D .dx 0 / SC .x x 0 / .x 0 / h 0C j 0 i; (3.3.25)
where SC .x x 0 / is given in (3.1.9).

In Sect. 4.6, on the Quantum Dynamical Principle, we will learn that the matrix
element h 0C j .x/ j 0 i, for the Dirac equation in (3.3.24), may be generated
by functionally differentiating h 0C j 0 i with respect to the external source .x/.
More precisely,
•
. i/ h 0C j 0 i D h 0C j .x/ j 0 i; (3.3.26)
•.x/
which from (3.3.25) leads to

Z
•
. i/ h 0C j 0 i D .dx 0 / SC .x x 0 / .x 0 / h 0C j 0 i: (3.3.27)
•.x/
Upon integration this gives

Z

h 0C j 0 i D exp i .dx/ .dx 0 / .x/SC .x x 0 /.x 0 / ; (3.3.28)
which should be compared with (2.6.13)/(2.6.10) and obviously verifies (3.3.27).

We have normalized h 0C j 0 i to unity for .x/ D 0, .x/ D 0, i.e., in the
absence of the external source, dispensing with any phase factor, signalling the fact
that the vacuum stays the same and nothing happens in the latter case.
Consider a source function .x/ given by (Fig. 3.4)
.x/ D 1 .x/ C 2 .x/; (3.3.29)
( η1 ) (η 2 )
x0
1 x0
2
Fig. 3.4 A causal arrangement with the source 2 switched on after the source 1 is switched off
with a causal arrangement such that 2 .x/ is switched on after 1 .x/ is switched
off. We will learn in Sect. 3.5, in the study of the Spin & Statistics Connection for
the Dirac field, that .x/, .x/ are Grassmann variables.
Thus we may write
h 0C j 0 i D h 0C j 0 i2 exp iW21 h 0C j 0 i1 ; (3.3.30)
where
Z
iW21 D i .dx1 /.dx2 /2 .x2 /SC .x2 x1 /1 .x1 /
Z
C i .dx1 /.dx2 /1 .x1 /SC .x1 x2 /2 .x2 /: (3.3.31)
In (3.3.30) we have emphasized the source-dependence by rewriting, for exam-

ple, (3.3.28) as h 0C j 0 i .
Using the corresponding two expressions for SC .x2 x1 /, SC .x1 x2 /,
in (3.1.10), (3.1.11), respectively, and the projection operations in (I.21), (I.22),
we obtain
Z X
m d3 p
i W21 D Œi 2 . p/ u.p; / Œi u.p; /1 . p/

p 0 .2/3
Z X
m d3 p
C Œi v.p; / 2 .p/ Œi 1 .p/v.p; /: (3.3.32)

p 0 .2/3
In writing the second integral, we have used the anti-commutativity of 1 .p/,

2 .p/, to move 2 .p/ to the left of 1 .p/, thus picking another minus sign.
For bookkeeping purposes, we use an ingenious notation due to Schwinger,
which facilitates further the analysis as well as the physical interpretation of the
underlying theory. We write
p p
2md!p . p/u.p; / D p ; 2md!p ./v.p; /.p/ D p C
(3.3.33)
p p
2md!p u.p; /. p/ D p ; 2md!p .p/./v.p; / D p C
providing four different entries, where
d3 p
d!p D : (3.3.34)
2p 0 .2/3
Don’t let the notation in (3.3.33) scare you. It is for bookkeeping purposes, and
it provides a simple worry free formalism for getting correct numerical factors such
as .2/3 and so on.
Let r D .p; ;
/,
D
, corresponding to all possible values that the variables
take in a convenient discrete values notation. This allows us to rewrite i W21 as a
formal sum
X
i W21 D Œ i r 2 i r 1 : (3.3.35)
.r/
One may also rewrite h 0C j 0 i in the form of a unitarity sum

X X
h 0C j 0 i D h 0C j nI n1 ; n2 ; : : :i 2 h nI n1 ; n2 ; : : : j 0 i1 ;
n n1 Cn2 C:::Dn
(3.3.36)
with h nI n1 ; n2 ; : : : j 0 i1 denoting the amplitude that the source 1 has emitted
n particles, n1 of which have the value r1 , and so on. The Fermi character of the
anti-commuting source requires that n1 D 0 or 1, and so on. Similarly, h 0C j
nI n1 ; n2 ; : : :i 2 , denotes the amplitude that the source 2 has absorbed (detected)
n particles, and so on.
Upon expanding the exponential of (3.3.35), using in the process (3.3.30), and
comparing with (3.3.36), we obtain for any given source :
.ir /n1 .ir2 /n2

h nI n1 ; n2 ; : : :j 0 i D h 0C j 0 i p 1 p ; (3.3.37)
n1 Š n2 Š
.ir /n2 .ir1 /n1

h 0C j nI n1 ; n2 ; : : :i D h 0C j 0 i p 2 p ; (3.3.38)
n2 Š n1 Š
where the ni D 0 or 1.
Physically, we note that h 0C j 0 i corresponds to processes where arbitrary
number of particles are emitted and re-absorbed through the time of operation of
.; /, with no particles (antiparticles) present after the sources cease to operate,
and before they begin to operate. Accordingly, in scattering theory, we define
amplitudes for emission of a given particle (antiparticle) by a given source, that
escapes this parent source as a free particle, by dividing the corresponding above
expression by h 0C j 0 i. Thus one eliminates particles absorbed and then
re-absorbed that do not participate in dynamical scattering process of interest.
Similarly, the amplitude of absorption of a given particle (antiparticle), way after the
particle (antiparticle) emerges, say, from a scattering process, is defined by dividing
the corresponding above expression by h 0C j 0 i. Accordingly, we introduce the
amplitudes of either absorption or emission of a particle (antiparticle) by a given
3.4 Coulomb Scattering of Relativistic Electrons 93
source as follows:
p
2m d!p Œ i . p/ u.p; / D h 0C j p ; i; (3.3.39)
p
2m d!p Œ i v.p; / .p/ D h 0C j p ; Ci; (3.3.40)
p
2m d!p Œ i u.p; / . p/ D h p ; j 0 i; (3.3.41)
p
2m d!p Œ i .p/ v.p; / D h p ; C j 0 i; (3.3.42)
where h 0C j p ; i denotes the amplitude that a source absorbs a particle, thus

playing the role of a detector, and h 0C j p ; Ci, denotes the amplitude that
also absorbs an antiparticle. Given that h 0C j p ; i denotes the amplitude for the
event just mentioned, i.e., that absorbs a given particle (electron), the analysis
leading to (3.3.18), shows that h 0C j p ; Ci necessarily represents the amplitude
that absorbs an antiparticle (the positron). Similarly, h p ; j 0 i represents the
amplitude that emits a particle, and h p ; C j 0 i represents the amplitude that
emits an antiparticle. p
It is advisable to keep the factors 2m d!p in the amplitudes in applications
in order not to waste time guessing numerical multiplicative factors in physical
applications.
The moral of the above analysis is this. For a general interacting theory, we
obtain an expression for h 0C j 0 i as in (3.3.30), where now the expression
corresponding to W21 may be expanded in powers of the source functions.
Transition amplitudes for arbitrary processes are then extracted, from a unitarity
sum, and are given by the coefficients of these source functions written as
in (3.3.39), (3.3.40), (3.3.41), and (3.3.42), with the sources appropriately arranged
causally to reflect the actual process where the particles in question are emitted
and detected by these sources. The procedure will be spelled out in detail in
Sect. 5.8, (5.8.14), (5.8.15), (5.8.16), (5.8.17), (5.8.18), (5.8.19), (5.8.20), and
(5.8.21). Applications of these will be given when considering interacting theories.
See also (3.2.16), (3.2.19), (3.2.26), (3.3.15), (3.4.1) and Problem 3.6 for simple
demonstrations. More involved non-linear theories will be dealt with later.
3.4 Coulomb Scattering of Relativistic Electrons
We have seen in (3.2.19) that, to the leading order in an external electromagenetic

potential A .x/, the transition amplitude of an electron scattering off such a field,
with p ! p 0 0 , for p 0 ¤ p , is given by
hp 0 0 j p iA D i e u.p 0 ; 0 / A . p 0 p/ u.p; /; (3.4.1)

where
Z
A .Q/ D .dx/ e i Q x A .x/: (3.4.2)
For Coulomb scattering (A0 D )

Z e2 Z e2
.e A .x// D ;0 ; e A .Q/ D Œ 2 ı.Q 0 / 0 ; (3.4.3)
4jxj jQj2
Q D p 0 p, and where Zjej is the charge of the scattering center. Also
1 1
D ; p p 0 D jpj jp 0 j cos : (3.4.4)
0
jp p j 2
4jpj sin2 2
2
R
Using the formal Fermi rule Œ2 ı. p 0 p 0 0 /2 ! Œ2 ı. p 0 p 0 0 / dx 0 , we
find for the transition probability for the process in question from (3.4.1), per unit
time, the expression
m d3 p m d3 p 0 Œ 2 ı. p 0 0 p0 / Z 2 e4
j u.p 0 ; 0 / 0 u.p; / j2 : (3.4.5)
p 0 .2/3 p 0 0 .2/3 16jpj4 sin4 2
Averaging over the initial electron spin projections =2 D ˙1=2, and summing
over the final ones, we obtain for the last term in the above equation, for p 0 D p 0 0 ,6
1 X 1
j u.p 0 ; 0 / 0 u.p; / j2 D Tr Œ . p C m/ 0 . p 0 C m/ 0
2 0
8m2

1
D 2 . p 0 /2 cos 2 C m 2 sin2 ; (3.4.6)
m 2 2
where in writing the first equality, we have used Eq. (I.21) for the projection
operation, and in the second equality we have used the properties of the gamma
matrices (see Appendix I).
Since the initial velocity of the electron may be written as v D p=p0 , the incident
flux is given by .d3 p=.2/3 /.jpj=p 0/ in a unit of volume. Using the relations
p00
d3 p 0 D d˝jp 0 j2 djp 0 j; ı. p 0 0 p0 / D ı.jp 0 jjp j/; jp 0 j2 D . p 0 0 /2 m2 ;
jp 0 j
(3.4.7)
6
See Problem 3.7.
3.4 Coulomb Scattering of Relativistic Electrons 95
we have for the differential cross section of the process

Z
d 1
djp 0 j ı.p 0 p/ 2 Z 2 e4 h 0 2 i
D 3 4
. p / cos 2 C m 2 sin2 ; (3.4.8)
d˝ 0 .2/ 16 jp j4 sin 2 2 2
which integrates out to

2 h
d Z 2 ˛2 p0 m2 i
D cos 2 C sin2 ; (3.4.9)
d˝ 4 sin4 2 jp j2 2 0
.p / 2 2
and
e2
˛ D ; (3.4.10)
4
defines the fine-structure constant. The expression in (3.4.9) is known as the Mott
differential cross section. It may be equivalently rewritten as
d ˇˇ Z 2 ˛2 h
2 2
i
ˇ D 1 ˇ sin ; (3.4.11)
d˝ Mott 4 p 2 ˇ 2 sin4 2 2
where we have used
m mˇ
p0 D p ; jp j D p : (3.4.12)
1 ˇ2 1 ˇ2
with ˇ denoting the speed of the electron .c D 1/.

We note that since d˝ D 2 sin d, we encounter a divergence in the total
cross section for ' 0, corresponding to the forward direction p ' p 0 . Also
note that (3.4.11) develops a singularity at low energies jpj ' 0, referred to as an
infra-red divergence. We will re-consider these points together with the so-called
radiative corrections to this scattering process within QED later on in Sect. 5.12.
The familiar Rutherford formula follows directly from (3.4.9) by considering the
non-relativistic (NR) limit

d Z 2 m2 .e2 =4/2
D ; (3.4.13)
d˝ NR 4jp j4 sin4 2
as expected.
3.5 Spin & Statistics and the Dirac Quantum Field;

Anti-Commutativity Properties Derived
Every student who has ever heard of spin 1/2 is aware that the corresponding field
theory should be quantized with anti-commutations relations. This is a special case
of the Spin & Statistics Connection which is the subject matter of Sect. 4.5. As
we will see, this property of Spin 1/2 is already inherited in the Dirac equation
itself, and establishing the underlying statistics of the Dirac quantum field is
straightforward and follows by a simple application of the main result obtained in
Sect. 4.5, and it is conveniently treated here at this stage. We work in the Majorana
representation of the gamma matrices (see (2.3.5), (2.3.6)).
One of the most important facts of the Majorana representation is that . =i/ =
. =i/, i.e., the matrices =i are real. (See Sect. 2.3).

Define the 8 8 matrices

0
D ; (3.5.1)
0
and write the Dirac quantum field in terms of a real and imaginary parts as
1 1 2

a D p a Ci a ; a D 1; 2; 3; 4: (3.5.2)
2
We introduce the 8 component real field

> D 1
1
1
2
1
3
1
4
2
1
2
2
2
3
2
4 : (3.5.3)
Similarly, we write the external source as
1
a D p .1a C i 2a /; a D 1; 2; 3; 4; (3.5.4)
2

> D 11 12 13 14 21 22 23 24 : (3.5.5)
The Dirac equation then takes the form

@
C m D ; . / D ; (3.5.6)
i
3.5 Spin & Statistics and the Dirac Quantum Field; Anti-Commutativity. . . 97
and note the complex conjugate property of just stated. This is a first order
equation. In this reformulation, the following replacements are then made

a Œ 0 ab b ! A Œ 0 AB B ; (3.5.7)

a Œ 0 ab b ! A Œ 0 AB B ; (3.5.8)
a; b D 1; 2; 3; 4; A; B D 1; : : : ; 8.
We note that the matrices, Q i 0 , and 0 satisfy the following key
properties (the i factor in Q is chosen for convenience)
.Q / D Q .Q /> D C Q ; . 0 /> D 0 : (3.5.9)
Note the signs on the right-hand sides of these equations. The underlying statistics
would be different if the signs of the last two equalities were reversed. According
to the underlying theory of the Spin & Statistics Connection (Sect. 4.5), these
inherited signs in the Dirac equation imply that the field components A satisfy
the following anti-commutation relations for x 0 0 D x 0 :
.Q0 /AB fC .x/; B .x 0 /g D i ıCA ı 3 .x x 0 /; where Q0 D i I; (3.5.10)
and, the anti-symmetry property of 0 , in particular, implies that is a Grassmann

variable and anti-commutes with .
From the definition of in terms of its components in (3.5.3), these anti-
commutations, in turn, lead to the anti-commutation relations of the Dirac field
components for x 0 0 D x 0 :
0 0
f a .x/; b .x /g D ıab ı 3 .x x 0 /; f a .x/; b .x
0
/g D 0; f
a .x/; b .x /g D 0:
(3.5.11)
One may, of course, use Pauli’s Fundamental Theorem (see, e.g., [6]), to infer
that the anti-commutation relations in (3.5.11) are valid in other representations, by
the application of the necessary G matrices as follows: ! G G (see (2.3.2),
(2.3.4), (I.1)) to the quantum field. The Grassmann variable property of and its
anti-commutativity with , in turn, imply the Grassmann variable property of , ,
and their anti-commutativity with the Dirac field and its adjoint.
In the absence of external sources, we may also use the general expression for
.x/ in Problem 3.8, with the coefficients a.p; /; b .p; / promoted to quantum
variables, and the first anti-commutation relation in (3.5.11) to find the following
anti-commutation relation valid at all times (see Problem 3.8)

0 @ 1
f a .x/; b .x /g D C m .x x 0 /; (3.5.12)
i ab i
where
Z
0 d3 p ip.xx 0 / sin p 0 .x 0 x 0 0 /
.x x / D e ; (3.5.13)
.2/3 p0
and reduces to the previous relation for x 0 0 D x 0 , upon multiplying it by 0bc .
3.6 Electromagnetic Current, Gauge Invariance

and h 0C j 0 i with External Electromagnetic Field
As a preparation for the main study of this section, re-consider the Dirac equation
in the presence of an external source .x/

@
Cm .x/ D .x/: (3.6.1)
i
As we have seen in Sect. 3.1, (3.1.8), the matrix element h 0C j .x/ j 0 i, is given
by
Z
h 0C j .x/ j 0 i D .dx 0 / SC .x x 0 /.x 0 / h 0C j 0 i; (3.6.2)
where SC .x x 0 / is the fermion propagator in (3.1.9). As discussed at the

of Sect. 3.3, we will learn from a direct application of the Quantum Dynamical
Principle in Sect. 4.6, that the matrix element h 0C j .x/ j 0 i, may be generated
by a functional differentiation of h 0C j 0 i with respect to .x/ as follows
•
. i/ h 0C j 0 i D h 0C j .x/ j 0 i; (3.6.3)
•.x/
which upon comparison with (3.6.2) gives

Z
•
. i/ h 0C j 0 i D .dx 0 / SC .x x 0 /.x 0 / h 0C j 0 i: (3.6.4)
•.x/
The solution is given by

h Z i
h 0C j 0 i D exp i .dx/.dx 0 / .x/ SC .x x 0 /.x 0 / : (3.6.5)
This is normalized to unity for .x/ D 0, .x/ D 0, in which case h 0C j 0 i

! h0 j 0i, where j 0i is the so-called Fock vacuum state , i.e., the vacuum state in
the absence of the external source (in the absence of any interaction).
3.6 Electromagnetic Current, Gauge Invariance and h 0C j 0 i with External. . . 99

@
eA .x/ C m .x/ D .x/: (3.6.6)
i
The purpose of this section is to derive the expression for h 0C j 0 iA , for this
system. As will be seen throughout this book, the vacuum persistence amplitude
h 0C j 0 iA , contains much useful information and gives rise to the actual solution
of an underlying quantum field theoretical problem.
There are several reasons why we consider this system at this stage. It provides a
simple interacting system involving fermions. Also it provides an excellent training
for the direct application of the rules of the Quantum Dynamical Principle developed
in Sect. 4.6, and to appreciate its power in applications and witness the simplicity
of its underlying rules before getting involved in its intricate details. Before spelling
out these rules, as we apply them to the system described by (3.6.6), we introduce
the electromagnetic current of the system.
To the above end, we write the equations for , and , in terms of components,
as follows
.@ /a i . /a e A C i m a D i a ; (3.6.7)

.@ /a i . /a e A C i m a D i a : (3.6.8)
The combination [ a .3.6.7/ .3.6.8/ a .3.6.7/ a C a .3.6.8/], gives

e
@ j .x/ D i e .x/.x/ .x/ .x/ ; j .x/ D Œ a .x/; .

/a .x/; (3.6.9)
2
where j .x/ defines the electromagnetic and is conserved in the absence of the
external source. In writing (3.6.9), we have used the Grassmann variable property
of , , that the latter anti-commute with the Dirac field and its adjoint (Sect. 3.5).
The commutator in the definition of the current is of significance, as under
charge conjugation (see Sect. 3.2), it changes sign, as an electromagnetic current
should. Recall that under charge conjugation, the Dirac quantum field transforms as
(see (3.2.29))
> >
! C ; ! C; a . /a $ . /a a: (3.6.10)
where C is the charge conjugation matrix, defined in (I.3), satisfying (2.3.8), (2.3.9).
Hence
j .x/ ! j .x/; (3.6.11)

under charge conjugation. It is common in the literature to write the current naïvely
as e with the understanding that it should be defined as a commutator as
given above in (3.6.9) to ensure the correct transformation given in (3.6.11).
We now spell out some basic facts that we will learn in applying the Quantum
Dynamical Principle (Sect. 4.6) to the system described by (3.6.6):
Lessons from the Quantum Dynamical Principle:
•
. i/ h 0C j 0 i D h 0C j .x/ j 0 i; (3.6.12)
•.x/
•
.i/ h 0C j 0 i D h 0C j .x/ j 0 i; (3.6.13)
•.x/
Z
@
e h 0C j 0 i D i .dx/ h 0C j j .x/ j 0 i A .x/; (3.6.14)
@e
• •
. i/ .i/ 0
h 0C j 0 i D h 0C j .x/ .x 0 / C j 0 i; (3.6.15)
•.x/ •.x /
where .: : :/C defines the time-ordered product of Fermi fields

0

a .x/ b .x C
D ™.x 0 x 0 0 / a .x/ b .x
0
/ ™.x 0 0 x 0 / b .x
0
/ a .x/; (3.6.16)
and ™.x 0 / D 1 for x 0 > 0, ™.x 0 / D 0 for x 0 < 0. Note the minus sign
between the two terms on the right-hand side of the equation that arises due to
the Fermi character of the Dirac field (Sect. 3.5). The anti-commutativity of Fermi
fields within a time-ordered product is directly inferred from the Grassmann variable
character of the sources and the functional derivative operations with respect to them
in (3.6.15).
Using (3.6.12), and recalling (3.2.13), we may write
Z
•
. i/ h 0C j 0 i D h 0C j .x/ j 0 i D .dx 0 / SC
A
.x; x 0 / .x 0 /h 0C j 0 i;
•.x/
(3.6.17)
where we have omitted the “subscript” A in h 0C j .:/ j 0 iA for simplicity of the
notation, and (see (3.2.11))

@
e A .x/ C m SC
A
.x; x 0 / D ı .4/ .x x 0 /: (3.6.18)
i
As in (3.6.3), (3.6.4), (3.6.5), we may integrate (3.6.17), to obtain

h Z i
h 0C j 0 i D h 0C j 0 i.e/ exp i .dx/.dx 0 /.x/SC
A
.x; x 0 /.x 0 / ; (3.6.19)
3.6 Electromagnetic Current, Gauge Invariance and h 0C j 0 i with External. . . 101
where the amplitude
h 0C j 0 i.e/ D h 0C j 0 i jD0;D0 ; (3.6.20)
and depends on eA .

Now we use (3.6.15) and (3.6.19), to conclude that for D 0, D 0,
0

h 0C j a .x/ b .x /
C
j 0 i.e/
D i .SC
A
/ab .x; x 0 /: (3.6.21)
h 0C j 0 i.e/
where .x/ (similarly ) within this time-ordered product satisfies the Dirac
equation (3.6.6) in the external electromagnetic field, in the absence of external
source and, as already indicated in (3.6.20), h 0C j 0 i.e/ denotes the vacuum
persistence amplitude in the absence of (i.e., for D 0; D 0).
We already know from the anti-commutation relation of the Dirac field and
its adjoint in (3.5.12), singularities may develop in defining an expression like
Œ a .x/; . /a .x 0 / at coincident points x D x 0 . Accordingly, one may formally
define the current7 as a limit of
! 0

e

j .x;
/ D a xC . /a x . /a x xC ;
2 2 2 2 a 2
(3.6.22)
Obviously the right-hand side of (3.6.22) reduces to the commutator in (3.6.9) in the
just mentioned limit. By initially choosing the arguments of .x/; .x 0 / in (3.6.22)
at non-coincident points, however, one destroys gauge invariance as we now discuss.
This, in turn, will allow one to define a gauge invariant current at the outset.
To ensure the generation of a gauge invariant electromagnetic current, we
proceed as follows. Under a gauge transformation of the electromagnetic potential,
A ! A C @ .x/, where .x/ is arbitrary, the field equation (3.6.6)
remains invariant provided the spinor .x/ transforms simultaneously as .x/ !
exp i Œe.x/ .x/, and, in turn, a .x/ b .x 0 / transforms as
0 0

a .x/ b .x / ! a .x/ b .x / exp ie .x/ .x 0 / : (3.6.23)
Accordingly, one may define the electromagnetic current as the limit of8
Z

xC
j .x/ D j .x;
/ exp i e d A ./ ;
! 0; (3.6.24)
x
7
In order to distinguish between an external, i.e., c-number, current and a quantum current, as given
in (3.6.22), we use the notation with small j for the latter, and capital J for the former in this
book.
8
See Problem 3.10 where the line integral over is carried out showing its dependence on
, and
for more details see also Appendix IV at the end of the book.
where x˙ D x ˙
=2, and is gauge invariant at the outset. This method is known as
Schwinger’s point splitting method. The exponential of the line-integral involving
the vector potential is referred to as the Schwinger line-integral.
The point of interest is the matrix element of the current between the vacuum
states in (3.6.14). From (3.6.21), (3.6.23) we may, in a limiting sense, write
Z xC
h 0C j j .x/ j 0 i.e/

D i e Tr Av Œ SC .x ; x C / exp i e
A
d A ./ ;
h 0C j 0 i.e/ x
(3.6.25)
where Av stands for an average over
0 > 0 and
0 < 0, with
in a space-like
direction. Here Tr denotes a trace over the spinor indices of gamma matrices.
To emphasize the dependence of SC A
.x; x 0 / on the combination eA , it is
0
convenient to use the notation SC .x; x I eA/ for the former. With the boundary
condition, h 0C j 0 i.e/ D 1; for e D 0, we may use (3.6.14) to integrate over e
to obtain
h 0C j 0 i.e/ D exp Œ i W ; (3.6.26)

Z Z
e
xC
iW D de 0 .dx/ Tr Œ SC .x ; x C I e 0 A/ A exp i e d A ./ ;
0 x
(3.6.27)
and an average taken of the limits
0 ! C0,
0 ! 0 is understood.
In Appendix IV, at the end of the book, i W is determined to arbitrary orders
in e A in the limit
! 0. It is shown that the
! 0 limits give rise to a
modification from the naïve expression only up to fourth order in the charge e.
In particular, to second order in e A (IV.27), in the just mentioned appendix,
gives
ˇ Z
ˇ e2 .dQ 1 / .dQ 1 /
i Wˇ D .2/4 ı .4/ .Q 1 C Q 2 /
2nd order 2 .2/4 .2/4
A 1 .Q 1 / A 2 .Q 2 / L 1 2 .Q 1 ; Q 2 /; (3.6.28)
Z
.dp/
Q Q
L1 2 .Q ; Q/ D 4
Tr 1 SC . p / 2 SC . p C /
.2/ 2 2

1 2 1 @ @ @ 1 @ 1
CQ Q Tr Œ SC . p/ C Tr Œ SC . p/ ;
24 @p1 @p 2 @p 2 @p 2
(3.6.29)
which with (3.6.27) providing a contribution to the vacuum-to-vacuum transition

amplitude in (3.6.26), in the presence of the external electromagnetic field A .
3.7 h 0C j 0 i.e/ in the Presence of a Constant F Field and Effective Action 103
The following tensor is referred to as the vacuum polarization tensor, to lowest

order in e,
˘ .Q/ D i e2 L1 2 .Q ; Q/; (3.6.30)
and will find applications in Chap. 5. It is evaluated in Appendix A of this chapter,

and the following expression emerges for it
˘ .Q/ D . Q2 Q Q / ˘.Q2 /; (3.6.31)
Z Z 1 !
e2 1 .dp/ 1 h Q2 i
2
˘.Q / D C 6 dz z.1 z/ ln 1 C 2 z.1 z/ :
12 2 i 2 Œ p 2 C m 2 2 2 0 m
(3.6.32)
This expression will be used in Sect. 5.10.2 in studying the photon propagator.
The presence of the last two terms in (3.6.29) ensure the gauge invariance of
the formulation and leads necessarily to a transverse expression for ˘ .Q/, in
response to a gauge transformation A .Q/ ! A .Q/ C Q .Q/ in (3.6.28),
Q ˘ .Q/ D 0; ˘ .Q/ Q D 0: (3.6.33)
In the next section, we also provide another expression for h 0C j 0 i.e/ , for
special external electromagnetic potentials A , which turns out to be useful for
applications as to the one given in Sect. 3.8 in pair production by a constant electric
field.
3.7 h 0C j 0 i.e/ in the Presence of a Constant F Field

and Effective Action
An expression for h 0C j 0 i.e/ is readily obtained for the Dirac equation in the
presence of a classical constant F Field, or effectively constant over sufficiently
extended regions of spacetime. An interesting application of this is that the quantum
nature of the fermion field adopted here leads to a modification of the Maxwell
Lagrangian density giving rise to an effective one involving higher powers of the
Faraday tensor F and of its dual to be defined shortly. This will be the subject
matter of this section. Another interesting application is that it provides an explicit
expression for the decay of the system initially in the vacuum state leading to eC e
pair creation by a constant electric field. This will be developed in the next section.
Consider the response of h 0C j 0 i.e/ to variations of Œ eA .x/ , which

from (3.6.14) is given by
Z
1
•h 0C j 0 i.e/ D i .dx/ h 0C j Œ a .x/; .

/a .x/ j 0 i •Œ e A .x/; (3.7.1)
2
where the current is defined in (3.6.9) by j D e Œ a .x/; . /a .x/=2.

At this stage it is convenient to introduce a matrix notation in spacetime as well.
For the propagator SA .x; x 0 / S.x; x 0 I eA/, we write
A
SC .x; x 0 / D h x j SC
A
j x 0 i; h x j 1 j x 0 i D ı .4/ .x x 0 /: (3.7.2)
The differential equation satisfied by SC .x; x 0 I eA/ in (3.6.18) may be rewritten in

matrix form as
1 @
Œ ˘ C m SC
A
D 1; A
SC D ; ˘ D p eA ; p D :
˘ C m i
(3.7.3)
From the definition h 0C j Œ a .x/; . /a .x/ =2 j 0 i.e/ D i SC A

h 0C j 0 i.e/ ,
in (3.6.21), (3.6.22), and the fact that •Œ e A D • Œ ˘ , (3.7.1), in matrix form,
becomes
•h 0C j 0 i.e/ D Tr Œ • .˘ / SC
A
h 0C j 0 i.e/ : (3.7.4)
where Tr stands for a trace over spinor indices as well as spacetime coordinates.
From (3.7.3), the propagator may be re-expressed as
Z 1
1
A
SC D . ˘ C m/ D . ˘ C m/ ds exp Œ s.m2 . ˘ /2 / :
m2 . ˘ /2 0
(3.7.5)
Here it turns out more convenient to use this representation than the one with
i s.m2 i
/ in the exponential. We use the fact that the trace of an odd number
of gamma matrices is zero. Also for a matrix M, • eM ¤ •M eM , in general,
2
but Tr Œ • eM D Tr Œ •M eM .9 Finally note that for a matrix A: Tr Œ • eA D
2 2
Tr Œ .•A A C A •A/ eA D 2 Tr Œ •A A eA . Thus we may write
h i
Tr • .˘ /. ˘ C m/ exp Œ s.m2 . ˘ /2 /
1 h
i
D Tr • .˘ /2 exp s.m2 . ˘ /2 / ; (3.7.6)
2
9
See Appendix B of this chapter.
from which (3.7.4) takes the convenient form

Z 1
•h 0C j 0 i.e/ 1 2 2 2
D Tr • .˘ / ds exp Œ s.m . ˘ / / : (3.7.7)
h 0C j 0 i.e/ 2 0
Again from the property Tr Œ • eM D Tr Œ •M eM , this finally gives us the useful

expression
Z 1
•h 0C j 0 i.e/ 1 ds 2 2
D • Tr exp Œ s.m . ˘ / / : (3.7.8)
h 0C j 0 i.e/ 2 0 s
The formalism will be gauge invariant if the results obtained are expressed in
terms of F and possibly in terms of its dual
1
F D
F : (3.7.9)
2
Here it is important to recall the expressions for the two invariants, as the scalar and
the pseudo-scalar, respectively, of the Maxwell field,
1 1 1
F F D . E2 B2 /; F F D E B: (3.7.10)
4 2 4
The subsequent analysis simplifies quite a bit if we first work in a frame, where E
and B are parallel. In this case the second expression above would read j E j j B j.
The general case may be then readily inferred.
To the above end, we consider a vector potential in the form
A D .0; 0; x1 B; x 0 E/; (3.7.11)
giving an electric and magnetic fields E and B, respectively, in the z-direction.

Using the identity D C Œ ; =2, we may write
Œ ; Œ ˘ ; ˘
m2 . ˘ /2 D m2 C .˘ /2 ; Œ ˘ ; ˘ D i eF :
2 2
(3.7.12)
On the other hand, with
F30 D F03 D F 03 D E; F 12 D B D F 21 ;
as the only non-vanishing components, we have

Œ ; 0 3 3 0
ise F D i s e E C se B; (3.7.13)
4 3 0 0 3
involving two commuting matrices.

In reference to the second term on the right-hand side of the first equality
in (3.7.12), the trace Tr, over spinor indices, of the exponential of the expression
in (3.7.13) is easily carried out by expanding the exponential leading to
n h 3 io
0 3 0
Tr exp i s e 3 E C se B D 4 cos.s j eE j/ cosh.s j eB j/:
0 0 3
(3.7.14)
Also
h p 2 2 2 2 i h p 3 2 2 2 i
˘ ˘ D . p12 C x1 e B C . p 0 /2 C x 0 C eE ; (3.7.15)
eB eE
which, with p0 as the generator of translation of x0 and p1 as the generator of

translation of x1 , may be rewritten as
h p 2 p1 i h p 3p 0 i
exp i exp i
eB eE
h i
. p1 /2 C .x1 /2 e2 B 2 . p0 /2 C .x0 /2 e2 E2
h p p0 i h p p i
3 2 1
exp i exp i : (3.7.16)
eE eB
Accordingly, in reference to (3.7.8) we may write
h i 2
Tr exp Œ s.m 2 . ˘ /2 / D 4 e s m cos.s j eE j/ cosh.s j eB j/ G.s I eE; eB/;
(3.7.17)
where
Z
D ˇ h
ˇ p 2 p1 i h p 3p 0 i
G.s I eE; eB/ D.dx/ x ˇ exp i exp i
eB eE
h i
exp s . p1 /2 C .x1 /2 e2 B 2 . p0 /2 C .x0 /2 e2 E2
h p p0 i h p p iˇ E
3 2 1 ˇ
exp i exp i ˇx : (3.7.18)
eE eB
By using completeness relations (Fourier transforms)

Z
.dp/ .dp 0 /
hx j pih p j j p 0 ih p 0 j xi; (3.7.19)
.2/4 .2/4
with hx j pi D exp Œ i xp , and noting that the expressions within the square brackets
multplying s in the exponential in (3.7.18) are independent of p 2 ; p 3 ; p 02 ; p 30 , we
may use the Fourier normalization condition h p 2 ; p 3 j p 20 ; p 03 i D .2/2 ı. p 2
p 20 / ı. p 3 p 30 /, we may integrate over p 2 ; p 3 ; p 02 ; p 30 to obtain for (3.7.18)
Z Z
jeEjjeB j dp0 dp1
G.s I eE; eB/ D 2
.dx/
.2/ 2 2
D ˇ h iˇ E
ˇ ˇ
p0 ; p1 ˇ exp s . p1 /2 C .x1 /2 e2 B 2 . p0 /2 C .x0 /2 e2 E2 ˇ p0 ; p1
(3.7.20)
where .x 0 /2 D .d=dp 0/2 , .x1 /2 D .d=dp1/2 , in the exponential.

We may use the well known results from the one dimensional harmonic oscillator
problem dealing with two independent systems. For one system, we have formally
a mass 1=2 e2B2 and an angular frequency 2 j eB j, and a second one with mass
2 2
1=2
R e0 E and 1angular frequency 2 j eE j=i, from which we carry out the traces
.dp =2/.dp =2/hp 0; p1 j : jp 0 ; p1 i in (3.7.20) to obtain
Z X1 1
X
jeEjjeB j
G.s I eE; eB/ D .dx/ expŒ 2s i jeEj.n C 1=2/ expŒ 2sjeBj.m C 1=2/
.2/2 nD0 mD0
Z
jeEjjeB j i 1
D .dx/ : (3.7.21)
.2/2 2 sin sjeEj 2 sinh sjeB j
From (3.7.17), the following expression then emerges for (3.7.8)

Z Z 1
•h 0C j 0 i.e/ i ds s m 2

D .dx/ e sjeEj cot sjeEj sjeB j coth sjeB j :
h 0C j 0 i.e/ 8 2 0 s3
(3.7.22)
For a reason that will become clear shortly we normalize h 0C j 0 i.e/ for
e D 0, in this section, by the phase factor
ˇ hZ 1 i hZ 1 i
h 0C j 0 i.e/ ˇeD0 D exp i .dx/ F F D exp i .dx/ E2 B2 ;
4 2
(3.7.23)
in terms of the Maxwell action.

The power series expansions
x2 x4 2x 6
x cot x D 1 ; (3.7.24)
3 45 945
y2 y4 2y 6
y coth y D 1 C C ; (3.7.25)
3 45 945
lead to the expansion
x 2 y 2 .x 2 y 2 /2 C 7x 2 y 2
.x cot x/ .y coth y/ D 1 C : (3.7.26)
3 45
For the subsequent analysis, we rewrite (3.7.22) by adding and subtracting the
second order contribution .se/2 .E2 B2 /=3 within its square brackets, giving
Z Z Z 1
•h 0C j 0 i.e/ .E 2 B 2 / 1 ds sm2
D i .dx/ C i .dx/ e
h 0C j 0 i.e/ 2 8 2 0 s3
h .s e/2 .E 2 B 2 / i
sjeEj cot sjeEj sjeB j coth sjeB j C ; (3.7.27)
3
where is the divergent constant
Z 1
e2 ds sm2
D e : (3.7.28)
12 2 !0 s
Using the boundary condition in (3.7.23), the functional integral of (3.7.27) gives
h Z Z Z 1
.E 2 B 2 / i 1 ds sm2
h 0C j 0 i.e/ D exp i .1 C / .dx/ exp i .dx/ e
2 8 2 0 s3
n .s e/2 .E 2 B 2 / o
sjeEj cot sjeEj sjeB j coth sjeB j C 1 ;
3
(3.7.29)
where the expression between the curly brackets goes to zero for e ! 0.
The expression in the first exponential above multiplied by i is nothing but the
Maxwell action scaled by the factor 1 C , which amounts in redefining .1 C
/A .x/A .x/ ! A .x/A .x/ provided one also redefines, .1 C /1 e2 ! e2 since
everything else in h 0C j 0 i.e/ depends on the product e A, with this product
remaining invariant, i.e., e A ! e A. One may use different notations for these two
re-scaled quantities if one wishes. Needless to say, this process of re-parametrization
is a reconciliation with the initial definition of the Maxwell Lagrangian density. It is
3.8 Pair Creation by a Constant Electric Field 109
necessary even if were finite. Accordingly, the following final expression emerges
.e/
h 0C j 0 i.e/ D e i W ; (3.7.30)
Z Z Z 1
.E 2 B 2 / 1 ds s m 2
W .e/ D .dx/ .dx/ e
2 8 2 0 s3
n .s e/2 .E 2 B 2 / o
sjeEj cot sjeEj sjeB j coth sjeB j C 1 : (3.7.31)
3
In this section, we consider the real part of W .e/ . The imaginary part and its
physical consequences are considered in the next section. As seen from (3.7.31), the
real part of W .e/ gives rise to a modification of the Maxwell Lagrangian density
which, to lowest order, is given by
Z 1
e4 Œ .E 2 B 2 /2 C 7.EB/2 2
L4 D s ds es m : (3.7.32)
8 2 45 0
The integration is elementary, and in terms of arbitrary directions of E, and B, this

becomes
2 ˛2 e2
L4 D Œ .E2 B2 /2 C 7.E B/2 ; ˛ D ; (3.7.33)
45 m4 4
expressed in terms of the invariants in (3.7.10). The new modified Lagrangian
density is known as the Euler-Heisenberg effective Lagrangian (Euler-Heisenberg,
1936), and much additional detailed work on this was also carried out by Schwinger
[7]. The technique used in this section based on a parametric representation of a
Green function is referred to as the Schwinger parametric representation.
3.8 Pair Creation by a Constant Electric Field
The vacuum-to-vacuum transition amplitude h 0C j 0 i.e/ , for the system described

by the Dirac equation in the presence of a constant F field was derived in the
previous section in (3.7.30), (3.7.31). The probability that the vacuum stays the
same in the presence of this external field is j h 0C j 0 i.e/ j2 . It reads
.e/
j h 0C j 0 i.e/ j2 D e 2 ImW ; (3.8.1)
and W .e/ given in (3.7.31).

We may define the decay rate of the vacuum, that is the probability of decay
of the vacuum per unit time by
ImW .e/
D 2 (3.8.2)
T
due to the intervening F field. A decay of the vacuum, means that the system,
initially in the vacuum state, emerges eventually to some other state, and is
accompanied by particle production. Here T, in a limiting sense T ! 1, denotes
the time of operation of the electric field. Since this is a charge conserving process,
eC e pairs may be created by the electric field. More precisely, we determine the
decay rate per unit volume =V, where V, in a limiting sense V ! 1, denotes
the volume of space where the electric field operates. Accordingly, the physical
quantity of interest is
ImW .e/
D 2 (3.8.3)
V VT
with corresponding anticipated small measure of the probability, per unit volume,
per unit time, for pair production. R
Since Re i W .e/ D Im W .e/ , (3.7.31) leads to ( .dx/ D V T)
ImW .e/ 1 n Z 1 ds 2
2 D Re i es m
VT 4 2 0 s3
h .s e/2 .E 2 B 2 / io
sjeEj cot sjeEj sjeB j coth sjeB j C 1 : (3.8.4)
3
For a pure electric field, i.e., E ¤ 0, B D 0, this reads
2 ImW .e/ 1 n Z 1 ds
.s eE/2 o
sm2
D Re i e sjeEj cot sjeEj C 1 :
VT 4 2 0 s3 3
(3.8.5)
There are several ways of evaluating the integral. One way is given in Prob-
lem 3.12. An easier way is to note that cot.sjeEj/ has singularities at s D n=jeEj
coming from the zeros of the sine function. At all other points the integral is real
and hence do not contribute after applying the Re .i/ operation. Note that s D 0
is not a singularity of the integrand, as is easily verified by considering the s ! 0
limit of the integrand, hence n D 1; 2; : : : . Near a singularity s ' n=jeE j, the
integrand multiplying the exp.s m2 / factor behaves like
1
.seE/2 e2 E 2 1
3
sjeEj cot sjeEj C 1 ' 2 2 n :
(3.8.6)
s 3 n s jeEj
3.9 Fermions and Anomalies in Field Theory: Abelian Case 111
The replacement m2 ! m2 i
, is equivalent to rewriting s i
instead of s,
since m2 as well as s are non-negative. Setting s i
D u and using the fact that
n o
1 n
Re i Dı u ; (3.8.7)
u n
jeEj
C i
jeEj
leads to a decay rate per unit time, per unit volume, corresponding to a measure of
pair production in (3.8.5), given by10
1
2 Im W ˛ E 2 X exp Œ n m2 =jeEj
D : (3.8.8)
VT 2 n D 1 n2
The phenomenon described by the above equation is often referred to as the

Schwinger effect. For jeEj m2 , the expression on the right-hand side of (3.8.8)
reduces to .˛ E 2 = 2 / exp. m2 =j eE j/ which is extremely small in practice,
difficult to observe. It is easily verified that 2 Im W=VT D 0 for a pure magnetic
field, i.e., B ¤ 0, E D 0, as the integral (3.8.4) is real with no singularities on the
real axis, and, as expected, no creation of such pairs is possible.
3.9 Fermions and Anomalies in Field Theory: Abelian Case
A classical symmetry does not necessarily hold true quantum mechanically. This is
well illustrated by considering the divergence of the so-called axial vector current
defined by j5 .x/ D 12 Œ a .x/; . 5 /a .x/. This symmetry breaking is due to
the fact that the fermions give rise to closed loops, represented by the trace of
products of Dirac propagators, referred to as radiative corrections, which are absent
classically.
We will learn, in particular, that for a massless fermion that although classically
the current j5 .x/ is conserved, i.e., @ j5 .x/ D 0, quantum mechanically it is not
@ j5 .x/ ¤ 0, and the right-hand of the latter equation is referred to as an anomaly.
What is most interesting is that although the current j5 .x/ leads to divergent
integrals, its derivatives gives rise to a finite expression for anomaly. Moreover, we
will see that the presence of the anomaly, a purely quantum mechanical effect, may
be tested experimentally. We first derive the explicit expression of the just mentioned
anomaly and then show how the anomaly is tested experimentally.
10
For an equivalent contour-integration derivation in the s-complex plane of this result see
Problem 3.12.
3.9.1 Derivation of the Anomaly

5
Consider first a global phase transformation of the Dirac field .x/ ! e .x/
D 0 .x/, where is a constant, in the Dirac equation (3.6.6) in the presence of an
external electromagnetic A .x/ field (in the absence of external sources ; /).
Upon multiplying the Dirac equation from the left by exp. 5 /, we obtain
h i
@
eA .x/ C m exp Œ 2 5 0
.x/ D 0; (3.9.1)
i
where we have used the relation f 5 ; g D 0. That is, the equation does not
remain invariant for m ¤ 0.
In connection to the above observation, consider the divergence of the axial
vector current. To this end, multiply the Dirac equation for from the left by
5 and the Dirac equation for (see, e.g., (3.6.8) for D 0) from the right also
by 5 . The resulting equations may be conveniently written as

5 @ .x/ D i e 5 A .x/ C m 5 .x/ ; (3.9.2)

@ .x/ 5 D i .x/ e 5 A .x/ m 5 ; (3.9.3)
which naïvely lead to @ j 5 .x/ D 2 i mj 5 .x/, where
1 1
j 5 .x/ D Œ a .x/; .
5
/a .x/ ; j 5 .x/ D Œ a .x/; .

5 .x//a :
2 2
For m D 0 the above, formally, leads to a seemingly conservation law @ j5 .x/ D
0. We will see that due to the singularity of the current j5 .x/, defined at coincident
spacetime points of the fields, its divergence is modified giving rise to an anomaly
and, for m D 0, it reads
e2 ˛ˇ
@ j 5 .x/ D " F F˛ˇ : (3.9.4)
16 2
The singularity of the axial vector current arises due to the quantum nature of the
fermion field. To show this, we use Schwinger’s point splitting method to generate
a gauge invariant expression for j 5 .x/.11
To the above end, consider the derivative of the product (x ˙ D x ˙
=2)
1
h Z xC i
P 5 .x;
/ D .x
a C / ; . 5
/ a .x / exp ie d A ./ ; (3.9.5)
2 x
11
See (3.6.24) as well as Appendix IV at the end of the book which will be quite useful to follow
the present section.
An application of (3.9.2), (3.9.3) gives for m D 0,

5

@ a .x C / ; . /a .x / D i e a .x C / ; . 5 /a .x /

A x C A x : (3.9.6)
and

A x C A x D
@ A .x/ C O.
3 /: (3.9.7)
On the other hand, upon setting D x C .

=2/, 1 1, this leads to
Z xC Z xC
3
@ exp Œ i e d A ./ D i eŒ
@ A .x/ C O.
/ exp Œ i e d A ./ :
x x
(3.9.8)
Thus we obtain,

@ P 5 .x;
/ D i e P 5 .x;
/
F C O.
3 / : (3.9.9)
For the matrix element of the divergence of the axial current between the vacuum
states we may then write for m D 0, as a limit
h 0C j @ j 5 .x/ j 0 i.e/

.e/
D e Av
F C O.
3 /
h 0C j 0 i
Z xC
Tr Œ SC A
.x ; x C / 5 exp Œ i e d A ./ ; (3.9.10)
x

0 > 0 and
0 < 0, with
in a space-like
direction. As for the electromagnetic current in (3.6.25), it is expressed in terms of
A
the propagator SC in (3.2.11).
We will show that only the single power of
contributes in (3.9.10) in the limit
! 0.
To the above end, we may infer from (3.2.12) in a power series in eA
Tr Œ SC
A
.x ; x C / 5 D Tr Œ SC .x x C / 5 C
X Z
C .e/N1 .dx 2 / : : : .dx N /A 2 .x2 / : : : A N .xN /
N 2
Tr Œ SC .x x N / N SC .xN xN1 / N1 : : : 2 SC .x 2 x C / 5 :

(3.9.11)
Since Tr Œ SC
A
.x ; x C / 5 D Tr Œ 5 SC
A
.x ; x C /, we may equivalently
rewrite (3.9.11) as
Tr Œ 5 SC
A
.x ; x C / D Tr Œ 5 SC .x x C / C
X Z
C .e/N1 .dx 2 / : : : .dx N /A 2 .x2 / : : : A N .xN /
N 2
Tr Œ 5 SC .x x 2 / 2 SC .x 2 x 3 / 3 : : : N SC .x N x C / :
(3.9.12)
In (3.9.11), we use the fact that
Tr Œ . : / D Tr Œ .:/> D Tr Œ C 1 .:/> C ;
where C is the charge conjugation matrix (see (I.3)), and the key property 5
D C 1 . 5 /> C , as well as the basic properties
C 1 . /> C D ; C 1 .SC .x x 0 //> C D SC .x 0 x/; (3.9.13)
to rewrite (3.9.11) as the average of the resulting expression for it, just introduced,
and of the one in (3.9.12):
h i 1 h i
A
Tr SC .x ; x C / 5 D Tr 5 Œ SC .xxC / C SC .x C x /
2
X Z
C .e/N1 .dx 2 / : : : .dx N /A 2 .x 2 / : : : A N .xN /
N 2
1h i
F 5 2 N x ; x 2 ; : : : ; x N ; x C C .1/N 1 F 5 2 N x C ; x 2 ; : : : ; x N ; x ;
2
(3.9.14)

F 5 2 N x; x2 ; : : : ; xN ; y
D Tr Œ 5 SC .x x 2 / 2 SC .x 2 x 3 / N SC .x N y/ : (3.9.15)
The importance of the .1/N1 factor in (3.9.14) cannot be overemphasized.

Upon taking Fourier transforms, the following expression results for (3.9.14)
Z
.dp/ Œ ei p
C ei p

Tr Œ SC
A
.x ; x C / 5 D Tr Œ 5 SC . p/ C
.2/4 2
X Z
.dQ2 / .dQN /
C eN1 ::: A .Q2 / : : : A N .QN / ei.Q 2 C:::CQ N / x
N 2
.2/ 4 .2/4 2
Z
.dp/ 1
ip
i P0 Q i
=2 P0
e e C.1/N1 eip
ei Q i
=2
.2/ 2
4

Q 2 2
5 Q2 N Q2
Tr SC . p C / SC . p / : : : SC . p Q 3 : : :Q N / ;
2 2 2
(3.9.16)
properly symmetrized. In reference to (3.9.10), the rest of the analysis rests on the
following:
P0 P0

Œeip
e i Q i
=2
C .1/N1 eip
ei Q i
=2

@ ip
i P0 Q i
=2 P0
D i Œe e .1/N1 eip
ei Q i
=2 ; (3.9.17)
@p
and for a function f . p/, as a product of Dirac propagators,

R with p 0 ! ip 0 , such

that f .p/ ! 1= , 4, for ! 1, .dp/ @f . p/=@p D 0 by Gauss’
Theorem, since the surface area grows like 3 . The equality in (3.9.17) is used as
follows. One integrates by parts over p by transferring i@=@p to the last line
in (3.9.16) involving the trace of products of Dirac propagators.
We multiply (3.9.16) by
, as indicated in (3.9.10), integrate over p by parts
by using, in the process, (3.9.17), to conclude from Gauss’ Theorem that no
contribution to (3.9.10) arises for N 4, for
! 0, since S.p/ D O.1 /. On
the other hand, no contribution arises for N D 3, in the limit
! 0. This is as a
consequence of a net minus sign (.1/N1 ) that results in the middle within the
square brackets, on the right-hand side of (3.9.17) in the process of integrating by
parts. For two powers of
, one of the
’s introduces one additional power of p in
the denominator of p - integral. Hence the term with N D 3 does not contribute.
The very first term in (3.9.16) does not contribute for the simple reason that
TrŒ 5 S. p/ D 0. For all those terms where Gauss’ Theorem leads to no
contribution to the anomaly, one does not have to set m D 0 in their initial
expressions. Clearly only the N D 2 term contributes.
All told, in the limit
! 0, the right-hand side of (3.9.10) becomes
Z
2 .dQ/ iQ x
e F .x/ e A .Q/ I .Q/; (3.9.18)
.2/4
Z h
.dp/ @ 5 Q Q i
I .Q/ D i Tr S. p C / S. p / ; (3.9.19)
.2/4 @p 2 2
in which m is set equal to zero.

By using the identity Tr Œ 5 D 4 i " , where the latter
is totally anti-symmetric, " 0123 D C1, as well as the Feynman parameter
representation given in (II.27) in Appendix II at the end of the book, we may write
h Z
Q Q i 1
4 i " Q p
Tr S. p C / S. p / D
5
dx ;
2 2 0 Œ. p 2 .1
Q
2x//2 C Q2 x.1 x/2
(3.9.20)
where we have used the anti-symmetry of " . If it weren’t for the derivative
@=@p in (3.9.19), the p-integral will diverge. By the application of the derivative
just mentioned, however, renders the integral finite and by shifting the integration
of variable p to p C Q.1 2x/=2 (see Appendix II, (II.14) at the end of the book),
the x - integrand in (3.9.20) then reduces to
4 i " Q 16 i " Q p p
: (3.9.21)
Œ p 2 C Q2 x.1 x/2 Œ p 2 C Q2 x.1 x/3
With p p effectively given by p2 ı =4, (3.9.19) then simplifies to

Z 1 Z
.dp/ " Q Q2 x.1 x/ i
I .Q/ D 4 dx D " Q ; (3.9.22)
0 .2/4 Œ p 2 C Q2 x.1 x/3 8 2
where we have used the integral (II.8) in Appendix II.

From (3.9.18), (3.9.22), we finally obtain for (3.9.10),
h 0C j @ j 5 .x/ j 0 i.e/ e2

D " F F ; (3.9.23)
h 0C j 0 i.e/ 16 2
which is (3.9.4) in the vector space of fermions and anti-fermions. The non-zero
expression on the right-hand side of the above equation is referred to as an anomaly.
It corresponds to the so-called abelian case where the gauge transformation of the
Dirac field, as discussed below (3.6.22), is implemented by a phase transformation
.x/ ! U.x/ .x/; with U.x/ D exp Œ i .x/ ; Œ U.x/; U.x 0 / D 0;

(3.9.24)
i.e., with commuting group elements. The underlying symmetry group of such
phase transformations is referred to as the U.1/ group. In the next section, we
encounter more generalized groups of gauge transformations, with non-commuting
group elements, referred to as non-abelian.
The equation of anomaly in (3.9.23) appeared in an early classic paper of
Schwinger [7, Eqs.(5.24), (3.30)]. Important work on anomalies was also carried
out later in the late sixties in Adler [1], Bell and Jackiw [3], Jackiw and Johnson [5],
and Adler and Bardeen [2].
The additional factor involving the line integral in defining gauge invariant
currents, systematically used above, was introduced by Schwinger [8, 9], and, as
mentioned earlier, is referred to as the Schwinger line-integral.
γ μγ 5 γ μγ 5
S p − Q2 S p + Q2 S p − Q2 S p + Q2
+
γσ γν γν γσ
S p + Q2 S p − Q2
Fig. 3.5 The integrand in (3.9.18)/(3.9.25) contributing to the anomaly may be represented by the
above triangles with vertices as shown connected by massless fermion propagators
Before providing an experimental verification of the anomaly in the next

subsection, we may gain further insight on the nature of the anomaly obtained above
in the following manner. We carry out explicitly the partial derivative @=@p in the
integrand of the anomalous term in (3.9.19). This allows us to write
Z h
.dp/ Q Q Q
I .Q/ D i 4
Tr 5 S. p C / S. p C / S. p /
.2/ 2 2 2
Q Q Q i
C 5 S. p C / S. p / S. p / ;
2 2 2
(3.9.25)
where we have used the identity
@
S. p/ D S. p/ S. p/; (3.9.26)
@p
and the integrand in (3.9.25) may be represented diagrammatically as shown in

Fig. 3.5, involving three vertex points identified by the matrices . 5 ; ; /,
. 5 ; ; /, respectively, in the sum, and the anomaly in (3.9.23) is referred
as an abelian triangle anomaly.
3.9.2 Experimental Verification of the Anomaly: 0 !

Decay
An experimental verification of the anomaly as given in (3.9.23) may be carried out

by considering the decay of the neutral pion to two photons: 0 ! .
To the above end, we present a simple description by replacing 5 in the
axial-vector above by 5 a =2, carrying SU.2/ isospin, where 1 ; 2 ; 3 are
the Pauli matrices, and replace in the vector current by Q, that matches the
2 2 nature of the Pauli matrices, with Q denoting the charge matrix

10
QD : (3.9.27)
00
Since now the matrix, a =2 in each of the two terms in the integrand in (3.9.25)
appears once, while the charge matrix Q appears twice, the anomaly associated
with the divergence of the isotopic axial current @ j5a becomes simply replaced
by
e2 h a i 2
2 a3 e
" F F Tr Q D ı " F F : (3.9.28)
16 2 2 32 2
Hence only the third component develops an anomaly, which is what is needed to
investigate the 0 ! decay. It also multiplies the earlier anomalous factor by
1=2. For a D 3, the anomaly may be rewritten as .e2 =8 2 /" @ A @ A .
Consider a state j k1 e1 ; k2 e2 i with two photons of momenta k1 ; k2 and
polarization vectors12 e1 ; e2 . Working to lowest order in the electromagnetic
charge13 for the process in question, we may write
h k e j A .x/ D e .k/ eikx hvacj; (3.9.29)
and the matrix element of the anomaly hk1 e1 ; k2 e2 j @ j5 3 .x/ j vaci becomes
e2
h k1 e1 ; k2 e2 j @ j5 3 .x/ j vaci D k1 e k2 e " ei.k1 Ck2 /x ; (3.9.30)
4 2
where the 8 2 factor has been replaced by 4 2 in the above coefficient because of
the indistinguishability of the two photons thus introducing a multiplicative factor
of 2.
Using the fact that the quantum field @ J 53 is a pseudo-scalar field like the
neutral pion field 3 itself, the field @ j53 may be considered a source term to the
pion field, i.e.,14
1
h k1 e1 ; k2 e2 j . C m2 / 3 .x/ j vaci D hk1 e1 ; k2 e2 j @ j53 .x/ j vaci;
f
(3.9.31)
12
Polarization vectors for photons will be considered in Sect. 5.2. The underlying details are not
needed here.
13
This allows us to identify F ; F as free quantum fields in the anomaly in (3.9.28)
14
Upon Fourier transform (x ! p) of this equation, the latter reduces formally to what one refers
to as the PCAC relation for p2 0, where PCAC stands for partially conserved axial current.
where 1=f is a coupling parameter.15 Hence

Z
1
hk1 e1 ; k2 e2 j 3 .x/ j vaci D .dx 0 / C .x x 0 / hk1 e1 ; k2 e2 j @ j53 .x 0 / j vaci;
f
(3.9.32)
where
Z 0
.dq/ ei p .xx /
4C .x x 0 / D ;
! C0: (3.9.33)
.2/4 p 2 C m2 i
With the necessary causal arrangement that the pion exists prior to its decay into two
photons, i.e., by taking x 0 < x 0 0 and applying (3.1.9)/(3.1.11), with . p C m/
in them simply replaced by one, and the expression in (3.9.32) becomes
Z Z
i d3 p
h k1 e1 ; k2 e2 j 3 .x/ j vaci D .dx 0 /
f .2/3 2p 0
0
eip .xx / hk1 e1 ; k2 e2j @ j53 .x 0 / j vaci: (3.9.34)
Using the expression in (3.9.30), we may explicitly integrate over x 0 to obtain

Z
e2 d3 p
hk1 e1 ;k2 e2 j 3 .x/ j vaci D i eip x
4 2 f .2/3 2p 0
.2/4 ı .4/ . p k1 k2 / k1 e k2 e " : (3.9.35)
The neutral pion, prior to its decay into the two photons, was created by the pion
field via the equation:
Z
d3 p
3 .x/ j vaci D jp i eip x ; (3.9.36)
.2/3 2p 0
leading from (3.9.35) to

Z
d3 p
h k1 e1 ; k2 e2 j pi eip x
.2/3 2p 0
Z
e2 d3 p
Di eip x .2/4 ı .4/ . p k1 k2 / k1 e k2 e " :
4 2 f .2/3 2p 0
(3.9.37)
15
The reader should recognize, at this stage, that this is just the Klein-Gordon equation with a
source term. See, e.g., Manoukian [6], p. 937.
On the left-hand side of this equation, we recognize hk1 e1 ; k2 e2 j pi as the

amplitude A. 0 ! / of interest, and thus the following expression emerges
for it
2
0 4 .4/ e
A. ! / D i .2/ ı . p k1 k2 / k1 e k2 e " ; (3.9.38)
4 2 f
which is, independently, a general expression for the process in question.

Experimentally,16 e2 =.4 2 f / 0:025 GeV1 , from which a non-vanishing
value is obtained
p for f 93 MeV. What is interesting here is that the decay
constant 2f ' 131MeV may be measured, e.g, from the C ! C decay,
and thus independently determines the anomaly factor e2 =.4 2 f / as obtained from
applying the anomaly in deriving the decay rate 0 ! .
It is quite interesting that a singular behavior, i.e., of divergence, of the axial
vector current, leads to a finite non-zero result (an anomaly) which can be verified
experimentally. This is an example of a good anomaly.17
3.10 Fermions and Anomalies in Field Theory:

Non-Abelian Case
We generalize the equation of anomaly derived in the last section to the so-called
non-abelian case to be defined shortly. Non-abelian gauge fields will be introduced
and discussed in detail in Sect. 6.1 and in other sections following it. Here we
introduce only the very basics to see how the equation of anomaly just mentioned is
generalized to the non-abelian case.18 As before, the result embodied in this section
shows that a classical symmetry does not necessarily hold in the quantum world.
We consider several Dirac fields D . 1 .x/; : : : ; n .x//, satisfying Dirac
equations,

@
g A .x/ C m .x/ D 0; (3.10.1)
i
written in matrix form, where A .x/ is an n n matrix. The mass, m, may, in

general, have a matrix structure as well. We consider a general transformation, by a
16
The value of this constant may be inferred from the decay rate 0 ! [4].
17
There are good anomalies and bad anomalies. An example of a bad anomaly is one which would
destroy renormalizability. For example this would happen in non-abelian gauge theories if quarks,
in general, come in other than three colors (see Sects. 6.4 and 6.15).
18
Some readers may wish to read Sect. 6.1 simultaneously with the present one. In the present
section we supply, however, enough details for the reader to be able to follow it.
3.10 Fermions and Anomalies in Field Theory: Non-Abelian Case 121
unitary matrix U.x/,
.x/ ! U.x/ .x/; (3.10.2)
and infer from (3.10.1), that the latter remains invariant, if the so-called gauge field
A .x/ transforms simultaneously as19
i
A .x/ ! U.x/A .x/U 1 .x/ C U.x/@ U 1 .x/: (3.10.3)
g
The gauge field A .x/ being an n n matrix, may be expanded in terms of a basis
set of matrices, assumed to have the following properties
A .x/ D ta Aa .x/; Œ ta ; tb D i fabc tc ; (3.10.4)
where the fabc are called the structure constants of the group of transformations,
and are totally anti-symmetric in their indices. If the structure constants are zero
the group of transformations is referred to as abelian, otherwise it is called non-
abelian. Non-abelian gauge groups will be studied in detail in Chap. 6. In order not
to confuse the indices just introduced with spinor indices we suppress the latter ones
in this section.
We may pose to consider a concrete case. For example, the group SU(3), consists
of 3 3 unitary matrices of determinant one. Here U in SU.3/ stands for unitary,
n D 3 for the size of the matrices, and S for special, i.e., of determinant one.
In quantum chromodynamics, corresponds to a quark field with three different
colors specified by fields 1 ; 2 ; 3 . The Dirac field will also carry another index
to specify the different flavors of quarks which we will also suppress here. The
mass m will carry a flavor index. The index a introduced above in Aa .x/ would
take 8 values corresponding to eight gluons.
The matrices U.x/ will be expressed as
U.x/ D exp Œ i g ta #a .x/ : (3.10.5)
Due to the nature of the SU.3/ matrices, for example, the so-called generators
ta of transformations, are not only Hermitian but also traceless since det U.x/ D
expŒi g #a .x/ Tr ta .
The matrix elements of the massless free Dirac propagator, i.e., for g D 0, in the
momentum representation, are simply given by
p
S. p/ D : (3.10.6)
p2 i
19
Note that @ .UU 1 / D 0 implies that U.@ U 1 / D .@ U/ U 1 .
As in the abelian case, one naïvely obtains the conservation of the axial vector
current j5 .x/ D Œ .x/; . 5 /.x/ =2 from (3.10.1) for m D 0.
Consider the expression .x ˙ D x ˙
=2/
Z
1
xC
P5 .x;
/ D .x C /; 1 C i g d A ./ . 5 /.x / ; (3.10.7)
2 x
The infinitesimal gauge transformations of the fields are given by
.x/ ! .x/ C i g •#a .x/ ta .x/; (3.10.8)

.x/ ! .x/ i g .x/•#a .x/ ta ; (3.10.9)
A .x/ ! A .x/ C @ •#a .x/ ta C ig •#a .x/ Œ ta ; A .x/ : (3.10.10)
The line integral is proportional to

, and is given by (see (3.9.8))
Z xC
d A ./ D
A .x/ C O.
3 /: (3.10.11)
x
We will see that this expression, with the single power of

will be sufficient
to derive the full expression of the anomaly, and that higher powers of
do
not contribute. It is readily verified that P5 .x;
/ is gauge invariant for an
infinitesimal gauge transformation implemented by an infinitesimal ı# a .x/ up to
first order in
, which will eventually be taken to go to zero.
Upon taking the divergence of P5 .x;
/, we obtain20
@ P5 .x;
/ D i g Œ .xC /; K .x;
/ . 5 /.x / ; (3.10.12)

K .x;
/ D
@ A .x/ ig
A .x/A .x/ A .x/A .x/ C
@ A .x/:
(3.10.13)
Therefore upon introducing the matrix field
G .x/ D @ A .x/ @ A .x/ i g Œ A .x/; A .x/ ; (3.10.14)
the following equation results

@ P5 .x;
/ D i g
.x C /; G .x/ . 5 /.x / : (3.10.15)
From (3.10.4), G .x/ may be also expanded in terms of the matrices ta ,
G .x/ D ta Ga .x/;

Ga .x/ D @ Aa .x/ @ Aa .x/ C gfabc Ab .x/Ac .x/: (3.10.16)
20
See also (3.9.8), (3.9.9).
3.10 Fermions and Anomalies in Field Theory: Non-Abelian Case 123
The matrix element between the vacuum states of the divergence of the axial
vector current then takes, in a limiting sense, the form
h 0C j @ j5 .x/ j 0 i.g/

5
D g G a .x/ Av
Tr Œ S A
C .x ; x C / ta ;
h 0C j 0 i.g/
(3.10.17)
expressed in terms of the Fourier transform of the propagator in (3.10.6), where Av
stands for an average over
0 > 0 and
0 < 0, with
in a space-like direction. In
h 0C j 0 i.g/ , the products A .x/, in the abelian theory, being multiplied by the
free Dirac propagators, are simply replaced by ta Aa .x/, in the non-abelian case,
and the coupling parameter (charge) e by g.
Due to the simple diagonal structure of the free Dirac propagator in (3.10.6)
with respect to the indices that label the matrix elements of the ta matrices, and
that the ta matrices are of different nature than of the Dirac matrices, i.e., they
commute with them, we can directly use the expression obtained for the abelian
case in (3.9.16), (3.9.17), and factor out the trace of the products of the ta matrices
as follows, to obtain, in a limiting sense for
! 0,
Z
.dp/ Œeip
eip
@

Tr Œ SC
A
.x ; x C / ta 5 D i Tr.ta / TrŒ 5 SC . p/
.2/ 4 2 @p
X Z
.dQ 2 / .dQN /
i g N1 Aa .Q2 / : : : AaN N .QN / ei .Q2 C:::CQN /x
N2
.2/4 .2/4 2 2
Z
.dp/ 1 P0 P0
4
Œ Tr. ta ta2 : : : taN /eip
e i Qi
=2 Tr. taN : : : ta2 ta /.1/N1 eip
ei Qi
=2
.2/ 2
@ h Q2 2 Q2 Q2 i
Tr 5 SC p C SC p : : : N SC p Q3 : : :QN ;
@p 2 2 2
(3.10.18)
symmetrized properly. Here we have already included the

multiplicative factor,
as shown on the left-hand side of the equation. Moreover, we have integrated over p
by parts to replace
by i @=@p , with the latter operating to the right, and thus
introduced an additional minus sign in the middle within the square brackets in the
first and third lines involving exp.ip
/ and exp.ip
/.
Clearly, we may invoke Gauss’ Theorem to infer that all the terms with N 4
are zero, because of the @=@p factor and by simple power counting of powers of p.
The first term is obviously zero .Tr Œ ta D 0/.
The N D 2 term may be inferred directly from the abelian case in (3.9.16)
and (3.9.17) to be
" "
g Tr.ta ta2 / 2
@ Aa2 .x/ D g Tr.ta ta2 / .@ Aa2 .x/ @ Aa2 .x//:
8 16 2
(3.10.19)
For two powers of

, a net minus sign arises within the square brackets involving
exp.ip
/ and exp.ip
/, giving a null result. Gauss’ Theorem guarantees that
higher powers of
do not contribute.
A new contribution, not present in the abelian case, now emerges from the N D
3 term. First note that for two and higher powers of
, one may invoke Gauss’
Theorem to infer that they do not contribute. For the single power of
, as given
in (3.10.18), the N D 3 term reads21
Z
.dQ2 / .dQ3 /
i g2 Aa .Q2 / Aa3 3 .Q3 / ei .Q 2 CQ 3 /x
.2/4 .2/4 2 2

Tr ta Œ ta2 ; ta3 K 2 3 .Q2 ; Q3 /; (3.10.20)
Z
2 3 .dp/ @ 5 Q2 2 Q2
K .Q2 ; Q3 /D Tr S C p C S C p
.2/4 @p 2 2

3 Q2
SC p Q3 :
2
(3.10.21)
The product of the denominators of the Dirac propagators may be combined by

using the Feynman parameter representation given in (II.35) of Appendix II, at the
end of the book, for m D 0
Z 1 Z x1
1 1
D2 dx1 dx2 ;
.pC Q2 2
2
/ .p Q2 2
2
/ .p Q2
2
Q 3 /2 0 0 Œ. p k/2 C M.Q2 ; Q3 ; x1 ; x2 /3
(3.10.22)
1
kD ŒQ2 .1 2x 2 / C 2 Q3 .1 x1 /; (3.10.23)
2
M.Q2 ; Q3 ; x1 ; x 2 / D Q22 x 2 .1 x 2 / C Q23 x1 .1 x1 / C 2 Q2 Q3 x 2 .1 x1 /:
(3.10.24)
By carrying out the differentiation with respect to p , using, in the process,

the integrals in (II.8), (II.9), (II.10), (II.11), and by now the familiar identity
Tr Œ 5 D 4 i " , K 2 3 .Q2 ; Q3 / readily works out to be
" 2 3
K 2 3 .Q2 ; Q3 / D : (3.10.25)
16 2
21
By carrying the differentiation with respect to p in (3.10.21), we generate a diagram involving
four connected fermion propagators referred to as a square diagram contributing to the anomaly as
opposed to the triangle diagram contribution to the anomaly in (3.10.19) involving three fermion
propagators. See also (3.10.25) which, however has exactly the same factor " 2 3 =.16 2 /, with
2 D ; 3 D , as the one in (3.10.19).
Appendix A: Evaluation of L1 2 125
From (3.10.17), (3.10.19), (3.10.20), (3.10.25), and the expression for G
in (3.10.14), we have
h 0C j @ j5 .x/ j 0 i.g/ g2

.g/
D " Tr .G G /; (3.10.26)
h 0C j 0 i 16 2
holding in the vector space of fermions and anti-fermions.

Note that this anomaly is the same as the abelian one in (3.9.23) with F F
replaced by Tr .G G /, and e by g.
The gauge transformation of A .x/ in (3.10.3), implies, in turn, the gauge
transformation of G .x/ in (3.10.14) to be (see Problem 3.14)
G .x/ ! U.x/ G .x/ U 1 .x/: (3.10.27)
Hence Tr .G G / is gauge invariant, and so is the equation of anomaly

in (3.10.26).
The above analysis of the anomaly will be crucial later on in Chap. 6 for the
renormalizability of the standard model. The key point to realize about this at this
point is that the standard model involves axial vector currents, and renormalizability
of the model hinges on the assumption of the absence of such anomalies.
Appendix A: Evaluation of L1 2
The integral L1 2 .Q; Q/ is defined in (3.6.29). Clearly, L1 2 .Q ; Q/ is zero
for Q D 0. Also we will show that
Q1 L1 2 .Q ; Q/ D 0; Q2 L1 2 .Q ; Q/ D 0; (A-3.1)
as required by gauge invariance under the transformation Aj .Qj / ! Aj .Qj / C
1
.Qj / Qj j in (3.6.28). To this end, using the notation SC . p/ D p C m, and the
1 1
identity Q D ŒSC . p Q=2/ SC . p C Q=2/, we note
that when (3.6.29) is
multiplied by Q 2 we obtain for the integrand Tr 1 Œ : , where
Q Q @ 1 @ 3
Œ : D Œ SC p C SC p C Q SC . p/ C Q SC . p/
2 2 @p 24 @p
!
2 Q @ 5
D O SC . p/; (A-3.2)
5Š 2 @p
and note that .Q@=@p/SC . p/ D SC . p/ Q SC . p/. We have used the useful
identities
Z 1
ex 1 D d e x x;
0
::
:
Z 1 Z 5 Z 4 Z 3 Z 2
x2 x3 x4
ex 1 x D d5 d4 d3 d2 d1 e1 x x5 ; (A-3.3)
2Š 3Š 4Š 0 0 0 0 0
to infer that

Q @ Q @ @ 1 @ 3
exp. / exp. / C Q C Q
2 @p 2 @p @p 24 @p
Z 1 Z 5 Z 4 Z 3 Z 2
Q @ Q @ 5
D 2 d5 d4 d3 d2 d1 cosh 1 : (A-3.4)
0 0 0 0 0 2 @p 2 @p
We may thus invoke Gauss’ Theorem to infer that the remainder in (A-3.2)
integrates out to zero. A similar analysis applies for the case when the integral is
multiplied by Q1 thus establishing (A-3.1). This allows us to write

i e2 L1 2 .Q ; Q/ D 1 2 Q2 Q1 Q2 ˘.Q2 /: (A-3.5)

The i factor is chosen for convenience. Also 1 2 Q2 Q1 Q2 ˘.Q2 / is the
lowest order contribution to the so-called vacuum polarization tensor, with L
given (3.6.29). To evaluate ˘.Q2 /, we contract the indices 1 and 2 in (A-3.5).
In particular, (3.6.29), leads us to evaluate the following integral
h Z Q Q i
3.2/ Q ˘.Q / D i e .dp/ Tr SC p
4 2 2 2
S C p C
2 2

1 @ 2 @

C Q Tr SC . p/ Tr SC . p/ SC . p/ : (A-3.6)
24 @p @p
The following integral is easily evaluated from the integrals in (II.8)–(II.9) in

Appendix II at the end of the book:
Z
1 @ 2 @

.dp/
Q Tr SC . p/ D
24 @p @p
Z
m2 m2 p 2 i 2 2
2 Q2 .dp/ D Q: (A-3.7)
. p 2 C m 2 /3 . p 2 C m 2 /4 3
Appendix A: Evaluation of L1 2 127
For the first term in the integral in (A-3.6), we may use the Feynman parameter
representation of the product of two denominators defined in (II.27) to write it, after
carrying out the trace, up to the i e2 factor, as
Z Z 1
8p 2 C 2Q2 16m2
I.Q2 / D .dp/ dz 2 : (A-3.8)
0 Œ p C Q2 .1 2z/ C Q2 z.1 z/ C m2 2
We may refer to the integral (II.23) and make a contraction over and there, to
obtain, in reference to the term with p2 in the numerator in the integrand in (A-3.8)
the following integral
Z
.dp/ p2
D
Œ. p k/2 C M 2 .k2 /2
Z Z
3 i 2 2
2 .dp/ .dp/
k C k M 2 .k2 / 2 2 2 2
C ; (A-3.9)
2 Œ p C M .k / Œ p C M 2 .k2 /
2
Q
k D .1 2z/; M 2 .k2 / D Q2 z.1 z/ C m2 : (A-3.10)
2
We here define the degree of divergence of a 4D-integral as the power of p in the
numerator minus the power of p in the denominator in the integrand plus four. Thus
the degree of divergence of the integral in (A-3.8) restricted to the .2Q2 16m2 /
term divided by the denominator is zero. This allows us (see Appendix II) to shift
the integration variable p, restricted to this term, to p Q.1 2z/=2. Accordingly,
we may use (A-3.9), (A-3.10), to rewrite (A-3.8) as
Z 1
3 Q2
I.Q2 / D 8 i 2 dz .1 2z/2 C
2 4 0
Z Z 1
Œ 2Q2 8m2 4Q2 .1 2z/2 C 8Q2 z.1 z/
.dp/ dz
0 Œ p 2 C Q2 z.1 z/ C m2 2
Z Z 1
1
8 .dp/ dz 2 : (A-3.11)
0 Œ p C Q z.1 z/ C m2
2
Now we may integrate the last integral by parts over z

Z ˇ1
1
1 z ˇ
dz 2 D 2 ˇ
0
2 2
Œ p C Q z.1 z/ C m Œ p C Q z.1 z/ C m ˇ0
2 2
Z 1
z Q2 .1 2z/
dz 2 ; (A-3.12)
0 Œ p C Q2 z.1 z/ C m2 2
to combine its second part with the second integral in (A-3.11) to obtain
Z Z 1
z dz
I.Q2 / D i 2 Q2 C 8 Q2 .dp/
2 C Q2 z.1 z/ C m2 2
0 Œ p
Z Z 1 Z
dz .dp/
8 m2 .dp/ 2 C Q2 z.1 z/ C m2 2
8 2 C m2
: (A-3.13)
0 Œ p Œ p
Integrating by parts one more time over z, using the integral (II.8), and writing
m2 D .m2 C Q2 z.1 z/ Q2 z.1 z//, we obtain
Z
7 .8p 2 16m2 /
I.Q2 / D i 2 Q2 C .dp/
3 Œ p 2 C m2 2
Z Z 1
.dp/ .3z2 2z3 /Q2 .1 2z/
C 4 Q2 C 4 i 2 Q2 dz : (A-3.14)
Œ p C m2 2
2
0 Q2 z.1 z/ C m2
Here we recognize the Q2 -independent part as I.0/ corresponding to the third term
in (A-3.6). The last integral in (A-3.14) may be integrated by parts over z to finally
obtain from (A-3.6), (A-3.7), (A-3.8),
Z Z 1 !
e2 1 .dp/ 1 h Q2 i
2
˘.Q / D C 6 dz z.1 z/ ln 1 C 2 z.1 z/ ;
12 2 i 2 Œ p 2 C m2 2 2 0 m
(A-3.15)
R
involving the logarithmically divergent integral .dp/=Œ p2 C m2 2 =.i 2 /. We will
have ample opportunities later on to interpret this constant (see Sect. 5.10.2). The
additive factor 1=2 to it depends on the rooting of the momentum Q in the
definition of the integral in (A-3.6).
Appendix B: Infinitesimal Variation of the Exponential

of a Matrix
For a matrix M, • eM ¤ • M eM , in general, but Tr Œ • eM D Tr Œ • M eM . To

see how the situation changes from one case to the other, and in passing derive an
expression for • eM , integrate the following, involving two matrices N, M, from
D 0 to D 1:
d N M
e e D e N .N M/ e M ; (B-3.1)
d
Problems 129
to obtain
Z 1
eN eM D 1 C d e N .N M/ e M ; (B-3.2)
0
or
Z 1
e e
N M
D d e N .N M/ e M eM : (B-3.3)
0
Let N D M C •M, then the latter equation reads

Z 1
e MC•M
e M
D d e .MC•M/ .• M/ eM eM : (B-3.4)
0
For an infinitesimal variation, we have the useful expression

Z 1
• eM D d e M .• M/ eM eM : (B-3.5)
0
If we perform a trace of the above, we simply have

Z 1
Tr Œ • eM D Tr Œ d • M eM ;
0
D Tr Œ •M eM : (B-3.6)
Problems
3.1 Show that this remarkable positive function:

8
ˆ
ˆ 0; t < a;
ˆ
< 1 exp
2 a 1;
ˆ
a t < 0;
f .t/ D 2 1
tCa a
ˆ
ˆ1 2 exp 2 tCa 1 ; 0 t < a;
ˆ
:̂
1; a t;
with a > 0, is not only continuous at t D a, t D 0, t D a, it vanishes at

t D a, and is equal to 1 at t D a, but has a continuous derivative as well. This
is a particular example of a function vanishing at a point and being also continuous.
The keen reader will realize that this provides a continuous function representation
for the (discontinuous) step function ™.t/ D 1 for t > 0 and ™.t/ D 0 for t < 0.
3.2 Verify that the propagator SC .x 0 x/ also satisfies (3.1.13).

3.3 Verify (3.2.15)/(3.2.16) for the vacuum expectation value of the Dirac field
.x/, in the presence of a weak external electromagnetic potential A , for x0
corresponding to a time much later after the source has been switched off.
3.4 Derive the expression for h 0C j .x/ j 0 i=h 0C j 0 i in (3.3.16), where x0
correspond to a point before the source .x 0 / is switched on.
3.5 Work out the details leading to the expression in (3.3.14)/(3.3.15) corre-
sponding to the scattering of a positron eC . p/ ! eC . p 0 / off a weak external
electromagnetic potential.
3.6 Show that the probability that a particle of spin 1/2 emitted by a source has
a momentum p within a range and spin is given by
Z Z X
1 m d3 p m d3 p
j u.p; /. p/ j2 ; where N D j u.p; /. p/ j2 :
N p
p 0 .2/3
p 0 .2/3
3.7 By averaging over the initial spins and summing over the final ones,
derive (3.4.6) for p 0 D p 0 0 .
3.8 Derive the anti-commutation relation in (3.5.12)/(3.5.13) for the Dirac quantum
field valid for all times. Hint: Note that the Dirac field at any time may be expressed
in terms of the time equal to zero, via the equation
ZX Z h
m d3 p 0
.x/ D d3 x 0 ei.xx / p u.p; / u .p; /

p 0 .2/3
0/ p
i
C ei.xx v.p; / v .p; / .x 0 /;
x 0 D .0; x 0 /. On the Rother hand, the orthogonal projections in (I.21), (I.22) allow
one to write .x/ D d3 y K.x; y/ .y/; y D .0; y/,
Z
m d3 p h i.xy/p 0 i.xy/p 0
i
K.x; y/ D e P C . p/ e P . p/ ;
p0 .2/3
where PC . p/ D . p C m/=.2m/, P . p/ D . p C m/=.2m/ as given,

respectively, in (I.21), (I.22).
3.9 Show that at equal times x 0 D x 0 0 , the anti-commutation relation obtained in
the previous problem reduces to the one in (3.5.11).
References 131
R R xC.
=2/
3.10 Consider the integral I D Œ.dQ/=.2/4 A .Q/ x.
=2/ d eiQ . Setting
D x C .
=2/, with as an integration variable, express the integral as
a power series in
. This is a typical integral which occurs in Schwinger’s point
splitting method.22

:B i E
3.11 Show that 4 Œ ; F D
i
, where F D @ A @ A .
i E B
3.12 Derive the expression in (3.8.8) for the Schwinger effect by working in the
complex s-plane, and elaborate on the importance of the term .sjeEj/2 =3 in (3.8.5).
3.13 Prove the identity in (3.9.25) with the propagator S. p/ D 1=. p C m/.
3.14 Derive the transformation rule for G in (3.10.27) for non-abelian gauge
theories.
3.15 Show that

1
exp Œ iF.z/ SC .z x/ C ı .4/ .z x/ .@=@x /F.x/
1
D SC .z x/ exp Œ iF.x/ ;
for a numerical function F.z/.23
References
1. Adler, S. L. (1969). Axial-vector vertex in Spinor electrodynamics. Physical Review, 177, 2426–
2438.
2. Adler, S. L., & Bardeen, W. (1969). Absence of higher-order corrections in the anomalous axial-
vector divergence equation. Nuovo Cimento A, 182, 1517–1536.
3. Bell, J. S., & Jackiw, R. (1969). A PCAC Puzzle: 0 ! in the -Model. Nuovo Cimento
A, 60, 47–61.
5. Jackiw, R., & Johnson, K. (1969). Anomalies of the axial-vector current. Physical Review, 182,
1459–1469.
7. Schwinger, J. (1951). On Gauge invariance and vacuum polarization. Physical Review, 82, 664–
679.
8. Schwinger, J. (1959). Field theory commutators. Physical Review Letters, 3, 296–297.
9. Schwinger, J. (1962). Exterior algebra and the action principle I. Proceedings of the National
Academy of Sciences of the United States of America, 48, 603–611.
22
See Appendix IV at the end of the book.
23
This result will be important in the analysis of Problem 5.20 in Chap. 5.
Recommended Reading

2. Schwinger, J. (1951). On gauge invariance and vacuum polarization. Physical Review, 82, 664–
679.
3. Schwinger, J. (1970). Particles, Sources, and Fields (Vol. I). Reading: Addison-Wesley.
Chapter 4
Fundamental Aspects of Quantum Field Theory
At sufficiently high energies, relativity becomes the indispensable language to do

physics and explain the exchange that takes place between energy and matter
allowing the creation of an unlimited number of particles such that the number
of particles need not be conserved. An appropriate description of such physical
processes for which a variable number of particles may be created or destroyed
is provided by the concept of a quantum field which we denote, generally, by .x/
as a function of a spacetime variable x D .x 0 ; x/, for which the Dirac quantum field
theory, as described in Chap. 3, is one example.
As the disturbance created by a measurement process cannot propagate faster
than the speed of light, different regions of space, at any given fixed time, are
dynamically independent in the sense that a measurement made in any region of
space is incompatible with any other in a different region. Space, at any given
time, defines a special case of a space-like (hyper-) surface with the latter defined
in such a manner that any two points x1 , x 2 lying on it are space-like separated
.x1 x2 /2 > 0, and hence cannot be connected by any signal. The quantum fields as
dynamical variables, providing degrees of freedom of a system, are thus labeled, by
infinitesimal three dimensional regions of space, with the latter being independent
in the sense just described, or in a local description they are labeled by the space
coordinate points x. At any given time x 0 D t, the space variable x together with
any indices a that independent fields may carry, such as spinor and/or vector indices,
specify the infinite (uncountable) degrees of freedom as x is made to vary in space
as well as the indices a take on their specific values. Upon writing a .x/ D x a .t/,
we note that .xa/ replaces the index i in the dynamical variables qi .t/, in well known
situations, dealing with finite number of degrees of freedom.
Independent dynamical variables a .x/ thus satisfy locality conditions, and for
the elementary fields, for examples, consisting of a (non-Hermitian) scalar '.x/,
and the Dirac field .x/, are specified by a commutation relation Œ '.x/; ' .x 0 / D
0 for the scalar field (Sects. 4.5 and 4.7), and by anti-commutation relations

DOI 10.1007/978-3-319-30939-2_4
134 4 Fundamental Aspects of Quantum Field Theory

f a .x/; b .x 0 /g D 0 for the Dirac field components (Sect. 3.5) for any two space-like
separated points x, x 0 .
The significant physical role that a quantum field .x/ acquires and the particle
content of a theory emerges from the examination of the commutation relations
of the field with the energy-momentum operator components P (Sect. 4.1) corre-
sponding to the underlying physical system.
The field concept, its significance and the meaning of wavefunction renormaliza-
tions are introduced in Sect. 4.1. The underlying algebra satisfied by the generators
of the inhomogeneous Lorentz transformations (Poincaré algebgra) is worked out
in Sect. 4.2. The action principle of quantum fields, together with its related
consequences, is the subject matter of Sect. 4.3. The principle of stationary action
encapsulates the dynamics of a system as the system develops between two space-
like surfaces. The principle of stationary action, together with generators emerging
from variations of these space-like boundary surfaces, summarize basic properties of
an underlying theory. The mere fact that generators are to be introduced to generate
field variations lead automatically to field equations – the so-called Euler-Lagrange
equations. These field equations emerging from the theory, involve dynamical
variables, as functions of time (and space), and describe the time evolution of theory.
Section 4.4 deals with Lorentz invariance and the energy-momentum tensor, as well
as the angular momentum density, providing, in particular, of a local description
of the energy-momentum distribution of matter in space. The celebrated Spin
and Statistics connection is studied in Sect. 4.5. The quantum dynamical principle
(QDP) of quantum field theory is treated in Sect. 4.6 from which physical processes
may be investigated.
A functional Fourier transform of expressions directly obtained from the QDP
leads directly to the equivalent path integral formalism which is applied to gauge
theories in subsequent chapters. The basic fields one encounters in field theory
are covered in Sect. 4.7, with further applications of the QDP, involving some
of these fields, in Sect. 4.8. Some intricacies with Lagrangians are spelled out in
Sect. 4.9. Section 4.10 deals with the celebrated CPT Theorem, where, in particular,
we will see that particle/antiparticle detectors and antiparticle/particle emitters are
interchanged in a CPT “transformed” world.
One of the major difficulties in quantum field theory, is that not all components
of the underlying theories are necessarily dynamically independent. This requires
special attention to be given to them in carrying out a dynamical description of
the theories. All of our present fundamental field theories are gauge theories, and
involve such fields.
In remaining chapters we establish the connection between the fundamental the-
ory developed here and the various interactions observed in nature. As emphasized
in the introductory chapter to the book, a major theme of modern quantum field
theory is to provide a unified description of nature. This calls for a look at the
various approaches and contributions to the subject that appeared in the literature
4.1 The Field Concept, Particle Aspect and Wavefunction Renormalization 135
since the pioneering work of Dirac in the mid twenties, and several of them are
treated in coming chapters and the accompanying book by the author.1
4.1 The Field Concept, Particle Aspect and Wavefunction

Renormalization
The significant role that a quantum field .x/ plays emerges from the examination
of its commutation with the energy-momentum operator P , corresponding to
the underlying physical system, and by considering, in particular, the vacuum
state j vaci assumed to be an eigenstate of P of zero energy and momentum
P jvaci D 0. As will be seen in the next section, the different components P
commute. The latter components consist of generators of spacetime translations,
and they imply that

.x/ D ei x P .0/ ei x P : (4.1.1)
The latter leads to the equation
i @ .x/ D Œ .x/; P : (4.1.2)
Upon introducing the Fourier transform

Z
.dp/ i xp
.x/ D e . p/; (4.1.3)
.2/4
Eq. (4.1.2) gives
p . p/ D . p/ P P . p/: (4.1.4)
Let j p 0 ; i denote an eigenstate of the components P . That is, P j p 0 ; i D p 0 j

p ; i, where stands for other labels needed to specify a state j p 0 ; i. By working
0
conveniently in terms of the bra hp 0 ; j, and by rewriting (4.1.4) as . p/ P D

P . p/ C p . p/, we obtain
h p 0 ; j . p/ P D . p 0 C p / h p 0 ; j . p/: (4.1.5)
That is, for p0 ? 0, . p/ injects/absorps a quantity of energy jp0 j. The idea

that one learns in elementary treatments of (free) field theory that a field may be
expanded in terms of creation and annihilation operators comes from this relation.
The situation we are now in is far more complex, in general, and the physical
1
Quantum Field Theory II: Introductions to Quantum Gravity, Supersymmetry, and String Theory,
(2016), Springer.
interpretation of .x/ in terms of creation and annihilation of just single isolated

free particles (anti-particle) becomes incomplete.
In general, the field .x/ satisfies coupled non-linear equations which in turn may
lead to non-trivial reactions on the field leading to the creation of multi-particles of
unlimited number altering the initial expected excitation that the field may provide
by a mere examination of the free field theory counterpart. One may even encounter
an extreme case that the particle expected to be created by the field is altogether
absent. That is, the creation of the particle occurs with zero probability in the
standard quantum mechanical language. All of this is best described by examining
the effect of a field .x/ on h vac j in the light of the spectrum of the energy-
momentum operator P .
The spectral representation of the energy-momentum operator P , expressed
through a unitary operator, is given by
Z Z
ei xP D .dp/ ei xp . p/; 1D .dp/ . p/; (4.1.6)
where . p/ is a projection operator, which together with (4.1.1) and the condition
hvacj P D 0, allow one to write
Z
hvacj .x/ D .dp/ ei xp hvacj .0/. p/: (4.1.7)
We may decompose the spectral representation in terms of a single-particle plus

multi-particle contributions to (4.1.7) and rewrite the latter as
ZX Z
d3 p
hvac j .x/ D ei xp hvac j .0/ j p; ihp; j C .dp/ei xp hvac j .0/0 . p/;

2p 0 .2/3
(4.1.8)
q
where p0 D p2 C m2 , m is the mass of the particle associated with the field ,
hp; j 0 . p 0 / D 0; hp; j p0 ; 0 i D .2/3 2p 0 • 0 •3 .p p0 /: (4.1.9)
Upon setting
p
hvacj .0/ j p; i D Z U.p; /; (4.1.10)
p
up to a phase factor multiplying the real coefficient Z, and where U.p; / is the
wavefunction of the particle in question, in particular, in the momentum description,
we obtain
Z X Z
d3 p p
hvac j .x/ D ei xp Z U.p; /hp; j C .dp/ ei xp hvac j .0/0 . p/;

2p 0 .2/3
(4.1.11)
4.1 The Field Concept, Particle Aspect and Wavefunction Renormalization 137
p
hvacj .x/ j p; i D Z U.p; / ei xp : (4.1.12)
and where U.p; / eipx denotes the wavefunction of the particle in question, and
0 . p/ is a projection operator on the remaining spectrum of the energy-momentum
of the system under study.
In the presence of multi-particle states, described by the second p integral
in (4.1.11), quantum mechanics implies the presence of a coefficient Z in the
expansion (4.1.11), to describe a particle just described, with Z as the probability of
the creation of the particle in question by the field .x/ out of the vacuum.
Situations may arise that a single isolated particle cannot be described, or that
the underlying theory predicts only composite particles, in which cases the first
term in (4.1.11) will be altogether absent. On the other hand, when dealing with a
scalar particle .x/ for which no symmetry arguments may be applied to dismiss
its vacuum expectation value hvac j .0/ j vaci, then an additional term of the
form hvac j .0/ j vacihvac j will appear on the right-hand of (4.1.11). Such a non-
vanishing vacuum expectation value of a scalar field is what happens in describing
the Higgs mechanism.
In studying a physical process and confronting theory with experiments, one
may ask the question: given that a certain particle was observed, that is, this event
happened with probability one, then what is the probability that it has emerged,
say, with a specific momentum and a given spin projection?. In the language of
probability theory, the underlying probability in such a question is referred to as a
conditional probability:
given that something happened, i.e., with probability
one, then what is the probability that the system
under consideration has some given characteristics?
p
In technical terms, this amounts to divide the field .x/ by Z, leading to the
definition of the field
p
ren .x/ D .x/= Z; (4.1.13)
thus isolating the particle in question with probability one, i.e., with certainty,
bringing the observed particle in evidence.
The process just described is called a wavefunction renormalization, which is
done independently of any perturbation theory, and Z is referred to as a wavefunction
renormalization constant. The scaled field ren .x/ is the so-called renormalized field.
Let us go one step further to see how a wavefunction renormalization is carried
out when one actually computes transition amplitudes. We will even generalize
the above further, by bringing the treatment closer to an experimental set-up,
where one would ask the question as to what is the amplitude that the particle is
actually observed, i.e., detected?. The analysis, in turn, shows the role played by a
field in computing transition amplitudes. As we are already familiar with the spin
1/2 propagator derived in Sect. 3.1, we consider for definiteness the field .x/ to
correspond to a Dirac field .x/ in a general interacting field theory.
To the above end, let O.x1 ; : : : ; xn / j vaci denote the product of n arbitrary
operators of spacetime arguments x1 ; : : : ; xn acting on the vacuum state j vaci, and
consider the vacuum-expectation of the time-ordered product:

hvacj .x/O.x1 ; : : : ; xn / C
j vaci; (4.1.14)
in which the operators are arranged from right to left in order of increasing time
argument with a minus sign inserted for each permutation of Fermi-Dirac fields
(Sect. 3.5). Here .x/ is the quantum field in question under investigation. We
emphasize that this matrix element is computed in a theory involving no external
sources coupled to the fields such as ; given in Chap. 3.
Now we multiply the above vacuum expectation value by a Fermi source .x/
(Sect. 3.1) from the left and integrate over x to obtain
Z

.dx/ .x/hvacj .x/O.x1 ; : : : ; xn / C
j vaci: (4.1.15)
By doing this, we will concentrate on the particle in question associated with the
field .x/. A similar process may be carried out for the particles associated with the
other fields in the time-ordered product by multiplying (4.1.14) by corresponding
source functions. We recall that the field .x/, in particular, is expected to satisfy a
non-linear equation. Let SC .x x 0 / denote the free spin 1/2 propagator derived in
Sect. 3.1. We use the property
Z
.dx 0 / SC .x x 0 / SC
1
.x 0 x 00 / D ı .4/ .x x 00 /; (4.1.16)
1
where SC .x 0 x 00 / is the inverse of the propagator. Consider the source .x/ to
be switched on, i.e., operate, at large positive x 0 values, thus one may “follow” the
particle for a sufficient long time to be able to measure its energy, momentum and
other attributes. For all x 0 > x 0 0 , at which times the source .x/ operates, it is easily
derived that
Z Z X 3
d p m i x 0p
.dx/ .x/SC .x x 0 / D e Œ i . p/u.p; / u.p; /;

.2/3 p 0
(4.1.17)
by using, in the process, the relations in (3.1.10) and (I.21). From (4.1.16), (4.1.17),
and the identity
@ 0
1
SC .x 0 x 00 / D C m ı .4/ .x 0 x 00 /; (4.1.18)
i
4.2 PoincarKe Algebra and Particle States 139
the following expression emerges for (4.1.15):

Z X 3 p
dp m
3 0
Œ i . p/ u.p; / Z

.2/ p
hZ 0
@ 0 i
.dx 0 / ei x p u.p; / C m hvacj ren .x
0
/ O.x1 ; : : : ; xn / C
j vaci :
i
(4.1.19)
One may readily continue this process as applied to the other fields in the time
ordered product in (4.1.14) by multiplying the latter by corresponding sources and
noting that for sources switched on and operating at large positive times, one is
dealing with particles outgoing from a process, while sources operating only at large
negative times, one is dealing with ingoing particles in a process. The keen reader
will realize that we have essentially obtained a formalism for computing transition
amplitudes in field theory. This will be taken up later, dealing with the vacuum-to-
vacuum transition amplitudes for various interactions, and specific applications will
be carried out.
The expression in (4.1.19) brings us into contact with a formalism, referred to as
the LSZ formalism,2 except that we have now an additional piece of physics dealing
with the detection (emission) of particles expressed by the first line in (4.1.19)
depending on the source function . p/. From (3.2.8), we recognize the term
Œi . p/ u.p; / as an amplitude for detection, i.e., of the actual observation of the
particle in question. Having scaled the field in (4.1.13) to define the renormalized
field
p ren .x/, we now have to define the renormalized source function ren .x/ D
Z .x/ as well for the physical interpretation of the latter corresponding to an
amplitude of particle detection. When one adds source terms to a Lagrangian
density such as .x/ .x/, corresponding, in particular, to the field .x/, one has
the invariance property .x/ .x/ D ren .x/ ren .x/.
Introduction of sources coupled to the fields facilitates tremendously doing field
theory. In practice, the transition amplitudes for particles (anti-particles) ingoing
and outgoing in a process is obtained from the coefficient of the source terms in
expressions like in (4.1.19) (see, e.g., Sects. 5.8 and 5.9).
4.2 PoincarKe Algebra and Particle States
PoincarKe (inhomogeneous Lorentz) transformations in Minkowski spacetime were

given in (2.2.7), Sect. 2.2. Such a transformation being connected to the identity
one, induces, according to Wigner’s Theorem (Sect. 2.1), a unitary operator to
act on particle states and quantum fields. To find the algebra satisfied by the
2
See Lehmann et al. [2].
generators of such a transformation, we consider successive infinitesimal PoincarKe

transformations forming a closed path as follows
.2 ; b2 /1 .1 ; b1 /1 .2 ; b2 /.1 ; b1 / D .; b/; (4.2.1)
represented pictorially by
2 2
emphasizing the reversal of the transformations in the third and the fourth segments
of the path. The matrix elements of are explicitly given in (2.2.13), (2.2.14),
(2.2.15), (2.2.16), (2.2.11), and for infinitesimal ones in (2.2.19), (2.2.20), (2.2.21)
and (2.2.22).
The infinitesimal coordinate changes •x associated with such transformations
may be spelled out as
•x D x x 0 D •b •! x ; •! D •! ; (4.2.2)
where D C •! .
From the group properties below (2.2.9) in Sect. 2.2, (4.2.1) readily leads to
•! D •! 2 •!1 •! 2 •! 1 ; (4.2.3)
•b D •! 2 •b 1 •! 1 •b 2 : (4.2.4)
The unitary operator corresponding, for example, to .; b/, for infinitesimal
transformations, has the structure
1
U D 1 C i G; G D •b P C •! J ; (4.2.5)
2
with P , J denoting the energy-momentum and angular momentum operators,
respectively, generating spacetime translations and homogeneous Lorentz transfor-
mations. The latter, corresponding to J , consist of Lorentz boosts and spacial
rotations.
Referring to the closed path in (4.2.1), one has
U21 U11 U2 U1 D U; (4.2.6)
which leads to
1
G D Œ G1 ; G2 ; (4.2.7)
i
with the parameters of the generator G given by .•! , •b / spelled out

in (4.2.3), (4.2.4), and those of G1 , G2 are denoted by .•!1 ; •b1 /, .•!2 ; •b2 /.
Upon comparison of the coefficients of identical products of the above parame-
ters on both sides of (4.2.7) gives the PoincarKe algebra
Œ P ; P D 0 (4.2.8)

Œ P ; J D i P P ; (4.2.9)

Œ J ; J D i J J C J J ; (4.2.10)
satisfied by the generators P , J .

The following operators P P and W W , where W is the (pseudo-) vector
defined by
1
W D
P J ; (4.2.11)
2
referred to as the Pauli-Lubanski (pseudo-) vector and whose physical significance
will become clear shortly, are invariants of the corresponding group. That is, they
commute with the PoincarKe generators P , J . Hence particle states may be
labeled by their eigenvalues. Here
is totally anti-symmetric specified by
0123 D C 1. Note that in (4.2.11): ¤ , ¤ , hence the operators P

and J in W commute (see (4.2.9)).
W satisfies the following commutation relations
Œ W ; P D 0 (4.2.12)

ŒW ;W D i
P W ; (4.2.13)
Œ W ; J D i . W W /; (4.2.14)
and the conditions
W P D 0; W W 0: (4.2.15)
In detail, the components of W are given by
W 0 D P J; W D P 0 J P N; Nk D J0 k : (4.2.16)
The representations of physical interest are those for which P
P P P 2 D m2 0; (4.2.17)
with sgn P 0 > 0 corresponding to positive energy, for massive (m ¤ 0), i.e.,
P time-like, and massless (m D 0), i.e., P light-like, particles. P D 0
corresponds to the vacuum state. Consider first massive particles.
I m ¤ 0:
Since W W is an invariant, it may be computed in any Lorentz frame. For a
massive particle, we may consider its rest frame for which P D .m; 0; 0; 0; 0/.
From (4.2.16), then
W W D m2 J2 : (4.2.18)
In the rest frame of such a particle, (4.2.10) leads to the usual commutation
relations of angular momentum components
Œ J i ; J j D i
ij k J k ; (4.2.19)
and
W W D m2 j. j C 1/; (4.2.20)
since W i becomes simply m J i and W 0 reduces to zero.

Consider the eigenvalue problem
P h p j D p h p j; (4.2.21)
where particle labels denoting other degrees of freedom are initially suppressed. In
particular, in the rest frame of the particle, we have from (4.2.9)
ŒP 0 ; J i h p j D 0: (4.2.22)
Accordingly, from (4.2.18), (4.2.19), (4.2.20), (4.2.21) and (4.2.22), we may

label a massive particle state, in the rest frame of the particle, by its mass m, and
the spin j, involving .2j C 1/ components. We will simply use the notation h 0 j
for these states, with 0 corresponding to the momentum of the particle at rest, and
corresponding to the values j. All other labels necessary to specify such a state
will be suppressed.
To find particle states for arbitrary momenta, we proceed as follows. Consider
the state
h 0 j exp Œ i ˛ J 03 exp Œ i n J ; (4.2.23)
n D .sin ; cos ; 0/; sinh ˛ D jp j=m: (4.2.24)

We will apply the operators P and W 0 to the state in (4.2.23). To this end, note
that the vector operator
V./ D exp Œ i n J P exp Œ i n J ; (4.2.25)
satisfies the equation
dV./
D n V./; V.0/ D P: (4.2.26)
d
Hence from (2.2.10), (2.2.11) and (2.2.12),
V i D R ij P j ; (4.2.27)
with [R ij ] denoting the rotation matrix spelled out in (2.2.11), and the unit vector
n is defined in (4.2.24).
On the other hand, defining
Y D exp Œ i ˛ J03 P exp Œ i ˛ J03 ; (4.2.28)
we obtain
dY d2 Y
D 3 Y 0 0 Y 3 ; D 3 Y 3 0 Y 0 : (4.2.29)
d˛ d˛ 2
The solution of (4.2.29) is elementary and is given by
Y 3 D P 3 cosh ˛ C P 0 sinh ˛; Y 0 D P 0 cosh ˛ C P 3 sinh ˛; (4.2.30)
and
Y a D P a; for a D 1; 2: (4.2.31)
From (4.2.23), (4.2.24), (4.2.25), (4.2.26), (4.2.27), (4.2.28), (4.2.29), (4.2.30)

and (4.2.31), the facts that h0 j P D 0, and h0 j P 0 D mh0 j, we may infer
that3 (4.2.23) defines a state with momentum-energy p ,
h 0 j exp Œ i ˛ J03 exp Œ i nJ P D p h 0 j exp Œ i ˛ J03 exp Œ i nJ ;

(4.2.32)
with
p
p0 D C p 2 C m2 ; p D jp j.sin cos ; sin sin ; cos /: (4.2.33)
3
See Problem 4.1.
p
n = (sin φ , − cos φ , 0)
θ p = |p|(sin θ cos φ , sin θ sin φ , cos θ )
φ 2
φ
n
Fig. 4.1 Rotation of a vector initially along the 3-axis c.w. by an angle about the unit vector n
lying in the 1–2 plane
Here we note that a rotation of vector p, initially along the 3-axis, by an angle ,
about the unit vector n in (4.2.24), gives the vector in (4.2.33) (Fig. 4.1).
A similar analysis4 shows that W 0 applied to the state in (4.2.23) gives
h 0 j exp Œ i ˛ J 03 exp Œ i n J W 0
D m sinh ˛ h 0 j exp Œ i ˛ J 03 exp Œ i n J : (4.2.34)
Accordingly, up to a normalization factor, we denote the states in (4.2.23) by

h p j, providing an orthogonal basis of vectors of single-particle states. The
orthogonality property follows from the facts that
0 D h p0 0 j P P j p i D .p 0 p/h p 0 0 j p i;
0 D h p 0 0 j W 0 W 0 j p i D m . 0 sinh ˛ 0 sinh ˛/h p 0 0 j p i:
(4.2.35)
From (4.2.24), sinh ˛ D jpj=m, hence the first equality in (4.2.35) shows that
sinh ˛ D sinh ˛0 . The two equalities in (4.2.35) together then imply that hp0 0 j p i
is proportional to ı 3 .p p 0 / ı 0 . Apart from the labels m and j, i.e., of , due
to the invariance of P P and W W , and other labels, we will not dwell upon
such additional labels here.
Using the Lorentz invariance of the measure5 d3 p=.2/3 2p0 , we may set
hp 0 0 j p i D .2/3 2p0 •3 .p p0 / ı 0 : (4.2.36)
I m D 0:
Consider an eigenvalue problem as in (4.2.21). In the massless case, we may
initially consider the three momentum of the particle to be along the 3-axis. In such
a frame, we may write the energy-momentum of the particle as k D .!; 0; 0; !/.
4
See Problem 4.2.
5
See (3.1.18).
As a consequence of (4.2.9), (4.2.16), (4.2.12) the following relations are realized
ŒP ; J 12 h k j D 0; W 0 h k j D ! J 3h k j; W 3 h k j D ! J 3 h k j; (4.2.37)
W 1 h k j D ! T 1 h k j; W 2h k j D ! T 2h k j (4.2.38)
where .J 12 D J 3 /
T 1 D J1 C N 2; T 2 D J 2 N1; N k D J 0 k; (4.2.39)
Œ T 1 ; T 2 D 0; Œ T 1; J 3 D i T 2; Œ T 2; J 3 D i T1: (4.2.40)
In the above, we have suppressed other labels in the states h k j.

Upon comparison of the set of commutation relations in (4.2.40) with (4.2.8),
(4.2.9), we learn that they correspond to translations in the 1–2 plane, via generators
T 1 ; T 2 , and rotations about the 3 - axis, via the generator J 3 . T 1 and T 2 have
continuous unbounded spectra. Incidentally, .T 1 /2 C .T 2 /2 D .W W /=! 2 , with
! > 0, is an invariant of the underlying group, with seemingly a continuous non-
negative spectrum. Unlike the energy and momentum of a particle, which possess
continuous spectra, there are no physical states with such spectra of ‘continuous’
spins. This leaves room only for the cases that T ih k j D 0 for i D 1; 2,
corresponding to W W D 0.
The interesting thing to note is that according to (4.2.15), we now have two
orthogonal light-like vectors:
W P D 0; W W D 0; P P D 0; p 0 > 0; (4.2.41)
which means that W , P are proportional to each other,6 i.e.,
W D P : (4.2.42)
In particular, this equation applied to W 0 in (4.2.16) gives P 0 D P J.

That is, , referred to as the helicity, is determined from the component of spin
along the momentum. Since W is a (pseudo-) vector and P is a vector, is a
(pseudo-) scalar. On account of (4.2.12), (4.2.14) and (4.2.42), commutes with the
generators P , J of the PoincarKe group, hence it is an invariant and the physical
states may be labeled by .
The invariance property of the helicity means that Nature would exhibit only
a unique value for the helicity of a given massless particle. On the other hand, if
parity is conserved for the interaction of a system under consideration, then this
pseudo-scalar can also take the value . Such a property is shared by photons
6
See Problem 4.3.
( D ˙ 1) for the electromagnetic interaction, and by the graviton ( D ˙ 2) for

the gravitational interaction.
We may label the initial states considered above as h k j . To find the states for
arbitrary momenta is now straightforward since is an invariant.
With ! D jpj, it is evident that, up to a normalization factor,
h p j D h k j exp Œ in J ; (4.2.43)
where n is defined in (4.2.24), and we verify, by using (4.2.25), (4.2.26) and

(4.2.27), that
h p j P D p h p j; (4.2.44)
with p 0 D jp j, p given in (4.2.33). The orthogonality of the states h p j follows

as before,
h p 0 0 j p i D .2/3 2p0 ı 3 .p p 0 / ı0 ; (4.2.45)
written in an invariant manner.

Before closing this section, we note that due to the rotational invariance of the
scalar product P J, i.e., Œ P J; J i ; D 0, a rotation induced on a state in (4.2.43)
about the momentum by an angle #, leads to
h p j exp Œ i# P J=jp j D exp Œ i# h p j: (4.2.46)
A rotation by an angle # D 2 gives C 1 or 1 for single or double valued

representations corresponding to integer and half-odd integer values for .
4.3 Principle of Stationary Action of Quantum Field Theory:

The Rationale Of
This section is involved in considering general variations, to be specified below, of

transformation functions h 2 j 1 i, associated with dynamical systems, between
two space-like surfaces 1 ; 2 . This brings us into the realm of a fundamental
theory referred to as the “Principle of Stationary Action”.
The Principle of Stationary Action provides the basic tool for deriving field
equations. The latter field equations emerge from the variation of the action under
the variations of the dynamical variables within specified boundaries. Variations
of dynamical variables call for generators to implement the corresponding trans-
formations. By introducing, in turn, such generators, the stationary aspect of the
action under dynamical field variations directly follows giving automatically the
field equations – the so-called Euler-Lagrange equations. Understanding this step
should be comforting to the student who would, otherwise, blindly write down
4.3 Principle of Stationary Action of Quantum Field Theory: The Rationale Of 147
the Euler-Lagrange equations. Equally important, is that the concept of the action
and of a Lagrangian density arise as a consequence of studying variations of
transformation functions describing the evolution of a dynamical system in time,
and hence are of physical interest. The principle also provides the way of expressing
various generators in terms of the variables underlying the Lagrangian density and
deals with conservation laws as well. Dynamics, beyond the field equations, such
as providing a method for obtaining expressions for transition amplitudes, is the
subject of Sect. 4.6 under the heading of the Quantum Dynamical Principle. The
celebrated Spin and Statistics Connection is treated in Sect. 4.5.
A state defined on a space-like surface will be denoted by h j, and labels
needed to specify such a state will, in general, be suppressed. Since there is always
a coordinate system for which such a space-like surface is the space itself, at a given
time, i.e., represented by a surface x 0 D t fixed, we may equally consider a state
denoted by h t j . We are interested in changes that may occur in transformation
functions h t2 jt1 i corresponding, in general, to different values of quantum numbers
associated with an evolving system at the different times t1 and t2 , under some
basic transformations.
Under an infinitesimal transformation via a unitary operator U D 1Ci G, where
G defines the generator for such a transformation, an operator O changes to O 0 D
U O U and with the change defined by
•O D O O 0 ; (4.3.1)
we have
1
•O D Œ O; G : (4.3.2)
i
Hence for a wavefunction , equivalence of the “primed” and “unprimed”
depictions, implies that •h jO j i D h jO j i h 0 jO 0 j 0 i D 0, and
with j 0 i D U j i gives j i j 0 i D •j i D i G j i. The invariance
0 0
P h j
of the scalar product iPD h j UU i D h j i, and the resolution of
the identity 1 D j aiha jD U j aiha j U , for bases vectors j ai, imply that

•ha jD ha j ha j U is given by
•h a j D i h a jG: (4.3.3)
We consider two types of changes of the transformation functions h t2 j t1 i:

(i) One type is in response to, a priori, imposed variations of the underlying
dynamical variables of the theory at all times t1 t t2 , by keeping t1
and t2 fixed. In such cases, we will use the notation • .x/, as applied to the
fields, for the variation that one may, a priori, impose on a field with a bar
under •. Here the variations • .x/ are chosen at will and may be chosen to
be numericals for unambiguous analyses, up to the statistics obeyed by the
fields .x/ and otherwise arbitrarily. Non-numerical changes of the fields will
be discussed later.
(ii) A second type corresponds to variations of just the boundaries, i.e., of the
surfaces x 0 D t1 and x 0 D t2 in response to a given transformation, taking
into account of the induced variations that arise on the underlying dynamical
variables themselves from such a transformation on the boundaries. For such
induced variations on the fields, we will simply use the notation •.x/ to
distinguish them from ones one may, a priori impose on them.
The first type of variations and its consequence are considered in Sects. 4.3.1,
4.3.2, and 4.3.3, while the second type is considered in Sect. 4.3.4.
4.3.1 A Priori Imposed Variations of Dynamical Variables and

Generators of Field Variations: Field Equations
We here consider variations of the fields a .x/, at any given time t in t1 t t2 ,

with the space-like surfaces x 0 D t1 and x 0 D t2 , kept fixed. Let x a .t/ a .x/
denote the dynamical variables. This is an extension of a theory involving dynamical
variables of a countable degrees of freedom qi .t/, to a theory involving an infinite
uncountable ones, with .x a/ now replacing the index i, as discussed earlier in the
introduction to this chapter.
An infinitesimal variation of a dynamical variable calls for a generator to
carry out such a variation. We recall that in quantum mechanics, an infinitesimal
coordinate translation •x i is carried out via a generator written as P i •x i , where
•x i are numerical constants that commute with the operator P i , and the latter
is then identified with the momentum operator. By extending,P in the process, the
generator for the countable number of degrees of freedom case i pi .t/•qi .t/, to an
infinite uncountable ones, the generator for infinitesimal variations, at a given time
t, may be expressed as
Z
G.t/ D d3 x a .x/ • a .x/; x 0 D t: (4.3.4)
In reference to the first type of variation with time fixed, we denote the changes
of the dynamical variables by • a .x/. The operators a .x/ are identified as
the canonical conjugate momenta of the fields a .x/ and are to be determined.
Following Schwinger, the variations • a .x/ are taken as numericals. Non-numerical
changes will be discussed later. The adjoints of the fields, as obtained from the fields
themselves, will be suppressed for simplicity.
For the time derivative of (4.3.4), evaluated at any given time t: t1 t t2 , we
have
Z h i
d
G.t/ D d3 x P a .x/ • a .x/ C a .x/ • P a .x/ : (4.3.5)
dt
The integrand in (4.3.5) may be expressed in terms of the total infinitesimal

variation •H of the Hamiltonian, at any given time, as follows. The latter, as a
functional of a .x/ and a .x/, is given by
Z 1
1
•H D d3 x • a .x/ Œ a .x/; H Œ a .x/; H • a .x/
i i
Z
D d3 x • a .x/ P a .x/ P a .x/ • a .x/ : (4.3.6)
R
Accordingly, upon solving for d3 x P a .x/• a .x/ from the above equation, we may
rewrite dG.t/=dt as
Z
d
G.t/ D d3 x • a .x/P a .x/ H .x/ ; (4.3.7)
dt
where H .x/ is the Hamiltonian density

Z
H.t/ D d3 x H .x/; x 0 D t: (4.3.8)
At this stage the concept of a Lagrangian density arises. Upon identifying the
Lagrangian density L .x/ as follows
L .x/ D a .x/P a .x/ H .x/; (4.3.9)
we obtain
Z
d
G.t/ D d3 x • L .x/: (4.3.10)
dt
Thus from (4.3.3), with t1 , t2 fixed, we have

Z t2
d
•h t2 j t1 i D i h t2 jŒ G.t2 / G.t1 / j t1 i D i h t2 j dt G.t/j t1 i; (4.3.11)
t1 dt
and from (4.3.10), (4.3.11), we obtain in detail

Z x 0 D t2
˝ ˇ ˇ ˛
•h t2 j t1 i D i h t2 jŒ G.t2 / G.t1 / j t1 i D i t2 ˇ .dx/ • L .x/ ˇ t1 ;
x 0 D t1
(4.3.12)
.dx/ D dx 0 dx1 dx 2 dx 3 , where we have used the notation • to remind us that we are
considering variations of the dynamical variables only for fixed times t1 , t2 . • L .x/
in (4.3.12) means variations at any given t in t1 t t2 , that is including on the
space-like surfaces x 0 D t1 , x 0 D t2 .
From the last equality in (4.3.12), we may infer that:

Only the variations of the fields on the boundaries x 0 D t1 ; x 0 D t2 in
the Lagrangian density may contribute to • h t2 j t1 i, as spelled out by the two
surface terms in ŒG1 .t2 / G1 .t1 /, and variations of the fields in • L within
the boundaries do not contribute. The latter variations lead to the Euler-Lagrange
equations.
How do we isolate these two different parts in • L ? To do this, suppose that the
Lagrangian density L .x/ depends on a .x/ and, on the first derivative @ a .x/.
Accordingly, quite generally, we may write
• L .x/ D F a .x/ • a .x/ C F a .x/ • .@ a .x//

D ŒF a .x/ @ F a .x/ • a .x/ C @ F a .x/ • a .x/ ; (4.3.13)
thus separating the two contributions mentioned above, and the variations within the
boundaries x 0 D t1 x 0 D t2 , do not contribute, i.e., for arbitrary • a .x/
Œ F a .x/ @ F a .x/ D 0; t1 < x 0 < t2 : (4.3.14)
These are the celebrated Euler-Lagrange Equations. On the other hand (4.3.13) also
gives from (4.3.12)
Z x0 D t2
Œ G.t2 / G.t1 / D .dx/ @ F a .x/ • a .x/ : (4.3.15)
x0 D t1
Upon introducing the hyper-surface elements
d D .dx1 dx 2 dx 3 ; 0/; (4.3.16)
we may apply Gauss’ Theorem to rewrite (4.3.15) as

Z h iˇ Z h iˇ
ˇ ˇ
Œ G.t2 / G.t1 / D d3 x F a 0 .x/ • a .x/ ˇ d3 x F a 0 .x/ • a .x/ ˇ :
x 0 D t2 x 0 D t1
(4.3.17)
Hence upon comparison with (4.3.4), we obtain
a .x/ D F a 0 .x/: (4.3.18)
The above analysis, in turn, introduces the action integral

Z x 0 D t2
AŒ a D .dx/ L .x/; (4.3.19)
x 0 D t1
a functional of the fields a .x/, as an integral over spacetime bounded by space-

like surfaces x 0 D t1 , x 0 D t2 . Here the Lagrangian density L .x/ is assumed to
depend on a .x/ and on the derivatives @ a .x/. It may be rewritten as
L .x/ D L Œ a .x/; @ a .x/: (4.3.20)
The indices a may stand for different types of fields and for any vector and/or spinor
indices they may carry. The Lagrangian density is assumed to be written as the sum
of products of the fields and their first derivatives.
To determine F a .x/ and F a .x/, we proceed as follows. The variation of the
action •AŒa , with fixed space-like surfaces x 0 D t1 , x 0 D t2 , in response to the
variation of the Lagrangian density •L .x/ is given by by
Z x 0 D t2
•AŒ a D .dx/ •L .x/; (4.3.21)
x 0 D t1
where

• L .x/ D L Œ a .x/; @ a .x/ L Œ a .x/ • a .x/; @ a .x/ • a .x/ ;
(4.3.22)
arising solely from arbitrary variations • a .x/ that one may a priori impose on the
fields a .x/, where recall that the notation • a .x/, with a bar under •, is used
to distinguish these variations from ones that are induced on the fields by other
given transformations such as by the rotation of a coordinate system or by a Lorentz
transformation.
According to the Spin & Statistics Connection, investigated in Sect. 4.5, once
the nature of a field has been established as being a Bose-Einstein or a Fermi-
Dirac one, we consider infinitesimal numerical, now, c-number or Grassmann,
variations • a .x/, respectively, as the case may be, such that a variation commutes
with all the fields and their derivatives, if the corresponding field a is a Bose-
Einstein one, or anti-commutes with all the Fermi-Dirac fields and their derivatives,
and also commutes with all the Bose-Einstein fields and their derivatives, if the
corresponding field a is a Fermi-Dirac one.
Field Equations
For infinitesimal • b .x/, typical terms in (4.3.22) may have the structures

a1 .x/ @ a2 .x/ : : : • a .x/ : : : D . "/ n a1 .x/ @ a2 .x/ : : : • a .x/;
(4.3.23)

@ b1 .x/ b2 .x/ : : : @ • b .x/ : : : D . "/ m @ b1 .x/ b2 .x/ : : : @ • b .x/;
(4.3.24)
up to product of numerical matrices. Here " D 1 if the corresponding field

a .x/, to • a .x/ is a Bose-Einstein field, and " D C 1 if it is a Fermi-Dirac one.
n , m denote the number of Fermi-Dirac fields to the right of • a .x/, @ • b .x/,

respectively. Each term contains an even number of fermion fields to satisfy the
Bose character of a Lagrangian density. Therefore, if the number of fermion fields
to the right of a fermion field is odd, then the number of fermion fields to its left
must be even and so on.
It becomes clear that we obtain the same expression for •L .x/ in (4.3.13),
(4.3.22), as the one that emerges from applying (4.3.23), (4.3.24) to all the terms
in •L .x/, if we use instead the following alternative formal procedure.
We introduce partial derivatives @=@ a .x/, @=@.@ a .x// to operate on the
fields and their first derivatives, respectively, applied both from the right of the
Lagrangian density in conformity with the identities in (4.3.23), (4.3.24), and both
obeying the same commutativity or anti-commutativity with the fields as • a .x/.
A straightforward application of the chain rule gives

@L .x/ @L .x/
• L .x/ D .• a .x// C .@ • a .x//: (4.3.25)
@ a .x/ r @.@ a .x// r
Here r stands for “from the right”. The partial derivatives in this equation are defined
by @b .x/=@a .x/ D •ba , @.@ b .x//=@.@ a .x// D • •ba , and so on, as opposed
to functional derivatives which also involve Dirac delta functions (see below).
One may rewrite • L .x/ in (4.3.13)/(4.3.22) as

@L .x/ @L .x/ @L .x/
• L .x/ D @ • a .x/ C @ • a .x/ :
@ a .x/ @.@ a .x// r @.@ a .x// r
(4.3.26)
Hence from (4.3.13), we have

@L .x/ a @L .x/
F .x/ D
a
; F .x/ D ; (4.3.27)
@ a .x/ r @.@ a .x// r
with all partial derivatives taken from the right.

Upon setting,

@L .x/
a .x/; (4.3.28)
@.@ a .x// r
the variation of the action AŒ a in (4.3.19) in response to the variation in (4.3.22)

is then from (4.3.14), (4.3.15), (4.3.16), (4.3.17) and (4.3.18)
Z t2

•AŒ a D .dx/ @ a .x/ • a .x/ ; (4.3.29)
t1

@L .x/ @L .x/
@ D 0: (4.3.30)
@ a .x/ @.@ a .x// r
Using the Bose character of a Lagrangian density, we may replace the partial
differentiations in (4.3.30) from the right by ones from the left since this means
simply to multiply the whole equation, in the just mentioned equation, by 1 for
a Fermi field. Accordingly the field equations (Euler-Lagrange Equations) may be
rewritten as
@L .x/ @L .x/
@ D 0; (4.3.31)
@ a .x/ @.@ a .x//
with the partial derivatives taken from the left, which is the conventional definition.
The Gauss’s Theorem, of course, gives from (4.3.29)
Z h ˇ
ˇ
•AŒ a D d3 x 0 a .x/• a .x/ˇ .t2 ! t1 / D Œ G.t2 / G.t1 / : (4.3.32)
x 0 D t2
A far simpler way of obtaining (4.3.31) is to take the functional derivative of the
action AŒ b , with respect to a .x/ , with AŒ a as a functional of b , in the
following manner.
To the above end, the functional derivative of b .x 0 / with respect to a .x/ is
defined by
• b .x 0 /
D •b a •.4/ .x 0 x/: (4.3.33)
• a .x/
For the functional derivative of a field @ 0 b .x 0 / with respect to a .x/, we note

that
Z
• @ 0 b .x 0 / • b .x 00 / 00 .4/ 00
D .dx 00 / @ • .x x 0 /
• a .x/ • a .x/
Z
D .dx 00 / •b a •.4/ .x 00 x/ @ 00 •.4/ .x 00 x 0 / D •b a @ 0 •.4/ .x 0 x/;
(4.3.34)
The product rule then follows to be
•
b .x 0 / @ 0 c .x 0 / D •b a .@ 0 c .x 0 // " b .x 0 / •c a @ 0 •.4/ .x 0 x/;
• a .x/
(4.3.35)
where " D 1, if at least one the fields a , b is a Bose-Einstein one, and
" D C 1, if they are both Fermi-Dirac ones.
The Euler-Lagrange equations in (4.3.31) may be then simply replaced by the
functional differentiation of the action AŒ b with respect to a .x/ expressed as
follows
Z
• •
AŒ b D .dx 0 / L Œ b .x 0 /; @ 0 b .x 0 / D 0; (4.3.36)
• a .x/ • a .x/
using the simple rules in (4.3.33), (4.3.34) and (4.3.35) and the usual rule of
commutativity or anti-commutativity with the fields as the case may be of the nature
of the fields a .x/ in question applied. The integration in (4.3.36) is taken over all
of spacetime in order not to introduce a surface integral since the latter is already
accounted for by the integral in (4.3.29). Equation (4.3.36) treats the fields b .x 0 /
and their derivatives @ 0 b .x 0 / on an equal footing.
Box 4.1: Key points in the stationary action principle for fixed boundaries
•h t2 j t1 i D h t2 j Œ G.t2 / G.t1 / j t1 i; with t1 ; t2 fixed; where

Z
G.t/ D d3 x a .x/• a .x/; x 0 D t; for arbitrary infinitesimal
variations of the fields a .x/; denoted by • a .x/; for any given

t in t1 t t2 :

@L .x/
a .x/ D a 0 .x/; a .x/ D ; where the a .x/
@.@ a .x// r
are the canonical conjugate momenta of the fields a :

Z
AŒ b D .dx/L Œ b .x/; @ b .x/; denotes the action integral
extended over all of spacetime, and

Z
• •
AŒ b D .dx 0 / L Œ b .x 0 /; @ 0 b .x 0 / D 0
•a .x/ • a .x/
gives rise to the Euler - Lagrange equations upon functional
differentiations, using the simple rules in (4.3.33), (4.3.34) and (4.3.35):
In conclusion, the following should be also noted:

1. Not all the field equations following from (4.3.31) or (4.3.36) are necessarily
equations of motion as such they may not involve the time derivative of a field in
question and may lead to constraints for the latter field. For examples, working in
the Coulomb gauge in QED and QCD, the field equations obtained by varying the
zeroth components A0 , Aa 0 of the vector potentials, respectively, do not involve
their time derivatives and lead to constraints for them, and not all the remaining
components are independent. The canonical conjugate momentum density of a
dependent field, expressed in terms of independent fields, is zero, and its variation
is induced by the generators of the independent fields just mentioned.
2. Later we will see how field equations are derived from Lagrangian densities for
higher spin fields involving arbitrary orders in the space derivative @, due to
constraints, but still of first order in the time derivative @0 , of the fields.
3. It is important to realize that all the results derived were obtained by a priori
numerical variations of the fields obeying their respective statistics as established
by the Spin & Statistics Connection in Sect. 4.5. Quantum variations applied
a priori to a Lagrangian are certainly generalizations and as such they often
involve additional assumptions and consequences not always of practical value,
and further generalizations lead, for example, to generalized statistics such as
para-statistics.
4.3.2 Commutation/Anti-commutation Relations
The generator for inducing infinitesimal changes • a , considering independent

quantum fields a .x/, is given in (4.3.4), and from (4.3.2), this change is given by
i • a .x/ D Œ a .x/; G.x 0 / ; (4.3.37)
where [ , ] is the commutator. In detail,

Z
i • a .x/ D d3 x 0 Œ a .x/; b .x 0 /• b .x 0 / x 0 0 D x 0: (4.3.38)
We use the fact that
• b .x 0 / a .x/ D " a .x/ • b .x 0 /; (4.3.39)
with " D 1 if at least one of the fields a .x/, b .x 0 / is a Bose-Einstein one, and
" D C 1 if both are Fermi-Dirac ones. We may thus rewrite (4.3.38) as
Z
i • a .x/ D d3 x0 a .x/ b .x 0 / C " b .x 0 / a .x/ • b .x 0 /; x0 0 D x 0; (4.3.40)
This equation implies the following equal-time anti-commutation/commutation

rules:
f a .x/; b .x 0 /g D i •a b •3 .x x 0 /; (4.3.41)
Œ a .x/; b .x 0 / D i •a b •3 .x x 0 /; (4.3.42)
for Fermi-Dirac and Bose-Einstein fields, respectively.

4.3.3 Generators for Quantum Responses to Field Variations:

Internal Symmetry Groups
Consider the following commutation relation for x 0 D x 0 0 ,
Œ c .x/; a .x 0 / b .x 0 / D c .x/ a .x 0 / b .x 0 / a .x 0 / b .x 0 / c .x/

D c .x/ a .x 0 / C " a .x 0 / c .x/ b .x 0 /
D i •c a •3 .x x 0 / b .x 0 /; (4.3.43)
where we have used (4.3.40), and, as before, " D

1 for Bose-Einstein/Fermi-
Dirac fields, respectively. Let Ma b .x/, a .x/ denote arbitrary numericals but
infinitesimal depending, in general, on the spacetime coordinate. For x 0 0 D x 0 ,
we introduce the operator
Z
0 1 3 0
G.x / D d x Ma b .x 0 / a .x 0 / b .x 0 / " b .x 0 / a .x 0 /
2

C .1 "/ a .x 0 / a .x 0 / ; (4.3.44)
where note the minus sign multiplying ", in this expression. Using the Fermi or
Bose character of the fields, we may use (4.3.41) or (4.3.42) to infer from (4.3.44)
that
1
Œ c .x/; G.x 0 / D i Mc b .x/ b .x/ C .1 "/c .x/ ; (4.3.45)
2
and hence the relation • c .x/ D .1=i/Œ c .x/; G.x 0 / , induces a quantum variation
to the field c .x/ given by
1
• c .x/ D Mc b .x/ b .x/ C .1 "/c .x/; (4.3.46)
2
as a linear combination of quantum field components b .x/ with, in general, space-

time dependent coefficients plus, in general, of a spacetime dependent numerical
shift for Bose-Einstein fields (" D 1).
In particular, the infinitesimal gauge transformation of the non-abelian vector
gauge field in (3.10.10) with
• A .x/ D i g •#a .x/ Œ ta ; A .x/ @ #a .x/ ta ; (4.3.47)
may be expressed as
• Ac .x/ D Ac .x/ A 0c .x/ D g •#a .x/ fabc Ab .x/ @ #c .x/; (4.3.48)
where A .x/ D tc Ac .x/, which is precisely of the form in (4.3.46), involving now
a Lorentz index .
4.3.4 Variations of Boundary Surfaces
In reference to the second type of variations corresponding only to change of the

boundary surfaces, in response to a transformation t ! t •t, the corresponding
generator GB .t/ is simply H.t/ •t, where H.t/ is the Hamiltonian. In terms of
the Hamiltonian density H .x/, in (4.3.8), this generator is explicitly given by
Z
GB .x 0 / D d3 x H .x/ •x 0 ; (4.3.49)
and
•h t j D ih t jGB .t/; (4.3.50)
at a given time t, and for the transformation function ht2 j t1 i,
•h t2 j t1 i D i h t2 jŒ GB .t2 / GB .t1 / j t1 i: (4.3.51)
The latter may be equivalently rewritten as

Z t2
d
•h t2 j t1 i D i h t2 j dt GB .t/ j t1 i; (4.3.52)
t1 dt
or as
Z t2
d
•B h t2 j t1 i D i h t2 j dt H.t/ •t j t1 i; (4.3.53)
t1 dt
where the subscript B is to remind us that only the boundary surfaces are being
varied. In terms of the Hamiltonian density,
Z x 0 D t2
•B h t2 j t1 i D i h t2 j .dx/ @ 0 •x 0 H .x/ j t1 i: (4.3.54)
x0 D t1
We may re-express the above in terms of the Lagrangian density

Z x 0 D t2
•B h t2 j t1 i D i h t2 j .dx/ @0 •x 0 a .x/@ 0 a .x/ •x 0 L .x/ j t1 i:
x 0 D t1
(4.3.55)
Under an infinitesimal time translation x 0 ! x 0 •x 0 with x 0 D .x 0 •x 0 ; x/

and hence 0a .x/ D a .x/ C • x 0 @ 0 a .x/, gives
• a .x/ D a .x/ 0a .x/ D •x 0 @ 0 a .x/: (4.3.56)
We also recall that a canonical conjugate momentum a .x/ is the zeroth component
of the field a .x/ as we have seen in (4.3.18), (4.3.27), (4.3.28). Accord-
ingly, (4.3.55) may be rewritten as
Z x 0 D t2
•B h t2 j t1 i D i h t2 j .dx/ @0 •x 0 L .x/ C a 0 .x/ •a .x/ j t1 i: (4.3.57)
x 0 D t1
We may use Gauss’ Theorem, to rewrite the above expression as

Z
ˇˇ
•B h t2 j t1 i D i h t2 j d3 x •x 0 L .x/ C a 0 .x/ • a .x/ ˇ .t2 ! t1 / j t1 i:
x 0 D t2
(4.3.58)
Accordingly, for a constant of motion given by

Z

QD d3 x •x 0 L .x/ C a 0 .x/ • a .x/ ; (4.3.59)
i.e., for Q, independent of time, •B h t2 j t1 i D 0.

With the Lagrangian density L .x/ as a Lorentz scalar, we may use the
invariance of a scalar product in spacetime, to readily generalize this result further
for states associated with space-like surfaces 1 , 2 corresponding to the change of
the boundaries under a Lorentz transformation x ! x • x to
Z 2

•B h 2 j 1 i D i h 2 j .dx/ @ •x L .x/ C a • a .x/ j 1 i: (4.3.60)
1
A local conservation law, corresponding to the conservation of the operator Q

in (4.3.59) may be also given, from the above integral, by considering an infinitesi-
mal deviation of the hypersurface 2 from 1 , about a point x, which leads to
@ j .x/ D 0; j .x/ D •x L .x/ C a • a .x/: (4.3.61)
This is the content of Noether’s Theorem . Of course the volume integral of j 0 .x/
simply gives the expression for Q in (4.3.59).
Finally, we note from (4.3.15), (4.3.28), (4.3.29), where • a .x/ corresponds to, a
priori, imposed arbitrary infinitesimal transformation, together with the expression
for the generator in (4.3.60) that the generator which induces transformation of the
fields for x ! x •x as well as other transformations, such as involved with
4.4 Inhomogeneous Lorentz Transformations and Energy-Momentum Tensor 159
internal symmetries, is given by the expression:

Z

G./ D .dx/ @ •x L .x/ C a • a .x/ ; (4.3.62)

where now • a .x/ is the infinitesimal response to the general transformations

mentioned above.
4.4 Inhomogeneous Lorentz Transformations

and Energy-Momentum Tensor
Under an infinitesimal inhomogeneous Lorentz transformation x ! x •x D x 0 ,

as discussed in Sect. 2.2, where
•x D •b •! x ; •! D •! ; (4.4.1)
a field a .x/ transforms as

i
0a .x 0 / D •a b C •! S a b b .x/; (4.4.2)
2
since the deviation from a .x/ must be proportional to •! . The matrix ŒS a b
defines the spin of the field. This gives
i
• a .x/ D a .x/ a 0 .x/ D • x @ a .x/ C •! S a b b .x/ : (4.4.3)
2
Accordingly, from (4.3.60) we may write
Z 2

•h 2 j 1 i D h 2 j .dx/@ •x L .x/ a •x @ a .x/

1
i
•! a S .x/ a j1 i: (4.4.4)
2
The antisymmetry property of •! allows us to write

•! a S .x/ a D •! a S .x/ a C a S .x/ a C a S .x/ a ;
(4.4.5)
which as a consequence of the symmetry in the last two terms in .; / is like
adding zero to the left-hand side. This suggests to set
i a
˝ .x/ D .S .x//a C a .S .x//a C a .S .x//a ; (4.4.6)
2
which is antisymmetric in the indices .; / since S D S . On the other hand,
one may write
1
•! D Œ @ •x @ •x ; (4.4.7)
2
and hence
1
@ •! ˝ D @ Œ @ •x @ •x ˝ : (4.4.8)
2
Now

@ Œ@ •x ˝ D @ ! ˝ D @ ! ˝ : (4.4.9)
On the other hand

@ Œ@ •x ˝ D @ @ Œ•x ˝ •x @ ˝

D @ @ •x ˝ C @ •x @ ˝ : (4.4.10)
The first term on the right-hand of the last equality is zero since ˝ is
antisymmetric in .; /, while @ @ is symmetric. Hence from (4.4.9), (4.4.10),
we may rewrite (4.4.8) as
•! •b •!

@ •! ˝ D @ ˝ C @ ˝ x @ ˝ :
2 2 2
(4.4.11)
The first term on the right-hand side is 1=2 of the left-hand side, and hence using,
in the process, the anti-symmetry property of •! , we get
•!
@ •! ˝ D @ •b @ ˝ C Œ x @ ˝ x @ ˝ :
2
(4.4.12)
Finally note that

•x L .x/ a •x @ a .x/ D •x L .x/ a @ a .x/ : (4.4.13)
Hence from (4.4.5)/(4.4.6), (4.4.12), (4.4.13) and (4.4.1), we may rewrite (4.4.4)
in the form
Z 2
•h 2 j 1 i D h 2 j .dx/ @ j .x/ j 1 i; (4.4.14)
1
4.4 Inhomogeneous Lorentz Transformations and Energy-Momentum Tensor 161
•!

j .x/ D •b T .x/ C x T .x/ x T .x/ ; (4.4.15)
2
where

T .x/ D L .x/ a @ a .x/ @ ˝ .x/: (4.4.16)
The first term on the right-hand side between the round brackets is referred to as
the canonical energy-momentum tensor.
We may set
M .x/ D x T .x/ x T .x/; (4.4.17)
to re-express j .x/ as
•!
j .x/ D •b T .x/ C M .x/: (4.4.18)
2
Invoking invariance under inhomogeneous Lorentz transformations, and consid-
ering the hypersurface 2 deviating infinitesimally from the hypersurface 1 , about
a point x, gives, in a limiting sense, the local conservation law
@ j .x/ D 0: (4.4.19)
For arbitrary independent infinitesimal variations •b , •! , we may infer

from (4.4.18), the conservations laws
@ T .x/ D 0; (4.4.20)
@ M .x/ D 0: (4.4.21)
On the other hand, applying @ directly to (4.4.17) and using (4.4.20) give
0 D @ M .x/ D T .x/ T .x/; (4.4.22)
establishing the symmetry of T . The latter is referred to as the symmetric

energy-momentum tensor. This is unlike the canonical one which is not necessarily
symmetric. This symmetry property is important as it is necessary for the conserva-
tion of the angular momentum density M . Also general relativity requires that
the energy-momentum tensor be symmetric.
The momentum and the angular momentum may be now written as
Z

P D d T .x/; (4.4.23)
Z
J D d M .x/: (4.4.24)
They may be also rewritten as integrals over space

Z

P D d3 x T 0 .x/; (4.4.25)
Z
J D d3 x M 0 .x/: (4.4.26)
The generator for Lorentz transformations , may be also spelled out as
•!
G D •b P C J ; (4.4.27)
2
expressed in terms of the variables of the Lagrangian density with infinitesimal
variations •b , •! .
For applications to the Maxwell field and the Dirac field, see Problems 4.6, 4.7
and 4.8. See also Problem 4.9.
4.5 Spin and Statistics Connection
The unique and important role that the Spin and Statistics Connection plays not
only in physics but in all of the sciences, in general, and in the stability of matter
and hence in our own existence, cannot be overemphasized. The Periodic Table of
Elements in chemistry is based on the Pauli Exclusion Principle as a special case.
Without it matter would be simply unstable and collapse.7
The Spin & Statistics Connection arises naturally and self consistently by the
analysis of the action integral.
Suppose we are given a Lagrangian density having the structure
1

L .x/ D a .x/ Qa b @ b .x/ @ a .x/ Qa b b .x/ C F Œ .x/ ; (4.5.1)
4
where the field .x/ D . a .x// is real, L is of first order in the derivative of the
field and of first degree, i.e., linear in @ as shown, Q is a numerical matrix,
and F Œ .x/ depends on .x/. Hermiticity of L , as the latter is related to the
Hamiltonian, implies that
.Q / D Q : (4.5.2)
7
See also the introductory chapter to the book.
4.5 Spin and Statistics Connection 163
A complex field may be re-expressed as a real field with double the number
of components consisting of its real and imaginary parts. For example, a four-
component Dirac field may be re-expressed as a real one having eight components
as we have seen in (3.5.3), Sect. 3.5. By appropriately choosing the fields in the
theory to have even more components, such a first order Lagrangian density may, in
general, be constructed. For the Dirac field, the structure given in (4.5.1) is obvious.
The situation involved with other fields will be treated later. Further generalizations
of (4.5.1) will be also discussed at the end of this section dealing with constraints.
Any square matrix C may be written as the sum of a symmetric and an anti-
symmetric matrices S and A, i.e., C D S C A, with
Sa b D .Ca b C Cb a /=2; A a b D .Ca b Cb a /=2; (4.5.3)
and S> D S, A> D A. For a matrix C such that C D C, we also have
S D S, A D A.
Accordingly, we consider two systems for which
.Q /> D Q or .Q /> D Q ; (4.5.4)
together with the condition in (4.5.2).

As before we define the action integral, a functional of the field ,
Z x0 D t2
AŒ D .dx/ L Œ .x/; @ .x/ ; (4.5.5)
x0 D t1
as an integral over spacetime bounded by space-like surfaces x 0 D t1 , x 0 D t2 .

We investigate the nature of the variation of the action in response to the variation
of the field defined by .x/ ! .x/ • .x/. Following Schwinger, we consider
variations of AŒ arising from c-number variations • .x/. We will then encounter
that such variations we were set to achieve with • .x/ commuting with the field
itself is possible only for systems satisfying the second condition in (4.5.4), while
• .x/ as a c-number Grassmann variable anti-commuting with the field is possible
only for systems satisfying the first one.
We define the variation
• L Œ .x/ D L Œ .x/ L Œ .x/ • .x/; @ .x/ @ • .x/ ; (4.5.6)
paying special attention to the order in which a variation • a .x/ appears relative to
the fields b .x/. As in (4.3.26), one has
Z 1
1 t2

•AŒ D .dx/ @ Œ a .x/ Qa b • b .x/ • a .x/ Qa b b .x/ C
2 t1 2
Z t2 1 Z t2

.dx/ Œ • a .x/ Qa b @ b .x/ @ a .x/ Qa b • b .x/ C .dx/ •F Œ.x/ ;
t1 2 t1
(4.5.7)
with the order in which the variations • .x/ appear kept intact. Note that the
integrand of the first term is the integral of a total differential.
We set

Qa b D " Q b a ; (4.5.8)
with " D ˙1 corresponding to the two cases in (4.5.4). By using a property of the
Dirac delta function, the second integral on the right-hand side of (4.5.7) may, up to
the addition of a term proportional to the first integral, be expressed as
Z Z h i
1 t2 t2

.dx/ .dx0 / Qa b • a .x/ b .x0 /" b .x0 / • a .x/ @ 0 •.4/ .x0 x/: (4.5.9)
2 t1 t1
From this we learn that a c-number variation • a .x/ commuting with the fields
b .x/ is possible only for " D 1, i.e., for .Q /> D Q , while a variation
• a .x/ as a c-number Grassmann variable anti-commuting with the fields b .x/ is
possible only for " D C1, i.e., for .Q /> D Q . Otherwise the expression within
the square brackets in the integrand in (4.5.9) would be zero and no equations of
motion for the fields involving their derivatives will result.
Therefore we may write,
Œ • a .x/; b .x 0 / D 0 for .Q /> D Q ;

D 1; (4.5.10)
f• a .x/; b .x 0 /g D 0 for .Q /> D C Q ;
D C1: (4.5.11)
That is, • a .x/ b .x 0 / D " b .x 0 / • a .x/. In both cases, we have

Z h i
• b .x/
t2
•AŒ D .dx/ @ a .x/ Qa b
t1 2
Z t2

C .dx/ • a .x/ Qa b @ b .x/ C • F Œ .x/ ; (4.5.12)
t1
and • F Œ .x/ D F Œ .x/ F Œ .x/ • .x/ , by definition.

A typical term in • F Œ .x/ may be of the form
a1 : : : an • a : : : D ."/n • a a1 : : : an : : : ; (4.5.13)
up to numerical matrices, and where on the left-hand side of this equation we have
the product of n field components appearing to the left • a . Accordingly, we will
get the same expression for • F Œ .x/ as prescribed above, if we set up the rule
that
@
• F Œ .x/ D • a .x/ F Œ .x/ ; (4.5.14)
@ a .x/
with the partial derivative operating from the left and anti-commuting with the field
components for the case .Q /> D Q , and commuting for .Q /> D Q . The
partial derivative is simply defined by @ b .x/=@ a .x/ D •b a as opposed to a
functional derivative which also involves a Dirac delta function.
The principle of stationary action for arbitrary variations • a .x/ then gives the
field equations
@
Qa b @ b .x/ C F Œ .x/ D 0: (4.5.15)
@a .x/
An application of Gauss’ Theorem to the first integral on the right-hand side

of (4.5.12) gives
Z t2
dx 0 @0 G.x 0 / D G.t2 / G.t1 /; (4.5.16)
t1
Z
• b .x 0 ; x/
G.x 0 / D d3 x a .x 0 ; x/ Q 0a b ; (4.5.17)
2
and G.t1 /, G.t2 / correspond to the variations • c .x/ of the field components on the
space-like boundary surfaces x 0 D t1 , x 0 D t2 , respectively, i.e., for fixed times.
As in Sect. 4.3.1, and as we will see at this section, G.x 0 / defines the generator for
such a transformation, at a given time x 0 , and accounts now for a change • c .x/=2
via a unitary operator U D 1 C i G, and U c .x/U D c .x/ • c .x/=2 leads to
• c .x/ 1
D Œ c .x/; G : (4.5.18)
2 i
From (4.5.10), (4.5.11) and (4.5.17), the above gives, at any given times x 0 D
00
x ,
Z • .x 0 /
• c .x/ 1
D Q 0a b d3 x 0 c .x/ a .x 0 / C " a .x 0 /c .x/
b
: (4.5.19)
2 i 2
and the following commutator or anti-commutator, in turn, emerge
Œc .x/ ; a .x 0 / Q0a b D i •cb •3 .x x 0 /; .Q /> D Q ; (4.5.20)
f c .x/ ; a .x 0 /g Q 0a b D i •cb •3 .x x 0 /; .Q /> D C Q : (4.5.21)
One may, for example, consider the case where F Œ .x/ in (4.5.1) is of the form
1
F Œ .x/ D a Mab b ; (4.5.22)
2
where M is a Hermitian numerical matrix.
According to (4.5.10), (4.5.11), we must have
.Q /> D Q ; M> D M; (4.5.23)

.Q /> D Q ; M > D M: (4.5.24)
4.5.1 Summary
Given a Lagrangian density as in (4.5.1), the statistics obeyed by the field in question
depends on the anti-hermitian matrix Q satisfying one of the conditions spelled out
in (4.5.4). In particular, we learn that a variation • a .x/ commuting with the fields
b .x/ is possible only for " D 1, i.e., for .Q /> D Q , while a variation
• a .x/ as a Grassmann variable anti-commuting with the fields b .x/ is possible
only for " D C1, i.e., for .Q /> D Q . In the first case the theory is quantized
via commutators derived in (4.5.20), while for the second case it is quantized via
anti-commutators derived in (4.5.21). The field equations are given in (4.5.15), and
the properties displayed in (4.5.23), (4.5.24) should be noted.
The method of investigation of the Spin and Statistics connection developed in
this section is referred to as Schwinger’s Constructive Approach.8
The above results have already been applied to the Dirac quantum field in
Sect. 3.5, and special cases are discussed below. Applications to other fields are
given in the next section. We first consider the Hamiltonian
4.5.2 The Hamiltonian
The Hamiltonian density H is directly obtained from (4.5.1) by subtracting off, in

the process, the net term in it involving the time derivative. This gives
1
H .x/ D @ k a .x/ Q ka b b .x/ a .x/ Q ka b @ k b .x/ F Œ .x/ ; (4.5.25)
4
and for the Hamiltonian
Z
H D d3 x H .x/: (4.5.26)
Under an infinitesimal variation of the fields • a .x/, as before,

Z
@
•H D d3 x • a .x/ Q kab @ k b .x/ C F Œ .x/ ; (4.5.27)
@b .x/
8
Schwinger [11–15], See also Manoukian [6], pp. 953–964.
and from the field equations (4.5.15), this simply becomes,

Z
•H D d3 x • a .x/ Q0a b @0 b .x/: (4.5.28)
The expression for the variation induced on the fields • a .x/, under a time
displacement x 0 ! x 0 •x 0 D x 0 0 , as opposed to the ones • a .x/ that one may,
a priori, impose on them as done above, follows from the consideration
0 .x 0 / D .x/; x 0 D .x 0 0 ; x/; x D .x 0 ; x/; (4.5.29)

0 .x 0 ; x/ D .x 0 C •x 0 ; x/ D .x/ C •x 0 @0 .x/; (4.5.30)
•.x/ D .x/ 0 .x/ D •x 0 @0 .x/: (4.5.31)
With H as the generator for time displacement via the unitary operator:
.1 i •x 0 H/, one has the induced variation • a .x/ D i Œ H; a .x/ •x 0 , or
i @ 0 a .x/ D Œ a .x/; H : (4.5.32)
Upon multiplying the latter by Q0a b • b .x/=2, remembering the properties

in (4.5.10), (4.5.11), using the Bose-Einstein character of the Hamiltonian, and
finally integrating over x gives from (4.5.28)
1
i • H D ŒH; G ; (4.5.33)
2
where G is given in (4.5.17). Since the integrand of • H in (4.5.28) is linear in

• a .x/, we see that this generator G accounts for half • a .x/=2 for the variation
of the fields • a .x/.
4.5.3 Constraints
Finally we note that for higher spins because of constraints, such as ones arising
from gauge constraints, Lagrangian densities of the form in (4.5.1) may arise, where
the matrix Q may become dependent on the space derivative @ of arbitrary orders
and so does F , which would involve arbitrary orders of the space derivatives of the
fields. That is, they become replaced by Q .@/, F Œ.x/; @.x/ and are even in @:
Q .@/ D Q .@/ and the same for F . In particular, this means that @ operates
equally to the left or right. In such cases the relations in (4.5.20), (4.5.21), with
x0 D x 0 0 , become simply replaced by
Œ c .x/ ; Q0a b .@ 0 / a .x 0 / D i •c b •3 .x x 0 /; .Q /> .@/ D Q .@/;

(4.5.34)
fc .x/ ; Q0a b .@ 0 / a .x 0 /g D i •c b •3 .x x 0 /; .Q /> .@/ D C Q .@/: (4.5.35)
and the field equations are readily handled as will be seen later.
4.6 Quantum Dynamical Principle (QDP) of Field Theory
The Quantum Dynamical Principle (QDP) gives rise to a powerful and elegant
formalism for studying dynamics in field theory and for carrying out all sorts of
computations. It is particularly easy to apply and provides, in a compact way,
a method for considering variations of so-called transformation functions, such
as transition amplitudes, with respect to parameters that may appear in a theory
as coupling parameters, as well as of variations with respect to external sources
coupled to the underlying fields.
In the above sense, out of the QDP a functional differential formalism emerges
for quantum field theory, where the dynamics is investigated and physical aspects
of the theory are extracted simply by functional differentiations of some given
generating functionals. In turn, it provides tremendous physical insight and an easy
way for studying basic processes in a theory as particles/anti-particles move from
emission to detection regions in a way similar to experimental situations.
By a functional Fourier transform (§ 2.6), the QDP also gives rise to its path
integral counterpart, appropriately referred to as the path integral formalism,
where continual integrations are to be carried out instead of differentiations of the
differential QDP formalism, as will be seen in Chaps. 5 and 6.
The purpose of this section is to derive the QDP from first principles. Applica-
tions of this will be given in various stages, and some applications were already
carried out in Sect. 3.3 and in Sect. 3.6. The reader will conveniently find a direct
summary suitable for practical applications of the QDP at the end of this section.
Suppose one has a Hamiltonian H.t; / depending on some parameters, denoted
collectively by , such as coupling parameters and external sources that we wish
to vary. The Hamiltonian may have an explicit time dependence assumed to come
from such parameters. That is, the explicit time dependence is of numerical nature
unaffected by transformations applied to operators in the underlying vector space in
which H.t; / operates. We may, in turn, introduce the Hamiltonian H.t; 0/, and
corresponding time development unitary operators U .t; /, U .t; 0/, associated
with the two Hamiltonians, satisfying
d d
i U .t; / D U .t; / H.t; /; i U .t; 0/ D U .t; 0/ H.t; 0/: (4.6.1)
dt dt
The Hamiltonian H.t; 0/ may also depend on some other parameters that we do not
wish to vary. H.t; 0/, in general denotes the renormalized free Hamiltonian, that is
the free Hamiltonian expressed in terms of physically measurable quantities.
4.6 Quantum Dynamical Principle (QDP) of Field Theory 169
Given a state h a; D 0 I t j h a; D 0 j U .t; 0/, corresponding to a system

governed by the Hamiltonian H.t; 0/, i.e., for D 0, we define a state
h a ; j D h a; D 0 j U .; 0/ U.; /; (4.6.2)
developing in time via the unitary operator U .t; / i.e.,
h a ; I t j D h a ; j U .t; /: (4.6.3)
Such a state, satisfies the Scrödinger equation
d
i h a ; I t j D h a ; I t j H.t; /; (4.6.4)
dt
and has the remarkable property that for t D , it simply becomes
ˇ
ˇ
h a ; I t j ˇ D h a; D 0 j U .; 0/ U.; / U .; / D h a; D 0 I j :
tD
(4.6.5)
In practice, one may choose ! ˙1, with the states h a ˙ 1; I t j describing
scattering states. The reason why they are referred to as scattering states is that
from (4.6.5),
h a ˙ 1; j U .t; / D h a ˙ 1; I t j! h a; D 0I t jD h a; D 0 j U .t; 0/; t ! ˙1:
The notation h a out j b ini for h a C 1; j b 1; i in such limits is often used.

Quite generally, however, we may introduce states:
h b 2 ; I t j; j a 1 ; I ti; and consider 1 t 2 : (4.6.6)
It is thus convenient to introduce the operator
V.t; / D U .t; 0/ U.t; /; (4.6.7)
and rewrite (4.6.2) as
h a ; j D h a; D 0 j V.; /: (4.6.8)
The operator V.t; / satisfies the equation
d
i V .t; / D U .t; /ŒH.t; / H.t; 0/U.t; 0/; (4.6.9)
dt
with ŒH.t; /H.t; 0/ denoting the part of the Hamiltonian depending on coupling
parameters and external sources that we wish to vary, which, as mentioned above,
are denoted collectively by . The Hamiltonian H.t; 0/ need not even to be

specified and may depend on other parameters that we do not wish to vary.
We are interested in investigating the dynamics in a time interval 1 < t <
2 . The QDP is particularly involved with the study of the variation of amplitudes
h b 2 ; j a 1 ; i with respect to one or more of the components of , such as
coupling parameters and of external sources. A derivative with respect to standing
for an external source will then denote a functional derivative (see, e.g., Sect. 3.6).
Using (4.6.9), the following identity is easily derived for ¤ 2 , ¤ 1 and
¤ 0 :
dh i
i V.2 ; /V .; /V.; 0 /V .1 ; 0 /
d
h i
D V.2 ; / U .; / H.; / H.; 0 / U.; 0 / V .1 ; 0 /: (4.6.10)
We integrate the latter over from 1 to 2 to obtain

h i
V.2 ; /V .1 ; / V.2 ; 0 /V .1 ; 0 /
Z 2 h i
D i V.2 ; / d U .; / H.; / H.; 0 / U.; 0 / V .1 ; 0 /:
1
(4.6.11)
For 0 infinitesimally close to , the following variation arises for infinitesimal

•,
h i h Z 2 i
• V.2 ; /V .1 ; / D i V.2 ; / d U .; /•H.; /U.; / V .1 ; /:
1
(4.6.12)
We take the matrix element of the above between the - independent states
hb; D 0 j, j a; D 0i, i.e., •hb; D 0 jD 0, • j a; D 0i D 0, and use the
expression for the - dependent states in (4.6.2)/(4.6.8). This gives
Z 2
•h b 2 ; j a 1 ; i D i h b 2 ; j d U .; / •H.; / U.; / j a 1 ; i:
1
(4.6.13)
The Hamiltonian H.; / is to be expressed in terms of the independent fields,
denoted collectively by , and their canonical conjugate momenta, denoted by
, as well as of the parameters . The independent fields are those for which
their canonical conjugate momenta do not vanish. The Hamiltonian H.t; / is then
written as
H.t; / D H.; ; ; t/: (4.6.14)

We may define the Heisenberg picture of H.t; / by
H.t; / D H..t/; .t/; ; t// D U .t; / H.; ; ; t/ U.t; /; (4.6.15)
relating the Schrödinger to the Heisenberg pictures, where now .t/, .t/, may
depend on . The parameters and the numerical t- dependence in H.:; :; ; t/ are
not affected by the transformation. Accordingly, (4.6.13) may be rewritten as
Z 2
•h b 2 ; j a 1 ; i D i h b 2 ; j d •H.; / j a; 1 ; i; (4.6.16)
1
provided it is understood that in the variation •H.; /, with respect to any of the
components of , the fields ./, ./ are kept fixed as dictated by the expression
in the integrand in (4.6.13) and by the steps going from (4.6.13), (4.6.14), (4.6.15)
and (4.6.16).
To simplify the notation we will use the notation h b t j for the - dependent
states h b t; j as no confusion may arise.
Equation (4.6.16) is the celebrated Schwinger Dynamical Principle also known
as the Quantum Dynamical Principle, for parameters and external source variations,
expressed in terms of the Hamiltonian and of the - dependent interacting states
h b j; j a i. The interesting thing to note is that although these states depend on ,
in the variation of h b 2 j a 1 i, with respect to , the entire variation on the right-
hand side of (4.6.16) is applied directly to the Hamiltonian (with the fields ./,
./ kept fixed). This is thanks to the key identity derived in (4.6.10).
We also consider variations of the matrix elements of an arbitrary function of the
variables indicated
B..t/; .t/; ; t/ B.t; /: (4.6.17)
To the above end, we subsequently use the identity
V.2 ; / B.; / V .1 ; /
D V.2 ; / V .; / U .; 0/ B.; ; ; / U.; 0/ V.; / V .1 ; /; (4.6.18)
where B.; ; ; t/ is in the Schrödinger picture.

By successive applications of (4.6.10), we have from (4.6.18), for 1 < < 2
Z 2

•Œ V.2 ; / B .; / V .1 ; / D i V.2 ; / d 0 •H. 0 ; / B.; / V .1 ; /

Z
C V.t2 ; / •B.; / V .1 ; / i V.t2 ; / d 0 B.; / •H. 0 ; / V .1 ; /:
1
(4.6.19)
where •B.; / is defined the same way as •H.; /, i.e., by varying and keeping
the fields .t/, .t/ fixed.
Using the definition of the time-ordered product

•H. 0 ; /B.; / C
D •H. 0 ; /B.; / . 0 / C B.; /•H. 0 ; / . 0 /;
(4.6.20)
and taking the matrix element of (4.6.19) between the states ha; D 0 j, j b; D 0i,
as before, the following equation emerges
•h b 2 j B.; / j a 1 i
Z 2
D i h b 2 j d 0 •H. 0 ; / B.; / C j a 1 i C h b 2 j •B.; / j a 1 i:
1
(4.6.21)
We may introduce an effective Lagrangian related to H..t/; .t/; t/ by

Z

L ..t/; .t/;
P ; t/ D d3 x .x/ .x/
P H..t/; .t/; ; t/; (4.6.22)
with a summation over the fields in the first term on the right-hand side
understood. The canonical conjugate momenta .x/ of the fields are defined
through
Z
L ..t/; .t/;
P ; t/ L ..t/; .t/
P •.t/;
P ; t/ D d3 x .x/•.t/:
P (4.6.23)
The variation of the expression in (4.6.22), with respect to , obtained by

keeping .t/, .t/ fixed is then given by
Z Z

3

d x .x/ •.t/
P C •L j D d3 x .x/ •.t/
P •H j; (4.6.24)
where the first term on the left-hand side comes from the variation of L , with
respect to .t/,
P as given in (4.6.23), and hence
•L j D •H j: (4.6.25)
Here it is important to note that •L j denotes variation with respect to keeping
.t/, .t/
P fixed, while •H j denotes variation with respect to keeping .t/, .t/
fixed.
Consider the following two types of dependent fields:
1. Fields for which their canonical conjugate momenta vanish and for which the
Euler-Lagrange equations give rise to constraints for them allowing one to
eliminate them in favor of the independent fields .x/ and the corresponding
canonical conjugate momenta .x/ as well as the parameters , in general.

When a Lagrangian does not involve the time derivative of a given field, its
canonical conjugate momentum vanishes. The important thing to note here is that
due to such constraints, these dependent fields may have an explicit dependence
on the parameters . These dependent fields will be denoted collectively by
%, and the constraints are derivable ones. Examples of such fields, working in
the Coulomb gauge, are the zeroth components A0 .x/ and A 0a .x/ of the vector
potentials in QED and QCD, respectively.
2. Consider dependent fields arising from constraints, one may, a priori, impose on
functions of the fields, involving no time derivatives, and applied directly to the
Lagrangian. That is, some fields or field components are written as functions of
other fields or their components and may involve their space derivatives. In short,
these dependent fields are expressed in terms of independent fields and perhaps
their space derivatives. Since the variations encountered above, with respect to
, are carried out with the independent fields kept fixed, such dependent fields
are automatically kept fixed in these variations unless they are also chosen to
have explicit - dependence. Such constraints are, a priori, imposed constraints.
Examples of such constraints are the Coulomb gauges r A D 0, r A a D 0
in QED and QCD, respectively. Here we see that one of the three components
may be expressed in favor of the other two, and, needless to say, not all the
components may be varied independently in deriving Euler-Lagrange equations.
After the dependent fields %, as just discussed in point 1. above, are eliminated
from the above constraints, the Lagrangian L.t; / L..t/; .t/;
P %.t/; ; t/ of the
system becomes replaced by the effective Lagrangian L ..t/; .t/;
P ; t/ considered
earlier in (4.6.22), i.e.,
L..t/; .t/;
P %.t/; ; t/ D L ..t/; .t/;
P ; t/: (4.6.26)
R 0
The variation of the action dt L.t; / with respect to , with .t/, .t/
P kept
fixed, comes from the explicit dependence of L.t; / on , and may also come from
the dependence of % on , as arising from the corresponding constraint equations.
The fields % satisfy Euler-Lagrange equation obtained by carrying out functional
derivatives of the action
Z x00 D 0
•
E% .x/ .dx0 / L .x0 ; / D 0; (4.6.27)
•%.x/ x00 D
for fixed , 0 , involving no surface terms, since, by definition, the canonical

conjugate momenta 0 D @L .x; /=@.@ 0 %.x// D 0, due to the fact that L .x; /
contains no time derivatives of the fields %.x/. Here L .x; / is the Lagrangian
density given through
Z
L.t; / L..t/; .t/;
P %.t/; ; t/ D d3 x L .x; /: (4.6.28)
Accordingly the variation of the action in (4.6.27), with respect to , keeping

.t/, .t/
P fixed, is from (4.6.26), (4.6.27)
Z x 0 D 0 Z 0 Z 0

.dx/ •%.x/E% .x/ C dt •L.t; /j D dt •L .t; /j: (4.6.29)
x 0 D
Hence,
Z 0 Z 0
dt •L.t; /j D dt •L .t; /j; (4.6.30)

where •L.t; /j stands for variations with respect to , with .t/, .t/ P as well as
the dependent fields %.t/ all kept fixed.
In particular, upon taking the limits 1 ! 1; 2 ! C1, and using the
notations ja i for ja 1 i; hbCj for h b 2 j, in these limits, (4.6.16), (4.6.25), (4.6.28),
(4.6.30), lead to
Z
•h bC j a i D i h bC j .dx/ •L .x; / j a i: (4.6.31)
The variations in •L .x; / in (4.6.31), with respect to , are from (4.6.28) carried
out by keeping the independent fields .x/ and the dependent fields %.x/ as well as
the derivatives @ .x/, r%.x/ all kept fixed.
Equation (4.6.31) is another version of the Schwinger Dynamical Principle, also
known as the Quantum Dynamical Principle (QDP), for parameters and external
source variations, expressed in terms of the Lagrangian density L .x; / and the
-dependent interacting states hbC j, j a i.
The expression for the QDP in (4.6.31) takes on a particular simple and a very
useful form by choosing H.t; 0/ H0 to be the free time-independent Hamiltonian,
and j a; D 0i, j b; D 0i its ground (vacuum) state: H 0 j 0i D 0. This gives for
the variation of the vacuum-to-vacuum transition amplitude
Z
•h 0C j 0 i D i h 0C j .dx/ •L .x; / j 0 i: (4.6.32)
One should note that other standard notations for the vacuum-to-vacuum tran-
sition amplitude h 0C j 0 i one may encounter are h 0 out j 0 ini, h vac.out/ j
vac.in/i. As we have already seen in Sect. 3.6, h 0C j 0 i is not necessarily just a
phase factor.
The very compact equation in (4.6.32) provides a wealth of information in
quantum field theory and is of central importance in applications as we have
witnessed in Sect. 3.3 and in Sect. 3.6, and as will be seen again and again in later
sections and subsequent chapters.
As an example, consider the Lagrangian density of a neutral scalar field9 .x/
L .x; / D L0 .x/ C LI .x/ C K.x/.x/; (4.6.33)
where L0 .x/, LI .x/ may denote the free and interacting parts, K.x/ is an external
(c-number) source, and a coupling parameter. Hence
Z
@
h bC j a i D i h bC j .dx/ LI .x/ j a i; (4.6.34)
@
and, as a functional derivative,

Z
• •
.i/ h bC j a i D h bC j .dx 0 / L .x 0 ; / j a i D h bC j .x/ j a i;
•K.x/ •K.x/
(4.6.35)
by using the functional derivative
•
K.x 0 / D •.4/ .x 0 x/: (4.6.36)
•K.x/
The QDP in terms of the vacuum-to-vacuum transition amplitude h 0C j 0 i

in (4.6.32) should be especially noted for applications. For earlier applications, see
Sects. 3.3 and 3.6, and many others will follow in coming sections and chapters as
mentioned above. A summary of the first steps needed for applications are spelled
out below.
4.6.1 Summary
Now we provide a summary of how the QDP is applied to the vacuum-to-vacuum

transition amplitude which is the starting point in many applications:
We are given a Lagrangian density L .x; / depending on parameters consisting
of coupling constants and external sources, denoted collectively by . The variation
of the vacuum-to-vacuum transition amplitude •h 0C j 0 i with respect to any of
the components of is given by:
Z
•h 0C j 0 i D i h 0C j .dx/ •L .x; / j 0 i; (4.6.37)
9
The scalar field is just a field which remains invariant under Lorentz transformations: 0 .x 0 / D
.x/. Details on basic fields are given in the next section.
where the variation •L .x; / is carried out on the explicit dependence of L .x; /
on , that is all the fields and their derivatives in it are kept fixed.
We may also be given a function F..x/; .x/; %.x/; / of the independent
fields .x/, their canonical conjugate momenta .x/, and, as denoted, perhaps of
dependent fields %.x/ and of . Suppose through constraint equations as obtained,
for example, from the Euler-Lagrange equations, the dependent fields may be
expressed in terms of ; , and at most of their spatial derivatives, and may be
of (the components) of . Let the resulting expression for F be denoted simply
by B.x; /. Then the variation of h 0C j B.x; / j 0 i with respect to any of the
components of is given by
Z

•h 0C j B.x; / j 0 i D ih 0Cj .dx 0 / •L .x 0 ; /B.x; / Cj 0 i C h 0C j •B.x; / j 0 i:
(4.6.38)
where the variation •B.x; / is carried out with ; , together with their spatial
derivatives if any, kept fixed. Also . : /C , as usual, denotes time-ordering. The
second term will give a contribution only if B.x; / depends explicitly on . In such
cases the variation of the above matrix element will pick up a contribution coming
from the second term on theright-hand side in addition to the first integral involving
the time ordered products •L .x 0 ; /B.x; / C . This causes some complications
in the quantization of non-abelian gauge theories, and some abelian ones as well,
but is easily handled by the application of (4.6.38), and is the origin of the so-
called Faddeev-Popov factor, and its generalizations. These intricate details will
be worked out later for such theories. An elementary application of this is given
in Sect. 4.8 as well as other applications are given in coming chapters.10 The path
integral formalism will be also considered.
4.7 A Panorama of Fields
This section deals with the basic quantum fields one often encounters in quantum
field theory/high-energy physics. The fields in question are of spin 0, 1, 3/2 and 2, for
massive and massless particles. The relevant technical details are relegated to their
corresponding subsections that follow. The technical details concerning spin 1/2,
the Dirac field, however, are spelled out in Chap. 3, while that of the massless spin
1, the photon, are the subject matter of Chap. 5. Needless to say, the corresponding
subsections to follow of all these fields, form an integral part of this section and
10
There is a long history with the quantum dynamical principle, see, Schwinger [10, 12, 14, 15],
Lam [1], and Manoukian [3, 4].
4.7 A Panorama of Fields 177
should be read simultaneously with the text material. We now proceed to summarize
some of the salient features of all these fields.
4.7.1 Summary of Salient Features of Some Basic Fields
Spin 0:
The Lagrangian density of a Hermitian scalar field ' with corresponding field
equation are
1 m2 2
L .x/ D @ '.x/@ '.x/ ' .x/ W . C m2 / '.x/ D 0; (4.7.1)
2 2
while the propagator is given by
Z 0
.dp/ eip .xx /
i hvac j '.x/'.x 0 / C j vaci D C .x x 0 / D ; (4.7.2)
.2/4 p2 C m2 i
and satisfies
. C m2 / C .x x 0 / D ı .4/ .x x 0 /: (4.7.3)
The local commutativity relation reads

Z
1 .dp/ ip .xx 0 /
Œ '.x/; '.x 0 / D i .xx 0 /; .xx 0 / D e ı. p2 Cm2 / ". p0/;
i .2/3
(4.7.4)
where ". p 0 / D C1 for p 0 > 0, D 1 for p 0 < 0, and .x x 0 / vanishes for
space-like separations .x x 0 /.
In the presence of an external source K, described by the substitution L .x/ !
L .x/ C K.x/ '.x/, the vacuum-to-vacuum transition amplitude is given by
hiZ i
h 0C j 0 i D exp .dx/.dx 0 /K.x/ C .x x 0 /K.x 0 / : (4.7.5)
2
For a non-Hermitian scalar field,
L .x/ D @ ' .x/@ '.x/ m2 ' .x/'.x/; (4.7.6)

2 2
. C m / '.x/ D 0; . C m / ' .x/ D 0; (4.7.7)
Œ '.x/; '.x 0 / D 0; Œ '.x/; ' .x 0 / D i .x x 0 /; (4.7.8)
Z 0
.dp/ eip .xx /
i hvac j '.x/ ' .x 0 / C j vaci D C .x x 0 / D :
.2/4 p 2 C m2 i
(4.7.9)
For L .x/ ! L .x/ C K .x/ '.x/ C ' .x/ K.x/,

h Z i
h 0C j 0 i D exp i .dx/.dx 0 /K .x/ C .x x 0 /K.x 0 / : (4.7.10)
Spin 1/2:
The Lagrangian density and the field equations are
1
L .x/ D .x/ @ .x/ .@ /.x/ .x/ m .x/ .x/; (4.7.11)
2i
@ @
C m .x/ D 0; .x/ C m D 0; (4.7.12)
i i
while the local anti-commutativity relation is

0 @ 1
f a .x/; b .x /g D Cm .x x 0 /; (4.7.13)
i ab i
where .x x 0 / is defined in (4.7.4), and the propagator is given by

Z
.dp/ . p C m/ ip .xx 0 /
i hvac j .x/ .x 0 / C
j vaci D SC .x x 0 / D e ;
.2/4 p 2 C m2 i
(4.7.14)
satisfying
@ @
0 .4/ 0 0
C m SC .x x / D ı .x x /; SC .x x/ C m D ı .4/ .x 0 x/:
i i
(4.7.15)
In the presence of external sources ; : L .x/ ! L .x/ C .x/ .x/ C .x/.x/,
the vacuum-to-vacuum transition amplitude is given by
h Z i
h 0C j 0C i D exp i .dx/.dx 0 / .x/ SC .x x 0 / .x 0 / : (4.7.16)
Spin 1:
˘ m ¤ 0:
For a Hermitian massive vector field V the Lagrangian density may be defined
as
1 m2
L .x/ D F .x/F .x/ V .x/ V .x/; F .x/ D @ V .x/ @ V .x/;
4 2
(4.7.17)
with corresponding field equations and a derived constraint
@ F C m2 V D 0; ) @ V D 0; . C m2 / V D 0; (4.7.18)
and
@@
Œ V .x/; V .x 0 / D i .x x 0 /; (4.7.19)
m2
where .x x 0 / is defined in (4.7.4). The propagator

Z
e ip .xx / p p
0
.dp/
4 C .x x 0 / D C 2 ; (4.7.20)
.2/4 p 2 C m2 i
m
satisfies

. C m2 / C @ @ 4 C .x x 0 / D ı ı .4/ .x x 0 /: (4.7.21)
For L .x/ ! L .x/ C K .x/V .x/,

hi Z i

h 0C j 0 i D exp .dx/.dx 0 /K .x/4 C .x x 0 /K .x 0 / : (4.7.22)
2
˘ m D 0:
For a massless (Hermitian) vector field in the presence of an external source K ,
the Lagrangian density in covariant gauges may be taken as
1
L D F F @ V C 2 C K V ; F D @ V @ V ; (4.7.23)
4 2
and is a field, referred to an auxiliary field, which may be eliminated and specifies
the gauge. The field equations are:
@ V D ; @ F D K C .1 /@ ; D @ K ; (4.7.24)
and the vacuum-to-vacuum transition amplitude takes the form

Z

i
h 0C j 0 i D exp .dx/.dx 0 /K .x/D .x x 0 /K .x 0 / ; (4.7.25)
2
where
Z
.dp/ h p p i eip .xx /
0
0
D .x x / D .1 / 2 ; (4.7.26)
.2/4 p p2 i
and is the covariant photon propagator for K D 0. Also the gauges defined by
D 1; D 0; D 3, are, respectively, referred to as the Feynman gauge, the
Landau gauge and the Yennie-Fried gauge.
In the so-called Coulomb gauge: @i V i .x/ D 0 as applied to the lagrangian
density
1
L D F F C K V ; (4.7.27)
4
the field equations are:
1
@ F D j 2 @j @ K ; (4.7.28)
@
and
hi Z i

h 0C j 0 i D exp .dx/.dx 0 /K .x/DC .x x 0 /K .x 0 / ; (4.7.29)
2
where
Z
0 .dQ/ iQ .xx 0 /
DC .x x /D e DC .Q/; (4.7.30)
.2 /4
1 Qi Qj 1
D00 ; D0Ci .Q/ D 0;
ij
C .Q/ D ; DC .Q/ D ı ij (4.7.31)
Q2 Q 2 Q2 i
defines the propagator in the Coulomb gauge for K D 0. For a conserved external
current source @ K D 0,
hi Z i
h 0C j 0 i D exp .dx/.dx 0 /K .x/DC .x x 0 /K .x 0 / ; (4.7.32)
2
Z 0
0 .dp/ eip .xx /
DC .x x / D : (4.7.33)
.2/4 p2 i
Spin 3=2:
˘ m ¤ 0:
The Lagrangian density of a massive spin 3/2 field, also referred to as the Rarita-
! !
Schwinger field,11 may be taken as . @ @ , @ D @ @ /
1 ! ! ! !
L D @ @ C @ C . @ /
2i
h i
m C ; (4.7.34)
11
Schwinger [18, p. 191] and Rarita and Schwinger [8].
with a carrying both a spinor index and a Lorentz one. The field equations are
given by
h @ @ @ @ i
C m C C C m .x/ D 0;
i i i i
(4.7.35)
or equivalently
@

Cm .x/ D 0; .x/ D 0; ) @ .x/ D 0; (4.7.36)
i
and with local anti-commutation relation
0 1 @
f a .x/; b .x /g D .x x 0 /; (4.7.37)
i i
@ @ @ @ 1 @ @ @
D C m C C C m
C ;
i i m2 3 im i im
(4.7.38)
where .x x 0 / is defined in (4.7.4). The propagator is given by

Z 0
0 .dp/ eip .xx /
C a b .x x /D . p/; (4.7.39)
.2/4 p2 C m2 i
a b
h 1 p p 2 i
. p/ D . p C m/ C C C 2 p p ;
3 m m m
(4.7.40)

p p
1 p p
D . p C m/ C C C . p C m/
C ;
m2 3 m m
(4.7.41)
and satisfies
h @ @ @ @ i
C m C C C m C bc .x x 0 /
i i i i ab
D ıac ı ı .4/ .x x 0 /:
(4.7.42)

In the presence of external sources K ; K , with L .x/ ! L .x/CK .x/ .x/

C .x/K .x/,
Z

h 0C j 0 i D exp i .dx/.dx 0 /K a .x/ C ab .x x 0 / Kb .x 0 / : (4.7.43)
˘ m D 0:
The Lagrangian density for m D 0, in the presence of external sources, may be
taken as . @ @ , r i @i , D @ @ /12
1 ! ! ! !
L D @ @ C @ C . @ / CK C K :
2i
(4.7.44)
We work in a Coulomb-like gauge @ i a .x/

i
D 0. This leads to the field equations:
r 0
i i
D i K 0; (4.7.45)
r2
0 1 @ 0 1 h@ i r ii 0
D K C K i; (4.7.46)
2 r2 i 2r 2 i i
@ ij j h @i r ii 0
i
D ˛ K 2
C K 0; (4.7.47)
i 2r i i
1 i k @ i @ k ` k `j @ ` @ j
˛ ij D ı 2 ı 2 : (4.7.48)
2 r r
The vacuum-to-vacuum transition amplitude becomes
Z

h 0C j 0 i D exp i .dx/.dx 0 /K a .x/DC ab .x x 0 /Kb .x 0 /; (4.7.49)
where
Z
.dp/ ip .xx 0 /
DC ab .x x 0 / D
e DC ab . p/; (4.7.50)
.2/4
. p/ ij 1 i k p i p k ` p k `j p` p j
DC ij . p/ D 2 ˛ . p/ D 2 ı 2 ı 2 ;
p i
p i
p 2 p
(4.7.51)
1 i k p i p k ` k `j p ` p j 1
˛ ij . p/ D ı 2 ı 2 ; DC 0 0 . p/ D 2 . p/;
2 p p 2p
(4.7.52)
1 i 1 i
DC 0 i . p/ D 2
p C p i 0 ; DC i 0 . p/ D 2
p C p i 0;
2p 2p
(4.7.53)

defines the propagator for K D 0; K D 0.
12
The massless Rarita-Schwinger field treatment is based on Manoukian [7].
Spin 2:
˘ m ¤ 0:
The Lagrangian density of a massive Hermitian spin 2 field may be taken as
1
L D @ U @ U C @ U @ U @ U @ U
2
1 m2
C @ U @ U U U U U ; (4.7.54)
2 2
where U D U , U U , and leads to the field equations
. C m2 /U˛ˇ C @˛ @ U ˇ C @ˇ @ U ˛ @˛ @ˇ U

˛ˇ . C m2 /U C @ @ U D 0: (4.7.55)
The propagator is given by

Z
˛ ˇI .dp/ ip .xx 0 / ˛ ˇI
C .x x 0 / D e C . p/; (4.7.56)
.2/4
1 1 1
˛ ˇI
C . p/ D 2 ˛ . p/ˇ . p/ C ˛ . p/ˇ . p/
p C m2 i
2 2
1 p p
˛ˇ . p/ . p/ ; . p/ D C 2 :
3 m
(4.7.57)
In the presence of an external source T , L .x/ ! L .x/ C T .x/U .x/,

hiZ
h 0C j 0 i D exp .dx/.dx 0 /T ˛ˇ .x/C ˛ˇI .x x 0 /T .x 0 /: (4.7.58)
2
˘ m D 0:
The Lagrangian density for the massless case in the presence of an external
source T , may be taken as
1
L D @ U @ U C @ U @ U @ U @ U
2
1
C @ U @ U C T U : (4.7.59)
2
A covariant gauge treatment of massless spin 2 dealing with the graviton will be
carried out in Sect. 2.3.1 of Volume II. On the other hand, in a Coulomb-like gauge
@i U i D 0, where U D U , U U , the propagator is given by
Z
I 0 .dp/ ip .xx 0 / I
DC .x x / D e DC . p/; (4.7.60)
.2/4
ij I k` .p/ p ip j
ij I k`
DC . p/ D ij .p/ D ı ij 2 ;
; (4.7.61)
p2 i
p

1 ik
ij I k` .p/ D .p/ j ` .p/ C i ` .p/ j k .p/ ij .p/ k ` .p/ ; (4.7.62)
2
ij I 00 1 00 I 00 1 p2
DC . p/ D ij .p/; DC . p/ D ; (4.7.63)
2 p2 2 p2 p2
0iI0k 1 00 I 0 i 0 i Ij k
DC . p/ D i k .p/; DC . p/ D 0; DC . p/ D 0: (4.7.64)
2 p2
and
hiZ i
I
h 0C j 0 i D exp .dx/.dx 0 /T .x/DC .x x 0 /T .x 0 / : (4.7.65)
2
For a conserved source @ T D 0,

h iZ 1 i
h 0C j 0 i D exp .dx/.dx 0 / T .x/T .x 0 / T .x/ T .x 0 / DC .x x 0 / ;
2 2
(4.7.66)
with DC .x x 0 / given in (4.7.33).

Now we carry out the details leading to the equations displaced above in this
section.
4.7.2 Spin 0
A scalar field '.x/ remains invariant under a Lorentz transformation, and hence, in
particular, under an infinitesimal one
' 0 .x 0 / D '.x/; x 0 D x C •! x ; •x D x x 0 : (4.7.67)
The latter may be rewritten as
' 0 .x/ D '.x C •x/ D '.x/ C •x @ '.x/; (4.7.68)

0
•'.x/ D '.x/ '.x / D •! x @ '.x/: (4.7.69)
Or finally as
i x @ x @

•'.x/ D •! ; (4.7.70)
2 i

where x @ x @ =i corresponds to the orbital angular momentum, and a spin
zero.
The field equation, in the presence of an external source K.x/, is defined by
. C m2 / '.x/ D K.x/; (4.7.71)
which may obtained from the Lagrangian density for a Hermitian scalar field '.x/,
1 1
L .x/ D @ '.x/ @ '.x/ m2 '.x/2 C K.x/'.x/: (4.7.72)
2 2
Upon taking the matrix elements h 0C j : j 0 i of (4.7.71), the following set of

equations emerge by finally carrying out a functional integration over K.x/ :
. C m2 /h 0C j '.x/ j 0 i D K.x/h 0C j 0 i; (4.7.73)

Z
h 0C j '.x/ j 0 i D .dx 0 /4C .x x 0 /K.x 0 / h 0C j 0 i; (4.7.74)
Z
•
.i/ h 0C j 0 i D .dx 0 /4C .x x 0 /K.x 0 /h 0C j 0 i; (4.7.75)
•K.x/
hiZ i
h 0C j 0 i D exp .dx/.dx 0 /K.x/4C .x x 0 /K.x 0 / ; (4.7.76)
2
where 4C .x x 0 / is the propagator

Z 0
.dk/ eik.xx /
4C .x x 0 / D ; .dk/ dk0 dk1 dk2 dk3 ;
! 0:
.2/ k C m2 i
4 2
(4.7.77)
The latter is defined as the vacuum expectation value of a time ordered product as
follows
ˇ
ˇ
i h 0C j '.x/'.x 0 / C j 0 iˇ
D 4C .x x 0 /: (4.7.78)
KD0;K D0
The equal-time commutation relation emerges by introducing, in the process,

D @L =@.@ '/ D @ ', D @ 0 ' D ',
P
P 0 / D i ı 3 .x x 0 /;
Œ '.x/; '.x x 0 D x 0 0: (4.7.79)
To discuss the spin and statistics connection, consider a first order formulation,
with Lagrangian density
1 1 1 m2 2
L D F @ ' ' @ F @ ' F @ F '/ C F F ' ;
4 4 2 2
(4.7.80)
which leads to
F D @ '; @ F D m2 ': (4.7.81)

p p
Define .a / D .F = m; m '/, having five components, and introduce the
following matrices, referred to as Duffin-Kemmer-Petiau matrices,
0 1 0 1
0 0001 0 0 00 0
B 0 0 0 0 0C B 0 0 0 0 1 C
B C B C
0 B C 1 B C
ˇ D iB 0 0 0 0 0C; ˇ D iB0 0 0 0 0 C; (4.7.82)
B C B C
@ 0 0 0 0 0A @0 0 0 0 0 A
1 0 0 0 0 0 1 0 0 0
0 1 0 1
00 0 0 0 000 0 0
B0 0 0 0 0 C B0 0 0 0 0 C
B C B C
B C B C
ˇ 2 D i B 0 0 0 0 1 C ; ˇ 3 D i B 0 0 0 0 0 C ; (4.7.83)
B C B C
@0 0 0 0 0 A @ 0 0 0 0 1 A
0 0 1 0 0 0 0 0 1 0
and D diag Œ1; 1; 1; 1; 1, to rewrite (4.7.80) as

1 m
L D Q @ @ Q C ; (4.7.84)
4 2
where
ˇ
Q D ; .Q0 iˇ 0 /; (4.7.85)
i
satisfying
.Q / D Q ; D ; (4.7.86)
ensuring the Hermiticity of the Lagrangian density, and
.Q /> D Q ; > D ; (4.7.87)
establishing the Bose character of the field.13 Here matrix multiplication is defined,
e.g, as D a a b b . The field equation and the equal-time commutation
relation that follow from this formalism are (see Sect. 4.5)
@

ˇ C m D 0; ı.x 0 x 0 0 / Œ a .x/; b .x 0 / Q0b c D i ıac ı .4/ .x x 0 /;
i
(4.7.88)
13
See (4.5.2), (4.5.20), (4.5.22), and (4.5.23).
and, in turn, lead to the field equation (for K D 0), and the equal-time commutation
relation given in (4.7.71) and (4.7.79), respectively. In addition it gives rise to the
local commutativity relation
ı.x 0 x 0 0 / Œ '.x/; '.x 0 / D 0: (4.7.89)
By an analysis very similar to the one carried out for the Dirac field leading
to (3.5.12), we also have the general commutation relation
Œ '.x/; '.x 0 / D i .x x 0 /; (4.7.90)
where .x x 0 / is defined in (3.5.13)) as

Z p
0 d3 p ip .xx 0 / sin p 0 .x 0 x 0 0 /
.xx / D e ; p0 D p 2 C m2 ; (4.7.91)
.2/3 p0
and is more conveniently rewritten as

Z
1 .dp/ ip .xx 0 /
.x x 0 / D e ı. p 2 C m2 / ". p0/; (4.7.92)
i .2/3
where ". p 0 / D C1 for p 0 > 0, D 1 for p 0 < 0. The Lorentz invariance

of (4.7.92) is evident since, in particular, a Lorentz transformation does not change
the sign of p0 . That is, since from (4.7.91), .x x 0 / D 0 for x 0 D x 0 0 , and x ¤
x 0 , the vector .x x 0 / is space-like: .x x 0 /2 > 0, we may infer from (4.7.92)
that .x x 0 / D 0 vanishes for all space-like separations .x x 0 / : .x x 0 /2 > 0.
.x x 0 / D 0 satisfies the equation
. C m2 / .x x 0 / D 0; (4.7.93)
as well as the key relation

ˇ
ˇ
@0 .x x 0 /ˇ D ı 3 .x x 0 /: (4.7.94)
x 0 Dx 0 0
For a non-Hermitian field, we may double the number of components consisting

of real and imaginary parts of F ; '. The commutation relations become simply
replaced by
ı.x 0 x 0 0 /Œ '.x/; ' .x 0 / D 0: (4.7.95)

ı.x 0 x 0 0 /Œ '.x/; 'P .x 0 / D i ı .4/ .x x 0 /; ı.x 0 x 0 0 /Œ '.x/; '.x 0 / D 0;
(4.7.96)

The Lagrangian density for a non-Hermitian field, in the presence of external

sources K, K in a second order formalism, may be written as
1 m2
L D @ ' @ ' C @ ' @ ' ' ' C ' ' C K ' C ' K;
2 2
(4.7.97)
with
h Z i
h 0C j 0 i D exp i .dx/.dx 0 /K .x/4C .x x 0 /K.x 0 / ; (4.7.98)
and for K D 0; K D 0,

i h 0C j '.x/' .x 0 / C j 0 i D 4C .x x 0 /; (4.7.99)
and the general commutation relations
Œ '.x/; '.x 0 / D 0; Œ ' .x/; ' .x 0 / D 0; Œ '.x/; ' .x 0 / D i .x x 0 /:

(4.7.100)
4.7.3 Spin 1
We introduce a vector field V satisfying the equations
. C m2 /V D 0; @ V D 0 (4.7.101)
We consider the case m ¤ 0. The massless case, is the subject matter of Chap. 5,
dealing with the photon. We will see that the above two equations describe a massive
spin 1 particle.
The above equations, for a Hermitian vector field, may be obtained from the
Lagrangian density
1 m2
L D F F V V ; F D @ V @ V ; (4.7.102)
4 2
leading to
@ F C m2 V D 0: (4.7.103)
Upon taking the partial drivative @ of the above equation, using the anti-symmetry
of F , and dividing, in the process, by m2 , lead to the two equations in (4.7.101).
In the presence of an external source K ,
1 m2
L D F F V V C K V ; (4.7.104)
4 2
the field equations

. C m2 / C @ @ V D K ; (4.7.105)
lead, in turn, to
Z

h 0C j V .x/ j 0 i D .dx 0 /4 C .x x 0 /K .x 0 /h 0C j 0 i (4.7.106)
Z
•
.i/ h 0C j 0 i D .dx 0 /4 C .x x 0 /K .x 0 /h 0C j 0 i: (4.7.107)
•K .x/
hiZ i

h 0C j 0 i D exp .dx/.dx 0 /K .x/4 C .x x 0 /K .x/ ;
2
(4.7.108)

where the propagator 4 C .x x 0 / is given by
Z
eik.xx / k k
0
.dk/
4 C .x x 0 / D C 2 : (4.7.109)
.2/4 k2 C m2 i
m
For K D 0, (4.7.103), gives, in particular, for D 0, the constraint equation
1
V0 D @ k F k0 : (4.7.110)
m2
Also
@L
D D F D F ; (4.7.111)
@.@ V /
hence
1
V0 D @k k ; (4.7.112)
m2
with k D F k0 denoting the canonical conjugate momenta of V k , and
ı.x 0 x 0 0 / Œ V j .x/; F k 0 .x 0 / D i ı j k ı .4/ .x x 0 /; (4.7.113)

1
ı.x 0 x 0 0 / Œ V j .x/; V 0 .x 0 / D i @ 0j ı .4/ .x x 0 /; (4.7.114)
m2
ı.x 0 x 0 0 / Œ V j .x/; V k .x 0 / D 0; (4.7.115)
ı.x 0 x 0 0 / Œ j .x/; k .x 0 / D 0; (4.7.116)

ı.x 0 x 0 0 / Œ V 0 .x/; V 0 .x 0 / D 0: (4.7.117)
Hence

ı.x 0 x 0 / Œ V .x/; F .x 0 / D i 0 0 ı .4/ .x x 0 /; (4.7.118)
@@
Œ V .x/; V .x 0 / D i .x x 0 /: (4.7.119)
m2
Under a Lorentz transformation x ! x 0 , x 0 D x C •! x , a vector field has

the transformation

V 0 .x 0 / D ı C •! V .x/
i
D ı C •! ˛ ˇ Œ S˛ ˇ V .x/; (4.7.120)
2
where
1
Œ S˛ ˇ D ı ˛ ˇ ı ˇ ˛ : (4.7.121)
i
From the submatrix
1 a
Œ Sij a b D .ı i ıj b ıj a ı i b ; i; j; a; b D 1; 2; 3; (4.7.122)
i
as the coefficient of • ! ij , corresponding to pure 3D rotations (see Sect. 2.2), one

may introduce the spin matrix Œ S k :
1 ij k
Œ S k a b D " Œ Sij a b ; (4.7.123)
2
where " ij k is totally anti-symmetric with "123 D C1, leading to
Œ S2 a c D Œ Sk a b Œ Sk b c D 2 ı a c ; (4.7.124)
establishing the spin s D 1 character of the vector field, with components satisfying
the well known commutations relations,
Œ S i ; Sj D i " ij k S k : (4.7.125)
Matrix representations of the spin components are

0 1 0 1 0 1
00 0 0 01 0 1 0
S1 D i @ 0 0 1 A ; S2 D i @ 0 0 0 A ; S3 D i @ 1 0 0 A : (4.7.126)
01 0 1 0 0 0 0 0
It is interesting to note that the rotation matrix in Sect. 2.2 for the c.w. rotation of a
3-vector by an angle about a unit vector n may be rewritten as
R D expŒ i n S : (4.7.127)

Polarization vectors may be defined for a massive vector particle e ; D
1; 0; 1, satisfying

e e 0 D ı0 ; p e D 0; p 2 D m2 ; (4.7.128)
as well as a completeness relation14

p p X p p X
D 2
C e e
D 2
C e e ; (4.7.129)
m m
D1;0;1 D1;0;1
having expressions
1 p
e C D .0; eC /; e D .0; e /; e0 D jpj; p 0 ; (4.7.130)
im jpj

e ˙ D e
; e0 D e0: (4.7.131)
A first order description of the vector field is given by the Lagrangian density
1 ! 1 ! 1 m2
L D F @ V C V @ F C F F V V : (4.7.132)
4 4 4 2
p p p F 01 F 02 F 03 p F 12 F 13 F 23
. a / D m V 1; m V 2; m V 3; p ; p ; p ; m V 0; p ; p ; p :
m m m m m m
(4.7.133)
In particular, the term involving the time derivative in (4.7.132) is given by
0 1
0 I
1 a 0 ! b 0
.Q /a b @ 0 ; Q 0 D @ I 0 A ; (4.7.134)
4
0 0
14
See also Problem 4.10.
where Q 0 is a 10 10 matrix, and I is the 3 3 unit matrix. We note that

.Q 0 / D Q 0 , for hermiticity, and .Q 0 /> D Q 0 , in particular, establishing the
Bose character of the vector field.
For a non-Hermitian vector field, the Lagrangian density may be written as

1
L D F F m2 V V C K V C V K ; (4.7.135)
2
and
h Z i

h 0C j 0 i D exp i .dx/.dx 0 /K .x/4 C .x x 0 /K .x 0 / ; (4.7.136)

where the propagator 4 C .x x 0 / is given in (4.7.109).
4.7.4 Spin 3=2

We consider a field carrying a Lorentz index and a Dirac spinor index a.
a
! !
We will see, in particular, that the set of equations . @ @ , @ D @ @ /
@

Cm .x/ D 0; .x/ D 0; (4.7.137)
i
suppressing the spinor index for simplicity of the notation, describes a spin 3=2
field. The latter is referred to as the Rarita-Schwinger field.
Upon multiplying the above equation by and using the identity D
2 leads to the condition

@ .x/ D 0: (4.7.138)
The first equation in (4.7.137) together with the constraint on its right-hand side,
may be obtained from the single equation15
h @ @ @ @ i
C m C C C m .x/ D K .x/;
i i i i
(4.7.139)
15
This is not the only equation which leads to the equations in (4.7.137). For example,
the
expressions within brackets in (4.7.139) replaced by .@=i C m/ .1=3/ @ =i C

@ =i C .1=3/ @=i C m , also leads to the equations in question. This merely
gives rise to adding a non-propagating term to the propagator in (4.7.144), i.e., a non-singular
contribution on the mass shell p2 D m2 , see Problem 4.11.
for K D 0. For greater generality and for easily deriving the expression for the
propagator, we have introduced a source term K in the above equation.
Upon comparison of the two equation obtained by multiplying the above
equation by and @ , respectively, one obtains for m ¤ 0
1 @
D 2 K; (4.7.140)
3m im
@ @ @ K
@ D 2 K C : (4.7.141)
3m im m
By substituting, in turn, these two equations in (4.7.139), gives
@ 1h @ @ 2 @ @ i

Cm D K C C C 2 K :
i 3 im im m i i
(4.7.142)
Hence for K D 0, the equations in question in (4.7.137), (4.7.138) are satisfied.

The expression for the propagator simply follows from (4.7.142) upon multiply-
ing, in the process, by . @=i C m/ leading to
Z

h 0C j a .x/ j 0 i D .dx 0 / C a b .x x 0 /Kb .x 0 /h 0C j 0 i; (4.7.143)
Z 0
0 .dp/ eip .xx /
C a b .x x /D . p/;
! 0; (4.7.144)
.2/4 p 2 C m2 i
ab
h 1 p p 2 i
. p/ D . p C m/ C C C 2 p p
3 m m m

p p
1 p p
D . p C m/ C C C . p C m/
C :
m2 3 m m
(4.7.145)
The field equations in (4.7.142) may be obtained from the Lagrangian density
1 ! ! ! !
L D @ @ C @ C . @ /

2i
h i
m C C K C K ; (4.7.146)
where D 0 as usual.

From the QDP, h 0C j a .x/ j 0 i D i •=•K a .x/ h 0C j 0 i, we may
integrate (4.7.143) to obtain
Z

h 0C j 0 i D exp i .dx/.dx 0 /K a .x/ C ab .x x 0 / Kb .x 0 / : (4.7.147)

For K D 0; K D 0, the field equation (4.7.142), together with the conditions
in (4.7.140), (4.7.141) lead to a constraint on the components of i as well (see
Problems 4.12 and 4.13), given by
h @i r i
C m i i
.x/ D 0; (4.7.148)
i i
involving no time derivatives. This constraint may be taken into account in setting
up the equal-time anti-commutator of the spatial components i .x/ by defining
1 i j @j @i 2 @i @j
ˇ ij D ı ij C i C j C 2 ; (4.7.149)
3 im im m i i
satisfying the constraint equation
h@ i r i
C m i ˇ ij D 0: (4.7.150)
i i
The term involving the time derivative in (4.7.146) is given by
ˇ 1 ij !
L ˇtime der: D i
ı C i j @ 0 j
: (4.7.151)
2i
But not all the components of i are independent due to the constraint in (4.7.148).
Thus we may introduce a field with components j .x/, and set
h@ i r i
i
.x/ D ˇ ij j .x/; C m i i
.x/ D 0; (4.7.152)
i i
from which (4.7.151) becomes
ˇ 1 !
L ˇtime der: D i
ı ij C i j ˇ j k @ 0 k .x/; i
D .ˇ ij j .x// ;
2i
(4.7.153)
and we may vary the components of k .x/ independently.
We now use the identity

ij h 1 j k @k @j 2 @ j @ k i
ı C i j ı jk C Cj k C 2
3 im im m i i

h @i r i 1 2 k
D ıi k Ci m k @ ; (4.7.154)
i i 3m 3 m2
i
together with the constraint (4.7.148) as applied to
h @i i
i r
i
C m D 0; (4.7.155)
i i
to infer from (4.7.153), that the canonical conjugate momentum of jc .x/ is simply
j j
i c . Thus the equal-time anti-commutation relations of kc and of i b is simply
given by
0
D i ıcb ı kj ı 3 .x x 0 /;
j
f kc .x/; i b .x /g (4.7.156)
and upon multiplying it by .i/ˇai kc , the following equal-time anti-commutation

relations emerge
0
D ˇa b ı 3 .x x 0 /; x0 D x 00
j ij
f a .x/;
i
b .x /g
j 0
f a .x/;
i
b .x /g D .ˇ ij 0 /a b ı 3 .x x 0 /; x 0 D x 0 0; (4.7.157)
generalizing that of the Dirac theory in (3.5.11) by taking into account of the con-
straint in (4.7.152). Since 0 D 0 i i this gives rise to the anti-commutations
relations of all the components .
Accordingly, as for the Dirac theory in (3.5.12), we have, from (4.7.154),
(4.7.155), (4.7.157), (4.7.145), the following general anti-commutation relations16
0 1 @
f a .x/; b .x /g D .x x 0 /; (4.7.158)
i ab i
@ @ @ @ 1 @ @ @
D C m C C C m
C :
i i m2 3 im i im
(4.7.159)
In the momentum description of the expression of the propagator in (4.7.145), let

p
PC . p/ D . p/; P . p/ D .p/; for p 0 D C p 2 C m2 :
(4.7.160)
We reconsider the polarization vectors introduced for spin 1 in (4.7.128), (4.7.129),
(4.7.130) and (4.7.131), and use the properties
p p X
X
C D e e D e e ; (4.7.161)
m2
D1;0;1 D1;0;1
16
Note that from (4.7.149)
h 1 i j @j @i 2 @ i @ j i
0
ˇ ij D 0 ı ij C C i j C 2
3 im im m i i
i j
@ @ 1 @ i @j
D 0 ı ij 2 C iC 0 jC :
m 3 im im
p X p
C D e e C . p C m/ ; (4.7.162)
m m2
D1;0;1
. p C m/ D 5 . p C m/ 5 ; (4.7.163)
. p C m/ X
D u.; p/ u.; p/; (4.7.164)
2m
with the last relation given in (I.21), to write

XX 1 XX
PC D e u./u./e e A.; /A.; 0 /e ;

3 0
;
(4.7.165)
suppressing the momentum variable for simplicity of writing, where

A.; / D e 5 u./; A.; / D u./ 5 e : (4.7.166)
The two terms in (4.7.165) may be combined to yield

X
PC D U ./U ./; (4.7.167)

where D 3=2; 1=2; 1=2; 3=2,

U .3=2/ D e C u.C1/; (4.7.168)
r
1 2
U .1=2/ D p e C u.1/ C e u.C1/; (4.7.169)
3 3 0
r
1 2
U .1=2/ D p e u.C1/ e u.1/; (4.7.170)
3 3 0
U .3=2/ D e u.1/; (4.7.171)
establishing the spin 3/2 character of the field. A similar analysis may be carried out
for P , with the Dirac spinors replaced by the v ones.
For a massless field .m D 0/, the Lagrangian density (4.7.146) becomes17
1 ! ! ! !
L D @ @ C @ C . @ / CK C K :
2i
(4.7.172)
17
The massless Rarita-Schwinger field treatment is based on Manoukian [7].
For K D 0; K D 0, the Lagrangian density, up to a total derivative, is invariant

under a gauge transformation

a .x/ ! a .x/ C @ a .x/: (4.7.173)
We may work in a Coulomb-like gauge defined by
@i a .x/
i
D 0: (4.7.174)
We may thus introduce fields U ib ; b , and set . r j @j /
ij j 1 i r @i
i
D ˛ ab U b b ; (4.7.175)
a
2 r2 ab
where
@ i @ j 1 i r @ i j r @ j
˛ ij D ı ij 2 C ; (4.7.176)
r 2 r2 r2
1 @i@s @`@j
D ı i s 2 ` s ı `j 2 ; (4.7.177)
2 r r
satisfying
@ i ˛ ij D 0; @j ˛ ij D 0; i ˛ ij D 0; ˛ ij j D 0; ˛ ij ˛ j k D ˛ i k : (4.7.178)
Also i defined in (4.7.175), satisfies the constraint in (4.7.174). One may now
vary U j ; ; 0 , and as seen below, this will lead not only to field equations but to
additional (derived) constraints.
To the above end, it is most convenient to express the field in the form
h 1 i r i i

D ı 0 0
C ı i ˛ ij U j @ ; (4.7.179)
2 r2
h j
0 1 i r @ i i
D 0 C U ˛
ji
i : (4.7.180)
2
r2
The variation of the field may be written in terms of the variation of

0 j
; U ; , by using in the process (4.7.180). This leads from the Lagrangian
density to . @ @ /
r 0
i i
D i K0 ; (4.7.181)
r2
0 1 @ 0 1 h@ i r ii 0
D K C Ki ; (4.7.182)
2r 2 i 2r 2 i i
@ ij j h@ i r ii 0
i
D ˛ K C K0 : (4.7.183)
i 2r 2 i i
0
Thus we derive, in particular, the constraints i i D 0, D 0, for K D 0.
Upon taking the matrix elements h 0C j : j 0 i of the expressions
in (4.7.182), (4.7.183), and using the fact that
•
h 0C j a .x/ j 0 i D .i/ h 0C j 0 i;
•K a .x/
we obtain by functional integration

Z

h 0C j 0 i D exp i .dx/.dx 0 /K a .x/DC ab .x x 0 /Kb .x 0 /; (4.7.184)

where the propagator DC ab .x x 0 / is given by
Z
.dp/ ip .xx 0 /
DC ab .x x 0 / D e DC ab . p/; (4.7.185)
.2/4
with . p D p , p D i pi /
. p/ ij 1 ik p i p k ` p k `j p ` p j
DC ij . p/ D ˛ . p/ D ı 2 ı 2 ;
p2 i
p2 i
p 2 p
(4.7.186)
1 ij p i p k ` k `j p ` p j
˛ ij . p/ D ı 2 ı 2 ; (4.7.187)
2 p p
1
DC 0 0 . p/ D . p/; (4.7.188)
2p 2
1 i
DC 0 i . p/ D 2
p C p i 0; (4.7.189)
2p
1 i
DC i 0 . p/ D p C p i 0: (4.7.190)
2p 2
! 0. Due to the presence of the denominator . p 2 i

/ in (4.7.186), only the
component DC ij . p/ has a propagation characteristic.

For K D 0; K D 0, we have, in particular, two constraints on the
spacial components i
given by i i D 0, and the adopted Coulomb-like
gauge @ i i
D 0. These constraints are taken care by the projection operator
˛ ij as is evident in (4.7.175), (4.7.176), (4.7.177) and (4.7.178). The equal-time
anti-commutation relations may be then expressed as
1 ij @ i @ k ` k `j @ ` @ j 3
f i
.x/; j
.x 0 /g D ı 2 ı 2 ı .x x 0 /; x 0 D x 0 0 :
2 r r
(4.7.191)
Consider the numerator of the propagator DC ij . p/ in (4.7.186), and set
p i p k ` p k `j p ` p j
ıi k ı 2 D PC ij . p/; p 0 D jp j: (4.7.192)
p2 2 p
By working in the chiral representation, the latter is explicitly given by (see

Problem 4.14)
X j
PC ij . p/ D U i ./ U ./; U i .˙3=2/ D ei˙ u.˙/: (4.7.193)
D˙3=2
4.7.5 Spin 2
We consider a symmetric Hermitian symmetric tensor field U D U , with m ¤

0, satisfying the equations
. C m2 / U .x/ D 0; @ U .x/ D 0; U .x/ U.x/ D 0:

(4.7.194)
We will show that this system describes a massive spin 2 particle. The above
equations may be derived from the single equation
. C m2 / U˛ˇ C @˛ @ U ˇ C @ˇ @ U ˛ @˛ @ˇ U

˛ˇ . C m2 / U C @ @ U D T˛ˇ ; (4.7.195)
when the external source T˛ˇ D Tˇ˛ is set equal to zero.

By comparing the equations obtained from the above equation, one by taking its
@˛ derivative, and the one obtained from multiplying it by ˛ˇ , we obtain
1 h1 ˛ 1 2 @ @ i
@ ˛ U˛ˇ D
˛ ˇ C
˛ ˇ @ @
ˇ @ˇ T ;
m2 2 3 3 m2
(4.7.196)
2 h1 ˛ @˛ @ˇ ˛ˇ i
UD T ˛ C T ; (4.7.197)
3 m2 2 m2
h1 1 1 i
. C m2 / U˛ˇ D ˛ ˇ C ˛ ˇ ˛ˇ T ; (4.7.198)
2 2 3
leading to the set of equations in (4.7.194) for T D 0, where

@ @
D 2 : (4.7.199)
m
The field equations in (4.7.195) may be obtained from the Lagrangian density
1
L D @ U @ U C @ U @ U @ U @ U
2
1 m2
C @ U @ U U U U U C T U : (4.7.200)
2 2
Upon writing the vacuum expectation values of (4.7.198) in the form
Z
I
h 0C j U .x/ j 0 i D .dx 0 / C

.x x 0 / T .x 0 /h 0C j 0 i; (4.7.201)
I
the propagator C .x x 0 / is readily extracted to be
Z
I .dp/ ip .xx 0 / I
C .x x 0 / D e C . p/; (4.7.202)
.2/4
where
1
I . p/ . p/ C 12 . p/ . p/ 13 . p/ . p/
2
C . p/ D ;
p 2 C m2 i
(4.7.203)

p p
. p/ D C 2 ; (4.7.204)
m
and the vacuum-to-vacuum transition amplitude takes the usual form
hiZ i
h 0C j 0 i D exp .dx/.dx 0 /T ˛ˇ .x/C˛ˇI .x x 0 /T .x 0 / : (4.7.205)
2
Consider the numerator of the propagator C˛ˇI . p/ in (4.7.203)/(4.7.205),
and set
1 ˛ 1 1
P ˛ˇI . p/ D . p/ˇ . p/ C ˛ . p/ˇ . p/ ˛ˇ . p/ . p/;
2 2 3
(4.7.206)
p
for a particle on the mass shell p0 D p2 C m2 . Using the completeness relation
in (4.7.129), the above expression may be rewritten as
Xh 1 ˇ ˇ
1 ˛ ˇ i
P ˛ˇI D e ˛ e e 0 e 0 C e ˛ e e 0 e 0 e e e 0 e 0 :
2 3
; 0
(4.7.207)

Finally the conditions e ˙ D e
; e 0 D e 0 given in (4.7.131), give rise to
the decomposition
X
P ˛ˇI D u˛ˇ ./ u ./; D 2; 1; 0; 1; 2; (4.7.208)

ˇ
u˛ˇ .2/ D e C
˛
e C; u˛ˇ .2/ D e
˛ ˇ
e ; (4.7.209)
1 1
ˇ ˇ ˇ
u˛ˇ .1/ D p e ˛0 e C C e ˛C e 0 ; u˛ˇ .1/ D p e ˛0 e ˇ
C e ˛ e 0 ;
2 2
(4.7.210)
r
˛ˇ 2 ˛ ˇ 1 ˛ ˇ 1 ˛ ˇ
u .0/ D e e p eC e p e e C; (4.7.211)
3 0 0 6 6
describing the 5 spin states of a massive spin 2 particle.

In the massless case, the Lagrangian density in (4.7.200) becomes
1 1
L D @ U @ U C @ U @ U @ U @ U C @ U @ U C T U :
2 2
(4.7.212)
For T D 0, the Lagrangian density is, up to a total derivative, invariant under the
gauge transformation
U ! U C @ C @ ; (4.7.213)
for arbitrary . A covariant gauge treatment of the massless spin 2 dealing with
the graviton will be carried out in Sect. 2.3.1 of Volume II. On the other hand, in a
Coulomb-like gauge we impose the constraint
@ i U i D 0; D 0; 1; 2; 3: (4.7.214)
The analysis becomes simplified by expressing the field in the following manner

U D i j U ij 0 i C i 0 U 0 i C 0 0 U 00 ; (4.7.215)
and set
1 ik j`
U ij D C i ` j k /H k ` ; U 0 i D ij ' j ; (4.7.216)
2
defined in terms of new unconstrained fields H k ` , ' j , and

@i@j
ij D ı ij 2 ; ij j k D i k; ; @ i ij D 0: (4.7.217)
r
The gauge condition (4.7.214) is automatically satisfied.
With the field components expressed as above, variations of the Lagrangian
density in (4.7.212) with respect to ' i , U 00 , H ij lead upon defining
1 ik j `
ij k ` D C i ` j k ij k` ; (4.7.218)
2
to the following equations
1 00
Ui i D T ; (4.7.219)
r2
1 h 00 i
U 00 D T C ij j
T ; (4.7.220)
2 r2 r2
1
U 0 i D 2 ik T 0k ; (4.7.221)
r
1
U ij D ij k` T k` C ij 2 T 00 : (4.7.222)
2 r
Thus we have derived additional constraints: U i i D 0, U 00 D 0, U 0 i D 0, for

T D 0.
Upon writing the vacuum expectation values of (4.7.220), (4.7.221) and 4.7.222)
in the form
Z
I
h 0C j U .x/ j 0 i D .dx 0 / DC .x x 0 / T .x 0 /h 0C j 0 i; (4.7.223)
I
the propagator DC .x x 0 / is readily extracted to be
Z
I .dp/ ip .xx 0 / I
DC .x x 0 / D e DC . p/; (4.7.224)
.2/4
I
where C . p/ is symmetric in .; /, and .; /, as well as in the interchange
.; / $ .; /, and
ij I k` .p/ p ip j
ij Ik `
D C . p/ D ;
! 0; ij
.p/ D ı ij
; (4.7.225)
p2 i
p2
1
ik
ij Ik` .p/ D .p/ j ` .p/ C i ` .p/ j k .p/ ij .p/ k ` .p/ ; (4.7.226)
2
ij I00 1 00I00 1 p2
D C . p/ D ij .p/; DC . p/ D ; (4.7.227)
2 p2 2 p2 p2
0i I 0k 1 00 I 0 i 0 i Ij k
DC . p/ D i k .p/; DC . p/ D 0; DC . p/ D 0: (4.7.228)
2 p2
The vacuum-to-vacuum transition amplitude has the usual form

hiZ i
I
h 0C j 0 i D exp .dx/.dx 0 /T .x/DC .x x 0 /T .x 0 / ; (4.7.229)
2
Consider the numerator ij I k` .p/ of the propagating part of the propagator

in (4.7.225) for p0 D jp j. Using the polarization 3-vectors e˙ in (4.7.130), with
X X
e e0 D ı0 ;
j j
p e˙ D 0; ij D ei e D e i e ; (4.7.230)
D˙ D˙
with the last relation defining a completeness relation in 3D, we may rewrite
1 X i k j `
ij I k ` .p/ D e e e 0 e0 C ei e` i j k `
j k
e 0 e0 e e e0 e0 : (4.7.231)
2 0 ;
Using the following properties e˙ D e

, given in the first relations in (4.7.131),
the above expression simplifies to
X ij
ij I k` .p/ D " " k ` ;
ij j
" ˙ 2 D e i˙ e ˙ ; (4.7.232)
D ˙ 2
describing the two helicity states of a massless spin 2 particle. The latter is identified
with the graviton.
For a conserved source @ T .x/ D 0, i.e., in the momentum description
p0 T 0 . p/ D p i T i . p/; (4.7.233)
the expression for h 0C j 0 i, simplifies to some extent and leads to a Lorentz

invariant expression for it. To this end the integrand multiplying the i factor, in the
exponent in (4.7.172) including the 1/2 factor, may be rewritten as
1 h ij ij I k` k` 0i I 0k 00 I 00 ij I 00
i
T D C T C 4 T0 i D C T0 k C T00 D C T00 C 2 Tij D C T00 :
2
(4.7.234)
In the momentum description, we have
ij k ` 1 1
ij I k` k`
T ij D C T k` D T ij T D T T T T 2
p2 i
2 p i
p 2 1
C T i i T 00 T 00 T 00 C T 00 2
T 00 C 2 T 0 i T 0 i ; (4.7.235)
2p p2
i k 0 k
4 T0 i D0Ci I 0 k T0 k D 4 T 0 i T
2p2
p2 1
D 2 T 0 k T 0 k 2 T 00 2 T 00 C 2 T 00 T 00 2 ; (4.7.236)
p p
00 I 00 p 2 00
T00 DC T00 D T 00 T ; (4.7.237)
2p4
ij ij i i 00 00 p
2
00 00 1
T00 D 2 T 00 00
ijI00
2 Tij DC T D T T C T T T T :
2p2 p2 p2
(4.7.238)
Hence from (4.7.234), (4.7.235), (4.7.236), (4.7.237), (4.7.238), and (4.7.229),

hiZ
1 i
h 0C j 0 i D exp .dx/.dx 0 / T .x/T .x 0 / T .x/ T .x 0 / DC .x x 0 / ;
2 2
(4.7.239)
Z ip .xx 0 /
.dp/ e
DC .x x 0 / D : (4.7.240)
.2/4 p 2 i
4.8 Further Illustrations and Applications of the QDP
Consider the simple Lagrangian density involving a neutral scalar field
1 m2 2
L .x; / D @ .x/ @ .x/ .x/ C 4 .x/ C K.x/ .x/; (4.8.1)
2 2 4
where K.x/ is an external source, is a coupling parameter, with the interaction

Lagrangian defined by the term
4
LI .x; / D .x/: (4.8.2)
4
Clearly, the canonical conjugate momentum .x/ of the field .x/ is given by
@0 .x/, and the field equation is obtained by taking the functional derivative of the
action with respect to .x/, as given in (4.3.36), and using the simple rules (4.3.33),
(4.3.34) and (4.3.35),
. C m2 / .x/ 3 .x/ D K.x/: (4.8.3)
The QDP in (4.6.37) for the derivative of the vacuum-to-vacuum transition

amplitude h 0C j 0 i, as given in the Summary at the end of Sect. 4.6., with respect
4.8 Further Illustrations and Applications of the QDP 205
to , gives
Z
@ 1
h 0C j 0 i D i h 0C j .dx/ 4 .x/ j 0 i: (4.8.4)
@ 4
Since .x/ is an independent field, repeated applications of (4.6.38) gives

• 4
h 0C j 4 .x/ j 0 i D i h 0C j 0 i; (4.8.5)
•K.x/
in a limiting sense of a time-ordered product whose nature will be discussed in the

next section. Here we have used the fact that
Z
•
.dx 0 / L .x 0 ; / D .x/: (4.8.6)
•K.x/
Upon setting
4
b I .x; / D i •
L ; (4.8.7)
4 •K.x/
we may use (4.8.4) and (4.8.5) to integrate the former equation over from 0 to a
specific value of to obtain
Z

b I .x; / h 0C j 0 iD0 ;
h 0C j 0 i D exp i .dx/ L (4.8.8)
where h 0C j 0 iD0 corresponds to h 0C j 0 i D h 0C j 0 i with set equal

to zero. The expression for h 0C j 0 iD0 for the scalar field is given in (4.7.5).
For the purpose of an illustration of a more involved case with a dependent field,
consider the Lagrangian density

L .x; / D L0 .V .x// C L0 . .x/; .x// .x/ .x/V .x/V .x/
2
C K .x/V .x/ C .x/ .x/ C .x/.x/; (4.8.9)
where L0 .V .x// is the Lagrangian density of a neutral vector field V .x/ of mass
>0
1 m2
L0 .V .x// D @ V @ V @ V @ V V V ; (4.8.10)
4 2
(see (4.7.17)), and L0 . .x/; .x// is the Lagrangian density of the Dirac field
in (4.7.11). Also K .x/, .x/, .x/ are external sources coupled to the correspond-
ing fields, with the latter two as Grassmann variables.
The interaction Lagrangian density of the coupling of the vector and the Dirac
field is given by

LI .x; / D .x/ .x/V .x/V .x/; (4.8.11)
2
and > 0 is a coupling parameter. To be more precise, we have to replace

1
a .x/ a .x/, for example, by an antisymmetric average 2. a .x/ a .x/ a .x/ a .x//
and such points will be elaborated upon in the next section.
We note that the Lagrangian density L0 .V .x// does not involves the time
derivative @ 0 V 0 of the field component V 0 of V , hence the canonical conjugate
momentum of V 0 is zero, and V 0 is a dependent field.
The field equation for the vector field V as obtained from (4.3.36), and from
the rules (4.3.33)–(4.3.35), is

@ @ V .x/ @ V .x/ C m2 C .x/ .x/ V .x/ D K .x/; (4.8.12)
and the canonical conjugate momenta k of the components V k , k D 1; 2; 3; are

given by
@L .x; /
k .x/ D D .@0 V k @k V 0 /: (4.8.13)
@.@0 V k /
Upon taking the derivative @ k of k , we note that

r 2 V 0 .x/ @0 .r V.x// D r .x/: (4.8.14)
On the other hand, picking up the D 0 – component of (4.8.12) gives

r 2 V 0 .x/ @0 .r V.x// C m2 C .x/ .x/ V 0 .x/ D K 0 .x/; (4.8.15)
where we have used the definition D r 2 @ 0 2 .

The last two equations then lead the following equation of constraint for V 0
2
m C .x/ .x/ V 0 .x/ D K 0 .x/ C r .x/: (4.8.16)
The QDP in (4.6.37) gives

Z
@ @
h 0C j 0 i D i h 0C j .dx/ L .x; / j 0 i
@ @
Z
i
D h 0C j .dx/ .x/ .x/V .x/V .x/ h 0 i: (4.8.17)
2
4.8 Further Illustrations and Applications of the QDP 207
The field V 0 appearing in LI .x; / in (4.8.11) is dependent. Since we will vary

the components of K .x/ independently no constraints, such as @ K .x/ D 0, may
be imposed on K .x/.
We use the properties
Z
•
.i/ .dx 0 / L .x 0 ; / D i .x/; (4.8.18)
•.x/
Z
•
.Ci/ .dx 0 / L .x 0 ; / D i .x/; (4.8.19)
•.x/
Z
•
.i/ .dx 0 / L .x 0 ; / D i V .x/; (4.8.20)
•K .x/
where we recall that the Grassmann variables , anti-commute with the Dirac
fields. Accordingly, by repeated application of (4.6.38), dealing with independent
fields, gives
• • • •
.i/ .i/ .i/ .i/ k h 0C j 0 i
•.x/ •.x/ •Kk .x/ •K .x/

D h 0C j .x/ .x/V k .x/Vk .x/ j 0 i; (4.8.21)
for k D 1; 2; 3, as a limiting case of a time ordered product whose nature again will
be elaborated upon in the next section.
On the other hand, an application of (4.6.38) for the dependent field V 0 .x/ gives
• • •
.i/ 0
.i/ 0 h 0C j 0 i D .i/ h 0C jV0 .x/ j 0 i
•K0 .x / •K .x/ •K0 .x 0 /
•
D h 0C j V 0 .x 0 /V0 .x/ C j 0 i C . i/ h 0C j V0 .x/ j 0 i;
•K0 .x 0 /
(4.8.22)
where .•=•K0 .x 0 //V0 .x/ is to be obtained from the constraint equation (4.8.16) by
keeping the fields .x/, .x/, and @k k .x/ fixed. To this end, we introduce the
following simplifying notations:
•
.i/ b
V ; (4.8.23)
•K .x/
• •
.i/ .i/ b
S.x/: (4.8.24)
•.x/ •.x/
From (4.8.16), we then have
• 1
h 0C j V0 .x/ j 0 i D h 0C j 0 i •.4/ .x x 0 /: (4.8.25)
•K0 .x 0 / m2 C b
S.x/
Upon carrying out the functional differentiations in b

S.x/, as applied to (4.8.22),
and using the above equation, we obtain

h 0C j .x/ .x/V 0 .x 0 /V0 .x/ j 0 i
C
•.4/ .x x 0 /
Db
S.x/ bV 0 .x 0 /b
V 0 .x/ C i h 0C j 0 i; (4.8.26)
m2 C b S.x/
where note that b S.x/, b

S.x 0 / commute, with each taken as the product defined
in (4.8.24).
Finally we set
b I .x; / D b
L V .x/b
S.x/b V .x/; (4.8.27)
2
and use (4.8.21) and (4.8.26) to rewrite (4.8.17) as
Z hZ
@ @ @ b
h 0C j 0 i D i h 0C j .dx/ LI .x/ j 0 i D i .dx/ L I .x; /
@ @ @
Z Z b i
i S.x/
.dx/ .dx 0 / •.4/ .x 0 x/ •.4/ .x x 0 / h 0C j 0 i:
2 Œ m2 C b
S.x/
(4.8.28)
By an integration over from 0 to a specific value for , the following

expression emerges for the vacuum-to-vacuum transition amplitude
Z
b I .x; / F b
h 0C j 0 i D exp i .dx/ L S.x/ h 0C j 0 iD0 ; (4.8.29)
where
1 Z Z

b
F S.x/ D exp .dx / .dx/ •.4/ .x 0 x/•.4/ .xx 0 / ln 1C 2 b
0
S.x/ ; (4.8.30)
2 m
b
S.x/ is defined in (4.8.24), and we have used the elementary integral
Z h

a d0 ai
D ln 1 C : (4.8.31)
0 Œb C a 0 b
The expression h 0C j 0 iD0 is the product of the corresponding one for the Dirac
field as was already derived in (3.3.28), and given (4.7.16), and of the corresponding
one for the vector field given in (4.7.22).
Upon comparison of the solution in (4.8.29) for the second example with the one
R
in (4.8.8) for the first example, we learn the following. The integral .dx/L b I .x; /
4.9 Time-Ordered Products, How to Write Down Lagrangians and Setting Up. . . 209
for the second example is modified due to the F factor in (4.8.30), and this is as a
consequence of the presence of the dependent field V 0 .x/. This means, in particular,
that the action integral gets modified. Such a property is shared by non-abelian
gauge theories in most gauges. We note that the double integral in (4.8.30) produces
a •.4/ .0/ factor due to the product of delta functions at coincident spacetime points.
Finally, we note that the presence of a dependent field is not necessary, in every
case, for the presence of a modifying factor F as encountered above beyond the
R
term exp.i .dx/ L b I .x; //. For example, if we replace the interaction Lagrangian
density in (4.8.11), say, by . /ab Œ a .x/; b .x/V =2, which is linear in the
vector field V , no such a modification arises (see Problem 4.16).
4.9 Time-Ordered Products, How to Write Down

Lagrangians and Setting Up the Solution
of Field Theory
This section deals with general aspects of the structure of Lagrangian densities in
quantum field theory and on the way one may then proceed to set up the solution of
an underlying theory. The way of actually constructing such Lagrangian densities
to describe the fundamental interactions of nature will be carried out in remaining
chapters. Before doing so, we recapitulate some aspects of time-ordered products.
To this end, recall that one may add coupling terms, to a given Lagrangian density
L .x/, consisting of the interaction of the underlying fields in a theory, say, of a
scalar field .x/ and Dirac fields .x/, .x/ with external sources
L .x/ ! L .x/ C K.x/.x/ C .x/ .x/ C .x/.x/; (4.9.1)
to generate matrix elements of the fields
•
.i/ h 0C j 0 i D h 0C j .x/ j 0 i; (4.9.2)
•.x/
•
.Ci/ h 0C j 0 i D h 0C j .x/ j 0 i; (4.9.3)
•.x/
•
.i/ h 0C j 0 i D h 0C j.x/ j 0 i: (4.9.4)
•K.x/
As a consequence of the Bose-Einstein character of a Lagrangian density, being

related to the Hamiltonian density, the Dirac field sources .x/, .x 0 / anti-commute
as well as the functional derivatives with respect to them, i.e.,
n • • o
; D 0: (4.9.5)
•.x 0 / •.x/
Hence from
• • 0

.Ci/ .i/ h 0C j 0 i D h 0C j a .x / b .x/ C j 0 i; (4.9.6)
•a .x 0 // •b .x/
• • 0

.i/ .Ci/ 0
h 0C j 0 i D h 0C j b .x/ a .x / C
j 0 i; (4.9.7)
•b .x/ •a .x /
we have the rule that within a time-ordered product, the Dirac quantum fields anti-
commute, i.e.,
0
0

a .x / b .x/ C D b .x/ a .x / C
: (4.9.8)
Similarly, one has

.x 0 /.x/ C D .x/.x 0 / C ; (4.9.9)

.x 0 / b .x/ C D 0
b .x/.x / C : (4.9.10)
The general rule is that within a time-ordered product, as far as their commu-
tativity is concerned, two fields may be commuted if at least one of them is a
Bose-Einstein field or may be anti-commuted if they are both Fermi-Dirac fields.
In particular, the following time-ordered products take the form
0
0
b .x/ a .x / C
D b .x/ a .x /.x 0 x 0 0 / a .x
0
/ b .x/.x
00
x 0 /;
(4.9.11)

.x/.x 0 / C
D .x/.x 0 /.x 0 x 0 0 / C .x 0 /.x/.x 0 0 x 0 /:
(4.9.12)
How to Write Down Lagrangian Densities:

1. Lorentz Invariance and Other Symmetries: In quantum field theory, a Lagrangian
density L .x/, not including couplings of the fields to external sources, is, a
priori, chosen to be Lorentz invariant, that is, it behaves as a Lorentz scalar:
L 0 .x 0 / D L .x/ under a Lorentz transformation x ! x 0 D x .
This, however, does not mean that given such a Lagrangian density, which is
formally a Lorentz scalar, one cannot subsequently use, for example, as may
be done in QED or in QCD, non-covariant gauges such as the Coulomb one
r A D 0, or r Aa D 0, respectively. The importance of using, for
example, the Coulomb gauge in such gauge theories is that only the physical
degrees of freedom associated with the vector fields are quantized ensuring the
necessary positive definiteness of the underlying physical Hilbert space. One
may then invoke the gauge invariance of transition probabilities associated with
the underlying physical processes to infer that the results should be the same
as the ones that may be obtained by using covariant gauges. By doing so,
4.9 Time-Ordered Products, How to Write Down Lagrangians and Setting Up. . . 211
one would finally recover a gauge and Lorentz invariant theoretical framework
for the computation of physical results. Accordingly, it is misleading to state
that a non-covariant gauge destroys the Lorentz invariance of a theory. In the
same manner, when one chooses a specific gauge, such as a covariant gauge,
the Lagrangian density becomes restricted to such a gauge and hence gauge
invariance is destroyed at the Lagrangian level, but the theory dealing with
physical processes is nevertheless gauge invariant. Symmetries relevant to the
underlying theory may also be implemented, as we will see, for example, when
supersymmetric field theories will be developed later. Symmetry breaking will
be also discussed.
2. Even in Fermi-Dirac Fields. The Bose-Einstein character of the Lagrangian
density implies that it must be even in the Fermi-Dirac fields constructed in such
a manner so that this aspect is maintained.
3. Limits of Time-Ordered Products. We have seen in the previous section that the
matrix element h 0C jLI .x/ j 0 i is generated by functional differentiations with
external sources and leads an expression for LI .x/ always as a limit of time-
ordered products. That is, the QDP self consistently requires that Lagrangian
densities be defined as the limits of time-ordered products. We have also seen
in (4.9.8), (4.9.9) and (4.9.10) that within a time-ordered product the fields
obey commutativity or anti-commutativity properties as spelled out in the just
mentioned equations. In reference to (4.9.11), for example, we consider the
average of the limits x ! x 0 C 0, x ! x 0 0, with initial spacelike
separation .x x 0 /2 > 0, and similarly for the scalar field in (4.9.12). This,
in particular, means that in writing down a Lagrangian density one is to take
the anti-symmetric average of the product of two Fermi-Dirac fields and the
symmetric average of the product of two Bose-Einstein fields. Hence, one makes
the following replacement for the product of two Dirac fields
1 1
a .x/ b .x/ ! a .x/ b .x/ b .x/ a .x/ D Œ a .x/; b .x/ ;
2 2
(4.9.13)
and for the interaction Lagrangian density part, for example, in QED,
1
eo a .x/. /ab b .x/ A .x/ ! f j .x/; A .x/g; (4.9.14)
2
where the electromagnetic current, having a Bose-Einstein character, is given by
1
j .x/ D eo Œ a .x/; b .x/ ; (4.9.15)
2 ab
and eo is the (unrenormalized) charge. We have already seen the definition
of the electromagnetic current defined as in (4.9.15) consistent with charge
conjugation transformation in (3.6.9) in Sect. 3.6. It is important to note, as
already emphasized in Sect. 3.6, that the functional derivative •=• c .x 0 /, for
example, should anti-commute with b .x/ giving from (4.9.13)
• 1 .4/
a .x/ b .x/ b .x/ a .x/ D •ac b .x/ • .x x 0 /; (4.9.16)
• c .x
0/ 2
otherwise one would get zero as a trivial result. The above analysis leads us to
emphasize the following:
“Students often wonder why one can commute (anti-commute) field compo-
nents in the interaction Lagrangian density for a theory dealing with operators.
With the interaction Lagrangian density defined as a limit of a time-ordered
product this becomes evident”.
4. Renormalizability. The Lagrangian density is so chosen so the theory becomes
renormalizable.
Setting Up the Solution of Field Theory:
The importance of deriving an expression for the vacuum-to-vacuum transition
amplitude h 0C j 0 i in field theory stems from the fact that one may carry out
a unitarity expansion for the latter in terms of multi-particle states from which
transition amplitudes for the underlying physical processes of a theory may be
readily extracted.18 At this stage the reader is urged to review the two examples
provided in the last section on the application of the QDP before considering the
general strategy for setting up the solution for h 0C j 0 i given below.
Let .x/ denote collectively the independent fields in a theory with canonical
conjugate momenta denoted collectively by .x/, and let %.x/ denote collec-
tively the dependent fields, i.e., with vanishing canonical conjugate momenta. The
Lagrangian density will be expressed as
L .x; / D L0 .x/ C LI .x; / C K.x/.x/ C S.x/%.x/; (4.9.17)
where
LI .x; 0/ D 0; (4.9.18)
and K.x/, S.x/ denote external sources coupled to these fields, respectively. The
dependence of L .x; / on the adjoints of the fields is suppressed for simplicity of
the notation.
A key equation for obtaining an expression for h 0C j 0 i is (see (4.6.37))
Z
@ @
h 0C j 0 i D i h 0C j .dx/ LI .x; / j 0 i: (4.9.19)
@ @
18
See, e.g., (5.8.20), (5.8.21), (5.9.16), (5.9.17), (5.9.18), (5.9.19), (5.9.20) and (5.9.21).
4.10 CPT 213
Now the basic idea is to re-express h 0C j .@=@/LI .x; / j 0 i as an operation

involving the functional derivatives .•=•K.x//, .•=•S.x// applied directly to
h 0C j 0 i.
Suppose:
(i) The Euler-Lagrange equations allow one to eliminate the dependent fields %.x/
in favor of .x/, .x/, , K.x/, S.x/, via equations of constraints, and, in
turn, allow one to rewrite LI .x; / as a sum of terms depending on the latter
variables, with coefficients depending on , and
(ii) Repeated applications of (4.6.38), as described below it, and as also done in the
second example of the last section, the above allows one, in turn, to rewrite
@
@
h 0C j LI .x; / j 0 i D b I .x; / C F
L b .x; / h 0C j 0 i;
@ @
(4.9.20)
where L b I .x; / denotes LI .x; /, with the fields (.x/, %.x/), simply replaced
by .˙i/ times (.•=•K.x//, .•=•S.x//), as the cases may be, (see, e.g., (4.9.2),
(4.9.3) and (4.9.4)), and Fb .x; / is an operation expressed in terms of these
functional derivatives and may depend on .
Upon integrating (4.9.20) over , the following expression for h 0C j 0 i D
h 0C j 0 i emerges
Z
Z

h 0C j 0 i D exp i .dx/ Lb I .x; / C b .x; 0 / h0C j 0 iD0 :
d0 F
0
(4.9.21)
b I .x; /, and hence also the action,
Here we see that, in general, the interaction term L
get modified due to the presence of the second term in the exponent. The amplitude
h 0C j 0 iD0 has usually a simple structure. The full amplitude h 0C j 0 i is
then obtained merely by functional differentiations of h 0C j 0 iD0 .
Applications of the above formula were given in the last section and other
applications will be given in coming chapters.
4.10 CPT
We recall (Sect. 2.1) that charge conjugation transformation C consists in replacing

every particle in a given process by its anti-particle, while parity P reverses the
direction of the three-momentum of every particle, and time T reversal involves
in reversing the direction in which a process evolves and as such, the particles
initially going into it, become outgoing, and vice versa, while reversing the direction
of their three-momenta and their spin projections. QFT predicts that the product
of the three transformations CPT (in any order), taken together, is a symmetry
of nature in the sense that any transition probability and its CPT counterpart are
equal, and no experiment seems to contradict this. Clearly, the product CPT, as a
net transformation, amounts to replacing every particle by its anti-particle, while
reversing their spin projections, and interchanging the initial and final states. Recall
also, due to the anti-unitary property of time reversal operation (Sect. 2.1), the CPT
transformation is implemented by an anti-unitary operator.
It should be noted that establishing the CPT invariance of the action corre-
sponding to a local Lagrangian density is only part of the analysis of the CPT
Theorem, one also has to show how the interchange of the initial and the final
states arises with the simultaneous replacements of particles by their antiparticles
and vice versa. In this respect, we will see that in the present analysis, detectors
of particles/antiparticles and emitters of antiparticles/particles are interchanged in a
CPT “transformed” world. The latter property will automatically take care of the
arbitrary the number of particles may be going in and those that may come out of a
scattering process in a direct and simple manner.
Let us first consider how basic fields transform under the above transformation.
To this end, note, in particular, that from (3.2.29), (3.2.31), (3.2.36), that for a Dirac
field:
>
C .x/ C1 D C .x/; P .x/ P1 D 0 .x 0 /; T .x/ T1 D 5 C .x 00 /;
(4.10.1)
with given fixed phases, where x 0 D .x 0 ; x/, x 00 D .x 0 ; x/, C D i 2 0 .

With a rather conventional choices of phases, we may thus introduce D CPT
transformations, and quite generally, of the following basic fields, consisting of
a Dirac field , a generic scalar field ', and a generic vector field V with a
direct generalization for the presence of several such fields for different species of
particles:
>
.x/ 1 D 5 0 .x/; .x/ 1 D >
.x/ 0 5 ;
(4.10.2)
'.x/ 1 D ' .x/; ' .x/ 1 D '.x/; (4.10.3)
V .x/ 1 D V .x/; V .x/ 1 D V .x/: (4.10.4)
Opposite phases of the scalar and vector fields are taken since @ '.x/ should
transform as a vector field.
Given a local Lagrangian density Lagrangian density L .x/, R it is not sufficient
to establish the symmetry of the action integral A D .dx/ L .x/ under a
CPT transformation. One also has to describe how incoming states and outgoing
states, for arbitrary processes, are interchanged and what happens to the spins of
participating particles. The latter is best described by coupling the underlying fields
to corresponding external sources to be able to identify all the particles that may be
ingoing or outgoing in any process in a straightforward manner. As we will see, the
latter avoids of introducing many-particle states for all possible processes that may
4.10 CPT 215
take, and, in turn, leads to examine as to what effectively happens to the external
source functions in a CPT “transformed” world.
To the above end, we consider a total action integral
Z
Atot D .dx/ L .x/ C LS .x/ ; (4.10.5)
where L .x/ a local Lagrangian density, and without loss of generality, restricting
the analysis to the fields in (4.10.2), (4.10.3) and (4.10.4), with the labeling of
different species suppressed,

LS .x/ D .x/ .x/ C .x/.x/ C K .x/.x/

C .x/K.x/ C K .x/V .x/ C V .x/K .x/ ; (4.10.6)
where K .x/; K .x/, K.x/; K .x/, .x/; .x/ denote external sources with the
latter two being Grassmann variables.
In the appendix to this chapter it is shown
Z Z
1
.dx/ L. x/ D .dx/ L .x/; (4.10.7)
Z
.dx/ Œ .x/ .x/ C .x/ .x/ 1
Z
D .dx/ Œ .x/ 5 .x/ .x/ 5 .x/; (4.10.8)
Z
.dx/ Œ' .x/ K.x/ C K .x/'.x/ 1
Z
D .dx/ Œ ' .x/ K.x/ C K .x/'.x/: (4.10.9)
Z
.dx/ ŒV .x/ K .x/ C K .x/V .x/ 1
Z
D .dx/ ŒV .x/ K .x/ K .x/V .x/: (4.10.10)
Therefore, only the external sources change which, after all, are responsible of
emitting and detecting all the particles in any given process. That is, CPT is a
symmetry of the action of the local Lagrangian density L .x/ in quantum field
theory
Z
A D .dx/ L .x/; (4.10.11)
while the sources “change their roles” in the following manner :
.x/ ! 5 .x/; .x/ ! .x/ 5 ; (4.10.12)

K.x/ ! K.x/; K .x/ ! K .x/; (4.10.13)
K .x/ ! K .x/; K .x/ ! K .x/: (4.10.14)
signalling the fact of changes have occurred with the ingoing and outgoing parti-
cles/antiparticles. More precisely, the second substitution in (4.10.12), for example,
implies that
. p/u.p; / ! .p/ 5 u.p; / D .p/v.p; /; (4.10.15)
and so on, where we have used the definition of v.p; / in (I.17). Referring
to (3.3.39), (3.3.42), we may infer that an outgoing particle of spin has been
replaced by an ingoing antiparticle of spin . Upon introducing polarization
vectors, as discussed in (4.7.130), (4.7.230), satisfying the properties in (4.7.131),
for massive and massless vector bosons, we may then prepare Table 4.1 which
shows the expected changes that occur with ingoing and outgoing states in the CPT
“transformed” world.
The CPT-symmetric nature of the action of local field theory and the fate of
the particles in a CPT “transformed” world in any given process, as described in
Table 4.1, is the content of the celebrated CPT Theorem.
Remarks
1. It should be emphasized that the theoretical analysis of the nature of particles
ingoing and outgoing in a physical process, as done above, under the CPT
transformation, and not just of establishing the transformation property of the
action of the local Lagrangian density in (4.10.11) in the absence of external
Table 4.1 Source functions and corresponding particles/anti-particles in physical processes

Type (IN) CPT (OUT)
Dirac Particle: u.p; /. p/ Anti-particle: v.p; /.p/
Anti-particle: .p/v.p; / Particle: . p/u.p; /
Spin-0 Particle: K. p/ Anti-particle: K.p/
Anti-particle: K .p/ Particle: K . p/

Vector Particle: e . p/K . p/ Anti-particle: e . p/K .p/

Anti-particle: e . p/K .p/ Particle: e . p/K . p/
Appendix A: Basic Equalities Involving the CPT Operator 217
sources, is of great importance. Establishing the invariance property in (4.10.7)

is only part of the story.
2. The introduction of the source functions simplifies the proof of the Theorem
tremendously as they tell us what happens to all possible ingoing and outgoing
particles/anti-particles in processes in a CPT “transformed” world. In particular,
we saw how naturally particle/antiparticle detectors and antiparticle/particle
emitters are interchanged in a CPT “transformed” world. They also provide
much physical insight on the preparatory and final stages of physical processes
in a theory.
3. From the way QFT has been developed over the years, the validity of the CPT
Theorem is of no surprise, and its theoretical development historically was
expected.
4. Although experiments do not contradict CPT invariance, they tell us, however,
that one or two of C, P, T taken at a time are violated. The violation of CP has
been particularly intriguing and, as pointed out by Sakharov [9] almost half a
century ago, may give us a clue on the reason for the imbalance of matter and
anti-matter with nature’s preference of matter over anti-matter which, together
with the Spin & Statistics Connection (Sect. 4.5), are contributing factors to our
existence.
5. Note that the argument of the source functions in the second column with oppo-
site sign of momentum to the right column, imply, through Fourier transform, that
the corresponding particle/antiparticle, going into (loss) or coming out (gain) of
the scattering process, as the case may be, carries the same energy-momentum
before CPT transformation, but spin and helicity state change as seen in
the Dirac spinors and polarization vectors, respectively.
Appendix A: Basic Equalities Involving the CPT Operator
We have the following explicit equalities for the CPT operator :

h i
.x/ .x/ C .x/ .x/ 1
h i
D . 0 /ab . 5 0 /bc c .x/ a .x/ C a .x/ b .x/. 0 5 /ba
h i
D . 5 /ac c .x/ a .x/ a .x/ b .x/. 5 0 /ba
h i
D .x/ 5 .x/ .x/ 5 .x/ : (A-4.1)
Equation (4.10.8) now follows upon changing the variable of integration x ! x.

Equations (4.10.9), (4.10.10) are self evident.
Finally, to establish (4.10.7), we consider transformation properties of the
product of fields and their derivatives that may appear in L .x/. Consider first Dirac
fields, and a general linear combination of, say, two Dirac fields
1 1 ::: k 1 2
M Œ a .x/; b .x/; (A-4.2)
2 ab
:::
where Mab1 k is of the general form, or a linear combination of such tensors as,
: : : k , 1 : : : k 5 . Partial derivatives will be considered in studying the
1
general structure of tensor components in (A-4.5) below. In view of the application

of a CPT transformation, note that
Œ . 5 0 / 1 : : : k . 0 5 / ab D .1/k Œ 1> : : : k> ab D .1/k Œ k : : : 1 ba ;

(A-4.3)
and similarly in the presence of an additional 5 matrix, where we have used the
properties: 5 D 5 , . 0 /2 D . 5 /2 D 1, 0 D > 0 , 0> D 0 ,
5> D 5 . Hence from the transformation rule of a spinor given in (4.10.2), we
obtain
1 ˇ 1
::: k 1 2 ˇ ::: k 1
Mab1 Œ a .x/; b .x/ 1 ˇ D .1/k Mab1 Œ a .x/;
2
b .x/ :
2 x!x 2
(A-4.4)
Note the importance of using the anti-symmetric average (see Sect. 4.9) in estab-
lishing this result.
On the other hand for any tensor component 1 ::: k .x/, such as constructed out
of products of vector fields, of scalar fields and their derivatives, we obviously have
under the CPT transformations (4.10.3) and (4.10.4),
1 ::: k .x/ 1 D .1/k 1 ::: k .x/: (A-4.5)
Upon defining
::: k 1 1 2
Mab1 Œ a .x/; b .x/ D ˝ 1 ::: k .x/; (A-4.6)
2
we may infer from (A-4.4) and (A-4.5) that
1 ˇ 1
ˇ
f˝ 1 ::: k .x/; 1 ::: k .x/g 1 ˇ D f˝ 1 ::: k .x/; 1 ::: k .x/g :
2 x!x 2
(A-4.7)
This establishes (4.10.7) since a Lagrangian density, must contain an even number
of Lorentz indices and all products in it are symmetrized with respect to Bose-
Einstein factors and anti-symmetrized with respect to Fermi-Dirac fields, and due to
the Hermiticity of the Lagrangian density, and finally followed by a change of the
variable of integration x ! x.
Appendix A: Basic Equalities Involving the CPT Operator 219
Problems
4.1 Use (4.2.23), (4.2.24), (4.2.25), (4.2.26), (4.2.27), (4.2.28), (4.2.29), (4.2.30)
and (4.2.31), with the definition of the unit vector, n as given in (4.2.24) and
the explicit expression of the rotation matrix Œ R ij in (2.2.11) with D to
derive (4.2.32).
4.2 Show that
exp Œ i˛J 03 exp Œ i n J W 0

D .W 3 sinh ˛ C W 0 cosh ˛/ exp Œ i˛J 03 exp Œ i n J ;
thus verifying (4.2.34).

4.3 We are given two orthogonal four vectors P W D 0, and P P D 0,
W W D 0, P 0 D jPj > 0. Show that these imply that, W D P for
some numerical .
4.4 Derive the explicit expressions for W 0 ; W in (4.2.16).
4.5 Derive the transformation law in (4.3.48) as it follows from (4.3.47).
4.6 Under a Lorentz transformation: x ! x •x D x 0 , •x D •! x , a vector
field A˛ .x/ has the following transformation law19
i
A0˛ .x 0 / D ı˛ ˇ C •! S /˛ ˇ Aˇ .x/;
2
where note that the argument of A0˛ on the left-hand-side is x 0 and not x, and the
spin structure is given by
ˇ 1
S ˛
D ˛ ˇ ˛ ˇ :
i
Given the Maxwell Lagrangian is: L D .1=4/F ˛ˇ F˛ˇ where F ˛ˇ D .@˛ Aˇ

@ˇ A˛ /: (i) find the (symmetric) energy-momentum tensor, (ii) show it is gauge
invariant, (iii) show directly that it is conserved. What is its trace ?. [Note that the
canonical energy momentum tensor is not gauge invariant.]

Use the formal expression of the Dirac Lagrangian density: L D @=i/
4.7 C
m , to show that the (symmetric) energy-momentum tensor is given by: T D
! ! ! !

.1=4 i/ @ C @ , where @ D @ @ .
19
See also (4.4.2).
4.8 Find the expressions of the components T 00 .x/ T 0k .x/ of the energy-
momentum tensor of the Dirac quantum field in Problem 4.7 involving no time
derivatives.
4.9 Derive the following fundamental equal-time commutation relation of the com-
ponent T 00 of the energy-momentum
tensor for the Dirac theory in Problems 4.7
and 4.8: Œ T 00 .x/; T 00 .x 0 / D i T 0k .x/ C T 0k .x 0 / @k ı 3 .x x0 /. This was derived
by Schwinger [16, 17]. Commutations relations of all the components of the energy-
momentum tensor in the full QED theory, taking into account of the gauge problem,
are derived in Manoukian [5].
4.10 Given three 3-vectors eC ; e ; p=jpj, satisfying
p pi pi i j
e˙ e
D 0; e˙ e˙ D 1;
j
e˙ D 0; ı ij D C e i
C e C C e e ;
jpj jpj jpj
with the last relation providing a completeness relation in 3D space, verify explicitly
the four dimensional completeness relation (4.7.129) in Minkowski spacetime.
4.11 Show that the equation
h @ 1 @ @ 1 @ i
Cm C C C m .x/ D K

.x/;
i 3 i i 3 i
also leads to the equations in (4.7.137), (4.7.138) for K D 0, and that it merely
adds the non-propagating term .2=3 m2 /Œ p p C . p m/ to
the propagator . p/=Œ p 2 C m2 i
. This is the equation used, for example, by
Takahashi [19].
4.12 Derive the constraint on the spatial components i
.x/ of the m ¤ 0 Rarita-
Schwinger field in (4.7.148).
4.13 Derive the basic Eq. (4.7.150) needed for the m ¤ 0 Rarita-Schwinger field
components to satisfy the constraint in (4.7.148).
4.14 Establish the spin 3/2 nature of the m D 0 Rarita-Schwinger field
in (4.7.193).
4.15 The propagator for a massive neutral vector field was derived in (4.7.106),
(4.7.107), (4.7.108) and (4.7.109), and may be obtained by functional differentiation
• • ˇ
ˇ
.i/ .i/ .i/ 0
h 0 C j 0 i ˇ D C .x x 0 /;
•K .x/ •K .x / K D0
directly from the vacuum-to-vacuum transition amplitude h 0C j 0 i given

in (4.7.108). Use the identity in (4.6.38), following from the quantum dynamical
principle, to show that the time ordered product of the vector fields h 0C j
References 221
ˇ
ˇ
V .x/V .x 0 / C 0 iˇ satisfies
K D0
ˇ 1 0 .4/
ˇ
i h 0C j V .x/V .x 0 / C 0 iˇ D C .x x 0 / ı 0 ı .x x 0 /:
K D0 m2
That is, (i time-ordered-product) of the vector fields not only does not represent
the propagator of the vector field, but is also not covariant.
4.16 Show that if the interaction Lagrangian density in (4.8.11) is replaced, say, by
LI .x/ D . /ab Œ a .x/; a .x/ V .x/=2, which is linear in V .x/, then there is
no modifying factor as in (4.8.29) in the action integral.
References
1. Lam, C. S. (1965). Feynman rules and Feynman integrals for systems with higher-spin fields.
Nuovo Cimento, 38, 1755–1765.
2. Lehmann, H., Symanzik, K., & Zimmermann, W. (1955). The formulation of quantized field
theories. Nuovo Cimento, XI, 205–222.
3. Manoukian, E. B. (1985). Quantum action principle and path integrals for long-range
interactions. Nuovo Cimento, 90A, 295–307.
4. Manoukian, E. B. (1986). Action principle and quantization of Gauge fields. Physical Review
D, 34, 3739–3749.
5. Manoukian, E. B. (1987). On the relativistic invariance of QED in the Coulomb Gauge and
field transformations. Journal of Physics G, 13, 1013–1021.
7. Manoukian, E. B. (2016). Rarita-Schwinger massless field in covariant and Coulomb-like
gauges. Modern Physics Letters A, 31, 1650047, (1–8).
8. Rarita, W., & Schwinger, J. (1941). On a theory of particles with half-integral spin. Physical
Review, 60, 61.
9. Sakharov, A. D. (1967). Violation of CP invariance, C asymmetry, and Baryon asymmetry of
the Universe. Soviet Physics JETP Letters, 5, 24–27.
10. Schwinger, J. (1951a). On the Green’s functions of quantized fields. I. Proceedings of the
National Academy of Sciences USA, 37, 452–455.
11. Schwinger, J. (1951b). The theory of quantized fields. I. Physical Review, 82, 914–927.
12. Schwinger, J. (1953). The theory of quantized fields. II, III. Physical Review, 91, 713–728,
728–740.
13. Schwinger, J. (1958). Addendum to spin, statistics and the CPT theorem. Proceedings of the
14. Schwinger, J. (1960). Unitary transformations and the action principle. Proceedings of the
15. Schwinger, J. (1962). Exterior algebra and the action principle I. Proceedings of the National
Academy of Sciences USA, 48, 603–611.
16. Schwinger, J. (1963a). Commutation relations and consevation laws. Physical Review, 130,
406–409.
17. Schwinger, J. (1963b). Energy-momentum density in field theory. Physical Review, 130,
800–805.
18. Schwinger, J. (1970). Particles, sources, and fields (Vol. I). Reading: Addison-Wesley.
19. Takahashi, Y. (1969). An introduction to field quantization. Oxford: Pergamon Press.
Recommended Reading
1. Manoukian, E. B. (1986). Action principle and quantization of Gauge fields. Physical Review
D, 34, 3739–3749.
3. Manoukian, E. B. (2016). Rarita-Schwinger massless field in covariant and Coulomb-like
gauges. Modern Physics Letters A, 31, 1650047, (1–8).
4. Schwinger, J. (1970). Particles, sources, and fields (Vol. I). Reading: Addison-Wesley.
5. Weinberg, S. (1995). The quantum theory of fields. I: Foundations. Cambridge: Cambridge
University Press.
Chapter 5
Abelian Gauge Theories
Quantum electrodynamis (QED), describing the interactions of electrons (positrons)

and photons, is, par excellence, an abelian gauge theory. It is one of the most
successful theories we have in physics and a most cherished one. It stood the test
of time, and provides the blue-print, as a first stage, for the development of modern
quantum field theory interactions. A theory with a symmetry group in which the
generators of the symmetry transformations commute is called an abelian theory.
In QED, the generator which induces a phase change of a (non-Hermitian) charged
field .x/
.x/ ! ei.x/ .x/; (5.1)
is simply the identity and hence the underlying group of transformations is abelian
denoted by U.1/. The transformation rule in (5.1) of a charged field, considered as
a complex entity with a real and imaginary part, is simply interpreted as a rotation
by an angle .x/, locally, in a two dimensional (2D) space, referred to as charge
space.1
The covariant gauge description of QED as well as of the Coulomb one are
both developed. Gauge transformations are worked out in the full theory not only
between covariant gauges but also with the Coulomb one. Explicit expressions of
generating functionals of QED are derived in the differential form, as follows from
the quantum dynamical principle, as well as in the path integral form. A relatively
simple demonstration of the renormalizability of QED is given, as well as of the
renormalization group method is developed for investigating the effective charge.
A renormalization group analysis is carried out for investigating the magnitude of
the effective fine-structure at the energy corresponding to the mass of the neutral
vector boson Z 0 , based on all of the well known charged leptons and quarks of
1
A geometrical description is set up for the development of abelian and non-abelian gauge theories
in a unified manner in Sect. 6.1 and may be beneficial to the reader.

DOI 10.1007/978-3-319-30939-2_5
224 5 Abelian Gauge Theories
specific masses which would contribute to this end. This has become an important
reference point for the electromagnetic coupling in present high-energy physics.
The Lamb shift and the anomalous magnetic moment of the electron, which have
much stimulated the development of quantum field theory in the early days, are
both derived. We also include several applications to scattering processes as well
as of the study of polarization correlations in scattering processes that have become
quite interesting in recent years. The theory of spontaneous symmetry breaking is
also worked out in a celebrated version of scalar boson electrodynamics and its
remarkable consequences are spelled out. Several studies were already carried out
in Chap. 3 which are certainly relevant to the present chapter, such as of the gauge
invariant treatment of diagrams with closed fermion loops, fermion anomalies in
field theory, as well as other applications.2
5.1 Spin One and the General Vector Field
Referring to Sect. 4.7.3, let us recapitulate, in a slightly different way, the spin 1
character of a vector field. Under an infinitesimal rotation c.c.w of a coordinate
system, in 3D Euclidean space, by an infinitesimal angle •# about a unit vector N,
a three-vector x, now denoted by x 0 in the new coordinate system, is given by
x0 D x •# .N x/ D x C •# .x N/; (5.1.1)
0i
x D ı ij C•# " ij k N k xj ; •! ij D •# " ij k N k D •! j i ; (5.1.2)
where " ij k is totally anti-symmetric with "123 D C1, from which the matrix
elements of the rotation matrix for such an infinitesimal rotation are given by3
R ij D ı ij C •# " ij k N k : (5.1.3)
Under such a coordinate transformation, a three-vector field Ai .x/, in the new

coordinate system, is given by

A 0 i .x 0 / D ı ij C •! ij Aj .x/; (5.1.4)
which may be rewritten as

i
A 0 i .x 0 / D ı ij C •! k` Œ S k` ij Aj .x/; (5.1.5)
2
ij i
D ı C •# " k`q N q Œ S k` ij Aj .x/; (5.1.6)
2
2
It is worth knowing that the name “photon” was coined by Lewis [43].
3
See also Eq. (2.2.11).
5.1 Spin One and the General Vector Field 225
where
1 ki `j
Œ S k` ij D ı ı ì kj ; i; j; k; ` D 1; 2; 3: (5.1.7)
i
One may then introduce the spin matrices Œ S q , q D 1; 2; 3,
1 k`q k` ij
Œ S q ij D " ŒS ; (5.1.8)
2
and rewrite (5.1.6) as

A 0 i .x 0 / D ı ij C i •# ŒS ij N Aj .x/: (5.1.9)
It is readily verified that
3
X 0 0j
Œ S2 ij D Œ Sq i i Œ Sq i D 2 ı ij D s.s C 1/ ı ij ; (5.1.10)
qD1
establishing the spin s D 1 character of the vector field, with the spin components
satisfying the well known commutations relations
0 0
Œ S q ; S q D i " qq k S k : (5.1.11)
As a direct generalization of (5.1.5), (5.1.6) and (5.1.7), a vector field A˛ .x/ has
the following transformation under a Lorentz transformation: x ! x •x D x 0 ,
•x D •! x ,

A0˛ .x 0 / D ı ˛ ˇ C •! ˛ ˇ Aˇ .x/ (5.1.12)
i
D ı ˛ ˇ C •! S /˛ ˇ Aˇ .x/; •! D •! ; (5.1.13)
2
for a covariant description,4 where we note that the argument of A ˛0 on the left-hand
side is again x 0 and not x,
ˇ 1
S ˛ D ˛ ˇ ˛ ˇ : (5.1.14)
i
4
See (4.7.120), (4.7.121), (2.2.17), (2.2.18), (2.2.19), (2.2.20), (2.2.21) and (2.2.22).
5.2 Polarization States of Photons
The polarization vectors e1 ; e2 of a photon are mutually orthogonal and are, in turn,
orthogonal to its momentum vector k. With the vector k chosen along the z-axis,
we may then introduce three unit vectors
k
nD D .0; 0; 1/; e1 D .1; 0; 0/; e2 D .0; 1; 0/; (5.2.1)
jkj
with the latter two providing a real representation of the polarization vectors,
satisfying
X
e i e ; ; 0 D 1; 2; i; j D 1; 2; 3;
j
n e D 0; e e0 D ı;0 ; ı ij D ni nj C
D1;2
(5.2.2)
where specifies the two polarization vectors, and the index i specifies the ith
component of the vectors. The equality, involving ı ij , is a completeness relation in
three dimensions for expanding a vector in terms of the three unit vectors n; e1 ; e2 .
One may also introduce a complex representation of the polarization vectors,
such as, .e˙ e˙ 1 /
1 1
eC D p .1; i; 0/; e D p .1; i; 0/; e e0 ; D ı;0 ; ; 0 D ˙ 1:
2 2
(5.2.3)
The completeness relation now simply reads
X j
X
ei
j
ı ij D ni nj C ei e D ni nj C e ; (5.2.4)
D˙ 1 D˙ 1
as is easily checked by considering specific values for the indices i; j specifying

components of the vectors.
One would also like to have the general expressions of the polarization vectors,
when the three momentum vector of a photon k has an arbitrary orientation
k D jkj .cos sin ; sin sin ; cos /: (5.2.5)
To achieve this, we rotate the initial coordinate system in which the vector k
is initially along the z-axis, c.c.w. by an angle about the unit vector N D
.sin ; cos ; 0/ as shown in (Fig. 5.1), by using the explicit structure of the
5.2 Polarization States of Photons 227
Fig. 5.1 The initial frame is

rotated c.c.w by an angle
about the unit vector N so
that k points in an arbitrary θ k y
direction in the new frame
φ
N
rotation matrix5 with matrix elements

R i k D ı i k " ij k N j sin C ı i k N i N k .cos 1/; i; j; k D 1; 2; 3; (5.2.6)
where " ij k is totally anti-symmetric with "123 D 1.

The rotation matrix gives the following general expressions for the polarization
vectors: (see Problem 5.1)
e1 D .cos 2 cos C sin2 ; sin cos .cos 1/; cos sin /;
(5.2.7)
e2 D .sin cos .cos 1/; sin2 cos C cos 2 ; sin sin /:
(5.2.8)
X
; 0 D 1; 2:
j
e e0 D ı0 ; k e D 0; ı ij D ni nj C e i e ; (5.2.9)
D1;2
for a real representation, and
1
eC D p . cos cos C i sin ; sin cos i cos ; sin / ei ;
2
(5.2.10)
1
e D p .cos cos C i sin ; sin cos i cos ; sin / ei ;
2
(5.2.11)
5
A reader who is not familiar with this expression may find a derivation of it in Manoukian [56],
p. 84. See also (2.2.11).
X j
X
ei
j
ı ij D ni nj C ei e D ni nj C e ; (5.2.12)
D˙ 1 D˙ 1
e e0 D ı0 ; e˙ D e

; k e D 0; ; 0 D ˙ 1: (5.2.13)
for a complex representation.

We need to introduce a covariant description of polarization. Since we have two
polarization states, we have the following orthogonality relations

e e
0 D ı0 ; k e D 0; ; 0 D ˙ 1; k2 D 0; (5.2.14)
for example working with a complex representation. The last orthogonality relation
implies that
k0 e0 D k e ; k0 D jkj; (5.2.15)
and with k e D 0, we take e0 D 0, and set

e D .0; e /: (5.2.16)
In order to write down a completeness relation in Minkowski spacetime, we may

introduce two additional vectors6 to eC ; e : k D .k 0 ; k/, k D .k 0 ; k/, where
we note that .k C k/ is a time-like vector, while .k k/ is a space-like one. Also
k e D 0; k e D 0: (5.2.17)
The completeness relation simply reads as
.k C k/ .k C k/ .k k/ .k k/ X

D C C e e
; (5.2.18)
.k C k/2 .k k/2
D˙ 1
which on account of the facts that k2 D 0; k2 D 0, this simplifies to
k k C k k X
D C e e
: (5.2.19)
kk
D˙ 1
5.3 Covariant Formulation of the Propagator
The gauge transformation of the Maxwell field A .x/ is defined by A .x/ !

A .x/ C @ .x/, and with arbitrary .x/, it leaves the field stress tensor F .x/ D
@ A .x/ @ A .x/ invariant. In particular, a covariant gauge choice for the
6
We follow Schwinger’s elegant construction [70].
5.3 Covariant Formulation of the Propagator 229
electromagnetic field is @ A .x/ D 0. We will work with more general covariant

gauges of the form
@ A .x/ D .x/; (5.3.1)
where is an arbitrary real parameter and .x/ is a real scalar field. The gauge
constraint in (5.3.1) may be derived from the following Lagrangian density
1
L D F F C J A @ A C 2 (5.3.2)
4 2
where J .x/ is an external, i.e., a classical, current. Variation with respect to ,

gives (5.3.1), i.e., the gauge constraint is a derived one. While variation with respect
to A leads to7
@ F .x/ D J .x/ C @ .x/: (5.3.3)
Using the expression F D @ A @ A , the above equation reads
A .x/ D J .x/ C .1 / @ .x/; (5.3.4)
where we have used the derived gauge constraint in (5.3.1).

Upon taking the @ derivative of (5.3.3), we also obtain
.x/ D @ J .x/: (5.3.5)
We consider the matrix element h 0C j : j 0 i of (5.3.4), to obtain

Z
h 0C jA .x/ j 0 i h 0C j.x 0 / j 0 i
D .dx 0 /DC .x x 0 / J .x 0 / C .1 /@0 ;
h 0C j 0 i h 0C j 0 i
(5.3.6)

where DC .x x 0 / is the propagator DC .x x 0 / D ı .4/ .x x 0 /
Z 0
.dk/ eik.xx /
DC .x x 0 / D ; .dk/ D dk0 dk1 dk2 dk3 : (5.3.7)
.2 /4 k2 i
7
Since the gauge constraint is now a derived one, one may vary all the components of the vector
field independently.
Taking Fourier transforms of (5.3.6), and using (5.3.5), the following expression
emerges
Z
h 0C jA .x/ j 0 i
D .dx 0 /D .x x 0 /J .x 0 /; (5.3.8)
h 0C j 0 i
Z
.dk/ h k k i eik.xx /
0
0
D .x x / D .1 / 2 ; (5.3.9)
.2 /4 k k2 i
defining covariant photon propagators8 in gauges specified by the parameter . The

gauge specified by the choice D 1 is referred to as the Feynman gauge, while the
choices D 0 as the Landau gauge, and D 3 as the Yennie-Fried gauge.
Upon using
•
i h 0C j 0 i D h 0C jA .x/ j 0 i; (5.3.10)
•J .x/
we may integrate (5.3.8) to obtain

hiZ i
h 0C j 0 i D exp .dx/.dx 0 /J .x/D .x x 0 /J .x 0 / (5.3.11)
2
normalized to unity for J .x/ D 0. The generating functional h 0C j 0 i

is determined in general covariant gauges specified by the values taken by the
parameter in (5.3.9)
The matrix element h 0C jF .x/ j 0 i of the field strength tensor F , is given
from (5.3.8) to be
Z
h 0C jF .x/ j 0 i
D .@ @ / .dx 0 /DC .x x 0 / J .x 0 /; (5.3.12)
h 0C j 0 i
and the gauge parameter in (5.3.9) cancels out on the right-hand side of the
equation.
It is important to note that for a conserved external current @ J .x/ D 0,
hi Z i
ˇ
h 0C j 0 iˇ@J D0 D exp .dx/.dx 0 /J .x/DC .x x 0 /J .x 0 / ; (5.3.13)
2
is independent of the gauge parameter , and DC .x x 0 / is defined in (5.3.7).
8
For J D 0.
5.4 Casimir Effect 231
Fig. 5.2 The parallel plates

in question are placed
between two parallel plates
+L
situated at large distances
+a/2
0
−a/2
5.4 Casimir Effect
The Casimir effect, in its simplest theoretical description, is an electromagnetic

force of attraction between two parallel perfectly conducting neutral plates in
vacuum. It is purely quantum mechanical, i.e., it is attributed to the quantum
nature of the electromagnetic field, and to the nature of the underlying boundary
condition imposed on it by the presence of the plates. It is one of those mysterious
consequences of quantum theory, i.e., an „-dependent result, that may be explained
by the response of the vacuum to external agents. By a careful treatment one may
introduce, in the process, a controlled environment, by placing the parallel plates
between two perfectly conducting plates placed, in turn, at very large macroscopic
distances (Fig. 5.2) from the two plates in question.9 This analysis clearly shows
how a net finite attractive arises between the plates.
The electric field components, in particular, tangent to the plates satisfy the
boundary conditions
ˇ
ˇ
ET .x0 ; xT ; z/ˇ D 0: (5.4.1)
z D ˙a=2; ˙L
Upon taking the functional derivative of (5.3.12), with respect to the external
current J ˛ , we obtain, in the process, for x 0 0 > x 0 ,
i hvacjF ˛ˇ .x 0 / F .x/jvaci

D @0˛ .@ ˇ @ ˇ / @0ˇ .@ ˛ @ ˛ / DC .x; x 0 /; (5.4.2)
9
Schwinger [72] and Manoukian [50].
where we have finally set J D 0, and, in the absence of the external current, we
have replaced j0˙ i by jvaci. Here DC .x; x 0 / satisfies the equation
DC .x; x 0 / D ı .4/ .x; x 0 /; (5.4.3)
with appropriate boundary conditions. Because translational invariance is broken

(along the z-axis), we have replaced the arguments of DC and ı .4/ by .x; x 0 /.
The Electric field components are given by E i D F 0 i , and the magnetic field
ones by B i .x/ D .1=2/" ij k F j k . Hence, in particular,10

i hvacjET .x 0 /ET .x/jvaci D 2 @0 0 @0 r 0T r T D< 0 0 00
C .x; x /; x < x : (5.4.4)
In reference to this equation, corresponding to the tangential components of the

electric fields, the boundary conditions in (5.4.1), implies a Fourier sine series for
ı.z ; z 0 /:
1
2 X n.z d/ n.z 0 d/
ı.z ; z 0 / D sin sin ; (5.4.5)
R nD1 R R
in (5.4.3), where
R D L a=2; d D a=2; for a=2 z ; z 0 L

RD a; d D a=2; for a=2 z ; z 0 a=2 : (5.4.6)
R D L a=2; d D a=2; for L z ; z 0 a=2
This leads to
1 Z
2 X d2 K exp Œ i K .xT x0T /
D< 0
C .x; x / Di
R nD1 .2/2 2E n .K; R/
n.z d/ n.z 0 d/

sin sin exp i E n .K; R/jx 0 x 0 0 j ; (5.4.7)
R R
r
n2 2
En .K; R/ D K2 C 2 : (5.4.8)
R
10
D< 0 0 0
C .x; x / stands for DC .x; x / for x < x
00
.
5.4 Casimir Effect 233
Similarly, we may consider the case x 0 0 < x 0 , and upon taking the average of
both cases, the following expression emerges
1 Z
1 2 X d2 K exp Œ i K .xT x0T / 2 n2 2
hvacj fET .x/; ET .x 0 /gjvaci D K C 2
2 R nD1 .2/2 2E n .K; R/ R2
n.z d/ n.z 0 d/

sin sin exp i E n .K; R/jx 0 x 0 0 j : (5.4.9)
R R
For the component E3 , ı.z ; z 0 / in (5.4.3) is to be expressed in a Fourier cosine

series expansion:
1
1 2 X n.z d/ n.z 0 d/
ı.z ; z 0 / D C cos cos : (5.4.10)
R R nD1 R R
This leads from (5.4.2) to

1 Z
1 2 X d2 K exp Œ i K .xT x0T / 2
hvacj fE3 .x/; E3 .x 0 /gjvaci D K
2 R nD1 .2/2 2E n .K; R/
n.z d/ n.z 0 d/

cos cos exp i E n .K; R/jx 0 x 0 0 j : (5.4.11)
R R
We may repeat a similar analysis for the magnetic field with corresponding
boundary conditions: B 3 .x 0 ; xT ; z/jz D ˙ a=2; ˙L D 0. The total vacuum energy of
the system may be then defined by ( x 0 x 0 0 T, L ! 1)
Z 1 Z
1 1 1
ED dTı.T/ d3 x hvacj fE.x 0 ; x/; E.x 00 ; x/gC fB.x 0 ; x/; B.x 00 ; x/gjvaci;
1 2 2 2
(5.4.12)
where the magnetic field contribution is identical to the electric one.
Upon using the elementary integrals
Z Z
2 R
nz 2 R
nz
dz sin2 D 1 D dz cos2 ; (5.4.13)
R 0 R R 0 R
we obtain the simple expression

r r
XZ 1 Z
d2 K X
1
n2 2 h n2 2 i
E DA dTı.T/ K C 2 exp i K2 C 2 T ;
2
R 1 .2/2 nD1 R R
(5.4.14)
R
where A D d2 xT , is the area of any of the plates, and the sum is over R D La=2
twice, and R D a once.
Now we use the following basic identity (see Problem 5.3)

r r
@ 2 n2 2 h
2 n2 2 i
K C exp i K C T
@a R2 R2

q
n2 2
2 n2 2 @R d2
2
@ exp i K C R2 T
D q ; (5.4.15)
R3 @a dT 2 @K2 2 2
K2 C nR2
and the integral (d2 K D 2jKjdjKj D dK2 )

q
Z 2 exp i
2 2
K2 C nR2 T
dK @ 1 exp Œ i n T=R
q D ; (5.4.16)
.2/2 @K2 2 n2 2
K C R2 4 n=R
to obtain
Z
@R d2 2 n h i n T i
1
1 @E 1 X 1 X
D dT ı.T/ exp ;
A @a 4 R 1 nD1
@a dT 2 R2 R
Z 1 h i n T i
i X 1 X @R=@a d3
D dT ı.T/ 3
exp ; (5.4.17)
2 R 1 nD1
R dT R
whose interpretation will soon follow. Upon carrying the elementary summations
over n and R, with the latter summation over R as described below (5.4.14), the
above equation becomes
Z 1
1 @E 1 d3
D dT ı.T/ F.a; T; L/; (5.4.18)
A @a 4a 1 dT 3
where
h 2i 2ia 1 i
F.a; T; L/ D : (5.4.19)
1 eiT=a L a=2 1 eiT=.La=2/
The expansion
2i 2 x x3
D i C C ; (5.4.20)
1 eix x 6 360
gives .L ! 1/
h 2 a T 3 T 3 2a aT i
F.a; T; L/ D C C : (5.4.21)
T 6a 360 a3 T 6L2
5.5 Emission and Detection of Photons 235
The interpretation of this expansion is now clear in view of its application

in (5.4.18). The first term 2a=.T/ within each of the round brackets above gives
each an infinite force per unit area on the plate when taken each separately, but these
forces are equal and in opposite directions, and hence cancel out. The expression
.d3 =dT 3 /F.a; T; L/ will then lead to a finite attractive force between the plates in
question, coming solely, for L ! 1, from the third term . 3 T 3 =360a3/ within
the first round brackets in (5.4.21), leading from (5.4.18) to the final expression
1 @E 2 „ c
D ; (5.4.22)
A @a 240 a 4
where we have re-inserted the fundamental constants „ and c in the final

expression.
The above beautiful result was first obtained by Casimir [16], and an early
experiment by Sparnaay [73] was not in contradiction with Casimir’s prediction.
More recent experiments (Lamoreaux [41], and, e.g., Bressi et al. [13]), however,
were more positive and showed agreement with theoretical predictions of the
Casimir effect within a few percent. Casimir forces are not necessarily attractive and
may be also repulsive depending on some factors such as on underlying geometrical
situations.11
The Casimir effect may be also derived by the method of the Riemann zeta
function regularization,12 a method that we use, e.g., in string theory.13 The above
derivation, however, is physically more interesting, and clearly emphasizes the
presence of arbitrary large forces, in opposite directions, within and out of the plates
which precisely cancel out leading finally to a finite calculable result.
5.5 Emission and Detection of Photons

ˇ
We consider the vacuum-to-vacuum transition apmlitude h 0C j 0 iˇ@J D0 in
(5.3.13) for the interaction of photons with a conserved external current
@ J .x/ D 0. To simplify the notation only, we will simply write this amplitude
as h 0C j 0 i. It may be rewritten as
h i Z .dk/ 1 i

h 0C j 0 i D exp J .k/ J .k/ ; (5.5.1)
2 .2 /4 k2 i
where note that the reality of J .x/ implies that J .k/ D J .k/.
11
See, e.g. Kenneth et al. [36] and Milton et al. [64].
12
See, e.g., Elizalde et al. [20] and Elizalde [19].
13
See Vol. II: Quantum Field Theory II: Introductions to Quantum Gravity, Supersymmetry, and
String Theory, (2016), Springer.
To compute the vacuum persistence probability jh 0C j 0 ij2 , we note that
i 2 Œ ı.k0 jkj/ C ı.k0 C jkj/

Re D ı.k / D : (5.5.2)
k2 i
2 jkj
This gives
Z X

d3 k
jh 0C j 0 ij2 D exp 3 0
.J .k/ e / .J .k/ e
/ 1; k0 D Cjkj;
.2 / 2k
D˙
(5.5.3)
where we have used the completeness relation, expressed
P in terms of the Minkowski

metric, in (5.2.19): D .k k C k k /=kk C D˙ 1 e e , with, e.g., complex
representation of polarizations vectors, and used the conservation laws: J .k/k D
0, k J .k/ D 0. This gives the consistent probabilistic result that the vacuum
persistence probability does not exceed one.
We use the convenient notation for bookkeeping purposes14
s
d3 k
iJ .k/ e D i Jk ; (5.5.4)
.2 /3 2k0
and don’t let the notation scare you. We may then repeat the analysis given
through Eqs. (3.3.28), (3.3.29), (3.3.30), (3.3.31), (3.3.32), (3.3.33), (3.3.34),
(3.3.35) and (3.3.36), now applied to photons, and using the expressions in
Eqs. (5.5.1), (5.5.3), (5.5.4), to infer from Eqs. (3.3.37) and (3.3.38), that the
probability that an external source J emits N photons, Nk of which have each

momentum k, and polasrization e , and so on, is given by
Nk1 1 Nk2 2
X jJk1 1 j2 jJk2 2 j2
Prob Œ N D jh 0C j 0 ij2 :
Nk1 1 Š Nk2 2 Š
N D .Nk1 1 CNk2 2 C /
(5.5.5)
Now we use the multinomial expansion15
X .jx1 j/ N1 .jx2 j/ N2 .jx1 j C jx2 j C / N

D ; (5.5.6)
N1 Š N2 Š NŠ
ND.N1 CN2 C /
14
This was conveniently introduced by Schwinger [70].
15
See, e.g., Manoukian [57].
5.5 Emission and Detection of Photons 237
and thanks to the convenient bookkeeping notation introduced above in (5.5.4), we

also have
Z X
2 2 d3 k
jJk1 1 j C jJk2 2 j C D jJ .k/ e j2 : (5.5.7)
.2 /3 2k 0

This allows us to rewrite (5.5.5) as

N
hN i
Prob Œ N D ehN i ; (5.5.8)
NŠ
where
Z
d3 k
hN i D J .k/J .k/; (5.5.9)
.2 /3 2k 0
and we have used (5.5.3) to write jh 0C j 0 ij2 D ehN i . We recognize Prob Œ N ,

in (5.5.8), as defining the Poisson distribution16 with hN i denoting the average
number of photons emitted by the external source.
In evaluating hNi, it is often more convenient to rewrite its expression involving
integrals in spacetime. This may be obtained directly from (5.5.9) (see also (5.5.2)
and (5.5.3)) to be
Z
hN i D .dx/ .dx 0 / J .x/ Œ Im DC .x x 0 / J .x 0 /: (5.5.10)
where, as in (5.5.9), @ J D 0, and DC .x x 0 / is defined in (5.3.7). Equa-

tion (5.5.10) may be equivalently rewritten as
Z Z
0 0 d3 k 0
hN i D .dx/ .dx / J .x/ J .x / 3 0
eik.xx / ; k0 D jkj; (5.5.11)
.2/ 2k
using the reality condition of the current. For an application of the above expression
in deriving the general classic radiation theory see Manoukian [58].
We may infer from Eqs. (3.3.39) and (3.3.41), that the amplitude of a current

source, as a detector, to absorb a photon with momentum k and polarization e ,
and the amplitude of a current source, as an emitter, to emit a photon with the same
attributes which escapes this parent source, to be used in scattering theory, are given,
16
See, e.g., op. cit.
respectively, by
s
d3 k
h 0C jki D i J .k/ e : (5.5.12)
.2 /3 2k0
s
d3 k
h k j 0 i D i J .k/ e ;; (5.5.13)
.2 /3 2k0
where we have omitted photons which are emitted and reabsorbed by the same
source as they do not participate in a scattering process, by dividing by the
corresponding vacuum-to-vacuum amplitudes h 0C j 0 i.17
5.6 Photons in a Medium
We consider a homogeneous and isotropic medium of permeability , and per-

mittivity ". To describe photons in such a medium, one simply scales F0 i F 0 i !
" F0 i F 0 i , and Fij F ij ! Fij F ij =, in the Lagrangian density, where F D
@ A @ A .18 That is, the Lagrangian density becomes
1 "
L D Fij F ij F0 i F 0 i C J A : (5.6.1)
4 2
We take advantage of our analysis in Sect. 5.5, and reduce the problem to the one
carried out in that section. To this end, we consider the following scalings:
p 1
x0 D " x 0 ; x D x; @ 0 D p @ 0; r D r ; (5.6.2)
"
1 1=4
A0 .x/ D A0 .x/; A.x/ D A.x/; (5.6.3)
1=4 "3=4 "1=4
"1=4 0 1
J 0 .x/ D J .x/; J.x/ D J.x/: (5.6.4)
1=4 3=4 "1=4
17
See discussion above Eq. (3.3.39).
18
Note that the scaling factors are not "2 , 1=2 , respectively, as one may naïvely expect. The
reason is that functional differentiation of the action, with respect to the vector potential, involving
the quadratic terms " F0 i F0 i , Fij Fij =, generate the linear terms corresponding to the electric and
magnetic fields components which are just needed in deriving Maxwell’s equations.
5.6 Photons in a Medium 239
The action then simply becomes

Z h 1 i
WD .dx / @ A .x/@ A .x/ @ A .x/@ A .x/ CJ .x/ A .x/ :
4
(5.6.5)
up to a gauge fixing constraint. Note that the argument of A .x/ is x and not x ,
also that
@ J .x/ D 3=4 "1=4 @ J .x/: (5.6.6)
The field equation is given by
h 0C jA .x/ j 0 i
D J .x/: (5.6.7)
h 0C j 0 i
up to gauge fixing terms, proportional to @ , which do not contribute when one

finally imposes current conservation. Thus upon defining the propagator DC .x x 0 /
satisfying
DC .x x 0 / D ı .4/ .x x 0 /; (5.6.8)
we have
Z
h 0C jA .x/ j 0 i
D .dx 0 / DC .x x 0 / J .x 0 /; (5.6.9)
h 0C j 0 i
where note that ı .4/ .x x 0 / .dx 0 / D ı .4/ .x x 0 / .dx 0 /.

Hence from (5.5.10), we have for the average number of photons emitted by a
current source in the medium
Z
hN i D .dx / .dx 0 / J .x/ Œ Im DC .x x 0 / J .x 0 /; (5.6.10)
hiZ i
h 0C j 0 i D exp .dx / .dx 0 /J .x/ DC .x x 0 / J .x 0 / ; (5.6.11)
2
for a conserved current @ J .x/ D 0 (see (5.6.6)). From (5.6.2) and (5.6.4), the
h i 1=2 Z 1 0 i
0 0 0 0 0
h 0C j 0 i D exp .dx/ .dx / J.x/J.x / J .x/J .x / D C .x x / ;
2 "1=2 "
(5.6.12)
Z p
ik0 .x 0 x 0 0 /= "
.dk/ 0 e
DC .x x 0 / D eik.xx / : (5.6.13)
.2 /4 k2 i
Since the current J .x/ is real, we may rewrite (5.6.10) as

Z
1=2 1 0
hN i D 1=2 .dx/ .dx 0 / J.x/J.x 0 / J .x/J 0 .x 0 /
" "
Z
d3 k 0 0 00 p
3
ei k.xx / ei jk j.x x /= " : (5.6.14)
.2 / 2 jk j
Upon inserting the identity

Z
1
jkj
1D d! ı ! p ; (5.6.15)
0 "
R1
in (5.6.14), we get hNi D 0 d!N.!/, ( k D jkjn )
p Z
" 0 0 1 0 0 0
N.!/ D ! .dx/ .dx / J.x/J.x / J .x/J .x /
16 3 "
Z p
0 00 0
ei!.x x / d˝ ei " ! n.xx / : (5.6.16)
Consider a charged particle, of charge e, moving along the z-axis with constant
speed v. Then
J i .x/ D e v ı i 3 ı.x 1 / ı.x 2 / ı.x 3 vx 0 /; J 0 .x/ D e ı.x 1 / ı.x 2 / ı.x 3 v x 0 /:

(5.6.17)
Then n .x x 0 / D .x 3 x 03 / cos , d˝ D 2 d cos . Upon integrating over x 0 ,
x 0 0 , and then over x 0 3 , we obtain, per unit length,
ˇ Z
ˇ e2 1 1 1
N.!/ˇ D 1 d cos ı cos p : (5.6.18)
unit length 4 " v2 1 v "
R
where the unit length arises from the integral dx 3 . The average number of photons
emitted with angular frequency in the interval .!; ! C d!/, as the charged particle
traverses a unit length, is then
ˇ e2 1
ˇ
N.!/ˇ D 1 ; (5.6.19)
unit length 4 " v2
and, on account the property of a cosine function, the latter does not vanish only for
p
v > 1= ". For a large number of charged particles this number need not be small.
5.7 Quantum Electrodynamics, Covariant Gauges: Setting Up the Solution 241
The expression in (5.6.19) is constant in ! and hence cannot be integrated over !

for arbitrary large !. A quantum correction treatment, however, provides a natural
cut-off in ! emphasizing the importance of the inclusion of radiative corrections.19
This form of radiation is referred to as C̆erenkov radiation. It is interesting that
astronauts during Apollo missions have reported of “seeing” flashes of light even
with their eyes closed. An explanation of this was attributed to high energy cosmic
particles, encountered freely in outer space, that would pass through one’s eyelids
causing C̆erenkov radiation to occur within one’s eye itself.20
5.7 Quantum Electrodynamics, Covariant Gauges: Setting

Up the Solution
We apply the functional differential formalism (Sects. 4.6 and 4.8), via the quantum
dynamical principle (QDP), to derive an explicit expression of the full QED
vacuum-to-vacuum transition amplitide h 0C j 0 i in covariant gauges (Sect. 5.3).
By carrying the relevant functional differentiations coupled to a functional Fourier
transform (Sect. 2.6), the path integral form of h 0C j 0 i is also derived. The
corresponding expressions in the Coulomb gauge will be derived in Sect. 5.14.
Gauge transformations of h 0C j 0 i between covariant gauges as well as between
covariant gauges and the Coulomb one will be derived in Sect. 5.15.
5.7.1 The Differential Formalism (QDP) and Solution of QED

in Covariant Gauges
The purpose of this subsection is to find explicit expressions for h 0C j 0 i of

QED in covariant gauges. The vacuum-to-vacuum transition amplitude h 0C j 0 i
is a generating functional for all possible underlying physical processes and also
for extracting various components of the theory, such a propagators and Green
functions.
For the Lagrangian density of QED, we consider
1 1 @ @
L D F F C m0
4 2 i i
1
C e0 Œ ; A C C C A J @ A C 2 ; (5.7.1)
2 2
19
For these additional details, see, e.g., Manoukian and Charuchittapan [59].
20
See, e.g., Fazio et al. [24], Pinsky et al. [66], and McNulty et al. [62].
where the field leads to a constraint on the vector potential A . That is, the
underlying constraint is derived from the Lagrangian density. As we will also see ,
in turn, may be eliminated in favor of @ A . We have written the parameters e0 ; m0 ,
with a subscript 0, to signal the facts that these are not the parameters directly
measured, as discussed in the Introduction of the book. The electromagnetic current
j D e0 Œ ; =2, has been also written as a commutator, consistent with charge
conjugation as discussed in Sect. 3.6 (see (3.6.9)).
The field equations together with the constraint are
@ F D J C j C @ ; (5.7.2)
h @ i

e 0 A C m0 D ; (5.7.3)
i
h @ i

C e0 A C m0 D ; (5.7.4)
i
@ A D : (5.7.5)
We take the expectation value of (5.7.2) between the vacuum states h 0C j : j 0 i,

take its @ derivative to derive the following equation

h 0C j.x/ j 0 i D @ J .x/h 0C j 0 i C h 0C j j .x/ j 0 i ; (5.7.6)
and use it in (5.7.5), to obtain the following equation of constraint

Z

@ h 0C jA .x/ j 0 iD .dx0 /DC .xx0 /@0 J .x0 /h 0C j 0 iCh 0C j j .x0 / j 0 i :
(5.7.7)
This, in turn, eliminates the field.
Upon taking the matrix element of (5.7.3) between the vacuum states, as
indicated by the following notation h 0C j : j 0 i, we obtain
h @ • i

e0 .i/
C m0 h 0 C j j 0 i D h 0C j 0 i; (5.7.8)
i •J
where we have used the quantum dynamical principle to write
•
h 0C j.A .x 0 / .x//C j 0 i D .i/ h 0C j .x/ j 0 i; (5.7.9)
•J .x 0 /
with the time ordered product in the limit x 0 $ x understood as an average of the
product of the two fields, as discussed in Sect. 4.9.
The vector potential A has been eliminated in favor of the “classical field”,
represented by, .i/•=•J in (5.7.8). Thus introducing the “classical field”
•
.i/ b
A .x/; (5.7.10)
•J .x/
and the Green function SC .x; x 0 I e0b

A/ satisfying the equations
h @ i

e0 b
A C m0 SC b c .x; x 0 I e0b
A/ D ı .4/ .x; x 0 / ıa c ; (5.7.11)
i ab
h @ i

C e0 b
A C m0 SC c b .x 0 ; x I e0b
A/ D ı .4/ .x 0 ; x/ ıa c ; (5.7.12)
i ba
we may, from (5.7.8), conveniently write

Z
h 0C j a .x/ j 0 i D .dx 0 / SC ab .x; x 0 I e0b
A/ b .x 0 /h 0C j 0 i: (5.7.13)
Similarly we have
Z
h 0C j a .x/ j 0 i D .dx 0 / SC c a .x 0 ; x I e0b
A/ c .x 0 /h 0C j 0 i: (5.7.14)
As we have already seen in Sect. 3.2, (3.2.12), or directly from (5.7.11)

and (5.7.12), the Green function SC .x; x 0 I e0b
A / satisfies the integral equation
Z
SC .x; x 0 I e0b
A/ D SC .xx 0 /Ce0 .dy/ SC .xy/ b
A.y/ SC .y; x 0 I e0b
A/; (5.7.15)
as is readily verified, where SC .x x 0 / is the free Dirac propagator satisfying,

@ @
C m0 SC .x x 0 / D ı .4/ .x x 0 /; SC .x 0 x/ C m0 D ı .4/ .x 0 x/:
i i
(5.7.16)
To obtain an expression for h 0C j 0 i in QED, we use the quantum dynamical
principle, and take its derivative with respect to e0 to obtain21 (see also (3.6.12),
(3.6.13), (3.6.14) and (3.6.15))
Z
@
.i/ h 0C j 0 i D .dx/h 0C j A .x/ .x/ .x/ C j 0 i (5.7.17)
@e0
Z
•
D .dx/.i/ h 0C j .x/ .x/ C j 0 i (5.7.18)
•J .x/
Z
• • •
D .dx/.i/ .i/ .i/ h 0C j 0 i:
•J .x/ •.x/ •.x/
(5.7.19)
21
Recall from Sect. 3.9, that within a time ordered product the Dirac fields anti-commute.
Equation (5.7.19) may be readily integrated with respect to e0 , to obtain

h Z • • • i
h 0C j 0 i D exp e0 .dx/ h 0C j 0 i0 ; (5.7.20)
•J .x/ •.x/ •.x/
where from (3.3.28) and (5.3.11),

h Z i
h 0C j 0 i0 D exp i .dx/ .dx 0 / .x/SC .x x 0 /.x 0 /
hi Z i
exp .dx/ .dx 0 / J .x/D .x x 0 /J .x 0 / ; (5.7.21)
2
where SC .x x 0 / is given in (3.1.9), and D .x x 0 / is given in (5.3.9). Thus

h 0C j 0 i may be obtained by functional differentiations of the explicitly given
expression of h 0C j 0 i0 .22
A far more interesting expression for h 0C j 0 i, and a more useful one for
practical applications, is obtained by examining (5.7.18). To this end, upon taking
the functional derivative of h 0C j a .x/ j 0 i, as given in (5.7.13), with respect to

i•=•c .x/ and multiplying it by c a , gives
h 0C j j .x/ j 0 i D i Tr Œ SC .x; x I e0b A/ h 0C j 0 i

Z
.dx 0 / SC ab .x; x 0 I e0b
A/ b .x 0 / c a h 0C j c .x/ j 0 i; (5.7.22)
using, in the process, the anti-commutativity of Grassmann sources, which

from (5.7.14) leads to

h 0C j .x/ .x/ C
j 0 i D i Tr Œ SC .x; x I e0b
A/ h 0C j 0 i
Z
C .dx 0 /.dx 00 / .x 00 / SC .x 00; x I e0b
A/ SC .x; x 0 I e0b
A/ .x 0 / h 0C j 0 i:
(5.7.23)
Now we multiply this equation by b

A .x/, as defined in (5.7.10), and integrate over
x, to obtain
Z Z
.dx/h 0C j.A .x/ .x/ .x//C j 0 i D i .dx/b
A .x/TrŒ SC .x; x I e0b
A/ h 0C j 0 i
Z
C .dx/ .dx 0 / .dx 00 / .x 00 / SC .x 00; x I e0b
A/ b
A.x/ SC .x; x 0 I e0b
A/ .x 0 /h 0C j 0 i:
(5.7.24)
22
The path integral version of (5.7.20) is derived in the next subsection.
In Problem 5.5 it is shown that

Z
@
SC .x ; x I e0 A/ D .dx / SC .x 00; x I e0b
00 0 b A/ b
A.x/ SC .x; x 0 I e0b
A/; (5.7.25)
@e0
which from (5.7.24) and (5.7.17) gives

Z
@
.i/ h 0C j 0 i D i .dx/ b A .x/ Tr Œ SC .x; x I e0b A/ h 0C j 0 i
@e0
@ Z
C .dx 00 / .dx 0 / .x 00 /SC .x 00; x 0 I e0b
A/.x 0 / h 0C j 0 i: (5.7.26)
@e0

We may now integrate over e0 to obtain23 b
A .x/ .i/•=•J .x/
hZ i
h 0C j 0 i D exp i .dx/.dx 0 / .x/ SC .x; x 0 I e0b
A/ .x 0 /
h Z Z e0 i
exp .dx/ de0 Tr Œ b
A.x/ SC .x; x I e0b
A/ h 0C j 0 i0 ; (5.7.27)
0
with the normalization condition

ˇ
ˇ
h 0C j 0 iˇ D 1; (5.7.28)
J ; ; D 0; e0 !0
and where the Trace operation in (5.7.27) is over gamma matrix indices, and now
h 0C j 0 i0 is given by
hiZ i
h 0C j 0 i0 D exp .dx/ .dx 0 / J .x/D .x x 0 /J .x 0 / ; (5.7.29)
2
Z
.dk/ h k k i e i k.xx /
0
D .x x 0 / D .1 / : (5.7.30)
.2/4 k2 k2 i
We note that from (5.7.22), that i Tr Œ SC .x; x I e0b

A/ , with coincident points,
which is to be applied to h 0C j 0 i, is nothing but the vacuum expectation value of
the current j .x/, (for ; D 0), up to the factor e0 , and must, by itself, be gauge
invariant under the replacement of b A .x/ by bA .x/ C @ .x/. It is studied in detail
in Sect. 3.6 and Appendix IV at the end of the book, and may be explicitly spelled
out by simply making the obvious substitution SC .x; x 0 I e0 A/ ! SC .x; x 0 I e0bA/ in
23
The expressions in (5.7.20) and (5.7.27) are appropriately referred to as generating functionals.
As we have no occasion to deal with subtleties in defining larger vector spaces to accommodate
covariant gauges, we will not go into such technicalities here.
there, for general e0 . In Problem 5.6, it is shown that the constraint equation (5.7.7)
is automatically satisfied, as expected and as it should.
Finally, we note that (5.7.27) may be also simply rewritten as
hZ i
A/ .x 0 / h 0C j 0 i ;
(5.7.31)
where h 0C j 0 i represents the full QED theory with no external electron lines,
and involves all the closed Fermion loops, with or without external photon lines, to
all orders of the theory.
Before closing this section, we note that the analysis carried out through (3.6.7),
(3.6.8) and (3.6.9), as applied to (5.7.3)/(5.7.4), shows that
h i 1
@ j .x/ D i e0 .x/.x/ .x/ .x/ ; j .x/ D e0 Œ ; : (5.7.32)
2
That is, in the absence of external Fermi sources, the current j .x/ is conserved.
In the next section, we examine the explicit expression for h 0C j 0 i in (5.7.27),
in some detail in view of applications in QED.
5.7.2 From the Differential Formalism to the Path Integral

Expression for h 0C j 0 i
From (2.6.18), (2.6.19), (2.6.23), we explicitly have

Z @

exp i SC D N11.D D/ exp i C m0 C C ;
i
(5.7.33)
Z

@
N1 D .D D/ exp i C m0 ; (5.7.34)
i
with SC now defined in terms of m0 .

On the other hand, in the momentum description, the free photon propagator, in
a covariant gauge specified by the parameter , is from (5.3.9) given by
Q Q 1
D .Q/ D .1 / 2 ; (5.7.35)
Q Q2 i
from which its inverse follows to be

1
D1
.Q/ D Q 2
1 Q Q ; D1 2 2
.Q / D .Q / D : (5.7.36)

We may now refer to (2.6.31) to write

hiZ i
exp.dx/ .dx 0 / J .x/D .x x 0 /J .x 0 /
2
Z h 1 i
1
D N21 .Da/ exp i a Œ C @ @ @ @ a C a J ;
2
(5.7.37)
Z

1 1
N2 D .Da/ exp i a Œ C @ @ @ @ a : (5.7.38)
2
where it is understood that .Da/ involves a product over the indices of a as well.
By formally integrating by parts .f D @ a @ a /, we obtain
Z

1 1 i
.dx/ a Œ C @ @ @ @ a
2
Z h 1 i
1
D .dx/ f .x/ f .x/ @ a .x/ @ a .x/ : (5.7.39)
4 2
At this stage, we may introduce a scalar field ' and, up to an overall unimportant
multiplicative factor, write
h 1 i Z h i
expi @ a @ a D .D'/ ı.@ a '/ exp i ' @ a C ' 2 :
2 2
(5.7.40)
Finally from (5.7.20) and (5.7.21) we thus obtain

Z Z
1
h 0C j 0 i D .D D/ .Da/ .D'/ ı.@ a '/ exp Œ i .dx/ Lc .x/;
N1 N2
(5.7.41)
up to an overall unimportant multiplicative constant. Here Lc is the Lagrangian

density, including the external sources J ; ; , obtained from the one in (5.7.1)
upon carrying out the substitutions
! ; ! ; A ! a ; (5.7.42)
and with the auxiliary field ! ', where we recall that ; are Grassmann
variables, i.e., are anti-commuting.
This equation is quite interesting as it explicitly shows the gauge constraint via
the delta functional ı.@ a '/. Needless to say, it is far easier to apply the
differential formalism, given in (5.7.20), involving functional differentiations, than
carrying out the functional integrations in (5.7.41), say, in a power series in e0 .
5.8 Low Order Contributions to lnh 0C j 0 i
In the present section we obtain low order contributions to ln h 0C j 0 i in e0 in

QED. More precisely, upon writing
h 0C j 0 i

eiW D exp i a0 C a1 e0 C a2 e20 C a3 e30 C : : : ; (5.8.1)
h 0C j 0 ijJ ;; D 0
we determine the coefficients a1 ; a2 ; a3 , containing a wealth of information on

QED, consisting only of connected components of the theory, that is, having at
least one propagator connecting any of its subparts. And from the normalization
condition in (5.8.1), only so-called diagrams with external lines, connected to their
respective sources, occur. These coefficients are analyzed and applied in the next
couple of sections and their physical consequences are spelled out. The power of the
formalism, is that all correct multiplicative factors in integrals, describing various
components of the theory, such as physical processes, occur automatically and no
guess work is required about such numerical factors, as powers of ’s, and proper
normalization constants.
To the above end, we recall the basic equations needed for determining the above
coefficients: (see (5.7.27)), .b
A .x/ D .i/•=•J .x//
hZ i
A/ .x 0 /
h Z Z e0 i
exp .dx/ de0 TrŒ b
A.x/ SC .x; x I e0b
A/ h 0C j 0 i0 ; (5.8.2)
0
Z
i

h 0C j 0 i0 D exp .dx/.dx 0 /J .x/D .x x 0 /J .x 0 / ; (5.8.3)
2
Z
A/ D SC .x x 0 / C e0 .dx 00 / SC .x x 00 / b
SC .x; x 0 I e0b A.x 00 /SC .x 00 ; x 0 I e0b
A/:
(5.8.4)
The latter may be expanded as follows
Z
SC .x; x 0 I e0b
A/ D SC .x x 0 / C e0 .dx1 / SC .x x1 / b
A.x1 / SC .x1 x 0 /
Z
C e20 .dx1 / .dx2 / SC .x x1 / b
A.x1 / SC .x1 x2 / b
A.x2 / SC .x2 x 0 /
Z
C e30 .dx1 / .dx2 / .dx3 / SC .x x1 / b
A.x1 / SC .x1 x2 / b
A.x2 / SC .x2 x3 /
b
A.x3 / SC .x3 x 0 / C :
(5.8.5)
5.8 Low Order Contributions to lnh 0C j 0 i 249
Also we have the expansion24

Z Z e0
.dx/ de0 Tr Œ b
A.x/SC .x; x I e0b
A/
0
Z
e2 .dQ1 / .dQ2 /
D 0 .2/4 ı .4/ .Q1 C Q2 / b
A1 .Q1 / b
A 2 .Q2 / L1 2 .Q1 ; Q2 / C
2 .2/4 .2/4
Z
e20
.dx1 / .dx2 / K 1 2 .x1 ; x2 / b
A1 .x1 / b
A 2 .x2 / C ; (5.8.6)
2
as given through (3.6.26), (3.6.27), (3.6.28) and (3.6.29), and worked out in a gauge
invariant manner, with (see Eqs. (3.6.30), (3.6.31) and (3.6.32))
L1 2 .Q ; Q/ D Œ 1 2 Q2 Q1 Q 2

Z Z 1 !
i 1 .dp/ 1 h Q2 i
C 6 dz z.1 z/ ln 1 C 2 z.1 z/ : (5.8.7)
12 2 i 2 Œp2 C m2 2 2 0 m
With expansions given above in (5.8.5) and (5.8.6), it is straightforward to

obtain iW, up to third order in e0 , as indicated in (5.8.1) involving external
sources.
The coefficients a1 ; a2 ; a3 are worked out in Problem 5.7, and are given in detail
through
a0
Z Z
1
D .dx/ .dx 0 / .x/ SC .xx 0 / .x 0 / C .dx/.dx 0 / J .x/ D .xx 0 /J .x 0 /;
2
(5.8.8)
a1
Z

D .dy/.dy 0 /.dx1 /.dz/ .y/ SC .y x1 / SC .x1 y 0 / .y 0 / D .x1 z/ J .z/;
(5.8.9)
a2
(Z
h i
D .dx/.dx 0 /.dx1 /.dx2 / .x/ SC .x x1 / 1 SC .x1 x 2 / 2 SC .x 2 x 0 / .x 0 /
24
Note that Rthe first order
contribution involves the expression: Tr Œ SC .0/ D
Tr Œ .dp/=.2/4 p =. p2 C m20 / D 0, as it has an odd integrand in p.
Z )
h i
1 2
.i/D1 2.x1 x 2 /C .dz1 /.dz2 /D1 1 .x1 z1 /J .z1 /D 2 2 .x2 z2 /J .z2 /
( Z
1 h i
C .dy1 /.dy2 /.dx1 /.dx2 /.dy 01 /.dy 02 / .y1 / SC .y1 x1 / 1 SC .x1 y 01 / .y 01 /
2
) ( Z
h i i
2 0 0
D1 2 .x1 x2 / .y 2 / SC .y 2 x 2 / SC .x 2 y 2 / .y 2 / C .dx1 /.dx2 /
2
)
.dz1 /.dz 2 / K 1 2 .x1 ; x 2 / D1 1 .x1 z1 / J 1 .z1 / D 2 2 .x 2 z 2 / J 2 .z 2 / ;
(5.8.10)
a3
(Z
h
D .dy1 /.dy 01 /.dx1 /.dx2 /.dx3 / .y1 /SC .y1 x1 / 1 SC .x1 x2 / 2 SC .x2 x3 /
ih Z
3 SC .x3 y 01 / .y 01 / .i/ .dz/ D1 2 .x1 x2 /D 3 3 .x3 z/J 3 .z/

C D1 3 .x1 x3 /D 2 2 .x2 z/J 2 .z/ C D 2 3 .x2 x3 /D1 1 .x1 z/J 1 .z/ C
Z )
i
1 2 3
.dz1 /.dz 2 /.dz 3 /D11 .x1 z1 /J .z1 /D 2 2 .x2 z2 /J .z2 /D 3 3 .x3 z3 /J .z3 /
(Z
C .dy1 /.dy 01 /.dy 2 /.dy 02 /.dx1 /.dx 2 /.dx3 /
h i
.y1 /SC .y1 x1 / 1 SC .x1 y 01 /.y 01 /
h i
.y 2 /SC .y 2 x2 / 2 SC .x2 x3 / 3 SC .x3 y 02 /.y 02 /
Z )

.dz/ D1 2 .x1 x2 /D 3 3 .x3 z/J 3 .z/CD1 3 .x1 x3 /D 2 2 .x 2 z/J 2 .z/
( Z
i h i
C .dy/.dy 0 /.dx1 /.dx 2 /.dx3 /K 1 2 .x1 ; x2 / .y/SC .yx3 / 3 SC .x3 y 0 /.y 0 /
2
Z )
h i
2 1
.dz/ D1 3 .x1 x3 /D 2 2 .x2 z/J .z/ C D 2 3 .x2 x3 /D1 1 .x1 z/J .z/ ;
(5.8.11)
5.8 Low Order Contributions to lnh 0C j 0 i 251
where K .x; y/ is defined in (5.8.6), with

Z
.dp/ p C m0 ip .xx 0 /
SC .xx 0 / D e ; (5.8.12)
.2/4 p2 C m20 i
Z
.dk/ h k k i eik.xx /
0
0
D .xx / D .1 / 2 : (5.8.13)
.2 /4 k k2 i
Remember that there are only connected parts in Eqs.(5.8.10) and (5.8.11). For
example, there is no D1 2 .x1 x 2 / within the very last square brackets on the last
line of (5.8.11) which will otherwise leave Œ .y/ SC .yx3 / 3 SC .x 3 y 0 / .y 0 /
disconnected from the other part multiplying this factor.
To extract transition amplitudes for physical processes from the above expres-
sions, we here recall the amplitudes of emission and detection by the external
sources (acting as emitters or detectors) of the particles involved:
s
m d3 p
h p ; j 0 i D Œ i u.p; / . p/ ; (5.8.14)
p 0 .2/3
s
m d3 p
h p ; C j 0 i D Œ i .p/ v.p; / ; (5.8.15)
p 0 .2/3
s
m d3 p
h 0C j p ; i D Œ i . p/ u.p; / ; (5.8.16)
p 0 .2/3
s
m d3 p
h 0C j p ; Ci D Œ i v.p; / .p/ ; (5.8.17)
p 0 .2/3
s
d3 k
h k j 0 i D i J .k/ e .k/ ; (5.8.18)
.2 /3 2k0
s
d3 k
h 0C jk i D i e .k/ J .k/ : (5.8.19)
.2 /3 2k0
according to (3.3.39), (3.3.40), (3.3.41), (3.3.42), (5.5.12),(5.5.13).

In extracting transition amplitudes, the sources are eventually withdrawn, i.e.,
one finally sets D 0; D 0; J D 0. Here we recall, for example, that
h p ; j 0 i denotes the amplitude that the source emits an electron, thus acting
as an emitter, and h 0C j p ; i denotes the amplitude that a source absorbs an
electron, thus playing the role of a detector. Don’t let the notations in these equations
scare you. They provide a simple worry free formalism for getting correct numerical
factors such as .2/3 , and so on, as mentioned above, and facilitate further the
analysis in obtaining transition amplitudes of physical processes.
The moral of introducing these amplitudes of emissions and detections, above

is the following. Using the explicit expressions in (5.8.8), (5.8.9), (5.8.10) and
(5.8.11), with the sources appropriately arranged causally to reflect the actual
process where the particles in question are emitted and detected by these sources,
the transition amplitudes for arbitrary processes are then simply given by the
coefficients of these amplitudes of emissions followed by detection. This may be
simply represented as follows:
0+ | particles (out) detected by sources

particles (out) on their way to detection|particles (in) emitted by sources
particles (in) emitted by sources|0 − ,
VACUUM VACUUM
DETECTORS TRANSITION AMPLITUDE EMITTERS (5.8.20)
as obtained from (5.8.8), (5.8.9), (5.8.10) and (5.8.11), corresponding to what

happens experimentally, and where the transition amplitude is given by
h particles (out)jparticles (in)i

D h particles (out) on their way to detectionjparticles (in) emitted by sourcesi:
(5.8.21)
This ingenious method which allows one to quickly extract transition amplitudes
with correct multiplicative numerical factors is due to Schwinger, and no guessing
is required as to what correct multiplicative factors should be in the formalism.
Before closing this section, we note that the interactions in processes are
mediated by the propagators. In the momentum description, the photon propagator
of momentum Q develops a singularity at Q2 D 0, i.e., for Q as a light-like
vector, corresponding to the masslessness-shell constraint of a real, i.e., detectable,
photon. In general, we may have cases for which Q2 ¤ 0 corresponding to
a space-like or a time-like Q. In the latter two cases, the four momentum of
the photon is said to be off the masslessness-shell condition Q2 D 0, and the
photon thus cannot be detected by a detector. It is then referred to as a virtual
particle since the photon does not have the appropriate relation between energy and
momentum to be detectable. Let us investigate the meaning of this last inequality.
For Q space-like, i.e., for Q2 > 0, for example, we have Q2 > .Q0 /2 , which
means the photon lacks the appropriate energy, in comparison to its momentum,
to be detectable. On the other hand, for Q2 < 0 i.e., for Q2 < .Q0 /2 , the
photon has a surplus of energy over its momentum to be detectable. To see how
actually such a virtual photon arises, with a space-like or time-like momentum, in
5.9 Basic Processes 253
Fig. 5.3 (a) Consider the diagram describing the scattering process of an electron, experiencing a
change of its three-momentum, via the exchange of a (virtual) photon with the remaining part of
the diagram (denoted by the shaded area). Conservation of momentum implies that the momentum
of the virtual photon is space-like. (b) Consider an electron and a positron annihilating, e.g. in the
c.m., into a (virtual) photon. The momentum of the virtual photon is time-like. (c) Consider the
scattering of an electron and a photon to a (virtual) electron. The momentum p of the virtual
electron is off the mass-shell satisfying the relation p2 < m2 . (d) Consider an electron becoming
virtual in the scattering of the electron with the emission of a photon. The momentum p of the
virtual electron is off the mass shell satisfying the relation p2 > m2 . The photon is denoted by
a wavy line, while the electron (positron) by a solid one. A virtual particle which has too much
energy or not enough energy to be on the mass shell, as the case may be, may, respectively, give
off energy to another particle or absorb energy from another one and may eventually emerge as a
real particle in an underlying Feynman diagram description of fundamental processes
the light of these observations, consider the diagrams depicting some processes in
Fig. 5.3.25
Now let us proceed, move to the next section, and see how things work out,
compute transition amplitudes by applying the above equations, and witness the
wealth of information stored in them.
5.9 Basic Processes
Experimentally, scattering processes are quantified and their likelihood of occur-

rence are determined in terms of what one calls the cross section. This quantity
arises in the following manner. Consider the scattering, say, of two particles of
masses m1 and m2 leading finally to an arbitrary number of particles that may
be allowed by the underlying theory. The transition probability per unit time
Probjunit time for the process to occur is first calculated. Consider a frame in which,
say, the particle m2 , is (initially) at rest. Such a frame may be referred to as
the laboratory or target (TF) frame. To determine the likelihood of the process to
occur, and simultaneously obtain a measure of the interaction of the two particles,
Probjunit time , in turn, is compared to, i.e., divided by, the probability per unit time
25
These processes are assumed to involve no external sources or external potentials.
Fig. 5.4 Diagram which

facilitates in defining the m1
incident flux
×
A
per unit area, denoted by F, that particle m1 crosses at the position of particle of
mass m2 , represented by in Fig. 5.4, as if the latter particle is absent, and hence
no interaction is involved.
The probability per unit area per unit time F, defined above, is called the incident
flux. Most importantly, the ratio defined above leads, upon integration, to a Lorentz
invariant expression for the cross section, and thus may be computed in any inertial
frame. By definition, the unit of cross section is that of an area and, intuitively, the
cross section provides an effective area for the process to occur. The differential
cross section d is then simply defined by the above mentioned ratio, i.e., by
Probjunit time per incident flux F: d D Probjunit time =F.
The probability of finding, the particle of mass m1 , of speed v, within a cube of
cross sectional area A and width vT, during a time T, is given by A v T=V, where
V denotes the volume of the 3D space in which the scattering process occurs. Thus
the probability that this particle crosses a unit area per unit time, at the position
of particle m2 , as if the latter is absent (see Fig. 5.4), is given by
v 1 jp1 TF j
D F: (5.9.1)
V V p01
For the initial particle of mass m2 at rest, we note that the above expression
allows us to write
. p01 p02 V/ F D m2 jp1 TF j: (5.9.2)
On the other hand, for arbitrary

qsharp momenta p1 , p2 , of the initial particles, the
Lorentz invariant expression . p1 p2 /2 m21 m22 , may be computed in any frame,

and, in particular, in the TF frame to be . p1 p2 p1 p2 /
q
. p1 p2 /2 m21 m22 D m2 jp1 TF j; (5.9.3)
The transition probability per unit time of a given process that emerges from the
theory for the scattering, say, of two particles into an arbitrary number of particles,
is given by an expression of the form26
ˇ 0 0 0 0 ˇ
ˇhp ; p ; : : : jp1 1 ; p 2 2 iˇ2
1 1 2 2 unit time
X
D .˘j d!j0 / jM j2 .2/4 ı .4/ p j0 p1 p2 V d!1 d!2 ; (5.9.4)
j
d!j D d3 pj =Œ.2/3 2p j0 ; (5.9.5)
where jM j2 is an invariant quantity, the symbol stands for any label needed to
specify a particle, in addition to its momentum, and V denotes the volume of the
3D space in which the scattering process occurs, as before. With the momenta of the
initial particles prepared to have sharp values, box normalization implies to replace
each of d3 p1 =.2/3 , d3 p 2 =.2/3 , corresponding to the initial particles, by 1=V.
The differential cross section then takes the form
P
.2/4 ı .4/ 0
j p j p1 p2
d D .˘j d!j0 / jM j2 ; (5.9.6)
4 p 10 p 02 VF
which from (5.9.2) and (5.9.3), may be rewritten in the form

P
.2/4 ı .4/ 0
j p j p1 p2
d D .˘j d!j0 / jM j2 q ; (5.9.7)
4 . p1 p2 /2 m21 m22
and is independent of V.
To obtain an expression of the differential cross section in the center-of-
momentum (CM) frame, it is convenient to introduce the invariant variable s D
. p1 C p2 /2 . Then some algebra gives
q q
1
. p1 p2 /2 m21 m22 D s2 2 .m21 C m22 / s C .m21 m22 /2 : (5.9.8)
2
p
In the center-of-momentum (CM) frame, p1 C p2 D 0 and p1 0 C p2 0 D s,
which by evaluating the expression on the left-hand side of (5.9.8), in the CM frame,
26
The momentum conserving delta function occurs as follows. Invoking translational invari-
ance, and the Hermiticity of the total momentum P which may equally operate to right
or left in hp 01 10 ; p 02 20 ; : : : jPjp1 1 ; p2 2 i, leading to the two equal expressions: . p1 C
p2 /hp 01 10 ; p 20 20 ; : : : jp1 1 ; p2 2 i D . p 01 C : ::/hp 01 10 ; p 20 20 ; : : : jp11 ; p2 2 i. Hence upon subtract-
ing one expression from the other, we obtain p1 Cp2 . p 01 C: : :/ hp 10 10 ; p 20 20 ; : : : jp1 1 ; p2 2 i D
0. On the other the relation x f .x/ D 0 implies that f .x/ is proportional to ı.x/.
allows us to write
q
1
jp1 CM j D p s2 2 .m21 C m22 / s C .m21 m22 /2 : (5.9.9)
2 s
Consider a process involving initially and finally two particles of masses

.m1 ; m2 /, .m 01 ; m 02 /, and momenta . p1 ; p 2 /; . p 10 ; p 02 /, respectively. We explicitly
have s D . p1 C p 2 /2 D . p 10 C p 20 /2 , and in the CM frame we also have
0 0 p
p 01 C p 02 D 0, p1 0 C p2 0 D p1 0 C p 2 0 D s. From the expression of jp1 CM j
in (5.9.9), we similarly have
q
1 0 0 0 0
jp 01 CM j D p s2 2 .m12 C m22 / s C .m12 m22 /2 : (5.9.10)
2 s
In reference to CM frame, the following useful expression then emerges

Z Z
1 jM j2 d3 p 01 d3 p 02 .4/ 0
d D p ı . p 1 C p 20 p1 p 2 /;
64 2 jp1 CM j s p 01 0 p 20 0
Z
1 jp 01 j2 djp 01 j p
D d˝ CM jM j2 p ı. p 10 0 C p 02 0 s /
64 2 p 10 0 00
p 2 jp1 CM j s
Z
1 jM j2 jp 01 CM j
D d˝ CM ; (5.9.11)
64 2 s jp1 CM j
q
where in the first line, we have replaced . p1 p2 /2 m21 m22 in (5.9.11) by
p
jp1 CM j s, referring to (5.9.8), (5.9.9) and (5.9.10). In the second line we have
merely integrated over p 02 . In the last line we have used the fact that
p ı.jp 0 j jp 01 CM j/
ı. p 10 0 C p 20 0 s / D 0 1p ; (5.9.12)
jp 1 CM j s=. p 10 0 p 02 0 /
to integrate over jp 01 j. Finally, note that d˝ CM D sin d d, where is the
angle between p1 CM and p 01 CM , and jp 01 CM j, jp1 CM j, may be replaced by the
expressions on the right-hand sides of (5.9.9), and (5.9.10). Thus in the CM frame,
we have the following expression for the differential cross section
d 1 jM j2 jp 01 CM j
D : (5.9.13)
d˝CM 64 2 s jp1 CM j
For future reference, we introduce the Mandelstam variables associated with a

scattering process: p1 p 2 ! p 10 p 20 , including the invariant s defined above:
s D . p1 C p 2 /2 D . p 10 C p 20 /2 ; t D . p1 p 01 /2 D . p 20 p 2 /2 ;
2 2
u D . p1 p 20 /2 D . p 10 p 2 /2 ; s C t C u D m21 C m22 C m 0 1 C m 0 2 :
(5.9.14)
In particular, in the CM frame, d t D 2jp 01 CM j jp1 CM jd cos , which

from (5.9.13) leads to
d ˇˇ 1 1
ˇ D jM j2 : (5.9.15)
d t CM 64 s jp1 CM j2
In this section, we consider several processes of interest to low orders, including

polarizations correlations studies of outgoing particles in such processes which have
been quite important in recent years. The derivations underlying the scattering pro-
cesses are obtained directly from the explicit expression obtained for h 0C j 0 i in
the previous section, to lowest orders. The formalism is powerful enough that all the
numerical factors in transition amplitudes appear directly as well as various relations
concerning the statistics of the underlying particles, and no guessing is necessary.
For the convenience of the reader we here recall the transition made between box
normalization and that of the infinite extension as arising from the Fourier transform,
in the complex form, and that of the Fourier integral, used often in scattering theory:
1 < x < 1; L=2 x L=2;

1 < k < 1; k D .2n=L/; n D 0; ˙1; ˙2; : : : ;
R 1 dk ikx R L=2
f .x/ D 1 2 e ef .k/; f .x/ D L1 L=2 dk eikxe f .k/;
R1 R L=2
e ikx
f .k/ D 1 dx e f .x/; e ikx
f .k/ D L=2 dx e f .x/;
R1 R L=2
.2/ ı.k/ D 1 dx e ; ikx
.L/ ık;0 D L=2 dx eikx ;
dn k=.2/n ; 1=L1 : : : L n ;
n .n/ .n/
.2/ ı .k/ : .L1 : : : L n / ık;0 :
5.9.1 e e ! e e , eC e ! eC e
We consider the processes e e ! e e , eC e ! eC e , referred to,

respectively, as Møller scattering and Bhabha scattering, to second order. To this
Fig. 5.5 Diagram

corresponding to Eq. (5.9.16)
including the presence of the
external sources denoted by
the half circles. Such a
diagram describes several
processes
order, we may replace the parameters e0 ; m0 in the lagrangian density by their

zeroth order, i.e., by the corresponding physical values e; m. These processes are
described in terms of the expression within the curly brackets in the second term of
a2 in (5.8.10)27 multiplied by i e2 as given in the exponent in (5.8.1) (see Fig. 5.5)
( Z
ie2

.dy1 /.dy 2 /.dx1 /.dx2 /.dy 01 /.dy 02 / .y1 / SC .y1 x1 / 1 SC .x1 y 01 / .y 01 /
2
)

2 0 0

D1 2 .x1 ; x 2 / .y 2 / SC .y 2 x 2 / SC .x 2 y 2 / .y 2 / :
(5.9.16)
We may write the source as .x/ D 1 .x/ C 2 .x/, where 1 .x/ is switched
on and then off in the remote past after the electrons are emitted, while 2 .x/
is switched on in the distant future and then switched off after the outgoing
electrons are detected (absorbed), with the interaction taking place later in time
than the emissions and earlier than detections. Thus for the first scattering process
in question, only the term involving the part 1 .x/ of .x/, and the part 2 .z/ of
.z/ will contribute.28
For all y 0 > x 0 , we have (see (4.1.17))
Z Z X 3
d k m i kx
.dy/ 2 .y/SC .y x/ D 3 k0
e Œ i 2 . p/ u.k; / u.k; /;

.2/
(5.9.17)
27
This expression involves also the process eC eC ! eC eC .
28
For the process eC eC ! eC eC , only the term involving the pair 2 .x/ and 1 .x/ will
contribute.
and similarly for all x 0 > y 0 ,

Z Z X 3
d p m i px
.dy/SC .x y/1 .y/ D e u.p; / Œ i u.p; / 1 . p/ :

.2/3 p0
(5.9.18)
Accordingly, in reference to the first process in question, we obtain for (5.9.16)
Z X
i e2
d!.k01 / d!.k02 / d!.k1 / d!.k2 / 16 m4 .2/4 ı 4 .k 02 C k 01 k1 k2 /
2 0 0
2 ;2 ;1 ;1
./Œ i 2 .k 02 / u.k02 ; 20 / Œ i 2 .k 01 / u.k01 ; 10 / Œi u.k2 ; 2 / 1 .k2 / Œ i u.k1 ; 1 / 1 .k1 /
u.k02 ; 20 / 1 u.k2 ; 2 / D1 2 .k2 k 02 / u.k01 ; 10 / 2 u.k1 ; 1 /;
(5.9.19)
where the overall () sign arises because we have (anti-)commuted 2 .k 01 / and
1 .k2 /.
Now suppose that the momenta and spins of the initial electrons are prepared to
be .p1 ; 1 /; .p2 ; 2 /. We encounter the two possibilities: (k1 D p1 ; k2 D p2 ) or
(k1 D p2 ; k2 D p1 ) (and similarly for the corresponding spins). Clearly, since one
has the integrations over .k02 ; k01 /, and sums over . 01 ; 02 /, we are free to carry out
the relabelings .p 01 ; 01 / , .p 02 ; 02 /, we have the complete symmetric relation29
without change of the overall sign in (5.9.19) by simultaneously commuting
1 .k1 /; 1 .k2 / and .k 01 /; .k 02 /. This leads simply to multiplying the expression
in (5.9.19) by a factor of 2.
We have to be also careful when considering the outgoing momenta.30 If the
momenta and spins of the outgoing (two identical particles) electrons are .p 02 ; 02 /,
.p 01 ; 01 / with p 01 ¤ p 02 , then the integral, say over k01 will “meet” both
values .p 02 ; p 01 /.
Similarly in the case for the integral over k02 , the product Œ i 2 . p 20 / u.p 02 ; 02 /
Œ i 2 . p 01 / u.p 01 ; 01 / will occur twice with .p 02 ; 02 / $ .p 01 ; 01 /.
Accordingly, for the process, in question, with the momenta of the electrons
taking values within infinitesimal intervals around the just specified non-coincident
values, and for the values taken by their respective spins, we obtain
i e2 d!.p 01 / d!.p 02 / d!.p1 / d!.p2 / 16 m4 .2/4 ı 4 . p 02 C p 01 p1 p 2 /

Œ i 2 . p 02 / u.p 02 ; 02 / Œ i 2 . p 01 / u.p 01 ; 10 / Œ i u.p 2 ; 2 / 1 . p 2 / Œ i u.p1 ; 1 / 1 . p1 /
h
u.p 02 ; 02 / 1 u.p 2 ; 2 / D1 2 . p1 p 01 / u.p 01 ; 01 / 2 u.p1 ; 1 /
i
u.p 01 ; 01 / 1 u.p2 ; 2 / D1 2 . p1 p 02 / u.p 02 ; 02 / 2 u.p1 ; 1 / :
(5.9.20)
29
Also note that D .k2 k 02 / D D .k1 k 01 /, and the symmetry of D in .; /.
30
On the other hand, if you fix the final momenta and spins first, then the argument is reversed.
Here we recall that we have two identical particles (electrons) in the final state,
and that the minus sign between the two terms within the square brackets, in
the last line, arises because we have (anti-)commuted Œ i 2 . p 10 / u.p 01 ; 01 / and
Œ i 2 . p 02 / u.p 02 ; 02 / , in the process of derivation, and thus it emerges as a
consequence of the anti-commutativity of the Grassmann sources. The Fermi-Dirac
statistics is automatically recovered (Fig. 5.6).
Upon comparing the above expression in (5.9.20) with the term in a unitarity sum
(with connected terms in scattering) (see also (5.8.20)):
h 0C jp 20 02 ; p 10 01 i h p 10 01 ; p 02 20 jp1 1 ; p2 2 i h p22 ; p1 1 j 0 i; (5.9.21)
we may infer from (5.9.20), that the amplitude of scattering in question is given by
q q p p
h p 10 01 ; p 20 02 jp1 1 ;p 2 2 i D i e2 2md! 01 2md! 02 2md! 1 2md! 1
h i
.2/4 ı .4/ . p 20 C p 10 p1 p2 / (5.9.22)
where
h i h
D u.p 02 ; 02 / 2 u.p 2 ; 2 / D 2 1 . p 2 p 02 / u.p 01 ; 10 / 1 u.p1 ; 1 /
i
u.p 01 ; 01 / 2 u.p 2 ; 2 / D 2 1 . p 2 p 01 / u.p 02 ; 02 / 1 u.p1 ; 1 / ;
(5.9.23)
(a) (b)
p2 p1 p1 p2 p2 p1 p1 p2
p2− p 2 p2− p 1 p2− p 2 p2− p 1
p2 p1 p2 p1 p2 p1 p2 p1
Fig. 5.6 (a) Møller scattering: e . p1 / e . p 2 / ! e . p01 / e . p 02 /. The Fermi-Dirac statistics

is automatically taken care of by the formalism. (b) Bhahba scattering: e . p1 / eC .q/ !
e . p1 / eC .q/, obtained from the Møller one, in the process, by the substitutions p02 ! q,
p 2 ! q 0 , u.p 02 ; 02 / ! v.q; /, u.p 2 ; 2 / ! v.q0 ; 0 /, in reference to (5.9.41), with an
overall change of sign of the amplitude
p p
h p2 2 ; p1 1 j 0 iDŒ i u.p2 ; 2 / 1 . p2 / 2m d! 2 Œ i u.p1 ; 1 / 1 . p1 / 2md!1 ;
(5.9.24)
q q
h 0C jp 20 20 ; p 01 01 iDŒ i 2 . p 20 /u.p 02 ; 02 / 2md! 02 Œ i 2 . p 10 / u.p 01 ; 01 / 2m d! 01 ;
(5.9.25)
and the latter two equations denote, respectively, the amplitudes of emissions and
detection of these particles. (See (5.8.14) and (5.8.16)).
From the Dirac equations
u.p; /. p C m/ D 0; . p C m/u.p; 1 / D 0; (5.9.26)
we obtain
. p2 p 20 / u.p 02 ; 02 / u.p2 ; 2 / D 0; . p2 p 10 / u.p 01 ; 01 / u.p2 ; 2 / D 0;

(5.9.27)
and, in turn, conclude that only the part of the photon propagator D proportional
to will contribute to the amplitude in (5.9.22), thus establishing its gauge
invariance.
Using the elementary property31 Œ.2/4 ı .4/ . p/2 D VT.2/4 ı .4/ . p/, where V is
the normalization volume, T is the total interaction time, we obtain for the transition
probability per unit time for the process for unpolarized electrons
d3 p 02 d3 p 01 X ˇh iˇ2
4 1 ˇ ˇ
Probjunit time D e4 16 m ˇ : ˇ
.2/3 2p 0 02 .2/3 2p 0 01 4 spins
d3 p 2 d3 p1
.2/4 V ı .4/ . p 20 C p 10 p1 p 2 /: (5.9.28)
.2/3 2p 2 .2/3 2p01
0
where the 1=4 factor arises because one is averaging over the initial spins of the
ingoing electrons.
For sharp initial momenta p1 ; p 2 , box normalization versus continuum
normalization,
P via Fourier series
R (in complex form) versus Fourier transform:
.1=V/ p . : / , .1=.2/3 / d3 p. : /, allows us to replace .d3 p1 =.2/3 / .d3 p2 =
.2/3 / by 1=V 2 . If F denotes the incident flux, then according to (5.9.2)
and (5.9.3),
ˇ p
ˇ
p01 p 02 V F ˇ D . p1 p 2 /2 m4 : (5.9.29)
TF
31
R This property is usually attributed to Fermi, and follows from the formal equality ı .4/ . p/jpD0 D
.dx/=.2/4 D VT=.2/4 .
The cross section for the process for unpolarized electrons, is then given,
from (5.9.6) and (5.9.7), by the Lorentz invariant expression
Z X ˇh iˇ2 ı .4/ . p 0 C p 0 p p /
e4 d3 p 02 d3 p 01 4 1 ˇ ˇ 1 2
D 0 0
16 m ˇ ˇ p2 1
;
4 2 2p 0 2 2p 0 1 4 spins 2
4 . p1 p 2 / m 4
(5.9.30)
R
RThe Lorentz invariance of this expression is evident since d3 p=2p0 Œ D
.dp/ ™. p0/ ı. p2 C m2 /Œ .
In reference to (5.9.7), jM j2 , and its average jM j2 , may be written as
ˇh iˇ2 1 X ˇh iˇ2
ˇ ˇ ˇ ˇ
jM j2 D e4 16 m4 ˇ : ˇ ; jM j2 D e4 16m4 ˇ : ˇ ; (5.9.31)
4 spins
After some algebra, the summation over spins leads to (see Problem 5.8) . p1 p 2 D
p1 p 2 /

1 X ˇˇh iˇˇ2 8 . p1 p 2 /2 C . p1 p 02 /2 C 2m2 . p1 p 10 C m2 /
ˇ : ˇ D
4 spins .2m/4 . p 2 p 02 /4

4
m C . p1 p 2 /2 C m2 . p1 p 2 C p1 p 01 C p1 p 20 /
C2
. p 2 p 02 /2 . p 2 p 01 /2

!
. p1 p 2 /2 C . p1 p 01 /2 C 2m2 . p1 p 20 C m2 /
C :
. p 2 p 01 /4
(5.9.32)
In the CM frame,
p1 D p 2 ; p 01 D p 02 ; p1 0 D p 2 0 D p 01 0 D p 01 0 ; (5.9.33)
1
p1 p 01 D 2 ˇ 2 m2 cos D 2 1/ m2 cos ; 2D : (5.9.34)
.1 ˇ 2 /
Using the general expression in (5.9.13), worked out in terms of CM variables,

remembering that jp 01 j=jp1 j D 1, and using (5.9.32), the following expression for
the differential cross section emerges
d ˛2 1 h 1 1 2 2 3 1 1 i
2 2
D 2 2 2 .2 1/ 2
2
C ;
d˝CM m . 1/2 sin 4 4 sin 4
(5.9.35)
where ˛ D e2 =.4/ is the fine-structure constant.
At high energies 2 1, the differential cross section takes the simple form
d ˛2 4 2
' 1 : (5.9.36)
d˝CM 4m2 2 sin2
For the second process eC e ! eC e , clearly only the following terms

contribute to (5.9.16)
n o
2 Œ 2 SC SC 1 D Œ 1 SC SC 2 C Œ 1 SC 1 D Œ 2 SC SC 2 ;
(5.9.37)
where the factor 2 arises because of the symmetry that arises between the indices
1 , 2 of the sources that occurs in (5.9.16) upon writing D 1 C 2 . Here we
note that for x 0 < x 0 0 , with x 0 arbitrarily large32
Z Z
m d3 p ip x 0 X
.dx/ 1 .x/ SC .x x 0 / D i 0 3
e 1 .p/v.p; / v.p; /;
p .2/
(5.9.38)
while for y 0 < y 0 0 , with y 0 0 arbitrarily large,

Z Z
0 0 0 m d3 p ip y X
.dy / SC .y y / 2 .y / D i e v.p; / v.p; /2 . p/ :
p 0 .2/3
(5.9.39)
Using the Grassmann character of the sources, we obtain for the amplitude in
question
n
A D i e2 Œ v.q; / v.q 0 ; 0 / D .q q 0 / Œ u.p 01 ; 10 / u.p1 ; 1 /
o
Œ u.p1 0 ; 10 / v.q; 0 / D .q C p1 / Œ v.q; / u.p1 ; 1 /
q
2m d!.p 01 / 2m d!.q 0 / 2m d!.p1 / 2m d!.q/ .2/4 ı .4/ . p 10 C q 0 p1 q/:
(5.9.40)
Upon comparison of this amplitude of Bhabha scattering, with the Møller one
in (5.9.22)/(5.9.23), we may infer that the latter amplitude is related to the former
32
Note the sign change in the exponent of the following two equations.
one by the substitutions:
p 02 ! q; p 2 ! q 0 ; u.p 02 ; 02 / ! v.q; /; u.p 2 ; 2 / ! v.q 0 ; 0 /; (5.9.41)
with an overall change of sign of the amplitude.

The above substitution rule leads, for unpolarized particles, to the following
expression for the differential cross section, at high energies, in the CM frame
d ˛2 4 2
' 2 2 2
1 cos4 : (5.9.42)
d˝CM 4m sin 2
5.9.2 e ! e ; eC e ! and Polarizations

Correlations
We consider the process e ! e , well known as Compton scattering, and the

related one of electron-positron annihilation into two photons: eC e ! . These
processes are described in terms of the expression
Z
2

ie .dx/ .dx 0 / .dx1 / .dx2 / .x/SC .x x1 / 1 SC .x1 x2 / 2 SC .x2 x 0 /.x 0 /
Z

.dz1 / .dz 2 / D11 .x1 z1 / J 1 .z1 / D 2 2 .x 2 z 2 / J 2 .z 2 / ;
(5.9.43)
which is given by the second term within the curly brackets in the first term of a2
in (5.8.10), multiplied by i e2 as given in the exponent in (5.8.1).

To describe these processes, we write D 1 C 2 , J D J1 C J2 , where 1 , J1
are switched on and then off in the remote past after the electron and the photon are

emitted, while 2 , J2 , are switched on in the distant future and then switched off
after the outgoing electron and photon are detected (absorbed), with the interaction
taking place later in time than the emissions and earlier than detections. Clearly, the
first process, is described by the explicit expression
Z
i e2 .dx/ .dx 0 / .dx1 / .dx 2 / .dz1 / .dz 2 /

2 .x/ SC .x x1 / 1 SC .x1 x 2 / 2 SC .x 2 x 0 / 1 .x 0 /
"
D11 .x1 z1 / J21 .z1 / D 2 2 .x 2 z 2 / J1 2 .z 2 /
#
1
C D1 1 .x1 z1 / J 1 .z1 / D 2 2 .x 2 z 2 / J 2 2 .z 2 / : (5.9.44)
Here z10 < x10 , z2 0 > x20 , with z1 0 , z2 0 arbitrarily large, hence invoking, in
the process, gauge invariance with respect external photon lines, we may make the
substitutions
Z Z X
d3 k
.dz1 / D1 1 .x1 z1 /J11 .z1 / ! i 0 3
e i kx 1 e1 e J1 .k/ ;
2k .2/

Z Z (5.9.45)
2 d3 k X
i kx 2
.dz 2 / D 2 2 .x 2 z 2 / J2 .z 2 / ! i e J 2 .k/ e e 2 :
2k 0 .2/3

(5.9.46)
For the initial electron and photon with respective momenta p, k, and final
momenta with p 0 , k 0 , with the latter two taking values within infinitesimal intervals
around the just specified values, the above equation gives the amplitude (Fig. 5.7)
s s
d3 p 0 d3 k 0 1 1
A D iM p .2/4 ı .4/ . p 0 C k 0 p k/;
.2/3 2p 0 0 2k 0 0 3
.2/ V 4 p 0 k 0
(5.9.47)
h . p C k/ C m . p k 0 / C m i
M D e2 2m u.p 0 ; 0 / e00 2 2
e C e 0 2 2
e00 u.p; /;
. p C k/ C m .p k / C m
(5.9.48)
1=2
where we recall that i e J1 .k/ d3 k=.2 /3 2k0 is the amplitude that the
(emitter) source J1 emits a photon of momentum k and polarization specified by
the parameter , and similarly, for the amplitude of detection of a photon (see (5.5.4)
and (5.5.5)) for a real polarization vector.
p k p
p+k + p−k
p k p
Fig. 5.7 Compton scattering showing only the momenta of the particles involved. The photon is
shown as a wavy line
We consider the initial electron at rest. Then upon squaring the momentum
conserving equation p 0 C k 0 D p k, using in the process p 0 0 C k 0 0 D m C k 0 ,
and setting k k 0 D k 0 k 0 0 cos , we obtain the well known expression relating the
photon energy before and after the collision
k00 1
D : (5.9.49)
k0 1C k0
.1 cos /
m
We consider an experiment where the electron spins are not observed. Accord-
ingly averaging over the initial spin of the electron and summing over the final spin,
we obtain
1X k 00 k0
jM j2 D e4 0 C 0 0 C 2 Œ 2 .e e00 /2 1 : (5.9.50)
2 spins k k
The p 0 integral, upon taking the absolute value squared of (5.9.47), is readily
obtained by using in the rest frame of the electron
Z Z
d3 p 0 .4/ 0
ı . p C k p k/ D .dp 0 /™. p 0 0 /ı. p 0 2 C m2 /ı .4/ . p 0 C k 0 p k/
0
2p0 0

D ™.mCk 0 k 0 0 /ı . p C k k 0 /2 Cm2 D .mCk0 k 0 0 /ı 2kk 0 2m.k0 k 00 / ;
(5.9.51)
and the k0 -integration may be now easily carried out. This leads for the differential
cross section for polarized photons, in the rest frame of the electron (TF frame), the
expression
d ˛ 2 k 0 0 2 k 0 0 k0 0 2

D C C 2 Œ 2 .e e 0/ 1 : (5.9.52)
d˝TF 4 m2 k 0 k0 k00
This is referred to as the Klein-Nishina formula.

For unpolarized photons, we average over the initial photon polarizations and
sum over the final ones, to obtain
d ˛ 2 k 0 0 2 k 0 0 k0 2

D C sin : (5.9.53)
d˝TF 2 m2 k 0 k0 k00
In the low energy limit k0 ! 0, we see from (5.9.49) that k 0 0 =k 0 ! 1, and the
above equation leads, upon integration over ˝TF , to the following cross section
8 2
D r ; (5.9.54)
3 0
referred to as Thomson scattering cross section, and r0 D ˛=m is the classical

electron radius of the order 2:8 1013 cm.
Now we turn to the process of electron-positron annihilation into two photons.

As an application, we will consider the polarizations correlations aspects of the
emerging two photons. From (5.9.43), this process is extracted from the expression
Z

i e2 .dx/.dx 0 /.dx1 /.dx 2 / 1 .x/SC .x x1 / 1 SC .x1 x 2 / 2 SC .x 2 x 0 /1 .x 0 /
Z

.dz1 /.dz 2 /D1 1 .x1 z1 /J21 .z1 /D 2 2 .x 2 z 2 /J22 .z 2 / : (5.9.55)
Here z 01 > x 01 , z 02 > x 02 , with z 01 ; z 02 arbitrarily large, x 0 < x 01 , x 0 0 < x 0

2
with x 0 ; x 0 0 arbitrarily large.
The amplitude for the process in question is readily extracted to be
s
d!.k1 / d!.k2 / 1
A D iM .2/4 ı .4/ .k1 C k2 p1 p2 /; (5.9.56)
4 p 10 p 02 V
h i
.2/ .1/ .1/ .2/
M De2 2mv.p 2 ; 2 / e 2 SC . p1 k1 / e1 C e1 SC . p1 k2 / e 2 u.p1 ; 1 /:
(5.9.57)
We work in the CM frame. To investigate the polarizations correlations of the
outgoing photons, suppose that they have emerged in the experiment perpendicular
to the direction of the electron positron pair (Fig. 5.8). Averaging over the initial
z −k
χ2
p
−p
y
χ1
k
Fig. 5.8 Photon polarization correlation experiment in eC e ! with unpolarized electron

and positron, in which the outgoing photons have emerged and moved along the x-axis as shown.
The polarization three-vectors lie in planes parallel to the yz - plane. Only one of the possible
outcomes of the experiment is shown here
spins of the electron and positron, we have in the CM frame corresponding to such
a process, with
p1 p D m ˇ.0; 1; 0/; k1 D p 0 .1; 0; 0/; (5.9.58)

.1/ .1/
e1 D .0; 0; cos 1 ; sin 1 /; e2 D .0; 0; sin 1 ; cos 1 /; (5.9.59)
.2/ .2/
e1 D .0; 0; cos 2 ; sin 2 /; e2 D .0; 0; sin 2 ; cos 2 /; (5.9.60)
.j / .j /
where e2 is obtained from e1 by the substitutions
j ! j C =2; j D 1; 2; (5.9.61)
and for 1 D 1; 2 D 1, .0 ˇ 2 1/,
Xˇˇ ˇˇ2 h .1/ .2/

.1/ .2/
.1/ .2/ .e p/.e p/ .e p/2 .e p/2 i
.1/ .2/
m2 ˇ : ˇ / 1 .e1 e 2 /2 C 4.e1 e 2 / 1 0 2 2 4 1 0 4 2 ;
spins
.p / .p /

2
D 1 cos.1 2 / 2 ˇ 2 cos 1 cos 2 ;

2
1 .1 ˇ 2 / cos.1 2 / ˇ 2 cos.1 C 2 / ; (5.9.62)
up to an unimportant numerical proportionality factor at hand for the investigation in

question. Therefore the properly normalized polarization correlations probability of
the two emerging photons with their polarization vectors making angles 1 ; 2 , as
shown above, is

2

1 cos.1 2 / 2 ˇ 2 cos 1 cos 2
P 1 ; 2 D ; (5.9.63)
2 Œ 1 C 2 ˇ 2 .1 ˇ 2 /
for all 0 ˇ 1. The normalizability of this joint distribution is verified since

P 1 ; 2 CP 1 C ; 2 CP 1 ; 2 C CP 1 C ; 2 C D 1: (5.9.64)
2 2 2 2
If the polarization of only one of the photons is measured, we have, e.g.,

P 1 ; D P 1 ; 2 C P 1 ; 2 C =2 ;
and hence

1 C 4ˇ 2 .1 ˇ 2 / cos2 1
P 1 ; D ; (5.9.65)
2 Œ 1 C 2 ˇ 2 .1 ˇ 2 /

1 C 4ˇ 2 .1 ˇ 2 / cos2 2
P ; 2 D : (5.9.66)
2 Œ 1 C 2 ˇ 2 .1 ˇ 2 /
The fact that the spin of one of the photons depends on the measurement of the spin
of the other photon is expressed by the basic relation

P 1 ; 2 ¤ P 1 ; P ; 2 : (5.9.67)
Re-iterating, this fundamental relation, in general, signifies the obvious correlation

that exists between the measurement of one of the spins on the other.
The speed ˇ dependence of the probabilities cannot be overlooked. The moral
of this is that polarization correlations probabilities, in general, depend on the
underlying dynamics of the theory and, in particular, on the underlying energy
(speed) and it is certainly misleading to attempt to derive them from just by
combining spins.33
At low energies ˇ ! 0, the probabilities in question become34
1 1
PŒ 1 ; 2 D sin2 .1 2 /; PŒ 1 ; D PŒ ; 2 D : (5.9.68)
2 2
5.9.3 e ! e
As another application we consider the scattering process e ! e , to the

leading order (Fig. 5.9), by taking into account the fact that the mass of the muon M
is much larger than the mass of the electron. To this end, we add a QED Lagrangian
for the muon to the QED Lagrangian of the electron.
Here the pertinent variables are given by
Q2 Q2
Q D k k 0 D p 0 p; pQ D ; kQ D ;
2 2
Q2 Q2 Q2
kk 0 D m2 ; pp 0 D M 2 ; (5.9.69)
2 2 2
neglecting the mass of the electron, i.e., for Q2 m2 . We will also use the rather
standard notations for the ingoing and outgoing electron energies:
k 0 E; k 0 0 E 0: (5.9.70)
33
This subsection is based on Manoukian and Yongram [61]. For the significance of these
polarizations correlations in their role in support of the monumental quantum theory against so-
called hidden variables theories, see, e.g., the just mentioned paper, Manoukian [56], Bell [5], and
Bell and Aspect [6]. For a detailed review, with applications to QED, the electroweak theory and
strings, see Yongram and Manoukian [83]. We have observed the speed dependence of polarization
correlations in actual quantum field theory calculations early (2003) (see [82])
34
See also Problem 5.9.
Fig. 5.9 The process

e ! e . The
momentum transfer carried p μ
by the photon is denoted by
Q. The electromagnetic
p e
charge e is shown for μ Q
convenience
e k
k
From our earlier treatments in the last two subsections, the expression of the
differential cross section is, by now standard, and in the rest frame of the muon, is
given by
d3 k 0 d3 p 0 .2/4 ı .4/ . p C Q p 0 /
d D jM j2 ; (5.9.71)
.2/3 2E 0 .2/3 2p 0 0 4 ME

where jM j2 D .e4 =Q 4 /.1=2/.1=2/Œ : 1 Œ : 2 , and in an arbitrary frame

X

Œ : 1 D 4 m2 Tr u.k 0 ; 0 / u.k; / u.k; / u.k 0 ; 0 /
spins

D 4 k 0 k C k 0 k kk 0 m2 ; (5.9.72)

X

Œ : 2 D 4 M2 Tr u.p 0 ; 0 / u.p; / u.p; / u.p 0 ; 0 /
spins

D 4 p p C p 0 p pp 0 M 2 :
0
(5.9.73)
In particular in the rest frame of muon (p 0 D M), we may use (5.9.69), to obtain
h Q2 Q2 Q2 i

Œ : 1 Œ : 2 D 64 M 2 E E 0 1 C ; (5.9.74)
4EE 0 2M 2 4EE 0
where we have neglected the mass m of the electron in comparison to the mass M
of the muon inside the square brackets. We note that
#
Q2 D 2 pQ D 2 M .E E 0 /; Q2 D 4 EE 0 sin2 ; (5.9.75)
2
# Q2 # Q2
sin2 D ; cos 2 D1 : (5.9.76)
2 4EE 0 2 4 EE 0
5.10 Modified Propagators 271
Now we use the integral

Z Z
d3 p 0 .4/
ı . p C Q p 0/ D .dp 0 /™. p 0 0 / ı. p 0 2 C M 2 / ı .4/ . p C Q p 0 /
2p00

D ı . p C Q/2 C M 2 D ı.2p Q C Q2 /; (5.9.77)
and in the rest frame of the muon, this gives

Z
d3 p 0 .4/
0 0
ı . p C Q p 0 / D ı 2 M Q0 C Q2
2p
1 Q2 1 E E0 0 E
D ı E0 E D ı E0 D ı E ;
2M 2M 2MD D 2ME D
(5.9.78)
where
E #
DD 1C2 sin2 ; (5.9.79)
M 2
and we have used the relation E 0 D .E Q2 =2M/, and the second equality
in (5.9.75) to solve for E 0 .
Finally, we note that d3 k 0 D E 02 dE 0 d˝, by neglecting the mass m of the
electron, to integrate over E 0 by using, in the process, the constraint set by the
delta function ı.E 0 E=D/ obtained above. All told, using, in particular (5.9.74)
and (5.9.76), we obtain
d ˛2 E0 h # Q2 #i
D cos 2 C sin2 : (5.9.80)
d˝ 4E 2 sin4 #
2
E 2 2M 2 2
5.10 Modified Propagators
To second order, consider the term in the exponential in (5.8.1) of h 0C j 0 i

involving only the bilinear form : : : , beyond the one in a0 , which naturally
leads to a modification of the propagator from its “bare” counterpart SC due to the
interaction. The corresponding expression is obtained from i a2 in (5.8.10). Using
a matrix multiplication notation in spacetime (as well as in Dirac indices), the total
net contribution may be written as35

i SC C e2 SC SC SC .i/ D : (5.10.1)
R
35
For a simple structure we have .dx/ .x/ SC .x x 0 / .x 0 /.dx 0 / D SC , in matrix
multiplication form.
In the momentum description this reads

Z
.dp/
i . p/ e
S. p/ . p/; (5.10.2)
.2/4
where eS. p/ is the modified fermion propagator due to the interaction, and is
given by
h Z i
e .dk/
S. p/ D SC . p/ 1 i e2 S C . p k/
D .k/S C . p/ : (5.10.3)
.2/4
It is more convenient to obtain the equation for the inverse eS1 of the modified
e1 e
propagator .S S D 1/. To second order, it satisfies the equation
Z
e .dk/
S1 . p/ D SC
1
. p/ C i e2 SC . p k/ D .k/: (5.10.4)
.2/4
We will investigate this equation and its physical interpretation in the first subsection
below.
Now we also consider terms depending on the bilinear form J : : : J which,
again, naturally leads to the modified photon propagator, due to the interaction, from
its “bare” counterpart D . To this end, again using a matrix multiplication notation
in spacetime, this is given from (5.8.8) and (5.8.10), to be
i 1

J D1 2 C i e2 D1 K D 2 J 2 ; (5.10.5)
2
where K is defined in (5.8.6) and (5.8.7). In the momentum description, the above
equation reads
Z
i .dk/ e .k/ J .k/;
J .k/ D (5.10.6)
2 .2/4
e is the modified photon propagator, due to the interaction, given by

where D

e .k/ D D1 .k/ ı 1 C i e2 L1 2 .k; k/D 2 .k/ ;
D (5.10.7)
where L .k; k/ was explicitly computed in Appendix A of Chap. 3, and
from (3.6.30), (3.6.31) and (3.6.32), it is given by
L1 2 .k ; k/ D Œ 1 2 k2 k1 k 2

Z Z 1 !
i 1 .dp/ 1
k2
C 6 dz z.1 z/ ln 1 C 2 z.1 z/ :
12 2 i 2 Œ p 2 C m2 2 2 0 m
(5.10.8)
Using the transversality condition of L .k; k/ in k , i.e., k L .k; k/ D
0; k L .k; k/ D 0, we have
e .k/ D D .k/ C 1 i e2 L .k; k/ 1 ;

D (5.10.9)
k2 k2
This equation will be analyzed in Sect. 5.10.2.
The quantum corrections to the propagators are referred to as radiative correc-
tions. In general, a diagram depicting a process, involving no integrations over
closed loops, i.e., no radiative corrections, is referred to as tree diagram.
5.10.1 Electron Self-Energy and Its Interpretation
To determine the modified electron propagator e S. p/, due to the interaction, and
provide its physical interpretation, we evaluate the self energy-part ˙. p/, given by
the integral (Fig. 5.10)
Z
.dk/ . p k/ C m
˙. p/ D i e2 D .k/; (5.10.10)
.2/4 . p k/2 C m2
where we have suppressed the i

factor in the denominator for convenience of
the notation. Also we work in arbitrary covariant gauge specified by the parameter
in the photon propagator, which we write as
k k 1
D .k/ D .1 / 2 : (5.10.11)
k k2 C 2
Here we have introduced a small fictitious mass ! 0 for the photon to avoid
divergences at low energies – the so called infrared divergence. Also note that due
to the interaction, as we will see below, the actual mass m of the electron is such
that m ¤ m0 , m0 D mŒ1CO.e2 /. Thus it is convenient to introduce the fee electron
propagator S. p/, depending on the physical mass, and set in the sequel
S1 . p/ D p C m; 1
SC . p/ D p C m0 : (5.10.12)
S−1 (p) = γ p + m0 −
p−k
Fig. 5.10 The inverse of the modified electron propagator, due to the interaction, to second order.
The corresponding mathematical expression will be denoted by e S1 . p/ D SC 1
. p/ ˙. p/. Note
the minus sign between the latter two terms
Needless to say, we have used S. p/ in the integrand in (5.10.10) to the leading

order.
The numerator N of the integrand multiplying .dk/=.2/4 is given by

N D . C 1/. p C m/ 2m C 2 k
k k
.1/ 2
Œ . p k/2 C m2 .1 / 2 Œ p2 C m2 : (5.10.13)
k k
The factor Œ . p k/2 C m2 in the third term cancels out with the denominator and
leads to an odd integral in k, and hence the integral of the third term gives zero. We
may now break the integral in question into three parts: ˙. p/ D ˙1 . p/ C ˙2 . p/ C
˙3 . p/, where
Z

.dk/ 1
˙1 . p/ D i e2 . C 1/. p C m/ 2m ;
.2/4 Œ . p k/2 C m2 .k2 C 2 /
(5.10.14)
Z
.dk/ k
˙2 . p/ D i e2 .2/ ; (5.10.15)
.2/ Œ . p k/ C m2 .k2 C 2 /
4 2
Z
2 2 2 .dk/ k
˙3 . p/ D i e .1 /Œ p C m :
.2/4 k2 Œ. p k/2 C m2 .k2 C 2 /
(5.10.16)
The
inverse electron
propagator e S1 has the general structure A C B. p C m/
2
CO . p C m/ where, A and B are constants. It diverges logarithmically
at high energies, referred to as an ultraviolet divergence. Clearly from the
dimensionality of e S1 , only the
constants A and B may have such a divergence.
Any correction of the form O . p C m/2 involves one additional power of k
in the integrand of ˙. p/ and hence necessarily is ultraviolet finite by simple
power counting of powers of k in its denominator and numerator. The main term
A C B. p C m/ is the key term for examining the ultraviolet divergence of e S1 .
Using the Feynman parameter representations to combine the denominators in
these integrals as given in (II.27), (II.35), in Appendix II at the end of the book, and
the momentum integrations given in (II.15), (II.16), (II.24), we obtain near the mass
shell p2 ' m2 (see Problem 5.13)
4 h i
˙1 . p/ ' . p C m/ .1 C / Cuv C 2.3 C / 4 Cir 2 m Cuv C 2 ;
˛
(5.10.17)
4 5 1
˙2 . p/ ' C . p C m/ Cuv C m Cuv C ; (5.10.18)
˛ 2 2
4
˙3 . p/ ' .1 /. p C m/ 1 2 Cir ; (5.10.19)
˛

up to ultraviolet finite terms O . p C m/2 , where we have used the identity
. p2 C m2 / D 2m. p C m/ . p C m/2 ; (5.10.20)

Z
1 1
Cuv D .dk/ ; (5.10.21)
i 2 .k2 C m2 /2
and for 2 =m2 ! 0 (see Problem 5.12)

Z 1
x dx 1 2
Cir D
D ln 2 : (5.10.22)
0 x2 C 2
2 .1 x/
2 m
m
Here (uv) and (ir) stand for ultraviolet and infrared, respectively. The above
Eqs. (5.10.17), (5.10.18) and (5.10.19) are given up to ultraviolet finite terms which
approach zero near the mass shell. From (5.10.17), (5.10.18), (5.10.19), (5.10.20),
(5.10.21) and (5.10.22), the inverse electron propagator, to second order, then has
the following behavior near the mass shell
. p C m0 C ım/
e
S1 . p/ ' C O . p C m/2 ; (5.10.23)
Z2
where
3˛ 3
ım D m Cuv C ; (5.10.24)
4 2
1 ˛ h 9i
D1C Cuv 2.3 / Cir C C : (5.10.25)
Z2 4 2
and ım is called the self-mass of the electron. Thus the electron’s mass is
necessarily given by
m D m0 C ım; (5.10.26)
and is gauge independent, i.e., it is independent of the parameter . It is referred to

as the renormalized mass, i.e., the experimentally observed electron mass, while m0
is referred to as the unrenormalized mass or bare mass. We will elaborate further on
this below.
The constant Cuv , may be defined in terms of an ultra-violet cut-off or by means
of dimensional regularization.36 We discuss both approaches. To this end, we may
36
See Appendix III, at the end of the book, for dimensional regularization.
define Cuv in terms of an ultraviolet cut-off as

Z h 2 i
1 1 2
Cuv D .dk/ D ln 2 1 ; (5.10.27)
i 2 .k2 2 2 2 2
C m / .k C / m
for 2 ! 1, where we have combined the denominators using (II.34), integrated

over k, using (II.8), and finally carried out the elementary integral over the Feynman
parameter x (see Problem 5.11). One may also define Cuv with dimensional
regularization, by working in .4 "/ – dimsional spacetime, with " > 0, by
introducing, in the process, an arbitrary scale parameter, say, , and using the
definition of the dimensional-regularized integral (III.8) to obtain
1 2 "=2 "
Cuv D : (5.10.28)
"=2 m2 2
Here the limit of the ultraviolet cut-off 2 ! 1, has been replaced by the limit of
the parameter " ! C 0, where we note that
" 2
D E C O."/: (5.10.29)
2 "
with E D 0:5772157 : : : denoting Euler’s constant.
The renormalization constant Z 2 being the residue of the modified propagator,
is gauge dependent and is, respectively, ultraviolet finite and infrared finite, in the
following stated gauges
1 ˛ h9 i
D 1C 6 Cir ; D 0; (5.10.30)
Z2 4 2
1 ˛ h 15 i
D 1C 3 Cuv C ; D 3; (5.10.31)
Z2 4 2
referred to as the Landau gauge and the Yennie-Fried gauge, respectively. In the
Feynman gauge D 1, it is ultraviolet cut-off and infrared cut-off dependent
1 ˛ h 11 i
D 1C Cuv 4 Cir C ; D 1: (5.10.32)
Z2 4 2
What the theory is telling us is that if jp; i denotes an electron state with
momentum p, and spin , normalized as
p 0 .3/ 0
hp 0 0 jp i D .2/3 ı .p p/ ı 0 ; (5.10.33)
m
then
p
hp j .x/jvaci D Z 2 u.p; / eip x ; (5.10.34)
defined in terms of the physical mass m. Moreover, with x 0 0 > x 0 ,

Z X
m d3 p
i hvacj .x 0 /jp i hp j .x/jvaci

p 0 .2/3
Z
d3 p 0
D Z2 i . p C m/ eip .xx / D Z 2 S.x 0 x/: (5.10.35)
2 p 0 .2/3
In reference to (5.10.34), the particle aspect appropriate for describing scattering

states dictates to scale .x/ ! .x/=.Z 2 /1=2 D ren .x/, thus introducing the
concept of the renormalized electron field, and carrying a wavefunction renormal-
ization.37
It is interesting to relate the wavefunction renormalization constant Z 2 in the
arbitrary gauge, specified by the parameter , to the one in the Landau gauge
in (5.10.30). This is obtained from (5.10.25) to be
ˇ
1 ˇˇ ˛ h 2 2 9 i
D 1 C ln C .3 / ln C
Z 2 ˇ 4 m2 m2 2
h i ˇ
˛ 2 2 9 1 ˇˇ ˛ 2
D1C ln 2 C 3 ln 2 C D C ln 2 :
4 m 2 Z 2 ˇD0 4
(5.10.36)
This will be compared with the full theory later on in Sect. 5.15 (see (5.15.30)).
Formally, Z 2 represents the probability that the (fermion) field creates a particle
out of the vacuum. The electron field, however, is a gauge dependent object and
one must be careful in such an interpretation here. The second point concerning
Z 2 is, in general, its infrared cut-off dependence. This is due to the fact that an
electron, being a charged particle, and the fact that the photon is massless, it cannot
be rigorously defined, in isolation, without the presence of an arbitrary number of
photons of arbitrarily small (soft) energies accompanying it, while Z 2 above is
introduced in terms of an electron without such a cloud of soft photons. Finally, we
note that this constant was introduced by going on the mass shell of the electron.
This corresponds to defining scattering states of an electron as observed in emission
and detection regions. This necessitates to scale the electron propagator by 1=Z 2 ,
and introduce, in turn, the renormalized electron propagator via the identification
e 1 e
Sren . p/ D S. p/; (5.10.37)
Z2

e
S1
ren . p/ D . p C m/ C O . p C m/
2
near the mass shell: (5.10.38)
37
See also Sect. 4.1 for details on the concept of wavefunction renormalization.
In reference to (5.10.34), the above condition defines a boundary condition for the
renormalized propagator near the mass shell, for a proper description of electron
in scattering states, up to the presence of soft photons accompanying it. The ever
presence of such soft photons will be discussed later in Sect. 5.12.
Now we come to the concept of mass renormalization. Using the ultraviolet cut-
off dependence of Cuv in (5.10.27), we may, from (5.10.24), write
3 ˛ h 2 1 i
ım D m ln 2 C : (5.10.39)
4 m 2
From (5.10.26), this suggests to define an effective mass which depends on an
energy scale by

2 3 ˛ h 2 1 i
m. / D m 1 ln 2 C ; (5.10.40)
4 m 2
with the so-called bare mass given by m0 D lim 2 !2 m. 2 /. The latter limit, for
2 ! 1, corresponds to going all the way to the very core of the electron for
which, presumably, no justification can be given with our present physical theories.
Through the process of renormalization, the parameter m0 has been eliminated
in favor of the physically observed mass m, as it should be done for a sensible
description of the theory.
Equation (5.10.40) is still remarkable in the sense that even at high energies but
much smaller than m expŒ.2=3˛/1=4, so that .3 ˛=4/ ln. 2 =m2 /C1=2 1,
m. 2 /, in the just mentioned equation, is not drastically different from the renormal-
ized ( physically observed) mass m measured at large distances in the lab.38 At
small distances, at the Planck length, for example, gravitation is expected to play
an essential role in formulating quantum dynamical theories. Extrapolating to such
distances, presumably, requires detailed generalizations in our physical theories.
5.10.2 Photon Self-Energy and Its Interpretation; Coulomb

Potential
The modified photon propagator, from (5.10.8), (5.10.9) and (3.6.30), may be
written as (Fig. 5.11)
e .k/ D D .k/ 1 ˘ .k/ 1 ;

D (5.10.41)
k2 k2
38
This is due to the logarithmic nature on the cut-off dependence in QED. On the other hand, for
the mass of the Higgs Boson, in the electroweak theory, one has a quadratic dependence on the
cut-off , and its unrenormalized and renormalized masses are drastically different.
p − 2k
= Dμν (k) 1 γ γν −
Dμν (k) −−
k2
μ
1
k2
p + 2k
Fig. 5.11 The modified photon propagator, due to the interaction, to second order, up to terms
obtained from the expression of a gauge invariant current. The corresponding mathematical
e .k/ D D .k/ .1=k2 / ˘ .k/.1=k2 /
expression will be denoted by D
where ˘ .k/ is given in (3.6.31) and (3.6.32) to be
˘ .k/ D Œ k2 k k ˘.k2 /; (5.10.42)
and
Z 1
2 e2 1 h k2 i
˘.k / D Cuv C 6 dz z.1 z/ ln 1 C z.1 z/ ; (5.10.43)
12 2 2 0 m2
where Cuv is given in (5.10.27). ˘ .k/ is referred to as the vacuum polarization
tensor. Note the transversality of the latter: k ˘ .k/ D 0, k ˘ .k/ D 0 which,
as mentioned in Sect. 3.6, is attributed to gauge invariance. Also note that
e2 1 ˛ h 2 1 i
˘.0/ D Cuv C D ln 2 ; (5.10.44)
12 2 2 3 m 2
Z 1 h i
2˛ k2
˘.k2 / ˘.0/ D dz z.1 z/ ln 1 C 2 z.1 z/ ˛ c .k2 /; (5.10.45)
0 m
and the latter is a finite expression.

In an arbitrary covariant gauge, specified by the parameter ,
h k k i 1 k k
D .k/ D 2 C 4 ; (5.10.46)
k k2 k
the modified photon propagator takes the form
h k k i
e .k/ D 1 1 ˘.k2 / C ;
k k
D 2 2
(5.10.47)
k k k4
and the longitudinal part of the propagator (coefficient of ) does not change with
the interaction. Later, we will show that the latter condition is true in the full theory
as well.
Near the mass shell k2 ' 0,
h i
e .k/ D k k 1 Z3 C O.k2 / C k k ;
D (5.10.48)
k 2 k 2 k4
where
˛ h 1i ˛ h 2 1 i
Z3 D 1 Cuv C D1 ln 2 ; (5.10.49)
3 2 3 m 2
is gauge invariant, and is defined with an ultraviolet cut-off , while with
dimensional regularization,39 it is given by

˛ 1 2 "=2 " 1
Z3 D 1 C : (5.10.50)
3 "=2 m2 2 2
Again from our study of the field concept, Z3 is interpreted as the probability
that the electromagnetic field creates a photon out of the vacuum. A clear picture
of the meaning of the renormalization constant Z3 further arises from a so-called
spectral decomposition of the photon propagator, which has also a counterpart in
the full theory, and is obtained as follows.
To the above end, note that in reference to the integral in (5.10.45), we may
generate the following chain of equalities
Z 1 h i Z 1 Z 1 h i
k2 @ k2
dz z.1 z/ ln 1 C 2 z.1 z/ D dz z.1 z/ d2 2 ln 1 C 2 z.1 z/ ;
0 m 0 m2 @
Z 1 Z 1
d2 ™.2 m2 /
D k2 dz z 2 .1 z/2
0 0 C k2 z.1 z/
2 2
Z 1 Z 1 Z 1 2
d ı.M 2 z.1 z/ 2 / ™.M 2 z.1 z/ m2 /
D k2 dz z 2 .1 z/2 dM 2 ;
0 .2m/2 0 2 k2 C M 2
(5.10.51)
where the last equality is easily verified by integrating over M 2 and using, in the
process, the delta function constraint. The lower limit .2m/2 of the M 2 -integral
39
In the next chapter, the gluon propagator, the counterpart of the photon propagator in QCD, will
be directly analyzed with dimensional regularization.
follows from the step function constraint and the fact that the roots of the equation
M 2 z.1 z/ m2 D 0 are
1 p 1 p
z1 D .1 1 4m2 =M 2 /; z2 D .1 C 1 4m2 =M 2 /;
2 2
requiring that M 2 4m2 . Upon integrating over 2 , we obtain for the above
integral
Z 1 Z
2 dM 2 1 z2
k 2 k2 C M 2
dz z.1 z/
.2m/ 2 M z1
Z r
k2 1
dM 2 1 2m2 4m2
D 2 2 2
1C 2 1 2: (5.10.52)
6 .2m/2 M k CM M M
All told, we have from (5.10.45), (5.10.47), (5.10.49) the following expression
e .k/ emerges40
for D
k k k k
e .k/ D D
D eC .k/ C ; (5.10.53)
k 2 k4
Z 1
eC .k/ D .M 2 /
D dM 2 2 ; (5.10.54)
0 k C M2 i
r
˛ 2 2m2 4m2
2 2 2 1
.M / D Z3 ı.M / C ™ M .2m/ 1 C 1 ;
3 M2 M2 M2
(5.10.55)
.M 2 / 0, and is referred to as the spectral function of the photon propagator, to

second order. In reference to (5.10.55), the first term in (5.10.55), corresponds to
the photon, with residue Z3 , and the second to the electron-positron pair, having an
invariant mass squared going from .2m/2 up to arbitrary large values due to their
arbitrary relative motion.
In the absence of radiative corrections, the counterpart of D eC .k/ is simply
DC .k/ D 1=k2 . The Coulomb interaction between two static charges,41 each of
charge q, may be extracted in the latter case as follows:
Z
2 d3 k ikx q2
U.jxj/ D q e D C .k/j k 0 D0 D : (5.10.56)
.2/3 4 jxj
40
The presence of i
in .k2 C M 2 i
/ in the denominator is understood.
41
A process involving no energy transfer.
eC .k/, the Coulomb

In the presence of radiative corrections, with DC .k/ ! D
potential, in turn, using (5.10.54), is modified to
Z Z 1
e d3 k ikx e q2
U.jxj/ D q2 e DC .k/jk0 D0 D dM 2 .M 2 / eM j x j ;
.2/3 4jxj 0
Z 1 r
q2 h ˛ dM 2 2m2 4m2 M j x j i
D Z3 C 1 C 1 e ;
4jxj 3 .2m/2 M 2 M2 M2
(5.10.57)
thus leading to the Coulomb potential, at large distances jxj ! 1, with modified
renormalized charges,
q2 p
e
U.jxj/ ! ren ; qren D q Z3 : (5.10.58)
4jxj
We will see later that the bare charge e0 is renormalized in this way as well, and
other renormalization constants such as Z2 , and a corresponding one for the vertex,
do not contribute to charge renormalization. That is,
p
e0 Z3 D e; ˛ 0 Z3 D ˛; (5.10.59)
where e and ˛, are the experimentally observed parameters measured at large

distances from the electron.
As for
p the electron field renormalization, the Maxwell field is also renormalized

as A = Z3 D Aren , corresponding to scattering states, thus introducing the concept
of a renormalized Maxqwell field. The renormalized photon propagator, then takes
the form
e 1 e e D e2ren D

e
D ren D D ; e2 D ren ; (5.10.60)
Z3

e 1 k k
D .k/ D D .k/ ˘ren .k/; (5.10.61)
ren
k2 k2
Z h i
2˛ 1 k2
˘ren .k/ D ˘.k/ ˘.0/ D dz z.1 z/ ln 1 C 2 z.1 z/ ; (5.10.62)
0 m
2 2
˛ k k 2
˘ren .k/ D 2
CO : (5.10.63)
15 m m2
The first relation defines the renormalized photon propagator whose transverse part
behaves as the free propagator near the mass shell, corresponding to scattering
states.
5.11 Vertex Part 283
From (5.10.59), we may write

h ˛ 2 1 i
˛0 D ˛ 1 C ln 2 ; ! 1: (5.10.64)
3 m 2
This, in turn, suggests to define an effective fine-structure by the relation
h ˛ 2 1 i
˛. 2 / D ˛ 1 C ln 2 ; (5.10.65)
3 m 2
which increases as 2 increases, i.e., it increases as the distance between the charge
decreases. This is physically expected, as from the diagram above, for example,
one may infer that the creation of an electron-positron pair leads to screening of
a charge when one is far away from the charge in question, and vice versa as one
approaches the charge as discussed in the introductory chapter of the book. The
bare fine-structure constant is then formally given by ˛0 D lim 2 !1 ˛. 2 /, which
corresponds to moving all the way to the core of the electron through the cloud of the
pairs of particles. The divergence encountered in this limit is of no surprise as our
theories are not justified to describe nature at absolute zero distances. One, however,
may re-express the theory in terms of physically observed quantities instead of
limiting experimentally unattainable quantities. The good news about QED, and all
renormalizable theories, is that when all unrenormalized quantities are eliminated
in favor of renormalized ones, the theories become ultraviolet finite defined in terms
of realistic physical quantities as the actual value of the fine-structure constant and
the actual mass of the electron.
In (5.10.58)/(5.10.59) we have seen that the Coulomb potential is simply
modified, at large distances by simply renormalizing the charges. On the other hand,
for small distances mjxj 1=2, the Coulomb potential is also modified, and as
given in Problem 5.14, to

e q2 2˛ h 1 5i
U.x/ ! Z3 C ln E ; (5.10.66)
4jxj 3 mjxj 6
where E D 0:5772157 : : : denotes Euler’s constant.
5.11 Vertex Part
We consider, up to third order, the combination of sources : : : J within the

exponential factor in h 0C j 0 i, in (5.8.1) which generate all diagrams with exactly
two electron lines and one photon lines that are attached to these sources. The lowest
order is given by a1 in (5.8.9), and the corresponding generated diagram is shown
in part (A) of Fig. 5.12. The other diagrams (B)–(E) are generated from the first and
third expressions of a3 in (5.8.11), within the curly brackets.
(A) (B) (C) (D) (E)
Fig. 5.12 The set of all diagrams (A)–(E), up to third order, which may be connected to other
diagrams by two electron lines and one photon line. This set of diagrams are all generated, with
exact numerical factors, from the terms a1 , a3 in (5.8.9), (5.8.11) involving the combination of
sources : : : J
Diagram (C)–(E) are obtained from our investigation of the self energy parts of
the electron and photon and are thus generated from the insertions of the latter in
external electron and photon lines, respectively and will be treated in Sect. 5.11.2.
The diagram in (A) is generated from a1 , while the diagram in (B) is generated
from the second term, within the first pair of curly brackets of a3 , leading, together,
to the combination
Z Z Z
.dp 0 / .dp/ .dQ/

i e0 4 4 4
. p 0 /SC . p 0 / . p 0 ; p/SC . p/. p/
.2/ .2/ .2/
D .Q/J .Q/.2/4 ı .4/ . p 0 p Q/; (5.11.1)
where note that the overall coefficient multiplying the integral is the bare charge e0 ,
. p 0 ; p/ D C . p 0 ; p/; (5.11.2)
defining the vertex part,

Z
0 2 .dk/ . p 0 k/ C m . p k/ C m
. p ; p/ D i e D .k/ :
.2/4 . p 0 k/2 C m2 . p k/2 C m2
(5.11.3)
to second order, and here we have used the renormalized charge to this lowest order.
The diagram in Fig. 5.13 shows the rooting of the momenta taken to run through
. p0 ; p/, and the direction of the external momenta.
To investigate the ultraviolet divergence of this integral, it is sufficient to
consider the external electron lines on the mass shell. We also consider initially the
expression up to the linear term in Q. The resulting expression will be also useful
in applications that will follow. The numerator coming from the part of the
photon propagator (see (5.10.11)), with the understanding that it is to be sandwiched
between u.p0 ; 0 / and u.p; /, is
N D 2 k k C 4 m k C 4. p C p 0 / k 4 . p 0 C p/ k C 4 pp 0 :
(5.11.4)
Fig. 5.13 The rooting of the

internal photon momentum p
taken in . p 0 ; p/, where
the vertex part is given by p −k
. p 0 ; p/ D Q
C . p 0 ; p/ Λ μ (p , p) = k γμ
p−k
We combine the three propagators in (5.11.3) with A D k2 , B D . p 0 k/2 C m2 ,

C D . p k/2 C m2 , by using the Feynman parameter representation in (II.35) of
Appendix II at the end of the book:
Z 1 Z
1 x
1
D 2 dx dz : (5.11.5)
ABC 0 0 ŒA.1 x/ C Bz C C.x z/3
Since by power counting the integral is only logarithmically divergent, we carry

out, in turn, a change of variable: k ! k C p 0 z C p.x z/, leading effectively to
A.1 x/ C Bz C C.x z/ ! k2 C m2 x 2 C 2 .1 x/: (5.11.6)
On the other hand for the resulting expression of the numerator N , we have
Z x
dz N D .k2 C m2 x 2 /x C mx 2 .1 x/Œ ; Q C m2 x.x 2 C 4x 4/ :
0
(5.11.7)
Integrating over k and x, give for the =.k2 C 2 / part of the photon
propagator contribution to (5.11.3)
˛ h 11 i 1
Cuv 4Cir C C Œ ; Q : (5.11.8)
4 2 2m
For the .1 /k k =k2 .k2 C 2 / term (see (5.10.11)) in the photon propagator,
note that the corresponding integrand in the vertex part (5.11.3) is simply
k k
. 1/ S. p 0 k/ S. p k/
k2 .k2 C 2 /
1 h i h i
D . 1/ 1 S 1 . p 0 / S. p 0 k/ 1 S. p k/ S 1 . p/ :
k2 .k2 C 2 /
(5.11.9)
When the latter is sandwiched between u.p 0 ; 0 /; u.p; /, it leads to an additional
contribution (see (5.10.22), (5.10.27), and Problem 5.12)
Z
.dk/ 1 2
i e2 .1 /
.2/ k .k C / k C 2
4 2 2 2 2
˛ 2 ˛ h i
D .1 / ln 2 D .1 / Cuv C 2 Cir C 1 ; (5.11.10)
4 4
with an ultraviolet cut-off 2 , and an infrared cut-off 2 . All told, the leading
contribution in Q to . p 0 ; p/, on the mass shell, is given by
1 ˛
. p 0 ; p/ D C Œ ; Q ; Q D p 0 p; (5.11.11)
Z1 8 m
1 ˛ h 9i
D 1C Cuv 2.3 / Cir C C : (5.11.12)
Z1 4 2
The renormalized vertex is then defined by

ren . p 0 ; p/ D Z1 . p 0 ; p/: (5.11.13)
From (5.10.25), we note that the vertex function renormalization constant Z 1 is

equal to Z 2 in all gauges. This is no accident, and we will see later that this is true
to all orders.
An infrared cut-off may be also introduced by dimensional regularization by
working in 4 C " dimensions of spacetime with " > 0. Still another approach,
for example, is to simply restrict the momentum carried by a photon jkj > K, with
K m. That is, in the latter approach, we may express the photon propagator part
DC as
Z Z
d3 k dk0 1 0
DC .x x 0 / D 3 2
eik.xx / ; (5.11.14)
jkj>K .2/ .2/ k i
which may be rewritten as

"Z
.dk/ 1 0
DC .x x 0 / D Lim!0 eik.xx /
.2/4 k2 C 2 i
Z Z #
d3 k dk0 1 ik.xx 0 /
3
e : (5.11.15)
jkj<K .2/ .2/ k2 C 2 i
The first expression within the square brackets is what we have been using thus
far. The second term, when substituted in the integral equation of the vertex part, in
addition to the first one, allows one to eliminate the fictitious photon mass cut-off
in favor of the infrared cut-off K, via the equation
Z 1 2
x dx 1 m 5
Cir D
2 D ln D ln C : (5.11.16)
0 x2 C .1 x/ 2 m2 2K 6
m2
This transformation rule will be useful later on.

The consequences of the renormalization constants Z1 ; Z 2 ; Z3 is discussed next.
5.11.1 Charge renormalization and External lines
To second order, the inverse of the modified electron propagator is given by
e
S 1 . p/ D SC
1
. p/ ˙. p/; (5.11.17)
where ˙. p/ is defined in (5.10.10), and, as emphasized earlier, SC . p/ depends

on the bare mass m0 , while S. p/, given in (5.10.12), is defined in terms of the
physical one m. Since m0 D m ım, we conveniently have
1
SC . p/ D S 1 . p/ ım: (5.11.18)
Referring to Fig. 5.10, this allows us to re-express e

S 1 . p/ as

e
S 1 . p/ D S 1 . p/ ˙. p/ C ım ; e
S. p/ D S. p/ C S. p/ ˙. p/ C ım S. p/;
(5.11.19)
with the following boundary condition
ˇ
˙. p/ C ım S. p/ˇmass shell ! Z 2 1 D .Z 1
2 1/; (5.11.20)
and the latter equality holds true to second order. Accordingly, if the mass term
m0 in the lagrangian density is simply rewritten as m C ım , and
we use the propagator S. p/, depending on the physical mass m, as we should,
as the zeroth order and treat ım in a perturbation expansion, thus generating
another, though simple, vertex part, then we may simply replace ˙. p/ in our earlier
expressions, wherever they appear, in a1 ; a2 ; a3 , in h 0C j 0 i, by ˙. p/ C ım, and
SC . p/ by S. p/. Thus the totality of all diagrams in Fig. 5.12, with two external
electron line and one photon line, up to the order considered, are generated by the
corresponding terms in a1 , a3 , including now the perturbation term ım , and
is given in terms of momentum variables as

Z Z Z
.dp 0 / .dp/ .dQ/
i 4 4
.2/4 ı.Q p 0 C p/
.2/ .2/ .2/4
. p 0 / S. p 0 / ˝ . p 0 ; p/ S. p/ . p/ D .Q/ J .Q/; (5.11.21)
h
˝ . p 0 ; p/ D e0 C ˙. p 0 / C ım S. p 0/ C . p 0 ; p/
1 i
C S. p/ ˙. p/ C ım 2 ˘ .Q/ ; (5.11.22)
Q
where note that the overall charge multiplying the expression for ˝ is e0 .
The coupling terms of the fields to their external sources in the Lagrangian
density may be rewritten as
C C A J D ren ren C ren ren

C Aren Jren ; (5.11.23)
p p
ren D = Z 2; ren D Z 2 ; (5.11.24)
p p
ren D = Z 2; ren D Z 2 ; (5.11.25)
p
p
Aren D A = Z 3 ; Jren D Z 3 J; (5.11.26)
and the external sources, now associated with renormalized states, i.e., states
described by spinors u; u, v; v, expressed in terms of the physical mass and satisfy
the earlier normalizations conditions, and with the photons described by appropriate
polarization vectors as before.
The external sources now become expressed in terms of renormalized
sourcespand hence ˝ . p 0 ; p/ becomes replaced by .Z 2 Z 3 Z 2 /1=2 ˝ . p 0 ; p/
.Z 2 Z3 /1 ˝ . p 0 ; p/, which will be used below in (5.11.27).
The identity k D . p 0 C m/ . p C m/, implies the vanishing of the latter
expression when sandwiched between the spinors u.p 0 ; 0 / and u.p; /. When this
is applied to the last term within the square brackets in (5.11.22), this term simply
becomes
u.p 0 ; 0 / u.p; /˘.Q2 /;

(see Problem 5.15). We introduce ˝ren . p 0 ; p/ as the renormalized counter-
part of ˝. p 0 ; p/ with renormalized external lines. By using (5.11.19)/
(5.10.23), (5.11.11)/(5.11.2), (5.10.44)/(5.10.49), we obtain from (5.11.22) on
the mass shell, the expression
1 i h
u.p 0 ; 0 / ˝ren

. p 0 ; p/ u.p; / D
p u.p 0 ; 0 / ˝ . p 0 ; p/ u.p; /
.Z 2 Z 3
h 1 i
D e0 p u.p 0 ; 0 /
Z 2 Z3
1 1
12 1 C 1 . 1 Z 3 / C O.Q2 / CO.Q / u.p; /;
Z2 Z1
(5.11.27)
for Q ' 0.
Now we may take the limit Q ! 0, to infer that the right-hand side of the
h 1 i
u.p; 0 / u.p; / e0
p
Z3
h 1 ih 1 1 i
1C 1 12 1 C 1 .1 Z 3 /
Z2 Z2 Z1
p
D Z3 e0 u.p; 0 / u.p; /; (5.11.28)
where we have deliberately written Z 21 D 1 C.Z 21 1/ in the external electron
lines renormalization constant, multiplied this factor with the expression within the
square brackets following it, working to second order, thus simply adding the second
order term .Z 21 1/ inside the square brackets, and finally used the fact that
Z 1 D Z 2.
Thus ˝ , due to the external lines renormalization, is consequently renormal-

ized as ˝ren D .Z 2 Z 3 Z 2 /1=2 ˝ , satisfying the boundary condition
ˇ
ˇ
u.p 0 ; 0 / ˝ren

. p 0 ; p/ u.p; /ˇ D e u.p; 0 / u.p; /; (5.11.29)
QD0
defining the renormalized charge, i.e., the experimentally observed charge given by
p
e D Z3 e0 ; (5.11.30)
and the wavefunction renormalization constant Z 2 of the electron, and that of the
vertex function Z 1 do not contribute to it, as a consequence of their equality - a
result which also holds true to all orders, as we will see later.
The zero momentum transfer of the electron provides information only on the
charge of the electron. To see any structure and obtain additional information on the
electron, a non-zero momentum Q transfer is necessary.
5.11.2 Anomalous Magnetic Moment of the Electron
We have seen earlier in (3.2.22), that the matrix element corresponding to the
scattering amplitude of a Dirac particle off an external electromagnetic field Aext .x/
has the following form
e Œ u.p 0 ; 0 / u.p; / Aext

.Q/
h . p C p 0 / Œ ; i
D e u.p 0 ; 0 / C Q u.p; / Aext
.Q/; (5.11.31)
2m 4m
0
.Q/ is the Fourier transform of A .x/, Q D . p p/. On the other
where Aext ext
hand, radiative corrections, modify the vertex e0 , including the renormalization

of charge, to the renormalized expression: e C ˛ Œ ; Q =.8 m/ at low
momentum transfer (see (5.11.11)/(5.11.13)). Thus according to QED, the above
amplitude becomes simply modified to

0 0 . p C p 0 / Œ ; e ˛
u.p ; / e C Q 1C u.p; / Aext
.Q/;
2m 4 m 2
(5.11.32)
providing a correction factor .1 C ˛=2/ to the electron magnetic moment,42

a celebrated result first derived by Schwinger.43 The additional factor ˛=2 is
referred to as the anomaly. Needless to say higher order effects of this correction
may be also carried out and have been successfully done over the years.44
5.12 Radiative Correction to Coulomb Scattering and Soft

Photon Contribution
We consider radiative corrections to the Coulomb scattering for non-zero momen-

tum transfer Q D p 0 p, where p; p 0 denote the initial and final momenta of
the electron. To this end, for Q2 ¤ 0, we extend the analysis of the vertex part in
Sect. 5.11 beyond the linear term in Q to second order Q2 .
42
See (3.2.22), (3.2.23), (3.2.24), (3.2.25) and (3.2.26).
43
Schwinger [68].
44
When the electron is treated non-relativistically, this correction

has been also
carefully computed
(see Manoukian [56], p. 453 for a textbook treatment) to be: 1C .˛=2/ , where D .16=9/
2 ln.3=2/ ' 0:97, and compares well with the QED result.
5.12 Radiative Correction to Coulomb Scattering and Soft Photon Contribution 291
In this case, we note that the expression in (5.11.6) in the denominator becomes
replaced by
A.1 x/ C Bz C C.x z/ ! k2 C m2 x2 C 2 .1 x/ C Q2 z.x z/; (5.12.1)
where
Q2
2 p 0p D Q2 2m2 ; Qp D ; (5.12.2)
2
and, in particular, that
Z
1 .dk/ Q2 1
D Cir C Cir : (5.12.3)
i 2 AB C 3 m2 2
to second order in Q. The modification of e0 u.p 0 ; 0 / ˝ . p 0 ; p/ u.p; /, after

mass and charge renormalization, is then readily obtained to second order in Q
to be
e u.p 0 ; 0 / ˝ren

. p 0 ; p/ u.p; /
" ! #
0 0 ˛ Q2 h 1 2 3 1 i ˛
D e u.p ; / 1 ln 2 C Œ ; Q u.p; /;
3 m2 2 m 8 5 8m
(5.12.4)
where, in the process, we have, from (5.10.62) and (5.10.63), used the fact that
d˘ren .Q/=dQ2 jQ 2 D 0 D .˛=15 m2 / in (5.11.22) and (5.11.27), after sandwiching

˝ren . p 0 ; p/ between u.p 0 ; 0 /, u.p; /.
Accordingly, for the scattering of an electron off a Coulomb potential
Ze2 Ze2
eA .x/ D ;0 ; eA .Q/ D Œ2ı.Q0 / 0 ; (5.12.5)
4jxj jQj2
we have from (5.12.4) and this equation, that the transition probability per unit time,
normalized in a unit volume, is given by .Q D p 0 p/
ˇ
ProbŒ p; ! p 0 ; 0 ˇunit time
m d p m 2 ı.Q /Z e ˇˇ
3 0 0 2 4 ˇ2
D 0 0 3 0 4
u.p 0 ; 0 / ˝ren
0
. p 0 ; p/ u.p; /ˇ ; (5.12.6)
p .2/ p jQj
for jQj2 =m2 1. Upon averaging over the initial spin of the electron and summing
over its final spin, it easily follows,45 that the differential cross section for Coulomb
45
See also (3.4.5), (3.4.6), (3.4.7), (3.4.8) and (3.4.9).
scattering, including radiative corrections, in the non-relativistic limit is given by
d d ˇˇ h 2 ˛ jQj2 1 2 1 i jQj2
D ˇ 1 ln 2 ; 1; (5.12.7)
d˝ d˝ 0 3 m2 2 m 5 m2
where d=d˝j0 is the differential cross section for Coulomb scattering without
radiative corrections (see (3.4.9), (3.4.10), (3.4.11), (3.4.12) and (3.4.13)).
At this point, it is wise to use the transformation rule in (5.11.16), to eliminate
the fictitious photon mass in favor of the infrared cut-off and rewrite the differential
cross section as
d d ˇˇ h 2 ˛ jQj2 m 5 1 i jQj2
D ˇ 1 ln C ; 1; (5.12.8)
d˝ d˝ 0 3 m2 2K 6 5 m2
involving a rigorously zero mass photon. This expression still depends on the
unwanted infrared cut-off K. A charged particle, however, is accompanied by very
soft photons, i.e., of arbitrary small energies and, due to the resolution involved in
any experimental set up, such very soft photons cannot be experimentally detected.
Thus to the lowest order considered above, we consider the emission of a very soft
photon in the Coulomb scattering problem of Sect. 3.4. The experimental resolution
will be defined by setting an upper bound for the energy of such a very soft by .
The diagrams in Fig. 5.14 show the emission of a photon of momentum k from
external electron lines in Coulomb scattering, involving no radiative corrections.
The corresponding matrix element for the scattering process is given by
h . p k/ C m . p 0 C k/ C m 0 i
A0 . p p 0 k/ e u.p 0 ; 0 / 0 2 2
C e u.p; /;
. p k/ C m . p 0 C k/2 C m2
(5.12.9)
which for a very soft photon, we may set k to be zero in the numerator, p 0 0 '
p0 , and the matrix element above, upon using the Dirac equations for the spinors,
reduces to
h p 0 p i
A0 . p p 0 / e u.p 0 ; 0 / 0 e u.p; /: (5.12.10)
p 0k pk
k k
p p p p
× + ×
γ Aext (p p k) γ Aext (p p k)
Fig. 5.14 The emission of a photon of momentum k, respectively, by the ingoing and outgoing
electron in Coulomb scattering not involving radiative corrections
5.13 Lamb Shift 293
For a non-relativistic electron p k D p 0 k D m k0 , and the differential cross

section, with unpolarized electrons, corresponding to the emission of a very soft
(unpolarized) photon, referred to as soft bremsstahlung (SB), is simply given by
Z
d ˇˇ d ˇˇ 2 d3 k 1 X j
ˇ D ˇe 3 2
Q i Q j ei e ; (5.12.11)
d˝ SB d˝ 0 K<jkj< .2/ 2 jkj m k
2

Qi D p 0i p i . Using the completeness relation and the angular average given below,
X k i kj D k i kj E 4 ij
j
ei e D ij 2
; 2
D ; (5.12.12)
k k 3

we obtain
d ˇˇ d ˇˇ 2 ˛ Q2 jQj2
ˇ D ˇ ln ; 1: (5.12.13)
d˝ SB d˝ 0 3 m2 K m2
Therefore, for the inclusive process, for the scattering amplitude, with radiative
corrections, and including the emission of a very soft photon, which would escape
its detection, is, from (5.12.8) and (5.12.13), given by
i
d ˇˇ d ˇˇ h 2 ˛ jQj2 m 5 1
ˇ D ˇ 1 ln C ln ; (5.12.14)
d˝ incl d˝ 0 3 m2 2K 6 5 K
or

d ˇˇ d ˇˇ h 2 ˛ jQj2 m 5 1 i jQj2
ˇ D ˇ 1 ln C ; 1; (5.12.15)
d˝ incl d˝ 0 3 m2 2 6 5 m2
and the infrared cut-off K cancels out eliminating the so-called infrared divergence.
This cancelation occurs to all orders when emissions of soft photons are appropri-
ately taken into consideration.46
5.13 Lamb Shift
Two experiments which have much stimulated the development of quantum field
theory in the earlier days, were the observation of anomalous magnetic moment of
the electron,47 which was already derived in Sect. 5.11.2, and of the Lamb shift.48
46
See, e.g., Yennie and Suura [81]. See also Manoukian [52].
47
Kusch and Foley [39], cf., Foley and Kusch [25].
48
See Lamb and Retherford [40] for the early experiment.
The Lamb shift in consideration here, is the splitting between the 2 S1=2 and the
2 P1=2 levels of the hydrogen atom and is due to the interaction of the atomic
electron with radiation, that is, with the photon.49 These levels are degenerate
according to the Dirac theory.50
We provide a derivation of the Lamb shift which is physically easy to grasp,
and parallels methods often used in formulating the eigenvalue problem in quantum
mechanics courses and the reader will, hopefully, be able to easily follow. The basic
idea is that with an infrared cut-off m˛ 2 K m, as discussed in Sect. 5.11, we
may use the expression of the vertex correction given in that section, by restricting
the spatial photon momentum k as follows jkj > K, and apply straightforward first
order perturbation theory with the latter as a perturbation to the Dirac theory. On
the other for jkj < K, one is necessarily involved with a low energy non-relativistic
regime, and a straightforward, though modified perturbative treatment in this region,
may be developed and readily carried out. In this latter, case, special care needs to be
taken as the analysis leads to a logarithmic expression in the fine-structure constant,
and the Coulomb potential has to be treated exactly, while radiative corrections may
be treated perturbatively. If we denote the contributions of the Lamb shift by •E >
and •E < , corresponding to both regions : jkj > K, jkj < K respectively, then
the total shift is given by •E D •E > C •E < , and the infrared cut-off K cancels
out between the two expressions.51 This leads to a residual logarithmic dependence
on the fine structure constant squared. Here we should remember that the infrared
problem, in the radiative corrections to Coulomb scattering, arises from low energy
photons, and in the atom, the energy of a photon is of the order of the difference
between the energy levels of the Hydrogen atom, that is, of the order m˛ 2 .
• Derivation and Evaluation of •E > :
Our starting point is the bound state problem of the Dirac equation in the presence
of a Coulomb potential given by
@

e 0 A 0 .x/ C m n .x/ D 0; (5.13.1)
i
with
iEn x 0 ˛
n .x/ D n .x/ e ; eA0 .x/ D ; (5.13.2)
jxj
49
Sometimes I have the feeling that there are as many derivations of the Lamb shift as there are
field theory practitioners.
50
A careful derivation of the Lamb shift was particularly carried out by Erickson and Yennie [22],
see also Fox and Yennie [26]. There are also other careful derivations in the literature- too many to
mention. A highly sophisticated and a very careful derivation, spelling out the finest of details, is
also given by Schwinger [71].
51
For a treatment and a pedagogical presentation of the Lamb shift with the electron treated non-
relativistically, see Manoukian [56], pp. 391–403, which is in the spirit of Bethe’s [8] classic
treatment.
5.13 Lamb Shift 295
for an attractive (spherically symmetric) potential. The former equation may be

rewritten as
r
0
eA 0 .x/ C 0 m n .x/ D En n .x/: (5.13.3)
i
Upon writing D . '/> , we have
1

'D 1 C O.˛ 2 / p ; (5.13.4)
2m
in the hydrogen atom, where eA0 m ˛ 2 , E n D mŒ1 C O.˛ 2 /. The normalization
R 3
condition
R 3 d x n .x/ n .x/ D 1, dictates to set D .1 p2 =8 m2 / with
d x .x/.x/ D 1. That is, we may write
0 1 s 0 1
p2 n
.1 / n
B 8m2 C p0 C m B C
n D@ A' @ A; (5.13.5)
p 2p0 p
n p0 Cm
n
2m
with the latter expression properly normalized. We recall that the fine-structure of
the hydrogen atom, as a perturbation, is of the order of ˛ 4 and non-relativistic
wave-functions corresponding to n above are sufficient to derive it. The Lamb
shift is even of smaller magnitude, hence we may still consider non-relativistic
wavefunctions.
The relativistic corrections arising from (5.13.3), necessitates combining
spin and orbital momentum in an atom, thus introducing the total angular
momentum by J D L C S. For the Lamb shift in consideration, j D 1=2
and we need to consider the two values ` D 0; 1.52 This leads to the
construction of wavefunctions for the treatment of relativistic corrections,
which for the problem in consideration, with quantum number n D 2, we
have,53
X X 1=2
1 ' 2;`;C p
2;` D ; N0 D 2; N1 D jam j C jbm j ;
N` ' 2;`;
mD0;1 mD0;C1
(5.13.6)
' 2;0;˙ .r; ˝/ D R20 .r/ Y0;0 .˝/; (5.13.7)
52
Note that mj D m` C ms , and for ms D C1=2 W mj D C1=2, mj D 1=2, and we have,
respectively, m` D 0; 1. Similarly, for ms D 1=2 W we have m` D C1; 0.
53
For details on constructions of wavefunctions corresponding to the addition of spin and orbital
angular momentum see, e.g., Manoukian [56], pp. 383, 384, 407.

X
' 2;1;C .r; ˝/ D R21 .r/ am Y1;m .˝/ ; (5.13.8)
mD0;1

X
' 2;1; .r; ˝/ D R21 .r/ bm Y1;m .˝/ ; (5.13.9)
mD0;C1
where a linear combination over spherical harmonics is involved for ` D

1. The radial parts R2` .r/ vanish exponentially e˛mr=2 for r ! 1.
The four component expressions may be then taken as (see (5.13.5) and
(5.13.6))
s 0 1
X p0 Cm B

C' 2;`; X rm ' 2;`; .p/
2;` .p/ ' @ A D u.p; / ;
2p 0 p N` p0 N`
D˙
p0 Cm
D˙
(5.13.10)
conveniently denoted in a momentum description, where u.p; / is the Dirac

spinor, satisfying
p0
. pCm/u.p; / D 0; normalized as u .p; /u.p; 0 / D ı 0 ; (5.13.11)
m
and the solutions 'n;`; .p/ vanish rapidly for jpj m˛, which are both important
conditions for the applicability of the explicit expression of the perturbation to be
just discussed.
In the presence of a perturbation •U.x/,
eA0 ! eA0 C •U; n ! n C• n; En ! En C •En ; (5.13.12)
leading from (5.13.3) to

r
0 eA 0 .x/ C 0 m • n .x/ C •U.x/ n .x/ D En• n .x/ C n .x/ •E n :
i
(5.13.13)

Upon multiplying this equation by n .x/, and integrating with respect to x, we
obtain the simple expression
Z
•En D d3 x
n .x/ •U.x/ n .x/; (5.13.14)
where we have, in the process, took the adjoint of (5.13.3), and used the fact that
. i / 0 D 0 i , i 0 D 0 i to cancel out the first term on the left-hand
side of (5.3.13) with the first term on its right-hand side after multiplying the latter

equation by n .x/ and integrating with respect to x by parts.
5.13 Lamb Shift 297
In the presence of radiative corrections, the interaction of the electron with the
external attractive Coulomb potential is provided by the replacement of e 0 A 0 .Q/,
in the momentum description, to the leading order, by the explicit expression
(see (5.12.4))
h
˛ Q2 m 5 3 1 i ˛
0 1 ln C C Œ 0
; i
Q i eA0 .Q/;
3 m2 2K 6 8 5 8 m
(5.13.15)
as a modification of the coefficient of 0 , where K is an infrared cut off discussed
in Sect. 5.11. In the atom, jQj, is of the order of the energy difference m ˛ 2
between its levels, hence, in particular, Q2 =m2 1, justifying the fact that only the
order Q2 =m2 is retained in (5.13.15). Also note that n;` .p/ in (5.13.10) satisfies
approximately the free, i.e. on mass shell, Dirac equation.
In the coordinate description, then . r i @ i /
h ˛ m 5 3 1 r 2 ˛ r i ˛
•U.x/ D ln C 2 : (5.13.16)
3 2K 6 8 5 m 4 m i jxj
Using the relation r 2 .1=jxj/ D 4 ı .3/ .x/, the first term contribution to •E >
in (5.13.14), and
using the notation discussed
in the introductory paragraphs of this
section, gives j'n` .0/j2 D .m˛/3 ı` 0 =n3
4 ˛ m 5 3 1 ˛
• .1/E > D ln C j' 2` .0/j 2
3 2K 6 8 5 m2
˛ .˛/4 m 5 3 1
D m ln C ı` 0 ; (5.13.17)
6 2K 6 8 5
where we have used
ı` 0
j'n` .0/j2 D .m˛/3 :
n3
For the second term, the relevant matrix element reduces to the evaluation
0 0 1 1
Z 0 p 2;`
˛2 p 1
d3 x 2;` 2;` @ A @ A
4 m 2m jxj p
p 0 2m 2;`
Z ( )
˛2 1 1 1
3
D d x i j 2;` @i @j 2;` 2;` @j @i 2;` C 2;` .@i /@j 2;`
8 m2 jxj jxj jxj
Z ( )
˛2 1

D d3 x 4 2;` ı .3/ .x/ 2;` C 2 i "ij k 2;` k @i @j 2;` ; (5.13.18)
8 m 2 jxj
where in the first equality we have integrated by parts, and in the second equality we
have used the identity: i j D ıij C i "ij k k . As before, the first term contributes for
` D 0 only. On the other hand, the second term is zero for ` D 0.54 For ` D 1, the

second term, within the curly brackets, may be written as 4 2;` S L=jxj3 2` ,55
defining a spin-orbit coupling, whose expectation value is simply 4 h1=jxj3 i D
4 .m ˛/3 =24, where we have used the fact that h1=jxj3 i D .m ˛/3 =3n3 , with n D
2, for the hydrogen atom. All told, we have
˛.˛/4 ˛.˛/4
• .2/E > D m ı` 0 m ı `1 : (5.13.19)
16 48
Equations (5.13.17) and (5.13.19) then give
˛.˛/4 h 1 m 5 3 1 1 i ˛.˛/4
•E > D m ln C C ı` 0 m ı` 1 : (5.13.20)
2 3 2K 6 8 5 8 48
• Derivation and Evaluation of •E < :

Our starting point for the derivation of the expression for •E < , which is
involved with very low energy much smaller than m, is the non-relativistic
Hamiltonian for the interaction of the electron with radiation in the presence
of the Coulomb potential. This is given by the expression
1 2 ˛
HD p eARAD C H 0;RAD ; (5.13.21)
2m jxj
where H 0;RAD is the free Hamiltonian of the radiation field whose detailed
structure, as we will see below, is not needed. The Hamiltonian may be rewritten
in a more convenient form as
H D H C C H 0;RAD C HI ; (5.13.22)
where H C is the Coulomb Hamiltonian
p2 ˛
HC D ; (5.13.23)
2m jxj
54
A careful demonstration that this term is zero for ` D 0, .L2 D 0/, can be found on pp. 381–382
in Manoukian [56] in the analysis of the atomic fine-structure.
2 2 2

may use the identity S L D J L S =2, and the fact that j.j C 1/ `.` C
55
Here one
1/ 3=4 =2 D 1 for j D 1=2; ` D 1.
5.13 Lamb Shift 299
and H I is the interaction Hamiltonian of the electron-photon system
e e2 2
HI D ARAD p C p ARAD C A : (5.13.24)
2m 2m RAD
We treat H I as a perturbation to the Coulomb Hamiltonian. As we will see below,
the Coulomb potential has to be treated exactly.
To zeroth order, that is for no electron-photon interaction, let jvaci, jk; i denote,
respectively, the vacuum
state and the single photon state of momentum k and (real)
polarization e .k/, k e .k/ D 0 . Then, in particular,
hvacjARAD .x/jvaci D 0; hvacjARAD .x/jki D eikx e .k/; (5.13.25)

hvacjARAD .x/jk1 1 ; : : : ; ka a i D 0; for a 2; (5.13.26)
h kjARAD .x/ D eikx e .k/ hvacj; (5.13.27)
where
jk1 1 ; : : : ; ka a i D ˘ia jki i i:
The above expressions allow us to evaluate hvacjA2RAD .x/ through the following
chain of equalities
hvacjA2RAD .x/ D hvacjARAD .x/jvaci hvacjARAD .x/

Z X
d3 k
C hvacjARAD .x/jki hkjARAD .x/
.2 /3 2 jkj
D1;2
Z X
d3 k
D eikx e .k/ e .k/ eikx hvacj
.2 /3 2 jkj
D1;2
Z X
d3 k
D hvacj; (5.13.28)
.2 /3 2 jkj
D1;2
where we have used the completeness relation for free photons consisting of
projection on the vacuum state, plus one photon contribution, plus two photons
contribution, and so on, given by
Z X !
d3 k
jvacihvacj C jk ihk j C D 1: (5.13.29)
.2 /3 2 jkj
D1;2
Also note that H 0;RAD j k i D j k j j k i.

To zeroth order, in the absence of the electron-photon interaction, the Hamil-

tonian is given by H C C H 0;RAD , and the following equations, in particular, are
satisfied

H C C H 0;RAD j' n i jvaci D E n j' n i jvaci (5.13.30)

H C C H 0;RAD j' n i jki D .E n C jkj/j' n i jki; (5.13.31)
where the E n denote the hydrogen atom energy levels with corresponding eigen-
vectors j' n i.
In the presence of the perturbation, described by H I , the bound-state eigenvalue
equation reads
en j n i:
H j n i D E (5.13.32)
Upon multiplying the latter equation by h ' n j hvacj h'n0 j, and using (5.13.30), give
h ' 0n jH I j n i
E n D ; E n D e
E n E n: (5.13.33)
h ' 0n j n i
Let us first consider the contribution of e2 A2RAD =2m in H I to the latter equation. To
this end, we have from (5.13.28)
Z X
h 'n0 jA2RAD .x/jn i h'n0 jn i d3 k
D
h 'n0 jn i h'n0 jn i .2 /3 2 jkj
D1;2
Z X
d3 k
D ; (5.13.34)
.2 /3 2 jkj
D1;2
which is independent of the quantum number n; ` and hence is common to all levels
and does not contribute in computing to the energy differences.56
On the other hand for the first term in (5.13.24), we first multiply (5.13.32) by the
single photon state h k; j, generating a state corresponding to H C only, given by
1
h kjn i D h kjH I j n i: (5.13.35)
e
.E n H C jkj/
From (5.13.28), the matrix element of ARAD between the vacuum states as
well as between the vacuum states and multiphoton states are zero, and hence the
contribution of the first term in (5.13.24), (5.13.25), (5.13.26), (5.13.27), (5.13.28),
56
Although this common constant is infinite, it may be defined with a cut-off.
5.13 Lamb Shift 301
(5.13.29), (5.13.30), (5.13.31), (5.13.32) and (5.13.33) is, from (5.13.35), given by
Z X
e 4 e2 d3 k
h'n0 jp ARAD .x/ C ARAD .x/ p jn i D
2m m2 .2 /3 2 jkj

D ˇ 1 ˇ E
ˇ ˇ
' n ˇp e .k/ eikx eikx e .k/ pˇ' n ; (5.13.36)
.En H C jkj/
where we have used the fact that k e .k/ D 0, and replaced the denominator
in (5.13.33) by one, and E 0n by En , to the leading order, and obviously replaced
h'n0 j by h'n j. Using the fact that eik:x is a translation operator of p to p k we
have
1 1
eikx eikx D .pk/2
.En HC jkj/ .En UC jkj/
2m
1
D .pk/2 p2
; (5.13.37)
.En 2m HC C 2m jkj/
where UC is the Coulomb potential. Accordingly, we may rewrite (5.13.36) as

Z X
e 4 e2 d3 k
h'n0 jp ARAD .x/ C ARAD .x/ pjn i D
2m m2 .2 /3 2 jkj

D ˇ 1 ˇ E
ˇ ˇ
'n ˇp e .k/ .pk/2 p2
e .k/ pˇ'n : (5.13.38)
.En 2m
HC C 2m
jkj/
The expression in (5.13.38) is not the final one for •E < . The reason is that in the
absence of the Coulomb interaction, such a term is not involved with the electron
in the atom, i.e., the electron in the Coulomb potential. Accordingly, we have to
subtract the corresponding conribution, with the substitutions H C ! p2 =2m, En !
p2 =2m, in the denominator in (5.13.38), from its original expression as it appears in
the latter equation. This leads to the replacement57
1
.pk/2 p2
!
.En 2m
HC C 2m
jkj/
1 1
.pk/2 p2
2 p2
: (5.13.39)
.En 2m HC C 2m jkj/ . .pk/
2m C 2m jkj/
57
This simply gives rise to a mass renormalization, see Manoukian [56], pp. 401–43. Also note that
the matrix element h'n jŒ : j'n i, in (5.13.38) for H C En ! 0, simply factors out as h'n j'n i Œ : D
Œ : , giving an expression independent of the Coulomb functions and its related quantum numbers.
This dictates, in turn, to modify the expression in (5.13.38) to the following one
Z X 3
˛ d k 1
h' n jp e .k/ h i
4 2 m2 jkj p
. m / k jkj 1 C jkj
2m
1
.H C En / h i e .k/ p j' n i: (5.13.40)
jkj
E n HC C . mp / k jkj 1 C 2m
In the atom, jpj is of the order m˛, while jkj is of the order m˛ 2 , which is of the
order of the difference between the energy levels of the hydrogen atom. Taking this
into account and finally recalling that •E < is to be defined with an upper cut-off
jkj < K, we have
Z X d3 k D ˇˇ ˇ E
˛ 1 ˇ
•E < D 'n ˇp e .k/.HC En / h i e .k/ pˇ'n :
4 2 m2 jkj<K jkj 2
H C En C jkj

(5.13.41)
We note that .HC En / is of an order comparable to jkj and should be kept in the
denominator. As a matter of fact, the latter provides a lower bound cut-off in the
infrared region jkj ! 0.
Upon using the completeness relation and the following angular integration,
respectively,
Z
X j ki kj ki kj 8 ij
ei .k/ e .k/ D ı ij 2
; d˝ ı ij 2 D ı ; (5.13.42)
k k 3

we get
ˇ E
< 2 ˛ D ˇˇ K ˇ
•E D ' n ˇŒp; H C ln pˇ' n ; (5.13.43)
3m2 jHC En j
D ˇ
ˇ
where we have replaced p.H C En / by Œ p; H C since ' n ˇ.H C En / D 0, and
noted that .H C En / is of the order m˛ 2 . The right-hand side of the above equation
involves the Coulomb Hamiltonian H C , the hydrogenic atomic functions ' n , and
the eigenvalues En , which are all known.
The expression in (5.13.43) may be rewritten, with n D 2, as
" ˇ E D ˇ #
< 2˛ D ˇˇ Ry ˇ ˇ
ˇ E K
ˇ
•E D 2; `ˇŒp; HC ln pˇ2; ` C 2; `ˇŒp; HC p ˇ2; ` ln ;
3m2 jHC En j Ry
(5.13.44)
5.13 Lamb Shift 303
where Ry stands for the Rydberg energy equal to m˛ 2 =2, and note that
D ˇ ˇ E m3 4
ˇ ˇ
2; `ˇŒ p; HC p ˇ2; ` D 2 ˛j' 2 .0/j2 D ˛ ı`;0 ; (5.13.45)
4
as is readily shown.
The following expression, involving the above mentioned known quantities, has
been evaluated numerically by several authors58
D ˇ ˇ E
ˇ Ry ˇ
2; `ˇŒp; HC ln jHC En j pˇ2; `
D ˇ ˇ E D 2:81 ı`;0 C 0:03 ı`;1 : (5.13.46)
ˇ ˇ
2; 0ˇŒ p; HC p ˇ2; 0
Hence

˛m 4 2K
•E< D .˛/ 2:81 C ln ln.˛ 2 / ı`;0 C 0:03 ı`;1 : (5.13.47)
6 m
• The Total Shift •E D •E > C •E < :

From (5.13.20) and (5.13.47), we have

˛m 19 ˛m 1
•E2;` D .˛/4 2:81 ln ˛ 2 ı`;0 C .˛/4 C 0:03 ı`;1 :
6 30 6 8
(5.13.48)
• The Lamb Shift:

The Lamb shift is then given by
•E2;0 •E2;1
D 1052:1 MHz; (5.13.49)
h
and is in good agreement with the experimental value59 which is about 1057.85
MHz. There are other small corrections which contribute to the Lamb shift, such
as contributions coming, for example, from the proton recoil as well as from the
finite structure of the proton, and one has to dwell on some phenomenological
aspects (see also Problem 5.16).
58
See, in particular, Bethe et al. [9] and Schwartz and Tiemann [67].
59
See, e.g., Kinoshita and Yennie [37], p. 7.
5.14 Coulomb Gauge Formulation
In the present section, the vacuum-to-vacuum transition amplitude in the full QED
is derived in the Coulomb gauge, in the differential form, by the application of the
quantum dynamical principle. The path integral expression then also readily follows
from the differential formalism. Gauge transformations of the vacuum-to-vacuum
transition amplitude between the Coulomb and covariant gauges is given in the next
section.
5.14.1 h 0C j 0 i in the Coulomb Gauge in the Functional

Differential Form
The Coulomb gauge is defined by
r A.x/ D @ i A i .x/ D 0; (5.14.1)
for which only physical transverse degrees of freedom of the photon are quantized.
One may apply the Coulomb gauge directly to the Maxwell Lagrangian density,
by solving one of the components of A in terms of the (derivatives) of the other
components, and then varying the other two independent components. We may, for
example, solve for A3 as follows
A3 D @1
3 @ A ;
a a
(5.14.2)
with a sum over a D 1; 2. Accordingly, we may simply write

a
•Ai D ı i a ı i 3 @1
3 @
a
•A ; (5.14.3)
expressed solely in terms of variations of the components A1 ; A2 , as one should.

Now consider the Lagrangian density
1
L D F F C A J : (5.14.4)
4
Variations with respect to Aa , and A0 , respectively, give

@ F a D @31 @ a @ F 3 C J 3 J a ; (5.14.5)
@ F 0 D J 0 ; or @ i F i 0 D J 0 : (5.14.6)
We note that (5.14.5) also holds true if we replace a by 3, since this simply gives the
empty statement @ F 3 D @ F 3 . Accordingly, we may rewrite (5.14.5) as

@ F j D @1
3 @ @ F
j 3
C J3 Jj; .j D 1; 2; 3/: (5.14.7)
5.14 Coulomb Gauge Formulation 305
One may combine (5.14.6) and (5.14.7) as follows

@ F D j @ j @1
3 @ F
3
C J 3 J : (5.14.8)

equation, and using @ @ F D 0, we may, in turn,
By applying @ to the above
1 j 3 3
solve for @3 @ @ F C J :
1
@1
3 @ F
3
C J 3 D 2 @ J : (5.14.9)
r
Upon substituting this equality in (5.14.8) gives
@ j @
@ F D j 2 J : (5.14.10)
r
This leads to the basic equations60
r 2 A0 D J 0 ; (5.14.11)
@i @ j j
Ai D ı ij J: (5.14.12)
r2
These two equations lead further to the equation for the matrix element of the
vector potential
h 0C jA .x/ j 0 i h @ @ @ @ i i @ @ i i i 1
D C 2
2
2 J .x/;
h 0C j 0 i r r r . i
/
(5.14.13)
giving
Z
h 0C jA .x/ j 0 i
D .dy/DC .x y/J .y/; (5.14.14)
h 0C j 0 i

with DC .x y/, specifying the photon propagator in the Coulomb gauge, given by
Z
.dQ/ iQ .xx 0 /
DC .x x 0 / D e DC .Q/; (5.14.15)
.2 /4
1 Q i Qj 1
D00 D0Ci .Q/ D 0:
ij
C .Q/ D 2
; D C .Q/ D ı ij ; (5.14.16)
Q Q 2 Q2 i
60
It is important to realize that the component A0 is not a dynamical variable, i.e., its equation
does not involve its time derivative, as is derived directly from the Lagrangian and need not, a
priori assumed.
Only the components D C .xy/ D D C .xy/ propagate. D 00

ij ji
C .xy/ is proportional
to ı.x 0 y 0 /.
Upon using the fact that
•
h 0C jA .x/ j 0 i D .i/ h 0C j 0 i;
•J .x/
we may functionally integrate (5.14.14) to obtain

hiZ i

h 0C j 0 i D exp .dx/.dx 0 /J .x/DC .x x 0 /J .x 0 / ; (5.14.17)
2
satisfying the B.C. h 0C j 0 i D 1 for J D 0 .

In the presence of the interaction with the fermion field, the Lagrangian density
is taken to be
1 1 @ @
L D F F C m0
4 2 i i
1
C e0 Œ ; A C C C A J ; (5.14.18)
2
from which the following field equations emerge
1 j
@ F D j @ @ .J C j / .J C j /; (5.14.19)
r2
h @ i

e 0 A C m0 D ; (5.14.20)
i
h @ i

C e0 A C m0 D ; (5.14.21)
i
with j D e0 Œ ; =2. Also as in (5.7.32), we have the following useful

equation
h i
@ j .x/ D i e0 .x/.x/ .x/ .x/ : (5.14.22)
From (5.14.17), in particular, we then have, by following the procedure carried

out for the corresponding covariant gauge formulation in Sect. 5.7 .e0 ¤ 0/ :
h Z • • • i
h 0C j 0 i D exp e0 .dx/
•.x/ •.x/ •J .x/
hi Z i

exp .dx/ .dx 0 / J .x/ DC .x x 0 / J .x 0 /
2
Z

exp i .dx/ .dx 0 / .x/ SC .x x 0 / .x 0 / ; (5.14.23)
for the solution of QED in the celebrated Coulomb gauge, where SC . p/ D

1=. pCm0/, in the momentum description, expressed in terms of the bare mass m0 .
The corresponding path integral expression of the vacuum-to-vacuum amplitude is
derived in the next subsection.
The canonical conjugate momenta @L =@.@ 0 Aa / of the fields A1 ; A2 , are given
61
by
a D F a 0 @1 a 30
3 @ F : (5.14.24)
In particular, we note that this is also zero if we replace a by 3: 3 D 0,

corresponding to the dependent field A3 . Also the canonical conjugate momentum
of A0 is zero, i.e., the latter is also a dependent field. Hence we may quite generally
write
D F 0 i @ i @1 30
3 F ; D 0; 1; 2; 3: (5.14.25)
Upon solving for i @ i @1 30

3 F , and taking the @ derivative of the resulting
equation leads to
r 2 @1
3 F
30
D @ @ F 0 D @ C .J 0 C j 0 /; (5.14.26)
where, with D 0, we have used (5.14.19) in writing the last equality. This equation
allows us to solve for F 0 from (5.14.25)
@i 0
F 0 D i 2
J C j 0 C @a a ; (5.14.27)
r
expressed in terms of canonical conjugate momenta, their space derivatives, the
current j 0 and the component of the external source J 0 . Note that F ij may be
expressed in terms of the independent fields A1 ; A2 . For an application of this useful
expression see Problem 5.17.
Equal-time commutation rules of the field components, may be derived as
follows. We recall that
0 D 0; 3 D 0; a D @31 @ a @ 0 A3 @ 0 A a ; a D 1; 2;
and hence
Œ Aa .x 0 ; x/; b .x 0 ; x 0 / D i ı ab ı .3/ .x x 0 /: (5.14.28)
61
Note that the canonical conjugate momentum of Aa is not @0 Aa , for a D 1; 2.
This leads to the equal time commutator (see Problem 5.18)

h @i @ j i
Œ Ai .x 0 ; x/; @ 0 Aj .x 0 ; x0 / D i ı ij 2 ı .3/ .x x 0 /: (5.14.29)
r
In the next section, we consider the transformations rules that relate the full
theory, via the vacuum-to-vacuum transition amplitudes, between the different
covariant gauges as well as with the Coulomb one carried out above.62
5.14.2 From the QDP to the Path Integral of h 0C j 0 i

in Coulomb Gauge
To derive the path integral expression of h 0C j 0 i in the Coulomb gauge we use

the elegant method of delta functionals developed in Sect. 2.7. To this end, we first
establish that for the external current J ,63
• @i @

˘ ı JT D ı i @ k k ˘ ı.J /; .where JT J i 2 J / (5.14.30)
•J r
up to a proportionality constant. Through a series of steps, the left-hand side of the
above equation may be rewritten as
@ i @
˘ ı J i 2 J
r
Z h Z @i @ j i
D ı.J 0 / ˘i Da i exp i .dx/ ai 2 aj J i
r
Z h Z
0
@i i
D ı.J / ˘x d.x/ ˘i Da ı.@ a / exp i .dx/ a i 2 J i
i j j
r
Z Z h Z
• @i i
D ı.J 0 / ı i @ k k ˘x d.x/ ˘i Da i exp i .dx/ ai 2 J i
•J r
Z Z h Z i
•
D ı.J 0 / ı i @ k k ˘x d.x/ ˘i Da i exp i .dx/ a i J i
•J
•
D ı i@ k k ˘ ı.J /; (5.14.31)
•J
62
This section is based on Manoukian [48]. For the relativistic invariance of QED in the Coulomb
gauge, see Manoukian [51]. For the gauge invariance of transition amplitudes, see Manoukian [53].
63
Recall the notation ˘ ı.J / ˘ ˘x ı.J .x//.
R
up to the proportionality constant ˘x d.x/,64 and where in the step before the
last one, we have made a change of variables a i ! a i C @ i =r 2 .
Set
Z
b D 1 .dx/ b
A F .x/ b
F .x/; where b A .x/ i ;
•
(5.14.32)
4 •J .x/
and note that

h • i h • • i
0
; J .x / D ı ı .4/ .x x 0 /; ; D 0: (5.14.33)
•J .x/ •J .x/ •J .x 0 /
The following commutators then easily follow
b ; J .x/ D @ b
ŒiA F .x/; b ; J .x/ D 0;
@ Œ i A (5.14.34)

b ; J .x/ D @ b
ŒiA F .x/; b;ŒiA
iA b ; J .x/ D 0: (5.14.35)
T T
Now we use the convenient notation h 0C j 0 ie0 D0; for the right-hand side
of (5.14.17). In reference to (5.14.10), we have

@ b
F .x/h 0C j 0 ie0 D 0; D J T h 0C j 0 ie0 D 0; : (5.14.36)
Hence the first commutator in (5.14.35) leads to

b ; J .x/ C J .x/ h 0C j 0 ie0 D 0; D 0:
ŒiA (5.14.37)
T T
We use the elementary relation that for two operators A; B, if Œ A; Œ A; B D 0,

then eA B eA D Œ A; B C B. The second commutator in (5.14.35) then allows us
to rewrite (5.14.37) as
b J .x/ exp Œ i A
exp Œ i A b h 0C j 0 ie0 D 0; D 0;
T

JT .x/ exp Œ i A b h 0C j 0 ie0 D0; D 0; (5.14.38)
for all x. Using the property that from xf .x/ D 0, one may infer that f .x/ D
Const. ı.x/, (5.14.38) then leads to
b ˘ ı.J /;
h 0C j 0 ie0 D 0; D exp Œ i A

(5.14.39)
T
64
Don’t let such an infinite proportionality constant scare you. We always divide by proportionality
constants in h 0C j 0 i when determining Green functions.
up to a multiplicative constant. From (5.14.30), we obtain the key equation

b ı i @k • ˘ ı.J /:
h 0C j 0 ie0 D 0; D exp Œ i A (5.14.40)
•J k
The full QED h 0C j 0 i, in the Coulomb gauge, then follows from (5.14.23)
to be
Z Z
h 0C j 0 i D .D D/ .Da/ ı.@k a / exp Œ i .dx/Lc .x/;
k
(5.14.41)
up to an unimportant multiplicative constant. Here Lc is the Lagrangian density,

including the external sources J ; ; , obtained from the one in (5.14.18) upon
carrying out the substitutions
! ; ! ; A ! a ;
where we recall that ; are Grassmann variables, i.e., are anti-commuting (see
also (5.7.41) for the covariant gauges).
This equation is quite interesting as it explicitly shows the gauge constraint via
the delta functional ı.@k a k /.
Note that the expression for h 0C j 0 i in (5.14.23), is a functional of the sources
J ; ; , while the coefficient of
Z

exp i .dx/ .x/.x/ C .x/.x/ C J .x/a .x/ ;
in the functional integrand in (5.14.41) is a functional of .a ; ; /. Thus the path
integral defines a functional Fourier transform of variables .J ; ; / to .a ; ; /.
5.15 Gauge Transformations of the Full Theory
Let us recall the expression for the vacuum-to-vacuum transition amplitude, or the
generating functional, of QED in covariant gauges (see (5.7.20)/(5.7.21)), denoted
here conveniently by FŒ; ; J; :
h Z • • • i
FŒ; ; J ; D exp e0 .dx/
•.x/ •.x/ •J .x/
hi Z i
exp .dx/ .dx 0 / J .x/ D .x x 0 I / J .x 0 / (5.15.1)
2
h Z i
exp i .dx/ .dx 0 / .x/ SC .x x 0 / .x 0 / ; (5.15.2)
5.15 Gauge Transformations of the Full Theory 311
Z
.dk/ h k k i ei k.xx /
0
D .x x 0 I / D .1 / ; (5.15.3)
.2 /4 k2 k2 i
where note that the exponential first factor involving the functional derivatives
above, acts simply as a translation operator:

• •
J ! J C e0 ;
• •
on any functional of J , via the functional derivative •=•J . This basic property
will be used throughout this section.
Note that throughout SC is expressed in terms of the bare mass m0 , i.e. SC . p/ D
1=. p C m0 / in the momentum description.
On the other hand, from the last section, the corresponding expression in
the Coulomb gauge, and with corresponding sources . ; ; K /, is given
from (5.14.23) by
h Z • • • i
FC Œ ; ; K D exp e0 .dx/
•.x/ •.x/ •K .x/
hiZ i

exp .dx/ .dx 0 / K .x/ DC .x x 0 / K .x 0 / (5.15.4)
2
h Z i
exp i .dx/ .dx 0 / .x/ SC .x x 0 / .x 0 / ; (5.15.5)
Z
0 .dk/ 0
DC .x x /D D .k/ ei k.xx / ;
.2 /4 C
h ki kj i 1 1
; D 0C0 .k/ D 2 ; D iC0 .k/ D 0 D D 0Ci .k/:
ij
D C .k/ D ij 2
k k2 i
k
(5.15.6)
Note the identity relating the free photon propagator in these gauges65
@ i @j ˛ˇ
˛ i 2 @˛ ˇ j 2 @ˇ D .x x 0 I / D DC .x x 0 /: (5.15.7)
r r
Also note the following invariance property of the expression
• • •

; (5.15.8)
•.x/ •.x/ •J .x/
65
See Problem 5.19.
under the transformations
.x/ ! .x/ ei.x/ ; .x/ ! ei.x/ .x/; (5.15.9)
for an arbitrary numerical function .x/.

We first derive the transformation rule relating the full theory in the Coulomb
gauge to the full one in covariant gauges specified by the parameter . This reads:
h 0 iˇˇ
FC Œ ; ; K D eiW FŒ; ; J ; ˇˇ ; (5.15.10)
D 0; D 0;J D 0
Z h i
0 • •
W D .dx/ .x/ exp i e0 c
i•J .x/ i•.x/
Z h i
• •
C .dx/ exp i e0 c .x/
.i/•.x/ i•J .x/
Z h i •

C .dx/ c @ K .x/ ; (5.15.11)
i•J .x/
@j
c D j : (5.15.12)
r2
To derive this, we begin with the right-hand side of (5.15.10). In a matrix
notation, first note that
0
hi i
eiW exp i SC exp J D ./ J
2
h i h i
• •
D exp i C exp i e0 c SC C exp i e0 c
i•J i•J

i
exp J C c @ K D ./ J C ˇ c @ˇ K ˇ : (5.15.13)
2
On the other hand, since we eventually, and in particular, set D 0; D 0, we may

consider the transformations in (5.15.9), and use the fact that
h Z • • • i
exp e0 .dx/
•.x/ •.x/ •J .x/
h Z • • • i
D exp e0 .dx/ ei.x/ ei.x/ ; (5.15.14)
•.x/ •.x/ •J .x/
for an arbitrary numerical .x/. Hence for D 0; D 0, in view of applying

the analysis to the right-hand side of (5.15.10), we may, from (5.15.13), write the
right-hand side of (5.15.10), before setting J D 0, as

h Z • • • i
exp e0 .dx/
•.x/ •.x/ •J .x/
h i h i
• •
exp i exp i e0 c SC exp i e0 c
i•J i•J

i ˇ
exp J C c @ K D ./ J C ˇ c @ˇ K ; (5.15.15)
2
rewritten in terms of the source variables ; . We eventually have to set J D

0 in this equation to compare the resulting expression with the one on the left-hand
side of (5.15.10). To do this, we have to carry out the functional derivatives •=•J
in (5.15.15) explicitly first. To this end, we may use the identity
h Z • • i
exp .dx/ @ i .x/ exp Œ i SC
•.x/ •.x/

D exp i ei SC ei ; (5.15.16)
for an arbitrary numerical .x/, established in Problem 5.20, and making the
identification
i D i e0 c .•=i •J /;
to rewrite the expression in (5.15.15), for J D 0, as

h Z • • • i

exp e0 .dx/ c @ exp i SC
•.x/ •.x/ •J .x/
ˇ
i
ˇ ˇˇ
exp J C c @ K D ./ J C ˇ c @ˇ K ˇ
2 ˇ
J D0
h Z i
• • •

D exp e0 .dx/ exp i SC
•.x/ •.x/ •K .x/

i
exp c @ K D ./ ˇ c @ˇ K ˇ : (5.15.17)
2
Now we may use (5.15.7) to conclude the statement made in (5.15.10).

To establish gauge transformations, say, from the Landau gauge, to covariant
gauges specified by the parameter , we note that
@ @ @ @ˇ ˇ
D ./ ˇ D Dˇ ./ˇD0 ; @˛ @˛ ; (5.15.18)

where
ˇ
D ./ˇD0 D L ;

is the celebrated free photon propagator in the Landau gauge:

Z
.dk/ i k.xx 0 /
k k 1
D L .x x 0 / D e 2 : (5.15.19)
.2 /4 k k2 i
An almost identical analysis as above, establishes the rule relating the gauge
transformations of the full theory from covariant ones, specified by , to Landau’s.
This reads:
h 0 iˇ
ˇ
FŒ; ; J ; D 0 D eiW FŒ; ; K ; ˇ
; (5.15.20)
D 0; D 0;K D 0
where W 0 is the functional operator defined in detail by

Z h • i •
W0 D c
.dx/.x/ exp i e0e
i•K .x/ i•.x/
Z h
• • i
C .dx/ exp i e0e c .x/
.i/•.x/ i•K .x/
Z h i •

C .dx/ e c @ J .x/ ; (5.15.21)
i•K .x/
@
c D
e : (5.15.22)

By putting D 0; D 0; K D 0 in (5.15.10), we immediately learn
that in the absence of external sources, FC Œ 0; 0; 0 D FŒ 0; 0; 0; . Similarly,
from (5.15.20) and (5.15.21), we learn that FŒ 0; 0; 0; 0 D FŒ 0; 0; 0; . Thus
we have the following gauge invariance normalization conditions valid when the
external sources are set equal to zero
FC Œ 0; 0; 0 D FŒ 0; 0; 0; D FŒ 0; 0; 0; 0 : (5.15.23)
In the absence of external Fermi sources only, as shown in Problem 5.21, we also
have
h i i
FŒ 0; 0; J ; D 0 D exp @ J / G @ J FŒ0; 0; J ; ; (5.15.24)
2
Z 0
.dk/ ei k.xx / 2
G.x x 0 / D ; (5.15.25)
.2/4 k2 .k2 C 2 / k2 C 2
where we have defined G.x x 0 / with infrared and ultraviolet cut-offs (see
also (5.10.11) and (5.10.27)).66
The relation of the exact electron propagator in covariant gauges, specified by the
parameter , to the exact one in the Landau gauge, in the presence of the external
current J , is readily worked out and, as shown in Problem 5.22, is given by
ˇ
S.x; y I J /ˇ D exp i e20 Œ G.0/ G.x y/
e
h Z
i
ˇ
exp i e0 .dz/J .z/ @z G.z x/ G.z y/ S.x; y I J /ˇD0 :
e
(5.15.26)
In particular, in the absence of the external current J D 0,

ˇ

ˇ
S.x; y/ˇ D exp i e20 G.0/ G.x y/
e S.x; y/ˇD0 :
e (5.15.27)
For jx y j ! 1 in Euclidean region we have from (5.11.10) and (5.15.25),
˛0 2
i e20 Œ G.0/ G.x y/ ! ln 2 ; (5.15.28)
4
and near the mass shell of the electron propagators, we then have for the wavefunc-
tion renormalization constants in question
ˇ 2 ˛0 =4 ˇ
ˇ ˇ
Z 2ˇ D Z 2ˇ : (5.15.29)
2 D0
Let us evaluate this to lowest order and compare it with our earlier result. To lowest
order, this gives
ˇ ˇ ˛ 2
ˇ ˇ
Z 2ˇ D Z 2ˇ C ln 2 ; (5.15.30)
D0 4
which coincides with our earlier low order result in (5.10.36) expressed there for
1=Z 2 .
This section on gauge transformations, is based on Manoukian and Siranan [60],
which includes extensive references on the gauge problem. It also includes detailed
treatments of the so-called axial gauge n A D 0, where n is a fixed vector, as
well as the Fock-Schwinger gauge x A D 0, in the full theory.67
66
This is the Fourier transform of the longitudinal part of the free photon propagator D , in the
momentum description, with cut-offs, written as k k 2 =Œk2 .k2 C 2 /.k2 C 2 /.
67
For the gauge invariance of transition amplitudes see Manoukian [53].
Before closing this section, we also provide the following useful and easily
derived68 identity of the generating functional in covariant gauges, specified by the
parameter . It reads in matrix form

i i

F %; %; e
J ; D exp Œ i @ ' e
J exp ' ' F ; ; J ; ; (5.15.31)
2
e @
% D ei e0 ' ; % D ei e0 ' ; J D J '; (5.15.32)

where '.x/ is a real numerical dimensionless function. We will make use of this
identity in the next section.
5.16 Vertex Function and Ward-Takahashi Identity; Full

Propagators
In this section, going beyond perturbation theory, a classic identity, referred as

the Ward-Takahashi identity, is derived which relates the derivative of the vertex
function to the exact electron propagater, and establishes, in particular, the
equality of the renormalization constants Z1 D Z 2 to all orders. This is followed
by deriving the equations satisfied by the electron propagator and vertex function
without recourse to perturbation theory. The spectral decomposition of the photon
propagator is then derived, generalizing the lowest order derivation given in
Sect. 5.10.2, Eq. (5.10.55), with much emphasis put on charge renormalization.
Finally an integral equation is derived for the vacuum polarization tensor.
5.16.1 Ward-Takahashi Identity
The generating functional of QED, in covariant gauges, specified by the parameter

in the presence of an arbitrary scalar function '.x/ was given above in (5.15.31).
The electron propagator follows from this identity, by functional differentiations
defining the propagator, and is given by
ˇ

e S.x; y/ˇ' D 0 exp i e0 '.x/ '.y/ ;
S.x; y/ D e (5.16.1)
where we have set D 0; D 0. Also note that under a variation •' of '
Z

•e
S.x; y/ D i e0 .dz/ ı .4/ .z x/ ı .4/ .z y/ e
S.x; y/ •'.z/: (5.16.2)
68
The derivation follows very closely to the ones in Problems 5.21 and 5.22.
5.16 Vertex Function and Ward-Takahashi Identity; Full Propagators 317
Similarly69 we have
ˇ
hA .x/i D hA .x/iˇ'D0 C @ '.x/; •hA .x/i D @ •'.x/: (5.16.3)
The vertex function in the full theory is defined through

Z
• e
S.x; y/ D e0 .dx 0 / .dy 0 / e
S.x; x 0 / .x 0 ; y 0 I z/ e
S.y 0 ; y/; (5.16.4)
•hA .z/i
which will be justified in Sect. 5.16.2, and coincides with to zeroth order. Hence
from the chain rule, we obtain
Z
•
•e
S.x; y/ D .dz/ e
S.x; y/ •hA .z/i
•hA .z/i
Z
D e0 .dx 0 / .dy 0 / .dz/ e S.x; x 0 / .x 0 ; y 0 I z/ e
S.y 0 ; y/ @z •'.z/
Z
D e0 .dx 0 / .dy 0 / .dz/ e
S.x; x 0 / @z .x 0 ; y 0 I z/ e
S.y 0 ; y/ •'.z/: (5.16.5)
Upon comparing this equation with that in (5.16.2), and multiplying, in the process,
both equations from the right and the left by the inverse of the propagator e S, we
obtain
@
1

.x; y I z/ D i ı .4/ .z x/ ı .4/ .z y/ e
S .x y/; (5.16.6)
@z
where all the external sources as well as ' have been set equal to zero.
Introducing the Fourier transform
Z
.dp/ .dk/ i . pCk/x ipy i kz
.x; y I z/ D e e e . p C k; p/; (5.16.7)
.2/4 .2/4
we obtain the identity
k . p C k; p/ D e
S1 . p C k/ e
S1 . p/; (5.16.8)
referred to as the Ward-Takahashi Identity.70 Upon taking the derivative of this

equation with respect to k , and taking the limit k ! 0, we obtain the zero

69
Recall that hA .x/i .i/ • F=•J .x/ =F.
70
Takahashi [78]. The identity in differential form, that is for zero momentum transfer of the vertex
function, is due to Ward [79]. Takahashi generalized it to non-zero momentum transfer.
momentum transfer version of the above identity given by
@ e1
. p; p/ D S . p/; (5.16.9)
@p
referred to as the Ward Identity. The wavefunction renormalization constant Z 2

of the electron propagator, and the vertex renormalization constant Z 1 , may be
defined through boundary conditions on the mass-shell p 2 D m2 , where m is
the renormalized mass, as follow

@ 1 1 1
u.p; / e
S . p/ u.p; / D ; u.p; / . p; p/u.p; / D :
@p Z2 Z1
(5.16.10)
These give the equality of the two renormalization constants
Z1 D Z 2: (5.16.11)
We have seen earlier in Sect. 5.11, that to lowest order, these two renormalization
constants are equal. This is now seen to be true to all orders, and is a consequence
of the Ward Identity. Due to this equality, charge renormalization occurs via the
wavefunction renormalization constant Z 3 of the photon only. This universality
property will be discussed in the next section.
In Sect. 5.16.2, the equations satisfied by the electron propagator and the vertex
function are derived. Here the definition of the vertex function through (5.16.4)
will be used. We have found it convenient to carry out a similar analysis for the
vacuum polarization and the photon propagator, right in this section, in the following
two subsections. In Sect. 5.16.3, the photon spectral decomposition of the photon
propagator is carried out, generalizing the low order one in (5.10.55). The equation
satisfied by the vacuum polarization is derived in Sect. 5.16.4.
5.16.2 Equations for the Electron Propagator and the Vertex

Function
Upon taking the matrix element of the electron field between the vacuum states, we
have
h @ 1 • i

e0 C m 0 h 0C j .x/ j 0 i D .x/h 0C j 0 i: (5.16.12)
i i •J .x/
The functional derivative •=•.y/ of this equation gives

h @ 1 • i

e0
C m 0 ih 0C j .x/ .y/ C j 0 i D ı .4/ .x y/h 0C j 0 i;
i i •J .x/
(5.16.13)
where now we have set the external Fermi sources equal to zero. Using the identity
1 • h 1 • i
h 0 C j 0 i D h 0 C j 0 i hA .x/i C ; (5.16.14)
i •J .x/ i •J .x/
we may rewrite the above equation as
h @ 1 • e

e0 hA .x/i C m0 eS.x; y/ D ı .4/ .x y/ C e0 S.x; y/:
i i •J .x/
(5.16.15)
Upon multiplying the latter from the right by Se1 .y; x 0 /, and integrating over y,
gives
h @
e
S1 .x; x 0 / D e0 hA .x/i C m0 ı .4/ .x x 0 /
i
Z
1 •
C e0
.dy/ e S.x; y/ e S1 .y; x 0 /; (5.16.16)
i •J .x/
where the equality

Z Z
.dy/ .•e
S.x; y// e
S1 .y; x 0 / D .dy/ e
S.x; y/ •e
S1 .y; x 0 /;
has been used. Using the definition of the exact photon propagator
• e .x; z/;
hA .z/i D D (5.16.17)
•J .x/
which will be treated in the next section, and the chain rule
Z
• e1 e .x; z/ • e
S .y; x 0 / D .dz/D S1 .y; x 0 /; (5.16.18)
•J .x/ •hA .z/i
we obtain from (5.16.16)

h @

e
S1 .x; x 0 / D e0 hA .x/i C m0 ı .4/ .x x 0 /
i
Z
•
i e0 .dy/.dz/ e S.x; y/ e e .x; z/:
S1 .y; x 0 /D (5.16.19)
•hA .z/i
This provides an iterative equation in terms of the exact electron and photon
propagators, starting with
h @

e
S1 .x; x 0 / D e0 hA .x/i C m0 ı .4/ .x x 0 /
i
Z
C i e20 .dy/.dz/ e e .x; z/ C :
S.x; y/ ı .4/ .y z/ ı .4/ .y x 0 /D
(5.16.20)
It is easily verified that this coincides with the lowest order expression in (5.10.4),
for
J ! 0; hA .x/i ! 0; e0 ! e;
and with the substitutions
e
S.x; y/ ! SC .xy/; e ! D ;
D
made inside the integral.

Using the definition of the vertex function in (5.16.4), and the Fourier transform
in (5.16.7), we obtain, in the absence of the external sources, the exact equation
e
S1 . p/ D p C m0 ˙. p/; (5.16.21)
Z
.dk/ e e .k/:
˙. p/ D i e20 S. p C k/ . p C k; p/ D (5.16.22)
.2/4
where from (5.16.4), we have used, in the process, the relation
• e
S1 .x; y/ D e0 .x; y I z/: (5.16.23)
•hA .z/i
In analogy to ˙ in (5.16.22), in the absence of external sources, we will still denote

the second term in (5.16.19), in the presence of the external current J by ˙ as
well :
Z
˙.x; x 0 / D i e20 .dy/.dz/ e e .x; z/:
S.x; y/ .y; x 0 I z/D (5.16.24)
˙.x; x 0 / is a functional of the exact propagators e e . By the chain rule

S, D
Z
• •D .y; y 0 / •
˙.x; x 0 / D .dy/.dy 0 / ˙.x; y/
•hA .z/i •hA .z/i •D .y; y 0 /

Z e y 0/
•S.y; •
C .dy/.dy 0 / ˙.x; y/: (5.16.25)
e
•hA .z/i •S.y; y 0 /
From the definition of the exact photon propagator in (5.16.17), the first term
necessarily includes an odd number (three) of external photon lines, for J D
0; hA i D 0. Hence from Furry’s Theorem,71 the first term in the above equation is
zero in the absence of external sources. At this stage, we introduce the e eC kernel
K defined by
•
˙.x; x 0 / D e20 K.x; x 0 I y 0 ; y/: (5.16.26)
•e
S.y; y 0 /
The factor e20 is introduced to keep track of the order when carrying out a perturba-
tion analysis. Upon taking the function derivative, via .•=•hA .z0 /i/, of (5.16.19),
i.e., of
h @

e
S1 .x; x 0 / D e0 hA .x/i C m0 ı .4/ .x x 0 / ˙.x; x 0 /; (5.16.27)
i
and using (5.16.23) and (5.16.25), we obtain, after dividing the resulting expression
by . e0 /
.x; x 0 I z/ D ı .4/ .z x/ ı .4/ .x x 0 /

Z
i e20 .dy/.dy 0 /.dx 00 /.dy 00 /K.x; x 0 I y 0 ; y/ e
S.y; x 00 / .x 00 ; y 00 ; z/ e
S.y 00 ; y 0 /:
(5.16.28)
Upon defining the following Fourier transform, and using (5.16.7)

Z
.dp/ .dp 0 / .dQ/ 0 0 0 0
K.x; x 0 I y 0 ; y/ D 4 4 4
K. p; p 0 I Q/ ei x. pCQ/ ei y. p CQ/ ei x p ei y p ;
.2/ .2/ .2/
(5.16.29)
we obtain from (5.16.28) the integral equation for the vertex function (Fig. 5.15):
Z
.dp0 /
. p C Q; p/ D i e20 K. p; p0 I Q/ e
S. p0 C Q/ . p0 C Q; p0 / e
S. p0 /:
.2/4
(5.16.30)
By definition, the e eC -kernel K is connected and the external electron lines are
removed. Some graphs which are excluded in the kernel K are given in Fig. 5.16.
The e eC -kernel times e20 :e20 K may be expanded in terms of the exact
propagators and the exact vertex function as shown in Fig. 5.17. Such an expansion
is quite useful in investigating aspects of the renormalized theory.
71
See Appendix IV, below (IV.19), at the end of the book.
p Q p Q
p Q p Q
Q = Q − ie20 K Q
p
p
p p
Fig. 5.15 The integral equation for the vertex function. Note the momentum flow into the e eC -
kernel
(a) (b) (c)
p p p p p p
Fig. 5.16 The kernel K cannot be composed of two parts connected by a single photon line as
shown in part (a). It cannot also be composed of two parts connected by one electron-positron pair
as shown in part (b). The diagram in part (c), as well as those connected by an even number of
photon lines, cannot contribute to the integral equation of the vertex function as they will give rise
to subgraphs involving an odd number of external photon lines to the vertex in Fig. 5.15 which
vanish by Furry’s theorem
p+Q p +Q p+Q p +Q
p+Q p +Q e0 e0 e0
e20 = +
p e0 e0 e0
p
p p p p
Fig. 5.17 The e eC - kernel times e20 :e20 K expanded in terms of the exact propagators and the
exact vertex function, referred to as a skeleton expansion. The arrows show the direction of
momentum flow, and recall that the external electron-positron lines are removed in the definition
of K
Finally we note from (5.16.24) and (5.16.6), we may also represent the integral
equation for e
S1 diagrammatically as in Fig. 5.18 (for J ! 0)72 :
72
Functional methods in deriving equations for propagators and vertex functions were particularly
used over the years by Schwinger and by his former students, notably by K. Johnson (see, e.g.,
[34]), and by many others.
S−1(p) = γ p + m0 −
p k
Fig.
P 5.18 Integral
e1
P equation of the inverse of the exact electron propagator S . p/ D p C m0
. p/, where . p/ is defined in (5.16.22)
5.16.3 Spectral Representation of the Photon Propagator,

Charge Renormalization, Coulomb Potential in Full
QED, Unordered Products of Currents
In order to derive73 a spectral representation74 of the full photon propagator, as we

did in (5.10.53), (5.10.54) and (5.10.55) to lowest order, we begin with the theory in
the Coulomb gauge. The reason why we consider first the Coulomb gauge, is that in
this gauge only physical degrees of freedom are quantized and one is dealing with
an underlying vector space with a positive definite metric. From such an analysis,
one may then infer the structure of the photon propagator in covariant
gauges.

We first note that due to the simple fact that Œei.x/ •=•.x / Œei.y/ •=•.y/
is independent of for x D y, we learn from (5.15.10), (5.15.11), (5.15.12),
(5.15.16), (5.15.17), and similarly, from (5.15.20), (5.15.21), (5.15.22), (5.15.18),
as well as from (5.15.23), that
ˇ
1 Y • • ˇ
n
j FC ˇˇ
FC j D1 •.xj / •.xj / D0;D0;J D0
Y ˇ
• ˇˇ
n
1 • j
D F ˇ : (5.16.31)
F j D1 •.xj / •.xj / D0;D0;J D0;
In particular, this means that the current-current correlation function satisfies the
relation75
ˇ ˇ
ˇ ˇ
h j .x/j .y/ iˇ D h j .x/j .y/ iˇ : (5.16.32)
C Coulomb C GaugeŒ
73
The details of the derivation of the spectral representation may be omitted in a first reading.
74
The spectral representation of a propagator, in general, is usually attributed to Källen [35] and
Lehmann [42], and is referred to as a Källen-Lehmann representation.
75
h : i D h 0C j : j 0 i=h 0C j 0 i, with now all the external sources set equal to zero.

That is, h j .x/j .y/ i is both covariant and gauge invariant.
C
Consider the vacuum expectation value of the product of two currents j .x/j .x0 /,
written in terms of a completeness expansion
X 0
h vacj j .x/j .x 0 /jvaci D eip n .xx / h vacj j .0/jnihnj j .0/jvaci
n
Z
.dQ/ iQ .xx 0 /
D e ™.Q0 /˙ .Q/; (5.16.33)
.2/3
X
˙ .Q/ D .2/3 ı .4/ .Q pn /h vacj j .0/jnihnj j .0/jvaci
n
h i
D Q Q Q2 H.Q2 /; (5.16.34)
where H.Q2 / is a scalar and gauge invariant, Q2 0; Q0 0. In writing the

last equality above, we have used the fact that the current j is conserved in the
absence of external sources (see, in particular, (5.7.32) and (5.14.22)).
Now consider the matrix element hvacjAi .x/Aj .x 0 /jvaciC , in the Coulomb gauge,
in the absence of external sources, to relate it to the vacuum expectation of the
currents above. Using a completeness relation, we may also express the latter matrix
element as
X 0
h vacjAi .x/Aj .x 0 /jvaciC D eip n .xx / h vacjAi .0/jniChnjAj .0/jvaciC
n
Z
.dQ/ ij Q i Qj iQ .xx 0 /
D ı e ™.Q0 /.Q/;
.2/3 Q2
(5.16.35)
where we have used the transversality of the vector field A : @i Ai .x/ D 0, and
where
.2/3 X .4/
.Q/ D ı .Q pn /h vacjA.0/jniC hnjA.0/jvaciC 0: (5.16.36)
2 n
The important point to note here is that the matrix element h A.x/ A.x 0 /iC , is
computed in a positive definite metric space.
To discuss the physical content of this equation, we write
X Z X X
d3 Q 0
1D jnihnj D jQ ihQ j C jnihnj; (5.16.37)
2 jQj.2/3
n n
where the first term on the extreme right-hand side of the equation stands for a
single photon contribution, and the parameter here, specifies the polarization of a
photon,
h Q jQ0 0 i D .2/3 2 jQj ı .3/ .Q Q0 / ı 0 : (5.16.38)

P0
Also n jnihnj corresponds to all possible remaining states. In particular, we may
write
p
h Q jA.0/jvaci D Z 3 e .Q/: (5.16.39)
with the e .Q/ representing polarization vectors, satisfying
Q e .Q/ D 0; e .Q/ e 0 .Q/ D ı 0 ; ; 0 D 1; 2; (5.16.40)

X Q i Qj
j
ei .Q/ e .Q/ D ı ij : (5.16.41)
D1;2
Q2
The single photon contribution to .Q/ in (5.16.37), (5.16.38) and (5.16.39),

follows from (5.16.36), to be simply Z 3 ı.Q2 /.
That is, we may write
.Q/ D Z 3 ı.Q2 / C .Q/; (5.16.42)
where .Q/ 076 denotes the remaining contribution to .Q/.

Now we use the field equation in the absence of external sources in the Coulomb
gauge
@i @k
Ai .x/ D ı ik 2 j k .x/; (5.16.43)
r
to relate .Q/ to H.Q2 / in (5.16.34). The above field equation leads from
(5.16.35) to
./. 0 / h vacjA.x/ A.x 0 /jvaciC

Z
.dQ/ iQ .xx 0 /
D2 e ™.Q0 /.Q2 /2 .Q/
.2/3
@i @` 0 @ 0i @ 0`
0
0
D ıi ` 2 ıi ` h vacj j ` .x/j ` .x 0 /jvaciC : (5.16.44)
r r 02
76
See (5.16.36).
Hence upon using the identity

Q i Q ` i ` 0 Q i Q ` ` ` 0
0
2 `` 0
ıi ` ı Q Q Q ı D 2 Q2 ; (5.16.45)
Q2 Q0 2
Eqs. (5.16.33) and (5.16.34) lead to .Q2 0; Q0 0/
0 .Q2 /2 .Q/ D .Q2 /2 .Q/ D .Q2 /H.Q2 /; (5.16.46)

or .Q/ D H.Q2 /=.Q2 / .Q2 /; (5.16.47)
where in the first equality in (5.16.46), we have used the fact that .Q2 /2 ı.Q2 / D 0.
Thus .Q2 / 0 is a scalar and gauge invariant.
We may rewrite (5.16.33) as
Z 1 Z
0 2 2 .dQ/
hvacj j .x/j .x /jvaciC D dM H.M /
0 .2/3
ih 0
ı.M C Q / Q Q Q2 eiQ .xx / ™.Q0 /:
2 2
(5.16.48)
Upon multiplying (5.16.48) by ™.x 0 x 0 0 / and repeating the derivation of this

equation for hvacj j .x 0 /j .x/jvaciC , as well as multiplying it by ™.x 0 0 x 0 /, and
finally adding the two contributions, we obtain

h vacj j .x/j .x 0 / jvaciC
C
Z 1 Z h i
.dQ/
D dM 2 H.M 2 / Q
Q Q 2
ı.Q2 C M 2 /
0 .2/4
h 0 0
i
™.Q0 / eiQ .xx / ™.x 0 x 0 0 / C eiQ.x x/ ™.x 0 0 x 0 / ;
Z 1 Z
.dQ/ h i
D dM 2 H.M 2 / Q Q Q 2
ı.Q2 C M 2 /
0 .2/4
0
h i
eiQ .xx / ™.Q0 /™.x 0 x 0 0 / C ™.Q0 /™.x 0 0 x 0 / ; (5.16.49)
or77

h vacj j .x/j .x 0 / jvaciC
C
Z Z
1
.dQ/ h i eiQ .xx 0 /
D i dM 2 H.M 2 / Q Q Q 2
: (5.16.50)
0 .2/4 Q2 C M 2 i
77
Note that .Q2 C M 2 i
/ in the denominator, means that Q2 and M 2 may be interchanged,
in a function of Q2 or M 2 multiplying 1=.Q2 C M 2 i
/, when applying the residue theorem
for x0 7 x 0 0 . Also note that the integrands in (5.16.50), (5.16.51) are even function in the variable
Q . See also Brown [14], pp. 461–463.
From (5.16.35) and (5.16.47), we may similarly write the corresponding expres-
sion for the field components Ai :
Z
.dQ/ ij Qi Qj iQ .xx 0 /
h vacj Ai .x/Aj .x 0 / jvaciC D i 2 e
C .2/4 Q
Z 1
.M 2 /
dM 2 2 : (5.16.51)
0 Q C M 2 i
In particular for x 0 > x 0 0 ,
Z Z
@i@ j 1 0
0 2 2 d3 Q eiQ .xx /
h A .x/A .x /iC D ı 2
i j ij
dM .M / p ;
r 0 .2/3 2 Q2 C M 2
(5.16.52)
p
Q0 D Q2 C M 2 . Hence
Z Z
i ij @i @ j 1
d3 Q iQ .xx 0 /
h Ai .x/AP j .x 0 /iC D ı 2 dM 2 .M 2 / e : (5.16.53)
2 r 0 .2/3
Similarly, for x 0 0 > x 0 , we have

Z Z
i ij @i @ j 1
d3 Q iQ .x 0 x/
h AP j .x 0 /Ai .x/iC D ı 2 dM 2 .M 2 / e : (5.16.54)
2 r 0 .2/3
Upon taking the limit x 0 0 ! x 0 0 in (5.16.53), and the limit x 0 0 ! x 0 C 0

in (5.16.54), and subtracting the resulting latter equation from the former resulting
one, we obtain
Z 1
@i @ j
h ŒAi .x0 ; x/; AP j .x0 ; x 0 /iC D i ı ij 2 ı .3/ .xx 0 / dM 2 .M 2 /: (5.16.55)
r 0
Hence from the equal-time commutation relations (5.14.29), this gives

Z 1 Z 1
2 2
dM .M / D 1; Z3 C dM 2 .M 2 / D 1; 0 Z3 1;
0 0
(5.16.56)
with Z3 denoting the probability that the field A creates a photon out of the vacuum.
At this stage, we note that the free photon propagator, in the Coulomb gauge,
may be written as (see (5.14.13)) .c D i @i =r 2 /
h @ @ i 1

D C .x y/ D C 2
@
c @
c ı .4/ .x y/; (5.16.57)
r . i
/
h @ @ @ i @ i @ i @ i i 1
h 0C jA .x/ j 0 iC D C
r2 r2 r2 i

J .x/h 0C j 0 iC C h 0C j j .x/ j 0 iC : (5.16.58)
The matrix element of the vector potential between the vacuum state, in the
presence of external sources, which will eventually be set equal to zero, then follows
to be
Z h i

h 0C jA .x/ j 0 iC D .dy/ D C .x y/ J .y/h0C j 0 iC C h 0C j j .y/ j 0 iC :
(5.16.59)
We recall the equation giving the equation for @ j .x/ in (5.14.22), expressed in
terms of the expectation values between the vacuum state, in the presence of external
Fermi sources:

h 0C j@ j .x/ j 0 iC D i e0 h0C j .x/.x/ .x/ .x/ j 0 iC : (5.16.60)
Hence
• • ˇ
ˇ
e0 .i/ .i/ h 0 C j@ j .x/ j 0 i C ˇ
•.x 0 / •.x 0 / D 0; D 0
h i

D i e20 ı .4/ .x x 0 /h 0C j 0
b .x / a .x/ C 0
a .x / b .x/ j 0 iC ab D 0:
C C
(5.16.61)
Upon carrying out the functional differentiations of (5.16.59), with respect to

• •
.i/ .i/ ;
•.x 0 / •.x 0 /

and upon using, in the process, the expression for DC in (5.16.57), the above
equation leads for J ! 0 to .c D i @i =r 2 /

h 0C j A .x/j .x 0 / j 0 iC
C
1
D ı @ c h 0C j j .x/j .x 0 / j 0 iC : (5.16.62)
. i
/ C
On the other hand, in the absence of external Fermi sources, we may use the
conservation of the current @ j D 0, to rewrite (5.16.59) as
Z

h 0C jA .x/ j 0 iC D .dy/ DC .x y/J .y/h 0C j 0 iC
1
C ı @ c h 0C j j .x/ j 0 iC : (5.16.63)
. i
/
Therefore by taking the functional differentiation of this equation, with respect to,
•=•J .y/, and setting J D 0, and then using (5.16.62), we finally obtain for the
full photon propagator in the Coulomb gauge the expression
e
D 0 0
C .x x / D DC .x x /

1
1
C ı ˛ @ c˛ ı ˇ @ cˇ i h j ˛ .x/j ˇ .x 0 / iC :
. i
/ . i
/ C
(5.16.64)
Using the identity

h @ @ i
ı ˛ @ c˛ ı ˇ @ cˇ @˛ @ˇ ˛ˇ D ./ C 2
@ c @ c ;
r
(5.16.65)
and
Z
i .dQ/ h i 0
hvacj j .x/j .x 0 / jvaci D Q Q Q 2
eiQ .xx /
C .2/4
Z 1
M 2 .M 2 /
dM 2 2 C 2 ; (5.16.66)
0 Q .Q M i
/
we obtain
Z 1

1
e .x x 0 / D C @ @ @ c @ c
D dM 2
C
r2 ./ 0
h i Z
.dQ/ iQ .xx 0 / M 2 .M 2 /
Z3 ı.M 2 / C .M 2 / ı .4/ .x x 0 / e ;
.2/4 .Q2 C M 2 i
/
(5.16.67)
where we have
conveniently multiplied ı .4/ .x x 0 /, within the curly brackets, by
R1
1 D 0 dM 2 Z 3 ı.M 2 / C .M 2 / (see (5.16.56)), or .c D i @i =r 2 /
h @ @ i
e
D .x x 0
/ D
C @
c @
c h.x x 0 /; (5.16.68)
r2
C
Z Z
.dQ/ iQ .xx 0 / 1 .M 2 /
h.x x 0 / D 4
e dM 2 2 ; (5.16.69)
.2/ 0 .Q C M 2 i
/
where we have used the definition .M 2 / D Z3 ı.M 2 /C.M 2 / and M 2 ı.M 2 / D 0.
The free photon propagator in a covariant gauge, specified by a parameter ,
may be simply written from (5.3.9), as
h @ @ i 1
D .x y/ D .1 / ı .4/ .x y/: (5.16.70)
. i
/
The matrix elements of the vector potential between the vacuum states, in
covariant gauges, then follows from (5.7.2)/(5.7.6), to be
Z h i

h 0C jA .x/ j 0 i D .dy/ D .x y/ J .y/h0C j 0 i Ch 0C j j .y/ j 0 i :
(5.16.71)
Upon
carrying out the functional
differentiations of this equation with respect to
.i/ •=•.x 0 / .i/ •=•.x 0 / , the above equation leads for J ! 0 to
1
h 0C j A .x/j .x 0 / j 0 i D h 0C j j .x/j .x 0 / j 0 i:
C . i
/ C
(5.16.72)
Hence by taking the functional differentiation of (5.16.71), with respect to, •=•J .y/,
and setting J D 0, and then using (5.16.72), we finally obtain for the full photon
propagator in covariant gauges
1 1
e .x x 0 / D D .x x 0 / C
D i h j .x/j .x 0 / i:
. i
/ . i
/ C
(5.16.73)
From (5.16.50), (5.16.47), (5.16.70), and by following a similar method as

carried out in (5.16.66) and (5.16.67), the photon propagator in the different
covariant gauges takes the form
h i
e .x x 0 / D @ @ h.x x 0 / C @ @
D
1
ı .4/ .x y/;
. i
/
(5.16.74)
where h.x x 0 / is defined in (5.16.69). Note that the longitudinal part of the
propagator does not change with the interaction.
In reference to (5.16.69), we may set
Z Z 1
0 .dk/ iQ .xx 0 / d.Q2 / d.Q2 / .M 2 /
h.x x / D e ; D dM 2 :
.2 /4 Q2 Q2 0 .Q2 C M 2 i
/
(5.16.75)
Equation (5.16.56) then leads to the following formal boundary conditions for
Q2 ! 1 W
˛
d.Q2 / ! 1; ˛ dren .Q2 / d.Q2 / ! ˛ 0 : (5.16.76)
Z3
Charge Renormalization and the Coulomb Potential
In the absence of radiative corrections, the Coulomb interaction between static

charges,78 each of charge q, may be extracted, in the Coulomb gauge, from the
static part D C0 0 .Q/, of the photon propagator as follows:
Z
d3 Q iQx 0 q2
U.jx x 0 j/ D q2 3
e D C 0 .Q/jQ 0 D 0 D : (5.16.77)
.2/ 4jx x 0 j
In the presence of radiative corrections, in the Coulomb gauge, D 0C 0 .Q/

is modified to D e 0C 0 .Q/, in the full theory, and its Fourier transform is,
from (5.16.68), (5.16.69), (5.16.56), given by
Z 1
e 0C 0 .Q/ D Z 3 1 C Q2
D dM 2 .M 2 / ; (5.16.78)
Q2 0 Q2 .Q2 C M 2 /
leading to a potential
Z
e x 0 j/ D q2 d3 Q iQ.xx 0 / e 0
U.jx e D 0 .Q/jQ0 D0
.2/3
h Z 1 i
q2 2 2 M jxx0 j
D Z 3 C dM .M / e ; (5.16.79)
4jx x 0 j 0
78
A process involving no energy transfer.
and, at large distance seaparation jx x 0 j ! 1, we recover the Coulomb potential

from (5.16.56)
q2ren p
e x 0 j/ !
U.jx ; qren D q Z 3 ; (5.16.80)
4jx x0 j
with modified renormalized charges. That the charge is renormalized via the photon
renormalization constant Z 3 only and not by the renormalization constants Z 2 , and
Z1 , as well, is a consequence of the Ward identity, established in Sect. 5.16, implying
that Z1 D Z 2 . This point will be further discussed in the next section.
An integral equation of the photon polarization tensor, to supplement those of
the electron and vertex functions derived in the previous subsection, is derived in
the following one.
Fourier Transform of the Unordered Product of Two Currents
From (5.16.73) and (5.16.75), we may solve for the vacuum expectation value of
the time-ordered product of two currents as follows
˝ ˛ Z .dQ/ Z 1
.M 2 /
i j .x/j .0/ D 4
e iQ x
. 2
Q Q
Q /Q 2
dM 2 2 :
C .2/ 0 Q C M2 i
(5.16.81)
This means that

Z Z
˝ 1
.dQ/
.x 0 / j .x/j .0/i D 2 2
dM .M /2
ı.Q2 C M 2 /.Q0 /
0 .2/4
. Q2 Q Q / Q2 ei Qx ™.x 0 /:

(5.16.82)
and
Z Z
0
˝
1
2 2 .dQ/
™.x / j .0/j .x/i D 2 dM .M / ı.Q2 C M 2 /.Q0 /
0 .2/4
. Q2 Q Q / Q2 ei Qx ™.x 0 /;

(5.16.83)
where we have made a change of the variable of integration Q ! Q in the last

integral.
Invoking hermiticity of the current operators and the normalization of the vacuum
state, Eq. (5.16.83) above, upon taking its complex conjugate, becomes
Z 1 Z
˝ ˛ .dQ/
™.x 0 / j .x/j .0/ D 2 dM 2 .M 2 / ı.Q2 C M 2 /.Q0 /
0 .2/4
. Q2 Q Q /Q2 / eiQ x ™.x 0 /:
(5.16.84)
Finally adding (5.16.82) and (5.16.84), and using the relation ™.x 0 / C ™.x 0 / D
1, gives for the vacuum expectation of the unordered product of two currents the
relation
Z 1 Z
˝ ˛ .dQ/
j .x/j .0/ D 2 dM 2 .M 2 / 4
ı.Q2 C M 2 / ™.Q0 /
0 .2/
. Q2 Q Q / Q2 eiQ x : (5.16.85)
A Fourier transform, with k0 > 0, then leads to the following useful expression
Z
˝ ˛
.dx/ ei ky j .x/j .0/ D 2 .k2 / k2 . k2 k k /; (5.16.86)
where the part of the spectral function ı.M 2 / of .M 2 / does not contribute as
it is multiplied by M 2 in (5.16.85), (5.16.86) with M 2 D Q2 , M 2 D k2 ,
respectively (see (5.16.42) and (5.16.56)).79
This equation will be used to study the fundamental process eC e ! hadrons
in Sect. 6.4 in quantum chromodynamics.
5.16.4 Integral Equation for the Vacuum Polarization Tensor
To derive an integral equation for the photon polarization tensor, note that
from (5.16.71), we may, by functional differentiation of hA .x/i with respect
to J .x 0 / and using the chain rule, write
Z
e .x; x 0 / D D .x; x 0 / C • e .z; x 0 /:
D .dy/.dz/ D .x; y/ h j .y/i D
•hA .z/i
(5.16.87)
79
See also the QED lowest contribution to .M 2 / in (5.10.55).
and we set the external sources equal to zero. This suggests to define the polarization
tensor80
•
˘ .y; z/ D h j .y/i: (5.16.88)
•hA .z/i
On the other hand, note that

h j .y/i D e0 ab

h b .y/ a .y/ i D i e0 Tr Œ e
S.y; y/ ; (5.16.89)
C
and from (5.16.4) and (5.16.88), we obtain, in the absence of external sources,
Z

˘ .x y/ D i e20 .dx 0 /.dy 0 / Tr e
S.x x 0 / .x 0 ; y 0 I y/ e
S.y 0 x/ :
(5.16.90)
Current conservation implies from (5.16.88) that .@=@y /˘ .y; z/ D 0. On the

hand, from (5.16.3) for •hA .z/i D .@=@z k /•'.z/ for an arbitrary •'.z/, (5.16.1)
gives •hj .y/i D 0,
Z Z

0 D •hj .y/i D .dz/˘ .y; z/•hA .z/i D .dz/˘ .y; z/.@=@z /.z/;
(5.16.91)

and hence we also have .@=@z / ˘ .y; z/ D 0. That is, the Fourier transform
˘ .Q/ has the general structure

˘ .Q/ D Q2 Q Q ˘.Q2 /: (5.16.92)
Compare with (5.10.42) to lowest order. Fourier transforming (5.16.90), and

using (5.16.7) give
Z
.dp/
e
˘
.Q/ D i e20 Tr S. p C Q/ . p C Q; p/ e
S. p/ : (5.16.93)
.2/4
Upon multiplying this equation by Q and using the Ward-Takahashi identity

in (5.16.8), we obtain
Z
.dp/
e
˘ .Q/ Q D i e20 4
Tr S. p/ e
S. p C Q/ : (5.16.94)
.2/
One is tempted to shift the variable of integration p ! p Q, in the second term to

obtain zero for the net contribution. This, however, is not justified as the integral is
at least quadratically divergent. Since the transversality of the vacuum polarization
80
The minus sign is introduced to be consistent with the notation of the lowest order.
5.17 The Full Renormalized Theory 335
tensor, giving rise to the structure in (5.16.92), has to be satisfied, on grounds of

gauge invariance, one may introduce a strong ultraviolet cut-off in (5.16.94) so that
the shift of the variable of integration is justified, thus eliminating the integral in
question. Or one may carry out a careful definition of a gauge invariant current to
define ˘ .Q/, itself, as done in Sect. 3.9,81 or one may equivalently extract the
transverse part (5.16.93) directly from (5.16.93), as is often done in the literature.
Finally, we note from (5.16.87), (5.16.88), (5.16.92), that in the momentum
description,
e .Q/ D D .Q/ 1 e .Q/;

D . Q2 Q Q /˘.Q2 /D (5.16.95)
Q2 i
or

1 Q Q
e .Q/ D Q Q
D C : (5.16.96)
Q2 .Q2 C Q2 ˘.Q2 / Q4
5.17 The Full Renormalized Theory
The renormalized basic components of the theory are defined by

p
e
Sren D e e
S=Z 2 ; D e
ren D D =Z 3 ; ren D Z1 ; e D Z 3 e0 ; m D m0 C ım:
(5.17.1)
Renormalization is the process in which the experimentally unattainable param-
eters e0 and m0 , introduced earlier as limits at absolute zero distances going all
the way to the core of the electron, are eliminated in favor of the experimentally
observed physical quantities e and m.
We, certainly, cannot claim that our present theories may be extended all the way
to absolute zero distances. I find it rather surprising that there are still practitioners
who are surprised when pathologies (divergences) are encountered when dealing
with the physically unattainable unrenormalzed quantitites.
On the other hand, the wavefunction renormalization constants emerge, irrespec-
tive of perturbation theory, in the following manner.82
Consider the matrix elements of the fields, say, and A , between single particle
states and the vacuum as follows
p
h p ; j .x/jvaci D Z 2 eip x u.p; /; (5.17.2)
p
h k jA .x/jvaci D Z 3 ei k x e .k/: (5.17.3)
81
One may also equivalently subtract a Taylor expansion in Q, of the integrand in (5.16.93), in the
process, in finally evaluating ˘ .Q/.
82
See also Sect. 4.1.
This is what quantum theory says, irrespective of perturbation theory,83 since the
fields may create other particles as well out
pof thep vacuum, which necessitates the
presence of the normalization coefficients Z 3 ; Z 2 in a probabilistic context.
One may, in turn, introduce the renormalized fields
p p p
ren D = Z 2; ren D = Z 2; Aren D A = Z3 ; (5.17.4)
which, in particular, now satisfy the conditions

hp ; j ren .x/jvaci D eip x u.p; /; hkjAren .x/jvaci D eik x e .k/
(5.17.5)
creating, in addition to other particles, an electron and a photon, respectively, out of

the vacuum with unit amplitudes, emphasizing the particle aspect of the theory as
“free” particles, described by properly normalized wavefunctions, when emerging
into the detection regions by counters and when released by emitters.
Thus carrying out wavefunction renormalizations results in the proper normaliza-
tion of particles’ wavefunctions in describing asymptotic states. It is a remarkable
property of QED, as well as of a very special class of other cherished theories to
be discussed later, that when the process of renormalization is carried out, in which
unrenormalized expressions are eliminated in favor of the renormalized ones, the
theory is finite.
An important step in studying the full renormalized QED theory, is to eliminate
in favor of the full vertex , by using the integral equation for the latter given
in (5.16.30). This shows that we may, in general, write
Z
.dp 0 /

D . p C Q; p/ C i e20 K. p; p 0 I Q/ e
S. p 0 C Q/ . p 0 C Q; p 0 / e
S. p 0 /:
.2/4
(5.17.6)
Let us see how the basic components of the theory respond to the scalings of the
propagators and and vertex function defined in (5.17.1).
From Fig. 5.17 given earlier in Sect. 5.16.2, and Fig. 5.19, we note that for the
e eC - kernel K, the combination 2
p e40 K may be2 expressed in terms of renormalized
quantitites multiplied by .1= Z 2 / D .1=Z 2 / . That is,
1
e20 K D e2 Kren ; (5.17.7)
.Z 2 /2
where e2 Kren may be expressed in terms of the full renormalized electron and
photon
p propagators, the renormalized vertex and the renormalized charge e D
Z3 e0 .
83
See also the Introductory chapter of the book and Sect. 4.1.
5.17 The Full Renormalized Theory 337
p+Q p +Q p+Q p +Q
e0 √Z 2 √Z 2 e0
p+Q p +Q Z1−1 √
e0 Z −1 Z1−1
√1 Z3 √ Z 3
Z3 √
e20 = √
+ √ Z 3 Z3
Z3
p e0 Z1−1 Z1−1 Z1−1
p √ √
e0 Z 2 Z 2 e0
p p p p
Fig. 5.19 Upon writing e SDZ 2e e D Z3 D

eren , D ren =Z1 , and using the Ward identity

Sren , D
relation Z1 D Z 2 , we note that we may rewrite e20 K as e2 Kren =.Z 2 /2 , where e2 Kren may be
expressed in terms of the full renormalized electron
p and photon propagators, the renormalized
vertex function and the renormalized charge e D Z3 e0
The integral equation for the renormalized vertex may be written as

ren1. p C Q; p/
Z
.dp 0 /
D Z1 i e2 Kren . p; p 0 I Q/ e
Sren . p 0 C Q/ ren

. p 0 C Q; p 0 / e
Sren . p 0 /:
.2/4
(5.17.8)
as follows from (5.16.30), where we have used the Ward identity relation Z1 D Z 2 .
From (5.16.21)/(5.16.22), an integral equation for the inverse of the renormalized
electron propagator may be also written. For studying its renormalizability, however,
one may use the Ward-Takahashi identity in (5.16.8) to write
ˇ
e
S1 0 0 0 ˇ
ren . p / D . p p/ ren . p ; p/ p!m ; (5.17.9)

since eS1 . p/j p!m D 0, where, in particular, j p!m attached to ren . p 0 ; p/
means that p2 is set equal to m2 in it, and that p on its extreme right-hand side
is replaced by m.
Finally we note that for the vacuum polarization tensor, we may, from (5.16.93)
and (5.17.6), represent it symbolically as

˘ D i e20 Tr Œe
Se Se
S C i e20 Tr ŒK e Se
S e
S ; (5.17.10)
and may be represented diagrammatically as in Fig. 5.20, and its renormalized

counterpart will be discussed in Sect. 5.18.2.
In Sects. 5.8 and 5.9, we have seen that the coupling of the external sources to
external lines readily lead, in a straightforward manner, to the extraction of the
scattering amplitude of any process desired. Specifically, upon using the explicit
Π μν = − i e20 + i e20 K
Fig. 5.20 Graphical representation of Eq. (5.17.10)
expression of the vacuum-to-vacuum amplitude h 0C j 0 i, as given in (5.7.20),

and writing h 0C j 0 i D eiW , as in (5.8.1), we may extract any amplitude
(a connected process) A from W, with the correct normalization factors, as we
have carried out in Sect. 5.9, for specific processes, and use, in the process, the
expressions in (5.9.17), (5.9.18), (5.9.38), (5.9.39), (5.9.45), (5.9.46), which signify
the role of the external sources in generating a scattering amplitude.
The external sources are coupled to the fields in the Lagrangian density in the
following manner:
C C A J ren ren C ren ren

C Aren Jren ; (5.17.11)
p p p
where D ren = Z 2 ; D ren = Z 2 , J D Jren = Z3 . The scalings of the
propagators, the vertex function and the charge, as defined above, lead simply to the
replacement of the sources by their renormalized scaled counterparts, emitting and
detecting “free” particles, making the particle aspect of the theory evident.
Various vertex connections that may occur in processes are shown in the
figure below, with the amplitudes expressed in terms of the full propagators and
the full vertex function. By introducing the renormalization scaling factors, in
defining the renormalized theory, one may, in turn, express amplitudes in terms of
the full renormalized propagators,
p the full renormalized vertex function, and the
renormalized charge e D Z3 e0 .
Incidently, the above analysis, also shows that due to the Ward identity,
p the charge
is renormalized by the photon wavefunction renormalzation only: e D Z3 e0 .
5.18 Finiteness of the Renormalized Theory; Renormalized

Vertex Function and Renormalized Propagators
The present section deals with the finiteness of renormalized theory order by order
in perturbation theory by a direct graphical analysis. A different approach will be
taken when dealing with non-abelian gauge fields.
5.18 Finiteness of the Renormalized Theory; Renormalized Vertex Function. . . 339
1/ Z 2 1/ Z 2 1/ Z3 1/ Z 2
√ √ √ √
Z2 Z2 Z3 Z2
√ √ √ √
Z2 Z2 Z3 Z2
e0 Z −1 e0 Z −1
√1
√1
√ Z3 √ Z2
Z3 Z2
√ √
1/ Z3 1/ Z2
√ √ 1/Z1
√ Z3 √ Z2 e0
Z3 Z2 √ √
−1 −1 Z2 Z2
√ e0 Z√ √ e0 Z1√ √ √
Z 2 √Z3 √
1
Z2 Z2 Z Z3 Z2
√ √ √ 2 √ Z3
Z2 Z2 Z2 Z3
Fig. 5.21 These are the types of vertex connections that may occur. Expressing the propagators,
the vertex function and the charge by their renormalized counterparts, simply lead to the
replacement of the sources by their renormalized scaled counterparts as defined below (5.17.11),
and the particle aspect of the theory, discussed above, becomes evident. The sources are denoted by
the half circles corresponding to electron/positron and photon emitters and detectors. By removing
the renormalized sources and the renormalized external propagators attached to them, in the
manner that was done in Sect. 5.9, as discussed above, the amplitudes A emerge expressed in
terms of the full renormalized components of the theory
5.18.1 Finiteness of the Renormalized Theory
We establish the finiteness of the renormalized QED theory (Fig. 5.21) by a direct
simple way. We then consider some applications of so-called renormalization group
methods as applied to it in the next section.
To lowest order, we have encountered ultraviolet divergences in the theory when
dealing with the electron and photon self energies, as well as the vertex function.
These are shown in the figure below. With this in mind, we define the degree of
divergence of a graph as:
the number of integration variables plus the highest power of momenta in the numerator
minus the highest power of momenta in the denominator.
Due to the transversality property, in the external photon lines, of the light-light
scattering graph, a result that follows by carefully invoking gauge invariance,84 its
naïve degree of divergence of 0 is reduced to 4, and hence is finite (Fig. 5.22).
84
See Appendix IV at the end of the book.
(a) (b) (c) (d)
Fig. 5.22 The naïve degrees of divergence of the above graphs are, respectively, C1; 2; 0; 0.
Due to the factor p in the electron propagator, where p is the momentum of the electron on
the external lines, the degree of divergence of the graph in (a) is reduced to 0. On the other
hand, due to the very welcome two powers of momenta of the external lines of the photon, in the
polarization tensor in (b), as follows by gauge invariance, its degree of divergence is also reduced
to 0 (see Appendix IV, (IV.27) at the end of the book). Again due to gauge invariance, the degree
of divergence of the light-light scattering graph in (d) is reduced to 4, and hence is finite (see
Appendix IV, (IV.26) at the end of the book)
(a) (b)
Fig. 5.23 Some improper graphs
The coupling parameters in QED is dimensionless and a remarkable property

of the theory is that the degree of divergence of a proper connected graph remains
the same independently of the order of perturbation theory. To this end we define a
graph to be proper if it cannot be broken into two disjoint graphs by cutting a single
photon or a single electron line. Some improper graphs are shown in Fig. 5.23.
The degree of divergence d.g/ of a connected and proper graph in QED is easily
checked to be given by
d.g/ D ìnteg ìnt

e 2 ` ;
int
(5.18.1)
ìnteg D number of integration variables;
e D number of internal electron lines;

ìnt (5.18.2)
ìnt
D number of internal photon lines:
The remarkable property of this theory that the degree of divergence of a connected
and proper graph with a fixed number of external lines does not change with the
order of perturbation theory, that is with the increase of the number of internal lines,
becomes evident when d.g/ is expressed in terms of the number of external lines
(a) (b)
Fig. 5.24 (a) Here ìnteg D 4 3, ìnte D 6, ` D 2. (b) In this case, ìnteg D 4 7, è D 14,
int int
` D 6. The degree of divergence of the graphs are d.g/ D 2, and are equal, as expected, and this
int
is effectively reduced to zero after subtracting two powers of momenta needed in the definition of
the polarization tensor as discussed before
of a graph (e.g., Fig. 5.24). It is easily verified that this is simply given by
3 ext
d.g/ D 4 ` èxt
; (5.18.3)
2 e
e ; ` denote the number of external electron ( positron) and photon lines,

where èxt ext
respectively.
To establish the finiteness of the renormalized theory to any order in the
renormalized charge, we may proceed by induction.
To the above end, we have seen in Sects. 5.10 and 5.11, that to lowest order,
the renormalized electron and photon propagators as well as the renormalized
vertex functions are finite. Now we assume, as an in induction hypothesis, that
in perturbation theory, the renormalized propagators and the renormalized vertex
function are finite to all orders .n2/ in e. This will readily allow us to study
the finiteness of e2 Kren , and any connected and proper graph G , with four more
external lines, to a given order n.
To the above end, we note, in reference to Figs. 5.19 and 5.21, e2 Kren , and a
connected and proper graph G , with at least four external lines, may be expanded in
terms of the full renormalized propagators and the renormalized full vertex function,
and have strictly negative degrees of divergence. The only divergence that may
occur are from subgraphs within them which are self-energy insertions and vertex
corrections. These, however, have been rendered finite, by the renormalization
procedure, to any order .n 2/, and has reduced their effective degrees of
divergence to strictly negative values, according to the induction hypothesis. Also
recall that a subgraph with only three external photon lines within them vanishes
by Furry’s theorem, that the effective degree of divergence of light-light scattering
is equal to 4, and finally note that all other subgraphs with four or more external
lines within them have strictly negative degrees of divergence. Hence e2 Kren and
G , with at least four external lines, have all their subintegrations, as well as their
overall integrations finite. That is, to any given order85 n in e, e2 Kren , G are finite.
Thus it remains to establish the finiteness of the propagators and the vertex function
to order n as well. This is carried out in the following two subsections.86
5.18.2 The Renormalized Vertex and the Renormalized

Electron Propagator
From (5.17.8), the renormalized vertex has the structure

ren e . p 0 ; p/:
. p 0 ; p/ D Z1 C e2 (5.18.4)
e . p 0 ; p/ is expressed in terms of renormalized components. Also

where
ˇ ˇ
ˇ e . p; p/ˇˇ
ren . p; p/ˇ D D Z1 C e2 ; (5.18.5)
p!m p!m
from which, by solving for .Z1 1/ , we may conveniently write

ˇ

ren . p C Q; p/ D C e2 e . p; p/ˇˇ
e . p C Q; p/ : (5.18.6)
p!m
The nth order contribution of the latter is given by
ˇ.n/ ˇ ˇ.n2/

ren . p C Q; p/ˇ D e2 e . p; p/ˇˇ
e . p C Q; p/ ˇ
ˇ : (5.18.7)
p!m
In the induction hypothesis, we have assumed that the renormalized vertex and the
renormalized electron and photon propagators are finite, and all subgraphs in them,
are finite for all orders .n 2/. In this subsection, we will see that this is true for
the renormalized vertex and renormalized electron propagator to order n as well. In
the next subsection we also establish this for the renormalized photon propagator.
In Fig. 5.15 of Sect. 5.16.2, a graphical representation of the vertex function was
given, and in Fig. 5.16, some graphs not contributing to the e eC -kernel K, in the
expression for the vertex function was also shown. A graphical representation of
ˇ
e . p 0 ; p/ˇ.n2/ is then given by Kren , ren

e
, Sren as shown in Fig. 5.25, which may
85
This is the content of a power counting theorem, due to Weinberg [80], which states that a graph
is finite if all of its corresponding subintegrations, including the overall integration, are finite. For
a pedagogical treatment of this technical problem see the author’s book on renormalization [45].
86
Finiteness is understood to refer to ultraviolet finiteness. The infrared problem is treated by the
inclusion of soft photons in computing transition amplitudes. In the study of propagators and the
vertex function, a non-zero photon mass is included in the analysis, as done earlier to the lowest
order, to avoid infrared divergences in intermediate steps.
p Q
p Q (n−2)
(n−2)
Λ μ (p + Q, p) =− i Kren Q
ren
p
p
Fig. 5.25 Graphical representation of e

. p C Q; p/. Only the overall integration diverges. The
two electron propagators in the graph also denote renormalized ones. Note also the relation
in (5.17.7)
contain subgraphs of orders .n 2/ and the renormalization procedure has

reduced their effective degrees of divergence to strictly negative values, and the
only divergence87 in e . p C Q; p/, may come from the overall integration with a
0 degree of divergence, and the single subtraction in (5.18.7) renders this overall
integration finite, giving rise to a renormalized vertex finite to order n as well.
The finiteness of the renormalized electron propagator, to order n, then also follows
from (5.17.9), which we recall resulted from an elementary application of the Ward-
Takahashi identity.
It remains to establish the finiteness of the renormalized photon propagator to
order n. This is the subject of the next subsection.88
5.18.3 The Renormalized Photon Propagator
From (5.16.96), the full photon propagator, in covariant gauges, may be written as
2
e .Q/ D Q Q d.Q / C Q Q :
D (5.18.8)
Q 2 Q 2 Q4
We set
1
d.Q2 / D ; ˘.Q2 / ˛ 0 .Q2 /: (5.18.9)
1 C ˛ 0 .Q2 /
87
Light-light scattering subgraphs, as noted before, are convergent.
88
Finiteness of renormalized QED by induction was carefully treated in Bjorken and Drell [11],
which was also applied to the renormalized photon propagator by a very elaborate method. We
have preferred to treat the photon propagator by a method developed in a remarkable paper by
Baker and Lee [4].
By definition,
1
Z3 D ; (5.18.10)
1 C ˛ 0 .0/
and we may rewrite
1

d1 .Q2 / D 1 C ˛ 0 .0/ C ˛ 0 .Q2 / .0/ D 1 C ˛ c .Q2 / ; (5.18.11)
Z3
where

c .Q2 / D .Q2 / .0/ ; ˛ c .0/ D 0; ˛ D Z 3 ˛ 0; (5.18.12)
and the latter relation for the fine-structure constant was defined earlier.
Using the renormalized version of the photon propagator
Q Q dren .Q2 / 1 Q Q
e
D .Q/ D
C ; (5.18.13)
ren
Q2 Q2 Z3 Q4
and (5.18.8), (5.18.11), we have

1

dren .Q2 / D 1 C ˛ c .Q2 / D Z 3 1 C ˛ 0 .Q2 / : (5.18.14)
To establish the finiteness of the photon propagator to order n as well, we first

define the vacuum polarization tensor by factoring out the unrenormalized fine-
structure constant ˛ 0 , as done in (5.18.9), and write
˘ .Q/ D ˛ 0 .Q/ D ˛ 0 . Q2 Q Q / .Q2 /; (5.18.15)
where from (5.17.10), (5.17.7), .Q/ has a skeleton expansion in terms of

renormalized components of the theory given by

D 4 i Tr Œe e
Sren ren
Sren ren C i e2 Tr ŒKren e e e
Sren ren e
Sren Sren ren Sren ;
(5.18.16)
and has the graphical representation in Fig. 5.26.

It is easily verified that .Q2 / is worked out to be given by
Q Q ˛ @ @
.Q2 / D ;˛ .Q/; ;˛ ˛ .Q/ .Q/: (5.18.17)
6 Q2 @Q˛ @Q˛
π μν =− 4 π i + 4π α i Kren
Fig. 5.26 Graphical skeleton expansion representation of .Q/, where note that on account of
the scaling factor as given in (5.17.1), and (5.17.7) for e20 K D e2 Kren =.Z 2 /2 , it is expressed in
terms of the renormalized components of the theory. Accordingly, all the propagators, the vertex
function, as well as the fine-structure constant, denote renormalized components
Upon scaling Q by a parameter , we then have from (5.18.12)

ˇD1 ˇ
ˇ Q Q ˛ ˇ
c .Q2 / D . 2 Q2 /ˇˇ D 2
;˛ .Q/ ;˛ ˛ .Q/ˇ : (5.18.18)
D0 6 Q 0

Note that up to order n, ˛ c is expanded in terms of the renormalized
propagators and the renormalized vertex, up to order n 2 in e.
To discuss the finiteness of the latter, we differentiate in (5.18.16),
with respect to Q, and thus differentiate all the (renormalized) components in it
with respect to this variable Q˛ , wherever it appears in them. To this end, the
renormalized vertex function may be written symbolically and conveniently as

ren D Z1 ie2 Kren e
Sren e
Sren ren ; (5.18.19)
e e
D Z1 ie2 ren Sren Sren Kren : (5.18.20)
Upon differentiating (5.18.19), we have,

1 C ie2 Kren e
Sren e
Sren ren ;˛ D ie2 Kren e
Sren e
Sren ;˛ ren : (5.18.21)

Multiplying this by ren e
Sren e
Sren gives

e e e
ren Sren Sren C ie2 ren Sren e
Sren Kren eSren e
Sren ren ;˛
e e

D ie2 ren Sren Sren Kren e
Sren e
Sren ;˛ ren
: (5.18.22)
From (5.18.20), we also have

e e
ie2 ren
Sren Sren Kren D .ren Z1 /: (5.18.23)
Thus the two terms within the round brackets on the left-hand side of (5.18.22) may
be combined to obtain the useful identity

Z1 eSren e
Sren ren ;˛ D ie2 ren eSren e
Sren Kren eSren e
Sren ;˛ ren

D i e2 ren e Sren Kren e
Sren e Sren e
Sren ;˛ ren e e
i e2 ren Sren Sren Kren ;˛ e
Sren e
Sren ren :
(5.18.24)
[π μν ] = − 4 π i − 4π i α Kren
Fig. 5.27 Graphical skeleton expansion representation of .Q/;˛ expressed in terms of renor-
malized components. The prime 0 denotes differentiation with respect to the momentum Q˛ in
.Q/. The prime 0 in the first graph on the right-hand
side stands for the differentiation of the
two renormalized electron propagators e Srene
Sren ;˛
Using (5.18.23), the above equation may be equivalently rewritten as

Z1 e
Sren e
Sren ren ;˛ C Z1 e Sren e
Sren ;˛ ren

e e
e e
D ren
Sren Sren ;˛ ren i e2 ren Sren Sren Kren ;˛ e
Sren e
Sren ren : (5.18.25)
From (5.16.93) and (5.19.9), one may also express symbolically in the
convenient form
D 4 i Z1 e
Sren e
Sren ren ; (5.18.26)
which upon using (5.18.25) gives

h i
e e e e
;˛ D 4 i ren
Sren Sren ;˛ ren 4i ˛ ren Sren Sren Kren ;˛ e
Sren e
Sren ren :
(5.18.27)
This may be represented graphically as shown in Fig. 5.27.

By power counting, i.e., by simple dimensional analysis, we note that the func-
tion ;˛ ˛ in (5.18.17), has only an overall logarithmic divergence contributing
to its nth order, by invoking, in the process, of finiteness to lower orders by the
induction hypothesis. This, in turn, means that .Q 2 / has also only an overall
logarithmic divergence. Clearly then, with one differentiation of ;˛ ˛ with
respect to Q not only the sub-integrations of the latter are finite, by the induction
hypothesis, but also the overall integration, as the single differentiation effectively
decreases its degree of divergence to minus one from zero. Using the facts that
@ Q Q @ Qˇ @
Q2 D 0; Q2 D ; (5.18.28)
@Q 2 Q 2 @Q 2 2 @Qˇ
as obtained, in the process, from the chain rule, gives from (5.18.17)
@ Q Q ˇ @ ˛ Q2
Q2 .Q 2
/ D Q ;˛ .Q/ D a finite expression;
@Q 2 12Q 2 @Qˇ m2
(5.18.29)
5.19 Effective Charge and the Renormalization Group 347
and we may write

Z Q 2 =m2
2
2
d
˛ c .Q / D ˛ .Q / .0/ D ˛ ./; (5.18.30)
0
which is finite up to order n as well. This completes the demonstration of the

renormalized theory to order n as well.89
5.19 Effective Charge and the Renormalization Group
5.19.1 Renormalization Group Analysis
We introduce the concept of an effective charge which interpolates between the

renormalized charge e and the unrenormalized charge e0 . To this end, we note
from (5.18.9) that we may write
˛0 ˛0
˛ 0 d.Q2 / D 2
D
: (5.19.1)
1 C ˛ 0 .Q / 1 C ˛ 0 . / C ˛ 0 .Q2 / . 2 /
2
In analogy to the fine-structure constant: ˛ D ˛ 0 =Œ1 C ˛ 0 .0/, we may define an

effective, energy dependent, fine-structure coupling ˛. 2 / by
˛0
˛. 2 / D ; (5.19.2)
1 C ˛ 0 . 2 /
with the conventional fine-structure constant ˛ corresponding to D 0, i.e.

measured at large distance from an electron, given by
˛ ˛.0/: (5.19.3)
From (5.19.1), (5.19.2) and (5.19.3), we obtain the basic equation

˛ dren .Q2 ; ˛/ D ˛. 2 / dren Q2 ; 2 ; ˛. 2 / ; (5.19.4)
where
1
dren Q2 ; 2 ; ˛. 2 / D ; (5.19.5)
1 C ˛. 2 / Œ.Q2 / . 2 /
dren .Q2 ; 0; ˛.0// dren .Q2 ; ˛/: (5.19.6)
89
Note from (5.18.14), that to order n, dren depends on .˛ c / to order < n and on the nth order
as well, and the latter has been also established to be finite.
Hence, in general, we have the following basic property encompassing the

change under the redefinition of renormalized coupling parameter
˛.12 / dren .Q2 ; 12 ; ˛.12 // D ˛. 22 / dren .Q2 ; 22 ; ˛. 22 //: (5.19.7)
Because of this basic property, such a procedure in renormalization theory has been
referred to as “renormalization group” which is an unfortunate name as it doesn’t
really have to do with a group.90
The effective coupling ˛. 2 / may be obtained from (5.19.5) by choosing Q2 D
2
to give
˛. 2 / D ˛ dren . 2 ; ˛/; (5.19.8)
and from (5.16.76), satisfies the boundary conditions
˛. 2 / ! ˛ for 2 ! 0; ! ˛ 0 for 2 ! 1; (5.19.9)
where the latter limit may diverge.

For energies of the order (91 GeV), corresponding to the mass of the vector
boson Z 0 , this effective coupling is experimentally of the order of 1/128.91 An
estimate of the effective fine-structure at such an energy is carried out in Sect. 5.19.2
taking into account those contributing quarks and the charged leptons of the three
generations by using the renormalization group.92
To study the high-energy asymptotic behavior of ˛ dren .Q2 ; ˛/, and hence of the
effective coupling and thus investigate the small distance behavior of the theory,
we first summarize some of the properties associated with the vacuum polarization
tensor to lowest order in ˛ (see (5.10.45), (5.10.49), also (5.18.9)). To this end,
˘.Q2 / D ˛ .Q2 /
h 1 i 2˛ Z 1 h Q2 i
D 1 dz z.1 z/ ln 1 C 2 z.1 z/ ; (5.19.10)
Z3 0 m
1 ˛ h 2 1 i
D1C ln 2 : (5.19.11)
Z3 3 m 2
90
The renormalization group was introduced by Stueckelberg and Peterman [74], who also coined
the name “renormalization group”. This was followed by other important work by Gell-Mann and
Low [27], Ovsyannikov [65], Bogoliubov and Shirkov [12] and by many others.
91
See, e.g., Beringer et al. [7], Mele [63], and Erler [23].
92
I remember when I was a graduate student, I used to wonder, together with my fellow students,
as to why nature chooses the value 1=137 rather than some other value. Now with the rapid
development of quantum field theory, together with higher energy experiments, we understand
this value to be a reflection of the energies “at which we were then”. Effectively, this coupling
changes as we move to higher energies attaining different numerical values as will be investigated
in the next subsection. Although understanding why Nature chooses the low energy numerical
value 1=137 is still important, it does not seem to be as mysterious now as it was then.
From these two equations, we note, in particular, that to lowest order
m d 2˛
ˇ.˛/ Z3 D ; (5.19.12)
Z 3 dm 3
d d 4 ˛ m2
m ˘.Q2 / D m ˛ .Q2 / ; Q2 m2 ; (5.19.13)
dm dm Q2
˛ h Q2 5i
˛ c .Q2 / D ˛ Œ.Q2 / .0/ ln 2 2
c .Q /;
˛ as Q2 m2 :
3 m 3
(5.19.14)
where “as” in as
c stands for asymptotic.
Now consider the full theory. We differentiate the equation
1

dren .Q2 / D 1 C ˛ c .Q2 / D Z 3 1 C ˛ 0 .Q2 / : (5.19.15)
with respect to the mass m of the electron by keeping the unrenormalized fine-
structure constant ˛ 0 and the ultraviolet cut-off 2 fixed with the latter arbitrarily
large.93 To this end we set94
1 d d
m Z 3 D ˇ.˛/; m .Q2 / D S : (5.19.16)
Z 3 dm dm
With ˛ 0 ; 2 fixed, the parameters that vary in .Q2 / on the extreme right-hand
side of (5.19.15) with m are then ˛, and m0 . This leads to
d
m Z 3 Œ1 C ˛ 0 .Q2 / D ˇ.˛/ dren
1
.Q2 / C ˛ S : (5.19.17)
dm
On the other hand, the parameters that vary in the first expression in (5.19.15) are ˛
and m itself. That is,
d 1 2 h @ m d @ i
m dren .Q / D m C ˛ ˛ d1 .Q2 /
dm @m ˛ dm @˛ ren
h @ @ i 1 2
D m C ˛ ˇ.˛/ d .Q /: (5.19.18)
@m @˛ ren
93
In carrying the differentiation of this equation with respect to m, we follow the elegant approach,
and to some extent the notation, of Adler [1].
94
Since .Q 2 / depends on the mass m through m0 , we may write: m.d=dm/ .Q 2 / D
.m=m0 / .dm0 =dm/ m0 .@ .Q 2 /=@m0 /.
From (5.19.15), (5.19.17), (5.19.18), we obtain95

h @ @ i
1
m C ˇ.˛/ ˛ 1 dren .Q2 / D ˛ S .Q2 /: (5.19.19)
@m @˛
Now we let Q2 =m2 become arbitrary large. Because of dimensional reasons,

order by order in perturbation theory S .Q2 / vanishes like m2 =Q2 up to powers
of ln.Q2 =m2 /,96 thus giving
h @ @ i
2 2
m C ˇ.˛/ ˛ ˛ d as
ren .Q =m ; ˛/ D 0; (5.19.20)
@m @˛
2 2 2 2
where the function d as ren .Q =m ; ˛/ is obtained from dren .Q =m ; ˛/, by neglect-
2 2
ing all the terms in the latter which vanish for Q =m ! 1. The func-
2 2 2 2
ren .Q =m ; ˛/, will then depend on powers of ln.Q =m / (see (5.19.33)). We
tion d as
2 2
note that due to the finiteness of d ren .Q =m ; ˛/, this equation implies that ˇ.˛/ is
as
finite.
2 2
We will solve (5.19.20) for d as ren .Q =m ; ˛/ with the boundary condition
ren .1; ˛/ D q.˛/:

˛ d as (5.19.21)
Up to second order in ˛, q.˛/ may be directly read from (5.19.14), (5.19.15) to be
5 2
q.˛/ D ˛ ˛ C : (5.19.22)
9
In reference to (5.19.20), we may use the chain rule, to replace the differentiation
with respect to ˛ by differentiation with respect to q.˛/, in the following manner
@ @ 1 @ 1 @
D q 0 .˛/ ; or ˛ ˇ.˛/ D ˛ ˇ.˛/ q 0 .˛/ : (5.19.23)
@˛ @q.˛/ 2 @˛ 2 @q.˛/
where the factor 1=2 is introduced for convenience. This suggests to introduce the
function
1
.q.˛// D ˛ ˇ.˛/ q 0 .˛/; (5.19.24)
2
95
Such a scaling equation is referred to as a Callan-Symanzik scaling equation [15, 75–77].
96
This is fully justified by the application of a theorem due to Weinberg [80] when it is applied to
the renormalized theory order by order [45], and essentially reduces to a dimensional analysis. The
m2 =Q 2 factor is already seen in (5.19.13) to lowest order.
known as the Gell-Mann and low function. Equation (5.19.20) may be now
rewritten as
h @ @ i
.q.˛// ˛ d as
ren D 0; D ln.Q2 =m2 /; (5.19.25)
@ @q.˛/
which, in particular, explains the choice of the factor 1=2 in (5.19.23). Finally, we
introduce the variable
Z q.˛/
dz @ @
y D ; D .q.˛// ; (5.19.26)
c .z/ @y @q.˛/
where c is an arbitrary constant, to rewrite (5.19.25) simply as

h @ @ i as
˛ d ren D 0: (5.19.27)
@ @y
We will obtain two useful expressions from this equation. The solution of the
differential equation is elementary and is given by
h Z
2 2 Q2 q.˛/
dz i
˛ ren .Q =m ; ˛/
d as D FŒ C y D F ln 2 C ; (5.19.28)
m c .z/
where F is an arbitrary function, but with the normalization condition
˛ d as
ren .1; ˛/ D q.˛/; (5.19.29)
it must satisfy the condition
hZ q.˛/
dz i
F D q.˛/: (5.19.30)
c .z/
As a translation operation in the variable y, we may rewrite (5.19.28) as

h @ i
2 2
˛ d as .Q =m ; ˛/ D exp FŒy
ren
@y
h @ i
D exp .q.˛// q.˛/; (5.19.31)
@q.˛/
where we have used (5.19.30). Expanding the exponential factor in (5.19.31), gives
1 n
2 2
X d n ˇˇ
˛ ren .Q =m ; ˛/
d as D .z/ zˇ ; (5.19.32)
nD0
nŠ dz z D q.˛/
leading to the useful expression

1 n
2 2
hX ln.Q2 =m2 / d d n1 ˇˇ i
˛ ren .Q =m ; ˛/
d as D q.˛/ C .q.˛// .z/ zˇ ;
nD1
nŠ dz dz zDq.˛/
(5.19.33)
as an explicit expansion in powers of ln.Q2 =m2 /.

Some expansions of ˇ.˛/ and .z/ which go beyond the lowest orders, are
given, respectively,97 by
2 ˛ 1 ˛2 121 ˛ 3
ˇ.˛/ D C 2
C ; (5.19.34)
3 2 144 3
h1 z 1 z2 18 101 z 3 i
.z/ D z C C .3/ C ; (5.19.35)
3 4 2 8 3 36 3
where .3/ 1:202, with denoting the Riemann zeta function.

Another useful expression obtained from (5.19.25), in addition to the one
in (5.19.33), for further analysis is obtained upon introducing the integral
Z ˛ d as 2 2
ren .Q =m ;˛/ Q2
dz
I.; ˛/ D ; with B.C. I.0; ˛/ D 0; D ln ;
q.˛/ .z/ m2
(5.19.36)
which from (5.19.25), it satisfies the equation
h @ @ i
.q.˛// I.; ˛/ D 1: (5.19.37)
@ @q.˛/
With y defined in (5.19.26), the solution of the above equation is given by98
I.; ˛/ D C GŒ C y ; with GŒ y D I.0; ˛/ D 0; (5.19.38)
where GŒ : is arbitrary, but is to satisfy the just given B.C. Also GŒ C y D

expŒ @=@y GŒ y D 0, which leads from (5.19.36), (5.19.38) to
Q2 Z ˛ d as 2 2
ren .Q =m ;˛/ dz
ln D ; (5.19.39)
m2 q.˛/ .z/
satisfying the proper B.C.
97
de Rafaël and Rosner [18] and Baker and Johnson [3]. See also Gorishnii et al. [30]. Note that
we use Adler’s ˇ.˛/ function notation which is ˇ.˛/=˛ of de Rafaël’s and Rosner’s.
98 eŒ C y D . C y/ C G
The solution may be also written as: y C G eŒ C y D C GŒ C y
where G eŒ C y . C y/ D GŒ C y. Hence (5.19.38) is the general solution of (5.19.37).
1
Finally from the boundary condition on dren .Q2 / ! Z3 for Q2 ! 1, as inferred
from (5.16.76), we also have from (5.19.20),
h @ @ i
m C ˇ.˛/ ˛ 1 Z 3 D 0: (5.19.40)
@m @˛
A very attractive feature of the expansion in (5.19.33), which may provide a hint
to determine the value of the fine-structure constant ˛, if the latter is a zero of
the beta function: ˇ.˛/ D 0,99 i.e., formally from (5.19.24), .q.˛// D 0, then
2 2
from the just mentioned expansion we may infer that ˛ d as ren .Q =m ; ˛/ D q.˛/ D
˛ 0 . Such a solution would mean that 1=Z 3 is also finite. This solution, in turn,
implies that the interval of integration in (5.19.39) degenerates to a point at which
.˛ 0 / vanishes. Not considering all possibilities, another situation that may arise
2
is in which the integral in (5.19.39) would not diverge until ˛ d as ren .Q / ! 1 as
2
Q ! 1 and 1=Z 3 , as well as ˛ 0 , would be infinite. The fine-structure constant
would then, of course, be undetermined. Finally, one may also consider the situation,
2
where ˛d as ren .Q / ! ˛ 0 , with the latter finite, and the Gell-Mann and Low function
.z/ develops a zero at such a point for consistency with (5.19.39). In this case ˛
would be undetermined, and the series would sum up to a non-trivial function of
ln.Q 2 =m2 / which approaches ˛ 0 as Q 2 ! 1. A key property of QED is that the
slope of ˇ.˛/ is positive near the origin, unlike non-abelian gauge theory ones,
such as QCD, in which it is negative leading to a decrease of the effective charge
with increase of energy, instead of increasing as in QED, a concept which is referred
to as asymptotic freedom . This will be investigated in the next chapter.100
Unfortunately, the Gell-Mann and Low function of QED may not have a non-
trivial zero,101 and one would then expect that e0 , which is formally defined as the
physically unattainable charge “measured” right at the core of an electron, to be
infinite. The divergence encountered in this limit is of no surprise as our theories are
not justified to describe nature at absolute zero distances. It is rather surprising that
some practitioners, let alone the others, still find it surprising that one may encounter
such an infinity when our theories are extended all the way with no limit. On the
other hand, to see how robust QED is, note from (5.19.33), (5.19.22), (5.19.35), or
1 as
directly from (5.19.14), (5.19.15), that to lowest order in ˛, .dren / .Q 2 =m2 ; ˛/ D
1 .˛=3/ln.Q =m /, leading to an unphysical pole at Q D m2 exp.3 =˛/,
2 2 2
referred as a Landau ghost. This seems to validate QED down to small distances
corresponding to enormous energies < m e645 . Of course this is an estimate
obtained by truncating an infinite series. But even with such a naïve estimate, this
indicates, nevertheless, that QED may be valid even to corresponding experimen-
tally unattainable small distances. At such distances, gravitation, in any case, may
99
See Adler [1], and, e.g., Manoukian [44].
100
One may similarly carry out an analysis of the wavefunction renormalization constant Z 2 and
of the unrenormalized mass m0 (see, e.g., [1]).
101
See, e.g., Krasnikov [38].
play an important role, and extrapolating to such distances, presumably, requires

detailed generalizations of our physical theories. Operationally, QED is a very
successful theory and is one of our most cherished ones. By eliminating the
physically unattainable parameters .m0 ; e0 / in favor of the physically measured
ones .m; e/, as it should be done for a sensible description of the theory, and
simultaneously carrying out wavefunction renormalizations for the corresponding
particles, as a proper normalization of their wavefunctions, for their appropriate
descriptions as “free particles” when impinging on detectors, the theory turns out to
be finite.
5.19.2 The Fine-Structure Effective Coupling at High Energy

Corresponding to the Mass of the Neutral Z 0 Vector
Boson
At present available high energies, a standard of the fine-structure coupling is taken

at an energy equal to the mass of the neutral Z 0 vector boson: Mz D 91:188 GeV.
An estimate of its value is given here based on all the charged leptons and all those
contributing quarks of the three generations. The effective fine-structure at an energy
squared Q2 may be written from (5.19.5), (5.19.6), (5.19.7) and (5.19.8) as
˛
˛.Q2 / D ˛dren .Q2 / D : (5.19.41)
1 C ˛ Œ.Q2 / .0/
or
1 1 2

D 1 C ˛ Œ.Q / .0/ : (5.19.42)
˛.Q2 / ˛
As long as ˛ ln.Q2 =m2 / 1, for m2 Q2 , we may use (5.19.14) to write

1 1 ˛ h Q2 5i
D 1 ln : (5.19.43)
˛.Q2 / ˛ 3 m2 3
We may take the contributions of the quark loops, with quarks corresponding to
the three colors, as well as of the charged lepton loops, of the three generations, with
charges and approximate masses given in Table 5.1 which contribute for Q2 D MZ 2 :
Table 5.1 Scaled charges and approximate masses (in GeV) of the quarks of masses MZ 2 and
the charged leptons
Particle u d s c b e
e 2=3 1=3 1=3 2=3 1=3 1 1 1
Mass 0:0024 0:0049 0:095 1:275 4:2 0.00051 0.1057 1.7768
5.20 Scalar Boson Electrodynamics, Effective Action and Spontaneous. . . 355
where e denotes the (scaled) charge, and note that the mass of the top quark
Mz D 91:188 GeV, and decouples from the analysis.102 We may then write the
contributions of the above quarks and leptons to the fine-structure constant at Q2 D
MZ 2 , with a factor 3 for the three quark colors, as
1 1 1 n X h M 2 5 i X h M 2 5 io
2 Z 2 Z
D 3 e ln C e i ln ;
˛.MZ 2 / ˛ 3 i;quarks i m2i 3 i;leptons
m 2
i 3
(5.19.44)
leading to the approximate anticipated value 1=˛.MZ 2 / ' 128.103
5.20 Scalar Boson Electrodynamics, Effective Action

and Spontaneous Symmetry Breaking
When the symmetry of the action is not shared by the vacuum, the symmetry is
said to be spontaneously broken. An interesting consequence of this is that for
a spontaneously broken continuous symmetry, the theory involves massless spin
zero boson(s), referred as Goldstone bosons. As we will see in a spontaneously
broken symmetry, the vacuum is degenerate. Considering a continuous global
transformation, we investigate in Sect. 5.20.2 to see how these Goldstone bosons
arise. In gauge theories, abelian or non-abelian, with a priori massless vector
bosons, we have invariance under local transformations, and the broken symmetry
is local, with underlying gauge transformations existing, in particular, between
charged scalar fields and the vector bosons. In these cases, the Goldstone bosons
combine with the massless vector bosons resulting in a theory in which the vector
bosons become massive borrowing the zeroth component of their helicities from
the Goldstone bosons themselves, and the latter now become unobservable. This
is worked out in a variation of scalar boson electrodynamics. The underlying
mechanism for this transmutation is referred to as the Higgs mechanism. In the
final theory, the number of degrees of freedom, associated with the fields, are the
same as the original version. The importance of this result in the renormalization
program, should be emphasized and the underlying idea is quite simple. We recall
that the propagator of a massless vector boson is given by
Q Q 1
D .Q/ D ; (5.20.1)
Q2 Q2 i
102
This is a consequence of the decoupling theorem and one carries out the analysis as if heavy
masses in comparison to the energy of concern are not part of the dynamics. For intricacies,
conditions and a proof of the decoupling theorem, see Manoukian [46, 47, 49] and other references
therein.
103
See below (6.16.22) for experimental value.
up to gauge terms, while for a massive one with mass M, we have from (4.7.20)
Q Q 1

C .Q/ D C 2
: (5.20.2)
M Q C M2 i
The massless propagator gives rise to damping 1=Q2 at high energies and is a
very welcome property in establishing renormalizability of a theory. The massive
one goes to a constant at high energies and renders the renormalizability of a theory
obscure. On the other hand, if the theory is gauge invariant, one may hope to infer
renormalizability of the theory from its massless vector boson counterpart. This has
been a key result in the success of present non-abelian gauge theories with massive
vector bosons.
In the following subsection, we consider a change of scalar field variables of
integrations in a path integral of a generating functional, needed to establish the so-
called Goldstone Theorem. Spontaneous symmetry breaking in the abelian case is
treated in the following subsection with the generation of a massive vector boson.
Spontaneous symmetry breaking in non-abelian gauge theories is considered in the
next chapter.
5.20.1 Change of Real Field Variables of Integration in a Path

Integral
Suppose we are given a Lagrangian density L , which among other fields, depends
on n real scalar fields 1 .x/; : : : ; n .x/, n > 1, and is invariant under the
transformation

i .x/ ! ı ij C i ı" .t/ij j .x/ D e
j .x/; .t/ij D .t/ji ; (5.20.3)
where t is a Hermitian matrix and, due to the reality of the fields 1 ; : : : ; n , it is

anti-symmetric as stated above, and ı" is infinitesimal.
We couple the fields 1 ; : : : ; n to external sources J 1 ; : : : ; J n , and introduce
the lagrangian density
L1 .x/ D L .x/ C J i .x/ i .x/: (5.20.4)
Under the transformation of the real fields in (5.20.3),
e1
L1 ! L1 C i ı
J i .t/ij j D L (5.20.5)
We investigate the nature of the path integral representation of the amplitude

given below, associated with the Lagrangian L1 , under a change of the variables
of the field variables of integrations 1 ; : : : ; n , up to first order in ı", as defined
in (5.20.3):
• h 0C j j .x/ j 0 i
h 0C j 0 i D eiWŒ J ; WD h j .x/i: (5.20.6)
•J j .x/ h 0C j 0 i
where J D .J 1 ; : : : ; J n /. In particular, we note that due to the anti-symmetry of

the matrix t, the Jacobian of the transformation will involve the determinant of a
matrix with ones along the diagonal and with all the elements off thediagonal being
of order ı
. That is, in particular, de 1 .x/ de
2 .x/ D .1 C O ı"/2 d 1 .x/ d 2 .x/.
Hence for part of the measure of the path integral, as a factor, involving the scalar
fields above, we have, up to first order in ı
,
Y Y
Œ De
i D Œ D i : (5.20.7)
i i
Also from (5.20.5),

R h Z i R
i .dy/e
e L 1 .y/
D 1 .dx/ ı" J i .x/ tij j .x/ ei .dy/e
L 1 .y/
: (5.20.8)
Hence, invoking invariance, to first order in ı", we have

h Z i
eiWŒ J D 1 ı
.dx/ J i .x/ tij h j .x/i eiWŒ J ; (5.20.9)
leading to the equation

Z
.dx/ J i .x/ tij h j .x/i D 0: (5.20.10)
5.20.2 Goldstone Bosons and Spontaneous Symmetry Breaking
At this stage we carry out a so-called Legendre transform of WŒ J , as a functional

of J, to a functional Œ hi of hi, which is defined by
Z
Œhi D WŒ J .dx 0 / J j .x 0 /h j .x 0 /i; (5.20.11)
R
referred to as an effective action, as it coincides with the classical action .dx/ Lc .x/
written, in particular, in terms of hi, when radiative corrections are neglected.
From the chain rule
Z 0
•WŒ J 0 •WŒ J •J .x /
j
D .dx / ; (5.20.12)
•h i .x/i •J j .x 0 / •h i .x/i
from which we obtain
• Œhi
D J i .x/; (5.20.13)
•h i .x/i
and the basic Eq. (5.20.10) takes the form

Z
• Œhi ij j
.dx/ t h .x/i D 0: (5.20.14)
•h i .x/i
We note that
Z Z
i k .4/ •J i .x/
0 •J i .x/ •h j .y/i •J i .x/
ı ı .x; x / D D .dy/ D .dy/ j k .y; x0 /;
•J k .x 0 / •h j .y/i •J k .x 0 / •h j .y/i
(5.20.15)
where the j k .y; x 0 / denote propagators (correlations functions) of the fields i ,
and the above equation leads to
•J i .x/
D 1
ij .x; y/: (5.20.16)
•h j .y/i
Hence (5.20.13) gives
•2 Œhi
D 1
ij .x; y/: (5.20.17)
•h i .x/i•h j .y/i
For a time independent sources J.x/, h 0C j 0 i is described through a time

evolution from time 1 to C1. Accordingly, wŒ J would then necessarily
involve the multiplicative factor T ! 1, corresponding to the total time just
described, due to the unitarity of the time evolution operator. In the other extreme
when J is independent of the space variable as well, then wŒ J will involve
the multiplicative factor ˝, denoting the spacetime volume, and we recover a
translational invariant theory. Finally, in the limit J ! 0, we may write
Œ hi D ˝ Veff ; (5.20.18)
where Veff is referred to as the effective potential, and (5.20.14) takes the form104
@ Veff ij j
t c D 0; (5.20.19)
@ci
104
Note that functional derivatives have now become simply partial derivatives.
where due to translational invariance, the matrix elements h i .x/i become simply
expressions independent of x, which we have conveniently denoted by ci . Upon
taking another derivative of (5.20.19), with respect to ck , we obtain
@ 2 Veff ij j ˇˇ
t c ˇ D 0; (5.20.20)
@ck @ci c Dc
where
@ Veff ˇˇ
ˇ D 0; i D 1; : : : ; n; (5.20.21)
@ ci c Dc
and the c are at the minimum of the effective potential. On the other hand
Z
.dx/.dy/ 1 e1
ij .x y/ D ˝ ij .0/; (5.20.22)
where the e1 .0/ denote the inverse of the propagator functions in the momentum
ij
description at zero momentum.
From (5.20.17), (5.20.18), (5.20.22), we may then simply replace (5.20.20) by
e1
ij j
k i .0/ t c D 0: (5.20.23)
j
Accordingly, for non-vanishing tij c , for some i, corresponding to spontaneous
symmetry breaking, the e1 .0/ vanish, indicating the presence of zero mass
ki
particles in the theory, referred to as Goldstone bosons. This is the content of the
celebrated Goldstone Theorem.105
It is interesting to see how the symmetry of the Lagrangian density L is
broken by the vacuum. To this end, from Wigner’s symmetry transformations
(see Sect. 2.1), the continuous transformation in (5.20.3) is implemented, in the
vector space in which the field operators act, via a unitary operator, which for
infinitesimal transformations would read as U D I i ı" G, where G is the
generator of transformation - an operator. Then in matrix form we have the relation:
U 1 U D .I C i ı" t/ which from (5.20.3) yields the relation
Œ G; D t ; (5.20.24)
where we recall that t is a (numerical) matrix. Upon taking the vacuum expectation
value of the above, we obtain
hvacj G G jvaci D t hvacjjvaci D t c : (5.20.25)
105
Goldstone [28] and Goldstone et al. [29].
Hence for a vector t c which is non-zero in (5.20.23), we cannot have G jvaci

equal to zero for all of its matrix elements, showing that the symmetry is broken by
the vacuum.
As an illustration of the above result, the example of interest here is the classic
one with the following Lagrangian density involving a complex scalar field:
L D @ ˚ @ ˚ C 2 ˚ ˚ .˚ ˚/2 ; (5.20.26)
V D 2 ˚ ˚ C .˚ ˚/2 : (5.20.27)
First note that the coefficient of 2 in the Lagrangian density in (5.20.26), is a “C”
rather than a “”, and this is what we need for spontaneous symmetry breaking,
i.e., for the field ˚ to develop a non vanishing expectation value in the vacuum.106
Also to avoid problems with unboundedness of the corresponding Hamiltonian p from
below, is taken to be positive definite. Upon writing ˚ D .1 C i 2 /= 2, the
above Lagrangian density becomes
1 1
L D @ 1 @ 1 @ 2 @ 2 V; (5.20.28)
2 2
where now
2 2 2
VD 1 C 22 C 12 C 22 : (5.20.29)
2 4
One may rotate the system to bring h2 i 2c to zero, and h1 i 1c to some
value v, which is determined by minimizing the potential. By neglecting radiative
corrections, for simplicity, we have107
@V 2
D 2 1c C 1c 1c C 22c D 0; v 2 D 2 : (5.20.30)
@1c
Upon writing 1 D '1 C v, 2 D ' 2 , with h' i i D 0, the quadratic part of the
Lagrangian density in (5.20.28) is easily worked out to be
1 1 .2v 2 / 2
Lquad D @ '1 @ '1 @ ' 2 @ ' 2 '1 C 0:' 22 ; (5.20.31)
2 2 2
with ' 2 representing the Goldstone boson, involving no mass.
106
We will neglect radiative corrections in the subsequent analysis. Although we are not consid-
ering radiative corrections here, the .˚ ˚/2 term in the Lagrangian density is important for
renormalizability of the theory.
107
Note that if the coefficient of 2 is of opposite sign to the one in (5.20.26), one obtains v D 0.
Now we consider scalar boson electrodynamics with the sign of the coefficient
of the parameter 2 is as given in (5.20.26). We now have a local broken symmetry,
and, as we will see below, the long range force, associated with the gauge field,
becomes screened in which the Goldstone boson combines with the gauge field to
create a massive one. To this end, the Lagrangian density in question is given by108
1
L D F F @ C i eA ˚ @ i eA ˚ V; (5.20.32)
4
where V is given in (5.20.27), F D @ A @ A . Upon defining the field
V D A C @ 2 =ve; G D @ V @ V ; (5.20.33)
the quadratic part of the above Lagrangian density is easily worked out to be
1 e2 v 2 1 .2v 2 / 2
Lquad D G G V V @ '1 @ '1 '1 ; (5.20.34)
4 2 2 2
and the Goldstone boson combined with the gauge field generating a massive vector
boson. The resulting theory describes a massive neutral scalar particle and a massive
vector boson. The scalar field 1 .x/ with non-vanishing vacuum expectation value
v is usually referred to as a Higgs field, and the resulting massive neutral spin zero
particle as a Higgs boson. Finally the mechanism for the generation of the massive
particles, as just described, is referred as the Higgs mechanism.109 Also note that the
spin zero massive neutral particle is described by the field '1 , i.e., after the vacuum
expectation value of 1 has been subtracted out from the latter.
It is remarkable, that even if 2 D 0, radiative corrections lead to spontaneous
symmetry breaking generating again a theory with one massive neutral scalar
particle and a massive vector boson.110
Problems
5.1 Derive the general expressions of the polarization vectors in (5.2.7), (5.2.8),
(5.2.9), (5.2.10), (5.2.11), (5.2.12) and (5.2.13).
5.2 Verify directly from (5.3.8) and (5.3.5) that @ h 0C jA .x/ j 0 i D
h 0C j.x/ j 0 i, consistent with the gauge condition (5.3.1).
5.3 Derive the key identity (5.4.15) in the Casimir Effect problem.
108
Since we are not considering renormalizability aspects here, we use the notation for e.
109
Englert and Brout [21] and Higgs [31–33]. See also Anderson [2].
110
For the relevant details see Coleman and Weinberg [17].
5.4 Consider two independent, but identical, assumed non-interacting, sources,

emitting photons in uniformly arbitrary directions, situated at different positions
R1 , R2 , and described by two current distributions giving rise to a resultant current
distribution given by
X Z .dk/ ik .xRj / ik0 x0
J .x/ D e e J .k/; where
j D 1;2
.2 /4
1
R1 D .0; 0; a=2/; R2 D .0; 0; a=2/; and J .k/ e p f .jkj/:
2
The angular correlation of the emerging photons is defined by111

Z Z
hci D d˝1 d˝2 n1 n2 fka .n1 ; n2 /:
where ni D ki =jki j i D 1; 2, jk1 j D jk2 j k, and fka .n1 ; n2 / is the corresponding

probability density. Evaluate hci as a function of k a. Argue that for small distance
separation a, the two sources may not be considered as independent and non-
interacting, and on other hand, for large separations, the correlation of photons
is not a meaningful concept either. In the intermediate meaningful range, a large
correlation, implies that the photons tend to travel in the same direction. (See
[10, 54, 55].)
5.5 Establish the identity in (5.7.25).
5.6 Show that (5.7.7) may be rewritten as @ h 0C jA .x/ j 0 i being equal to
Z Z
0 0
.dx /DC .x x /@ 0 J .x 0 /h 0C j 0 i C i e0 .dx 0 /.dx 00 / DC .x; x 0 /
h i
.x 00 / SC .x 00 ; x 0 I e0b
A/ .x 0 / .x0 / SC .x 0 ; x 00 I e0b
A/ .x 00 / h 0C j0 i;
and verify that this equality also follows upon the application of the expression for
h 0C j 0 i in (5.7.27).
5.7 Derive the explicit expressions for a0 ; a1 ; a2 ; a3 in the expansion of
h 0C j 0 i in terms of connected parts with external lines as given, respectively
in (5.8.8), (5.8.9), (5.8.10), (5.8.11).
5.8 Show that the summation over spins of the expression in (5.9.31) is given
by (5.9.32).
111
This was introduced by Bialinicki-Birula and Bialinicka-Birula [10].
5.9 Consider the polarization correlations of the outgoing electrons in e e !

e e scattering, in which the initial state has been prepared with one of the ingoing
electrons moving with momentum p1 D p in the y-direction, with spin along the
z-axis, and the other moving in the opposite direction with momentum p2 D p,
and spin along the z direction. Suppose that the outgoing electrons have emerged
with momenta in opposite direction of the x-axis. The spins are measured along
unit vectors n1 , n2 , specified by angles 1 , 2 , as shown in the figure for some
possible outcomes:
z χ2 −p z χ2 −p
n2 n2
p −p p −p
y y
χ1 χ1
n1 n1
p p
x x
Show that the probability of

simultaneous
measurements of the spins along n1 n2 is
given by PŒ1 ; 2 D sin2 .1 2 /=2 =2. And if only one of the spins is measured,
then PŒ1 ; D PŒ; 2 D 1=2. Interpret these results.
5.10 Establish the identity . p2 C m2 / D 2m. p C m/ . p C m/2 .
5.11 The logarithmically ultraviolet divergent constant, that typically occurs in
QED is given by the integral Cuv in (5.10.21). The latter may be rigorously defined
with an ultraviolet
cut-off
as in
(5.10.27).
Show as indicated in that equation, with
the cut-off, Cuv D ln 2 =m2 1 for 2 ! 1.
5.12 Two basic infrared divergent constants that occur in QED, are given by the
following two integrals, defined in terms of a fictitious mass of the photon,
Z 1 Z 1
x dx x3 dx
Cir D
2 ; Dir D
2 2 :
0 x2 C m2
.1 x/ 0 x2 C m2
.1 x/

Show thatfor .2 =m2 / ! 0, Cir D .1=2/ ln 2 =m2 /, Dir D Cir .1=2/ D
.1=2/Œln 2 =m2 / C 1.
5.13 Evaluate ˙1 . p/ in (5.10.14) to obtain the expression in (5.10.17).
5.14 Use (5.10.57), to derive the expression of the modification of the Coulomb
interaction, to second order, due to radiative corrections, at small distances, i.e., for
m jxj 1=2, as given in (5.10.66).
5.15 Show that u.p 0 ; 0 / ˘ .k/ .1=k2 /u.p; / D u.p 0 ; 0 / u.p; / ˘.k2 / ,
which ! u.p 0 ; 0 / u.p; / .1 Z3 /, for k2 ! 0, where ˘ .k/ is given
in (5.10.42).
5.16 Suppose that the charge distribution of the proton is given by .x/ D
.1=8 3 / ejxj= , where 1= 2 D 0:81M 2p , where Mp is the mass of the proton.
Estimate the energy shift in a hydrogen ` D 0 – energy level due to this distribution.
Work in units with the Coulomb potential given by ˛=jxj.

5.17 Express the matrix element h 0C j F k .x/A .y/ j 0 i as functional deriva-
C
tives acting on h 0C j 0 i in the Coulomb gauge.
5.18 Derive the equal-time commutation relations of the field components Ai .x/ in
the Coulomb gauge, using the equality in (5.14.24).
5.19 Establish the identity in (5.15.7).
5.20 Derive the identity in (5.15.16), written here in a convenient matrix multipli-
cation form:
• •

i
exp i e0 @ exp i SC D exp i ei e0 SC eie0 :
• •
5.21 Show that in the absence of the external Fermi sources, the generating
functionals in the Landau gauge, and in arbitrary covariant gauges, specified by
the parameter , are simply related by (5.15.24).
5.22 Show that the exact electron propagator, in an arbitrary covariant gauge,
specified by the parameter , is related to the one in the Landau gauge, in the
presence of a non-zero external current J , is as given in (5.15.26).
References
1. Adler, S. L. (1972). Short-distance behavior of quantum electrodynamics and an eigenvalue

condition for ˛. Physical Review D, 5, 3021–3047.
2. Anderson, P. W. (1963). Plasmons, gauge invariance, and Mass. Physical Review, 130, 439–
442.
3. Baker, M., & Johnson, K. (1969). Quantum electrodynamics at small distances. Physical
Review D, 183, 1292–1299.
4. Baker, M., & Lee, C. (1977). Overlapping-divergence-free Skeleton expansion in non-Abelian
gauge theories. Physical Review D, 15, 2201–2234.
5. Bell, J. S. (1989). Speakable and unspeakable in quantum mechanics. Cambridge: Cambridge
University Press.
6. Bell, J. S., & Aspect, A. (2004). Speakable and unspeakable in quantum mechanics: Collected
papers on quantum philosophy (2nd ed.). Cambridge: Cambridge University Press.
8. Bethe, H. (1947). The electromagnetic shift of energy levels. Physical Review, 72, 339–341.
References 365
9. Bethe, H. A., Brown, L. M., & Stehn, J. R. (1950). Numerical value of the lamb shift. Physical
Review, 77, 370–374.
10. Bialinicki-Birula, I., & Bialinicka-Birula, Z. (1990). Angular correlations of photons. Physical
Review, A42, 2829–2838.
11. Bjorken, J. D., & Drell, S. D. (1965). Relativistic quantum fields. New York/San Fran-
cisco/London: McGraw-Hill.
12. Bogoliubov, N. N., & Shirkov, D. V. (1959). Introduction to the theory of quantized fields.
New York: Interscience.
13. Bressi, G., Carugno, G., Onofrio, R., & Ruoso, G. (2002). Measurement of the Casimir force
between parallel metallic surfaces. Physical Review Letters, 88, 041804:1–4.
14. Brown, L. S. (1995). Quantum field theory. Cambridge: Cambridge University Press.
15. Callan, C. G. (1970). Broken scale invariance in scalar field theory. Physical Review, D2,
1541–1547.
16. Casimir, H. G. (1948). On the attraction between two perfectly conducting plates. Proceedings
of the Koninklijke Nederlandse Akademie van Wetenschappen, B 51, 793–795.
17. Coleman, S., & Weinberg, E. (1973). Radiative corrections as the orign of spontaneous
symmetry breaking. Physical Review, D7, 1888–1910.
18. de Rafaël, E., & Rosner, J. L. (1974). Short-distance behavior of quantum electrodynanics and
the Callan-Symanzik equation for the photon. Annals of Physics (NY), 82, 369–406.
19. Elizalde, E. (1995). Ten physical applications of spectral zeta functions. Berlin: Springer.
20. Elizalde, E., Odintsov, S. D., Romeo, A., Bytsenko, A. A., & Zerbini, S. (1994). Zeta
regularization techniques with applications. Singapore: World Scientific.
21. Englert, F., & Brout, R. (1964). Broken symmetry and the mass of gauge vector bosons.
22. Erickson, G. W., & Yennie, D. R. (1965). Radiative level shifts, I. Formulation and lowest
order lamb shift. Annals of Physics (NY), 35, 271–313.
23. Erler, J. (1999). Calculation of the QED coupling ˛.MO Z / in the modified minimal subtraction
scheme. Physical Review D, 59, 054008, 1–7.
24. Fazio, G. G., Jelly, J. V., & Charman, W. N. (1970). Generation of Cherenkov light flashes by
cosmic radiation within the eyes of the Apollo astronauts. Nature, 228, 260–264.
25. Foley, H. M., & Kusch, P. (1948). The magnetic moment of the electron. Physical Review, 74,
250–263.
26. Fox, J. A., & Yennie, D. R. (1973). Some formal aspects of the lamb shift problem. Annals of
Physics (NY), 81, 438–480.
27. Gell-Mann, M., & Low, F. E. (1954). Quantum electrodynamics at small distances. Physical
Review, 95, 1300–1312.
28. Goldstone, J. (1961). Field theories with superconductor solutions. Il Nuovo Cimento,
19, 154–164.
29. Goldstone, J., Salam A., & Weinberg, S. (1962). Broken symmetries. Physical Review, 127,
965–970.
30. Gorishnii, S. G., et al. (1991). The analytic four loop corrections to the QED beta function in
the MS scheme and the QED Psi function: Total reevaluation. Physics Letters, B256, 81–86.
31. Higgs, P. W. (1964a). Broken symmetries, massles particles and gauge fields. Physics Letters,
12, 132–133.
32. Higgs, P. W. (1964b). Broken symmetries and the masses of gauge bosons. Physical Review
Letters, 13, 508–509.
33. Higgs, P. W. (1966). Spontaneous symmetry breaking without massless particles. Physical
Review, 145, 1156–1163.
34. Johnson, K. (1968). 9th Latin American Scool of Physics, Santiago de Chile. In K. Johnson
& I. Saavedra (Eds.), Solid state physics, nuclear physics, and particle physics. New York: W.
A. Benjamin
35. Källen, G. (1952). On the definition of renormalization constants in quantum electrodynamics.
Helvetica Physica Acta, 25, 417–434.
36. Kenneth, O., Klich, I., Mann, A., & Revzen, M. (2002). Repulsive Casimir forces. Physical
Review Letters, 89, 033001:1–4.
37. Kinoshita, T., & Yennie, D. R. (1990). High precision tests of quantum electrodynamics – an
overview. In T. Kinoshita (Ed.), Quantum electrodynamics: Advanced series on directions in
high energy physics (Vol. 7). Singapore: World Scientific.
38. Krasnikov, N. V. (1989). Is finite charge renormalization possible in quantum electrodynamics?
39. Kusch, P., & Foley, H. M. (1947). Precision measurement of the ratio of the atomic ‘g values’
in the 2 P3=2 and 2 P1=2 states of Gallium. Physical Review, 72, 1256.
40. Lamb, W. E., Jr., & Retherford, R. C. (1947). Fine structure of the Hydrogen atom by a
microwave method. Zeitschrift für Physik, 72, 241–243. Reprinted in Schwinger (1958).
41. Lamoreaux, S. K. (1997). Demonstration of the Casimir force in the 0.6 to 6 m range.
42. Lehmann, H. (1954). Properties of propagation functions and renormalization constants of
quantized fields. Il Nuovo Cimento, 11, 342–357.
43. Lewis, G. N. (1926). The conservation of photons. Nature, 118, 874–875.
44. Manoukian, E. B. (1975). Fundamental identity for the infinite-order-zero nature in quantum
electrodynamics. Physical Review, D12, 3365–3367.
45. Manoukian, E. B. (1983). Renormalization. New York/London/Paris: Academic.
46. Manoukian, E. B. (1984). Proof of the decoupling theorem of field theory in Minkowski space.
Journal of Mathematical Physics, 25, 1519–1523.
D34, 3739–3749.
49. Manoukian, E. B. (1986b). Generalized conditions for the decoupling theorem of quantum field
theory in Minkowski space with particles of vanishing small masses. Journal of Mathematical
Physics, 27, 1879–1882.
50. Manoukian, E. B. (1987). Casimir effect, the gauge problem and seagull terms. Journal of
Physics A, 20, 2827–2832.
51. Manoukian, E. B. (1987). On the relativistic invariance of QED in the Coulomb gauge and
field transformations. Journal of Physics, G13, 1013–1021.
52. Manoukian, E. B. (1988a). Charged particle emission and detection sources in quantum field
theory and infrared photons. Fortschritte der Physik, 36, 1–7.
53. Manoukian, E. B. (1988b). Gauge invariance properties of transition amplitudes in gauge
theories. I. International Journal of Theoretical Physics, 27, 787–800.
54. Manoukian, E. B. (1992). Field-theoretical view of the angular correlation of photons. Physical
Review, A46, 2962–2964.
55. Manoukian, E. B. (1994). Particle correlation in quantum field theory. Fortschritte der Physik,
42, 743–763.
57. Manoukian, E. B. (2011). Modern concepts and theorems of mathematical statistics.
New York: Springer. Paperback Edition.
58. Manoukian, E. B. (2015). Vacuum-to-vacuum transition amplitude and the classic radiation
theory. Radiation Physics and Chemistry, 106, 268–270.
59. Manoukian, E. B., & Charuchittapan, D. (2000). Quantum electrodynamics of C̆erenkov
radiation at finite temperature. International Journal of Theoretical Physics, 39, 2197–2206.
60. Manoukian, E. B., & Siranan, S. (2005). Action principle and algebraic approach to gauge
transformations in gauge theories. International Journal of Theoretical Physics, 44, 53–62.
61. Manoukian, E. B., & Yongram, N. (2004). Speed dependent polarization correlations in QED.
European Physical Journal, D 31, 137–143.
62. McNulty, P. J., Pease, V. P., & Bond, V. P. (1976). Role of Cerenkov radiation in the eye-flashes
observed by Apollo astronauts. Life Sciences and Space Research, 14, 205–217.
References 367
63. Mele, S. (2006). Measurements of the running of the electromagnetic coupling at LEP. In
XXVI Physics in Collision, Búzios, Rio de Janeiro, 6–9 July.
64. Milton, K. A., et al. (2012). Repulsive Casimir and Casimir-Polder forces. Journal of Physics,
A 45, 374006.
65. Ovsyannikov, L. V. (1956). General solution to renormalization group equations. Doklady
Akademii Nauk SSSR, 109, 1112–1115.
66. Pinsky, L. S., et al. (1974). Light flashes observed by astronauts on Apollo 11 through Apollo
17. Science, 183, 957–959.
67. Schwartz, C., & Tiemann, J. J. (1959). New calculation of the numerical value of the lamb
shift. Annals of Physics (NY), 6, 178–187.
68. Schwinger, J. (1948). On quantum-electrodynamics and the magnetic moment of the electron.
Physical Review, 73, 416.
69. Schwinger, J. (Ed.) (1958). Selected papers on quantum electrodynamics. New York: Dover
Publications.
70. Schwinger, J. (1969). Particles and sources. New York: Gordon and Breach.
71. Schwinger, J. (1973). Particles, sources, and fields (Vol. II). Reading: Addison-Wesley.
72. Schwinger, J. (1975). Casimir effect in source theory. Letters in Mathematical Physics, 1,
43–47.
73. Sparnaay, M. J. (1958). Measurement of attractive forces between flat plates. Physica, 24,
751–764.
74. Stueckelberg, E. C. G., & Peterman, A. (1953). La Normalisation des Constantes dans la
Théorie des Quanta. Helvetica Physica Acta, 26, 499–520.
75. Symanzik, K. (1970). Small distance behaviour in field theory and power counting. Commu-
nications in Mathematical Physics, 18, 227–246.
76. Symanzik, K. (1971). Small distance behavior in field theory. In G. Höhler (Ed.), Springer
tracts in modern physics (Vol. 57). New York: Springer.
77. Symanzik, K. (1971). Small distance behaviour analysis in field theory and Wilson expansions.
Communications in Mathematical Physics, 23, 49–86.
78. Takahashi, Y. (1957). On the generalized ward identity. Nuovo Cimento, 6, 370–375.
79. Ward, J. C. (1950). An identity in quantum electrodynamics. Physical Review, 78, 182.
80. Weinberg, S. (1960). High-energy behavior in quantum field theory. Physical Review, 118,
838–849.
81. Yennie, D. R., & Suura, H. (1957). Higher order radiative corrections to electron scattering.
Physical Review, 105, 1378–1382
82. Yongram, N., & Manoukian, E. B. (2003). Joint probabilities of photon polarization
correlations in eC e annihilation. International Journal of Theoretical Physics, 42, 1755–
1764.
83. Yongram, N., & Manoukian, E. B. (2013). Quantum field theory analysis of polarizations
correlations, entanglement and Bell’s inequality: Explicit processes. Fortschritte der Physik,
61, 668–684.
Recommended Reading
Press.
2. Kinoshita, T. (Ed.) (1990). “Quantum electrodynamics”: Advanced series on directions in high
energy physics (Vol. 7). Singapore: World Scientific.
3. Manoukian, E. B. (1983). Renormalization. New York/London/Paris: Academic.
4. Manoukian, E. B. (1986). Action principle and quantization of gauge fields. Physical Review,
D34, 3739–3749.

6. Schwinger, J. (1973a). Particles, sources, and fields (Vol. II). Reading: Addison-Wesley.
7. Schwinger, J. (1973b). A report on quantum electrodynamics. In J. Mehra (Ed.), The physicist’s
conception of nature. Dordrecht/Holland: D. Reidel Publishing Company.
8. Weinberg, S. (1996). The quantum theory of fields. II: Modern applications. Cambridge:
Cambridge University Press.
9. Yongram, N., Manoukian, E. B. (2013). Quantum field theory analysis of polarizations
correlations, entanglement and Bell’s inequality: Explicit processes. Fortschritte der Physik,
61, 668–684.
Chapter 6
Non-Abelian Gauge Theories
The present chapter1 deals with the intricacies of non-abelian gauge field theories.
We consider the extension of local gauge transformations of QED, with the gauge
group U.1/ of phase transformations, to SU.N/ groups. We recall that the group
SU.N/ means the group of symmetry transformations consisting of N N unitary
matrices that are “special” in that they have determinant one. Unlike QED, SU.N/
involves .N 2 1/ non-commuting generators, and hence field theories involving
SU.N/ groups are referred to as non-abelian gauge field theories, with QED being
an abelian one. Local gauge invariance imposed on a theory, under such an SU.N/
group, leads to the introduction of .N 2 1/ interacting vector fields in a theory,
referred to as Yang-Mills fields, which mediate interactions between particles
carrying specific quantum numbers associated with the group. This happens in a
similar way as in QED, where local gauge invariance leads to the introduction of
the photon into a theory involving charged particles, and mediates the interaction
between charged particles. A photon, however, is electrically neutral and does not
have a direct self interaction. The vector particles of an SU.N/ group, however, do
carry specific quantum numbers, such as charge, color, . . . , and do respond directly
to each other, i.e., they interact. We will see in Sect. 6.1, how such local symmetry
groups of transformations are generated by insisting that one has the freedom in
carrying transformations, depending explicitly on the spacetime point, at which a
transformation is being carried out. These local gauge symmetry groups do not
only assign specific quantum numbers to the particles in consideration but also give
rise to the underlying dynamical theory of these particles, by being guided, in its
development, by the QED theory.
The quantization of Yang-Mills field theories in the presence of matter is treated
in Sects. 6.2 and 6.3. The first non-abelian gauge theory we consider is that
of quantum chromodynamics as the modern theory of strong interaction and is
1
It is a good idea to review the content of the introductory chapter of the book, especially of those
aspects dealt with in this chapter.

DOI 10.1007/978-3-319-30939-2_6
370 6 Non-Abelian Gauge Theories
developed in Sects. 6.4, 6.5, 6.6, 6.7, 6.8, 6.9, 6.10, 6.11, 6.12, and 6.13. It is based
on the color group SU.3/C , often denoted simply by SU.3/, involving quarks,
coming in three colors, and 8 (massless) vector particles (the gluons). We will learn
that there are many reasons why quarks should carry such a quantum number, and
why three of them (Sect. 6.4). The renormalization group, which shows how the
theory responds to scale transformations, in turn, leads to the description of the high-
energy behavior of the theory simply in terms of an effective coupling which, thanks
to the non-abelian gauge character of the theory, becomes weaker, with the increase
of energy, and eventually becomes zero, a phenomenon referred to as asymptotic
freedom. This is the subject of Sects. 6.6 and 6.7. We will see how the non-abelian
gauge character of the theory with interacting gluons is responsible for this very
welcome behavior of the theory at high energies. The smallness of the effective
coupling constant at high energies allows one to carry out perturbation expansions in
such a coupling at high energies such as in describing e eC annhilation (Sect. 6.8).
Asymptotic freedom is a key result in explaining several experiments such as the
deep inelastic experiment (Sects. 6.9, 6.10, and 6.11) where a virtual photon imparts
high energies to a proton, showing that a proton is considered to be consisting
of weakly interacting quarks, quark-antiquark pairs and gluons, collectively called
partons, within the extension of the proton. Some interesting aspects of quark and
gluon splittings, as well as the formations of QCD jets, in e eC annhilation, are
investigated as well (Sect. 6.10). Most commonly two jets are observed in opposite
directions to a quark-antiquark produced in the CM frame of the annhilating e eC
pair. Low energy (soft) gluons radiated by a parent quark (antiquark), give rise
to an enhanced probability, due to an inherited infrared divergence as in QED
(Sect. 5.12), when they are emitted in the same direction (collinear) as the parent
quark (antiquark). Such a result, based on the observation of the 2 back-to-back
jets events, not only gives a strong support of the existence of quarks but due to
the orientations of the axis of the back-to-back jets imply the spin 1/2 character of
quarks. As a consequence of asymptotic freedom, occasionally, i.e., with a smaller
probability, due to the vanishing of the effective coupling at high energies, a high
energy (hard) non-collinear gluon may be emitted as well leading to a three jets
events and so on. As quarks are not observed, it is expected that the effective
coupling between them becomes large (infrared slavery) when they separate beyond
1 fm, the approximate extension of a proton, and are thus confined within the proton.
When a coupling becomes large, one has the tendency to carry out perturbation
theory in the inverse of the coupling. Lattice gauge theory, where spacetime is
considered to have a lattice structure, is quite suitable to do this and, in turn,
allows one to investigate the nature of quark confinement. This underlying theory is
developed in Sects. 6.12 and 6.13.
The electroweak theory, as a unification of the electromagnetic and weak
interactions, based on the product group SU.2/U.1/ of leptons and the underlying
vector bosons is treated in Sect. 6.14. Special attention is given to an experimental
determination of sin2 W , where W provides a measure of the mixing occurring
between the electromagnetic and weak couplings. This section includes the study of
the theory of spontaneous symmetry breaking, where masses are generated for some
6.1 Concept of Gauge Fields and Internal Degrees of Freedom: From. . . 371
of the particles via the Higgs mechanism. Here by spontaneous symmetry breaking
it is meant that a symmetry of the Hamiltonian is not obeyed by the vacuum. The
problem of masses of neutrinos and the theory of “neutrino oscillations”, very much
relevant to neutrino masses and mass differences, are also presented. Quarks are
incorporated into the theory in Sect. 6.15. Special attention is also given to the
important aspect of the absence of anomalies in the standard model of elementary
particles which is based on the product group SU.3/SU.2/U.1/, as investigated
in Sect. 6.15, and is a key result for its renormalizability.
The standard model of elementary particles, based on the product of the above
three groups, just mentioned, involves three different couplings, one for each group.
The embedding of this product group into a larger group, such as SU.5/, leads to
a unification of these couplings into a theory with a single coupling at some high
energies, and the quantization of the electromagnetic fractional charges of quarks in
units of e=3 emerges naturally. The underlying theory of this simplest grand unified
theory is developed in Sect. 6.16 and its limitations are discussed.
6.1 Concept of Gauge Fields and Internal Degrees

of Freedom: From Geometry to Dynamics
In a geometrical setting, gauge fields tell us something about systems without

actually being considered as dynamical variables and, in turn, allow one to describe
interactions between charged fields, or between fields carrying other internal
quantum numbers, such as isospin and color, or may describe the interaction of
all fields simply due to their energy-momentum content. In the present section, we
consider aspects of gauge fields associated with internal quantum numbers. The
gauge field corresponding to the last case will be treated in an accompanying volume
by the author,2 and is due to gravity felt by all particles.
We first consider the concept of a charged field, i.e., a non-Hermitian field .x/,
which, in turn, may be rewritten as D 1 C i2 in charge space (see Fig. 6.1).
Upon a phase transformation by a phase factor #.x/ which may, in general, depend
on spacetime coordinates, it is transformed into a field 0 .x/, as shown in Fig. 6.1,
and is given by
.x/ ! ei#.x/ .x/ D 0 .x/; (6.1.1)
and is referred to as a gauge transformation. This may be equivalently achieved by

the rotation of the coordinate system, of charge space, “set up” at spacetime point
x, in opposite direction. Thus local gauge transformations of a charged field may
2
Quantum Field Theory II: Introductions to Quantum Gravity, Supersymmetry, and String Theory,
Springer (2016).
χ (x)
χ2 (x) χ (x)
ϑ (x) χ 1 (x) = χ 1 (x) cos ϑ (x) − χ 2(x) sin ϑ (x)
χ 2 (x) = χ 1 (x) sin ϑ (x) + χ 2(x) cos ϑ (x)
χ1 (x)
Fig. 6.1 Phase transformation of a charged field in charge space with the field, at a spacetime
point x, represented in terms of its real and imaginary parts
be described by “setting up”, at every point of spacetime x, a coordinate system

representing charge space.
The above is easily generalized to fields .x/ with several internal degrees of
freedom, represented, say, by N components corresponding to various degrees of
freedoms, which transform as
i .x/ ! .ei#.x/t /i j j .x/; #.x/ D .# 1 .x/; : : : ; # M .x//; i; j D 1; : : : ; N:

(6.1.2)
where the indices i,j specify the various components of a field in internal space,
with M matrices ta : t D .t1 ; : : : ; tM /, represented by N N matrices with basic
properties to be spelled out below. By considering local transformations, where
#.x/, is real, and as indicated, may depend on the spacetime point x, and one,
in particular, may describe both the spacetime position of a field and its internal
(charge, isospin, color,. . . ) state.
In order to describe dynamical theories, the derivative of a field should be
also considered in the general case, in such a way that everything turns out to
be invariant under the transformation rule in (6.1.2) with #.x/ depending on the
spacetime points. Consider a curve parametrized by a parameter , and coordinate
labels x ./ in Minkowski spacetime. The field .x/ will, in particular, depend on
the parameter through the dependence of x on .
At every point x of spacetime, we set up an N dimensional Euclidean space,
associated with internal space, with basis vector fields fei .x/g, and expand the field
.x/ at the point x as follows
.x/ D i .x/ ei .x/: (6.1.3)
We need a structure to be able to compare the relative generalized “orientation” of a

coordinate system, associated with internal space, at a spacetime point x from one
infinitesimally close to it as we move along such a curve (Fig. 6.2).
Fig. 6.2 Relative coordinate

rotations are considered and
formally depicted in an
abstract N dimensional
Euclidean space
corresponding to two
neighboring points on a curve
in Minkowski spacetime. This
leads to the concept of
connection needed to relate
these relative rotations λ
λ +dλ
Minkowski Spacetime
Such a structure is called the connection and arises naturally in the following way.
The change of the field .x.// as it moves along the curve may be represented
from (6.1.3) by
d.x.// D @ i .x/dx ei .x/ C i .x/ dei .x/: (6.1.4)
Now dei .x/ at point x ./, must vanish for dx ! 0, and it may be expanded in
terms of the basis vectors ej , i.e.,
dei D i i j .x/ ej .x/dx (6.1.5)
The totality of the expansion coefficients fi j .x/g is referred to as the connection
telling us how “fast” the coordinates get “rotated”, locally, as we move from a
spacetime point x to an infinitesimally close one. The i factor in (6.1.5) is chosen
for convenience.
From (6.1.4)), (6.1.5), we obtain

d.x/ D ıi j @ i i j .x/ i .x/ej .x/dx : (6.1.6)
Hence d.x/, as the field itself, may be expanded in terms of the basis vectors
ei .x/ in the local coordinate system of internal space. And we learn that the partial
derivative @ has been replaced by a new structure r in the general case, referred
to as the gauge-covariant derivative, which in matrix form, may be written as
.r / D .@ i /: (6.1.7)
With the operator of transformation given in (6.1.2) written as
V.x/ D exp Œ i# .x/ t ; (6.1.8)

.r /.x/ would transform the same way as .x/ ! .V/.x/, i.e.,
r .x/ ! V.x/ r .x/; (6.1.9)
provided transforms as
h @ i 1
.x/ ! V.x/ .x/ V .x/; (6.1.10)
i

where note that @ V.x/ V 1 .x/ D V.x/ @ V 1 .x/.
By carrying, in general, different generalized “rotations” at different spacetime
points, the field .x/ is experiencing a myriad of different twisting and providing a
complicated induced structure of the space associated with the integral degrees of
freedom (charge, isospin, color,. . . ) and which manifest their existence. In turn, one
may hold a structure “responsible” for such twisting of the field, in charge space,
by introducing the concept of curvature, locally defined at each spacetime point,
as a measure of the strength on such twisting. This gives rise to a geometrical
way of introducing fields that interact with charges and other physical entities
corresponding to other internal degrees of freedom.
The concept of curvature may be introduced in the following way. We first define
the gauge-covariant derivative of .x/ along a Lorentz vector V
V r .x/ V r .x/: (6.1.11)
Consider now the transfer of .x/ along a closed path in Minkowski spacetime

defined by the parallelogram shown below along the Lorentz vectors •1 v1 ,

•2 v2 , •1 v1 , •2 v2 , with v1 , v2 two different Lorentz vectors to form a
parallelogram, with infinitesimal •1 , •2 (Fig. 6.3).
At the point 2, we may write

.x2 / D I .x1 / C •1 v1 r1 I .x1 / C : : :
D .1 C •1 v1 r1 C /I .x1 /; (6.1.12)
ϑ1
ϑ3 3
− ξ 1v 1
4
4 ξ2 v2
− ξ2
1 v2
− − ξ2 v 2
ξ1 v ξ1 v
1 1 3 2
ξ1v 1
2 ξ 2v 2 1
Minkowski Spacetime
Fig. 6.3 Diagrams corresponding to the transfer of .x/ along a closed path in Minkowski
spacetime defined by the parallelogram as described in the text
where I .x1 / denotes the initial state, and r1 stands for r with x D x1 . Let F .x1 /
denote the final state, then
F .x1 / D .1 •2 v2 r4 C : : :/.1 •1 v1 r3 C : : :/

.1 C •2 v2 r2 C : : :/ .1 C •1 v1 r1 C : : :/I .x1 /: (6.1.13)
From which the following expression emerges for the change in twisting by going
around the closed path

F .x1 / I .x1 / D •1 •2 v1 v2 Œr ;r I .x1 /; (6.1.14)
where [.,.] stands for the commutator. The curvature is defined by
R .x/ D i Œr ;r

1
D @ .x/ @ .x/ C .x/ .x/ .x/ .x/ ; (6.1.15)
i
responsible for the alteration of the state of I .x/ by going around a closed path.
Note that the differential operators @ , @ cancel out in the final expression for
the commutator Œr ;r . R .x/ in (6.1.15)) is written in matrix form. Its matrix
elements are given by
1
R i j .x/ D @ j i .x/ @ j i .x/ C k i .x/ j k .x/ k i .x/j k .x/ :
i
(6.1.16)
Here i; j; k D 1; : : : ; N, where N corresponds to the number of the components of
the field .x/, while ; D 0; 1; 2; 3 correspond to spacetime indices.
From (6.1.10) and the definition in (6.1.15)), R , as a matrix, has the following
gauge transformation rule
R .x/ ! V.x/ R .x/ V 1 .x/; Tr R .x/R .x/ ! Tr R .x/R .x/; (6.1.17)
with the second rule implying an important invariance property, where Tr.:/ .:/i i .
The operator in (6.1.8), corresponding to the transformation rule given in (6.1.2),
gives rise to a representation of a continuous group whose elements are labeled by
a set of smooth parameters referred to as a Lie group. The group multiplication is
defined by
exp.i # 1 t/ exp.i # 2 t/ D exp.i # t/: (6.1.18)
In particular for any two components ta , tb of t, and a real number , this group
property implies that
exp.i tb / exp.i ta / exp.i tb / exp.i ta / D exp.i #c tc /; (6.1.19)

Clearly the coefficients #c depend on ta , tb and ! 0 for ! 0. We may use the

classic Baker-Campbell-Hausdorff formula for two operators A, B
1 1 1
eA eB D eACBCC ; C D ŒA; B C ŒA; ŒA; B C ŒB; ŒB; A C : : : ; (6.1.20)
2 12 12
to infer from (6.1.19) that

exp .i/2 Œta ; tb C O. 3 / C : : : D exp.i #c tc /; (6.1.21)
leading to #c D 2 fabc C : : : for a given pair .a, b/, with a choice of real numericals
fabc , and for ! 0, the following commutation relation emerges
Œ ta ; tb D i fabc tc ; (6.1.22)
providing a representation of an algebra, referred to as a Lie algebra, with

generators3 ta ; a D 1; : : : ; N. The numericals fabc are called the structure constants
of the algebra, and due to the commutator on the left-hand side of (6.1.22),
they are anti-symmetric in the indices a, b. The following identity, obtained by
multiplying (6.1.22) by td from the right, taking the trace, and using the property of
a trace
i fabc Tr.tc td / D Tr.Œ ta ; tb td / D Tr.td Œ ta ; tb / D Tr.ta Œ tb ; td / D Tr.tb Œ td ; ta /;

(6.1.23)
is easily established. We will be exclusively involved with groups satisfying the
normalization condition
1
Tr.tc td / D •c d : (6.1.24)
2
From the last two equalities in (6.1.23), we may then infer that the structure
constants fabd are totally anti-symmetric in all of its three indices, as a result of
the odd property that results in the interchange of any two generators, ta ; tb ; td in
their corresponding commutation relations.
Another representation of the algebra in (6.1.22) may be provided from the
following consideration. For any three matrices, and hence, in particular, for
matrices ta , tb , tc , we may use the easily derived Jacobi identity
Œ ta ; Œ tb ; tc C Œ tb ; Œ tc ; ta C Œ tc ; Œ ta ; tb D 0; (6.1.25)
to conclude from (6.1.22) that
fcbd fdae C fcad fdbe fabd fcde D 0; (6.1.26)
3
For simplicity of the notation, we use the same symbol for a representation matrix, in general, as
the abstract group element and no confusion should arise.
where in writing this equation, we have invoked the anti-symmetry nature of

the structure constants fabc . Accordingly, we may define matrices Ta with matrix
elements given below and satisfying the commutations relations in (6.1.22)
.Tb /ac D i fabc ; Œ Ta ; Tb D i fabd Td : (6.1.27)
These matrices provide a representation of the Lie algebra called the adjoint
representation and is generated by the structure constants themselves.
An algebra for which all the generators commute is called an abelian algebra and
otherwise it is called non-abelian.
A gauge vector field A .x/, referred to as a Yang-Mills field with corresponding
field strength G .x/, may be materialized and defined in terms of the connection
in (6.1.7) and the curvature in (6.1.15), respectively, by introducing, in the process,
a coupling parameter go
1
A .x/ D .x/; r D @ i go A .x/; (6.1.28)
go
1
G D R .x/ D @ A .x/@ A .x/ igo ŒA ; A : (6.1.29)
go
This, from a geometrical setting, provides a way of introducing vector fields, as

dynamical variables, that interact with charge and physical entities corresponding
to other internal degrees of freedom carried by a given field .x/, referred to as
a matter field, with go defining an overall coupling parameter between the fields.
The vector fields may themselves possess internal degrees of freedom themselves
and the geometrical description with an associated curvature in charge (integral
degrees of freedom) space produces naturally and in a simple way means to describe
interactions between these curious entities we call gauge fields, and all with the same
coupling parameter go .
For the subsequent treatment, we consider the parametrization # .x/ D go ƒ.x/ .
These vector fields together with the given field .x/ in question, have from (6.1.10)
and (6.1.17) the following transformations
h i i
.x/ ! V.x/.x/; A .x/ ! V.x/ A .x/ C @ V 1 .x/; (6.1.30)
go
G .x/ ! V.x/G .x/V 1 .x/; Tr G .x/G .x/ ! Tr G .x/G .x/: (6.1.31)
Of particular interest are infinitesimal transformations of the fields .x/, A .x/,

which are obtained from the above transformations to be
.x/ ! .x/ C i go •a .x/ ta .x/; (6.1.32)

A .x/ ! A .x/ C @ •a .x/ta C igo •a .x/ Œta ; A .x/: (6.1.33)
From the second term on the right-hand side of (6.1.33), we make use of the matrix
nature of A .x/ to write
A .x/ D Ac .x/ tc ; Ac .x/ ! Ac .x/ C @ •c .x/ go fcab •a .x/Ab .x/:
(6.1.34)
Here we see that in addition to a rather familiar change @ •c .x/, internal degrees
of freedom are associated with Aa .x/ as well. It also transforms according to the
adjoint representation.
The field strength G may be then rewritten as
G .x/ D Gc .x/ tc ; (6.1.35)

Gc .x/ D @ Ac .x/@ Ac .x/ C go fcab Aa .x/ Ab .x/: (6.1.36)
One may set up a so-called kinetic energy term for the set gauge vector fields,
and, due to the invariance property on the right-hand side of (6.1.31), we are bound
to consider an expression like
1 1
Tr G .x/G .x/ D Tr.ta tb / Ga .x/Gb .x/; (6.1.37)
2 2
up to a normalization factor, involving a direct interaction between the gauge vector
fields.
For the group U.1/ corresponding to phase transformations, describing, in
particular, electrodynamics for which the field strength G reduces to the field
strength tensor F , the coupling go is identified with the electronic charge eo , and
one is dealing with an abelian gauge group, with properties
F D @ A .x/ @ A .x/; A .x/ ! A .x/ C @ •.x/; (6.1.38)

r D @ i eo A .x/; .x/ ! .x/ C i eo •.x/.x/: (6.1.39)
The group SU.N/, is defined in terms of N N unitary and unimodular matrices,

i.e., of determinant 1. This implies that for the transformations with group elements
as given in (6.1.8), the corresponding matrices ta must be Hermitian to satisfy the
unitary property of the group element, and must be of zero trace so that the group
elements have determinant equal to one. The latter property is easily verified since
referring to (6.1.8)

det exp . i #a ta / D exp i#a Tr Œ ta D 1:
This involves of M D N 2 1 non-commuting generators ta represented by

N N Hermitian and traceless matrices. As shown at the end of this section,
one has .N 1/ Hermitian diagonal matrices in this set. The number .N 1/
of diagonal matrices is referred to as the rank of the group. The construction
of a set of generators for SU.N/ is straightforward and is spelled out below in
subsection 6.1.1. For the SU.3/ color group of quantum chromodynamics, the
quarks come in N D 3 different colors, and the theory involves N 2 1 D 8 gluons.
For the group SU.2/, the lowest representation is given by a doublet with
generators represented, as discussed at the end of this section by the three (N 2 1 D
22 1 D 3) Pauli matrices, as 2 2 Hermitian traceless matrices, divided by 2,

1 .x/
.x/ D ; t D ; Œ ta ; tb D i
abc tc ; D .1 ; 2 ; 3 /; (6.1.40)
2 .x/ 2
where
abc is totally anti-symmetric with
123 D C1. Here we have only one
diagonal matrix, that is the group is of rank one, it is given by t3 D diagŒ1=2; 1=2,
and a corresponding particle state may be labeled by its eigenvalues ˙1=2. The
adjoint representation is provided by the generators Ta , with matrix elements
(see (6.1.27)) .Tb /a c D i
abc , i.e.,
0 1 0 1 0 1
00 0 0 0 i 0 i 0
T1 D @ 0 0 i A ; T2 D @ 0 0 0 A ; T3 D @ i 0 0 A ; (6.1.41)
0 i 0 i 0 0 0 0 0
which also satisfy the commutation relations
ŒTa ; Tb D i
a b c Tc : (6.1.42)
The next lowest representation of SU.2/ is a triplet

0 1
1 .x/
.x/ D @ 2 .x/ A : (6.1.43)
3 .x/
A more convenient and equivalent representation of the above commutation rela-

tions by 3 3 Hermitian matrices in which T30 is diagonal and hence may explicitly
display the eigenvalues of such a multiplet is obtained by a similarity transformation
of these generators. That is, by multiplying (6.1.41) from the left by a matrix S and
from the right by its inverse S1 , (STa S1 D T 0a ;) giving rise to the generators
0 1 0 1 0 1
010 0 1 0 10 0
1 i
T 01 D p @1 0 1A ; T 02 D p @1 0 1A ; T 03 D @0 0 0 A ; (6.1.44)
2 010 2 0 1 0 0 0 1
displaying the eigenvalues .1; 0; 1/ of a triplet, where

0 1
1i 0p
S D@ 0 0 2A: (6.1.45)
1i 0
The lowest representation of SU.3/ is defined by a triplet as in (6.1.43), referred

to as the fundamental representation. SU.3/ is a group of rank 2 : .31 D 2/ that is,
two of its generators ta ; a D 1; : : : ; 8; .M D N 21 D 8), may be in diagonal form.
Generators are obtained directly from the analysis to be carried out at the end of the
section. They are 3 3 Hermitian traceless matrices and are given by ta D a =2,
where a are called the Gell-Mann matrices
0 1 0 1 0 1
0 1 0 0 i 0 1 0 0
1 D @ 1 0 0 A ; 2 D @ i 0 0 A ; 3 D @ 0 1 0A;
0 0 0 0 0 0 0 0 0
0 1 0 1 0 1
0 0 1 0 0 i 0 0 0
4 D @ 0 0 0 A ; 5 D @ 0 0 0 A ; 6 D @ 0 0 1 A;
1 0 0 i 0 0 0 1 0
0 1 0 1
0 0 0 10 0
1
7 D @ 0 0 i A ; 8 D p @ 0 1 0 A : (6.1.46)
0 i 0 3 0 0 2
The structure constants are easily evaluated and are given by
1
f123 D 1; f147 D f156 D f246 D f257 D f345 D f367 D ;
2
p
3
f458 D f678 D : (6.1.47)
2
Due to the Hermitian property of the matrices ta D a =2, with the a =2 defined
in (6.1.46), the complex conjugates are given by ta D ta> , and the complex conju-
gate of the unitary matrix .exp.igo a ta // becomes simply .exp.igo a .ta /> //.
Also as there is no matrix S such that S a S1 D > a ; a D 1; : : : ; 8, we have
two inequivalent representations. This is easily seen by noting that > 8 D 8 , and
the existence of such a matrix S would imply that 8 and 8 have the same
eigenvalues thus leading to a contradiction. This new transformation is defined for
the complex conjugate of the fields. One then introduces the complex conjugate
transformation as
.x/ ! .exp.igo a .ta /> // .x/; (6.1.48)
thus generating an inequivalent, that is new, representation. For SU.2/, we may

define the antisymmetric matrix S D .
ab / with
ab D
ab ;
12 D C1, with
the result that for the Pauli matrices, S S1 D > , where we note that
ab b
transforms the same way as a . That is, the complex conjugate provides an
equivalent description for SU.2/.
The above treatment will be very useful in setting up the formalism of gauge
theories in later parts.
We close this section by considering some additional notions related to Lie

algebras, as well as elaborate on the structure of the generators of SU.N/.
We recall from above, that two representations are equivalent, if the correspond-
ing generators ta , ta0 are related by a similarity transformation: S ta S1 D ta0 .
A subalgebra is said to be invariant if the commutator of any one of its elements
with any element of the full algebra yields an element of the subalgebra itself (or
may be equal to zero).
An algebra which has no non-trivial invariant subalgebra is called simple. [The
full algebra and the empty set are, by definition, trivial invariant subalgebras.]
A representation is said to be reducible if it is equivalent to one involving two
(or more) sets of generators fta1 g, fta2 0 g, and corresponding structure constants fa1b c ,
fa20 b 0 c 0 , such that
Œta1 ; tb2 0 D 0; Œ ta1 ; tb1 D ifa1b c tc1 ; Œ ta2 0 ; tb2 0 D ifa20 b 0 c 0 tc2 0 : (6.1.49)
The latter equivalent representation is then said to be the direct sum of the two
(or more) representations, with generators fta1 g, fta2 0 g which would act, in turn, in
different vector spaces. Otherwise a representation is said to be irreducible.
6.1.1 Generators of SU(N)
We need to generate a complete set of N 2 1 Hermitian traceless N N matrices.

We may start by constructing elementary matrices which have the single element 1
in every possible way skipping the diagonal elements. Thus we generate N 2 N D
N.N 1/ matrices, with necessarily zeros on their diagonal. From this we will
construct Hermitian, properly normalized, matrices. We then proceed to generate
N 1 diagonal (and hence commuting) Hermitian matrices thus completing the set
of generators.
Introduce the set of matrices:
E.a; b/ D1 at the ath row and bth column for a ¤ b;

and zero otherwise; a; b D 1; : : : ; N; (6.1.50)
thus generating N 2N matrices each with a single element 1 and zeros everywhere
else, including zeros for their diagonal elements. From these we will introduce
below properly normalized Hermitian matrices. We now generate N 1 diagonal
Hermitian traceless matrices as follows.
We define these N N matrices as

1
Lk D p diagŒ 1; 1; : : : ; 1; k; 0; ::; 0 ; k D 1; : : : ; .N 1/: (6.1.51)
2k.k C 1/
where the number 1 appears k number of times for each k. Clearly, for k D N 1
there are no zeros after the element k D N C1. We note that
1
Lj L k / Lj for j < k; Tr.L i Lj / D •ij i; j; k D 1; : : : ; .N 1/: (6.1.52)
2
From (6.1.50), we introduce the following two sets of N.N 1/=2 Hermitian
traceless matrices
1 1
D1 .a; b/ D .E.a; b/ C E.b; a//; D 2 .a; b/ D .E.a; b/ E.b; a//;
2 2i
(6.1.53)
1 a < b N, which we may denote by LN ; : : : ; L.N 2 1/ , thus completing the
set of generators.
It is easy to verify that the generators L1 ; : : : ; L.N 2 1/ satisfy the normalization
condition in (6.1.24). The number .N1/ of diagonal matrices is called the rank of
the group.
SU.2/ is of rank 1, that is, it has one diagonal matrix and L1 given above
coincides with the Pauli matrix 3 =2. Also L 2 ; L3 coincide with 1 =2; 2 =2.
SU.3/ is of rank 2, and L1 , L 2 coincide with t3 D 3 =2, t8 D 8 =2, where
the Gell-Mann matrices a are defined in (6.1.46), and L3 ; : : : ; L8 coincide with the
remaining Gell-Mann matrices divided by two.
6.2 Quantization of Non-Abelian Gauge Fields

in the Coulomb Gauge
The gauge fields in non-abelian gauge theories Aa , unlike the gauge field in abelian
gauge theories, the photon which has no charge, necessarily involves a direct self-
interaction due to a “charge” it carries, and the simplest Lagrangian density for the
gauge fields which is invariant under non-abelian gauge transformations reflects this
fact.4 A simple Lagrangian density for the gauge fields is given by
1
L .x/ D Ga .x/ G
a .x/; (6.2.1)
4
4
A review of the content of Sect. 6.1 will be beneficial for the reader.
6.2 Quantization of Non-Abelian Gauge Fields in the Coulomb Gauge 383
described in terms of the fields strengths Ga .x/ in (6.1.36), involving two
derivatives of the fields, as a direct generalization of the corresponding one in
electrodynamics, but also involves the interaction of the gauge field. The field
strength is given in detail in terms of the gauge fields by the relation in (6.1.36)
Gc D @ Ac @ Ac C go fabc Aa Ab : (6.2.2)
where fabc are totally anti-symmetric structure constants of the underlying group of
transformations, with generators ta satisfying (see (6.1.22), (6.1.24))
1
Œta ; tb D ifabc tc ; Tr.ta tb / D ıab : (6.2.3)
2
The non-abelian symmetry gauge transformations are implemented by operators
V.x/ D expŒ i go a .x/ ta , and upon setting Aa ta D A , Ga ta D G , we have
the following gauge transformation rules (see (6.1.30), (6.1.31)), for these fields
h i i
A .x/ ! V.x/ A .x/ C @ V 1 .x/; (6.2.4)
go
G .x/ ! V.x/G .x/V 1 .x/; Tr Œ G .x/G .x/ ! Tr Œ G .x/G .x/:
(6.2.5)
The last transformation rule for Tr Œ G G Ga Ga =2, shows the

gauge invariance of the Lagrangian density in (6.2.1) for pure gauge fields.
Upon defining

rab D ıab @ C go facb Ac ; (6.2.6)
with matrix elements just defined here, we have the basic relations
Œ r ; r cb D go fcab G

a ; rab rbc G
c D 0; (6.2.7)
as shown in Problems 6.2, 6.3, with the latter equality generalizing the equation
@ @ F D 0 in the abelian case, to the non-abelian one.
To develop the field theory dynamics, we couple the gauge fields Aa to external

sources Ka thus introducing the Lagrangian density
1
L D Ga G
a C Ka Aa : (6.2.8)
4
Now we must pick up a gauge to describe the dynamics. In the present section we
work in the Coulomb gauge. Covariant gauges will be considered in the following
section. The Coulomb gauge is defined by (i D 1; 2; 3)
@ i A ai .x/ D 0; (6.2.9)
which allows us to solve, say, A3a in terms of the other fields and their space
derivatives
A3a D @1
3 @ A a ; D 1; 2; (6.2.10)
and due to the dependence of A3a on the other fields, the variation of the fields takes
the form5

•A a i D ıi ıi 3 @1
3 @

•A a ; (6.2.11)
with a sum over D 1; 2, in the latter two equations understood.

The field equations then follow to be
j0
rabj Gb D rac Gc 0 D K 0a ; (6.2.12)
j
rab Gb D @1 3 3
3 @ rac Gc C Ka Ka ;
j j
(6.2.13)
where we note that for j D 3 the latter equation is trivially true.

The above two equations may be combined to

rab Gb D j @ j @1 3 3
3 rac Gc C Ka Ka : (6.2.14)
By applying the operator rda to this equation, and using the second equation
in (6.2.7), we obtain

rdaj @ j @1 3 3
3 rac Gc C Ka D rda Ka ; rdaj @ j ıda r 2 C go fdca Acj @j :
(6.2.15)
3
This suggests to introduce the Green function Dab , to solve for @1 3
3 .rac Gc C Ka /
6
in (6.2.14), satisfying the equation

ıda r 2 C go fdca Acj .x/@j Dab .x; x 0 / D ı .4/ .x; x 0 / ıdb : (6.2.16)
Thus from (6.2.15), (6.2.16), we may rewrite our field equation (6.2.14) finally
as

rab Gb D ıac ı j @ j Dab rbc Kc : (6.2.17)
The canonical conjugate momenta of Aa , . D 1; 2/, A0a are readily worked

to be
a D Ga 0 @1 30
3 @ Ga ; a0 D 0: (6.2.18)
5
Compare this with the Coulomb gauge problem in QED in Sect. 5.14.
6
Note that since this equation involves no time derivative, Dab .x; x 0 / involves a ı.x 0 ; x 0 0 / factor.
Therefore A0a is a dependent field. Also the first expression a gives zero if is
replaced by 3 (a3 D 0), as it should for a dependent field. Accordingly we may
combine these components and write
a D Ga 0 j @1 j 30

3 @ Ga ; (6.2.19)
as is easily verified.
0
As we will see below, the expression for Ga is necessary for finding the solution
of the non-abelian gauge theory. To this end, we apply the operator rba to the
above equation to obtain
rb a a D rba Ga 0 rbaj @1 j 30 0 1 j 3 0

3 @ Ga D K b rb aj @3 @ Ga ; (6.2.20)
where we have used (6.2.12). Therefore

rb aj @1 j 30 0
3 @ Ga D rba a C K b : (6.2.21)
From Eq. (6.2.16) of the Green function Dab , we may then solve for j @1 j 30
3 @ Ga
in (6.2.21), to finally obtain from (6.2.19)

Ga 0 D a j @ j Dab rbc c C K 0b ; (6.2.22)
expressed in particular, in terms of canonical conjugate momenta and their space

derivatives (no time derivatives). Since A0a is a dependent field so is Ga 0 . The
functional derivative of the latter with respect to the external source Kc is then
given by
• •
Ga 0 .x/ D ı 0 j @ j Dac .x; x 0 /; G ka l .x/ D 0: (6.2.23)
•Kc .x 0 / •Kc .x 0 /
by keeping the independent fields A1 ; A2 , and their canonical conjugate momenta,

fixed, where we recall that A3 is expressed in terms of A1 ; A2 , and recall that
0 D 0; 3 D 0.
From the explicit expression of Gc in (6.2.2), the partial derivative of the
Lagrangian density in (6.2.8) with respect to the coupling parameter go , is given by
@ 1 1
L D fabc G k0
a Ab Ac D fabc G a Ab k Ac 0 a Ab k Acl :
fabc G kl (6.2.24)
@go 2 2
Now we use the basic functional differentiation rule, developed in Sect. 4.6, with
a summary provided at the end of that section, that for an operator O.x/
• ˝ ˇ ˇ ˛
.i/ h 0C jO.x/ j 0 i D 0C ˇ Aa .x 0 /O.x/ C ˇ 0
•Ka .x 0 /
˝ ˇ • ˇ ˛
i 0C ˇ O.x/ ˇ 0 ; (6.2.25)
•Ka .x 0 /
where .: : :/C , denotes the time-ordered product, and the functional derivative of
O.x/, in the second term on the right-hand side, is taken, as in (6.2.23), by keeping
the (independent) fields and their canonical momenta fixed. Here recall that A3a is
expressed in terms of the independent fields and hence it is automatically kept fixed,
and it involves no time derivatives.

Upon setting i•=•Ka b Aa , an immediate application of (6.2.25), (6.2.23),
leads to the following two key equations:

h 0C j Ac 0 Ab k G ka 0 /C j 0 i D b
Ac 0 h 0C j Ab k G ka 0 /C j 0 i
•
C i h 0 C j Ab k G k 0 /C j 0 i
•Kc0 a
Db
Ac 0b G ka 0 h 0C j 0 i i b
Ab k b Ab k @ k b
Dac .x; x/h 0C j 0 i; (6.2.26)
h 0C jG ak l Ab k Ac l j 0 i D b
G ka lb
Ab kb
Acl h 0C j 0 i: (6.2.27)
Hence from (6.2.24)
@ h @ i
h 0C j L .x/ j 0 i D b .x/ ifabcb
L Ab k @ k b
Dca .x; x/ h 0C j 0 i:
@go @go
(6.2.28)
where we have used the anti-symmetry of fabc to rewrite the second expression
within the square brackets. On the other hand, the quantum dynamical principle
gives
Z
@ @
h 0C j 0 i D i .dx/h 0C j L .x/ j 0 i; (6.2.29)
@go @go
and hence from (6.2.28)

Z h @ i
@ b .x/ C fabcb
h 0C j 0 i D .dx/ i L Ab k @k b
Dca .x; x/ h 0C j 0 i:
@go @go
(6.2.30)
We use the matrix notation as obtained from (6.2.16)

D ˇ ˇ E
ˇ 1 ˇ 0
Dab .x; x 0 / D xˇ ˇx ; (6.2.31)
r 2 C go . fA k /@ k ab
where . f b
Ak /da D fdcab
Ack , the notation
Z
Tr Œ E D .dx/Eaa .x; x/; (6.2.32)
to rewrite the second term within the square brackets in (6.2.30), integrated over x,
as

1
Tr . f b
Ak /@k 2 : (6.2.33)
Œr C go . f b
Al /@l
An elementary integration of the latter over go from 0 to go gives the multiplicative

factor
h 1 i
det ıab C go facb 2 b
Ack @k ; (6.2.34)
r
to h 0C j 0 i in (6.2.30). All told the following expression emerges for h 0C j 0 i:

h i Z
2 b b I .x/h 0C j 0 i0 ;
h 0C j 0 i D det ıab r C go facb Ack @ exp Œ i .dx/L
k
(6.2.35)
up to an unimportant multiplicative constant factor, where in detail
g2
b I D go fabc @b
L Aa b
Aa @b Ab b
Ac o fabc b

Ab b
Ac fade b
Adb
Ae ; (6.2.36)
2 4
and from (5.14.16), (5.15.17),
hi Z i
h 0C j 0 i0 D exp .dx/.dx 0 /Ka .x/Dab
C
.x x 0 /Kb .x 0 / ; (6.2.37)
2
Dab
C
D ıab D
C
,
1 Qi Qj 1
D00
C
D ; DijC .Q/ D ıij 2 ; D0Ci .Q/ D 0; (6.2.38)
Q2 Q Q2 i
in the momentum description. The determinant factor in (6.2.35) is referred to as a

Faddeev-Popov determinant.7
The inclusion of matter field is straightforward, and with the field as a column
vector,8 with corresponding gauge transformation rules: ! V , ! V 1 ,
one may write the Lagrangian density as
1 1 @ @
L D Ga G
a C Mo
4 2 i i
1
C go Œ ; A C C C Aa Ka ; (6.2.39)
2
where Mo is a mass matrix for the Fermion field. Such generalizations will be
considered as we go along in coming sections.9
6.3 Functional Fourier Transform and Transition

to Covariant Gauges; BRS Transformations
and Renormalization of Gauge Theories
In the present section, dealing with the vacuum-to-vacuum transition amplitude, we

develop a functional Fourier transform, working in the Coulomb gauge which then
allows us to make a transition to covariant gauge. The final subsection deals with
the renormalizability of the non-abelian gauge theories set up in the last section via
an elegant approach carried out via so-called BRS transformations.10
6.3.1 Functional Fourier Transform and The Coulomb Gauge
The path integral expression for h 0C j 0 i, in the Coulomb gauge in (6.2.35) for

non-abelian gauge fields, as a functional Fourier transform Ka ! Aa , with the
latter now a classical field, may be then inferred from (5.14.41), in the absence of
7
Faddeev and Popov [37]. It is interesting to note that the Faddeev-Popov factor also appears in the
functional differential equation satisfied by h 0C j 0 i, see Manoukian [86].
8
See, for example, (6.1.43).
9
This section is based on Manoukian [84].
10
BRS stands for Becchi, Rouet and Stora: [15].
6.3 Functional Fourier Transform and Transition to Covariant Gauges; BRS. . . 389
matter, and may be denoted by

Z

h 0C j 0 iC D ˘ax DAa .x/ı @k Aka .x/ MC ŒA
Z Z

exp Œ i .dx/Wcl .x/ exp Œ i .dx/Kb .x/Ab .x/ ; (6.3.1)
up to an unimportant multiplicative constant factor, with C in h 0C j 0 iC

emphasizing that this expression corresponds to the Coulomb gauge, where 11
h i
MC Œ A D det @k ıab @k C go facb Ac k ; (6.3.2)
1
Wcl .x/ D G .x/Ga .x/; G
a D @ Aa @ Aa C go fabc Ab Ac : (6.3.3)
4 a
now defined in terms of classical fields. We here see both the Coulomb gauge
constraint via the delta functional ˘ax ı.@k Aka .x// explicitly, as well as a Faddeev-
Popov determinant in the Coulomb gauge displayed in the functional integrand
in (6.3.1). The expression for h 0C j 0 i in (6.2.35), (6.2.36), (6.2.37) is a

functional of Ka , and we here recognize the path-integralRexpression in (6.3.1) as

a functional Fourier integral with the coefficient of expŒ i .dx/Kb .x/Ab .x/ as a

functional of Aa .
To find the corresponding expression to the one in (6.3.1) in other
gauges we proceed
as follows. To this end, let us concentrate on the factor
˘ax ı @k Aka .x/ MC ŒA , in the path integrand in (6.3.1), which originated from
working in the Coulomb gauge. Consider all possible gauge transformations on

the field Aa .x/ ! .Aa .x//. / , via the group g of transformations: V.x/ D
expŒ i go a .x/ta for all . a .x/; a/, with a .x/ real. In turn consider the sum of

all these gauge transformations as applied to a given function f .Aa .x//, which we
may express as
X
f .Aa .x//. / : (6.3.4)
all . b .x/;b/
Taking a product over all spacetime points x, and all the values taken by the group
indices a, we have
X Z
˘ax f .Aa .x// ! ˘x dg.x/ ˘ax f .Aa .x//.g/ ;
. /
(6.3.5)
all . b .x/;b/
where the expression on the right-hand side stands for the formal mathematical
expression for the left-hand side.
11
Note that in the determinant @k Akc D .@k Akc / C Akc @k .
Let us use the notation

Z
˘x dg.x/ ˘ax f .Aa .x//.g/ D Bf ŒA : (6.3.6)
Since the left-hand sides includes all possible transsformation defined by the group
0
g for all . b .x/; b/, a further gauge transformation .Aa .x//.g/ ! ..Aa .x//.g/ /.g /
necessarily leaves the sum (integral) invariant as it is already included in the sum.
That is, Bf ŒA is gauge invariant, which we record here as
0
Bf Œ.A /.g / D Bf ŒA : (6.3.7)
In view of application to the path integral in (6.3.1), consider the product

Z
˘ax ı.@k Aka .x// ˘x dg.x/ ˘ax ı @k Aka .x//.g/ : (6.3.8)
Since the expression on the left-hand side of the integral is the same as the one
in the integrand corresponding to the identity transformation, i.e., for .x/ ! 0,
it suffices to consider the integral for in this limit. From Problem 6.1, for
infinitesimal , we have

Aa ! Aa C rab b ; rab D ıab @ C go facb Ac ; (6.3.9)
hence for @k .Aka / D 0, on account of the delta function ı.@k Aka .x// on the left-hand
side of the integral in (6.3.8) for all group indices a, and all spacetime points x,
we may write
@k .Aka /. / D @k Œ.ıab @k C go facb Akc /b : (6.3.10)
This allows us to rewrite (6.3.8) as

Z
˘ax ı.@k Aka .x// ˘x dg.x/ ˘ax ı @k Aka .x//.g/
Z
D ˘ax ı.@k Aka .x// ˘e x de .x/ ˘ax ı @k Œ.ıab @k C go facb Akc /b
Z
1
D ˘ax ı.@k Aka .x// ˘e x de .x/ı.e .x//
MC Œ A
1
D ˘ax ı.@k Aka .x// ; (6.3.11)
MC Œ A
where MC Œ A is the determinant defined in (6.3.2).

On account of the multiplicative factor ˘ax ı.@k Aka .x// in (6.3.8), we may write
from (6.3.6) and (6.3.11),
h i 1
˘ax ı.@k Aka .x// det @k .ıab @k C go facb Akc / D ˘ax ı.@k Aka .x// ; (6.3.12)
Bf ŒA
f .Aa .x// D ı.@k Aka .x//; (6.3.13)
and Bf ŒA is gauge invariant.

Now let us choose another function f which we conveniently denote by

fQ .Aa .x// D ı @ Aa .x/ 'a .x/ ; (6.3.14)
where the 'a .x/ are some arbitrary functions, and is an arbitrary parameter. Then
almost an identical analysis leading to (6.3.12), gives
h i
˘ax ı @ Aa .x/ 'a .x/ det @ .ıab @ C go facb Ac /
1
D ˘ax ı @ Aa .x/ 'a .x/ ; (6.3.15)
BQf ŒA
where BQf ŒA is gauge invariant, defined by the general expression in (6.3.6) with
f now replaced by fQ . In particular we note the trivial identity
Z
1
˘x dg.x/ ˘ax ı @ .Aa .x//.g/ 'a .x/ D 1: (6.3.16)
BQf ŒA
Upon multiplying the integrand in the path integral in (6.3.1) by the above
expression, i.e., by one, the path integral, by using in the process (6.3.12), may
be rewritten as
Z Z
h 0C j 0 iC D ˘ax DAa .x/ ˘x dg.x/ ˘ax ı @ .Aa .x//.g/ 'a .x/
Z Z
1
ı @k Aka .x/ exp Œ i .dx/Wcl .x/ exp Œ i .dx/Kb .x/Ab .x/ :
BQf ŒA Bf ŒA
(6.3.17)
Using the invariance of the measure ˘ax DAak .x/ under gauge transforma-
tions, we carry out an inverse gauge transformation of the field variables:
1 /
Aa .x/ ! .Aa .x//.g to obtain for the above path integral
Z
h 0C j 0 iC D ˘ax DAa .x/ ı @ Aa .x/ 'a .x/
h i Z
det @ .ıab @ C go facb Ac / exp Œ i .dx/Wcl .x/ H ŒK; A; (6.3.18)
where
Z
1 .g1 /

H ŒK; A D ˘x dg.x/ ˘ ax ı @k .A k
.x//
Bf ŒA a
Z
1
exp Œ i .dx/Kb .x/ .Ab .x//.g / ; (6.3.19)
and we have used the gauge invariance of Wcl .x/, Bf ŒA , BQf ŒA , and finally
used (6.3.15).

In the absence of the external sources Kb .x/, H Œ0; A is simply one. We may
conveniently rewrite the expression in (6.3.18) simply as
h Z
•i
h 0C j 0 iC D H K; i ˘ax DAa .x/ı @ Aa .x/ 'a .x/
•J
h i Z Z ˇ
ˇ
det @ .ıab @ C go facb Ac / exp Œ i .dx/Wcl .x/ exp Œ i .dx/Jb .x/ Ab .x/ ˇ :
JD0
(6.3.20)
Since h 0C j 0 iC on the left-hand of the above equation is independent of the

functions 'a .x/, we may multiply the above by any functional differential operator
h i • i ˇ
˘ax exp N ; N .w/ˇwD0 D 0: (6.3.21)
2 •'a .x/
In particular for N Œwa .x/ D w2a .x/, (6.3.20) becomes (see Problem 6.4)
h Z
•i
h 0C j 0 iC D H K; i ˘ax DAa .x/
•J
h i Z Z ˇ
e ˇ
det @ .ıab @ C go facb Ac / exp Œ i .dx/W cl .x/ exp Œ i .dx/Jb .x/ Ab .x/ ˇ ;
JD0
(6.3.22)
up to an unimportant multiplicative constant, where

Z Z

1 1
ecl .x/ D
.dx/W .dx/ G
a .x/Ga .x/ @ Aa @ Aa
4 2
(6.3.23)
Z h 1 i Z
1
D .dx/Aa .x/ ./ 1 @ @ Aa .x/ C .dx/LI cl .x/;
2
(6.3.24)
go g2
LI cl .x/ D fabc @ Aa @ Aa Ab Ac o fabc Ab Ac fade Ad Ae : (6.3.25)
2 4
The generating functional h 0C j 0 iC , in the Coulomb gauge, may be denoted

by FC Œ K , and from (6.3.22), the generating functional FŒ J; , in covariant
gauges specified by the parameter , by comparison, may be then explicitly given
by
Z Z

FŒ J; D ˘ax DAa .x/ det @ rab exp i .dx/ Wecl .x/ C Jb .x/ Ab .x/ ;
(6.3.26)
and (6.3.22) relates the theory formulated in the Coulomb gauge with those in
covariant gauges specified by the parameter , and is discussed further in the next
subsection.
It is evident from the free part in the Lagrangian density in (6.3.24) that the
gauge field free propagator in the momentum description, in the covariant gauges,
is given by
h Q Q i 1

Dab .Q/ D ıab .1 / 2 2
; Q Q Dab .Q/ D ıab :
Q .Q i
/
(6.3.27)

Because of the presence of the Faddeev-Popov determinant det @ rab , in the

quantization procedure, in covariant gauges, derived in (6.3.26), we may use the
functional integral
Z Z
det M D D! D! expŒ i .dx/ !.x/M!.x/; (6.3.28)
in (2.6.19), up to an unimportant multiplicative constant, corresponding to Grass-

mann variables, to write
Z Z

det Œ@ rab D ˘ax D! a .x/ D!a .x/ expŒ i .dx/ ! a .x/@ rab !b .x/ ; (6.3.29)
generating a spin 0 field in the theory obeying Fermi-Dirac statistics. Because of this
unusual property of a spin 0 field, it is referred to as a ghost. The mere fact that they
emerge here is that they cancel out those contributions in the theory which would,
otherwise, destroy gauge invariance. That is, in the present quantization procedure,
because of these ghost fields a gauge invariant theory results.12 As a matter of fact
we will explicitly see in studying the gluon propagator in the next section how these
ghost fields are responsible for restoring gauge invariance of the propagator. The
ghost fields are not observable.
Now we are ready to write down the explicit effective Lagrangian density of the
theory and work in covariant gauges, with the bare propagator of the gluon field
given in (6.3.27):
1 1
Leff .x/ D G .x/Ga .x/ @ Aa @ Aa C ! a .x/@ rab !b .x/; (6.3.30)
4 a 2

where rab is defined in (6.3.9). The vacuum-to-vacuum transition amplitude, as the

generating functional, depending on the external source Jb , in covariant gauges,
specified by the parameter , is from (6.3.26), (6.3.29), (6.3.30), written in a
convenient form, is given by
Z Z

h 0C j 0 i D ˘ DA D! D! exp i .dx/ Leff .x/ C Jb .x/ Ab .x/ :
(6.3.31)

It is worth recording how the ghost contribution ! a .x/@ rab !b .x/ arises in the
Lagrangian density in (6.3.30). It is obtained in the following manner, by varying,

in the process, the gauge constraint functional Fa D @ Aa 'a , in (6.3.14), in
response to a gauge transformation as given in (6.3.9), leading to
• •

Fa ! @ .Aa C rab b 'a / ! ! a Fa !b D ! a
Œ@ .rac c / !b ;
•b •b
(6.3.32)
with the right-hand side coinciding with the ghost contribution. The transformation
law from the Coulomb to covariant gauges of the theory is discussed further in the
following subsection. The final subsection deals with the renormalizability of the
theory.
12
This also means that one may choose a specific gauge to make the ghost fields disappear. But
there is no need to go into it here. Feynman, in his monumental work on the quantization problem
of the gravitational field in 1963 [40], realized the presence of such ghost fields in his non-abelian
gauge theory.
6.3.2 Trasformation Law from the Coulomb Gauge

to Covariant Gauges in Non-abelian Gauge Field
Theories
The transformation law from the Coulomb gauge to the covariant is expressed by
the relation
h • i ˇ
ˇ
FC Œ K D H K; i FŒ J; ˇ : (6.3.33)
•J JD0
where FC Œ K and FŒ J; denote the generating functionals in the Coulomb and

covariant gauges, specified by the parameter
h i, respectively, in the presence of the
external sources K and J, and H K; i •J• is defined in (6.3.19). What the delta
1
function ı @k .Aka /.g / in (6.3.19) is telling us is that we have to find .Aa /. / , as
a function of Ab , for which @k .Aka /. / D 0. This may be achieved by expanding

.Aa /. / in powers of Ab as shown in Problem 6.5. Using (6.3.6), i.e.,
Z
1 1
˘x dg.x/ ˘ax ı @k .Aka .x//.g / D 1; (6.3.34)
Bf ŒA
we obtain from Problem 6.5

Z h i
@ @k
H ŒK; A D exp i .dx/Ka .x/ 2 k Aa .x/ C O.A2 / ; (6.3.35)
r
h Z
•i h @ @ k k • • 2 i
H K; i D exp i .dx/Ka .x/ .i/ C O ;
•J r2 •Ja .x/ •Ja
(6.3.36)
which should be compared with the corresponding expression in QED in (5.15.10),

(5.15.11), (5.15.12), and is admittedly more complicated.
6.3.3 BRS Trasformations and Renormalization

of Non-abelian Gauge Field Theories
Due to the particular covariant gauge chosen in the theory, described by the
expression in (6.3.31) specified by a parameter , the effective Lagrangian density
Leff given in (6.3.30) is not gauge invariant. There exists, however, residual
transformations, referred to as the BRS transformations,13 for which Leff , as well

as the functional integration measure in (6.3.31), are invariant, given by

•Aa .x/ D • rab ! b .x/; (6.3.37)
go
•!a .x/ D • fabc !b .x/ !c .x/; (6.3.38)
2
1
•! a .x/ D • @ Aa .x/; (6.3.39)

as is readily verified, where • is an infinitesimal constant parameter which anti-

commutes with !a , ! a , while it commutes with Ac . The ghost fields !a and ! a
are treated independently. We couple the ghost fields !a , ! a to external sources a ,
and a , respectively. It is also convenient to introduce external sources Qa and La

coupled to rab !b and .go =2/fabc !a !b appearing in (6.3.37), (6.3.38) as follows
go
Qa rab !b ; L a fabc ! b !c ; (6.3.40)
2
where it may be verified that the coefficients of the latter two external sources are
invariant under the BRS transformations (6.3.37), (6.3.38), i.e.,

•.rab b / D 0; •. fabc ! b !c / D 0: (6.3.41)
The total Lagrangian density of the theory involving external sources may be
now written as
go
Ltot D Leff C Ja Aa C a !a C ! a a C Qa rab !b L a fabc ! b !c ;
2
(6.3.42)
where the sources a , a , Qa are Grassmann-like sources. Thus under the BSR
transformations in (6.3.37), (6.3.38), (6.3.39), we have
h go 1 i

• Ltot D Ja rab !b .x/ C a fabc !b .x/ !c .x/ C a @ Aa •: (6.3.43)
2
Hence upon making infinitesimal changes of the functional integration variables in
the generating functional
Z Z

h 0C j 0 i D ˘ DA D! D! exp i .dx/Ltot .x/ exp i WŒJ; ; ; Q; LI go ; ;
(6.3.44)
13
BRS stands for Becchi, Rouet and Stora: [15].
introduced in (6.3.37), (6.3.38), (6.3.39), we obtain14

Z h • • 1 • i
.dx/ Ja C a a @ WŒJ; ; ; Q; LI go ; D 0:
•Qa •La •Ja
(6.3.45)
At this stage, we may establish the transversality of the polarization tensor

associated with the vector particle in analogy to the polarization tensor of the photon
in QED in (5.16.92), (5.16.96). To this end, we note that the field equation for !b
is given by

@ rab !b D a ; (6.3.46)
which upon taking its vacuum expectation h 0C j Œ : j 0 i value and dividing by

h 0C j 0 i, gives the functional differential equation
•
@ W D a : (6.3.47)
•Qa
Hence upon taking the functional derivatives: •=•a .x 0 /@ .•=•Jb .y// of (6.3.45),
y
and setting a D 0, Ja D 0 give

h • • 1 0 • •
@y C @x @y
•a .x 0 / •Qb .y/ •Ja .x 0 / •Jb .y/
Z
1 • • • i ˇˇ
.dx/ c .x/ 0
@x @y Wˇ D 0: (6.3.48)
•a .x / •Jc .x/ •Jb .y/ JD0
Upon using (6.3.47), then setting a D 0, Qa D 0, and using the definition of

propagator
the ˇ field, here in a convenient notation, Dab .x y/ D
of the vector
ˇ
•=•Ja .x/ •=•Jb .y/ W J D 0 , we obtain

@x @y D ab .x y/ D ıab ı .4/ .x y/: (6.3.49)
Upon Fourier transform this gives

k k Dab .k/ D ıab ; (6.3.50)
from which the transversality of the corresponding polarization tensor, as in QED

in (5.16.92), (5.16.96), easily follows.15 Again, as in QED, elementary power
14
This identity is referred to as a Slavnov-Taylor identity: Slavnov [110] and Taylor [117].
15
This transversality property is explicitly established and worked out, e.g., to lowest order, with
dimensional regularization, in (6.6.17), in the presence of matter in (6.6.18).
counting, in particular, shows that the ultraviolet divergences of the theory are only
of logarithmic types.16
We introduce a Legendre transform as in (5.20.11), and recast the entire theory
in terms of the effective action defined by
Z

˝ŒA; !; !; Q; LI go ; D WŒJ; ; ; Q; LI go ; .dx/ Ja Aa C a !a C ! a a ;
(6.3.51)
which takes into account all radiative corrections, generating full propagators and
Green functions of the theory. We note that
•˝ •˝ •˝ •˝
D Ja .x/; D a .x/; D a .x/ D @ ;
•Aa .x/ •!a .x/ • ! a .x/ •Qa .x/
(6.3.52)
where in writing the last equality we have used (6.3.47), which gives the clas-
sical field equation for !a .x/ corresponding to (6.3.46). Accordingly, we may
rewrite (6.3.45) as
Z h i
•˝ •˝ •˝ •˝ 1 •˝
.dx/ C C @ @ Aa .x/ D 0:
•Aa .x/ •Qa .x/ •!a .x/ •La .x/ •Qa .x/
(6.3.53)
The first and the last terms within the square brackets may be combined, upon partial
integration, giving rise to the integral corresponding to those two terms
Z
•˝ h •˝ 1 i
.dx/ @ @ Aa .x/ : (6.3.54)
•Qa .x/ •Aa .x/
This suggests to define an effective action by removing the gauge fixing term
from ˝ as follows
Z
1
ŒA; !; !; Q; LI go ; D ˝ŒA; !; !; Q; LI go ; C .dx/ @ Aa @ Aa :
2
(6.3.55)
In terms of , Eq. (6.3.53) now simply becomes replaced by

Z h • • • • i
.dx/ C D 0: (6.3.56)
•Aa .x/ •Qa .x/ •!a .x/ •La .x/
16
See also below at the end of this subsection.
We define the following renormalized (scaled) fields and parameters through the
equations
p q q
Aa D Z3 Aren a ; !a D ZQ 3 !ren a ; ! a D ZQ 3 ! ren a ;
p q
La D Z3 Lren a ; Qa D ZQ 3 Qren a ; go D Z gren ; D Z3 ren ; (6.3.57)
We note, in particular, that for the gauge fixing term in (6.3.55),
1 1
@ Aa @ Aa D @ Aren a @ Aren a : (6.3.58)
2 2ren
The above allows us to simply rewrite
ŒA; !; !; Q; LI go ;
p q q q p
Œ Z3 Aren ; ZQ 3 ! ren ; ZQ 3 ; !ren ; ZQ 3 Qren ; Z3 Lren I Z gren ; Z3 ren ; (6.3.59)
as expressed in terms of renormalized components. Since

q q
Z3 ZQ 3 Aren a Qren a D Aa Qa ; Z3 ZQ 3 Lren a !ren a D La !a ; (6.3.60)
we may equivalently replace (6.3.56) by

Z h i
• • • •
.dx/ C D 0; (6.3.61)
•Aren a .x/ •Qren a .x/ •!ren a .x/ •Lren a .x/
From the last equation in (6.3.52) we may supplement this equation by
• •
@ D 0: (6.3.62)
•Qren a .x/ •! ren a .x/
By using the notation

Z h •1 •2 •1 •2 i
1 2 D .dx/ C ; (6.3.63)
•Aren a .x/ •Qren a .x/ •!ren a .x/ •Lren a .x/
the identity in (6.3.61) may be then denoted by
D 0: (6.3.64)
Working with the renormalized (scaled) fields and parameters, we carry out a
loop expansion of (in powers of „) expressed as
X
D .N/ : (6.3.65)
N0
In particular, for N D 0, involving no loops, .0/ is by construction just the

classical action written in terms of renormalized (scaled) fields and the renormalized
parameter gren , not involving the gauge fixing term, and is explicitly given by
.0/ D IŒAren a ; ! ren a ; !ren a ; Qren a ; Lren a I gren

Z h 1 i
gren
D .dx/ Gren a G C.Qren a C! @
ren a /r !
renab ren b L f ! !
ren a abc ren b ren c :
4 ren a
2
(6.3.66)
Up to N th order, (6.3.64) reads
X
N
.n/ .Nn/ D 0: (6.3.67)
nD0
To zeroth order, .0/ is given in (6.3.66) and is obviously finite since it involves no
loop integrations. As an induction hypothesis, working with the renormalized fields
and parameters, then suppose that all the infinities in (6.3.67) up to the loop order
.N 1/ have been cancelled out by appropriate choices of renormalization constants
up to that order, and (6.3.67) holds true with the .n/ , for n D 1; : : : ; .N 1/, in
it replaced by their finite parts. Hence the infinities
ˇ in (6.3.67) can come only from
.N/ˇ
the n D N; n D 0 terms. We denote by ˇ the divergent part of .N/ , i.e.,
ˇ div
ˇ
.N/ .N/ˇ denotes its finite part, where we note, by the induction hypothesis,
div
that divergences in subdiagrams of .N/ are removed up to .N 1/ loops. If
ˇ ˇ
.0/
ˇ
.N/ ˇ
ˇ
.N/ ˇ .0/
• .N/ ˇdiv • .N/ ˇdiv
C D 0; @ D 0;
div div •Qren a •! ren a
(6.3.68)
then (6.3.67) will be satisfied by the finite parts of all the .n/ , for n D 1; : : : ; N,
and all the infinities are cancelled out to the order of N loops as well.
From (6.3.63), the first equation above may be conveniently rewritten as
ˇ
ˇ
O .N/ ˇ D 0; (6.3.69)
div
where17
• .0/ • • .0/ • • .0/ • • .0/ •

OD C C C ;
•Qren a •Aren a •Aren a •Qren a •Lren a •!ren a •!ren a •Lren a
(6.3.70)
and .0/ is given in (6.3.66).

At this stage, we consider the N loops contribution to the changes from the
classical action IŒAren ; ! ren ; ! ren ; Qren ; Lren I gren with the latter written in terms of
renormalized fields and renormalized coupling gren , with changes of ˇthe renormal-
ization constants as discussed below. To this end, if we let .Z3 /1=2 ˇN represents
.Z3 /1=2 determined up to N loops, then we have the following expressions for the
exact N loops contributions:
p p ˇ p ˇ q q ˇ q ˇ ˇ ˇ
• Z3 D Z3 ˇN Z3 ˇN1 ; • ZQ 3 D ZQ 3 ˇN ZQ 3 ˇN1 ; •Z D Z ˇNZ ˇN1 :
(6.3.71)
From the chain rule
p • p h •Aren a • •Lren a • i
• Z3 p D • Z3 p C p
• Z3 • Z3 •Aren a • Z3 •Lren a
p h
• Z3 • • i
D p Aren a C Lren a ; (6.3.72)
Z3 •Aren a •L ren a
p p
p process, that Aren a D Aa = Z 3 , Lren a D La = Z 3 . By
where we have used, in the
a similar application to ZQ 3 and Z, the following expression emerges for the N
loops contribution:
p h
• • i
•IŒAren ; ! ren ; ! ren ; Qren ; Lren I gren .N/ D • Z 3 Aren a C Lren a
•Aren a •Lren a
q h
• • • i @
C • ZQ 3 ! ren a C !ren a C Qren a C •Z gren
•! ren a •!ren a •Qren a @gren
IŒAren ; ! ren ; !ren ; Qren ; Lren I gren ; (6.3.73)
17
This operator was conveniently used, for example, in Lee [77] and Itzykson and Zuber [67],
where it is understood that not only one sums over the repeated index in (6.3.70), but also one
integrates over the coincident arguments x of the functional differentiation variables in it.
p p
where • Z 3 , • ZQ 3 , •Z are given in (6.3.71)) involving each N loops. Accord- p
p p p
ingly, we have replaced • Z 3 = Z 3 by • Z 3 , and have done similarly for ZQ 3
and Z, where recall that these renormalization constants are all equal to one for zero
loops.
By commuting the operator Œ@ .•=•Qren a / •=.•! ren a / with the operator f : g
above in (6.3.73), and applying it directly to IŒAren ; ! ren ; !ren ; Qren ; Lren I gren , we
verify that the second equation in (6.3.68) is automatically satisfied on account of the
combination of the structure .Qren aC! ren a @ / appearing in I as given in (6.3.66). ˇ
ˇ
On the other hand it is straightforward, though quite tedious, to verify that .N/ ˇ
div
identified with the above expression, i.e.,
ˇ
ˇ
.N/ ˇ D •IŒAren ; ! ren ; ! ren ; Qren ; LI gren .N/ ; (6.3.74)
div
defined on the right-hand side (6.3.73), satisfies (6.3.69), by direct application of

the functional differential operator O to theˇ right-hand side of (6.3.73). That is,
ˇ
with (6.3.69) now satisfied, i.e., with .N/ ˇ determined and hence subtracted
div
from .N/ , we infer the finiteness of the theory up to N loops as well.18
The transversality condition of the polarization tensor, as a consequence
of (6.3.50), reduces the (naïve) degree of divergence of the graph with two
external vector lines by 2, thus leading to a degree of divergence equal to zero.
In particular, the derivative in one of the vertices involving an external ghost line
can be transposed to !, thus reducing an associated graph involving such a vertex
by one.
We will see that when fermions are included, one may encounter anomalies
of the type shown in Sect. 3.10. Such possible anomalies will be investigated in
Sect. 6.15.2, and their absence, as will be clear, are important for the renormaliz-
ability of the standard model.
Graphs of some one loop contributions to the theory of Yang-Mills fields
in the presence of matter (spinor) fields with utmost four external lines are
given in Figs. 6.4 and 6.5. For explicit one loop computations, including that of
renormalization constants, in the presence of matter (spinor) fields, with a regular-
ization procedure to justify intermediate steps in the analysis, see Sects. 6.6 and 6.7.
18
For additional details and other aspects, such as the inclusion of matter fields, see Lee [77],
Itzykson and Zuber [67], Weinberg [121], and Joglekar and Lee [68].
6.4 Quantum Chromodynamics 403
= + + + + •••
= + •••
= + •••
Fig. 6.4 The Yang-Mills vectors, the ghosts and the spinors self energy parts. A curly line denotes
a Yang-Mills field, a dashed line denotes a ghost and a solid line denotes a spinor
= + + + + •••
= + + •••
= + + •••
= + + + + + •••
Fig. 6.5 One loop contributions to vertex parts with three and four vector particles external lines,
as well as with two spinor-one vector, and two ghost-one vector external lines
6.4 Quantum Chromodynamics
Quantum chromodynamics (QCD) is the modern theory of strong interaction

with massless non-abelian gauge vector fields, referred to as Yang-Mills fields,
interacting with quarks as well as with themselves. It is based on the symmetry
color gauge group SU(3).19 The renormalization group, which shows how the
theory responds to scale transformations, in turn, leads to the description of the
high-energy behavior of the theory simply in terms of an effective coupling which,
thanks to the non-abelian gauge character of the theory, becomes weaker, with
the increase of energy, and eventually becomes zero, a phenomenon referred to as
asymptotic freedom. This has far reaching consequences as it allows one to develop
perturbation theory at high energies, in the effective coupling, and carry out various
applications which were not possible before the development of the theory, and
19
Classic references for the quark description of hadrons are: Gell-Mann [52], Zweig [125], Gell-
Mann [53], and for introducing color: Fritzsch and Gell-Mann [48], Fritzsch et al. [49], see also
Han and Nambu [63] and Greenberg [58].
Fig. 6.6 Elastic e p

scattering with a single P
virtual photon exchange
p
P
between them, where Q is p
spacelike, i.e., Q2 > 0 Q
γ∗
e−
k
e− k
explain the results of several experiments such as the deep inelastic scattering (DIS)
experiments20 which confirm that a nucleon contains point-like constituents and are
identified with quarks.
In the rest of this section we discuss some of the salient features of QCD and
provide a general view of some topics which will be treated in some detail in several
sections to follow.
A typical way to probe the internal structure of the proton is through electron-
proton scattering. The composite nature of the proton, as having an underlying
structure, becomes evident when one compares the differential cross sections for
elastic electron-proton scattering with the proton described as having a finite
extension to the one described as a point-like particle. The corresponding differential
cross sections, with a single virtual photon exchange between the electron and
proton, are given respectively, by (Fig. 6.6)
" Q2 ! #
˛2 E0 G2E C G2M Q2 2
d 4Mp2 2# 2#
D cos C G sin ; (6.4.1)
d˝ 4E2 sin4 #
2
E 1C Q2 2 2Mp2 M 2
4Mp2

d .point) ˛2 E0 # Q2 2#
D cos2 C sin ; (6.4.2)
d˝ 4E2 sin4 #
2
E 2 2Mp2 2
where E; E0 denote the energies of the electron before and after scattering, respec-
tively, and # denotes the scattering angle: kk0 D jkjjk0 j cos #. The first expression
above is referred to as the Rosenbluth formula and is derived in Problem 6.6 in
the target frame. The second one is obtained directly from the one carried out
for electron-muon elastic scattering in Sect. 5.9.3 (see (5.9.80)) by changing the
mass of the muon to the proton mass Mp . In both expressions (6.4.1), (6.4.2),
the mass of the electron is neglected in comparison to the mass of the proton.
Here GE .Q2 /, GM .Q2 / are the electric and magnetic form factors of the proton
normalized as GE .0/ D 1, GM .0/ D 1 C , where D 1:79 is its anomalous
magnetic moment. The obvious difference between the corresponding cross sections
for large momentum transfer Q2 , for which the form factors GE .Q2 /, GM .Q2 /
20
Panofski [95], Bloom et al. [23], and Friedman and Kendall [47].
vanish rapidly,21 establishes the finite extension of the proton when compared with
experiment.22 A recent value for the root-mean-square charge radius of the proton
is ' :88 1013 cm.23
The proton is described as composite of three quarks: u; u; d. The three standard
generations of quarks are given by
u charge C 2=3jej; c charge C 2=3jej; t charge C 2=3jej;

d charge 1=3jej; s charge 1=3jej; b charge 1=3jej;
making up six flavors all in all, with masses satisfying: mu < md < ms < mc <
mb < mt .24
The quarks also, necessarily, carry a quantum number called color.25 There are
several reasons for this. For example the baryon resonance CC , of spin 3/ 2 and
charge C2jej, is described by the triplet of quarks .u u u/, all with spin aligned in
the ground state, and is thus given by a symmetric wavefunction in contradiction
with the Spin & Statistics theorem. On the other hand, if the quarks carry an
additional quantum number, one may form anti-symmetric linear combinations
of the product wavefunctions of .u a u b u c / which certainly saves the day. Also
the cross section of the process eC e ! hadrons, with a single photon exchange
between the .eC ; e / - pair and the product, where the virtual photon creates quarks-
antiquarks pairs, as charged particles, with the three possible colors of SU(3),
relative to the cross section for the process eC e ! C , as we will see in
Sect. 6.5, is simply derived to be (see also Problem 6.7)
.eC e ! hadrons/ X nf
R.s/ D C C
D 3 .e2i =e2 /; : (6.4.3)
.e e ! / iD1
with single photon exchange as shown in part (a) in Fig. 6.7, by neglecting radiative
corrections as well as neglecting the mass of the electron and the masses of the
quarks contributing to the sum. Here s denotes the CM energy squared, and nf
denotes the number of flavors to be retained in the sum with corresponding quarks
having masses squared much less than s, for justifying the neglect of their masses,
and the ei denote their charges. The critical factor 3 accounts for the three SU(3)
21
See, e.g., Lepage and Brodsky [79, 80].
22
For some recent experiments on these form factors see Puckett et al. [102], Punjabi et al. [103],
and Andivahis et al. [9]. In earlier years a dipole fit was made for these form factors: GE .Q2 / D
O.1=Q4 /, GM .Q2 / D O.1=Q4 /.
23
Mohr et al. [93].
24
See, for example, Beringer et al. [17] and the tabulated estimated values.
25
Fritzsch and Gell-Mann [48] and Fritzsch et al. [49], see also Han and Nambu [63] and Greenberg
[58].
(a) (b)
s = −Q2
P
p
Q, Q2 < 0 Q, Q2 > 0
γ∗ γ∗ e−
e+ k
e− k
e−
(c) d (d)
u d
G u
ϑ
proton
u u
s
d s e− e+
γ∗ d
e− e−
Fig. 6.7 (a) The process eC e ! “anything”. (b) The deep inelastic scattering process e p !
e “anything”. (c) Quantum field theory at work: From valence quarks .u; u; d/ of the proton to
partons consisting in addition all possible quark-antiquark pairs and gluons as constituents of a
proton. (d) Production of two narrow jets of hadrons in eC e - annihilation, in the CM frame, with
jets axis angular distribution consistent with the distribution .1 C cos2 #/, corresponding to two
outgoing spin 1/2 particles, confirming the spin 1/2 character of the initial quark and the antiquark
(This will be investigated in Sect. 6.10.2)
p
colors. For example, in an earlier experiment26 R.s/ D 3:55 for s D 10:52 GeV.
For such an energy, we may include the 5 quarks u; d; s; c; b since mb < 5:0 GeV,
in the sum above obtaining
h4 1 1 4 1i 11
R.s/ D 3 C C C C D ; (6.4.4)
9 9 9 9 9 3
which compares well with the experimental result. Without the three factor, the
result will be quite contradictory. We will also see in Sect. 6.15.2 that the assignment
of three colors for quarks is essential for the internal consistency of the standard
model, as it leads to the elimination of anomalies which is important for the
renormalizability of the model.
The concept of color turns up to be of great importance. QCD is a theory
invariant under transformations in color space, by the group generators of SU(3),27
i.e., hadrons are necessarily colorless. Thus there is a built in constraint in the
theory that single quark states, carrying a color index, are not invariant under such
transformations and quarks are thus confined within the hadrons.
26
Ammar et al. [8].
27
See Sect. 6.1 for such transformations (rotations).
Upon the increase of the energy imparted to the proton in electron-proton

scattering, beyond the resonance region,28 experimentally the reaction changes
“character” and the differential cross section for the underlying inelastic processes
may be described (Sect. 6.9) by the expression (Sect. 6.9)
d2 ˛ 2 cos2 #=2
D Œ 2 W1 .; Q2 / tan2 #=2 C W2 .; Q2 / ; (6.4.5)
dE0 d˝ 4 E2 sin4 #=2
involving a single photon exchange, evaluated in the fixed target frame. Here
k k 0 D jkjjk 0 j cos #, and W1 ; W2 are referred to as structure functions of the
proton, have dimensionality mass1 , and are functions of two invariant variables
.Q2 ; / which in the rest frame of the proton are given by: Q2 D .k k 0 /2 D
4EE0 sin2 .#=2/ > 0, D PQ=Mp D E E0 > 0. Such processes are shown in
part (b) of Fig. 6.7, giving rise to a general hadronic final state. Because of the rapid
vanishing property of the electric and magnetic form factors with Q2 as discussed
below (6.4.2), it was expected that these structure functions will behave in a similar
fashion at large momentum transfer squared Q2 . Early experiments,29 mentioned
before, have shown that for constant x D Q2 =.2Mp / < 1, the dimensionless
functions Mp W1 .Q2 ; /, and W2 .Q2 ; / remain approximately constant, indepen-
dently of Q2 instead.30 This is reminiscent of scattering by point-like particles,
and thus provided evidence of point-like constituents of the proton. This led to
the development of the so-called parton model31 in which these point-like particles
within the proton are free and the virtual photon interacts with each of its charged
constituents independently,32 instead of interacting with the proton as a whole. That
is, it assumes that the deep inelastic scattering cross section may be calculated from
incoherent sums of elastic scattering processes of the electron with the constituents
of the proton (Sect. 6.10).
Experimentally,33 there are scaling violations of logarithmic type in Q2 , at
present attainable high energies, and are accounted for through QCD radiative
corrections. The parton model together with the scaling violation observed provides
a picture of the proton as consisting of point-like particles which are nearly free
inside the proton and have some interaction between them, at least at present high
energies. Such an interaction between quarks supports the idea of the existence of
other constituents within the proton which we call gluons mediating the interaction
28
A typical resonance is the C particle of mass 1.232 GeV, consisting of a proton p and a 0
meson.
29
Panofski [95], Bloom et al. [23], and Friedman and Kendall [47].
30
This was predicted by Bjorken [18]. A strict Q2 - independence of Mp W1 .Q2 ; /; W2 .Q2 ; /,
for constant x is referred to as Bjorken scaling.
31
Feynman [41, 42] and Bjorken and Pachos [22].
32
For an analogy to this, see part (b) of Fig. 1.7 in the introductory chapter to the book.
33
See, e.g., Eidelman et al. [35] and Beringer et al. [17].
between the quarks in analogy to QED and bind the quarks through them to
form hadrons. But unlike QED, the resulting theory would necessarily be a non-
abelian gauge theory one to ensure its asymptotic freedom nature consistent
with experiments. Such interactions, in turn, introduce the concept of partons as
constituents of a proton in a very general way, as shown in part (c) in the Fig. 6.7,
where in addition to the three quarks u u d, referred to as valence quarks, the
proton may contain other quark-antiquark pairs and gluons. All these constituents
are referred to as partons.34 Later we will see that about half of the momentum of
the proton is carried by the gluons.
The virtual photon impinging on the proton in the deep inelastic scattering
process shown in part (b) of Fig. 6.7, is, by definition, not on the mass shell thus it
acquires a longitudinal polarization in addition to the transverse ones. Thus one may
define two cross sections T , L associated with the transverse and longitudinal
polarizations, respectively. A measurement of the ratio R D L =T provides
overwhelming evidence of the spin 1/2 structure of the quarks (Sect. 6.9) rather than
that of spin 0.35
Before going into technical aspects of QCD and develop the underlying dynam-
ics, we would like to comment on an intriguing observation related to quarks. In
our presentation of the annihilation process eC e ! qNq in (6.4.3), in the CM,
one may naively expect that the q, qN , produced, travel back to back and finally hit
detectors and are observed. But this is not what happens. As soon as the products q
and qN move apart by a separation distance of the order of 1 fm, color confinement
prevents them to move further apart, and the increase36 in potential energy with
separation distance breaks the pair by creating two q; qN pairs instead, with one
pair moving essentially in the direction of the parent q and the other essentially
in the direction of the parent qN . Such a proliferation may continue generating more
such pairs moving almost in the same directions as the parents q, and qN . As the
transverse momenta are limited such break ups into pairs transversally are expected
to be limited.37 An emerging quark (antiquark) may, in turn, emit low energy
(soft) gluons with an enhanced probability, due to an inherited infrared divergence
as in QED (Sect. 5.12), to be emitted in the same direction as the parent quark
(antiquark). Colored particles join together into color-singlets, and two narrow jets
of such hadrons, are finally produced of well collimated hadrons, moving essentially
in the same direction as, and as if sprayed by, the parents q, and qN , instead of
34
Feynman [41, 42].
35
Riordan et al. [105], Abe et al. [1], and Airapetian et al. [3].
36
See Sect. 6.13.
37
As a consequence of asymptotic freedom, a large momentum transfer to a constituent particle
through the exchange of a high energy gluon is suppressed by the smallness of the effective
coupling parameter. Also due the infrared-divergence problem, as in QED, the total rate for
emitting a collinear gluon, due to its masslessness, by a quark is enhanced (see, e.g., the end
of Sect. 6.10).
having a final state of hadrons distributed isotropically (see part (d) of Fig. 6.7).38
This has not only provided further overwhelming evidence of the existence of
quarks dynamically, but due to the fact that the jet axis angular distribution is
consistent with the distribution .1 C cos2 #/, corresponding to two outgoing spin
1/ 2 particles,39 confirmed their spin 1/2 character as well. This is investigated in
Sect. 6.10.2. As a consequence of asymptotic freedom, occasionally, i.e., with a
smaller probability, due to the vanishing of the effective coupling at high energies,
a high energy (hard) gluon may be emitted as well leading to a three jet event and
so on.
To develop the dynamics of QCD, we spell out the form of its Lagrangian density,
as directly obtained from (6.2.39), to be40
X nf @
1
L D Ga G
a j go A C moj j ; (6.4.6)
4 jD1
i

where nf denotes the number of flavors, A D Aa ta , and for the SU(N) gauge
symmetry group, the generators of the group fta ; a D 1; : : : ; N 21g are represented
by N N matrices (see Sect. 6.1 for details),
Œ ta ; tb D ifabc tc ; TrŒ ta tb D ıab =2; (6.4.7)
G
c D @ Ac @ Ac C go fabc Aa Ab : (6.4.8)
Now we are ready to write down the explicit effective Lagrangian density of the
theory and work in covariant gauges, specified by a gauge parameter , with the
bare propagator of the gluon field given in (6.3.27) with D 1, for example,
representing the Feynman gauge. This effective Lagrangian density which takes
into account the ghost fields as generated from the Faddeev-Popov determinant,
is from (6.4.6) and (6.3.30) given by
1
Leff D L @ Aa @ Aa @ ! a rab !b : (6.4.9)
2

with the last term obtained by partial integration, where rab is defined in (6.2.6).
38
The formation of jets was suggested by Bjorken and Brodsky in [19]. For an early experiment
confirming this see Hanson et al. [64].
39
See Problem 6.7.

!
40
It is understood that @ is to be replaced by . @ @ /=2 in (6.4.6), see (6.2.39). We have also
set the external sources equal to zero in (6.4.6).
From (6.4.6) and (6.4.9), the interaction part of the effective Lagrangian density
becomes
go g2
LI eff D fabc @ Aa @ Aa Ab Ac o fabc Ab Ac fade Ad Ae
2 4
1
C go Œ j ; A j go facb @ ! a Ac !b : (6.4.10)
2
In high-energy QCD analysis, at an energy specified, say, by an energy scale
parameter , one may set all the masses mq of all those quarks, antiquarks q, qN ,
for which, mq << equal to zero. On other hand, the propagator of a quark q with
sufficiently heavy mass mq , behaves like . p C mq /=. p2 C m2q / D O.1=mq /. The
latter is an elementary example of the decoupling41 of particles with heavy masses
and if mq >> one carries out the analysis as if they are not part of the dynamics
and neglect their contributions. As a working hypothesis, in high-energy QCD, one
may then restrict the theory only to those quark flavors with masses mq << ,
and set these masses equal to zero in comparison to the energy scale in question,
as a leading contribution. The number of relevant flavors nf is thus determined
accordingly. For example in the application of eC e annihilation in (6.4.4), we have
neglected the contribution of the top quark whose mass is much larger than the
specified energy, and nf D 5.
In Sect. 6.5, eC e annihilation is studied in view of applications of the theory
with QCD radiative corrections in Sect. 6.8. Renormalized propagators and vertex
functions are determined in Sect. 6.6, and the expression of the renormalized quark-
gluon coupling is derived in Sect. 6.7.
The renormalization group is the subject matter of Sects. 6.8. DIS is considered
in general in Sect. 6.9, with application to the parton model in Sect. 6.10, and with
QCD corrections in Sect. 6.11. As the effective coupling is expected to become large
at low energies, a perturbative treatment of it at low energies is not very promising
in investigating the fundamental problem of quark confinement from this approach.
An alternative approach to quark confinement is through the Wilson loop and lattice
gauge theory, and is the subject matter of Sect. 6.12 and 6.13.
The effective interaction of the effective Lagrangian density defined in (6.4.10)
generates the (bare) vertices in Fig. 6.8.
41
For intricacies, conditions and a proof of the decoupling theorem, see Manoukian [82, 83, 85]
and other references therein.
6.5 eC e Annihilation 411
(a) (b) aα
aμ k1
k3
cγ
k2
bβ
gota γμ igo f abc (k1 − k2 )γ ηαβ
+ (k2 − k3 )α ηβ γ + (k3 − k1 )β ηγα
(c) aα (d) a
dδ
cμ
bβ
cγ
b kμ
−g2o f abe f cde ηαγ ηβ δ − ηαδ ηβ γ igo f acb kμ
+ f ace f dbe ηαδ ηγβ − ηαβ ηγδ
+ f ade f bce ηαβ ηδ γ − ηαγ ηδ β
Fig. 6.8 The bare QCD vertices: (a) fermion-gluon vertex, (b) three-gluon vertex, (c) four-gluon
vertex, (d) gluon-ghost vertex. The solid lines denote fermions, the curly, spring-like ones, denote
the gluons, while the dashed ones denote the ghosts
6.5 eC e Annihilation
A general expression is derived for eC e annihilation with a single photon exchange

between the annihilated pair and the product particles as shown below in view
of applications to the parton model in this section and in the investigation to the
contribution of QCD radiative corrections in Sect. 6.8 after having introduced the
renormalization group. At sufficiently high energies, we eventually neglect the mass
of the electron (positron). To the above end, the cross section for the process in the
CM frame, may be written as (jp1 p2 j D s=2, see (5.9.8), also (5.9.7))
h i
2
1 1 X 4 m Tr v. p1 1 / u. p2 2 // v. p1 1 / u. p2 2 /
D e2 ı
.Q2 /2 4 spins 4 .s 2/
X ˇ ˇ ˇ ˇ
.2/4 ı .4/ .Pn Q/ hvacˇj .0/ˇn; Pn ihn; Pn ˇ j .0/ˇvaci;
n;Pn
(6.5.1)
s D . p1 C p2 /2 (see Sect. 5.9). The various factors in this equation are spelled out,
in turn, in the following equalities given below
h iˇ
ˇ
m2 Tr v. p1 1 / u. p 2 2 / v. p1 1 / u. p 2 2 / ˇ
m!0
D p1 p 2 Cp1 p 2 p1p 2 ; (6.5.2)

X ˇ ˇ ˇ ˇ
.2/4 ı .4/ .Pn Q/ hvacˇj .0/ˇn; Pn ihn; Pnˇ j .0/ˇvaci
n;Pn
Z
˝ ˇ ˇ ˛
D .dx/ eiQ x vacˇj .x/j .0/ˇvac : (6.5.3)
From (5.16.86), the latter is simply equal to .Q2 < 0/
2 .Q2 / Q2 . Q2 Q Q /; (6.5.4)
where .M 2 / is the photon spectral function (see (5.16.42), (5.16.56)) for M 2 > 0.
Since
Œ p1 p2 Cp1 p2 p1 p2 Œ Q2 Q Q D .Q2 /2 ; (6.5.5)
with m set equal to zero, the following expression emerges for the cross section
.eC e ! hadrons/ D e2 .s/; s D Q2 : (6.5.6)
where s is the CM-energy squared.

It is customary to compare the hadronic case with the process eC e ! C .
To this end for Q2 >> m2 , we may refer to the QED lowest contribution given in
the second term on the right-hand side of (5.10.55), with m ! m , and neglect the
mass of muon, giving for the spectral function .s/C D ˛=.3s/ for muons
production. This leads to
4 ˛2
.eC e ! C / D ; (6.5.7)
3s
and
.eC e ! hadrons/ .s/

R.s/ D C C
D 12 2 s 2 : (6.5.8)
.e e ! / e
In particular, if we neglect the QCD radiative corrections, only quark-antiquark

q; qN pairs, as charged particles, contribute to .s/, (Fig. 6.9). Gluon production
necessarily involve the QCD coupling. Again at sufficiently high energies, we
may neglect the masses of the quarks, and simply make the replacement of ˛
in the expression ˛=3s, given above Eq. (6.5.7) for the muons, as follows:
6.6 Self-Energies and Vertex Functions in QCD 413
Fig. 6.9 (a) Expression for

the process eC e ! qNq, for (a) qi 2 (b) qi
nf flavors each with 3 colors.
ei e+
(b) The corresponding photon 3 Σi
self-energy part for each γ e ei ei
¯qi
quark qi with charge ei
e− ¯qi
˛ ! 3 e2i =4, where ei is the charge of a given quark, with three different colors,
and sum over flavors, i D 1; : : : ; nf to obtain by using, in the process, (6.5.6) with
.s/ ! qNq .s/,
ˇ .eC e ! qNq / X nf
R.s/ˇqNq D C C
D 3 .e2i =e2 /: (6.5.9)
.e e ! / iD1
This is the expression for the parton model for the hadronic process. The energy s
signifies how many flavors one may include in the sum. The higher the energies, the
more flavors may be included. QCD corrections to R.s/ above may be determined
after we investigate the nature of the effective coupling and the associated renormal-
ization group as applied to the theory. This is done in Sect. 6.8. An assessment of
the expression in (6.5.9) was done in (6.4.4).
6.6 Self-Energies and Vertex Functions in QCD
In this section we evaluate the self energies and vertex functions needed for
determining the renormalized quark-gluon coupling to lowest order. The integrals
given in Box 6.1 with dimensional regularization, the details of which are given
Appendix III at the end of the book, are quite useful and straightforward to use in
evaluating self energies.
Box 6.1: Integrals needed to evaluate the self-energies required for determining the renormalized
quark-gluon coupling to the leading order. Details on the integrals are given in Appendix III
Z ˇ
.dk/ 1 ˇ
ˇ D i F.2D =Q2 ; "/
.2/ .Q k/ k Reg
4 2 2
Z ˇ
.dk/ k ˇ Q
ˇ Di F.2D =Q2 ; "/
.2/4 .Q k/2 k2 Reg 2
Z ˇ
.dk/ k k ˇ i
ˇ D Œ Q2 C .4 "/Q Q F.2D =Q2 ; "/
.2/4 .Q k/2 k2 Reg 4.3 "/
1 2 "=2 2 .1 " / . " /
2 2
F.2D =Q2 ; "/ D D
.4/D=2 Q2 .2 "/
The self energy of the ghost, being of order g2o and having two external ghost
lines, does not contribute in determining the lowest contribution the renormalized
quark-gluon coupling, but only contributes to higher orders. Also the tadpole
obtained from the four-gluon vertex by joining two of its lines is zero in the
dimensional regularization scheme
(dk)
: 1
(2π )4 (Q−k)2
= 0,
Q Q Reg
as obtained directly from the third integral from the table above by contracting over
the and indices. We also need the following formulas involving the SU.N/
generators and color factors42 :
1 1
Tr Œta tb D TR ıab ; Œta ta ij D CF ıij ; Œtb ta tb D ta Œtb tb CA D ta ;
2 2N
(6.6.1)
1
facd fbcd D CA ıab ; i fabc tb tc D CA ta ; (6.6.2)
2
N2 1
TR D 1=2; CF D ; CA D N: (6.6.3)
2N
The self energies and vertex functions needed for determining the renormalized
fermion-gluon coupling to lowest order are given below. We work in the Feynman
gauge. All computations are carried out with massless quarks which will be justified
below Eq. (6.7.15) in the next section by choosing a renormalization scale
much larger than the neglected masses of these quarks in defining renormalization
constants.
All the self energies below are obtained directly from the integrals in Box 6.1.
Corresponding integrals for determining the vertex parts will be given later.
6.6.1 Fermion Inverse Propagator
The inverse of the fermion propagator is obtained from that of the electron in
Sect. 5.10.1 by replacing e by g, for a massless particle, and multiplying the self
energy part by .ta ta /ij :
Z ˇ
.dk/ . p k/ ˇˇ
Sij1 . p/ 2
D ıij p C ig .ta ta /ij ˇ ; (6.6.4)
.2/4 . p k/2 k2 Reg
42
Here F stands for fundamental, while A stands for adjoint.
.. p k// D .2 "/ .k p /; (6.6.5)
which from Box. 6.1 gives43

h " i
Sij1 D ıij 1 C g2 .1 / CF F. p2 ; "/ p; CF D .N 2 1/=2N: (6.6.6)
2
6.6.2 Inverse Gluon Propagator
This consists of three parts (see Fig. 6.8, Fig. 6.10)
.1/ .2/ .3/

D1 2
ab .Q/ D Q ıab C ˘ab .Q/ C ˘ab .Q/ C ˘ab .Q/; (6.6.7)
Z ˇ
.1/ 2 .dk/ Tr Œ .Q k/ . k/ ˇˇ
˘ab .Q/ D ig ıii Tr Œta tb ˇ ; (6.6.8)
.2/4 .Q k/2 k2 Reg
TrŒ .Q k/ k D 4 Œ2 k k C Q k C Q k C .k2 Qk/ ;

(6.6.9)
k Q−k Q−k
taγ μ tb γ ν i fcda i fcdb
ta γ μ ta γ μ
p p aμ bν aμ bν
p−k Q Q Q Q
k k
p2 p2
Q−k γα
tb γ σ tb
i fdca i fcdb p2 −k k−p2
Q Q
aμ bν aμ k aμ
k
Q Q ta γ μ i fbca
p1 −k p1 −k
k p1 tcγσ p1 tcγ
β
Fig. 6.10 Self energies and vertex functions needed for determining the renormalized quark-
gluon coupling to lowest order. The tadpole contribution to the gluon self-energy vanishes with
dimensional regularization as shown above. These are expressed here in the Feynman gauge
43
To simplify the notation we simply write F. p2 ; "/ for F.2D =p2 ; "/ given in Box. 6.1.
which give
.1/ .2 "/
˘ab .Q/ D g2 nf ıab Œ Q2 Q Q F.Q2 ; "/: (6.6.10)
.3 "/
For the second part, we have

Z ˇ
.2/ 21 .dk/ iV˛ˇ .Q C k; k; Q/ iV ˛ˇ .Q k; k; Q/ ˇˇ
˘ab .Q/ D ig Œfcda fcdb ˇ ;
2 .2/4 .Q k/2 k2 Reg
(6.6.11)
V1 2 3 .k1 ; k2 ; k3 / D .k1 k2 /3 1 2 C .k2 k3 /1 2 3 C .k3 k1 /2 3 1 ;
(6.6.12)
where the 1=2 overall factor arises from having two identical bosons in a closed
loop,
i V˛ˇ .Q C k; k; Q/ i V ˛ˇ .Q k; k; Q/

D .2k2 2kq C 5Q2 / .2 C "/Q Q C.10 4"/k k C.2" 5/.Q k C Q k /;
(6.6.13)
which give
.2/ 1 1
˘ab .Q/ D g2 CA ıab Œ.19 6"/ Q2 .22 7"/Q Q F.Q2 ; "/:
2 2.3 "/
(6.6.14)
The third part involves the ghost particle, and is given by
Z ˇ
.3/ 2 .dk/ .i.Q k/ /.ik / ˇˇ
˘ab .Q/ D ig fdca fcdb ˇ ; (6.6.15)
.2/4 k2 .Q k/2 Reg
which is explicitly integrated to give
.3/ 1
˘ab .Q/ D g2 CA ıab Œ Q2 C .2 "/Q Q F.Q2 ; "/: (6.6.16)
4.3 "/
In particular, we note that
.2/ .3/ .10 3"/

˘ab .Q/ C ˘ab .Q/ D g2 CA ıab Œ Q2 Q Q F.Q2 ; "/:
2.3 "/
(6.6.17)
Box 6.2: Integrals needed to evaluate the vertex parts required for the determination of the
renormalized quark-gluon coupling to the leading order. Details on the integrals are given in
Problems 6.8, 6.9, with " > 0 and ı > 0, corresponding, respectively, to ultraviolet and infrared
cutoffs, and appropriate mass scales D , D 0
Z ˇ
.dk/ 1 1 2 0 ı 2 ı .1 ı
ˇ i 2 2 2
ˇ D D
.2/4 . p1 k/2 . p2 k/2 k2 Reg .4/D 0 =2 Q 2 Q2 .1 C ı/
Z ˇ

.dk/
i . p1 C p2 / 1 D 0 22 ı ı 1 ı .1 C ı
k ˇ 2 2 2
ˇ D
.2/4 . p1 k/2 . p2 k/2 k2 Reg .4/D 0 =2 Q2 Q2 .2 C ı/
Z ˇ
2 " " 2 1 " 1
.dk/ k k ˇ i D 2 2 2 i
ˇ D C 0
.2/4 . p1 k/2 . p2 k/2 k2 Reg .4/D=2 Q2 2 .3 "/ .4/D =2 Q2
ı
2 0 ı ı
ı
2 1 C 2ı
2 2 2C 2
D
1 . p1 p1 C p2 p2 / C . p1 p2 C p1 p2 /
Q2 2 .3 C ı/ .3 C ı/
Therefore for the inverse of the gluon propagator, we finally have

Q Q 2.2 "/nf .10 3"/CA
D1
ab .Q/
2 2
D ıab Q C g F.Q 2
; "/ ;
Q2 2.3 "/
(6.6.18)
and most importantly, we note that the contribution of the ghost, coming from the
third part, has restored the transversality of the gluon propagator, consistent with
gauge invariance, when added to the second part.
In order to evaluate the vertex functions in Fig. 6.10, we need the integrals in
Box 6.2, worked out in Problems 6.8, 6.9. Here p21 D 0 D p22 .
Here we encounter infrared divergences, which are rigorously taken care of by
continuing the dimensionality of spacetime from 4 to 4 C ı with ı > 0 in integrals
involving these divergences, and by introducing, in the process, a mass scale D 0
to define the corresponding integrals, as done in ultraviolet regularization.44
6.6.3 Fermion-Gluon vertex
The fermion-gluon vertex consists of two parts:
.1/ .2/
aij . p1 ; p2 / D Œta ij C aij . p1 ; p2 / C aij . p1 ; p2 /; (6.6.19)
44
Dimensional regularization as a way for handling infrared divergences is not new, see, e.g., Field
[44].
The first part is given by (see Fig. 6.8, Fig. 6.10)

Z ˇ
.1/ 2 .dk/ . p2 k/ . p1 k/ ˇˇ
aij . p1 ; p2 / D ig Œtb ta tc ij ıbc ˇ :
.2/4 . p1 k/2 . p2 k/2 k2 Reg
(6.6.20)
Only the quadratic part in k is relevant in the ultraviolet region. Using the identity
k k D ˛ ˇ k˛ kˇ , gives
" 2
.1/ g2 2D "=2 2 2 1 2
"
aij . p1 ; p2 / D Œtb ta tc ij ıbc .2 "/
.4/D=2 Q2 2 .3 "/
2
.1/ D 0
2
g
C h ; ı ;
.4/2 IR Q2
(6.6.21)
.1/
where hIR .2D 0 =Q2 ; ı/ is a function of the infrared regularization parameters, and
is independent of the ultraviolet ones. Its infrared singular structure is spelled out in
Problem 6.10. Its explicit expression will not be needed in the sequel. The second
part is defined by
Z ˇ
.2/ .dk/ i V˛ˇ .k p2 ; p1 k; Q/ ˛ . k/ ˇ ˇˇ
aij . p1 ; p2 / D ig2 Œ fbca tb tc ij ˇ :
.2/4 . p1 k/2 . p2 k/2 k2 Reg
(6.6.22)
Again the ultraviolet relevant part is the quadratic part in k in the numerator in the
integrand, i.e.,
V˛ˇ .k p2 ; p1 k; Q/ ˛ . k/ ˇ ! 2 Œ k2 C .2 "/ k k; (6.6.23)
.2/
which leads for aij . p1 ; p2 / the expression

g2 2D "=2 2" 2 .1 2" g2 .2/ 2D 0
Œi fbca tb tc ij 2 .3 "/C h ; ı ;
.4/D=2 Q2 .3 "/ .4/2 IR Q2
(6.6.24)
.2/
independent of the ultraviolet ones. Its infrared singular structure is spelled out in
6.7 Renormalization Constants, Effective Coupling, Asymptotic Freedom, and. . . 419
Problem 6.10. Hence

.1/ .2/ g2 2D "=2
aij . p1 ; p2 / C aij . p1 ; p2 / D
.4/D=2 Q2
.2 "/2 " 2 .1 "
2 2
Œtb ta tc ij ıbc Œi fbca tb tc ij 2 .3 "/
2 .3 "/
2
g 2 0
C 2
Œta ij hIR D2 ; ı ; (6.6.25)
.4/ Q
is independent of the ultraviolet ones. Its infrared singular structure is spelled out
in Problem 6.10. The explicit expression of hIR .2D 0 =Q2 ; ı/, however, will not be
needed again.
6.7 Renormalization Constants, Effective Coupling,

Asymptotic Freedom, and What is Responsible
for the Latter?
A renormalized quantity cannot depend on the ultraviolet regularization parameters

."; 2D /. Referring to the expression for Sij1 , in (6.6.6), we note that

" 1 2D "=2 " 2 1 2"
1
2 .4/D=2 p2 2 .2 "/
1 h 2 2 2 i
D E ln C ln C c 1 ; for " ! 0; (6.7.1)
.4/2 " 42D p2
where c1 is some constant. Hence we may write

2 i
1h g2
Sij1 . p/ D ıij 1C CF ln 2 C a1 p; (6.7.2)
Z2 .4/2 p
1 g 2 h 2 2 i
D1C CF ln C a 0 ; (6.7.3)
Z2 .4/2 " 2D
where a0 ; a1 are constants, and a0 Ca1 D c1 Cln.4/. Hence a given choice for the
value of a0 determines the constant a1 . Note that the coefficient of 1=Z2 in (6.7.2),
is independent of the ultraviolet regularization parameters. A given choice for the
value of the constant a0 defines a renormalization scheme.
The simplest choice is to choose a0 D 0 which defines the so-called minimal

subtraction scheme .MS/. A more popular choice is to choose
a0 D E C ln.4/;
as suggested from the expression on the right-hand side of (6.7.1), and is referred
to as the modified minimal subtraction scheme .MS/, which we adopt here.45 The
wavefunction renormalization constant Z2 of a fermion may be then written in this
renormalization scheme, to second order, as
1 g2 h2 2 i N2 1
D1C CF E ln ; CF D : (6.7.4)
Z2 .4/2 " 42D 2N
One may refer to as a renormalization scale. An identical analysis gives

directly from (6.6.18)
h Q Q 1
i
D1 .Q/ D Q 2
ı ab C 1 C A.2
=Q 2
/ 1 ;
ab
Q2 Z3
(6.7.5)
g 2 2 5 h 2 i
A.2 =Q2 / D 2
nf CA ln 2 C a2 ; (6.7.6)
.4/ 3 3 Q
1
Q Q 1
Dab .Q/ D ıab 2 C ıab Z 3 1 A.2 =Q2 / 1 ;
Q Q2 Q2
(6.7.7)
where
1 g2 2 5 h 2 2 i
D 1C n f CA E ln ; CA D N: (6.7.8)
Z3 .4/2 3 3 " 42D
It also follows directly from (6.6.25) that
1h g2
2 g2 2D 0 i
aij D Œ ta ij 1C t t
b b C CA ln C a 3 C h IR . ; ı/ ;
Z1 .4/2 Q2 .4/2 Q2
(6.7.9)
1 g2 h2 2 i
D1C Œ CF C CA E ln ; (6.7.10)
Z1 .4/2 " 42D
45
Higher order loops contributions depend on the subtraction scheme adopted.
6.7 Renormalization Constants, Effective Coupling, Asymptotic Freedom, and. . . 421
N2 1
tb tb D CF I; CF D ; CA D N; (6.7.11)
2N
independent of the ultraviolet ones. Its infrared singular structure is spelled out
in Problem 6.10. That the vertex function is infrared divergent and requires an
infrared cut-off should not be surprising from our earlier QED treatment. The
explicit expression of hIR .2D 0 =Q2 ; ı/ will not be needed.
p p p
Therefore with D Z 2 ren , D Z 2 ren , A D Z3 Aren , D ren =Z1 ,
the renormalized quark-gluon coupling is given by
p
2 Z3 Z2
gs . / D go ; ˛ s .2 / D g2s .2 /=4; ˛ o s D g2o =4; (6.7.12)
Z1
2 i
2 ˛ s .2 / h 11 N 2nf ih 2
˛s . / D ˛ os 1 C E ln : (6.7.13)
4 3 3 " 42D
Upon setting
11N 2n 33 2n
f f
ˇ0 D ; ˇ0 D for SU(3) ; (6.7.14)
12 12
we may isolate the 2 -independent part of (6.7.13) by rewriting it as

h2 i 1 1
E C ln.42D / D C ln.2 /: (6.7.15)
" ˇ 0 ˛ os ˇ 0 ˛ s .2 /
Since the left-hand side of this equation is independent of 2 , this means that for all
2 such that nf of the quark flavors have masses squared << 2 ,46 the combination
on the right-hand side of the above equation, with each of the two terms depending
on the 2 ’s, is an invariant, which we denote by ln.2 /. In particular, the latter
depends on nf . This gives
1
˛ s .2 / D : (6.7.16)
ˇ 0 ln.2 =2 /
In general .33 2nf / > 0, and hence ˇ 0 is strictly positive. As 2 increases, the
effective coupling ˛ s .2 / becomes weaker and eventually vanishes. This property
of non-abelian gauge theories is referred to as asymptotic freedom. This QCD mass
scale , is obtained experimentally as will be discussed below. The above equation
46
This justifies the neglect of the masses of the nf flavors in comparison to 2 .
also gives
d
2 ˛ s .2 / D ˇ 0 ˛ 2s .2 /; (6.7.17)
d2
and the minus sign on the right-hand side of the equation is the origin of asymptotic
freedom. This is unlike the situation in QED, where (5.10.65) gives, for the effective
fine-structure constant, the relation 2 .d=d 2 / ˛. 2 / D C ˛ 2 =3 C . Higher
order loops contribution to the right-hand side of (6.7.17), may be also evaluated.
For example next to the leading term on the right-hand side of (6.7.17), due to two
loops, is given by ˇ1 ˛s3 .2 /,47 where ˇ1 D .153 19nf /=.24 2 /, and depends
on nf .
When 2 corresponds to an energy associated with a given process, ˛ s .2 /
provides an effective coupling for the theory and denotes the strength of the strong
interaction at that energy. This effective coupling must be small enough to carry out
reliable perturbative computations and the choice of the value of is important.
It is a function of the number of flavors nf retained in the theory. The effective
coupling ˛ s .2 / for 2 D MZ2 , corresponding to the mass MZ ' 91 GeV, of the
Z 0 vector boson, is experimentally of the order48 0.12. The mass MZ provides a
typical standard renormalization scale. The coupling parameter ˛ s is determined
at some momentum transfer from various experiments such as from DIS ones,
where violation of Bjorken scaling is sensitive to the value of ˛ s (as well as
from other experiments such as eC e -annihilation). Multi-loop analysis49 of the
renormalization group is then carried out to evolve such values of ˛ s to the energy
scale MZ . Such analyses, in turn, require corresponding values for to be used
consistently when solving the renormalization group equation involving the beta
function.
If, Q2 corresponds to a typical momentum squared associated with a given pro-
cess, then from the invariance property established in (6.7.15), we may use (6.7.16)
to express the effective coupling ˛ s .Q2 /, also referred to as running coupling, in
terms of the renormalized coupling ˛ s .2 / as follows
˛ s .2 / 1
˛ s .Q2 / D 2 2 2
D : (6.7.18)
1 C ˇ 0 ˛ s . / ln.Q = / ˇ 0 ln.Q2 =2 /
47
Jones [69] and Caswell [29].
48
Typically, recommendations for the value of are: 205–221 MeV, 286–306 MeV, 329–
349 MeV, for nf D 5; 4; 3, respectively. For the underlying phenomenology, as well as the
determination of ˛ s .M 2Z / from experiments, see Beringer et al. [17].
49
When considering more than one loop contribution to the beta function, the renormalization
scheme becomes important. The modified minimal subtraction method .MS/ mentioned above, is
the most widely used in this case.
6.8 Renormalization Group and QCD Corrections to eC e Annihilation 423
The second equality gives rise to a transformation of the dependence of ˛ s .Q2 / on

the dimensionless coupling ˛ s .2 / to a dependence on an energy scale instead.50
6.7.1 What Part of the Dynamics is Responsible for Asymptotic

Freedom?
Unlike QED, the vector particle, the gluon, has a direct self-interaction. Therefore
it is natural to investigate the contribution of the gluon self-interaction to ˇ 0
in (6.7.14) for the SU.3/ symmetry group. To isolate this contribution, let us
scale the three-gluon vertex V1 2 3 contribution to it by a parameter, say ,
which we can later set equal to one. This vertex function appears in the gluon self-
.2/
energy part ˘ab in (6.6.11). We have seen in (6.6.17), that the self-energy part
.3/
˘ab , involving the ghost field in (6.6.15), must be added to it to ensure gauge
.2/ .3/
invariance. Thus we must scale the expression for .˘ab C ˘ab / in (6.6.17) by
. The three-gluon vertex V1 2 3 also appears in the fermion-gluon vertex part
.2/
aij in (6.6.22). Thus we must also scale its expression given in (6.6.24) by as
well. An elementary algebra gives
1 ˇ
ˇ
ˇ0 D 14 CA 3 CA 2 nf ˇ ; CA D 3: (6.7.19)
12 D1
The gluon self-interaction is thus responsible for making ˇ0 strictly positive and
responsible for asymptotic freedom. It is interesting to note that for nf D 6, the
gluon-self interaction contribution to ˇ0 is precisely twice as big as the remainder
in magnitude.
6.8 Renormalization Group and QCD Corrections to eC e

Annihilation
Consider a dimensionless51 unrenormalized quantity G.Q2 ; ˛os /, and for simplicity

of the presentation, suppose it is a function of a single momentum (squared)
variable Q2 , and of the unrenormalized coupling ˛os , and is gauge independent,
suppressing the cut-offs in the theory. The unrenormalized theory is independent
of the renormalization scale we choose to define the renormalized theory. One
chooses at will depending on the energy range one is interested in defining the
50
This is, in general, referred to as dimensional transmutation.
51
An expression which has a non-zero dimensionality is readily taken care of by a scaling
argument.
renormalized theory. Hence we have the elementary relation
d
2 G.Q2 ; ˛os / D 0: (6.8.1)
d2
Suppose that this expression is multiplicatively renormalized via a renormalization

constant Z.2 ; ˛ s /, with ˛ s ˛ s .2 / denoting the renormalized coupling. That is,
Q2 1
Gren ln 2 ; ˛s D G.Q2 ; ˛os /; (6.8.2)
Z.2 ; ˛ s /
d
2 ˛ s D ˇ.˛ s /; (6.8.3)
d2
1 d
2 Z.2 ; ˛ s / D .˛ s /: (6.8.4)
Z.2 ; ˛ s/ d2
Upon substituting (6.8.2), in (6.8.1), using (6.8.3), and the fact that
d @ @
2 D 2 2 C ˇ.˛ s / ; (6.8.5)
d2 @ @˛ s
we obtain from (6.8.4), (6.8.2) and (6.8.1)

h @ @ i Q2
2 C ˇ.˛ s / C .˛ s / G ren ln 2 ; ˛s D 0: (6.8.6)
@2 @˛ s
This suggests to set
Q2 h Z ˛s dz .z/ i 2
Gren ln 2 ; ˛s D exp eren ln Q ; ˛s ;
G (6.8.7)
c ˇ.z/ 2
where c is some constant, to obtain

h @ @ i e Q2
2 C ˇ.˛ s / Gren ln 2 ; ˛s D 0: (6.8.8)
@2 @˛ s
Upon introducing the variable

Z ˛s
dz
yD ; (6.8.9)
c0 ˇ.z/
where c 0 is a constant, (6.8.8) may be rewritten as

h@ @ i 2
eren ln Q ; ˛s D 0;
G (6.8.10)
@y @ ln.Q2 =2 / 2
where we have used the chain rule @=@˛ s D .@y=@˛ s /.@=@y/. The solution
of (6.8.10) is of the form
2 h Z ˛s dz Q2 i
eren ln Q ; ˛s D
G C ln 2 ; (6.8.11)
2
c 0 ˇ.z/
where we have used the definition of the variable y in (6.8.9). From (6.8.8), this
gives
Q2 h Z ˛s dz .z/ i h Z ˛s dz Q2 i
Gren ln 2 ; ˛s D exp C ln 2 : (6.8.12)
c ˇ.z/ c 0 ˇ.z/
On the other hand, from (6.8.3), we may introduce the effective coupling measured
at Q2 : ˛.Q2 /, where recall that ˛ s ˛ s .2 /, and write
Q2 Z ˛s .Q 2 / dz
ln 2 D ; (6.8.13)
˛s ˇ.z/
by simple integration, to re-express (6.8.12) as
Q2 h Z ˛s dz .z/ i h Z ˛s .Q 2 / dz i
Gren ln 2 ; ˛s D exp C0 ; (6.8.14)
c ˇ.z/ c0 ˇ.z/
obtained by combining ln.Q2 =2 / with the first integral within the square brackets
in Œ : in (6.8.12). Comparing the above equation with the one in (6.8.12), leads
us to the inescapable conclusion that
hZ ˛ s .Q2 /
dz i hZ ˛s .Q 2 / dz .z/ i
C 0 D exp Gren 0; ˛s .Q2 / : (6.8.15)
c0 ˇ.z/ c ˇ.z/
Upon substituting this back in (6.8.14), gives the following remarkable solution of
the renormalization group equation in (6.8.6)
Q2 h Z ˛s .Q 2 / dz .z/ i
Gren ln 2 ; ˛s D exp Gren 0; ˛s .Q2 / : (6.8.16)
˛s ˇ.z/
The fact that the effective coupling ˛ s .Q2 / becomes weaker at high energies
allows one to use perturbation theory in this coupling. The moral in this is that
the high energy behavior of the theory, described by Q2 becoming large, has now
shifted from the dependence of the theory on the momentum Q to a dependence on
the effective coupling instead which becomes small in the high energy limit. This, as
we have seen in the previous section, is due to the non-abelian character of the gauge
field and its inherited remarkable property one refers to as asymptotic freedom.
An elementary application of the renormalization group is to obtain the QCD

corrections to eC e annihilation to that of the parton model. The underlying result
has become quite important as the theory shows constancy, as one goes to higher
energies, of R.s/,52 defined by the ratio of the cross section eC e ! hadrons to
that of eC e ! C , consistent with experiments.
To the above end, the transverse part of the photon propagator has the structure,
Z 1
Q Q Z 3. / 2 .M 2 /
D .Q/ D C dM ; (6.8.17)
Q2 Q2 i
0 Q2 C M 2 i
where the photon wavefunction renormalization constant Z 3 . / is given by

Z 1
Z 3. / D 1 dM 2 .M 2 /: (6.8.18)
0
Upon substituting this expression in (6.8.17) gives

Q Q d.Q2 /
D .Q/ D ; (6.8.19)
Q2 Q2 i
Z 1
M 2 .M 2 /
d.Q2 / D 1 dM 2 2 : (6.8.20)
0 Q C M2 i
The renormalized counterpart of d.Q2 /, is given by the expression
1
dren .Q2 / D d.Q2 /: (6.8.21)
Z 3. /
Sice .M 2 / has an overall factor ˛ (the fine-structure constant), and is small, one
may write
Z 1
1
D1C dM 2 .M 2 /; (6.8.22)
Z 3. / 0
and we have
Z Z
2
1
2 2
1
M 2 .M 2 /
dren .Q / D 1 C dM .M / 1 dM 2
0 0 Q2 C M 2 i
Z 1 Z 1
M 2 .M 2 /
D1C dM 2 .M 2 / dM 2 ; (6.8.23)
0 0 Q2 C M 2 i
52
See also Sect. 6.5.
leading to
Z 1
.M 2 /
dren .Q2 / D 1 C Q2 dM 2 ; (6.8.24)
0 Q2 C M2 i
which is different from the expression of d.Q2 / in (6.8.20). When we take their
imaginary parts, however, they coincide, as seen from (6.8.24) and (6.8.20),
Im dren .Q2 / D Im d.Q2 / D Q2 .Q2 /; (6.8.25)
as a consequence of the assumption of working to the leading order in ˛ with the

corresponding approximation made in (6.8.22).
That is
d d
2 Œ Im dren .Q2 / D 2 2 Œ Im d.Q2 / D 0; (6.8.26)
d2 d
where is a renormalization energy scale. Upon writing 53

Im dren .Q2 / D ˛ G ln s=2 ; ˛ s ; s D Q2 > 0; (6.8.27)
using the relation
d @ d˛ @ d˛ s @
2 D 2 2 C 2 2 C 2 2 ; (6.8.28)
d2 @ d @˛ d @˛ s
and the facts that

. /
d˛ 2 dZ
2 2
D ˛ . / 3 2 D O.˛ 2 /; (6.8.29)
d Z3 d
d˛ s @
@
2 2
˛ G ln s=2 ; ˛ s D ˛ ˇ.˛ s / G ln s=2 ; ˛ s C O.˛ 2 /;
d @˛ s @˛ s
(6.8.30)
lead to the simple equation

h @ i
2 d˛ s @
2 2
C 2
G ln s=2 ; ˛ s D 0: (6.8.31)
@ d @˛ s
53
Note that since Q is time-like, i.e., Q2 < 0, means that Q2 > 0.
From (6.8.4), (6.8.8), (6.8.16), the solution is simply given by

G ln s=2 ; ˛ s D G.0; ˛ s .s//; (6.8.32)
from which one may carry out standard perturbation theory in ˛ s and then replace
it by ˛ s .2 /.
From the relation in (6.8.25), (6.5.9), for single photon exchange, with s denoting
the CM energy squared, and with
.11N 2nf / .33 2nf /

ˇ0 D D ;
12 12
for the SU(3), color group (see (6.7.14)), the following expression then emerges, up
to additional vanishing terms,
.eC e ! hadrons/ Xnf h 1 i

2 2
R.s/ D D 3 .e =e / 1 C c 1 C ;
.eC e ! C / iD1
i
ˇ 0 ln.s=2 /
(6.8.33)
and is properly normalized to the one in (6.5.9) in the absence of radiative

corrections, where the perturbative expansion coefficients are numerical constants,
the computations of which are quite tedious. For example, c1 D 1=,54 and is
obtained from the last three photon self-energy parts, of order ˛ g2 , shown in
Fig. 6.11, together with the first part corresponding to the parton model.
The main thing is that the parton model is consistent with the above description
at sufficiently high energies. The fact that c1 > 0 means that the approach to the
parton model is from above.
The simplicity of the parton model should, however, be emphasized as it provides
a convenient starting point for straightforward computations in QCD. The next
section is involved with the fundamental theory of deep inelastic scattering.
Fig. 6.11 Photon self energy parts giving rise to the leading QCD contribution to eC e
annihilation
54
The value of this constant may be inferred from a computation carried out many years ago by
Jost and Luttinger [70] in QED, see also Bjorken and Drell [21, p.344], Appelquist and Georgi
[11], Zee [124] as well as Baikov et al. [13].
6.9 Deep Inelastic Scattering: Differential Cross Section and Structure Functions 429
6.9 Deep Inelastic Scattering: Differential Cross Section

and Structure Functions
The progress that has been made over the years in physics through scattering
experiments has been quite significant. The emergence of the endless variety of
particles by accelerators, the extraction of information on their interactions and
the determination of the structure of matter are a few of the examples concerned
with the analysis of scattering processes. Remarkable experimental and theoretical
studies of resolving the structure of nucleons through the scattering of electrons
and neutrinos off them have been, in particular, carried out. And, in the process,
dynamical evidence for the existence of point-like55 constituents was established.
To probe and unravel the structure of a proton, for example, by e p collision, via
the exchange of a photon of spacelike momentum squared Q 2 , imparted to the
proton, the latter should be large enough.56
Here we are interested in the above inelastic e p scattering beyond the reso-
nance region,57 in the neighborhood of a couple of GeVs, at which experimentally
the reaction changes “character” and the scattering process may be appropriately
represented by e p ! e C “anything”, where “anything” refers to any particles
that may be created in the process with a final large invariant mass Mn , referred
to as deep inelastic scattering, a process in which the proton loses its identity. The
interest in such a process involving the proton (or a neutron) is to choose Q2 .> 0/
sufficiently large to unravel its underlying structure, as mentioned above, confirm its
“quark description” with quarks as point-like constituents, and establish consistency
of asymptotic freedom with experiments, with quarks inside the proton being nearly
free at small distances. The process may be represented diagrammatically as shown
in Fig. 6.12, where a one photon exchange is considered as a leading contribution to
the process in question.
In the target frame, in which the ingoing proton is at rest, the momentum
transferred squared Q2 > 0 is given by Q2 D 2EE 0 .1 cos #/ D 4EE 0 sin2 #=2,
Fig. 6.12 The deep inelastic

e p ! e “anything”
scattering process
p
P } Anythin
g
γ∗ Q, Q2 > 0
e−
k
k
e−
55
Point-like at least at the energies in which such experiments were carried out.
56
To obtain an order of magnitude of such a momentum squared needed, wep note that since the
size pof the proton is of the order of 1 fm .1013 cm/, we ought to have h= Q 2 < 1 fm, i.e.,
2= Q 2 < 1 fm 5=GeV or 1:6 GeV2 < Q 2 or even larger.
57
A typical resonance is the C of mass 1.232 GeV, consisting of a proton p and a 0 meson.
where # is the angle through which the electron is scattered: k k 0 D jkjjk0 j cos #.
If P, as indicated in the diagram, denotes the (four) momentum of the proton, the
variable of the structure functions in (6.4.5), written in an invariant manner, is
defined by
PQ
D ; Q D k k 0; (6.9.1)
M
where Mp M is the mass of the proton. We note that if Pn denotes the total
momentum of all the particles created in the process corresponding to “anything”
above, with an invariant mass squared Mn2 D .P C Q/2 , then PQ D .Q2 C Mn2
M 2 /=2. In the rest frame of the proton, we then have
Q0 D .Q2 C Mn2 M 2 /=2M > 0: (6.9.2)
The expression for the differential cross section of the process is readily obtained.
By averaging the initial spins of the electron and proton, and summing over the spin
of the emerging electron, the differential cross section may be then simply written,
in proton rest frame, as
.e/4 1 X X 2 jhk0 0 I n; Pn jk; Pij2 d3 k0

d D 4
.4m 2M/ .2/4 ı 4 .P C Q Pn / ;
Q 4 spins 4ME .2/3 2E 0
.n;Pn /
(6.9.3)
where E; E 0 are the initial and final energies of the electron, respectively, and Pn
is the total momentum of the particles in state jni,
hk 0 0 I n; Pn jk; Pi D u.k 0 ; 0 / u.k; /hn; Pn j j .0/jPi; (6.9.4)
j .0/ is the electromagnetic current associated with the proton.

We use the facts that ˛ D e2 =4, and for m ! 0, d3 k 0 D E 02 dE 0 d˝ 0 , the
relations
X ˇ
ˇ
m2 TrŒu.k 0 ; 0 / u.k; /u.k; / u.k 0 ; 0 /ˇ D Œk 0 k C k 0 k kk 0 ;
m!0
spins
(6.9.5)
and (see Problem 6.11)
1X 1 X
.2/4 ı 4 .P C Q Pn /hP; j j .0/jn; Pn ihn; Pnj j .0/jP; i
2 2 n;P
n
X Z
1 1
D .dy/ei Q y hP; j j .y=2/j .y=2/jP; i W .; Q2 /;
2 2
(6.9.6)
Fig. 6.13 The process

p ! “anything” for
determining general
p
P } Anythin
g
expressions of the tensor W Q

γ∗
R
averaging over the spin of the proton, where .dy/ corresponds to an integration
over spacetime, and the multiplicative factor 1=2 above is chosen for conve-
nience.
To find the general expression for W , we note that the corresponding relevant
process for it is given in Fig. 6.13, involving a virtual photon (Q2 ¤ 0/. Here
we have at our disposal two vectors Q; P. Also current conservation @ j D 0,
implies that Q W D 0, Q W D 0. Hence W has the general structure
Q Q 1 PQ Q PQ Q
W D . /W1 .; Q2 / C 2 .P /.P /W2 .; Q2 /:
Q2 M Q2 Q2
(6.9.7)
W1 ; W2 are referred to as structure functions of the proton, and have dimensionality
mass1 . They are functions of two invariant variables. It is more convenient to
write the structure functions as functions of the invariant variable Q2 and the
dimensionless one
Q2
x D : (6.9.8)
2M
Upon neglecting the mass of the electron in comparison to that of the proton, the
differential cross section for the process in question, is, from (6.9.3), (6.9.4), (6.9.5),
(6.9.6), (6.9.7), then given by
d2 ˛ 2 cos2 #=2
D Œ 2 W1 .x; Q2 / tan2 #=2 C W2 .x; Q2 / : (6.9.9)
0
dE d˝ 0
4 E2 sin4 #=2
In particular, for elastic scattering, “anything” is replaced by an emerging proton,

and the expression within the square brackets may be then expressed in terms of the
electric and magnetic form factors of the proton as we have encountered in (6.4.1).
The invariant mass squared Mn2 associated with final state may be conveniently
written as

2 2 2 1
Mn D M C Q 1 : (6.9.10)
x
The above equality implies that x 1. For elastic scattering x D 1. Since > 0,
we then have, in general,
0 x 1: (6.9.11)
Experimentally, as the available invariant mass Mn of the final state is made to

increase, by high momentum transfers squared Q2 , and large , for fixed values
of x < 1, defining the so-called Bjorken limit, the dimensionless functions
MW1 .x; Q2 /; W2 .x; Q2 /, become functions of the dimensionless ratio x and
depend on logarithms of Q2 , at present available high energies, depending on
the value of x,58 instead of rapidly vanishing like the elastic form factors, as
one may naïvely expect. Note from (6.9.10), that as Q2 ! 1, x < 1, one is
moving “deeply” into the inelastic region Mn2 ! 1, involved with “deep inelastic
scattering”.
One may directly derive positivity properties of the structure functions. Consider
the target frame in which the proton is at rest: P D .M; 0/. p Suppose the virtual
photon three-momentum is in the z direction Q D .; 0; 0; Q2 C 2 /, where we
have used the fact that Q2 D Q23 2 . The polarization vectors of the virtual photon
(Q2 > 0) may be then taken as

p p

1 D .0; 1; 0; 0/;
2 D .0; 0; 1; 0/;
0 D . Q2 C 2 ; 0; 0; /= Q2 :
(6.9.12)

A transversely polarized photon, leads to a cross section T
W
, D W1 ,
for D 1; 2. Hence W1 0. On the other hand, a longitudinal polarization,59 leads
to a cross section (see Problems 6.12)
Q2 C 2
L
0 W
0 D Œ W2 W1 0: (6.9.13)
Q2
Accordingly, upon multiplying the above expression, by 2 x M, we are led to the

basic inequalities

4 M 2 x2
0 2 x M W1 1C W2 : (6.9.14)
Q2
One often sets
M W1 .x; Q2 / F1 .x; Q2 /; W2 .x; Q2 / F2 .x; Q2 /; (6.9.15)
expressed in terms of dimensionless functions. The basic inequalities in (6.9.14)

then read

4 M 2 x2
0 2 x F1 .x; Q2 / 1C F2 .x; Q2 /: (6.9.16)
Q2
58
See, e.g., Eidelman et al. [35].
59
Longitudinal refers to the fact that the 3-vector of the polarization vector is parallel to Q.
The necessity of having two structure functions is due to transverse as well as

longitudinal polarization states available for virtual photons.60
The above also suggests to define transverse and longitudinal structure functions
as follows:
FT .x; Q2 / D 2 x F1 .x; Q2 /; (6.9.17)

4 M 2 x2
FL .x; Q2 / D 1 C F2 .x; Q2 / 2 x F1 .x; Q2 /: (6.9.18)
Q2
For future reference, we also define the ratio of these structure functions
FL .x; Q2 /
R.x; Q2 / D : (6.9.19)
FT .x; Q2 /
To find the proportionalities coefficients of the transverse and longitudinal parts

of the cross section above, we define the flux factor (5.9.8), p corresponding to
the p
initial proton, with averaged spin, and virtual pair by .PQ/2 C Q2 M 2 D
2
M 1 C 2xM= M D Q =2x. Hence in the deep inelastic region, we have
from (6.9.6), (6.9.9), (5.9.7) for a given polarization
e2
D .2M/
.2W /
; (6.9.20)
4 .Q2 =2x/
where in reference to (5.9.7), jM j2 averaged over the spin of the proton jM j2 D

e2 .2M/
.2W /
. In particular, from (6.9.13), (6.9.15))
F2 .x; Q2 / Q2 Q2
D 2
T C L ; F1 .x; Q2 / D T; (6.9.21)
x 4 ˛ x 8 2 ˛ x
for Q2 ! 1, ! 1, and x ¤ 0 fixed.

We next consider the celebrated parton model to describe the above process in
question, spell out the underlying assumptions involved in it, and derive the explicit
expressions for the structure functions which show dependence only on the variable
x, and are independent of Q2 . The latter condition is referred to as Bjorken scaling.
As mentioned above, experiments also show logarithmic dependence on Q2 . As
we will see later, a QCD treatment of the problem, leads to explicit calculable
dependence on the logarithms of Q2 . The simpler description of deep inelastic
scattering via the parton model is discussed next.
60
The longitudinal polarization state arises as a consequence of the fact that for a virtual photon
Q2 ¤ 0, and the photon ceases to be massless.
6.10 Deep Inelastic Scattering, The Parton Model, Parton

Splitting; QCD Jets
6.10.1 The Parton Model
As the available invariant mass Mn of the final state in e p scattering is made

to increase beyond the resonance region, by high momentum transfers squared Q2 ,
and large , for fixed values of x < 1, experimentally the dimensionless functions
MW1 .x; Q2 /; W2 .x; Q2 /, become functions of the dimensionless ratio x, defined
in (6.9.8), and depend on logarithms of Q2 , at present available high energies, for
given values of x,61 instead of rapidly vanishing like the elastic form factors, as
one may naïvely expect. This non-vanishing property of the structure functions is
reminiscent of scattering off point-particles. The parton model consists in assuming
that the partons within the proton, as point-like, are free and the virtual photon
interacts with each of its charged constituents independently, instead of interacting
with the proton as a whole. That is, it assumes that the deep inelastic scattering cross
section may be calculated from incoherent sums of elastic scattering processes of
the electron with the charged constituents of the proton. If we represent interactions
between partons by people holding hands, then if you pull one person you drag
everybody else along. In the parton model the situation is illustrated by part (b) of
the Fig. 1.7, in the introductory chapter, rather than by its part (a).
We will see below, that the parton model leads to the dependence of the structure
functions on the dimensionless ratio x, but are independent of Q2 , a property
referred to as Bjorken scaling, and the physical significance of this parameter
naturally arises. QCD corrections studied in the next section accounts for the actual
dependence of the structure functions on the logarithm of Q2 as well.
The momentum carried by a quark, as a fraction of the momentum of the proton,
is almost collinear with the momentum of the proton. The process for such a parton
to acquire a significant transverse momentum through the exchange of a high energy
gluon is suppressed by the smallness of the effective coupling at large momentum
transfer. On the other hand, although the effective coupling becomes small at large
momentum transfer, a quark may emit a collinear gluon, and due to the inherit
infrared divergence associated with the masslessness of the gluons, as in QED, this
compensates the smallness of the effective coupling and such a process becomes
significant. This is discussed at the end of the section, and in detail in the next section
in which QCD corrections, to the naïve parton model, are taken into account.
Let P (0 1), denote the fraction of momentum of the proton carried
by a parton. The differential cross section of elastic scattering of such a parton
with the electron, with one photon exchange, as shown below, is simply obtained as
61
6.10 Deep Inelastic Scattering, The Parton Model, Parton Splitting; QCD Jets 435
p parton
ξP ei
parton Q
γ∗ e−
e k
k
e−
Fig. 6.14 Elastic scattering of the electron, via a one photon exchange, with a parton of charge
ei carrying a fraction P of the momentum of the proton. The gluon, of course, has no
electromagnetic charge
follows. Let ei denote the charge of an i th parton (Fig. 6.14)).62 The structure tensor

Wi contribution due to the i th parton is then simply given by (see Problem 6.13)
Z
e2i =e2 d3 p 0 .4/
Wi D ı .P C Q p 0 / Œ 2 2 P P C .P Q C P Q / PQ :
M 2 p 0 0
(6.10.1)
Now we use the integral
Z
d3 p 0 .4/ 1
ı .P C Q p 0 / D ı. x/; (6.10.2)
2p00 2 jPQj
established in Problem 6.14, where x is defined in (6.9.8). Using the facts that
QP=jQPj D 1, D x, according the delta function ı. x/ constraint, the
definition of the variable x, and conveniently grouping the various terms in (6.10.1),
gives for the latter
e2i =e2
Q Q
Wi .; Q2 / D ı. x/
2M Q2
e2i =e2 1
PQ Q
PQ Q
C x ı. x/ 2 P P ; (6.10.3)
M Q2 Q2
where is defined in (6.9.1). We may infer that x denotes the fraction of momentum
carried by a quark (antiquark). Let fi ./ denote the number densities of partons of
type i having momenta in the range .; C d/P. Then the structure tensor W
62
Since gluons, as neutral partons, do not carry electromagnetic charge, they will have no role in
this analysis. We will say more about this at the end of the section and elaborate on how they are
taken into account in a proper QCD treatment which is the subject of the following section.
of the deep inelastic process, according to the parton model, is given by
XZ 1

Q Q 1 X 2 2
dfi ./Wi .; Q2 / D .e =e / fi .x/
i 0 Q2 2M i i
1
PQ Q
PQ Q 1 X 2 2
C P P .e =e / x fi .x/: (6.10.4)
M2 Q2 Q2 i i
The structure functions are then explicitly given by

X
2 M x W1 .x/ D .e2i =e2 / x fi .x/ 2 x F1 .x/; (6.10.5)
i
X
W2 .x/ D .e2i =e2 / x fi .x/ F2 .x/; (6.10.6)
i
where we have used the definitions of the dimensionless structure functions F1 ; F2

in (6.9.15). Thus Bjorken scaling holds in the parton model, with the structure
constants depending only on x, and are independent of Q2 . It also gives the
following equality involving structure functions63
2 x F1 .x/ D F2 .x/: (6.10.7)
The above equality is quite significant as it provides ample dynamical evidence of

the spin 1/ 2 character of quarks. To this end, using the definitions of the transverse
and longitudinal structure functions in (6.9.17), (6.9.18), we obtain for the ratio R,
defined in (6.9.19), the expression R D 4M 2 x 2 =Q2 0 for Q2 ! 1. On the other
hand, if quarks are of spin 0, it is easily shown, as worked out in Problem 6.15, that
FT D 0, leading to exactly the opposite behavior R ! 1. Although the strict
equality in (6.10.7) does not hold in practice, the numerical data 64 for 2 x F1 =F2 is
concentrated around the value one, and hence from the definition of R, this implies
consistency with a spin 1/ 2 character for quarks rather than of spin 0.65
It is worth noting that if one takes into account only the three basic quarks
(u; u; d), associated with the proton, into consideration,66 and P assumes that
2 2
they occur P with the same frequency, then
P we may write .e
i i =e / x fi .x/
D x f .x/ Pi .e2i =e2 / D x f .x/ D .1=3/ i x fi .x/, where we have used the
2 2
fact that
R1 i .ei =e / D .4=9/ C .4=9/ C .1=9/ D 1. It leads to the relation
R1 P
0 dx F2 .x/ D .1=3/ 0 dx i x fi .x/. This is consistent with rough estimates
63
This equality is referred to as the Callan-Gross relation [28].
64
See, e.g., Bodek et al. [24].
65
The presence of such spin 0 bosons, if any, in addition to spin 1/2 ones, is possibly not of
significance.
66
These quarks are referred to as valence quarks and the remaining as sea quarks, and antiquarks.
R1 P R1
0 dx i x fi .x/ 0:54, and 0 dx F2 .x/ 0:18 obtained by integrating over deep
inelastic scattering experimental data.67 They provide evidence of fractional charges
of the quarks. Similar results hold for the deep inelastic scattering associated with
the e n system (see Problem 6.16).
In the development of the parton model above, gluons, as neutral partons, did
not play any role as they have no electromagnetic charge. Hence it is expected
that the sum of the fractions of the proton’s momentum, carried by the quarks
and antiquarks, do not add up to match the momentum of the proton. The
R1 P
experimental order of magnitude rough estimate 0 dx i x fi .x/ 0:54 indicates
that approximately the other half of the momentum of the proton is carried by neutral
partons - the gluons. Hence such a result is hardly surprising. In the next section,
dealing with QCD corrections, gluons are naturally taken into account as a result of
their emissions by quarks and from the underlying QCD interactions.
Experimentally, Bjorken scaling is violated by logarithmic terms in the momen-
tum transfer squared Q2 , for high values attainable in present experiments, and are
accounted for in a QCD treatment of the problem as investigated in the next section.
6.10.2 Parton Splitting
In preparation for the analysis to be carried out in the next section, which takes into
account QCD corrections to the naïve parton model, in deep inelastic scattering,
at large Q2 , we investigate the nature of a quark emitting a gluon, as well as a
gluon splitting, derive the probabilities of such events and see how these modify the
corresponding parton distributions with a variation in Q2 .
To the above end, we recall the first expression
F2 .x; Q2 / Q2
D ; (6.10.8)
x 4 2 ˛ x
in (6.9.21) in deep inelastic scattering, where is the cross section for the scattering
of a virtual photon off the proton: . p ! “anything"/. A quark with fractional
momentum of that of the proton, upon the emission of a gluon, say, acquires a
momentum with a fractional longitudinal momentum z of the parent quark, thus
having a final fractional longitudinal momentum z x of that of the proton. For
the naïve parton model, we may write
Z Z
1 1
Q2 h 4 2 ˛ z i
fi .x/ D d fi ./ dz ı.z x/ ı.z 1/ ; (6.10.9)
0 0 4 2 ˛ z Q2
67
Halzen and Martin [61, p.203]. See also Field and Feynman [45] for similar data.
easily verified upon integrations. For the emission of a gluon by a quark, the
distribution function will change fi .x/ ! fi .x; Q2 /, and from the above expressions
given in (6.10.8), (6.10.9), we may write
Z Z
1 1
Q2 h 4 2 ˛ z e2 i
fi .x; Q2 / D d fi ./ dz ı.z x/ ı.z 1/ C 2

O i .z; Q 2
/ ;
0 0 4 2 ˛ z Q2 ei
(6.10.10)
where O i .z; Q2 / is the cross section of the scattering: q ! qG, where G

denotes a gluon, and if the momentum of the initial quark is p, then the fractional
longitudinal momentum of the final quark is zp. In the next section, we will see
how to modify the above expression to take into account other relevant processes
and also to reflect the dependence of these processes on Q2 at high energies. To
lowest order, there are two diagrams associated with the above process with the
same final products of a quark and a gluon. They are given by
p
γ∗ : 1 = ei g u(p , σ )γ μ −(pγ (p−Q)
−Q) ν
2 γ tc u(p, σ ) μ eν ,
p
G Q
p
G −γ (p+Q) μ
p : 2 = ei g u(p , σ )γ ν tc (p+Q)2
γ u(p, σ ) μ eν ,
γ∗
Q
where the e are polarization vectors of the gluon, and the

are polarization
vectors of the virtual photon.68 We define Mandelstam variables (5.9.14) relevant to
the parton scattering above in the following way:
p
G :
p
Q
sO D . p C Q/2 ; Ot D . p p 0 /2 ; uO D .Q p 0 /2 ; sO C Ot C uO D Q2 :

(6.10.11)
The masses of all the partons are set equal to zero. On the other hand, the photon
being virtual means that it is not on the mass shell and Q2 ¤ 0, in general. Actually
we are interested in the large Q2 behavior. Upon using the identity TrŒtc tc D 4, for
the SU.3/ group, and averaging over spins, polarizations, and the three colors of
the latter group, give for the average jM j2 of jM j2 , where M D 2mi .M1 C M2 /,
68
See (6.9.12).
the expression .˛s D g2 =4/

ˇ 128 2 e2 h Ot sO uO Q2 i
ˇ
jM j2 ˇ D ˛ ˛ s i2 C 2 ; (6.10.12)
mi D0 3 e sO Ot sO Ot
where mi is the mass of the quark. The definition of the parameter z as the
fractional longitudinal momentum of the daughter quark relative to the parent
quark momentum, and the definition of the invariant sO, given above, lead from
the equalities sO C Q2 D 2pQ, z D Q2 =2pQ (see (6.9.8)), to the relations
Q2 sO .1 z/
zD ; D 1 z; sO D Q2 : (6.10.13)
sO C Q2 sO C Q2 z
Since uO D .Os C Ot C Q2 /, we may use the relations just given above, to rewrite the
expression in the square brackets in (6.10.12) as
1 C z2 1
Q2 z2 2 z2
Œ: D O
t C C : (6.10.14)
z Q 4 .1 z/ 1 z Ot Q2 .1 z/
In the CM frame, (5.9.15) gives
dO i 1
D jM j2 ; (6.10.15)
dOt 64 sO p2CM
.Os C Q2 /2 Q4 Q2
sO p 2CM D D 2; Ot D .cos 1/; (6.10.16)
4 4z 2z
and 0 j Ot j Q2 =z . For Q2 large, (6.10.12), (6.10.13), (6.10.14), (6.10.15),

(6.12.16) lead to
Z 2
8 ˛z ˛ s e2i h 1 C z2 i Q =z dj Ot j
O i .z; Q2 / D : (6.10.17)
3 Q2 e 2 1 z 0 j Ot j
The integral involves an infrared divergence coming from its lower end of inte-
gration, corresponding to emission of a collinear gluon, which in here may be
rigorously defined by introducing an infrared cut-off. As we are interested in the
variation of the distribution function in (6.10.10) with Q2 , we do not have to carry
out this procedure here. Accordingly, we may write
4 2 ˛ z e2i ˛ s 4 h 1 C z2 ih 2
i
O i D ln.Q / C ; (6.10.18)
Q2 e2 2 3 1 z
where the remaining terms are independent of Q2 for sufficiently large Q2 .

From (6.10.10), the following expression emerges
Z Z
2
1 1
˛ s 4 h 1 C z2 i
•fi .x; Q / D d fi ./ dz ı.z x/ •ln.Q2 /: (6.10.19)
0 0 2 3 1 z
The denominator .1 z/ in Œ.1 C z2 /=.1 z/ is not integrable at z D 1. This,

however, may be defined by a “C” operation, in the sense of distributions, which
for a given smooth function f .z/, one sets
Z 1 Z 1
f .z/ f .z/ f .1/
dz D dz : (6.10.20)
0 .z 1/C 0 .z 1/
Also to the same leading order above, we have to take into account the radiative
(self-energy and vertex) corrections, where the quark reabsorbs the gluon and the
quark remains a quark, it does not split into quark plus a gluon. This contribution
may be represented by an expression A ı.z 1/ where A is a constant, which
modifies the expression
within the square brackets in (6.10.19) to Œ.1 C z2 /=.1
z/C C Aı.z 1/ . Conservation of probability then leads from (6.10.10) to
Z
1
˛ s 4 h 1 C z2 i
dz ı.z 1/ C C Aı.z 1/ lnQ2 D 1; (6.10.21)
0 2 3 .1 z/C
giving A D 3=2, with Q2 expressed relative to any scale. Thus we introduce the
so-called splitting factor of a quark into a quark and a gluon, where the emerging
quark has a fractional longitudinal momentum in the range .z; z C dz/ relative to
the momentum of the parent quark:
4 h .1 C z2 / 3 i
Pq!qG .z/ D C ı.z 1/ : (6.10.22)
3 .1 z/C 2
The same analysis gives PqN !NqG .z/ D Pq!qG .z/, and a similar analysis leads
to the splitting factors Pq!Gq .z/ D PqN !GNq .z/, where now the gluon carries a
fractional longitudinal momentum z of the momentum of the quark (antiquark),
and is recorded in (6.10.41).
Now we come to the interesting case of the splitting of a gluon. To the same
order, the gluon may split to a quark antiquark pair, or split into two gluons, denoted
by PG!qNq .z/ D PG!Nqq .z/, and PG!GG .z/. We evidently have the constraint
bβ
z p2
z
p2
p1
p aμ p p1 a α p
1−z 1−z cγ
Z 1 Z 1 h i
dz z ı.z 1/ D 1; dz z nf PG!qNq .z/ C nf PG!Nqq .z/ C PG!GG .z/ D 0;
0 0
(6.10.23)
corresponding to the splitting gluon with momentum p, where nf denotes the

number of quark flavors considered, and note the directions of the momenta taken.
To obtain PG!GG , we recall the expression for the three-gluon vertex in
Fig. 6.8:

abc
V˛ˇ D ig fabc . p C p2 / ˛ˇ . p2 p1 /˛ ˇ . p1 C p/ˇ ˛ ; (6.10.24)
As in (6.10.16), (6.10.17) a singularity arises in the zero mass limit for collinear
momenta of the gluons with the parent one. The momenta are taken as
k2 k2
p D .k; 0; 0; k/; p2 D zk C ? ; k? ; zk ; p1 D .1 z/k ? ; k? ; .1 z/k ;
2zk 2zk
(6.10.25)
q
where note that p02 D z2 k2 C k2? ' zkCk2? =2zk, and p01 follows by conservation
of energy. Also p22 D O.k4? /, that is the corresponding gluon is nearly on the mass
shell, while
k2?
p21 ' ; (6.10.26)
z
and the corresponding gluon is to be treated as a virtual particle, the expression of

which we record here for future reference.
ˇ
Upon introducing polarization vectors e˛ , e2 , e1 , as in (6.9.12), multiplying the
latter vectors by the three-gluon vertex, taking its absolute value squared, and
summing over the final polarizations states, and averaging over the initial ones, give
4g2 k2? h z 1z i

jVj2 D CA C C Cz.1 z/ ; (6.10.27)
z.1 z/ 1 z z
where CA is the group factor equal to 3 for the SU.3/ group given in (6.6.2), (6.6.3).
We consider, in turn, the cross section of a process : AG ! BG, via the exchange
of a gluon, with the momentum of the incoming gluon given by p in (6.10.25). The
flux factor of the process as given in (5.9.7), (5.9.8) becomes
1 1 .1 z/
D D
4 ppA 4 k p0A .1 pzA =p0A / 4 .1 z/k p0A .1 pzA =p0A /
.1 z/ .1 z/
' ' ; (6.10.28)
4 p01 p0A .1 0
pA =pA /
z
4p1 pA

where note that p1 ' .1 z/k; 0; 0; .1 z/k D .1 z/p. By using
d3 p2 k dz dk2?
' ; (6.10.29)
.2/3 2p02 .2/3 2 z k
and the expression for the inverse of the squared of the gluon momentum squared
p21 as obtained from (6.10.26)
1 z2
D ; (6.10.30)
p41 k4?
the above mentioned cross section may be represented for k2? ! 0, by

z p2
B
p1
p
1−z
Z Z i
1
k dk2? g2 k2? h z 1z
GA ! GB D dz 12 C C Cz.1 z/
0 .2/3 2 z k z.1 z/ 1 z z
2
z
4
Œ.1 z/ GA ! B ; (6.10.31)
k?

where GA ! B denotes the cross section with an incoming real transverse gluon
having a longitudinal momentum .1 z/k, where the factor .1 z/ multiplies
the cross section just mentioned occurring on the right-hand side of (6.10.31) as a
consequence of the numerator in the last equality in (6.10.28). The above equation
simplifies to
Z Z
1
dk2? ˛ s
GA ! GB D dz 2 2
PG!GG .z/ GA ! B ; (6.10.32)
0 k?
displaying the mass zero singularity occurring in the integration over transverse
momentum for a gluon with almost collinear momentum with the parent gluon, and
.˛ s D g2 =4/
h z 1z i
PG!GG .z/ D 6 C C z.1 z/ ; (6.10.33)
.1 z/C z
where we have used the distributional definition of the term 1=.1 z/ as given
in (6.10.20). We have to add a term Cı.z1/ to the above expression corresponding
to the emission of a soft gluon. The constant C is determined from the constraint
equation (6.10.23). Before determining the constant C, it is straightforward to work
out, by an almost identical treatment as the one just given, and obtain the splitting
factor of a gluon to a quark-antiquark pair :
1 2
PG!qNq .z/ D Œ z C .1 z/2 : (6.10.34)
2
In view of determining the constant C, note that
Z 1
1
dz z Œz 2 C .1 z/2 D ; (6.10.35)
0 3
Z 1 h i
z 1z 11
dz z C C z.1 z/ D ; (6.10.36)
0 .1 z/C z 12
from which (6.10.23) gives C D .11=12 nf =18/. Note the following equalities
hold between the splitting factors:
Pq!qG .z/ D PqN !NqG .z/; (6.10.37)

Pq!Gq .z/ D PqN !GNq .z/; (6.10.38)
PG!qNq .z/ D PG!Nqq .z/: (6.10.39)
The last equality follows from that in (6.10.34) since the latter is invariant under
the exchange z $ .1 z/. In Pa!bc .z/, parton b has a fractional momentum z
relative to that of parton a. Accordingly, note that Pq!qG .z/ ¤ Pq!Gq .z/.
The splitting factors are then explicitly given by
4 h 1 C z2 3 i
Pq!qG .z/ D PqN !NqG .z/ D C ı.z 1/ ; (6.10.40)
3 .1 z/C 2
h
4 1 C .1 z/ 2i
Pq!Gq .z/ D PqN !GNq .z/ D ; (6.10.41)
3 z
1
PG!qNq .z/ D PG!Nqq .z/ D Œz 2 C .1 z/2 ; (6.10.42)
2
h z 1z 11 n i
f
PG!GG .z/ D 6 C C z.1 z/ C ı.z 1/ :
.1 z/C z 12 18
(6.10.43)
As before, nf denotes the number of quark flavors, and 1=.1 z/C is defined
in (6.10.20).
A unified notation generalizing the one in (6.10.32) may be given as follows:
Z Z
1
dk2? ˛ s
aA ! bB D dz 2 2
Pa!bc .z/ c A ! B : (6.10.44)
0 k?
6.10.3 QCD Jets
One of the successful applications of QCD is in providing consistent descriptions of

jet events observed in the process e eC ! hadrons.69 When a quark-antiquark
pair is initially produced in such a process, these emerging particles develop
bremsstrahlung cascades of narrowly collimated gluons, which in turn lead to the
further productions of quark-antiquark pairs. These processes continue until all
colored particles are combined into color singlet hadrons leading to the generation
of jets of hadrons at large distances. The formation of hadrons from the q; qN pair
is called hadronization.70 When a gluon, for example, is radiated from an initially
emerging quark (antiquark) with a sufficiently large (hard) energy that one is able
to detect it experimentally, and with a non-zero transverse momentum, such that
its momentum is well separated (non-collinear) from the direction of the emerging
quark (antiquark) momentum after it emits the gluon,71 a system evolves into a
system of 3 jets and so on for the generation of higher jets events. By the analysis
of the angular distributions of the angular orientations of these jets, the spins of the
initial particles in the annihilation process may, in principle, be predicted.
The purpose of this subsection is to investigate formally such jet structures
as they follow from lowest order QCD. In the figure below the process of e eC
annihilation, via a single photon exchange, is given, for the production of a qNq pair,
as well as the emission of a gluon G, and the emission and re-absorption of a gluon
in the pair production process.
In the CM system of e eC , Q2 D .Q0 /2 , and the longitudinal momentum of
an emerging quark q in the upper diagram in part (b) of Fig. 6.15, is a fraction z
of the intermediate (virtual) quark momentum, while the longitudinal momentum of
an emerging gluon G is of fraction .1 z/, and similarly for the lower diagram
69
See, e.g., Hanson et al. [64] for experimental observations of two jet events, and Brandelik et al.
[25], Berger et al. [16], and Bartel et al. [14] for three jet events.
70
For specific illustrations of this, cf., Field and Feynman [46], Hoyer et al. [66], and Ali et al. [4].
71
Such a gluon is referred to as a hard non-collinear gluon.
Fig. 6.15 (a) Diagrams corresponding to the process e eC ! q qN , including radiative correc-
tions, to lowest order, via the exchange of a photon. (b) Diagrams corresponding to the process
e eC ! q qN G, to the leading order, via the exchange of a photon
involving the antiquark qN . From (6.10.44), and the expression for the splitting of a
quark to a quark and a gluon in (6.10.40) for z < 1, we may write for k2? ¤ 0,
d2 ˛ s 4 1 C z2 1
2
.e eC ! q qN G/ D 2 .e eC ! q qN /; (6.10.45)
dz dk? 2 3 1 z k2?
where the overall factor 2 is due to the equivalence of the two diagrams in part (b)
in Fig. 6.15 with q $ qN , and from (6.5.9)/6.5.7),
4 ˛ 2 X 2 2
.e eC ! q qN / D .ef =e / 0 ; E2 D .Q0 /2 : (6.10.46)
E2 f
In particular, note that the expression for the cross section in (6.10.45) will
become arbitrarily large for k2? ' 0, and all 0 .1 z/ 1, i.e., whether the
gluon is soft or hard, with the gluon moving in the same direction (collinear gluon)
as the parent quark (antiquark). On the other hand, as the value of k2? deviates
from a non-zero value, defined by the condition, say, k2? =S & ı 2 , where ı is a
small parameter: 0 < ı 1, as a measure of departure from a collinear behavior
of a radiated gluon, and S is some scale, the cross section will take large values
for a very soft gluon .1 z/ ' 0. Such a soft gluon will not be observable if
its energy is beyond the experimental resolution, i.e., say, for .1 z/ < , for a
small parameter: 0 < 1. In both cases where k2? ' 0, 0 .1 z/ 1,
or k2? =S & ı 2 , .1 z/ < , one will be dealing with a 2-jet event where either
a gluon is moving together with a quark (antiquark) and both in almost opposite
direction to the antiquark (quark), or the quark and antiquark are moving in almost
opposite directions, and in addition a non-collinear soft gluon is radiated of energy
too low to be detected, respectively. In the CM of the e eC system, the momenta
of the external particles, in reference to part (b) of Fig. 6.15, may be represented as
follows:
p
p
k
p
The above in turn, leads to define a 3-jet event, for k2? =S ı 2 , .1 z/ , i.e.,
z 1 , with the radiated gluon being necessarily hard, and non-collinear.
To the leading order, one may define the total cross section by
.e eC ! hadrons/ D 2jets C 3jets ; (6.10.47)
where
.e eC ! hadrons/ D .e eC ! q qN / C O.˛ s /: (6.10.48)
The expression for a 2-jet event may be the obtained by subtracting the 3-jet
event contribution from the total cross section in (6.10.48). Since the integral of the
expression in (6.10.45), over z and k? 2
is singular for z ! 1, k2? ! 0, its leading
behavior for the 3-jet event, will come from values of z . 1 , and of k2? =S & ı 2 ,
corresponding to points very close to these singular points, and for which (6.10.45)
is valid. Using the elementary integrals
Z Z 1 C z2 h3 i
1
dk2? .1 /
D ln ı 2 ; dz D C 2 ln C O. /;
ı2 k2? 0 1z 2
(6.10.49)
to integrate the expression in (6.10.45), to obtain
4˛ s
3jets D 0 Œ.3 C 4 ln / ln ı C O.1/; (6.10.50)
3
where O.1/ is well defined for ! 0; ı ! 0. From (6.10.47), (6.10.48), (6.10.46)

and the just derived equation, we have to the leading order
n 4˛ s
o
2jets D 0 1 3 C 4 ln ln ı C O.1/ ; (6.10.51)
3
where 0 is given in (6.10.46).
6.11 Deep Inelastic Scattering: QCD Corrections 447
Actually the expression within the curly brackets in the above equation is
independent of the angle the jet axis makes with the one made by the annihilating
e eC pair in the CM,72 hence directly from Problem 6.7 we may write the expres-
sion for differential cross section corresponding to the cross section in (6.10.51) as
follows
d2jets .1 C cos2 #/ X 2 2 n 4˛ s
o
D ˛2 3 .e =e / 1 3 C 4 ln ln ı C O.1/ :
d˝ 4E2 f
f
3
(6.10.52)
The angular distribution: / .1 C cos2 / of the initially emerging pair q; qN , is

that of particles of spin 1/2. This jet axis angular distribution is consistent with
experiment and is in good support of the spin 1/2 character of quarks.73 The 3-jet
event is consistent with a spin 1 for the gluon.
The advantage of deriving the expression for the 2-jet event from the 3-jet event
one, to the leading order, is that there is a detailed cancelation of infrared divergence
in part (a) of the diagrams in Fig. 6.15 of the same nature we have encountered in
Sect. 5.12, while the 3-jet one involves no such infrared divergences.
6.11 Deep Inelastic Scattering: QCD Corrections
In deep inelastic scattering, one is, in particular, working at large momentum transfer
squared Q2 to the proton. At such a given momentum transfer squared, a quark
may emit a gluon, as discussed at the end of the last section, and, in turn, a gluon
may emit other gluons or a quark-antiquark pair. Accordingly a parton density
distribution of finding a parton with fractional momentum in a range .x; x C dx/
will change, and will depend on Q2 , which we denote by fi .x; Q2 /, with i standing
for quarks .q/, antiquarks .Nq/, or gluons (G). Secondly, the coupling parameter
will not be simply ˛ s but becomes ˛ s .Q2 /, that is, it will also depend on Q2 .
Examples of the splitting of partons into other partons are shown in Fig. 6.16.
Upon the increase of the momentum transfer Q2 to Q2 C •Q2 , a parton, say of
type a, of longitudinal fractional momentum of the proton in .; C d/ may emit
partons b and c : a ! b c, where parton b, say, may have a longitudinal fractional
momentum of that of the proton in a range .x; xCdx/. We denote by z the fractional
longitudinal momentum carried by daughter parton b relative to that of the parent
72
See part (d) of Fig. 6.7 in Sect. 6.4.
73
The explicit 2-jet structure in (6.10.51) was first obtained in Sterman and Weinberg [111]. See
also Stevenson [112].
(a) (b) (c)
Fig. 6.16 Splitting of partons into other partons. The corresponding measures of probabilities of
splittings at each point of splitting, as sub-processes, are each of the order ˛ s
parton a. That is74
Q2 x
z D D ; z D x: (6.11.1)
2 PQ
The additional contribution to the structure function W2 in (6.10.6), to the

parton model, due to a splitting of a parton a into partons b and c, may be extracted
from the processes in Fig. 6.16, with the point of splitting denoting a vertex in the
processes.
As we have done at the end of the last section, we introduce a splitting factor,
denoted by Pa!b c .z/, corresponding to this “break up” of a into b and c as
follows:
“Given that a parton a, of fractional longitudinal momentum in .; C d) of
that of the proton, is found in the proton, then Pa!b c .z/ corresponds to the process
where parton a splits into partons b and c : a ! b c, with parton b carrying a
fractional longitudinal momentum z of the parent parton a.”
It modifies the parton distribution of type b with the infinitesimal change Q2 !
Q C •Q2 , as discussed below. Pa!b c .z/, we recall is referred to as a splitting
2
factor, and for various a; b; and c partons, they were investigated in the last section
and are spelled out in (6.10.40), (6.10.41), (6.10.42), (6.10.43).
The above allows us to write down a completeness probabilistic equation for the
change in the parton distribution fb .x; Q2 /, of parton of type b, with infinitesimal
increase •lnQ2 as follows: (see also (6.10.19), (6.10.44))
Z 1 Z 1 X
˛ s .Q2 /
•fb .x; Q2 / D dz d ı.x z/ fa .; Q2 / Pa!bc .z/ •ln.Q2 /:
2 0 0 a;c
(6.11.2)
74
Here it is worth recalling that in the parton model, a parton in the proton has a reduced momentum
x, compared to that of the proton, given by x D Q2 =.2 PQ/.
Read this equation as follows:

C “ For all possible partons a and c, the change of a parton b distribution,
upon infinitesimal change • ln Q2 , is given by the splitting of a parton a, with
fractional longitudinal momentum of the proton in .; C d/ and with splitting
factor Pa!bc .z/, into a parton b and a parton c, where parton b has fractional
longitudinal momentum in .z; z C dz/ of the momentum of the parent a, for all
0 1, 0 z 1, such that z D x ”. Here we have also used the effective
coupling strength to be ˛s .Q2 / at the momentum transferred squared Q2 in question
considered. B
For example, for a quark q, the following differential-integral equation emerges
directly from (6.11.2)
Z Z 1
d ˛ s .Q2 / 1
Q2 fq .x; Q 2
/ D d dz ı.x z/
dQ2 2 0 0

fq .; Q2 /Pq!qG .z/ C fG .; Q2 /PG!qNq .z/ : (6.11.3)
Needless to say, a similar equation may be then written down for an antiquark qN .
Eqn. (6.11.3), may be also rewritten in a more familiar form as
Z
d ˛ s .Q2 / 1 d
Q2 fq .x; Q 2
/ D
dQ2 2 x
h x x i
fq .; Q2 /Pq!qG C fG .; Q2 /PG!qNq ; (6.11.4)

where, in the process of writing the above equation, we have used the fact that
.x=/ D z 1 implies that x . 1/.
It remains to adjoin to the equation of the quark distribution function in (6.11.3),
an equation for the gluon distribution function fG .x; Q2 /. It is simply given by a
completeness probabilistic equation to be
Z Z 1
d ˛ s .Q2 / 1
Q2 fG .x; Q 2
/ D d dz ı.x z/
dQ2 2 0 0
hX X i
fq .;Q2 /Pq!Gq .z/ C fqN .; Q2 /PqN !GNq .z/ C fG .; Q2 /PG!GG .z/ :
q qN
(6.11.5)
In PG!GG .z/, one of the emerging gluons has a momentum reduced by a fraction
z relative to the parent gluon.75
75
The evolution equations (6.11.3), (6.11.4), (6.11.5) in Q2 , are often referred to as Altarelli-Parisi
equations [5].
The splitting factors Pq!qG .z/, PG!qNq .z/, Pq!Gq .z/, PG!GG .z/, are spelled
at the end of the last section. They describe the breaking of partons into other
R1
partons. What we will need, however, are their moments 0 dz z n1 Pa!bc .z/.
The evolution equations may be solved for the moments corresponding to an i th
parton distribution functions defined by
Z 1
M ni .Q2 / D dx x n1 fi .x; Q2 /: (6.11.6)
0
by defining, in the process, the moments of the splitting factors, referred to as

anomalous dimensions:
Z 1
Anbc D dz z n1 Pa!bc .z/: (6.11.7)
0
Upon multiplying (6.11.3), (6.11.5) by x n1 , and integrating over x from 0 to 1,

give the simultaneous equations:
d ˛ s .Q2 /
Q2 2
M nq .Q2 / D Œ M nq .Q2 / AnqG C M nG .Q2 / AnqNq ; (6.11.8)
dQ 2
d ˛ s .Q2 /
Q2 M n
.Q 2
/ D Œ M nqN .Q2 / AnqN G C M nG .Q2 / AnqN q ; (6.11.9)
dQ2 qN 2
d ˛ s .Q2 / h
X n 2 X n i
Q2 M n
.Q 2
/ D M .Q / C M n
qN .Q 2
/ A C M n
.Q 2
/ A GG ;
n
dQ2 G 2 q Gq G
q qN
(6.11.10)
R1
where we have simply noted that 0 dx x n1 ı.x z/ D n1 z n1 , and where we
have used 76 Pq!qG .z/ D PqN !NqG .z/, implying that AnqG D AnqN G , also Pq!Gq .z/ D
PqN !GNq .z/, implying that AnGq D AnGNq , and finally PG!qNq .z/ D PG!Nqq .z/,
implying that AnqN q D AnqNq .
The moments of the splitting factors in (6.11.7) are readily obtained by explicitly
integrating the splitting factors in (6.10.40), (6.10.41), (6.10.42), (6.10.43), multi-
plied by z n1 as defined in (6.11.7), over z . They are given by
4h X 1i
n
1 1
AnqG D AnqN G D 2 ; (6.11.11)
3 n.n C 1/ 2 j D2
j
4 n2 C n C 2
AnGq D AnGNq D ; (6.11.12)
3 n.n2 1/
76
See (6.10.37), (6.10.38), (6.10.39), (6.10.40), (6.10.41), (6.10.42), and (6.10.43).
1 n2 C n C 2
AnqNq D AnqN q D ; (6.11.13)
2 n.n C 1/.n C 2/
h 2 2 1 nf X n
1i
AnGG D 3 C 2 ; (6.11.14)
.n C 1/.n C 2/ n.n 1/ 6 9 j D2
j
where nf denotes the number of quark flavors.

The following values and rigorous bounds of these moments, established in
Problem 6.17, are quite useful:
16 16 1 nf
A2qG D A2qN G D ; A2Gq D A2GNq D ; A2qNq D A2qN q D ; A2GG D ;
9 9 6 3
(6.11.15)
Pn
and for n D 3; 4; : : : < 1, #4n D jD4 .1=j/,
26 8 n 25 8 25
#4 < AnqG D AnqN G #4n < ; (6.11.16)
9 3 9 3 9
7 7
0 < AnGq DAnGNq ; 0 < AnqNq D AnqN q ; (6.11.17)
9 60
11 nf 21 nf
6 #4n < AnGG 6 #4n ; (6.11.18)
2 3 5 3
and where #4n should be set equal to zero for n D 3.

To solve the evolution equations (6.11.8), (6.11.9), (6.11.10), we introduce the
variable below, and re-express Q2 d=dQ2 in terms of a derivative with respect to
it as follows (see Problem 6.18) (for the SU.3/ color group)
1 h ˛ s .Q2o / i 1 h ln.Q2 =2 / i 33 2nf

D ln D ln ; b0 D 2ˇ 0 D ;
b0 ˛ s .Q2 / b0 ln.Q2o =2 / 6
(6.11.19)
d ˛ s .Q2 / d 1
Q2 D ; ˛ s .Q2 / D ; (6.11.20)
dQ2 2 d ˇ 0 ln.Q2 =2 /
where Q2o denotes some fixed reference point, Q2 > Q2o , and where ˇ 0 is defined
in (6.7.14).
We may then rewrite the evolution equations (6.11.8), (6.11.9), (6.11.10) in the
following convenients forms77
d n 2
M .Q / D M nq .Q2 / AnqG C M nG .Q2 / AnqNq ; (6.11.21)
d q
d n 2
M .Q / D M nqN .Q2 / AnqN G C M nG .Q2 / AnqN q ; (6.11.22)
d qN
d n 2 hX X i
M G .Q / D M nq .Q2 / C M nqN .Q2 / AnGq C M nG .Q2 / AnGG : (6.11.23)
d q qN
Note the strict negativity of AnqG D AnqN G , and of AnGG , for all n D 2; 3; : : : < 1,
as shown in (6.11.15), (6.11.16), (6.11.17), and (6.11.18).
A particular simple equation results for the difference ŒM nq .Q2 / M nqN .Q2 /,
which follows from (6.11.21), (6.11.22) to be given by78
d
ŒM nq .Q2 / M nqN .Q2 / D ŒM nq .Q2 / M nqN .Q2 /AnqG ; (6.11.24)
d
whose solution is given by
h ln.Q2 =2 / i b1 AnqG
ŒM nq .Q2 / M nqN .Q2 / D ŒM nq .Q2o / M nqN .Q2o / 0
: (6.11.25)
ln.Q2o =2 /
The strict negativity condition of the anomalous dimensions AnqG , established above
(see (6.11.15), (6.11.16)), implies that
ŒM nq .Q2 / M nqN .Q2 / ! 0; Q2 ! 1; (6.11.26)
for n 2. In particular, from (6.11.6), this means that for n D 2, the total fraction
of momentum carried by the quarks and antiquarks, are asymptotically equal.
On the other hand, we may consider the combination79
hX X i
M nq .Q2 / C M nqN .Q2 / M n.S/ .Q2 /: (6.11.27)
q qN
77
The equalities of some of the moments of the splitting factors in (6.11.11), (6.11.12), (6.11.13)
should be noted.
78
This equation is referred to as the non-singlet equation.
79
S here stands for singlet.
From (6.11.21), (6.11.22), (6.11.23), we then have the following simultaneous

equations
d n.S/ 2
M .Q / D M n.S/ .Q2 / AnqG C 2 nf M nG .Q2 / AnqNq ; (6.11.28)
d
d n 2
M .Q / D M n.S/ .Q2 / AnGq C M nG .Q2 / AnGG ; (6.11.29)
d G
where the factor nf in the second term, on the right-hand side of (6.11.28) arises
because we are summing over quarks in (6.11.27) corresponding to nf flavors.
These equations may be represented in an elegant matrix form as
!
d M n.S/ .Q2 / AnqG 2 nf AnqNq M n.S/ .Q2 /
D : (6.11.30)
d M nG .Q2 / AnGq AnGG M nG .Q2 /
To solve this equation, we need to diagonalize the 2 2 matrix on its right-hand

side. To this end, we use the notations
AnqG An ; 2 nf AnqNq Bn ; AnGq Cn ; AnGG Dn : (6.11.31)
The matrix in question may be then diagonalized via the equation

An Bn C
n 0
M M 1 D ; (6.11.32)
Cn Dn 0
n
where

1 .n An / Bn
M Dp ;
.C
n n /Bn
.C
n A n / B n

1 1 Bn Bn
M Dp ; (6.11.33)
.C
n n /Bn
.C
n An / .n An /
and the eigenvalues ˙

n , are given by
1
p
˙
n D An C Dn ˙ .An C Dn /2 4 .An Dn Bn Cn /
2
1
p
D An C Dn ˙ .An Dn /2 C 4 Bn Cn : (6.11.34)
2
The eigenvalues are real for all n. In particular for n D 2, we have
16 nf
C
2 D 0;
2 D ; i.e., C
2 2 > 0; (6.11.35)
9 3
.
2 A2 / D nf =3; .C
2 A2 / D 16=9; B2 D nf =3; (6.11.36)
and for n D 3; 4; : : : < 1, it is shown in Problem 6.19 that80 C

n < n < 0 and
˙
.n An / ¤ 0 as well. Hence we may state
C
n n > 0; .˙
n An / ¤ 0; Bn > 0; for n D 2; 3; : : : < 1; (6.11.37)
where the positivity of Bn D 2 nf AnqNq follows from (6.11.17).

Upon multiplying (6.11.30) from left by the matrix M, we obtain the equation
C
d .
n An /M
n.S/
CBn M nG n 0 .
n An /M
n.S/
CBn M nG
D :
d C.C
n An /M
n.S/
Bn M nG 0
n C.C
n An /M
n.S/
Bn M nG
(6.11.38)
whose solution is readily obtained for n D 2, on account that C
2 D 0, and is
given by
M 2.S/ .Q2 / C M 2G .Q2 / D M 2.S/ .Q2o / C M 2G .Q2o /; (6.11.39)

16 2.S/ 2 nf 2 2 16 nf
Œ M .Q / M G .Q / D Œ M 2.S/ .Q2o / M 2G .Q2o /
9 3 9 3

h ln.Q2 =2 / i b 169 C n3f
1
0
: (6.11.40)
ln.Q2o =2 /
On the other hand, for n 3, we obtain after elementary algebraic steps
1 h ln.Q2 =2 / i b1 C h ln.Q2 =2 / i b1

M n.S/ .Q2 / D C
Œ : 1 2 2
0 n
C Œ : 2 2 2
0 n
;
.n n / ln.Qo = / ln.Qo = /
(6.11.41)
1 h ln.Q2 =2 / i b1 C h ln.Q2 =2 / i b1
M nG .Q2 / D C Œ : 3
0 n
C Œ : 4
0 n
;
.n
n/ ln.Q2o =2 / ln.Q2o =2 /
(6.11.42)
80
A strictly negative upper bound for C
n , and a lower bound for n are also given in
Problem 6.19.
involving the very welcome strictly negative anomalous dimensions: C

n < n < 0,
and where
Œ : 1 D .
n An /M
n.S/
.Q2o /CBn M nG .Q2o /; (6.11.43)
Œ : 2 D C .C
n An /M
n.S/
.Q2o /Bn M nG .Q2o /; (6.11.44)
.C
n An /
Œ : 3 D C Œ : 1 ; (6.11.45)
Bn
.
n An /
Œ : 4 D C Œ : 2 : (6.11.46)
Bn
The above analysis provides us a lot of information. Let us elaborate on our

findings.
We learn that for n < 1, the moments of the parton density distributions
become independent of Q2 , for Q2 ! 1, or more precisely that their limits exist,
asymptotically for truly asymptotic Q2 ! 1.
For 3 n < 1, (6.11.27)/(6.11.41), (6.11.42), show that
hX X i
M nq .Q2 / C M nqN .Q2 / ! 0; Q2 ! 1; (6.11.47)
q qN
M nG .Q2 / ! 0; Q2 ! 1: (6.11.48)
P 2
This does not mean that the underlying densities : q Œ fq .x; Q / C
fqN .x; Q2 / ; fG .x; Q2 /, vanish for large momentum transfer squared, but that they
become mainly concentrated near x D 0, such that when multiplied by x n1 , for
n 3, they give very small contributions to the corresponding moments in the sense
of distributions. Such a behavior of these densities having large contributions near
x D 0, at high momentum transfer squared Q2 , is well supported experimentally.
It signals the fact that correspondingly a large number of partons may acquire only
infinitesimally small fractions of the momentum of the proton to sum up to one, at
high momentum transfer squared Q2 .
We note that (6.11.25) is also valid for n D 1, for which A1qG D A1qN G D 0
(see (6.11.11)). From the definition of the moments in (6.11.6), this then gives
Z 1 Z 1
dx Œ fq .x; Q2 / fqN .x; Q2 / D dx Œ fq .x; Q2o / fqN .x; Q2o / : (6.11.49)
0 0
That is, the net number of quarks-antiquarks, of any given flavor, is conserved. In
addition to the three basic quarks in a proton, there may be an arbitrary number
of quark-antiquark of various flavors, within the proton, as the momentum transfer
squared Q2 is increased. We may also sum over all quarks and antiquarks to obtain
from the above equation

Z 1 hX X i Z 1 hX X i
dx fq .x; Q2 / fqN .x; Q2 / D dx fq .x; Q2o / fqN .x; Q2o / :
0 q qN 0 q qN
(6.11.50)
For the contribution of all quarks and antiquarks in a proton, we then have a
conservation law and a normaliation condition:
Z 1 hX X i Z 1 hX X i
•Q 2 dx fq .x; Q2 / fqN .x; Q2 / D 0; dx fq .x; Q2 / fqN .x; Q2 / D 3:
0 q qN 0 q qN
(6.11.51)
Equations (6.11.39)/(6.11.27), give the conservation of total momentum carried

by the partons
Z 1 hX X i
dx x fq .x; Q2 / C fqN .x; Q2 / C fG .x; Q2 /
0 q qN
Z 1 hX X i
D dx x fq .x; Q2o / C fqN .x; Q2o / C fG .x; Q2o / D 1; (6.11.52)
0 q qN
normalized to one. On the other hand, (6.11.40) gives
16 2.S/ 2 nf
M .Q / MG2 .Q2 / ! 0; Q2 ! 1; (6.11.53)
9 3
which together (6.11.26), (6.11.27), lead finally from (6.11.52) to
Z 1 X 3 nf
dx x fq .x; Q2 / ! ; (6.11.54)
0 q
32 C 6 nf
Z 1 X 3 nf
dx x fqN .x; Q2 / ! ; (6.11.55)
0 32 C 6 nf
qN
Z 1
32
dx x fG .x; Q2 / ! ; (6.11.56)
0 32 C 6 nf
for Q2 ! 1, and the gluons carry a fraction 16=.16 C 3 nf / of the momentum of

the proton.
6.12 From the Schwinger Line-Integral to the Wilson Loop and How the latter. . . 457
6.12 From the Schwinger Line-Integral to the Wilson Loop

and How the latter Emerges
We hold a quark at a space point x, in the presence of a given classical gluon

vector fields Aa .x/.81 This means that for the quark propagator S.x; x 0 /, with x,
x 0 denoting possible positions of such a quark, we must have S.x; x 0 / / ı .3/ .x
x 0 /. That is, the effective Lagrangian density is not to involve a space derivative @.
Gauge invariance, in turn, then says that it does not depend on the space component
A as well. Hence, by inspection, the Lagrangian density must have the form
1 h @ 0 @0 i
0
L D 0 M C g 0 A0 ; (6.12.1)
2 i i
where is an appropriate spinor necessary to satisfy the requirement spelled out
above, with S.x; x 0 / / ı .3/ .x x 0 / for large M.
Technically, the above may indeed be achieved by a familiar Foldy-Wouthuysen-
Tani82 transformation in QCD, which simply allows one to consider the large mass
behavior of the theory for the quarks. As we have seen above, this large M behavior
is easily written down by inspection.
To isolate field variables associated with a quark and its antiparticle the antiquark,
we note that charge conjugation via C D i 2 0 transforms a Dirac field, defined
below, as follows83
0 1 0 1
q.x/ q.x/
>
D@ A ! C
.x/ D C .x/ D @ A; (6.12.2)
2 > 2 >
i .q .x// i .q .x//
simply exchanging particle with antiparticle q.x/ $ q.x/, where we have used the
identity 2 . 2 /> D I.
81
The need of considering only a classical such field will become clear below as we will be working
with functional derivatives with respect to external sources or equivalently in the path integral
version.
82
This was suggested in a remarkable paper by Brown and Weisberger [26]. For details of the
Foldy-Wouthuysen-Tani transformation see Bjorken and Drell [20, pp. 46–52] and Manoukian
[87, pp. 944–947].
83
C D CD , 0 are given, respectively, in (I.16), (I.8) in the Dirac representation in Appendix I at
the end of the book.
Now by using the anti-commutativity of spinor fields, within a Lagrangian

density, and integrating by parts with respect to @0 , (6.12.1) takes the simple form
@ @
0 0
L D q gA 0 C M q q gA 0 C M q; (6.12.3)
i i
A0 D Ac0 tc ; A 0 D Ac0 .tc> /; tc D c =2; (6.12.4)
with c denoting the Gell-Mann matrices of Sect. 6.1.

The equation of the propagator of a quark field is given by
@
0
gA0 C M S.x; y/ D ı .4/ .x y/; (6.12.5)
i
whose solution is readily verified to be
Z x0 0 0
S.x; y/ D i expŒCi g d A0 .; x/ ı .3/ .x y/ eiM.x y / .x 0 y 0 /;
y0 C
(6.12.6)
where, with x 0 > y 0 , the Schwinger line-integral, which is a priori introduced to

generate gauge invariant expressions, is defined with a time-ordering operation
Z x0
expŒCi g d A0 .; x/
y0 C
X Z x0 Z n Z 2
D .Ci g/n d n d n1 : : : d1 Œ A0 . n ; x/ : : : A0 .1 ; x/ : (6.12.7)
n0 y0 y0 y0
We recall that in a time-ordered product, one orders the product with the term at the
latest time first, from the left. Also note that x0 n n1 : : : 1 y0 .
Note that although components of the fields Ac0 .; x/, as classical fields in the path
integral expression or as functional derivatives, commute for different values of ,
the time ordering “C” in (6.12.6), is necessary due to the non-commutativity of the
group matrices t c .
Differentiation with respect to x 0 gives (see Problem 6.20)
Z x0
@0 expŒ C i g d A0 .; x/
y0 C
Z x0
D i g A0 .x 0 ;x/ expŒ C i g d A0 .; x/ ; x 0 > y 0; (6.12.8)
y0 C
Therefore using the relations @ 0 ™.x 0 x 0 0 / D ı.x 0 x 0 0 /, the solution of the

propagator S.x; x 0 / in (6.12.6) is then easily verified, satisfying the basic relation
S.x; x 0 / / ı .3/ .x x 0 / sought.
The propagator S.x; y/ for the antiquark satisfies the same equation as the one
in (6.12.5) with A0 .x/ in it simply replaced by A 0 . That is,
Z x0 0 0
S.x; y/ D i expŒCi g d A 0 .; x/ ı .3/ .x y/ eiM.x y / ™.x 0 y 0 /;
y0 C
(6.12.9)
and again the time-ordering is necessary.

The following two elementary identities emerge from the definition of time-
ordered product in (6.12.7), the fact that the group matrices tc are Hermitian, and
the definition of A 0 , for x 0 > y 0 :
Z x0 h Z x0 i>
exp Œ i g dA 0 .; x/ D exp Œi g dA 0 .; x/ ; (6.12.10)
y0 C y0
h Z x0 i Z x0
exp Œ i g dA0 .; x/ D exp Œi g dA0 .; x/ ; (6.12.11)
y0 C y0
where
Z x0
expŒi g d A0 .; x/
y0
X Z x0 Z n Z 2
D .i g/n d n d n1 : : : d1 Œ A0 .1 ; x/ : : : A0 .n ; x/ : (6.12.12)
n0 y0 y0 y0
is defined with a time anti-ordering operation. That is in time anti-ordering, one

orders a product with the factor earliest first from the left, and note that x 0 n
n1 : : : 1 y0 .
Gauge transformations of Schwinger line-integrals, relevant to the subsequent
work, are given in Box 6.3.
Box 6.3: We have introduced a unified notation for ordering and anti-ordering along a line .0; R/
in the same way as done for the time variable. Here
A .t; x/ A .t; a/; V.t; a/ D expŒ i c .t; a/ tc ; x D .a; 0; 0/
Z T
h˙ .T; aI 0; a/ expŒ ˙ i g d A0 .; a/ ; T >0
0 ˙
Z R
h˙ .t; RI t; 0/ expŒ ˙ i g d A1 .t; / ; R>0
0 ˙
hC .T; aI 0; a/ ! V.T; a/ hC .T; aI 0; a/ V 1 .0; a/ hVC .T; aI 0; a/
h .T; aI 0; a/ ! V.0; a/ h .T; aI 0; a/ V 1 .T; a/ hV .T; aI 0; a/
hC .t; RI t; 0/ ! V.t; R/ hC .t; RI t; 0/ V 1 .t; 0/ hVC .t; RI t; 0/
h .t; RI t; 0/ ! V.t; 0/ h .t; RI t; 0/ V 1 .t; R/ hV .t; RI t; 0/
To prove the first transformation rule, define
QC .t/ D V.t; a/ hC .t; aI 0; a/; QC .0/ D V.0; a/; t 0: (6.12.13)
Using the facts that
d
hC .t; aI 0; a/ D igA0 .t; a/hC .t; aI 0; a/; (6.12.14)
dt
dV.t; a/ 1
V .t; a/ .@ 0 V.t; a//V 1 .t; a/ D V.t; a/@ 0 V 1 .t; a/; (6.12.15)
dt
we have
d h d i
QC .t/ D V.t; a/ C V.t; a/ ig A0 .t; a/ V 1 .t; a/V.t; a/ h.t; aI 0; a/
dt dt
h @ 0 i 1
D ig V.t; a/ A0 .t; a/ V .t; a/ QC .t/ igAV0 .t; a/ QC .t/;
ig
(6.12.16)
thus defining AV0 .t; a/. Integrating from 0 to T, gives
Z T
QC .T/ D exp i g d AV0 .t; a/ QC .0/ hV .t; aI 0; a/ V.0; a/; (6.12.17)
0 C
thus defining hV .t; aI 0; a/. Upon multiplying the above equation from the right by
V 1 .0; a/ and using the definition of QC .T/ in (6.12.13), the first transformation
rule in Box 6.3 follows. The proof of the other three transformations are very similar
(see Problem 6.21).
We are interested in deriving an expression of the static potential for the

interaction between a quark and an antiquark held (i.e., for M ! 1) at some
distance R apart. To this end, we place an antiquark at the origin x D .0; 0; 0/,
of a coordinate system, and a quark at a distance R from it at y D .R; 0; 0/.
Accordingly, we introduce the following corresponding state for this configuration,
and its corresponding time-developed one at time T:
j 0 i D q .0; R/ C.0; R/ q .0; 0/jvaci; (6.12.18)
j T i D q .T; R/ C.T; R/ q .T; 0/jvaci; (6.12.19)
where the kernel C.0; R/ will be specified below consistent with gauge invariance.
Consider the amplitude
G.T 0; T RI 0 R; 0 0/ D h T j 0 i (6.12.20)

D hvacj q.T; 0/D Œ C .T; R/ DC qC .T; R/ qB .0; R/ Œ C.0; R/ BA q A .0; 0/ jvaci:
C
(6.12.21)
The basic idea is to ensure that this provides a gauge invariant expression.
From the explicit expressions of the amplitudes for T > 0 obtained, respectively,
from the propagators84 in (6.12.6) and (6.12.9)

hvacjqC .T; R/ qB .0; R/jvacinorm ! ŒhC .T; RI 0; R/CB ; (6.12.22)

hvacjq D .T; 0/ q A .0; 0/jvacinorm ! Œh .T; 0I 0; 0/AD ; (6.12.23)
properly normalized to one85 for T ! 0, and from the transformation rules in

Box 6.3, it is evident one needs to set
Œ C.0; R/ BA D Œ hC .0; RI 00/ BA; Œ C .T; R/ DC D Œ h .T; RI T0/ DC ;
(6.12.24)
to obtain a gauge invariant expression
G.T 0; T RI 0 0; 0 R/
! Œ h .T; 0I 00/ AD Œ h .T; RI T0/ DC Œ hC .T; RI 0; R/ CB Œ hC .0; RI 00/ BA
I
Tr exp i g dx A .x/ ; (6.12.25)
C C

84
Recall that the propagator for a fermion field is defined by S.x; x0 / D ihvacj .x/ .x0 / jvaci,
0 00
for x > x , and for a properly normalized vacuum state, see, e.g., (3.6.21).
85
These expressions will be applied in a discrete formulation of spacetime.
Fig. 6.17 The Wilson loop,

represented by the curve C.
Here C is along the sides of a
rectangle of area R T T W [C, A] = Tr expig C dx
μA
μ (x)
+
where
I
WŒC; A D Tr exp i g dx A .x/ ; (6.12.26)
C C
is the celebrated Wilson loop and is automatically gauge invariant since the initial
and end points of the path C coincide (a closed loop) due to the nature of the
trace operation. We see how the Wilson loop emerges from Schwinger line-integrals
associated with the quark and antiquark propagators and by finally invoking gauge
invariance, giving rise to four Schwinger line-integrals going around in a closed
loop. The “C” operation, now, simply orders the group matrices ta around the
loop. The closed loop, described by the curve C, is shown in Fig. 6.17.
We were in the course of defining the transformation function h T j 0 i, where
the states j 0 i, j T i are defined in (6.12.18), (6.12.19), corresponding to an
antiquark set at .0; 0; 0/ and a quark set at .R; 0; 0/. We recall that fields may
create various particles out of the vacuum in addition, in general, to creating the
particle corresponding to it. Thus the state h T j, for example, in addition to a
quark and an antiquark, held at R and 0, respectively, at time T, may also involve
other complicated configurations. With this in mind, we may insert a completeness
relation in the definition of the above Green function in (6.12.20), to write
X
h T j 0 i D eiEn T h 0 jnihnj 0i: (6.12.27)
n
To extract the total energy of the “heavy” quark-antiquark pair,86 as just described
above, we may make the replacement T ! i T, and subsequently let T ! 1,
isolating the contribution of the lowest-energy intermediate state. In this limit, we
obtain from (6.12.27)
h T j 0 i ! eE.R/T ; T ! 1; h 0 j 0 i D 1; (6.12.28)
86
Recall that the masses were taken to be large, and hence the kinetic energy p2 =2M of each
particles goes to zero, as obtained from the limit of the relativistic energy, and only the mass M
survives in the latter.
6.13 Lattices and Quark Confinement 463
where E.R/ provides the leading contribution to the sum over the intermediate
states having the expected expression
E.R/ D E2 D 2M C V.R/; (6.12.29)
to be consistently verified, involving no kinetic energies, where the 2M are the

contributions of the rest masses of the pair, and V.R/ denotes the static potential
between the pair separated by a distance R. That h 0 j 0 i D 1, follows from the
fact that
Z R Z R
h 0 j 0 i D hvacjŒ exp i g dA1 .0; / AC Œ exp C i g dA1 .0; / CA jvaci;
0 0 C
(6.12.30)
and from (6.12.11),
Z R Z R
exp i g dA1 .0; / D exp C i g dA1 .0; / ; (6.12.31)
0 0 C
and the result follows from the unitarity of the matrices. Since in the final analysis
to follow, the problem is formulated in terms of pure gauge fields, the additive 2M
term in (6.12.29), remnant of the Fermi fields, may be omitted subsequently, and one
may introduce a corresponding normalization factor in the definition of the vacuum
state.
To extract the potential in question, we are led to consider the expression
I ˇ
1 ˇ ı
Tr exp i g dx A0 .x/ h 0C j 0 iJ ˇ ; A0 .:/ tc .i/ ;
h 0C j 0 iJ C C JD0 ıJc .:/
(6.12.32)
with A represented by the functional derivative A0 given above. Here h 0C j

0 iJ denotes the full vacuum-to-vacuum transition amplitude of the pure gauge
field theory of gluons, including self energies and closed gluon loops as well as

ghosts,87 and depends on external sources Ja .z/ coupled to the gluons. The problem
of quark confinement via the Wilson loop is the subject of the next section.
6.13 Lattices and Quark Confinement
In this section, we consider a variation of quantum field theory we have been

considering so far, where a 4D Euclidean spacetime continuum is now replaced by
a discrete one described by a lattice, with a finite small lattice spacing playing the
87
In the limit M ! 1 closed fermion loops vanish.
role of a cut-off. We will see that the lattice description gives a rise to a method for
studying the strong coupling limit of QCD. The investigation of the strong coupling
is relevant physically as it is expected that the effective coupling of QCD becomes
strong at distances large enough for the confinement of quarks within hadrons. Thus
this method provides a self-consistent method for the investigation of the problem
of quark confinement. From the extrapolation between this strong coupling limit
and the weak coupling one, emerging from the asymptotic free theory description,
one may be optimistic that one is dealing with the same theory observed at different
energy scales with a corresponding effective coupling.
To construct a lattice, with lattice spacing a, pick a point x in 4D Euclidean
space and draw a line of length a in a given direction and another one in its opposite
direction. Repeat this in the remaining mutually perpendicular directions. The end
points of every such lines will then provide a starting point, as x was, for repeating
the same procedure in a self evident way. This is shown in parts (a) and (b) of
Fig. 6.18. The points of intersections of the lines generated are referred to as lattice
sites. A line connecting two lattice sites is called a link. A 2D dimensional square
of area a2 bounded by four links is referred to as a plaquette as shown in parts (b)
and (c) of Fig. 6.18. No additional details on lattice description will be needed in the
sequel.
We introduce an appropriate gauge invariant action integral, in Euclidean space,
for the gauge field in lattice theory. To this end, we define four ordered Schwinger
line-integrals U.ì / along the links `1 ; `2 ; `3 ; `4 , in given fixed directions of
vector components specified by the indices, , and made evident by the definition
of the links:

U.ì / D expŒ i a g A.ì / C ; (6.13.1)
where the A.ì / are integrals defined below. We suppress the other two variables
along directions perpendicular to the directions specified by the indices and .
(b) (c)
ν
(a) μ
+a
−a 3
−a x +a 4 2
+a A Link 1
−a A Plaquette
A Plaquette
Lattice Site
Fig. 6.18 Construction of a lattice, with part (b) showing part of a 3 D version of a lattice
In particular for a ! 0, and fixed and ,

Z xCa
a2 @A .x; y/
A.`1 / D dx A .x; y/ ! a A .x; y/ C ; (6.13.2)
x 2 @x
Z yCa
@A .x; y/ a2 @A .x; y/
A.`2 / D dy A .x C a; y/ ! a A .x; y/ C a2
C ;
y @x 2 @y
(6.13.3)
Z xCa
@A .x; y/ a2 @A .x; y/
A.`3 / D dx A .x; y C a/ ! a A .x; y/ a2 ;
x @y 2 @x
(6.13.4)
Z yCa
a2 @A .x; y/
A.`4 / D dy A .x; y/ ! a A .x; y/ ; (6.13.5)
y 2 @y
where no summation is implied over repeated indices here.

Upon using the Baker-Campbell-Hausdorff formula for two matrices A and B:
exp A exp B D exp ŒACBCŒA; B=2 C: : :, (6.13.2), (6.13.3), (6.13.4), (6.13.5) give,
for a ! 0,

Tr ŒU.`4 /U.`3 /U.`2 /U.`1 / ! Tr exp Œ i g a2 G ; (6.13.6)
where recall that G D @ A @ A igŒA ; A .

Thus if we set
Tr ŒU.`4 /U.`3 /U.`2 /U.`1 / D Wp ; (6.13.7)
and define

1 X
Sp D ˇ 1 Re.Wp / ; SEuc D Sp ; (6.13.8)
N p
where the latter sum is over lattice points, and where ˇ is a parameter to be specified
below, we obtain for a ! 0
g2
Sp ! ˇ a4 Tr ŒG2 ; (6.13.9)
2N
where we have used the fact that TrŒ tc D 0, Tr I D N for the SU.N/ group, and
recall that TrŒ tc td D ıcd =2. Hence upon approximating the sum over all plaquettes
P
a4 p with a spacetime integral in Euclidean space, we see that for a ! 0,
Z
g2 1
SEuc ! ˇ .dx/Euc Tr ŒG .x/G .x/; (6.13.10)
2N 2
where the 1=2 factor in the integral arises because of the symmetry under the
interchange of and . Thus SEuc in (6.13.8) defines an appropriate expression
for the action integral in lattice theory, with
ˇ D 2N=g2 : (6.13.11)
With the strategy of obtaining a gauge invariant procedure for evaluating the
expression in (6.12.32), involving the Wilson loop, integrations over the vector fields
are replaced by integrations over Schwinger line-integral associated with all the
links in the lattice, involving the vector field. That is, we introduce a measure of
integration defined as a product involving all the links in the lattice normalized to
unity
Z
dŒ U D ˘ì dU.ì /; dŒ U D 1; (6.13.12)
and develop integration theory over such links by a gauge invariant procedure. This
is done in a straightforward manner as follows. Consider the integral
Z
dŒ U U.`j /: (6.13.13)
This is clearly not gauge invariant unless it is identically equal to zero. According to
Box 6.3, U.`j / has a well defined gauge transformation, and hence gauge invariance
alone implies that the above integral must be zero. On the hand, consider the integral
Z
dŒ U Œ U.ì / AB Œ U .`j /CD : (6.13.14)
According to the previous integral, this is zero unless ì and `j denote the same
link. Thus the above integral must be proportional to ıì `j . The above integral, in
turn, is gauge invariant for B D C or for D D A. For B D C, summed over,
Œ U.ì / AB Œ U .ì / BD D ıAD . Accordingly the above integral must be proportional
to ıBC ıAD ıì `j . Since the trace of the identity matrix is equal to N, for the SU.N/
group, the above integral is equal to ıBC ıAD ıì `j =N. Summarizing then, we have the
following elementary integration rules for integrations over the links in the lattice
Z Z Z
1
dŒ U D 1; dŒ U U.ì / D 0; dŒ U Œ U.ì / AB Œ U .`j / CD D ıBC ıAD ıì `j :
N
(6.13.15)
These are the only three integrations rules we will need in the sequel. In particular,
we learn:
(i) The integral over every link vanishes.
(ii) For every integral involving a link factor U.`j /, we need a compensating factor
U .`j / to obtain a non-zero value for the integral.
(iii) The integral involving the pairing of two such factors, for each link, gives rise
to a factor of 1=N.
Thus we are led to consider the integral
Z Z
1
h WŒ C; A i D dŒ U WŒ C; A exp Œ SEuc ; ZD dŒ U exp Œ SEuc ;
NZ
(6.13.16)
where WŒ C; A is defined in Fig. 6.17, and the factor 1=N is a conventional factor,
whose significance will emerge naturallyPbelow.
Clearly the constant factor exp Œ p ˇ in exp Œ SEuc , with SEuc defined
in (6.13.8), will cancel out in the numerator and the denominator in (6.13.16).
Hence we may replace exp Œ SEuc by ˘p exp Œ .ˇ=2N/.Wp C Wp1 / in the latter
equation, where we have used the fact that Re Œ Wp D .Wp CWp1 /=2, and note that
the plaquette is traversed in Wp1 in opposite direction to that of Wp , ˇN D 2N=g2 .
In the strong coupling limit
h 1
1
i 1
exp Wp C W ! 1 C .Wp C Wp1 /: (6.13.17)
g2 p
g2
The area of each plaquette, being equal to a2 , the Wilson loop going around the
sides of a rectangle of area R T, together with the elementary integration rules in
(i) - (iii), spelled out above, dictate that we must have L M plaquettes filling the
area enclosed by the Wilson rectangle, where L D R=a, M D T=a. We may break
the Wilson loop into the product of ordered Schwinger line-integrals going around
c.c.w., and every link, specified by U.ì / in the Wilson loop, will then have exactly
a partner U.ì / , coming from one of the sides of a plaquette touching the contour
of the Wilson loop at the link ì in question, as shown in part (a) of Fig. 6.19. Hence
to the leading order in the strong coupling, only the product of L M plaquettes,
arranged as shown in part (a) in Fig. 6.19, will contribution to (6.13.8).
Since we have precisely the product of L M plaquettes, we may infer
from (6.13.17), that h WŒ C; A i in (6.13.16), involves a factor .1=g2/LM . Also at
each link, due to the pairings U.ì / U .ì / for each link, as given by the rule of
integration (ii) earlier, we get a factor .1=N/.2LMCLCM/ , where .2LM C L C M/
denotes the number of links in the two dimensional rectangle of L M plaquettes
within the Wilson loop (see Fig. 6.19). Due to the fact that the group generators are
N N matrices, and hence the trace of the identity matrix, obtained by contracting
the Kronecker delta indices in rule (ii) of integration above, is equal to N, i.e.,
(a)
(b)
Number of plaquettes = LM
(3 × 4 = 12)
Number of sites = (L + 1)(M + 1)
(4 × 5 = 20)
Number of links = (2LM + L + M )
(24 + 3 + 4 = 31)
Number of plaquettes =
Number of links − Number of sites + 1
Fig. 6.19 Part (a) shows the pairings U.ì /, U .ì / that occur for each link by an example. Part
(b) facilitates counting the number of links and sites encountered in the problem by an example.
The area of the Wilson rectangle, in this presentation, is actually equal to R T, with R units
horizontally and T units vertically, L D R=a, M D T=a
equivalently there are N “colors”, per quark, associated with each lattice site, we
also get a factor N for each site, i.e., a factor N .LC1/.MC1/ . That is in the strong
coupling limit, on account of Z ! 1, (6.13.16) becomes
h WŒ C; A i ! .1=N/.1=g2/LM .1=N/.2LMCLCM/ .N/.LC1/.MC1/

2
D .1=Ng2 /LM D .1=g2 N/R T=a D exp Œ ln.g2 N/=a2 R T ; (6.13.18)
from which we conclude that .N 3/
1
V.R/ D KR; KD ln.Ng2 /; N D 3 for the SU.3/ group; (6.13.19)
a2
giving a linearly rising potential in the separation distance R between the pair. It
takes an unlimited energy to separate the pair by an unlimited distance. K is referred
to as the string tension. This is in analogy to the situation where the two particles
are joined by a string, with K D V.R/=R denoting the fixed energy per unit length
of the string, which is its tension. A rough estimate of K may be obtained by taking
V to be of the order of the rest energy of a proton 1 GeV, and R to be its size
1 fm 5=GeV, giving K 0:2 GeV2 .
With the inverse of the lattice spacing 1=a formally playing the role of an
ultraviolet cut-off, one may use (6.7.17) for the beta function, to relate the coupling,
in the weak coupling case, to the lattice spacing:
dg2.a/ ˇ0 4
a D g .a/; (6.13.20)
da 2
6.14 The Electroweak Theory I 469
where ˇ 0 is given in (6.7.14), whose solution is
1
aD exp Œ 2=ˇ 0 g2 ; (6.13.21)
L
where 1=L is an integration constant characteristic of the lattice. With a as the

dimensional parameter in lattice theory, we may write a2 K D f .g/. The string
tension, being a physical quantity, should not depend on the cut-off a. That is,
dK=da D 0. By the chain rule, this gives the following equation for f .g/
dg d
a f .g/ D 2 f .g/; with solution f .g/ D C exp Œ4=ˇ 0 g2 :
da dg
(6.13.22)
Hence (6.13.21), and the relation K D f .g/=a2 give
2
K D C exp Œ 2=ˇ 0 g2 =a D C2L ; (6.13.23)
where the coefficient C is characteristic of the string tension.

It is not a a priori clear that the continuum limit as obtained from lattice theory
will join smoothly its strong coupling limit, through intermediate coupling regimes,
to the weak coupling case, obtained already via the phenomenon of asymptotic
freedom, and establish that one is dealing with the same QCD theory all along. For
example, the continuum limit theory from the strong coupling limit, and through
intermediate coupling regimes, may finally lead to a constant non-vanishing value
for its limiting coupling parameter, instead of approaching zero as given by the QCD
asymptotic freedom. Analytical and numerical methods give strong support that the
latter does not happen, and that the strong coupling limit and the weak coupling case
are joined smoothly, through intermediate coupling regimes, indicating that one is
dealing with one theory at different scales,88 and QCD is not only asymptotically
free but is confining as well. Although confinement may be established for the
Abelian gauge theory through lattice theory in the strong coupling limit, as done
above for the non-Abelian case, there are some investigations89 indicating that the
strong coupling limit and the weak coupling one may be separated by a phase
transition and do not pertain to the same theory.
6.14 The Electroweak Theory I
In the present section, we develop the celebrated electroweak theory, starting from
the early classic Fermi theory of weak interactions. The electroweak theory is based
on the product symmetry groups SU.2/ U.1/ which is spontaneously broken to
88
See, e.g., Kogut [72], Kogut et al. [73], and Creutz [32].
89
See, e.g., Guth [60] and Lautrup and Nauenberg [76].
the U.1/ group. We then discuss how a key parameter “sin2 W ” in the electroweak
theory, where W is referred to as the Weinberg angle or the weak-mixing angle,
is experimentally determined. In the classic electroweak theory the neutrino masses
are set equal to zero. Because of this, we elaborate on how the masses of neutrinos
may be generated, and develop as well a straightforward theoretical description
of “neutrino oscillations” experiments which have become quite popular in recent
years providing a clear indication of neutrino mass differences.
6.14.1 Development of the Theory: From the Fermi Theory

to the Electroweak Theory
The pioneering work on the weak interaction theory goes back to Fermi [38, 39] in
an attempt to describe ˇ decay: n ! p C e C Q e , incorporating Pauli’s postulated
particle earlier in (1930), which Pauli90 called neutron, while Fermi referred to it
later as the neutrino. Now this particle is referred to as the anti-neutrino due to the
fact that different lepton numbers are assigned to the neutrino and anti-neutrino,
consistent with other reactions, and due the opposite helicities carried by them. The
energy of the electron produced in the reaction, from a decaying neutron at rest,
varied significantly experimentally up to the maximum energy Ee D .Mn2 Mp2 C
m2e /=2Mn , where the latter expression corresponds to the (hypothetical) reaction
involving no (anti-) neutrino, and it was accordingly expected that the neutrino is
extremely light. Thus Pauli considered it as small but actually non-zero.91
Fermi described weak interactions, by taking the interaction Hamiltonian density
to involve the product of four fermions at a point92 having the structure HI
GF Œ. 1 Oi 2 /. 3 Oi 4 / C h:c: where the Oi are, in general, expressed in terms
of gamma matrices, and GF is the Fermi coupling consistently given by GF D
1:166 105 =GeV2 . The latter constant has dimensionality of Œmass2 . Thus on
dimensional grounds, a typical cross section for the scattering of four fermions will
have the behavior .GF /2 E2 , where E is the energy associated with the process,
while a formal partial wave analysis gives . 1=E2 . This unitarity bound would
then be violated for E & .GF /1=2 300 GeV. On the other hand consider some
low orders radiative contributions to the scattering of four fermions in part (a) of
Fig. 6.20.
By assigning a dimension E1 to every fermion propagator, and a dimension E4
to every loop integration, in part (a) of Fig. 6.20, we see that even when there are no
derivative couplings, the degree of divergences of the diagrams, including radiative
90
See Pauli [96] for the historical account. The experimental discovery of the neutrino was made
by Reines and Cowan [104].
91
In this respect, see also Perrin [97] and Fermi [38, 39].
92
A quick review of the part relevant to the weak interaction in the introductory chapter in the book
is recommended.
(a)
+ + +
(b)
+ + +
Fig. 6.20 (a) The degrees of divergence increase with order of perturbation theory in this four
fermion scattering with a four-fermi-interaction at a point without a bound. (b) Unless the
longitudinal part of the propagator of the massive vector boson (represented by a wavy line) hitting
a vertex vanishes, the same behavior occurs as in (a). The longitudinal part will not contribute if
the vector boson is coupled to a conserved current
corrections, increase as 2 ; 4 ; : : : with the orders of the interaction, as functions

of an ultraviolet cut-off , with no bound. The theory becomes out of control to
give it any sense at all. To improve this divergence problem, an attempt was then
made to describe weak interactions by introducing, in the process, intermediate
massive vector bosons as mediators of interactions with two fermions and one vector
boson meeting at a point, similar to what one encounters in QED with a (massless)
vector boson being the photon, instead of four fermions meeting at a point of the
original Fermi theory.
Massive vector bosons have one additional polarization state than that of a
massless vector boson such as the photon. This extra polarization state, in turn,
gives rise to a longitudinal component in the massive vector propagator which, as
mentioned in the introductory chapter of the book, has, for example, the undesirable
property that the propagator does not vanish at high energies,93 unlike that of their
massless counter parts which vanish rapidly. This makes the degree of divergences
in such theories, involving non-vanishing longitudinal parts, in general, increase
with the orders of perturbation theory without bound as seen above in part (b)
of Fig. 6.20. These theories thus become again uncontrollable with no much hope
to give them any sense and carry out practical computations as one does in QED
through the renormalization programme. Details of this sort of such theories were
particularly investigated by Salam and Ward [108]. The longitudinal polarization
also behaves like k =M at high energies also giving rise to problems with the high
energy behavior of cross sections of processes involving external vector bosons. In
QED, even if a longitudinal component arises, because of a specific choice of a
gauge, this problem is eliminated in computing gauge invariant expressions because
the photon is coupled to a conserved current.
A different approach had to be taken to deal with the weak interactions as
discussed next. The emergence of a gauge invariant theory, such as QED or its
93
It behaves like k k =k2 M 2 for k2 ! 1.
generalization, where the vector bosons of the theory are a priori massless so that
problems arising from longitudinal components of massive ones are avoided, is to
be expected.
The idea of weak and electromagnetic unification was first advanced by
Schwinger [109] involving vector bosons, and that vacuum expectation values of
scalar fields could provide a way of breaking symmetries and generate masses for
fermions.94 In 1961, Glashow95 proposed an SU.2/ U.1/ gauge symmetry for
such a model. To deal with the problem of the masses of the vector bosons, Weinberg
[120] and independently Salam [107] proposed a renormalizable theory in which the
vector bosons are taken to be massless in the action integral, and acquire masses,
by spontaneous symmetry breaking, via the so-called Higgs mechanism, where the
Higgs boson field has a non-vanishing vacuum expectation value. The associated
photon field, however, remains massless, as it should, and the SU.2/ U.1/
gauge symmetry is spontaneously broken to the U(1) gauge symmetry of phase
transformations of QED. This theory is popularly known as the Weinberg-Salam
electroweak theory. The meaning of all this seemingly complicated “language”, but
hopefully and eventually comprehensible for the reader, will be spelled out, together
with some of the extensions of this theory, in the remaining sections of this chapter.
With the discovery of violation of parity96 and the observation in the 1950s
that neutrinos are left-handed (and hence the anti-neutrinos are right-handed),97
the leptonic current corresponding to the outgoing electron and the anti-neutrino
in ˇ decay was taken as J D e .1 5 /. To explain some processes which
were observed not to happen,98 such as ! e C , and describe other
processes, which happened, such as ! e C C , Q it was inferred that there
is a different99 neutrino associated with e , , and different quantum labelings
(lepton numbers) (Table 6.1) were assigned to the pairs .e ; e /, . ; /, with
electronic lepton-number : Le D C1, and muonic lepton-number : L D C1, “1”
for their respective anti-particles. Separate conservation laws of electronic-type
leptons and muonic-type leptons were then assumed. This, for example, explained
why processes such as Q C p ! eC C n were not observed, while processes such
as Q C p ! C C n did happen. Landau [74], Lee and Yang [78], Salam [106],
and Wu et al. [122] assumed that the neutrino is massless. The left-handedness of
the neutrino was concluded in an experiment by Goldhaber et al. in 1958 [57]. By
the late fifties, Feynman and Gell-Mann [43] and Sudarshan and Marshak [113]
suggested that all the particles massive and massless are to be taken to be left-handed
94
Martin and Glashow [88], p.16. That such a unification may lead to a renormalizable theory was
expressed by Glashow in his Ph.D. thesis in 1958 as a graduate student of Schwinger.
95
Glashow [55].
96
For the experiment, see Wu et al. [122].
97
The experiment was carried by Goldhaber et al. in 1958 [57].
98
An upper bound for the branching ratio for the decay rates is . ! e C /= . !
all/ < 2:4 1012 , at 90% confidence level (see [17]).
99
The observation that there are different types of neutrinos was made by Danby et al. [33].
Table 6.1 Lepton numbers Lepton Le L L

of the three generations of
leptons e 1 0 0
e 1 0 0
0 1 0
0 1 0
0 0 1
0 0 1
in weak interactions theory generating what is called the V A theory involving

vector minus axial vector (charge changing) currents.
The effective Fermi interaction Lagrangian density, based on four-fermions at a
point, having the structure of a V A theory, incorporating parity violation, turns
out to provide a good description for decay . ! e C Q e C /, and may be
written as
GF
LF D p J˛ J ˛ ; J ˛ D e .1 5 /e C .1 5 / ; (6.14.1)
2
from which the decay process may extracted, and the Fermi constant may be
determined experimentally.100
After the discovery of parity violation, as discussed above, the charge exchanging
interaction between leptons and hadrons, prior to the birth of the electroweak theory,
was taken to have the structure
1

p GF e .1 5 /e C .1 5 / J h C h:c:; (6.14.2)
2

involving left-handed leptons (and right-handed for their anti-particles), where J h
is a hadronic current to be specified later, with an inherit universality of the processes
involving the pairs .e ; e / and . ; / with equal strength, as shown by the
common coupling constant GF . Upon including the pair . ; /, using the anti-
commutativity of 5 with the gamma matrices, and using the notation

e
`1 D ; `2 D ; `3 D ; (6.14.3)
e L
L
L
100
The decay rate of the muon is investigated in Problem 6.24.
where L stands for left-handed, we may write the charge leptonic-current as101
X 1 i
J D ì T ì ; T˙ D T1 ˙ i T2; Ti D ; (6.14.4)
i
2
with the i denoting the Pauli matrices,102 taking into account of the fact that the
neutrinos have no charges. Invoking different lepton-number conservation, and due
to the very structure of the charge current in (6.14.4), these left handed lepton pairs
are considered, in a simplest possible way, as doublets of an SU.2/ group, referred
to as weak iso-doublets, with the latter group often denoted by SU.2/L to remind
us that these lepton doublets are left-handed. In setting up the original electroweak
theory, the neutrinos were assumed to be massless, as we do here,103 thus parity
violation allows only one helicity for them and the underlying theory would not
involve right-handed neutrinos. Accordingly, the right-handed leptons e
R , R , and

R , respectively, would not have right-handed neutrinos to be paired with and they
are considered as singlets under the group. It is convenient to denote the charged
right-handed leptons by ri ; i D 1; 2; 3. The transformation rules for a (left-handed)
iso-doublet `, is given by
1
` ! ei #T `; TD ; (6.14.5)
2
and, of course, the right-handed ones do not respond to the transformations, i.e., r !
r. The weak isospin quantum number of the doublets is given by IL D 1=2, with
the third components of weak isospin for the upper and lower entries given by I3 D
˙1=2, respectively. On the other hand, IR D 0. In terms of the electromagnetic
charge Q and the third components of the weak isospin I3 , we have the following
equation for the left-handed (L) and right-handed (R) leptons:
1
L W Q D I3 ; R W Q D I3 1 .I3 D 0/: (6.14.6)
2
This set of equations “cries out loud” to introduce a new quantum number, which
we denote by Y, and rewrite the above equations simply as (Table 6.2)
1
Q D I3 C Y; L W Y D 1; R W Y D 2: (6.14.7)
2
101
In this section, we avoid using the standard notation D =2 to avoid confusion with the
notation used for the tau particle.
102
In Sect. 6.1, we have seen that the Pauli matrices (divided by two) are generators of SU.2/
transformations.
103
The problem of the neutrino masses, together with the fact they are small, is addressed in
Sect. 6.14.3 of the present section.
Table 6.2 Electroweak quantum numbers of the leptons in the Standard Model. The anti-particles
have the same I but upposite Q, I3 (for Q ¤ 0; I3 ¤ 0), and Y D 2.Q I3 /
Particle I I3 Q Y
.` L ; `
L / 1=2 .C1=2; 1=2/ .0; 1/ 1
`
R 0 0 1 2
This new quantum number is called weak hypercharge. In turn, this suggests to
introduce the new symmetry group U.1/Y to account for this quantum number.
The combined symmetry group generated thus far is then SU.2/L U.1/Y , with Y
corresponding to the weak hypercharge. Since the weak hypercharge of the right-
handed leptons ri is twice as that of the left-handed ones ì , we have the following
transformation rules corresponding to the U.1/Y group as phase transformations
r ! ei ˇ r; ` ! ei ˇ=2 `: (6.14.8)
Promoting the transformations in (6.14.5), (6.14.8) to local ones necessitates

the introduction of four gauge vector fields: three vector bosons W1 ; W2 ; W3 for
SU.2/L group, and one vector boson B for U.1/Y , group to restore the symmetry
in a lagrangian density involving the above mentioned leptons.
From what we have learnt from QCD, except now we are working with an SU.2/
group and we are dealing with three vector bosons instead of eight, and from what
we have learnt from QED, except now the electromagnetic charge is replaced by
weak hypercharge, we may immediately write down the Lagrangian density L1
for the leptons and the vector bosons, not considering, at this stage, the quarks,
interactions with other fields and corresponding mass terms. The nature of such
terms will be investigated and added later :
1X i 1 1 X
L1 D G G i F F ì D ì C ri D 0 ri C h:c: ;
4 i 4 2 i
(6.14.9)
@ g0 @
D D C B g T W ; D 0 D C g 0 B ; F D @ B @ B ;
i 2 i
(6.14.10)
necessarily involving two coupling parameters .g; g 0 / for the product of two
groups,104
G D @ W @ W gW W ; G D .G1 ; G2 ; G3 /: (6.14.11)
104
Recall that the structure constants of SU.2/ are
abc .
To extract the physics from the Lagrangian density, we use, in the process, the
identities105
1 1
T W D p T W C T C W C T3 W3 ; W D p W1 C i W2 ;
2 2
(6.14.12)
1
W D p W1 i W2 ; D L

L C R

R: (6.14.13)
2
By expressing the two fields B , W3 , in terms of two other fields, together with
the definitions that follow
B D A cos W C Z sin W ; W3 D A sin W C Z cos W ;

(6.14.14)
tan W D g 0 =g; e1 D e; e2 D ; e3 D ; 1 D e ; 2 D ; 3 D ;
(6.14.15)
where T ˙ are defined in (6.14.4), the Lagrangian density (6.14.9) that describes the
interaction of the leptons with the vector bosons may be written as
h
1 X g0 i
L10 D ì g T W B ì g 0 ri ri B C h:c:
2 i 2
X
D g sin W ei ei A C
i

g sin W ei ˛e ˇe 5 /ei C ˛ i 1 5 /i Z

g h i
C p eiL iL W C iL eiL W : (6.14.16)
2
The first term in the summand on the right-hand side of the above equation,
describes the interaction of a neutral vector field A with the right-handed as well
as the left-handed components of the electron, muon and the tau particles, and does
not interact with the neutrinos. Accordingly, the latter field denotes the photon field,
provided it does not acquire mass by symmetry breaking, which will be investigated
next, and the combination g sin W necessarily denotes the electron charge e. This
together with the expressions of the parameters ˛e ; ˇe ; ˛ in (6.14.16) are given by
1
e D g sin W D g 0 cos W ; .3 tan W cot W /;
˛e D (6.14.17)
4
1 1
ˇe D ˛ ; ˛ D .tan W C cot W / D : (6.14.18)
4 4 sin W cos W
0
105
Note that L=R D. L=R / .
Since g is given by g D e= sin W , we see that everything so far may be determined

in terms of the angle W . This angle is referred to as the weak-mixing angle or
Weinberg angle. Also note that Z couples to both left-handed and right-handed
leptons ei , that is, the neutral current is not a pure VA current for these leptons,106
and there is a flavor universality of neutrinos Z - couplings.
The experimental determination of sin W is discussed in the next subsection.
There are aspects of the theory we have to investigate. First note that a lepton
mass term of the general structure , for a spinor , cannot be added to the
Lagrangian density since it would break the gauge symmetry. This is easily seen by
noting that
D L R C R L; (6.14.19)
and since left-handed and right-handed components have different transformations,

this expression is not invariant under the underlying symmetry transformation. It
suggests, however, how one may use it to generate leptonic masses. To make the
above expression invariant, we may multiply it by another doublet, referred to as a
Higgs field, having the structure
C

˚D ; (6.14.20)
0
with necessarily a weak hypercharge Y D 1 for invariance,107 and introduce, in

turn, the gauge invariant simple Yukawa term in the Lagrangian density
X
i

C

L`˚ D i eiR C h:c: : (6.14.21)
i
ei L
0
From the doublet transformation rules, under SU.2/L U.1/Y , we must have

C C
! ei'.x/n =2 ; (6.14.22)
0 0

C C
!e iˇ.x/=2
: (6.14.23)
0 0
Now we introduce the potential for the scalar field in (6.14.20),
m2 1
V.˚˚/ D ˚ ˚ C .˚ ˚/2 ; m2 > 0; > 0: (6.14.24)
2 4
106
It is precisely because of this that the pure neutral leptonic current, for example, involving the
electron, as discussed in Sect. 6.14.2, facilitates one to determine sin2 W experimentally.
107
See (6.14.7), (6.14.8).
and subtract it from the Lagrangian density. The negative sign of the first term is
important for spontaneous symmetry breaking and generation of masses of particles.
Working in the tree approximation (i.e., with no radiative corrections), the stationary
point of the potential is given by
r
v2 v m2
hvacj˚ jvacihvacj˚jvaci ; and p D : (6.14.25)
2 2
As a matter of fact, we can always108 find a real field '.x/, a unit vector n, and
a real field ˇ.x/, such that the transformations in (6.14.22), (6.14.23) reduce the
doublet in (6.14.20) to the form

C Q C
! ; (6.14.26)
0 Q 0
where Q 0 is a Hermitian field, referred to as the Higgs boson field, and the vacuum
expectation values of Q 0 , and Q C are given by
˝ ˇ ˇ E v ˝ ˇ ˇ ˛
vacˇQ 0 ˇvac D p ; and vacˇQ C ˇvac D 0; (6.14.27)
2
corresponding to spontaneous symmetry breaking, with v a non-negative number.

This special gauge attained by the choice of the above transformation to achieve this
structure is referred to as the unitary gauge.
From (6.14.21), (6.14.26), (6.14.27), we may infer that the leptons e; ;
acquire masses given by109
v
m i D i p : (6.14.28)
2
Since a priori no right-handed neutrinos were included in the theory, the neutrinos
do not acquire a mass.
The interaction of the vector bosons with the Higgs field arises through the
standard expression
ˇ @ g 0 ˇˇ2
ˇ
Lvec:bosons;ˆ D ˇ gW T C B ˚ ˇ : (6.14.29)
i 2
108
See Problem 6.22.
109
Here we have used the facts that eR=L eR=L D 0, as a consequence of the facts that f 5 ; 0 g D
0, and . 5 /2 D I.
From the latter term in the Lagrangian density (in the unitary gauge), the following
expression emerges for the mass terms of the vector bosons
!
ˇ @ g0 0 ˇ2
ˇ ˇ
ˇ gW T C B v ˇ
i 2 p
2
mass
v 2 g2 1 2
D W W C .g C g 0 2 /Z Z : (6.14.30)
2 2 4
The acquired masses of the vector bosons are then given by .g2 Cg 0 2 D g2 sec2 W /
v v
MW D jgj; MZ D jg sec W j; M D 0; (6.14.31)
2 2
where we note that W is a complex field, while Z is a real one, which explains
the presence of the overall 1=2 factors in the above equations. As the photon field
A does not appear on the extreme right-hand side of (6.14.30), the photon remains
massless.
The numerical value of the vacuum expectation value of the Higgs boson field
p
v= 2 may be readily estimated as follows. By using the fact the propagator of the
˛ˇ
now massive vector boson W propagator 4C .q2 / approaches110: ˛ˇ =MW 2
at low
2 2
energies jq j << MW , we may infer from the last expression on the right-hand side
of (6.14.16), that the exchange of W vector boson at low energy between, eL and
L type leptons, may be described by the effective Lagrangian density
g 2 1 h 5 ih
15 i
1
Le;eff D p 2
e e C h:c: : (6.14.32)
2 MW 2 2
By extracting the equivalent expression from the effective Lagrangian density

in (6.14.1) to the above one, and by making a direct comparison gives111
g2 p
2
D 2 GF : (6.14.33)
4M W
From this equation and the first equality in (6.14.31), we obtain the following key
estimate in the electroweak theory
1=2
1
vD p ' 246 GeV: (6.14.34)
2 GF
110
See (4.7.20).
Note that each .1 5 / in (6.14.32) is divided by a factor of 2 which accounts for the factor
111
1=4 on the left-hand side of (6.14.33).

Eq. (6.14.31) and the first equality in (6.14.17), give the following masses for the
vector bosons
v ˇˇ e ˇˇ ˇ e ˇ
ˇ ˇ
MW D ˇ ˇ; MZ D v ˇ ˇ: (6.14.35)
2 sin W sin 2W
Simultaneous measurements of differential cross sections of processes involving

electron-neutrinos scattering,112 for example, as conveniently discussed in the next
subsection, give the value
sin2 W D 0:2324 ˙ 0:0083: (6.14.36)
This provides the following estimates for the masses of the vector bosons
in (6.14.35): MW ' 77:4 GeV, MZ ' 88:3 GeV. On the other hand, taking
the value of the effective fine-structure to be ˛ ' 1=128, as determined 113
at energies of the order 91 GeV, as discussed earlier below Eq. (5.19.9) and
investigated theoretically in Sect. 5.19.2, gives for the masses estimates of the
order 80:1 GeV; 91:4 GeV, respectively, which are comparable to experimental
values.114
An expression for the Higgs boson mass may be obtained from the p potential
in
˝ ˇ(6.14.24)
˛ by working in the unitary gauge. Upon writing ˚ . C v/= 2, with
vacˇjvac D 0, we then have
ˇ ˇˇ 1
ˇ
V.˚ ˚/ˇ D . C v/2 2 C 2v v 2 ˇ D v 2 2 ; (6.14.37)
mass 16 mass 4
from which we may extract the mass of the Higgs boson to be
r

MH D v ; (6.14.38)
2
where is undetermined.115
In setting up the electroweak theory, we recall that it was originally assumed
that the neutrinos were massless, left-handed, and we have seen that the absence
of right-handed components in the theory precludes the neutrinos from acquiring
a mass. The problem of neutrino masses and their small values is investigated
in Sect. 6.14.3. This is followed by Sect. 6.14.4 dealing with neutrino oscillations
which is relevant to the masses of neutrinos. First, however, we consider an
experimental determination of sin2 W .
112
Vilain et al. [119].
113
See, e.g., Mele [89], Erler [36], and Beringer et al. [17].
114
115
If one is to rely on the validity of perturbation, and restricts
.0; 1/, then formally for '
1=2, for example, (6.14.34)/(6.14.38) give the value MH ' 123 GeV. The coupling is, however,
constrained by renormalization restrictions.
6.14.2 Experimental Determination of sin2 W
The interaction Lagrangian density for the description of purely leptonic neutral
current, i.e., via the exchange of only the Z vector boson, processes involving the
following neutrinos-electron scattering : C e ! C e , Q C e ! Q C e ,
e C e ! e C e , Q e C e ! Q e C e , may be extracted from (6.14.16) to be

L.Z/
e e
D g sin W e ˛e ˇe 5 /e C ˛ 1 5 / C ˛ e 1 5 /e Z :
(6.14.39)
From (6.14.17), (6.14.18) this may be rewritten as

g
L.Z/
e e
D 5
e .gV gA /e C L L C eL eL Z ;
2 cos W
(6.14.40)
1 1
gV D 2 sin2 W ; gA D : (6.14.41)
2 2
On the other hand, (6.14.32), (6.14.33), (6.14.34) give
g p
D MZ . 2GF /1=2 ; (6.14.42)
2 cos W
which leads to
p
L.Z/
e e
D MZ . 2GF / 1=2 5
e .gV gA /e C L L C eL eL Z :
(6.14.43)
The scattering processes:̇ e Ce ! e Ce , e Ce ! e Ce from (6.14.16)

also involve the exchange of a W vector boson. From the just mentioned equation,
the interaction Lagrangian density responsible for this exchange is then given by
p p h i
L.W/
ee
D 2MW . 2GF /1=2 eL eL W C eL eL W : (6.14.44)
For low energy scattering of the above scattering processes in question, we may
replace the Z propagator by ˛ˇ =MZ2 , and the W propagator by ˛ˇ =MZ2 . Thus the
following effective interaction Lagrangian density emerges for corresponding to the
exchange of the vector bosons Z and W :
ˇ p

ˇ
L.Z/
e e
ˇ D 2 GF e .gV gA 5 /e L L C eL eL ; (6.14.45)
eff
ˇ p

.W/ ˇ
Le e ˇ D 2 2 GF eL eL eL eL ; (6.14.46)
eff
where in writing the last equation, we have used the classic Fierz identity (see, e.g.,
Appendix A of Chapter 8 in Vol. II), and the anti-commutativity of the fermion
fields, to write (see Problem 6.23)

eL eL eL eL D eL eL eL eL : (6.14.47)
Therefore the total effective interaction Lagrangian density for the processes
mentioned in the beginning of this subsection is given by
p

Le
I
eff D 2 GF e .gV gA 5 /e L L
p h i

C 2 GF e ŒgV C 1 ŒgA C 1 5 e eL eL ; (6.14.48)
where gV and gA are defined in (6.14.41).

Precise simultaneous measurements116 of the corresponding differential cross
sections of . e/, .Q e/, .e e/, .Q e e/, and a comparison with the ones dictated117
by the above expression in (6.14.48), allow the determination of the constants
gV ; gA ,118 and, in turn, give the value of sin2 W as recorded in (6.14.36).
6.14.3 Masses of the Neutrinos and the “Seesaw Mechanism”
There is an overwhelmingly large number of experiments119 which are in favor

of no-zero neutrino masses. In the original electroweak theory, no right-handed
neutrinos where involved and thus the neutrinos remain massless under spontaneous
symmetry breaking. By assuming their existence, the neutrinos may acquire masses
through the Higgs mechanism as the charged leptons (and quarks Sect. 6.15). This
method, however, fails to account for the expected small masses of the neutrinos.
Perhaps, a far more interesting way to account for the small masses, and which
is physically appealing, is through the so-called “seesaw mechanism”120 which we
now consider.
In the “seesaw mechanism”, one assumes that lepton numbers are violated
at some high-energy scales well beyond that of the electroweak energy scale.
A simple way to describe this is to assign to each charged lepton, in the process, a
116
Vilain et al. [119].
117
These are obtained in the same way, e.g, as the differential cross section in Sect. 5.9.3. See
’t Hooft [116].
118
The value of gA is consistent with the theoretical value of 1= 2.
119
See, e.g., Davis et al. [34], Hirata et al. [65], Fukuda et al. [50], Ambrosio et al. [7],
Athanassopoulos et al. [12], and Cleveland et al. [31]. The latter, in particular, is consistent with
neutrino oscillations experiments which have been acknowledged by the 2015 Nobel Prize.
120
See, e.g., Minkowski [90], Gell-Mann et al. [54], Yanagida [123], and Mohapatra and
Senjanović [92].
right-handed neutrino. The underlying mechanism leads to two sets of mas-

sive Majorana particles, each with three particles, with one set corresponding to
presently observed neutrinos, and another set consisting of neutrinos with very large
masses beyond the presently available energies. We will see how these latter heavy
neutrinos are to be handled in making explicit computations in field theory. We
recall that a Majorana field is invariant under charge conjugation.
Recall that the charge conjugate of a spinor field , is defined by C D C >
(see Sect. 2.5), where121 C is the charge conjugation matrix. For a left-handed
spinor L , and a right-handed one R , we have the following relations
C > C >
L;R DC L=R ; L=R D Œ. L=R/ C 1 : (6.14.49)
It is important to note that
1
5 1
5 0 1˙5 1 ˙ 5 > >
L=R D 0 D . / D D ;
2 2 2 2
(6.14.50)
on account that f 5 ; 0 g D 0, with 5 ; 0 as symmetric and Hermitian. That is,

C C
L , . L / are right-handed, while R , . R / are left-handed. These immediately
imply that
L L D 0; R R D 0: (6.14.51)
From the properties of the charge conjugation matrix, and the definition of charge
conjugate spinor fields, we also have, in particular,
C C
1R 2L D. 2L / . 1R / : (6.14.52)
For simplicity of the presentation, we consider a one generation neutrino with

field L , i.e., a one flavor neutrino. We introduce a right-handed neutrino, denoted
by NR , and define, in turn, the following mass-Lagrangian density
1
Lmass D mD NR L C L NR MR NR NR /C C .NR /C NR ; (6.14.53)
2
where we will eventually take the mass parameters to satisfy the constraint MR
mD . The reason for the 1=2 factor in the second term is to avoid double counting,
since we may, e.g., rewrite NRC NR D .NR /> C 1 NR and the same field NR
appears twice which should be taken into account in the application of the Euler-
Lagrange equation. The first term in (6.14.53) is referred to as a Dirac mass
term, and is assumed to have been generated, via spontaneous symmetry breaking,
121
Remember : C D C > D C , .C /2 D I, ŒC ; 5 D 0.
through the Higgs mechanism, or simply as just an electroweak symmetry breaking

term introduced into the theory. The second one is of different nature, referred to as
a Majorana mass term, and with lepton number C1 for a neutrino, we note that this
term does not conserve lepton number, and violates it by two units.
Using the basic properties in (6.14.49), (6.14.50), (6.14.51), (6.14.52), the mass-

Lagrangian density in (6.14.53) may be conveniently rewritten as C > D . /C
1n h > > i h > io

Lmass D mD C ..NR /C LCC L .NR /C CMR C .NR /C .NR /C C h:c:;
2
(6.14.54)
where we have suitably grouped together terms in the first expression f : g. The
above may be rewritten in simple matrix form as
>
1
L 0 mD L
Lmass D C C h:c: (6.14.55)
2 N CR mD MR N CR
It is easy to check that with

p p
i cos # i sin # m1 m2
VD ; sin # D p ; cos # D p ;
sin # cos # m1 C m2 m1 C m2
(6.14.56)
q q
1 1
m1 D 4m2D C MR2 MR ; m2 D 4m2D C MR2 C MR ; (6.14.57)
2 2
the following transformation emerges

> m1 0 0 mD
V V D : (6.14.58)
0 m2 mD MR
Hence upon setting

L 1 L
V D ; (6.14.59)
.NR /C 2 L
>

we obtain C D . /C
h i
1 L > m1 0 L
Lmass D C V V C h:c:
2 .NR /C 0 m2 .NR /C
m1
m2

D .1L /C 1 L C 1 L .1 L /C C .2 L /C 2 L C 2 L .2 L /C :
2 2
(6.14.60)
Using the basic properties in (6.14.51), and defining the Majorana fields i D i L C
.i L /C , i D 1; 2, the following mass-Lagrangian density emerges
1
Lmass D m1 1 1 C m2 2 2 : (6.14.61)
2
Thus, we have obtained two massive Majorana fields, and for MR mD , we have
from (6.14.57), one of mass m1 ' m2D =MR which would be small for sufficiently
large MR , and one with corresponding large mass m2 ' MR .122 Note that cos # '
1; sin # ' 0. The 2 does not participate in the dynamics, it is referred to as a
“sterile” particle and decouples from the theory.
It is interesting that one may achieve small neutrino masses by the “seesaw
mechanism”. The analysis involving more than one generation of neutrino follows
the same pattern as above.
6.14.4 Neutrino Oscillations: An Interlude
There are by now several experiments123 which provide clear evidence of neutrino
oscillations,124 implying that transitions occur between different neutrino flavors,
and that a given neutrino flavor may disappear, in a neutrino beam, and another
neutrino flavor appears in an oscillatory manner depending on the distances traveled
by the neutrinos. This is understood to be caused, in general, by nonzero neutrino
masses and by so-called neutrino mixing. By neutrino mixing it is meant that the
neutrino fields may be expressed as linear combinations of neutrinos, which we
denote by .1 ; 2 ; : : :/, with specific masses .m1 ; m2 ; : : :/.
The underlying aspect of the above physical phenomenon is that the transition
probability between, say, two neutrino flavors .x / ; .y / : Prob Œ .x / ! .y / , is of
an oscillatory nature, depending on the distance traveled by the .x / neutrino, and
a value of one taken by this probability, at a given distance, for example, specifies
the “disappearance” of the .x / neutrino.125 Considering ultra-relativistic massive
neutrinos, that is, with masses negligible in comparison to their momenta, we will
see, by a simple derivation that for transitions involving two kinds of neutrinos,
the amplitude of oscillations of the just mentioned probability is determined by the
122
MR is often taken as large as 1015 GeV according to a scale set up by GUT.
123
See, e.g., Anselmann et al. [10], Cleveland et al. [31], Fukuda et al. [50, 51], Hampel et al. [62],
and Altmann et al. [6]. It is rather interesting to point out that the theory of neutrino oscillations
was written up in this book much earlier than the 2015 Nobel Prize in Physics was announced.
124
Pontecorvo [99, 100] and Maki et al. [81]. See also Pontecorvo [101] and Gribov and Pontecorvo
[59].
125
That is, reduction in beam intensity of .x / neutrinos, and hence a reduction of neutrinos, in
general, will be observed if one is “looking” only for this flavor.
mixing of the neutrinos, while its oscillatory nature depends on the distance traveled
by the .x / neutrino and its energy.
In a theoretical description of a neutrino oscillating experiment, we have causally,
an emission source of an ultra-relativistic neutrino flavor, followed by its propaga-
tion, described by the propagator taking into account of neutrino flavor mixing, and
finally a detection source of a given flavor type : emitter propagation detection.
Before discussing the above process, let us quickly review the way one may
describe the combined process of emission, propagation and detection of a single
ultra-relativistic massive spin 1/2 particle of mass m. We know from § 3.3.1, that
given two sources > .x/, < .x 0 / with x 0 > x 0 0 for all x 0 ; x 0 0 , for which, the
respective sources do not vanish, this amplitude is given by
Z hZ
.dp/ ip .xx 0 / . p C m/ i < 0
i W12 D i .dx/.dx 0 / > .x/ e .x /
.2/4 p 2 C m2 i
Z hZ i
d3 p ip .xx 0 /
D .dx/.dx 0 / .i/ > .x/ e . p C m/ .i/ < .x 0 /;
2p0 .2/3
(6.14.62)
p
p0 D p2 C m2 . For an ultra-relativistic particle,
p m2
p0 D p2 C m2 ' j p j C ; (6.14.63)
2j p j
p C 0 p 0 C m
' N C 0 ; N D p=j p j: (6.14.64)
p0
We also note that

1
N C 0 D .1 N/: (6.14.65)
N
p
Accordingly by defining u.p/ D j p j .1 N/,
Z Z
0 0
.p; x 0 / D d3 x u.p/ .x/ ei.p xj p jx / ; .p; x 0 / D d3 x .x/u.p/ ei.p xj p jx / ;
(6.14.66)
the amplitude i W12 becomes simply

Z
d3 p 2 0 00

i W12 D 3
Œ dx 0 i > .p; x 0 / ei.m =2jpj/.x x / dx 0 0 i < .p; x 0 0 / :
2 j p j.2/
(6.14.67)
Here given that a particle was emitted by an emitter (source < ) and detected by
2 0 00
a detector (source > ), then ei.m =2jp j/.x x / denotes the amplitude that the
particle has moved from the emitter to the detector with probability one, indicating
the stability of the particle.
Now consider two spinor fields 1 ; 2 with corresponding masses m1 ; m2
respectively, with Lagrangian density
L D L0 C 1 1 C 1 1 C 2 2 C 2 2 ; (6.14.68)
where L0 are their free Lagrangian densities, and 1 ; 2 are external sources.
In experiments, involving neutrino oscillations, involving, for example, the
transition ! e the latter are considered to correspond to states as linear
combinations of states of definite masses m1 ; m2 . Accordingly, if we let the sources
of emission of be denoted by ./ , and that of the detection of e by .e/ ,
then these sources may be defined as linear combinations of the ones in (6.14.68)
corresponding to definite mass states, i.e., with normalized coefficients
.e/ D 1 cos C 2 sin ; ./ D 1 sin C 2 cos ;

or;
1 D .e/ cos ./ sin ; 2 D .e/ sin C ./ cos : (6.14.69)

Also by readily extracting the coefficient of Œ > <
.e/ ./ in
2 0 x 0 0 / 2 0 x 0 0 /
Œ >
1 e
i.m1 =2j p j/.x
1< C Œ >
2 e
i.m2 =2j p j/.x
2< ; / (6.14.70)
where the sources denoted by are defined as in (6.14.66), we may immediately

write down the amplitude of occurrence for the process : “emission of by the
emitter, transition amplitude . ! e / amplitude of detection of e ” as
follows:
Z
d3 p
dx 0 dx 0 0 i > 0 < 00
.e/ .p; x / A. ! e / i ./ .p; x /; (6.14.71)
2 j p j.2/3

where > <
.e/ is “switched on” after ./ is “switched off”,
m2 2 0 m1 2 0
A. ! e / D sin cos expŒi .x x 0 0 / expŒi .x x 0 0 / :
2j p j 2j p j
(6.14.72)
Upon taking the absolute value squared of the above gives the probability
m2 m2 L
Prob Œ ! e D sin2 .2/ sin2 1 2
; (6.14.73)
4 E
where we have taken E ' j p j, and, due to the fact that the ultra-relativistic particle
travels near with the speed of light x= t ' 1, identified .x 0 x 0 0 / ' L , with
L denoting the distance traveled by the ultra-relativistic neutrino.126 It is customary
to define the oscillation length
.4E /
L D ; (6.14.74)
.m22 m21 /
and rewrite the probability in question as

L
Prob Œ ! e D sin2 .2/ sin2 : (6.14.75)
L
We note that the oscillations of the probability are governed by the distance L
traveled by the ultra-relativistic muon neutrino and its energy E , while the
amplitude sin2 .2/ of the oscillations is governed by the neutrino mixing. It is
important to note that the probability of no transition : Prob Œ ! , for example,
referred to as the survival probability of the neutrino , as obtained directly from

the coefficient of Œ > <
./ ./ in (6.14.70), with emission and detection sources
causally arranged, is given by

L
Prob Œ ! D 1 sin2 .2/ sin2 ; (6.14.76)
L
and coincides with that obtained from a completeness relation
Prob Œ ! D 1 Prob Œ ! e ; (6.14.77)
for consistency, as is easily checked.

Note that for L L , the neutrino has not traveled long enough to
“disappear”, and its survival probability Prob Œ ! will attain a value not
far from one.
6.15 Electroweak Theory II: Incorporation of Quarks;

Anomalies and Renormalizability
This section deals with the incorporation of quarks in the electroweak theory,
generating what has been called the standard model based on the product symmetry
groups SU.3/SU.2/U.1/. The electroweak theory as well as the standard model
126
Needless to say, in a relativistic setting, one should not confuse distance traveled with length
measurement, in general, with the latter, for example, involved in coincident time measurements
of the end points, say, of a moving ruler.
6.15 Electroweak Theory II: Incorporation of Quarks; Anomalies and. . . 489
involve axial vector currents and the problem of anomalies studied in Sects. 3.9 and
3.10 arises and their absence is critical for the renormalizability of the standard
model. Based on the work in Sects. 3.9 and 3.10, we show that there are actually no
anomalies in the standard model.
6.15.1 Quarks and the Electroweak Theory
In order to take into account strangeness changing interactions such as: 0 ! p C

, and, e.g., suppress a decay (strangeness changing) process such as: 0 !
p C e C Q e in comparison to the classic beta decay process: n ! p C e C Q e ,
and take into account other processes involving flavor changing quarks, the three
generations of quarks, in analogy to the leptons’ generations, but now guided by the
just mentioned processes, are taken of the form

u c t
q1 L D ; q 2L D ; q 3L D ;
d0 L s0 L b0 L (6.15.1)
uR ; dR cR ; sR tR ; bR
with left-handed quarks placed in weak iso-doublets, while the right-handed ones in
weak iso-singlets. Also .d0 ; s0 ; b0 / are, on general grounds, linear combinations127
(allowing necessarily cross generations transitions), of the form
0 1 0 10 1
d0 Vud Vus Vub d
@ s 0 A D @ Vcd Vcs Vcb A @ s A ; (6.15.2)
b0 Vtd Vts Vtb b
where the matrix VCKM D ŒVab is referred to as the Cabibbo-Kobayashi-Maskawa

matrix . Vud is a measure of the coupling of u to d, and so on. The electroweak
quantum numbers of the quarks are given in Table 6.3.
Table 6.3 Electroweak quantum numbers of the quarks in the Standard Model: q> D u; c; t;
q< D d; s; b. The antiquarks have the same I, but opposite Q; I3 (for I3 ¤ 0), and Y D 2.QI3 /.
Each of the quarks comes in three different colors
Particle I I3 Q Y
.q> L ; q<L / 1=2 .C1=2; 1=2/ .C2=3; 1=3/ 1=3
.q>R ; q<R / 0 .0; 0/ .C2=3; 1=3/ .C4=3; 2=3/
127
Cabibbo [27], Glashow et al. [56], and Kobayashi and Maskawa [71].
The hadronic current may be then written as

0 1 0 01
u d
@ A 5 @ 0A
Jh D c .1 / s ; (6.15.3)
t b0
while an electromagnetic one is given by
2 X 1 X
J em D q> q> q< q< : (6.15.4)
3 q 3 q
> <
A standard representation of the unitary Cabibbo-Kobayashi-Maskawa matrix is

given by
0 1
c12 c13 s12 c13 s13 eiı
VCKM D @ s12 c23 c12 s23 s13 eiı c12 c23 s12 s23 s13 eiı s23 c13 A ; (6.15.5)
s12 s23 c12 c23 s13 eiı c12 s23 s12 c23 s13 eiı c23 c13
parametrized by three mixing angles 12 ; 13 ; 23 , and a phase ı,128 where sij D
sin ij ; cij D cos ij . Experimentally, approximate values of the magnitudes of the
matrix elements are129
!
ˇ 0:974 0:225 0:004
VCKM ˇ jVab j
D 0:225 0:973 0:041 ; (6.15.6)
0:009 0:040 0:999
and the matrix deviates slightly from the identity.

In particular, in using the approximations 13 D 0; 23 D 0; ı D 0, the CKM
matrix reduces to
0 1
cos 12 sin 12 0
@ sin 12 cos 12 0A; (6.15.7)
0 0 1
making contact with earlier investigation of Cabibbo involving fermions mixing

between the first two generations, with the single angle 12 C ; sin C ' 0:225,
referred to as the Cabibbo angle. This, in particular, leads to d0 D d cos C Cs sin C ,
and it was originally proposed to suppress strangeness changing processes in
comparison to strangeness non-changing ones by choosing sin C small enough.
Examples of a strangeness changing-processes and a strangeness non-changing
128
This phase is responsible for flavor changing CP-violating processes in the standard model.
129
Beringer et al. [17].
one, involving now the W boson, are .s W C ; u/ and .d W C ; u/, respectively. By

intoducing, in the process, the Cabibbo angle, there was no need to introduce a new
coupling to deal with strangeness changing processes and the Fermi coupling GF
was taken as a universal weak interaction coupling.
To generate masses for the quarks, we may introduce three quark fields Qi > ; i D
1; 2; 3, of charges 2=3, and three quark fields Qi < ; i D 1; 2; 3 of charges 1=3.
We will then eventually obtain an expression of the fields associated with the quark
fields u; d; c; s; t; b in terms of these fields. In a unitary gauge, a Yukawa interaction
may be defined by the interaction of quarks with the Higgs field as follows
p
X Qi > . C v/= 2
LYuk D ij Qj > R
i;j
Qi < L
0
X
Qi > 0 p
ij Qj < R C h:c:; (6.15.8)
Qi < L
. C v/= 2
i;j
wit i; j D 1; 2; 3, leading to a mass-Lagrangian density having the structure
Lmass
Q
D Q> L M1 Q> R Q< L M2 Q< R C h:c:; (6.15.9)
written in matrix form. We may now introduce unitary matrices U1 ; V1 ; U2 ; V2 ,

which diagonalize the matrices M1 ; M2 as follows

U1 M1 V1 D m.1/ ; U2 M2 V2 D m.2/ ; (6.15.10)
where m.1/ ; m.2/ are diagonal and real with non-negative entries. Accordingly, by
setting U1 Q> L D Q 0> L , V1 Q> R D Q 0> R , U2 Q< L D Q 0< L , V2 Q< R D Q <
0
R , we
obtain

Lmass
Q
D Q 0> L C Q 0> R m.1/ .Q >
0 0
L C Q > R/

Q 0< L C Q 0< R m.2/ Q 0< L C Q 0< R ; (6.15.11)
where due to the facts that m.1/ ; m.2/ are diagonal, we have used the properties
R=L R=L D 0, for a spinor . With . Q 0> L C Q 0> R / interpreted as a column
matrix with three elements, the latter may be identified with the quarks u; c; t.
Similarly, d; s; b are identified with . Q 0< L C Q 0< R /.
6.15.2 Anomalies and Renormalizability
We first consider the standard model with massless particles. Gauge invariance
is a key criterion for the renormalizability of the theory. In particular, gauge
invariance requires that currents coupled to gauge fields must be conserved and
hence implying the absence of anomalies of the type encountered in Sect. 3.10. The
renormalizability of Yang-Mills field theories was considered in Sect. 6.3.3. Once
we include fermions, anomalies associated with the divergence of currents may
appear of the type investigated in Sect. 3.10. We will see that there are no such
anomalies in the standard model.
To the above end, we first recall that the weak hypercharges assigned to the left-
and right-handed fermions involved in the theory are as given Table 6.4 below as
obtained from Tables 6.2 and 6.3.
We refer to the diagrams in Fig. 6.21 below, to investigate the nature of the
anomalies that one may encounter in the theory, in general.
Here we recall that in the electroweak theory, the SU.2/ gauge bosons couple
only to the left-handed currents and the U(1) gauge boson couples to the left- and
right-handed currents but with different U(1) charges. On the other hand in the
color SU.3/ theory, the gluons couple to the left- and right-handed fermions in
the same way.
For further analysis, recall the following properties: f ; 5 g D 0, S. p/ D
p=p2 , fS. p/; 5 g D 0, . 5 /2 D 1, TrŒ 5 D 0, TrŒ 1 D 4. With the group
structure SU.3/ SU.2/L U.1/Y , with the standard model in mind, we define the
following set of matrices:
n o
fa g t1 ; t2 ; t3 ; 1 ; 2 ; 3 ; I ; (6.15.12)
where the identity I corresponds to the generator Y, and the ai ; bi are some
numerical constants. Recall that the generators of SU.3/, and of SU.2/ in Fig. 6.21
are given, respectively, by ta D a =2, with the a denoting the Gell-Mann matrices
in (6.1.46), and a D a =2. The a matrices commute with the gamma ones. In
reference to the above figure, diagram (b) is obtained from diagram (a) by carrying
out the elementary substitutions
h i
Tr 1 5 1 fb ; c gS 2 S 3 S )
h i
Tr 1 5 .a1 C b1 5 /a fb ; c gS 2 .a2 C b2 5 /S 3 .a3 C b3 5 /S :
(6.15.13)
Tr Œfc ; b g ) 4 Œa1 .a2 a3 C b2 b3 / C b1 .a2 b3 C a3 b2 / Tr Œa fc ; b g:
(6.15.14)
where we have used, in the process, that 1 5 .a1 C b1 5 / D .a1 b1 5 / 1 5 .
Table 6.4 Weak hypercharge and charge classifications of the left-handed and right-handed
Fermi particles in the first generation in the standard model. The corresponding quantum numbers
of the other generations are identical. We use the definition Y D 2.Q I3 /
Particles L eL uL dL eR uR dR
Y 1 1 1=3 1=3 2 4=3 2=3
Q 0 1 2=3 1=3 1 2=3 1=3
(a)
γ μ1 γ 5 1
(b)
γ μ1 γ 5(a1 +b1γ 5)Γa
S S S S
γ μ3 Γc S γ μ2 Γb γ μ3 (a3 +b3 γ 5 )Γc S γ μ2 (a2 +b2γ 5)Γb
Fig. 6.21 (a) Vertices in the triangle associated with the anomaly that may be extracted from the
investigations in Sect. 3.10. (b) The vertices that may appear in the standard model, where the
matrices a are defined below.
Y, for the weak hypercharges, may be also written in matrix form. Using the
basic properties, TrŒta D 0, TrŒa D 0, we have
Tr Œ ta fb ; c g D Tr Œta Tr Œ fb ; c g D 0;
Tr Œ a ftb ; tc g D Tr Œa Tr Œ ftb ; tc g D 0;
1
Tr Œ a ; fb ; c g D ıbc Tr Œa D 0: (6.15.15)
2
and we immediately infer the absence of anomalies for vertices corresponding to
• SU.3/ SU.2/ SU.2/,
• SU.2/ SU.3/ SU.3/,
• SU.2/ SU.2/ SU.2/,
• Two U.1/s.
For
• SU.3/ SU.3/ SU.3/, a1 D 0; b1 D 1; a2 D a3 D 1, b2 D b3 D 0, imply
from (6.15.14) the absence of an anomaly, where we have also used the fact
that . 5 /2 D I. Refer to Fig. 6.21b.
For
• One U.1/ : For b1 D y" ; a1 D y""; " D
1; a2 D a3 D 1; b2 D b3 D 1:
where the coefficients y" will be specified below by using Table 6.4, and " D
1
correspond to left-handed and right-handed particles, respectively. First note that
h i
Œa1 .a2 a3 C b2 b3 / C b1 .a2 b3 C a3 b2 / D y" ".1 C 1/ C .1 1/ D 2 y" ." 1/;
(6.15.16)
and hence only " D 1 might contribute. From Table 6.4, we have y1 D
1; 1, and three times for the 3 colors of quarks: y1 D 1=3; 1=3. Hence the
anomaly is proportional to
X h 1 1 i
4 y1 4 1 1 C 3 C D 0: (6.15.17)
y1
3 3
The reader should realize how critical it is that the quarks come in 3 colors for
P The above is also true for b2 D b3 D 0, as well for
the absence of the anomaly.
b2 D 1; b3 D 0, since yC1 yC1 is zero as well.
For
• U.1/U.1/U.1/. With the weak isospin written in a matrix form, the anomaly
is proportional to
1 3 h 4 3 2 3 i
Tr Œ YL3 Tr Œ YR3 D 2.1/3 C 3 2 1.2/3 C 3 C3 D 0;
3 3 3
(6.15.18)
and we note again that they vanish only if the quarks come in three colors.
This provides further evidence of three color charges needed for quarks for
consistency of the theory.
Therefore the standard model is anomaly-free, and lies on the fact that the theory
involves several fermions whose separate contributions to the anomalies cancel out,
and that quarks and leptons appear in Nature in equal numbers organize themselves
into successive generations.
Gauge invariance of transition probabilities implies that one may work with
the gauge fields in any gauge one pleases. To prove the renormalizability of
spontaneously broken electroweak theory, where the vector particles, in particular,
become massive, this gauge freedom may be invoked to choose a gauge where the
propagator of a massive vector particle of mass M, may be expressed as

k k k2 .1= 1/ 1
C .k; I M/ D 2 ; (6.15.19)
k .k2 = C M 2 / .k2 C M 2 /
as introduced by ’t Hooft [114, 115]. In particular for finite parameter ,

C .k; I M/ has the very welcome high-energy behavior O.1=k2 / for k2 ! 1.
Thus leading a class of gauges, denoted by R , for which renormalizability of the
theory may be inferred as in QED. Thus renormalizability only results from the

underlying gauge invariance. On the other hand for ! 1, C .k; I M/ simply
becomes the propagator of a massive vector particle, and one may formally infer,
by invoking gauge invariance, the renormalizability of the spontaneously broken
gauge theory with massive vector particles.
6.16 Grand Unification 495
6.16 Grand Unification
A theory based on the product of the three symmetry groups SU.3/ SU.2/ U.1/
necessarily involves three different couplings, with SU.3/ corresponding to the
color group of quarks, and SU.2/ U.1/ corresponding to the electroweak theory.
We may consider a larger group, such as SU.5/ which contains the product of
the above three group just mentioned, and would necessarily involve one single
coupling. The covariant derivative would then, in general, take the form
24
X 1
r D @ i gunif Va Ta ; Tr ŒTa Tb D ıab : (6.16.1)
aD1
2
Here we recall that SU.5/ stands for all 5 5 unitary matrices of determinant one.
Thus it involves 52 1 D 24 generators and 24 gauge fields (see Sect. 6.1). We will
explicitly obtain the generators corresponding to the fundamental representation and
see, in the process, how the product group, given above, is contained in the SU.5/
group. The only quarks and leptons that arise in this unifying scheme are those of
the electroweak theory.
Eleven of the twenty four generators denoted by T1 ; : : : ; T11 for the lowest
dimensional representation of SU.5/ are given at the end of this section, cor-
responding simply to SU.3/, SU.2/ transformations, consisting of 8 plus 3
generators, respectively.
To construct the remaining generators, one may proceed as follows. It is easy to
find the lowest five dimensional (the fundamental) representation represented by
0 1
a1
B a2 C
B C
B C
D B a3 C : (6.16.2)
B 4C
@a A
a5
Now consider, in turn, infinitesimal SU.3/; SU.2/ transformations via the opera-
tors
8
X X
U D1Ci ı
a Ta C i ı"a Ta ; (6.16.3)
aD1 aD9;10;11
applied to the above column vector. From the expression of the infinitesimal
transformation, we note that, .a1 ; a2 ; a3 / form an SU.3/ triplet, while it is
unaffected by the SU.2/ transformation, and are thus singlets under the latter group.
Referring to quarks from the first generation, we see from Table 6.3 in the last
section that the right-handed quarks uR ; dR form the SU.2/ singlets. On the other
hand, .a4 ; a5 /, are unaffected by the SU.3/ are thus singlets under this group, but
form a doublet under SU.2/. From Table 6.2 in Sect. 6.14, one may be tempted
to choose the SU.2/ doublet to be .eL ; eL /. Since the generators for SU.N/,
have to be traceless, the eigenvalues of the charge operator Q add up to zero. This
will the case if .a1 ; a2 ; a3 / are chosen to be the right-handed d quarks fields, of
different colors, of charges 1=3, and, in turn, the SU.2/ doublet is taken to be the
following as charge conjugate of fields .eL /C ; .e L /C , where we recall that the
charge conjugate of a left-handed field is right-handed.130 This incidently shows the
consistency of the fractional charge 1=3 assigned to the (colored) d quarks.131
It is important to realize that a left handed state may be equivalently described in
terms of its right-handed charge conjugate state. That is, we may write
2 3
0 1 Q Y=2
d1 6 1=3 1=3 7
B d C 6 7
B 2 C 6 7
B C 6 1=3 1=3 7
D B d3 C ; 6 7; (6.16.4)
B C 6 1=3 1=3 7
@ .eL /C A 6 7
4 1 1=2 5
.eL /C R 0 1=2
denoted by Œ5, in a group theoretical notation. The minus sign multiplying .eL /C
above arises in the following manner. In order to assign the correct weak isospin
quantum numbers to .eL /C and .e /C , a rotation of 180o was made as follows :

2 =2 .eL /C .eL /C
ei D :
.eL /C .eL /C
The 15 left-handed fermions, per generation, with the first generation consisting
of the fermions:
.d1 ; d2 ; d3 ; e ; e; u1 ; u2 ; u3 ; u1 ; u2 ; u3 ; d1 ; d2 ; d3 ; eC /L ;
are fitted into the fundamental conjugate132 representation as follows

.d1 ; d2 ; d3 ; e ; e /L , denoted by Œ 5 , and the remaining 10 left-handed ones in
a next higher dimensional representation of SU.5/ denoted by Œ10 . The latter,
consisting of 10 independent components, is represented by an anti-symmetric
5 5 matrix.
By referring to (6.16.4), the expression of this representation state allows us
to introduce the weak hypercharge as a generator of SU.5/, restricted to the
above state , given by T12 j D Const: Y D Const: diag Œ 2=3; 2=3; 2=3; 1; 1 .
130
See below Eq. (6.14.50).
131
Assignment of fractional charges to the quarks consistently as encountered here, lead eventually
to such equalities as of the magnitudes of the charge of the electron and that of the proton.
132
Conjugate representations are discussed in Sect. 6.1.
The normalization factor is obtained from Tr Œ .T12 /2 D 1=2. This leads to the
expression for T12 matrix given at the end of the section.
Therefore, the remaining generators are of the form
!
1 0 .U i /
TiC12 D ; i D 1; : : : ; 12; (6.16.5)
2 .U i / 0
where the matrices .U i /32 may be readily defined with each having only one of
their elements taking the value 1 or i, with the remaining elements equal to zero,
as follows, going for example c.w. :

001 00 i 000 000
U1 D ; U2 D ; U3 D ; U4 D ;
000 000 001 00 i

5 000 6 000 11 010 12 0 i 0
U D ; U D ; ::: ; U D ; U D ;
010 0 i 0 000 000
(6.16.6)
leading finally to the 24 generators as given at the end of this section.

One may now compare the SU.5/ covariant derivative of
24
X
r D @ i gunif Va Ta ; (6.16.7)
aD1
with the corresponding SU.3/ SU.2/ U.1/ expression written conveniently as
8 r
X 5 X
0
r D @ i gs Aa Ta C i g B T12 i g Wi T8Ci ; (6.16.8)
aD1
3 iD1;2;3
with Va; D Aa ; a D 1; : : : ; 8 denoting the gluons, and V.8Ci/ D Wi ; i D 1; 2; 3

from which the electroweak vector bosons, in (6.14.12), (6.14.14) are defined. From
the comparison of the above two expressions, we may then infer that the equalities
of the SU.3/ SU.2/ U.1/ .couplings/2 :
5 02 2
g 2s .M 2 / D g .M / D g 2 .M 2 /; (6.16.9)
3
at a unifying SU.5/ energy scale M which will be determined. This will be investi-
gated below. Before doing this, we note that the SU.5/ symmetry group introduces
twelve additional vector bosons, which consist of 6 vector bosons, P denoted by,
X1 ; X2 ; X3 Y1 ; Y2 ; Y3 , and their 6 anti-particles, via the term 24
aD13 a Ta in the
V
covariant derivative. This term may be written as

0 1
XN 1 YN 1
B 0 XN 2 YN 2 C
B C
B XN 3 YN 3 C; (6.16.10)
@X X X A
1 2 3
0
Y1 Y2 Y3
where the four-vector indices of the boson fields are suppressed.

Application of (6.14.15), (6.16.9): tan W D g 0 =g, gives
ˇ
sin2 W ˇM 2 D 3=8; (6.16.11)
at an energy scale M of the order of, say, of the mass of the X vector boson.
For energies much less than MX , specified by a parameter , we may use the
renormalization group equations, which relate the magnitudes of the couplings2
g2s ; g2 ; g 0 2 at different energy scales, to continue to lower energy scales available at
present experiments. For the SU.N/ gauge group, we have from (6.7.17), (6.7.14),
d 1 .11N 4ng / g2s

2 D ˇ0 ; ˛s ; (6.16.12)
d2 ˛ s .2 / 12 4
expressed in terms of the number of generations of quarks ng .

To obtain, the corresponding expression for U.1/, Table 6.5, in this regard, spells
out the weak hypercharges and charges of all the particles in the first generation and
is useful. Accordingly for ng generations, with 15 fermions per generation, we have:
X Yi 2
1 2 2

1 2 2 2 1 2

10
D ng 2 C 1 C3 2 C C D ng :
i
2 2 6 3 3 3
(6.16.13)
We may infer from Sect. 5.19.1 in QED, which is simply based on the U.1/ group,
that
d 1 1 X 2
2 2 2
D Qi ; (6.16.14)
d ˛ QED . / 3
Table 6.5 Weak hypercharge and charge classifications of all the left-handed quarks and Leptons
in the first generation. The corresponding quantum numbers of the other generations are identical
Particles e eC u uC d dC
Y=2 1=2 1=2 1 1=6 2=3 1=6 1=3
Q 0 1 1 2=3 2=3 1=3 1=3
to lowest order in the charges, where the Qi (in units of jej) are now the charges of
quarks, taking into account color multiplicity, and of the charges of the charged
leptons, and also the contributions of the underlying number of generations P is2
understood to be taken. For the present
P U.1/ Y group, we have to replace i Qi
in (6.16.14) by the expression .1=2/ i .Yi =2/2 , with an overall 1=2 factor because
the expression in (6.16.13) takes into account the anti-particles as well.
The renormalization group equations then become133 dealing with the incorpo-
ration of supersymmetry in the standard model.
d 1 1 g2s
2 DC .33 4 ng /; .˛ s D for SU.3//; (6.16.15)
d2 ˛ s .2 / 12 4
d 1 1 g2
2 DC .22 4 ng /; .˛ D for SU.2//; (6.16.16)
d2 ˛.2 / 12 4
d 1 1 .20 ng/ g02
2 2 0 2
D ; .˛ 0 D for U.1//; (6.16.17)
d ˛ . / 12 3 4
where in writing the right-hand sides of (6.16.15), (6.16.16), we have used the
expression for ˇ 0 in (6.16.14), with N D 3; N D 2, respectively.
In particular, these equations lead to
d 1 1 11
2 D ; (6.16.18)
d2 ˛ s .2 / ˛.2 / 12
d 1 3 1 22
2 2 D : (6.16.19)
d ˛.2 / 5 ˛ 0 .2 / 12
Upon subtracting Eq. (6.16.19) from twice of Eq. (6.16.18), integrating with
respect to 2 from M 2Z , corresponding to the mass MZ of the Z vector boson
as a reference point, to M, the unifying energy specified by a mass parameter M,
and finally using the unifying boundary conditions (6.16.9), we simply obtain
2 3 3
C D 0: (6.16.20)
˛s .M 2Z / ˛.M 2Z / 5 ˛ 0 .M 2Z /
From the first two defining equations in (6.14.17) we may also write
1 cos2 W 1 sin2 W
D ; D : (6.16.21)
˛0 ˛ em ˛ ˛ em
133
In the beta functions, on the right-hand sides of (6.16.16), (6.16.17), we have neglected tiny
contributions due to the Higgs boson field. These will be spelled out in Sect. 3.13 of Volume II.
The latter two equations then lead to

ˇ 1 5 ˛ em .M 2Z /
ˇ
sin2 W ˇ D C : (6.16.22)
M 2Z 6 9 ˛ s .M 2Z /
Experimentally,134 ˛ˇs .MZ 2 / D 0:1184 ˙ 0:0007, 1=˛em .MZ 2 / D 127:916 ˙ 0:015.

These give, sin2 W ˇM 2 ' 0:21, which compares well with the experimental result
Z
in (6.14.36).
On the other hand, the second defining equation in (6.16.21), (6.16.18),
and (6.16.22) give upon integration,
2
M 2 8 ˛ em .M 2Z /
ln D 1 ; (6.16.23)
M 2Z 11˛em .M 2Z / 3 ˛ s .M 2Z /
where we have used the boundary condition ˛ s .M 2 / D ˛.M 2 /. From the

experimental values of ˛em .M 2Z /, ˛ s .M 2Z / quoted above, we may determines the
scale M ' 1:1 1015 GeV.
Finally (6.16.15) leads, upon integration with respect to 2 , the solution
2
1 1 .33 4 ng / M
D 2
C ln ; (6.16.24)
2
˛ s .M / ˛ s .M Z / 12 M 2Z
which from (6.16.23) and the experimental input at M 2Z , give ˛ s .M 2 / '

0:024, and thus from (6.16.9) we also have ˛.M 2 / ' 0:024, ˛ 0 .M 2 / '
0:0144. Of particular interest is the value of ˛ em .M 2 / which is obtained
from (6.16.11), (6.16.21) to be ˛ em .M 2 / ' .3=8/ 0:024 D 0:009.
To assess the departure from the strict equality of the couplings at a unifying
scale, as given in (6.16.9), i.e., ˛ s .M 2 / D .5=3/˛ 0 .M 2 / D ˛.M 2 /, one may define
the critical parameter135
˛ 1 .M 2Z / ˛s1 .M 2Z /
D ; (6.16.25)
.3=5/˛ 01 .M 2Z / ˛ 1 .M 2Z /
which taking into account the experimental values quoted above, gives exp '
0:74. On the other hand upon integrating (6.16.15), (6.16.16), (6.16.17) from
134
Beringer et al. [17].
135
Peskin [98].
2 D M 2Z to 2 D M 2 , denoting ˇs D .33 4ng /=12, ˇ D .22 4 ng /=12,

.3=5/ˇ 0 D 4ng =12, and assuming the strict unifying equality stated above at
2 D M 2 , we have for for the right-hand side of in (6.16.25),
ˇs ˇ 11=12
theor D 0
D D 0:5; (6.16.26)
ˇ .3=5/ˇ 22=12
and the departure is significant. A supersymmetric treatment of this grand unifying

theorem provides, in turn, excellent agreement for this critical parameter between
theory and experiment as we will see in Sect. 3.13 of Volume II.
The new bosons X and Y are referred to as leptoquarks as they carry both color
and lepton numbers. The mere fact that the SU.5/ representations contain both
quarks and leptons, implies that baryon number is not conserved, and the proton can
decay. For example, u C u ! X, where, in turn, X may convert to eC and dC , i.e.,
the charge of X is 4=3. Similarly, u C d ! Y, where, in turn, Y may convert to eC
and uC , i.e., the charge of Y is 1=3.
An order of magnitude of the lifetime of the proton is readily obtained from
the expression of the muon lifetime as extracted from Problem 6.24 which is
given by D 192 3 =GF 2 M 5 , and from (6.14.33), G 2F D g4 MW 4
=32. By
formally replacing M by M p , and M W , g , in the expression for G F2 , by
2 2 2 2
M 2 MX2 , g2unif , respectively, one obtains the rough expression for the proton
lifetime p M 4 =g4unif M 5p . A more precise calculation gives136 p!eC 0
10x years, 28:5 x 31:5 which is too low in comparison to the experimental
lower bound > 2 1033 years at 90% confidence limit.137 Increasing the scale
of unification138 increases the theoretical estimate of the proton lifetime as it is
proportional to M 4 . Consistency with such a large experimental lower bound
for the proton decay is certainly a basic problem in developing grand unified
theories.
Although this unifying theory, predicts, for example, the key estimate for the
value of sin2 W .MZ2 / close enough to the experimental value, a discrepancy
between theory and experiment that arises means that the strict equality of the
fundamental constants as given in (6.16.9), does not hold rigorously. This situation
is much improved in a supersymmetric setting.
136
See Langacker [75].
137
See Miura [91] and Olive et al. [94].
138
In a supersymmetric extension M 1016 GeV, see § 2.13 in Volume II.
The fundamental energy scale that arises in the standard model from the vacuum
expectation value of the Higgs boson sets the scale for the masses in the theory,
such as for the masses of the vector bosons.139 The enormous mass difference
between the electroweak scale 102 GeV and all the way up to the grand unified
scale 1015 GeV, referred to as the hierarchy problem, should be of concern, as it
is such an enormous shift in energy scale providing no hint as to what happens
to the physics in between. The radiative corrections to the mass squared of the
Higgs boson, as a scalar particle, are quadratically divergent,140 and hence with
a cut-off of the order of the grand unified scale, or of the Planck energy scale
1019 GeV, at which gravity is expected to be significant, it is necessary that the
bare mass squared of the Higgs boson be correspondingly large to cancel such a
quadratic dependence on the enormous cut-off to finally obtain a finite physical
mass141 for the Higgs boson of the order of the low electroweak energy scale, in
comparison, and is highly unnatural.142 This unnatural cancelation of enormously
large numbers has been termed a facet of the hierarchy problem. Supersymmetry
is of significance in dealing with the hierarchy problem, as in supersymmetric
field theories, in general, cancelations of such large quadratic corrections, a priori,
occur between graphs involving particles and their supersymmetric counterparts in
a supersymmetric version of a non-supersymmetric field theory, up to possibly of
divergences of logarithmic type which are tolerable, thus protecting a scalar particle
from acquiring a large bare mass.
Finally, we note that in the above unifying theory, the quarks and leptons are
those of the electroweak theory involving no right-handed neutrinos to generate
masses to the presently known neutrinos by the “seesaw mechanism” as discussed
in Sect. 6.14.3. The larger group SO.10/, which has been also much studied over
years as a grand unified theory, accommodates a right-handed neutrino and the
“seesaw mechanism” may be then applied. Several other grand unified theories have
been also proposed in recent years and we refer the reader to the vast literature
on the subject.143 The remarkable simplicity of the SU.5/ grand unifying scheme
should be noted, and that it provides the very first steps for the development of
more complicated grand unifying schemes that may meet the challenge set up by
experiments.
139
We recall how accurate were the masses of the massive vector bosons obtained theoretically
directly at the tree level via the vacuum expectation value of the Higgs field given through
Eq. (6.14.34) which, in turn, is expressed simply in terms of the Fermi coupling.
140
Veltman [118].
141
Aad et al. [2] and Chatrchyan et al. [30].
142
As mentioned above, the question, in turn, arises as to what amounts for such an enormous
difference between the energy scale of grand unification and the energy scale that characterizes the
standard model.
143
See, e.g., Beringer et al. [17] and references therein.
In closing this section, we provide the fundamental representation of the SU.5/

group generators:
0 1
1 0 1 0 0
T1;:::;8 D @ 1;:::;8 A ; T9;10;11 D ;
2 0 0 2 0 1;2;3
0 1
r 2=3 0 0
1 3 B 0 2=3 0 0 C
T12 D @ 0 0 2=3 A;
2 5 1 0
0 0 1
0 0 0
1 0 0 0
1
1B 0 0 0
C 1B 0 0 0
C
T13 D @ 1 0 A ; T14 D @ i 0 A;
2 0 0 1 2 0 0 i
0 0 0
0 0 0 0
0
0 0 0
1 0 0 0
1
1B 0 0 0
C 1B 0 0 0
C
T15 D @ 0 1 A; T16 D @ 0 i A;
2 0 0 0 2 000
0 0 1
0 0 0 i
0
0 0 0
1 0 0 0
1
1B 0 0 1
C 1B 0 0 i
C
T17 D @ 0 0 A ; T18 D @ 0 0 A;
2 0 0 0 2 0 0 0
0 1 0
0 0 i 0
0
0 0 1
1 0 0 i
1
1B 0 0 0
C 1B 0 0 0
C
T19 D @ 0 0 A; T20 D @ 0 0 A;
2 0 0 0 2 000
1 0 0
0 i 0 0
0
0 1 0
1 0 i 0
1
1B 0 0 0
C 1B 0 0 0
C
T21 D @ 0 0 A ; T22 D @ 0 0 A;
2 1 0 0 2 i 0 0
0 0 0
0 0 0 0
0
0 0 0
1 0 0 0
1
1B 0 1 0
C 1B 0 i 0
C
T23 D @ 0 0 A ; T24 D @ 0 0 A;
2 0 1 0 2 0 i 0
0 0 0
0 0 0 0
0
where the Gell-Mann matrices 1 ; : : : ; 8 are defined in Sect. 6.1.
Problems

6.1 Find the infinitesimal gauge transformation of the non-abelian gauge fields Aa
via the unitary operators V.x/ D expŒi go c .x/tc .

6.2 Show that Œr ; r cb D go fcab Ga , where .r /ac D ıac @ C go fabc Ab .

6.3 Show that rab rbc Gc D 0.
6.4 Show that
Z
i h • • i ˇˇ
exp .dx/ ˘ax ı @ Aa .x/ 'a .x/ ˇ
2 •'c .x/ •'c .x/ 'D0
Z

i
D exp .dx/@ Aa .x/ @ Aa .x/ ;
2
up to an unimportant multiplicative constant.

6.5 Obtain the leading contribution of the gauge transformed non-abelian gauge
field as a function of the original field if the former satisfies the Coulomb gauge.
6.6 Derive Rosenbluth formula in (6.4.1), for elastic electron-proton scattering,
for a finite extended description of the proton, via a single photon exchange, by
neglecting the electron mass.
6.7 Derive the differential cross section for e eC ! qNq, via a single photon
exchange, in the CM, by neglecting the mass of the electron (positron) as well as of
the quark (antiquark). Also neglect radiative corrections.
6.8 Evaluate the following infrared regularized integrals, where the dimensionality
of spacetime is replaced by
D 0 D 4 C ı; with ı > 0; p21 D 0 D p 22 ; 2p1p2 Q2 ;
Z
i.2D 0 /ı=2 1 1ı=2 2 2ı .1 2ı
0
dD k 1
(i) D :
.2/D 0 . p1 k/2 . p2 k/2 k2 .4/D 0 =2 Q2 .1 C ı/
Problems 505
Z
i.2D 0 /ı=2 1 1ı=2
0
dD k k
(ii) D
.2/D 0 . p1 k/2 . p2 k/2 k2 .4/D 0 =2 Q2
ı
2 1 ı2 .1 C 2ı

p1 C p2 :
.2 C ı/
6.9 Show by dimensional regularization of the following integral in its ultraviolet

and infrared divergent parts, that it may written as:
Z ˇ
.dk/ k k ˇ i 2D 2" 2" 2 1 2"
ˇ D
.2/4 . p1 k/2 . p2 k/2 k2 Reg .4/D=2 Q2 2 .3 "/
1 2 0 ı
i 2 ı
ı
2 2C 2
ı
C D
1 . p1 p1 Cp2 p2 /
.4/D 0 =2 Q2 Q2 2 .3 C ı/
2
ı

1C 2
C . p1 p2 Cp1 p2 / ; where D D 4 "; D 0 D 4 C ı; " > 0; ı > 0:
.3 C ı/
6.10 Derive the explicit general structure of the infrared regularized function
.1/ .2/
hIR .2D 0 =Q2 ; ı/ in (6.6.25), hIR .2D 0 =Q2 ; ı/ in (6.6.21), hIR .2D 0 =Q2 ; ı/ in (6.6.24).
6.11 Derive the equality in (6.9.6).
6.12 Prove the expressions in (6.9.13), (6.9.14), (6.9.16).

6.13 Show that the structure tensor Wi contribution due to the i th parton may be
written as in (6.10.1).
R 3 0 2

1
6.14 Show that 2d pp0 0 ı .4/ .P C Q p 0 / D 2 jPQj ı C 2QPQ , where p 0 2 D
p2 D 2 M 2 .
6.15 Show that if quarks (anti-quarks) are of spin 0, then W1 D 0. This leads to
the results that F T D 0, for the transverse structure function in (6.9.17)/(6.10.5),
and R ! 1, defined in (6.9.19).
6.16 Show that if one takes into account only the three basic quarks (u; d; d),
associated with the neutron, into consideration, and assumes that they occur
R1
with the same frequency, then this leads to the relation 0 dx F2en .x/ D
R1 P
.2=9/ 0 dx i x fi .x/, consistent with approximate experimental values
Z 1 Z 1 X
dx F2en .x/ 0:12; dx x fi .x/ 0:54:
0 0 i
This, again, provides evidence of fractional charges of the quarks.

6.17 Derive the rigorous bounds in (6.11.16), (6.11.17), (6.11.18) for the moments
of the splitting functions in (6.11.11), (6.11.12), (6.11.13), (6.11.14), for n D
3; 4; : : : < 1.
6.18 Establish the relation: Q2 d=.dQ2/ D Œ˛s .Q2 /=.2 / d=.d/ in (6.11.20).
6.19 (i) Verify equation (6.11.32) which diagonalizes the 2 2 matrix on the
left-hand side of (6.11.32). For all n D 3; 4; : : : < 1,
(ii) Show that C C
n < n < 0, for the eigenvalues.
(iii) Derive the inequality:
p 11 nf 1
50 q
2 Œ 6#4n < < C
< C .98=135/ nf :
2 3 n
2 9
(iv) Show that C ˙

n An > 0, n An < 0. That is, n An ¤ 0.
6.20 Derive the relation in (6.12.8).

6.21 Establish the gauge transformation rules for the remaining three line-integrals
in Box 6.3 in Sect. 6.12.
6.22 Show by a combined SU.2/L U.1/Y transformation, as in (6.14.22),
(6.14.23), one may always find one to reduce the Higgs field doublet to the form
in (6.14.26) satisfying (6.14.27).
6.23 Show that Œ eL L Œ L eL D Œ eL eL Œ L L .
6.24 According to (6.14.1), the amplitude for muon decay: ! e C Q e C
may be taken in a tree approximation as
GF
A D p Œu .k2 / .1 5 /u . p/ Œue .k/ .1 5 /ve .k1 /:
2
By neglecting the masses of the electron and neutrinos in comparison to the mass
M of the muon, and working in the rest frame of the muon, show that its decay
rate is given by D G2F M 5 =192 3 .
References
1. Abe, K. et al. (1999). Measurement of R D L =T for :03 < x < 0:1 and fit to world data.
2. Aad, G. et al. (2012). Observation of a new particle in the search for the Standard Model
Higgs Boson with the ATLAS detector at the LHC. Physics Letters, B716, 1–29.
3. Airapetian, A. et al. (2002). Measurement of R D L =T in deep ineastic scattering.
arXiv:hep-ex/0210068.
4. Ali, A. et al. (1980). A QCD analysis of the high energy eC e data from PETRA. Physics
Letters, 93B, 155–160.
5. Altarelli, G., & Parisi, G. (1977). Asymptotic freedom in parton language. Nuclear Physics,
B126, 298–318.
6. Altmann, M. et al. (2005). Complete results for five years of GNO solar neutrino observations.
References 507
7. Ambrosio, M. et al. (1998). Measurement of the atmospheric neutrino induced upgoing muon
flux using MACRO. Physics Letters, B434, 451–457.
8. Ammar,
p R. et al. (1998). Measurement of the total cross section for eC e ! Hadrons at
s D 10:52 GeV. Physical Review, D57, 1350–1358.
9. Andivahis, L. et al. (1994). Measurements of the electric and magnetic form factors of the
proton from Q2 D 1.75 to 8.83 .GeV/c/2 . Physical Review, D50, 5491–5517.
10. Anselmann, P. et al. (1992). Solar neutrinos observed by GALLEX at Gram Sasso. Physics
Letters, B285, 376–389.
11. Appelquist, T., & Georgi, H. (1973). eC e annihilation in gauge theories of strong interac-
tions. Physical Review, D8, 4000–4002.
12. Athanassopoulos, C. et al. (1998). Results on ! e neutrino oscillations from the LSND
experiment. Physical Review Letters, 81, 1774–1777.
13. Baikov, P. A., Chetyrkin, K. G., & Khun, J. H. (2008). Order ˛s4 QCD corrections to Z and
decays. Physical Review Letters, 101, 012002.
14. Bartel, W. et al. (1980). Observations of planar three-jet events in e eC annihilation and
evidence for gluon bremsstrahlung. Physics Letters, 91B, 142–147.
15. Becchi, C., Rouet, A., & Stora, R. (1975). Renormalization of the Abelian Higgs-Kibble
model. Communications in Mathematical Physics, 42, 127–162.
16. Berger, C. et al. (1979). Evidence for gluon bremsstrahlung in eC e annihilation at high
energies. Physics Letters, 86B, 418–425.
18. Bjorken, J. D. (1969). Asymptotic sum rules at infinite momentum. Physical Review, 179,
1547–1553.
19. Bjorken, J. D., & Bodsky, S. J. (1970). Statistical model for electron-positron annihilation
into hadrons. Physical Review, D1, 1416–1420.
20. Bjorken, J. D., & Drell, S. D. (1964). Relativistic quantum mechanics. New York/San
Francisco/London: McGraw-Hill.
21. Bjorken, J. D., & Drell, S. D. (1965). Relativistic quantum fields. New York/San Fran-
cisco/London: McGraw-Hill.
22. Bjorken, J. D., & Pachos, E. A. (1969). Inelastic electron-proton and y-proton scattering and
the structure of the nucleon. Physical Review, 185, 1975–1982.
23. Bloom, E. D. et al. (1969). High- Energy Inelastic e-p Scattering at 6ı and 10ı . Physical
24. Bodek, A. et al. (1979). Experimental studies of the neutron and proton electromagnetic
structure functions. Physical Review, D20, 1471–1552.
25. Brandelik, R. et al. (1979). Evidence for planar events in eC e annihilation at high energies.
Physics Letters, 86B, 243–249.
26. Brown, L. S., & Weisberger, W. I. (1979). Remarks on the static potential in quantum
chromodynamics. Physical Review, D20, 3239–3245.
27. Cabibbo, N. (1963). Unitary symmetry and leptonic decays. Physical Review Letters, 10,
531–533.
28. Callan, C. G., & Gross, D. J. (1969). High-energy electroproduction and the constitution of
the electric current. Physical Review Letters, 22, 156–159.
29. Caswell, W. E. (1974). Asymptotic behavior of Non-Abelian gauge theories to two-loop order.
30. Chatrchyan, S. et al. (2012). Observation of a new boson at mass 125 GeV with the CMS
Experiment at LHC. Physics Letters, B716, 30–61.
31. Cleveland, B. T. et al. (1998). Measurement of the solar electron neutrino flux with the
homestake chlorine detector. Astrophysics Journal, 496, 505–526.
32. Creutz, M. (1983). Quarks, gluons and lattices. Cambridge: Cambridge University Press.
33. Danby, G. et al. (1962). Observation of high-energy neutrino reactions and the existence of
two kinds of neutrinos. Physical Review Letters, 9, 36–44.
34. Davis, R. et al. (1968). Search for neutrinos from the sun. Physical Review Letters, 20, 1205–
1209.
35. Eidelman, S. et al. (2004). Particle data group. Physics Letters, B592, 1.
36. Erler, J. (1999). Calculation of the QED coupling ˛.M
O Z / in the modified minimal subtraction
scheme. Physical Review D, 59, 054008, 1–7.
37. Faddeev, L. D., & Popov, V. N. (1967). Feynman diagrams for the Yang-Mills field. Physics
Letters, B25, 29–30.
38. Fermi, E. (1934a). Tentativo di una teoria dei raggi ˇ. Nuovo Cimento, 11, 1–19.
39. Fermi, E. (1934b). Versuch einer Theorie der ˇ - Strahlen. Zeitschrift fur Physik, 88, 161–171.
40. Feynman, R. P. (1963). Quantum theory of gravitation. Acta Physica Polonica, 24, 697–722.
41. Feynman, R. P. (1969a). The behavior of hadron collisions at extreme energies. In Proceed-
ings of the 3rd Topical Conference on High Energy Collisions, Stony Brook. New York:
Gordon & Breach.
42. Feynman, R. P. (1969b). Very high-energy collisions of hadrons. Physical Review Letters, 23,
1415–1417.
43. Feynman, R. P., & Gell-Mann, M. (1958). Theory of fermi interaction. Physical Review, 109,
193–198.
44. Field, R. D. (1989). Applications of perturbative QCD. Redwood City: Addison-Welry.
45. Field, R. D., & Feynman, R. P. (1977). Quark elastic scattering as a source of high-transverse-
momentum mesons. Physical Review, D15, 2590–2616.
46. Field, R. D., & Feynman, R. P. (1978). A parametrization of the properties of quark jets.
47. Friedman, J. I., & Kendall, H. W. (1972). Deep inelastic electron scattering. Annual Review
of Nuclear and Particle Science, 22, 203–254.
48. Fritzsch, H., & Gell-Mann, M. (1972). Quatks and what else? In J. D. Jackson & A. Roberts
(Eds.), Proceedings of the XVI International Conference on High Energy Physics (Vol. 2).
Chicago: Chicago University Press.
49. Fritzsch, H., Gell-Mann, M., & Leutwyler, H. (1973). Advantages of the color octet gluon.
50. Fukuda, Y. et al. (1998). Evidence for oscillations of atmospheric neutrinos. Physical Review
Letters, 81, 1562–1567.
51. Fukuda, Y. et al. (2002). Determination of solar neutrino oscillation parameters using 1496
days of Super-Kamiokande-I data. Physics Letters, B539, 179–187.
52. Gell-Mann, M. (1964). A schematic model of baryons and mesons. Physics Letters, 8, 214–
215.
53. Gell-Mann, M. (1972). Quarks. Acta Physica Austriaca Supplement IX, 9, 733–761.
54. Gell-Mann, M., Raymond, P., & Slansky, R. (1979). Complex spinors and unified theories. In
P. van Nieuwenhuizen & D. Z. Friedman (Eds.), Supergravity. Amsterdam: North-Holland.
55. Glashow, S. L. (1961). Partial symmetries of weak interactions. Nuclear Physics, 22, 579–
588.
56. Glashow, S. L., Iliopoulos, J., & Maiani, L. (1970). Weak interactions with Lepton-Hadron
symmetry. Physical Review, D2, 1285–1292.
57. Goldhaber, M., Grodzins, L., & Sunyar, A. W. (1958). Helicity of the neutrinos. Physical
Review, 109, 1015–1017.
58. Greenberg, O. W. (1964). Spin and unitary spin independence in a paraquark model of baryons
and mesons. Physical Review Letters, 13, 598–602.
59. Gribov, V. N., & Pontecorvo, B. (1969). Neutrino astronomy and lepton charge. Physics
Letters, B616, 174–190.
60. Guth, A. H. (1980). Existence proof of a nonconfining phase in four-dimensional U(1) lattice
theory. Physical Review, D21, 2291–2307.
61. Halzen, F., & Martin, A. D. (1984). Quarks and leptons: An introductory course in modern
particle physics. New York: Wiley.
62. Hampel, W. et al. (1999). GALLEX solar neutrino observations: Results for GALLEX IV.
63. Han, M. Y., & Nambu, Y. (1965). Three-triplet model with double SU(3) symmetry. Physical
Review, 139, B1006–B1010.
References 509
64. Hanson, G. et al. (1975). Evidence for jet structures in hadron production by eC e
annihilation. Physical Review Letters, 35, 1609–1612.
65. Hirata, K. S. et al. (1996). Solar neutrino data covering solar cycle 22. Physical Review
Letters, 77, 1683–1686.
66. Hoyer, P. et al. (1979). Quantum chromodynamics and jets in eC e . Nuclear Physics, B161,
349–372.
67. Itzykson, C., & Zuber, J.-B. (1980). Quantum field theory. New York/Toronto: McGraw-Hill.
68. Joglekar, S. D., & Lee, B. W. (1976). General theory of renormalization of gauge invariant
operators. Annals of Physics, 97, 160–215.
69. Jones, D. R. T. (1974). Two-loop diagrams in Yang-Mills theory. Nuclear Physics, B75, 531–
538.
70. Jost, R., & Luttinger, J. M. (1950). Vacuum polarization and e4 charge renormalization for
electrons. Helvetica Physica Acta, 23, 201.
71. Kobayashi, M., & Maskawa, K. (1973). CP violation in the renormalizable theory of weak
interaction. Progress of Theoretical Physics, 49, 652–657.
72. Kogut, J. B. (1980). Progress in lattice theory. Physics Reports, 67, 67–102.
73. Kogut, J. B., Pearson, R. P., & Shigemitsu, J. (1981). The string tension, confinement and
roughening in SU(3) Hamiltonian lattice gauge theory. Physics Letters, 98B, 63–68.
74. Landau, L. D. (1957). On the Conservation Laws for Weak Interactions. Nuclear Physics, 3,
127–131.
75. Langacker, P. (1981). Grand unified theories and proton decay. Physics Reports, 72, 185–385.
76. Lautrup, B., & Nauenberg, M. (1980). Phase transition in four-dimensional compact QED.
Physics Letters, 95B, 63–66.
77. Lee, B. (1976). In R. Balian & J. Zinn-Justin (Eds.), Methods in field theory. Amsterdam:
North Holland.
78. Lee, T. D., & Yang, C. N. (1956). Question of parity conservation in weak interactions.
Physical Review, 104, 254–258. See also ibid., 106, 1671 (1957).
79. Lepage, G. P., & Brodsky, S. J. (1979). Exclusive processes in quantum chromodynamics: The
form factors of baryons at large momentum transfer. Physical Review Letters, 43, 545–549.
80. Lepage, G. P., & Brodsky, S. J. (1980). Exclusive processes in perturbative quantum
chromodynamics. Physical Review, D22, 2157–2198.
81. Maki, Z., Nakagawa, M., & Sakata, S. (1962). Remarks on the unified model of elementary
particles. Progress of Theoretical Physics, 28, 870–880.
82. Manoukian, E. B. (1984a). Proof of the decoupling theorem of field theory in Minkowski
space. Journal of Mathematics and Physics, 25, 1519–1523.
D34, 3739–3749.
85. Manoukian, E. B. (1986b). Generalized conditions for the decoupling theorem of quantum
field theory in Minkowski space with particles of vanishing small masses. Journal of
Mathematics and Physics, 27, 1879–1882.
86. Manoukian, E. B. (1987). Functional differential equations for gauge theories. Physical
Review, D35, 2047–2048.
88. Martin, P. C., & Glashow, S. L. (2008). Julian Schwinger 1918-1994: A biographical memoir.
National Academy of Sciences, Washington, DC, Copyright 2008.
89. Mele, S. (2006). Measurements of the running of the electromagnetic coupling at LEP. In
XXVI Physics in Collision, 6–9 July 2006, Búzios, Rio de Janeiro.
90. Minkowski, P. (1977). ! ey at a rate of one out of 109 muon decays? Physics Letters, B67,
421–428.
91. Miura, M. (2010). Search for nucleon decays in Super-Kamiokande. ICHEP, Paris, Ses-
sion, 10, 408–412.
92. Mohapatra, R. N., & Senjanović, G. (1980). Neutrino mass and spontaneous parity noncon-
servation. Physical Review Letters, 44, 912–915.
93. Mohr, P. J. et al. (2008). CODATA recommended values of the fundamental physical
constants. Reviews of Modern Physics, 80, 633–730.
94. Olive, K. A. et al. (2014). Particle data group. Chinese Physics C, 38, 090001.
95. Panofski, W. (1968). In Proceedings of the 14th International Conference on High Energy
Physics (pp. 23–39). CERN Scientific Information, Vienna.
96. Pauli, W. (1957). On the earlier and more recent history of the neutrino. In: Winter, K. (Ed.),
Neutrino physics. Cambridge: Cambridge University Press (1991).
97. Perrin, F. (1933). Possibilité d’Emission de Particules Neutres de Masse Nulle dans les
Radioactivités ˇ. Comptes Rendus Academie des Sciences Paris, 197, 1625.
98. Peskin, M. (1997). Beyond the Standard Model. In N. Ellis & M. Neubert (Eds.), European
School of High-Energy Physics 1996, CERN-97-03, Genève.
99. Pontecorvo, B. (1957). Mesonium and Anti-Mesonium. Soviet Physics JETP, 33, 549–551.
Original Russian Version: Zhurnal Experimental’noi i Teoreticheskoi Fiziki, 6, 429–431
(1957).
100. Pontecorvo, B. (1958). Inverse beta processes and nonconservation of lepton charge. Soviet
Physics JETP, 34, 247–248. Original Russian Version: Zhurnal Experimental’noi i Teoretich-
eskoi Fiziki, 7, 172–173 (1957).
101. Pontecorvo, B. (1968). Neutrino experiments and the problem of conservation of Leptonic
charge. Soviet Physics JETP, 53, 1717–1725. Original Russian Version: Zhurnal Experimen-
tal’noi i Teoreticheskoi Fiziki, 26, 984–988 (1967).
102. Puckett, A. J. R. et al. (2012). Final analysis of proton form factor ratio at Q2 D 4:0, 4.8 and
5.6 GeV2 . Physical Review, C85, 045203.
103. Punjabi, V. et al. (2005). Proton elastic From factor ratios to Q2 D 3; 5 GeV2 by polarization
transfer. Physical Review, C71, 055202.
104. Reines, F., & Cowan, C. L. (1956). The neutrino. Nature, 178, 446–449.
105. Riordan, E. M. et al. (1974). Extraction of R D L =T from deep inelastic ep and ed cross
sections. Physical Review Letters, 33, 561–564.
106. Salam, A. (1957). Parity consevation and a two-component theory of the neutrino. Nuovo
Cimento, 5, 299–301.
107. Salam, A. (1968). Weak and electromagnetic interactions. In N. Svartholm (Ed.), Elementary
Particle Theory, Proceedings of the 8th Nobel Symposium, Almqvist and Wiksell, Stockholm.
108. Salam, A., & Ward, J. (1964). Electromagnetic and weak interactions. Physics Letters, 13,
168–170.
109. Schwinger, J. (1957). A theory of the fundamental interactions. Annals of Physics, 2, 407–
434.
110. Slavnov, A. A. (1972). Ward identities in gauge theories. Theoretical and Mathematical
Physics, 10, 152–160. English Translation: Theoretical and Mathematical Physics, 10, 99–
108 (1972).
111. Sterman, G., & Weinberg, S. (1977). Jets from quantum chromodynamics. Physical Review
Letters, 39, 1436–1439.
112. Stevenson, P. M. (1978). Comments on the Sterman-Weinberg jet formula. Physics Letters,
78B, 451–454.
113. Sudarshan, E. C. G., & Marshak, R. (1958). Chirality invariance and the universal fermi
interaction. Physical Review, 109, 1860–1862.
114. ’t Hooft, G. (1971a). Renormalizable of massless Yang-Mills fields. Nuclear Physics, B33,
173–199.
115. ’t Hooft, G. (1971b). Renormalizable Lagrangians for massive Yang-Mills fields. Nuclear
Physics, B35, 167–188.
116. ’t Hooft, G. (1971c). Prediction for neutrino-electron scattering cross sections in weinberg’s
model of electroweak interaction. Physics Letters, B37, 195–196.
117. Taylor, J. C. (1971). Ward identities and charge renormalization. Nuclear Physics, B33, 436–
444.
Recommended Reading 511
118. Veltman, M. J. G. (1981). The infrared-ultraviolet connection. Acta Physica Polonica, B12,
437.
119. Vilain, P. et al. (1994). Precision Measurement of Electroweak Parameters from the Scattering
of Muon-Neutrinos on Electrons. Physics Letters B, 335, 246–252.
120. Weinberg, S. (1967). A Model of Leptons. Physical Review Letters, 19, 1264–1266.
121. Weinberg, S. (1996). The Quantum Theory of Fields, II: Modern Applications. Cambridge
University Press, Cambridge.
122. Wu, C. S. et al. (1957). Experimental tests of parity conservation in beta decay. Physical
Review, 105, 1413–1415.
123. Yanagida, Y. (1980). Horizontal symmetry and masses of neutrinos. Progress of Theoretical
Physics, 64, 1103–1105.
124. Zee, A. (1973). Electron-positron annihilation in stagnant field theories. Physical Review, D8,
4038–4041.
125. Zweig, G. (1964). An SU3 Model for strong interaction symmetry and its breaking. CERN
Preprint, TH-401, 1–24.
Recommended Reading
Press.
3. Manoukian, E. B. (1981). Generalized decoupling theorem in quantum field theory. Journal of
4. Manoukian, E. B. (1984). Proof of the decoupling theorem in Minkowski space. Journal of
5. Manoukian, E. B. (1986a). Generalized conditions for the decoupling theorem of quantum field
theory in Minkowski space with particles of vanishingly small masses. Journal of Mathematics
and Physics, 27, 1879–1882.
6. Manoukian, E. B. (1986b). Action principle and quantization of gauge fields. Physical Review,
D34, 3739–3749.
7. Manoukian, E. B. (1987). Functional differential equations for gauge theories. Physical Review,
D35, 2047–2048.
8. Olive, K. A. et al. (2014). Particle Data Group. Chinese Physics C, 38, 090001.
9. Ross, G. G. (1985). Grand Unified Theories. Reading: Benjamin/Cummings Publishing.
10. Weinberg. S. (1996). The Quantum Theory of Fields. II: Modern Applications. Cambridge:
Cambridge University Press.
11. Yongram, N., Manoukian, E. B., & Siranan, S. (2006). Polarization correlations in muon pair
production in the electroweak model. Modern Physics Letters A, 21, 979–984.
General Appendices
Appendix I
The Dirac Formalism
For the convenience of the reader we gather here some important equations dealing
with the Dirac formalism for the description of spin 1/ 2 particles. For derivations
and detailed presentation, we refer the reader to Chapter 16 of Manoukian [6].
The Dirac equation is given by

@
Cm D 0; f ; g D 2 ; Œ D diagŒ1; 1; 1; 1 D Œ :
i
(I.1)
h i
@
eA .x/ Cm .x/ D 0; (I.2)
i

from which one obtains the equation . D 0/
h i
@ C C >
C eA .x/ Cm .x/ D 0; .x/ D C .x/; C D i 2 0;
i
(I.3)
thus introducing, in turn, the charge conjugation matrix C , and the charge conjugate
spinor C .
Under a homogeneous Lorentz transformation which may include a 3D rotation
(see (2.2.1), (2.2.3), (2.2.6))
x 0 D x ; @ D @ 0 ; @ 0 D @ : (I.4)
the Dirac equation reads

h @ 0 i
0
C m K .x/ D 0; K .x/ D .x 0 /; (I.5)
i

DOI 10.1007/978-3-319-30939-2
516 I The Dirac Formalism
where the matrix K satisfies the relations
K 0 K D 0; D K 1 K: (I.6)
For infinitesimal transformations, we recall from (2.2.18), that ' ı C

ı! , ı! D ı! , and we may set K ' I C .i=2/ ı! S , where S is to

be determined. By substituting this expression in the last equation in (I.6) gives
i
Œ S ; D i ; with solution S D Œ ; : (I.7)
4
Some of the properties of the gamma matrices, based on their anti-commutations
relations in (I.1) are given in Box I.1.
Box I.1: Some properties of the gamma matrices
1
D I C Œ ; ; Tr Œ D 0;
2
0 2 2
D I; i D I; i D 1; 2; 3:
D 4 I;

D 2 ;

D 4 ;

D 2 ;
h h ii
; ; D 4 ;
Tr Œ D 4 ;

Tr ˛ ˇ D 4 ˛ˇ ˛ ˇ C ˛ ˇ ;

Tr 5 ˛ ˇ D 4 i "˛ˇ ;
"˛ˇ totally anti-symmetric with "0123 D C 1:

Tr Œ odd number of ’s D 0:
5 D i 0 1 2 3 ; Tr Œ 5 D 0; Tr Œ 5 D 0;
5 2
D I; f 5 ; g D 0;
2
2
a D I a2 .a0 /2 ; a D I a2 ;

a D a1 ; a2 ; a3 ; a0 D a0 ; ai D ai ; i D 1; 2; 3:
I The Dirac Formalism 517
The Dirac representation of the matrices, in particular, is defined by

I 0 0 0 I
0D ; D ; 5 i 0 1 2 3 D ; (I.8)
0 I 0 I 0
and 1 ; 2 ; 3 are the Pauli matrices

1 0 1 2 0 i 3 1 0
D ; D ; D ; (I.9)
1 0 i 0 0 1
satisfying, the relations f i ; j g D 2 ı ij . p

In the momentum description, on the mass shell p0 D p2 C m2 , where m is
the mass of a particle, we have two sets of solutions: u.p; /, v.p; /, with D ˙1
specifying spin states, satisfying the equations
. p C m/u.p; / D 0; u.p; /. p C m/ D 0; (I.10)

. p C m/v.p; / D 0; v.p; /. p C m/ D 0; (I.11)
and u.p; / may be taken as

s 0 1

p0CmB C
u.p; / D @ A; D ˙ 1; (I.12)
2m p

p 0 Cm

involving two normalized two component spinors satisfying 0 D ı 0 , which,
for an arbitrary unit vector N D .cos sin ; sin sin ; cos /, may be taken as
0 i=2 1 0 i=2 1
e cos 2 e sin 2
CN D @ A ; N D @ A; (I.13)
i=2 i=2
e sin 2 e cos 2
N ˙N D ˙ ˙N ; i 2 ˙N

D

N : (I.14)
We note that the spin matrix is defined by

1 i 1 1 0
S D " ij k S j k D " ij k Œ j ; k ;
i
SD †D : (I.15)
2 8 2 2 0
In the Dirac representation, the charge conjugation matrix C D i 2 0 is

ˇ
2 0ˇ 0 i 2
C jD D i ˇ D ; (I.16)
D i 2 0
with the charge conjugate of u.p; / given by C u>.p; ˙1/ D ˙ 5 u.p;

1/.
Thus we may define the second set of spinors in the momentum description by
v.p; ˙1/ D 5 u.p;

1/; v.p; ˙1/ D u.p;
1/ 5 ; (I.17)
up to phase factors, with v.p; / as charge conjugate of u.p; /.

Under space reflection: x 0 D .x 0 ; x/ the Dirac equation reads
h @ 0 i

C m 0 .x/ D 0; 0
.x 0 / D 0 .x/; (I.18)
i
up to a phase factor for the latter. Hence under space reflection, u.p; / !
0 u.p; / D C u.p; /, v.p; / ! 0 v.p; / D v.p; /, having
opposite (intrinsic) parities.
We have the following normalization conditions
u.p; /u.p; 0 / D ı 0 ; v.p; / v.p; 0 / D ı 0 ; u.p; / v.p; 0 / D 0;

(I.19)
0
p
u .p; / u.p; 0 / D v .p; / v.p; 0 / D ı 0 ; u .p; / v.p; 0 / D 0;
m
(I.20)
and the completeness relations

X . p C m/
u.p; / u.p; / D PC . p/; (I.21)

2m
X . p C m/
v.p; / v.p; / D P . p/: (I.22)

2m
Note that for
N D p=jpj; (I.23)
I The Dirac Formalism 519
in (I.13),
s 0 1
˙
p0CmB C
u.p; ˙/ D @ A: (I.24)
2m
˙ p 0jpj
Cm ˙
For a massless Dirac particle, it is convenient to work in the chiral representation

(see (2.3.3)), and with the unit vector N chosen along the momentum of the
particle, as given in (I.23), we have spinors1

C IC5 C 0 I5 C
u.p; C1/ D D ; u.p; 1/ D D ;
0 2 2
(I.25)
u .p; ˙1/ u.p; ˙1/ D 1. We may conveniently write

˙ 0
u.˙p; ˙1/ D ; u.˙p;
1/ D ; (I.26)
0
where ˙ with ˙ means spin projection in the same direction as of the momentum,
while ˙ with
means in the opposite directions. With an appropriately chosen
phases, the latter are related by a parity transformation: 0 u.p; C1/ !
u.Cp; C1/, 0 u.Cp; 1/ ! u.p; 1/, with spin and momentum in opposite
directions to spin and momentum in the same direction, and vice versa. Accordingly,
if parity is not conserved, nature picks up only one of the helicities.
The interest in the equations in (I.25) is that .I C 5 /=2 projects out a state
corresponding to a right-handed particle, with spin along its momentum, while
.I 5 /=2 projects out a left-handed one, with spin in opposite direction to
its momentum. The corresponding particles are referred to, respectively, as right-
handed and left-handed. The situation may be demonstrated as shown in Fig. I.1
p p
Right-Handed Left-Handed
Fig. I.1 Diagrams defining the handedness of a particle, where p is its 3-momentum
1
For the relevant details, see Manoukian [6], p. 912.
The eigenvectors u.p; ˙1/ then satisfy the simultaneous eigenvalue equations:
0 p u.p; ˙1/ D jpj u.p; ˙1/; (I.27)

5 u.p; ˙1/ D ˙ u.p; ˙1/; (I.28)
S N u.p; ˙1/ D ˙ .1=2/ u.p; ˙1/; (I.29)
where f 5 ; S N; 0 pg is a commuting set of operators,2 and specify the state

of a particle. The eigenvalues of 5 are referred to as the chiralities (handedness)
of a particle, and S N .†=2/ N defines the helicity operator, with u.p; ˙/
corresponding to a massless particle with two-helicity states. As seen above, if parity
is not conserved a massless particle is to be considered to have only one helicity.
2
Note that 5 does not commute with 0 . p C m/ for m ¤ 0.
Appendix II
Doing Integrals in Field Theory
For a real, consider the following two integrals

Z 1 r Z 1 r
2 2
dy cos.a y / D ; dy sin.a y / D sgn a: (II.1)
1 2jaj 1 2jaj
p
Using the fact that sin =4 D cos =4 D 1= 2, they may be combined into an
exponential form as
Z 1 r h i
2
dy ei a y D exp i sgn a : (II.2)
1 jaj 4
In view of applications in n-dimensional spacetime, we apply this integral to a

typical expression of the denominator of a propagator, which may be first rewritten
as
Z 1 X
n1
1
D i ds exp Œis.k2 C M 2 i
/; k2 D ki2 .k0 /2 :
.k2 C M 2 i
/ 0 iD1
(II.3)
In n-dimensional spacetime, our metric is defined by Œ D diagŒ1; 1; : : : ; 1.
From (II.2) we obtain the following integrals , with s > 0,
Z 1
r
i
dk0 exp Œ i s .k0 /2 D e 4;
1 s
(II.4)
Z Z Z r
1 1 1 i n1
dk1 dk2 : : : dk n1 exp Œ i s .k1 /2 C C .kn1 /2 D e 4 ;
1 1 1 s
(II.5)

DOI 10.1007/978-3-319-30939-2
522 II Doing Integrals in Field Theory
and hence
Z 1 Z Z 1 1 1 n2 n2 X
n1
2 2
dk dk : : : dkn1 e i s k D
0 1
; k2 D .ki /2 .k0 /2 ; s > 0:
1 1 1 i s iD1
(II.6)
In the remaining part of this appendix, we work only with n D 4, i.e., in 4-
dimensional spacetime. We also suppress the i
factor in (II.3).
If we integrate the expression in (II.3) over k, we encounter a singularity in the
s-integral at s D 0 . We may, however, differentiate (II.3) twice with respect to M 2
to obtain
Z 1
1 i

D ds s2 exp is.k2 C M 2 / ; (II.7)
.k2 C M 2 /3 2 0
leading from (II.6) to the useful integral

Z
.dk/ i 2 1
D ; .dk/ D dk0 dk1 dk2 dk3 ; (II.8)
.k2 C M 2 /3 2 M2
R1
where we have used the fact that 0 ds exp is.M 2 i
/ D i=.M 2 /.
A further differentiation of (II.8) with respect to M 2 , for example, leads to
Z Z
.dk/ i 2 1 k2 .dk/ i 2 1
D ; D : (II.9)
.k2 C M 2 /4 6 .M 2 /2 .k2 C M 2 /4 3 M2
We note that quite generally one may write,

Z
k k
.dk/ D A ; (II.10)
.k2 C M 2 /4
Z
k1 k 2 k 3 k 4
.dk/ D B .1 2 3 4 C 1 3 2 4 C 1 4 2 3 /: (II.11)
.k2 C M 2 /5
To determine the coefficient A, for example, we may make a contraction over , ,

to obtain from the second equation in (II.9), A D i 2 =12M 2 , using the fact that
D 4. Similarly, contractions over 1 ; 2 , and 3 ; 4 , give 24B to the right-
hand side of (II.10), leading to B D i 2 =96M 2 . These values are worth recording
here:
i 2 i 2
A D ; B D : (II.12)
12M 2 96M 2
II Doing Integrals in Field Theory 523
Thus we are bound to encounter the following readily evaluated integrals:

Z
.k2 /m2 i 2 .m/ .n m/
.dk/ 2 2
D ; (II.13)
.k C M / n .M 2 /nm .n/
which obviously exist for n > m > 0, where .z/ is the gamma function.
Now we consider more complicated integrals involving an additional (external)
momentum. For convenience, we enumerate the type of integrals considered.
1.
Z
1 1
I. p 2 / D .dk/ 2 2 2 2
2 D 0; (II.14)
Œ.k p/ C M . p / Œk C M 2 . p 2 /2
where, as indicated, M 2 . p 2 / may be a function of p 2 . To show that this integral

is indeed zero, we take its derivative with respect to p . It is easily checked
that the derivative of the integrand is zero giving @I. p 2 /=@p D 0. With
the boundary condition that I.0/ D 0, gives I. p 2 / = 0. Here we note that if
we define the degree of divergence of an integral as the of power of k in the
numerator minus the power of k in the denominator plus four, then the degree
of divergence of the above integral restricted to the first term is zero. This then
allows one to make a shift of the integration variable k ! k C p in the first
part of the integral just mentioned leading to the net result zero for the integral
I. p 2 /. Of course such a shift of an integration variable is obviously valid if the
degree of divergence of an integral is negative.
The degree of divergence of the following integrals are negative, thus by
making a shift of the integration variable lead to the stated results by using, in
the process, (II.8):
Z Z
.dk/ .dk/ i 2
D D ; (II.15)
Œ .k p/2 C M 2 . p 2 / 3 Œ k2 2 2
C M .p /3 2M 2
Z Z
k .dk/ i 2 p
.dk/ D p D ; (II.16)
Œ .k p/2 C M 2 . p 2 / 3 Œ k2 C M 2 . p 2 / 3 2M 2
where in the latter equation, we have used the property that an integral which is
odd in k is zero. Another integral, where the shift of the integration variable in
the first term is obviously permissible is
Z
k k k k p p
.dk/ 2 2 2 3
2 2 2 3
2 D0:
Œ .k p/ C M . p / Œk C M .p / Œ k C M 2 . p 2 / 3
(II.17)
2. To obtain the following integral

Z
k k k 1 p C p C p
.dk/ 2 2 2 3

Œ .k p/ C M . p / 4 Œ k2 C M 2 . p 2 / 2
5 i 2 i 2
D . p C p C p / C p p p ; (II.18)
24 2M 2 . p 2 /
we denote the result of the integration by
C. p 2 / . p C p C p / C D. p 2 /p p p ; (II.19)
and determine C. p 2 / and D. p 2 /. To this end, note that the integral is zero for
p D 0 since the resulting integrand is an odd function of k. Secondly, note
that the degree of divergence of the integral restricted to the first term involving
k k k is positive and hence we cannot simply make a shift of the integration
variable k to k C p in evaluating this term. What we do instead is take the
derivative of the integral with respect to p as shown below.
The derivative of the first term in the integrand, for example, is

@M 2 . p 2 / k k k k k k k
.3/ 2p C C 6 :
@p Œ .k p/2 C M 2 . p 2 / 4 Œ .k p/2 C M 2 . p 2 / 4
The first term leads to an integral of degree of divergence 1, and the second to a
degree of divergence 0. Hence we can make a shift of the integration variable p to
p C k here. Continuing in this manner and using any of the integrations obtained
before (II.18), we obtain for the derivative of the integral on the left-hand side
of (II.18) explicitly
5 i 2 i 2 @M 2 . p 2 /
. C C / 2 2 2
p p p
24 2.M . p // @p
i 2
C . p p C p p C p p /: (II.20)
2M 2 . p 2 /
This is to be compared with the derivative of the expression in (II.19) with respect
to k ,
@C. p 2 /
. p C p C p / C C. p 2 / . C C /
@p
@D. p 2 /
C p p p C D. p 2 / . p p C p p C p p /; (II.21)
@p
from which we obtain
5 i 2 i 2
C. p 2 / D ; D. p 2 / D ; (II.22)
24 2M 2 . p 2 /
as given on the right-hand side of (II.18).

The following integrals is obtained in a similar manner
Z !
p p k k 1 p 2
.dp/ 2 2 2 2
2 2 2 2

Œ . p k/ C M .k / Œ p C M .k / 4 Œ p C M 2 .k2 / 2
2
5 i 2 i 2 2
D k k k : (II.23)
6 6
Here we have used p as the integration variable for a direct application of this in
Appendix A of Chap. 3. The following integral is also quite useful
Z !
k p i 2
.dk/ D p : (II.24)
Œ .k p/2 C M 2 . p 2 / 2 Œ k2 C M 2 . p 2 / 2 2
The reader is strongly encouraged to go through the above details and realize the
simplicity in evaluating these integrals by the method just developed.
Integrands appearing in (II.8) . . . (II.24) arise when one combines the product of
two (or more) factors such as
1 1
; (II.25)
.k p/ C M k C m2
2 2 2
into one factor in the following manner. To this end, for two c-numbers A; B,
consider the integral
Z 1
d 1 1 1
dx D : (II.26)
0 dx Œ Ax C B.1 x/ A B
By carrying out the differentiation with respect to x and by dividing the resulting
integral by .B A/ gives the useful formula
Z 1
1 1
D dx : (II.27)
AB 0 Œ Ax C B.1 x/ 2
This is referred to as a Feynman parameter representation of the product on the

left-hand side of (II.27).
As an interesting example for applying this formula and comparing it with

the method developed above for evaluating some integrals, we will evaluate the
following
Z
1 1
I. p 2 / D .dk/ 2 2
2 ; I. p 2 /jpD0 D 0: (II.28)
.k p/ C m k C m2
Let us apply the method developed earlier first to evaluate this integral by differen-
tiating it with respect to p . This gives
Z
@ p k
I. p 2 / D 2 .dk/ D i 2 p ;
@p Œ .k p/2 C m2 2 Œ .k p/2 C m2 2
(II.29)
i 2 2
I. p 2 / D p ; (II.30)
2
where we have used the fact that the degree of divergence of the integral in (II.29)
restricted to the first term is zero to make a shift of the integration variable in
the first, and then used (II.24), to finally obtain I. p 2 /. Now follow the explicit
evaluation of (II.28).
Let us use (II.27). This leads from the latter to the evaluation of the integral
Z 1 Z
2 p 2 2 kp
I. p / D dx .dk/
0 Œ .k px/2 C p 2 x.1 x/ C m2 2
Z Z 1
i 2 2 2 .2x 1/ dx
D p Cp .dk/ 2 C k2 x.1 x/ C m2 2
; (II.31)
2 0 Œ k
where in obtaining the latter equality, we have, in the process, used (II.24). By noting
that p 2 .1 2x/ dx D d . p 2 x.1 x//, we may carry out the integration over x
in (II.31) by parts, and then carry out the resulting integral over k using (II.8). The
integral on the right-hand side of (II.31), as the coefficient of p 2 , then becomes
Z 1
2 x.1 x/p 2 .1 2x/
i dx
0 Œ m2 C p 2 x.1 x/
Z 1
d h p2 i
D i 2 dx x.1 x/ ln 1 C 2 x.1 x/
0 dx m
Z 1 h i
p2
D C i 2 dx .1 2x/ ln 1 C 2 x.1 x/ : (II.32)
0 m
The last integral is zero since the change of variable x ! 1 x gives minus the
same integral. That is, the integral on the right-hand side of (II.31) is zero and we
obtain the result given in (II.30).
Another example that is worth considering is

Z
k 1 p k 2 i 2 2
.dk/ 2 2
D p p ; (II.33)
.k p/ C m 2 .k C m2 /2
2 3
and we leave it as an exercise for the reader to verify it by both methods given above.
A useful formula that follows from (II.27) is obtained by differentiating the latter
with respect to A. This gives
Z 1
1 1
D 2 x dx : (II.34)
A2 B 0 Œ Ax C B.1 x/ 3
More generally,
Z 1 Z
1 x
1
D 2 dx dz : (II.35)
AB C 0 0 Œ A.1 x/ C Bz C C.x z/ 3
The latter may be generalized for the product of N, not necessarily identical,
factors, and more conveniently written, as follows
Z 1 Z Z
1 x1 xN2
D .N 1/Š dx1 dx 2 dxN1
A1 A2 AN 0 0 0
1
; (II.36)
Œ A1 xN1 C A2 .xN2 xN1 / C C AN .1 x1 / N
which, for N D 3, coincides with (II.35) upon setting A1 D B; A2 D C; A3 D A,

x1 D x; x 2 D z .
Appendix III
Analytic Continuation in Spacetime Dimension
and Dimensional Regularization
Upon differentiation of the expression in (II.3)
Z 1 X
n1
1
D i ds expŒi s.k2 C M 2 i
/; k2 D ki2 .k0 /2 :
.k C M 2 i
/
2
0 iD1
(III.1)
. 1/ number of times, we obtain
Z 1
1 .i/
2 2
D ds s1 expŒi s.k2 C M 2 i
/: (III.2)
.k C M i
/ ./ 0
Hence from (II.6)

Z Z 1
.i/ 2 C1
n
d nk 1 2 C
/
ds s 2 1 es .i M
n
D : (III.3)
.2/n .k2 C M 2 i
/
n
.4/ 2 ./ 0
Upon using the integral

Z
1
2 C
/ 1 n
ds s 2 1 es .i M
n
D n ; (III.4)
0 .i M 2 C
/ 2 2
we obtain
Z
dn k 1 i 1 . n2 / n
D ; > :
.2/n .k2 C M 2 i
/ .4/ 2 .M 2 i
/ 2 ./
n n
2
(III.5)

DOI 10.1007/978-3-319-30939-2
530 III Analytic Continuation in Spacetime Dimension and Dimensional Regularization
From the above integral it easily follows that

Z
dn k k2 i n 1 . n2 1/
D ;
.2/n .k2 C M 2 i
/ .4/ 2 2 .M 2 i
/ 2 1 ./
n n
(III.6)
valid for > .n=2/ C 1.
One then defines integrals by an analytic extension of spacetime dimension to D
as follows
Z
dD k 1 i 1 . D2 /
D ; (III.7)
.2/D .k2 C M 2 i
/ .4/ 2 .M 2 i
/ 2 ./
D D
Z
dDk k2 iD .M 2 i
/ . D2 /
D :
.2/D .k2 C M 2 i
/ .4/ 2 .M 2 i
/ 2 .2 D 2/ ./
D D
(III.8)
Dimensional regularization consists of choosing
D D 4 "; " > 0; (III.9)
with " not to be confused with

earlier. The so-called ultraviolet divergences in
quantum field theory, i.e., at high energies, are equivalently expressed in terms of
singularities in ", with the latter kept > 0.
We may use Feynman parameters representation, as given in (II.27), to write
Z 1
1 1
D dx ; (III.10)
.k p/2 k2 0 Œ .k p x/2 C p 2 x.1 x/ 2
which from (III.7) leads to

Z 1 Z Z 1
dD k 1 i 1
dx D dx
"=2 ."=2/:
0 .2/D .k p/2 k2 D
.4/ 2 0 p 2 x.1 x/
(III.11)
Upon using the integral
Z 1
.˛/ .ˇ/
dx x ˛1 .1 x/ˇ1 D ; ˛ > 0; ˇ > 0; (III.12)
0 .˛ C ˇ/
we obtain
Z
dD k 1 i 1 "=2 . 2" / 2 .1 2" /
D : (III.13)
.2/D .k p/2 k2 .4/ 2 p 2
D
.2 "/
III Analytic Continuation in Spacetime Dimension and Dimensional Regularization 531
By shifting the variable k: k ! k C p x, using the expression on the right-hand

side of (III.10), and (III.12), in the process, we also obtain
Z
dD k k p i 1 "=2 . 2" / 2 .1 2" /
D : (III.14)
.2/D .k p/2 k2 2 .4/ D2 p 2 .2 "/
Here we have also used the property that the integral of an odd function of k is zero.
Another useful integral is obtained again by shifting the variable k: k ! k C p x, by
using in the process (II.8) and noting that in D dimensions D D to obtain
Z
dD k k k 1 i 1 "=2 . 2" / 2 .1 2" /
D
.2/D .k p/2 k2 4.3 "/ .4/ D2 p 2 .2 "/
Œ p 2 C.4 "/p p : (III.15)
In particular, using the fact that D .4 "/, dimensional regularization gives

from the above equation the interesting result
Z
dD k 1
D 0: (III.16)
.2/D .k p/2
The following properties of the - function should be noted

" "
.z C 1/ D z .z/; 1 D 1 C E C O."2 /;
2 2
" 2
D E C O."/; (III.17)
2 "
where E D 0:5772157 : : : is Euler’s constant.
Finally, we note that due to dimensional reasons, when the dimension of space-
time is analytically continued to D, an integral in four dimensions is then defined
by introducing an arbitrary mass parameter D , and by making, in the process, the
replacement .dk/ ! .D /" dD k. Therefore, a four dimensional regularized integral
may be defined as
Z ˇ Z
.dk/ ˇ " dD k
F.k; p/ ˇ D . D / F.k; p/: (III.18)
.2/4 Reg .2/D
We note that the integrals (III.13), (III.14), (III.15), using the definition
in (III.18), all involve the following factor, when expanded in powers of ",
1 2 "=2 " 1 h2 i
2 2
D
D E `n. p = / C `n.4/ C O."/ ;
.4/
D
2 p2 2 16 2 " D
(III.19)
532 III Analytic Continuation in Spacetime Dimension and Dimensional Regularization
up to overall multiplicative finite constant terms coming from the other gamma
functions occurring in their factors, for " ! 0. The parameter " has now replaced
the ultraviolet cut-off.
It is also worth knowing that the following two integrals obtained by the same
method as above, and by power counting, are ultraviolet finite (UV finite):
Z
dD k 1
D UV finite; (III.20)
.2/D . p1 k/2 . p 2 k/2 k2
Z
dD k k
D UV finite; (III.21)
.2/ . p1 k/ . p 2 k/2 k2
D 2
while with Q D p 2 p1 ¤ 0,
Z
dD k k k i 1 "=2 . 2" / 2 .1 2" /
D C UV finite:
.2/D . p1 k/2 . p 2 k/2 k2 .4/ D2 Q2 .3/ .3 "/
(III.22)
In D D 4" dimensions, the gamma matrices are defined to satisfy the following
relations:
Box III.1: Some properties of the gamma matrices in D D 4 " dimensions
f ; g D 2 I; D .4 "/; D .4 "/I;

D .2 "/ ; Tr I D 4; Tr Œ D 4 ;

Tr ˛ ˇ D 4 ˛ˇ ˛ ˇ C ˛ ˇ ;
Tr Œ odd number of ’s D 0:
Appendix IV
Schwinger’s Point Splitting Method of Currents:
Arbitrary Orders
The vacuum-to-vacuum transition amplitude corresponding to Dirac’s equation

Œ .@ =i eA / C m D 0, in the presence of an external electromagnetic field
A , is given in (3.6.26) to be given by
h 0C j 0 i.e/ D exp i W; (IV.1)
which is normalized to unity for e D 0. Here .x ˙ D x ˙

=2/
Z e R
0 xC
iW D de 0 .dx/ Tr Œ S.x ; x C I e 0 A/ A ei e x d A ./ ;
! 0;
0
(IV.2)
as given in (3.6.27), with a priori dependence on
arising from Schwinger’s point
splitting of the current as given in (3.6.24) and spelled out to be
Z xC
h 0C j j .x/ j 0 i.e/
D i Tr Av Œ SC .x ; x C I eA/ exp i e d A ./ ;
h 0C j 0 i.e/ x
(IV.3)
0 > 0 and
0 < 0, with
in a space-like
direction. Here Tr denotes a trace over the spinor indices of gamma matrices. We
also note that SA .x ; x C / SC .x ; x C I eA/.
The purpose of this appendix, is to develop a perturbation expansion of (IV.2)
to arbitrary orders in eA in the limit
! 0. We will see, in particular, that the
perturbative expansion, due to the presence of Schwinger’s line-integral, is modified
from its naïve expression, only up to fourth order in the external electromagnetic
field.
We first consider the expression
Tr Œ S.x ; x C I eA/ A .x/ Tr Œ S.x ; x C I eA/ A .x/; (IV.4)

DOI 10.1007/978-3-319-30939-2
534 IV Schwinger’s Point Splitting Method of Currents: Arbitrary Orders
multiplying the Schwinger line-integral in (IV.2) for a coupling e. From the integral
equation for S.x ; xC I eA/ SC
A
.x ; xC /, in (3.2.12), we may carry out an expansion
in powers of the external potential A as
Tr Œ S.x ; x C I eA/ A .x/ D A1 .x/ Tr ŒSC .x x C / 1 C

X Z
C .e/N1 .dx 2 / .dxN / A1 .x/A 2 .x 2 / A N .x N /
N2
Tr Œ SC .x x N / N SC .x N x N1 / N1 2 SC .x 2 x C / 1 ; (IV.5)
where SC .x x 0 / is the fermion propagator in (3.1.9). Equivalently for the

expression on the right-hand side of (IV.5), we have
Tr Œ S.x ; x C I eA/ A .x/ D A1 .x/ Tr Œ 1 SC .x x C / C

X Z
C .e/ N1
.dx2 / .dxN / A1 .x/A 2 .x 2 / : : : A N .x N /
N2
Tr Œ 1 SC .x x 2 / 2 : : : N1 SC .x N1 x N / N SC .x N x C / : (IV.6)
We may rewrite the TrŒ. : / term within the multiple integral in (IV.5) as
Tr Œ . : /> TrŒ C 1 . : /> C , where C is the charge conjugation matrix in (I.3),
satisfying, in particular,
C 1 . /> C D ; C 1 .SC .x x 0 //> C D SC .x 0 x/: (IV.7)
Accordingly, (IV.5) may be re-expressed as
Tr Œ S.x ; x C I eA/ A .x/ D A1 .x/ Tr Œ 1 SC .x C x / C

X Z
C .1/ .e/
N N1
.dx2 / .dxN / A1 .x/A 2 .x 2 / AN .x N /
N2
Tr Œ 1 SC .x C x 2 / 2 N1 SC .x N1 xN / N SC .x N x / : (IV.8)
Eqs. (IV.6), (IV.8) allow us to write (IV.5) as the average of the just mentioned
two equivalent expressions. That is
1 h i
Tr Œ S.x ; x C I eA/ A .x/ D A1 .x/ Tr 1 Œ SC .x x C /SC.x C x /
2
X Z
C .e/N1 .dx 2 / : : : .dx N / A1 .x/A 2 .x 2 / AN .x N /
N2
1
Œ F 1 2 N .x ; x 2 ; : : : ; x N ; x C / C .1/N FN 1 2 N .x C ; x 2 ; ; x N ; x / ;
2
(IV.9)
IV Schwinger’s Point Splitting Method of Currents: Arbitrary Orders 535
where
F 1 2 N.x; x2 ; : : : ; xN ; x 0 / D Tr Œ 1SC .x x2 / 2SC .x2 x3 / NSC .xN x 0 / :

(IV.10)
Note that SC .x x C /, SC .x C x /, are independent of x.
By carrying out Fourier transforms,
Z Z
.dQ/ iQx 0 .dp/ ip.xx 0 /
A .x/ D e A .Q/; SC .x x / D e SC . p/; (IV.11)
.2/4 .2/4
(IV.9) becomes
Tr Œ S.x ; x C I eA/ A .x/

Z
.dp/ .dQ1 / iQ1 x 1 ip
i Q1
=2
D e Œe e eip
ei Q1
=2 A1 .Q1 /Tr Œ 1 SC . p/ C
.2/4 .2/4 2
X Z .dp/ .dQ1 / .dQN / i .Q1 C C QN /x
C 4 .2/4
::: e
N2
.2/ .2/4
1 N1 ip
i P0 Q i
=2 P0
e Œe e C .1/N eip
ei Q i
=2 A1 .Q1 / AN .QN /
2

Q1 2 Q1 Q1
Tr 1 SC . p / SC . p Q2 / : : : N SC . p Q2 : : :QN / :
2 2 2
(IV.12)
Now we consider the contribution of the gauge compensating factor, i.e., of the
Schwinger line-integral in (IV.2), for a given coupling e. Since we will eventually
take the limits of
going to zero, we may carry out an expansion as follows
Z xC
exp Œ i e d A ./ D 1 C I 1 C I 2 C I 3 C : (IV.13)
x
To carry out the - integration, we make a change of variable ! W D

x C .
=2/ , 1 C1, make a Fourier transform of A .x/, and expand in
powers of
to obtain for
' 0
Z h i
ie .dQ1 / i Q1 x .Q1
/2
I1 D 2 e A 1 .Q 1 / 1 C
1 ; (IV.14)
2 .2/4 24
2 Z
ie 1 .dQ1 / .dQ2 / i.Q1 CQ 2 /x
I2 D 4 e A1 .Q1 / A 2 .Q2 /Œ 1 C
1
2 ;
2 2Š .2/4 .2/4
(IV.15)
3 Z
ie 1 .dQ1 / .dQ2 / .dQ3 / i .Q1 CQ 2 CQ 3 /x
I3 D 8 e A1 .Q1 / A 2 .Q2 /
2 3Š .2/4 .2/4 .2/4
A 3 .Q 3 /Œ 1 C
1
2
3 : (IV.16)
We will show that all the terms represented by : : : in (IV.14), (IV.15), (IV.16), as
well as I4 ; I5 ; : : : in (IV.13), will not contribute in the limiting procedure.
We multiply (IV.12) by the expression Œ1 C I1 C I2 C I3 C : : : in (IV.13),
extract the .N 1/th order terms in e of the product, integrate over e as indicated
in (IV.2), and P integrate over x as well. P0
Here we note the basic property of the
0
factor Œeip
e i Qi
=2 C .1/N eip
e i Qi
=2 =2 in (IV.12), in reference to (IV.13),
(IV.14), (IV.15), (IV.16):
1
ip
i P0 Q i
=2 P0

1 : : :
k
e e C .1/Nk eip
ei Q i
=2
2
@ @ 1
P0 P0
D i ::: i eip
e i Q i
=2 C .1/N eip
ei Q i
=2 : (IV.17)
@p1 @p k 2
We may then integrate by parts over p to apply these derivatives with respect
to the pj , replacing the
j in (IV.14), (IV.15), (IV.16), as indicated above, to the
Tr Œ : factor in (IV.12), and take the limit
! 0. The process is straightforward,
and we obtain for the Nth- order term in e for i W in (IV.2)
ˇ Z
ˇ .e/N .dQ1 / .dQN / i .Q1 C C QN /x
i Wˇ D .dx/ 4
::: e
.N/ N .2/ .2/4
A1 .Q1 / A N .QN / L1 ::: N .Q1 ; : : : ; QN /; (IV.18)
where
Z
1
.dp/
L1 ::: N .Q1 ; : : : ; QN / D 1 C .1/N ˘ 1 ::: N . p I Q1 ; : : : ; QN /;
2 .2/4
(IV.19)
and we immediately note that only terms even in N, i.e., for N D 2; 4; : : : ;

contribute. This is known as Furry’s Theorem . It is a consequence of charge
conjugation invariance of electrodynamics (if A is also made to transform to A
consistently with Maxwell’s equations).
Before spelling out the explicit expressions for ˘ 1 ::: N . p I Q1 ; : : : ; QN /, for the
various N, we first recall Gauss’ Theorem in 4D conveniently written in the way
it presents itself in the above formulae for a function f . p/ involving a product of
Dirac propagators. It reads
Z Z
@
.dp/ f . p/ D d f . p/; (IV.20)
@p
with the right-hand side defined as a surface integral, where d is an element of

surface. With p0 ! ip0 , d D n d, n a unit vector along p , if d scaled
by a parameter , it grows like 3 . Accordingly, if f .p/ vanishes faster than 3 ,
then the integral on the right-hand side is zero.
According to the rule in (IV.17), the gauge compensating terms In in (IV.13),

each of them, lead to n or more derivatives @=@p , corresponding to the

appearing in the In ,3 and acting on the Tr Œ : factor in (IV.12). That is, a gauge
compensating term In leads explicitly to integrals of the form
Z
@ @ @
.dp/
@p 1 @p 2 @p n

Q1 2 Q1 Q1
Tr 1 SC . p / SC . p Q2 /: : : k SC . p Q2 Qk / ;
2 2 2
(IV.21)
and may involve even more derivatives with respect to p. By Gauss’ Theorem, such
integrals would vanish for n 1 C k 4, since S.p/ D O.1 /. Accordingly, for
k D 1, only, I1 ; I2 ; I3 may contribute, and for k D 2, only I1 ; I2 may contribute,
and finally for k D 3 only I1 may contribute. Hence we may infer that only ˘ 1 2 ,
and ˘ 1 2 3 4 would involve an addition to the expressions in (IV.2) coming from
gauge compensating terms. Needless to say ˘ 1 2 3 is zero, corresponding to N D
3.
Accordingly for N D 6; 8; : : : ,
˘ 1 ::: N . p I Q1 ; : : : ; QN /

Q1 2 Q1 Q1
D Tr 1 SC . p / SC . p Q2 / : : : N SC .p Q2 QN / sym ;
2 2 2
(IV.22)
where Œ : sym means to symmetrize over the .N 1/ pairs of “generalized indices”

.2 ; Q2 / , . . . ,.N ; QN /.
For N D 4, we have
˘ 1 2 3 4 . pI Q1 ; : : : ; Q4 / D

Q1 2 Q1 Q1
Tr 1 SC . p / SC . p Q2 / : : : 4 SC . p Q2 Q3 Q4 /
2 2 2
@
Q1 2 Q1 Q1
C Tr 1 SC . p / SC . p Q2 / 3 SC . p Q2 Q3
@p 4 2 2 2
1 @ @
Q1 2 Q1
C Tr 1 SC . p / SC . p Q2 /
2 Š @p 3 @p 4 2 2
1 @ @ @

C Tr 1 SC . p/ : (IV.23)
3Š @p 2 @p 3 @p 4
3
See, e.g., (IV.14), (IV.15), (IV.16).
A symmetrization is necessary. However much simplification of this expression may

be done first. To this end, note that for any given q, we may write
1 1 1 1
D C Q : (IV.24)
. p Q/ C m p C m . p Q/ C m p C m
Clearly, the second term introduces one more power of the momentum p in the
denominator, and hence invoking Gauss’ Theorem, we may infer that we may set
the Q momenta in the second and third terms on the right-hand side of (IV.23)
equal to zero. On the other hand symmetrizing the second term on the right-hand
side of (IV.23) over 2 ; 3 , after setting the Q momenta equal to zero, it takes the
simple form
1 @
1 @

Tr 1 SC .p/ 2 SC . p/ 3 SC . p/ C Tr 1 SC . p/ 3 SC . p/ 2 SC . p/
2 @p 4 2 @p 4
1 @ @

Tr 1 SC . p/ 2 SC . p/ ; (IV.25)
2 @p 3 @p 4
which exactly cancels the third term on the right-hand side of (IV.23) (after setting
the Q’s equal to zero). Hence (IV.23) finally reduces to
˘ 1 2 3 4 . p I Q1 ; : : : ; Q4 /

Q1 2 Q1 Q1
D Tr 1 SC . p / SC . p Q2 / : : : 4 SC . p Q2 Q3 Q4 / sym
2 2 2
1 @ @ @

C Tr 1 SC . p/ : (IV.26)
3 @p 2 @p 3 @p 4
For N D 2, we simply have (Q2 D Q1 )

Q1 2 Q1
˘ 1 2 . p I Q1 ; Q1 / D Tr 1 SC . p / SC . p C /
2 2
1 @ @ @ @
C Q11 Q1 2 Tr Œ 1 SC . p/ C Tr Œ 1 SC . p/ : (IV.27)
24 @p 1 @p 2 @p 2 @p 2
Thus we have obtained all the terms in (IV.18), (IV.19), for the Nth order
contribution to i W in (IV.2). After integrating (IV.18) over x, the following
expression emerges for i W
X .e/N Z .dQ1 / .dQN /

iW D 4
::: 4
.2/4 ı 4 .Q1 C C QN /
ND2;4;::
N .2/ .2/
Z
.dp/
A1 .Q1 / : : : AN .QN / ˘ 1 ::: N . p I Q1 ; : : : ; QN /; (IV.28)
.2/4
where ˘ 1 :::N . p I Q1 ; : : : ; QN / is given in (IV.22) for N D 6; 8; : : :, and in (IV.26)

for N D 4, and in (IV.27) for N D 2. The expression in (IV.27) for N D 2 was
possibly first derived in details by K. Johnson4, and by Schwinger himself.
4
See, e.g., K. Johnson (1965). Brandeis University Summer Institute in Theoretical Physics. In
Lectures on particles and fields. Englewood Cliffs: Prentice Hall.
Appendix V
Renormalization and the Underlying
Subtractions
In this appendix, we learn how to define the renormalized Feynman integral,

associated with a graph, given its unrenormalized Feynman integrand.
One defines a graph by specifying a set of vertices V D .1 ; : : : ; r /, a set of
lines L D .`1 ; : : : ; `s /, together with a rule describing which of these lines join
which of the vertices. By removing all the external lines of a graph, one generates
an amputated graph. Vertices fi g in an amputated graph, to which external lines
were attached are referred to as external vertices, while their remaining vertices
are referred to as internal vertices. From now on we will consider only amputated
graphs and amputated subdiagrams with the latter to be defined below. By specifying
a subset V 0 V of the vertices and all the lines in L of a graph G that join these
vertices, but not those lines joining vertices not in V 0 , one defines a subgraph G 0 of
G. By specifying a subset V 0 V of the vertices and not necessarily all the lines
in L of the original graph that join these vertices, one defines a subdiagram g of
G. Examples of an (amputated) graph, together with a subgraph and a subdiagram
are given in Fig. V.1.
A vertex of an amputated subdiagram at which an external line was attached
previous to the removal of the latter is referred to as an external vertex, while
the other vertices are referred to as internal vertices. All the lines of an amputated
subdiagram are, by definition, internal lines.
A subdiagram is called disconnected if it is connected out of two or more
subdiagrams in which any two of them have no vertices and no lines in common.
If upon cutting or removing an internal line of a subdiagram, the number of the
connected parts of the latter is increased, it is called an improper line. A subdiagram
is called a proper connected subdiagram, if it is connected and has no improper
lines. A subdiagram is called proper but not connected, if it is constructed out of two
or more subdiagrams none of which have improper lines. Two subdiagrams g1 ; g2
specified by the pairs .V1 ; L1 /, .V2 ; L2 /, for which V1 V2 , and L1 L2 , we
write g1 g2 . In this work, the symbol may include equality, while ¦ is used
to exclude an equality.

DOI 10.1007/978-3-319-30939-2
542 V Renormalization and the Underlying Subtractions
(a) (b) (c)

a b a b a b
e c
f d f f
Fig. V.1 A graph is introduced in part (a). The dashed lines stand for external lines that have
been removed defining an amputated graph. Vertices at f and d represent external vertices for the
graph, while the remaining ones represent internal vertices. A subgraph is introduced in part (b)
with four external lines that have been removed. In this case f and b represent external vertices of
the subgraph, while a represents an internal vertex. A subdiagram is introduced in part (c) where
one of the lines joining the vertices at a and b has been cut with the resulting two external lines
removed. The vertices at a, b and f represent external vertices for the subdiagram in part (c)
Since the subtractions of renormalization involve a sequence of Taylor expan-

sions in the momenta associated with the external lines of a proper graph and in the
momenta of the external momenta of proper subdiagrams, we have to spell out how
such momenta are defined for the proper graph and the proper subdiagrams. This is
considered next.
Suppose that a proper and connected graph G has m external vertices with
total external momenta fq1 ; : : : ; qm g q G , qr D .q0r ; q1r ; q2r ; q3r /. The integration
variables associated with the graph are denoted by fk10 ; k11 ; : : : ; kn3 g k G . A line
labeled `, joining a vertex i to a vertex j , carries a momentum denoted by Qij `
which may be written as
X
n X
m
Qij ` D kij ` C qij ` ; kij ` D asij ` ks ; qij ` D brij ` qr ; Qij ` D Qj i ` :
sD1 rD1
(V.1)
At each external vertex j of G,
X X
G
Qij ` D qG
j ; and by momentum conservation j D 0;
qG (V.2)
i` j
where the sum is over all i corresponding to all the vertices with lines ` joining
them to the external vertex j . For an internal vertex, the right-hand side of the
above first equality is set equal to zero.
A line joining a vertex i to a vertex j in G, will be represented by
Cij ` .Qij ` ; ij ` / D fij ` .Qij ` ; ij ` /ŒQ2ij ` C 2ij ` 1 ; (V.3)
involving a familiar multiplicative factor fij ` .Qij ` ; ij ` / when one is dealing with
particles of non-zero spin. Let j be fixed, and consider the set fi.j/ g1 i rj of
V Renormalization and the Underlying Subtractions 543
vertices attached by lines to the vertex j in G, and consider the set L G .j /
of all lines joining the vertices i.j/ to the vertex j . Moreover, let fQij ` g1 ` sij
be the set of momenta carried by these lines. Then we assign to the veretx j a
polynomial Pj D Pj .Qij1 ; : : : ; Qrj j srj j /. The unrenormalized Feynman integrand
associated with the proper and connected graph G, up to an overall multiplicative
constants (involving couplings, etc.), is of the form
Y
IG D G
ij `;i<j Pj Cij ` : (V.4)
We introduce canonical variables,4 by choosing

X
G
qij ` D qG
j ; and qij ` D ui uj ; (V.5)
i`
where ui ; uj are four vectors, and the second equality above means that the external
variables of the lines joining a vertex i to the vertex j are all chosen to be equal.
In particular, we also note that (V.2), (V.5) imply that
X X
G
kij ` D 0; G
kij ` D 0; (V.6)
i` ij`
with the latter defining a constraint. Equation (V.5) provides .# V 1/ independent

solutions of the .# V 1/ independent differences .uj ui /, where # V denotes
the number of vertices. We may write the variables qij ` as a linear combination of
the elements in fqG j g q : qij ` D qij ` .q /.
G G
If 4n denotes the number of independent integration variables associated with

G then n D # L G # V G C 1. Eqs. (V.6) imply that only n of the # L G kij ` are
independent, and we may write kij ` D kij ` .k/. We provide examples of canonical
decompositions of some graphs and some diagrams obtained from them.
Example V.1 Consider the graph in Fig. V.2a, with external vertices at 1 and 2. At
the vertex 1
1 1
q D q211 C q212 C q213 ; q; i.e. q21 ` D q; ` D 1; 2; 3:
or u2 u1 D
3 3
(V.7)
At vertex 1 , we also have: k211 C k212 C k213 D 0. Let k211 D k1 , k212 D k2 , then
k213 D k1 k2 . Therefore a canonical choice of variables are
1 1 1
Q211 D k1 C q; Q212 D k2 C q; Q213 D k1 k2 C q: } (V.8)
3 3 3
4
Canonical variables were introduced by Zimmermann [9].
(a) 1
(b) 1
2 q2 q1 1 3 q3
1
q1 1 2 2
3
2
2
3
Fig. V.2 (a) A graph with external vertices at 1 and 2, and external momenta q1 q, q2 q.
(b) A graph with external vertices at 1 and 3 and an internal vertex at 2
Example V.2 Consider the graph in Fig. V.2b with q1 q, q3 q. At the
vertices, 1 , 2 , 3 , we have, respectively, with 2 an internal vertex
q D q211 C q212 C q213 C q311 C q312 D 3.u2 u1 / C 2.u3 u1 /; (V.9)

0 D q121 C q122 C q123 C q32 D 3.u1 u2 / C .u3 u2 /; (V.10)
q D q131 C q132 C q23 D 2.u1 u3 / C .u2 u3 /: (V.11)
The unique solutions of the differences are then: u1 u2 D q=11, u1 u3 D

4 q=11, u2 u3 D 3 q=11. That is, q12 ` D q=11, ` D 1; 2; 3 ; q13 ` D 4 q=11,
` D 1; 2 ; q23 D 3 q=11. For 16 integration variables, we may write for the four
vectors: k121 D k1 , k122 D k2 , k123 D k3 , k131 D k4 . For the remaining integration
variables, we note that at 2 , and 3 , respectively, we have
k121 C k122 C k123 C k32 D 0; k131 C k132 C k23 D 0; (V.12)
from which k32 D k1 k2 k3 , and k132 D k1 k2 k3 k4 . Accordingly a

canonical set of variables are
1 1 1 4
Q121 D k1 q; Q122 D k2 q; Q123 D k3 q Q131 D k4 q;
11 11 11 11
(V.13)
4 3
Q132 D k1 k2 k3 k4 q; Q32 D k1 k2 k3 C q: }
11 11
(V.14)
Example V.3 Consider the graph G in Fig. V.3 below, with q1 D q, q4 D q. Let
k12 D 3 .k1 C 3 k2 /=8, k34 D 3 .3 k1 C k2 /=8. Following the procedure in the
above two examples gives rise to the following canonical decompositions:
3 9 1 3 9 1 3 3
Q12 D k1 C k2 C q; Q13 D k1 k2 C q; Q32 D k1 k2 ;
8 8 2 8 8 2 4 4
(V.15)
9 3 1 9 3 1
Q34 D k1 k2 C q; Q24 D k1 C k2 C q: }
8 8 2 8 8 2
(V.16)
Canonical decomposition corresponding to a proper and connected subdiagram

g of G, will be defined in a similar manner. We write the momentum associated
with a line in g as
g g
Qij ` D kijG ` C qG
ij ` D k ij ` C q ij ` ; (V.17)
and at each external vertex j of g,

X g g
X g g
g
q ij ` .kG ; qG / D q j .kG ; qG /; g
q j .kG ; qG / D 0 and q ij ` D wi wj ;
i` j
(V.18)
where wi ; wj are four vectors, and the sum over i ` pertains to the subdiagram g.
g g
If j is an internal vertex of g then we set q j D 0. Once the q ij ` are determined,
g g g
then k ij ` D Qij ` q ij ` . We show that the k ij ` are linear combination of the
elements in kG only and are independent of the elements in qG . To see this, set
the integration variables in kG , equal to zero, then
g g
X g g
X X g
g
q ij ` .0; qG / D q j .0; qG /; q ij ` .0; qG / D 0;
i` j ı`
ˇ
D .wi wj /ˇkG D0 D wi 0 wj 0 :
g
q ij ` .0; qG / (V.19)
g g
But for k G D 0, Qij ` D qG
ij ` D k ij ` .0; q / C q ij ` .0; q / and
G G
X X X
g 0 0
g G
qij ` D g
Qij ` D q j .0; qG /; g
ij ` D 0;
qG ij ` D ui uj ;
qG
i` i` i`
(V.20)
for i; j; ` pertaining to the subdiagram g. Uniqueness of the solutions of the

.# V g 1/ conditions of the .# V g 1/ independent differences ui 0 uj 0 or
wi 0 wj 0 , implies that ui 0 uj 0 D wi 0 wj 0 . That is k ij ` .0; qG / D 0 and hence
g
g g
k ij ` D k ij ` .k /.
G
0 0
Similarly for two proper connected subdiagram g 0
g g
¤ g, k ij ` C q ij ` D
g g g0 g0 g0 g0 g0
k ij ` C q ij ` , and k ij ` D k ij ` .kg /, q ij` D q ij ` .kg ; qg /. In particular, the q j are
linear combinations of the kij ` in g=g0 . Here we note that once the unrenormalized
g
Feynman integrands Ig and Ig 0 have been defined, Ig=g 0 is defined through Ig D

Ig=g 0 Ig 0 . Note that
X
g0 0
ij ` D q .k ; k /;
qG g g g
at each vertex j of g 0 ; (V.21)
i`
with the sum over i ` pertaining to the subdiagram g 0 . Also

X X X
g g0 g=g 0
qj D g
ij ` D
qG Qij ` C Qij ` ; (V.22)
i` i` i`
and hence the above two equations imply that
g0
X g
g g=g 0 g
q j .kg ; qg / D q j k ij ` C q ij ` ; (V.23)
i`
which establishes the stated result.
Example V.4 Consider the graph G and the two subdiagrams g1 and g2 in Fig. V.3.
From (V.15), (V.16), (V.17), at the vertices 2 and of 3 of subdiagram g1 , respectively,
g g g 3 9 1
q 21 D q 421 C q 321 D Q42 C Q32 D w4 w2 C w3 w2 D k1 k2 q;
8 8 2
(V.24)
g g g 3 9 1
q 31 D q 431 C q 231 D Q43 C Q23 D w4 w3 C w2 w3 D C k1 C k2 q:
8 8 2
(V.25)
g
These lead to .q 41 D q/
g 1 3 1 g 1 3 1 g 1 3
q 341 D k1 k2 C q; q 241 D k1 C k2 C q; q 321 D k1 k2 :
8 8 2 8 8 2 4 4
(V.26)
For the internal variables we then have

g g g g g g
k 341 D Q34 q 341 D k1 ; k 241 D Q24 q 241 D k1 ; k 321 D Q32 q 321 D k1 :
(V.27)
2
1 2 3 4 1 3 2 3 4
G g2 g1
Fig. V.3 A graph G with external vertices at 1 and 4, showing only two of its proper subdiagrams
g1 and g2 having each three external vertices. The dashed lines are just to remind us of the external
lines that have been removed.
g g
Similarly for the subdiagram g2 we have q12 D q; q22 D .9=8/ k1 C .3=8/ k2 C
g2
.q=2/; q 3 D .9=8/ k1 .3=8/ k2 C .q=2/
g 3 1 g 3 1 1 g 3 1 1
q 322 D k1 C k2 ; q 132 D k1 k2 C q; q 122 D k1 C k2 C q;
4 4 8 8 2 8 8 2
(V.28)
g g g
k 322 D k2 ; k 132 D k2 ; k 122 D k2 : } (V.29)
Once the external momentum variables qG ij ` of the graph G in question, as well as

g
the external variables q ij ` of all proper and connected subdiagrams of G have been
defined, one may define Taylor operations TG , Tg in their external momenta about
the origin up to their (superficial) degree of divergences. The (superficial) degree
of divergence d.g/ of a proper connected subdiagram g involving 4 t integration
variables is defined by
d.g/ D deg Ig C 4 t: (V.30)
where “deg” is here defined by scaling the integration variables in Ig by a

parameter and by considering the power of for ! 1. For a proper not
necessarily connected subdiagram g involving proper and connected subdiagrams
gP1 ; g2 ; : : : ; gs , the (superficial) degree of divergence of g is defined by d.g/ D
s
iD1 d.gi /.
We are particularly interested in the consecutive application of two or more
Taylor operations. To this end, consider two proper subdiagrams g 0 ¤ g, we define
Tg Tg 0 Ig by the following procedure. We first note that
0 0
Tg 0 Ig 0 D F.Q; k g ; q g /; (V.31)
0 0
where the function Ig=g 0 F.Q; kg ; qg / should be now expressed in terms the internal
k g , and external q g variables of g. To do this, we use (V.17), and the functional
0 0 0 0
relations k g D k g .k g /, q g D q g .k g ; q g /. The application of consecutive several
Taylor operation Tg 0s : : : Tg 01 Ig 0s for proper subdiagrams g 01 0
¤ : : : ¤ g s is similarly
handled.
Now we are ready to define the renormalized R Feynman integrand involving
Taylor operations, about the origin, applied to the external variables of a proper
and connected graph G, and to its proper, not necessarily connected subdiagrams.
The external variables of such subdiagrams constitute the totality of the external
variables of its connected parts.
The renormalized Feynman integrand R, involving subtractions, is defined as
follows:
h XY i
R D 1C .Tg / I G ; (V.32)
D g
D
where the sum is over all non-empty sets D such that

(i) If g
D, then g is a proper, but not necessarily connected, subdiagram of G
with d.g/ 0. If d.G/ 0, then one of the elements of D may be G itself.
(ii) If g1 ; g2
D, then either g1
¤ g2 or g2 ¤ g1 . If g1 ¤ g2 , then the ordering
of the Taylor operations in (V.32) is as : : : .Tg2 / : : : .Tg1 / : : : .
Note that the Taylor operations are applied directly to the integrand I G in the
momentum representation, and no questions of divergences arise in (V.32). The
sets D are called renormalization sets.
We see that an important task is to find the renormalization sets D associated
with a graph G. For example, for the graph G in Fig. V.3, the normalization
sets, say for simplicity for scalar field self-coupling in 6 dimensional spacetime,
are: fGg; fG; g1 g; fG; g2 g; fg1 g; fg2 g. Note that the subdiagram obtained from the
graph G by cutting (i.e, removing) line 23 has a degree of divergence -2 in 6
dimensions. Note also that fg1 ; g2 g is not a D set, since neither g1 ¤ g2 nor
g2 ¤ g 1 . The renormalized integrand then becomes
h i
1 C .T G / C .T G /.T g1 / C .T G /.T g2 / C .T g1 / C .T g2 / I G
h i
D .1 T G / 1 C .T g1 / C .T g2 / I G ;
(V.33)
with d.G/ D 2; d.g1 / D d.g2 / D 0.

Remarks
(1) The key point in establishing the finiteness of the renormalized Feynman
integral in Euclidean space for massive particles 2ij ` ¤ 0 is to show that when
any subset of the integration variables become large, one can always group the
Taylor operations in a particular way, and show that the renormalized integrand
has just right behavior and vanishes rapidly asymptotically as a function of these
variables and guaranties its integrability. A proof of this is given in Manoukian
[3]). For a proof of the finiteness of the renormalized in Minkowski space, in
the sense of distributions, see also the latter reference, and for a simpler and
direct proof see Manoukian [4, 5], including for zero-mass particles.
(2) Once the finiteness of the renormalized theory is established, with subtractions
at the origin, it may be then normalized at other points.
(3) The above subtraction procedure may be reduced to the one referred to as
the Bogoliubov-Parasiuk-Hepp-Zimmermann scheme, and for the proof of the
underlying theorem, referred to as the “Unifying Theorem of Renormaliza-
tion”,5 see Manoukian [3].
(4) The renormalization procedure with subtractions is equivalent to adding terms
to the original Lagrangian, referred to as counter-terms, and adjusting their
numerical coefficients in such a way that parameters appearing in the theory
5
See also Zeidler [8, p. 972] and Figueroa and Gracia-Bondia [1].
(masses, couplings,. . . ) are taken from experiments. For the equivalence of the
subtraction scheme and the counter-term formalism, see Manoukian [2, 3]. This
establishes the criterion of renormalizabilty and is expressed as follows. One, a
prioi, may generalize the original Lagrangian density by adding to it new terms.
If the counter-terms needed to make the (modified) theory finite and consistent,
have the same structures as of terms in the (modified) Lagrangian density and
are finite in number, then the (modified) theory is called renormalizable. In the
latter case only a finite number of parameters are taken from experiments. For
example in QED, the counter-terms have exactly the same structures of terms in
the Dirac-Maxwell Lagrangian density and the values of the two parameters the
(renormalized) mass and (renormalized) charge of the electron are taken from
experiments. If the number of counter-terms needed are infinite in number, then
one may need to fix an infinite number of parameters and the theory loses its
predictive power.
(5) Asymptotic behavior of the renormalized theory, such as at high-energy, or for
small masses, or for large masses, and other variations, as well as of the proof
of the decoupling theorem consistently used in QCD, see Manoukian [3].
(6) Regarding the author’s work in the completion of the renormalization program
stemming that of Salam’s, Streater [7] writes: “It is the end of a long chapter
in the history of physics”. He also states: “Physicists found Salam’s [method]
easier than the BPH one”.
(7) When you write down a local Lagrangian density, the parameters (couplings,
masses,. . . ) are introduced at infinite energies (specified by large ultraviolet cut-
offs) and are unattainable. Renormalization theory eliminates these parameters
in favor of physically measurable parameters.
Recommended Reading
1. Figueroa, H., & Gracia-Bondia, J. M. (2004). The uses of Connes and Kreimer’s algebraic
formulation of renormalization. International Journal of Modern Physics, A19, 2739–2754.
hep–th/0301015v2.
2. Manoukian, E. B. (1979). Subtractions vs Counterterms. Nuovo Cimento, 53A, 345–358.
3. Manoukian, E. B. (1983a). Renormalization. New York/London/Paris: Academic Press.
4. Manoukian, E. B. (1983b). Elementary proof of
! C0 limit of Renormalized Feynman
amplitudes. Journal of Physics: Mathematical and General, 16, 4131–4133.
5. Manoukian, E. B. (1984b). Elementary proof of
! C0 limit of Renormalized Feynman
amplitudes. II: Theories involving zero mass particles. Journal of Physics: Mathematical and
General, 17, 1931–1935.
7. Streater, R. F. (1985). Review of Renormalization by E. B. Manoukian. Bulletin of London
Mathematical Society, 17, pp. 509–510.
pp. 972–975.
9. Zimmermann, W. (1969). Convergence of Bogoliubov’s method of renormalization in momen-
tum space. Communications in Mathematical Physics, 15, 208–234.
Solutions to the Problems
Chapter 2
p p
2.1. Note that cos 2 D 1= jaj2 C 1, sin 2 D jaj= a2 C 1, where we have used
f 0 ; g D 0, . a/ 2 D jaj2 . Also note that .jaj a/
h a i
G 0 G1 D 0 cos 2 sin 2
jaj
0 1 0

G a G D a cos 2 C a sin 2 :
h
Hence G 0 . a C 1/ G1 D 0 a cos 2 .sin 2=jaj/ C .cos 2 C
i
a sin 2 .
The statement of the problem then follows from the expressions of cos 2
and sin 2 given above. p p
2.2. Let a D p=m, cos D . p0 C m/=2p p
0 , sin D jpj= 2p0 . p0 C m/ in
p
Problem 2.1. Then G H G1 D m 0 R p2 =m2 C 1 D 0 p2 C m2 .
2.3. TheR integral in question is given by: dj .j ˇj /.˛ 0 C c1 j / which reduces
to dj j .˛ 0 C c1 ˇj / D ˛R 0 C c1 ˇj D f .ˇj /, where
R c1 is a c number.
2.4. The integral is given by d .˛ 0 C c1 / D d c1 D c1 , since c1 is a c
number. On the other
R Rhand, .@=@/.˛
0 C c1 / D c1 .@=@/ D c1 .
R
2.5. (i) dR R D 1 D dR R . The latter, in turn, is equal to R .dR / D
R
.dR / R . That is, .dR / D dR , and similarly .dI / D dI . From
the definition d D .dR C i dI / we obtain .d/ D . dR .i/.1/dI /,

from which the condition .d/ D d follows. (ii) Using the latter
R R R
result, we have d D . d / D d . Recall that complex
conjugation reverses the order in a product.

DOI 10.1007/978-3-319-30939-2
552 Solutions to the Problems
2.6. The integral in question may be rewritten as:

Z
d1 1 Œ.i/2 .1 ˛1 /.1 ˛ 1 / 1 d1 : : : dn n Œ.i/2 .n ˛n /.n ˛ n / n d n :
R R
The result then follows from the integrals d D 1, d D 1. Note that
every factor like j .j ˛j /, .j ˛ j /j , .j ˛j /.j ˛ j / commutes with
everything for any j. R
2.7. Œ•=•a .x/ exp Œ i A D i exp Œ i A .dx 0 / Aab .x; x 0 /b .x 0 /. A further
application of Œ •=•c .y/ to the latter gives: exp Œ i A
h Z Z i
00 00 00
.i/.i/ .dx /d .x /Adc .x ; y/ .dy/Aad .x; y/d .y/ C i Aac .x; y/ ;
which coincides with (2.6.15), where we note that exp Œ i A commutes with
everything.
2.8. Using the identities in (2.6.11), (2.6.12), we may write to the leading order
h Z
• • i
ZŒ; D 1 C i e .dx/.i/ A .x/.i/
•.x/ •.x/
Z h @ i

D D exp i C m C C :
i
But the functional integral is from (2.6.27) is equal to C expŒi SC , where the
multiplicative factor C is independent of .; / and, of course, independent
of e as well. Hence from (2.6.15), we have
h Z
• • i
ZŒ; D C 1 C i e .dx/.i/ A .x/.i/ expŒ i SC
•.x/ •.x/
h Z Z
D C 1 e .dx/TrŒ A .x/SC .x; x/ C i e .dx/.dx 0 /.dy/
i
.x/ SC .x; x 0 / A .x 0 /SC .x 0 ; y/ .y/ expŒ i SC :
Therefore to the leading order, exp Œ i SC

A
D ZŒ; =ZŒ0; 0 is equal to
h Z i
1 C i e .dx/.dx 0 /.dy/.x/ SC .x; x 0 / A .x 0 /SC .x 0 ; y/ .y/ expŒ i SC :
Solutions to the Problems 553
In particular carrying the functional derivatives: .i/Œ•=•.y/Œ•=•.x/ of

exp Œ i SC
A
and then setting D 0; D 0, we obtain from the above
equation, to the leading order in A ,
Z
A
SC .x; y/ D SC .x; y/ C e .dx 0 / SC .x; x 0 / A .x 0 /SC .x 0 ; y/;
1
where SC .x; x 0 / D . @ =i C m/ ı .4/ .x x 0 /. SC is worked out in Sect. 3.1
with an appropriate boundary condition.
Chapter 3
3.1. At t D a, the second expression is .1=2/exp.1/ D 0, and f .t/ is
continuous at this point and vanishes. The continuity at t D 0 is obvious
too. For t ! a from below, write t D a
;
! C0. Then the third
expression

is
1 .1=2/ exp 2.a=
/ 1 ! 1 for
! C0, and f .t/ is continuous at
t D a as well and is equal to 1.
f (t)
1
−a a
0
The derivative is given by

8
ˆ
ˆ0; t < a;
ˆ
ˆ
a
< a exp 2 tCa 1 ; a t < 0;
0 .tCa/2
f .t/ D
a
ˆ
ˆ .ta/2 exp 2 tCa 1 ; 0 t < a;
a
ˆ
:̂
0; a t:
The continuity of f 0 .t/ is immediate by noting, in the process, that at t D ˙ a

we have .1=
/2 exp.1=
/ ! 0 for
! C0, and is obviously continuous
at t D 0. It is easy to see that f 0 .t/ provides a continuous representation of a
Dirac delta for a ! 0, which peaks at t D 0 and vanishes for t < a; t a.
3.2. The left-hand side of the equation in question is given by
Z
.dp/ p C m ip.x0 x/ @
e C m D ı .4/ .x0 x/;
.2/4 p2 C m2 i
i
where we have used the identity . p C m/. p C m/ D . p2 C m2 /.

3.3. For a weak external electromagnetic potential we have seen in (3.2.14) that
Z
A
SC .x; x 0 / ' SC .x x 0 / C e .dx 00 / SC .x x 00 / A .x 00 /SC
A
.x 00 ; x 0 /:
Using the expression for SC .x x 0 / in (3.1.10) for x0 > x 00 , and the Fourier
transform of the field A .x/ in (3.2.18), we obtain for the e-dependent part
of h .x/iA in (3.2.13)
Z
d3 p0 d3 p 0 0
e2 .i/2 3 0 0 3 0
. p0 C m/ . p C m/A . p0 p/eip x eip x ;
.2/ 2p .2/ 2p
for x 0 0 x 00 0 x 0 . Using the projections in (I.21) for . p 0 C m/,

. p C m/, as well as the contribution to h .x/iA coming from the first term
A
on the right-hand side of the equation for SC .x; x0 / above, as given in (3.2.17),
the result in (3.2.15)/(3.2.16) follows.
3.4. We explicitly have
Z
h0C j .x/ j 0 i d3 p
Di . p C m/ .p/ eipx :
h0C j 0 i .2/3 2p 0
Upon using the projection for . p C m/ in (I.22), the result follows.

3.5. The e-dependent part of SCA
.x 0 ; x/, with x0 x 000 x 00 is explicitly given
by
Z
d3 p 0 d3 p 0 0
e.i/2 . p C m/ . p 0 C m/ ıA . p 0 p/ eipx eip x
.2/ 2p .2/3 2p 0
3 0 0
Z X
d3 p 0 d3 p 0
D 3 0 0 3 0
Œi v.p; / eipx Œi e v.p; / v.p 0 ; 0 /
0
.2/ 2p .2/ 2p

0
v.p 0 ; 0 / eip x ıA . p 0 p/;
where we have used the projection in (I.22). Moreover by writing the e-

A
independent part of SC .x 0 ; x/, i.e., SC .x 0 ; x/, as
Z X
0 m d3 p 0 m d3 p 3 p
0
SC .x ; x/ D 0 0 .2/3 p 0 .2/3
.2/ ı 0 ı .3/ .p 0 p/
0
p m
;
0 0
Œi v.p; / eip x v.p 0 ; 0 / eip x ;
the result stated for h 0C j .x/ j 0 iıA =h0C j 0 iıA in (3.3.14) follows,
by multiplying SC A
.x 0 ; x/ by .x 0 /, and integrating over x 0 as indicated
in (3.3.13).
3.6. This is simply obtained by taking the absolute value squared of (3.3.41), and
restrict the p-integration as indicated in the region specified by , picking
up the spin , and dividing by the normalization factor N.
3.7. By using the projection given in (I.21), we have
1 X ˇˇ ˇ2 1

u.p 0 ; 0 / 0 u.p; / ˇ D Tr . p 0 C m/ 0 . p C m/ 0
2 0 8m2

1 ˚ 0
D p p Tr Œ 0 0 C m2 Tr Œ 0 0 ;
8m2
where we have used the fact that the trace of an odd number of gamma
matrices is zero. Finally using the identities

Tr Œ 0 0 D 4 0 0 0 0 C 0 0 ; Tr Œ 0 0 D 4;
and pp 0 D jp j2 cos . p 0 /2 , jp j2 D . p0 /2 m2 , .1Ccos / D 2 cos2 =2; .1

cos / D 2 sin2 .=2/, the result follows.
3.8. PC . p/, P . p/, defined in (I.21), (I.22), satisfy, in particular, the following
equations:
.P˙ . p/ 0 / D P˙ . p/ 0 ; .2m/2 PC . p/ 0 PC . p/ D 2p 0 . p C m/;

.2m/2 P . p/ 0 P . p/ D 2p 0 . p C m/; P˙ . p 0 ; p// 0 P . p 0 ; p/ D 0:

Using the anti-commutation relation f a .y/; b .y 0 /g D ıab ı 3 .y y 0 /, for
y 0 D y 0 0 .D 0/, in (3.5.11), we note that due to the last identity in the above
set of equations that the cross term in f a .x/; b .x 0 /g, for arbitrary x; x 0 ,
vanishes and we obtain from the remaining identities above
Z
0 @ d3 p h ip .xx 0 / ip .xx 0 /
i
f a .x/; b .x /g D Cm e e :
i ab .2/3 2p 0
Upon changing the variable p ! p in the second term in the above

integrand, the statement in the problem follows.
3.9. Note that . @=i C m/ab expŒ i p .x x 0 / Œsin. p 0 p 0 0 / becomes
simply
0ab p 0 =i at equal times, and the p 0 =i factor cancels the product of
i and 1=p 0 in the integrand in .x x 0 /, giving 0ab ı 3 .x x 0 /. The
result then follows by multiplying by 0 .
3.10. The integral may be expressed in terms of the sine function, or equivalently
Z Z C1
" .dQ/
ID A .Q/ e iQ x
d ei .
Q=2/
2 .2/4 1
Z h
.dQ/ 1
Q 2 1
Q 4 i
D
A .Q/ 1 C eiQ x ;
.2/4 3Š 2 5Š 2
and is an odd function in

.
0 i
3.11. From (I.15) 4i Œ i ; j " i j k D ˙k . Moreover Œ 0 ; i D 2 . Hence
i 0
the expressions F0 i D F 0 i D E i , Fij D " i j k B k , lead from
i i i
Œ ; F D 2 Œ 0 ; i F 0 i C Œ i ; j Fij ;
4 4 4
to the stated result, consistent with (3.7.13) for E; B along the third axis.
3.12. The formal substitution s ! i s amounts in replacing (3.8.5) by
Z
2 ImW .e/ 1 1
ds ism2 h .seE/2 i
D 2 Re i e sjeEj coth sjeEj 1 :
VT 4 0 s3 3
The expression within the square brackets is real. Accordingly, we may make
2
the replacement ReŒ i eism D sin.sm2 /, leading to an integrand which is an
even function of s. Thus the above integral may be rewritten as
Z 1 h i
2 ImW .e/ 1 ds 2
.seE/2
D 2 3
sin.sm / sjeEj coth sjeEj 1 ;
VT 8 1 s 3
Z 1
1 ds ism2 h .seE/2 i
D Im e sjeEj coth sjeEj 1 :
8 2 1 s
3 3
It is precisely because of the .seE/2 =3, term within the square brackets that
the point s D 0 is not a pole of the integrand. By closing the integral from
below as shown in the c.w. direction in the complex s-plane, we enclose
all the poles in==jeEj; n D 1; 2; : : :, and note that i i .Ims/ m2 D
2
m2 .Ims/, thus em .Ims/ ! 0 for .Ims/ ! 1.
s−plane
× − iπ /|eE|
×−2iπ /|eE|
For s '
in=jeEj,
with n D 1; 2; : : :, the integrand approaches:
2
eism = s2 s C jeEj
in
, and the residue theorem gives
X1 1
˛ E 2 X e n m =jeEj
2
2 ImW .e/ 1 jeEj2 n m2 =jeEj
D Im .2i/ e D ;
VT 8 2 nD1
n2 2 2 nD1 n2
coinciding with the one in (3.8.8).

3.13. The identity follows by considering the chain of equalities
1 1

.. p C •p/ C m/ . p C m/
1 h i 1
D p C m . p C •p/ C m
.. p C •p/ C m/ . p C m/
1 1
D . •p/ ;
.. p C •p/ C m/ . p C m/
and the identity follows upon taking the limit •p ! 0.

3.14. Due to the matrix nature of U.x/, @ U.x/ and U.x/ do not commute. From
the gauge transformation in (3.10.3) A ! A0 D U A U 1 C.i=g/U@ U 1 ,
hence using the fact that @ .U 1 U/ D 0 i.e., U 1 .@ U/ D .@ U 1 /U,
then following expression emerges for @ UA U 1 C.i=g/ U@ U 1 D
h i i i
U.@ A /U 1 U@ U 1; UA U 1 .U@ U 1 / .U@ U 1 /C U@ @ U 1 :
g g
Accordingly,
h i
@ A0 @ A0 D U.@ A @ A /U 1 U@ U 1; UA U 1
h i ih i
C U@ U 1; UA U 1 U@ U 1; U@ U 1 ;
g
h i
i g Œ A0 ; A0 D i g UŒ A ; A U 1 U@ U 1; UA U 1
h i ih i
C U@ U 1; UA U 1 C U@ U 1; U@ U 1 :
g
The last two equations give

@ A0 @ A0 ig Œ A0 ; A0 D U @ A @ A ig ŒA ; A U 1 ;
which provides the gauge transformation of G D @ A @ A

ig Œ A ; A .
3.15. .@=i@z /Cm SC .zy/ D ı.4/ .zy/. Upon multiplying the latter, from the
1 1
right, by SC .yx/, and integrating with respect to y, we obtain : SC .zx/

.4/ .4/

D .@=i@z / C m ı .z x/ D ı .z x/ @ =i @ x C m . Hence

1 @
exp Œ iF.z/ SC .z x/ C ı .4/ .z x/ F.x/
@x
h @ i
!
D exp Œ iF.x/ ı .z x/
.4/
C @ F.x/ Cm
i
h @ i

D ı .4/ .z x/ C m exp Œ iF.x/ D SC 1
.z x/ exp Œ iF.x/ ;
i
which is the statement in the problem.
Chapter 4
4.1. Using the facts that h 0 j P D 0, and h 0 j P 0 D m h 0 j, we may infer
from (4.2.23)–(4.2.31) that
h 0 j exp Œ i˛J 03 exp Œ i n J P i D R ij h 0 j Y j exp Œ i˛J 03 exp Œ i n J

D R ij m sinh ˛ ıj 3 h 0 j exp Œ i˛J 03 exp Œ i n J :
With n given in (4.2.24), we may refer to (2.2.11) to infer that R33 D cos ,
R23 D sin sin , R13 D cos sin , and use the fact that sinh ˛ D
jp j=m, to conclude that the state in (4.2.23) becomes multiplied by p by
the action of P, where p is given in (4.2.33). On the other p hand, Y 0 D
0 3
P cosh ˛ C P sinh ˛, as given in (4.2.30), and m cosh ˛ D jp j2 C m2 ,
which verifies (4.2.32) for the application of P 0 as well.
4.2. From (4.2.14) we note that W 0 commutes with J. On the other hand,
consider the expression: K.˛/ D exp Œ i˛ J03 W 0 exp Œ i˛ J03 , with B.C.
K.0/ D W 0 . We explicitly have K 0 .0/ D W 3 , K 00 .˛/ D K.˛/, with the
latter two equalities following from (4.2.14), from which we obtain K.˛/ D
W 3 sinh ˛ C W 0 cosh ˛. This leads to the statement of the problem by finally
noting that h 0jW 0 D 0, h 0jW 3 D mh 0; j.
4.3. If W D 0, then we may write W D 0 P , and there is nothing to
prove. Accordingly suppose that W ¤ 0, that is, jW 0 j D jWj > 0. From
the orthogonality of the two vectors: P 0 W 0 D jPj jWj cos or jW 0 jjP 0 j D
jPj jWj j cos j from which j cos j D 1. That is, W D P. On the other
hand, the relation W 0 P 0 D W P D jPj2 also implies that we also have
W 0 D P 0 , since P 0 D jPj > 0, which completes the proof.
4.4. W 0 D .1=2/ " ij k Pi Jj k . But .1=2/ " ij k Jj k D J i , hence W 0 D P: J. On the

other hand, W i D .1=2/" iP J . The latter may be rewritten as P 0 J i C
" 0 ij k Pj J0 k . But J 0 k D N k , from which W D P 0 J P N.
4.5. Upon writing A .x/ D tc Ac .x/, and using the commutation rule Œta ; tb D
i fabc tc , as given in (3.10.4), the result follows.
4.6. We note that ˛ D @L =@.@ A˛ .x// D F ˛ . Hence from (4.4.6)
i ˛ ˇ
˝ D F .S /˛ Aˇ C F ˛ .S /˛ ˇ Aˇ C F ˛ .S /˛ ˇ Aˇ D F A:
2
Maxwell’s equations: @ F D 0, give @ ˝ D F @ A . Accordingly,

from (4.4.16)
T D ..1=4/ F ˛ˇ F˛ˇ F ˛ @ A˛ / F @ A
D .1=4/ F ˛ˇ F˛ˇ C F ˛ F ˛ ;
which is obviously symmetric and gauge invariant, i.e., under the transfor-
mation A˛ ! A˛ C @˛ , for the latter. Using the elementary identity
@ F˛ˇ D @˛ F ˇ @ˇ F ˛ , and the asymmetry of F ˛ˇ the conservation law
follows @ T D 12 F ˛ˇ @ F˛ˇ C F ˛ @ F ˛ D 0. T is traceless.

4.7. @L =@.@ / r D i . Recall from (I.7) that S D .i=4/Œ ; . Hence
from the definition of ˝ in (4.4.6), taking into account the adjoint field
contribution,
1 1
˝ C .˝ / D f ; S g C f ; S g C f ; S g
2 4
i
D :
8
Upon using the Dirac equations: @ D i m , @ D i m , we

obtain
1 i ! !
@ ˝ C .˝ / D @ @ ;
2 4
L D 0, and from (4.4.16)
1 i ! !
T D i @ C i @ C @ @
2 4
1 ! !
D @ C @ ;
4i
which is obviously symmetric.
4.8. The general expression for the energy-momentum

! tensor, given in the

!
previous problem is T D .1=4 i/ @ C @ . In particular,
1 !0 1 0 !k !
T 00 D
@ ; T0 k D @ C k @0 :
2i 4i
!

The Dirac equations give @0 D 0 . r Ci m , @0 D r i m 0 .
The right-hand sides of these two equations together with the identity
k j D .1=2/Œ k ; j kj , lead to the following expressions
1 ! 1 1 !k
T 00 D mC r ; T0 k D @j
Œ j ; k C
@ :
2i 8i 2i
4.9. This commutation relation follows by the application of the general equal
commutation rule of the commutator of bilinear forms:
h i
Œ a .x/; b .x/; Œ c .x 0 /; d .x 0 /
D 2 ıbc ı 3 .x x0 / Œ
a .x/; d .x
0
/ 2 ıad ı 3 .x x0 / Œ 0
c .x /; b .x/: . /
Note that the equal-time anti-commutation relations in (4.3.41) imply that

since b .x/ D i b 0 , f a .x/; b .x 0 /g D ı 3 .x x0 /, f a .x/; b .x 0 /g D 0.

To prove the identity . /, note that the following ones, in turn
0 0 0 0
a .x/ b .x/ c .x / d .x / D c .x / d .x / a .x/ b .x/
ıad ı 3 .x x0 / 0
c .x / b .x/ C ıbc ı 3 .x x0 / 0
a .x/ d .x /;
0 0 0 0
a .x/ d .x / b .x/ c .x / D d .x / c .x / a .x/ b .x/
ıbc ı 3 .x x0 / d .x
0
/
a .x/ C ıad ı 3 .x x0 / 0
b .x/ c .x /;
are sufficient to establish the identity in question, by a mere relabeling of the

spinor indices, and by the fact that
h i
@k ı 3 .x x0 / c .x 0 / b .x/ C c .x/ b .x 0 /
h i
D @k ı 3 .x x0 / c .x 0 / b .x 0 / C c .x/ b .x/ :
4.10. The right-hand side of (4.7.129) is given by:
. p 0 pi /=m2 C e 00 e i0 D 0; for D 0; D i;

0 0 2
p p =m C e 0 0
0 e0 D .p C m2 /=m2 C p2 =m2 D 1; for
2
D D 0;

pi pj =m2 C e i 0 2 i j 2 2
j
C e C C e e C . p / p p =.m jpj / D ı ; for D i; D j;
i j ij
using, in the process, the 3D completeness relation for the last relation.
4.11. By multiplying the given equation, in turn, by and @ and by considering

the two resulting equations simultaneously, we are led to the following
equivalent two equations . @ @ , @K D @ K , K D K /
m@ D . @ K/ C @K; D 2 . @ K C @K/=.im2 / 3 K=m;
which when are substituted back in the initial equation in the problem give
@ h 1 @ @ 2 @ @ i

Cm D C C C 2 K
i 3 im im m i i
2 @ h @
@ @

i
C m C m K :
3 m2 i i i i

Upon multiplying this equation by @=iCm , the statement in the problem
then follows from an analysis similar to the one leading to (4.7.143), and by
making
note that the first term in the above equation multiplied by @=i C
m is nothing but K , where is given in (4.7.145) in the momentum
description, while the coefficient of .Cm2 / of the second term is the non-
propagating, non-singular term, mentioned in the statement of the problem.
The equations in (4.7.137), (4.7.138) are also satisfied for K D 0.
4.12. Consider the equation for 0 in (4.7.137) . @=i C m/ 0 D 0 for K D 0.
It may be rewritten as
0@0 0 r @i i r
0
0D C Cm D 0 C C m 0 i i
;
i i i i
where we have used, in turn, the constraint in (4.7.137) and (4.7.138) to
eliminate 0 . The constraint equation in (4.7.148) follows upon multiplying
the above equation by 0 .
4.13. The following two equalities are explicitly verified
@j r j @j
i ˇ ij DC C2 ;
im 3im im
@ i ij @j 1 r j @j r @j r2 j @j
ˇ D C C2 C2 :
i i 3 i im i im 3m im
The equation in question follows directly from combining these two
identities.
4.14. By working in the chiral representation of the gamma matrices, given
in (2.3.3), u.˙1/ may be taken (see (I.25)) and conveniently normalized, as
p C p 0 X
u.C1/ D jpj ; u.1/ D jpj ; u./u./ D p;
0
where ˙ are defined in (I.13). We may then write
1 X i k `
e e u.C/u.C/ C u./u./ k e ` 0 e 0 :
j
PC ij D ./
2 0
;
By considering,
p for example, the vector
p p to be along the 3-axis, and eC D
.1; i; 0/= 2, e D .1; i; 0/= 2, the following are easily established
X
e 0 e0 D ı j 1 1 C ı j 2 2 ;
j
0
X j
p
e 0 e0 u.C/ D 2 e j u./;
0
X p
u.C/ e i e D 2 e i u./;

and with an almost identical analysis carried out for the u./u./ in (*)
completes the verification of (4.7.193).
4.15. In the presence of the external source, (4.7.103), for D 0 becomes replaced
by @ k F k 0 C m2 V 0 D K 0 , and with F k 0 D k , denoting the canonical
conjugate momenta of the components V k , the dependent field V 0 satisfies
the equation
1 0
V0 D K C @ k k /:
m2
By keeping the canonical conjugate momentum k and its space derivative

fixed, we have
• 1 0 .4/ 0
V 0 .x 0 / D ı .x x/; or equivalently
•K .x/ m2
• 1 0 .4/ 0
V .x 0 / D ı 0 ı .x x/:
•K .x/ m2
On the other hand, .i/•=•K .x 0 /h 0C j 0 i D h 0C j V .x 0 / j 0 i, and hence

from (4.6.38),
• •
.i/ .i/ 0
h 0C j 0 i D h 0C j V .x/V .x 0 / C 0 i
•K .x/ •K .x /
•
i h 0C j V .x 0 / j 0 i
•K .x/
i
D h 0C j V .x/V .x 0 / C j 0 i 2 ı 0 0 ı .4/ .x 0 x/h0C j 0 i;
m
from which the statement in the problem follows upon multiplying by i, and
is a consequence of the presence of a dependent field.
4.16. Although V0 .x/ is a dependent field we note that
• • •
.i/ .i/ .i/ h 0C j 0 i
•K .x/ •.x/ •.x/
•
D .i/ h 0C j .x/ .x/ C j 0 i
•K .x/

D h 0C j .x/ .x/V .x/ C j 0 i D h 0C j LI .x/ j 0D i
where the functional derivative •=•K 0 .x/ does not generate an additional
term as in (4.6.38) because .x/ .x/ in the first equality consists only of
independent fields.
Chapter 5
5.1. The solution immediately follows by writing down the rotation matrix
ŒR i k explicitly and the polarization vectors as 3 1 (column) matrices. The
expression for ŒR i k follows from (5.2.6) to be
0 1
cos2 cos C sin2 sin cos .cos 1/ cos sin
ŒR i k D @ sin cos .cos 1/ sin2 cos C cos2 sin sin A :
cos sin sin sin cos
R R
5.2. @ .dx 0 /D .x x 0 / J .x 0 / D .dx 0 /DC .x x 0 /@0 J .x 0 / D .x/:
5.3. Going through these various steps, we have: En .K; R/ exp Œ iEn .K; R/T
d d2 exp Œ i En .K; R/T

D i exp Œ i En .K; R/T D 2 ;
dT dT En .K; R/
r
exp Œ i En .K; R/T 2i @ h n2 2 i
2
D exp i K C T :
En .K; R/ T @K2 R2
Therefore
r
d2 n 1 @ h n2 2 io
En .K; R/ exp Œi En .K; R/T D 2 i 2 2
exp i K2 C 2 T :
dT T @K R
An elementary application of @=@a to the above equation gives the stated

result, where recall that R is expressed in terms of a.
5.4. Clearly
1 ˇˇ i j k j n1 R1 i j k j n2 R2 ˇ2
ˇ
fka .n1 ; n2 / D ˇe e C ei j k j n1 R2 ei j k j n2 R1 ˇ :
C
independently of f .j k j/ where
Z Z ˇ ˇ2
ˇ ˇ
CD d˝1 d˝2 ˇ ei j k j n1 R1 ei j k j n2 R2 C ei j k j n1 R2 ei j k j n2 R1 ˇ :
Upon using the elementary integrals,

Z h cos ka
1
sin ka i h sin2 ka i
dx x ei k a x D 2 i ; C D 32 1 C 2 2 ;
1 ka k 2 a2 k a
we obtain the following expression

cos ka sin ka 2 . sin2 ka
hci D 1 C :
ka k 2 a2 k 2 a2
5.5. For a matrix S D 1=ŒA e0 B, we have
@ 1 1 h 1 1 i
D ; ıe0 ! 0
@e0 A e0 B ıe0 A .e0 C ıe0 / B A e0 B
1 h A e0 B A .e0 C ıe0 / B i 1
D
A .e0 C ıe0 / B ıe0 A e0 B
1 1
! B ;
A e0 B A e0 B
and the result follows by matrix multiplication with matrix elements indices
specified by spacetime variables.
5.6. From (5.7.3) and (5.7.4), we have

@ h 0C j j j 0 i D i e0 h 0C j j 0 i h 0C j j 0 i ;
upon taking matrix element between vacuum states. The first result then
follows from the explicit equations for h 0C j .x/ j 0 i in (5.7.13),
h 0C j .x/ j 0i in (5.7.14), and (5.7.7). Now write h 0C j 0 i D
FŒ•=•J h 0C j 0 i0 in (5.7.27). Then
Z

@ h 0C jA .x/ j 0 i D FŒ•=•J .dx 0 /DC .x x 0 /@0 J .x 0 /h 0C j 0 i0 :

ˇ
Upon using the identity: FŒ•=•J J .x/ D J .x/C•FŒT=•T .x/ ˇT D•=•J , the
expansion
Z
SC .y; y I e0 T/ D SC .xx /ie0 .dy/ SC .yy1 / T .y1 / SC .y1 y 0 /C ;
0 0
and the identity

Z
0 •
.dx 0 /.dy1 /DC .x x 0 /@x SC .y y1 / T .y1 / SC .y1 y 0 /
•T .x 0 /
D i Œ DC .x y/ SC .y y 0 / DC .x y 0 / SC .y y 0 / ;
O factor leads to the stated result.

which applied to the expŒ i SC .: I e0 A/ R
O
By the same method just developed, the other factor exp Œ de0 TrŒ A.:/S C
0O
.:I e A/ gives no contribution (see also Appendix IV).
5.7. Write the exponential in (5.8.2) involving .i/•=•J , acting on h 0C j
0 i0 , as
exp i Œ A0 C e0 A1 C e20 .A2 C iB2 / C e30 A3 C D

h i 1 i
1 Ci e0 A1 Ci e20 A2 C iB2 C A21 C i e30 A3 A31 C iA1 A2 Ci B2 C ei A0;
2 3Š
up to third order, and where B 2 is the coefficient of e20 in (5.8.6). Obviously,

a0 D A0 as the zeroth order involving no functional differentiations. Also
a1 D Œ S S D J in a matrix notation. Clearly, A31 involves only
disconnected parts as it involves only three photon propagators since to third
order they cannot connect together three different expressions of the form
Œ S S . B 2 leads to .1=2/ Œ .i/ D1 2 C D1 1 J 1 D 2 2 J 2 K 1 2 , and
the first term in the latter .1=2/.i/D1 2 K 1 2 involves no external lines
and should be omitted due to the normalization condition of h 0C j 0 i.
For A21 , the only connected term is Œ S 1 S Œ S 2 S .i/D1 2 . The
application of A2 is straightforward and leads only to connected parts, and so
is for A3 . The application of iA1 A 2 leads to

i S 1 SŒS 2 S 3 .i/D1 3 D 2 2 J 2 C

C D1 1 J 1 D 2 2 J 2 D 3 3 J 3 C .i/D1 2 D 3 3 J 3 C .i/D 2 3 D1 1 J 1 :
Clearly the second and the fourth terms within the round brackets give rise to
disconnected parts and are to be omitted. Finally, A1 B 2 leads to
i
h i
Œ S 3 S K 1 2 D1 3 D 2 2 J 2 C D 2 3 D1 1 J 1 ;
2
retaining only connected parts.
5.8. This involves four terms. The term obtained by multiplying the first term
in (5.9.23) by its complex conjugate is given by
1
Œu.p 02 ; 02 / u.p 2 ; 2 /u.p 01 ; 01 / u.p1 ; 1 /
. p 2 p 20 /4
u.p1 ; 1 / u.p 01 ; 01 /u.p 2 ; 2 / u.p 02 ; 02 /
1 1
D Tr Œ . p 01 C m/ . p1 C m/
.2m/4 . p 2 p 20 /4
Tr Œ . p 20 C m/ . p 2 C m/ ;
P
where we have used the relation ua .p; /ub .p; / D . p C m/ab =2m.
Application of the properties of the gamma matrices readily gives the first
term in (5.9.32). It is easy to see that the two cross terms are identical, and
direct applications of the method just given for the first term leads to the other
two.
5.9. The four-spinors of the ingoing electrons are given by
0 1 0 1
s 1 s 0
p0 Cm B 0 C
B C ; u.p 2 ; / D p 0Cm B C
B 1 C ;
u.p1 ; C/ D
2m @ 1 A 2m @ 0 A
i i
0 1
p
where D ˇ=. C 1/ D ˇ=.1 C 1 ˇ 2 /. The four spinors of the outgoing
electrons may be written as
s ! s !
p0 C m 1 p0 Cm 2
u.p 01 ; 01 / D p 01 ; u.p 02 ; 02 / D p 01 ;
2m p 0 Cm 1
2m p 0 Cm 2
where
n1 D .0; sin1 ; cos 1 /; n 2 D .0; sin 2 ; cos 2 /; p 01 D m ˇ .1; 0; 0/ D p 02 ;
j D .ei=4 cos.j =2/ ei=4 sin.j =2/> ; j D 1; 2:
From (5.9.23), we have for the amplitude of the process

h0 1 i
1 0

A/ 1 2 :
1 1 0 2 0 1 1 2
This leads to jAj2 =2 D PŒ1 ; 2 as given in the problem, satisfying the

completeness relation:
PŒ1 ; 2 C PŒ1 C ; 2 C PŒ1 ; 2 C C PŒ1 C ; 2 C D 1:
If only one spin is measured: PŒ1 ; D PŒ1 ; 2 C PŒ1 ; 2 C D 1=2,

and similarly PŒ; 2 D 1=2. The obvious correlation between the two
electron spins, that is the measurement of the spin of one is correlated
with the spin value of the other, is expressed by the fundamental general
relation PŒ1 ; 2 ¤ PŒ1 ; PŒ; 2 . It is interesting to note that the above
probabilities are independent of the speed ˇ of the particles. This is due to
the special choice of axes of measurements and spin orientations chosen in
the problem. More generally polarization correlations depend on the speed of
the underlying particles (see, e.g., Sect. 5.9.2 and references therein).
5.10. . p C m/2 D p2 C 2 m p C m2 D . p2 C m2 / C 2m. p C m/ from which
the identity follows.
5.11. From the method of the Feynman parameter representation, as given
in (II.34), in Appendix I at the end of the book,
Z Z 1 Z
1 2 .dk/
.dk/ 2 D 2 2 x dx
3
.k C m2 /2 .k2 C 2 / 0 k C C .m2 2 /x
2 2
Z 1 Z 1 h i
2 2 x dx 2 2
D i D 2 dx 1 ;
0 2 C .m2 2 / x m 2 0 2 C .m2 2 /x
using the integral (II.8). The stated result then follows upon carrying out the
elementary x-integral, and then considering the limit 2 ! 1.
5.12. Using the integral
Z
x dx 1
2 a 1 x a=2
D ln x axCa C p arctan p ;
x 2 ax C a 2 2 a .a=2/2 a .a=2/2

for a > 0, we obtain with a D 2 =m2 ! 0, Cir D .1=2/ ln 2 =m2
C.=2m/Œ.=2/CO.=m/. On the other hand 2 @Cir =@2 D Cir CDir .
5.13.
Z Z Z 1
.dk/ dx
I D .dk/ ;
Œ . p k/ C m2 Œ k2 C 2
2
0 F.k2 ; x/

2
where F.k2 ; x/ D k2 C . p 2 C m2 /x.1 x/ C m2 x 2 C 2 .1 x/ , after
having made a shift of the (integration) variable k ! kCpx. Upon integration
over x by parts, and using the elementary integral (II.8), we obtain I D
i 2 Œ Cuv C I 2 , where Cuv is given in (5.10.21), and

Z 1
2 2 x .1 2x/ dx
I2 D . p C m /
0 Œ.p2 C m2 / x.1 x/ C m2 x 2 C 2 .1 x/
Z 1
x .2 m2 x 2 / dx
C :
0 Œ . p 2 C m2 / x.1 x/ C m2 x 2 C 2 .1 x/
.1/
In the first integral I 2 , we may set . p 2 C m2 / D 0, in the denominator,
.1/
obtaining, I 2 D Œ. p 2 C m2 /=m2 Œ Cir 2 , for 2 =m2 ! 0, where Cir
is given in (5.10.22). The second integral may be written, up to first order in
. p 2 C m2 /, as
Z 1 Z 1
.2/ x 2 .1 x/.2 m2 x 2 / dx x .2 m2 x 2 / dx
I 2 D . p 2 C m2 / C :
0 Œ m2 x 2 C 2 .1 x/ 2 0 Œ m2 x 2 C 2 .1 x/
Both integrals are elementary except, perhaps, the integral

Z 1 Z 1
x 2 .1 x/ dx x 2 dx
2 ! 2 ;
0 Œ m2 x 2 C 2 .1 x/ 2 0 Œ m2 x 2 C 2 .1 x/ 2
where we have used, in the process, the value of the second integral Dir in
Problem 5.12, to set the second term to zero. The above integral, multiplied by
.2/
2 , vanishes
like O.=m/. Accordingly, I 2 D 2 C Œ . p 2 C m2 /=m2 Œ 3
2 Cir . Hence I 2 D 2 C Œ . p 2 C m2 /=m2 Œ Cir 2 C 3 2 Cir ,
Z h i
.dk/ 2
p C m 2
D i Cuv C2 2 Cir 1 CO . pCm/ ;
Œ . p k/2 C m2 Œ k2 C 2 m
where we have used the identity in Problem 5.10. The statement of the
problem follows upon multiplying this integral by ie2 Œ. C 1/. p C m/
2m=.2/4 .
5.14. From (5.10.57), the modified
Coulomb potential,
to second order may be
0 2 0
written as U .x/ D q Z3 C .˛=3/ U .x/ =.4jxj/, where
Z Z r
0
1
dM 2 M jxj 1
dM 2 2 m2 4 m2
U .x/ D 2
e C 2
1C 1 1 eM jxj :
.2m/2 M .2m/2 M M2 M2
The first integral is readily expressed in terms of the exponential integral

function,6 and it is given by the equivalent integral with asymptotic
6
See, e.g., I. S. Gradshteyn and I. M. Ryzhik (2000). Tables of Integrals, Series and Products (6th
ed.). San Diego/San Francisco: Academic Press. pp. 875–877.
behavior
Z 1 1
et
I1 D 2 dt D 2 Ei.2mjxj/ ' 2 E C 2 ln C O.mjxj/;
2mjxj t 2mjxj
where E D 0:5772157 : : : denotes Euler’s constant. On the other hand,

it is justifiable to take the limit mjxj ! 0 inside the second integral. This
amounts to evaluate the integral
Z r
1
0 dM 2 2 m2 4 m2
I 2 2
1 C 1 1 :
.2m/2 M M2 M2
Upon introducing the integration variable z D 1 .1 4m2 =M 2 /1=2 , the

above integral simplifies to
Z h
0
1
22 i 1 5
I D
2 dz 1 2z C z D 2 ln C :
0 z2 2 6
All told, we obtain for mjxj 1,

0 q2 2˛ h 1 5i
U .x/ D Z3 C ln E :
4jxj 3 mjxj 6
5.15. From (5.10.42), u.p 0 ; / ˘ .k/ .1=k2 /u.p; / is equal to
k
u.p 0 ; / u.p; / ˘.k2 / u.p 0 ; /. p 0 p/u.p; / :
k2
The Dirac spinors satisfy the equations . pCm/u.p; / D 0, u.p 0 ; /. p 0 C

m/ D 0. That is the last expression above is zero. The statement of the
problem then follows from (5.10.44), (5.10.49), by taking the limit k2 ! 0
in this order, where ˘.0/ D 1 Z 3 .
5.16. Using the normalization condition of .x/, the change in potential energy is
Z Z
.x 0 / 1 i 1 1
U.x/ D ˛ d3 x0 D ˛ d3 x0 .x0 / :
jx x0 j jxj jx x0 j jxj
leading to an approximate energy shift

Z Z
2 3 0 0 3 1 1
•E D ˛j'n0 .0/j d x .x / d x :
jx x0 j jxj
The x0 - integrand, multiplying .x0 /, is easily worked out, for example, by

a Legendre polynomial expansion,7 to be
Z Z
1 1 2 2
d3 x D 2 djxj jxj2
C 0 D 2jx0 j2 =3;
jx x0 j jxj jxj<jx 0 j jx0 j jxj
R
and d3 x0 .x0 / jx0 j2 D 12 2 . This gives
8 ˛4 m 2 2
•E D m :
n3
5.17. F k A D F k0 A0 C F ki Ai . Using the notation .i/•=•J b

A , note that
b b
A 0 .y/ F .x/h 0C j 0 i is equal to
k0
@k
b
A0 .y/h 0C j F k 0 .x/ j 0 i D h 0C j F k 0 .x/A0 .y/ j 0 iCi ı .4/ .xy/h 0C j 0 i;
C r2
where we have used (5.14.27) and (4.6.38). Hence
h @k i
h 0C j F k .x/ A .y/ C j 0 i D bF k .x/ b
A .y/ i 2 ı .4/ .x y/ h 0C j 0 i:
r
5.18. We note from the just mentioned equation that
a D @ 1 a 0 3 0 a
3 @ @ A @ A ; a D 1; 2;
@ a a D @ 1 0 a 3 3 1 0 2 3
3 @ .@ a @ A @ @ a A / D @ 3 @ r A :
a
where we have used the relation @ a Aa D @ 3 A3 , in writing the last equality.

From the equal-time commutation relations:
ŒAa .x 0 ; x0 /; b .x 0 ; x/ D i ı ab ı .3/ .x0 x/; a; b D 1; 2;
and the above two equations in turn, lead to
ŒAa .x 0 ; x0 /; @31 @ b @ 0 A3 .x 0 ; x/ ŒAa .x 0 ; x0 /; @ 0 Ab .x 0 ; x/ D i ı ab ı .3/ .x0 x/;

@ a .3/ 0
ŒAa .x 0 ; x0 /; @ 1 0 3 0
3 @ A .x ; x/ D i ı .x x/;
r2
P1
7
Recall: .jx x0 j/1 D .1=r> / nD0 .r< =r> /n Pn .cos /, in a standard notation, P 0 .cos / D
R1
1; P1 .cos / D cos , 1 d cos Pn .cos / Pn 0 .cos / D 2 ın n 0 =.2n C 1/.
from which the following key equal-time commutation relations emerges

@a@b
ŒAa .x 0 ; x0 /; @ 0 A b .x 0 ; x/ D i ı ab 2 ı .3/ .x0 x/; a; b D 1; 2:
r
The expression A 3 D .@ a =@3 /Aa then leads to the equal time commutation
relations in question.
5.19. The Fourier transform of the left-hand side of the identity reads ˛ˇ
˛ˇ
.˛j kj kˇ =k2 / .ˇj kj k˛ =k2 / C .k˛ kˇ =k2 / =k2 , and coincides with DC .k/
(see (5.14.13).
5.20. Denote the left-hand side by K. This gives
• • • h i

.i/ K D exp i e0 @ SC .x :/ .:/ exp i SC
•.x/ • •
h • i
D SC .x :/ .:/ C i e0 @ .:/ K;
•.:/
1
which upon multiplying by SC .z x/ and integrating over x, gives
Z h
1 • • i
.dx/ SC .z x/.i/ K D .z/ C i e0 @z .z/ K
•.x/ •.z/
which, after multiplying it by exp Œ i e0 .z/ , may be rewritten as

Z h i
1 •
.dx/ exp Œ i e0 .z/ SC .z x/ C ı .4/ .z x/ e0 @ .x/ .i/ K
•.x/
D exp Œ i e0 .z/ .z/K:
But
1

exp Œ i e0 .z/ Œ SC .z x/ C ı .4/ .z x/ e0 @ .x/
1
D SC .z x/ exp Œ i e0 .x/ ;
(see Problem 3.15). Hence upon multiplying the former equation by:
exp Œ i e0 .y/ SC .y z/, and integrating over z , we obtain
Z
•
.i/ KD .dz/ exp Œ i e0 .y/ SC .y z/ exp Œ i e0 .z/ .z/ K:
•.y/
Functionally integrating over .y/ leads to the right-hand side of the equation
stated in the problem, and incidentally satisfies the appropriate boundary
condition for e0 ! 0.

5.21. Upon setting e0 •=•.z/ •=•.z/ D fO .z/, with ! , ! ,
J ! K , we have from (5.15.1),
hZ
@ @ • i ˇ
ˇ
exp .dx/ J .x/ FŒ ; ; K
; ˇ
•K .x/ K D0
hi i
i
D exp .fO C J /D ./.fO C J / exp .@ J / G .@ J /
2 2
h @ i

˛ ˛
exp ifO @ 2
J exp i SC ;

where G is defined in (5.15.25). From Problem 5.20, we also have, with
h @ i
˛ ˛
D e0 J ;
2
h @ i

˛ ˛
exp ifO @ J exp i SC D expŒ i ei SC ei :
2
Since we eventually have to set the external Fermi sources to zero, we may
make a change of these source variables,
! ei ; ! ei ;
and use the invariance of fO , under such a transformation, to reach the

statement made in the problem by finally using, in the process, (5.15.20)–
(5.15.22).

5.22. Set e0 •=•.z/ •=•.z/ D fO .z/. Then from (5.15.20),
• • ˇ
ˇ
.i/ .i/ FŒ ; ; J ; D 0 ˇ
•.x/ •.y/ D 0; D 0
• • ˇ
D .i/ .i/ O FŒ ; ; K ; ˇˇ
exp Œ Q ; ./
•.x/ •.y/ D 0; D 0; K D 0
h • • i Z h i •
O D e0 aQ y
Q Q
a
C .dx 0 / aQ 0 @ 0 J .x 0 / :
•K .y/
x
•K .x/
•K .x 0 /

O may be more conveniently rewritten as

The first term in Q
Z
•
e0 .dx 0 / aQ 0 Œ ı .4/ .x 0 x/ ı .4/ .x 0 y/ :
•K .x 0 /
O generating translations in K , the right-hand side of the former

With exp Œ Q
equation ./ is given, in matrix multiplication notation in spacetime, by
• • hi i
exp Œ i .J / .i/ .i/ exp .fO C J /D .fO C J /
•.x/ •.y/ 2

ˇ
ˇ
exp ifO @ exp i SC ˇ ;
D 0; D 0; K D 0
Z h i
.z I x; y/ D .dz 0 / G.z z 0 / e0 ı .4/ .z 0 x/ ı .4/ .z 0 y/ @ J .z 0 / ;
1
.J / D J .:/ @P . : I x; y/ C g. : I x; y/ G.: :/ g. : I x; y/;
2

g.z I x; y/ D e0 ı .4/ .z x/ ı .4/ .z y/ @z J .z/:
where G.z z 0 /
is defined in (5.15.25)

with
an ultraviolet

cut-off. From

Problem 5.20: exp ifO @ exp i SC D exp i.ei / SC .ei / .
Hence upon defining sources
T D ei ; T D ei ;
using the chain rule: .•=•/ D ei .•=•T/, and simply evaluating the
functional .J / above, we obtain
• • ˇ
ˇ
.i/ .i/ .i/ FŒ ; ; J ; D 0 ˇ
•.x/ •.y/ D 0; D 0;

ˇ
2 • • ˇ
D ei$ ŒJ ei e0 G.0/G.xy/ .i/ .i/ .i/ FŒT; T; J ; ˇ ; ./
•T .x/ •T.y/ TD0;TD0;
Z
1
$ ŒJ D .@ J / G .@ J / e0 .dz/J .z/@z Œ G.z x/G.z y/ :
2
Upon dividing (**) by FŒ 0; 0; J ; D 0 , as given in (5.15.24), the

statement of the problem follows.
Chapter 6
6.1. For infinitesimal transformations
V.x/ ' I C ig o c .x/tc , and the transfor-
mation rule A ! VA V 1 C i .V/@ V 1 =go , defined in (6.2.4), gives
A !
A Cigo Œ tc ; tb c Ab Ctc @ c , where we have used the relation A D tb Ab .
This leads to the following infinitesimal transformation, upon using the anti-
symmetric nature of the structure constants,

Aa ! Aa C rac

c ;
rac D ıac @ C go fabc Ab :
6.2. We explicitly have

Œr ; r cb D go fcab .@ Aa @ Aa / C g2o fcda faeb fcea fadb Ad Ae :
Using the identity fcda faeb fcea fadb D fcba fade D fcab fade , the result follows

upon factoring out go fcab , and using the definition of Ga .
6.3. From the commutation relation in (6.2.7), established in the previous prob-
lem, we have
rab rbc G

c D rab rbc Gc C go fabc Gb Gc
D rab rbc G

c D rab rbc Gc ;
where in going from the first line to the second we have used the anti-

symmetry of fabc . In the last equality we have used the fact that Gc D Gc ,
and relabeled $ , thus establishing the equality.
6.4. By an integral representation of the delta functional, up to an unimportant
multiplicative constant, the left-hand side becomes
Z Z Z

˘bx Db .x/ exp i Œ .dx/ a .x/@ Aa C a .x/a .x/ D ˘bx Db .x/
2
Z Z
i i
exp Œ .dx/a .x/a .x/ exp Œ .dx/@ Aa .x/ @ Aa .x/;
2 2
upon completing the squares in the exponential, and shifting the variable a .
The result follows after integration over the latter variable.
. /
6.5. From Problem 6.1: Aa ' Aa C.ıab @ C go facb Ac /b . The constraint :
. /k
@k Aa D 0 gives a ' .@k =@2 /Aka C O.A2 /. When the latter is
. /
substituted back in the expression for Aa , we obtain

A.
a
/
' Œ.@ k @k /=@2 Aa C O.A2 /:
6.6. This expression is obtained from the corresponding differential cross section
for e ! e in Sect. 5.9.3, by replacing the expression
1 X

M2 Tr u.p 0 ; 0 / u.p; / u.p; / u.p 0 ; 0 / in it by
. p 0 =M/. p 0 0 =M/ spins
1 X

M2 Tr h p 0 ; 0 j j .0/jp; ih p j j .0/jp 0 ; 0 i ; where
. p 0 =Mp /. p 0 0 =Mp / p spins
h Œ ; ˛ i
h p 0 ; 0 j j .0/jp i D u.p 0 ; 0 / F1 .Q2 / C Q˛ F2 .Q2 / u.p; /;
4 Mp
as readily follows by invariance arguments and application of the Dirac

equation . p C Mp /u.p; / D 0. Mp denotes the mass of the proton. This
gives
" #
d ˇˇ ˛2 E0 2 Q2 2 2 2 # Q2 2
2
ˇ D 2 4 #
F1 C F2 cos C F1 CF2 sin :
d˝ TF 4E sin 2
E 4 Mp2 2 2Mp2 2
The result follows upon setting: GE D ŒF1 .Q2 =4Mp2 /F2 , GM D ŒF1 C

F2 .
6.7. Let .k; k 0 /, denote the momenta of .e ; eC /, and . p; p 0 / denote the
momenta of q; qN . Using the fact that
X h i
m2e Tr v.k 0 ; 0 / u.k; / u.k; / v.k 0 ; 0 / ! Œk 0 k C k 0 k kk 0 ;
spins
for me ! 0. The corresponding expression for the quarks is then
Œp 0 p C p 0 p pp 0 and hence;
Œ k 0 k C k 0 k kk 0 Œ p 0 p C p 0 p pp 0
D 2 Œk p0 0 kp C k p0 kp 0 / .1 C cos2 #/;
in the CM frame, where k p=jk pj D cos #. That is, # is the angle made by
the momentum of an emerging quark q relative to that of the electron. Hence
X
d=d˝ / 3 e4 .e2f =e2 /.1 C cos2 #/;
f
where ef is the charge of the quark of a given flavor, and the factor 3 is
P three different colors. The cross section then works out to be /
for the
3 e4 f .e2f =e2 /.16 /=3. Upon comparison of this expression with the cross
section for eC e ! C , with masses set equal to zero in (6.5.7), we
obtain
d .1 C cos2 #/ X 2 2 p
D ˛2 3 .ef =e /; s D CM energy:
d˝ 4s f
6.8. (i) By using a Feynman parameter representation and shifting the variable of
integration k, the integrand becomes replaced by:
Z 1 Z x
2 dx dz 1=Œk2 C Q2 .1 x/z3 ;
0 0
which by using the integral representation over k in (II.7) in Appendix II, at

the end of the book, gives
i 1 .1ı=2/ 1 ı Z 1 Z x
2 .1Cı=2/
2 dx .1 x/ dz z .1Cı=2/ :
.4/D 0 =2 Q2 .3/ 0 0
Finally carrying out the z-integral, followed by the use of the integral (III.12),
involving gamma functions, the stated result follows.
(ii) As in part (i), except the xz integrands become simply multiplied by

Œ p1 .1 x/ C p2 z , after the shift of the integration variables k and setting,
in the process, an odd integral in k equal to zero. Finally the xz – integrals
are readily carried out as above leading to the stated result.
6.9. By using the Feynman parameter representation in Problem 6.8 above, and
shifting the k-integration variable again, the denominator of the integrand
becomes simply

k k C . p1 .1 x/ C p2 z/. p1 .1 x/ C p 2 z/;
after setting an odd integral in k equal to zero. The integral involving the k k
part may be ultraviolet-regularized using the integral in (III.8), while the

integral involving . p1 .1 x/ C p2 z/. p1 .1 x/ C p2 z/ may be infrared-
regularized as in Problem 6.8. Finally, the .x; z/ – integrations yield the stated
result in a straightforward manner as in Problem 6.8.
6.10. It is sufficient to spell out, the general infra-red singular structure of the
function hIR .2D 0 =Q2 ; ı/. To this end, we refer to the right-hand sides of the
integrals in Box 6.2 of the regularized integrals in Sect. 6.6. If an integral

depends on p1 , then multiplying it, by p1 gives zero, and if multiplied
by either, p2 , or Q , give a factor Q which cancels out the factor 1=Q2
2
multiplying .2D 0 =Q2 /ı=2 . Similar statements follow if the integral depends

on p2 . On the other hand, if we multiply the first integral by either p1 p2 or Q2 ,
these terms cancel out again the 1=Q2 factor just mentioned. That is, in all the
terms contributing to the fermion-gluon vertex, the 1=Q2 factor multiplying
.2D 0 =Q2 /ı=2 is canceled out in the infra-red regularized part. The most infra-
red singular part in evaluating the vertex function comes from the first integral
ı=2 2
in the Table now involving the factor 2D 0 =Q2 ı=2 . Therefore the
infra-red singular structure of hIR .2D 0 =Q2 ; ı/ is given by a linear combina-
tion of the following terms : 1=ı 2 ; 1=ı; .1=ı/ ln.Q2 =2D 0 /; ln2 .Q2 =2D 0 /:
6.11.
X
.2/4 ı 4 . p C Q Pn /h P; j j .0/jnPn ihnPn j j .0/jP; i
nPn
XZ
D .dy/ ei. pn PQ/y h P; j j .0/jnPn ihnPnj j .0/jP; i
nPn
Z
D .dy/ eiQ y h P; j j .y=2/j .y=2/jP; i
where we have used the fact h P; j j .0/jnPn i
D ei.y=2/P ei.y=2/Pn h P; jeŒi.y=2/Mom:Op: j .0/ eŒi.y=2/Mom:Op: jPn i;
and a similar expression for the other factor, and we finally summed over
(n; Pn ).
6.12. Since P i D 0, Qi D Q3 ı i3 , we explicitly have W11 D W22 D W1 . Hence for
a transversal photon

T
W
D W1 0;

for D 1; 2. On the other hand,
0 W
0 , involves the following three
terms:
Œ.Q2 C 2 /=Q2 2 Œ.Q2 C 2 /W2 =Q2 W1 ;

Œ2 2.Q2 C 2 /=Q2 ŒW1 .Q2 C 2 /W2 =Q2 ;
. 2 =Q2 /2 Œ.Q2 C 2 /W2 =Q2 W1 :
Their sum gives

L
0 W
0 D ŒW2 .Q2 C 2 /=Q2 W1 0;
which establishes (6.9.13). Equations (6.9.14), (6.9.16) follow upon multiply-

ing (6.9.13) by 2 x M, with x D Q2 =2M, and finally using, in the process,
the definitions in (6.9.15).
6.13. We explicitly have (see also (6.9.7))
Z
1 e2i d3 p 0
Wi D Œ : .2/4 ı .4/ .P C Q p 0 /;
2M e2 2p 0 0 .2/3
where Œ : is obtained from (6.9.5) by making the substitutions: k ! P,

k 0 ! p 0 , and finally using the conservation law p 0 D P C Q.
6.14. The integral on the left-hand side is equal to

Z
.dp 0 / . p 0 0 / ı. p 0 2 C 2 M 2 / ı .4/ .P C Q p 0 /

D .P 0 C Q 0 /ı .P C Q/2 C 2 M 2 D ı.2PQ C Q2 /:
The result in question then follows upon taking 2QP outside the argument
of ı.2PQ C Q2 /.
6.15. The vertex function V for a spin 0 boson going from momentum p to
p 0 after interacting with the virtual photon must be of the form p C p 0 ,
with equal coefficients due to gauge invariance: Q . p C p 0 / D 0, where
Q D p 0 p. On the other hand, from the definition of Q, we may rewrite
. p C p 0 / D 2. p C Q =2/:
Also pQ D Q2 =2, i.e., Q =2 D p Q Q =Q2 . Thus the vertex function for
the spin 0 boson, consistent with gauge invariance, is simply proportional to
. p pQ Q =Q2 /. This in turn gives rise to a structure function contribution
proportional to
. p pQ Q =Q2 /. p p Q Q =Q2 /;
involving no . Q Q =Q2 / term. This leads to W1 D 0 and the results

stated in the problem follow.
6.16. We may write
X X X
e2i x fi .x/ D x f .x/ e2i D .2=3/ x f .x/ D .2=3/.1=3/ x fi .x/;
i i i
P 2
where we have used the fact that i ei D .4=9/ C .1=9/ C .1=9/ D 2=3.
Upon integration over x, this gives the relation stated in the problem.
6.17. For AnqG , we note that n 3 implies that n.n C 1/ 3 4 D 12, and hence
0 < 1=n.n C 1/ 1=12. Also, we may write
X
n X
n
1=j D 1=2 C 1=3 C 1=j:
jD2 jD4
The inequality for AnqG , then follows. The lower bound is easy to obtain,
just omit the positive part. The same reasoning leads to the inequalities
in (6.11.18) for AnGG , and AnGq , AnqNq in (6.11.17), where note, for example,

that AnGq in (6.11.12) may be rewritten as: .4=3/Œ 1=.n1/ C 2=n.n2 1/ .

6.18. Let t D ln.Q2 =2 /, then D .1=b0 / ln ln.Q2 =2 /=ln.Q2o =2 / , or
D .1=b0 / ln.t=to /. This gives .d=dt/ D .1=b0 t/ D ˛s .Q2 /=.2 /, where
we have used the relation ˛s .Q2 / D 1=.ˇ0 t/, to lowest order. From the chain
rule d=dt D .d=dt/.d=d/, the relation follows.
6.19. (i) This directly follows by noting that .An Dn Bn Cn / D C
n n , and that
C
An C Dn D n C n , upon carrying out the multiplication of the three matri-
ces on the left-hand side of (6.11.32). (ii) From (6.11.17), 0 < 8nf AnqNq AnGq
p
4 Bn Cn .98=135/nf . Thus C 2
n n D .An Dn / C 4Bn Cn
is real and positive. Also from (6.11.16), (6.11.18) we establish the positivity
of .An Dn / .AnqG AnGG / :
59 nf 10 49 nf 10
C C #4n < .An Dn / < C C #4n :
45 3 3 18 3 3
p
Hence using the fact that for any two positive numbers a; b : a2 C b2
.a C b/, we obtain from (6.11.34), (6.11.16)
1
q
C
An C Dn C .An Dn / C .98=135/nf ;
n
2
1
q 1
50 q
D 2 An C .98=135/nf < C .98=135/nf < 0;
2 2 9
with the upper bound, as shown, is strictly negative for unusually
p large
nf < 43. On the other hand .An Dn / > 4Bn Cn , and .An CDn / > 2.An CDn /
with the latter being negative. Hence
p

n D .1=2/ŒAn C Dn .An Dn /2 C 4Bn Cn
p p
> . 2=2/ Œ2Dn > 2 Œ .11=2/ C .nf =3/ C 6 #4n ;
as follows from (6.11.18). (iv) Finally

p
.C
n An / D .1=2/Œ.An Dn / C .An Dn /2 C 4Bn Cn > 0;
p
.
n An / D .1=2/Œ.An Dn / C .An Dn /2 C 4Bn Cn < 0:
R x0
6.20. For x 0 > y 0 , expŒCig y 0 d A 0 .; x/
C
Z x0 Z x0 Z 2
D 1 C .ig/ d1 A0 .1 ; x/ C .ig/2 d 2 d1 A0 .2 ; x/A0 .1 ; x/ C ;
y0 y0 x0 0
R x0
and @ 0 . : /C D igA0 .x 0 ; x/ Œ1 C .ig/ y 0 d1 A0 .1 ; x/ C .
6.21. For T > 0, consider the following expression, as a function of t 0 :
Q .t/ D h .t; a; 0; a/V 1 .t; a/

with condition Q .0/ D V 1 .0; a/. Using the fact that
dh .t; a; 0; a/=dt D igh .t; a; 0; a/A0 .t; a/;
we obtain dQ .t/=dt

D igh .t; aI 0; a/V 1 .t; a/V.t; a/ A0 .t; a/ .1=ig/.d=dt/ V 1 .t; a/;
which is just i g Q .t/AV0 .t; a/. Upon integration from 0 to T, gives:
h .T; aI 0; a/V 1 .T; a/ D V 1 .0; a/hV .T; aI 0; a/;
from which the transformation rule in question follows. The transformation

rules of the last two are almost identical to the first two by simply exchanging
time variables with space variables.
6.22. If a priori C is zero, then by a specific choice of the transformation
in (6.14.23), as a phase factor, we may remove any phase that 0 may have
upon the transformation in. Otherwise, suppose that C ¤ 0. Using the
identity
expŒi'n =2 D cos.'=2/ C i n sin.'=2/;
the transformation in (6.14.22) gives

cos.'=2/ C i n3 sin.'=2/ .i n1 C n2 / sin.'=2/ C
:
.i n1 n2 / sin.'=2/ cos.'=2/ i n3 sin.'=2/ 0
Upon considering the expression 0 = C , and writing the latter as

. 0 = C / D .a C i b/, with a and b real, it is easily checked, by equating
the resulting upper entry to zero, i.e., by setting

cos.'=2/ C i n3 sin.'=2/ C .i n1 C n2 / sin.'=2/.a C i b/ D 0;
that this equation has always a solution for all real a and b, by appropriate
choices of n1 ; n2 ; n3 , and '. This makes the resulting upper entry equal to
zero in vacuum expectation value. Any phase that may arise from the second
row of the transformation above may be removed by the appropriate choice
of the transformation in (6.14.23), giving finally a real non-negative field for
the lower entry in vacuum expectation value. Thus such transformations give
rise to a field as given on the right-hand side of (6.14.26), with its components
satisfying Eq. (6.14.27).
6.23. Using the anti-commutativity of the fermion fields, we have

ŒNeL L ŒŒN L eL D .1=4/Nea eD N c B Œ .1 5 /aB Œ .1 5 /cD :
Multiplying the Fierz identity8
1 1
. /ab . /cd D ıad ıcb . /ad . /cb . 5 /ad . 5 /cb C . 5 /ad . 5 /cb ;
2 2
by .1 5 /bB .1 5 /dD and using the identities f 5 ; g D 0, . 5 /2 D 1,

give Œ .1 5 /aD Œ .1 5 /cB , and the identity immediately follows.
6.24. Following the method in Sects. 5.9.3, 5.9.1, and averaging over the spin of
the muon and summing over the spins of the product particles, we have
1X ˇ
ˇ
.2M /.2me /.2me /.2m / jA j2 ˇ D 64 GF 2 . p k /.k1 k2 /;
2 spins me ;me ;m !0
which we conveniently denote by X. The decay rate is then given by

. p D .M ; 0//
Z
1 d3 k d3 k1 d3 k2
d D X .2/4 ı .4/ . pkk1 k2 /:
16M .2/3 jkj .2/3 jk1 j .2/3 jk2 j
Also note that

Z
3
p k d k1 d3 k2 =jk1 jjk2 j k1 k2 ı .4/ . pkk1k2 / D .M2 =6/Œ3M jkj4jkj2 :
Accordingly, using the fact that d3 k=jkj D Ee dEe d˝, and noting by
conservation of energy and momentum that the maximum value of the energy
Ee attained by the electron corresponds to the neutrinos moving in the same
direction leading to Ee jmax D M =2, we readily obtain, upon carrying the
k-integration, the decay rate stated in the problem.
8
For many details on Fierz identity and some of its generalizations, see Appendix A to Chapter 2
in Volume II.
Index
Z 2 in QED, 275 Bjorken limit, 432

Z 3 in QED, meaning of, 280, 327 Bjorken scaling, 407, 433, 434, 436
PCAC, 118 BRS transformation, 388, 396
SU.2/, 379
SU.3/, 380
SU.N/, 369, 378 Cabibbo-Kobayashi-Maskawa matrix, 489
U.1/, 223, 369 Callan-Symanzik equation, 350
sin2 W , experimental determination, 481 Casimir Effect, 231
C̆erenkov radiation and Apollo missions, charge conjugation matrix
241 properties of, 55
charge of the electron, 80
charge renormalization, 282, 289, 332, 338
AdS/CFT correspondence, 30 charge space, 371
anomalies, 74, 111, 402 chirality, 520
0 ! decay, 117 closed path, 375
experimental verification, 120 color, QCD, 405
non-abelian, 120 colors of quarks, 405, 406, 494
abelian, 111, 117 conditional probability, 137
absence of anomalies in the standard connection, 373
model, 494 connection, non-abelian gauge theories, 373
bad anomalies, 120 Coulomb gauge formulation of QED, 304
good anomalies, 120 Coulomb gauge, non-abelian gauge theories,
triangle anomaly, 117 388
anomalous magnetic moment of the electron, Coulomb potential in QED, 282, 332
290 Coulomb scattering, 94
anti-electron, 88 for relativistic electron, 95
anti-linear operator, 47 of non-relativistic electron, 95
anti-unitary operator, 47 covariant formulation of QED, 241
asymptotic freedom, 16, 353, 370, 403, 421 CPT, 31, 214
what is responsible for it, 423
decay rate of muon, 506

bare electron mass, meaning of, 278 decoupling theorem, 410
Bekenstein-Hawking Entropy formula, 23 deep inelastic scattering, 404, 429, 432
beta function in QED, 350, 352 degree of divergence, 127, 339
beta function, QCD, 422 delta functional, 69

DOI 10.1007/978-3-319-30939-2
584 Index
dimensional regularization, 275, 530 from the Coulomb gauge to covariant

dimensional transmutation, 423 gauges, 395
dimensional-regularized integral, 276 from the QDP to path integrals, 308, 388
Dirac equation of anti-commuting fields, 67
with external electromagnetic field, 80 of complex scalar fields, 67
Dirac propagator, 77 of real fields, 68
in external electromagnetic field, 80 functional integral, 65
Dirac quantum field, 79, 80, 85 Furry’s Theorem, 536
donkey electron, 88
double valued representation, 146
Duffin-Kemmer-Petiau matrices, 186 g-factor of electron, 83
gamma matrices
chiral representation, 54, 519
effective action, 357, 398 Dirac representation, 517
effective charge, 347, 348 gauge transformations
effective coupling, 370, 403, 422 non-abelian gauge fields, 383
effective fine-structure, meaning of, 283 of the full QED theory, 310
electron-positron kernel, 321, 336 Schwinger line-integrals, 459
electron-positron relative intrinsic parities, gauges
88 ;
t Hooft gauges, 494
Euler-Heisenberg effective Lagrangian, 73, Feynman gauge, 230, 276
109 Landau gauge, 230, 276
Euler-Lagrange field equations, 150, 153 Yennie-Fried gauge, 230, 276
how do they arise ?, 150 Gauss’ Theorem, 536
Gell-Mann and Low function, 351, 352
Gell-Mann matrices, 380
Faddeev-Popov determinant, 388 generating functional, non-abelian gauge
fermion propagator, 77 theories in covariant gauges, 393
Feynman diagram description of fundamental generations of quarks, 405
processes, 253 ghost fields, 394
Feynman gauge, 230, 276 ghosts and gauge invariance, 417
field equations, 151 gluon splitting, 440
fields, 135, 176 Goldstone bosons, 359
concept of, 135 Goldstone Theorem, 359
particle content, 135 Gordon decomposition, 82
Spin 0, 177, 184 Grassmann variables, 56, 74
Spin 1, 178, 188
Spin 1/2, 178
Spin 2, 183, 199 hadronization, 444
Spin 3/2, 180, 192 handedness, 519
and wavefunction renormalization, 137 Hawking radiation, 22
fine-structure constant, definition of, 95 helicity, 145, 520
Fock vacuum, 98 hierarchy problem, 17, 25, 26, 502
Foldy-Wouthuysen-Tani transformation, 457 Higgs boson, 361
form factors, 405 Higgs field, 361
Fourier Transform Higgs mechanism, 355, 361
Grassmann variables, 61 holographic principle, 31
fractional charges of quarks, 496
functional derivatives with respect to
anti-commuting functions, 63 incident flux, definition of, 254
functional determinants, 66 inequivalent representations of SU.3/., 380
functional Fourier Transforms, 45, 66, 67 inertial frames, 50
and Non-abelian gauge theories, 388 infrared divergence, 293
and path integrals, 310 infrared slavery, 370
Index 585
inhomogeneous Lorentz transformations parton splitting, 437, 447

group properties of, 51 partons, 408
integration over commuting complex variables, path integral expression for QED, 246
60 Pauli-Lubanski (pseudo-) vector, 141
integration over complex Grassmann variables, componens of, 141
60 photon propagator in covariant gauges, 230
integrations over links, 466 photon spectral representation, 281, 323
plaquette, 464
Poincaré algebra, 49, 141
Jacobian of transformation for Grassmann Poincaré Transformations, 51
variables, 58 polarization correlations, 267, 269
jets, QCD, 406, 408, 444 polarization vectors for photons, 226, 227
positron, re-discovery of, 88
principle of stationary action, 146
Källen-Lehmann representation, 323
QCD jets, 444

Lagrangian density Quantum Dynamical Principle, 171, 174, 386
non-abelian gauge fields, 394 Quantum Field Theory, 2
QCD, 409 quark confinement, 464
QED, 241 quark flavors, 405
Lamb shift, 294, 303 quarks, 404
Landau gauge, 230, 276
Landau ghost, 353
lattice, 463 radiative corrections, 273
lattice spacing, 463 Rarita-Schwinger field, 180, 192
Legendre transform, 357, 398 renormalizability, criterion of, 549
linear operator, 47 renormalization
link, 464 canonical variables, 543
Lorentz scalar, 53 renormalization group, 8
Lorentz transformations, 49, 50 renormalized electron field, 277
infinitesimal ones, 52 Rosenbluth formula, 404
LSZ formalism, 139 rotation matrix, 52, 227
running coupling, 422
Majorana spinor, 55
definition of, 55 Schwinger Dynamical Principle, 171, 174
metric Schwinger effect, 73, 111
Minkowski, 50 Schwinger line-integral, 102, 116, 458, 464,
minimal subtraction scheme .MS/, 420 534
Minkowski metric, 50 Schwinger parametric representation, 109
Minkowski spacetime, 51 Schwinger’s constructive approach, 166
modified minimal subtraction scheme (MS), Schwinger’s point splitting method, 73, 102,
420 131, 533
moments of splitting factors, 450 Schwinger-Feynman boundary condition, 77,
Mott differential cross section, 95 78
self-mass of the electron, 275
sites, lattice, 464
Noether’s Theorem, 158 skeleton expansion, 322
soft bremsstahlung, 293
solution of QED in Coulomb gauge, 307
particle concept in quantum field theory solution of QED in covariant formulation, 244,
how it emerges, 134 245
parton model, 14, 407 spin 1/2 character of quarks, 436
586 Index
Spin & Statistics Connection, 32, 55, 162, 166 vacuum polarization tensor, 103, 279
splitting factor, quark, 440 valence quarks, 408
splitting factors of partons, 443, 448 vertices, QCD, 410, 411
spontaneous symmetry breaking, 357, 359, virtual particles, 252
360, 371, 478 volume element in Minkowski spacetime,
structure functions, 407, 431 invariance of, 53
substitution law, 264
Super-Poincaré algebra, 49 Ward Identity, 318
survival probability of neutrino, 488 Ward-Takahashi Identity, 317
wavefunction renormalization, 134, 137
meaning of, 137
time anti-ordering, 459
and conditional probability, 137
time-ordering, 459
and quantum mechanics, 137
transition amplitudes in quantum field theory,
Wigner’s Theorem of symmetry
89
transformations, 47
tree diagram, 273
Wilson loop, 462
unitary gauge, 478 Yang-Mills field, 369, 377

unitary operator, 47 Yennie-Fried gauge, 230, 276

Quantum Field Theory I

Uploaded by

Copyright:

Available Formats

Quantum Field Theory I

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

Quantum Field Theory I

Uploaded by

Copyright:

Available Formats

Graduate Texts in Physics

More information about this series at http://www.springer.com/series/8431

Quantum Field Theory I

ISSN 1868-4513 ISSN 1868-4521 (electronic)

Library of Congress Control Number: 2016935720

© Springer International Publishing Switzerland 2016

Printed on acid-free paper

This Springer imprint is published by Springer Nature

• The very elegant functional differential approach of Schwinger, referred to

• The fine-structure effective coupling ˛ ' 1=128 at high energy corresponding

In Volume II, the reader is introduced to quantum gravity, supersymmetry, and

In the beginning of it all, I was introduced to the theoretical aspects of quantum

Nattapong Yongram, Siri Sirininlakul, Tukkamon Vijaktanawudhi (aka Kanchana

3.2 The Dirac Quantum Field Concept, Particle Content,

4.6 Quantum Dynamical Principle (QDP) of Field Theory .. . . . . . . . . . . . . 168

5.12 Radiative Correction to Coulomb Scattering and Soft

6.3 Functional Fourier Transform and Transition

Appendix I: The Dirac Formalism . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 515

Appendix II: Doing Integrals in Field Theory . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 521

Appendix III: Analytic Continuation in Spacetime Dimension

Appendix IV: Schwinger’s Point Splitting Method of Currents:

Appendix V: Renormalization and the Underlying Subtractions . . . . . . . . . 541

Solutions to the Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 551

1J D 6:242 109 GeV

(Masses) Mp D 938:3 MeV=c2 , Mn D 939:6 MeV=c2 ,

© Springer International Publishing Switzerland 2016 1

Field Theory” or just “Quantum Field Theory”. Quantum Electrodynamics is an

be of no surprise as one is assuming that our theories are valid up to infinite

A process was, in turn, then carried out, referred to as “renormalization”,

In units of „ D 1, c D 1, ŒMass D ŒLength1 . Roughly speaking, in a

necessarily a vanishingly small range of the interaction.25 Upon comparison of both

Evidently, the Fourier transform of the propagator in the energy-momentum descrip-

where k0 is its energy, and k D .k1 ; k2 ; k3 / its momentum. Formally for M 2W !

signalling, in a limiting sense, a short range interaction for a heavy-mass particle.

in the energy-momentum description.

providing no damping in such a limit. Moreover, as one goes to higher orders in

Glashow, a former graduate student of Schwinger, eventually realized [96]31 the

Fig. 1.5 A process involving

photons in quantum electrodynamics since photons do not carry a charge. These

Perturbatively renormalizable theories are distinguished from the non-renormaliz-

DS : Dyson Salam Weinberg Author

BP : Bogoliubov - Parasiuk Hepp - Zimmermann

The development of the general theory of renormalization from the DS and

referred to as the Bekenstein-Hawking Entropy formula76 of a BH. This relation

Supersymmetry is now over 40 years old. Supersymmetry provides a symmetry

small and methods, referred to as compactifications,99 have been developed to deal

the microscopic level, and their consequences on a macroscopic scale is certainly

10. Streater, R. F. (1985). Review of Renormalization by E. B. Manoukian. Bulletin of London

© Springer International Publishing Switzerland 2016 45

2.1 Wigner’s Symmetry Transformations in the Quantum

Invariance of physical laws under some given transformations leads to conservation

and fj 0 i ; j 0 ig provide an equivalent description as fj i ; jig. One may scale

To understand Wigner’s Theorem, one needs to define the following operators:

L Œ a j i C b ji  D a L j i C b L ji ; (2.1.2)

L Œ a j i C b ji  D a L j i C b L ji ; (2.1.3)

respectively, with * denoting complex conjugation.

2.1.1 Wigner’s Symmetry Transformations

Under a symmetry transformation, there exists a unitary or anti-unitary operator U

The unitarity condition U U D I D UU implies that • G is a self-adjoint

Moreover for a rotation of a coordinate system by an infinitesimal angle • about a

In units of „ D 1, c D 1, ŒMass D ŒLength1 . Roughly speaking, in a

and fj 0 i ; j 0 ig provide an equivalent description as fj i ; jig. One may scale

L Œ a j i C b ji D a L j i C b L ji ; (2.1.2)

L Œ a j i C b ji D a L j i C b L ji ; (2.1.3)

The unitarity condition U U D I D UU implies that • G is a self-adjoint

Moreover for a rotation of a coordinate system by an infinitesimal angle • about a

•b D .•; •x/; •! ij D " ij k nk • ; •! 0 i D •v i ; •! 00 D 0: (2.1.16)

where is independent of x, x 0 . Before we spell out the structure of the matrix

.x 0 y 0 / .x 0 y 0 / D .x y / .x y /; (2.2.4)

D ; .1 / D ; (2.2.5)

Thus a Poincaré transformation may be specified by a pair .; b/, with D

•! i j D " ij k n k •; (2.2.19)

satisfying the anti-commutation relations: f ; g D 2 . Various representa-

introduces via the transformation G G 1 the following representation of gamma

We note that 5 C ; 5 C , together with C , are anti-symmetric matrices.

Consider n real Grassmann variables 1 ; : : : n , i.e., variables that satisfy fi ; k g D

d.i j / D di j C i dj : (2.4.4)

d.i j / D di j dj i ; (2.4.5)

dj i D i dj : (2.4.6)

i.e., . / D . /, leads to the following definition

.I R / D I R or .I R / D R I ; (2.4.9)

.1 : : : n / D n : : : 1 : (2.4.10)

Imposing translational invariance of these integrals under i ! i C ˛ i D i0 ,

•.i ˇ i / D .i ˇ i /; (2.4.18)

i1 : : : in D " i1 ::: in 1 : : : n ; (2.4.21)

As before, out of two real Grassmann variables R ; I , we may define a complex

Using the fact that I R D =2i, we may infer that

For integrations over n complex Grassmann variables 1 ; : : : ; n , consider the

D .1 ˛1 /.1 ˛ 1 / : : : .n ˛n /.n ˛ n /; (2.5.1)

.1 ˛ 1 / : : : .n ˛n /.n ˛ n / D .˛ 1 /1 .˛1 /"1 : : : .˛ n /n .˛n /"n ; (2.5.2)

./ D ı 0 C ı1 ; ./" D ı"0 C ı"1 : (2.5.3)

dn .ın 0 n ın 1 /.n ˛n /.n ˛ n / dn .ı"n 0 C ı"n 1 /: (2.5.4)

i d1 d1 : : : i dn dn exp Œ i. ˛ C ˛ / ;

written in terms of variables .; /.

fa .x/; b .x 0 /g D 0; fa .x/; b .x 0 /g D 0; fa .x/; b .x 0 /g D 0; (2.6.5)

a .x/; a .x/ anti-commute with b .x 0 /; b .x 0 / as well as with the functional