Advanced

Particle Physics
Volume I

© 2011 by Taylor and Francis Group, LLC

 Particles, Fields, and
Quantum Electrodynamics

O. M. Boyarkin

Boca Raton London New York

CRC Press is an imprint of the

Taylor & Francis Group, an informa business
A TA Y L O R & F R A N C I S B O O K

© 2011 by Taylor and Francis Group, LLC

 Taylor & Francis
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2011 by Taylor and Francis Group, LLC
Taylor & Francis is an Informa business

No claim to original U.S. Government works

Printed in the United States of America on acid-free paper

10 9 8 7 6 5 4 3 2 1

International Standard Book Number: 978-1-4398-0414-8 (Hardback)

This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the valid-
ity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or uti-
lized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopy-
ing, microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com

© 2011 by Taylor and Francis Group, LLC

 To all
those whom
I sweetly
was deceived by.

© 2011 by Taylor and Francis Group, LLC

 Contents

Introduction xi

Designations xv

I Mathematical Prelude 1
1 Relativistic invariance 3

2 Three-dimensional world 9
2.1 Orthogonal transformation group O(3) . . . . . . . . . . . . . . . . . . . . 9
2.2 Tensor representations of the SO(3)-group . . . . . . . . . . . . . . . . . . 14

3 The four-dimensional Minkowski space 21

3.1 The homogeneous Lorenz group . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Classification of irreducible representations of the homogeneous Lorentz
group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Tensor representations of the limited homogeneous Lorentz group . . . . . 26
3.4 Spinor representations of the homogeneous Lorentz group . . . . . . . . . 26
3.5 Poincare group and its representations . . . . . . . . . . . . . . . . . . . . 29

4 Lagrangian formulation of field theory 37

4.1 Principle of least action. Lagrange-Euler equations . . . . . . . . . . . . . 37
4.2 Noether theorem and dynamic invariants . . . . . . . . . . . . . . . . . . . 39
4.3 Energy-momentum tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4 Tensor of angular momentum . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.5 Electromagnetic current vector and electric charge . . . . . . . . . . . . . 45
4.6 Isotopic spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5 Discrete symmetry operations 53

5.1 Spatial inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Time inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Charge conjugation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.4 G-parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.5 CP T theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

Problems 65

References 69

II Birds’s-Eye View on the Microworld 71

6 Fundamental interactions 73
6.1 Species of interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.2 On the path to a unified field theory . . . . . . . . . . . . . . . . . . . . . 87

vii
© 2011 by Taylor and Francis Group, LLC
 viii Contents

7 Atoms–Nuclei–Nucleons 101
7.1 Atomism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2 Rutherford model of the atom . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.3 Structure of the atomic nucleus . . . . . . . . . . . . . . . . . . . . . . . . 111

8 From muon to gluon 113

9 Hadron families 125

9.1 Yukawa hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2 Isotopic multiplets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
9.3 Unitary multiplets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

10 Quark “atoms” 167

10.1 Hypothesis of fundamental triplets . . . . . . . . . . . . . . . . . . . . . . 167
10.2 X-ray photography of nucleons . . . . . . . . . . . . . . . . . . . . . . . . 173
10.3 Parton model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
10.4 Quantum chromodynamics. The first acquaintance . . . . . . . . . . . . . 197
10.5 Heavy quarks and flavor symmetries . . . . . . . . . . . . . . . . . . . . . 207

11 Passing glance on the theory of electroweak interaction 219

11.1 Spontaneous symmetry breaking . . . . . . . . . . . . . . . . . . . . . . . 219
11.2 Glashow–Weinberg–Salam theory . . . . . . . . . . . . . . . . . . . . . . . 223

12 Fundamental particles of the Standard Model 237

13 Technical equipment of particle physics 245

13.1 Accelerators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
13.2 Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
13.3 Neutrino Telescopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

Problems 263

References 269

III Quantum Idyll—Free Fields 273

14 Scalar field 275

14.1 Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
14.2 Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
14.3 Production and destruction operators . . . . . . . . . . . . . . . . . . . . 280
14.4 Wave-corpuscle dualism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
14.5 Commutation relations. Properties of commutator functions . . . . . . . . 287
14.6 Microcausality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
14.7 Relativistically covariant scheme of the canonical quantization . . . . . . . 291
14.8 Green function of the scalar field . . . . . . . . . . . . . . . . . . . . . . . 293
14.9 Causal Green function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
14.10 Chronological product. Convolution of operators . . . . . . . . . . . . . . 297
14.11 First-order equation for scalar particles . . . . . . . . . . . . . . . . . . . . 300

© 2011 by Taylor and Francis Group, LLC

 Contents ix

15 Particles with spin 1/2 309

15.1 Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
15.2 Calculating γ-matrix tracks . . . . . . . . . . . . . . . . . . . . . . . . . . 313
15.3 Relativistic covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
15.4 Solutions of the free Dirac equation . . . . . . . . . . . . . . . . . . . . . . 321
15.5 Once more about particles spin . . . . . . . . . . . . . . . . . . . . . . . . 327
15.6 Polarization density matrix for Dirac particles . . . . . . . . . . . . . . . . 332
15.7 Dirac equation in the external electromagnetic field . . . . . . . . . . . . . 335
15.8 Charge conjugation, spatial inversion, and time inversion for the Dirac field 339
15.9 Relativistic quantization scheme . . . . . . . . . . . . . . . . . . . . . . . 342
15.10 Dirac field quantization (momentum space) . . . . . . . . . . . . . . . . . 344
15.11 Dirac field quantization (configuration space) . . . . . . . . . . . . . . . . 348
15.12 Majorana equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
15.13 Quantization of Majorana field . . . . . . . . . . . . . . . . . . . . . . . . 355
15.14 Second quantized representation of discrete operations C, P , and T for
particles with spin 1/2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
15.15 Neutrino in the Dirac and Majorana theory . . . . . . . . . . . . . . . . . 363

16 Massive vector field 371

16.1 Proca equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
16.2 Dynamic invariants of the vector field . . . . . . . . . . . . . . . . . . . . 374
16.3 Commutation relations in the Proca theory . . . . . . . . . . . . . . . . . 379
16.4 Duffin-Kemmer equation for the vector field . . . . . . . . . . . . . . . . . 380

17 Electromagnetic field 383

17.1 Maxwell equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
17.2 Dynamic invariants of the electromagnetic field . . . . . . . . . . . . . . . 387
17.3 Electromagnetic field quantization . . . . . . . . . . . . . . . . . . . . . . 391
17.4 Quantization with use of an indefinite metric . . . . . . . . . . . . . . . . 394
17.5 Photon polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399

Problems 403

References 407

IV Quantum Electrodynamics 409

18 S-matrix 411
18.1 Equations and dynamic variables in the Heisenberg representation . . . . 411
18.2 Interaction representation . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
18.3 Scattering matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
18.4 Representation of the S-matrix as the sum of normal products . . . . . . 423
18.5 Calculation of S-matrix elements . . . . . . . . . . . . . . . . . . . . . . . 425
18.6 Feynman rules in coordinate space . . . . . . . . . . . . . . . . . . . . . . 430
18.7 Furry’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
18.8 Feynman rules in the momentum representation . . . . . . . . . . . . . . . 438
18.9 Crossing symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
18.10 Cross section for unpolarized particles . . . . . . . . . . . . . . . . . . . . 447
18.11 Cross section for scattering by external fields . . . . . . . . . . . . . . . . 451
18.12 Decays of elementary particles . . . . . . . . . . . . . . . . . . . . . . . . . 455
18.13 Cross section for polarized particles . . . . . . . . . . . . . . . . . . . . . . 458

© 2011 by Taylor and Francis Group, LLC

 x Contents

18.14 Statistical hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462

19 Transmission of γ-radiation through matter 465

19.1 Photoeffect (nonrelativistic case) . . . . . . . . . . . . . . . . . . . . . . . 465
19.2 Photoeffect (relativistic case) . . . . . . . . . . . . . . . . . . . . . . . . . 470
19.3 Compton-effect (unpolarized case) . . . . . . . . . . . . . . . . . . . . . . 474
19.4 Compton-effect (polarized case) . . . . . . . . . . . . . . . . . . . . . . . . 482
19.5 Electron-positron pair production by the photon in the nucleus field . . . 487
19.6 Resume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495

20 Scattering of electrons and positrons 499

20.1 Bremsstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
20.2 Radiation of long-wave photons. Infrared catastrophe . . . . . . . . . . . . 504
20.3 e− e+ -pair annihilation into two γ-quanta. e− e+ -pair production under
annihilation of two γ-quanta . . . . . . . . . . . . . . . . . . . . . . . . . . 511
20.4 Positronium lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
20.5 Möller scattering e− + e− → e− + e− . . . . . . . . . . . . . . . . . . . 522
20.6 Processes e+ + e− → l + l (l = e− , µ− , τ − ) . . . . . . . . . . . . . . . 524

21 Radiative corrections 529

21.1 Calculations of integrals over virtual momenta . . . . . . . . . . . . . . . . 531
21.2 Regularization methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536
21.3 Self-energy of the electron . . . . . . . . . . . . . . . . . . . . . . . . . . . 543
21.4 Vacuum polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 549
21.5 Vertex function of the third order . . . . . . . . . . . . . . . . . . . . . . . 558
21.6 Radiative corrections to scattering of an electron by the Coulomb
nucleus field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562
21.7 Anomalous magnetic moment of leptons . . . . . . . . . . . . . . . . . . . 571
21.8 Lamb shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580

22 Renormalization theory 587

22.1 Primitive divergent diagrams . . . . . . . . . . . . . . . . . . . . . . . . . 588
22.2 Effective elements of Feynman diagrams . . . . . . . . . . . . . . . . . . . 593
22.3 Ward identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 601
22.4 Extraction of divergences . . . . . . . . . . . . . . . . . . . . . . . . . . . 604
22.5 QED renormalizability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609
22.6 Renormalization group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618
22.7 Consistency problems of the QED . . . . . . . . . . . . . . . . . . . . . . . 620

Problems 623

References 627

Appendix: Natural system of units 631

Index 633

© 2011 by Taylor and Francis Group, LLC

 Introduction

Among the natural sciences, only physics assumes the role of an all-embracing discipline
because its investigative subject is the universe as a whole. Physics enables a unified
approach to all objects of the universe from the elementary particles that constitute atoms
to giant astronomical structures. The greatest discovery of the twentieth century is the
realization that the physical world surrounding us was nonexistent at one time. There
is no other problem of science that could be so challenging as an effort to explain the
origin of the universe and to establish the reasons why and how the world arrangement has
been regulated. Possibilities to develop the first realistic models for the evolution of the
universe have been opened only with a breakthrough in the theory of elementary particles
because this theory provides the principal way to understand the laws of nature. The
greatest achievements in this section of physics are at first well-known and understandable
only for a narrow group of the experts and later become cultural elements alongside with
masterpieces of art. The latest particular discoveries in elementary particle physics enables
one to describe all natural phenomena within the scope of a unified descriptive scheme and,
hence, to establish a link between the macrocosm, where the galaxies and their aggregations
are scattered like scarce particles of dust and the microcosm of elementary particles. There
are two poles of the world: the giant universe, on the one hand, and the fundamental
particles, invisible despite the use of any available microscope, on the other hand. And it
has been found that a young universe possessed the properties of a microparticle, whereas
some microobjects, for example, microscopic black holes, are liable to harbor whole galactic
worlds.
Elementary particle physics is a relatively new sphere of human knowledge. Its birth
year is 1897, when the first elementary particle, an electron, was discovered. De Broglie’s
hypothesis and its confirmation during the electron diffraction experiments have formed
one of the principal statements of microscopic physics, wave–corpuscle dualism, that has
provided the foundation for nonrelativistic quantum mechanics (NQM). Being the first
theory in microworld physics, the NQM is sufficiently successful to explain the structure of
matter at an atomic and molecular level.
Merging a special relativity theory (SRT) and the NQM has culminated in the derivation
of quantum field theory (QFT) that is the foundation for elementary particle physics. At
first, particle physics seemed to represent a set of the formulations spun out of thin air and
of intuitive assumptions. Each form of the fundamental interactions was studied separately
and almost independently of all others. The only feature in common that unified all the
interactions was associated with the divergences present in series of the perturbation the-
ory. A theory of electromagnetic interactions between electrons and positrons, quantum
electrodynamics (QED), was a pleasant exception to this gloomy landscape. However, the
situation had changed drastically by the 1970s. Because of a series of experiments, the
chain of molecules–atoms–nuclei–nucleons was extended by the inclusion of a further layer
in the structure of matter: the quark-lepton level. And it was found that the quark and
lepton dynamics may be described using quantum theories with a local gauge symmetry.
The use of the gauge symmetry group SU (2)EW × U (1)EW together with a spontaneous
symmetry breaking hypothesis allows for the unification of electromagnetic and weak in-
teractions. Then, the inclusion of a theory of strong interaction based on the gauge color

xi
© 2011 by Taylor and Francis Group, LLC
 xii Introduction

group SU (3)c has resulted in the creation of a so-called standard model (SM). The SM
perfectly illuminates not only the microworld phenomena, but numerous cosmological phe-
nomena as well, for example, the Big Bang theory. However, by this time there are existing
experimental results, which will possibly demand going out beyond the SM. And it is not
improbable that, with the increase of accelerator energies, better accuracy of observations
in γ- and neutrino-astronomy, the number of phenomena at variance with the SM will be
increased resulting in the development of a new theory. Then, according to the correspon-
dence principle reading that every new and more exact theory incorporates, in the limiting
case, an old, less exact theory, all formulae of the SM could be derived from the appropriate
expressions of a new theory in a certain limiting procedure. In other words, the SM might
be a low-energy fragment of a new theory with the extended gauge group relative to the
SM. Further advance to the unified picture of the world consists in the unification of strong
and electroweak interactions, that is, in the framing of grand unification theory (GUT). We
have every reason to believe that the solution will be again found in the searches of the
gauge group including the gauge group of the SM as a subgroup. No doubt that for the
formation of a unified field theory, that involves both the GUT and gravitational interaction
theory, gauge symmetries will play a significant role.
This textbook is intended for anyone who wants to know how the world in which we live
really works. Having gained this knowledge, he or she might as well add his or her own new
page to this story. Particle physics is the main method to get to know laws of the nature.
The way to this science is thorny and complicated. Particle physics is similar to the arts
since it takes as much inspiration and hard work as literature, painting, and music.
Microworld physics is one of the “most challenging territories” of human knowledge. Us-
ing several textbooks under its study leads to the fact that the process of immersion into
the subject noticeably loses its efficacy because of the changing presentation style, the vari-
ation of complexity of the mathematical apparatus, the passage to differing designations,
and so on. And each textbook has its specific “language” and manner of presentation of the
material under study. My experience as a lecturer in elementary particle physics shows that
students or beginning post graduates are, as a rule, rarely willing to use several textbooks,
rather preferring a single favorite. This textbook contains all the particle physics founda-
tions and, as a result, the necessity of consulting any additional textbooks and monographs
is minimized.
To ease mastering the computation procedures in a quantum field theory, the majority
of the calculations have been performed without dropping the complex intermediate steps.
Therefore, the “as is easy to see”-type phrases, which may be associated with the omission of
the most subtle computational parts in some monographs, here denote the actual situation
and should not give anxiety to the reader.
A physicist-theorist has not only to perform complicated and skilled calculations and to
propose new theories, but he or she has to know experimental procedures and measuring
methods as well. Because of this, I have tried, as far as possible, to follow the links between
theory and experiment. Moreover, this textbook gives descriptions for modern devices
used in the physics of the microworld: accelerators, detectors of elementary particles, and
neutrino telescopes. Besides, I include such actual divisions of the modern particle physics
as the SM extensions and physics of the massive neutrinos that are usually contained in
monographs.
The textbook is constructed in the following manner. The first volume consists of four
parts. Part I covers the mathematical basis of modern quantum field theory. The most
necessary knowledge of the group theory are provided for, the Noether theorem is proved,
and the major motion integrals to be connected with both space and internal symmetry
are deduced. In Part II, fundamental interactions and the ways of their unifications are
discussed. The main theoretical preconditions and experiments that allowed one to estab-

© 2011 by Taylor and Francis Group, LLC

 Introduction xiii

lish matter structure at the quark-lepton level are considered. In Part III, the secondary
quantized theories of free fields with spin 0, 1/2, and 1 are investigated. Special attention is
attached to the neutrino field. Part IV is devoted to the first successfully operating quan-
tum field theory, quantum electrodynamics. In this part the methods of calculations with
polarized and unpolarized particles with and without the inclusion of radiative corrections
are yielded. Different renormalization schemes of the quantum field theory are discussed.
The second volume includes three parts. In Part V a quantum chromodynamics (QCD) is
expounded. The quantization scheme with the help of functional integrals is considered and
the renormalization problems are investigated. Cross sections of basic hard processes are
calculated. Nonperturbative methods (lattice approach, QCD vacuum) are also presented.
Part VI is devoted to electroweak interactions. The description of the Glashow–Weinberg–
Salam theory is given. Composite models and a left-right symmetric model are viewed as
the standard model extensions. In Part VII, massive neutrino physics is propounded.
Every part of the first volume contains problems that are an integral part of the text.
The goal of a majority of them is mastering the calculation technique to be described in the
textbook. There are also the problems that aim at generalizations of the results obtained.
This textbook could be used for teaching the following courses: (i) Introduction to Physics
of Elementary Particles; (ii) Foundations of the Quantum Field Theory; (iii) Quantum
Electrodynamics; (iv) Electroweak Interaction; (v) Quantum Chromodynamics; (vi) Physics
beyond the SM; and (vii) Physics of Massive Neutrinos.
This two volume-textbook is mainly meant for students and post-graduate students who
are specialized in particle physics. It will also be useful for post-doctors, researchers, and
academics. In order to understand the textbook contents, the reader must have knowledge of
nonrelativistic quantum mechanics, the special theory of relativity, and higher mathematics
fundamentals only. The textbook will help any attentive reader to be confident, when sailing
in the ocean of specialized literature devoted to the problems of strong- and electroweak-
interaction physics, and to be adequately prepared to begin research activities.
In conclusion, I would like to thank all physicists for personal contacts that contributed
greatly to my scientific horizons. Among them are Zygmunt Ajduk, James Daniel Bjorken,
Alexander Pankov, Stefan Pokorski, Dieter Rein, Alexander Studenikin, Dieter Schild-
knecht, Goran Senjanovic, and Victor Tikhomirov. I especially thank my wife and Dr.
B. Abai for the constant support during the long course of writing this book, and for their
inexhaustible belief that it would be completed.

© 2011 by Taylor and Francis Group, LLC

 Designations

Let us mark three-dimensional indices with Latin letters, four-dimensional indices running
values 0,1,2,3 by Greek ones. All components of four-dimensional vectors are real numbers.
We introduce two kinds of four-dimensional tensors. Thus, for four-dimensional coordinates
by definition we have
xµ = (x0 , x1 , x2 , x3 ) = (ct, x, y, z),

xµ = (x0 , x1 , x2 , x3 ) = (ct, −x, −y, −z).

Four-dimensional vectors with upper (low) index we name contravariant (covariant) vectors.
In the same way the difference is made for covariant and contravariant tensors with the rank
higher than one. Let us define the metric tensor
 
1 0 0 0
 0 −1 0 0 
gµν = .
0 0 −1 0
0 0 0 −1

Since the determinant in this matrix is not equal to 0, there is its inverse matrix gµν for
which takes place
gµν = g µν .
To raise and lower indices the metric tensor is employed. Thus, for example,

xµ = gµν xν , T µν = g µλ g νσ Tλσ , etc.

One calls twice-repeated indices dummy ones and on them summarizing is meant. Note,
that only space components change its sign under transition from covariant to contravariant
four-dimensional vectors. The scalar product of two four-dimensional vectors aµ and bµ is
defined as follows:
aµ bµ = a0 b0 − ab,
where
ab = ak bk = ak bk = a1 b1 + a2 b2 + a3 b3 .
Four-dimensional vector of energy-momentum has the following form:

pµ = (E/c, p)

and for it
p µ p µ = m 2 c2
is true.
Four-dimensional generalization for the Nabla operator is given by the expression:

∂
∂µ ≡ = (∂0 , ▽).
∂xµ

xv
© 2011 by Taylor and Francis Group, LLC
 xvi Designations

The symbol  is used for the D’Alembert operator:

1 ∂2
 = ∂µ ∂ µ = − △,
c2 ∂t2
where △ is the Laplace operator. Quantity εijk is a completely antisymmetric tensor:

 1, n − even
εijk = −1, n − odd

0, two and more indices coinside

where n is the number of transpositions that leads indices ijk to the sequence 123. Symbol
εµνλσ denotes the four-dimensional generalization of tensor εijk with ε0123 = 1 (while
ε0123 = −1).
Upper signs ∗, T , and † mean operations of complex conjugation, transposition, and
Hermitian conjugation respectively. A continuous line above spinors indicates the operation
of the Dirac conjugation:
u = u† γ4 ,
where γµ are Dirac matrices.
For basic vectors of representations and state vectors we use the Dirac bra- (|... >) and
ket- (< ...|) vectors. Thus, for example:

Ψ(p, s3 ) ≡ |p, s3 >, and Ψ† (p, s3 ) ≡< p, s3 |.

Let pus mark the three-dimensional radius vector with r, and its module with r, where
r = x21 + x22 + x23 .
In this book the Heaviside system of units are used, in which e2 /(4π~c) = αem . This very
normalization of electric charge is adopted in periodic literature on quantum field theory.
In the Gauss system of units, the normalization e2 /(~c) = αem is used (the value of electric
charge corresponding to it is usually given in tables of physical constants). In the Heaviside
system of units, the equations of electromagnetic fields have a more convenient form, since
the multiplier 4π does not enter there. The Coulomb law, however, in this system has the
following form:
q1 q2
Fc = .
4πr2
In contrast to this, field equations in the Gauss system of units contain the factor 4π, and
the Coulomb law has the simple form Fc = q1 q2 /r2 . It is obvious that the value αem is the
same in all systems of units, while the magnitude of the elementary charge e takes different
values.

© 2011 by Taylor and Francis Group, LLC

 Part I

Mathematical Prelude

1
© 2011 by Taylor and Francis Group, LLC
 1
Relativistic invariance

Any constitution will lose its ground

against good accoutrements.
Kozma Prutkov, “Thoughts about Democracy”

The classical physics description of all nature phenomena consists of two essentially dif-
ferent components—matter particles and fields. Movements of mass points, from which as
it seemed at that time, one could build all the variety of objects existing in nature, were
completely defined by laws of Newton mechanics. A classical particle is a small object lo-
calized in a restricted region of space. In the most general case, the position of the particle
is defined by three spatial coordinates, that is, its maximum number of degrees of freedom
equals three. Having set initial conditions and having solved the evolution equation (the
second Newton law), we obtain exhaustive information about the particle.
Description of the electromagnetic field, which represented the sole field known by that
time, was much more difficult, since it was necessary to indicate magnitudes and directions of
electric and magnetic field strengths at any point in space and at every instant of time. Thus
a field, unlike a mass point, had not three but an infinite number of degrees of freedom.
Interference, dispersion, and diffraction constituted other important field characteristics,
inaccessible by classical particles. Moreover, already at the time of Poisson (1811), the
field concept to describe electric and magnetic phenomena was created as an alternative to
the long-range interaction theory. So, since particles and fields are bearers of essentially
different individual characteristics, the universe picture in classical physics possesses clear
features of duality. However, the description of particles and fields has some things in
common. First, for particles and fields classical (Laplace’s) determinism operates. Second,
dynamic variables, used for their description, are the ordinary c-numbers (c1 c2 = c2 c1 ).
Nonrelativistic quantum mechanics changed the dual scheme of classical physics for a
partly more consecutive picture of wave-corpuscle dualism. According to the de Broglie
hypothesis, all the matter particles possess not only corpuscular, but also wave properties.
On the other hand, an electromagnetic field alongside with wave properties exhibits discrete
characteristics, which before were ascribed only to particles. The wave-corpuscle dualism
demanded reconsideration of both motion laws and every method of objects description.
In nonrelativistic quantum mechanics dynamical observables are no longer associated with
c-numbers, now they are associated with operators (q-numbers) satisfying definite commu-
tation relations. The Schrödinger equation is the evolution equation of a quantum system:

∂
i~ Ψ(r, t) = HΨ(r, t), (1.1)
∂t
where H is a Hamiltonian operator, Ψ(r, t) is a wavefunction of the system. In the quan-
tum theory a wavefunction is the main source of information about corpuscular and wave
properties of a microobject. Employing the wavefunction the average values of dynamical
observables are obtained and | Ψ(r, t) |2 dV defines the probability of a particle location
in the volume element dV . The description of states in classical and quantum regions is
essentially different. Coordinates and momenta of classical particles are the quantities that

3
© 2011 by Taylor and Francis Group, LLC
 4 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

are precisely and simultaneously measured by experiments. The wavefunction Ψ(r, t) de-
scribing the state of quantum particles cannot be directly measured by experiment, that
is to say, it has no physical meaning (only the square of its module has physical mean-
ing). From it follows that in the quantum theory the notion of determinism finds different
meaning. In this case the links of cause and effect reveal itself only at the probability
level, namely, the specifying of interaction form and wavefunction describing a system at
some moment of time, must uniquely determine wavefunction in subsequent moments of
time. The procedure of correlation between dynamical observables and operators, at which
the wavefunction remains to be the c-number is called formalism of primary quantization.
However, in quantum mechanics, an electromagnetic field had been keeping on to be apart.
It was described by classical Maxwell equations, that is, it was considered to be a classical
continuous field. Nonrelativistic quantum mechanics allows the description of movements of
electrons, nucleons, mesons, and other particles but not their production and annihilation,
that means, it is only applicable to systems with a constant number of particles. However,
the universal mutual convertibility of particles is one of the most characteristic features of
the microworld. Thus, the conservation law of the particles number is approximate, to be
exact, it is valid in the low energy region (of the order of atomic electrons energy) only. To
describe production and annihilation of particles, merging of quantum mechanics and the
special theory of relativity was needed.
The necessity for further development of quantum theory was as well dictated by a more
serious task—to quantize fields, corresponding to the known elementary particles, that is,
to build a quantum theory for systems with infinite numbers of degrees of freedom. The
solution of the above-mentioned problems led to the generalization of quantum mechanics,
which is called a quantum field theory. Quantized wave field is a fundamental physical con-
cept, logically reflecting the unified wave-corpuscle viewpoint at objects of nature. Each
kind of elementary particles is correlated with the corresponding wave field. Wave field
properties are determined by spin, charge, and other characteristics of these particles. To
quantize a wave field means to set up a correspondence between field and discrete energy
quanta (we call them “quanta of the given field”). Interactions between particles are car-
ried out by exchange of quanta of a certain field. For example, electromagnetic quanta
(photons) are particles, which describe interactions between electrically charged particles.
Thus, within the quantum field theory properties of elementary particles and their inter-
actions are formulated and described. A field is also characterized by a certain function
ψ(x) (x means plurality of four-dimensional coordinates), for which one uses the name a
field function, and sometimes, the name a wavefunction just as in nonrelativistic quantum
mechanics. In the quantum field theory field functions describe the population of particles,
with the theory containing mutual conversion processes of particles in the explicit form. In
correspondence with it, field functions acquire operator sense and fall into production and
annihilation operators of particles, between which commutation relations are established.
The procedure of imposing anticommutative relations (for fermions) and commutative ones
(for bosons) on field functions derived the name “formalism of secondary quantization”.
Field functions are no longer classical functions as it used to be in quantum mechanics.
They become operators acting upon a state vector | ΦN >, which characterizes completely
a physical system consisting of N particles. The vector | ΦN > is represented by a beam
in the complex Hilbert space , that is, in the complex infinite-dimensional space consisting
of a set of elements. In this space besides operations of addition and multiplication by a
number, the scalar product (Φ, Ψ) is also specified, which meets the following demands:

(Ψ, Ψ) ≥ 0; (Ψ, Ψ) = 0 only when | Φ >= 0, (1.2)

(Φ, Ψ + Θ) = (Φ, Ψ) + (Φ, Θ), (1.3)

© 2011 by Taylor and Francis Group, LLC

 Relativistic invariance 5

(Φ, αΨ) = α(Φ, Ψ), α ∈ C1, (1.4)

∗
(Φ, Ψ) = (Ψ, Φ) . (1.5)
If N is known with assurance, then the module square | ΦN > defining probability of the
given state evidently turns into 1. It means, that a state vector with any fixed N is nor-
malized to one. The form of an evolution equation for the system state vector depends on
a representation used in the formulation of a theory. The most convenient representation
is an interaction one, in which (as we demonstrate further) field functions satisfy free equa-
tions of motion and the evolution of a system state vector is defined by a Schrödinger-like
equation:
∂
i~ | ΦN (t) >= Hint | ΦN (t) >, (1.6)
∂t
where Hint is an interaction Hamiltonian. While looking at (1.6) some reasons for anxiety
comes into being. Einstein’s relativity principle reads, that all physical laws must have the
same form in all inertial frames of reference (IFR). Since the IFR are connected with each
other through the Lorentz transformation, then the relativity principle demands invari-
ance of a physical theory with respect to the Lorentz transformations, that is, relativistic
covariance. The equation (1.6) as it has been written is not relativistically covariant (co-
variance occurs only when all the four coordinates enter an equation at the equal foots).
However, it is possible to attach Eq. (1.6) the so-called “seeming” covariance [1,2]. The
idea is to change the nonrelativistic consideration of the whole space at some moment of
time by introducing the space-like arbitrary hypersurface σ. The hypersurface σ is de-
fined by only one condition, namely, it must contain no points that could be connected
by a light signal, that is, an interval between any two points x and y at σ is always a
space-like one ((x − y)2 < 0). Since the concept of space-like hypersurface is relativisti-
cally invariant, then setting of the system state vector as a functional on σ would be also
relativistically invariant. A state vector can be determined by independent observations
at all points of σ, because perturbations caused by observations cannot spread faster than
light speed. Due to the aforesaid, we are not limited by a certain system of coordinates
always working with regions where all the points are physically independent. Let us iden-
tify an element of volume dV with the projection of the space-like hypersurface element
dσ = (dσµ ) = (dx1 dx2 dx3 , cdx2 dx3 dt, cdx1 dx3 dt, cdx1 dx2 dt) on a hyperplane σt , for which
t=const. The hypersurface in four-dimensional space is an ordinary two-dimensional plane
analog in three-dimensional space, that is to say, it is some three-dimensional manifold in
the four-dimensional world. In the three-dimensional space the volume of a parallelepiped,
built on three vectors dr, dr′ , and dr′′ , is equal to a determinant of the third order composed
from components of these vectors. In four-dimensional space the projections of the volume
of a “parallelepiped” (that is, the area of the hyperplane), built on four-vectors dxµ , dxµ′ ,
and dxµ′′ , are expressed in the same way. They constitute an antisymmetric tensor of the
third rank, whose components are defined by the expression:
 µ 
 dx
 dx′µ dx′′µ 
dσ µνλ =  dxν dx′ν dx′′ν  .
 dxλ dx′λ dx′′λ 

As an element of integration over the hypersurface it is more convenient to use the four-
vector dσ µ , which is dual to the tensor dσ νλρ :
1
dσ µ = − εµνλρ dσνλρ .
6
Geometrically dσ µ is a four-vector that is equal in magnitude to the area of the hypersurface
element and normal to this element (that is, dσ µ is perpendicular to all lines drawn in the

© 2011 by Taylor and Francis Group, LLC

 6 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

hypersurface element). It is possible to ascribe local time t(r) to any point r of the space-like
hypersurface σ. A unit normal vector to the hypersurface σ at any point x is a time-like
vector, that is, nµ (x)nµ (x) = 1. The transition to the hyperplane σt takes place when
nµ = (1, 0, 0, 0).
Now let us consider Eq. (1.6) as a result of infinite series of equations, obtained by intro-
ducing local time for every point at the hypersurface σ. Exhibit the interaction Hamiltonian
as the sum over small three-dimensional cells ∆V , belonging to σ:
X
Hint (t) = Hint (x)∆V. (1.7)
σ

The equation that is fulfilled in the small cell the around space-time point with coordinates
(t(r), r) can be represented in the following form:

∂ | ΦN (t(r)) >
i~ = Hint (x)∆V | ΦN (t(r)) > . (1.8)
∂t(r)

The result obtained is quite obvious: since the change of | Φ(t) > at the infinitesimal
replacement of the hyperplane σt as a whole is given by the integral:
Z
Hint (x)dr,
σt

then the variation of | ΦN (t(r)) > at the point (t(r), r) must be defined by the density of
the interaction Hamiltonian Hint (x)∆V in the infinitesimal vicinity of the point x. In Eq.
(1.8) we shall switch over from the noninvariant differentiation with respect to time to the
covariant differentiation with respect to σ. For this purpose we consider the value of the
state vector | ΦN (t(r)) >=| ΦN (σ) > at two space-like hypersurfaces σ and σ ′ , which are
distinguished by an infinitesimal only in the vicinity of the space-time point x (see Fig.
1.1).

FIGURE 1.1
Two space-like hypersurfaces σ and σ ′ .

Then we have:
δ | ΦN (σ ′ ) > − | ΦN (σ) >
| ΦN (σ) >= lim =
δσ(x) ∆Ω(x)→0 Ω(x)

| ΦN (t(r)) + ∆t(r)) > − | ΦN (t(r)) >

= lim R =
∆t∆V →0 c∆t(r) ∆V dr
| ΦN (σ ′ ) > − | ΦN (σ) >
= lim , (1.9)
∆t∆V →0 c∆t(r)∆V

© 2011 by Taylor and Francis Group, LLC

 Relativistic invariance 7

where Ω is a space-time volume (an invariant quantity) compressed into the point x in the
limit. Using Eqs. (1.8) and (1.9) one can arrive at the relativistically invariant evolution
equation for the system state vector:
δ
i~c | ΦN (σ) >= Hint (x) | ΦN (σ) > . (1.10)
δσ(x)
The total Hamiltonian is connected with the total Lagrangian by the relation:
δL
H = ψ̇j − L, (1.11)
δ ψ̇j

where j = 1, 2, ..n, ψ̇j ≡ ∂t ψj , and n is the freedom degrees number of a quantum field
system. Since the Lagrangian is a relativistically invariant quantity, then for the interaction
Hamiltonian to be invariant, the derivatives with respect to four-coordinates must be absent
in the part of the Lagrangian, which is responsible for the interaction. For such interactions
we used the term “couplings without derivatives.” In case of couplings without derivatives
we simply have
Hint = −Lint . (1.12)
Thus, for example, the Lagrangian of interaction between particles with spin 1/2 and an
electromagnetic field contains no derivatives, while the electromagnetic interaction between
massive particles with spin 0 and 1 is an example of couplings with derivatives. However,
paradoxical though it may seem, in the latter case the theory finds the relativistic covariance
at the final stage. This transformation, to be similar to the metamorphosis of an ugly
duckling into a beautiful swan, is caused by the following circumstances. For the solutions
of Eq. (1.10) to exist, the integrability condition must be fulfilled:
δ δ δ δ
| ΦN (σ) >= | ΦN (σ) >, (1.13)
δσ(x) δσ(x′ ) δσ(x′ ) δσ(x)

where x and x′ are two points on the hypersurface σ. In its turn, for the condition to be
carried out, it is necessary, that:
 
′ ∂ ′ ∂
[Hint (x : n), Hint (x : n)] = i~c Hint (x : n) − Hint (x : n) , (1.14)
∂σ(x) ∂σ(x′ )

where ∂/(∂σ(x)) is the differentiation with respect to arguments, explicitly depending on σ

and the designation Hint (x : n) emphasizes the dependence of the interaction Hamiltonian
on the normal vector to the space-like hypersurface nµ (x). By means of Eq. (1.14), which
is the basis to define the interaction Hamiltonian in case of couplings with derivatives (see
for details [3]), it is possible to demonstrate, that

Hint (x : n) = −Lint (x) + W µν (x)nµ (x)nν (x) + ..... (1.15)

Besides Hint (x : n), Green functions of equations for vector and scalar particles also dis-
played dependence on nµ (x). In quantum field theory we call these functions “propagators”
and use them to describe the spreading of particles in virtual states. While calculating
the experimentally observed quantities (for example, cross sections), the so-called σ(x)-
dependence compensation procedure comes into play. It works as follows [4]. Terms, which
depend on the space-like hypersurface, that is, on nµ (x), appear in both an interaction
Hamiltonian and in propagators. Further, the exact mutual cancellation of these terms
in the S-matrix elements in all orders of the perturbation theory takes place. Thus, the
correct computational procedure in the theory of electromagnetic interactions of scalar and

© 2011 by Taylor and Francis Group, LLC

 8 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

vector bosons is to completely ignore the terms, which depend on the orientation of the
space-like hypersurface both in an interaction Hamiltonian and in particle propagators. In
the modern theory, which unifies strong, weak, and electromagnetic interactions, couplings
with derivatives for fundamental particles occur for particles with spin 1 and 0. In this
case such couplings are always contained for the vector particles and may be held for the
scalar particles only in some theory modifications. Then, we can claim that the evolution
equation for the state vector (1.10) is relativistically covariant in our picture of the world.
The fulfillment of Eq. (1.12) instills pleasant confidence in us, that the refusal from the
Hamiltonian formalism in favor of the Lagrangian one in the field theory is possible. Notice
that the Hamiltonian formalism is not relativistically invariant due to the singled-out role
of time. Let us underline that there is no proof that compensation of the σ(x)-dependence
for an arbitrary spin theory, which contains couplings with derivatives, takes place. Con-
sequently, if an experiment, being the last judge in the controversy, adds any new, say,
charged particles with spin 3/2 to the existing fundamental particles, we would be obliged
to check whether the compensation of the σ(x)-dependence takes place in the theory.
One may come to relativistic invariance of a theory from the other more constructive posi-
tion, namely, from the viewpoint of investigating transformation properties of field functions
under the Lorentz group transformations. To put this another way, it is necessary to de-
termine the Lorentz group representation for field functions. As this takes place, unitary
representations of the Lorentz group are of principal interest for us. The importance of ob-
taining such representations is due to the fact that they completely define a wave equation
for a physical system. To carry out this program we must supplement our mathematical
arsenal with the necessary information about the group theory. Let us make one remark
concerning our terminology. We shall resort to the theory of secondary quantization only in
Part III and in the first two Parts we shall work within the primary quantization formalism.
It allows us to view wavefunctions as classical objects in Parts I and II.

© 2011 by Taylor and Francis Group, LLC

 2
Three-dimensional world

2.1 Orthogonal transformation group O(3)

Let us start our excursus in the group theory from studies of transformations of the three-
dimensional Euclidean space. We consider transformations:

x′ = Rx, (2.1)

leaving a scalar product (x · y) invariant, for which the relation:

(x′ · y′ ) = (x · y) (2.2)

is true. As follows from (2.2), these transformations are carried out through orthogonal
matrices
RRT = RT R = I.
Since:
det(RRT ) = det RT det R = (det R)2 = 1,
then det R = ±1. In case det R = +1 we have proper orthogonal transformations or
rotations, while when det R = −1 we obtain improper orthogonal transformations . The
latter is exemplified by space inversion, which is represented by the matrix:
 
−1 0 0
R− =  0 −1 0  . (2.3)
0 0 −1
Any rotation can be represented as a product of three consecutive rotations around orthog-
onal axes x1 , x2 , x3 about angles θ1 , θ2 , θ3 . Under the rotations around axis x1 coordinates
x are transformed in the following way:

x′1 = x1 + 0 × x2 + 0 × x3 ,

x′2 = 0 × x1 + cos θ1 x2 + sin θ1 x3 ,

x′1 = 0 × x1 − sin θ1 x2 + cos θ1 x3 .
In the matrix form it looks as follows:
 ′     
x1 x1 1 0 0 x1
 x′2  = Rθ1  x2  =  0 cos θ1 sin θ1   x2  .
x′3 x3 0 − sin θ1 cos θ1 x3
In the same way we can represent transformations of rotations around axes x2 and x3 :
   
cos θ2 0 sin θ2 cos θ3 sin θ3 0
Rθ2 =  0 1 0 , Rθ3 =  − sin θ3 cos θ3 0  .
− sin θ2 0 cos θ2 0 0 1

9
© 2011 by Taylor and Francis Group, LLC
 10 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The matrix of a rotation around an arbitrary axis is defined by the product of matrices:
Rθ1 Rθ2 Rθ3 . Pay attention to the fact that the final result depends on the order of the
multiplication of Rθ matrices, that is to say, rotation transformations are not commuta-
tive. Totality of all the orthogonal transformations in the three-dimensional Euclidean
space makes up a group, namely, an orthogonal group O(3). Transformations with the
determinant equal to 1 are called either “unimodular” or “special” (the name of a group is
indicated by the letter S). Thus, the rotation group is the group SO(3). Groups consisting
of commutative elements are called Abelian groups, while ones with noncommutative ele-
ments are called non-Abelian groups . The group SO(3) is a non-Abelian group. The group
SO(3), supplemented with reflections, makes up the group O(3). It means, that SO(3) is a
subgroup of O(3). At this point it is high time to introduce the rules of the game, adopted
in the group theory.
A group is a set of elements G, which meets the following demands:
1. On the set G the group action is defined (call it “multiplication” conventionally),
which puts in correlation to every pair of elements f and g a certain h element from the
same set. The element h is a product of elements f and g, that is, h = f g. In the most
general case f g 6= gf .
2. The set contains such a unit element e, for which the relation ef = f e = f is true,
where f is any element from the set G.
3. Alongside with any element f , the set G contains an inverse element f −1 , which
possesses the following property:

f f −1 = f −1 f = e.

The group is finite (infinite) when the number of its elements is finite (infinite). The three-
dimensional rotation group is infinite while the reflection group is finite. The group is
called discrete (continuous) when its elements take discrete (continuous) values. The three-
dimensional rotation group is continuous, while the reflection group is discrete. The order of
group is a number of independent parameters, which defined a group. The three-dimensional
rotation group is the group of the third order (its independent parameters are θ1 , θ2 , θ3 ).
The order of the reflection group is the same. The order of the transformation matrix (two-
column, three-column, etc.) defines dimensionality of a group (d). The dimensionality of the
three-dimensional rotation group and of the reflection group is three. The dimensionality
of the O(n)-group is defined by the formula d = n(n − 1)/2. Let us note, that the order
of a group differs from its dimensionality. In the case of the three-dimensional orthogonal
group this coincidence is accidental. Continuous groups of the finite order are Lie groups.
The three-dimensional rotation group is an example of a Lie group.
Now our purpose is to define all representations of an orthogonal group. In general,
representation of a certain G group is a mapping, which compares every element g of G
with linear operator Ug , acting in some vector space V . In such a mapping a multiplication
table for a group is maintained, and the unit element e of the group G is represented by
identity transformation of I in V :

Ug1 Ug2 = Ug3 (g3 = g1 g2 ), Ue = I.

The most simple representations, of which one can build all the rest of group represen-
tations by means of multiplication, are named fundamental. If the group representation is
formed by the group generators themselves then that is regular or adjoint.
Subspace V1 of the space V is called the invariant subspace with respect to representation
Ug in case if all the vectors v in V1 are transformed by any Ug into vectors v ′ belonging
again to V1 . A representation is irreducible, if the only subspace V , which are invariant

© 2011 by Taylor and Francis Group, LLC

 Three-dimensional world 11

with respect to representation g → Ug , are the whole space and subspace, consisting of one
zero vector. Otherwise a representation is reducible.
Studies of representations could be limited to studies of irreducible representations. More-
over, only nonequivalent irreducible representations are of a main interest for us. Two
representations Ug and Ug′ are equivalent, if there is a unitary operator M , which ensures

Ug′ = M Ug M −1 v ′ = M v,

where v and v ′ are vectors of representation spaces. Thus, two equivalent representations
could be seen as realizations of one and the same representation in terms of two different
bases in a vector space.
All irreducible representations are finite-dimensional and any representation is a direct
sum of irreducible finite-dimensional representations. The direct sum of two square matrices
D1 and D2 is by definition a square matrix D3 for which
 
D1 0
D3 = .
0 D2
is true. Symbolically, it is set down as follows:
M
D3 = D1 D2 .

Further on, infinitesimal rotations and matrices, corresponding to them will play a fun-
damental role. Their importance is caused by the fact, that they produce one-parameter
subgroups and any finite rotation can be composed of a sequence of infinitesimal rotations.
To obtain matrices for infinitesimal rotations, we expand every element of the matrix of
finite rotations into the Taylor series and constrain ourselves by the terms of the first order:

d 

Ai = R(θi ) , (i = 1, 2, 3) (2.4)
dθi 
θi =0

The obtained matrices (2.4) will be called generators of rotations around i-axis. Their
obvious form is given by
   
0 0 0 0 0 −1
A1 =  0 0 1, A2 = 0 0 0 ,
0 −1 0 1 0 0
 
0 1 0
A3 =  −1 0 0. (2.5)
0 0 0
From (2.5) it is clear, that infinitesimal rotations commute with each other. A rotation
about a finite angle θi can be viewed as a result of n rotations about angle θi /n:
 n
θi
R(θi ) = lim 1 + Ai = exp (Ai θi ). (2.6)
n→∞ n
By direct calculations it is possible to check that generators Ai satisfy the commutation
relation:
[Ai , Ak ] = −εikl Al (2.7)
and commute with the reflection operator:

[Ai , R− ] = 0. (2.8)

© 2011 by Taylor and Francis Group, LLC

 12 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

A set of A generators constitute the so-called Lie algebra of the three-dimensional rotation
group.
The set of elements N is a Lie algebra, corresponding to a certain group, if it meets the
following demands:
1. If X and Y are elements of the set N , then both the sum of X + Y and the product
αX (where α is an arbitrary number) belong to N once again.
2. Commutator of two elements X and Y of the set N is again expressed in terms of
elements of the set N .
3. Commutators of elements of the set N satisfy the relation:

[X, Y ] + [Y, X] = 0, [X, (Y + Z)] = [X, Y ] + [X, Z],

[X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y ]] = 0, (Jacobi identity). (2.9)
The representation of the Lie algebra is a correspondence X → T (X), which confronts
every element X with the linear operator T (X) acting in a vector space in such a way, that:

T (X + Y ) = T (X) + T (Y ), T (cX) = cT (X),

T ([X, Y ]) = [T (X), T (Y )].

By the example of the rotation group, we show that the Lie algebra representation
uniquely determines the representation of the group itself. Infinitesimal rotation about
the angle δθ is put down as follows:

R(δθ) = I + Ai δθi + O(δθ2 ).

The corresponding representation operator has the form:

U (δθ) = I + Mi δθi + O(δθ2 ), (2.10)

where Mi constitute the representation of the Lie algebra generators and obey the same
commutation relations as Ai :
[Mi , Mk ] = −εikl Ml . (2.11)
Let us consider two rotations about the finite angles aθ and bθ. Because rotations about
one and the same axis commute:

R(aθ)R(bθ) = R((a + b)θ)

then as a result:
U (aθ)U (bθ) = U ((a + b)θ). (2.12)
Differentiating (2.12) with respect to a, using (2.10) and taking into consideration that
d d
↔ ,
da db
at a = 0 we obtain the equation
d d
U (aθ)U (bθ)|a=0 = U (bθ) = (Mi θi )U (bθ). (2.13)
da db
To integrate Eq. (2.13) and use the initial condition U (0) = 1 produces the expression for
operators of the representation of the three-dimensional rotation group:

U (θ) = exp (Mi θi ). (2.14)

© 2011 by Taylor and Francis Group, LLC

 Three-dimensional world 13

Since in unitary representations operators U are unitary, then operators Mi are anti-
Hermitian. Now we proceed to Hermitian operators Ji = −iMi ,, which satisfy ordinary
permutation relations for angular moment operators:

[Jk , Jm ] = iεkmn Jn . (2.15)

The problem of finding all the irreducible rotation group representations is equivalent to
that of finding all the possible sets of matrices J1 , J2 , J3 , complying with the permutation
relations (2.15). The Shur lemma has the fundamental importance in representations theory
of groups, realized by complex matrices. It reads: “for representation to be irreducible, it is
necessary and sufficient, that the only matrices, which commute with all the representation
matrices are the matrices which are multiple of the identity matrix.” Such matrices being
polynomials of generators, are called Casimir operators UiK . According to the Shur lemma,
a representation is irreducible in the case of

UiK = λI

and as a result
UiK ψ = λψ,
where ψ denotes the field function. Since the Casimir operators enter the total set∗ of
commuting operators, then conserved physical quantities correspond to them. The number
of the Casimir operators defines a group rank.
So, if we have found all the Casimir operators, have chosen one eigenvalue from each
operator, and built up a representation that acts in space stretched on corresponding eigen-
functions, such a representation is irreducible, according to the Shur lemma. In other words,
to classify irreducible representations of a group one must find the spectrum of eigenvalues
and eigenfunctions of the Casimir operators.
Because the three-dimensional rotations group has one Casimir operator:

J2 = J12 + J22 + J32 , (2.16)

then the rank of this group is equal to 1. In this case the Casimir operator is nothing else but
the squared angular moment operator with eigenvalues j(j+1), where j = 0, 1/2, 1, 3/2, 2, ....
Consequently, every irreducible representation of the three-dimensional rotations group is
characterized by the positive integer or half-integer number j, which also defines a repre-
sentation dimensionality by the formula 2j + 1. From quantum mechanics we know that
the squared angular moment operator commutes with a projection of the angular moment
operator onto a certain direction singled out in space, for example, the x3 -axis. Thus, the
basic representation functions are the eigenfunction of the operators J2 and J3 and could
be marked by their eigenvalues.
Passing on to the classification of irreducible representations of the orthogonal group we
keep in mind that the linear operator U− corresponding to the operation of space inversion
R− commutes with all the operators of the rotations group representation. According to the
Shur lemma, in every irreducible representation U− must be a multiple of the unit operator.
Thus, the irreducible representation of the orthogonal group is classified by a pair of indices
(j, ηp ), where the latter is the eigenvalue of U− corresponding to the given representation.
Representations with integer j are called one-valued or tensor representations, in the case
of half-integer j representations are called two-valued or spinor representations.

∗ It undoubtedly contains the Hamiltonian as well.

© 2011 by Taylor and Francis Group, LLC

 14 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

2.2 Tensor representations of the SO(3)-group

Since at integer values j the relation

(U− )2 ψ = Iψ

takes place, then eigenvalues of the reflection operator are +1 and −1. Thus, there are two
different representations of the orthogonal group, in the former U− = I and in the latter
U− = −I.
At j = 0 the representation is one-dimensional, each group element is expressed in terms
of the unit operator, and generators are identically equal to zero. Let us call representation
{0,1} as a scalar and {0,-1} as a pseudoscalar one. Quantities transformed according to
(pseudo)scalar representation are (pseudo)tensors of zero-rank or (pseudo)scalars.
At j = 1 the representation is three-dimensional. One might use matrices of generators
of the three-dimensional rotations group as the matrix representation of generators Mi . For
the representations {1,-1} and {1,1} we use the terms vector representation and pseudovec-
tor representation, respectively. Three-dimensional quantities transformed with respect to
(pseudo)vector representations are called (pseudo)tensors of first rank or (pseudo)vectors.
Very often pseudovectors are called axial vectors and vectors are called polar vectors. All
the representations with j = 0, 1 are irreducible, while the representations with j ≥ 2
are reducible. To summarize the aforesaid, we can give the following definition: a three-
dimensional tensor (pseudotensor) of the n-rank is a quantity, which is transformed under
the representation {n, (−1)n } ({n, (−1)n+1 }) of the group O(3). More convenient for prac-
tical use language, the tensor and pseudotensor of the n-rank are the quantities whose
components, when rotating, are transformed according to the law:
′
ψik..m (r′ ) = Rip Rkl ...Rms ψ pl..s (r). (2.17)
|{z} |{z}
n n

However, under reflections we have:

′
ψik..m (r′ ) = (−1)n ψik..m (r), (2.18)
|{z} |{z}
n n

for a tensor
′
ψik..m (r′ ) = (−1)n+1 ψik..m (r), (2.19)
|{z} |{z}
n n

and for a pseudotensor, respectively.

The ability to expand reducible representations in irreducible ones will make one more
element of our mathematical culture. As an example, we consider a field function ψik , which
is a tensor of the second rank with respect to the three-dimensional rotations group and
consequently it has nine components. A tensor of the second rank is a product of two three-
dimensional vectors, which are transformed according to three-dimensional representations.
Schematically it can be written in the form:
O
3 3. (2.20)

Let us represent ψik as a sum of symmetric and antisymmetric tensors:

s 1 a 1
ψik = (ψik + ψki ), ψik = (ψik − ψki ).
2 2

© 2011 by Taylor and Francis Group, LLC

 Three-dimensional world 15

By the direct check one can be sure, that

s′ (1) s (1) a′ (1) (1)
a
ψik = Rim Rkn ψmn ψik = Rim Rkn ψmn ,

that is, under the rotations the components of the symmetric and antisymmetric tensors are
not mixed with each other. In other words, under transformations of the group SO(3) the
nine components of the tensor ψik fall into two independent totalities: three-dimensional
a s
one ψik and six-dimensional one ψik . Further we remember, that the scalar product of three-
dimensional vectors is invariant under the rotations. Then the symmetric tensor could be
also expanded into two independent totalities, namely, a scalar
1 1 1
ψ0 = Sp{ψmn } = ψnn = (ψ11 + ψ22 + ψ33 )
3 3 3
and five components, which constitute a matrix with a zero-spur:

s s 1
ψik = ψik − δik ψnn .
3
So, from the viewpoint of the rotation transformations, the second rank tensor ψik is the
sum of the three independent quantities: the one-dimensional ψ 0 , the three-dimensional ψka
(k = 1, 2, 3), and the five-dimensional ψls (l = 1, 2, 3, 4, 5). The corresponding representation
of the rotations group consists of matrices of a less dimensionality and has the box-diagonal
form:
 a   a
ψ1 U11 U12 U13 0 0 0 0 0 0 ψ1
a
 ψ2   U21 U22 U23 0 0 0 0 0 0   ψ2a 
 a 
0   ψ3a 
 
 ψ3   U31 U32 U33 0 0 0 0 0
 s 
U44 U45 U46 U47 U48 0   ψ1s 
 
 ψ1   0 0 0
 s   
U  ψ2  =  0 0 0 U54 U55 U56 U57 U58 0   ψ2s  . (2.21)
 s 
U64 U65 U66 U67 U68 0   ψ3s 
 
 ψ3   0 0 0
 s   
 ψ4   0 0 0 U74 U75 U76 U77 U78 0   ψ4s 
 s   s 
ψ5 0 0 0 U84 U85 U86 U87 U88 0 ψ5
0
ψ 0 0 0 0 0 0 0 0 1 ψ0
If a representation matrix can be written down in the box-diagonal form, the correspond-
ing representation is reducible (otherwise it is irreducible). Then, Eq. (2.21) means
O M M
3 3=9=5 3 1. (2.22)

Let us try to present the sequence of manipulations from Eqs. (2.20) to (2.22) in a more
obvious form. In Fig. 2.1 we plot basis vectors of the representations entering the left-hand
side of Eq. (2.22), where values of the angular momentum projection mj are denoted with
dots. To obtain the totality of the basis vectors, composing representations in the right-

õ
-1 0 1 -1 0 1

FIGURE 2.1
The basis vectors of the representations 3 and 3.

hand side of Eq. (2.22), one might use the moments multiplication rule, which is well known

© 2011 by Taylor and Francis Group, LLC

 16 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

from quantum mechanics. In order to get the products of two moments, we superimpose
the center mj = 0 of the second diagram with each of the three states mj = 1, 0, −1 of the
first diagram. Further, we should calculate frequency ni with which every summation result
mj appears (in our example mj = +1 occurs twice, mj = 0 does triply). From the obtained
states we single out diagrams with the same value of j. This procedure is displayed in Fig.
2.2. Then, if one takes into account that the states number with the angular momentum

õ =
-1 0 1 -1 0 1 -2 -1 0 1 2

FIGURE 2.2
Graphical interpretation of the moments summation rule.

eigenvalue j is equal to 2j + 1, that is,

O O
(2j1 + 1) (2j2 + 1) ≡ j1 j2 ,

the generalization of the relation obtained takes the familiar form:

O X
j1 j2 = j, (2.23)
L

L j-representation occurs once, j =| j1 − j2 |, | j1 − j2 + 1 |, .... | j1 + j2 | and

where every
subscript indicates that the sum is direct, not normal.

§ 1.2.3. Spinor representations of the three-dimensional rotations group

At j = 1/2 representations of the rotations group is two-dimensional and generators

(1/2)
Jk might be chosen in the form of iσk /2, where σk are Pauli matrices
     
0 1 0 −i 1 0
σ1 = , σ2 = , σ3 = , (2.24)
1 0 i 0 0 −1

that satisfy the relation:

σk σl = δkl + iεklm σm . (2.25)
The operator to transform a wavefunction under the rotations about θ angle around the
axis with the unit vector n = (n1 , n2 , n3 ), is written down as follows:
∞  k  2  2
θ X 1 i 1 θ 1 θ
U (1/2) (n, θ) = exp [i(n · σ) ] = (n · σ)θ = [1 − + −
2 k! 2 2! 2 4! 2
k=0
 3  5
θ 1 θ 1 θ θ θ
....] + i(n · σ)[ − + − ...] = I cos + i(n · σ) sin . (2.26)
2 3! 2 5! 2 2 2
From (2.26) it follows that the rotation matrices are unitary and their determinant equals
1. Objects transformed on the representation j = 1/2 shall be called spinors of the first

© 2011 by Taylor and Francis Group, LLC

 Three-dimensional world 17

rank. As a basis in space of such spinors, one might choose eigenvectors of the matrices
(1/4)σ2 (spin square) and (1/2)σ3 (spin projection onto the x3 -axis) which have the form:
   
1 0
ω+ = , ω− = .
0 1
Since:
1 2 3
σ ω± = ω±
4 4
and
1 1 1 1
σ3 ω+ = + ω+ , σ3 ω− = − ω− ,
2 2 2 2
it becomes obvious, that spinors of the first rank are used to describe particles with the
spin 1/2.
The system rotation around the angle 2π causes quite unexpected results:
U (1/2) (n, θ + 2π)ψ = −U (1/2) (n, θ)ψ, (2.27)
that is, the system does not return to the starting state. It means that observable quantities
cannot be represented by a spinor, since spinors have no definite transformation properties
under the rotations. From a physical point of view, it makes the main difference between
the spinor and tensor representations. It should be noted that in the tensor representations
observables can be connected with field functions directly (electromagnetic field strengths
may serve as an example).
A Hermitian conjugation of a spinor is carried out in the ordinary way, that is, it consists
of the transposition and the complex conjugation:
ξ † = (ξ1∗ , ξ2∗ ) .
An observable can be represented as a bilinear combination of spinor quantities. Needless
to say, it will possess single-valued transformation properties. Thus, for example, under the
rotations the quantity ξ † χ behaves as the scalar:
ξ ′† χ′ = ξ † U (1/2)† (n, θ)U (1/2) (n, θ)χ = ξ † χ.
It can be shown in the same way, that quantity ξ † σχ is a vector and so on.
From (2.27) it follows that the representation with the weight j = 1/2 is two-valued,
namely,
R(n, θ) → ±U (1/2) (n, θ). (2.28)
It brings us to the idea, that from the point of view of the SO(3)-group, the existence of
two kinds of spinors is possible.
Let us slightly alter the above-mentioned facts in form. In two-dimensional complex space
S (2) (spinor space) we introduce basis e1 and e2 by which any spinor ξ can be expanded:
ξ = ξ 1 e1 + ξ 2 e2 . (2.29)
At the rotation about the angle θ around an axis with the unit vector n spinors belonging
to S (2) are transformed according to the formula:
ξ ′α = U (1/2)α β ξ β , (2.30)
where α and β are spinor indices taking the values 1 and 2. The operation of raising and
lowering the indices is carried out by means of an antisymmetric “metric tensor”:
 
αβ 0 1
ǫαβ = ǫ = (2.31)
−1 0

© 2011 by Taylor and Francis Group, LLC

 18 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

that is, ǫ = iσ2 .

Now let us examine another example of a spinor space S (2̇) whose spinors we shall consider
to be different from ones of S (2) . For any basis (e1 , e2 ) of the spinor space S (2) one may
uniquely determine the dual basis connected with the initial one by the relation:

α̇ α̇ 0, α̇ 6= β
e e β = δβ = (2.32)
1, α̇ = β,

In (2.32) we have supplied the spinor indices of the space S (2̇) with dots to emphasize the
fact that the transformation law of the spinors on these indices differs from that of the
spinors of the space S (2) . An arbitrary spinor η belonging to S (2̇) is expanding in dual basis
in the following way:
η = η1̇ e1̇ + η2̇ e2̇ . (2.33)
At the three-dimensional rotations the transformation law of the spinors belonging to the
space S (2̇) is given by the expression:
(1/2)α̇ β̇
η α̇′ = Ũβ̇ η . (2.34)

A matrix of the transformation Ũ (1/2) meets the choice of the rotation generators Mk in the
form −iσk∗ /2 to ensure the commutative relations fulfillment (2.11) as well. Passing to the
spinors with the low indices by means of the metric tensor ǫα̇β̇ = ǫα̇β̇ = ǫαβ in Eq. (2.34)
and with allowance made for the relation:

σ2 σ ∗ σ2 = −σ. (2.35)

we arrive at:
(1/2)α̇†
ησ̇′ = ǫσ̇α̇ Ũ (1/2)α̇ β̇ ǫβ̇ τ̇ ητ̇ = Uσ̇ ηα̇ .
A two-component quantity  
ξ1
ξα = ,
ξ2
which is transformed by the law (2.30) at the three-dimensional rotations we call a con-
travariant spinor of the first rank. Spinors whose transformation law has the form:

ηα′ = Uα(1/2)β† ηβ (2.36)

we call covariant spinors. In Eq. (2.36) we consciously omitted dots above the indices, but
further on we shall always imply that the transformation law on low (covariant) indices is
determined by the formula (2.36). Now it is possible to obtain a bilinear form from covariant
and contravariant spinors:

(ξ, η) = ξ α ǫαβ η β = ξ 1 η1 + ξ 2 η2 , (2.37)

to be invariant under the three-dimensional rotations. We can interpret it as the scalar

product.
Now we proceed to a discussion about spinor properties at the space inversion. Let us
introduce an arbitrary plane P with the unit normal vector n. Decompose the vector x into
two components to be parallel and perpendicular to the vector n:

x = xk + x⊥ = (x · n)n − [n × [n × x]]. (2.38)

© 2011 by Taylor and Francis Group, LLC

 Three-dimensional world 19

We introduce the matrix-vector R whose components are three inversion matrices of the
coordinate axes:
     
−1 0 0 1 0 0 1 0 0
R−1 =  0 1 0  , R−2 =  0 −1 0  , R−3 =  0 1 0  .
0 0 1 0 0 1 0 0 −1

Transition to a vector x′

x → x′ = −(x · n)n − [n × [n × x]] = x − 2n(x · n), (2.39)

which is in fact a specular reflection of the vector x in the plane P , is carried out by the
k k
inversion matrix: R− = (R− · n)n. By direct calculations it is possible to check that R−
satisfies the following relations:
k k k
{R−i , A⊥
k } = 0, [R−i , Ak ] = 0, (2.40)

where i, k = 1, 2, 3. In the expression (2.40) braces denotes the anticommutator, and the
rotation generators are also expanded into Ak and A⊥ :

A = Ak + A⊥ = (A · n)n − [n × [n × A]]. (2.41)

k
In any representation permutation relations of an operator, corresponding to R− , and
generators of this representation must have the same form as the relation (2.40), that is,
k k k
{T−i , Mk⊥ } = 0, [T−i , Mk ] = 0, (2.42)

Consistency of the conditions (2.42) and allowing for (Tk )2 = 1 result in the conclusion,
that under the reflection (2.39) the spinor ξ is transformed by the law:

ξ α → ξ ′α = (n · σ)α β α β
β ξ = Nβ ξ , (2.43)

where from two possible signs in front of the reflection matrix N the sign “plus” was chosen
arbitrarily (two reflections might be considered as the rotation about the angle 2π). As the
result of matching the transformation laws for the spinors ξ and η at the space rotations
(2.30) and (2.36), we obtain the relation:

ηβ ǫβα ξ α∗ (2.44)

where ξ ∗ is a column with components ξ 1∗ and ξ 2∗ , and a sign denotes “transformed

as follows.” Using (2.43) and (2.44) we discover, that at the reflection in the plane P the
following expression is true for the covariant spinor η:

ηβ → ηβ′ ǫβα Nτα∗ ξ τ ∗ = −ǫβα Nτα∗ ǫτ κ ǫκγ ξ γ∗ ,

With (2.35) and (2.44) taken into account, we finally have:

ηβ′ = −Nβα ηα (2.45)

So, the transformation laws for the η- and ξ-spinors are different at the reflections. It once
again proves the fact, that these two types of spinors may not be reduced to each other.
To put it differently, a linear transformation that would transform the spinors of the space
S (2̇) into ones of the space S (2) is absent, that is, such a transformation is identically equal
to zero. Actually, if such a transformation D existed, the following relations would be true:

η = Dξ, ηN = DξN , (2.46)

© 2011 by Taylor and Francis Group, LLC

 20 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where ηN = N η and ξN = N ξ, while the matrix D had to anticommute with N . Then

orienting the plane P sequentially along the coordinate axes x1 , x2 , and x3 we would have
obtained anticommutation of the matrix D with all the matrices σi . However, the Pauli
matrices and the unit matrix I constitute a system of linear independent matrices (it means,
that any 2 × 2-matrix can be expanded in terms of them). Hence, a 2 × 2-matrix that would
commute with all the Pauli matrices is absent, that is, we arrive at

{D, N } = 0 only at D = 0.

Spinors of the highest ranks are built by analogy with the tensors theory. The set 2n of
complex numbers α1 , α2 , ...αn transformed by the law:
 ′ ′ ′ ′
ψ α1 α2 ..αn = Uα′1 α1 Uα′2 α2 ...Uα′n αn ψ α1 α2 ..αn . (2.47)

is called a contravariant spinor of the rank n. Similarly, a covariant spinor of the rank n is
defined through its transformation properties:

′ ′ ′ ′
ψβ1′ β2′ ..βn′ = U β1 β1 † U β2 β2 † ...U βn βn † ψβ1 β2 ..βn . (2.48)

The set of quantities transformed by the rule:

 ′ ′ ′ ′ ′ ′ ′
α α ..α
ψβ ′1β ′2..β ′n = Uα′1 α1 Uα′2 α2 ...Uα′n αn U β1 β1 † U β2 β2 † ...U βl βl † ψβα11βα22..β
..αn
l
. (2.49)
1 2 l

is named a mixed spinor of the rank n + l. By means of simplification (summation) over

covariant and contravariant indices, one can obtain spinors of lowest ranks.
All representations with j > 1/2 are reducible. Let us consider, for example, a field
function, which is a mixed spinor of the second rank ψβα , that is, it is transformed by
the representation j = 3/2. Using the above-formulated diagram rule for expansion into
irreducible representations, we obtain:
O M
2 2=3 1, (2.50)

where 2 (2) denotes covariant (contravariant) components of the rotations group representa-
tion. Hereinafter the term “group representation” is used in two meanings: as an operator
transforming field functions, and as field functions transformed by a given representation.
Since the quantum field theory is a relativistic theory, it operates with quantities, defined
in the four-dimensional Minkowski space, hence, we have to attach the four-dimensional
aspect to the developed classification. The three-dimensional classification, however, will
not prove to be useless, since it can be successfully used to study properties of tensors and
spinors in abstract spaces (for example, in the space of an isotopic or color “spins”).

© 2011 by Taylor and Francis Group, LLC

 3
The four-dimensional Minkowski space

3.1 The homogeneous Lorenz group

Let us consider homogeneous Lorenz transformations, that is, linear transformations:

x′µ = Λµν xν , (3.1)

which leave the scalar product of four-dimensional vectors xµ yµ invariant. Since the squares
of three variables enter xµ yµ with the plus sign and one variable square enters it with the
minus sign, the Lorentz group is denoted by O(3.1). All the transformation coefficients Λµν
are real. From the condition:
xµ yµ = inv.

it results:
Λνµ Λµσ = Λνµ Λσµ = δσν . (3.2)

Thus, the following condition:

det Λ = ±1 (3.3)

must be true. A subgroup with the determinant to be equal to +1 is denoted by SO(3, 1).
Eq. (3.2) means, in its turn, that there exists the reverse Lorenz transformation. Since the
product of two Lorentz transformations is again a Lorentz transformation, then the set of
these transformations composes the group.
Having set ν = σ = 0 in Eq. (3.2), we obtain:

3
X

2
2
Λ00 =1+ Λi0 ≥ 1. (3.4)
i=1

From (3.4) it follows:

| Λ00 |≥ 1.

By this fact the Lorentz transformations , depending upon the signs of det Λ and Λ00 , are
divided into four categories:

1) proper orthochronous transformations L↑+ with det Λ = 1, Λ00 > 0);

2) nonproper orthochronous transformations (L↑− with det Λ = −1, Λ00 > 0);
3) proper nonorthochronous transformations (L↓+ with det Λ = 1, Λ00 < 0);
4) nonproper nonorthochronous transformations (L↓− with det Λ = −1, Λ00 < 0).

Besides continuous transformations (the rotations in the four-dimensional space), the

Lorentz transformations also contain two discrete operations, namely, the space and time

21
© 2011 by Taylor and Francis Group, LLC
 22 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

inversions (P and T ). They are defined as follows:

x0
    0   0 
1 0 0 0 x x
 x1   0 −1 0 0   x1   −x1 
P  2 =   2  =  , (3.5)
x 0 0 −1 0 x −x2
x3 0 0 0 −1 x3 −x3

x0
 0  
−x0
   
−1 0 0 0 x
1 1 1
x   0 1 0 0x   x 
T  2 =   2  =  2 . (3.6)
x 0 0 1 0 x x
x3 0 0 0 1 x3 x3
Hereinafter to write down 4 × 4-matrices we follow the agreement:

Λ ≡ Λµν ,

where the superscript numerates a column, and the subscript indicates a row. The transfor-
mations L↓± involve the time reflections, while L↑± do not. Analogously, the transformations
L↓↑ ↓↑
− contain the space inversion, while L+ do not. The homogeneous Lorentz group can be
symbolically represented as the sum:

L = L↑+ + P T L↑+ + P L↑+ + T L↑+ (3.7)

Since any Lorentz transformation can be presented as the product L↑+ , P , T , P T , we can
limit ourselves to studying the space-space rotations (the rotations in the planes [x1 x2 ],
[x1 x3 ] and [x2 x3 ]) and the purely Lorentz transformations (the transformations connecting
two inertial reference frames that move with the relative speed v∗ ) only. The totality of these
transformations constitutes a group of the limited homogeneous Lorentz transformations L↑+
(the proper orthochronous Lorentz group or, simply, the proper Lorentz group).
If the speed of relative movement of the inertial reference frames (IRFs) under consid-
eration is directed along the common axis x1 , then we have the following relation between
their coordinates:
 0 ′ 
(1 − β 2 )−1/2 −β(1 − β 2 )−1/2 0 0
 0 
x x
1 2 −1/2 2 −1/2
 x  −β(1 − β ) (1 − β ) 0 0   x1 
 2 =  2 , (3.8)
x 0 0 1 0 x
x3 0 0 0 1 x3

where β = v/c. Setting:

(1 − β 2 )−1/2 = cosh u, β(1 − β 2 )−1/2 = sinh u,

we can interpret the Lorenz boost along the axis x1 as a rotation in the plane x1 x0 by the
angle u = arctanh β. Actually,
′
x0
  0   0 
x cosh u − sinh u 0 0 x
 x1   x1   − sinh u cosh u 0 0   x1 
 2  = Λ(10; u)  2  =   2 . (3.9)
x x 0 0 1 0 x
x3 x3 0 0 0 1 x3

It is not out of space to emphasize the conventional character of such a interpretation of

the boost. Since tan u = iβ, the rotation proceeds by the imaginary angle. This approach,

∗ We shall also use for these transformations the term “boosts.”

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 23

however, allows one to interpret the limited homogeneous Lorenz transformations as the
group, which consist only of the rotations in the four-dimensional Minkowski space.
Let us determine the explicit form for generators of the Lorentz transformations. A
generator for the rotation in the plane [x1 x0 ] is defined by the formula:

d 
10 
J = Λ(10; u)
du 
u=0

and has the form  

0 −1 0 0
 −1 0 0 0 
J 10 = . (3.10)
0 0 0 0
0 0 0 0
Analogously, the generators J 20 and J 30 , corresponding to the boosts along the coordinate
axes x2 and x3 , are given by the expressions:
   
0 0 −1 0 0 0 0 −1
 0 0 0 0  0 0 0 0 
J 20 = , J 30 = . (3.11)
−1 0 0 0 0 0 0 0
0 0 0 0 −1 0 0 0

The generators of the space rotations in the four-dimensional notations look as follows:
   
0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0
J 12 = , J 23 = , (3.12)
0 −1 0 0 0 0 0 1
0 0 0 0 0 0 −1 0
 
0 0 0 0
0 0 0 −1 
J 31 = . (3.13)
0 0 0 0
0 1 0 0
Since the rotation in the plane [xµ xν ] by the angle ϕ is identical to that in the plane [xν xµ ]
by the angle −ϕ, then the equality: J µν = −J νµ takes place. An arbitrary infinitesimal
transformation of the homogeneous Lorentz group can be written down in the following
way:
1
Λ(ω) = I + ωµν J µν , (3.14)
2
where ωµν = −ωνµ . By means of direct calculations one can be convinced, that the gener-
ators J µν of the Lorentz group obey the permutation relations:

[J µν , J ρσ ] = g µρ J νσ + g νσ J µρ − g µσ J νρ − g νρ J µσ . (3.15)

3.2 Classification of irreducible representations of the homogeneous

Lorentz group
Suppose, there exists two IRFs: K and K ′ with the coordinates x and x′ , respectively. Let
one and the same wave field be described by a K-frame observer with the help of functions

© 2011 by Taylor and Francis Group, LLC

 24 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

ψi (x) and a K ′ -frame observer with the help of functions ψl′ (x′ ). Then values of ψi (x)
completely define those of ψl′ (x′ ), since x and x′ correspond to one and the same event in
the space-time. Thus, ψl′ (x′ ) must be expressed in terms of ψi (x). Consequently, under
transition from the first observer to the second one, transformation of the fields, described
by both, occurs according to the law:

ψ n′ (x′ ) = Dnk (Λ)ψk (x). (3.16)

It should be emphasized, that the matrix D depends on the Lorentz transformation matrix
and the indices n and k may have tensor dimensionality. For the sake of simplicity, the law
of the field transformation (3.16) can be written down in a more convenient form:

ψ n′ (x) = Dnk (Λ)ψk (Λ−1 x), (3.17)

The totality of matrices (or, what is the same, operators) D(Λ), which specify transforma-
tions of field components, constitute the Lorentz group representation. The matrix of any
representation of the limited homogeneous Lorenz group D(Λ) can be set down as follows:
1
D(Λ) = I + ω µν Mµν , (3.18)
2
where M µν compose a representation for the Lee-algebra generators, satisfying permutation
relations, which are similar to the permutation relations for the Lorentz group generators
(3.15):
[Mµν , Mρσ ] = gµρ Mνσ + gνσ Mµρ − gµσ Mνρ − gνρ Mµσ . (3.19)
Finding representations of the limited Lorentz group is reduced to determining all the
irreducible matrices, suiting to the relations (3.19). The group under consideration has
finite-dimensional and infinite-dimensional irreducible representations. From the point of
view of physical applications we are interested in classification of quantities with the fi-
nite number of components transformed according to a finite-dimensional representation
of the homogeneous Lorentz group. However, one should not go so far to lose an inter-
est in infinite-dimensional representations in general. Thus, unitary infinite-dimensional
irreducible representations of the Poincare group deal with the classification of elementary
particles. A thoughtful reader can guess at once, that in the last case we are dealing with
the presence in a complete set, classifying the representation basis, of such operators that
possess a continuous spectrum, not limited from above (as, for example, the momentum
operator). In other words, basis vectors of a representation have the finite number of com-
ponents, while the number of basis vectors as well as eigenvalues of an operator with a
continuous spectrum is infinite.
Now we consider finite-dimensional representations of the limited homogeneous Lorentz
group, namely the situation, when transformation matrices contain the finite number of
rows (or columns, since matrices are square). Let us define the operators

M = (M32 , M13 , M21 ), N = (M01 , M02 , M03 ). (3.20)

They obey the following permutation relations:

[Mi , Mj ] = εijk Mk , [Ni , Nj ] = −εijk Mk ,

[Mi , Nj ] = εijk Nk . (3.21)

Next, we pass on to new quantities according to the formulae:
1 1
X= (M + iN), Y= (M − iN). (3.22)
2 2

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 25

It is easy to show that they satisfy commutation relations to be already quite clear for
physical interpretation:
[Xi , Xj ] = iεijk Xk , [Yi , Yj ] = iεijk Yk , [Xi , Yj ] = 0. (3.23)
Thus, we have demonstrated that infinitesimal operators of the Lorentz group M µν could
be reduced to two independent vector operators X and Y to obey the permutation relations
for the moment of momentum projection operators. Therefore, we can obtain all irreducible
finite-dimensional representations of the homogeneous Lorentz group when the matrices for
the moments are taken as X and Y. Since the angular momentum square commutes with
its every projection
[X2 , Xi ] = 0, [Y2 , Yi ] = 0, (3.24)
then the operators X2 , Y2 are the Casimir operators. It means, as we already know, that
they are multiple to the unit operator:
X2 = j(j + 1)I2j+1 , Y2 = k(k + 1)I2k+1 , (3.25)
where j and k are arbitrary positive integer or half-integer numbers, I2j+1 (I2k+1 ) is the
identity matrix, containing 2j + 1 (2k + 1) rows and columns. So, every irreducible rep-
resentation of the limited homogeneous Lorentz group is characterized by two integer or
half-integer numbers j and k. This representation denoted by (j, k), has the dimensionality
(2j + 1)(2k + 1), since it is stretched over the totality (2j + 1)(2k + 1) of the basis vectors
| j, mj ; k, mk > satisfying the eigenvalue equations:
X2 | j, mj ; k, mk >= j(j + 1) | j, mj ; k, mk >, (3.26)
X3 | j, mj ; k, mk >= mj | j, mj ; k, mk >, (3.27)
2
Y | j, mj ; k, mk >= k(k + 1) | j, mj ; k, mk >, (3.28)
Y3 | j, mj ; k, mk >= mk | j, mj ; k, mk > . (3.29)
Let us find out for what representations complex-conjugated field functions are trans-
formed. Taking into account that
M∗ = M, N∗ = N,
and making use of (3.22), we conclude that at the complex conjugation the replacement:
X → Y, Y→X (3.30)
takes place. So, if | j, mj ; k, mk > is transformed under representation (j, k), then its
complex-conjugated field function is transformed under the mirror representation (k, j).
Under the space inversion the speed, entering into a boost, reverses its sign v → −v.
Then:
N → −N, M → M.
As a result we come to recognize that under the space inversion the transformations (j, k)
and (k, j) are interchanged by their places. A similar phenomenon takes place for the
time inversion. Thus, for the theory to be invariant under the P - or T -transformations,
the representation must either have the form (j, j) or must be the direct product of the
representations (j, k) and (k, j).
Depending on whether the number j + k is integer or half-integer, irreducible represen-
tations (j, k) of the limited homogeneous Lorentz group may fall in two categories. In the
former case, the space rotation by the angle 2π corresponds to the identical transformation
I (one-valued, or the tensor representations), and in the latter case, a field function reverses
its sign (two-valued, or the spinor representations).

© 2011 by Taylor and Francis Group, LLC

 26 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

3.3 Tensor representations of the limited homogeneous Lorentz

group
Let us give a definition of four-dimensional tensors. Quantity, which is not changed under
the transformations of the limited Lorentz group (3.1), is called a zero-rank tensor or a
scalar. The representation (0, 0) corresponds to the scalar. The totality of quantities which
transformation law at (3.1) has the form:

a′µ = Λµν xν , (a′µ = Λνµ xν ), (3.31)

where
∂x′µ
. Λµν =
∂xν
is called a contravariant (covariant) first rank tensor or a vector. The representation
(1/2, 1/2) corresponds to the vector.
Now we are quite ready to introduce a mixed tensor, which is contravariant with respect
to indices µ1 µ2 ...µn and covariant with respect to indices ν1 ν2 ....νk . It is a totality of
quantities with the following transformation properties:

Tν′µ1 ν12µ....ν
2 ...µn
k
= Λµσ11 Λµσ22 ...Λµσnn Λρν11 Λρν22 ...Λρνkk Tρσ11ρσ22...ρ
...σn
k
. (3.32)

The representations of the limited homogeneous Lorentz group (j, k)Nare irreducible.
N The
same representations, but composed by direct products like (j1 , k1 ) (j2 , k2 ) ... are re-
ducible. The recipe of the expansion of such products follows from the obvious generalization
of the formula (2.23): O X
(j1 , k1 ) (j2 , k2 ) = (j, k), (3.33)
L

where j =| j1 −j2 |, | j1 −j2 +1 |, .... | j1 +j2 |, k =| k1 −k2 |, | k1 −k2 +1 |, .... | k1 +k2 |, and
every representations (j, k) occurs once. Let us apply (3.33) to analyze a representation of
a contravariant tensor of the second rank T µν . This tensor is a product N of two four-vectors
and, consequently, is transformed by the representation (1/2, 1/2) (1/2, 1/2). Since:
O M M M
(1/2, 1/2) (1/2, 1/2) = (1, 1) (1, 0) (0, 1) (0, 0),

we could state, that the representation for T µν is reducible and falls into three irreducible
L
ones: a symmetric tensor with zero spur ((1, 1)), an antisymmetric tensor ((1, 0) (0, 1)),
and a scalar to be equal to the spur of the initial tensor ((0, 0)).

3.4 Spinor representations of the homogeneous Lorentz group

Let us now turn to the two-valued or spinor representations. In case of the first rank spinors,
the permutation relations for the generators Mk and Nk are satisfied both by the choice:

Mk = iσk /2, Nk = σk /2,

and by the choice:

Mk = −iσk∗ /2, Nk = σk∗ /2.

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 27

Consequently, as in the three-dimensional case, we can speak about two variants of a spinor
space or about two kinds of spinors, which we denote by ξ and η, as in the case of the
SO(3)-group. The operation of raising and lowering all spinor indices is again carried out
by means of an antisymmetric “metric tensor” ǫαβ . A spinor of the first kind will be a spinor
ξ α , which under the space rotations is transformed like the contravariant spinor under the
three-dimensional rotations, that is, according to the law:
  α
i(n · σ)
ξ ′α = exp θ ξβ , (3.34)
2 β

and under the hyperbolic rotations by the angle ϕ in the plane (x0 , n) (Lorentz boosts), its
transformation law has the form:
  α
(n · σ)
ξ ′α = exp ϕ ξβ . (3.35)
2 β

Thus, a choice of the generators for this kind of the spinors is as follows:

Y = 0, X 6= 0. (3.36)

It means, in its turn, that they are transformed under representation (1/2, 0). Choosing the
generators Mk and Nk in the form:
iσk∗ σk∗
Mk = − , Nk = , (3.37)
2 2
we obtain Y 6= 0 and X = 0, that is, the representation (0, 1/2). Under the SO(3, 1)-
rotations the transformation law for spinors, belonging to this representation, has the shape:
 α
′α (n · σ ∗ ) (n · σ ∗ )
η = exp [−i θ+ ϕ] η β . (3.38)
2 2 β

From Eq. (3.38) emerges, that:

 β
(n · σ) (n · σ)
ηα′ = exp [i θ− ϕ] ηβ , (3.39)
2 2 α

that is, the transformation laws for the spinors ξ α and ηα follow from each other at the
Hermitian conjugation. The spinors, whose transformation properties at the SO(3, 1)-
transformations are given by the formula (3.39) hereinafter we call the spinors of the second
kind. It is important to stress, that we deal with nonequivalent representations of the
limited homogeneous Lorentz group, that is, there exists no matrix S, for which

N = SMS −1

(N and M are the transformation matrices for the first- and second-kind spinors under the
SO(3, 1) rotations) is true.
At transformations of the space inversion the representations (j, 0) and (0, j) are inter-
changed by the places, and, as a result

ξ α ↔ ηα .

The inversion inclusion means the transition from the proper Lorentz group SO(3, 1) to
the Lorentz group 0(3, 1). Because of this the simplest irreducible spinor representation of

© 2011 by Taylor and Francis Group, LLC

 28 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
N
0(3, 1) is four-dimensional and composed by bispinor ξ α ηα , usually written down in the
form of a column matrix:
 α
ξ
Ψ= .
ηα

Thus, under the homogeneous Lorentz transformations the four-dimensional spinor Ψ is

transformed as follows:
 
  (nσ) (nσ)  
ξ exp [i θ + ϕ] 0  ξ

→ 2 2 . (3.40)
η (nσ) (nσ)  η
0 exp [−i θ+ ϕ]
2 2

Since matrices of transformation of the Lorentz boosts are not unitary, then finite dimen-
sional representations of the homogeneous Lorentz group are also nonunitary. It is a very
sad circumstance, since only for unitary representations of symmetry groups a probabil-
ity of a transition between two states does not depend on a reference frame, chosen for
measurements. Thus, we are facing the conflict with the relativity principle. Unlike the
three-dimensional rotations group, the proper Lorentz group is noncompact, which results
in nonunitarity of its representations. For the rotations group θ is varying from 0 up to 2π,
and these extreme points are identified in such a way, that the line is closed in the circle.
Therefore, the group space of the SO(3)-group is finite. The Lorentz group noncompactness
is caused by properties of the Lorentz boosts. Since the speeds, which are the Lorentz boost
parameters, take values in the open interval from v = 0 up to v = c, then the group space
of the SO(3, 1)-group is infinite and, as a consequence, this group is noncompact. Notice,
all of this is true only for finite-dimensional representations of the homogeneous Lorentz
group, while infinite-dimensional representations are unitary. There is something, however,
that forces us to search for other groups of space-time symmetry under describing elemen-
tary particles. The Casimir operators and group generators that commute with each other,
must fix a relativistic particle state. At the first glance on the homogeneous Lorentz group,
we are convinced that there are no such quantities among its invariants and generators,
which could define such important characteristics of elementary particles as mass, spin, and
momentum.
A fundamental group in elementary particle physics is a nonhomogeneous Lorentz group
(Poincare group) rather than the limited homogeneous Lorentz group. The Poincare group
consists of the Lorentz boosts, the three-dimensional rotations, and space-time translations.
We shall study it in the next subsection and now it is not out of place to dwell on terminology
we are going to use at a later time. So, as we have learned, there are two kinds of tensor
quantities in physics, namely, three- and four-dimensional ones. Three-dimensional tensors
are defined in the three-dimensional Euclidian space R3 with respect to the rotations group
SO(3), while four-dimensional tensors are defined in the four-dimensional Minkowski space
M4 with respect to the proper Lorentz group SO(3, 1). In the same way, there are two
different spinors, related to the groups SO(3) and SO(3, 1), respectively. If for R3 one
uses the ordinary coordinate space, then tensor and spinor quantities, given in it, describe
non-relativistic physics. One would think that it is more pertinent to call them “non-
relativistic tensors and spinors” and use the terms “relativistic tensors and spinors” for
the corresponding quantities, given in space-time continuum of Minkowski. However, such
a terminology will result in misunderstandings if we work not in coordinate, but in some
inner, abstract space. For the same reason we shall avoid the terms “relativistic” and “non-
relativistic” in tensor and spinor algebra. Instead, we shall use the terms “tensors” and
“spinors” in R3 or in M4 .

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 29

3.5 Poincare group and its representations

It turns out that the nonhomogeneous Lorentz group, the Poincare group, to perform trans-
formations between any IRFs, satisfies the relativity principle. Notice, that in quantum
theory the IRF is connected with the totality of macroscopic devices (laboratory), neces-
sary for the complete description of elementary particles system under study. The devices
participate in the quantum mechanic description twice: at first, by means of some device
a quantum object state is prepared and then measurements of objects are performed with
another device. The Poincare group transformation may be assigned a double meaning. On
the one hand, it can describe a change of space-time location of a physical system during
two measurements in the same IRF, that is: difference in location of two identical physical
systems with respect to the given IRF (active viewpoint). On the other hand, this trans-
formation can characterize the difference between two IRFs, in which one and the same
system is being studied (passive viewpoint). We shall stick to the active interpretation of
the Poincare group transformation.
The Poincare transformations P = {a, Λ} leave the square of a distance between two
four-dimensional vectors (x − y)µ (x − y)µ invariant. P are defined by:

x′µ = (Px)µ = Λνµ xν + aµ , (3.41)

that is, as the product of operations of the shift by the real vector aµ and of the homogeneous
Lorentz transformation Λ. It is important that the shift takes place after the homogeneous
Lorentz transformation. Product of two Poincare transformations P1 and P2 is equal to:

{a1 , Λ1 }{a2 , Λ2 } = {a1 + Λ1 a2 , Λ1 Λ2 }. (3.42)

Under active interpretation of the Poincare transformation with the element g = (a, Λ),
(in)
every initial physical state Ψ(in) is compared with a transformed state Ψg , which differs
from the initial state Ψ(in) by arrangement of preparing devices. At a creation of the state
(in)
Ψg , a facility with preparing devices has been either shifted, turned, or is evenly moving
with regard to the previous position (according to geometrical sense of g). In the same way,
every state Ψ(f ) , registered by measuring devices, can be compared with a transformed state
(f ) (in) (f )
Ψg . To find the probability of a transition Ψg → Ψg , which we denote by Wig →fg , we
must repeat the experiment Ψ(in) → Ψ(f ) (with the same or an identical physical system).
In so doing, all the preparing and measuring devices must be transformed with regard to
the initial state according to the geometrical sense of g. The existence of the space-time
symmetry exhibits in the equality:

Wi→f = Wig →fg ,

that is, in an invariance of the theory with respect to the Poincare group transformations.
Analogously, as in the case of the homogeneous Lorentz transformation, the Poincare group
can be divided in four classes differing according to signs of the quantities det Λ and Λ00 :
↑
1) P+ (det Λ = 1, Λ00 > 0);
↑
2) P− (det Λ = −1, Λ00 > 0);
↓
3) P+ (det Λ = 1, Λ00 < 0);
↓
4) P− (det Λ = −1, Λ00 < 0).

© 2011 by Taylor and Francis Group, LLC

 30 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Thus, we can write down a symbolic equality:

↑ ↑ ↑ ↑
P = P+ + P T P+ + P P+ + T P+ . (3.43)

Classification of elementary particle states deals, first of all, with studies of the proper
↑
orthochronous group P+ . This group contains ten generators: three generators M corre-
spond to the space-space rotations, three generators N denote boosts, and four generators
Pµ correspond to the space-time translations. We have already obtained expressions for
M and N in a matrix form. It turns out that to perform the same operation for transla-
tion operators we are forced to introduce a fifths fictitious coordinate without any physical
meaning. It remains unchanged under the Poincare transformations. Then transformations
(3.41) can be presented as follows:
′  0
x0 Λ01 Λ02 Λ03 a0
  0 
Λ0 x
1
x   Λ10 Λ11 Λ12 Λ13 a   x1 
1
 2
 x  =  Λ20 Λ21 Λ22 Λ23 a2   x2  .
  
(3.44)
 3  3
x Λ0 Λ31 Λ32 Λ33 a 3 3 
x
1 0 0 0 0 1 1
In order to make the generators be Hermitian quantities we introduce the factor −i into
Eq. (2.4). Then, for shift generators we obtain 5 × 5-matrices with the elements:
λ
(Pµ )β = −iδµλ δβ5 . (3.45)

By means of direct calculation one may be convinced that the following permutation rela-
tions:
[Mµν , Mρσ ] = −i(gµρ Mνσ + gνσ Mµρ − gµσ Mνρ − gνρ Mµσ ). (3.46)
[Pµ , Pν ] = 0, (3.47)
[Pσ , Mµν ] = i(gσµ Pν − gσν Pµ ) (3.48)
are valid. Eqs. (3.47)–(3.48) bring us to the idea that a shift generator must be identified
with a system momentum operator.
However, introduction of the fifths coordinate can cause vague anxiety in our readers:
“Maybe we are dealing with hidden dimensions of the space-time?” For this reason we shall
obtain Eqs. (3.46)–(3.48) in a less radical way. Recall about the existence of two versions
of quantum mechanics, namely, the Schrödinger wave mechanics and the Heisenberg matrix
mechanics. In the former case the differential form of operators is used while in the latter
case operators are represented in the matrix form. Let us also turn to the differential form
of writing the Poincare group generators. A generator, corresponding to parameter α(k)
(k = 1, 2 . . . , n, where n is a number of parameters in a group), is defined by its action on
the function ψ(x):  
ψ(x′ ) − ψ(x)
Lα(k) ψ(x) = −i lim ,
α(k) →0 α(k)
(to introduce the factor i ensures the Hermitian character of generators). For an infinitesimal
Poincare transformation we have:
6
!
X
′ ν (i) (i)ν
xµ = δµ + ǫ λµ xν + δaµ , (3.49)
i=1

(k) (k)
where λνµ = −λµν , and ǫ(i) and δaµ are infinitesimal parameters of the four-dimensional
rotations and the four-dimensional translations, respectively. By means of (3.49) it is pos-
sible to obtain a working formula to define the generator Lα(k) (by α(k) we mean all ten

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 31

parameters of the transformation (3.49) ):

   
∂x′1 ∂ ∂x′2 ∂
Lα(k) = −i + +
∂α(k) α(k) =0 ∂x1 ∂α(k) α(k) =0 ∂x2
      ′µ 
∂x′3 ∂ ∂x′0 ∂ ∂x ∂
+ − =i . (3.50)
∂α(k) α(k) =0 ∂x3 ∂α(k) α(k) =0 ∂x0 ∂α(k) α(k) =0 ∂xµ
By using the obtained formula, we find the expression for generators of the Poincare group

∂ ∂ ∂ ∂
Mk = −iεklm xl , Nk = i(t + xk ), Pµ = i . (3.51)
∂xm ∂xk ∂t ∂xµ
From the relations derived it follows that ~Mk is the orbital angular momentum projection
operator whereas the translation operator, multiplied by ~, represents the four-dimensional
energy-momentum vector pµ . One can be easily convinced of the fact, that the operators
Mµν and Pµ satisfy the commutation relations (3.46)–(3.48).
Relations (3.47) and (3.48) demonstrate, that the generators M and P commute with
the Hamiltonian P0 , what allows us to classify physical states according to eigenvalues of
these generators. The eigenvalues of the Lorentz boosts generators N cannot be used for
this purpose because they do not commute with P0 .
In case of the finite-dimensional Poincare transformations, the law of fields transformation
has the form:
ψl′ (x) = Bli (Λ)ψi [(Λ−1 (x − a)],
where the set of matrices B(Λ) forms representations of the Poincare group. Once again
to classify all irreducible unitary representations of the group one must find all the repre-
sentations of the permutation relations (3.46)–(3.48) in the form of Hermitian operators.
To achieve this goal we fix Casimir operators. Let us define the so-called Pauli-Lubanski
pseudovector :
1
Wσ = εσµνλ Mµν pλ . (3.52)
2
In Eq. (3.52) Mµν is a generator of representation of the proper Lorentz group, and
consequently, it is already a sum of both orbital and spin moments of a wave field. In the
three-dimensional vector notations Wσ has the form:

W 0 = p · M, W = p0 M − [p × N ]. (3.53)

It is obvious that the four-dimensional pseudovector Wσ is orthogonal to the four-dimensional

vector pσ
Wσ pσ = 0. (3.54)
The permutation relations for the Pauli-Lubanski pseudovector have the shape:

[Mµν , Wσ ] = i(gνσ Wµ − gµσ Wν ), [Wµ , pν ] = 0,

[Wµ , Wν ] = iεµνσλ W σ pλ . (3.55)

Now it is possible to check, that the scalar operators:

p µ p µ = m2 , w = −W µ Wµ , (3.56)

commute with all the generators Mµν and pµ , that is, they are the Casimir operators. Thus,
their eigenvalues can be used to classify irreducible representations of the Poincare group.

© 2011 by Taylor and Francis Group, LLC

 32 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

At m2 ≥ 0 one more invariant quantity, composed of momenta, exists. It is an energy sign

operator:
p0
ǫ= (3.57)
| p0 |
whose eigenvalues are ±1. Conservation of ǫ is caused by the fact that the proper or-
thochronous transformations of the Poincare group do not change a sign of a time compo-
nent of a time-like vector. Thus, a complete set of states, composing basis for the Poincare
group representations, breaks up on six different classes depending on the values of m2 and
sign of ǫ:
I. m2 > 0, ǫ > 0,
II. m2 > 0, ǫ < 0,
µ 2
III. p 6= 0, m = 0, ǫ > 0,
µ 2
IV. p 6= 0, m = 0, ǫ < 0,
V. pµ = 0,
V I. m2 < 0.
The first and the third classes correspond to a massive and massless physical particles
respectively, the fifth class—to vacuum, the sixth class—to virtual particles (which can
possess space-like momentum). The remaining classes are probably not physical ones. In
case of a massive particle we can perform a transformation to the reference frame where a
particle is at rest in which p = 0, p0 = m, and

W µ = m(0, M23 , M31 , M12 ) = m(0, S1 , S2 , S3 ).

It is easy to check, that the operators Sk (k = 1, 2, 3) satisfy the ordinary commutation

relations for the moment of momentum

[Sk , Sl ] = iεkln Sn . (3.58)

The quantity W 2 /m2 is equal to the moment of momentum square at rest, that is, the spin
square. Eigenvalues of S2 are J(J + 1) with J = 0, 1/2, 1, 3/2, 2, ..., and Sk are generators
of the irreducible (2J + 1)-dimensional representation of the three-dimensional rotations
group. Since W 2 is an invariant operator, then in any reference frame for the state | p, m >,
↑
transformed by irreducible representation P+ , the equality:

W 2 | p, m >= m2 J(J + 1) | p, m > .

will be true.
To establish covariant connection between Wµ and the spin operator Sk , we introduce
(k) (k) (k)
space-like normals nµ , with properties: nµ n(i)µ = δki , nµ pµ = 0 (k, i = 1, 2, 3). Together
(0)
with the speed vµ = pµ /m = nµ they make a complete system of the four-dimensional
normals:
n(λ) µν (σ)
µ g nν = g
λσ
. (3.59)
(0)
Since W µ nµ = 0, then there are only three independent space-like components of W µ ,
that is
Wµ = mSi n(i)
µ . (3.60)
(k)
Multiplying (3.60) by g µν nν , we obtain an expression for the spin operator in the covariant
form:
1
Sk = W µ n(k) µ . (3.61)
m

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 33

Besides the main invariants W 2 , m2 , and ǫ (at m2 > 0) there is one more operator in the
Poincare group:
2M
z = exp (2πiM3 ) = (−1) 3 , (3.62)
↑ J
which commutes with all the generators of the group P+ . Eigenvalues z are equal to (−1) ,
that is, for particles with an integer spin value (we call them bosons) z = +1, and for the
particles with a half-integer spin value (they are called fermions) z = −1. Basic vectors of
representations with a different z value are mutually orthogonal, so that superposition of
states with integer and half-integer spin value is impossible.
Thus, the irreducible unitary Poincare group representations are singled out by the values
of mass m, spin J, and the energy sign (at m2 ≥ 0)). Basic vectors of an irreducible
representation are also classified according to operator eigenvalues, entering one of the
complete sets, which are composed in turn, out of the generators Mµν and Pµ . For example,
in the canonical basis the complete set of mutually commuting operators contains operators:
1 µ (3)
m2 , S2, pi , S3 = W nµ . (3.63)
m
The vectors | +, m, J; p, J3 > constitute a basis of representation (“+” corresponds to the
energy sign). At fixed value of p for every pair (m, J) the representation basis consists
of 2J + 1 vectors. The fact that irreducible representation is infinite-dimensional, means
that any elementary system can take an infinitely large number of linearly independent
states. For each pair (m, J) and a given energy sign there is one, and only one irreducible
representation of the Poincare group, if we do not take into consideration unitary-equivalent
representations. At integer J a representation is one-valued, while at half-integer J it is
two-valued.
Any elementary particle can be defined as an elementary quantum mechanical system, all
the possible states of which are governed by functions, which realize in their totality some
invariant space of the Poincare group representations, in other words, by functions, which
are transformed under an irreducible representation of this group. A particle is elementary
in case it has no inner coordinate, which can divide its states in two or more groups in
the invariant manner. So, to describe elementary particles one must choose only relativistic
equations, based on the usage of the field functions, which realize irreducible representations
of the Poincare group. Important to note, that it is a very strong statement, because it
forces us to reject quite a wide class of equations. By way of example we cite Stückelberg
equations for the neutral [5] and charged [6] particles:

∂ µ ψµν + m2 ψν + ∂ν ψ = 0, 
ψµν = ∂µ ψν − ∂ν ψµ , (3.64)
ψ = ∂ µ ψµ .


These equations describe a particle which can dwell either in the state with spin 1 or with
spin 0.
However, we still have some ambiguity in choosing motion equations, since the homoge-
neous Lorentz group is in the mean time the subgroup of the Poincare group. Thus, for
example, a particle with spin 1 and mass m can be described by two different representations
of the homogeneous Lorentz group, namely, by:
M M M
(0, 1) (1, 0) (1/2, 1/2) or (0, 1) (1, 0). (3.65)

The well-known Proca equations [7]:


∂ µ ψµν + m2 ψν = 0,
(3.66)
ψµν = ∂µ ψν − ∂ν ψµ .

© 2011 by Taylor and Francis Group, LLC

 34 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

correspond to the first choice. In the second case we arrive at equations [8]:

∂ µ ψµνλ + m2 ψνλ + ∂ν ψλ − ∂λ ψν = 0, 
ψµνλ = ∂µ ψνλ − ∂ν ψµλ + ∂λ ψµν , (3.67)
ψν = ∂ µ ψµν .


As we can guess from these examples, under formulation of motion equations the homoge-
neous Lorentz group still plays an important role. The transformation law for field functions
under Poincare group transformations has a nonlocal character (which is caused by the
translation generator, the energy-momentum operator), while for the homogeneous Lorentz
group this law is local. However, it is clear that we should like to deal with quantities trans-
formed by the local law. It turns out that elimination of nonlocality in the transformation
law could be fulfilled by means of separating the transformation matrix Bli (Λ) into three
matrices, which already belong to representation of the homogeneous Lorentz group [9].
Below we give the traditional rules, which allow to further narrow the class of permissible
equations of motion for quantum fields.
1. Under the Lorentz transformations the field functions Ψk (x) must be transformed
by finite-dimensional (and consequently nonunitary) representation of the homogeneous
Lorentz group. In this case representation might be reducible and the number of components
could exceed 2J + 1.
2. The field functions must admit build-up of an invariant bilinear Hermitian form:
′ ′
Ψ†k (x)η kk Ψk′ (x) = Ψ†k (Λx)η kk Ψk′ (Λx), (3.68)
′
where elements of the Hermitian operator η kk are constants.
3. The field functions Ψk (x) must satisfy not only motion equations but must meet some
additional demands as well (they follow from motion equations), which ensure reduction of
independent components Ψk (x) up to 2J + 1.
The vector field described by Eqs. (3.68) contains six independent components. The
doubling of field function components is caused by the fact that a particle can be in states
with either positive or negative space parity. Then, the combination of states with the
opposite space parity will correspond to a particle without definite space parity. For the
first time the possibility, that such particles exist in the Nature, was mentioned in Ref. [10].
However, due to a very radical rule 3, such theories, similar to the one, described by Eqs.
(3.68) further on will be out of our consideration.
Now let us switch over to the analysis of representations belonging to the IIIth class.
In this case there are two different types of representations: 1) w = 0; 2) w = a2 , where
a is a real number. To characterize the first type of representations, we must introduce a
new quantum number, because w = m2 = 0. Since the isotropic vectors Wµ and pµ are
orthogonal to each other, the relation
W µ = −λpµ . (3.69)
(0)
must be fulfilled. Multiplying the both sides of Eq. (3.69) by the vector nµ being normal
to arbitrary space-like hypersurface we arrive at the relation:
(0) (0)
W µ nµ 1 εµνλσ Mνλ pσ nµ
λ=− (0)
= (0)
(3.70)
pµ nµ 2 pµ nµ
(0)
If we choose the normal vector as nµ = (1, 0, 0, 0), what corresponds to the transition to
the hyperplane t =const, then Eq. (3.70) produces:
M·p S·p
λ= = . (3.71)
p0 |p|

© 2011 by Taylor and Francis Group, LLC

 The four-dimensional Minkowski space 35

It follows from (3.71), that λ is a particle spin projection onto its motion direction, or
in other words, it is a helicity. Physical states of a particle with zero mass are classified
according to its helicity. One independent state corresponds to each value of helicity.
Under the space inversion P the helicity sign is reversed, because of
P P
S −→ S, p −→ −p.

Then, in the case of the P -invariance of the theory, for any nonzero spin value a particle
with zero mass has two polarization states (at J = 0, λ is also equal to zero). When the
P -invariance does not take place, the number of polarization states can be either equal to
two, or reduced to one. It is worthwhile to notice that the statements “the spin is parallel or
antiparallel to motion direction” are relativistically covariant for zero-mass particles, since
expression (3.71), determining λ, does not change its form at any Lorentz transformations.
Spin moment of a particle with nonzero mass can be also either parallel or antiparallel to
its velocity. These notions, however, are not Lorentz-invariant. If a velocity and a spin are
parallel in one reference frame it does not necessarily mean that they are parallel to each
other in an other reference frame. In particular, it is obvious, that in the very reference
frame, in which a particle is at rest, the spin moment is not parallel to the velocity, since the
latter is equal to zero. Any particle with a nonzero mass can be described in the reference
frame, in which it rests. For such a particle a statement, that its spin is parallel to its
velocity is not Lorentz-invariant. However, for a zero-mass particle moving with the light
velocity there is no reference frame where it would be at rest. It explains the difference
between zero-mass particles with only two polarization directions for J 6= 0, and particles
with the nonzero mass, which have 2J + 1 states of polarization for the spin J. For m = 0
and w = a2 the Poincare group representations are infinite-dimensional on the spin variable.
That is why, if corresponding particles existed, they would possess continuous spin. Even
if such particles exist, nature is successfully hiding them from us.
In general, there are other invariant operators too (they are related to the invariance of the
system with respect to the transformations in inner spaces, for example, operators of electric
charge, operators of baryon and lepton charge, and so on). Such operators commute with
the Casimir operators of the Poincare group and their eigenvalues characterize a physical
system state as well. Therefore, the more general writing for the basis vectors of space
of an irreducible representation is | +, m, J; p, J3 , α >, where α denotes eigenvalues of the
Casimir operators of the groups, related to dynamic symmetries.

© 2011 by Taylor and Francis Group, LLC

 4
Lagrangian formulation of field theory

4.1 Principle of least action. Lagrange-Euler equations

Components of a field function are functions of coordinates and time and satisfy some
definite differential equations, which are called field equations. Field equations can be given
a very general variational or Lagrangian form, if one accepts, that a field as the dynamic
system must be characterized by a definite Lagrangian function. For a mechanical system
this function is written down as the sum over all the material points of the system. For a
continuous system like a wave field, Lagrangian function is expressed by the space integral
from Lagrangian function density:
Z
L(t) = Ld3 x. (4.1)

However, the variational principle deals not with the Lagrangian function but with an action
S obtained from it by means of integration over x0 = ct
Z
S = c L(t)dt. (4.2)

Therefore, noncovariant expression (4.1) in Lagrangian formalism is in fact intermediate

and it is sufficient to consider the Lagrangian function density L, which depends on the
four space-time coordinates x in an implicit way. Hereinafter, we shall simply call L by the
Lagrangian. Analogously, the Hamiltonian function density H will be named the Hamil-
tonian. The Lagrangian depends on components of a field wavefunction ψk and on their
first derivatives with respect to the space-time coordinates ∂ψk /∂xν . In so doing we assign
the meaning of generalized field coordinates to ψk , and the meaning of generalized field
velocities to ∂ψk /∂xν .
The demand of lacking derivatives with respect to field functions above the first order in
the Lagrangian is caused by the following reason. Dynamics of all known quantum fields
is described either by equations of the first or the second order. On the one hand, these
equations follow from the principle of a least action:

δS = 0, (4.3)

and under action variation the order of derivatives, entering L, is raised by one. The La-
grangian must be a relativistically covariant quantity, namely, a relativistic scalar L′ (x′ ) =
L(x). Thanks to the invariance under the space-time translations, we conclude that the
Lagrangian must not manifestly contain any space-time coordinates x. Since dynamical
observables (four-dimensional vector of the energy-momentum, the moment of momentum
tensor, and so on) are real quantities, then the Lagrangian must be also a real quantity.
We shall also impose the locality condition on the Lagrangian, that is, we assume that the
Lagrangian depends only on the state of fields in the infinitely small vicinity of the point

37
© 2011 by Taylor and Francis Group, LLC
 38 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

x. So:
∂ψk
L(x) = L(ψk , ) = L (ψk (x), ψk;ν (x)) .
∂xν
General variation of an action, related with varying both field wave functions and of
integration domain boundaries, is equal to:
Z Z Z
′ ′
δS = δ L(x)dx = L(x )dx − L(x)dx, (4.4)

where dx = dx0 dr. For L′ (x′ ) we have the expression:

dL
L′ (x′ ) = L ψk′ (x′ ), ψk;ν
′
(x′ ) = L(x) + δL(x) = L(x) + δL(x) + λ δxλ , (4.5)
dx
In Eq. (4.5) we denote a form variation of corresponding quantities by the symbol δ. This
way δL(x) represents varying the Lagrangian only on the account of the variation of field
functions and field function derivatives:
∂L ∂L
δL(x) = δψk + δψk;ν , (4.6)
∂ψk ∂ψk;ν

and δψk (x) does a variation of a field function form:

δψk (x) = ψk′ (x) − ψk (x).

Thus, we have:
Z   Z Z
dL λ ′
δS = δL(x) + δx dx + L(x)dx − L(x)dx. (4.7)
dxλ
For an infinitesimal coordinate transformation:

xλ → x′λ = xλ + δxλ , (4.8)

volume elements in the new and old coordinate systems are connected by the relation:
 
∂(x′0 , x′1 , x′2 , x′3 ) ∂δxλ
dx′ = dx′0 dx′1 dx′2 dx′3 = dx ≈ 1 + dx. (4.9)
∂(x0 , x1 , x2 , x3 ) ∂xλ
Then, in the final form the action variation becomes:
Z Z
∂ λ

∂L ∂ ∂L
δS = [δL(x) + λ L(x)δx ]dx = [ − ]δψk dx+
∂x ∂ψk ∂xλ ∂ψk;λ
Z  
∂ ∂L λ
+ δψ k + Lδx dx. (4.10)
∂xλ ∂ψk;λ
Let us consider the action variation at fixed integration boundaries, in the mean time
assuming, that wavefunctions variation at these boundaries is equal to zero. With the help
of the four-dimensional Gauss theorem the second integral in Eq. (4.10) can be rewritten
as follows: Z Z
∂ λ λ
D (x)dx =
λ (n)
D(n) (x)dσλ , (4.11)
Ω ∂x Σ
where
λ ∂L
D(n) (x) = δψk + Lδxλ
∂ψk;λ

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 39

and Σ is an arbitrary closed space-like hypersurface, restricting four-dimensional volume Ω.

As integration domains one should choose space-like hypersurfaces Σi and Σf corresponding
to the initial and final states, and, as a closing surface, one should take an arbitrary space-
like hypersurface Σ0 , connecting Σi and Σf . Then, we obtain:
Z Z Z Z
λ λ (i) λ (f ) λ
D(n) (x)dσλ = D(n) (x)dσλ + D(n) (x)dσλ + D(n) (x)dσλ . (4.12)
Σ Σi Σf Σ0

Taking into consideration arbitrariness in choice of hypersurface Σ0 , we carry it away on

infinity. Then, for all points belonging to Σ0 we have

x2ν = c2 t2 − r2 → −∞

or
| r |→ ∞.
However, on infinity there are no fields, what means:
Z
λ
ψ(x) |x∈Σ0 = 0, or D(n) (x)dσλ = 0.
Σ0

As a result Eq.(4.12) takes the form:

Z Z Z
λ λ (i) λ (f )
D(n) (x)dσλ = D(n) (x)dσλ + D(n) (x)dσλ .
Σ Σi Σf

Since δψk and δxλ at the initial and final space-like hypersurfaces are equal to zero, the
right-hand side of Eq.(4.11) turns into zero. Then the Lagrangian-Euler equations:
∂L ∂ ∂L
− = 0, (4.13)
∂ψk ∂xλ ∂ψk;λ

follow from condition (4.3) and arbitrariness of integration domain. These equations are
interpreted as field motion equations. Since physical properties of the system are governed
by the action (4.3) then the Lagrangian L(x) is defined up to a four-dimensional divergence.
It means, that Lagrangians:
∂
L(x) and L(x) + Fλ
∂xλ

result in the same field motion equations, since the variation of the term ∂F λ /∂xλ gives
zero contribution due to the four-dimensional Gauss theorem.

4.2 Noether theorem and dynamic invariants

For the complete description of physical systems one must know not only the motion equa-
tions, but the main characteristics of these systems as well. As such characteristics it is
convenient to use quantities, constant in time, which we call either integrals of motion or
dynamic invariants. It turns out, that dynamic invariants are connected with symmetry
of a physical system with respect to a definite transformation group. This connection is
described by a theorem, proved by E. Noether [11].

© 2011 by Taylor and Francis Group, LLC

 40 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Let us assume that there is some finite-parametric (depending on l constant parameters)

continuous transformation of coordinates and field functions under which action variation
vanishes. Then l dynamic invariants—l combinations of field functions and field function
derivatives that are conserved in time—correspond to such a transformation.
To prove this, we consider infinitesimal coordinates transformation:

λ
xλ → x′λ = xλ + δxλ , δxλ = Xnλ δω n = I(n) ν xν δω n . (4.14)

Under (1.14) the field functions are transformed by the law:

j
ψi (x) → ψi′ (x′ ) = ψi (x) + δψi (x), δψi (x) = Yi(n) δω n = K(n) i
ψj δω n . (4.15)

In Eqs. (4.14) and (4.15) index n takes the values 1, 2, ...l, ω n are transformation param-
eters, I(n) are generators of a coordinate transformations group, and K(n) are generators
of representation of this group. Indices of field functions i and transformation parameters
n can have either tensor or spinor contents (it will be specified in every particular case).
Notice, that a transformation law of field function derivatives:
′
ψi;k (x) → ψi;k (x′ ) = ψi;k (x) + δψi;k (x) (4.16)

contains variations δψi;k (x), which are not derivatives with respect to δψi . In other words,
the operations δ and ∂/∂x are not commutative. It is caused by the fact that δψi is the
variation of a field function due to both change of its form and change of the argument.
The form variation of a field function δψi (x) up to infinitesimal of the second order can be
presented as follows:
 
δψi (x) = δψi (x) − ψi;k δxk = Yi(n) − ψi;k X(n)
k
δω n . (4.17)

By definition the operation δ is commutative with that ∂/∂x.

Now we turn to the general expression for the action variation. We consider Eq. (4.10)
for true motion, that is, when the Lagrangian-Euler equations are fulfilled. Further we
suppose, that an integration domain Ω together with a wave field as a whole is subjected
either to infinitesimal translation or to infinitesimal rotation. Then, from (4.10) follows:
Z
∂
δS = −δω n Dλ (x)dx = 0,
λ (n)
(4.18)
Ω ∂x
where quantities:

λ ∂L(x)  ν

λ
D(n) (x) = ψi;ν X(n) − Yi(n) − L(x)X(n) (4.19)
∂ψi;λ
hereinafter we will call the Noether currents. Using arbitrariness of integration domain over
four-dimensional volume and Eq. (4.18), we obtain the continuity equation:
∂
Dλ (x) = 0. (4.20)
∂xλ (n)
Integration of Eq. (4.20) over the volume V results in the expression:
Z Z Z I
0 l 0 l
∂0 D(n) (x)dr + ∂l D(n) (x)dr = ∂0 D(n) (x)dr + D(n) (x)dSl =
V V V S
Z
0
= ∂0 D(n) (x)dr = 0, (4.21)
V

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 41

where we take into account the three-dimensional Gauss theorem and the fact that any
fields are absent at infinity.
In its turn, from (4.21) follows, that the integrals:
Z
0
C(n) = D(n) (x)dr = const (4.22)

do not depend on time. It is possible to rewrite the expression (4.22) in manifestly rela-
tivistically invariant form. To reach this purpose let us use the definition of the functional
derivative of the functional F (σ) (see Eq.(1.9)). In the case when Fµ (σ) is represented by
an integral over space-like hypersurface
Z
Fµ (σ) = F (x)dσµ ,
σ

calculations give:
Z Z 
δ F (x)dσµ
Fµ (σ) = lim − = ∂µ F (x). (4.23)
δσ(x) Ω→0 σ σ′ Ω

Because of this the relation:

δ
F (σ) = ∂µ F µ (x) (4.24)
δσ(x)

is carried out for the functional:

Z
F (σ) = F µ (x)dσµ
σ

where F µ (x) is an analytical function. Then, from Eq. (4.24) follows, that the functional
F (σ) does not depend on the choice of the hypersurface σ (δF (σ)/δσ = 0) if Fµ (x) satisfies
λ
the continuity equation. Since for D(n) (x) Eq.(4.20) is fulfilled, the equivalent form of
writing (4.22) is
Z
µ
C(n) = D(n) (x)dσµ = const, (4.25)
σ

as due to arbitrariness of σ, it could be always chosen as hyperplane with t=const.

Thus, we proved, that l dynamic invariants C(n) (n = 1, 2, ..l) correspond to every con-
tinuous l-parametric transformation of coordinates (4.14) and field functions (4.15) that
converts the action variation to zero. The relations (4.20) and (4.25) present mathematical
writing of the conservation laws in the integral and differential forms, respectively. It is
λ
worthwhile to notice, that the Noether currents D(n) (x) are ambiguously defined. Actually,
λ
the continuity equations will be satisfied for both D(n) (x) and

′λ λ ∂ [νλ]
D(n) (x) = D(n) (x) + f (x), (4.26)
∂xν (n)
[νλ]
where f(n) (x) are arbitrary functions, which are antisymmetric in indices λ and ν. Let us
proceed to the concrete definition of dynamic invariants and symmetry groups connected
with them.

© 2011 by Taylor and Francis Group, LLC

 42 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

4.3 Energy-momentum tensor

The opportunity of selecting the natural laws among the chaos of phenomena surrounding
us is based on two reasons. First, in most cases it is possible to determine the plurality
of initial conditions amenable to adjustment that is important for the phenomenon under
study. Such a definition is not a trivial one, it makes the foundation of art in experimental
investigation of phenomenon. Second, under the same given initial conditions, the experi-
mental result will be always the same, regardless of place and time of experiment. In the
language of initial conditions it appears as follows: absolute location and absolute time are
never essential initial conditions. In other words, the laws of physics do not depend on the
choice of the origin of coordinates both on the space and on the time axes, that is, they
are invariant under translations along the coordinates axes in four-dimensional space. If
correlations between events of the same events sequence were changing day after day and
were different at different points in space, then the existence of physics would be impossible.
However, it may turn out that invariance with respect to the space-time translations, which
we connect with homogeneity of the time and space, could bear an approximate charac-
ter. The application of the translational invariance postulate on cosmic scales presumes
the homogeneous and stationary character of the universe. Nowadays, the Big Bang theory
succeeded the stationary universe model, which was keeping a delicate silence about the
genesis of our world. The Big Bang theory does not rule out the possibility of homogeneity
violation in the space-time continuum. At this point we should calm too impressionable
readers, that such regions of the universe have not been discovered by now. Space and time
homogeneity means the Lagrangian invariance under the space and time translations group.
We shall demonstrate that the conservation law of the energy-momentum four-dimensional
vector corresponds to such an invariance. As this takes place, the three-dimensional mo-
mentum conservation law follows from space homogeneity, while the energy conservation
law follows from the time homogeneity.
The invariance of a theory with respect to the four-parametric group of the four-dimensional
translations:
x′ν = xν + aν ,
will be ensured automatically if one takes into account that under these transformations
field functions do not change their form, that is:

ψk (x) = ψk′ (x′ ) = ψk′ (x + a). (4.27)

In the case of infinitesimal translations:

x′ν = xν + δaν = xν + δxν , (4.28)

we have
δω µ = δxµ , Xµν = δµν ; δψk = 0, Yi(n) = 0;
ψk (x) = ψk′ (x) + ψk;ν (x)δxν , δψk = −ψk;ν (x)δxν .
λ
Thus, in the case under consideration, the Noether current D(n) (x) proves to be the second
rank tensor:
∂L ∂ψk
Tµ′ν = − Lδµν , (4.29)
∂ψk;ν ∂xµ
satisfying the continuity equation:
∂ν Tµ′ν = 0. (4.30)

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 43

The corresponding dynamic invariant is:

Z
Pν = T ν0 dx. (4.31)

In classical mechanics the time component of this vector represents the Hamiltonian func-
tion, that is, the energy. From reasons of the relativistic covariance it follows that the
four-dimensional vector P ν is the energy-momentum vector. The tensor defined by Eq.
(4.29) shall be called the canonical tensor of the energy-momentum. In the most general
case it is not symmetric. However, with allowance made for ambiguity in its definition
(see Eq. (4.25) ), it is always possible to make the energy-momentum tensor symmetric by
[λν]
adding a specially chosen quantity of the form: ∂λ fµ .

4.4 Tensor of angular momentum

As stated before, it would be impossible to discover the laws of nature, if they were not
invariant under the space-time translations. The same would be true in the case of lacking
an invariance under the space rotations. As experiments show, three-dimensional space
surrounding us is not only homogeneous, but also isotropic. Space isotropy means that
properties of the space are equivalent in all directions, that is, properties of a closed system
do not change under coordinate system rotations. Of course, it is difficult to check the
space isotropy by direct experiment in conditions of the earth’s gravitation or any other
gravitation. In this case there exists obvious difference between “up” and “down” direc-
tions, as well as between “left” and “right” directions. Notwithstanding the fact, Newton
formulated the gravitation theory in such a way that the motion equations were invariant
under rotations. A conservation law, following from three-dimensional space isotropy, is
known to us as a conservation law of three-dimensional angular momentum L. In this case
the connection established by the Noether theorem, has the following form

space isotropy→SO(3) invariance → L=const

However, from the relativity principle it follows, that physical laws are also invariant under
rotations in the space-time planes. Due to this reason the four-dimensional space of the
universe, in which we have the luck to live, is isotropic or invariant with respect to the
six-parametric group of the four-dimensional rotations, the SO(3, 1)-group.
Under infinitesimal four-dimensional rotations the coordinate transformation law has the
form:
x′ν = xν + xµ δω [µν] = xν + δxν , (4.32)
where, as we see, the index (n) fell into indices µ and ν (µ < ν) denoting the plane, in
which the rotation with the parameter δω [µν] takes place. δxν is given as:

δxν = X[µσ]
ν
δω [µσ] = xµ δω [µν] = xµ δω [µσ] δσν =

= δω [µσ] (xµ δσν − xσ δµν ). (4.33)

From Eqs. (14.33) and (4.14) it follows
ν


X[µσ] = xµ δσν − xσ δµν , I νλ σµ
= δµλ δσν − δµν δσλ . (4.34)

© 2011 by Taylor and Francis Group, LLC

 44 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

λ
So, now the Noether current Dµν (x) is the third rank tensor having the form:

∂L(x)
λ
Mµν (x) = {ψi;σ (x)[xν δµσ − xµ δνσ ] − (Kµν )ji ψj (x)} + L(x) xµ δνλ −
∂ψi;λ (x)

∂L(x) j
− xν δµλ = xν Tµ′λ − xµ Tν′λ − (Kµν )i ψj (x). (4.35)
∂ψi;λ (x)
j λ
For a scalar field (Kµν )i = 0 the relation between the tensor Mµν (x) and the energy-
momentum tensor takes the same form as in the classical mechanics of the material point.
For this reason the quantity:

L[µν]λ = xν T ′µλ − xµ T ′νλ = T ′σλ (I µν )στ xτ (4.36)

should be interpreted as the intrinsic angular moment of the field. Since in the case of a
j
multicomponent field (Kµν )i 6= 0, the second quantity in the expression (4.35):

λ ∂L(x) j
S[µν] (x) = − (Kµν )i ψj (x). (4.37)
∂ψi;λ (x)

is logically identified with the spin moment tensor of particles to be described by the wave
field. Thus, the Lagrangian invariance with respect to the four-dimensional rotations group
in the Minkowski space resulted in the conservation of the sum of the spin and orbital
moments of the field:
Z Z  
[µν] [µν]0
M = M dr = L[µν]0 + S [µν]0 dr = L[µν] + S [µν] . (4.38)

Carrying out the convolution of space components of the spin moment tensor with the
three-dimensional antisymmetric third rank tensor εαβγ , we obtain components of the three-
dimensional spin pseudovector:
1
Sα = εαβγ Sβγ , (4.39)
2
in which, as it should be expected, we recognize a classical analog of the spin operator,
namely, the Pauli-Lubanski pseudovector.
As it follows from (4.36) for any field the orbital moment tensor is defined by the same Her-
mitian generators of the Lorentz group (I µν )στ , while the spin moment tensor is expressed
j
by means of infinitesimal operators (Kµν )i to form the Lorentz group representation for
functions of a given field.
λ
Now we should like to get rid of arbitrariness in definition of the Noether currents D(n) (x).
For this purpose it would be very attractive to use some physical principles. Let us turn to
the general relativity theory where one of the most important elements of the theory is the
energy-momentum tensor Tµν . The Einstein equations have the form:

1 8πGN
Rµν (η) − R(η)ηµν (x) = Tµν (η), (4.40)
2 c4
where ηµν (x) is a metric tensor, Rµν (η) is the Ricci tensor, R(η) = Rµν (η)η µν and GN is
the Newton constant. Since the tensors Rµν (η) and ηµν are symmetric, then the energy-
momentum tensor must be of the same property. Further, remembering the fact that the
energy-momentum tensor is automatically symmetric for a scalar field, we understand that
the symmetrization procedure of this tensor for fields with the spin is connected with usage of
λ
the spin moment tensor combinations instead of f[µν] . Following these decisive statements,

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 45

after straightforward calculations we arrive at the symmetric (metric) energy-momentum

tensor:
(metr) ′ (metr)
Tµν = Tµν + ∂λ f[µλ]ν = Tνµ , (4.41)
where
1

f[µλ]ν =
Sµ[νλ] + Sλ[µν] + Sν[µλ] .
2
Then connection between the total moment tensor and the energy-momentum tensor is of
the same form as in material point mechanics:
M [µν]λ = xν T (metr)µλ − xµ T (metr)νλ . (4.42)

This writing allows us to interpret the obtained results as follows. The SO(3.1)-invariance
of the theory results in conservation of the quantity:
Z  
M [µν] = xν T (metr)µ0 − xµ T (metr)ν0 dr, (4.43)

where invariance with respect to the three-dimensional rotations is associated with conser-
vation of the space components M [12] , M [23] , and M [31] (the total moment of momentum
of the field), whereas invariance with respect to the Lorenz boosts leads to conservation of
the space-time components M [01] , M [02] , and M [03] , describing motion of the system mass
center.

4.5 Electromagnetic current vector and electric charge

Thus, we considered dynamic invariants relating with symmetry in the Minkowski space.
Besides them, other dynamic invariants are introduced into particle physics, so-called inner
quantum numbers, caused by symmetries of physical systems with respect to transforma-
tions in abstract spaces. Corresponding symmetries are called “nongeometrical, inner” or
“dynamic”. The electric charge conservation law is an excellent example illustrating the
connection:
inner symmetry → invariance → conservation law .
Let us consider the arbitrary field, described by N-component complex functions ψk (x)
and ψk† (x) = (ψk∗ )T (k=1,2,. . . ,N). The demand for the field Lagrangian L(x) to be real
means, that it contains only bilinear combinations of field functions and their first derivatives
of the following form:

ψk† (x)Akl ψl (x), ∂ µ ψk† (x)Bµν

kl ν
∂ ψl (x), (4.44)

∂ µ ψk† (x)Cµkl ψl (x), ψk† (x)Cν′kl ∂ ν ψl (x), (4.45)

where Akl , Bµν
kl
, Cµkl , Cν′kl are quantities that do not depend on x, and the indices k and l can
have either tensor or matrix dimensionality. Thus, for example, for a free electron-positron
field we have:
i 0 µ kl
Akl = −m(γ 0 )kl , kl
Bµν = 0, Cµ′kl = −Cµkl = (γ γ ) ,
2
where γµ is 4 × 4 Dirac matrices, satisfying relations of Clifford algebra:

{γ µ , γ ν } = 2g µν , (4.46)

© 2011 by Taylor and Francis Group, LLC

 46 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

k and l are matrix indices. From Eqs. (4.44) and (4.45) it follows that the field functions are
defined with a precision of an arbitrary phase factor. It means, that under transformation:

ψk (x) → ψk′ (x) = U (α)ψk (x) = exp (iα)ψk (x),
(4.47)
ψk† (x) → ψk′† (x) = U † (α)ψk† (x) = exp (−iα)ψk† (x),

where α is an arbitrary constant, the system Lagrangian remains invariant. In other words,
physical reality, corresponding to the description in terms of the old (ψk (x)) and new (ψk′ (x))
field functions, is the same. Transformations U (α) make up one-parametric group of the
global gauge transformations (transformations are global, since phase α does not depend
on x). Sometimes such a group is called the gauge transformation group of the first kind.
The U (α)-group is unitary, that is:

U (α)U † (α) = I,

where I is a unit matrix. This group is Abelian as well, since all its elements commute with
each other. Then, under infinitesimal transformation of field functions:

ψk′ (x) = (1 + iδα)ψk (x) = ψk (x) + δψk (x), (4.48)

from the Noether theorem it follows the continuity equation for four-dimensional current
vector:
∂µ j µ (x) = 0, (4.49)
where !
∂L(x) ∂L(x)
µ
j (x) = i ψk† (x) − ψk (x) . (4.50)
∂[∂µ ψk† (x)] ∂[∂µ ψk (x)]
It leads, in its turn, to conservation of the corresponding charge:
Z
Q = j 0 (x)dx = const. (4.51)

It is obvious, that if the current j µ satisfies the continuity equation, the same is true for
the current to be equal to const × j µ . In the method we used, a unit of measurement of
the charge was not fixed. It can be fixed only by means of additional physical assumptions.
Setting α = qα0 , where q is an electric charge of particles, corresponding to a wave field,
we arrive at the electric charge conservation law:
Z Z !
0 † ∂L(x) ∂L(x)
Qem = jem (x)dr = iq ψk (x) − ψk (x) dr = const. (4.52)
∂[∂0 ψk† (x)] ∂[∂0 ψk (x)]

The electric charge belongs to additive characteristics. The experimental test of the electric
charge conservation law is based on a test of electron stability and the zero mass of a
photon. Analysis of possible events of atmospheric electricity, which could appear as a
result of electron decays in the atmosphere gives for the lowest limit of the electron life
time > 1021 years. The existence of a large-scale magnetic field in disk components of the
Galaxy leads to the most strict upper limit on the photon mass ≤ 10−27 eV/c2 . Strict
equality in the absolute value of the proton and electron charges also confirms the electric
charge conservation law. All of these facts make rather problematic the existence of theories
to admit violation of (4.52) and give us all the grounds to consider the electric charge
conservation law to be exact. That means that any processes, which are proceeding with
its violation, are forbidden.

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 47

It is worthwhile to note, that in the case of the real fields

ψk† (x) = ψkT (x),

and the expression for j µ (x) turns into zero. In the heat of the moment one may state
that complex field functions describe charged particles, while real functions are only used
for neutral particles. What actually happens is that the real functions are only used for
true neutral particles, that is, in the case when a particle is equivalent to its antiparticle
(photon, π 0 meson, and so on). Clearly, simply neutral particles generate no electric cur-
rents. However, such particles possess another sorts of “currents” and “charges” related
with these currents. We shall turn back to this topic later.
For further analysis it is useful to give geometrical form to the gauge transformation
(4.47). For the sake of simplicity, let us consider one-component field φ(x) (such fields
describe spinless particles). The field functions φ(x) and φ∗ (x)(x) can be presented in the
form:
φ1 (x) + iφ2 (x) φ1 (x) − iφ2 (x)
φ(x) = √ , φ∗ (x) = √ , (4.53)
2 2
where φ1 (x) and φ2 (x) are real quantities. Then the gauge transformations (4.47) will look
like: 
φ′1 (x) + iφ′2 (x) = exp (iα)[φ1 (x) + iφ2 (x)],
(4.54)
φ′1 (x) − iφ′2 (x) = exp (−iα)[φ1 (x) − iφ2 (x)].
Since Eqs. (4.54) could be shaped to the form:
 ′   
φ1 (x) cos α sin α φ1 (x)
= , (4.55)
φ2 (x) − sin α cos α φ2 (x)

then it is obvious, that the gauge transformations (4.47) may be interpreted as rotations
of the vector φ(x) = (φ1 (x), φ2 (x)) about the angle α. In group theory slang the above
said means, that the group U (1) is locally isomorphic to the orthogonal rotations group
SO(2) in two-dimensional space of the real functions φ1 (x) and φ2 (x). As α = const, this
transformation must be the same for all points of the space-time continuum, that is, it
is the “global” gauge transformation. To put this another way, when a rotation in one
point by the angle α in the inner space of the field φ(x) is made, the same rotation in all
other points must be simultaneously carried out. We would have no reason to worry, if
a conserved quantity related to invariance under this transformation were not the source
of physical field. The electric charge, however, is producing the electromagnetic field in
space and the electromagnetic interaction is spreading at the finite speed, not instantly. To
avoid conflict with the short-range interaction theory (or the field concept), we are forced
to localize the gauge transformation U (1). It means that we consider the phase α to be
different for various space-time points, that is, α = α(x). The corresponding transformation
is called the “local” gauge transformation or the gauge transformation of the second kind.
Let us investigate the physical consequences of theory invariance under the local gauge
transformations. As an example, we refer to the electron-positron field ψ(x), for which the
Lagrangian function density is given by the expression:
i 
L= ψ(x)γ µ ∂µ ψ(x) − ∂µ ψ(x)γ µ ψ(x) − mψ(x)ψ(x). (4.56)
2
where the line above the field function means the Dirac conjugation, that is: ψ(x) =
(ψ ∗ (x))T γ 0 . The field function transformation law under the local gauge transformations of
the U (1)-group has the form:

ψ(x) → ψ ′ (x) = exp [iα(x)]ψ(x). (4.57)

© 2011 by Taylor and Francis Group, LLC

 48 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Due to the presence of the derivative the Lagrangian (4.56) is not invariant under these
transformations, since:

∂µ ψ(x) → exp [iα(x)][∂µ ψ(x) + iψ(x)∂µ α(x)]. (4.58)

For the Lagrangian to be invariant, we must introduce a new derivative in such a way that
a derivative of the field function is transformed in the same manner as the field function
itself:
Dµ ψ(x) → exp [iα(x)]Dµ ψ(x). (4.59)
The relation (4.59) will be fulfilled under the condition that

Dµ = ∂µ − igAµ (x), (4.60)

where under the local transformations (4.57) the introduced vector field Aµ (x) must behave
by the following way:
Aµ (x) → Aµ (x) + g −1 ∂µ α(x). (4.61)
We call new derivative Dµ a covariant derivative. Since, besides the fermions field our
system contains the vector field as well, the Lagrangian (4.56) must be supplemented with
a free Lagrangian of the field Aµ (x), which must not violate the local gauge invariance and
be relativistically covariant at the same time. If we call for fulfillment of the superposition
principle, then only a quantity like:

aFµν (x)F µν (x) + bF̃µν (x)F̃ µν (x), (4.62)

where
1
Fµν (x) = ∂µ Aν (x) − ∂ν Aµ (x), εµνλσ F λσ (x),
F̃µν (x) =
2
is possible. Now, to obtain the Lagrangian of quantum electrodynamics (the theory of
electromagnetic interaction between electrons and positrons) it is sufficient to identify g
with the electric charge of the electron e, Aµ (x)—with the four-dimensional potential of the
electromagnetic field and choose the coefficients in (4.62) as follows:

a = −1/4, b = 0.

The corresponding Noether current (the electromagnetic current) and the conserved charge
(the electric charge) are defined by the expressions (4.51) and (4.52) multiplied by e. Thus,
the Lagrangian invariance with respect to the local gauge transformation group not only
ensures the electric charge conservation, but results in harmony with the special relativity
theory as well. The appearance of the gauge boson related to the gauge group U (1), the
electromagnetic interaction carrier, which is called a photon, is one more consequence of the
locality of this group. Thus, if one distracts from some arbitrariness in the choice of the
free field Lagrangians, then the aforesaid is a strong argument in favor of the fact that the
local gauge invariance represents a fundamental principle laying in the basis of a theory of
any interaction.

4.6 Isotopic spin

It is known from the experiment, that if one neglects electromagnetic interaction, then it
is possible to unify some elementary particles in families, the so-called isotopic multiplets.

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 49

Particles entering these multiplets have identical properties and approximately the same
masses. The well-known examples are the isotopic doublet of nucleons (p, n) and the
isotopic triplet of pions (π + , π 0 , π − ). The observed mass spectrum for members of the
above-listed multiplets has the form:

mp = 938.27231 ± 0.00028 MeV/c2 , mn = 939.56563 ± 0.00028 MeV/c2

mπ± = 139.56995 ± 0.00035 MeV/c2 , mπ0 = 134.9764 ± 0.0006 MeV/c2 .

Believing, that under the absence of electromagnetic interaction masses of members of the
same multiplet coincide, we can consider particles of a multiplet as different states of one
and the same particle, every state being marked by a definite charge value. By analogy with
the ordinary spin, we confront every multiplet with so-called isotopic spin, whose number
of projections is equal to the number of particles in the multiplet; we also confront every
projection with a particle of definite charge. We also introduce abstract (inner) three-
dimensional space, which we call the isospace.
Since the isospin operators must satisfy the same commutation relations as the ordinary
spin operators, then the isospin square and one of its projections, for example, the third
one, can be simultaneously measured. Further on we shall have to introduce a wide variety
of operators, similar to the spin operator, namely, a unitary spin operator, a color spin
operator, and a weak isospin operator. For spin-like operators let us use the symbol S with
a superscript defining an operator class (S (is) , S (un) , S (col) , and S W ). As far as operator
eigenvalues are concerned, we use notations, adopted by “Review of Particle Physics.” Thus,
eigenvalues of the isotopic spin S is (the ordinary spin S) are marked by I (J). So, a field
function of particles of a given multiplet must depend not only on the coordinates and the
ordinary spin, but also on the isospin and its projection, that is, it must be a product of an
isotopic function, a coordinate function, and a spin function. As in the case of an ordinary
spin, an isospin dependence is described by multicomponent isotopic wavefunctions. For
example, the nucleon doublet (p, n) is related with an isospinor (I = 1/2), while the pion
triplet is associated to an isovector (I = 1). A correlation of the sign of the projection I3
with the charge sign is arbitrary. A common choice is the sign’s coincidence of the electric
charge and I3 . Thus, for the proton I3 = 1/2, for the neutron I3 = −1/2, for the π + , π 0 , π − -
mesons I3 = +1, 0, −1, respectively. If the electromagnetic interaction is switched off, then
multiplet particles are identical, in other words, the system of isomultiplet particles is
degenerate on the electric charge. In mathematical language this degeneracy is formulated
as an invariance with respect to the three-dimensional rotations group in the isospace. The
degree of coincidence of particle masses of the same multiplet could be viewed as validity
criteria for isotopic invariance. For the pions:
mπ ± − mπ 0
= δπ ≈ 3%, (4.63)
m˜π

that is, much more than the corresponding value for nucleons. In the case of the proton and
neutron, the mass difference is caused not only by the Coulomb forces, but also by magnetic
ones (magnetic moments of nucleons are not equal to zero), that is, certain compensation
takes place. The mass difference inside of the pion triplet is caused only by the Coulomb
forces since both the pion spins and the magnetic moments are equal to zero. Obviously, the
quantity δi (i = N, π, ...) represents the limit of a precision that should be waited under the
fulfillment of predictions based on the isotopic invariance within one or an other multiplet.
As we shall demonstrate later, the isotopic invariance (or charge independence) leads to
the isotopic spin conservation law. However, before we analyze the consequences of this
invariance, we must broaden our outlook a little in the group theory.

© 2011 by Taylor and Francis Group, LLC

 50 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Let us address the simplest one-parametric unitary group U (1) again. In the n-dimensional
vector space of complex functions ψk (x) one may define the set of linear homogeneous trans-
formations:
ψk′ (x) = U ψk (x), ψk′† (x) = (U ψk (x))† = ψk† (x)U † , (4.64)
which leave a scalar product invariant and satisfy the unitary demand:

U U † = U † U = I. (4.65)

The set of all the square n × n-matrices U with the above-mentioned properties constitutes
a unitary group U (n). To impose on unitary matrices U the unimodular condition:

det U = 1, (4.66)

will single out from U (n) its subgroup, namely, a special unitary group SU (n). In this case
O
U (n) = U (1) SU (n), (4.67)

that is, the transformations U ∈ U (n) differ from those belonging to SU (n) by means of a
phase factor of the form exp (iα) ∈ U (1). The theory of the SU (n)-groups is in many ways
similar to that of the SO(n)-groups. The SU (n)-group goes over into the SO(n)-group
if one passes from a complex n-dimensional space into a real one. The SU (n)-group is a
finite-parametric continuous Lie group. The relations (4.65) and (4.66) defining its elements
impose n2 + 1 conditions on 2n2 real parameters. Therefore, the number of independent
parameters of the SU (n)-group is n2 − 1. The simplest example of a special unitary group
is a group SU (2). Recall that any unitary unimodular n × n-matrix can be written down in
an exponential form:
U = exp (iH),
where due to demands (4.65) and (4.66) a n × n-matrix H satisfies the conditions:

H † = H, Sp {H} = 0. (4.68)

It is an easy matter to check that at n = 2 a special unitary group is formed by the already
familiar matrices:
U (θ) = exp [i(σ · θ)/2], (4.69)
which realize the spinor representation of the three-dimensional rotations group. In prin-
ciple, we could have guessed this from the very beginning. Since the generators of the
three-dimensional rotations group and generators of any corresponding representation sat-
isfy the same relations, then the SO(3)- and SU (2)-groups are locally isomorphic to each
other. Thus, we came to the conclusion, that to solve a problem of constructing irreducible
representations of the SU (2)-group and its Lie algebra is equivalent to the solution of a
similar problem for the SO(3)-group. It is therefore concluded that both the generators Tα
(α = 1, 2, 3) of an arbitrary representation for the SU (2)-group and the generators Kα of
the SO(3)-group representation are governed by the same permutation relations, that is,

[Tα , Tβ ] = iεαβγ Tγ , (α, β, γ = 1, 2, 3). (4.70)

Now we are quite ready to study the consequences of the isotopic invariance. Assume
that the Lagrangian is invariant with respect to the SU (2)-group. Ordinary coordinates xµ
are not changed under the isotopic transformations Xnλ = 0, and the law of the infinitesimal
transformation of the wavefunction has the form:
j
ψk′ (x) = ψk (x) + δψk (x), δψk (x) = (Tγ )k ψj δθγ , (4.71)

© 2011 by Taylor and Francis Group, LLC

 Lagrangian formulation of field theory 51

where for the sake of simplicity we turned from δω [αβ] to element δθγ , which represents an
infinitesimal angle of rotation about an axis γ in the isotopic space defined by:

δθγ = εαβγ δω [αβ] .

Since (Tγ )jk = (Kγ )jk , then the Noether current corresponding to the isotopic invariance is
completely equivalent to the three-dimensional part of the spin moment tensor:

∂L(x) j
Nγλ (x) = − (Tγ )i ψj (x). (4.72)
∂ψi;λ (x)

Spatial integrals Z
Sγ(is) = Nγ0 dr, (4.73)

determining an isospin pseudovector, are constant in time. The total isotopic spin of a
particles system is given by a vector sum of the isotopic spin vectors of all the particles (in
full analogy with the rules for composition of moments). The isotopic invariance means the
(is)
conservation of S (is)2 and S3 . Worthwhile to note, when the electromagnetic interaction is
switched on, the theory loses invariance with respect to arbitrary rotations in the isospace.
In this case the theory remains invariant only under rotations around the x3 -axis. As a
(is)
consequence, S (is)2 ceases to be motion integral and only S3 is conserved, that is, the
electric charge remains the motion integral.

© 2011 by Taylor and Francis Group, LLC

 5
Discrete symmetry operations

5.1 Spatial inversion

Rotations are the examples of continuous transformations, since a rotation by any finite
angle can be viewed as an infinite sequence of infinitesimal rotations. There exist also
discrete transformations, which cannot be presented as a series of infinitesimal steps. The
space inversion r → −r is an example of such a transformation. The corresponding group
is finite and contains only two elements: the unit element I and an element P , satisfying
a condition P 2 = 1 (to use the space inversion transformation twice is equivalent to the
identical transformation).
Laws of classical physics are invariant under the space inversion. It means, that a mirror
image of any process, which obeys laws of classical physics and takes place in nature, is
subject to the same laws and its existence is physically possible. In other words, symmetry
with respect to the space inversion means, that two physicists, one of which is using a
left-coordinate system (a lefthander) while the other is using a right-coordinate system
(a righthander), describing the corresponding law of the nature, will arrive at the same
expression. In classical physics, however, the existence of mirror symmetry does not produce
any conserved quantity.
In quantum theory strong, electromagnetic, and gravitation interactions are conserving
an invariance with respect to the mirror reflection, while a weak interaction does not possess
such a property. Distracting for a while from weak interaction, we show that the mirror
symmetry results in a conserved quantum number, a so-called space parity. First of all
we demonstrate, that when the theory is invariant with respect to the operation: r → −r
(Hamiltonian commutes with the operator P ), then it is always possible to choose stationary
states of a physical system in such a way, that they are orthogonal eigenstates of the
operator P (without degeneration this result is obvious). For the sake of simplicity we
restrict ourselves to a field obeying to superposition principle. In the most general case a
wave field of an arbitrary spin is described by a system of linear equations. These equations
can be always represented in the form of a matrix equation of the first order:
 
∂ mc
iΛµ − ψ(x) = 0, (5.1)
∂xµ ~

where the algebra of square matrices Λµ is defined by a choice of a representation according

to which the multicomponent wavefunction ψ(x) is transformed. Such a conclusion follows
from the fact that derivatives of any order

∂k
ψj (x)
(∂xµ )k

can be always reduced to derivatives of the first order, if one sequentially introduces new

53
© 2011 by Taylor and Francis Group, LLC
 54 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

functions viewing them as additional components of the initial wavefunctions:

′ ∂ ′′ ∂ ′
ψj;µ = ψj (x), ψj;µ,ν = ψ (x), ...
∂xµ ∂xν j;µ

In its turn, any equation of the first order can be always represented in the Schrödinger-like
form:
∂ψ(x)
i~ = Hψ(x). (5.2)
∂t
Applying the P -operator to Eq. (5.2) and taking into account

[H, P ] = 0, (5.3)

one obtains:
∂ψ ′ (x)
i~= Hψ ′ (x), (5.4)
∂t
where ψ ′ (x) = P ψ(x). Consequently, combinations:
1
ψ± (x) = (ψ(x) ± P ψ(x)) ,
2
are also solutions of the Schrödinger equation. Furthermore, these combinations are or-
thogonal eigenstates of the P -operator with or without degeneration in the Hamiltonian
spectrum
P ψ± (x) = ±ψ± (x).
Under r → −r the momentum reverses its sign, in other words, the relation:

{P, p} = 0

takes place. Therefore, the space inversion operator commutes with the moment of momen-
tum operator L:
[P, L] = 0, (5.5)
that is, the eigenvalues of these operators are simultaneously and therewith exactly mea-
surable.
Let us define the eigenvalues of the space inversion operator (space parity) for a particle
with the given moment of momentum L and zero-spin. Recall, that an eigenfunction of the
|m|
operator L2 is the spherical function Yl (θ, ϕ), having the form
s
(2l + 1)(l− | m |)! |m|
m
Yl (θ, ϕ) = sin|m| θPl (cos θ)eimϕ , (5.6)
4π(l+ | m |)!

|m|
where Pl (cos θ) are the generalized Legendre polynomials

|m| 1 d|m|+l 2
Pl (x) = (x − 1)l .
2l l! dx|m|+l
Allowing for that under the space inversion the Decart coordinates:

x = r sin θ cos ϕ, y = r sin θ sin ϕ, z = r cos θ,

reverse their signs we find without trouble:

P P P
r → r, θ → π − θ, ϕ → ϕ + π.

© 2011 by Taylor and Francis Group, LLC

 Discrete symmetry operations 55

Then:
P
Ylm (θ, ϕ) → (−1)l Ylm (θ, ϕ)
that is, the moment of momentum also defines the particle state parity. When a particle
spin equals 0, then its space parity is still determined as (−1)l (l = j +s, j +s−1, ... | j −s |,),
where j is the total particle moment.
The theory invariance with respect to the space inversion demands invariability of quan-
tities measured by experiment, for example:
Z Z
ψ (r, t)pψ(r, t)dr = ψ ′† (r, t)p′ ψ ′ (r, t)dr =
†

Z Z
=− ψ † (r, t)P † pP ψ(r, t)dr = ψ † (r, t)pP † P ψ(r, t)dr. (5.7)

From Eq. (5.7) follows, that the parity operator must be unitary
P † P = P P † = I.
Up to now we spoke of the parity as a quantum number, characterizing a wavefunction
of a particle in the coordinate space, that is, we considered only that part of the parity,
which does not depend on particles type. In so doing we did not touch upon a subject,
whether quantum system components possess any inner parity. However, if a number of
particles can be changed, every particle must be ascribed an inner parity. In classical
theory it is impossible, because there exists no notion concerning production and destruction
of particles. For this reason the inner parity has no sense in the classical theory and,
consequently, the parity conservation law has no classical analog, unlike the conservation
laws of energy, momentum, and angular momentum.
Let us denote a wavefunction of a quantum system, consisting of particles a, b... by the
expression ψab... (ra , ra , ...). To act the parity operator is defined by the relation:
P ψab... (ra , xa , ...) = ǫa ǫb ...ψab... (−ra , −ra , ...),
where ǫa , ǫb ,.. are inner parities, chosen in such a way, so that P commutes with H. Further
on we shall speak of the resultant space parity of a particle, which is equal to the product of
its inner and orbital (space) parities. System invariance with respect to the space inversion
means, that if a system is in a state with a definite parity value, then only transitions
to states with the same parity value are possible. It is obvious that the parity operator
eigenvalues are multiplicative quantum numbers. Recall, that the majority of quantum
numbers is not multiplicative. For example, the electric and baryon charges are additive
quantum numbers. The question arises, how does one define the inner parity of object,
whose inner structure is unknown and which may not possess any structure in general. One
does not always manage to define inner parities of elementary particles in an unambiguous
way. It forces us to adhere to a certain agreement, that is, to speak about relative inner
parity. Let us explain what we mean by an example. Assume that we want to define the
inner parity of the particle c by using a reaction:
a + b → a + b + c, (5.8)
which takes place with the parity conservation. Then we have:
ǫa ǫb (−1)li = ǫa ǫb ǫc (−1)lf , (5.9)
where li (lf ) are the moment of momentum eigenvalues in the initial (final) state. From
Eq. (5.9) follows:
ǫc = (−1)lf +li .

© 2011 by Taylor and Francis Group, LLC

 56 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In this example the weak point in the definition procedure of the inner parity of the particle
c is very obvious: the inner parity can be unambiguously defined only for particles whose
conserved quantum numbers (electric charge, baryon charge, strangeness, charm, etc.) equal
to zero. Otherwise, we can speak only of relative parity. Thus, for example, if the particle
c has the baryon charge, then only the following reaction is possible:

a+b→a+b+c+d (5.10)

where d is baryon, and the parity conservation law defines only the relative parity:
ǫc
= (−1)lf +li .
ǫd
Now we can define inner parities of all the baryons with respect to parities of some main
baryons. Paying respect to the Sakata model [12] (it is assumed that all hadrons consist of
p, n, and Λ) we take as main baryons p, n, Λ and let their parities be positive. The inner
parity of a particle defines a kind of its wavefunction, namely, whether wavefunction is a
true quantity (scalar, vector, etc.) or a pseudoquantity (pseudoscalar, pseudovector, etc.).

5.2 Time inversion

In the previous subsection we got acquainted with the space inversion transformation that
had an obvious physical meaning. Really, it is logical to compare two physical systems
with each other if they follow from each other by the mirror reflection operation (r ↔ −r).
The operation of the time inversion t → −t has no so evident meaning in our world,
where time always flows in one direction. Nevertheless, the time inversion is a very useful
transformation and has a certain physical meaning, if one interprets it correctly. A simple
task from the classical mechanics can serve as an example. Suppose, we have a system A
and its Hamiltonian explicitly does not depend on time. The dynamics of A is defined by
the Hamiltonian equations:
∂ ∂
q̇(t) = H(q(t), p(t)), ṗ(t) = − H(q(t), p(t)), (5.11)
∂p(t) ∂q(t)

and specification of initial conditions for the generalized coordinates q(t) and momenta p(t):

q(ti ) = q (i) , p(ti ) = p(i) . (5.12)

Solving (5.11) and using (5.12), we can determine the values q(t) and p(t) at any moment
of time tf > ti (tf is the final moment of time)

q(tf ) = q (f ) , p(tf ) = p(f ) . (5.13)

A trajectory on phase plane Ati →tf corresponds to transition of the system A from the
initial to the final state. Let us introduce new functions:

q ′ (t) = q(tf − t), p′ (t) = −p(tf − t),

which satisfy the initial conditions:

q ′ (0) = q(tf ) = q (f ) , p′ (0) = −p(tf ) = −p(f )

© 2011 by Taylor and Francis Group, LLC

 Discrete symmetry operations 57

and the motion equations:

∂ ∂
q̇ ′ (t) = −q̇(tf − t) = − H(q(tf − t), p(tf − t)) = ′ H(q ′ (t), −p′ (t)), (5.14)
∂p(tf − t) ∂p (t)

∂ ∂
ṗ′ (t) = ṗ(tf − t) = − H(q(tf − t), p(tf − t)) = − H(q ′ (t), −p′ (t)). (5.15)
∂q(tf − t) ∂q ′ (t)
At t = tf the functions q ′ (t) and p′ (t) meet the following demands:

q ′ (tf ) = q(ti ) = q (i) , p′ (tf ) = −p(ti ) = −p(i) , (5.16)

where we set ti = 0. From the obtained relations follows, that the functions q ′ (t) and
p′ (t) characterize another mechanical system AT , which passes from the initial state with
t = tf to the final state with t = ti , and the evolution of the system is described by
the trajectory ATtf →ti . Since under the time inversion the components of the generalized
momentum reverse the sign p(t) = (d/dt)mq → p′ (t) = −p(t), then boundary conditions of
the system AT are inverted in time with respect to boundary conditions of the system A.
If one does not take into consideration irreversible processes (friction, for example), then
we have
H(q(t), −p(t)) = H(q(t), p(t)), (5.17)
that is, both systems are described by the same equations. All this means, that the phase
trajectory of the inverted system ATtf →ti follows from Ati →tf by the replacement p(t) →
−p(t) and the system AT moves back in time. In other words, the Hamiltonian equations
allow two solutions, which are connected with each other by the time inversion operation.
It is also obvious, that the time inversion should be understood from the viewpoint of the
motion conversion.
If one records some physical process with a video camera and plays it back, then one can
observe a process, which is also possible. In some cases such a reversed-in-time process may
appear to be of extremely low-probability, however, none of the classical physics laws would
be violated during this process. Thus, for example, a reversed-in-time dive from a tower
would be put into practice, if water molecules, moving suitably, gave back the diver, who
had fallen into a pool, his initial energy and momentum, and, as a result, he could go up
to the tower again. However, our life experience tells us that the probability of such events
is extremely small. The principle of the entropy increase applied to macroscopic systems
establishes that time flows in the direction of realization of the most probable processes,
however, no law of classical physics prohibits the realization of any low-probability process.
Since the principle of the entropy increase is only applicable to macrosystems (with large
numbers of particles), but not to microworld events (processes, in which individual particles
are participated), then we cannot use it to define time direction under the investigation of
the microworld phenomena.
Now let us discuss the subject of the time inversion in quantum theory. All physical
quantities are divided in two classes with respect to the time inversion. The first class
consists of quantities that contain time in even powers and, consequently, are not changed
under the time inversion (coordinates, total energy, kinetic energy, etc.). Quantities of the
second class contain time in odd powers and reverse the sign at t → −t (momentum, total
moment, spin, etc.). The transformation group of the time inversion is finite and contains
two elements: the time inversion operator T and the identical transformation. The time
reflection operation, applied to the system twice, returns it back to the initial state.
Let us discover the main properties of a wavefunction of a particle with arbitrary spin
with respect to the time inversion. As we already know, for this purpose it is sufficient to

© 2011 by Taylor and Francis Group, LLC

 58 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

consider the Schrödinger equation:

∂
i~ ψ(r, t) = Hψ(r, t). (5.18)
∂t
The time inversion operation T yields the state ψ(r, −t), in which all physical quantities of
the first class have the same values as in state ψ(r, t), and physical quantities of the second
class have the opposite sign. Even if the Hamiltonian is invariant under the time inversion,
T
H −→ H,

the corresponding equation in the system with inverse time, nevertheless, has the form:
∂
−i~ ψ(r, −t) = Hψ(r, −t), (5.19)
∂t
that is, the motion equations for the functions ψ(r, t) and ψ(r, −t) are not identical.
Let us consider the equation complex conjugated with respect to Eq. (5.19)

∂ ∗
i~ ψ (x, −t) = H∗ ψ ∗ (x, −t). (5.20)
∂t
Since H is the Hermitian operator, then H and H† have the same eigenvalues (but, broadly
speaking, different eigenfunctions). It means that there exists such an unitary operator V ,
that:
V H∗ V −1 = H.
With regard to this relation, Eq. (5.20) takes the form:

∂
i~ V ψ ∗ (x, −t) = HV ψ ∗ (x, −t), (5.21)
∂t
that is, V ψ ∗ (x, −t) is a solution inverted in time. Thus, an operator T ′ , which transforms
a wavefunction under the time inversion:

ψ(x, −t) = T ′ ψ(x, t), (5.22)

is an antiunitary operator.
Let us recall the main properties of an antiunitary operator B:

B(c1 ψ1 (x) + c2 ψ2 (x)) = c∗1 Bψ1 (x) + c∗2 Bψ2 (x), (5.23)
Z Z
(Bψ1 (x)) (ψ2 (x))dx = ψ2 (x)† ψ1 (x)dx.
†
(5.24)

Since any antiunitary operator can be presented as a product of an unitary operator and
complex conjugation operator K, then Eq. (5.22) acquires the form:

ψ(r, −t) = T ′ Kψ(r, t) = T ψ ∗ (r, t), (5.25)

where an explicit form of unitary operator T depends on representation wherein a wave-

function is specified. To replace a wavefunction by a complex conjugated one under the
time inversion means, that the physical state of a system cannot be the eigenstate of the
operator T . Due to this reason there is no corresponding quantum number. That is the
main difference between the space and time inversion. A more detailed description of the
time inversion transformation can be found in the book by Wigner [13].

© 2011 by Taylor and Francis Group, LLC

 Discrete symmetry operations 59

Until 1964 it had been considered that in the microscopic scale the time direction was
not measurable, that is, all laws of microworld were invariant under the time inversion. Ob-
servation of a decay KL0 → π + π − became a first indirect demonstration of the T -invariance
violation [14] (we supply the long-living K 0 -meson with the symbol L while its short-living
fellow is denoted by KS0 ). The reaction:

KL0 → π 0 π 0 ,

which also confirms the existence of this effect, was registered later.

5.3 Charge conjugation

Besides the space and time inversions we can define the charge conjugation operation. This
operation leaves invariable space-time particle characteristics (coordinates, momenta, spins,
etc.), but for all that, changes every particle for antiparticle

Cψ(x) = ηc ψ c (x), (5.26)

where ψ(x) (ψ c (x)) is a wavefunction of a particle (antiparticle), ηc is a phase factor, for

which | ηc |2 is true, and C is a linear unitary transformation with properties:

C = C † = C −1 . (5.27)

We emphasize, that a particle differs from an antiparticle not only in the sign of the electric
charge (Q) and consequently, in the sign of the magnetic moment, but also in signs of
other quantum numbers, such as the baryon charge (B), lepton number (L), strangeness
(s), charm c, and so on. Therefore, the charge conjugation operator anticommutes with
operators of corresponding “charges”

CQ = −QC, CB = −BC, CL = −LC, ... (5.28)

Eqs. (5.28) are in conflict with the conditions of the simultaneous and precise eigenvalues
measurement of the operators C, Q, B, S, ..:

[C, Q] = [C, B] = [C, L] = [C, S] = .... = 0. (5.29)

Consequently, a charge parity (eigenvalue of the operator C) can be defined only for true
neutral particles and systems, for example: π 0 , γ, e− e+ , µ− µ+ ... The charge parity is
a multiplicative quantum number and it can take values ±1. If an initial system has a
definite C-parity, then from the invariance under the charge conjugation it follows that
a final system must possess the same parity. Experimental data show that strong and
electromagnetic interactions are symmetric with respect to the C-operation, while a weak
interaction is not symmetric.
By way of exercise let us define the charge parities of some particles. We start with the
photon. Since electric and magnetic fields are caused by static and moving charges, then:
C C
E −→ −E, B −→ −B. (5.30)

From Eq. (5.30) it is evident, that:

C
Aµ −→ −Aµ , (5.31)

© 2011 by Taylor and Francis Group, LLC

 60 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

that is, the charge parity of the photon is ηγ = −1. Recalling, that the π 0 -meson decay:

π 0 → 2γ

is motivated by an electromagnetic interaction, which conserves the C-parity, we obtain

ηπ0 = 1.

5.4 G-parity
So, we have learned, that eigenstates of the charge conjugation operator can be only states
with the total electric charge being equal to zero. Taking into account the usefulness of
selection rules with respect to the C-parity for neutral systems, we should like to have a
conserved quantum number, somehow connected with the C-conjugation for charged states
too. It appears to be possible for strong interactions because they conserve isotopic spin.
We shall demonstrate this on the example of the isotopic triplet of the π-mesons.
Let us represent the triplet wavefunction in the form of a column matrix:
 
ψπ1 (x)
 ψπ2 (x)  ,
ψπ3 (x)

To define the neutral and charged pion fields we use the expressions:

1
ψπ± (x) = √ (ψπ1 (x) ± iψπ2 (x)) , ψπ0 (x) = ψπ3 (x). (5.32)
2

Fulfillment of the following relations is obvious:

Cψπ0 (x) = ψπ0 (x), Cψπ± (x) = ψπ∓ (x). (5.33)

Consequently,
   
ψπ1 (x) ψπ1 (x)
C  ψπ2 (x)  =  −ψπ2 (x)  . (5.34)
ψπ3 (x) ψπ3 (x)
Although the states ψπ± are not eigenstates of the charge conjugation operator in them-
selves, and yet, it is possible to build linear combinations from them, which will be eigen-
states of the operator C. However, such states do not possess a definite value of the electric
charge, and this situation does not suit us at all. Earlier we were dealing with the isotopic
spin operators Ti (i = 1, 2, 3) that were the rotation generators in the isospace. Looking at
Eq. (5.34) we can clearly see the way of obtaining the eigenfunctions of such an operators
combination that must contain the operator C. We should make a rotation in the isospace
in such a way, that the first and third components of the wavefunction of the π-meson
triplet reverse the signs. It can be reached by means of the operator of the rotation in the
isotopic space about the second axis by the angle 1800 , that is, by operator exp (iπT2 ):
   
ψπ1 (x) −ψπ1 (x)
exp (iπT2 )  ψπ2 (x)  =  ψπ2 (x)  . (5.35)
ψπ3 (x) −ψπ3 (x)

© 2011 by Taylor and Francis Group, LLC

 Discrete symmetry operations 61

So, the required combination of the operators is G = C exp (iπT2 ). From the relations:
       
ψπ1 (x) −ψπ1 (x) ψπ+ (x) ψπ+ (x)
G  ψπ2 (x)  =  −ψπ2 (x)  , G  ψπ− (x)  = −  ψπ− (x)  , (5.36)
ψπ3 (x) −ψπ3 (x) ψπ0 (x) ψπ0 (x)
it follows that all the π-mesons have the G-operator eigenvalues (G-parity) equal to −1.
Therefore, the G-operator eigenvalues serve as a convenient generalization of the charge
parity concept in the case with “charged” (having nonzero quantum “charges”) particles.
The operation of the G-conjugation for an arbitrary state is defined as the charge conju-
gation followed by the rotation by the angle π about the second axis in the isotopic space.
From this it follows that the Hamiltonian, being invariant under the charge conjugation and
rotations in the isotopic space, is also invariant under the G-conjugation. Thus, G is the
“good” quantum number only if the isospin is conserved, that is, the G-parity conservation
law works only in the case of a strong interaction.
Since the charge parity is multiplicative and isotopic spin is additive, then the G-parity is
a multiplicative quantum number. Thus, for example, a state, containing n of the π-mesons,
has the G-parity equal to (−1)n . For this reason transitions between odd and even numbers
of the π-mesons under the action of strong interaction are forbidden. From the existence of
the decays:
ρ0 → π + + π − , ω → π+ + π− + π0 ,
(here we are dealing with the ρ0 - and ω-meson with the mass 770 and 782 MeV/c2 ) it
follows immediately, that the G-parity of the ρ0 -meson equals +1, and the G-parity of the
ω-meson equals -1.

5.5 CP T theorem
CP T theorem concerns properties of the physical systems with respect to the product of
the transformations C, P , and T , whose result is:
r → −r, t → −t, particles → antiparticles.
(sometimes the CP T transformation is called a strong reflection of the space-time). This
theorem had been proven before 1956. At that time there were no doubts about physical
theories invariance with respect to every discrete operation taken separately. First it was
established [15], that theories, invariant under the proper Lorentz transformations and the
space inversions were also invariant under the time inversion or the charge conjugation.
Then G. Luders managed to show [16], that the P invariant relativistic quantum theory
of an arbitrary field was automatically invariant under the combined operation CT . In
1957 Pauli gave the general proof of the CP T theorem [17]. Another name of this theorem
is the Pauli-Luders theorem. In the Lagrangian formulation of quantized field theory the
contents of this theorem is as follows.“The theory is invariant under the combined operation
CP T if and only if the quantum theory of the given field meets the following demands: 1)
locality; 2) invariance of the Lagrangian with respect to the proper Lorentz transformations;
3) hermicity of the Lagrangian; and 4) standard connection between the spin and statistics
(particles with integer spin are subject to Bose-Einstein statistics, particles with half-integer
spin are subject to Fermi-Dirac statistics).” We underline, that the combined operator CP T
is not proportional to the identity operator, that is, speech does not go about the identical
transformation.

© 2011 by Taylor and Francis Group, LLC

 62 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Violation of the P - and C-invariance, discovered in 1956, as well as the CP -asymmetry,

discovered in 1964 hardly caused any changes in theoretical apparatus of quantized field
theory, which could naturally involve these discoveries without violating any fundamental
principles of theory. The probable experimental detection of the CP T -invariance violation
will have tragic consequences for modern quantum field theory and will result in the radical
revision of its main principles.
One consequence of the CP T -invariance is the fact that if the interaction is invariant
with respect to one of the transformations C, P , or T it must be also invariant with respect
to the product of two remaining operators. Analogously, if the interaction is not invariant
with respect to one of the three transformations, it must not be invariant with respect to
the product of two other ones. Really, already the first experiments on detection of the
space parity violation in a weak interaction also demonstrated the violation of the charge
symmetry. Thus, for example, at β − -decay:

X → Y + e− + ν e , (5.37)

where X (Y ) is the mother (daughter) nucleus, the electrons have predominantly left-handed
polarization, while in the charge conjugated decay the positrons have predominantly right-
handed polarization.
For several years after the discovery of the charge and mirror symmetry there was a hope,
that at least the CP -symmetry and, on the strength of the CP T -theorem, T -reversibility
would remain the firm laws of the nature. However, as we already mentioned before, in
1964 the decay of the long-living neutral KL0 -meson through the channel:

KL0 → π + + π − , (5.38)

was discovered. It proved the absence of the T -symmetry and, consequently, the absence
of the CP -invariance of the world we live in. Further detailed studies of (5.38) and other
KL0 -meson decays proved a violation of the CP - and T -invariance, but did not detect any
hints on the ruin of the CP T -symmetry.
Unlike the C- and P -asymmetries, all the measured CP asymmetric effects are restricted
only by the KL0 -meson decay through the channels:

KL0 → π + + l− + ν l , KL0 → π − + l+ + νl ,

KL0 → π + π − , KL0 → π 0 + π 0 ,
where l = e− , µ− , τ − . As a measure of the CP -violation one may use the amplitudes
ratio for the decays of the KS0 - and KS0 -meson (recall, that the KS0 decay goes with the
CP -conservation):
A(KL0 → π + π − )
ǫ= ≈ 2.3 × 10−3 . (5.39)
A(KS0 → π + π − )
The CP asymmetric effects are so little in other kaon decays as well. However, we have
all reasons to expect that the effects of the CP -asymmetry turn out, to put it mildly,
more significant during studies of the B 0 - and B 0 -meson decays. Thus, the standard model
predicts, that these effects can reach 0.9.
To find out the mechanism of the CP -violation it would be very important to measure
the value of the neutron static (intrinsic) electric dipole moment (EDM) (not to be confused
with a dynamic EDM, which arises on a particle in a strong electric field). Both the neutral
and charged particles can possess the static EDM. The appearance of the static EDM of
the neutron may be caused by the fact that the positive charge Q is distributed over the
surface of one part of the neutron hemisphere, while the negative charge, equal to −Q, is

© 2011 by Taylor and Francis Group, LLC

 Discrete symmetry operations 63

spread over the other part of the neutron hemisphere. Then, by measuring a deformation,
caused by such a charge distribution, we would actually measure the intrinsic EDM. Due
to its electric neutrality, stability and a large size (compared to leptons) the neutron is a
very convenient object to measure this moment. An interaction energy of a dipole with an
external electric field has the form:

Hint = (d0 · E) ∝ (S · E),

where d0 is the intrinsic EDM value of the particle, S is the particle spin operator and E
is the strength of the external electric field. Under the time inversion S reverses its sign,
while E does not, that is, this interaction is not CP -invariant. Thus, the CP -invariance
implies, that the intrinsic EDM of the neutron and other particles must be equal to zero.
The latest experiments with ultracold neutrons gave the following result:

| dn |< 0.97e × 10−25 cm

One of the significant consequences of the CP T -invariance is that particles and antipar-
ticles must have absolutely identical masses, life-times, and their magnetic dipole moments
must be identical in value, but opposite in sign. The CP T -theorem also predicts, that
particles and antiparticles interact in the same way with a gravitation field. Up to now
there is no experiment confirming violation of the CP T -invariance. The mass equality of
particles and antiparticles has been most precisely measured for the K 0 - and K 0 -mesons:
|mK 0 − mK 0 |
< 6 × 10−19 .
mK 0
The best accuracy in measuring the life-time (τ ) for particles and antiparticles has been
reached for muons:
Γµ+ →all − Γµ− →all
< 3 × 10−5
Γµ+ →all
(recall, that τ = ~/Γ). The anomalous magnetic moments equality for the electron and
positron has been checked with an accuracy of ∼ 10−12 , while for µ− and µ+ —with an
accuracy of ∼ 10−7 .
In conclusion we address once more the invariance with respect to the charge conjugation.
The C-invariance means that physical laws and equations, corresponding to them, are not
changed if all particles are replaced by antiparticles. Another way of putting it is that
the laws in the “world” and in the “antiworld” are equivalent. At this point the most
experienced reader could make the remark: “Since all the matter in our galaxy consists of
protons, neutrons, and electrons rather than of antiprotons, antineutrons, and positrons,
our world is not C-invariant.” Once again we stress, the C-invariance means, that particles
and antiparticles interact in an identical way. From the C-invariance it does not in any way
follow that in some region of the universe the equal amounts of particles and antiparticles
are present. Really, in our local galactic cluster the amount of antimatter is overwhelmingly
small: the antimatter fraction constitutes < 10−4 . To extrapolate this fact over the whole
universe bears a name “baryon asymmetry of the Universe” (BAU). The quantity:
nB − nB
δ= , (5.40)
nγ

where nB , nB , nγ are the density of baryons, antibaryons, and relic photons, is the quanti-
tative measure of the BAU. According to the current data [18] its value is given by:

δ ≈ 6 × 10−10 .

© 2011 by Taylor and Francis Group, LLC

 64 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Explanation of the BAU origin and the value of δ is one of the key problems of modern
cosmology and elementary particle physics. Needless to say, one may accept the point
of view that the BAU is obliged by the role of an accident at the early stages of world
evolution. Then, the value δ should be considered as an initial condition. There is nothing
to contradict such an explanation, however, it is not solely possible.
It is more attractive to assume that at its birth moment the universe was still symmetric
nB = nB while the asymmetry in the observed part of the universe appeared at some
evolutionary stage. In this case two versions are possible: 1) the conservation law of the
baryon number B is exact; 2) there are interactions leading to the nonconservation of B.
In the first case for the BAU to appear either separation of matter and antimatter in
macroscopic scales is needed (which is considered to be feasible with difficulty) or burial
of antibaryons into black holes that could separate matter and antimatter when the CP -
violation takes place. The second version is realized in the so-called great unification theories
(GUTs). In such theories there exist superheavy particles, X-bosons, which are interaction
carriers. Their decays are caused by the interaction to violate both the CP -invariance and
the baryon number conservation law. For such particles the decays into the u-quarks
r r
X → u + u, X → u + u, (5.41)

and into the d-quark and a lepton

1−r 1−r
X → d + l, X → d + l, (5.42)

prove to be allowed. In the case of the CP -violation we have r 6= r. By this reason

an excess of quarks over antiquarks (or baryons over antibaryons) appears under r ≥ r.
However, that is possible only in the case of thermodynamic nonequilibrium, because in the
equilibrium case to increase baryons production velocity leads to the growth of antibaryons
production velocity. Thus, the demand, so that the X-decays are nonequilibrium, represents
an obligatory condition. Otherwise, the CP T -symmetry conservation would ensure a system
neutrality over all conserved charges, that is, in the thermodynamic equilibrium B = 0.
Under expansion and cooling, according to the Big Bang theory, the universe passes the
phase when the CP nonconserving interactions go out of equilibrium. Obviously, during a
period that is equilibrium over interactions with the B nonconservation any initial value of
B will be wiped out. Under exiting from this period the universe gains B 6= 0 at the sacrifice
of the above-mentioned microprocesses. The GUT, based on the SU (5)-gauge group and
supplemented by the Big Bang theory, leads to the value of δ in the interval from ∼ 10−6
to ∼ 10−12 .

© 2011 by Taylor and Francis Group, LLC

 Problems

1.1. Examine four successive infinitesimal rotations:

(a) the rotation about the angle δϕ1 around the 1th axis;
(b) the rotation about the angle δϕ2 around the 2th axis;
(c) the rotation about the angle −δϕ1 around the 1th axis;
(d) the rotation about the angle −δϕ2 around the 2th axis.
Reveal that the sequence of these rotations is equivalent to the rotation of the second order
about the angle δϕ1 ×δϕ2 . Using this circumstance prove that the corresponding generators
satisfy the relation
[J1 , J − 2] = iJ3 .
1.2. Show that the product of two Levi-Chivita symbols is defined by the relation
 
δii′ δij ′ δik′
ǫijk ǫi′ j ′ k′ =  δji′ δjj ′ δjk′  .
δki′ δkj ′ δkk′

1.3. Making use of the induction method prove the following expression for theconvolution
of the Levi-Chivita symbols on several indices
 
δal+1 bl+1 δal+1 bl+2 ... δal+1 bn
 ... ... ... ... 
ǫa1 ...al al+1 ...an ǫa1 ...al bl+1 ...bn = l!  .
... ... ... ...
δan bl+1 δan bl+2 ... δan bn
1.4. Let a G-group representation be irreducible. Show that in this case any linear operator
U commuting with all the G-group representation operators is multiple to the unit one (Shur
lemma). To put it differently, using the relations

T (g)U = U T (g), g∈G

prove the equality

U = λI,
where λ is a constant. Take into account the fact that in a complex space any linear oper-
ator has one eigenvalue at least.
1.5. Find all possible states of the composite particle made of the particles with the spin
S = 1/2 and S = 3/2.
1.6. Consider the particle to be described by the mixed spinor of the third rank ψγαβ . Find
the irreducible representations corresponding to it.
1.7. Find the representation laws for the components of the symmetric Sµν and antisym-
metric Aµν four-tensors under the homogeneous Lorentz group transformations.
1.8. Write down the homogeneous Lorentz group transformation formulae for the four-
momentum pµ when the relative velocity of motion of the reference frames under investiga-
tion has an arbitrary orientation.
1.9. Two particles having the masses m1 = m2 = m and momenta p1 , p2 come into
collision. Determine the value of the Lorentz invariant s = (p1 + p2 )µ (p1 + p2 )µ in the

65
© 2011 by Taylor and Francis Group, LLC
 66 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

center-of-mass frame (p1 + p2 = 0). Find as well the first particle energy in the system in
which the second particle is at rest (laboratory system).
1.10. A particle a collides with a rest particle b:

a+ b → ...

The particle masses (ma , mb ) and the total energy of the a-particle (Ea ) are given. Find:
(a) the laboratory frame reference
p velocity βc with regard to the center-of-mass system;
(b) the Lorentz factor γc = 1/ 1 − βc2 ;
(c) the total energy of the particles in the center-of-mass system.
Assuming the colliding particle masses to be equal, obtain the formula to relate the a-
particle Lorentz factor with the Lorentz factor of the center-of-mass frame

γa = 2γc2 − 1.

How does this formula transform for relativistic energies?

1.11. For second-order matrices, the matrices σµ and the unit matrix constitute a basis.
Let us confront every real four-vector xµ with the Hermitian matrix Y according to the
following rule:  0 
x + x3 x1 − ix2
Y = xµ σµ = .
x1 + ix2 x0 − x3
Introduce the transformation
Y ′ = U Y U †, (I.1)
where U is unimodular matrix (detU =1), and find the relation between the matrix U and
the homogeneous Lorentz group transformation matrix.
1.12. Find the explicit form of the transformation matrices U (see Eq. (P.1) ) in two-
dimensional space when:
(a) the U -matrix is Hermitian;
(b) the U -matrix is unitary.
Using the formula obtained in the previous problem
1
Λµν (U ) = Sp(σ µ U σν U † ),
2
determine the explicit form of the corresponding Lorentz transformation for the four-vector
xµ .
1.13. The Poincare group representation is characterized by the total set of the mutually
commuting operators. The Casimir operators pµ pµ = m2 and w = −Wµ W µ (Wµ is the
Pauli-Lubanski vector) always enter to this set. Along with them the momentum operator
pµ and the third spin projection operator S3 to be defined by Wµ are also included in the
total set. In the rest frame let the spin vector be equal to

WR
SR = ,
m
and in an arbitrary reference frame represent a linear combination of the Wµ vector com-
ponents with the coefficients depending on pν only.
Prove that the single axial-vector linear combination of the Wµ -operators, which turn to
the vector SR in the rest frame, is given by the expression
   
1 W0 pj 1 (p · M)
Sj = Wj − = p0 Mj − [p × N ]j − pj , (I.2)
m m + p0 m m + p0

© 2011 by Taylor and Francis Group, LLC

 Problems 67

where M0j = Nj , M12 = M3 , ... and Mµν is the proper Lorentz group representation
generator (the sum of the orbital and spin field moments). For proof one should take
advantage of the following commutation relations

[Mj , Sk ] = iǫjkl Sl (I.3)

which are fulfilled owing to the fact that S is a three-vector.

1.14. Show that the relation (I.2) is nothing else than the spatial components of the Pauli-
Lubanski vector to be transformed in the rest frame reference


Sj = Λ−1s W j,

where Λs is the limited homogeneous Lorentz transformation, which works as follows

(Λ−1 ν
s )µν p = (m, 0, 0, 0).

1.15. There is a scalar field with a self-action λΦ4 . The theory Lagrangian is defined by
the expression
1 λ
L = [(∂µ Φ)(∂ µ Φ) − m2 Φ2 ] − Φ4 .
2 4!
Find the motion equations, the energy-momentum tensor and the angular tensor.
1.16. The equations system

∂ µ ψµνλ + m2 ψνλ + ∂ν ψλ − ∂λ ψν = 0, 
ψµνλ = ∂µ ψνλ − ∂ν ψµλ + ∂λ ψµν , (I.4)
ψν = ∂ µ ψµν .


Find the Lagrangian to result in the equations.

1.17. Using the Lagrangian obtained in the previous problem write down the expressions
for the energy-momentum tensor and the electric current.
1.18. The free Lagrangian of the charged scalar field is given by
1
L= [(∂µ Φ)∗ (∂ µ Φ) − m2 Φ∗ Φ].
2
Having demanded the theory invariance with respect to the local gauge transformations
build up the total Lagrangian describing the electromagnetic interaction of the charged
scalar particles.
1.19. The Pauli principle for identical particles reads:“a fermion system wavefunction is
antisymmetric under rearrangement of any two particles while a boson system wavefunction
is symmetric relative to the same rearrangement.” Making use of this principle find:
(a) the isotopic spin of the ρ-meson to decay through the channel

ρ → π+ + π−

(Sρ = 1);
(b) the isotopic spin of the deuteron regarding it as the 3 S1 state of the (np) system.
1.20. Show that the recharge reaction of the slow π mesons under collision with the
deuterium
π− + d → n + n + π0
is forbidden when the orbital moment of the π 0 meson regarding the (nn) system and the
orbital moment of the π − meson regarding the deuterium are equal to zero. Under proof
set the deuterium parity to be equal to +1.

© 2011 by Taylor and Francis Group, LLC

 68 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

1.21. Determine the positronium charge parity. Point out the more probable decay channels
of the parapositronium (1 S0 ) and the orthopositronium (3 S1 ).
1.22. Using the G parity conservation law in the strong interaction establish whether the
following reactions
p + p → π+ + π− ,
p + p → π0 + π0
are possible.

© 2011 by Taylor and Francis Group, LLC

 References

[1] S. Tomonaga, Progr. Theor. Phys. 1, 27 (1946).

[2] J. Schwinger, Phys. Rev. 75, 651 (1949).
[3] H. Umezawa, Quantum field theory, (North-Holland PC, Amsterdam, 1956).
[4] F. Rohrlich, Phys. Rev. 80, 666 (1950); P.T. Matthews, Phys. Rev. 76, 1657
(1949).
[5] E. Stuckelberg, Helv. Phys. Acta, 11, 225 (1938).
[6] O. M. Boyarkin, JETP, 48, 13 (1978).
[7] A. Proca, J. phys. et rad. 7, 347 (1936).
[8] O. M. Boyarkin, J. Phys. G: Nucl. Phys. 8, 161 (1982).
[9] H. P. Stapp, Phys. Rev. 125, 2139 (1962).
[10] E. P. Wigner, G. C. Wick, A. S. Wightman, Phys. Rev. 88, 101 (1952).
[11] E. Noether, Nachr. Ges. Wiss. Gottingen, 171 (1918).
[12] S. Sakata, Progr. Theor. Phys. 16, 686 (1956).
[13] E. P. Wigner, Group Theory, (New York, 1959).
[14] J. H. Christenson et al., Phys. Rev. Lett. 13, 138 (1964).
[15] L. C. Biedenharn, M. E. Rose, Phys. Rev. 83, 459 (1951); H. A. Tolhoek, de
Groot, Phys. Rev. 84, 151 (1951).
[16] G. Luders, Zeit. Phys. 133, 325 (1952).
[17] W. Pauli, in Niels Bohr and the Development of Physics, (Pergamon Press,
London, 1957).
[18] D. N. Spergel et al., Astrophys. J. Suppl. 170, 377 (2007).

69
© 2011 by Taylor and Francis Group, LLC
 Part II

Birds’s-Eye View on the

Microworld

71
© 2011 by Taylor and Francis Group, LLC
 6
Fundamental interactions

Once, however, there was an eagle who became terribly

bored with living all alone. The eagle began to brood.
And the more he brooded, the more he began to think
how wonderful it would be to live as the serf-owners
of old lived. Why not gather a company of varlets and
live like a lord? Let the crows bring him the latest
gossip, the parrots turn somersaults, magpies cook
porridge, horned owls, screech-owls and barn owls keep
watch at night, and hawks, falcons and vultures supply
him with food. And he, the eagle, would devote himself
to bloody tyranny.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

6.1 Species of interactions

The main purpose of physics is to explain all natural phenomena by means of a few simple
fundamental principles. Since all matter consists of particles, then elementary particles
physics must give the final answer. By this, one of the main subjects is the question about
the nature of interactions. It appears that despite of the diversity of the world surrounding
us, all interactions are reduced to four fundamental types, which differ greatly in intensity
of their proceeding. Such a classification is not free from some conventionality due to the
fact that the relative role of different interactions is changed while the energy of interacting
particles grows. It means that the separation of interactions into classes, based on the
comparison of processes intensity, can be made reliably only for not very high energies. Let
us consider the types of fundamental interactions in order of decreasing their intensity.
Strong interaction. This is an overwhelming type of interaction in nuclear physics of
high energies. Particles, which participate in strong, weak, and gravitation interactions, are
called hadrons. Hadrons with half-integer spin are called baryons, while hadrons with integer
spin are called mesons. In addition, the charged hadrons participate in electromagnetic
interaction. Just the strong interaction causes couplings between protons and neutrons
in the atomic nuclei and ensures an exclusive strength of these formations lying at the
heart of matter stability in Earth conditions. The strong interaction is also responsible for
the confinement of quarks inside hadrons. This interaction may be manifested not only
as an ordinary attraction in the nucleus, but as a force, which causes instability of some
elementary particles (particles to be decaying by means of the strong interaction are called
resonances). Due to its great value, the strong interaction is a source of huge energy. In
particular, the main part of heat in the Sun is produced by the strong interaction, when
deuterium nuclei, made by the weak interaction, together with protons are synthesized into
helium nuclei. The strong interaction is a short-range interaction with a radius of 10−13

73
© 2011 by Taylor and Francis Group, LLC
 74 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

cm.
Only after the quark structure of hadrons had been discovered, the strong interaction
theory ceased to resemble the plots of Russian popular tales. This theory, called quan-
tum chromodynamics (QCD), resembles quantum electrodynamics (QED) in construction.
However, the QED has a local gauge symmetry with respect to the group U (1)em , while a
local gauge symmetry of the QCD is associated with the SU (2)c -group. Note, that both
symmetries are internal ones, that is, they are connected with a system symmetry not in the
ordinary space-time but in abstract spaces. The lower index c in the QCD symmetry group
is caused by the fact, that quarks, beside ordinary quantum numbers, have three additional
degrees of freedom, for which we use the conventional term “color” or color charge R (red),
G (green), and B (blue). For both the QCD and QED internal symmetries are exact.
The subject of internal symmetry violation is of great importance in quantum field theory.
There are two mechanisms of symmetry violations, namely, explicit and spontaneous. Under
explicit violation the Lagrangian contains terms, which are not invariant with respect to
a symmetry group. The value of these terms characterizes the degree of corresponding
symmetry violation. Thus, for example, the Lagrangian of the strong interaction is variant
under isotopic transformations, but the total Lagrangian also contains the electromagnetic
and weak interactions, which explicitly violate the isotopic symmetry. For this reason the
complete theory does not possess the exact isotopic invariance.
Under the spontaneous symmetry violation the Lagrangian possesses the invariance with
respect to the transformations of the internal symmetry group; vacuum (vacuum is the
state with the minimum energy), however, loses this invariance. Vacuum noninvariance
reveals itself through the fact, that one or more components of a quantized field (as a
rule these components correspond to scalar particles) acquire the nonzero vacuum averages
< 0|ϕi |0 > (or vacuum expectation values), which define various energy scales of the theory.
Under the spontaneously violated local symmetry the corresponding gauge bosons, which
are interaction carriers, prove to be massive particles, while under the exact symmetry these
gauge bosons are massless. Thus, carriers of the strong interaction between quarks, which
we call gluons, are massless particles.
In all observable hadrons the color charges of the quarks are compensated, that is, hadrons
are colorless (white) formations. Hadron colorlessness can result either from mixing three
main colors (true for baryons) or from the mixture of color and anticolor (true for mesons).
In the strong interaction the color charges of the quarks play the same role as the electric
charges of particles do in the electromagnetic interaction. The color charge is a source of
the gluon field. As this takes place the gluon carries on itself both the color and anticolor
charges, that is, its color composition is a product of color and anticolor. When the quark
emits the gluon its color changes, depending on a gluon color. For instance, the red quark,
emitting the red-antiblue gluon, turns blue. Analogously, the blue quark, absorbing the
red-antiblue gluon, turns red, etc. A total of 3 × 3 = 9 “color-anticolor” combinations are
possible. Among them there is one corresponding to the colorless state:

g0 = RR + BB + GG.

Since under its emitting or absorbing the quark state is not changed, then this combination
cannot play the role of the gluon bearing interaction between quarks. Thus, only eight
gluons are left. The gluons are electrically neutral, have zero-mass and spin equal to 1.
All this makes them similar to the photons. But unlike the photons, the gluons have a
“charge” of the field whose interaction they bear. The gluon can emit or absorb other
gluons, changing its own color in so doing. That is, the gluons create the new gluon field
around themselves. The photons are deprived of such a property, they have no electric
charge and no new electric field is created around them. The electromagnetic field is the

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 75

most intensive near the charge, which causes the field and further away it is permanently
dispersed in space and weakened. Color-charged gluons produce around themselves new
gluons, which produce new gluons and so on. The result is, that the gluon field is not
decreasing, but increasing further away from the quark creating this field. In other words,
effective color charges of the quarks and gluons are increasing as the distance grows. At
the distances of the hadron size order (∼ 10−13 cm) the color interaction becomes really
strong. The perturbation theory, the main mathematical apparatus in microworld physics,
is not applicable in this domain, so there are no reliable calculations. However, one could
expect on qualitative grounds, that strengthening the interaction with distance must result
in the fact that separating of isolated quarks at large distances becomes impossible. To
put this another way, it brings to life imprisonment of the quarks in hadron prison. This
phenomenon is called confinement.
The baryon consists of three differently colored quarks. The quarks are constantly ex-
changing the gluons and changing their own color. These changes, however, are not arbi-
trary. Mathematical apparatus of the QCD restricts severely this play of colors. At any
moment of time the summarized color of three quarks must represent the sum R + G + B.
Mesons consist of quark-antiquark pairs, every pair is colorless. Then, no matter what
gluons the quark-antiquark pairs are exchanging the mesons also remain white formations.
So, from the QCD standpoint, the strong interaction is nothing else but the tendency to
maintain the SU (3)c -symmetry, resulting in conservation of the white color of hadrons,
while their components change their colors.
The strong interaction intensity is characterized by the so-called QCD running coupling
constant:
g2
αs (q) = s ,
4π
where gs is a gauge constant of the SU (3)c -group. The term “running” reflects dependence
of αs on a distance or on a transferred momentum q. We are reminded, that to estimate a
quantity order in the microworld one may use the Heisenberg uncertainty relation. There-
fore, the transition to short (large) distances means transition to large (small) values of a
transferred momentum q ∼ ~/r.
Evolution of the running coupling constant of the QCD is governed by the equation:
12π
αs (q) = , (6.1)
(33 − 2nf ) ln(q 2 /Λ2QCD )

where nf is a number of quark kinds (at given color) or a number of quark flavors, and
ΛQCD is a scale parameter of the QCD. Derivation of this equation is based on the use of
the perturbation theory and the structure of the total Lagrangian describing the theory. At
q 2 ≫ Λ2QCD the effective constant αs (q) is small and, consequently, the perturbation theory
describes the behavior of weakly interacting quarks and gluons successfully. At q 2 ∼ Λ2QCD
it is impossible to use the perturbation theory while strongly interacting gluons and quarks
start to form coupled systems, hadrons. Obviously, the parameter ΛQCD defines the border
between the world of quasifree quarks and gluons and the world of real hadrons. The
value of ΛQCD is not predicted by the theory. It is a free parameter that is determined
from experiments. Nowadays, despite the joint efforts of experimentalists and theorists, the
exact value of ΛQCD remains unknown (its approximate value lays between 100 and 200
MeV). The formula (6.1) leads to the decrease of an effective interaction as a momentum
grows and in the asymptotic ultraviolet limit the effective interaction tends to zero. Then
the fields, participating in interaction, become free. The phenomenon of self-switching off
interaction at short distances that is a reverse side of confinement, is called asymptotic
freedom.

© 2011 by Taylor and Francis Group, LLC

 76 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Since the hadrons are neutral with respect to the color, then at distances, bigger than
the hadron size, there is no strong interaction between the hadrons at all. It is similar
to the absence of electromagnetic forces between atoms at big distances, since they are
electrically neutral. However, when two or more atoms approach at a distance, when their
electron clouds are overlapping, the so-called Van der Waals forces, or chemical forces come
into action. Their radius of action is of the atom size order. Molecular bond is caused
by these forces. Its mechanism is based on the exchange of electrons between atoms, that
is, the molecular bond is a complicated manifestation of the fundamental electromagnetic
interaction between two spatially-distributed charged systems. Analogously, the hadron
interaction also can be viewed as a complicated manifestation of the fundamental strong
interaction between the color quarks, which becomes observable only under approaching
the quark cores of the hadrons.
Electromagnetic interaction. The electromagnetic interaction keeps electrons within
atoms and binds atoms in molecules and crystals. This interaction lies at the basis of nearly
all the phenomena around us—chemical, physical, and biological. Elementary particles with
the electric charge take part in the electromagnetic interaction. Neutral particles could
also interact with the electromagnetic field due to multipole moments (dipole, quadrupole,
anapole, etc.). However, only particles with a composite structure can have such moments.
Thus, the electromagnetic interaction is not as universal as the gravitation one. It is easy to
observe the electric and magnetic forces, acting between macroscopic bodies. These forces,
just as the gravitation forces, are subjected to the inverse square law, that is, they are
long-range forces. For instance the Earth magnetic field extends far into cosmic space, and
the Sun magnetic field fills all the solar system.
Unification of the electric and magnetic forces into the classical theory of the electromag-
netic field, made by Maxwell in the 1850s, is an example of the first unified field theory.
The idea of unification can be illustrated by the example of the Lorentz force:
v
F = q{E + [ × H]}. (6.2)
c
Let us assume, that |E| ∼ |H|. Then from Eq. (6.2) it follows, that at |v| ≪ c the magnetic
forces are very small compared to the electric ones and approach in their value to them only
at |v| → c. Thus, the relative intensity of the forces is defined by the particle velocity, that
is, there is the scale on which unification of the electric and magnetic fields takes place and
this scale is determined by the light velocity. Since the energy is also growing at |v| → c, we
can state, that unification occurs in the region of the ultrarelativistic energies of particles.
The quantum theory of the electromagnetic interaction of the electrons and positrons,
called the QED, had been built up at the beginning of the 1950s. The QED is the most
exact of all physical theories. Here the electromagnetic interaction is exhibited in its pure
form. Unprecedented accuracy of calculations in the QED is caused by usage of apparatus
of the perturbation theory on the small dimensionless parameter:
e2
αem (q) = ,
4π~c
which is called a fine structure constant. αem (q) is also a function of a distance or a
transferred momentum. Its macroscopic value, which is defined at q = me c, equals to
1/137.0359895(61). The exact symmetry of the QED with respect to the local gauge group
U (1)em , whose gauge constant is equal to the electron charge, leads to the fact that the
electromagnetic interaction carrier, the photon, has zero mass.
The QED perfectly describes not only the electrons, but electromagnetic properties of
other charged leptons as well. Contrary to this, electromagnetic properties of the hadrons
are not amenable to calculations because the hadrons are basically controlled by the strong

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 77

interaction. We stress, that the QED is not only the first model of the quantum field theory
(QFT), but it is also the simplest and the most extensively studied version of the QFT.
Within the framework of the QED many fundamental concepts of the QFT were discovered
and formulated. All this allows to build up more complicated quantum field theories in the
image and similarity of the QED.
In the QED the phenomenon of vacuum polarization results in the screening of the elec-
tron charge by vacuum positrons. When polarizing the vacuum the electron attracts virtual
positrons and repulses virtual electrons. As a result, the electron charge is partly screened,
if one sees it from a large distance. If one penetrates deep inside of a cloud of virtual pairs,
then screening would decrease and the effective electron charge would increase. In other
words, contrary to the QCD running constant that is increasing with the distance, the QED
running constant is decreasing as the distance grows. The calculations, made within the
QED scope, define the evolution of αem (q) by means of the equation:

3παem (me c)
αem (q) = . (6.3)
3π − αem (me c) ln [q 2 /(4m2e c2 )]

From Eq. (6.3) it follows that at q ∼ 80 GeV/c the value of αem is approximately equal to
∼ 1/128.
Weak interaction. The weak interaction is destructive in its character because it is
not able to create stable states of matter in the way, as, for example, the gravitation force
maintains the existence of the solar system or the electromagnetic interaction ensures atom
stability. In other words, the main destination of the weak interaction is to regulate a life-
time of inanimate matter. It is responsible for nuclear β ± -decays, for decays of particles not
belonging to the resonance class (we call such particles “stable”). Only a few stable parti-
cles, for example, π 0 -meson, η 0 -meson, and Σ0 -hyperon decay due to the electromagnetic
interaction.
If electromagnetic multipole moments of a neutrino are equal to zero, then all the pro-
cesses with the neutrino participation are caused by the weak interaction only. The weak
interaction is also responsible for nuclear and atomic processes going with parity viola-
tion. Particles, participating in the weak interaction and with the availability of the electric
charge in the electromagnetic interaction but not participating in the strong ones, are called
leptons.
In some cases the weak interaction also influences macroscopic objects. For example, it
plays a key role in the Sun energy release, because deuterium nucleus production from two
protons is caused by just this interaction:

p + p → 2 D + e+ + νe .

Neutrino emission in weak interaction processes defines stars evolution, especially at their
final stages, initiates supernova explosions and pulsar production. If it were possible to
switch off the weak interaction, then the matter around us would acquire quite another
structure. It would contain all particles to decay due to the weak interaction (muons,
π ± -mesons, K-mesons, etc.).
Intensity of the weak interaction is defined by a Fermi constant

GF = 1.16639(1) × 10−5 (~3 c3 ) GeV−2 ,

which is dimensional as we see. In the reference frame, where a particle rests the probability
of the decay Γ due the weak interaction turns out to be proportional to G2F m5 (m is a mass
of a decaying particle). In virtue of the Heisenberg uncertainty relation, an elementary
particle lifetime τ is inversely proportional to Γ. For particles, decaying due to the weak

© 2011 by Taylor and Francis Group, LLC

 78 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

interaction, the value of τ is quite large by microworld scales and lies in the interval 103 –
10−10 s. The lifetime of a particle decreases as the intensity of interaction, causing decay,
grows. For particles, which instability is caused by the electromagnetic interaction, τ is of
the order of ∼ 10−16 s, while for particles decaying because of the strong interaction, τ is
of the order of 10−23 –10−24 s.
Notwithstanding the fact that the first process, caused by the weak interaction, the ra-
dioactive β − -decay of the nucleus, had been discovered by A. Becquerel in 1896, attempts
of constructing the weak interaction theory was crowned with success only in the 1960s. For
the construction of this theory Glashow, Salam, and Weinberg were awarded the Nobel prize
in 1979. In this theory both the electromagnetic and weak interactions are the manifesta-
tions of one and the same interaction, which is called an electroweak
N(EW) interaction. The
local symmetry with respect to transformations of the SU (2)EW U (1)EW -gauge group,
possessing the two gauge coupling constants g and g ′ , makes the base of the theory. In this
case, there are two peculiarities, which make the EW interaction different from both the
QED and QCD.
First, the local gauge symmetry of the EW interaction is spontaneously violated up to
the local gauge symmetry of the QED:
O
SU (2)EW U (1)EW → U (1)em .

Second, from the very beginning the theory is not invariant with respect to the operation
of the space inversion. N
The nonviolated local symmetry SU (2)EW U (1)EW demands the existence of four
massless particles with spin 1, two of which are neutral and the remaining two are charged.
It was known from experiments, that the action radius of the weak interaction RW is ex-
tremely small ∼ 10−16 cm. Consequently, carriers of this interaction must have masses of
the order of ∼ ~/(RW c). To give the mass to the gauge bosons of the weak interaction, a
doublet of massless scalar fields, Higgs bosons, consisting of neutral and charged compo-
nents is introduced into the theory. In so doing, the neutral Higgs boson is not properly
neutral. Due to the spontaneous symmetry violation (the nonzero vacuum average from a
neutral component
N of the Higgs doublet is chosen) three of the gauge bosons of the group
SU (2)EW U (1)EW acquire masses, while the forth one remains massless. The massive
gauge bosons, W ± and Z, are identified with the gauge bosons of the weak interaction and
the massless gauge boson, γ, is identified with the photon. Out of four massless scalar fields,
one neutral field acquires mass and the remaining three leave physical sector, as if they were
eaten by the gauge bosons while they are gaining their masses. From the massless vector
field with two spin states and the massless scalar field the massive vector particle with three
spin projections is produced, so, the number of degrees of freedom is conserved. The idea
of generating the mass through interaction with the filled vacuum was stated for the first
time by J. Schwinger [1]. In this work nucleons were the investigation subject. Schwinger’s
idea was transferred by P. Higgs on vector particles [2]. The mass production of the gauge
field due to the spontaneous violation of the local symmetry is called a Higgs mechanism.
The neutral currents discovery (1973), that is, observing the processes going through the
virtual Z-boson exchange, should be considered as the first confirmation of the Glashow-
Weinberg-Salam (GWS) theory. The detection of the W ± - and Z-bosons at the proton-
antiproton collider in CERN (1983) was the genuine triumph of the theory. Up-to-date
values of the gauge boson masses are:

mW = 80.41 ± 0.10 GeV/c2 , mZ = 91.187 ± 0.007 GeV/c2 .

The predicted coincidence of the generations number of quarks and leptons (quark-lepton
symmetry) also should be attributed to the success of the GWS theory.

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 79

We emphasize that, although the GWS theory unifies the weak and electromagnetic
interactions, the gauge constants of the SU (2)EW - and U (1)EW -groups, g and g ′ , are not
connected with each other. These constants satisfy the relations:

tan θW = g ′ /g, g = e sin θW , (6.4)

where there are no theoretical predictions for the angle θW (Weinberg angle) and it is
defined from experiments only.
By now lots of data have been accumulated, which prove, that experiments fit the the-
ory perfectly. However, the main problem in the EW interaction
N theory is not solved yet,
namely, the mechanism of violating the original SU (2)EW U (1)EW -symmetry is not es-
tablished. The most direct way to solve this problem is experimental searching for the Higgs
boson. Since the theory does not predict its mass mH , then the range of researching for it
is rather wide.
The fact, that in the world surrounding us, we discriminate the electromagnetic and weak
interactions, only means that their unification scale or the boundary of the spontaneous
symmetry violation in the EW interaction theory lies on the higher energy scale ∼ mW c2 =
80.4 GeV, that corresponds to the distances of the order of 10−16 cm.
To compare different interactions it is convenient to use dimensionless quantities. For
this purpose we introduce a quantity α2 , which characterizes the intensity of the weak
interaction according to the relations:

g2 g2 GF
α2 (q) = , 2 = √ . (6.5)
4π 8mW 2(~c)3
Gravitation interaction. Although the gravitation interaction is the weakest of all,
it possesses a cumulative effect. So, the gravitation interaction between two bodies is a
cumulative sum of interactions between elementary masses that form these bodies. Since
in the microworld the contribution of the gravitation interaction is very small compared to
other interactions, it does not result in measurable effects on the subatomic level. However,
on the macroscopic level the gravitation interaction is dominating: it keeps together parts
of the terrestrial globe, unifies the Sun and planets into the solar system, connects stars in
galaxies and controls the evolution of the whole universe.
Since the gravitation interaction was discovered in the first place, then just with its help
the term “force” appeared in physics. The gravitation interaction is universal, because it
is in operation between all bodies having the mass. It belongs to long-range interactions.
Building the nonrelativistic gravitation theory was completed by I. Newton in 1687. Ac-
cording to this theory two mass points, having the masses m1 , m2 and lying at a distance
r, are attracted with the force, whose value and direction are given by the expression:
m1 m2 r
F = GN ,
r3
where GN is a Newton constant, GN = 6.67259(85) × 10−8 cm3 g−1 s−2 . In this theory the
force depends on the particles position at a given time only and so the gravitation interaction
propagates instantly.
At an arbitrary mass distribution the gravitational force, operating on any point mass
m0 in the given spatial point, can be expressed as a product of m0 on a vector Eg which
is called the gravitation field strength. In the Newton theory the superposition principle is
valid for the gravitation field. As this field is potential it is possible to introduce, by the
usual fashion, the gravitation potential

Eg = −grad ϕg .

© 2011 by Taylor and Francis Group, LLC

 80 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The potential of a continuous distribution of a matter density ρ(r) satisfies the Poisson
equation
∆ϕg = −4πGN ρ(r). (6.6)

The Newton gravitation theory has allowed us to describe with great precision an exten-
sive range of phenomena, including the motion of natural and artificial bodies in the solar
system, the motion of celestial bodies in other systems: in binary stars, in stellar clusters,
and in galaxies. On the basis of this theory the existence of the planet Neptune and the
satellite of Sirius has been predicted. In modern astronomy the gravitational law of Newton
is the foundation on the basis of which the motions, the structure of celestial bodies, their
masses and evolution are calculated. The precise definition of the Earth’s gravitational field
allows us to define a mass distribution under Earth’s surface and, hence, to directly solve
the important applied problems.
As the Newton theory assumes the instantaneous propagation of the gravitation it cannot
be made consistent with the special relativity theory, stating that the interaction propaga-
tion velocity cannot exceed c. It means, that this theory is inoperable when gravitational
fields are so strong, that they accelerate bodies, moving in them, up to the velocities of the
order of c. The velocity v, up to which the body falling freely from infinity (v |t=0 ≈ 0 ) as
far as some point with a gravitational potential ϕg (r) has been accelerated, can be found
from the relation
mv 2
= mϕg (r)
2
(we have put ϕ(∞) = 0). Hence, the Newton theory of gravitation is applicable only in the
case when
|ϕg | ≪ c2 .
For the gravitational fields of usual celestial bodies this requirement is fulfilled. For example,
on the Sun surface we have |ϕg |/c2 ≈ 4 × 10−6 , and on the surface of the white dwarfs |ϕg |
is about 10−3 .
Besides the Newton theory is inapplicable under calculation of the particles motion even
in a weak gravitational field with |ϕg | ≪ c2 if particles flying near massive bodies, had the
velocity v ∼ c already far from them. Hence, it will lead to the improper answer under
calculation of the light trajectory in a gravitational field. The Newton theory also is not
used under investigation of the varying gravitational fields created by moving bodies (for
example, binary stars) at distances r > cτ , where τ is a period of revolution in a system
of a binary star. Really, the Newton theory, based on the instantaneous propagation of the
interaction, is unable to take into account the retardation effect that appears to be essential
in this case.
The relativistic gravitation theory, that is, the general relativity theory (GRT) was built
by Einstein in 1915. It changed drastically understanding gravitation in the classical, New-
tonian, physics. In the Einstein theory gravitation is not a force, but a manifestation of the
curvature of the space-time. The flat metric of Minkowski gµν = diag(1, −1, −1, −1) in the
space of the GRT is deformed into a metric

ηµν (x) = gµν + hµν (x),

where hµν (x) describes a massless spin-2 field whose source is a matter. The two postulates
make the foundation of the GRT. The first one defines the form of the Lagrangian density
Lg , describing a propagation and a self-action of the gravitation field. On the basis of the
second postulate, namely, an equivalence principle, the gravitation interaction is introduced
by means of substitution gµν → ηµν (x) into the Lagrangians of all the existing fields, that

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 81

is, into LQCD + LEW . Variation of the total Lagrangian Lg + LQCD + LEW with respect
to the gravitation potentials ηµν (x) leads to the Einstein gravitation equations

1 8πGN
Rµν (η) − R(η)ηµν (x) = Tµν (η), (6.7)
2 c4
where Rµν (η) is a Ricci tensor
β
Rµν (η) = ∂α Γα α α β α
µν − ∂ν Γµα + Γαβ Γµν − Γνα Γµβ ,

Γλµν are Christoffel symbols, which play the role of the gravitation field strength

1 λσ
Γλµν = η (∂µ ηνσ + ∂ν ηµσ − ∂σ ηµν ) (6.8)
2
R(η) = Rµν (η)η µν and Tµν is a symmetric energy-momentum tensor of matter.
Outwardly the Einstein equations (6.7) are similar to Eq. (6.6) for the Newtonian poten-
tial. In both cases the quantities characterizing the field stand in the left-hand side and the
quantities characterizing matter that creates this field do in the right-hand side. However,
between these equations there is a number of essential differences. Eq. (6.6) is linear and,
consequently, satisfies the superposition principle. It allows computing a gravitational po-
tential for any distribution of masses moving arbitrarily. The Newton gravitation field does
not depend on the masses motion, therefore Eq. (6.6) in itself does not define immediately
their motion. To describe the mass motion we must invoke the second Newton law. In the
Einstein theory a pattern is quite different. As Eq. (6.7) are nonlinear, the superposition
principle does not work any more. Further, in this theory it is impossible to set arbitrarily
a right-hand side of Eq. (6.7) (that is, Tµν ), which depends on a matter motion, and then
to calculate the gravitational field ηµν . The solution of the Einstein equations leads both to
the definition of the motion of matter generating the field and to the evaluation of the field
itself. In so doing it is essential, that the gravitation field equations also contain the mass
motion equations in the gravitation field. From the physical point of view it is equivalent to
the fact, that in the Einstein theory the matter creates the space-time curvature which, in
its turn, influences the motion of matter originating this curvature. In the GRT all particles
move along extremal lines, called geodesic curves. In the flat space-time geodesic curves
degenerate into straight lines. Notice, that in the ordinary field theory operating the flat
space-time, the motion equations are also obtained by means of the extremum condition,
however, this condition is imposed on the system action. The stronger is the gravitation
field, the more appreciable is the curvature of the space-time. Thus, the nonrelativistic
gravitation theory is not applicable, when the gravitation fields are very strong, as it oc-
curs near collapsing objects like neutron stars or black holes. On the other hand, in weak
fields one may be restricted by the calculation of small corrections to the Newton equations.
The effects corresponding to these corrections, allow testing the GRT experimentally in the
ordinary gravitational fields as well.
For today the experimental status of the basic statements of the Einstein theory is as
follows. To check the principle of equivalence of the gravitational and inert masses is carried
out with a precision of 10−12 . The theoretical formula for changing the light frequency (red
shift) in the gravitational field that also is a consequence of the equivalence principle, is
verified with a precision of 2 × 10−4 . The constancy with time of GN being postulated
by the theory was tested by radar observations of the motions both of planets (Mercury,
Venus) and of spaceships, by measuring the Moon motion with the help of the laser, by
observations of the motion of a neutron star, namely, pulsar PSR 1913+16, which forms
part of a double star-shaped system. All the collection of the experimental data confirms

© 2011 by Taylor and Francis Group, LLC

 82 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the constancy of GN with a precision of

 
 1 dGN  −11
 GN dt  < 10
  years−1 .

The GRT predicts bending the light ray when it is passing near the heavy mass. The
analogous bending follows from the Newton theory as well, however, in the Einstein theory
this effect is twice more. Numerous observations of this effect being done under passage
of the light coming from the stars near the Sun (during the complete solar eclipse) have
confirmed the GRT predictions with a precision up to ∼ 11%. The much more accuracy
(∼ 0.3%) has been already reached under the observation of the extraterrestrial point
radiation sources.
The Einstein theory also predicts the slow rotation of the elliptic orbits of the planets
spinning around the Sun. It should be emphasized that this rotation is not explained by the
gravitational fields of other planets. The effect has the greatest magnitude for the Mercury
orbit—43′′ in a century. At present the verification precision of this prediction (precession
of the Mercury perihelion) reaches 0.5%.
The GRT effects should be rather considerable when the stars are moving in a tight
double system. With the greatest precision the motion of the pulsar PSR 1913+16 forming
a part of the double star is explored. Here the orbit rotation due to the GRT effects attains
4.2% in one year, and for 14 observations years (1975–1989) has given ∼ 600 .
One further the GRT effect is the prediction, that the bodies, moving with variable
acceleration, will radiate gravity waves. Despite numerous attempts one has not managed
to register gravity waves as far. However, there are the serious grounds in support of their
existence now already. For example, the observations of the pulsar PSR 1913+16 have
confirmed an energy loss of the double system due to the radiation of the gravity waves.
As a consequence of the effect, the period of the star circulation should decrease with time.
The observations confirm the GRT prediction with the precision of 1%.
The GRT describes not only the large-scale phenomena in the universe, it also determines
the evolution of the universe itself. And just here the final experimental checkout of the
theory is being awaited. The first relativistic cosmological model, based on Eqs. (6.7), was
built by Einstein in 1917. Proceeding from the considerations conventional for classical
science, Einstein suggested that the universe, as a totality, should be eternal and invariable.
However, Eqs. (6.7) were unable to describe the stationary universe. Because of this,
Einstein has introduced the Λ-term, now known as the cosmological constant, in Eqs. (6.7)
1 8πGN
Rµν (η) − R(η)ηµν (x) = Tµν (η) − Ληµν (x), (6.9)
2 c4
where Λ > 0. As a result, the last term in (6.9) describes the repulsive gravitational forces
complementary to the attractive gravitational forces of the normal matter (Tµν ). Nominally,
the cosmological term is equivalent to the additional term of the energy-momentum tensor.
Recalling the analogy between the Poisson equation for the gravitational potential in the
Newtonian theory and the Einstein equation, the emergence of the similar term in the
Newtonian gravitation theory were equivalent to the introduction of an additional force
acting on the body from an object having a negative mass M0
mM0
F0 = GN r.
r3
From the beginning, the function of the cosmological constant was to create or, what is
more accurate, to describe antigravitation. Einstein assumed that in this way it is possible
to balance the gravitation of the universe matter and ensure an immovability of matter

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 83

distribution, that is, stationarity of the universe. Such a model cannot answer the ques-
tion:“How and from where had the universe originated?” As regards this, the theory keeps
a delicate silence. Nevertheless, no longer than 15 years later the astrophysical observations
made the scientists give up the model of the stationary universe.
At the beginning of the 1920s A. Friedmann demonstrated that by the appropriate selec-
tion of a metric the equations of the GRT have nonstationary solutions with the cosmological
term present as well. In Friedmann models based on the homogeneous and isotropic uni-
verse matter is considered as a continuous medium, uniformly filling the space and having
specific values of the density ρ and pressure P at every instant of time. To analyze the
motion of such a medium, the co-moving frame of reference is usually used, similar to the
Lagrangian coordinates in the classical hydrodynamics. In this system, the matter is mo-
tionless, deformation of the matter being reflected by that of the reference system, and
hence the problem is reduced to the description of the reference system deformation. The
three-dimensional space of the co-moving frame of reference is referred to as a co-moving
space. For a homogeneous and isotropic space the square of the four-dimensional interval
ds may be represented in the form:

dx2 + dy 2 + dz 2
ds2 = ηµν dxµ dxν = c2 dt2 − a2 (t) , (6.10)
1 + k(x2 + y 2 + z 2 )/4

where x, y, z are dimensionless space coordinates, a(t) is a radius of a space curvature and
k = −1, 0, 1. It should be noted that under selecting the metric we have already assumed
that the universe is nonstationary. The space curvature is positive at k = 1 and negative
at k = −1. Provided k = 0, the space is Eucledian (flat), and a(t) has the meaning of a
scale factor. Variation of a(t) in time describes expansion or compression of the co-moving
reference frame and hence the matter. The metric in (6.10) is known as the Friedmann–
Robertson–Walker metric that forms a basis for modern cosmology. It is convenient to
rewrite the expression (6.10) in the spherical coordinates:
 
2 2 2 2 dr2 2 2 2 2 2
ds = c dt − a (t) + r dθ + r sin θdφ , (6.11)
1 + kr2

that is, the nonzero components of the metric tensor ηµν (µ, ν = t, r, θ, φ) have the form

a2 (t)
ηtt = 1, ηrr = − , ηθθ = −a2 (t)r2 , ηφφ = −a2 (t)r2 sin2 θ. (6.12)
1 + kr2
Using
ηµν η νσ = δµσ ,
we obtain
1 + kr2
η tt = 1, η rr = − , η θθ = −a−2 (t)r−2 ,
a2 (t)

η φφ = −a−2 (t)r−2 sin−2 θ. (6.13)

To solve the problem about the deformation of the reference frame, it remains only to find
the unknown function a(t). The dynamics of the homogeneous and isotropic universe may
be described similarly to a model for ideal liquid with the density ρ(t) and pressure P (t),
averaged over all the galaxies, their clusters, and superclusters. Then, a hydrodynamic
energy-momentum tensor for the matter is given by:

Tµν = P ηµν + (P + ρc2 )Uµ Uν , (6.14)

© 2011 by Taylor and Francis Group, LLC

 84 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where U t = 1, U i = 0. As seen from the calculations, for the metric (6.12) the following
components of the Christoffel symbols are nonzero:
 
aȧ ȧ 1 ∂ηlj ∂ηlk ∂ηjk
Γtij = 2 ηij , Γitj = 3 δji , Γijk = η il + − . (6.15)
a a 2 ∂xk ∂xj ∂xl

Taking into account Eq. (6.15) we obtain the following expression for the components of
the Ricci tensor
3ä 1

Rtt = , Rti = 0, Rij = 2
aä + 2ȧ2 + 2k ηij . (6.16)
a a
The time components of the Einstein’s equation give
 
ä 4πGN 3P Λc2
=− ρ+ 2 + (6.17)
a 3 c 3
while the purely space components of that lead to the relation
 
ä 2ȧ2 2k P
+ 2 + 2 = 4πGN ρ − 2 + Λc2 . (6.18)
a a a c

Omitting ä from Eqs. (6.17), (6.18) we arrive at the first-order differential equation for a(t)

ȧ2 k 8πGN Λc2

+ = ρ + . (6.19)
a2 a2 3 3
Eq. (6.17) describes changes in the expansion speed of the universe under the effect of
gravitation. From this equation it follows that gravitation is due not only to the matter
density but also to its pressure in the combination ρc2 + 3P , referred to as the effective
gravitating energy of the matter ρG mat . It is obvious that in this case the cosmological term
will result in antigravitation since its effective gravitating energy is negative. To find the
function a(t) and determine a cosmological model by this means, it is necessary to know
for some t the values of density ρ(t0 ) = ρ0 as well as the cosmological constant Λ(t0 ) = Λ0 .
Usually, instead of ρ0 one uses the quantity Ω = ρ0 /ρc , where ρc = 3H 2 /(8πGN ) is a
critic matter density in the universe and H is a Hubble constant whose value is defined by
experiments.
In whatever the scale the mass calculations for the universe should be performed, the
deficiency of mass will be always revealed. Dynamically, a behavior of the galaxies them-
selves as well as (super)clusters is so as if they contain much more matter than is really
available in their visible components, referred to as a luminous matter or baryon matter.
The present-day value of the cosmic density (average over the whole observable world) for
the baryon matter is determined by:
ρB
ΩB = = 0.02 ± 0.01. (6.20)
ρc
Apart from the baryon matter in the universe there is a hidden mass (lately referred to
as dark matter). Two special types of dark matter exist: a cold dark matter and a hot dark
matter. The cold dark matter (CDM) is composed of nonrelativistic objects and its present
density is given by:
ρD
ΩD = = 0.3 ± 0.1. (6.21)
ρc
The CDM forms a vast invisible corona, or halo, around the stellar disk of the Milky Way.
Similar dark halos seem to be present in all sufficiently massive isolated galaxies. The CDM

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 85

is also contained in galactic clusters and superclusters. As with our galaxy, it makes up
about 90% and sometimes more of the total mass for all these systems. There is no emission
or absorption of electromagnetic waves by the CDM that manifests itself exclusively through
the created gravitation. Unfortunately, the nature of the CDM has not been conclusively
established up to the present. A wide variety of the possibilities is considered: from weakly
interacting massive elementary particles to massive (exceeding a mass of the Sun) black
holes, etc.
The third component of the cosmological medium is hot dark matter (HDM). The HDM
comprises ultrarelativistic particles with masses equal to zero or of the order of ∼ eV. The
density of this medium is determined by the expression:
ρR
ΩR = = 0.8 × 10−5 α, (6.22)
ρc
where the constant factor 1 < α < 10–30 includes the contribution of neutrinos, gravitons,
and other possible ultrarelativistic particles.
In the final years of the twentieth century one more component of the cosmological
medium, a cosmic vacuum, was discovered as a result of a series of experiments conducted
by two big research collaborations of astronomers. As it turned out, a vacuum (a cosmic
vacuum is often referred to as dark energy) predominates in the universe, with the energy
density making it superior over all the “ordinary” forms of the cosmic matter taken together
ρV
ΩV = = 0.73 ± 0.01. (6.23)
ρc

This means that 67% of the total matter of the world falls on the vacuum, 30%—on the cold
dark matter, about 3%—on the baryon matter, and all the rest is connected with radiation.
The modern observations also indicate that cosmological expansion is proceeding with
acceleration. On the other hand, acceleration may be exclusively due to antigravitation,
whose origin is the cosmological term Λ in the Einstein equations. It is presently universally
acknowledged that the cosmic vacuum is described by the cosmological constant. The
vacuum density is related to the value of this constant by the following relation:
Λ
ρV = .
8πGN
From this point of view, we can relate the evolution of the universe to its genesis: expan-
sion of matter stems from antigravitation of the cosmological vacuum, the matter per se
appearing as a result of quantum fluctuations of the same vacuum.
Nevertheless, despite the impressive successes, the GRT has some unresolved problems.
For example, the quantization of the GRT faces serious difficulties. So, it follows from the
Einstein field equation, that the gravitation field theory does not belong to the class of
renormalizable theories. Let us explain what we mean talking about a renormalization. As
we know, the mathematical apparatus of quantum theory is mainly based on the usage of
the perturbation theory series. In the four-dimensional field theories these series contain
infinitely large quantities, which one must be removed in one way or another. Normally that
is reached by means of the redefinition of finite number of physical parameters, such as the
mass, the charge, etc. This procedure is called renormalization and the theories, in which
it eliminates divergences, are called the renormalizable theories. For non-renormalizable
theories there is no procedure to ensure convergence of the perturbation theory series.
The presence of a dimensional interaction constant makes the ordinary renormalization
procedure impossible. To eliminate divergences in the theory, we must summarize all the
terms in the corresponding series of the perturbation theory. As a result, some divergences

© 2011 by Taylor and Francis Group, LLC

 86 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

are canceled, and the remaining infinities are eliminated by the renormalization of the
physical parameters of the theory. However, if the interaction constant is dimensional,
then the terms in the perturbation theory series have the different dimensions and their
summation has no sense.
In the GTR under expanding the metric tensor ηµν in a power series near the flat space
with the metric gµν , the interaction constant κ appears

ηµν ≈ gµν + 2κhµν , (6.24)

√
where κ is equal to 2GN . Then in the first nonvanishing approximation the gravitation
field equations have the form
hµν = GN Tµν

and, as it is easy to see, in the stationary case they transfer to Newton’s equations of the
gravitation field. However, the interaction constant κ proves to be a dimensional quantity
and, as a result, the ordinary theory of the renormalization does not work. Another difficulty
connecting with the quantization of the GRT has the experimental nature. A particle,
creating the gravitation field, a graviton, has not been yet discovered. The theory predicts
for it zero mass, zero electric charge, and spin being equal to 2.
The Minkowski space considered in the special relativity theory (that is, in absence of
gravitating bodies) has the high level of a symmetry that is described by the Poincare group.
According to the relativity principle this group generates isomorphic sequences of events.
In the space where the gravitation field presents, the symmetry disappears completely.
Therefore, in such a space the relativity principle is not fulfilled. For this reason the title,
the general relativity theory, belonging to Einstein, is inadequate and gradually disappears
from literature, replaced with “the gravity theory.”
Just as for other interactions, it is possible to introduce a nondimensional intensity of the
gravitation interaction αg (q). It is defined as follows:

GN m2
αg (q) = . (6.25)
4π~c

The main difference in the above-enumerated interactions is the strength of their mani-
festation in nature. There are different ways to compare the interaction’s intensity. One of
them is based on values of corresponding energy effects. Thus, for example, the electromag-
netic interaction can be characterized by the binding energy of an electron in the ground
state of a hydrogen atom, Eem ≈ 10 eV, while the energy effect of the strong interaction
can be determined from the binding energy of nucleons in a nucleus, Es ≈ 10 MeV.
Another way is to compare the running coupling constants, which describe different in-
teractions. However, since these quantities are energy functions, we must point out the
energy value, at which the comparison takes place. One should remember, that the running
constants of the groups SU (2)EW and U (1)EW (α2 and α1 = g ′2 /4π) cannot be identified
with the running constants of the weak and electromagnetic interactions. The operation is
legal only at energies much less than the energy, at which the spontaneous violation of the
local symmetry of the electroweak interaction takes place. At the scale 1 GeV the running
coupling constants of the strong, electromagnetic, and weak interactions are connected by
the relation
αs : αem : αW ≈ 1 : 10−2 : 10−6 .

As soon as the gravitation interaction is switched on, a confusing indefiniteness appears.

What elementary particle should be taken as a standard? Now, the mass is the “charge”

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 87

of the gravitation interaction, but the mass spectrum of elementary particles is continuous.
So, for example, the ratio of the Coulomb and gravitation forces has the form

Fc
≈ 1036 (6.26)
Fg

for the protons and

Fc
≈ 1043 (6.27)
Fg
for the electrons. Using Eqs. (6.26) and (6.27) we arrive at two different intensity hierarchies

αs : αem : αW : αG ≈ 1 : 10−2 : 10−6 : 10−38 ,
(6.28)
αs : αem : αW : αG ≈ 1 : 10−2 : 10−6 : 10−45 .

6.2 On the path to a unified field theory

Grand Unification Theory. Extension of the electroweak theory to quarks and inclusion
of the QCD resulted in unification of the strong and electroweak interactions. This created
scheme was called the standard model (SM). The picture of fundamental forces in this model
is charmingly simple. The strong, weak, and electromagnetic interactions are caused by the
existence of the local gauge symmetry group
O O
SU (3)c SU (2)EW U (1)EW

with its three gauge constants gs , g, and g ′ and twelve gauge bosons being the carriers of the
strong and electroweak interactions. At sufficiently small distances all these forces mainly
resemble each other and lead to the potential of the Coulomb type ∼ g 2 /r. Scale of short
distances for the strong interaction represents distances much smaller than the hadron size,
that is, more small than 10−13 cm. For the electroweak interaction the scale of the small
distances is the distance that is much smaller than the Compton wavelength of the W ± -
and Z-bosons (λc = ~/mc), that is, more small than 10−16 cm. It is obvious, that at such
short distances the existence of the masses on the gauge bosons is becoming inessential.
Since the SM gauge group is the production of three unbound sets of gauge transfor-
mations: the groups SU (3)c , SU (2)EW and U (1)EW , then the three gauge constants of
these groups are not connected with each other. The gauge constants will be bound, if the
SM gauge group proves to be embedded in a more wide group of gauge transformations G.
Symbolically, it is written as follows:
O O
G ∋ SU (3)c SU (2)EW U (1)EW .

As a result, all the interactions will be described by the unified gauge theory, the Grand
Unification theory (GUT), with one gauge constant gGU and at the same time all the other
gauge constants are connected with gGU in an unambiguous way, defined by the choice of the
group G. The GUT symmetry must be spontaneously violated at supershort distances being
many orders smaller than those, at which unification of the electromagnetic and weak inter-
actions takes place. In other words, the strong interaction with the
N local SU (3)c -symmetry,
as well as the electroweak interaction with the local SU (2)EW U (1)EW -symmetry turn
out to be the low energy fragments of the gauge interaction with the group G.

© 2011 by Taylor and Francis Group, LLC

 88 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

To estimate the distance scale, at which the Grand Unification takes place, one should
turn to the equations defining evolution of the running constants of the strong and elec-
troweak interactions. In so doing, it is necessary to represent these equations in such a form
so that they determine the constants variation not as a function of the transferred momen-
tum q, but as a function of variation of the mass scale µ. So, we suppose that the G-group
exists. The charges of the SUc (3)-, SUEW (2)-, and UEW (1)-subgroups are indicated by the
symbols g3 , g2 , and g1 respectively. When µ = MGU these three constants are merged into
the one constant of the Grand Unification, that is,

g3 (µ) = g2 (µ) = g1 (µ) = qGU , at µ ≥ MGU .

Once µ < MGU , then the constants gi (µ) are being separated and, at long last, shall pass
into the phenomenological constants g, g ′ , and gs describing the observed interactions under
energies less than mW c2 .
Let us find the link between the interaction constants in the models with the groups
SUc (3) × SUEW (2) × UEW (1) and G under µ ≥ MGU . The Lagrangian determining the
interaction between the quarks and gluons is connected with the SUc (3)-group while the
Lagrangian describing interaction between the fundamental fermions (the quarks and lep-
tons) and the gauge bosons of the electroweak interaction is connected with the SUEW (2)-
and UEW (1)-groups. Apart from the γ-matrices, the wavefunctions of bosons and fermions,
these Lagrangians also contain Ngenerators and N the interaction constants of the correspond-
ing groups. When G ⊃ SU (3)c SU (2)EW U (1)EW , then the generators of the SUc (3)-,
SUEW (2)-, and UEW (1)-subgroups must be the generators of the G-group too. It means
that the normalization of all the generators must be the same. For non-Abelian groups
the normalization is fixed by the nonlinear commutation relations between the generators
defining their Lie algebra. For this reason, the normalization of the G-group generators is
just the same as that of the SUc (3)- and SUEW (2)-group generators. From this follows

gs (µ) = g3 (µ), g(µ) = g2 (µ). (6.29)

Since there are no nonlinear constraints on the generator of the Abelian group UEW (1),
then this generator should be multiplied by the coefficient c in order to ensure the right
normalization. Thus, we have
g ′ (µ) = cg1 (µ), (6.30)
where c is defined by the G-group structure. The relations (6.29), (6.30) are fulfilled in the
limit of the G-group, that is, under µ ≥ MGU . Now we should consider the region µ < MGU .
The cause of changing the gauge coupling constants is the vacuum polarization, that is, it is
stipulated by the processes of creation and consequent destruction of the virtual particles.
If one neglects the contribution coming from the Higgs bosons then the coupling constant
evolution in the SU (3)c -, SU (2)EW -, and U (1)EW -gauge groups shall be described by the
equations (see, for detail, Ref. [3]):

dgi
= −bi gi3 , (6.31)
d(ln µ)

where
nf 22 − 2nf 33 − 2nf
b1 = − , b2 = , b3 = .
24π 2 48π 2 48π 2
To integrate the relations (6.31) leads to the result

gi−2 (µ) = gi−2 (µ0 ) + 2bi ln(µ/µo ). (6.32)

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 89

Thus, according to the theory, the dependence of 1/αi = 4πgi−2 on ln M is linear, its slope
value being defined by the polarization effect of the relevant vacuum. Discrepancy of the
paths on which the three gauge coupling constants 1/αi (µ) converge to 1/αGU is caused
by discrepancy of the coefficients bi in Eq. (6.32). It is easy to guess that the larger value
of the slope of 1/αs compared to the slope of 1/αW is caused by the fact that the number
of the gluons is larger than the number of the carriers of the weak interaction (the W ± -
and Z-bosons) by a factor of two and, as a result, the gluons give the bigger antiscreening
effect. In 1/α the screening effect predominates (the tangent of the slope angle is negative
by now) and for this reason the value of 1/α drops with the growth of M .
On the unification scale µ = MGU we have
−2
gGU = gi−2 (µ0 ) + 2bi ln(MGU /µo ). (6.33)

To find MGU we should exclude nf and gGU from Eqs. (6.33). For this purpose we produce
a linear combination
c2 1 1 + c2 MGU
2 + 2− 2 = 2[c2 b1 + b2 − (1 + c2 )b3 ] ln , (6.34)
g1 g2 g3 µ0

where gi = gi (µ0 ). With allowance made for Eq. (6.4) the left-hand side of Eq. (6.34) takes
the form
1 1 1 + c2 1 1 + c2
+ 2− 2 = 2− . (6.35)
g ′2 g g3 e g32
To substitute (6.35) into (6.34) gives the desired equation for determining MGU
   
MGU 48π 2 1 1 + c2 6π 1 1 + c2
ln = − = − , (6.36)
µ0 22(1 + 3c2 ) e2 g32 11(1 + 3c2 ) α αs

where the absence of subscript “em” by α underlines the circumstance that the case in
point is the fine structure constant not in the QED, but already in the more precise theory,
namely, in the theory of the electroweak interaction. Setting the values of µ0 , α(µ0 ) and
αs (µ0 ) one may evaluate both the value MGU at which the relation (6.36) is fulfilled and
the value of the unified constant αGU (MGU ). Analogous calculations give one of the basic
parameters of the GWS theory, sin θW ,
 
1 2 α
sin2 θW = 1 + 2c . (6.37)
1 + 3c2 αs

Set µ0 = mW and take into account that α(mW ) ≈ 1/128 and αs (mW ) ≈ 0.1. Further
we choose the SU (5)-group as the G-group. That gives the value 5/3 for the quantity
c. Then for sin2 θW we obtain ∼ 0.23 that is amazingly close to the experimental value.
This suggests that to estimate the unification scale order one may use the SU (5)-group.
Then Eq. (6.36) gives ln(MGU /mW ) ≈ 29 and we obtain MGU /mW ≈ 4 × 1012 . It is not
difficult to find αGU (MGU ), it proving close to 0.02. One may carry out the analogous
calculations not concretizing the G-group, but under the condition that all the known
fermions (or every fermion families individually) form the total representation of the G-
group [4]. At µ0 = mW and with an allowance made for the contribution in vacuum
polarization coming from the SM Higgs boson the dependence of the running constants αs ,
α, and α2 on µ is represented in Fig. 6.1. As it follows from Fig. 6.1, in this case the
Grand Unification takes place at MGU ≈ 2 × 1014 GeV/c2 . The more detailed analysis
leads to the result: MGU ≈ 2 × 1015ΛQCD . The obtained values correspond to the distances
LGU ≈ ~c/EGU ∼ 10−28 cm. At distances, shorter than LGU , the initial symmetry is

© 2011 by Taylor and Francis Group, LLC

 90 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

50

40 3/(8a)
1/a GU
30
1/a 2
20

1/a S
10
mW M GU
L QCD

1 10
2 4
10 10
6 8
10
10
10
12
10
14
10 10
16 m , GeV

FIGURE 6.1
The µ-dependence of the running coupling constants αs , α, and α2 .

restored and the interaction is described by the single constant αGU , which evolutionary
law is defined by the structure of the G-group.
At this point it is natural to pose a question: “whether it is possible to trust calculations,
altogether ignoring the gravitation interaction.” Gravitation effects are of the order of 1,
when the masses of interacting particles have such values, that the potential gravitation
energy is comparable to the particle rest energy

GN MP2
≈ M P c2 . (6.38)
r
If we set the distance between masses equal to the Compton wavelength, that is, to quantity
~/MP c, then the condition (6.38) is realized at
 1/2
~c
MP = = 1.22 × 1019 GeV/c2 . (6.39)
GN

The obtained mass value is called the Plank mass. The time and the length, corresponding
to it r r
~GN −44 ~GN
tP = ≈ 5.4 × 10 s, LP = ≈ 1.6 × 10−33 cm, (6.40)
c5 c3
are called the Plank time and length. Since LGU ≫ LP , then on the Grand Unification scale
we have the right to neglect gravitation effects.
N The choice
N of a GUT group G is defined
by the demand: it must contain SU (3) SU (2)EW U (1)EW as a subgroup and has
representations in which all the known quarks and leptons may be included. Since the SU (5)-
group is the minimal group satisfying this demand the first to be investigated was the SU (5)-
model (Georgi-Glashow model [5]). The schema based on the SO(10)-group [6] is the direct
generalization of the SU (5)-model. Probably, in laboratory conditions we shall never be able
to produce energies, corresponding to the unification scale, consequently, the experimental
check of the GUT is a very complicated task. Among the GUT consequences being available
for observation we note the predictions of such effects as proton instability and neutron-
antineutron oscillations (neutron transformation into antineutron in the vacuum and the

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 91

reverse process). In the electroweak interaction the gauge constants g and g ′ are not bound
together and their ratio g/g ′ = tan θW is an experimentally obtained parameter. Opposite
to that, the GUT allows us to calculate the Weinberg angle. The GUT models explain
naturally electric charge quantization, manifested through the fact that quark charges are
multiples of e/3, while lepton charges are equal either to e or to 0.
The GUTs have some cosmological consequences as well. According to an adopted point
of view, our universe came into being approximately 2 × 1010 years ago as a result of the
Big Bang and it is still expanding. This expansion is described by the GRT equations. The
universe size changed from the value of the Plank length order to a contemporary value,
which is of the order of 1028 cm. A substance, compressed in a volume of the order of
L3P , began its evolution with the energy of the Plank order, that is, the early universe is
a gigantic laboratory, where the GUT consequences could be checked. Within the GUT
framework it is possible to get an explanation of the fact, that the matter at the moment
is prevailing over the antimatter in the universe (baryon asymmetry). The value of the
ratio of the baryons concentration nB to the photons concentration nγ in cosmic microwave
background can be also obtained in the context of the GUT.
Along with the above-mentioned achievements, there are some weak points in the existing
GUT. Let us enumerate some of them. The models have a rich variety of free parameters,
which number exceeds the number of those in the SM. It is not possible to make any state-
ment concerning the number of fermion generations within the framework of the models.
The gravitation is excluded from the unification scheme. Serious difficulties are produced
under explanation of difference by twelve orders of the distance scales at which the Grand
Unification symmetry G and the electroweak interaction symmetry are broken (the hierar-
chies problem).
Supersymmetry. Is there a more “grandiose” unification, a unified field theory, which
includes both the gravitation and the SM? Before we discuss the directions in which the con-
struction of the theory is going, let us first get acquainted with one more type of symmetry—
a supersymmetry (see, for review, [7]).
Up to now we have been considering the space-time and internal symmetries. Geometrical
translations and rotations do not change the nature of a particle: the photon remains the
photon after any space-time transformations. The internal symmetries can change the
nature of a particle but not its spin value. So, under the action of the isotopic rotations the
proton can turn into the neutron, but it cannot transform into the π 0 -meson, for instance.
Unlike the above-mentioned symmetries, supersymmetry transformations can change not
only the space-time coordinates of a particle and its nature, but its spin value as well. In
other words, the supersymmetry implies the invariance of a physical system under fermion-
boson transitions, that in its turn, allows us to call it the Fermi-Bose symmetry. The
basis for supersymmetric theories is the extension of the space-time up to the superspace,
which besides the normal space-time coordinates xµ , also includes the spinor coordinates
θα . The spinor variables θα anticommute with each other and commute with the space-time
coordinates xµ

{θα , θβ } = 0 α, β = 1, 2, 3, 4,
(6.41)
[θα , xµ ] = 0 µ = 0, 1, 2, 3.

It follows from (6.41) that θα and xµ can be viewed as odd and even generatrices of a Grass-
mann algebra, respectively. θα is usually called Grassmann coordinates. Anticommutation
of θα is necessary to ensure the right connection between the spin and statistics. The desire
to describe bosons and fermions on the same level became the reason of introducing θα . To
simplify the matter, we may say, that the classical coordinates correspond to bosons, while
the Grassmann coordinates θα correspond to fermions. The important consequence of the

© 2011 by Taylor and Francis Group, LLC

 92 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

anticommutation of the Grassmann coordinates is their nilpotency

θα2 = 0.

In the most general case, points in superspace are characterized by four even coordinates
xµ and 4N odd coordinates θαj , where j = 1, 2, .., N (N-extended supersymmetry). Let
us restrict ourselves to the supersymmetry with N = 1. In the ordinary four-dimensional
space-time there is the Poincare transformations group with 10 parameters, while in the
superspace the expanded Poincare group with 14 parameters comes into action, where
besides the rotations and translations in ordinary space the supertranslations

x′µ = xµ + 2i ǫγ µ θ,
(6.42)
θ′ = θ + ǫ,

are added. In Eq. (6.42) θ is represented in the form of 4-columns consisting of θα , the
spinor ǫ defined by 4 real parameters sets the supertranslation and ǫ = ǫ† γ0 . Under the
ordinary Poincare transformations θ and ǫ behave as ordinary spinors. Generators Qα
correspond to the supertranslations
∂ i ∂
Qα = − (γ µ θ)α µ . (6.43)
∂θα 2 ∂x
Besides the ordinary relations of the Poincare algebra (1.3.46)–(1.3.48), the supersymme-
try algebra also includes
[Qα , Pµ ] = 0, (6.44)
1
[Qα , Mµλ ] = (σµλ )βα Qβ , (6.45)
2
{Qα , Qβ } = −(γ µ C)αβ Pµ , (6.46)
where σµλ = i/2[γµ , γλ ], and C is a charge conjugation matrix. The first of these relations
(6.44) means, that the supertranslation does not influence shifts in the classical space-time,
the second one means, that Qα is transformed under the Lorentz transformations as a
ordinary spinor. Due to the anticommutation relation (6.46) internal degrees of freedom
of a particle are connected with external space-time degrees of freedom. Both continuous
transformations, carried by the generators Pµ and Mµλ , and the discrete ones, corresponding
to the generators Qα , turn out to be unified in superalgebra.
The procedure of theory construction is as follows. All the fields are substituted for a
superfield Φ(x, θ) and expansion in terms of the Grassmann variables θα is done. In so
doing, it appears, that instead of an infinite series we obtain the finite number of expansion
members of Φ(x, θ). It is caused by properties of the Grassmann algebra, in which prod-
ucts can be composed of no more components than the number of algebra generatrices. As
soon as two identical generatrices occur during expanding, by means of anticommutation
relations (6.41) they can be placed together, and their square is equal to zero. At every
member of the obtained series some tensor or spinor function appears according to whether
the degree of θ is even or odd. The Lagrangian of the theory is defined in the form of
invariant square expressions that consist of the superfield and first derivatives of the su-
perfield with respect to both the classical and Grassmann coordinates. The superaction
is defined as an integral of the Lagrangian over all variables. As a result of the integra-
tion over θα , the four-dimensional action is obtained, which is expressed only by means of
four-dimensional spinor and boson functions, the expansion coefficients of Φ(x, θ). Thus,
the Grassmann variables are not included in the final result. Like fields in the Minkowski
space, the superfields are classified according to eigenvalues of the Casimir operators, which

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 93

are built out of supersymmetry group generators. Quantum numbers are the superspin and
the superisospin, which generalize notions of the spin and isospin for ordinary fields. To
single out irreducible representations from the superfields, either additional conditions are
imposed (to eliminate the redundant superspins) or a gauge invariance demand is invoked.
As we see, the prescriptions are the same as for ordinary fields.
Since every boson is associated with a supersymmetric fermion and vice versa, then the
number of particles in the theory is doubled. The supersymmetric partners get their names
either with the prefix “s” for scalar partners of normal fermions (for example, squark,
selectron) or with the ending “ino” for fermionic partners of normal bosons (for example,
photino, gravitino).
In supersymmetric theories divergences in perturbation theory series, corresponding to
bosons and fermions have the opposite signs and mutually compensate each other. Thus,
there is no need for the renormalization at all. In other words, the supersymmetry allows
constructing finite, divergence-free theories. The inclusion of the supersymmetry in the
SM led to come into being the minimal supersymmetric standard model. Discovery of the
superpartners of the known fundamental particles will experimentally prove the existence
of the supersymmetry in nature. By now, search of the superpartners have not given the
positive results.
Supergravitation. In the 1960s some papers appeared in which the GTR was reformu-
lated in the form of a gauge gravitation theory. For this purpose the symmetry group of
the flat space-time, the 10-parameters Lorentz group, was chosen and its localization was
carried out (that is, the transformation parameters became the coordinate functions). As
a result, the gauge fields appeared, which were associated with the gravitation field of the
GTR. The theory obtained was completely identical to the GTR and in fact, produced no
new results. The development of the gauge interpretation of the gravitation was defined
mainly by hopes to use it in the coming time for unification of the gravitation interaction
with the other ones. The star hour of the gravitation gauge variant came in the 1970s when
the supersymmetry theory had appeared. The theory, appearing as a result of the merging
of two origins, the supersymmetry and the gauge principle, was called the supergravitation
(see, for review, [8]). The supergravitation geometry is as simple and elegant as that of
Einstein’s GTR (the latter corresponds of the supergravitation with N = 0). The basis of
supergravitation is a relativity principle, which reads, that “the form of physical laws does
not depend on the choice of a coordinate system in the superspace.”
In the simple supergravitation (N = 1) the fourteen-parametric Lorentz group, extended
by the supertranslation transformations is localized. In this variant of supergravitation
the graviton and its superpartner gravitino (with spin 3/2) are the carriers of the gravi-
tation field. In the simplest supergravitation extension (N = 2) a symmetry group with
18 parameters is localized, and consequently, there are more interaction carriers, they are:
the graviton, two gravitinos, and a graviphoton. The N = 2 supergravitation represents
the first supergravitation theory, which unifies the gravitation with the electromagnetism
in principle. It becomes possible to unify particles with spin 2 and 1 due to the presence
of the intermediate stage, the particle with the spin 3/2. In supergravitation the number
of divergences is much less than the one in the ordinary gravitation theory. Many viable
supersymmetric field theories contain supergravitation as an important component, which
helps to spontaneously violate the supersymmetry. In such models the hierarchy problem
finds its solution.
The 0(8)-supergravitation is the most extensive and promising supergravitation theory.
It is probably nonrenormalizable. In addition to this confusing fact, the 0(8)-gauge group
is too small to contain the minimal gauge group of the SM. In transition to groups, more
extensive, than 0(8), particles with spin ≥ 3 appear. By now all the attempts to construct a
consistent theory of interacting particles with the spin 3 and higher have been not successful.

© 2011 by Taylor and Francis Group, LLC

 94 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It makes us suspect, that the 0(8)-gauge group is the limit of the supergravitation. Notice,
that the SM group might appear within the frameworks of the N = 8 supergravitation. It
turns out that on a mass shell the 0(8)-symmetry extends to the SU (8)-symmetry, which
already contains the gauge group of the SM.
Kaluza-Klein five-dimensional world. To get acquainted with another direction in
construction of an unified field theory, let us consider the Kaluza concept [9] later developed
in the works by K. Klein [10]. The idea was to unify the GRT and the Maxwell electrodynam-
ics on the basis of a hypothesis that our universe is the curved five-dimensional space-time.
One of the coordinates denotes the time, while the other four are spatial coordinates. In
such a space the five-dimensional interval square dS 2 is defined by the relation

dS 2 = gAB (x)dxA dxB , (6.47)

where A, B=0,1,2,3,4, and gAB (x) is a metric tensor with fifteen independent components.
Then ten combinations gAB (x)
g4µ (x)g4ν (x)
gµν (x) +
g44 (x)
are associated with ten components of the metric tensor of the GRT ηµν (x). The following
four combinations:
g4ν (x)
p
−g44 (x)
are connected with four components of the electromagnetic potential Aµ (x). Let us explain
why this operation is legal. Remember that the Christoffel symbols ΓAC,B are field strengths
in the curved space-time. We set one of the indices C equal to 4, A and B equal to 0,1,2,3.
Then, using (6.8), we obtain
 
1 ∂gµν (x) ∂gν4 (x) ∂gµ4 (x)
Γµ4,ν = + − . (6.48)
2 ∂x4 ∂xµ ∂xν
When one assumes, that the fifth coordinate is cyclic, then the expression (6.48) takes
the form  
1 ∂gν4 (x) ∂gµ4 (x)
Γµ4,ν = − , (6.49)
2 ∂xµ ∂xν
or, having set
c2 c2
Fµν (x) = √ Γµ4,ν , Aµ (x) = √ g4µ (x), (6.50)
GN 2 GN
for (6.49) we arrive at  
∂Aν (x) ∂Aµ (x)
Fµν (x) = − , (6.51)
∂xµ ∂xν
that is, g4ν (x) and Γµ4,ν can be really identified with the potential and the tensor of the
electromagnetic field, respectively. The equations, governing system evolution, follow from
the least-action principle
δ(Sm + Sf ) = 0, (6.52)
where Sm and Sf are actions for the matter and the field, respectively. Variation of the
first term in Eq. (6.52) results in five equations for geodesic lines, four of which coincide
with the known four-dimensional equations for charged particles moving in the gravitation
and electromagnetic fields
d2 xν dxµ dxλ e dxµ
= −Γµλ
ν + Fνµ , (6.53)
ds2 ds ds m ds

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 95

and the fifth equation

dx4 1 q
=− √ (6.54)
ds 2 GN m
shows, that while a body is moving in the gravitation and electromagnetic fields, its electric
charge is conserved.
Varying only the potentials of the five-dimensional space-time, we obtain fifteen equations,
which break down into the system out of ten of the ordinary four-dimensional GRT equations
(6.7) and the system out of four of the Maxwell equations

∂F µν ∂F µν ∂F νσ ∂F σµ
= j µ, + + = 0. (6.55)
∂xν ∂xσ ∂xµ ∂xν
In so doing, the equation for the scalar component g44 (x) again is out of use.
Despite obvious merits, the Kaluza-Klein theory leaves two questions to be completely
open:

(i) What is the physical meaning of the fifth coordinate x4 and why it is not observable?
(ii) Why all physical quantities are cyclic with respect to x4 ?

The obvious answer could be that the manifestation region of the additional dimensionality
is beyond the existing experimental technology. At the end of the twentieth such state-
ments become the rule of good form and the mighty imagination of the theorist-physicists
produces a great oasis of exotic phenomena in the energy scale close to the Plank energy.
However, by 1938 only A. Einstein and P. Bergmann could come up with such an idea [11].
They suggested that the fifth coordinate can change from 0 to some value L, that is, the
five-dimensional world is confined in a layer with thickness L. The assumption was also
made, that any function Ψ(x), related to physics, changes little along x4 over a length of
the layer, so that
dΨ(x)
L ≪ Ψ(x)
dx4
and in the average Ψ(x) may be considered as a function only of the four-dimensional
coordinates. Actually, instead of restricting x4 values to quantity L, it is possible to assume,
that the fifth coordinate varies within infinite limits, however, only functions, periodic in
x4 with the period L are under consideration. It means, that it is possible to glue together
all the points, being distant from each other along x4 on interval L, without any harm for
generality done. As a result, we arrive at the five-dimensional space-time being closed by
x4 . The world with such a property we shall call cyclic, closed, or compactified in the fifth
coordinate. In such a theory there is no need for postulating the cyclic character of x4 , since
the gauge principle of switching on electromagnetic interaction, which had been armed by
us, resulted in the conclusion, that wavefunctions of charged particles have the form
 
iec
Ψ(x) = Ψ(xν ) exp √ x4 , (6.56)
2 GN ~

where Ψ(xν ) is an ordinary wavefunction in the four-dimensional space. The expression

(6.56) describes the cyclic dependence of Ψ(x) on x4 with the period
√
4π GN ~
L= ≈ 10−31 cm. (6.57)
ec
Thus, the cyclic period or the world compactification scale in the fifth coordinate is in-
finitesimally small compared to the scale of phenomena, studied by contemporary physics.

© 2011 by Taylor and Francis Group, LLC

 96 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Consequently, it is not surprising, that the fifth dimensionality has been skillfully hiding it-
self from experimentalists up to now. Of course, the extension of space-time dimensionalities
can be generalized on larger dimensionalities as well.
Since the early universe was a substance compressed in the volume with the radius of the
Plank length order, then it is easy to understand that the evolution of universe is completely
defined by elementary particle physics. Consequently, the idea of the closed dimensionalities
must find its place in the Big Bang theory, too. The universe is thought to have carried
out the space-time compactification in additional dimensionalities at the early stages of
evolution when its energy was rather high. At present, this hypothesis is the necessary
attribute of the contemporary multidimension theories.
Superstring theory. The quantum field theory (QFT) and the GTR contain all hu-
man knowledge of the fundamental forces of nature. The QFT brilliantly explains all the
microworld phenomena up to the distances 10−15 cm. The GTR, in its turn, is beyond any
competition in describing both large-scale events in the universe, as well as the evolution
history of the universe itself. The striking success of these theories is caused by the fact,
that taken together, they explain both the behavior and structure of matter from subnu-
clear to cosmological scales. All attempts to unify these theories, however, have been always
facing two main difficulties. The former deals with the appearance of divergences, while the
latter demands to refuse one or more ideas about the nature of universe, which are highly
respected nowadays. The QFT and the GTR are based on few postulates, which appear
to be mandatory and natural. The postulates, the abandonment of which helps to unify
two theories, are as follows: (i) space-time continuity; (ii) causality; (iii) theory unitarity
(the absence of the states with the negative norm); (iv) interaction locality; and (v) point
(structureless) particles. The first four items are rather serious. Against their background
the abandonment of the fifth item looks like a child’s prank.
Any local theory, operating with structureless objects contains divergences at energies,
higher than the Plank energy EP . It is caused, as we know, by the fact, that EP is that very
energy value, when the gravitation quantum theory is needed, while it is not renormalizable.
To achieve renormalizability, another model can be introduced, in which at “low” energies
(EP ) fundamental objects behave as point ones and their composite structure reveals itself
only at energies, higher than EP . If one adds to this hypothesis some latest inventions,
concerning unification of all interactions (supersymmetry, Kaluza-Klein theory, etc.), then
one more way is being opened to the unified field theory, called the superstring theory (the
reader could find the superstring theory presentation in the book [12]).
String theories were used by hadron physics as early as the 1960s. Their creation was
inspired by the fact, that at the distances of the order of 10−13 cm the gluon fields, binding
the quarks, are concentrated in space not evenly but along the lines connecting the quarks.
It resulted in the interpretation of the hadrons as nonlocal objects, one-dimensional strings,
on the ends of which either the quarks for the baryons (fermion strings) or the quarks
and antiquarks for the mesons (boson strings) are placed. Thus, the infinitely thin tube of
the gluon field is modeled by a relativistic string. Relativistic string theories managed to
explain many anomalies of the quark universe. Thus, for example, since the string energy is
proportional to its length L, and the string mass square is m2 ∼ L2 , then the string angular
moment of the string having the form of the linear segment is proportional to L2 . Really,
experiments confirm linear dependence between the hadron spin J and the square of its
mass J ≈ αm2 , where α ≈ 4 × 10−28 cm2 /(~c)2 . The relativistic string, which binds the
quark and antiquark, generates potential, linearly growing with distance, that is, it explains
the quark confinement. The theory is built in such a way, that after the string breaks no
free quarks appear, since at both newly formed string ends a pair “quark-antiquark” is
produced.
However, the string relativistic theories have some very serious deficiencies, such as, for

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 97

example, tachyon existence, massless hadrons with spin 2, etc. It also turned out, that
consistent quantum theories could be only formulated in 26-dimensional space-time for
bosonic strings and in 10-dimensional space-time for fermion strings.
The renaissance of interest in string theories took place after 1986 when M. Green and
J. Schwarz proved [13], that the 10-dimensional
N ∗ gauge superstring theory, based on the
internal symmetry group SO(32) or E8 E8 (subscript indicates the group rank) can
be used to unify all the interactions. Superstrings are one-dimensional in a spatial sense
(two-dimensional, if the time is taken into account) objects with the typical length of the
order of LP . They are put in an n-dimensional (n ≥ 10) space-time manifold. The observ-
able space-time dimensionality is achieved by compactifying unnecessary dimensionalities
at distances of the order of the Plank length. The theory contains a mechanism, ensuring a
spontaneous compactification of additional dimensionalities. The initial symmetry is bro-
ken up to a symmetry group, involving the supergravitation and the supersymmetric
N GUT
with fixed parameters and given particles content. If the gauge group E8 E8 is used then
one E8 -group contains all the low energy physics, while the other E8 manifests itself only in
the gravitation interaction. Phenomenological properties of the superstring theory depend
a lot on a compactification mechanism. As an example, let us pay attention to the following
circumstance. Since the division into the normal space-time dimensionalities and compact-
ified ones is not very strict, then it is possible, that some universes with nonconventional
dimensionalities of the space-time exist.
The important difference between the superstring theory and the local field theory is
that in the former theory the free superstring is characterized by an infinite number of
supermultiplets, while in the latter one every field describes particles of only one kind. The
superstrings have the same number of fermion and boson degrees of freedom. Superstring
excitations (which are rotations, vibrations, or excitations of internal degrees of freedom)
are associated with elementary √ particles observed. Particle mass scale is regulated by the
superstring tension T with T ≈ MP c2 . The number of states with masses, smaller than
the Plank mass, is finite. It defines the number of elementary particles existing in nature.
There is also a great number of excitations with masses, bigger than the Plank mass.
The majority of these modes are unstable, however, there are also stable solutions with
exotic characteristics (a magnetic charge, for example). It is remarkable, that in particle
spectrum, which corresponds to superstring theory solutions, one massless state with the
spin 2 appears, which is described by the GTR equations in a low energy limit, that is, it
is the graviton.
Strings appear in two topologies: as open strings with free ends and as closed loops.
Besides this, they can possess an internal orientation. Quantum numbers of the open
strings are located at their ends, while in the closed loops quantum numbers are evenly
spread along the string. The string interaction has the local character, despite the fact,
that they are extended objects. When interacting the strings can scatter, produce new
strings, and emit point particles as well.
It is evident that the choice of one or another model in the role of a candidate in an unified
theory is the most principal question in the superstring theory. It is proved that all the
known superstring theories are connected with each other by a duality transformation [14].
Discovery of the duality allows to reveal nonperturbative superstring properties that are
connected with the existence of new stretched objects, D-brans.† The assumption was stated

∗ E together with the groups G , E , E , and E constitute the exceptional group class. The rank
8 2 4 6 7
of the exceptional group is fixed, while the normal (regular) group can have any rank (for instance,
SU (2), SU (3), SU (5), and so on).
† The bran is a three-dimensional manifold with ordinary matter enclosed in a comprehensive multi-

dimensional space. The D-brans include not only the matter fields but the gauge fields as well.

© 2011 by Taylor and Francis Group, LLC

 98 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

that all superstring theories are the special case of the unified fundamental theory, the m-
theory, which lives in 11-dimensional space-time and describes 11-dimensional supergravity
on rather small distances. The development of the superstring theory showed that it was a

FIGURE 6.2
The building of twenty-first century physics.

fruitful generalization of the local field theory. However, nowadays the superstring theory
still undergoes its development stage and has no experimental confirmations yet. Let us note
that the experimental check of superstring models is very difficult due to many unknown
parameters they contain. We would like to believe that the completed superstring theory
would contain only two fundamental parameters: the tension and the superstring interaction
constant.
In Fig. 6.2 the building of twenty-first century physics is pictured. Unfinished floors
are designated by the dashed lines. Physics as a science originated from astronomy at the
end of the sixteenth century. Two persons were at the cradle of physics, Tycho Brahe and
Johannes Kepler. The former was the first experimentalist, whereas the latter was the first
theorist. The majestic building of physics has been under construction for 400 years. And

© 2011 by Taylor and Francis Group, LLC

 Fundamental interactions 99

now, for its construction to be completed joint operations of already millions of physicists
are required.

© 2011 by Taylor and Francis Group, LLC

 7
Atoms–Nuclei–Nucleons

The bullfinch was a sharp little fellow and a born misin-

-former. He received his education at a school for soldiers’
children and afterwards served in the army as a regimental
clerk. Having tackled the rules of punctuation, he began
to publish an uncensored periodical entitled “The Wood-
land Herald.” Yet, try as he might, he somehow just
couldn’t fall in with official requirements. Either he
touched on something forbidden, or he would omit a point
which not only could but should have been mentioned. And
naturally, he got rapped on the knuckles for this.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

7.1 Atomism
The idea of fundamental particles has always symbolized the deepest aim of scientific
cognition–to express the complexity of nature through its simplest notions. Twenty-five
hundred years ago Greek philosophers laid the foundations of our understanding of the
nature of matter, when they made the first attempts to reduce the variety of the world to
the interaction of a few initial components, fundamental particles, or elements. In 4 B.C.
Phales suggested that water was the single primary element, of which all the existing world
is made up. Later Anaximen from Milet extended the list of primary elements to four,
such as soil, air, fire, and water. It is generally agreed that Democritus (460–370 B.C.) is
the creator of the atom idea, however, history also mentions his teacher Levkipus in this
connection.
A legend states that one time Democritus sat on the seashore with an apple in his hand.
His stream of thought was as follows: “Suppose, I shall cut this apple in halves, and then
every half I shall cut in halves again, so that there will be a quarter of an apple left, then
one eighth, one sixteenth, etc. The question is, if I continue to cut the apple to the end, will
the cut parts always possess properties of an apple? Or, maybe, at a certain moment the
remaining part of the apple will have no properties of an apple any more?” Upon serious
consideration the philosopher arrived at the conclusion that there was a limit for such a
division and he called the last indivisible part an “atom.” His conclusions mentioned below,
are stated in his book Great Diacosmos.
“The beginning of the Universe is atoms and emptiness, other things exist only in our
imagination. There exists infinite number of worlds, and they all have their beginning and
end in time. And nothing appears from non-existence, nothing goes into non-existence.
Atoms are uncountable both in amount and in variety, whirling like the wind they rush in
the Universe and so all composite substances create: fire, water, air, soil. The point is that
these substances are compounds of some atoms. Atoms are unaffected by any influence, they

101
© 2011 by Taylor and Francis Group, LLC
 102 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

are unchangeable due to their hardness.”

Democritus could not prove his statements, so he suggested his contemporaries trust
his words. The majority of contemporaries did not believe him, and Aristotle was among
them. Aristotle was the author of the opposite teaching. According to him, the process
of the apple partition could be infinitely extended. Natural philosophy is not based on
experiments and mathematics, it used the element of faith as the single criterion of truth.
So it is not surprising, that both teachings for the ancients seemed to be equally reasonable
and acceptable. It is difficult to tell, what has weighed down the weights bowl in favor of
Aristotle’s philosophy. Maybe, it was the gleam of military glory of Alexander the Great,
whose teacher was Aristotle. Anyway, the teachings of Aristotle became dominant, while
Democritus had been forgotten for many centuries.
In the 1640s Democritus’s idea about atoms had been restored to life by the French
philosopher Gassendet. When spring comes, all violets bloom at once. It was so with
the atomic hypothesis. After twenty centuries of oblivion all the contemporary advanced
scientists believed in the atomic theory, including the great Isaak Newton, whose credo was
“not to build any hypothesises.” One of the burning questions of atomism was, undoubtedly,
the question whether the variety of bodies in nature means, according to Democritus, the
same variety of atoms? If the answer is positive, then the atomic hypothesis does not bring
us closer to an understanding the world. Luckily, the answer is negative. The variety of
substances in nature is caused not by a variety of different types of atoms, but by the variety
of different compounds of these atoms (nowadays, these compounds are called molecules).
In 1808 D. Dalton, upon studying many chemical reactions, precisely formulated the notion
of a chemical element: “Chemical element is a substance, consisting of atoms of the same
type.” It turned out, that there were not so many chemical elements. In 1869 D. Mendeleev
managed to place all of them in one periodic table (at that time only 63 elements were
discovered, while now their number reaches 120). The work by the botanist R. Brown
(1827) may be considered as the first experimental proof in support of the atomic theory.
He observed chaotic motions of flower pollen in water (Brownian motion). The discovery
by Brown did not attract scientists’s attention immediately and for a long time its nature
remained unclear. Only seventy eight years later, the atomistic theory of Brownian motion
was established in works by A. Einstein and M. Smoluchovski. In 1908 J. Perrin carried out
a series of experiments to study Brownian motion. Not only did he prove experimentally
the works by Einstein and Smoluchovski but he measured sizes and masses of atoms as well.
The last and, probably, final proof of atomic matter structure was the work by E. Rutherford
and Royds on measuring the number of α-particles in radium. By that time it had been
known, that in minerals, containing α-radioactive elements (radium, thorium), helium is
accumulated. The task was to determine the number of α-particles emitted by a sample
and measure the volume of helium produced on the sample. In a second 13.6×1010 particles
are emitted by one gram of radium. Having captured two electrons all these α-particles turn
into helium atoms and occupy the volume of 5.32 × 10−9 cm3 . Consequently, 1 cm3 contains
L = 2.56 × 1019 atoms. Let us compare the value obtained with the Loschmidt number
calculated as early as 1865 on the basis of the molecular-kinetic theory. One mole of helium
(or any other gas) occupies a volume 2.241×10−2 cm3 /mole and contains 6.02×1023 atoms,
that is, 1 cm3 contains 2.68 × 1019 atoms. The coincidence is impressive. So, the existence
of atoms got the final experimental proof. That has completed the ascent of physics on the
first step of the “Quantum Stairway.” A picture of the world on this step was very simple:
the matter in our universe consists of indivisible atoms of different types, which make all
the elements of the Mendeleev periodic table.
The variety of elements and the explicit systematization in the periodic table by Mendeleev
inspire into us belief that far from being utopian, the result obtained is not final yet. The
existence of the next step of the “Quantum Stairway” became obvious after a series of

© 2011 by Taylor and Francis Group, LLC

 Atoms–Nuclei–Nucleons 103

experiments, which may be called the “Roentgen” of atom according to Rutherford (1909–
1911).

7.2 Rutherford model of the atom

The fact that all atoms contain electrons was the first important piece of information about
the internal structure of an atom. The electrons have a negative electric charge, while atoms
are electrically neutral. Consequently, every atom must contain enough of positively charged
matter to compensate the negative charge of the electrons. In 1903 in the book Electricity
and Matter, D. D. Thomson introduced a model of a radiating atom, which was satisfying
to the totality of chemical and spectroscopic experiments existing at that time. Thomson’s
predecessor in the model was his famous namesake William Thomson (Lord Kelvin), who in
1902 suggested coming back to the atomic theory of F. Epinus (1759). The Epinus atom is
a sphere, being uniformly charged with positive electricity, in the center of which a negative
charged corpuscle is located. D. D. Thomson developed this model by assuming that the
electrons rotate inside of a sphere, the number of the electrons and their orbit configuration
depending on the atom’s nature. The electrons are spaced as concentric rings (shell) and
perform periodic motions, which cause the observed atom spectrum. Investigating the
stability of the electron combination, Thomson gave the physical interpretation of valence.
The atom model by Thomson was the first real attempt to explain the chemical properties of
a substance and the periodic law by Mendeleev. His model, called “pudding with raisins”
was warmly accepted by the scientific public. One should not think that the spirit of
contradiction, being so characteristic for scientific creativity, has not led to the construction
of some alternative atom models. So, for example, in 1901 A. Perren published the article
titled “Nuclear planetary structure of the atom” in the scientific popular journal Scientific
Review. Two years later the Japanese physicist Ch. Nagaoka (a disciple of Bolzman),
suggested the model according to which an atom consists of a positive charged nucleus and
of the electrons ring rotating around the nucleus (atom of Saturn type). First Nagaoka
published his results in even less popular among physicists journal Proceedings of Tokyo
physical-mathematical society. Then in 1904 he published the article on the same topic in
the itPhilosophical Magazine. However, physicists paid no attention either to the Perren or
Nagaoka models, although these models could be regarded as the predecessors of the atom
nuclear model later suggested by Rutherford.
Nine more years passed before the atom model by Thomson was brought under serious
experimental testing, results of which showed that the model had a single drawback, namely,
it had nothing in common with physical reality. Creators of quantum theory liked to say:“in
order to check what is inside of a pudding with raisins one must simply put a finger in it.”
In reality Rutherford used the method that was rather similar with the one described
above. α-particles from radioactive sources were used as a probe, and thin foils of different
substances with a typical thickness of about 104 atomic layers were used as targets. In
Fig. 7.1, the scheme of these experiments done by G. Geiger and E. Marsden under the
guidance of Rutherford is represented. An ampoule with a radioactive source (radium C,
Po-214) was placed behind a lead screen with a small opening. α-particle beam was passing
through a lead collimator and was directed at a target. After interacting with target atoms,
α-particles got into a mobile screen of sulphureous zinc causing scintillations on it, which
were registered by means of a microscope.
As we know from optics, to see an object, one must use the electromagnetic waves having

© 2011 by Taylor and Francis Group, LLC

 104 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 7.1
The scheme of Rutherford’s experiments.

wavelengths of the same order as its size. The same rule takes place in the quantum world,
only here we are dealing with de Broglie wavelength λ. Of course, in 1909 nobody knew
about de Broglie waves, but nature once again proves to be favorable to man, and α-particles
with energy 7.7 MeV (typical energy values in Rutherford experiments) had a wavelength
of ∼ 5 × 10−13 cm, which was quite enough to “see” a nucleus in an atom.
The Thomson atom model predicted that the main part of α-particles flies straight
through a foil, while the remaining particles are subjected to the small deviation. In reality
the behavior of most α-particles fitted in the frames of this model, but parallel with that
some particles deviated on big angles. Let us show, quantitatively, that the Thomson model
is not able to explain the results of these experiments. When, according to Thompson, one
assumes that a positive charge Q is spread evenly all over the volume of a gold atom and
does not take into account the influence of the electrons, then the electric field strength
inside of an atom at a distance r from its center is given as:
Qr
E+ (r) = ,
4πR3
where R is an atom radius. The α-particle maximum deviation takes place, if it is slightly
touching the atom surface where E+ (r) reaches values of the order of 1013 V/cm. Thus,
the maximum value of the force is determined by the expression:
Q|e|
Fmax = .
2πR2
Since this force rapidly decreases with a distance, then for an approximate estimation, one
can consider the interaction on the small interval L, which includes the distances before
and after the contact between the α-particle and the atom. We accept L ∼ 2R, and take
the force value being maximum, that is, during the time interval ∆t = L/vα ≈ 2R/vα the
α-particle is subjected to Fmax . In this case, the transverse momentum, transferred to the
α-particle, is equal to
Q|e|
∆pα = Fmax ∆t ∼ .
πRvα
Consequently, the maximum deviation angle is as follows:
Q|e|
θ+ < . (7.1)
πRmα vα2

© 2011 by Taylor and Francis Group, LLC

 Atoms–Nuclei–Nucleons 105

Using the values 6.6 × 10−27 kg and 2 × 107 m/s for the mass and the velocity of the
α-particle respectively, we obtain in the case of the gold atom (Q = 79|e|)

θ+ < 0.020 .

Now let us take into account the influence of the atomic electrons on the α-particles
motion. We assume that their initial velocity is equal to zero. Then the momentum transfer
to an atomic electron is maximum under the head-on collision. From the conservation laws
of the momentum and the energy (in a nonrelativistic case) it follows, that after a collision
an electron has acquired the velocity
2mα
ve = vα ≈ 2vα ,
mα + me
where vα is the initial velocity of the α-particle. Certainly, at such a collision the α-particle
does not deviate. At a sliding collision the electron momentum change ∆p would be already
less than 2me vα . To obtain a value of a maximum possible deviation θ− under the scattering
of the α-particle by the electron, we assume that the electron after a collision flies out at the
right angle to the initial direction of the α-particle motion and has the momentum equal to
2me vα . Then, the result follows
∆pα 2me
θ− ∼ < ∼ 0.020 . (7.2)
pα mα
Notwithstanding the fact that the deviation of the incident α-particle caused by both the
atomic electron and the positive charged sphere is as low as the order of 0.020 , whether a
series of such deviations could give rise to a big scattering angle. Let us suppose, that an
average deviation about the angle θ ∼ 0.010 , caused either by the positive charged sphere
or by the electron, occurs under a transition through one atomic layer. Using the statistical
methods for obtaining the result of a sequence of random deviations, we find, that after
passing through the N = 104 atomic layers, the total average deviation is equal to
√
θ t = θ N ∼ 10 . (7.3)

Really, the experimentally measured average deviation represented about 10 . However,

some part of the α-particles were scattered through much bigger angles. For example, one
of 8000 α-particles deviated through angle θ ≥ 900 . The probability, that the α-particle
is subjected to summarized scattering through the angle larger than θ under the average
deviation θt is !
θ2
P (≥ θ) = exp − 2 . (7.4)
θt
For the Thomson atom model P (≥ 900 ) = exp (−8100) ∼ 10−3500 , that is, only one of 103500
α-particles can be scattered through the angle ≥ 900 that contradicts the experimental data.
To explain the α-particles scattering results Rutherford suggested the planetary atomic
model, which the essence is as follows. Practically all the atom mass is concentrated in its
nucleus that is located in the center and has the size of 10−13 –10−12 cm. The electrons are
rotating around the nucleus at the distance of the order of 10−8 cm. Soon they found out,
that the nucleus electric charge exactly equals the element number in the periodic table by
Mendeleev. In the beginning of 1913 this idea was introduced by the Dutch physicist Vander
Broek and some months later a Rutherford disciple G. Moseley produced its experimental
proof. Moseley performed a set of experiments to measure the X-ray spectrum for various
elements. It turned out that the X-ray wavelength systematically decreases while the atomic

© 2011 by Taylor and Francis Group, LLC

 106 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

number Z in the periodical table increases. Moseley got the conclusion that this regularity
is caused by increasing the atomic nucleus charge. The charge increases from atom to
atom by one electronic unit and the number of such units coincides with the number of the
element position in the Mendeleev table. Since the atom is electrically neutral, it means
that the total number of the electrons in the atom is equal to Z as well.
Then, in the Rutherford atom at the nucleus surface the electric field strength is > 1021
V/cm, which is almost eight orders higher than that at the atom surface. In Fig. 7.2
(R is the atom radius) the distribution of the electric field strength E+ (r) in the atomic
models by Thomson (Fig. 7.2(a)) and Rutherford (Fig. 7.2(b)) are showed for comparison.
It is obvious, that the strong field in the planetary model can cause a big deviation and

(a) (b)

FIGURE 7.2
The electric field strength in the Tomson atom model (a) and in the Rutherford atom model
(b).

even scatter the α-particle backwards, when it is flying too close to the nucleus. However,
the secret of the rapid development of physics lies in the fact that for confirming a physical
hypothesis not only qualitative, but also quantitative, coincidences are needed. Let us prove
that the Rutherford model correctly describes the scattering of α-particles by atoms.
To analyze both elastic and nonelastic collisions either the laboratory reference frame
(“lab” frame) or the center-of-mass frame is used. The lab frame corresponds to the standard
performance of experiments, namely, a beam of particle of type I strikes on a fixed target,
built up from particles of type II. In the center-of-mass frame, however, equations, describing
scattering processes, are much more simple, since the total system momentum equals zero
p1 + p2 = 0. In the center-of-mass frame particle collision is reduced to a motion of one
particle with a reduced mass
m1 m2
M=
m1 + m2
in the field U (r) of an immovable force center, located in a particles inertia center. In
the lab frame scattering angles θ1 and θ2 (the second particle rested before collision) are
connected with the scattering angle in the center-of-mass frame by the relations
π−χ m2 sin χ
θ2 = , tan θ1 = . (7.5)
2 m1 + m2 cos χ
Notice that the center-of-mass frame can be practically realized under the performance of
experiments with colliding beams.

© 2011 by Taylor and Francis Group, LLC

 Atoms–Nuclei–Nucleons 107

In classical physics the collision of two particles is completely defined by their speeds
and an impact parameter ρ. However, during real experiments we deal not with individual
deviation of a particle but with the scattering of a beam, consisting of identical particles,
striking on a scattering center at the same speed. Various particles in a beam have different
impact parameters and, consequently, are scattered through different angles χ. Let dN be
a number of particles, scattered in a unit of time through angles belonging to an interval
χ and χ + dχ. Since dN is a function of a falling beam density, that is not convenient to
characterize a scattering process. For this reason we use a quantity
dN
dσ = ,
n
where n is a number of particles passing in the time unit through the unit of area of a beam
cross section (we assume beam homogeneity over the whole section). In a given interval
of angles only those particles are scattered, which fly with impact parameters enclosed in
the interval between ρ(χ) and ρ(χ) + dρ(χ). The number of such particles is equal to the
product of n and the area of the ring between circles with the radii ρ(χ) and ρ(χ) + dρ(χ),
that is, dN = 2πρ(χ)dρ(χ)n. Thus, the effective cross section of scattering within the
interval of the flat angles dχ (it is also called the differential cross section) is defined by the
expression:
dρ(χ)
dσ = 2πρ(χ) | | dχ. (7.6)
dχ
Bearing in mind that the derivative dρ(χ)/dχ also could be negative, we used its absolute
value only. Passing to the solid angle dΩ, we obtain
ρ(χ) dρ(χ)
dσ = | | dΩ. (7.7)
sin χ dχ
To integrate the differential cross section over all values of the solid angle produces the
total cross section σ. The cross section has the dimension of the area: it is the “useful”
area of the interacting system, consisting of the incident particle and the target-particle.
Thus, for the incident particle a target is like an area, and the hit in this area results in an
interaction. The effective cross sections can be smaller or bigger than the geometrical cross
sections of target particles, and can coincide with them as well. To obtain in the lab frame
the scattering cross sections for the incident beam, one should express χ in terms of θ1 and
θ2 by means of Eqs. (7.5).
In quantum theory the scattering problem is considered from other grounds since the
conception of trajectories and impact parameters make no sense under the motion with a
definite speed. Here the aim of the theory is to calculate the probability, that in the result
of the collision the particles are scattered through one or another angle. Once again, we can
introduce the conception of the effective cross section, which characterizes the transition
probability of a system, consisting of two colliding particles, as a result of their elastic or
nonelastic scattering to the definite final state.
The differential cross section of scattering within angles interval dΩ is equal to the ratio
between the probability of such transitions Pi→f per the time unit to the incident particles
flux j0
Wi→f
dσ = dΩ,
j0
where Wi→f = Pi→f /∆t. Integration over the entire interval of the solid angle variation
gives the total cross section
Z 4π
Wi→f
σ= dΩ.
0 j0

© 2011 by Taylor and Francis Group, LLC

 108 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the CGS of units the square centimeter cm2 is the unit of the effective cross section.
This unit, however, is very large for the microworld and we use a unit having the order of
a geometrical cross section of a nucleus 1 barn = 10−26 cm2 .
Let us calculate the differential cross section of the α-particles scattering in the Born
approximation. In the center-of-mass frame the Born formula has the following form:
Z
M2 i
dσ = | U (r) exp ( qr)dr |2 dΩ, (7.8)
4π 2 ~4 ~

where dΩ = 2π sin χdχ, U (r) = Ze2 /2πr, Z is an atomic number of a target, q = p − p′ ,

p and p′ are the momenta of the particles before and after collision. Using the Poisson
equation for the potential of a point charge, located in the beginning of the coordinate
system e
∆ = −4πeδ(r),
r
and the Fourier transformation for the delta function
Z
1
δ(r) = exp (ikr)dk,
(2π)3

we obtain the following expression for the Fourier image of the potential energy
Z Z Z
2 1 i Ze2 dk (p − p′ )
U (q) = Ze exp[ (qr)]dr = exp{i[ + k]r}dr
2πr ~ 4π 3 k2 ~
Z
Ze2 (2π)3 1 p − p′ 2Ze2 ~2
= δ( + k)dk = .
4π 3 k2 ~ | p − p′ |2
Taking into account that
χ
| p − p′ |= 2p sin ,
2
where | p |=| p′ |= p, we arrive at the result:
 2
Ze2 M dΩ
dσ = . (7.9)
4πp2 sin4 (χ/2)

Notice some interesting peculiarities of the formula (7.9), which is called the Rutherford
formula. When we do not take into consideration any relativistic effects, then solving the
task exactly, we obtain the expression (7.9) as well (see, for example, [15]). The solution
obtained does not depend on the potential energy sign, that is, the solution is the same
for both attracting and repulsing force centers. Since the differential cross section does not
contain the Plank constant, then one can analyze the Rutherford scattering by classical or
quantum mechanical methods with the same success. The latter facts follow from the rule,
which reads:“if forces of an interaction between particles depend on a distance as rn , then
the cross section of scattering of such particles at each other is proportional to ~4+2n .”
If we try to integrate the expression (7.9) over the scattering angle, then we obtain
infinity. It is caused by the long-range character of the electrostatic forces. For this reason
the particles are scattered, no matter how far away from a scattering center they fly. To
obtain a reasonable, that is, a finite result, we must take into consideration a screening
effect of the electron shell. This is achieved under use of the potential

Ze2 r
U (r) = exp (− ),
2πr a

© 2011 by Taylor and Francis Group, LLC

 Atoms–Nuclei–Nucleons 109

where a has the value of the order of the atom radius.

Now we must take the last step, namely, to compare the theoretical formula (7.9) with
the experimental results. Let us turn to the lab frame and consider the cross section for the
incident α-particles, taking into account that mN ≫ mα (mN is a nuclear mass). In this
case θ ≡ θ1 ∼ χ and M ∼ mα , so that
 2
Ze2 dΩ
dσ = . (7.10)
4πmα vα2 sin4 θ2

In Fig. 7.3 the results of calculations according to the formula (7.10) are plotted, which
define the scattered particles number in relation to the angle θ at Z = 79. The experimental

FIGURE 7.3
The θ angle dependence of the number of the scattered α-particles.

data obtained by Geiger and Marsden are displayed by small triangles. Excellent agreement
between the theoretical and experimental results testifies that the internal structure of the
atom indeed includes the hard and massive core, which is the nucleus.
It should be stressed that the formula (7.10) is valid under fulfillment the following con-
ditions.

(i) The atom nucleus must have the mass exceeding the α-particle mass to such a de-
gree that the recoil energy of the nucleus may be neglected safely.
(ii) The nucleus potential consists of the Coulomb part and the part that is responsible for
nuclear attraction forces ϕN . For the α-particle not to be influenced by the nuclear forces
the collision diameter d, defining the minimal distance on which the incoming particle can
approach the scattering center, must greatly exceed the action radius of the potential ϕN .
In this case the collision diameter is determined by the help of the energy conservation law

mα vα2 Ze2
= . (7.11)
2 2πd
By a lucky chance both of these conditions has proved to be satisfied in experiments with
gold targets. However, in later experiments with aluminum targets (Z = 13) deviations
from the Rutherford formula were especially noticeable in the region of the large angles.

© 2011 by Taylor and Francis Group, LLC

 110 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Using the Rutherford formula, we can approximately estimate the size of the atom nu-
cleus. Let us assume, that the geometrical cross section of the nucleus is of the same order
as the differential scattering cross-section of the α-particles deflected by the angle being
greater than 900 . Then for Z = 79 and Eα = 7.68 MeV we have dσ/dΩ = 6.87 × 10−28
m2 , which produces the wholly plausible result R = 1.5 × 10−14 m. If the core were absent
in the atom center, that is, the Thomson model were true, then decreasing the α-particles
deflected by big angles θ would take place along the shortest curve, which is a straight
line (see Fig. 7.3). Thus, the excess of the differential cross section over the straight line
crossing the horizontal axis provides direct confirmation of the fact that atoms are not in-
divisible elements of matter, but represent in themselves composite structures, consisting
of the positive charged nucleus and the electrons.
Thus, according to Rutherford, the atom is similar to the solar system. The character
of resemblance to the solar system represents as not only qualitative but also quantitative,
which is very strange and has not found an exhaustive explanation till now. If one takes
the ratio between the diameters of the Sun and the solar system, then this ratio proves to
be approximately equal to one between the diameters of the nucleus and the atom. Further
on, according to quantum theory, the electrons in the atom are not located at arbitrary
distances from the nucleus. Their orbital radii are defined by the relation
4π~2 2
rn = n , n = 1, 2, 3...
me2
It turns out, that the planets have analogous behavior, namely, distances between the
planets and the Sun are not changed in a random manner but are subjected to a definite
law. This fact was known to J. Kepler, but it was first mathematically formulated by D.
Titius in 1772. Later on I. Bode made some corrections and the law was given the title the
Titius-Bode law. If the distance from the Sun to Mercury is adopted as an 0.4 arbitrary
unit, then the formula for the planet radii takes the form
Rn = 0.3n + 0.4,
where n = 0 (Mercury), 1 (Venus), 2 (Earth), 3 (Mars), 5 (Jupiter), 6 (Saturn), 7 (Uranus),
8 (Neptune), and 9 (Pluto). It is remarkable, that in this scheme a planet with the number
n = 4, which should be spaced between Mars and Jupiter, is missing. Exactly in this place
an asteroid belt is located. Astronomers suppose that this asteroid belt represents the
fragments of the planet Phaeton that existed in former times.
As a result of the triumphal success of the Rutherford atom model the traditional picture
of this model with its precise figure of the electron orbits becomes the universally recognized
emblem of the past twentieth century. It has appeared in books covers, exhibitions and
conferences booklets, and stamped papers of institutes and universities. Although nowadays
we know that in reality there are no electron orbits in classical understanding, these figures
have remained the deserved tribute of respect to the Rutherford atom model.
In the experiments described above Rutherford discovered not only the atomic nucleus
he also established the main type of experiment in the microworld physics, which is parti-
cles scattering at each other. By the example of the Rutherford experiments we see that
the scattering experiments are characterized by three time intervals. During the first in-
terval the systems are prepared and brought into contact. During the second interval an
interaction takes place. During the third interval the systems appeared in the result of an
interaction (reaction products) moving towards measuring devices (detectors). Since during
the first and the third intervals the systems are isolated from each other, then the systems
investigations are reduced to the analysis of free motion equations. The description of the
second interval demands deriving the solutions of the interacting systems equations and as
a result this is one of the most important and fundamental physics problems.

© 2011 by Taylor and Francis Group, LLC

 Atoms–Nuclei–Nucleons 111

7.3 Structure of the atomic nucleus

The discovery of the first isotopes in 1919 by F. Aston and establishing the integer numbers
rule (all atomic or molecular masses are integer numbers within the limits of observation pre-
cision) set in turn the question about the composite structure of the nucleus. The long ago
forgotten Proud hypothesis (1815), stating that hydrogen is the part of all atoms, became
popular once again. Analogously to the term “protil,” introduced by Proud, Rutherford
suggests using the term “proton” for the hydrogen atomic nucleus (1920). As far back as
World War I Rutherford started to investigate the collisions of the α-particles with the nu-
clei of such light elements as hydrogen, nitrogen, oxygen, and air. At the central collisions
the nuclei emitted particles, causing scintillations at a screen made of sulphurous zinc. At
that time a deep belief existed, that only nucleus components can be knocked out of the nu-
cleus. Due to a lucky coincidence it really took place in the experiments on the α-particles
scattering, because the α-particle energy Eα was not enough to produce even the lightest
hadron π 0 -meson (Eα < 135 MeV). Based on studying the behavior of knocked-out particles
in the magnetic field, in 1924 E. Rutherford and D. Chedvick presented in their paper the
final proof that those particles were protons. They also suggested, that the α-particle was
stuck in the nucleus. Using the Wilson camera, in 1925 P. Blackett made a photograph of
the reaction, which showed that the α-particle is captured by the nucleus and its track is
ended by the typical “fork,” in which the short fat track belongs to the residual nucleus
while the long thick track belongs to the knocked-out proton. The reaction, in which one
was shown for the first time that the proton enters into the composition of the nucleus
4 14
2 He + 7 N → 17
8 O + p,

was defined as an example of the first two-particle reaction.

Since nuclei masses are always larger than Z proton masses, then it is obvious, that some
more particles must enter into the composition of the nucleus. It is natural to assume that
the nucleus contains the electrons and that the protons exceeding the nuclear electrons cause
the total positive nuclear charge. However, this hypothesis, despite its charming simplicity
(no need to introduce new particles), ran into a sequence of serious contradictions. Let us
enumerate them.

(i) The protons and the electrons are fermions having the spin 1/2. Then, as follows from
the rule of summing up the spin moments vectors, the nuclei with an even number of the
protons np and the electrons ne must have an integer spin, while the nuclei with odd np and
ne must have a half-integer spin. For example, for a deuteron (the nucleus of a deuterium
atom) the spin should equal either 3/2 or 1/2 depending on the orientation of the electron
and proton spins. In reality the observed deuteron spin is equal to 1, which is in no way
reconciled with the hypothesis of the nuclear electrons.
(ii) If the electrons were present in the nucleus then the nuclear magnetic moment would be
of the Bohr magneton order (µB = e~/2me c). However, as experiments exhibit, the nuclear
magnetic moments are of the nuclear magneton order (µN = e~/2mp c), which is 1000 times
less than µB .
(iii) From the Heisenberg uncertainty relation

∆p · ∆x ≥ ~/2

it follows that in the nucleus (∆x ≈ 10−14 m) ∆p for the electron has of the order of
~/∆x ≈ 10−20 kg · m/sec. Since the momentum must be no less than this value, then for

© 2011 by Taylor and Francis Group, LLC

 112 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the energy of such an electron we obtain the value ∼ 20 MeV. The maximum energy of
the electrons emitted in the β-decay of various nuclides is in the interval 0.011–6.609 MeV
that is much less than the energy of the electrons inside of the nucleus, if the electrons were
inside of it.
In 1930 W. Bote and H. Becker obtained deeply penetrating radiation, which arises
under the α-particles bombardment of light elements nuclei 94 Be, 11 19
5 B, and 9 F . Since this
radiation went through a lead layer several centimeters thick and was not deflected by the
magnetic or electric fields, for several years unsuccessful attempts had been made to identify
it with high energy photons. In February, 27, 1932 I. Chadwick published an article in the
journal Nature [16], where he showed that the Bote-Becker penetrating radiation is a flux
of neutral particles having masses close to the proton mass (nowadays these particles are
called neutrons, n). After Chadwick’s article, events developed at a striking quickness. In
the same journal D. Ivanenko published a letter [17], in which he proposed the hypothesis
of a proton-neutron nuclear structure. Soon the proton-neutron model was also introduced
by W. Heisenberg in his article, published in the journal Zeitschrift für Physik (1932, June)
[18]. Due to numerous similarities of the proton and the neutron (neutron spin equals 1/2
and mn ≈ mp ) a common name “nucleon” is used for both of them.
In the nucleus the nucleons form not such a hard lattice as the atoms in a crystal do, but
rather a liquid structure in which the nucleons can move like molecules in a liquid. The
successful explanation of an element’s structure in the Mendeleev periodic table by means of
only three types of particles, namely, the protons and the neutrons contained in the nucleus,
which the electrons surround, provided all the reasons (at that time) to consider that e− ,
p, and n are the structureless matter blocks, of which all the universe is built. Establishing
this level of the matter structure was the third step of the “Quantum Stairway.” The list of
the “elementary” particles, known at that time, was rather short: electron, proton, neutron,
photon, and positron (the electron antiparticle that was discovered in 1932 by K. Anderson
in cosmic rays). So, the world used to be described by a fascinating simple scheme. One
could think, that the third step, reached with such difficulty, was the last one on the way to
the Creator. However, it was nothing else, but the illusion of an understanding that is like
Kipling’s cat who walks by herself and makes visits in accordance with her desire only.

© 2011 by Taylor and Francis Group, LLC

 8
From muon to gluon

The woodpecker was a humble scholar and lived a life

of strict seclusion. He never saw anyone (many even
thought that he drank as deep as all deep scholars), and
sat on a pine-branch all day, pecking away. He’d pecked
out a series of works on woodland history: “Goblin Gene-
alogy,” “Did Baba-Yaga the Witch Ever Wed?” “The Sex
of a Sorceress and the Census,” etc. But no matter how
much he pecked, he couldn’t find a publisher for his
peckings. And so he, too, finally made up his mind to
offer his services to the eagle as Courtyard Chronicler.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

In 1936 muons µ± , discovered by K. Anderson and S. Niedermayer in cosmic rays, were

added to the list of elementary particles. Before 1953, when the first accelerator was con-
structed (Brookhaven proton synchrotron with maximum energy 3 Gev), elementary particle
investigation were intimately connected with cosmic ray investigation. In 1947 the group by
S. Powell discovered π + and π − -mesons. The situation with elementary particles already
became not entirely simple. The electrons and the nucleons are necessary to build atoms.
The photons and the pions play the role of the carriers of the electromagnetic and nuclear
forces, respectively. An electron antiparticle, a positron, can be viewed as a delicate hint
(for the experienced mind) on the existence of antimatter searching, which should be con-
tinued until antinucleons are found. However, what does one do with muons, which do not
find their place in this world scheme?
In the late 1940s–early 1950s a real demographic explosion had occurred in the elementary
particle world. A whole zoo of new particles, called “strange” particles, had been discovered.
A main peculiarity of those particles and the ones, discovered later, is that they are not
the component of matter observed. They live for a very short time and decay into stable
particles (protons, electrons, photons, and neutrinos). First particles from this group, K + -
and K − -mesons, Λ-hyperons were discovered in cosmic rays, the next ones—in accelerators.
From the early 1950s accelerators become the main tool to investigate matter microparticles.
Accelerators’s energy is growing and the tendency for increasing the number of fundamental
particles becomes more and more apparent. Nowadays the list of elementary particles has
become tremendously large ∼ 400. Properties of discovered elementary particles prove to
be unusual in many respects. To describe them, characteristics taken from classical physics,
such as electric charge, mass, momentum, angular moment, and magnetic moment, proved
to be insufficient. It was necessary to introduce many new quantum numbers, having
no classical analogs, which we call internal quantum numbers. The first reason for their
introduction deals with additional degrees of freedom of elementary particles. The second
reason is caused by striving to explain the nonobservation of some “acceptable” reactions∗

∗ “Acceptable”
reactions are reactions that are allowed by the conservation laws of energy, momentum,
angular moment, and electric charge.

113
© 2011 by Taylor and Francis Group, LLC
 114 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

by means of the existence of an internal symmetry that leads to the conservation of a

corresponding charge. The latter circumstance carries out the generous dispensation of the
conserved charges to some groups of particles, according to the following principle: “dynamic
symmetry corresponds to a closed channel of an acceptable reaction.” It is possible, that a
reader, being experienced by any sort of direct and inverse theorems about the existence
and uniqueness of the solutions, will feel no deep satisfaction. As quieting reason, one may
call attention to aesthetic attractiveness of the symmetric approach in all spheres of our
life (it is hardly probable that the statue of Venus from Milos being deprived of symmetry
could find a place within the walls of the Louvre).
Let us proceed to the classification of known particles. Elementary particles are divided
into three categories:

(i) Hadrons, which participate in the strong, weak, and gravitation interactions. Being
electrically charged they participate in electromagnetic interaction too.
(ii) Leptons, which do not participate in the strong interaction.
(iii) Field quanta, which carry the strong, electromagnetic, and weak interactions.

The hadrons are divided into baryons with a half-integer spin and mesons with an inte-
ger spin. Maximum values of the spin for hadrons are as follows: 6 for the mesons a6 (2450)
and f6 (2510), and 11/2 for the baryons N (2600) and ∆(2420) (in parenthesis the hadrons
masses are given in MeV/c2 ). Electron (e), muon (µ), tay-lepton (τ ), and their neutrinos
(νe , νµ , ντ ), belong to the lepton class. They all have the spin 1/2. If one subtracts from
the total number of discovered particles 12 interaction carriers (8 gluons, W ± , Z, and the
photon) and 6 leptons, then the total number of the hadrons is obtained.
For the division of particles with the spin 1/2 into the leptons and the baryons to make
sense, transitions between these particle kinds must be impossible. For example, the neutron
must not decay into the electron-positron pair and the electron neutrino

n → e− + e+ + νe . (8.1)

In reality, this decay has been never observed. Let us introduce two quantum numbers,
namely, baryon B and lepton L charges connecting them with dynamic symmetries with
respect to global∗ gauge transformations. For the baryons B = 1, for the antibaryons
B = −1, for nonbaryons B = 0. In the case of the leptons, it is accepted to speak of
not a lepton charge, but a lepton flavor. One discriminates the total lepton flavor L and
individual lepton flavors Le , Lµ and Lτ . For e− (e+ )—Le = 1 (Le = −1), for µ− (µ+ ) —
Lµ = 1 (Lµ = −1), for τ − (τ + )—Lτ = 1 (Lτ = −1) and
X
L= Li .
i=e,µ,τ

All nonlepton particles have L = 0. By now no reactions with violation of either the total
or individual lepton flavors have been observed. However, there are no serious reasons
in support of the lepton flavor conservation law. Consequently, many vital electroweak
interaction theories predict the existence of processes, in which either the total or individual
lepton flavor is not conserved. Scientists are intensively searching for reactions, which can
help to establish upper limits on their cross sections. Some lepton decays going with the
violation of Li are given below:

µ− → e− + γ (4.9 × 10−11 ), (8.2)

∗ When the transformation parameters do not depend on coordinates the transformation is called global.

© 2011 by Taylor and Francis Group, LLC

 From muon to gluon 115

µ− → e− + e+ + e− (1.0 × 10−12 ), (8.3)

− − −6
τ →µ +γ (3.0 × 10 ), (8.4)
− − 0 −6
τ →e +π (3.7 × 10 ), (8.5)
where in brackets the upper bounds on their branchings∗ have been pointed. Comparison of
theoretical expressions for partial decay width with experimental ones results in establishing
the bounds on parameters of the theories in which these decays are allowed.
Despite the fact that the baryon charge conservation law ensures matter stability, its
correctness is subjected to question, too. Within the framework of some GUTs B is not a
conserved quantum number and that leads to proton instability. Some decay channels with
the fixed upper limit on the proton lifetime with respect to the given channel are as follows:

p → e+ + π 0 (5.5 × 1032 yrs), (8.6)

p → e+ + νl + νl′ (1.1 × 1031 yrs), (8.7)

− + + 30
p→e +µ +µ (6 × 10 yrs) (8.8)
Since the age of our universe is as short as ∼ 109 years, there are no special reasons for
inconsolable grief over possible proton instability.
There is an important difference between the electric charge conservation and internal
quantum numbers conservation. The electric charge is not a simple number, similar, for
example, to the baryon charge, which is ascribed to various particles. The electric charge
governs the system dynamics and is a source of the electromagnetic field in itself. The
interaction between charged particles is carried out by means of the electromagnetic field
whose quanta are nothing but the massless photons. Since with the help of the corresponding
devices the electric field is easily measured, then it is possible to measure the electric charge
of an object from a large distance, that is, without close contact with this object. Nothing
similar takes place with the baryon charge. There exists no baryonic field connected with
the baryon charge. For this reason it is impossible to measure the baryon charge (number)
of an object at some distance away.
In a consistent description of interacting systems, the above-mentioned difference between
the electric charge and “charges,” not being the sources of physical fields, is in a different
character of gauge transformations, responsible for the conservation laws. For a conserved
quantity to be a field source, the theory must be invariant under local gauge transformations.
Thus, the electric charge conservation corresponds to the invariance of the theory under the
following transformations:
Uem (1) = exp [ieα(x)]. (8.9)
Conserving the charges, not producing physical fields, are bound up with the invariance of
the theory under global gauge transformations

UN (1) = exp [iN αN ], (8.10)

where N = B, Le , Lµ , Lτ , .....
Let us proceed to the introduction of inexact internal quantum numbers, that is, numbers
that are already conserved but not in all interactions. Above we mentioned the so-called
strange particles that are produced by pairs (one or more) under colliding π-mesons with
nucleons. Since the production of these particles was caused by a strong interaction the
probability of their birth was large. However, they decayed into ordinary hadrons or leptons

∗ Branchingis the ratio between the widths of the given decay channel (partial decay width) and the total
decay width.

© 2011 by Taylor and Francis Group, LLC

 116 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

at the expense of only the weak interaction and as a result the probability of their decays
was very small. The uncommon behavior of these particles once again reminded physicists
of the famous phrase by F. Bacon: “The perfect beauty without touch of strangeness is not
available in the world.” When in 1954 at the Brookhaven Cosmotron these particles were
obtained for the first time, among other processes the following one was observed:

π− + p → Λ + K 0 .

The large value of its cross section indicated, that it is going on exclusively due to the strong
interaction. On the other hand, the long lifetimes (∼ 10−10 s) of the particles Λ and K 0
with respect to the decays

Λ → p + π− , K 0 → π+ + π−

testified that these decays are caused by the weak interaction. For some reasons Λ and K 0
decays into lighter hadrons are forbidden due to the strong interaction and as a result, they
live a long time. M. Gell-Mann and K. Nishijima independently from each other introduced
a new additive quantum number, a strangeness s. They postulated its conservation in the
strong and electromagnetic interactions and nonconservation in processes, caused by the
weak interaction (for the weak interaction |∆s| = 1). For already known hadrons s takes
the values -3, -2, -1, 0, 1. Further extension of the hadron sector demanded the introduction
of such quantum numbers as a charm (c) and a beauty (b). They are also additive numbers
and are conserved in the strong and electromagnetic interactions. In hadron decays due to
the weak interaction they vary according to the rule

|∆c| = |∆b| = 1.

Using characteristics introduced by us, we can divide known baryons and mesons into the
following families:

I. normal (s = c = b = 0) hadrons (nucleons, π-mesons);

II. strange (c = b = 0, s 6= 0) hadrons (Λ-, Σ-, Ξ- and Ω− -hyperons, K-mesons);
III. charmed (s = b = 0, c 6= 0) hadrons (Λ+ ±,0
c - and Σc - hyperons, D-mesons);
IV. beautiful (s = c = 0, b 6= 0) hadrons (Λb -baryons and B-mesons).

There are also mixtures of the last three families:

V. strange and charmed (b = 0, s 6= 0, c 6= 0) hadrons (Ξ+ 0
c - and Ωc -baryons, Ds -mesons);
VI. strange and beautiful (c = 0, s 6= 0, b 6= 0) hadrons (Ξb -baryons, Bs+ -mesons);
VII. charmed and beautiful (s = 0, c 6= 0, b 6= 0) hadrons (Bc -mesons).

From all the plurality of the elementary particles only eleven are stable. They are:
three neutrinos (νe , νµ , ντ ), three antineutrinos (ν e , ν µ , ν τ ), the photon, the electron, the
positron, the proton, and the antiproton. Other particles are unstable. Unstable particles
can be divided in two classes: metastable particles and resonances. Metastable particles
decay due to the weak or electromagnetic interactions, that is, they are tolerant to the
decay caused by the strong interaction. Normally these particles are included to the class
of stable particles. Particles-resonances decay predominantly due to the strong interaction
(there may be the channels caused by the electromagnetic and weak interactions, but these
channels are greatly suppressed). A typical resonance lifetime belongs to the interval 10−23 –
10−24 (this time is necessary for a relativistic particle to cover the distance of the order of
the hadron size ∼ 10−13 cm). Such short lifetimes do not enable one to register resonance

© 2011 by Taylor and Francis Group, LLC

 From muon to gluon 117

traces in track detectors. Resonances are not observed in a free state, they reveal themselves
while scattering in the form of quasi-stationary states of two or three strongly interacting
particles. They possess such particle characteristics as the spin, the electric charge and they
can be specified by internal quantum numbers, conserved in the strong interaction (isospin,
parity, hypercharge, etc.). Resonances, however, have no definite mass value, unlike stable
particles. They are described by a mass spectrum of a dispersion type, the maximum of this
spectrum is called the itresonance mass m. The resonance mass spectrum width Γ supplies
information about the probability of a resonance decay and must not exceed mc2 /2.
The nature of an unstable particle is the most transparent when one uses the concept
of a quasi-stationary state. Considering an unstable particle f we can write down its total
Hamiltonian H in the form:
H = Hf + Hd ,
where Hd is the part of the Hamiltonian, which is responsible for the decay. When neglect-
ing Hd , the particle becomes stable and, as this takes place, its states are eigenstates of
the operator Hf . In the case of the metastable particle, Hd contains the weak and electro-
magnetic interactions while in the case of the resonances it contains, in addition, the strong
interaction. With Hd , taken into account, the states of the particle f are quasi-stationary.
Let us find out the connection of the quasi-stationary state decay law with the function
of energy distribution or, what is one and the same, with the mass spectrum of this state.
Let Ψ(α, t = 0) be the initial state of a system. Here α denotes the plurality of variables,
according to which the system states are classified. Now we expand Ψ(α, t = 0) ≡ Ψ(α, 0)
in terms of the eigenfunctions of the energy operator ψ(α, E) (continuous spectrum)
Z
Ψ(α, 0) = a(E)ψ(α, E)dE. (8.11)

Then, at the time moment t, the state of a system is determined by the expression:
Z
iEt
Ψ(α, t) = exp ( )a(E)ψ(α, E)dE. (8.12)
~
From Eqs. (8.11) and (8.12) we obtain the probability of finding the system in the initial
state after the time t
Z 2  Z 2
   iEt 
W (t) =  Ψ† (α, 0)Ψ(α, t)dα =  exp ( )|a(E)|2 dE  =
   
   ~ 
Z 2
 iEt 
 
=  exp ( )w(E)dE  , (8.13)
 ~ 

where w(E)dE = |α(E)|2 dE is a function of the energy distribution for the initial Ψ(α, 0)
state and, consequently, for the final Ψ(α, t)) one. So, the decay probability of the state
Ψ(α, 0) is only defined by the function of the energy distribution in this state.
In the case of a resting unstable particle, the energy distribution dW (E) = w(E)dE is
nothing else, but a particle mass spectrum. The states ψ(α, E) therewith include the decay
product states as well. It becomes obvious from (8.13) that for a quasi-stationary state to
decay it is necessary and sufficient that the energy distribution integral function W (E) is
continuous. Thus, the discrete mass spectrum is excluded. Really, the decay probability
[1 − W (t)] tends to unity under t → ∞ only when W (E) is a continuous function.
To obtain the radioactive decay law, familiar to us from nuclear physics
Γt
W (t) = W0 exp (− ), (8.14)
~

© 2011 by Taylor and Francis Group, LLC

 118 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where Γ is a total decay width of an unstable particle f , it is enough to assume, that the
function a(E) has the form:
Γ/2
a(E) = , (8.15)
E − E0 + iΓ/2
where E0 = mf c2 . Indeed with the help of the residues theory we have
Z
Γ/2 iEt
W (t) =| exp ( )dE |2 =
E − E0 + iΓ/2 ~

2πiΓ Γt iE0 t 2 Γt
=| exp (− ) exp ( ) | = (πΓ)2 exp (− ). (8.16)
2 2~ ~ ~
Thus, from (8.15) it follows, that the mass distribution of a resting unstable particle has
the dispersion character

(Γ/2)2
w(mc2 )dm = dm. (8.17)
(mc2 − mf c2 )2 + Γ2 /4

Since unstable particles are characterized by the mass spectrum, then from the point
of view of the Poincare group, they must be described by reducible representations. But
the group theoretical definition of elementary character presupposes that the corresponding
representation must be irreducible, so, from this point of view, unstable particles are not
elementary particles. Before the quark structure of the hadrons was established, it was the
reason to give concern. Moreover, the situation been has aggravated by the fact that in all
of the known experiments, stable and unstable particles participate in interactions almost
in the same way. Consequently, we have no reason to consider unstable particles to be less
fundamental objects compared to stable particles. Moreover, if all hadrons (resonances and
stable particles) are entered on a diagram, where the spin J and the mass square m2 are
plotted along the axes, then the states with the same B, I, and S are located on straight
lines (Fig. 8.1) which are called Regge trajectories.∗ Regge [19], Chew and Frautschi [20]
(see, for review, [21] as well) developed a strong interaction theory, in which hadrons are
compared with poles of a scattering amplitude in a complex plane of spin values. These
poles are called Regge poles. Regge trajectories are simply showing that the spin value is
a function of the energy or, which is equivalent, of the mass J(m2 ). On every trajectory
hadrons appear with an interval ∆J = 2. Stable hadrons are the lowest located state on the
Regge trajectories and that constitutes the only difference between them and resonances.
Notice that just the striking linearity of the Regge trajectories detected at experiments
gave impetus to creating the string models of hadrons. In these models the hadrons are
considered as stretched objects, strings, the quantization of which results in the appearance
of a particles sequence to be placed at linearly growing Regge trajectories.
Nowadays the difference between the physical and group theoretical concepts of the ele-
mentary character for resonances must be accepted with philosophical calmness since they
are not structureless formations but made of the quarks.
If one neglects the interaction causing a decay, then formalism of relativistic quantum
theory can be generalized to unstable states. Such an approximation is necessary to describe
unstable particles scattering. As a rule, in scattering processes, particles in the initial and
final states are not supposed to interact, that can be realized when they are separated at
infinitely great distances from each other. In this case, one does not follow to forget that
in reality unstable particles have been decayed long before asymptotical separation was
achieved.

∗ The graph is called the Chew–Frautschi diagram.

© 2011 by Taylor and Francis Group, LLC

 From muon to gluon 119
y

ry
J or

to
ct

jec
e
r aj

tra
15 2
/ -t

L-
N
13 2
/ N(2600)
11 2
/
9/ 2
7/ 2 L(2100)

5/ 2 N(2190)
N(1520)
3/ 2 L(1690)
1/ 2

0 2 4 6 8 10
m2, GeV 2

FIGURE 8.1
The Regge trajectories for baryon states with the negative parity.

Basic methods of resonances detection are based on the fact that resonances have the mass
spectrum of the dispersion type. The first method deals with investigating the maxima in
the total scattering cross section. To make it definite, let us assume, that we deal with the
reactions:
a + b → Y → a + b, (8.18)
d + f → Y → d + f, (8.19)
a + b → Y → d + f, (8.20)
where for Y both decay channels are possible

Y → a + b, Y → d + f. (8.21)

Due to fulfillment of Eq. (8.21), these reactions are going through a s-channel.∗ The exis-
tence of such s-channel diagrams is a necessary condition for the resonance to be observed.
This resonance peak, connected with Y , appears in all the above-mentioned reactions. If
one presumes, that the reaction (8.19) is going on only by means of the production of the
resonance-particle Y in a virtual state (the s-channel is only one of the reactions), then
the total cross section of the elastic df -scattering as a function of the energy E near the
resonance is defined by the Breit-Wigner formula† [22]:

(Γ/2)2
σ(E) = σ0 , (8.22)
(E − E0 )2 + Γ2 /4

As we can see, this expression coincides with the masses distribution w(E) to an accuracy
of the kinematic factor. The energy E0 , corresponding to the cross section maximum

∗ A s-channel diagram is the Feynman diagram, where annihilation of initial particles takes place at one

space-time point, while final particles production takes place at the other one.
† If other reaction channels are present, the possibility is not eliminated that even at resonance the contri-

bution into the cross section coming from the s-channel diagram proves to be smaller (or the same order)
as compared with one of the remaining diagrams. As a result the resonance peak in the cross section of the
reaction in question is absent.

© 2011 by Taylor and Francis Group, LLC

 120 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

σ(E) = σ0 being divided by c2 , defines what we agreed to consider the resonance mass.
The maximum width informs us about the resonance decay probability. In this case the
particles in the final state appear with retardation ∆t ∼ ~/Γ as compared to scattering
without the resonance production. The main drawback of this method is that it does not
allow us to determine resonance quantum numbers completely.
The next method is the phase analysis method that is more universal because with its
help, it is possible to define all the resonance characteristics (mass, width, spin, parity,
isotopic spin, and so on). The method is based on measuring the differential cross section
of elastic scattering dσ = 2π|f (θ, E)|2 sin θdθ (θ is a scattering angle). If particles having
the spin participate in scattering, then the scattering amplitude f (θ, E) is expanded as a
series in the spherical functions Yml (θ, ϕ). For spinless particles this expansion has the form:
Xp X
f (θ, E) = 4π(2l + 1)fl (E)Yl0 (θ) = (2l + 1)fl (E)Pl (θ), (8.23)
l l

where the coefficients fl (E) are partial waves of scattering with the moment l, which are
determined from experimental data as complex functions of E. The resonance with the spin
J=l reveals itself as the Breit-Wigner contribution to fl (E). If the spins of two particles are
equal to 0 and 1/2 respectively, then instead of Eq. (8.23) we have
r r
Xp l+1 1/2 l 1/2
f (θ, E) = 4π(2l + 1)[ f 1 (E)Yl+ 1 (θ, ϕ) − f 1 (E)Yl− 1 (θ, ϕ)].
2l + 1 l+ 2 2 2l + 1 l− 2 2
l
(8.24)
In the case when there are three or more particles in the final state, the method of
maxima in mass distributions is used to search the resonances. Let us consider the inelastic
scattering reaction
a + b → c1 + ... + Y → c1 + ... + d + f, (8.25)
where a resonance Y decays into stable particles d and f . During such scattering, the
momentum and the energy of particles a and b are distributed between groups of particles
c1 + ... and d + f in the final state. If the resonance Y were the stable particle, then in its
proper system (rest system) its energy, and the mass×c2 as well, would have the definite
value. But the resonance is unstable and is specified by the distribution function w(E),
which in the particle rest system is directly connected with the mass spectrum of particle
decay products. In other words, the total set of states in (8.24) consists of two-particle
states ψdf (E) to have the same quantum numbers as does the Y -particle.∗ In distribution
of the invariant mass square of the particles d and f
2 4
Mdf c = (Ed + Ef )2 − (pd + pf )2 c2 , (8.26)

the brightly expressed maximum will be observed. Thus, studying the distribution of the
masses in complexes of particles belonging in the final states, one may obtain directly the
distributions on the masses of unstable particles, which are the resonance states in such
complexes. Of course, it does not necessarily mean that every peak in the mass distri-
butions of the particles being the reaction products can be identified with the resonance,
because kinematic peaks also occur, which are inherent in the given reaction only.† The
difference between the real resonance and the ghost peak is that the energy, corresponding
appearance of the real resonance in different experiments, is always the same, while the

∗ Ifthe decay channel Y → d+f is not the only one, and there exists another one, for example Y → k +m+l,
then three-particle states ψkml (E) will enter into Eq. (8.24) and so on.
† Such peaks are called ghost ones.

© 2011 by Taylor and Francis Group, LLC

 From muon to gluon 121

energy, connected with the ghost peak, is changed from one experiment to another. The
distribution of the invariant mass square in the system Σ+ π − , arising in the reaction
π − + p → Σ+ + K 0 + π − (8.27)
at the momenta of the incident π − -meson from 2.2 to 2.4 GeV/c is shown in Fig. 8.2 (N is
a number of events on (50 MeV)2 . Three peaks are distinguished against the background

FIGURE 8.2
The distribution of the invariant mass square in the Σ+ π − -system. The dashed line repre-
sents the expected distribution defined by the relativistic phase volume only.

of uncorrelated events. All of them turn out to be the real resonances.

First resonances (∆-resonances) were discovered in 1952 by E. Fermi under scattering of
the π-mesons by protons
π + p → ∆ → π + p. (8.28)
The total cross section of the pion-nucleon scattering versus the center-of-mass frame energy
is displayed in Fig. 8.3. The solid line corresponds to σπ+ p , while the dashed one describes
σπ− p . As it follows from Fig. 8.3, at the π-mesons kinetic energy T = 195 MeV the
resonance peak appears both in π + p-scattering (∆++ -resonance) and in π − p-scattering
(∆0 -resonance). From the Gell-Mann-Nishijima formula
(is) B
Q = S3 +
2
it is evident that when the ∆++ - and ∆0 -resonance enter one and the same isotopic multiplet
they have S (is) = 3/2.
Further, by the example of the reaction (8.28), we show how to define such resonance
characteristics as the mass and the isotopic spin from the experimental data. In the lab
frame the proton rests, and consequently, the three-dimensional momentum and the energy
of the ∆-resonance are defined by:
1p r
p∆ = pπ = (Tπ + mπ c2 )2 − m2π c4 , (8.29)
c

© 2011 by Taylor and Francis Group, LLC

 122 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 8.3
The total cross sections of the π + p- and π − p- scattering.

E∆ = Tπr + mπ c2 + mp c2 . (8.30)
From Eqs. (8.29) and (8.30) it follows
q
1 2 − p2 c2 = 1236 MeV/c2
m∆ = 2 E∆ ∆
c
(contemporary data lessen down the ∆ mass value to 1232 MeV/c2 ).
To find the isotopic spin value we use the isotopic invariance of the cross sections of the
reactions (8.28). Since the isotopic spins of the nucleon and the π-meson are equal to 1/2
and 1 respectively, their sum could be equal either to 3/2 or 1/2. In order to do it we need
the pion-nucleon states expansion in terms of the isotopic spin eigenstates. This expansion
is given by the formulae (9.82), which we shall deduce later. Then, from the isotopic spin
conservation law it follows that only the following scattering amplitudes exist:

(Ψfπ+ p , Ψiπ+ p ) = (Ψf3/2,3/2 , Ψi3/2,3/2 ) = A3/2 , (8.31)

1 2
(Ψfπ− p , Ψiπ− p ) = A3/2 + A1/2 , (8.32)
3 3
√ √
2 2
(Ψfπ0 n , Ψiπ− p ) = A3/2 − A1/2 . (8.33)
3 3
The total cross sections of scattering in the region, where the multiple π-mesons production
is insignificant, are given by:
σπ+ = ρ | A3/2 |2 ,
 
1 2
σπ− = σπ− p→π− p + σπ− p→π0 n = ρ | A3/2 |2 + | A1/2 |2 , (8.34)
3 3
where ρ is a kinematic factor (which is constant for all the three processes, if we neglect
masses difference in isomultiplets). If we assume, that the isotopic spin of the ∆-resonance

© 2011 by Taylor and Francis Group, LLC

 From muon to gluon 123

is equal to 3/2, then the second term in Eq. (8.34) at E = Tπr goes to zero and we arrive
at the relation:
(σπ+ /σπ− ) |E=Tπr = 3,
which is in excellent accord with the experiment.
However, looking at this megapolis of the hadrons, we understand, that the above-
mentioned division is nothing else, but a scheme of streets and squares that does not bring us
closer to understanding the idea of the architect creating this elementary particles Babylon.
Moreover, this picture is far from perfection and we do not experience reverential delight
under a sight on it. Let us be reminded of the guiding thread, which brought us from the
first step of the Quantum Stairway to the third one. The Mendeleev periodic table was
built on the basis of an accidentally discovered periodic alteration of chemical properties
of elements alongside with increasing their atomic masses (nucleus electric charge, to be
exact). Sixty three years later after the creation of this table, with its help, we managed to
construct hundreds of elements from only three fundamental (as it seemed then) particles
p, n, and e− . Let us concentrate our efforts to find a similar table for the hadrons.

© 2011 by Taylor and Francis Group, LLC

 9
Hadron families

The bullfinch was the biggest success. By way of salu-

tation he read an article from the morning paper so
fluently that even the eagle thought he understood it.
Take life easy, the bullfinch read, and the eagle nodded:
jusso! All that he ever wanted, said the bullfinch,
was that his paper should sell well, and the eagle
repeated: jusso! The flunkey’s life is better than his
master’s, he chirped; the master has worlds of care,
the flunkey none–the good master looks after them.
Jusso! again said the eagle. They say that when he
had a conscience he had no trousers, but now that
it is dead he wears two pairs at once. And again the
eagle came in with his jusso!
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

9.1 Yukawa hypothesis

In the language of quantum field theory, force field influence on a particle is interpreted
as emission and absorption of the quanta of this field. For example, the electromagnetic
interaction between two electrons appears due to photons exchange. Photon masslessness
reveals itself in the long-range action of the Coulomb forces, that is, the Coulomb potential
decreases with a distance as 1/r. It is natural to expect that the massive carrier of an
interaction will produce a short-range potential. In an analogy with the exchange nature
of electromagnetic forces, in 1935 H. Yukawa [23] introduced a hypothesis about massive
quantum exchange, between nucleons in a nucleus. The idea was, that the interaction
between this massive quantum (π-meson) and nucleon N appears as a result of combination
of virtual process
N ↔ N ′ + π. (9.1)

In other words, the Hamiltonian, causing the interaction between nucleons in a nucleus Hint
must be of the form:
Hint ∼ N (x)N (x)π(x). (9.2)

Connections of this kind often occur in elementary particle physics and they are called
Yukawa couplings.
Now we should more carefully consider the concept of virtual states. Physical laws de-
scribe only experimentally measurable quantities. In quantum theory the uncertainty rela-
tions for dynamical observables, represented by noncommutating operators, impose restric-
tions on a precision of simultaneous measurements of these quantities. Thus, the Heisenberg

125
© 2011 by Taylor and Francis Group, LLC
 126 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

uncertainty relations impose precision limits on simultaneous measurements of particle co-

ordinates and momenta
~
∆pi ∆xi ≥ , (i = 1, 2, 3), (9.3)
2
as well as its energy and time
~
∆E∆t ≥ . (9.4)
2
The uncertainty relation (9.4) can be interpreted by the way in which it is somewhat unusual
from the viewpoint of classical physics. In the microworld the processes, going on with the
violation of the energy conservation law on the value of ∆E are allowed, provided that this
violation lasts no longer than ∆t ∼ ~/∆E. In this case the violation cannot be registered
by any physical devices. Analogously, the momentum conservation law could be violated
on the value ∆p in the region ∆x ∼ ~/∆p. We emphasize that the uncertainty relations
reflect the inner nature of the microworld and has nothing to do with the imperfection of
our measuring devices.
When considering interactions between particles by means of field quanta exchange, one
naturally thinks of ∆E to denote brought in or taken away quantum energy and ∆t to
denote the exchange duration or the lifetime of this quantum. The particles states with
such lifetimes are called the virtual states. All the elementary particles can occur both
in real and virtual states. Inherent for real particles the connection between energy E,
momentum p and mass m
E 2 = p2 c2 + m2 c4 (9.5)
is violated in the case of the virtual particles due to the appearance of ∆E. Keeping in
mind the violation of this equality, they say the virtual particles lay beyond the mass shell.
So, according to the Yukawa hypothesis, the interaction between nucleons in the nucleus is
carried out by means of the virtual particle exchange. Emitting and absorbing the virtual
π-mesons, protons, and neutrons turn into each other. It is easy to guess that for p ↔ n
coupling to exist, π + and π − -mesons are necessary. Being based on the experimentally
determined principle of the nuclear forces charge independence

VN = Vpp = Vnn = Vnp , (9.6)

where VN is nucleons interaction potential, one might conclude that a neutral π 0 -meson is
needed to describe p ↔ p and n ↔ n interactions. Of course, p ↔ p and n ↔ n interactions
can take place due to two charged π-mesons as well. Thus, for example, the interaction
between two protons takes place as follows. Both protons emit one π + -meson each, in
so doing they turn into neutrons. Then, these neutrons absorb π + -mesons and turn into
protons again. A corresponding mutual conversion chain for each proton has the form:

p → π + + n, n + π + → p, (9.7)

and p ↔ p interaction on its own is displayed by the reaction:

p + p → n + π + + n + π + → p + p. (9.8)

It is quite obvious that emission of two π-mesons necessitates the larger value of ∆E and
consequently, lessens ∆t. In a shorter lifetime π-mesons cover a shorter distance and that
reduces (approximately twice) the action radius of the nuclear forces between identical
nucleons, and, as a result, the condition (9.6) will be violated.
Let us estimate the π-meson mass by means of the uncertainty relation (9.4). We consider
the interaction between a proton and a neutron. The proton emits the π + -meson and turns

© 2011 by Taylor and Francis Group, LLC

 Hadron families 127

into the neutron, while the initial neutron having absorbed the π + -meson becomes the
proton
p → n + π+ , n + π + → p. (9.9)
The chain of the reactions (9.9) can be represented as follows:
π+
p + n −→ p + n. (9.10)
If one neglects the proton recoil momentum, then the emission of the virtual π + -meson
leads to the energy violation by the value ∼ mπ c2 at the minimum. The virtual meson
exists during the time ∆t ∼ ~/∆E ∼ ~/(mπ c2 ). In ∆t time the virtual ∆t-meson, even if
it moves at a maximum possible speed c, covers the distance, which defines the maximum
value of the force field action radius
~c ~
R ∼ c∆t ∼ 2
∼ . (9.11)
mπ c mπ c
Substituting in Eq. (9.11) the experimentally obtained value of the nuclear force action
radius Rn ∼ 10−13 , we obtain mπ ≈ 280 me (me is the electron mass).
In reality, Yukawa predicted, that the mass of the nuclear interaction carrier might be
equal to 206 me . That was caused by using the improper data concerning the value of Rn .
Being guided by this number, experimentalists began to search and have found the particle
with the mass 207 me , which was called a muon (µ). However, the muons turned out to be
weakly interacting particles. Consequently, they are not suitable for the role of the nuclear
field quantum. As late as 1947, predicted by Yukawa π-mesons were discovered in cosmic
rays.
The notion about the interaction law, caused by the π-meson exchange, is given by the the
Yukawa potential, that is, the potential energy that is derived under the assumption that
the interacting particles are immovable (particles masses are so large that one may precisely
fix their positions and neglects their recoils under emitting and absorbing quanta). Since
the π-mesons have zero spin, their behavior is described by the Gordon-Klein equation. In
a free case this equation can be obtained from the relativistic relation between the four-
dimensional momentum and the mass (9.5) by substitution:
∂ ∂
E → i~ , p → −i~ ,
∂t ∂r
which gives  
1 ∂2 m2π c2

− ∆ + ψ(r, t) =  + k02 ψ(r, t) = 0, (9.12)
c2 ∂t2 ~2
where k0 = mπ c/~ = λ−1 π and λπ is the Compton wavelength of the π-meson. Under
switching on an interaction, a source density of the π-meson field must stand in the right-
hand side of Eq. (9.12). Since nucleons, whose density is designated by ρ(r, t), provide such
a source, then Eq. (9.12) takes the form:


 + k02 ψ(r, t) = ρ(r, t). (9.13)
In case of a static field (9.13) turns into


∆ − k02 ψ(r) = −ρ(r). (9.14)
The Green functions method must be used to write down the solution of Eq. (9.14)
without specifying its right-hand side form. The solution will look like
Z
ψ(r) = G(r − r′ )ρ(r′ )dr′ , (9.15)

© 2011 by Taylor and Francis Group, LLC

 128 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where G(r) is the Green function, satisfying the equation:

∆ − k02 G(r) = −δ(r). (9.16)

Inserting in Eq. (9.16) the Fourier expansions of the delta-function

Z
1
δ(r) = exp (ip · r)dp (9.17)
(2π)3

and the Green function

Z
1
G(r) = G(p) exp (ip · r)dp, (9.18)
(2π)3

we obtain the following expression for the Fourier image of the Green function:
1
G(p) = . (9.19)
p2 + k02

To calculate the Green function in the expression (9.18) it is convenient to turn to the
spherical coordinate system with the axis z oriented along p. Then we arrive at
Z ∞ Z π Z 2π
1 p2
G(r) = dp sin θ exp [ipr cos θ]dθ dϕ =
(2π)3 0 p + k02
2
0 0
Z ∞ Z ∞
1 p sin(pr) 1 p exp [ipr]
= dp = 2 dp, (9.20)
2π 2 0 r(p2 + k02 ) 4π i −∞ r(p2 + k02 )
where p = |p|, r = |r|. The remaining integral over p is most convenient to calculate by
the method of the residues theory. The function under the integral sign is an analytical
function of p, which exponentially decreases in the upper half-plane and has poles in points
±ik0 . Consequently, the integral is equal to the residue in the point p = +ik0 multiplied
by 2πi. Thus, the Green function is equal to
1
G(r) = exp [−k0 |r|] (9.21)
4π|r|

and the solution of Eq. (9.16) takes the form:

Z
1 exp [−k0 |r − r′ |] ′ ′
ψ(r) = ρ(r )dr . (9.22)
4π |r − r′ |

Notice, in particular, that at mπ = 0 the expression (9.22) turns into a very familiar
expression for the electromagnetic potential, created by given charges distribution ρ(r′ ). As
it is known the energy of the system under consideration is described by the Hamiltonian
function H Z
H = Hdr,

where H is the Hamiltonian function density. Let us obtain H by using its connection with
the Lagrangian L. It is easy to check, that the Lagrange-Eulerian equations lead to Eq.
(9.15) if L is defined by:
 
1 ∂ψ(x) ∂ψ(x) 2
L= − k0 ψ(x)ψ(x) + ρ(x)ψ(x), (9.23)
2 ∂xµ ∂xµ

© 2011 by Taylor and Francis Group, LLC

 Hadron families 129

where we do not specify a π-meson field in question, since the theory is charge independent.
Then, applying the usual rules, we obtain the following expression for H:
Z  2  2
1 ∂ψ(x) ∂ψ(x)
H= [ + + k02 ψ(x)2 − 2ρ(x)ψ(x)]dr. (9.24)
2 ∂x0 ∂xk
Passing to a stationary case, taking into account the three-dimensional Gauss theorem and
using Eq. (9.22), we arrive at
Z  2 Z
1 ∂ψ(r) 1
H= [ + k02 ψ(r)2 − 2ρ(r)ψ(r)]dr = {ψ(r)[−∆ + k02 ]ψ(r) − 2ρ(r)ψ(r)}dr =
2 ∂xk 2
Z Z Z
1 1 ρ(r)ρ(r′ )
=− ψ(r)ρ(r)dr = − dr exp (−k0 |r − r′ |)dr′ . (9.25)
2 8π |r − r′ |
At mπ = 0 the expression obtained is nothing more nor less than the electrostatic interaction
energy of two charge distributions. Then it follows from Eq. (9.25) that the interaction
potential of two nucleons separated by the distance r is determined by the equation:
C
VN (r) = − exp (−k0 |r|), (9.26)
4π|r|
where C is a constant connected with the nucleon “nuclear” charge. From the expression
for VN (r) follows the same result as from the uncertainty relation: the nuclear forces have
the finite action radius, approximately equal to the Compton wave length of the π-meson.
Notice, that the negative sign in the expression for the Yukawa potential, testifying the
attraction character of the nuclear forces, is caused by the choice of the sign in front of the
nucleon field density function ρ(r). One should emphasize, that the above-given discussion
is classical in its essence. However, it is possible to show, that the same result will be
obtained, when the π-mesons will be considered in the quantum mechanic way, while the
nucleons will be described by the classical function of the source as before.
The process of emitting and absorbing the π-meson lasts no longer than 10−23 s. In all
modern experiments such a process may be considered as an instant one. Roughly speaking,
the protons spend one part of their life in a nucleus being the protons, while during the
second part they are the neutrons. So, it is natural to consider the proton and the neutron
as two different states of the same particle given the title nucleon. All nucleons consist of
identical cores, surrounded by a cloud of virtual π-mesons. The only difference between p
and n lies in the character of such a cloud. When two such clouds are approaching each other
at the distance of the order of the π-meson Compton wave length, the π-meson exchange
between clouds takes place.∗ . So the π-mesons are constantly scurrying between interacting
nucleons. The situation partly reminds of the covalent coupling carried out by the electrons
in the molecular ion H2+ . Since, in this case, the electron can transfer from one to another
proton, the exchange forces appear and they are added to the ordinary Coulomb forces.
Obviously, the π-meson clouds (fur coat), surrounding neutrons and protons, contribute
to nucleon magnetic moments. Since such virtual conversion chains

p → π + + n, n → π− + p

are possible for a proton and a neutron, then anomalous parts of magnetic moments of the
particles, caused by the π-meson field, must be approximately equal in value and opposite
in sign.

∗ Atsmaller distances exchanging the heavier particles (vector mesons ρ , ϕ, ω, etc.) begins to be more
substantial.

© 2011 by Taylor and Francis Group, LLC

 130 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Let us consider the process of originating the neutron magnetic moment due to its virtual
dissociation into π − and p. Since the π-meson has the zero-spin, it also has the zero intrinsic
magnetic moment. Then only the π-meson with a nonzero orbital moment, for example,
in p-state (l = 1), contributes to the neutron magnetic moment. For the angular moment
conservation law in this virtual process to be fulfilled, the following demands must be met.

(i) The direction of the orbital moment of the virtual π-meson being equal to 1, must
coincide with the neutron spin direction.
(ii) The virtual proton spin must be directed opposite to the neutron spin.

Since the π − -meson is negatively charged, then the neutron magnetic moment induced
by the π − -meson, is negative as well. To estimate the neutron anomalous magnetic mo-
ment value, we must know how long the neutron exists in the dissociated state or, what
amounts to the same thing, the transition probability into this state. Since mp ≈ 6.72mπ ,
the magnetic moment of the system p + π − is in value of the order equal to (−6.72 + 1)µN
(µN = e~/(2mp c)). Then the neutron magnetic moment observed is

µn = W0 µsn − 5.72(1 − W0 )µN , (9.27)

where µsn is the neutron intrinsic magnetic moment, that is, the quantity to be equal to zero,
W0 is the probability that the neutron will be found in the naked neutron state, (1 − W0 ) is
the probability that the neutron will be found in the dissociated p+ π − -state. Experimental
value of the neutron magnetic moment µn = −1.913 µN is obtained if one sets W0 ≈ 0.665.
Analogously, for the proton anomalous magnetic moment we have

µp = W0 µN + 6.72(1 − W0 )µN , (9.28)

where we have taken into consideration that the intrinsic magnetic moment of the proton is
equal to µN and assumed that the probabilities of the proton and neutron virtual dissocia-
tions are equal. Then the total proton magnetic moment is pretty close to its experimentally
measured value µp = 2.793 µN .
The experimentally proved Yukawa theory of the nuclear forces is a milestone in the
development of physics. It has finally strengthened the assurance that the quantum in-
terpretation of the interaction as an exchange of virtual quanta is a correct one. Such an
interaction interpretation underlies the foundation of modern physical theories. Taking into
consideration virtual particles changes our concepts of the physical vacuum under transition
from the classical to quantum theory. The example of the electrodynamics is very signifi-
cant in this case. The electromagnetic field in the classical theory is defined by the values
of the strengths of the electric and magnetic fields (E and H) given in all points of space
and in all moments of time. Under transition to the quantum electrodynamics in places
of those strengths operators appear which, in particular, do not commute with operators,
defining the number of photons in a given state. However, only physical quantities, to which
commuting operators correspond, can simultaneously have definite values. If operators do
not commute, then the more precisely the quantity corresponding to one of these operators
is defined, the less information can be obtained for the second quantity. At the exact def-
inition of E and/or H the number of photons is absolutely undefined. In the same way, if
the number of photons is exactly defined, then field strengths are not defined.
In quantum theory we determine a vacuum as a state without real particles or as a state
with the least energy. Then, since for the photon vacuum (electromagnetic vacuum) the
number of particles is zero, that is, is exactly defined, the field strengths are not defined, and
this fact, in particular, does not accept these strengths being equal to zero. Impossibility
to simultaneously set both the field strengths and the photons number equal to zero make

© 2011 by Taylor and Francis Group, LLC

 Hadron families 131

us consider the vacuum state in quantum theory not as the field absence, but as one of
the possible field states having the definite properties that are displayed in real physical
processes.
Virtual production and absorption of the photons should be viewed as the manifestation of
the photon vacuum, or, in other words, as taking account of the photon vacuum effects. The
concept of a vacuum being the lowest field energy state can be analogously introduced also
for other particles. Considering interacting fields, the lowest energy state of all the systems
could be called a vacuum state. If sufficient energy is supplied to a field in a vacuum state,
then the field is exited, that is, the field quantum is produced. Thus, particle production
can be described as a transition from an “unobserved” vacuum state to a real one. Without
real particles and external fields a vacuum, as a rule, does not reveal itself through any
phenomena due to its isotropy. The presence of real particles and/or external fields leads
to the vacuum isotropy violation, since production of virtual particles and their subsequent
absorption results in changes of the state of the real physical system. Virtual particle
production and absorption of particles is limited by conservation laws, the electric charge
conservation law among them. For this reason virtual production of a charged particle is
impossible without charge change of real particles (if there are any). If real particle charges
are invariable, then in virtual processes charged particles are created and destroyed in pairs
only (particle-antiparticle). Thus, in the case of charged particles one can speak only of
the particles-antiparticles vacuum: electron-positron vacuum, proton-antiproton vacuum,
etc. From this it also follows, that since, for example, positrons and electrons can be
produced only in pairs, one cannot speak of electrons as isolated and solitary types of
matter, just as one impossibly draws a demarcation line between the electric and magnetic
fields. The electron and positron fields make up the unified electron-positron field, and
this circumstance remains imperceptible, provided the processes of pairs productions and
pairs destructions may be neglected. Analogously, as in the case of the photon-antiproton
vacuum, the electron-positron vacuum or any other particles-antiparticles vacuums will lead
to observable effects, one of which is a change in the physical properties of particles. The
above-mentioned effects of the charge screening and appearance of the nucleons anomalous
magnetic moments can serve as an example.
So, in quantum theory every particle is enclosed by the fur coat consisting of clouds of
virtual quanta, produced and subsequently absorbed by the particle. Quanta can belong to
any field (electromagnetic, electron-positron, meson, etc.), with which the particle is inter-
acting. The fur coat contains many layers with different densities. For example, since the
meson interactions of nucleons are a hundred times more intensive than the electromagnetic
ones, the meson fur coat of the proton should be several orders thicker than the electromag-
netic one. The fur coat is not something like a solid state, since quanta, its components, are
continuously produced and annihilated. One can say, that in the quantum theory a particle
is suffering from striptease-mania, since one part of its lifetime it spent in the dressed state,
while during the rest of the time it is naked.

9.2 Isotopic multiplets

Experimental confirmation of the Yukawa hypothesis has one more important consequence.
So nuclear forces, acting between proton and neutron, proton and proton as well as between
two neutrons, are practically the same (of course, particles are in the same states). In other
words, the peculiarity of the interaction between nucleons is that switching off the proton

© 2011 by Taylor and Francis Group, LLC

 132 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

charge liquidates the difference between proton and neutron. All this allowed us to consider
a neutron and a proton as two states of one and the same particle, a nucleon. Different
values of the isotopic spin operator projections correspond to the two possible states of the
nucleon. Let us agreed to ascribe I3 = 1/2 to the proton and I3 = −1/2 to the neutron.
Thus, a nucleon wavefunction can be presented in the form:
O
ΨN = N ψ(x; J), (9.29)

where    
X 1 0
N= ai ψi , ψp = , ψn = , (9.30)
0 1
i=p,n

|ap |2 and |an |2 define probability that the nucleon will be found in the proton and neutron
states, respectively (|ap |2 + |an |2 = 1) and ψ(x; J) is a part of a wavefunction that includes
coordinate and spin dependence. At this point we should recollect some facts, related to the
ordinary spin. Although the spin describes particle behavior with respect to rotations in
ordinary three-dimensional space, it cannot be connected with any particle spatial rotations.
The spin can be related with indestructible particle rotation only in the internal space.
Despite belonging to different kind of spaces, the spin and the orbital moment of momentum
can be summarized, their sum being the total moment of momentum of a particle. A
more convenient and quantum streamlined spin definition is simply indicating the existence
on particles of the new (spin) degrees of freedom, which are the eigenvalues of the spin
projection operator S3 , their number being equal to 2J + 1. The isospin is also connected
with particle behavior in the internal space, where the third axis is correlated with the
electric charge. The number of the isospin degrees of freedom, that is, the number of
(is)
possible values of S3 , is again equal to 2I + 1. However, unlike the ordinary spin, the
isospin does not contribute to quantities, which define the particle behavior in the ordinary
space. It can be checked easily, that with our agreements concerning the eigenvalues of the
(is)
operator S3 for the nucleon, Gell-Mann-Nishijima formula
B
Q = I3 + , (9.31)
2
where Q is a nucleon charge, expressed in units of |e|, takes place. From the aesthetic point
of view it is attractive to explain the aforesaid by the existence of an isospin symmetry,
which is the mathematical copy of the spin symmetry. It means that the isospin generators
satisfy the same permutation relations as the operators of the ordinary angular moment,
that is,
[Ik , Im ] = iεkmn In . (9.32)
Thus, in the states space there is the SU (2)-group of special (det U = 1) unitary (U † U = I)
transformations, for which the states Φ and MΦ (M is a transformation matrix of the
SU (2)-group) describe one and the same phenomenon when only the strong interaction is
taken into account. A nucleon is the most simple (spinor) representation of the rotations
group in the isotropic space. In this case the generators of the SU (2)-group themselves form
the group representation, that is, in the isospace the transformation matrix of a nucleon
state at the rotations through the angle θ is as follows:
U (θ) = exp [i(σ · θ)/2]. (9.33)
(is)
Thus, the Pauli matrices, multiplied by 1/2, play roles of the representation generators Sα
(α = 1, 2, 3). Since
    
(is) 1 1 0 1 1 1 1
S3 ψp = = = ψp , (9.34)
2 0 −1 0 2 0 2

© 2011 by Taylor and Francis Group, LLC

 Hadron families 133
    
(is) 1 1 0 0 1 0 1
S3 ψn = =− = − ψn , (9.35)
2 0 −1 1 2 1 2
(is) (is) (is)
then the operator S3 really is the isospin projection operator. From S1 , S2 one may
compose the operators
(is) (is) (is) 1
S± = S1 ± iS2 = (σ1 ± iσ2 ). (9.36)
2
They influence the proton and neutron states in the following way:
     
(is) 0 1 1 (is) 0 1 0
S+ ψp = = 0, S+ ψn = = ψp , (9.37)
0 0 0 0 0 1
     
(is) 0 0 1 (is) 0 0 0
S− ψp = = ψn , S− ψn = = 0. (9.38)
1 0 0 1 0 1
(is) (is)
From Eqs. (9.37) and (9.38) it is obvious, that the operator S+ (S− ) raises (lowers) the
(is)
isospin projection value by 1 for the nucleon states. In what follows we shall call S+ and
(is)
S− as the raising and lowering operators, for short. Their meaning can be understood
without resorting to their obvious form (not to be related to concrete representation), but
by using only the commutation relations
(is) (is) (is) (is) (is) (is)
[S+ , S− ] = 2S3 , [S3 , S± ] = ±S± , (9.39)
(is)
For this purpose it is enough to find the result of acting the operator S3 on the state
(is)
S± ΨI,I3 ,, where in the wavefunction ΨI,I3 for the sake of simplicity the spin and spatial
variables are neglected. Eqs. (9.37) and (9.38) are the special cases of the relations
(is)
p
S± ΨI,I3 = (I ∓ I3 )(I ± I3 + 1)ΨI,I3 ±1 . (9.40)

Since the obtained relations play an important role in obtaining the isospin or spin parts
of wavefunctions for compound systems, we shall give their proof. We start from putting
down the operator of the isotopic spin square as follows:
(is) (is) (is) (is)
(S(is) )2 = S− S+ + (S3 )2 + S3 . (9.41)

Using definitions of the matrix elements of the operator A

Z
(A)mn = Ψ†I,I3 =m AΨI,I3 =n dr, (9.42)

we find the matrix elements for both parts of the operator equation (9.41). With allowance
(is)
made for the diagonal form of the operators (S(is) )2 , S3 and taking into account the
condition of the wavefunctions normalization, we obtain the result:
(is) (is)
I(I + 1)Imm = (S− )mk (S+ )km + (m2 + m)Imm , (9.43)

where Imm are elements of the identity matrix. By means of the commutation relations
(is)
it is possible to find nonzero matrix elements (S± )km . Calculations for matrix elements
of the commutators (9.39) between states with I3 = m and I3 = k leads to the following
equations: )
(is) (is) (is) (is) (is)
(S3 S+ )mk − (S+ S3 )mk = (S+ )mk ,
(is) (is) (is) (is) (is) (9.44)
(S3 S− )mk − (S− S3 )mk = −(S− )mk ,

© 2011 by Taylor and Francis Group, LLC

 134 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Since
(is) (is) (is) (is) (is)
(S3 S± )mk = (S3 )ml (S± )lk = m(S± )mk ,
then the relations (9.44) take the form:
)
(is) (is)
(m − k)(S+ )mk = (S+ )mk ,
(is) (is) (9.45)
(m − k)(S− )mk = −(S− )mk .

(is) (is)
From Eqs. (9.45) it follows, that the matrix (S+ )mk ((S− )mk ) has nonzero elements
only for transitions, corresponding to the increase (decrease) of the quantum number I3 by
unity I3 → I3 + 1 (I3 → I3 − 1). Then the right-hand side of the expression (9.43) can be
presented as follows:
(is) (is) (is)
(S− )m,m+1 (S+ )m+1,m + (m2 + m)Imm == |(S− )m,m+1 |2 +
(is)
+(m2 + m)Imm = |(S+ )m+1,m |2 + (m2 + m)Imm , (9.46)
where we have taken into account
(is) (is) (is) (is)
(S+ )km = (S− )∗mk , (S− )km = (S+ )∗mk

By means of (9.46) Eq. (9.43) leads to the result:

(is)
p p
(S+ )I3 +1,I3 = I(I + 1) − I3 (I3 + 1) = (I − I3 )(I + I3 + 1), (9.47)
(is)
p
(S− )I3 −1,I3 = (I + I3 )(I − I3 + 1), (9.48)
which proves to be equivalent to the relations (9.40). Notice, that the proof of the relation
(9.40) could be done with purely military straightforwardness. For this purpose one should
build the “raising” and “lowering” operators from the orbital moment operators

L± = Lx ± iLy , Lk = −i~εklm xl ∂m ,

and act by them on the eigenfunctions of the operators L2 and L3 , which are the spherical
functions Ylml (θ, ϕ)

L2 Ylml (θ, ϕ) = l(l + 1)~2 Ylml (θ, ϕ), L3 Ylml (θ, ϕ) = ml ~Ylml (θ, ϕ).

As a result, we arrive at the required relation

p
L± Ylml (θ, ϕ) = ~ (l ∓ ml )(l ± ml + 1)Ylml ±1 (θ, ϕ). (9.49)

For antinucleons the scheme of building the wavefunction must be changed. It is caused
by the fact that the wavefunctions of the form ψ p C and ψ n C (C is the charge conjugation
matrix) correspond to antinucleons. Thus, the relation

N ′ = U (θ)N, (9.50)

means, that
′
N = U † (θ)N . (9.51)
Having used the obvious form of the transformation matrix U (θ), one can be easily per-
suaded, that

U ∗ (θ) = exp [−i(σ∗ · θ)/2] = (iσ2 ) exp [i(σ · θ)/2](−iσ2 ) = (iσ2 )U (θ)(−iσ2 ). (9.52)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 135

Consequently, if we write down for the antinucleon

   
−ψ n C −ψn
ξ = −iσ2 (N C) = = , (9.53)
ψpC ψp

then we get the transformation law

ξ ′ = U (θ)ξ. (9.54)

Since in the isospace, the transformation properties of ξ and N are the same, then working
with ξ and not with the antinucleon doublet directly, we do not make any difference between
particles and antiparticles. Thereby, there is no need to modify the theory of moments with
an eye to extending this theory in the case of antiparticles.
Since nucleons are fermions, then, according to the Pauli principle, no more than one
nucleon can present in one state, in this case the state now is characterized by one more
additional number, the eigenvalue of the third component of the isospin operator. Conse-
quently, a wavefunction of two nucleons must be completely antisymmetric with respect to
the nucleons transposition.
Let us consider, as a case in point, a two-nucleon system, a deuteron, which is described
by the production of two nucleon wavefunctions ψA and ψB . Since the isotopic spin is an
additive quantum number, then the two-nucleon state can possess the isotopic spin being
equal either to 1 or to 0. The isospin operator of a compound system with the wavefunction
ΨAB is defined by the expression:
(is) (is)
S(is) = SA + SB , (9.55)

where the subscripts indicate, on which wavefunction of the subsystem the operator acts.
It is convenient to represent the square of the operator S(is) in the form:
(is) (is) (is) (is) (is) (is) (is) (is) (is)
(S(is) )2 = (SA )2 + (SB )2 + 2SA · SB = (SA )2 + (SB )2 + (SA1 + iSA2 )(SB1 −
(is) (is) (is) (is) (is) (is) (is) (is) (is) (is) (is)
−iSB2 ) + (SA1 − iSA2 )(SB1 + SB2 ) + 2SA3 SB3 = (SA )2 + (SB )2 + SA+ SB− +
(is) (is) (is) (is)
+SA− SB+ + 2SA3 SB3 . (9.56)
Two-nucleon wavefunction in states with the definite values of I and I3 is a linear combi-
nation of the states

ψp(A) ψp(B) , ψp(A) ψn(B) , ψn(A) ψp(B) , ψn(A) ψn(B) . (9.57)

It is absolutely obvious, that the states

ψp(A) ψp(B) and ψn(A) ψn(B) (9.58)

enter into the isotopic triplet (I = 1), and the values I3 = 1 and I3 = −1 are associated
(A) (B) (A) (B)
with them. A certain combination of the remaining states ψp ψn and ψn ψp must play
the role of the third component of the triplet with I3 = 0. Finally, the other orthogonal
(A) (B) (A) (B)
combination of the states ψp ψn and ψn ψp can only belong to the singlet state with
I = 0 and I3 = 0. To single out the triplet state with I3 = 0, the state with I3 = 0 must
(is) (is) (is)
be acted upon by the lowering operator S− = SA− + SB− , which, as we see, is symmetric
(A) (B)
with respect to the indices A and B. Since the state ψp ψp is also symmetric with respect
to these indices, then the emerging state has the form:
 
(is)
S− ψp(A) ψp(B) = const × ψp(A) ψn(B) + ψn(A) ψp(B) , (9.59)

© 2011 by Taylor and Francis Group, LLC

 136 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
√
where from the normalization condition it follows that const equals 1/ 2. Because the
state with I = I3 = 0 must be orthogonal to that defined by Eq. (9.59), then it is given by
the expression:
1  
√ ψp(A) ψn(B) − ψn(A) ψp(B) . (9.60)
2
Using the eigenvalue equations

(S(is) )2 ΨAB = I(I + 1)ΨAB , 

(is)
(SA )2 ψA = IA (IA + 1)ψA , , (9.61)
(is) 
(SB )2 ψB = IB (IB + 1)ψB , 

and the expression for (S(is) )2 (9.56), one can easily check, that the symmetric states

pp, I3 = 1 
√1 (pn + np) , I3 = 0 I = 1, (9.62)
2 
nn, I3 = −1

really form the isotriplet, and the antisymmetric state

1
√ (pn − np) (9.63)
2

is the isosinglet (in the formulas (9.62) and (9.63) we switched over to more economic and
obvious notations). The deuteron, which is the spatially symmetric two-nucleon state, must
have such an isospin, that its total wavefunction is antisymmetric. Thus, the conclusion is,
that the deuteron has an isospin equal to zero.
For a nucleon-antinucleon system (N N ) the isotriplet and the isosinglet states now have
the form: 
−pn, I3 = 1 
√1 (pp − nn) , I3 = 0 I = 1, (9.64)
2 
np, I3 = −1
1
√ (pp + nn) . I=0 (9.65)
2
Isodoublets are also formed by some strange particles, for example, by K-mesons
 +  0
K K
(s = 1), (s = −1),
K0 K−

cascade Ξ-baryons  
Ξ0
(s = −2)
Ξ−
and so on. Since the masses of the strange particles, constituting the isodoublet are very
close to each other

mK ± = 493.577 ± 0.016 MeV/c2 mK 0 ,K 0 = 497.672 ± 0.031 MeV/c2 ,

mΞ− = 1321.32 ± 0.13 MeV/c2 mΞ0 = 1314.9 ± 0.6 MeV/c2 ,

then it is possible to speak of the confirmation of conserving the isospin under the strong
interaction between strange particles. However, the appearance of a new quantum number,

© 2011 by Taylor and Francis Group, LLC

 Hadron families 137

a strangeness, forces us to change the Gell-Mann-Nishijima formula (9.33) in the following

way:
B+s Y
Q = I3 + = I3 + , (9.66)
2 2
where Y is a particle hypercharge. The formula (9.66) can be viewed as nature’s delicate
hint at the structure of the gauge model of the electroweak interaction, which many years
later will be destined to be discovered by Glashow, Weinberg, and Salam. Actually, the
first term in the right-hand side of Eq. (9.66) is connected with the SU (2)-group, while the
second one is associated with the U (1)-group. Both the isospin and hypercharge symmetries
are not exact ones. However, the quantity in the left-hand side of Eq. (9.66), the electric
charge, belongs among the exactly conserving quantities. So, if one suggests, that the Cre-
ator was using the same rules of the game to produce both the strong and electroweak
interactions, then the electroweak interaction theory must contain the following elements:

(i) the weak isospin group SU (2)EW and the weak hypercharge group U (1)EW ;
(ii) the SU (2)EW - and U (1)EW -symmetries must be violated to the level of the U (1)em -
symmetry.

In Part II we shall learn that these very elements made the foundation of the Glashow-
Weinberg-Salam model.
According to the Yukawa hypothesis, the nuclear forces between nucleons are caused by
the π ± -, π 0 -meson exchanges (although, as we know at present, it is a very approximate
statement). Consequently, the pion-nucleon interaction must be isotopically invariant as
well. From the reaction
N → N ′ + π,
which is the basis for the Yukawa hypothesis, follows that the π-meson must have the
isospin equal either to I = 1 or I = 0. Since there are three π-mesons having the same
spins, parities and almost equal masses, then it is natural to assume, that nature chose the
possibility with I = 1. Thus, the π-mesons can be viewed as the different charge states of
one and the same particle, whose wavefunction is transformed as a vector in the isospace,
that is, has the following form:
O
Ψπ = Π Φ(x; J), (9.67)

where  
3 3 1
X X α+
Π= bi ψi′ , | bi |2 = 1, ψπ′ + = √  i  ,
i=1 i=1
2 0
   
0 1
α−
ψπ′ 0 =  0  , ψπ′ − = √  −i  ,
1 2 0
and α± are phase factors, which we choose later. Now the matrices
     
0 0 0 0 0 i 0 −i 0
(is)′ (is)′ (is)′
S1 = 0 0 −i  , S2 =  0 0 0, S3 =  i 0 0  .
0 i 0 −i 0 0 0 0 0

play the role of the representation generators. It is easy to check correctness of the relations:
(is)′ (is)′ (is)′
S3 ψπ′ + = ψπ′ + , S3 ψπ′ 0 = 0, S3 ψπ′ − = −ψπ′ − .

© 2011 by Taylor and Francis Group, LLC

 138 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Raising and lowering operators in this case are of the form:

   
0 0 −1 0 0 1
(is)′ (is)′
S+ =  0 0 −i  , S− =  0 0 −i  .
1 i 0 −1 i 0

Their action on the isotopic parts of the wavefunctions of the pion triplet is governed as
follows: √ √ )
(is)′ (is)′
S+ ψπ′ 0 = − 2(α+ )−1 ψπ′ + , S+ ψπ′ − = 2α− ψπ′ 0 ,
(is)′ √ (is)′ √ (9.68)
S− ψπ′ + = − 2α+ ψπ′ 0 , S− ψπ′ 0 = 2(α− )−1 ψπ′ − .
The choice of phase factors in the form of

α+ = −1, α− = 1

makes the right-hand parts in Eq. (9.68) positive to turn the relations (9.68) in the par-
ticular case of the formula (9.40). However, it is often more convenient to work with the
(is)′
representation, in which the matrix S3 is diagonal. The transition to such a representa-
tion is carried out by the transformation:

ψπ = Oψπ′ ,

where a matrix O has the form:

 
1 −i √0
1
O = √ 0 0 2.
2 1 i 0

In the new representation the generators are defined according to

     
0 −1 0 0 i 0 1 0 0
(is) 1 (is) 1 (is)
S1 = √  −1 0 1  , S2 = √  −i 0 −i  , S3 =  0 0 0 ,
2 0 1 0 2 0 i 0 0 0 −1
(9.69)
and the isospin parts of the π-meson wavefunctions have the form:
     
1 0 0
ψπ+ =  0  , ψπ0 =  1  , ψπ− =  0  . (9.70)
0 0 1

Besides, there are isomultiplets formed of one particle only, so-called isosinglets. The iso-
topic parts of wavefunctions for such particles are isoscalars, which are invariant under the
isotopic transformations. The Λ-hyperon can serve as an example.
To illustrate the power of the formula (9.40) under obtaining the spin or isospin parts of
a wavefunction of composite systems, let us define the isotopic wavefunction of the pion-
nucleon system (ΠN ). According to the rule of the vector composition, the total isospin of
the system can take values I = 3/2 or I = 1/2. Consequently, there are six states

(3/2, 3/2), (3/2, 1/2), (3/2, −1/2), (3/2, −3/2),
(9.71)
(1/2, 1/2), (1/2, −1/2).

which must be expressed in terms of the π-meson and nucleon states. Obviously, the
“highest” state is
(3/2, 3/2) = π + p, (9.72)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 139
(is)
since it is the only one with I3 = 3/2. Now let the operator S− act on both sides of Eq.
(9.72). Then, according to (9.40), in the left-hand side we obtain
(is)
√
S− (3/2, 3/2) = 3(3/2, 1/2). (9.73)

In so doing, the right-hand side of Eq. (9.71) takes the form:

(is) (is) (is) (is) (is)
√ 0
S− (π + p) = Sπ− (π + p)+SN − (π + p) = (Sπ− π + p)+(π + SN − p) = 2(π p)+(π + n). (9.74)

Combining Eqs. (9.73) and (9.74) we get

r r
2 0 1 +
(3/2, 1/2) = (π p) + (π n). (9.75)
3 3
(is)
Action of the operator S− on Eq. (9.75) results in
(is)
S− (3/2, 1/2) = 2(3/2, −1/2) (9.76)

and
r r r r
(is) 2 0 1 + 2 (is) 0 0 (is) 1 (is) + (is)
S− [ (π p)+ (π n)] = [(S π p)+(π SN − p)]+ [(S π n)+(π + SN − n)] =
3 3 3 π− 3 π−
r r r
2√ − 0 1 √ 0 2 − 2 0
= [ 2(π p) + (π n)] + [ 2(π n) + 0] = √ (π p) + 2 (π n). (9.77)
3 3 3 3
Eqs. (9.76)–(9.77) produce
r r
1 − 2 0
(3/2, −1/2) = (π p) + (π n). (9.78)
3 3
(is)
Now build-up of the state (3/2, −3/2) is possible without using the operator S− because
(is)
the only state with S3 = −3/2 is the combination:

(3/2, −3/2) = (π − n). (9.79)

The remaining two states with (1/2, 1/2) and (1/2, −1/2) can be easily obtained, according
to their orthogonality to the states (3/2, 1/2) and (3/2, −1/2) respectively, that is, they
have the form: r r
2 + 1 0
(1/2, 1/2) = (π n) − (π p), (9.80)
3 3
r r
1 0 2 −
(1/2, −1/2) = (π n) − (π p). (9.81)
3 3
Solving Eqs. (9.75), (9.78), (9.80) and (9.81), we obtain the final answer

(π + p) = (3/2, 3/2), (π − n) = (3/2, −3/2), 

q q 

0 2 1 
(π p) = 3 (3/2, 1/2) − 3 (1/2, 1/2) 


q q 

+ 1 2
(π n) = 3 (3/2, 1/2) + 3 (1/2, 1/2), (9.82)
q q 

(π − p) = 13 (3/2, −1/2) − 23 (1/2, −1/2), 


q q 

2 1


0
(π n) = 3 (3/2, −1/2) + 3 (1/2, −1/2). 

© 2011 by Taylor and Francis Group, LLC

 140 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Thus, we have learned the mechanism of defining the isospin (or any spinlike) part of
the wavefunction for any compound system, that is, in this situation we have known the
answer for questions “how”, “where”, ... and “from.” Thereafter, it is high time to become
aware of the existence of tabulated values of the numeric coefficients, which appear in the
composition theory of moments (orbital, spinlike, total). By means of these coefficients the
isospin function of the pion-nucleon system ΦI,I3 is expressed in terms of the isofunctions
of the π-meson and nucleon N as follows:
X I ′ ,I ′ ;I ′′ ,I ′′
ΦI,I3 = CI,I33 3
ΠI ′ ,I3′ NI ′′ ,I3′′ ,
I3′′ ,I3′

I ′ ,I ′ ;I ′′ ,I ′′
where CI,I33 3
are Clebsch-Gordan coefficients, which can be found in the Review of
Particle Physics published every two years.

9.3 Unitary multiplets

By using the isospin symmetry we can reduce a number of independent elementary particles.
For example, three π-mesons are simply three states of one and the same particle and so on.
However, by the 1960s so many isotopic multiplets had been discovered, that the necessity
of searching for a higher symmetry to unify particles into more densely populated families
became natural. Since such a symmetry has thoroughly escaped observation up to now,
it was obvious that somehow and somewhere this symmetry has passed a violation stage,
that is, today it is approximate. As a violation criteria of this symmetry, once again they
can take mass difference for particles, entering into new multiplets, which we are going
to call unitary multiplets. Unitary multiplets must contain isomultiplets, that is, unitary
symmetry group contains in itself a subgroup of the isotopic transformations. Thus, among
the Casimir operators of an unitary symmetry group, the isotopic spin square is present.
What other operators can apply for the role of invariants in a new group? If we agree once
again to denote particle states with dots in some abstract space, then the dimensionality of
a new space of states is defined by the number of invariants of an unitary symmetry group.
From the physical point of view it is reasonable to demand, that “coordinates” of this space
must be dependent (in the same way as space and time are interconnected in the special
theory of relativity). Experiments, which had been carried out by that time, demonstrated,
that the strangeness s and the isospin I are not independent. Let us explain what is meant
by this. In the strong interaction s and I are the conserved quantities, and as a result, the
following relation takes place:
∆s = ∆I3 = 0. (9.83)
Behavior of the isospin and the strangeness is not so faultless as far as the weak interaction
is concerned. Since the isospin and the strangeness are not defined for leptons, then it
makes sense to analyze only semilepton and nonlepton weak interaction. In the former case
the final state is formed of both leptons and hadrons, while in the latter case leptons are
absent in the final state. Below we give some examples of typical semilepton decays and
indicate the changes of the strangeness and the isotopic spin projection

n → p + e− + ν e , ∆s = 0, ∆I3 = 1 (9.84)

π + → π 0 + e+ + νe , ∆s = 0, ∆I3 = −1 (9.85)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 141

Λ → p + e− + ν e , ∆s = 1, ∆I3 = 1/2, (9.86)

+ 0 +
K → π + µ + νµ , ∆s = −1, ∆I3 = −1/2. (9.87)
For nonlepton decays the selection rules according to s and I3 can be illustrated by the
example of the reactions:

Λ → p + π− , ∆s = 1, ∆I3 = −1/2, (9.88)

K + → π+ + π0 , ∆s = −1, ∆I3 = 1/2. (9.89)

In the electromagnetic interaction the strangeness remains a good quantum number. Two
typical decays
π 0 → 2γ, η 0 → 2γ, (9.90)
demonstrate this circumstance

∆s = 0, ∆I3 = 0. (9.91)

From the aforesaid it follows that in all the existing processes for a closed system the change
of the strangeness entails the strictly defined change of the isotopic spin projection

| ∆s |= 0 −→ | ∆I3 |= 1, 0, (9.92)

| ∆s |= 1 −→ | ∆I3 |= 1/2. (9.93)

Thus, the choice of two quantum numbers, the isospin projection and the strangeness, for
dependent coordinates of the unitary spin space is quite well grounded. In Fig. 9.1 we

FIGURE 9.1
The unitary spin plane.

plotted the plane with the coordinates (s, I3 ) and on it we shall specify the states of unitary
multiplets. From the mathematical point of view, transitions between states must be carried
out by means of operators, which we agree to denote by vectors. The raising and lowering
(is) (is)
isospin operators S+ and S− allow crossing from one state to another inside of one and
(is)
the same isomultiplet. Since acting S± on a state changes the values of I3 by ±1 and in the
process ∆s = 0, then the vectors with module 1, parallel to axis I3 , are correlated with these
operators. It is obvious, that in order to ensure the transitions between isomultiplets, that
is, the transitions with a strangeness change, the similar raising and lowering operators

© 2011 by Taylor and Francis Group, LLC

 142 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

must exist. It is natural that these operators must be spinlike, that is, their algebra is
determined by the same commutation relations as for ordinary spin algebra. The raising
U
operator S+ is associated with the selection rules

∆s = 1, ∆I3 = −1/2, (9.94)

U
while the lowering operator S− describes the situation with

∆s = −1, ∆I3 = 1/2. (9.95)

U U
The operators S+ , S− , and S3U are the group generators of the so-called U -spin, and for
them the following commutation relations:
U U
[S+ , S− ] = 2S3U (9.96)

are valid. Obviously, without the introduction of one more type of generators it is impossible
to obtain on the plane (s, I3 ) the closed and symmetric (with respect to axis s) figures, which
we are going to associate with unitary multiplets. The transitions with the selection rules

∆s = 1, ∆I3 = 1/2,
(9.97)
∆s = −1, ∆I3 = −1/2,
V V
are carried out by the operators S+ and S− . These operators alongside with S3V constitute
a V -spin group, and for them the ordinary commutation relations for the moments are
fulfilled
V V
[S+ , S− ] = 2S3V . (9.98)
Notice, that in the chosen√scale the modules of the vectors, we correlate with the operators
U V
S± and S± , are equal to 5/2, and tangents of the angles, to be formed by the operators
U V
S+ and S+ with the strangeness axis, are equal to 1/2.
(is) V U
Nine generators S1,2,3 , S1,2,3 , and S1,2,3 form the closed system and generate a group
of the second rank SU (3), an unitary spin group, which was proposed by Y. Neeman [24]
and, regardless of him, M. Gell-Mann [25] to classify hadrons. In this scheme, just as in
the Mendeleev periodic table, objects are placed in the order of increasing their masses and
separation into families, unifying elementary particles with similar properties, is used. The
proposed theory is very often called the octal way, since, according to its statements, the
majority of particles is grouped in 8-plets (octets).
With regard to hadron classification over isotopic multiplets, practically all hadrons that
were reliably established by the 1960s, could be unified in four families. These families are
characterized by the baryon charge, the spin and the parity (J p ). Let us take the following
notations. After a isomultiplet symbol the approximated value of average mass in MeV/c2
will be given in parentheses, while the hypercharge value will be indicated under every
multiplet. Then these families are as follows:

8 baryons with 1/2+ and B = 1:

N (939) Λ(1115) Σ(1193) Ξ(1317)
p, n Λ0 Σ+ , Σ0 , Σ− Ξ0 , Ξ− , (9.99)
+1 0 0 −1
8 pseudoscalar mesons and meson resonances with 0− :
π(137) K(495) K(495) η(548)
− 0
π , π0 , π−
+
K +, K 0 K ,K η , (9.100)
0 +1 −1 0

© 2011 by Taylor and Francis Group, LLC

 Hadron families 143

8 vector resonances with 1− :

∗
ρ(770) K ∗ (992) K (992) ω(782)
∗− ∗0
+ 0 −
ρ ,ρ ,ρ K ∗+ , K ∗0 K ,K ω0 , (9.101)
0 +1 −1 0
3+
9 baryon resonances with 2 and B = 1:

∆(1232) Σ∗ (1385) Ξ∗ (1530)

++
∆ , ∆+ , ∆0 , ∆− Σ , Σ∗0 , Σ∗−
∗+
Ξ∗0 , Ξ∗− . (9.102)
1 0 −1

Subsequently, we shall find the explicit form of both the operators of the U -, V -spin and
of the operators of the YU -, YV -hypercharge. We shall be also convinced, the commutation
relations (9.96) and (9.98) are correct. Forestalling events we want to point out, that all of
these operators are in fact definite combinations of the isospin and hypercharge operators.
V U
However, at the moment we are not interested in the explicit form of the operators S± , S± .
(is)
Our task is, first of all, to be sure that they are able together with the operators S± to
place all the known, by that time, hadrons in the superfamilies (9.99)–(9.102).
Let us start from the supermultiplet including the nucleons. For the sake of convenience
on the abscissa we shall plot the hypercharge values and not the strangeness ones. In
the plane (Y, I3 ), which we are going to call the unitary spin plane, the point with the
coordinates Y = 1, I3 = 1/2 corresponds to the proton (Fig. 9.2). Let us assume, that p

FIGURE 9.2
The baryon octet.

occupies the highest state, that is,

(is) V U
S+ p = S+ p = S+ p = 0. (9.103)

© 2011 by Taylor and Francis Group, LLC

 144 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

To obtain other particles, the lowering operators must act on the proton state. It gives

(is)
S− p = n, (Y = 1, I3 = −1/2), (9.104)

V
S− p = Λ, Σ0 (Y = 0, I3 = 0), (9.105)

U
S− p = Σ+ (Y = 0, I3 = 1). (9.106)
U V
Action of S− on n brings forth the same result as the action of S− on p, that is, it yields the
state with Y = 0 = I3 = 0 which we have already identified with the Λ- and Σ0 -hyperons. A
new state with the quantum numbers Y = 0 and I3 = −1, the Σ− -hyperon, can be obtained,
V
if the operator S− acts on n. It is important to remember that our final goal is to obtain a
closed symmetrical figure. Consequently, the lowest state must have the quantum numbers
with opposite signs compared to the highest state, that is, it has Y = −1, I3 = −1/2. The
Ξ− -hyperon is such a state and transition to it from the Σ− -state is carried out by the
U (is)
S− -operator. Then, by means of the S+ -operator we arrive at the last state with Y = −1,
0
I3 = 1/2, which is the Ξ -hyperon. As one can see, the particle masses in the baryon octet
are much more different from each other than those in isomultiplets. Thus, for example,

m(Ξ) − m(N )
≈ 17%.
m(Ξ) + m(N )

In other words, the unitary symmetry has been violated stronger than the isotopic one.
(is) U V
In the same way by means of the operators S± , S± and S± , the 0+ -mesons of (9.100)
−
and the 1 -meson resonances of (9.101) can be grouped into octets. There are also unitary
singlets, for example, the η ′ (957)-meson forms the 0− -singlet. Unlike mesons (where par-
ticles and antiparticles enter into one and the same families), antibaryons form individual
families, which are the same as the baryon ones (9.102).
So, we introduced the new quantum number, the unitary spin, to be a generalization of
the isospin, and involve both the isospin and the strangeness. Our world is made in such
a way, that the strong interaction is approximately invariant with respect to the rotations
in the unitary spin space. For this reason hadrons are grouped into the unitary multiplets.
This is an axiom of the unitary symmetry theory. All particles of such superfamilies can
be viewed simply as a set of the states of one and the same particle, which are degenerated
in the electric charge and the hypercharge. According to concepts of the SU (3)-symmetry,
baryons with spin 1/2 must be unified in the unitary octet, while baryons with spin 3/2
must be grouped into the unitary decuplet.
Nine baryon resonances with 3/2+ (9.102) might be placed in the decuplet with one
vacant lower place (Fig. 9.3). From Fig. 9.3 it follows, that the difference of masses
between neighboring isomultiplets is constant and it is approximately equal to 146 MeV/c2 .
Thus, it is possible to predict the mass, the strangeness, and the electric charge of a missing
member of the baryon decuplet 3/2+, which we are going to denote as Ω−

mΩ = 1676 MeV/c2 , s = −3.

The strangeness conservation law forbids the decay of this particle through the strong
interaction. Actually, a decay channel with the least mass of strange particles

Ω− → Ξ0 + K − (9.107)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 145

FIGURE 9.3
The baryon resonances decuplet.

turned out to be closed due to the energy arguments (mΩ < mΞ0 + mK − ). The weak
interaction with the following decay channels


 Λ + K−
 Ξ− + π 0


Ω → Ξ0 + π − (9.108)
 − + −


 Ξ + π + π
 0
Ξ + e− + ν e

can give the only chance for the Ω− -hyperon to maintain its status of an unstable particle.
Then the calculations of the Ω− -hyperon lifetime gave the value of the order of 10−10 s. On
the microworld scale it was a long-lived particle and its track had to be seen in a bubble
chamber that had already existed by that time. In 1964 at the Brookhaven Accelerator the
particle was discovered, which characteristics exactly coincided with the predicted ones. It
was one of those miraculous enlightenments, which are so rare in the history of the human
mind. Discovery of the Ω− -hyperon was similar to that of the planet Neptune by Leverrie—
the discovery made on the pen’s edge. It was regarded to be a brilliant proof of the hadrons
classification according to the SU (3)-symmetry.
A certain isotopic multiplet is described by an isotopic wavefunction, being one of an
irreducible representations of the SU (2)-group. In the same way a certain unitary multiplet
can be described by a multi component unitary wavefunction, which is an irreducible rep-
resentation of the SU (3)-group. Let us study in detail the SU (3)-group and in particular
its Lie-algebra and its irreducible representations.
The number of linearly independent vectors, defined in the linear space in which the
transformations matrix act, is called the representation dimensionality. In the case of an
internal symmetry the number of particles in the corresponding multiplet is the representa-
tion dimensionality. The most simple representations, from which all the rest of the group
representations can be built with the help of multiplication, are named by the fundamen-
tal ones. For the SU (n)-group those are n-component spinors. Thus, the triplet is the

© 2011 by Taylor and Francis Group, LLC

 146 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

fundamental representation of the SU (3)-group. To be exact, there are two such represen-
tations: covariant and contravariant triplets, but we will discuss it later. Let us consider
the transformations, which leave the three-component SU (3)-spinor invariant

χk −→ χ′k = Ukl χl , (9.109)

where U are arbitrary and unimodular 3 × 3-matrices. Canonic representation of a matrix

U has the form  
i
U = exp αa λa , (9.110)
2
where a = 1, 2, ..8 and λa are 3 × 3-matrices satisfying the relations

Sp (λa λb ) = 2δab , Sp λa = 0 λ†a = λa . (9.111)

We are reminded that the number of independent parameters and the number of generators
of the SU (n)-group is equal to n2 − 1. Here the matrices λa play the same role as the
Pauli matrices do in the case of the SU (2)-symmetry, that is, λa /2 are the generators of
the fundamental representation of the unitary spin. The standard writing of these matrices,
introduced by Gell-Mann is as follows:
     
0 1 0 0 −i 0 1 0 0
λ1 =  1 0 0  , λ2 =  i 0 0  , λ3 =  0 −1 0  ,
0 0 0 0 0 0 0 0 0
     
0 0 1 0 0 −i 0 0 0
λ4 =  0 0 0  , λ5 =  0 0 0  , λ6 =  0 0 1  ,
1 0 0 i 0 0 0 1 0
   
0 0 0 1 0 0
1
λ7 =  0 0 −i  , λ8 = √  0 1 0  . (9.112)
0 i 0 3 0 0 −2

The choice of the generators in the form (9.112) is convenient, because the first three
matrices λ1,2,3 are the Pauli matrices (they form Lie algebra of the SU (2)-group). What
this means is that the SU (3)-group contains the isotopic spin group as the subgroup. The
matrices λ3 and λ8 commute with each other, that is, SU (3) is really the group of the
second rank. The permutation relations to characterize the group and to be satisfied by the
matrices λa , resemble in form those for the matrices σl

[λa , λb ] = 2ifabc λc . (9.113)

Structural constants fabc are real and antisymmetric with respect to all indices. They can
be determined by means of the relations:
1
fabc = Sp([λa , λb ]λc ). (9.114)
4i
The components of fabc being different from zero have the following values:
√
1 3
f123 = 1, f147 = f246 = f345 = f257 = −f156 = −f367 = , f458 = f678 = .
2 2
(9.115)
For the matrices (9.112) antipermutation relations exist as well
4
{λa , λb } = 2dabc λc + δab , (9.116)
3

© 2011 by Taylor and Francis Group, LLC

 Hadron families 147

where constants dabc are completely symmetric with respect to transpositions of indices.
By means of relations
1
dabc = Sp({λa , λb }λc ), (9.117)
4
the nonzero components of dabc can be obtained

d118 = d228 = d338 = −d888 = √13 , d448 = d558 = 
1
= d668 = d778 = − 2√ 3
, d146 = d157 = −d247 = . (9.118)
1 
= d256 = d344 = d355 = −d366 = −d377 = 2

The matrices λa are usually called the unitary spin operators. As in the case of the
normal spin, the sum of the matrix squares is proportional to an identity matrix
 
1 0 0
16 
λa λa = 0 1 0 (9.119)
3
0 0 1

and that gives us the right to use the above-mentioned term. The λa -matrices can be also
viewed as components of the eight-dimensional SU (3)-vector. The aforesaid is also valid
for generators of any representation of the SU (3)-group. So, from generators of the unitary
(un)
spin Sa the operator of the unitary spin square can be obtained

S(un)2 = Sa(un) Sa(un) , (9.120)

whose eigenvalues characterize the given representation. However, it is not the only Casimir
operator in the SU (3)-group. If we introduce the SU (3)-vector

2 (un)
Da = dabc Sb Sc(un) , (9.121)
3
then it is easy to see that the quantity

F = Sa(un) Da , (9.122)
(un)
is the second Casimir operator. Although the Da -vector is made of the Sa -generators,
(un)
however, Da and Sa are linearly independent. It follows from the fact that after multi-
(un)
plying by Sa they form two different Casimir operators of the SU (3)-group. These eight-
dimensional SU (3)-vectors satisfy the commutation relations typical for moments theory:
(un)
[Sa(un) , Sb ] = ifabc Sc(un) , (9.123)

(un)
[Da , Sb ] = ifabc Dc . (9.124)
Notice, that in the case of the SU (2)-group the well-known relation:

[pk , Ml ] = iεklm pm , (9.125)

(un)
characterizing vector property of p, is analog of (9.124). Representing Sa and Db in the
form of 8 × 8-matrices
(Sa(un) )bc = ifabc , (9.126)
(Da )bc = dabc . (9.127)

© 2011 by Taylor and Francis Group, LLC

 148 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

one could be convinced of their linear independence. Equality (9.126) is proved by substi-
tution of λa , λb , λc into the Jacobi identity

[[A, B], C] + [[B, C], A] + [[C, A], B] = 0, (9.128)

by a consequent multiplication by every λ and by calculating the trace. Further, the validity
of the choice of the matrix representation for the Da -operator could be checked by the
fulfillment of (9.124).
(un)
The connection of Sa both with the isospin operators and with the U -, V -spin operators
is given by:

(is) (un) (un) U (un) (un)
S± = S1 ± iS2 , S± = S6 ± iS7 , 
(un) (un) (is) (un) (un) 
(9.129)
SV = S
± 4 ± iS 5, S =S
3 ,3 Y = √2 S .
3 8

Since the hypercharge commutes with the third projection of the isospin, then the hyper-
(un)
charge operator can differ
√ from S8 only in a constant factor. If in Eq. (9.129) we set
this factor equal to 2/ 3, then we arrive at correct expressions for hadron hypercharges.
Taking into account (9.129), it is easy to check the validity of the following commutation
relations:
U U 3 (is)
[S+ , S− ] = Y − S3 ≡ 2S3U , (9.130)
2
V V 3 (is)
[S+ , S− ]= Y + S3 ≡ 2S3V . (9.131)
2
The relations (9.130) and (9.131) prove, that the SU (3)-group besides the isospin subgroup
SU (2) really contains two more SU (2)-subgroups, the subgroups of the U - and V -spin.
Since the charge operator
(un) 1 (un)
Q = S3 + √ S8 , (9.132)
3
satisfies the commutation relations
U
[Q, S± ] = [Q, S3U ] = 0, (9.133)

then the operator YU = const × Q plays the role of the hypercharge operator for the U -spin.
To obtain the correct values of hadron quantum numbers, the constant in definition of the
U -hypercharge must be set to −1. Similar considerations concerning the operator YV lead
to the result:
(is) 1
YV = S3 − Y. (9.134)
2
It is obvious, that classification of particles in the SU (3)-multiplet can be done according to
both isomultiplets and multiplets of the U - and V -spin. In other words, states of particles
in the unitary multiplet can be presented by points on the planes (YU , U3 ) and (YV , V3 ).
The particles distribution of the baryon and meson octets according to the U - and V -spins
is shown in Tables 9.1 and 9.2. Notice, that the U -multiplets contain particles with the
same value of the electric charge.
The next step is to find an irreducible representations of the unitary spin group. A
unitary scalar is the most simple irreducible representation. It describes particles forming
unitary singlets. Thereupon the fundamental representation of the SU (3)-group, the unitary
triplet, about which we discussed above, follows. Let us add some mathematical details to
the aforesaid.

© 2011 by Taylor and Francis Group, LLC

 Hadron families 149
TABLE 9.1
The baryon octet.
U YU Baryons Mesons
√ √
0 0 − 23 Σ0 − 12 Λ0 - 3 0
2 π − 12 η 0
1 − −
2 1 Σ ,Ξ π , K−
−

1
2 -1 p, Σ+ K +, π+
√ √ 0
1 0 n, − 21 Σ0 + 3 0
2 Λ ,Ξ
0
K 0 , − 12 π 0 + 3 0
2 η ,K

TABLE 9.2
The meson octet.
V YV Baryons Mesons
√ √
3 0 1 0 3 0
0 0 2 Σ − 2Λ 2 π − 21 η 0
1 + 0 + 0
2 1 Σ ,Ξ π ,K
1
2 -1 n, Σ− K 0 , π−
√ √
1 0 p, 12 Σ0 + 3 0
2 Λ ,Ξ
−
K + , 12 π 0 + 3 0
2 η ,K
−

In a space E1 , transformed with respect to the fundamental representation, we introduce

the orthonormalized basis vectors ek (k = 1, 2, 3), which can be chosen in the form:
     
1 0 0
e1 =  0  , e2 =  1  , e3 =  0  . (9.135)
0 0 1
For objects, defined in unitary spin spaces we are going to use the term “vector.” The
concrete nature of a vector is decoded in writing down its components. Thus, vectors in
the space E1 are spinors of the first rank and arbitrary spinor Ψ is defined by the formula:

Ψ = Ψ k ek . (9.136)

Under the SU (3)-transformations components of the spinor Ψk are transformed by the law:

Ψ′k = Ukl Ψl . (9.137)

Just as in the case of other groups, we continue to call spinors that are transformed by the
fundamental representation contravariant spinors of the first rank. Thus, in the space E1
the contravariant spinors of the first rank are set.
Using the obvious form of the fundamental representation generators, it is easy to check
the validity of the following relations:

(is)
S3 e1 = 21 e1 , Y e1 = 13 e1 , 


(is) 1 1
S3 e 2 = − 2 e 2 , Y e2 = 3 e2 , . (9.138)


(is) 2 
S3 e3 = 0, Y e3 = − 3 e3 ,
It is convenient to display the basis vectors of the space E1 on the unitary spin plane, as
in Fig. 9.4. Clearly, according to our understandings, the unitary contravariant spinor χ
with the components χ1 = χ2 = χ3 = 1 describes the unitary triplet of particles, in which
the first two elements form the isodoublet (I = 1/2) and the third one forms the isosinglet
(I=0). Let us agree to call it the fundamental triplet and display it as the triangle in the
plane (I3 , Y ).

© 2011 by Taylor and Francis Group, LLC

 150 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
Y

e2 e1

−1/2 1/2
I3

e3
−2/3

FIGURE 9.4
The basic vectors in the E1 space.

The space E1 , transforming with respect to the representation that is conjugate (contra-
gradient) relative to the fundamental representation of the SU (3)-group, is also irreducible.
In this space, just as in the previous case, arbitrary spinor Φ can be presented in the form:

Φ = Φk e k , (9.139)

where the Φk -spinor components are transformed by means of the law:

†
Φ′k = Ukl
∗
Φl = Φl Ulk . (9.140)

The spinors, defined in the space E1 shall be called the covariant spinors of the first rank.
Thus, under transition from contravariant spinors to covariant ones, in the transformation
matrix U the change
λa → −λ∗a (9.141)
(is)
takes place. Inserting (9.141) into the expression for the operators S3 , Y and choosing
the basis vectors of the space E1 also in the form (9.135), we obtain the following eigenvalue
equations: 
(is)
S3 e1 = − 21 e1 , Y e1 = − 31 e1 , 


(is) 2 1 2 2 1 2
S3 e = 2 e , Y e = −3e , (9.142)


(is)
S3 e3 = 0, Y e3 = 23 e3 .


The basis vectors of the space E1 on the unitary spin plane are displayed in Fig. 9.5.
The components of the Hermitian conjugated covariant spinor Ψ†k are transformed as the
components of the contravariant spinor Ψk . Since a Hermitian conjugated wavefunction is
connected with antiparticles, then the covariant spinor η with the components η1 = η2 =
η3 = 1 describes the triplet of antiparticles. In what follows we are going to call it the
antitriplet and depict it as the triangle on the unitary spin plane (Fig. 9.5).

© 2011 by Taylor and Francis Group, LLC

 Hadron families 151
Y

2/3
e3

−1/2 1/2
I3

e1 e2
−1/3

FIGURE 9.5
The basic vectors in the E1 space.

In the case of the isotopic spin group two doublets, transforming according to the rules:

κ′ = U κ, κ′ = U ∗ κ, (9.143)

are equivalent, since U and U ∗ or, which is equivalent, σi and −σi∗ are connected through
the similarity transformation. This statement breaks down in the case of the matrices λa
and −λ∗a . For this reason the representations of the SU (3)-group are characterized by a
number of the contravariant p and covariant q indices D(p, q).
So, we got acquainted with the most simple representations: D(0, 0), D(1, 0), and D(0, 1).
More complicated representations of the SU (3)-group are no longer irreducible. Conse-
quently, a recipe is needed to divide these representations into direct sums of irreducible
representations.
Let us first study some concrete examples, and
N then we shall try to summarize the results
obtained. We start with the representation U U in the space E2 with the basis ekl . The
arbitrary contravariant spinor of the second rank with components
 11 
Φ Φ21 Φ31
Φkl =  Φ12 Φ22 Φ32 
Φ13 Φ23 Φ33

can be expanded into the symmetric part with six independent components
1 kl

Φ(s)kl = Φ + Φlk (9.144)
2
and into the asymmetric part with three independent components
1 kl

Φ(a)kl = Φ − Φlk . (9.145)
2
Notice, that Sp(Φkl ) is not a SU (3)-scalar. A scalar can be obtained by convolutions of
covariant and contravariant indices. According to this, the space E2 expands into the two

© 2011 by Taylor and Francis Group, LLC

 152 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
(6) (3)
subspaces E2 and E2 , whose bases are formed from six vectors
 1
(6) √ (ekl + elk ) ,
2
k 6= l,
e{kl} = ekl = (9.146)
ekl , k = l,
and three vectors
(3) 1
e[kl] = ekl = √ (ekl − elk ) (9.147)
2
(6)
respectively. Using (9.146) and (9.147) we can write down any vector in the subspaces E2
(3) (6)
and E2 . For example, an arbitrary vector in E2 has the form:
X (6) 1 X kl X
Φ(6) = Φkl ekl = √ Φ (ekl + elk ) + Φkk ekk . (9.148)
2 k>l k

The vector Φ, belonging to the space E2 , can be viewed as a direct product of two con-
travariant spinors of the first rank, or in other words, the components of the Φkl -spinor is
the direct product of the components of the Φk - and Φl -spinors. The operation of antisym-
metrization can be reduced to multiplying the starting spinor either by the quantities εkmn ,
εkmn or by their production. Thus, for example, since the relations:
1
V p = εpmn Φ(a)
mn Φ(a)
mn = εmnp V p (9.149)
2
(a)
are fulfilled, then the symmetric spinor of the second rank having the components Φmn is
equivalent to the three-dimensional vector Vp . Consequently, from the outset we can deal
(a)
not with Φmn but with the quantity εpmn Φmn . Notice, that symmetrization of spinors of the
highest ranks does not cause a transition from covariant to contravariant indices and vice
versa. It is not true, however, in the case of spinors antisymmetrization (see, for example
(9.149). Thus, we demonstrated that the product Φk and Φl is reducible and breaks down
into the two irreducible representations, that is
O M
3 3=6 3.
N ∗
Let us consider now the representation U U in the space E11 with the basis ekl . In
this space vectors are mixed spinors of the second rank with components Φkl
 1 
Φ1 Φ21 Φ31
Φkl =  Φ12 Φ22 Φ32  . (9.150)
Φ13 Φ23 Φ33

The track of this spinor Sp(Φkl ) = Φ11 + Φ22 + Φ33 is a unitary scalar, which is not changed
under the SU (3)-transformations. Subtracting the quantity δlk Φnn /3 from (9.150), we obtain
the eight-component spinor
1
Φkl − δlk Φnn =
3
2 1 1 2 3 2

Φ
3 1 − 3 (Φ 2 + Φ 3 ) Φ 1 Φ31
2 2 1
= Φ12 1 3
3 Φ2 − 3 (Φ1 + Φ3 ) Φ32 . (9.151)
2 3 1
Φ13 Φ23 Φ
3 3 − 3 (Φ 1
1 + Φ 2
2 )
which already is irreducible. According to this, the space E11 is decomposed into the two
(9) (1) (1)
irreducible subspaces E11 and E11 . The basis vector of the one-dimensional subspace E11
is given by:
(1)k 1

el = √ e11 + e22 + e33 . (9.152)
3

© 2011 by Taylor and Francis Group, LLC

 Hadron families 153

The basis vectors in the second subspace must be orthogonal to the vector (9.152). Obvi-
ously, six quantities ekl with k 6= l meet this demand. The remaining two basis vectors have
the form of linear combinations of the vectors e11 , e22 , and e33 . They will be orthonormalized
(1)
and orthogonal to the basis vector of the subspace E11 , provided the following choice:

1
1

√ e11 − e22 , √ e11 + e22 − 2e33
2 6
(9)
is made. Thus, the basis vectors of the space E11 are as follows:

1
1

e12 , √ e11 − e22 , e21 , e13 , e23 , e31 , e32 , √ e11 + e22 − 2e33 . (9.153)
2 6
Since the vector Φ, belonging to the space E11 , is a direct product of the contravariant and
covariant spinors of the first rank, then the representation to correspond to it breaks down
into two irreducible ones O M
3 3=1 8.
Thus, from the very beginning we managed to find among irreducible representations of
the SU (3)-group the eight-dimensional representation, which can be used to describe the
particle octets.
Let us determine the isospin content of the octet obtained, considering, that a mixed
spinor with components Φkl is composed of the product of the fundamental triplet and the
fundamental antitriplet. If we denote an isospin multiplet by (I, Y ), then a triplet and
antitriplet are presented as (1/2, 1/3) + (0, −2/3) and (1/2, −1/3) + (0, 2/3), respectively.
The product of the two isospin multiplets (I1 , Y1 ) and (I2 , Y2 ) contains multiplets with the
hypercharge Y = Y1 + Y2 and the isospins, which according to the moment summation rule,
take the values
I =| I1 − I2 |, | I1 − I2 | +1, .., I1 + I2 .
Thus, the isomultiplet multiplication rule has the form:

(I1 , Y1 )(I2 , Y2 ) = (| I1 − I2 |, Y1 + Y2 ) + (| I1 − I2 | +1, Y1 + Y2 ) + .....+

+(I1 + I2 , Y1 + Y2 ). (9.154)
Using (9.154), we arrive at the relation
O M M M M
3 3=1 8 = (1/2, 1) 2(0, 0) (1, 0) (1/2, −1). (9.155)

With the help of Eqs. (9.154) and (9.155) we finally obtain

M M M
8 = (1/2, 1) (0, 0) (1, 0) (1/2, −1), (9.156)

that is, the octet consists of the two isodoublets with Y = 1 and Y = −1, the isotriplet
with Y = 0, and the isosinglet with Y = 0.
For symmetrization and antisymmetrization of the covariant or contravariant spinors of
the rank n ≥ 2 it is convenient to use projection operators. Quantities to symmetrize and
antisymmetrize spinors on two indices can be built from the Kronecker symbols

kl 1 k l

Smn = δm δn + δnk δm
l
(9.157)
2
1 k l

Akl
mn =
l k
δ δ − δm δn . (9.158)
2 m n

© 2011 by Taylor and Francis Group, LLC

 154 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It is easy to see that the quantities Akl kl

mn and Smn form the set of the projection operators.
Actually they meet all the demands, imposed on these operators

A2 = A, S 2 = S, A + S = I, AS = 0, (9.159)

where for the sake of simplicity we neglected spinor indices. Antisymmetrization can be
reduced to multiplication of the initial spinor by the quantities εkmn , εkmn or by their
product. Thus, for example, an antisymmetric spinor of the second rank with components
(a)
Φmn
Φ(a) kl
mn = Amn Φkl

is equivalent to a three-dimensional vector Vp , since the following relations:

1
V p = εpmn Φ(a)
mn Φ(a)
mn = εmnp V p (9.160)
2
(a)
take place. Consequently, from the very beginning we can deal not with Φmn , but with the
quantity εpmn Φmn . Analogously, we could use the quantity εpmn Φpmn instead of the spinor
Φm̃ñp̃ , where the tilde above the indices indicates the spinor antisymmetry with respect to
transpositions according to these indices.
Let us use the methods of symmetrization and antisymmetrization over spinor indices to
obtain irreducibleNrepresentations
N of the space E3 , which is transformed by means of the
representation U U U . We consider an arbitrary contravariant spinor of the third rank
Φpkl (27 components). First, with the help of the symmetrizing projector we single out from
it all the components, symmetric in all three indices. Performing this operation by stages
with every pair of indices, we obtain

Φṗk̇l = Smn
pk
Φmnl , (9.161)

Φpk̇l̇ = Smn
kl
Φpmn , (9.162)
Φṗkl̇ = Smn
pl
Φmkn , (9.163)
where the dot over the indices denotes the spinor symmetry with respect to transposition
on those indices. Summarizing Eqs. (9.161)–(9.163), we get the spinor to be symmetric in
all the three indices:
1 pkl

Φṗk̇l̇ = Φ + Φplk + Φlpk + Φlkp + Φklp + Φkpl . (9.164)
6
There are ten independent components of this spinor: three with p = k = l, six with only
two coinciding indices, and one with three different p, k, l. The same sequence of operations
with the antisymmetrizing projector A produces a completely antisymmetric spinor:
1 pkl

Φp̃k̃l̃ = Φ − Φplk + Φlpk − Φlkp + Φklp − Φkpl . (9.165)
6

The spinor Φp̃k̃l̃ has only one independent component, since the spinor being antisymmetric
in all the three indices is equivalent to a scalar V
1
V = εpkl Φpkl .
6
In the initial spinor let us single out components to be symmetric in p, k and antisymmetric
in k, l
Φṗk̇l = Smn
pk
Φmnl , (9.166)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 155

Φpk̃l̃ = Akl
mn Φ
pmn
. (9.167)
However, not all components in Eqs. (9.166), (9.167) are independent. Some of them
are already taken into consideration at definition of the spinors, which are symmetric and
antisymmetric in all the three indices (for example, the first two terms in Eq. (9.164), the
first and the last terms in Eq. (9.165) ). Summing up (9.166) and (9.167) and subtracting
from the obtained result the parts found before, we arrive at:
˙ 1 pkl

Φṗk̃l̃ = Φ + Φkpl − Φplk − Φlpk . (9.168)
3
Continuing manipulations with the projectors (9.157) and (9.158), we obtain the expression
for the spinor, which is antisymmetric with respect to p, k and symmetric with respect to
k, l indices:
˙ 1 pkl

Φp̃k̃l̇ = Φ + Φplk − Φkpl − Φklp . (9.169)
3
˙ ˙
It is clear that the number of independent components q of the Φṗk̃l̃ - and Φp̃k̃l̇ -spinors
coincide and it is consequently equal to:
27 − 10 − 1
q= = 8.
2
Thus, the spinor Φpkl falls into the following irreducible representations:
O O M M M
3 3 3=1 8 8 10. (9.170)

Let us determine (I, Y )-content of the representations in (9.170). According to (9.154)

we have O O M M M M
3 3 3 = (3/2, 1) 2(1/2, 1) 3(1, 0) 3(0, 0)
M M
3(1/2, −1) (0, −2). (9.171)
Taking into consideration the isotopic structure of the octet (9.156) obtained before, we
define the (I, Y )-content of the decuplet from Eq. (9.171)
M M M
10 = (3/2, 1) (1, 0) (1/2, −1) (0, 2). (9.172)

Now we demonstrate how to fulfill splitting on irreducible representations in the case

of mixed spinors. Recall, that in the SU (3)-group there are three invariants, namely, δlk ,
εijk , and εijk . The Kronecker symbol invariance follows from the unitarity of the SU (3)-
transformations:
δl′k = Ulm Un∗k δm
n
= δlk . (9.173)
The antisymmetric quantities εijk and εijk are invariant due to the unimodularity of the
SU (3)-group:
ε′ijk = Uim Ujn Ukp εmnp = (detU )εijk = εijk . (9.174)
pk
Thus, the projectors, which symmetrize Smn and antisymmetrize Akl
mn , where

1 klp
Akl
mn = ε εmnp , (9.175)
2
are built of the SU (3)-group invariants. Besides these operations, there exist the operation
of convolution in the upper as well as in the low indices and the operation of a track
vanishing for a mixed spinor. They are both carried out by means of the invariant δnm .

© 2011 by Taylor and Francis Group, LLC

 156 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The mixed spinor for which the above-mentioned operations can not be fulfilled is already
irreducible.
By the example of the spinor Φmnpq we illustrate how to obtain irreducible spinors by
means of the SU (3)-group invariants.

(i) Multiplicating of Φmn p q

pq by δm and δn produces a scalar:

Ψ = Φpq
pq .

(ii) Multiplicating of Φmn p

pq by δm and vanishing of the track results in the eight-components
quantity:
1 n pm
Ψnq = Φpn pq − δq (Φpm ).
3
Analogously, multiplicating of Φmn
pq by δ q
n and vanishing of the track leads to:

1 m nq
Ψm mq
p = Φpq − δp (Φnq ).
3
(iii) Multiplication of Φmn
pq by εmnl and symmetrization with respect to the indices lpq gives
ten-components quantity:
Ψl̇ṗq̇ = S(lpq)εmnl Φmn
pq ,

where S(lpq) symbolizes symmetrization with respect to the indices lpq, that, as we know,
is reduced to the multiplication of the initial spinor by the Kronecker symbols combinations.
Analogously, the multiplication of Φmnpq by ε
pql
and symmetrization with respect to indices
lmn produces another irreducible spinor:

Ψl̇ṁṅ = S(lmn)εpql Φmn

pq .

(iv) Symmetrization with respect to the upper and low indices and vanishing of the track
results in the 27-components quantity:
1 m ṡṅ
1 m n

Φṁ ṅ
ṗq̇ − δ Φ + δqm Φṡṗṅṡ + δpn Φṁ ṡ n ṁṡ
ṡq̇ + δq Φṗṡ + δ δ + δqn δpm Φṙṙṡṡ .
5 p ṡq̇ 20 p q
If one takes into consideration that the spinor Φmn
pq can be presented as the product of the
spinors Φmp and Φn
q then all the above-mentioned splitting into irreducible spinors can be
symbolically expressed as follows:
O M M M M M
8 8=1 8 8 10 10 27.

Acquired experience allows us to come to generalizations, namely, to formulate the irre-

ducibility condition for the highest dimension representations (representations, not belong-
ing to: D(0, 0), D(1, 0), and D(0, 1)). A mixed spinor is irreducible, if being multiplied
by all the three invariants of the SU (3)-group—δlk , εijk , εijk —it produces zero. We are
reminded that we replace antisymmetrization with multiplication by εijk , εijk or by their
product. If one takes into account that the product of a symmetric spinor on the com-
pletely antisymmetric quantity εijk or εijk is equal to zero, then the aforesaid means, that
the components of an irreducible spinor must be separately symmetric with respect to the
upper and low indices and its track must turn to zero.
Let us connect the number of covariant and contravariant indices of an irreducible repre-
sentation with the number of particles in a multiplet n, and in so doing we determine the
form of a function n(p, q). Obviously, n(p, q) is also a number of independent components of
k ···k
a corresponding irreducible SU (3)-spinor Ψl11···lqp . Sometimes we use the term “spinor” to

© 2011 by Taylor and Francis Group, LLC

 Hadron families 157

denote vector components in the unitary spin space, only if it causes no misunderstandings.
Consider a contravariant spinor Ψk1 ···kp symmetrized with respect to all indices (conse-
quently, irreducible), or representation D(p, 0). Find the quantity n(p, 0). By p1 , p2 , p3 we
denote the number of ones, twos, and threes among the indices p. Since p1 + p2 + p3 = p,
then with the given p2 the number p1 can run the p − p2 + 1-values from 0 to p − p2 .
Therefore, we have
p
X 1
n(p, 0) = (p − p2 + 1) = (p + 1)(p + 2). (9.176)
p2 =0
2

Analogously, the number of components of a symmetrized covariant spinor of the rank q is

equal to
1
n(0, q) = (q + 1)(q + 2). (9.177)
2
k ···k
Then the number of components of the spinor Ψl11···lqp be given by:

1
n(p, 0)n(0, q) = (q + 1)(q + 2)(p + 1)(p + 2), (9.178)
4
if all the tracks:
k ...m..k
Ψl11...m..lqp

would be arbitrary. With allowance made for Eq. (9.178) the number of such tracks or
conditions:
k ...m..k
Ψl11...m..lqp = 0 (9.179)
is equal to the number of components of a mixed spinor with q − 1 contravariant indices
and p − 1 covariant ones

1
n(q − 1, 0)n(0, p − 1) = q(q + 1)p(p + 1). (9.180)
4
Thus, the multiplicity of the representation D(p, q) is defined by the expression:

1
n(p, q) = n(p, 0)n(0, q) − n(q − 1, 0)n(0, p − 1) = (p + 1)(q + 1)(p+
2

+q + 2). (9.181)
Now we can establish the multiplicity of any spinor. For example, using the relation
(9.170), we obtain for the spinor Φpmq :
O O M M
D(1, 0) D(1, 0) D(1, 0) = D(0, 0) 2D(1, 1) D(3, 0), (9.182)

where it has been allowed for that the transition from contravariant indices to covariant
indices has taken place during antisymmetrization.
The next step is to determine the explicit form of wavefunctions of unitary multiplets.
Let us consider a space transformed under the representation to be the product of the
fundamental representations
O O O O
U ··· U U∗ · · · U∗ . (9.183)
| {z }| {z }
p times q times

© 2011 by Taylor and Francis Group, LLC

 158 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
l ···l
We introduce the orthonormalized basis ek11 ···kqp . Then an arbitrary vector in this space is
defined by the formula:
k ···k l ···l
Ψ = Ψl11···lqp ek11 ···kqp . (9.184)
The components of the vector Ψ are transformed according to the law
′k ···k m ···m
1
Ψl1 ···l q
p
= Uk1 m1 · · · Ukp mp Ul∗1 n1 · · · Ul∗q nq Ψn11···nq p . (9.185)

The action of the representation generators on the basis vectors is determined by the for-
mula:
q
X p
X
l ···l l ···l l′ lr+1 ···lq
l ···l
Sa(un) ek11 ···kqp = (λa )lr l′r ek11 ···kr−1
p
r
− λTa kr kr′
ek11 ···kqr−1 kr′ kr+1 ···kp , (9.186)
r=1 r=1

where we took into consideration the explicit form of the generators for the contravariant
and covariant spinors of the first rank. Let us denote the numbers of the covariant (con-
travariant) indices, equal to one, two and three by q1 , q2 and q3 (p1 , p2 and p3 ), respectively.
Then from Eq. (9.186) and the particular form of the λa -matrices one can see, that the
basis vector with the given numbers of indices, equal to one, two, and three correspond to
the following eigenvalues of the SU (3)-group invariants

(is) l ···l 1 1 l ···l

S3 ek11 ···kqp = { [q1 − p1 ] − [q2 − p2 ]}ek11 ···kqp , (9.187)
2 2
l ···l 1 1 1 l ···l
Y ek11 ···kqp = { √ [q1 − p1 ] + √ [q2 − p2 ] − √ [q3 − p3 ]}ek11 ···kqp . (9.188)
2 3 2 3 3
(is)
As before, we shall consider the eigenvalues of the operators S3 and Y to be components
l ···l
of two-dimensional vectors on the unitary spin plane, that is, the end of the vector ek11 ···kqp
corresponds to the definite state of the SU (3)-multiplet.
As an example, we examine the irreducible representation D(1, 1), according to which
the mixed spinor of the second rank ψlk with the zero-track is transformed. As we already
know, the vectors
f a = (ψlk )a elk , a = 1, 2, ...8, (9.189)
making up the basis in the space D(1, 1) can be chosen in the form:

)
f 1 = e12 , f 2 = √12 e11 − e22 , f 3 = e21 , f 4 = e13 ,

(9.190)
f 5 = e23 , f 6 = e31 , f 7 = e32 , f 8 = √16 e11 + e22 − 2e33 ,

Then the nonzero components of the vectors (ψlk )a of the basis f a have the following values

1
2 
2
ψ12 = 1; ψ11 = √1 ,
= − √12 ; ψ22 

3

2

3 4 3 5 1 6
ψ21 = 1; ψ1 = 1; ψ2 = 1; ψ3 = 1; (9.191)

7
8
8
8 
ψ11 = ψ22 = √16 , ψ33 = − √26 .

ψ32 = 1;

Using Eqs. (9.187) and (9.188), it is easy to demonstrate that the vectors f a describe the
following isomultiplets of the baryon SU (3)-octet:

I = 1, Y = 0 : Σ+ = f 1 , Σ0 = f 2 , Σ− = f 3 ; 

I = 21 , Y = 1 : p = f 4 , n = f 5; 
1 − 6 0 7
(9.192)
I = 2 , Y = −1 : Ξ = f , Ξ =f ; 

0 8 
I = 0, Y = 0 : Λ = f .

© 2011 by Taylor and Francis Group, LLC

 Hadron families 159

Now we introduce the quantity
a

Bkl = ψlk f a,
whose matrix components represent the wavefunction of the baryon octet
 1 0 
√ Σ + √1 Λ0 Σ+ p
2 6
(Blk ) = 
 Σ− − √12 Σ0 + √16 Λ0 n  . (9.193)
Ξ− Ξ0 − √26 Λ0

To obtain a normalized wavefunction, corresponding to the definite particle of the octet, in

Eq. (9.193) its symbol should be replaced by 1 and all the matrix elements Blk unrelated
to this particle must be set equal to zero.
Meson octets have a similar form. Thus, for example, the wavefunctions of the pseu-
doscalar and vector meson octets are defined by the expressions:
 1 0 
√ π + √1 η 0 π+ K+
2 6
(Plk ) = 
 π− − √12 π 0 + √16 η 0 K0   (9.194)
K− K0 − √26 η 0
 
√1 ρ0 + √16 ω 0 ρ+ K ∗+
2
1 √1 ω 0
(Vlk ) = 
 ρ− 0
− √2 ρ + 6
K ∗0  , (9.195)
K ∗− K ∗0 − √26 ω 0
respectively.
It is obvious that any matrix with the zero-track could be expanded in terms of the total
system of the linearly independent matrices λa

Φ = λa Φa ,

where the coefficients Φa present the SU (3)-vector components. This, in its turn, exhibits
the wavefunctions of the baryon or meson octets in the form of the eight-dimensional vectors,
as for example:
1
Ba = Sp(Bλa ). (9.196)
2
Notice, that the P - and V -matrices are Hermitian ones, that is, they transfer to themselves
under the transposition and replacement of particles by antiparticles. Obviously, the baryon
matrices do not have such a property, since baryons and antibaryons form independent
multiplets. The computations, similar to the above, produce the following expression for
the antibaryons octet matrix
 − 
√1 Σ0 + √1 Λ0 Σ− Ξ
2 6
k
− √12 Σ0 + √1 Λ0
 
(B l ) =  Σ+ 6
Ξ0 . (9.197)
p n − √26 Λ0

With expressions for unitary wavefunctions near at hand, one can use the Lagrangian
formalism to describe the behavior of the baryon and meson superfamilies. However, there
is one “but” connected with particle masses in an unitary multiplet. The free Lagrangian
of the baryon octet has the form:
ih k k
i k
L= B l (x)γµ ∂ µ Bkl (x) − ∂ µ B l (x)γµ Bkl (x) − m0 B l (x)Blk (x), (9.198)
2

© 2011 by Taylor and Francis Group, LLC

 160 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where we assumed, that all the baryons possess the same mass m0 . However, in reality
the SU (3)-symmetry has been violated and the particle masses in the multiplet differ from
each other. Consequently, this factor should be necessarily taken into consideration under
accomplishing the exact calculations.
The world is made in such a way, that the weaker the interaction, the less symmetric it is.
The stronger interaction behaves as though it does not observe slight violations of definite
conservation laws. As a result, such an interaction conserves the given physical quantity and
consequently, it is more symmetric. The total interaction between hadrons can be presented
as a sum of a hypothetical super strong interaction (with the SU (3)-symmetry group), the
strong interaction (which violates the unitary symmetry, but conserves the isotopic one), and
last the electromagnetic and weak interactions (which violate the isotopic invariance). The
division of the strong interaction into the super strong and the normal strong interaction, is,
of course, rather conditional. In fact, there is no super strong interaction at all. There are
only high energy regions, where mass differences of particles in multiplets are insignificant.
When the strong, electromagnetic and weak interactions are switched off, the exact SU (3)-
symmetry takes place and all the particles in the unitary multiplet are degenerated in the
mass (m0 is a degeneracy mass). Switching on the strong interaction makes the mass
operator in the free Lagrangians dependent on the isospin and the hypercharge

(Ψ(n) (n) (n) (n)

α , MΨα ) = (Ψα , M0 Ψα )+

+(Ψ(n) (n)
α , ∆MΨα ) ≡ M = m0 · 1
(n)
+ ∆m(n) (I, Y ), (9.199)
(n)
where Ψα denotes the unitary wavefunction written in the form of the column matrix,
the indices n and α characterize the representation and the particle state in a multiplet,
respectively.
Let us consider the concrete case, the baryon octet. From the multiplication rules for the
representations:
O M M M M M
8 8=1 8 8 10 10 27. (9.200)

it follows that the masses of the octet members can be expressed in terms of 6 constants
that are connected with contributions coming from the following multiplets: one from a
singlet, two from octets, two from decouplets, and one from a 27-representation. It is
known from experiments that states with fixed Y , belonging to the given isomultiplet, are
still degenerate on a mass. For this reason, in expansion of the mass matrix in terms of
irreducible representations, only operators with Y = T = 0 appear. Since the 10- and
10-representations contain no members with Y = T = 0, then they do not contribute to M .
Further, following Gell-Mann [25] we assume (and every assumption is good if it brings the
theory in accordance with experiments) that the value of contribution to the mass operator
decreases while the representation dimension grows. In other words in (9.199) the term
m0 · 1(n) should be interpreted as the contribution from the singlet representation (zero-
order approximation on the strong interaction) while that ∆m(n) (I, Y ) should be interpreted
as the contribution from the octet representation (first-order approximation on the strong
interaction). Since the operator ∆M enters into the total set of operators, defining a
system state, then it must commute with operators of the hypercharge and the isospin. It
is possible, provided it is either the third or eighth component of the SU (3)-vector. In the
(un)
unitary spin group there are two independent vectors Sa and Da . Consequently, Eq.
(9.199) can be presented as follows:

(un)
(Ψ(9) (9)
α , MΨα ) = m0 · 1
(9)
+ c1 (Ψ(9)
α , S8 Ψ(9) (9) (9)
α ) + c2 (Ψα , D8 Ψα ), (9.201)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 161

where c1 , c2 are real constants and for ∆M we stopped our choice on the eighth component
of the SU (3)-vector. Employing Eqs. (9.118), (9.121), and (9.129), it is easy to obtain
 
1 (is)2 3
D8 = √ 3Sk − Y 2 − Sa(un)2 . (9.202)
3 3 4

Inserting (9.202) into (9.201), we arrive at Gell-Mann-Okubo mass formula [25,26]

1 2
(Ψ(9) (9)
α , MΨα ) = m0 + AY + B[I(I + 1) − Y ], (9.203)
4
which defines particle masses in all unitary multiplets according to the values of real con-
stants m0 , A, and B. Notice, that the formula (9.203) is valid when particles are described
by the first-order equations. The first-order equations formalism is employed both for par-
ticles with the spin 1/2 and for particles with the spin 0 and 1. However, as a rule, the
second-order equations are most commonly used for bosons. In this case, the Gell-Mann-
Okubo formula acquires the form:
1
M = µ20 + A′ Y + B ′ [I(I + 1) − Y 2 ].
4
There are four different masses in the baryon octet: mN , mΞ , mΣ , and mΛ . Writing Eq.
(9.203) for these masses and excluding the parameters m0 , B and C, we obtain the formula

mΣ + 3mΛ = 2(mN + mΞ ), (9.204)

which is fulfilled with an accuracy of 0.5%.

Switching on the electromagnetic interaction initiates further masses splitting in a unitary
multiplet. To define an electromagnetic addition to particle masses, the division of unitary
superfamilies into the U -spin multiplets proves useful to us. At this point it is appropriate
to recollect the first work in the teaching strategy, which so magnificently covered the very
essence of this problem. It was written in the twelfth century by canon Chue (St. Victoria
Abbey, Paris). Unlike a modern “opus magnum” devoted to the teaching strategy, the work
contained only one sentence: study everything, afterwards you shall see—there is nothing
unnecessary. Since the U -spin multiplets group particles with the equal value of the electric
charge, one can assume that the electromagnetic interaction conserves the U -spin, that is, it
acts identically on all members of a U -multiplet. Let us denote mass corrections to charged
and neutral particles by the symbols δ± and δ0 , respectively, and use the Table 9.1, which
defines the baryon octet division into the U -multiplets. Switching on the electromagnetic
interaction modifies the masses of the members of the two U -doublets
(0) (0)
mp = mN + δ + , mΣ+ = mΣ + δ+ , (9.205)
(0) (0)
mΣ− = mΣ + δ− , mΞ − = mΞ + δ − , (9.206)
where the upper index 0 denotes the particle mass in the isomultiplet before the electro-
magnetic interaction is switched on. Analogously, for the member of the U -spin triplet we
get
(0) (0)
mn = mN + δ 0 , mΣ0 = mΣ + δ0 . (9.207)
(0)
Excluding δ± , δ0 , and mN,Σ,Ξ from Eqs. (9.205)–(9.207), we arrive at Coleman-Glashow
formula [27]:
mΞ− − mΞ0 = mΣ− − mΣ+ + mp − mn , (9.208)
which is fulfilled with an accuracy of ∼ 1.4%.

© 2011 by Taylor and Francis Group, LLC

 162 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Now we proceed to the build-up of the SU (3) invariant interaction Lagrangian for the
baryon octet. Let us restrict ourselves to the case of a baryon-meson interaction. Of course,
it must be a generalization of the isotopically invariant interaction between nucleons and
π-mesons:
L(is) = igN π N α (σk )αβ γ5 Nβ πk , (9.209)
where α, β = 1, 2, k = 1, 2, 3 and γ5 ≡ γ 5 = iγ 0 γ 1 γ 2 γ 3 . The obvious analog of (9.209) is
the expression:
igf B a (Sc(un) )ab γ5 Bb Pc . (9.210)
However, contrary to the SU (2)-group where there was only one independent SU (2)-vector,
(un)
the SU (3)-group has two independent SU (3)-vectors, Sa and Da . Thus, the SU (3)
invariant interaction Lagrangian has the form:
(un) (un)
L(un) = L1 + L2 = igf B a (Sc(un) )ab γ5 Bb Pc + igd B a (Dc )ab γ5 Bb Pc =
i i
= gf fcab B a γ5 Bb Pc + igd dcab B a γ5 Bb Pc = − Sp([Bγ5 , B]P ) + Sp({Bγ5 , B}P ). (9.211)
4 4
In the next to last line of Eq. (9.211) we have used matrix representations of the operators
(un)
Sa and Da .
Everything, we have yet known about the SU (3)-symmetry and about symmetry in gen-
eral, concerns only kinematic aspects of symmetry. However the real power of symmetry
reveals itself under investigation of physical systems evolution. For example, symmetry
allows us to obtain relationships between the cross sections of different reactions without
using the motion equations. It will be very instructive to show how to find such relationships
by the example of the SU (3)-symmetry.
Suppose, we study the reaction:

1 + 2 → 3 + 4. (9.212)

We shall denote the wavefunction of the initial (final) state by Φi (Φf ). Since particles are
free before and after interaction, then:

Φi = Φ1 (p1 , J1 )Φ2 (p2 , J2 ), Φf = Φ3 (p3 , J3 )Φ4 (p4 , J4 ),

where the wavefunction Φn (pn , Jn ) (n = 1, 2, 3, 4) depends on the momentum and the spin
of nth-particle. An amplitude of a transition from the initial to the final state is given by
the expression:
Mi→f = Φ†4 Φ†3 SΦ1 Φ2 , (9.213)
where the operator S derived the name the S-matrix is defined by the form of the interaction
Lagrangian for particles 1, 2, 3, and 4. In the case when we deal not with particles scattering,
but with the scattering of SU (3)-multiplets, the wavefunctions and the S-matrix contain
dependence on the unitary spin as well. Since the spin matrices and the unitary spin ones
act in different spaces, they commute with each other. It leads to the fact that the spatial
and unitary parts are separated. Thus, the wavefunction of the l-multiplet is the production
of the spatial part Φl (p, J), which is common for all multiplet particles, and the unitary
(un)
part Φl . Then, the amplitude of the reaction (9.212) can be presented in the following
form: XX
Mi→f = Rj Γi Tji , (9.214)
j i

where Γi are SU (3)-invariant combinations, Rj are relativistically invariant spin combi-

nations and Tji are relativistically invariant functions on the momenta of the initial and

© 2011 by Taylor and Francis Group, LLC

 Hadron families 163

final particles. To define the set Γi , all the possible independent unitary invariants (SU (3)-
scalars) must be composed out of the unitary wavefunctions of particles, entering into the
reaction. The number of independent SU (3)-scalars might be reduced in case we demand
that the amplitude of the reaction (9.212) is invariant under other transformations (P , C,
T , etc.).
Let us come back to the case of baryons and consider elastic scattering of the pseudoscalar
meson octet by the baryon octet:

P + B → P ′ + B′, (9.215)

where for the sake of convenience the final states of the meson and baryon octets are marked
with the primes. To define the number of independent
N unitary scalars, corresponding to the
reaction (9.215) we use the fact that the 8 8-products of the wavefunctions, both of the
initial state and of the final one, are separated into irreducible representations according
to the formula (9.198). Since the unitary symmetry allows only transitions between equal
representations,∗ then from (9.200) it follows that in the amplitude of the reaction (9.215)
the matrix elements of only the following eight transitions:

1 → 1, 8 → 8, 8′ → 8, 8 → 8′ ,
(9.216)
8′ → 8′ , 10 → 10, 10 → 10, 27 → 27.

are not equal to zero. In other words, the amplitude of the process (9.215) contains eight
independent unitary scalars Γi , which must be built from products:
′k ′m
B l P n Bpi Pjr

by means of summation over all the possible pairs of the upper and down indices. Summa-
tion over one pair of indices for two matrices means multiplication of these matrices:
m
Blk Pkm = (BP )l ,

while summation over the second pair of indices gives the track of this product:
m
(BP )m = Sp (BP ) .

From products of four unitary wavefunctions eight independent invariants, differing from
each other by summation order, can be constructed
′ ′ ′ ′ 
Γ1 = Sp(B B)Sp(P P ), Γ2 = Sp(B P )Sp(BP ), 

′ ′ ′ ′ 

Γ3 = Sp(B P )Sp(BP ), Γ4 = Sp(B BP P ),
′ ′ ′ ′ (9.217)
Γ5 = Sp(B BP P ), Γ6 = Sp(BB P P ), 

′ ′ ′ ′ ′ ′ 
Γ7 = Sp(BB P P ), Γ8 = Sp(B P BP ) − Sp(B P BP ).

The number of independent quantities Γi decreases, if the theory invariance under the time
inversion is taken into account. The process (9.215) contains the same baryon and meson
octets in the initial and final states, that is, B ′ = B and P ′ = P . Consequently, at the
T -inversion it turns into itself. Since at the time inversion the initial particles become final
ones and vice versa, then the unitary wavefunctions are transformed according to the law:
l k l k
Blk → B k , B l → Bkl , Plk → P k , P l → Pkl . (9.218)

∗ This is caused by the orthogonality of representations.

© 2011 by Taylor and Francis Group, LLC

 164 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Let us find out, how the Γi quantities behave under the T -inversion. We start with the
last combination in Eq. (9.217). Since:
k
j m m j m j
Sp(BP BP ) = B i P k Bm Pji → Bki Pm
k k i
B j P i = B j Pm Bk P i = Sp(BP BP )

and
Sp(BP BP ) → Sp(BP BP ),
takes place, then under the time inversion Γ8 reverses its sign and, as a result, proves to
be forbidden by the theory T -invariance. It can be shown in a similar way, that all the
remaining combinations in (9.217) are T -invariants. Now we can proceed to establishing
connections between the cross sections both of elastic and of nonelastic scattering of particles
entering into the baryon and meson octets. Consider the processes:

π + + p → π + + p, (9.219)

π − + p → π 0 + n, (9.220)
+ + +
π +p→K +Σ , (9.221)
+ +
K + p → K + p, (9.222)
K − + p → K − + p, (9.223)
− −
π + p → π + p, (9.224)
K − + p → π − + Σ+ . (9.225)
From the explicit form of the wavefunctions of the baryon and meson octets (see Eqs.
1 1
(9.193)–(9.195)) it follows, that for the process (9.219) B13 = B 3 = P12 = P 2 = 1, and all
the other components are equal to zero. As a result, from seven independent unitary scalars
Γi only two are different from zero

Γ1 = Γ7 = 1.

Consequently, the amplitude of the process (9.219) is defined by the expression:

X
M (π + p → π + p) = (T1j + T7j ) Rj . (9.226)
j

In an analogous way the expressions for the amplitudes of the processes (9.220)–(9.225)
can be obtained. Excluding quantities Tij Rj from them we arrive at the following relations
between amplitudes:
√
M (π + p → π + p) − M (π − p → π − p) = 2M (π − p → π 0 n), (9.227)

M (K − p → K − p) − M (π − p → π − p) = M (K − p → π − Σ+ ), (9.228)
+ + + + + + +
M (K p → K p) − M (π p → π p) = M (π p → K Σ ). (9.229)
Eq. (9.227) could be obtained even from the isotopic invariance while Eqs. (9.228) and
(9.229) are consequences of only the unitary symmetry. Passing in (9.228)–(9.229) from the
amplitudes to the cross sections, we obtain the so-called triangle inequalities:
p p p
σ(K − p → π − Σ+ ) ≥| σ(K − p → K − p) − σ(π − p → π − p) |, (9.230)
p p p
σ(π + p → K + Σ+ ) ≥| σ(π + p → π + p) − σ(K + p → K + p) | . (9.231)

© 2011 by Taylor and Francis Group, LLC

 Hadron families 165

It is also possible to demonstrate, that


M (K − p → π + Σ− ) = M (K − p → K 0 Ξ0 ),
(9.232)
M (π − p → K + Σ− ) = M (K 0 p → K + Ξ0 ).

takes place. This, in its turn, leads to the equalities between the cross sections:

σ(K − p → π + Σ− ) = σ(K − p → K 0 Ξ0 ),
(9.233)
σ(π − p → K + Σ− ) = σ(K 0 p → K + Ξ0 ).

However, the above-mentioned considerations apply only to the idealized case of the
exact unitary symmetry. In reality, the SU (3)-symmetry is very approximate, as there
is sizeable mass splitting in the SU (3)-multiplets. Because of this, to compare relations
obtained with experiments, symmetry violation must be taken into account. In the first
approximation one could allow for only mass splitting, that is, substitutes physical masses
into the Lagrangians, but considers the interaction Lagrangians to be SU (3) invariant as
before. This procedure leads to a quite good accord with the experiment. The next stage in
taking into account unitary symmetry violation, is to introduce corrections to the interaction
Lagrangians, which brings in the appearance of additional parameters in the theory. Due
to a large number of such parameters, calculations become ineffective.
(un)2
The wavefunctions of unitary multiplets are eigenfunctions of the Casimir operators Sa
(un)
and Sa Da . Thus, every irreducible representation can be marked with the eigenvalues of
these operators. Labeling of the SU (3)-multiplets with only the eigenvalues of the operator
of the unitary spin square is generally accepted. From Eq. (9.186) it is easy to obtain
1
Sa(un)2 D(p, q) = gD(p, q) = [p + q + (p2 + pq + q 2 )]D(p, q). (9.234)
3
Then, for the unitary multiplets of the lowest dimensions we have


 0 for 1,
4
for 3, 3,
g= 3 (9.235)

 3 for 8,

6 for 10.

The classification of hadrons according to unitary multiplets resembles one of chemical

elements in the Mendeleev periodic table. Like the Mendeleev table, the hadrons classi-
fication unostentatiously points up (but again for the experienced mind) to a composite
structure of hadrons. In this sense the SU (3)-group of the isospin and the hypercharge
has served its historical mission. It has prepared all the conditions for the next step up
the Quantum Stairway, going out on the quark-lepton level of matter structure. However
unlike the Moor,∗ who had to disappear after his task was realized, the SU (3)-symmetry,
as we shall see later, settles down firmly in physics of the strong interaction. True, it must
slightly change its role.

∗ Here a personage of the Othello tragedy is born in mind.

© 2011 by Taylor and Francis Group, LLC

 10
Quark “atoms”

It was now the nightingale’s turn. But he came a cropper

at the very first note. He sang of a servant’s delight
at God having sent him a master, of noble eagles that
never grudged servants money for vodka. In a word, the
nightingale did his best to sing in the servile vein, but
try as he would, the “art” that lived in him refused
to be curbed. The bird himself was servile from beak to
tail (he even picked up a second-hand necktie somewhere
and curled his hair) but his “art” just could not be kept
within the bounds of servility and broke through again
and again.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

10.1 Hypothesis of fundamental triplets

In 1964 M. Gell-Mann [29] and G. Zweig [30] independently from each other hypothesized
that all hadrons are built of three particles of the unitary triplet. Wishing to emphasize
the unusual properties of new blocks of matter, Gell-Mann called them “quarks.” The term
was borrowed from Finnegans Wake by James Joyce. If one compares this novel with War
and Peace by L. Tolstoy, then the first thing that comes to mind is that Finnegans Wake
was written in at least 26-dimensional space-time, which never experienced the joy of com-
pactification. During the act a protagonist Humphrey Chimpden Earwicher is constantly
changing his appearance. He is reincarnated at one moment into Mark, the King of Corn-
well, at another into his sons, Sham and Shaun, and so on. Earwicher’s children (he has
a daughter as well) are far from being simple and they also can be transformed into their
father. There is an episode in the novel when the protagonist being reincarnated into King
Mark, sends his nephew, the knight Tristan, by a wedding boat to bring the king’s bride
Isolde. As expected, during their travels Tristan and Isolde happen to have been struck
down by Cupid’s arrow practically on the spot. The seagulls circling above the ship are
completely informed about the events taking place on the ship, as demonstrated by their
song starting with the words: “Three quarks for Mister Mark.” If one distracts from the
remaining part of the song, then the above-mentioned phrase can be unambiguously viewed
as a prediction of the fundamental triplet. Setting the imagination free, one can assume
that Mister Earwicher with his transformations chain reproduces the hadrons spectrum and
his children are nothing else but three quarks. However, the whole content of the bird’s
opus suggests that the phrase “three quarks” may be treated as evidence that the old king
was deceived triply. Time will show whether that is the excited indication on the analogous
fate of the quark hypothesis, the development of which physicists have been devoted to
already for half a century.

167
© 2011 by Taylor and Francis Group, LLC
 168 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Earlier we demonstrated how to build singlets, octets, and decouplets from the unitary
triplets with precisely the same isotopic structure as experimentally observed unitary mul-
tiplets of hadrons. It appeared that the hypothesis of the fundamental triplets was obvious
even for philosophers who are constantly loitering around the building of modern physics
and whose basic activities consist in abusing the terminology specially invented by just
them. However, such models were constantly neglected due to the inertia of thinking be-
cause they demanded fragmentation of the electric and baryon charges of particles entering
into the fundamental triplet.
Originally a quark family included only three particles and three corresponding antipar-
ticles. Two fundamental triplets, which we denote by 3 and 3, had the following form:
   
q1 u
q =  q2  =  d  , q = (q 1 , q2 , q 3 ) = (u, d, s), (10.1)
q3 s

where symbols u, d, and s (“up,” “down”, and “strange”) are used for components of the
quark triplet qα (α = 1, 2, 3). It is also possible to say, that a quark dwells in three aroma
(flavor) states and attribute the flavor meaning to the index α. u- and d-quark form the
isotopic doublet (I3u = 1/2 and I3d = −1/2), while s-quark forms the isotopic singlet. Only
the s-quark has a nonzero strangeness (s = −1). Since we are going to construct particles
with arbitrary spin values, then the quarks must have spin 1/2. We also ascribe to all
the three quarks the baryon charge 1/3. Then it becomes obvious, that in the unitary spin
plane the quark and antiquark triplets are denoted by the same triangles as the fundamental
triplet and antitriplet of the SU (3)-group, respectively (see Figs. 10.1). In its turn, from the

Y Y
2/3
s

1/3
d u

I3 I3
−1/2 1/2 −1/2 1/2

u d
−1/3

s −2/3

FIGURE 10.1
The quark and antiquark triplets.

Gell-Mann-Nishijima formula it follows that the quark charges in units of |e| are fractional
quantities
Qu = 2/3, Qd = Qs = −1/3.
Further, we should define the quark contents of hadrons. It is evident, that mesons must
contain an even number of quarks, while baryons must contain an odd number of quarks.

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 169

Let us assume that all mesons are built from quark-antiquark pairs

Mik = q i q k ,

and all baryons are built from three quarks

B ikl = q i q k q l .

Now hadrons made of the quarks should be placed into the corresponding unitary multi-
plets. In our forthcoming design activity we are going to use the composition law of the
unitary spins as our basic instrument. This law must be a generalization of the composition
rule for the ordinary spins (2.23). Let us formulate this law in such a way that it become
applicable to any spin, whether it is the ordinary, isotopic, unitary spin, and so on. Thus,
to obtain all the possible spin states of a coupled two-particle system, it is necessary:

(i) to superpose the center of a spin diagram of the first particle on every state of a spin
diagram of the second particle;
(ii) to mark the states obtained;
(iii) to single out diagrams with the equal spin value out of all states.

Let us begin the hadrons construction process with a meson sector. Making the center of
diagram 3 coincident with every out of three states of the diagram 3, we obtain nine possible
states of the 33-system (Fig. 10.2). These states are separated into the SU (3)-octet and

Y Y Y

ds us

d u ss, dd, uu ss, dd, uu

du ud
I3 I3 I3
s

su sd

3⊗3 = 8 ⊕ 1

FIGURE 10.2
The meson multiplets.

the SU (3)-singlet. Three of them, namely, I,II,III being linear combinations uu, dd, and
ss falls in one and the same point of the diagram with I3 = Y = 0. Consequently, they
demand our great attention. The unitary singlet I is nothing else but
 

u
I = const u, d, s  d  ,
s

© 2011 by Taylor and Francis Group, LLC

 170 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where const is defined from the normalization condition. It gives

1

I = √ uu + dd + ss . (10.2)
3

Two remaining states belong to the octet. We assign the state II to the isospin triplet,
whose contents could be reconstructed by the replacement p → u and n → d in the formulas
(9.62). So,
(du, II, −ud),

where
1

II = √ uu − dd .
2
Next, we find that the isospin singlet, the state III, being orthogonal to I and II is defined
by the expression:
1

III = √ uu + dd − 2ss .
6
The spin J of a composite meson is defined by

J = S + L,

where S is the spin of a (qq)-system being equal 1 or 0, L is a relative orbital moment of

quarks. Since q and q have opposite internal parities, the meson parity is determined by
the expression
P = −(−1)l ,

where l is an orbital quantum number and we have taken into account that under the space
inversion θ → π − θ, ϕ → ϕ + π the wavefunction angular part Ylml (θ, ϕ) of a qq-pair
acquires the factor (−1)l . It should also take into consideration that meson multiplets
possess the definite charge parity (the qq-system is neutral). To find a proper value of the
charge conjugation operator C we must carry out the replacement q ↔ q, interchange the
quarks, and exchange their spin directions. This operation results in

C = −(−1)s+1 (−1)l = (−1)s+l ,

where s is a spin quantum number and the common minus sign is connected with the per-
mutation of fermions, while the factor (−1)s+1 is associated with the symmetry properties
of the qq-pair spin states (see Eqs. (9.62) (9.63) ). In such a way under l = 0, 1 the
three quarks model predicts the existence of meson octets and singlets with the following
quantum numbers

0−+
l = 0, j = 0, 1 → −−
1

1+− 

2++

l = 1, j = 0, 1, 2 → ++
1  
0++


where the symbol J P C is used for marking the states. The states with l = 0 can be viewed
as the basic ones, while the states with l > 0 can be viewed as the orbital excitations. It

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 171

should be stressed that the quark filling of meson nonets with different values of J P C is the
same. Below we give the meson nonet with J P C = 0−+ consisting of the octet
0

K 0 = ds, K + = us, π + = ud, K = sd,  



1

K − = su, π − = du, π 0 = √ uu − dd , (10.3)
2 

1


η = √ uu + dd − 2ss . 

6
and the singlet r
′ 1

η = uu + dd + ss . (10.4)
3
So, all the discovered mesons can be placed into the SU (3)-multiplets qq. Up to 1971, when
the first reliable data confirming the existence of quarks into hadrons were obtained, such
tests have been the main argument in favor of the quark hypothesis.
We are coming now to the investigation of the quark structure of baryons. First, we
combine two quarks. We plot the already familiar to us result
O M
3 3=6 3

in Fig. 10.3. We remind that the sextet 6 is symmetric with respect to transposition of two

Y
dd ud uu ud
2/3
d u d u
1/3
0 ⊗ = ⊕
−1/3 ds us
ds us
−2/3
s s
−1
−4/3
ss

FIGURE 10.3
N L
The graphical interpretation of the formula 3 3 = 6 3.

quarks while the triplet 3 is antisymmetric. In Fig. 10.3 we specify only the quark filling
in every point of the diagram. For the wavefunction of particles to be exactly defined we
must take into account the symmetry properties of multiplets. So, the state ud belonging
to the sextet is described by:
1
ψs (ud) = √ (ud + du) ,
2
whereas the wavefunction of the analogous state in the triplet has the form:
1
ψt (ud) = √ (ud − du) .
2

© 2011 by Taylor and Francis Group, LLC

 172 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Let us add one more quark. The final result of decomposition

O O M O O M O M M M
3 3 3 = (6 3) 3 = (6 3) (3 3) = 10 8 8 1, (10.5)

is displayed in Fig. 10.4, where the octet following the decuplet appears under summarizing
N
the unitary spins of the sextet and the triplet, while the second octet comes from 3 3.

Y
ddd udd uud uuu udd uud udd uud
1

dds uus
0 dds uus dds uus
uds ⊕ uds ⊕ uds ⊕
uds
−1 dss uss
dss uss dss uss

−2
sss

FIGURE 10.4
N N L L L
The graphical interpretation of the formula 3 3 3 = 10 8 8 1.

Eqs. (9.164), (9.165), (9.168), and (9.169) prove to be useful under finding the wave-
functions of the multiplets corresponding to the division (10.5). These equations could be
presented in the following form:
1 i k l

B ikl (10) = q q q + qi ql qk + ql qi qk + ql qk qi + qk ql qi + qk qi ql , (10.6)
6
1 i k l

B ikl (1) = q q q − qi ql qk + ql qi qk − ql qk qi + qk ql qi − qk qi ql , (10.7)
6
1 i k l

Bi[kl] (8) = q q q + qk qi ql − qi ql qk − ql qi qk , (10.8)
3
1 i k l

B[ik]l (8) = q q q + qiql qk − qk qi ql − qk ql qi , (10.9)
3
which is more convenient for practical use. For the multiplets with mixed symmetry we are
constrained by the indication of only antisymmetry on indices and it was done with the help
of the square brackets. The choice of the quantities B [ik]l (8) and B i[kl] (8) is not unique,
that is, they are ambiguous. Moreover, one may be convinced by the simple check that
B [ik]l (8) and B i[kl] (8) are not orthogonal to each other. It is evident that this circumstance
could become a source of trouble at a later time. Let us correct this deficiency. With the
help of Jacobi identities:

B i[kl] (8) + B l[ik] (8) + B k[li] (8) = 0, (10.10)

B [ik]l (8) + B [li]k (8) + B [kl]i (8) = 0, (10.11)

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 173

one may build the mutually orthogonal representations:

2

B ikl (8) = √ q i q k q l + q i q l q k + q k q i q l + q k q l q i − 2q l q i q k − 2q l q k q i , (10.12)
3 3
1 i k l

B ′ikl (8) =
q q q − qi ql qk + qk qi ql − qk ql qi . (10.13)
3
Notice, that the constant quantities standing in front of the parenthesis are not the normal-
izing coefficients. They are no more than the coefficients of splitting B αβγ into irreducible
tensors, that is,
1
2 1
B ikl = B ikl (10) + B ikl (1) + √ B ikl (8) + B ′ikl (8).
6 3 3 3

To put this another way, to obtain the unitary spin wavefunctions from the quantities (10.6),
(10.7), (10.12), and (10.13) one should additionally allow for the normalization condition.
As an example, we use Eqs. (10.7), (10.12), and (10.13) in order to get the wavefunctions
of baryons with the quark filling udd (i = 1, k = l = 2). These functions have the form:
1
ψ(udd)(10) = √ (udd + dud + ddu) , (10.14)
3
for the decuplet,
1
ψ(udd)(8) = √ (2udd − dud − ddu) , (10.15)
6
for the first octet, and
1
ψ ′ (udd)(8) = √ (dud − ddu), (10.16)
2
for the second octet.
We denote the relative angular moment of two quarks by L1 and the angular moment
of the third quark as related to the masses center of the first two quarks by L2 . Then the
total angular moment of three quarks is determined by the expression:

L = L1 + L2 .

Assuming the quark parities to be positive we find that for the low-laying baryon states
(L = L1 = L2 = 0) the parity is equal to +1. According to the vector summation rules the
resulting spin of the three quarks may equal either 3/2 or 1/2. Thus, the low-laying baryons
multiplets are characterized by the values 3/2+ and 1/2+ . Baryons having the lowest mass
values are precisely placed into the decuplet 3/2+ and the octet 1/2+ . Heavier baryons must
have l = 1, that is, either l1 = 1 and l2 = 0 or l1 = 0 and l2 = 1 (in both cases their parity
is negative −1). And really, existing baryon resonances are finely placed into the following
− − − − −
multiplets: singlets and decuplets with 21 and 32 , octets with 12 , 32 and 25 .

10.2 X-ray photography of nucleons

Today, building the quark model of hadrons could be viewed as some kind of analog of
establishing the nucleon structure of atomic nuclei. However, initially the quark model was
considered the only formal scheme that was very convenient to systematize hadrons. At the

© 2011 by Taylor and Francis Group, LLC

 174 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

end of the 1960s physicists had obtained, at their disposal, new possibilities for investigating
the structure of hadrons. The created sources of high-energy electrons probed distances up
to 10−15 cm, that is, on two orders smaller than the hadron size. To use electrons is
convenient for two reasons. First, they are structureless particles and, second, they do not
participate in a strong interaction. Since the electromagnetic interaction of point particles
has been studied thoroughly, the theoretical analysis of the experimental results is greatly
facilitated. At that time it was already known that hadrons have specific structures that
are described by electromagnetic formfactors. By formfactors we understand the function
characterizing the space distribution of the electric charge and multipole moments inside
hadrons (further, for the sake of simplicity, we shall talk about the magnetic dipole moment
only). Thus, carrying out the “Roentgen” of a proton with the help of scattered electrons
one may investigate the proton structure more carefully and, by doing so, one gets down to
the experimental checkout of the quark hypothesis.
The first stage in similar kinds of investigations consists of the analysis of elastic scatter-
ing. Consequently, we should start with the reaction

e− + p → γ ∗ , Z ∗ → e− + p, (10.17)

where the intermediate step in (10.17) means that interaction is performed by exchanges
of the virtual photon and Z-boson. For the sake of simplicity, we shall take into account
the photon exchange only and shall be constrained by the second order of the perturbation
theory. So, our task is to define a differential cross section of the process (10.17).
To calculate cross sections of elementary particle processes in quantum field theory a
method of Feynman diagrams [30] exists. The successive presentation of this method will
be given in Part IV. Here we set forth some preliminary information to understand the
calculation technique.
Feynman diagrams, or Feynman graphs, outwardly resemble displays of trajectories of all
particles taking part in an interaction. The classical description, however, is not applicable
in this case, so that interpretation of graphs as classical trajectories is incorrect, one can
speak only of outer similarity. By the example of nonrelativistic quantum mechanics we
became aware of the fact, that only in the case of the most simple potentials we managed
to obtain exact solutions for quantum systems. For this reason, the perturbation theory
method is the main mathematical method in the quantum theory. The essence of the
perturbation theory method is most transparent in the Green function formalism. So, if
a system wavefunction in an initial state ψ(ri , ti ) is known, then its value at an arbitrary
moment of time t = tf in the point with coordinates rf is given by the expression:
Z
ψ(rf , tf ) = G(rf − ri , tf − ti )ψ(ri , ti )d4 xi . (10.18)

The Green function in nonrelativistic quantum mechanics is represented by the series in the
perturbation Hint (r, t)

G(rf − ri , tf − ti ) = G(0) (rf − ri , tf − ti )−

Z
i
− G(0) (rf − r1 , tf − t1 )Hint (r1 , t1 )G(0) (r1 − ri , t1 − ti )d4 x1 +
~
 2 Z Z
i
+ − d4 x1 G(0) (rf − r1 , tf − t1 )Hint (r1 , t1 )G(0) (r1 − r2 , t1 − t2 )Hint (r2 , t2 )×
~

×G(0) (r2 − ri , t2 − ti )d4 x2 + ....., (10.19)

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 175

where G(0) (r, t) is the Green function in zero approximation. The expression (10.19) brings
us to the idea that the evolution of a quantum system can be plotted as a chain of transitions
from the initial state to the final one through the totality of points (vertices) corresponding
to interaction acts. In so doing the number of vertices corresponds to the perturbation
theory order. These are nonrelativistic prerequisites for appearance of the Feynman diagram
method. This method is also unambiguously connected with the step-by-step inclusion of
every order of the perturbation theory. The diagrams topology is completely defined by the
form of the interaction Lagrangian. By the aspect of the Feynman diagram, it is immediately
possible to write down the corresponding expression for a scattering amplitude, that is, there
is a set of rules, the so-called Feynman rules connecting any diagram element with particle
characteristics. To depict a free particle one or another line is introduced (which is, of course,
only a graphic symbol of particles propagation), while lines knot (vertex) corresponds to
particles interaction. External lines describe real initial and final particles and internal ones
do virtual particles. Since particles in initial and final states are considered to be free (it
is guaranteed by experiment conditions), then real particles correspond to wavefunctions,
which are solutions of free equations of corresponding fields. Virtual particles, describing
interaction, are put in correspondence of the Green functions of the equations, which they
would satisfy, if they were real. These functions will be referred to the propagators. A
coupling constant, characterizing thep given interaction, must be in every diagram vertex.
In the case of the QED it equals e2 /(4π~c). Moreover, in every vertex the momentum
conservation law must be taken into consideration (interaction, taking place at a vertex,
can take place at any space point, so that ∆xi = ∞, and it means, that a momentum
is precisely defined). In every vertex of “charge” (electric, baryon, lepton, strange, and
so on) conservation laws that are valid for the given interaction, must be fulfilled. The
Feynman diagrams can be plotted in the coordinate and momentum spaces, in the latter
case the four-dimensional momentum of a corresponding particle is ascribed to every line.
Let us display bosons with a wavy line, fermions with a solid line. Sometimes a symbol of
a particle will be also mentioned near every line. We agree to direct time axes from the
left to the right in all the diagrams. In fermion lines and ones corresponding to charged
bosons we put arrows that denote a particle propagation direction. Arrows, directed to the
time axis denote particles, while antiparticles are indicated by arrows being opposite to the
time axis direction, that is, in the Feynman diagram formalism antiparticles are moving
backwards in time. The easiest way to understand that is to address the holes theory that
was proposed by Dirac to overcome difficulties connected with the appearance of negative
energies. In this theory the states with the positive energy and whose dependence on time
has the form exp (−iEt), are identified with the electrons, while the negative energy states
with the wavefunctions ψ ∼ exp (iEt) are identified with the positrons. It is obvious that
the change of the time direction leads to changing the sign in the exponent exp (iEt), that
is, to solutions with the positive energy.
In the QED the interaction Hamiltonian is given by the expression:

Hint = eψ l (x)γµ ψl (x)Aµ (x), (10.20)

(l = e, µ, τ ), that is, Hint has the trilinear structure. It means that in every vertex three
lines are encountered, namely, two fermion (electron and/or positron) and one photon lines.
As this takes place, the electric charge conservation law demands direction invariability of
the fermion line over its length.
The Feynman diagram corresponding to the process (10.17) is given in Fig. 10.5. The
circle in the vertex, describing the interaction of the virtual photon with the proton, stresses
the circumstance that the proton, unlike the electron, has an internal structure. When
the proton were the point particle with the charge |e| and the magnetic moment e~/2mpc

© 2011 by Taylor and Francis Group, LLC

 176 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 10.5
The Feynman diagram for the process e− p → γ ∗ → e− p.

predicted by the Dirac theory, then the cross section of the elastic electron-proton scattering
would follow from the cross section of the process
e− + µ+ → e− + µ+ , (10.21)
under changing the muon mass by the proton one. It proves to be convenient to calculate
the cross section of the reaction (10.21) and then, using the obtained result, define the
cross section of the elastic electron-proton scattering taking into consideration the proton
composite structure.
Before proceeding to researching the reaction (10.21), let us agree about the choice of
measurement units. The reasonably chosen units comprise a convenient tool under the
description of a definite phenomena region. The fundamental constants of the quantum
field theory (QFT) ~ and c enter into the majority of theory equations. One may “unload”
the QFT formulae when one sets these constants to be equal to 1, that is, one chooses the
action quantum ~ as an action unit and does the light velocity c as a velocity unit. The
system of units with ~ = c = 1 derives the name of the natural system of units (NSU). The
description of the NSU is given in Appendix 1. Further, unless otherwise specified, we shall
use the NSU over the course of the book. It is self-evident that when comparing between
theory and experiment we should pass to one of an ordinary system of units in formulae
obtained.
According to the formulae to be obtained in Section 18, in the momentum representation
the differential cross section of the process (10.21) is given by the expression
(e) (µ)
2
d3 pf d3 pf (µ) (e) (µ) (e)
dσ =| Mif | δ (4) (pi + pi − pf − pf ), (10.22)
(2π)2 | v(e) |
(e)
where j0 is an initial density of the flux of particles participating in the reaction,∗ pi and
(e) (µ) (µ)
pi (pf and pf ) are four-dimensional momenta of initial and final electrons (muons), and
Mi→f is a matrix element defined by the Feynman rules
   " #
s s
(e) m e ν (e) m e −ig ντ
Mi→f = ψ(pf ) (e)
 (−ieγ ) ψ(pi )
(e)

(e) (e)
×
Ef Ei (pf − pi )2

∗ In the lab frame (the initial muon is in rest) j0 =| v(e) | (v(e) is an initial electron velocity).

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 177
   
s s
(µ) mµ  τ  (µ) mµ 
× ψ(pi )

(µ)
(−ieγ ) ψ(pf ) (µ)
. (10.23)
Ei Ef
p
Let us elucidate the origin of the factors belonging to Mi→f . The factors ψ(pf ) m/Ef and
p
ψ(pi ) m/Ei are associated with the leptons in the final and initial states, respectively. The
factors −ieγν and −ieγτ correspond to two vertices in which the electromagnetic interaction
of the point-like leptons occurs; ν is the polarization index of the photon created in the
electron vertex, τ is the polarization index of the photon annihilated in the muon vertex.
The factor standing in the square bracket describes the virtual photon propagation between
the electron and muon vertices. The order of writing the spinor matrices in (10.23) is
determined by the direction of the diagram detour, which is opposite to the fermion lines
direction, that is, the diagram detour is performed from the final fermion state to the initial
one (for antifermions, on the contrary).
When we are not interested by the particles polarization in the initial and final states,
then in | Mif |2 one should fulfill averaging over the initial polarizations and summing over
the final ones. It means that we must pass to the quantity
1X X (e) (e) (µ) (µ)
|Mif |2 = g0 | [ψ α (pf )(γ ν )αβ ψβ (pi )][ψ m (pf )(γν )mn ψn (pi )] |2 , (10.24)
4
i,f i,f

where
e4 m2e m2µ
g0 = (µ) (e) (µ) (e)
,
4q 4 Ei Ei Ef Ef
(e) (e) (µ) (µ)
qν = (pi − pf )ν = (pf − pi )ν and the factor 1/4 appears at the cost of averaging
over polarizations of the initial electron and muon. Since in Eq. (10.24) the matrix indices,
which define the multiplication order, are represented in explicit form, then the factors,
entering into this expression, may be interchanged arbitrarily. Grouping the electron and
muon parties separately, we obtain
1X X (e) (e) (e) (e)
|Mif |2 = g0 [ψ α (pf )(γ ν )αβ ψβ (pi )ψ α′ (pi )(γ τ )α′ β ′ ψβ ′ (pf )]×
4
i,f i,f

(µ) (µ) (µ) (µ)

×[ψm (pf )(γν )mn ψn (pi )ψ m′ (pi )(γτ )m′ n′ ψn′ (pf )] =
X (e) (e) (e) (e)
= g0 [ψβ ′ (pf )ψ α (pf )(γ ν )αβ ψβ (pi )ψ α′ (pi )(γ τ )α′ β ′ ]×
i,f

(µ) (µ) (µ) (µ)

×[ψn′ (pf )ψ m (pf )(γν )mn ψn (pi )ψ m′ (pi )(γτ )m′ n′ ]. (10.25)
To summarize over the spin lepton states (the factor 1/2 converts such summarizing into
averaging) is fulfilled with the help of the relations
 
X ǫ ǫp̂ + m
ψαǫ (p)ψ β (p) = , (10.26)
2m αβ

where p̂ = γν pν , ǫ = 1 for particles and ǫ = −1 for antiparticles (the derivation of (10.26)

will be done in Part III). Thus, Eq. (10.25) may be put down as:

1X
|Mif |2 = g0 Lντ Mντ , (10.27)
4
i,f

© 2011 by Taylor and Francis Group, LLC

 178 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where we shall call Lντ and Mντ as the electron and muon tensor, respectively. For the
electron tensor we have
X (e) (e) (e) (e)
Lντ = ψβ ′ (pf )ψ α (pf )(γ ν )αβ ψβ (pi )ψ α′ (pi )(γ τ )α′ β ′ =
i,f

(e)
! !
(e)
p̂f + me ν p̂i + me
= (γ )αβ (γ τ )α′ β ′ =
2me 2me
β′ α βα′

1 (e) (e) 1 (e) (e)

= Sp[(p̂f + me )γ ν (p̂i + me )γ τ ] = Sp[γ ν (p̂i + me )γ τ (p̂f + me )], (10.28)
4m2e 4m2e
where the symbol “Sp” means the operation of taking the matrices track and the cyclicity
property
Sp[ABCD] = Sp[DABC] = Sp[CDAB] = Sp[BCDA]
has been allowed for. As far as the muon tensor is concerned, the analogous operations give

Mντ → Lντ (p(e) → −p(µ) , me → mµ ). (10.29)

The track in the expressions (10.28) and (10.29) could be found by the application of the
formulae

Sp(γ µ γ ν ) = 4g µν , Sp(γ µ γ ν γ λ γ σ ) = 4(g µν g λσ + g µσ g νλ − g µλ g νσ ) (10.30)

and by considering the fact, that the track of the odd number of the γ-matrices is equal to
zero. The calculations give
(e) (e) (e)ν (e)τ (e)τ (e)ν
Lντ = [g ντ (m2e − pi pf ) + pi pf + pi pf ]. (10.31)

Multiplying the lepton tensors and neglecting the electron mass, we obtain for (10.27) the
following expression in the lab frame:

1X e4 h
(e) (e)
|Mif |2 = (µ) (e) (µ) (e)
m2µ (q 2 + 4Ei Ef )−
4 4
2q Ei Ei Ef Ef
i,f

 
2 (e) (e)
i 2e4 m2µ 2 θ q2 2 θ
−q mµ (Ei − Ef ) = cos − sin , (10.32)
(µ)
q 4 Ei Ef
(µ) 2 2m2µ 2

where θ is the angle between the momenta of the impacting and outgoing electrons.
Under the deduction of (10.32) we have made use of the relations
(e) (e) (e) (e)
q 2 = (pi − pf )2 ≈ −2(pf pi ) ≈

(e)2
(e) (e) (e) (e) θ 2 4Ei sin2 θ2
≈ −2Ef Ei (1 − cos θ) = −4Ef Ei sin =− (e)
, (10.33)
2 2E
1 + i sin2 θ mµ 2
(µ) (e) (e)
q 2 = −2(qpi ) = −2(Ei − Ef )mµ . (10.34)
(µ)
The latter is the consequence of the operation of taking the square of the identity q + pi =
(µ)
pf .
Let us substitute (10.32) into the expression for the differential cross section (10.22).
The presence of the delta functions presents an opportunity to make the integration over

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 179

the three-dimensional momenta of the final particles. Taking into consideration the delta
function property
1
δ(x2 − a2 ) = [δ(x − a) + δ(x + a)], (10.35)
2a
and accounting that the muon is on the mass shell and its energy is positive, we can rewrite
(µ) (µ)
the integral over pf for the cross section part, depending on pf , in the form:

(µ)
d3 pf
Z Z
(4) (µ) (µ) (µ) (µ) (µ) (µ) (µ) (µ)2
Ip(µ) = (µ)
δ (pi +q−pf ) = d3 pf dEf δ (4) (pi +q−pf )θ(Ef )δ(pf −m2µ ),
f
2Ef
(10.36)
where 
1, at x > 0,
θ(x) =
0, at x < 0.
Further calculations of Ip(µ) are trivial and the final result is given by the expression:
f

 

(µ) 2
 
(µ)
 1 (e) (e) q2
Ip(µ) = δ (pi + q) − m2µ = δ 2pi q + q 2 = δ Ei − Ef + =
f 2mµ 2mµ
(e) (e)
! !
(e)
1 (e) (e) 2Ei Ef θ 1 (e) E
= δ Ei − Ef − sin2 = δ Ef − i , (10.37)
2mµ mµ 2 2bmµ b
where
(e)
2Ei θ
b=1+ sin2 . (10.38)
mµ 2
So, our cross section acquires the form
  !
(e)
dσ e4 2 θ q2 2 θ (e) Ei
= 4 cos − sin δ Ef − . (10.39)
d3 pf
(e) q (2π)2 b 2 2m2µ 2 b

(e)
Further, the passage to the spherical coordinate system with the vector pi directing along
(e)
the axis z allows us to present d3 pf in the form:

(e) (e) (e) (e) (e)

d3 pf =| pf |2 d | pf |2 dϕd(cos θ(e) ) = (Ef )2 dEf dΩ.

Using the expression obtained, one may carry out integrating the expression (10.39) without
any trouble and obtains the Mott formula [15]:
!2 (e)  
dσ α Ef θ q2 2 θ
= cos2 − sin . (10.40)
dΩ (e)
2Ei sin2 (θ/2) Ei
(e) 2 2m2µ 2

We would do well to discuss the case of the elastic scattering of the electron by a spinless
(scalar) particle h, that is, to consider the reaction

e− + h → e− + h.

The electromagnetic interaction Hamiltonian of the scalar particle described by the wave-
function ϕ(x) has the form
S=0
Hint = ie[ϕ∗ (x)∂µ ϕ(x) − ∂µ ϕ∗ (x)ϕ(x)]Aµ (x) + e2 ϕ∗ (x)ϕ(x)Aµ (x)Aµ (x).

© 2011 by Taylor and Francis Group, LLC

 180 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the momentum representation the vertex corresponding to the interaction between the
photon with the polarization ν and the scalar particle is determined by

e(pν + p′ν ), (10.41)

where pν (p′ν ) is the four-dimensional momentum of the h-particle entering (coming out of)
the vertex. Then in order to define the cross section in the expression (10.27) one would
make the replacement:
Mντ → (pi + pf )ν (pi + pf )τ ,
which would lead to the following value of the cross section:
  !2 (e)
dσ α Ef θ
= (e) (e)
cos2 . (10.42)
dΩ 0 2
2E sin (θ/2) E 2
i i

Thus, the factor −ieγµ corresponds to the Feynman diagram vertex describing the elec-
tromagnetic interaction of the point particles with the spin 1/2. For particles having the
composite structure, such as the proton, the form of the vertex has a more complicated view
and this form is not known on the whole. However, one may define the form of the proton
electromagnetic vertex (Fig. 10.6) from conditions that are natural enough for the quan-
tum theory. We bear in mind the following conditions: the relativistic covariance and the
four-dimensional current conservation law or, what is the same, gradient invariance. The

FIGURE 10.6
The electromagnetic proton vertex.

number of elements from which one may construct the Lorentz covariant vector operator of
vertex is not quite large. They are:
four-vectors
p±ν = (pf ± pi )ν ,
two four-dimensional tensors
εµνλσ , g µν ,
and the following Dirac matrix combinations
vectors pseudovectors
γµ γ5 γµ
tensors pseudotensors
σµν γ5 σµν
γµ γν γλ γ5 γµ γν γλ

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 181

where
i
σµν =
[γµ , γν ], γ5 = iγ 0 γ 1 γ 2 γ 3
2
(all the other Dirac matrix products may be reduced to one of the listed above fourteen
combinations by application of the anticommutation relations {γµ , γν } = 2gµν ). From these
quantities one can build twelve independent four-dimensional vectors viν :

1. p+ν 2. p−ν 3. γν
4. σνµ p+µ 5. σνµ p−µ 6. p̂+ p−ν
7. p̂− p+ν 8. p̂+ p+ν 9. p̂− p−ν
10. γ5 ενµλτ γµ p+λ p−τ 11. p̂+ p̂− p+ν 12. p̂+ p̂− p−ν

(multiplying every vector by γ5 , one may also introduce twelve pseudovectors that would
prove to be claimed when the parity violation effects are being taken into account, that is,
when the weak interaction is switched on). We insert the same amount of the analytical
functions Fi of invariant variable q 2 = p2−ν and represent the proton electromagnetic vertex
in the form:
X12
−i|e|Λν = −i|e| Fi (q 2 )viν . (10.43)
i=1

From Fig. 10.5 it is obvious that the vertex operator, entering into the reaction amplitude,
is confined between the wavefunctions of the free proton. Consequently, we are interested
in the quantity
Jν = |e|ψ(pf )Λν ψ(pi ), (10.44)
which represents the four-dimensional vector of the electromagnetic transition current for
the proton. Then, using the free Dirac equation for the spinors ψ(p) and ψ(p)

(p̂ − mp )ψ(p) = 0, ψ(p)(p̂ − mp ) = 0, (10.45)

one may reduce the number of the four-dimensional vectors viν at the cost of redefining
arbitrary functions Fi (q 2 ). Really, to apply Eqs. (10.45) reduces 6,8 to 1,2 and converts
7,9 into zero. Further, in consequence of the relations

ψ(pf )p̂+ p̂− ψ(pi ) = ψ(pf )[−2m2p + 2(pi pf )]ψ(pi ),

11,12 and 4,5 are also transformed to the kind of 1,2. The inclusion of the identity
1 λτ ν
ενµλτ γ5 γµ = (σ γ + σ τ ν γ λ + σ νλ γ τ ) (10.46)
3
and the Dirac equation reduces 10 to 1 and 2. All this allows us to represent the transition
current in the form:

|e|ψ(pf )Λν ψ(pi ) = |e|ψ(pf )[F1 (q 2 )p+ν + F2 (q 2 )p−ν + F3 (q 2 )γν ]ψ(pi ). (10.47)

Since the current Jν is conserved, then multiplying the expression (10.47) by p−ν gives

ψ(pf )F2 (q 2 )q 2 ψ(pi ) = 0.

This relation, in its turn, means that F2 (q 2 ) is equal to zero for all real values of q 2 6= 0.
Then, from the analyticity condition of the formfactors it follows

F2 (q 2 ) = 0

© 2011 by Taylor and Francis Group, LLC

 182 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

for any q 2 .
Carrying out the Hermitian conjugation of the space components of the expression (10.47),
we obtain

{ψ(pf )[F1 (q 2 )p+ + F3 (q 2 )γ]ψ(pi )}† = {ψ † (pf )γ0 [F1 (q 2 )p+ + F3 (q 2 )γ]ψ(pi )}† =

= ψ † (pi )[F1∗ (q 2 )p+ − F3∗ (q 2 )γ]γ0 ψ(pf ) = ψ(pi )[F1∗ (q 2 )p+ + F3∗ (q 2 )γ]ψ(pf ), (10.48)
where we have chosen the γ-matrices representation with anti-Hermitian γ-matrices and
taken into consideration that the matrices γ and γ0 anticommute with each other. When
pi = pf the transition current is nothing more nor less than the electromagnetic current Jν .
Since the three-dimensional electromagnetic current is Hermitian (J† = J), then comparing
the left-hand and right-hand sides of Eq. (10.48), we have drawn the conclusion:

F1 (q 2 ) = F1∗ (q 2 ), F3 (q 2 ) = F3∗ (q 2 ), (10.49)

that is, F1 and F3 are real quantities. Further, the relation

pν ψ(p) = (iσνµ pµ + mγν )ψ(p) (10.50)

appears to be useful. To prove it one should multiply the equation

γµ pµ ψ(p) = mψ(p)

by γν on the left and transform the left-hand side by the following manner
1
γν γµ pµ ψ(p) = ([γν , γµ ] + {γν , γµ })pµ ψ(p) = (−iσνµ + gνµ )pµ ψ(p).
2
It is convenient to represent the quantities F1 (q 2 ) and F3 (q 2 ) as follows:

a(p) F2 (q 2 )
F1 (q 2 ) = − , F3 (q 2 ) = F1 (q 2 ) + a(p) F2 (q 2 ), (10.51)
2mp

where a(p) is the value of the proton anomalous magnetic moment expressed in terms of the
nuclear magneton units. Then, with an allowance made for Eq. (10.50), the expression for
the transition current takes the form:

Jν = |e|ψ(pf )[F1 (q 2 )p+ν + F3 (q 2 )γν ]ψ(pi ) = |e|ψ(pf ){[F1 (q 2 ) + a(p) F2 (q 2 )]γν −

 
a(p) F2 (q 2 ) a(p)
− p+ν }ψ(pi ) = |e|ψ(pf ) F1 (q 2 )γν + i F2 (q 2 )σνµ q µ ψ(pi ) =
2mp 2mp
= |e|ψ(pf )Λν ψ(pi ), (10.52)
where  
a(p)
Λν = F1 (q 2 )γν + i F2 (q 2 )σνµ q µ . (10.53)
2mp
So, the proton electromagnetic vertex −i|e|Λν has been represented in terms of two formfac-
tors F1,2 that hold all information concerning the proton structure. From Eq. (10.53) the
sense of the transition Fi (q 2 ) → Fi (q 2 ) has become clear as well. Now, the electromagnetic
vertex operator consists of two terms where the former describes the Dirac-type interaction
and the latter does the Pauli-type interaction. Under Fi (q 2 ) → 1 the quantity Λν transfers
into the vertex operator of the point particle having the spin 1/2 and the magnetic moment
(1 + a(p) )µN , as demands the correspondence principle.

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 183

When q 2 → 0, then gamma-raying of the proton is performed by long-wavelength photons

that do not “see” the internal proton structure. In this case we simply observe the point
particle. For this reason the nucleon formfactors must be chosen in such a way that the
following conditions are fulfilled:

F1 (0) = 1, F2 (0) = 1 for proton,
.
F1 (0) = 0, F2 (0) = 1 for neutron.

When calculating the differential cross section of the elastic electron-proton scattering,
we shall make use of the expression (10.53) as the proton electromagnetic vertex. Then,
the proton tensor is defined by:
1
P λσ = Sp[Λλ (p̂f + mp )Λσ (p̂i + mp )]. (10.54)
4
Multiplying (10.31) by (10.54), we work out the final result, namely, the Rosenbluth formula
[31]:
  ( )
dσ dσ 2 2 a(p)2 q 2 2 2 q2 h 2 (p) 2
i2
2 θ
= F1 (q ) − F (q ) − F1 (q ) + a F2 (q ) tan . (10.55)
dΩ dΩ 0 4m2p 2 2m2p 2

The factor, standing in the braces of Eq. (10.55), describes the manner in which the
scattering process is changed because of the proton structure. When the proton were the
point particle, like the muon, then at any q 2 one would have

a(p) = 0 and F1 (q 2 ) = 1

and the expression (10.55) would coincide with the Mott formula. As one should expect,
deviations from the Mott formula are larger in the region of small wavelengths of the virtual
photon, that is, in the region of the big values of q 2 . For example, when the electric charge
distribution inside the proton is described by the exponential law, then we obtain at q 2 → ∞
 
dσ dσ
= q −8 .
dΩ dΩ 0
In order to measure the formfactors experimentally it is convenient to redefine them
in such a way that the interference term F1 (q 2 )F2 (q 2 ) will be absent in the cross section
(10.55). With this object in mind we introduce the electric and magnetic nucleon formfactors
[32] (they are called Sachs formfactors as well):

a(N ) q 2 N
GN 2 N
E (q ) = F1 + F , GN 2 N
M (q ) = F1 + a
(N ) N
F2 , (10.56)
4m2p 2

where N = n, p. Then, the formula (10.55) takes the form:

   p 2 
dσ dσ (GE ) − q 2 (GpM )2 /(2mp )2 q2 p 2 2 θ
= − (G ) tan . (10.57)
dΩ dΩ 0 1 − q 2 /(2mp )2 2m2p M 2

From definition of GN 2 N 2
E (q ) and GM (q ) follows

GpE (0) = 1, GpM (0) = 1 + a(p) = 2.79µN ,
(10.58)
GnE (0) = 0, GpM (0) = a(n) = −1.91µN .

This suggests that GN 2 N 2

E (q ) and GM (q ) must be related with distributions of the charge
and magnetic moment of a nucleon.

© 2011 by Taylor and Francis Group, LLC

 184 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The experimental method of determining the formfactors is simple in its description at

least. Let us fix q 2 and plot the quantity
 −1
2 θ dσ dσ
f (tan ) =
2 dΩ dΩ 0

along the ordinate, whereas the values of tan2 (θ/2) are along the abscissa. Then the function
f is represented as the straight line (Rosenbluth straight line) whose slope is

q 2 [GpM (q 2 )]2
− √
( 2mp )2

and whose ordinate in the point

θ 1
tan2 =−
2 2[1 − q 2 /(2mp )2 ]

equals
(GpE )2 [1 − q 2 /(2mp )2 ]−1 .
In other words, the straight line slope gives the value of (GpM )2 while the ordinate does the
value of (GpE )2 . Repeating this procedure under the different values of q 2 one may define the
formfactors as the functions of q 2 . True enough, since we are dealing with the formfactor
squares, then the confused ambiguity, concerning the formfactor signs, is left. However, it
will disappear if we take into account the formfactor values at q 2 = 0 (the sole exception is
provided by GnE (q 2 ) because GnE (0) = 0).
Note some more peculiarities of measuring the proton formfactors. With an allowance
made for the value of q 2 , which is given by Eq. (10.36), we obtain for (dσ/dΩ)0
  (e)
!2
dσ α2 Ef θ
=− 2 (e)
tan−2 .
dΩ 0 q Ei 2

Then, at the electron backward scattering (θ = 1800 ) the differential cross section (10.55)
takes the form:
(e) 2
!  2  (e)
!2
dσ α2 Ef −2 θ q 2 θ p αEf
= 2 tan 2 2
tan [GM (q )] = √ [GpM (q 2 )]2 ,
dΩ q E
(e) 2 2m2p 2 2mp E
(e)
i i

which defines GpM (q 2 ) directly. At great values of q 2 the contribution coming from GpM (q 2 )
is essentially√suppressed so far as the factor [1 − q 2 /(2mp )2 ]−1 stands in front of [(GpE (q 2 )]2
while −q 2 /( 2mp )2 stands in front of [GpM (q 2 )]2 .
The neutron formfactors are found from the data of scattering off electrons by the
deuteron. For example, under the analysis of the reaction

e− + d → n + p + e−

one should subtract the contribution coming from the electron-proton scattering and make
the small correction on the nucleon coupling. The obtained data are in accordance with the
assumption
GnE (q 2 ) ≈ 0, GnM (q 2 ) ≈ a(n) GpE (q 2 ).
In Fig. 10.7 we display the proton formfactor dependence on the square of the transferred
momentum. For point particles the formfactor is a constant. The formfactor dependence

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 185

FIGURE 10.7
The transferred momentum square dependence of the proton formfactors.

on q 2 , which is observed in experiments means the nucleons have a specific structure. The
experimental data (see, for review, [33]) are agreed with the so-called scale relations, the
scaling law
Gp (q 2 ) Gn (q 2 )
GpE (q 2 ) = M = M = G(q 2 ), (10.59)
2.79 −1.91
where in the region q 2 ≤ 0.5 (GeV)2 the unified formfactor G(q 2 ) is well described by the
empiric dipole formula [34]:

−2
G(q 2 ) = 1 − q 2 /m20 , (10.60)
m0 = 0.71 GeV. At q 2 ≥ 0.5 GeV2 deviations of ≤ 20% from the dipole formula are
observed. Notice, all the experimental data concerning the elastic ep-scattering may be
described if one assumes that the scaling law is valid and the unified formfactor has the
form of the sum of two poles
b 1−b
G(q 2 ) = + , (10.61)
1 − q 2 /m21 1 − q 2 /m22
where
b = −0.33, m1 = 1.31 GeV, m2 = 0.64 GeV.
If the proton were the static target, then the Fourier images of the formfactors would be
interpreted as a distribution of the electric charge and the magnetic moment. Unfortunately,
a proton recoil spoils all the state of affairs. However, one could find a Lorentz frame of
reference in which the physical meaning of the formfactors becomes more transparent. The
Breit reference system (the “brick wall” reference system):
pi = −pf .
has such a property. In this reference frame the proton behaves like a ball jumping aside
a brick wall. As in the Breit reference frame the four-dimensional vector qν is remaining
spacelike as before
qν2 = −q2 < 0,

© 2011 by Taylor and Francis Group, LLC

 186 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

this reference frame really exists and, consequently, could be connected with the lab frame
by the Lorentz transformation. Using (10.52) we can find the components of the proton
transition current in the Breit reference frame

pµi = (E, p), pµf = (E, −p).

If one denotes the eigenvalues of the proton helicity operator in the initial and final states
by αi,f and takes into account the fact that at p → −p the helicity changes signs, the
normalization condition:
′ E (α) ′
ψ (α)† (p)ψ (α ) (p) = δαα′ , ψ (p)ψ (α ) (p) = δαα′ ,
mp

leads to the result:

E (−αi )
ψ (−αi )† (pf )ψ (αi ) (pi ) = , ψ (pf )ψ (αi ) (pi ) = 1.
mp

Then the time component of the current is different from zero if and only if αf = −αi and
its connection with the formfactors has the following form:
 
p 2 (p) p 2 E (p) p 2 
J0 ≡ ρ = |e|ψ(pf ) [F1 (q ) + a F2 (q )]γ0 − a F2 (q ) ψ(pi ) = |e| F1p (q 2 )+
mp
  
E2
+a (p)
1 − 2 F2 (q ) = |e|GpE (q 2 ),
p 2
mp
that is,
ρ
GpE (q 2 ) = . (10.62)
|e|
Analogous manipulations with the space component of the current lead to the relation:
J
GpM (q 2 ) = . (10.63)
|e|ψ(pf )γψ(pi )

From (10.62) it is obvious that we achieved the goal to be thought–in the “brick wall”
reference frame the formfactor GpE (q 2 ) could be interpreted as the Fourier image of the
static distribution of the electric charge
Z
GpE (q 2 ) = ρ(r) exp (iq · r)dr, (10.64)

(for a pointlike charge ρ(r) = δ(r) and the formfactor is simply equal to 1). Notice, that
such a interpretation has not been assigned a literal meaning. Since every value q corre-
sponds to its reference frame, then the function ρ(r) does not refer to the definite reference
frame. However, in the nonrelativistic limit q2 ≪ m2p changing of the proton energy during
scattering may be neglected. Then the Breit reference frame coincides with the proton rest
reference frame and the function ρ(r) gains the real meaning of the space charge distribution.
Based on the experimental data by measuring the formfactors, it is easy to determine the
space size and the charge distribution character of the proton. At small values of | q | the
exponent in Eq. (10.64) could be expanded into the series:
Z  
p 2 (q · r)2
GE (q ) ≈ ρ(r) 1 + i(q · r) − + .... dr.
2

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 187

Assuming the charge distribution to be spherically symmetric (ρ(r) = ρ(r)) and directing
the z axis along the vector q, we obtain
Z Z
1 q2 q2
GpE (q 2 ) ≈ 1 − ρ(r)(q · r)2 dr ≈ 1 − ρ(r)r2 (4πr2 dr) ≈ 1 − < r2 >,
2 6 6
whereR < r2 > is an average value of the proton radius square and we have taken into account
that ρ(r)dr = 1. Measurements of the nucleon formfactors lead to the conclusion that
the average radius of the proton and the neutron has the order of 0.8 Fermi. Roughly the
same radius is found from the experimental data for the magnetic moment distribution.
To establish the electromagnetic structure of the proton we should check what forms of
the charge distribution lead to the dipole formula (10.60), which works well in the region
of small values of q 2 . It turns out that the positive result is provided by the exponential
distribution
m3
ρ(r) = ρ(r) = 0 exp (−m0 r), (10.65)
8π
where m0 is the mass of a particle carrying an interaction between nucleons. Really, sub-
stituting (10.65) into (10.64) and choosing the spherical coordinate system with the axis z
along the vector q, we arrive at the result (θ is the azimuthal angle!):
Z Z 2π Z π
p 2 m30 ∞
−2
GE (q ) = dr dϕ dθr2 sin θ exp (i | q | r cos θ − m0 r) = 1 − q 2 /m20 .
8π 0 0 0
(10.66)
So, the distribution of the charge and the magnetic moment for the proton is described by
the sufficiently simple function, the exponent. Since the quantity ρ(r) tends to the constant
under r → 0, then it is obvious that a nucleon does not have any hard core, that is, there
are no the congestion of charges in the center, as it was in the case of an atom. From this,
it does not follow in any way that structural elements are generally absent inside a nucleon.
On the contrary, the distribution nonhomogeneity of the charge and the magnetic moment
testifies to presence of such objects. In order to “see” and investigate the properties of
blocks that constitute a nucleon one needs to increase the resolution of our devices. This
means in the quantum language that we should further decrease the de Broglier wavelength
of the virtual photon, that is, the value of q 2 should be increased.
Without doubt the energy growth of the incoming electrons leads to changing the char-
acter of the collisions with nucleons. So, at definite values of q 2 the proton may be excited
into one of the nucleons resonances (∆(1238), N ∗ (1520), N ∗ (1688), and the like) and emits
the π 0 -meson under returning to the final state (quasi-elastic scattering). On further in-
creasing the values of q 2 , multiple hadrons generation, a proton break-up into numerous
fragments (deep inelastic scattering), will take place. In such a situation, identification of
the final proton state is out of the question. In Fig. 10.8 we represent the Feynman diagram
for this case. Here, P is an initial four-dimensional momentum of the proton and Pf is a
four-dimensional momentum of a final hadron state. It appears, that now we already have
two independent variables, as distinguished from the elastic scattering. To make sure of
this we introduce the quantity
(q · P )
ν= , (10.67)
mp
which simply is an electron energy loss in the proton rest system, that is,
(e) (e)
ν = Ei − Ef .
Let us see what role plays the quantity ν in the electron-proton collisions. The mass of a
final state is defined according to
Mf2 = Pf2 = (q + P )2 = m2p + q 2 + 2mp ν. (10.68)

© 2011 by Taylor and Francis Group, LLC

 188 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 10.8
Deep inelastic collisions of electrons with hadrons.

From Eq. (10.68) we find the values of q 2 for the elastic scattering

q 2 = −2mp ν, (10.69)

for exciting a nucleon resonance with the mass mN

q 2 = −2mp ν − m2p + m2N , (10.70)

and for the deep inelastic scattering (the proton converts into a continuous spectrum through
the resonances region)
q 2 = −2mp ν − m2p + Mf2 , (10.71)
where Mf2 forms a continuous spectrum. We select the scattered electron in the final state
and do not concretize the hadrons system on which the proton is decayed. Such reactions
are called inclusive ones, since they include all that is possible into the hadrons system.
Then the measured cross section is the cross sections sum including different final states
of hadrons. From (10.71) follows, since Mf2 may take any values, then the quantities q 2
and ν are already independent variables in the case of the deep inelastic scattering. There-
fore, the additional kinematic degree of freedom appears under the inelastic electron-proton
scattering. One is clearly expecting, thanks to this circumstance, the inelastic scattering
investigation to give more information than the elastic processes investigation. In the case
of the inelastic scattering the electron-proton interaction dynamics will already be defined
by the formfactors W1,2 (ν, q 2 ) (since they determine the proton structure they are called
the structure functions). Their precise form can be established only within the sequential
theory of the strong interaction, that is, the QCD. However, in order to find W λσ we again
call to the relativistic and gauge invariance for help. Let us go in this way.
There are two vectors Pλ and qλ in our disposal. As we are going to parameterize the
cross section, which has already been summed up and averaged over spins, the matrices γλ
are not included in consideration. The set of independent tensors that are built from the
vectors Pλ and qλ are as follows:

P λP σ, qλ qσ , P λqσ , qλP σ . (10.72)

Based on the fact that the metric tensor g λσ can also participate in our constructions, the
most general expression for the hadron tensor takes the form:

W λσ = a1 g λσ + a2 P λ P σ + a3 q λ q σ + a4 P λ q σ + a5 q λ P σ , (10.73)

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 189

where ai is a function of the scalars q 2 and ν. From the current conservation law follows

qλ W λσ = qσ W λσ = 0. (10.74)

Then, multiplying (10.73) by qλ (qσ ) and equaling the coefficients at q σ and P σ (q λ and
P λ ), we arrive at four equations:

a1 + a3 q 2 + a4 (P q) = 0, (10.75)

a2 (P q) + a5 q 2 = 0, (10.76)
a1 + a3 q 2 + a5 (P q) = 0, (10.77)
2
a2 (P q) + a4 q = 0. (10.78)
Subtracting (10.76) from (10.78), we obtain a4 = a5 . This, in its turn, means that Eqs.
(10.75) and (10.77) coincide, that is, there are only two equations to define four quantities.
We choose a1 and a2 as the independent quantities. From Eqs. (10.77) and (10.78) we find
 2
a2 (P q) a1 (P q)
a4 = − , a3 = − 2 + a2 . (10.79)
q2 q q2

Using the relations (10.79) and introducing the designations

W2 (ν, q 2 )
a1 = −W1 (ν, q 2 ), a2 = ,
m2p

we deduce for the hadron tensor W λσ the final expression:

    
λσ λσ qλqσ 2 1 λ (P q)q λ σ (P q)q σ
W = −g + 2 W1 (ν, q ) + 2 P − P − W2 (ν, q 2 ),
q mN q2 q2
(10.80)
where W1,2 (ν, q 2 ) are functions defining the nucleon structure. Multiplying (10.80) with
the electron tensor Lλσ (Eg. (10.31) ) and neglecting the electron mass we obtain for the
doubly differential cross section of the inclusive ep-scattering in the lab frame the following
expression:  
d2 σ 2 2 θ 2
(e)
= σ0 2W1 (ν, q ) tan + W2 (ν, q ) , (10.81)
dΩdE 2
f

where !2
α θ
σ0 = (e)
cos2 . (10.82)
2Ei 2
sin (θ/2) 2

From (10.81) it is evident that the functions W1,2 have the dimensionality of length. The
comparison of (10.82) with (10.42) makes obvious the fact that σ0 is nothing else but the
cross section of the elastic scattering off electrons by the point spinless particle having an
infinitely large mass.
The structure functions W1,2 (ν, q 2 ) represent the inelastic analog of the formfactors of
the elastic scattering. The technique of their experimental determination is also simple.
When one fixes the variables q 2 , ν and plots the values of
 θ d2 σ
f tan2 = σ −1
(e) 0
2 dΩdEf

© 2011 by Taylor and Francis Group, LLC

 190 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

along the ordinate whereas the values of tan2 (θ/2) along the abscissa, then f (tan2 (θ/2)) is
displayed as a straight line. Its slope equals 2W1 (ν, q 2 ) and its ordinate equals W2 (ν, q 2 ) at
the point tan2 (θ/2) = 0.
The functions W1,2 (ν, q 2 ) could be determined under investigating the process for different
scattering angles as well. Obviously, at small angles W2 (ν, q 2 ) dominates but when the
angles are close to 1800, W1 (ν, q 2 ) will prevail. In Fig. 10.9 we represent W2 (ν, q 2 ) as a
function of ν for different values of q 2 at θ = 60 . At small values of ν the peaks of W2 (ν, q 2 )

FIGURE 10.9
The W2 (ν, q 2 )-dependence of ν for different values of q 2 .

are associated with the elastic formfactor and the resonances excitation. At ν ≥ 3 (GeV)2
the final states hit on the continuous spectrum region and the curve is flattened. When
ν ≥ 4 (GeV)2 all the experimental points of Fig. 10.9 lay on the same curve for any values
of q 2 . Let us try to understand all the ensuing consequences.
We define the nondimensional variable
q2
x=− , 0 ≤ x ≤ 1, (10.83)
2mp ν
which is connected with the invariant mass of a hadron system Mf by the relation:
!
1
Mf2 = m2p − q 2 −1 .
x

It is clear that the case x = 1 corresponds to the elastic eN -scattering. One logically passes
to the dimensionless structure function as well
νW2 (ν, q 2 ) = G2 (x, q 2 /m2p ). (10.84)

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 191

At ν ≥ 4 GeV2 the quantity G2 (x, q 2 /m2p ) becomes the function of the variable x only. In
other words, when one fixes x, then the plot of G2 (x, q 2 /m2p ) versus x is represented by
the straight line being parallel to the ordinate. Experiments show that the other structure
function has the analogous behavior

mp W1 (ν, q 2 ) = G1 (x, q 2 /m2p ), (10.85)

where we have again passed to the dimensionless variable. Thus, in the deep inelastic region

| q 2 |≫ m2N , ν ≫ mN , (10.86)

the structure functions G1,2 (x, q 2 /m2p ) become independent on any scale or are scale invari-
ant.
Recall, that the scale invariance, the scaling, is the invariance of physical theory with
respect to space-time transformations

x → ρx, t → ρt, (10.87)

where ρ > 0 is a numerical parameter of transformation. In quantum theory the transfor-

mations (10.87) are supplemented by ones

p → ρp, E → ρE. (10.88)

Physical quantities are changed in accordance with their dimensionalities under the scale
transformations. So, the vector potential of the electromagnetic field and the current are
transformed by the law
A → ρ−1 A, j → ρ−3 j.
It is evident that dimensionless quantities are scale invariants. Particle masses also fall into
this category. When masses or other dimensional quantities, which are not changed under
the scale transformations, do not enter into motion equations or boundary conditions, then
the corresponding theories are scale invariant. Free Lagrangians of the photon and gluon
fields possess the scale invariance. Clearly, in the real world where the gravitational, weak,
electromagnetic, and strong interactions are switched on, the scaling does not take place.
In the first place the absence of the scale invariance is caused by the fact that for physical
particles the relation
E 2 = m2 + p2
must be fulfilled. It is obvious that this relation is not invariant with respect to the transfor-
mations (10.88). On the other hand, there are no reasons that would obstruct exhibiting the
scaling in nature. As we saw, the scale invariance of the dimensionless structure functions,
taking place in the deep inelastic ep-scattering, is one of such examples.
Even before the first experiments for studying the inclusive electron-proton scattering,
J. Bjorken [35] predicted the scaling of the functions G1,2 (x, q 2 /m2p ) in the deep inelastic
region, that is )
G1 (x, q 2 /m2p ) → F1 (x),
(10.89)
G2 (x, q 2 /m2p ) → F2 (x),
when
| q 2 |→ ∞, ν→∞
and x is fixed. For this reason, the phenomena of the scale invariance of the structure
functions is named by Bjorken scaling. Later experiments, fulfilled on electrons, muons,
and neutrinos beams, displayed that the Bjorken scaling is not exact (true, violation is

© 2011 by Taylor and Francis Group, LLC

 192 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

insignificant and may be considered as a correction to the basic effect). We shall not go
into causes of scaling violation since that is beyond the framework of the book (successive
explanation could be obtained within the QCD only). At a given stage the most important
thing for us is to understand the conclusions that follow from the scale invariance.
The behavior of the functions G1,2 (x, q 2 /m2p ) is greatly distinguished from the correspond-
ing behavior of the elastic formfactors GE,M (q 2 ). Whereas GE,M (q 2 ) sharply fall down with
the increase of | q 2 |, at | q 2 |→ ∞ the functions GE,M (q 2 ) do not depend on q 2 at all. This
looks like the proton would not have the electromagnetic formfactors in the deep inelastic
scattering region. In other words, in this case the proton behaves as a point particle. How-
ever, the proton is not the point particle and its size, defined from the elastic ep-scattering,
is far from small (∼ 1 Fermi). The only reasonable explanation of the experiments on the
ep-scattering resides in the fact that the electric charge inside the proton is concentrated
in several points, that is, in particles entering into the composition of the proton. All this
may be formulated in a somewhat different way. The presence of the scaling means that
such dimensional parameters as the proton mass mp and the corresponding length ∼ 1/mp
do not play any significant dynamical role in the deep inelastic scattering processes, that is,
there is no distances scale in this case for the proton. We shall explain the aforesaid in the
example of an atom. In an atom, aside from its own size, there is one more distances scale,
the size of the atom nuclear. Thanks to the uncertainty relation, the scale of distances is
inversely proportional to that of energies. The existence of the two scales in an atom leads
to the fact that processes, taking place with it at low and high energies, are distinguished
from each other in a radical way. When the distances are bigger than the nuclear size
RN , the system is approximately described by the Coulomb potential. However, when the
distances have the order of RN or are smaller than RN , the Yukawa potential works. In
that case when there is only one scale, the qualitative change of the processes character is
not in progress under the energy increase. To put it otherwise, for the proton in particular,
and for hadrons in general, the second scale is equal to infinity in the case of the inclusive
scattering. This means, if a hadron really represents the composite particle, then particles,
entering the hadron, should have the negligibly small size, that is, they should be pointlike
or, what is the same, structureless. At collisions with these hadron blocks the electrons can
be often scattered through large angles just as α-particles were scattered in the Rutherford
experiments when they were finding their way into an atom’s nuclei.

10.3 Parton model

To prove the scale invariance of the structure functions G1,2 (x, q 2 /m2p ) Bjorken used the
hadron current algebra. One naturally expected the appearance of the model that could
explain the scaling from the point of view of the hadron structure. With this in mind, in
1969 R. Feynman [36] suggested the parton model, the brief content of which we shall give
below (the reader could find the detailed description of the model in the book [37]).
An extended nucleon (and any hadron as well) does not represent a formation that is
smeared by the continuous way in the space, but it consists of weakly confined point particles
(partons). The scattering of the high energy electron occurs on partons incoherently, that
is, the electron is elastically scattered on one of partons, not affecting others. Every nucleon
participates in the inelastic reaction by only one parton (active parton), which transfers the
fraction xi (i is an index of parton kind) of the hadron four-dimensional momentum Pµ .

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 193

Thus, for the ith parton we have

pi = xi P, Ei = xi E, (10.90)

When one denotes the partons number density, the parton distribution function, by fi (xi ),
then fi (xi )dxi defines the number of partons with the four-dimensional momentum xi Pµ
in the range from xi to xi + dxi . However, under fulfillment of (10.90) the parton mass
proves to be a variable quantity that seems strange at least. The situation is clearing up
in the reference frame where the time component of the vector qµ is equal to zero (such
a system really exists since q 2 < 0). To find it we consider the reference frame K, which
moves as related to the lab frame with the velocity v being parallel to the vector q. Using
the Lorentz transformation, we obtain
q0′ − v | q′ |
q0 = √ , (10.91)
1 − v2
where we have supplied the quantities in the lab frame by the prime. When one chooses
the velocity v to be equal to
q′
v = 0′ , (10.92)
|q |
then from Eq. (10.91) follows that we have achieved our goal—q0 = 0. It is clear that in
the reference frame K the nucleon momentum is given by:
mN v
| P |= √ . (10.93)
1 − v2
Using (10.92) and taking into consideration q0′ = ν, we arrive at
mN ν
| P |= p (10.94)
Q2

and
m2N ν 2 mN νx
P2 = = , (10.95)
Q2 2
where we have passed to the positive quantity Q2 = −q 2 for the reasons of convenience. In
the deep inelastic region the relations (10.86) take place, therefore

| P |≫ mN P → ∞. (10.96)

Thus, in the Lorentz system where q0 = 0 and the relation (10.96) is fulfilled (the system
of infinity momentum (SIM) ), one may escape the question about a variable mass of a
parton, if one assumes both mi and mN being equal to zero. The notion of partons makes
sense only in the reference frame where a nucleon moves with the relativistic velocity. This
circumstance is the reflection of the already known fact that only the high energy virtual
photon (| q 2 |≥ 1—2 GeV2 ) may discern a parton. In the SIM the transverse component
(as related to the direction of the nucleon motion) of the parton momentum appears to be
negligibly small. Really, in the rest system of a nucleon the mean square longitudinal and
transverse parton momenta are equal to each other. It is clear, that in the p SIM, whose
velocity as related to the lab frame is close to the light velocity (v = ν/ ν 2 − q 2 ≈ 1)
the longitudinal parton momentum is much bigger than the transverse parton momentum.
The transition to the SIM has one more advantage, namely, it sheds a light on a parton
behavior in a nucleon. In this system the interaction acts frequency of partons with each
other decreases, by virtue of the relativistic delay of the time. Thus, in the short time

© 2011 by Taylor and Francis Group, LLC

 194 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

interval between interactions of the virtual photon with the parton, the parton behaves
as a nearly free particle. Scattered partons (active partons) and residues of the initial
hadrons, which did not take part in interactions (the set of passive partons), turn into final
hadrons thanks to the strong interaction. The final hadrons produce two jets, one in the
direction of a scattered parton, and the other in the direction of an initial nucleon. A
hadron jet represents the set of hadrons having small (the order of 300 GeV) transverse
momenta relative to the motion of the parent particle. The jet’s existence already on its
own serves as evidence of the weakness of hadron matter interaction on the small distances.
Indeed, if the hadron matter produced what amounts to dense high excited cluster, then the
isotropic configuration with a large value of an average transverse momentum (the order of
an collision energy) would be natural for outgoing secondary particles.
Now we show how the parton model explains the scaling phenomenon. Incoherence (prob-
abilities are summarized rather than scattering amplitudes) of the lepton-parton scattering
means that every parton with the corresponding distribution function fi (xi ) gives indepen-
dent contribution into the lepton-parton scattering cross section, that is,
X
dσ = dσi ∼ Lλσ W λσ .
i

The hadron tensor W λσ could be represented in the following form:

XZ 1
fi (xi )
W λσ = dxi W λσ , (10.97)
i 0 2mN pi0 (i)

λσ
where W(i) is a parton tensor, and the factor (2mN pi0 )−1 was introduced for reasons of
convenience. Assume that partons have the spin equal to 1/2. As we are not interested
in a parton polarization we should carry out an average over the initial spin parton states
and a summation over the final ones. Thus, to find the parton tensor one may use the
corresponding expression for the pointlike fermion (see, Eq. (10.31) ) and presents (10.97)
in the form:
Z Z
λσ 1 X 2 1 1
W = Qi dxi fi (xi ) d4 p′i θ(p′i0 )δ(p′2 2 (4)
i − mi )δ (pi + q − p′i ) ×
mN i 0 4

×Sp[γ λ (p̂′i + mi )γ σ (p̂i + mi )], (10.98)

where pi and p′i are theith-parton momentum in the initial and final states, respectively,
and Qi is an electric charge of ith-parton. In the expression (10.98) the first delta function
δ(p′2 2
i − mi ) ensures locating a parton on the mass shell while the second one δ
(4)
(pi + q −
p′i ) guarantees fulfillment of the four-dimensional momentum conservation law under the
interaction between partons and virtual photons. Having taken the track and having carried
out the integration over four-dimensional momenta of final partons, we obtain the following
expression for the hadron tensor:
Z 1
λσ
X 1
W = Q2i dxi fi (xi ) δ(xi − x)[2pλi pσi + q λ pσi + pλi q σ − g λσ (pi q)], (10.99)
i 0 2m2N ν

where we have neglected a parton mass and taken into account that
 2 
2 1 q xi P q 1
δ(q + 2pi q) = δ + = δ (xi − x) .
2mN 2mN mN 2mN ν

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 195

The comparison of (10.99) with (10.80) allows us to determine the contribution coming from
the ith-parton to the structural nucleon functions W1,2 :

XZ 1
(i)
W1,2 = dxi fi (xi )W1,2 ,
i 0

where
(i) 1 (i) Q2i x2i
W1 = Q2 xi δ(xi − x), W2 = δ(xi − x). (10.100)
2mN i ν
Thus, due to the momentum conservation law expressed by the function δ(xi −x), the scaling
variable x proves to coincide with a parton fraction of a nucleon momentum. The relations
(10.100) express the extremely instructive fact—a virtual photon can be absorbed by ith-
parton only at the right value of the variable xi = −q 2 /(2mN ν). Substituting (10.100) into
(10.97) and considering the definitions (10.84) and (10.85), we get the Bjorken scaling for
the structural nucleon functions:

1 X 2
mN W1 = C1 (x, q 2 /m2p ) = x Qi fi (x), 


2 i
X (10.101)
νW2 = C2 (x, q 2 /m2p ) = x2 Q2i fi (x). 


i

From (10.101) it follows Callan-Cross formula [38]:

C2 (x, q 2 /m2p ) = 2xC1 (x, q 2 /m2p ). (10.102)

The relations (10.101) establish a link between the structural nucleon functions and the
distribution functions of partons interacting with a photon.
So, the parton model proves to be extremely fruitful. First and foremost, with its help
one managed to prove the scaling behavior of the structure functions G1,2 (x, q 2 /m2p ). It
gets worse and worse as it goes on. This model gives an opportunity to establish the spin
of partons participating in the electromagnetic interaction. Really, the scaling relations
(10.101) takes place only at the parton spin being equal to 1/2. For example, if the parton
has zero spin, the replacement
1
Sp[γ λ (p̂′i + mi )γ σ (p̂i + mi )] → (p′i + pi )λ (p′i + pi )σ
4
would be done in the expression for the parton tensor (10.98). Then, as is easy to see, the
function F2 remains unchanged while F1 = 0.
Evidently, it would be rather attractive to identify the partons with the quarks. From the
quantum laws, the quarks in hadrons can exist both in real and in virtual states. We agree
to call the real quarks by the valent quarks.∗ So, three valent quarks enter baryons whereas
mesons consist of valent quark-antiquark pairs. Just the valent quarks define additive
quantum numbers of hadrons (electric charge, strange, baryon charge, and so on). Thanks to
the uncertainty relation, quark-antiquark pairs could be supplemented to the valent quarks
in a short time. One naturally calls the quarks forming the sea of virtual quark-antiquark
pairs by the sea quarks. For the belief to the quark-parton model to be strengthened, the
electric charges of the quarks should be defined. Measuring the nonelastic formfactors and
using the relations (10.101) one may determine the quark charges. Generally speaking, this
is not so simple because a nucleon consists of a large number of partons interacting with

∗ Such quarks are also named by the block or constituent quarks.

© 2011 by Taylor and Francis Group, LLC

 196 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

each other. As a consequence of detailed measurements and comparing the experiments

on various hadron targets they managed to select the contributions coming from different
kinds of partons and define the electric charges of partons. It appeared that they coincide
with the electric charges of the quarks, namely, are equal to 2/3 and −1/3.
To investigate the electron-nucleon scattering gives the opportunity to define the nucleon
momentum fractions, which are transferred by quarks and antiquarks, when the nucleon
moves with the big velocity. The quark and antiquark contributions to the nucleon momen-
tum could be expressed in terms of the quark and antiquark
XZ
x[fi (x) + f i (x)]dx = 1 − ǫ, (10.103)
i

where we have taken into account the contributions coming both from valent quarks and
from the sea ones. If one assumes that the quarks and the antiquarks are the sole pretenders
on the partons role, then ǫ must take the value 0. By the end of the 1960s of the twentieth
century, the experiments on probing the nucleons by means of the electrons were fulfilled
at SLAC (Stanford Linear Accelerator Center) and some years later (in 1973) the similar
raying of nucleons, with the help of the neutrino beams, was carried out at CERN (Conseil
Europeen pour la Recherche Nucleaire). These experiments gave

ǫ ≈ 0.5.

This result means that nearly 50% of the nucleon momentum is transferred by partons that
do not take part in both the electromagnetic and weak interactions. In the QCD only the
gluons, which are the carriers of the strong interaction between the quarks, may pretend to
the role of such particles.
As discussed earlier, the distribution functions of the valent and sea quarks in a nucleon
may be measured in the processes of the deep inelastic ep- or µp-scattering. However, the
best processes for the experimental definition of quark distributions are the processes of
the inclusive scattering of neutrino or antineutrino from protons. This is caused by the
fact that the neutrino interacts only with the d- and u-quarks while the antineutrino does
with the d- and u-quarks. The quark distributions are described by sufficiently simple
functions that once again confirms the truth every physicist must know:“The Creator is not
ill-intentioned.” For example, the distribution functions of the valent quarks in the proton
have the form:
√
xfuV (x) ≈ 2.04 x(1 − x)2.5 , fdV (x) ≈ 0.57(1 − x).

At present we discuss the question about the quarks masses. The particle mass could
be exactly determined by its energy only for the free particle. Since the free quarks were
not discovered up till now, then one assigns the precise meaning to their masses with
difficulty. The quark mass problem is very similar to the problem concerning the electron
mass in solid state physics. The electron, when it is moving in a solid, behaves as a
particle with the “effective” mass mef , which is significantly distinguished from its true
mass m. Moreover, mef may depend on the motion features, because in the reality the
masses difference ∆m = mef − m is caused by the interaction of the electron with objects
surrounding it. In this sense the masses of all the quarks are “effective,” because they
are defined in the processes in which the quarks are interacting with other particles. For
this reason the quark mass values, found under analyzing the energy levels location of the
bound states (quarkoniums.∗), may appreciably differ from those values that correspond to

∗ Bound state to consist of heavy quarks and a corresponding antiquark is named the quarkonium.

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 197

the weak decay of the quarks. They distinguish the current and constituent (block) quark
masses. Since, in the quantum field theory interactions are formulated on the language of
currents and potentials, it is natural to call the quarks entering into Lagrangians by the
current quarks. Thus, the current masses concern to naked quarks and they do not take
into account the contributions coming from their gluons and quark-antiquarks fur coats.
In the QCD the current quark mass depends on the momentum transferred to the quark
and is decreased with the growth of this momentum. So, at the scale ∼ 2 GeV the current
quark masses are confined into the intervals

1.5 MeV < mu < 3.3 MeV, 3.5 MeV < md < 6.0 MeV, ms = 104+26
−34 MeV.
(10.104)
Thanks to the fur coats contribution, the block masses exceed the corresponding current
masses on 300 GeV approximately. From (10.104) it becomes clear that the success of the
SU (2)-symmetry is basically caused by a closeness of the masses of the u- and d-quarks.
The symmetry with respect to the SU (3)-group, which includes the more heavier s-quark, is
already violated much stronger. However, by the irony of fate, just the approximate SU (3)-
symmetry found the exact symmetry status in the same quark walk of life, provided the
condition that it is being used to describe interactions between quarks in themselves.

10.4 Quantum chromodynamics. The first acquaintance

So, late in the 1960s, the hypothetical quarks have been acquiring the status of objects
the reality of which, while indirectly, manifests in experiments. However, the quark model
has a lot of unresolved problems, as before. One of these problems is connected with the
quarks statistics. As an example, we consider the Ω− -hyperon entering into the 3/2+ baryon
decuplet. Its total wavefunction is the product of three wavefunctions which express the
dependence on the space-time variables, the ordinary and unitary spins, that is,

ΨΩ = Ψ(r)Φ(S)Θ(S(un) ), (10.105)

where the quark filling of the Ω− -hyperon may be schematically represented in the following
way
Θ(S(un) ) =| s ↑ s ↑ s ↑>, (10.106)
(the arrow on the quark symbol defines the direction of the spin projection). It is clear that
Θ(S(un) ) is completely symmetric with respect to any transposition of the s-quarks. In the
case of parallelism of three spins Φ(S) is also symmetric. Since the symmetry character of the
space part of the wavefunction is determined by the factor (−1)l (for the 3/2+ -decuplet, l =
0), then Ψ(r) proves to be symmetric under the three quarks transposition as well. Thus, the
total wavefunction of three s-quarks system appears symmetric. However, the quarks have
the half-integer spin, consequently, they obey to the Fermi-Dirac statistics and the Pauli
principle is valid for them. As a result, the total wavefunction ΨΩ must be antisymmetric.
To prove the Pauli theorem about the connection of the spin with the statistics is based
on such fundamental concepts of the quantum theory as microcausality and locality. The
refusal from this theorem would be tantamount to the volley of Aurora cruiser that has
announced the beginning of the October revolution in Russia in 1917. But the evolutions
way always was more preferable than the mysterious ways of revolutions. The painless
solution of the conflict with the statistics proves to be possible by introducing the new
discrete variable [39], called a color [40], which is assigned to all the quarks independently

© 2011 by Taylor and Francis Group, LLC

 198 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

of the flavor. By the example of the Ω− -hyperon, it is clear that the minimum number
of the new variable values should be equal to 3. We are restricted by three values for the
color degree of freedom and shall designate them as R (red), G (green), and B (blue). To
introduce the color allows putting three quarks into one and the same quantum mechanical
state inside a hadron. For the Pauli principle to be fulfilled, the baryon wavefunctions
must be antisymmetrized on the color variables. So, the Ω− -hyperon wavefunction part,
connected with the unitary spin, has the form:
3
1 X
Θ(S(un) ) = √ εijk si sj sk , (10.107)
6 i,j,k=1
√
where the coefficient 1/ 6 is related with the normalization, and not entirely convenient
indices R, G, B are simply replaced by 1,2,3.
When we are talking about the kinematic aspects of the quark systems, then three internal
degrees of freedom, the colors, are conveniently considered as the eigenvalues of the color
spin operator (Fig. 10.10). It should be stressed that by definition the color spin operator
has nonzero values for quarks and gluons only. If the hypothesis of the quarks confinement

FIGURE 10.10
The color spin operators.

is valid then the color spin belongs to the number of the hidden degrees of freedom. This,
in its turn, means that all the particles, which are being observed in the free state, must
have the color spin equal to zero. As it follows from Fig. 10.10 there are two ways to obtain
a colorless (or white) hadron. The first way is to mix three different colors (or anticolors)
while in the second way every color is mixed with the corresponding anticolor. Then, for
any baryons, whose wavefunction with consideration both for flavor and for color will be
written in the form
3
1 X
Bαβγ = √ εijk qαi qβj qγk , (10.108)
6 i,j,k=1

the color spins are mutually compensated and the total baryon color equals zero. Analo-
gously, mesons prove to be colorless because they consist of a mixture of color and corre-
sponding anticolor in equal proportion
3
1 X αi
Mβα = √ q qβi . (10.109)
3 i=1

Let us express the hadron colorless on the strict mathematical language. Assume that a
quark changes its color, to say, it goes from one color state to another. Moreover, we shall

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 199

consider that this new state is the linear combination of all the possible old states (for the
sake of simplicity, we omit the flavor indices of the quarks)
3
X
qi′ = Uij qj . (10.110)
i=1

Now we require hadrons to have one and the same view both in the old and new variables,
that is, hadrons should be the invariants of the transformation (10.110)
3
X 3
X
qi′ q ′i = qi q i , (10.111)
i=1 i=1

3
X 3
X
εijk qi′ qj′ qk′ = εijk qi qj qk . (10.112)
i,j,k=1 i,j,k=1

As the antiquarks q are transformed by the complex conjugated matrices U ∗ , then from
(10.111) follows
X3
∗
Uij Uim = δjm ,
i=1

that is, the transformation (10.111) is performed by the unitary matrices. In its turn Eq.
(10.112) leads to the condition:
det U = 1 (10.113)
which means the transformation (10.111) is special. So, the transformations in question
constitute the special unitary group in three-dimensional color space, the SU (3)-group,
that is, the explicit form of (10.110) is as follows:

qj′ = [exp (iαa Ta )]jk qk , (10.114)

where a = 1, 2, ...8, αa are real parameters and Ta = λa /2 are group generators. Further,
in order to distinguish this group from the group on flavors we shall be talking about
it as the color group SU (3)c . The observed hadrons are invariant with respect to the
transformations of the SU (3)c -group or, what is the same, represent the color singlets. The
color spin introduction liquidates not only the conflict with the statistics, it also forbids
the existence of bound qq-states (of the type RB, GR, and so on) on the strength of the
postulate: only the colorless hadrons are observable. Thus, the color scheme explains the
exceptional role of the quark combinations qqq, qqq, and qq in nature.
Now we address the dynamical aspects of the color hypothesis. It is evident that when we
are considering the quarks dynamics the quark interactions should be taken into account.
In this case, the role of the color degree of freedom is most of all similar to the role of the
electric or gravitational charges and, let them forgive us for this terminological liberty, we
shall be speaking of the color charge rather than the color spin.
The quantum field theory describing interactions between quarks, the QCD, is built by
the analogy with the quantum electrodynamics (QED). Recall that the electromagnetic field
appears as the compensated field that ensures the charged fields invariance with respect to
the local one-parametric Abelian group U (1)em . In 1954 C. N. Yang and R. L. Mills [41]
investigated the local generalization of the non-Abelian groups by an example of three-
parametric group SU (2). As a result they came to recognize that in this case the local
gauge invariance of the theory already demands introducing a three-parametric compen-
sated field which is interpreted as a field of strongly interacting vector bosons. An obvious

© 2011 by Taylor and Francis Group, LLC

 200 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

generalization of this fact lies in a statement: in the case of n-parametric local gauge group
the theory invariance demands introducing n-parametric compensated field.∗ However the
Yang-Mills theory faced two troubles there and then. First, these vector bosons, similar
to the photon, would have zero mass and lead to long-range forces. But in this case, such
particles should be already observed. Second, due to the big value of the interaction cou-
pling constant the perturbation theory does not work, which does not allow making any
conclusions from the theory at that time. In this connection the Yang-Mills theory has
attracted purely academic interest first and it was practically impossible to make out the
prototype of the future theory of the strong interaction in its outlines. There were not such
notions as quark and gluon. They appeared a decade later and, in the beginning, they were
not connected with a mathematical apparatus of non-Abelian theories in any way. Still
ten years were necessary in order for the synthesis of these two ideas to led to the QCD
formulation.
The QCD is based on developing the formalism stated in Section 4.5. But now, in place
of the U (1)em -gauge group, we are dealing with the phase transformations group of the
color quark fields, the SU (3)c -group. We shall consider for simplicity that the quarks have
one flavor only (one flavor approximation). Then the free quarks Lagrangian is given by
the expression:

i
L0 = [qk (x)γµ ∂ µ qk (x) − ∂ µ qk (x)γµ qk (x)] − mqk (x)qk (x), (10.115)
2
where k = R, B, G. Let us investigate the consequences of the invariance of L0 with respect
to the local gauge transformations of the non-Abelian group SU (3)c

qk′ (x) = Ukj (x)qj (x) = {exp[−igs αa (x)Ta ]}kj qj (x), (10.116)

where Ta = λa /2 and λa are the Gell-Mann matrices (9.112). So, our task is to ensure the
local SU (3)c invariance of the Lagrangian L0 . Pass in (10.116) to infinitesimal transforma-
tions
qk′ (x) = [δkj + iαa (x)(Ta )kj ]qj (x). (10.117)
In this case the derivative of the quark field function is transformed by a law

∂µ qk′ (x) = [δkj + iαa (x)(Ta )kj ]∂µ qj (x) + i(Ta )kj qj (x)∂µ αa (x) (10.118)

and violates the invariance of L0 . To rescue the situation we introduce eight gauge fields
Gaµ (x) and build covariant derivatives
µ
Dkj (x) = δkj ∂ µ + igs (Ta )kj Gaµ (x), (10.119)

where gs is a gauge constant of the SU (3)c group. Further, we replace the ordinary deriva-
tives with the covariant ones in L0
i µ µ†
L= [qk (x)γµ Dkj (x)qj (x) − Dkj qk (x)γµ qj (x)] − mqk (x)qk (x) =
2

∗ Gaugefields which are introduced to ensure the local non-Abelian gauge invariance are now called the
Yang-Mills fields and the equations they satisfy in the free case

∂ ν Gaµν + gs εabc Gνb Gcµν = 0,

where εabc are structural constants of the local gauge group under consideration, are given the title Yang-
Mills equations.

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 201
i
= [qk (x)γµ ∂ µ qk (x) − ∂ µ qk (x)γµ qk (x)] − mqk (x)qk (x) − gs [qk (x)γµ (Ta )kj qj (x)]Gµa (x).
2
(10.120)
The invariance of the Lagrangian will be provided under a condition:
µ′ µ
qk ′ (x)Dkj (x)qj′ (x) = qk (x)Dkj (x)qj (x),

what, in its turn, gives

µ′ µ †
Dkj (x) = Uki (x)Dil (x)Ujl (x). (10.121)
By analogy with the QED we demand that the transformation law of the gauge fields Gaµ (x)
has the form:
1
G′aµ (x) = Gaµ (x) − ∂µ αa (x). (10.122)
gs
However, in this case, the last quantity in (10.120) is not the invariant with respect to
the SU (3)c transformations. Really, taking into account the algebra of the λa -matrices we
obtain
[qk (x)γµ (Ta )kj qj (x)]′ = [qk (x)γµ (Ta )kj qj (x)] + iαb (x)qk (x)γµ (Ta Tb −
−Tb Ta )kj qj (x) = [qk (x)γµ (Ta )kj qj (x)] − fabc αb (x)(qk (x)γµ (Tc )kj qj (x)),
From the obtained expression follows that the gauge invariance of the Lagrangian (10.120)
will be restored if we replace the transformation law (10.122) with
1
G′aµ (x) = Gaµ (x) − ∂µ αa (x) − fabc αb (x)Gcµ (x). (10.123)
gs
Now we should supplement the Lagrangian L by that of the free gauge bosons LG . Thanks
to presence of the last term in Eq. (10.123), the field tensor Gaµ (x) has a more complicated
form than its analog in the QED. It is not difficult to show that the gauge invariance will
be ensured by the following choice of LG
1
LG = − Gaµν (x)Gµν
a (x), (10.124)
4
where
µ
Gµν µ ν ν µ ν
a (x) = ∂ Ga (x) − ∂ Ga (x) − gs fabc Gb (x)Gc (x). (10.125)
Thus, using only the requirement of the Lagrangian invariance with respect to the local
gauge group SU (3)c , we get the total Lagrangian of the QCD
i
LQCD = [qk (x)γµ ∂ µ qk (x) − ∂ µ qk (x)γµ qk (x)] − mqk (x)qk (x)−
2
µ 1
−gs [qk (x)γµ Gkj (x)qj (x)] − Gµν (x)Gµν (x) (10.126)
4
where
1 µ
Gµkj (x) = G (x) (λa )kj , Gµν (x) = ∂ µ Gν (x) − ∂ ν Gµ (x) − igs [Gµ (x), Gν (x)],
2 a
Gµkj (x) is a four-dimensional gluon field potential in a point x (every component of Gµkj
represents a 3 × 3 Hermitian matrix in the color space). The number of quark field phases,
we can change by arbitrary ways, is equal to eight. Consequently, to compensate all changing
of the phases we need eight gluons as well. Since the introduction of the gluon mass term
leads to the violation of the local gauge invariance, the carriers of the strong interaction,
the gluons, are massless. From (10.126) follows that the wavefunctions of the gluon field

© 2011 by Taylor and Francis Group, LLC

 202 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

possess two color indices j and k, that is, the two color gluons bear the color and anticolor
charge. It is clear, that the combination deprived with the color charge g0 = RR+BB +GG
is a color singlet. Exchange of this singlet changes by no means the color state of the quark,
and consequently, g0 cannot pretend on the role of particle bearing the color interaction
between the quarks. In order to establish the wavefunction color parts of eight gluons left
we appeal to the original version of the quark theory, the fundamental triplets hypothesis.
Let us set up a correspondence between three quark flavors and three quark colors. Then
the gluons by their own structure are analogous to the mesons. Therefore, to determine
the color parts of the gluon wavefunctions we can use the quark filling of the meson octet.
Having fulfilled the replacements

u → R, d → G, s → B,

in Eq. (10.3), we arrive at the following expressions for the color parts of the gluon wave-
functions: 
g1 = GB, g2 = RB, g3 = RG, g4 =
BG, 

g5 = BR, g6 = GR, g7 = √12 RR − GG , (10.127)


g8 = √16 RR + GG − 2BB . 

It is obvious from the form of LQCD that in the QCD the kinetic energy of the gluons is
not already purely kinetic, since it contains interactions between the gluons (∼ gs Gaµ Gbν Gcλ
and ∼ gs2 Gaµ Gbν Gcλ Gdσ ). Thus, in the QCD the Feynman diagrams include vertices in which
only the gluons are met. In other words, the gluons possess nonlinear self-interaction (self-
interaction is the typical feature of non-Abelian gauge theories) and this circumstance is
caused by the fact that they have the color charge. The inclusion of the gluons contribution
into the vacuum polarization explains the quarks behavior specificity at large transfers of
the momentum. With penetrating in the gluon fur coat, which surrounds the quark, the
quark color charge is being decreased. This means that in the limit of infinitely small
distances separating the quarks, the color interaction between them is switched off and
they behave very similar to free particles (asymptotic freedom). In the case of the deep
inelastic scattering off electrons by protons the quarks, which exhibit themselves as partons,
are in the protons just in the same condition.
To build the successive quantum theory of quark-gluon interactions we need to use the
so-called interaction representation in which quarks and gluons are described by the free
equations of motion. It is apparent that in the QCD such an operation is not absolutely
lawful because quarks and gluons are not observed in the free states. Quarks in hadrons
might be considered as free particles at small distances and only in this case to use the
QCD, based on the perturbation theory methods, (perturbative QCD) is lawful. With the
growth of the distances between quarks their effective coupling constant increases and, as
a result, the perturbative QCD does not cease to work. By now the quarks confinement
has not received the final understanding within the QCD, that is, it has remained the only
hypothesis confirmed by experiments.
We will briefly discuss some models that explain the confinement of quarks and gluons
inside hadrons. So, one may account for the confinement because the hadrons in the color
states are much heavier than those in the colorless states and, for this reason, the latter
are not observed in up-to-date experiments. The analogy with atoms helps to understand
this idea. Let the neutral (nonionized) atoms correspond to the white hadrons while the
charged ions correspond to color states of the hadrons. It is clear that the ions have a larger
energy and are going to come back in the neutral atom state. Such a tendency is explained
by the fact that the electromagnetic interaction, which could be approximately described
by the Coulomb potential Vc ∼ αem /r in the atom case, acts between the electric charges.

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 203

One naturally assumes that there also exists color interaction to hinder the quark flying out
of the hadron. This idea has found embodiment in the descriptive bag model. The simplest
variant, the MIT bag model [42] (Massachusetts Institute of Technology), is based on the
assumption: a hadron represents the bag with the sharp borders that hinders all color objects
fly out. The hadron system is described by the Lagrangian function
Z
L = dr[LQCD − f (r)], (10.128)

where f (r) defines the walls pressure and, consequently, makes a provision for the confine-
ment of both quarks and gluons. If the quarks become widely separated, then the gluon
fields, propagating between the quarks, are stretched into straight lines and the bag takes
the form of the tube. In the case of interaction of quarks with antiquarks the picture looks
the most simple. When one forgets about the nonlinearity for the time being, then the
system qq would be similar to the electric dipole whose distribution of field lines in the
space is displayed in Fig. 10.11(a). The presence of an interaction between gluons results

e+ e− q q

(a) (b)

FIGURE 10.11
The field lines distribution in the qq-system in (a) and (b).

in compressing these lines into the color field tube, the gluon guide (Fig. 10.11(b)). The
interaction structure is such that the tube cross section area S is kept constant under the
growth of the distance between the members of the pair qq (r). But the field line numbers
depend on the source charges only. Consequently, the field strength in the gluon guide is
being kept constant and the field energy increases proportionally to the tube volume Sr.
Since the cross section area S is fixed the energy of the system qq is linearly increasing
with the growth of r. It means that any process, in which a finite amount of the energy is
transferred to this system, is not able to separate the quark from the antiquark. In reality
the flux tube cannot be stretched infinitely since the production of the quark-antiquark
pair from the vacuum has become energetically more profitable, which corresponds to the
transition of the one-hadron state into the two-hadron state. In the first approximation the
effective potential of interaction between q and q has the form:
4 αs
V c (r) = − + Ar + V0 , (10.129)
3 r
where A and V0 are positive constant quantities, that is, it looks like a whirlpool (Fig.
10.12). The first Coulomb-like term ∼ 1/r accords with the one-gluon exchange (the analog
of the one-photon exchange in the electrodynamics), and the linear term ∼ r providing

© 2011 by Taylor and Francis Group, LLC

 204 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 10.12
The effective potential of interaction between q and q.

the quark confinement inside the hadron is explained by the contribution of multigluons
exchanges. In the next approximation the relativistic corrections, the spin-spin and spin-
orbital interactions, are taken into account in the expression for the potential V c (r). As
we see, with growing the distance to the source V c (r) is increased. Just the same reasons
can be applied to any composite color singlet. If one tries to disjoin some color part of this
system (say, a qq-pair in a baryon) from others, then its energy will linearly be increased
resulting in the confinement of color components.
One more sufficiently perspective attempt of the confinement explanation is the so-called
Wilson lattice theory [43]. In lattice theories one is assumed that the space and time do not
make up a continuum but represent the points discrete set that resemble the crystal lattice
(most commonly, the cubic one). The quarks are placed in the lattice sites and the field
lines of the gluons field connect the lattice arbitrary site with its nearest neighbors only.
Further, it is assumed that the energy of the interaction between quarks or between quarks
and antiquarks is proportional to a length of a string to connect them. One may draw an
uncounted set of strings between two points of a such a space-time. In the lattice theory the
quantum mechanics average is fulfilled on these strings. Integrals appearing in the process
may be analytically calculated in the region of so-called strong coupling when the lattice
step is much bigger than the typical scale of the quantum fluctuations of the gluon fields
(∼ 10−13 cm). To pass to the continuous space-time is realized by the way of decreasing
the lattice step. Since in this case the lattice sites are merged with each other, then the
problem of calculating the large multiplicity integrals appears. The problem is usually
resolved with the Monte Carlo method. As this method could be applied to the finite
multiplicity integrals the lattices with the finite number of the sites along every one of four
axes are only considered. In the long run one may show that under certain conditions the
quarks confinement taking place in the strong coupling region is persisting under decreasing
the lattice step as well. Detailed description of the lattice QCD could be found in the book
by Creutz [44].
Since the gluons bear the color they cannot exist in free states as well. If the QCD is
true, then one should expect the existence of hadrons containing the gluons only. The most
simple bound gluon state is two gluons forming the color singlet. Of course one may build
the color singlets from three and more gluons. Hadrons of these kinds are called gluonium or
glueball. Different glueballs may be discriminated by the spin and the mass only. From the
theoretical point of view the gluonium identification seems to be very difficult because it is

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 205

impossible to point to the decays or other properties of the gluonium that would certainly
discriminate the gluonium from the quarkonium having the same quantum numbers. On
the other hand, calculations show that the most intensive gluonium formation must occur
in those reactions and decays in which the gluons rather than the quarks are produced at
small distances. Examples are found in decays of heavy mesons ψ and Υ. However, up to
now the gluonium has not been discovered. The numerical calculations on a computer being
done within the QCD allow us to obtain the definite predictions concerning the masses of
the most light glueballs. In doing so the typical masses scale proves to be of the order of
1.5 GeV.
One such confirmation of the existence of the quarks and the gluons is the process of
detecting the hadron jets. To understand it we recall how elementary particles are observed
in the Wilson chamber. The track left by the particle is not a manifestation of the particle
itself at all. It is the result of the interaction between the particle and matter filling the
chamber. This interaction leads to the production of great numbers of ions along the particle
trajectory. It is evident that the interaction distorts the elementary particle motion. For
high energy particles (only such particles have left the track) such distortions (for example,
the track jitter in a transverse plane) are negligibly small and we could state the trajectory
of the particle itself as being observed directly. Analogously, the multiparticle cluster of
fast hadrons with small transverse momenta, the hadron jet, is the track for a parton that
is outgoing from the deep inelastic scattering region. The jet is not only the visualization
way of the “free” quark or gluon but the form of their existence as well. Here a vacuum,
a structure of which is set by the QCD, plays the role of an environment filling the track
chamber. In this vacuum, just the same as in the QED vacuum, small-scale fluctuations
(SCF) are present thanks to the asymptotic freedom phenomenon. Apart from the SCF,
gluon field long-wavelength fluctuations (LWF) caused by nonlinear character of the QCD
exist. At this time, the notions of “small” or “large” are determined from the viewpoint
of a parton, that is, the LWF are realized on the distances of the quark universe radius
(∼ 1 Fermi). The LWF corresponds to the really strong interquarks interaction. They play
the role of “matter” with which a high-energy parton interacts. The average transverse
momentum < pT > of creating hadrons just conforms to the LWF scale: < pT >≈ 300
MeV ≈ 1 Fermi−1 . Thus the hadron jet created by the quark or the gluon may be identified
with the quark itself or gluon itself in just the same sense as the chain of drops in the Wilson
chamber is considered by the particle trajectory.
In 1975, based upon the results of analyzing the e+ e− -annihilation into hadrons at the
electron-positron storage ring SPEAR, the Stanford research group announced the discovery
of quark jets. It appeared that when the jet energy grows the average angle of the jet
spread decreases, that is, hadrons are increasingly gathered round the direction of flying
away q and q. At the jet energy of the order of 18 GeV the hadrons that constitute
the jet occupy only 2% of the total solid angle. Investigations have also shown that the
correspondence quark↔jet has a universal character. This means the composition of hadrons
in jets (relationships between p, n, π, K, and so on) and hadrons distribution over momenta
do not depend on what concrete reaction the given flavor quark, the jet primogenitor, is
produced.
The first indirect manifestations of gluon jets (DESY, PETRA, 1979) were connected
with researching the decays of the Υ-meson∗ into hadrons. According to the QCD the three
gluons must be produced at annihilation of the bb-pair, that is, the decays in question must
have the three-jets nature. True enough, in the case of the Υ-mesons it is difficult to observe
three gluon jets directly because the gluons energy is too small as yet. For this reason two

∗ This meson represents the bound state of two quarks b and b to be dealt within the next subsection.

© 2011 by Taylor and Francis Group, LLC

 206 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

indirect methods are used. The essence of the former is as follows. Usually, at large energies
in the result of the electron-positron annihilation two quark jets in the opposite directions
are created. If one increases the energy Ecm to an extent that the Υ-meson begins to be
born, then the situation will be changed. In the final state instead of the quark-antiquark
pair three gluons will be created, their momenta being allocated now on the whole space. As
a consequence, at Ecm ≥ mΥ the two-jets structure of the created hadrons, which existed
at Ecm < mΥ , should disappear. Just that effect was registered at DESY.
The second method of the indirect observing of the gluon jets was based on kinematics
of the events. At the electron-positron pair annihilation the Υ-meson is created in the rest.
Consequently, the total momentum of the created gluons must be equal to zero as well.
But three three-dimensional vectors, the sum of which equals zero, must lay in the same
plane. With good precision this should also be fulfilled for the particle momenta of the final
hadron state. Of course this plane is varied from one decay to another. The analysis of the
Υ-decays showed the final hadron momenta do lay in the same plane.
In the same year, a little bit later, under increasing the PETRA energy up to 20 GeV
the bremsstrahlung of the gluon by the quark was observed in the following process:

e+ + e− → q + q + g.

The additional hadrons created by the bremsstrahlung gluon should lead to the azimuthal
asymmetric thickening of one of the quark jets. With the energy increasing and the en-
hancement of the statistics set, one managed to select the gluon from the basic jet and
measure the distributions both over energy and over the angle of flying out of the gluon jet.
Measurements have shown the new jet behaves similarly to the bremsstrahlung photon in
the reaction:
e+ + e− → µ+ + µ− + γ,
that is, just as the particle with the spin 1 should behave.
Once one has managed to “see” the quark, the next task was to count the quarks number
(with allowance made for the color and the flavor). For this purpose the process of the
e+ e− -annihilation into hadrons appeared to be the most suitable

e+ + e− → q + q → 1 jet + 2 jet. (10.130)

The quark and the antiquark, which are produced at the second stage of the reaction,
cannot exist in the free states since they are the color objects. They extract the color
quark-antiquark pair suitable to them from a vacuum and recombine with this pair into
colorless hadrons that fly apart as two jets directed oppositely. At sufficiently high energies
the contributions coming from created quark-antiquark pairs with different colors and flavors
are incoherent, that is, the total cross section of the reaction (10.130) is presented in the
form of the cross sections sum over quarks of all the colors and flavors. The cross section
obtained may be compared with that of the e+ e− annihilation into the other point particle,
muons
e+ + e− → µ+ + µ− . (10.131)
Both the quarks and charged leptons are described by the Dirac equation. However, mass
and charge entering into this equation are quite different for quarks and leptons. In the
case of high energies one may neglect their masses to get the following expression for the
ratio of the cross sections of the reactions (10.130) and (10.131):
P
σe+ e− →hadrons σe+ e− →qi qi X  eq 2
R= = i = i
. (10.132)
σe+ e− →µ+ µ− σe+ e− →µ+ µ− i
e

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 207

If the quarks did not have the color degree of freedom, this ratio would be∗
 e 2  e 2  e 2 4 1 1 2
u d s
R= + + = + + = . (10.133)
e e e 9 9 9 3
The experiment gave three times as much value. But such trebling must appear with the
regard for the color degrees of freedom. Really, every quark-antiquark pair of the given
flavor may be created in three different color states.
The origin of the ordinary strong interaction between real colorless hadrons (for example,
between nucleons in the atom nuclear) is explained by approximately the same way as the
origin of the chemical couplings of the electromagnetic nature between electrically neutral
atoms and molecules in matter, that is, by means of interaction of the charged structural
components. In particular, the π-mesons exchange between nucleons† may be connected
with the quark-antiquark pair production inside a nucleon and a subsequent conversion of
these pairs into the π-mesons that can fly out of the nucleons because they are colorless
objects.

10.5 Heavy quarks and flavor symmetries

Thus, we have managed to build all the hadrons, which were known by 1975, with the help
of the ui , di , and si -quarks. And the understanding illusion visited us once again: all matter
in the universe consists of combinations of nine quarks and four leptons (electron, electron
neutrino, muon, and muon neutrino). It seemed that the problem of matter structure was
close to the completion. It only remained for us to make more precise the properties of
the fields describing the weak and strong interactions. True enough one “but” was again,
namely, the so-called quark-lepton symmetry.
Stable matter consists of the electrons, the u and d-quarks. We have every reason to
believe that the electron neutrino is a stable particle as well. We call this totality by the
first generation of the quarks and leptons. Further, we introduce a new quantum number
S W with the same algebra as the ordinary spin has and consider the representation with
the weight 1/2. Let us place the first generation particles into the weak isospin doublets
   
νe u
, . (10.134)
e− d

Notice, the charges difference of neutrinos and charged leptons are equal to those of the up
and down quarks while the algebraic sum of the charges of quarks (with allowance made
for trebling over the color) and leptons is equal to zero. There exists the second lepton
generation too  
νµ
, (10.135)
µ−
whereas the second quark generation is kept unfilled
 
?
. (10.136)
s

∗ Running ahead we stress, it is valid only when energies are smaller than the creation threshold of hadrons,
which consist of more heavier quarks, that is Ee− e+ < 3 GeV.
† Such a treatment of nuclear forces hold good when the distance between nucleons exceeds 8 × 10−14 cm.

© 2011 by Taylor and Francis Group, LLC

 208 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

If one assumes that the quark-lepton symmetry is the immovable law of nature, then the
existence of the quarks of a new flavor is needed.
In the autumn of 1974 in Brookhaven National Laboratory the group under the super-
vision of C. Ting began investigating the process of the electron-positron pair production
at the collisions of the protons with the helium target (pp-collisions) in the mass region
between 2 and 4 GeV. It was discovered that majority of the e− e+ -pairs has the mass being
approximately equal to 3.1 GeV, that is, the pronounced maximum takes place in the cross
section of the process
p + He → e+ + e− + X, (10.137)
where we have designated the undetected particles plurality by the symbol X. If this
maximum corresponds to the true resonance, then it should be present in the cross sections
of some other reactions. And really, at about the same time the group working at the
SPEAR facility (B. Richter as a supervisor) discovered this resonance under investigating
the processes

e+ + e− → hadrons, e+ + e− → e+ + e− , e+ + e− → µ+ + µ− . (10.138)

Both groups simultaneously reported about the discovery of the new particle with the mass
∼ 3.1 GeV. Since it was highly difficult to decide who bore the palm, then the new particle
was called the double name J/ψ (the symbol J means Ting’s name in Chinese and the
name ψ was proposed by the SPEAR Collaboration). The name J/ψ has been saved not
only as a tribute of respect to its ground breakers but also due to a play on words, J/ψ=
gi/psi=gipsy, which has been so amusing to the romantic soul of a physicist.
The particle J/ψ has the spin 1, the negative parity and it is long-lived by the microworld
standards, that is, it has the anomalously small decay width Γ ≈ 70 keV whereas ordinary
resonances have Γ = 100—200 MeV. To clarify, the true nature of the new particle was being
proceeded by about three years. Among the working hypotheses that have appeared, there
was even the hypothesis: J/ψ is not a hadron but the long-awaited neutral intermediate
boson, one of carriers of the weak interaction. With the passage of time larger and larger
arguments arise what begin to turn the scale in favor of the hypothesis: J/ψ consists of a
quark and a antiquark, cc, with a new flavor. From the simplest estimation mc ≃ mJ/ψ /2 =
1.55 GeV it follows at once that the c-quarks should be much more heavier than the u, d, and
s-quarks. To match the theory and experiments one should make an assumption that the
c-quark is the carrier of the new quantum number, the charm, c = 1 for c-quark and c = −1
for c. Then J/ψ represents the particle with the hidden charm and, on its structure, is very
similar to ϕ-meson which is constituted of the strange quark and antiquark (the particle
with the hidden strange). This resemblance has served as a key to understand the small
decay width of the discovered particle. In Figs.10.13(a) and 10.13(b) the quark diagrams
corresponding to the ϕ-meson decay are displayed. The distinction from Feynman diagrams
resides in the fact that here the quarks, when t = ±∞, are not free particles because hadrons
hold them captive as before. Besides, in these diagrams the strong interaction between
quarks are not usually displayed. Just the same as in the case of the Feynman diagrams
the arrow directed backwards the time∗ corresponds to the antiparticle. In Fig. 10.13(a)
the s-quarks entering into the ϕ-meson composition get over to the K-mesons composition.
In the second case (Fig. 10.13(b)) the strange quarks are annihilated and instead of them
the pairs of the u- and d-quarks appear. The experiments show the ϕ-meson predominantly
decays through the channel K + K (the relative probability, the branching, Br ≈ 84%) and
very reluctantly decays through the channel π + + π − + π 0 (Br ≈ 15%). At this example

∗ On the diagrams the time flies from left to right.

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 209

u
π+
s d s
K−
d s u
π0
d s u
K+
s d s
π−
u
ϕ → π+ + π0 + π− ϕ → K+ + K−

(a) (b)

FIGURE 10.13
The quark diagrams (a) and (b) of the ϕ-meson decay.

we see how to work the approximate semi-phenomenological rule by Okubo–Zweig–Iizuka

(OZI) [45], which assumes the systematization of relative amplitudes of reactions of hadrons
interaction depending on a topology of quark diagrams to display these processes. The
largest degree of suppression is present in the diagrams where the quarks and antiquarks
lines going out of one and the same hadron are connected with each other and represent the
block that is not related with the remaining part of the diagram. In this case the quark-
antiquark pair belonging to one and the same hadron disappears. An alternative to such a
process is the process where the same quark and antiquark pass into the different hadrons
of the final state. All this shows once again the behavior uncommonness of the quarks. In
processes of inclusive scattering the created quark-antiquark pairs behaved in such a way as
though they beforehand deduced in what groups on two (mesons) or three (baryons) they
should be unified. The OZI rule may be also understood as the manifestation of the specific
quark “thinking,” namely, the quarks choose the possible variant of the future hadron prison
already at their creating. The OZI rule, as applied to the J/ψ-particle, means J/ψ would
predominantly decay through the particles containing the c-quark and the light u-and d-
quarks (Fig. 10.14(b)). These particles having the explicit charm are called the D-meson.

u
π+
c d c
D+
d c u
π0
d c u
D−
c d c
π−
u
+
ψ →π +π +π 0 − ψ9D+ + D−

(a) (b)

FIGURE 10.14
The quark diagrams (a) and (b) of the J/ψ-particle decay.

© 2011 by Taylor and Francis Group, LLC

 210 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

However, by a lucky chance, the decay

J/ψ → D+ + D− (10.139)
proved to be forbidden energetically (2mD > mJ/ψ ). The decays into the π- and K-mesons
were allowed energetically but they were connected with the annihilation of the c-quarks
and, as a result, were greatly suppressed by virtue of the OZI rule (Fig. 10.14(a)). Thus,
the small value of the J/ψ-meson decay width served as the indication of the existence of
the new kind of quark c.
About a week later after the J/ψ-meson discovery, at SPEAR the narrow resonance ψ ′
placed at slightly more high energy was detected. With further increasing the energy one
had found some resonances both with the spin 1 and with the spin 0 in the neighborhood of
4 GeV. If one displays the mass spectrum of these particles (we call them the ψ-particles)
graphically, then this spectrum will resemble the atom spectral lines picture (Fig. 10.15).
According to contemporary concepts, this mass spectrum can be related to the energy levels

m, M eV

Ψ(2S)
3700 γ
ηc (2S)

γ
3500 χc2 (1P )
hadrons χc1 (1P )

γ χc0 (1P )
γ
3300
γ γ
hadrons
3100
J/Ψ(1S)

2900 ηc (1S)

JPC = 0−+ 1−− 0++ 1++ 2++

FIGURE 10.15
The charmonium spectrum.

spectrum of the system cc. By analogy with the well-known positronium the bound system
cc is called the charmonium. However, unlike the positronium where the binding forces have
the electromagnetic nature, the charmonium owes its existence to the strong interaction.
The particles denoted by a prime represent the radial excitations of the underlying states.
To classify the charmonium levels we make use of the spectroscopic designations 2S+1 LJ .
Of course, this is a nonrelativistic classification but the nonrelativistic approach is lawful
because of the c-quark large mass (by now experiments give the current c-quark mass value
in the interval from 1.15 to 1.35 GeV).
From the data concerning only ψ-particles it was impossible to make the final conclusion
in favor of the existence of the c-quark. The hypothesis about the charm quark is ultimately

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 211

confirmed only after the discovery of hadrons with the explicit charm (1976): mesons

D+ = cd, D0 = cu,
0
D− = dc, D = uc,
F + = cs, F − = sc,
and baryons
Λ+ = cdu, Σ+
c = udc, Σ++
c = uuc, . . .
These discoveries furnished the genuine triumph of the quark theory of hadrons structure.
Since the mass of the discovered fourth quark was much bigger than the masses of the u-,
d-, and s-quarks it was already hard to say of the existence of the SU (4)-symmetry (flavor
symmetry) in the hadron world. Notice, the SU (2)-symmetry violation is ∼ 1% while the
SU (2)-symmetry violation is 10÷20%. According to a contemporary point view, a violation
of a SU (nf )-symmetry is caused by the quark masses difference; namely, the more apparent
this difference the stronger the violation of the corresponding SU (nf )-symmetry.
However, flavor symmetries are being used as before to classify hadrons. Now, to graph-
ically picture the SU (4)-group representations one should draw on a three-dimensional co-
ordinate system. Let us agree to plot the charm value on the axis z, the isospin projection
value on the axis y, and the strange value on the axis x. In Figs. 10.16, and 10.17 the SU (4)
16-plet for pseudoscalar mesons and the SU (4) 20-plet baryons are shown, respectively. The
shaded areas correspond to the SU (3)-multiplet with the equal value of the charm. Soon

FIGURE 10.16
The 16-plet for the pseudoscalar mesons made of the u, d, s, and c quarks.

after the J/ψ-meson discovery in 1975 one more charged lepton, a τ -lepton, was detected
in the independent experiments series. The direct confirmation of the existence of the ντ -
neutrino will have taken place much later (2001). However, even in 1975, the majority of
physicists based upon the idea of the unity of nature, was considering that the τ -lepton also
has the neutrino satellite. Thus, the quark-lepton symmetry that had been rebuilt with the
c-quark discovery was violated again. The new lepton pair should correspond a new quark
pair. The first confirmation of this hypothesis happened in 1977. In the Fermi laboratory
(FERMILAB) at the proton accelerator with the energy up to 400 GeV, a new particle
with the mass 9.45 GeV and J P = 1− was discovered. It was called Υ-particle (upsilon).

© 2011 by Taylor and Francis Group, LLC

 212 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 10.17
The 20-plet of baryons made of the u, d, s, and c quarks.

Subsequently, in DESY (Hamburg) and at Cornell University (USA) at investigation of the

electron-positron collisions the existence of the Υ-particle was confirmed and its excited
states (Υ′ , Υ′′ and so on) were detected. Since the properties of the Υ-particle are very
similar to those of the J/ψ-meson then one assumed that it is the bound state of the quark-
antiquark pair of the fifth kind, Υ = bb. As the particles belonging to the Υ family (we call
it upsilonium) have masses around 10 GeV, then the mass of the quarks entering into them
must be by no means smaller than 5 GeV (according to the contemporary data the mass
of the current b-quark lays in the interval from 4 to 4.4 GeV). The investigations of the
Υ-meson properties have shown the b-quark has I = 0 and Q = −1/3. Physicists were going
on the already well-paved road for the small width of the Υ-meson decay to be explained.
The new quantum number, the beauty b = 1, which is conserved in the strong interaction
only, was assigned to the b-quark. Then the OZI rule suppressed decays into hadrons that
do not contain b-quarks and nature, for its part, in order not to lead the researchers into
temptation, gave orders in such a way that the decays into a pair of mesons with the explicit
beauty proves to be forbidden energetically (the most light beauty mesons B + = ub and
B 0 = db have the mass 5270 MeV). As we know the quantity
σe+ e− →hadrons
R=
σe+ e− →µ+ µ−

is a sensitive tool to define the quark flavors number. Between thresholds of a qi q i -pairs
production the quantity R is constant and when the next i threshold had been achieved
it was abruptly increased on the value 3Q2i . In Fig. 10.18 we display the experimentally
measured values of Rexp as a function of energy in the center-of-mass frame. The resonant
levels of the charmonium and upsilonium are put on the stepped monotonous behavior.
Upon subtracting the resonant contributions Rexp is well matched with the experimental
value, which is the powerful argument in favor of the b-quark existence.
To date, several tens of mesons and several baryons with the explicit beauty are known.
Contrary to them the upsilonium represents the state with the hidden beauty.
Thus, in the late eighties physicists have at their disposal the two quarkoniums, the
charmonium and upsilonium. In the QCD the quarkonium role is similar to the hydrogen

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 213

E cm , GeV

FIGURE 10.18
The experimental values of Rexp as a function of the electron-positron beam energy.

role in Atomic Physics. The quarkoniums serve as probes of the strong interaction at
small distances. In particular quarkonium decays are the gluons source, that is, here gluon
properties and specificity of gluons conversion into hadrons could be investigated.
After the b-quark discovery the hunting season is opening on its partner on the weak
isospin doublet, which was called the t-quark. Its name originates from the first letter of the
word top, though one may consider that the origin is the word truth. At first, the indirect
evidences for the existence of the sixth quark were obtained. Among those, the results of
investigating the process
e+ + e− → b + b (10.140)
at the CERN electron-positron collider called LEP (Large Electron Positron) were the most
convincing. To study the asymmetry in the scattering of the b-quark confirmed the b-quark
does represent the member of the SU (2) doublet of the weak isospin, as the electroweak
interaction model by Glashow, Weinberg, and Salam (GWS) [46] predicted. Recall, by
that time the GWS theory had gone through the basic stage of its experimental checkout.
Moreover, the detailed measurements of the properties of the W ± - and Z-bosons that were
fulfilled at LEP, SLC (Stanford Linear Collider, e+ e− -collider), CERN SppS (pp-collider),
and FERMILAB Tevatron (pp-collider) gave the unquestionable proofs in favor of the exis-
tence of the heavy quark connected with the b-quark by means of the electroweak interaction.
In April 1994 the first experiments, which directly point the t-quark existence, appeared.
They were carried out at FERMILAB Tevatron, which represents the six-kilometer storage
ring where the protons and antiprotons having the energy 1.8 TeV in the center-of-mass
frame are rotated in opposite directions. The pp collisions occur in two points of the ring
where two detectors system, CDF and D⊘ detectors, are placed. The collider luminosity∗
was being constantly increased and in 1994 it reached the value L = 1031 cm−2 · s−1 , which
allowed one to detect the t-quark. At looking for the t-quarks one is guided by the fact

∗ The luminosity L is the collisions number in a second on the unit section, that is, L, being multiplied by
the cross section of the given process σ that expressed in cm2 , gives the number of the corresponding events
in a second.

© 2011 by Taylor and Francis Group, LLC

 214 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

that the cross sections both of their productions and of their consequent decays depend on
the t-quark mass value. The first to raise the low bound on mt to ∼ 175 GeV was CDF
Collaboration. Already in 1994 they observed twelve events with mt ∼ 175 GeV that corre-
spond to the selection criterions for the t-quark. But the expected background constituted
roughly six events and, although the interpretation of the obtained results as the t-quark
observation offered the most probable, the statistical providing of the result was insufficient
to recognize it as the t-quark discovery. And only in February 1995, in one and the same
day, both Collaborations, CDF, and D0 sent off the reports about the t-quark detection to
the press.
According to the QCD, at energy in the center-of-mass frame E = 1.8 TeV, the tt pair
production in pp collisions mainly occurs at the cost of the subprocesses:

q + q → t + t, g + g → t + t, (10.141)

where we have designated the gluons by the symbol g. We choose the axis z along the
proton beam direction and shall study such variables of the final products of a reaction
as the transverse momentum relative to z axis pT and connected with it the transverse
energy ET . The standard model (SM) predicts that the t-quark decays through the channel
t → W + b just about always. Distribution over the transverse momentum for one of the
decay products has the peculiarity that enables one not only to detect such decays but to
define the mass of the unstable particle as well. This method based on merely kinematic
arguments became very popular after it had been used under discovering the W -boson
in pp collisions (in that case, the electrons and the positrons with very high transverse
momenta, pT ∼ 40 GeV, were detected). At sizeable excess of the decaying particle mass
mi over the decay product masses, the events concentration maximum is observed close
to the transverse momentum value of one of the decay products kT to be approximately
equal to mi /2. However, this method of the t-quark identification did not work since the
W -boson was the unstable particle with a very small lifetime while the b-quark served as
a source of producing the hadron jets. The W -boson decays into qq ′ -pairs (ud or s) with
the probability 2/3 and decays into one of three lepton families (lνl ) with the probability
1/3. Thus, there are 6 high-energy point fermions that could be either charged and neutral
leptons or quarks giving rise to the production of the hadron jets. It was very difficult to
select the decays with the τ -leptons from the hadron background and, for this reason, they
were not taken into consideration. High background (signal-to-noise correlation was smaller
than 10−4 ) made an inclusion in analysis of final states with six hadron jets impossible. As
a result the following channels decay of the tt-pair were available to observe

tt → e+ νe be− ν e b, (1/81), (10.142)

tt → µ+ νµ bµ− ν µ b, (1/81), (10.143)

± ∓
tt → e νe bµ ν µ b, (2/81), (10.144)
± ′
tt → e νe bqq b, (12/81), (10.145)
tt → µ± νµ bqq′ b, (12/81), (10.146)
where the numbers in the parentheses denote the branching predicted by the SM. First three
dilepton channels prove to be the most clear but they have very small statistics. Last two
channels, lepton+hadron jets, have high statistics (close to 30% on the total decay width
tt) but suffer from extremely large background. At pp collisions the total cross section of
the tt-pair production depends not only on the parton distributions inside the proton and
the antiproton but on the t-quark mass as well. So, at mt = 100, 155 GeV it has the values
∼ 102 and 10 pb, respectively. It should be stressed that these values are very small by

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 215

the strong interaction standards. When E = 1.8 TeV the pp collisions also lead to the
production of hadron jets and lepton pairs in the final state at the cost of the production of
the W - and Z-gauge bosons in virtual states. It has been just these events that constitute
the main background for the tt-pair production. To compare, we give the values of the
competed cross sections:

(i) for the W -production—σW ∼ 20 nb;

(ii) for the Z-production—σZ ∼ 2 nb;
(iii) for the W W -production—σW W ∼ 10 pb;
(iv) for the W Z-production—σW Z ∼ 5 pb.

Let us consider the criterions used by CDF and D0 Collaborations to select the signal
from the background. The heavy quark pair production (mt = 173.8 ± 5.2 GeV) and
its consequent decay generates the final states with more large average energy than the
background events. To investigate the distribution of the final hadrons and leptons over
the transverse momenta proves to be useful in this case as well. High-energy electrons,
muons, and hadrons were recorded with the help of different detectors and were easily
distinguishable from each other. As for the neutrino is concerned the disbalance of the
total transverse energy ET or the missing transverse momentum pT testify to its existence
in the final state. It is evident that the quantity
Np
X
HT = ETi ,
i=1

where the summation is realized over all hadron jets and basic electron clusters by which
are meant a leptons plurality, including even if one electron, is very useful for the kinematic
analysis. At the investigation of the final states lepton+hadron jets, two methods were used
by D0 Collaboration to select the signal from the background. The former based upon
the kinematic analysis (KA) resided in the demand HT > 200 GeV and in the presence of
at least four hadron jets with ET > 15 GeV. The second method was connected with the
b-quark identification (B tagging-out) through the decays
b → µ− + ν µ + X, b → µ+ + νµ + X. (10.147)
Studying the data on lepton+hadron jets the CDF Collaboration already used other meth-
ods. One of them was also connected with the b-quarks and received the name: “tagging-out
of the second vertex (TSV).” The tracks of charged products of the b-quarks decay are de-
tected in the drift chamber. Only the tracks with pT > 1.5 GeV (decays of the b-quarks to
be produced from the tt-pairs and the W -bosons are their sources) were of interest. The b-
quark decay point in the silicon stripped detector was rebuilt by the extrapolations method.
It allowed selecting the signal from a background by means of a comparison of a different
events intensity. The second method received the name “tagging-out of soft leptons (TSL),”
was based on detecting the low-energy (pT ∼ 2 GeV) muons and electrons near hadron jets.

In Table 10.1 we give the expected number of the background events and the number of
the events connected with the t-quark production in the observation results by CDF and D0
Collaborations. As it follows from Table 10.1, in all channels both Collaborations observed
the sizeable excess of the signal over a background that undoubtedly testifies to the tt-pairs
production in the pp collisions.
To account for the experimental data one should assign to the t-quark a new quantum
number t (the top or the truth) equal to 1 which along with the strange, the charm, and the

© 2011 by Taylor and Francis Group, LLC

 216 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
TABLE 10.1
The observation results for the t-quark production at
CDF and D0.
Sampling Background Signal
dileptons (CDF) 1.3 ± 0.3 6
dileptons (D0 0.65 ± 0.15 3
leptons+jets (D0 KA) 0.93 ± 0.5 8
leptons+jets (D0 B-tagging-out) 1.21 ± 0.26 6
leptons+jets (CDF TSV) 6.7 ± 2.1 27
leptons+jets (CDF TSL) 15.4 ± 2.3 23

beauty is conserved in the strong interaction but is not conserved in the electromagnetic
and weak interactions.
Below in Table 10.2. we give the additive quantum numbers of all the six quarks. It was

TABLE 10.2
The quantum numbers of the quarks.
q Q J I I3 B s c b t
u 2/3 1/2 1/2 1/2 1/3 0 0 0 0
d -1/3 1/2 1/2 -1/2 1/3 0 0 0 0
c 2/3 1/2 0 0 1/3 0 1 0 0
s -1/3 1/2 0 0 1/3 -1 0 0 0
t 2/3 1/2 0 0 1/3 0 0 0 1
b -1/3 1/2 0 0 1/3 0 0 -1 0

natural to wait for the existence of a quarkonium consisting of the t-quark and t-antiquark
(t), the toponium. However, by now the toponium has not been discovered. It is not
inconceivable that nature presented us only two kinds of the simplest “quark” atoms with
the great numbers of energy levels. Recall, that in Atomic Physics the hydrogen atom was
such a present. But in Nuclear Physics, mildly speaking, nature already was a little bit
miserly and constrained itself by the deuteron only which has got none of the excited levels.
Introducing the b- and t-quarks extends the flavor symmetry of the strong interaction
up to the SU (6)-group. However, thanks to the sharp gradation of the quark masses
this symmetry is strongly violated. On the other hand, since up to the present hadrons
containing the constituent t-quark have not been discovered, then for practical calculations
we could successfully use the SU (5) flavor symmetry whose violation degree is much less
(mb ≪ mt ). The flavor symmetry plays a double role, it not only defines the classification of
hadrons on various multiplets but it also establishes a series of dynamical relations between
amplitudes of different processes of hadrons interaction. In the low energy region (E ≤
m0 , where m0 is an average mass value in a multiplet) strongly violated flavor symmetry
possesses a weak predictive force and is unlikely applicable for practical use. However, at
high energies (E ≫ m0 ) it becomes the useful tool of investigation.
Among unsolved problems of quark mechanics the problem of existing quarks in the free
state appears to be principal. There is nothing to prevent us from conjecturing that the
negative results on the quarks production are caused only by the insufficient energy of

© 2011 by Taylor and Francis Group, LLC

 Quark “atoms” 217

contemporary accelerators to create the quarks in a free state. This, in turn, means that at
the hadrons production the giant quark mass is eaten by such huge binding energies. And
only the fact of the free quarks existence can prove this statement.
Looking for the free quarks is being carried on two directions. The former is a geophysical-
chemical approach, which is based on the assumption: the stable quarks occur in matter
surrounding us. In this case the negative charged quarks will be captured by nuclei and form
either quark atoms or quark ions with the noninteger charge. It is evident that the created
compounds should have specific physical and chemical properties. Thus, the experiments of
this approach are aimed for discovering the characteristic manifestation of the fractionally
charged particles existence: the lowered ionization constituting 1/9 or 4/9 of the integer
charged particle ionization; an unusual value of e/m in mass-spectroscopic experiments;
an anomalous behavior of a levitating matter grain in an electrostatic field; a nonstandard
position of spectral lines in quark atoms, and so on. The quarks were being sought in
terrestrial matter, in lunar soil, in meteorites. To look for quark atoms in solar matter with
the help of spectroscopic methods was also carried out. It is natural that the most reliable
constraints on the free quarks existence have been obtained during searching for the quarks
in the stable matter of the Earth. Different variants of experiments result in that the upper
limit values of a possible quarks concentration in matter lay in the interval from 5 × 10−15
quarks
to 5 × 10−28 nucleon .
The second direction includes the attempts of detecting the quarks in cosmic rays and
accelerators directly. In experiments on accelerators the free quarks are not discovered up
to the masses 250 GeV in pp collisions (CDF, 1992) and 84 GeV in e− e+ collisions (LEP,
DELPHI, 1997) at the production cross sections higher than 10 and 1 pb, respectively.
Experiments on detecting the quarks in collisions of cosmic rays with particles in the upper
atmosphere layers, which were fulfilled in a wide interval of energies and, consequently, did
not have severe constraints on the mass of created quarks, have not also brought to success
and have set the constraint on the quarks flux from the cosmos: < 2.1×10−15 cm2 ·quarks
steradian·s
(KAM2, 1991).
Meanwhile, if the free quarks existence is not forbidden in principle, then they would
be created at early stages of the universe evolution when the temperature was very high,
say kT > 2mq . At such a temperature the quarks are in the state of the thermodynamic
equilibrium with other fundamental particles (number of created quarks is equal to that
of annihilating quarks). At kT ∼ mq the equilibrium is violated: reactions of the quarks
production have been switched off and the quarks start to burn away. This burning away
takes place at the cost of reactions of the kind

q + q → mesons, q + q → q + baryon. (10.148)

Since the reactions (10.145) are exothermic, then their cross sections tend to constant
values whose sum is denoted by σ0 . If one assumes that the cross section of the quarks
destruction has the typical nuclear scale, say σ0 ∼ m−2 π , one may show the quark-to-
proton-concentration ratio in the present universe must be
nq
∼ 10−12 .
np
This number is greater than the gold abundance on the Earth but the quark Klondike never
has been opened to date. One would assume the quarks are unstable particles and all this
relic quark sea has had time to disappear by now. However, on the strength of the electric
charge conservation law, at least one of the quarks must be stable and should live till the
present day. So, either the free quarks are really absent in nature or their production cross
section has as minimum the atom scale rather than the nuclear one.

© 2011 by Taylor and Francis Group, LLC

 218 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It should be noted that some experimental groups make reports concerning the free quark
observations every now and then. So, the Stanford University group investigated a behavior
of a niobium ball ∼ 10−4 g that levitated in the nonuniform magnetic field. They observed
the cases when the electric charge of the ball was equal to ±e/3. As this takes place, the
quarks
corresponding quarks concentration has the order of ≈ 10−20 nucleon [47]. However, these
data are not confirmed by other investigations yet. And the result is included in a category
of the reliable one, if and only if, it is independently obtained by several different groups
using different experimental methods.
Physics development shows that nature gives answers to correctly posed questions only.
There is no sense in the question that tormented ancient scholastics: “How many angels
could be place into a sword tip?” The question, physicists tortured themselves at the very
outset of Quantum Theory, appeared to be senseless: “What is the electron—the particle
or the wave?” Maybe, when we are trying to detect the quarks in a free state we are in
the analogous situation? It is not expected that the quarks represent the specific kind of
quasi-particles, field quanta, which describe collective oscillations of corresponding freedom
degrees of a hadron. We faced such formations in other regions of physics and before. Among
these are: the magnon, the quantum of the spin oscillations in magneto-ordered systems;
the plasmon, the quantum of the charge density oscillations in conductive mediums; and the
phonon, the quantum of the elastic oscillations of the atoms or molecules in crystal lattice.
At switching off interaction similar particles are pulled down into compound parts and stop
their existence. For example, the phonon decays and turns into a plurality of independent
motions of particles, which constitute a crystal. Then the quarks have the sense only as
dynamical essences inside hadrons in just the same way as the phonons that cannot exist
outside a crystal. However, be it as it may, looking for the free quarks are continued.
The problems of their detection at LHC (Large Hadron Collider) and accelerators of the
next generation, NLC (Next Liner Collider), FMC (First Muon Collider), and so on, are
intensively discussed.

© 2011 by Taylor and Francis Group, LLC

 11
Passing glance on the theory of electroweak
interaction

The point is that the owl and the hawk who had taken joint
leadership of education made a gross blunder. They took it
into their heads to teach grammar to the eagle himself.
But a still greater blunder was that like all tutors, neither
one nor the other gave their pupil a breather. The owl
dogged his footsteps, shouting out bbb, kkk, zzz, while
the hawk unceasingly tried to impress on him that without
the four rules of arithmetic one cannot divide one’s booty.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

11.1 Spontaneous symmetry breaking

The QCD is not the sole descendant of the Yang-Mills theory. The electroweak interaction
theory was also built on the basis of the non-Abelian gauge theory. However, the way of
its production was already more complex. The point is that gauge fields are long-range
by its nature. This fact immediately leads to the zero mass of the interaction carriers as
it takes place both in the QCD and in the QED. The weak interaction exists only at very
small distances and the weak interaction carriers, the W ± - and Z-bosons, must have a huge
mass. So, it was necessary to combine two incompatible things, namely, the local gauge
invariance and the nonzero mass of the W ± - and Z-bosons. It appeared that one needs to
use the idea of a spontaneous symmetry breaking to solve this problem.
An evolution of any physical system is defined by two factors: (i) a form of the Lagrangian
(or the Hamiltonian), and (ii) initial conditions. As an example, we consider a smooth
surface in the form of a peaked hat resting on a horizontal ground (Fig. 11.1). We place
a ball on the hat top and shall consider that the gravitational force is the sole force acting
on the system. It is clear that the system possesses the explicit symmetry with respect to
rotations about a horizontal axis passing through the hat center. However, the system is
not stable. Really, if we move the ball out of position, that is, change the initial condition,
it will be rolled down and the system symmetry will be violated. Having stopped on the
fixed place of the hat brim, the ball sets the selected direction from the central axis. The
system has found a stability at the cost of the symmetry violation. Since the state with
the violated symmetry has more low energy, the ball prefers to roll on the hat brim. In
the stable configuration the initial rotational symmetry of the gravitational force exists as
before but now it exists in the hidden form. The observed system state does not reflect the
symmetry of the interaction that is present in the system.
In quantum field theory two analogous conditions define a character of a symmetry man-
ifestation: (i) a form of the Lagrangian (or the Hamiltonian); and (ii) a vacuum form. In

219
© 2011 by Taylor and Francis Group, LLC
 220 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 11.1
The example of the spontaneous symmetry violation.

the QCD and the QED both the Lagrangian and the vacuum state were invariant under
the symmetry transformations. In such cases we say the symmetry of the theory has not
been violated. In 1960 Y. Nambu and J. Goldstone [48] showed that there exist theories
whose Lagrangian is invariant with respect to symmetry transformations while a vacuum is
not invariant. In this case we use the term “spontaneous symmetry breaking.” Since we do
not want spontaneously to break the Lorentz invariance, spontaneous symmetry breaking
can be realized only in the presence of scalar fields (fundamental or composite). With other
words the vacuum is filled with scalars (with zero four-dimensional momenta) carrying the
quantum number of the broken symmetry.
We investigate the spontaneous breaking of the local gauge symmetry by the example of
the U (1)-group. Consider the world that consists of charged scalar particles only and is
defined by the Lagrangian

L ≡ T − V (ϕ) = [∂µ ϕ(x)]∗ [∂ µ ϕ(x)] − µ2 ϕ∗ (x)ϕ(x) − λ[ϕ∗ (x)ϕ(x)]2 (11.1)

with λ > 0. When µ2 > 0, then the Lagrangian will describe a self-interacting (according
to the law [ϕ∗ (x)ϕ(x)]2 ) scalar field with the mass µ2 . In this case the value ϕ(x) = 0
corresponds to the vacuum (the minimum of V (ϕ). In other words, the average over the
vacuum of ϕ(x) turns into zero (< 0|ϕ(x)|0 >= 0). However, we wish to study the case
with µ2 < 0. If one introduces the field functions according to the relation

ϕ1 (x) + iϕ2 (x)

ϕ(x) = √ ,
2

the Lagrangian (11.1) takes the form:

1
L= {[∂µ ϕ1 (x)]2 + [∂µ ϕ2 (x)]2 − µ2 [ϕ21 (x) + ϕ22 (x)] − λ[ϕ21 + ϕ22 (x)]2 /2}. (11.2)
2
From this writing it is evident that the minima of the potential V (ϕ) lie on the circle of the
radius v in the plane ϕ1 , ϕ2 (see, Fig. 11.2)

µ2
ϕ21 + ϕ22 = v 2 , v2 = − . (11.3)
λ

This means the constant scalar field, the so-called scalar field vacuum condensate, exists in
the vacuum. This quantity, that is, the energy shift of the ground state, cannot be measured

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 221

FIGURE 11.2
The potential of the complex scalar field.

directly since the quantity under test is the difference between the given energy and the
vacuum energy.
In order to make the Lagrangian (11.2) be invariant with respect to the local gauge
transformations of the
ϕ′ (x) = exp[iα(x)]ϕ(x),
we introduce the covariant derivative
Dµ (x) = ∂µ + ieAµ (x), (11.4)
where Aµ (x) is transformed by the law
1
A′µ (x) = Aµ (x) − ∂µ α(x). (11.5)
e
Having supplemented the obtained Lagrangian with the free gauge bosons Lagrangian we
arrive at
L = [∂µ − ieAµ (x)]ϕ∗ (x)[∂ µ + ieAµ (x)]ϕ(x) − µ2 ϕ∗ (x)ϕ(x) − λ[ϕ∗ (x)ϕ(x)]2 −
1
− Fµν (x)F µν (x). (11.6)
4
Within elementary particle physics in the majority of instances we cannot obtain exact
solutions. We most commonly have to use the perturbation theory series expansion and
calculate fluctuations close to a minimum energy. If we try to carry out the expansion in a
neighborhood of the unstable point ϕ = 0, the perturbation theory series will not converge.
The correct activity method is to carry out the expansion in a neighborhood of the minimum
of the potential V (ϕ), that is, in a neighborhood of the stable vacuum. For the minimum we
choose the point ϕ1 = v, ϕ2 = 0. Notice, the vacuum is not already invariant with respect
to the U (1)-group , that is, the symmetry appears to be spontaneously broken. To expand
L in a neighborhood of the vacuum we introduce real fields η(x) and ξ(x), which describe
quantum fluctuations around this minimum
η(x) + iξ(x) + v
ϕ(x) = √ . (11.7)
2

© 2011 by Taylor and Francis Group, LLC

 222 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

To substitute (11.7) into the Lagrangian (11.6) leads to the expression

1 1 1
L′ = {[∂µ ξ(x)]2 + [∂µ η(x)]2 } − v 2 λη 2 (x) + e2 v 2 Aµ (x)Aµ (x) − Fµν (x)F µν (x) + Lint ,
2 2 4
(11.8)
where Lint denotes the terms describing interactions of η(x), ξ(x), and Aµ (x) fields. From
the Lagrangian (11.8) follows
√
mξ = 0, mη = 2λv 2 , mA = ev. (11.9)

In summary, we have attained the goal to be sought. Our gauge bosons have found the
mass. However, as this takes place, the other problem connected with the occurrence of
a massless scalar particle has appeared. Such particles are called Goldstone bosons. Let
us gain an understanding of the situation. Having given the mass to the field Aµ (x), we
thereby increased the number of polarization degrees of freedom from 2 to 3, because now
the field Aµ (x) can have the longitudinal polarization too. But the simple shift of the field
variables that is given by Eq. (11.7) cannot create new degrees of freedom in any way. It
is obvious that not all the fields entering into L′ correspond to the physical particles. It is
beyond doubt that just the Goldstone boson arose suspicion. Let us show that it does not
really belong to the physical sector. Since the theory is gauge invariant, we can carry out
any gauge transformation (fix the gauge∗ ) and the physical contents of the theory is kept
invariable. The approximate equality
 
η(x) + iξ(x) + v η(x) + v iξ(x)
ϕ(x) = √ ≈ √ exp (11.10)
2 2 v

which is true to lowest order in ξ(x) can suggest the required transformation form. It is
clear that we should introduce new fields of the kind
 
′ iξ(x) η(x) + v
ϕ (x) = exp − ϕ(x) = √ ,
v 2
1
A′µ (x) = Aµ (x) + ∂µ ξ(x).
ev
Further, taking into account
   
iξ(x)  ′ ′ ′
 iξ(x) 1
Dµ ϕ(x) = exp ∂µ ϕ (x) + ieAµ (x)ϕ (x) = exp √ {∂µ η(x)+ieA′µ (x)[η(x)+v]},
v v 2
and
1
|Dµ ϕ(x)|2 = |∂µ η(x) + ieA′µ (x)[η(x) + v]|2 , Fµν = ∂µ A′ν (x) − ∂ν A′µ (x),
2
we rewrite the Lagrangian (11.6) in the following form:

1 µ2 λ 1
L′′ = |∂µ η(x)+ieA′µ (x)[η(x)+v]|2 − [η(x)+v]2 − [η(x)+v]4 − Fµν (x)F µν (x) = L0 +Lint ,
2 2 4 4
(11.11)
where
1 µ2 2 1 1
L0 = [∂µ η(x)]2 − η (x) − [∂µ A′ν − ∂ν A′µ ]2 + (ev)2 A′µ (x)Aµ′ (x),
2 2 4 2

∗ The gauge that excludes the Goldstone boson is called the unitary one.

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 223
1 2 ′ λ
Lint = e Aµ (x)Aµ′ (x)η(x)[η(x) + 2v] − λv 2 η 3 (x) − η 4 (x).
2 4
To sum up, the field ξ(x) disappears from the Lagrangian, that is, the seeming additional
degree of freedom connected with the Goldstone boson proved spurious in practical situ-
ations. Obviously, this degree of freedom corresponds to the liberty of the choice of the
gauge transformation. The Lagrangian describes a mere two interacting massive particles,
the vector gauge
√ boson Aµ (x) with the mass M = ev and the scalar particle η(x) with
the mass m = 2µ (the Higgs boson). Having absorbed the Goldstone boson the massless
gauge field became massive. The number of degrees of freedom has been conserved since
the massive vector particle with three spin projections appeared from the massless vector
field, having two spin states, and from the massless scalar particle. In field theory this
phenomenon was discovered in 1964 and named the Higgs mechanism.

11.2 Glashow–Weinberg–Salam theory

The symmetry SU (2)L × U (1)Y was first proposed by S. Glashow (1961) and then it was
expanded to include the massive vector bosons (W and Z) by S. Weinberg and A. Salam
(1967–1968). At building the electroweak interaction theory by Glashow, Weinberg, and
Salam (GWS) [46] one may single out four basic stages. It should be stressed that the
analogous division takes place for any version of an electroweak theory as well.
First, one should choose a gauge group G in such a way that it includes all the necessary
vector particles. The gauge symmetry SU (2)L × U (1)Y ∗ is the base of the GWS theory.
Here SU (2)L represents a group of the weak isospin while U (1)Y represents a group of
the weak hypercharge (the term “weak” is used in order to stress the difference from the
corresponding characteristics of the strong interaction). Thus, the weak isospin SW and the
weak hypercharge Y W (more precisely, Y W /2) are generators of the gauge transformations
of SU (2)L and U (1)Y , respectively. The weak hypercharge is assigned to every field so that
the analog of the formula by Gell-Mann-Nishijima is fulfilled

YW
Q = S3W + . (11.12)
2

The unbroken local SU (2)L × U (1)Y -symmetry demands the existence of four massless
vector bosons. Three of them W 1 , W 2 , W 3 represent gauge bosons of the non-Abelian
group SU (2)L and their interaction is characterized by the gauge constant g. B describes
the gauge field of the Abelian group U (1)Y and its interaction is determined by the gauge
constant g ′ .
In the second stage one should choose the representation of the symmetry group for the
matter particles (leptons and quarks). Since the theory will describe weak processes which,
as it is well known, do not conserve the parity, then the theory must be explicitly mirror-
asymmetric from the beginning. This asymmetry is realized as follows. The left-hand
components of the fermions
1
ψL (x) = (1 − γ5 )ψ(x)
2

∗ We used the designation SU (2)EW × U (1)EW for this group before.

© 2011 by Taylor and Francis Group, LLC

 224 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

form the weak isospin doublets with respect to the SU (2)L -group
     
νeL νµL ντ L 1
, , , S W = , Y W = −1 (11.13)
e−
L µ−
L τL
−
2
     
uL cL t 1 1
, , L , SW = , YW = , (11.14)
dL sL bL 2 3
while the right-hand components of all fermions excepting the neutrinos
1
ψR (x) = (1 + γ5 )ψ(x)
2
represent the weak isospin singlets

e− − −
R , µR , τR , S W = 0, Y W = −2,

4
u R , c R , tR , S W = 0, Y W =
, (11.15)
3
2
dR , sR , bR , S W = 0, Y W = − .
3
The absence of the neutrino singlets in the GWS theory was connected with the fact that
the neutrino was considered the massless particles by the time of this theory production.
At the SU (2)L × U (1)Y global transformations the transformation law of the left-hand
and right-hand components of the field ψ(x) has the form:

′ βY W iβY W
ψL (x) = exp[i(α · SW + )]ψL (x), ′
ψR (x) = exp[ ]ψR (x). (11.16)
2 2
We begin our consideration with the leptons. The lepton sector of the GWS theory is
described by the Lagrangian
X
Ll = i [ψ lL (x)γ µ ∂µ ψlL (x) + ψ lR (x)γ µ ∂µ ψlR (x)]. (11.17)
l=e,µ,τ

To introduce the mass terms into the Lagrangian (11.17) directly

ml [ψ lL (x)ψlR (x) + ψ lR (x)ψlL (x)]

violates the gauge invariance. This makes us use the mechanism of the mass generation at
the cost of the spontaneous symmetry breaking not only for the weak interaction carriers
but for leptons as well. For this purpose the Lagrangian of the Yukawa type that describe
interactions between leptons and Higgs fields is used
X
LY = − fl [ψ lL (x)ψlR (x)ϕ(x) + ϕ† (x)ψ lR (x)ψlL (x)]. (11.18)
l=e,µ,τ

In order to make the neutrino massive it is sufficient to introduce the neutrino singlets in
the theory. Notice, there is no theoretical principle that fixes the Yukawa constants fl and
they unfortunately remain arbitrary parameters of the theory.
In the third stage one needs to localize the gauge group in question, that is, carry out
the replacement
α → α(x), β → β(x).

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 225

This, as it is known, demands the transition to the covariant derivatives and an introduction
of the free gauge bosons Lagrangian. In the case of the SU (2)L × U (1)Y -gauge group the
covariant derivatives for the fields entering into the Lagrangian have the form:

YW
Dµ = ∂µ − igSW · Wµ − ig ′ Bµ . (11.19)
2
Then, recalling that at S W = 1/2 the matrices σk /2 are the generators of the SU (2)
transformations, we obtain for the fields ψL (x) and ψR (x)
 
ig ig ′
Dµ ψlL (x) = ∂µ − σ · Wµ (x) + Bµ (x) ψlL (x), (11.20)
2 2

Dµ ψlR (x) = [∂µ + ig ′ Bµ (x)] ψlR (x). (11.21)

The free gauge bosons Lagrangian is determined by the expression
1 1
L0 = − Waµν (x)Waµν (x) − Bµν (x)B µν (x),
4 4
a
where Wµν (x) is a tensor of a non-Abelian field

Waµν (x) = ∂µ Waν (x) − ∂ν Waµ (x) + gεabc Wbν (x)Wcµ (x), a, b, c = 1, 2, 3,

Bµν (x) is a tensor of an Abelian field

Bµν (x) = ∂µ Bν (x) − ∂ν Bµ (x).

To pass to the covariant derivative leads to the appearance of two basic interactions in the
total Lagrangian: interaction of the weak currents isotriplet Jµ (x) with three vector bosons
Wµ (x)
gJµl (x) · Wµ (x) = gψ lL (x)γ µ SW · Wµ (x)ψlL (x), (11.22)
and interaction of the weak hypercharge current jlY µ (x) with fourth vector boson Bµ (x)

g′ Y µ YW
jl (x)Bµ (x) = g ′ ψ l (x)γ µ ψl (x)Bµ (x), (11.23)
2 2
where
ψl (x) = ψlL (x) + ψlR (x).
In the fourth stage we must give the mass both to the weak interaction carriers and to the
leptons. For this purpose we shall make use of the mechanism of the spontaneous symmetry
breaking on the chain
SU (2)L × U (1)Y → U (1)em .
The supplementary two-component complex scalar Higgs field ϕ(x) (four degrees of freedom)
is introduced
   + 
1 iΦ1 (x) + Φ2 (x) ϕ (x)
ϕ(x) = √ = , S W = 1/2, Y W = 1, (11.24)
2 H(x) − iΦ3 (x) ϕ0 (x)

where Im Φi (x) = 0 and Im H(x) = 0. The Lagrangian describing the Higgs fields doublet,
in addition to the kinetic energy, contains the potential energy of self-interaction as well

LH = |Dµ ϕ(x)|2 − V (ϕ), (11.25)

© 2011 by Taylor and Francis Group, LLC

 226 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where
V (ϕ) = µ2 ϕ† (x)ϕ(x) + λ[ϕ† (x)ϕ(x)]2 (11.26)
2
with λ > 0 and µ < 0. The spontaneous symmetry breakingp is realized by the shift of the
neutral Higgs field component on the real constant v = −µ2 /λ
 
1 iΦ1 (x) + Φ2 (x)
ϕ(x) = √ = ϕ′ (x) + ξ0 , (11.27)
2 H(x) − iΦ3 (x) + v
where  
1 0
ξ0 = √
2 v
and < 0|ϕ′ (x)|0 >= 0. Then the parametrization of fluctuations close to ξ0 in the lowest
order in powers of Φ takes the form:
 
1 0
ϕ(x) = √ exp[iσ · Φ(x)/v] . (11.28)
2 v + H(x)

Notice, at any choice of ϕ(x) the symmetry violation inevitably results in the appearance
of the mass on the corresponding gauge bosons. But when the invariance, both of the
Lagrangian and of the vacuum with respect to some gauge transformation subgroup, is
conserved, then the gauge bosons connected with these subgroups are kept massless. Under
the choice  
1 0
< 0|ϕ(x)|0 >= √
2 v
with S W = 1/2, S3W = −1/2 and Y W = 1 both the SU (2)L - and U (1)Y -gauge symmetries
are violated. Since the generators of the groups U (1)em , SU (2)L and U (1)Y satisfy the
relation (11.12), then
Qξ0 = 0, (11.29)
or
ξ0′ = exp[iα(x)Q]ξ0 = ξ0 . (11.30)
Thus, both the final Lagrangian and the vacuum are invariant with respect to the U (1)em -
group transformations what ensures the zero-mass photon.
As a result of the shift in |Dµ ϕ(x)|2 the terms which are bilinear on the components of
Wµa and Bµ appear. They give the contribution to the mass matrix of the gauge bosons
 2
 ig
 ig ′ 

[− σ · Wµ (x) − Bµ (x)]ϕ(x) =
 2 2 
  
1 2
 gWµ3 (x) + g ′ Bµ (x) g[Wµ1 (x) − iWµ2 (x)] 0 
=   =
8  g[Wµ1 (x) + iWµ2 (x)] −gWµ3 (x) + g ′ Bµ (x) v 

g2 v2  1 v2  2
= [Wµ (x)]2 + [Wµ2 (x)]2 + gWµ3 (x) − g ′ Bµ (x) . (11.31)
8 8
As it follows from Eq. (11.31) the fields Wµ3 (x) and Bµ (x) prove to be mixed. This is no
surprise, since they have the identical quantum numbers. For diagonalization of the last
term in (11.31) we pass to a new basis
    3   3
Zµ 1 g −g ′ Wµ cos θW − sin θW Wµ
= p ′ = , (11.32)
Aµ 2
g +g ′2 g g B µ sin θ W cos θ W Bµ

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 227

where
g′
tan θW = .
g
In the new basis the term v 2 [gWµ3 (x) − g ′ Bµ (x)]2 /8 takes the form:

1 2
m Zµ (x)Z µ (x),
2 Z
where p
g 2 + g ′2
v
mZ = . (11.33)
2
Further crossing to complex self-conjugate fields

Wµ1 ∓ iWµ2
Wµ± = √ , Wµ ≡ Wµ− , Wµ∗ ≡ Wµ+ (11.34)
2
and taking into account (11.33), we rewrite Eq. (11.31) in the form:
1
m2W Wµ∗ (x)W µ (x) + m2Z Zµ (x)Z µ (x), (11.35)
2
where
gv
mW = . (11.36)
2
Thus, the weak interaction carriers have acquired the mass whereas the photon is kept
massless. The massless gauge bosons had two polarization states. After they had acquired
the mass the number of their polarization states increased by one. They borrow these three
additional degrees of freedom from the Higgs bosons. However, the Higgs field had four
degrees of freedom. What destiny has the last Higgs component? It appeared that the
remaining Higgs boson becomes massive and passes into the physical particles sector (the
physical Higgs boson).
Substituting (11.28) into (11.27), one may be convinced that the field H(x) has the mass
m2H = 2λv 2 while the fields Φi (x) have been kept massless, that is, they represent the
Goldstone fields. To eliminate the Goldstone bosons we use the parametrization of the
field ϕ(x) in the form (11.29). We carry out the SU (2)L -gauge transformation in the total
Lagrangian written in terms of ϕ(x)
 
′ 1 0
ϕ (x) = U(x)ϕ(x) = √ ,
2 v + H(x)
 
σk ′ i σk
W (x) = U(x) ∂µ + Wkµ (x) U −1 (x),
2 kµ g 2
′
ψL (x) = U(x)ψL (x), Bµ′ (x) = Bµ (x), ′
ψR (x) = ψR (x)
where
U(x) = exp[−iσ · Φ(x)/v].
The transformation law for Wkµ (x)
follows from the Lagrangian invariance with respect to
U(x), what is ensured by the condition

U −1 (x)Dµ′ (x)U(x)ψL (x) = Dµ (x)ψL (x).

It is not difficult to show the fields Φ1 (x), Φ2 (x), and Φ3 (x) are not already contained in
the final Lagrangian. To sum up, three Goldstone bosons are eliminated by the gauge

© 2011 by Taylor and Francis Group, LLC

 228 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

transformation from the theory (to be gauged) while the liberated three degrees of freedom
cross into the transverse components of the W ± - and Z-bosons, which became massive.
The physical Higgs boson is not isolated from the rest of the part of the model. It interacts
both with the leptons and with the gauge bosons. The corresponding Lagrangian is given
by the expression:
 
X g2 ∗ µ 1  
LH = − fl l(x)l(x)H(x)+ Wµ (x)W (x) + 2
Zµ (x)Z (x) H 2 (x) + 2vH(x) .
µ
4 2 cos θW
l
(11.37)
For the massive neutrinos, as stated before, we should introduce the neutrino singlets in
the theory. This leads to the appearance of the following term in LH :
X
− fνl ν l (x)νl (x)H(x).
l

The leptons also get the masses in the result of the shift of the field ϕ on the constant
(11.28)
fl v fν v
ml = √ , mνl = √l . (11.38)
2 2
As it follows from Eqs. (11.37) and (11.38) the coupling constants describing the interaction
of the Higgs bosons with the W - and Z-bosons are proportional to gmW and gmZ respec-
tively, that is, they are much larger than the coupling constants determining the interaction
between the Higgs bosons and the fermions.
Let us consider the sum of the terms (11.22) and (11.23) that describe the interaction
between the leptons and the gauge bosons of the SU (2)L × U (1)Y -group. Taking into
account the explicit form of the Pauli matrices, we may present an interaction of the weak
currents isotriplet Jµ (x) with three Wµ (x)-bosons in the form:
  
g 0 Wµ1 (x) − iWµ2 (x) νlL (x)
gJµl (x) · Wµ (x) = (ν lL (x), lL (x))γ µ +
2 Wµ1 (x) + iWµ2 (x) 0 l(x)
 3  
g Wµ (x) 0 νlL (x)
+ (ν lL (x), lL (x))γ µ =
2 0 −Wµ3 (x) lL (x)
g  
= √ ν lL (x)γ µ lL (x)Wµ∗ (x) + lL (x)γ µ νlL (x)Wµ (x) + gJl3µ (x)Wµ3 (x), (11.39)
2
where
g
gJl3µ (x)Wµ3 (x) = [ν lL (x)γµ νlL (x)Wµ3 (x) − lL (x)γµ lL (x)Wµ3 (x)]. (11.40)
2
Interaction of the weak hypercharge current jlY µ (x) with the fourth vector boson Bµ (x), in
its turn, may be written as:
g′ Y µ g′
jl (x)Bµ (x) = − ψ lL (x)γ µ ψlL (x)Bµ (x) − g ′ ψ lR (x)γ µ ψlR (x)Bµ (x) =
2 2
g′
=− [ν L (x)γ µ νlL (x) + lL (x)γ µ lL (x) + 2lR (x)γ µ lR (x)]Bµ (x). (11.41)
2
Now we should unify the last term in (11.39) with (11.41). It is evident at a glance that
the choice of the basis in the form (11.32) provides the absence of the electromagnetic
interaction of the neutrino. Taking into consideration Eq. (11.32), we get without trouble
g′ Y µ g′g  
gJl3µ (x)Wµ3 (x) + jl (x)Bµ (x) = − p lL (x)γ µ lL (x) + lR (x)γ µ lR (x) Aµ (x)+
2 g 2 + g ′2

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 229
"
g 2 + g ′2 g 2 − g ′2
+ p ν lL (x)γ µ νlL (x) − p lL (x)γ µ lL (x)+
2 g 2 + g ′2 2 g 2 + g ′2
#
g ′2 µ
+p lR (x)γ lR (x) Zµ (x). (11.42)
g 2 + g ′2
Since Eq. (11.12) takes place, then the following relation:
1
(jµ )em = Jµ3 + jµY
2
µ
must be fulfilled, that is, the electromagnetic interaction −jem (x)Aµ (x) has to be in-
cluded into interactions (11.22) and (11.23). Then it becomes clear that the first term in
(11.42)
p describes the electromagnetic interaction of the charged leptons and the multiplier
gg ′ / g 2 + g ′2 is nothing more nor less than the electric charge
gg ′
|e| = p . (11.43)
g 2 + g ′2
Taking into consideration this circumstance and allowing for Eqs. (11.39), (11.42), we
obtain the final expression for the interaction Lagrangian of the gauge bosons with the
leptons
g  
LG = √ ν l (x)γ µ (1 − γ5 )l(x)Wµ∗ (x) + l(x)γ µ (1 − γ5 )νl (x)Wµ (x) − el(x)γ µ l(x)Aµ (x)+
2 2
g 
+ ν l (x)γ µ [1 − γ5 ]νl (x) + l(x)γ µ [4 sin2 θW − 1 + γ5 ]l(x) Zµ (x). (11.44)
4 cos θW
Since all the terms in the Lagrangian (11.44) represent the quantities of the type current×po-
tential, then the weak interaction caused by the exchanges of the W - and Z-bosons is
commonly called an interaction of the charged and neutral currents, respectively.
To develop the electroweak interactions theory for quarks is performed in full analogy
with the above-mentioned scheme for the leptons. There is the following correspondence
between the quarks and the leptons:
   α    α    α
νeL uL νµL cL ντ L tL
↔ , ↔ , ↔ , (11.45)
e−
L dα
L µ−L sαL τL− bα
L

νeR ↔ uα R, e−R ↔ dR ,
α
νµR ↔ cα R, (11.46)
µ− α
R ↔ sR , ντ R ↔ tα R, τR− ↔ bα R,
where α is a color quark index.∗ But what actually happens is that, instead of the d-, s-
and b-quarks, their linear combinations enter into singlets and doublets of the weak isospin.
We shall designate them by symbols d′ , s′ , and b′ . They are connected with the unprimed
quarks by the relations:
 ′  
d d
q ′d =  s′  = MCKM q d = MCKM  s  =
b′ b
  
c12 c13 s12 c13 s13 e−iδ13 d
=  −s12 c23 − c12 s23 s13 eiδ13 c12 c23 − s12 s23 s13 eiδ13 s23 c13   s  , (11.47)
s12 s23 − c12 c23 s13 eiδ13 −c12 s23 − s12 c23 s13 eiδ13 c23 c13 b

∗ As the electroweak interaction does not change the quark flavor, further the flavor index will be neglected.

© 2011 by Taylor and Francis Group, LLC

 230 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where MCKM is the Cabibbo–Kobayashi–Maskawa matrix (CKM), cij = cos θij CKM
, sij =
CKM CKM
sin θij , i, j are generation indices to equal 1, 2, 3, θij are mixing angles and a phase
multiplier e±iδ13 describes the CP parity violation. So in Eqs. (11.45) and (11.46) one
should make the replacement:

qid → qi′d = MCKM

ij qjd .

The CKM matrix is unitary. Its elements could be determined from the weak decays
of hadrons and from the experiments on the deep inelastic scattering of the neutrino by
hadrons. Since the matrix MCKM is the product of three noncommuting matrices of the ro-
tations in three planes [x1 x2 ], [x1 x3 ], and [x2 x3 ] of the abstract space, then the parametriza-
tion (11.47) is not unique.∗
In place of Eqs. (11.22) and (11.23) we have for the quarks

g′ Y µ YW
gJµq (x)·Wµ (x)+ jq (x)Bµ (x) = gψ qL (x)γ µ SW ·Wµ (x)ψqL (x)+g ′ ψ q (x)γ µ ψq (x)Bµ (x).
2 2
(11.48)
Passing in (11.48) from the gauge basis (W 1 , W 2 , W 3 , B) to the mass eigenstates one
(W, W ∗ , Z, γ) with the help of formulae (11.32), (11.34), we arrive at the following ex-
pression for the Lagrangian describing the interaction between quarks and gauge bosons:
g  
Lq = √ q uiL (x)γ µ MCKM ik
d
qkL Wµ∗ (x) + q dkL (x)γ µ MCKM∗ki
u
qiL Wµ (x) +
2
   
g 8 4
+ q ui (x)γ µ 1 − sin2 θW + γ5 qiu (x) − q di (x)γ µ 1 − sin2 θW +
4 cos θW 3 3
 e  
+γ5 ) qid (x) Zµ (x) + 2qui (x)γ µ qiu (x) − q di (x)γ µ qid (x) Aµ (x). (11.49)
3
We define the Yukawa Lagrangian for the quarks by analogy with the lepton case. Then,
after spontaneous symmetry breaking we get the Lagrangian determining the interaction of
quarks with the physical Higgs boson in the form:
X
L=− fqi q i (x)qi (x)H(x),
i

where the Yukawa constant fqi determines the mass of the qi -quark

fq v
mqi = √i .
2
The Fermi theory was a predecessor of the GWS theory. This theory most advantageously
managed with the description of the weak interaction in the low energy region. Then,
according to the correspondence principle, at low energies the GWS theory must reproduce
the results of the former theory. We take advantage this circumstance to define the linkage
between the parameters of the new and old theories. It may be beneficial to recall some
essential points of the Fermi theory.
By the early 1930s it was established that the beta-radioactivity of all nuclei is caused
by a pair of fundamental reactions where the proton and neutron are interconverted

n → p + e− + ν e , (11.50)

∗ We have used the parametrization accepted in Review of Particle Physics.

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 231

p → n + e+ + νe . (11.51)
The electron and the antineutrino or the positron and the neutrino, which appear as a
result of the beta-decay, are created because they do not exist in the radioactive nucleus.
This phenomenon is analogous to the process of the photon emitting by the electron when
it transits with the one orbit on the other orbit located close to the nucleus. E. Fermi
used this analogy and already known mathematical apparatus of the quantum theory of the
electromagnetic interaction in his weak interaction theory proposed in 1934. The interaction
Lagrangian by Fermi has the form of the product current×current
GF
LF = √ jµ† (x)j µ (x). (11.52)
2
The weak current entering into (11.52) is built out of the wavefunctions of the particles which
are pairwise unified: neutron—proton, electron—electron antineutrino and so on. So, for
the process (11.50) the one current is nucleon and it transfers the neutron to the proton. The
other current is lepton and it creates the pair, the electron and the electron antineutrino.
These currents belong to the class of the charged currents since they change the electric
charge of the particles to interact. In the both currents the charge is increased by |e|: the
positive charged proton arises from the neutral neutron while the electron antineutrino does
from the electron. The interaction (11.52) gave the name four-fermion contact interaction
∗
. The structure of the charged weak currents has been finally established in the middle
(19)50s. Analysis of experiments being done by that time led to the conclusion that the weak
current represents the difference of the vector V and the axial vector A (V − A structure).
Thus the lepton part of the current is defined by the expression:
jlσ (x) = e(x)γ σ (1 − γ5 )νe (x) + µ(x)γ σ (1 − γ5 )νµ (x) + τ (x)γ σ (1 − γ5 )ντ (x). (11.53)
After determination of the composite structure of nucleons the quarks occupy the nucleons
place in the nucleon part of the charged weak current
′ ′
jqσ (x) = d (x)γ σ (1 − γ5 )u(x) + s′ (x)γ σ (1 − γ5 )c(x) + b (x)γ σ (1 − γ5 )t(x). (11.54)
If one is constrained by the first order of the perturbation theory in the weak interaction
constant GF under calculations, then the Fermi theory gives a fine accordance with an
experiment. However the corrections of the high orders in GF represent the integrals which
become infinite at the large energies, that is, physically meaningless. Consequently the
Fermi theory must be suitably reconstructed. The refusal of interaction locality is the most
evident way. In other words, the analogy between the quantum electrodynamics and the
weak interaction theory should be deeper, namely, the weak interaction is also carried by
gauge bosons.
To connect the GWS theory parameters with the Fermi constant GF we consider the
muon decay through the channel
µ− → e− + ν e + νµ . (11.55)
The corresponding diagrams in the momentum representation for the Fermi theory and the
GWS theory are displayed in Figs.10.3 and 10.4. In the GWS theory the amplitude of the
decay (11.55) is given by the expression:
r  
g 2 mµ mνµ me mνe gσβ − qσ qβ /m2W
AW SG = ν µ (pνµ )γ σ (1 − γ5 )µ(pµ ) ×
8 Eµ Ee Eνe Eνµ m2W − q 2

∗ Sincefour fermion wavefunctions enter into the Lagrangian we call the interaction by four-fermion. As the
interaction takes place in one and the same point x we name it by contact.

© 2011 by Taylor and Francis Group, LLC

 232 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 11.3 FIGURE 11.4

The Feynman diagram for the muon The Feynman diagram for the
decay in the Fermi theory. muon decay in the GWS theory.

×e(pe )γ β (1 − γ5 )νe (pνe ), (11.56)

where q = pµ − pνµ and the expression standing in the parenthesis describes the propagation
of the virtual W − boson, that is, is the propagator of this particle. In the Fermi theory the
decay amplitude has the form:
r
GF mµ mνµ me mνe
AF = √ ν µ (pνµ )γ σ (1 − γ5 )µ(pµ )e(pe )γσ (1 − γ5 )νe (pνe ). (11.57)
2 Eµ Ee Eνe Eνµ

Under small q the expressions (11.56) and (11.57) must coincide. As in this case the W -
boson propagator is reduced to gσβ /m2W , the required linkage is given by:

g2 GF
= √ . (11.58)
8m2W 2

The constant g characterizes emitting and absorbing the W ± -bosons, much as e defines
emitting and absorbing the photons. From (11.43) follows that e > g and, therefore, the
weak interaction is in essence stronger than the electromagnetic one. However, as it was
shown, the weak process amplitudes are proportional to g 2 /m2W at low energies.∗ . So, owing
to that the W -bosons are very heavy, the weak interaction processes appear to be much
orders of magnitude weaker than the electromagnetic processes.
Not only does the GWS theory unify the electromagnetic and weak interactions, but it
also predicts the existence of new phenomena in the weak interaction physics, the neutral
currents. In 1973 the first reactions caused by the neutral currents were observed

νµ + p → νµ + p + π + + π − . (11.59)

Information confirming the neutral currents existence also follows from experiments on
observing the parity violation in atom physics. The interaction constant of the neutral
currents proves to be approximately the same as that of the charged currents.
The GWS theory predicts the linkage between the masses of the W - and Z-bosons as
well. From the relations (11.33) and (11.36) follows
p
mW g 2 + g ′2 mW
mZ = = . (11.60)
g cos θW

∗ Recall,
that the separation into the electromagnetic and weak interactions has a sense only at energies
< 100 GeV.

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 233

Using Eqs. (11.43) and (11.58), we obtain

r
παem
mW = √ sin−1 θW . (11.61)
GF 2
Having done three independent experiments one may define the constants GF , sin θW , αem ,
the knowledge of which determines not only the masses of the W - and Z-bosons but the
vacuum average v (vacuum expectation value) of the Higgs field as well
mW sin θW
v= √ .
παem
The Weinberg angle value may be found out of different experiments concerning nuclear
physics, physics of the weak interactions at low energies, and high energy physics. By 1983
the results of sin2 θW determination have come to be matched. The averaged value was
given by:
sin2 θW ≈ 0.23. (11.62)
Then substituting the values
1
αem ≈ , GF = 1.17 × 10−5 GeV−2 ,
137
into Eqs. (11.60), (11.61) and using the value of sin2 θW , we find

mW ≈ 80 GeV, mZ ≈ 91 GeV. (11.63)

It is evident that the discovery of the W - and Z-bosons would be the deciding step on the
way of the GWS theory checkout. For these purposes the proton-antiproton collider was
built in CERN. It began operation in summer 1981. The direct production of the W -boson
with the subsequent decay through the electron and electron antineutrino

u + d → W −∗ → e− + ν e (11.64)

is displayed in Fig. 11.5. The cross section of the reaction (11.64) is a function of the

FIGURE 11.5
The Feynman diagram for the process u + d → W −∗ → e− + ν e .

colliding quarks energy and as soon as the energy in the center-of-mass frame is approaching
to mW , the W -boson is exhibiting as a resonance. In the resonance region the cross section
has the sharp maximum of the height and the width of which are predicted by the GWS
theory. At the resonance the cross section value can be calculated by application of the
Breit-Wigner formula (8.22).

© 2011 by Taylor and Francis Group, LLC

 234 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It is beyond doubt that to directly observe the quark collisions is impossible, since the
quarks in the free states are unavailable for us. The proton-antiproton collisions are just
the most changing. A monochromatic proton beam may be considered as the quark beam
with the wide distribution over momenta, that is, when P is a proton momentum then
xi P is a momentum of an ith-quark (0 ≤ xi ≤ 1). An antiproton beam looks analogous.
In the process of the W production the quark picks for itself the antiquark with suitable
momentum. In the center-of-mass frame the energy of the impacts of the quark with the
antiquark sqq is connected with spp by the relation:

sqq = spp xq xq .

The distribution functions of the quarks both in the proton and in the antiproton are such
that in order to provide the right correlation between the proton quark and the antiproton
antiquark the following condition:

xq ≈ xq ≥ 0.25 (11.65)

must be fulfilled. Thus, there is one wide region of optimal energies for the pp impacts at
the given W -boson mass. For mW = 80 GeV it is given by:
√
400 ≤ spp ≤ 600 GeV.

In actual fact the following processes are investigated at the CERN pp-collider:

p + p → W ± + X, (11.66)

where X is arbitrary hadrons plurality. To detect the W ± -bosons was carried out through
the lepton decays
W + → e+ + νe , W − → e− + ν e . (11.67)
Such processes are represented by diagrams that include the elements both of the quarks
diagrams and of the Feynman ones. So the diagram pictured in Fig. 11.6 corresponds to
the process
p + p → W −∗ → e− + ν e + X. (11.68)
To obtain the cross section of the reaction (11.68) one should integrate the cross section
of the reaction (11.64) at the resonance over the distributions of the quarks in the proton
and the antiproton.
Since the W -boson mass is large, then charged leptons l± , appearing under the W boson
decay, have a large transverse momentum. Thus, the detection of l± has not caused any
troubles. The neutrino recording was based on purely kinematic reasons. For this purpose
the special detector was used. Its sensitivity relative to all charged or neutral particles
that were created in the impact process, was uniform in the whole volume over the whole
solid angle. Since interactions are observed in the center-of-mass frame any appreciable
disbalance of a momentum signals the presence of one or more numbers of noninteracting
particles (presumably, neutrinos). Calorimeters are ideally suited for the role of such detec-
tors because their energy registration efficiency may be done to be sufficiently homogeneous
for different particles to hit in them. Notice, in the reference frame where the W -boson
rests the momentum carried away by the neutrino pν is equal to mW /2, that is, it is very
large.
In January 1983 two independent Collaborations UA1 and UA2 working at the CERN
pp-collider presented the first results on detecting the W -bosons in the reactions (11.66),
(11.67).

© 2011 by Taylor and Francis Group, LLC

 Passing glance on the theory of electroweak interaction 235

FIGURE 11.6
The Feynman diagram for the process p + p → W −∗ → e− + ν e + X.

In June 1983 the group UA1 had reported the observation of the first five cases of the
creation and the decay of the Z-bosons. In August of the same year the group UA2 had
detected the eight analogous events. The Z-bosons are created in the reaction

p+p→Z +X (11.69)

and are detected through the decays

Z → e− + e+ , Z → e− + e+ + γ, Z → µ− + µ+ . (11.70)

These experiments have led not only to the discovery of the W - and Z-bosons, they have
also shown that their properties are exactly described by the GWS theory. It should be
noticed that the mass values (11.63) are approximate rather than precise. To obtain the
precise values of the gauge boson masses we must take into consideration the interaction
of particles with the vacuum or, what is the same, incorporate the higher orders of the
perturbation theory (radiative corrections) under calculating the cross sections. To influence
the radiative corrections (RC) on the values of αem and sin2 θW is especially significant.
The calculations showed that the inclusion of the RC changes the gauge bosons mass values
in the formulae (11.63) on 5% approximately.

© 2011 by Taylor and Francis Group, LLC

 12
Fundamental particles of the Standard Model

Once more the blackbirds chanted “The Fruits of Science Feed the
Young.” But all realized that the “Golden Age” was drawing to a close.
The sole prospect now was the descending darkness of ignorance, accom-
panied by its inevitable companions—dissension and strife.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

The evolution of our notions about the matter structure, the four steps on the Quantum
Stairway, may be schematically represented by Fig. 12.1. Every stage of this Stairway

E
cm

cm
d
-16

-15
d
10

d u
u
10

pn
cm

n n pn
p p n
-13

n nnp
pn p
10
10 cm
-8

FIGURE 12.1
The Quantum Stairway.

237
© 2011 by Taylor and Francis Group, LLC
 238 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

presents a separate region of physics and, consequently, the phenomena range described by
it is characterized first and foremost by particle velocities and a region size that is available
for particle motions. An object having an intrinsic structure from the viewpoint of the
higher stage is considered as a structureless one in all phenomena at any underlying stage.
So an atom is supposed to be a point particle in Classical Physics, a nucleus—in Atomic
Physics, a nucleon—in Nuclear Physics.
Before when we talked about the quark-lepton symmetry we were pointing the way of
its storing, placing the quarks and the leptons into the weak isospin doublets. In this case
the down quarks proved to be subjected to mixing. The absence of similar mixing in the
lepton sector would shake not only our belief in the quark-lepton symmetry, but it would
also make us refuse the belief in the unity of nature. The analogous phenomenon proves to
has taken place in the lepton sector as well. But we have learned about it much later, in
2001. The experiments showed that the electron, muon, and tau-lepton neutrinos are not
the states with definite mass value but they represent the mixtures of the physical states
ν1 , ν2 , and ν3 . The corresponding mixing matrix in the neutrino sector MN M has the same
form as the CKM matrix, that is,

MN M = MCKM (θij
CKM NM
→ θij ), (12.1)
NM
where θij are neutrinos mixing angles. Moreover, the mixing angles in the lepton sector
are connected with those in the quark sector by the relations:

NM CKM π CKM NM π NM CKM

θ12 + θ12 = , θ23 + θ23 = , θ13 ∼ θ13 .
4 4
Let us look aside from the existence of the right-handed weak isospin singlets for the
time being and represent the quarks and the leptons by means of three generations of the
SU (2)L -doublets    α
νeL uL
, (12.2)
e−
L d′L
   α
νµL cL
, , (12.3)
µ−L s′L
   α
ντ L tL
, . (12.4)
τL− b′L
The processes of annihilating the electrons-positrons into hadrons have set the upper bound
on the quarks size. By now this bound is ≤ 10−16 cm. Such a limitation exists for leptons as
well. Within the accuracy of the modern experiment the particles entering into both groups
are considered as point particles, that is, at present we have all the reasons to believe that
these particles are fundamental.
It is convenient to present the fundamental particles of the SM in the three-dimensional
coordinate system M, in which the operator eigenvalues of the weak isospin projection on
the third axis S3W are plotted along the z axis while the plane xy is used for the definition
of the eigenvalues of the color spin operator S (col) . Draw a regular-shaped triangular prism
in M (Fig. 12.2). We take into account the quark-lepton symmetry and start the process
of placing the particles with the first fermion generation. The upper (lower) components of
the weak isospin doublets will be put in the upper (lower) base of the prism. Three color
states of the u-quark are placed in the vertices of the upper base (the points corresponding
to S3W = 1/2) and three color states of the d′ -quark are placed in the vertices of the low
base (S3W = −1/2). The colorless leptons will be housed in the straight line x = y = 0:
the electron neutrinos in the center of the upper base while the electrons in the center

© 2011 by Taylor and Francis Group, LLC

 Fundamental particles of the Standard Model 239

FIGURE 12.2
The fundamental particles of the SM.

of the lower base. Two analogous prisms will correspond to the fundamental fermions of
the second and third generations. The quarks and the leptons form the first group of
the fundamental particles, the matter particles group. The next group, the interaction
carriers group, constitute four gauge bosons of the electroweak interaction W − , W + , Z, γ,
eight gluons of the QCD gi and yet-undiscovered carrier of the gravitational interaction,
the graviton G. We call this group the gauge bosons group. Let us be distracted from the
graviton existence and display the remaining interaction carriers in the form of the arrows,
meaning the result of interactions between the bosons and the fermions, at the same figure.
Emitting or absorbing the charged W ± -bosons leads to the transitions between the vertices
of the fermion triangles with ∆S3W = ±1, that is, it produces the motion along the z axis.
From eight existing gluons two

1
1

g7 = √ RR − GG , g8 = √ RR + GG − 2BB
2 6

© 2011 by Taylor and Francis Group, LLC

 240 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

are responsible for the processes without a change of the quarks color whereas six

g1 = GB, g2 = RB, g3 = RG, g4 = BG, g5 = BR, g6 = GR

—for the processes changing the quarks color. The clockwise motions and counterclockwise
motions along the perimeters of the equilateral fermion triangles correspond to the gluons
to change the color.
We are coming now to displaying the gauge bosons, which do not effect the change of
the electric charge, of the weak isospin, of the color and, consequently, do not alter the
fermion position in the M space. Such particles are the photon, the neutral gauge boson,
and two remaining gluons. Apart from the above-mentioned fundamental particles, there is
one more particle that has the common features with both groups but still stands aside from
them, that is, it actually forms the third group of the fundamental particles. This is the
Higgs boson having the zero-spin and electric charge to be equal to zero. It does not enter
into the matter composition and its interaction with all the fundamental fermions is not
attached to the definite class. However, according to the SM all the fundamental particles
acquired their masses thanks to the Higgs boson. It may be an echo of the spontaneous
symmetry breaking, which happened (if any) in the epoch of the early Universe.∗ Under
emitting and absorbing the Higgs boson the fermion does not alter its position in the M
space as well. Let us agree the bosons, whose interactions with the matter particles result
in
∆Q = ∆S3W = ∆S (col) = 0,
will be displayed by the arrows that begin and end on the fermions.
The first generation plays a particular role. In effect, all we see around us in nature
consists of the first generation fermions. All the members of the second and third generations
are unstable, the exception is probably provided by the neutrinos. They appear only in
accelerators and in phenomena produced by the cosmic rays. The fermions of the second
and third generations played the important role in the early universe, in the first instant of
the Big Bang. In particular, the neutrino flavors number has defined the quantitative ratio
between hydrogen and helium in the universe. The second and third generations made an
impact on the mass values of the first generation particles.
We pay attention to the trick nature has used choosing the mass values of the first
generation particles. Let us consider the simplest but the most important atom of our
universe, the hydrogen atom. The stability of the atom is caused by the fact that the
reaction
p + e− → n + νe (12.5)
is energetically forbidden because the masses of the electron and the proton, which constitute
it, satisfy an inequality
me < ∆m, (12.6)
where ∆m = mn −mp ≈ 1.3 MeV (we have neglected the neutrino mass). It is clear that the
fragile equilibrium expressed by Eq. (12.6) may be violated even by an insignificant change
of mu , md , and me . Consequences of the hydrogen instability would be catastrophic. If the
hydrogen that represents the main fuel for the stars of universe would be absent, then the

∗ InNovember 2000 the reports about the Higgs boson observation with the mass equal to 115 GeV came
from two groups L3 and ALEPH (LEP) that investigated the reaction
e+ + e− → Z + H.
However, two other Collaborations working at LEP, OPAL, and DELPHI did not confirm the results of
their colleagues. In 2001 LEP ceased the operation, leaving the Higgs boson history unwritten.

© 2011 by Taylor and Francis Group, LLC

 Fundamental particles of the Standard Model 241

ordinary stars did not exist and the universe took a totally another form. Then in order
to make our world stable, say in store, why not increase the value of ∆m in the relation
(12.6). However, the other problem connected with the deuterium is waiting for us in this
way. Its nucleus, the deuteron, possesses the most small binding energy Ec ≈ 2.24 MeV. A
guarantor of the deuteron stability is the fact that in it the decay of the neutron through
the channel
n → p + e− + ν e (12.7)
is energetically unprofitable. In this case, the energy conservation law demands

mp + mn − Ec = mp + mp + me + T, (12.8)

where T is a kinetic energy of the decaying particles. From the positivity of T it follows
that the decay is forbidden under condition

Ec + me > ∆m. (12.9)

Thus, if we made ∆m too large and violated (12.9), then the deuterium would be unstable
that led to its complete lack in nature. However, the deuterium production is the first step
in the chain of nuclear transformations tracing from hydrogen to more heavy elements that
were not in the early universe. In the case of deuterium lack, the routine way of producing
the elements that are heavier than the hydrogen would become impossible and the universe
would acquire an entirely different appearance.
The inequality (12.9) demands ∆m must be sufficiently small. However, ∆m is nothing
else than masses difference of particles entering the same isomultiplet. If ∆m is compared
with masses difference of particles belonging to other isomultiplets, it proves to be minimal.
For example,

mπ± − mπ0 = 4.6 MeV, and mΞ− − mΞ0 = 6.48 MeV.

A smallness of ∆m is supplied with the fact that the masses of the u and d quarks are
closely related.
The small ratio of the electron and the proton masses is the cause of such an important
phenomenon as the exact localization of the nucleus in the electrons cloud that, in its turn,
defines a molecules architecture. Otherwise the stable configurations, characterizing the
life itself and much in the world surrounding us, would not exist. All this looks as though
nature was finely tuning the masses of the first generation fermions.
Nature often resorts to a similar fine-tuning of physics laws. Here the example with such
composite systems as stars will be indicative. Their existence is caused by all four inter-
actions. The gravitational interaction tends to compress a star. As the star consistence
increases, the temperature in the star center grows until a thermonuclear reaction, burning
hydrogen into helium, is begun. The strong and weak interactions govern this reaction. The
appearance of new energy source slows down the compression process since radiation exerts
pressure on the star’s outward layers. At that, the main effect is caused by electromagnetic
radiation. Finally, the release rate of the thermonuclear energy rises to an extent that the
electromagnetic radiation balances the gravitational forces action in all volume of a star’s
matter. At these conditions the system structure critically depends on the interactions mag-
nitude and thereby on numerical values of fundamental constants. Equilibrium between the
gravitational and electromagnetic interactions within stars proves to be kept with striking
accuracy. Calculations fulfilled by B. Carter [49] show that changing any interaction only
on 10−40 of its value would result in catastrophe for stars of the Sun’s type.
Let us return to Fig. 12.2. Generally speaking, the prism presented in it displays only the
weak isospin left-handed doublets. To represent the right-handed fermion singlets we need

© 2011 by Taylor and Francis Group, LLC

 242 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

one more figure in the M space. Having combined the upper and lower bases of the prism
in the point S W = 0 we obtain the sought figure. In so doing the u and d quarks of the
identical color fall into the same vertex of the obtained triangle. Now transitions between
the right-handed quarks of different (transitions associated with the W ± -bosons) flavors are
absent. As for the right-handed neutrino, its position on the figure depends on a way that
is used to impart the mass to this particle. In the most simple case, when a right-handed
neutrino SU (2)L -singlet is introduced into the theory, the right-handed neutrino falls into
the same point as for the right-handed electron. Inasmuch as the symmetry between the
right-handed singlets plane and the prism is absent, the presented picture loses some portion
of its attraction.
To rectify the situation one should remove the asymmetry between the left and the
right, that is, apply to models possessing the mirror symmetry (P -invariance) before the
spontaneous breaking symmetry. The SM gauge group extension on the factor SU (2) is the
primary choice. The model with the gauge group SU (2)L ×SU (2)R ×U (1)B−L was proposed
in [50] and later was generalized for the case of arbitrary orientation of the SU (2)R -generator
in the group space [51]. In this model all the fundamental fermions enter into the theory in
the symmetric way, that is, they constitute the left-handed and right-handed doublets
     α  α
νeL NeR uL uR
, , , , (12.10)
e−
L e−
R d′L d′R
     α  α
νµL NµR cL cR
, , , , (12.11)
µ−L µ−
R s′L s′R
     α  α
ντ L Nτ R tL tR
− , − , , (12.12)
τL τR b′L b′R
Apart from three additional right-handed neutrinos two additional charged WR± -bosons
and one neutral Z ′ -boson are present in the theory. Then, in the M-space two rectilinear
triangular prisms correspond to every fermion generation. The prism associated with the
left-handed doublets is analogous to that displayed in Fig. 12.2. In the prism associated
with the right-handed doublets the motion along z-axis is generated by the WR± -bosons
while transitions not to affect the fermions position are caused by the Z ′ -boson. In the
simplest version of this model [52], after a spontaneous symmetry violation we are left with
8 physical Higgs bosons : two singly charged scalars (a motion along prism edges correlates
with them) and six neutral scalars (arrows to begin and finish on fermions may again be
used for these particles).
At a sight in Fig. 12.2 it is difficult to get rid of the temptation to declare the quarks
and the leptons, which belong to the base triangles, by the quark-lepton octet of some
symmetry group. The fermions of the second and third generations would also constitute
the analogous octets. Having stood on this point of view, we reduce all the plurality of the
fundamental matter particles to three superparticles, each of them could be in eight states.
However, such a simple and elegant scheme has its own exotics. The term a “multiplet”
has a sense, if and only if, the transitions between all its states are allowed. But in our
case only the quark↔quark and lepton↔lepton transitions exist whereas the lepton↔quark
transitions are forbidden. This circumstance is a consequence of a lack of interconnections
between the electroweak and QCD fields within the SM. But if we choose the symmetry
group, which allow us to place all the fermions of each generation into a fundamental
octet, and subsequently gauge this symmetry and spontaneously break it, then we obtain
frighteningly huge numbers, both of the Higgs bosons and of the gauge bosons. It is clear
that such a scheme of the universe hardly has the right to life, since, as Aristotle said:
“Nature always realizes the best of possibilities.”

© 2011 by Taylor and Francis Group, LLC

 Fundamental particles of the Standard Model 243

The SU (5)-group is a minimum group of the GUT that includes the SM group as a
subgroup. In this model one cannot manage to place all the known fundamental fermions
of each generation into one representation. However, it could be done with the help of two
representations, the quintet and decuplet representations of the SU (5)-group. The quintet
for the first generation has the form:
′ ′ ′
(dR , dG , dB , e− , νe ),

while the corresponding decuplet is represented by an antisymmetric matrix

−d′R
 
0 uB −uG −uR
′
 −uB 0 uR −uG −dG 
1 
−d′B  .

√  uG −uR 0 −uB
2 u uG uB 0 −e +
R
d′R d′G d′B e+ 0

In so doing all the fermion fields are considered by the left-hand chiral fields, that is, the
wavefunctions of all the fields are multiplied by the quantity (1 − γ5 )/2. Evidently, if one
displays the fermion multiplets of the SU (5)-group in the coordinate system M, then the
obtained picture will not yet hold the same aesthetic appeal as it was in the case of Fig.
12.2. As the SU (5)-group has 24 generators, then the corresponding gauge transformation
is achieved by 24 gauge bosons. Twelve of them are the gauge bosons of the SM. The
remaining bosons, the Xi± and Yi± bosons (i = 1, 2, 3), have the masses ∼ MGU and the
charges ±4e/3, ±e/3. There is also a great deal of physical Higgs bosons. For example,
this number equals 16 in the version proposed by Georgi and Glashow.
Under increasing the GUT group dimensionality, the number of both of the physical Higgs
bosons and of the gauge bosons grow. So, in the SO(10)-group the number of the gauge
bosons reaches 45.
In these examples we see one of possibilities—the achieved stage on the Quantum Stairway
is the last stage but the list of the fundamental particles may be wider. Increasing the list
may occur both through the physical Higgs bosons and through the gauge bosons. There are
also no reasons to believe that the number of fermion generations may not be greater than
three. The other possibility remains to be open, namely, that all the fundamental particles,
or some of them at least, for example, the quarks and the leptons actually represent an
objects consisting of subparticles, preons. Then some new forces responsible for unifying the
preons into the quarks and the leptons must exist. In this case the strong and electroweak
forces appear to be no more fundamental than the chemical or nuclear forces and the
attempt of building the GUT from the elementary quarks and leptons will be doomed to
failure. Are the preons the last undivided matter blocks or is climbing to the next stage
of the Quantum Stairway waiting for us? It is clear only that as early as the first century
bfB.C. Lucretius said: “But without doubt, the well-known limit of breaking in pieces has
been set.”

© 2011 by Taylor and Francis Group, LLC

 13
Technical equipment of particle physics

In less than a month not a trace remained of the recent

“Golden Age.” To vindicate themselves the falcon and
the vulture joined hands temporarily and put all the
blame on education. Science, they said, was, of course,
beneficial, but only if applied at the proper time. After
all, our grandsires did without it, and so could we. And
to prove that all the harm really lay in education they
began to discover conspiracies, invariably such as
involved at least a prayer-book or an elementary school
reader. An interminable nightmare of searches, investiga-
tions and trials followed.
M.E. Saltykov Shchedrin, “The Eagle-Patron of Arts”

13.1 Accelerators
From the time of the atomic nucleus discovery by Rutherford it has been clear that a study
of the nuclei structure requires the production of accelerated beams. Natural sources of
accelerated particles, radioactive elements, exhibit very low intensities, restricted energies,
and are absolutely uncontrollable. To produce high-energy particles, the developments of
special accelerating facilities were being started. Presently, giant accelerators symbolize
modern elementary particle physics. Experiments involving elementary particles are impos-
sible without the use of accelerators, and the progress of physics is inconceivable without
such experiments. During the twentieth century typical accelerator energies were varying
from a few eV to several TeV. In other words, the attainable energies were growing ex-
ponentially, being doubled every 2.5 years. It is obvious that such a rapid energy growth
cannot be expected in the twenty-first century.
Modern accelerator facilities include three basic blocks:

(i) an accelerator that effects the kinetic energy increase, formation, and ejection of high-
intensity particle beams;
(ii) a detector that comprises a system for registration of the interaction processes and
analysis of the reaction products; besides, a particular section of the detector may be used
as a target;
(iii) equipment for input/output, storage and processing of the experimental information;
a unit for automatic control of the whole accelerating system.

According to the acceleration principle, facilities may be classed as synchrotron and linear
accelerators; depending on the collision method, they may be subdivided into the stationary-

245
© 2011 by Taylor and Francis Group, LLC
 246 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

target and colliding-beam machines. Both stable, for example, electron, proton, neutrino,
and unstable particles, for example, muons, may be accelerated in the process.
The first accelerator designed by P. Van de Graaff in 1931 was of the high-voltage linear
type. It represented a combination of the high-voltage source (generator) and acceleration
vacuum tube, where the charged particles, moving between the generator poles, acquired
an energy corresponding to the voltage at the poles. High-voltage accelerators are intended
to accelerate light (electrons) as well as heavy particles (protons, ions). Their merits are as
follows: continuous operation, high energy stability of the accelerated particles, and small
energy spread in the beam (∆E ≤ 0.01%). High-voltage accelerators with energies up to
10–20 MeV are still in current use for preliminary acceleration in large accelerators.
As the attainment of high potential difference (for example, 106 V) presents technical
difficulties, it is expedient to cause the accelerated particles to rotate in the magnetic field,
progressively accelerating them by low-voltage pulses applied to the accelerating electrodes
at a certain instance of time and at a frequency equal to the particle rotation frequency. In
this way acceleration is imparted only to the particles entrapped into the accelerating gap
at proper times. Because of this, only parts of the beam are accelerated (particle clusters)
rather than the whole beam. This principle was used for the creation of the first cyclic
accelerator (cyclotron) constructed by E. Lawrence in 1931. In a cyclotron the charged
particles are accelerated from zero to maximum energy under the effect of an alternating
electric field with the constant period Ta . Curving of the orbits is provided by a constant
magnetic field directed perpendicular to the orbit plane. The rotation period of a particle
is determined by the expression:
2πmc 2πE
Tf = p = , (13.1)
2
eB 1 − v /c 2 ecB

where B is a magnetic field induction, v is a particle motion velocity on an orbit, and E

is a total particle energy. It should be noted that Eq. (13.1) is written in the CGS. In
this section, for the sake of obviousness, it is convenient to use an ordinary system of units.
For nonrelativistic velocities E ≈ mc2 the period is constant. Provided in this case T f is a
multiple of Ta , a prolonged resonance may be observed between the particle rotation in the
magnetic field and variations in the accelerating voltage. Accelerated particles are moving
along the spiral orbits with an ever-growing radius
mcv
R= . (13.2)
eB
And acceleration takes place until the particle motion is in resonance with the accelerating
field. When relativistic velocities are attained by the accelerated particles, the total particle
energy begins to grow, causing the resonance disturbance, and hence acceleration of the
particles is terminated.
Independently of one another, V. I. Vecsler (USSR, 1944) and E. McMillan (USA, 1945)
have proposed the phase stability principle, which ensures the resonance condition at any
relativistic velocity. By Eq. (13.1), the relationship between the total particle energy and
accelerating field frequency ωa should be as follows:
ecBq
E= , (13.3)
ωa
where q = 1, 2, 3 . . .. According to the stable-phase mechanism, the particle energy au-
tomatically takes the value close to the resonance one, with a relatively slow variation in
time of the accelerating electric field frequency and magnetic field induction. The finding
of the stable phase principle has resulted in the advent of new types of accelerators. As

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 247

follows from (13.3), an increase in the equilibrium (resonance-associated) energy of a par-

ticle requires a decrease of the accelerating field frequency (synchrocyclotron) or increased
induction of the twisting magnetic field (synchrotron), or else variation of both the indi-
cated frequency and induction (synchrophasotron) or, finally, an increase in the acceleration
multiplicity, that is, in the value q (microtron).
The stable phase ensures stability of the particle motion in the azimuthal direction (in
the direction of a particle orbit). The transverse (orbit-perpendicular) motion stability, or
focusing, is realized by an adequate selection of the radius-variation of the magnetic field.
To achieve stability in both transverse directions, it is necessary to attain a minor radial
decrease of the magnetic field.
Synchrotrons are used for the formation of a beam of relativistic electrons. Synchropha-
sotrons serve as accelerators of heavy particles (protons, ions). In accelerators of both types
the radius of the equilibrium orbit is constant. This circumstance allows one to make the
magnetic system in the form of the ring. The principal limitations of ring accelerators are
as follows: (i) accelerated particles in the twisting magnetic field sustain the energy loss due
to magneto-bremsstrahlung (synchrotron radiation) whose power is given by the formula:
Cγ cE 4
P= , (13.4)
R2
where
2e2
Cγ = ;
3(mc2 )4
and (ii) bulky magnetic system.
To increase the energy of an accelerated particle considering these limitations, it is re-
quired to enlarge the ring radius of the accelerator. Indeed, with growing R losses by
radiation drop, the magnetic field value necessary for retention of the particle on the orbit
is decreased. So, proton synchrophasotron with a maximum energy of 500 GeV, constructed
in the FERMILAB in 1972, was 2 km in diameter.
Proton and ion linear accelerators are based on the same principle as cyclic: a particle
moving in the resonance falls within the accelerating voltage phase in every gap. Never-
theless, the particle motion proceeds along a straight line, and the gaps along this line are
arranged at certain intervals in order that the particle transit time from gap to gap be equal
to the period of the accelerating electric field Ta or be a multiple of this period. Here the
phase stability is also necessary for matching of the transit time between the gaps with Ta
and for focusing in the transverse direction.
Linear electron accelerators are considerably different from the proton ones. Considering
that the velocity of relativistic electrons is practically constant
pc2 pc2
v= =p ≈ c,
E p2 c2 + m2e c4
synchronism is ensured because the accelerating electromagnetic wave propagates at a ve-
locity of light thus excluding the necessity for the phase-stability mechanism. As
dp⊥ mv⊥
= 0, then p = const,
dt 1 − v 2 /c2
and the transverse motion velocities v⊥ are rapidly falling with an increase in v. Conse-
quently, there is no need in focusing too. The transverse Coulomb repulsion of electrons
in the beam is insignificant due to a nearly absolute compensation, owing to the magnetic
attraction of the currents. The energy losses by synchrotron radiation in a linear acceler-
ator are also insignificant. To transit to high energies, however, calls for increasing of an

© 2011 by Taylor and Francis Group, LLC

 248 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

accelerator’s length. For example, the linear electron accelerator at the Stanford Linear
Accelerator Center (SLAC) constructed in 1966 and having a maximum energy of 25 GeV
was 3 km in length.
For the most part, the primary electron source in accelerators is represented by the so-
called electron gun including a thermionic cathode and electron-optical system. A source
of protons and weakly-ionized heavy ions is plasma, from where they are pulled by the
external electric field. Positrons, antiprotons, and greatly-charged ions are generated due
to interactions between the primary electron, proton, or ion beam and matter. Vacuum
within the volume, where the particle motion takes place, in all accelerators is of the order
of 10−5 –10−7 mm Hg to lower particle scattering from the residual gas.
The concept of colliding-beam accelerators put forward by D. Kernst in 1956 has led
to revolution in the technology. Its realization enables one to attain a critical energy in-
crease for the colliding particles and hence to investigate the matter structure at still closer
distances.
With a stationary target the kinetic energy of an incoming particle (shell) is only partly
transferred to the reaction energy; some part of the kinetic energy is spent for the target
recoil energy. In case the target is more massive than a shell the recoil energy is low,
otherwise the collision efficiency decreases drastically. Because of a relativistic increase of
mass, the energy loss by recoil is growing with the particle velocity approaching the velocity
of light. A character of interaction is determined by the particle energy in the center-of-
mass frame rather than in the lab frame as the major part of the energy in the lab frame
is transformed into the kinetic energy of the reaction products.
In the present-day accelerators the colliding beams are not strictly opposite, intersecting
at a small angle. During processing the results of such experiments all kinematic character-
istics are transformed to the center-of-mass frame for subsequent analysis. On collision of
two particles with arbitrary momenta pa and pb the transition to the center-of-mass frame
is realized by the Lorentz transformation at the appropriate velocity
(pa + pb )c2
v= . (13.5)
Ea + Eb
Actually, directing the axis x towards the sum of momenta pa + pb = p and using the
Lorentz transformation for the four-dimensional momentum, in a new coordinate system
(marked with asterisk) we get:
px − (Ea + Eb )v/c2
p∗x = p , p∗y = py , p∗z = pz . (13.6)
1 − v 2 /c2
Setting the vector p∗ to zero, we can find the center-of-mass frame velocity relative to the
reference frame, where
pc2
p 6= 0, v= ,
Ea + Eb
to coincide with Eq. (13.5). The center-of-mass frame velocity with respect to the lab frame
(b particle rests) may be determined by the expression:
pa c2
v= . (13.7)
Ea + mb c2
With the use of (13.7) it is possible to relate the particle energies and momenta for both
systems. For instance, in case of the particle a we have:
pa − Ea v/c2 mb pa
p∗a = p =p , (13.8)
2
1 − v /c 2 ma + m2b + 2Ea mb /c2
2

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 249

m2a c2 + mb Ea
Ea∗ = p . (13.9)
m2a + m2b + 2Ea mb /c2
The relation (13.9) enables one to gain some insight into the energy advantage of accelerators
of the collider type. For simplicity, we consider the collision of identical particles. Passing
to kinetic energy in the formula (13.9) we find:

2(T ∗ )2
T = 4T ∗ + . (13.10)
m b c2

From (13.10) it follows that to produce the energy T ∗ = 200 GeV one can realize the
variant of a stationary-target electron accelerator at an energy of T ≈ 1.6 × 108 GeV. As
regards the energy, the merits of colliders are obvious. At the same time, one should never
forget that cheese may be free-of-charge in a mousetrap only. The principal drawback of
colliders is the low frequency of the reactions. This peculiarity is easily comprehended
when we compare shooting at a large stationary target and shooting at the bullets that fly
towards each other. The efficiency of colliders may be improved using a higher particles
density of the beam. This is attained by the use of storage rings, where the accelerated
particles are stored throughout many acceleration cycles. Moreover, the focusing system
may be constructed so that maximum beam compression can be provided at the point of
collision, contributing to the enhanced beam density and higher probability of the reaction.
We introduce the quantity known as accelerator luminosity that defines the number of
events in a unit time at the unit interaction cross section (1 cm2 ). For stationary-target
accelerators the luminosity L is equal to

L = n0 lN, (13.11)

where n0 is particles density in the target, l is a target thickness along the beam, and N is
particles flux outgoing from the accelerator. In the case of colliding beams the accelerator
luminosity is given by the expression:

N1 N2 lf
L= . (13.12)
lc S
Here N1 and N2 is the total particles number in the beams, l—cluster extent, lc —collision
length (lc > l), f —rotation frequency of the particles in the accelerator in terms of s−1 ,
S—cross section area of a larger cluster measured in cm2 . Luminosity dimensionality is
cm−2 · s−1 . The luminosity multiplied by the process cross section σ in cm2 gives the
relevant number of events per second

R = σL. (13.13)

Modern high-energy accelerators are colliders. Also, these accelerators may be operated
in the stationary-target mode. All of them represent synchrotrons, with the exception of
the linear electron-positron collider at SLC (Stanford). This collider obtains energy up to
100 GeV at the luminosity L = 2.5 × 1030 cm−2 s−1 . In linear accelerators there is a single
region for interaction of colliding particles, and therefore its cross section area S should
be exceptionally small with a diameter of the order of ∼ 10−4 cm. At SLC the beam of
accelerated electrons (positrons) is divided into hyperdense clusters (4 × 1010 particles) 0.1
cm in length, with a cross section width of 1.5 × 10−4 cm and height of 0.5 × 10−4 cm. The
SLC is about 1.5 km in length.
The operation of the most powerful cyclic electron-positron collider LEP located at
Geneva was terminated in 2001. Its maximum energies were in excess of 200 GeV. To

© 2011 by Taylor and Francis Group, LLC

 250 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

decrease synchrotron radiation, the perimeter of LEP was increased to 26.66 km. Its lu-
minosity amounted to L = 5 × 1031 cm−2 s−1 . In this collider the clusters of accelerated
electrons and positrons were denser by the order of magnitude (6 × 1011 particles), whereas
their space dimensions were much greater: extent of 1 cm, height—8×10−4 cm and width—
2 × 10−2 cm. The rotation period was 22 × 10−6 s. During acceleration time of 550 s the
particles covered a distance of the ring approximately 2.5×107 times, however, keeping their
orbits to an accuracy of 2 × 10−3 cm. The energy of LEP is a limit for circular electron-
positron colliders. Further energy enhancement for e− e+ -machines is possible only in case
of linear facilities. Such colliders are already at the stage of conception [53]; their putting
into service is expected in 10–12 years. The typical energy for such a linear accelerator (so-
called linac) will reach 500–1000 GeV. The energy of linear colliders cannot be increased
without limit. This is associated with the material structure of the accelerator. In order
for a particle to be accelerated to energies of the order of 1000 TeV or higher on the typical
distances of order of 100 km, it is required to create an accelerating gradient of the electric
field in the region of 108 V/cm. Unfortunately, such strong fields will break away electrons
from atoms, altering the structure of any material. An effort to realize acceleration using
a field of this strength will result in the destruction of the accelerator. It is hoped that the
exit from this deadlock may be found with the help of nanotechnologies. There is reason to
believe that advances in nanotechnology will enable the creation of microscopic accelerator
cells with the requisite accelerating gradient. In this case, these cells will have such property
that after their disruption they could be regenerated in a short time.
It is clear that for the effective acceleration the gain in energy of a particle per cycle
should be higher than the total radiation loss P determined by the formula (13.4). Since
P ∼ m−4 , in case of protons the attainable acceleration energies will be much higher.
Presently, the highest energy (2 TeV) is provided by the FERMILAB proton-antiproton
collider. Its performance is as follows: luminosity—2.1 × 1032 cm−2 s−1 , rotation period—
3.8 × 10−6 s, total acceleration time—10 s, proton-antiproton cluster extent—38 cm, radius
of p(p)-beam—34 × 10−4 cm (29 × 10−4 cm). The ring length of this accelerator is equal to
6.9 km.
A proton LHC (Large Hadron Collider, 2009) has been constructed on the basis of LEP
to reach fantastic energies 14 TeV by today’s standards. However, this is not the limit as
in principle the circular proton machines are capable of providing the energies from 100
to 1000 TeV. Because of this, the creation of another proton supercollider is technically
possible. At the present time this idea is put forward for discussion. A tentative name
for this machine is VLHC (Very Large Hadron Collider). Its contemplated running into
operation should not be expected earlier than in 20-30 years.
The basis of the scientific program of LHC is search experiments. In the first place it is
an experimental looking for Higgs bosons and superpartners of ordinary particles. Also, it
is planned to realize a search for preons and heavy gauge bosons W ±′ and Z ′ , additionally
to the SM gauge bosons. Another problem is associated with a search for the formation of
quark-gluon plasma using the potentialities of LHC. The detection of such processes is a
real challenge for researchers as the cross sections of the expected processes are extremely
small. Moreover, some of the background events possess the intensities by milliards higher
than that of an event under study. Reliability of the obtained results will be ensured
by the simultaneous use of two different detectors operated by different research teams.
Two all-purpose detectors ATLAS and CMS are constructed to facilitate the solution of
the principal tasks of LHC. It should be noted that the detectors are constructed on the
basis of dissimilar conceptions. Their magnetic systems, the general construction, detecting
devices are considerably different. It is obvious that the coincidence of the results obtained
at ATLAS and CMS should be indicative of their maximum reliability.
A new ALICE detector that is also created at the present time is intended for investiga-

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 251

tion of collisions with ultrarelativistic energies: nucleus-nucleus (Pb-Pb, Ca-Ca) as well as

proton-proton and proton-nucleus. Since these collisions exhibit a hyperhigh energy density
(5.5 TeV per each pair of colliding nucleons), the occurrence of quark deconfinement and
the formation of quark-gluon plasma may be expected.
Electron and proton beams may be ejected from accelerators and directed to the exter-
nal targets, both hydrogen and nuclear. The produced charged π ± -mesons, K-mesons, or
p may be focused into the secondary beams for their further use during the experiments.
This process may continue to produce muon and neutrino beams. Since the decay of the
π − results in two weakly interacting particles µ− and ν µ , after its passage through the
absorber the π-meson beam is turned to the muon beam with the neutrino contaminant.
At a sufficiently large thickness of the absorbing material muons disappear, leaving a pure
muon antineutrinos beam. The K + - and π + -mesons (the K − - and π − -mesons) resultant
from the proton bombardment of a beryllium oxide target are used as the neutrinos (an-
tineutrinos) source. Basically, the beam comprises muon neutrinos and antineutrinos, the
electron component being strongly suppressed. The energy of neutrinos within the beam is
uniformly distributed from zero to Emax = rK,π pµ c, where pµ is a muon momentum and

m2K,π − m2µ
rK,π = ,
m2K,π

(rK = 0.954, rπ = 0.427). The Earth shield serves as a muon absorber. An absorber
1 km in length provides absorption of muons with the energy up to 200 GeV. Increasing
the maximum energy of muons necessitates further growth of the absorber length. As the
neutrino is electrically neutral, focusing of the neutrino beam is accomplished indirectly.
In Fig. 13.1 the scheme of the K − - or π − -meson decays into the muon and the muon
antineutrino is displayed. The angle φ the muon antineutrino trajectory makes with the

FIGURE 13.1
The scheme of producing the neutrino beams.

original meson trajectory is determined uniquely by the muon antineutrino energy and the
muon momentum: r
mK,π c rK,π pµ c
φ= − 1. (13.14)
pµ Eν
Thus, it is evident that changing the muon beam momentum with the help of quadrupoles
and band-pass magnets one can gain focusing of the neutrino beam. Muon storage rings
are used as neutrino fabrics, where electron neutrinos and muon antineutrinos (or electron
antineutrinos and muon neutrinos) are produced approximately in equal proportions already
within the beam. The production of focused high-energy neutrino beams is required mainly

© 2011 by Taylor and Francis Group, LLC

 252 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

for so-called “long-baseline” oscillation experiments, in the process of which a neutrino

beam created by the accelerator penetrates the Earth’s thickness and is detected by an
underground detector. The principal objective of such experiments is a study of neutrino
oscillations (transitions νi ↔ νj6=i , where i, j = e, µ, τ ).
Compton scattering of high-energy electrons from laser photons makes it possible to
produce γ-beams with the energy amounting to 80% of that for the primary electrons (this
procedure was first realized at SLAC in 1963). This opens up possibilities for the creation of
γγ-colliders with the luminosity to 10%Le , where Le is the luminosity of a parent electron
collider (e+ e− or e− e− ). Besides, a γ-source may be provided by the classical photon
bremsstrahlung of e− or e+ beam. The next generation electron-positron colliders, for
example, NLC (Next Linear Collider) with a maximum energy of 500 GeV, are so designed
that they can be operated both in the γγ- and e± γ-modes. In this way combined colliders
may be constructed in addition to the available electron and proton colliders. Presently, this
class of accelerators is represented by the DESY synchrotron (Hamburg), version HERA
(e− p-collider), that is used to study collisions of electrons (with the energy 30 GeV) and
protons (with the energy 820 GeV).
Probably, our reader is of the opinion that only stable particles may be the candidates
for the function of collider-accelerated particles as the limited lifetime of unstable particles
will preclude their acceleration to very high energies. However, this is a mere delusion, so
to speak, a tribute to the nonrelativistic style of thinking. Just recall the time deceleration
phenomenon associated with the moving clock. In the lab frame, an unstable particle cov-
ers to its decay the distance that is much greater than may be derived from nonrelativistic
considerations by means of simple multiplication of its velocity by its proper lifetime. Pre-
cisely this principle is the basis for the construction of muon colliders (MC). The greatest
difficulties during the construction of the MC are associated with the fact that over the
lifetime of a muon, that is equal to τµ =2.2 ms in the proper reference frame, muon beams
should be stored, cooled, accelerated, and brought into interaction with each other. In the
lab frame the muon lifetime is increased due to the relativistic factor by the value

1 m c2 Eµ
κ0 = p = pµ = 2
.
2
1 − v /c 2 2 2
mµ c 1 − v /c 2 m µc

Then the intensity of muon decay along the beam trajectory may be written in the form:

dN N
=− , (13.15)
dl L µ κ0

where Lµ = cτµ ≈ 660 m and it is assumed that v ≈ . In the absence of the acceleration
mode this leads to the ordinary exponential form of the beam loss:

l
N = N0 exp (− ). (13.16)
L µ κ0

In the acceleration section κ is not a constant

eVg
κ = κ0 + κ′ l = κ0 + l, (13.17)
m µ c2

where eVg is a acceleration gradient. Substituting expression (13.17) for κ0 into (13.15), we
obtain
 1/(Lµ κ′ )
κ0
N (l) = N0 ,
κ0 + κ′ l

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 253

to give the relation

 1/(Lµ κ′ )
N (l) Ei
= , (13.18)
N0 Ef
where Ei (Ef ) is an initial (final) muon energy. From (13.18) it follows that the decay losses
in the muon beam are minimal for
Lµ κ′ ≫ 1. (13.19)
Substitution of the numerical values into (13.19) demonstrates that the normal functioning
of a muon collider requires the existence of the acceleration gradient
eVg ≫ 0.16 MeV/m
throughout the whole muon system.
It is well known that the creation of e− e+ colliders with multiTeV energy is restricted by
two factors: (i) increasing losses by synchrotron radiation; and (ii) drastic increase in the
material costs as two linear accelerators will be required to avoid considerable synchrotron
radiation in the storage rings.
The bremsstrahlung of muons is negligible. They may be accelerated and stored in the
rings, whose radius is considerably less as opposed to hadron colliders with comparable
energies. Unlike hadron colliders, where the background appears at the point of particles
interaction and comes from the accelerator as well, the background of the MC may be found
in the detectors only. Also, the MC exhibits high monochromaticity. The roof-mean-square
deviation R from the Gaussian energy distribution in the beam falls within the interval
from 0.04% to 0.08%. Owing to cooling of the muon beam, R may be decreased down to
0.01%. Thus, the energy resolution of the beam in the MC is much higher than that in
e− e+ colliders. Another advantage of the MC is its fast rearrangement for operation in
the µ− µ− - or µ+ µ+ -mode. Since the construction of the MC includes special storage rings
to provide optimization of the luminosity for some energy, the MC is an ideal instrument
for investigation of resonances with an extremely small decay width (e.g., the Higgs boson
of the SM). A quantity determining the energy spread in the beam σ√s is an essentially
important characteristic for the collider. In the case of the MC this quantity is given by
the expression:   √ 
√
R s
σ s = (7 MeV) . (13.20)
0.01% 100 GeV
Note that the detection and examination of a particular particle in the s-channel may be
successful provided σ√s is of the same order as the total decay width of this particle.
At present two projects are investigated for the construction of the MC. The first,
√ First
Muon Collider, allows for building a MC having the center-of-mass frame energy s = 0.5
TeV and luminosity L ∼ 1033 cm−2 s−1 . The second, √ Next Muon Collider, is a collider of
much higher power with the following characteristics: s ∼ 4 TeV, L ∼ 1035 cm−2 s−1 .

13.2 Detectors
For progress in experimental particle physics the development of accelerators is an essen-
tial but insufficient requirement. Another prerequisite is the development of experimental
methods, the creation of more and more sophisticated detectors, computerization, and other
modern technologies. In order to propose an experiment for the testing of a particular hy-
pothesis or explaining experimental results, a physicist-theorist must necessarily know the

© 2011 by Taylor and Francis Group, LLC

 254 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

advanced elementary particle detection techniques. Below is a short summary of the most
extensively employed types of detectors.
The functioning of a detector is based on ionization or excitation of the atoms within the
detector material by the accelerated charged particles. A charged particle leaves a track
of the ionized or excited atoms that enables one to judge the form of a particle path. Un-
charged particles (photons, neutrons, neutrinos, etc.) do not ionize the material, exhibiting
themselves by the secondary charged particles, produced as a result of interactions between
the uncharged particles and detector material. For instance, photons are detected by the
produced electron-positron pairs and Compton electrons of recoil; neutrinos—by the gener-
ation of the µ+ -mesons upon their collision with protons and electrons, by the formation of
recoil electrons or due to the inverse β-decay within the detector material. Passing of the
particles through the material may be also accompanied by the formation of free electrons,
ions, positrons subsequently annihilating with electrons to photons, and also by different
reactions, thermal phenomena, and the like. Because of this, the particles may be detected
by electric pulses appearing at the detector’s exit, by photoemulsion blackening or else by
changes in the structure of solid material in the detector. Electric signals of the particles
are generally so weak that an additional system is required for the signal amplification.
The most important characteristics of a detector are as follows: efficiency, that is, detec-
tion probability for a particle finding its way into the active volume of a detector; detector
memory time, that is, the time period when the changes within the detector volume due
to transit of the detected particle are retained; dead time of a detector, that is, the time
of returning to the initial sensitivity after the regular action; energy, spatial and temporal
resolution, that is, the energy, position, and time determination accuracy for a particle
detected. According to the information obtained about the particles, all detectors may be
conventionally subdivided into two classes. The first class comprises detectors signaling
about the particles by short electrical pulses. This class, in turn, may be further subdivided
into two subclasses: (a) detectors of single particles; and (b) detectors of electromagnetic
and nuclear cascades. The second class is represented by track registration detectors pro-
viding direct observation of particle tracks.

1 a. Single Particle Detectors

Ionization Chamber (IC). IC is the simplest form of such a detector. It comprises

two electrodes, and the interelectrode space is filled with gaseous, liquid, or solid material.
Under the effect of moving charged particles, the material within the chamber is ionized
to produce free electrons and ions. Migrating between the electrodes, electrons and ions
create a current pulse in the external circuit of the chamber. IC may be used to detect
both the particle fluxes (measurement of the average output current) and single particles
(measurement of output pulses). Actually, IC may be used for the detection of all particle
types, although by the appropriate selection of the detector material and electric field one
is enabled to adjust the chamber for the detection of a certain particle type.
Proportional Multiwire Chambers (PMC). PMC represent a modern prototype
of proportional counters, gas-discharge detectors, where the electric signal amplitude at
the output is proportional to the energy spent by the particle on gas ionization. PMC
consist of numerous parallel, small-diameter (∼ 2 × 10−3 cm) anodic wires fixed between
two flat cathodes, solid or wire-type but with wires of greater diameter. Each of the anodic
wires is functioning as an independent detector. Under the effect of an electric field, the
primary electrons formed by the particle entering in the detector are moving to the anode
to get into a high-strength field, where they are greatly accelerated causing the secondary
gas ionization, that is, electron avalanches are observed. The spatial resolution of PMC is

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 255

minor: ≥ 7 × 10−2 cm. However, short dead time ≥ 3 × 10−8 s make them most widely
used position detectors.
Drift Chambers (DC). DC are used as position detectors. These gas-discharge devices
involve wire electrodes, where the particle positions are determined by the drift time of the
electron in a homogeneous and constant electric field, from the place of their origination
to anodic wires. The field around the anode is inhomogeneous, resulting in electron accel-
eration and hence in the formation of electron avalanches. The spatial resolution of DC
achieves 10−6 cm. As the dead time (electron drift time) of DC is long (∼ 10−6 s), these
chambers cannot function in intensive loading conditions.
High-precision Position Detectors (HPD). HPD are used to reconstruct the particle
positions and paths at the vertex of the event under study or in its neighborhood. A typical
task for HPD is a search for the “second vertex” resultant from the decay of the short-lived
(with a lifetime of 10−12 –10−13 s) particle that was produced at the first vertex.
Most popular are microstrip semiconductor detectors structured as follows: strips of a
conducting material are deposited as electrodes on one of the surfaces of a silicon monocrys-
tal, whereas the other surface is metallized. The voltage applied to these electrodes makes up
a few Volts. An ionizing particle in transit through the crystal is forming the electron-hole
pairs migrating towards the electrodes to create the current pulses. The spatial resolution d
of microstrip detectors is determined by the strip width and interstrip gaps, reaching 10−7
cm (higher resolution is demonstrated only by nuclear photoemulsions d ∼ 10−8 cm). The
temporal resolution of these detectors is of the order of 10−8 s.
Scintillation Counter (SC). SC comprises a scintillating material (special liquids, plas-
tics, crystals, noble gases), where a charged particle effects both ionization and excitation of
the atoms and molecules forming this scintillator. Recovering to ground state, these atoms
and molecules are emitting photons incident on the cathode of the photomultiplier (PM) to
knock off photoelectrons from the cathode. Owing to these electrons, a pulse with amplitude
proportional to the energy transferred by the particle to the scintillator is formed at the an-
ode of the PM. The accuracy of the measured particle energy provided by the SC is within
10%. Since under the effect of charged particles the majority of scintillators reveal the
characteristic luminescence time of about 2 × 10−8 s, the transit time of a particle through
the counter may be determined with high accuracy. The detection efficiency for charged
particles is close to 100%. Neutrons may be detected (by recoil protons) with the use of
hydrogen-containing scintillators, and photons—with the use of higher-density scintillators
like iodide sodium. Detection of neutrinos requires the use of combined scintillators. So,
in their experiment (1953) F. Reines and C. L. Cowan have used the hydrogen-containing
scintillator with an addition of a cadmium salt for the detection of electron antineutrino in
the reaction
p + ν e → n + e+ . (13.21)

The first scintillation flare was caused by annihilation of the positron with the electron of
scintillator, while the second flare occurring in (5 − 10) × 10−6 s was due to the cadmium
atom returning into the ground state after absorption of a neutron.
Cherenkov Counters (CC). The operation principle of CC is based on the detection
of Cherenkov radiation. This radiation is generated when charged particles are moving in
a transparent medium at the speed v exceeding the speed of light in the medium
c
v≥ , (13.22)
n

where n is a refractive index of this medium. Light is emitted only in a forward direction
along the motion path of a particle, forming a cone with an axis in the direction of v, and

© 2011 by Taylor and Francis Group, LLC

 256 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the cone angle, emission angle, is determined by the relation

c
cos θ = . (13.23)
vn
This phenomenon is similar to the acoustic cone of the airplane flying at a supersonic speed.
A light flare following the particle motion in the medium is detected by the PM. The counter
is used for the detection of relativistic particles as well as estimation of their charge, speed,
and motion direction (to within 10−4 ). Measuring the particle momentum by deflection
in a magnetic field, one is able to measure its speed with the help of the CC and hence
to determine a mass of this particle. Relation (13.22) forms the basis for the operation of
threshold or integrating CC that can detect all particles having the speeds above threshold
v > vt = c/n. And the counter provides radiation measurements over the whole range of
angles from 0 to θmax = arccos(/vn). The operation of angular or differential CC is based on
relation (13.23). These counters detect particles at the speeds from v0 to v0 + ∆v. Emission
of the particles parallel to the optical axis of the counter is collected only in a narrow range
of angles, from θ0 to θ0 + ∆θ. The CC comprise a radiation-generating medium, collecting
optics that directs this radiation to the photocathode of PM, PM and recording system. A
new type of gas CC, RICH (Ring Image Cherenkov detector), has been proposed recently.
This detector provides detection and imaging of the particles by their Cherenkov radiation.
The energy region, where the CC offer mass separation for the particles, has an upper limit
as the difference in the speeds of particles distinguished by their masses is decreased with
growing energy. For instance, the speeds separation of the π- and K-mesons by threshold
gas CC is possible up to the energies amounting to several dozens of GeV, whereas by
differential gas CC with a compensation of radiation dispersion—up to several hundred
GeV.
Transition Radiation Detectors (TRD). The operation of these detectors is based
on the formation of an electromagnetic field by the moving charged particle. The elec-
tromagnetic field of this particle changes its configuration when the particle goes from a
medium with the dielectric permeability ǫ1 to that with ǫ2 . This process is accompanied by
the emission of transient radiation. The distinguishing feature of this radiation is the fact
that its properties are determined by the relativistic factor
1 E
γ=p = 2
.
2
1 − v /c 2 mc

In this case the radiation intensity is proportional to the particle energy, and radiation
is concentrated within a cone of angle θ = γ −1 . In layered structures, where a particle
crosses the interface repeatedly, the intensity of transient radiation may be resonantly am-
plified providing a means for the detection and identification of ultrarelativistic particles
with γ > 103 . The Lorentz factor γ ∼ m−1 and intensity of transient radiation at constant
energy E will be greater for particles with lower mass. This allows for mass separation over
the energy range hardly accessible for gas CC. TRD consist of a layered medium, usually
comprising a multitude of foils from lightweight materials (Li, Al), which are stretched per-
pendicular to the particle direction, and a radiation detector. In some TRD the radiator
may be represented by ordinary porous materials like foamed plastics. The radiation de-
tection is most commonly done by proportional wire chambers filled with heavy gases.

1 b. Electromagnetic and Hadron Cascade Detectors

The operation principle of these detectors, also referred to as Total-Absorption Detectors

(TAD), is based on total absorption of the cascades created by the detected particles within

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 257

the detector material. TAD enable detection of the integrated Cherenkov radiation for
all the particles forming the electron-photon shower (total-absorption CC) or integrated
energy spent by all the particles for ionization (calorimeters). In electromagnetic cascades
this energy is practically equal to the energy of the primary electron or photon. In hadron
cascades ionization requires the major part of the energy possessed by the primary particle
but some part of its energy (up to 20–30%) is spent for nuclear disintegration, then carried
away by neutrinos formed as a result of particle decays, and is not detected by calorimeters.
Total-Absorption Cherenkov Counters. These detectors provide detection of pho-
tons and electrons with estimation of their energy. The blocks of lead glass serve as radiators.
Their size must be sufficient for absorption of the main part of the shower produced by the
primary particle. Cherenkov radiation is detected by the PM.
Calorimeters (C). C are intended to measure the energy of particles, both charged and
neutral, beginning from 102 GeV and higher. Interacting with the nuclei within the material
of C, a high-energy particle produces a cascade of the high-energy secondary particles, in
turn, interacting with the material to generate new particles. Because of this, electron-
nuclear showers occurring in the active volume of the device rapidly moves in a direction
of the primary particle, and its energy is spent for ionization of the material. Provided a
layer of the material in C is sufficiently large, all shower particles are left in the material,
and the number of created ions is proportional to the energy of the primary particle. Then
these ions are collected at the calorimeter electrodes, and their total charge is measured to
determine the primary particle energy with an accuracy of 10–15%. The use of C makes
it possible to locate the shower origination and determine its spatial development. The
simplest C are constructed as “sandwiches” consisting of alternated layers of heavy material
and ionization detectors. In electromagnetic C such a sandwich comprises thin layers of lead
and scintillator. Its total thickness may reach a few dozen centimeters.
Hadron cascades are slowly developing in the majority of materials, penetrating deeper
than the electromagnetic ones. Because of this, hadron C possess a much greater thickness,
up to several meters, with thicker layers of the material (commonly iron) and scintillator,
that may be replaced by other ionization measuring detectors.

2. Track Detectors

These detectors enable observation of particle tracks. Track detectors, being subjected to
magnetic fields, make it possible to determine a sign of the electric charge for the particles
and to measure their momenta by the path curvature to a high degree of accuracy. The
first track detector was constructed by Ch. Wilson in 1912 and received the name the
Wilson chamber. Its operation is based on condensation of supersaturated vapor and the
formation of visible little drops of liquid at the ions originating along the path of a fast
charged particle. The device comprises a closed vessel, having windows intended for the
track observation, filled with gas and saturated vapors of some liquid substance, for example,
methyl alcohol. On rapid adiabatic rarefying this gas is cooled, whereas the vapor becomes
supersaturated. After photographing the tracks, the chamber regains its initial state due
to fast gas compression, causing evaporation of the droplets at the ions with the formation
of saturated vapor, and the former ions influenced by the electric field are collected from
the tracks at the electrodes.
Bubble Chamber (BC). BC is one of the main types of track devices in high-energy
physics. It comprises a large container several meters in diameter filled with a transparent
superheated liquid. Its boiling is delayed owing to the high pressure that is 5–20 times
higher than the atmospheric pressure. The abrupt decrease of the pressure results in su-
perheating of the liquid, and in case an ionizing particle is passing through the chamber

© 2011 by Taylor and Francis Group, LLC

 258 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

an additional heating leads to drastic boiling of the liquid in a narrow channel along the
particle path. Its trajectory is marked by a chain of vapor bubbles. These bubbles are
allowed to grow for a period of 10 ms, afterwards they are photographed by stereoscopic
cameras. Subsequently, the starting pressure is applied to the liquid causing collapse of the
bubbles, and the chamber is ready for operation again. In the BC the most common is liquid
hydrogen. Deuterium, propane, and such heavy liquids as xenon and Freon are more rarely
used. The latter are of special interest, especially for the detection of neutrino interactions.
BC is usually placed in a strong magnetic field, and by the track curvature one can measure
the particle momenta to a high accuracy. The spatial resolution of BC comes to 10−2 cm.
The main advantage of BC is the possibility to use its working liquid both as a target for
the incoming particles and as a detector for the reactions proceeding upon particle collisions
with electrons and nuclei of the liquid. This advantage is especially marked in studies of
complex processes involving large numbers of particles. The principal disadvantage of BC
is its “uncontrollability”: it is impossible to realize its response on a signal of fast detectors
that previously select the required events. The response of BC with a period of ∼ 1 s is
synchronized with points in time of fast beam ejection from the accelerator.
Spark Chambers (SC). SC are the controlled gas discharge detectors including a se-
ries of parallel metallic plates placed into the container filled with inert gas. These plates,
alternately, are connected to the high-voltage source or have an earth connection. Provided
an ionizing particle crosses the working volume of SC, by command of the monitor counters
a short high-voltage pulse (10–20 kV/cm) is applied. Originating at the points of particle
pass, spark discharges parallel to the electric field are photographed or located by the mag-
netostriction method.
Streamer Chambers (StC). StC are the counter-controlled gas discharge detectors,
where discharges are formed exclusively along the particle tracks. These chambers contain
two flat parallel electrodes positioned at a distance measuring several dozens of centimeters.
To the electrodes a very short (≤ 10−8 s) high-voltage (10–50 kV/cm) pulse is applied. In
these conditions the discharges, originating at an ionizing particle passage, are terminated
and take the form of short (∼ 10−1 cm) luminous channels (streamers) aligned with the
field. Their photographs are made to obtain the track images. As contrasted to SC, StC
are isotropic, that is, they are capable of reproducing the tracks of every spatial orientation
and can measure the particles ionization.
As a rule, all the available particle detectors are combined detector systems (CDS) featur-
ing a series of detectors integrated in one detecting unit. CDS represent the major element
of the modern accelerator. Their size measures dozens of meters, mass amounts to ∼ 104 t,
and the number of information channels may be as great as 106 . The personnel required for
their operation runs into hundreds of people, whereas the construction expenditures com-
prise a significant part of the total cost for the whole accelerating complex. The majority of
CDS are similar in structure, though the choice, amount, dimensions, and arrangement of
elements are dependent on the specific task at hand. Most typical elements are as follows:
target, vertex detector surrounding the target that indicates the reaction products and de-
termines their escape direction; position detectors localizing trajectories of primary and
secondary particles; spectrometric detectors measuring the momenta of secondary particles
or their energy; and identifiers of secondary particles. Large-scale CDS are given proper
names such as ATLAS, ALICE, DELPHI, etc. Now the particle fluxes passing through CDS
are as great as 108 s−1 . Unfortunately, the difficulties in processing the measurement results
in the case of numerous information channels and high detection rates generally prevent a
real-time analysis. Considering this situation, information is recorded and processed on
completing the experiment.

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 259

13.3 Neutrino Telescopes

Among the fundamental particles, the neutrino holds a special place since it plays an im-
portant role in large-scale events of the universe. The finding that the neutrino has a mass
makes this particle a worthy candidate for the role of a particle constituting the hot dark
matter. This, in turn, enables one to evaluate the average matter density in the Universe,
age of the Universe, and its further fate. The problems associated with detection of cosmic
neutrinos are the subject matter of the neutrino astrophysics. The neutrino astrophysics
may be considered as a compound part of the elementary particle physics. And this is related
not only with the fact that both physics divisions are concerned with the universe structure.
The other important aspect is that solution of the problems concerning the generation and
detection of the neutrinos depends upon the character and intensity of interaction between
the elementary particles. Because of this, it is obvious that neutrino telescopes (NT), being
basic instruments of the neutrino astrophysics, are useful for studies of particle physics too.
Depending on the detection technique, all available NT, may be subdivided into two
classes: NT operating in the continuous counting mode and NT operating in the discrete
counting mode. The first class includes NT using the radiochemical methods. The second
class involves NT intended for real-time detection of the particles, the production of which
is initiated by the interaction between the neutrinos and the working medium of NT.
The operation of radiochemical NT is based on investigating the process of the inverse
β-decay due to the incoming neutrino

νe + X → e− + Y, (13.24)

where X are nuclei of the elements determining the initial composition of NT detector. As a
rule, nuclei of Y originating in the detector are radioactive, their half-life T1/2 determining
the duration of an active measurement stage ta ∼ (2—3)T1/2 . The formation rate of
daughter nuclei is given by the expression:
Z
R = N Φ(E)σ(E)dE, (13.25)

where Φ(E) is the neutrino flux incident on the detector, N are numbers of detector atoms,
and σ is the cross section of the process (13.24). At the incident neutrino flux 1010 cm−2 ·s−1
(approximately amounting to the flux of solar neutrinos incident on the Earth) and σ of
the order of 10−45 cm2 the provision of a single useful event a day necessitates about 1030
atoms. Consequently, a mass of this detector should be in the region of several kilotons.
Chemical analysis of the detector material takes place in time ta , and the nuclei number Y
is indicative of the capture rate for the neutrinos. The advantage of radiochemical NT is the
possibility for varying the reaction energy threshold with changes in the detector material.
This makes them indispensable in studies of low-energy neutrino fluxes. At the same time,
radiochemical NT feature an inability to measure such neutrino characteristics as hitting
time for the detector, energy, and trajectory direction. The latter is rather discouraging as
we have no chances to distinguish between, for example, the solar neutrino and the neutrino
produced by the terrestrial source.
As an example of a radiochemical NT, we consider the Homestake facility that was the
first to study the fluxes of the solar neutrinos (1967–2001). The neutrinos were detected
with the use of the chlorine-argon method, that is, operation of this NT was based on the
chemical reaction
37
Cl + νe → 37 Ar + e− . (13.26)

© 2011 by Taylor and Francis Group, LLC

 260 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Isolation of the useful event was realized using the decay of

37
Ar → 37 Cl + e− + ν e , (13.27)

whose half-life is 35 days. This facility, representing a vessel with a capacity of 390 l filled
with 610 t of perchloroethylene (C2 Cl4 ), was located in the gold-bearing mine (Homestake,
South Dakota, USA) at a depth of 1 480 m. As argon atoms are produced in the form of
a volatile compound, they were isolated approximately once a month. The obtained argon
was subjected to multistage processing. At the final stage, special small-size proportional
chambers were filled with this argon. Then the chambers were shielded with lowest-activity
lead to provide observation of decays (13.27).
The operation of the second-type NT may be based, for example, on the detection of
elastic scattering
νl + e− → νl + e− , (13.28)
where l = e, µ, τ . The spectrum of recoil electrons is given by the following expression:
"  2 #
dσ 2 2 T me T
= σ0 g1 + g2 1 − − g1 g2 2 . (13.29)
dT Eν Eν

Here T is a kinetic energy of recoil electrons, σ0 = 8.8 × 10−45 cm2 ,

1
g1 = sin2 θW ± , g2 = sin2 θW ,
2
where “+” sign is associated with the νe -scattering, while “-” sign—with the νµ - and ντ -
scattering. Because of this, in the second case the scattering cross section is approximately
one-sixth of that in the first case making it possible to distinguish between the neutrino
kinds. The cross section of the νe -scattering integrated with respect to the energy is simple
in form
Eν
σ(νe e) = 9 × 10−44 cm2 . (13.30)
10 MeV
The angular distribution of recoil electrons is characterized by a sharp maximum
p in the
forward direction (relative to the incoming neutrino) whose width is ∆θ ∼ mν /Eν . This
enables one to determine the direction of a neutrino source by the trajectories of scattered
electrons. Another advantage of such a facility is the possibility for the detection of indi-
vidual acts when the neutrino is hitting the detector, permitting measurements of time and
energy for the neutrinos (real-time operation). A limitation of NT based on recoil electrons
resides in the presence of a considerable background as the detected electrons may be pro-
duced during the processes of elastic scattering involving any neutral particles. However,
in the case of the high-energy neutrinos the background elimination is greatly facilitated.
Also, the operation of the second-type NT may be based on the detection of particles
produced in reactions with the high-energy neutrino, for example,

νl (ν l ) + N → l(l) + N ′ , νl (ν l ) + N → l(l) + N ′ + X, (13.31)

νl (ν l ) + 16 O → l(l)16 O + π + (π − ), (13.32)
where N = p, n and X denotes hadron collections. Since the secondary high-energy elec-
trons and hadrons are responsible for the nuclear-electromagnetic shower and the main free
path of these electrons within the detector material is very short, they are practically in-
distinguishable from hadrons. Compared to electrons, the mean free path of high-energy
muons is very long as their energy losses for bremsstrahlung, formation of the e− e+ -pairs

© 2011 by Taylor and Francis Group, LLC

 Technical equipment of particle physics 261

and nuclear interactions are small. Of special importance is the fact that muons are mov-
ing practically in the same direction as the neutrinos producing them. The average angle
between the νµ - and µ-trajectories expressed in degrees is determined by the expression
s
100
< θ >≈ 2.6 × .
Eµ (GeV)

In the active volume of NT, bremsstrahlung photons, the e− e+ -pairs and hadrons originate
along the muon trajectory to initiate the nuclear-electromagnetic showers. With Eµ ≈ 100
TeV particular minor showers are overlapping, and the whole muon trajectory is glowing
due to the Cherenkov radiation of these showers. Based on the direction and intensity of
the Cherenkov radiation, one can determine the trajectory and energy for muon.
Hadrons generated in the reactions described by (13.31) and (13.32) also initiate nuclear-
electromagnetic showers, the direction and energy of which is determined by the Cherenkov
radiation. The detection of showers may be performed using the acoustic method as well
[51]. In this case the detected signal is represented by the pressure pulse in the active
volume of NT conditioned by the drastic heating of a narrow channel within the shower
due to ionization energy losses of the electrons. For instance, in water an acoustic signal is
propagating in the form of a thin disk, with a thickness of about the shower length s ∼ 5 m
and characteristic radius R ∼ 1 km. By this method the detecting element is represented
by hydrophones detecting signals perpendicular to the shower axis.
In NT based on detection of the secondary muons the effective volume of the detector is
considerably greater than the physical volume owing to the detection of muons generated
within a thick layer of the material surrounding the detector. When the neutrinos are
detected with the use of a detection mechanism on the basis of hadron showers, however,
this is not the case. Short lengths of hadron showers enable their detection by the Cherenkov
radiation only within the physical volume of NT.
As a working material (detector) for NT of the second type one can use water or arctic
ice. Arctic ice represents a sterile medium with lower concentration of radioactive elements
than in sea or lake water. The use of arctic ice as a detector contributes considerably to the
sensitivity of NT. For instance, an NT positioned at a depth of about 1 km makes it possible
to separate the background muon (atmospheric) flux that is 100 times greater compared
to the limiting flux for the deep-sea NT DUMAND (Deep Underwater Muon and Neutrino
Detector), which was positioned in sea water at a depth of 4.5 km. NT AMANDA located
at the South Pole is intended for studies of high-energy neutrino fluxes. Deep-water NT on
muons, BAIKAL NT-200 (the Baikal Lake), is an example too.
Among NT of the second type one may name the Super-Kamiokande facility (Japan,
Kamioka) constructed jointly with the U.S. specialists (1996). This NT is located in the
mine with a shielding depth of 2700 m in water equivalent; absorption of particle fluxes by
the rock is equivalent to a water thickness of 2700 m. The principal element of this facility
is a water Cherenkov detector in the shape of a cylinder, 39 m in diameter, and 41 m in
height, that contains 50000 t of water and provides ring imaging of the detected particles.
The detector is optically subdivided into the internal (working) volume scanned by 11200
photoelectronic multipliers (PM), and also the outer (shielding) volume containing 2200
PM and operating in the anticoincidence mode. This NT can investigate the fluxes of both
the solar and atmospheric neutrinos.
The main source of the solar neutrinos is a series of thermonuclear fusion reactions at the
central part of the Sun, resultant in hydrogen-to-helium transformation without catalysts,
the hydrogen cycle. This chain may be represented as a multistage process

4p → 4 He + 2νe + 2e+ + 26.73 MeV − Eν , (13.33)

© 2011 by Taylor and Francis Group, LLC

 262 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where Eν is an energy carried away by the electron neutrino, its average value being ∼
0.6 MeV (Emax < 18.8 MeV). In Super-Kamiokande the detection of the solar neutrino
is realized using the neutrino elastic scattering reaction from electrons with the energy
threshold 5.5 MeV.
Cosmic rays interacting with atomic nuclei initiate in the atmosphere surrounding the
Earth the production of pions, kaons and muons, the decay channels of which involve
electron and muon neutrinos as well as antineutrinos

π ± → µ± + νµ (ν µ ) → e± + νe (ν e ) + νµ (ν µ ),
(13.34)
K ± → µ± + νµ (ν µ ) → e± + νe (ν e ) + νµ (ν µ ).
The neutrino flux is formed in the region of 10–20 km altitudes above sea level, its energy
varying from 100 MeV to 1 000 GeV. Since the dominant interaction type for the neutrinos
with such high energies is interaction with the target nuclei, in case of Super-Kamiokande
the detection of the atmospheric neutrinos is performed using reactions (13.31) and (13.32).
The second-type NT, Sudbury Neutrino Observatory (SNO), came into use in Canada
in May 1999. At this facility the detector is represented by 1000 t of heavy water (D2 O)
enabling investigation of the solar neutrino with the help of the following processes:
νe + d → p + p + e− , (13.35)
− −
νl + e → νl + e , (13.36)
νl + d → νl + n + p, (13.37)
Reaction (13.35) is sensitive to νe -neutrino, whereas reactions (13.36) and (13.37) are sen-
sitive to the neutrinos of all three kinds. For reactions (13.35) and (13.36) the energy
threshold equals 5 MeV, and that for reaction (13.37) is 2.225 MeV.
The multipurpose NT named KamLAND (Japan, Kamioka) using 1000 t of an ultra-
pure liquid scintillator as a detector came into operation in Spring 2001. Although the solar
neutrino may be detected by KamLAND, its main function is to observe the oscillations in
the total neutrino flux coming from ten reactors localized in the region at 80–350 km from
the detector.
The flux of the solar neutrinos originating in reaction
7
Be + e− → 7 Li + νe
(so-called beryllium neutrinos) is especially sensitive to the neutrino characteristics. Real-
time measurements of this monoenergetic (Eν = 0.86 MeV) flux are the principal objectives
of NT named BOREXINO (Gran Sasso, Italy) that is based on recoil electrons with the
threshold 250 keV and came to operation early in 2002. Recoil electrons caused by νl e-
scattering (the cross sections for νµ and ντ are smaller than for νe ) produce light flare
in the bulk of the liquid scintillator, that is detected by the PM. Nylon sphere contains
300 t of ultra pure pseudocumene, and 100 t of pseudocumene contained in the central
region comprise the effective (sensitive) volume. The nylon sphere is, in turn, surrounded
by pseudocumene filling the corrosion-proof steel sphere 13.7 m in diameter that contains
optical elements surrounding the nylon sphere. The whole construction is submerged into
the reservoir with purified water having a mass of 2500 t. The experimental threshold
is set as 0.25 MeV because the energy spectrum of recoil electrons is continuous up to
0.66 MeV. At these low energies the control of natural radioactivity caused by radioactive
isotopes being present everywhere is the greatest problem. By the present time extensive
research has been conducted with the aim to select materials and realize their purification
to extremely high levels of radioactive purity. Simultaneously, the measuring techniques for
ultra low radioactivity levels have been developed. The attained results are quite impressive:
10−16 − −10−17 (gram of contaminant per gram of material) for 232 T h and 238 U .

© 2011 by Taylor and Francis Group, LLC

 Problems

2.1. A µ− -meson has been captured by an Al27 atom. Define the orbit radius of the µ− -
meson and its residence time inside the nucleus.
2.2. When one tries to integrate the expression (7.10) over the scattering angle then infinity
will be obtained. The reason consists in the long-range character of the electrostatic forces.
The particles are scattered, no matter how far away from a force center they fly. To get
a finite result one should allow for a screening effect of the electron shell. This is attained
under using the potential
Ze2 r
U (r) = exp (− ),
2πr a
where a has the value of the order of the atom radius. Using this potential, find the total
cross section of the α-particles scattering.
2.3. Consider the reactions mentioned below

1) Σ− + p → Λ0 + n, 2) Σ+ + p → K + + p, 3) p → n + e+ + ν e ,
0
4) Ω− → n + π + , 5) Ξ0 → Λ0 + K , 6) π − + p → Σ− + K + ,
7) π − + p → K − + Σ+ , 8) η → π + + π − + γ, 9) µ+ → e+ + e− + e+ .
Point out which of them are forbidden by conservation laws. Name these laws.
2.4. Examine the resonance production processes

π + + p → ∆++ → π + + p,

π − + p → ∆0 → π − + p.
Thinking ∆++ and ∆0 to enter into one and the same isotopic multiplet and taking into
consideration the experimental results

σπ+ p 
= 3,
σπ− p E=T r
π

where Tπr is the kinetic energy of the π-mesons at which the ∆-resonances appear, determine
the ∆-resonances isotopic spin.
2.5. Protons being bombarded by K − -mesons with the kinetic energy TK = 790 MeV, the
following reaction
K − + p → (π − Λ) → Λ + π + + π −
is observed. At the same time the emergent π + -mesons have the kinetic energy Tπ∗ = 300
MeV. Find the mass of the (π − Λ)-resonance and its decay energy.
2.6. Protons are bombarded by π − -mesons with the momentum 1.14 GeV, with the reaction
events
π − + p → n + 2γ
being registered. Two peaks in distribution on the angle of divergence of two γ-quants
in the center-of-mass frame are observed. The former is located near the angle 250 while

263
© 2011 by Taylor and Francis Group, LLC
 264 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the latter—near the angle 1000 . Find for each of these peaks the masses of corresponding
unstable particles.
2.7. Find the angular distribution of secondary particles in two-body decay in flight X →
1 + 2 when in the rest frame of the X-particle this distribution is isotropic.
2.8. For the decay
X → 1+2
find the distribution of secondary particles on the angle of their divergence in the lab frame
if in the rest frame of the X-particle at m1 = m2 the decay has an isotropic character.
2.9. Two particles with the isospins I = 1/2 and I = 1 make up a coupled state. Us-
(is) (is) (is)
ing the properties of the lowering and rasing isospin operators S± = S1 ± S2 build
isospin wavefunctions for the following states: Ψ(3/2, 1/2), Ψ(3/2, −1/2), Ψ(1/2, 1/2) where
Ψ(I, I3 ) is a wavefunction of a system with the total isospin I and the projection I3 .
2.10. Taking into account the f1 (1420)-resonance spin to be equal to zero, define isotopic
relations between strong decay probabilities of the f1 (1420)-resonance through the channels:
0
f → K K + + π− , f → K 0 K − + π+ ,
0
f → K +K − + π0 , f → K 0K + π0 .
Thinking the C-parity of f (1420) to equal +1, point out the basic peculiarities of detecting
the third decay channel.
2.11. Allowing for the isospin conversation law in the strong interaction, find the ratio of
the total cross sections for the following reactions:

p + p → d + π+ , n + p → d + π0 .

2.12. Ascertain on what isospin channels the following reactions:

π − + p → K 0 + Σ0 , π − + p → K + + Σ− ,

π + + p → K + + Σ+
could be proceeded. Assuming the channel with the definite isospin to dominate, find the
ratio of the total cross sections for these processes.
2.13. Define an isospin change and a third isospin projection change in the following weak
decays
0
K + → π+ + π0 , K → π+ + π− .
2.14. Let us consider a single π-meson production processes to be observed under neutrino
scattering by nucleons:
νµ + p → π + + p + µ− , (II.1)
νµ + n → π 0 + p + µ− , νµ + n → π + + n + µ− . (II.2)
If we assume the first pion-nucleon ∆(1232)-resonance contribution to dominate, then a
π − N -system will have the fixed isospin I = 3/2. The strangeness not being changed in the
weak processes (∆s = 0), if that is the case the hadron state isospin is varied by 1 (∆I = 1).
Show, then the cross section of the reaction (II.1) is three times bigger than that of the
reaction (II.2). Assuming ∆I = 2, find the ratio of these reactions as well. [NOTE: If one
assumes a certain fictitious particle with the isospin equal ∆I, a spurion, to be added to the
left-hand side of the equations describing the decay processes, then the above-mentioned
decays may be considered as the processes going with the isospin conservation].

© 2011 by Taylor and Francis Group, LLC

 Problems 265

2.15. Proceeding from the rule ∆I = 1/2, prove the correctness of the following relations
between three-pion decay probabilities of charged and neutral kaons:
3
Γ(KL → 3π 0 ) = Γ(KL → π + π − π 0 ), Γ(K + → π + π − π + ) = 4Γ(KL → π + π 0 π 0 ).
2
NOTE: In obtaining the given relations, one has assumed all three π-mesons to be in the
S relative state. Then any pair of π-mesons must be in a symmetric isospin state, that is,
either in the state with I = 0, or in the state with I = 2, or in the mixed state with I = 0
and I = 2. The third π-meson with I = 1 must be arranged with that pair so as in the case
of the K + -decays the resultant three-pion state has I = 1 and I3 = 1 while in the case of
the K − -decays it has I = 1 and I3 = 0].
0
2.16. Taking into account the parities of K 0 - and K -mesons to be negative, build from
them two states with a definite CP -parity.
2.17. Under K 0 -meson decays an uncommon phenomenon is observed. One half of the
K 0 -mesons decays during the short time τS through the channels:

K 0 → π− + π+ , K 0 → π0 + π0 ,

while the latter half of the K 0 -mesons does during the time τL = 651τS = 5.77 × 10−8 s
through the channels:

K 0 → π0 + π− + π+ , K 0 → π0 + π0 + π0 .

Being based on the CP -parity conservation law, explain this phenomenon.

2.18. (Regeneration of K10 -mesons) Consider the K 0 -mesons beam moving in a vacuum
during the time much bigger than the K10 lifetime. Assume this beam hits on a target
consisting of heavy nuclei with the mass M that is much more than the incoming particles
energy. The number of nuclei per the target length unit is equal to N . Find the K10 -mesons
production probability on coming out of the target. Consider contributions of different
target nuclei to be added coherently.
2.19. Being based both on the U -spin conservation law and on the energy reasons, deter-
mine which of the reactions mentioned below

∆− → Σ− + K 0 , Σ∗− → Σ− + π 0 ,

Ξ∗− → Ξ− + η, Ξ∗− → Ξ− + π 0
are allowed and find the relations between allowed reaction amplitudes.
2.20. If one considers a photon as a scalar relative to the U -spin, then which of the following
decays are allowed in the SU (3)-symmetry scheme

Ξ∗− → Ξ− + γ, Σ∗− → Σ− + γ,

∆+ → p + γ, Σ∗+ → Σ+ + γ.
2.21. Allowing for the quark filling of a meson octet 0−+ and setting mu = md =, show
the validity of the following mass relation

2(mK + mK ) = 3mη + mπ .

2.22. Taking into account the quark filling of a baryon decuplet 3/2+ and setting mu =
md =, obtain the equal intervals rule

mΩ − mΞ = mΞ − mΣ = mΣ − m∆ .

© 2011 by Taylor and Francis Group, LLC

 266 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

2.23. Taking into account the quark filling of a baryon octet 1/2+ and setting mu = md =,
prove the validity of the relation

2(mN + mΞ ) = 3mΛ + mΣ .

2.24. What amount of gluons could appear under annihilation of neutral mesons 0−+ and
1−− ?
2.25. Making use of the parton model, find the ratio of the total cross sections of πN -
and N N -interactions and compare it with the experimental value. Think the quark and
antiquark total cross sections to be connected with the relations

σqq = σqq = σqq .

2.26. Taking into consideration the quark filling of vector ρ-, ω- and φ-mesons
1

ρ = √ uu − dd ,
2
1

ω = √ uu + dd , φ = ss,
2
compute the probabilities ratio of their decays through the e− e+ -pair. Under computations
use the Feynman diagram for the subprocess

q + q → γ ∗ → e− + e+ .

Compare results obtained with the experimental data.

2.27. Find a quark wavefunction of a proton with the up spin.
2.28. Compute proton and neutron magnetic moments in the quark model.
2.29. Consider a electron-positron bound state, a positronium. Analyze possible decay
channels of the positronium S-state with the spin 0 (parapositronium) and 1 (orthopositro-
nium).
2.30. Using spectroscopic designations (n2S+1 Lj ) propose a classification of low energy
levels for charmonium and bottonium. Find quantum characteristics J P C of these states
and establish a correspondence of charmonium and bottonium to experimentally observable
states of J/Ψ- and Υ-families.
2.31. A collision of a high-energy electron with a proton is described by an energy ν,
transmitted to a proton in the proton rest frame, and a quantity q 2 = p2 − ν 2 , where p
is a momentum transmitted to the proton. Neglecting the electron mass, show that the
following relations:
q 2 = 2EE ′ (1 − cos θ), ν = |E ′ − E|,
where E and E ′ is the electron energy before and after collision and θ is a scattering angle,
are fulfilled.
2.32. Using the previous problem result consider the interaction of a relativistic electron
with a proton in the center-of-mass frame. Consider that a momentum of each particle is
equal to p and the quantities ν and q are known. Find the energy transmitted to the proton
in the center-of-mass frame. Neglect the electron mass and think that p ≫ M , where M is
the proton mass. Estimate the value of the electron interaction time with the proton.
2.33. An electron flies at a proton having a momentum p in the center-of-mass frame.
Calculate the effective mass of a quark entering into the proton at the quark interaction
with the electron. Base calculations on the following assumptions. Before collision the quark
possessed the transverse momentum p⊥ to be perpendicular to the proton center-of-mass

© 2011 by Taylor and Francis Group, LLC

 Problems 267

direction. Under flying out the quark gets the longitudinal proton momentum xp, where
0 < x < 1. By definition
mef f = p∆E,
where ∆E is an energy spent on the free quark production. Making calculations, assume

p2 x2 ≫ m2q + p2⊥ , p2 (1 − x)2 ≫ M + p2⊥ ,

where M is the proton mass.

2.34. Using the parton model, show at what conditions one may consider a quark to be
free under its interaction with a high-energy electron. Think the effective quark mass to be
known.
2.35. Determine the energy distributions of muons and neutrinos resultant the decay

π + → µ+ + νµ

in the reference frame where π + -meson is moving with a momentum p. Take into account
that in the π + -meson rest frame the angle distribution of muons (neutrinos) is isotropic.
2.36. What distance does a beam µ, π + , π 0 , K + , n pass in a vacuum before its intensity
is halved? Take on the neutron energy to be equal to 14 MeV and the rest of particles
energy—1 GeV.
2.37. Proton and electron circular accelerators have the equal radius 350 m. Estimate the
maximum allowable energy (neglecting synchrotron emission losses) electrons and protons
could get under H = 20 kGs. What are synchrotron emission losses at one rotation under
maximum allowable energy of particles?
2.38. Provide a comparative characteristic of proton accelerators: (i) accelerator with col-
liding beams (proton and antiproton currents ip = ip = 2.8 mA; clusters rotation frequency
f = 104 Hz; the number of clusters on an orbit N = 1; a beam size rx = ry = 0.02 mm;
momenta pp = pp = 32 GeV/c); and (ii) accelerator with a fixed target (a momentum
pp = 400 GeV/c, a proton beam intensity I = 1012 /, a target—Be, ρ = 1.85 g/cm3 , a
beam target size—10 cm).
Compare: carrying out experiments time; an energy available for new particles produc-
tion; possibilities for generating new particle beams; and possibilities for detecting collision
products.
2.39. Evaluate the speed of a set of statistics in a process of the Z-boson resonance pro-
duction at e− e+ -collider under the following conditions: electron and positron currents are
ie = ie = 3 mA, clusters rotation frequency is f = 104 Hz, the number of clusters to be
simultaneously found on an orbit equals N = 1, and a beam cross section is S = 0.04 mm2 ,
mZ = 91.2 GeV/c2 . In order to solve the problem use the Breit-Wigner formula

gπ Γ(Z → e+ e− )Γ(Z → µ+ µ− )
σ= ,
p2 (E − mz )2 + Γ2tot /4

where
2JZ + 1
g= , E = Ee− + Ee+ ,
(2Je− + 1)(2Je+ + 1)
JZ,ee are spins of Z, e− , e+ , Ee− (Ee+ ) is the electron (positron) energy, Γ(Z → e+ e− ) and
Γ(Z → µ+ µ− ) are the partial widths of decays, Γtot is the Z-boson total decay width being
equal to 2.49 GeV and Γ(Z → e+ e− ) = Γ(Z → µ+ µ− ) = 0.0336 Γtot .
2.40. At colliding e− e+ -beam rings the reaction

e− + e+ → π + + π −

© 2011 by Taylor and Francis Group, LLC

 268 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

is investigated. It is ascertained that the total cross section takes on a value 1.8 × 10−30 cm2
near the resonance e− + e+ → ρ → π + + π − . Estimate the amount of such events for one
hour when the ring radius is r = 10 m, the electron and positron beam currents are i = 10
mA, the cross section area of the beams is S = 0.1 cm2 and at every rotation the electron
and positron clusters are collided twice (N = 2).

© 2011 by Taylor and Francis Group, LLC

 References

[1] J. Schwinger, Ann. Phys. 2, 407 (1957).

[2] P. Higgs, Phys. Lett. B12, 132 (1964).
[3] P. Langacker, Phys. Repts. 72, 185 (1981).
[4] H. Georgi, H. Quinnn, S. Weinberg, Phys. Rev. Lett. 33, 451 (174).
[5] H. Georgi, and S. L. Glashow, Phys. Rev. Lett. 32, 438 (1974).
[6] H. Georgi, Particles and Fields, (AIP, New York, 1975); H. Fritzcsh and P.
Minkowski, Ann. Phys. 93, 193 (1975).
[7] J. Wess and J. Bagger, Supersymmetry and Supergravity, (Princeton University
Press, Princeton, New Jersey, 1983).
[8] Supergravity’81, Proc. the 1st School on Supergravity, 22 April-6 May, 1981,
Trieste, Italy (ed. by S. Ferrara and J. G. Taylor).
[9] Th. Kaluza, Zitzungsber. Preuss. Acad. Wiss. Berlin, Math. Phys. K1, 966
(1921).
[10] K. Klein, Zeit. Phys. 37, 895 (1926); Nature, 118, 516 (1926).
[11] A. Einstein, P. Bergmann, Annals Math. 39, 683 (1938).
[12] M. Green, J. Schwartz and E. Witten, Superstring theory, (Cambridge Univer-
sity Press, Cambridge, 1987).
[13] M. Green and J. Schwarz, Phys. Lett. B151, 21 (1985).
[14] J. Polchinski, Phys. Rev., Lett. 75, 4274 (1995); E. Witten, Nucl. Phys., B443,
85 (1995).
[15] N. F. Mott, H.S. W. Massay, The Theory of Atomic Collisions, (Clarendon,
1965).
[16] J. Chadwick, Nature 129, 312 (1932).
[17] D. Ivanenko, Nature 129, 798 (1932).
[18] W. Heisenberg, Zeit. Phys. 77, 1 (1932).
[19] T. Regge, Nuovo Cimento 14, 957 (1959).
[20] G. Chew, S. Frautschi, Phys. Rev. Lett. 5, 12 (1960).
[21] P. D. Collins and E. J. Squires, Regge Poles in Particle Physics, (Springer-
Verlag, New York, 1968).
[22] G. Breit, E. P. Wigner, Phys. Rev. 49, 519 (1936); 642 (1936).
[23] H. Yukawa, Proc. Phys. Math. Soc. Japan, 17, 48 (1935).
[24] Y. Neeman, Nucl. Phys. 26, 222 (1961).

269
© 2011 by Taylor and Francis Group, LLC
 270 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

[25] M. Gell-Mann, Phys. Rev. 125, 1067 (1962).

[26] S. Okubo, Progr. Theor. Phys. 27, 949 (1962).
[27] S. Coleman, S. L. Glashow, Phys. Rev. Lett. 6, 423 (1961).
[28] M. Gell-Mann, Phys. Lett. 8, 214 (1964).
[29] G. Zweig, CERN preprints 8182/Th. 401 and 8419/Th. 412, 1964.
[30] R. P. Feynman, Phys. Rev. 76, 769 (1949).
[31] M. N. Rosenbluth, Phys. Rev. 79, 615 (1950).
[32] R. G. Sachs, Phys. Rev. 126, 2256 (1962).
[33] R. E. Taylor, in Proc. 1975 Int. Symp. on Lepton and Photon Interactions at
High Energies, (ed. W.T.Kirk), (Stanford, California, 1975).
[34] J. R. Dunning at al. Phys. Rev. 141, 1286 (1966).
[35] J. D. Bjorken, Phys. Rev. 179, 1547 (1969).
[36] R. P. Feynman, Phys. Rev. Lett. 23, 1415 (1969).
[37] R. P. Feynman, Photon-hadron interaction, (Reading, Massachusets, Benjamin,
1972).
[38] C. G. Jr. Callan, D. I. Cross, Phys. Rev. Lett. 22, 156 (1969).
[39] O. W. Greenberg, Phys. Rev. Lett. bf13, 598 (1964); M. Han, Y. Nambu, Phys.
Rev. B139, 1006 (1965).
[40] H. Fritzsch, M. Gell-Mann, H. Leutwyler, Phys. Lett. B47, 365 (1973).
[41] C. N. Yang, R. L. Mills, Phys. Rev. 96, 191 (1954).
[42] A. Chodos, R. L. Jaffe, K. Johnson, C. B. Thorn, V. E. Weiskopf, Phys. Rev.
D9, 3471 (1974); W. Bardeen, M. S. Chanowitz, S. D. Drell, M. Weinstein, T.
M. Yan, Phys. Rev. D11, 1094 (1974).
[43] K. G. Wilson, Phys. Rev. D10, 2445 (1974).
[44] M. Creutz, Quarks, Gluons and Lattices, (University Press, Cambridge, 1983).
[45] S. Okubo, Phys. Lett. 5, 165 (1963); G. Zweig, CERN preprint Th/401, 1964;
J. Iizuka, K. Okada, O. Shito, Prog. Theor. Phys. 35, 1061 (1965).
[46] S. L. Glashow, Nucl. Phys. 22, 579 (1961); S. Weinberg, Phys. Rev. Lett.
19, 1264 (1967); A. Salam, in Elementary particle physics, (ed. N.Svartholm),
(Almqvist and Wilsell, Stockholm, 1968).
[47] G. S. Larue et al., Phys. Rev. Lett. 38, 1011 (1977); 42, 142 (1979).
[48] Y. Nambu, Phys. Rev. Lett. 4, 380 (1960); J. Goldstone, Nuovo Cimento 9,
154 (1961).
[49] B. Carter, in Proceedings of the International Astronomical Union Symposium
63: Confrontation of Cosmological Theories with Observational Data, (ed. M.
S. Longair), (D. Reidel, Boston, 1974), p. 291.
[50] R. N. Mohapatra, J. C. Pati, Phys. Rev. D11, 566 (1975); R. N. Mohapatra,
G. Senjanovic, Phys. Rev. D23, 165 (1981).

© 2011 by Taylor and Francis Group, LLC

 References 271

[51] O. M. Boyarkin, Phys. Rev. D50, 2247 (1994).

[52] G. Senjanovic, Nucl. Phys. B153, 334 (1979).
[53] R. W. Assmann et al., “A3-TeV e+ e− linear collider based on CLIC technol-
ogy,” SLAC-REPRINT-2000-096.

© 2011 by Taylor and Francis Group, LLC

 Part III

Quantum Idyll—Free Fields

273
© 2011 by Taylor and Francis Group, LLC
 14
Scalar field

Bruin the First had a long career behind him;

he could build a den and root up trees; hence, with certain
allowances, he could pass for an expert engineer. His greatest
merit, however, was his ardent desire to figure in the annals
of History. For this reason he preferred the glories of
bloodshed to everything else world. And so, whatever he
spoke of, be it commerce, industry or science, he always
ended the same: “Blood, sirs, blood, that’s what we need”.
M.E. Saltykov Shchedrin, “Bears in Government”

14.1 Klein-Gordon equation

Interest in a free-field theory is by no means purely academic. Before and after the particle
collision in the process of scattering (known as the key experiment in the microworld)
the distance between these particles is so great that their interaction is practically zero.
Because of this, we can consider the particles at the initial and final states to be free. To
describe such a physical situation, an adequate mathematical operation—the representation
of interaction—may be used for going from equations of interacting fields to the free-field
equations. Of course, we have to do with a certain idealization; there is no possibility
to exclude interactions all together, while interactions between the scattered particles are
excluded. In other words, exact equations for particles at the initial and final states have
to include the terms determining the interactions with the environment.
Let us consider the simplest example of a relativistically invariant wave equation

( + m2 )ψ(x) = 0, (14.1)

proposed by E. Schrödinger in 1926 [1] along with his famous non-relativistic equation.
Later this equation was considered by Gordon [2], Klein [3], Fock [4], and Kudar [5]. Now
one is customary to call Eq. (14.1) a Klein-Gordon equation. From the beginning, there was
a tendency to use Eq. (14.1) for the description of a relativistic electron. But later with a
deeper insight into the problem it was accepted that the equation is associated with spinless
particles and should be related to scalar and pseudoscalar mesons (π ± , π 0 , K ± , K 0 , ...).
After the discovery of a compound hadron structure, a role of the Klein-Gordon equation
was reduced to the description of Higgs bosons. It should be noted that in the majority of
SM extensions, the Higgs sector includes neutral as well as charged scalar and pseudoscalar
bosons. Owing to its remarkable simplicity, the Klein-Gordon equation still remains a
proving ground for verification of the principal statements of the quantum field ideology.
The wavefunction of a spinless particle ψ(x) is a single-component quantity that in the

275
© 2011 by Taylor and Francis Group, LLC
 276 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

process of inhomogeneous Lorentz transformations

x′ = Λx + a

is transformed as
ψ ′ (x′ ) = ψ(x) (14.2)
or
ψ ′ (x) = ψ(Λ−1 (x − a)). (14.3)
Provided the wavefunction is of positive (negative) parity, Eq. (14.1) describes a scalar
(pseudoscalar) particle. For a truly neutral particle the wavefunction is real, being complex
otherwise. The generalities of the second quantization are most fully reflected for a field
that includes both particles and antiparticles. In this case Eq. (14.1) may be derived from
the Lagrangian density

L(x) = ∂µ ψ ∗ (x)∂ µ ψ(x) − m2 ψ ∗ (x)ψ(x). (14.4)

Using a common procedure, we find the following dynamic variables:

Tµν (x) = ∂µ ψ ∗ (x)∂ν ψ(x) + ∂ν ψ ∗ (x)∂µ ψ(x) − gµν L(x), (14.5)

the energy-momentum tensor,

λ
Mµν (x) = ∂ λ ψ ∗ (x)[xν ∂µ ψ(x) − xµ ∂ν ψ(x)] + ∂ λ ψ(x)[xν ∂µ ψ ∗ (x)−

−xµ ∂ν ψ ∗ (x)] + L(x)(xµ δνλ − xν δµλ ) (14.6)

the angular momentum tensor, and (α = e, s, c, b, ..)

jµ(α) (x) = (ρ(x), j(x)) = iα[ψ ∗ (x)∂µ ψ(x) − ∂µ ψ ∗ (x)ψ(x)] (14.7)

the “current” four-dimensional vector (α = e, s, c, b, ..).

For neutral particles the four-dimensional vector of the electromagnetic current (α = e)
becomes zero. But in the case when particles are not truly neutral, the “current” associated
with a quantum number distinguishing particle and antiparticle is nonzero. For instance,
with the use of Eq. (14.1) for the description of K 0 -meson the current related to the
(s)
quantum number called the strangeness, jµ (x), is different from zero.
According to de Broglie, a free particle is described by a plane monochromatic wave (de
Broglie wave). In consequence, the solution to the Klein-Gordon equation is given by:

φk (x) = A exp (−ikx). (14.8)

Solutions of the plane-wave type will be always understood as a limiting case for the corre-
sponding solutions of the wave-packet type, meeting the orthonormality requirements. For
the solutions of (14.8) orthonormalization is as follows:
Z

i d3 r φ∗k (x)[∂ 0 φk′ (x)] − [∂ 0 φ∗k (x)]φk′ (x) ≡

Z ←→
≡i d3 r[φ∗k (x) ∂ 0 φk′ (x)] = 2ωk δ(k′ − k), (14.9)
Z ←→ Z ←→
i d3 r[φ∗k (x) ∂ 0 φ∗k′ (x)] = i d3 r[φk (x) ∂ 0 φk′ (x)] = 0. (14.10)

© 2011 by Taylor and Francis Group, LLC

 Scalar field 277

It should be noted that the description of a free particle by a plane wave (a complete set
includes the momentum and energy operators) is not unique. For a complete set one may
take the operators of energy, angular momentum square, and angular momentum projection
onto the chosen direction in space. In the latter case the eigenfunction of these operators
and hence the solution for the free Klein-Gordon equation will be given by the quantity:
exp (−iωt + ikr) |mz |
φlmz (r, θ, ϕ) = A Yl (θ, ϕ), (14.11)
r
|m |
where Yl z (θ, ϕ) is the spherical function.
Substituting (14.8) into (14.1), we obtain:
p
E = ω = ± k2 + m2 .
And the appearance of the negative-energy solutions is like a bolt from the blue. Negative
values of the energy in the free case are physically meaningless. Within the scope of classical
physics we could ignore their appearance altogether. Indeed, in the classical physics the
energy of a particle may vary only continuously. Because of this, a particle with a positive
energy at the initial moment of time has no chance of going to the negative-energy state,
due to the 2mc2 interval of “unobservable energies” separating the positive-energy and
negative-energy zones. At the same time, in a quantum theory it is natural to deal precisely
with discrete transitions from one level to the other. And, in particular, the transitions
between the states with negative and positive energies are not forbidden. In this way the
solutions with negative energy could not be simply neglected without violation of the basic
principles inherent in the quantum theory. For example, if we would discard a partial
solution associated with the negative energy and retain only the positive-energy solution,
then the obtained system of functions will be incomplete.
There is one more difficulty connected with the negative probabilities. Let us derive
a relativistic equation of continuity. To this end, we write an equation for the complex-
conjugate wave function:
( + m2 )ψ ∗ (x) = 0. (14.12)
Multiplying (14.1) by ψ ∗ (x) on the left, multiplying (14.12) by ψ(x) on the right, and then
subtracting one equation from the other, we have:
∂ µ jµ (x) = 0,
where the probability current is determined by the following expression:
jµ (x) = ψ ∗ (x)∂µ ψ(x) − ψ(x)∂µ ψ ∗ (x).
It is obvious that the probability density
j0 (x) ≡ ρ(x) = ψ ∗ (x)∂t ψ(x) − ψ(x)∂t ψ ∗ (x)
is not always positive. Since Eq. (14.1) involves the second derivative with respect to
time, in a particular instant of time ψ(x) and ∂t ψ(x) may be given both independently and
arbitrarily, that may yield negative values of ρ(x). Besides, it is clear that if ρ(x) will be
greater than zero for the states with a positive energy, then the solutions with negative
energies result in a negative probability density. Thus, there is no way for interpretation
of the Klein-Gordon equation as a single-particle equation with a wavefunction ψ(x), that
is, we have to turn down the ideology of a nonrelativistic (primary quantized) theory.
The difficulties, resultant from the appearance of the physically-meaningless solutions with
negative energies, are due to “illegality” of considering the problem of a single-particle
motion at relativistic energies. At such energies we have to deal with the particle-antiparticle
pair production (and their annihilation), and this problem concerns a variable number of
the particles, that is, we have to solve the problem using methods of a quantum field theory.

© 2011 by Taylor and Francis Group, LLC

 278 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

14.2 Quantization
In 1934 the Klein-Gordon equation was reanimated by Pauli and Weisskopf who have used
the second quantization method. By this method, the field functions should be considered as
the operators governed by certain commutation relations rather than as c-numbers. These
operators influence the state vector Φ that is describing the state of a field as a particular
quantum-mechanical system with a variable number of the particles. In nonrelativistic
quantum mechanics (NQM) the wavefunction of a system may have been specified in any
representation, no matter coordinate, momentum, or of some other type. In a second
quantized theory the state vector is determined in the space of the particle numbers (quanta
of a given field) or, simply speaking, in the occupation numbers space (Fock space).
Recall how the states of physical systems are described in NQM. As is known, two methods
may be used. In the first method, the operators relating to the physical quantities are
assumed to be independent of time, whereas the state wavefunction is considered variable
in time, its variation being described by the Schrödinger equation. The time dependence of
the average values of the physical quantities appears only through the time dependence of
the state wavefunction according to the formulae
Z
f (t) = Ψ∗ (q, t)fˆΨ(q, t)dq.

This representation was named the Schrödinger representation. However, the quantum
mechanical apparatus may be equivalently formulated in some other way, carrying over the
time dependence from the wavefunctions to the operators (Heisenberg representation). The
state vector and operators of the both representations are connected by the relations:
ΨS (q, t) = ŜΨH (q, 0), fˆS = Ŝ fˆH Ŝ −1 ,
where  
i
Ŝ = exp − Ĥt .
~
Formally, equations of motion for the operators in Heisenberg’s representation are similar in
form to the classical mechanics equations for the corresponding quantities. On account of
this reason, the procedure of second quantization is conveniently considered in Heisenberg’s
representation.
Let the field described by the function ψ(x) be confined within a cube of volume V = L3 ,
where L is the cube edge length. We expand the operator field functions in a complete set
of solutions for the Klein-Gordon equation:
1 X 1
ψ(x) = √ √ [a(+) (−)
k exp [i(kr) − iωt] + ak exp [i(kr) + iωt]], (14.13)
V k 2ω
1 X 1
ψ ∗ (x) = √ √ [a(+)∗
k
(−)∗
exp [−i(kr) + iωt] + ak exp [−i(kr) − iωt]], (14.14)
V k 2ω
(+) (−)
where the quantities ak and ak are operators,∗ and ψ (+) (x) (ψ (−) (x)) represent a sum
of the solutions with a positive (negative) frequency, satisfying the orthonormalization con-
ditions: Z ←→
(q)∗ (q′ )
i d3 r[ψk (x) ∂ 0 ψk′ (x)] = 2ωk δqq′ δk′ k . (14.15)

∗ Hereinafter, the cap over operators that is usual for the NQM will be omitted.

© 2011 by Taylor and Francis Group, LLC

 Scalar field 279

Now we find the permutation relations for the operator field functions. As we still believe
in the unity of nature, a change over to quantum field theory should be a natural con-
tinuation of that thread, which has led us to the NQM from classical physics. Recall that
dynamic variables in the NQM have acquired the status of operators, and their commutation
relations were found with the use of quantum Poisson brackets:
X  ∂a ∂b ∂a ∂b

′
[a, b] = − , (14.16)
r
∂qr ∂pr ∂pr ∂qr

where a and b are arbitrary operators. The Poisson brackets in turn were simply related to
the commutator
[a, b]′ = −i[a, b]. (14.17)
Specifically, for the coordinate and momentum operators the relations were as follows:
[qr (t), ks (t)] = iδrs , (14.18)
where we have written down the time argument for the operators to emphasize that they
were determined at the same instant of time. In quantum field theory, the wavefunctions are
playing the role of the coordinates, and the canonically conjugate momenta are determined
with the help of the Lagrangian just the same way as in the classical physics
∂L ∂L
π(x) = = ∂0 ψ ∗ (x), π ∗ (x) = = ∂0 ψ(x). (14.19)
∂(∂ 0 ψ(x)) ∂(∂ 0 ψ ∗ (x))
Quite logically it may be assumed that the commutative law of the operator field func-
tions in its form is similar to (14.18). It should be remembered, however, that the field
functions are dependent on the spin, leading to different statistics for fermions and bosons.
Without proof we accept the Pauli theorem that relates the transformational properties of
a field and its quantization method: “The fields describing the integer-spin particles are
quantized with the use of commutators (Bose-Einstein quantization); whereas the fields de-
scribing the half-integer-spin particles are quantized using anticommutators (Fermi-Dirac
quantization).” Note that for bosons as well as fermions the permutation relations of the
field functions are c-numbers.
In case of a spinless particle, as a field analogue of (14.18), we have the commutation
relations:
[ψ(r′ , t), π(r, t)] = [ψ(r′ , t), ∂0 ψ ∗ (r, t)] = iδ(r′ − r), (14.20)
[ψ ∗ (r′ , t), π ∗ (r, t)] = [ψ ∗ (r′ , t), ∂0 ψ(r, t)] = iδ(r′ − r). (14.21)
All other commutators from the pairs of “coordinate” and “momentum” operators are zero.
The above commutation relations were determined at equal times only. Under incoincident
moments of time their determination necessitates the knowledge of solutions for the motion
(+) (−)
equations. Using (14.20), we may find the commutators of the operators ak and ak .
Substitute the expansions (14.13) and (14.14) into (14.20)
i X ω ′  (+) (+)∗
[ψ(r, t), ∂0 ψ ∗ (r′ , t)] = √ [ak , ak′ ]×
V 4ωω ′
kk′

(−) (−)∗
× exp {i[(kr) − (k′ r′ ) − (ω − ω ′ )t]} − [ak , ak′ ] exp i{[(kr) − (k′ r′ ) + (ω − ω ′ )t]} ,
(14.22)
(+) (−)
where we have took into account that ak and ak are describing the independent degrees
of freedom of a field and hence must commutate
(+)∗ (−)∗ (+) (−)
[ak′ , ak ] = [ak′ , ak ] = 0. (14.23)

© 2011 by Taylor and Francis Group, LLC

 280 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It is easily seen that to obtain the delta function on the right-hand side of (14.22), the
following quantization rules should be applied:
(+) (+)∗ (−) (−)∗
[ak , ak′ ] = −[ak , ak′ ] = δkk′ . (14.24)

Actually, in this case the relation (14.22) takes the form:

i X
[ψ(r, t), ∂0 ψ ∗ (r′ , t)] = exp {i[(kr) − (kr′ )}. (14.25)
V
k

To exchange integration with summation in the right-hand side of (14.25), we should re-
member that the condition of vanishing the field function and its first space derivative at
the boundaries of a cube with the edge L results in the following relations:

ki L = 2πni , i = x, y, z.

Then the total number ∆N of k-vectors over the interval d3 k is equal to

V d3 k
∆N = .
(2π)3
Provided the volume V is large, and the functions under the summation sign are slowly
varying, a change from summation to integration is realized according to
Z
1 X 1
→ d3 k. (14.26)
V (2π)3
k

By substitution of (14.26) into the right-hand side of (14.25) with subsequent comparison
with (14.20) we are convinced that the commutation relations (14.24) are valid.
The above-stated quantization method is referred to as a canonical quantization. With
this formalism the time and space coordinates are not equal in rights. This circumstance
casts some doubt on the relativistic invariance of a second-quantized theory based on this
approach. But in actual fact the doubts are unfounded, and by the canonical field quanti-
zation we obtain the expressions possessing the relativistic invariance. At the same time,
the canonical formalism may be “shaped” into the covariant form without difficulties. And
as the reader could guess, this may be performed using the same method that has provided
a relativistic covariance for the evolution equation of the state vector in Chapter 1. We
revert to this problem at the end of Section 14.5.

14.3 Production and destruction operators

The next step of the second quantization procedure for a field should be establishing a
(+) (−)
physical meaning for the operators ak and ak . We introduce the quantities Nk+ =
(+)∗ (+) (−)∗ (−)
ak ak and Nk = ak ak . An equation for the eigenvalues of the operator Nk+ is of
−

the form:
(+)∗ (+)
ak ak Φn+ = n+ k Φn+ . (14.27)
k k

In Eq. (14.27) Φn+ is the eigenfunction meeting the normalization condition:

k
Z
Φ∗n+ Φn+ dτ ≡ (Φn+ , Φn+ ) = 1, (14.28)
k k k k

© 2011 by Taylor and Francis Group, LLC

 Scalar field 281

where integration is performed over all the variables on which Φn+ is dependent. From Eq.
k
(14.27) it follows that:
(+)∗ (+) (+) (+)
n+
k = (Φn+ , ak ak Φn+ ) = (ak Φn+ , ak Φn+ ) ≥ 0. (14.29)
k k k k

(−1) (+)
Let us define the function Φn+ = ak Φn+ . For n+
k > 0 this function is nonzero as
k k

(+)∗ (−1) (+)∗ (+)

ak Φn+ = ak ak Φn+ = n+
k Φn+ 6= 0. (14.30)
k k k

It is seen that it is the eigenfunction of the operator Nk+ :

(−1) (+)∗ (+) (+) (+) (+)∗ (+)
Nk+ Φn+ = ak ak ak Φn+ = (ak ak − 1)ak Φn+ =
k k k

(−1)
= (n+
k − 1)Φn+ . (14.31)
k

The relation (14.31) makes it possible to write

(+)
ak Φn+ = αΦn+ −1 , (14.32)
k k

where α is a constant whose form will be stated later. In similar way, we can show that
the operator Nk+ also has n+ +
k − 2, nk − 3,... as its eigenvalues. But it is obvious that such
a series should be broken because negative eigenvalues n+ k are impossible. This is possible
only in the case when n+k represents a series of positive integers, n+
k = 0, 1, 2, ... Clearly,
these numbers may be as great as you like. For demonstration we consider the following
functions:
(1) (+)∗ (2) (+)∗ (+)∗ (3) (+)∗ (+)∗ (+)∗
Φn+ = ak Φn+ , Φn+ = ak ak Φn+ , Φn+ = ak ak ak Φn+ , .....
k k k k k k

(1)
For Φn+ we have an equation for the eigenvalue:
k

(1) (+)∗ (+) (+)∗ (+)∗ (+)∗ (+)

Nk+ Φn+ = ak ak ak Φn+ = ak (1 + ak ak )Φn+ =
k k k

(1)
= (n+
k + 1)Φn+ . (14.33)
k

Then it is readily seen that the eigenvalues of the operator Nk+ are represented by a numer-
ical series n+ + +
k + 1, nk + 2, nk + 3..... Also, from (14.33) it follows that

(+)∗
ak Φn+ = α′ Φn+ +1 , α′ = const. (14.34)
k k

(−1) (1)
The properties of the functions Φn+ and Φn+ enable one to find the matrix elements for
k k
(+) (+)∗
the operators ak and ak :

(+) (−1) (+)
< n+′
k | ak | n+
k >≡ (Φn+ , ak Φn+ ),

k
(+)∗
k
(1) (+)∗ (14.35)
< n+′′
k | ak | n+
k >≡ (Φn+ , ak Φn+ ). 
k k

Clearly, these matrix elements are nonzero only in the case when n+′ + +′′
k = nk − 1 and nk =
+
nk + 1. From (14.35) it follows immediately that:
q
(+)
< n+k − 1 | a k | n +
k >= n+
k, (14.36)

© 2011 by Taylor and Francis Group, LLC

 282 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
q
(+)∗
< n+ k + 1 | a k | n +
k >= n+
k + 1, (14.37)
where we ignore the phase term that is unity in absolute value. Now we can find the
constants α and α′ , that leads to:
q q
(+) (+)∗
ak Φn+ = n+ Φ +
k n −1 , a k Φ n+ = n+
k + 1Φn+ +1 . (14.38)
k k k k

(+)∗ (+)
From Eqs. (14.38) the physical meaning of the operators ak and ak becomes quite
transparent. Let the field under consideration contain n+
k positive-energy particles with
the momentum k. If Φn+ denotes the field state vector, then, according to (14.38), the
k
(+)∗
operation of ak reduce to increasing the quanta number n+ k by 1, whereas the operation
(+) + (+)∗
ak —to decreasing the quanta number nk by 1. In this way ak represents the pro-
(+)
duction operator of a particle with the momentum k and the positive energy ω, and ak
represents the annihilation operator of the same particle. It should be emphasized that
in the quantum theory, as distinct from the classical one, a particle is not understood as
something localized in space. Here it means nothing else but field excitations. But these ex-
citations are discrete objects satisfying the proper relativistic relations between the energy
and momentum. Therefore, they are termed as particles.
Let us return back to expressions (14.13) and (14.14). The form of the terms involved
in the first sums of both expressions enables one to infer that annihilation of the parti-
cles having the positive energy is associated with the time factor exp (−iωt), and their
production—with exp (iωt). Then the way to reconstruction of the remaining terms be-
comes quite obvious. To make their form meeting the norms of a quantum behavior, it is
(−) (−)∗
necessary to interpret ak as a production operator and ak —as an annihilation operator
for some other particles, which, as it is found out later, are antiparticles. Consequently,
(−) (−)∗
ak is replaced by the antiparticle production operator b∗−k , whereas ak —by the an-
tiparticle annihilation operator b−k . Also, for simplicity, we omit the frequency index of
(+) (+)∗
the operators ak , ak . Replacing the notation for the summation variable k by −k (in
so doing the exponential factor takes the relativistically-covariant form) in the second sums
of expressions (14.13) and (14.14), we have the following expansion of the field functions:

1 X 1
ψ(x) = ψ (+) (x) + ψ (−) (x) = √ √ [ak exp [i(kr) − iωt] + b∗k exp [−i(kr) + iωt]],
V k 2ω
(14.39)
∗ (+)∗ (−)∗ 1 X 1 ∗
ψ (x) = ψ (x) + ψ (x) = √ √ [ak exp [−i(kr) + iωt] + bk exp [i(kr) − iωt]],
V k 2ω
(14.40)
where the operators ak and bk satisfy the commutation relations, this time symmetric

[ak , a∗k′ ] = δkk′ , (14.41)

[bk , b∗k′ ] = δkk′ , (14.42)

Thus, all the operators ak , bk become multiplied by the exponents having the “proper” time
dependence (∼ exp (−iωt)), that is, all the solutions obtained have the “correct” energy
sign. As a result, in “charged” fields we come to the understanding that two kinds of the
particles (particles per se and antiparticles) are possible, which are acting jointly and on
equal footing. It is evident that the sign changes in the exponent exp (iωt) (going to the
solutions with the positive energy) may be attained by varying the direction in the course of
time. In other words, a solution with negative energy provides the description for a particle

© 2011 by Taylor and Francis Group, LLC

 Scalar field 283

moving backwards in time or, equivalently, an antiparticle with the positive energy, whose
motion in time is in the forward direction. Note that from the mathematical viewpoint,
the relations (14.39) and (14.40) represent nothing else but the Fourier series expansion of
the ψ(x)- and ψ ∗ (x)-functions. Possibly, an enlightened reader has already guessed that
expressions (14.39) and (14.40) may be derived on the basis of the relativistic covariance
requirement. The Lorentz transformations, the rotation of a four-dimensional coordinate
system, do not take the axis t beyond the boundaries of the corresponding light-cone cavity,
which is a manifestation of the existence of the limit signals propagation velocity. But
from the pure mathematical viewpoint, as a rotation we can also consider the four-inversion
(P T -transformation):
r → −r, t → −t, (14.43)
because the determinant of this transformation equals +1. However, considering that in
this case the time axis is shifted from one cavity of the light cone into the other, such a
transformation is physically unrealizable. Then any expression invariant under the Lorentz
transformations should be invariant under the four-inversion. For the operator of a scalar
field this implies:
ψ(r, t) = ψ(−r, −t), (14.44)
and that is only possible when a charge conjugation is performed in ψ(−r, −t), meaning
interchangeability of particles and antiparticles:

ak → b∗k , bk → a∗k . (14.45)

A set of (14.43) and (14.45) is nothing else but CP T -transformation. To obtain expansion
of the field functions for a free scalar field in terms of the solution of the Klein-Gordon
equations in the form of (14.39) and (14.40), it is enough to demand these expansions be
meeting the CP T -invariance requirement.

14.4 Wave-corpuscle dualism

The commutation conditions imposed on the operators a(k) and b(k) have to result in
the adequate corpuscular pattern of a scalar field. This implies that the dynamic field
invariants (energy, momentum, angular momentum square, angular momentum projection,
electric charge, etc.) should be equal to a sum of the dynamic invariants for separate
particles, quanta of the scalar field. Provided these requirements are met, we will have a
direct support for the validity of the derived commutation relations.
We assume that Eq. (14.1) describes a field of the charged particles and then restrict
ourselves to finding the energy, momentum, and electric charge of the field. From (14.5)
the energy is determined as:
Z Z X
E = T d x = d3 x[
00 3
∂µ ψ ∗ (x)∂µ ψ(x) + m2 ψ ∗ (x)ψ(x)] =
µ

1 XZ 1
= d3 x{ √ [∂µ ψ (+)∗ (k)∂µ ψ (+) (k) + ∂µ ψ (+)∗ (k)∂µ ψ (−) (k)+
V 4ωω ′
kk′

+∂µ ψ (−)∗ (k)∂µ ψ (+) (k) + ∂µ ψ (−)∗ (k)∂µ ψ (−) (k) + m2 [ψ (+)∗ (k)ψ (+) (k)+
+ψ (+)∗ (p)ψ (−) (k) + ψ (−)∗ (k)ψ (+) (k) + ψ (−)∗ (k)ψ (−) (k)]},

© 2011 by Taylor and Francis Group, LLC

 284 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where we have introduced the designations


ψ (+) (k) = ak exp [i(kr) − iωt] = ψ (+) (k) exp (−iωt),
(14.46)
ψ (−) (k) = bk exp [−i(kr) + iωt] = ψ (−) (k) exp (iωt).

It may be easily shown that the terms, which contain the products of the functions ψ (±) (k)
and ψ (±)∗ (k) having the same sign of frequency, make no contributions into the dynamic
invariant E. For example,
Z
1 X d3 x
√ [∂µ ψ (+)∗ (k)∂µ ψ (−) (k ′ ) + m2 ψ (+)∗ (k)ψ (−) (k ′ )] =
V ′ 4ωω ′
kk

1 X 1
= √ exp [i(ω + ω ′ )t](m2 − kµ kµ′ )a∗k bk′ ×
V 4ωω ′
kk′
Z
1 X 1
× d3 x exp [−i(k′ + k)r] = √ exp [i(ω + ω ′ )t](m2 − kµ kµ′ )×
V 4ωω ′
kk′
X 1
×a∗k bk′ (2π)3 δ(k + k′ ) = exp (2iωt)(m2 − ω 2 + k2 )a∗k b−k = 0. (14.47)
2ω
k

In the derivation of (14.47) we have allowed for the relation:

(2π)3
δkk′ = δ(k − k′ ), (14.48)
V
the validity of that may be easily demonstrated using a definition of the δ-function and
assuming k = k′ . Finally, for the field energy we have:
X
E= ωk [a∗k ak + bk b∗k ]. (14.49)
k

By the substitution of (14.13), (14.14) into (14.5) and (14.7) we can determine the momen-
tum and electric charge of the field by the following relations:
P )
P = k k[a∗k ak + bk b∗k ],
P (14.50)
Q = k q[a∗k ak − bk b∗k ],

where q is an electric charge of particles. Note that Q is not positively defined.

With the use of the commutation relations we can rewrite the expression for the field
energy (14.49) as:
X X 1 1
E= ωk [a∗k ak + b∗k bk + 1] = ωk [(Nka + ) + (Nkb + )]. (14.51)
2 2
k k

Two very important conclusions may be drawn from (14.51). First, the operators Nka = a∗k ak
and Nkb = b∗k bk are associated with the particles and antiparticles number, respectively.
Second, in (14.51) we recognize a well-known expression from the NQM for the energy of a
two-dimensional harmonic oscillator, which enables us to treat the field under consideration
as a set of the infinite number of harmonic oscillators. Also, expression (14.51) reveals that
the field energy has the lowest value when the particle and antiparticle numbers equal zero.
Such a lowest-energy state of the field is called the vacuum state. Adding a constant to E,
if required, we can shift the reference point for the energy so that a minimum value of the

© 2011 by Taylor and Francis Group, LLC

 Scalar field 285

energy can be zero. Now we assume that there is a single zero-energy state Φ0 ≡| 0 >, the
vacuum state. It should satisfy the conditions:

Di Φ0 = 0, ak Φ0 = 0, bk Φ0 = 0, (14.52)

where Di is a set of the dynamic field invariants. It is supposed that a vector of the vacuum
state is normalized to unity:
(Φ0 , Φ0 ) = 1.
An arbitrary normalized state, involving n+ −
k1 particles with the momentum k1 , nk′ antipar- 1
ticles with the momentum k′1 , n+ −
k2 particles with the momentum k2 , nk′2 antiparticles with
the momentum k′2 and so on, may be written in the following form:

Φn+ − ≡| n+ + − −
k1 , nk2 , ...; nk′ , nk′ .... >=
k1 nk′ ... 1 2
1

1 + + n−′ n−′
=q [a∗k1 ]nk1 [a∗k2 ]nk2 [b∗k′1 ] k1 [b∗k′2 ] k2 ..... | 0 > . (14.53)
n+ + − −
k1 nk2 ...nk′ nk′ 1 2

Considering the commutation relations, the momentum and charge of this field may be
represented as:
X 1 1
P= k[(Nka + ) + (Nkb + )], (14.54)
2 2
k
X
Q= q[(Nka − (Nkb − 1)]. (14.55)
k

As seen from (14.55), the electric charges of particles and antiparticles are opposite in sign.
But if in the quantization process we used anticommutators rather than commutators, the
energy and momentum would not be positively defined, whereas the charges of particles
and antiparticles would not have the opposite signs.
The vacuum average of E: X
(Φ0 , EΦ0 ) = ωk
k

proves to be equal to infinity. Similar infinities are encountered when a vacuum average is
taken for the other dynamics invariants as well. These difficultiesPmay be obviated in two
ways: to subtract an infinite constant (for the energy, for example, k ωk is such a constant)
or to use the freedom of changing the operators order in the expressions for the dynamic
invariants. Taking the latter approach, we write the quantities expressed in terms of the
free-field operators in the form of normal products (N -products). This means an expansion
in terms of the production and annihilation operators with their subsequent reordering: all
annihilation operators should be to the right of all the production ones. Since separation
of the field function ψ(x) into positive and negative frequencies (expansion in terms of the
production and annihilation operators) is relativistically invariant (kµ xµ ≡ kx = inv), the
definition of the normal product is relativistically invariant too. The normal product of
the operators will be denoted either by the symbol N or by the colons that are placed
on each side of the expression containing these operators. Conventionally, all the dynamic
variables quadratically dependent on the operators with identical arguments (for example,
Lagrangian, momentum energy tensor, electromagnetic current, etc.) by definition are
written in the form of the normal product. In this case the Lagrangian density for a scalar
field is of the following form:

L(x) =: ∂µ ψ ∗ (x)∂ µ ψ(x) − m2 ψ ∗ (x)ψ(x) : . (14.56)

© 2011 by Taylor and Francis Group, LLC

 286 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Then a four-dimensional vector of the field momentum and charge contain no infinite ex-
pressions at all:
X
Pµ = kµ [Nka + Nkb ],
k

X
Q= q[Nka − Nkb ].
k

Thus, the dynamic field invariants are actually expressed as sums of the dynamic invari-
ants for the field quanta, that is, the corpuscular properties of a scalar field are adequately
described by the quantization conditions (14.41) and (14.42).
For better understanding of the second quantization procedure we proceed from the
established equivalence between a wave field and a set of harmonic oscillators. All the
results associated with the second quantization procedure—commutation relations (14.41)
and (14.42), matrix elements of the production (14.36) and annihilation (14.37) operators,
expression for the field energy (14.51), and the like—may be derived from the problem for
a harmonic oscillator. For this purpose one should specify the Hamiltonian function:

1 2
H= (p + ω 2 qj2 ), j = a, b,
2 j

where ω is a frequency, pj and qj are momenta and coordinates of the oscillator satisfying
the commutation relations
[qa , pb ] = iδab ,

and, for simplicity, the mass of the oscillator is made equal to unity. And then with the use
of the production and annihilation operators we construct for oscillators their coordinate
and momentum operators as follows:

1 1
ak = √ (pa + iωqa ), a∗k = √ (pa − iωqa ),
2ω 2ω

1 1
bk = √ (pb + iωqb ), b∗k = √ (pb − iωqb ).
2ω 2ω

Next, it is necessary to represent the oscillator in the excited state (Nka,b ≥ 1) as a collection
of excitation quanta, each having the energy ωk . The equidistant energy levels result in
a very important property: impossibility to distinguish between the excitation quanta for
each of the oscillators, of course without reference to the values of other quantum numbers.
Note that the particles are associated with the excitation quanta of the oscillator and not
with the oscillator itself. In other words, the appearance of the particles is due to the
quantization procedure, or more precisely due to quantization of the oscillator energy. The
wavefunctions of the oscillator, the eigenfunctions of the Hamiltonian operator, may be
derived from each other with the use of the operators a∗k and b∗k

p q
a∗k ΦNka = 1 + Nka ΦNka +1 , b∗k ΦNkb = 1 + Nkb ΦNkb +1 .

Thus, the state vector of a system is dependent on the numbers determining a degree of
the quantum states occupation with the field particles, that is, it turns to be given in the
occupation numbers space.

© 2011 by Taylor and Francis Group, LLC

 Scalar field 287

14.5 Commutation relations. Properties of commutator functions

Let us find the commutation relations between the field functions at arbitrary times. As
it has been mentioned previously, this is only possible when we know the solutions for
equations of motion. In the case of a free-field they are given by expressions (14.39) and
(14.40). It should be noted that, if we were interested in the commutation relations for
a particle located, for example, in a homogeneous magnetic field, expansion of the field
function would involve the solutions for the Klein-Gordon equation in a magnetic field
(Laguerre polynomials) rather than de Broglie plane waves. With the use of (14.39) and
(14.40) we obtain:
1 X 1
[ψ(x), ψ ∗ (x′ )] = √ {[ak , a∗k′ ] exp (ik ′ x′ − ikx) + [b∗k , bk′ ] exp (−ik ′ x′ + ikx)} =
V 4ωω ′
kk′

1 X 1
= (exp [−ik(x − x′ )] − exp [ik(x − x′ )]) = −i∆0 (x − x′ ; m2 ), (14.57)
V 2ω
k

where Z
2 2 1 sin ωt 3
∆0 (x; m ) ≡ ∆0 (r, t; m ) = exp (−ikr) d k. (14.58)
(2π)3 ω
As seen, the function ∆0 (x; m2 ) (called Pauli-Jordan function) is a relativistic invariant.
To illustrate, we represent it as:
Z
i
∆0 (x; m2 ) = exp (−ikx)ε(ω)δ(k 2 − m2 )d4 k, (14.59)
(2π)3

where d4 k = dk1 dk2 dk3 dω, ε(ω) = θ(ω) − θ(−ω) and


1, when x > 0,
θ(x) =
0, when x < 0.

It is obvious that all factors under the integral sign, except for ε(ω), are relativistically
invariant. The presence of the delta function δ(k 2 − m2 ) in (14.59) tell us that integration
is carried out on the surface k 2 − m2 = 0. Therefore, the frequency sign (ω)in this case is
relativistically invariant. This suggests a relativistic invariance of the function ε(ω).
From the viewpoint of mathematics, the Pauli-Jordan function is of the same class as the
δ-function, that is, the class of the improper or generalized functions. As distinct from the
normal functions, the generalized ones are determined by setting the integration rules for
their products with sufficiently regular functions rather than by setting a correspondence
between the values of the function and its argument.
It is useful to divide the Pauli-Jordan function into the frequency components:

∆0 (x; m2 ) = ∆+ 2 − 2
0 (x; m ) + ∆0 (x; m ), (14.60)

where Z
i
∆± 2
0 (x; m ) = ± exp (−ikx)θ(±ω)δ(k 2 − m2 )d4 k. (14.61)
(2π)3
They may be used to express the nonzero commutators associated with the negative and
positive frequency components of the field function:

[ψ (+) (x), ψ (+)∗ (x)] = −i∆+ 2

0 (x; m ),

© 2011 by Taylor and Francis Group, LLC

 288 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

[ψ (−) (x), ψ (−)∗ (x)] = −i∆− 2

0 (x; m ).
Later it will be demonstrated that ∆± 2
0 (x; m ) may be used to express not only the Pauli-
Jordan function but other commutator functions as well. Because of this, we begin our
study of the above-mentioned class of functions with the calculations of the integrals (14.61).
Carrying out the integration with respect to k0 and angular variables in space k, we represent
these expressions in the form
1 ∂ 1 ∂ ∗
∆+ 2
0 (x; m ) = F (x), ∆− 2
0 (x; m ) = F (x), (14.62)
4πr ∂r 4πr ∂r
where Z ∞
i exp[i(ωt + kr)]
F (x) = dk, k = |k|, r = |r|.
2π −∞ ω
By this means the problem of finding ∆± 2
0 (x; m ) is reduced to calculation of the function
F (x). Having done the replacement of the variable
k = m sinh y, ω = m cosh y,
we for F (x) obtain
Z ∞
i
F (x) = exp[im(t cosh y + r sinh y)]dy. (14.63)
2π −∞

For four cases possible: 1) t > 0, t > r; 2) t > 0, t < r; 3) t < 0, |t| > r; 4) t < 0, |t| < r,
we use the following substitutions:
√ √ √ √
1) t = √λ cosh y0 , r = √λ sinh y0 , 2) t = √−λ sinh y0 , r = √−λ cosh y0 ,
3) t = − λ cosh y0 , r = λ sinh y0 , 4) t = − −λ sinh y0 , r = −λ cosh y0 ,
where λ = x2µ . Then taking into consideration the integral representation of cylindrical
functions [6], for four different time domains we get:
Z ∞ √ √ √
i 1 i
1) exp[im λ cosh(y + y0 )]dy = − J0 (m λ) − N0 (m λ),
2π −∞ 2 2
Z ∞ √ √
i i
2) exp[im −λ sinh(y + y0 )]dy = K0 (m −λ),
2π −∞ π
Z ∞ √ √ √
i 1 i
3) exp[−im λ cosh(y − y0 )]dy = J0 (m λ) − N0 (m λ),
2π −∞ 2 2
Z ∞
i √ i √
4) exp[−im −λ sinh(y − y0 )]dy = K0 (m −λ),
2π −∞ π
or  1 √ √
 2i N0 (m λ) − 21 ε(t)J0 (m λ) when λ > 0,
F (x) = F (t, λ) = √ (14.64)
 i
π K0 (m −λ), when λ < 0,
where J0 (z) is a Bessel function of the zero order, N0 (z) is a Neumann function of the zero
order, K0 (z) is a Hankel function of an imaginary argument of the zero order. At z ≈ 0
these cylindrical functions are represented by the series:

2 
J0 (z) = 1 − z2 + O(z 4 ), 



h
i 

2 z 2 z 2 2
N0 (z) = π 1 − 2 ln 2 + π C + O(z ), (14.65)


h
2 i 


K0 (z) = − 1 + z2 ln z2 − C + O(z 2 ) 

© 2011 by Taylor and Francis Group, LLC

 Scalar field 289

(C is the Euler constant to be equal to 0.577215...), and in the region of great values of z
they are described by the asymptotic formulae:

J0 (z) = − √2πz sin z, 




2
N0 (z) = √πz sin z, . (14.66)


q 

2 −z
K0 (z) = πz e . 

In (14.62) we proceed from the differentiation with respect to r to that with respect to λ
and take into consideration a jump of the function F (t, λ) at the point λ = 0. Then, having
regard to the following relations for the cylindrical functions

∂
Z1 (x) = − Z0 (x), (14.67)
∂x
where Z = J, N, K, we derive the required expressions:

ε(t)δ(λ) mθ(λ) h √ √ i imθ(−λ) √

∆± 2
0 (x; m ) = − √ ε(t)J1 (m λ) ± iN1 (m λ) ∓ √ K 1 (m −λ).
4π 8π λ 4π 2 −λ
(14.68)
It should be noted that all singularities of the functions ∆± 0 (x; m 2
) are positioned at the
light cone only (λ = 0), as at the space and time infinitiesp these functions, in accordance
with (14.66) and (14.67), are decreased as |λ|−3/4 exp(−m |λ|) and |λ|−3/4 , respectively.
Under investigating a regularity of different commutator functions we need the expansions
of ∆± 2
0 (x; m ) in a light con neighborhood. Using Eqs. (14.65) and (14.67), we arrive at

1 i im2 m|λ|1/2 m2 p
∆± 2
0 (x; m ) = ε(t)δ(λ) ± 2 ∓ 2 ln − ε(t)θ(λ) + O( |λ| ln |λ|). (14.69)
4π 4π λ 8π 2 16π

For the expressions (14.69) there are four singularity types at the light cone: the pole λ−1 ,
ln |λ|, δ(λ) and the jump θ(λ). The term ε(t) creates no additional singularities beyond the
light cone because, due to factors δ(λ) and θ(λ) which stand before it, jumps of ε(t) appear
at the origin only, that is, again at the light cone. So the functions ∆± 2
0 (x; m ) are singular.
Using (14.60) and (14.68), we can represent the Pauli-Jordan function in the form, ex-
hibiting both its covariant properties and its singularities

1 m √
∆0 (x; m2 ) = ε(t)δ(λ) − √ θ(λ)ε(t)I1 (m λ). (14.70)
2π 4π λ

It follows from (14.70) that this function is zero beyond the light cone (λ < 0). Because of
this, all (anti)commutators of the field operators, the arguments of which are separated by
a space-like interval, go to zero.
Making use of (14.69), we get that in the neighborhood of the light cone the commutator
function ∆0 (x; m2 ) is determined by

1 m2
∆0 (x; m2 ) = ε(t)δ(λ) − ε(t)θ(λ) + O(λ). (14.71)
2π 8π

According to (14.71), the function ∆0 (x; m2 ) has at the light cone a δ-shaped singularity
and a jump of m2 ε(t)/8π under the transition from the space-like domain to the time-like
one.

© 2011 by Taylor and Francis Group, LLC

 290 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

14.6 Microcausality
The equality of the ∆0 (r, t; m2 )-function to zero over the whole domain exterior to the
light cone may be demonstrated from more common considerations. Really, recall that
∆0 (r, t; m2 )-function is relativistically invariant and that each world point of the space-like
domain, with the use of the Lorentz transformations, may be reduced to a point with t = 0.
And since we have:
∆0 (r, t; m2 )|t=0 = 0,
then for the Pauli-Jordan function the following is true:
∆0 (r, t; m2 ) = 0, | t |<| r | . (14.72)
As a result, commutators of the operator field functions disappear beyond the light cone.
This property of the commutators (and anticommutators in the case of Fermi fields) was
called microcausality or local commutativity. As is known, the commutation relations allow
the following physical interpretation. Provided the operators A and B associated with some
physical quantities a and b do not commute with each other and meet the commutation
relations
[A, B] = iC,
where C in the most general case is an operator, then a system has no states, for which a
and b may have exact values simultaneously. Errors of measuring the quantities a and b are
in the intervals ∆a and ∆b obeying the uncertainty relations:
1
∆a · ∆b ≥ <C >.
2
In this case from the commutation relations (14.57) it follows that

∗ ′ ≥ ∆0 (x − x′ ; m2 ) for time-like interval,
∆ψ(x)∆ψ (x ) = (14.73)
0 for space-like interval.
It seems that microcausality, reflecting the independence of physical events separated by the
space-like interval, represents a certain modernization of the causality principle (future has
no influence on the past). However, in the quantum field theory the notion of a point event
is devoid of a direct physical meaning. This is due to the Heisenberg uncertainty relations
setting the lower limit to the extent and duration of any act associated with the fields
interaction, measuring the dynamic invariants of a field, etc. To illustrate, the coordinate
of a particle at rest may be accurately fixed merely within its Compton wavelength. Local
fields represent mathematical idealization, and only the fields weight-averaged over a small
space-time domain are physically meaningful. In other words, the microcausality conditions
should be considered as extrapolation of a physical causality condition to the domain of
short distances and small time intervals. Verification of the quantum electrodynamics has
provided support for the microcausality conditions up to the distances on the order of 10−16
cm and time intervals on the order of 10−26 s. Let us return back to the relation (14.73).
Since ∆0 (x − x′ ; m2 )-function has a δ-like singularity at the light cone (x − x′ )2 = 0 and
the δ-function is physically meaningful only after integration with respect to the domain
containing the δ-like singularity, then, to clarify a physical meaning of (14.73), we should
take the field values not at the discrete world points but averaged over the space-time
domains Ω and Ω′ . Now instead of (14.73), we have

∗ ′ ≥ {∆0 (x − x′ ; m2 )}ΩΩ′ for time-like interval,
∆{ψ(x)}Ω ∆{ψ (x )}Ω′ = (14.74)
0 for space-like interval,

© 2011 by Taylor and Francis Group, LLC

 Scalar field 291

where {}Ω ({}Ω′ ) denotes the operation of averaging over the domain Ω (Ω′ ). So, when the
operators representing a measured value at different points of the Minkowsky space x and
x′ commute with each other, the associated quantity may be measured at these points so
that one measurement cannot influence the other. This is due to the fact that in this case:

| r − r′ |>| t − t′ |,

and so the field perturbation caused by measurement procedure in one domain, even with
the light-speed propagation, there is no possibility to reach the other domain. If Ω and Ω′
are coupled by the light signals, at least partially, one measurement act may influence the
results of the other. No doubts that in this place a natural question on the reader appears:
“What observables may be associated with the operators ψ(x) and ψ ∗ (x′ )?” We know that
in the NQM the wavefunction itself has no physical meaning, as distinct from its squared
modulus. Because of this, it would be strange to expect that in the quantum field theory
the wavefunction, even in the operator form, acquires a status of the observed quantity. Of
course, this is not the case. In the quantum field theory the observables (charge, energy,
etc.) are in essence bilinear combinations of the field functions, with the electromagnetic
field strengths making a nice exception. However, bilinear combinations meet the causality
principle only at a condition that for the field operators the microcausality principle is
fulfilled.

14.7 Relativistically covariant scheme of the canonical

quantization
As follows from the integral representation of the function ∆0 (x; m2 ) (14.58), it is a singular
solution for the Klein-Gordon equation

( + m2 )∆0 (x; m2 ) = 0, (14.75)

satisfying the initial conditions:

∂
∆0 (r, t; m2 )|t=0 = 0, ∆0 (r, t; m2 )|t=0 = δ (3) (r). (14.76)
∂t
By definition, this function is an even function of the space coordinates and an odd time
function:

∆0 (−r, t; m2 ) = ∆0 (r, t; m2 ), ∆0 (r, −t; m2 ) = −∆0 (r, t; m2 ).

Also, it is easy to see that its positive- and negative-frequency components are the solutions
for the Klein-Gordon equation:

( + m2 )∆+ 2
0 (x; m ) = 0, ( + m2 )∆− 2
0 (x; m ) = 0. (14.77)

So, an explicit form of the quantities to characterize the field demonstrates that, de-
spite a selected role of time in the canonical quantization procedure, the resultant theory
is relativistically invariant. Based on this reasoning, the existence of a canonical quantiza-
tion procedure possessing an explicit relativistic invariance at all the intermediate stages is
apparent [7]. We describe its principal points.

© 2011 by Taylor and Francis Group, LLC

 292 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Consider the space-like hypersurface σ, whose normal vector is nµ . Any four-dimensional

vector may be resolved into the longitudinal and transverse components with respect to the
vector nµ :
aµ = aµT + nµ an , (14.78)
where
aµT = (δνµ − nµ nν )aν , an = n ν aν .
All scalar products in this representation are of the form:
aµ bµ = an bn − aµT bTµ .
The “momentum” π(x) canonically conjugate to the “coordinate” of the field ψ(x) may be
determined in the relativistically invariant form
∂L
π(x) =  . (14.79)
∂ψ(x)
∂ ∂xn

From the Lagrangian (14.4) we can find

π(x) = ∂n ψ † (x) = ∂µ ψ ∗ (x)nµ = πµc (x)nµ . (14.80)
Next, we suppose that the commutation relations between the field coordinates and mo-
menta are given as follows:
Z
[ψ(x), πµc (x′ )]dσ µ (x′ ) = i, (14.81)
σ
Z
[ψ † (x), πµc† (x′ )]dσ µ (x′ ) = i, (14.82)
σ
where x and x′ belong to one and the same space-like hypersurface σ. Substituting the
expression for πµc (x′ ) into (14.81), we have:
Z
∂ψ ∗ (x′ )
[ψ(x), ]dσ µ (x′ ) = i. (14.83)
σ ∂x′µ
The relation of (14.83) is nothing else but a covariant generalization of the ordinary canonical
commutation relation:
∂ψ ∗ (t, r′ )
[ψ(t, r), ] = iδ (3) (r′ − r), (14.84)
∂t
having been integrated with respect to the three-dimensional volume dV ′ including the
point r. To prove this statement, we show that the left-hand side of the expression (14.83)
is independent on the choice of the space-like hypersurface σ, that is, its derivative δ/δσ(x′ )
goes to zero. Taking into consideration that ψ ∗ (x′ ) obeys the Klein-Gordon equation, we
actually obtain:
Z
δ ∂ψ ∗ (x′′)
′
[ψ(x), ]dσµ (x′′ ) = [ψ(x), ′ ψ ∗ (x′ )] = m2 [ψ(x), ψ ∗ (x′ )] = 0. (14.85)
δσ(x ) σ ∂x′′µ

Then for hypersurface σ in the left-hand side of (14.83) we can select the hyperplane t′ =
t =const thus proving our statement.
When r′ 6= r, the right-hand side of the expression (14.83) goes to zero. In this case the
corresponding covariant generalization of (14.83), similar to other field commutators to be
equal to zero, represents simply the relation:
[ψ(x′ ), ψ ∗ (x) = 0, (x′µ − xµ )2 < 0, (14.86)

© 2011 by Taylor and Francis Group, LLC

 Scalar field 293

that means commutation of the field variables associated with two different points of the
space-like hypersurface.
Let us show that the function ∆0 (x; m2 ) makes it possible to solve a Cauchy problem for
the Klein-Gordon equation. It means that using this function, we can find a Klein-Gordon
equation solution at the arbitrary space-like hypersurface σ in terms of the values of the
solution and its derivative at any earlier space-like hypersurface σ0 :
Z  
µ ′ ∂∆0 (x − x′ ; m2 ) ′ ′
′
2 ∂ψ(x )
ψ(x) = dσ (x ) ψ(x ) − ∆0 (x − x ; m ) . (14.87)
σ0 ∂x′µ ∂x′µ

To demonstrate the validity of Eq.(14.87) we write its right-hand side in the form:
Z
Iσ = dσ µ Fµ
σ0

and act upon that by the operator δ/δσ. Since both ψ(x′ ), and ∆0 (x − x′ ; m2 ) are the
solutions for the Klein-Gordon equations, we can derive:
R µ
R µ
δ σ dσ Fµ (x) − σ0 dσ Fµ (x) ∂Fµ (x)
Iσ = lim = = 0. (14.88)
δσ(x) Ω→0 Ω(x) ∂xµ

As follows from (14.88), Iσ is independent on a choice of the space-like hypersurface σ.

Then nothing prevents us from choosing as such the hyperplane t = const again. Owing to
the properties of ∆0 (x; m2 ), from (14.87) we obtain the identity ψ(x) = ψ(x), which was to
be proved.
Now finding of the commutation relations at an arbitrary instant of time becomes fairly
trivial. To find the value of [ψ(x), ψ † (x′ )], one should express the operator ψ(x) in terms
of the field variables at the space-like hypersurface passing through the point x′ , and, next,
makes use of the well-known commutation relations at this surface. By application of Eqs.
(14.87) and (14.83) we get
Z 
∂∆0 (x − x′′ ; m2 )
[ψ(x), ψ † (x′ )] = dσ µ (x′′ ) [ψ(x′′ ), ψ † (x′ )]−
σ ∂x′′µ

∂ψ(x′′ ) † ′
′′ 2
−∆0 (x − x ; m )[ , ψ (x )] . (14.89)
∂x′′µ
Since x′ ∈ σ, x′′ ∈ σ0 and the integral in the right-hand side is independent on a choice of
σ, then, passing to the hyperplane t=const, we obtain the same commutation relations as
in the ordinary canonical formalism:

[ψ(x), ψ † (x′ )] = −i∆0 (x − x′ ; m2 ). (14.90)

14.8 Green function of the scalar field

For a scalar field the Green function is a solution of the equation:

( + m2 )G(x) = δ(x). (14.91)

© 2011 by Taylor and Francis Group, LLC

 294 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Performing an integral Fourier expansion of G(x) and using an integral representation of

the delta-function, with the help of Eq. (14.91) we obtain the following formal expression
for the Green function:
Z
1
G(x) = G(k) exp (−ikx)d4 k, (14.92)
(2π)4
where
1
G(k) = .
m2 − k 2
√
This expression is indefinite because no rules for detour of the poles (k0 )1,2 = ± k2 + m2
are set. Indefiniteness reveals the fact that a complete solution for equation (14.91) is
given as a sum of a partial solution for the inhomogeneous equation and solutions of the
corresponding homogeneous equation, ∆+ −
0 (x) and ∆0 (x), taken with arbitrary factors. We
shall demonstrate that, setting the rules for detour of the poles or, which is equivalent,
imposing the boundary conditions for G(x), we uniquely determine these factors.
Let us consider a retarded Green function, for that the following initial condition is true:

∆ret (x) = 0, t < 0. (14.93)

In order to represent ∆ret (x) in the form close to (14.92), we note that on multiplication of
this function by exp (−ǫt), because of (14.93), it assumes no additional properties:

∆ret (x) exp (−ǫt) = Gǫ , (14.94)

and may be represented as a limit:

∆ret (x) = lim Gǫ . (14.95)

ǫ→+0

In accordance with the definition of (14.94), the function Gǫ satisfies the equation:
" 2 #
∂
+ ǫ − △ + m2 Gǫ = δ(x)
∂t

taking in the momentum representation in the limit ǫ → +0 the following form:

1 1
→ 2 .
m2 − (k0 + iǫ)2 + k2 m − k 2 − 2iǫk0

And according to (14.94), the retarded Green function may be represented as follows:
Z
1 d4 k
∆ret (x) = exp (−ikx). (14.96)
(2π)4 m2 − k 2 − 2iǫk0

It is obvious that expression (14.96) meets the condition of (14.93). To make sure that this
is true, it suffices to integrate with respect to the variable k0 using a theory of residues.
An infinitesimal addition
√ 2iǫk0 , in the denominator of (14.96) indicates that detour of both
poles (k0 )1,2 = ± k2 + m2 = ±ω in a complex plane of the variable k0 should be from
above, that is, integration follows the contour LC (Fig. 14.1). In accordance with Jordan’s
lemma, the contour L may be always closed by semicircles L⌣ and L⌢ of radius R (Fig.14.2)
so that in the limit R → ∞ the additional contour integrals along L⌣ and L⌢ turn to zero.
To this end, it is sufficient to require the fulfillment of the condition:

exp (−izt)|R→∞ → 0, (14.97)

© 2011 by Taylor and Francis Group, LLC

 Scalar field 295

Im k 0

LC

−ω +ω Re k 0

FIGURE 14.1
The integration contour for the function ∆ret .

Im k0
Lr+
−ω LC Re k0
L⌢
Lr− +ω
t<0 R→∞
r+ → 0 R→∞ t>0
−ω Lr+ L⌣
Re k0
r− → 0 Lr− LC +ω Im k0

FIGURE 14.2
The closed integration contours in the complex plane of k0 .

where z = Re k0 + iIm k0 , under R → ∞. In this case from (14.96) it follows that for t < 0
the integration should be over the upper half-plane with a counterclockwise detour of the
closed contour, and for t > 0—clockwise over the lower half-plane. At t < 0 there are no
poles within the contour of integration, and we obtain (14.93). On the other hand, using a
theory of residues, at t > 0 we find:
Z  
1 exp (ik0 t) − exp (−ik0 t)
∆ret (x) = d3 k exp (−ikr) = ∆0 (x).
(2π)3 i 2k0 √
k0 = k2 +m2
(14.98)
Thus
∆ret (x) = θ(t)∆0 (x). (14.99)

Performing similar calculations for the advanced Green function

∆adv (x) = 0, at t>0 (14.100)

we get
Z
adv 1 d4 k
∆ (x) = exp (−ikx) = −θ(−t)∆0 (x). (14.101)
(2π)4 m2 − k 2 + 2iǫk0

© 2011 by Taylor and Francis Group, LLC

 296 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

14.9 Causal Green function

Causal link of the production and annihilation processes at various space-time points is
established by the causal Green function. Since in a theory of interacting fields this function
gives a description for the virtual particles propagation, it is often called the propagator.
The process of a scalar particle production at the point x and its subsequent annihilation
at the point x′ (t < t′ ) is associated with the matrix element:

Φ∗1 (x′ )Φ1 (x) = Φ∗0 ψ (−) (x′ )ψ (−)∗ (x)Φ0 = −i∆+ ′ 2
0 (x − x ; m ). (14.102)

But when a particle is produced at the point x′ and annihilated at the point x, instead of
(14.102) we obtain:

Φ∗1 (x)Φ1 (x′ ) = Φ∗0 ψ (−) (x)ψ (−)∗ (x′ )Φ0 = i∆− ′ 2
0 (x − x ; m ). (14.103)

To characterize both processes, we introduce the causal function in accordance with the
following relation:
∆c (x) = θ(t)∆− 2 + 2
0 (x; m ) − θ(−t)∆0 (x; m ). (14.104)
As seen from (14.104), the causal function represents a sum of the partial solution for
inhomogeneous equation (14.91) and general solution for equation (14.1) because:

∆c (x) = ∆ret (x) − ∆+ 2

0 (x; m ). (14.105)

Thus, it satisfies all the requirements for the general solution of the equation:

( + m2 )∆c (x) = δ(x). (14.106)

Let us find an expression for the causal function in the momentum representation. For
this purpose we rewrite the expression for ∆ret (x) in the following form:
Z  Z 0
i
∆ret (x) = lim d4
k exp (−ikx) − dβθ(−k0 )×
(2π)4 ǫ→+0 −∞
Z ∞ 
2 2 2 2
× exp [−iβ(m − k − 2iǫk0 )] + dβθ(k0 ) exp [−iβ(m − k − 2iǫk0 )] =
0
 Z 4 Z 
1 d k
= P exp (−ikx) + iπ d4 kδ(m2 − k 2 ) exp (−ikx) , (14.107)
(2π)4 m2 − k 2
where the symbol P denotes that the principal-value integral is taken. The principal value
exists for the improper integral under consideration because: (i) the function f (k) =
1/(m2 − k 2 )√is continuous over the interval (−∞, ∞), going to zero only at the points
(k0 )1,2 = ± k2 + m2 ; (ii) the first-order and second-order derivatives, (∂f (k)/∂k0 ) and
(∂ 2 f (k)/∂k02 ), at singular points are nonzero. In this way the integral:
Z
d4 k
exp (−ikx)
m2 − k 2
is divergent, but there are limits for both singular points. For instance, for the point (k0 )2
it is of the form:
"Z (2) Z k0(1) −η #
k0 −η
dk0 dk0
lim exp (−ikx) + exp (−ikx) .
η→+0 −∞ m2 − k 2 2
k0 +η m − k
(2) 2

© 2011 by Taylor and Francis Group, LLC

 Scalar field 297

Using (14.61), (14.105) and taking into account

1 1
P + iπδ(a) = ,
a (a − iǫ)

we get
Z
c 1 d4 k
∆ (x) = exp (−ikx). (14.108)
(2π)4 m2 − k 2 − iǫ
To study a behavior of the causality function, we establish its explicit form. This op-
eration presents no difficulties as we have already obtained the associated expressions for
∆± 2
0 (x; m ). With allowance made for (14.68) and (14.104), we find:

1 m h √ √ i
∆c (x) = δ(λ) − √ θ(λ) J1 (m λ) − iN1 (m λ) +
4π 8π λ

im √
+ √ θ(−λ)K1 (m −λ). (14.109)
4π 2 −λ
So, the causal function is other than zero both within and beyond the light cone.
With the help of (14.69), we can obtain expansion of ∆c (x) in the neighborhood of the
light cone:

1 1 m2 im2 m|λ|1/2 p
∆c (x) = δ(λ) + 2 − θ(λ) + ln + O( |λ| ln |λ|). (14.110)
4π 4π iλ 16π 8π 2 2
As follows from (14.110), at the light cone the causal function has all four singularities
possible for commutator functions: δ(λ), pole λ−1 , jump θ(λ) and ln |λ|.

14.10 Chronological product. Convolution of operators

Except for the normal product of the operators, an important role is played by the chronolog-
ical product, the T -product. The factors of a T -product have chronological order, namely:
the factor on the right is associated with a lower value of time, and that on the left—with
the greater one. In the case of a scalar field this means that:

ψ(x)ψ ∗ (x′ ), t > t′
T (ψ(x)ψ ∗ (x′ )) = (14.111)
ψ ∗ (x′ )ψ(x), t < t′ .

Let us show that definition of the T -product is relativistically invariant. Indeed, if x and
x′ are separated by a time-like interval, this statement is apparent because in this case the
notion of earlier and later time is absolute in character. But if x and x′ are separated by a
space-like interval, and a sign of t−t′ is varying in the process of the Lorentz transformation,
the operators ψ(x) and ψ ∗ (x′ ) commute with each other. Because of this, in all frames of
reference they may be positioned in the order set by the chronological product. Since bosons
are quantized by the commutation relations, if required, reordering of the factors both in
the normal and chronological products leaves the sign unaltered. As may be seen later, for
fermions, where quantization involves anticommutators, this is not the case.
Consider the difference between chronological and normal products of the field functions
ψ(x) and ψ ∗ (x′ ). This difference, referred to as convolution or pairing of the operators, is

© 2011 by Taylor and Francis Group, LLC

 298 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

denoted by the superscript si , where i takes on values 1,2,3... and its value is the same for
each pair of the operators,

T (ψ(x)ψ ∗ (x′ )) − N (ψ(x)ψ ∗ (x′ )) ≡ ψ s1 (x)ψ ∗s1 (x′ ). (14.112)

It is easily demonstrated with direct calculations that, both for t > t′ and for t < t′ ,
convolution of the operators ψ s1 (x)ψ ∗s1 (x′ ) is given by one and the same expression:

1 X 1
ψ s1 (x)ψ ∗s1 (x′ ) = [exp [ik(r − r′ ) − iω | t − t′ |], (14.113)
V 2ω
k

that contains no operators, that is, is a c-number. Going from the sum to the integral, we
obtain
ψ s1 (x)ψ ∗s1 (x′ ) = ∆′ (x − x′ ), (14.114)

where Z
1 d3 k
∆′ (x) = θ(k0 ) exp (ikr − ik0 | t |) (14.115)
2(2π)3 k0
and k0 ≡ ω. Further, it is essential to know a Fourier-component of the ∆′ (x) function
Z
1
∆′ (x) = d4 k∆′ (k) exp (−ikx). (14.116)
(2π)4

It may be found by the formula:

Z
exp (−ik0 | t |) 1 exp (−ik0 t)
= dk0 , (14.117)
ω πi LI m2 − k 2

with integration over the complex plane k0 along the contour LI shown in Fig. 14.3. The

k0

−ω
LI
+ω

FIGURE 14.3
The integration contour LI .

validity of (14.117) is apparent as for t > 0 the integral is determined by the residue at the
point (k0 )1 = ω and for t < 0—by the residue at the point (k0 )2 = −ω. In Eq. (14.117)
from integration along the contour LI we may proceed to integration along the real axis,
with shifting of the ω-pole to the lower half-plane and −ω-pole—to the upper half-plane
(Fig. 14.4). To this end, we replace m2 − k 2 by m2 − k 2 − iǫ in the denominator of the

© 2011 by Taylor and Francis Group, LLC

 Scalar field 299

k0

−ω + i0
LR

ω − i0

FIGURE 14.4
Crossing to integration along the real axis.

integrand, where ǫ is an infinitesimal positive number. Substituting (14.117) into (14.115)

and comparing with (14.116), we derive the following expression for ∆′ (k):
Z
−i d4 k
∆′ (x) = 4
exp (−ikx), (14.118)
(2π) m − k 2 − iǫ
2

that is a clear demonstration of the coincident (within a constant factor) of operators

convolution and causal Green function. As follows from the definition of the operators
convolution (14.112), it is related to the vacuum average of the T -product for these operators

ψ s (x)ψ ∗s (x′ ) =< T (ψ(x)ψ ∗ (x′ )) >0 = −i∆c (x − x′ ). (14.119)

Also, it is obvious that convolution of the two commutating operators becomes zero:

ψ s1 (x)ψ s1 (x′ ) = ψ ∗s1 (x)ψ ∗s1 (x′ ) = 0. (14.120)

So far on the expansion of the field functions into the solutions for free equations of
motion we have used the discrete momentum representation. It turned to be very useful for
consideration of the field from the viewpoint of the wave-particle dualism. It was helpful
for making clear a field model based on a set of an infinite number of harmonic oscillators.
At the same time, in practice a ordinary (continuous) momentum representation is more
convenient for the calculations. This becomes particularly apparent under a changeover
from the coordinate representation to the momentum representation in matrix elements
associated with the Feynman diagrams. In what follows we use the continuous momentum
representation. A changeover from the discrete representation to the continuous one is
as follows. First, in the expressions (14.13) and (14.14) we must replace summation with
integration that is achieved by the use of (14.26). Next, we go from the operators with
(±)
respect to the discrete variable ak to new operators a(±) (k). Recall that in the case of
a discrete spectrum we have selected the normalization corresponding to a single particle
within the volume V = 1. Retention of the previous normalization requires the fulfillment
of the condition: Z
X (±)
|ak |2 = d3 k|a(±) (k)|2 ,
k

whence the desired relation is found:

√
V (±)
a → a(±) (k).
(2π)3/2 k

© 2011 by Taylor and Francis Group, LLC

 300 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Because of this, modification involves also the commutation relations for the production
and annihilation operators. Here we need replacement of the Kronecker symbols by the
delta-functions. As a result, we have:
Z
(+) (−) 1 d3 k
ψ(x) = ψ (x) + ψ (x) = √ [a(k)e−ikx + b∗ (k)eikx ], (14.121)
(2π)3/2 2k0
Z
∗ (+)∗ (−)∗ 1 d3 k ∗
ψ (x) = ψ (x) + ψ (x) = √ [a (k)eikx + b(k)e−ikx ] (14.122)
(2π)3/2 2k0
and
[a(k), a∗ (k′ )] = [b(k), b∗ (k′ )] = δ(k − k′ ). (14.123)
With the use of the orthonormality conditions the expansions (14.121) and (14.122) may
be inverted, giving the following equations:
Z  
i d3 x ikx
↔
a(k) = √ e ∂0 ψ(x) , (14.124)
(2π)3/2 2k0
Z  
i d3 x ikx
↔
∗
b(k) = √ e ∂0 ψ (x) . (14.125)
(2π)3/2 2k0
The equations for a∗ (k) and b∗ (k) follow from (14.124) and (14.125) under complex con-
jugation. Using the obtained relations, we can rewrite the system state vector (14.53) in
terms of the operator field functions in the configuration space as follows:
Z Z
Φn+ n−′ ... = ... f (x1 , ..., xn )ψ α1 (x1 )...ψ αn (xn )d4 x1 ...d4 xn Φ0 , (14.126)
k1 k1

where αi = ±, ψ + (x) ≡ ψ(x), ψ − (x) ≡ ψ ∗ (x) and the weight function f (x1 , ..., xn ) has a
meaning of the quantum-mechanical wavefunction for a particles system in the configuration
space. It should be noted that by definition of the vacuum state, the operator field functions
in (14.126) contain only the parts associated with the particles and antiparticles production.
Definition of the state vector in the form of (14.126) is referred to as the Fock state-vector
representation.
For a truly neutral scalar field the above second quantization formalism has no appreciable
changes. The Lagrangian density in this case is given by:
1 m2 ∗
L(x) =: ∂µ ψ(x)∂ µ ψ(x) − ψ (x)ψ(x) : . (14.127)
2 2
Expansion of the field function with respect to the solutions for the Gordon-Klein equation
involves the particles operators only:
Z
1 d3 k
ψ(x) = 3/2
√ [a(k) exp (−ikx) + a∗ (k) exp (ikx)]. (14.128)
(2π) 2k0
The principal difference from the case of a complex scalar field is the fact that all “currents”
of the real scalar field are zero.

14.11 First-order equation for scalar particles

The Klein-Gordon equation is inferred from the relativistic relations between the energy,
momentum, and mass. It means that all the components of the operator field functions

© 2011 by Taylor and Francis Group, LLC

 Scalar field 301

describing the particles with an arbitrary spin have to obey this equation, or else, as they
often say, must obey the Klein-Gordon condition. A relativistic wave equation for an
arbitrary spin field may be represented as:

Λαβ (∂)Qβ (x) = 0, (14.129)

where β = 1, 2, ....r (r > 2J + 1),

Λαβ (∂) = (iΓµ ∂µ − m)αβ ,

Γµ are r-rank square matrices, the properties of which are determined by the transformation
laws of the Qβ (x)-function and by the covariance conditions of equation (14.129). The
existence of a ”conjugate” field function is needed for the Lagrangian density to be built.
This function is determined by:
Qβ (x) ≡ Q∗α ηαβ
where η is a nonsingular matrix. An equation for Qβ (x) is of the form:

Λαβ (∂)Qα (x) = 0, (14.130)

where Λ(∂) ≡ ηΛ(∂). The Klein-Gordon condition restricts Eqs. (14.129) and (14.130) as
regards the existence of such a dτ α (∂) operator that would meet the relations:

dτ α (∂)Λαβ (∂) = ( + m2 )δτ β , (14.131)

Λαβ (∂)dβτ (∂) = ( + m2 )δτ β , (14.132)

(d(∂) ≡ d(∂)η −1 ), and would be dependent on the operators ∂µ

d(∂) ≡ [dτ α (∂)] = α + αµ ∂ µ + ..... + αµ1 ....µn ∂ µ1 ....∂ µn + ..., (14.133)

where the coefficients α, αµ .... are matrices of the r-order. Note that in the most general
case the quantities Qβ (x) belong to the reducible representation.
Substitution of (14.133) into (14.131) leads to the following recurrent relations:

P=(P−mI,
α
)
iαΓµ − mαµ = 0, i(αν Γµ + αµ Γν ) − 2mαµν = 2gµν ,
(14.134)
(iΓµ1 αµ2 ,...µl − mαµ1 ,...µl ) = 0 l > 2,
P (P )
where the symbol denotes summation over all the terms resultant from various rear-
rangement of the indices.
To have an explicit form of the dαβ (∂)-operator, we need to know a maximum order of
the involved derivatives (nmax ). We prove a theorem that states that nmax is given by the
relation [8]:
nmax = 2J, m 6= 0, (14.135)
where J is a maximum spin value of different fields described by the field function Qα .
In the process Nof the Lorentz transformation the operator dαβ (∂) is transformed in the
same way as Qα Qβ . Since the quantities Qα describe the fields with a maximum spin
J, they form an invariant space for the representation of a three-dimensional rotations
group that is decomposed
N into irreducible representations DJ , DJ−1 ,.... Because of this,
the representation Qα Qβ may be decomposed into D2J , D2J−1 ,... Similarly, ∂µ1 ...∂µl are
associated with a family of the irreducible representations Dl , Dl−1 , .. until some of the oper-
ators ∂µ begin to combine into the scalar operator . So, in d(∂) the term αµ1 ....µl ∂ µ1 ....∂ µl
may appear only in the following form:
l−2J
αµ1 ....µl ∂ µ1 ....∂ µl = α′µ1 ....µ2J ∂ µ1 ....∂ µ2J () 2
, (14.136)

© 2011 by Taylor and Francis Group, LLC

 302 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where l − 2J > 0. It is evident that the relation (14.136) goes to zero for odd l − 2J. Then
from the last line of the recurrent formulae (14.134) we obtain that the quantity αµ1 ....µl is
also zero for even positive values of l − 2J, that is,

αµ1 ....µl = 0 l > 2J, (14.137)

which was to be proved.

Let us perform the procedure of transformation from the second-order equation to the
first-order one (linearization) for a spinless particles field. We introduce the following quan-
tities:
i
ψµ (x) = ∂µ ψ(x).
m
Then Eq. (14.1) is replaced by the system:

i∂µ ψ(x) − mψµ (x) = 0,
(14.138)
i∂µ ψ µ (x) − mψ(x) = 0,

that may be represented in the form of the first-order matrix equation:

(iβ µ ∂µ − m)Ψ(x) = 0, (14.139)

where  
ψ0 (x)
 ψ1 (x) 
 
Ψ(x) =  ψ2 (x) 
 
ψ3 (x)
ψ(x)
and the β-matrices have the following form:
    
0 0 0 0 1 0 0 0 0 0 

0 0 0 0 0 0 0 0 0 −1  
    
β0 =  0 0 0 0 0, β1 =  0 0 0 0 0 , 

    
0 0 0 0 0 0 0 0 0 0 



1 0 0 0 0  0 1 0 0 0  (14.140)
0 0 0 0 0 0 0 0 0 0 

0 0 0 0 0  0 0 0 0 0  

    
β2 =  0 0 0 0 −1  , β3 =  0 0 0 0 0 .

    

0 0 0 0 0 0 0 0 0 −1 


0 0 1 0 0 0 0 0 1 0

Eq. (14.139) was proposed in [9] and called the Duffin-Kemmer equation.
Since the resultant representation includes the representation (0, 0), according to which
the scalar ψ(x) is transformed, and the representation (1/2, 1/2) defining the transformation
law of the vector ψµ (x), then:
nmax = 2.
Solving the system of Eqs. (14.134) for nmax = 2, we obtain both the expression for d(∂):
1 1
d(∂) = (β µ β ν + β ν β µ )∂µ ∂ν − iβ µ ∂µ − ( + m2 ), (14.141)
2m m
and the permutation relations determining the Duffin-Kemmer matrix algebra:

β µ β ν β λ + β λ β ν β µ = g µν β λ + g λν β µ . (14.142)

© 2011 by Taylor and Francis Group, LLC

 Scalar field 303

It is easily seen that the determinant of each βµ -matrix is zero, that is, there are no matrices
inverse to βµ . This may be easily demonstrated without recourse to the explicit form of
matrices βµ , proceeding from their algebra only. Actually, from (14.142) it follows that:

βµ βν βµ = 0 (µ 6= ν),
(14.143)
βµ3 = βµ , (no summation on µ).

As a consequence, in case the matrices inverse to βµ were existing, the second relation
(14.143) would give βµ2 = I, and hence from the first relation it would follow that βµ βν = 0
(µ 6= ν) and βµ = 0.
It is clear that the Duffin-Kemmer equation and relations of (14.142) are invariant with
respect to the transformation:

Ψ′ (x) = U Ψ(x), βµ′ = U βµ U −1 , (14.144)

where U is an arbitrary unitary matrix. Thus, the physical results obtained by the Duffin-
Kemmer theory should be expressed in terms of the quantities to be invariant with respect
to the transformations (14.144). For the expressions involving β-matrices this means that
they may contain only such combinations of β-matrices, which may be reduced to tracks.
Indeed, for the product of an arbitrary number of β-matrices the only possible quantity
being invariant under the transformation (14.144) is a track as:

Sp {βµ1 βµ2 .....βµk } = Sp {U −1 U βµ1 βµ2 .....βµk } = Sp {U βµ1 βµ2 .....βµk U −1 } =

= Sp {U βµ1 U −1 U βµ2 U −1 .....U βµk U −1 } = Sp {βµ′ 1 βµ′ 2 .....βµ′ k } = inv,

where we have used:
Sp (ABC) = Sp (CAB) = Sp (BCA). (14.145)
To calculate the tracks of β-matrices, we use the following procedure. Introduce the
projective operators:
1 µ 1
P = (β βµ − 1) , P = (4 − β µ βµ ) ,
3 3
separating in the five-dimensional space of the Ψ(x)-functions one-dimensional space of the
ψ(x)-scalars and four-dimensional space of the ψµ (x)-vectors, respectively. Using (14.142),
we make sure that the following relations:

P βν = βν P , P βν βσ = βν βσ P ,
(14.146)
P βν βσ = βν βσ P, P βν P = P βν P = 0.

are valid. Now with the help of (14.146) we may find the formula of a track for the product
of any number n of β-matrices, So, for even n, after simple transformations, we have:
2
Sp{βµ1 βµ2 .....βµn } = Sp{βµ1 βµ2 .....βµn P 2 } + Sp{βµ1 βµ2 .....βµn P } =

= Sp{P βµ1 βµ2 .....βµn P } + Sp{P βµ1 βµ2 .....βµn P } =

= Sp{P βµ2 βµ3 .....βµn βµ1 P } + Sp{P βµ1 βµ2 .....βµn P } =
= Sp{(P βµ2 βµ3 P ).....(P βµn βµ1 P )} + Sp{(P βµ1 βµ2 P ).....(P βµn−1 βµn P }. (14.147)
Taking into account that for the five-row β-matrices the following is satisfied:

P βµ βσ P = gµσ ,

© 2011 by Taylor and Francis Group, LLC

 304 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

finally we have:

Sp{βµ βν βλ βκ .....βρ βσ βτ } = gµν gλκ ....gστ + gνλ ...gρσ gτ µ . (14.148)

In a similar way we may show that a track for an odd number of β-matrices equals zero.
As is seen in the first-order equation, the number of components of the spinless-particle
wavefunction Qα is found to be redundant. From Section 3.6 it follows that for the reduction
in the number of independent components down to 2J + 1 the field functions should meet
the additional conditions contained in the equation of motion themselves. As follows from
(14.143), with the use of 1 − βµ2 we can eliminate some of the β-matrices in Eq. (14.139).
It is clear that the matrix β0 is the best candidate for the elimination, since in this case we
arrive at the relation containing no time derivative. Multiplication of Eq. (14.139) on the
left by the matrix 1 − β02 results in the following:
i
(1 − β02 )Ψ(x) = βk ∂k β02 Ψ(x). (14.149)
m
Because 1 − β02 = diag(0, 1, 1, 1, 0) and β02 = diag(1, 0, 0, 0, 1), expression (14.149) makes
it possible to express three components of the field function (1 − β02 )Ψ(x) in terms of the
two components β02 Ψ(x), that could be thought to be dynamically independent. These two
components, however, prove to be equal in the rest frame of a particle, that is apparent from
the additional condition of (14.149) written in the momentum representation. Then their
relationship may be established in any Lorentz system, that is, the number of independent
components is actually equal to 1.
Let us derive an equation for the ”conjugate” field function Ψ(x). Performing Hermitian
conjugation from (14.139), allowing for the properties of β-matrices, we get:


i∂µ Ψ† (x)β µ† + mΨ† (x) = i ∂0 Ψ† (x)β 0 + ∂k Ψ† (x)βk + mΨ† (x) = 0. (14.150)

The obtained equation is not of the correct form as it is deliberately relativistically non-
invariant. To eliminate this deficiency, we introduce the matrix 2β02 − 1 = η0 with the
apparent properties:
[β0 , η0 ] = 0, {βk , η0 } = 0, (14.151)
and define
Ψ(x) = Ψ† (x)η0 .
Taking into consideration Eq. (14.151), we get an equation for the conjugate field function:

∂Ψ(x) µ
i β + mΨ(x) = 0, (14.152)
∂xµ
which has the correct form due to the manipulations undertaken. Inclusion of Ψ(x) allows
for definition of the Lagrangian:
1 1
L =: Ψ(x)(iβµ ∂ µ − m)Ψ(x) − (i∂ µ Ψ(x)βµ + mΨ(x))Ψ(x) : (14.153)
2 2
and construction of the dynamic variables on its basis. Of particular importance is a feature
of the Lagrangians leading to the first-order wave equations: when the wave equations are
satisfied, these Lagrangians are going to zero.
We consider a relativistic covariance of the Duffin-Kemmer equation. The physics ex-
pressed with the help of any relativistic equation should be independent of the selected
Lorentz frame of reference. Because of this, an adequate description of physical phenomena
necessitates that the equation itself be similarly invariant relative to the selection of a frame

© 2011 by Taylor and Francis Group, LLC

 Scalar field 305

of reference. Let us demonstrate that the Duffin-Kemmer equation is in a form invariant

with respect to the Poincare transformations:

x′ = Λx + a,

on condition that the transformed field function is determined by:

Ψ′ (x′ ) = S(Λ)Ψ(x) = S(Λ)Ψ[Λ−1 (x′ − a)], (14.154)

and the matrix Λ meets special requirements, the finding of which is our principal objective.
Invariance of this equation means that in a new frame of reference the field function Ψ′ (x′ )
obeys the equation:
(iβ µ ∂µ′ − m)Ψ′ (x′ ) = 0. (14.155)
It should be noted that the Lorentz transformations do not “touch upon” the matrices β µ .
Using the relations (14.154) and

∂ ∂x′ν ∂
= = Λνµ ∂ν′ , (14.156)
∂xµ ∂xµ ∂x′ν
Eq. (14.139) may be rewritten as:

iΛνµ β µ ∂ν′ [S −1 (Λ)Ψ′ (x′ )] − mS −1 (Λ)Ψ′ (x′ ) = 0. (14.157)

After the multiplication of Eq. (14.157) by S(Λ) on the left we have:

iS(Λ)Λνµ β µ S −1 (Λ)∂ν′ Ψ′ (x′ ) − mΨ′ (x′ ) = 0. (14.158)

Consequently, the Duffin-Kemmer equation will be relativistically covariant when the trans-
formation matrix for the field function (14.154) satisfies the condition:

Λνµ β µ = S −1 (Λ)β ν S(Λ). (14.159)

With the use of the Poincare transformation property

Λνµ Λσµ = δσν ,

we can easily verify that the matrices:

β ′ν = Λνµ β µ ,

obey the same permutation relations as the matrices βµ.

By a system of four Eqs. (14.159) we can determine the transformation matrix for the
field functions of the Duffin-Kemmer equation in the case of the Poincare transformations.
To illustrate, we perform the procedure for the case of space inversion, that is, in the case
when the matrix Λ is of the following form:
 
1 0 0 0
 0 −1 0 0 
Λ= . (14.160)
0 0 −1 0
0 0 0 −1

Multiplying the equality (14.159) on the left by S(Λ), we bring it into the form:

βν S(Λ) = S(Λ)Λµν βµ . (14.161)

© 2011 by Taylor and Francis Group, LLC

 306 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Now with allowance made for the explicit form of the Λ matrix we get:

β0 S(Λ) = S(Λ)β0 , β1 S(Λ) = −S(Λ)β1 ,
(14.162)
β2 S(Λ) = −S(Λ)β2 , β3 S(Λ) = −S(Λ)β3 .

The relations (14.162) are satisfied if S(Λ) is selected as follows: S(Λ) = η0 .

It is possible that the reader is inspired with the illusion that under the transition from
the second- to the first-order equation the number of roots of the associated characteristic
equation is reduced to one, with elimination of the unwanted negative root. But this is
not the case, since “Nothing originates from nothing.” Indeed, the substitution of the de
Broglie wave:
Ψ(x) = ϕ(k) exp (−ikx),
into Eq. (14.139) gives
(k̂ − m)ϕ(k) = 0, (14.163)
where k̂ = β µ kµ . By Eq. (14.142), for any vector aµ the following relation is satisfied:

â(â2 − a2 ) = 0, (14.164)

then from (14.163) and (14.164) follows:

k 2 − m2 = 0,

As before, there are two solutions with opposite signs of the energy.
Proceeding from the same ideology as in the case of the second-order equation, we can
perform a series expansion of the operator field functions in terms of the solutions for the
free equation:
Z
1 d3 k
Ψ(x) = √ [a(k)ϕ(+) (k) exp (−ikx) + b∗ (k)ϕ(−) (k) exp (ikx)], (14.165)
(2π)3/2 2k0
Z
1 d3 k ∗
Ψ(x) = 3/2
√ [a (k)ϕ(+) (k) exp (ikx) + b(k)ϕ(−) (k) exp (−ikx)], (14.166)
(2π) 2k0
where the particles (antiparticles) production a∗ (k) (b∗ (k)) and annihilation a(k) (b(k))
operators satisfy the commutation relations (14.123), and ϕ(±) (k) obey the normalization
conditions:
ϕ(±) (k)β0 ϕ(±) (k) = ±2k0 . (14.167)
It is easy to derive nonzero commutation relations for the operator field functions:

[Ψα (x), Ψβ (x′ )] = −idαβ (∂)∆0 (x − x′ ; m2 ) (14.168)

and the expression for the causal Green function:

Z
c 1 dαβ (k)
Dαβ (x) = dαβ (∂)∆c0 (x) = d4 k 2 exp (−ikx). (14.169)
(2π)4 m − k 2 − iǫ

The fact that the operator dαβ (∂) is so frequently used in our expressions indicates its
exclusive role in the Duffin-Kemmer theory. Note that this operator is also very important
for the particles with a nonzero spin value. Point to one more activity sphere of the operator
dαβ (∂). It is connected to computing of the bilinear combinations of the form:

P (±) (k) = ϕ(±) (k)ϕ(±) (k), (14.170)

© 2011 by Taylor and Francis Group, LLC

 Scalar field 307

which is quite often found in calculations of the scattering cross sections. As P (+) (k) and
P (−) (k) represent solutions of the equations:

(k̂ − m)P (+) (k) = 0, (14.171)

(k̂ + m)P (−) (k) = 0, (14.172)

they should be coincident, within a numerical factor, with the operators d(k) and d(k):

P (+) (k) = c+ d(k), P (−) (k) = c− d(k). (14.173)

The factors c+ and c− may be determined from the normalization conditions (14.167) as
follows:
ϕ(+) (+) (+) (+)
α (k)(β0 )ατ ϕτ (k) = Sp{ϕα (k)(β0 )ατ ϕτ (k)} =

= Sp{(β0 )ατ ϕ(+) (+)

τ (k)ϕα (k)} = Sp{(β0 )P
(+)
(k)) = 2k0 , (14.174)
(−) (−) (−)
ϕ (k)β0 ϕ (k) = Sp(β0 P (k)) = −2k0 . (14.175)
Using the track calculation rules for the Duffin-Kemmer matrices, with the help of (14.174)
and (14.175) we get
1
c+ = c− = − .
2
Considering that the particles are at the mass surface, we finally arrive at:

k̂(k̂ + m) k̂(k̂ − m)
ϕ(+) (k)ϕ(+) (k) = , ϕ(−) (k)ϕ(−) (k) = . (14.176)
2m 2m
With the use of (14.153), (14.165), and (14.166), we can calculate the energy, momentum,
charge, and other dynamic invariants of a spinless particle. In so doing we obtain the same
expressions as in the case of the Klein-Gordon equation. Because of this, it may be inferred
that the descriptions of a free scalar field both in the Duffin-Kemmer and Klein-Gordon
form are equivalent. Also, this equivalence is retained on inclusion of the interactions as
both formalisms result in the same values of the quantities measured in the process of
experiments.
Thus, even though the equation order was lowered, the Duffin-Kemmer formalism has
not given a simpler pattern of a scalar field compared to the Klein-Gordon formalism.
However, our time was not wasted. It was demonstrated that the characteristics of the
arbitrary-spin particles are expressed in terms of the corresponding quantities of a scalar
field. Moreover, this makes it possible: (i) to simplify, using the formalism of the first-order
equations, a derivation procedure for the commutation relations, causal Green function,
and other quantities associated with the arbitrary-spin particles; and (ii) to formulate the
relativistic correspondence principle that is very useful in practical calculations: scattering
cross sections of spin particles σs include the cross sections of the scalar particles σ0 , that
is,
lim σs = σ0 .
s→0

© 2011 by Taylor and Francis Group, LLC

 15
Particles with spin 1/2

And sure enough, the forest subjects hadn’t winked

twice, when Bruin the First, the new governor, arrived. He
reached his residence early in the morning of St. Michael’s Day
and immediately resolved: tomorrow would see bloodshed. What
made him adopt this decision is unknown: for as a matter of
fact he was not really spiteful—just a brute, that’s all.
M.E. Saltykov Shchedrin, “Bears in Government”

15.1 Dirac equation

As it was shown, the single-particle Klein-Gordon equation suffers from two disadvantages:
(i) the probability density is not positively defined; and (ii) the existence of the solutions
with a negative value of the energy. Trying to overcome these problems, in 1928 P. Dirac
proposed an equation that was later named after him. The Dirac equation is of special
importance because it allows for the description of such fundamental particles of matter as
leptons and quarks.
At least now one easily understands the train of reasoning that led Dirac to his equation
[10]. To avoid the appearance of negative probabilities, it is essential that an expression for
the probability density be free from the time derivatives. Consequently, a wave equation
should involve the time derivatives of the first order, and not higher. Allowing for that a
relativistic covariance requires complete symmetry with respect to all the space and time
coordinates, the equation should be of the first order for the space derivatives as well. If, in
so doing, one takes on arms the principle of superposition, then the conclusion follows—the
wavefunction should satisfy the first-order linear differential equation for all four coordi-
nates. Then with due regard for the Klein-Gordon equation, we lead to the problem that
is already known
( + m2 )ψr = drs (∂) (iγµ ∂ µ − m)sκ ψκ = 0. (15.1)

Its solution necessitates specification of the order for the operator drs (∂). It is natural
that in this situation unity has the greatest attractiveness and Dirac’s choosing fell upon
it. Then it becomes clear that the operator drs (∂) should be looked for in the form

drs (∂) = − (iγµ ∂ µ + m)rs .

Since we have:
(iγµ ∂ µ + m) (−iγν ∂ ν + m) = γµ γν ∂ µ ∂ ν + m2 =

1
= (γµ γν + γν γµ )∂ µ ∂ ν + m2 ,
2

309
© 2011 by Taylor and Francis Group, LLC
 310 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

then Eq. (15.1) is satisfied when the matrices γµ prove to be meeting the anticommutation
relations:
{γµ , γν } = 2gµν . (15.2)
A set of hypercomplex numbers, for which the relation (15.2) is true, forms the Clifford
algebra that exists for any space supplied by the metric gµν . If the space dimension is
n, the dimension of algebra (that is, the number of independent elements involved in the
expansion of any n × n-matrix) is 2n . Note, that considering γµ -matrices, there is no need
to assume that they possess a certain property with respect to the Hermitian conjugation
[11].
By Dirac it was supposed that a relativistic free electron is described by the equation:

(iγµ ∂ µ − m)ψ(x) = 0. (15.3)

It seems that the electron may be equally well described by the following equation:

(iγµ ∂ µ + m)ψ(x) = 0.

However, a physical meaning of the solutions and the conclusions made on their basis are
the same for both equations. This is conditioned by the fact that the relation of Clifford’s
algebra (15.2) is invariant in respect of the transformation:

γµ → −γµ .

The condition (15.2) defines the matrices γµ ambiguously. When the wavefunction ψ is
subjected by the transformation:
ψ ′ = U ψ, (15.4)
and for the matrices γ µ a transformation of similarity is performed:

γµ′ = U γµ U −1 , (15.5)

where U is an arbitrary unitary matrix, the form of both the Dirac equation and relation
(15.2) remains unaltered. Therefore, the γ-matrices are determined to within a unitary
transformation, and a particular representation of these matrices may be chosen by different
ways. At the same time, the opposite is true [12] (the fundamental theorem for the γ-
matrix): if two sets of matrices γµ and γµ′ satisfy the relations (15.2), then there exists the
matrix U relating them by the relation (15.5).
Let us define a rank of the matrices γµ . To this end, we find a determinant of the product
γµ γν for µ 6= ν. Considering that:

γµ γν = −Iγν γµ ,

where I is the unit matrix, we get:

det(γµ γν ) = det(γµ )det(γν ) = det(−I)det(γν )det(γµ ). (15.6)

Since the determinants represent ordinary c-numbers, then from Eq. (15.6) follows

det(−I) = 1,

or
(−1)n = 1.
It is obvious that the value n = 2 is excluded because there are no linearly independent
two-row matrices satisfying the condition (15.2). Linearly independent two-row matrices are

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 311

represented by three Pauli matrices σ1,2,3 and the unit matrix that naturally commutes (and
not anticommutes) with all σ1,2,3 . In case n = 4 there is a possibility to construct matrices
with the desired properties. By a straightforward checking procedure we can see that the
following two choices of the gamma-matrices, namely the Dirac-Pauli representation:
   
k 0 σk 0 I 0
γ = , γ = , (15.7)
−σk 0 0 −I

and Weyl representation (chiral representation):

   
k 0 −σk 0 0 I
γ = , γ = , (15.8)
σk 0 I 0

satisfy all the above-mentioned requirements, being coupled with each other by the unitary
transformation. In both representations γ0 is the Hermitian matrix, whereas the matrices
γk are anti-Hermitian.
It should be noted that the matrix γ5 ≡ γ 5 ≡ iγ 0 γ 1 γ 2 γ 3 is diagonal in the representation
(15.8):  
I 0
γ5 = , (15.9)
0 −I
while in the representation (15.7) it has the form:
 
0 I
γ5 = . (15.10)
I 0

From the γ-matrices we may form 16 linearly independent elements:

Γj = I, γµ , σµν , εµνλσ γ ν γ λ γ σ , γ5 , (15.11)

where
i
σµν = (γµ γν − γν γµ ).
2
Using the same arguments as in the case of the Duffin-Kemmer equation, we can obtain
an equation for the “conjugate” wavefunction:

i∂ µ ψ(x)γµ + mψ(x) = 0, (15.12)

where ψ(x) = ψ † (x)γ 0 . Equations of motion follow from the Lagrangian density:

1 1
L= ψ(x)(iγµ ∂ µ − m)ψ(x) − (i∂ µ ψ(x)γµ + mψ(x))ψ(x), (15.13)
2 2

with independent variations in terms of ψ(x) and ψ(x). It should be noted that the La-
grangian becomes zero when the equations of motion hold.
Multiplying Eq. (15.12) on the right by ψ(x) whereas Eq. (15.3) on the left by ψ(x) and
adding the obtained equations, we get the continuity equation:

∂ µ jµ (x) = 0, (15.14)

where jµ (x), the four-dimensional vector of the current density, is given by the following
expression:
jµ (x) = ψ(x)γµ ψ(x). (15.15)

© 2011 by Taylor and Francis Group, LLC

 312 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

As:
3
X
j0 (x) ≡ ρ(x) = ψ(x)γ0 ψ(x) = | ψi |2 , (15.16)
i=0

the probability density turns to be positively defined, and the single-particle Dirac equation
overcomes one of the barriers that the Klein-Gordon equation could not leave behind. But,
as far as the appearance of negative energies is concerned, this trouble has remained common
for both equations.
We rewrite the Dirac equation (15.3) in the Schrödinger-like form:

i∂t ψ(x) = H ′ ψ(x), (15.17)

where
H ′ = i(α · ∇) + mγ0 , αk = γ0 γk .
The plane de Broglie wave
ψ(x) = ψ(p) exp (−ipx) (15.18)
should be the solution of Eq. (15.3) in a free case. This is true on condition that the
bispinor ψ(p) obeys the equation:

Hψ(p) = p0 ψ(p), (15.19)

with the Hamiltonian function:

H = (α · p) + mγ0 . (15.20)

From (15.19) it follows that for a Dirac particle at rest the equation in eigenvalues is fulfilled:

mγ0 ψ(p) = p0 ψ(p).

With the use of a diagonal representation for the γ0 -matrix (15.7) we come to the conclusion
that we have two solutions with the energy equal to m and two solutions with the energy
equal to −m. To obtain a result in an arbitrary reference frame, we have to use the
solution existence condition for a system of homogeneous linear equations (15.19), that is,
the determinant composed of the coefficients under unknown functions ψ(p) should become
zero. Calculating the determinant, we get:

p20 = p2 + m2 ,

or
p0 = ±Ep ,
p
where Ep = p2 + m2 . Thus, in the Dirac equation the solutions with the negative energy
are also present.
At the same time, the achievements of the Dirac equation are so impressive (it gave
the correct values for: spin, spin magnetic-dipole moment for the electron, fine-structure
of energy levels for hydrogen, etc.) that it was impossible to bury it in oblivion. The
only possibility was to reconstruct its ideology. In 1930 Dirac has put forward a so-called
theory of holes [13]. According to this theory, the vacuum is not actually emptiness, as
we perceive it, being rather a system, where all the negative energy levels have two elec-
trons with differently directed spin projections—similar to the situation in dielectrics with
their valence zone completely filled by the electrons. Since electrons are governed by the
Pauli principle, none of them is able to make a series of transitions from a positive level
to negative-energy levels, falling to the values of E = −∞ and donating infinitely high

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 313

energy in the form of electromagnetic radiation. Following Dirac’s example, the vacuum
becomes a complex system, in a sense including a part of the universe—electrons occupying
all the levels with negative energy. Dirac’s vacuum is no longer a complete emptiness of
Democritus; it becomes a prototype of the vacuum of the modern quantum field theory.
What happens when an electron from a negative-energy level is excited to go to the
positive-energy one? A vacant place, referred to as a hole, with the energy −Ep should be
immediately filled by the electron having the positive energy Ep and emitting the energy
2Ep in the process. Then a hole is considered as the absence of the electron or presence
of the particle with the positive charge +|e|. Considering that normally all the negative-
energy levels are filled, a hole is associated with a decrease of the negative or increase of
the positive energy. So, a hole behaves like a positively charged particle with the positive
energy. If such a particle is found, it may be filled by the positive-energy electron. Such
a transition should represent the process when positively- and negatively-charged particles
mutually destroy each other. At first Dirac posited that a positively charged “electron”
(antielectron) was represented by the proton. As atoms include both electrons and protons,
this statement could be verified without difficulty. It suffices to calculate a probability of
the transition: electron+hole→vacuum,
derive the lifetime of the normal matter tm , and compare the results obtained with the
experiment. The value of tm calculated by Oppenheimer [14] was on the order of 10−10 s,
that, to put it mildly, is at variance with the stability of our world. Fortunately, soon all
these problems were solved due to the experimental results. C. Andersen [15] has reported
about the discovery of a particle in cosmic rays, possessing the same mechanical charac-
teristics (spin, mass, etc.) as an electron but having the electromagnetic characteristics
equal in the value and opposite in sign. This particle is actually an antielectron and later
it has been called the positron. The prediction of the positron, and antiparticles in general,
was a historic triumph for the Dirac theory. Symmetry between the matter and antimatter
was for the first time accepted in physics to become one of the major ideas in a theory of
elementary particles.
Presenting the multiple-particle formalism, the holes theory leads to significantly compli-
cated calculations in practice. To illustrate, a wavefunction describing at least one electron
should take into account all the filled states with the negative energy. Though the holes
theory is close in spirit to the quantum field theory we still have to refuse it. And this
is caused not only by the elaborate computations. For example, in the case of bosons the
holes theory is useless because the reasoning of Dirac was based on the Pauli principle. On
the other hand, even for fermions, the concept of the unobserved infinite charged sea is
a bit mystical. Because of this, the use of the second quantization to interpret the wave
equations of the quantum theory is physically more reasonable.

15.2 Calculating γ-matrix tracks

As the Dirac equation is invariant under the transformations (15.5) and (15.6), the dynamic
observables should be also invariant with respect to these transformations. In other words
all the products of γ-matrices in the dynamic observables should be reduced to the tracks
(spurs), which are the only invariants relative to the transformation of similarity (15.6).
Let us find the formulae to calculate the spurs of the γ-matrices. First, we demonstrate
that the spur of the odd number of these matrices is zero. For simplicity, we perform this

© 2011 by Taylor and Francis Group, LLC

 314 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

with the help of the matrix γ5 . Using (14.145) and

{γµ , γ5 } = 0, γ52 = 1,

we get:

Sp {γµ1 γµ2 .....γµn } = Sp {γ5 γ5 γµ1 γµ2 .....γµn } = Sp {γ5 γµ1 γµ2 .....γµn γ5 } =

= (−1)n Sp {γµ1 γµ2 .....γµn } = 0. (15.21)

If this product contains an even number of matrices that is equal to n, then the permuta-
tion relations enable us to reduce the spur of this product to a sum of the product spurs
containing n − 2 matrices only. For instance, for n = 2 we have:

Sp (γµ γν ) = Sp (−γµ γν + 2gµν ).

whence:
Sp (γµ γν ) = 4gµν . (15.22)
Next, we find Sp (γµ γν γλ γσ ). Taking account of:

γµ γν γλ γσ = 2gµν γλ γσ − γν γµ γλ γσ , 
γν γµ γλ γσ = 2gµλ γν γσ − γν γλ γµ γσ , (15.23)

γν γλ γµ γσ = 2gµσ γν γλ − γν γλ γσ γµ ,

we obtain:
γµ γν γλ γσ = 2gµν γλ γσ − 2gµλ γν γσ + 2gµσ γν γλ − γν γλ γσ γµ . (15.24)
In this way the final result is as follows:

Sp (γµ γν γλ γσ ) = 4gµν gλσ − 4gµλ gνσ + 4gµσ gνλ . (15.25)

To find the spur of the arbitrary number of matrices, we use the equation:
X
Sp {γµ1 γµ2 .....γµ2n } = 4 (−1)N gνλ gσκ ...., (15.26)

where ν, λ, σ, κ, .... is a particular combination of the indices µ1 , µ2 , ....., µ2n with the follow-
ing prescription [16]. Each matrix γµi is associated with a point of a circle, and the matrices
have the same order as in the left-hand side of the equality (15.26). Next, all the points are
pairwise joined by the straight lines, and each line joining the points, for instance, ν and
λ is associated with the factor gνλ . All possible connections of the points are taken, each
connection in the sum (15.26) being associated with the term: (−1)N gνλ gσκ .... where N is
the number of intersection points for the lines. From (15.26) we can easily derive:

Sp (γ5 γµ γν γλ γσ ) = 4iεµνλσ . (15.27)

15.3 Relativistic covariance

Let us investigate the properties of the Dirac wavefunction ψ(x) under the Poincare trans-
formations:
x′µ = Λνµ xν + aµ .

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 315

In a new reference system ψ(x) is determined by the relation:

4
X
ψr′ (x) = Srκ (Λ)ψκ [Λ−1 (x − a)], (15.28)
κ=1

where the matrix indices are written in the explicit form. Carrying out the similar reasoning
as in the case of the Duffin-Kemmer equation, we lead to the conclusion that the Dirac
equation is relativistically covariant if and only if the following condition is fulfilled:

S −1 (Λ)γµ S(Λ) = Λνµ γν . (15.29)

Evidently, the matrices γµ′ = Λνµ γν are also satisfying the relations of Clifford’s algebra. A
system of Eqs. (15.29) is determining the matrix S to within the numerical factor, whose
value is significant for finding of the matrix that transforms the wavefunction ψ. This factor
is calculated as follows. Performing the Hermitian conjugation procedure for both sides of
the relation (15.29), we have

3
!† 3
†
X X
−1 µ †
(S γ S) = (Λµν γ ν ) = µ0 0
Λ γ − µl l
Λ γ = Λµ0 γ 0 + Λµl γ l . (15.30)
l=1 l=1

Multiplying (15.30) by γ0 on the left and on the right, and considering that the properties
of γ µ with respect to the Hermitian conjugation may be written in the compact form:

(γ µ )† = γ 0 γ µ γ 0 ,

we get: !
3
X
0 µ0 0 µl l
γ Λ γ + Λ γ γ 0 = Λµν γ ν = S −1 γ µ S (15.31)
l=1

and
γ 0 (S −1 γ µ S)† γ 0 = (γ 0 S † γ 0 )γ µ (γ 0 S † γ 0 )−1 . (15.32)
Combining the obtained relations, as a result we have:

(Sγ 0 S † γ 0 )γ µ (Sγ 0 S † γ 0 )−1 = γ µ ,

that is, the matrix Sγ 0 S † γ 0 commutes with all the matrices γ µ and hence it is divisible by
the unit matrix:
Sγ 0 S † γ 0 = aI. (15.33)
Due to the Hermitian properties of γ 0 , the constant a is real. Later it will be shown that
for the matrix S the normalization det S = 1 is always possible. In this case a4 = 1 and
a = ±1. Next, we consider the identity:
3
!
X
S † S = S † γ 0 γ 0 S = aγ 0 S −1 γ 0 S = aγ 0 Λ0ν γ ν = a Λ00 I − Λ0l γ 0 γ l . (15.34)
l=1

As the matrix S is nonsingular and nonzero, the matrix S † S > 0 and its eigenvalues are
real and positive numbers, that is, Sp (S † S) > 0. Then, taking account of Sp (γ 0 γ l ) = 0,
we get:
Sp (S † S) = 4aΛ00 > 0. (15.35)

© 2011 by Taylor and Francis Group, LLC

 316 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

As follows from (15.35), at orthochronous transformations (Λ00 ≥ 0)

a = 1,

while at the nonorthchronous ones (Λ00 ≤ 0)

a = −1.

Using the obtained results, we can state a law for the transformation of the conjugate
bispinor ψ. And taking into consideration (15.33), we have:
′
ψ = (ψ ′ )† γ 0 = ψ † S † γ 0 = aψ † γ 0 S −1 = aψS −1 . (15.36)

Let us demonstrate how to work the definition scheme of the matrix S for different types
of the Poincare transformations. We begin with the improper transformations. According
to (15.29), the matrix S(P ) of the space inversion should satisfy the relations:

S −1 (P )γ 0 S(P ) = γ 0 , S −1 (P )γ k S(P ) = −γ k . (15.37)

Consequently, for S(P ) we can choose ±γ 0 or ±iγ 0 .

It is not so easy to find the matrix for the time inversion with the help of (15.35) as the
time-reversal operation is described by the antiunitary operator (see, Section 5.2). The def-
inition of this matrix, as well as the selection of phase in the space-inversion transformation,
will be treated after the consideration of the charge conjugation operation.
Let us establish the form of S matrix at four-dimensional rotations. We consider the case
of infinitesimal transformations:

x′µ = xµ + ǫωµν xµ , (15.38)

where ǫ ≪ 1 and ωνµ = −ωµν . The representation associated with the transformation
(15.38) is infinitesimally different from the identity one, in the first-order with respect to ǫ
being of the form:

S (I + ǫω) = I + ǫG, S −1 (I + ǫω) = S (I − ǫω) = I − ǫG. (15.39)

Substituting (15.39) into (15.29), we obtain

S −1 γµ S = (I − ǫG)γµ (I + ǫG) = γµ + ǫ(γµ G − Gγµ ) = Λνµ γν = γµ + ǫωµν γν . (15.40)

Because of this, the matrix G should have the following property:

γµ G − Gγµ = ωµν γν . (15.41)

Normalization det S = 1 requires det(I + ǫG) = 1 + ǫSp G = 1, that is, Sp G = 0. By the

addition of a matrix being divisible by the unit matrix to G, we can always reduce the spur
of G to zero. The fulfillment of (15.41) and Sp G = 0 is guaranteed, provided the matrix G
is determined by:
i
G = ω µν σµν . (15.42)
4
In the case of a infinitesimal space rotation by an angle ǫ about the axis xi , it is clear that:

ωjk = ǫεjki ni ,

where ni is an unit vector. According to (15.42), the transformation generator in this case
is as follows:
i
G(ni ) = Σi ni , (15.43)
2

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 317

where  
1 σi 0
Σi = εijk σjk = (15.44)
2 0 σi
(note that the matrix Σi is diagonal in any representation of the γ-matrix). Proceeding
from all the above, we can write:
 
∂
ψ(xj ) = S(ǫ, ni )ψ(xj − ǫεjki ni xk ) ≈ S(ǫ, ni ) 1 − ǫεjki ni xk ψ(xj ) =
∂xj
  
1
= 1 + iǫni Σi + Mi ψ(xj ), (15.45)
2
where Mi = −iεikj xk ∂j is a projection of the angular momentum on axis xi .
Let the rotation be realized about the given axis, with the direction characterized by the
unit vector n at a finite angle θ. Then the matrix associated with a particular wavefunction
transformation is given by:
     
θ θ θ
S(θ, n) = exp i(n · Σ) = cos + i(n · Σ) sin , (15.46)
2 2 2
where (to go from the exponent to trigonometric functions) we use the Taylor series expan-
sion and properties of the γ-matrices. From (15.46) it follows that:

S(θ + 2π) = S(θ),

that is, under the rotation by 2π a sign of the wavefunction is changing. This, in its turn,
points to the fact that we are dealing with spinor objects. Since a wavefunction of the Dirac
equation is a four-component quantity, this function is represented by two spinors, that is,
it is a four-dimensional spinor. Similar objects will be further referred to as bispinors.
Obviously, the operator Σ plays a role of the Dirac-particle spin operator. Considering
 
1 0 0 0
3 0 1 0 0
Σ2 =  
4 0 0 1 0
0 0 0 1
and taking into account that the eigenvalues of the operators specified in the form of diagonal
matrices are coincident with the values of diagonal elements, we obtain J(J + 1) = 3/4.
Thus, the Dirac equation is really describing the particles with the spin J = 1/2.
Now we consider the Poincare transformations relating different inertial frames of ref-
erence. Let a relative motion of the systems be along the axis x1 , that is, we have an
infinitesimal rotation through the imaginary angle iǫ in the plane (x0 x1 ). Only two com-
ponents of ω µν , ω 01 = −ω 10 = 1, are different from zero, and we have
i 1 1
G[10] = σ10 = − γ1 γ0 = α1 .
2 2 2
The finite transformation is associated with the matrix:
 ϕ ϕ ϕ
S(ϕ, n1 ) = exp α1 = cosh + α1 sinh , (15.47)
2 2 2
where tanh ϕ = v. Provided the motion is not along the axis x1 , but along an arbitrary
direction characterized by the unit vector n, then we have to replace the expression (15.47)
by: h ϕi ϕ ϕ
S(ϕ, n) = exp (n · α) = cosh + (n · α) sinh . (15.48)
2 2 2

© 2011 by Taylor and Francis Group, LLC

 318 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The final formula determining a law of the bispinor transformation at four-dimensional

rotations is as follows:
  
′ −1 i µν 1
ψ (x) = S(Λ)ψ(Λ x) = exp ω σµν + Lµν ψ(x), (15.49)
2 2

where Lµν = i(xµ ∂ν −xν ∂µ ) is the angular momentum and we have taken into consideration
that pµ = (E, −p).
If we use the chiral representation for the γ-matrices, in (15.46) and (15.48) we can rec-
ognize the equations for transformations of the first- and second-kind spinors, well-known
from Section 3.4. Let us recall the course of our previous reasoning:
Dirac equation→relativistic covariance→transformation laws of spinors.
It goes without saying that the reverse is also possible, namely: proceeding from the trans-
formational properties of spinors with respect to the transformations of the Poincare group,
we can come to the Dirac equation. Note that this is a distinctive feature of any equation
for the particles with nonzero spin as opposed to the Klein-Gordon equation that represents
only the relativistic relations between the energy, momentum, and mass. One will be very
instructive to illustrate such a way of obtaining an equation taking the Dirac equation as
an example.
In the chiral representation the bispinor ψ comprises the spinors ξ and η, transformed by
the representations (1/2,0) and (0,1/2), respectively. According to (15.48), the spinors ξ(0)
and ξ(p) are related by the boost transformation:
h ϕi h ϕ  ϕ i
ξ(p) = exp (n · σ) ξ(0) = cosh + (n · σ) sinh ξ(0) =
2 2 2
"s  s #
β+1 β−1 p0 + m + (p · σ)
= + (n · σ) ξ(0) = p ξ(0), (15.50)
2 2 2m(p0 + m)
where we introduce β = p0 /m and allow for:
ϕ rβ + 1 ϕ r
β−1
|p| · n = p, cosh = , sinh = .
2 2 2 2
Similarly, using the explicit-form Lorentz boost, for the spinor η(p) we get:

p0 + m − (p · σ)
η(p) = p η(0). (15.51)
2m(p0 + m)

Recall that in the canonical basis the representation is characterized by eigenvalues of the
operators:
1
m2 , S2, pi , S3 = wµ n(3) µ ,
m
that is, the representation basis comprises the vectors | +, m, J; p, J3 >. It means that in a
system of the particle at rest the state of the particle with the spin J = 1/2 may be equally
well described both by the spinor ξ(0) and spinor η(0), that is, the following:

ξ(0) = η(0)

takes place. Acting on Eq. (15.51) by the operator:

p0 + (p · σ)
m

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 319

and taking into account Eq. (15.50), we obtain:

p0 + (p · σ)
ξ(p) = η(p). (15.52)
m
Similar operations with (15.50) lead to:
p0 − (p · σ)
η(p) = ξ(p). (15.53)
m
From the derived equations it follows that one of the spinors η(p) or ξ(p) may be chosen
at will, whereas the second is expressed in terms of the first spinor. In other words, the
bispinor describing the particle with the spin 1/2 has two linearly independent components
only. For the bispinor:  
ξ(p)
ψ(p) =
η(p)
a system of Eqs. (15.52)–(15.53) takes the form:

(p̂ − m)ψ(p) = 0. (15.54)

So we have demonstrated that the Dirac equation actually represents the relation between
the spinors transformed in terms of the representations (1/2,0) and (0,1/2).
In the chiral representation the spinors comprising the bispinor ψ are independently
transformed under the Poincare transformations, that is the main advantage of this trans-
formation. We introduce the projective operators as follows:
1 + γ5 1 − γ5
ΛR = , ΛL = . (15.55)
2 2
For the states:
ψR,L = ΛR,L ψ,
we have the equations for eigenvalues of the operator γ5 subsequently referred to as the
chirality operator
γ5 ψR,L = ±ψR,L .
Using the projective operators of the right (positive) chirality ΛR and left (negative) chirality
ΛL , from the bispinor ψ we can obtain the two-component Weyl spinors:
 
ξ
ψR = , (15.56)
0

and  
0
ψL = . (15.57)
η
Another very popular representation of the γ-matrices, the Dirac-Pauli representation, is
very useful in the analysis of the nonrelativistic limit for the Dirac equation. We show that
this is actually so. Represent the bispinor ψ(p) in the form:
 
ϕ(p)
ψ(p) =
χ(p)

Besides, for definiteness, we consider the case of positive energies. Then Eq. (15.54) is
broken into a system of two matrix equations:

(σ · p)χ(p) + mϕ(p) = Ep ϕ(p), (15.58)

© 2011 by Taylor and Francis Group, LLC

 320 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

(σ · p)ϕ(p) − mχ(p) = Ep χ(p). (15.59)

From (15.58) and (15.59) we find:

(σ · p)
χ(p) = ϕ(p), (15.60)
Ep + m2

where Ep + mc2 6= 0 and we for some time return to the normal units. Since for v ≪ c we
have
c2 p2 1  v 2 †
χ† (p)χ(p) = 2 2
ϕ† (p)ϕ(p) ≈ ϕ (p)ϕ(p),
(Ep + mc ) 4 c
in the nonrelativistic limit the components χ(p) in the order of magnitude are equal to
(v/c)ϕ(p) and hence are small. Because of this, the lower and upper bispinor components
in the Dirac-Pauli representation are identified as big and small components, respectively.
For m = 0 Eqs. (15.52) and (15.53) are broken into two equations called the Weyl
equations [17], having the following form in the nonstationary case:

i∂t η(p) = −(p · σ)η(p),
(15.61)
i∂t ξ(p) = (p · σ)ξ(p).

Having studied the properties of Dirac’s spinors under the Poincare transformations, we
can define a geometric character of the bilinear combinations constructed on the basis of
these spinors. So, the quantity ψψ represents a scalar with respect to the orthochronous
transformations as we have:
′
ψ ψ ′ = ψS −1 Sψ = ψψ,
and in the case of the nonorthochronous transformations it behaves itself as a pseudoscalar.
Similarly, the bilinear combination ψγ µ ψ is transformed as a four-dimensional vector under
the orthochronous transformations:
′
ψ γ µ ψ ′ = ψS −1 γ µ Sψ = Λµν ψγ ν ψ (15.62)

and as a four-dimensional pseudoscalar in the case of the nonorthochronous transformations:

′
ψ γ µ ψ ′ = −Λµν ψγ ν ψ. (15.63)

It is easy to establish that the bilinear combinations: ψσ µν ψ and ψγ µ γ ν γ λ ψ (µ < ν < λ)

represent the second- and third-rank tensors, respectively. Using the matrix γ5 , from the
above-mentioned quantities we can compose the pseudoquantities in respect to the space
reversal operation. For example, the bilinear combination ψγ5 ψ is transformed as:

′ i µ ν λ σ
ψ γ5 ψ ′ = aψS −1 γ5 Sψ = −a Λ Λ Λ Λ εµνλσ ψγ α γ β γ ǫ γ κ ψ = a(det Λ)ψγ5 ψ,
4! α β ǫ κ
where we have used
i
γ5 = − εµνλσ γ µ γ ν γ λ γ σ . (15.64)
4!
Knowing the transformational properties of the bilinear combinations ψΓj ψ, we can con-
struct the invariant quantities to describe various interactions between the Dirac particles.
For example, an invariant, describing the electromagnetic interaction of the Dirac particles,
may be obtained through the multiplication of ψγ µ ψ with the electromagnetic-field vector
potential Aµ . Besides, interaction with an electromagnetic field may be governed by the
quantity ψσ µν ψFµν (Pauli interaction for the magnetic moment).

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 321

15.4 Solutions of the free Dirac equation

Considering that any free particle should be described by the de Broglie wave, the solutions
for the Dirac equation should be of the form:

ψ (+) (x) = u(p) exp (−ipx) for positive energy, (15.65)

ψ (−) (x) = v(p) exp (ipx) for negative energy. (15.66)

To substitute (15.65) and (15.66) into the Dirac equation gives the equations for definition
of the bispinors u(p) and v(p):
(p̂ − m)u(p) = 0, (15.67)
(p̂ + m)v(p) = 0. (15.68)
We establish a procedure that makes it possible to distinguish the states of particles and
antiparticles. It seems that this may be realized with the use of the operator:
H
Πǫ = √ ,
H2
that commutes with the Hamiltonian. It is evident that the operator Πǫ is Hermitian and
unitary. In the momentum representation it has the simple form:

(α · p) + mγ0
Πǫ = . (15.69)
Ep

Since Π2ǫ = 1, then the eigenvalues of this operator are as follows:

p0
ǫ= = ±1.
Ep

Then, using the operator Π+ , we could derive the electron states from the solutions of
the Dirac equation, while the Π− -operator could be used for the finding of the positron
components of the field function. Such a procedure is quite acceptable in the nonrela-
tivistic theory. Unfortunately, the operator Πǫ could not be used because its form is not
relativistically covariant. Let us try to look for the covariant projection operators instead.
Considering Eqs. (15.67), (15.68), we come to the conclusion that the projection operator
to the electron states may be represented as:
p̂ + m
P+ = , (15.70)
2m
because, due to (15.67) and (15.68), it features the following properties:

P+ u(p) = u(p), P+ v(p) = 0. (15.71)

If we are interested in the negative-energy states, then in a similar way we can define the
operator:
−p̂ + m
P− = , (15.72)
2m
which has the properties:

P− v(p) = v(p), P− u(p) = 0. (15.73)

© 2011 by Taylor and Francis Group, LLC

 322 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Since the standard relations for the projection operators:

Pǫ Pǫ′ = δǫǫ′ , P+ + P− = I, (15.74)

where
ǫp̂ + m
Pǫ = , (15.75)
2m
are fulfilled, then P+ and P− are actually the projection operators to be found.
Eqs. (15.67) and (15.68) give no unique definitions for the bispinors u(p) and v(p) as
each of the equations has two linearly independent solutions. It is clear that these solutions
should conform to two possible values of the electron polarization. The polarization state of
a nonrelativistic electron may be characterized by setting the direction, the spin projection
onto which has a well-defined value. This approach is inapplicable for a relativistic electron
as the spin projection operator onto the arbitrary direction given by the unit vector n
commutes with the Hamiltonian only in the case when n k p. To describe the polarization
state of a relativistic electron, we use the Pauli-Lubanski pseudovector:
1
Wµ = − εµνλρ σ νλ pρ , (15.76)
4
where we have taken into consideration that the total momentum is given by the following
expression:
1
J νλ = σ νλ + M νλ .
2
Next, we introduce the four-dimensional vector p̃ (very useful for studies of polarization
phenomena) that is characterized by

p̃2 = −m2 , pµ p̃µ = 0. (15.77)

In a system of the particle at rest the vector p̃µ is of the form:

p̃µ = m(0, n). (15.78)

Then in the moving reference frame with p k n it has the components:

 
p0 p
p̃ = | p |, . (15.79)
|p|
For the vector nµ = p̃µ /m the following relation:
1
Wµ nµ = − γ5 n̂p̂ (15.80)
2
is true (recall, that ε0123 = −1). Since

[Wµ nµ , p̂] = −γ5 nµ pµ p̂ = 0,

then the operators Wµ nµ and p̂ have the common total eigenfunctions system. In other
words, the bispinors u(p) and v(p), apart from Eqs. (15.67) and (15.68), must satisfy the
equations: 
′ ′
−m−1 Wµ nµ u(α ) (p) = α′ u(α ) (p),
′ ′ (15.81)
−m−1 Wµ nµ v (α ) (p) = α′ v (α ) (p).
Because (m−1 Wµ nµ )2 = 1/4, then it is clear that
1
α′ = ± .
2

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 323

So, two linear independent solutions of Eqs. (15.67) and (15.68) are characterized by the
eigenvalues of the operator −Wµ nµ /m:

Wµ nµ (α) 1 1
− u (p) = γ5 n̂u(α) (p) = αu(α) (p), (15.82)
m 2 2
Wµ nµ (α) 1 1
− v (p) = − γ5 n̂v (α) (p) = αv (α) (p), (15.83)
m 2 2
where α = ±1. As follows from the derived relations, the operator selecting the states with
a defined value of α for the electrons takes the form:
1
Λ(α) (n) = (1 + αγ5 n̂), (15.84)
2
And the corresponding operator for the positrons is derived from (15.84) with the substi-
tution:
α → −α.
Let us clarify the physical meaning of the quantum number α. Start with the simplest
situation, namely, with the rest particle case. For the bispinors we accept the following
orthonormalization condition:
′ Ep
w(κ)† (p)w(κ ) (p) = δκκ′ , (15.85)
m
where
w(1) (p) = u(+) (p), w(2) (p) = u(−) (p),
w(3) (p) = v (+) (p) w(4) (p) = v (−) (p),
while u(+) (0) (v (+) (0)) and u(−) (0) (v (−) (0)) describe the states of the electron (positron)
with the spin being parallel and antiparallel to the vector n, respectively. Under fulfill-
ment (15.85) four linearly independent solutions of the Dirac equation in the Dirac-Pauli
representation for the rest particle take the form:
 ±  
ς 0
u(±) (0) = N0  0  , v (±) (0) = N0  0  , (15.86)
0 ς±

where
   
1 + n3 + 1 − −n1 + in2
N0 = p , ς = , ς = . (15.87)
2(n3 + 1) n1 + in2 n3 + 1

Supposing p = 0, we obtain for the operator −Wµ nµ /m:

 
Wµ nµ 1 1 1 σ 0
− = γ5 γ 0 γ = Σ = . (15.88)
m 2 2 2 0 σ

Owing to the use of this operator for the states of (15.86), it is clear that in a primary
quantization theory α/2 has the meaning of an eigenvalue of the spin projection operator
onto the direction n.
For a moving particle we have:
2
ˆp̃p̂ = − m (p · γ)γ0 . (15.89)
|p|

© 2011 by Taylor and Francis Group, LLC

 324 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Then, taking into account Eqs. (15.82), (15.83), we arrive at the relations:

1 (α) 1 (α)
S(p)u(α) (p) = αu (p), S(p)v (α) (p) = αv (p), (15.90)
2 2
where
(Σ · p)
S(p) =
|p|
represents the chirality operator, which commutes with the Hamiltonian. At p 6= 0 the
number α/2 is again having the meaning of an eigenvalue for the spin projection operator,
however this time onto the direction of particle motion. In this case it is customary to speak
of the state with a definite helicity, positive for α = 1 or negative for α = −1. Provided
the particle spin is directed opposite to its momentum (similar to the left-hand screw), this
particle is defined as a particle of the left-handed helicity or left-hand polarization. When
the spin is in the direction of the particle momentum, we can speak about a right-hand
polarized particle. Thus, the operators projecting onto the electron states with the spin
projection 1/2 and −1/2 are as follows:

Λ(+) (n) ≡ Λ(α=1) (n) and Λ(−) (n) ≡ Λ(α=−1) (n). (15.91)

By contrast, in the case of the positron the operators Λ(−) (n) and Λ(+) (n) separate the states
with the spin projection onto the direction n that is equal to +1/2 and −1/2, respectively.
For p 6= 0 the solutions for the Dirac equation may be derived from the solutions of (15.86)
with the help of Lorentz transformations to a system moving at the velocity p/Ep . Having
written the formula (15.48) in the Dirac-Pauli representation
 
cosh(ϕ/2) (n · σ) sinh(ϕ/2)
S(ϕ, n) = =
(n · σ) sinh(ϕ/2) cosh(ϕ/2)
 p p 
(Ep + m)/2m (n · σ) (Ep − m)/2m
= p p , (15.92)
(n · σ) (Ep − m)/2m (Ep + m)/2m
we obtain the solution sought:
 
(±) (±) p̂ + m ς±
u (p) = S(ϕ, n)u (0) = p u(±) (0) = N (p·σ ) ± , (15.93)
2m(Ep + m) Ep +m ς
 (p·σ ) ± 
−p̂ + m Ep +m ς
v (±) (p) = p v (±) (0) = N ±
, (15.94)
2m(Ep + m) ς
where s
Ep + m
N= .
4Ep (n3 + 1)
Thus, we have established that in a primary quantization theory the field functions, de-
scribing the states with a definite sign of the energy, and the spin projection are defined as
follows:
u(α) (p) = P+ Λ(α) (n)ǫ(p), v (α) (p) = P− Λ(−α) (n)ǫ(p), (15.95)
where ǫ(p) is an arbitrary bispinor satisfying only the normalization condition ǫ(p)ǫ(p) = 1.
Accepting of the orthonormalization condition (15.85) for Dirac spinors was dictated by the
requirement of relativistic invariance because in this case both the left-hand and right-hand
sides of relation (15.85) are transformed as the fourth component of vector. The advantages

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 325

of such an orthonormalization are quite obvious under the transition to conjugate spinors.
Multiplying Eq.
p0 w(κ) (p) = [(α · p) + γ0 m]w(κ) (p) (15.96)
by w(κ)† (p)γ0 on the left, while Eq.

p0 w(κ)† (p) = w(κ)† (p)[(α · p) + γ0 m] (15.97)

by γ0 w(κ) (p) on the right and summarizing the results obtained, we get
m (κ)†
w (κ) (p)w(κ) (p) = w (p)w(κ) (p) (15.98)
p0
or ′
w(κ) (p)w(κ ) (p) = ε(p0 )δκκ′ . (15.99)
Using the explicit form of the Dirac equation solutions (15.93) and (15.94), one may show
that they satisfy the relations:

X 4
X
[u(α) (α) (α) (α)
r (p)us (p) − vr (p)v s (p)] = ε(p0 )wr(κ) (p)w (κ)
s (p) = δrs , (15.100)
α κ=1

where r and s are matrix indices. The N order of factors in (15.100) corresponds to the
direct product of bispinors w and w (w w), which represents the 4 × 4-matrix. Note that
(15.100) may be proved without the use of the solutions in the explicit form. With this aim
in view we recall that the bispinors w(κ) (p) form a complete system, and any bispinor U (p)
may be resolved into them, that is, we have:
4
X 4
X
U (p) = aκ w(κ) (p) = w(κ) (p)aκ (15.101)
κ=1 κ=1

′
(aκ is an ordinary c-number!). To multiply (15.101) by w(κ ) (p) on the left and allow for
(15.99) gives
′
aκ′ = ε(p0 )w (κ ) (p)U (p).
Substituting the obtained equation into (15.101), we arrive at the relation (15.100).
With the use of relation (15.100) we can perform commonly occurring operations of the
summation and averaging over the spin states of an electron or positron. Let us suppose
that we are concerned with the calculation of the expression:

2 2 4
! 4 
X X X X
(f M w(κ) )(w (κ) N g) = f r Mrs ws(κ)  w (κ)
ǫ Nǫβ gβ
=
κ=1 κ=1 r,s=1 ǫ,β=1

4 2
!
X X
= f r Mrs ws(κ) w(κ)
ǫ ) Nǫβ gβ , (15.102)
r,s,ǫ,β=1 κ=1

where M and N are some operators (products of the γ-matrices), f and g are bispinors
and, as may be seen, the sum is taken only over the electron states. To multiply (15.100)
by the operator P+ on the left and right gives the relation to be necessary for definition of
(15.102):
X 2
wr(κ) (p)w (κ)
s (p) = (P+ )rs . (15.103)
κ=1

© 2011 by Taylor and Francis Group, LLC

 326 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

When the summation of (15.102) is over κ = 3, 4 that is, over the positron states, we have
to use the following equation:
4
X
wr(κ) (p)w (κ)
s (p) = −(P− )rs , (15.104)
κ=3

which also follows from (15.100).

Finally, we consider zero-mass particles. This may be of particular interest because in
the ultrarelativistic limit massive particles actually behave themselves as massless objects.
And for the particles of so small a mass as a neutrino the energies on the order of a few kV
are considered as ultrarelativistic. Unfortunately, there is no possibility to derive a solution
for the massless Dirac equation:
ˆ
i∂ψ(x) = 0, (15.105)
from the solutions of (15.92) and (15.93) by a simple changeover m → 0. This is caused
by the fact that now, as distinct from the massive particles, due to commutation of the
ˆ the field function is an eigenfunction of the chirality operator:
operators γ5 and ∂,

γ5 ψq (x) = qψq (x),

where q = ±1. Consequently, the independent solution of Eq. (15.105) must be represented
in the following form:

uq (p) exp (−ipx) for positive energy,
ψ(x) = (15.106)
vq (p) exp (ipx) for negative energy.

To multiply (15.105) by γ5 γ0 for the case E > 0 leads to the equation:

(p · Σ)uq (p) = q|p|uq (p), (15.107)

from which follows that the chirality is equal to the helicity of a particle. As for the
negative energy solutions, the chirality and helicity have the opposite signs. In the Dirac-
Pauli representation the bispinors uq (p) and vq (p) have the form:
   
1 gq (p) 1 dq (p)
uq (p) = √ , vq (p) = √ , (15.108)
2 qgq (p) 2 qdq (p)
where gq (p) and dq (p) satisfy the eigenvalue equations

(n · σ)gq (p) = qgq (p), (n · σ)dq (p) = qdq (p) (15.109)

and n = p/|p|. Eqs. (15.108) and (15.109) lead us to the idea that the functions uq (p)
and vq (p) may be distinguished by the phase factor only. Let us show that this is the case.
Rewrite the chirality operator in the spherical coordinate system

(n · σ) = σx sin θ cos ϕ + σy sin θ sin ϕ + σz cos θ =

 
cos θ sin θe−iϕ
= . (15.110)
sin θeiϕ − cos θ
Using (15.110), we find the eigenfunctions of the chirality operator for E > 0
   
cos(θ/2) − sin(θ/2)e−iϕ
g+ (p) = , g− (p) = . (15.111)
sin(θ/2)eiϕ cos(θ/2)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 327

From the relations (15.111) follows

∗
gq (p) = iqσ2 g−q (p). (15.112)

Obviously, at the charge conjugation operation the Dirac bispinors behave as follows:
c
uq (p) = v−q (p) = Cv T−q (p), vq (p) = uc−q (p) = CuT−q (p).

Later we shall show that in the Dirac-Pauli representation the charge conjugation matrix
has the form:  
0 −iσ2
C= .
−iσ2 0
Then, it is easy to show the fulfillment of the following relations:
 ∗
 
−iσ2 g− (p) = −g+ (p)
v+ (p) = CuT− (p) = √12

∗ = −u + (p), 

iσ2 g− (p) = g+ (p)
(15.113)


v− (p) = CuT+ (p) = −u− (p).


Thus, when m = 0 for the given momentum value, an equation for a particle with the spin
of 1/2 has just two independent solutions. In other words, in this case a fermion is described
by the two-component spinors with a specific chirality value, the Weyl’s spinors ψL and ψR .
The Dirac equation therewith is equivalent to a pair of the Weyl equations (15.61). Recall
that the necessity to use bispinors for particles with the spin of 1/2 is determined by the
requirement of the theory invariance with respect to the space inversion operation. When a
particle is described by a single spinor, this symmetry is lost. And in the case of a neutrino
this argument causes no concern. As is known at present, neutrinos take part in the weak
interaction to be not invariant with respect to P. The Lagrangian construction for the weak
interaction is so that it involves the field function of a neutrino preceded by the operator
ΛL .
Considering that normalization is dictated by the form of an equation of motion, Lorentz-
invariant normalizations of the solutions (15.87) obtained for the nonzero masses in the case
of massless particles take the following form:

u(α) (p)γ0 u(β) (p) = (2m)−1 u(α) (p){p̂, γ0 }u(β) (p) = 2Ep δαβ ,
(15.114)
v (α) γ0 (p)v (β) (p) = 2Ep δαβ .

15.5 Once more about particles spin

The idea that an electron possesses spin (spinning electron) was first put forward by R.
Kronig in 1921 for explanation of the anomalous Zeeman effect. However, as fate has willed,
the first to listen to him was V. Pauli. Pauli’s reaction to this hypothesis was negative.
Being dispirited by the opposition of Pauli and cold reception given to the spinning electron
idea by his colleagues, Kronig did not dare to publish a paper with his hypothesis. Four
years later fate appeared to be more favorable to the graduates of the Leiden University S.
Gaudsmit and G.E. Uhlenbeck. P. Ehrenfest was the first listener of their idea, according
to which an electron represents a spinning sphere with a constant angular momentum.
Ehrenfest has immediately recommended them to make a relevant publication in the journal
Naturwissenschaften. By this concept, a spinning electron should possess an intrinsic dipole

© 2011 by Taylor and Francis Group, LLC

 328 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

magnetic moment µs and intrinsic angular momentum S, that is called the spin moment,
or simply the spin. In an atom the orbital moment of an electron may be compared to the
angular momentum of the Earth upon its revolution around the Sun, accompanied by the
seasonal changes. And the spin moment may be compared with the angular momentum of
the Earth rotating about its axis, accompanied by the alternation of day and night.
However, the calculations of Uhlenbeck have demonstrated that for the required intrinsic
mechanical moment (S2 = 3~2 /4) a linear velocity of the rotational motion of an electron
should be higher than the speed of light. Confused by the obtained result, Uhlenbeck was
willing to address the editorial board and recall the paper. When he informed Ehrenfest
about his decision Ehrenfest had remarked: “It is a bit late. Besides, you both are young
enough to afford stupid things.”
The concept of the electron spin was successfully used for deeper understanding of the
anomalous Zeeman effect and also of numerous atomic effects. In 1928 this hypothesis
was given conclusive evidence when Dirac, with the use of the relativistic quantum theory,
demonstrated that particles with the mass and charge of an electron should have an intrinsic
angular momentum of the value predicted by S. Gaudsmit and G.E. Uhlenbeck. Naturally,
the representation of electrons in the form of charged rotating spheres is acceptable only
within the scope of classical physics, having no sense in quantum theory. Nevertheless, in
most cases the classical interpretation is very useful. To illustrate, we consider polarization
of particles. Let us suppose that there is a certain preferential direction in space along the
axis z, and the conditions of the problem are those that the eigenvalues of the operators S2
and Sz are simultaneously and exactly measurable, that is,

[S2 , Sz ] = 0.

As for the spin operators, the commutation relations:

[Sj , Sk ] = iǫjkl Sl , j, k, l = x, y, z,

are true, the values of Sx and Sy 3 are not exactly defined. Then, similar to classical
mechanics, we may speakp of the precession of the spin vector S about the z-axis at the
angle θs (cos θs = Jz / J(J + 1), where Jz is the eigenvalue of the operator Sz ). In other
words, the statement that “the spin is along the z direction” should be understood as
follows. The spin vector lies somewhere at a particular cone surface in such a manner that
its component in the z direction equals 1/2, and two other components are unknown, while
it is known to do nothing but that:
3
Sx2 + Sy2 + Sz2 = .
4
The only preferential direction of a free particle is the direction of its motion. In the
process the intrinsic angular momentum of the particle with zero mass is oriented parallel
to the direction of its motion. If, as it has been agreed, the spin is associated with some
intrinsic motion, it should be realized within the plane perpendicular to the velocity—the
so-called “transverse polarization.” It should be emphasized that for particles with the zero
mass the helicity value, the spin projection onto the direction of motion, Σ(p) = ±1 is a
relativistically invariant quantity, in contrast to particles with the nonzero mass. Also, for
a particle with the nonzero mass the spin may be parallel to the velocity, that is, for this
particle the transverse polarization states are also possible. But these notions are Lorentz-
noninvariant since the helicity value in this case is dependent on the fact in what coordinate
system the observation is fulfilled. For example, an observer outrunning the particle will
observe even such a phenomenon as the helicity overturn (spin flip). With growing the
velocity of a particle, the process of its outrunning is not within the capacity of every

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 329

observer, and helicity acquires the feature of stability for a greater number of the observers.
In the ideal case when the particle velocity is equal to the speed of light, the helicity is a
good quantum number in any reference frame. For a nonzero-mass particle in a reference
frame, where it is at rest, the spin moment is not parallel to the velocity because the latter
is zero. In the case of such a particle, the statement concerning the spin parallelism to the
velocity is true not for every observer and hence the particle should have other polarization
states as well. Then for the spin J the number of polarization states becomes 2J + 1. On
the other hand, a zero-mass particle is always moving at the speed of light. This is the
reason why it has no system of reference, in respect to which this particle can be at rest.
Consequently, the particles of this type have only two polarization states at all values of
J. In this way the particle polarization is a characteristic of its state, associated with the
spin and its direction in space. When a particle is absolutely nonpolarized, its properties
are identical in all directions, similar to the particle with J = 0. In the general case the
polarization determines a degree of symmetry (or asymmetry) of the particle’s properties in
space. A particle is identified as polarized if its symmetry characteristic includes the screw
axis, like a rotating solid body. Polarization of particles in the general case is determined
by (2J + 1)2 − 1 parameters. For zero-mass particles the number of these parameters is
equal to two.
Polarization of particles may be realized by various methods. For instance, for neutrons
we can take an iron target magnetized up to the saturation that most actively absorbs
the neutrons, whose spins are parallel to the F e magnetization. In this case a neutron
beam emerging from the target is enriched in the state with the spin that is antiparallel
to the magnetization. Polarized electrons may be produced by the following methods: (i)
β radioactive sources (polarization due to V − A-interaction); (ii) polarization owing to
scattering of electrons (e.g., Mott scattering); (iii) optical pumping of alkali atoms by the
polarized light; (iv) ionization of polarized metastable 4 He due to the gas discharge, and
so on.

However, the most popular method of the particle polarization is transmission of the
particles through the high-intensity electromagnetic fields. Relating the spin with intrinsic
motion, we thereby proceed to the quasi-classical approximation. Let us find a relativistic
equation determining the spin dynamics in this case. We introduce the spin four-dimensional
vector sµ . In the rest frame it coincides with the spin three-dimensional vector ξ, that is,
sµ = (0, ξ). Then, since sµ and pµ are orthogonal in the rest reference frame they must
keep this property in an arbitrary reference frame

sµ pµ = 0. (15.115)

Besides, in any reference frame the relation:

sµ sµ = −ξ2 . (15.116)

takes place as well. To obtain the four-dimensional vector sµ for the electron moving
with the velocity v = p/Ep , one should use a Lorentz transformation to convert the four-
dimensional vector (m, 0, 0, 0) to pµ . The matrix of such a transformation represents the
product of three matrices of rotation in planes (x0 , xi ) through angles ϕi = arctanh(pi /Ep ).

© 2011 by Taylor and Francis Group, LLC

 330 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The explicit form of this matrix is as follows:

 
Ep p1 p2 p3
 m m m m 
 p
 1 p21 p1 p2 p1 p3 

 1+ 
µ
 m m(Ep + m) m(Ep + m) m(Ep + m) 
[L(p)]ν =  p

p2 p1 p22 p2 p3
.
 (15.117)
2

 m 1 + 
 m(Ep + m) m(Ep + m) m(Ep + m)  
 p3 p3 p1 p3 p2 p23 
1+
m m(Ep + m) m(Ep + m) m(Ep + m)
Thus, in a moving reference frame the components of the four-dimensional spin vector are
defined by the expressions:
(p · ξ)p (p · ξ)
s=ξ+ , s0 = . (15.118)
m(Ep + m) m
It is obvious that the left-hand side of the spin evolution equation is nothing else but a
derivative of sµ with respect to the proper time τ (dτ = ds). To have the right-hand side
of the desired equation, we demand it to be linear and homogeneous both in terms of the
electromagnetic field tensor Fµν and in terms of sµ . Also, the right-hand side may involve
the four-dimensional vector of velocity uµ = pµ /m, for which the following:

sµ u µ = 0

is true. Then, demanding the relativistic invariance of the required equation, we arrive at
dsµ
= αFµν sν + βuµ Fνσ uν sσ , (15.119)
dτ
where α and β are constant coefficients, The constant α could be found with the help of the
correspondence principle. A classical analog for the spin is a mechanical angular momentum
to obey the equation:
dM
= [r × F]. (15.120)
dt
In the case of a magnetic field Eq. (15.120) takes the form:
dM e
= [M × H] = 2[µ × H], (15.121)
dt m
where we have taken into account
e
µ= M.
2m
By the correspondence principle, at v → 0 the spin moment should satisfy the same equation
as does the mechanical angular moment. Setting in Eq. (15.119) sµ = (0, ξ), uµ = (1, 0, 0, 0)
and τ = t, we obtain:
dξ
= α[ξ × H]. (15.122)
dt
Further on we take into consideration that

µs = µs ξ,

where the subscript s stresses the fact that we are dealing with the magnetic moment
connected with the spin. Then, to compare Eqs. (15.121) and (15.122) leads to the result:

α = 2µs .

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 331

Differentiating the relation

sµ u µ = 0
with respect to the proper time τ and using the classical equation for a point charge moving
in a external field
duν
m = eF νλ uλ ,
dτ
we obtain:
dsµ duµ e e
uµ = −sµ = − sµ F µλ uλ = F µλ uµ sλ . (15.123)
dτ dτ m m
Multiplying the both sides of Eq. (15.119) by uµ and taking into account the relation

uµ uµ = 1,

we find the value of β  e 

β=2 − µs = −2µan s ,
2m
where µan s is an anomalous magnetic moment connected with the spin. Therefore, the
relativistic spin motion equation, Bargmann–Michel–Telegdi (BMT) equation [18], has the
following form:
dsµ
= 2µs Fµν sν − 2µan ν σ
s uµ Fνσ u s . (15.124)
dτ
From Eq. (15.124) follows that sµ sµ =const, that is, even if a particle moves in a external
field its polarization ξ is left untouched.
In practice it is more convenient to operate with the quantity ξ, directly characterizing the
particle spin direction in its instantaneous rest frame, rather than with the four-dimensional
vector of the spin. For the spatial components of Eq. (15.124) we have:
ds 2µs m 2µan ǫ
= {[s × H] + (sv)E} + s {v(v[s × H]) − v(sE) + v(sv)(vE)} . (15.125)
dt ǫ m
Substituting the relation
p(ξp)
s=ξ+ ,
m(ǫ + m)
into (15.125) we get:
dξ 2µs m + 2µan
s (ǫ − m) 2µanǫ 2µs m + 2µan
s ǫ
= [ξ×H]+ s (vH)[v×ξ]+ [ξ×[E×v]]. (15.126)
dt ǫ ǫ+m ǫ+m
where when we differentiated p and ǫ the motion equations:
dp dǫ
= eE + e[v × H], = e(vE)
dt dt
have been taken into account.
At the same time, we are interested not in a change of the absolute spin vector direction
in space but in changing a spin orientation with respect to the direction of motion. We
introduce the unit vector n = v/|v| and decompose ξ with its help into the perpendicular
and parallel components:

ξ = ξ⊥ + ξk , ξ k = n(nξ), ξ ⊥ = [n × [ξ × n]].

Then, the module of the parallel component of the spin vector is defined by the equation:
 
dξk an 2 µs m2 an
= 2µs (ξ ⊥ [H × n]) + − µs (ξ ⊥ E). (15.127)
dt v ǫ2

© 2011 by Taylor and Francis Group, LLC

 332 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The applicability condition of a quasi-classical approximation calls for a relatively slow

varying of the particle momentum, in turn leading to the field smallness condition (for
example, in a magnetic field the Larmor radius R = pc/eH should be great compared to
the particle wavelength) and to the condition of small changing the fields for the distances
on the order of the particle wavelength. Provided a field is rapidly changed, the BMT
equation should incorporate the additional components involving the field derivatives with
respect to the coordinates.

15.6 Polarization density matrix for Dirac particles

In modern practice the accelerators are operating in the incoming beam-target mode as well
as in the colliding-beam mode, and usually the beams are formed of particles with the spin
1/2. And the particle scattering experiments are statistical in character. The beam and a
stationary or movable target comprise a great number of the particles. An experimenter
measures the average scattering over a considerable number of pair collisions between the
beam particles and the target. Because of this, we can represent the beam and target as
a statistical ensemble of the systems, each of which contains a pair of interacting particles.
And the experimentally measured polarization is an average over the statistical ensemble.
Of course, each particle is always polarized in a particular manner. When considering
an aggregation of particles, however, it makes sense to look at polarization of the whole
system. It is clear that two variants are possible: (i) completely polarized system with all the
constituting particles having the same spin state, that enables one to describe this system
in terms of the field function; and (ii) system in a mixed state when all the particles are
differently polarized and associated with different field functions. In this case the system
represents a statistical mixture of various pure spin states with specific weights, that is,
there is no common field function.
Let us define the qualitative characteristic for system polarization degree. To this end,
we consider a non-relativistic electron beam. Let the vector n be in the z-direction, whereas
N+ and N− denote the number of electrons in the states with Jz = 1/2 and Jz = −1/2,
respectively. If N+ 6= N− the beam be polarized. Polarization ξz in a direction of the z-axis
is determined be the expression:
N+ − N−
ξz = . (15.128)
N+ + N−
At ξz = ±1 the beam is completely polarized while at ξz = 0 one is completely unpolarized.
In the general case the beam polarization has the longitudinal and transverse components,
that is, it is characterized by a polarization vector ξ. This vector defines the direction along
which the electron spins are mainly orientated. The module of the vector ξ gives a numerical
value of a polarization. Let us define this vector for the case of a completely polarized beam.
The coordinate dependence of the electron wavefunction describing a free motion with the
momentum p is reduced to the common factor exp (−ipx), and the amplitude ϕ(p) (two-
component spinor in the nonrelativistic case) plays a role of a spin part of the wavefunction.
In such a (pure) state an electron is completely polarized. We normalize the spinor ϕ(p)
 
c1
ϕ(p) = (15.129)
c2

as follows:
ϕ† (p)ϕ(p) =| c1 |2 + | c2 |2 = 1.

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 333

The electron states with Jz = 1/2 and Jz = −1/2 are described by the spinors:
   
1 0
ϕ1 = , ϕ2 = , (15.130)
0 1

to form the spin basis by which any spinor could be expanded. Using ϕ1,2 one may rewrite
the electron wavefunction (15.129) in the shape:

ϕ(p) = c1 ϕ1 + c2 ϕ2 . (15.131)

The quantities | c1 |2 and | c2 |2 are equal to the probability of finding an electron in the
states with Jz = 1/2 and Jz = −1/2, respectively. Therefore, the expression (15.128) could
be represented in the form:

ξz =| c1 |2 − | c2 |2 = ϕ† (p)σz ϕ(p),

which is naturally generalized on the three-dimensional case:

ξ = ϕ† (p)σϕ(p). (15.132)

In this way for a nonrelativistic electron beam the polarization vector is an average value of
the spin operator σ. The polarization vector projections ξx = c∗1 c2 +c1 c∗2 , ξy = i(c1 c∗2 −c∗1 c2 ),
and ξz are called the Stokes parameters of an electron.
Let us analyze a mixed electron beam. We assume that ξ i is the value of electron
polarization in the pure state i, and wi —relative probability with which this pure state
enters a mixed beam. Then the polarization ξ of the mixed state is determined by the
equation: X
ξ= wi ξ i , (15.133)
i
P i
where i w = 1. To substitute (15.131) and (15.132) into (15.133) leads to the expression:

X 2
X X X
†
ξ= wi ϕi† (p)σϕi (p) = wi ciα ci∗
β ϕβ σϕα = ραβ ϕ†β σϕα = Sp(ρσ). (15.134)
i α,β=1 i i

The matrix X
ραβ = wi ciα ci∗ i i∗
β ≡ cα cβ (15.135)
i

derives the name “the polarization density matrix of a mixed state.” From its definition it
follows that in a pure state (special case of a mixed state) ραβ is reduced to the product:

ραβ = cα c∗β .

Consequently, the matrix ραβ meets the normalization conditions:

Sp(ρ) = 1. (15.136)

A polarization matrix is built of the coefficients of decomposing the wavefunction of the

pure state in terms of the basis vectors and of the relative probabilities with that the pure
state enters into the mixed one. From (15.134) it is clear that with the help of ραβ we
can calculate the average value for the spin operator, that is, a polarization density matrix
completely characterizes the spin properties of a system. We can obtain a definition of the

© 2011 by Taylor and Francis Group, LLC

 334 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

ραβ -matrix that is more convenient for practical applications. It suffices to develop equation
(15.135):
X  | ci |2 ci ci∗ 
ραβ = wi 1 1 2
(15.137)
ci∗
1 c2 | ci2 |2
i

and by the direct calculations to demonstrate that the expression:

1
[δαβ + (ξ · σ)αβ ]
2
is exactly equal to the right-hand side of (15.137). Thus, the expression:

1
ραβ = [δαβ + (ξ · σ)αβ ] (15.138)
2
represents the other equivalent form of writing the polarization density matrix. Assuming
that ξ is equal to αn, in the right-hand side of (15.138) we recognize the projection operator
on the state with the spin projection onto the direction n being equal to α.
In analogy with a nonrelativistic case, the polarization density matrix of a relativistic
electron is given by the following expression:

ρ(+)
rκ = ur (p)uκ (p), (r, κ = 1, 2, 3, 4). (15.139)

Then, in a pure state it has the form:

ρ(+)
rκ = ur (p)uκ (p), (15.140)

and, consequently, it is normalized by the condition:

Sp(ρ(+) ) = 1.

Using (15.95) one could rewrite the expression (15.140) as follows:

 
ρ(+)
rκ = P + Λ (α)
(n) . (15.141)
rκ

For the mixed state in Eq. (15.141) we need an averaging procedure over the parameters
characterizing the system that incorporates this electron:


ρ(+)
rκ = P+ Λ
(α) (n)
rκ
. (15.142)

The only possible result of such an averaging may be a change in the spin projection value.
To take this into consideration, in Eq. (15.142) it is enough to set α = 1 and replace the
unit vector nµ by the vector sµ , that now takes the responsibility for the sign and degree of
the beam polarization upon oneself. Thus, the polarization density matrix for a relativistic
electron is determined by the expression:
 
1 p̂ + m
ρ(+)
rκ = (1 − γ5 ŝ) . (15.143)
2 2m rκ

The case s2µ = −1 corresponds to the pure state and the one sµ = 0—to the unpolarized
state  
(+) 1 p̂ + m
ρrκ = . (15.144)
2 2m rκ

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 335

The last equation

P may be derived by summation with respect to the polarization states of
the matrix 1/2 α vrα v α
κ . If the polarization density matrix be given, the vector sµ would
be found with help of the relation:

sµ = Sp (ργ5 γµ ). (15.145)

Similarly, with the use of Eq. (15.95), we can obtain an expression for the polarization
density matrix of positron:
 
(−) 1 p̂ − m
ρrκ = (1 + γ5 ŝ) . (15.146)
2 2m rκ

At the modern electron-positron colliders in use the energies in a center-of-mass system

are as high as hundreds of GeV, that is, leptons are ultrarelativistic particles. The energy
values of the designed muon colliders (> 400 GeV) also allow one to believe muon beams are
ultrarelativistic objects too. Let us consider how the spin projection operator is modified
in these conditions. In the ultrarelativistic limit we have:
 
m2
| p |≈ p0 1 − 2 .
2p0
Then in the electron case the calculations give:
 
p0
m−1 p̃µ γ µ u(p) = m−1 γ0 | p | − (γ · p) u(p) ≈
|p|
 
−1 m2 m2
≈m γ0 p0 − γ0 − (γ · p) + (γ · p) 2 u(p) ≈ u(p). (15.147)
2p0 2p0
From the relation (15.147) follows
1
Λ(α) (n) → (1 − αγ5 ) , (15.148)
2
that is, at high energies the helicity projection operator passes into the chirality projection
operator. Note, if for an electron the operators (1 − γ5 )/2 and (1 + γ5 )/2 are the projectors
onto the states with the positive and negative chirality, respectively, then in the case of a
positron these operators exchange their roles.
It should be emphasized that in the relativistic quantum theory the spin of a particle is
not representing an integral of motion. As a result, in an arbitrary reference frame it is
impossible to use a density matrix for characterization of a relativistic electron, the density
matrix being physically meaningful in the electron rest frame only.

15.7 Dirac equation in the external electromagnetic field

In quantum electrodynamics there are some problems, whose solutions may be obtained
within the scope of a single-particle theory. In these problems the number of particles is
fixed, and interaction may be introduced using the notion of an external (classical) electro-
magnetic field. Since in certain conditions external fields may generate the electron-positron
pairs from the vacuum, there is a need to know the limiting values of their strengths when
the single-particle approximation is applicable.

© 2011 by Taylor and Francis Group, LLC

 336 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

For a while let us come back to ordinary units. Due to the uncertainty relation:

~
∆t∆E ≥ ,
2
a virtual e− e+ -pair with the lifetime:

~
∆t ≈
2mc2
may appear from the vacuum. In this time the electron and positron may be separated by
a distance:
~
∆r ≈ c∆t ≈ ,
2mc
that is, by a distance on the order of the Compton wavelength of the electron. If an
external electric field is able to do the work of 2mc2 on the electron at the distance ∆r,
then production of a pair from the vacuum becomes real. To accomplish this the field must
have the order:
4m2 c3
eEc ∆r = 2mc2 ⇒ Ec = . (15.149)
e~
At such values of the field E = E0 ( E0 is the critical electric field) the vacuum is unstable in
respect to the production of the electron-positron pairs. The critical value of the magnetic
field H0 is established from the equality condition of the electron rotation energy quantum
~ωH (ωH is cyclotron frequency, ωH = eH/mc) and the electron rest energy:

m 2 c3
~ωH = mc2 ⇒ Hc = . (15.150)
e~
But the magnetic field does no work and hence the vacuum remains stable with respect to
the magnetic field. In other words, the magnetic field allows for a one-particle interpretation
of the problem of charged particle motion in the field even for H > Hc .
The solutions for the Dirac equation in the external electromagnetic field based on a
single-particle interpretation prove to also be relevant for the many-particle problems of
the quantum field theory when the particles in the initial and/or final state are located in
such a field. The Furry representation [19], the method using exact solutions of the Dirac
equation, was named after its author who has shown that an apparatus of the Feynman
diagrams could be generalized to the case when the electron is in the bound state rather
than in the free one. This method is fairly simple. The four-dimensional potential Aaµ is
divided into two parts:
Aaµ = Aex
µ + Aµ ,

where Aex µ describes an external field and is precisely included into the wavefunction of the
electron, and Aµ —a quantized electromagnetic field that is allowed for by the perturbation
theory. The calculation of the particular physical phenomena takes place with the following
modification of the Feynman diagram technique: in the initial and final states the electron
is located in the external field and described by the solution for the Dirac equation in this
field, the electron propagator being the Green function of this electron in the external field.
However, despite the importance of exact solutions for the Dirac equation in the external
field, we leave them in peace (detailed discussion of the problem may be found in Ref.
[20]), focusing our attention on a more modest task—the study of the Dirac equation in the
external electromagnetic field and definition of an intrinsic (spin) magnetic moment of the
electron.

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 337

Consider the motion of a Dirac particle in the external electromagnetic field that is
specified by the potential Aex
µ (x). We introduce an electromagnetic interaction by a minimal
way, that is, with the help of the recipe:

∂µ → Dµ = ∂µ + ieAex
µ (x). (15.151)

Then the Dirac equations for the bispinors ψ and ψ take the form:

(iγµ Dµ − m)ψ(x) = 0, (15.152)

(iγµT Dµ∗ + m)ψ(x) = 0. (15.153)

Such a recipe provides the invariance of equations (15.152) and (15.153) with respect to the
gauge transformations:

ψ(x) → ψ(x) exp (ieα(x)), Aex ex

µ (x) → Aµ (x) − ∂µ α(x). (15.154)

The electron rotating in a closed orbit with the momentum L has the orbital magnetic
moment:
e
µl = L. (15.155)
2m
Believing that the Creator who had constructed our world was acting by analogy, we might
expect that the value of a spin magnetic moment should be as follows:
e
µs = g s S (15.156)
2m
with gs = 1 (gs is the Lande factor or the gyromagnetic ratio). But as is evidenced by
the experiments, for the electron gs = 2 (or actually somewhat lower, aex = (gs − 2)/2 =
0.0011596521884(43) [21], that may be explained by the inclusion of radiative corrections
taking into account the interaction between a particle and vacuum fluctuations [22]). When
calculating the value of µs , it seems very useful to pick out a term in Eq. (15.152), which
describes the interaction between a spinless particle and the electromagnetic field. This
may be done by quadrating the Dirac equation, that is, by introduction of the bispinor Φ
in the following way:
1
ψ = (i∂ˆ − eÂex + m)Φ. (15.157)
m
Substituting (15.157) into (15.152), we obtain

[(i∂ˆ − eÂex )(i∂ˆ − eÂex ) − m2 ]Φ = 0. (15.158)

With allowance made for

1 µ ν 1
γ µγ ν = [γ , γ ] + {γ µ , γ ν }
2 2
and
[(i∂µ − eAex ex
µ ), (i∂ν − eAν )] = −ieFµν ,

we find that the bispinor Φ obeys the equation:

h e i
Dµ Dµ + m2 + σ µν Fµν Φ = 0. (15.159)
2
An addition (relative to the Klein-Gordon equation) term
e µν
σ Fµν = −e[(Σ · H) − i(α · E)], (15.160)
2

© 2011 by Taylor and Francis Group, LLC

 338 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where Hi = −εijk Fjk /2, Ei = F0i , defines the interaction of the spin magnetic moment
µs with the magnetic field H and the interaction of the spin electric moment ds with
the electric field E. The value of µs will be established later, and now we discuss some
mathematical details. At rising the order of the Dirac equation the number of solutions
is doubled. This means that among the solutions of Eq. (15.159) some are superfluous.
Because of this, we have to select the required solutions. In this case the situation is saved
due to the fact that Eq. (15.158) involves only the matrices σ µν commuting with the matrix
γ5 . Consequently, the solutions of Eq. (15.159) satisfy the condition:

γ5 Φ± = qΦ± , (15.161)

where q = ±1. If we restrict ourselves to the solutions (15.159) associated with only one
specific eigenvalue of q, we can establish a unique correspondence between these solutions
and those for the Dirac equation. Multiplying, for example, the equality (15.157) by

1/2(1 − γ5 )

and using Eq. (15.161) with q = 1, we obtain:

1
Φ+ = (1 − γ5 )ψ, (15.162)
2
that is, each function ψ corresponds to only one Φ+ function.
Provided for the γ-matrices the chiral representation is chosen, then from (15.162) it
follows that Φ+ is a two-component spinor. Then taking into account the relations:

σ10 = iγ1 γ0 = γ2 γ3 (iγ0 γ1 γ2 γ3 ) = iσ23 γ5 ,

σ20 = iσ31 γ5 , σ30 = iσ12 γ5

and using for the γ-matrices the representation (15.7), we obtain the following equation for
the spinor Φ+ :  
Dµ Dµ + m2 − eσ · (H + iE) Φ+ = 0. (15.163)
It should be noted that because
+ 1
Φ Φ+ = ψ(1 + γ5 )(1 − γ5 )ψ = 0,
4
then we have no possibility to construct the Lagrangian from which Eq. (15.163) might
follow.
For simplicity, we assume: E = A0 = 0 and come back to the ordinary units. Then Eq.
(15.163) takes the shape:
"  2 #
~2 2 ie e~
∂ − ~∂k + Ak + m c − (σ · H) Φ+ = 0.
2 2
(15.164)
c2 t c c

Being a classical analog of the orbital moment, the spin was first introduced into nonrel-
ativistic quantum mechanics. Because of this, the most common (and illustrative) way to
determine a value of the spin magnetic moment is finding of the Hamiltonian in a nonrela-
tivistic approximation. The reference points of the energy in the nonrelativistic theory and
in the special theory of relativity are distinguished by mc2 and hence it is more convenient
to introduce for the spinor Φ+ the following transformation:
 
+ ′ imc2 t
Φ (r, t) = Φ (r, t) exp − . (15.165)
~

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 339

Substituting (15.165) into (15.164) and neglecting the terms higher than the first order in
terms of v/c, we derive:
∂Φ′
i~ = HΦ′ , (15.166)
∂t
where
1  e 2 e~
H= p− A − (σ · H). (15.167)
2m c 2mc
Eq. (15.166) is distinguished from the nonrelativistic Schrödinger equation by the term
taking the form of the potential energy of a magnetic dipole in the external field. Thus, the
electron magnetic moment is defined by the expression:
e~ e~
µs = σ=2 S, (15.168)
2mc 2mc
from which follows that the gyromagnetic ratio gs really equals 2.

15.8 Charge conjugation, spatial inversion, and time inversion for

the Dirac field
The holes theory implies the existence of electrons and positrons satisfying the same equa-
tion and having the identical mechanical and opposite electromagnetic characteristics. Con-
sequently, the Dirac equation should possess a symmetry about the sign changing operation
for the electron charge, the charge conjugation operation. To put this another way, when
in an electromagnetic field the electron is described by the equation:
[γ µ (i∂µ − eAµ ) − m]ψ(x) = 0, (15.169)
then the relevant equation for the positron has the form:
[γ µ (i∂µ + eAµ ) − m]ψ c (x) = 0 (15.170)
and there is one-to-one correspondence between ψ(x) and ψ c (x). Direct our efforts to
determining the connection between Eqs. (15.169) and (15.170). Carrying out the complex
conjugation operation in Eq. (15.169), we obtain:
[−γ µ∗ (i∂µ + eAµ ) − m]ψ ∗ (x) = 0. (15.171)
Obviously, if one could manage to find a matrix of the shape Cγ0 with properties:
−(Cγ0 )γ µ∗ = γ µ (Cγ0 ), (15.172)
Eq. (15.171) might be represented in the form of Eq.(15.170), namely,
[γ µ (i∂µ + eAµ ) − m]Cγ0 ψ ∗ (x) = 0.
Then the connection sought takes the form:
T
ψ c (x) = Cγ0 ψ ∗ (x) = Cψ . (15.173)
The matrix C is called the charge conjugation matrix and ψ c (x)—the charge conjugate
bispinor. With the help of the relation
γ0 γ µ γ0 = γ µ† ,

© 2011 by Taylor and Francis Group, LLC

 340 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the condition determining the matrix C could be rewritten in the form that is more conve-
nient for practical applications:
C −1 γ µ C = −γ µT . (15.174)
Since the matrices γµT satisfy the same anticommutation relations as the matrices γµ , then,
according to the fundamental theorem about the Dirac matrices (see, Section 15.2) the
matrix C really exists. Using (15.174) and taking into consideration the γ-matrix algebra,
one could show that the matrix C possesses the following properties:
C † = C −1 = C = −C T = C ∗ . (15.175)
Let us establish how the spinor ψ(x) is transformed under the charge conjugation. Since:
T
ψ(x) → ψ c (x) = Cψ (x) = C[ψ † (x)γ0 ]T = Cγ0T ψ †T (x), (15.176)
then having done the Hermitian conjugation in Eq. (15.176), we obtain:

ψ † (x) → (ψ c (x))† = ψ T (x)γ0†T C † . (15.177)

To multiply the both sides of Eq. (15.177) by γ0 and account for (15.174), (15.175) gives:

ψ(x) → (ψ c (x))† γ0 = ψ T (x)γ0†T C † γ0 = ψ T (x)γ0 (−γ0 C † ) = −ψ T (x)C,

that allows one to conclude:
c
ψ (x) = −ψ T (x)C. (15.178)
So, due to the invariance with respect to the charge conjugation operation, a change in
sign for the four-dimensional potential of the electromagnetic field:
Aµ (x) → Acµ (x) = −Aµ (x), (15.179)
with simultaneous antilinear transformations (15.173) and (15.178) for the bispinors ψ(x)
c
and ψ(x), leads to the equations for the bispinors ψ c (x) and ψ (x) in the field Acµ (x), which
in their form are identical to the Dirac equations for ψ(x) and ψ(x) in the field Aµ (x). This
means that the total Lagrangian of this system electromagnetic field + Dirac field retains
its form under the transformations (15.173), (15.178), and (15.179). Considering that the
electromagnetic field potential is included into the Dirac equations with the coefficient e,,
the transformation (15.179) should be interpreted as a change in the charge sign. Invariance
of the Dirac equation with respect to the transformation of the charge conjugation points
to the equal rights of two signs of the charge: for each solution ψ(x) associated with a
T
certain sign of the charge there is the solution Cψ (x) having the same properties but
corresponding to the opposite-sign charge.
In the Dirac-Pauli representation the matrix C has the form:
 
0 −iσ2
C = −iγ2 γ0 = . (15.180)
−iσ2 0
When passing on to other representations of γ-matrices, a new matrix of the charge conju-
gation, C ′ , may be obtained by means of the formula:
C ′ = U CU T . (15.181)
In the assumption of the relativistic covariance of the charge conjugation operation the
bispinors ψ c (x) and ψ(x) should have the same transformational properties under the
Poincare transformations, that is,
ψ c′ (x′ ) = Sψ c (x), (15.182)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 341

ψ ′ (x′ ) = Sψ(x). (15.183)

Substituting Eq. (15.173) into (15.182) and allowing for
′
ψ (x′ ) = aψ(x)S −1 ,

we arrive at the result:


C −1 S −1 C, at Λ00 > 0,
ST = (15.184)
−C −1 S −1 C, at Λ00 < 0.

Next, we show that the conditions (15.184) make it possible to decrease the number of
the admissible values for the matrices performing the space-inversion and time-inversion
transformations.
Before we have found out that under the space inversion the bispinor ψ(x) is transformed
by the following way:
ψ ′ (−r, t) = P ψ(r, t) = ηp γ0 ψ(r, t), (15.185)
where ηp is an intrinsic parity of a spinor particle equal either to ±1 or to ±i. It is an easy
matter to convince that the fulfillment of (15.184) is guaranteed by the choice ηp = ±i.
To establish the transformation law of the Dirac’s conjugate bispinor we carry out the
Hermitian conjugation on (15.185) and multiply the result obtained by γ0 on the right
′
ψ (−r, t) = ηp∗ ψ(r, t)γ0 . (15.186)

Therefore, the intrinsic parities of the Dirac particles and antiparticles are opposite.
Consider the time-inversion transformation. From Section 5.2 we know that this transfor-
mation represents the product of the complex conjugation operation of a wavefunction and
some unitary transformation which we denote by T . Thus, for a time-inverted wavefunction
ψ ′ (r, −t) we have:
ψ ′ (r, −t) = T ψ(r, t). (15.187)
To find an explicit form of a matrix T we write the Dirac equations to be satisfied by the
functions ψ ′ (r, t) and ψ(r, −t): ψ ′ (r, t) and ψ(r, −t)

(iγ0 ∂t − iγ∇ − m)ψ ′ (r, t) = 0, (15.188)

(−iγ0T ∂t − iγ T ∇ + m)ψ(r, −t) = 0. (15.189)

To multiply the last equation by −T on the left gives

(iT γ0T ∂t + iT γ T ∇ − mT )ψ(r, −t) = 0. (15.190)

Now we demand the function T ψ(r, −t) to obey the same equation as ψ ′ (r, t):

(iγ0 ∂t − iγ∇ − m)T ψ(r, −t) = 0. (15.191)

Comparing Eqs. (15.188) and (15.191), we have drawn the conclusion that the matrix T
must satisfy the conditions:

T γ0T = γ0 T, T γ T = −γT. (15.192)

It is easy to verify that in the Dirac-Pauli and Weyl representations the matrix

T = ηt γ 3 γ 1 γ 0 , (15.193)

© 2011 by Taylor and Francis Group, LLC

 342 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

fulfills these conditions.∗ So we finally have

ψ ′ (r, −t) = T ψ(r, t) = ηt γ 3 γ 1 γ 0 ψ(r, t). (15.194)

Analogous manipulation shows that under time inversion the conjugate bispinor is trans-
formed by the law:
′
ψ (r, −t) = ηt∗ ψ(r, t)γ 0 γ 1 γ 3 . (15.195)
Using the explicit form of the spin projection operator it is easy to display that under the
time inversion the helicity changes its sign.
Let us find the changeover of the field function of a spinor field under the P T and CP T
operations in the Dirac-Pauli and Weyl representations. Calculations lead to the result:

P T ψ(r, t) = ηp ηt γ 0 γ 3 γ 1 ψ ∗ (−r, −t), (15.196)

CP T ψ(r, t) = iγ 2 γ 0 [ηp ηt γ 3 γ 1 ψ ∗ (−r, −t)]∗ = −η0 γ5 ψ(−r, −t)], (15.197)

where η0 = ηt∗ ηp∗ .

15.9 Relativistic quantization scheme

The canonically conjugate momenta in the case of the Dirac field are determined by the
following expressions:
i i
π(x) = ψ † (x), π † (x) = − ψ(x),
2 2
from where it follows that they are no longer independent of the “coordinates” of the
field ψ(x) and ψ † (x). This makes us doubtful whether the use of the canonical quantization
procedure is appropriate. From the simultaneous commutation relations for ψ(x) and π(x′ ):

{ψ(x), π(x′ )}t=t′ = iδ (3) (r − r′ ),

it follows that
1
{ψ(x), ψ † (x′ )}t=t′ = δ (3) (r − r′ ). (15.198)
2
On the other hand, from similar relations for the canonically conjugate pair ψ † (x), π † (x′ )
we have
1
− {ψ † (x), ψ(x′ )}t=t′ = δ (3) (r − r′ ),
2
that is contrary to Eq. (15.198).
Thus, the formalism of the canonical quantization is directly inapplicable, and we have
a nice opportunity to broaden our range of interests owing to the introduction of the so-
called relativistic quantization procedure. It is based on a theory of representations requiring
correspondence between the transformation laws for the quantized and classical fields, both
under transformations of the coordinate systems and the fields themselves. Its essence is as
follows.
The transformations of the coordinates and field functions:

x′ = Λx + a, ψ ′ (x′ ) = S(Λ)ψ(x).

∗ (According to (15.184) ηt must equal ±i)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 343

induce a change in the state vector of the system Φ that is determined by the relation:

Φ′ = Uv Φ. (15.199)

Invariance of the norm of the state vector requires unitarity of the operator Uv while the
superposition principle imposes the linearity condition on it. We consider this situation
from an experimenter’s viewpoint. Let us think of some dynamic observable (for example,
momentum Pµ (x)) as being measured in the states described by the vectors Φ and Φ′ . Then
a relativistic covariance requires the following relations to be fulfilled:

(Φ′ , Pν (x′ )Φ′ ) = Λµν (Φ, Pµ (x)Φ). (15.200)

This in turn leads to the requirement that is relating expressions for the momentum operator
in both systems:
Uv−1 Pµ (x′ )Uv = Λνµ Pν (x). (15.201)
Similarly, when computing the average over the operator field function, the resultant ex-
pression retains a geometric nature of the nonquantized field function, that is, the relation:

(Φ′ , ψ(x′ )Φ′ ) = S(Λ)(Φ, ψ(x)Φ) = (Φ, ψ ′ (x′ )Φ). (15.202)

takes place. Thus, for the field with arbitrary spin value we obtain from Eq. (15.202)

Uv−1 ψ(x′ )Uv = S(Λ)ψ(x) (15.203)

or
ψ ′ (x′ ) = Uv−1 ψ(x′ )Uv . (15.204)
In the case of infinitesimal transformations of the Poincare group the transformation matrix
of a state vector could be represented in the shape:
X
Uv = 1 + δUv , δUv = i Un δω n , (15.205)
n

where δω n are transformation parameters. Then for infinitesimal transformations from Eq.
(15.204) follows:

ψ ′ (x′ ) = (1 − δUv )ψ(x′ )(1 + δUv ) ≈ ψ(x′ ) + [ψ(x′ ), δUv ], (15.206)

which leads to the following expression:

δψ(x) = ψ ′ (x) − ψ(x) = [ψ(x), δUv ]. (15.207)

Using the explicit form of the form variation for the field function δψ(x) (1.4.17), we get the
conditions on the operator field functions that at the heart of the relativistic quantization
procedure:  
k
i Yi(n) − ψi;k X(n) = [Un , ψi ]. (15.208)

In order to make the direct application the formula (15.208) should be rewritten for
concrete transformations. In the case of space-time translation:
k
X(n) = δnk , Yi(n) = 0, Un = Pn ,

the relation (15.208) takes the form:

∂ψi (x)
i = [ψi (x), Pn ]. (15.209)
∂xn

© 2011 by Taylor and Francis Group, LLC

 344 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Recall that in the nonrelativistic quantum mechanics Heisenberg’s operators N (x) obeyed
the motion equations:
∂N (x)
i = [N (x), H], (15.210)
∂t
where H is a system Hamiltonian. Since in the quantum field theory we deal with operators
dependent on space-time coordinates in the same sense as the Heisenberg’s operators of non-
relativistic quantum mechanics are time-dependent, then a generalization of Eq. (15.210) for
the space coordinates must exist as well. Such a relativistic generalization of Eq. (15.210)
is represented by Eq. (15.209) derived by us for the partial case, when the field function
itself plays a role of the operator N (x).
Under four-dimensional rotations the condition (15.208) leads to the relation:
n o
j
i ψi;σ (x)(xν δµσ − xµ δνσ ) − (Kµν )i ψj (x) = [ψi (x), Mµν ], (15.211)

where Mµν is a tensor operator of a total angular momentum and we have taken into
account Eqs. (1.4.15) and (1.4.34). From obtained formulae it is also obvious that under
the Poincare transformations the state vector is transformed by the law:
 
′ µ i µν
Φ = exp 1 + iPµ a + Mµν ω . (15.212)
2
As an example of the transformations not touching upon space-time coordinates we con-
sider the gauge transformations of the first kind.∗ In this case the state vector transforma-
tion law has the form:
Φ′ = exp (iαQ)Φ, (15.213)
where Q should be interpreted as an charge operator (electric, baryon, etc.) of a secondary
quantized field. Now the formula (15.208) leads to the relations:
ψ(x) = [ψ(x), Q], ψ ∗ (x) = −[ψ ∗ (x), Q]. (15.214)
Eqs. (15.209), (15.211), and (15.214) form the base of the relativistic quantization pro-
cedure. With their use one can establish both a physical meaning of the operator field
functions and the permutation relations associated with them.

15.10 Dirac field quantization (momentum space)

Let us apply the above-mentioned procedure to the Dirac particles. The form of the per-
mutation relations may be most conveniently determined by Eq. (15.209). In this manner
we should begin with the construction of an energy-momentum tensor. The standard ma-
nipulation leads to the expression:
i 
T µν = ψ(x)γ ν ∂ µ ψ(x) − ∂ µ ψ(x)γ ν ψ(x) . (15.215)
2
Though asymmetric, this tensor results in the same value for four-dimensional vector of the
field momentum as does the symmetrized tensor:
Z Z
i  
P µ = T µ0 d3 r = d3 r ψ(x)γ 0 ∂ µ ψ(x) − ∂ µ ψ(x)γ 0 ψ(x) =
2

∗ Sometimes they are simply called gradient transformations.

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 345
Z Z
i 
= d3 r 2ψ(x)γ 0 ∂ µ ψ(x) − ∂ µ [ψ(x)γ 0 ψ(x)] = i d3 r[ψ(x)γ 0 ∂ µ ψ(x)]. (15.216)
2
During deriving (15.216) we have used the Gauss theorem and allowed for the electric
current conservation law
∂ µ jµ(em) = 0, (15.217)
where
jµ(em) = eψ(x)γµ ψ(x). (15.218)
Further on we expand the operators ψ(x) and ψ(x) in terms of the c-number solutions of
the Dirac equation:
ψ(x) = ψ (+) (x) + ψ (−) (x) =
Z r 2
1 3 m X
= d p [ar (p)u(r) (p)e−ipx + b†r (p)v (r) (p)eipx ], (15.219)
(2π)3/2 Ep r=1

ψ(x) = ψ (+) (x) + ψ (−) (x) =

Z r 2
1 m X †
= d3 p [a (p)u(r) (p)eipx + br (p)v (r) (p)e−ipx ], (15.220)
(2π)3/2 Ep r=1 r

where ar (p) and br (p) are operators. To substitute (15.209) and (15.210) into the ex-
pression for the four-dimensional momentum vector (15.217) and take into account the
orthonormalization condition for bispinors (15.99) gives the following result:
Z 2
X
Pµ = d3 p pµ [a†r (p)ar (p) − br (p)b†r (p)]. (15.221)
r=1

The form of expression (15.221) is almost identical to that of the four-dimensional momen-
tum vector of a scalar field. However, a change of the sign before the second term is very
dramatic within the scope of the primary-quantized theory. The energy is not positively
defined (indefinite), that is, the energy spectrum has no lower limit. Recall that in the case
of a scalar field the charge is indefinite instead. In this case the situation may be improved
by the use of the commutation relations capable of providing a change of the sign in front
of the second term in (15.221) when taking it to the traditional form of the antiparticle
number operator. Obviously, quantization with the commutators presents no definiteness
of the energy, whereas the problem is completely solved due to the use of anticommutators.
Proceeding further, we should remember this fact. With the use of the operator expan-
sion of the field functions (15.219), (15.220) and with the help of (15.209) we can establish
that the translational invariance is liable to take place under fulfillment of the following
conditions: 
[Pµ , as (p)] = −pµ as (p), [Pµ , bs (p)] = −pµ bs (p), 
    (15.222)
Pµ , a†s (p) = pµ a†s (p), Pµ , b†s (p) = pµ b†s (p).


Insertion of the expression for Pµ into the first of the relations (15.222) yields the result:
" #
X

† ′ ′ ′ †
ar (p )ar (p ) − br (p )br (p ) , as (p) = −δ (3) (p − p′ )as (p).
′
(15.223)
r

As the operators as (p) and bs (p) are associated with different degrees of freedom of the
field, we should have:
[br (p′ )b†r (p′ ), as (p)] = 0. (15.224)

© 2011 by Taylor and Francis Group, LLC

 346 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Then the relation (15.223) takes the form:

X

a†r (p′ ){ar (p′ ), as (p)} − {a†r (p′ ), as (p)}ar (p′ ) = −δ (3) (p − p′ )as (p). (15.225)
r

The obvious solution of (15.225) is as follows:

{ar (p′ ), as (p)} = 0, (15.226)

{ar (p′ ), a†s (p)} = δ (3) (p − p′ )δrs . (15.227)

Analogously, from the remaining relations (15.223) we can obtain:

{br (p′ ), bs (p)} = 0, (15.228)

{br (p′ ), b†s (p)} = δ (3) (p − p′ )δrs . (15.229)

If the state ΦE is the energy operator eigenstate with the eigenvalue E, then, with allowance
made for the relations (15.222), we have:

Har (p)ΦE = ar (p)HΦE + [H, ar (p)]ΦE = (E − k0 )ar (p)ΦE , (15.230)

Ha†r (p)ΦE = (E + k0 )a†r (p)ΦE . (15.231)

Similar relations takes place for the states br (p)ΦE and b†r (p)ΦE .
Thus, the operators ar (p)
and br (p) have the meaning of annihilation operators because they decrease the system’s
energy, while a†r (p) and b†r (p)—the meaning of production operators as they increase the
energy of this system.
To establish a physical meaning of the operators a†r (p) and b†r (p) definitively, we should
have to use the expressions (15.211) and (15.214). But since the anticommutation relations
between the operators have been already found, the situation may be simplified. Using
the expression for the spin moment tensor, we arrive at the spin projection operator on a
motion direction:
Z
1
S3 = d3 p[a†1 (p)a1 (p) − a†2 (p)a2 (p) + b1 (p)b†1 (p) − b2 (p)b†2 (p)]. (15.232)
2
where the vector p is directed along the axis z. Then, with consideration for the relations
(15.227) and (15.229) we shall have
1 †
 2 a1 (p), r = 1,
[S3 , a†r (p)] = (15.233)
 1 †
− 2 a2 (p), r = 2,
 1 †
 − 2 b1 (p), r = 1,
[S3 , b†r (p)] = (15.234)
1 †
2 b2 (p), r = 2.
Using the Lagrangian (15.13), we calculate the expression for the electric charge of a field:
Z 2
X
Q= d3 p e[a†r (p)ar (p) + br (p)b†r (p)]. (15.235)
r=1

Note that the charge is positively defined even before the application of the quantization
rules. This is inferred as a partial corollary to the Pauli theorem about the connection

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 347

of the spin with statistics. By the classical field theory, the energy for the particles with
a half-integer spin is indefinite and for those with an integer spin the charge is indefinite
instead.
To obviate these inconsistencies, fermions are quantized with the use of the anticom-
mutators, and bosons—with commutators. Using the anticommutators, we find without
trouble:
[Q, a†r (p)] = ea†r (p), [Q, b†r (p)] = −eb†r (p). (15.236)
Our further reasoning is quite clear. We take the state with a definite value of the charge
and spin projection ΦQ,S3 . Acting by the operators Q and S3 on the states a†r (p)ΦQ,S3 and
b†r (p)ΦQ,S3 , and using the relations (15.234), (15.235) and (15.237), we obtain the following
ultimate answer. The quantities a†r (p) and ar (p) represent the production and annihilation
operators for particles with the momentum p, mass m (p2 = m2 ), charge −|e|, and spin
projection onto the z-axis that is equal to 1/2 (r = 1) or -1/2 (r = 2). The operators b†r (p)
and br (p) are associated with antiparticles, being distinguished from the foregoing by the
signs of a charge (+|e|) and spin projection -1/2 for r = 1 and 1/2 for r = 2. Such sign
reversals are in line with the predictions of Dirac’s theory of holes. From this viewpoint, a
positron is understood as the absence of an electron with the negative energy. If the missing
electron has a positive spin projection 1/2 and is charged negatively, its absence results in
the negative spin projection −1/2 and the positive charge.
(+)
We introduce the operators of the particles number Nr (p) and the antiparticles number
(−)
Nr (p):
Nr(+) (p) = a†r (p)ar (p), Nr(−) (p) = b†r (p)br (p). (15.237)
With their help an energy and a charge of a field are represented in the form:
Z X2
H = d3 p Ep [Nr(+) (p) + Nr(−) (p)] + E0 , (15.238)
r=1
Z 2
X
Q= d3 p e[Nr(+) (p) − Nr(−) (p)] + Q0 . (15.239)
r=1
In the formulae (15.238), (15.239) E0 and Q0 denote the energy and total charge of the
vacuum: Z Z
E0 = −2 d3 pEp , Q0 = −2e d3 p.

By the holes theory, we can assume that the observables represent the differences H − E0
and Q − Q0 only. But in the spirit of quantum field theory another approach (normal
products method) seems to be more appropriate, being also helpful in avoidance of such
infinities. As we pass to the normal products, beginning just from the Lagrangian, we get
the expressions for dynamic variables involving no infinite vacuum contributions. From the
anticommutation relations it follows:
Nr(+)2 (p) = a†r (p)ar (p)a†r (p)ar (p) = a†r (p)[1 − a†r (p)ar (p)]ar (p) = a†r (p)ar (p) = Nr(+) (p),
where we have allowed for that according to (15.226) ar (p)ar (p) = 0. In this way for
fermions the particles number in one state may be only zero or unity in agreement with the
Pauli principle. We characterize the vacuum state by the conditions:
Nr(+) (p)Φ0 = Nr(−) (p)Φ0 = 0,
to be true for all r and p. It is obvious that the state a†r (p)Φ0 is the single particle state with
the momentum p, the mass m and the spin projection ±1/2 for r = 1, 2; a†r2 (p2 )a†r1 (p1 )Φ0
is two particle state ..., b†r (p)Φ0 is the single antiparticle state and so on.

© 2011 by Taylor and Francis Group, LLC

 348 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Taking into account of Eqs. (15.254) and (15.265), in the expansion of the field function
(15.219) we can go from the sum over r to the sum over the eigenvalues of the spin projection
operator J3 = ±1/2 (α = ±):
Z r
1 3 m X
ψ(x) = d p [aα (p)u(α) (p)e−ipx + b†α (p)v (α) (p)eipx ], (15.240)
(2π)3/2 Ep α=±

where the bispinors u(α) (p) and v (α) (p) are now defined the following way:
 
ς±
! (p · σ) ∓
(p · σ) ± , ς
u(±) (p) = N ς v (±) (p) = N  Ep + m  . (15.241)
Ep + m ∓
ς

Using the orthonormality properties of the solutions for the Dirac equation, we can realize
the inversion of Eq. (15.240). Calculations give
Z
1
aα (p) = d3 xeipx u(α) (p)γ0 ψ(x), (15.242)
(2π)3/2
Z
1
bα (p) = d3 xψ(x)γ0 v (α) (p) (15.243)
(2π)3/2
and so on. Based on the derived relations, we can write the state vector of this system in
the Fock representation
Z Z
Φn+ n−′ ... = ... F (x1 , ..., xn )ψ b1 (x1 )...ψ bn (xn )d4 x1 ...d4 xn Φ0 . (15.244)
k1 k1

where bi = ±, ψ + (x) ≡ ψ(x), ψ − (x) ≡ ψ(x) and F (x1 , ..., xn ) has a meaning of the ordinary
quantum-mechanical wavefunction of the system including n particles in the configuration
space.

15.11 Dirac field quantization (configuration space)

In the configuration space a role of the production and annihilation operators of the electrons
(positrons) is played by the quantities ψ (+) (x) (ψ (−) (x)) and ψ (+) (x) (ψ (−) (x)), respectively.
Through the use of these operators the vacuum is set by the condition:

ψ (+) (x)Φ0 = ψ (−) (x)Φ0 = 0.

Nonzero anticommutation relations for the number operators have the form:

{ψα(+) (x), ψ (+) β (x′ )} =

s
Z Z 2
1 m2 X (s)
= d3 p d3 p′ {ar (p), a†s (p′ )}vα(r) (p)v β (p′ )×
(2π)3 Ep Ep′ r,s=1
Z
1
′ ′ d3 p
× exp [−ipx + ip x ] = (p̂ + m)αβ exp [−ip(x − x′ )] =
2(2π)3 Ep

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 349
Z
1 d3 p
= (iγ µ ∂µ + m)αβ exp [−ip(x − x′ )] =
2(2π)3 Ep
= −i(iγ µ ∂µ + m)αβ ∆+ ′ 2 + ′
0 (x − x ; m ) = −iSαβ (x − x ). (15.245)
Analogous manipulations for the production and annihilation operators of the positrons
lead to the result:
−
{ψα(−) (x), ψ (−) β (x′ )} = −iSαβ (x − x′ ), (15.246)
where
−
Sαβ (x) = (iγ µ ∂µ + m)αβ ∆− 2
0 (x; m ). (15.247)
With allowance made for the definition
+ −
Sαβ (x) = Sαβ (x) + Sαβ (x) = (iγ µ ∂µ + m)αβ ∆0 (x; m2 ), (15.248)

we arrive at the Dirac fields anticommutator for arbitrary space-time points:

{ψα (x), ψ β (x′ )} = −iSαβ (x − x′ ). (15.249)

The function Sαβ (x) possesses the following properties:

(iγ µ ∂µ −m)S(x) = (iγ µ ∂µ −m)(iγ µ ∂µ +m)∆0 (x; m2 ) = −(+m2 )∆0 (x; m2 ) = 0 (15.250)

and
S(x)|t=0 = iγ0 δ (3) (x). (15.251)
Differential equation (15.250), together with the initial condition (15.251), defines the func-
tion S(x) unambiguously. This singular function plays a part of the Green function in the
process of solving the problems with the initial conditions as we have at hand:
Z
ψ(x) = −i dσµ (x′ )S(x − x′ )γ µ ψ(x′ ), (15.252)

where t > t′ . To prove the relation (15.252) is trivial. The derivative δ/δσ(x′ ) from the
right-hand side of (15.252) goes to zero because both S(x − x′ ) and ψ(x′ ) satisfy the Dirac
equation. This means that the integral in the right-hand side of (15.252) is independent on a
choice of the hypersurface σ ′ . As a result, we can take as that the hyperplane t′ = t =const,
that, with considering the initial condition for the function S(x), turns (15.252) into the
identity.
Comparison between Eqs. (15.248) and (15.1) gives additional evidence that the singular
functions related to the wave fields, which are described by the first-order equations, may be
expressed with the use of the operator drs (∂) in terms of the singular functions characterizing
the Klein-Gordon field. In particular, the causal function of a Dirac particle is given by the
expression:
S c (x) = (iγ µ ∂µ + m)αβ ∆c (x; m2 ). (15.253)
The Fourier series expansion of the function S c (x) has the form:
Z
c 1 (p̂ + m)αβ
Sαβ (x) = d4 p 2 exp (−ipx). (15.254)
(2π)4 m − p2 − iǫ

Knowing S c (x), we can calculate the convolution of the field operators:

s
ψαs (x)ψ β (x′ ) =< 0|T (ψα (x)ψ β (x′ )|0 >= −iSαβ
c
(x − x′ ). (15.255)

© 2011 by Taylor and Francis Group, LLC

 350 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It should be noted that in the case of the Dirac field, under the T and N -ordering due to
quantization with the use of anticommutators, permutation of the two operators leads to a
change of the sign. For example,

′

ψα (x)ψ β (x′ ), t > t′ ,
T ψα (x)ψ β (x ) = ′ (15.256)
−ψβ (x )ψα (x), t < t′ .

and


N ψα (x)ψ β (x′ ) = −N ψ β (x′ )ψα (x) .
It is interesting that the anticommutation relations (15.249) may be also obtained with
the use of a somewhat modified canonical formalism. Recall that for the Dirac field the
simultaneous anticommutation relations of the pairs ψ(x), π(x′ ) and ψ † (x), π † (x′ ) were at
variance with each other. For this problem a Solomonian decision may be based on the
principle borrowed from politics: “no person, no problem.” One of the canonical momenta
may be removed by changing of the Lagrangian. The only “painless” modification of the
Lagrangian possible is an addition or subtraction of the terms having the form of the four-
dimensional divergency. By adding of the quantity:
i
L= ∂µ (ψγ µ ψ),
2
to the Lagrangian (15.13), π † (x) becomes zero and π(x) receives the value iψ † (x), that in
turn leads to the following anticommutation relation for the coincident times:

{ψ(x), π(x′ )}t=t′ = i{ψ(x), ψ † (x′ )}t=t′ = iδ (3) (r − r′ ). (15.257)

Next, into the anticommutator:

{ψ(x), ψ † (x′ )}
we substitute the function ψ † (x′ ) represented with the help of the Green function (15.252).
Then, selecting the hyperplane t = t′ =const as the hypersurface σ and using (15.257), we get
the same anticommutation relations in arbitrary time as within the scope of the relativistic
quantization procedure. It should be emphasized that the difficulties encountered with the
canonical formalism are characteristic for the first-order equations of a particle theory for
any spin. Nevertheless, the above procedure for establishing the anticommutation relations,
that is (mildly speaking) somewhat inconsistent as the reason for the exclusion of π(x) and
not π † (x) is not clear, invariably gives the correct result.
To conclude this consideration, we present a number of useful relations for the charge
conjugate operator field functions. In Section 15.14 it will be shown that the effect exerted
by the second-quantized charge-conjugation operator C on the operator field function is
defined by the following relation:
T c
Cψ(x)C −1 = ηc ψ c (x) = ψ (x), Cψ(x)C −1 = ηc∗ ψ (x) = C −1 ψ(x), (15.258)

where |ηc |2 = 1. This in turn means that the linkage between the charge conjugate operator
field functions is of the same form as in a primary quantization theory. In this case, using
the expansion of ψ(x) (15.219) and allowing for the properties of the charge conjugation
matrix, we get
c
{ψαc (x), ψ β (x)} = −iSαβ (x − x′ ), (15.259)
{ψα (x), ψβc (x)} = −iCτ β Sατ (x − x′ ), (15.260)
c
{ψ α (x), ψ β (x)} = −iCβτ Sτ α (x − x′ ). (15.261)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 351

15.12 Majorana equation

The charged leptons and quarks are described by the Dirac equation. The same equation
may be used to describe electrically neutral particles with the spin 1/2, on condition that
the particles and antiparticles are not identical. And for the truly neutral particles with the
spin 1/2 the Dirac equation should be modified. This necessity is clear because the quantity
iγ µ ∂µ in the Dirac-Pauli and Weyl representations is not actually real. Because of this, to
describe a truly neutral particle, we have to choose the representation with the property:

Re(γµ ) = 0.

This requirement is met by the Majorana representation:

    
0 0 σ2 1 σ3 0 
γ = , γ = −i ,

σ2 0 0 σ3 

     (15.262)
0 σ2 σ1 0 

γ2 = , γ3 = i ,

−σ2 0 0 σ1

that is related to the Dirac-Pauli representation by the unitary transformation:

1 1
γµ () = √ (1 − γ2 )γµ √ (1 + γ2 ), (15.263)
2 2
where the γ-matrices in the right-hand side of (15.263) are given in the Dirac-Pauli rep-
resentation. A truly neutral particle with an arbitrary spin is conventionally called the
Majorana particle. The presently known particles of this class have an integer spin. We
cannot rule out the existence of the truly neutral fermions in nature. A theory of truly
neutral particles with the spin 1/2 has been first introduced by E. Majorana [23]. Their
role may be played primarily by the neutrinos. The experiments on Super-Kamiokande [24]
and Sudbury [25] have demonstrated that a neutrino is not a massless particle as it was
supposed by the standard model. There is reason to believe that the neutrino mass is in the
region of several tenths of eV. Naturally, during typical collider experiments the neutrino
with such a small mass practically behaves as a massless particle. As will be shown later,
for mν = 0 the behavior of the Majorana and Dirac neutrinos is identical. The difference
between these two types of neutrino is exhibited only in the processes, where the mass has
a significant role. Recording of double neutrinoless beta-decay (0ν2β)

(N, Z) → (N − 2, Z + 2) + e− + e− . (15.264)

may be one of the possible experiments to support the Majorana nature of the neutrino. In
the beginning of 2002 detecting this process was reported by the Heidelberg-Moscow (HM)
Collaboration [26]. However, it is early to classify these results among the reliable data as
some other collaborations have serious doubts (see, for example, Ref. [27]). We shall come
back to this problem in Part 7.
If we were concerned with the particles described by the tensor quantities, the fact of
true neutrality of the particles could be expressed by the apparent relation:

ψ(x) = ψ ∗ (x). (15.265)

But in the case of the spinor particles the behavior of ψ(x) and ψ ∗ (x) under the Poincare
transformations is different. In the case when the orbital moment of momentum equals zero

© 2011 by Taylor and Francis Group, LLC

 352 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the corresponding transformation laws have the form:

 
′ i µν
ψ (Λx + a) = exp σµν ω ψ(x) (15.266)
4
and  
i ∗ µν
ψ ′∗ (Λx + a) = exp − σµν ω ψ ∗ (x). (15.267)
4
Certainly, if we impose the condition (15.265) in a particular Lorentz system, this con-
dition is invalid in some other Lorentz system. As has been previously established, the
field function transformation law dictates the equation form. In this manner, for the rela-
tivistically invariant definition of a truly neutral spinor particle both sides of Eq. (15.265)
should involve the quantities obeying the same equation. This condition is met by ψ(x) and
ψ c (x). Really, at e = 0 Eqs. (15.169) and (15.170) takes the same form. Consequently, the
relativistically invariant definition of a truly neutral particle is as follows:

ψ(x) = λ⊙ ψ c (x), (15.268)

where λ⊙ is a phase factor (|λ⊙ |2 = 1). The relation (15.268) is customarily called the
Majorana condition.
First, we consider the case of massless neutrinos and demonstrate the possibility to de-
scribe the Majorana neutrino by the Weyl spinors using the chiral representation. We shall
work in the chiral representation. Then with regard to:
T
ψ c = ψ = iγ 2 ψ ∗ , (15.269)

we obtain for the two-components Weyl’s spinors ψR,L :

  ∗    
c 1 − γ5 0 iσ2 η ∗ 1 + γ5 iσ2 η ∗
ψL ≡ (ψL )c = iγ 2 = = , (15.270)
2 η 0 2 0
  ∗    
c c 2 1 + γ5 ξ 0 1 − γ5 0
ψR ≡ (ψR ) = iγ = = . (15.271)
2 0 −iσ2 ξ ∗ 2 −iσ2 ξ ∗
The massless neutrino field Lagrangian is determined by the expression:

L = iψγµ ∂ µ ψ = i[ψ L γµ ∂ µ ψL + ψ R γµ ∂ µ ψR ]. (15.272)

With the help of the Weyl’s spinors we can define the Majorana spinors χ and ζ, for
which the condition of (15.268) proves to be fulfilled:
c
χ = ψL + ψL , (15.273)
c
ζ = ψR − ψR . (15.274)
To find the connection between the Weyl’s and Majorana spinors presents no special prob-
lems. Using Eqs. (15.273) and (15.274) we arrive at:
1 − γ5 c 1 + γ5
ψL = χ, ψL = χ, (15.275)
2 2
1 + γ5 c 1 − γ5
ψR = ζ, ψR =− ζ. (15.276)
2 2
The substitution of Eqs. (15.275) and (15.276) into (15.272) leads to the Lagrangian:
i
L= [χγµ ∂ µ χ + ζγµ ∂ µ ζ]. (15.277)
2

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 353

To describe the Majorana neutrinos, we use the spinor χ. In the case of massive neutrinos
the Lagrangian will look like:
1
L= [iχγµ ∂ µ χ − mχχ], (15.278)
2
where m is a real quantity. Now instead of (15.273), we assume:
c
χ = ψL exp (iβ1 ) + ψL exp (iβ2 ). (15.279)

This, in its turn, fixes the phase in the Majorana condition:

λ∗ = exp (iβ1 + iβ2 ).

Rewriting the Lagrangian (15.278) in terms of ψL , we get:

1 c
L = iψ L γµ ∂ µ ψL − M ψ L ψL + conj.. (15.280)
2
where ML = m exp [i(β2 − β1 )]. Note that the phase may be at will transferred from ML
to ψR .
The apparent generalization of the Lagrangian (15.280) on the four-component spinors is
given by the expression:

L = i(ψ R γµ ∂ µ ψR + ψ L γµ ∂ µ ψL ) + LM
mass , (15.281)

where
1 c c
LMmass = − {MR ψ R ψR + ML ψ L ψL + conj.}. (15.282)
2
It will be recalled that for Dirac particles the mass term in the Lagrangian had the form:

LD
mass = −MD {ψ L ψR + conj.}. (15.283)

Thus, in the Lagrangians for particles with the spin 1/2 two types of the mass terms are
possible. And further they are referred to as the Dirac and Majorana mass, respectively.
In the most general case the mass term in the Lagrangian is defined by the expression:

1   c

c c ML MD ψL + ψL
Lm = − ψ L + ψ L , ψ R + ψ R c . (15.284)
2 MD MR ψR + ψR

When introducing into the Lagrangian the Majorana-mass terms, the Lagrangian becomes
noninvariant with respect to the gauge transformations providing the conservation of the
lepton flavor:
ψ( x) → ψ( x) exp (iLβ). (15.285)
Because of this, the Majorana theory predicts the existence of the processes with the lepton
flavor violation |∆L| = 2. The 0ν2β-process furnishes an excellent example of such an
exotics.
Let us introduce the designations:

σ µ ≡ (1, −σi ), σ̃ µ ≡ (1, σi )

and pass to the two-components spinors η in the Lagrangian (15.280):

im T

L = iη † (σ µ ∂µ )η − η σ2 η − η † σ2 η ∗ , (15.286)
2

© 2011 by Taylor and Francis Group, LLC

 354 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where for the sake of simplicity we have set β1 = β2 . By application of the variational
principle, the Majorana equation in the form of the two-components spinor is followed from
the Lagrangian (15.286) 
σ µ ∂µ η + mσ2 η ∗ = 0,
(15.287)
σ̃ µ σ2 ∂µ η ∗ − mη = 0.
Eliminating η ∗ from Eq. (15.287) we arrive, as we might expect, at the Klein-Gordon
equation:
( + m2 )η = 0.
The representation of a truly neutral spinor field in terms of ψL or ψR is not the only
possibility. For example, the Majorana spinors may be also represented as:
ψ + ψc
φ+ = λ+ √ , (15.288)
2
or
ψ − ψc
φ− = λ− √ . (15.289)
2
Since ψ and ψ c have the identical transformation properties, then the Majorana condition:

φc± = λ± φ±

appears to be relativistic covariant. It goes without saying that physics should be indepen-
dent of the construction method for Majorana’s spinors, that is, the representations used by
us are unitary equivalent [28]. Within the scope of a second-quantized theory, the unitary
equivalence points to the existence of the operator U that secures the fulfillment of the
relations:
U φ− U −1 = ζ, U φ+ U −1 = χ. (15.290)
With due regard for the definition of the spinors χ and ζ from Eq. (15.290) follows:
1 − γ5 1 + γ5
U −1 ψL U = φ+ , U −1 ψR U = φ− , (15.291)
2 2
1 + γ5 1 − γ5
U −1 ψL
c
U= φ+ , U −1 ψRc
U =− φ− . (15.292)
2 2
We direct our attention to the fact that the fulfillment of (15.291) and (15.292) makes it
possible to describe the Majorana neutrino simply with the use of the Weyl spinors ψL and
ψR rather than their combinations satisfying the Majorana condition. To prove relations
(15.291) and (15.292) we consider the simplest case, namely, when mν = 0. This enables
us to use the following expansion for the operator field function of the neutrino:
Z 2
1 X
ψ(x) = d3 p [ar (p)u(r) (p)e−ipx + b†r (p)v (r) (p)eipx ], (15.293)
(2π)3/2 r=1

where the bispinors u(r) (p) and v (r) (p) satisfy the equations (see, (15.113) )

γ5 u(r) (p) = qu(r) (p), γ5 v (r) (p) = −qu(r) (p) (15.294)

with q = 1 for r = 1 and q = −1 for r = 2. Under the proof we also need the expansions
of the field operators entering into Eqs. (15.291) and (15.292) in terms of the plane waves.
With allowance made for
ψ c = ψ(ar (p) ↔ br (p))

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 355

and Eqs. (15.258), (15.294), we get:

Z
1
ψL (x) = d3 p[a1 (p)u(1) (p)e−ipx + b†2 (p)v (2) (p)eipx ], (15.295)
(2π)3/2
Z
1
ψR (x) = d3 p[a2 (p)u(2) (p)e−ipx + b†1 (p)v (1) (p)eipx ], (15.296)
(2π)3/2
Z
1
c
ψL (x) = d3 p[b2 (p)u(2) (p)e−ipx + a†1 (p)v (1) (p)eipx ], (15.297)
(2π)3/2
Z
1
c
ψR (x) = d3 p[b1 (p)u(1) (p)e−ipx + a†2 (p)v (2) (p)eipx ], (15.298)
(2π)3/2
Z 2
1 X
φ± (x) = √ d3 p {[ar (p) ± br (p)]u(r) (p)e−ipx + [±a†r (p)+
(2π)3/2 2 r=1

+b†r (p)]v (r) (p)eipx }. (15.299)

Intuitively we feel that the matrix U should take the form:
X
U = exp [θ (b†r ar − a†r br )],
r

assuming by this Im(θ) = 0. Carrying out simple manipulation, we obtain:

X θ2 X X
U −1 as U = as + θ[ (b†r ar − a†r br ), as ] + [ (b†r ar − a†r br ), [ (b†r′ ar′ − a†r′ br′ ), as ]]+
r
2! r ′ r

+... = as cos θ + bs sin θ (15.300)

and
U −1 bs U = as cos θ − bs sin θ. (15.301)
When θ equals π/4, from the formulae obtained follows:
√ −1 √ −1 
√2U −1 a1 (p)U = a1 (p) + b1 (p), √ 2U−1 a2 (p)U = a2 (p) − b2 (p), (15.302)
2U b1 (p)U = b1 (p) − a1 (p), 2U b2 (p)U = b2 (p) + a2 (p).

By means of Eqs. (15.302) we can easily establish the validity of the relations (15.291) and
(15.292), and our task is accomplished.

15.13 Quantization of Majorana field

At first, we carry out the quantization of the Majorana neutrino field with the use of the
Weyl’s spinors η(x) and η ∗ (x) [29]. In the most general case the operator expansion of the
spinors η(x) and η ∗ (x) in terms of the plane waves have the form:
Z
1
η(x) = d3 pN {[a(p)β(p)+b(p)α(p)]e−ipx +[c(p)β(p)+d(p)α(p)]eipx ], (15.303)
(2π)3/2

© 2011 by Taylor and Francis Group, LLC

 356 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
Z
1
η ∗ (x) = d3 pN {[a† (p)β ∗ (p) + b† (p)α∗ (p)]eipx + [c† (p)β ∗ (p) + d† (p)α∗ (p)]e−ipx ],
(2π)3/2
(15.304)
where N is a normalization factor to be determined, and the two-components spinors α(p)
and β(p) obey both the eigenvalue equations:
(p · σ) (p · σ)
α(p) = α(p), β(p) = −β(p) (15.305)
|p| |p|
and the normalization conditions:

α† (p)α(p) = β † (p)β(p) = 1,
(15.306)
α† (p)β(p) = β † (p)α(p) = 0.

Substitute the expressions (15.303) and (15.304) into the motion equations (15.287). These
equations will be converted to identities under fulfillment of the following conditions:

σ2 α∗ (p) = iβ(p), σ2 β ∗ (p) = −iα(p), (15.307)

m m
d(p) = a† (p), b(p) = − c† (p). (15.308)
Ep + |p| Ep + |p|
Introducing the designations

a(p) ≡ a− (p), c† (p) ≡ a+ (p),

we could rewrite the expansion of η(x) in the form:

Z
1 m
η(x) = d3 pN {[a− (p)β(p) − a+ (p)α(p)]e−ipx +
(2π)3/2 Ep + |p|
m
+[a†+ (p)β(p) + a† (p)α(p)]eipx }. (15.309)
Ep + |p| −
Since “coordinates” and “momenta” of the Majorana field are not independent, we have
to use the relativistic quantization scheme. At the first step we define the dynamical ob-
servables. The energy-momentum tensor will look like:

T µν = iη † σ µ ∂ ν η − g µν L. (15.310)

With its help the energy-momentum vector could be found

Z
i
Pµ = d3 x{2η ∗ (x)∂ µ η(x) − g 0µ ∂t [η ∗ (x)η(x)]}. (15.311)
2
To insert the expansions of η(x) and η ∗ (x) into Eq. (15.311) gives the expression for the
field energy: Z
2Ep
E = N 2 d3 p p0 [a† (p)a− (p) − a+ (p)a†+ (p)]. (15.312)
Ep + |p| −
From Eq. (15.312) it has become obvious that N should be chosen in the form:
s
Ep + |p|
N= . (15.313)
2Ep

Further, by the standard way one may establish nonzero anticommutators:

{a± (p), a†± (p′ )} = δ (3) (p − p′ ) (15.314)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 357

where a+ (p) (a− (p)) and a†+ (p) (a†− (p)) have meaning of the production and annihila-
tion operators of particles with the positive (negative) helicity, mass m and momentum p,
respectively.
Let us proceed to the definition of singular functions of the Majorana field. From Eq.
(15.305) and the orthonormalization conditions for the spinors α(p) and β(p) it is easy to
obtain:
1 1
αr (p)α∗s (p) = (|p| + σj pj )rs , βr (p)βs∗ (p) = (|p| − σj pj )rs . (15.315)
2|p| 2|p|

By application of the formulae (15.315) we could establish the values of the anticommutators
for the operator field functions at an arbitrary instant of time:

{ηr (x), ηs∗ (x′ )} = (σ̃ µ ∂µ )rs ∆0 (x − x′ ; m2 ), (15.316)

{ηr (x), ηsT (x′ )} = m(σ2 )rs ∆0 (x − x′ ; m2 ). (15.317)

This in turn allows us to find nonzero convolutions of the Majorana field operators:

< 0|T [ηr (x)ηs† (x′ )]|0 >= (σ̃ µ ∂µ )rs ∆c (x − x′ ), (15.318)

< 0|T [ηr (x)ηsT (x′ )]|0 >= m(σ2 )rs ∆c (x − x′ ). (15.319)
Again expression (15.319) supports the conclusion that the lepton flavor violation in the
case of the Majorana neutrino takes place for nonzero mass of the neutrinos only.
The above formalism of the two-component spinors is convenient for the calculation of
matrix elements when the Majorana neutrinos are the sole fermions involved in the inter-
action Lagrangian Lint . Provided Lint also includes, for instance, the charged leptons, we
are obliged to go to the two-component spinors for them as well. This procedure is not only
devoid of aesthetic attractiveness but also results in a considerable increase of the terms in
the amplitude of the corresponding reaction. Because of this reason to describe Majorana’s
neutrinos we shall use the most habitual formalism of the four-component spinors.
In the interaction Lagrangian of any electroweak theory, no matter the SM or its ex-
tension, the neutrino field function is accompanied by chirality projection operators. For
example, in the SM the Lagrangian describing the interaction of neutrinos with gauge bosons
has the form:
g g
Lνint = ν l (x)γµ (1−γ5 )νl (x)Z µ (x)+ √ l(x)γµ (1−γ5 )νl (x)W µ (x)+conj., (15.320)
4 cos θW 2 2
where we are not concretizing the neutrino nature. In the SM for every generation we have
the only left-handed neutrino (or more precisely, preferentially left-handed neutrino, but
this aspect will be considered later). In some extensions of the SM to this neutrino a heavy
right-handed neutrino is added. For now, let us direct our attention to the SM only. Then
for the Majorana neutrino it is reasonable to select the representation:

ν M (x) = λ⊙ [ψL (x) + ψL

c
(x)]. (15.321)

Taking into account the formulae (15.249) and (15.259), we get:

{ναM (x), ν M ′ ′
β (x )} = −iSαβ (x − x ). (15.322)

As directly follows from (15.322), convolution of the operators ναM (x) and ν M ′
β (x ) for the
Majorana field has the same value as for the Dirac field, that is,


< 0|T ναM (x)ν M ′ c
β (x ) |0 >= −iSαβ (x − x ).
′
(15.323)

© 2011 by Taylor and Francis Group, LLC

 358 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

But unlike the Dirac case, owing to the relation (15.260), the following quantity is also
different from zero:


< 0|T ναM (x)νβM (x′ ) |0 >= −iλ2⊙ mαβ ∆c (x − x′ ). (15.324)

It should be noted that at will we may remove the charge conjugation operator from the
reaction amplitude. To this end, the operator C should be set up on the field function of the
lepton forming with the neutrino a neutral or charged current (see the expression (15.320) ).
Obviously, it is more convenient to use the expansion of the Majorana field operator simply
in terms of the solutions of the Dirac equation. Such an expression may be obtained if
we recall that the constant orthonormalized spinors ς ± , contained in the operator function
expansion for the electron-positron field, are determined with an accuracy of the phase.
Then the desired expansion for ν M (x) could be represented in the shape:
Z r
1 m X
ν M (x) = 3/2
d3
p [aα (p)u(α) (p)e−ipx + λ⊙ a†α (p)vm
(α)
(p)eipx ], (15.325)
(2π) Ep α=±

where the bispinor u(α) (p) is defined by the formula (15.241) as before while the bispinor
(α)
vm (p) is given by the expression:
 
(p · σ) ±
ς
(±)
vm (p) = N  Ep + m m  , (15.326)
±
ςm
   
+ 0 − 1
ςm =− , ςm = .
1 0
In Eq. (15.325) we have placed the phase factor before the Majorana particle production
operator. This choice makes it possible to call λ⊙ the phase factor of the production.
It should be emphasized that nothing interferes with shifting of λ⊙ to the annihilation
operator. A direct checking demonstrates that when ν M (x) is represented as (15.325) the
Majorana condition appears to be fulfilled:
Z r
M c 1 3 m X †
[ν (x)] = 3/2
d p [aα (p)Cγ0 u(α)∗ (p)eipx + λ∗⊙ aα (p)Cγ0 vm
(α)∗
(p)e−ipx ] =
(2π) Ep
α)
Z r
1 m X
= λ∗⊙ d3 p [aα (p)u(α) (p)e−ipx + λ⊙ a†α (p)vm
(α)
(p)eipx ],
(2π)3/2 Ep α
where we have taken into consideration the relations:
(α)
[vm (p)]c = C[v (α) T
m (p)] = u
(α)
(p), [u(α) (p)]c = C[u(α) (p)]T = vm
(α)
(p). (15.327)

15.14 Second quantized representation of discrete operations C, P ,

and T for particles with spin 1/2
Second quantized operators of discrete transformations are operators in the Fock space
and, consequently, do not commute with production and annihilation operators. In order
to distinguish them from operators transforming unquantized wavefunctions, we shall use

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 359

the following designations: C for the charge conjugation operator, P for the space inversion
operator, and T for the time inversion operator. It is obvious that the vacuum must be
invariant with respect to these transformations, that is,

CΦ0 = PΦ0 = T Φ0 = 0. (15.328)

The action of the operator U (U = C, P, T ) on a system state vector that is given in the
Fock representation (15.244) leads to the result:
Z Z
U Φn+ −
p n ′ ...
= ... F ′ (x1 , ..., xn )U ψ b1 (x1 )U −1 ...U ψ bn (xn )U −1 d4 x1 ...d4 xn Φ0 ,
1 p1

where F ′ = F when U is unitary, and F ′ = F ∗ when U is nonunitary. Thus, calling

for invariance of the theory with respect to the U -transformation, we reduce our problem
to establishing of the transformation laws for the operator field functions, that is, to the
determination of the quantities: U ψ(x)U −1 and U ψ(x)U −1 . In so doing, in accordance
with the correspondence principle, we have to take into account the transformation laws
for the unquantized field functions stated by us in Section 15.8, on the one hand. And
on the other hand, since the operators U have no effect on the spinor components of the
operator field functions,∗ use of the motion equations (as in case of a unquantized theory)
does not result in the definitive establishment of the transformation laws. In other words,
in a second quantized theory, when searching for the transformation laws, we are in need
of a new criterion. It is clear that only invariance of the permutation relations (PR) with
respect to the transformations C, P and T could be such a criterion.
Let us consider the charge conjugation operation. According to the results of Section
15.8, we define this operation as:
c
Cψ(x)C −1 = ηc ψ c (x), Cψ(x)C −1 = ηc∗ ψ (x). (15.329)

Direct calculations indicate that, with such a definition of the C-operation, invariance of
the PR is provided under the condition that the C-operator is unitary, as distinct from a
primary quantized theory. The invariance with respect to the C-operation takes place for
an electrically neutral system only. In this case the state vectors of the initial system (Φ)
and the transformed one (Φc ) may be distinguished from each other only in a phase factor:
N
Y
Φc = ηC Φ = (ηc )i Φ.
i

To display this circumstance we have introduced the phase factor of the charge conjugation
operation |ηc | = 1.
Now, by application (15.329) we establish transformation properties of the production
and annihilation operators. Using the expansion of the operator field function (15.240), we
get:
1+α 1+α
Caα (p)C −1 = ηc (−1) 2 bα (p), Cbα (k)C −1 = ηc∗ (−1) 2 aα (p), (15.330)

where the following identities for the bispinors:

1+α 1+α
Cu(α) (p) = (−1) 2 v (α) (p), Cv (α) (p) = (−1) 2 u(α) (p) (15.331)

∗ Tobe more specific u(α) (p) and v(α) (p) commute with the unitary operator U , whereas under permutation
with antiunitary operator U these bispinors are replaced with complex conjugate ones.

© 2011 by Taylor and Francis Group, LLC

 360 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

have been taken into consideration. For the derivation of (15.331) it is convenient to test
them first in the Dirac-Pauli representation, and, next, to make sure that the derived equa-
tions do not depend on the representation choice. For the production operators the corre-
sponding transformation laws may be found from the relations (15.331) with an allowance
made for the condition:
(U cU −1 )∗ = U c∗ U −1 ,
that is valid both for the unitary and antiunitary operator U as well as for an arbitrary
operator c.
We may take the opposite course, namely, we define the charge conjugation operation
as a transformation of the fermion with the given spin orientation into the antifermion
with the same spin orientation. In this case for the fermion annihilation operators we
1+α
immediately obtain the relations (15.240) with the omitted factor (−1) 2 , from where the
transformation laws of the operator field function ψ(x) may be established.
Reasoning in a similar way, we can determine the space and time inversion operations:

Pψ(r, t)P −1 = ηp γ0 ψ(−r, t), Pψ(r, t)P −1 = ηp∗ γ0 ψ(−r, t), (15.332)

T ψ(r, t)T −1 = ηt γ 3 γ 1 ψ(r, −t),
(15.333)
T ψ(r, t)P −1 = ηt∗ γ 1 γ 3 ψ(r, −t),
where |ηp | = 1, |ηt | = 1. In this case the operator P must be unitary and the operator
T —antiunitary. Using the expansion of ψ(x) in the form (15.240), we define the behavior
of the annihilation operators under the P- and T -transformations

Paα (p)P −1 = ηp aα (−p), Pbα (p)P −1 = −ηp∗ bα (−p), (15.334)

1+α
)
T aα (p)T −1 = −ηt (−1) 2 a−α (−p),
1+α (15.335)
T bα (p)T −1 = ηt∗ (−1) 2 b−α (−p).
When deriving the relations obtained the identities
1+α
)
γ 3 γ 1 u(α)∗ (p) = −(−1) 2 u(−α) (−p), γ0 u(α) (p) = u(α) (−p),
1+α (15.336)
γ 3 γ 1 v (α)∗ (p) = (−1) 2 v (−α) (−p), γ0 v (α) (p) = −v (α) (−p)

have been allowed for. From (15.334) and (15.335) it follows that, while the space inversion
results only in the sign changing for the three-dimensional particle momentum, in the case
of the time inversion the sign is changed both for the momentum p and spin projection α.
The product of all three operators Θ = CPT gives:

Θψ(x)Θ−1 = ηcpt γ5 γ0 ψ(−x), Θψ(x)Θ−1 = ηcpt

∗
γ5 γ0 ψ(−x), (15.337)

where ηcpt = ηc ηp ηt . So the Θ = CPT -operation leads to the replacement of a particle by

the antiparticle, with the same momentum and opposite spin projection.
Let us consider Majorana neutrinos behavior under the discrete transformations. For the
operator field function of the neutrino the expansion (15.325) will be used. We start with
the charge conjugation operation. Now, instead of (15.329), we have:

Cν M (r, t)C −1 = (ηc∗ λ⊙ )∗ ν M (r, t),

that, in its turn, leads to the relations:

Caα (p)C −1 = (ηc∗ λ⊙ )∗ aα (p), Ca†α (p)C −1 = (ηc∗ λ⊙ )∗ a†α (p). (15.338)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 361

On the strength of unitarity of the operator C from Eqs. (15.338) one may conclude that
the quantity ηc′ = ηc∗ λ⊙ is real. Then, considering the C-invariance of the vacuum, we get
the relations:
C|p, α >= ηc′ |p, α >, (15.339)
where |p, α >≡ a†α (p)|0 >, that lead us to the conclusion: the free Majorana particle is
the charge conjugation operator eigenstate. However, this statement is not valid for the
physical Majorana neutrino as the weak interaction disturbs the C-invariance.
From Eqs. (15.329) and (15.330) it follows that under the CP-transformation the operator
field function is developed by the law:

CPψ(r, t)(CP)−1 = ηcp γ0 ψ c (−r, t). (15.340)

For the Majorana neutrino the expression (15.340) is converted to the form:

CPν M (r, t)(CP)−1 = (ηcp

∗
λ⊙ )∗ γ0 ν M (−r, t).

Using the relation obtained and with allowance made for the identities

γ 0 u(α) (p) = u(α) (−p), γ 0 vm

(α) (α)
(p) = −vm (−p), (15.341)

one may establish the behavior of the operators aα (p) and a†α (p) under the CP-transformation:

CPaα (p)(CP)−1 = (ηcp ∗
λ⊙ )∗ aα (−p),
(15.342)
CPa†α (p)(CP)−1 = −(ηcp ∗
λ⊙ )∗ a†α (−p).

Compatibility of Eqs. (15.342) would be ensured when the condition:

∗
(ηcp λ⊙ )∗ = −ηcp
∗
λ⊙

be fulfilled. In this way the free Majorana particle may be defined as the eigenstate of the
operator CP with the eigenvalue:
∗
−(ηcp λ⊙ )∗ ≡ ηcp
′
= ±i.

This is very unusual because for the well-known elementary particles the intrinsic CP-parity
is a real number. It is interesting to elucidate whether this statement is at variance with
the reported experimental results. Let us consider the decay:

Z → νl + νl (15.343)

assuming the Majorana nature for the neutrino. We take neutrinos as nonrelativistic par-
ticles, and the CP-invariance as an inviolable law of nature. In the reference frame of the
rest Z-boson the total angular momentum of the initial state equals 1, and, as a result, the
following final states are possible:
3 1

S1 (l = 0, J = 1), P1 (l = 1, J = 0),
3 3 (15.344)
P1 (l = 1, J = 1), D1 (l = 2, J = 1).

Since in the final state we have two identical fermions, the total wavefunction of the system
is asymmetric. This condition is satisfied only by the 3 P1 -state (the space part of the
wavefunction is antisymmetric and the spin part—symmetric). Acting on the final state by
the operator CP we get:
′2
CP|νν; l = 1, J = 1 >= ηcp (−1)l |νν; l = 1, J = 1 >=

© 2011 by Taylor and Francis Group, LLC

 362 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
′2
= −ηcp |νν; l = 1, J = 1 > . (15.345)
Because the CP-parity of the initial state is equal to 1, then from (15.345) it follows:
′2
−ηcp = 1,
′
that is, for the Majorana neutrino ηcp may be really equal to ±i.
In the case of the space inversion a sign of the helicity is changed and hence two states
of the Majorana neutrino are associated with the opposite intrinsic CP-parity. It is known
that in the quark sector the CP-invariance could be violated. There are also doubts about
conservation of the CP-invariance for leptons. Consequently, definition of the Majorana
neutrino as the CP-eigenstate is equally liable not to have a nature of the absolute statement.
Let us turn our attention to the CPT -transformation. Allowing for the Majorana condi-
tion, we can represent the relation (15.337) in the form:

Θν M (x)Θ−1 = ηcpt λ∗⊙ γ5 γ0 C −1 ν M (−x). (15.346)

From (15.346) one may obtain transformation rules for the production and annihilation
operators of the Majorana neutrino:
1+α
Θaα (p)Θ−1 = −ηcpt λ∗⊙ (−1) 2 a−α (p), (15.347)
1+α
Θa†α (p)Θ−1 = ηcpt λ⊙ (−1) 2 a†−α (p), (15.348)
(α)
where the following identities for the bispinors u(α) (p) and vm (p):
1+α 1−α
γ5 γ0 C −1 u(α) (p) = (−1) 2 u(−α)∗ (p), γ5 γ0 C −1 vm
(α)
(p) = (−1) 2 (−α)∗
vm (p). (15.349)

have been taken into account. Carrying out the charge conjugation operation in Eq.
(15.348), we arrive at
1+α
Θaα (p)Θ−1 = ηcpt
∗
λ∗⊙ (−1) 2 a−α (p).
Compatibility of the obtained equation with that (15.347) demands the fulfillment of the
relation:
∗
ηcpt = −ηcpt ,
that is, ηcpt must take the values ±i. Since the vacuum is CPT -invariant, then for one-
particle state of the Majorana neutrino the equation:

CPT |p, α >= η∗′ |p, −α >, (15.350)

1+α
′
where ηcpt = ηcpt λ⊙ (−1) 2 , should be obeyed. Let us note the dependence of the CPT -
operator eigenvalue on the helicity of the initial state. In a system at rest the CPT -
transformation applied to the Majorana neutrino simply rotates its spin. Thus, in accor-
dance with (15.350), we may state that the Majorana neutrino is identical to its antiparticle
with due regard for the following. It passes into itself upon the CPT -transformation accom-
panied by the rotation through 1800 . As at the present time the CPT -invariance is believed
to be a universal truth, Eq. (15.350) should be regarded as a definition for both the free
and physical Majorana particles.
With respect to the discrete transformations, the above-mentioned properties have a
number of very important consequences. We limit ourselves to the consideration of their
effect only on the electromagnetic structure of the neutrino. Let us assume that the neutrino
is in a homogeneous and static electromagnetic field with the magnetic induction B and the

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 363

electric field strength E. Provided the neutrino has a magnetic dipole moment, then the
interaction energy of this neutrino with the external electromagnetic field is determined as:

Eint = −µm < S · B > −µe < S · E > . (15.351)

Taking the neutrino as the Majorana particle, let us find the result of the CPT -invariance of
a theory. We give due consideration to the fact that, with respect to the CPT -transformation,
B and E are invariant, whereas the spin direction is changed. Then as a result of the CPT -
transformation, Eint changes over to -Eint . In this case we can save the CPT -invariance
only on the assumption that the Majorana neutrino has neither magnetic nor electric dipole
moments. As is assumed by the CPT -invariance of the theory, the Majorana neutrino has
only one multipole moment that is referred to as an anapole moment µa [30]. This moment
may be observed when placing the particle into an external vortex magnetic field. And the
arising interaction energy of the neutrino in the nonrelativistic limit is given by:

Eint = −µa < S · rotH > . (15.352)

To cite a single example of the source for this moment, we take the toroid distribution of
the electric current (toroidal winding).
As regards the Dirac neutrino, it can possess both the electric and magnetic dipole mo-
ments as well as the anapole moment.

15.15 Neutrino in the Dirac and Majorana theory

If the neutrino is a Majorana particle, in some processes involving the neutrino the lepton
charge conservation law will be violated. Recall that this law was introduced not only for
the successive subdivision of the particles into leptons and baryons. Also, it was called up
to explain why in the process of interactions between the neutrino and matter the reactions:

νl + (A, Z) → l + (A, Z + 1), (15.353)

where l = e− , µ− , τ − , were allowed, whereas the reactions:

ν l + (A, Z) → l + (A, Z + 1) (15.354)

were found forbidden. Taking for the leptons (νl and l) the lepton charge L being equal to
1, for antileptons (ν l and l) L = −1, and assuming:
X
Li = const,
i

we have solved this problem without any difficulties. In so doing νl and ν l represented a
particle and antiparticle, respectively, each of which may be in two states with different
helicity values. In such an approach the neutrino is considered to be the Dirac particle.
But this interpretation is not the only possible one. We could find an explanation in speci-
ficity of the interaction of the neutrino with matter as well. We assume the P-noninvariant
weak interaction is so structured that each right-hand neutrino in the process of interaction
with a matter produces l, and left-handed neutrino—l. In the SM the Lagrangian describing
the interaction of the neutrino with a matter has just this very form (see Eq. (15.320) ).
Using this Lagrangian, there is no need to “provide” the neutrino with any new quantum

© 2011 by Taylor and Francis Group, LLC

 364 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

number. Of course, nothing interferes with the assumption that the neutrino is a truly neu-
tral particle, that is, νl and ν l are simply two states of one particle with different helicity
values. The lepton number of such neutrinos is zero and hence we consider them as particles
having the Majorana nature. If mν = 0 then, according to the Lagrangian (15.320), the
Dirac neutrino represents the left-handed particle while the Dirac antineutrino represents
the right-handed particle. The validity of this statement is obvious. A produced neutrino
is described by the bispinor u(α) (p) followed by the factor (1 + γ5 ). In a massless case the
bispinor u(α) (p) is an eigenfunction of the chirality operator that in turn is coincident with
the helicity operator. Taking account of the identity:
u(α) (p)(1 + γ5 ) = u(α)† (p)(1 − γ5 )γ0 = [γ0 (1 − γ5 )u(α) (p)]† ,
we come to the conclusion that a helicity value of the neutrino is negative. The production
of an antineutrino is associated with a factor in the matrix element (1 − γ5 )v (α) (p). Since
in the case of antiparticles the operator (1 − γ5 ) projects the states with the spin oriented
in the direction of motion, the produced antineutrino proves to be a right-handed particle.
Inclusion of mass leads to the appearance of the right-handed (left-handed) helicity on
the neutrino (antineutrino). We took no notice of these components, for example, in the
phenomena of β-decay because the neutrino mass is extremely small. To illustrate, assuming
m/E ≪ 1 and using expansion for the neutrino operator wavefunction (15.240), in the case
of the bispinor u(α) (p) we get:
 r 
E+m α r 
r
m (α) ς ςα
  α 
≈ 1 m ς
 2E 
u (p) =  r α + √ α . (15.355)
E  E−m
α
 2 ας 2 2E −ας
(n · σ)ς
2E
Set α = −1 and act the operator ΛR on (15.355). Then, as expected, the first term goes to
zero and the contribution of the second term is of the form:
 −
m ς
√ − .
2 2E ς
The derived relation means that the amplitude of the “wrong” (right-handed) helicity of the
neutrino is proportional to m/E. In a similar way we can demonstrate that, when the mass
is included, for the antineutrino the emerging left-handed component is also proportional to
m/E. Thus, in the processes described by the Lagrangian (15.320), the ultrarelativistic neu-
trinos are produced mainly in the left-helicity state, whereas for the produced antineutrinos
the preferential helicity is right-handed. We should emphasize that this helicity character
is true for the case of the Majorana neutrino too.
At first glance it seems that establishing whether a neutrino is the Majorana or Dirac
particle is an easy task. To this end, it is sufficient to create a beam of, for instance,
left-handed neutrinos νl . Next, it suffices to realize the neutrino spin-flip with the help
of some mechanism, and then to direct these right-handed neutrinos to the target. If
during interaction with the target the processes described by (15.354) are also detected, the
neutrino has the Majorana nature; otherwise it has the Dirac nature. We emphasize this is
true for the standard model supplemented with neutral left-handed leptons NlR (l = e, µ, τ ).
In this case NlR interact with the Higgs boson only and are actually sterile particles.
Unfortunately, under the state of the art in technology such experiments may be assigned
to the category of mental ones only. Nevertheless, numerous experiments involving the
neutrino may be currently conducted. Let us try to elucidate why the nature of the neutrino
has not been established until the present time. As an example, we investigate the reaction:
νl + l → νl + l. (15.356)

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 365

The corresponding Feynman diagrams are shown in Fig. 15.1, where the waved lines display
the propagations of the virtual gauge bosons (W and Z). First, we consider the diagram
connecting with the Z-boson exchange. The matrix element associated with it is given by

νl l νl νl

W− Z

l νl l l

FIGURE 15.1
The Feynman diagram describing the process νl + l → νl + l.

the expression:
Mνl l→νl l =< lf ; νf |A|li ; νi >, (15.357)
where  
g2 µ ∂µ ∂ν
A= ν l (x)γ (1 − γ5 )νl (x) gµν + ∆c (x − y; m2Z )×
16 cos2 θW m2Z
×l(y)γ ν [γ5 + 4 sin2 θW − 1]l(y), (15.358)
|li ; νi >= a†αν (pνi )a†αl (pli )|0 >, |lf ; νf >= a†αν (pνf )a†αl (plf )|0 > .
i i f f

and the factor (gµν +∂µ ∂ν /m2Z )∆c (x−y; m2Z ) corresponds to the Z-boson propagator. When
writing Eq. (15.357), we have taken into account the form of the Lagrangian describing the
interaction of leptons with the Z-bosons (11.44).
In the matrix element (15.357) we are interested only in the part associated with the
neutrino current:

Jµ ≡< lf ; νf |jµ |li ; νi >=< lf ; νf |ν l (x)γµ (1 − γ5 )νl (x)|li ; νi > . (15.359)

First, we demonstrate that for the Majorana neutrino a vector part of the neutrino current
jµM becomes zero. With this in mind we consider the quantity ν cl (x)γµ νlc (x). Taking into
account the γ-matrices behavior under the charge conjugation and the fact of anticommuting
the operator field functions, we find:

ν cl (x)γµ νlc (x) = νlT (x)γµT ν Tl (x) = (νl )α (γµ )βα (ν l )β =

= −(ν l )β (γµ )βα (νl )α . (15.360)

On the other hand, due to the Majorana condition, we have

ν cl (x)γµ νlc (x) = ν l (x)γµ νl (x). (15.361)

© 2011 by Taylor and Francis Group, LLC

 366 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Comparison of the obtained relation with Eq. (15.360) leads to the conclusion:

ν l (x)γµ νl (x) = 0, (15.362)

which is what we set out to prove.

Inserting the operator field functions expansion (15.325) into (15.359) and allowing for
the anticommutation relations (15.314), we get

(JµM )if = −u(αf ) (pf )γµ γ5 u(αi ) (pi ) + v (αi ) (pi )γµ γ5 v (αf ) (pf ). (15.363)

Because  c
v (α) (p) = u(α) (p) = Cu(α)T (p), (15.364)

then
v (α) (p) = −u(α)T (p)C −1 . (15.365)
Now the last term in the right-hand side of (15.363) may be rewritten in the form:

v (αi ) (pi )γµ γ5 v (αf ) (pf ) = −u(αi )T (pi )C −1 γµ γ5 Cu(αf )T (pf ) =

= −u(αi )T (pi )(γµ γ5 )T u(αf )T (pf ) = −u(αf ) (pf )γµ γ5 u(αi ) (pi ),
which finally gives
(JµM )if = −2u(αf ) (pf )γµ γ5 u(αi ) (pi ). (15.366)
Having done the analogous calculations for the case of the Dirac neutrino, we obtain

(JµD )if = u(αf ) (pf )γµ (1 − γ5 )u(αi ) (pi ). (15.367)

In the ultrarelativistic limit both the initial and final neutrinos are predominantly left-
handed particle, that is, the relation:
m
γ5 u(−) (p) ≈ −u(−) (p) + O
2E
m
(the same is true for the Majorana neutrino as well) takes place. Then, at E → 0 we have
for the Dirac current

u(αf ) (pf )γµ (1 − γ5 )u(αi ) (pi ) −→ −2u(αf ) (pf )γµ γ5 u(αi ) (pi ), (15.368)

that allows us to conclude

(JµD )if ≈ (JµM )if . (15.369)
Again, we see that a value of the effect enabling one to distinguish between the Majorana
and Dirac neutrino is totally determined by the neutrino mass value. In other words, in
experiments with the real neutrino the chances to establish the neutrino nature are growing
as the energy decreases. The result of (15.369) holds not only for the exchange processes
with virtual Z-bosons (processes with neutral currents) but also for those with virtual W -
bosons (processes with charged currents). To prove this statement, it suffices to represent
the matrix element corresponding to the diagram with the W -exchange pictured in Fig.
15.1 as a product of two currents, one of which is the neutrino current of the form given by
(15.359). This may be always realized with the use of the so-called Fierz transformation

(ua Qα ub )(uc Qα ud ) = −(ua Qµ ud )(uc Qµ ub ),

where Qα = γµ (1 − γ5 ), that will be proven later on.

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 367

FIGURE 15.2
The diagram for the process 0ν2β.

Should the processes involving real neutrino fail to settle the dilemma with the nature of
the neutrino, we have to consider the reactions involving the virtual neutrino. One of the
examples is furnished by the process 0ν2β whose diagram is shown in Fig. 15.2. Emission of
two W -bosons by the initial nucleus is described by the matrix element calculated within the
scope of the nuclear physics. Subsequent production of an electron pair due to the exchange
of virtual neutrino is already associated with the sphere of the elementary particle physics.
Proceeding further, we base our reasoning on the SM, that is, the theory comprises only the
left-handed charged currents. In this case the Lagrangian form (15.320) dictates us that the
neutrino emitted at the vertex 1 looks like ν and that absorbed at the vertex 2 looks like ν.
In consequence, the process proceeds only on condition that ν and ν are identical particles.
In other words, for the Dirac neutrinos it is forbidden, being allowed for the Majorana ones.
The constituent part of the diagram in Fig. 15.2 is a diagram associated with the sub-
process:
W ∗− W ∗− → e− + e− , (15.370)
where W ∗ denotes the virtual W -boson. Note that process (15.370) in itself is of no special
interest, while the time-reversal process:

e− + e− → W − + W , (15.371)

that is referred to as inverse double neutrinoless β-decay, attracts lively interest and will be
studied at next-generation electron-positron colliders (for example NLC) operating in the
e− e− -mode. It is planned to study a similar process, also supporting the Majorana nature
of the neutrino:
µ− + µ− → W − + W − (15.372)
at the muon colliders (FMC and NMC).
Let us demonstrate that in the case of 0ν2β the effect is also determined by the neutrino
mass. The neutrinos involved in the Lagrangian (15.320) belong to a definite lepton family,
that is they are defined on the flavor basis. At the same time, both the Dirac and Majorana
equations are written for particles with a specific mass νj (j = 1, 2, 3). Then, in the
Lagrangian (15.320) we should change from the flavor basis to the physical one, that could
be done with the help of the relation:

νl (x) = UljN M νj (x), (15.373)

© 2011 by Taylor and Francis Group, LLC

 368 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where U N M lj is the neutrino mixing matrix defined by the formula (12.1) (further super-
scripts on this matrix will be omitted).
The matrix element of the reaction (15.370) is given by:

MW − W − →e− e− ≡< e− − − −
1 ; e2 |A|W1 ; W2 >,

where
g2
A= e(x)γµ (1 − γ5 )Uej νjs1 (x)W µ (x)e(y)γσ (1 − γ5 )Uej ′ νjs′1 (y)W σ (y) =
8
g2 µ
W (x)W σ (y)L(eν)
= µσ , (15.374)
8
and only the convolutions with j ′ = j are different from zero:

νjs1 (y)νjs1 (x) =< 0|T (νj (y)νj (x))|0 > . (15.375)

For (15.375) we may directly use expression (15.324). But it is difficult to keep from tempta-
tion to demonstrate a trick reducing such quantities to the calculation of the Dirac particle
propagators [31]. So the problem consists in the transition from < 0|T (νj (y)νj (x))|0 > to
< 0|T (νj (y)ν j (x))|0 >. This may be performed using the charge conjugation operation and
Majorana condition. The straightforward calculation yields:

eγµ (1 − γ5 )νj = (ec )c γµ (1 − γ5 )(νjc )c = −(ec )T C −1 γµ (1 − γ5 )C(ν c j )T =

= −(ec )T [−γµ (1 + γ5 )]T (ν c j )T = −νjc γµ (1 + γ5 )ec = −λ⊙j ν j γµ (1 + γ5 )ec . (15.376)

Considering (15.376), the expression (15.374) takes the desired form:

g2
A = −λ⊙j e(x)γµ (1 − γ5 )Uej νjs1 (x)W µ (x)ν sj 1 (y)γσ (1 + γ5 )Uej ec (y)W σ (y)
8
where
νjs1 (y)ν sj 1 (x) =< 0|T (νj (y)ν j (x))|0 >= −iSjc (y − x).
Then, using the explicit form of Sjc (y − x), the orthogonality condition of the operators
(1 − γ5 ) and (1 + γ5 ), we arrive at the same expression for A:

g2 2
A = 2iλ⊙j e(x)γµ (1 − γ5 )Uej mνj ∆c (y − x; m2νj )W µ (x)γσ ec (y)W σ (y), (15.377)
8
which would follow if the relation (15.324) is accounted for.
When calculating the matrix element MW − W − →e− e− , we are interested just in the lepton
factor, that is,
− − (eν) − −
M(eν)
µσ =< e1 ; e2 |Lµσ |W1 ; W2 > . (15.378)
Using the operator field function expansions of an electron (15.219), (15.220) and a neutrino
(15.325), the Fourier representation for the causal function of the Dirac particle (15.254) and
allowing for the anticommutation relations for the production and annihilation operators,
we obtain, upon straightforward algebraic manipulation, the expression:

< e− − − − eν µσ
1 ; e2 |A|W1 ; W2 >= Mµσ B , (15.379)

where
2
(eν)
λ⊙j Uej mνj
Mµσ = 2u(α1 ) (p1 )γµ γσ (1 + γ5 )v (α2 ) (p2 ) (15.380)
k + m2νj
2

© 2011 by Taylor and Francis Group, LLC

 Particles with spin 1/2 369

p1,2 are three-dimensional electron momenta, k is four-dimensional virtual neutrino mo-

mentum, and B µσ denotes the factor that is insignificant for our line of reasoning. In this
way the matrix element of the 0ν2β-process is actually determined by the neutrino mass,
that is: X
2
M0ν2β ∝ λ⊙j Uej mνj . (15.381)
j
2
We can demonstrate that on the assumption of CP-invariance the quantity λ⊙j Uej appears
′ 2
to be equal to −iηcp (νj )|Uej | and hence is a real number. All information concerning
the neutrino mass obtained from a study of the 0ν2β-process or of the inverse double
neutrinoless beta-decay is contained in the quantity:
(ν )
X
′
mefef = −i ηcp (νj )|Uej |2 mνj . (15.382)
j

If the neutrino sector were free from mixing, the right-hand side of the expression (15.382)
(ν )
were simply equal to mνe . Because of this, mefef may be interpreted as the effective mass
of an electron neutrino. Similarly, studying the process,

µ− + µ− → W − + W −
(ν )
we get information about the effective mass of a muon neutrino mefµf . Note that the theory
structure may give the following variant:

mef f ≪ mj .
(ν )
The modern data [32] gives the following value of mefef on 99.73% CL:
(ν )
mefef = (0.1 − 0.9) eV. (15.383)

Among the available explanations to smallness of the neutrino mass with respect to
the masses of charged leptons and quarks, most popular is a so-called seesaw mechanism.
The idea is as follows. Recall how the energy conversion is occurring in the process of
the mathematical pendulum oscillation. Its kinetic energy is maximal at some instant of
time when the equilibrium point is passed through, while the potential energy is zero. As
the point of a maximum deviation from the equilibrium is approached, the kinetic energy
decreases and potential energy is growing. Then let us imagine a system of two particles,
where a role of the energy is played by mass. In this case the heavier one of the particles,
the lighter the other. With the help of the Dirac spinors νl (x) and ν l (x) in each generation
two Majorana spinors are constructed, which are associated with two neutrino states. A
light particle is associated with the ordinary neutrino (ν) while the term “heavy neutrino”
is used for its partner. So the total number of neutrinos in such schemes equals 6. The
masses of light and heavy neutrino are related by the seesaw relation:

mνl mNl = m2l , (15.384)

that is a key to explanation of a small mass of the νl -neutrino. A representative model

with such a neutrino sector is a model based on the SU (2)L × SU (2)R × U (1)B−L -gauge
group [33] (the left-right model). The lower limit for the mass of a heavy electron neutrino,
obtained from analysis of the available experiments, is on the order of 1 TeV [34].
In models with additional neutrinos the chances of establishing the neutrino nature by
the collider experiments, for example, inverse double neutrinoless beta-decay are drastically
increased. For instance, modifying the SM so that the neutrino can be a massive Majorana

© 2011 by Taylor and Francis Group, LLC

 370 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

particle, in the ultrarelativistic case the cross section of the process (15.371) is determined
by [35]:
 2
−3 |Uej |2 mνj
σe− e− →W − W − ≈ 4.16 × 10 pb
GeV
and hence appears to be a small value. In the case of the left- right model the cross section
of the process (15.371) is much greater. For example, when mNe ≈ 1 TeV and Ecm = 200
GeV, it may be as much as several tenths of pb [36].

© 2011 by Taylor and Francis Group, LLC

 16
Massive vector field

Bruin the second inquired if the forest hadn’t

at least a university or some academy which he might burn down.
But in this respect, too, as in others, it turned out that
his predecessor had anticipated his intentions, enrolling all
university students as soldiers of remote garrisons and confining
the academicians in a hollow tree, where, submerged in
lethargy, they remain to this day.
M.E. Saltykov Shchedrin, “Bears in Government”

16.1 Proca equation

To describe a particle with the spin 1 (vector particle), we need a field function having three
independent components only. For a free-moving particle of nonzero mass the rest reference
frame is always available. Obviously, just in this system the intrinsic symmetry properties of
a particle itself are revealed. In this case we should consider symmetry regarding all possible
rotations around the center, that is, symmetry regarding the whole group of spherical
symmetry. The symmetry properties of a particle relative to this group are characterized
by its spin J, determining the number 2J + 1 of the wavefunction components transformed
through each other. When the four-dimensional vector is used as a wave function, in a rest
reference frame the vector particle is associated with three space components of this vector.
In order to make certain that this is the case, we introduce the projection operators:
∂ µ ∂ν ∂ µ ∂ν
Λ(1)µ
ν (∂) = δνµ + , Λ(0)µ
ν (∂) = − , (16.1)
m2 m2
having the properties:
)
(1)µ (0)µ (1)µ (1)ν (1)µ
Λν (∂) + Λν (∂) = δνµ , Λν (∂)Λσ (∂) = Λσ (∂),
(0)µ (0)ν (0)µ (0)µ (1)ν (16.2)
Λν (∂)Λσ (∂) = Λσ (∂), Λν (∂)Λσ (∂) = 0.
The Pauli-Lubanski pseudovector for a field with the spin 1 is given by the expression:
1
(wσ )τ κ = εσµνλ (Mµν )τ κ pλ , (16.3)
2
where the generator of the vector representation of the proper Lorentz group (Mµν )τ κ
coincides with that of the Lorentz group itself:
(Mµν )τ κ = (I µν )τ κ = δτµ δκν − δκµ δτν . (16.4)
Acting by the square of the operator (16.3)
ρ ρτ
(W )γ = (wσ ) (wσ )τ γ , (16.5)

371
© 2011 by Taylor and Francis Group, LLC
 372 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

on the states
ψρ(1) (p) = Λ(1)µ
ρ (p)ψµ (p), ψρ(0) (p) = Λ(0)µ
ρ (p)ψµ (p), (16.6)
where
Λ(s)µ
ρ (p) = Λ(s)µ
ρ (∂) (∂ → −ip) , s = 0, 1,
we get
(W )ργ ψρ(1) (p) = 2m2 ψγ(1) (p), (W )ργ ψρ(0) (p) = 0. (16.7)
2
Since the eigenvalues of the operators (16.5) are equal to m J(J + 1)δγρ , then from (16.7) it
(1) (0)
follows that ψρ (x) describes the states of a particle with the spin 1 whereas ψρ (x) does
those of a particle with the spin 0. In its turn, from the expressions (16.6) it is not difficult
(1)
to understand that in the rest reference frame only the space components of ψρ (x) are
different from zero.
In a similar way we can show that, due to its four-dimensional nature, the field function of
a vector particle may be the components of the antisymmetric second-rank four-dimensional
tensor ψµν (x) whose mixed components ψ0k (x) in a rest reference frame become zero.
Thus, the procedure of constructing an equation of motion for a vector particle is in princi-
ple clear. In the most general case we should take the quantities ψµ (x) and ψµν (x) and then
combine them in a particular manner to provide the fulfillment of both the Klein-Gordon
condition and additional condition that reduces the number of independent components of
the field function to three. In the case of the complex vector field the equation sought has
the form [37]: 
∂ ν ψµν (x) − m2 ψµ (x) = 0, 
ψµν (x) = ∂µ ψν − ∂ν ψµ , (16.8)

(+conj.)
To differentiate the first line of Eq. (16.8) with respect to xµ and allow for a antisymmetry
of ψµν (x) leads to the condition:
∂ ν ψν (x) = 0. (16.9)
In the rest reference frame, where ψν (x) does not depend on the space coordinates, we find:
k 0 ψ0 (x) = mψ0 (x) = 0
What this means is:
ψ0 (x) = 0,
that is, the equations system really describe the particle with the spin 1. Eqs. (16.8) could
be derived from the Lagrangian:
  
1 † 1 µν µ ν ν µ 1 1 †
L =: ψµν (x) ψ (x) − ∂ ψ (x) + ∂ ψ (x) + ψ (x)−
2 2 2 2 µν

−∂µ ψν† (x) + ∂ν ψµ† (x) ψ µν (x) + m2 ψ ν† (x)ψν (x) :, (16.10)
where ψ ν (x), ψ ν† (x), ψ µν (x) and ψ µν† (x) should be considered as the independent general-
ized coordinates. In this case for the description of a vector particle we use the representation
that is a sum of the two irreducible representations, namely, the first: L (1/2,1/2) by which
the four-dimensional vector ψµ (x) is transformed, and the second: (0,1) (1,0) whereby the
antisymmetric tensor ψµν is transformed.
The components ψ0 (x) are of no importance for the dynamics, and at will they may be
painlessly excluded from the theory. Using Eqs. (16.8), we can express them in terms of
the dynamically independent components:
1 µ 1 µ †
ψ0 (x) = − ∂ ψµ0 (x), ψ0† (x) = − ∂ ψµ0 (x), (16.11)
m2 m2

© 2011 by Taylor and Francis Group, LLC

 Massive vector field 373

and the derived relations may be used to determine the dynamic invariants. The momenta,
canonically conjugate to ψµ (x) and ψµ† (x), are given by the expressions:

∂L ∂L
πµ (x) = = ψ µ0† (x), 𵆠(x) = = ψ µ0 (x). (16.12)
∂(∂0 ψ µ (x)) ∂(∂0 ψ µ† (x))

As π0 (x) and π0† (x) are identically equal to zero, we have no problems with lack of the
associated variables ψ0 (x) and ψ0† (x). Taking energy as an example, we can demonstrate
that the dynamic invariants may be expressed in terms of only the generalized “coordinates”
and “momenta.” We have
Z Z
  
H =: d3 x π k (x)∂0 ψ k (x) + π k† (x)∂0 ψ k† (x) : −L =: d3 x (π(x) · π † (x))+

1
+(∇ × ψ(x)) · (∇ × ψ † (x)) + m2 (ψ(x) · ψ † (x)) + (∇ · π(x))(∇ · π † (x))+
m2

1  
+ 2 ∇ · π(x)(∇ · π † (x)) + π † (x)(∇ · π(x)) :=
m
Z h
=: d3 x (π(x) · π † (x)) + (∇ × ψ(x)) · (∇ × ψ † (x)) + m2 (ψ(x) · ψ † (x))+

1 †
+ (∇ · π(x))(∇ · π (x)) . (16.13)
m2
From Eq. (16.13) it follows that the vector field energy H is a positive definite quantity.
We can show that without ruling out the components ψ0 (x) and ψ0† (x), the condition of
positiveness of H were impossible.
At the same time, a vector particle is much more conveniently described by just one
four-dimensional vector ψ ν (x). In so doing, the equation:

(Bµλ,σν ∂ µ ∂λ + m2 δνσ )ψν (x) = 0, (16.14)

where
Bµλ,σν = δµλ δσν − δµν δλσ ,
is the equivalent writing of the Proca equation. Eq. (16.14) contains the additional condition
(16.9), that may be directly demonstrated by acting the operator ∂ σ on this equation and
taking account of the fact that:

Bµλ,σν = −Bµν,σλ = −Bσλ,µν .

The Lagrange function density, of which Eq. (16.14) follows, has the shape:

L =: −Bµλ,σν ∂ µ ψ σ† (x)∂λ ψν (x) + m2 ψµ† (x)ψ µ (x) : . (16.15)

At first sight, it seems that the situation with the vector particle may be simplified still
further. The system of Eqs. (16.8) may be replaced by the equation:

( + m2 )ψµ (x) = 0, (16.16)

and fulfillment of the condition:

∂ µ ψµ (x) = 0

© 2011 by Taylor and Francis Group, LLC

 374 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

may be required independently, or as they say, imposed manually. Of course, Eq. (16.16)
together with the additional condition is equivalent to Eq. (16.14). The Lagrangian corre-
sponding to (16.16) is given by:
L′ =: −∂ µ ψ ν† (x)∂µ ψν (x) + m2 ψµ† (x)ψ µ (x) : . (16.17)
Let us pay attention to signs in the Lagrangians of a vector field. When comparing, for
instance, (16.17) to the Lagrangian of a scalar field LJ=0 (see, (14.4)), we can observe that
the signs of L′ and of LJ=0 are the same for ν = 1, 2, 3 while the signs of both Lagrangians
are opposite for ν = 0. Of course, a choice of the sign in the Lagrangian is not arbitrary.
It is dictated by the energy positiveness requirement. Because the time components of the
field function of the vector particle make no contribution to the total energy, a sign in front
of the space part of the four-dimensional scalar product ψ † (x) · ψ(x) is the decisive factor
in the Lagrangian.
The Lagrangian (16.17) differs from that (16.15) on the quantity:
1 σ 1
L′ − L = − : ∂ [∂µ ψσ† (x)ψ µ (x) + ψµ† (x)∂ µ ψσ (x)] + [(∂µ ∂ σ ψσ† (x))ψ µ (x)+
2 2
+ψµ† (x)(∂ µ ∂ σ ψσ (x))] : . (16.18)
The first term in the right-hand side of (16.18) representing the four-dimensional divergence
excites no apprehension, while the second one suggests an idea that both formalisms may
be nonequivalent. During the calculation of dynamic variables we assume the fulfillment
both of the motion equations and of consequences resulting from these equations. In this
way the second term in the right-hand side of (16.18) becomes zero, and in a free case the
dynamic invariants prove to be equal in the theories involving the Lagrangians L′ and L.
This circumstance could give cause for the statement that both these theories are equivalent.
This is a rather frequent error, though one may easily guess that the indicated equivalence is
retained only in a free case. Indeed, let us consider the introduction of the electromagnetic
interaction:
∂µ → Dµ ≡ ∂µ − ieAµ .
In this case, (16.16) is converted into four independent Klein-Gordon equations, where nat-
urally there is no place for interaction between the spin and electromagnetic field. Besides,
it is not clear what to do with the additional condition: to keep it in the form of (16.9) or
replace it by:
Dµ ψ µ = 0.
It is obvious that an ultimate choice of one or another theory may be made only with the
help of the experiment. Using a complex vector field, we can describe the charged carriers
of the electroweak interactions, the W ± -bosons. The experimental data accumulated by the
present time (measuring cross sections, determining the multipole electromagnetic moments
and so on) support Eq. (16.14) (to be more specific, the Yang-Mills equation, whose linear
approximation is just the Proca equation).

16.2 Dynamic invariants of the vector field

Since in a free case the Lagrangians L′ and L lead to the same expressions for the field
invariants, for this consideration we will use L′ . By the usual way we find the energy-
momentum tensor:
Tµν = −∂µ ψλ† (x)∂ν ψ λ (x) − ∂ν ψλ† (x)∂µ ψ λ (x) − gµν L′ , (16.19)

© 2011 by Taylor and Francis Group, LLC

 Massive vector field 375

the electric current vector:

jµ = ie[∂µ ψλ† (x)ψ λ (x) − ψλ† (x)∂µ ψ λ (x)], (16.20)

and the spin moment tensor:

λ
Sµν = ψν† (x)∂ λ ψµ (x) + ∂ λ ψµ† (x)ψ ν (x) − ψµ† (x)∂ λ ψν (x) − ∂ λ ψν† (x)ψ µ (x). (16.21)

Recall, in order to describe spin properties of particles one uses three space components
0
of the tensor Sµν which, in its aggregate, form a three-dimensional vector S with the
components: Z
Sm = εmnl d3 xSnl 0
, l, m, n = 1, 2, 3. (16.22)

The expansion of the operator field functions in terms of the Proca equation solutions
will look like:
Z 3
1 d3 k X λ
ψµ (x) = √ ϑ (k)[aλ (k) exp (−ikx) + b†λ (k) exp (ikx)], (16.23)
(2π)3/2 2k0 λ=1 µ

Z 3
1 d3 k X λ∗
ψµ† (x) = √ ϑµ (k)[a†λ (k) exp (ikx) + bλ (k) exp (−ikx)], (16.24)
(2π)3/2 2k0 λ=1

where the polarization vectors ϑλµ (k) may be complex quantities in the most general case.
Substituting (16.23), (16.24) into (16.19), (16.20), (16.21) and carrying out the integration
over the space coordinates, we derive the energy-momentum four-dimensional vector:
Z Z
P µ = T 0µ d3 x = − d3 kk µ [ψν(+)† (k)ψ (+)ν (k) + ψν(−)† (k)ψ (−)ν (k)], (16.25)

the electric charge:

Z Z
Q = j d x = e d3 k[ψν(+)† (k)ψ (+)ν (k) − ψν(−)† (k)ψ (−)ν (k)],
0 3
(16.26)

and the spin vector:

Z
S=i d3 k{[ψ(−)† (k) × ψ (−) (k)] − [ψ (+)† (k) × ψ (+) (k)]}, (16.27)

where we have introduced the designations:

3
X 3
X
ψµ(+) (k) = ϑλµ (k)aλ (k), ψµ(−) (k) = ϑλµ (k)b†λ (k).
λ=1 λ=1

Let us consider the polarization vectors ϑλµ (k) in greater detail. Our next objective is
diagonalization of the dynamic invariants. This may be performed with the relevant choice
of the vectors ϑλµ (k). For a while, we forget an operator nature of the field function, assuming
it to be an ordinary wavefunction of nonrelativistic quantum mechanics. For the solution
of Eq. (16.14) we may take the quantity:

ψµ (x) = N ϑµ (k) exp (ikx),

where N is a normalization constant depending on a spin projection on the chosen direction

in space. Direct the particle momentum along the axis x3 . Due to the additional condition

© 2011 by Taylor and Francis Group, LLC

 376 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

(16.9), only three of four components involved in the polarization vector ϑµ are independent.
Let us exclude the time component and use for the remaining space components the following
two choices of three-unit vectors (local reference point):
e1 (k) = (1, 0, 0), e2 (k) = (0, 1, 0), e3 (k) = (0, 0, 1), (16.28)
and
1 1
e′1 (k) = √ (1, i, 0), e′2 (k) = √ (1, −i, 0), e′3 (k) = (0, 0, 1). (16.29)
2 2
Suppose that we work with the local reference point eλ (k) and the particle polarization
vector is equal to e1 (k). In this case the vector ψ(x), varying in time, invariably remains to
be directed along the x1 -axis. Because of this, they say that a particle with the wavefunction
ψ(x) is linearly polarized along the x1 -axis. And if we were concerned with photons, the
essence of this statement would be more clear for a beam of photons being situated in
the same state with the ψ(x)-function. Then in the electromagnetic field under study
the electric field strength vector E = −∂0 ψ(x) would be making the oscillation along the
x1 -axis, that is, we deal with the case of a field linearly polarized in the x1 direction.
However, the strength vectors E and H are observable for the electromagnetic field only.
And unfortunately, the classical physics obviousness speaking to the heart, has no analogs
for massive vector fields.
When the vector ψ(x), varying in time, permanently remains to be directed along the
x2 - or x3 -direction, we have a vector particle linearly polarized along the axis x2 or x3 ,
respectively. The states with ψ(x) k e3 (k) correspond to the longitudinal (relative to the
momentum) polarization while those with ψ(x) k e1 (k) or ψ(x) k e2 (k) correspond to the
transverse polarizations.
Now we assume that in the case of the local reference point (16.29) the polarization vector
equals e′1 (k), that is,
ψ1 (x) 6= 0, ψ2 (x) 6= 0, ψ3 (x) = 0.
Then, in the plane z = 0 we arrive at the following expressions for the transverse components
ψ1 (x), ψ2 (x):
N N  π
ψ1 (t) = √ exp (ik0 t), ψ2 (t) = √ exp ik0 t + ,
2 2 2
i.e., for the resultant oscillation the component along the x2 -direction passes ahead of the
component in the x1 -direction by π/2. All this means that the vector ψ(t) is rotating, while
being constant in value. This situation is referred to as the circular polarization of the field.
Provided the polarization vector is e′2 (k), then the particle is again circularly polarized but
this time the rotation direction of the vector ψ(t) is opposite to the previous one.
Now we present this consideration in the relativistically covariant form. It is easily
seen that the normalization condition and the additional condition (16.9) result in the
polarization four-dimensional vectors, which should satisfy the following relation:
3  
X kµ kν
ϑλ†
µ (k)ϑλ
ν (k) = − g µν − . (16.30)
m2
λ=1

By the direct calculations we can check that at kµ = (k0 , 0, 0, |k|) the condition (16.30) is
satisfied by the following two sets of the polarization vectors:

ϑ1µ (k) = (0, 1, 0, 0), ϑ2µ (k) = (0, 0, 1, 0),
(16.31)
ϑ3µ (k) = m−1 (|k|, 0, 0, k0 ),

© 2011 by Taylor and Francis Group, LLC

 Massive vector field 377

and √ −1 √ −1 
ϑ′1
µ (k) = ( 2) (0, 1, i, 0), ϑ′2
µ (k) = ( 2) (0, 1, −i, 0),
(16.32)
ϑ′3
µ (k) = m
−1
(|k|, 0, 0, k0 ),
representing the four-dimensional generalization of the local reference points of (16.28) and
(16.29), respectively.
In the expression for the energy the term with µ = 0 proves to be negative and hence
the energy is not positively defined. However, the time components of the field functions
(±) (±)†
ψ0 (k) and ψ0 (k) are not the dynamically independent field components and should be
excluded with the use of the additional conditions:

k µ ψµ(±) (k) = 0, k µ ψµ(±)† (k) = 0. (16.33)

Note that in the case of the Lagrangian L′ the situation is somewhat burdened with the
presence of nonzero momenta canonically conjugate to ψ0 (x) and ψ0† (x), whereas for L
these momenta are identically equal to zero. With the help of the expressions resulting
from (16.33):
(±) 1 (±)† 1
ψ0 (k) = kn ψn(±) (k), ψ0 (k) = kn ψn(±) (k)
k0 k0
it is an easy matter to get

1
ψν(±)† (k)ψ (±)ν (k) = −ψ(±)† (k) · ψ (±) (k) + (k · ψ (±)† (k))(k · ψ (±) (k)). (16.34)
k02

Perform the expansion of the vectors ψ (±) (k) in terms of the local reference point (16.31):

k k0
ψ (+) (k) = ϑ1 (k)a1 (k) + ϑ2 (k)a2 (k) + a3 (k), (16.35)
|k|m

k k0 †
ψ (−) (k) = ϑ1 (k)b†1 (k) + ϑ2 (k)b†2 (k) + b (k). (16.36)
|k|m 3
Using (16.34)–(16.36), we get the following expressions for the four-dimensional momentum
and the electric charge of the vector field:
Z
P µ = d3 xk µ [(a† (k) · a(k)) + (b† (k) · b(k))], (16.37)

Z
Q = −e d3 x[(a† (k) · a(k)) − (b† (k) · b(k))]. (16.38)

Thus, in the new variables the energy proves to be positively defined.

However, in a coordinate system with the unit vectors ϑλ (k) the expression for the spin
vector projection onto the direction of a particle momentum is not diagonal:
Z
S3 = i d3 k[a†1 (k)a2 (k) − a†2 (k)a1 (k) − b†1 (k)b2 (k) + b†2 (k)b1 (k)]. (16.39)

We can improve the situation by going to the local reference point (16.32). Inserting the
expressions:
X X
ψ (+) (k) = ϑ′λ (k)aλ (k), ψ (−) (k) = ϑ′λ (k)b†λ (k)
λ λ

© 2011 by Taylor and Francis Group, LLC

 378 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

into Eqs. (16.25)–(16.27), we must assure ourselves that in the new local reference point
the quantities P µ , Q maintain their form and the spin vector projection is brought into the
diagonal form:
Z
S3 = d3 k[a†1 (k)a1 (k) − a†2 (k)a2 (k) − b†1 (k)b1 (k) + b†2 (k)b2 (k)]. (16.40)

Let us establish the physical meaning of the operators: a†n (k), an (k), b†n (k) bn (k).
The form of the dynamic variables points to the fact that we have obtained the correct
corpuscular structure of the vector field. Experience in quantization of the scalar field
suggests that this pattern was obtained on the basis of the commutation relations for the
operators. Natural generalization of the scalar field commutators in the case of a vector
field is provided by the relations:
[am (k), a†n (k′ )] = [bm (k), b†n (k′ )] = δmn δ(k′ − k). (16.41)
And in the next section their validity will be proved on the basis of the canonical formalism.
By application of (16.41) it is not difficult to fix:
[an (k), H] = k0 an (k), [a†n (k), H] = −k0 a†n (k), (16.42)
[bn (k), H] = k0 bn (k), [b†n (k), H] = −k0 b†n (k). (16.43)
If ΦE is the energy operator eigenstate with the eigenvalue E, then:
Han (k)ΦE = (E − k0 )an (k)ΦE , Ha†n (k)ΦE = (E + k0 )a†n (k)ΦE . (16.44)
and the same takes place for the states bn (k)ΦE and b†n (k)ΦE . Therefore, an (k) and bn (k)
have a meaning of the annihilation operators whereas a†n (k) and b†n (k) do a meaning of the
production operators.
Using Eq. (16.41), we find
[Q, a†n (k)] = −ea†n (k), [Q, b†n (k)] = eb†n (k), (16.45)
 †
 a1 (k), n = 1,
[S3 , a†n (k)] = −a†2 (k), n = 2, (16.46)

0, n = 3,
 †
 −b1 (k), n = 1,
[S3 , b†n (k)] = b†2 (k), n = 2, (16.47)

0, n = 3.
It should be emphasized that Eqs. (16.46) and (16.47) are valid only in case S3 to be
diagonal, that in turn is associated with the circular polarization of a vector particle. To
act the operators Q and S3 on the one-particle state Φakn = a†n (k)Φ0 leads to the equations:
QΦakn = Qa†n (k)Φ0 = [Q, a†n (k)]Φ0 , (16.48)
S3 Φakn = [S3 , a†n (k)]Φ0 . (16.49)
From Eqs. (16.44), (16.48), and (16.49) it becomes apparent that a†n (k)
represents the
production operator of the particle with the momentum k, the energy k0 , the charge −e,
and the spin projection onto the motion direction being equal to 1,-1,0 for n = 1, n = 2
and n = 3, respectively. The physical meaning of the operator b†n (k) follows from the
aforementioned under changing the charge sign and the spin projection sign. We accepted
the terminology, according to which particles possess the negative charge and antiparticles
possess the positive charge. This is nothing else but preservation of a historic tradition as
the first elementary particle discovered was the electron, the positron being the herald of
the antiworld.

© 2011 by Taylor and Francis Group, LLC

 Massive vector field 379

16.3 Commutation relations in the Proca theory

The Lagrangian L′ helps considerably in the simplification of deriving the dynamic invari-
ants of a vector field. But as the physical reality is approached (quantization, switching on
interaction and so on), it becomes less and less applicable. This makes us to leave it and
address the Lagrangians (16.10) and (16.15) instead. Both these Lagrangians give equiva-
lent physical results and hence a choice of one of them is simply a matter of our preference.
Further, we shall use the Lagrangian (16.15).
The quantization procedure with respect to the scalar field is somewhat complicated
due to the presence of an additional condition. The simultaneous canonical commutation
relations have the shape:

[ψ m (r′ , t), π n (r, t)] = [ψ m (r′ , t), (∂ n ψ 0† (r, t) − ∂ 0 ψ n† (r, t))] = iδmn δ(r′ − r), (16.50)

[ψ m† (r′ , t), π n† (r, t)] = [ψ m† (r′ , t), (∂ n ψ 0 (r, t) − ∂ 0 ψ n (r, t))] = iδmn δ(r′ − r), (16.51)
and all the remaining commutators containing: ψ m (r′ , t), π n (r, t), ψ m† (r′ , t) and π n† (r, t)
convert to zero. Similar to the case of the scalar field, from (16.50) and (16.51) we can derive
the commutation relations at an arbitrary instant of time, however, only for the space
components of the operator field functions. The derived relations are not relativistically
covariant, that naturally causes complications in the process of their manipulation. It is
necessary to direct our efforts to establishing the commutation relations including all four
components of the field function, formally considered at the same base. To that end, we
previously specify the commutators involving the field function component with µ = 0:
1 l ′ i
[ψ l (r′ , t), ψ0† (r, t)] = [ψ (r , t), ∂n π n (r, t)] = 2 ∂ l δ(r′ − r), (16.52)
m2 m

[ψ l (r′ , t), ψ0 (r, t)] = 0. (16.53)

With their use we can determine a set of the commutators involving the derivatives from
the field:  
l ′ n† ln ∂l∂n
[ψ (r , t), ∂0 ψ (r, t)] = −i g + δ(r′ − r), (16.54)
m2

[ψ l (r′ , t), ∂ 0 ψ0† (r, t)] = 0, (16.55)

l ′ n
[ψ (r , t), ∂ ψn† (r, t)] = 0, (16.56)
i n
[ψ 0 (r′ , t), ∂ 0 ψ0† (r, t)] = − ∂ ∂n δ(r′ − r), (16.57)
m2
where we have allowed for (16.9), (16.11) and

∂0 ψn† (r, t) = πn (r, t) + ∂n ψ0† (r, t). (16.58)

Substitution of the the operator field function expansions in terms of the solutions of the
free Proca equation (16.23), (16.24) into the left-hand side of the relations (16.52)–(16.57)
converts these relations into identities under fulfillment of the following commutation rela-
tions:
 
(+) (+)† ′ (−) (−)† ′ kµ kν
[ψµ (k), ψν (k )] = [ψµ (k), ψν (k )] = gµν − δ(k′ − k). (16.59)
m2

© 2011 by Taylor and Francis Group, LLC

 380 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Then, with allowance made for the fact that the polarization vectors obey the condition
(16.30), we find the commutation relations for the operators am (k) and bm (k):

[am (k), a†n (k′ )] = [bm (k), b†n (k′ )] = δmn δ(k′ − k). (16.60)

In its turn, Eqs. (16.60) allow us to establish the commutators for the operator field
functions at an arbitrary instant of time:
±
[ψ (±)ν (x), ψµ(±)† (y)] = iΛ(1)ν 2
µ (∂)∆0 (x − y; m ), (16.61)

or
[ψν (x), ψµ† (y)] = iDµν (x − y; m2 ) (16.62)
where  
2 ∂µ ∂ν
Dµν (x; m ) = Λµ(1)σ (∂)gσν ∆0 (x; m2 ) = gµν + ∆0 (x; m2 ).
m2
It is no trouble to verify that the commutation relations (16.61) are consistent both with the
(1)ν
field equations and the additional condition. Also it is obvious that the operator Λµ 9∂)
in the Proca theory plays the same role as does the operator d(∂) in the formalism of the
first-order equations. For example, with its help the causal Green function for the vector
field could be written
c
Dµν (x; m2 ) = Λ(1)σ
µ gνσ (∂)∆c (x). (16.63)
The above formalism may be easily generalized to the real vector field for m 6= 0 through
which the neutral carrier of the electroweak interaction, the Z-boson, is described. The
corresponding Lagrangian is given by the expression:

1 m2
L=−: Bµλ,σν ∂ µ ψ σ (x)∂λ ψν (x) + ψµ (x)ψ µ (x) : . (16.64)
2 2

16.4 Duffin-Kemmer equation for the vector field

The equations system (16.8) could be represented in the form of a matrix first-order equa-
tion. Let us introduce the ten-dimensional wavefunction:
 
ψ01 /m
 ψ02 /m 
 
 ψ03 /m 
 
   ψ23 /m 
 
ψνσ (x)/m  ψ /m 
Ψ(x) = {ΨA } = =  31 , (16.65)
−iψµ (x)  ψ12 /m 
 
 −iψ1 
 
 −iψ2 
 
−iψ3
−iψ0

where A = 1, 2, 3...10. Through the use of it Eqs. (16.8) are represented in the form:

(iβ µ ∂µ − m)Ψ(x) = 0, (16.66)

© 2011 by Taylor and Francis Group, LLC

 Massive vector field 381

where the ten-row matrices βµ have the following nonzero components:


(β0 )17 = (β0 )71 = (β0 )28 = (β0 )82 = (β0 )39 = (β0 )93 = 1, 


(β1 )110 = −(β1 )101 = (β1 )59 = −(β1 )95 = (β1 )86 = −(β1 )68 = 1,
(16.67)
(β2 )210 = −(β2 )102 = (β2 )94 = −(β2 )49 = (β2 )67 = −(β2 )76 = 1, 


(β3 )310 = −(β3 )103 = (β3 )48 = −(β3 )84 = (β3 )75 = −(β3 )57 = 1.

The equations (16.66) are called the Duffin-Kemmer equations for vector particles [9,38].
Since the resultant representation comprises a sum of the representations (1/2, 1/2), (0, 1),
and (1, 0), then:
nmax = 2.
This in turn is responsible for the fact that both the expression for the operator d(∂) and
commutation relations determining the β-matrices algebra are similar to those for the first-
order equations of the scalar field. In this manner all the formulae derived by us for the
scalar field in Section 14.11 may be used to describe the vector particle. Although, with the
exception of the formula determining the spurs of the β-matrices. We address the relation
(14.147) to have been obtained without using an explicit form of the matrices βµ , P , and
P . In just the same way as in the case of the spin 0, we introduce the projection operators
(10)
P (10) and P :
(10)
P (10) = (3 − β µ βµ )/2, P = (β µ βµ − 2)/3, (16.68)
(10)
where P (10) (P ) selects the tensor (vector) part of the field function Ψ(x). Then, by
application of (16.67) and (16.68), it is an easy matter to verify the validity of the relation:

{P (10) βµ βν P (10) }α α α α
τ = Bµν , τ = gµν δτ − gµτ δν . (16.69)

With the help of Eqs. (16.69) and (3.1.147) we find that the spur of an even number of
β-matrices is defined by the expression [39]:

Sp{βµ1 βµ2 .....βµn } = Bµ2 µ3 , α τ Bµ4 µ5 , τ ρ ...Bµn µ1 , γ α +

+Bµ1 µ2 , α τ Bµ3 µ4 , τ ρ ...Bµn−1 µn , γ α . (16.70)

Along similar lines it is shown that the spur of an odd number of β-matrices is equal to zero.
From the general formula (16.70) we obtain, without trouble, the following expressions for
the spurs from the product of two and four β-matrices:

Sp{βµ βν } = 3gµν , Sp{βµ βν βλ βσ } = 3(gµν gλσ + gµσ gνλ ). (16.71)

© 2011 by Taylor and Francis Group, LLC

 17
Electromagnetic field

The point is that as Bruin the third lay in his

den things in the forest went on as usual. It wouldn’t be right
to call this state of affairs welfare. But the governor’s task,
after all, is not to establish an ideal order, but to support
and safeguard, despite its deficiencies, the existing order
which has been sanctified by ages. Neither does it consist in
perpetrating any manner of atrocities, whether minor, major or
medium! Nay, he should be satisfied with the “natural” ones.
And if ancient custom ordains that the wolves should skin
rabbits and the hawks and owls pluck crows’ feathers, although
such order is not quite synonymous with welfare, it is,
nevertheless, an order of a kind, and as such must be obeyed.
M.E. Saltykov Shchedrin, “Bears in Government”

17.1 Maxwell equations

An electromagnetic field theory was formulated by J. Maxwell in 1865 in the form of a
system of differential equations generalizing all the fundamental laws of electromagnetic
phenomena. Maxwell equations in a vacuum in the presence of the electric current with the
density j and the electric charge with the density ρ are of the following form:

∂H ∂E 
rot E = − , rot H = + j,
∂t ∂t (17.1)
div E = ρ, div H = 0. 

The first equation is associated with the Faraday induction law: changing of a magnetic
field leads to the beginnings of an electric field. The second equation represents the Ampere
law:
rot H = j,
where the Maxwell displacement current is introduced to take into consideration that a
change of the electric field results in inducing the magnetic field. The third equation is an
equivalent of the Coulomb law. The fourth equation states that there are no sources of
the magnetic field other than currents, that is, magnetic charges (magnetic monopoles) in
nature are nonexistent.
At first sight on the Maxwell equations it becomes apparent there is one opportunity that
was either missed by nature or we are not able to detect that by our devices up till now.
In a free case j = ρ = 0 these equations are invariant with respect to the so-called dual
transformations:
E → H, H → −E, (17.2)

383
© 2011 by Taylor and Francis Group, LLC
 384 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

that is, there is the mutual symmetry of the electric and magnetic fields. Nevertheless,
the presence of the electric currents and charges disturbs the equivalence of the electric
and magnetic fields. Symmetry of the Maxwell equations may be restored only by the fact
of existing magnetic charges with the density ρm and currents jm connected with these
charges. We denote the magnetic charge density by ρm and the currents associated with
the magnetic charges by jm . In this case, the Maxwell equations would be of the form:

∂H ∂E 
rot E = − − jm , rot H = + j,
∂t ∂t (17.3)
div E = ρ, div H = ρ , 
m

whereas the dual transformations would be complemented by:

ρ → ρm , ρm → −ρ, j → jm , jm → −j. (17.4)

The hypothesis that a magnetic monopole is existent in nature was put forward by P. Dirac
in 1931 [40] in an effort to explain the electric charge quantization law. And since then such
a hypothetic magnetic charge has been called Dirac’s monopole. Using the Schrödinger
equation for description of the interaction between the electric Q and magnetic g charges,
Dirac has obtained the charge quantization condition:

Qg = 2πn~c, n = 0, ±1, ±2, ±3, .. (17.5)

where, for a while, we use the conventional units. In case the smallest magnetic charge g1 is
existent, from relation (17.5) it follows that the electric charge may take on only the values:

2π~c 4π~c 6π~c

Q=± ,± ,± , ....
g1 g1 g1

On the other hand, from the charge quantization condition it follows that the magnetic
charge is also quantized:
2π~c ne
gn = n = . (17.6)
e 2αem
As we see the minimal magnetic charge of the monopole should be equal to ≈ 68.5e.
It appears that the condition (17.5) may be derived directly from the Maxwell equations,
written in the form of Eq. (17.3), using the semiclassical and semiquantum Bohr theory [41].
Let an electric charge be slowly moving in the closed orbit and a magnetic charge be slowly
moving along the loop enclosing this orbit. Motion of the magnetic charge generates the
electric field, whose integral along the magnetic charge path is, in accordance with (17.3),
given by the expression: Z 
I
1 dΦ
Edl = − jm dS + , (17.7)
c dt
where Z
Φ= HdS

and S is an arbitrary surface restricted by the contour l. To find a change in the momentum
p of the electric charge due to the electric field, Eq. (17.7) should be multiplied by Q and
integrated with respect to time. As a result we have
I
e
∆pdl = − (∆g + ∆Φ) . (17.8)
c

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 385

Using the Bohr quantization rule:

I
∆pdl = 2πn~ (17.9)

and assuming that the magnetic charge comes back to the initial point (∆g = g and ∆Φ =
0), we finally obtain
Qg = 2πn~c.
The possibility to explain the charge quantization in a theory describing both the electric
and magnetic sources still remains one of the theoretical arguments to have us treat the
magnetic monopole hypothesis with the proper respect. We could list some other theoretical
arguments in its support. However, a decisive role as always belongs to the experiment.
But by the present time numerous efforts of experimental search for the Dirac monopole
have been in vain. But from this does not follow in any way that such objects are absent
in nature since so far a theoretical principle forbidding their existence has not been known.
We rewrite the Maxwell equations (17.1) in the explicit relativistically invariant form. Of
course, they are Lorentz-invariant in the three-dimensional form as well. This is so because
revealing the covariance of the Maxwell equations relative to the Lorentz transformations
by Einstein has led to the creation of the special theory of relativity. Our objective is to go
to the notation, where the covariance is expressed in the way that gives a clearer insight of
the problem. We introduce the real covariant four-dimensional vector of an electromagnetic
potential Aµ so that:
∂A
E = −grad A0 − 0 , H = rot A. (17.10)
∂x
The components of the four-dimensional rotor of the four-dimensional potential form the
electromagnetic-field antisymmetric tensor:

Fµν = ∂µ Aν − ∂ν Aµ , (17.11)

the components of which are related to the components of the electric and magnetic field
strength vectors as follows:
1
Hi = − εijk Fjk , Ei = F0i . (17.12)
2
In the free case the Maxwell equations are written by means of the tensor Fµν in the form:

∂ µ Fµν = 0, (17.13)

∂ µ Fνλ + ∂ ν Fλµ + ∂ λ Fµν = 0. (17.14)

Next, proceeding from Fµν to the potential Aµ , we see that four Eqs. (17.14) are the
consequence of the definition (17.11), not resulting in any equation for Aµ . Whereas four
Eqs. (17.13) lead to:
Aµ − ∂ ν ∂µ Aν = 0, (17.15)
that is an analog of the Proca equation for the massless neutral vector field. The four-
dimensional form of Eqs. (17.13) and (17.15) is a direct proof for the relativistic covariance
of the Maxwell equations, both in terms of the electromagnetic field tensor and potential.
The procedure of introducing the potential Aµ is ambiguous. In the Maxwell theory
the observables—the vectors E and H, together with the motion equations and all the
relations for the electromagnetic field—are invariant with respect to the second-kind gauge
transformation (or second-kind gradient transformation):

Aµ (x) → A′µ (x) = Aµ (x) + ∂µ χ(x), (17.16)

© 2011 by Taylor and Francis Group, LLC

 386 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where χ(x) is an arbitrary function having partial derivatives of the first and second order.
This ambiguity means that the dynamic description of an electromagnetic field in terms of
Aµ (x) involves the degrees of freedom, nonphysical in character.
Ambiguity in a choice of the potentials allows for imposition of some additional condition.
The only invariant condition, linear in Aµ (x), is the so-called Lorentz condition:

∂ µ Aµ = 0, (17.17)

with due regard for that Eq. (17.15) takes the form:

Aµ (x) = 0. (17.18)

The principal advantage of the condition (17.17) is the fact that both in the free case and
in the case with the included interaction the quantity ∂ µ Aµ satisfies the same equation, the
D’Alembert equation. Really, acting on the equation:

Aµ (x) = jµ (x) (17.19)

by the operator ∂ µ and taking into account the electric current conservation law:

∂ µ jµ (x) = 0,

we get
∂ µ Aµ (x) = 0. (17.20)
It is an easy matter to check that χ(x) may be always chosen so that the condition
(17.17) can be fulfilled. It is also clear that the relation (17.17) and the electromagnetic field
equations for the potentials (17.18) remain to be invariant with respect to the specialized
gradient transformation of the second kind:

Aµ (x) → A′µ (x) = Aµ (x) + ∂µ χ0 (x), (17.21)

where χ0 (x) is an arbitrary function satisfying the D’Alembert equation:

χ0 (x) = 0. (17.22)

In any partial Lorentz reference frame the function χ0 (x) may be chosen in such a way that
one of the components of Aµ (x), for example, the scalar potential, A0 (x), becomes equal to
zero. Then the Lorentz condition takes the shape:

div A(x) = 0. (17.23)

To establish the physical meaning of the relation (17.23) we take advantage of the expan-
sion of the four-dimensional potential in terms of the free equation solutions or, just the
same, in terms of the de Broglie plane waves:
Z
1 d3 k
Aµ (x) = √ [A+µ (k)e
−ikx
+ A−
µ (k)e
ikx
]. (17.24)
(2π)3/2 2k0
To insert (17.24) into (17.23) gives

(k · A± (k)) = 0, (17.25)

that represents the condition of the electromagnetic field transversality. Even though an
electromagnetic field is described by the four-dimensional potential, only two linearly-
independent components, orthogonal to the wave vector, have the physical meaning. Note

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 387

that, in spite of the noncovariance of the transversality condition (17.25), one could always
attain its fulfillment in any partial Lorentz system by means of the suitable choice of the
function χ0 (x) in the transformation (17.21).
Thus, a massless vector field could not be considered as the limiting case of a massive
vector field at m → 0 because this leads to erroneous physical results. This is seen even
from the manner of how mass is introduced into the particular formula for a massive vector
field (for example, in the expression for the commutator function Dµν (x; m2 )). Massless
fields significantly differ from massive ones. In the given case the difference is in the number
of polarization states that for photons is less by one compared to the massive neutral vector
particles. Such a decreased number of degrees of freedom is caused by zero mass of the
photon, as only in this case the Proca equations for a neutral vector field become gradiently
invariant.
The Maxwell equations (17.13) and the definition of the tensor Fµν may be obtained from
the Lagrange function density:
 
1 1
L = F µν Fµν − ∂µ Aν + ∂ν Aµ . (17.26)
2 2
The wave equations for the potential follows from the Lagrange function density:
1
L′ = − ∂ ν Aµ ∂ν Aµ . (17.27)
2
Now in order to derive the Maxwell equations we should add the Lorentz condition (17.17)
to the wave equations. It is interesting that this condition is associated only with a theory of
the classical electromagnetic field and, as will be shown later, upon quantization is replaced
by the particular conditions for the system state vector, which are equivalent to the Lorentz
condition exclusively from the viewpoint of the average values.

17.2 Dynamic invariants of the electromagnetic field

By application of the Lagrangian (17.26) the canonical energy-momentum tensor has the
form:
Tµ′ν = F νσ ∂µ Aσ − δµν L. (17.28)
Subtracting from (17.28) the insignificant term F νσ ∂σ Aµ to satisfy the relation:

∂σ (F νσ ∂σ Aµ ) = 0,

and going to the gradiently-invariant quantities in the Lagrangian, for the metric (sym-
metrized) energy-momentum tensor we obtain the following expression:
1
Tνµ = Fνσ Fµσ − gνµ F λσ Fλσ . (17.29)
4
From (17.29) it follows that the momentum and energy of the electromagnetic field are
determined by the expressions: Z
P = d3 x[E × H], (17.30)
Z
1
H= d3 x(E2 + H2 ). (17.31)
2

© 2011 by Taylor and Francis Group, LLC

 388 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

With the help of (17.29) we find the total angular momentum tensor for the electromagnetic
field: Z Z
Mµν = d3 x(xµ Tν0 − xν Tµ0 ) = d3 x(xµ sν − xν sµ ), (17.32)

where
1 2
s0 = (E + H2 ), s = [E × H].
2
The space components of Mµν form the field angular momentum vector to be given by:
Z
1
M= d3 x[r × [E × H]]. (17.33)
c
The spin moment tensor is defined by the expression:
σ
Sµν = Fµσ Aν − Fνσ Aµ .

Therefore, the spin vector has the form:

Z
S= d3 x[E × A]. (17.34)

Note that the division of the photon angular momentum into the orbital and spin parts
is formal. Only the total photon angular momentum has the physical meaning. This may
be explained as follows. First, the ordinary definition of the spin as a moment of a particle
at rest is inapplicable to a photon, since there is no rest reference frame for particle with
zero mass—in any reference frame this particle is moving with the speed of light. Second,
an expression for the spin proves to be gradiently-noninvariant, whereas the observables
for an electromagnetic field should possess this feature. In consequence, the states with
certain values of the orbital and spin moments fail to satisfy the transversality condition in
the general case. But formally, the representation of the total angular momentum for the
electromagnetic field as a sum of two terms appears to be very useful.
It should be noted that the photon presents another quantum-mechanical surprise. Since
the photons are relativistic objects, a minimal measuring error for their coordinates, ∆q,
is coincident with the de Broglie wavelength. This means that it makes sense to speak
of the photon coordinates only in the cases when the characteristic sizes of the problem
are greater than the wavelength. In the quantum case, however, the wavelength is not
considered small, so the notion of the photon coordinates becomes objectless. In other
words, the wavefunction of the photon is not determined by the field value at a particular
point, being rather dependent on the field distribution in the region with a size on the
order of the wavelength. This suggests that localization of the photon in a smaller region
is impossible and hence the notion of the probability density for the photon localization at
a particular space point is absurd. Mathematically, this situation is reflected by the fact
that in the coordinate representation for the photon there is no way to set up a quantity
that may be interpreted as a probability density. Such a quantity should be expressed by a
positive bilinear combination of the photon wavefunctions, representing the time component
of the current four-dimensional vector Jµ , whose continuity equation:

∂ µ Jµ = 0

just provides the particle number conservation law. It is impossible to compose a bilinear
combination forming the four-dimensional vector with zero divergence from the vectors of
the electromagnetic field. Such a vector could not be constructed with the use of the four-
dimensional potential Aµ too. Indeed, Aµ may be introduced into the current Jν only as

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 389

a component of the Fµν -tensor due to the gauge invariance. This means that Jν should
be bilinearly formed from Fµν , Fτν and four-dimensional vector k τ . However, such a vector
will deliberately be equal to zero because of the transversality condition:

k τ Fτν = 0.

Recalling what stands in the gauge invariance, we come to the conclusion that the above
complication is associated with the zero mass of the photon.
The Lagrangians L and L′ are distinguished by the quantity that, upon integration with
respect to the four-dimensional space and with due regard for the Lorentz condition, be-
comes zero. In a similar way the dynamic characteristics of the electromagnetic field prove
to be the same for both Lagrangians with allowance made for the equations of motion and
condition (17.17). A reader has the opportunity to find whether the expressions for the
energy-momentum metric tensor:
1
Tνµ = −∂ν Aσ ∂µ Aσ + gνµ ∂ λ Aσ ∂λ Aσ (17.35)
2
and the spin momentum tensor:
σ
Sµν = Aν ∂ σ Aµ − Aµ ∂ σ Aν , (17.36)

following from the Lagrangian L′ are equivalent to the relations (17.29) and (17.33).
In the momentum space we introduce the local reference point eλµ (k) to be connected
with the unit polarization vectors of the photon. Further, in Eq. (17.24) we carry out the
(±)
expansion of Aµ (k) in terms of the vectors eλµ (k):
Z3
1 d3 k X λ
Aµ (x) = A(+)
µ (x) + A(−)
µ (x) = √ [e (k)a+
λ (k)e
−ikx
+ eλ∗ −
µ (k)aλ (k)e
ikx
],
(2π)3/2 2k0 λ=0 µ
(17.37)
where a±λ (k) have a sense of the de Broglie waves amplitudes in the nonquantized theory.
Thus, at the given momentum value there are four linearly independent solutions to differ
with polarizations. In the most simple case one may chose the vectors eλµ (k) in such a way
that they coincide with the unit vectors of the coordinate system nµ , that is,

eλµ (k) = δµλ . (17.38)

Then, if one directs n3 along k, λ = 1, 2 will correspond to the transverse polarization,

λ = 3 to the longitudinal one, and λ = 0 to the scalar (or time) one. In this coordinate
system the three-dimensional vectors e1 (k), e2 (k), and k/ | k |, for which the relation:

(em (k) · en (k)) = δmn , [em (k) × en (k)] = εmnk ek ,

is valid, form the orthonormalized system with the completeness condition:

kr kl
e1r (k)e1l (k) + e2r (k)e2l (k) + = δrl , (17.39)
| k |2

where the indices r, l = 1, 2, 3 denote the vector components.

For a free massless particle the only direction is preferential in space, the direction of
the momentum vector. It is obvious that in this case symmetry of the wavefunction of a
particle with respect to the whole group of three-dimensional rotations (spherical symmetry)
is lacking, and we can consider only the axial symmetry about the preferred axis. In other

© 2011 by Taylor and Francis Group, LLC

 390 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

words, for a given k the photon may have only two states with different values of the spin
vector projection described by the polarization vectors e1µ and e2µ . As such we can choose:

e1µ = (0, 1, 0, 0), e2µ = (0, 0, 1, 0), (17.40)

that is in accord with a linear polarization about the axis 1 or 2. Another choice is:

1 1
e1µ = √ (0, 1, −i, 0), e2µ = √ (0, 1, i, 0), (17.41)
2 2

that is associated with a circular polarization of the electromagnetic field.

As we have (17.38), the total sum (and not a scalar product!) over the polarizations
equals:
X 3
eλµ eλ∗
ν = δµν . (17.42)
λ=0

At the same time, the summation should be performed only over the physical states, that
is, over the transverse polarizations. This may be done using Eq. (17.39) to give the
noncovariant formula:
2
X km kn
eλm eλ∗
n = δmn − . (17.43)
| k |2
λ=1

In practice it is more convenient to use a relativistically covariant expression for the

summation over the photon spin states that includes all four photon polarizations possible.
Using the explicit form of the vectors eλµ (k), we arrive at

3
X
eλµ (k)eλ∗
ν (k) = −gµν . (17.44)
λ=0

Now we should display that the usage of (17.44) also leads to the inclusion of only the phys-
ical photon states. Substituting the expansion (17.37) into the Lorentz condition (17.17),
we get two relations:
| k | a± ±
3 (k) − k0 a0 (k) = 0, (17.45)
to be valid for any choice of the transverse polarization vectors. Since kµ k µ = k02 − k2 = 0,
we have
k0 =| k | .
Then, from Eqs. (17.45) it follows

a+ − + −
3 (k)a3 (k) − a0 (k)a0 (k) = 0. (17.46)

The obtained result indicates that, due to the Lorentz condition, the densities of the average
numbers of longitudinal a+ − + −
3 (k)a3 (k) and temporal photons a0 (k)a0 (k) are equal, their
contributions into the dynamic invariants being opposite in sign. Thus, longitudinal and
temporal photons, hereinafter referred to as pseudophotons, are as if compensating each
other. Thus, for example, using Eq. (17.44), we can show that the quadratic form involved
in the definitions of the majority of dynamic variables includes no pseudophoton states:
3
X 2
X
′
−Aµ (k)Aµ (k) = − g λλ a+ −
λ (k)aλ′ (k) = a+ −
λ (k)aλ (k). (17.47)
λ,λ′ =0 λ=1

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 391

Let us define the dynamical variables in the case of the circular polarization of the photons.
The energy-momentum four-dimensional vector is given by the expression:
Z X Z
P µ = − d3 k(k µ A+ ν (k)A−ν
(k)) = d3 k(k µ a+ −
λ (k)aλ (k)). (17.48)
λ=1,2

Similar to the case of a massive vector field, the energy turns to be positively defined due
to the Lorentz condition only.
For the spin vector projection onto the momentum direction we find
Z
S3 = d3 k[a+ − + −
1 (k)a1 (k) − a2 (k)a2 (k)]. (17.49)

From the expressions (17.48) and (17.49) it follows that the quantities a+ −
λ (k)aλ (k) represent
the average number of the photons with the momentum k, the energy k0 , possessing the
spin projection onto the motion direction that is equal to +1 (λ = 1) and -1 (λ = 2). Note,
that in the case of using the vectors of the linear photon polarization, all the dynamical
invariants except the spin projection operator are reduced to the diagonal form as well.
Since the potential is a real quantity, then the relations:

(a± ∗ ∓
µ (k)) = aµ (k).

take place. This allows us to take the designations:

a+
µ (k)) ≡ aµ (k), a− †
µ (k) ≡ aµ (k)

17.3 Electromagnetic field quantization

In the case in question quantization is performed so that the requirements of the relativis-
tic covariance, energy positiveness, Lorentz and transversality conditions can be satisfied
simultaneously. When proceeding from the Lagrangian density (17.26), the canonical quan-
tization procedure proves to be unsuitable. Actually, as

πµ (x) = Fµ0 (x),

the canonical momentum π0 (x) is identically going to zero and the canonical quantization
procedure is inapplicable to the time component of the field function. This is radically
different from the situation in the case of a massive vector field, where the Lorentz condition
was the operation condition included into the Lagrangian and when the component ψ0 (x)
could be directly eliminated from the theory. For the situation to be remedied we use the
Lagrangian (17.27) and shall take the vector potential components as being the independent
quantities.∗ Now the canonical momentum is given by:

π µ (x) = −∂ 0 Aµ (x), (17.50)

what allows us to write the simultaneous commutation relations for all the field components

[πµ (r, t), Aν (r′ , t)] = igµν δ (3) (r − r′ ) (17.51)

∗ It means that we discard the Lorentz condition in the operator form.

© 2011 by Taylor and Francis Group, LLC

 392 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

(all the remaining commutators are equal to zero). For the four-dimensional potential
Aµ (x), which we consider the Hermitian operator, an expansion in terms of plane waves is
of the form:
Z 3
1 d3 k X λ †
Aµ (x) = √ [e (k)aλ (k)e−ikx + eλ∗
µ (k)aλ (k)e
ikx
], (17.52)
(2π)3/2 2k0 λ=0 µ

where aλ (k) and a†λ (k) are operator quantities. For simplicity, we take eλµ in the form of δµλ
to demonstrate that the canonical quantization rules may be fulfilled when the operators
aλ (k) and a†λ (k) are subjected to the commutation relations:

[aµ (k), a†ν (k′ )] = gµν δ (3) (k − k′ ). (17.53)

From (17.53), in its turn, one may find both the commutators for the potentials in an
arbitrary instant of time:

[Aµ (x), Aν (x′ )] = igµν D0 (x − x′ ), (17.54)

where
D0 (x) ≡ ∆0 (x; m2 = 0)
and the convolution for the potentials:

Asµ1 (x)Asν1 (x′ ) = igµν D0c (x − x′ ), (17.55)

where D0c (x) = ∆c (x) at m2 = 0.

Now it is high time to recall the gauge ambiguity of the potentials Aµ . Can it have
an influence on the form of the operators introduced in the process of quantization of the
electromagnetic field? And if so, what is this influence? We are interested in particular
in the form of the photon causal function. To find the answer, we replace the Lagrangian
(17.26) by the expression:

1 1
L = − Fµν F µν − ∂µ Aµ ∂ν Aν . (17.56)
4 2ξ

The last term in (17.56) violates the gauge invariance. The quantity ξ, which may be chosen
arbitrary, is called a gauge parameter. At ξ = 1 the Lagrangian (17.56) differs from that
given by the expression (17.27) on insignificant four-dimensional divergence. Variation of
the potentials leads to the following motion equations:
 
ξ−1
dσν (∂)Aσ = [∂µ ∂ µ gσν − ∂ν ∂σ ]Aσ = 0. (17.57)
ξ

In the case under investigation the photon propagator satisfies the equation:
 
µ ξ−1
[∂µ ∂ gσν − ∂ν ∂σ ]Dρσc (x − x′ ) = gνρ δ(x − x′ ). (17.58)
ξ

It immediately follows that the propagator in the momentum representation is defined by

the expression:
 
σc ξ − 1 kν kσ σc gνρ
dσν (k)Dρ (k) = [gσν − 2
]Dρ (k) = − 2 . (17.59)
ξ k k

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 393

Since the matrix operator dσν (k) has nonzero eigenvalues on the vector kν :
 
ξ − 1 kν kσ ν 1
[gσν − ]k = k ν , (17.60)
ξ k2 ξ

the matrix dσν (k) possesses a reverse matrix and, as a result, Eq. (17.58) has a single-valued
c
solution. Carrying out the division of Dµν (k) on the transverse and longitudinal parts:

c
Dµν (k) = D(tr) (k 2 )(gµν − kµ kν /k 2 ) + D(l) (k 2 )kµ kν /k 2 (17.61)

and substituting (17.61) into (17.59), we get

       
ξ − 1 kν kσ ξ − 1 kν kσ
k 2 g σν − D µc
σ = k 2
g σν
− [(δσµ −
ξ k2 ξ k2
   
k µ kσ (tr) 2 k µ kσ (l) 2 µν kµ kν 1 k µ k ν (l) 2
− 2 D (k ) + 2 D (k ) = g − 2 D(tr) (k 2 ) + D (k ).
k k k ξ k2
(17.62)
Comparing the obtained expression with Eq. (17.59), we arrive at the expressions for the
transverse and longitudinal parts of the photon propagator:
1 ξ
D(tr) (k 2 ) = − , D(l) (k 2 ) = − ,
k2 k2
that finally gives  
c ξ − 1 kµ kν 1
Dµν (k) = − gµν + . (17.63)
ξ k2 k2
In the limit ξ → ∞ the gauge-fixed term ξ −1 (∂µ Aµ )2 disappears. This suggests the
following remedy to “expulsion” ξ from the final results. In the intermediate calculations
we have to keep the finite value of ξ, while in the final stage of calculations we should turn
ξ to infinity. But there is no need for this procedure as, due to the gauge invariance of
the theory, none of the physical results is dependent on a choice of ξ. Really, the photon
propagation function is introduced into the physical quantities, scattering amplitudes, being
multiplied by the transitional currents of two electrons, that is, in the combinations of
µ c ν µ
the form: j12 Dµν j34 , where jik = eψ i (x)γ µ ψk (x). But, on the strength of the current
conservation law:
µ
∂µ jik = 0,
we have
kµ j µ ik = 0,
where k = pi − pk . So, to introduce the gauge parameter into the theory is dictated by the
convenience requirements only. However, as we shall see in the second volume, terms fixing
the gauge are an indispensable tool under quantization of non-Abelian fields.
The limiting case ξ → ∞ corresponds to the Landau gauge:
 
c kµ kν 1
Dµν (k) = − gµν − 2 . (17.64)
k k2

It should be noted that such a choice of the parameter ξ is analogous to the Lorentz gauge
of the potentials. Further, unless otherwise specified, we shall take ξ = 1 (Feynman gauge):
c gµν
Dµν (k) = − , (17.65)
k2

© 2011 by Taylor and Francis Group, LLC

 394 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

what is associated with (17.55).

With the help of the four-dimensional vector of the energy-momentum expressed in terms
of the longitudinal, transverse, and time components aν (k)
Z
P = − d3 k[k µ aν (k)aν† (k)],
µ
(17.66)

we can demonstrate that the operators am (k) and a†m (k) (m = 1, 2, 3) represent the annihi-
lation and production operators for the transverse and longitudinal photons. For the time
components we have the commutation relations with the wrong sign:
[a0 (k), a†0 (k′ )] = δ (3) (k − k′ ). (17.67)
When going from µ = 1, 2, 3 to µ = 0 the operators a†µ (k) and aµ (k) as if change their
roles: a0 (k) behaves itself as a†m (k) rather than as am (k), whereas a†0 (k) is similar to
am (k). In other words, the quantization procedure is disturbed. It seems that the situation
may be improved by the assumption that a0 (k) are the production operators and a†0 (k)—
the annihilation operators of temporal pseudophotons. But this situation proves to be in
conflict with the fact of reality of the electromagnetic field. To show that this is the case,
we take the vacuum average for expression (17.67):

Φ†0 [a0 (k), a†0 (k′ )]Φ0 = −Φ†0 a†0 (k)a0 (k′ )Φ0 = δ (3) (k − k′ ). (17.68)
Multiplying the left-hand side of (17.68) by F † (k′ )F (k) (F (k) is an arbitrary analytical
function) and integrating over k and k′ , we arrive at the result:
Z Z Z
− d3 k ′ Φ†0 a†0 (k′ )F † (k′ ) d3 kF (k)a(k)Φ0 = −Φ†0 | d3 kF (k)a(k) |2 Φ0 < 0. (17.69)

Carrying out the same mathematical operations with the right-hand side of (17.68), we get
Z Z Z
d3 k ′ F † (k′ ) d3 kF (k)δ (3) (k − k′ ) = d3 k | F (k) |2 > 0,

that contradicts the relation (17.69).

It is also easy to see that the norm of the field state vector with an odd number of scalar
photons is negative. For example:
(Φ1k0 , Φ1k0 ) = (a†0 (k)Φ0 , a†0 (k)Φ0 ) = (Φ0 , a0 (k)a†0 (k)Φ0 ) = −1.

17.4 Quantization with use of an indefinite metric

To restore a sequence of the quantization procedure—all aµ (k) and a†µ (k) have the meaning
of the annihilation and production operators—one could use the procedure proposed by
Gupta [42] and Bleuler [43]. The content of this procedure is as follows.
To provide a physically interpreted theory, the Hilbert space H should include the sub-
space H+ with a positively defined metric, whose vectors describe the physical states of a
system. An operator in the complete space H corresponds to the physical observable if it
acts as a Hermitian operator on the physical states of H+ . In the Hilbert space we introduce
a linear and unitary metric operator η with the properties:
[η, am (k)] = {η, a0 (k)} = 0, ηΦ0 = 0, η2 = 1 (17.70)

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 395

Then, having determined the metric conjugation operation:

L′ = ηLη = qL, q = ±1,

one may speak of that the operators connected with the vector field part have the positive
metric parity and those corresponding the scalar field part possess the negative metric
parity. The fulfillment of (17.70) may be realized if we select η in the following form:

η = (−1)n0 ,

where n0 is a number of the scalar pseudophotons in the given state.

Conjugation of the operators by Gupta is understood as a set of the Hermitian and metric
conjugations:
L⋆ = ηL† η = qL† . (17.71)
Next we assume that, unlike other components, the scalar field component A0 (x) is anti-
Hermitian, that is,
A†µ (x) = −Aµ (x),
and this is equivalent to going to the Hermitian component A4 (x) = −iA0 (x). To realize
self-conjugation (by Gupta) of the field operators, we have to redefine the scalar product of
the state vectors as follows:
Z Z
(Φ, Φ′ ) = Φ† ηΦ′ dτ = Φ⋆ Φ′ dτ. (17.72)

At η = 1 we obtain the ordinary definition of the scalar product. In the general case
(17.72) may result in a positive as well as negative norm of the vector Φ, not excepting
the situation when Φ 6= 0 but (Φ, Φ) equals zero. Because of this, such spaces are thought
of as possessing an indefinite metric. The definition (17.71) is actually providing the self-
conjunction of operators as we have
Z Z
[ Φ⋆ Aµ (x)Φdτ ]† = Φ⋆ Aµ (x)Φdτ.

The above-mentioned procedure brings about the commutation relations:

[aµ (k), a⋆ν (k′ )] = −δµν δ(k − k′ ), (17.73)

[Aµ (x), A⋆ν (x′ )] = −iδµν D0 (x − x′ ). (17.74)

Let us choose, for the sake of simplicity:

e(λ) λ
µ (k) = δµ ,

that is, let the polarization index coincide with the Lorentz one. Then, the relations (17.73)
allow us to treat aµ (k) (a⋆µ (k)) as being the annihilation (production) operators of the
photons with the transverse, longitudinal, and time polarizations. Thus, the relativistic
quantization scheme is ensured.
The indefinite metric owes its origin to the problems arising during quantization of the
electromagnetic field. Subsequently, its sphere was widened to theories of a massive charged
vector field [44–46] due to the attempts to construct a renormalizable theory of the elec-
tromagnetic interaction for this field. ∗ The idea lay in the fact that in order to describe

∗A theory based on the Proca equations is not renormalizable and becomes such only upon the involvement
of the weak interaction and Higgs mechanism.

© 2011 by Taylor and Francis Group, LLC

 396 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the particles with the spin 1, one should use a collection of fields: vector field+auxiliary
field. Most often a role of the auxiliary field was played by a scalar L field. Since the norms
of the vector and scalar field states had different signs (H = H+ H− ), at calculation of
the scattering cross sections the subtractive procedure was working: the contributions of
scalar particles were partly canceling those of the longitudinal vector particles. As a result,
a theory became renormalizable. But the existence of the negative-metric subspaces was al-
ways a source of serious trouble. The reasons for that are obvious. The transitions between
the states with the different metric parity may lead to negative probabilities, that is, to the
unitarity violation of the theory. True enough, if the total number of the negative-metric
particles in the initial and final states is even, the probability appears to be positive for any
number of the positive-metric particles. There is a simple approach to verify the unitarity
of a theory with the indefinite metric. With this aim in view one should calculate the cross
section σH± →H∓ for processes of the type:

A+ + B+ → C− + D+ ,

where a subscript on a particle points to the fact that a particle belongs to the positive- or
negative-metric space. If the theory is unitary, then the quantities σH± →H∓ must vanish.
Unfortunately, all efforts to retain the unitarity in theories of a massive charged vector
field with the auxiliary scalar field were unsuccessful. It seems that the situation with the
indefinite-metric theories may be rescued by the presence of a symmetry that takes place
both in the free case and upon inclusion of the interaction. Indeed, according to the Noether
theorem, in this case a conserved quantity should be in existence. And this motion integral
may forbid the transitions between the subspaces with differing metrics. However, this hope
proved to be in vain because in the presence of the indefinite metric the symmetry does
not lead to a motion integral [47]. It may be that a golden time of the indefinite metric is
in the near future, but now it is used only as some skillful device providing the relativistic
covariance of the electromagnetic field quantization.
In the case of the electromagnetic field we have another problem associated with the
Lorentz condition. As for quantization we assume that all the operators Aµ (x) are inde-
pendent, we have no right to require the fulfillment of the Lorentz condition directly for
Aµ (x). It is an easy matter to see that the way of imposing the Lorentz condition on the
allowed state vectors:  
∂
Aµ (x) Φ = 0, (17.75)
∂xµ
appears to be blocked as well. Really, assuming Φ = Φ0 , we have
   
∂ ∂ −
Aµ (x) Φ0 = A (x) Φ0 = 0. (17.76)
∂xµ ∂xµ µ

Multiplying the left-hand side of (17.76) by A−⋆ ′

ν (x ) we get
 
∂ − ∂

A−⋆
ν (x′
) Aµ (x) Φ0 = A−⋆ ′ −
ν (x )Aµ (x) Φ0 =
∂xµ ∂xµ

∂
∂ ∂
= A− −⋆ ′
µ (x)Aν (x ) Φ0 + iδµν D− (x − x′ )Φ0 = i D− (x − x′ )Φ0 6= 0,
∂xµ ∂xµ 0 ∂xν 0
which is in contradiction with the statement (17.76).
For the agreement with the classical field, it is necessary to fulfill the Lorentz condition
whether in the operator form or on the average. It appears that the latter may be realized

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 397

by changing the Lorentz condition; it should concern only the part of the operator Aµ (x)
to annihilate photons, that is,
 
∂ +
A (x) Φ = 0, for all x. (17.77)
∂xµ µ

Since, both in the free case and at including the interaction the operator ∂ µ Aµ (x) obeys
the equation:
∂ µ Aµ (x) = 0,
then the division of this operator on the positive- and negative-frequency parts ∂ µ A±
µ (x) is
relativistically covariant. Carrying out the Gupta conjugation of the condition (17.77), we
arrive at  
∂ −
Φ⋆ Aµ (x) = 0, for all x. (17.78)
∂xµ
The relations (17.77) and (17.78) ensure the fulfillment of the Lorentz condition on the
average:  
⋆ ∂
Φ Aµ (x) Φ = 0. (17.79)
∂xµ
The next problem is the energy positiveness. From the auxiliary conditions, written in
the momentum representation:

(| k | a3 (k) − k0 a0 (k))Φ = 0, (17.80)

Φ⋆ (| k | a⋆3 (k) − k0 a⋆0 (k)) = 0, (17.81)

follows that:

Φ⋆ [a⋆3 (k)a3 (k) − a⋆0 (k)a0 (k)]Φ = Φ⋆ a⋆3 (k)[a3 (k) − a0 (k)]Φ = 0 (17.82)

or
Φ⋆ [N3 (k) − N0 (k)]Φ = 0, (17.83)
where N3 (k) and N0 (k) are number operators number longitudinal and temporal pseu-
dophotons, respectively. As is seen, the Lorentz condition in a reference frame with k k e3 (k)
permits only the states of a system that involves equal numbers of the longitudinal and tem-
poral pseudophotons having the same momentum. The temporal pseudophotons make a
negative contribution into the dynamic invariants that is canceled by the contribution from
the longitudinal pseudophotons. Thus, the total energy of the field:
Z Z 2
X
< H >= d3 kk0 < −a⋆µ (k)aµ (k) >= d3 kk0 < a⋆m (k)am (k) > (17.84)
m=1

proves to be positive defined. Note that the concept of the electromagnetic field vacuum
should be updated because of the relation (17.83). The vacuum of the electromagnetic
field represents such a state of the field, where the real transverse photons are lacking. As
regards the longitudinal and scalar pseudophotons, their numbers N3 (k) and N0 (k) are
always nonzero. But, similar to any other physically realized state, the state vector for the
vacuum is proportional to the value N3 (k) − N0 (k).
A final accord in history with the indefinite metric will be the proof of the fact that in
concrete calculations we can do without its introduction. We show that the average values
over the observables are the same both for the ordinary norm definition and for definition

© 2011 by Taylor and Francis Group, LLC

 398 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

of the norm with the indefinite metric. To this end, we rewrite an expansion for a positive
frequency part of the potential in terms of the local reference point eλµ (k) = δµλ in the form:

2  
X kµ
A+
µ (k) = em
µ (k)am (k) + − δµ a3 (k) + δµ0 a0 (k) = Atr,+
0
µ (k)+
m=1
| (k) |

+kµ Λ(k) + δµ0 B + (k), (17.85)

where Atr,+
µ (k) is a transverse component of the potential and B + (k) obeys the condition:

B + (k)Φ = 0.

Due to the gradient invariance the potentials Aµ (x) and A′µ (x) = Aµ (x) + ∂µ f (x) (f (x)
(f (x) is an arbitrary function) lead to the equivalent physical results. Because of this,
in the expressions for the potential we can neglect the quantities with the structure of
the four-dimensional gradient or, when an analysis is performed for the momentum rep-
resentation, the quantities proportional to kµ . It should be noted that this procedure is
also valid when the interaction is included as the electromagnetic interaction Lagrangian
Lem = ieAµ (x)j µ (x), by virtue of the current conservation law:

∂µ j µ = 0,

retains its form when going from Aµ (x) to Aµ (x)′ . Thus, we obtain

A+ tr,+
µ (k)Φ = Aµ (k)Φ, Φ⋆ A− ⋆ tr,−
µ (k) = Φ Aµ (k), (17.86)

which is equivalent to the statement:

Φ⋆ Aµ (k)Φ = Φ⋆ Atr
µ (k)Φ.

Let us show the validity of the more general statement:

Φ⋆ P(A)Φ = Φ⋆ P(Atr )Φ, (17.87)

where P(A) (P(Atr )) is an operator being polynomial on A (Atr ). Substitute the expression
for Aµ (k) (17.85) into the relation (17.87). To use the Lorentz conditions (17.77), (17.78)
there is a need to perform the commutation of the quantities Atr and B = B + + B − . Since:

[Atr,± , B ± ] = [Atr,∓ , B ± ] = 0,

then, as a result, only the quantities Atr,− and Atr,+ remain under the average symbol in
the left-hand side of (17.87), as was to be shown.
Now, we proceed to the main stage of our proof, that is, we are going to show the validity
of of the relation:
Φ⋆ P(Atr )Φ = Φ†tr P(Atr )Φtr , (17.88)
where Φtr is a vector of the pure photon state (pseudophotons are absent). A vector of
an arbitrary state could be represented as a superposition of the Φtr -state and Φps -states
containing different numbers of pseudophotons. Due to the condition (17.83) Φps -states
may incorporate pseudophoton operators only in the combination B − , that is,
X Y
Φ = {1 + Cn [a⋆3 (ki ) − a⋆0 (ki )]}Φtr , (17.89)
n 1≤i≤n

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 399
X Y
Φ⋆ = Φ⋆tr {1 + Cn [a3 (ki ) − a0 (ki )]}. (17.90)
n 1≤i≤n

To insert (17.89) and (17.90) into the left-hand side of the expression (17.88) produces the
result: X Y
Φ⋆ P(Atr )Φ = Φ⋆tr {1 + Cn [a3 (ki ) − a0 (ki )]}P(Atr ){1+
n 1≤i≤n
X Y
+ Cm [a⋆3 (kj ) − a⋆0 (kj )]}Φtr = Φ⋆tr {1+
m 1≤j≤m
X Y X Y
+ Cm [a⋆3 (kj ) − a⋆0 (kj )]}P(Atr ){1 + Cn [a3 (ki ) − a0 (ki )]}Φtr =
m 1≤j≤m n 1≤i≤n

= Φ⋆tr P(Atr )Φtr = Φ†tr ηP(Atr )Φtr = Φ†tr P(Atr )ηΦtr =

= Φ†tr P(Atr )Φtr , (17.91)
which confirms the validity of (17.88). Because of this, for the subsequent descriptions of
the electromagnetic field we can use the conventional formulae for the Hilbert space with
the positive (definite) metric. Using (17.91) it is readily shown that the norm of physically
allowed states proves to be positively definite. However, note that this is not a consequence
of the introduction of the indefinite metric. The true reasons are: the Lorentz condition and
the invariance of the electromagnetic field theory with respect to the specialized gradient
transformation of the second kind.
In conclusion, we have to indicate a possibility to reformulate the Maxwell equations into
the matrix Duffin-Kemmer equations in the same manner as it was done for the theory of
the massive vector field. The only difference is in matching of the dimensionalities for the
components of the vector Aµ and the tensor Fµν that is realized due to the introduction of
an arbitrary constant with the dimensionality of the inverse length.

17.5 Photon polarization

We start with the three-dimensional description of the nonquantized electromagnetic field.
At a given momentum the photon state is characterized by the unit polarization vector eλ
that plays the role of the spin part of the photon wavefunction. An arbitrary polarization
eλ may be represented as the imposing of two mutually orthogonal polarizations e(1) and
e(2) chosen by some particular way:

eλ = aλ1 e(1) + aλ2 e(2) , (17.92)

where in the most general case aλ1 and aλ2 are complex quantities satisfying the normalization
condition:
|aλ1 |2 + |aλ2 |2 = 1.
Since the common factor standing ahead of eλ is arbitrary, then a vector of any polarization
could be written in the form:

eλ = e(1) cos α + e(2) sin αeiβ , (17.93)

where α and β are real numbers. Let e(i) = δri , where i = 1, 2 and r = 1, 2, 3, then β = 0
would correspond to the linear polarization at an angle α to the x1 -axis. At β = ±π/2 and

© 2011 by Taylor and Francis Group, LLC

 400 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

α = π/4 we have the circular polarizations while the elliptic polarization correlates with
arbitrary values of β and α.
Note that in the case of a three-dimensional transverse gauge and with the proviso that
the z-axis has been chosen along k, the vectors e(1) and e(2) involved in (17.92) have zero
projection onto the z-axis, that is, they are two-dimensional objects. Proceeding from all
the above, we can establish a formal analogy between the photons and massive particles with
the spin 1/2. Both of them have two polarization states. True enough, in the case of the
electron we consider the spin projection onto the momentum, whereas for the photons—
spin projection onto the axis perpendicular to the momentum direction. Similar to the
electrons, these photon states may be represented with the use of two-dimensional column
vectors under the condition that the k-direction (the z-axis) still remains the quantization
axis.
The state of a photon with a definite momentum and polarization is a pure state; it
is described by the wavefunction and associated with the complete quantum-mechanical
description of the particle state. At the same time, there are some situations when we cannot
assign a certain wavefunction to the photon. For example, at the photon scattering by the
electron there exists only the wavefunction of a system electron+photon, whose expansion
in terms of the wavefunctions of the free photon already includes the wavefunctions of the
electron. In other words, there is a possibility for the mixed photon states corresponding
to a less complete description using the density matrix rather than the wavefunction.
Let us consider a case when the photon momentum is given, while its polarization state
is indeterminate because the factors aλ1 and aλ2 , or the quantities α and β, depend on the
parameters characterizing another system. Such a state of the photon is called the partial
polarization state that may be described by the density matrix. The polarization density
matrix of the photon represents the second-rank tensor ργmn (m, n = 1, 2) specified in the
plane perpendicular to the k-vector:
X
ργmn = g λ aλm aλ∗ λ λ∗
n = am an , (17.94)
λ

λ
where g is a relative probability with which the λ- polarization enters into the mixed
beam, and an overline denotes averaging over the relevant parameters characterizing a total
system. From the density matrix definition it follows that this matrix is Hermitian:

ργmn = ργ∗
nm , (17.95)

and is normalized by the condition:

Sp ργ = ργ11 + ργ22 = 1. (17.96)

Thanks to (17.95) the diagonal components ργ11

and ργ22 are real, whereas the nondiagonal
ones are complex and they satisfy the condition:

ργ12 = ργ∗
21 .

Thus, the density matrix is defined by three real parameters.

Let us find out the physical meaning of elements of the photon density matrix. When one
chooses e(1) k Ox and e(2) k Oy, then from (17.94) it follows that ργ11 = aλ1 aλ∗
1 characterizes
the probability of the photon polarization along the x-axis, while ργ22 —along the y-axis. For
the case of the circular polarization the vectors e(1)′ and e(2)′ , which are connected with
the vectors e(1) and e(2) by the relations:
1   1  
e(1) = √ e(1)′ + e(2)′ , e(2) = √ e(1)′ − e(2)′ ,
2 i 2

© 2011 by Taylor and Francis Group, LLC

 Electromagnetic field 401

enter into the formula (17.92). In the new unit vectors this formula takes the form:
1
1

eλ = √ aλ1 − iaλ2 e(1)′ + √ aλ1 + iaλ2 e(2)′ = a1λ′ e(1)′ + aλ′
2 e
(2)′
.
2 2
Then, the probability of the left and right circular polarization (complying with the vectors
e(1)′ and e(2)′ ) is equal to

γ′ 1 γ
ρ11 = aλ′ λ′∗
1 a1 = [ρ + ργ22 + i(ργ12 − ργ21 )] = [1 + i(ργ12 − ργ21 )],
2 11
1
ργ′
22 = [1 − i(ργ12 − ργ21 )].
2
So, knowing the density matrix, we can find for the photon the probability of its having a po-
larization associated with the vector e(i) . This probability dw(i) is defined as a “projection”
of the tensor ργmn onto the direction of the vector e(i) , that is,

dw(i) = ργmn e(i) (i)∗

m en . (17.97)

In the general case a partial polarization may be conveniently described by three real
Stokes parameters ξ1 , ξ2 , and ξ3 , representing the density matrix as:
 
3  
γ 1 X 1 1 + ξ3 ξ1 − iξ2
ρ =  1+ ξj σj =
 . (17.98)
2 j=1
2 ξ1 + iξ2 1 − ξ3

This formula may be converted to determine the Stokes parameters through the matrix
density:
ξ = Sp(ργ σ). (17.99)
The parameter ξ3 characterizes a linear polarization along the directions of the x- or
y-axis. The probabilities that the photon is linearly polarized in these axes are equal to
(1 + ξ3 )/2 and (1 − ξ3 )/2 respectively, the values ξ3 = 1 or ξ3 = −1 corresponding to the
complete polarizations in these directions.
The parameter ξ1 characterizes a linear polarization in the directions for which α = π/4 or
α = −π/4 (β = 0). The probabilities that the photon is linearly polarized in these directions
√
are found by the projection of the tensor ργmn onto the vectors e(i) = (1, ±1, 0)/ 2 and
proves to be equal to (1 + ξ1 )/2 or (1 − ξ1 )/2, respectively. Again, at ξ1 = +1 or ξ1 = −1
we have the case of the complete polarization.
The parameter ξ2 defines the degree of the circular polarization. The projection of the
density matrix onto the circular polarization vectors gives the following expressions for the
probability of the right and left polarization:
1 1
(1 + ξ2 ) , (1 − ξ2 ) .
2 2
As the probabilities must be positive and less than 1, then we conclude:

|ξn | ≤ 1.

For the unpolarized photon we have

ξ1 = ξ2 = ξ3 = 0, ργ = 1/2.

In the case of the complete polarized photon the relation:

ξ12 + ξ22 + ξ32 = 1. (17.100)

© 2011 by Taylor and Francis Group, LLC

 402 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

should be carried out. Let us demonstrate that this is so indeed. To fulfill the relation
(17.100) the Stokes parameters may be taken as:
ξ1 = sin 2α cos β, ξ2 = sin 2α sin β, ξ3 = cos 2α. (17.101)
At these values of ξn the matrix ργ coincides with that to be built up with the help of the
polarization vector (17.93) and, consequently, be associated with the state of the complete
polarization
In the theory of the quantized electromagnetic field the polarization vectors represent the
factors standing ahead of the production and annihilation operators in the operator expan-
sion of the four-dimensional potential. Similar to a case of the electron, the polarization
density matrix is determined as:
(ργ )mn = eλm eλ∗
n , (17.102)
where eλ (λ = 1, 2) are the three-dimensional polarization vectors. Recall that for the
three-dimensional transverse gauge (k · A) = 0 in the reference frame with k k Oz the
vectors eλ are two-dimensional. Obviously, representation of the density matrix in terms of
the Stokes parameters (17.98) is valid in the second quantized theory as well. This is not
surprising as, unlike the electron, the photons have no nonrelativistic region, and quantum
mechanics of the photon is a relativistic theory from the start.
One may represent the photon density matrix in the four-dimensional form as well. It is
natural to define this matrix in the following way:
(ργ )µν = eλµ eλ∗
ν , (17.103)
where eλµ are four-dimensional polarization vectors, and λ takes the values equal to 1,2 as
before. When we choose the three-dimensional gauge eλ = (0, eλ ), then in the reference
frame with k k Oz the nonzero components of the tensor (ργ )µν will coincide with (17.102).
In order to obtain the four-dimensional representation of the photon density matrix in
(1) (2)
terms of the Stokes parameters, we introduce two unit space-like real vectors nµ and nµ ,
which are orthogonal both to each other and to the four-dimensional vector of the photon
momentum kµ :
′ 
(n(l) · n(l ) ) = −δll′ , l, l′ = 1, 2,
(17.104)
(k · n(l) ) = 0.
From (17.104) we find
eλσ = −(eλ · n(l) )n(l)
σ . (17.105)
(l)
Since the vectors n are space-like and orthogonal, then one may seek the reference frame
in which they have the form:
n(1) 1
µ = (0, e ), n(2) 2
µ = (0, e ).

The relation (17.105) is conceptually the relativistic generalization of this fact. By applica-
tion of (17.105) one may deduce the required expression for the density matrix:
 
3
′ 1 X
(l′ )
ργµν = (eλ · n(l) )(eλ∗ · n(l′ ) )nµ(l) n(l
ν
)
= 1 + ξj σj  n(l)
µ nν . (17.106)
2 j=1
ll′
(1)
Note that the conditions (17.104) provide no unambiguous definition for the vectors nµ
(2)
and nµ . Indeed, we can add to each of them a four-dimensional vector of the form Λkµ ,
where Λ is an arbitrary analytical function, to make them still meet the conditions (17.104).
This arbitrariness is due to the gauge ambiguity of the photon polarization density matrix
stating that matrices related by the gauge transformation are describing the same physical
reality.

© 2011 by Taylor and Francis Group, LLC

 Problems

3.1. Let a scalar field be confined in a spatial cube with the volume V = L3 , where
L is a cube edge length. Using the periodicity condition of a wavefunction ψ on every
spatial coordinate with the period L, write down an expansion of ψ in terms of frequency
components. Tending the cube size to infinity, turn to continuous representation. Write
down the four-dimensional energy-momentum vector and show that now the field confined
in the volume V may be considered as the harmonic oscillators totality. Find the energy
and momentum of these oscillators.
3.2. Determine the charge and space parity of two scalar particles (particle and antiparticle)
with the relative orbital moment equal l.
3.3. Find the explicit form of the field functions transformation matrix in the case of purely
spacial rotations for the Duffin-Kemmer equation describing scalar particles.
3.4. Working within the first order equations formalism, find the values of the following
quantities:
ϕ(±) (k)ϕ(±) (k).
3.5. When one works within the first quantized theory, the Klein-Gordon equation leads to
negative probabilities. In order to avoid the appearance of these probabilities the motion
equation does not include the time derivatives higher of the first order. Proceeding from
that and taking into account that the following relation
p
E = p2 + m2

must be fulfilled for relativistic particles, obtain the Dirac equation.

3.6. Consider the totality of transformations consisting both of purely spacial rotations
and of spacial axis inversions for a particle with the spin 1/2. Show, that the field function
frequency parts ψ (+) and ψ (−) are transformed independently.
3.7. Prove the validity of the spin summation rules (15.103) and (15.104) with the help of
the Green function to obey the inhomogeneous equation

(p̂ + m)G(p) = 1.

3.8. Show that for the particles with the spin 1/2 the energy-momentum tensor obeys the
continuity equation.
3.9. Find the values of the following expressions:

Sp{p̂1 p̂2 }, Sp{(p̂1 + m1 )(p̂2 + m2 )}

Sp{k̂1 (p̂1 + m1 )k̂2 (p̂2 + m2 )}, Sp{k̂1 k̂2 (p̂1 + m1 )(p̂2 + m2 )},
γ ν γ µ γν , γ ν γ µ γ σ γν , γ ν γ µ γ σ γ ρ γγν .
3.10. Write down the Dirac equation in such a representation where it has no imaginary
coefficients.
3.11. In the rest reference frame the spin of a free Dirac particle is conserved and its
wavefunction in the Dirac-Pauli representation has only two components for which the spin
projections onto the given axis are equal to ±1/2. Find such a representation where the

403
© 2011 by Taylor and Francis Group, LLC
 404 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

wavefunction (plane wave) possesses only two components corresponding to the definite
values of the spin projection in the rest reference frame.
3.12. Consider the totality of 16 linearly independent four-row matrices ΓA (A = 1, 2...16)

1, γν , γ5, iγ 5 γ ν , σ µν

and obtain the completeness condition for them. Carry out the same for 4 linearly inde-
pendent two-row matrices ΣB (B = 1, 2, 3, 4)

1, σx , σy , σz .

3.13. Employing the bispinors ψa and ψb , build up 16 bilinear combinations to be grouped

into 5 different Lorentz-covariant quantities. Making use of the field functions transforma-
tion law, check upon their transformation properties.
3.14. Build up Lorentz scalars from four bispinors ψa , ψb , ψc , and ψd .
3.15. In calculating scattering reactions, the following formulae
X
(ψ a ΓA ψb )(ψ c ΓA ψd ) = CAA′ (ψ a ΓA′ ψd )(ψ c ΓA′ ψb )
A′

(Firz relations) are very useful. Find all the coefficients CAA′ .
3.16. Find the energy levels and the wavefunctions of the electron in the constant magnetic
field.
3.17. Derive the Pauli equation
 
∂ϕ 1  e 2 e~
i~ = p − A + eA0 − (σ · H) ϕ
∂t 2m c 2mc

from the Dirac one.

3.18. Find the Dirac equation solution in the field of the arbitrary polarized plane wave

Aµ = Aµ (θ), θ = k µ xµ , kµ Aµ = 0,

where kµ is a wave vector. Define the effective electron mass m2∗ = qµ q µ , where qµ is a time
average of the kinetic momentum density Qµ (the kinetic momentum is pµ − eAµ .)
3.19. Define a change of a particle polarization direction moving in the plane perpendicular
to a homogeneous magnetic field (v ⊥ H).
3.20. Find a change of a particle polarization direction moving in a homogeneous electric
field.
3.21. In supersymmetric theories a particle γ with the spin 1/2 is a superpartner of γ. At
the same time γ possesses the Majorana nature. Assuming that in the process

γ+γ →γ+γ

the C-parity is conserved, find the intrinsic parity of γ.

3.22. In the general form the matrix element of the conserved electromagnetic current
(em)
Jµ for the neutrino is as follows:

< ν(pf , sf )|Jµ(em) |ν(pi , si ) >= u(pf , sf ){FQ (q 2 )γµ + FA (q 2 )(q 2 qµ − q̂qµ )γ5 −

−i[FM (q 2 ) + FE (q 2 )γ5 ]σµν q ν }u(pi , si ),

where q µ = pµf − pµi , Fn (q 2 ) are formfactors determining distributions of the electric charge
(n = Q), axial charge (n = A), magnetic dipole moment (n = M ), and electric dipole

© 2011 by Taylor and Francis Group, LLC

 References 405

moment (n = E), respectively. Demanding the CP T -invariance of the theory, show that for
the Majorana neutrino the single nonvanishing formfactor is the axial one (this formfactor
describes an anapole moment).
3.23. For an arbitrary field with the spin 1/2 the CP -transformation is determined by the
following way:
CPΨ(r, t)(CP)−1 = ηCP ∗
γ4 γ2 Ψ∗ (−r, t),
where ηCP is a phase factor. Using this definition and assuming the Dirac neutrino nature,
prove the validity of the relation:

CP|ar (p) >= ηCP |br (−p) >,

where |ar (p) > (|br (p) >) designates the state to involve a particle (antiparticle) with the
momentum p and polarization r.
3.24. If the vector field causal function is defined by the relation:
c′ (−) (+)
Dkl (x) = θ(x0 )Dkl (x) − θ(−x0 )Dkl (x) (A)

where Z  
(±) ±1 kl kn
Dkl (x) = eikx θ(±k0 )δ(k 2 − m2 ) gkl − dk,
(2π)3 i m2
the direct calculations of the expression (A) using the integral representation of the θ-
function Z +∞ ix0 τ
1 e
θ(±x0 ) = dτ,
2πi −∞ τ ∓ iǫ
leads to the expression different from
 
c 1 ∂2
Dkl (x) = gkl + 2 k Dc (x)
m ∂x ∂xk
(±)
in neighborhood of the point x = 0. Find Dkl (x) and point out the reason of this contra-
diction.
3.25. In the photon case find the vector functions (spherical vectors) to be eigenfunctions
both of the square angular momentum operator and of the angular momentum projection
operator onto the given axis. [NOTE: Make use of the relation known from nonrelativistic
quantum mechanics
[L̂k , Âk ] = iǫklm Âm ,
where Âm is some vector physical quantity to characterize a system.]
3.26. Show that the theory to be described by the Lagrangian of the problem 1.16 is
quantized with the use of the indefinite metric.

© 2011 by Taylor and Francis Group, LLC

 References

[1] E. Schrödinger, Ann. Phys. 79, 361 (1926).

[2] W. Gordon, Zeit. Phys. 40, 117 (1926).
[3] O. Klein, Zeit. Phys. 37, 895 (1926).
[4] V. Fock, Zeit. Phys. 38, 242 (1926).
[5] J. Kudar, Ann. Phys. 81, 632 (1926).
[6] I. S. Gradschtein, I. M. Ryzhik, Tables of Integrals, Sums and Series, (Fiz. Mat.
Glz., Moskow, 1963).
[7] J. Schwinger, Phys. Rev. 74, 1439 (1948); 75, 651 (1949).
[8] H. Umezawa, Prog. Theor. Phys. 7, 551 (1952).
[9] R. J. Duffin, Phys. Rev. 54, 1114 (1938); N. Kemmer, Proc. Roy. Soc. A173,
91 (1939).
[10] P. A. M. Dirac, Proc. Roy. Soc. A117, 610 (1928).
[11] W. Pauli, Ann. de l’Institut Henri Poincare, 6, 137 (1936).
[12] W. Pauli, Handbuch der Physik, (Bd. 24/1, Berlin, 1933).
[13] P. A. M. Dirac, Proc. Cambr. Phil. Soc. 26, 376 (1930).
[14] J. R. Oppenheimer, Phys. Rev. 35, 562 (1930).
[15] C. D. Anderson, Phys. Rev. 43, 491 (1933).
[16] E. R. Caianello, S. Fubini, Nuovo Cimento 9, 1218 (1952).
[17] H. Weyl, Zeit. Phys. 56, 330 (1929).
[18] V. Bargman, L. Michel, V. Telegdi, Phys. Rev. Lett. 2, 435 (1959).
[19] W. H. Furry, Phys. Rev. 81, 115 (1951).
[20] V. B. Berestetskii, E. M. Lifshitz, and L. P. Pitaevskii, Relativistic Quantum
Theory, (Pergamon Press, Oxford, 1971).
[21] R. S. van Dyck Jr., P. B. Schwinberg, and H. G. Dehmelt, Phys. Rev. Lett. 59,
26 (1987).
[22] P. J. Mohr and B. N. Taylor, Rev. Mod. Phys. 72, 351 (2000).
[23] E. Majorana, Nuova Cimento 14, 171 (1937).
[24] S. Fukuda et al., Super-Kamiokande Collaboration, Phys. Rev. Lett. 86, 5656
(2001).
[25] Q. R. Ahmad et al., SNO Collaboration, Phys. Rev. Lett. 87, 071301 (2001).

407
© 2011 by Taylor and Francis Group, LLC
 408 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

[26] H. V. Klapdor-Kleingrothaus, A.Dietz, H.L.Harney and I.V.Krivosheina

(Heidelberg-Moscow Collaboration), Mod. Phys. Lett. A16, 2409 (2002).
[27] C. E. Aalseth et al., Mod. Phys. Lett. A510, 1475 (2002).
[28] B. Touschek, Zeit. Phys. 125, 108 (1949); F. Gursey, Nuovo Cimento, Serie X
7, 411 (1958).
[29] R. V. Case, Phys. Rev. 107, 307 (1957).
[30] I. B. Zel’dovich, JETF 33, 1531 (1957).
[31] B. Kayser, F. Gibrat-Debu and F. Perrier, The physics of massive neutrinos,
(World Scientific, Singapore, 1989).
[32] H. V. Klapdor-Kleingrothaus et al. Phys. Lett. A586, 198 (2004).
[33] J. C. Pati and A. Salam, Phys. Rev. D10, 275 (1974); R. N. Mohapatra and
G. Senjanovic, Phys. Rev. D23, 165 (1981).
[34] O. M. Boyarkin, G. G. Boyarkina, T. I. Bakanova, Phys. Rev. D70, 113010-1
(2004).
[35] D. London, G. Belanger, and J. N. Ng, Phys. Lett. B188, 155 (1987).
[36] O. M. Boyarkin, D. Rein, Phys. Rev. D53, 361 (1996).
[37] A. Proca, J. Phys. et Rad. 7, 347 (1936).
[38] N. Kemmer, Proc. Roy. Soc. A166, 127 (1938).
[39] P. Roman, Theory of elementary particles, (N. H. Pub. Co. Amsterdam, 1964);
I. Corson, Theory of tensors, spinors and wave equations, (Benjamin Press,
NY, 1953).
[40] P. A. M.Dirac, Proc. Roy. Soc. A133, 821 (1931).
[41] M. G. Galkin, Phys. Lett. A28, 45 (1968).
[42] S. Gupta, Proc. Phys. Soc. A63, 681 (1950).
[43] K. Bleuler, Helv. Phys. Acta. 23, 567 (1950).
[44] T. D. Lee, C. N. Yang, Phys. Rev. 128, 885 (1962); T. D. Lee, G. C. Wick,
Nucl. Phys. B9, 209 (1969).
[45] J. Hsu, Lett. Nuovo Cimento 12, 503 (1975).
[46] O. M. Boyarkin, JETP 48, 13 (1978).
[47] O. M. Boyarkin, J. Phys. G: Nucl. Phys. 8, 161 (1982); O. M. Boyarkin, Sov.
Phys. J. 10, 828 (1989).

© 2011 by Taylor and Francis Group, LLC

 Part IV

Quantum Electrodynamics

409
© 2011 by Taylor and Francis Group, LLC
 18
S-matrix

You are not wrong, who deem

That my days have been a dream;
Yet if hope has flown away
In a night, or in a day,
In a vision, or in none,
Is it therefore the less gone?
All that we see or seem
Is but a dream within a dream.
Edgar Allan Poe, “A Dream within a Dream”

18.1 Equations and dynamic variables in the Heisenberg

representation
Having used the principle of the local gauge invariance with respect to the U (1)-group
in Section 4.5 we have established the form of the Lagrangian to describe the interaction
between the electromagnetic and electron-positron fields:
i  1
L= ψ(x)γ µ Dµ ψ(x) − Dµ† ψ(x)γ µ ψ(x) − mψ(x)ψ(x) − Fµν (x)Fµν (x), (18.1)
2 4
where Dµ = ∂µ + ieAµ (x) and Fµν (x) = ∂µ Aν (x) − ∂ν Aµ (x). Here the operators of the
electromagnetic and electron-positron fields are denoted by a bold font to stress that they
are set in the Heisenberg representation. Attention is drawn to the fact that the local gauge
invariance is kept only when an electromagnetic field quanta mass is identical equal to zero.
If the photons had a mass m, then the Lagrangian of a free electromagnetic field would
include the mass term −m2 Aµ Aµ that is gauge noninvariant.
From the Lagrangian (18.1) follows the basic equations of the QED:

[γ µ (i∂µ − eAµ (x)) − m]ψ(x) = 0, (18.2)

[γ µT (i∂µ + eAµ (x)) + m]ψ(x) = 0, (18.3)

∂ µ Fµν (x) = jν (x). (18.4)
The electromagnetic current is defined by the expression:

jν (x) = eψ(x)γν ψ(x), (18.5)

which formally coincides with that for the current in the case of the free electron-positron
field. However, it is not exactly the same since this expression contains the quantities ψ(x)
and ψ(x) to obey the equations for interacting fields rather than those for free fields. The

411
© 2011 by Taylor and Francis Group, LLC
 412 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

field operators act on the state vectors in the particles number space Φ. Since we work in
the Heisenberg representation, then the following relation:

∂
Φ=0
∂t
is valid.
In Eq. (18.4) one may cross to the electromagnetic potentials:

Aν (x) = jν (x), (18.6)

In so doing, we should supplement this equation, as in the case of free fields, with the
condition on the state vectors:
 
∂
A+ (x) Φ = 0. (18.7)
∂xµ µ

Thanks to the current continuity, the operator ∂ µ Aµ (x) satisfies one and the same equation
both in a free case and in the presence of interaction:

∂Aµ (x)
 = 0. (18.8)
∂xµ

Therefore, the additional condition (18.7) is relativistically invariant.

Eqs. (18.2), (18.3), and (18.6) follow from the Lagrangian that one conveniently writes
down as a sum of the Lagrangians of free and interacting fields:
(e) (γ)
L = L0 + L0 + Lint , (18.9)

where
(e) i 
L0 = ψ(x)γ µ ∂µ ψ(x) − ∂µ ψ(x)γ µ ψ(x) − mψ(x)ψ(x), (18.10)
2
(γ) 1
L0 = − ∂µ Aν ∂ µ Aν , (18.11)
2
Lint = −jµ (x)Aµ (x), (18.12)
Recall, that the Lagrangians (18.1) and (18.9) differ from each other by an insignificant
term having the four-dimensional divergence form. The interaction Lagrangian Lint does
not contain derivatives of the field operators. We speak of such interactions as couplings
without derivatives. In this case the interaction Hamiltonian density Hint is simply equal
to −Lint and the interaction Hamiltonian is given by:
Z
Hint (t) = − Lint d3 x. (18.13)

A system of differential equations (18.2)–(18.4) contains only a part of the information

about the interacting fields. These equations should be amplified by permutation relations
(PR) between the field operators. As evolution of the operators in time is defined by
equations of motion, the PR in the case of interacting fields may be given for the initial
instant of time only. It is clear that the determination of the PR at some time is equivalent
to their calculation at the coincident instants of time. Having realized what PR we should
find, we can easily guess the way of their searching. We remember that precisely the
simultaneous PR are at the base of the canonical quantization procedure. Then, with the

© 2011 by Taylor and Francis Group, LLC

 S-matrix 413

help of the canonical formalism one may find the following nonvanishing simultaneous PR
(see, the formulae (15.257) and (17.51) ):

[Aµ (x), ∂0 Aν (x′ )]t0 =t′0 = igµν δ (3) (r − r′ ). (18.14)

{ψ(x), ψ † (x′ )}t0 =t′0 = δ (3) (r − r′ ). (18.15)

As it seems, a change to the PR valid for any time intervals may be realized in the same way
as for free fields. So, for example, in the free case for the electromagnetic field the solution
Aν (x) on arbitrary space-like hypersurface σ0 could be determined by application of the
Green function D(x) in terms of the solution on any more early space-like hypersurface
Z  
µ ′ ∂D(x − x′ ) ′
′
′ ∂Aν (x )
Aν (x) = dσ (x ) Aν (x ) − D(x − x ) . (18.16)
σ0 ∂x′µ ∂x′µ

Further, selecting the hypersurface σ0 by the corresponding way, one can obtain the PR
sought. If the formula similar to (18.16) existed in the case of the interacting fields, then
we would manage to define the PR at arbitrary instants of time. However to derive such a
formula demands solutions knowledge of Eqs. (18.2), (18.3), and (18.6), which represent the
partial differential inhomogeneous equations system. It is obvious that the task of solving
this system is rather complicated. Moreover, in the majority of cases solutions may be
obtained by means of approximate methods only. Because approximate PR do not suit us,
we have to look for other ways to solve this problem.

18.2 Interaction representation

Having carried out the unitary transformation on the Heisenberg operators N :

N → N(s) = e−iHt NeiHt , (18.17)

where H is a total Hamiltonian of a system, we pass to the Schrödinger representation in

which, as it is obvious from (18.17), operators are not changed as time goes on:
d (s)
N = 0. (18.18)
dt
The operator transformation (18.17) is associated with the state vector transformation:

Φ → Φ(s) (t) = e−iHt Φ. (18.19)

In the Schrödinger representation the state vector Φ(s) (t) satisfies the Schrödinger equation:
∂ (s)
i Φ (t) = H(s) Φ(s) (t). (18.20)
∂t
From (18.17) it follows that in the representations by Heisenberg and Schrödinger the Hamil-
tonians coincide:
H(s) = H. (18.21)
However, in interacting fields theory an interaction representation proves to be more
convenient. This representation is sometimes named after the authors who shaped it into
an explicit covariant form [1], the Tomonaga-Schwinger representation.

© 2011 by Taylor and Francis Group, LLC

 414 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Let us divide the total Hamiltonian H(s) into the free and interaction Hamiltonians:
(s) (s)
H(s) = H0 + Hint . (18.22)

Pass to an interaction representation connected with the Schrödinger one by a unitary

transformation: (s) (s)
N(s) → N = eiH0 (t−t0 ) N(s) e−iH0 (t−t0 ) , (18.23)
(s)
Φ(s) (t) → Φ(t) = eiH0 (t−t0 )
Φ(s) (t). (18.24)
From Eqs. (18.23) and (18.24) it follows that both operators and state vectors belonging
to the Schrödinger and new representations coincide at t = t0 .
Let us find out how to change operators and state vectors in the interaction representation
with the passage of time. With allowance made for (18.18) from (18.23) follows:

∂ (s) (s) (s) (s)

(s) (s)
N = iH0 eiH0 (t−t0 ) N(s) e−iH0 (t−t0 ) − eiH0 (t−t0 ) N(s) H0 e−iH0 (t−t0 ) = i[H0 , N ],
∂t
(18.25)
(s)
where we have taken into account that H0 = H0 . Analogously, for the state vectors we
have:
∂ (s) (s) (s)
(s) (s)
i Φ(t) = −H0 eiH0 (t−t0 ) Φ(s) (t)+eiH0 (t−t0 ) (H0 +SHint )Φ(s) (t) = Hint Φ(t). (18.26)
∂t
Thus, in the interaction representation both operators and state vectors are varied, the
evolution law of operators being defined by the free fields Hamiltonian and that of state
vectors—by the interaction Hamiltonian.
Now we establish a connection between the interaction representation and the Heisenberg
one. Setting in (18.24)
(s)
Φ(s) (t) = e−iH (t−t0 ) Φ, (18.27)
we obtain
Φ(t) = S(t, t0 )Φ, (18.28)
where (s)
(t−t0 ) −iH(s) (t−t0 )
S(t, t0 ) = eiH0 e . (18.29)
Note, that usage in the exponent of the quantity t − t0 ensures the coincidence of state
vectors in the interaction and Heisenberg representations at the initial moment of time.
From (18.29) it follows that the operator S(t, t0 ) obeys the equation:

∂
i S(t, t0 ) = Hint S(t, t0 ) (18.30)
∂t
and the condition S(t0 , t0 ) = 1. It is also an easy matter to verify that this operator relates
operators of the Heisenberg representation with those of the interaction representation.
With this aim in view one should substitute the expression
(s) (s)
N(s) = e−iH (t−t0 )
NeiH (t−t0 )
,

into (18.23). As a result we get

N = S(t, t0 )NS −1 (t, t0 ). (18.31)

Now we are quite ready to feel all the advantages of the interaction representation. First,
we show that in this representation operators of the electromagnetic and electron-positron

© 2011 by Taylor and Francis Group, LLC

 S-matrix 415

fields satisfy equations for the free fields. For this purpose equations for field operators in
the commutator form (18.25) are needed. In so doing to find the four-dimensional vector
of field momentum we shall use the expression:
Z Z X ∂L ∂ψ k
Pµ = dσ ν Tµν = dσ ν [ − Lgµν ]. (18.32)
∂ψ k;ν ∂xµ
k

The usefulness of such a definition implies that on the strength of the conservation law:

∂ µ Tµν = 0, (18.33)

the expression (18.32) does not depend on a choice of a space-like hypersurface σ. To obtain
an equation describing behavior of the operator ψ(x) one must set N (x) equal to γ0 ψ in
(18.25). Then, all our efforts are reduced to the calculation of the commutator between
R (0) (0)
γ0 ψ and H0 = dσ ν T0ν (Tµν is an energy-momentum tensor for the free electromagnetic
field). Since a hypersurface σ is arbitrary, there is nothing to prevent us to take it as a
hyperplane with t = t′ . In this case H0 acquires the view:
Z  
1 ′ ∂Aν (x′ ) ∂Aν (x′ ) ∂Aν (x′ ) ∂Aν (x′ ) ′ ∂
H0 = dr −2 ′ ′
+ µ′
+ ψ(x ) iγk ′k +
2 t=t′ ∂t ∂t ∂x ∂xµ′ ∂x
← ! #
∂
−iγk ′k + 2m ψ(x′ ) , (18.34)
∂x
where the arrow, above the differential operator ∂k , directed in the left means the operator
acts on a field function standing in the left. Further on, one should take into consideration
that at coinciding times operators in the interaction representation satisfy the same PR
as do operators in the Heisenberg representation (18.14), (18.15). Then, allowing for the
operator relation:
[A, BC] = {A, B}C − B{A, C},
we result in
Z   ′

1 ′ † ′ k ∂ψ(x )
[H0 , γ0 ψ(x)] = dr {ψ (x ), ψ(x)} iγ +
2 t=t′ ∂x′k
† ′

k ∂ψ (x ) ′ † ′ ′
−iγ { , ψ(x)}ψ(x ) + 2m{ψ (x ), ψ(x)}ψ(x ) =
∂x′k
Z   ′

1 ′ (3) ′ k ∂ψ(x ) ∂δ (3) (r − r′ )
=− dr δ (r − r ) iγ ′ − iγ k ψ(x′ )+
2 t=t′ ∂xk ∂x′k
Z   
o 1 ∂ψ(x′ )
+2mδ (3) (r − r′ )ψ(x′ ) = − dr′ 2δ (3) (r − r′ ) iγ k −
2 t=t′ ∂x′k
  
k ∂ ∂ψ(x)
−iγ ′
(3)
δ (r − r )ψ(x ) + 2mδ (r − r )ψ(x ) = −iγ k
′ ′ (3) ′ ′
− mψ(x), (18.35)
∂xk ∂xk
where to reach the last line we have made an allowance for the three-dimensional Gauss
theorem. Connecting (18.25) and (18.35) we get

[iγ µ ∂µ − m]ψ(x) = 0, (18.36)

For N (x) = ψ † (x) and N (x) = ∂0 Aµ (x) analogous operations lead to the equations:

[iγ µT ∂µ + m]ψ(x) = 0. (18.37)

© 2011 by Taylor and Francis Group, LLC

 416 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Aµ (x) = 0. (18.38)

So, in the interaction representation field operators do satisfy equations for free fields. In
order to make the picture complete these equations should be supplemented by the general
character PR, that is, by the PR at incoincident moments of time. To change from the
simultaneous PR to the general ones could be realized with the help of formulae determining
the operators evolution in time, that is, by means of (18.16) and
Z
ψ(x) = −i dσµ (x′ )S(x − x′ )γ µ ψ(x′ ). (18.39)

The insertion of the expression (18.39) into an anticommutator for electron-positron field
operators:
{ψ(x), ψ(x′ )},
choice of a hypersurface in the form of a hyperplane t = t′ , and usage of the simultaneous
PR, give us the PR for an arbitrary moment of time:

{ψ(x), ψ(x′ )} = −iS(x − x′ ), (18.40)

in which we immediately recognize the PR for the free case. In a similar manner, by ap-
plication of the formulae (18.14) and (18.16) we must assure ourselves that for the electro-
magnetic field the PR in the interaction representation has the same view as the Heisenberg
representation for the free case:

[Aµ (x), Aν (x′ )] = igµν D0 (x − x′ ). (18.41)

18.3 Scattering matrix

In quantum field theory, similar to quantum mechanics, two problems are of principal
importance. The first of them is reduced to finding a spectrum of the bound states (for
example, spectrums of the hydrogen atom, of positronium, of muonium, and so on), and the
second problem is connected with investigating both elastic and inelastic particles scattering.
Here we shall centre on the latter problem.
Recall statements of the problem about particle collision in nonrelativistic quantum me-
chanics. It has been assumed that in the initial state at t = −∞, the particles are too
distant from each other so that interaction between them is absent, that is, Hint = 0. Then
the particles are drawing together and coming into an interaction. Further, they again
move apart, with an infinitely large distance between them, to become free at t = ∞. The
scattering problem consists in finding the wavefunction for a system of particles at t = ∞
provided the wavefunction of a system of free particles at t = −∞ is known. An expan-
sion of the final state wavefunction in terms of the wavefunctions of free particles makes it
possible to determine the amplitudes for different scattering processes.
When we use this procedure in the quantum electrodynamic (QED), the situation is
greatly complicated. Since we take into account particle interaction effects with a vacuum,
the particles are not free both before and after the collision. The only formal possibility
to switch off the interaction between particles in the QED is to set a coupling constant
between the electron-positron and electromagnetic fields (electron charge) equal to zero.
But varying the charge of the electron, we change its other characteristics as well (mass,
magnetic moment, and so on). For example, the electron mass is the lowest energy of

© 2011 by Taylor and Francis Group, LLC

 S-matrix 417

the state with the charge e in the case of interacting fields. It is naive to think that this
quantity will be the same as the bare electron mass. A particle is referred to as “bare”
when it has no interaction with the vacuum of the QED whose structure is given by Hint .
And remaining within the scope of the proposed procedure, we have a possibility to study
scattering processes only of bare and not real particles. However, the situation may be
improved owing to a convenient method called the adiabatic hypothesis. By this hypothesis,
the interaction Hamiltonian should be additionally multiplied by a converging factor of the
form exp (−α|t|) that, at the end of our calculations, should be set equal to unity by the
limiting transition α → 0. And the scattering process per se is represented as follows.
For t = −∞ a distance between the particles is too large and there is no interaction

exp (−α|t|)Hint = 0. (18.42)

In this manner we have a system of bare free particles, that is, the system is described by
the state vector that is an eigenvector of the operator H0 . Then adiabatic switching-on
of the interaction between these particles and vacuum takes place and, as a result, the
particles “dress themselves” in fur coats consisting of virtual particles. Despite the fact
that at this stage bare particles are transformed into the real ones, as before the particles
have no interaction with each other. Following the next full switching-on of the interaction,
scattering of the particles takes place. Subsequently, as the particles become more and more
distant from each other, they regain their freedom, however keeping their fur coats due to
the interaction with the vacuum. At t = +∞ the interaction is adiabatically switched off,
whereas the particles (losing their fur coats and being more and more separated) are trans-
formed into free bare particles. Since both switching-on and switching-off of the interaction
is adiabatic, the states of a system of fields are subjected only to an adiabatic change that
does not lead to the appearance of new states or the disappearance of old ones. This, in
turn, is indicative of a complete agreement between the states of bare and real particles and
hence between the processes of their scattering. We will come back to the characteristics of
real particles later (under discussion of the QED renormalizability), and now we consider
real and bare particles, making no distinction between them.
Thus, the scattering problem statement in the QED proves to be the same as in nonrel-
ativistic quantum mechanics. In the process of solving this problem we use the interaction
representation. It is easily seen that a formal solution for the state vector equation (18.26)
is of the form:
Φ(t) = S(t, t0 )Φ(t0 ). (18.43)
Assuming t0 = −∞ and t = +∞, we can find the linkage between the initial and final state
vectors of the field system we are interested in as follows:

Φ(∞) = S(∞, −∞)Φ(−∞). (18.44)

In (18.44) the operator S(∞, −∞) ≡ S that changes the initial state vector Φ(−∞) to the
final state vector Φ(∞) is called the scattering matrix or the S-matrix. This matrix has
been first introduced in Ref. [2] devoted to the problems of nuclei scattering. In 1943 it was
again studied in detail by Heisenberg [3] but as applied to a theory of elementary particles.
The S-matrix obeys an integral equation:
Z t
S(t, t0 ) = 1 − i Hint (t′ )S(t′ , t0 )dt′ , (18.45)
t0

which is equivalent to the differential equation (18.30) with the boundary condition

S(t0 , t0 ) = 1.

© 2011 by Taylor and Francis Group, LLC

 418 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Having set
Φ(−∞) = Φi ,
where the index i denotes an initial state of a system consisting of electrons and photons,
we rewrite (18.44) in the form:
Φ(∞) = SΦi . (18.46)
Now we should expand a final state in terms of an orthonormalized set of free field state
vectors: X
Φ(∞) = a f Φf . (18.47)
f

Using the orthonormalization condition, from (18.47) we obtain the following expression for
the coefficients af ≡ ai→f :

ai→f = (Φf , Φ(∞)) = (Φf , SΦi ) ≡< f |S|i > . (18.48)

So, probability amplitudes for different scattering processes are defined by the S-matrix
elements.
To determine the explicit form of the S-matrix we address Eq. (18.30) where the inter-
action Hamiltonian is given by:
Z
Hint (t) = jµ (x)Aµ (x)dr. (18.49)

We shall seek a solution of this equation in the form of series in powers of e:

∞
X
S(t, t0 ) = S (n) (t, t0 ), (18.50)
n=0

where the matrix S (n) is proportional to en . To substitute (18.50) into (18.30) leads to the
equation:
∂  (0)  
i S + S (1) + S (2) + S (3) + ... = Hint (t) S (0) + S (1) +
∂t

+S (2) + S (3) + ... . (18.51)

Allowing for that Hint ∼ e, we equate terms at the same powers of e:

∂S (0) ∂S (1) ∂S (2)

i = 0, i = Hint (t)S (0) , i = Hint (t)S (1) ,
∂t ∂t ∂t
(18.52)
∂S (3)
= Hint (t)S (2) , ...
i
∂t
The solutions of the system (18.52) are yielded by the expressions:
Z t
(0) (1)
S (t, t0 ) = 1, S (t, t0 ) = −i dt1 Hint (t1 ), S (2) (t, t0 ) =
t0

Z t1 Z t Z t1
(1) 2
= −i dt2 Hint (t2 )S (t2 ) = (−i) dt1 dt2 Hint (t1 )Hint (t2 ),
t0 t0 t0

................. (18.53)

© 2011 by Taylor and Francis Group, LLC

 S-matrix 419
Z t Z t1 Z t2 Z tn−1
S (n) (t, t0 ) = (−i)n dt1 dt2 dt3 ... dtn Hint (t1 )Hint (t2 )Hint (t3 )...Hint (tn ).
t0 t0 t0 t0
The upper integration limits of these integrals are different and lay in the interval (t, t0 ). It
would be more attractive to transform the integrals in (18.53) so that integration in each
of them is carried out over one and the same interval (t0 , t) rather than over different ones
(t0 , t), (t0 , t1 ), ...(t0 , tn−1 ). The first to make this was F. Dyson [4]. For this purpose to be
achieved, it is sufficient to choose the time axis direction, that is, for example, to demand
t > t1 > t2 > ...tn−1 > tn > t0 .
We show how that is being done by the example of the expression for S (2) (t, t0 ).
In S (2) (t, t0 ) the integration domain is simply the triangle, which is placed below the
bisectrix of the coordinate angle in the plane (t1 , t2 ) (Fig. 18.1), and for which t1 > t2 .
Having fulfilled replacing the variables

t2

t2 = t1
t

t0

t0 t t1

FIGURE 18.1
The integration domain for S (2) (t, t0 ).

t1 → t2 , t2 → t1 ,
we obtain for S (2) (t, t0 ) the expression:
Z t Z t2
S (2) (t, t0 ) = − dt2 dt1 Hint (t2 )Hint (t1 ). (18.54)
t0 t0

Let us change the integration order in S (2) (t, t0 ), namely, first we shall integrate over t1 . It
leads to modification of the integration limits:
Z t Z t
(2)
S (t, t0 ) = − dt1 dt2 Hint (t2 )Hint (t1 ). (18.55)
t0 t1

At present, for the integration domain we have the triangle, placed above bisectrix of the
coordinate angle, for which t2 > t1 . To combine the transformed expression (18.55) with
the original gives:
Z Z t1 Z t 
(2) 1 t
S (t0 , t) = − dt1 dt2 Hint (t1 )Hint (t2 ) + dt2 Hint (t2 )Hint (t1 ) . (18.56)
2 t0 t0 t1

© 2011 by Taylor and Francis Group, LLC

 420 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the case of the commutation of the operators Hint (t2 ) and Hint (t1 ) the expression (18.56)
would be simply rewritten down in the form of the double integral:
Z Z t
1 t
S (2) (t0 , t) = − dt1 dt2 Hint (t1 )Hint (t2 ),
2 t0 t0

where both integrations be already fulfilled in one interval (t0 , t), that is, over the whole
quadrate pictured in Fig. 18.1. However, the operators Hint (t2 ) and Hint (t1 ) do not
commute and we should think out some trick allowing us to simplify the expression (18.56).
Matters could be straightened out by the use of the chronological product operator T . Since
the fermion factors enter into the operator Hint (t) by pairs in which both factors refer to
the same moment of time, we have:

Hint (t1 )Hint (t2 ), t 1 > t2 ,
T (Hint (t1 )Hint (t2 )) = (18.57)
Hint (t2 )Hint (t1 ), t 2 > t1 ,
or

T (Hint (t1 )Hint (t2 )) = Θ(t1 − t2 )Hint (t1 )Hint (t2 ) + Θ(t2 − t1 )Hint (t2 )Hint (t1 ). (18.58)

Then the expression for S (2) (t, t0 ) (18.56) could be written in the following compact form:
Z Z t
1 t
S (2) (t, t0 ) = − dt1 dt2 T (Hint (t1 )Hint (t2 )). (18.59)
2 t0 t0

Once we had understood the basic idea, we consider the S-matrix element of the nth-
order. Allowing for the chosen time direction, we can represent the expression for S (n) (t, t0 )
in the view:
Z t Z t1 Z tn−1
(n) n
S (t, t0 ) = (−i) dt1 dt2 ... dtn Θ(t1 − t2 )Θ(t2 − t3 )...Θ(tn−1 − tn )×
t0 t0 t0

×Hint (t1 )Hint (t2 )...Hint (tn )). (18.60)

Assuming, next, that the integration order rearrangement is allowed, we put down S (n) (t, t0 )
as:
Z Z t Z t
(n) (−i)n X t
S (t, t0 ) = dt1 dt2 ... dtn Θ(tα1 −tα2 )...Θ(tαn−1 −tαn )Hint (tα1 )...Hint (tαn ),
n! t0 t0 t0
P
P (18.61)
where the symbol P denotes the summation over all permutations (t1 , t2 , ...tn ). Taking
into account that the Hamiltonian, which contains only bilinear combinations of fermion
fields, satisfies the relation:
X
T (Hint (t1 )Hint (t2 )...Hint (tn )) = Θ(tα1 − tα2 )...Θ(tαn−1 − tαn )Hint (tα1 )...Hint (tαn ),
P
(18.62)
we finally get
Z t Z t Z t
(n) (−i)n
S (t, t0 ) = dt1 dt2 ... dtn T (Hint (t1 )Hint (t2 )...Hint (tn )) . (18.63)
n! t0 t0 t0

Thus, we have shown that the S-matrix is defined by the following series:
∞ Z Z t Z t
X (−i)n t
S(t, t0 ) = dt1 dt2 ... dtn T (Hint (t1 )Hint (t2 )...Hint (tn )) . (18.64)
n=0
n! t0 t0 t0

© 2011 by Taylor and Francis Group, LLC

 S-matrix 421

The formal summation of this series leads to the expression:

 Z t 
′ ′
S(t, t0 ) = T exp [−i dt Hint (t )] . (18.65)
t0

Now, it is high time to stop and check whether in the process of finding S(t, t0 ) we have
done some unlawful things after which S(t, t0 ) is not the formal solution of Eq. (18.30).
With this aim in view, we differentiate the expansion (18.64) with respect to t:
∞ Z Z t Z t
∂S(t, t0 ) X (−i)n t
= dt1 dt2 ... dtn−1 nHint (t)T (Hint (t1 )Hint (t2 )...Hint (tn−1 )) .
∂t n=1
n! t0 t0 t0
(18.66)
Under writing the right-hand side of the equality (18.66) we have used the integrand expres-
sion symmetry and the fact that t > t1 , ... > tn−1 . All of this taken together has allowed
to take out Hint (t) for the sign of the T -product and put it on the left of the remaining
factors. Thus, the equality (18.66) may be represented in the following form:
∞ Z Z t Z t
∂S(t, t0 ) X (−i)n−1 t
i = Hint (t) dt1 dt2 ... dtn−1 (t)×
∂t n=1
(n − 1)! t0 t0 t0

∞ Z Z t Z t
X (−i)n t
×T (Hint (t1 )Hint (t2 )...Hint (tn−1 )) = Hint (t) dt1 dt2 ... dtn ×
n=0
n! t0 t0 t0

T (Hint (t1 )Hint (t2 )...Hint (tn )) = Hint (t)S(t, t0 ), (18.67)

which is what we set out to prove.
To obtain the S-matrix in the scattering problem, in the expansion (18.65) t0 must tend
to −∞ while t must tend to +∞. It gives
 Z ∞ 
S = T exp [−i Hint (t)dt] (18.68)
−∞

or, after substitution of the explicit expression for Hint (t) (18.12),
 Z ∞ 
µ 4
S = T exp [−i N (jµ (x)A (x))d x] . (18.69)
−∞

It is obvious that the Sn -element of the scattering matrix represents the T -product of the
N -ordered factors. The T -product of such a kind will be called the mixed T -product.
Let us focus our attention on some mathematical aspects in calculating the S-matrix.
First, transition to the limits t = ±∞ in the S (n) -element may be realized by n! different
methods. Second, without the use of wavepackets for the description of the initial and final
states, it is necessary to find the way for averaging of the terms periodically dependent
on t0 and t. Nevertheless, all these problems are removed due to the adopted adiabatic
hypothesis that necessitates the following substitution in the S-matrix:

Hint (t) → Hint (t) exp [−α|t|], (18.70)

and the transition to the limit α → 0 only after all the integrations are performed. We
shall demonstrate that this mathematical procedure also allows painlessly working with
unnormalized states of the plane-wave type.
Let us take S (2) from the expansion (18.53) and tend t0 to −∞ while t to +∞. Further
on, for the obtained expression we find the matrix element between the initial and final

© 2011 by Taylor and Francis Group, LLC

 422 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
(2)
states: Si→f ≡< f |S (2) |i > . Passing to the Schrödinger representation and introducing the
total system of the eigenstates of H0 between Hint (t1 ) and Hint (t2 ), we arrive at:

(2)
X Z +∞ Z t1
Si→f = − dt1 dt2 ×
n −∞ −∞

(s) (s)
×ei(Ef −En )t1 ei(En −Ei )t2 < f |Hint |n >< n|Hint |i >, (18.71)
where for the sake of simplicity we have assumed that the Schrödinger and interaction
representations coincide at t0 = 0. If |i > is the plane wave state, then the integral over t2
on the lower limit represents a fast oscillating quantity. To obtain the finite result would
be possible when the |i >-state be the wave package:
Z Z
i(Et−p·r) 3
|i >= ψp e d p with |ψp |2 d3 p < ∞.

Then, the contribution from the lower limit t2 = −∞ would be simply equal to zero accord-
ing to the Riemann-Lebesgue lemma. The replacement (18.70) proves to lead to the finite
result as well. Really, after its performance we have:

(2)
X < f |H (s) |n >< n|H (s) |i >
int int
Si→f = −2πiδ(Ei − Ef ) . (18.72)
n
Ei − En + iα

Using the similar line of reasoning, one may show the validity of the following formula:
(
(s)
X < f |H (s) |n >< n|H (s) |i >
int int
< f |S|i >= −2πiδ(Ei − Ef ) < f |Hint |i > + +
n
Ei − E n + iα
)
(s) (s) (s)
X < f |Hint |n1 >< n1 |Hint |n2 > ... < nk |Hint |i >
+... + , (18.73)
n1 ,n2 ...nk
(Ei − En1 + iα)(En1 − En2 + iα)...(Enk − Ef + iα)

in which we immediately get to know the result of the “ancient” (preceded S-matrix) per-
turbation theory.
Let us demonstrate that the S-matrix defined by the expression (18.69) is covariant. By
virtue of the fact that jµ (x)Aµ (x) and d4 x are scalars, and the normal product operation
is covariant, the exponent in the expression (18.69) is an invariant. Since the chronological
product operator is defined by the invariant way as well, then the relativistic covariance of
the S-matrix becomes obvious.
From the definition (18.69) it also follows one more important property of the S-matrix,
namely, its unitarity. Actually, due to hermicity of L(x) the matrix S † is expressed in the
product form of the same factors as S with only two differences, namely, the factors in S †
are taken with the opposite signs and installed in the reverse chronological order. By this
reason, under multiplication of S and S † all the factors pairwise give 1.
The S-matrix must also satisfy the causality condition that reads: “some event having
occurred in a system could have an impact on a system evolution course only in the future
and could not have any influence on a system behavior in the past.” To mathematically
formulate this condition it is necessary to introduce the function g(x), which has the value
in the interval (0,1) and characterizes intensity of switching-on an interaction in the point
x. In the regions where g(x) = 0 an interaction is absent, where g(x) = 1 it is switched
on completely, and under 0 < g(x) < 1 it is switched on only in part. Replacing Lint with
Lint g(x), we arrive at an interaction having been switched on with the intensity g(x) and,
accordingly, the scattering matrix S(g) depending on g(x).

© 2011 by Taylor and Francis Group, LLC

 S-matrix 423

Let us illustrate the idea of obtaining the causality condition in the explicit view. To do
this, we shall take up the following case. The space-time region G, in which the function
g(x) is different from zero, breaks up into two separate subregions G1 and G2 , all points of
G1 laying in the past as related to some moment of time t = τ and all points of G2 laying
in the future as related to t = τ (in our designations it will look like G2 > G1 ). In this case
the function g(x) is defined as a sum of two functions:

g(x) = g1 (x) + g2 (x), (18.74)

one of them g1 (x) not being zero only in G1 while the other g2 (x) nonvanishing only in G2 .
Then, at the moment of time τ the state vector Φτ not depending on an interaction in the
region G2 will be defined as:
Φτ = S(g1 )Φi , (18.75)
where S(g1 ) is the scattering matrix that describes switching-on an interaction with the
intensity g1 (x), Φi is the initial state vector of the system. The final state Φf could be
obtained from Φτ with the help of the operator S(g2 ) governing an interaction in the region
G2 :
Φf = S(g2 )Φτ . (18.76)
Comparing Eqs. (18.74)–(18.76) with

Φf = S(g)Φi , (18.77)

we get
S(g1 + g2 ) = S(g1 )S(g2 ) at G2 > G1 . (18.78)
If all the points of the region G1 are space-like to all the ones of the region G2 (G1 ∼ G2 ),
then the time order of the regions can always be changed by a corresponding Lorentz
transformation, that is:

S(g1 )S(g2 ) = S(g2 )S(g1 ) G1 ∼ G2 . (18.79)

The formulae (18.78) and (18.79) are the essence of the integral principle of the causality.
One may formulate the causality condition in a differential form as well. However, these
formulae are not a source of any additional information about the S-matrix and in sub-
sequent discussions we shall not use both the integral and differential formulations of the
causality condition. We refer to the reader taking an active interest in this question to the
book [5] where these questions are considered in detail.

18.4 Representation of the S-matrix as the sum of normal products

Since the operators ψ(x), ψ(x) and Aµ (x) represent the sum of the production and annihi-
lation operators of the corresponding particles, then every term of the S-matrix expansion
consists of the products sum of the production and annihilation operators of electrons,
positrons, and photons. Let us define conditions under which products of such a kind have
nonvanishing matrix elements for the process we are interested in |i >→ |f >. We assume
that one photon is present in the initial state |i >= |1γ >, and there are electrons and
positrons in the final state |f >= |1e− ; 1e+ >. As this takes place, in the appropriate
nonzero matrix element:
Si→f =< i|S|f > (18.80)

© 2011 by Taylor and Francis Group, LLC

 424 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

one of the annihilation operators should destroy a photon in the state |i >, two production
operators should create an electron and a positron in the state |f >, and all other oper-
ators should be in pairs, the operators in each pair generating and annihilating the same
particle. As a source of virtual particles, these paired operators make the calculation of
matrix elements for a scattering matrix too complicated. Fortunately, there is a method to
take into account similar virtual processes, relating the number of paired operators to the
perturbation degree in which a process under study is evaluated. The idea is fairly simple:
the scattering matrix should be represented as a sum over normal products of the particle
production and annihilation operators. Then in the process of calculations the annihilation
operators destroy only the particles in the initial states, whereas the production operators
create only the particles to be in the final states. So, our task is to exhibit the mixed
product as a sum of the N -products. It could be done employing two Wick’s theorems [6].
The first Wick’s theorem. The T -product of n linear operators is equal to a sum of
their N -products that includes all possible combinations of pairings and the term without
pairings, that is,

T (B1 B2 B3 B4 ...Bn ) = 

= N (B1 B2 B3 B4 ...Bn ) + (a) 

 

s1 s1 s1 s1 
N (B1 B2 B3 B4 ...Bn ) + N (B1 B2 B3 B4 ...Bn ) + 

s1 (b) 
+... + N (B1 B2 B3 B4 ...Bns1 ) + ... 



s1 s1 s2 s2 s1 s2 s1 s2
+N (B1 B2 B3 B4 ...Bn ) + N (B1 B2 B3 B4 ...Bn ) + (18.81)
(c)
+... + N (B1s1 B2 B3s2 B4s2 ...Bns1 ) + 



+...................................................................+ 



s1 s2 s2 sn/2 sn/2 s1
 
+N B1 B2 B3 ...Bk ...Bn−1 Bn +


s1 s1 s2 s n/2 s1 sn/2 (d) 

+... + N B1 B2 B3 ...Bk ...Bn−1 Bn + ...,

where: (a) is a normal product without any pairings, (b) is all possible normal products
with one pairing, (c) is all possible normal products with two pairings, (d) is a term with all
paired operators. Here, by linear operators are meant operators being linear combinations
of the production and annihilation operators. When Bose-field operators are paired, then
one may place them simply alongside each other. The Fermi-field operators being paired
may be placed alongside each other as well, however, in so doing, multiplication on the
value (−1)p , where p is a number of operator transpositions to have been carried out, must
be fulfilled. One may factor out pairings from the sign of the normal product.
Let us prove the first Wick’s theorem. If we perform in each term of the formula (18.81)
identical simultaneous permutations of the factors, the equality is not violated. Because of
this, without loss of generality, it can be assumed that the operators on the left and right
sides of (18.81) are already in chronological order. And the symbol of the T -product on
the left side of (18.81) may be omitted. Next, we arrange the operators so that all the
annihilation operators be positioned to the right of the production ones, that is, we perform
N -ordering of the expression (18.81). To this end, we sequentially rearrange the farthermost
left N -unordered production operator with all the annihilation operators that are situated
to the left of it. In the process some additional terms with pairing between the rearranged
operators according to formula:

AB = T (AB) = N (AB) + As1 B s1 = ±BA + As1 B s1 , (18.82)

where the sign + (−) is associated with Bose (Fermi) operators, will appear. A similar
ordering operation is performed for other unordered production operators too. As a result,
the initial T -product is represented as a sum over N -products. These N -products may
have the positive or negative sign. At the same time, rearranging the factors under the N -
product sign to make them T -ordered again, we obtain the positive sign for all N -products.

© 2011 by Taylor and Francis Group, LLC

 S-matrix 425

A sum of the derived N -products includes only pairings between the pairs of N -unordered
operators rather than all the pairings possible. Considering that pairings between the N -
ordered operators, being simultaneously the T -ordered ones, are equal to zero, we can state:
a sum of N -products in (18.81) involves all the pairings possible. Thus, the first Wick’s
theorem has been proved.
The second Wick’s theorem. A mixed product of the operators equals a sum of their
N -products, where the operators are related by all the pairings possible, except for the
pairings between the operators within one and the same N -product. The proof is found in
an analogy with the first Wick’s theorem. It should be remembered that there is no need
to interchange the positions of the operators under the same sign of N -product as these
operators are already N -ordered and hence their associated pairings are zero. Subsequently,
we will use the second Wick’s theorem that will be referred to simply as the Wick’s theorem.

18.5 Calculation of S-matrix elements

Obviously, we should above all choose the definite normalization of state vectors just as
we do in the scattering problems of quantum mechanics. The normalization on the volume
unit proves to be the most convenient. To find a state vector with such a normalization
we address the limiting process, namely, consider the sequence of state vectors with an
unlimitedly growing ordinary norm. First, we carry out this operation for the case of one
particle. Let us take the one-particle state vector in the view:
Z
Φ1 = d3 pφσ (p)a†σ (p)Φ0 , (18.83)

where the symbol σ is used to set the plurality of such quantum numbers as mass, spin
projection, and electric and lepton charges. It is clear that the norm of this state equals
Z
Φ†1 Φ1 = d3 p|φσ (p)|2 = N. (18.84)

Then, assuming N = 1, one may state that the expression:

|φσ (p)|2 d3 p

gives the probability that the particle characterized by the quantum number σ possesses
the momentum in the interval d3 p about the average value p. It, in its turn, means that the
function φσ (p) is a wavefunction in the momentum representation. Therefore, its Fourier
image: Z
1
ψσ (r) = d3 pφσ (p)eipr (18.85)
(2π)3/2
is nothing but the wavefunction in the coordinate representation. The norm of ψσ (r) appears
to be equal to: Z Z
d3 x|ψσ (r)|2 = d3 p|φσ (p)|2 = N. (18.86)

It is clear, at N = 1 the quantity dwx = |ψσ (r)|2 d3 x should be treated as the probability
of finding a particle with quantum numbers σ in an infinitesimal element of volume d3 x.
When N being equal to the particles number (N ≫ 1), dwx defines a mean number of

© 2011 by Taylor and Francis Group, LLC

 426 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

particles in the volume d3 x. Now, we shall unlimitedly increase the norm N in such a way
as to tend φσ (p) to the expression:

(2π)3/2 δ (3) (p − p0 ). (18.87)

As a result, we obtain the wave package with more and more decreasing spread of momentum
about the central value p0 . Then, from (18.85) we find:

|φσ (p)| → 1, (18.88)

and as a consequence:
|φσ (p)|2 → 1. (18.89)
The relation (18.89) means that in the limit the volume unit contains one particle. Passing
in (18.83) to the limit, we get the following expression for the one-particle state vector
normalized by the volume unit:

Φ1 = (2π)3/2 a†σ (p)Φ0 . (18.90)

In the case of particle k-kinds one should consider the expression:

Z Z
Φk = d p1 φσ1 (p1 )aσ1 ... d3 pk φσk (pk )a†σk Φ0 ,
3 †
(18.91)

rather than Eq. (18.85). The norm of such a state vector represents the k product of the
one-particle states norm:
k Z
Y k
Y
Φ†k Φk = 3 2
d pj |φσj (pj )| = Nj . (18.92)
j=1 j=1

It is obvious that the above-stated reasoning scheme could be transferred to every factor
in (18.91) and (18.92). Then the multiparticle state vector, which is normalized by the
volume unit for every particle being in this state, is determined by the expression:

Φk = (2π)3k/2 a†σ1 (p1 )a†σ2 (p2 )...a†σk (pk )Φ0 . (18.93)

After all these preparations we could pass at once to the determination of the S-matrix el-
ements. For this purpose we use Wick’s theorem. As an example, we consider the Compton-
effect on a free electron
e− (p) + γ(k) → e− (p′ ) + γ(k ′ ), (18.94)
where four-dimensional momenta of particles are denoted by letters in parentheses. The
final and initial state vectors normalized by the volume unit have the form:

|i >= Φi = (2π)3 a(e)†

r (p)a(γ)†
s (k)|0 >,

(e)† (γ)†
|f >= Φf = (2π)3 ar′ (p′ )as′ (k′ )|0 >,
where the production operators of the electrons and photons are supplied by the superscripts
e and γ, respectively. Thus, the matrix element is set down in the form:
!
X (1) (2) (3)
(n)
Si→f =< f | S |i >= I + Si→f + Si→f + Si→f + ... (18.95)
n

© 2011 by Taylor and Francis Group, LLC

 S-matrix 427

In virtue of (18.69) and Wick’s theorem, the expression for the matrix element in the first
order of the perturbation theory will look like:
Z
(1) 6 (e) (γ)

Si→f = −ie(2π) d4 x < 0|ar′ (p′ )as′ (k′ )N ψ(x)γ µ Aµ (x)ψ(x) a(e)†
r (p)a(γ)†
s (k)|0 > .
(18.96)
Taking into account the vacuum state vector normalization condition and commutativity
of the operators of the electron-positron and photon fields, one could report the expression
(18.96) as a product of two parts:
Z h i
(1) (e)

Si→f = −ie(2π)6 d4 x < 0|ar′ (p′ )N ψ(x)γ µ ψ(x) a(e)† r (p)|0 > ×
h i
(γ)
× < 0|as′ (k′ )Aµ (x)a(γ)†s (k)|0 > =
Z n
6 (e)
= −ie(2π) d4 x < 0|ar′ (p′ )[ψ (+) (x)γ µ ψ (+) (x) + ψ (−) (x)γ µ ψ (+) (x)+
o
+ψ (+) (x)γ µ ψ (−) (x) − ψ (−) (x)γ µT ψ (−) (x)]a(e)†
r (p)|0 > ×
h i
(γ)
× < 0|as′ (k′ )Aµ (x)a(γ)† s (k)|0 > , (18.97)
where we have allowed for both the electron-positron field operators expansion into the
positive and negative frequency parts and the relation:
(−) (−)
N (ψ (−) α (x)(γ µ )αβ ψβ (x)) = −ψβ (x)(γ µ )Tβα ψ (−) α (x) = −ψ (−) (x)γ µT ψ (−) (x).
Let us calculate the individual contributions from all the four terms entering into the
electron-positron part. Consider the first term. Since the relations:

ψ (+) (x)|0 >=< 0|ψ (+) (x) = 0,

take place, we should carry the operator ψ (+) (x) to the right until it acts on |0 > and the
operator ψ (+) (x) to the left up to the state < 0|. The calculation gets:
Z Z s
(e) ′ (+) µ (+) (e)† 1 XX 3 3 m2
< 0|ar′ (p )ψ (x)γ ψ (x)ar (p)|0 >= 3
d p1 d p2 ×
(2π) m n Ep1 Ep2
 
(e) (e)
[< 0| {ar′ (p′ ), a(e)†
m (p 1 )} − a (e)†
m (p )a
1 r ′ (p ′
) u(m) (p1 )eip1 x ]γ µ ×
 
×[u(n) (p2 )e−ip2 x {a(e)n (p 2 ), a (e)†
r (p)} − a (e)†
r (p)a (e)
n (p2 ) |0 >] =
s
1 m2 i(p′ −p)x (r′ ) ′ µ (r)
= e u (p )γ u (p), (18.98)
(2π)3 Ep Ep′
where we have used the expansions of the operators ψ(x) and ψ(x) into the Fourier integrals
(e) (e)†
and anticommutation relations for the operators an (p) and ar (p).
The similar manipulation leads to the conclusion that all the remaining terms in the
electron-positron factor give zero contribution.
Now we calculate the photon factor entering to (18.97). Using the expansion of Aµ into
the Fourier integral, we obtain
Z 3
(γ) 1 d3 k1 (γ)
X
< 0|as′ (k′ )Aµ (x)a(γ)†
s (k)|0 >= √ < 0|as′ (k′ ) ×
(2π)3/2 2k10 λ=0

© 2011 by Taylor and Francis Group, LLC

 428 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
(γ) (γ)†
×[eλµ (k1 )aλ (k1 )e−ik1 x + eλ∗
µ (k1 )aλ (k1 )eik1 x ]a(γ)†
s (k)|0 >=
Z 3
1 d3 k1 X n λ (γ)

(γ)
= √ eµ (k1 ) < 0|as′ (k′ )e−ik1 x [aλ (k1 ), a(γ)†
s (k)]+
(2π)3/2 2k10 λ=0
 
(γ) (γ) ′ (γ)†
+a(γ)†
s (k)aλ (k1 ) |0 > +eλ∗
µ (k1 ) < 0| [as′ (k ), aλ (k1 )]+
 o
(γ)† (γ)
+aλ (k1 )as′ (k′ ) eik1 x a(γ)†
s (k)|0 > = 0. (18.99)

Thus, the matrix element of the process (18.94) in the first order of the perturbation
theory becomes zero. To put this another way, in the first order of the perturbation theory
the Compton-effect does not proceed.
Now we pass to the definition of the matrix element of the process (18.94) in the second
order of the perturbation theory. According to Wick’s theorem we have
Z Z
(2) (−ie)2 

< f |Si→f |i >=< f | d4 x1 d4 x2 N ψ(x1 )γ µ ψ(x1 )ψ(x2 )γ ν ψ(x2 ) +
2
 s 
+N ψ (x1 )γ µ ψ(x1 )ψ(x2 )γ ν ψ s1 (x2 ) +
1

 s1

+N ψ(x1 )γ µ ψ s1 (x1 )ψ (x2 )γ ν ψ(x2 ) +

s2 s1
o
+ψ (x1 )γ µ ψ s1 (x1 )ψ (x2 )γ ν ψ s2 (x2 ) N (Aµ (x1 )Aν (x2 )) + Asµ1 (x1 )Asν1 (x2 ) |i > .
(18.100)
In the factor containing only electron-positron operators the second and third terms are
equivalent. To make sure in this we shall do the following operations: (i) rearrange the
operators ψ(x2 )γ ν ψ(x2 ) and ψ(x1 )γ µ ψ(x1 ) in the second term under the sign of the normal
product; (ii) carry out the replacement of the integration variables x1 ↔ x2 ; (iii) rename
the summation indices µ ↔ ν; and (iv) take into account the relation:

N (Aµ (x1 )Aν (x2 )) = N (Aν (x1 )Aµ (x2 )).

As a result we get the expression to coincide with the third term.

Using the expansion of the operators of the electron-positron field and the anticommuta-
tion relations one could show that the terms
s1 s2
N (ψ(x1 )γ µ ψ(x1 ) × ψ(x2 )γ ν ψ(x2 )) and ψ (x1 )γ µ ψ s2 (x1 )ψ (x2 )γ ν ψ s1 (x2 )

give zero contribution. The term Asµ1 (x1 )Asν1 (x2 ) entering into the photon factor yields no
contribution too. Thus, all that remains is to evaluate the term:
Z Z n  
(2) (e) s1
Si→f = −e2 (2π)6 d4 x1 d4 x2 < 0|ar′ (p′ )N ψ(x1 )γ µ ψ s1 (x1 )ψ (x2 )γ ν ψ(x2 ) ×

on o
(γ)
×a(e)†
r (p)|0 > < 0|as′ (k′ )N (Aµ (x1 )Aν (x2 ))a(γ)†
s (k)|0 > . (18.101)

From the calculation of the matrix element in the first order of the perturbation theory it fol-
s1
lows that the factor N (ψ(x1 )γ µ ψ s1 (x1 )ψ (x2 )γ ν ψ(x2 )) will produce only one nonvanishing
term:
s1
ψ (+) (x1 )γ µ ψ s1 (x1 )ψ (x2 )γ ν ψ (+) (x2 ).

© 2011 by Taylor and Francis Group, LLC

 S-matrix 429

To substitute the Fourier-expansions of the electron-positron field operators and the pairings
between the operators leads to the following expression for the first factor in (18.100):
s Z
i m2 (r′ ) ′ ip′ x1 µ p̂1 + m −ip1 (x1 −x2 ) ν (r)
u (p )e γ d4 p1 2 e γ u (p)e−ipx2 . (18.102)
(2π)7 Ep Ep′ p 1 − m2

Integration over the four-dimensional momentum is obliged by the chronological pairing of

operators. In the general case of the n-pairings of arbitrary particle operators n-multiple
integral over momenta of corresponding particles will appear.
Having taken advantage of the electromagnetic field operators expansion in terms of the
positive and negative frequency parts, we find the expression for the photon factor:

Aµ (x1 )Aν (x2 ) = A(+) (+) (+) (−) (−) (+)

µ (x1 )Aν (x2 ) + Aµ (x1 )Aν (x2 ) + Aµ (x1 )Aν (x2 )+

+A(−) (−)
µ (x1 )Aν (x2 ). (18.103)
(+) (+) (−) (−)
The operators Aµ (x1 )Aν (x2 ) and Aµ (x1 )Aν (x2 ) do not yield any contribution to
(±)
the matrix element. Allowing for the Fourier-integral expansion of Aµ (x) and the com-
mutation relations for the production and annihilation photon operators, we obtain the
contributions from the second and third terms in the photon factor (18.100):
1 1  ′ ′
(γ)
< 0|as′ (k′ )Aµ (x1 )Aν (x2 )a(γ)†
s (k)|0 >= 3
p esµ ∗ (k′ )eik x1 esν (k)e−ikx2 +
(2π) ′
2k0 k0
′ ′

+esν ∗ (k′ )eik x2 esµ (k)e−ikx1 . (18.104)

Insert (18.102) and (18.104) into (18.100). As a result we arrive at the expression main-
taining the coordinates x1 and x2 in the index of the exponent only. Then integration over
the variables x1 and x2 leads to the appearance of the δ-functions:
Z Z
′ ′
d4 x1 d4 x2 eix1 (p +k −p1 ) eix2 (p1 −p−k) = (2π)8 δ (4) (p′ + k ′ − p1 )δ (4) (p1 − p − k), (18.105)
Z Z
′ ′
4
d x1 d4 x2 eix1 (p −k−p1 ) eix2 (p1 −p+k ) = (2π)8 δ (4) (p′ − k − p1 )δ (4) (p1 − p + k ′ ), (18.106)

where the expression (18.105) arises at multiplication of the electron factor by the first term
in (18.104) while the one (18.106) does at multiplication of the electron factor with the sec-
ond term from (18.104). Every δ-function entering either into (18.105) or into (18.106)
ensures the four-dimensional momentum conservation in the diagram vertices x1 and x2 .
To integrate over the space-time coordinates means the passage to the momentum represen-
tation. Therefore, having substituted (18.102), (18.104) into (18.100) and having integrated
over the variables x1 and x2 , we come to the expression for the Compton-effect matrix ele-
ment in the second order of the perturbation theory in the momentum representation:
Z s 
(2) 2 4 4 m2 ′ p̂1 + m ν
Si→f = −ie (2π) d p1 u(r ) (p′ )γ µ δ (4) (p′ + k ′ − p1 ) 2 γ ×
4Ep Ep′ k0 k0′ p 1 − m2

′ ′ p̂1 + m ν
×δ (4) (p + k − p1 )u(r) (p)esµ ∗ (k′ )esν (k) + u(r ) (p′ )γ µ δ (4) (p′ − k − p1 ) γ ×
p21 − m2
′
o
×δ (4) (p − k ′ − p1 )u(r) (p)esν ∗ (k′ )esµ (k) =

© 2011 by Taylor and Francis Group, LLC

 430 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
s ( "
2 4 m2 (r ′ ) ′ p̂ + k̂ + m s
= −ie (2π) u (p ) esµ ∗ (k′ )γ µ
′
e (k)γ ν +
′
4Ep Ep′ k0 k0 (p + k)2 − m2 ν
# )
′
s µ p̂ − k̂ + m s′ ∗ ′ ν
+eµ (k)γ e (k )γ u (p) δ (4) (p + k − p′ − k ′ ).
(r)
(18.107)
(p′ − k)2 − m2 ν
In the obtained expression the remaining δ-function expresses the conservation law of the
four-dimensional momentum. Note that in the second order of the perturbation theory one
manages to get rid of the integration over the four-dimensional momentum but this is not
the case when n ≥ 3.
One may continue the process of finding the Compton-effect matrix elements in the
higher-orders of the perturbation theory. However, we have already achieved the main goal.
We have learned the process of obtaining the matrix elements of the scattering matrix on
the basis of general analytical expression for the matrix element in the given order of the
(n)
perturbation theory by application of the Wick’s theorem. The scheme of finding Si→f
is as follows: (i) the initial and final state vectors (|i > and |f >) are built; (ii) with
the help of the Wick’s theorem the scattering matrix S (n) is reduced to the normal form;
(iii) the matrix element of the process is found and the terms giving nonzero contributions
are selected; (iv) commutating the annihilation (production) operators entering to the S-
matrix with the production (annihilation) operators belonging to |i > (< f |) is fulfilled; (v)
integration over four-dimensional coordinates is carried out (it is equivalent to the passage
to the momentum representation); and (vi) to integrate over four-dimensional momenta of
virtual particles is accomplished.
The above-stated procedure may be shaped to a more convenient form that allows, not
addressing to Wick’s theorem each time, to find the S-matrix elements at once. To ac-
complish this it is necessary to formulate rules that put in correspondence to each normal
product a particular graph, a diagram.

18.6 Feynman rules in coordinate space

Let us set the correspondence rules between the scattering matrix elements and graphical
images. In the QED the analytical expression for the scattering matrix elements S (n)
represents a combination including functions corresponding to final and initial particles,
pairings of the electron-positron and electromagnetic field operators, and some number of
γ-matrices. It is clear that we ought to specify the graphical image of the above-mentioned
elements.
Let us agree to correlate four-dimensional vectors x1 , x2 , ...xn , on which the integration in
S (n) is accomplished, with diagram points (diagram vertices). As far as field operators are
concerned, they will be associated with lines passing through these vertices. On the strength
of the local character of the interaction Lagrangian (18.12), in every diagram vertex one
photon line, one incoming and one outgoing electron or positron lines come together. Since
the photon field operator Aµ (x) by its geometrical nature is the four-dimensional vector,
then a vertex is also characterized by a corresponding four-dimensional index µ. We agree
to correlate the vertex with the factor −ieγ µ . Obviously, the number of vertices coincides
with the order of the perturbation theory. The operator Aµ (x) will be displayed by the
waved line without a definite direction with the origin in the vertex x and the index µ; the
operator ψ(x)—by the directed solid line incoming to the vertex x; the operator ψ(x)—by
the directed solid line outgoing from the vertex x. All these lines are called the external

© 2011 by Taylor and Francis Group, LLC

 S-matrix 431

ones and they describe real particles. Since Aµ (x) represents the sum of the production and
annihilation operators for photons, then a waved external line displays a photon emitted or
absorbed as a result of the scattering process. In a similar way, because ψ(x) is the sum of
the electron annihilation operator and the positron production operator, then the solid line
directed to a vertex will denote an electron existing before scattering or a positron to come
into being as a result of the scattering process. As ψ(x) comprises the sum of the electron
production operator and the positron annihilation operator, then a solid line outgoing from
a vertex will display either a positron in a initial state or an electron in a final state.
Let us agree to confront pairings between operators with internal lines of a diagram and
consider that these lines correspond to the propagation of virtual particles. The pairing of
the photon operators:
Aµ (x)s Asν (y) = igµν D0c (x − y)
will be depicted by a nondirectional waved line connecting the vertices x and y in which
the matrices γ µ and γ ν are operating, respectively. As this takes place, we consider this
line as a display of photon motion between the points x and y. Nonsymmetric pairing:
s
ψ s (x)ψ (y) = −iS c(x − y)

is confronted with a solid diagram line to connect the vertices x and y. From the meaning
of the operators ψ(x) and ψ(y) it follows that this line could be interpreted as an electron
production in the point y and its subsequent annihilation in the point x (and vice versa for
a positron). Therefore, it is expedient to choose the direction of this line from the point y
to the point x.
In the case of an interaction with an external (classical) electromagnetic field the La-
grangian contains an additional term:

Leint = −e : ψ k (x) (γ µ )kl ψl (x) : A(e)

µ (x). (18.108)

Interaction with an external field will be displayed by a waved line to go out of a diagram
vertex and come to an end in a shaded circle that symbolically denotes the scattering center.
(e)
In the scattering matrix element the factor Aµ (x) (µ is a vertex index) is associated with
this line. Let us consider, for example, electron scattering by an external field in the
first order of the perturbation theory. The corresponding matrix element is given by the
expression:
Z s
4 m2 ′
Si→f = −ie d x ur′ (p′ )Aeµ (x)γ µ ur (p)ei(p −p)x . (18.109)
Ep Ep′

If we substitute the Fourier expansion of an external potential:

Z
1
Aeµ (x) = d4 qAeµ (q)e−iqx , (18.110)
(2π)3/2

into (18.109), then, having integrated over x, we get

s Z
3/2 m2
Si→f = −ie(2π) d4 qur′ (p′ )Aeµ (q)γ µ ur (p)δ(p + q − p′ ). (18.111)
Ep Ep′

The presence of the δ-function leads to equality q = p′ − p. Since q 2 = 0 for a photon

and the square (p′ − p)2 certainly differs from zero, then such a equality for real particles is
impossible in principle. This suggests that an external field may be replaced by the effective

© 2011 by Taylor and Francis Group, LLC

 432 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

virtual photon exchanged by the interacting particle with a source of the external field. It
is important that such an interpretation for interaction with an external electromagnetic
field is in line with physics of the phenomenon. An external field differs from a quantized
one only in that it is fixed, its state remaining unchanged in the process of interactions. An
external field may be represented as a limit of the average value for some quantized field
when its intensity is infinitely increasing. Then interaction with this field and in the limit is
realized by the virtual photons, for which the process of propagation in the limiting case is
described by the given potential Aeµ (x) and not by the causal photon function D0c (xc − xd ),
where xc (xd ) are the coordinates of the photon production (destruction) point.
The above correspondence rules between the factors of the S (n) -matrix element and Feyn-
man diagrams are known as the Feynman rules in the coordinate representation [7]. Using
these rules and constructing all possible diagrams for the process under study in the partic-
ular order of the perturbation theory, one may obtain an analytical expression for S (n) , and
(n)
then go to the calculation of Si→f . But there are a few aspects to be taken into consideration.
First, a situation is possible when several normal products are distinguished only by the
places of imposing couplings between the field operators and make identical contributions to
the scattering matrix element. We shall designate them as the equivalent normal products.
The number of the equivalent N -products associated with given diagram is found as follows.
Vertices of the diagram are labeled with numbers 1,2,..n. Then we count the permutations
number for these numbers, leaving the diagram form unaltered. Provided the count gives
the value g, the number of the equivalent N -products r is determined by the relation:

r = n!/g. (18.112)

Second, each normal product associated with a particular physical process may be rep-
resented as a sum of several addends, including the products of the production and an-
nihilation operators of the particles, which take part in the process studied, and being
distinguished only by the arrangement order of the operators. These addends are repre-
sented as the diagrams differing only by the positions of electron, positron, and photon
lines. Such diagrams are known as the topologically nonequivalent diagrams (TND).
To illustrate, let us consider the Compton-effect in the second-order of the perturbation
theory. As follows from (18.107), a normal product is divided into two parts associated
with two diagrams, which differ from each other only by the positioning of the photon
lines representing the initial and final photons. When the process involves n photons,
after the normal product is divided into the addends including the photon production and
annihilation operators, we have n! addends corresponding to n! TND.
Similarly, for the process involving several electrons and positrons its normal product is
represented as a sum over the terms including the same electron and positron production
and annihilation operators, being distinguished only by the arrangement order of these
operators. For example, two TND displayed in Fig. 18.2, where p1,2 (p′1,2 ) are the four-
dimensional momenta of the initial (final) electrons, k and k ′ are the four-dimensional
momenta of the virtual photons, corresponds to the process:

e− + e− → e− + e− (18.113)

As distinct from the processes involving several photons, for which TND are associated
with the matrix elements of the same sign, for the processes involving several electrons
(positrons), TND are associated with the matrix elements that may have opposite signs.
This is due to the fact that the operators of an electron-positron field satisfy the anticom-
mutation relations.

© 2011 by Taylor and Francis Group, LLC

 S-matrix 433

(a) (b)

FIGURE 18.2
The Feynman diagrams (a) and (b) corresponding to the process e− + e− → e− + e− .

Next, we show that matrix elements of TND (TND).for the process (18.113) have the
opposite signs. The corresponding matrix element in the second order of the perturbation
theory has the form:
Z Z
(2) −ie2 (2π)6 (e) (e)
Si→f = < 0|ar′ (p′1 )ar′ (p′2 )[ dx1 dx2 gµν D0c (x1 − x2 )×
2 1 2

×N ψ(x1 )γ µ ψ(x1 )ψ(x2 )γ ν ψ(x2 ) ]a(e)† (e)†
r2 (p2 )ar1 (p1 )|0 > . (18.114)
Substituting the Fourier-series expansions of the electron field operators and allowing for
the anticommutation relations, we arrive at:
s Z Z
(2) −ie2 m4
Si→f = dx1 dx2 gµν D0c (x1 − x2 )×
2 Ep1 Ep′1 Ep2 Ep′2
n ′ ′ ′ ′
× v (r2 ) (p′2 )γ µ v (r2 ) (p2 )v (r1 ) (p′1 )γ ν v (r1 ) (p1 )ei(p2 −p2 )x1 +i(p1 −p1 )x2 +
′ ′ ′ ′
+v (r1 ) (p′1 )γ µ v (r1 ) (p1 )v (r2 ) (p′2 )γ ν v (r2 ) (p2 )ei(p2 −p2 )x2 +i(p1 −p1 )x1 −
′ ′ ′ ′
−v (r1 ) (p′1 )γ µ v (r2 ) (p2 )v (r2 ) (p′2 )γ ν v (r1 ) (p1 )ei(p1 −p2 )x1 +i(p2 −p1 )x2 −
′ ′ ′ ′
o
−v (r2 ) (p′2 )γ µ v (r1 ) (p1 )v (r1 ) (p′1 )γ ν v (r2 ) (p2 )ei(p2 −p1 )x1 +i(p1 −p2 )x2 . (18.115)

Note that the common sign of this sum is conditional and is defined by the assignment
way of the initial and final states (or, what is the same, by the arrangement order of
external electron operators). So, for example, there is nothing to prevent us to take |i >=
(e)† (e)† (e)† (e)†
(2π)3 ar1 (p1 )ar2 (p2 )|0 >) instead of |i >= (2π)3 ar2 (p2 )ar1 (p1 )|0 >. Since the matrix
element enters as the square modulo into the cross section to measure in experiment, then
the common sign of the matrix element for scattering of identical fermions is arbitrary at
all. However, the relative sign of the addends in (18.115) does not depend on the accepted
order of the external operators arrangement.
The first and second addends in (18.115) are different from each other by simultaneous
rearrangement of the indices µ, ν and the arguments x1 , x2 . Because both these addends
are multiplied by the expression that is symmetric both regarding µ, ν and regarding x1 , x2 ,

© 2011 by Taylor and Francis Group, LLC

 434 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

then they are equal to each other and correspond to the diagram displayed on Fig. 18.2a.
The similar way one demonstrates that the third and forth addends which, in turn, are
associated with the diagram of Fig. 18.2(b), are equivalent. Thus, as claimed, the matrix
elements corresponding to both diagrams have the opposite signs. In other words, in the
case of the presence of identical fermions in the initial or final states, the expression for the
matrix element must be antisymmetric with respect to identical fermions (in the case of
bosons the matrix elements should be symmetric in relation to rearrangement of identical
bosons).
Generalize this result on the case when l identical fermions (electrons or positrons) par-
ticipate in scattering. Let us denote momenta of particles before and after scattering by
p1 , p2 , ...pl and p′1 , p′2 , ...p′l , respectively. l TNDs correspond to the process. If on some
two diagrams fermion lines after scattering (which are continuation of fermion lines before
scattering) are designated as p′1 , p′2 , ... and p′j1 , p′j2 , ... respectively, then the relative sign of
the matrix elements conforming to these diagrams is defined by the transposition parity
 
1, 2, ..., l
δp = . (18.116)
j1 , j2 ..., jl

Since matrices enter into the interaction Lagrangian:

Lint = −e : ψ k (x) (γ µ )kl ψl (x)Aµ (x) :, (18.117)

(n)
(k and l are matrix indices), then the order of factors in Si→f becomes to be significant as
well. From the expression (18.115) we start to guess that the order of matrices placement
correlates with the motion along an electron line of a diagram from the final to initial states
(in the case of positrons the motion occurs in inverse direction). To ultimately become
stronger in this statement, we consider a more complicated diagram with a continuous
electron line. The diagram of the third order displayed on Fig. 18.3 to describe the process
(e)
of the electron scattering by the external field Aµ (x) is a good candidate for this purpose.
Note that in the QED the anomalous magnetic moment of leptons is evaluated by application

q = p1 − p 2

p1 − k p2 − k

p1 p2

FIGURE 18.3
The process of the electron scattering by the external field.

of just these very diagrams.

© 2011 by Taylor and Francis Group, LLC

 S-matrix 435
(e)
The normal product corresponding to the electron scattering by the field Aµ (x) has the
form:  s s2
A = N ψ m (x1 )(γ µ )mn ψn (x1 )Asµ3 (x1 )ψ r (x2 )(γ ν )rj ψjs1 (x2 )×
1


s2
×A(e) τ s3
ν (x2 )ψ k (x3 )(γ )kl ψl (x3 )Aτ (x3 ) . (18.118)
Once the matrix indices to determine the matrices multiplication order have been written
out, we have the right to move matrices. Using the pairing definition for the electron and
photon operators, we obtain the final expression for (18.118):
A = −iψk (x3 )(γ τ )kl Slr
c
(x3 − x2 )(γ ν )rj Sjm
c
(x2 − x1 )(γ µ )mn ψn (x1 )gµτ D0c (x1 − x3 )A(e)
ν (x2 ),
(18.119)
that is, the matrices placement really corresponds to the motion against the electron line
direction. The correspondence rules must be supplemented by one more rule, the signs
rule, which takes into account the specificity of diagrams containing even numbers of closed
fermion loops. As an example, we shall deal with a diagram with one electron loop. The
factor in the normal product corresponding to this loop proves to contain the minus sign:

TABLE 18.1
The Feynman rules for the QED.
No Factor in Element in Feynman State
scattering matrix diagram
1 Operator ψ(x) under External solid Electron (positron)
sign of normal product line incoming in in initial (final) state
vertex x
2 Operator ψ(x) under External solid Electron (positron)
sign of normal product line outgoing from in final (initial) state
vertex x
3 Operator Aµ (x) under External waved Photon in initial
sign of normal product line at vertex x or final state
having index µ
4 Quantity Aeµ (x) External waved External
line connecting electromagnetic field
shaded circle
with vertex x
having index µ
5 Quantity −ieγ µ Vertex with index Interaction act
µ
6 Pairing of electron- Internal solid line Motion of virtual
positron field operators electron (positron) from
−iS c (x − y) x to y (from y to x)
7 Pairing of photon field Internal waved Motion of virtual
operators line photon between
igµν D0c (x − y) vertices x and y
with indices µ and ν

 s s2

N ψ m (x1 )(γ µ )mn ψns2 (x1 )ψ r (x2 )(γ ν )rj ψjs1 (x2 ) = −(γ µ )mn Snr
c
(x1 −x2 )(γ ν )rj Sjm
c
1
(x2 −x1 ) =

= −Sp (γ µ S c (x1 − x2 )γ ν S c (x2 − x1 )) .

© 2011 by Taylor and Francis Group, LLC

 436 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the most general case the signs rule is formulated as follows: if a diagram contains l
closed fermion loops, then a matrix element must be multiplied by the factor (−1)l .
So, the matrix element conforming to the definite physical process could be presented in
the form: X (n)
(n)
Si→f = Mi→f , (18.120)
(n)
where Mi→f ≡< f |M(n) |i > and individual addends in the right-hand side differ from
each other by the placement order of the operators creating and annihilating the particles
to participate in the process. This expression may be obtained by two ways: from the
general analytical expression for the matrix element by application of Wick’s theorem, or
with the help of the Feynman diagrams. To accomplish this operation within the diagram
technique by Feynman we should first draw all the possible topologically different diagrams
of a process in the given order of the perturbation theory. Next, to determine the form
of M(n) we must correlate every diagram element with a factor in the scattering matrix
in accordance with the Feynman rules gathered in Table 18.1 and multiply the obtained
expression by the quantity:
η = (−1)l rδp . (18.121)
(n)
Passing from M(n) to Mi→f and using the formula (18.120), we deduce the matrix element
under consideration. There is no doubt that this way is more simple and more convenient
than the method based on permanent usage of the Wick’s theorem.

18.7 Furry’s theorem

We shall prove the following theorem belonging to W. Furry [8]: a total matrix element which
corresponds to diagrams containing closed internal fermion loops to consist of odd number
lines turns into zero. To prove it, we will look at a process describing the contribution of
the third order in an amplitude of a transition vacuum→vacuum under the influence of an
external field. There are two diagrams differing only in a detour direction of a closed electron
(3)
loop (Fig. 18.4). Thus, the matrix element of the reaction in question MV includes two

x1 x1

x3 x2 x3 x2
(a) (b)

FIGURE 18.4
The Feynman diagrams (a) and (b) describing the vacuum→vacuum-transition under the
external field influence.

© 2011 by Taylor and Francis Group, LLC

 S-matrix 437
(3) (3)
addends Ma and Mb to accord with these diagrams:
Z Z Z
(3) (3) i
MV = M(3)
a +Mb = − (−ie)3 dx1 dx2 dx3 {Sp[Âe (x1 )S c (x1 −x2 )Âe (x2 )S c (x2 −x3 )×
3!

×Âe (x3 )S c (x3 −x1 )]+Sp[Âe (x1 )S c (x1 −x3 )Âe (x3 )S c (x3 −x2 )Âe (x2 )S c (x2 −x1 )]}, (18.122)
where Âe (x) = Aeµ (x)γ µ .
(3)
Using the charge conjugation matrix C we rewrite the spur to enter into Ma in the
following view:

Sp[...] = Sp[C −1 Âe (x1 )CC −1 S c (x1 − x2 )CC −1 Âe (x2 )CC −1 S c (x2 − x3 )×

×CC −1 Âe (x3 )CC −1 S c (x3 − x1 )C]. (18.123)

From the charge conjugation matrix property:

C −1 γ µ C = −γ µT ,

follows Z
1 −p̂T + m −ip(x1 −x2 )
C −1 S c (x1 − x2 )C = − d4 p e =
(2π)4 p 2 − m2
Z
1 p̂T + m −ip(x2 −x1 )
=− d4 p e = S cT (x2 − x1 ). (18.124)
(2π)4 p 2 − m2
This, in its turn, allows to set down (18.123) as:

Sp[...] = (−1)3 Sp[ÂeT (x1 )S cT (x2 − x1 )ÂeT (x2 )S cT (x3 − x2 )ÂeT (x3 )S cT (x1 − x3 )] =

= −Sp[S c (x1 − x3 )Âe (x3 )S c (x3 − x2 )Âe (x2 )S c (x2 − x1 )Âe (x1 )] =
= −Sp[Âe (x1 )S c (x1 − x3 )Âe (x3 )S c (x3 − x2 )Âe (x2 )S c (x2 − x1 )]. (18.125)
(3) (3) (3)
From (18.125) it is evident that Ma = −Mb , and, as a consequence, MV = 0.
Let us interpret the obtained result. It may be assumed that the first term associated with
the diagram in Fig. 18.4(a) describes propagation of an electron in one direction, whereas
the second term associated with the diagram in Fig. 18.4(b) defines propagation of an
electron in the reverse direction. Upon the reversal of its direction, an electron behaves like
a positron. Because of this, its charge is changing the sign, and the sign of each interaction
with an external electromagnetic field is also changed. Since the vertices number is odd,
both diagrams compensate each other. In a similar manner, if there is a diagram that
includes an internal closed electron loop with an arbitrary number of electron lines, along
with this diagram, we have to consider another diagram that is distinguished from the first
one only by the direction of a detour around the closed electron loop. And in case the
number of vertices is odd the matrix element corresponding to the sum of these diagrams
is zero.
Note that using a change of variables, we can transform the right-hand side of equality
(18.125) to the initial form—the left-hand side of equality (18.123). In the process the term
under consideration proves to be equal to itself with a minus sign, that is, equals zero. The
same is true for the second term in (18.122). In this way each individual term proves to be
zero, as one would expect based on the charge parity conservation law. A closed electron
loop with k vertices, that is, with k adjoined photon lines, is associated with a neutral
system of real and virtual photons, the total number of which equals k. As the photon
parity is equal to −1, for odd k the initial and final states possess a different charge parity,
and this is impossible as the experiment goes.

© 2011 by Taylor and Francis Group, LLC

 438 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

18.8 Feynman rules in the momentum representation

(n)
It is more convenient to calculate the matrix elements Si→f in the momentum representation
for the following reasons. First, in this representation the operators and the causal functions
have extremely simple structures. Second, in the coordinate representation, with the use of
the diagram technique, one finds the expression for an element of the scattering matrix S (n)
(n)
and only afterwards the corresponding matrix element Si→f may be calculated, whereas
(n)
in the momentum representation Si→f may be directly written due to the correspondence
rules. To find the matrix element in the momentum representation we sufficiently carry
out three operations in the scattering matrix S (n) written in the coordinate representation:
(n)
(i) to pass on to the matrix element Si→f ; (ii) to decompose the operators and the causal
functions in terms of the Fourier integrals; and (iii) to integrate the obtained expressions
over the space-time coordinates corresponding to the diagram vertices.
(n)
Let us fulfill these operations step-by-step. The matrix element Si→f may be represented
in the form:
Z Z Z
(n) (−ie)n
Si→f = dx1 dx2 ... dxn K...µj ...µl ... (x1 , ...xn )×
n!



× < f |N ...ψ(xj )γ µj ...γ µl ψ(xl )... N ...Aµj−1 (xj−1 )...Aµl+1 (xl+1 )... |i >, (18.126)
where K...µj ...µl ... (x1 , ...xn ) comprises the product of the field operator pairings. As we have
already known to calculate the expression:



< f |N ...ψ(xj )γ µj ...γ µl ψ(xl )... N ...Aµj−1 (xj−1 )...Aµl+1 (xl+1 )... |i > (18.127)

we should commute the production operators from the normal products (PON) with the
annihilation operators involved in the states < f |, and also the annihilation operators from
the normal products (AON) with the production operators from |i > until one of PON
affects < 0| or one of AON turns to be alongside |0 >, that gives zero as a result. The
matrix element of (18.127) is nonzero only in the case when for each operator from the
normal product we can find an operator in < f | or |i >, a commutation with which results
in the δ -function, that is, in the pairwise “cancelation” of both operators. In this way
the expression (18.127) is nonzero when a sum of the particle numbers in each field, in the
initial and final states, is precisely equal to the number of the operator functions for this
field in the N -product. It should be noted that expression (18.127) proves to be nonzero
when < f | and |i >, apart from the operators “canceling” the normal-product operators,
include “redundant” operators that are canceled pairwise. For this purpose it is necessary
that the momenta of the “redundant” particles be equal in the < f | and |i > states. In this
case the total number of particles in the initial and final states are exceeding the number
of the operators in the N -product by some even number. But this is possible only if for
whatever reasons the indicated “redundant” particles do not take place in the interaction.
Because of this, such a situation is beyond our consideration. And in the experimental
conditions the particles with an invariable momentum are considered as belonging to the
principal initial beam, and only those changing their momenta are regarded as scattered.
So, the matrix element (18.127) is exhibited in the form of the product of the commutation
(e)†
results. In this case, the result of commuting ψ(x) with (2π)3/2 ar (p), which is equal to:
r
m (r)
u (p)e−ipx (18.128)
Ep

© 2011 by Taylor and Francis Group, LLC

 S-matrix 439

corresponds to the electron in the initial state with the momentum p and polarization r,
(e)
the result of commuting ψ(x) with (2π)3/2 ar (p):
r
m (r)
u (p)eipx (18.129)
Ep

—to the electron in the final state with the momentum p and polarization r, the result of
(e)†
commuting ψ(x) with (2π)3/2 br (p):
r
m (r)
v (p)e−ipx , (18.130)
Ep

—to the positron in the initial state with the momentum p and polarization r, the result
(e)
of commuting ψ(x) with (2π)3/2 br (p):
r
m (r)
v (p)eipx . (18.131)
Ep

—to the positron in the final state with the momentum p and polarization r.
Note that in the case of the positron an incoming (outgoing) line has the arrow directed
from a vertex (toward a vertex) and the momentum to point against the arrow. Analogously,
we have the factor: r r 
1 λ −ikx 1 λ∗ ikx
e (k)e e (k)e (18.132)
2k0 µ 2k0 µ
for the photon in a initial (final) state with the momentum k and polarization λ at a vertex
with an index µ.
So, the negative-frequency exponent exp(−ipx) always corresponds to particles incoming
into a point x with the momentum p while the factor exp(ipx) is associated with particles
outgoing from a point x with the momentum p.
Substituting the integral representations of the field operator pairings into (18.126), we
see that the integration over variables x1 , x2 , ..., xn is reduced to the integrals of the form:
Z !
P X
ixl n pn 4 (4)
dxl e = (2π) δ pn ,
n

that is, this integration leads to the appearance of the δ-functions at every vertex. To sum
up the aforesaid, we come to the Feynman rules in a momentum representation.
To calculate a matrix element of a process i → f one should choose the time direction
(recall, our choice is from left to right) and picture all topologically nonequivalent diagrams
(diagrams distinguished only by vertex indices rearrangement are not considered to be
different) in accordance with the form of the interaction Hamiltonian in the QED, that
is, in every vertex one photon and two fermion lines must come across. In the vertices
of the obtained diagrams one should arrange four-dimensional indices (µ, ν...). On lines
one should place momentum variables, that is, every external and internal line is assigned
its four-dimensional momentum. It might be well to point out that though a photon line
usually has no direction but once momentum has been assigned to it, then this line gains
the momentum direction, inasmuch as the momentum comes out from one vertex and enters
(n)
into the another. Then, one may get a matrix element Mi→f conforming to any diagrams
of nth-order using the following rules.
1) The factors: r
m (r)
u (p),
Ep

© 2011 by Taylor and Francis Group, LLC

 440 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
r
m (r)
v (p).
Ep
correspond to external solid lines to go out of the vertices in the positive and negative
direction of the time axis, respectively. The first factor describes the electron production
with the momentum p and polarization r while the second one does the positron annihilation
with the momentum p and polarization r.
2) The factors: r
m (r)
u (p),
Ep
r
m (r)
v (p).
Ep
correspond to external solid lines to go into vertices in the positive and negative direction
of the time axis respectively. The first factor describes the electron annihilation with the
momentum p and polarization r while the second one does the positron production with
the momentum p and polarization r.
3) The factor to describe the production or annihilation of a photon with the momentum
k and polarization λ: r
1 (λ)
e (k)
2k0 µ
corresponds to an external waved line at a vertex with an index µ.
4) The factor to describe either virtual electron propagation in the diagram arrow direc-
tion or virtual positron propagation in the opposite direction:
i p̂ + m
,
(2π)4 p2 − m2
corresponds to an internal solid line.
5) The factor to describe virtual photon propagation between vertices with indices µ and
ν:
−i gµν
,
(2π)4 k 2
corresponds to an internal waved line.
6) The factor to describe interaction between particles:

−ieγ µ (2π)4 δ (4) (p − p′ ± k),

(p and p′ are four-dimensional momenta of external or internal fermion lines, k is four-

dimensional momentum of an external or internal photon line, and the δ-function expresses
the four-dimensional momentum conservation law for all particles meeting at this vertex)
corresponds to every vertex.
7) All matrices acting upon spinor indices are placed in that order as they occur at the
diagram detour along an electron line from a final state to an initial one.
8) To integrate over four-dimensional momenta of internal lines displaying virtual particles
is fulfilled.
9) A matrix element is multiplied by the numerical factor:
r
(−1)l δp .
n!
(n)
In the ending we should carry out the summation over matrix elements Mi→f to match
all diagrams of a process under investigation. As far as an interaction with a external

© 2011 by Taylor and Francis Group, LLC

 S-matrix 441

electromagnetic field is concerned, then, in the most general case, for it to be covered the
Feynman rules in the coordinate space should be used. Feynman rules for a static external
field in the momentum representation will be viewed in Section 18.11.
At the absence of an interaction, when both four-dimensional momenta and quantum
numbers of particles in an initial state are equal to those of particles in a final state, the
scattering matrix S evidently coincides with the identity matrix I. It is convenient to
separate the identity matrix. To do it we report the scattering matrix elements in the form:
Si→f = δif + iRi→f , (18.133)
where Ri→f ≡< f |R|i > are matrix elements of the so-called reaction matrix R. Beyond
doubt, a phase factor in front of R is arbitrary. The choice made by us in (18.133) fits to
that of the phase factor in nonrelativistic quantum scattering theory. In nondiagonal matrix
elements the first term in (18.133) falls out, and thus matrix elements of the matrices S
and R for a transition i → f are connected with each other by a simple relation:
Si→f = iRi→f . (18.134)
In this case we have
X X X
Ri→f = −i(2π)4 δ (4) ( pi − pf ) Mi→f . (18.135)
i f
P
We shall call the quantity Mi→f , which enters into (18.135) by the reaction ampli-
tude Ai→f . In the case of an interaction with an external field not all components of
four-dimensional momentum will be conserved quantities (for example, only the forth com-
ponent, that is, a system energy, could be conserved). This, in its turn, leads to changing
a dimensionality of the δ-function in (18.135).
The rules obtained above could easily be generalized on the case of other point particles
with the spin 1/2—the charged leptons µ, τ and the quarks uα , dα , sα , cα , bα , tα . In so
doing to modify the Feynman rules resides in changing both a charge and a propagator of
an electron with those of particles in question:
i p̂ + m i p̂ + mf
e → q, → , (18.136)
4 2
(2π) p − m 2 (2π) p − m2f + imf Γf
4 2

where Γf is decay width of a particle with a flavor f .

18.9 Crossing symmetry

To make use of the diagram technique for the determination of scattering amplitudes, Ai→f ,
renders existence of a definite symmetry evident. This symmetry will be called the crossing
symmetry.
We shall concern ourselves with binary processes with the participation of particles a, b, c,
and d. Denote the particles motion direction by arrows and choose the time axis direction
from left to right. In this case the following reactions are possible:∗
a + b → c + d, (18.137)

∗ Recall,that one may handle reaction equations as one does with ordinary equations, with one exception
only. Now, at passage from one side of an equation to another we have the changes: particle→antiparticle,
antiparticle→particle.

© 2011 by Taylor and Francis Group, LLC

 442 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

a + c → b + d, (18.138)
a + d → c + b. (18.139)
and the inverse reactions as well.
The reactions (18.137)–(18.139) represent three different channels of some reaction (let
us call it the generalized reaction). The definite linkage between amplitudes of these chan-
nels proves to exist. This linkage could be established by means of introducing invariant
variables.
In the case of any scattering of the kind (18.137) we have two independent variables, for
example, an incoming particle energy and a scattering angle. Exhibit the process amplitude,
and the cross section itself, as a function of variables to be invariant with respect to the
Lorentz transformation. So, there are four four-dimensional momenta of particles at our
disposal. Therefore, all that could be done is to build of them scalar products:

(pa pb ), (pa pc ), (pa pd ), (pb pc ), (pb pd ), (pc pd ). (18.140)

Owing to the four-dimensional momentum conservation law

pa + pb = pc + pd (18.141)

and the relations p2j = m2j (j = a, b, c, d), only two of six scalar products in (18.140) are in-
dependent. It is commonly accepted to not use these independent scalars but Mandelstam’s
variables to connect with them:

s = (pa + pb )2 , t = (pa − pc )2 , u = (pa − pd )2 . (18.142)

It is an easy matter to verify that these variables satisfy the relation:

s + t + u = m2a + m2b + m2c + m2d . (18.143)

Set down the four-dimensional momentum conservation law and the Mandelstam variables
definition for the remaining two channels (18.138), (18.139)

pa + pc = pb + pd ,
(18.144)
s = (pa − pb )2 , t = (pa + pc )2 , u = (pa − pd )2 ,

pa + pd = pb + pc ,
(18.145)
s = (pa − pb )2 , t = (pa − pc )2 , u = (pa + pd )2 .
When one fulfills the change:

pa → pa , pb → pc , pd → pb , pc → pd (18.146)

into (18.141), (18.142) and simultaneously passes to new designations:

s → t, t → u, u → s, (18.147)

then one leads to the expressions (18.144) for the second channel. Analogously, carrying
out the replacement:

pa → pa , pb → pd , pc → pb , pd → pc (18.148)

and passing to new designations:

s → u, t → s, u → t, (18.149)

© 2011 by Taylor and Francis Group, LLC

 S-matrix 443

we get the expressions (18.145) for the third channel.

Thus, to describe all the three channels one may use the same variables. To put it in
another way one may take any of the three reactions as an original one and thereafter passes
from it to others. It is convenient to fix for every channel a definite title. One may easily
be convinced that in the center-of-mass frame the quantities:

s = (Ea + Eb )2 , t = (Ea + Eb )2 u = (Ea + Ed )2

are total energy squares in the first, second, and third channels, respectively. For this reason
the first channel is named the s-channel, the second—the u-channel, and the third—the t-
channel. Now, according to accepted terminology, we denote the first process amplitude as
As (s, t, u) and then keep the previous designations for the Mandelstam variables. As far as
the remaining channels are concerned we rewrite their characteristics in the following view:

At (st , tt , u), st = (pa + pc )2 , tt = (pa − pb )2 , u = (pa − pd )2 (18.150)

and

Au (su , t, uu ), su = (pb + pc )2 , t = (pa − pc )2 , uu = (pa − pb )2 . (18.151)

Now the crossing symmetry may be expressed with the help of the relations:

As (s, t, u) = At (st = t, tt = s, u) = Au (su = u, t, uu = s). (18.152)

Thanks to the crossing symmetry all three channels may be pictured by one generalized
diagram (see, Fig. 18.5). Let us show that physical regions of changing the variables for

a c
u

t
b d

FIGURE 18.5
The s-, u-, and t-channels of the binary process.

the three processes under consideration are not mutually overlapped. To map these regions
we employ a scalene coordinate system (Fig. 18.6) to be symmetric with respect to the
variables s, u, t (the triangle is equilateral). Every point on this plane called Mandelstam’s
plane [9] is defined by three coordinates to be equal to the distance of the point up to the
sides (or their prolongations) of the triangle abc. The positive direction for reading s, u,

© 2011 by Taylor and Francis Group, LLC

 444 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

u s

b
u

h
t
t

a c

FIGURE 18.6
Mandelstam’s plane.

and t is pointed by arrows guided inside the triangle. The triangle height has been chosen
equal to: X
h=s+u+t= m2i .
i=a,b,c,d

Values that may be responsible for a physical scattering process must satisfy definite
conditions being consequences of the four-dimensional momentum conservation law and the
fact that momenta of initial and final particles lay on the mass shell p2i = m2i .
As the product of two four-dimensional momenta obeys an inequality:

(pa pb ) ≥ ma mb , (18.153)

then

(pa +pb )2 ≥ (ma +mb )2 , (pa −pc )2 ≤ (ma −mc )2 , (pa −pd )2 ≤ (ma −md )2 . (18.154)

From (18.154) it follows that the following inequalities take place for the s-channel:

(ma + mb )2 ≤ s ≥ (mc + md )2 , 
(ma − mc )2 ≥ t ≤ (md − mb )2 , (18.155)
(ma − md )2 ≥ u ≤ (mc − mb )2 .


The similar inequalities could be obtained for the u- and t-channels.

To find remaining conditions we, following Ref. [10], introduce a four-dimensional vector:

Kµ = εµνλσ pνa pλb pσc . (18.156)

In the rest frame of one of the particles, say, the particle a (pa = (ma , 0, 0, 0)) the vector
Kµ has only space components Ki = ma εi0kl pkb plc , and, consequently, Ki2 ≤ 0, that is, the
vector Kµ is space-like. The square of the vector Kµ defines the so-called boundary Kibble’s
function:  
 p2 (pa pb ) (pa pc ) 
 a
φ = 4Kµ2 =  (pb pa ) p2b (pb pc )  ≥ 0. (18.157)
 (pc pa ) (pc pb ) p2c 

© 2011 by Taylor and Francis Group, LLC

 S-matrix 445

The condition (18.157) may be expressed by way of the invariants s, t, u in the form unified
for all channels:
stu ≥ As + Bt + Cu, (18.158)
where 
Ah = (m2a m2b − m2c m2d )(m2a + m2b − m2c − m2d ), 
Bh = (m2a m2c − m2b m2d )(m2a + m2c − m2b − m2d ), (18.159)
Ch = (m2a m2d − m2b m2c )(m2a + m2d − m2b − m2c ).


As an illustration we consider the case when the s-channel describes the elastic scattering

ma = mc ≡ m, mb = md ≡ µ

with m > µ. Since:

h = 2(m2 + µ2 ), A = C = 0, B = (m2 − µ2 )2 ,

then the inequality (18.158) takes the form:

sut ≥ (m2 − µ2 )2 t. (18.160)

The boundary of the region governed by the inequality (18.160) consists of a straight line
t = 0 and a hyperbola
su = (m2 − µ2 )2 , (18.161)
which has two branches laying in sectors u < 0, s < 0 and u > 0, s > 0, and two
asymptotes s = 0 and u = 0. Next, from the condition (18.155) one must additionally
allow for the inequality s ≥ (m + µ)2 in the s-channel and the one u ≥ (m + µ)2 in
the u-channel. Thereafter, all the remaining inequalities are satisfied automatically. As a
result we arrive at Fig. 18.7 to display physical regions for all the three channels. Since
the physical domains of variables variation for the three processes under study are not
overlapping, the amplitudes may be coincident only in nonphysical domains (in Fig. 18.7
they are unshaded). In other words, the crossing relations are meaningful exclusively for
the amplitudes specified over the whole domain of the variables variation, that is, they
represent the analytical functions. Therefore, substitution of the variables in the relation
(18.152) should be understood not formally but rather as an analytical continuation from the
physical domain to the nonphysical one, for example, with respect to the variable s for the
process (18.137). Then the values of the invariant amplitudes As (s, t, u), Au (su , t, uu ), and
At (st , tt , u) in various physical domains are nothing else but an analytical continuation of the
function specified in one of the domains. N. Bogolyubov was the first to prove a possibility
for analytical continuation of the amplitude in the process of finding the dispersion relations
for the πN -scattering (see, for review, the monograph [11]). Based on the microcausality
principle, the existence of a unified analytical function of the complex variable s, whose
boundary values constitute the cross process amplitudes, has been demonstrated. The
main difficulty of the proof stems from the fact that the amplitudes include the generalized
functions calling for analytical extension too. The most general and classical outcome here
is a theorem (known as the “edge of the wedge” theorem) on the possible unification of
the advanced and retarded Green functions into a unified analytical function. Afterwards
this theorem was considered in Ref. [12]. Based on this theorem, the dispersion relations
may be proved for different cases. So we come to the conclusion that a crossing symmetry
follows from the scattering amplitude analyticity and hence scattering matrix analyticity.
And now let us touch upon the dispersion relations. Formally, they are Cauchy theorem
corollaries from a complex variable function theory applied to the scattering amplitude

© 2011 by Taylor and Francis Group, LLC

 446 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 18.7
The physical regions corresponding to the reactions I, II, and III.

A(s). The dispersion relations are usually represented with integral relations of the form:
Z
1 ∞ Im A(s′ ) ′
Re A(s) = ds + ..., (18.162)
π −∞ s′ − s
between the real and imaginary parts of the amplitude for a two-particle forward scattering:

A(s) = A(s, θ = 0) = Re A(s) + iIm A(s), (18.163)

each being related to the observables: total cross section σ(s) and zero-angle differential
cross section dσ(s, θ = 0).
R. Kronig and H. Kramers were the first to use the relations similar to (18.162) in their
study of the refractive index of light in a medium [13]. With the use of (18.162) the re-
lationship between dispersion and absorption has been established. Afterwards, the term
“dispersion relations” was introduced for analytical continuations of the scattering ampli-
tudes. The dispersion relations have been tested experimentally to reveal their applicability
up to the scales of ∼ 10−16 cm. The dispersion relations formed the basis for a great number
of approaches to the description of the strong interaction (see, for example, Ref. [14]). As
a matter of fact, they have lost their exclusive role to a considerable degree, due to the
progress in the quantum chromodynamics as the dynamic theory for the strong interaction.

Note that a crossing symmetry is a little more than a simple property of the S-matrix
associated with CP T -invariance. And CP T -invariance necessitates equality of the ampli-
tudes for the processes following from each other by permutation of the initial and final
states, with substitution of antiparticles for all the particles. The cross-invariance enables
such a transformation to be performed not only for all particles at once but also for any
particle individually.
We shall also use the term a cross-universality. Its meaning is as follows. For simplicity,
we consider a binary reaction, limiting ourselves to the second-order of the perturbation

© 2011 by Taylor and Francis Group, LLC

 S-matrix 447

theory. Let us assume that it is described by two Feynman diagrams, and in these diagrams
the propagators of virtual particles have denominators including the variables s and u,
respectively. In this case we can state that the considered reaction proceeds through the
s- and u-channels. Then, if we use a particular permutation of particles in the initial (or
final) state, one diagram follows from the other. Obviously, there exists a substitution of
the momenta and the polarization vectors of the particles making it possible to couple the
matrix elements corresponding to both diagrams. Invariance of the amplitude with respect
to such a substitution is called a cross-universality. This feature proves to be very useful
for the calculation of scattering cross sections.

18.10 Cross section for unpolarized particles

Let us establish the relations between elements of the S-matrix and cross sections of the
scattering processes. For the initial and final states normalized to the unit volume the
probability of transition from the initial state to the final one is determined as follows:
′
dwi→f = |Ri→f |2 . (18.164)

Changing one of the δ-functions in Ri→f in accordance with the formula:

 
  Z Z
(4)
X X 1
δ ( pi − pf ) = lim lim dt d3 x×
 P P V →∞ T →∞ (2π)4
i f
pi = pf
    
 X X  VT
× exp i  pi − pf  x = , (18.165)

i
P (2π)4
f P
pi = pf

we get X X X
′
dwi→f = (2π)4 δ (4) ( pi − pf )| Mi→f |2 V T. (18.166)
i f

Actually, the quantum-mechanical nature of the elementary particles and also the experi-
mental conditions lead to the situation when a detector is recording the particles with, in
principle, the unavoidable spread of the values measured for each of the three momentum
components of particles in the final state over some intervals. As the final (as well as initial)
state is associated with a continuous spectrum, of practical interest is the knowledge that
the three-dimensional momenta of the final state particles p′j are lying in the given intervals
d3 p′j . To find this probability, Eq. (18.166) should be multiplied by the element of a phase
volume (states number of particles with definite quantum numbers and momenta falling in
the specified intervals):
3 ′
d3 p′1 V d3 p′2 V d pnf V
... , (18.167)
(2π)3 (2π)3 (2π)3
where V is a normalization volume and nf is a particles number in the final state. Dividing
the obtained expression by V T , we get the differential probability of the process in a unit
of time and in a unit of volume
′ nf
dwi→f nf
Y d3 p′j
dwi→f = V =
VT j=1
(2π)3

© 2011 by Taylor and Francis Group, LLC

 448 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
nf
X X X Y d3 p′j
= (2π)4 δ (4) ( pi − pf )| Mi→f |2 V nf . (18.168)
(2π)3
i f j=1

Of course, it is preferable to have a quantity that is independent on the particles number in

a beam or in the target, characterizing only the intensity of interaction between the colliding
particles. In the lab frame the necessary features will be displayed by the quantity that is
equal to the ratio of the probability dwi→f to the density of the target particles ρt , and to
the density of the incoming particles flux jb :
P nf
′
X X | Mi→f |2 Y d3 p′j
dσi→f = (2π)4 δ (4) ( pi − pf ) , (18.169)
i
ρt jb j=1
(2π)3
f

where we have set the normalization volume equal to 1. However, when considering expres-
sion (18.169), a reader (even inexperienced) can raise a question: How can it be possible
that the experimentally measured quantity involves the δ-function? Indeed, this function
is not one of the ordinary (analytical) functions. It is classed with the generalized func-
tions specified in terms of their effect on the ordinary ones. And the δ-function appears
only at the intermediate steps of calculations, not reaching the final physical result. In Eq.
(18.169) the δ-function expressing the four-dimensional momentum conservation law may
be eliminated by two integrations over the three-dimensional momenta of any two particles
belonging to the final state. Thus, the differential probability of scattering into the final
state is given by the expression:
Z ′
′
dσi→f = dσi→f , (18.170)

where the prime above the integral sign denotes the twice integration eliminating the δ-
function to be fulfilled. In so doing the following terminology is adopted. Provided in the
final state we have two particles with the momenta p′1 and p′2 , then we can take a set of the
final states, where a three-dimensional momentum, for example, of the first particle p′1 is
lying within the given solid angle dΩ. After integrations, we obtain a quantity proportional
to the element of the solid angle dΩ (beyond doubt one may transform to another variable
connecting with the solid angle). This quantity is called the differential cross section into
the element of the solid angle. The total cross section follows from the differential one after
integration with respect to dΩ over the interval from 0 to 4π.
When the final state involves three particles, after the required integrations we are left
with two variables, for example, y and z (and an element of the solid angle dΩ may be
selected as one of them too). Then we have a double differential cross section with respect
to the variables y and z. To obtain the total cross section, integration must be done with
respect to y as well as z. Also, in real experiments the particles of more than three kinds
may be generated. But, as a rule, we confine ourselves to detection only and hence to the
description of the final states involving no more than three particles.
Next, we are constrained by discussion only situations being the most important from the
physical point of view, namely, when there are only one or two particles in a initial state.
In the former we are dealing with a decay while in the latter—with a particles collision.
We shall address decays later, but now we proceed to the determination of a flux density of
initial particles in a case of two particles collision. In the QED one could find this quantity
considering a current density of the electron-positron field and the Poynting vector for the
electromagnetic field. First, we define the flux density for the case when one of the initial
particles is in rest (such a particle will be called a target). We assume that the incident flux
consists of electrons. Then, the average value of the current operator in the initial state:
|i >= Φ1 = (2π)3/2 a†σ (p)Φ0 , (18.171)

© 2011 by Taylor and Francis Group, LLC

 S-matrix 449

is set down as:

m σ
< i|jµ (x)|i >= e(2π)3 < 0|aσ (p)ψ(x)γµ ψ(x)a†σ (p)|0 >= e u (p)γµ uσ (p) =
Ep
pµ
=e = evµ , (18.172)
Ep
where vµ is a speed of the initial electron. It is evident that we shall come to the analogous
result for a positron flux too. Looking at the Poynting vector one may also establish that
in the case of photons the flux density of initial particles appears to be equal to 1. Thus,
we have ascertained that at the accepted normalization the flux density both for a fermion
and for a photon incoming on a motionless target is simply equal to its speed.
In the general case, at the particles density of a beam and a target different from 1, we
have for initial particles flux density:
(0) (0) (0)
J (0) = ρb ρt vb , (18.173)
(0) (0)
where ρb and ρt are flux densities of moving particles and target ones in the reference
(0)
frame K (0) where target particles rest vt = 0. To find an expression for the initial particles
flux density in an arbitrary Lorentz reference frame K where both initial particles move,
(0) (0) (0)
we present ρb ρt and vb in a relativistically invariant form. Since:
(0) (0) (0) (0)µ
ρb ρt = jbµ jt , (18.174)

then in the system K the obvious relation:

(p1 p2 )ρb ρt
jbµ jtµ = , (18.175)
E1 E2
where p1,2 are momenta of collided particles, takes place. Further, we have:
q
(0) (0) (0)2 p
(0) |p1 | m |p
2 1 | m22 E1 − m21 m22 (p1 p2 )2 − p21 p22
vb = (0) = (0)
= (0)
= , (18.176)
E m2 E m2 E (p1 p2 )
1 1 1

(0) (0) (0) (0)

where we have accepted designations p1 ≡ pb , p2 ≡ pt . From (18.175) and (18.176)
we get the expression for the flux J in an arbitrary Lorentz reference frame:
p
ρb ρt (p1 p2 )2 − p21 p22
J= . (18.177)
E1 E2
Using this formula and the definitions (18.169), (18.170) and taking into account that at
accepted normalization ρt = ρt = 1, we arrive at the following expression for the differential
cross section of the scattering process:
Z nf nf
4
′
(4)
X X E1 E2 Y d3 p′j
dσi→f = (2π) δ (p1 +p2 − p′j )| Mi→f | p 2
. (18.178)
j=1 (p1 p2 )2 − m21 m22 j=1 (2π)3

We shall show that the expression for dσi→f may be rewritten in such a way that√ it will
contain
P invariant quantities only.
P To do this we select noninvariant factors 1/ Ek from
Mi→f , that is, we exhibit Mi→f in the form:
X Y 1
Mi→f = Ai→f √ , (18.179)
Ek

© 2011 by Taylor and Francis Group, LLC

 450 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where multiplication should be extended over all particles both in the initial and in final
states and Ai→f is a relativistic invariant. Inserting this expression into (18.178) and
allowing for the relation:
Z 3 Z
d pf
= 2 d4 pf δ(p2f − m2f )θ(Ef ),
Ef

we obtain
Z nf nf " 4 ′ #
4
′
(4)
X Y d pj δ(p′2
2nf 2 ′
j − mj )θ(Ej )
dσi→f = (2π) δ (p1 +p2 − p′j )|Ai→f |2 p .
j=1 (p1 p2 )2 − m21 m22 j=1 (2π)3
(18.180)
Now the relativistic invariance of dσi→f will not create any doubts.
Let us come back to the expression (18.178) and eliminate the δ-functions entering in it
nf nf nf
X X X
δ (4) (p1 + p2 − p′j ) = δ (3) (p1 + p2 − p′j )δ(E1 + E2 − Ej′ ). (18.181)
j=1 j=1 j=1

Carry out this operation in the explicit view for the two particles state. To integrate over
d3 p′1 gives rise to the disappearance of δ (3) (p1 + p2 − p′1 − p′2 ). In order to remove the
second δ-function we rewrite d3 p′2 in the spherical coordinate system:

d3 p′2 = |p′2 |2 d|p′2 | sin θdϕdθ = |p′2 |2 d|p′2 |dΩ = |p′2 |E2′ dE2′ dΩ, (18.182)

where ϕ is an azimuthal angle of the vector p′2 , θ is a zenith angle of the vector p′2 (the
angle between p′2 and the third coordinate axis) and we have used the relation:

|p′2 |2 = E2′2 − m22 .

Integration over E2′ may be fulfilled when one replaces dE2′ with d(E1′ + E2′ ). Since E1′ + E2′
is a function of E2′ , then the relation:

∂ (E1′ + E2′ ) ′
d (E1′ + E2′ ) = dE2 (18.183)
∂E2′

takes place. With allowance made for (18.182), we obtain

 −1
∂(E1′ + E2′ )
d3 p′2 = |p′2 |E2′ d (E1′ + E2′ ) dΩ. (18.184)
∂E2′

To substitute (18.184) into (18.180) and integrate over E1′ + E2′ gives the final expression
for the differential cross section of the reaction:
"  −1 X #
−2 ′ ′ ∂(E1′ + E2′ ) 2
dσi→f = (2π) |p2 |E2 | Mi→f | ×
∂E2′ ′ ′ p1 +p2 −p1 −p2

E1 E2
×p dΩ. (18.185)
(p1 p2 )2 − m21 m22
Assume that a beam of the initial state particles is not polarized, whereas devices detect-
ing the scattered particles respond to the particles with all values of the spin projection.
In other words, we are not interested in a particular polarization state of both the initial
and final particles. In this case, the cross section should be summed up over all possible

© 2011 by Taylor and Francis Group, LLC

 S-matrix 451

polarization states of the final particles

P and averaged over polarizations of the initial ones.
We denote both operations as spin , that is, the cross section for unpolarized particles is
represented as
X
dσi→f .
spin

Summation over the polarizations of photons is carried out with the use of Eq. (17.44), and
the summation in the case of electrons and positrons—with the use of Eqs. (15.103) and
(15.104). And in the case of averaging over the polarizations, we use the same equations
with a preceding factor of 1/2.

18.11 Cross section for scattering by external fields

When using the Feynman momentum-space rules in studies of the particle scattering pro-
cesses by external electromagnetic fields, we need to know the Fourier images for all the
“familiar” potentials. Generally, for their calculation the classical electrodynamics equa-
tions are required. In so doing, the fact whether an external field depends or not depends
on time is important as well.
To illustrate the basic pattern for the description of interactions with an external field,
we consider scattering of an electron by the external electromagnetic field. It should be
noted that nuclear scattering of electrons used in studies of the nuclear charge distribution
is a very effective instrument to reveal the structure of nuclear levels. Fig. 18.8 presents
the relevant Feynman diagram in the first order of the perturbation theory. As the Fourier

q = p2 − p 1

p2 p1

FIGURE 18.8
The Feynman diagram describing the electron scattering by the external electromagnetic
field.

image for the potential of an external field is not known, we have to use the Feynman rules
in the coordinate representation. In this representation the matrix element is as follows:
s Z
m2 ′
Si→f = −ie d4 xur′ (p′ )Aeµ (x)γ µ ur (p)ei(p −p)x . (18.186)
Ep Ep′

© 2011 by Taylor and Francis Group, LLC

 452 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The external field Aeµ (x) may be expressed in terms of the sources density. For instance,
when we have an external field created by the moving nucleus, then for the scalar potential
we get:
Ae0 (r, t) = Zeρ(r, t), (18.187)
where ρ(r, t) characterizes a nucleus charge distribution density. From (18.187) follows the
well-known formula: Z
Ze ρ(r′ , t − |r − r′ |)
Ae0 (r, t) = d3 x′ . (18.188)
4π |r − r′ |
Of course, expression (18.188) may be substituted into the matrix element (18.186), reducing
our task to the computation of integrals. But these integrals are rather sophisticated to
force us to use the momentum representation again.
To substitute the Fourier expansion of the nucleus potential
Z
1
Aeµ (x) = d3 qAeµ (q)e−iqx
(2π)3/2

into (18.186) allows to carry out integration over the variable x and this, in its turn, leads to
the appearance of the δ-function. In problems of particles collision the δ-function expresses
the conservation law of the four-dimensional momentum of the particles system. Because
we change an external field with an effective virtual photon, a pseudophoton (q 2 6= 0),
herein presence of the δ-function leads to the fact that the four-dimensional pseudophoton
momentum taken with the opposite sign equals the four-dimensional momentum to be
transferred to the nucleus as a consequence of the electron scattering

−q = p − p′ .

So, −q makes up the three-dimensional momentum of nucleus recoil while −q0 makes up
the energy imparted to the nucleus. All this means that at scattering by a fixed source
the four-dimensional momentum is not conserved by this time. A source can absorb or
give back a momentum without change of its state. Thus, for example, in the case of
a stationary source the three-dimensional momentum is not conserved and only particles
energy is conserved. Really, the variable q0 is absent in the stationary potential expansion:
Z
1
Aeµ (r) = d3 qAeµ (q)eiqr (18.189)
(2π)3/2

From (18.189) it is evident that in the given case, the pseudophotons transfer the three-
dimensional momentum q and their energy q0 equals zero. Now the matrix element (18.186)
is rewritten in the following manner:
Z s
5/2 3 m2
Si→f = −ie(2π) d q ur′ (p′ )Aeµ (q)γ µ ur (p)δ(q + p − p′ )δ(Ep′ − Ep ) =
Ep Ep′
s
5/2 m2
= −ie(2π) ur′ (p′ )Aeµ (p − p′ )γ µ ur (p)δ(Ep′ − Ep ). (18.190)
Ep Ep′

We shall determine the cross section of the electron on Coulomb’s field of the motionless
nucleus considering the nucleus field to be point charge, that is,

Ze
Aeµ (r) = gµ0
4πr

© 2011 by Taylor and Francis Group, LLC

 S-matrix 453

(In this case, one may say that unlike a real photon polarization vector, a pseudophoton
polarization vector is directed along the time axis). To seek the Fourier image of the
potential Ae0 (r) we direct our attention to the Poisson equation for a point charge placed
at the coordinates origin:
∆A0 (r) = −eδ(r).
The consequence of this equation is the relation:
1
∆ = −4πδ(r). (18.191)
r
With the help of (18.189) and (18.191) we easily obtain the required quantity:

Ze
Ae0 (q) = . (18.192)
(2π)3/2 |q|2

Thus, the probability of the electron scattering by the Coulomb’s field of the nucleus into
one selected final state ur′ (p′ ) is given by the expression:

Z 2 α2 m2 (2π)2 16π 2 1 T
dWi→f = |ur′ (p′ ) γ 0 ur (p)|2 δ(Ep′ − Ep ) , (18.193)
Ep Ep′ |p − p′ |2 2π

where the appearance of the factor T /(2π) is due to the δ-function

Z T /2
1 T
δ(Ep′ − Ep ) = lim dtei(Ep −Ep′ )t =
Ep →Ep′ 2π −T /2 2π

When the initial beam of the electrons is not polarized and we are not interested in the
spin states of the scattered beam, then we should sum up the obtained expression over the
final spin states and average over the initial ones. Then, we arrive at the following expression
for the probability of the electron scattering into the final states group d3 p′ /(2π)3 in a unit
of time:
 ′ 
4Z 2 α2 m2 1 p̂ + m 0 p̂ + m 0 1
dwi→f = Sp γ γ δ(Ep′ − Ep )d3 p′ . (18.194)
Ep Ep′ 2 2m 2m |p − p′ |4

In order to get the differential cross section of scattering we should divide the expression
(18.194) into initial particles flux density (in our case this quantity is simply |p|/Ep = v)
and get rid of the δ-function. Integration over Ep′ will be fulfilled without trouble if we
pass in the spherical coordinate system

d3 p′ = |p′ |Ep′ dEp′ dΩ.

Since Ep = Ep′ , then |p′ | is equal to |p| as well. Next, we calculate the spur of the γ-matrices
entering into (18.194). Using the formulae of Section 15.2, we obtain
 ′ 
p̂ + m 0 p̂ + m 0 1

Sp γ γ = 2
Sp p̂′ γ 0 p̂γ 0 + m2 =
2m 2m 4m
1 1
=[2Ep Ep′ − (p′ · p) + m2 ] = 2 [E 2 + (p′ · p) + m2 ] =
m2 m
 
1 2 2 2 1 2 2 2 2E 2 2 2 θ
= 2 [E + p cos θ + m ] = 2 [(E − m )(1 + cos θ) + 2m ] = 2 1 − v sin ,
m m m 2
(18.195)

© 2011 by Taylor and Francis Group, LLC

 454 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where θ is an angle between momenta of the incoming and scattered electron, E = Ep = Ep′
and p = |p′ | = |p|. Gathering the obtained results, we find the final expression for the
differential cross section of the electron scattering by the Coulomb’s field of the nucleus
[15]:
 
Z 2 α2 θ
dσi→f = 2 2 4 1 − v 2 sin2 dΩ. (18.196)
4p v sin (θ/2) 2
A few words about the formula (18.196). It is distinguished from Rutherford’s formula by
the factor [1 − v 2 sin2 (θ/2)]. This factor serves as the correction allowing for the electron
spin. Really, it appears from the expression for the spur, and hence must include interaction
of an electron magnetic moment with a magnetic field. In the reference frame where the
electron rests (v = 0) this factor becomes equal to 1. As calculations show, the cross
section for scattering a spinless particle by the Coulomb’s field of the nucleus is defined by
the Rutherford’s formula. Thus, we are again convinced that the cross sections of particles
possessing the spin involve the ones of scalar particles.
Computing the scattering cross section in subsequent orders of the perturbation theory
leads to a divergent result to be caused by the infinite range of a Coulomb field. This problem
has been solved by Dalitz [16], who has demonstrated that the scattering amplitude in a
screened Coulomb field
Ze e−ηr
Ae0 (r) = ,
4π r
where η −1 is a screening radius (η > 0), in the nonrelativistic limit is reduced to the lower-
order scattering amplitude (in the first order on Zη) multiplied by the phase factor to go to
infinity as η −1 when η → 0. Divergences in the higher-order of the perturbation theory are
emerging just due to the infinite contribution of this phase factor. By the method proposed
by Dalitz, the screened Coulomb potential Ae0 (r) is used at all stages of the intermediate
calculations, while at the final stage it is assumed that η is zero. In this way the final result
includes no screening radius, divergences associated with a long-range interaction of the
Coulomb forces being absent. This problem will be discussed in greater detail in Section
21.6 devoted to the radiation corrections for the electron scattering by the Coulomb nucleus
field.
Let us proceed to the description of a particle scattering from a static external electro-
magnetic field in the higher-orders of the perturbation theory, that is, in the case when a
matrix element involves the integrals over the virtual momenta. In Fig. 18.9 we display the
Feynman diagram describing the electron scattering by a static external field in the second
order of the perturbation theory. Using the Feynman rules in the coordinate representation,

p1 p2

FIGURE 18.9
The Feynman diagram describing the electron scattering by the external electromagnetic
field (second order of the perturbation theory).

© 2011 by Taylor and Francis Group, LLC

 S-matrix 455

we arrive at the expression for the corresponding matrix element:

s
2 m2
Si→f = −ie (2π)δ(Ep − Ep′ ) ur′ (p′ )×
Ep Ep′
Z
k̂ + m e ′
× d3 kAeν (p − k)γ ν 2 A (p − k)γ µ ur (p),
k − m2 µ
where the integration already occurs only over the space components of the virtual elec-
tron momentum. As seen, such a disappearance of an integration with respect to the time
component of the virtual particle momentum takes place only when at one of the vertices,
between which this particle is moving, there is an interaction with an external field. Owing
to the experience gained, the Feynman rules may be extended in the momentum repre-
sentation to include an interaction with a static external electromagnetic field too. These
additions are as follows.
10) The factor:
Aeµ (q)
(2π)3/2
corresponds to an interaction with an external static electromagnetic field at a vertex with
the index µ.
11) The factor:
ieγ µ (2π)4 δ (3) (p − p′ − q)δ(p0 − p′0 ).
corresponds to every vertex to describe an interaction with an external static electromag-
netic field.
12) Integration over variables q relating to external potentials is carried out.
Now we could give a general formula for the differential cross section of a particle scat-
tering by an external static electromagnetic field. It has the following manner:
nf
2π h ′ ′ X 2
i Y d3 p′j
dσi→f = |p1 |E1 | Mi→f | P ′ dΩ . (18.197)
v E− Ej
j=2
(2π)3

where v and E are a speed and an energy of an incoming particle, and a matrix element
Mi→f is found by means of the Feynman rules in the momentum representation with an
allowance made for the supplements done above. If scattering happens on a point Coulomb
field having the charge Ze, then the external potential Fourier-images entering into Mi→f
are determined by the expression (18.192).
The above-stated scheme may be applied to finding a differential cross section of scattering
by any external electromagnetic field. Note, when an external field depends on time, elastic
scattering of a particle occurs with changing an energy (frequency), however, when an
external field is constant in time, under elastic scattering changing an energy does not take
place (coherent scattering).

18.12 Decays of elementary particles

We are coming now to the analysis of elementary particle decay processes. We have the
following expression for the decay probability in a unit of time:
nf nf
4 (4)
X X Y
2 d3 pj
dΓ = (2π) δ (k − pj )| Mi→f | , (18.198)
j=1 j=1
(2π)3

© 2011 by Taylor and Francis Group, LLC

 456 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
P
where k is four-dimensional momentum of a decaying particle. Since | Mi→f |2 involves
the factor E −1 , where E is an initial particle energy, then the decay probability is decreased
as the speed of a decaying particle grows. The total decay probability of an unstable particle
in a unit of time is found by integrating (18.198) over all momenta of secondary particles.
If in this case there are k-modes (channels) of a decay, then the total probability Γt will be
given by a sum of decay probabilities on every channel
X
Γt = Γk .

The total decay probability in a unit of time Γt is connected with a lifetime of an unstable
particle by the relation:  
1 ~
τ= = . (18.199)
Γt Γt
The lifetime τ defines a time interval, during which the number of initial unstable particles is
decreased by a factor equal to e (e = 2.7183..). The lifetime of a resting unstable elementary
particle τ0 is a physical characteristic of a particle, it being given in tables of elementary
particle properties. In a reference frame where an unstable particle moves with a speed
v = p/E, the lifetime is increased by a factor equal to E/m as compared with τ0 .
As an example, we consider a two-particle decay of the W -boson

W → a + b. (18.200)

The differential decay probability dΓW →ab in a unit of time for the resting W -boson is
determined as
1 X d3 pa d3 pb
dΓ = (2π)4 δ (4) (k − pa − pb ) | M′W →ab |2 , (18.201)
mW Ea Eb (2π)6

where for the sake of convenience we have selected factors mW , Ea , and Eb , which are
connected with the normalization, from |MW →ab |2 . Next, we assume that a Lagrangian
form describing the W ab-interaction is unknown for us. Invoke the relativistic covariance
for help and try to find the expression for the total decay probability. Obviously, the matrix
element MW →ab may be presented in the form:

MW →ab = gϑλµ (k)M µ , (18.202)

where g is a constant of W ab-interaction and ϑλµ (k) is a four-dimensional vector of the

W -boson polarization. To average over the spin states of the W -boson leads to the result:
 
1X 2 g2 kµ kν
|MW →ab | = − gµν − 2 M µ M ν† . (18.203)
3 3 mW
λ

From the relativistic invariance demand follows that the quantity M µ M ν† must be a four-
dimensional tensor of the second rank. We can construct such a tensor using independent
quantities that are at our disposal. In the case in question we have, in the capacity of
those, two vectors k µ , (pa − pb )µ and the metric tensor g µν . The vector k µ is out, since the
transversality condition:
k µ ϑλµ (k) = 0
is valid for a massive vector field. Then, we obtain for M µ M ν†

M µ M ν† = A(pa − pb )µ (pa − pb )ν + Bg µν , (18.204)

© 2011 by Taylor and Francis Group, LLC

 S-matrix 457

where in the most general case A and B are analytical functions of invariant variable
q 2 = (pa + pb )2 to be taken in the point q 2 = m2W . Substitution of (18.204) into (18.201)
with allowance made for the relations
Ea2 − Eb2 m2 − m2b
Ea − Eb = = a , (18.205)
Ea + Eb mW

(m2a − m2b )2
4Ea Eb = (Ea + Eb )2 − (Ea − Eb )2 = m2W − , (18.206)
m2W
1

p2 = Ea2 + Eb2 − m2a − m2b , (18.207)
2
where | pb |=| pa |= p, leads to the expression:

g2  −2
dΓ = (2π)4 δ (4) (k − pa − pb ) mW A[w(m2W , m2a , m2b )]2 −
3mW Ea Eb

d3 pa d3 pb
−3B} , (18.208)
(2π)6
where the function w(A, B, C) being completely antisymmetric with respect to all its vari-
ables has the form:

w(A, B, C) = [A2 + B 2 + C 2 − 2(AB + AC + BC)]1/2 . (18.209)

To determine the total decay probability we should integrate the expression (18.208) over
momenta of decay products. In this case a knowledge of the following integral:
Z
d3 pa d3 pb
I = δ (4) (k − pa − pb ) . (18.210)
4Ea Eb
proves to be useful. To integrate this expression over the three-dimensional momentum of
the a particle gives: Z
d3 pb
I = δ(mW − Ea − Eb ) . (18.211)
4Eb Ea
Introduce the spherical coordinate system

d3 pb = p2 dpdϕd(cos θ)

and allow for that all directions of the vector p are equiprobable. Then (18.211) takes the
view:
Z 2 Z  −1
p dp p2 d(Ea + Eb )
I=π δ(mW − Ea − Eb ) = π δ(mW − Ea − Eb )×
Ea Eb Ea Eb dp
πp π
×d(Ea + Eb ) = = m−2 w(m2W , m2a , m2b ), (18.212)
mW 2 W
where at the final stage of calculations we have taken into account the relations (18.205)–
(18.207). Using (18.212) gets the following expression for the total cross section of the decay
(18.200):

g 2  −2
ΓW →ab = 3 mW A[w(m2W , m2a , m2b )]3 − 3Bw(m2W , m2a , m2b ) . (18.213)
6πmW

© 2011 by Taylor and Francis Group, LLC

 458 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

It is clear that to define the quantities A and B we should have additional sources of infor-
mation. If a successive and internally consistent theory describing the W -bosons interaction
were absent, we would have to address experiments. However, nowadays we have the SM to
be in perfect accord with modern experimental data. With its help we could get the explicit
expression for the matrix element of the reaction (18.200). For example, when a = l and
b = ν l , A (B) proves to be equal to -1/2 -(m2W − m2l )/2, where for reasons of simplicity we
have neglected the antineutrino mass.

18.13 Cross section for polarized particles

In this section we take into consideration the polarization states of interacting particles.
We start with a case when one electron is present both in an initial state and in a final one.
Then the process amplitude is given by

Mi→f = uf Oui , (18.214)

where ui ≡ u(pi ), uf ≡ u(pf ), and O is a 4 × 4-matrix whose concrete form is unessential

in this context. The process probability is proportional to the amplitude module square:

dwi→f ∼ |Mi→f |2 = (uf Oui )(ui Ouf ) = uiα uiα′ O α′ β ′ uf β ′ uf β Oβα , (18.215)

where O = γ0 O† γ0 and we have set down spinor indices in the explicit form. When an
electron in a initial state is completely polarized, then the formula (18.215) does not undergo
any changes. In the case of a partially polarized initial electron one should carry out the
following change in the formula (18.215):
(+)
uiα uiα′ → (ρi )αα′ ,
(+)
where ρi is a polarization density matrix∗ of the electron in the initial state. As far as
the definition of the electron polarization in a final state is concerned, then, from the point
of view of experiment, two situations are possible here: (i) measurement of the probability
that the electron in a final state will possess definite polarization; and (ii) finding of all
probable polarization states, that is, establishing the density matrix of the electron in a
final state. In the first case, a detector selects the definite polarization state of the electron.
If this state is characterized by a four-dimensional vector of polarization sd , then one must
(+)
replace the bilinear combination uf β uf β ′ with the density matrix (ρd )ββ ′ in (18.215). Here
(+)
ρd represents the density matrix to define a detector and have the same structure as the
(+) (+)
matrix ρi : ρi
 
(+) 1 p̂f + m
ρd = (1 + γ5 ŝd ) , (18.216)
2 2m
where (pf · sd ) = 0. So, if the initial electron is partially polarized and the definite po-
larization state d of the final electron is fixed, then the probability of the process under
investigation is defined as:
(+) (+)
dwi→f ∼ Sp(ρi Oρd O). (18.217)

∗ We shall further call it by the density matrix for brevity sake.

© 2011 by Taylor and Francis Group, LLC

 S-matrix 459

It is obvious that interaction with a detector changes the electron state. Therefore, the
matrix ρd and with it the probability dwi→f characterizes not only the electron scattering
process itself but detector properties too, that is, its capabilities for selection of one or
an other electron polarization. When we are interested in a polarization state of the final
electron in which the electron is brought by the interaction process itself, then we come
to the second setting up a problem. The probability, that the electron described by the
bispinor uf possesses the definite polarization to be associated with the bispinor ud , is
proportional to |ud uf |2 = udα udβ uf β uf α . Since the case in point is the states with partial
polarization rather than a definite one, then the changes
(+) (+)
uf β uf α → (ρf )βα , udα udβ → (ρd )αβ (18.218)

must be done in the expression for the probability. Thus, the required probability proves
(+) (+)
to be proportional to the expression Sp(ρd ρf ). As the same probability is defined by
the expression (18.217), then the conclusion follows
(+) (+) (+) (+)
Sp(ρi Oρd O) ∼ Sp(ρf ρd ). (18.219)

So, we have for the density matrix of the electron in the final state:
(+) (+)
ρf ∼ Oρi O. (18.220)

However, our detector has been tuned only on the electron recording, therefore, the matrix
O must have somehow stored information about this fact. Accounting for:

P+ u f = u f ,

where P+ = (p̂f + m)/2m represents the projection operator onto the electron states, the
matrix element (18.214) may be reported as:

Mi→f = uf P+ Oui .
(+)
Then, the matrix ρf will be given by:
(+) (+)
ρf = N P+ Oρi OP+ . (18.221)

Normalization factor N is determined by the normalization condition of the density matrix

Sp(ρf ) = 1:
1 h i
(+)
N −1 = Sp Oρi O(p̂f + m) . (18.222)
2m
To replace (18.222) into (18.221) gives the final expression for the density matrix of the
electron in the final state:
(+)
(+) 1 [(p̂f + m)Oρi O(p̂f + m)]
ρf = (+)
. (18.223)
2m Sp[(p̂f + m)Oρ O] i

This matrix is defined only by the four-dimensional polarization vector of the electron in
the final state sf :  
(+) 1 p̂f + m
ρf = (1 + γ5 ŝf ) , (18.224)
2 2m
(+)
where (sf · pf ) = 0. Knowledge of ρf allows to find sf
(+)
sf µ = Sp(ρf γ5 γµ ),

© 2011 by Taylor and Francis Group, LLC

 460 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

which, in its turn, unambiguously defines the three-dimensional polarization vector ξ.

In a similar spirit we may also consider cases when initial and final states are positron or
one of the states is electron while the other is positron. For example, the expression for the
density matrix of the positron in the final state has the form:
(−)
(−) 1 [(−p̂f + m)Oρi O(−p̂f + m)]
ρf = (−)
. (18.225)
2m Sp[(−p̂f + m)Oρ O] i

The results obtained are generalized without trouble in cases of many electrons and positrons.
Now we shall pass to the investigation of processes with the participation of partially
polarized photons. Let us consider a process when the initial state contains one photon and
the final state contains one photon as well. In this case the matrix element of the process
could be presented in the form:
′
Si→f = eλf µ∗ Oµν eλiν , (18.226)

where eλ are photon polarization vectors. If the initial photon is partially polarized, then
in the matrix element square
′ ′
|Si→f |2 = eλf µ∗ Oµν eλiν eλf τ Oτ σ∗ eλ∗
iσ (18.227)

the replacement:
γ
eλiν eλ∗
iσ → (ρi )νσ

should be fulfilled. As this takes place, when three-dimensional polarization vectors figure
in the relation (18.226), then the density matrix is given by:

1
(ργi )ll′ = [1 + (ξ i · σ)]ll′ , (18.228)
2
where ξ i are Stokes’s parameters for the photon in the initial state and l, l′ = 1, 2. If four-
dimensional polarization vectors enter into the relation (18.226), then the density matrix
will be defined as follows:
2
1 X (l) (l′ ) 1 (1) (1) (2) (2)
(ργi )µν = [1 + (ξ i · σ)]ll′ (ni )µ (ni )ν = [(ni )µ (ni )ν + (ni )µ (ni )ν ]+
2 ′ 2
l,l =1

ξ1 (1) (2) (2) (1) iξ2 (1) (2) (2) (1)

+ [(ni )µ (ni )ν + (ni )µ (ni )ν ] − [(ni )µ (ni )ν − (ni )µ (ni )ν ]+
2 2
ξ3 (1) (1) (2) (2)
+ [(ni )µ (ni )ν − (ni )µ (ni )ν ], (18.229)
2
(l)
where four-dimensional vectors (ni )µ (l = 1, 2) satisfy the relations:
′
)
(l)
(ni · n(l )i ) = −δll′
(l) (18.230)
(ki · ni ) = 0

and ki is a four-dimensional momentum of the photon in the initial state.

Next, we shall discuss questions of determining the photon polarization in a final state.
Here, in just the same way as in the case of any particles with the spin, two setting up
problems are possible: (i) finding the probability that the final photon possesses definite
polarization; and (ii) determining the density matrix of the final photon. To solve the first

© 2011 by Taylor and Francis Group, LLC

 S-matrix 461
′ ′
problem is trivial and resides in the replacement of the bilinear combination eλf ν eλf µ∗ involved
(γ)
in (18.227) with the density matrix (ρd )νµ to characterize a detector:
(l) (l)
ργd = ργi (ξi → ξd , ni → nd ),
(l)
where ξ d are Stokes’s parameters for a detector, nd are two vectors satisfying the relations:

(l) (l′ ) (j)

(nd · nd ) = −δll′ , (kf · nd ) = 0

and kf is a four-dimensional momentum of the photon in the final state. The probability,
that the photon whose polarization in the initial state was being set by the matrix ργi will
have polarization set by the matrix ργd , is defined by the expression:
X ′ ′ ′ γ γ †
dw(sd ) ∼ g λ g λ eλdµ∗ Oµν eλiν eλdµ′ Oµ∗ ′ ν ′ eλ∗
iν ′ = Sp(ρd Oρi O ), (18.231)
λ,λ′

where g λ is a relative probability with which λ-polarization goes into a mixed beam.
To solve the second problem we take into consideration this fact: if the photon in the final
state is characterized by the four-dimensional polarization vector eλfµ , then the probability
′
that the state with the four-dimensional polarization vector eλdµ will be recorded with the
help of a detector is given by the quantity:
′ ′ ′
dW ∼ |eλfµ eλdµ∗ |2 = eλfµ eλdµ∗ eλ∗ λ
f µ′ edµ′ . (18.232)

When we are dealing with partially polarized states rather than with pure ones, then in the
expression (18.232) we should carry out the changes:
γ ′ ′
eλfµ eλ∗
f µ′ → (ρf )µµ′ , eλdµ eλdµ∗′ → (ργd )µµ′ , (18.233)

where ργf is a density matrix of the photon in the final state. Then, for the required
probability we get:
dw(sd ) ∼ Sp(ργf ργd ). (18.234)
To compare (18.234) with (18.231) gives the relation:

ργf = N (Oργi O† ). (18.235)

Having defined the coefficient N from the normalization condition, we come to the final
expression for the density matrix of the photon in the final state:

Oργi O†
ργf = . (18.236)
Sp(Oργi O† )

Substituting the definition of the matrices ργd and ργf , where

1  (l) (l′ )
(ργf )µν = 1 + (ξ f · σ) ll′ (nf )µ (nf )ν , (18.237)
2
(l) (l)
and nf ≡ nd , into the relation (18.234), we arrive at:

1 
dw(sd ) ∼ Sp(ργf ργd ) = 1 + (ξ f · ξ d ) . (18.238)
2

© 2011 by Taylor and Francis Group, LLC

 462 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The meaning of the obtained formula is as follows. If the cross section summed up over
polarizations of the photon in the final state equals σ, then the production cross section of
the photon with polarization parameters ξ d will be given by the expression:
1  
σd = σ 1 + (ξ f · ξ d ) . (18.239)
2
The reader could find more detailed information about the density matrix and its usage
in calculating probabilities of different processes in elementary particles physics in the book
[17].

18.14 Statistical hypothesis

As follows from the formulae (18.169) and (18.198) the differential cross section dσ and the
differential decay probability in a unit of time dΓ are defined by the phase volume element
dJ
nf
4 (4)
X X Y d3 p′j
dJ = (2π) δ ( pi − pf ) (18.240)
i j=1
(2π)3
f

and several known factors. It is clear that dJ is not a relativistic invariant. However,
the wavefunctions of free particles involve the normalization factors (2E)−1/2 each of them
being squared also appears in the expression describing the quantum transition. Taking into
consideration this circumstance, we can always pass to the relativistically invariant phase
volume element of secondary (final) particles
nf
4 (4)
X X Y d3 p′j
dI = (2π) δ ( pi − pf ) (18.241)
i j=1
2Ej (2π)3
f

The formulae (18.169), (18.198) could be simplified by the assumption that the quantity
|Mi→f |2 is constant. Then dσ and dΓ become proportional to the relativistic phase vol-
umes of secondary particles. This assumption is called a statistical hypothesis or a phase
volume approximation. In the context of this approach, taking into consideration the four-
dimensional momentum conservation laws and not concretizing process dynamics, one could
estimate momentum-angle distributions of secondary particles.
The statistical hypothesis corresponds to such mechanisms of producing secondary par-
ticles in which dynamic correlations both between four-dimensional momenta of secondary
particles and between four-dimensional momenta of initial and final particles are absent.
Since the statistical hypothesis does not allow for interaction dynamics, then in the general
case its predictions are not justified in the experiments. However, under lack of knowl-
edge concerning a process dynamics such predictions give preliminary information about
momentum-angle distributions of secondary particles. In a number of cases the similar
approach allows formulation of the radically new and most effective methods: (i) for ex-
perimental search of new particles; (ii) for investigating properties of new particles and
processes. Thus, for example, calculating the group spectra of secondary particles in their
effective mass is realized on the basis of the statistical hypothesis, methods of experimental
resonances identification extensively relying on results of these calculations. The effective
mass distributions of the secondary particles group found in such a way effectively repro-
duce spectrum of background events or, at least, allow one to predetermine abilities of an
experimental facility to be required for solving problems appearing in physics of resonances.

© 2011 by Taylor and Francis Group, LLC

 S-matrix 463

As an example, let us consider the effective mass distribution of two particles in a three-
body system. The relativistically invariant phase volume element of a three-body state has
the form
d3 p1 d3 p2 d3 p3
dI3 ∼ δ (4) (p − p1 − p2 − p3 ) , (18.242)
E1 E2 E3
where pi and Ei (i = 1, 2, 3) is a three-dimensional momentum and an energy of the ith
particle in a final state. Since the phase volume model does not pretend to predict the
absolute values of the cross sections and the decay probabilities, numerical factors in the
expressions for the relativistic phase volumes are omitted. Comparison of the effective mass
distribution of two secondary particles, which was obtained on the basis of Eq. (18.242),
with the experimental data is put into practice only on the distribution form and not on
the absolute value. By this reason, it is convenient to normalize the distribution (18.242)
on a unit value of a total phase volume, that is, to calculate the quantity
dI3
dW3 = , (18.243)
I3
where Z
I3 = dI3 .

Introducing the effective mass of particles 1 and 2

p
s12 = (p1 + p2 )2

and integrating (18.242) over d3 p1 and d3 p2 , we obtain

d3 p3 √
dI3 ∼ W2 ( s12 , m1 , m2 ),
E3
where p
√ 2π [s12 − (m1 + m2 )2 ][s12 − (m1 − m2 )2 ]
W2 ( s12 , m1 , m2 ) = .
s12
As the phase volume element is a relativistic invariant, calculating the integral over d3 p
could be fulfilled in an arbitrary reference frame. Let us work in the center-of-mass system
of all three secondary particles. Inasmuch as in this system
√
E1 + E2 + E3 = s,
√
where s is the effective mass of three particles system, then E3 is defined by the equation
q √
E3 + E32 + s12 − m23 = s, (18.244)

whose root is as follows

s − s12 + m23
E3 = √ . (18.245)
2 s
Expressing components of the vector p3 through the spherical coordinates and changing
the variables from E3 to s12 with the help of Eq. (18.245), we get
d3 p3 |p3 |ds12 dΩ3
= |p3 |dE3 dΩ3 = √ , (18.246)
E3 2 s
where p √ √
q [s − ( s12 + m3 )2 ][s − ( s12 − m3 )2 ]
2 2
|p3 | = E3 − m3 = √ =
2 s

© 2011 by Taylor and Francis Group, LLC

 464 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
p √ √
[( s + m3 )2 − s12 ][( s − m3 )2 − s12 ]
= √ .
2 s
To integrate over the solid angle dΩ3 leads to the following expression for the effective mass
square events distribution to normalize on a unit
dW3 p
= [(s12 − (m1 + m2 )2 ][s12 − (m1 − m2 )2 ]×
ds12
p √ √
2π 2 [( s + m3 )2 − s12 ][( s − m3 )2 − s12 ]
× √ , (18.247)
ss12 I3 ( s, m1 , m2 , m3 )
where Z q √
√ 2π 2 ds12
I3 ( s, m1 , m2 , m3 ) = [( s + m3 )2 − s12 ]×
s s12
q √
× [( s − m3 )2 − s12 ][s12 − (m1 + m2 )2 ][s12 − (m1 − m2 )2 ]. (18.248)
The integration limits in Eq. (18.248) are given in the following way
√
(m1 + m2 )2 ≤ s12 ≤ ( s − m3 )2 ,

where the upper integration bound follows from the inequality E3 ≥ m3 , E3 being de-
termined by the expression (18.245). Normalization of (18.247) on a unit area under the
distribution
√ curve is usually performed by means of numerical calculations of the function
I3 ( s, m1 , m2 , m3 ) for the specific parameter values. From (18.247) it appears that the
function dW3 /ds12 turns into zero under the upper and lower limits of changing s12 and
has a smooth monotonous behavior in a region placed between zeros. The maximum of
dW3 /ds12 is situated somewhere in the middle of a possible values interval of s12 . As a
matter of principle, in the maximum region the form of dW3 /ds12 distribution has little in
common with the effective mass Breit-Wigner resonance distribution
dN 1
∼ , (18.249)
dm2 [(m2 − m2res )2 + (mres Γ/2)2 ]
P
where m2 = ( pf )2 , pf is a four-dimensional momenta of decay products (when comparing
Eq. (18.249) with Eq. (18.247), one should set m2 = s12 ). Moreover, the Breit-Wigner
peak always is at the same value of the effective mass equal the resonance mass s12 =
mres . The maximum of the phase curve dW3 /ds12 changes its position depending on s. In
that way, there is a fundamental capability to identify resonances against a background of
nonresonance statistical mass distributions.

© 2011 by Taylor and Francis Group, LLC

 19
Transmission of γ-radiation through matter

“Over the Mountains Of the Moon,

Down the Valley of the Shadow,
Ride, boldly ride,”
The shade replied,—
“If you seek for Eldorado!”
Edgar Allan Poe, “Eldorado”

The physics of elementary particles originally has dealt with electrons and γ-radiation, so
that here the greatest amount, both of the key ideas and of technical methods, have been
stored. This and the next sections will be devoted to the technical side of the QED, the
calculations of particular processes. For today the QED lost the status of the precise theory
in the sense that it is already not a complete theory. It, together with the weak interaction
theory, forms the electroweak interaction theory. If one considers some pure electrodynamic
process (there are only charged leptons and photons in the initial and final states) in the
electroweak theory, then, as a rule, diagrams, additionally in relation to the QED, appear
with virtual gauge bosons (W ± and Z), neutral leptons and quarks appear. For example,
in the second order of the perturbation theory, every time while the Feynman diagram has
an internal photon line, then the Z-boson contribution is also possible. Such processes as

e− e+ → e− e+ , µ+ µ− , τ + τ − ,

having the s-channel Feynman diagram will feel the Z-boson effect more strongly as the
energy in the center-of-mass frame is approached in the value mZ . There are, however,
restricted numbers of processes that proceed at the cost only of the electromagnetic inter-
action in the second order of the perturbation theory. Under transition to the third and
higher orders of the perturbation theory all electrodynamic processes include contributions
of the weak interaction. However, the QED is the most simple example of a gauge theory
of interacting fields. Understanding its structure will help us to further advance.

19.1 Photoeffect (nonrelativistic case)

In 1887 the phenomenon of electron emission by metallic sodium irradiated with a mercury-
vapor lamp was revealed by H. Hertz.∗ The process of light absorption with subsequent
electron emission was given the name photoeffect. For this phenomenon A. Stoletov (1888)
and F. Lenard (1902) have stated the following empirical laws: (i) a maximum kinetic
energy of photoelectrons Tmax is independent on the light intensity, (ii) Tmax is linearly

∗ Recall,
that H. Hertz has experimentally confirmed a wave nature of the electromagnetic radiation and
hence the validity of Maxwell’s electrodynamics.

465
© 2011 by Taylor and Francis Group, LLC
 466 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

dependent on the light frequency ω, and (iii) the photoeffect has a threshold frequency ω0 .
However, these laws could not be explained within the scope of classical physics.
A theory of the photoeffect was proposed in 1905 by Einstein who had logically developed
the Planck theory of light quanta. According to Einstein, light is emitted by quanta (in line
with the Planck hypothesis), its propagation being due to the quanta as well. Based on a
theory of the electromagnetic field energy quantization and assuming that at the photoeffect
a photon is interacting with an individual electron, Einstein has put forward an explanation
for the photoeffect mechanisms using a single equation:

~ω = ~ω0 + T, (19.1)

where ~ω0 is the photoelectric work function of electrons from matter, and, for the time
being, we cross to the ordinary system of units. Verifying Einstein’s equation, in 1916 R.
Millikan defined the quantity ~ that proved to be identical to the quantity found by Planck
in a theory of thermal radiation.
A free electron is unable to absorb a photon. This transpires from the fact that for
the transition of an electron from the state with the momentum pµ = (E/c, p) to the
state with the momentum p′µ = (E ′ /c, p′ ) in the absence of the third body (for example,
condensed medium, atom, etc.), the conservation laws for the energy and three-dimensional
momentum:
E − E ′ = ~ω, p − p′ = ~ω/c
are inconsistent at all electron speeds v < c. By the conventional terminology, the pho-
toeffect in a condensed medium is called the photoemissive effect, and the transition of an
electron from one of the atomic or molecular bound states to a continuous spectrum is
referred to as photoionization. The energy required to carry an electron from the ground
state to a continuous spectrum state with T = 0 is called atom ionization energy, and, when
it is measured in eV, this energy is designated ionization potential I. In a hydrogen-like
atom we have
mZ 2 e4
I= = Z 2 I0 .
2(4π)2 ~2
In multi electron atoms high energetic photons could knock off electrons from different
electron shells. Such atoms have several ionization potentials.
Another characteristic of photoeffect is the photoionization cross section σ. By definition
this is the ratio of ionization acts number per one atom in a unit of time and volume to an
intensity of the coming photon flux (flux is supposed to be monochromatic). The quantity
σ may be analytically calculated for the hydrogen atom and for a hydrogen-like atom with
Z ≪ 137.
First, we shall work out the cross section in a nonrelativistic case. So, in the initial state
there is one electron in the bound state with the energy Ei = −I and a photon with the
momentum k. The final state, belonging to a continuous spectrum, is defined by the mo-
mentum p and the energy Ef ≡ E. It is obvious that the process could be described with
the help of the concept of an external field entering into the electron Hamiltonian. The
electron-positron field operator may be expanded into the eigenfunctions of this Hamilto-
nian, the electron state concept in itself taking into account the electron interaction with
environmental objects. Under such an approach investigating the processes of emitting
and absorbing a photon using the scattering matrix of the first order becomes possible. It
should be emphasized that only the S-matrix element of the first order differs from zero for
the photoeffect. Therefore, if we use the exact wavefunctions of the electron in the nucleus
field, then the photoeffect cross section obtained in the first order of the perturbation theory
represents the exact expression.

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 467

Let ψi (x) = ψi (r)eiIt and ψf (x) = ψf (r)e−iEt be the wavefunctions of the initial and
final electron, Aµ (x) = Aµ (r)e−iωt be the photon wavefunction (we have come back to the
natural system of units). Then the matrix element of the operator S (1) could be represented
in the form:
(1)
Si→f = −2πiAi→f δ(−I + ω − E), (19.2)
where the amplitude of the photon absorption Ai→f is connected with the transition current:

jµ (x) = eψf (x)γµ ψi (x) = jµ (r)ei(I+E)t (19.3)

by the relation: Z
Ai→f = j µ (r)Aµ (r)d3 x, (19.4)

We choose the photon wavefunction in the three-dimensional transverse gauge, that is,
the four-dimensional polarization vector has the form eλµ (k) = (0, eλ (k)). Since the emitted
electron belongs to a continuous spectrum, then the photoeffect cross section will be given
by the formula:
d3 p
dσ = 2π|Ai→f |2 δ(−I + ω − E) , (19.5)
(2π)3
where the absorption amplitude has the appearance:
Z
1
Ai→f = −e √ ψ f (r)(γeλ (k))ψi (r)eikr d3 x. (19.6)
2ω

Under writing of Eq. (19.5) we have also allowed for the fact that the wavefunctions, both
of the electron and the photon, are normalized on one particle in the volume V = 1.
Let us fulfill calculations for two regions of the photon energy: for ω ≪ m and for ω ≫ I.
As I ∼ mZ 2 α2 ≪ m, then these two regions are partially overlapped under I ≪ ω ≪ m.
Therefore, investigation of these regions gives the complete pattern of the photoeffect in the
nonrelativistic approximation.
When ω ≪ m, then the electron speeds are nonrelativistic both in the initial state and
in the final one. That allows us to replace the quantity γ 0 γ = α with the nonrelativistic
speed operator v = −i∇/m in Eq. (19.6). Assuming that the photon wavelength is much
greater than the atom size, we may change the quantity eikr with 1. We also carry out the
transition to the spherical coordinate system with Oz||k:

d3 p = p2 d|p|dϕ sin θdθ = E|p|dEdΩ

and integrate (19.5) over E. After these operations we arrive at the expression for the
differential cross section:
m|p|
dσ = e2 2 |(eλ (k)Li→f )|2 dΩ, (19.7)
8π ω
where Z
1
Li→f =− (ψf† ∇ψi )d3 x (19.8)
m
(we have saved the minus sign in (19.8) for the sake of convenience only).
We shall be constrained by investigating the photoeffect for the ground level of a hydrogen
atom or a hydrogen-like ion. Then, the wavefunction of the initial electron is defined by
the expression:
ψi (r) = N e−ηr , (19.9)

© 2011 by Taylor and Francis Group, LLC

 468 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where s
Z 4π 1 Z3
η= , a0 = = , N= .
a0 me2 mα πa30
As far as the electron wavefunction in the final state is concerned ψf , then it must be
the electron wavefunction in the Coulomb’s nucleus field but belongs to continuous spec-
trum already. Recall, that in the centrally symmetric field alongside the system of the
wavefunctions for the discrete spectrum:

ψnlm = Rnl (r)Ylm (θ, ϕ), (19.10)

(Rnl (r) is the radial part of the wavefunction and Ylm (θ, ϕ) is the angle part of that) which
corresponds to the states with the definite energy, angular moment and its projection, we
have the system of the functions for the continuous spectrum associated with the state that
has the definite energy but does not possess any definite values both of the angular moment
and of its projection:
r ∞
1 πX l
ψp = i (2l + 1)e−iδl Rpl (r)Pl (nn1 ), (19.11)
p 2
l=0

where δl are scattering phases, Pl (nn1 ) are the Legendre’s polynomials, n = p/p, and
n1 = r/r. At infinity the system (19.11) describes the plane wave plus the convergent
spherical wave. In our case, thanks to the selection rules on l, the transition from the
ground state (the s-state) only to the p-state is possible. Keeping only the term with l = 1
and omitting insignificant phase factors, we rewrite the expression (19.11) in the form:
r
3 π
ψf = (nn1 )Rp1 (r). (19.12)
p 2
Now the quantity (eλ (k)Li→f ) to be of interest will look like:
Z Z Z
3η 5/2 3η 5/2
(eλ (k)Li→f ) = √ d3 x(nn1 )(eλ (k)n1 )e−ηr Rp1 (r) = √ dΩ′ dr×
2mp 2mp
Z
12πη 5/2 λ
×r2 (nn1 )(eλ (k)n1 )e−ηr Rp1 (r) = √ (e (k)n) drr2 e−ηr Rp1 (r). (19.13)
2pm
The radial wavefunctions of the continuous spectrum could be represented in terms of the
degenerated hypergeometric function (see, for example, Ref. [18]). We have for Rp1 (r):
s
2η 1 + ν2
Rp1 (r) = pre−ipr F (2 + iν, 4, 2ipr), (19.14)
3 ν(1 − e−2πν )

where ν = η/p. Taking into account the formulae

Z ∞
dze−βz z γ−1 F (α, γ, kz) = Γ(γ)β α−γ (β − k)−α , (19.15)
0
 iν
ν +i
= e−2ν/ arctan ν (19.16)
ν −i
and the fact that at n = 3, 4, ... Γ(n) = (n − 1)!, we get

27/2 πν 3 (eλ (k)n)e−2ν/ arctan ν

(eλ (k)Li→f ) = √ √ . (19.17)
pm(1 + ν 2 )3/2 1 − e−2πν

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 469

Gathering the obtained results, we find that the differential cross section of the photoeffect
with emitting the electron within the element of the solid angle dΩ is given by the expression:
 4 −4ν/ arctan ν
27 παa20 I e
dσ = (eλ (k)n)2 dΩ, (19.18)
Z2 ω 1 − e−2πν

where we have given the account of:

p2

ω =T +I = 1 + ν2 . (19.19)
2m
From Eq. (19.18) follows that the angle distribution of the photoelectrons is defined by the
factor (eλ (k)n)2 . It is maximum in the directions parallel to falling photons polarization
direction and turns to zero both in the directions to be perpendicular to eλ (k) and in the
incidence direction as well.
In the case of nonpolarized photons the formula (19.18) must be averaged over the po-
larization directions of falling photons that could be done with the help of the relation:
2  
1X λ λ 1 kj kl
ej el = δjl − 2 , (19.20)
2 2 ω
λ=1

where j, l = 1, 2, 3. To integrate the obtained cross section gives the total cross section of
the photoeffect for unpolarized particles [19]:
 4 −4ν/ arctan ν
29 π 2 αa20 I e
σ= . (19.21)
3Z 2 ω 1 − e−2πν

From Eq. (19.21) we notice that the photoeffect cross section tends to constant directly
under threshold. At ω → I the limiting value of σ is equal to:

29 π 2 αa20
σ= (19.22)
3e4 Z 2
(here e = 2.718...).
Under small excess ω above I, when still ω − I ≪ I (ν ≫ 1), we have:

σ ∼ (I/ω)8/3 .

When ω − I ∼ I (ν ∼ 1):
σ ∼ (I/ω)3 .
Away from the red border of the photoeffect ω ≫ I (however, ω ≪ mc2 is still fulfilled)
the Born’s approximation is applied (ν ≪ 1) and the formula (19.21) assumes the form:
 7/2
28 παa20 Z 5 I0
σ= . (19.23)
3 ω

In the case of nonrelativistic speeds of the photoelectrons the exact formulae have been
obtained for σ of excited atoms (transitions from states with n > 1, n1 is the main quantum
number). In the approximated calculations when n ≥ 1 and ω ≤ I one may use the quasi-
classical Kramers’s formula:  3
26 παa20 I
σK = 3/2 2 5 . (19.24)
3 Z n ω

© 2011 by Taylor and Francis Group, LLC

 470 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In multi electron atoms with the mean and large value of Z, the photoeffect takes place on
the electrons of internal shells mainly on K-electrons, when the photons energy lays in the
interval from 102 to 105 eV (roentgen radiation and γ-radiation). To get the total cross
section of the photoeffect on K-shell one should multiply the expression (19.21) by 2 as
this shell has two electrons. When we are increasing ω, then under reaching ω = IL (IL is
ionization potential of L-electrons) the photoeffect starts to go on for the L-shell electrons
and, as a result, σ exhibits jump-like growth.

19.2 Photoeffect (relativistic case)

We shall investigate the photoeffect in the relativistic region, that is, in the case when
the photon energy is great as compared with the K-electron energy (ω ≫ I). Now we
would take the solution of the Dirac equation in the Coulomb’s nucleus field as the electron
wavefunctions. However, since in the case under consideration Ef = ω −I ≫ I, then there is
no need to solve this problem exactly and it is not unreasonable to allow for the Coulomb’s
field using the perturbation theory (it is valid for nuclei with Zα ≪ 1).
Because the photoelectron could be relativistic, then its unperturbed wavefunction must
have the form of the relativistic plane wave, that is, ψf is represented by the expression:
r
(0) (1) m  
ψf = ψf + ψf = uf (pf )eipf r + ψ (1) (19.25)
Ef

(for the sake of convenience hereafter we supply all quantities concerning the initial and final
electron states by the corresponding index). In the initial state the electron is nonrelativistic,
nevertheless, as will be shown later, a relativistic correction of the order Zα must also be
included in its wavefunction ψi . We shall look for ψi in the following shape:
 
(0) (1) (1)
ψi = ψi + ψi = ui (0)ψS + ψi , (19.26)

where ψS is the Schrödinger equation solution for a bound state, ui (0) is a bispinor of the
view  
w
ui (0) = ,
0
and w is a bispinor to describe polarization state of an electron. Next, we set down the
Dirac equation for the electron in the Coulomb’s nucleus field:
 
Zα 0 0
Ei + − mγ + iγ (γ∇) ψi = 0, (19.27)
r

one acting on that by the operator:

 
Zα 0 0
Ei + + mγ − iγ (γ∇) .
r

The resulting equation will look like:

 
2Ei Zα Zα Z 2 α2
(∆ + p2i + 0
)ψi = iγ (γ∇) − 2 ψi . (19.28)
r r r

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 471

Substituting (19.26) into (19.28), neglecting the term being quadratic on the perturbation
and allowing for that ψS satisfies the Schrödinger equation, we arrive at the relation for
(1)
ψi :
 
1 Zα (1) iZα 0 1
∆ − |Ei | + ψi = γ (γ∇) ui (0)ψS , (19.29)
2m r 2m r
where Ei is the energy level of the initial state predicted by the Schrödinger theory. As is
easy to see the solution of Eq. (19.29) is given by the expression:

(1) i 0
ψi =− γ (γui (0)∇ψS ). (19.30)
2m
So, the wavefunction of the initial electron will look like:
 
i 0
ψi = 1 − γ (γ∇) ui (0)ψS . (19.31)
2m

The function ψi has been obtained for distances r ∼ (mZα)−1 on which the correction to
it had the order of Zα. Since the derivative of purely exponential function (19.9) (and the
correction term with it in Eq. (19.31) ) is always proportional to Zα, then the formula
obtained is suitable for any s-state in the case of any values of r. As will be illustrated
subsequently, only the small value of r proves to be essential in the problem in question.
This circumstance allows us to courageously use the formula (19.31).
Insert the obtained functions of the initial and final electrons into the transition amplitude
(19.6)
s Z     
1 i 0
Ai→f = −e d3 x uf (pf ) (γeλ (k)) 1 − γ (γ∇) ui (0)ψS ×
4ωEf 2m

(1)
o
×e−i(pf −k)r + ψ f (γeλ (k))eikr ui (0)ψS . (19.32)

Recall, that we are working in the first approximation on Zα. Then, to find the first
correction in the second addend of Ai→f , one is sufficiently constrained by only the first
term in the series expansion of ψS . That is equivalent to the replacement:
s
Z3
ψS → . (19.33)
πa30

Unfortunately, such a straightforward way does not achieve the object in the first addend
of the expression for Ai→f . Really, in this case this addend appears to be proportional to
δ (3) (pf − k) and to turn into zero at pf 6= k. At v ∼ 1 the relativistic correction to be
proportional to Zα gives the contribution to the cross section that has the same order as
the next term of the ψS expansion in terms of Zα. Using the Gauss theorem and passing
from configuration space functions to their Fourier images, we may rewrite the expression
for the photon absorption amplitude in the form:
s s
1 Z3 
Ai→f = −e uf (pf )(γeλ (k)) [1+
4ωEf πa30
 
1 0 −ηr

(1) λ
+ γ γ(pf − k) ui (0) e (pf −k)
+ (ψ f )k (γe (k))ui (0) , (19.34)
2m

© 2011 by Taylor and Francis Group, LLC

 472 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where the vector index denotes the corresponding Fourier image. From the equation:

e−ηr
(∆ − η 2 ) = −4πδ (3) (r), (19.35)
r
follows:  
e−ηr 4π
= . (19.36)
r q q2 + η2
To differentiate the obtained result with respect to the parameter η results in:
8πη
(e−ηr )q = . (19.37)
(q2 + η 2 )2
(1)
To find (ψf )k we set down the Dirac equation in the Coulomb’s field, substitute ψf into
that, and restrict ourselves only by the first order of smallness

(1) Zα 0
[γ 0 Ef + i(γ∇) − m]ψf = − γ uf (pf )eipf r . (19.38)
r
Acting upon both sides of this equation by the operator [γ 0 Ef + i(γ∇) + m], we get:

(1) eipr
(∆ + p2f )ψf = −Zα[γ 0 Ef + i(γ∇) + m]γ 0 uf (pf ) . (19.39)
r
In order to transfer to the Fourier images we multiply Eq. (19.39) by e−ikr and integrate it
over r. In doing so, we use the Gauss theorem in the terms involving the operators ∇ and
∆. The result will be as follows:
(1) 4π
(p2f − k2 )ψf = −Zα[2Ef γ 0 − (k − pf )γ]γ 0 uf (pf ) , (19.40)
(k − pf )2

where we have taken into consideration that the bispinor uf (pf ) obeys the equation:

[γ 0 Ef − (pf γ) − m]uf (pf ) = 0.

From (19.40) it follows:

 (1)
 2Ef γ 0 + (k − pf )γ 0
ψf = 4πZαuf (pf ) γ . (19.41)
q (k2 − p2f )(k − pf )2

Substituting the obtained Fourier images of the electron wavefunctions into the absorption
amplitude (19.34) leads to the relation:
r
π (Z/a0 )5/2
Ai→f = 4 uf (pf )Bui (0), (19.42)
ωEf (k − pf )2

where
B = a(γeλ (k)) + (γeλ (k))γ 0 (γb) + (γc)γ 0 (γeλ (k)),
1 Ef pf − k k − pf
a= + , b= , c= .
(k − pf )2 m(k2 − p2f ) 2m(k − pf )2 2m(k2 − p2f )
For the sake of simplicity we shall not be interested in the electron polarizations. Then,
we should sum up the cross section over the spin states of the final electron (sf ) and average

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 473

over the spin states of the initial electron (si ). As a result of such an operation the quantity
appears:
1 X 1
|uf (pf )Bui (0)|2 = Sp[(p̂f + m)B(γ0 + 1)B].
2 s ,s 8m
i f

To determine it, one must be able to find spurs of products of the γ 0 - and γ-matrices. Recall,
the mechanism of deriving the four-dimensional formulae for calculating spurs is based on
two facts only: (i) anticommutation of γ5 with all the γ-matrices; and (ii) commutation
relations. Using these in the three-dimensional case, we arrive at the following conclusions.
Only the spurs of products with even number of the factors γ 0 and γ are different from
zero. To calculate the spur one should collect all γ 0 in one place (that gives 1 for these
products) and use the four-dimensional formulae of Section 15.2 with replacing the metric
tensor g with the Kronecker symbol with the minus sign. With allowance made for these
simple modifications, we get:
1 X 1
|uf (pf )Bui (0)|2 == [apf − (b − c)(Ef + m)]2 +
2 s ,s 2m(Ef + m)
i f

2(beλ (k))
+ [(Ef + m)(ceλ∗ (k)) + a(pf eλ∗ (k))], (19.43)
m2
where B = γ 0 B † γ 0 .
To define the cross section behavior character depending on the photon polarizations, we
direct the axis z along k and choose the plane (k, eλ (k)) as the one xOz. Note, to transit to
the cross section for unpolarized photons, to overage over the photon polarizations, is put
into effect by integration over the angle ϕ from 0 to 2π and multiplication of the obtained
result by 1/2. Taking into consideration the relation:

(pf eλ (k)) = |pf | cos ϕ sin θ,

and the fact that in the relativistic case

ω = I + Ef − m ≈ Ef − m,

we get without trouble the formulae:

k2 − p2f = −2m(Ef − m), (k − pf )2 = 2Ef (Ef − m)(1 − v cos θ),

where v ≡ |v| is a photoelectron speed. By application, the obtained relations we find that
the differential cross section of the photoeffect for the polarized photons and unpolarized
electrons is given by the expression [20]:

3 2 2 2
( √
5 4 2 v (1 − v ) sin θ (1 − 1 − v 2 )2
dσ = Z α r0 √ (1−
(1 − 1 − v 2 )5 (1 − v cos θ)4 2(1 − v 2 )3/2
" √ # )
(1 − 1 − v 2 )(1 − v cos θ) 2
−v cos θ) + 2 − cos ϕ dΩ, (19.44)
1 − v2

where r0 = e2 /(4πm) is the classical radius of the electron.

In the ultrarelativistic case v√
tends to 1 and the cross section of the photoeffect has sharp
maximum at small angles θ ∼ 1 − v 2 , that is, the photoelectrons preferentially fly in the
direction of photons falling.

© 2011 by Taylor and Francis Group, LLC

 474 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

To integrate the expression (19.44) over angles and multiply the obtained result by 1/2
gives the total cross section of the photoeffect for unpolarized particles [20]:
" p !#
2 3/2 2−1
(γ − 1) 4 γ(γ − 2) 1 γ + γ
σ = 2πZ 5 α4 r02 + 1− p ln p , (19.45)
(γ − 1)5 3 γ+1 2γ γ 2 − 1 γ − γ 2 − 1
where
1
γ= √ .
1 − v2
In the ultrarelativistic case this formula takes the simple view:
σ = 2πZ 5 α4 r02 γ −1 . (19.46)
In the limit of small values of γ − 1 ≪ 1 we get, as was to be expected, the nonrelativistic
value of the cross section (19.23).

19.3 Compton-effect (unpolarized case)

One of the first to support Einstein’s idea concerning the light quanta was Compton, who in
1922 discovered a special type of electromagnetic radiation scattering. As a source, Comp-
ton has used an X-ray tube, where the thermal electrons outgoing from the cathode are
accelerated up to high speeds to be decelerated when impinging the anticathode. Electron
deceleration results from interactions between the electrons and a Coulomb nucleus field.
And the resultant electromagnetic radiation received the name bremsstrahlung. Recall that
the difference in the approaches to a nature of bremsstrahlung on the basis of the classi-
cal electromagnetic theory (CET) and the quantum theory (QT) is as follows. The CET
predicts initiation of continuous radiation on each collision in the process of the electron
deceleration, while the QT predicts the generation of a single photon with the energy ~ω
that is generally different for every collision act. In Compton’s experiment the radiation
emitted by the molybdenum anticathode of an X-ray tube was scattered by graphite. Scat-
tered radiation was incident upon the crystal of an X-ray spectrometer and next—upon an
ionization chamber (or photographic plate). Compton has studied the intensity of scatter-
ing spectra for various angles between the direction of the initial and scattered beams θ.
As it turned out, the lines of a scattering spectrum are shifted relative to the initial lines to
higher values of θ, that is, a change in the wavelength of scattered beams was growing, with
an increase of θ. The experimental results of Compton were first published in the Bulletin
of the National Research Council of the USA Academy of Sciences in October 1922. The
essence of this paper may be rendered as follows: X-ray beams represent a flux of photons
having a particular momentum, similar to any other particles, and the scattering act is an
elastic collision between a photon and an electron. Subsequently, not only is the scattering
process of X-ray radiation by matter electrons, but the scattering process of a photon with
any energy by an electron has been named the Compton effect as well.
Describing the Compton effect, for the sake of simplicity, we shall consider the electrons
being in graphite to be free. It is quite natural because the binding energies of the electrons
have the order of several eV, while X-ray beam photons used in the Compton experiment
had the energy equal to 17.5 keV.
We now proceed to the calculation of the cross section for the Compton effect
e− + γ → e− + γ. (19.47)

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 475

The corresponding Feynman diagrams in the second order of the perturbation theory are

p1 k2
k1 k2

f1
f2

p1 p2
k1 p2

FIGURE 19.1
The Feynman diagrams corresponding to the Compton effect.

displayed in Fig. 19.1. It should be noted that in the electroweak theory any additional
diagrams in relation to ones pictured in Fig. 19.1 do not appear. In the higher orders
of the perturbation theory the corrections of weak interaction come about from insertions
of vacuum loops both in the external and in the internal lines of the QED diagrams. The
W − νe -, He− and Ze− -loops represented in Fig. 19.2 may be placed in electron lines whereas
the W − W + , q i qi -loops displayed in Fig. 19.3—in photon lines. In accordance with the

W− H

e e e e
νe e

e e
e

FIGURE 19.2
The loops with the W − νe -, He− and Ze− -virtual pairs.

Feynman rules the amplitude of the process (19.47) takes the form:
r
2 m2
Ae− γ→e− γ = −ie u′ (p′ )e′∗ ′ µν
µ (k )O eν (k)u(p), (19.48)
4EE ′ ωω ′

© 2011 by Taylor and Francis Group, LLC

 476 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

W− qι

γ γ γ γ

W+ qι

FIGURE 19.3
The loops with the W − W + - and qi q i -virtual pairs.

where
fˆ1 + m ν fˆ2 + m µ
Oµν = γ µ 2 2
γ + γν 2 γ , (19.49)
f1 − m f 2 − m2
p = (E, p), p′ = (E ′ , p′ ), k = (ω, k), k ′ = (ω ′ , k′ ), eν (k) (e′ν (k′ )) is a four-dimensional
polarization vector of the initial (final) photon, and

f1µ = (p + k)µ = (p′ + k ′ )µ , f2µ = (p − k ′ )µ = (p′ − k)µ .

The four-dimensional vectors f1µ and f2µ represent the momenta of the virtual electrons.
The squares of these momenta are nothing but the Mandelstam’s variables:
2 2
f1µ = s, f2µ = u. (19.50)

To obtain the transition probability we must take the module square of the quantity
Ae− γ→e− γ . We are constrained by the case when the initial electron is unpolarized and we
are not interested in the spin projection of the final electron. Then the expression:

|Ae− γ→e− γ |2

should be summed up over the spin states of the final electron (s′ ) and averaged over the
ones of the initial electron (s), that is, the situation is reduced to calculating the quantity:
1X 2
|A − − | .
2 ′ e γ→e γ
s,s

Using the formula (15.103), we find:

1X e4
|Ae− γ→e− γ |2 = Sp[M(p̂ + m)M(p̂′ + m)], (19.51)
2 ′ 32EE ′ ωω ′
s,s

where M = e′∗ ′ µν
µ (k )O eν (k) and M = γ0 M γ0 .
†

With allowance made for antihermicity of γ and hermicity of γ0 , the expression under
the spur sign could be represented in the view:

1 ˆ 1
′
M(p̂ + m)M(p̂ + m) = γ µ
2
e′∗ ′
µ (k )(f1 + m)eν (k) + eµ (k)(fˆ2 +
s−m u − m2
 
′∗ ′ σ 1 ∗ ˆ ′ ′ 1 ˆ
+m)eν (k )] (p̂+m)γ e (k)(f1 + m)eτ (k ) + e (k )(f2 + m)eτ (k) γτ (p̂′ +m).
′ ′ ∗
s − m2 σ u − m2 σ
(19.52)

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 477

We next assume that the incoming and scattered photons also are unpolarized. Then the
expression (19.51) should be summed up over the final photon states (λ′ ) and averaged over
the initial ones (λ). Carrying out this operation with the help of formula (17.44), we get:

1 X e4
|Ae− γ→e− γ |2 = Sp F, (19.53)
4 64EE ′ ωω ′
s,s′ ,λ,λ′

where  
1 1 µ ˆ ν 1 ν ˆ µ
F = γ (f 1 + m)γ + γ (f 2 + m)γ ×
s − m2 s − m2 u − m2
×(p̂ + m)γν (fˆ1 + m)γµ (p̂′ + m)+
 
1 1 µ ˆ 1
+ γ (f1 + m)γ +ν
γ (f2 + m)γ (p̂ + m)γµ (fˆ2 + m)γν (p̂′ + m).
ν ˆ µ
u − m2 s − m2 u − m2
(19.54)
It is easy to verify that the second addend in this expression follows from the first one
under replacement:
k ↔ −k ′ , (19.55)
which, in its turn, is equivalent to:

f1 ↔ f2 , s ↔ u. (19.56)

The pointed linkage between the addends in the expression (19.54) is the crossing-universality
consequence of the transition amplitude. In this context it means that in the case of un-
polarized photons the matrix element corresponding to the diagram pictured in Fig. 19.1a
Ma follows from the one pictured in Fig. 19.1(b) Mb under changing:

k ↔ −k ′ , eµ (k) ↔ e′µ (k′ ). (19.57)

Having set down the amplitude square:

|Ae− γ→e− γ |2 = Ma Ma + Ma Mb + Mb Ma + Mb Mb , (19.58)

| {z } | {z }
I II
we are convinced that at the replacement (19.56) the third and forth addends pass into
the second and first ones, respectively. This property of the cross section still stands for
identical polarized photons too. In the case when the process is described by 3 topolog-
ically nonequivalent diagrams possessing the crossing-universality, the cross section could
be represented as a sum of three addends passing into each other under the particular
transformation of particle momenta.
It is obvious that the both addends in (19.53) must be functions of two invariants s and
u only. Having realized this circumstance, we can exhibit Sp F in the following form:

Sp F = P (s, u) + P (u, s), (19.59)

where
P (s, u) = h1 (s, u) + h2 (s, u), (19.60)
1
h1 (s, u) = Sp[γ µ (fˆ1 + m)γ ν (p̂ + m)γν (fˆ1 + m)γµ (p̂′ + m)], (19.61)
(s − m2 )2
1
h2 (s, u) = Sp[γ ν (fˆ2 + m)γ µ (p̂ + m)γν (fˆ1 + m)γµ (p̂′ + m)]. (19.62)
(s − m2 )(u − m2 )

© 2011 by Taylor and Francis Group, LLC

 478 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In what follows the formulae for convolutions over four-dimensional indices will be useful:

γ µ γµ = 4, 

γ µ γ ν γµ = −2γ ν ,



µ ν τ ντ
γ γ γ γµ = 4g , . (19.63)
γ µ γ ν γ τ γ σ γµ = −2γ σ γ τ γ ν ,




γ µ γ ν γ τ γ σ γ ρ γµ = 2(γ ρ γ ν γ τ γ σ + γ σ γ τ γ ν γ ρ ).


Using (15.22), (15.25), and (19.63), we come to the expression for h1 (s, u):
 
4 1 ′
h1 (s, u) = Sp fˆ1 p̂fˆ1 p̂′ − 4m2 (fˆ1 p̂ + fˆ1 p̂′ − f12 − 4
p̂p̂ ) + 4m =
(s − m2 )2 4

4m4 + 2m2 (s − m2 ) − (s − m2 )(u − m2 )

=8 . (19.64)
(s − m2 )2
The similar calculations yield:

2m2 + s + u
h2 (s, u) = 8m2 . (19.65)
(s − m2 )(u − m2 )

Gathering together the necessary relations, we get the final expression for the spur F :
    
m2 m2 m2 m2 u − m2 s − m2
SpF = 32 + + +1 −8 + .
s − m2 u − m2 s − m2 u − m2 s − m2 u − m2
(19.66)
Using the general formula (18.185), one may now write the differential cross section of
the Compton effect:
 −1
e4 ω ′2 ∂(ω ′ + E ′ )
dσ = (2π)−2 Sp F dΩ, (19.67)
64E ′ ω ′ (pk) ∂ω ′

where dΩ is a element of the solid angle in which the vector k′ lays (the z-axis is directed
along the vector k).
We shall investigate the expression:

∂(ω ′ + E ′ )
.
∂ω ′
Since the momentum p′ is connected with k′ by the conservation law:

E ′2 = m2 + (p + k − k′ )2 , (19.68)

then E ′ is a function of ω ′ . Calculations are easiest fulfilled in the lab frame, in which the
electron is initially at rest (p = 0, and E = m). Then we have:

∂(ω ′ + E ′ ) ω ′ − ω cos θ E ′ + ω ′ − ω cos θ m + ω(1 − cos θ)

= 1 + = = =
∂ω ′ E′ E′ E′
mω ′ + ω ′ ω(1 − cos θ) mω ′ + kk ′ (pk ′ ) + (kk ′ ) (p′ k ′ ) s − m2
= = = = = . (19.69)
ω′E ′ ω′E ′ ω′E ′ ω′E ′ 2ω ′ E ′
To substitute (19.69) into the expression (19.67) results in:

ω ′2 dΩ
dσ = r02 Sp F. (19.70)
4(s − m2 )2

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 479

Note, that the obtained cross section is a relativistic invariant really. The invariance of
Sp F is evident while the invariance of the factor ω ′2 dΩ follows from the identity:
Z 3 Z
d k
ω ′2 dΩ = = 2 d4 kδ(k 2 ). (19.71)
ω′
In the center-of-mass frame E = E ′ , ω = ω ′ and the Mandelstam’s variables takes the
values: 
s = (E + ω)2 , t = k 2 + k ′2 − 2kk ′ = −2ω 2 (1 − cos θ),
(19.72)
u = m2 − 2pk ′ = m2 − 2ω(E + ω cos θ).
From s + t + u = 2m2 and the relations (19.72) we find without trouble:

(s − m2 )
ω= √ , (19.73)
2 s

that allows us to express d(cos θ) in terms of dt

1 2s
d(cos θ) = dt = dt. (19.74)
2ω 2 (s − m2 )2

As a result, we arrive at the formula for the differential cross section:

  
dσ 8πr02 m2 m2 m2 m2 m2
= + + +1 −
dt (s − m2 )2 s − m2 u − m2 s − m2 u − m2
 
1 s − m2 u − m2
− + . (19.75)
4 u − m2 s − m2
By application of (19.75) it is an easy matter to express the Compton effect cross section
in terms of the collision parameters in any particular reference frame.
Let us find the dependence of scattered photon frequency on incoming photon frequency.
Multiplying the four-dimensional momentum conservation law:

pµ + kµ = p′µ + kµ′ , (19.76)

by kµ , we get:
(pk) = (p′ k) + (kk ′ ). (19.77)
Having rewritten (19.76) as:
pµ − kµ′ = p′µ − kµ (19.78)
and having raised to the second power both sides of the obtained expression, we get:

(p′ k) = (pk ′ ). (19.79)

Insertion of (19.79) into (19.77) leads to the Compton formula:

ω
ω′ = , (19.80)
1 + ω(1 − cos θ)/m

where θ is an angle between k and k′ . At θ = 0, ω ′ = ω and scattering, consequently,

is absent. Changing the scattered photon frequency is maximum when θ = π, and, in
accordance with (19.80), proves to be equal to:
∆ω 2ω
= . (19.81)
ω′ m

© 2011 by Taylor and Francis Group, LLC

 480 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

From (19.81) follows that the frequency shift becomes significant only at the photon energies
to be commensurable with m = 0.511 MeV.
The Compton effect cross section takes the most simple view in the lab frame. Allowing
for that in this case
s = m2 + 2mω, u = m2 − 2mω ′ ,
from (19.75) we obtain the well-known Klein-Nishina formula [21]:
 2  
dσ r2 ω′ ω′ ω
= 0 2
+ ′ − sin θ . (19.82)
dΩ 2 ω ω ω

When m → ∞ the cross section (19.82) turns into zero, that is, scattering does not
happen. This situation is very instructive. If we are using this series of the perturbation
theory, then we shall have to make some conclusions about the character of the physical
process only after some terms of this series have been investigated. For the case under
study it means just one—the photon scattering process by an infinitely heavy charge with
the spin 1/2, exemplified by the atom nucleus, does not go on in the second order of the
perturbation theory. The analysis shows, the scattering matrix of the forth order of the
perturbation theory already contains elements to describe such a photon scattering.
If the experiment is aimed at measuring the angle distribution of scattered electrons,
then the expression for the cross section (19.82) should be transformed so that it will be
a function of incoming photon frequency and the photon scattering angle only. Using the
Compton formula, we obtain
 
dσ r02 1 + cos2 θ ǫ2 (1 − cos θ)2
= 1+ , (19.83)
dΩ 2 [1 + ǫ(1 − cos θ)]2 (1 + cos2 θ) [1 + ǫ(1 − cos θ)]

where ǫ = ω/m. One is easily convinced that at small energies of the photon (ǫ ≪ 1) the
formula passes to the classical expression by Thomson for the differential cross section on
a free rest electron:
dσ r2
= 0 (1 + cos2 θ). (19.84)
dΩ 2
One may measure the dependence of the cross section on the angle between the momenta
of the scattered electron and the incoming photon (β). It is readily shown that the angles
β and θ are connected by the relation:
s
1 − cos θ
cos β = (1 + ǫ) . (19.85)
2 + ǫ(ǫ + 2)(1 − cos θ)

Thus, β is changed from π/2 at θ = 0 up to 0 at θ = π. By means of (19.85) the differential

cross section is rewritten in the form:

dσ (1 + ǫ)2 cos β (1 + ǫ)2 sin2 2β
= 4r02 1 − +
dΩe (1 + 2ǫ + ǫ2 sin2 β)2 2[1 + ǫ(ǫ + 2) sin2 β]2

2ǫ2 cos4 β
+ , (19.86)
(1 + 2ǫ + ǫ2 sin2 β)[1 + ǫ(ǫ + 2) sin2 β]
where dΩe is an element of the solid angle in which the scattered electron momentum lays.
To get the total cross section through invariant quantities we should integrate the differ-
ential cross section (19.75) over t from −(s − m2 )2 /s to 0. We emphasize that the energy

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 481

of colliding particles is considered to be given, that is, s = const. Elementary integration

gives the result:
  
2πr02 4 8 1 8 1
σ= 1 − − 2 ln(1 + x) + + − , (19.87)
x x x 2 x 2(1 + x)2
where
s − m2
x= .
m2
In the ultrarelativistic region (x ≫ 1) the cross section has the shape:
  
2πr02 1 ln x
σ= ln x + + O . (19.88)
x 2 x
In the nonrelativistic case (x ≪ 1) the expression (19.87) is reduced to the classical formula
by Thomson:
8πr02
σ= (1 − x). (19.89)
3
In the lab frame we have x = 2ω/m and the formulae (19.87)–(19.89) define the de-
pendence of the Compton effect cross section on the photon energy in the case of resting
electron.
The Compton effect may be investigated not only in experiments with a fixed target but in
e− γ-colliders as well, that is, in the colliding beams mode. In the first case all measurements
are fulfilled in the lab frame while in the second one—in the center-of-mass frame. To be
more specific, in the latter the colliding e− - and γ-beams intersect at a small angle. Under
data processing of such experiments all kinematic characteristics of particles are transferred
by the Lorentz transformation into the center-of-mass frame where data analysis is carried
out. So, we are assured once again that writing of a cross section in terms of invariant
quantities is expedient. Having determined the values of s, u, t in the lab frame and in the
center-of-mass frame, we obtain the cross sections both for colliding beams mode and for
the fixed target mode.
Let us find out what laws of the Compton effect will be registered by observers to be
situated in the lab frame and in the center-of-mass one. In this case we are limited by the
ultra-relativistic region. From the formula (19.88) it follows that the behavior character of
the total cross section in the lab frame and in center-of-mass one is identical. In the both
frames the total cross section is decreased as the energy grows. However, in the lab frame
this decrease occurs according to the law:
ω
σ ∼ ω −1 ln
m
while in the center-of-mass frame (x ≈ 4ω 2 /m2 ) the cross section is decreased somewhat
faster
ω
σ ∼ ω −2 ln .
m
The analysis of the formula (19.75) shows that in the ultrarelativistic case the angle
distribution in both of these frames has an absolutely different character. In the lab frame
the differential
p cross section has a sharp maximum in the forward direction. In a narrow
cone θ  m/ω the relation:
ω ≈ ω′
2
p ≈ r0 at θ → 0. Beyond this
is fulfilled and the cross section goes into its maximum value
cone the cross section is decreased and in the region θ ≫ m/ω, where
m
ω′ ≈ ,
1 − cos θ

© 2011 by Taylor and Francis Group, LLC

 482 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

it becomes equal to:

dσ r2 m
= 0 , (19.90)
dΩ 2 ω(1 − cos θ)
that is, it is lowered by a factor equal to ∼ ω/m.
In the center-of-mass frame the situation is quite the reverse. Here, the differential cross
section possesses the maximum in the back direction. With the help of Eq. (19.72) one
may demonstrate that at π − θ ≪ 1 the invariants entering into (19.75) are defined by the
formulae:
s − m2 4ω 2 s + t − m2 ω2
≈ , ≈ 1 + (π − θ)2 . (19.91)
m2 m2 m2 m2
Then, we have:

m2 dt r2 dΩ
dσ ≈ 8πr02 = 0 . (19.92)
4(s − m2 )(s 2
+t−m ) 2 1 + (π − θ)2 ω 2 /m2

From (19.92) it is apparent that into a narrow cone π − θ  m/ω:

dσ
∼ r02 ,
dΩ
and beyond it the differential cross section is decreased in ∼ ω 2 /m2 times.

19.4 Compton-effect (polarized case)

In the case of arbitrary polarization states the matrix element square of the Compton
scattering |Me− γ→e− γ |2 must be subjected to the replacement:

e 4 m2 h
µν (+) γ λτ
i
|Me− γ→e− γ |2 → ′ ′
Sp ρ(+)′ ργ′
τ µO ρ ρνλ O . (19.93)
4EE ωω
We shall use the photon density matrix in the four-dimensional form. We introduce four
mutually orthogonal four-dimensional vectors:

Pτ , Nτ , Kτ , Mτ , (19.94)

where
2pK
Pτ = pτ + p′τ − Kτ , Nτ = ετ µνλ P µ M ν K λ ,
K2
Kτ = kτ + kτ′ , Mτ = kτ′ − kτ .
It is easy to check that the vectors Pτ , Nτ are orthogonal to the ones Kτ , Mτ , and, as a
result, they are orthogonal to k, k ′ as well. Since K is the time-like vector (K 2 = 2kk ′ > 0),
then in the reference frame, in which K = 0, it follows from KP = 0 and KN = 0 that

P0 = 0, N0 = 0,

that is, the vectors Pτ and Nτ are space-like. From Pτ and Nτ one may construct the two
unit vectors:
Nµ Pµ
n(1)
µ = √ , n(2)
µ = √ , (19.95)
−N 2 −P 2

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 483

the former is the pseudovector while the latter does the (proper) vector. On the strength
of these properties the vectors (19.95) may be used to build the four-dimensional density
matrix of the initial and final photon states. It should be stressed that in the most general
case there must be four such vectors rather than two ones.
Of course, one may also use the three-dimensional form of writing the photons density
matrix. In this case one conveniently chooses the vectors e1 and e2 being transverse in
reference to the photon momenta:

[kk′ ] 1 1 ′ ′1
e1 = , e2 = [ke1 ], e′1 = e1 , e′2 = [k e ]. (19.96)
|[kk′ ]| ω ω′

Then the vectors e1 and e′1 will be perpendicular to the scattering plane whereas the vectors
e2 and e′2 will lay in this plane. However, in the three-dimensional case, which is without
doubt more simple from the point of view of calculations, the final result will not have the
invariant form.
To determine the photon density matrices being related to the four-dimensional unit
(1) (2)
vectors nµ and nµ we must first establish the linkage of these vectors with the photon
polarizations. Independent directions of polarizations for each of photons are defined by the
components of the three-dimensional vectors n1 and n2 , which are transverse in relation
both to k and to k′ . Both in the center-of-mass frame and in the lab one the vector P lays in
the plane (k, k′ ) while the vector N is perpendicular to this plane. Therefore, the direction
n1 makes sense of the polarization to be perpendicular to the scattering plane whereas the
direction n2 —of polarization in the scattering plane (see, for comparison, Eq. (19.96) ).
One should also take into consideration that the Stoke’s parameters are defined with respect
to the Cartesian coordinates system with the z-axis to be parallel to the three-dimensional
photon momentum. The vectors N, P⊥ , k constitute such a system for the initial photon,
and the vectors N, −P′⊥ , k′ (P⊥ and P′⊥ are components of P to be perpendicular to k
and k′ , respectively)—for the final photon. The sign change of n(2) in the photon density
matrix (18.229) is equivalent to the one of ξ1 and ξ2 . Thus, the required density matrices
of the initial and final photons are given by the expressions:
1 (l′ )
(ργi )µν = [1 + (ξ i · σ)]ll′ n(l)
µ nν , ξ i = (ξi1 , ξi2 , ξi3 ), (19.97)
2
1 (l′ )
(ργd )µν = [1 + (ξ d · σ)]ll′ n(l)
µ nν , ξ d = (−ξd1 , −ξd2 , ξd3 ), (19.98)
2
where l, l′ = 1, 2, and we have emphasized the fact that the polarization matrix of the final
photon is dictated by detector characteristics (ρ′γ ≡ ργd ).
We should also like to expand the second rank tensor Oµν entering into the matrix element
(1) (2)
in terms of nµ , nµ . This tensor could always be built from products of four-dimensional
vectors to govern reaction kinematics. For such vectors we can take the four mutually
orthogonal vectors (19.94). However, the tensors ργ′ and ργ contain only the components
N and P . Then, under construction of Oµν it is sufficient to make use of the vectors N
and P only (addends to be proportional to M and K will simply turn into zero). Sorting
(1) (2)
out all possible combinations of nµ and nν , we arrive at the expression for the tensor
µν
O , which is similar, in its structure, to the expression for the photon density matrix in
the four-dimensional form (18.229):

Oµν = O0 (n(1) (1) (2) (2) (1) (2) (2) (1) (1) (2) (2) (1)
µ nν + nµ nν ) + O1 (nµ nν + nµ nν ) − iO2 (nµ nν − nµ nν )+

+O3 (n(1) (1) (2) (2)

µ nν − nµ nν ). (19.99)

© 2011 by Taylor and Francis Group, LLC

 484 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The quantity Oµν enters into the square of the matrix element as u′ (p′ )Oµν u(p). The
inclusion of this circumstance proves to be very productive under the determination of the
explicit form of the quantities Oi (i = 0, 1, 2, 3). Since:

u′ (p′ )(p̂ + p̂′ )u(p) = 2mu′ (p′ )u(p)

takes place, then Oi may involve the matrices K̂ rather than the ones (p̂+ p̂′ ). The quantities
O0 and O3 are scalars in the same sense as Oµν is a four-dimensional tensor. Therefore,
they have the following structure:

O0 = c0 + c′0 K̂, O3 = c3 + c′3 K̂. (19.100)

In its turn, O1 and O2 must be pseudoscalars and contain the matrix γ5 alongside the
matrix K̂:
O1 = γ5 (c1 + c′1 K̂), O2 = γ5 (c2 + c′2 K̂). (19.101)
Taking into account orthonormalization of the vectors n(1) and n(2) , we may express Oi
in terms of projections of the tensor Oµν . For example,
1 µν (1) (1) 1 µν (1) (1)
O0 = O (nµ nν + n(2) (2)
µ nν ), O3 = O (nµ nν − n(2) (2)
µ nν )
2 2
an so on. In order to facilitate calculations the tensor Oµν should be rewritten in terms of
the mutually orthogonal four-dimensional vectors (19.94):
′
p̂ + k̂ + m ν ν p̂ − k̂ + m µ
Oµν = γ µ γ + γ γ =
s − m2 u − m2
P̂ + 2m ν ν P̂ + 2m 1 µ 
= γµ γ + γ γ µ
− γ K̂γ ν
− γ ν
K̂γ µ
. (19.102)
2(s − m2 ) 2(u − m2 ) t
After all these preliminary operations we can embark on determination of Oi . To define
O0 and O3 we have the equations:
 
u′ (p′ )(c0,3 + c′0,3 K̂)u(p) = −u′ (p′ )Oµν N −2 Nµ Nν ± P −2 Pµ Pν u(p), (19.103)

where Oµν is given by Eq. (19.102). Let us use the relations:

γ µ γ ν cµ cν = c2 , γ µ γ τ γ ν cµ cν = 2ĉcτ − γ τ c2 , (19.104)

where cµ is an arbitrary four-dimensional vector, in the right-hand side of Eq. (19.103).

Further, we allow for orthogonality of the four-dimensional vectors Pµ , Nµ , Kµ and the
Dirac equation. Then, the comparison of the left-hand and right-hand sides of Eqs. (19.103)
yields:
1
O0 = −ma+ , O3 = ma+ + a− K̂, (19.105)
2
where
1 1
a± = 2
± .
s−m u − m2
The similar calculations result in:
i
O1 = a+ γ 5 K̂, O2 = −ma+ γ 5 . (19.106)
2
With the help of equalities:

γ µ = γµ , γµ γν ...γλ = γλ ...γν γµ , γ 5 = −γ5 , γ5 γµ = γ5 γµ ,

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 485

it is easy to verify that the tensor components O µν (O µν = γ 0 Oµν

†
γ0 ) follow from the ones
Oµν by means of the replacement Oi → O i , where

O0 = O0 , O 1 = −O1 , O2 = −O2 , O 3 = O3 , (19.107)

and simultaneous rearrangement of the indices µν.

Usage of the expressions (19.97)–(19.99), (19.105)–(19.107) allows one to get the result:

e 4 m2 h
(+)′ γ µν (+) γ T λτ
i
|Me− γ→e− γ |2 = Sp ρ (ρ )
d τµ O ρ (ρ )
i νλ O =
4EE ′ ωω ′
e 4 m2 n
= Sp (ρ(+)′ O0 ρ(+) O0 + ρ(+)′ Oρ(+) O) + (ξ i + ξ d )(ρ(+)′ O0 ρ(+) O+
4EE ′ ωω ′
+ρ(+)′ Oρ(+) O 0 ) − i(ξi − ξ d )[ρ(+)′ O × ρ(+) O] + (ξ i ξ d )(ρ(+)′ O0 ρ(+) O0 − ρ(+)′ Oρ(+) O)+
+ρ(+)′ (ξ d O)ρ(+) (ξ i O) + ρ(+)′ (ξ i O)ρ(+) (ξ d O) − i[ξi × ξ d ](ρ(+)′ O0 ρ(+) O−
o
−ρ(+)′ Oρ(+) O0 . (19.108)

Substituting (19.108) into the general expression for the scattering cross section (18.185)
with dΩ = dϕ sin θdθ, where ϕ is a scattering plane azimuth, we deduce the formula making
it possible to find the scattering cross section of polarized photons by polarized electrons.
We shall use it to calculate the scattering cross section of polarized photons by unpolarized
electrons. In this case we should set
p̂ + m p̂′ + m
ρ(+) = , ρ(+)′ =
2m 2m
in the expression (19.108) and double the result accordingly summation over the final elec-
tron polarizations. In the lab frame the differential cross section becomes the form [22]:
 2
dσ r2 ω ′
= 0 {F0 + F3 (ξi3 + ξd3 ) + F11 ξi1 ξd1 + F22 ξi2 ξd2 + F33 ξi3 ξd3 } , (19.109)
dϕd cos θ 4 ω
where
ω ω′
F0 = + − sin2 θ, F3 = sin2 θ,
ω′ ω
 
ω ω′
F11 = 2 cos θ, F22 = + cos θ, F33 = 1 + cos2 θ.
ω′ ω
The expression (19.109) holds implicit dependence on the scattering plane azimuth through
the Stokes parameters, since these parameters are defined with respect to the x−, y−, z-
axes (recall, that the x-axis is common for both photons and perpendicular to the scattering
plane while the y-axis lays in the scattering plane). Having set ξ d = 0 and having doubled
the result, we could obtain the differential cross section of scattering for the polarized initial
photon (all other particles to participate in the process are unpolarized). Denoting it as
dσ(ξ i ), we arrive at:
 2
r02 ω ′
dσ(ξ i ) = F dΩ, (19.110)
2 ω
where
ω ω′
F = F0 + ξi3 F3 = + − (1 − ξi3 ) sin2 θ. (19.111)
ω′ ω
From the obtained expression follows that the photons polarized perpendicular to the scat-
tering plane (ξi3 = 1) are scattered stronger than the ones polarized in the scattering plane

© 2011 by Taylor and Francis Group, LLC

 486 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

(ξi3 = −1). Since the cross section does not include ξi2 , then the dependence on the circular
polarization is absent. When the linear polarization of the initial photon in relation to the
x- or y-axes is missed or, again, when there is polarization in relation to directions to make
the angle equal to 450 with these axes, then the scattering cross section coincides with the
one for the unpolarized photons.
The scattering cross section of the unpolarized initial photon with detecting the final
polarized photon, which follows from (19.109) at ξi = 0:
 2
r02 ω ′
dσ(ξ d ) = [F0 + ξd3 F3 ]dΩ, (19.112)
4 ω
possesses the similar properties. One may also set the task of determining the photon
polarizations that arise precisely due to scattering by the electron. To do it we should first
find the photon polarization matrix in the final state, and thereafter ξf could be obtained
by application of the formula:
ξ f = Sp(ργf σ).
However, one may avoid new calculations and use the available expression (19.109).
Polarization states of particles are completely defined by the Stokes parameters ξ. Let
us assume that we are interested in the initial photon polarization and the final photon
polarization to be selected by a detector. It is obvious that the square |Mi→f |2 is linear
on each of the vectors ξ i and ξd . Then, |Mi→f |2 being the function of ξd should have the
following form:
|Mi→f |2 = a + (bξ d ), (19.113)
where a and b, in turn, are linear functions of ξ i . On the other hand, according to (18.238)
the quantity |Mi→f |2 is represented in terms of the density matrices of the initial and final
photons by the formula:

|Mi→f |2 ∼ Sp(ργd ργf ) = 1 + (ξ d ξ f ). (19.114)

To compare the expressions (19.113) and (19.114) leads us to the conclusion:

b
ξf = . (19.115)
a
Thus, having calculated the cross section as a function of the parameters ξ d , we can easily
determine the photon polarization state in which the photon is brought directly by the
scattering process. It should be stressed that the formula (19.115) remains valid for electrons
too.
Using (19.115), we find the Stokes parameters for the final photon:
2ξi1
ξf 1 = cos θ,
F
 
ξi2 ω ω′
ξf 2 = cos θ + , (19.116)
F ω′ ω
1
[sin2 θ + (1 + cos2 θ)ξi3 ].
ξf 3 =
F
From the obtained formulae one may make the conclusion that in the scattering result
the unpolarized photon goes partially polarized. Really, setting ξi1 = 0 into (19.116), we
get:
sin2 θ
ξf 1 = ξf 2 = 0, ξf 3 = ω ω′ 2 . (19.117)
ω ′ + ω − sin θ

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 487

As ξf 3 > 0, then the photon is polarized perpendicularly to the scattering plane. In the
nonrelativistic approximation:
sin2 θ
ξf 3 = (19.118)
1 + cos2 θ
and the radiation is completely polarized at the scattering angle equal to 900 .
From (19.116) it also follows that the circularly polarized photon is produced only in the
case when the initial photon was circular polarized.
The spurs calculation in the formula (19.108) are considerably complicated for polarized
electrons. Since these calculations do not involve any principal difficulties, then, when
desired, the reader can independently carry them out addressing the original literature to
verify obtained results (the complete summary of the results may be found in the review
articles [23]).

19.5 Electron-positron pair production by the photon in the

nucleus field
One more process, which happens under photon motion in matter, is photon “materializa-
tion” through e− e+ -pair production. It is clear that the energy to be necessary for this
must be no less than 2m. One photon having the energy even greater than 2m is not in a
position to produce the e− e+ -pair, as the conservation laws of the momentum and energy
are not being fulfilled simultaneously. To create the e− e+ -pair by one photon the presence
of an extraneous particle is needed. The role of such a particle may play an atomic nucleus.
Since an action of the heavy nucleus could be allowed for as one of an external field, then
the amplitude of the given process is defined by the expression (19.2) as in the case of the
photoeffect. The final state is described by the product of the electron and positron wave-
functions belonging to a continuous spectrum. The electron wavefunction in its asymptotic
form must contain a convergent spherical wave in parallel with a plane wave while the
positron wavefunction, apart from a plane wave, must involve a divergent spherical wave.
In view of the complexity of these functions one manages to exactly calculate the cross
section of the pair production in two cases only: (i) in a region of small energies, when the
nonrelativistic approximation is applicable; (ii) in the ultrarelativistic case.
However, if the problem conditions are such that the Born’s approximation applicability
criterion for the Coulomb’s field is met:
Zα
≪ 1, (19.119)
v±

where v− (v+ ) is a velocity of the electron (positron), then in order to find the cross section
of the e− e+ -pair production process one sufficiently calculates the corresponding element
of the scattering matrix in the second order of the perturbation theory. We shall be limited
by the investigation of just that very case.
Since in our approach the electric nucleus field is considered as a pseudophotons plurality,
then this process could schematically be represented in the form:

γ + γ̃ → e− + e+ , (19.120)

where hereafter the symbol γ̃ is used for designation of an external electromagnetic field.
Remember that an external field is a nonquantized object, that is, it is not associated with

© 2011 by Taylor and Francis Group, LLC

 488 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

any production and annihilation operators. The diagrams pictured in Fig. 19.4, where k
is the four-dimensional photon momentum, p− and p+ are the four-dimensional momenta
of the electron and the positron, and q is the momentum transferred nucleus, correspond
to the process (19.120). Attention must be given to the minus sign in front of the four-

q p− k p−

k −p+ q −p+

FIGURE 19.4
The Feynman diagrams describing the e+ e− -production by the photon in the nucleus field.

dimensional positron momentum. Its appearance is obligated to the sign in the exponent
that accompanies the production operator of the positron in the final state. We emphasize
that in the momentum representation on the Feynman diagrams a negative momentum is
assigned to all antiparticles not only in the QED but also in any gauge theory.
In quantum field theory prior to calculating the cross section of some process, one must
check whether this process is a crossing-channel of the other one that has been investigated
before. At a single glance on the diagrams displayed in Fig. 19.4 it becomes clear that
they differ from the diagrams of Fig. 19.1 describing the Compton effect by the following
rearrangements only:

k → k, k ′ → q, p → −p− , p′ → p+ , (19.121)

where q is a four-dimensional pseudophoton momentum. Then the amplitude of the e− e+ -

pair production in the nucleus field follows from that of the Compton effect under replace-
ment (19.121) and is defined by the expression:
s
m2
Mγ γ̃→e− e+ = −ie2 u(p− )eλµ (k)Oµν Aeν (q)v(p+ ), (19.122)
2E− E+ ω

where
fˆ1 + m ν fˆ2 + m µ
Oµν = γ µ 2 2
γ + γν 2 γ ,
f1 − m f 2 − m2
Ae (q) is a Fourier component of the nucleus field, and

f1µ = (p− − k)µ , f2µ = (−p+ + k)µ , q = k − p− − p+ .

The e− e+ -pair production process alongside the Compton effect belong to the processes to
be described by the same diagrams both in the QED and in the electroweak theory when
one is constrained by the second order of the perturbation theory.

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 489

We shall neglect the recoil momentum, that is, the nucleus will be considered to be
fixed (q0 = 0). Then, according to (18.192), its Coulomb field is covered by the Fourier
component:
Ze
Ae0 (q) = . (19.123)
(2π)3/2 q2
Let us find the cross section of the e+ e− -pair production for unpolarized particles, that
is, average over polarizations of the incoming photon and sum up over spin orientations
of the electron and positron to be emitted. As we know, this operation is reduced to the
following change in the amplitude square of the process (19.120):

1
|u(p− )eµ (k)Oµ0 Ae0 (q)v(p+ )|2 → (Ae (q))2 Sp F, (19.124)
8m2 0
where
F = Oµ0 (−p̂+ + m)Oµ0 (p̂− + m).
It is clear that under rearrangement of the photon and pseudophoton lines on the diagrams
of Fig. 19.4 these diagrams convert to each other. Pointed rearrangement is associated with
the replacement of the four-dimensional momenta of the particles:

p− ↔ −p+ , q ↔ −q, k ↔ −k. (19.125)

The diagrams symmetry existence must lead to splitting of Sp F into two addends to
transform to each other under replacement (19.124). It is easy to see that the required
splitting has the view:
Sp F = P1 + P2 , (19.126)
where
1
P1 = Sp[Oµ0 (−p̂+ + m)γ 0 (fˆ1 + m)γµ (p̂− + m)], (19.127)
m2 κ 1

1
P2 = Sp[Oµ0 (−p̂+ + m)γµ (fˆ2 + m)γ 0 (p̂− + m)], (19.128)
m2 κ 2
and we have introduced invariant quantities

m2 κi = fiµ
2
− m2 , i = 1, 2.

If one fulfils the replacement (19.124) into the second term of the expression (19.126) and
take into account that in this case

f1 ↔ f2 . κ1 ↔ κ 2 ,

then, under the spur sign, we shall obtain the matrices identical to those entering in the
expression for P1 but placed in inverse order. Thus, we work out P1 enough while the
relation:
P2 (k, p− , p+ ) = P1 (−k, −p+ , −p− ) (19.129)
allows us to find P2 . Using the formulae of Section 15.2, we obtain
  2 
1 1 q2 4E+ 4E+ E− κ1
P1 = 2 κ1 κ2 + 2 2 − (2 − κ1 ) + +
4 κ1 m m2 m2
 2   2   2  
1 q q2 2E+ q2 2E− q2
+ κ1 + κ2 + 2 + 2 − κ2 + 2 − κ1 + 2 −
κ 1 κ 2 m2 m m2 m m2 m

© 2011 by Taylor and Francis Group, LLC

 490 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

2E− E+
− (κ 1 + κ 2 − 4) . (19.130)
m2
By application of (19.129) we define P2 /4 and arrive at the following expression for Sp
F:   
2 2κ1 κ2 q2 q2 2 2 2
Sp F = 2 2 2 + κ 1 + κ 2 + − (E + E ) +
κ1 κ2 m2 m2 m2 − +

  
2 2 2q2 (κ1 E− − κ2 E+ )2
+(κ1 + κ1 ) κ1 κ2 + 2 − 8 . (19.131)
m m2
We next draw our attention to the formula (18.197) to deduce the differential cross section.
Allow for that unlike the Compton effect, in the case in question the electron momentum
p− and the positron one p+ are independent, that is,
 −1
∂(E− + E+ )
= 1.
∂E−
We also make use of the relations:

m2 κ1 = −2p− k = 2p− k − 2E− ω = 2ω(|p− | cos θ− − E− ),

m2 κ2 = −2p+ k = 2ω(|p+ | cos θ+ − E+ ),

q 2 = −m2 (κ1 + κ2 + 2) − 2(E− E+ + p− p+ ),
All this taken together allows one to get the differential cross section of the e+ e− -pair
production by the photon in the Coulomb nucleus field [24]:
(   2
4Z 2 α3 | p− || p+ | E+ p− E− p+
dσ = −4 k + +
(2π)2 ω 3 q4 m4 κ1 κ2
  2 )
2 p− p+ 2ω 2 2
+q k − + [k(p− + p+ )] dE+ dΩ− Ω+ , (19.132)
κ1 κ2 κ1 κ2
where dΩ− and dΩ+ are elements of the solid angles in which the momenta p− and p+ lay,
E− and E+ are the electron and positron energies for which the relation

E− + E+ = ω

takes place. The differential cross section (19.132) may be rewritten in the form to be more
convenient for investigation
(
Z 2 α3 p− p+ dE+ sin θ− sin θ+ dθ− dθ+ dϕ p2− sin2 θ− 2
dσ = (4E+ − q2 )+
(2π)ω 3 q4 (E− − p− cos θ− )2

p2+ sin2 θ+ 2 2 2 p2+ sin2 θ+ + p2− sin2 θ−

+ (4E − − q ) − 2ω +
(E+ − p+ cos θ+ )2 (E+ − p+ cos θ+ )(E− − p− cos θ− )

2p− p+ sin θ+ sin θ− cos ϕ
+ (4E− E+ + q2 − 2ω 2 ) , (19.133)
(E+ − p+ cos θ+ )(E− − p− cos θ− )
where θ± are angles between p± and k, ϕ is an angle between the planes (k, p− ) and
(k, p+ ), p± = |p± |.
As we can see, the angular distribution for the produced electron and positron has a
very complicated character. It contains two high peaks in directions of the speeds v± =

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 491

cos−1 θ± and vanishes at θ− = θ+ = 0. The expression for dσ is greatly simplified in the

ultrarelativistic region (E± ≫ m) only, when a leading role is played by small angles θ± . In
this case the electron and the positron are emitted in the forward direction predominantly,
that is, in the narrow cone with of the angle θ± ∼ m/ω along the photon motion direction.
Carrying out the replacements:
1 2
sin θ± → θ± , cos θ± → 1 − θ±
2
into (19.133), we deduce the angular distribution in the small angles region:

4Z 2 α3 E−2 2
E+ 2 2
dσ = [ω (u + v 2 )ξη − 2E− E+ (u2 ξ 2 + v 2 η 2 )+
πm ω q4
2 3

2 2
+2(E− + E+ )uvξη cos ϕ]θ− θ+ dE+ dθ− dθ+ dϕ, (19.134)
where
p − θ− p + θ+ 1 1
u= , v= , ξ= , η= .
m m 1 + u2 1 + v2
To integrate (19.133) over the angles is a rather intricate operation. For the calculation
details we refer the reader to Ref. [25] and give the final result only:
 
Z 2 α3 p− p+ 4 2E− E+ (p2− + p2+ ) 2 E+ E−
dσ = 3 2
− − 2 2 + m l− 3 + l+ 3 −
ω m 3 p+ p− p− p+
 
1 8E− E+ ω2 2 2
−l− l+ +L − + 3 3 (E− E+ + p2− p2+ − m2 E− E+ )−
p− p+ 3p− p+ p− p+
 
m2 ω E− E+ − p2+ E− E+ − p2−
− l+ + l − dE+ , (19.135)
2p− p+ p3+ p3−
where
E± + p± E− E+ + p− p+ + m2
l± = ln , L = ln .
E± − p± E− E+ − p− p+ + m2
Recall that all the formulae obtained above are based on the Born’s approximation
(Zα/v± ≪ 1). The symmetry of the cross sections (19.131)–(19.135) with respect to the
electron and the positron corresponds to just that very circumstance. This symmetry dis-
appears in more higher orders of the perturbation theory.
In the ultrarelativistic region, when all energies are significantly larger than m (ω, E± ≫
m), the expression (19.135) is appreciably simplified:
  
4Z 2 α3 2 2 2 2E− E+ 1
dσ = 3 2 E− + E+ + E− E+ ln − dE+ . (19.136)
ω m 3 ωm 2

Integration of this cross section over E+ from m to ω yields the total cross section of the
e− e+ -pair production by the photon:
 
Z 2 α3 28 2ω 218
σ= ln − , ω ≫ m. (19.137)
m2 9 m 27

We call attention to the fact that for ω → ∞ the total cross section, though slowly varying
by the logarithmic law, is still tending to infinity. Such a behavior is a symptom that a
theory suffers from a serious disease: invasion of divergences. Sometimes this disease may
be cured by redefinition of the finite number of the theoretical parameters (renormalizable

© 2011 by Taylor and Francis Group, LLC

 492 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

theories), while in other cases this procedure is not a success (nonrenormalizable theories).
Fortunately, the QED falls into the category of the renormalizable theories.
Close to the reaction threshold, the electron and positron velocities may be too low so
that the relation (19.118) becomes invalid, making the Born’s approximation inapplicable.
In this case we have to apply an exact solution of the problem, that is, operate with the
wavefunctions of the electron and positron in the Coulomb field of the nucleus. It is obvious
that in this case there is no symmetry with respect to the electron (attracted to the nucleus)
and to the p positron (repelled by the nucleus).
If v± ≈ (ω − 2m)/ω remaining more less than 1 is of this sort that v± ≫ Zα, then
the Born approximation is applicable at the reaction threshold too. In the case of non-
relativistic energies of the pair we have

ω ≈ 2m ≫ p± , q ≈ ω.

Then one may set everywhere in (19.133)

E± = m, E± − p± cos θ± = m, ω = 2m,

what leads this relation to the form:

Z 2 α3 p− p+ 2
dσ = (p+ sin2 θ+ + p2− sin2 θ− )dΩ− dΩ+ dE+ . (19.138)
64π 2 m7
To integrate the cross section obtained over the angles results in

Z 2 α3
dσ = p− p+ (p2− + p2+ )dE+ =
6m7
2Z 2 α3 p
= 5
(ω − 2m) (E+ − m)(E− − m)dE+ . (19.139)
3m
From (19.139) one may derive the total cross section of the e− e+ -pair production in the
nonrelativistic region
Z ω−2m  3
Z 2 α3 π ω − 2m
σ= dσ = . (19.140)
0 12m2 m
If the relative velocity of the electron and the positron v0 is small, then we must allow
for their Coulomb interaction. This interaction is essential when v0 has the same order as
the velocity of a particle in the positronium

v0 ≤ α.

The cross section of the e− e+ -pair production having been calculated, we assumed that
the nucleus field is pure Coulomb. It is true when distances from the nucleus are less
than the radius of the atom K-shell. At great distances the nucleus field is partially or
completely screened by the atom electrons field. As a result, the cross section of the e− e+ -
pair production is decreased. Let us evaluate that region of distances from the nucleus that
is the most significant for the process under investigation.
From the view of the integral defining the Fourier component for the Coulomb field of the
nucleus it follows that the main contribution in the integral is coming from the distances of
the order of
1
R0 ∼ . (19.141)
|q|
Really, the great values of r are unessential because of the fast oscillating function exp(iqr),
whereas the small ones—due to smallness of corresponding integration volume 4πr2 dr.

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 493

These circumstances allow one to consider that R0 defines the most significant values of the
impact parameter in the order of magnitude. Obviously, the screening effect will not play
any role when the maximum value of R0 , which is associated with the minimum value of
q = |q|, is markedly less than the effective atom size Ra (R0max ≪ Ra ). On the contrary, at
fulfillment of inequality R0max ≫ Ra the complete screening of the Coulomb nucleus field
must take place.
Find qmin in the nonrelativistic and ultrarelativistic regions. In the former case ω ≈ 2m
and we get
qmin ≈ 2m.
In the ultrarelativistic region (ω, E± ≫ m) the result is
m2 ω
q q
qmin = ω − E− 2 − m2 − 2 − m2 ≈
E+ .
2E− E+
So, the maximum value of R0 is given by the formulae:
 1 p
 m, ω ≈ 2m, (ω − 2m)/ω ≪ 1,
R0max ∼ (19.142)
 2E− E+
m2 ω , ω, E± ≫ m.
According to the Thomas-Fermi atom model the effective atom radius is determined by
the expression:
1
Ra = . (19.143)
mαZ 1/3
From comparison of (19.142) with (19.143) one becomes clear that the condition

R0max > Ra
may not be realized in the nonrelativistic region but it may take place in the ultrarelativistic
region. Therefore, the inclusion of the screening effect is always needed in the high-energy
region. It is obvious that the above-derived formulae for the e− e+ -pair production cross
section without inclusion of the screening effect in the ultrarelativistic case are valid only
under realization of the condition:
2E− E+ 137
≪ 1/3 . (19.144)
mω Z
To take into consideration the nucleus field screening effect by external electrons one
should find the Fourier component of summarized potential produced both by the Coulomb
nucleus field and by the electrons field. In the stationary case we have
(e)
∆A0 (r) = −ρ(r), (19.145)
where ρ(r) is a charge density,

ρ(r) = Zeδ(r) − eN (r)

and N (r) is an electron density in an atom. By the standard way we get

(e) e[Z − F (q)]

A0 (q) = , (19.146)
(2π)3/2 q2
where F (q) is an atom formfactor,
Z
1
F (q) = N (r)e−iqr d3 x.
(2π)3/2

© 2011 by Taylor and Francis Group, LLC

 494 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

So, to allow for the screening effect one should make the following replacement in the
e− e+ -pair production reaction amplitude:
Z → Z − F (q) = Z[1 − f (qZ −1/3 )]. (19.147)
In the Thomas-Fermi model f (x) is an universal function that is identical for all atoms and
whose values could be defined by application of numerical integration.
Introducing the atom formfactor one could report the e− e+ -pair production cross section
in the ultrarelativistic region as follows [24]:
    
Z 2 α3 2 2

4 2 4
dσ = 3 2 E− + E+ Φ1 (ζ) − ln Z + E− E+ Φ2 (ζ) − ln Z dE+ . (19.148)
ω m 3 3 3
where
Ra mω
ζ = 200α = 100 ,
R0max E− E+ Z 1/3
Φ1 (ζ) and Φ2 (ζ) are functions of ζ shown in Fig. 19.5. To be sure that the quantity ζ

FIGURE 19.5
The graphs of the functions Φ1 (ζ) and Φ2 (ζ).

determines the screening effect. When ζ is decreasing (increasing) the screening effect is
growing (lowering). When ζ = 0 one may speak of the complete screening that obviously
takes place under the fulfillment of inequality:
2E− E+
≫ 137Z −1/3.
mω
In this case
2
Φ1 (0) = 4 ln 183, Φ2 (0) = 4 ln 183 −
3
and the energies distribution take the form:
  
4Z 2 α3 2 2 2 −1/3 E− E+
dσ = 3 2 E− + E+ + E− E+ ln(183Z )− dE+ . (19.149)
ω m 3 9

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 495

Integrating (19.149) over E+ , we deduce the e− e+ -pair production cross section in the
case of the complete screening
 
Z 2 α3 28 −1/3 2
σ= ln(183Z )− . (19.150)
m2 9 27
In Fig. 19.6 we represent the plot of σ versus the photon energy for lead and aluminium
with and without inclusion of the screening effect. Polarizations effects for the e− e+ -pair

FIGURE 19.6
The total cross section of the process γγ̃ → e− e+ versus the photon energy.

production process could be investigated by the same method to have been used in Section
19.4 under discussing the Compton effect. Here we shall not touch this problem and refer
an interested reader to the original literature [26].

19.6 Resume
When γ-quanta pass through the matter (target), some quanta leave the flux due to their
interaction with electrons and nuclei. The beam intensity J is decreased with the growing
thickness of a target x as:
J = J0 e−µx , (19.151)
where J0 is a flux intensity of photons falling down on a target, µ is an absorption coefficient.
In the formation of µ three processes have a decisive role: (i) photoeffect at the electron
shell of an atom; (ii) Compton scattering of photons by quasi-free electrons; and (iii) the
e − e+-pair production in the electrostatic field of the atom nucleus. When N is the number
of atoms within 1 cm3 of a target, σi is the cross section of the above processes per atom
of the target, then we have:
X3
µ=N σi . (19.152)
i=1

© 2011 by Taylor and Francis Group, LLC

 496 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the case of the photoeffect its cross section is decreased with the growing photon energy:
close to the red threshold it is falling as Z 5 ω −7/2 and in the ultrarelativistic region—as
Z 5 ω −1 . In the process the principal contribution of ∼ 80% into the total photoeffect cross
section is made by the K-shell, whereas the filled L-shell contributes ∼ 16%, the contribution
of the M -shell being ∼ 4% only.
As distinct from the photoeffect, in the case of the Compton effect a photon is not
absorbed, an incident photon beam with the energy ω is transformed to the scattered beam
of photons with the energy ω ′ that is dependent on the scattering angle θ. The energy of a
scattered photon is varying within the interval from ω for θ = 0 to ω[1 + 2ω/m]−1 for θ = π.
When the energy ω is considerably higher than the binding energy of the K-electron, the
total cross section of the Compton effect for an atom may be considered proportional to
the electrons number, that is, in the case of neutral atoms it is proportional to the nucleus
charge Z. With growing energy of the incoming photon the Compton effect cross section
falls down as ∼ Zω −1 ln(ω/m).
The e− e+ -pair production leads to the absorption of a photon, its energy is distributed
between the electron and positron, some part being donated to the nucleus. With increasing
photon energy the e− e+ -pair production cross section is growing as ∼ Z 2 ln(2ω/m).
A relative role of the three principal photon absorption processes in the formation of the
coefficient µ is dependent on Z and on the photon energy ω. In Fig. 19.7 we display the
plot of the function µ(ω) for different substances with allowance made for the screening.
The behavior of the components caused by the photoeffect, the Compton effect, and the

FIGURE 19.7
The plot of the function µ(ω).

e− e+ -pair production effect is also given for lead.

© 2011 by Taylor and Francis Group, LLC

 Transmission of γ-radiation through matter 497

Since the photoeffect depends on Z 5 it is most important for heavy elements; that is why
Pb is used for protection from X-rays and γ-radiations. The Compton effect is the most
sizable for light elements. As is demonstrated by the relationship between the absorption
coefficient µ, photon energy and matter charge Z, in a small-energy region ω < ω1 the
photoeffect is the basic mechanism for the interaction between a photon and matter, in
the intermediate region ω1 < ω < ω2 the Compton effect is more important, and in the
high energy region ω > ω2 the key role is played by the the e− e+ -pair production. The
boundary values of the energy separating the preferential domains of these effects for various
substances are different. In the case of lead they are

ω1 = 0.5 MeV, ω2 = 5 MeV.

Aside from these main processes, there exist several mechanisms associated with the
photon departure from the flux. The first is known as the Delbrück scattering [27], coherent
photons scattering by a Coulomb nucleus field, when the e− e+ -pair is formed due to an
incoming photon and then annihilated to emit a photon with the initial energy

γ + γ̃ → γ + γ̃. (19.153)

The lowest order of the perturbation theory in which the Delbrück scattering process pro-
ceeds is the fourth one. The corresponding Feynman diagram is shown in Fig. 19.8. The

FIGURE 19.8
The Delbrück scattering process.

Delbrück scattering may be regarded as a special case of light-by-light scattering

γ + γ → γ + γ. (19.154)

However, the cross section of the process (19.153) in Z 4 α2 is greater than that of the process
(19.154), facilitating its experimental observation.
Besides, contributions to γ-radiation absorption are given by: (i) Thomson elastic scat-
tering by the structureless nucleus, (ii) Compton scattering by the nucleus nucleons, and
(iii) absorption in the nuclear reactions of the type (γ, n),∗ (γ, p), and (γ, α). The latter

∗ The first symbol in parentheses denotes an incoming particle while the second one does a new outgoing
particle

© 2011 by Taylor and Francis Group, LLC

 498 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

are important in the region of the so-called giant dipole resonance that is associated with
the photon energies of the order of 10–20 MeV. In case of photons with the energies within
this region the photonuclear processes may result in 5–10%-contributions into µ.

© 2011 by Taylor and Francis Group, LLC

 20
Scattering of electrons and positrons

Then this ebony bird beguiling my sad fancy into

smiling,
By the grave and stern decorum of the countenance
it wore,
“Though thy crest be shorn and shaven, thou,” I said,
“art sure no craven,
Ghastly grim and ancient Raven wandering from
the Nightly shore—
Tell me what thy lordly name is on the Night’s
Plutonian shore!”
Quoth the Raven “Nevermore.”
Edgar Allan Poe, “The Raven”

20.1 Bremsstrahlung
Most important for retardation of the high-energy charged particles is a bremsstrahlung—
the process accompanying every change in the state of a charged particle. According to
classical electrodynamics, describing the principal features of the bremsstrahlung with a
good approximation, its intensity is proportional to the squared acceleration and inversely
proportional to the particle mass. As a result, within the same field the bremsstrahlung
of an electron is greater than that of a proton by a factor of several millions. Therefore,
the bremsstrahlung caused by scattering of electrons from the Coulomb field of the atomic
nuclei and electrons is most often observed and used in practice. For example, this process
is used to generate high-energy photons due to the transmission of electrons through the
high-density targets.
Let us calculate the cross section of the electron bremsstrahlung by the Coulomb nucleus
field. In so doing we shall assume that the field may be allowed for by the perturbation
theory methods, that is, the applicability criterion of the Born approximation is fulfilled
Zα
≪ 1,
v
where v is an electron velocity. The diagrams corresponding to the process:

e− + γ̃ → e− + γ (20.1)

are pictured in Fig. 20.1. It is clear that the process (20.1) is another cross channel of
the e− e+ -pair production process in the Coulomb nucleus field. Therefore, the amplitude
Me− γ̃→e− γ follows from that of the e− e+ -pair production by means of the replacement:

p+ → −p1 , p− → p2 , k → −k. (20.2)

499
© 2011 by Taylor and Francis Group, LLC
 500 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

q p2

q p2

p2 + k
p1 − k

p1 k

p1 k

FIGURE 20.1
The Feynman diagrams for the electron bremsstrahlung by the Coulomb nucleus field.

So, we may not work out the cross section anew, and use the results that we have already
obtained for the e− e+ -pair production cross section. Carrying out the replacement (20.2)
into (19.133) and taking into consideration changing both of the incoming particle flux
value and of the final states number, we get the following expression for the differential
cross section of the bremsstrahlung in the case of unpolarized particles [24]:

Z 2 α3 p2 p22 sin2 θ2
dσ = (4E12 − q 2 )+
(2π)2 p1 ωq 4 (E2 − p2 cos θ2 )2

p21 sin2 θ1 2 2 2 p21 sin2 θ1 + p22 sin2 θ2

+ (4E2 − q ) + 2ω −
(E1 − p1 cos θ1 )2 (E1 − p1 cos θ1 )(E2 − p2 cos θ2 )

2p1 p2 sin θ1 sin θ2 cos ϕ
− (4E1 E2 − q 2 + 2ω 2 ) dωdΩdΩ2 , (20.3)
(E1 − p1 cos θ1 )(E2 − p2 cos θ2 )
where dΩ and dΩ2 are elements of the solid angles in which the vectors k and p2 lay, ϕ is
an angle between the planes (k, p1 ) and (k, p2 ), p1,2 = |p1,2 | and

q = |q| = |p2 + k − p1 |.

Mark some typical features of the angle distribution (20.3).

(i) By virtue of the fact that the expression (20.3) is proportional to ω −1 , dσ turns into
infinity under ω → 0 (infrared catastrophe). This effect is caused by the zero photon mass.
We shall discuss in detail the infrared catastrophe in the next Section.
(ii) In the angle distribution two peaks in the velocity directions of the initial and final
electrons, which are called forth by the factors (1 − v1 cos θ1 )−1 and (1 − v2 cos θ2 )−1 , are
present. In the ultrarelativistic case (v1 ∼ 1 and v2 ∼ 1) the angle distribution has a sharp
maximum near the momentum direction of the incoming electron. In this case the radiation
is basically centred into a narrow cone around this direction with the angle of the order of
θ1 ∼ m/E1 (the momentum of the scattered electron lays in the same cone).

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 501

To find the spectral distribution of the bremsstrahlung we carry out the replacement
(20.2) in the expression (19.135). The result is
 
Z 2 α3 p2 4 2E1 E2 (p21 + p22 ) 2 E2 E1
dσω = − + m l1 3 + l2 2 −
p1 ωm2 3 p21 p22 p1 p3
 
1 8E1 E2 ω2
−l1 l2 +L + 3 3 (E12 E22 + p21 p22 + m2 E1 E2 )+
p1 p2 3p1 p2 p1 p2
2
 
m ω E1 E2 + p21 E1 E2 + p22
+ l1 − l 2 dω, (20.4)
2p1 p2 p31 p32
where
E1,2 + p1,2 E1,2 + p1,2
l1,2 = ln = 2 ln ,
E1,2 − p1,2 m
E1 E2 + p1 p2 − m2 E1 E2 + p1 p2 − m2
L = ln 2
= 2 ln .
E1 E2 − p1 p2 − m mω
The allowed values of the photon energy in the formulae obtained are limited by the con-
dition imposed on the final electron velocity (Zα/v2 ≪ 1), namely, the electron does not
have to lose nearly all its energy.
In the nonrelativistic case (p1 ≪ m) the spectral distribution is greatly simplified:
√ √
16Z 2α3 dω p1 + p2 8Z 2 α3 dω ( T1 + T1 − ω)2
dσω = ln = ln , (20.5)
3p21 ω p1 − p2 3mT1 ω ω

where T1 = E1 − m = p21 /(2m) is a kinetic energy of the initial electron. Thus, in the
nonrelativistic region the radiation cross section approximately varies in inverse proportion
to the photon energy and vanishes under the greatest possible photon energy ω = T1 .
In the ultrarelativistic region (E1,2 ≫ m) dσω takes the form:
  
4Z 2 α3 E2 dω E1 E2 2 2E1 E2 1
dσω = + − ln − =
ωE1 m2 E2 E1 3 m 2
"  2  #  
4Z 2 α3 dω ω 2 ω 2E1 (E1 − ω) 1
= 1 + 1 − − 1 − ln − . (20.6)
ωm2 E1 3 E1 mω 2
From the expression (20.6) it is evident that the radiation probability by the electron of
a particular part of its energy, that is, the radiation probability at the given ratio ω/m,
approximately grows as ln(E1 /m). As far as the radiation intensity J = ωdω is concerned,
it is logarithmically diverging both in the relativistic and nonrelativistic regions. However,
both anomalies for the process in question are removed by means of the inclusion of the
nucleus field screening by the atom electrons. To take into account the screening may be
carried out in exactly in the same manner as for the case of the e− e+ -pair production in the
nucleus field. In the ultrarelativistic region we derive the formula for the bremsstrahlung
cross section [24]:
    
Z 2 α3 2 2

4 2 4
dσω = E 1 + E 2 Φ 1 (ζ) − ln Z − E1 E2 Φ 2 (ζ) − ln Z dω. (20.7)
ωE12 m2 3 3 3

In the case of the complete screening (ζ = 0), that is, when

2E1 E2
≫ 137Z −1/3, E1 ≫ 137mZ −1/3,
mω

© 2011 by Taylor and Francis Group, LLC

 502 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the cross section (20.7) acquires the form:

    
4Z 2 α3 2 2 2 −1/3 E1 E2 ω
dσω = E1 + E2 − E1 E2 ln(183Z )+ d . (20.8)
ωE12 m2 3 9 E1 − m

In these conditions at the given ratio ω/E1 the spectral distribution of the radiation does
not depend on the initial electron energy E1 . Under great values of ζ (small screening) this
circumstance breaks down.
Having found the bremsstrahlung cross section, one could determine the average energy
loss of the electron when it moves through matter. This quantity attributed to the unit of
the electron path equals
  Z E1 −m  
dE1 dσω
− =N ω dω = N E1 Φr , (20.9)
dx r 0 dω

where N is the number of atoms per the unit of the matter volume, and Φr is the cross
section of the energy loss on radiation:
Z E1 −m  
1 dσω
Φr = ω dω. (20.10)
E1 0 dω

Inserting the formula (20.4) into (20.10) and fulfilling the integration, we get the expres-
sion for Φr :
"  2
12E12 + 4m2 E1 + p1 m2 (8E1 + 6p1 ) E1 + p1
Φr = Φ ln − ln −
3E1 p1 m 3E1 p21 m
 
4 2m2 2p1 (E1 + p1 )
− + F , (20.11)
3 E1 p1 m2
where Φ = Z 2 α3 /m2 , and the function F (x) has been defined as:
Z x
ln(1 + y)
F (x) = dy.
0 y
In order to find Φr in the nonrelativistic and relativistic regions we use the following
expansions of the function F (x):
 
π2 1 2 1
F (x) = + ln x − F , for any x, (20.12)
6 2 x

x2 x3 x4
F (x) = x − + − + ... x ≪ 1. (20.13)
4 9 16
Allowing for (20.12) and (20.13), we arrive at:
16
Φr = Φ, E1 ≪ m, (20.14)
3
 
2E1 1
Φr = 4Φ ln − , E1 ≫ m. (20.15)
m 3
From the derived formulae it follows that in the nonrelativistic region the ratio of the
emitted energy ω to the energy of the initial electron E1 does not depend on E1 , while in
the ultrarelativistic case Φr is logarithmically increased with E1 . As one would expect, this

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 503

increase is removed by the inclusion of the screening. At the complete screening, in place
of (20.15) we have the constant cross section:
 
−1/3 2
Φr = Φ 4 ln(183Z )+ , E1 ≫ 137mZ −1/3.
9
The energy-dependence of the cross section Φr with the screening inclusion in the logarith-
mic scale is represented in Fig. 20.2, where, for the comparison, the straight line associated
with the cross section to be determined by Eq. (20.15) is also displayed. Under motion of

FIGURE 20.2
Φr E1 − m
as a function of .
Φ m

the electrons in matter one should also bear in mind that the radiation occurs not only on
the nucleus but on the electrons as well. In the ultrarelativistic case the bremsstrahlung
cross section of the electron from the electron is unlike that of the electron from the nucleus
by the absence of the factor Z 2 only [28]. Therefore, the presence of Z atom electrons may
be approximately taken into consideration by the replacement:
Z 2 → Z(Z + 1).
There exist one more source of energy loss by the electron when it moves in matter,
namely, inelastic collisions with atoms. In this case, the average lost energy along the
length unit is given by the expression [29]:
 
dE1 2πα2 N Z E13
− = ln , (20.16)
dx i m 2mI 2
where I is an average ionization potential (I ≈ 13.5Z eV). Thus, the ratio of the loss due
to radiation to that due to collisions is equal to
   −1
dE1 dE1 1 E1 Z
≈ , (20.17)
dx r dx i 1600 m

© 2011 by Taylor and Francis Group, LLC

 504 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where, for the sake of simplicity, we have assumed that the logarithmic term in the formulae
(20.15) and (20.16) has approximately an equal value. It is evident that the losses due to
radiation and those due to collisions become equal when the relation:

(E1 )0 ≈ 1600mZ −1

takes place. When E1 > (E1 )0 the losses due to radiation prevail over the ones due to
collisions ((E1 )0 equals 10 MeV for lead, 55 MeV—for copper, 200 MeV—for air). A reader
could find the detailed discussion of the bremsstrahlung polarization in Refs. [30,31].

20.2 Radiation of long-wave photons. Infrared catastrophe

In the previous Section we have faced the problem of turning into an infinity of the proba-
bility of emitting the photons with ω → 0, the infrared catastrophe (IC). The IC proves to
be caused by the fact that the ordinary perturbation theory, based on expanding the scat-
tering matrix into a series in powers of e, is not applied to the processes in which the long
wavelength photons (soft photons) participate. To make this circumstance obvious we shall
investigate the inclusion of emitting the soft photons within the diagrammatic technique
[32].
Let us consider an arbitrary scattering process of a charged particle that may be accom-
panied by the emission of any number of the ordinary photons (in the text referred to as
the hard photons). Along with this process, we consider a process that is distinguished only
by the emission of an additional soft photon. The cross sections of both processes dσab and
dσabγ have a relatively simple relationship, the determination of which presents a problem
at hand.
The diagrams for the process with an additional photon follow directly from those for the
basic process by the “joining” of an external photon line to some electron line (external or
internal), that is, by the substitution:

is replaced by
p p p−k

FIGURE 20.3
The inclusion of a soft photon in an electron line.

Obviously, the main role will be played by the diagrams, where such a joining is realized at
the external electron line. In fact, if p—a momentum of the external line (p2 = m2 ), then
at small k we also have (p − k)2 ≈ m2 , that is, a contribution of the electron propagator
S(p − k) introduced in the reaction amplitude is considerable due to the position close to
its pole. This makes it possible to neglect variations in the momenta of the lines associated
with the emission of a soft photon for all the remaining diagram parts.

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 505

Joining the photon line to the initial electron line p induces the following replacement in
the reaction amplitude:

p̂ − k̂ + m λ∗
u(p) → eS(p − k)êλ∗ (k)u(p) = e ê (k)u(p) ≈
(p − k)2 − m2

p̂ + m λ∗ (peλ∗ (k))
≈ −e ê (k)u(p) = −e u(p), (20.18)
2(pk) (pk)
where the Dirac equation has been allowed for at the last stage of the transformations.
The similar way joining the soft photon to the final electron line with p′ is reduced to the
replacement:
(p′ eλ∗ (k))
u′ (p′ ) → eu′ (p′ ) . (20.19)
(p′ k)
For the sake of definiteness, we assume that the cross section dσab describes the electron
scattering by the fixed nucleus with emitting arbitrary numbers of hard photons. The
corresponding amplitude is as follows:

Mab = u′ (p′ )Mu(p).

In the expression for Mab we first fulfill the replacement (20.18) and then the replace-
ment (20.19). Next, summing the results obtained, we arrive at the amplitude of the
bremsstrahlung for the same hard photons plus one more soft photon:
 ′ λ∗ 
(p e (k)) (peλ∗ (k))
Mabγ = eMab − . (20.20)
(p′ k) (pk)

We draw attention to the gauge invariance of the derived expression, namely, the reaction
amplitude is not modified under changing of the four-dimensional vector of the photon
polarization:
eλµ → eλµ + const · kµ .
To find the connection between the cross sections it is convenient to come back to the
three-dimensional designation. Assuming eλ (k) = (0, eλ (k)), we get

 (p e (k)) (peλ∗ (k)) 2 d3 k

 ′ λ∗ 
2
dσabγ = e dσab 
 −  =
(p′ k) (pk)  2(2π)3 ω
 2
2
 p′ eλ∗ (k) peλ∗ (k)  d3 k
= e dσab  ′
 − , (20.21)
E ω(1 − v′ n) Eω(1 − vn)  2(2π)3 ω
where v and v′ are velocities of the initial and final electrons, and n = k/ω. Summing over
the polarizations of the soft photon, we finally obtain
 2
[v′ n] [vn] dωdΩ
dσabγ = α − dσab . (20.22)
1 − v′ n 1 − vn 4π 2 ω

From (20.22) follows that with a precision of ω −1 the expression standing in front of dσab
coincides with the radiation intensity dI calculated in the classical radiation theory, that is,

dI
dσabγ = dσab (20.23)
ω

© 2011 by Taylor and Francis Group, LLC

 506 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

takes place. However, in order to consider the electron motion to be given and use the
classical electrodynamics the following conditions:
ω ω
≪ 1, ≪ 1, (20.24)
E |∆p|
where ∆p is a momentum transferred to the nucleus, must be fulfilled. Besides, the photon
wave length λ must be significantly bigger than r0 . All this means that the obtained
above formulae taking into account emitting a soft photon are valid under fulfillment of the
conditions (20.24) only.
In a real experiment the final particles energy is known with some spreading ∆E inside
that one cannot separate processes proceeding with and without emitting soft photons.
(∆E)
Then the quantity of interest for us, dσabγ , will be given by the expression:
Z ∆E
(∆E) dI ∆E
dσabγ = dσab ∼ α ln dσab . (20.25)
ωs ω ωs
Since the values ωs had no limitations from below, then at enough small ωs the parameter
α ln(∆E/ωs ) may become greater than 1 to lead to dσabγ  dσab . But it means the inap-
plicability of the perturbation theory as, according to its ideology, the quantity dσabγ must
be of a more higher smallness order than dσab . To put it differently, at the investigation of
the soft photon radiation processes the perturbation theory expansion parameter should be
considered by the product α ln(∆E/ωs ) rather than α.
The formula (20.20) could be generalized without trouble in the event of simultaneously
emitting several soft photons. For every one of the photon the factor having the same
form as that standing at Mab in (20.20) is added to the amplitude Mabγ . Let us convince
ourselves by the example of emitting two soft photons that this is the case. The lines of
both soft photons must be joined at external electron lines in two different sequences. Thus,
for example, the diagram with a initial external line p is changed by two ones pictured in
Fig. 20.4. Then the additional factors:

k1 k2 k2 k1

or
p p − k1 p − k1 − k2 p p − k2 p − k2 − k1

FIGURE 20.4
The diagrams with two soft photons.

(peλ1 ∗ (k1 ))(peλ2 ∗ (k2 ) (peλ1 ∗ (k1 ))(peλ2 ∗ (k2 )

and
2[(pk1 ) + (pk2 )]2(pk1 ) 2[(pk1 ) + (pk2 )]2(pk2 )
appear in the amplitudes corresponding to the first and second diagrams, respectively.
Further we should join first the lines k1 and k2 at a final electron line p′ , subsequently
k1 —at a final electron line while k2 —at a initial electron line, and lastly k2 —at a final
electron line while k1 —at a initial electron line. To sum the results obtained really leads to
the formula of the kind (20.20):
 ′ λ1 ∗  
2 (p e (k1 )) (peλ1 ∗ (k1 )) (p′ eλ2 ∗ (k2 )) (peλ2 ∗ (k2 ))
Mabγ = e Mab − − . (20.26)
(p′ k1 ) (pk1 ) (p′ k2 ) (pk2 )

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 507

The cross section of a process is factorized in a similar way. Thus, soft photons are
emitted independently. The cross section of a process with emitting n soft photons could
be represented in the form:

dσabγ = dσab dw1 dw2 ...dwn , (20.27)

where dw1 , dw2 ... are probabilities of the individual emitting of photons with the four-
dimensional momentum k1 , k2 ... Under integration of this formula over the finite interval
of frequencies and directions one should introduce the factor 1/n!, which takes into account
the identity of these particles.
The formula (20.27) may be rewritten in a somewhat different form too. Since emitting
the soft photon does not influence an electron motion (this is ensured by the fulfillment
of the conditions (20.24) ), then the processes of subsequent emitting the soft photons are
statistically independent. Hence, it follows that the probability of emitting such photons is
defined by the Poisson formula:
e−n
w(n) = (n)n , (20.28)
n!
where n is an average number of the photons emitted:
Z ∆E
n= ω −1 dI. (20.29)
ωs

Now the cross section of the process with emitting the n photons is represented in the view:

dσabγ = dσab w(n). (20.30)

If ωs = 0, then according to (20.25) and (20.29) n is equal to infinity, and, consequently,

w(n) vanishes. It means that the probability of electron scattering with emitting the finite
number of the soft photons equals zero. In the same time, because
∞
X
w(n) = 1,
n=0

then ∞
X
dσabγ = dσab
n=0

and dσab comprises the cross section of scattering, which is accompanied by emitting any
number of the soft photons. This cross section may be defined by application of the pertur-
bation theory in the case of a sufficiently weak external field. On the other hand, the cross
section, accompanied by emitting one soft photon or finite number of the soft photons, is
incorrectly given by the perturbation theory. In accordance with the perturbation theory
this cross section is equal to infinity whereas it vanishes in a real situation.
As regards the cross sections for the emission of hard photons, which is naturally calcu-
lated by the perturbation theory, then the emission cross section, for example, of a single
hard photon should be interpreted as a cross section describing the emission both of the
hard photon and of an arbitrary number of soft photons.
Because the perturbation theory leads to incorrect values of the cross sections for the
interactions between electrons, protons and soft photons, we should pick out the dangerous
border ωmin , below which the perturbation theory methods are inapplicable, assuming that
for a real photon the following is true ω > ωmin . However, in practice it is more convenient
to use the equivalent condition: setting no limit to the photon energy, we assume that a

© 2011 by Taylor and Francis Group, LLC

 508 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

photon has some small mass λ 6= 0. A change from the noninvariant quantity ωmin to the
relativistically invariant quantity (mass) is called for the maintenance of the relativistic
invariance of the theory. Note that the introduction of the photon mass is nothing else
but a formal approach to consideration of the infrared divergences. Of course, the photon
mass must not enter into any formulae for the observables. To ensure the fulfillment of this
circumstance we would simply go to the limit λ → 0 after all the intermediate calculations.
But we need not do this. As has been demonstrated previously, at sufficiently small ∆E an
experimenter can hardly distinguish between the reaction associated with the emission of
a soft photon and the corresponding elastic scattering reaction. Because of this, the cross
section dσabγ should be considered together with the cross section dσab calculated in the
same order of the perturbation theory. In the next section, we shall find the cross section of
(2)
the pure Coulomb scattering in the second order of the perturbation theory dσab and show
that it also contains the infrared divergence. In so doing the infrared divergences entering
into the observed quantity:
(2) (2)
dσobs = dσab + dσabγ
are canceled, that is, the infrared cut-off parameter disappears. So, the fictitious photon
mass λ is required only with the aim to make the intermediate calculations meaningful.
Note that, since the above procedure takes into account the experimental resolution for the
energy ∆E, the quantity ∆E is involved in the final result. This, in its turn, reveals the fact
that for the low-frequency region of particular importance is the total energy carried over
by the emitted soft photons, as related to ∆E, rather than the number of these photons.
Our next task is to establish the linkage between ωmin and λ. To solve it we consider
the soft photon bremsstrahlung process by the electron in the Coulomb field of the nucleus
[32]. Using the replacements:

−2(pk) → −2(pk) + λ2 , 2(p′ k) → 2(p′ k) + λ2 ,

we introduce the photon mass into Eqs. (20.18), (20.19). Choose the soft photon polariza-
tion vector eνµ (k) in the form δµν . Then the bremsstrahlung amplitude is determined by the
expression:
r  
2 m2 ′ e 2pµ 2p′µ
Me− γ̃→e− γ = −ie [u(p )Â (q)u(p)] + . (20.31)
2EE ′ ω λ2 − 2(pk) λ2 + 2(p′ k)

Since E, E ′ ≫ λ, then
p p
λ2 − 2(pk) = λ2 − 2E k2 + λ2 + 2pk ≈ −2E k2 + λ2 + 2pk = −2(pk),

λ2 + 2(p′ k) ≈ 2(p′ k).

Further, we assume that the photon energy does not exceed the value ∆E that defines
the experimental energy resolution and satisfies the demand:

λ ≪ ∆E ≪ E.

For the sake of simplicity, we shall not be interested in the polarizations of the initial and
final particles. Then, with allowance made for Eqs. (20.21) and (4.1.197), we arrive at the
following expression for the differential cross section:

(∆E) B
dσe− γ̃→e− γ = e2 dσ0 , (20.32)
2(2π)3

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 509

where
  4 Z  2
Z 2 α2 2 2 θ
X p′µ pµ d3 k
dσ0 = 2 2 4 1 − v sin dΩ, B= ′
− ,
4p v sin (θ/2) 2 µ=1 k0 ≤∆E (p k) (pk) ω

dΩ is a solid angle element in which the scattered electron momentum lays.

To estimate the connection between ωmin and λ, we should calculate the quantity B
entering into (20.32) by two expedients: (i) λ 6= 0 while k0 is changed from k0 = λ to
k0 = ∆E; and (ii) λ = 0 while k0 is changed from k0 = ωmin to k0 = ∆E. To compare
the obtained expressions that we denote as Bλ and Bωmin will lead us to the sought-for
relation. The calculation of Bλ results in
Z |k|=∆E 2 
k d|k|dΩγ m2 m2
Bλ = − √ √ + √ +
|k|=0 k2 + λ2 (E ′ k2 + λ2 − p′ k)2 (E k2 + λ2 − pk)2

2(pp′ )
+ √ √ =
(E k2 + λ2 − pk)(E ′ k2 + λ2 − p′ k)
Z |k|=∆E 2 
k d|k|dΩγ m2 m2
=− √ √ + √ +
|k|=0 k2 + λ2 (E ′ k2 + λ2 − p′ k)2 (E k2 + λ2 − pk)2
Z 1 
dz
+(pp′ ) √ , (20.33)
2 2 2
−1 (Ez k + λ − pz k)

where dΩγ is a solid angle element in which the photon momentum lays,

1 1
pz = [(1 + z)p + (1 + z)p′ ], Ez = [(1 + z)E + (1 − z)E ′ ],
2 2
and we have used the identity:
Z 1
1 1 dz
√ √ = √ .
(E k2 + λ2 − pk)(E ′ k2 + λ2 − p′ k) 2 −1 (Ez k2 + λ2 − pz k)2

The integrals over dΩγ could easily be worked out in the spherical coordinates whose polar
axis is directed first along the vector p, then along p′ , and lastly along pz . Thus, for
example,
Z θγ =π
dΩγ 2π 

√ = √  =
2 2
(Ez k + λ − pz k)2 2 2
| pz || k | (Ez k + λ − | pz || k | cos θγ ) θ =0
γ

4π
= .
Ez2 λ2 + k2 (Ez2 − p2z )
Taking into consideration the integral value:
Z ∆E
k2 d|k| 1 2∆E E E + |p|
√ = 2 2
ln − 2 − p2 )
ln ,
0
2 2 2 2 2 2
[k (E − p ) + λ E ] k + λ2 E − p λ 2|p|(E E − |p|

we deduce the following expression for Bλ :

 Z 1  
′ dz 2∆E 1 E E + |p|
Bλ = 4π −2 + (pp ) 2 2
ln + ln +
−1 Ez − pz λ 2 |p| E − |p|

© 2011 by Taylor and Francis Group, LLC

 510 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
Z 1 
E′ E ′ + |p′ | ′ dz Ez Ez + |pz |
+ ′ ln ′ − (pp ) ln . (20.34)
|p | E − |p′ | 2 2
−1 Ez − pz |pz | Ez − |pz |

For more compact writing it is convenient to introduce a new integration variable ξ:

r
|pz | θ θ
ξ= = cos2 + z 2 sin2
vEz 2 2

and the functions Φ and G(v, θ) determined by the relations:

q2 p2 sin2 (θ/2) v2
sinh2 Φ = 2
= 2
= sin2 (θ/2), (20.35)
4m m 1 − v2
Z 1
dξ 1 + vξ
G(v, θ) = p ln . (20.36)
cos(θ/2) (1 − v2 ξ 2 ) 2 2
ξ − cos (θ/2) 1 − vξ

Then Bλ takes the ultimate appearance:

 
2∆E 1 1+v 1 − v2
Bλ = 4π 2(2Φ coth 2Φ − 1) ln + ln − cosh 2Φ G(v, θ) , (20.37)
λ v 1−v v sin(θ/2)

where we have allowed for the relation:

(pp′ ) = m2 cosh 2Φ.

The gained experience allows one to calculate Bωmin without any trouble
 Z 1 Z ∆E
dz d|k| ∆E
Bωmin = 4π −2 + (pp′ ) = 8π(2Φ coth 2Φ − 1) ln . (20.38)
−1 Ez − p2z
2
ωmin |k| ωmin

To compare the expressions (20.37) and (20.38) gives the linkage between the infrared
cut-off parameter and the photon mass:
 
2ωmin 1 1 1+v 1 − v2
ln = ln −− cosh 2Φ G(v, θ) . (20.39)
λ 2(1 − 2Φ coth 2Φ) v 1 − v v sin(θ/2)

In the nonrelativistic case this formula results in the relation:

5
ln 2ωmin = ln λ + . (20.40)
6

Knowledge of Bλ permits one to obtain the differential cross section of the electron
scattering in the Coulomb nucleus field with emitting the soft photon whose energy does
not exceed ∆E [32]:
  
(∆E) Z 2 α2 θ α 2∆E
dσe− γ̃→e− γ = 1 − v 2 sin2 2(2Φ coth 2Φ − 1) ln +
4p v sin4 (θ/2)
2 2 2 π λ


1 1+v 1 − v2
+ ln − cosh 2Φ G(v, θ) dΩ. (20.41)
v 1−v v sin(θ/2)

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 511

p− k p− k′

p p′

−p+ k′ −p+ k

FIGURE 20.5
The Feynman diagrams describing the process e− e+ → γγ.

20.3 e− e+ -pair annihilation into two γ-quanta. e− e+ -pair

production under annihilation of two γ-quanta
We shall investigate the process:

e− + e+ → γ + γ, (20.42)

whose diagrams are displayed in Fig. 20.5. These diagrams have precisely the same form
as do the Compton effect diagrams with one exception only: one fermion line and one
photon line interchange their positions. Therefore, the process (20.42) and the Compton
effect represent two crossing-channels of one and the same generalized reaction. In this
case the Compton effect is the s-channel while the process (20.42) does the t-channel. This
circumstance allows us to take advantage the of Compton effect calculation results to find
the cross section of the reaction (20.42).
To match Fig. 20.5 to Fig. 19.1 shows that the diagrams of the process (20.42) follow
from those of the Compton effect under replacement:

p → p− , p′ → −p+ , k → −k1 , k ′ → k2 . (20.43)

In our case the Mandelstam variables are defined as follows:


st = (p− − k1 )2 = (k2 − p+ )2 , 

tt = (p− + p+ )2 = (k1 + k2 )2 , (20.44)

u = (p− − k2 )2 = (k1 − p+ )2 .


Let us find the cross section of the process (20.42) when the initial and final particles
are unpolarized. Then the amplitude square |Ae− e+ →γγ |2 being expressed in terms of the
invariants st , tt , and u, coincides with the similar quantity for the Compton effect under
changing the significance of the invariants only (recall, that the polarization states number

© 2011 by Taylor and Francis Group, LLC

 512 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

for the electron and the photon proves to be equal, and by this reason, the factor 1/2
connected with averaging is conserved). Therefore, all that remains is to anew enumerate
the factors standing in front of |Ae− e+ →γγ |2 in the expression for the differential cross
section. Allowing for
1
(p− p+ )2 − m4 = tt (tt − 4m2 )
4
and coming back to the standard designations:

st = t, tt = s, u = u,

we get the following expression for the differential cross section of the e− e+ -pair annihilation
into two photons [33]:
  
dσ 8πr02 m2 m2 m2 m2 m2
= + + +1 −
dt s(s − 4m2 ) t − m2 u − m2 t − m2 u − m2
 
1 t − m2 u − m2
− + . (20.45)
4 u − m2 t − m2
In the center-of-mass frame the expression (20.45) takes on the simple form:
 
m2 r02 E 2 + p2 + p2 sin2 θ 2p4 sin4 θ
dσ = − dΩ, (20.46)
4pE E 2 − p2 cos2 θ (E 2 − p2 cos2 θ)2

where E = E− = E+ , p = |p− | = |p+ | and θ is an angle between p− and k1 .

Note some inherent features of the cross section in the center-of-mass frame. As long as
two identical particles are present in the final state, then the angle distribution is symmetric
with respect to the angle 900 . Under high energies the peaks caused by the presence of the
denominators E 2 (1 − v cos θ)(1 + v cos θ) are observed both for forward-scattering and for
back-scattering.
In order to evaluate the total cross section of the e− e+ -pair annihilation one must inte-
grate dσ (20.46) over θ from 0 to π while over azimuth ϕ from 0 to π (this corresponds
to that circumstance that the photons appearing in the final state are indiscernible). It is
more convenient, however, to deduce the total cross section in the invariant variables. To
fix the integration limits over t we find t in the center-of-mass frame:
s 1p
t = m2 − 2(p− k1 ) = m2 − 2Eω + 2pω cos θ = m2 − + s(s − 4m2 ) cos θ. (20.47)
2 2
From (20.47) it follows
s 1p s 1p
m2 − − s(s − 4m2 ) ≤ t ≤ m2 − + s(s − 4m2 ).
2 2 2 2
The straightforward integration gives
 √ √ 
2πr02 2 ǫ+ ǫ−4 p
σ= 2 (ǫ + 4ǫ − 8) ln √ √ − (ǫ + 4) ǫ(ǫ − 4) , (20.48)
ǫ (ǫ − 4) ǫ− ǫ−4

where ǫ = s/m2 .
In the nonrelativistic case we obtain from Eq. (20.48)

πr2
σ= √ 0 . (20.49)
ǫ−4

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 513

In the ultrarelativistic region the total cross section acquires the form:

2πr02
σ= (ln ǫ − 1). (20.50)
ǫ
We present the total cross section value in the center-of-mass frame as well
 
πr02 (1 − v 2 ) 3 − v 4 1 + v 2
σ= ln − 2(2 − v ) , (20.51)
4v v 1−v

where v = |p|/E is a velocity of colliding particles.

The formulae describing the positron collision process with the resting electron are of
practical interest too. In this case the invariant ǫ takes the view:
E+
ǫ = 2(1 + γ), γ= ,
m
where E+ is a positron energy in the reference frame where the electron is in rest, and the
total cross section is given by the expression:
" #
πr02 γ 2 + 4γ + 1 p
2
γ+3
σ= ln(γ + γ − 1) − p . (20.52)
γ+1 γ2 − 1 γ2 − 1

In the nonrelativistic region the total cross section is inversely proportional to the positron
velocity:
πr2
σ = 0, (20.53)
v+
while the positron annihilation probability in matter w does not depend on the velocity:

w = Znv+ σ = Znπr02 ,

where n is a number of atoms in the volume unit. So, for the positron flying into lead the
lifetime τ is equal to
1
τ= ≈ 10−10 s.
w
At high energies of the positron (E+ ≫ m) we have
 
πmr02 2E+
σ= ln −1 , (20.54)
E+ m

the positron lifetime tending to zero when v+ → 1.

Further, without going into details we discuss some features of polarization effects for
the process under investigation. To do this we should above all find out the wavefunction
properties of two photon systems. It is convenient to carry out an investigation in the
center-of-mass frame of the photons [34]

k1 = −k2 = k.

Such a system could be determined in all cases apart from the one when two photons are
moving with respect to each other in a parallel way in the same direction. Really, for such
photons the summarized momentum K = k1 + k2 and the total energy ω = ω1 + ω2 must
satisfy the relation:
K2 = ω 2 ,

© 2011 by Taylor and Francis Group, LLC

 514 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

that is, there is no reference frame with K = 0.

The wavefunction of the two photon system in the momentum representation is repre-
sented in the form of three-dimensional second rank tensor Φim (n). This tensor is the
product of the coordinate and spin wavefunctions. The coordinate part of the wavefunction
comprises the spherical function Ylm (n) (n = k/k), which is simply a scalar from the point
of view of the geometrical properties. As distinct from it, the spin part being a bilinear
combination of the spin photon wavefunctions is multi component. Let the index i corre-
spond to the first photon, whereas the one m corresponds to the second photon. Then,
from the wavefunction transversality condition follows:

ni Φim (n) = nm Φim (n) = 0. (20.55)

The mutual rearrangement of the photons means the rearrangement of the indices of the
tensor Φim (n) along with the simultaneous change of the sign n. Taking into consideration
that the photons obey the Bose statistics, we get

Φim (−n) = Φmi (n). (20.56)

Divide the tensor Φim (n) into the symmetric and antisymmetric parts with respect to the
indices i and m:
Φim (n) = Sim (n) + Aim (n).
Obviously, both the symmetric and antisymmetric parts of the tensor Φim (n) must satisfy
the relations (20.55) and (20.56). Then, from (20.56) it is evident that

Sim (−n) = Sim (n), (20.57)

Aim (−n) = −Aim (n). (20.58)

The space inversion on its own does not change the components sign of the second rank
tensor but the one changes the sign of n. Therefore, it follows from (20.57) that the
wavefunction is symmetric with respect to the space inversion, that is, the one corresponds
to the even states of the system. Analogously, it stems from (20.58) that the wavefunction
Aim (n) describes the odd states of the two-photons system.
The antisymmetric tensor on its transformation properties is equivalent to the axial vector
a(n) with the components:
1
al (n) = εlim Aim (n).
2
As
Aim (n) = εiml al (n),
then it follows from the orthogonality condition (20.55)

Aim (n)nm = εiml al (n)nm = [n × a(n)]i = 0,

that is, the vectors n and a(n) are parallel. Consequently, the vector a(n) may be repre-
sented in the form:
a(n) = nϕ(n). (20.59)
The relation (20.59) permits us to correlate the antisymmetric tensor Aim (n) to the zero
value of the system spin, that is, this tensor describes the state with l = j. From (20.58) it
is evident that
a(−n) = −a(n),
what, in its turn, gives
ϕ(−n) = ϕ(n). (20.60)

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 515

The equality (20.60) points to the fact that the scalar ϕ(n) is being built from the spherical
functions only of the even order l (including 0). Thus, we have shown that the tensor
Aim (n) accords with the odd states of the two-photons system with the even moment j.
As far as the symmetric tensor Sim (n) is concerned, then, to find out its spin nature,
it must be expanded into irreducible parts. This operation is reduced to the expansion
′
into Sp[Sim (n)] (the scalar) and the symmetric tensor Sim (n) with the zero spur (five
components). The scalar is associated with the spin zero, and, hence, it governs the states
′
with the even value of j = l. It is clear that the tensor Sim (n) corresponds to the spin value
to equal 2. Allowing for that in this case l must be even, we find:
(i) at even j 6= 0 three states are possible
l = j ± 2, j;
(ii) at odd j 6= 1 two states are possible
l = j ± 1.
The exception are provided by j = 0 with one state l = 2 and j = 1 with one state l = 2.
However, the orthogonality condition of the tensor Sim (n) to the vector n has not been
taken into account in our calculation. Because of this, one must subtract the number of the
states to be associated with a symmetric tensor of the second rank, which is parallel to the
vector n from the obtained states number. Such a tensor has the form:
′′
Sim (n) = ni bm (n) + nm bi (n), (20.61)
where b(n) is a vector which, according to (20.57), satisfies the demand:
b(−n) = −b(n). (20.62)
′′
From (20.61) and (20.62) it follows that the tensor Sim (n) is equivalent to an odd vector
which, in its turn, may be expressed through the spherical functions only of the odd orders
l. Inasmuch as the vector is associated with the spin 1, then one may state: (i) at even
j 6= 0 two states with l = j ± 1 are possible; and (ii) at odd j one state with l = j is possible.
One state with j = 0 and l = 1 is of a special case.
The derived results are gathered into the Table 20.1:

TABLE 20.1
The number of states of the two-photons system.
j Number of even states Number of odd states
0 1 1
1 0 0
2s 2 1
2s+1 1 0

where s is an integer positive nonzero number. Attention is drawn to the fact that the
two-photons system cannot be in the state with j = 1 at all. Therefore, a system having
the moment equal to 1 cannot decay into two photons.
The wavefunction Φij (k) controls the polarizations correlation of photons. The probabil-
ity that two photons simultaneously have the definite polarizations eλ1 and eλ2 is propor-
tional to
Φij (k)eλi 1 ∗ eλj 2 ∗ .

© 2011 by Taylor and Francis Group, LLC

 516 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Thus, if the polarization of one photon is given, then that of a second photon is defined as
follows:
eλj 2 ∼ Φij (k)eλi 1 ∗ . (20.63)
In odd states of the system Φij (k) coincides with the antisymmetric tensor Aij (k). Then
we have
eλ2 eλ1 ∗ ∼ Aij (k)eλi 1 ∗ eλj 1 ∗ = 0,
that is, the polarizations of both photons are mutually orthogonal. In the case of the linear
polarization it means the perpendicularity of their directions while in the case of the circular
polarization it signifies the direction opposite of their rotations.
Now we can discuss the qualitative features of the polarization effects in the center-of-
mass frame for the nonrelativistic and ultrarelativistic regions of the particles to collide.
At v → 0 the process is proceeding only in the case when the system electron+positron is
in the S-state. Since in this case the system possesses the negative parity, then the system
parity of the photons to come into being is negative too. For the odd states, as we have
learned earlier, the photon polarization vectors are mutually orthogonal.
When the electron and the positron are polarized, in the nonrelativistic region the process
(20.42) will proceed if and only if their spins are antiparallel. Actually, as the annihilation
happens into the S-state, then the total moment coincides with the total spin of the particles
which is equal to 1 when their spins are parallel. However, the two-photons system can in
no way possess the states with the total moment to equal 1.
In the ultrarelativistic case the chirality projection operator becomes the helicity projec-
tion operator (see, Section 15.6):
1
Λ(α) (n) → (1 + αγ5 ) ,
2
Then the analysis of the amplitude
s
m2 p̂− − k̂1 + m
Ae− e+ →γγ = −ie2 v(p+ )eλν 1 ∗ (k1 ) γ ν γ µ+
4E− E+ ω1 ω2 (p− − k1 )2 − m2
!
µ p̂− − k̂2 + m ν
+γ γ eλµ2 ∗ (k2 )u(p− ), (20.64)
(p− − k2 )2 − m2
leads us to the conclusion that in the case when one may neglect the electron mass the
reaction will only proceed for the electron and the positron with the opposite signs of the
helicity (in the center-of-mass frame the parallel spins are associated with the different
values of the helicity).
In the ultrarelativistic case the annihilation of the electron and the positron with the
same helicities is possible solely at the inclusion in the process amplitude of the terms
containing m. In order of the magnitude the amplitude of this process differs from that of
the annihilation of the e− e+ -pair with the parallel spins by the factor m/E (a distinction
between corresponding cross sections are proportional to (m/E)2 ).
A more detail analysis of the polarization effects shows that the process in question is
extremely sensitive to the positron polarization. This circumstance is used for determining
the helicities of the positrons to be emitted in a β-decay.
A comprehensive investigation of the polarization effects at the e− e+ -pair annihilation
into the two photons could be found in Refs. [35].
Sometimes the process of the e− e+ -pair annihilation into the two photons is called the
annihilation in the flight. This process is not the dominated one under the positrons ab-
sorption by matter. In the vicinity of 80% the positrons stop to form a positronium, a
nonstable bound state of the electron and the positron.

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 517

In parallel with the conversion of the e− e+ -pair into the photons the process:

γ + γ → e− + e+ (20.65)

is also possible. The corresponding Feynman diagrams are pictured in Fig. 20.6. It is

k p− k′ p−

p p′

k′ −p+ k −p+

FIGURE 20.6
The Feynman diagrams for the process γγ → e− e+ .

obvious that the diagrams of the electron-positron pair production follow from those of the
process (20.42) under the time inversion. Therefore, the amplitude squares of both processes
are identical and their linkage with the cross section differs only in that now we have
t2t s2
I = (k1 k2 )2 = = .
4 4
Thus, the differential cross section of the e− e+ -pair production under the two-photons
annihilation for the case of unpolarized particles is given by the expression [33]:
  
dσ 8πr02 m2 m2 m2 m2 m2
= + + +1 −
dt s2 t − m2 u − m2 t − m2 u − m2
 
1 t − m2 u − m2
− + . (20.66)
4 u − m2 t − m2
Let us cross over to the center-of-mass frame. Taking into consideration the relations:

ω1 = ω2 = ω0 , E− = E+ = ω0 ,

we can represent the expression for the cross section in the form:
p 
r02 m2 ω02 − m2 2ω02 − m2 + (ω02 − m2 ) sin2 θ
dσ = −
4ω03 m2 cos2 θ + ω02 sin2 θ

2(ω02 − m2 )2 sin4 θ
− 2 dΩ. (20.67)
(m cos2 θ + ω02 sin2 θ)2

© 2011 by Taylor and Francis Group, LLC

 518 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

At integration of (20.67) one should allow for that because the two final particles (electron
and positron) are nonidentical, the final result must not be divided into 2 as it took place
in the case of the e− e+ -pair annihilation. Thus, the total cross section is equal to
"  r ! s  #
πr02 m2 2m2 m4 ω0 ω02 m2 m2
σ= 2 + 2 − 4 ln + −1 − 1− 2 1+ 2 . (20.68)
ω02 ω0 ω0 m m2 ω0 ω0

20.4 Positronium lifetime

In the preceding section we have considered the e− e+ -pair annihilation neglecting the
Coulomb force. However, the latter may manifest itself as the appearance of a neutral
system (positronium), a bound state of the electron and the positron. The positronium
represents a lepton analog of atomic bound states, but in contradistinction to them, the
positronium is unstable. We shall solve the problem of finding the lifetime of the positron-
ium in the ground state (S-state).
In the positronium the electron and positron spins may be parallel and antiparallel.
The first case is associated with the triplet state 3 S called the orthopositronium, and the
second case—with the singlet state 1 S referred to as the parapositronium. The s size of the
positronium is approximately twice as big as that of a hydrogen atom (as the reduced mass
of the positronium equals m/2, where m is the electron mass), whereas its binding energy
is only one half of that for the hydrogen atom. An energy level of the ground state of the
parapositronium is lower than that of the orthopositronium by 8.41 × 10−4 eV, and in a
magnetic field there is a possibility for the transitions between these positronium states. It
is noteworthy that the positronium is the simplest system coupled by pure electromagnetic
forces and hence the properties of the free positronium are of special interest for testing the
validity of the QED. As a system of two photons could not be in the state with the total
momentum equal to one, the orthopositronium could not break down into two photons.
Moreover, because the positronium state 3 S is an odd-charge state, we can state that
breaking-down into an even number of photons for the positronium is impossible at all. On
the other hand, the parapositronium is an even-charge state and hence its breaking-down
into any odd number of photons proves to be forbidden. So the principal process determining
the positronium lifetime is a two-photons annihilation in the case of the parapositronium and
a three-photons annihilation in the case of the orthopositronium. Besides, the corresponding
probabilities of the positronium decay may be associated with the annihilation cross sections
of a free e− e+ -pair [36].
From the uncertainty relation it follows that in order of magnitude the momenta of the
electron and the positron in the positronium are

∼ me2 /~

where we have gone to the ordinary system of units. By virtue of the fact that they are
more less than mc, then in order to calculate the annihilation probability one may transfer
to the limit of two particles resting in the origin of coordinates. Then, having multiplied
the annihilation cross section σ by the flux density to be equal to v|ψ(0)|2 , we obtain the
annihilation probability. Here v is the relative velocity of the particles and ψ(r) is the
wavefunction of the ground state of the positronium
1 2~2
ψ(r) = √ e−r/a , a= . (20.69)
πa3 me2

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 519

Let us denote the decay probability of the orthopositronium and the parapositronium by w1
and w0 while the decay probability of the positronium averaging over the spin orientations
by w. Since the relative weights of the states with the spins 1 and 0 are equal to 3/4 and
1/4 respectively, then
3 1
w = w1 + w0 . (20.70)
4 4
takes place. It is evident that the quantity w has the view:

w = w2γ + w3γ + ..., (20.71)

where w nγ is the probability of converting the e− e+ -pair into n photons, which have been
averaged over the spin states. From (20.70) and (20.71) follows

1 3
w0 = w2γ + w4γ + ..., w1 = w 3γ + w 5γ + ...,
4 4
that is,
1 3
w0 ≈ w 2γ , w1 ≈ w 3γ . (20.72)
4 4
Using the formula (20.53) for the total cross section of the e− e+ -pair annihilation into
the two photons at small velocities, one may get the lifetime of the parapositronium from
(20.72)
1 1 2~
τ0 = = = ≈ 1.23 × 10−10 .
w0 4|ψ(0)|2 (v+ σ)v+ →0 mc2 α5
As the level width Γ0 = ~/τ0 is small compared with the energy of the ground state

me4 2α
2
|E0 | = = mc ,
4~2 4
then our treatment of the parapositronium as the quasistationary system is well founded.
Since the lifetime of the orthopositronium is defined by the expression:
1 3
τ1 = = 2
, (20.73)
w1 4|ψ(0)| (v+ σ3γ )v+ →0

our next task is to determine the cross section of the process:

e− + e+ → γ + γ + γ. (20.74)

The process is figured by six diagrams, one of them is shown in Fig. 20.7 (the remaining
five are distinguished from that by rearrangements of the photon momenta k1 , k2 , and k3
only). The corresponding amplitude is as follows:
s
3 m2
A3γ = −ie eλ1 ∗ (k1 )eλν 2 ∗ (k2 )eλρ 3 ∗ (k3 )v(p+ )Oµνρ u(p− ), (20.75)
8E− E+ ω1 ω2 ω3 µ

where
X k̂3 − p̂+ + m p̂− − k̂1 + m
Oµνρ = γµ γν γρ
(k3 − p+ )2 − m2 (p− − k1 )2 − m2
P

and the sum is taken over all rearrangements of the photon numbers 1,2,3 along with
simultaneous rearrangements of the indices of the corresponding vertices µ, ν, and ρ.

© 2011 by Taylor and Francis Group, LLC

 520 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

k3
−p+

k2

k1

p−

FIGURE 20.7
The Feynman diagram for the process e− e+ → γγγ.

To sum over the spin states of the final particles and average over the spin states of the
initial particles leads to the expression:
1X e6
|A3γ |2 = Sp[(p̂− + m)Oµνρ (−p̂+ + m)O µνρ . (20.76)
4 spin 128E− E+ ω1 ω2 ω3

In the case of small velocities of the electron and the positron one may set their three-
dimensional momenta equal to zero to give: p− = p+ = p = (m, 0, 0, 0). Then the expres-
sions both for the propagators and for the density matrix become the simple forms:
k̂3 − p̂+ + m k̂3 + m(1 − γ0 )
2 2
≈ ,
(k3 − p+ ) − m −2mω3
1 1
(p̂− + m) ≈ (γ0 + 1).
2m 2
Adduce the scheme of further calculations. An appreciable simplification is achieved
above all due to using the cross-universality. The amplitude square module involves 36
terms of the type M∗ijk Mi′ j ′ k′ (where the indices i, j, k take the values 1,2,3 and define the
order of following of the photon vertices at the motion along the diagram from the bottom
upwards). It is sufficient to calculate only six terms from 36 ones, for example, the group:
M∗123 (M123 + M132 + M321 + M312 + M231 + M213 ), (20.77)
and, thanks to the cross-universality, all the remaining terms will follow from this group
under the corresponding transformation of the photon momenta. Next, in the expression
(20.77) one may separate the terms that go into each other under different rearrangements of
the photons. The products (pki ) (i=1,2,3) appearing under the taking of spurs are expressed
in terms of the electron mass and the photon energies ω1 , ω2 , ω3 . The products (k1 k2 ), ...
are defined by the conservation law of the four-dimensional momentum:
2p = k1 + k2 + k3 . (20.78)
Thus, for example, having rewritten the expression (20.78) in the form 2p − k3 = k1 + k2
and having squared it, we get
(k1 k2 ) = 2m(m − ω3 ).

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 521

PIt is clear2 that one of the validity tests of the obtained result will be the symmetry of
spin |A3γ | with respect to ω1 , ω2 , and ω3 . In response to straightforward but extremely
cumbersome operations we get
" 2  2  2 #
1X m − ω 1 m − ω 2 m − ω 3
|A3γ |2 = 16e6 + + . (20.79)
4 spin ω2 ω3 ω1 ω3 ω2 ω1

Making use of (20.79), we arrive at the following expression for the differential cross section
of the e− e+ -pair annihilation into three photons [37]:
" 2  2  2 #
e6 m − ω1 m − ω2 m − ω3
dσ3γ = + + ×
(4π)3 π 2 m2 v ω2 ω3 ω1 ω3 ω2 ω1

d3 k1 d3 k2 d3 k3
×δ (3) (k1 + k2 + k3 )δ(ω1 + ω2 + ω3 − 2m) . (20.80)
ω1 ω2 ω3
The first of δ-functions in (20.80) is eliminated by means of the integration over k3

δ (3) (k1 + k2 + k3 )d3 k3 → 1.

To liquidate the second δ-function we first rewrite the differentials in the spherical system
of coordinates
d3 k1 d3 k2 = 4πω12 dω1 2πω22 d(cos θ12 )dω2 ,
where θ12 is an angle between k1 and k2 , and it is meant that the integrations both over
the directions k1 and over the azimuth k2 with respect to k1 have already been fulfilled.
Further, differentiating the equality
q
ω3 = ω12 + ω22 + 2ω1 ω2 cos θ12 ,

we find
ω3
d(cos θ12 ) =
dω3 . (20.81)
ω1 ω2
Now the integration over ω3 is carried out without any trouble and we arrive at the following
expression for the cross section:
" 2  2  2 #
8e6 m − ω1 m − ω2 m − ω3
dσ3γ = + + dω1 dω2 . (20.82)
(4π)3 v ω2 ω3 ω1 ω3 ω2 ω1

Each of the frequencies ω1 , ω2 , ω3 are changed within the limits from 0 to m (the value
m is reached by two frequencies when the third is equal to 0). At a given value of ω1 the
frequency ω2 varies between m − ω1 and m. To integrate over ω1 and ω2 with allowance
made for the photons identity (it is taken into account by multiplying the cross section into
1/6) yields the total cross section of the e− e+ -pair annihilation into the three photons:
Z m Z m Z m 
dσ3γ 8e6 ω1 (m − ω1 ) 2m − ω1
σ3γ = dω1 dω2 = 3 3v
dω1 + +
0 m−ω1 dω 1 dω 2 (4π) 0 (2m − ω1 )2 ω1
  
2m(m − ω1 ) 2m(m − ω1 )2 m − ω1 4αr02 (π 2 − 9)
+ 2 − 3
ln = . (20.83)
ω1 (2m − ω1 ) m 3v
Therefore, the lifetime of the orthopositronium equals
9π~
τ1 = ≈ 1.4 × 10−7 s. (20.84)
2(π 2 − 9)mc2 α6

© 2011 by Taylor and Francis Group, LLC

 522 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

We call attention to the fact that the inequality Γ1 ≪ |E0 | is fulfilled for the orthopositro-
nium even to a greater extent than for the positronium.
Comparison of experimental and theoretical data for the positronium reveals their com-
plete agreement, which serves as one more argument in support of the QED. By the present
time, the positronium is a perfectly understood system. Clearly, its properties in a matter,
specifically its lifetime, are different from the characteristics of a free positronium, being de-
termined by the matter structure. This makes it possible to use the positronium in studies
of the physical and chemical properties of the matter. For example, the positronium may be
used to study high-rate chemical reactions, with the rates comparable to the positronium
lifetime. To this end, one measures a change in the positronium lifetime or splitting of the
energy levels in the ortho- and para states [38].
Another example of a bound leptonic system presents the muonium (e− µ+ ), the bound
metastable state of the electron and the positively charged muon. As leptons take no part
in the strong interaction, the muonium properties are in a very good agreement with the
QED. However, it is impossible to estimate the muonium lifetime on the basis of the QED
because its instability is conditioned by the muon decay due to the weak interaction
µ+ → e+ + ν µ + νe . (20.85)
Measurements of the muonium polarization in various substances became the basis of a
new efficient method of investigating the structure of condensed media, kinetic phenomena,
chemical reactions, phase changes and so on [39]. Also, an interest to muonium is caused
by a search for its transitions to antimuonium, (µ+ e− ) → (µ− e+ ). The detection of such
transitions would be evidence of the violation of the electron or muon flavor.

20.5 Möller scattering e− + e− → e− + e−

We are coming now to the investigation of processes disconnected with the emission and
absorption of the photons. In the electroweak theory these processes are supplemented by
diagrams with exchanges of the gauge Z-boson already in the second order of the perturba-
tion theory. In this sense, calculations of the cross sections of similar processes within the
QED scopes have an approximate character.
Let us examine a reaction of the elastic scattering of the electron by the electron:
e− + e− → e− + e− , (20.86)
that was investigated for the first time in Ref. [40]. The Feynman diagrams for this process
are displayed in Fig. 20.8. The corresponding amplitude is given by:
s 
2 m2 [u(p′2 )γµ u(p2 )][u(p′1 )γ µ u(p1 )]
Ae− e− →e− e− = −ie −
E1 E2 E1′ E2′ (p1 − p′1 )2

[u(p′1 )γµ u(p2 )][u(p′2 )γ µ u(p1 )]
− , (20.87)
(p1 − p′2 )2
where p1 , p2 (p′1 , p′2 ) is the electron momentum in the initial (final) state.
In accordance with the general rules the differential cross section of the process (20.86)
for the unpolarized particles is determined by the expression:
e2 dt
dσ = [P1 (t, u) + P2 (t, u) + P1 (u, t) + P2 (u, t)] , (20.88)
s(s − 4m2 )

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 523

p1 p′1 p1 p′2

k k′

p2 p′2 p2 p′1

FIGURE 20.8
The Feynman diagrams for the process e− e− → e− e− .

where
1
P1 (t, u) = Sp[(p̂′2 + m)γ µ (p̂2 + m)γ ν ]Sp[(p̂′1 + m)γµ (p̂1 + m)γν ],
16t2
1
P2 (t, u) = − Sp[(p̂′2 + m)γ µ (p̂2 + m)γ ν (p̂′1 + m)γµ (p̂1 + m)γν ],
16tu
s = (p1 + p2 )2 , t = (p1 − p′1 )2 , u = (p1 − p′2 )2 .
Simple calculations result in
 
2 2 ′ 2 2 2 ′ 1 s2 + u 2 2 2
P1 (t, u) = {(p p
1 2 ) + (p p
1 2 ) + 2m [m − (p p
1 1 )]} = + 4m (t − m ) ,
t2 t2 2
(20.89)
2 2 2 s 2
s
2

P2 (t, u) = [(p1 p2 ) − 2m ](p1 p2 ) = −m − 3m . (20.90)
tu tu 2 2
To substitute (20.89) and (20.90) into (20.88) yields the following expression for the differ-
ential cross section:
  2 
4πr02 m2 dt 1 s + u2 2 2
dσ = + 4m (t − m ) +
s(s − 4m2 ) t2 2
  
1 s 2 + t2 4 s s
+ 2 + 4m2 (u − m2 ) + − m2 − 3m2 . (20.91)
u 2 tu 2 2
In the center-of-mass frame the Mandelstam variables take the values
θ θ
s = 4E 2 , t = −4p2 sin2 , u = −4p2 cos2 . (20.92)
2 2
Note that in this frame the momentum p and the energy E of the electrons are invariable.
The cross section (20.91) in terms of the invariant variables becomes the form (Möller
formula):
"  2  #
m2 r02 (E 2 + p2 )2 4 3 p2 4
dσ = − + 1+ dΩ. (20.93)
4p4 E 2 sin4 θ sin2 θ E 2 + p2 sin2 θ

As follows from (20.93) there are two peaks in the cross section, one peak for the forward-
scattering and another for the back-scattering.

© 2011 by Taylor and Francis Group, LLC

 524 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the nonrelativistic limit (v ≪ 1, E/m ≈ 1) the expression (20.93) goes into the
Rutherford formula with the inclusion of the exchange (in the Born approximation):
 
r2 1 1 1
dσ = 0 4 + − dΩ, (20.94)
16v sin4 (θ/2) cos4 (θ/2) cos2 (θ/2) sin2 (θ/2)

where v = 2p/m is a relative velocity of the electrons. In the ultrarelativistic region we

have
r2 m2 (3 + cos2 θ)2
dσ = 0 2 dΩ. (20.95)
E 4 sin4 θ
We choose, as the lab frame, the frame in which the second electron was in rest before
collision. Introduce the quantity ∆
E1 − E1′ E′ − m
∆= = 2 , (20.96)
m m
to define the energy transferred by the incoming electron in units of m. Rewriting the
Mandelstam variables in the lab frame

s = 2m(m + E1 ), t = −2m2 ∆, u = −2m(E1 − m − m∆) (20.97)

and inserting their values into (20.91), we obtain

 
2πr2 d∆ ǫ2 (ǫ − 1)2 2ǫ2 + 2ǫ − 1
dσ = 2 0 − + 1 , (20.98)
ǫ − 1 ∆2 (ǫ − 1 − ∆)2 ∆(ǫ − 1 − ∆)

where ǫ = E1 /m. This formula defines the energy distributions for the secondary electrons
generated during passage of the relativistic electron through the matter. Note that the
relation (20.98) is symmetric with respect to the kinetic energies of the final electrons m∆
and m(ǫ − 1 − ∆) due to the identity of both particles.
When trying to obtain the total cross section for the reaction (20.86), we may have an
unpleasant surprise: an integral with respect to the angles turns out to be divergent because
of the propagator denominators. This is caused by the fact that with zero photon mass the
electromagnetic forces have an infinite range of action and, no matter how far separated are
the charge particles, they are scattered from each other anyway. Similar to the asymptotic
properties of the field operators, the whole modern scattering theory may be formulated
only in the case when the forces possess a finite range of action or when all the particles
involved have no matter how small a mass that is finite anyway. But it is known that
the situation may be improved by screening. In reality, experimenters never have to deal
with two isolated charges because there always exist other charges causing screening of the
Coulomb field at long distances.
A reader can find the discussion of the polarization effects for the process (20.86) in Refs.
[41].

20.6 Processes e+ + e− → l + l (l = e−, µ−, τ −)

We shall consider the other cross channel of the reaction (20.86), namely, the electron-
positron pair annihilation with the subsequent production of the electron-positron pair:

e− + e+ → e− + e+ . (20.99)

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 525

p− p′−

p− p−
k k′

−p+ −p′+
−p+ −p′+

FIGURE 20.9
The Feynman diagrams for the process e− e− → ll.

The corresponding Feynman diagrams are pictured in Fig. 20.9. We draw attention to
the different characters of both diagrams. In the former the lines of the initial and final
electron are intersected in the top vertex while the ones of the initial and final positron—in
the bottom vertex. In the latter the lines of the electron and the positron are intersected
in every vertex; in the first vertex the e− e+ -annihilation with emitting a virtual photon
occurs (time flows from left to right), whereas in the second one the e− e+ -production from
a virtual photon takes place. By this reason the second diagram is called the annihilation
one and the first diagram—the scattering one. It should be marked that for the annihilation
diagram the squared four-dimensional momentum of the virtual photon kµ2 = (p− + p+ )2µ is
greater than zero (time-like) while for the scattering diagram kµ2 = (p− − p′− )2µ is less than
zero (space-like). It is also clear that the scattering diagram leads to the big peak in the
amplitude for the forward-scattering.
The transition from the reaction (20.86) to the reaction (20.99) is performed by the
replacement:
p1 → −p′+ , p2 → p− , p′1 → −p+ , p′2 → p′− . (20.100)
In the case of the reaction (20.99) the Mandelstam variables are defined by the expression:

s = (p− + p+ )2 , t = (p+ − p′+ )2 , u = (p− − p′+ )2 .

Again, we are not interested in the particle polarizations. Then in the formula (20.91) we
must fulfill the replacement:

s → u, t → t, u → s. (20.101)

As a result we arrive at the following expression for the differential cross section [42]:
  2 
4πr02 m2 dt 1 s + u2 2 2
dσ = + 4m (t − m ) +
s(s − 4m2 ) t2 2
  
1 u 2 + t2 2 2 4 u 2
 u
2
+ + 4m (s − m ) + − m − 3m . (20.102)
s2 2 st 2 2
In the center-of-mass frame
θ θ
s = 4E 2 , t = −4p2 sin2 , u = −4p2 cos2 (20.103)
2 2

© 2011 by Taylor and Francis Group, LLC

 526 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

and the cross section takes the form:


m2 r02 (E 2 + p2 )2 8E 4 − m4
dσ = 2 4 − 2 2 2 +
16E p sin (θ/2) p E sin (θ/2)
4


12E 4 + m4 4p2 (E 2 + p2 ) 2 θ 4p4 4 θ
+ − sin + sin dΩ, (20.104)
E4 E4 2 E4 2
where E and θ are an energy and a scattering angle in the center-of-mass frame. The
expression (20.104) does not already possess the symmetry with respect to the replacement
θ → π − θ to be typical for scattering of identical particles.
In the nonrelativistic region this formula is reduced to the Rutherford one without ex-
change:
 α 2 dΩ
dσ = 4 , (20.105)
mv 2 sin (θ/2)
where v = 2p/m is a relative velocity of the electron and the positron.
It is instructive to find out the reasons of absence of the exchange addends in (20.105).
In the nonrelativistic approximation the field operators of the electron-positron field may
be considered by two component spinors. Since the relation

u(p)γµ u(p) ≈ ϕ†− (p)ϕ− (p), (20.106)

where ϕ− (p) denotes the big components of the electron spinor, takes place (the similar
relation is valid for the positron as well), then the amplitude corresponding to the scattering
diagram becomes the form:
1 † ′
Asc ∼ [ϕ (p )ϕ− (p− )][ϕ†+ (p′+ )ϕ+ (p+ )]. (20.107)
t2 − −
Allowing for that in the center-of-mass frame t2 = −q2 , we come to the conclusion that
the contribution from the scattering diagram exactly gives the Rutherford formula. In this
approximation the contribution coming from the annihilation diagram is equal to zero. The
mathematical explanation of this fact is as follows. At finding of |Ae− e+ →e− e+ |2 the terms

A†sc Aan + A†an Aan

where
Aan ∼ [ϕ†+ (p+ )ϕ− (p− )][ϕ†− (p′− )ϕ+ (p′+ )],
vanish at the cost of the orthogonality of the electron and positron spinors.
In the ultrarelativistic limit we get

r02 m2 (3 + cos2 θ)2 cos4 (θ/2)

dσ = dΩ, (20.108)
E2 4 sin4 θ
that is, this cross section differs from that of the electron-electron scattering by the factor
cos4 (θ/2) only.
Now we investigate the processes of the µ− µ+ -, τ − τ + -pair production under the electron-
positron annihilation. This task is analogous to the previous one but significantly simpler.
Because the QED Lagrangian contains the electron, muon, and tau-lepton currents in the
form of the individual addends, then the Feynman diagrams do not have any vertices in
which the µ− → e− -, τ − → e− -, τ − → µ− -conversions or the µ− e+ -, τ − e+ -, µ− τ + -pair
annihilations would occur. Every chain of diagram lines (from entrance to exit or whole

© 2011 by Taylor and Francis Group, LLC

 Scattering of electrons and positrons 527

closed loop) conserves the lepton flavor, that is, the conservation law of the individual lepton
flavor:
XN XN XN
Lei = const, Lµi = const, Lτ i = const (20.109)
i=1 i=1 i=1

is valid in the QED. Because of this, for example, the process of the muon pair production
under the e− e+ -pair annihilation in the second order of the perturbation theory is repre-
sented by only one diagram as opposite to the process (20.99) (see, Fig. 20.10). In the

e− µ−
γ

e+ µ+

FIGURE 20.10
The Feynman diagram for the process e− e+ → µ− µ+ .

invariant variables

s = (pe− + pe+ )2 , t = (pe− − pµ+ )2 , u = (pe− − pµ− )2

the differential cross section for the case of the unpolarized particles is provided by the
expression [43]:
 2 
e4 dt t + u2 2 2 2 2
dσ = + (m e + m µ )(2s − m e − m µ ) . (20.110)
4πs3 (s − 4m2e ) 2
Neglecting the electron mass, we find the limits of the integration over t. For this purpose
we carry out the transition to the center-of-mass frame and next return to the Mandelstam
variables in the obtained expression for t:
s 1q
t ≈ m2µ − 2E 2 + 2pe pµ cos θ = m2µ − + s(s − 4m2µ ) cos θ. (20.111)
2 4
So, at a given s the value of t lays within the interval:
s 1q s 1q
m2µ − − s(s − 4m2µ ) ≤ t ≤ m2µ − + s(s − 4m2µ ). (20.112)
2 4 2 4
Fulfilling the integration over t into (20.110), we gain the total cross section of the muon
pair production: r " #
4πα2 4m2µ 2m2µ
σ= 1− 1+ . (20.113)
3s s s
√ √
The cross section (20.113) goes into the maximum when s ≈ 1.7 s0 , where s0 is a value
of s on the reaction threshold (in the center-of-mass frame s0 = 4m2µ ). In this case the
value of the cross section proves to be 20 times as small as that of the cross section of the
electron-positron annihilation into the two photons under the same energy.
As follows from (20.113), if the relative velocity of the created particles tends to zero
(s → 4m2µ ), the cross section becomes zero. However, this inference is invalid. The formula

© 2011 by Taylor and Francis Group, LLC

 528 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

(20.113) is no longer valid close to the reaction threshold. As in this case the Coulomb in-
teraction becomes considerable, the created muons could not be considered as free particles.
In the nonrelativistic quantum scattering theory it is demonstrated that the cross section
of the reaction describing two slowly attracted particles at v → 0 tends to the constant
limit. By the detailed balancing principle —for an isolated system the probability of a di-
rect transition n → m is equal to that of the reverse transition m → n—the cross section of
this reaction is associated with the cross section of the reverse reaction, that is, the reaction
of producing the same charged particles. In consequence, when two charged particles are
created, the cross section close to the threshold is tending to the constant limit. This means
that upon the transition through a threshold such a reaction takes place immediately with
the finite value of the cross section.

© 2011 by Taylor and Francis Group, LLC

 21
Radiative corrections

But the Raven, sitting lonely on the placid bust, spoke

only
That one word, as if his soul in that one word he did
outpour.
Nothing father then he uttered—not a feather then
he fluttered—
Till I scarcely more than muttered “Other friends have
flown before—
On the morrow he will leave me, as my Hopes have
flown before.”
Then the bird said “Nevermore.”
Edgar Allan Poe, “The Raven”

So far the calculations for scattering cross sections have been performed being restricted
to the Born approximation. The matrix elements were fairly simple in form involving
no integration with respect to the four-dimensional momenta of virtual particles. The
corresponding diagrams, including a single internal line, looked just like the trees we have
drawn in early childhood. In analogy, Born diagrams are usually called the tree diagrams,
whereas the Born approximation is called the tree approximation.
In this section we investigate higher orders of the perturbation theory. This is motivated
not only by the necessity to improve the accuracy of the calculation but also by our wish to
make sure of the consistency and correctness of the QED. In the process we have to consider
two problems: (i) elucidation of the meaning for the parameters m and e involved in the
Lagrangian, and (ii) appearance of divergent integrals.
In concept, in the original Dirac equations m and e represent a mass and a charge of
a free electron, i.e., the electron completely isolated from the effect of an electromagnetic
field. In other words, m and e are the characteristics of a particular hypothetical object—
the bare electron. It is clear that an energy spectrum of the interacting fields is no longer
coincident with that of free fields. Because of this, the mass of the bare electron should be
different from that of the physical electron, i.e., different from a minimum energy of the
singly charged state of the interacting fields. It makes sense to assume that the electron
mass is given by the quantity:

(Ψp , HΨp )
mexp ≡ mR = |p=0 ,
(Ψp , Ψp )

where H is a total Hamiltonian of a system, calculated within the required accuracy in

higher orders of the perturbation theory. In the case of a physical charge of the electron
we proceed in a similar way. To illustrate, a physical charge may be determined in a
investigation of low-frequency electromagnetic waves scattered from the electron at rest.
As the process under study is purely classical, in the limiting case p → 0 no quantum
correction should occur. Then a physical charge of the electron eexp ≡ eR is determined
as a factor entering into the Thomson formula for the effective scattering of the photon at

529
© 2011 by Taylor and Francis Group, LLC
 530 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

p → 0,
 2
8π e2R
σ= .
3 4πmR
Thus, we should find the relations:
e = e(eR ), m = m(mR )
and eliminate the parameters m and e having expressed them in terms of the experimental
values of the mass and the charge. Such a redefinition of the theory parameters is called
the renormalization and makes a compound part of any calculation to be compared with
the experimental results.
The appearance of the divergences stems from a singular character of the Green functions
at relatively short distances. In the momentum space this is equivalent to the fact that the
associated Fourier images at infinity are decreasing rather slowly. We shall illustrate the
divergences appearance by the example of the electron scattering from the Coulomb field of
a nucleus. The Feynman diagrams describing this process in the first order and the radiation
correction in the third order of the perturbation theory are shown in Fig. 21.1. The matrix

q = p2 − p1

q = p2 − p1

p1 p2 p1 p2
(a) (b)

q = p2 − p1

p1 − k p2 − k

p1 k p2

(c)

FIGURE 21.1
The Feynman diagrams (a)-(c) describing the electron scattering from the Coulomb field of
the nucleus.

element corresponding to the diagrams of Fig.21.1(a) is proportional to the integral:

Z
p̂2 − k̂ + m p̂1 − k̂ + m 1
Iν = γµ 2 2
γν 2 2
γ µ 2 d4 k. (21.1)
(p2 − k) − m (p1 − k) − m k

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 531

In the integrand (21.1) the power of the momentum k, with respect to which the integration
is performed, is six both in the numerator and denominator. Since the integration over k is
going from 0 to ∞, at the upper limit the integral as well as the matrix element is infinite
or divergent, and in this case the divergence is logarithmic. The divergences deriving from
great momenta are referred to as ultraviolet. Recall that the integrals may be divergent at
small values of k as well as due to zero mass of the photon, and in this case the divergences
are termed as infrared.
The appearance of the divergences and the results of their subsequent elimination are
suggestive that quantum field theory is similar to the land described by the famous Russian
tale-teller E. Schwartz where people are living according to the principle: “No matter what
happens, we always have a holiday.” The appearing divergences force us to redefine (reg-
ularize) the Green functions. On the other hand, inclusion of the interactions necessitates
a change from the characteristics of bare particles to those of the physical ones. Owing
to the regularized Green functions and also due to the redetermined mass and charge of
the electron, ultraviolet divergences disappear from all the observables, whereas the infrared
divergences, as noted above, are canceled with the allowance for higher orders of the pertur-
bation theory. In this manner by the solution of two a priori independent problems we can
derive the finite results of the QED, being in an excellent agreement with the experimental
data.

21.1 Calculations of integrals over virtual momenta

Let us consider a diagram having an arbitrary number of external lines to correspond to
virtual particles. In the most general case the number of the vertex δ-functions may happen
to be insufficient for the convolution of all integrals. The necessary condition for this is the
presence of closed loops in the diagram. So, if the number of the diagram vertices is less than
the one of the internal lines, then the corresponding matrix element contains the integral
over virtual four-dimensional momenta of the kind:
Z
Il = d4 k1 ...d4 kl F (k1 , ...kl ), (21.2)

where l is a number of topologically independent loops, and F (k1 , ...kl ) represents the
product of vertex factors multiplied by the electron and photon propagators. In the QED
the integral (21.2) may be reported in terms of the product of integrals, each of them has
the following structure: Z
F (k)
I = d4 k , (21.3)
α1 ...αn
where F (k) is a polynomial of some power in k, and α1 , ...αn are polynomials of the second
power with respect to k. Therefore, our prime goal is to get acquainted with the technique
of calculating the integrals of the kind (21.3).
To compute the integrals (21.3) we make use of the parametrization method by Feynman
[7]. Here the starting point is the formula:
Z 1 Z 1 Z 1
1 δ(ξ1 + ξ2 + ... + ξn − 1)
= (n − 1)! dξ1 dξ2 ... dξn =
α1 ...αn 0 0 0 [α1 ξ1 + α2 ξ2 + ... + αn ξn ]n
Z 1 Z x1 Z xn−2
dxn−1
= (n − 1)! dx1 dx2 ... ,
0 0 0 [α x
1 n−1 + α (x
2 n−2 − xn−1 ) + ... + αn (1 − x1 )]n
(21.4)

© 2011 by Taylor and Francis Group, LLC

 532 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

or the equivalent one:

Z 1 Z 1 Z 1
1
= (n − 1)! un−2
1 du1 un−3
2 du2 ... dun−1 ×
α1 ...αn 0 0 0

1
× , (21.5)
[α1 u1 ...un−1 + α2 u1 ...un−2 (1 − un−1 ) + ... + αn (1 − u1 )]n
where x1 = u1 , x2 = u1 u2 , ..., xn−1 = u1 u2 ...un−1 . One may differentiate the equality
(21.4) with respect to αn and gain the identity of the type:
Z 1 Z x1 Z xn−2
1
= n! (1 − x1 )dx1 dx2 ... dxn−1 ×
α1 ...α2n 0 0 0

1
× . (21.6)
[α1 xn−1 + α2 (xn−2 − xn−1 ) + ... + αn (1 − x1 )]n+1
The advantage of the last formula implies that when there are two identical factors it takes
one integration less than the factors number.
Let us prove the validity of the relation (21.5). At n = 2 we may write
  Z α2
1 1 1 1 1 dx
= − = . (21.7)
α1 α2 α2 − α1 α1 α2 α2 − α1 α1 x2

Transferring to new variable x = α1 u + α2 (1 − u), we obtain

Z 1
1 1
= du . (21.8)
α1 α2 0 [α1 u + α2 (1 − u)]2

which is what we set out to prove. Note that the equality (21.8) is true for all values of
α1 and α2 . If α1 and α2 have, however, the opposite signs, then u should be regarded as a
complex variable and the integration contour must depart from the real axis to bypass the
singularity at the point u = α2 /(α2 − α1 ).
Next, we prove that if the relation (21.5) is fulfilled under some n, it will be fulfilled under
n+ 1 as well. Since the operations of integrating over different ui in (21.5) are commutative,
then Z 1 Z 1 Z 1
1 n−2 n−3 dun−1
= (n − 1)! u1 du1 u2 du2 ... nα
, (21.9)
α1 ...αn αn+1 0 0 0 A n+1

where An is a denominator being in the right-hand side of the formula (21.5). On the other
hand Z 1
1 un−1
0 du0
n
= n+1
. (21.10)
nA αn+1 0 [Au0 + αn+1 (1 − u0 )]

From (21.9) and (21.10) it immediately follows

Z 1 Z 1 Z 1
1 dun−1
= n! un−1
0 du0 un−2
1 du1 ...
α1 ...αn αn+1 0 0 0 [Au0 + αn+1 (1 − u0 )]n+1

which proves our statement.

Now, by application (21.4) the integral (21.3) could be represented in the view:
Z Z 1 Z x1 Z xn−2 Z
F (k) F (k)dk
d4 k = (n − 1)! dx1 dx2 ... dxn−1 , (21.11)
α1 ...αn 0 0 0 [(k − a)2 + b]n

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 533

q0 q0
C
C

(a) (b)

FIGURE 21.2
The rotation of the integration path in (a) and (b).

where a four-dimensional vector a and a scalar b depend on parameters x1 , ...xn−1 . Here

the integration over the variable k0 is carried out in accordance with the rule of the poles
detour, that is, along the contour C shown in Fig. 21.2(a). In doing so, every of the factors
α−1
i in the left-hand side of (21.11) has two poles over the variable k0 which are bypassed
along the contour C. The integrand in (21.11) has 2n simple poles over the variable k0
whereas, after the Feynman parametrization, it has only two nth order poles with the same
detour rule along the contour C. Since the singular points of the integrand place in the
real axis, then one may swing the contour C through π/2, that is, bring it into coincidence
with the imaginary axis (Fig.21.2(b)). As this takes place, k0 is changed by ik4 (k4 is a real
quantity) and k02 − k2 converts to −(k42 + k2 ). Thus, the rotation of the integration path
around the angle π/2 counterclockwise (Wick rotation) corresponds to the transition in the
k-space from the pseudo-Euclidean metric to the Euclidean one.
Using the spherical system of coordinates in the four-dimensional Euclidean space

k1 = k cos ϕ sin θ sin ξ, k2 = k sin ϕ sin θ sin ξ, 


k3 = k cos θ sin ξ, k4 = k cos ξ, (21.12)


2
d4 k = k 3 dk sin θdθdϕ sin ξdξ,


we may get
d4 k = dk0 d3 k → d4E k = idk4 d3 k = ik 3 dk sin2 ξdξdΩ,
where dΩ is a four-dimensional solid angle in the Euclidean space. Having fulfilled the
integration over the angles, we arrive at
d4 k → 2iπ 2 k 3 dk. (21.13)
If the integral (21.11) is divergent, then one conveniently carries out the integration first
over some finite invariant region Σ characterized by the value N , where the limit N → ∞ is
associated with the whole infinite four-dimensional k-space. Such a finite invariant region
may be specified, for example, by the inequalities:
(ks)2
|k 2 | ≤ N 2 , ≤ N 2,
s2
where s is an arbitrary time-like four-dimensional vector [44]. After the Wick rotation the
region Σ is transformed to a four-dimensional sphere.
It is convenient to begin calculating the integrals of the kind (21.11) with the integral:
Z
1
I2 (p, l) = d4 k 2 . (21.14)
Σ [k − 2(pk) + l]2

© 2011 by Taylor and Francis Group, LLC

 534 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The shift of the origin of coordinates for divergent integrals does not leave these integrals
unchanged. As a result of such a shift an additive constant is added to divergent integrals.
However, as will be shown later, the similar transformation of coordinates does not lead to
changing the values of the logarithmically divergent integrals. After shifting the coordinates
origin k − p → k, the integral (21.14) takes the form:
Z
d4 k
J2 (p, l) = 2 2 2
. (21.15)
Σ′ [k + l − p ]

Carrying out the Wick rotation and realizing the integration, we get
Z  
d4E k 2 L′2
J2 (p, l) = = iπ ln 2 −1 , (21.16)
[−k 2 + l − p2 ]2 p −l
where L′ is an integration sphere radius associated with the region Σ′ . When we took the
integral (21.15) over the region Σ, then the quantity L would appear in Eq. (21.16) rather
than L′ . However L′ differs from L by a finite value. Therefore, if L′ is sufficiently big,
then ln L′ is distinguished from ln L by a small value of the order of L−1 . Thus, we have
the right to set ln L′ = ln L in the formula (21.16).
Further, we compute the logarithmically divergent integral:
Z
µν k µ k ν d4 k
I3 (p, l) = 2 3
.
Σ [k − 2(pk) + l]

After the coordinates origin shift k − p → k, it becomes the form:

Z
µν (k µ + pµ )(k ν + pν )d4 k
I3 (p, l) = . (21.17)
Σ′ [k 2 + l − p2 ]3
Breaking (21.17) into parts and yielding the Wick rotation, we obtain
Z Z Z  
kµ kν d4 k 1 k 2 d4 k 1 k 2 d4E k ig µν π 2 L′2 3
2 ′ 3
= gµν 2 ′ 3
= − gµν = ln ′ − ,
Σ′ (k − l ) 4 Σ′ (k − l ) 4 (−k 2 − l′ )3 4 l 2
Z Z
k µ d4 k µ ν d4 k iπ 2 pµ pν
2 ′ 3
= 0, p p 2 ′ 3
= ,
Σ′ (k − l ) Σ′ (k − l ) 2l′
where l′ = p2 − l and we have allowed for that integrals of an odd function over symmetric
limits vanish. Thus, the integral I3µν (p, l) appears to be equal to
 
ig µν π 2 L2 3 iπ 2 pµ pν
I3µν (p, l) = ln 2 − + . (21.18)
4 p −l 2 2(p2 − l)
Integrating the formula (21.18) over pν , we find
Z  
k µ d4 k L2 3
I2µ (p, l) = 2 − 2(pk) + l]2
= ip µ 2
π ln 2−l
− + g µ (L2 , l). (21.19)
Σ [k p 2
To determine the quantity g µ (L2 , l) we substitute pµ = 0 into Eq.(21.19). Then, in the left-
hand side we obtain the integral of an odd function over the symmetric limits and (21.19)
gives the result:
g µ (L2 , l) = 0. (21.20)
According to (21.20) the formula (21.19) acquires the final form:
Z  
µ k µ d4 k µ 2 L2 3
I2 (p, l) = 2 2
= ip π ln 2 − . (21.21)
Σ [k − 2(pk) + l] p −l 2

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 535

Integrating (21.16) over l, we come to the following integral:

Z
d4 k  
I1 (p, l) = 2
= −iπ 2 l ln L2 + (p2 − l) ln(p2 − l) − p2 + g(L2 , p2 ) , (21.22)
Σ k − 2(pk) + l

where g(L2 , p2 ) is some function of L2 and p2 . Differentiating (21.22) with respect to pµ ,

we get
Z  
µ k µ d4 k 2 µ 2 ∂g(L2 , p2 )
J2 (p, l) = 2 2
= −iπ p ln(p − l) + . (21.23)
Σ [k − 2(pk) + l] ∂p2
To compare (21.23) with (21.21) results in:

∂g(L2 , p2 ) 3
2
= − ln L2 ,
∂p 2
that is,
3 2
p − p2 ln L2 + g0 (L2 ).
g(L2 , p2 ) =
2
To substitute the obtained value of g(L2 , p2 ) into (21.22) gives
Z  
d4 k 2 2 2 2 1 2 2 2 2
I1 (p, l) = 2
= −iπ l ln L + (p − l) ln(p − l) + p − p ln L + g 0 (L ) .
Σ k − 2(pk) + l 2
(21.24)
To find g0 (L2 ) we set p2 = 0 in the both sides of the expression (21.24). Then, with
allowance made for Z  
d4 k 2 2 L2
2
= −iπ L + l ln ,
Σ k +l −l
we gain finally  
2 2 2 L2 1 2
I1 (p, l) = iπ L + (p − l) ln 2 − p . (21.25)
p −l 2
It should be noted that the integral (21.25) allows us to find practically all integrals to
appear under calculation of radiative corrections (RC). Thus, for example, the elementary
differentiation of (21.25) over pµ and l leads to the following expressions (recall, the quantity
p2 − l enters into (21.25) under the module sign):
Z  
kµ d4 k 1 ∂I1 (p, l) 2 L2 3
2 2
= = iπ p µ ln − , (21.26)
Σ [k − 2(pk) + l] 2 ∂pµ p2 − l 2
Z  
kµ kν d4 k 1 ∂ 2 I1 (p, l) iπ 2 L2 3 iπ 2
2 3
= = g µν ln − + pµ pν , (21.27)
Σ [k − 2(pk) + l] 8 ∂pµ ∂pν 4 p2 − l 2 2(p2 − l)
Z
kν d4 k 1 ∂ 2 I1 (p, l) iπ 2
2 3
= − = pν , (21.28)
Σ [k − 2(pk) + l] 4 ∂pν ∂l 2(p2 − l)
Z
d4 k 1 ∂ 2 I1 (p, l) iπ 2
2 3
= 2
= . (21.29)
Σ [k − 2(pk) + l] 2 ∂l 2(p2 − l)
Z  
d4 k ∂I1 (p, l) 2 L2
2 2
=− = iπ ln 2 −1 . (21.30)
Σ [k − 2(pk) + l] ∂l p −l
The quadratically divergent integral:
Z
(2) kµ kν d4 k
Iµν (p, l) = .
Σ [k 2 − 2(pk) + l]2

© 2011 by Taylor and Francis Group, LLC

 536 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

demands some additional efforts. To estimate it one may integrate (21.27) over l. A constant
appearing under this integration could be defined with the help of the integral:
Z
µνλ k µ k ν k λ d4 k
J3 (p, l) = 2 3
.
Σ [k − 2(pk) + l]

It may be done by application of the formula:

Z
1 1 n(b − a)dx
n
− n
= , (21.31)
a b [(a − b)x + b]n+1

where J3µνλ (p, l) has been taken as a and the vanishing integral
Z
k µ k ν k λ d4 k
Σ [k 2 + l]3

has been used as b. We give the final result of calculations:

Z  
kµ kν d4 k iπ 2 2 L2 5p2 − 3l L2
2 2
= g µν (l − p ) ln + + +
Σ [k − 2(pk) + l] 2 p2 − l 6 2
 
L2 11
+iπ 2 ln 2 − pµ pν . (21.32)
p −l 6

21.2 Regularization methods

The appearance of divergent integrals into the matrix elements induce us to do violence to
the conventional mathematical apparatus, namely, we have to sum up and multiply these
matrix elements as if they are finite. Though we have succeeded in driving the divergences
out of a theory at the final stages of the calculations, we are not saved from regularization
of the theory at the intermediate stages. This procedure comprises in the replacement of
the original expressions by more smooth ones, and, as a result, all the integrals become
finite.
As has been established in Section 14.9, the causal Green functions are the generalized
singular functions possessing singularities at the light cone. The divergent integrals consid-
ered in the previous section are associated with closed loops of the diagrams representing
the products of the causal Green functions with the coincident arguments. These products
include the squared generalized singular functions of the form δ(x2 ), x−2 and hence are
not well defined. In other words, the products of the Green functions require a supplement
to the definition, as one would expect from their membership to the generalized functions.
Indeed, the generalized function is specified by the integration rules for its product with
sufficiently regular functions, giving no special receipt for the integration of the product of
several generalized functions.
Thus, our task is to formulate the operation of replacing the causal Green functions
∆c (x) or integrals of their products by some approximating regularized expressions not
including any ultraviolet divergences or, what is the same when we work in the coordinate
representation—the singularities at the light cone. As it takes place, two approaches are
possible.

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 537

By the first, the so-called covariant regularization is based on the derivation of regular ap-
proximations to the causal functions themselves, i.e., on the substitution of some regularized
approximations for the generalized singular functions

∆c (x) → reg ∆c (x).

It is obvious that the products of new propagators become regular automatically.

In the second approach the original expressions for the propagators are unchanged while
both the propagator products and integrals over the virtual four-dimension momenta are
redefined by means of regularized approximations. Among such a regularization kind is
above all a cut-off regularization (CR). In its simplest and well-known variant, after the
Wick rotation, the integration for each loop is limited to a compact sphere of radius L with
k 2 ≤ L2 that makes all the integrals convergent. One more way consists in the change-
over to the discrete space-time (DR), i.e., in the assumption that xµ takes on only discrete
values associated with, for example, sites of a regular lattice. Of course, cut-off at short
distances is equivalent to that at large values of the momenta. But these two approaches
are not relativistically covariant, being unable to retain the initial gauge symmetry of the
theory. In contrast to the CR and DR, the dimensional regularization, despite its belonging
to the same class, retains the symmetry properties of the corresponding nonregularized
expressions.
Covariant regularization. This regularization was suggested in Ref. [45] and has
come to be known as the Pauli-Villars regularization. This scheme possesses the relativistic
covariance and conserves the gauge invariance of a theory under appropriate modification.
The Pauli-Villars regularization resides in the replacement of singular Green functions of a
free field Gc (x; m) by a linear combination:
X
Gc (x; m) → reg Gc (x; m) = Gc (x; m) + ci Gc (x; Mi ) (21.33)
i

The quantities Gc (x; Mi ) are causal functions of fictitious fields with the masses Mi and
the coefficients ci standing in front of them sorted out so that the regularized function
reg Gc (x; m) becomes sufficiently regular in the vicinity of the light cone or, what is the
same, its Fourier image reg Gc (p; m) is sufficiently fast decreasing in the region of big values
of |p2 |.
So, it follows from the formula (14.110) that in the zero spin case the most strong singu-
larities at the light cone δ(λ) and λ−1 enter into ∆c (x; m2 ) with coefficients not depending
on the mass m. Because of this, setting ci
X
1+ ci = 0, (21.34)
i

we arrive at the expression for reg ∆c (x; m2 ), which does not contain the singularities δ(λ)
and λ−1 . In the region of big values of |p2 | its Fourier image reg ∆c (p; m2 ) will be decreasing
as |p2 |−2 . In a similar manner, the condition:
X
m2 + ci Mi2 = 0 (21.35)
i

ensures the singularities absence of the kind θ(λ), ln |λ| into reg ∆c (x; m2 ) and leads to the
fact that reg ∆c (p; m2 ) is decreasing as |p2 |−6 in the ultraviolet region. It is remarkable
that the conditions (21.34) and (21.35) ensure the Green functions regularization in the
QED as well.

© 2011 by Taylor and Francis Group, LLC

 538 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The Pauli-Villars regularization represents a specific modification of the one-particle prop-

agator. In the simplest variant, it is reduced to the introduction of one supplementary mass
M with the factor c = −1. Thus, for example, in the momentum representation we have
for the scalar field
1 1 m2 − M 2
reg ∆c (k) = − = . (21.36)
k 2 − m2 k2 − M 2 (k 2 − m2 )(k 2 − M 2 )

As follows from (21.36), the regularization procedure results in the considerably changed
behavior of the propagator in the ultraviolet region or, equivalently, in the neighborhood
of the light cone. This particular case reveals the essence of the regularization operation.
Any elimination of infinities is related to the existence of a subtraction procedure. On the
other hand, propagators are bilinear combinations of the field functions, and the negative
sign appearing in the propagator results from quantization of a theory with the use of an
indefinite metric. Consequently, in the case of a regularization of the photon propagator:
gµν gµν
reg ∆cµν (k) = 2
− 2 (21.37)
k k − M2
the second addend in (21.37) should be considered as a result of an additional interaction
between the electron-positron field and vector field, whose quanta possess the mass M and
whose norm is not positively defined.
After the intermediate calculations are completed, we have to get rid of fictitious fields,
that is, to take away the regularization. As follows from Eq. (21.36), with M tending
to infinity we arrive at the initial propagator. This allows us to state that removal of the
Pauli-Villars regularization is realized when going to the limit Mi → ∞. However, it is naive
to believe that application of this procedure for the expressions involving the propagator
products may bring back the same results as are possible without the regularization. Let
us estimate the effect of the regularization taking the one-loop diagram with two scalar
internal lines in Fig. 21.3 as an example. The corresponding matrix element contains the

p p

p−k

FIGURE 21.3
The loop with two virtual scalar particles.

product of two propagators with coincident arguments:

Z Z
M ∼ d4 xeikx d4 yeiky ∆c (x − y; m2 )∆c (y − x; m2 ).

After the transition to the momentum representation we obtain the integral:

Z
1 1
I(p2 ) = d4 k . (21.38)
[(k − p) − m ] [k − m2 ]
2 2 2

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 539

This logarithmically divergent integral will be the subject of our investigation. It is clear
that in order to make this integral finite, one sufficiently carries out the replacement (21.36)
in one of the propagators. This results in
Z  
1 1 1
reg I(p2 ) = d4 k − =
(k − p)2 − m2 [k 2 − m2 ] [k 2 − M 2 ]
Z
m2 − M 2
= d4 k .
[(k − p)2 − m2 ][k 2 − m2 ][k 2 − M 2 ]
Obviously, reg I(p2 ) could be expanded into such two addends, that, after tending M to
the infinity, the former diverges at the integration over the four-dimensional momenta while
the latter is convergent. The required splitting is given by the formula:

reg I(p2 ) = ID (0) + IC (p2 ), (21.39)

where Z
m2 − M 2
ID (0) = d4 k , (21.40)
(k 2 − m2 )2 (k 2 − M 2 )
Z  
2 4 m2 − M 2 1 1
IC (p ) = d k 2 − 2 . (21.41)
(k − m2 )(k 2 − M 2 ) (k − p)2 − m2 k − m2
Because of the addend IC (p2 ) does not contain any divergencies, in the integrand we may
transform to the limit M → ∞ what results in:
Z
2(pk) − p2
IC (p ) = d4 k 2
2
. (21.42)
(k − m2 )2 [(k − p)2 − m2 ]

The expressions (21.40) and (21.41) represent the first two terms of the expansion of the
function reg I(p2 ) into the Taylor series in the vicinity of the point p2 = 0. Since only the
divergent part ID (0) involves the fictitious mass M , then one may connect the quantity M
with the cut-off parameter.
So, we have shown that the finite part of the regularized integral ID (p2 ) is given by the
difference of the regularized integral and the first term of its expansion into the Taylor
series. It should be emphasized, that the procedure of obtaining the finite expressions
for the matrix elements by application of expanding into the Taylor series and subsequent
subtraction of divergent integrals is at the heart of the foundation of the renormalization
theory.
Making use of Eq. (21.6), one may present the integral (21.42) in the following form:
Z 1 Z
2(kp) − p2
IC (p) = 2 (1 − x)dx d4 k =
0 [(k − xp)2 − m2 + x(1 − x)p2 ]3
Z 1 Z
(2x − 1)p2
=2 (1 − x)dx d4 k . (21.43)
0 [k 2 − m2 + x(1 − x)p2 ]3
At the last stage we have fulfilled the variables change-over k → k+xp and dropped the term
that is linear on k because its contribution equals zero due to the oddness of the integrand.
To estimate the integral over momenta we carry out the Wick rotation and transit to the
spherical coordinates. As a result we get
Z Z ∞
1 k 2 d(k 2 )
d4 k 2 = iπ 2
. (21.44)
[k − m2 + x(1 − x)p2 ]3 0 [−k 2 − m2 + x(1 − x)p2 ]3

© 2011 by Taylor and Francis Group, LLC

 540 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Allowing for the well-known formula for the beta-function:

Z ∞ m−1
z dz 1 Γ(m)Γ(n − m)
2 )n
= 2 n−m , (21.45)
0 (z + a (a ) Γ(n)

one can calculate the remaining integral

Z
1 iπ 2
d4 k 2 2 2 3
=− . (21.46)
[k − m + x(1 − x)p ] 2[m − x(1 − x)p2 ]
2

Then we arrive at the following expression for IC (p2 ):

Z ∞
2 2 (1 − x)(2x − 1)p2
IC (p ) = −iπ dx 2 . (21.47)
0 [m − x(1 − x)p2 ]

The evaluation of the divergent part ID (0) is fulfilled in the similar way
Z 1
2 2 xdx 2 M2
ID (0) = iM π ≈ iπ ln (21.48)
0 x(m2 − M 2 ) + M 2 m2

(in the numerator we have neglected by the term m2 compared with M 2 ). Thus, we
ultimately gain
Z ∞
2 2 M2 2 (1 − x)(2x − 1)p2
reg I(p ) = iπ ln 2 − iπ dx 2 , (21.49)
m 0 [m − x(1 − x)p2 ]

where the finite part could be integrated to completion. Realizing the limiting transition
M → ∞, we do not get rid of the auxiliary mass M . To put this another way, we cannot
manage to take off the regularization directly. However, the regularization enables us to
perform splitting a matrix element into a finite part and a part containing the ultraviolet
(UV) divergencies in the relativistically invariant form; in so doing the mass M enters into
a divergent part only.
Dimensional regularization. Since the UV divergencies appear under integrating over
virtual momenta in the four-dimensional space-time, then, lowering the dimensionality of
the space-time one could make these integrals finite. This idea is at heart of the basis of
the dimensional regularization [46]. In this case one may define the Feynman integrals as
analytical functions of a parameter n, a space-time dimensionality, and the UV divergencies
will be exhibited as singularities at n → 4. As in the previous case, the finite part of
the integral can be obtained by subtracting some first terms of the expansion into the
Taylor series. The technical advantage of the dimensional regularization lies in the fact
that it maintains the symmetry properties and the corresponding invariance of irregularized
expressions.
We shall elucidate the details of this method by an example of calculating the integral
(21.38). To define the integral over n-dimensional space we shall believe that the internal
momentum has n-components—kµ = (k0 , k1 , ..., kn−1 )—while the external one has only four
nonvanishing components—pµ = (p0 , p1 , p2 , p3 , 0, ...0). Then, in the n-dimensional space,
the integral (21.38) takes the form:
Z
1
I(p2 ; n) = dn k . (21.50)
[(k − p)2 − m2 ](k 2 − m2 )

It is obvious that it is converging when n < 4. Let us try to assign a specific meaning to
this expression under non-integer n. With this in mind we first transform (21.50) using

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 541

the formula (21.4), then carry out the Wick rotation and make the change-over of variables
k → k − xp. As a result we get
Z 1 Z Z 1 Z
2 dn k dn k
I(p ; n) = dx = i dx .
0 [(k − xp)2 − m2 + x(1 − x)p2 ]2 0 [k 2 − m2 + x(1 − x)p2 ]2
Getting over to the n-dimensional spherical coordinate system and fulfilling the integration
over the angle variables, we obtain
Z 1 Z ∞ Z 2π Z π Z π Z π
sinn−2 θn−1 dθn−1
I(p2 ; n) = i dx k n−1 dk dθ1 sin θ2 dθ2 sin2 θ3 dθ3 ... 2 2 2 2
=
0 0 0 0 0 0 [k − m + x(1 − x)p ]
Z 1 Z ∞
2iπ n/2 k n−1 dk
= dx , (21.51)
Γ(n/2) 0 0 [k 2 − m2+ x(1 − x)p2 ]2
where we have used the formula:
Z √

π
m πΓ m+1
2

sin θdθ = .
0 Γ m+2
2

Thus, we have extracted the dependence on n in the explicit view. The integral (21.51) is
the well-defined quantity in the complex plane of n at 0 < Re(n) < 4. The limitation from
the bottom is caused by the existence of the infrared (IF)-divergency that is not a source of
our interest at this time. Completing the integration of the expression (21.51) by parts, we
may extend the analyticity region of the function I(p2 ; n). So, making use of the expression:

nΓ(n) = Γ(n + 1) (21.52)

and accomplishing a single integration by parts, we have

Z ∞
1 k n−1 dk
n

=
Γ 2 0 [k − m + x(1 − x)p2 ]2
2 2

Z ∞  
2 d 1
=−
k n dk . (21.53)
Γ n2 + 1 0 dk [k 2 − m2 + x(1 − x)p2 ]2
The integral (21.53) is the analytical function even at −2 < Re(n) < 4. Iterating the
procedure of the integration by parts d times we extend the analyticity region to −2d <
Re(n) < 4 and, further, up to −∞ < Re(n). This, in its turn, means that the integral
(21.53) represents the analytical function in the region Re(n) < 4.
To clear up the integral behavior at n → 4 we estimate it using the formula (21.45) when
n < 4: Z
2 n/2
 n 1 dx
I(p ; n) = iπ Γ 2 − . (21.54)
2 0 [m − x(1 − x)p2 ]2−n/2
2

It follows from the relation (21.52)

 n  Γ 3 − n2 2
Γ 2− = n → when n → 4, (21.55)
2 2− 2 4−n

that is, the singularity at n = 4 is a simple pole. If we shall expand all the quantities into
the series in the vicinity of the point n = 4:
 n 2
Γ 2− = + A + (n − 4)B + ..., (21.56)
2 4−n

© 2011 by Taylor and Francis Group, LLC

 542 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

[m2 − x(1 − x)p2 ]n−4 = 1 + (n − 4) ln[m2 − x(1 − x)p2 ] + ..., (21.57)

where A and B are some constants, then we get in the limit

Z 1
2 2iπ 2 2
lim I(p ; n) = I(p ) = − iπ 2 dx ln[m2 − x(1 − x)p2 ] + iπ 2 A. (21.58)
n→4 4−n 0

To single out the finite part in (21.58) it is sufficient to subtract the first term of the
expansion of the function I(p2 ) into the Taylor series in the vicinity of the point p2 = 0:

I(p2 ) = I(0) + IC (p2 ), (21.59)

where
 
2 2i 2 2iπ 2
I(0) = π − i ln m + iA ≈ , (21.60)
4−n (4 − n)

Z 1   Z 1
m2 − x(1 − x)p2 (1 − x)(2x − 1)p2
IC (p2 ) = −iπ 2 dx ln 2
= −iπ 2
dx 2 , (21.61)
0 m 0 m − x(1 − x)p2

and in (21.61) we have done the integration by parts. Thus, we have obtained the finite part
of the integral (21.38), which has precisely the same form as in the case of the covariant
regularization. Therefore, as could be expected, the finite part of the Green functions
product does not depend on the regularization way. It depends on the subtraction point
only. The divergency of the term I(0) is related to the simple pole at n = 4, which is
associated with the term ln(M 2 /m2 ) appearing at the covariant regularization.
Successive calculation of the matrix elements containing the ultraviolet (UV) divergen-
cies consists of three stages. First, we transit to the regularized Green functions Re Gc .
At the second stage, the parts GcC that are left finite under removing the regularization
(M → ∞) are extracted from the obtained expressions for Re Gc . Finally, the divergencies
associated with the divergent parts of Re Gc are eliminated from the theory. The quantity
GcC is obtained by subtracting several terms of the Taylor series from Re Gc and does not
depend on the regularization way. This allows us to technically accept the most convenient
strategy that provides the finite result for every diagram not demanding any regularization
under an intermediate stage. This strategy is based on the regularization by means of the
cut-off. To calculate divergent integrals is carried out using the Feynman technique where
the integration sphere radius is identified with the boundary momentum. Further, to split
divergent integrals into the finite and infinite parts, following Dyson [4], we fulfill the ex-
pansion of integrand R(p, t) (p is an external momentum and t is an internal one) into the
Taylor series in external momenta:
   
∂R(p, t) 1 ∂ 2 R(p, t)
R(p, t) = R(0, t) + pµ + pµ pν + ... (21.62)
∂pµ p=0 2 ∂pµ ∂pν p=0

Every differentiation increases the difference of the power t in the denominator and nu-
merator by one. Because of this, if the integral is diverging logarithmically (linearly or
quadratically), then only one term (two or first three terms) will be diverging under term-
by-term integration of the expansion R(p, t). As will be shown later, to liquidate divergent
expressions, the counter-terms, whose physical meaning reside in the redefinition of particle
parameters, are introduced into the theory.

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 543

21.3 Self-energy of the electron

In the interaction representation the time evolution of the state vector

Φ(t0 ) = Φ(p; s) = (2π)3/2 a†s (p)|0 >

to describe a free electron at the instant of time t0 (at the remote past) is determined by
the equation:
Φ(t) = S(t, t0 )Φ(t0 ), (21.63)
where
∂S(t, t0 )
i = Hint S(t, t0 ).
∂t
The energy of the state Φ(p; s) is equal to E0 , where E0 is the eigenvalue of the Hamiltonian
operator for a free particle. However, the energy of an electron state is changed as the result
of the interaction of the electron with the quantized electron-positron and electromagnetic
fields. This, in turn, leads to modifying a phase in the amplitude of the transition from the
state at the instant of time t0 to the state with the time t:
(Φ(p; s), S(t, t0 )Φ(p; s))
(Φ(p; s), Φ(p; s))e−i∆E(t−t0 ) = . (21.64)
(Φ0 , S(t, t0 )Φ0 )
In Eq. (21.64) the time difference t − t0 must be taken very big and the average over all
oscillations from t to t0 must be fulfilled.
Let us explain the presence of the quantity (Φ0 , S(t, t0 )Φ0 ) in Eq. (21.64). Among
the addends entering into (Φ(p; s), S(t, t0 )Φ(p; s)), there are such ones that are associated
with the diagrams for a transition: vacuum→vacuum (Fig. 21.4). We shall show that

Γ1 Γ2 Γ3 Γ4

FIGURE 21.4
The Feynman diagrams describing the transition vacuum→vacuum.

the total contribution from such vacuum processes is given by just the quantity: AV =
(Φ0 , S(t, t0 )Φ0 ). Denote the elementary connected vacuum diagrams (diagrams that may
be bypassed moving only along internal lines) by Γ1 , Γ2 , Γ3 ,... Then, in the most general
case, the diagram of the vacuum fluctuations is derived from the n1 ones of the kind Γ1 , from
the n2 ones of the kind Γ2 , and so on. The contribution of the diagram Γ = n1 Γ1 +n2 Γ2 +...
into the matrix element will be equal to:

(Φ0 , SΓ Φ0 ) = (Φ0 , SΓ1 Φ0 )n1 (Φ0 , SΓ2 Φ0 )n2 ..., (21.65)

© 2011 by Taylor and Francis Group, LLC

 544 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where (Φ0 , SΓ1 Φ0 ) is a matrix element corresponding to the diagrams Γ1 . The fact, that the
contributions of the diagrams Γ1 , Γ2 ... are simply multiplied, follows from the disconnect-
edness of the diagrams under consideration. If the diagram contained some identical parts,
then family terms would not all be different, because the indices rearrangement of their
vertices resulted in the same diagrams. It immediately follows that, in the most general
case, the amplitude of the transition vacuum→vacuum is given by the expression:
X 1
AV = (Φ0 , SΓ1 Φ0 )n1 (Φ0 , SΓ2 Φ0 )n2 ... =
n1 ,n2 ,...
n 1 !n 2 !...

= e(Φ0 ,SΓ1 Φ0 ) e(Φ0 ,UΓ2 Φ0 ) ... = e<0|SΓ1 +SΓ2 +...|0> = e<0|S|0>c = e−N , (21.66)
where < 0|S|0 >c denotes the contribution of all the connected diagrams. Inasmuch as
the quantity N is purely imaginary, then the probability that the vacuum will remain the
vacuum is equal to 1.
Investigate some real process described by the plurality of diagrams that are associated
with an amplitude A. In higher orders of the perturbation theory the same diagrams will
appear in accompaniment of the vacuum diagrams disconnected with them. To sum over
all the vacuum processes gives rise simply to the multiplication of the amplitude A by the
phase factor exp(−N ) whose modulus equals 1.
Thus, the appearance cause of the factor (Φ0 , S(t, t0 )Φ0 ) into the denominator of (21.64)
is correlated to the inclusion of diagrams corresponding to the transition vacuum→vacuum.
This factor is canceled with the analogous one, which is held in the numerator. Next, we
decide not to take into account the diagrams that describe the vacuum fluctuations and,
hence, the quantities of the kind (Φ0 , S(t, t0 )Φ0 ) will not appear in our consideration.
In both sides of the expression (21.64) we carry out the expansion into the series restricted
to two nonvanishing terms. As a result we arrive at the equation:
Z Z
−i∆E T = (−ie) < Φ(p; s)| d x1 d4 x2 T (: ψ(x1 )γ µ Aµ (x1 )ψ(x1 ) : ×
2 4

m
× : ψ(x2 )γ ν Aν (x2 )ψ(x2 ) :)|Φ(p; s) >= T us (p)Σ(2) (p)us (p), (21.67)
E0
or
im
∆E = us (p)Σ(2) (p)us (p), (21.68)
E0
where m denotes the mass of the bare electron now, and
Z
e2 p̂ − k̂ + m 1
Σ(2) (p) = − d4 kγµ γµ 2 . (21.69)
(2π)4 2
(p − k) − m 2 k

Under writing Eq. (21.69) we have allowed for that δ(0) = V T /(2π)4 and V = 1. The matrix
element standing in the right-hand side of the expression (21.67) is in line with the diagram
Fig. 21.5(a) to describe the emission and the subsequent absorption of the virtual photon
by the free electron. The diagram of Fig. 21.5(a) determines a self-energy of the electron in
the second order of the perturbation theory. It represents the simplest case of the electron
self-energy diagram (ESED), which is considered to mean a block confined between two
electron lines. The more complicated ESEDs of the forth order of the perturbation theory
are pictured in Figs. 21.5(a)–(e).
The only measurements possible for a free particle are associated with the determination
of its energy E0 and momentum p. Proceeding from these quantities, we can form a single
invariant combination, the particle mass, m2 = E02 − p2 . Thus, the effect of a relativistically

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 545

(a) (b)

(c) (d)

(e)

FIGURE 21.5
The Feynman diagrams (a)–(e) for the electron self-energy.

invariant interaction of a particle with the vacuum is limited to a change in its mass by
some amount δm. In this case the change-over of a bare particle energy is defined by the
expression: p
E0 + ∆E = (m + δm)2 + p2 . (21.70)
Expanding the expression (21.70) into the series in δm and being constrained by the first
order of smallness, we get
mδm
∆E = . (21.71)
E0
Now it becomes obvious that the operator Σ(2) (p) has a meaning of the mass operator of
the electron in the second order of the perturbation theory.
The effect of the self-energy for a free particle may result only in changing its mass. This
principle forming the basis for a renormalization theory was first used by Kramers [47] in
classical electrodynamics. If one recalls the phrase of M. Twain: “clothes adorn a person
and naked people have little if any influence in the society,” then Kramers’idea may be
regarded simply as an asymptotic extension of the social observations to the phenomena
of the microworld. In his work Kramers also ignores any effect of the bare mass by the
argument that it could not be measured experimentally and the only observable mass is
mR = m + δm = Zm m. In this way, from a theoretical viewpoint, the main part is played
by the observable mR rather than by the quantities m or δm taken separately. As this takes
place, a change in the mass δm may prove to be infinite. But this is of no importance as in

© 2011 by Taylor and Francis Group, LLC

 546 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the physical observables, for example, the energy levels or the cross sections expressed in
terms of the observable mass mR are finite.
The operation of the mass redefinition is called the mass renormalization. It may explic-
itly be included into the theory by means of introducing a counter-term into the Lagrangian.
The Lagrangian of a free field contains a term:

−m : ψ(x)ψ(x) :,

which could be presented as follows:

−(mR − δm) : ψ(x)ψ(x) : .

Then the term

δm : ψ(x)ψ(x) :
may be considered as a part of the interaction Lagrangian along with

−e : ψ(x)γ µ Aµ (x)ψ(x) : .

In so doing, the contributions coming from δm : ψ(x)ψ(x) : must always be examined with
the ones coming from the diagrams of the electron self-energy. This statement is illustrated
in Fig. 21.6, where the circles at the diagrams are associated with the contribution from
the term
δm : ψ(x)ψ(x) : .
The infinite quantity δm is so defined that in all orders of the perturbation theory the

−δm

−δm −δm
+ + + + + + ...

−δm

FIGURE 21.6
The Feynman diagrams illustrating the mass renormalization.

contributions of the self-energy diagrams for a free electron are mutually canceled with the
ones coming from the diagrams corresponding to the counter-term −δm : ψ(x)ψ(x) :. The

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 547

amplitude that the electron being in the state |Ep , p; s > will be remaining in the same
state is proportional to the quantity:
us (p)[1 + Σ′ (p)]us (p),
where Σ′ (p) denotes the contribution of all the self-energy diagrams. The demand of the
mass renormalization implies that at p̂ = mR the quantity Σ′ (p) must be equal to zero,
because we include the contribution, coming from the term δm, in Σ′ (p).
The diagrams in Fig. 21.6 may also be considered as a modification of the internal
electron line. Then the original propagator S c (p) = (p̂ − m)−1 is transmuted into
1 1 1
S c′ (p) = + Σ′ (p) . (21.72)
p̂ − m p̂ − m p̂ − m
Now the mass renormalization principle can be formulated as the following demand: “the
pole of the modified propagator to be regarded as a function of p̂ must be positioned at the
point associated with the experimental value of the particle mass.” It is obvious that this
demand fixes the value of δm in every order of the perturbation theory.
We now proceed to calculating the mass operator of the electron in the second order of
the perturbation theory [7]. This operator, apart from the UV divergencies, contains the
IF divergencies as well. To liquidate the latter we introduce as usual the photon mass λ
into the electron propagator, that is, carry out the replacement:
1 1
D0c (k) = → 2 . (21.73)
k2 (k − λ2 )
Substituting (21.73) into the expression (21.69), we obtain
Z Z
(2) e2 4 p̂ − k̂ + m µ 1 α p̂ − k̂ − 2m 1
Σ (p) = − d kγµ γ = d4 k .
(2π)4 (p − k)2 − m2 k 2 − λ2 2π 3 (p − k)2 − m2 k 2 − λ2

Taking advantage of the Feynman parametrization, we present Σ(2) (p) in the form:
Z 1 Z
α p̂ − k̂ − 2m
Σ(2) (p) = 3 dx d4 k 2 .
2π 0 {k + [−2(kp) + p2 − m2 + λ2 ]x − λ2 }2
At present we may use the formulae (21.26) and (21.30) to give:
Z 
iα 1 L2
Σ(2) (p) = dx [p̂(1 − x) − 2m] ln 2 +
2π 0 p x(x − 1) + m2 x + λ2 (1 − x)
  
3
+p̂ x − 1 + 2m . (21.74)
2
(2)
To extract the finite part ΣC (p) two first terms of the expansion of Σ(2) (p) into the
Taylor series in powers (p̂ − mR ) should be subtracted from the expression for Σ(2) (p):
 (2) 
(2) (2) (2) ∂Σ (p)
ΣC (p) = Σ (p) − Σ (mR ) − (p̂ − mR ) . (21.75)
∂ p̂ p̂=mR

In accordance with (21.68) and (21.71) the quantity Σ(2) (mR ) coincides with the electro-
magnetic mass of the electron δm(2) within the constant factor i:
Z 1   
α L2 3
δm(2) = iΣ(2) (mR ) = dx mR (1 + x) ln 2 2 − m R x + 1 .
2π 0 mR x − λ2 (x − 1) 2
(21.76)

© 2011 by Taylor and Francis Group, LLC

 548 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Having designated δm as the electromagnetic mass, we emphasized its origin. It is apparent

that the analogous corrections to the bare electron mass will come from the weak and
gravitational interactions as well.
Having fulfilled the integration and neglecting the quantity λ as compared with mR , we
arrive at the following expression for the electromagnetic mass of the electron:
 
(2) 3α L2 1
δm = mR − m = mR ln 2 + . (21.77)
4π mR 2

Thus, in the second order of the perturbation theory the physical meaning of the electron
mass is given as:
mR = Zm m, (21.78)
where the singular factor Zm :
 
3α L2 1
Zm = 1 + mR ln 2 +
4π mR 2

is logarithmically diverging under removing the regularization (L2 → ∞).

Differentiating (21.74), we gain
  Z 1 
∂Σ(2) (p) (2) iα L2
= Σ1 = dx (1 − x) ln +
∂ p̂ p̂=mR 2π 0 m2R x2 − λ2 (x − 1)
  
3 2x(1 − x2 )m2R iα 1 L2 9 λ2
+ x−1− 2 2 = ln + + ln 2 . (21.79)
2 mR x − λ2 (x − 1) 2π 2 m2R 4 mR
(2)
To insert Σ(2) (mR ) and Σ1 into (21.76) leads to the following expression for the finite part
of the mass operator in the second order of the perturbation theory:
Z 
(2) iα 1 m2R x2 − λ2 (x − 1)
ΣC (p) = dx [p̂(1 − x) − 2mR ] ln 2 +
2π 0 p x(x − 1) + m2R x − λ2 (x − 1)

2m2R x(1 − x2 )
+(p̂ − mR ) . (21.80)
m2R x2 − λ2 (x − 1)
In the most general case, the diagram for the electron self-energy may be a part of a more
complex diagram and hence the outgoing electron lines (one or both) are associated with
the electron propagation functions. Then, to allow for the counter-term proportional to δm
in the Lagrangian will lead to subtraction of the divergences due to the presence of Σ(p) in
matrix elements in every order of the perturbation theory. This means that, for example,
during calculations of any process, the diagram of which includes the second-order diagram
for the electron self-energy, we can directly use the finite expression for the mass operator
(21.80), disregarding the counter-term ∼ δm.
Integration of (21.80) may be carried out in the analytical form under the assumption:

m2R − p2 λ
ρ= ≫ and ρ > 0. (21.81)
m2R mR

The final result is as follows:

  
(2) iα 1 2 − 3ρ
ΣC (p) = (p̂ − m)2 1− ln ρ −
2πmR 2(1 − ρ) 1−ρ

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 549
   
p̂ + m 1 ρ2 + 4ρ − 4 λ
− 2−ρ+ ln ρ + 1 + 2 ln . (21.82)
ρmR 2(1 − ρ) 1−ρ mR
At the analytical continuation of the expression (21.82) into the region ρ < 0 the logarithm
phase is defined by the replacement mR → mR − iε in accordance with the poles detour
rule. As this takes place, ρ → ρ − iε, so that, when ρ < 0, ln ρ should be understood as:

ln ρ = ln |ρ| − iπ.
(2)
The expression for ΣC (p) is greatly simplified in the region of big momenta (p2 ≫ m2R ):

(2) αp̂ p2
ΣC (p) = ln 2 . (21.83)
4π mR

21.4 Vacuum polarization

We shall examine the electron scattering by the external electromagnetic field taking into
account the radiative corrections (RC). Introduce the counter-term to be connected with
the mass renormalization into the interaction Lagrangian Lint . Then Lint is determined by
the expression:

Lint = −e{N [ψ(x)γ µ ψ(x)(Aµ (x) + Aeµ (x))]} + δmN [ψ(x)ψ(x)]. (21.84)

In Fig. 21.7 we represent the corresponding Feynman diagrams up to the third order of the
perturbation theory inclusive. Here we shall centre on the diagram shown in Fig. 21.7f1.
Let us agree to identify the block confined between two photon internal or external lines
as a photon self-energy diagram (PSED). Obviously, in the case under consideration we
deal with the PSED of the second order of the perturbation theory. In the diagram of Fig.
21.7f1 the external photon line, the electron-positron loop and the internal photon line may
be changed by a single effective photon line with the momentum q:

aeµ (q) = Aeµ (q) + Dµν

c
(q)Π(2)νσ (q)Aeσ (q) = Aeµ (q) + δAeµ (q), (21.85)

where Z  
(−ie)2 1 1
Π(2)
µν (q) = 4
d kSp γµ γν , (21.86)
(2π)4 Σ q̂ + k̂ − m k̂ − m
q = p2 − p1 . Note that the minus sign connected with the existence of the fermion loop has
(2) (2)
been included into the definition of Πµν (q). The operator Πµν (q) is called the polarization
operator of the second order. From (21.85) it follows that thanks to the gauge invariance
the relation:
q ν Π(2)
νσ (q) = 0 (21.87)
must be fulfilled (beyond doubt this relation still stands in any order of the perturbation
(2)
theory). Apparently, the tensor Πνσ (q) can be represented in the form:

Π(2) 2 2 2
νσ (q) = gνσ q Π1 (q ) + qν qσ Π2 (q ), (21.88)

where owing to the gauge invariance Π1,2 (q 2 ) are functions of q 2 only. For the relation
(21.87) to be met the polarization operator must be transverse with respect to q, that is, it
must have the form:
Πνσ (q) = (qν qσ − gνσ q 2 )Π(q 2 ), (21.89)

© 2011 by Taylor and Francis Group, LLC

 550 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

q q1 q
q2
p1 + q1 p1 − k p2 − k
p2
p1 p1 p2 p1 p2
k
a b c

k
q q
−δm

p2 p2 − k p2 p2 p2
p1 p1

d1 d2

k
q q
−δm

p1 p1 − k p1 p2 p1 p2
p1
e1 e2

q q

p+q p −δe

p1 p2 p2
p1

f1 f2

FIGURE 21.7
The Feynman diagrams describing the RC for the electron scattering by the external elec-
tromagnetic field.

where −Π2 (q 2 ) = Π2 (q 2 ) = Π(q 2 ) and in order to emphasize the validity of the expansion
(21.89) for the polarization tensor of any order we have omitted the superscript 2 in Πµν (q).
In accordance with (21.85) the quantity aeµ (q) is the sum of the given external electromag-
netic field Aeµ (q) and an additive δAeµ (q), which may be interpreted as an electromagnetic
field caused by a polarization of the electron-positron field vacuum under switching on the
field Aeµ (q). The field potential Aµ (x) is associated with the current Jµ (x), which creates
it by means of the Maxwell equation:
Aµ (x) = Jµ (x).

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 551

In the Fourier components this linkage looks so:

q 2 Aµ (q) = −Jµ (q). (21.90)

Using (21.85) and (21.90), we could write

jµe (q) = Jµe (q) + δJµe (q) = [1 + iΠ(2) (q 2 )]Jµe (q). (21.91)

From the obtained expression it is evident that the closed electron-positron loop in this
diagram induces the current jµe (q) being the source of the photons with the momentum q,
which influence the electron and scatter it.
We shall calculate the polarization operator in the second order of the perturbation
theory. Since in the expression (21.86) the numerator contains the fourth power of the
integration variable while the denominator contains the second one, then the integral is
(2)
diverging quadratically. However, in reality Πνσ (q) possesses only the logarithmic diver-
gency. It results from the gauge invariance. To make sure of it, first one should regularize
the expression (21.86) on the electron mass and then use the transversality condition (21.87)
[46,50]. On the other hand, the same result could be got by a more simple way too. Making
use of the Feynman parametrization and calculating the spur of the γ-matrices product, we
present Πµν (q) in the following form:
Z
α Sp[γµ (k̂ + q̂ + m)γν (k̂ + m)]
Πµν (q) = − d4 k =
4π 3 [(k + q)2 − m2 ][k 2 − m2 ]
Z 1 Z
α Sp[γµ (k̂ + q̂ + m)γν (k̂ + m)]
=− dx d4 k =
4π 3 0 [(k + qx)2 + q 2 (x − x2 ) − m2 ]2
Z 1 Z
α Sp[γµ (k̂ + q̂(1 − x) + m)γν (k̂ − q̂x + m)]
=− 3 dx d4 k =
4π 0 [k 2 + q 2 (x − x2 ) − m2 ]2
Z 1 Z
α (gµν q 2 − 2qµ qν )(x − x2 ) + gµν (m2 − k 2 /2)
=− 3 dx d4 k . (21.92)
π 0 [k 2 + q 2 (x − x2 ) − m2 ]2
Because, as already stated, the integral (21.92) is diverging only logarithmically, then the
variable change-over k → k − qx being made in (21.92) is quite legal. Acting upon the last
line of (21.92) by the operator qµ , we come to recognize that the polarization tensor will
obey the transversality demand under the fulfillment of the condition:
Z
−q 2 (x − x2 ) + m2 − k 2 /2
d4 k 2 = 0,
[k + q 2 (x − x2 ) − m2 ]2

that is, the quadratically divergent part of the integral (21.92) proves to be equal to zero.
With allowance made for the preceding we obtain for the polarization tensor

Πµν (q) = (qµ qν − gµν q 2 )Π(q 2 ), (21.93)

where Z Z
1
2α d4E k
Π(q 2 ) = dx(x − x2 ) . (21.94)
π3 0 [−k 2 + q 2 (x− x2 ) − m2 ]2
Carrying out the integration into (21.94), we arrive at the final expression for the polariza-
tion tensor in the second order of the perturbation theory [7,45]:
 
(2) 2 iα 2 1 L2 4m2 + 2q 2 1
Πµν (q ) = (qµ qν − gµν q ) ln + (1 − ϑ cot ϑ) − , (21.95)
π 3 m2 3q 2 9

© 2011 by Taylor and Francis Group, LLC

 552 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where q 2 = 4m2 sin2 ϑ.

Inasmuch as the polarization tensor is diverging logarithmically, then in the expansion
into the Taylor series in powers of q 2 :

Π(2) (q 2 ) = Π(2) (0) + Π(2)′ (0)q 2 + ..., (21.96)

where 
(2)′ ∂Π(2) (q 2 ) 

Π (0) =  ,
∂q 2 
q2 =0

all terms starting with the second one are finite. Really, in subsequent terms of the expansion
in powers of q 2 the exponent of the integration variable k is decreasing as much as the one
of the external variable q is increasing. In the expansion of the function Π(2) (q 2 ) the term
not depending on q 2 contains the integral over k, whose integrand behaves as dk/k. On
this reason Π(2) (0) is diverging logarithmically. R The next term of the expansion has the
power q 2 and contains the integral of the manner dk/k 3 , which is already converging under
big values of k. Therefore, Π(2)′ (0) and all the higher terms of the expansion (21.96) are
reported by the expressions convergent on the variable k. Thus, the regularized operator
Π(2) (q 2 ) is determined by the expression:
 
(2) (2) iα 4m2R + 2q 2 1
ΠC (q 2 ) = Π(2) (q 2 ) − ΠD (q 2 ) = (1 − ϑ cot ϑ) − , (21.97)
π 3q 2 9
(2)
where ΠD (q 2 ) = Π(2) (0).
(2)
To define the behavior of ΠC (q 2 ) in the limit of small momenta we are sufficiently
constrained by the second term in the expansion (21.96). To calculate gives:
 Z Z
∂Π(2) (q 2 ) 
 4α 1 2 2 d4E k iα
 =− 3 dx(x − x ) = ,
∂q 2 2 π 0 [−k − m2R ]3
2 15πm2R
q =0

that is,
(2) iα q 2
ΠC (q 2 ) = , when q 2 ≪ m2R . (21.98)
15π m2R
The value q 2 = 4m2R is the threshold for the production of the single electron-positron
(2)
pair by the virtual photon. At q 2 > 4m2R the expression for ΠC (q 2 ) becomes complex and
(e)
the matrix element M has both the real and imaginary parts. The imaginary part is
associated with the time decrease of the scattering amplitude without the real pair produc-
tion. On the other hand, the amplitude of the transition probability into the final state,
which includes specified real pairs, will accordingly be increasing in time. Due to the S-
matrix unitarity the increase and accompanied decrease of these matrix elements are such
that the total probability of the transition from the initial state equals to 1. To define the
analytical continuation (21.96) in the region q 2 > 4m2R one should recall the poles detour
rule (mR → mR − iε) to lead to the following replacements:
π
[1 − q 2 /(4m2R )]1/2 → −i[q 2 /(4m2R ) − 1]1/2 , ϑ→ + iu, cot ϑ → −i tanh u,
2
where sinh u = [q 2 /(4m2R ) − 1]1/2 . Then, one may show that
 
(2) 2 iα q2
ΠC (q ) = ln 2 − iπ , when q 2 ≫ m2R . (21.99)
3π mR

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 553

Analogously, for the case of big space-like values of q 2 we have

 
(2) 2 iα q2
ΠC (q ) = − ln − 2 , when q 2 → −∞. (21.100)
3π mR

Again we shall focus our attention on the process of the electron scattering by the Coulomb
nucleus field, assuming, in this case, the mass renormalization to be fulfilled. The matrix
element corresponding to the diagram shown in Fig. 21.7f1 is given by the expression:
s
(e) 3/2 m2R
M = −(2π) ie us (p2 )M (e) us (p1 ),
Ep1 Ep2

where
1 µν
M (e) = −iγµ Π (q)Aeν (q).
q2
Let us analyze the sum of the matrix elements M(a) + M(e) = A, restricted by two limiting
cases: (i) q 2 → 0; (ii) q 2 → −∞. Then, in the former case this sum is defined by the
expression:
s  
3/2 m2R µ e2 L2
A = −(2π) ie us (p2 )γ 1 − ln Aeµ (q)us (p1 ), (21.101)
Ep1 Ep2 12π 2 m2R

while in the latter one (L2 ≫ −q 2 ) we have:

s  
3/2 m2R e2 L2
A = −(2π) ie us (p2 )γ µ 1 − ln Aeµ (q)us (p1 ). (21.102)
Ep1 Ep2 12π 2 q2

In these expressions only the big logarithmic contributions are kept. It is clear that in terms
of the initial parameters the perturbation theory does not work, since the big corrections
(∼ e2 ln L2 ) already appear in the approximation that follows after the Born one. Let us
show that using the naive renormalization one may improve the situation. For the sake of
simplicity, for the time being, we are distracted from the existence of other divergencies
appearing in the QED. Further, in the formula (21.101) we redefine the charge and the
external field potential:
−1/2 −1/2
eR = Z0 e, AeRµ = Z0 Aeµ , (21.103)
where the singular factor Z0 in the second order of the perturbation theory is given by the
expression:
e2 L2
Z0−1 = 1 + iΠ(2) (0) = 1 − ln .
12π 2 m2R
Then, the amplitude Aq2 →0 expressed in terms of the variables eR and AeRµ takes the
same form as the Born one f (e, Ae ) in which the following replacements must be done:
e → eR , Aeµ → AeRµ :
s
e 3/2 m2R
Aq2 →0 = fB (eR , AR ) = −(2π) ieR us (p2 )γ µ AeRµ (q)us (p1 ).
Ep1 Ep2

Thus, we again include the infinity into the observable quantities. True, in the successive
approach resulting in the renormalization of the charge, the Green functions, the wave-
functions of the electron-positron and electromagnetic fields the infinite constants in the
relations (21.103) will take an other form.

© 2011 by Taylor and Francis Group, LLC

 554 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

From the expression (21.91) follows that the divergent part of the induced current jµe (q)
is exactly proportional to the inducing external current Jµe (q). Because of this, in the given
case the renormalization should be considered as the operation under which both the four-
dimensional density of the external current Jµe (q) (or the potential) and the charge are
renormalized. It is impossible to experimentally separate the external current Jµe (q) from
the induced current, which is proportional to it. Just as the electron is always surrounded
by the fur coat of virtual quanta (and as a result the physical mass is always measured),
the external current polarizes vacuum at all times. Therefore, the observable values of the
external current and the charge will always be determined by the expressions (21.103) when
we carry out their measurements at big distances.
So, if one uses from the outset the quantities eR and AeR as the expansion parameters,
then the diagram pictured in Fig. 21.7f1 may be excluded from consideration at all. To put
it otherwise, one must “by hand” distract its contribution at the point q 2 = 0. One could
perform the same operation in the next orders of the perturbation theory as well. Then,
after its fulfillment, the scattering amplitude taken at the point q 2 = 0 will coincide with
the Born amplitude fB (eR , AeR ) now in all orders of the perturbation theory. The exact
amplitude A proves as though it were normalized on the Born one at the point q 2 = 0. By
this reason, the value q 2 = 0 in the procedure in question is treated as the subtraction point
or the normalization point. It should be emphasized that to introduce the renormalized
quantities in accordance with the formula (21.103) makes the scattering amplitude finite at
any values of q 2 . Since the corresponding matrix element is logarithmically diverging, then
for the matrix element to be made finite, one distraction is sufficient at an arbitrary point.
In particular, after the substitution (21.103) the amplitude (21.102) is represented with an
accuracy of terms ∼ e4R in the form:
 
e2 m2
Aq2 ≫m2R = fB (eR , AeR ) 1 − R 2 ln 2R .
12π q

It is obvious that we may introduce the counter-term connected with the charge renormal-
ization into the starting Lagrangian just as we do in the case of the mass renormalization.
The vacuum polarization produces changing the Maxwell equations for a free field, that
is, changing the components Aµ (x). Hence, to take into consideration the vacuum polar-
ization effect by the gauge invariant way, in the total Lagrangian one must carry out the
replacement: √
Fµν (x) → Fµν (x) 1 + δe.
Here δe is sorted out in such a way that the polarization operator divergencies are compen-
sated in every order of the perturbation theory. Thus, for example, for this purpose to be
achieved in the second order of the perturbation theory, the fulfillment of the demand:
1 (2)
δe = iΠD (q 2 )
2
is needed. Accordingly, the interaction Lagrangian is defined as follows:

Lint = −eR N [ψ(x)γ µ ψ(x)Aµ (x)] + δmN [ψ(x)ψ(x)]+

1
− δeFµν (x)F µν (x)]. (21.104)
4
Then during calculations for each diagram including the irreducible self-energy photon part
we add a diagram, where the ordinary self-energy part of the photon is replaced by the
part given in Fig. 21.8. With the summation of these diagrams the infinities due to
the vacuum polarization are canceled giving as a result a matrix element, whose factors

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 555

δe

k k

FIGURE 21.8
The Feynman diagram connecting with inclusion of the charge renormalization counter-
term.

represent only the finite part of the polarization tensor, that is, the quantities of the kind
(2)
ΠCµν (q 2 ). From the qualitative point of view, varying of the mass or the charge is quite
natural for the particles under the effect of interactions. A similar phenomenon has been
encountered in the classical electrodynamics. In the medium a foreign charge creates around
itself a polarization cloud, being partially screened by this cloud. Because of this, for long
distances the effective value of the observed charge turns to be lower than the real one.
Such a particle is moving together with its polarization cloud and this leads to the effective
change in its inertia properties, that is, the mass. As seen, even in the absence of divergences
in the theory, the renormalization is a must. This is due to the fact that in the real world
it is impossible to turn off interactions between particles; we always observe the parameters
of bare particles plus the corrections to them. Nobody can abandon the rule: “observables
must be expressed in terms of the renormalized constants.”
Without doubt, the finite part of the polarization tensor ΠC (q 2 ) leads to the observed
effects. Let us consider, for example, the Möller scattering taking into account the vacuum
polarization effect in higher orders of the perturbation theory. The corresponding diagrams
are displayed in Fig. 21.9. As the series of the perturbation theory are convergent, then

e e0 2

= 1 − + − ...

FIGURE 21.9
The Feynman diagrams illustrating the charge renormalization.

we essentially deal with the decreased geometric progression whose denominator is equal to
iΠC (q 2 ) (q = p1 − p′1 ). Summing up the obtained series, we come to the following formula:

e2R
e2 (q 2 ) = .
1 − iΠC (q 2 )

So, the experimentally defined charge depends on the value q 2 in the experiment. The fine
structure constant:
e2 (q 2 )
α(q 2 ) ≡
4π

© 2011 by Taylor and Francis Group, LLC

 556 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

also proves to be depending on q 2 . By this reason it is called the running coupling constant
of the QED. By virtue of the fact that in our case q 2 < 0, the quantity ΠC (q 2 ) is determined
by the expression (21.100) in the limit of big values of Q2 = −q 2 . Then, the running coupling
constant of the QED will be define as:
3πα(mR )
α(Q2 ) = , (21.105)
3π − α(mR ) ln(Q2 /m2R )
where α(mR ) is the macroscopic value of the fine structure constant (≈ 1/137). Special
attention must be given to the following. First, the expression (21.104) maintains only
observed quantities that are finite without question. Second, the quantity α exhibits the
extremely moderate growth in the whole interval of the values Q2 available.
Using the expression (21.97), we can find the effective potential of the external electro-
magnetic field in the second order of the perturbation theory caused by vacuum polarization:
(2)νσ
[Aeµ (q)]ef f = Aeµ (q) + Dµν
c
(q)ΠC (q 2 )Aeσ (q) =
  
α 4m2R + 2q 2 1
= 1− (1 − ϑ cot ϑ) − Aeµ (q). (21.106)
π 3q 2 9
At q 2 = 0 we get the photon field, δAeµ (q) vanishing in this case. So, a single photon moving
in a space is not inducing the current. It means that the vacuum polarization effects are
absent for the electromagnetic waves free to propagate. However, this statement is valid
only until the effects of the lowest order is being considered. In Fig. 21.10 we present one
of the diagrams describing the scattering of two photons by each other as the result of the
vacuum polarization effects in the forth order of the perturbation theory. In the limit of

k1 p − k3 − k4 k3

p − k1 p − k3

k2 p k4

FIGURE 21.10
The Feynman diagram describing the scattering of two photons by each other.

small values of q 2 the expression (21.106) becomes the form:

 
α
[Aeµ (q)]ef f = 1 − q 2
Aeµ (x)
15πm2R
or  
α
[Aeµ )(x)]ef f = 1+  Aeµ (x). (21.107)
15πm2R

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 557

As follows from (21.107), the vacuum polarization causes smearing of the potential over
the region, the radius of which is of the order of the Compton wavelength of the electron.
The physical pattern of the phenomenon is well known. Due to the interaction with an
electron-positron field, the charge e becomes surrounded by a cloud of virtual electrons and
positrons. Some part of them, possessing the total charge δe with the same sign as e, goes
to infinity. And a part of the cloud with the total charge −δe is concentrated around the
initial charge in the region with the sizes λc = ~/mc. When a charge is observed from a
distance exceeding λc , this charge seems to be effectively equal to the renormalized charge
eR = e − δe. At the same time, were the observations from distances considerably below λc
possible, one could measure the bare charge e.
According to (21.107), in a hydrogen atom the effective potential for the electron, includ-
ing only the vacuum polarization effects, is given by the expression:
 
1 4α (3)
−Ze2 − δ (r) .
r 15m2R

Uehling has been the first [48] to note that due to the vacuum polarization the electrostatic
potential of a point charge in the vacuum is not described by the exact Coulomb law.
Deviation from the Coulomb law has been calculated by Uehling soon after the appearance
of the early Dirac [49] and Heisenberg [50] papers devoted to the vacuum polarization. It
should be recorded that modifying the Coulomb law is caused not only by means of the
PSEDs. The electron self-energy and vertex diagrams give contributions as well. We shall
investigate this question in detail under consideration of the Lamb shift in Section 21.8.
At q 2 = 0 the polarization tensor Πµν (q 2 ) describes the photon self-energy. In the second
order of the perturbation theory the photon self-energy appears as a result of the production
of the virtual electron-positron pair and its subsequent annihilation (Fig. 21.11). The

k k

p−k

FIGURE 21.11
The photon self-energy diagram.

matrix element associated with the diagram shown in Fig. 21.11 is defined by the expression:

M ∼ eλµ (q)Πµν (q 2 )eλν (q) = eλµ (q)(q µ q ν − g µν q 2 )eλν (q)Π(q 2 ) =

= −q 2 gµν eλν (q)eλµ (q)Π(q 2 ) = 0. (21.108)

The vanishing of the photon self-energy is a consequence of the gauge invariance of the
expression for Πµν . This circumstance, in its turn, is connected with the fact that the
photon mass is exactly equal to zero. Thus, owing to the gauge invariance there is no need
to introduce the counter-term for the photon mass renormalization. One may relate the
particle mass value with the pole position of the particle propagator. Let us define the exact
photon propagator by an infinite plurality of diagrams displayed in Fig. 21.12. Then, the

© 2011 by Taylor and Francis Group, LLC

 558 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

+ + +

+ + + ...

FIGURE 21.12
The Feynman diagrams defining the exact photon propagator.

zero-mass value of the photon follows from the statement: “The exact photon propagator
has the pole at q 2 = 0.”
The photon interacts not only with the electron-positron field, it interacts with fields of
other particles as well. An interaction with a vacuum of these fields is described by the same
self-energy diagrams whose internal lines are assigned to corresponding particles. In this
case it appears that contributions coming from different particles are simply summarized
[51]. Since in the second order of the perturbation theory the radiative effects caused by
the vacuum polarization are inversely proportional to a mass square of a particle giving rise
to a loop, then contributions from more light particles dominate. One important point to
remember is that the QED allows us to calculate the contributions coming from the charged
leptons only.

21.5 Vertex function of the third order

In the previous two sections we have considered the diagrams of the electron self-energy and
the vacuum polarization in the second order of the perturbation theory. The divergencies
corresponding to them exhaust all the possible ones of the second order scattering matrix
S (2) . Let us proceed now to the investigation of the elements matrix behavior in S (3) . A
part of divergent terms in S (3) incorporating the second order divergencies, such as the ones
associated with the diagrams of Fig. 21.13, is compensated by the second order counter-
terms simultaneously with removing the divergencies from S (2) . By this reason, specific
divergencies of the third order may be contained only in the terms of S (3) corresponding
to the diagrams pictured in Fig. 21.14. On the strength of the Furry theorem the matrix
elements appropriate to the diagrams presented in Figs. 21.14(a),(b) turn into zero. Thus,
in the third order of the perturbation theory, we shall have to examine only the matrix
element corresponding to the diagram shown in Fig. 21.14(c):

s
m2
M = −ie u(p2 )Λ(3) µ
µ (p1 , p2 )u(p1 )A (q), (21.109)
Ep1 Ep2

(3)
where Λµ (p1 , p2 ) is the third-order vertex function:

Z
ie2 p̂2 − k̂ + m p̂1 − k̂ + m 1
Λ(3)
µ (p1 , p2 ) = − d4 kγν γµ γν 2 , (21.110)
(2π)4 2
(p2 − k) − m 2 2
(p1 − k) − m 2 k − λ2

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 559

(b)

(a) (c)

FIGURE 21.13
The Feynman diagrams (a)–(c) of the third order containing divergences.

(a) (b) (c)

FIGURE 21.14
The vertex diagrams (a)–(c) of the third order.

and we have modified routinely the photon propagator to eliminate the infrared (IR) diver-
gencies. It is convenient to present the expression (21.110) in the form:

ie2
Λ(3)
µ (p1 , p2 ) = − {γν (p̂2 + m)γµ (p̂1 + m)γ ν J1 − [γν (p̂2 + m)γµ γ σ γ ν +
(2π)4

+γν γ σ γµ (p̂1 + m)γ µ ]Jσ + γν γ σ γµ γ τ γ ν Jστ }, (21.111)

where
Z
d4 k(1; σ; στ )
J(1; σ; στ ) = (21.112)
Σ [k 2 − 2(p1 k) + p21 − m2 ][k 2− 2(p2 k) + p22 − m2 ](k 2 − λ2 )

in the parenthesis (1; σ; στ ) one must maintain 1 for the integral J1 , kσ —for the integral Jσ ,
and kσ kτ —for the integral Jστ . Among the three integrals only the third integral contains
the UV-divergency (it is logarithmic). Therefore, under calculations of the integrals Jσ and
Jστ one may set λ equal to zero. Further, we shall be constrained by the investigation
of the case when p1 and p2 comprise the four-dimensional momenta of a free electron
(p21 = p22 = m2R ).

© 2011 by Taylor and Francis Group, LLC

 560 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Using the Feynman parametrization, we represent J1 , Jσ , and Jστ in the following form:
Z 1 Z 1 Z
d4 k
J1 = 2 dy xdx 2 3
, (21.113)
0 0 Σ [k − 2(px k) + lx ]
Z 1 Z 1 Z
d4 k(kσ ; kσ kτ )
J(σ;στ ) = 2 dy xdx , (21.114)
0 0 Σ [k 2 − 2(px k)]3
where
py = yp1 + (1 − y)p2 , px = xpy , lx = (x − 1)λ2 .
Integrating the obtained expressions over k is accomplished by application of the formulae
(21.27)–(21.29). Having done the integration over x, we come to the expressions:
Z
iπ 2 1
dy p2y
J1 = ln 2 , (21.115)
2 0 p2y λ
Z 1
dy
Jσ = iπ 2 p ,
2 yσ
(21.116)
0 py
Z 1    
iπ 2 L2 1 pyσ pyτ
Jστ = dy gστ ln 2 − −2 , (21.117)
4 0 py 2 p2y
where under calculation of J1 we have neglected the term:
Z 1
λ2 dx
.
2p2y 0 x2 p2y − λ2 (x − 1)

Let us express q 2 = (p2 − p1 )2 in terms of the angle ϑ in accordance with the relation:

q 2 = 4m2 sin2 ϑ. (21.118)

Introducing a new variable ξ instead of y

tan ξ
2y − 1 = , (21.119)
tan ϑ
we can rewrite p2y and pyσ as follows:

p2y = [p1 + q(1 − y)]2 = m2 − q 2 y(1 − y) = m2 [1 − 4y(1 − y) sin2 ϑ] =

cos2 ϑ
= m2 cos2 ϑ 1 + tg2 ξ = m2 , (21.120)
cos2 ξ
tan ξ 1
pyσ = (p1σ − p2σ ) + (p1σ + p2σ ). (21.121)
2 tan ϑ 2
Passing on to the new variable and integrating by parts, we get for J1
Z ϑ  "Z #
ϑ
2iπ 2 cos ξ λ 2iπ 2 λ
J1 = − 2 ln + ln dξ = − 2 ξ tan ξdξ + ϑ ln .
m sin 2ϑ 0 cos ϑ m m sin 2ϑ 0 m
(21.122)
The similar mathematics lead to the following values of two remaining integrals:

iπ 2 ϑ
Jσ = (p1σ + p2σ ), (21.123)
m2 sin 2ϑ

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 561
  
iπ 2 L 1 ϑ(p1σ + p2σ )(p1τ + p2τ )
Jστ = gστ ln − − +
2 m 4 2m2 sin 2ϑ
  
qσ qτ
+ gστ − 2 (1 − ϑ cot ϑ) . (21.124)
q
The next step is to simplify the coefficients that accompany the integrals J1 , Jσ , Jστ in
the expression for the vertex function. Since we calculate the vertex operator taken in
plates between the bispinors u(p2 ) and u(p1 ), then the progress in this direction should
be expected under invoking the Dirac equation. Using the relations (19.63), we rewrite the
coefficient standing in front of J1 in the following view:

γν (p̂2 + m)γµ (p̂1 + m)γ ν = −2p̂1 γµ p̂2 − 2m2 γµ + 2m(γµ p̂2 + p̂2 γµ + γµ p̂1 + p̂1 γµ ). (21.125)

Next, taking into consideration the relations:

p̂1 = p̂2 − q̂, p̂2 = p̂1 + q̂,

we arrive at

u(p2 ){−2(p̂2 − q̂)γµ (p̂1 + q̂) − 2m2 γµ + 2m[γµ (p̂1 + q̂) + p̂2 γµ + γµ p̂1 +

+(p̂2 − q̂)γµ ]}u(p1 ) = u(p2 )(4m2 γµ +2q̂γµ q̂)u(p1 ) = u(p2 )(4m2 γµ −2q 2 γµ )u(p1 ), (21.126)
where we have allowed for that
1
u(p2 )q ν q λ γν γµ γλ u(p1 ) = u(p2 )q ν q λ (γν γµ γλ + γλ γµ γν ) u(p1 ) =
2
1
= u(p2 )q ν q λ (4gµλ γν − 2gνλ γµ ) u(p1 ) = −u(p2 )q 2 γµ u(p1 ).
2
The analogous mathematical manipulation allow one to present the factors standing be-
fore the integrals Jσ and Jστ as:

γ ν (p̂2 + m)γµ γσ γν + γ ν γσ γµ (p̂1 + m)γν = 4mgσµ + 2(q̂γµ γσ − γσ γµ q̂), (21.127)

γ ν γσ γµ γτ γν = −2γτ γµ γσ . (21.128)
Gathering together the obtained results, we arrive at the following expression for the
third-order vertex function:
ie2 
Λ(3)
µ (p1 , p2 ) = (4m2 − 2q 2 )γµ J1 + [4im(p1 + p2 )µ + 2(p̂1 + p̂2 )γµ q̂−
(2π)4
 
1
−2q̂γµ (p̂1 + p̂2 ) − (p̂1 + p̂2 )γµ (p̂1 + p̂2 ) K1 − 2q̂γµ q̂K2 − 4γµ K3 , (21.129)
2
where
iπ 2 ϑ iπ 2
K1 = , K 2 = (1 − ϑ cot ϑ),
m2 sin 2ϑ 2q 2
 
iπ 2 L 1
K3 = 1 − ϑ cot ϑ + ln − .
2 m 4
Because we are dealing with the determination of the quantity:

u(p2 )Λ(3)
µ (p1 , p2 )u(p1 ),

© 2011 by Taylor and Francis Group, LLC

 562 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

then we may again use the Dirac equation in (21.129). Then, we gain the following expres-
sion for the vertex operator:
" Z ϑ
(3) e2 γµ 2ϑ  m  2
Λµ (p1 , p2 ) = − ln −1 − ξ tan ξdξ−
(2π)2 tan 2ϑ λ tan 2ϑ 0

ϑ 1 1 L2 e2 2ϑ
− tan ϑ − − ln 2 − 2
(γµ q̂ − q̂γµ ) . (21.130)
2 8 4 m 32π m sin 2ϑ
The derived expression is diverging logarithmically. Therefore, to extract the finite part
(3)
ΛCµ (p1 , p2 ) it is sufficient to subtract only the first term of the expansion into the Taylor
series in the vicinity of the point q = 0 from (21.130). Since
 
(3) (3) e2 γµ 1 L2 9 λ
Λµ (p1 , p1 ) = Λµ (p1 , p2 )|q=0 = ln + − ln , (21.131)
(2π)2 4 m2 8 m
(3)
where p21 = p22 = m2R , then the finite value of Λµ (p1 , p2 ) is given by the expression [7]:
(3)
ΛCµ (p1 , p2 ) = Λ(3) (3)
µ (p1 , p2 ) − Λµ (p1 , p1 ) =
" Z ϑ #
α  m  ϑ
R
=− (2ϑ cot 2ϑ − 1) ln − 1 − 2 cot 2ϑ ξ tan ξdξ − tan ϑ γµ −
π λ 0 2
α 2ϑ
− (γµ q̂ − q̂γµ ) . (21.132)
8πmR sin 2ϑ
Attention is drawn to the fact that the dependence on the fictitious photon mass λ is
(3) (3)
conserved in ΛCµ (q), and, as is easy to see, ΛCµ (q) is diverging logarithmically when λ → 0.
As will be shown in Section 21.7 this divergency is fictitious and it disappears under the
inclusion of the radiative corrections.
The formula (21.132) is true at p21 = p22 = m2R and q 2 ≤ 4m2R . When the values of the
(3)
momentum q are small, the expression for the operator ΛCµ (q) takes the simple form:
   
(3) α q̂γµ − γµ q̂ 2q 2 mR 3
ΛCµ (q) = + ln − γµ . (21.133)
2π 4mR 3m2R λ 8

If q 2 > 4m2R , then the external potential can create real pairs and the denominators of
Eqs. (21.113) and (21.114), containing the auxiliary variables x and y, have the poles. At
the same time, these integrals are well defined because the additive to the mass (iε) gives
the single-valued prescription to calculate them. At p21 6= m2R and p22 6= m2R the vertex
operator was first obtained in Ref. [52]. For the case p21,2 ≫ m2R and |q 2 | ≫ m2R it has the
form:  
(3) α |q 2 | mR 1 2 |q 2 |
ΛCµ (q) = − ln ln + ln γµ . (21.134)
π 4m2R λ 4 4m2R

21.6 Radiative corrections to scattering of an electron by the

Coulomb nucleus field
So, we have studied the basic technical receptions of the calculations in higher orders of the
perturbation theory. Possessing a stored arsenal of knowledge, we can rush to the storm of

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 563

some particular process already not being constrained by the tree approximation. Let us
deduce the cross section of the elastic scattering of the electron by the Coulomb potential
with inclusion of the RC. In so doing, we pose the following goals: (i) to prove the UV-
divergencies cancelation in a given order of the perturbation theory; (ii) to illustrate the
technique of calculating the cross sections under inclusion of the RC; and (iii) to show, that
the final result is free from the IR-divergencies and does not contain the fictitious photon
mass.
The Feynman diagrams describing the electron scattering by the Coulomb nucleus field
up to the third order of the perturbation theory inclusive are shown in Fig. 21.7. We
call attention to the RC absence for external lines. It is caused by the definition of the
regularized expressions for the mass electron operator ΣC (p) and the polarization tensor
Πµν 2
C (q ). Really, from Eqs. (21.75) and (21.97) follows
 
1  1 
 
ΣC (p) = ΣC (p)  = 0,
p̂ − mR  p̂ − mR 
p̂=mR p̂=mR

1 µν 2 
Π (q ) = 0,
q2 C 
q2 =0

ensuring the vanishing contribution from insertions of the self-energy diagrams into the
electron and photon external lines.
The amplitude of the Coulomb scattering is given by the expression:
m h
a) b) c) d ) d ) e1 )
A = −ie(2π)3/2 u(p2 ) M(1) + M(2) + M(3) + M(3)1 + M(3)2 + M(3) +
E
i
e2 ) f1 ) f2 )
+M(3) + M(3) + M(3) u(p1 ), (21.135)

where
a)
M(1) = γ µ Aeµ (p2 − p1 ), (21.136)
Z
b) e k̂ + m ν e
M(2) = − d3 kγ µ Aeµ (p2 − k) γ Aν (p1 − k), (21.137)
(2π)3/2 k 2 − m2
c)
M(3) = Λ(3)µ (p2 , p1 )Aeµ (p2 − p1 ), (21.138)
d ) 1
M(3)1 = iΣ(2) (p2 ) γ µ Aeµ (p2 − p1 ), (21.139)
p̂2 − m
d ) 1
M(3)2 = −δm γ µ Aeµ (p2 − p1 ), (21.140)
p̂2 − m
e1 ) 1
M(3) = iγ µ Aeµ (p2 − p1 ) Σ(2) (p1 ), (21.141)
p̂1 − m
e2 ) 1
M(3) = −δmγ µ Aeµ (p2 − p1 ) , (21.142)
p̂1 − m
f1 ) 1
M(3) = −iγν 2 Π(2)νµ (q 2 )Aeµ (p2 − p1 ), (21.143)
q
f2 ) i 1
M(3) = − δeγν 2 [q 2 g νµ − q µ q ν ]Aeµ (p2 − p1 ), (21.144)
2 q
q = p2 − p1 . We start our analysis with the addends (21.139) and (21.141). Since they
contain the factor (p̂−m)−1 , their denominators turn into zero when taken in plates between

© 2011 by Taylor and Francis Group, LLC

 564 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the bispinors u(p2 ) and u(p1 ). This difficulty is connected with the limiting transition—
the initial and final times are directed to ±∞ in the scattering matrix S(tf , ti ). To gain
the finite result one should address to the adiabatic hypothesis, that is, one introduces the
damped factors to be necessary for the right definition of the initial and final bare states
[4,7]. Then the interaction Lagrangian (21.84) is replaced with the expression:

Lint = −eg(t){N [ψ(x)γ µ ψ(x)(Aµ (x) + Aeµ (x))]} + δm[g(t)]2 N [ψ(x)ψ(x)], (21.145)

where the damped function g(t) is adiabatically switched on and switched off the interaction
between fields and we have also allowed for that δm is caused by the second-order effect. It is
assumed that the time, during which the function g(t) is significantly changed, considerably
exceeds the scattering process duration. When one denotes the Fourier image of the function
g(t) by G(Γ0 ), then
Z ∞ Z ∞
−iΓ0 t
g(t) = dΓ0 G(Γ0 )e = dΓ0 G(Γ0 )e−iΓx , (21.146)
−∞ −∞

where Γµ = (Γ0 , 0, 0, 0). We also choose the normalization:

Z ∞
g(0) = dΓ0 G(Γ0 ) = 1 (21.147)
−∞

and assume that the function G(Γ0 ) behaves in much the same way as the δ-function and
essentially differs from zero only for the values of Γ0 to equal T −1 in the order of magnitude.
When employing the interaction Lagrangian in the form (21.145), the matrix element
(21.141) is substituted for the expression:
Z ∞ Z ∞
e1 ) 1
M(3) = i dΓ0 dΓ′0 γ µ Aeµ (p2 − p1 − Γ − Γ′ ) ×
−∞ −∞ p̂1 + Γ̂ + Γ̂′ − m

×Σ(2) (p1 + Γ)G(Γ0 )G(Γ′0 ). (21.148)

When T → ∞ and Γ0 , Γ′0
→ 0, then the integrand could be transformed. In a reconstruction
process the expansion being valid for arbitrary operators A and B [7]:

1 1 1 1 1 1 1
= − B + B B − .... (21.149)
A+B A A A A A A
proves to be useful. The validity of (21.149) could be proved through the iteration of the
identity:
1 1 1 1 1 1 1 1 1 1
= − B = (A + B) − B = (A + B − B) .
A+B A A A+B A A+B A A+B A A+B

Expanding the operator Σ(2) (p1 + Γ) with the help of the relation (21.149) and keeping the
terms of the first power in Γ only, we get
Z
α 1 1
Σ(2) (p1 + Γ) = − 3 d4 kγ µ γµ = Σ(2) (p1 )+
4π p̂1 + k̂ + Γ̂ − m k 2 − λ2
Z
α 1 1 1
+ 3 d4 kγ µ γν γµ Γν = Σ(2) (p1 ) + iΛ(3)ν (p1 , p1 )Γν .
4π p̂1 + k̂ − m p̂1 + k̂ − m k 2 − λ2
(21.150)

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 565
1 e )
Thus, the matrix element M(3) becomes the form:
Z ∞ Z ∞
e ) 1
1
M(3) =i dΓ0 dΓ′0 γ µ Aeµ (p2 − p1 − Γ − Γ′ ) ×
−∞ −∞ p̂1 + Γ̂ + Γ̂′ − m
h i
× Σ(2) (p1 ) + iΛ(3)ν (p1 , p1 )Γν G(Γ0 )G(Γ′0 ). (21.151)
2 e )
After the similar operations the matrix element M(3) will look like:
Z ∞ Z ∞
e ) 1
2
M(3) = −δm dΓ0 dΓ′0 γ µ Aeµ (p2 − p1 − Γ − Γ′ ) G(Γ0 )G(Γ′0 ).
−∞ −∞ p̂1 + Γ̂ + Γ̂′ −m
(21.152)
e1 ) e2 )
Let us compare the expressions for the matrix elements M(3) and M(3) . Since δm =
(2) c1 ) e2 )
iΣ (p1 = m2R ),
then the addition of and M(3) M(3)
leads to the disappearance of the
infinities associated with the mass renormalization:
Z ∞ Z ∞
e1 ) e2 ) 1
M(3) + M(3) =− dΓ0 dΓ′0 γ µ Aeµ (p2 − p1 − Γ − Γ′ ) ×
−∞ −∞ p̂1 + Γ̂ + Γ̂′ − m

×Λ(3)ν (p1 , p1 )Γν G(Γ0 )G(Γ′0 ). (21.153)

c1 ) c )
Next, in order to symmetrize (21.153) we change Γν with (Γν + Γ′ν )/2. Since M(3) 2
+ M(3)
is being taken in plates between u(p1 ) and u(p2 ), then the operator:
1 ν
(γ p1ν − m) (21.154)
2
may be added to the expression:
1 ν
γ (Γν + Γ′ν )
2
All this allows us to present the expression (21.153) in the following manner:
Z ∞ Z ∞
e1 ) e2 ) 1
M(3) + M(3) = − dΓ0 dΓ′0 γ µ Aeµ (p2 − p1 − Γ − Γ′ ) ×
−∞ −∞ p̂1 + Γ̂ + Γ̂′ − m
1 1
×Λ(3)ν (p1 , p1 ) [Γ̂ + Γ̂′ + p̂1 − m]G(Γ0 )G(Γ′0 ) = − γ µ Aeµ (p2 − p1 )Λ(3)ν (p1 , p1 ), (21.155)
2 2
where in the final stage we have gone to the limit Γ0 , Γ′0 → 0.
Applying the similar mathematical operations to the matrix elements (21.139) and (21.140),
we obtain
d ) d ) 1
M(3)1 + M(3)2 = − γ µ Aeµ (p2 − p1 )Λ(3)ν (p1 , p1 ). (21.156)
2
To summarize (21.155) and (21.156) with (21.138) results in
e ) e ) d ) d ) c)
1
M(3) 2
+ M(3) + M(3)1 + M(3)2 + M(3) = Aeµ (p2 − p1 )[Λ(3)µ (p1 , p2 ) − Λ(3)µ (p1 , p1 )] =

(3)µ
= Aeµ (p2 − p1 )ΛC (p1 , p2 ), (21.157)
that is, the mutual cancelation of the divergencies associated with the self-energy and
vertex diagrams take place. Later we shall show that the similar cancelation occurs for
any processes in all orders of the perturbation theory and is the consequence of the gauge
invariance of the QED.

© 2011 by Taylor and Francis Group, LLC

 566 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
(2)
The addition of Eqs. (21.143), (21.144) and the inclusion of the relation δe = 2iΠD (q 2 )
also bring about the finite result:
f )
1 2 f ) 1 (2)νµ 2 e
M(3) + M(3) = −iγν Π (q )Aµ (p2 − p1 ). (21.158)
q2 C
So, we have demonstrated that although each of the matrix elements corresponding to
the diagrams displayed in Figs. 21.7(c)–(f) taken individually has the UV-divergencies, the
sum of these matrix elements is finite and equals
X j) m (3)µ iγν (2)νµ
M(3) = M(3) = −ie(2π)3/2 u(p2 )[ΛC (p1 , p2 )− 2 ΠC (q 2 )]Aeµ (p2 −p1 )u(p1 ).
E q
j=c,d,e,f
(21.159)
Now, we are coming to the determination of the cross section. According to the general
rules it is defined by the expression:
2π|p|E
dσ = |M(1) + M(2) + M(3) |2 dΩ, (21.160)
v
where |p1 | = |p2 | = |p|,
m
M(1) = −ie(2π)3/2
u(p2 )Aeµ (p2 − p1 )γ µ u(p1 ),
E
m b)
M(2) = −ie(2π)3/2 u(p2 )M(2) u(p1 ).
E
As we have agreed to calculate the cross section including the third order diagrams, then
we can obtain the complete picture considering the terms proportional to e6 but no higher
(thus, for example, with allowance made for the terms proportional to e8 we would be forced
to take into account the fourth-order diagrams). So, in the third order of the perturbation
theory the cross section is presented by the sum:

dσ = dσ(1) + dσ(2) + dσrad , (21.161)

where dσ(1) is the cross section in the first Born approximation, and the quantity:

4π|p|E
dσ(2) = dΩ Re{M∗(1) M(2) }, (21.162)
v
4π|p|E
dσrad = dΩ Re{M∗(1) M(3) } (21.163)
v
are the corrections to dσ(1) . It is obvious that dσ(2) and dσrad may be deduced independently
from each other.
b)
Let us consider the matrix element M(2) determined by the expression (21.137). In the
pure Coulomb field of the resting nucleus we have
Z|e|
Ae0 (r) = .
4πr
The integral in Eq. (21.137) is logarithmically diverging for such a potential. This diver-
gency is specific for the Coulomb field and is connected with the decrease slowness of this
field at large distances. However, we have known the prescription that allows us to get
rid of this difficulty. We must take into consideration the screening, that is, carry out the
replacement:
Z|e| Z|e| −ηr
→ e , (21.164)
4πr 4πr

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 567

where η is a positive constant, and pass on to the limit η → 0 in the end of calculations.
In the case of the screened Coulomb potential the quantity Aeµ (q) has the form:

Z|e|gµ0
Aeµ (q) = . (21.165)
(2π) (q2 +
3/2 η2 )

Making use of this expression, we find

2iZ 2 α2 m
M(2) = u(p2 )[(γ 0 E + m)J1 + (γJ)]u(p1 ), (21.166)
π E
where we have introduced the designations:
Z
d3 k
J1 = , (21.167)
[(p2 − k) + η ][(p1 − k)2 + η 2 ][p21 − k2 + iǫ]
2 2

Z
kd3 k p1 + p2
J= 2 = J2 (21.168)
[(p2 − k)2 + η 2 ][(p1 2 2 2
− k) + η ][p1 − k + iǫ] 2
and allowed for that so far as the integral J is symmetric with respect to p1 and p2 , then
it must be directed along p1 + p2 (the addend iǫ has been included so as to remind the
reader of the pole detour rule in the electron propagator). Eliminating the matrices γ by
application of the Dirac equation

(γp)u(p) = (γ 0 E − m)u(p),

we rewrite (21.166) in the following form:

2iZ 2 α2 m
M(2) = u(p2 )[γ 0 E(J1 + J2 ) + m(J1 − J2 )]u(p1 ). (21.169)
π E
In order to calculate the integrals J1,2 entering into (21.169), we shall employ the Feynman
parametrization. With the help of Eq. (21.4) the integral J1 is presented as:
Z 1 Z 1 Z 1
J1 = −2 dξ1 dξ2 dξ3 ×
0 0 0
Z
d3 kδ(1 − ξ1 − ξ2 − ξ3 )
× =
{[(p2 − k)2 + η 2 ]ξ1 + [(p1 − k)2 + η 2 ]ξ2 + [k2 − p21 − iǫ]ξ3 }3
Z 1 Z 1−ξ1
= −2 dξ1 dξ2 ×
0 0
Z 3
d k
× . (21.170)
[η 2 (ξ1 + ξ2 ) + p21 (2ξ1 + 2ξ2 − 1) − 2k(ξ1 p2 + ξ2 p1 ) + k2 − iǫ]3
Having introduced a new variable f = k − ξ1 p2 − ξ2 p1 , we can take without trouble the
three-dimensional integral: Z
d3 f π2
= i
(f 2 − a2 − iǫ)3 4a3
and arrive at the following value for J1 :
Z 1 Z 1−ξ1
iπ 2
J1 = − dξ1 dξ2 ×
2 0 0

© 2011 by Taylor and Francis Group, LLC

 568 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
1
× . (21.171)
[p21 (ξ12 + ξ22 − 2ξ1 − 2ξ2 + 1) + 2ξ1 ξ2 p1 p2 − η 2 (ξ1 + ξ2 ) − iǫ]3/2
Next, we pass on to new variables:
x = ξ1 + ξ2 , y = ξ1 − ξ2 .
The integration over y from 0 to x is easily fulfilled and gives the result:
Z 1
iπ 2 xdx
J1 = − 3
, (21.172)
2|p| 0 [bx2 − 2x + 1 − η 2 x|p|−2 − iǫ] [(1 − x)2 − η 2 x|p|−2 − iǫ]1/2

where
p21 + p1 p2 ϑ
b= = cos2 .
2p21 2
To work out the integral over x at η → 0 we divide the integration region into two parts:
Z 1 Z 1−η1 Z 1
...dx = ...dx + ...dx,
0 0 1−η1

where η/|p| ≪ η1 ≪ 1. Having set η equal to 0 in the first integral, we get

Z 1−η1  
1 (1 − x)2 1−η1 1 η12
...dx = ln | = ln + iπ . (21.173)
0 2(1 − b) (bx2 − 2x + 1 − iǫ) 0 2(1 − b) 1−b
Note, that the pole detour rule to be specified by the additive -iǫ allowed us to define
changing the logarithm argument under transition from 0 to 1 − η1 . In the second integral
one may set x = 1 everywhere aside from the addend (1 − x)2 and also put η = 0 in the
first square brackets of the denominator. Then, taking into account the pole detour rule
again, we obtain
Z 1 Z η1 " Z
η/|p|
1 dx′ 1 dx′
...dx = − = − i +
1−η1 1 − b 0 (x′2 − η 2 |p|−2 − iǫ)1/2 1−b 0 (η 2 |p|−2 − x′2 )1/2
Z η1 #  
dx′ 1 iπ 2|p|η1
+ ′2 2 −2 )1/2
= − + ln . (21.174)
η/|p| (x − η |p| 1−b 2 η
When summarizing the both integrals, the auxiliary quantity η1 falls out and we get
 
iπ 2 2|p| sin(ϑ/2)
J1 = ln . (21.175)
2|p|3 sin2 (ϑ/2) η
The analogous mathematical manipulations yield the following expression for the integral
J2 :
π 3 [1 − sin(ϑ/2)] iπ 2
J2 = J1 − − ln [sin(ϑ/2)] . (21.176)
4|p|3 cos2 (ϑ/2) sin(ϑ/2) 2|p|3 cos2 (ϑ/2)
It should be pointed out that the real parts of the deduced integrals do not depend on the
parameter η:
π 3 [1 − sin(ϑ/2)]
Re J1 = 0, Re J2 = . (21.177)
4|p|3 cos2 (ϑ/2) sin(ϑ/2)
After all these preparations we can find dσ(2) . Assuming the initial and final electrons
unpolarized, we arrive at [53]:
π(Zα)3 E
dσ(2) = [1 − sin(ϑ/2)] dΩ. (21.178)
4|p|3 sin3 (ϑ/2)

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 569

The screening constant does not enter into the final result. This is the consequence of the
fact that at the intermediate stages of calculations we have repeatedly used the smallness of
η. Inasmuch as the transition from the case of the electron scattering to that of the positron
scattering is realized by means of the replacement Z → −Z, then the similar expression for
the positron has the opposite sign. Recall that in the first Born approximation the cross
sections for the electron and the positron coincide.
We are coming now to the definition of dσrad . Employing Eqs. (21.96) and (21.132), one
may find the matrix element M(3) . However, in so doing, we must analytically continue the
(3) (2)
expressions for ΛCµ (p1 , p2 ) and ΠC (q 2 ) into a region of the negative values of q 2 . Putting
ϑ/2 → iΦ, where the angle Φ is now connected with the scattering angle by the relation:
θ
−q 2 = 4p2 sin2 = 4m2 sinh2 Φ, (21.179)
2
in Eqs. (21.96), (21.132) and allowing for (21.165) (here the screening constant may be set
equal to zero), we find
iZe2 α m nh  m
R

M(3) = u(p 2 )γ0 (1 − 2Φ coth 2Φ) ln − 1 +
π|q|2 E λ
Z Φ #
1 q̂ 1Φ
+2 coth 2Φ (u tanh u)du + Φ tanh Φ − −
0 2 2m R sinh 2Φ
  
1 2 1
− 1 − coth Φ (1 − Φ coth Φ) − u(p1 ). (21.180)
3 9
Further, we substitute M(3) , M(1) into (21.163) and carry out the summation and average
over the electron spin states. As a result we derive [54]

Z 2 α3 E 1
dσrad = − [1 − v sin (θ/2)] 2(1 − Φ coth Φ)(1 − coth2 Φ)−
2 2
4π|p|3 sin4 (θ/2) 3
  Z Φ
2 λ
− + Φ tanh Φ + 2(1 − 2Φ coth 2Φ) 1 + ln + 4 coth 2Φ u tanh udu+
9 mR 0

2Φv 2 sin2 (θ/2)
+ . (21.181)
sinh 2Φ[1 − v 2 sin2 (θ/2)]
The last thing that remains to be done is to get rid of the fictitious photon mass λ. As
follows from Section 20.2 we must be considering the pure elastic scattering along with the
scattering with emitting a soft photon whose energy ∆E is significantly less than that of the
electron. We add the cross section accompanied by emitting a soft photon (see Eq. (20.41)
) to the cross sections sum dσ(1) + dσ(2) + dσrad and take into account that the integral
entering into (21.181) may be converted to the form:
Z Φ Z
(1 − v 2 ) sinh 2Φ 1 ξ ln(1 − v 2 ξ 2 )dξ
u tanh udu = p +
0 4 sin(θ/2) 2 2 2
cos(θ/2) (1 − v ξ ) ξ − cos (θ/2)
2

Φ 1
+ ln . (21.182)
2 1 − v2
Then, as will readily be observed, the addends containing ln λ are canceled and we come
to the final formula for the cross section of the electron scattering by the Coulomb nucleus
field in the e6 approximation [55]:
 
dσ dσ
= [1 − δ(E, θ, ∆E)] , (21.183)
dΩ dΩ M

© 2011 by Taylor and Francis Group, LLC

 570 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where δ(E, θ, ∆E) = δrad − δ(2) , (dσ/dΩ)M is the Mott effective cross section,
 
dσ Z 2 α2
= [1 − v 2 sin2 (θ/2)]
dΩ M 4|p|2 v 2sin4 (θ/2)

and the quantities δ(2) and δrad are determined by the expressions:

1 − sin(θ/2)
δ(2) = πZαv sin(θ/2) , (21.184)
1 − v 2 sin2 (θ/2)
  
α 2∆E
δrad = 2(1 − 2Φ coth 2Φ) 1 + ln + Φ tanh Φ + 2(1 − Φ coth Φ) ×
π mR
 
1 2 1 1−v 1
× 1 − coth2 Φ − + ln + 2Φ coth 2Φ ln +
3 9 v 1+v 1 − v2
Z 
2Φv 2 sin2 (θ/2) (1 − v 2 ) cosh 2Φ 1 ln(1 + vξ)
+ 2 + −
2
sinh 2Φ[1 − v sin (θ/2)] v sin(θ/2) cos(θ/2) 1 − vξ
 )
ln(1 − vξ) dξ
− p . (21.185)
1 + vξ ξ 2 − cos2 (θ/2)
In a similar way the photon mass is removed in the cross section to be calculated in more
higher orders of the perturbation theory. So, considering any scattering process (as complex
as is wished), being the effect of the nth order of the perturbation theory, along with its RC
that are associated with the (n + 2)th order of the perturbation theory, we must take into
account the process of emitting a real soft photon as well. And in this case the summary
cross section does not include the photon mass [56]. From Eqs. (21.184) and (21.185) it
is evident that in the nonrelativistic region the quantities δ(2) and δrad tend to zero when
v → 0. Of course one may use the formulae (21.184) and (21.185) only in the case when
the quantities δ(2) and δrad are rather small. In particular, q 2 must not be very large and
the nucleus must be light enough (the second Born approximation gives the right results
at Z < 10—15 only). From the formula (21.185) it also follows that with decreasing ∆E,
that is, with improving the energy detector resolution, the quantity δrad becomes large and,
as a result, the expression (21.183) breaks down. Under these circumstances, the RC of
higher orders become important. In Ref. [57] it was shown that the cross section may
be represented in the form (21.183) under inclusion of the RC of a more higher order as
well. The expression standing in the right-hand side of (21.183) resembles expanding the
exponent into the series. By this reason J. Schwinger made the assumption [58] that, in the
most general case, the formula:
 
dσ dσ
= e−δ(E,θ,∆E) . (21.186)
dΩ dΩ M

takes place. Some time later it was demonstrated [59] that under small values of ∆E
(∆E ≪ E) the Schwinger assumption is asymptotically valid. In order to gain some insight
into the effect of the RC, in Table 21.1 we display the δ(2) and δrad values expressed in
percent versus E, Z, θ, and ∆E. We refer the reader, who wishes to gain more detailed
information concerning the RC calculation in the QED, to the excellent paper by Brawn
and Feynman [60] devoted to computing the cross section of the Compton scattering in the
e6 approximation.

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 571
TABLE 21.1
The δ(2) and δrad values versus E, Z, θ, and ∆E.
E 2.5 MeV 4 MeV 9.5 MeV
θ 450 900 1350 450 900 1350 450 900 1350
δ(2) /Z 0.61 0.89 0.86 0.62 0.92 0.99 0.63 0.94 1.07
δrad at ∆E=10 kEV 4.8 7.4 8.7 6.9 9.9 11.3 12.4 15.9 17.5
δrad at ∆E=25 keV 3.9 6.0 7.1 5.7 8.1 9.3 10.5 13.5 14.8
δrad at ∆E=50 keV 3.2 5.0 5.9 4.7 6.8 7.9 9.0 11.7 12.8

21.7 Anomalous magnetic moment of leptons

It is known that when a nonrelativistic particle with the magnetic moment is placed in an
external magnetic field, then it acquires an additional energy Eint = −(µH). Therefore, for
the intrinsic magnetic moment of a particle to be measured, one must investigate its inter-
action with a constant or slowly varying electromagnetic field and extract in the interaction
energy a term proportional to the magnetic field.
We now turn to the electron scattering by an external field taking into account the
diagrams up to the third-order inclusive (Fig. 21.8). In this case one may understand the
electron scattering as that by the effective potential (Aeµ )ef f . Assuming this field weak,
we can be limited by an approximation linear in (Aeµ )ef f . Then, the scattering amplitude
including the contributions coming from the diagrams a, b, d, e, and f pictured in Fig. 21.8
is given by the expression:
s
m2
A = −ie(2π)3/2 u(p2 )Λµ (q)u(p1 )Aeµ (p2 − p1 ), (21.187)
Ep1 Ep2

where
(2)τ µ (3)µ
Λµ (q) = [γ µ + γ ν Dντ
c
(p2 − p1 )ΠC (p2 − p1 ) + ΛC (p2 , p1 )]. (21.188)

It should be emphasized that, in accordance with Eqs. (21.96) and (21.132) at small values
(3)µ
of q,∗ only the quantity ΛC (p2 , p1 ) contains addends linear with respect to q.
The most general structure of the matrix element that follows from the relativistic and
gauge invariance (see, for example, deducing the formula (10.53) ) is as follows:
s  
3/2 m2 i
A = −ie(2π) u(p2 ) FE (q 2 )γ µ + FM (q 2 )σ νµ qν u(p1 )Aeµ (p2 − p1 ).
Ep1 Ep2 2m
(21.189)
Thus, the electron vertex function Λν is represented in terms of two form factors FE,M to
describe electromagnetic properties of the electron. If we examine the vertex in the first
order of the perturbation theory, then

FE0 (q 2 ) = 1, 0
FM (q 2 ) = 0. (21.190)

It means that deviations of the form factors FE,M (q 2 ) from FE,M

0
(q 2 ) should be interpreted
as electromagnetic corrections of higher orders. To give the physical explanation of these
corrections we address the nonrelativistic approximation.

∗ Small values of q correspond to a slowly varying field.

© 2011 by Taylor and Francis Group, LLC

 572 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

In the lowest order of the perturbation theory the relation:

 Z   Z 
3 3
e lim < p2 | d xψ(x)Λµ (x)ψ(x)|p1 > = lim < p2 | d xjµ (x)|p1 > (21.191)
p1 ,p2 →0 p1 ,p2 →0

takes place. Obviously, with including higher orders of the perturbation theory, the left-
hand side of (21.191) becomes equal to
eu(0)Λµ (0)u(0) = eu(0)γµ FE (0)u(0). (21.192)
Next, taking into account the definition of the total measured charge eexp :
 Z 
3
eexp = lim < p2 | d xj0 (x)|p1 > (21.193)
p1 ,p2 →0

and allowing for that u(0)γu(0) = 0, we arrive at

eexp
FE (0) = . (21.194)
e
Due to Eq. (21.194) the quantity FE (q 2 ) is sometimes called the charge form factor. How-
ever, this title is not absolutely exact because at q 2 6= 0 the magnetic moment is also
connected with the matrix element eu(p2 )γµ u(p1 ). To make sure that it is the case we
again access to the nonrelativistic approximation. Expand the matrix element of the cur-
rent into the series in powers of v/c and, then, multiply every term of the series by the
potential of the external electromagnetic field:
 
A(q) · (p1 + p2 ) σ · [A × q]
eAu(p2 )γu(p1 ) ≈ −ieϕ(p2 ) +i ϕ(p1 ), (21.195)
2m 2m
where q = p2 − p1 , ϕ(p) is the big component of the bispinor u(p) and we have accepted
the three-dimensional gauge of the potential
(A(q) · q) = 0.
Evidently, the first addend in (21.195) corresponds to the linkage of the current ep/m with
the potential A while the second one governs the linkage of the electron with the external
magnetic field H
σ · [A × q]
−ieϕ(p2 ) ϕ(p1 ) = µ0 ϕ(p2 )σ · Hϕ(p1 ),
2m
where µ0 = e/(2m). In the Dirac theory the electron automatically has the “normal”
magnetic moment ege /2m with ge = 2. Thus, the title “Dirac form factor” is best suited
to the form factor FE (q 2 ).
The form factor FM (q 2 ) may be named for Pauli. Really, by virtue of the relation:
1 µν
σ Fµν = −[(Σ · H) − i(α · E)],
2
one may state that the RC make the electron behave in such a way as if it possesses
distribution of an anomalous dipole (magnetic and electric) moment to be characterized
by FM (q 2 ). So, FM (0) measures contribution to the (intrinsic) magnetic moment of the
electron caused by electromagnetic corrections of higher orders. In the Bohr magneton
units the total (intrinsic) magnetic moment of the electron µe may be written as
 
(g − 2)e
µe = 1 + µB , (21.196)
2

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 573

where 1 standing in the parenthesis results from the term FE0 (q 2 )γν in Eq. (21.189), and
ae = (g − 2)e /2 is an anomalous magnetic moment (AMM).
Inserting (21.132) and (21.96) into (21.188), we find
" Z ϑ #
α  m  ϑ
2 R
FE (q ) = 1 + (1 − 2ϑ cot 2ϑ) ln − 1 + 2ϑ cot 2ϑ ξ tan ξdξ + tan ϑ −
π λ 0 2
 
4m2R + 2q 2 1
− (1 − ϑ cot ϑ) − , (21.197)
3q 2 9
α 2ϑ
FM (q 2 ) = . (21.198)
2π sin 2ϑ
Since
α
FM (q 2 )|q2 =0 =
,
2π
then we can conclude that owing to the RC in the second order of the perturbation theory
the electron placed in a slowly varying external electromagnetic field behaves as a particle
having the magnetic moment  α
µe = 1 + µB , (21.199)
2π
that is,  
(g − 2)e α
= . (21.200)
2 theor 2π
We call attention to the fact that the AMM is directly related to the vertex function. By
this reason, computing corrections to the magnetic moment it is sufficient to consider the
vertex diagrams only. For example, in order to compute the electron AMM with a precision
of terms of the order of α3 , we must deal with diagrams shown in Fig. 21.15. Calculations
lead to the result [61]:
(g − 2)e α  α 2
= − 0.328 . (21.201)
2 2π π
At present, within the QED there are the calculations of the electron AMM in higher
orders of the perturbation theory
(g − 2)e X  α n
(ae )QED = = an . (21.202)
2 n
π

For example, for n = 4 the result is [62]:

α  α 2  α 3  α 4
(ae )QED = − 0.32847844400 + 1.181234017 − 1.5098(384) . (21.203)
2π π π π
Emphasize a promptly increasing complexity of calculations under transition to more higher
orders of the perturbation theory. With computing a3 (the contribution is accounted for
of sixth-order diagrams) the diagrams number is equal 72. In the standard model the
contributions to the AMM coming from the electroweak sector aEW e and those caused by
corrections to the photon propagator from virtual hadrons ahad
e have also been calculated
[62]:
aEW
e + ahad
e = 1.66(3) × 10−12 . (21.204)
Now we address to experiments on observing the electron AMM. Nafe, Nelson, and Rabi
were the first [63] who detected the AMM existence investigating the hyperfine splitting of
the energy levels for the atomic hydrogen and deuterium. The idea of these experiments is
as follows.

© 2011 by Taylor and Francis Group, LLC

 574 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

p2 − k
p1 − k p2 − k p1 − k k′

p1 p2 p1 p2

k k k
p

(a) (b)

p2 − k
k′
p1 − k k′ p1 − k p2 − k

p1 p2 − k ′ p2 p1 p2
k k

(c) (d)

k′
p1 − k ′ p2 − k

k p2
p1

(e)

FIGURE 21.15
The Feynman diagrams (a)–(e) contributing to the electron AMM.

The hyperfine energy levels structure of atoms and molecules is due both to the interaction
of a magnetic nucleus field with a magnetic field created by electrons and to that of a
quadrupole nucleus moment with a nonhomogeneous intra-atomic electric field. In atoms
and ions the magnetic interaction plays the main role. Let us find the corresponding energy
for the hydrogen atom being constrained by the nonrelativistic case [64].

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 575

If the hydrogen atom nucleus has the magnetic moment µp = µp σ p , where σ p are the
Pauli matrices acting upon the nucleus wavefunctions, then it produces the magnetic field:
µp σ p
A = rot , H = rot A.
r
The magnetic field of the nucleus must act on the electron magnetic moment µe = −µe σ e
(σ e has an effect upon the electron wavefunctions), with the result that an additional
interaction between the nucleus and the electron appears. This interaction leads to the
hyperfine structure of the energy levels:
 σp 
Hint = −(µe H) = µe µp σ e rot rot =
r
 
1 1
= µe µp (σ e ∇)(σ p ∇) − (σ e σ p )∇2 . (21.205)
r r
In the first approximation one may assume that there are no chosen directions, and, hence,
the relations:
1
(σ e ∇)(σ p ∇) = (σ e σ p )∆, (21.206)
3
1
∆ = −4πδ(r). (21.207)
r
take place. Then, we get
8π
Hint = µe µp (σ e σ p )δ(r). (21.208)
3
From (21.208) it follows that in the first approximation the interaction of the magnetic
moments has an impact on the S-state only.
To find (σ e σ p ) we make use of the eigenvalues equation for the operator of the total spin:
1 2
~ (σ e + σ p )2 = ~2 I(I + 1), (21.209)
4
where I is an eigenvalue of the hydrogen atom spin (I = 0, 1). Since the left-hand side of
the expression (21.209) may be rewritten in the view:
1 2 2 1
~ [σe + σp2 + 2(σ e σ p )] = ~2 [6 + 2(σe σ p )], (21.210)
4 4
we get
(σ e σ p ) = 2I(I + 1) − 3. (21.211)
Allowing for Eqs. (21.208) and (21.211) we could determine the average of the operator
Hint Z
8π
∆EI = d3 xψ † (r)Hint ψ(r) = µp µe |ψ(0)|2 [2I(I + 1) − 3]. (21.212)
3
So
∆EI=0 = −8πµn µe |ψ(0)|2 , (21.213)
when the electron and proton spins are antiparallel, and
8π
∆EI=1 = µn µe |ψ(0)|2 , (21.214)
3
when the electron and proton spins are parallel. For the hydrogen atom in the S-state we
have
1
|ψ(0)|2 = ,
πn3 a30

© 2011 by Taylor and Francis Group, LLC

 576 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where
~2
a0 = .
me2
Then, the difference between the levels that defines S-therm splitting caused by the electron
interaction with the magnetic moment of the nucleus is governed by the expression:
∆EI=1 − ∆EI=0 32µe µn 1
∆ω = = , (21.215)
~ 3~ n3 a30
At n = 1 it follows from the formula (21.215)
∆ωth = 1417 MHz,
while in the experiments by Nafe, Nelson, and Rabi the obtained result was:
∆ωexp = 1420 MHz,
what has corresponded to 0.26% deviation from the theoretical result under condition (g −
2)e /2 = 0. It has been found that the frequency ∆ωth was not growing to ∆ωexp with
allowance for the relativistic corrections or nucleus mass finiteness. Since by that time a
magnetic moment of the proton has been measured within a sufficient accuracy, the only
doubts were about a magnetic moment of the electron. Because of this, an assumption put
forward by Breit [65]—the electron possesses not only the kinematic magnetic moment equal
to e~/mc but also the intrinsic magnetic moment the existence of which is not described by
the conventional Dirac theory—looked quite naturally.
Using the magnetic resonance method, in 1948 P. Kusch and H. Foley [66] performed
high-precision measurements for the frequencies of the Zeeman splitting in a spectrum of
gallium. Their direct measurements of the electron ge -factor provided support for Breit’s
assumption. Schwinger was the first to calculate the quantity of the AMM in the second
order with respect to e [67]:
α
a(2)
e =
2π
(sometimes the additive to the AMM equal to α/(2π) is called the Schwinger term). To
(2)
substitute the value µe = (1 + ae )µB into (21.215) gives the result being in the excellent
agreement with the experiment.
The (g − 2)e /2-anomaly considerably changes the spin dynamics of the Dirac electron.
To demonstrate this we consider the electron motion in a magnetic field that is constant in
time and homogeneous in space. In this case the Bargmann–Michel–Telegdi equation [68]
describing the spin evolution has the form:
dξ ec eae (vH)
= (1 + γae )[ξ × H] + [v × ξ], (21.216)
dt E Ec(1 + γ)
where γ = E/(mc2 ) and we have turned to the ordinary system of units. As a rule, of
particular importance is a change in the absolute spin direction relative to the direction of
motion rather than its changing in space. Present ξ in the view:
ξ = nξk + ξ ⊥ ,
where n = v/v and investigate the case when the magnetic field is directed perpendicular
to the electron motion orbit plane v⊥H = 0. In the right-hand side of (21.216) only the
first term is left and, hence, the particle spin is rotating about the direction H with the
frequency:  
1
Ωs = Ω a e + ,
γ

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 577

where Ω = eH/(mc), whereas the electron rotation frequency proves to be less

ecH Ω
Ωe = = .
E γ
The difference of these frequencies is proportional to the AMM value:

Ωs − Ωe = a e Ω (21.217)

and does not depend on a particle motion speed. Inasmuch as the perpendicular projection
of the spin ξ ⊥ is the motion integral

dξ ⊥
= 0, (21.218)
dt
then only ξk can precess with respect to the direction v. Really, allowing for the motion
equations
dp ec
= [p × H],
dt E
we get from (21.216)
d
(ξp) = −Ωae [ξ × p]h (21.219)
dt
(h = H/H), that is, ξk oscillates with the frequency Ωae . Exploring the variation in time
of the longitudinal polarization for an ensemble of polarized particles just comprises the
essence of a precession method used to find the AMM.
Besides, there exists an other method, a spin-resonance one. It is based on the spin-
resonance experiments conducted to study frequencies of the transitions between the Landau-
Rabi levels for the lepton in a homogeneous magnetic field. It is clear that a significant
progress in measurements of the AMM should be expected due to experiments with free
electrons. For the first time the AMM value for a free electron was measured in 1957 by
Dehmelt (see the review paper [69]) who, using a high-frequency field, has succeeded in
the induction and observation of the spin-resonance transitions of the electrons moving in
a magnetic field. Let us explain the essence of the spin-resonance method.
We take care of the Pauli equation
 
1  e 2 ge µB
p− A + (σH) ψ(r) = Eψ(r) (21.220)
2m c 2

where we are using the ordinary units as before. In a homogeneous magnetic field directed
along the axis z, the electron energy spectrum has the form (Landau-Rabi levels):
 
1 ms g e p2
En,ms = ~Ω n + + + z , (21.221)
2 2 2m

where ms = ±1/2 is associated with the spin projection onto the field direction. If we
had ge = 2, then the electron energy would have a spin degeneracy (thus, for example,
E1,1/2 = E2,−1/2 and so on). However, as ge = 2(1 + ae ), this degeneracy is absent and we
come to two sets of the energy levels in accordance with the spin projection onto the field
direction H: 
1 ae 
1. ms = , E1 = ~Ω n + 1 + ;
2 2
1  ae 
2. ms = − , E2 = ~Ω n − ,
2 2

© 2011 by Taylor and Francis Group, LLC

 578 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

where we are interested in only the cyclotron orbital motion, that is, pz = 0. The quantum
transitions between the adjacent levels in each series are realized without a spin flip and
are characterized by the frequency Ω. The transitions with a spin flip are corresponding to
the transitions from one series to another. In the process the frequency of these transitions
in combination with a change of the orbital quantum number turns to be proportional
to ae Ω. Observation using the spin-resonance method is based on using two magnetic
fields: a constant magnetic field to fix the projection of the particle spin moment, and an
alternating magnetic field to generate the spin-flip transitions. Creation of the Penning
trap has improved the potentialities of the spin-resonance method considerably. In this
trap, combination of stationary electric and magnetic fields provide a finite motion of the
electrons in the finite region of space. To illustrate, an isolated electron may be trapped
for a period of several days, offering a unique possibility for continuous observation of its
characteristics.
The electric potential of the Penning trap is chosen in the form:

e2 a(x2 + y 2 − 2z 2 )
ϕ=− ,
2
where a is an constant, and a homogeneous magnetic field H = (0, 0, H) is specified by a
vector potential
A = (−Hy/2, Hx/2, 0).
Since in a homogeneous magnetic field the space and spin parts of the wavefunction are
separated, then the Pauli equation for the electron put in the Penning trap could be easily
solved. The finiteness demand of the obtained solution leads to the following expression for
the electron energy spectrum:
       
1 1 1 g e ms Ω
E = ~ ωc′ n + + ω z nz + − ω m nm + + , (21.222)
2 2 2 2
where n, nz , nm = 0, 1, 2, 3... . From the viewpoint of physics, this spectrum is associated
with cyclotron orbital motion at p the modified frequency ωc′ = Ω − ωm , axial harmonic
oscillations 2
p at the frequency ωz = 2e a/m, magnetron oscillations at the frequency ωm =
Ω(1 − 1 − 2ωz2 /Ω2 )/2, and the last term in (21.222) defines the energy of interaction
between the spin and magnetic field. The radial and vertical motion is simultaneously
stable at the condition that
(1 − 2ωz2 /Ω2 )2 > 0.
Hence, it follows that, when
H2
0<a< ,
4me2
the electron is remaining inside the trap under ϕ0 = e2 a/2. Since the formula:
ωa − ωm
ae = , (21.223)
ωc′ + ωm

where ωa = Ωge /2 − ωc′ , takes place, the AMM definition is reduced to measuring the
frequencies in the case of the spin-resonance method too. In so doing, it is enough to find
only three frequencies since all the frequencies are coupled with the relation:

ωz2 = 2ωm ωc′ .

Note that both precession and spin-resonance methods enable one to measure the AMM
directly, independently of measuring the g-factor that specifies the gyromagnetic ratio.

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 579

Up to now, the electron AMM has most exactly been measured by application of the
spin-resonance method in experiments at the University of Washington [70]
ae = 0.0011596521884(43).
The experiments were realized using the Penning trap with a single electron during some
days to have ensured the high accuracy of measurements.
Our consideration of the AMM has concerned its kinematic part only. No doubt that the
AMM of any particle has a dynamic nature as well. The influence of an external magnetic
field on the electron AMM has been investigated in Ref. [71]. It was shown that the additive
to the electron MM induced by a homogeneous magnetic field is defined as
α
δae (E, H) = F (E, H),
2π
where H is a magnetic field strength, E is an electron energy, F (E, H) is a function of a
parameter κ = (H/H0 )(E/mc2 ), and H0 is the critical value of the magnetic field. In the
case κ ≪ 1, F (E, H) is given by the expression:
 
37 1 1
F (E, H) = 12κ2 − ln − C − ln 3 , (21.224)
12 κ 2
where C = 0.577.... In ordinary conditions of experiment the values of F (E, H) are vanish-
ingly small. However, for the electrons with the energy 10 MeV in the field with H ≈ 4×104
Gs,
F (E, H) ≈ 4 × 10−8
which is available to observe.
We are coming to the investigation of the muon AMM. In this case the correction in the
(2)
second order on e aµ has the same form as that for the electron AMM, that is,
α
a(2) (2)
µ = ae (me → mµ ) = µ0 , (21.225)
2π
where µ0 = e~/(2mµ c) is the normal magnetic moment of the muon. However, this is not
the case for the correction in the fourth order on e. The reason is that the vertex function
(2)
includes a polarization operator Πµν to be determined by contributions coming from vir-
tual e− e+ -, µ+ µ− - and τ + τ − -pairs for any charged leptons. Since these contributions are
inversely proportional to the mass square of particles forming a loop, then those of more
lighter particles dominate. It means that for the muon the contribution coming from the
electron-positron loop will be dominated. By now computing the muon AMM has been
fulfilled within the QED in the fifth order on α [62]:
α  α 2  α 3
aQED
µ = + 0.765857376(27) + 24.05050898(44) +
2π π π
 α 4  α 5
+126.07(41) + 930(170) . (21.226)
π π
Growing the coefficients in the (α/π) expansion reflects the presence of large quantity
ln(mµ /me ) ≈ 5.3, which comes from electron loops. The most exact determination of
the fine structure constant at the scale E = me c2 has been done in Ref. [72]
α−1 (ae ) = 137.03599958(52). (21.227)
To substitute (21.227) into (21.226) results in
aQED
µ = 116584705.7(2.9) × 10−11 . (21.228)

© 2011 by Taylor and Francis Group, LLC

 580 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Up to date the value of the muon AMM measured at Brookhaven National Laboratory
(BNL) on Alternating Gradient Synchrotron with the help of the precession method [73] is
given by the expression:
aexp
µ− = 11 659 214(8)(3) × 10
−10
µ0 , (BNL’01). (21.229)

The uncertainty of BNL 2001 result is equal to 0.7 ppm (ppm = parts per million or 1 ppm
= 10−6 ). In case of the electron the attained precision is much higher, being at a level
of ppb units. Nevertheless, the measurements of aexp µ− give us more information about the
short-range phenomena as the value of such effects proves to be proportional to m2l [74].
In this way we have the amplification factor of m2µ /m2e ≈ 40000 fully compensating for the
inadequate accuracy in the measurement of aexp µ− . All this results in the sensitivity of the
muon anomaly to hadron and electroweak corrections, providing a reliable instrument for
observation of the effects caused by still undiscovered particles having masses on the order
of several hundred GeV. As a consequence, with calculating the muon AMM contributions
coming from all model sectors must be taken into account. It means that in the SM, aSM µ
is defined as:
aSM
µ = aQED
µ + aEW
µ + ahad
µ , (21.230)
where the electroweak contribution is given by the expression [62]:
aEW
µ = 152(4) × 10−11 µ0 . (21.231)

The largest theoretical ambiguities in aµ are governed by the quantity ahad µ , which is mainly
defined by contributions of virtual hadrons to a photon propagator in the fourth and sixth
orders. Lower corrections cause vacuum polarization by hadrons ahad µ (V P 1), while correc-
tions of the sixth order along with vacuum polarization by hadrons ahad µ (V P 2) include in
itself light-by-light scattering ahad
µ (LbyL) as well. Since a had
µ contains integrals over virtual
particle energies including a region of soft energies, ahad
µ could not be reliably calculated
had
within the perturbative QCD. Only the existing value of the quantity aµ (V P 2) cause no
doubts. As for ahadµ (LbyL) all its estimations made so far are model dependent. The calcu-
lations are based on the chiral perturbation or on the extended Nambu-Jona-Lasinio model.
Also the vector meson dominance is assumed and the phenomenological parametrization of
the pion form factor πγ ∗ γ ∗ is introduced for regularizing the divergences. The value of
ahad
µ (V P 1) can be calculated on the basis of experimental data concerning the investiga-
tions of the cross section e+ e− → hadrons and/or on the basis of the analysis of data
concerning a τ -decay into hadrons. However, the inclusion of the latter introduces sys-
tematic uncertainties originating from isospin symmetry breaking effects that are difficult
to estimate. A recent analysis, performed in Ref. [75] showed, that ahad µ (V P 1), obtained
exclusively on the basis of e+ e− data differs a lot from ahad µ (V P 1), whose calculation is
based on the data, concerning τ → hadrons. Since all data concerning the QED effects are
in excellent agreement with the theoretical predictions, then only the corrections coming
from the electroweak and hadron sectors are under suspicion. We shall come back to the
detailed discussion of this problem in the second volume.

21.8 Lamb shift

The Dirac equation for the electron in the Coulomb field of a point nucleus predicts de-
generacy of the energy levels for the bound states, possessing the same values of the main

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 581

quantum number n and the quantum number of the total moment j and differing in the
values of the quantum numbers for the orbital moment l = j ± 1/2. Thus, for example, the
states 2S1/2 and 2P1/2 should have the same energy. With the use of radio spectroscopic
methods in 1947 Lamb and Retherford [76] demonstrated that in a hydrogen atom the level
2S1/2 is higher than the level 2P1/2 approximately by 1000 MHz. This discovery was an
impetus for all further development of the QED finally leading to the understanding of
this phenomenon. Removal of the degeneracy from the levels 2S1/2 and 2P1/2 is caused
by quantum fluctuations of the vacuum of the electromagnetic and electron-positron fields
modifying the potential energy ϕ(r) = −Ze2 /r associated with the electron-nucleus inter-
actions. Also, this process named the Lamb shift may be observed between other levels in
hydrogen or hydrogen-like atoms. This term is frequently used in a more wide sense as a
level shift of the bound electron states in an external field caused by the RC.
On examination of the Lamb shift we shall follow Ref. [77] in which the Dirac theory
is combined with the RC. Let us consider the electron scattering process in an external
field Ae (x). The corresponding Feynman diagrams are represented in Fig. 21.7. From
the expression (21.159) it is evident that the cumulative effect of the RC corresponding to
the diagrams pictured in Figs. 21.7(a)–(f ) can be regarded as changing the external field
potential. Then, the effective potential for the electron proves to be equal to
(2)τ µ (3)µ
γ µ [Aeµ (q)]ef f = [γ µ + γ ν Dντ
c
(p2 − p1 )ΠC (p2 − p1 ) + ΛC (p2 , p1 )]Aeµ (q), (21.232)

rather than γ µ Aeµ (q). It is convenient to split the change of γ µ δAeµ into three parts

γ µ δAeµ (q) = γ µ [δA(γ) AMM

µ (q) + δAµ (q) + δA(V
µ
E)
(q)], (21.233)
(γ)
where a source of the correction δAµ (q) is a vacuum polarization, δAAMM
µ (q) is associated
(V )
with the AMM of the electron, and the correction δAµ E (q) appears due to the Dirac form
factor in the expression for the vertex function. In the case of small transferred momenta
q 2 (see Eqs. (21.133) and (21.98) ) these quantities are as follows:

αq 2
γ µ δA(γ)
µ (q) = −γ
µ
Ae (q). (21.234)
15πm2R µ
 
α q̂γµ − γµ q̂
γ µ δAAMM
µ (q) = Aeµ (q), (21.235)
2π 4mR
 
αq 2 mR 3
γ µ δA(Vµ
E)
(q) = γ µ ln − Aeµ (q) (21.236)
3πm2R λ 8
Thus, in the coordinate representation the Lagrangian, describing the interaction between
the electron and an external field slowly varying in space and time, will be given by the
expression:
Z h i
Lint ≈ e d3 xψ(x)γ µ Aeµ (x) + δA(γ)
µ (x) + δAµ
AMM
(x) + δA(V
µ
E)
(x) ψ(x), (21.237)

where
α
γ µ δA(γ)
µ (x) = Aeµ (x), (21.238)
15πm2R
 
α mR 3
γ µ δA(V
µ
E)
(x) = −γ µ
ln − Aeµ (x), (21.239)
3πm2R λ 8
α
γ µ δAAMM
µ (x) = σ νµ ∂ν Aeµ (x). (21.240)
4πmR

© 2011 by Taylor and Francis Group, LLC

 582 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Using (21.237), we arrive at

 µ
γ [i∂µ − eAeµ (x) − δAeµ (x)] − m ψ(x) = 0, (21.241)

where
γ µ δAeµ (x) = γ µ [δA(γ) (VE )
µ (x) + δAµ (x) + δAAMM
µ (x)].
The equation obtained is nothing more nor less than the Dirac equation written with an
accuracy of the third order on e, which includes the mass operator in an external field with
inclusion of higher approximation effects [78]. We shall return to the deduction of this
equation in Section 22.2, and now we are coming to the investigation of (21.241) for the
hydrogen atom case assuming the interaction between the electron and the proton to be
purely Coulomb.
The quantity ψ(x) should be interpreted as the matrix element of the field operator
between the vacuum and the one-electron state in the presence of an external field. In
accordance with the perturbation theory rules the additional term in Eq. (21.241) must be
taken into account in the first order only. Unperturbed states are solutions of the Dirac
equation for a particle in an external field:

{γ µ [i∂µ − eAeµ (r)] − mR }ψ(x) = 0, (21.242)

where
Ze
Aeµ (r) = − gµ0 .
4πr
According to the general rules, the shift of the energy level caused by Hint is yielded by:
Z h i
(γ) (V )
δE = e d3 xψ † (r) δA0 (r) + δA0 E (r) + δAAMM 0 (r) ψ(r), (21.243)

where
(γ) α
δA0 (r) = − ∆Ae0 (r), (21.244)
15πm2R
 
(V ) α mR 3
δA0 E (r) = ln − ∆Ae0 (r), (21.245)
3πm2R λ 8
iα
δAAMM
0 (r) = (γ · E(r)). (21.246)
4πmR
To find δE one may make use of the solutions of Eq. (21.242), which have been obtained
by application of the perturbation theory. We take the Schrödinger equation solution for
the hydrogen atom:
|ml |
ψn,l,ml (r) ≡ ψn,l,j (r) = Nnl e−ρ/2 ρl L2l+1
n+l (ρ)Yl (θ, ϕ), (21.247)

where
2r 1
ρ= , a= ,
na mR e 2
Nnl is a normalization coefficient, L2l+1
n+l (ρ) is the generalized Laguerre polynomial, and
|m |
Yl l (θ, ϕ) are spherical functions, as the zeroth order approximation. Allowing for the
Coulomb character of the field and using
 
1
∆ = −4πδ (3) (r),
r

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 583

we deduce the following expression for the energy shift initiated by the first two addends in
Eq. (21.243):
Z  
Ze2 α 3 † mR 3 1 (3)
(δE1 )n,l,j = d xψn,l,j (r) ln − − δ (r)ψn,l,j (r) =
3πm2R λ 8 5
 
4mR α 4 mR 3 1
= δl,0 (Zα) ln − − . (21.248)
3πn3 λ 8 5
So, in the nonrelativistic approximation the addends (21.244) and (21.247) influence the
S-state only.
The expression for (δE1 )n,l,j includes the fictitious photon mass, therefore, we have to
gain an understanding of why it has happened. In the description of free-particle scattering
the IR-divergences are due to the ability of a charged lepton to emit and absorb soft photons
without much shifting from the mass surface p2 = m2R . As a result, our lepton propagators
have very small denominators and together with the photon propagators lead to the IR
divergences. But during the calculation of the Lamb shift no IR divergences appear as in
this case the four-dimensional momentum of the electron is beyond the mass surface. In
this way even at very low photon energies the denominator of the electron propagator is
m2R − p2 and not zero. The derived vertex operator is logarithmically dependent on this
value. Because of this, the factor 2 ln(λ/mR ) associated with the scattering of a free particle
in the case of the Lamb shift is replaced by ln(m2R − p2 )/m2R ). For the bound hydrogen-like
state we have
p0 = mR − En ,
where En is a binding energy in the state n and |p| ≈ ZαmR , to give
 
1 2 2 2 p2
(m − p ) ≈ En + .
m2R R mR 2mR

Then, in the case in question the factor ln λ2 /m2R is replaced with

2mR En + p2
ln .
m2R

Due to this factor the main contribution to changing the position of the level 2S1/2 is given
by the addend δA(VE ) . It shifts the level 2S1/2 approximately in 1010 MHz with respect to
that 2P1/2 . The vacuum polarization effects provide the additive to the shift E2S1/2 −E2P1/2
equal to -27 MHz.
Now we consider the contribution to the level shift induced by the electron AMM
Z
ieα †
(δE2 )n,l,j = d3 xψn,l,j (r)(γ · E(r))ψn,l,j (r). (21.249)
4πmR

If one neglects the Coulomb potential, then the big ϕ(r) and small χ(r) components of ψ(r)
will be coupled by the approximate relation:

(σ · ∇)
χ(r) = −i ϕ(r).
2mR

Inasmuch as eE = −Zαr/r3 and the relations:

ψ † (r)(γ · r)ψ(r) = ϕ† (r)(σ · r)χ(r) − χ† (r)(σ · r)ϕ(r) ≈

© 2011 by Taylor and Francis Group, LLC

 584 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics
1  † 
≈ ϕ (r)(σ · r)(σ · ∇)ϕ(r) + conj.
2imR
are valid, then after integration by parts the integral in (21.249) becomes the view:
Z Z  
(γ · r) i (σ · r)
d3 xψ † (r) 3 ψ(r) ≈ d3 xϕ† (r)(σ · ∇) ϕ(r) . (21.250)
r 2mR r3

Taking into consideration the identity:

[σj Aj , σk Bk ] = δjk (Aj Bk − Bk Aj ) + iεjkl σl (Aj Bk + Bk Aj ) (21.251)

and the relation

1 r
∇ = − 3,
r r
we arrive at
Z  
Zα2 (LS)
(δE2 )n,l,j = d3
xϕ†n,l,j (r) 4πδ (3)
(r) + 4 3 ϕn,l,j (r), (21.252)
8πm2R r

where L = −i[r × ∇] and S is the spin moment operator. The eigenvalues of the operator
(LS) are given by:
 
1 3
(LS) = (1 − δl,0 ) j(j + 1) − l(l + 1) − . (21.253)
2 4

To compute the average value for r−3 results in

2(ZmR α)3
< r−3 >n,l,j = , l ≥ 1. (21.254)
l(l + 1)(2l + 1)n3

With the help of (21.252), (21.253), and (21.254) we get the final result for (δE2 )n,l,j

α(Zα)4 mR
(δE2 )n,l,j = Cj,l , (21.255)
2πn3 (2l + 1)
where
j(j + 1) − l(l + 1) − 3/4
Cj,l = δl,0 + (1 − δl,0 ) =
l(l + 1)

(l + 1)−1 , when j = l + 1/2,
=
−l−1 , when j = l − 1/2, l ≥ 1.
The electron AMM leads to the levels shift approximately in 68 MHz. Thus, the predicted
value of the Lamb shift is in the neighborhood of 1052 MHz with an accuracy of α(Zα)4 .
In summary, we have obtained the approximate value of the Lamb shift using the average
values of the operators ΛµC (p1 , p2 ) and Πµν 2
C (q ) for hydrogen-like states under consideration.
Strictly speaking, the expression (21.132) for the vertex operator ΛµC (p1 , p2 ) is inapplicable
because it was deduced for a free electron (p21 = p22 = m2R ) rather than for a bound electron
(p2 6= m2R ). However, in our case |p| ≪ mR and the deviations from the formula (21.132)
are extremely small.
To upgrade the calculations, first we should go to an exact electron propagator in the
Coulomb field. As there is no closed expression for the electron propagator available, this
approach presents serious difficulties during the calculations and is still incompletely real-
ized. Besides, it is necessary to take into consideration all the corrections associated with
the nucleus: due to its magnetic moment, recoil, form factors, radius and mass finiteness,

© 2011 by Taylor and Francis Group, LLC

 Radiative corrections 585

polarizability, etc. Moreover, it is required to take into account the metastability of the
excited levels, that is, the natural width of the levels 2S1/2 and 2P1/2 .
By the present time, allowances have been made for the leading higher-order corrections
to the coupling constant Zα, second-order corrections to α in the electron self-energy, for
the electron AMM and the vacuum polarization, and also for the effects due to the proton
radius and mass finiteness. All this yields the following theoretical value for the Lamb shift
in the hydrogen [79]:
E2S1/2 − E2P1/2 = 1057.8640(140) MHz. (21.256)
When calculating these values, the main uncertainty results from the higher orders on
coupling (Zα) and the second order (α2 ) for the electron self-energy. The discrepancies
between the theoretical and experimental measurements for the Lamb shift are presently
insignificant. So, the experimental values obtained in Refs. [80,81] are

E2S1/2 − E2P1/2 = 1057.8450(90), (21.257)

E2S1/2 − E2P1/2 = 1057.8514(19). (21.258)

The survey of the consecutive theory of the Lamb shift in higher orders of the perturbation
theory could be found in Ref. [82]. A reader interested in details and calculations technique
is recommended to refer to the books [83].
An interest in the high-precision calculation and measurements of the Lamb shift in the
hydrogen atom is caused not only by the existing disparity between the theoretical results
and experimental data but also by the possibility of obtaining information concerning the
structure and properties of the corrections having no direct relation to the QED. The Lamb
shift defines the properties of the nucleus-bound electron, taking into consideration both the
QED effects and those due to the nucleus structure. For example, the Lamb shift correction
caused by a finite size of the proton δp :

m3 (Zα)4
δp = < rp2 >,
12
where < rp2 >—the root mean square proton radius, is within the accuracy limits of a
modern experiment. In the case of < rp2 >1/2 = 0.862 fm the correction amounts to 0.146
MHz. In principle, the experimental accuracy of ∼ 2 kHz attained in the estimation of the
Lamb shift by the use of an atomic interferometer [81] makes it possible to measure the
proton radius with an error of 0.007 fm, that is nearly one half of the experimental error in
the case of the elastic e− p-scattering.

© 2011 by Taylor and Francis Group, LLC

 22
Renormalization theory

“Prophet!” said I, “thing of evil!—prophet still,

if bird or devil!—
Whether Tempter sent, or whether tempest tossed thee
here ashore
Desolate yet all undaunted, on this desert land
enchanted—
On this home by Horror haunted—tell me truly, I
implore—
Is there—is there balm in Gilead?—tell me—
tell me, I implore!”
Quoth the Raven “Nevermore.”
Edgar Allan Poe, “The Raven”

With calculating the elements of the S-matrix in the QED, we may face three types of
divergences: the UV-divergences, the IR-divergences, and singularities due to the coinci-
dence of two or more poles in the integrand. Divergences of the last type may appear, for
example, when for some special values of the particle momenta a multiparticle process may
be subdivided into independent processes involving independent particle groups. Consider-
ation of such divergences is beyond the scope of this book, and an interested reader may
find the relevant information in Ref. [84]. The IR-divergences are also out of consideration
because they have put emphasis in the previous Sections. Now our attention is focused on
the UV-divergences, presenting much trouble indeed.
According to the terminology of quantum field theory (QFT), renormalization is un-
derstood as a procedure aimed precisely at the elimination of the UV-divergences. Such
divergences are known to appear, when integrals are evaluated over the momenta in the
matrix elements associated with the closed-loop Feynman diagrams. As it has been shown
earlier, some the UV-divergences in the QED may be removed by the redefinition of such
parameters as a mass and an electric charge of the electron. Here it will be demonstrated
that this procedure is sufficient to exclude all the singular components of the matrix ele-
ments and Green functions available in the QED. In other words, all the UV-divergences
may be reduced to the field additives δm and δe to the electron mass and charge. Formally,
these additives may be expressed in terms of the singular factors Zm and Ze to the bare
(initial) mass m and charge e:

m + δm = Zm m, e + δe = Ze e. (22.1)

1/2
The quantities measured in experiments include only the functions of Zm m and Ze e
products. The observables involve no bare mass or charge and no UV-singularities, taken
separately or in combination. This makes it possible to identify the products of (22.1) with
the physical meaning of the electron mass and charge:

mR = Zm m, eR = Ze e. (22.2)

587
© 2011 by Taylor and Francis Group, LLC
 588 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

An operation of redefining the physical parameters:

m → eR , e → eR , (22.3)
by means of their multiplication into singular factors, completely excluding the UV-divergen-
ces from the observables, is referred to as the renormalization operation. The full description
of such an approach to renormalization of the QED has been first proposed by Dyson [4].
This method has been further developed in Refs. [85–88].
The operation used to cancel the UV-divergences may be formalized without the relations
similar to (22.2) as well. The idea is as follows. The RC to the matrix elements are expressed
in terms of the propagator products. In the coordinate representation the propagators have
the light-cone singularities giving rise to the UV-divergences. Mathematically, this amounts
to the problem of finding the operation of multiplication for the generalized functions. The
theory of multiplying the generalized functions involved in quantum-field calculations has
been developed by Bogoliubov. By application of this theory, Bogoliubov and Parasyuk
have solved the problem of removing the UV-divergences [89]. They have proved a theorem
(Bogoliubov–Parasyuk theorem) enabling the derivation of unambiguous expressions for the
scattering matrix elements, within the scope of the perturbation theory but without the
recourse to the intermediate regularization, counter-terms, and singular expressions similar
to those involved in (22.2). A reader could find the extensive description of this procedure
in the book [90].
It is naive to suppose that a renormalization theory should be necessarily based on a
series expansion of the perturbation theory. The first attempts to formulate this theory
without any approximations on the basis of an axiomatic approach have been made in Refs.
[91]. A complete description of a renormalization theory in the axiomatic QFT may be
found in [92]. More detailed references to the theory are given in the monograph [93] that
is a good guide on the problem.
When considering a quantitative renormalization theory, we shall hold to historic se-
quence, that is, follow the method developed by Dyson [4] and Ward [94].

22.1 Primitive divergent diagrams

Let us review the types of UV-infinities found in the QED. To this end, we consider the
general-form diagram including internal and external electron and phonon lines. A matrix
element associated with this diagram involves integration over all the internal momenta pk .
As is known, such an integration may be performed conveniently after the Wick’s rotation.
According to Dyson [95], in the process no extra divergences are brought about. With using
such transformations of an integral, a study of the integral and its divergence is simplified.
When the momentum degree in the numerator is lower than that in the denominator, the
integral is convergent. At the same time, when the momentum degree in the numerator
is equal to or higher than that in the denominator, the integral may be divergent but not
necessarily in all cases. The integral dimension D is introduced as a difference in the total
degree of momenta in the numerator and denominator of the integrand. Then a sufficient
convergence condition may be written as:
D < 0, (22.4)
while the condition at which the divergences are possible takes the form:
D ≥ 0, (22.5)

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 589

D = 0 being understood as a logarithmic divergence. To characterize diagrams in the QED

we shall use the following notations: Fe is a number of external electron lines, Fi is a number
of internal electron lines, Be is a number of external photon lines, and Bi is a number of
internal photon lines, n is a number of vertices.
Specifying the number and type of external lines, we concretize the process. And assum-
ing the values of Fi and Bi , we define an approximation for the calculation of this process.
The momenta degree in the denominator and numerator is calculated in the following way:

(i) Every internal electron or photon line enters the factor d4 p into the numerator. It
gives the contribution in D equal to 4(Fi + Bi ).
(ii) In every vertex the four-dimensional δ-function appears after integration over the whole
space-time. Its presence is equivalent to the addition of the fourth power of a momentum
into the denominator. Since one of δ-functions is reserved for realization of the total four-
dimensional momentum conservation law, then the overall contribution of vertices in D is
equal to −4(n − 1).
(iii) Every internal electron line introduces the first power of a momentum into the denom-
inator to contribute −2Bi to D.
(iv) Every internal photon line introduces the second power of a momentum into the de-
nominator to give the contribution in D equal to −2Bi .

Therefore, the dimensionality of D will be defined by the equation:

D = 3Fi + 2Bi − 4(n − 1). (22.6)

We try to express this quantity in terms of only the external lines number. If we shall
manage to do it, then the divergences number in the theory proves to be finite.
Because two electron lines are converging at each vertex, a number of the electron line
ends is equal to the doubled number of the vertices. As each internal electron line has two
ends and each external line has only one, we get:

2Fi + Fe = 2n. (22.7)

Similarly, each vertex is the end of a photon line and hence a number of the photon line
ends equals n
2Bi + Be = n. (22.8)
To substitute (22.7), (22.8) into (22.6) leads to the encouraging result:
3
D = 4 − Fe − Be . (22.9)
2
The developed approach to the evaluation of the divergences order for the integrals is,
however, applicable only to matrix elements involving integration with respect to a single
four-dimensional momentum. In case of a higher-multiplicity integral the calculation of
degrees is insufficient to determine the integral divergence or convergence. This is illustrated
in Fig. 22.1. For both diagrams pictured in this figure we have D = −1. Nevertheless, the
diagram a is finite, while the diagram b is divergent. Because of this, the inequalities of
(22.4), (22.5) are directly applicable, as a divergence or convergence criterion, exclusively
to the diagrams associated with a single integration. Provided a diagram includes l internal
lines, the conditions of (22.4) and (22.5) are valid only in the case when a subintegration
result over an arbitrary selected (l − 1)th internal four-dimensional momentum is finite. In
other words, in the case of a divergent diagram its divergence should appear at the final
integration with respect to the four-dimensional momentum. Divergent diagrams featuring

© 2011 by Taylor and Francis Group, LLC

 590 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

(a) (b)

FIGURE 22.1
The diagrams (a) and (b) describing the process e− e+ → γγ in the second and fourth order.

this property are referred to as primitively divergent diagrams (PDD). Such diagrams may
be characterized as those leading to the finite matrix element if any of the internal lines
of a diagram is broken and replaced by two external lines. Consideration of the PDD is
of particular importance because they give the basic RC enabling the calculation of all
other RC by the appropriate embedding. This procedure has been demonstrated during
the calculation of the RC for electron scattering by an external field. At the same time,
since the RC are the source of divergences, for analysis of divergences in the theory it is
sufficient to take into consideration only the PDD and to study the properties of matrix
elements to be associated with them.
Let us determine a degree of the PDD. By definition, the integrals associated with such
diagrams are convergent when one of the integration variables is fixed, and hence divergent
integrals appear in the case when in the integrand the numerator degree 4(Fi + Bi − n + 1)
is equal to or higher the denominator degree 2Bi + Fi , that is,

4(Fi + Bi − n + 1) ≥ 2Bi + Fi . (22.10)

With allowance made for (22.7) and (22.8) we arrive at the condition:
3
D = 4 − Fe − Be ≥ 0. (22.11)
2
Inasmuch as D does not contain n, the divergence degree of the PDD does not depend on
the diagram vertices number. It means, that there are a finite number of the PDD in the
QED. So, if by application of renormalization we manage to do all matrix elements of the
PDD finite, then we lead to the finite (renormalizable) theory.
Using the formula (22.11), we define all kinds of the PDD:

(i) Fe = Be = 0, D = 4—divergence of the fourth order;

(ii) Fe = 0, Be = 4, D = 0—logarithmic divergence;
(iii) Fe = 0, Be = 3, D = 1—linear divergence;
(iv) Fe = 0, Be = 1, D = 3—cubic divergence;
(v) Fe = 0, Be = 2, D = 2—squared divergence;
(vi) Fe = 2, Be = 0, D = 1—linear divergence;
(vii) Fe = 2, Be = 1, D = 0—logarithmic divergence.

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 591

The diagram of Fig. 22.2 describing the transition vacuum→ vacuum corresponds to
the first case. This graph is eliminated from consideration on the basis of the reasons to

FIGURE 22.2
The Feynman diagram associated with the transition vacuum→vacuum.

have been brought forward in Section 20.3. An analysis of divergences in any QFT should
always be performed with due regard for the symmetry properties. Due to invariance of a
theory, some PDD are actually finite or zero. In the QED light-by-light scattering diagrams
(Fe = 0, Be = 4; Fig. 22.3), triangular diagrams (Fe = 0, Be = 3; Fig. 22.4), and
“tailless fish” diagrams (Fe = 0, Be = 1; Fig. 22.5) are referred to such PDD. In case

p − k3 − k4 p − k 4 − k3
k1 k3 k1 k4 k1 k3

p − k1 p − k3 p − k1 p − k4

k2 k4 k2 k3 k2 k4
p p

(a) (b) (c)

p − k2 − k3 p − k 3 − k2
k1 k4 k1 k3 k1 k4

p − k1 p − k2

k2 k3 k2 k4 k2 k3
p p

(d) (e) (f)

FIGURE 22.3
The Feynman diagrams (a)–(f) for the light-by-light scattering.

of the light-by-light scattering Eq. (22.11) gives D = 0 but in fact there is no logarithmic
divergence at all, and the corresponding matrix element is finite. Let us demonstrate that
this circumstance is caused by the gauge invariance of the QED.
Since the amplitude of the light-by-light scattering A(k1 , k2 , k3 , k4 ) is logarithmically
diverging, then in order to get its finite part AR (k1 , k2 , k3 , k4 ) one sufficiently subtracts
from AR (k1 , k2 , k3 , k4 ) only the first term of the expansion into the Taylor series, that is

AR (k1 , k2 , k3 , k4 ) = A(k1 , k2 , k3 , k4 ) − A(0, 0, 0, 0). (22.12)

© 2011 by Taylor and Francis Group, LLC

 592 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

FIGURE 22.4
The triangular diagrams.

FIGURE 22.5
The “tailless fish” diagrams.

Owing to the gauge invariance observables depend on only the electromagnetic field tensor

Fµν (k) = kµ Aν (k) − kν Aµ (k),

rather than on the potential Aµ (k). By this reason, observables must turn into zero at
kµ = 0, and, hence, A(0, 0, 0, 0) = 0. This result is confirmed by direct calculations [96]
yielding the following expression for the cross section in the ultrarelativistic limit:

α2 r02  m 2  ω 4
σ= ln . (22.13)
π2 ω m
Note that the four-photon interaction has no classical analog. It results from quantum
fluctuations of the virtual pairs of charged particles. When the photons energy becomes
equal to 2m a real pair is produced, and the cross section has the order of several µb.
Despite the fact that by modern standards the cross section is so great, this effect is hardly
observable due to a significant background. For example, in the interstellar space the
radiation scattering by galactic hydrogen (and dust) is much greater. But as has been
already mentioned in Section 19.6, another electromagnetic process with Be = 4, Fe = 0
is our object of experimental observation, light scattering from an electrostatic field in the
second order of the perturbation theory (Delbrück scattering).
As for the diagrams with Fe = 0, Be = 3 and Fe = 0, Be = 1, their matrix elements
vanish due to the Furry theorem. Recall, that this theorem is a consequence of the QED
invariance with respect to the charge conjugation operation.
Now we proceed to the discussion of the case Fe = 0, Be = 2. The only primitively-
divergent photon self-energy diagram (PSED) is shown in Fig. 22.11. Earlier it has been
demonstrated that, owing to the gauge invariance, in fact this diagram has merely a loga-
rithmic rather than quadratic divergence.
At Fe = 2, Be = 0 we have a single primitively-divergent electron self-energy diagram
(ESED) (see, Fig. 21.5(a)). According to Eq. (22.11) this diagram should have the linear
divergence. However, the divergence proves to be logarithmic only (see, Section 21.3). It
should be noted that in case of the odd dimensionality of D, as a rule, the closest lower
even-order divergence is realized.

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 593

The last case with Fe = 2, Be = 1 corresponds to a vertex diagram (VD). The simplest
example of such a diagram is represented in Fig. 22.6(a). In the given case we have the

(a) (b)

(c)

FIGURE 22.6
An example of vertex diagrams (a)–(c).

logarithmic divergence. But as shown in Section 21.6, these divergences are compensated by
those existing in the self-energy electron and photon diagrams. Note that this reduction is
due to the gauge invariance of the QED as in the case of a successive renormalization, only
the gauge-invariant diagram sets should be always considered. The number of primitively-
divergent VDs is infinite. Some of these diagrams are shown in Fig. 22.6(b),(c).
In this way the foregoing results may be defined as follows: any primitively-divergent
diagram should be the 5−, 6−, or 7−type diagram.

22.2 Effective elements of Feynman diagrams

Such notions as compact, noncompact, reducible, and irreducible diagrams are introduced
for the self-energy and vertex diagrams.
An ESED is compact when it could not be subdivided into parts connected only by one
electron line, being noncompact otherwise. In Fig. 21.5 all the diagrams are compact while

© 2011 by Taylor and Francis Group, LLC

 594 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

the a-diagram is noncompact.

In a similar way, it is impossible to divide a compact PSED into parts connected by a

single photon line, whereas the situation is opposite in case of a noncompact PSED. In Fig.
22.7 the b-diagram is noncompact and the remaining ones are compact. A VD is called

(a)

(b)

(c)

(d)

FIGURE 22.7
The PSED diagrams (a)–(d).

compact when it could not be subdivided into parts connected with each other only by an
electron or photon line, and otherwise a VD is noncompact. All VDs pictured in Fig. 22.6
are compact. An example of a noncompact VD is given in Fig. 22.8. If all the self-energy
and vertex parts are excluded from some diagram G, we have a reduced diagram referred to
as a skeleton of the diagram G. A diagram coincident with its skeleton is called irreducible.

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 595

FIGURE 22.8
An example of the noncompact VD diagram.

The ESED and PSED of the second order shown in Figs. 21.5(a) and 22.7(a) are the single
examples of the irreducible self-energy parts. All higher-order self-energy parts are reducible
and may be obtained by inserting the PSED, ESED, and VD into the lines and vertices
of the diagrams displayed in Figs. 21.5(a) and 22.7(a). There are an infinite number of
irreducible VDs. Among the diagrams pictured in Fig. 22.6 the first two diagrams are
irreducible.
Let us consider an arbitrary diagram containing an internal electron line. In the higher-
order approximations this diagram is completed by diagrams with various ESEDs included
in the line. Then the initial electron line, together with all the included ESEDs, is graphi-
cally represented as a heavy solid line called the effective internal electron line. The quantity
associated with this line is called the electron Green function G(e) (p) (for the sake of simplic-
ity, hereafter we drop spinor indices). In Fig. 22.9 an infinite sum of diagrams corresponding
to the effective electron line G(e) (p) is represented. It is obvious that the following relation

G(e) (p) Sc (p) Sc (p) Sc (p) Sc (p) Sc (p) Sc (p)

= + + + ...

FIGURE 22.9
The diagrams corresponding the effective electron line.

takes place

G(e) (p) S c (p) S c (p) Σ∗ (p) S c (p) S c (p) Σ∗ (p) S c (p) Σ∗ (p) S c (p)
= + + + ... =
i i i i i i i i i i
S c (p) S c (p) Σ(p) S c (p)
= + . (22.14)
i i i i
A graphical linkage between Σ∗ (p)/i and Σ(p)/i is displayed in Fig. 22.10, where Σ∗ (p)/i
denotes a sum of quantities associated with only compact ESEDs, and an operator Σ(p)/i
describes a sum of all possible compact and noncompact ESEDs with the given momentum
p (recall, the appearance of the factor i in Σ∗ (p) is produced by the Wick rotation). From
Fig. 22.10 it is evident that
Σ(p) Σ∗ (p) Σ∗ (p) S c (p) Σ∗ (p) Σ∗ (p) S c (p) Σ∗ (p) S c (p) Σ∗ (p)
= + + + ... =
i i i i i i i i i i

© 2011 by Taylor and Francis Group, LLC

 596 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Σ(p) Σ∗ (p) Σ∗ (p) Σ∗ (p) Σ∗ (p) Σ∗ (p) Σ∗ (p)

= + + + ···
Sc (p) Sc (p) Sc (p)

FIGURE 22.10
Σ(p) as a sum of Σ∗ (p).

Σ∗ (p) Σ∗ (p)
= [1 + Σ∗ (p)S c (p) + Σ∗ (p)S c (p)Σ∗ (p)S c (p) + ...] = [1 − S c (p)Σ∗ (p)]−1 .
i i
(22.15)
To insert this expression into (22.14) leads to the result

G(e) (p) = S c (p) + S c (p)Σ∗ (p)[1 − S c (p)Σ∗ (p)]−1 S c (p) = [1 − S c (p)Σ∗ (p)]−1 S c (p), (22.16)

from which the Dyson equation for the electron Green function follows

G(e) (p) = S c (p) + S c (p)Σ∗ (p)G(e) (p). (22.17)

It is obvious that the graphical representation of this equation has the form pictured in Fig.
22.11. And by the presentation of G(e) (p) in the form of (22.14) it is assumed that from the

= +
x y x y x y
x′ y′

FIGURE 22.11
The graphical representation of Eq. (22.17) in the coordinate space.

Feynman graphs we can select more simple blocks to be calculated using the general rules
of the diagram technique. Combining these blocks with each other, we can subsequently
obtain the proper expression for the graphs as a whole. The possibility of such a selection
is an important feature of the diagram technique. This is connected with the fact that a
numerical factor in the diagram is independent of its order. The same feature makes it
possible to use the function G(e) to simplify the calculation of the RC for the amplitudes
of different processes. Instead of considering the graphs with the necessary corrections to
the internal electron lines every time anew, we have a possibility to replace these lines by
the heavy ones, that is, we simply correlate them with the G(e) propagators in the required
approximation.
In a similar way one can determine an effective internal photon line as an internal photon
line with all possible PSEDs included in it. In a figure the internal effective line will be
displayed by the heavy waved line. The quantity associated with this line is called the
photon Green function G(γ) (k) (for simplicity, hereafter, we omit tensor indices). The
graphic presentation of G(γ) (k) is given in Fig. 22.12. From this figure it is obvious that
the formula:
G(γ) (k) Dc (k) Dc (k) Π(k) Dc (k)
= + (22.18)
−i −i −i i −i

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 597

G(γ) (k) Dc (k) Dc (k) Dc (k) Dc (k) Dc (k) Dc (k)

= + + + ...

FIGURE 22.12
The graphic presentation of G(γ) (k).

takes place.
Introducing a polarization operator Π∗ (k) to be a sum of all compact PSEDs with the
given momentum k, Π(k) can graphically be displayed by Fig. 22.13. From this figure it

Π(k) Π∗ (k) Π∗ (k) Π∗ (k) Π∗ (k) Π∗ (k) Π∗ (k)

= + + +...
Dc (k) Dc (k) Dc (k)

FIGURE 22.13
Π(k) as a sum of Π∗ (k).

follows that
Π(k) = Π∗ (k)[1 − Dc (k)Π∗ (k)]−1 . (22.19)
Substitution of the obtained expression into (22.18) results in the Dyson equation for the
photon Green function

G(γ) (k) = Dc (k) + Dc (k)Π∗ (k)G(γ) (k), (22.20)

whose graphical form is given in Fig. 22.14. Note, the relations (22.17) and (22.20) are the

Π∗

= +
x y x y x x′ y′ y

FIGURE 22.14
The graphical representation of Eq. (22.20).

integral equations. For example, in the coordinate representation the relation (22.17) looks
like:
Z Z
G(e) (x, y) = S c (x − y) + d4 x′ d4 y ′ S c (x − x′ )Σ∗ (x′ , y ′ )G(e) (y ′ , y). (22.21)

Next, we define the effective external electron and photon lines as external electron and
photon lines with all possible self-energy diagrams included in these lines. Their graphic
presentation is shown in Figs. 22.15 and 22.16. According to these figures we can write

© 2011 by Taylor and Francis Group, LLC

 598 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Σ(p)
Ψ(p) Ψ(p) Ψ(p) Sc (p)
= +

FIGURE 22.15
The effective external electron line.
Π(k)
A(k) A(k) A(k) Dc (k)
= +

FIGURE 22.16
The effective external photon line.

ψ(p) = ψ(p) + S c (p)Σ(p)ψ(p), (22.22)

c
Aµ (k) = Aµ (k) + Dµν (k)Πνσ (k)Aσ (k). (22.23)
Analogously, an effective line for an external electromagnetic field is defined (in this figure
it is displayed by the heavy waved line with the shaded circle at the end)

Aeµ (k) = Aeµ (k) + Dµν

c
(k)Πνσ (k)Aeσ (k). (22.24)

Prior to the definition of the effective vertex, note that since noncompact VDs represent
collections both of compact VDs and of the effective electron and photon lines, then using the
effective electron and photon lines allows us to omit noncompact VDs from consideration.
In this case a sum of the quantities corresponding to all possible compact VDs with the
specified momenta p1 , p2 , and k on the electron and photon lines represents the vertex
function Γµ,αβ (p1 , p2 ; k) (µ is a vector index while α and β are spinor indices). In diagrams
the vertex function is pictured by a heavy point, known as the effective vertex. Since p1 , p2 ,
and k are related by the four-dimensional momentum conservation law, in what follows k
will be omitted from the argument of the vertex function. It is convenient to extract the
ordinary vertex from the vertex function

Γµ (p1 , p2 ) = γµ + Λµ (p1 , p2 ), (22.25)

where we have ignored spinor indices. Using the effective lines and effective vertices, we
can replace complex diagrams encountered in calculations of the RC by the effective ones,
the skeleton diagrams, representing irreducible diagrams with the effective lines (instead of
the ordinary lines) and effective vertices (instead of the ordinary vertices). To calculate the
required matrix element, we have to know the vertex function and Green functions. The
Green functions may be found provided the mass and polarization operators are known.
Let us consider the way of finding these quantities.
Since only one irreducible ESED and a single irreducible PSED exist, the mass and
polarization operators have one skeleton diagram each (see, Fig. 22.17). It should be noted
that all the reducible self-energy parts are exhausted by the insertion of the vertex parts
into one vertex rather than into both vertices. Otherwise, the same self-energy part would
be taken into consideration more than once. The explanation is as follows. Consider the
internal structure of the graph Σ∗ in the coordinate space (Fig. 22.18). By the inclusion of
all self-energy parts into the internal electron and photon lines of the irreducible diagram

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 599

G(γ) (k) G(e) (p)

k p

γµ Γν (p, p − k) γµ Γν (p, p − k)

p−k p−k

G(e) (p − k) G(e) (p − k)
Σ(p) Πµν (k)

FIGURE 22.17
The skeleton diagrams for the mass and polarization operators.

= + y
x y x y x
τ

(a)

=
x y x y

(b)

FIGURE 22.18
The internal structure of the graph Σ∗ in the coordinate space in (a) and (b).

we have two heavy lines (solid and waved) outgoing from the point x. As this takes place,
two variants are possible: both branches are going, without meeting, to the point y (the
first term in the right-hand side of Fig. 22.18(a)) or only to some points z and τ (the second
term in the right-hand side of Fig. 22.18(a)). Then the remaining part of the graph in the
second term represents what we call the vertex part Λµ (p1 , p2 ). A sum of both terms on
the right-hand side of Fig. 22.18(a) just gives the desired result (Fig. 22.18(b)). Applying
the rules for writing the matrix elements to the skeleton diagrams associated with the mass
and polarization operators, we get
Z
∗ e2
Σ (p) = − γ µ G(e) (p − k)Γν (p, p − k)G(γ) 4
µν (k)d k, (22.26)
(2π)4

Z
e2
Π∗µν (p) = γµ G(e) (p)Γν (p, p − k)G(e) (p − k)d4 p. (22.27)
(2π)4

To insert the expressions obtained to the Dyson equation gives the integral equation for

© 2011 by Taylor and Francis Group, LLC

 600 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

determining the electron and photon Green functions

Z
(e) c e2
G (p) = S (p) − S (p) γ µ G(e) (p − k)Γν (p, p − k)G(γ)
c
µν (k)G
(e)
(p)d4 k, (22.28)
(2π)4
Z
e2 h i
G(γ)
µν (k) = D c
µν (k) + 4
D c
µλ (k) Sp γ λ G(e) (p)Γσ (p, p − k)G(e) (p − k) G(γ) 4
σν (k)d p.
(2π)
(22.29)
Our next step is finding the vertex function Γµ (p1 , p2 ). But it turns out that this func-
tion could not be represented in the closed form. Indeed, the number of irreducible VDs
is countless. So, to derive an integral equation for Γµ (p1 , p2 ), we have to draw all the irre-
ducible VDs and for each of them replace ordinary lines and vertices by the effective ones.
The whole series for Γµ is generated by just this very way
Z
e2
Γµ (p1 , p2 ) = γµ − Γσ (p1 , p1 − k)G(e) (p1 − k)Γµ (p1 − k, p2 − k)×
(2π)4
×G(e) (p2 − k)Γν (p2 − k, p2 )G(γ) 4
σν (k)d k + ... (22.30)
In such a manner an equation for the vertex function and hence equations for the electron
and photon Green functions could not be represented in the closed form using only dif-
ferentiation and integration operations with respect to the space-time coordinates. At the
same time, equations for these three functions may be obtained in the closed form when
we use the functional differentiation operation [78,97]. Since the main question that we are
interested in here is that of the existence of such equations rather than the technique of
their derivation, then we are not going to reproduce their deduction and refer an interested
reader to the monograph [98].
In the presence of an external electromagnetic field the internal electron line includes,
apart from the ESED, particular parts associated with the vacuum polarization by the
external field, that is, the parts joined to the remaining graph by a photon line. In this case
Fig. 22.11 should correspondingly be supplemented and we come to Fig. 22.19. Then the

Σ∗
= + +
x y x y x y x ξ y
x′ y′

FIGURE 22.19
G(e) (x, y) in the presence of an external electromagnetic field.

integral equation (22.21) is replaced by:

Z Z
(e) c 4 ′
G (x, y) = S (x − y) + d x d4 y ′ S c (x − x′ )Σ∗ (x′ , y ′ )G(e) (y ′ , y)+
Z
+eγ µ d4 zS c(x − z)Aeµ (z)G(e) (z, y). (22.31)

Here Aeµ (x) is the sum of the original external field Aeµ (x) and the RC to it δAeµ (x). In the
coordinate representation the operator Σ∗ (x, y) is given by:
Z Z
Σ∗ (x, y) = −e2 γ µ d4 z d4 τ G(e) (x, z)Γµ (z, y; τ )G(γ) (x, τ ), (22.32)

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 601

where
Γµ (z, y; τ ) = γµ δ (4) (z − τ )δ (4) (y − τ ) + Λµ (z, y; τ ). (22.33)
Acting by the operator i∂ˆ − m on Eq. (22.31), we arrive at the Schwinger equation [78]:
Z
[i∂ˆ − eÂe (x) − m]G(e) (x, y) + d4 y ′ Σ∗ (x, y ′ )G(e) (y ′ , y) = δ (4) (x − y), (22.34)

where we have allowed for

(i∂ˆ − m)S c (x − y) = δ (4) (x − y).

One can pass from the equation for the Green function to that for the electron wavefunction
in an external field with the RC inclusion. For this purpose it is enough to omit the δ-
function in the right-hand side of Eq. (22.34) to give:
Z
ˆ
[i∂ − e (x) − m]ψ(x) + d4 y ′ Σ∗ (x, y ′ )ψ(y ′ ) = 0.
e (22.35)

It should be recorded that the majority of the above manipulations are concerned with
infinite quantities. Obviously, a mathematical rigor necessitates regularization of all the
quantities prior to studying their properties and before the derivation of the equations they
are satisfying. With the use of the Pauli-Villars regularization the quantities, which were
finite without the regularization, remain unchanged, while, for example, linearly divergent
quantities change to the finite linear functions of the regularizing masses.

22.3 Ward identity

As follows from Eqs. (22.26) and (22.27), relations between the vertex function and the
electron as well as photon Green functions are complex in nature. But when the arguments
of the vertex function are coincident, this relation becomes quite simple. We can find it for
this particular case.
Having differentiated an identity

S c (p)[S c (p)]−1 = 1,

with respect to pµ , we obtain

∂S c (p) c ∂[S c (p)]−1 c

= −S (p) S (p). (22.36)
∂pµ ∂pµ

Substitution of the value

[S c (p)]−1 = m − p̂
into Eq. (22.36) results in
∂S c (p)
= S c (p)γµ S c (p). (22.37)
∂pµ
The right-hand side of the derived relation, to within a constant factor, is equal to the
matrix element associated with a part of the diagram, where an external photon line is
joined to the internal electron line (Fig. 22.20). Thus, formal differentiation of the electron

© 2011 by Taylor and Francis Group, LLC

 602 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

k=0

γµ
p p

FIGURE 22.20
The graphical representation of the right-hand side of Eq. (22.37).

propagator with respect to pµ corresponds to the situation when a photon line with zero
momentum is joined to the electron line.
Next, we consider a diagram W and insert an irreducible ESED in its internal electron
line, that is, carry out the replacement:
S c (p) → S c (p)Σ(W, p)S c (p). (22.38)
Hereinafter, we introduce the diagram symbol into the argument of the self-energy operators
to emphasize that now they are the components of the diagram W . It is clear that Σ(W, p)
differs from the operator Σ(2) (p)
Z
e2
Σ(2) (p) = − d4 kγ ν S c (p − k)γν Dc (k) (22.39)
(2π)4
by the factor i. Differentiating (22.39) and making use of the relation (22.37), we arrive at
the equation
Z
∂Σ(W, p) ie2
=− d4 kγ ν S c (p − k)γµ S c (p − k)γν Dc (k). (22.40)
∂pµ (2π)4
The expression in the right-hand side of Eq. (22.39) is equal to the lowest (third) order ver-
tex operator at the zero-momentum transfer Λ(3) (p, p). So, we can state that the performed
substitution (22.38) and subsequent differentiation of the obtained diagram with respect to
pµ is equivalent to the inclusion of the vertex into the internal electron line or replacement:

γµ → Λ(3) (p, p)
at the vertex of the initial diagram. It is seen that from the compact ESED associated
with the operator Σ′ (W, p) we can obtain a compact VD by joining the external photon line
to some internal electron line. Formally, an insertion of the zero-momentum photon line
into the electron line implies taking a derivative with respect to the external momentum
from the propagator associated with this electron line. Consequently, due to differentiation
of the operator Σ′ (W, p) with respect to pµ , a photon line is joined to each electron line
transferring the momentum p. Provided the ESED contains a closed electron line as in
Fig. 22.21, in case of the photon line joining to the electron loop we obtain a collection
of diagrams whose sum is equal to zero in accordance with the Furry theorem. Because of
this, the electron loop takes no part in the differentiation with respect to the momentum
of an external electron line.
If the sum of all the compact ESED’s associated with the operator Σ∗ (W, p) ≡ Σ∗ (p) is
differentiated with respect to pµ , then we shall lead to the identity
∂Σ∗ (p)
= Λµ (p, p), (22.41)
∂pµ

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 603

FIGURE 22.21
The ESED containing a closed electron line.

where Λµ (p, p) denotes the set of all the compact VDs, Ward has obtained for the first time
[99]. Using (22.17), one can write

[G(e) (p)]−1 = [S c (p)]−1 − Σ∗ (p) = [m − p̂ − Σ∗ (p)] . (22.42)

Then, differentiating (22.42) with respect to pµ yields another writing of the Ward identity:

∂[G(e) (p)]−1 ∂
= µ [m − p̂ − Σ∗ (p)] = −[γµ + Λµ (p, p)] = −Γµ (p, p). (22.43)
∂pµ ∂p

Let us show that the Ward identity is a consequence of the gauge invariance of the QED.
Under an infinitesimal gauge transformation we have

∂
Aµ (x) → Aµ (x) − δφ(x) = Aµ (x) − δAµ (x), (22.44)
∂xµ
)
ψ(x) → ψ(x) exp[−ieδφ(x)] ≈ [1 − ieδφ(x)]ψ(x),
(22.45)
ψ(x) → ψ(x) exp[ieδφ(x)] ≈ [1 + ieδφ(x)]ψ(x).
In this case the change-over of the electron Green function is

δG(e) (x, y) = ieG(e) (x − y)[δφ(y) − δφ(x)]. (22.46)

Emphasize, the gauge transformation (22.45) violates the space-time homogeneity and the
function δG(e) depends on the arguments x and y individually rather than on the difference
x − y. Introducing the Fourier representation for δG(e) (x, y) and δφ(x)
Z Z
δG (x, y) = d p1 d4 p2 δG(e) (p1 , p2 )ei(p1 y−p2 x) ,
(e) 4
(22.47)

Z
δφ(x) = d4 pδφ(p)e−ipx (22.48)

and substituting them into Eq. (22.46), we gain

δG(e) (p + q, p) = ieδφ(q)[G(e) (p) − G(e) (p + q)]. (22.49)

Because the theory is gauge invariant, then the obtained value of the electron propaga-
tor must coincide with the value an infinitesimal external field δAµ (x) generates. In the
momentum representation δAµ (x) takes the form:

δAµ (q) = −iqµ δφ(q). (22.50)

© 2011 by Taylor and Francis Group, LLC

 604 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

p p+q

FIGURE 22.22
The graphical representation of δG(e) (p1 , p2 ).

When one is limited by the first order of smallness in terms of δφ, then the quantity
δG(e) (p1 , p2 ) is graphically represented by the skeleton diagram of Fig. 22.22. However, in
this figure the effective external photon line associated with the factor
c
δAµ (q) + Dµσ (q)Πσν (q)δAν (q) (22.51)
must be changed by the ordinary line as the second term of (22.51) vanishes. Then, from
Fig. 22.22 it follows
δG(e) (p + q, p) = eG(e) (p + q)Γµ (p + q, p; q)G(e) (p)δAµ (q). (22.52)
Inserting δAµ (q) defined by (22.50) into Eq. (22.52) and comparing the obtained expression
with (22.49), we arrive at the relation:
G(e) (p + q) − G(e) (p) = qµ G(e) (p + q)Γµ (p + q, p; q)G(e) (p), (22.53)
which may also be represented in the form:
[G(e) (p + q)]−1 − [G(e) (p)]−1 = −qµ Γµ (p + q, p; q). (22.54)
Expanding both sides of the expression (22.54) into a series in the infinitesimal quantity qµ
and transiting to the limit q → 0, we again get the Ward identity in the form (22.43)
∂
[G(e) (p)]−1 = −Γµ (p, p), (22.55)
∂pµ
where Γ(p, p; 0) ≡ Γ(p, p).

22.4 Extraction of divergences

There are three kinds of the PDDs in the QED, namely, the self-energy and vertex dia-
grams. In Sections 21.3–21.5 we have already investigated the extraction of the divergences
in the lowest orders of the perturbation theory for the matrix elements corresponding to
these diagrams. In the process, we have used the invariant method by Dyson [4] based on
expanding the integrand associated with a divergent diagram into the Taylor series in terms
of the external momentum (see Eq. (21.62)).
We have found that the ESED of the second order is described by the expression:
 (2) 
∂Σ (p) (2)
Σ(2) (p) = Σ(2) (mR ) + (p̂ − mR ) + ΣC (p), (22.56)
∂ p̂ p̂=mR

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 605

where the first two terms are diverging linearly and logarithmically, respectively. As in this
case the second-order diagram is simultaneously an irreducible diagram, Eq. (22.56) is valid
for the divergences separation of the irreducible ESED, actually giving a contribution of all
the irreducible ESEDs. Introducing the designations
 (2) 
∂Σ (p)
Σ(2) (mR ) = An , = Bn,
∂ p̂ p̂=mR

we rewrite (22.56) in the view:

Σn (p) = An + B n (p̂ − mR ) + ΣnC (p), (22.57)

where ΣnC (p) vanishes for a free electron. The mass renormalization is determined in such
a way that the free electron mass (p̂ = mR ) is exactly equal to mR , that is,

ψ(p)Σn (p)ψ(p) = 0 at p̂ = mR , (22.58)

to give
An = −iδm. (22.59)
2
The polarization tensor Π(W, q ) appearing after replacement of the internal photon line:
c
Dµν (q) → Dµλc (q)Πλσ (W, q)Dνσc (q), (22.60)

is distinguished from Π(q 2 ) by the constant factor i. This is easily seen when we take
into account the explicit form of the operator Πλσ (q) and use the Feynman gauge for the
propagator of an electromagnetic field. Investigating the PSED of the second order, we
have established that splitting the polarization tensor into the finite and divergent parts
has the form:
(2)
Π(2) (q 2 ) = Π(2) (0) + ΠC (q 2 ), (22.61)
(2)
where Π(2) (0) contains the logarithmic divergence and the finite quantity ΠC (q 2 ) equals
zero at q 2 = 0. Since the diagram under consideration is the single irreducible PSED, then
the contribution of all irreducible PSEDs into Π(q 2 ) is given by the expression:

Πn (q 2 ) = C n + ΠnC (q 2 ), (22.62)

where we have taken the designation C n = Πn (0).

In the case of the VDs the situation is somewhat different. Here, apart from the third-
order diagram considered in Section 21.5, we have an infinite set of irreducible diagrams.
In this way the separation procedure for the divergences should be extended to all the
irreducible VDs.
A matrix element corresponding to some irreducible VD could be represented by an
integral Z
Λµ (p1 , p2 ) = d4 kRµVi (p1 , p2 , k).
Vi
(22.63)

As the contribution of the vertex parts has a logarithmic divergence, in the Taylor expansion
for the integrand (22.63) the terms with the derivatives are not required. As a result we
get:
ΛVµi (p1 , p2 ) = ΛVµi (p1 , p1 ) + ΛVCµ
i
(p1 , p2 ) (22.64)
where ΛVµi (p1 , p1 ) is logarithmically diverging and the finite quantity ΛVCµ
i
(p1 , p2 ) turns into
zero at p1 = p2 . The relativistic invariance demands

ΛVµi (p1 , p1 ) = L(Vi )γµ . (22.65)

© 2011 by Taylor and Francis Group, LLC

 606 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

At p̂1 = p̂2 = mR and p1 = p2 the operator ΛVCµ i

(p1 , p2 ) is reduced to the matrix γµ
multiplied by a constant value. This value may be included into the term L(Vi )γµ . Then
the operator ΛVCµi
(p1 , p2 ) will be equal to zero at p̂1 = p̂2 = mR and p1 = p2 , rather
than at p1 = p2 . By the way, such a choice is more justified from the physical viewpoint.
Indeed, the quantity ΛVCµ i
(p1 , p2 ) makes no contribution when the electromagnetic potential
is constant, because such a potential is transferring zero momentum. Very well, this is clear
as a constant electromagnetic potential may be eliminated any time with the use of the
gauge transformation.
Having summarized (22.64) over all the irreducible VDs, we obtain the required relation:
X
ΛVµi (p1 , p2 ) = Ln γµ + ΛnCµ (p1 , p2 ), (22.66)
Vi

where X
Ln = Λ(Vi ). (22.67)
Vi

To demonstrate a possibility for renormalization of a theory as a whole, we have to

consider diagrams of all types and all orders, both reducible and irreducible. Thus, the
next step necessitates demonstration of the divergences separation in the case of reducible
diagrams. It should be noted that separation of the matrix elements into finite and infinite
parts must be performed in the relativistically invariant and unambiguous way. Logically,
it may be assumed that the formulae providing the basis for separation of divergences in
the expressions associated with reducible diagrams are of the same form as for irreducible
diagrams. This statement may be supported if we consider the procedure of separating the
finite and infinite expressions in the general form for the ESEDs.
Let us look at a reducible electron self-energy part Σr (W, p), which is a compound element
of some diagram. It is clear on relativistic invariance grounds that this operator must have
the form:
Σr (W, p) = A′ + Bµ′ pµ + ΣrC (p), (22.68)
where the matrix operators A′ and Bµ′ are diverging linearly and logarithmically, respec-
tively. Inasmuch as Σr (W, p) represents a 4 × 4- matrix, then it may be expanded in terms
of 16 linearly independent matrices Γj (see, Eq. (15.11)):
1
Σr (W, p) = ΣS (p) + ΣV µ (p)γ µ + ΣT µν (p)σ µν + ΣP (p)γ5 + ΣAµ (p)γ µ γ5 . (22.69)
2
The parity conservation provides

ΣP (p) = ΣAµ (p) = 0. (22.70)

Because the vector pµ is the single one being in our disposal, then, due to the relativistic
invariance, the vector and tensor coefficients in (22.69) are of the form:

ΣV µ = pµ ΣV (p2 ), ΣT µν = pµ pν ΣT (p2 ). (22.71)

Note, the relativistic invariance also demands ΣS (p) ≡ ΣS (p2 ). Thus, the expression (22.69)
is represented in the form:

Σr (W, p) = ΣS (p2 ) + γ µ pµ ΣV (p2 ). (22.72)

Comparing the expressions (22.69) and (22.72), we find

A′ = ΣS (0), Bµ′ = γµ ΣV (0) = Bγµ , (22.73)

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 607

to give
Σr (W, p) = Ar + B r (p̂ − mR ) + ΣrC (p),
where Ar = A′ +B r mR . So, the contribution of the reducible ESEDs into the mass operator
X
Σ(Wi , p) = Σ(p)
i

has the same structure as in the case of the irreducible ESEDs, and we have finally

Σ(p) = A + B(p̂ − mR ) + ΣC (p). (22.74)

In a similar manner, one could show that the contributions of the reducible diagrams into
the total operators
X X
Π(Wi , q 2 ) = Π(q 2 ), Λµ (p1 , p2 , Vi ) = Λµ (p1 , p2 ),
i i

are governed by the expressions having the same form as in the case of irreducible diagrams.
All this results in
Π(q 2 ) = C + ΠC (q 2 ), (22.75)
Λµ (p1 , p2 ) = Lγµ + ΛCµ (p1 , p2 ), (22.76)
where C and L are logarithmically divergent constants. It should be stressed that now all
the constants entering into Eqs. (22.7), (22.75) are real. The relations (22.74)–(22.76) form
the basis for the divergences separation in expressions associated both with the irreducible
and reducible divergent diagrams. Note, the relation:

B = L, (22.77)

which connects the divergent quantities contained in Eqs. (22.74) and (22.76), is the direct
consequence of the Ward identity. This is a proof that in the QED in all orders of the
perturbation theory the vertex divergence is canceled with the divergence that remains in
the electron self-energy after the electron mass renormalization.
But we face unpleasant surprises under more detailed analysis of the divergences in re-
ducible diagrams. It appears that the G diagrams, resultant from the irreducible diagram
G0 by insertion of the PSEDs, ESEDs, and VDs into various lines and vertices of G0 , may
be subdivided into following three classes: (i) diagrams with absolutely non-overlapping
insertions; (ii) diagrams where the insertions are totally contained within each other; and
(iii) diagrams with overlapping insertions. The first two classes of divergences may be
separated by the invariant Dyson method (see, Eq. (21.62)). When the insertions are non-
overlapping, separation is performed independently for each insertion. Divergences of the
second class are separated step-by-step, beginning from divergences of the inner insertion
and ending with the divergence of the diagram G as a whole. To illustrate, in the case of
a diagram given in Fig. 22.23 the divergences associated with integration with respect to
virtual momenta are separated in the following sequence: k1 , k2 , and k3 . A combination
of divergences of the first two types is demonstrated in Fig. 22.24. Here the sequence of
separation is as follows: first we extract the divergence due to integration over k1 , then do
divergence of the self-energy photon part in line k2 , and finally do divergence associated
with integration over k3 .
The method of Dyson is inapplicable for overlapping insertions because in this case the
integrals are divergent with respect to more than one variable. They are divergent even if
integration is over one of the variables, whereas all others are fixed. As an example, we

© 2011 by Taylor and Francis Group, LLC

 608 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

k1 k2 k3

FIGURE 22.23
An example of the diagram with overlapping insertions.

k2

k1

k2

FIGURE 22.24
An example of the diagram with divergences of the (i) and (ii) type.

k1 k2

p p − k1 p − k1 − k2 p − k2 p

FIGURE 22.25
The overlapping ESED.

consider an overlapping ESED (Fig. 22.25), which may be obtained by insertion of the VD
into any of the vertices of the second-order ESED. It is associated with the matrix element
Z Z
ie4 1 p̂ − k̂1 + m
Σ(4) (p) = 8
d4
k1 d4 k2 2 γ µ ×
(2π) k1 (p − k1 )2 − m2

p̂ − k̂1 − k̂2 + m 1 p̂ − k̂2 + m

×γ ν 2 2
γµ 2 γν . (22.78)
(p − k1 − k2 ) − m k2 (p − k2 )2 − m2

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 609

From (22.78) it is evident that the double integral over k1 and k2 is linearly diverging.
To fix one of variables does not save the situation. Fixing k2 (k1 ), we get the integral
logarithmically divergent with respect to k1 (k2 ).
In much the same manner, overlapping divergences are found on insertion of the VD into
the vertices of the PSED. An example of such a diagram, where the insertion of the VD
has been made into one of the vertices of the second-order PSED, is given in Fig. 22.26. In

FIGURE 22.26
The PSED obtained by the insertion of the VD.

the general case a contribution made by such reducible diagrams into Π∗ or Σ∗ represents
an integral that possesses the divergences corresponding to all the diagram construction
procedures possible with the use of the VD insertion into one or both vertices of the initial
irreducible diagram. The number of possible methods to construct a reducible diagram
from the irreducible components in such a way is growing with the diagram order. Since for
each construction method divergences are introduced independently, then investigating the
appeared divergences becomes very complicated. Note that, when we look for an adequate
approach to the divergence separation of the overlapping diagrams, our main difficulty is a
correct enumeration of all the divergent parts, enabling interpretation of subtraction as a
renormalization and leading to an unambiguous separation of the finite parts. A method
enabling unambiguous separation of covariant and absolutely convergent residual terms in
the case of the overlapping divergent parts has been proposed in Refs. [100]. With its help
one can deduce the formulae (22.74)–(22.76) in the case of any diagrams.

22.5 QED renormalizability

Renormalizability of the QED may be proved without going into the problem of overlap-
ping divergences. Again, the main hero of this history is the gauge invariance, or more
precisely, one of its consequences, the Ward identity. We shall give the proof of the QED
renormalizability following the papers by Ward [94].
First we show the way to obviate the problem of overlapping divergences in case of the
ESED. We take a look at the mass operator Σ(W1 , p) corresponding to the diagram W1
with overlapping divergences (see, Fig. 22.25). Computing the derivative (∂Σ(W1 , p)/∂pµ ),
we get three addends associated with the diagrams displayed in Fig. 22.27. These three
diagrams are obtained upon inserting the zero-momentum external photon line to three
internal electron lines of the diagram in Fig. 22.25. Note that insertion of the external

© 2011 by Taylor and Francis Group, LLC

 610 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

( a) (b) (c)

FIGURE 22.27
The diagrams (a)–(c) associated with (∂Σ(W1 , p)/∂pµ ).

photon line is (i) primarily lowering a degree of divergence (the diagram of Fig. 22.25 is
diverging linearly while that of Fig. 22.27 is diverging only logarithmically), and (ii) most
important, in the diagrams found in Fig. 22.27 we have no overlapping divergences. Let us
(1)
denote the operators correlated with the diagrams pictured in Fig. 22.27 as Λ(p1 , p2 , Vj )
P (1)
(j = 1, 2, 3). Non-overlapping divergences in Λµ (p1 , p2 , Vj ) could be extracted by
application the Dyson method. With allowance made for
3
∂ X (1)
µ
Σ(W1 , p) = Λµ (p, p, Vj ) = Λµ (p, p, V (1) ), (22.79)
∂p j=1

we carry out integration (22.79) and arrive at

Z p
Σ(W1 , p) − Σ(W1 , p′ ) = dq µ Λµ(q, q, V (1) ). (22.80)
p′

Making use of the relation:



′ 
Σ(W1 , p ) = −iδm(W1 ),

p̂′ =m R

one can present the formula (22.80) in the form:

Z p
Σ(W1 , p) + iδm(W1 ) = dq µ Λµ (q, q, V (1) ), (22.81)
p′

where p̂′ = mR . In such a manner, if the expression for the operator Λµ (q, q, V (1) ) (corre-
sponding to the compact VD) with already separated divergences is used in Eq. (22.81),
then divergences of the operator Σ(W1 , p) also prove to be isolated.
In the most general case, we should go to the sum over all compact diagrams. Then,
integration of the Ward identity yields
Z p Z 1
∗ ∗ ′ µ
Σ (p) − Σ (p ) = dq Λµ (q, q) = dx(pµ − p′µ )Λµ (px , px ), (22.82)
p′ 0

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 611

where p′ is a free-electron momentum, and in order to reach the last line we have done the
replacement
q = px = px + (1 − x)p′ .
The term Σ∗ (p′ ) allows us to automatically take into account the mass renormalization
because we have by definition

Σ∗ (p′ )|p̂′ =mR = −iδm.

From this point on we shall omit the term Σ∗ (p′ ) supposing that δm has been included into
the definition of Σ∗ (p).
As follows from Section 22.2, an exact value of the operator Λµ is found when in the
expressions associated with all the irreducible VDs we perform the following substitutions:

γµ → Γµ , S c → G(e) , Dc → G(γ) (22.83)

and summarize the whole series. The operator Λµ has no overlapping as it is constructed
from irreducible diagrams. Consequently, separation of divergences may be carried out
without fail.
A similar procedure is available for the PSED as well. Here the problem of overlapping
divergencies may be obviated by reducing analysis of the PSED to an analysis of the dia-
grams resultant from the differentiation over the external photon momentum. For example,
let us consider the diagram W2 shown in Fig. 22.28. The operator associated with this di-

q p
q−p
k k
k−q k−p

FIGURE 22.28
The PSED of the fourth order.

agram is denoted as Π(W2 , k 2 ) while the symbol ∆µ (W2 , k, k) will be used for the operator
resulting from differentiation of Π(W2 , k 2 ) with respect to kµ . Since
 
∂ 1 1 1
=− γµ , (22.84)
∂kµ k̂ − q̂ − m k̂ − q̂ − m k̂ − q̂ − m
then the operator ∆µ (W2 , k, k) could be constructed by application of the diagrams of Figs.
22.29 obtained by means of insertions of an external photon line with the zero-momentum
to those electron lines of the diagram of Fig. 22.28 that transfer an external momentum.
It should be noted that the diagrams shown in Figs. 22.29 are nonzero, despite the Furry
theorem, as here a sum of the diagrams is not considered. Obviously, the divergence of
the obtained diagrams becomes weaker because differentiation leads to an increase in the
denominator degree in the integrand by unity. In more higher degrees the PSEDs may
emerge in which the external photon momentum is transferred by some of the internal
photon lines (see, for example, Fig. 22.30). In this case
 
∂ 1 kµ − qµ
2
= −4
∂kµ (k − q) (k − q)3

© 2011 by Taylor and Francis Group, LLC

 612 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

k−q k−p
k−q
k−q k−p

FIGURE 22.29
The diagrams obtained by means of insertions of an external photon line to the diagram
pictured in Fig. 22.28.

k k

FIGURE 22.30
An example of higher degrees the PSED.

and the integral convergence is again improved (the term being proportional to qµ vanishes
in accordance with the Lorentz invariance). But in both cases the principal feature is the
elimination of overlapping divergences due to differentiation with respect to kµ . Actually,
when a zero-momentum photon line is connected to some part having an overlapping diver-
gence, the diagram element with a joined up photon line ceases to be divergent, and only
one of the overlapping divergences is retained.
Let us define an operator ∆µ (it is analogous to the operator Λµ ) with the help of the
relation:
∂ ∗ 2
Π (k ) = ∆µ (k, k). (22.85)
∂kµ
Then, taking into account that the photon self-energy equal to zero (Π∗ (0) = 0), we find
from (22.85)
Z k Z 1
∗ 2 µ
Π (k ) = dq ∆µ (q, q) = dyk µ ∆µ (yk, yk). (22.86)
0 0
The definition of ∆µ as the derivative of Π∗ is in fact an implicit definition enabling one
to find the functions Π∗ and ∆µ with the help of a certain class of the specially selected
irreducible diagrams. Naturally, a question arises: if ∆µ is an analog of Λµ , what quantity
will be associated with Γµ ? This quantity is designated by Wµ and given by the relation:
Wµ (k) = −2kµ + ∆µ (k). (22.87)
Really, as
∂[Dc (k)]−1
= 2k µ ,
∂kµ
then
∂ ∂
[G(γ) ]−1 = {[Dc (k)]−1 − [Π∗ (k)]} = 2k µ − ∆µ (k, k) = −Wµ (k). (22.88)
∂kµ ∂kµ

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 613

By application of Eq. (22.88), we find

Z 1
[G(γ) ]−1 = − dyk µ Wµ (ky). (22.89)
0

In the formalism of Feynman diagrams, the function Wµ represents a contribution of

the irreducible diagrams following from all the PSEDs by connection of an external photon
line to the electron lines transferring an external photon momentum (see, for example, Fig.
22.31). To obtain an exact value of the operator Wµ , we have to make substitutions (22.83)

p−k p−k

k k k k

(a) (b)

FIGURE 22.31
An example of the irreducible diagrams (a) and (b) following from the PSEDs.

in the expressions associated with the starting Feynman diagrams and to sum the derived
series. This gives the result:
Z
2α
Wµ = −2kµ − Sp{Γν (p, p + k)G(e) (p + k)Γµ (p + k, p + k)×
3(2π)3

×G(e) (p + k)Γν (p + k, p)G(e) (p)}d4 p + ..., (22.90)

where we have written only the term related with the diagram shown in Fig. 22.31(a).
Since, in the most general case, the operator is diverging quadratically, then the operator
∆µ may have linear divergence and no higher. Therefore, its finite value is given as:
 
∂∆µ (yk, yk)
∆Cµ (yk, yk) = ∆µ (yk, yk) − ∆µ (0, 0) − (yk)ν . (22.91)
∂(yk)ν k=0

Actually, due to the Lorentz invariance, this divergence is reduced to the logarithmic one
as there is no invariant four-dimensional vector, that is, ∆µ (0, 0) = 0. Recall, the vertex
operator playing the similar role for the ESED, is logarithmically diverging too, that is

ΓCµ (p1 , p2 ) = γµ + Λµ (p1 , p2 ) − Λµ (p′1 , p′1 ) = γµ + ΛCµ (p1 , p2 ), (22.92)

where p′1 is a four-dimensional momentum for a free electron.

Demonstration of the QED renormalizability may be performed in two stages. At the
first stage the subtraction procedure is performed, making all the elements of the S-matrix
finite. The second stage is a proof that this subtraction is equivalent to renormalization,
that is, dropping of the infinite terms is equivalent to redefinition of the mass and charge
values for a bare electron.

© 2011 by Taylor and Francis Group, LLC

 614 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

First, we define the finite parts of the functions Γµ , G(e) and G(γ) G, which should be
inserted into skeleton diagrams during the calculation of exact matrix elements. By the
appropriate subtractions, the kernels in the Dyson equations (22.17) and (22.20) are made
finite
Z 1
(e) c c (e)
G (p) = S (p) + S (p) dx(pµ − p′µ )[Λµ (px , px ) − Λµ (p′ , p′ )]GC (p), (22.93)
0
Z 1  
(γ) c c µ ∂∆µ (yk, yk) (γ)
G (k) = D (k)+ D (k) dyk [∆µ (yk, yk)− (yk)ν ]GC (k). (22.94)
0 ∂(yk)ν k=0
Next, in the modified equations (22.92)–(22.94) the constant e is replaced by the quantity
eR (e, m) selected so that the modified functions:
 
 
(e) (e)  (γ) (γ) 
GR (p; eR ) = G (p) , GR (k; eR ) = G (k) ,
 
e=eR e=eR



ΓRµ (p1 , p2 ; eR ) = ΓCµ (p1 , p2 )

e=eR

are different from the initial diverging functions G (p; e), G(γ) (k; e) and Γµ (p1 , p2 ; e) only
(e)

by the common factors. This means that modification of the initial integral equations
(22.17), (22.20), and (22.25) by subtraction of the divergent terms is reduced to renormal-
ization of the charge. Thus, we should show the validity of the equations:
(e)
G(e) (p; e) = Z2 (eR )GR (p; eR ), (22.95)
(γ)
G(γ) (k; e) = Z3 (eR )GR (k; eR ), (22.96)
Γµ (p1 , p2 ; e) = Z1−1 (eR )ΓRµ (p1 , p2 ; eR ), (22.97)
where Z1,2,3 is a divergent constant and eR is a renormalized charge
1/2
eR = Z1−1 Z2 Z3 e. (22.98)
Since these equations make it possible to interpret subtraction of infinite constants as a
separation of infinite constant factors, they testify to a multiplicative character of the renor-
malization.
Let us prove the relations (22.95)–(22.97). To this end we substitute G(e) (p; e), G(γ) (k; e)
and Γµ (p1 , p2 ; e) obtained from these relations into Eqs. (22.17), (22.20), and (22.25). It is
evident that in (22.25) an every term with the coefficient (e2 )n contains n functions G(γ) ,
2n functions G(e) and 2n + 1 functions Γµ . Then, carrying out the replacement
(e) (γ)
)
G(e) (p; e) → GR (p; eR ), G(γ) (k; e) → GR (k; eR ),
, (22.99)
Γµ (p1 , p2 ; e) → ΓRµ (p1 , p2 ; eR ) e → eR , ,
in the expression for Γµ and using the relations (22.95)–(22.97), we lead to the result:
(e) (γ)
Λµ (e, Γµ , G(e) , G(γ) ) = Z1−1 Λµ (eR , ΓRµ , GR , GR ). (22.100)
Next, we take into consideration that in (22.90) an every term with the coefficient (e2 )n
incorporates n−1 functions G(γ) and 2n+1 functions G(e) and Γµ . Fulfilling the change-over
(22.99) into (22.90), we come to the conclusion:
(e) (γ)
Wµ (e, Γµ , G(e) , G(γ) ) = Z1−1 Z3−1 Z2 Wµ (eR , ΓRµ , GR , GR ),

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 615

or
(e) (γ)
∆µ (e, Γµ , G(e) , G(γ) ) = Z1−1 Z3−1 Z2 ∆µ (eR , ΓRµ , GR , GR ). (22.101)
By application of the obtained expressions, we can rewrite the integral equations for the
mass and polarization operators (22.82), (22.86) as:
Z 1
∗ −1 (e) (γ)
Σ (p; eR ) = Z1 dx(pµ − p′µ )Λµ (px , px ; eR , ΓRµ , GR , GR ) (22.102)
0
Z 1
(e) (γ)
Π∗ (k; eR ) = Z1−1 Z2 Z3−1 dyk µ ∆µ (yk, yk; eR , ΓRµ , GR , GR ). (22.103)
0
When substituting the expressions (22.102) and (22.103) the integral equations (22.17),
(22.20), and (22.25) become the form:
(e)
Z2 GR (p; eR ) = S c (p) + S c (p)Z1−1 Z2 ×
Z 1
(e) (γ) (e)
× dx(pµ − p′µ )Λµ (px , px ; eR , ΓRµ , GR , GR )GR (p; eR ), (22.104)
0
Z 1
(γ) (e) (γ) (γ)
Z3 GR (k; eR ) c
= D (k) + D c
(k)Z1−1 Z2 dyk µ ∆µ (yk, yk; eR , ΓRµ , GR , GR )GR (k; eR ),
0
(22.105)
(e) (γ)
Z1−1 ΓRµ (p1 , p2 ; eR ) = γµ + Z1−1 Λµ (p1 , p2 ; eR , ΓRµ , GR , GR ). (22.106)
These equations must be compatible with the integral ones (22.92)–(22.94). In this case the
compatibility conditions enable us to determine the divergent constants Z1,2,3 .
At the fulfillment of the equality

Z1 γµ = γµ − Λµ (p′ , p′ ; eR ), (22.107)
(e) (γ)
where Λµ (p1 , p2 ; eR ) ≡ Λµ (p1 , p2 ; eR , ΓRµ , GR , GR ), Eqs. (22.106) and (22.92) identically
coincide. To compare (22.107) with the relations (22.76) and (22.77) permits one to express
the constant Z1 in terms of divergent parts of the vertex and mass operators

Z1 = 1 + B = 1 + L. (22.108)

The compatibility of Eqs. (22.93) and (22.104) results in

Z 1
Z2 − 1 = S c (p) dx(pµ − p′µ )[Z1−1 Z2 Λµ (px , px ; eR ) − Λµ (px , px ; eR ) + Lγµ ], (22.109)
0

where we have allowed for Λµ (p′ , p′ ; eR ) = Lγµ . Integrating the third addend in the right-
hand side of Eq. (22.109) and considering (22.108) and p̂′ = mR , we get
Z 1
c
Z2 − 1 = Z1 − 1 + S (p) dx(pµ − p′µ )[Z1−1 Z2 Λµ (px , px ; eR ) − Λµ (px , px ; eR )]. (22.110)
0

From (22.110) it immediately follows

Z1 = Z2 . (22.111)

In a similar way, the identical coincidence condition of Eqs. (22.105) and (22.94) gives
Z 1  
2 µ ν ∂∆µ (yk, yk)
(Z3 − 1)k = k k dy y .
0 ∂(ykν ) k=0

© 2011 by Taylor and Francis Group, LLC

 616 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

The solution of this equation is

Z3 = 1 + k −2 k µ k ν Cµν
′
, (22.112)

where Z  
1
′ ∂∆µ (yk, yk)
Cµν = dy y . (22.113)
0 ∂(ykν ) k=0

From the relativistic invariance follows

′
Cµν = C0 gµν ,

where 0 = Cσσ . Taking into account the linkage between the operator ∆µ (yk, yk) and the
polarization tensor, we arrive at
C0 = C = iΠ(0).
Thus, Eq. (22.112) takes the final form

Z3 = 1 + C = 1 + iΠ(0). (22.114)

So, the relations (22.95)–(22.97) are really valid when the renormalized charge eR is given
by the expression (22.98) while the divergent constants Z1,2,3 are defined by Eqs. (22.108),
(22.111), and (22.114).
From the Dyson equation for the electron Green function we have

[G(e) (p; e)]−1 = [S c (p)]−1 − Σ∗ (p). (22.115)

After the mass renormalization (including the term : δmψ(x)ψ(x) : into the interaction
Lagrangian), the operator Σ∗ (p) becomes the view:

Σ̃∗ (p) = B(p̂ − mR ) + ΣC (p), (22.116)

where ΣC (p) obeys the relation:


1 

ΣC (p) . (22.117)
p̂ − mR 
p̂=mR

Employing (22.108) and (22.111), one could present (22.115) as follows:

 −1
(e) 1 ΣC (p)
G (p; e) = Z2 + . (22.118)
p̂ − mR p̂ − mR

Eq. (22.118) gives the result

(e)
lim (p̂ − mR )GR (p; eR ) = 1. (22.119)
p̂→mR

One could analogously demonstrate that

(γ)
lim k 2 GR (k) = 1. (22.120)
k2 →0



ΓRµ (p1 , p2 ) = γµ . (22.121)

p̂1 =p̂2 =mR

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 617

As follows from the relations (22.119), (22.120), and (22.121), for free motion of an
electron or photon there are no any observable RCs in all orders of the perturbation theory.
As, in essence, the relations (22.119) and (22.120) determine the poles of the renormalized
Green functions, these relations may be considered as a requirement for the physical electron
to have an experimental mass value mR , and for the photon to have zero mass.
To this point our consideration has been concerned only with the internal effective lines
and vertices of the Feynman diagrams. Now we are coming to a discussion of external
electron and photon lines. Accomplishing the mass renormalization in (22.22) and allowing
for the relation
Σ̃(p) = (Z2 − 1)(p̂ − mR ), (22.122)
which is valid at p̂ → mR , we could represent (22.22) as:

ψ(p) = ψ(p) + (Z2 − 1)S c (p)(p̂ − mR )ψ(p). (22.123)

Unfortunately, it is impossible to take the derived relation as a renormalization condition

for the electron function. The expression (22.123) is ambiguous because it is dependent on
the action direction of the operator (p̂ − mR ). When the operator (p̂ − mR ) effects upon
ψ(p), we have zero, and in case of its action upon S c (p) the result is −1. A similar situation
has been encountered in Section 21.6 when the indicated ambiguity has been attributed to
the limiting process under which the initial and final states of the scattering matrix are
directed to t = ±∞. There is no point in repeating the procedure of adiabatic interaction
switching on and off that allows one to come to the correct result. Instead, we demonstrate
that the proper renormalization of the electron and photon wavefunctions is of the form
[52]:
1/2
ψ(p) = Z2 ψR (p), (22.124)
1/2
ψ(p) = Z2 ψ R (p), (22.125)
1/2
Aµ (k) = Z3 ARµ (k). (22.126)
The relation similar to (22.126) exists for the potential of an external electromagnetic field.
Note that, since the divergent constant Z2 governs renormalization of the wavefunction, the
equality Z1 = Z2 should be interpreted as a cancellation in all orders of the vertex divergence
by the wavefunction renormalization divergence. As shown in Section 22.2, inclusion of
higher orders of the perturbation theory for a particular process necessitates replacement
of the ordinary Feynman diagram elements in the associated irreducible diagrams by the
effective elements. Thus, when considering some nth order irreducible diagram, we can
schematically represent the corresponding element of the S-matrix as follows:
Z
n
M∼e (Γ)n (G(e) )Fi (G(γ) )Bi (ψ)Fe (A)Be , (22.127)

where (X)n denotes that the quantity X enters the integrand n-times.
The matrix element M is expressed in terms of the nonrenormalized quantities. Having
done the transition to the renormalized quantities in accordance with the formulae:
−1/2 1/2
e = Z3 eR , mR = m + δm, ψ = Z1 ψR , (22.128)
1/2 (γ) (e)
A = Z3 AR , G(γ) = Z3 GR , G(e) = Z1 GR , Γµ = Z1−1 ΓRµ , (22.129)
we obtain
Z
−n+Fi +Fe /2 B +Be /2−n/2 n (e) (γ)
M ∼ Z1 Z3 i eR (ΓR )n (GR )Fi (GR )Bi (ψR )Fe (AR )Be . (22.130)

© 2011 by Taylor and Francis Group, LLC

 618 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Each vertex is a joining place of two electron lines and one photon line, an internal line
entering into two vertices and an external line entering into a single vertex. Therefore,
1
n = Fi + Fe = 2Bi + Be , (22.131)
2
and, as a consequence, the exponents of Z3 and Z1 proves to be equal to zero
Z
(e) (γ)
M ∼ eR (ΓR )n (GR )Fi (GR )Bi (ψR )Fe (AR )Be .
n
(22.132)

In this way the matrix elements are identically expressed in terms of nonrenormalized
(e) (γ)
quantities (e, m, Γ, G(e) , G(γ) , ψ, A) and renormalized ones (eR , mR , ΓR , GR , GR , ψR , AR ).
As the matrix element M is assumed to be perfectly general, renormalizability of the QED
is completely proved. It should be noted that the proof is based on the relation (22.131),
connecting the vertex and line numbers in irreducible diagrams, and also on the gauge
invariance (to be exact, on its consequence, the Ward identity). We can construct the proof
of renormalizability for any other gauge theory in much the same manner.
It is significant that to keep the physical meaning of the renormalized mass and the renor-
malized charge, arbitrary renormalization constants must satisfy the particular conditions.
To illustrate, the requirements for the finiteness and positivity of mass as well as finiteness
and reality of charge lead to the following restrictions:

0 < m + δm < ∞, 0 < Z3 < ∞.

22.6 Renormalization group

Let us consider another important aspect of the problem involving a choice of subtraction
points in the process of renormalization. When calculating a physical (renormalized) charge
in Section 21.4, we have made the matrix element of the diagram in Fig. 21.7(e) finite by
subtraction of its contribution at the point q 2 = 0. From the viewpoint of experiment,
this means that we determine a physical charge as a parameter that figures in the electron
scattering cross section at an external electromagnetic field for q 2 = 0. Obviously, the
adopted approach is rather ambiguous. A charge might be determined for scattering of the
electron with some nonzero value |q 2 | = m∗ . This new charge e(m∗ ) in its value is differing
from the conventional one. Expansion of the S-matrix in this case is of the same form but
with respect to the new charge e(m∗ ). According to the renormalization procedure, the
contributions of the diagrams involving the PSEDs should be subtracted at the point m∗ .
A similar selection of a subtraction point occurs for renormalization of mass as well. By
the invariant Dyson method, the subtraction points are nothing else but the Taylor series
expansion points for the divergent expressions. The use of different methods of selection for
the series expansion points results in differing theoretical definitions of physical parameters.
At the same time, all subtraction points are equivalent as physics must be independent
on the selection of renormalization conditions. In other words, the renormalization proce-
dure should be self-consistent. On the mathematical language it means as follows. If one
(e) (γ) (e) (γ)
transits from the quantities e1 , G1 , G1 , Γ1µ to the ones e2 , G2 , G2 , Γ2µ by means of
transformation: 
1/2 (e) (e)
e2 = z0 e1 , G2 = z −1 G1 , 
(22.133)
(γ) (γ)
G2 = z0−1 G1 , Γ2µ = zΓ1µ , 

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 619

where z, z0 are arbitrary quantities, then matrix elements M expressed in terms of the first
and second sets of the quantities will be identical
M1 = M2 .
Assuming in (22.133) that z = z3 , z0 = z2 = z1 and taking the quantities with indices 1
and 2 as renormalized and bare ones respectively, we see that Eq. (22.133) represents a
multiplicative part of the renormalization procedure. It is expedient to recall our steps.
First, it has been demonstrated that in order to eliminate the divergences in the theory
we must introduce into the Lagrangian a finite number of counter-terms. Then we have
proved that, generally speaking, this is equivalent to some renormalization of the Green
functions, vertex function, and charge. In the process the counter-term associated with the
mass renormalization does not result in multiplicative renormalizations.
The transformations of the multiplicative renormalization (22.133) are clearly possessing
the group property. The corresponding group is identified as the multiplicative renormaliza-
tion group or simply renormalization one (RG). By definition, the RG in the QED represents
a group of the multiplicative transformations that act in a abstract space with the Green
functions, vertex functions and charge as “coordinates,” and cause no changes in the ob-
served effects. It is clear that the presence of the RG is not peculiar to the QED and may
be set in any renormalized theory. To this end, among all the acceptable counter-terms of
the present theory, we should separate only those leading to multiplicative renormalization
of the Green functions, vertex functions and charges, to obtain the transformation relations
similar to (22.133). Moreover, such a group also may be found in other fields of theoreti-
cal physics associated with critical phenomena: theory of turbulence, physics of polymers,
transport theory, magnetic hydrodynamics, etc. In the most general case, the RG appears
when the mathematical description of a physical problem involves selection of a particular
solution satisfying the boundary condition for some value of the argument (at a certain
renormalization point). And invariance with respect to the RG transformations (renorm-
invariance) indicates that the physical meaning of the theory is independent on the choice
of the renormalization point.
In the QFT the RG was first investigated in Ref. [101]. Later this group has been used
to obtain the concrete information about UV asymptotic form of the Green functions in the
QED [102]. It has been found that the renorm-invariance of the Green functions enables
derivation of their associated equations, with the solutions that allow one to carry out the
calculations beyond the applicability domain of the perturbation theory. A theory of the
RG has been completely developed in the works by Bogolyubov and Shirkov (see, Ref. [90]).
We make no efforts to study the RG equations for the vertex function or Green functions,
restricting ourselves to the consideration of an equation for the running coupling constant
α.
In the QED the RG transformation can be written as:
R(t) = {m2 → m∗2 = m2 t, α → α′ = α(t, α)}, (22.134)
where t is a continuous transformation parameter changing the scale of the variable m2 ,
and α satisfies the functional equation:
α(t, τ ) = α[t, α(τ, α)]. (22.135)
On the strength of Eq. (22.135) the transformations R(t) possess the group property:
R(t)R(τ ) = R(tτ )
and form the continuous group, that is, the Lie group. Thus, the RG could completely
be characterized by its infinitesimal element. Therefore, in place of functional equations,

© 2011 by Taylor and Francis Group, LLC

 620 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

we may examine differential ones (they are called the Lie equations) corresponding to the
transformations R(t) with t being close to 1. Such an equation for α could be set down in
the form:
∂α(t, α)
t = β[α(t, α)]. (22.136)
∂t
Here the function β(α) representing the group generator is prescribed according to the
relation: 
∂α(ξ, α) 

β(α) =  . (22.137)
∂ξ 
ξ=1

To calculate the β-function, we use the formula (22.140), on the right-hand side of which an
approximate value α obtained from the perturbation theory is substituted. Let us constraint
ourselves by the one-loop approximation. Then, in accordance with (21.100), the relation:
h α i
α(t, α) = α 1 + ln t , (22.138)
3π
holds for the ultraviolet region, and with the help of (22.137) we get
α2
β(α) = . (22.139)
3π
Inserting the deduced expression β(α) into (22.136) and integrating the obtained equation,
we come to the already familiar result:
α
α(t, α) = , (22.140)
1 + bα ln t
where b = −1/(3π) and α = α(0, α). This expression is an exact solution for a differential
equation (22.136) or, equivalently, for the functional group equation (22.135). At the same
time, upon a series expansion in terms of α it is in line with the approximate expression
(22.138). Consequently, it can be stated that a method of the RG offers synthesis of the
perturbation theory and renorm-invariance. The derived expression (22.140) includes a sum
over all the principal logarithmic contributions of the kind α(α ln t)n and may be used up
to the infinitely great values of t. It is apparent that the parameter t should be identified
with a ratio Q2 /m∗2 , where Q2 is the square of the four-dimensional momentum. By this
reason, the limit t → ∞ is associated with the UV asymptotic form Q2 → ∞. Recall that
the formulae of the kind (22.140) for the effective coupling constants of the electroweak and
strong interactions are basic for the formulation of the GUTs. A mathematical apparatus of
the GUTs relies on a system of coupled Lie equations for several effective coupling constants
that is a generalization of Eq. (22.136).

22.7 Consistency problems of the QED

The fact, that in the QED infinite quantities are appearing, does not matter as long as
the observables are finite. The appearance of divergent quantities at the intermediate stage
again may be considered as an introduction of some new class of the mathematical quantities
disappearing from the end formulae for the physical observables. Examples are found in
complex numbers, the δ-functions, and so on. Complex numbers have been traditionally
used by physics for a long time. The sewing together of the wave functions in a quasi-
classical approximation, based on the introduction of complex space coordinates, leads to

© 2011 by Taylor and Francis Group, LLC

 Renormalization theory 621

the Bohr-Sommerfeld quantization rule whose validity is unquestionable. In other words, a

problem associated with the infinities, appearing in the course of events of the “scenario”
and happily disappearing in the denouement, may be apprehended as the a happy end of
some melodrama. Unfortunately, the list of the QED problems is far from being exhausted.
Under studying the QED renormalizability we have used the S-matrix series expansion
in terms of the coupling constant powers. In the process it has silently been implied that
the series is convergent after renormalization. But it is known that, provided a possibility
for trapping into the bound stationary states exists in the theory, after renormalization the
series are divergent [103]. Recall that as this takes place, a series of the perturbation theory
is divergent even in a nonrelativistic quantum mechanics [104]. In this way, even when we
leave a problem of convergence alone, the renormalization pattern used is applicable only
in the case of the scattering processes of free particle, when there are no bound states.
As for the series convergence in the perturbation theory, the situation is also not so good.
The amplitude of an arbitrary process may be represented as a sum over the terms with
increasing powers of the quantity α, every term of the series being lower that the preceding
one by about a factor of 100. Because of this, it seems that a contribution made by the
diagrams with higher αn -powers should be infinitesimal so they could be omitted. But we
have to point out that the number of such diagrams has a factorial growth (∼ n!). Then
for a sufficiently high order the diagrams number is so great that it exceeds a small factor
of αn due to the high power n!. As a result, the higher-order corrections are growing, and
a sum over the whole series is infinite. Such a series are known in mathematics, where they
are referred to as asymptotic. The asymptotic series
∞
X
F (p; α) = αn fn (p) (22.141)
n=0

may be used for the description of a behavior exhibited by the function F , with a very
high but always finite accuracy. As distinct from the convergent series, the terms of the
asymptotic series αn fn are first decreased with the growing number n and then, beginning
from some nc , they are increased, generally without bound. At that, a maximum accuracy,
with which the function F may be approximated, is determined by the quantity fnc (p).
The lower is fnc (p), the higher the accuracy.
Physically, this aspect of the problem has been first explained in Ref. [105]. An asymp-
totic character of a series for the S-matrix in the QED has been inferred with the following
reasoning. Let us suppose that we calculate a physical observable F (p; α) (α = e2R /(4π)) in
the form of a power series for the running coupling constant e2R . When this series is conver-
gent at a particular positive value of e2R , it should be convergent in some circle of radius e2R
with a center at the origin of a complex plane Z = e2R . Consequently, the series should be
also convergent for e2R = 0 and, besides, on some interval of a real negative semiaxis, that
is, at negative values of e2R . But F (p; −e2R ) allows for a simple physical interpretation. The
quantity F (p; −e2R ) would represent the quantity F under study if the interaction between
two charges were determined by the function −e2R Dc (x) rather than e2R Dc (x). Meaning
that negative values of e2R are associated with the pseudoworld, where like charges are
attracted to each other, while unlike charges repel one another. Then the conventional
definition of the vacuum as a state with the lowest energy is invalid. Indeed, imagine the
situation when N electron-positron pairs are produced. In the process, all electrons are
concentrated in one region of the space, whereas positrons occupy the other. If these re-
gions are small enough and sufficiently separated from one another, in the case of great N
a negative Coulomb energy of the attracted like charges is higher than the rest energy of
particles and their kinetic energy. We call these bound states pathological. Apart from this
state, let another normal state be existent that involves a number of particles. This state

© 2011 by Taylor and Francis Group, LLC

 622 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

is separated from the pathological one of the same energy by a potential barrier, the height
of which is equal to the energy required to form N pairs (2N me ). And the tunneling effect
leads to the transitions from the normal to the pathological state. Note that a pathological
state, wherein the ordinary state is transferred, is nonstationary as particles are generated
in ever-growing numbers. In other words, the vacuum disintegration occurs. Due to these
effects, we cannot suppose that the QED with the interaction function −e2R Dc (x) leads to
the analytical S-matrix. As a result, at e2R 6= 0 a series for S(e2R ) is asymptotic and not
convergent. However, this difficulty is rather mathematical in nature since in Ref. [105] it
has been demonstrated that nc ≈ 4πe−2 R = 137 and the associated maximum accuracy, to
within the series (22.141) is close to reality, equals exp(−4π/e2R ). Evidently, this is more
than enough for the QED calculations in practice.
Another problem of the QED is related with a behavior of the effective electron charge
eef f ≡ e(Q2 ). As follows from:

3πe2 (mR )
e2 (Q2 ) = , (22.142)
3π − e2 (mR ) ln(Q2 /m2R )

for the transferred momentum:

 
2 3π
Q = m2R exp 2
(22.143)
e (mR )

the denominator of the expression (22.142) goes to zero, the charge itself being infinitely
large. As a consequence, a physically meaningless restriction is imposed on the values of
the four-dimensional momentum transfer. Substituting the numerical values into (22.143),
we find that the effective charge goes to zero at the fantastically high energies ∼ 10566 eV
exceeding the energy of the universe as a whole. This circumstance is somewhat soothing
because there is a chance to attribute the revealed inconsistency of the QED to its openness.
With such high energies it is necessary to take into consideration the effects due to all other
interactions, the gravitational one including.
It turned out so that at our climbing the third stage of the Quantum Stairway the QED
was to be responsible for the matter structure, just as it became the first working model of
the quantum field theory. Its basic properties—the relativistic covariance, the local gauge
invariance, the convergence of perturbation theory series, changing the interaction constant
with a momentum transferred—got those strings from which the Gobelin tapestry of modern
physics was weaved. It is difficult to foresee what aspect would assume the elementary
particles theory if the carriers of interactions between the electrons were massive particles
with the spin 3/2 rather than the photons. We finish this volume by the words of the poet
Robert Burns:
Sad words:
“Might be,”
Can’t be forgotten
By mice
And people too.

© 2011 by Taylor and Francis Group, LLC

 Problems

4.1. Free-fields operators to be entered into normal products and ordinal pairings obey the
free equation of motion. So, for a scalar field we have

( − m2 ) : ψ(x)ψ(y) := 0, ( − m2 )ψ s (x)ψ s (y) = 0.

Find the equations to be valid for ψ(x)ψ(y) standing both under the sign of chronological
product and under that of the T -product.
4.2. Prove that the vacuum expectation value (VEV) of the chronological product of linear
operators A, B1 , B2 , ...Bn is equal to the sum of n VEVs taking from the same chronological
products with all the possible pairings between one of these operators (for example, A) and
all other operators, that is,
X
< 0|T (AB1 ...Bn )|0 >= < 0|T (As B1 ...Bis ...Bn )|0 >
i

(the third Wick’s theorem).

4.3. For charged scalar particles the Lagrangian taking into account the electromagnetic
interaction is given by the expression

Lem = −ie[ϕ∗ (x)∂µ ϕ(x) − ∂µ ϕ∗ (x)ϕ(x)]Aµ (x) − e2 ϕ∗ (x)ϕ(x)Aµ (x)Aµ (x).

Direct calculations gives the following results for the chronological pairings of scalar field
operators
< 0|T (∂µ ψ(x)ψ ∗ (x′ )|0 >= −i∂µ ∆c (x − x′ ),

< 0|T (∂µ ψ(x)∂ν′ ψ ∗ (x′ )|0 >= −i∂µ ∂ν′ ∆c (x − x′ ) + 2inµ nν δ (4) (x − x′ ).
(em)
The terms depending on δ (4) (x − x′ ) also appear in Hint . However, as computations
show, in the matrix elements the mutual cancellation of the similar terms takes place and
the rectilinear procedure is reduced to completely neglecting these terms (see discussion in
Chapter 1.1). With allowance made for this circumstance, build up the Feynman rules in
the coordinate and momentum space.
4.4. Will Furry’s theorem be valid in the scalar electrodynamics?
4.5. Calculate the decay probability of a muon through the channel

µ− → e− + ν e + νµ ,

setting the matrix element square equal to 8G2F m4µ /3.

4.6. The muon decay amplitude through the channel

µ− → e− + ν e + νµ

to be summed up over particle spins is given by the expression

X
|M|2 = 128G2F (pe · pνµ )(pµ · pνe ). (IV.1)

623
© 2011 by Taylor and Francis Group, LLC
 624 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

Thinking the amplitudes of the decay (IV.1) and the reaction

νµ + e− → ν e + µ− (IV.2)

to obey the crossing-symmetry condition, find the differential and total cross sections of the
reaction (IV.2). Represent the differential cross section as a function of the energy portion
transferred to the recoil lepton: y = Eµ /Eνµ . For the sake of simplicity, set the neutrino
masses equal to zero.
4.7. Make the same as in the previous problem for the reaction

ν e + e− → ν µ + µ− .

4.8. Using the Feynman rules, write down the expression for the invariant amplitude of the
process
e+ + e− → γ + γ.
Check up the gauge invariance of the expression obtained.
4.9. Making use of the result of the problem 4.4., ascertain to which energies the Born’s
approximation of the four-fermion contact interaction is correct (unitary bound). [NOTE:
In the general case the scattering amplitude could be represented in the form
∞
1X
f (θ) = (2l + 1)fl Pl (cos θ),
k
l=0

where fl is a partial wave amplitude.]

4.10. Find physically acceptable value regions of the variables u, s, and t for the reaction

KS0 + π 0 → π 0 + π 0 .

4.11. Make the same as in the previous problem for the reaction

p + γ → p + π0 .

4.12. The relativistically invariant phase volume element for secondary particles is as
follows  
3
d pf (4)  X X
dΦ = Πf δ pi − pf  ,
2Ef i f

where pi and pf are four-dimensional momenta of initial and final particles. Find the total
phase volume Z
P hi = dΦ
√
for pair of particles with the masses m1 , m2 and the effective mass equal to s (s =
(p1 + p2 )µ (p1 + p2 )µ ).
4.13. Using the formula of the problem 4.8. and the particle mass values, compare the
phase volumes of the following decays:

K10 → π − + π +
(IV.3)
K + → π+ + π+ ,

K + → µ+ + νµ
(IV.4)
K + → e+ + νe

© 2011 by Taylor and Francis Group, LLC

 Problems 625

4.14. A resting particle a is decaying into particles 1, 2, and 3. Find events distribution
density in variables E1cm and E2cm (Ecm is an energy in the center-of-mass frame), assuming
that to be given by the relativistically invariant phase volume.
4.15. Using the results of the problem 4.13. and taking into consideration experimental
values of decay branchings Br(K + → π + + π 0 ) = 0.21 and Br(K10 → π + + π − ) = 0.684,
obtain the ratio of matrix element module squares for the following decays

K + → π+ + π0 , K10 → π + + π − .

4.16. Find the isotopic relations between probabilities of the φ(1020)-decays through the
channels:
0
φ → K + + K −, φ → K0 + K .
Having introduced a correction for a phase volume value, compare the obtained value with
the experimental data.
4.17. The isospin is not conserved in the weak interaction. However, in most cases nonlep-
ton decays of strange particles (decays in which a final state isospin is not defined exactly)
are under the selection rule
1
∆I = .
2
Using this rule, find a ratio of probabilities for the Λ-hyperon decays through the channels:

Λ → p + π− , Λ → n + π0 .

With allowance made for correction for the phase volume, define a ratio

Γ(Λ → nπ 0 )
.
Γ(Λ → nπ 0 ) + Γ(Λ → pπ − )

Compare the obtained value with the experimental one. [NOTE: The rule ∆I = 1/2 is
easiest to be applied assuming that some fictitious particle (spurion) with the isospin ∆I is
added to the left side of equations describing decay processes (in our case Isp = 1/2). After
that it can consider these decays as processes in which the isospin is conserved.]
4.18. Calculate the differential and total cross sections of the Compton effect for a scalar
particle.
4.19. Find the differential cross section of the process

γ + γ̃ → h− + h+ ,

where h− (h+ ) is a negative (positive) charged scalar boson, in the ultrarelativistic case.
4.20. Calculate the differential cross section of the bremsstrahlung for an ultrarelativistic
scalar particle.
4.21. Find the differential cross section of the reaction

h− + h+ → γ + γ,

where h− (h+ ) is a negative (positive) charged scalar boson.

4.22. Determine the fine structure of the parapositronium levels.
4.23. Find the differential cross section of an electron scattering by an electron in the non-
relativistic case. Consider electrons to be polarized.
4.24. Find the spin flip probability of a complete polarized electron under its scattering by
a nonpolarized electron in the nonrelativistic case.
4.25. Find the differential cross section of an elastic electron scattering by a muon in the
center-of-mass frame.

© 2011 by Taylor and Francis Group, LLC

 626 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

4.26. Find the self-energy of a charged scalar particle.

4.27. Determine the vacuum polarization in quantum electrodynamics of scalar particles.
4.28. Using the method of counting the momenta power in integrands, prove the scalar
electrodynamics renormalizability.

© 2011 by Taylor and Francis Group, LLC

 References

[1] S. Tomonaga, Progr. Theor. Phys. 1, 27 (1946); J. Schwinger, Phys. Rev. 74,
1439 (1948).
[2] J. A. Wheeler, Phys. Rev. 52, 1107 (1937).
[3] W. Heisenberg, Zeit. Phys. 120, 513 (1943); 120, 673 (1943).
[4] F. J. Dyson, Phys. Rev. 75, 486 (1949); 75, 1736 (1949).
[5] N. N. Bogoliubov, D. V. Shirkov, Introduction to the Theory of Quantized
Fields, (Interscience Publishers Inc., New York, 1959).
[6] G. Wick, Phys.Rev. 80, 268 (1950).
[7] R. Feynman, Phys. Rev. 76, 749 (1949); 76, 769 (1949).
[8] W. H. Furry, Phys. Rev. 51, 125 (1937).
[9] S. Mandelstam, Phys. Rev. 112, 1344 (1958).
[10] T. W. B. Kibble, Phys. Rev. 117, 1159 (1960).
[11] J. G. Taylor, Lectures on Dispersion Relations in Quantum Field Theory and
Related Topics, (New issue, Univrsity of Maryland, 1959).
[12] H. Bremerman, R. Oehme, J. G. Taylor, Phys. Rev. 109, 2178 (1958).
[13] R. Kronig, J. Opt. Soc. Amer. 12, 547 (1926); H. Kramers, Atti. Congr. Intern.
Fisici Como. 2, 545 (1927).
[14] G. Barton, Introduction to Dispersion Techniques in Field Theory,
(W. Benjamin. Inc., New York, 1965).
[15] N. F. Mott, Proc. Roy. Soc. A124, 425 (1929).
[16] R. H. Dalitz, Proc. Roy. Soc. A 206, 509 (1951).
[17] K. Blum, Density Matrix Theory and Applications, (Plenum Press, New York
and London, 1981).
[18] L. D. Landau, E. M. Lifshitz, Quantum Mechanics. Nonrelativistic Theory,
(Pergamon Press, London, 1959).
[19] A. Zommerfeld, Atombau und Spectrallinlen, (Brauschweig, 1939).
[20] F. Sauter, Ann. d. Phys. 9, 217 (1931); 11, 454 (1931).
[21] O. Klein, Y. Nishina, Zeit. Phys. 52, 853 (1929).
[22] U. Fano, Phys. Rev. 93, 121 (1954).
[23] H. A. Tolhoek, Rev. Mod. Phys. 28, 277 (1956); H. Mc Master, Rev. Mod.
Phys. 33, 8 (1961).
[24] H. Bethe, W. Heitler, Proc. Roy. Soc. 146, 83 (1934).

627
© 2011 by Taylor and Francis Group, LLC
 628 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

[25] R. L. Gluckstern, M. H. Hull, Phys. Rev. 90, 1030 (1953).

[26] R. I. Gluckstern, M. H. Hull, G. Breit, Phys. Rev. 90, 1026 (1953); K. Mc Voy,
Phys. Rev. 106, 828 (1957); H. Banerjee, Phys. Rev. 111, 532 (1958).
[27] M. Delbrück, Zeit. Phys. 84, 144 (1933).
[28] V. A. Baier, V. S. Fadin, V. A. Khoze, 50, 1611 (1966); 51, 1135 (1966).
[29] G. Breit, E. Teller, Astrophys. J. 91, 215 (1940).
[30] K. Mc Voy, Phys. Rev. 110, 484 (1958).
[31] J. M. Jauch, F. Rohrlich, Helv. Phys. Acta 27, 613 (1954).
[32] R. P. Feynman, Phys. Rev. 74, 1430 (1953).
[33] P. A. Dirac, Proc. Cambr. Phil. Soc. 26, 361 (1930); I. Tamm, Zeit. Phys. 62,
545 (1930).
[34] L. D. Landau, Dokl. Akadem. Nauk: Ser. Fiz. 60, 207 (1948) (Also published
in Collected Papers of L. D. Landau (Edited by D. Ter Haar, Pergamon Press,
1965).
[35] L. A. Page, Phys. Rev. 106, 394 (1957); H. Mc Master, Rev. Mod. Phys. 33, 8
(1961).
[36] I. Ya. Pomeranchuk, Dokl. Akadem. Nauk: Ser. Fiz. 60, 218 (1947).
[37] A. Ore, J. Powell, Phys. Rev. 75, 1696 (1949).
[38] V. I. Goldansky, V. G. Firsov, Uspekhi Khimii, 40, 1353 (1971).
[39] A. Schenk, Muon Spin Rotation Spectroscopy, (Bristol, 1985).
[40] S. Möller, Ann. Phys. 14, 568 (1932).
[41] A. M. Bincer, Phys. Rev. 107, 1434 (1957); K. Böckman, G. Kramer, W. R.
Theis, Zeit. Phys. 150, 201 (1958).
[42] H. J. Bhabha, Proc. Roy. Soc. 154, 195 (1936).
[43] V. B. Berestetski, I. Ya. Pomeranchuk, JETP, 29, 864 (1955).
[44] A. I. Akhiezer, R. V. Polovin, Uspekhi Fiz. Nauk 51, 3 (1953).
[45] W. Pauli, F. Villars, Rev. Mod. Phys. 21, 434 (1949).
[46] G. t’Hooft, M. Veltman, Nucl. Phys. B44, 189 (1972); C. G. Bollini, J. J.
Giambiagi, Phys.Lett. B40, 566 (1972).
[47] H. A. Kramers, Rapports du 8e Conseil Solvay 1948, 241, Bruxelles, 1950.
[48] E. A. Uehling, Phys. Rev. 48, 55 (1935).
[49] P. A. M. Dirac, Proc. Cambr. Phil. Soc. 30, 150 (1934).
[50] W. Heisenberg, Zeit. Phys. 90, 209 (1934); 92, 692 (1934).
[51] D. Feldman, Phys. Rev. 76, 1369 (1949); H. Umezawa, R. Kawabe, Progr.
Theor. Phys. 4, 443 (1949).
[52] R. Karplus, N. M. Kroll, Phys.Rev. 77, 536 (1950).
[53] W. A. McKinley, H. Feshbach, Phys. Rev. 74, 1759 (1948).

© 2011 by Taylor and Francis Group, LLC

 References 629

[54] R. Dalitz, Proc. Roy. Soc. A206, 509 (1951).

[55] J. Schwinger, Phys. Rev. 75, 1912 (1949); L. Elton, H. Robertson, Proc. Roy.
Soc. A65, 145 (1952).
[56] J. Jauch, F. Rohrlich, Helv. Phys. Acta 27, 613 (1954).
[57] H. Suura, Phys. Rev. 99, 1020 (1955).
[58] J. Schwinger, Phys. Rev. 76, 790 (1949).
[59] D. R. Yennie, H. Suura, Phys. Rev. 105, 1738 (1957); Yung Su Tsai, Phys.
Rev. 120, 269 (1960); S. N. Gupta, Phys. Rev. 98, 1507 (1955).
[60] L. M. Brawn, R. P. Feynman, Phys. Rev. 85, 231 (1952).
[61] C. M. Sommerfield, Phys. Rev. 107, 328 (1957); A. Peterman, Helv. Phys. Acta
30, 409 (1957).
[62] A. Czarnecki and W. J. Marciano, Nucl. Phys. (Proc. Suppl) 76, 245 (1999);
P. J. Mohr and B. N. Taylor, Rev. Mod. Phys. 72, 351 (2000).
[63] J. E. Nafe, E. B. Nelson, I. I. Rabi, Phys. Rev. 71, 914 (1947).
[64] E. Fermi, Zeit. Phys. 60, 320 (1930).
[65] G. Breit, Phys. Rev. 72, 984L (1947).
[66] H. M. Foley and P. Kusch, Phys. Rev. 73, 412 (1948).
[67] J. Schwinger, Phys. Rev. 73, 416L (1948).
[68] V. Bargmann, L. Michel, V. Telegdi, Phys. Rev. Lett. 2, 435 (1959).
[69] J. Field, E. Picasso, F. Combley, Uspekhi Fiz. Nauk, 127, 553 (1979).
[70] R. S. van Dyck Jr., P. B. Schwinberg, and H. G. Dehmelt, Phys. Rev. Lett. 59,
26 (1987).
[71] I. M. Ternov, V. R. Khalilov, V. N. Radionov, Yu. I. Klimenko, JETP, 74, 1301
(1978); I. M. Ternov, V. R. Khalilov, Yad. Fiz. 35, 328 (1982).
[72] T. Kinoshita, Rept. Prog. Phys. 59, 1459 (1996).
[73] Muon g-2 Collaboration, G. W. Bennett at al., Phys. Rev. Lett. 92, 161802
(2004).
[74] T. Kinoshita and W. J. Marciano, in Quantum Electrodynamics, (edited by T.
Kinoshita, World Scientific, Singapore, 1990), 419–478.
[75] K. Hagiwara, A. Martin, D. Nomura, and T. Teubner, Phys. Rev. D69, 093003
(2004).
[76] W. E. Lamb, R. C. Retherford, Phys. Rev. 72, 241 (1947).
[77] H. A. Bethe, Phys. Rev. 72, 339 (1947).
[78] J. Schwinger, Proc. Nat. Acad. Sci. USA 37, 452 (1951).
[79] P. J. Mohr, Phys. Rev. Lett. 34, 1050 (1975).
[80] S. R. Lundeen, F. M. Pipkin, Phys. Rev. Lett. 46, 232 (1981).
[81] Yu. L. Sokolov, V. P. Yakovlev, JETP 83, 15 (1982).

© 2011 by Taylor and Francis Group, LLC

 630 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

[82] G. W. Erickson, D. R. Yennie, Ann. Phys. 35, 271 (1965); 35, 447 (1965).
[83] H. A. Bethe, E. Salpeter, Quantum Mechanics of One and Two Electron Atoms,
(Springer-Verlag, Berlin, 1957); T. Kinoshita, Perspective of Quantum Physics,
(Iwanami Shoten, Tokyo, 1978).
[84] R. I. Eden, Proc. Roy. Soc. A215, 133 (1952).
[85] P. T. Mathews, A. Salam, Rev. Mod. Phys. 23, 311 (1951).
[86] G. Takeda, Prog. Theor. Phys. 7, 359 (1952).
[87] J. C. Ward, Phys. Rev. 84, 497 (1951).
[88] J. Schwinger, Qauntum Electrodynamics, (Dower, New York, 1958).
[89] N. N. Bogoliubov, O. S. Parasiuk, Acta Math. 97, 227 (1958).
[90] N. N. Bogoliubov, D. V. Shirkov, Introduction to the Theory of Quantized
Fields, (John Wiley, 1982).
[91] G. Källen, Helv. Phys. Acta 25, 417 (1952); H. Lehmann, Nuovo Cimento 11,
342 (1954); H. Lehmann, K. Symanzik, W. Zimmermann, Nuovo Cimento 6,
319 (1957); A. S. Wightman, Phys. Rev. 101, 860 (1956).
[92] N. N. Bogoliubov, A. A. Logunov, I. T. Todorov, Introduction to an Axiomatic
Quantum Field Theory, (Benjamin, 1975); H. Epstein, V. Glaser, Annales
deL’Institute Poincare XIX, 211 (1973).
[93] K. Hepp, Theorie de la Renormalisation, (Springer-Verlag, Berlin, 1969).
[94] J. C. Ward, Phys.Rev. 73, 182 (1950); Proc. Phys. Soc. A64, 54 (1951).
[95] F. I. Dyson, Phys. Rev. 82, 428 (1951).
[96] A. Akhiezer, Phys. Zs. d.Sowjetunion 11, 263 (1937); R. Karlplus, M. Neuman,
Phys. Rev. 80, 380 (1950).
[97] J. Schwinger, Proc. Nat. Acad. USA 37, 455 (1951).
[98] A. I. Akhiezer, V. B. Berestetski, Quantum Electrodynamics, (Inter., New York,
1965).
[99] J. C. Ward, Phys. Rev. 77, 293L (1950); 78, 182L (1950).
[100] A. Salam, Phys. Rev. 82, 217 (1951); 84, 426 (1951).
[101] E. Stückelberg, A. Petermann, Helv. Phys. Acta 26, 499 (1953).
[102] M. Gell-Mann, F. Low, Phys. Rev. 95, 1300 (1954).
[103] B. Ferretti, Nuovo Cimento 8, 108 (1951); K. Nishijima, Progr. Theor. Phys.
6, 37 (1951).
[104] R. Jost, A. Pais, Phys. Rev. 82, 840 (1951).
[105] F. J. Dyson, Phys. Rev. 85, 631 (1952).

© 2011 by Taylor and Francis Group, LLC

 Appendix: Natural system of units

Typical velocities of elementary particles are close to the light velocity, angular momenta
represent multiples of ~/2, while energies even if they reach the order of 10−5 J, are ranked
among a category of superhigh ones. Consequently, when we are trying to imagine the
elementary particles world visually, one of the troubles in our consciousness is caused by
the fact that constants, whose values cannot be laid in macroworld standards, present in
the microworld physics formulae. On the other hand, it is easy to appreciate their values
compared with each other. So, as the natural step, we should break off the connection with
macroworld. It is achieved by the transition to the system of units where the fundamental
physical constants of microworld are used as the basic units. Such systems are called natural
system of units (NSU).
The first to suggest one of NSU kinds was M. Plank. He chose ~, c, G, and k (the
Boltzmann constant) as the basic units. Under the NSU construction the fundamental
constants, taking as the basic units, are formally assumed to be equal to 1. So, the Plank
NSU is defined by the relation:
~ = c = G = k = 1.
Because the quantum field theory (QFT) is symbiosis of the special theory of relativity
and nonrelativistic quantum mechanics, then the NSU of the QFT should be defined as
follows:
~ = c = 1.
In this system the dimensionality of any dynamical observable O is connected with the mass
dimensionality, that is,
[O] = [mn ] (n ∈ N ).
For example, for velocity, action and angular momentum n is equal to 0 while for energy
and momentum we have n = 1. From the Heisenberg uncertainty relations:
~
∆pi · ∆xi ≥ , (A.1)
2
~
∆E · ∆t ≥ , (A.2)
2
it follows that for coordinate and time n equals -1. In this system the electric charge is a
dimensionless quantity as its linkage with the fine structure constant α is given by:

e2
= α.
4π~c
Using the definition of the Lorentz force
e
F = eE + [v × H],
c
one may be convinced that the strengths of the electric E and magnetic H fields have the
dimensionality of m2 . In the NSU, the value “eV” (1 eV=1.6021892·10−19 J) and derivatives
from it (keV, MeV, GeV, TeV, etc.) are used as a mass unit.

631
© 2011 by Taylor and Francis Group, LLC
 632 Advanced Particle Physics: Particles, Fields, and Quantum Electrodynamics

To pass on to some ordinary system of units, we must have formulae connecting the basic
units of both systems. Let us chose “GeV” as a basic unit in the NSU. Then, for the CGS
system the definitions of “g,” “cm,” and “s” may be found from the relations:

1GeV = 1.6021892 · 10−10 J = 1.7826759 · 10−24 g · c2 , (A.3)

~c = 1.9732858 · 10−14 GeV · cm, (A.4)

−25
~ = 6.582173 · 10 GeV · s. (A.5)
Using Eqs. (A.3)–(A.5), one could express any derivative CGS unit in terms of “GeV.” For
example, in the CGS the force dimensionality is given by

g · cm ·10−12 Gev2
1dyn = = ,
s2 ~c
that allows us to state 1 dyne is equal to GeV2 in the CGS.
To compare some quantities, it is necessary that they have identical dimensionalities. One
of the advantages of the NSU implies that some quantities with incoincident dimensionalities
in ordinary system of units have one and the same dimensionality in the NSU. As an
example, it is instructive to consider the coupling constants GF and G defining the intensity
of the weak and gravitational interactions, respectively. The calculations give

NSU
GF ∼ 1.4 · 10−49 Erg · cm3 = 1.2 · 10−5 ~3 c3 GeV−2 −→ 1.2 · 10−5 GeV−2 ,

NSU
G ∼ 6.7 · 10−8 cm3 · g−1 · s−2 = 6.7 · 10−39 ~c5 GeV−2 −→ 6.7 · 10−39 GeV−2 .
The obtained result corresponds to the well known fact—in the region of energies reached
to date, the intensity of the weak interaction is as great as 1033 times than that of the
gravitational interaction.
After the formula for observable quantity O has been obtained, we should make the
transition from the NSU to some ordinary system of units, to say, to the CGS. For this
purpose, we must rebuild the right dimensionalities of all the physical quantities entering
into O (aside from energy, of course) with the help of the following replacements:

[m]E → [m]c2 , [l]E → [l](~c)−1 , [t]E → [t](~)−1 , (A.6)

where the subscript E shows that the quantity dimensionality is taken in the NSU.

© 2011 by Taylor and Francis Group, LLC