p
elie arieran
Field /Iheory
From Operators to
Path Integrals
Te
eur)
een
co) Compactified
2D space
Kerson HuangQUANTUM FIELD
THEORY
From Operators to
Path Integrals
Kerson Huang
Massachusetts Institute of Technology
Cambridge, Massachusetts
A WileyInterscience Publication
JOHN WILEY & SONS, INC.
New York = Chichester = Weinheim = Brisbane = Singapore = TorontoThis text is printed on acidfree paper. @
Copyright © 1998 by John Wiley & Sons, Inc. All rights reserved.
Published simultaneously in Canada.
No part of this publication may be reproduced, stored in a retrieval system or transmitted in any
form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise,
except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without
either the prior written permission of the Publisher, or authorization through payment of the
appropriate petcopy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA
(01923, (978) 7508400, fax (978) 7504744, Requests to the Publisher for permission should be
addressed to the Permissions Department, John Wiley & Sons, Inc., 605 Third Avenue, New York,
NY 101580012, (212) 850601, fax (212) 8506008, EMail: PERMREQ @ WILEY.COM.
Library of Congress Cataloging in Publication Data:
Huang, Kerson, 1928—
Quantum field theory / by Kerson Huang.
pcm
Includes bibliographical references and index.
ISBN 0471141208 (cloth : alk. paper)
1. Quantum field theory. I. Title
QC174.45,H82 1998
530.143—de 21 9813437
Printed in the United States of America
10987654321To RosemaryContents
Preface xv
Acknowledgment xvii
1. Introducing Quantum Fields 1
1.1, The Classical String, 1
1.2. The Quantum String, 5
1.3. Second Quantization, 7
1.4, Creation and Annihilation Operators, 10
1.5. Bose and Fermi Statistics, 12
Problems, 13
References, 16
. Scalar Fields 17
2.1. KleinGordon Equation, 17
Real Scalar Field, 18
Energy and Momentum, 20
Particle Spectrum, 22
Continuum Normalization, 22
Complex Scalar Field, 24
Charge and Antiparticle, 25
.8. Microcausality, 27
2.9. The Feynman Propagator, 28
2.10. The Wave Functional, 31
2.11. Functional Operations, 32
2.12. Vacuum Wave Functional, 33
2.13. The * Theory, 35
Problems, 37
nv
viioe
4
wv
a
. Dirac Equation
Contents
Relativistic Fields 40
3.1. Lorentz Transformations, 40
Minimal Representation: SL(2C), 43
3.3. The Poincaré Group, 44
3.4. Scalar, Vector, and Spinor Fields, 46
Relativistic Quantum Fields, 48
3.6. OneParticle States, 50
Problems, 52
Reference, 53
Canonical Formalism 54
4.1. Principle of Stationary Action, 54
4.2. Noether’s Theorem, 57
4.3. Translational Invariance, 58
4.4. Lorentz Invariance, 61
4.5. Symmetrized EnergyMomentum Tensor, 63
4.6. Gauge Invariance, 63
Problems, 65
Reference, 67
Electromagnetic Field 68
5.1. Maxwell’s Equations, 68
5.2. Covariance of the Classical Theory, 70
5.3. Canonical Formalism, 72
5.4. Quantization in Coulomb Gauge, 75
5.5. Spin Angular Momentum, 78
5.6. Intrinsic Parity, 80
5.7. Transverse Propagator, 81
5.8. Vacuum Fluctuations, 82
5.9. The Casimir Effect, 85
5.10. The Gauge Principle, 88
Problems, 90
References, 93
94
6.1. Dirac Algebra, 94
6.2. Wave Functions and Current Density, 98
6.3. Plane Waves, 99
6.4. Lorentz Transformations, 102
6.5. Interpretation of Dirac Matrices, 106
6.6. External Electromagnetic Field, 107
6.7. Nonrelativistic Limit, 109
6.8. Thomas Precession, 111
6.9. Hole Theory, 1147

a
6.10. Charge Conjugation, 117
6.11. Massless Particles, 118
Problems, 119
References, 122
The Dirac Field
7.1. Quantization of the Dirac Field, 123
7.2. Feynman Propagator, 126
7.3. Normal Ordering, 128
7.4. Electromagnetic Interactions, 129
7.5. Isospin, 130
7.6. Parity, 132
7.7. Charge Conjugation, 133
7.8. Time Reversal, 134
Problems, 136
Reference, 137
. Dynamics of Interacting Fields
8.1. Time Evolution, 138
8.2. Interaction Picture, 139
8.3. Adiabatic Switching, 142
8.4. Correlation Functions in the Interaction Picture, 144
8.5. S Matrix and Scattering, 147
8.6. Scattering Cross Section, 148
8.7. Potential Scattering, 151
8.8. Adiabatic Theorem, 154
Problems, 158
References, 159
Feynman Graphs
9.1. Perturbation Theory, 160
9.2. TimeOrdered and Normal Products, 161
9.3. Wick’s Theorem, 162
9.4. Feynman Rules for Scalar Theory, 165
9.5. Types of Feynman Graphs, 171
9.5.1. Vacuum Graph, 171
9.5.2. SelfEnergy Graph, 171
9.5.3. Connected Graph, 172
9.6. Wick Rotation, 172
9.7. Regularization Schemes, 173
9.8. LinkedCluster Theorem, 175
9.9. Vacuum Graphs, 175
Problems, 176
Reference, 181
Contents ix
123
138
160x
10.
12.
ae
Contents
Vacuum Correlation Functions 182
10.1. Feynman Rules, 182
10.2. Reduction Formula, 185
10.3. The Generating Functional, 188
10.4. Connected Correlation Functions, 189
10.5. Lehmann Representation, 190
10.6. DysonSchwinger Equations, 193
10.7. Bound States, 197
10.8. BetheSalpeter Equation, 200
Problems, 201
References, 203
. Quantum Electrodynamics 204
11.1, Interaction Hamiltonian, 204
11.2. Photon Propagator, 207
11.3. Feynman Graphs, 210
11.4. Feynman Rules, 215
11.5. Properties of Feynman Graphs, 217
Problems, 219
References, 219
Processes in Quantum Electrodynamics 220
12.1, Compton Scattering, 220
. Electromagnetic Form Factors, 224
. Anomalous Magnetic Moment, 229
12.4. Charge Distribution, 232
Problems, 234
References, 235
Perturbative Renormalization 236
13.1. Primitive Divergences in QED, 236
13.2. Electron SelfEnergy, 237
Vacuum Polarization, 243
Running Coupling Constant, 247
Full Vertex, 248
Ward Identity, 249
Renormalization to Second Order, 251
Renormalization to All Orders, 252
9. CallanSymanzik Equation, 257
13.10. Triviality, 259
Problems, 260
References, 263Contents xi
14, Path Integrals 264
14.1. Path Integrals in Quantum Mechanics, 264
14.2. Imaginary Time, 268
143, Path Integrals in Quantum Field Theory, 269
144, Euclidean SpaceTime, 271
14.5. Vacuum Amplitudes, 272
14.6. Statistical Mechanics, 274
14.7. Gaussian Integrals, 276
14,8, Perturbation Theory, 280
14.9. The Loop Expansion, 283
14.10. Boson and Fermion Loops, 284
14.11. Grassmann Variables, 287
ae
16.
Problems, 290
References, 293
Broken Symmetry 294
(Ea Be
15.2.
15.3.
15.4.
15.5.
15.6.
15.7.
15.8
Why Broken Symmetry, 294
Ferromagnetism, 296
Spin Waves, 300
Breaking Gauge Invariance, 303
Superfluidity, 306
GinsburgLandau Theory, 310
Effective Action, 312
Effective Potential, 314
Problems, 315
References, 318
Renormalization 320
16.1.
16.2.
16.3.
16.4.
16.5.
16.6.
16.7.
16.8.
16.9.
The Cutoff as Scale Parameter, 320
Momentum Space RG, 324
16.2.1. Designating Fast and Slow Modes, 325
16.2.2, Coarse Graining, 326
16.2.3. Rescaling, 326
Real Space RG, 327
163.1. Making Blocks, 328
16.3.2. Coarse Graining, 329
16.3.3. Rescaling, 330
Renormalization of Correlation Functions, 330
Relevant and Irrelevant Parameters, 331
The Free Field, 332
IR Fixed Point and Phase Transition, 334
Crossover, 336
Relation with Perturbative Renormalization, 337xii Contents
16.10. Why Correct Theories are Beautiful, 339
Problems, 341
References, 342
17. The Gaussian Fixed Point 343
17.1. Stability of the Free Field, 343
17.2. General Scalar Field, 344
17.3. Feynman Graphs, 345
17.4, WegnerHoughton Formula, 347
17.5. Renormalized Couplings, 350
17.6. The RG Matrix, 352
17.7. Nontriviality and Asymptotic Freedom, 356
17.8. The Case d= 2, 357
Problems, 359
References, 359
18. In Two Dimensions 360
18.1. Absence of LongRange Order, 360
18.2. Topological Order, 361
18.3. XY Model, 363
18.4. Kosterlitz—Thouless Transition, 367
18.5. Vortex Model, 368
18.6, 2D Superfluidity, 370
18.7. RG Trajectories, 372
18.8. Universal Jump of Superfluid Density, 376
Problems, 378
References, 378
19. Topological Excitations 380
19.1. Topological Soliton, 380
19.2, Instanton and Tunneling, 383
19.3. Depinning of Charge Density Waves, 385
19.4, Nonlinear Sigma Model, 389
19.5. The Skyrmion, 391
19.6. The Hopf Invariant, 395
19.7. Fractional Spin, 398
19.8. Monopoles, Vortices, and Anomalies, 401
Problems, 403
References, 405
A. Background Material 406
A.1. Notation, 406A.2. Classical Mechanics, 408
A.3. Quantum Mechanics, 408
B. Linear Response
References, 418
Index
Contents xiii
412
419Preface
Quantum field theory, the quantum mechanics of continuous systems, arose at the
beginning of the quantum era, in the problem of blackbody radiation. It became ful
ly developed in quantum electrodynamics, the most successful theory in physics.
Since that time, it has been united with statistical mechanics through Feynman's
path integral, and its domain has been expanded to cover particle physics, con
densedmatter physics, astrophysics, and wherever path integrals are spoken.
This book is a textbook on the subject, aimed at readers conversant with what is
usually called “advanced quantum mechanics,” the equivalent of a firstyear gradu
ate course. Previous exposure to the Dirac equation and “second quantization”
would be very helpful, but not absolutely necessary. The mathematical level is not
higher than what is required in advanced quantum mechanics; but a degree of matu
rity is assumed.
In physics, a continuous system is one that appears to be so at long wavelengths
or low frequencies. To model it as mathematically continuous, one runs into diffi
culties, in that the highfrequency modes often give rise to infinities. The usual pro
cedure is to start with a discrete version, by discarding the highfrequency modes
beyond some cutoff, and then try to approach the continuum limit, through a
process called renormalization.
Renormalization is a relatively new concept, but its workings were already evi
dent in classical physics. At the beginning of the atomic era, Boltzmann noted that
classical equipartition of energy presents conceptual difficulties, when one serious
ly considers the atomic structure of matter. Since atoms are expected to contain
smaller subunits, which in turn should composed of even smaller subunits, and so
ad infinitum, and each degree of freedom contributes equally to the thermal energy
of a substance, the specific heat of matter would be infinite.. The origin of this di
vergence lies in the extrapolation of known physical laws into the highfrequency
domain, a characteristic shared by the infinities in quantum field theory.
Boltzmann's “paradox,” however, matters not a whit when it comes to practical
calculations, as evidenced by the great success of classical physics. The reason is
that most equations of macroscopic physics, such as those in thermodynamics and
xvXvi Preface
hydrodynamics, make no explicit reference to atoms, but depend on coefficients
like the specific heat, which can be obtained from experiments. From a modern per
spective, we say that such theories are “renormalizable,” in that the microstructure
can be absorbed into measurable quantities.
One goal of this book is to explain what renormalization is, how it works, and
what makes some systems appear “renormalizable” and others not. We follow the
historical route, discovering it in quantum electrodynamics through necessity, and
then realizing its physical meaning through Wilson’s pathintegral formulation.
This book, then, starts with a thorough introduction to the usual operator for
malism, including Feynman graphs, from Chapters 110. This is followed by Chap
ters 11~14 on quantum electrodynamics, which illustrates how to do practical calcu
lations, and includes a complete discussion of perturbative renormalization. The last
part, Chapters 1519, introduces the Feynman path integral, and discusses “mod
ern” subjects, including the physical approach to renormalization, spontaneous
symmetry breaking, and topological excitations. I have entirely omitted non
Abelian gauge fields and the standard model of particle physics, because these sub
jects are discussed in another book: K. Huang, Quarks, Leptons, and Gauge Field,
2nd ed. (World Scientific, Singapore, 1992).
Ihave chosen to introduce path integrals only after the canonical approach is
fully developed and applied. Others might want them discussed earlier. To accom
modate different tastes, I have tried to make each chapter selfcontained in as much
as possible, so that a knowledgeable reader can pick and skip.
There is definitely a change in flavor when quantum field theory is conveyed
through the path integral. Apart from the union with statistical mechanics, which
immeasurably enriches the subject, it liberates our imagination by making it possi
ble to contemplate virtual but fantastic deformations, such as altering the structure
of spacetime. I am reminded of the classification of things as “gray” or “green” by
Freeman Dyson, in his book Disturbing the Universe (Harper & Row, New York,
1979). He classified physics gray (and I suppose that included quantum field theo
ry,) as opposed to things green, such as poems and horse manure. Ina private letter
dated August 3, 1983, Dyson wrote, “Everyone has to make his own choice of what
to call gray and green. I took my choice from Goethe:
Grau, tenerer freund, ist alle Theorie,
Und griin des Lebens Goldner Baum.
Dear friend, all theory is grey,
‘And green is the golden tree of life.
I must admit that Hilbert space does seem a bit dreary at times; but, with Feynman’s
path integral, quantum field theory has surely turned green.
Kerson Huanc
December, 1997
Marblehead, MassachusettsAcknowledgments
T wish to thank my colleagues and students for helping me to learn the stuff in this
book over the years, in particular Herman Feshbach, from whom I took the first
course on the subject. I thank Jeffrey Goldstone, Roman Jackiw, and Ken Johnson,
who can usually be relied on to provide correct and simple answers to difficult ques
tions; and Patrick Lee, who taught a course with me on the subject, thereby broad
ening my horizon. Last but not least, I thank my editor Greg Franklin for his under
standing and support.
K.H.
xviiQUANTUM FIELD
THEORYCHAPTER ONE +
Introducing
Quantum Fields
1.1 THE CLASSICAL STRING
We obtain a quantum field by quantizing a classical field, of which the simplest ex
ample is the classical string. To be on firm mathematical grounds, we define the
latter as the longwavelength limit of a discrete chain. Consider N + 2 masses de
scribed by the classical Lagrangian
NA
Ua. d= > [FB ZO su? a
FO
where m is the mass and « a force constant. The coordinate q,(¢) represents the later
al displacement of the jth mass along a onedimensional chain. We impose fixed
endpoint boundary conditions, by setting
Gol!) = Gnei() = 0 (1.2)
The equations of motion for the N remaining movable masses are then
mg,—Kqu24+9,)=0 CG +N) (3)
The normal modes have the form
g(t) = cos(wt) sin( jp) (4)
To satisfy the boundary conditions, choose p to have one of the N possible values
™
N+1
Pa= (m=1,...,N) (1.5)2 Introducing Quantum Fields
‘Substituting this into the equations of motion, we obtain N independent normal fre
quencies «,:
(n=1+N) (1.6)
wt= asin
where:
a7
This is a cutoff frequency, for the modes with n > N merely repeat the lower ones.
For N= 4, for example, the independent modes correspond to n = 1,2,3,4. The case
n= Sis trivial, since p = 7, and hence qt) = 0 by (1.4). The case n = 6 is the same
as that for n= 4, since wg = a,, and sin( jpg) = sin(jp,).
When Nis large, and we are not interested in the behavior near the endpoints, it
is convenient to use periodic boundary conditions:
en) = Ghd) (1.8)
In this case the normal modes are
gf)= er“ (19)
For N even, the boundary conditions can be satisfied by putting
2nn N
= (n=081,...# 3) (1.10)
The corresponding normal frequencies are
o2= a3sin( 7) (1.11)
Compared to the fixedend case, the spacing between normal frequencies is now
doubled; but each frequency is twofold degenerate, and the number of normal
modes remains the same. A comparison of the two cases for N = 8 is shown in
Fig, 1.1.
The equilibrium distance a between masses does not explicitly appear in the
Lagrangian; it merely supplies a length scale for physical distances. For example, it
appears in the definition of the distance of a mass from an end of the chain:
.N) (1.12)
The total length of the chain is then defined as1.1 The Classical String = 3
Fixedend Periodic
@p ®o
o4 N Ne2 0 ne
Mode number Mode number
Figure 1.1 Normal modes of the classical chain for fixedend and periodic boundary conditions.
R=Na (1.13)
In the continuum limit
a>0 N+@ (R=Na fixed) (1.4)
the discrete chain approaches a continuous string, and the coordinate approaches a
classical field defined by
qx, 1) = gi) (1.15)
The Lagrangian in the continuum limit can be obtained by making the replacements
, 1) 2
mia af A]
1
ad 7 fe (1.16)
Assuming that the mass density p and string tension «approach finite limits
m
p= (7)
o= Ka (1.18)
we obtain the limit Lagrangian4 Introducing Quantum Fields
Lolly (3587) am
This leads to the equation of motion
F(x, t) 1 Pq@s _
a et ae
0 (1.20)
which is a wave equation, with propagation velocity
(1.21)
The general solutions are the real and imaginary parts of
ge, 1) = eteen (1.22)
with a linear dispersion law
w=ck (1.23)
For fixedend boundary conditions
q(0, t) = q(R, )=0 (1.24)
the normal modes of the continuous string are
Gn(% t) = COS(pt) sin(k,x) (1.25)
with w, = ck,, and
(1.26)
The normal frequencies «, are the same as those for the discrete chain for n/N < 1,
as given in (1.6). However, the number of modes of the continuum string is infinite,
and only the first N modes have correspondence with those of the discrete string.
This is illustrated in Fig. 1.2 for N = 4. Thus, there is a cutoff frequency
oy= a (1.27)
This is of the same order, but not same as the maximum frequency defined earlier,
y= 2cla, for «, is based on a linear dispersion law. The continuum model is an ac
curate representation of the discrete chain only for @ < a.1.2 The Quantum String 5
(aoe AS
LAA, tS
B Lee 8 Soe
Figure 1.2. Normal modes of a discrete chain of four masses, compared with those of a continuous
string. The former repeat themselves after the first four modes. (After J.C. Slater and N. H. Frank, Me
chanics, McGrawHill, New York, 1947.)
For periodic boundary conditions
(0, 1) = 4(R, t) (1.28)
the allowed wave numbers are
ky = (n=0,41,42,...) (1.29)
We obtain the cutoff frequency «, by setting n = N/2.
The highfrequency cutoff is a theoretical necessity. Without it, the specific
heat of the string will diverge, since each normal mode contributes an amount kT.
The value of the cutoff cannot be determined from the longwavelength effective
theory, because only the combination c = aa/7 occurs. Absorbing the cutoff into
measurable parameters, as done in (1.17), is called renormalization. A theory for
which this can be done is said to be renormalizable.
Nonrenormalizable systems exhibit behavior that is sensitive to details on an
atomic scale, Such behavior would appear to be random on a macroscopic scale, as
in the propagation of cracks in materials, and the nucleation of raindrops.
1.2. THE QUANTUM STRING
We now quantize the classical chain, to obtain a quantum field in the continuum
limit. The Hamiltonian of the classical discrete chain is given by6 Introducing Quantum Fields
pp
Hp.) [FE + Fao? (1.30)
where pj = md. The system can be quantized by replacing p; and q, by Hermitian
operators satisfying the commutation relations
[Pp 4] = “15 (3th
We impose periodic boundary conditions, and expand these operators in Fourier se
ries:
1 om
 mui
D> DN Da Oe
1 8
aoe >, Perms (1.32)
where P,, and Q,, are operators satisfying
(PE, Qn) = 18m
PL=P.,
22, (1.33)
The system is reduced to a sum of independent harmonic oscillators:
ve 1 t 1 20r
= ¥ [Lopes
HD, [ag Piet ymei0t
4K ™m
2a A ina 7
o = sim v7) (1.34)
The eigenvalues are labeled by a set of occupation numbers {a,}:
N2
>, nly + 5) (1.38)
2
where @, = 0,1,2, .... The frequency «, is taken to be the positive root of w2, since
His positivedefinite.
In the continuum limit (1.14) the Hamiltonian becomes
_f® 1 @ [ q(x, t) \?
Heont = I a 55m n+ (Ae)  (1.36)
where, with x= ja,1.3. Second Quantization 7
— PO
P(x, t= a
=p 2D
ms (1.37)
The quantum field g(x, f) and its canonical conjugate p(x, 1) satisfy the equaltime
commutation relation
[PG 0, g@', 0) = 16(x  x") (1.38)
Just as in the classical case, we have to introduce a cutoff frequency «,,. General
properties of the quantum field will be discussed more fully in Chapter 2.
1.3. SECOND QUANTIZATION
Another way to obtain a quantum field is to consider a collection of identical parti
cles in quantum mechanics. In this case, the quantum field is an equivalent descrip
tion of the system. Identical particles are defined by a Hamiltonian that is (1) invari
ant under a permutation of the particle coordinates and (2) has the same form for
any number of particles. The quantizedfield description is called “second quantiza
tion” for historical reasons, but quantization was actually done only once.
Let Hy be the Hilbert space of a system of N identical nonrelativistic particles.
The union of all Hy is called the Fock space:
F=U Hy (1.39)
The subspace with N’= 0 contains the vacuum state as its only member. We assume
that N is the eigenvalues of a “number operator” Nop, which commutes with the
Hamiltonian. It is natural to introduce operators on Fock space that connect sub
spaces of different N. An elementary operator of this kind creates or annihilates one
particle at a point in space. Such an operator is a quantum field operator, since it is
a spatial function. This is why a quantummechanical manyparticle system auto
matically gives rise to a quantum field.
For definiteness, consider N nonrelativistic particles in three spatial dimen
sions, with coordinates {r,,... , ry}. The Hamiltonian is
Mez
V2 +H... 8y) (1.40)
where V? is the Laplacian with respect to r, and where V is a symmetric function
of its arguments. The eigenfunctions ¥, are defined by
AY, (0), 0.5 ty) = E(t 5 By) (1.41)8 Introducing Quantum Fields
For Bose or Fermi statistics, V,, is respectively symmetric or antisymmetric under
an interchange of any two coordinates r, and r). The particles are called bosons or
fermions, respectively.
We now describe the equivalent quantum field theory, and justify it later. Let
Urr) be the Schrédingerpicture operator that annihilates one particle at r. Its Her
mitian conjugate '(r) will create one particle at r. They are defined through the
commutation relations
(An), Wor]. = Br— 4’)
(YO), Wr’). = 0 (1.42)
where [4,B], = AB + BA, with the plus sign corresponding to bosons and the minus
sign to fermions. The Fockspace Hamiltonian is defined in such a manner that it re
duces to (1.40) in the Nparticle subspace.
‘A general Nparticle Hamiltonian has the structure
H=D fl) + 3 att» m+ Mo rte (1.43)
where the functions g, h, and so on are symmetric functions of their arguments. The
first term is a “oneparticle operator,” a sum of operators of the form f(r), which act
on one particle only. The second term is a “twoparticle operator,” a sum of opera
tors of the form g(r, r), over all distinct pairs. Generally, an “nparticle operator”
is a sum of operators that depend only on a set of n coordinates. To construct the
Hamiltonian on Fock space, we associate an nparticle operator with an operator on
Fock space, with the following correspondences:
Ysa > far vreau)
Yateany > 5 far din vendors
i
sp [ars Pra drs Uh Whaat
iJek
(1.44)
where for brevity we have written y = Ur), 212 = a(t, ¥), and so on.
As an example, suppose the potential in (1.40) is a sum of twobody potentials:
Vir .. 5 ty) => v(t 5) (1.45)
%
Then the corresponding Fockspace Hamiltonian, also denoted H, takes the form1.3 Second Quantization =
n= fennvrne
2m
1
+ Fen Prove nwrao(r, rdMrMe) (1.46)
The particle number is the eigenvalue of the number operator, defined as
Nop = far Worry (1.47)
By using (2.18), we can verify the relations
Nop#f] = 0
[YAr),Nop] = Yr)
[Y'@).Nop] = Wir) (1.48)
These imply that the action of {(r) on a eigenstate of Ny is to decrease its eigenval
ue by 1, while that of y'(1r) is to increase it by 1. Thus yr) is an annihilation opera
tor, while yt(r) is a creation operator. The vacuum state 0) is defined as the eigen
state of N,, with eigenvalue zero. It is annihilated by all annihilation operators:
ww(r)/0) = 0 (1.49)
By applying y(r) to the vacuum state repeatedly, it is easy to show that the eigen
values of N,, are nonnegative integers.
To demonstrate that the quantum field is equivalent to the manyparticle sys
tem, consider a complete set of states E,N) of the quantum field, which are simulta
neous eigenstate of H and Noy:
HEN)= EJB)
NoglE.N) = ME,N)
We define the Nparticle wave function ‘V(r,,.. .,ry) corresponding to E,N) by
1
Welt. tW) = eel) Wee IEN) (1.50)
which has the correct symmetry with respect to particle permutation. It tells us that
the probability amplitude for finding N particles at the positions r,,..., ry can be
found by annihilating the particles at the respective locations from the state E,N),
and evaluating the overlap between the resulting state and the vacuum state. We
leave it as an exercise to show that this wave function satisfies the Nparticle
Schrédinger equation (1.41). (See Problem 1.3.)10 Introducing Quantum Fields
1.4 CREATION AND ANNIHILATION OPERATORS
The field operator y(r) annihilates a particle at r. That is, it annihilates a particle
whose wave function is a 6 function. Since the latter can be written as a linear su
perposition of a complete set of wave functions, we can express Y(r) as a linear su
perposition of operators that annihilate particles with specific types of wave func
tions. Suppose that u(r) is a member of a complete orthonormal set of
singleparticle wave functions:
fr uteuce) = bux
> uduter’)= 81’)
7
An example of such a set is plane waves:
ul) = yee" (1.51)
‘We can expand the field operators with respect to such a basis:
vee) = 3, wlan,
wo)= > ut(rjal
The coefficient a, and aj are operators that satisfy the commutation relations
[ay abe = Bix
[aps ag]. = 0 (1.52)
where the + sign is for bosons and the — sign is for fermions. These relations follow
from (1.42) and the orthonormality of the functions u,(r).
It follows from the commutation relations that, for each , the eigenvalues of
a,'a, are integers n,, called the “occupation number of the singleparticle state k”:
ata\n) = nln)
(nlm) = Bim (1.53)
where we have omitted the label k for brevity. The allowed values of the occupation
number are given by
nal %b2..4% Bose statistics)
0,1 (Fermi statistics)1.4 Creation and Annihilation Operators. 11
The actions of a and at have the following results:
an) = Vnn — 1)
a‘n) = V1 £n)n+ 1) (1.54)
where the +sign corresponds respectively to Bose (+) and Fermi (~) statistics, which
show that a annihilates a particle in the state with wave function u(r), and a‘ creates
such a particle. We leave it as an exercise to derive these basic results. (See Problem
1.2)
The state 0) corresponding to n = 0 is the vacuum state, which satisfies
ald)=0 (1.55)
We assume that it is normalizable:
(0/0) = 1 (1.56)
Obviously all other states can be obtained by creating particles from the vacuum:
(1.57)
‘We can simultaneously diagonalize aja, for all k. The eigenstates are then la
beled by a set of occupation numbers {1, m,, .. . }, and they constitute a basis for
the Fock space. The total number of particles present is N= 2,1. We have
alaylto, «5 My.) = Meloy «= 5 Mis»)
Alto, «5 Mey) = (DV, «41,
allng, «  5 My )=(CLVTE ml,  5m +1...)
where
{S (Bose statistics). (1:58)
2p ®), with periodic
boundary conditions. Expand the field in terms of annihilation operators a, for freepar
ticle states of momentum k, and show that
kK 1
HD ay alent 2A 2, Raha
=S dr e&ty(r).
1.5. Consider a system of N nonrelativistic electrons and N positive ions with Coulomb in
teractions, enclosed in a periodic box of volume (2. The Hamiltonian is given by
& pt 7 & @ ee 4P
Pi i
H=> Piay > > yt
2 2m * 2 Ae Tn R) 2 IRR)” 2M
The ions are heavy. Hence consider R, to be fixed numbers, neglect P,, and drop the last
two terms,
(a) Label singleelectron states by momentum k and spin s, designated collectively as
a= {k,s}. The corresponding wave function isProblems 15
ef) et)
(b) To go to the quantizedfield representation, replace one and twoparticle operators
by the rules
x
2ke) > (one) ahag
Dae FD aap MaBolnayad
& By
(© Define Fourier transforms:
2 Bur 4tre?
Sexe AME ety R
(is FR OO FR
¢) = 8 Janene
Ane?
[ks — Kil
Buy 8
ee aklts + ky — ka — ky)
where 5, is the Kronecker delta.
(@ Obtain the Hamiltonian in quantizedfield form:
re ne? 1
He DF Mets ED Ds Cn tee) Apes
2 Pak 55"
Ane? 1 SS aR!
a XS Bae
(©) Show that the second term gives, for small k,
2re* 2me?N(N  1
Be LL ote ney
which is divergent at k = 0. Show that the divergent term proportional to N? is can
celed by the k = 0 limit of the third term. The O(V) term above remains divergent.
The source of this divergence is the periodic boundary conditions, by which the set
of coordinates {r},..., ry} is being repeated an infinite number of times in space.
‘Consequently, the Coulomb energy of an electron diverges, due to longrange inter
actions with an infinite number of distant copies.
(f) itis clear that this is a mathematical artifact, and to avoid it we should treat the ions
dynamically; but we do not wish to add that complication. The expedient way out is
to simply leave out the k = 0 contribution in both the second and third terms. Hav16 Introducing Quantum Fields
ing done this, we can ignore the third term altogether, because it sums to zero under
the assumption thatthe fons are uniformly distributed in space, Thus we take as ef
fective Hamiltonian
ae YF tes Ye (Apri sg bes) Mp sAgs”
fe Sat
Keo
This describes electrons immersed in a uniform positively charged background that,
makes the whole system electrically neutral.
1.6 Imposing periodic boundary conditions means filling infinite space with identical cells
that contain copies of our system. This problem illustrates the effect of longrange inter
actions among the cells. Consider a unit point charge at the center r= 0 of a cubic cell of
volume Z?, which contains a uniform negative charge density, so that the total charge in
the cell is zero. Impose periodic boundary conditions, and calculate the potential in the
neighborhood of the unit point charge.
(a). Show that the potential is given by
4nw et 2m,
os i
O° Be B
by showing
vere =4ed 8(r—nL)— *
(b) Forr/L <1, show?
Yn)= 1 5 7 +0)
= 2.837297 «+ 
REFERENCES
1, K. Huang, Statistical Mechanics, 2nd ed., Wiley, New York, 1987, Appendix.
2. K, Huang and C. N. Yang, Phys. Rev. 105, 767 (1957).
3. Liischer, Commun, Math. Phys. 105, 153 (1986).
4, C.N. Yang, Chin, J: Phys. 25, 80 (1987).
2Huang and Yang (2] first calculated c, but gave an incorrect value, 2.37. The correct value was ob
tained by Liischer [3]. An elementary derivation was given by Yang [4].Scalar Fields
2.1  KLEINGORDON EQUATION
A fast way to go from classical mechanics to quantum mechanics is to replace the
energy and momentum of a particle by operators, according to the prescription
Eait
a
p>iV (2.1)
Making the replacements in the nonrelativistic relation E = p?/2m, and applying the
result to the wave function, we obtain the Schrédinger equation:
1 a
> Vr, ) =i Wr, 1 2.2)
Im YE =i we (2.2)
Of course, this is not covariant under Lorentz transformations. For a covariant equa
tion, we use the same trick on the relativistic relation E? = p? + m?. The result is the
Klein—Gordon equation
(CP + m?) W(x) =0 (2.3)
where x stands for x# = (f, x), and
a
P=09,3V? (2.4)
in units with c = 1. Assuming that y is invariant under Lorentz transformations, we
have a covariant equation—one that has the same form in all Lorentz frames. What
is not clear is how to interpret yx).
By analogy with the Schrédinger equation, we might interpret ¢{x) to be the
1718 Scalar Fields
wave function of a relativistic particle. That would require the existence of a 4vec
tor probability current density j#, which should be conserved: 4,, j# = 0. However,
the obvious choice jy = #*ys is untenable, for ¥*w is Lorentzinvariant by assump
tion, and cannot be part of a 4vector.
As in the case of the Schrddinger equation, we can construct a conserved cur
rent as follows. Multiply (2.3) from the left by y* to obtain
(YA OW) — (YAY d,s) + mys = 0 (2.5)
Subtracting this from its complex conjugate leads to
4, j= 0 (2.6)
where
JE = Yon — Yroey (2.7)
However, the time component
ay Oe
jo = yr—— — ——_,
Pea a @8)
is not positivedefinite, and therefore cannot be a probability density.
The root of the difficulty lies in the second time derivative in the Klein—Gordon
equation. As we shall see, this leads to negative frequencies corresponding to an
tiparticles. The relativistic kinematics makes it impossible to have a oneparticle
theory, We shall regard 4(x) not as a wave function but as a classical wave field, and
as such should be quantized.
2.2 REAL SCALAR FIELD
Consider a real scalar field ¢(r, ‘), which is invariant under Lorentz transforma
tions. The current j# vanishes identically in this case. We enclose the system in a
large periodic box of volume 2, and expand the field in a Fourier series:
d= va > aude" (2.9)
where
GO = G4) (2.10)
because the field is real. Assuming that 4(r, £) satisfies the KleinGordon equation,
we have2.2 RealScalar Field == 19
Gxt fq, =0 (2.11)
where ,
oR =k +m? (2.12)
The system is equivalent to a collection of harmonic oscillators, and may be quan
tized by imposing the commutation relations
GLO), gu(0)] = uae
(4x(0), 4ue(0)] = 0 (2.13)
where qi(0) is the hermitian conjugate of q,(0). This fixes the normalization of the
field, left arbitrary by the KleinGordon equation.
Since the KleinGordon equation is invariant under time translation, the origin
of time is arbitrary. The commutations relations in fact hold at any time f:
GO, WO] = Sue
[als ae (= 0 (2.14)
which are equivalent to
de, 0, 60°, 0] = Fr’)
[h(r, 1), dr’, N= 0 (2.15)
These are called equaltime commutators, which serve as initial conditions for the
equation of motion. The unequaltime commutators must be calculated from the so
lutions, and contain dynamical information.
In the present freeparticle case, the equation of motion (2.11) is trivial to
solve. For given wave number k there are two frequencies +, with
Oy = +V + (2.16)
Taking into account the reality property (2.10), we write the solution in the form
1 rte
= aye + atyelow ;
a = Foeelne™ + abel] 17)
where a, and aj are operators, with commutation relations determined by (2.14):
Lays aL = Ba
[ax ay] =0 (2.18)20 Scalar Fields
The normalization factor (2w)~'? in (2.17) is chosen to make the commutators sim
ple. We recognize that a, is an annihilation operator, and aj a creation operator for a
boson, as introduced in Section 1.4. The timedependent quantizedfield operator
can be represented in the form
H.S agian talemreny 19)
x Vie,
The positivefrequency part (the first term) annihilates a particle, and the negative
frequency part creates a particle. The negativefrequency part is absent in a nonrela
tivistic field, because the kinematic relation E = p?/2m allows only one sign of the
frequency.
2.3 ENERGY AND MOMENTUM
Analogy with the harmonic oscillator suggests that the Hamiltonian of the free
scalar field should be
1
H= 7 dul? + lq?) (2.20)
In terms of the field (x) = (r, 2), it has the form
= [ar 2f(x)
e)=5](Fe) +19 ¢r +e  (2.21)
where (x) is called the Hamiltonian density. The Lagrangian of the system can be
obtained through the general relation L(q, q) = pq —L(p, 4):
L= fers (2.22)
where £(x) is the Lagrangian density given by
£@)=+ (4)  Vor mae] 3 edad i 223)
The last form shows that the Lagrangian density is Lorentzinvariant. In contrast,
the Hamiltonian density, which is an energy density, cannot be invariant. For this
reason, relativistic theories are usually specified via the Lagrangian density.
In terms of creation and annihilation operators, the Hamiltonian takes the diag
onalized form23 Energy and Momentum = 21
H=> wala, + +) (2.24)
©
The zeropoint energy diverges unless there is a cutoff; but the cutoff has no physi
cal relevance, since the energy of any state relative to that of the vacuum is indepen
dent of it. The energy of a particle is given by
0, = Vie +m? (2.25)
where k is its momentum and m is the rest mass. Accordingly, the total momentum
operator is
P=> kaja, (2.26)
f
According to the general principles of quantum mechanics, the Hamiltonian is
the generator of time evolution, through the Heisenberg equation of motion:
aH, 6¢¢, 9) AED 227)
‘The formal solution yields
Hr, t) = "g(r, Oe (2.28)
For consistency, we must show that this is consistent with the KleinGordon equa
tion, which we used to arrive at the solution (2.19). Substituting (r, 0) from (2.19)
into (2.28), we obtain
dr, = ae agent ale Te ttte (2.29)
For the freefield Hamiltonian, we have (see Problem 2.1)
eget = queria (2.30)
This demonstrates that (2.28) is the same as (2.19).
Again, according to general principles, the momentum operator P should gen
erate spatial translations:
=P, Hr, )] = V dtr, 2) (2.31)
with formal solution
Hr, ) =e? *H(0, DeP* (2.32)
Asa straightforward calculation shows, this is the same as (2.19).22 Scalar Fields
2.4 PARTICLE SPECTRUM
The vacuum state 0) is the state of lowest energy, defined ty
a(0)=0 (all k) (2.33)
and normalized to
(00) = 1 (2.34)
A oneparticle state is defined by
Ip) = afl0) (2.35)
Using the commutation relations, we find
aylp) = 444310) = (ayay + Byp) 0) = dypl0) (236)
The field operator has nonvanishing matrix elements only between a oneparticle
state and the vacuum:
loro)
<0 !p) = —=— (2.37)
Oo = aa
This is the wave function of a particle of momentum p, normalized to a density
(2e,0)". By successively creating particles from the vacuum, we can build a com
plete set of states:
Vacuum: — 0)
1particle states: Ip) = a,*0)
2particle states: [PiP2) = ap, "ap, "10)
2.5 CONTINUUM NORMALIZATION
In the limit © — ©, the allowed values of k approach a continuum, and we can
make the replacements
I &k
azo! Qn2.5 Continuum Normalization — 23
DB —(277H(kk’) (2.38)
We define continuum versions of the annihilation and creation operators by putting
atk) = Oa,
at(k) = Oat (2.39)
The commutators then have limiting forms:
[a(ho, a%(k’)] = 22P H(k k’)
[a(k), a(k’)] = 0 (2.40)
The field operator can be represented as a Fourier integral:
Pk 1
= kT0K) + qt(Ik)e lerox0
(rr, 1) J On? Vie [a(k)e" +al(kje“ ) (2.41)
As before, the vacuum state 0) is defined by a(k)0) = 0 with (00) = 1, and a
oneparticle state is defined by
Ip) = at(p)l0) (2.42)
Using the commutation relations, we find
a(k)p) = (277)8°(p — k)/0) (2.43)
The singleparticle wave function is
elorep)
0 = 2.44)
(ovb0anw) = Fae 4)
with a particle density (2w,)'. The normalization is such that the number of parti
cles in volume element d*r is the Lorentzinvariant combination d°r/(2w,). The
Hamiltonian and total momentum now take the forms
=(e Gap one toatl)
Bk
P= Jeeps ko") (2.45)
The choice between discrete or continuum normalization is a matter of nota
tion, since we always regard as large but finite in intermediate steps of calcula24 Scalar Fields
tions. The limit (2 — © is taken only in final answers. This is done to avoid irrele
vant concerns about mathematical rigor, such as how to define the Hilbert space
when the dimensionality is noncountably infinite. The continuum normalization
merely anticipates this limit
2.6 COMPLEX SCALAR FIELD
‘A complex scalar field is just two real scalar fields constituting the real and imagi
nary parts. What is new is a symmetry between the two fields, and this leads to a
conserved current. In physical terms, a complex field can have electric charge,
whereas a real field must be neutral.
‘We denote the complex scalar field by y(r, 1), and decompose it into real and
imaginary parts in the form
1
= qld + is) (2.46)
The Lagrangian density is taken to be
L(x) = HU ()A,. Wee) — mp)
2
DL (2) 52) — mbox) h(x] (2.47)
The normalization factor 1/V2 in (2.46) is chosen in order that ¢ has the same nor
malization as the real scalar field discussed previously.
To quantize the system, we impose the commutation relations
Tbe, , H(e", )) = 8,F01')
[hr 1), (0, 0] = 0 (2.48)
The complex field yr, 1) becomes a nonHermitian operator satisfying
Wr, ), WO", O)= 82’)
(Hr, 9, He’, 1 = LH, 0, Wr’, O1= Lr, ), Wr’, D]=0 (2.49)
Thus, the canonical conjugate to wis y*.
In accordance with (2.19), we can expand ¢; in terms of annihilation and cre
ation operators:
1
Tage anor + ahewerond) (j= 1,2) (2.50)
V20,
605.0" Se2.7. Charge and Antiparticle 25
with the commutation relations
[ats afi] = 8B.
[ans a']= 0 (2.51)
In the complex representation, we have
1 1
Ur, t)= Var Via, OH
ellen) + oifeitero4)
We )= Wad Vim (ble ro10 + celeron (2.52)
where
1
b= lain tn)
 
= alan —idy,) (2.53)
with commutation relations
[bus O53] = Ske
Lew ¢B] = xp
[bus Bp] = [us Cp) = [us ep] = 0 (2.54)
The total energy and momentum can be expressed as follows:
H=> ox(ahyay, + aban) = >, oy(bhdy + chew)
v. i
P=S ghklahon+aand= F kbli tle) 58)
There are two type of quanta, which can be designated either as a, and a, quanta, or
as b and c quanta. The energy and the momentum do not distinguish between these
descriptions. We shall see, however, that only the b and c quanta have definite
charge.
2.7 CHARGE AND ANTIPARTICLE
The current density for the complex scalar field is given by
IH= Wonk — ROW = 3 (dod — br) (2.56)26 Scalar Fields
which satisfies the conservation law 4, j# = 0, or
a jo 
ah tVi0 (2.57)
Integrating both sides over the spatial volume, we obtain
a d
By. yyy = 4 (gap) =
fa at (x) a fa ‘xf(x) = 0 (2.58)
or
do
a7 (2.59)
where Q is the total charge operator
0= faxPe)
= TeatheeT
=> @hahahaw)
K
=> Ox  chew) (2.60)
7
‘As we can see, a b quantum carries one unit of positive charge, and a c quantum car
ries one unit of negative charge. The aj, a) quanta, which are linear combinations of
those of b and c, do not have definite charge. By convention, we refer to a c quan
tum as an “antiparticle.” Thus, the positivefrequency part of annihilates a parti
cle, and its negativefrequency part creates an antiparticle. Similarly, yt either cre
ates a particle or annihilates an antiparticle. In light of this, we can say that for the
real field, the particle is its own antiparticle.
The term “charge” is used in a generic sense, and does not necessarily mean the
electric charge, since we have not turned on the electromagnetic coupling. The unit
of charge is arbitrary, because j* is defined only up to a multiplicative constant.
It is straightforward to verify that charge is conserved:
(0, H]=0 (2.61)
This implies that the number of particles NV, minus the number of antiparticles N_ is
a constant of the motion. In the freefield case, this conservation law is trivial, since
N, are separately conserved. It becomes significant when, in the presence of inter
actions, N, are no longer conserved. In that case, N, — N_is still conserved as long
as (2.61) holds.2.8 Microcausality 27
2.8 MICROCAUSALITY
A classical signal propagating according to the KleinGordon equation has a group
velocity
= 9% Ik!
Perour a * Vitam (2.62)
which never exceeds 1. This means that events at two spacetime points lying out
side of each other's light cone (or separated by a spacelike interval) cannot influence
each other. In the quantum theory, this means that two field operators at points sep
arated by a spacelike interval must commute with each other:
[d), 6@']}=0 if (ex'P <0 (2.63)
This condition is called microcausality. We must verify that our quantized field the
ory satisfies this condition.
To compute the commutator in (2.63), note that it is a cnumber,! and at fixed
x’ it satisfies the KleinGordon equation, because ¢(x) does. The initial condition at
Xo = x5 is the equaltime commutator (2.15), which is a cnumber. Therefore the
commutator remains a cnumber at all times, and we can equate it with its vacuum
expectation value:
[4(x), $0)] = (OGG), (110) = AG y) (2.64)
This defines a Lorentzinvariant correlation function A(x — y), which depends on x —
y, and not on x and y separately, because of the translational invariance of the vacu
uum state. (See Problem 2.1.) We use the expansion (2.41) to obtain
ar oyjoy = { —22E— eater
Ike, DON) =  So,
Pk wy
(odode, 010) = [ aay ewer (265)
Subtracting one from the other, we have
Pk sind) ee
Ga¥ a (2.66)
A(x) = i(O1L4(r, 0), (0110) = 
4A cnumber is “classical” number, a multiple of the identity operator.28 Scalar Fields
Since A(x) is Lorentzinvariant, it can depend only on the invariant
ex : (2.67)
If.x is spacelike, for which x? < 0, we can put = 0. By (2.66) this gives A(x) = 0. =
The proof of microcausality depends on the initial condition from the commu
tator in (2.15), which quantizes the system according to Bose statistics. Had we
used Fermi statistics by replacing commutator with anticommutator, microcausality
would have been violated. The particles here have spin 0, since there does not exist
discrete degrees of freedom corresponding to spin. Our result is partial demonstra
tion of the spinstatistics connection, which states that particles with integer spin
are bosons, while those with halfinteger spin are fermions. The second half of the
statement will be shown in Chapter 6 on the Dirac field.
2.9 THE FEYNMAN PROPAGATOR
The propagation of a free particle in the vacuum can be described by the correlation
function
AO (Xp — x1) = =H) ()10) (2.68)
in which y*(x,) creates a particle from the vacuum at x, which is annihilated by
Wx) at x2. This makes sense physically when , > f,. Similarly, the correlation func
tion
Ae, =x) = =O] Ce) O2)10) (2.69)
describes the propagation of a test antiparticle from x, to x,, and is physically mean
ingful when f, > 4. To obtain a correlation function that has physical meaning, we
use either A‘ or AO depending on the sign of the relative time. The result is the
Feynman propagator, or causal propagator:
Mex — x1) = HKO THO) WCx,)/0) (2.70)
where the timeordering operator T rearranges the operators, if necessary, such that
the operators stand in such order that time increases from right to left:
A(ty)B(t) if >t,
ae eas if ney ee
If 4 > f,, the Feynman propagator describes the propagation of a particle from x, to
%; if fy < ty, it describes the propagation of an antiparticle from x, to x;. This is the2.9 The Feynman Propagator 29
basis of Feynman's famous remark: “An antiparticle is a particle traveling back
wards in time.”
To calculate the propagator, we start with the expression
ae OldAx)yit(0)0) if x°>0
Arle) = { (Old (O)ykx))0) if x°<0 7)
and insert a complete set of states between the operators. Since the field operator
connects the vacuum to oneparticle states only, we have
(OlWCx)Ik) {kly*(0)0) if, 29> 0 (2.73)
A) =} Oop { (OW COVK) (KiyoOI0) if x°>0
Using Wx) = e”*yA0)e*”*, and changing the integration variable from k to —K in
the lower formula, we obtain
Ap(x) = =i f ei YA0)0)/Pe* reread (2.74)
Quy "
Using the following integral representation
cd ike
ioe = 2 f ee
ds
Gooltm (1) (2.75)
we obtain
ArQ) = fe On 2 lOWOOP Se (n> 0°) (2.76)
From (2.44) we have
Koimonone = = em
os
Therefore
=
&y)= S tn) 278)
Operating on both sides of this equation by (? + m? gives
(CP + mi?) p(x) = 54x) (2.79)
This shows that the Feyriman propagator is a Green's function of the KleinGordon
equation.30 Scalar Fields
The Fourier transform of the Feynman propagator is
5 1 :
\0 Bate 2.80)
which has poles at
k= VE (2.81)
corresponding to a particle or antiparticle of mass m. The residue 2c4{(0y(0)0)? =
1 reflects the normalization of the wave function. We may view the propagator as
the propagation amplitude of a virtual particle of 4momentum k“. The virtual parti
cle, whose squared mass #? ranges between —~ and ©, becomes a real particle when
it “goes on mass shell,” at (2 = m?.
To obtain Ay(x) explicitly, we integrate over the angles of k in (2.74) to obtain
2 sin kr
Ais) = G5 al a= Oe (2.82)
By Lorentz invariance A,(x) can depend only on
sex=PP 2.83)
For s > 0, we can put r = 0 to obtain the representation
a@)= of “ae envi" _ mV) (6>0) 2.84)
4 wy 8aVs
For s <0, we put = 0 to obtain
R sink V=s
dee= ef eS — = K(mV=) — (s<0) (2.88)
where H{" and K, are Bessel functions. In the timelike region s > 0 the function de
scribes an outgoing wave for large s. This corresponds to the in prescription in
(2.80). The in prescription would have yielded an incoming wave. In the spacelike
region s <0 it damps exponentially for large s. On the light cone s = 0 there is a
deltafunction singularity not covered by the preceding formulas:
Jim Axo) =F Tate) (2.86)2.10 The Wave Functional 31
2.10 THE WAVE FUNCTIONAL
In quantum mechanics, the coordinate representation is defined by basis states r)
satisfying
Toplt) = r[r) (2.87)
with the basic commutator realized through the replacements
Top (cnumber)
Pop > iV (2.88)
A state A) is represented by the wave function.
Walt) = (r/A) (2.89)
and inner products are defined by
(AB) = [aru ste) vale) (2.90)
In the analogous field representation in quantum field theory, we diagonalize
the field operator, thus representing it by its eigenvalue, which is a cnumber func
tion, For a real scalar field ¢,,(r) at a fixed time r= 0, we denote its eigenstates by
\):
dop(H)IP) = H1)4) (2.91)
where the eigenvalue (1) is a realvalued function of r. The commutation relations
(2.15) are realized through the replacements
dop(t) > a (cnumber function)
ae
al) 5 92)
where 8/84(r) is denotes the functional derivative with respect to the value of the
function gat r.
A state A) is represented in the field representation by the wave functional
Wyld] = (pla) (2.93)
which is a complexvalued function whose argument is a function; that is, its value
depends on the form of the function. Inner products between wave functional are
functional integrals:32 Scalar Fields
(He ¥0)= [D6 VIL6oI1 2.94)
where Dé denotes the measure on the space of functions. Writing the Hamiltonian
in the field representation, we have the Schrédinger equation for the wave function
al:
IV, #]
zl gag VHP +more] WEA a 295)
2.11 FUNCTIONAL OPERATIONS
We digress on functional operations on a functional F[¢]. First, the functional de
rivative 5F[4]/6¢(x) is defined as follows. We make a small change d—> # + 8¢,
where the function 8¢(x) is different from zero only in the neighborhood of x. Then
the functional derivative is given by
SIG] _ , [ Fld + 66] FId]
B(x) “ke 5px) ] 2.96)
To calculate any functional derivative, we need the elementary functional derivative
5(x)/8(y), which is obviously proportional to &(x — y). To determine the propor
tionality constant, we replace the continuous space of x by a lattice of spacing a, and
denote by ¢; the value of the function on site j. Clearly,
9b,
by og
se = 88 297)
Inthe continuum limit
Y safer (2.98)
5
5, > ade —y)
we have
Sd(x)
= . (2.99)
Bao) ED 2.99)
With this formula, we can calculate a general functional derivative. As illustration,
take FT] = J d?y{V h(y)?. Then2.12 Vacuum Wave Functional 33
FAS »,
In either of the preceding methods, the integral by itself may not have a contin
uum limit; but matrix elements of the form
(2.103)
usually has a definite continuum limit.
2.12 VACUUM WAVE FUNCTIONAL
We now calculate the wave functional for the free vacuum state. First let us express
the annihilation operator a(k) in the field representation. From (2.41) we have, at
t=0,
Jarre de) a [a(k) + at(k)]
7"
Jar erste) =1 tea at()] (2.104)34 Scalar Fields
Solving for a(k), we obtain
a(k)= Vaal e™*[0,4(0) + 16(0)] (2.105)
Now we write
ey = VIE + ne ht = VV im eet (2.106)
so that
a(k) = Tix. farlo@v=F +m + id(e*
_ vag lO + md(n)] + iG(n) fer (2.107)
The last step is obtained by expanding V—V? + m in a power series in V2, per
forming a partial integration in every term, and summing the series again. The sur
face integrals generated in the partial integrations vanish as a result of periodic
boundary conditions. Replacing id(r) by 5/5¢(r), we obtain
1
atk)
x far em [Vivre Fm b(n) + wel (2.108)
The virtue of this representation is that the Fourier coefficient in the integrand is in
dependent of k.
The wave functional of the free vacuum satisfies the equation
a(k)Vo[ 4] = 0 (2.109)
Thus, it must be annihilated by the Fourier coefficient in (2.108):
é
[v=o FH oH + sas Moldl=0 2.110)
The solution to this equation is
Wolw] = C exp[+ far $n) V=V7 = Fd) (2.111)
where C is a normalization constant. This gives the probability amplitude that the
field has the functional form ¢(r) in the vacuum state. The relative probability for
the field to have a functional form lying in the neighborhood of ¢ in the volume el
ement D¢ of function space is213 The ¢*Theory 35
IWlelPD¢
The most probable form is ¢ = 0, and deviations from it occur with a Gaussianlike
distribution.
‘The exponent in (2.111) can be rewritten in different forms. Introducing the
Fourier transform
Hk) = J Br (nr) (2.112)
we can write
Jar deyVV FHP on) = [ee am VE Fre do?
= [a'r ar $Kee—r') dee") 2.113)
where
Kw) = [eee = eer a me (2.114)
For a complex scalar field, there are now two coordinates, which can be taken
as either {¢,, 2} or {y, y+}. Inner products of wave functionals now take the form
(Ya Va) = fos, Do2V5ltos boVal 2. $2] (2.115)
or equivalently
(Va, Va) = Jowywau, wel w] (2.116)
The complex measure is defined in terms of the real and imaginary parts:
Di Di = Dd, Dd (2.117)
The vacuum wave functional for the free complex field is just the product of those
for the two independent real fields. Reexpressing the result in terms of the complex
field, we have
Wold] = Cexp[f a'r vA(VV? + mF U)] (2.118)
2.13 THE ¢* THEORY
As the simplest example of an interacting field theory, consider the Lagrangian den
sity of the socalled $* theory:36 Scalar Fields
L(x) = 5 bd, — 3? P  gdb* (2.119)
where g > 0. The quartic term makes the equation of motion nonlinear:
(P+ m)p+ge=0 (2.120)
The Hamiltonian density takes the form
He) =3](F) 1797] + Moe)
1
V(b) = ym dx) + gb") (2.121)
which suggests that V((x)) is a potential.
To quantize the theory, we impose the equaltime commutators (2.15), which
can be satisfied by taking as initial condition
= Ly jane + ate
or, 0) vad Valance + ale *
ety i,
S Vad 7a oan + a") (2.122)
where the creation and annihilation operators aj, a, are defined by the commutation
relations
[ay ah] = Bx
[a,, ay] = 0 (2.123)
The equation of motion is not soluble unless g = 0. We can always write, as a formal
solution,
(r,t) = edn, Oe (2.124)
but this is not simple unless g = 0.
To see the effects of the interactions, separate the Hamiltonian into a “free”
term and an “interaction” term, at some arbitrary time t = 0:
H=Ho+ Hix (2.125)
where
Jarno, 0 +1V 4, OF]Problems 37
Hige= g[°rd%(r, 0) (2.126
In terms of the creation and annihilation operators we have
Hy=Co+ >, wyalay
c
Hoy = & + Sxl He tH AMO Gt saya} +a,Nal+asyal +a,)
where Cy is an irrelevant zeropoint energy and 6, denotes the Kronecker 5. We use
the shorthand 0; = a,, 4) = ay, and so on. The interaction Hamiltonian Hix, de
scribes fourparticle processes that conserve momentum. Substitution of this expan
sion into (2.124) generates a complicated series for the timedependent field opera
tor. We shall learn how to organize such terms in a systematic manner in Chapter 9.
PROBLEMS
2.1 SpaceTime Translation Consider a free scalar field (x), which can be expanded in
terms of the annihilation operators a,. This problem illustrates the fact that the 4mo
mentum P+ = 3,é4a.a, is the generator of spacetime translations.
(a) Asa useful tool show that, for two operators 4 and B,
Be“ =B+ (4, B]+ 314,14, B+ 5
(b) Use this formula to show
Page = ayes
and the infinitesimal form
[P#, ay] = kay,
(©) Establish that P# is the generator of spacetime translations by showing
(PH, (x)] = thx)
(@)_ Let K) be an eigenstate of P#, satisfying PH\K) = KK) . Show that this state is
translationally invariant:
(KI) PODIK) = (Kb ~») AOI)
2.2. Charge Conjugation The designation of particle and antiparticle is a matter of con
vention, and we can freely reverse the labels. More specifically, for a complex scalar38 Scalar Fields
field (x), construct an operator C’that takes by to c, and vice versa, and commutes with
Hamiltonian
CYQ)C = Yi(x)
[4, C]=0
cc=1
Cel
The operation C is called charge conjugation, or particleantiparticle conjugation.
2.3 Lorentz Invariance We have calculated the function A(x) = (O\[¢(x), #(0)}0) in
(2.66), but not in a manifestly Lorentzinvariant form. Show that it can be put into the
desired form
{4
A(x) = anil @ay et8(12 — melo)
2.4 Spin and Statistics  Quantize the real scalar field according to Fermi statistics; in oth
er words replace the commutators in (2.15) by anticommutators. Show that this will vio
late microcausality.
2.8 External Source Consider a real scalar field ¢(x)\coupled to an external source func
tion J(x), with Lagrangian density
L)= FHbIG+ amide + Jb
(a) Obtain the Hamiltonian in terms of the creation and annihilation operators a4, aj,
for planewave states.
(b) Suppose that the source is static, that is, that J(x,t) is independent of t. Using per
turbation theory, show that there is no scattering from the fixed source to second
order.
(©) Show that there is no scattering at all, to any order. (Hint: Show that a linear canon
ical transformation of ay, reduces the Hamiltonian to the sourcefree case.)
2.6 Level Shift Suppose that the external source in the previous problem consists of a sin
gle static point source: J(x) = g%(r).
(a) Calculate the change in the energy of the vacuum state to order g?. The result will
be a divergent integral. Cut it off at a large momentum A. This illustrates a proto
type of divergence in quantum field theory.
(b) Show that all levels of the system shift by the same amount and therefore that the
divergence in this case has no physical relevance.
2.7 Yukawa Potential Continuing with the last two problems, suppose the source function
U(x) consists of two static point sources located at ry,r2:
Jr, = g(r ~ 11) + Hr ~r2)]
Treating g as a perturbation, calculate the change in the vacuum energy to second order
in g, and show that there is an attractive potential between the two point sources:28
Problems 39
where R = r, ~ r9. This is the Yukawa potential, originally proposed as the potential be
tween two nucleons due to interactions with scalar mesons.
‘Vacuum Fluctuations Consider a free scalar field in a large periodic box of volume
. Let the Fourier transform be denoted
@)
)
©
@
de = Fz Jarre)
Show that the vacuum expectation value of G(k) is zero.
By expanding the field in terms of creation and annihilation operators, show that
the meansquare fluctuation of the Fourier transform is given by
5 1
(OP W00) = Saree
‘The meansquare average can be expressed in the field representation as.
foowaovates
(016(&)0) =
foovae
where V4{ ¢] is the wave functional ofproduce the last result from this formula.
Calculate the meansquare fluctuation {0¢7(x)0) in coordinate space. The result is
divergent because of the highmomentum modes. Exhibit its dependence on the
cutoff momentum A.Relativistic Fields
3.1 LORENTZ TRANSFORMATIONS
Relativistic quantum fields can be classified according to the way they transform
under Lorentz transformations. More specifically, they transform according to irre
ducible representations of the Lorentz group. The different representations give rise
to particles with different values of the spin angular momentum.
‘According to the principle of special relativity, the laws of physics should be
covariant with respect to Lorentz transformations; that is, they should have the same
forms in all reference frames connected by Lorentz transformations. The simplest
Lorentz transformation is a “boost” of the reference frame with velocity v along
some axis, say, the x axis:
GB.)
This may be supplemented by a rotation of the coordinate system, say, about the z
axis through an angle 6:
x’ =x cos6+y sind
y' =x sind + y cosd (3.2)
Defining a boost “angle” by
coshg = 1/V1— 22
sinhd = V1 —0 (3.3)
403.1 Lorentz Transformations 41
we can write the matrices of these transformations as follows:
cosh sinh 0 0
; ~sinh ¢ 0 0
Lorentz boost: $ coshd 1 oO}.
0 001
1 0 00
Rotation, [2 C088 sind 0 G4)
0 sin@d cos@ 0
0 0 oo.
The inverses of these matrices can be obtained by reversing the signs of # and 6.
The rotation matrices are orthogonal matrices, while the Lorentz boosts are not, be
cause the invariant form — x? for the Lorentz boost is not positivedefinite.
The angles of rotation are not additive, unless the rotations are all made about
the same axis. Similarly, the velocities of successive Lorentz boosts are not additive,
unless the boosts are all made along the same direction.
We use a relativistic notation in which the coordinate 4vector is denoted by
x= (1, r) and the metric tensor is diagonal:
@.5)
1
A general Lorentz transformation is a linear transformation A on x that leaves
2? = P — r? invariant:
xH= AM XY (3.6)
with the requirement
Suv AMaA"s = Sap @B.7)
which ensures the invariance of x*. In shorthand, we write the transformation in the
form
x=Ax (3.8)
The transformations above form the continuous Lorentz group, which is character
ized by six parameters: three velocity components and three angles of rotation. As
we can see from (3.4), they are represented by matrices with determinant +1. In
contrast, the discrete transformations42 Relativistic Fields
Spatial reflection:
G9)
Time reversal:
have determinant —1. These discrete elements together with the continuous Lorentz
transformations form the general Lorentz group. We shall reserve the name
“Lorentz transformation” for the continuous Lorentz transformations.
Any element of the Lorentz group can be built up from infinitesimal ones, with
the general form
At, = gt, + of, (3.10)
We write in shorthand
A=l+o G.I)
Lorentz transformations generally do not commute with one other; but the infinites
imal transformations do, because their commutators are of secondorder smallness:
(1+ @)(1 + @) = 1+ @ + @ + Ow?) (@.12)
Thus, group multiplication is equivalent to addition of the ws.
An infinitesimal transformation of the coordinate system, characterized by
boosts with velocities v/ along the x/ axes, and rotations of angles 6* about the x*
axes, is described by the tensor
0 vt =P
0B
8 0 6.13)
# 0
By raising the lower index, we obtain an antisymmetric tensor
of 2 8
= ms
co gireon,= 2 0 3.14)
2 8 0 6
v7 P 6! 0
whose elements can be summarized as follows:
= ke
ol =
coll =e! = —eb'ge (3.15)3.2. Minimal Representation: SL(2C) 43
3.2. MINIMAL REPRESENTATION: SL(2C)
It is well known that the smallest faithful representation of the rotation group is
SU(2), the group of 2 x 2 unitary matrices of unit determinant. For the Lorentz
group, the minimal representation is SL(2C), the linear group of 2 x 2 complex ma
trices of unit determinant. To see this, let us organize the coordinates into a 2 x 2
complex matrix:
xerr@n=(CY, ae ) G.16)
where o* are the Pauli matrices, with the following properties:
fo, 08} = by
o'o? = io? (and cyclic permutations)
[o', 02] =2i03 (and cyclic permutations) @.17)
‘We see that
det ¥=x? (3.18)
A Lorentz transformation that takes X into X’ can be represented by the operation
X" = L(A)XL"(A) (3.19)
where L(A) is a2 x 2 complex matrix and L"(A) its Hermitian conjugate. Taking the
determinant of both sides, we have
det X" = det Xdet L(A)? (3.20)
To preserve x2, we must have det X” = det X, and hence
det L(A) = #1 21)
Consequently det A = +1. This is a more formal proof of a result stated earlier. The
matrices L(A) with det L =  constitute the group SL(2C).
Any 2 * 2 can be represented in the form
(3.22)
4+@0( A+B; a)
By +iBy ABy
where A and B, are complex numbers. The determinant of the preceding equation is44 Relativistic Fields
A? ~3,B}. Hence L(A) is a matrix of this form, with 4? 2,B} = 1. We leave it as
exercises to show that a pure boost and pure rotation are represented by the follow
ing: :
a: = ertob? = cosh? — (fo) sinh?
Boost along: L(A) = e##? = cosh — (ra) sinb
Rotation about: L(A) = elit 282 = cos +i(fro) sin 3.23)
where ft is a unit vector, @ is the boost angle defined in (3.3), and @ is the rotation
angle.
3.3 THE POINCARE GROUP
The laws of physics should be covariant with respect to spacetime translations as
well as Lorentz transformations. These transformations combined constitute the in
homogeneous Lorentz group, or the Poincaré group. The transformation law is as
follows:
x's al + ABx” (3.24)
where a is a 4vector. The infinitesimal version has the form
XH xh + Ql + aot, x” (3.25)
which contains 10 independent parameters: a” and w#” = w",
We can realize the Poincaré group on the space of functions /{x), through the
transformation
Se!) =fe+ a+ wx)
=I) + ay d,f G0) + OF, x”, I)
= [P+ ad 3 py XH” — x” NFR) (3.26)
where we have use the fact that w,,,, is antisymmetric. We can rewrite
Se"
() —iaMPH + yHaoht) ie) (3.27)
which defines the generators3.3 ThePoincaréGroup 45
Me? = x#P¥ — x¥Pe (3.28)
Of these, 10 are independent operators, constituting the Lie algebra of the Poincaré
group. An arbitrary element of the Poincaré group can be written in the form
exp (ia, P# —iw,,,MH") (3.29)
where a and wo” represent 10 real independent parameters.
From (3.28) we obtain the commutator
be PY
igh (3.30)
Although derived from an explicit representation, we consider the preceding equa
tions as abstract algebraic relations. Such a procedure is analogous to obtaining the
Lie algebra [J/, J*] = ie/*J! for angular momentum from the special representation
ir x V.. As abstract relations, the Lie algebra admits halfinteger representations.
The Lie algebra of the Poincaré group consists of the following commutators:
[Me, M28] = —~i(gua 8 — graiyub + gyB\ua — gubyra)
[M, P"]= i(gt*P' g”P#)
[P4, P*]=0 31)
which can be obtained through a straightforward calculation. In physical terms, the
four generators
PH = (H, P', P?, P?) (3.32)
make up the total 4momentum operator, and P° = H is the Hamiltonian. The six in
dependent components of M“” are generalized angular momentum operators made
up of the angular momentum J and the Lorentz boost K:
Mi = eiktyt
M%=Ki (3.33)
We can recast the Poincaré algebra as follows. The last two lines in (3.31) are equiv
alent to
[P’, P= [P), H] = [, H]=0
[), PX] =ie/t!P!
[k), H] =iP!
[B, P)] = 18, (3.34)46 Relativistic Fields
These relations all involve the inhomogeneous part of the group. The first equation
above expresses the independence of the spatial translations among themselves, of
spatial and time translations, and of rotations and time translations. The second
equation is what one can deduce from J = ir x V and P = iV. The other equa
tions above describe how energy and momentum change under a Lorentz. boost. In
addition to these, we obtain from (3.31) a closed set of commutation relations
among angular momentum and boost operators:
VJ = ie!
[K!, K4]=iet¥y!
WKN =iel'K! (3.35)
These form the Lie algebra of the Lorentz group.
3.4 SCALAR, VECTOR, AND SPINOR FIELDS
In quantum mechanics, the wave functions in a central potential can be classified
according to orbital angular momenta, which correspond to irreducible representa
tions of the rotation group, with possible dimensions 2/ + 1, (/ = 0,1,2, ...). Ina
similar way, relativistic fields transform according to irreducible representations of
the Lorentz group, which have definite dimensions. Accordingly, a relativistic field
has a definite number of components, related to the spin angular momentum of the
field.
The simplest relativistic field is a scalar field, which may have more than one
component, but each component x) must be invariant under Lorentz transforma
tions:
'(x') = @) (3.36)
This says that the transformed field called ¢’, at the transformed coordinate x’, is
the same as the original field called ¢, at the old coordinate x. It expresses the fact
that x’ and x are different labels that we use for the same physical point, and the
scalar field is unaffected by this; but for us the functional form of the field must
change:
'@)= GA'x) (3.37)
‘As we shall see, the spin of a scalar field is zero.
A vector field, such as the electromagnetic field 4“(x), is affected by a change
in the coordinate system, since by definition its four components transform among
themselves like x", The transformation law is3.4 Scalar, Vector, and Spinor Fields 47
A’#(x!) = AH AMX) (w= 0,1,2,3) (3.38)
The spin of a vector field is 1. This will be demonstrated in Section 5.5.
In general, a tensor field of rank n transforms like a product of n x“ terms, and
corresponds to spin n. For example, the gravitational field is a symmetric tensor of
rank 2.
There are “halfinteger” representations, analogous to those for the rotation
group. The latter are representations of SU(2), which generalizes to SL(2C) in the
present case. To accommodate spacetime reflections, we have to include two
copies of SL(2C), so that they transform into each other under a reflection. Accord
ingly, the minimal representation space is spanned by a fourcomponent complex
field, called the Dirac spinor field y(x), which transforms according to
Wi’) = SCA) (r= 1,2,3,4) (3.39)
where S(A) is a 4 x 4 complex matrix, discussed in more detail in Chapter 6. The
spin of a spinor field is +.
In general, a field forming a Kdimensional irreducible representation of the
Lorentz group has K components:
bx) (a= 1,2,...,K) (3.40)
which transform under a Lorentz transformation A according to
Pal") = San Mb) 3.41)
For an infinitesimal transformation A = 1 + @, we can put S(A) in the form
Sas = Sap + $k (3.42)
this defines the coefficients 3/4”, which, as we will show, constitute the spin matrix.
Under an infinitesimal Lorentz transformation, then, a general field transforms
according to
hale’) = dalx) + $y Xhs P40) (3.43)
The change in the functional form of the field can be found by writing
bale!) = bale + ex) = by(X) + Oy X”Iuha(X)
= bi x) — Oy Axd”  x9, )b4 (x) (3.44)
Thus48 Relativistic Fields
PAC) = ha!) + F Oy fXtd” — XH) a(x) (3.45)
Substituting {(x’) from (3.42), we obtain
bal) = Pax) + 3 Oy sL HO” — x” Eap + Zio bs(x) (3.46)
This identifies the K x K matrices ¥#” = 3" as spin matrices, since they are added
to the generalized orbital angular momentum.
The spin matrix for a scalar field is obviously zero. For the vector field, we can
find it from its transformation law under an infinitesimal Lorentz transformation
AL(a!) = Aah) + ap AMG) G47)
Putting wag = 2,28, We obtain for the vector field
Xap Sap 8hBa (3.48)
As we shall show in Section 5.5, this gives spin 1. The case of the spinor field will
be discussed in Chapter 6, and is included in the following summary for reference:
Scalar field: 2"”=0
Vector field: X45=ghes—geen
Spinor field: a= sor YY es (3.49)
where are the 4 x 4 Dirac matrices defined in Chapter 6.
3.5 RELATIVISTIC QUANTUM FIELDS
Since quantum fields are operators that act on a Hilbert space, we can represent
Lorentz transformations by transformations on the Hilbert space. Recall that a
Lorentz transformation changes the functional form of a classical field:
Pale} bale) (3.50)
In the quantized version, this means that the operator ¢, attached to point x is re
placed by ¢/. Since @, and ¢/ act on the same Hilbert space, the transformation is a
mapping of the Hilbert space into itself. Since ¢, and ¢/ are physically equivalent,
the transformation must be unitary. Thus, there should exist a unitary operator U(A)
on the Hilbert space, corresponding to the Lorentz transformation A, such that
Pale) = UA) b(X)U A) @.51)3.5 Relativistic Quantum Fields 49
The fact that the transformation is unitary means
UN(A) = UA) (3.52)
From the definition of the primed fields 5(x) = Sasds(x), we obtain the condition
Udslx)U*' = Sans A'x) (3.53)
The set of operators U(A) forms an infinitedimensional unitary representation of
the Lorentz group. In contrast to this, the finitedimensional representations of the
Lorentz group are nonunitary. As examples, we have
Scalar field: Ud(x)U“! = (Ax)
Vector field: UAM(x)U! = Ag A*(A“!x)
Spinor field: Uy,(x)U = S,,,(A“'x) (3.54)
We can immediately extend this consideration to Poincaré transformations:
Uebs ()U"! = Saab A“e a) (3.55)
For infinitesimal Poincaré transformations, U must be in the neighborhood of the
identity operator, and linear in the parameters of the Poincaré group:
U=1iaupe+ ad (3.56)
This defines the Hermitian operators P and M«”, which represent the generators of
the Poincaré group on the Hilbert space. In contrast, the generators denoted by the
same symbols in (3.28) are finite matrices, generally nonHermitian.
Substituting (3.56) into (3.55), we obtain
( iaP + you bse(1 +iaP  zoM)
= (5 + 73s) Oana) (3.57)
which is written in an obvious abbreviated notation. Expanding both sides to first
order in a“ and w”, and equating their coefficients, we obtain
HTPH, ba(x)] = bax)
HME”, ba(x)] = (x — xO") hale) + Eas ho(x) (3.58)
This shows that P* is the 4momentum operator, since it generates spacetime50 Relativistic Fields
translations, and Mé” is a generalized angular momentum operator, since it gener
ates spacetime rotations. The spin matrix Sy, induces a mixing of the field compo
nents undergoing spacetime rotation.
The generators P# and M+” can be constructed explicitly from the field opera
tors @,. Rather than doing this on a casebycase basis, we shall do it via a unified
approach in the next chapter.
3.6 ONEPARTICLE STATES
A oneparticle state is an eigenstate of P#, with energy eigenvalue E > 0, and mo
mentum eigenvalue p, such that the invariant mass squared
Ps Ep? =m? (3.59)
is a fixed number. Such a state corresponds to a particle of mass m. The P? of any
state generally lies in a continuum, but those of oneparticle states form a discrete
set. If there are no massless particles, the invariantmass spectrum of a field theory
consists of the vacuum value 0 as a lower bound, a discrete set of particle masses,
and a continuum separated from the vacuum value by a finite gap. The continuum
corresponds to states containing two or more particles, whose masses occur within
the gap. There can be particles whose mass occurs in the continuum, but only if
these particles are stable against decay, due to selection rules. The gap vanishes
when there are massless particles, such as the photon. In this case, there is a discrete
mass in the continuum corresponding to the electron, which cannot decay into pho
tons because of conservation laws.
The oneparticle states of a free field can be generated by applying creation op
erators a, to the vacuum state. If we do this for a nonfree field, we will not get one
particle states, because we will not get eigenstates of P+. Instead, we will have a
mixture of states involving interacting particles. Nevertheless, we can discuss prop
erties of a oneparticle state through general considerations, without constructing it
explicitly.
We confine our attention to massive particles, with m > 0. There exists a
Lorentz frame in which p = 0, called the rest frame. The spin operator S of the one
particle state is defined as the total angular momentum in the rest frame. The eigen
value of S? has the form S(S+ 1), where Sis called the spin of the particle. The pro
jection s of S along the momentum of the particle is called the helicity, which for
m> Ohas 2S +  possible values S, S—1,...,—S. We can label a oneparticle state
by momentum p and helicity s:
particle state) = p, s) (3.60)
The parameters m and S are suppressed, because they are constants.
For m > 0a oneparticle state in the rest frame is denoted by s), and we can ob
tain p, s) from s) through a Lorentz boost L(p):3.6 OneParticle States 51
Ip, s) = UL(p))\s) (3.61)
Applying a Lorentz transformation A to both sides, we obtain
U(A)Ip, s) = UA)UL(p))\s) (3.62)
Now insert in front of the right side the identity operator in the form
1= UL(p)U'L(p)) 3.63)
and regroup the factors in the following manner:
U(A)p, s) = UL@U"(LP) U(AUL(p))IIs) G.64)
By the group property, the operator within the square brackets can be rewritten as.
UL@)UAUL(p)) = UL(A p)AL(p)) (3.65)
which represents a pure rotation called the Wigner rotation:
R(A, p) = L!(Ap)AL(p) (3.66)
It boosts a particle from rest to momentum p, makes a Lorentz transformation A,
and then brings the particle back to rest. The operation has no effect on the state
vector except possibly multiplying it by a phase factor, which represents a rotation.
Thus, the general effect of a Lorentz transformation on a oneparticle state is to
boost the momentum, and rotate the spin by a Wigner rotation:
U(A)Ip, s) = UL(Ap))U(R)s) (3.67)
For a more explicit representation of R, we insert a complete set of helicity states to
obtain
UAylp, s) = 5, ULCAR))s") s"U)s)
= > Du(R)iAp,s’) (3.68)
where DJ..(R) are the rotation coefficients. An example of the Wigner rotation is the
Thomas precession discussed in Section 6.8.
Massless particles are special, in that there is no rest frame. A massless particle
of spin S can have the values +S only. We shall explicitly demonstrate this for pho
tons in Section 5.5. A general proof may be found in books on representations of the
Lorentz group (see, e.g., Tung [1]).52 Relativistic Fields
PROBLEMS
3.1 Verify the Poincaré algebra (3.31).
3.2. Verify the spin matrix (3.48) for the vector field.
3.3. Verify that the SZ(2C) matrices L(A) given in (3.23) correctly represent Lorentz trans
formations. It is necessary to verify them only for infinitesimal transformations.
3.4. Show the following identity, which is usefull when working with SL(2C) matrices:
(0: Ao: B)=ABtio AB
where the components of oare Pauli matrices and the components of A and B are num
bers.
3.5 Consider two infinitesimal successive infinitesimal Lorentz boosts with angles g, and
>. Show that the result is equivalent to a boost #; + > plus a rotation + ¢, x gs. Here,
@=¥ tanhr'», where v is the velocity of the boost. Lorentz boosts.
3.6 (a) Under the action of a Lorentz boost with velocity v, a 4momentum p# is trans
formed to p', show
a
PO = XPo VP)
vepty( >~ We)
where y= (17),
(b) Writing p"# = A+,p’, obtain the transformation matrix
ec %
Ms
” = Vor
yt aj& =f UR
(©) Let L(p) be the transformation matrix corresponding to a Lorentz boost that trans
forms the rest frame of a particle of mass m into a frame in which the particle has
momentum p and energy £. Show
Em
woy
Pe et a Pe
m mE + m)
3.7 (a) Consider a particle of mass m and helicity s, moving with momentum p along the z
axis. Make a Lorentz boost of velocity v along the x axis, so that p> p'. Find the
rotation matrix Ri for the Wigner rotation.
(b) Show that for an ultrarelativistic particle
iP opie
RehReference 53
That is, the Wigner rotation is the same as that taking the initial velocity p/E to the
final velocity p'/E'. This shows that the helicity of a massless particle such as the
photon is Lorentzinvariant.
REFERENCE
1. W.K. Tung, Group Theory in Physics, World Scientific, Singapore, 1985, Section 10.4.4.Canonical Formalism
4.1 PRINCIPLE OF STATIONARY ACTION
The equations of motion for a classical field can be derived from a Lagrangian
through the principle of stationary action. This approach gives a unified treatment
of topics discussed previously through special examples, It also makes clear that
symmetries of the system give rise to conservation laws. Consider a set of classical
fields collectively denoted by (x):
Hx) = {h1(2), , bx} 4.1)
We denote their spacetime derivatives by
lx) = 9,900) (4.2)
‘The Lagrangian density is assumed to depend on the fields and their first deriva
tives:
L(x) = L(4(x), bul) (43)
This will ensure that the equations of motion are secondorder differential equations
in the time, as in classical mechanics. We assume that, unless external fields are ex
plicitly introduced, spacetime is homogeneous. Thus, £(x) depends on x not explic
itly, but only implicitly through q(x) and $,(x). We consider only local field theo
ries, for which the Lagrangian density at x depends only on properties of the field at
x. Nonlocal terms of the form
Je'rb.00K ale  4.0) a4)
are ruled out, unless K,,(x — y) * &(x —y).
544.1 Principle of Stationary Action 55,
The classical action of the system is
=  dtxei 4.
SI] [i L(x) 4.5)
where 0 is the spacetime volume, which eventually goes to infinity. We impose
definite boundary conditions on the surface of 1, say, # = 0. The principle of sta
tionary action, which is a generalization of that in classical mechanics, states the
following:
Suppose that (x) is a solution to the equations of motion. If we vary its functional
form by adding an arbitrary infinitesimal function 6¢(x) that preserves the bound
ary condition:
H&) > Hx) + Sx) 6b(x) = 0 on boundary of O
then the variation of the action will be of secondorder smallness:
8S = Sb + 86]  ST] = 0
This means that S[@] is at an extremum when q(x) is a solution to the equation of
motion,
To find the equation of motion according to this principle, let us calculate the
variation of the Lagrangian density:
or spoe a0 (4.6)
Using the fact that
5, = (4,8) = 9,59) 472
we get
a
a&L=— 4, 5h + 4, —— 5. 4.8)
[aaa] Pera(ae) a8
The last term is a total 4divergence. It vanishes when integrated over the
spacetime volume, since it then becomes a surface integral by Gauss’ theorem, and
5d = 0 on the surface. Thus
eleSAg er56 Canonical Formalism
Since 5¢(x) is arbitrary, its coefficient must vanish. We thus obtain the equation of
motion
(4.10)
where
a
7 = — 4.11
a, G11)
The canonical conjugate to (x) is defined by analogy with classical mechan
ics:
aw
max) = W(x) = a (4.12)
where a dot denotes partial time derivative. The Hamiltonian density is defined by
H(x) = H( mx), Hx), V d(x) = 7b L(x) (4.13)
where d(x) should be reexpressed in terms of m(x) according to (4.12). The Hamil
tonian is given by
H= faal00) (4.14)
To quantize the system, we replace the field and its canonical conjugate by op
erators, which are defined by the equaltime commutators
i[m(x, 1), HY, Dl. = RX y)
Lax, ), my, De = (4, 0, Hy, D].=0 (4.15)
where for bosons we use the commutator
[4, BL = [4, B] = ABBA (4.16)
and for fermions we use the anticommutator
[A, B], = (4, B} = AB+BA (4.17)
If a(x, t) = 0, as is the case for the electromagnetic field, then the field conjugate to
a(x, 2) is not an independent dynamical variable, and should not be independently
quantized.4.2 Noether’s Theorem 57
4.2 NOETHER’S THEOREM
A transformation (x) > (x) + 5d(x) is called a symmetry transformation of the
system if it changes the Lagrangian density only by the addition of a 4divergence.
As we have seen, this does not change the equations of motion. More specifically,
the change must be of the form 8£(x) = 4,,//(x) for arbitrary (x), regardless of
whether it obeys the equation of motion. If the symmetry transformation is continu
ous, then there is an associated conserved current density. The formal statement is
as follows.
i NOETHER’S THEOREM [1] /f, under a continuous infinitesimal transfor
mation
Hx) > Hx) + SGX)
the change in the Lagrangian density is found to be of the form
BL(x) = YW,0)
without using the equations of motion, then there exists a current density
JER) = (0) h(x) — Wer) (4.18)
which, for fields obeying the equations of motion, satisfies
In JMx) = 0 (4.19)
Proof. We calculate the 8£(x) when the field changes by 54, using the equa
tion of motion, but without assuming that 8 comes from a symmetry transforma
tion:
“ow a
a= wo? 06, Oe woh d,(56)
= a,(m*5$) + (35 . a.n)50
= 4,75)
where the equation of motion was used in the last step. Specializing the preceding
to symmetry transformations, we equate it with d,,/7* to obtain
 3,18 — W) =0 .58 Canonical Formalism
The conserved current j# is called a Noether current, and is determined only up to
an arbitrary normalization factor.
Noether’s theorem was proved for classical fields, and one usually extends it to
quantum theory by replacing the fields in j# by the corresponding quantized fields.
This does not always give a conserved current in the quantum theory. When the
quantum current so obtained fails to be conserved, the nonzero divergence 4, is
called a “current anomaly.” Some examples of this are discussed in Section 19.8.
4.3 TRANSLATIONAL INVARIANCE
‘An important symmetry for any system is Poincaré invariance, which is called a
“spacetime symmetry,” because it is associated with the transformation of the field
under a change in the coordinate system. We discuss this symmetry by breaking up
the Poincaré group into the translation and Lorentz subgroups.
Invariance under the translation group should give rise to four independent
Noether currents corresponding to the four possible spacetime translations. Con
sider an infinitesimal translation
who xt at (4.20)
under which each component of (x) independently transforms according to
$'(x + a) = G(x). The functional form of the field changes according to
(x) = Px — a) = Hx) — a, 0" b(x) (4.21)
We shall choose a,, to have only one nonvanishing component, say, a, and take a=
0,1,2,3 in turn. Thus, the change in functional form of the field is
SG(x) = HHx) (4.22)
where we have dropped the proportionality constant —a,,, because it enters all subse
quent formulas only through the overall normalization of the Noether current. Dif
ferentiating the above with respect to x“ gives
5,2) = Fh, (x) (4.23)
The statement of translational invariance is that £(x) does not depend on x ex
plicitly. Hence
L(x) = I b+ = #6, = PL(x)
a
oh %,
= d,(g**L(x)) (4.24)43 Translational Invariance 59
from which we read off
Weer) = gh£(x) (4.25)
The four corresponding Noether currents are denoted by 7.“, where a labels the
direction of translation:
T B(x) = 1d P(x) — ghAL(x) (4.26)
They satisfy the conservation law
4,T (0) = 0 (4.27)
This is called the canonical energymomentum tensor. It is generally not a symmet
ric tensor. The subscript “c,” which stands for “canonical,” distinguishes it from a
symmetrized version to be discussed later.
The conservation law can be rewritten in the form
a
Gale =0 (4.28)
2 rms
ot
Integrating both sides over all space, and assuming that surface contributions van
ish, we obtain
d
arn (4.29)
where
Pox faxree (4.30)
is the total 4momentum. Thus, 7% is the energy density and T* the kth compo
nent of the momentum density. Their conservation laws are given respectively by
the time and spatial components of (4.28):
a
ot
Zry+ TH=0 (431)
It is clear that 7 is the kth component of the energy current and 7 is the kth com
ponent of the current of the “jth component of momentum.” The latter is called the
stressenergy tensor. These components are displayed in the following matrix:60 Canonical Formalism
Te Tol re 78s
THa= 73
(4.32)
The identification of T as the energy density is consistent with the definition of
the Hamiltonian density in (4.13).
The explicit expressions for the total energy and momentum are
P= fa®xfme)'4) — £09] = [d*eale0)
P= faremoatete) (4.33)
where a dot denotes time derivative and (x) is the Hamiltonian density. We go over
to the quantum theory by replacing (x) and (x) by the appropriate operators.
To show that P* generates spatial translations, we calculate the commutator of
(x) with the total momentum operator
Pea [aby my) AO(y) (434)
This does not depend on yo. We are therefore free to choose yp = xo =f, to take ad
vantage of the simplicity of the equaltime commutators. Thus, we have
1% dex a1 form. 9°, oe]
fay 80  yee «e (435)
To show that P° generates time translations, we calculate
[P®, bx, 1] = J Ay {Lmty, dy. Ds HX, D]  [L(Y 1), Hx, D)}
= bls, + faryt AY, HY, D, Hx, (L(y. ), OO, D]} (4.36)
The integral in the second term identically vanishes.
Proof. Use the representation = i8/51 to write
SOY) _ ans, dy)
Bate) aagay
[0(y), 6@)] =
Using = 46/46, we can calculate the integral as follows:
L(x) Ad(x) _ x
[ses amx) ma 
AL(y) SHY) AL(y)] _
fa] By dmx) ae44 Lorentz Invariance 61
Combining the preceding results, we have
i[P?, $x] = P(x) (4.37)
This shows that the total 4momentum generates spacetime translations. If F is a
function of ¢ and its derivatives, then
i[P*, F] = ®F (4.38)
Choosing, in particular, F = P#, we have i[P*, P#] = 9°P#. The right side vanishes
because P# is independent of space by construction, and independent of time by
4momentum conservation. Thus
[P*, P4]=0 (4.39)
which is part of the Poincaré algebra discussed in Chapter 3. It is realized here in
terms of the field operators.
4.4 LORENTZ INVARIANCE
Consider an infinitesimal Lorentz transformation in the direction labeled by {a,8}.
For example, a rotation about the x? axis corresponds to {1,2}, or a boost along the
x! axis corresponds to {0,1}. Under the transformation, the functional form of the
field changes by 5¢(x), as given by (3.46). We have, up to a multiplicative constant,
B(x) = (x4 — xa + BP) hx) (4.40)
where 52 is the spin matrix, antisymmetric in {a, 8}. Differentiating 5 with re
spect to x", we find
5d,(2t) = [x%9 — xP + BP] h(x) + (geGF  ghb%) (4.41)
Using the fact that £(x) has no explicit x dependence, we have
a“ a
8L= 9g
apo 3g, 7%
= (9 — Pa) + apne + LB, + gh mg" (4.42)
The statement of Lorentz invariance is that the preceding is a 4divergence. The first
term is of the desired form, for it can be rewritten (x) — %(x8L) = d,(ghPx2L —
g#x6C), Therefore the rest must vanish:
SpE + ALO, + 1G? — ag =0 (4.43)62 — Canonical Formalism
With this, we have
AL = d,(grPxel — grexPL) (4.44)
which gives
We(x) = (ghPxa — gh )L(x) (4.45)
This leads to six independent Noether currents labeled by {@, 8}:
MESB = ar c2® — xP + 3) bfx) — (ghPx — ger )L(x) (4.46)
which can be written in the form
MBF (x) = x°T#A(x) — xP TH(x) + THLE Hx) (4.47)
where T#” is the canonical energymomentum tensor. It satisfies the conservation
law
8, M#°(x) = 0 (4.48)
and is called the canonical angular momentum tensor.
The conservation law gives rise to six constants of the motion, the Lorentz
boost
W= [ae Mm) = [asf He)  TEI] (49)
and the angular momentum
STe= sell dx MON(x) = $e/#[dPxfxt 19x) —x/T MQ) + mBIGG)] (4.50)
where the first two terms represent orbital angular momentum and the last term is
the intrinsic spin.
It is straightforward to verify that K’ and J* generate Lorentz transformations by
showing
i[Jary moe, 0, des. 0]= 549 as
where 5¢ is as given by (4.40). It can also be verified that the commutation rela
tions among the operators P# and M22 realize the Poincaré algebra (3.31).4.5 Symmetrized EnergyMomentum Tensor 63
4.5 SYMMETRIZED ENERGYMOMENTUM TENSOR
The canonical energymomentum tensor 7 #“ is not unique, because the Lagrangian
density is defined only up to a 4divergence. We can replace it by any tensor of the
form
T= TH+ LOX ue (4.52)
where X;yq is antisymmetric in A:
X ua = XH (4.53)
The antisymmetry is a sufficient condition that the conservation law be unchanged:
9,T#* = 6,T#* + 4$3,9,Xsyar= 0 (4.54)
A possible change in the total 4momentum is
fareroe— fareroe= 5 [atea, x0" (4.55)
which vanishes for the following reasons: (1) the term with A = 0 vanishes because
Xa = () by antisymmetry and (2) the terms from A = k vanish because they give a
surface integral. From a physical point of view, therefore, 7 is equivalent to T¢*,
because they give the same total 4momentum,
The fact that 7#“ is not guaranteed to be a symmetric tensor poses a problem, if
it is to be used as a source of the gravitational field. We can, however, replace it with
a equivalent symmetric tensor 7#*. The condition for symmetry is
Twa Taw = [Tea Ten] + $9,(XMa— Yow) = 0) (4,56)
The term in brackets can be rewritten using the condition (4.43) for Lorentz invari
ance:
Tes Tee = mip — nig =p =H IA
=A m'3Heg) — (35  ames (437)
where the last term vanishes by the equation of motion. Substituting this result into
(4.56), we obtain the condition
XAwa — ran = 2DPSueds (4.58)64 Canonical Formalism
a solution to which is
Xue = (AS He — AEM — Se)” (4.59)
This term is needed only when the spin is nonzero.
Corresponding to the symmetrized energymomentum tensor, we can define a
new angular momentum tensor:
MueB == xa THB — x6 THe (4.60)
This is related to the canonical angular momentum tensor through
MuoB = MusB + 29, (xAXMM8 — xPX AHA) (4.61)
It is easy to show that M+? is conserved, and that it preserves the definition of the
boost and the angular momentum:
4,MHe8 = 0
f. PxMoo8 = f. d2xM 008 (4.62)
From a physical point of view, therefore, M*“8 and M4? are equivalent.
4.6 GAUGE INVARIANCE
In contrast to spacetime symmetries, there are internal symmetries, which are xin
dependent transformations of the field that leaves the Lagrangian density invariant.
‘A simple example is a change of phase in a complex scalar field:
x) > EF plx)
We) > eye) (4.63)
This is called a global gauge transformation, where the label “global” refers to the
fact that @ is independent of x. Invariance with respect to it means that the La
grangian density is independent of the phase. This is true for the free field, in which
the fields appear in the combination y"(x)ix) or y'(x)4,1Mx). The infinitesimal
form of the transformation is
Sx) = Wx)
Sy") = We) (4.64)
where we left out a proportionality constant ia.
In terms of the real and imaginary parts defined byProblems 65
He) = Fy ldale) + ibst0)]
Va) = Fylh e609] 465)
the transformation is a rotation in internal space:
/(&) = d(x) cosa + py(x) sina
2(x) = —d;(x) sina + (x) cosa (4.66)
The Noether current is just the conserved current mentioned in Chapter 2:
a
¥
© 45,0)
= yore — yr oy
2( doh, — bho) (4.67)
More generally, internal symmetries are linear transformations for a multicompo
nent field $,(x), (a= 1, ...,K), of the form
a(x) = Carbs) (4.68)
where C,, are elements of a K x K constant matrix. If the matrix belongs to a Kdi
mensional representation of some group G, we call G an internal symmetry group.
The group can be continuous or discrete. In the previous example G = U(1), the uni
tary group of dimension 1.
Physical examples of conserved charges are
© Electric charge = positive minus negative charge
© Baryon number = number of protons minus number of antiprotons
Electron number = number of electrons minus number of positrons
A important case is isospin, which is discussed in Section 7.5.
PROBLEMS
4.1 Consider a field (x) with Lagrangian density £o(x) +.£,(x), where the first term has a
certain symmetry, while the second term does not. That is, under a transformation 44x)
> Hx) + 640), 
Lox) = HW, (x),66 Canonical Formalism
whereas £,(x) cannot be put into this form. If £,(x) were absent, the system would have
a conserved Noether current, Show that, in the presence of L,(x), the divergence of the
wouldbe Noether current is 82; (x).
4.2 A condition for Lorentz invariance is (4.43). For scalar fields, for which 2° = 0, what
restriction does this place on the Lagrangian density?
43. Nonrelativistic System A nonrelativistic manyparticle system has a secondquan
tized Hamiltonian
H= fox V00(a5 vit )400
where jis the chemical potential. Usually one assumes the commutation relation [Y(x),
W(y)] = (x —y). We want to see whether this is consistent with the canonical formal
ism
(a) Find the equation of motion using i= [W, H].
(b) Regard H as a classical Hamiltonian, Show that the corresponding Lagrangian den
sity is
ca inragiars (3, 2H) o
Work out the equation of motion using the canonical formalism.
(©) Show that the usual commutation relation [¥(x), ¥#(y)] = &(x— y) is canonical.
(@) Work out the Noether current associated with spacetime translational invariance,
and global gauge invariance.
4.4. Field Representation Since y(r) and iy*(r) are canonical conjugates in the nonrela
tivistic system, it would be awkward to introduce the field representation by diagonaliz~
ing Yd). Show that in this case we can put
1 6
e)= " a 
6
ban)
veo al 
where g(r) is a cnumber function,
4.5. FirstOrder Lagrangian The nonrelativistic Lagrangian in the last problem differs
from a relativistic one, in that it is firstorder instead of secondorder in the time deriva
tive. The Dirac field discussed in Chapter 7 also has a firstorder Lagrangian. To fully
explore the consistency of the canonical formalism, let us strip the problem down to
bare essentials, and consider a classical system with two coordinates a and 6, which are
like iy* and y. Take the Lagrangian to be
L(a, b, b)=abVa, b)
The canonical rule says that a has no canonical conjugate, It is the conjugate to b. Is this,
completely consistent with the Lagrangian and Hamiltonian equations of motion?
(a) Find the Lagrangian equations for motion fora and.Reference 67
(b) Find canonical momenta and the Hamilton equations of motion, Check that they
are the same as the Lagrangian ones.
(©) Itis thus completely consistent to regard a and b as canonically conjugate. To quan
tize the system, impose (4, 6], =—i.
REFERENCE
1. E. Noether, Nachr. kgl. Wiss. Géttingen 235 (1918); E. L. Hill, Rev. Mod. Phys, 23, 253
951).Electromagnetic Field
5.1 MAXWELL EQUATIONS
The classical electromagnetic field is described by two 3vector fields, the electric
field E(x, t) and the magnetic field B(x, t), which obey Maxwell’s equations. In ra
tionalized units with c = 1, they read
VE=p
VB=0
(6.1)
where p(r, 1) and j(r, £) are respectively the external charge density and external cur
rent density, which must satisfy the continuity equation
4
+Z (5.2)
The second and third equations are solved by introducing the vector potential A(r, #)
and scalar potential (r, 1):
B=VxA
E=Vg— (5.3)
whereupon the remaining two equations become
685.1 Maxwell’s Equations 69.
$ 2 j At
( v)A iv(va+ Ze)
6.4)
The potentials are determined only up to a local gauge transformation, which in
volves an arbitrary function x(r, ‘)
ASAt+VX
Ox
b> 6 63)
The Lorentz gauge corresponds to the condition
VAt # =0 — (Lorentz gauge) (5.6)
In this gauge, both potentials satisfy the wave equation:
é 2
(a> 73
& 2 =
(W)ee 6.7)
The symmetric appearance of these equations is sometimes convenient, but it actu
ally obscures the physics. These equations seem to indicate that there are four inde
pendent propagating modes, but actually there only two—the transverse compo
nents of A. This can be shown by going to the Coulomb gauge.
In Coulomb gauge (or radiation gauge), A is purely transverse:
VA=0 — (Coulomb gauge) (5.8)
The equations for the potentials become
(re
V?b=p (3.9)
where j, is the transverse current density70 Electromagnetic Field
inj2V¢ 5.10)
in=iG 6.
which satisfies Vj; = 0. In this gauge, A describes transverse electromagnetic’radi
ation, whose source is the transverse current density, while describes the instanta
neous Coulomb interaction between charges. The potential between two unit
charges located at r, and r, is given by
_ i
Aqir,— 1)
or) (3.11)
To show that we can always impose the Coulomb gauge, suppose VA =/. To go to
Coulomb gauge, we make the gauge transformation A > A+ V x, with y satisfying
Vx =f. The solution corresponds to the statement that x is the electrostatic po
tential due to the charge distribution VA.
We are using rationalized instead of unrationalized units. The difference be
tween these systems arises from the normalization convention for the free fields,
and is tabulated as follows:
Rationalized __Unrationalized
4Current je 4aje
Coulomb’s law gir Paar
Energy density (E2 + B’y72 (E2+ B87
Field operator A 4nA.
5.2. COVARIANCE OF THE CLASSICAL THEORY
‘We postulate that the potentials form a 4vector
A= ($A) (5.12)
and this determines how Maxwell's equations transform under a Lorentz transfor
mation. Since we always impose a gauge condition, a Lorentz transformation must
be accompanied by a gauge transformation
Ab —> AB Ay (5.13)
in order to maintain the gauge condition. Under an infinitesimal Lorentz transfor
mation, therefore, the vector potential transforms according to
A'#= AH wp” — dy (6.14)a
5.2 Covariance of the Classical Theory 71
where y is such that 4'# satisfies the gauge condition.
The Lorentz gauge 0,,4# = 0 is covariant, and the equations of motion take the
form (5.7), which are manifestly covariant
ceae (6.15)
where jis the conserved 4vector current density
(D4, jn=0 (5.16)
In this gauge, however, the physical degrees of freedom are not manifest.
In Coulomb gauge, where physical degree freedom are made explicit, the equa
tions of motion (5.9) are not manifestly covariant; but they actually are, because
there always exists a gauge transformation to maintain the appearance of the equa
tions in all Lorentz frames. One has to choose between manifest covariance with
Lorentz gauge, or manifest transversality with Coulomb gauge, and we choose the
latter.
The electric and magnetic fields are components of the antisymmetric field
tensor
Fr gud” — AH (5.17)
which is gaugeinvariant. The dual field tensor is defined as
For = Len Fag (5.18)
In terms of the electric and magnetic fields, we have
PO=E Pia —eitpt
=~4el/P) (5.19)
The components of the field tensor and its dual can be displayed as matrices:
0 f P B
pow( BOO, OR (5.20)
BB Bl oO
0 Bl B 8
Fur= B 0B B (5.21)
 0 OE!
B EF E 02 Electromagnetic Field
We see that F*” is obtainable from F#” through the duality transformation
{E, B} > {B,E} (5.22)
From the field tensors we can form two independent Lorentz invariants:
4 F!F,,= 5 (BPE*) (scalar)
+ PMR, =BE (pseudoscalar) (5.23)
In terms of the field tensors, Maxwell's equations read
6,FHY =
(5.24)
which are gaugeinvariant and Lorentzcovariant, and are invariant under the duality
transformation when j” = 0. Since F#” = —F, the first equation is consistent only if
,j" = 0. The second equation is identically satisfied by putting F“”= d¥A— PAM.
5.3. CANONICAL FORMALISM
‘The Lagrangian density of the free electromagnetic field is
E? — B?) (5.25)
PYr,,
Apart from an overall factor, this is uniquely determined by the requirement that it
be Lorentz and gaugeinvariant, and does not contain higher derivatives of 4¥ than
first derivatives. The minus sign in front is chosen to give a positive energy density
for the free field, and the factor + sets the normalization of the fields. To obtain the
equations of motion from the action principle, we must use the potential 4” as the
field variable. The Lagrangian density then reads
= AMAA, (5.26)
from which we obtain
a
m= =Fe 5.27
0,4) oe
The equation of motion is5.3. Canonical Formalism 73
a
o,meY — oA =0 (5.28)
Since the last term on the left side is zero, we have
0,FH*= 0 (5.29)
The use of the potential makes d,, 74” = 0. Thus, we correctly recover Maxwell's
equations for free fields.
The canonical conjugate to 4” is
7
Fe (5.30)
which vanishes identically for v= 0, indicating that 4° is not a dynamical variable.
The dynamical fields are 4‘, with canonical conjugate ~F° = E*, However, the lon
gitudinal part of A has no physical significance, because it can be changed at will
through a gauge transformation (see Table 5.1.) The only dynamical degrees of free
dom are the two transverse components of A, and we can go to the Coulomb gauge
to make this explicit. In Coulomb gauge 4° satisfies the Poisson equation, and is de
termined by the external charges.
The canonical energymomentum tensor is, according to (4.26),
TES = He, — ghOL = FHA, — geal (531)
which can be rewritten using the equation of motion:
TE, = FMF ay BEL 0(F HAG) (5.32)
The last term is not symmetric in u and a, and not gaugeinvariant. However, it is a
total 4divergence antisymmetric in 1 and v, and is conserved because 4,,0,(FH¥A*)
= 0. As discussed in Section 4.5, such a term has no effect on the conservation law
and the definitions of total energy and total momentum, and may therefore be omit
ted. Thus we take as energymomentum tensor the symmetric and gaugeinvariant
tensor
Th =FYF ay — Bah (5.33)
TABLE 5.1 Fields and Canonical Conjugates
Field Canonical Conjugate Remark
a 0 Not dynamical variable
A  Only transverse part physical74 Electromagnetic Field
which satisfies
The trace of the tensor vanishes:
It is now straightforward to obtain the Hamiltonian density:
H= 1% =—F%Fy,—£ = + (BE? +B?)
The momentum density (the Poynting vector) is
Sk = [0 = FUP = eHiEB)
The total Hamiltonian H and total momentum P are given by
H=t Jarre +B?)
P= fa Sr Ex B
The conservation of energy and momentum correspond to the statements
2 H+TS=0
Sst+apt=0
where
Dk = (EVES + BIBS) + 5 8y(E? + B®)
is the stress tensor.
According to (4.47), the generalized angular momentum tensor is
Moab = xaTuB— x8 Tua
which satisfies the conservation law
6,Mu2b = 0
(5.34)
(6.35)
(5.36)
(537)
(5.38)
(6.39)
(5.40)
(5.41)
(5.42)
It follows that the total angular momentum J and the Lorentz boost K are given by5.4 Quantization in Coulomb Gauge 75
I= fatrrx (EB)
K=®rH (5.43)
5.4 QUANTIZATION IN COULOMB GAUGE
To quantize the electromagnetic field, we must first eliminate all unphysical de
grees of freedom by fixing the gauge, and in the following we shall use Coulomb
gauge.' In the absence of external charges, we can set 4° = 0, and write the Hamil
tonian in the form
t far (E+Vx AP) (VA=0) (5.44)
The canonical conjugate to A is E = 2 A/ét, and we would normally impose the
equaltime commutation relation [(r, #),A"(r’, #)] = 18,5°(r  r’). But this is incor
rect here, because the right side is not consistent with VA = 0, nor with one of
Maxwell’s equations VE = 0. We therefore replace 5,,5°(r — r’) by its transverse
projection, and take
[Pr 0, AM’, 0) = Bh (rr’) (5.45)
where the transverse delta function 8 1,(r— r’) is defined by
&k Kk
an0)=  Sap (5 je et (5.46)
and satisfies
aIBT(r) = T(r) = 0 (5.47)
‘A complete set of solutions to Maxwell’s equations in a periodic box of volume
are the transverse plane waves
(ker a = [kl
where €(k) is a unit polarization vector normal to k. For each k, there are two inde
pendent polarization vectors €,(k) and €,(k); and €,(k), €(k), k together form a
righthanded coordinate system:
ke,(k) = ke,(k) = 0
'For quantization in other gauges, see Huang [1].76 Electromagnetic Field
€,(k)€(k) = 5,
k
i(k) * €,(k) =
610) * el Ty (5.48)
Having chosen €;(k), €,(k), there is still arbitrariness in the choice of €,(k), €,(k).
By convention, we choose
(k= x(k)
€,(k) = €,(k) (5.49)
as illustrated in Fig. 5.1. The following sum over polarizations results in the trans
verse projection operator (see Problem 5.2):
Kei
md (5.50)
2
Ti(k) = Y(W)ef(k) = 5, —
A
We now expand the field in terms of the transverse plane waves:
2
el([a(kye™* + a,1(Ke*"]
1
4092 at
4,0)  Yettolatve*— aoe] 651)
The commutation relations (5.45) are satisfied by imposing the commutation rela
tions
E(k) Ex(k)
E,(k)
E,{k)
Figure 5.1 Polarization vectors of an electromagnetic wave.54 Quantization in Coulomb Gauge 77
{4,00}, @7()] = 845kp
[4,(),a,(p)] = 0 (5.52)
where a,(k) is the annihilation operator of a photon—a field quantum of momentum
k and linear polarization s, In the freefield case, the timedependent operator
A(t, t) is simply obtained by replacing (Kr by (kr — af) in the exponents in
(5.51), because
eltia(p)e = a,(p)e ">" (5.53)
The Hamiltonian and the total momentum operator of the electromagnetic field are
given respectively by
t 2 (kl[a3d04,0) + +]
P= kal(ka,(k) (5.54)
iG
These equations show that photons are boson with energymomentum relation @, =
k. The vacuum state 0) is the state with no photons. All other states of the system
can be generated by applying creation operators repeatedly to the vacuum state.
In the limit Q — , the expansion (5.51) becomes a Fourier integral:
ak 2
Air, 0) = ospsvaas 2 Ath s)[a(k, s)e®*™ + at(k, s)e*™] (5.55)
where we write e(k, s) = €,(k) for consistency in notation. The continuum form of
the annihilation operator is
a(k, s) = VQa,(k) (5.56)
which obeys the commutation relations
[a(k, s), at(k’, s')] = 2m)°5,.5(k — k') (8.57)
The Hamiltonian and total momentum now take the forms
H=~ fee Japs ila ak, 5)
P= S j= ea ark sdall) (5.58)78 Electromagnetic Field
5.5 SPINANGULAR IMENTUM
According to (5.43), the angular momentum density is
u=rx[Ex(V xA)] (5.59)
We define the spin density to be the part that is independent of the origin of r. To
find it, let us first rewrite the preceding in component form:
uf = etkelinennay/B!9,A9 (5.60)
Now combine the last two € symbols according to the rule
etimgnam = 5, Big — BigBin (5.61)
We then obtain
ui = eW'Q/Et9, A? — VEG, A*)
= eV IEIA, AY — dy(x/E"AS) + xI(d, EB") Ak + EA] (5.62)
The factor »/ in the first term cannot be removed by manipulations involving 4,, be
cause j # k. The second term is a total 3divergence, and can be ignored. The third
term vanishes because VE = 0. The last term is independent of r, and is identified
as the spin density:
s=EXA (5.63)
This is the spatial part of the tensor
sh? = AS ae (5.64)
where 223 is the spin matrix given in (3.49). The spin angular momentum is given
by
s=arExa (5.65)
Using the expansion (5.51), we obtain
S= d c{an(boy'ay(k)  af Qax(k)] (5.66)
where k = k/k. To diagonalize this, we make a linear transform to circularly polar
ized photons.5.5 Spin Angular Momentum 719
The polarization vectors for circularly polarized photons are
1
> AG tie) (5.67)
where the label k has been suppressed. As one can easily verify, the plane wave
Refe,ci0%] (5.68)
represents a traveling wave whose polarization vector rotates in a righthanded
sense about k. This is called a leficircularly polarized wave, because an observer
facing the incoming wave would see the polarization rotating to the left. Similarly,
€ corresponds to a rightcircularly polarized wave. The annihilation operators for a
circularly polarized photons are given by
Leftcircularly polarized: a,(k) = Slat ~ia(k))
Rightcireularly polarized: a.(k)= Zefayh)+iasQ] (5.69)
The commutation relations are
(a.(86), @2(p)] = Bap
[a,(k),a(p)] = [a.(k), a%(p)] = 0 (5.70)
In terms of these, the spin operator becomes diagonal:
S= 2 Kc[a,(e)'a, (06) ~ a1(k)a.(49)] (3.71)
This shows that the photon has spin 1, but there are only two helicity states. The he
licity +1 corresponds to leftcircular polarization, and —I corresponds to rightcircu
lar polarization:
a,(k) annihilates helicity state + 1 (6.72)
In terms of circular polarization, the field operator has the expansion
1
oo ler
A(r, 0) 2 Viati {[e.(k)a,(k) + €(k)a_(k) Je’
+ [e#(k)al(k) + €*(k)at(k)] er} (5.73)
In the convention (5.49) the sense of the circular polarizations remains unchanged
when k > k:80 Electromagnetic Field
€.(k) = ie,(k)
€(k) = i€(k) (5.74)
5.6 INTRINSIC PARITY
Let us make a coordinate transformation r — r’, with the transformation law
3
=> pitr’t (5.75)
fi
Since A* is a vector field, this induces the unitary transformation U according to
(3.54):
3
UAi(n)U"! = Spite’) (5.76)
A
where 4/(r) = 4(r, 0). For spatial reflection r’ = —r, we denote the unitary transfor
mation by ?:
PAM EYP! = Ar) (5.77)
This establishes the fact that the electromagnetic field has odd intrinsic parity.
To investigate how photon states transform, we substitute into the preceding the
expansion (5.51). Using the abbreviation
2
a(k) = 26004,(8) (5.78)
we have
PAP! = > Fggn PAVE e+ Pat(k)P tet]
= 2) aa Fla eB + at eT]
=> ca Fla We + atte] (5.79)
©
where the last relation is obtained by changing the summation variable from k to
~k. Thus
Pa(kyP"' = a(k) (5.80)5.7 Transverse Propagator 81
which gives
Pa,(k)P! = —a,(k)
Pa,(k)P"' = a,(k) (5.81)
In terms of circular polarization, we obtain
Pa,(k)P"' = —a(k) (5.82)
A onephoton state of momentum k, linear polarization s, is defined by
Ik, s) =a 3()/0) (5.83)
States with circular polarization are given by
Ik, +) = aJ()/0) (5.84)
which are linearly superpositions of states with linear polarizations:
1
4+) = =s5[k,1) +] e
Ik, +) Valk, ) + i]k,2)] (5.85)
Assuming that the vacuum state is invariant under reflection, we have
Ik, +) = Pa}(k)P0) = —k, *) (5.86)
Thus, under spatial reflection, left and right are interchanged, and the state vector
changes sign.
5.7 TRANSVERSE PROPAGATOR
We now calculate the photon propagator in Coulomb gauge:
Di(x) = i(OTA‘(x)4/(0)10) (5.87)
where the subscript “T” reminds us that the field is transverse: #4*= 0, Expanding
the field in creation and annihilation operators, we have
i d?kd?k! f Ola'(k, s)ait(k’, s‘)]0e** (vy > 0)
Died= a) Vaeat {tie "che >) 688)82 Electromagnetic Field
where
o=ky (5.89)
and we use the abbreviation
a'(k, s) = e(k, s)a(k, s) (5.90)
The vacuum expectation values are easily calculated:
(Ola‘(k, s)a"(k’, s°)]0) = (Ola’(k’, s")a't(k, s)0)
= (277)36,,5(k — k’)e(k, s)e(k’, s’) (5.91)
Therefore
Di) = @ a je EE eet i(k) (5.92)
where /! is defined in (5.50). This can be rewritten as a fourdimensional Fourier in
tegral, with the help of the identity
eo
ea .
sal Mam IM) (5.93)
20
The final form is
elke
Dil) = lee B+in
Mk) (5.94)
where 42 = ko? k?. The Fourier transform is
Bim =
Cad ) (5.95)
Fral® ae
This is not Lorentzcovariant, for it is in Coulomb gauge. To prove that the quan
tized field theory is covariant, we should exhibit the gauge transformation that will
maintain the form of the transverse propagator under Lorentz transformations.
However, this is unnecessary, as we shall show in Chapter 11. The point is that non
covariant part of the propagator is physically irrelevant, because, owing to current
conservation, it does not contribute to the scattering amplitude.
5.8 VACUUM FLUCTUATIONS
The vacuum state is neither an eigenstate of E nor B, since these operators annihi
late or create photons singly. Although the fields average to zero, their meansquare5.8 Vacuum Fluctuations 83
fluctuations are large. This can be shown via direct calculation, as in Problem 2.8.
We can also demonstrate it through the following argument. The energy density in
the vacuum state is
+ (0E? + B?0) = (OE?0) (5.96)
for the freefield theory is invariant under the duality transformation. Equating this
with the zeropoint energy per unit volume in (5.54), we have
1 1
(OIE) = 55D = Gop [HRI 697)
which diverges because of the shortwavelength modes. This divergence is harm
less, since only energy differences have physical significance; but the longwave
length part of the fluctuations gives rise to observable effects, including the Casimir
effect.
We illustrate the essence of the Casimir effect in a simple onedimensional ex
ample, leaving for the next section a more detailed treatment. Consider the modes
of a harmonic oscillator in a box of length L. The zeropoint energy is
E(L)= +> flo)
o=— = (n=1,2,...%) (5.98)
where we have introduced a cutoff function (w), with the properties
A0)=1
Ko) =>0 (5.99)
There is a cutoff frequency «,, above which /(w) decreases rapidly to zero, and we
take the limit w, — © eventually. Suppose that a partition is inserted, such that nor
mal modes are required to have a node at the wall. The modes near the cutoff fre
quency are hardly affected, because their wavelengths are vanishingly small. There
fore, there are now fewer normal modes below the cutoff, as illustrated in Fig. 5.2,
and the zeropoint energy decreases.
For definiteness, choose the cutoff function to be
Slo) = ewe (5.100)
The zeropoint energy for a box without partitions can be easily calculated, with the
result84 Electromagnetic Field
Figure 5.2. When a wall is inserted into a box, those normal modes that do not have a node at the wall
are suppressed. Consequently, the number of modes below a fixed frequency decreases, and the zero
point energy is lowered.
7 1 Loe 7
ie ne ae >
BL sinh(mlaL) = Qn? Dap POW) — 101)
Eg(L) =
Now insert two partitions centered about the midpoint, separated by distance a. The
box is divided into three compartments—one with length a and the others with
length (L  a)/2—and the zeropoint energy becomes
Lez 7
La 7
= F(a) + 2p — SE  ;
B(@)= Ela) + 2Eo>~ 8 Fe ~ GLa) ~ Tha (5.102)
In the limit L — &, the attractive force between the walls is given by
dE@)_
Ga (4a? 6.103)
which is independent of the cutoff.5.9 TheCasimir Effect 85
5.9 THE CASIMIR EFFECT
We now calculate the force between two metallic plates in the electrodynamic vacu
um. The first task is to obtain the normal modes of the electromagnetic field in a
perfectly conducting box of size a x b x c. We choose one corner of the box as ori
gin, and use Coulomb gauge. On each face of the box, the boundary condition is
Ej=0 B,=0 (8.104)
where the subscripts  and + denote respectively the tangential and normal compo
nents. We put B= V x A, E=A, to obtain
Aj=0 (VxA)L= (5.105)
On the yz plane, for example, the boundary conditions are
A,=A,=0 4,4,6.4,=0
The first says that A is normal to the surface, and therefore the second condition is
automatically satisfied. We must, however, satisfy the gauge condition
4,4, + 8,4, + 4,4, =0 (5.106)
which leads to
A (5.107)
Thus, the boundary conditions in Coulomb gauge are
Ay=0 (5.108)
For A,, for example, the conditions are
[44s ¥, 2) =0 = [84:0 ¥, Zena = 0
Axx, ¥, 0) = AxQ y, a) = 0
AG Y, 0) = ALG y, €) = 0 (5.109)
A complete set of solutions to the wave equation is given by
A,= + cos(k,x) sin(ky) sin(kz)86 Electromagnetic Field
in(k,x) cos(k,y) sin(k,z)
in(k,x) sin(k,y) cos(kz) (5.110)
where
mn, f= om
b marc
(5.111)
with n,= 1,2,..., %. The frequency is given by
«y= VET +R (6.112)
If all three components of k are nonzero, there are two independent solutions corre
sponding to the + signs in 4,. If any component of k vanishes, there is only one so
lution, For example, if &, = 0, then 4, = 4, = 0, and the + sign does not make any dif
ference. We can now obtain the zeropoint energy:
Eo(a,b,c)= 4 Y Vi +R F(VE +) (5.113)
oe
+S Vi+B+BRVE +R +R) (5.114)
Keoliyskz
where F(k) is a cutoff function.
Consider now a large cubicle box of edge L, which is divided into three com
partments as shown in Fig. 5.3, with two parallel metallic plates inserted normal to
the x axis, separated by a distance a, symmetric about the midpoint. The zeropoint
energy is the sum of those of the compartments. That of the middle compartment, of
dimensions L x L x a, with L > ©, is given by
L
Figure 5.3 Two metallic plates separated by distance a in the electrodynamic vacuum, which is repre~
sented by a cube of edge L > =.5.9 TheCasimir Effect 87
oe)
u@= Pe LL)
“x = > sh oie Fb) + 25,
(5.115)
where
oo [raver 2) (5.116)
We can rewrite the nsum using the EulerMacLaurin formula [2]
>  1 Br cng) Ba girs
Sate [am G(n) + 5 GO)~FG'O) FG" +++» 6.117)
where B, = 4, By =—ss. Using G’(0) = 0, G'""(0) = 4, and the fact that all higher
derivatives vanish at n = 0, we obtain
Som [a Gin) + + [ ovor( 22)  =
a
This leads to
n
Ut@)= 2 a+  es  (5.118)
where
C= afar [aeevere Fe
V4mlo 4
1
C= ge [deere (5.119)
The zeropoint energy in the box in Fig. 5.3 is given by
E(a) = Ula) + 2U(L  a2)
= ela +20, +O)
mal
This gives an attractive force per unit area between the plates:88 Electromagnetic Field
1 @)_
“Pa 20a" oe
or, in practical units,
whe 01
f= aa  “2 dyn/om? (5.121)
where a is in micrometers. Figure 5.4 compares this result and early measurements
[3], with reasonable agreement. More recent measurements of a similar force be
tween a plate and a sphere have achieved much greater experimental accuracy [4].
5.10 THE GAUGE PRINCIPLE
We now discuss how the electromagnetic field should be coupled to charged fields.
A nonrelativistic charged particle obeys the Schrédinger equation
1
(V +ieAP +e] iar, 1) cue. (5.122)
0.10
0.05
0.01
0 0.5 1.0 1.5 2.0
d (micron)
Figure 5.4 The Casimir attractive force between two metallic plates in the vacuum: x—chromium
steel, o—chromium; solid line—theory. [Data from M. J. Sparnay, Physica 24, 751 (1958).]5.10 The Gauge Principle 89
where A, ¢ are respectively the vector and scalar potentials of an external electro
magnetic field and e is the charge of the particle. We can derive the form of the in
teraction as follows. In the absence of external fields, the Schrédinger equation is
invariant under a global gauge transformation
Ur, )—> eur, t) (8.123)
where « is an arbitrary real constant. The invariance depends on the fact that dy)
transforms in the same manner as y. If we make a local gauge transformation, with
dependent on r, f, this condition will not hold, for we have
Hy Cae i( Hoy] (5.124)
To make the equation invariant, we must cancel the terms involving de. This can
done by introducing the fields 4 = (A,¢) through the replacement
2 ys) {2 + ieanca)] 4a) (6.125)
2 yo [© + ier] 100 ;
a, X,
The Schrédinger equation is now invariant under the local gauge transformation
AMX) > AMR) + dF Y(x)
Wx) > expliex) W(x) (5.126)
The quantity
Deft) = [a + ieA*(x) Wx) (5.127)
is called the covariant derivative, A» is called the gauge field, and the recipe for re
placing 4 by Dé is called the gauge principle.
‘Actually, the gauge principle works only for a fully relativistic theory. For the
nonrelativistic Schrédinger equation, it fails to produce magnetic moment terms of
the form —pV x A, which has to be put in by hand, with 2 arbitrary. In the rela
tivistic Dirac equation discussed in the next chapter, the gauge principle gives the
full electromagnetic interaction of the electron, with a completely determined mag
netic moment.
‘As a relativistic example, consider the complex scalar field with Lagrangian
density
Lox) = Bey ep mn? rp (5.128)
which is invariant under the global gauge transformation
Wax) > etx) (5.129)90 Electromagnetic Field
where w is a constant; but it is not invariant when w depends on x. To extend the
symmetry to local gauge invariance, we make the replacement
Di = 9 + iedH(x) (6.130)
where e is the electric charge. The Lagrangian density is generalized to
L(t) =4 FHF, + (Dey) "Dey (6.131)
which is invariant under the local gauge transformation
G(x) > eR POR)
AM(x) > AMX) + dEX(x) (5.132)
where x(x) is an arbitrary spacetime function. The Lagrangian density of the free
electromagnetic field is included to make the system selfcontained dynamically.
PROBLEMS
5.1 The Lagrangian density for the electromagnetic field in the presence of an external cur
rent density j# is
L=3 FOR, JHA,
‘What is the condition on j# for this to be gaugeinvariant?
Consider the symmetric tensor
u
5.
2
T=) eta
4
(@) Show that #7! = kT! =0, and T= 2.
(b) Using the preceding conditions show the statement in (5.50):
Kd
Hb) = 8, Ts
5.3. Verify that the field operators (5.51) satisfy the commutation relations (5.45). Show, in
particular, that the transverse delta function arises from the transversality of the polar
ization vectors expressed by (5.50) .
Rotations Apply the transformation law (5.76) to rotations. In particular, let R be a r0
tations of the coordinate system about z axis through g, and ¢ be that about the x axis
through 7.
5.
=Problems = 91
x=x' cos p+y' sing xx’
R:y y=x'sin 9+ y'c0s @ Ly yay
zaz'
‘Show that the creation operators transform as indicated in the following:
Ral(k)R* = e¥al(k) Eal(K)E" = al)
Ral(lQR* = elvalk) gale '= al)
Ral(k)R" = el¥al(k) gale?
Rat(K)R! = eiealk) al(K)E =
5.5 TwoPhoton States [5] We can obtain interesting information about a state of two
photons by examining its behavior under rotations and reflection. Consider two photons
with momenta k and ~k. There are four independent states of polarization, which can be
classified according to circular polarizations:
I++) = ala!)
1) =al(ia!(t0)0)
F) = al(kyalt9)0)
+) = al(keat(4))0)
(a) Verify that, in terms of states with linear polarization,
= [a{(d0a{(k)  a}(00a}(K)] (0)
A+) = [al doiaito) + a}(ka}(¥)] [0)
H+) = [al QoatK) + a}(k)ahKs) + ia} (k)a}(A0) ~ ia}(tepa}(K)] 0)
I) = [al Wal (A) + a}(he)a}(H) — ia}()a}(Ao) + ta}(k)a{(k)] 0)
HH) +S)
From this, note that the polarization of the two photons are correlated:
© Inthe state ++)+—) the planes are parallel.
© In the state ++)—) the planes are orthogonal.
In the states +) and +), the planes have equal probability of being parallel or
orthogonal.
Work out the transformation laws for the four polarization states under R,&P, using
results of the last problem, and the fact that the vacuum state is invariant. Verify the
results summarized in the table of eigenvalues (listed whenever the state in ques
tion is an eigenstate of the operation indicated):
(b)
Byte) KH) I) EY)
1 1 ete ete
él 1
Pol 1 1 o492 Electromagnetic Field
(©) From the preceding table, verify the following quantum numbers for a twophoton
state:
© The only state with odd parity is +)~). There are three states with even par
ity: }+4)+1), +), and [+).
For odd total angular momentum J= 1,3,5,... ., the only possible states are +~)
and +). The reason is as follows. The other two states are both eigenstates of R
and € with eigenvalue 1. However, an initial state that is an eigenstate of R with
eigenvalue 1 must have the rotation properties of the spherical harmonic Y/°,
and therefore changes sign under € for J = 1,3,5,. .
© For total angular momentum J = 0,1, the only possible states are ++) +~) and
[++}1), because the other two states have spin projections + 2 along the z
axis, values that are too large for J= 0,1.
(@)_ Verify that a twophoton state cannot have J=
This gives Yang's selection rule [6]: A spin J particle cannot decay into two photons, For
example, just by observing that the 7 meson decays into two photons, we can conclude
that its spin cannot be 1. (Its, in fact, a spin 0 particle.)
5.6 Dirac Monopole A magnetic monopole has a magnetic field Baie = gf/r?, with total
magnetic flux 47 g. Accommodate such a magnetic field into Maxwell’s equations in
the following manner. To keep VB = 0, postulate that there is a return flux 47 g con
centrated in an infinitely thin string attached to the monopole. The vector potential then
consists of a part due to the monopole, and a part due to the string:
A= Apa + Asring
where A,oie is any vector potential that satisfies VX pote = Byring, and is, of course, de
termined only up to a gauge transformation,
(a) Give one solution for Apo.
(b) The shape of the string can be changed through a gauge transformation. For a
straightline string leading from the monopole to infinity, show that the vector po
tential of the string is of the puregauge form
Ascing = 28 V8
where @ is the azimuthal angle around the string.
(©) Consider a quantummechanical particle of electric charge e in the field of the
monopole, with wave function y. Show that the string can be transformed away
through a gauge transformation
Ym erieety
()_ Since w has to be singlevalue, the coefficient of 8 in the exponent must be an inte~
ger n, and thus
ge=ni2References 93
‘This is the Dirac quantization condition. The mere possibility that a monopole can
exist quantizes the electric charge.
(©) Show that the total angular momentum of the system consisting of a charge e and a
monopole g points from the charge to the monopole, and has the magnitude ge. Ob
tain the Dirac quantization condition by quantizing the angular momentum.
5.7, Cutoff Functions
(a) Calculate the vacuum energy (5.98) for a onedimensional system using a sharp
cutoff, which corresponds to f(w) = (0,  @), and show
Lot 3
E(a)= [5 =F* (harp cutoff)
Since this is independent of a, there will be no force between inserted walls.
(b) Show, on the other hand, that any continuous cutoff function will have a nonzero
cutoffindependent force. To do this, write
cao 50S)
Since the argument of f approaches a continuous variable in the limit w, —> 2, we
can approximate the sum by an integral, using the EulerMacLaurin formula
6.117):
Se el
EelL)= Henna aE
 Let nae 2
= 2 Jarygy se + 00%
The cutoffindependent term is the same as that in (5.101).
REFERENCES
1. K. Huang, Quarks, Leptons, and Gauge Fields, 2nd ed., World Scientific, Singapore,
1992, Chapter 8.
2. M. Abramowitz and I. A, Stegun, Eds., Handbook of Mathematical Functions, National
Bureau of Standards, Washington, DC, 1964, p. 806.
M. J. Sparney, Physica 24, 751 (1958).
S. K, Lamoreaux, Phys. Rev. Lett. 78, 5 (1997).
. C.N. Yang, Phys, Rev, 77, 242 (1950).
. C. N. Yang, Phys. Rev. 77, 242 (1950); L. D. Landau, Dokl. Akad. Nauk (USSR) 60, 207
(1948),
awayDirac Equation
6.1 DIRAC ALGEBRA
A relativistic wave equation must treat space and time on the same footing. The
KleinGordon equation does that, but it involves second time derivatives, a feature
responsible for its failure as a oneparticle equation. Dirac tries to remedy this by
proposing a firstorder differential equation. To obtain a equation for the wave func
tion ythat is linear in the spacetime derivatives 2), Dirac writes
(éy#d,, — m) Wx) = 0 (6.1)
where the +" are numerical coefficients, so far undetermined. To satisfy the rela
tivistic kinematics, yx) must also satisfy the KleinGordon equation. Multiplying
from the left by (yd, +m), we have
0= (ya, + m)(éy*9,, — m2)
= rtd dy + MyM)
= LECH Yt YG ud, +m] Wx) (6.2)
This reduces to the Klein~Gordon equation
(CP + my Wx) =0 (6.3)
if and only if
HY + P= 2h" (64)
This algebraic relation defines four objects “, which anticommute with one anoth
er, with
946.1 Dirac Algebra 95.
oy
OP =1 (6.5)
Clearly +“ cannot be numbers. They can be represented by matrices, called Dirac
matrices.
According to (6.4) *y” should be a Hermitian matrix. Thus “ is either Her
mitian or antiHermitian. Putting « # v and taking the trace of both sides in (6.4),
we obtain
Try=0 (w= 0,1,2,3) (6.6)
This condition immediately rules out matrices of odd dimension. It also rules out
dimension 2, for there are only three independent traceless 2x2 matrices—the Pauli
matrices. Therefore, the dimension must be at least 4. That a 4x4 representation ex
ists can be shown by explicit construction.
Define the following 4x4 Hermitian matrices:
“(2 &( 3) ©
where 1 stands for the 2x2 unit matrix, and o* are the 2x2 Pauli matrices:
ol 04 1 0
=  =
a i 4) o fr 3) o (0 4) (68)
‘We shall not use different notations for 2x2 and 4x4 matrices, since the context usu
ally makes the meaning clear. It follows from the definitions that
(ap = p=
{at B}=0 (6.9)
A standard representation for the Dirac matrices is
y= =(6 if] (6.10)
0 l
0 ot
k= sgt =
yt= Pat (. a ) (6.11)
‘The matrix y’ is Hermitian with (y°)? = 1, and y*is antiHermitian, with (y= 1:
Y= OPFHI96 Dirac Equation
ayy OAP=1 6.12)
From these we can show
Aone = 6.13)
The representation given here is not unique. A unitary transformation SS! gives
an equally acceptable set of matrices, since such a transformation obviously pre
serves (6.4).
The and their products, together with the unit matrix, generate a set of 16 in
dependent 44 matrices, in terms of which any 4x4 matrix can be expanded. We in
troduce special symbols for some of their products:
ys= ivy yy
on = Sty) (6.14)
The “fifth” Dirac matrix ys is Hermitian, with square 1, and anticommutes with all
four ys
(95)" = Ys
(ys =1
{YY} =0 (6.15)
In our standard representation it has the form
oe G 0) (6.16)
The generalized Pauli matrices o#” = c#” have six independent members:
of = iat
ot = ett (6.17)
where o* denotes the matrix of 2x2 blocks made up of Pauli matrices along the di
agonal. It is straightforward to show that
of = ys
Ports = oH (6.18)
A complete set of 16 independent 4x4 matrices I’, is given in Table 6.1. By de
finition, we take I'y = 1. All the I, are traceless except for To:6.1 Dirac Algebra 97
TABLE 6.1 Matrices of Dirac Algebra
r, Number
1 1
~ 4
ys 4
on 6
Ys 1
Total 16
TY, =0 (n#0) (6.19)
The set is closed under multiplication and commutation, and is called Dirac alge
bra. The commutators are given in Table 6.2.
An arbitrary 4x4 matrix M can be expanded in the form
(6.20)
where
Tr(MT,)
oT?
(6.21)
TABLE 6.2 Commutators of Dirac Algebra
[1 A] = 274 2e"98 Dirac Equation
6.2 WAVE FUNCTIONS AND CURRENT DENSITY
Rewriting the Dirac equation in a 3vector notation, we have
Cia V+ Am) ye) =: (622)
This looks like a singleparticle wave equation with Hamiltonian
H=ap+ Bm (6.23)
where p is the momentum operator. The wave function (x) is a fourcomponent
column vector called a Dirac spinor:
dy
= %
wl a (6.24)
Ws
where yf, are complex numbers. The complex conjugate is the column vector
w( i (625)
ui
and there are other types of conjugates:
Hermitian adjoint: wt=(WF UF wf Wh)
Pauli adjoint: P= wrP=(Wh vt W Wa) (6.26)
The Hermitian conjugate of (6.1) reads
=i") yt — mypt(x) = 0 (6.27)
Now write yt = 7, and use 7°“t7? = y* to obtain the equation for the Pauli ad
joint:
(4,Diy" + mib=0 (6.28)
Another way of writing this is
Wiy"d, + m)=0 (6.29)63 Plane Waves 99
where the overhead arrow on d,, indicates that it acts to the left.
The conserved density current is given by
aye (6.30)
It is easy to see, with the help of (6.1) and (6.29), that
SF" = (Wd + Pye aw
=mibb+ mpb=0 (6.31)
Note that ° is positivedefinite:
P= b= Why + Po¥ dr + Ws ts + Yate (6.32)
As opposed to
Us = Wy + Yaa — Ws ls — Ua (6.33)
The current j# can therefore serve as a particle current density. As we shall see,
however, the Dirac equation fails to qualify as a singleparticle equation for a differ
ent reason; namely, the energy spectrum is not bounded from below. As we shall
discuss in Section 6.9, the remedy is a redefinition of the vacuum state known as
“hole theory,” which makes the system a manyparticle system. With this modifica
tion, j° will become an operator, whose expectation values are no longer positive
definite, but can be interpreted as charge density.
6.3 PLANE WAVES
Planewave solutions to the Dirac equation can be constructed by putting
Wx) = eP*u(p) (6.34)
where p“ = (p°, p), and u(p) is a column vector called a Dirac spinor:
4 (Pp)
w=  20) (635)
u4(P),
Since U(x) satisfies the KleinGordon equation, we have
pot —p?— mi? =0100 Dirac Equation
For given momentum, there are two roots for the energy p°, with opposite signs:
Do=tE (6.36)
where £ is defined as the positive quantity
E=+Vp tm (637)
The Dirac equation now takes the form
(pm)u(p) = 0 (6.38)
where p is a 4x4 matrix defined by
B= ¥DL= Pp? 7k (6.39)
It has the property
Bat PA =2pq (6.40)
which follows from (6.4).
To find explicit solutions, we note that
(p—m)\(h + m)= p? m2 =0 (641)
Thus, each column of the matrix (~ + m) satisfies the Dirac equation. The explicit
form of the matrix is
m+p? 0 pp
0 +p? " 3
Bem=! os m mye a (6.42)
Ps P 0 mp®
where
pi= pip (6.43)
The number of independent columns can be found by letting p*—> 0, since the ma
trix is a continuous function of p*. In that limit p° = +m, and the matrix becomes
proportional to
1000 0100
0000 ° 0000 9
0000) frre 001 0] frp 0, while columns 3 and 4
are independent for p® < 0. The independent solutions are then columns  and 2 of
(6.42) for p® = E, and columns 3 and 4 for p° = —E. We designate them as u(p, s).
The explicit solutions for p° = E are
1 0
0 1
c  P
up, 1)= Cl oe up, 2)= C] (6.44)
Ps oa
mt+E mt+E
The solutions for p? = E are
* P
m+E mt+E
=P Pe
= 4)= 4
wp.3)=C] P=  ip, )=c A (643)
1 0
0 1
where
E=+Vp+m
m+E
c (6.46)
For a given p, these solutions form an orthogonal set:
E
WN(p, s)u(p, 8’) = 783s (6.47)
For a given energy, the wave functions above resemble those of a nonrelativistic par
ticle of spin +, and it is natural to regard s as a spin label. We shall see that this is a
correct interpretation.
Taking the Hermitian conjugate of (6.38), we have
ul(p, spt —m)=0 (6.48)
Multiplying the equation from the right by 7, and using the identity
Poy =p (6.49)102 Dirac Equation
we find the Pauliadjoint equation
u(p, s\(p~m) =0 (6.50)
Multiplying (6.48) from the right by u, and (6.38) from the left by u‘, and writing
out p more explicitly, we have
ul(yp? — py —m)u=0
ul(yp? + y'p,—m)u = 0 (651)
Adding the two equations leads to the relation
0
ty= in 6.52
wu = iw (6.52)
We can restate the orthonormality of the solutions in the form
U(p, s)u(P, s')= + dy (6.53)
where the plus sign applies for the positiveenergy solutions, corresponding to s =
1,2, and the minus sign is used for the negativeenergy solution with s = 3,4.
6.4 LORENTZ TRANSFORMATIONS
Under a Lorentz transformation x’ = Ax, the Dirac equation in the new frame reads
(ey, — mV’) = 0 (6.54)
Note that remains unchanged, because it is just a numerical matrix, We relate the
new wave function to the old through a linear unitary transformation:
W(e')= Six) SIS=1 (6.55)
where Sis a nonsingular 44 unitary matrix. To demonstrate Lorentz covariance, we
shall show that there exists a nonsingular transformation on * that will restore the
Dirac equation to the old form. Putting 4, = A¥.d, and multiplying the equation by
‘S"' from the left, we obtain
S(ey#AZ A, — m)SWUx) = 0
which reduces to the original equation if (S"'yS)A¥ = y” or
S1yS= Mey” (6.56)64 Lorentz Transformations 103
The existence of S will be demonstrated by explicit construction.
It suffices to consider an infinitesimal Lorentz transformation
XH = xt + ot?” (6.57)
where « contains six infinitesimal parameters, the three rotations @ and the three
boosts of the coordinate frame:
co = af = yk
aii = y + ¥. The x mesons were ob
served after being stopped in matter, and found to be longitudinally polarized. The
preceding result shows that they had the same longitudinal polarization at the mo
ment of decay.6.7 NONRELATIVISTIC LIMIT
We shall study the nonrelativistic limit of (6.82), by first putting it in secondorder
form, Multiplying it from the left by #°D, + m, we obtain
(YD D, + mPyb=0 (6.93)
Writing the first term as half the symmetric part plus the antisymmetric part with
respect to the labels z and v, we can show
PYD,D, = gD Dy + 2 HY Dy Dy] (6.94)
A straightforward calculation gives
[D,, D,) = ieF (6.95)
‘We thus arrive at the secondorder equation
(>, + Sot ae + m) u=0 (6.96)
Consider a stationary solution of energy E, with di/ot = iE . We can rewrite the
equation in the form
[(p eA)?  eo B+ iea: EmJy= (E eh)? (6.97)
where we have used the relation
to"F,,=—o B+ ia E (6.98)
The equation displays a magneticmoment term @  B, with electricmoment term
a: E generated by the moving magnetic moment.
In the nonrelativistic limit the components y, and y, are small, and it is conve
nient to rewrite the above in twocomponent form by putting
Ue (3) (6.99)
where y and é are twocomponent column vectors. Substituting this into the Dirac
‘equation (6.82), we obtain the coupled equations
_Eebmyxo (p—eA)E=0
(Eeb + m)Eo p—eA)y=0 (6.100)110 Dirac Equation
Solving for the “small” component & we have
go: (peA)x
Eeb+m say)
which shows that it is of order p/E compared to the “large” component x.
‘The secondorder equation can be rewritten in the block form
(meet 1B) 2) 2e0r( 2) (6.102)
ieoE ieoE
where a= p — eA. We write the equation for x, and eliminate ¢ with the help of
(o E\(o 7)
[= co Bene ie Eeb+m
r ecore (6.103)
which is an exact equation. We go to the nonrelativistic limit by putting
E=mte (6.104)
and assume € 0:
p*=(E, p) (6.132)
In 3vector form, the equations read.
(@: p+ Bm)u(p, s) = Eu(p, s)
(a: p + Bm)o(p, s) = Ee(p, s) (6.133)
Note that the energy of o(p, s) is still negative, for all we did was reverse the mo
mentum, and write its energy as —E (with E > 0).
The orthonormality of the solutions is expressed by the relations116 Dirac Equation
U(P, S)u(P, 5’) = de
Up, sp, 8") = 3
Up, s)u(p, s')=0 (6.134)
which are equivalent to
E
ul(p, s)u(p, s') = 7 Bae
E
it = 5.
MP, SAP, 8') = TBs
ul(p, s)op, s’) = 0 (6.135)
The completeness of the solution is stated as
Sup (6.136)
2
DLP, s)us*(p, 8) + op, s)o#(p, 8)] =
&
where a and b are spinor indices. This is equivalent to the matrix equation
2
> lp, s)u(p, 5)  o(p, Jap, 5)] = 1 (6.137)
a
The terms above are respectively projection operators onto positiveenergy and neg
ativeenergy states:
A.(p)= up. site, s)= =P
2 
A.@)=Sa. yap, 9)= =F (6.138)
sl m
which have the properties
[APP= ALP) Alp) +A.(@)=1 (6.139)
Note that the 4vector p* in /p is defined to have positive time component p =
The Dirac equation cannot be a oneparticle equation, but it furnishes a finite
dimensional representation of the full Lorentz group. As such, it provides a com
plete set of oneparticle wave functions, in terms of which we can analyze the oper
ator of a spin+ field, as we shall do in the next chapter.6.10 Charge Conjugation 117
6.10 CHARGE CONJUGATION
An antiparticle should have opposite charge to a particle, since it represents the ab
sence of a particle in the negativeenergy sea. This is intuitively obvious; but let us
make certain that the formalism gives this result. In the presence of an external elec
tromagnetic field A¥(x), the Dirac equation is as given by (6.82). We denote the
wave function as (x) for positiveenergy plane wave states, and y(x) for negative
energy planewave states:
Ux) =e PXy(p, 5)
YE(x) = EP *(p, 5) (6.140)
where E = + Vp? + m2, Then (6.82) can be rewritten
Liy"(6,. + ieA,,) — m] Wx) =0
liv"(6,.—ieA,) — m] YF) = 0 (6.141)
which show that the charge indeed has opposite signs for particle and antiparticle.
The two equations above can transformed into each other through “charge conjuga
tion,” or “particleantiparticle conjugation.” To change the sign of the coupling
term in the first equation, we take the complex conjugate:
Liy*#(d,, — ied.) — m] YA(x) = 0 (6.142)
We then make a unitary transformation to bring it to the form of the second equa
tion. Thus
YO) = mY) (6.143)
where 7) is a 4x4 matrix with the properties
wel (6.144)
TW nyn=
The solution is, in our standard representation of the Dirac matrices,
n= (6.145)
(where 7? is the second Dirac matrix). In terms of the spinors, charge conjugation
corresponds to the transformation
v(p, s) = mu*(p, s) (6.146)118 Dirac Equation
Since {7, P} = 0. This shows that particles and antiparticles have opposite parity.
Like the time reversal discussed earlier, the charge conjugation here is an oper
ation on Dirac wave functions, and not on physical states, which are defined in
quantum field theory. The operation is relevant because we expand the quantum
field operators in terms of Dirac wave functions.
6.11 MASSLESS PARTICLES
For a massless Dirac particle, with m = 0, the equation for the Dirac spinor reduces
to pu(p) = 0, or
a@pu(p) = pou(p) (6.147)
where
po=tE E=p (6.148)
Since [a, ys] = 0, we can diagonalize +s, whose eigenvalue +1 is called “chirality.”
The solution with chirality +  is called “righthanded,” denoted up; one with chiral
ity 1 is called “lefthanded,” denoted u,:
‘Ystie(P) = Ua(P)
‘Y¥sui(P) = —u(P) (6.149)
Using the relation
yWa=o (6.150)
we have
o: pu(p)= o ‘ysu(p) (6.151)
which states that the helicity op is the chirality time the sign of the energy. Thus,
for a righthanded particle, the helicity is correlated with the sign of the energy, and
for a lefthanded particle it is anticorrelated. For a given momentum p, the four in
dependent solutions are uc(p, s), where C = R, L denotes chirality and s = + 1 de
notes helicity. Explicit solutions can be obtained from (6.45) by putting m = 0; but
obviously we cannot normalize them according to (13.105). Instead, we put
udp, S)uicAp, 8") = Beco 2E (6.152)
Itis easy to show Yic(P, s)uic(P, 5) = 0, it follows that for, since {5, 7} =0, it folProblems 119.
lows that # and u have opposite chirality. The oneparticle states p) have the proper
ties
(Pip) =280n)9@P)
Joie w= (6.133)
We must, of course, define the vacuum using hole theory. In analogy with the mas
sive case, we define antiparticle spinors:
Vc(P, 5) = uc(P, =5) (6.154)
For a given p, the independent solutions can be taken to be ug(P, 1), 2R(P, 1),
u,(p,—1), oL(p, 1). Thus, a righthanded particle is a righthanded screw, and a left
handed particle is a lefthanded screw. The correlation between handedness and he
licity is reversed for antiparticles.
PROBLEMS
6.1 Lorentz Boost
(a) The transformation matrix for an infinitesimal Lorentz transformation is of the
form S= 1 + iR, where R satisfies 6 60). Review the argument leading to the form
R= Cw,,0#, and show that
(b) Using the identity +P {oy = o#, show that
Srp =S
With this, verify the transformation law for i given in (6.72).
(©) Obtain the freeparticle solutions u(p, s) to the Dirac equation by applying a
Lorentz boost to the solutions in the rest frame:
up, s) = [eos £ + pesinh =p
where $= tanhr'y, and
b=
euoS
SoHs120
62
63
64
65
Dirac Equation
Intrinsic Parity Show (6.73) that
u(p, s) (s=1,2)
uno Mio) Gade
and therefore
Yup, s)=u(p,s) — (s=1,2)
YAp, 8) =2(p,8) (8= 1,2)
These relations indicate that particles and antiparticles have opposite intrinsic parity.
PauliTerm The Dirac equation describes a particle with g= 2. Physical particles have
g factors different from 2 because of interactions, which give rise to an “anomalous”
magnetic moment. The electron acquire the anomalous moment through interactions
with the quantized electromagnetic field. That for the proton and neutron are dominated
by the strong interactions. Suppose that the g factor is 2 + «. Show that this can be ac
commodated by taking the Dirac equation in external electromagnetic field to be
[may + ieA,) Set Fae n Ua)=0
The extra term is called the “Pauli term.” For the proton and the neutron, the experimen
tal values are xp = 1.79, ky =—1.91, respectively.
Chiral Current The chiral current density is defined by
F8) = Vey ysl)
Using the Dirac equation, show that
4,58 = 2mibys
The chiral current becomes conserved in the massless limit m — 0. In quantum field
theory with electromagnetic interactions turned on, this property is destroyed by the ax
ial anomaly [3].
Zitterbewegung The zitterbewegung [4] is a kinematic property of the spin+ repre~
sentation of the Lorentz group, the “clockwork” of the Dirac equation. To exhibit this
motion, construct a wave packet for a Dirac particle:
&p
Qay
te.) = [SP errtns(pye® + wipe
where £ = +V/p? + m?, and w,(p) are linear combinations of Dirac spinors with positive
(negative) energies #£. Calculate the expectation value of the velocity (v) = f drueah
Show that
&p
ny
(Vv) = Vo + 2Re f (wt anw_jere*References 121
where vo = (27)f p(w! aw. + w! anw_) . Integrate this to obtain the average position
(1) = to + vot + inf
The last term is the zitterbewegung, which arises from an interference between positive
and negativeenergy states. On dimensions grounds, we can conclude that the amplitude
of this oscillatory motion is of the order of the Compton wavelength I/m, and therefore
unobservable. In the hole theory, when all negative energy state are filled, the zitterbe
wegung becomes part of the vacuum fluctuations of the Dirac field, for it can happen
only when holes are momentarily created as a result of fluctuations.
6.6 Gordon Decomposition
(a) From the definition of the Dirac matrices, show that,
yey = gh iow
(b) Multiply the equation (iy“a,, — m)ys= 0 from the left by Jy", and use the identity to
©
rewrite the result in the form
= Lo =
Trty= 5 (Mond) — (OD + aor)
with spatial components
= Fp CVD HT 1+ T «Gow +5 Can}
This is the Gordon decomposition, which splits the current density into a “convec
tion” part, plus contributions from the spin. It suggests that the spin is the orbital
angular momentum of the zitterbewegung.
Let u,= u(p,, s,), = 1,2), be two Dirac spinors. Let
Prem pit + pt
k= ptpt
Show that
Beye, = HP" + oh uy
6.7 Massless Particles Consider massless Dirac particles.
(a)
Show iic(p, s)uc(P, s) = 0.
(b) Show that the projection operators A,(p) for positive and negative energies have
the properties 122 Dirac Equation
Aup)=Etep
AL(p) = EA.(p)
Ad(p) + A(p)=E
REFERENCES
1. K.M. Case, Phys. Rev. 106, 177 (1957).
2. RL. Garwin, L. M. Lederman, and M. Weinrich, Phys. Rev. 105, 1415 (1957).
3. K. Huang, Quark, Leptons, and Gauge Fields, 2nd ed. World Scientific, Singapore, 1992,
Chapter 11.
4, K. Huang, Am. J. Phys. 20,479 (1952),The Dirac Field
7.1. QUANTIZATION OF THE DIRAC FIELD
In hole theory, the Dirac equation describes a manyfermion system, and thus the
Dirac “wave function” x) should be regarded as a classical field to be quantized
according to Fermi statistics. To carry out the quantization in the canonical formal
ism, we take as classical Lagrangian density
L(x) = Wayliy"d,,— m) Wx) (7.1)
where (x) is a fourcomponent spinor and the independent field variables are the
components (x). We note that £(x) is Lorentzinvariant, and globally gaugein
variant. This is a firstorder Lagrangian density, involving first instead of second
derivatives with respect to time. We have illustrated theselfconsistency of the
canonical formalism in this case in Problem 4.5. Therefore, following strict canoni
cal procedures, we calculate
a“ ae
mt = Fado =i Ya ou, (7.2)
ait le
The equation of motion is
id,dne + mp=0 (73)
which correctly gives the Dirac equation in Hermitianconjugate form. The La
grangian density vanishes for fields satisfying the equation of motion:
£(x)=0 (for fields satisfying equation of motion) (74)
The canonical conjugate to y, is i, since
123124 The Dirac Field
T= 7a = iE (7.5)
The Lagrangian does not depend on y,*, and therefore y/# has no conjugate. One
must resist the temptation to make the Lagrangian more “symmetric” by replacing
ipy“d,,p with (i/2)[y“(9,.W) + (4, YW). This would be akin to “adding feet when
drawing a snake,” as a Chinese saying goes.
The canonical quantization rules lead to the following anticomutation relations:
(Uelet, ), WLC", D} = 55% = 1")
{Walt , y(t", )} = 0 (76)
where a and b denote spinor indices. The anticommutators serve as initial condi
tions for the Dirac equation. They also fix the normalization left arbitrary in the
Dirac equation.
The Lagrangian density is invariant under the global gauge transformation y—>
ew, where w is a constant. The associated Noether current is
FER) = Hx) 7)
which is conserved:
Gy iM(x) = 0 (7.8)
The canonical energymomentum tensor, which is associated with translational
invariance, is given by
(79)
with conservation law
,.T#%(x) = 0 (7.10)
The energy and momentum densities are respectively
TQ = WiPw= YCiaV + Bm)
To = Yrioy (7.11)
When integrate over space, they give the Hamiltonian H and total momentum P:
H= f. @r Wi(r, )(iaV + Bm)Kr, t)
P
ifar WEY ur, (7.12)
The generalized angular momentum tensor is7.1 Quantization of the Dirac Field 125
MEA) = YRDLix* — xP yd) + 5 oP) (7.13)
and the angular momentum and the boost operators are respectively
i
= eit dir MH = elf aProtlecta!—40)  Fo] y
, 1
Kiz [ar Mo = ifr Pai — IP + ye) uv (7.14)
From the angular momentum, we can read off the spin operator:
(7.15)
in agreement with what we found in the last chapter.
The oneparticle solutions obtained in Chapter 6 constitute a basis in terms of
which the field operators may be expanded. We normalize the wave functions in a
large periodic box of volume @, and write
7
We D> planet, 8) + byl, 8]
eV 08,
Wr, )= —" fa,,teHPEut(p, s) + bp,ert(p, s)] (7.16)
2 ne ul(p, s) + bye" rK(p, s
where
E,=+Vpetm (7.17)
D
The factor Vmi/E, appears because the Dirac spinors are normalized according to
(6.133) and (6.134):
Hp, utp, 3") = Fw", SMB. 8) = Bu
z
> , m ”
3(P, 8)Ap, 5!) =F oP, sp, 5") =
'p
2(p, s)u(p, s’)= 0 (7.18)
With this factor taken out, we have simple anticommutation rules
{aps ape} = {Booby} = Sav Bop
{apedp's'} = {bps bps} = (apspsr} = 0 (7.19)126 The Dirac Field
which lead to the interpretation that ap, annihilates a particle whose wave function
is u(p, s) and bp, annihilates an antiparticle whose wave function is o'(p, s).
Hole theory is implemented through the statement
y,!0) = by{0)=0 for all p,s (7.20)
This implies that there are neither particles nor antiparticles in the vacuum state 0).
In terms of the annihilation and creation operators we have
H= SE jlaitys— dyed fe) =D, Ea hetys + bfsbys— 1)
e e
P= DY Plahdp, — bp.b}.) =D Plapuays + bf.bpe) (721)
e fe
If we had not used hole theory, b,, would be creation instead of annihilation opera
tor, and 5,,bp,t would have eigenvalues 0,1. Consequently, the Hamiltonian would
not be bounded from below. The sign reversal that makes the Hamiltonian positive
definite, of course, depends on the fact that we quantized the system according to
Fermi statistics.
The charge operator is given by
o= far typu= arr vy
= > @htys — bpsbfe + 1) (7.22)
which shows that particles and antiparticles have opposite charge. The minus sign
above arises through rewriting b,.bj, as —b{,bp, + 1. This is dictated by the fact that
b{,bp, has positive eigenvalues in hole theory. The normalization of Q is arbitrary,
for the magnitude of the charge is determined only when there is interaction with
the electromagnetic field.
7.2 FEYNMAN PROPAGATOR
The Feynman propagator for the Dirac field is a 4x4 matrix
iSp(x) = (OTYAx)(O)0) (7.23)
where the timeordering operator T is defined to include a sign change when two
fermion operators A and B are interchanged:
_fAt)B) if, >t
ae) = [nny thee (724)7.2. Feynman Propagator 127
‘The propagator can be calculated straightforwardly, using the expansions (7.16). Let
x= (t,), For t> 0, the only contribution comes from terms in the expansion of the
form aat:
[Se)ap = (OlWel, 1)Us(0)10) = ae ett) u(P, s)us(P, S) (7.25)
For ¢ <0 we need only keep terms of the form bbt:
Sj 1 P . 7
SHC as = — OO, N10)=—H S FreHPPY nD, iP.) (7.26)
v Fp a
The sum over spin states results in the projection operators given in (6.137). Sup
pressing the spinor indices and going to the limit © — ©, we have
L_f(mtpeen — (¢>0)
ison= [55 Gn) 26, al (m— per) (<0) Me
We can make the replacement
m+ pam PE,— yipt> m+ ip Yt (7.28)
because this operator acts on the exponential factor. For ¢< 0, make the change of
variables p—> —p. Then we have
1 iyo — ht) ertEpllgloe
iS(x) = s&s zs (" + iy? vere pli (7.29)
Now use the representation
e POF iE,
othe el Seren rere CED
where 7 — 0°. Then, the operator id/4t in the previous formula can be replaced by
Po. The final result is
ptm
Pips ptm _ 7:
50) =  ioreo eer (731)
We leave it as an exercise to show that
Sex) = (ey#9,, + m)AelX) (7.32)
where A,(x) is the Feynman propagator for a free scalar field of mass m.128 ‘The Dirac Field
The Fourier transform of the propagator is
a
pPm+in pmt+in
5,(p) = (7.33)
where the right side is the inverse of a 44 matrix. The 4vector p is arbitrary, with
either p® > 0 or p® > 0. For p> 0, we have, according to (6.137),
2
(ptm
mo st
u(p, sju(p,s) (p> 0) (7.34)
For p? > 0, let us define g# = —p#, Then according to (6.137), we obtain
2
Cem) I) Sq, 554.8) (p90) (735)
This shows that the residue at the massshell pole at p? = m? contains the wave func
tions of an electron of momentum p, or those of a positron of momentum q = p.
7.3 NORMAL ORDERING
Both H and Q contain divergent contributions from the zeropoint energy and
charge of the vacuum state. These terms have no physical relevance since energy
and charge are measured relative to those of the vacuum state. They can be eliminat
ed by redefining the reference points, and this can be achieved by arranging the or
der of operators appropriately.
We first introduce the notion of normal ordering. Suppose that O is a product
of creation and annihilation operators. The corresponding normal product : O : is
defined as that obtained from O by rearranging the order of the factors, if necessary,
such that all creation operators stand to the left of all annihilation operators. In the
rearrangement process, an interchange of two fermion operators gives rise to a fac
tor —1. As an example:
(7.36)
Normal ordering can be naturally extended to a sum of products:
20, + O,:=:0,:+:0,: (731)
We now redefine the Hamiltonian and the current as
Pri, iaV + Bm), 1) :7.4 Electromagnetic Interactions 129
FE) = yx) = (7.38)
It is clear that these operators give zero when operating on the vacuum state, be
cause annihilation operators stand to the far right. This is just a formal way of stat
ing that the zeropoint energy and currents are to be omitted. As the notation is
somewhat cumbersome, we shall not explicit indicate normal ordering unless neces
sary.
7.4. ELECTROMAGNETIC INTERACTIONS
We consider systems of interacting fields with a Lagrangian density consisting of
the sum of the free Lagrangian densities of the participating fields, plus an interac
tion Lagrangian density that couple the fields together. This is not the most general
case conceivable, but it is what we can handle mathematically. We illustrate the
types of interactions commonly encountered, Consider a Dirac field, a complex
scalar field, and the electromagnetic field, which have free Lagrangian densities
given by
Loirac = Wiy"d,, — mip
Lecatar = HO*3,6 — PG
Lem =—3 FHF yy (7.39)
According to the gauge principle, the matter fields can be coupled to the electro
magnetic (em) field by replacing 4, by the covariant derivative
DH = 4, + ieAMx) (7.40)
where e is the electric charge. Assuming that both the Dirac field and the scalar
field have the same charge e, the electrodynamic Lagrangian density is
LA POE, + HiyrD,— mW [D*SHD,b85*S
= Lam + Loine + Lae + Li an
where
Ling = (Gh + PVA,
ja iebey
Je= ie b*($) — (Hd*)d] + Pb*h AM (7.42)
The matter fields are coupled through conserved currents, which are the Noether130 The Dirac Field
currents associated with global gauge invariance. For the scalar field, the current
has an e? term proportional to 4*. This becomes a mass term for the photon when
*¢ develops a vacuum expectation value, in spontaneous symmetry breaking.
(See Problem 15.5.)
The electromagnetic field couples to all charged fields through the gauge prin
ciple, and is universal in this sense. The vacuum fluctuations of the electromagnetic
field include the momentary creation of virtual particleantiparticle pairs and their
subsequent annihilation. The temporary charge separation makes the vacuum into a
dielectriclike medium, and all charged fields of the world participate in this “vacu
um polarization,” as their contributions being determined solely by charge and
mass.
7.5 ISOSPIN
The Dirac field can be used in a phenomenological description of protons and neu
trons, which are really made of quarks. The effective theory is useful in describing
the “chargeindependent” pionnucleon interactions at low energies. It is based on
the fact that proton and neutron are almost identical, and so are the three mesons,
and the strong nuclear forces respect the identities. By ignoring the electromagnetic
and weak interactions, we can regard the proton and neutron as different states of a
particle called the nucleon, and the a mesons as different states of the pion.
The nucleon field is represented by a twocomponent Dirac field
_ tice)
UG) = ( so} (7.43)
where i= I corresponds to proton, and i= 2 to neutron. Each y is a fourcomponent
Dirac spinor field. Writing out all the indices, we have eight complex fields yi2(),
with a=1,...,4 andi= 1, 2. By analogy with spin angular momentum, we define
the isospin 7/2 as generators of rotations in the twodimensional internal space
spanned by y, and ys:
w(t) a) m6 2 om
The proton and neutron states are eigenstates of 73/2 with respective eigenvalues ++
and—3:
w=(o) b=(;) 45)
which can be created from the vacuum by applying (x). We use a shorthand nota
tion in which the spinor and internal indices are suppressed. For example,75 isospin 131
DP TY = Vil P aol Miley (7.46)
where a, b are summed from  to 4 and i, j are summed from  to 2.
More generally we define isospin as an internal symmetry group whose gener
ators >/ obey angular momentum commutation relations (Lie algebra):
Un T= icy
Thus one can simultaneously diagonalize 7? = (+ 1) and J, and denote isospin
eigenstates by /, /;). The nucleon belongs to the fundamental representation with
I= 4, in which [= 7/2. The overhead arrow denotes a vector in isospin space, which
has three components because that is the number of generators of the group.
The pion field has /= 1, and is described by a threecomponent real field
iQ)
$x)= bx) (7.47)
30)
This the “adjoint representation” of the group, which has the same dimension as the
number of generators, and in which the generators are represented by matrices 7"
taken directly from Lie algebra:
(Tim = ie" (7.48)
Experimental evidence dictates that (x) be pseudoscalar, that is, that it change
sign under spatial reflection. We note that J; is not diagonal. The physical pion
fields, which are eigenstates of P, are related to ¢, through
2) = elds) +idsC0]
10) = Tpldi0is00]
P(x) = s(x) (7.49)
These operators create states with J; = +1, —1, 0 respectively, when they operate on
the vacuum state,
“Charge independence” in the pionnucleon system means that the interaction
conserves isospin. A Lorentzinvariant effective Lagrangian density, known as
“pseudoscalar coupling,” is given by
L(x) = Hiya, — M+ [Hb H— mS GB] + Cy W'S (7.50)
The vector notation makes manifest the rotational invariance in isospin space. A
competing model is the “pseudovector coupling” model, with132 The Dirac Field
L(x) = Wey"d,,— M+ 3[4,8 4,m°$B) + 2's FW)  3,6 (7.51)
Some consequences of isospin invariance are explored in Problem 7.4.
7.6 PARITY
We discuss the discrete symmetries, using as an example the electromagnetic cou
pling as contained in
L(x) =3 FWP + iy"(d,, + ieA,,) — m] (7.52)
Under a Lorentz transformation x —> Ax, the field operators ¢,(x) undergoe a uni
tary transformation U given by (3.54):
Udalx)U"' = Sapbs(A'®) (7.53)
where S,,. This can be extended to spatial reflection x —> x, t—> 1, for which the uni
tary operator U is denoted by ®. For the Dirac field, we have S = according to
(6.69) and (6.70), and thus
Pr, HP! = Pwr, t)
Pur, HP! =W4,jP (7.54)
Since A* transforms like a vector,
PAE, P =A, t)
PAK, P= Ar, t) (7.55)
Thus we have
PL(r, NP! = L(4, t) (7.56)
which show that the Lagrangian L = f d°x £(x) is invariant.
From the expansion (7.16) at ¢= 0:
We) =>  Gp lane Fup, 5) + erp, 3)] 31)
Ps Ey
we have
Pure'=S. Fe [Pay Pe Tu(p, s) + PLP? (p, s)]
a VE,7.7 Charge Conjugation 133
mn
PO >  Fp net PD.s)+ heh PAp. 3] (58)
Using the relations (Problem 6.2)
Pulp, s) = u(p, s)
Pup, s)=—(P, 5) (7.59)
we obtain the statement that particles and antiparticles in Dirac theory have oppo
site intrinsic parity:
Pay?! =a»,
PbyP =—b ps (7.60)
The transformation P may be accompanied by a rotation in spin space with respect
to the index s, as is clear from (7.58); but we leave it out for simplicity.
7.1 CHARGE CONJUGATION
Charge conjugation, or particleantiparticle conjugation, is defined as a unitary op
eration C on the Hilbert space that interchanges particle and antiparticle, and revers
es the sign of the electromagnetic field:
Cap,C1 = byg
CbpsC! = ap,
CAR(R)C = ~AK(x) (7.61)
The transformation of A%x) is not specified independently, because in Coulomb
gauge it is not an independent field. It is clear that £(x) is invariant under this trans
formation, because the freefield Lagrangian densities are invariant, and the Dirac
field is coupled to the electromagnetic field through the current density, which
changes sign.
To find how the Dirac field operator transforms, let us compare the following
expansions:
HO= Ep, ane. 9) + Be *0, 9]
vo=> ne [ate Tu*(p, s) + bp,e?*a*(p, s)] (7.62)
a V OE,134 ‘The Dirac Field
The expansion coefficients satisfy (6.146):
2(p, s)= mu*(p, s) (7.63)
where 1 = iy? is areal 4x4 matrix. Therefore
+ [TR
WO) =>  op Une up, 3) + ape", 9) (7.64)
w V OB,
which shows
Cur)C! = nb) (7.65)
Note that the Dirac wave functions undergo complex conjugation, which is a
nonlinear operation, because (Au)* = A*u*. The field operator, however, undergoes
a linear transformation, because C(AW)C"! = A Ci C!, The difference can be traced
to the fact that in the Dirac equation we have to change the sign of the coupling to
an external electromagnetic field, whereas in the field theory, the electromagnetic
field is part of the system, and changes sign under charge conjugation.
7.8 TIME REVERSAL
Time reversal is the operation of interchanging past and future, represented by a op
erator T on Hilbert space. Suppose that V, is a member of a complete set of state in
Hilbert space, where a stands for quantum numbers, such as momentum p and spin
projection s on a fixed axis. The timereversed state TV, must be a member of the
same set:
TV,=Vz (7.66)
where @ are the timereversed quantum numbers, defined by correspondence with
classical mechanics:
(7.67)
and the helicity is invariant. The basic property of T is
(TV, TVs) = (Us, Va) (7.68)
that is, it interchanges initial and final states. This can be rewritten
(TV, TVy) = (Yas Vo)* (7.69)
Replacing ¥, by AW;, where A is a complex number, we have7.8 Time Reversal 135
(T¥,, TAV;)) = Va, Vo)* (7.70)
Therefore
TAY, = TV, (1.7)
Thus, when acting on a number, 7 takes its complex conjugate. This makes T non
linear. More specifically, it is called an. “antilinear” operator. A general representa
tion of T is complex conjugation followed by a unitary transformation:
T=Us
T=Us (7.72)
where it is assumed that U commutes with complex conjugation. For the
Schrédinger equation
wv
rae (7.73)
time reversal means
H(IY)= ADD (7.74)
The system is invariant under time reversal if the timereversed equation is equiva
lent to the original. Taking the complex conjugate, we have
HUY)
r; (7.75)
HUY,
Thus, the system is invariant under time reversal if the Hamiltonian is real:
H=H* (7.76)
which implies that the Lagrangian must be real.
Without going through all the details, we can conclude that
TAM)T! =AK(r)
TWe)T = yrystr) (1.77)
The first equation follows from the requirement that 4* transform like the current
density, which must change sign, because classically it is a velocity. The second fol
lows from the fact that +s is the transformation that preserves the Dirac equation136 The Dirac Field
under time reversal, as shown in (6.70). It is straightforward to verify that the La
grangian is invariant, if the charge ¢ is real.
There is a theorem known as the PCT theorem, which states that a local field
theory that is Lorentz invariant is automatically invariant under the product PCT,
even though it may be separately invariant under PC, separately, We refer the read
er elsewhere [1] for proof.
PROBLEMS
7.1 EnergyMomentum Tensor The canonical energymomentum tensor 7. for the
Dirac field is not symmetric in wa. According to Section 4.5, we can construct an
equivalent symmetric tensor TH" = T#2 + £,X™<, Find X™#2,
7.2 Propagator Show that the propagator for the Dirac field is related to that of the scalar
field through
Selx) = (iy#d,, + mg)
7.3. Neutrinos Neutrinos are massless Dirac particles. Using the convention for wave
function given in Section 6.11, expand the field operator in terms of annihilation and
creation operators.
7.4 Isospin Transformations
(a) Show that under an infinitesimal isospin transformation, the nucleon field, and the
pion field transform according to
$ b+8~S
where the components of ~w are arbitrary infinitesimal real parameters.
(b) Let = yl 7, where ¥ is a 4x4 Dirac matrix. Show that / transforms like a vector
in isospin space:
Vs0+ ax
7.5 PionNucleon Scattering As far as isospin properties are concerned, the pion and nu
cleon states can be labeled by / and J
Im')= 11, 1) I) =1,=1) 7?)=1, 0) }
+) In)=14,1\)
(a) A state containing a pion and a nucleon is a direct product in isospin space, as, for
example, a"n) =1, 1) x [3 . However, this is not an eigenstate of total isospin
and thus not an eigenstate of the Hamiltonian of the system. Show that eigenstates
of the isospin are the following:Reference 137
13, 3)= a)
$.4)= [Fear — [Son
5,3)= fina) — [3
13,3)= /3lam®)—/3lpar)
(b) For interactions that conserve isospin, the pionnucleon scattering amplitude de
pends only on total isospin and not on J; (for the same reason that atomic energy
levels are independent of the magnetic quantum number, i.e., the WignerEckart
theorem.) Denote the transition amplitudes by ay and a, and show that they have
the form
(pa Tipm) = asa
(pr Tp)
(na?Tlpar)
5a + Fay2
3 asa 3 ay2
where T is the transition operator. The corresponding scattering cross sections are
proportional to the squares of these amplitudes.
(©) Neglect ay,2 compared to aya, and show that pionnucleon scattering cross sections
bear the ratio
Opa): opr) : (nm > par) =9 1:2
This is verified experimentally at low energies. The reason that ay72 dominates is
the resonance scattering 7+ N—> A—> m+ N, where A is a particle of spin 3,
isopin 3, with mass 1232 MeV, known at one time as the “33 resonance.”
REFERENCE
1. R. Streater and A. Wightman, PCT, Spin and Statistics, and All That, Benjamin, New York,
1964.Dynamics of
Interacting Fields
8.1 TIME EVOLUTION
The dynamics of a quantum mechanical system is governed by the Hamiltonian H,
which generates time translations. One may view the time development from differ
ent perspectives. In the Schrddinger picture, one regards the operators O, as time
independent objects, and the state vector V, changes with time according to the
Schrédinger equation
VO
ie = HY (8.1)
Assuming that H is timeindependent, we have the formal solution
WO) = V0) (8.2)
The matrix element of an operator O, evolves in time according to
(PO IV(D) = (®,(O)le"#O,0"¥ (0) (8.3)
The subscript “s” identifies states and operators in the Schrédinger picture.
In the Heisenberg picture, the state is assumed to be constant in time, but the
operators evolve. The matrix elements of an operator must be independent of the
picture, and this requirement relates the Heisenberg picture to the Schrodinger pic
ture:
¥,= V0)
O,(t) = Oe" (8.4)
138
Much more than documents.
Discover everything Scribd has to offer, including books and audiobooks from major publishers.
Cancel anytime.