(Documents On Modern Physics) Schwinger Julian - Particles and Sources-Gordon and Breach, Science Publishers (1969) PDF

DOCUMENTS ON MODERN PHYSICS
Edited by
ELLIOTT w. MONTROLL, University of Rochester

GEORGE H. VINEYARD, Brookhaven National Laboratory
MAURICE LEvY, Universite de Paris
A. ABRAGAM, L'Effet Mossbauer

H. BACRY, Lectures on Group Theory
S. T. BELYAEV, Collective Excitations in Nuclei
T. A. BRODY, Symbol-Manipulation Techniques for Physics
K. G. BUDDEN, Lectures on Magnetoionic Theory
J. W. CHAMBERLAIN, Motion of Charged Particles in the Earth's Magnetic Field
S. CHAPMAN, Solar Plasma and Geomagnetism and Aurora
H. Cmu, Neutrino Astrophysics
A. H. CoITRELL, Theory of Crystal Dislocations
J. DANON, Lectures on the Mossbauer Effect
BRYCE S. DEWITT, Dynamical Theory of Groups and Fields
R. H. DICKE, The Theoretical Significance of Experimental Relativity
P. FoNG, Statistical Theory of Nuclear Fission
M. GOURDIN, Laws of Symmetry Theorem of T. C. P.
D. HESTENES, Space-Time Algebras
JoHN G. KIRKwooD, Selected Topics in Statistical Mechanics; Macromolecules; Theory
of Liquids; Theory of Solutions; Proteins; Quantum Statistics and Cooperative
Phenomena; Shock and Detonation Waves; Dielectrics-Intermolecular
Forces-Optical Rotation
R. LATTES, Methods of Resolutions of Some Boundary Problems in Mathematical Physics
F. E. Low, Symmetries and Elementary Particles
P. H. E. MEDER, Quantum Statistical Mechanics
M. MosJDNsKY, Group Theory and the Many-body Problem
M. NIKOLIC, Kinematics and Multiparticle Systems; Analysis of Scattering and Decay
A. B. Pn>PARD, The Dynamics of Conduction Electrons
L. SCHWARTZ, Applications of Distributions to the Theory of Elementary Particles in
Quantum Mechanics
J. ScHWINGER, Particles and Sources
J. SCHWINGER AND D. SAXON, Discontinuities in Wave Guides
M. TINKHAM, Superconductivity
Additional Volumes in preparation

Particles
and
Sources
JULIAN SCHWINGER
Harvard University
Notes by Tung-mow Yan
GORDON AND BREACH, SCIENCE PUBLISHERS

New York London Paris.
Copyright© 1969 by Gordon and Breach, Science Publishers, Inc.
150 Fifth Avenue, New York, N. Y. 10011
Library of Congress Catalog Card Number: 69-19390
Editorial office for Great Britain:
Gordon and Breach, Science Publishers Ltd.

8 Bloomsbury Way
London W. C. 1
Editorial office for France:
Gordon & Breach

7-9, rue Emile Dubois
Paris 14e
Distributed in Canada by:
The Ryerson Press

299 Queen Street West
Toronto 2B, Ontario
All rights reserved. No part of this book may be reproduced or utilized in any form or by
any means, electronic or mechanical, including photocopying, recording, or by any in-
formation storage and retrieval system, without permission in writing from the Publishers.
(0206)
Printed in Switzerland by City-Druck AG, Zurich

Editors' Preface
Seventy years ago when the fraternity of physicists was smaller than the
audience at a weekly physics colloquium in a major university, a J. Willard
Gibbs could, after ten years of thought, summarize his ideas on a subject in
a few monumental papers or in a classic treatise. His competition did not
intimidate him into a muddled correspondence with his favorite editor nor
did it occur to his colleagues that their own progress was retarded by his
leisurely publication schedule.
Today the dramatic phase of a new branch of physics spans less than a
decade and subsides before the definitive treatise is published. Moreover,
modern physics is an extremely interconnected discipline and the busy
practitioner of one of its branches must be kept aware of breakthroughs in
other areas. An expository literature which is clear and timely is needed to
relieve him of the burden of wading through tentative and hastily written
papers scattered in many journals.
To this end we have undertaken the editing of a new series, entitled Docu-
ments on Modern Physics, which will make available selected reviews, lecture
notes, conference proceedings, and important collections of papers in
branches of physics of special current interest. Complete coverage of a field
will not be a primary aim. Rather, we will emphasize readability, speed of
publication, and importance to students and research workers. The books
will appear in low-cost paper-covered editions, as well as in cloth covers. The
scope will be broad, the style informal.
vi EDITORS' PREFACE
From time to time, older branches of physics come alive again, and forgotten
writings acquire relevance to recent developments. We expect to make a
number of such works available by including them in this series along with
new works.
ELLIOTT W. MONTROLL
GEORGE H. VINEYARD
MAURICE LEVY
Contents
1. NONINTERACTING PARTICLES 1
1.1 Introduction
1.2 Sources 3
1.3 Spinless Particles 4
1.4 An Application. Stimulated Emission 10
1.5 Spin-1 Particles. The Photon 12
1.6 Massless Spin-2 Particles. The Graviton 16
1.7 Spin-(1/2) Particles 18
1.8 Multispinor Formulation of Arbitrary Spin 25
General Connection of Spin and Statistics
2. ELECTRODYNAMICS 29
2.1 Combined System of Noninteracting Electrons and Photons. 29

Fields
2.2 Primitive Interaction and Interaction Skeleton. Extended Sources 32
2.3 Calculation of Matrix Elements 37
2.4 Two-Particle Exchange 39
2.5 Modified Photon Propagation Function 42
2.6 Electromagnetic Form Factors 46
Vlll CONTENTS
3. STRONG AND WEAK INTERACTIONS 53
3.1 Low-Energy re+ N System 53

3.2 Partial Symmetry. Chiral Invariance 56
3.3 Non-Abelian Vector Gauge Particles, p and A 1 • 60
3.4 Widths of p and A 1 68
3.5 Low-Energy rcrc Interactions 72
3.6 Spectrum of the Decay Process 17*(960)-->17 + 2rc 76
3.7 Pion Electromagnetic Mass 80
3.8 U4 x U4 Partial Symmetry 86
3.9 Concluding Remarks 91
PARTICLES AND SOURCES
1. NONINTERACTING PARTICLES
1.1 Introduction
Research in particle physics presently falls in two main classi-

fications: Quantum Field Theory and S-Matrix Theory. While quantum
field theory, like field theorists, comes in all sizes and shapes we regard its
basic characteristics to be as follows. It is a space-time formulated operator
theory. The fundamental dynamical variables, the fields, describe certain
localized excitations, which in particle language, correspond to all possible
combinations of particles with the prescribed quantum numbers. When the
physical couplings are weak, as in electrodynamics, the relation between field
and particle may appear to be close. For strongly interacting systems it is
certainly very remote. Current algebra was originally motivated by a reaction
against field theory, in which currents, supposedly more physical, were
introduced as basic operators. But the latest development in this area has
been a return to the phenomenological field as primary. The practical
difficulty in connecting field-theory hypotheses with raw experimental data
led to the phenomenological emphasis of S-matrix theory. Here the particle
is primary. Attention is concentrated on the momentum space description of
various collisions. Dynamical principles are sought in the analytic extension
of the momenta to complex values.
In these lectures we shall describe a totally new approach to particle physics.
It is intermediate in concept between the two previous formulations. It
1
2 J. SCHWINGER
shares with field theory the physical emphasis upon space and time, but it is
not an operator theory. Like S-matrix theory, it is phenomenological in its
emphasis upon the actual physical system, but there is no reference to
analyticity in momentum space. The results of quantum electrodynamics are
reproduced without the irrelevance of divergences, or renormalizations. This
means that we remove the unnecessary field-theory hypothesis that detailed
space-time description is possible, down to distances smaller than those
involved in presently accessible phenomena. Characteristic S-matrix features
such as dispersion relations are deduced, not from abstract mathematical
requirements, but through intuitive space-time arguments. Current algebra
results, which have lent credence to detailed speculations concerning the
inner structures of particles, are recovered by purely phenomenological
procedures. Thus, a unified method is now available, one that is economical
in hypothesis and effective in application.
PARTICLES AND SOURCES 3
1.2 Sources
The concept of particle has been steadily extended in the last few
decades. From the stable electron and proton, to the very long-lived neutron,
to the short-lived n and A, to the highly unstable p and N * it has progressed
to more and more short-lived excitations. Thus, in general, particles must be
created in order to study them, since most of them are unstable. In a general
sense this is also true of high-energy stable particles, which must be created
in that situation by some device, i.e. an accelerator. One can regard all such
creation acts as collisions, in which the necessary properties are transferred
from other particles to the one of interest. There will be a variety of collisions
that can serve to create a particle with specified properties. The other
particles in the collision appear only to supply these attributes. They are,
in an abstract sense, the source of the particle in question. In other words,
the source concept is the abstraction of all possible dynamical mechanisms
whereby the particular particle can be produced. We try to represent this
abstraction of realistic processes numerically. The fact that the collision will
have some degree of space-time localizability indicates the utility of a
numerical function S(x) to measure this aspect. The effectiveness of the
collision in liberating various momenta can be measured by another function
S(p). The quantum-mechanical requirement of complementarity can then
be imposed, as in
S(p) = f (dx)e-ipxS(x), (1.1)

where
and units h = c = 1 are used.
Unstable particles eventually decay and the decay process is a detection

device. More generally, any detection device can be regarded as removing or
annihilating the particle. Thus the source concept can again be used as an
abstraction of annihilation collisions, with the source acting negatively, as
a sink. We now have a total picture of any physical situation in which
sources are used to create the initial particles from the vacuum state, and
sources are used to detect the final particles resulting from some interaction,
and then we return to the vacuum state.
4 J. SCHWINGER
1.3 Spinless Particles
To make the concept the basis of a quantitative theory, we consider

first a stable spinless neutral particle of mass m. To begin with, we use a
weak source, i.e. one for which multiparticle production is negligible. To
specify a weak source, we consider its effectiveness in creating a particle
with momentum p, in the small range (dp). An invariant measure of momen-
tum space is
dw = (dp) _1_ po= +,Jp2+m2. (1.2)

P (2n) 3 2p 0 '
We now define the source K in terms of the creation and annihilation

probability amplitudes
(1PIO_ )K =.J dwPiK(p), (1.3)
(O+ I 1P)K =.J dwPiK(-p), (1.4)
which convey the idea that the source liberates or absorbs momentum p in
the respective processes. The square-root signs appear because the proba-
bilities are proportional to the momentum range (dp). The subscript on the
vacuum state indicates the time sense, IO_) is the vacuum state before the
source has operated. The factors of i are included here for later convenience.
There are probability requirements that express the weak source restriction:
(1.5)
p
1~1(0+10_)Kl 2 + .l:1(0+11p)Kl 2 • (1.6)

p
Since
IK(p) I2 =If (dx)e-ipxK(x) I 2
= If (dx)eipxK* (x) I 2
= IK*(-p)l 2 ,
these two conditions, (1.5) and (1.6), will be equivalent if K(x) is real. We
shall take this to be a general property of any source; a complex source is
regarded as a combination of two real sources which describes some physical
property of that multiplicity. We also notice, from the restriction (1.5) or
(1.6), that I (O+ IO_ )KI must differ from unity by terms of order K 2 • It is
consistent to assume that no terms of order K appear in (O+ IO_)\ as we
shall verify later.
Now consider a complete situation in which particles are created by K 2 ,
propagate in space and time, and then are detected by K 1 (Fig. 1.1). In this
FIGURE 1.1
--------------------- o_
case the causal situation is well-defined, the detection source K 1 (x) is cer-
tainly localized in time later than the emission source Ki(x). The overall
description is then expressed by the vacuum probability amplitude
(O+ 1o_y;;(O+ 1o_)K (0+ 1o_)K + _L(O+ 11pY (1PIO-Y

1 2 1 2
p
;; 1 + O(K1 2 ) + O(K2 2 )
+ f dwPf (dx)(dx')iK1(x)eip(x-x')iK2 (x'). (1.7)
We regard K 1 (x) and K 2 (x) as two different manifestations of the same
physical mechanism, that is, they are the values of one general source in
different space-time regions. Therefore the only possible combination that
can occur is the total source
K=K1 +K2 •
This is a fundamental postulate, the principle of the unity of the source,
which embodies the idea of the uniformity of nature. Then the vacuum
amplitude must have the general q~adratic form
(O+ Io_ y = 1+(i/2) f (dx)(dx')K(x)L1+(x-x')K(x'), (1.8)

where it is known that
J
x 0 >x 0 ': (1/2)[ Lf +(x-x')+L1+(x' -x) =if dwPeip(x-x'). (1.9)
Since only the part of L1+(x-x') that is symmetrical in x and x' can con-
tribute to the integral, we are permitted to define
L1+(x-x') = Lf +(x' -x), (1.10)

and
(1.11)
6 J. SCHWINGER
This function is familiar. It is the solution of the differential equation
(m 2 - a2 )L1 + (x- x') = t5(x- x') (1.12)
that obeys the boundary condition of outgoing waves in time, that is,
positive (negative) frequencies for positive (negative) time differences. It
should be remarked that we did not begin with these requirements, they
appear as a consequence of our theory. We record here the alternative
representation for L1 + (x- x'),
Ll+(x-x) =
,I (dp)
--4
(2n) p
eip(x-x')
2
+m
2 .
I
-u; s~+o
. (1.13)
The quadratic terms in K 1 and K 2 that appear in (O+ IO_ )Know reproduce
the structure of the product of ( 0 +IO_ )K 1 and ( 0 + IO_ )K2 • We can now
proceed to check the probability requirement, (1.5) or (1.6). From (1.8), we
get
I (O+ Io_ y 1
2
=1- f (dx)(dx')K(x) Re (1/i)L1 +(x-x')K(x') (1.14)
for a weak source. The notation "Re" means "the real part of". Now
Re(1/i)L1+(x-x') =Ref dwPeip(x-x'), (1.15)
which holds for all x-x', since the right-hand side is symmetrical in x and
x'. Thus
f (dx)(dx')K(x) Re (1/i)Ll+(x-x')K(x')
= Ref dwpK(-p)K(p)
= f dwPI K(p)l 2' (1.16)

and (1.14) becomes
I <o+ Io_ >K 12 = 1- LI <1pIo_>K1 2 , (1.17)

p
which is equivalent to (1.5) or (1.6). It can now be recognized that the

physically necessary minus sign in ( 1.17) is the consequence of the particular
way in which the factors of i in (1PIO_)K and (O+ l1P)K have been intro-
duced. These discussions show that the formalism is internally consistent.
To remove the restriction to weak sources, we make use of the possibility
of preparing directional sources to arrange an arbitrary number of pairs of
weak emission and weak absorption sources in such a way that a particle
emitted by the emission source of one pair will not be detected by the
detection source of another. This situation is illustrated in Fig. 1.2. Since
these processes are physically independent, we have simply
(O+ IO_)K =TI [t+(i/2)f (dx)(dx')Ka(x)Ll+(x-x')Ka(x')], (1.18)

IX
PARTICLES AND SOURCES
FIGURE 1.2
a=l a=2 a=3

o_
where
Ka= (K1 +K2)a'

refers to a particular source pair. Insisting again upon the unity of sources,
we require that only the combination
(1.19)
occurs. Since
f (dx)(dx')Ka(x)L1+(x-x')Kp(x')~O for r:t.=Ffi, (1.20)
the appropriate expression is
(O+ IO_ )K f
= exp[ (i/2) (dx)(dx')K(x)L1+(x-x')K(x')]. (1.21)
Conversely, under the condition {1.20) and that the sources are weak, (1.21)
reproduces (1.18), because
J
exp[ (i/2) (dx)(dx')K(x)Lf+(x-x')K(x')]
~ exp[(i/2)}2J (dx)(dx')Ka(x)Lf+(x-x')Ka{x')]
IX
= II exp[(i/2)f (dx)(dx')Ka(x)Lf+(x-x')Ka{x')]
IX
~ 11 [1+(i/2)J(dx)(dx')Ka(x)Lf+(x-x')Ka(x')].
IX
We have produced, under special circumstances, a situation in

which at a given time an arbitrary number of noninteracting particles can
be present. Let's now extend this result to all analogous situations in which
the physical interactions among the particles are not significant but the
particles need not be macroscopically isolated, so that microscopic quantum
8 J. SCHWINGER
FIGURE 1.3
--------------------- O+
o_
interference effects come into play. Thus consider more generally multipar-
ticle exchanges between sources under the noninteracting cau~al condition
(Fig.1.3)
Then
(O+ Io_ )K = (O+ IO_ )K 1 exp(,if (dx)(dx')K1(x)Lf+(x-x')K2(x')]
x (O+ IO_ )K 2 ' (1.22)
0 0
where, as before, keep in mind that x > x ', and therefore
if (dx)(dx')K1(x)Lf+(x-x')K2 (x')
= f dwiK1(-p)iK2 (p)
= l)K1 P *iK2 P (1.23)
p
in which we have defined
KP =.JdwPK(p). (1.24)
We now analyze the vacuum amplitude in terms of multiparticle states
(O+ IO_)K = }2(0+ I {n})K1({n} 1o_)K2. (1.25)

{n}
This is accomplished by writing
exp(}:iK1P*iK2 P)
p
=II exp(iK1P*iK P)
p
2
(1.26)
Thus, we identify
<{n} Io_ > =

K TI (iKp)np
r;;;--i
'\/ nP;
( O+ Io_ )K , {n} = {np}, (1.27)
p
and
(1.28)
p
which clearly describe multiparticle states with nP = 0, 1, 2, · · · particles in

each momentum space cell. Since the particles are not individually dis-
tinguishable, and there is no limit to the number of particles with a specified
property, we are describing identical particles obeying Bose-Einstein (BE)
statistics.
As a check of consistency, let's test the probability normalization condition,
or equivalently, the completeness of the multiparticle states,
}:(0-1{n})K({n}1o_y = 1, (1.29)
{n}
<
(0-1 {n} >K = [ {n} Io_ )K] *.
Inserting ( 1.27) and its complex conjugate, we get
}:(0-1 {n})K({n} IO_)K = 1(0+ jO_)KI 2 IJ exp(IKPI 2 )
w p
= 1(0+ IO_)Kl exp(}:IKPl 2 ).

2
(1.30)
p
But direct evaluation gives
1(0+ IO_)KI 2 =exp[ - f (dx)(dx')K(x) Re (1/i)Lf+(x-x')K(x')]
=exp[ -f dw~(-p)K(p)],
i.e.
(1.31)
p
and the completeness is verified. Notice how the structure of (O+ IO_ )K
enters in two different ways. On the one hand it gives 1(0+ IO_ )K 12 directly,
and on the other it is used to generate ({n}IO_)K / (O+IO_)K. The con-
sistency of these two procedures serves as a severe test of the formulation.
10 J. SCHWINGER
1.4 An Application. Stimulated Emission
The original definition of a creation source refers to the creation

of particles from the vacuum state. In connection with the identification of
BE identical particles, it is of some interest to generalize the meaning of a
creation source to a situation in which particles may be present initially,
as illustrated in Fig. 1.4. In fact, the general formula (1.21) will supply the
answer to this problem. Let the initial particles be created by a creation
FIGURE 1.4
~K
FIGURE 1.5
o_
source K 2 , and the final particles be detected by a detection source K1 • The
well-defined causal arrangement is indicated in Fig. 1.5. The total source,
now denoted by (K), is the sum of its constituents
(K) = K 1 +K2 +K.
Now (1.21) gives
(o+ Io_ )<K> = (O+ Io_ )K' +K2(o+ Io_ )K
x exp{if (dx)(dx')[ K 1(x)L1+(x-x')K(x')
+ K(x)L1+(x-x')K2 (x') ]}
= }2(0+ I {n} )K {n} IO_ )K2(0+ IO_ )K
1
(
{n}
x exp[_L(iK1 P *iKP+ iKP *iK2 p)]. (1.32)

p
If K is a weak source, only terms linear in K are significant, and

(O+ IO_)(K> = ,L(O+ I {n}Y ({n} jO_ )K
1 2
{n}
x [ 1+ _L(iK1 P*iKP + iKP *iK2 p)]. (1.33)

p
But
iKp({n} jO_)K =.JnP+ 1 ( {n} + 1Pj0_ )K, (1.34)
(O+ I {n} )KiKP * = (O+ I {n} + 1Py.Jnp+ 1, (1.35)
which follow from (1.27) and (1.28), so that the terms linear in Kare
III(O+ I {n} + 1P)K .Jnp+ 1iKP({n}10_ y

1 2
{n} P
.J
+ (O+ I {n} )K 1 nP+ 1 iKP *( {n} + 1P IO_ )K 2 J.
On the other hand, we have the general relation
(O+ 10_ )CK>= I(O+ l{n}Y 1 ({n} I {n'}Y( {n'} IO-Y2 •

{n},{n'}
On comparison, we infer the matrix elements
(1.36)
and
( {n} I {n} + 1P)K =.Jnp+ 1 iKP *. (1.37)
( 1.36) can be rewritten as
({n}+ 1pl {n})K =.Jnp+ 1(1PIO_)K (1.38)
which gives the characteristic BE connection of spontaneous and stimulated

emission, the total emission probability being proportional to nP + 1, with
nP the number of particles of momentum p present initially.
12 J. SCHWINGER
1.5 Spin-1 Particles. The Photon
The function Ll+(x-x') is an invariant one which depends only on

(x - x') 2 • Thus the fundamental structure for noninteracting spin-0 particles
(O+ IO_)K = exp[(i/2)f (dx)(dx')K(x)Ll+(x-x')K(x')]

has an invariant significance, if K(x) is transformed as a scalar. This is
characteristic of spin-0 particles. The exponential quadratic form is, in
fact, entirely general, although it is derived within the context of spin-0
particles. It simply describes the creation of particles, propagation of
particles in space and time, and the detection of particles. The extra multi-
plicities that the particle possesses will superimpose on this basic structure.
As a first step to remove the spin restriction, suppose the scalar is replaced
by a vector Jµ(x), and consider the provisional structure
(O+ Io_ y = exp[ (i/2) f (dx)(dx')Jµ(x)Ll+(x-x')Jµ(x')]. (1.39)
However, this by itself will violate physical requirements, for the vacuum
persistence probability implied by (1.39) is
1(0+ 10-/1 2 =exp[ - f (dx)(dx')Jµ(x) Re (1/i)Ll+(x-x')Jµ(x')]
=exp[ - f dwPJµ(-p)Jµ(p)]
= exp[-J dwp{IJ(p)l 2 -IJ 0 (p)l 2 }], (1.40)
which is not necessarily less than or equal to unity, as expressed by the
positiveness requirement
f dwP[IJ(p)l 2
- IJ 0 (p)l 2 ]~0.
Therefore, the time component of lµ, which appears with the wrong sign,
must be suppressed, in an invariant manner. This can be accomplished by
the observation that the time-like momentum p" ( - p 2 = m 2 > 0) supplies a
natural time-like direction. On writing
Jµ(p)* lµ(p) = Jµ(p)*gµvP(p)
we now replace the metric tensor gµv by the combination
PµPv
gµv + - 2 -
m
which, in the rest frame of p(p = ©, p 0 = m), takes the following values
l
µ,v=k,l : <>kl
gµv+ P~v = µ,v=O,k or k,O: 0 (1.41)
µ,v=O,O :0
and
J•(p)*(g.,+ p;;,, )r(p)= IJ(p)l 2 >0. (1.42)
Thus, the correct structure, in coordinate space, is
(O+ 10-)1 = expG f(dx)(dx')J"(x{g.,

- ~2 a.a}1+(x-x')J'(x') J. (1.43)
Insistence upon the physical positiveness property has left us, for each
momentum in its rest frame, with three independent sources, which are
transformed into each other by spatial rotations. We are clearly describing
a spin-1 particle. In an arbitrary coordinate frame we have
(1.44)
Here
(1.45)
(1.46)
where ep;. are the polarization vectors associated with momentum pµ, which
may be taken to be real. As a check of dimensionality, setµ= v in (1.44) and
sum overµ. We get
4-1=3,
which gives the right account of multiplicity. With the definition
Jp;. =.Jdwpeµp;.*Jµ(p), (1.47)
we have
(1.48)
As this expression suggests, the whole story of spin-0, such as identification

of multiparticle states, check of completeness, etc., can be repeated here
with A. added top as quantum numbers. For instance
(1.49)
and
(1.50)
14 J. SCHWINGER
Since this discussion depends heavily on the fact that m=FO, it

1
naturally raises a question concerning m = 0, where gµ 11 - - 2 aµav is no
m
longer meaningful. Let's return to the original expression (1.40). The only
alternative now is to restrict the source in a covariant manner so that I J{p) 12
- IJ 0 {p)l 2;?;:0. In particular, J{p) = 0 must imply J 0 (p) = 0. Since we are
looking for a scalar restriction among the components of Jll, a~d there is
only one vector pµ available, the only possibility is the scalar equation
(1.51)
or, in coordinate space,
(1.52)
We are obviously describing the photon. Its masslessness demands that its
source be a divergenceless or conserved vector. Incidentally, if the deri-
vatives in (1.43) are transferred to the sources, the mass-dependent part has
the structure--;. <\Jµ(x)Lf + (x- x')8J 11(x'), which disappears if the source
m
is restricted to be a conserved vector. And then one has a smooth transition
from m=FO tom= 0.
To see that the condition (1.52) is sufficient as well as necessary for the
correct physical interpretation of
(O+ 10-/ = exp[(i/2)f (dx)(dx')Jµ(x)D+(x-x')Jµ(x')], (1.53)

where
D+(x) = Lf+(x, m2 = 0), (1.54)
we calculate the vacuum persistence probability
(1.55)
by constructing a dyadic representation for gµv· For m=FO, pµ provides a

time-like vector, now pµ(p 2 = 0) is a null vector. Therefore let us define
another null vector
so that pµ+ pµ is a time-like vector, and pµ- pµ is a space-like vector. We

now introduce two other unit space-like vectors, ellp;., which obey
(1.56)
Hence
(p+ iJY'(p+ PY (p- p)µ(p- PY + ~ µ v *
(p+ p)2 + (p- p)2 .tft~ p;.e p;. '
or
~ µ p;.e vp;. * •
+ £..Je (1.57)
l=1,2
Again the dimensionality can be checked by letting µ = v and summing over

µ.One gets
4=2+2.
When the representation for gµv' (1.57), is inserted into (1.55), and the
conditions pJµ(p) = pµJµ(-p) = 0 are used (notice that pµlµ(p)#O), we get
l(O+ IO-YI 2 =exp(- }:llp;.l 2)~1, (1.58)
p).
as required. The necessary reduction to two polarizations which are charac-

teristic of the photon has also been accomplished through the use of the
divergenceless condition on Jµ, (1.52).
16 J. SCHWINGER
1.6 Massless Spin-2 Particles. The Graviton
As a further example to illustrate the universality of the source

description, let us discuss briefly another massless boson known (or con-
jectured) as the graviton. The graviton is a spin-2 particle. Its source is a
second-rank tensor obeying
Tµv(x) = Tvµ(x),
OµPv(x) = 0. (1.59)
The appropriate expression for the vacuum amplitude is

(O+ Io_ y f
= exp{(i/2) (dx)(dx')Pv(x)[gµ~VIC
-(1/2)gµvgA."]D+(x-x')T;."(x')}. (1.60)
We must, of course, verify that this expression satisfies the positiveness
requirement and at the same time that it involves only the two independent
polarizations characteristic of a massless particle.
It follows from (1.60) that
1(0+ IO_ )T 1 2 = exp{ - f dwPPv( -p)[gµlgvrc

-(1/2)gµvg).rc]T)."(p)}, {1.61)
where pv(p) obeys
PµTµv(p) = pµPv(-p) = 0.
If we insert the dyadic representation for gµv' (1.57) (for simplicity, real
e/s are used here), we have effectively, for a fixed p,
gµ).gv"-<112)gµvg)."__. I [eµaevpeA.ae"p-c112)eµaevae).pe",,J.
IX,P=1,2
Keeping in mind that symmetry in µ, v and A., K is required, we get finally

(1.62)
where
(1.63)
is symmetrical in IX and P(also symmetrical inµ and v), and obeys

(1.64)
Therefore, only two of the eµvaP, IX ,p = 1,2 are independent, and if one
defines
(1.65)
and
(1.66)
then
(1.67)
With the definition
TpA =~eµvpATwv(p), (1.68)

(1.61) becomes finally
= exp(-_LITpA..1 )~1,
2
l(O+IO-Yl 2 (1.69)
p,A=1,2
which confirms the physical positiveness requirement and exhibits the

two independent polarizations characteristic of massless particles.
18 J. SCHWINGER
1.7 Spin-(1/2) Particles
In order to discuss spin-( 1/2) particles, we will briefly review the

Dirac algebra in the Majorana representation which is particularly useful
for our purpose. The 4 x 4 matrices y" obey
{y", yv} = _ 2gµv, (1.70)
(y0 ) 2 = + 1, (yk) 2 = - 1, k = 1, 2, 3. (1.71)
These matrices possess simple algebraic properties. But as a set they do

not have a common symmetry, y0 is imaginary and antisymmetrical (and
therefore Hermitian), while Yk (k = 1, 2, 3) is imaginary and symmetrical
(and therefore skew Hermitian). However, out of these matrices one can
construct another set which do have a common symmetry. But their alge-
braic properties are no longer simple. These are y0 y" which are all real and
symmetrical. The fact that one cannot have a set of matrices with both
simple algebraic properties and a common symmetry is due to the inde-
finite nature of the space-time metric. The matrix y0 represents the fundamen-
tal metric tensor. In the same sense that Jµgµvr is a scalar, so is rJY 0 rJ if rJ
is a four-component spinor.
As I have remarked before, the basic exponential quadratic form of (O+ IO_)
is entirely general. Thus we ask: Is it correct to simply write
(O+ IO_)" f 0
= exp[ (i/2) (dx)(dx')rJ(x)y Ll+(x-x')rJ(x') J
in order to describe spin-(1/2) particles? The objection here is that all four
components of rJ~(x) contribute, giving a multiplicity of four rather than the
multiplicity of two appropriate to spin-(1/2). This is overcome by writing
(O+ 10-)" = exp[(i/2)f (dx)(dx')YJ(x)y 0 (m-yµ(1/i)oµ)

x A+(x-x')rJ(x')]. (1.72)
As usual, consider an emission and an absorption source,
Then
if (dx)(dx')rJ 1(x)y°(m-y(1/i)8)L1 + (x-x')rJi(x')
= f dwirJ 1(-p)y 0 (m-yp)rJ 2 (p). (1.73)
For a particular momentum p, this structure can be examined in its rest

frame (p = 0, p 0 = m), and
m-yp~m+y°m = m(1+y 0 ), (1.74)
which selects only y ' = + 1, and rejects y ' = -1. This produces the required
0 0
reduction to two independent sources, characteristic of spin-( 1/2) particles,

since y0 ' = + 1 is a property unaltered by spatial rotations.
We now define the 4 x 4 matrix function
(1.75)
and observe that
(y"(1/i)oµ+m)G+(x-x') = ~(x-x'), (1.76)
in virtue of the differential equation satisfied by Lf+(x-x'). Thus, G+(x-x')

is the Green's function of the Dirac equation. Transposition of the matrices,
combined with the exchange of x and x', has the following effect on
y0 G+(x-x'),
[y 0 G+(x' -x)Y = (y 0 m-y0 y"(1/i)o'µ)TLf+(x' -x)

= -[y 0 G+(x-x')], (1. 77)
since y0 is antisymmetrical and the y0 yµ are symmetrical (8' µ = -8µ). But then
if rJ(x) are ordinary numbers, the quadratic form in the exponential function
of (1.72) is zero identically! There are two possibilities to escape this con-
clusion. The first one is to introduce another matrix in G+(x-x'), which is
outside of space-time and therefore does not affect the spin description,
and is antisymmetrical. The simplest choice is
q = (~ -~} (1.78)
The source functions, 17(x), of course are now given an additional multi-
plicity of two. Keeping the previous definition of G+(x-x'), we then have
f
(O+ Io_)" = exp[ (i/2) (dx)(dx')rJ(x)y°qG+(x-x')rJ(x')]' (1.79)
which describes spin-(1/2) BE particles. But the physical requirement on the

vacuum persistence probability that it should never exceed unity, which is
a positiveness requirement,
f (dx)(dx')17(x)y q Re (1/i)G+(x-x')rJ(x')~O
0
(1.80)
can never be satisfied! The matrix q, being intrinsically indefinite, has

eigenvalues ± 1, and is independent of space-time matrices. Therefore a
matrix transformation on q can change q into -q, and the left-hand side of
(1.80) will change sign, in contradiction with the positiveness requirement.
20 J. SCHWINGER
The second possibility is to match the multiplication property of the sources

11,(x) to the antisymmetry of the structure y0 G+(x-x'), i.e. they must be
anticommuting numbers:
Now we have an identity, instead of a paradox,
=- f (dx)(dx') ,L11,(x)[y G+(x' -x) ]n11,,(x')

0
'·"
= + f (dx)(dx'),L11,,(x')[y G+(x' -x)]n1h(x),
0
'·"
which is simply a relabelling of variables. Just this kind of anticommuting
number is well-known as the elements of a Grassman, or exterior algebra.
The physical properties of the system can now be studied by the
standard procedure. Consider an emission and absorption source
and deduce
(O+ 10-)" = (O+ 10-)"'
J
x exp[i (dx)(dx')17 1(x)y 0 G+(x-x')11 2 (x') ](O+ I0_ )" 2 , (1.81)
since quadratic functions of the anticommuting numbers are commutative.

Now
x 0 >x0 ': G+(x-x') =if dcoPeip(x-x'>(m-yp), (1.82)
where the projection matrix has the property

2
m-yp ) = m-yp ' (1.83)
( 2m 2m
since
(1.84)
The projection matrix can be exhibited in the eigenfunction expansion
(1.85)
where
(1.86)
and A. is a polarization or spin index. The dimensionality can be checked by
taking the trace of (1.85).
(1.87)
in which the relation Tr yµ = 0 has been used. Thus,
x 0 > x 0 ': G+ (x- x') = if dwPeip(x-x'>2m "'fpp;.Up;. *y 0

• (1.88)
..1.=1,2
We now define
~dwP2m up;. *y 0 17(p) = 1'/p;., (1.89)
~ dwP2m17( -p)y0 up;. = 1'/p;. *.

It is taken for granted that we keep the convention of choosing 17(x) to be
real. Later we will learn more about the complex conjugation property
of the anticommuting numbers. As linear combinations of the 17(x), 1'/p;. and
1'/p;. * are totally anticommutative, i.e.,
(1.90)
which implies, in particular,
(1.91)
Hence
f
exp[i (dx)(dx')17 1(x)y 0 G+(x-x')17 2(x')]
= exp(~)1'/1p;. *i1'/2p;.)
p).
= IT (1 + i111p;. *i112p;.). (1.92)

p..1.
That is
"". *.l1'/zp;.) = IT""<.
exp(~.}1'/ip;. .t..J l1'/ip;. *.l1'/ 2 p;. )np;,. , (1.93)
p..1. p). np;..
where each np;., can take only on two values,

np;. = 0, 1.
We recognize the basic feature of Fermi-Dirac (FD) statistics, the exclusion
principle! A FD source cannot operate to create two particles with the
same properties. The vacuum amplitude has the general form
(O+ 10-)11 =(O+1 o_ )"'IT I011ip;. *i112p;.)"p)..(o+ 1o_ )112 • (1.94)
p). np;,.=0,1
22 J. SCHWINGER
Since the sources do have simple algebraic properties, each term of this
structure can be factored into two parts, one on the left referring entirely to
the detection source and the other on the right referring entirely to the
emission source. To illustrate the procedure, consider the example of two
states, a and b,
and in general
(1.95)
where TI is some standard multiplication order and ITT is its inverse. The
factor np;. ! = 1 is not necessary, it is included only to show that the results
are universal-BE particles and FD particles are exactly on the same
footing, only the implicit algebraic property of the sources conveying the
nature of the particles.
It is important to check the completeness property of the multiparticle
states, as in
L(O- ln)11(nl0-)11 = L[ (nl0-)11]*(nl0-)11 =1,

n n
(1.96)
n n
where n symbolizes the whole set of occupation numbers. The sense of

multiplication here is important, since the individual factors are functions
of noncommuting numbers. On inserting the expressions (1.95), these
conditions become
1=l(O+10-)11 l 2 I [IT<11p;.r).]*IT<11p;.)"PA,
n p). p).
1 = l(O+ Io_),, 12 .L ITT<11p;. *tp;.[ITT<11p;. *tp).]*.

n pJ.. p).
(1.97)
These two relations become identical, if complex conjugation reverses the

sense of multiplication, as conveyed by
(1.98)
This is a general source property, since transposition is ineffective for BE
sources. We infer from (1.98) that, while the product of two real BE sources
is real, the product of two real FD sources is imaginary. Now
= I1T(17p;.*)"PA.I1(17p;,)"PA.
p). p).
= II <11p). *Y/p;.tpA.,
p).
and ( 1.97) becomes

2
1=1(0+ Io_)" 1 exp(L11p;. *Y/p;,), (1.99)
p).
which is thus valid for both statistics. Now, direct evaluation gives
1(0+ 10-)"1 2 =exp[ - Ref (dx)(dx')17(x)y 0 (m-y(1/i)o)

x (1/i)Ll+(x-x')17(x')].
From what we have just learned, the combination 17(x)y0 17(x') is real since
y0 is imaginary. Also (1/i)yµ is real. Therefore, everything is real except
(1/i)Ll+(x-x'). But the real part of the latter is
Re (1/i)Ll+(x-x') = Ref dwPeip(x-x').
Under the integration sign, we have
m-y(1/i)o_.m-yp = 2m2:up;.up;,*y 0 •
).
Therefore,
(1.100)
But
(1.101)
is a reality statement, and the completeness requirement is verified.

Spin (1/2) is only compatible with FD statistics. We now turn to the question
of compatibility between integer spin and FD statistics. The actual integer
spin value is irrelevant in this consideration, and the essence of the whole
argument can be seen from the simplest case, spin 0. First we notice that we
cannot let the source functions, K(x), be anticommuting numbers, without
introducing an antisymmetrical matrix q to compensate the fundamental
symmetry of the function Ll+(x-x') in x and x'. This matrix q must be a
non-space-time one, in order not to affect the already established spin
24 J. SCHWINGER
description. We give K(x) a multiplicity of two, and the matrix q is that

introduced previously. Thus, spin-0 FD particles are described by
(O+ Io_ )K f
= exp[ (i/2) (dx)(dx')K(x)qLl+(x-x')K(x')]. (1.102)
Write
q = Lq'uq,uq,*, (1.103)
q'=±1
and define
Uq'*,Jdw;K(p) = Kpq''
.JdwPK(-p)uq' = Kpq'*, (1.104)
for the consideration of an emission and absorption source. Accordingly

if (dx)(dx')K 1(x)qLl+(x-x')K2(x')
= LiK1pq'*q'iK2pq'•
pq'
But
. [ ( 1 - q') ] ] 2
q'= [ e' - 4 - n: , (1.105)
only introduces a q' -dependent phase factor in the identification of multi-
particle states. These phase factors will be suppressed when we consider
the completeness requirement (1.29), since each matrix element is combined
with its complex conjugate. So it is required that
1=1(0+ IO-YI 2 exp(LKpq'*Kpq')· (1.106)

pq'
But since K(x)qK(x') is real, which follows from (1.98), we have directly
1(0+ 10-YI 2 =exp[ - f (dx)(dx')K(x)q Re (1/i)Lf+(x-x')K(x')]
=exp(- 2)'Kpq'*Kpq'), (1.107)
pq'
in which q' appears explicitly. The completeness requirement is violated

through the indefiniteness of q.
1.8 Multispinor Formulation of Arbitrary Spin.
General Connection of Spin and Statistics
We have seen in the specific examples of spin-0 and spin-(1/2) that

there is a definite connection between spin and statistics. We will now show
that the connection of spin and statistics can be established for arbitrary
spin. Indeed, there is a uniform treatment of all spins, which gives the
spin-statistics connection in a general way. It uses multispinors s, 1 ••• ,n(x),
each ( = 1, 2, 3, 4 being a Dirac index. Using the spin-(1/2) structure
repeated n times, one gets
n n
(O+ jO_)s = exp[(i/2)f(dx)(dx')S(x) IT Yrr. IT 0
a=1 a=1
x (m-yµ( 1/i)oµ)rr.Ll+(x-x')S(x')], (1.108)

where
[r/, r/] = o, ifa=l=P, (1.109)
since they act on different indices. The consideration of an emission and

absorption source
gives
f(dx)(dx')S (x) TI [r (m-yµ(1/i)8µ) ]aL1+(x-x')Si(x')
1
a
0
= f dwiS 1(p)* TI [r (m-yp)JiS (p).

a
0
2 (1.110)
For a specified momentum, in its rest frame we have
II [r (m-yp)]a =II [r (m+y m)]rr. =IT [m(l +y )]a,

0 0 0 0
(1.111)
which is the projection matrix to select y0 ' = + 1 for each index a. Thus, in
the rest frame, the sources reduce effectively to Su 1u2 ••• a"' each a being a
two-valued spin index. Now suppose s, 1, 2 ••• ," has definite symmetry property.
Specifically, let us consider the totally symmetrical one. Then Sa 1 u 2 ••• un is
totally symmetrical. As we all have learned from elementary quantum
mechanics, this produces the state of maximum resultant spin, that is, the
sources are those of a particle of spin
s = (1/2)n, (1.112)
and
s = (1/2), 1, (3/2), ... '
as
n = 1, 2, 3, ....
26 J. SCHWINGER
Only s = 0 is missing in this series. For this one has to take n = 2 and use an
antisymmetrical spinor source s,,, 2
(x). Incidentally, for n~3, the indepen-
dent components of a totally antisymmetrical multispinor vanishes identi-
cally in the rest frame, since S 111112 ... 11" = O(n~3). Note that more complicated
symmetry patterns give equivalent descriptions, to the extent that a definite
spin appears in the rest frame. Consider, for example, n = 3 with the require-
ment of antisymmetry in a pair of Dirac indices. The latter contributes zero
spin in the rest frame and we have a possible description of an s = ( 1/2)
particle.
The symmetry property of the matrix propagation function is expressed by
{fI [y 0 (m-yµ(1/i)8'µ)]aL1+(x' -x)Y

a
= fI(-)[y 0 (m-yµ(1/i)8µ)]aL1-r-(x-x'),
lX
or
{IJ [y 0 (m-yµ(1/i)8'µ)]aL1+(x' -x)Y
a
(1.113)
If the algebraic properties of the sources are to match the symmetry pro-
perties of the propagation function, we learn that
J
n =even, s =integer: [ S(x), S(x') = 0, BE
n =odd, s = integer+(1/2): {S(x), S(x')} = 0, FD (1.114)
which express the general connection between spin and statistics.

In the same way as for the case spin-(1/2), one can verify that the complete-
ness requirement
I(0-1 {n})S({n} IO_)s = 1, (1.115)

{n}
is satisfied, which is shown by alternatively using (O+ 10-l to calculate

directly 1(0+ IO_ )s 12 and indirectly by producing first ( {n} Io_ )s / (O+ Io_ )s.
The consistency test of the two procedures is important for attempts to
reverse the natural spin-statistics connection by introducing a non-space-
time antisymmetrical matrix q = G-~) . The completeness condition can
be rewritten as
1= I
{n)
(0-l{n}/
(0- IO+)s
(1.116)
As in the discussion of the previous section for spin 0, eigenvalues q'
appear as phase factors in the identification of multiparticle states, since
q' = 1, or q' = - 1 = (i) 2 • These phase factors disappear on forming
(0-l{n})s. ({n}IO_)s =[({n}IO-/]*

(0-IO+)s (O+IO_)s (O+IO_)s
({n}IO_)s]
(1.117)
[
x (O+ 10-/ .
The direct evaluation of 1(0+ IO_)sl 2 gives
1(0+ IO_ )sl 2 f
= exp{- (dx)(dx')S(x)q II [y0 (m-y(1/i)o) ]a
IX
x Re (1/i)Lf+(x-x')S(x')}, (1.118)
since S(x)q II y0 aS(x') is real with abnormal statistics, while S(x) II y0 aS(x')
IX IX
is real with normal statistics. In both cases, the reality property follows
simply from the Hermitian character of the matrices. Now the indefinite
eigenvalues of q are explicitly involved in (1.118), and the completeness
requirement is violated.
2. ELECTRODYNAMICS
2.1 Combined System of Noninteracting Electrons

and Photons. Fields
We shall now study the most familiar dynamical system, of

electrons and photons, in order to indicate how the interaction among
particles can be incorporated very naturally into the source theory.
Properties of noninteracting particles are described by individual vacuum
amplitudes. They are multiplied together to describe the combined system.
Write
(O+ Io_ )s = eiw(S)' (2.1)
where, for the system of electrons and photons under physical conditions of
noninteraction,
w(S) = w2 = (1/2)J(dx)(dx')17(x)y 0 G+(x-x')17(x')
+ (1/2) J(dx)(dx')Jµ(x)D+(x-x')Jµ(x'). (2.2)
It is convenient to introduce auxiliary quantities which express the effects at
one source of other sources, and combine the latter with the particular pro-
pagation function. Thus, we define
f
IJl(X) = (dx')G+(x-x')17(x'), (~(1/i)8µ +m)IJl(X) = 17(x),
Aµ(x) = f (dx')D+(x-x')Jµ(x'),-8 Aµ(x)
2
=Jµ(x), BµAµ(x) = 0. (2.3)

29
30 J. SCHWINGER
These are the fields of the corresponding sources. They are numerical
quantities of the same type as the sources, namely, Aµ(x) are totally com-
mutative numbers, and IJl(X) are totally anticommutative. The auxiliary
quantities, fields, are very useful, since they summarize the effects, in the
region of interest, of sources which may be very far away. Now, w2 can be
represented alternatively as
w2 = (1/2) f (dx)17(x)y°1J1(x) + (1/2) f (dx)Jµ(x)Aµ(x)

= (1/2) f (dx)IJl(x)y 0 17(x) + (1/2) f (dx)Jµ(x)Aµ(x), (2.4)
or
w2 = (1/2) f (dx)IJl(x)y 0 (y"(1/i)8µ + m)IJl(x)
J
+(1/2) (dx)Aµ(x)(-8 2 )Aµ(x). (2.5)
In the latter form only fields appear. A third form combines the two repre-
sentations,
w2 = f (dx)[ 1J1(x)y017(x)-(1/2)1J1(x)y0 (y(1/i)8 + m)IJl(X)]
+ f (dx)[ Aµ(x)Jµ(x)-(1/2)Aµ(x)(-8 2 )Aµ(x)]. (2.6)
All the three forms are equivalent in virtue of the relations between fields
and sources, (2.3). But the last form has the advantage that these connections
can be derived from it instead of being stated independently. Let us examine
the dependence of w2 on the sources 17, Jµ and the fields IJI, Aµ,
~w 2 =f(dx){1J1(x)y 0 J17(x)+~1J1(x)y 0 [17(x)-(y(l/i)8+m)1J1(x)]

+ All(x)Mµ(x) + JAµ(x)[ Jµ(x)-( - 82 )Aµ(x) ]} .
Now, if we insist that w2 depends only on the sources, that is, that it be
stationary with respect to variations of the fields, the field differential
equations are recovered. In this form w2 is an action, and we write
(2.7)
where !"£ is the Lagrange function of the system,
(2.8)
Since aµAµ(x) = 0, the second term can also be replaced by
-(1/2)[ Aµ(-8 2 )Aµ-(8µAµ) 2 ]

---> -(1/2)[ (8µA 11 ) 2 -(8µAµ) 2 ]
---> -(1/4)P 11Fµ 11 , (2.9)
as far as w2 is concerned, where
Fµv = oµAv-BvAµ. (2.10)
The Lagrange function thus acquires the new form
ff = -( 1/2)1Jfy0 (y(1/i)o + m)IJI - ( 1/4)PvFµv. (2.11)
Since the source lµ is conserved

oµJµ(x) = 0,
we now have the freedom of gauge transformation
Aµ(x)-->Aµ(x)+ 8µ).(x), A.(x) arbitrary,

which leaves the action w2 unchanged. Gauge invariance is an important
aspect of the masslessness of the photon.
32 J. SCHWINGER
2.2 Primitive Interaction and Interaction Skeleton.

Extended Sources.
So far sources are only called upon to liberate energy-momentum,

or mass, equal to that of the particle being described. This is too restrictive
for a dynamical theory. The notion of source is an abstraction of realistic
physical production and annihilation mechanisms. The simplest example is
provided by the creation of a charged particle in a collision, which involves
the transfer of charge to the particle from the source (idealizing all the
partners in the collision). An accelerated charge radiates, and the possible
creation of a photon along with the particle is an inescapable part of the
physical mechanism. Thus it would be artificial to separate the mechanism
that creates a charged particle alone from the mechanism that creates a
charged particle and a photon. Accordingly we regard it meaningful and
useful to define a charged particle source 17(p), not only for - p 2 = 111~ , but
also for - p 2 > m 2 • Thus, we now have two possible emission processes for a
charged particle source, as shown in Fig. 2.1, where the emission of an
FIGURE 2.1
£q J£:
-p2>m2
-p2=m2
electron and a photon is supposed to occur without interaction between

these particles. We shall call the latter the primitive interaction. A similar
discussion also applies to an absorption source. These considerations suggest
more generally that it is meaningful and useful to generalize a source to
include any combination of particles with the same properties, apart from
mass, as the specified particle. This is the general concept of an extended
source.
The primitive interaction, describing the dynamical process in which an
electron source 17(p), with - p 2 > m 2 , radiates an electron and a photon,
gives new possibilities for simple particle sources. For example, the sub-
sequent removal of the photon, or of the electron, gives a particular reali-
zation of an electron source, or of a photon source, as illustrated in Fig. 2.2.
Single particle exchanges between these sources give new phenomena such
as the ones shown in Fig. 2.3. These couplings involve four sources, 1717JJ,
or 11111111, which can be written in space-time forms in a way that permits
immediate generalization to a wider class of spatio-temporal arrangements.
In particular, the final forms can be used to give a first, skeleton description
of electron-photon scattering or electron (positron)-electron scattering
FIGURE 2.2
e
J I/
Effective electron source Effective photon source
FIGURE 2.3
,,
J
J
I/
FIGURE 2.4
J ,.,
I/
e - y scattering e - e scattering
(Fig. 2.4). This process can be continued. For example, from electron-
photon scattering, one gets an effective electron source of the form 11JJ.
Single electron exchange with an effective electron source of the form 17J
gives a new process which involves 1717JJJ, and so on. Thus, one gets a series
34 J, SCHWINGER
of dynamical processes involving 1717J, 17ytJJ, yt17JJJ, ... etc. This proliferation
introduces, at a first stage, an infinite variety of interactions, in skeletal
form, involving only single particle exchanges.
One can give an analytic expression to describe all these processes by
beginning with the primitive interaction term w3 which must have the form
f
w3 = (1/2) (dx)(dx')(d~)yt(x)y°G(xx', ~µyt(x')Aµ(~), (2.12)
from the consideration of an effective photon source. But we must also be

able to use the effective electron source point of view. Therefore each rt
must be associated with a G+ that describes the propagation of the electron
to the point of detection. Thus, w3 must have the form
W3"' f (dx)lfl(X)lfl(X)A(x)
omitting the finer details. Now, independently, suppose we do have an
additional term in w, say w', that depends on Aµ. Let us change J;, and
therefore Aµ, by an infinitesimal function. Thus
w = w2 +w'(A), (2.13)
and
c>w'(A)
c>w'(A) = f (d~)(d~')c>Jµ(~)D+(~ -0 c>Aµ(O , (2.14)
which can be interpreted as single photon exchange between c>P(~) and

c>w'(A)
an effective photon source . The latter, as a photon source, must
c>Aµ(O
be conserved,
o c>w'(A) = 0, (2.15)
µ c>Aµ(x)
or
c>w'
f (dx).A.(x)oµ c>Aµ(x) = o, (2.16)
where .A(x) is an arbitrary function. Therefore
(2.17)
which asserts that w'(A) must remain unchanged when Aµ(x) is displaced by
oµ.A.(x). In other words, w'(A) must be gauge invariant.
An elementary solution to both requirements is perfectly familiar. It is
obtained from w2 by the gauge-covariant substitution
aµ~aµ-ieqAµ
and has the structure

(2.18)
Here q = (~-~}s the charge matrix which must be introduced, since

lf/YoYµlfl =0. The symmetry of y°yµ is in conflict with the antisymmetry of
lfl~lfl,,.
Thus, in order to represent the charge of the electron, it is necessary
to introduce an additional multiplicity of two, and 1f1 is now an eight-
component object. The coupling constant e is the observed charge of the
electron, and may be identified through soft photon emission. The action w
is invariant under the gauge transformation
Aµ--+-Aµ + oµA., (2.19)
lfl--+-eieqllfl. (2.20)
It is necessary that 17(x) responds like lfl(x) to a gauge transformation. This

expresses the fact that 17(x) symbolizes the charged particles involved in
creating the electron, which also emit and absorb photons.
The logical progression in this development should be noted. We start
with the photon. Its masslessness demands that the source be a conserved
vector, from which follows the requirement of gauge invariance, which is
not an independent hypothesis. In quantum field theory the procedure is
exactly reversed. One begins with the requirement of gauge invariance and
deduces the existence of the photon. But this also involves a dynamical
restriction. If the coupling is sufficiently strong, there is the possibility that
a massless particle does not exist.
The action principle gives the field equation for 1f1,
[y((1/i)o-eqA)+m ]lfl = 11, (2.21)
or
(y(1/i)o+m)lfl = 17+yeqAlfl, (2.22)
which can be converted into the integral equation,

(2.23)
in which space-time coordinates are regarded as matrix indices. We shall use

the notation 1f1 A to distinguish the electron field influenced by the vector
potential A from the noninteracting field If/,
If/ A = If/+ G+ yeqA 1f1 A. (2.24)
The iterated solution is

lflA =If!+ G+eqyAlfl+ G+eqyAG+eqyAlfl+ ... , (2.25)
36 J. SCHWINGER
and
J
w = (dx)[ JµAµ-(1/4)PvFµv+(1/2)17y 01f1 A]
= f(dx)[JµAµ-(1/4)PvFµv]
0 0
+(1/2) f (dx)11Y 1/1+(1/2) f lf/Y eqyAlfl
+(1/2) f lflJJ°eqyAG+eqyAlfl+ .... (2.26)
The action principle also supplies the field equation for A (in the Lorentz
gauge, oµAµ = 0),
-o 2 Aµ =Jµ+(1/2)1f1y°eqyµlfl+.... (2.27)
Again, we use a special notation, At/t, for the vector potential under the
influence of 1f1 to distinguish it from the noninteracting field Aµ. Then
,.,
(2.28)
and
w = f (dx)[ (1/2)JµAµ+(1/2)17y 0 1f1] +(1/2) f (dx)lfly 0 eqyAlfl

+ (1/2) f (dx)(dx')(1/2)(1f1y°eq~lfl)(x)D+(x-x')(1/2)
x (lfly 0 eqyµl/l)(x')
f
+(1/2) (dx)(dx')lfl(x)y 0 eqyA(x)G+(x-x')eqyA(x')lfl(x')
+ ... , (2.29)
in which the interactions are now made explicit. The first term describes the
noninteracting system, the second term is the primitive interaction, the
third term describes e- -e- ore--e+ scattering, and the fourth term des-
cribes electron-photon scattering or pair annihilation. The infinite series in
(2.29) is a sequence of increasingly elaborate interaction skeletons. Later
terms in this series do not contain modifications of earlier ones.
It should be emphasized that the iterated solution is a classification of
processes in terms of increasing degree of complexity. It is not a perturbation
expansion. The physical electron mass m, and the physical electron charge e,
which are identified originally under specific physical circumstances, will
never change their significance when the class of phenomena under exami-
nation is enlarged.
2.3 Calculation of Matrix Elements
To show how to use the structure (2.29) to calculate a particular

matrix element, let us consider Compton scattering as an example. Write
(2.30)
where 1J1 2 and A 2 are the fields of weak emission sources, and I/Ii and A1
are the fields of weak detection sources. The experimental arrangement is
FIGURE 2.5
,, 1
112
shown in Fig. 2.5. When the substitutions (2.30) are made, in (2.29), Comp-
ton scattering is described by terms involving l/lil// 2 A1 A2 , which are
f
w = (dx)(dx')[ 1J1 1(x)y 0 eqyA 1 (x)G+(x-x')eqyAi(x')'/1 2(x')
+ 1/1 i (x)y 0 eqyAi(x)G+(x-x')eqyA 1 (x')1J1 2(x')]. (2.31)
In these two terms, the roles of initial and final photon are reversed. Therefore
w is invariant under the interchange A 1 ~A 2 • This symmetry, so called
crossing symmetry, is automatically incorporated in the source theory, and
is simply a consequence of particle statistics.
Now, the interaction region is far away from the sources, and since the
interaction certainly occurs later in time than the emissions and earlier than
the detections, we have, in the interaction region,
1/12(x) = L .JdwP2m ueipxif/ puq'
paq
2
puq
Aiµ(x) = L il1u *.Jdwk eµu *e-ikx. (2.32)
kA.
The relevant part of the vacuum amplitude is

( o+ Io_) w ~ iw, (2.33)
38 J. SCHWINGER
where w is the expression given by (2.31). On the other hand,

(O+ Io_ )"1 = L (O+ I{n} )" 1li( {n} I{n'} >< {n'} Io_ ) 11212 '
{n},{n'}
in which there is a contribution of the form
L(O+ I 1(paq)11(U))"1li( 1(paq)11(kA)111(paq)21(H)2)

X( 1(paq) z 1(u)2 IO+ )"2h. (2.34)
But we know that
(2.35)
Substituting the expressions (2.32) into (2.33), we identify, with the. aid of
(2.34) and (2.35),
( 1(paq)i 1(U)t I 1(paq)21(H)z)

= ie 2) dwP 1dwp 2(2m) 2dwk 1dwk 2 (2n)4<5(p 1 +k 1 -p2-k2)
(2.36)
Standard procedures can now be followed to calculate the differential cross
section, etc.
2.4 Two-Particle Exchange
The first stage of dynamical description introduces an interaction

skeleton, which brings various processes into existence, but is not yet a
realistic theory. We reach the second stage which is a realistic, and for
many purposes accurate description by taking into account two-particle
exchanges.
The introduction of the primitive interaction implies that extended 1'/
sources can interact by exchanging an electron, or by the exchange of an
electron and a photon (Fig. 2.6). The propagation function G+(x-x')
FIGURE 2.6
+
,,
-p2 =m2 -p2>m2
will be modified to account for the fact that mass values other than -p 2 = m 2
can now be exchanged. The new propagation function G+(x-x') must
have the form
G+ = G++ ... ,
where the additional terms refer to higher mass exchanges. If the sources
cannot supply sufficient mass, then G+ effectively reduces to G+, as it
should, since this is the situation under which G+ was originally introduced.
The primitive interaction also contains the possibility that a photon source
with -k 2 >(2m) 2 creates an electron-positron pair. Two such sources interact
by exchanging a photon, or an electron-positron pair (Fig. 2. 7), and the
FIGURE 2.7
J J
J J
-k2~(2m)2
40 J. SCHWINGER
propagation function D+(x-x') will be modified. The new propagation

function D+(x-x'), as in the case of electron sources, must have the form
D+=D++ ... ,
where the additional terms refer to higher mass exchanges. These sub-
stitutions (G+-+G+ and D+-+D+) must occur in all processes that can be
analyzed into single particle exchanges. However, this does not exhaust the
implications of two-particle exchanges. Idealized sources only partially
describe realistic processes and additional effects that are characteristic of the
specific interaction will appear. The simplest illustration is supplied by the
primitive interaction itself. The latter describes an extended J as the source
of an electron-positron pair, emitted under circumstances in which the
particles have no opportunity to interact. In the next stage we enlarge the
physical circumstances to permit such interactions to occur. The eiectron-
positron interaction is described by the third term of (2.29) which can be
symbolized as (lf/lfl)D+(lfllfl). Now if we write
If/ =lf/1 +lf/2,
there are two terms which describe the e + e - scattering, (1f1 1 lfl 2 )D+ ( 1f1 1 lfl 2 )
and (lf/ 11f/ 1)D+(lf/ 21f/2 ). These are represented in Fig.2.8. The first term is
FIGURE 2.8
ordinary Coulomb scattering and the second term describes pair annihi-
lation and subsequent pair recreation. If now the electron and positron
created by an extended photon source are permitted to interact, one recog-
nizes that the second mechanism just described is the one that produces the
modification D+ -+D+. The dynamical details are suggested in Fig. 2.9(a).
The first mechanism, the ordinary Coulomb scattering [Fig. 2.9(b)] leads
to a new phenomenon. The quantitative theory shows that the effect is
represented by the alteration of the electromagnetic properties of the
electron, namely, the introduction of an electric form factor, and an addi-
tional magnetic moment, with its form factor.
In order to get a complete picture of the second stage of description, one
must also consider multiple two-particle exchanges. The pair creation from
an extended photon source is viewed as the conversion of a virtual photon
FIGURE 2.9
(a) (b)
FIGURE 2.10
into an electron-positron pair. Now the process of recombination leads

again to a virtual photon. And the whole process can be repeated inde-
finitely (Fig. 2.10). When we consider electron-positron scattering, the
final particles may again interact, and the whole process can be indefinitely
repeated. The new physical possibility the latter implies is bound electron-
posi tron states, the positronium particles. Thus physics becomes more
realistic, although it need not be completely accurate, at this level of des-
cription.
J. SCHWINGER
2.5 Modified Photon Propagation Function
I would like now to give a general discussion of the modified

photon propagation function D+ • The coupling between two photon
sources is produced by the exchange of a photon, or any multiparticle
exchange, with - k 2 = M 2 > 0. Let's write the complete contribution of the
latter as
if (dx)(dx')J1 µ(x)a(x-x')J2 µ(x'), (2.37)
f
which is to be added to (dx)(dx')J1 µ(x)D+(x-x')Jl'(x').
A more complicated tensor structure cannot appear, in virtue of the general
property of a photon source,
-.
Relativistic invariance requires that a(x-x') be a scalar function, and we
can write
a(x-x') =
f (dk)
- - 3e 1
(2n)
"k(
x-x
')
a(-k )
2
'
(2.38)
where k 0 is the energy liberated by the source. Now, (-k 2 = M 2 )
(dk) = (dk) . d(ko)2 = dw dM2 (2.39)

(2n) 3 (2n) 3 2k0 k '
and (2.37) becomes

f dM 2 a(M 2 ) f (dx)(dx')J1 µ(x)if dwkeik(x-x'> J2 µ(x'). (2.40)
One recognizes
if dwkeik(x-x') = A+(x-x', M 2)
for the special circumstances x 0 > x 0 ' under consideration. Thus, inclusion
of multiparticle exchange gives the modified photon propagation function
D+(x-x') = D+(x-x')+ f dM 2 a(M 2 )L1 +(x-x', M 2
). (2.41)
Alternatively, it has the general momentum-space representation
D (x-x') =
+
f (dk) eik(x-x')"jj (k)
(2n) 4 + '
(2.42)
with
- 1
D+(k)= - 2- . +
f dM i a(M2)
.• (2.43)
k -ze k 2 +M 2 -ze
There are general physical restrictions on the function a(M 2 ) from vacuum
persistence probabiJity considerations. For a weak source, we have
f
(O+ Io_ y = 1+ (i/2) (dx)(dx')P(x)D+(x-x')Jµ(x')
= 1 +(i/2)J (dk) P(k)*D+(k)Jµ(k), (2.44)

(2n) 4
and
(2.45)
If the source has only space-like momenta, k 2 > 0, no particle creation is

--possible and the vacuum state must persist, that is, I( 0 + IO_ )11 2 = 1.
Therefore
z -
k >0: ImD+(k)= dM
f 2 Im a(M2)
=0. (2.46)
k 2 +M 2
A simple theorem on Fourier transforms (for this purpose, it is convenient
to introduce the new variables k 2 =ex, M 2 = eY, where x and y range from
- oo to + oo) shows that
Im a(M 2 ) = 0. (2.47)
Next, consider the situation in which the source contains only time-like
momenta. Since
(2.48)
and
I (dk)
-m5(k
4
(2n)
+M )=
2 2 I( dk) 1
- · -3 = f0 d w k
(2n) 2k '
(2.49)
we get
1(0+ 10_)11 2 = f
1- dwkJµ(k)*.fµ(k)/k2=0
- f dM 2
f
a(M 2 ) dwJµ(k)* Jµ(k)f-k2 =M2. (2.50)
In the rest frame of kµ, M#-0, we have J 0(k)=0 and Jµ(k)*Jµ(k)
= IJ(k) I 2 > 0. Therefore it is necessary that
(2.51)
that is, a(M 2 ) must be real and non-negative.

We now consider a simple mechanism of single pair exchange (Fig. 2. 7).
The pair creation by an extended photon source is described by the primitive
interaction. From its contribution to the vacuum amplitude
(2.52)
44 J. SCHWINGER
it is seen that Aµ appears as an effective electron-positron source. Comparing

with the description of the noninteracting propagation of two particles
if f
exp[ (dx)1f1(x)·J'°17(x) J~ -(1/2) (dx)(dx')lfl(x)y 0 17(x)
x 17(x')y0 lfl(x'), (2.53)
one identifies
(2.54)
where the left-hand side is regarded as a matrix. The consideration of two
such effective electron-positron sources give the following contribution to
the vacuum amplitude:
-(1/2) f (dx)(dx')A 1 µ(x)A 2"(x') Tr[ eqyµG+(x-x')eqyv
x G+(x' -x)], I"• (2.55)
when the substitution (2.54) is made for 17 1 (x)17 1 (x') and 17i(x)17 2 (x') in the
expression
exp[iJ (dx)(dx')17 1(x)y 0 G+(x-x')17i(x') J

~ -(1/2)[f (dx)(dx')17 1(x)y 0 G+(x-x')17 2(x')]2.
Under the well-defined causal circumstances (x 0 > x 0 '), the two electron
propagation functions have the form
G+(x-x') =if dwPeip(x-x'>(m-yp), (2.56)
G+(x' -x) =if dwP,eip'(x-x'>(m+yp'). (2.57)
Now (2.55) becomes
J,1 µ(- k)J2 "(k)
-(1/2)fdw dw ,
p p (k2)2
x Tr [ eqyµ(m-yp)eqyv( -m-yp') J, (2.58)

where
k=p+p',
is the total momentum liberated by the source. We transfer our attention

to the total momentum exchange, k, by introducing the unit factor
(-k2=M2)
f
(2n) 3 dM 2dwk<5(p+ p' -k). (2.59)
Thus, (2.58) reduces to
f
-(e 2/2) dM dwkJ/(-k)Jz"(k)(1/M 2 ) Iµv(k),
2 2
(2.60)
where
Iµv(k) = (2n) 3 f dwPdwp·~(p + p' -k)
x Tr [qyµ(m-yp)qy"(-m-yp')]. (2.61)
A simple evaluation in the rest frame of k gives the result
(2.62)
in which the kµk" term will not contribute, since the current Jµ is conserved.
With the definition
a(M 2 ) =
3n M
1J1- (2m)
ex - · -
2
-
M
2
( 1+
M2 '
2
2m-) M>2m
'
(2.63)
e1
where ex = - is the physical fine-structure constant, we get the contribution
4n
to the vacuum amplitude
- f dM 2 a(M 2 )dwkJ1 µ(-k)J2 µ(k)
=if dM 2 a(M 2 ) f (dx)(dx')J1 µ(x)if dwkeik(x-x')J2µ(x'). (2.64)
Again we recognize that L1+(x-x', M 2 ) is the appropriate generalization
for the idwk integral. The modified photon propagation function appears as
- 1
D+(k)= - 2- - . +
f dM z a(M2)
• ,
k -ze k 2 +M 2 -ze
which is a particular realization of the general form (2.43). The function
a(M 2 ) given by (2.63) is real and non-negative. For M»2m, the integral
behaves like
dM 2 1
f Mz kz+M2
and there is no question about its existence.

46 1. SCHWINGER
2.6 Electromagnetic Form Factors
I would like now to discuss the modifications of the electromagnetic

properties of the electron, following the same pattern of discussion as in the
simpler case of the modified photon propagation function. Firstly we
consider the general structure that one should expect on the basis of kine-
matics and causality, and secondly we will calculate a specific mechanism
within that frame work.
FIGURE 2.11
"'•
We are interested in the interaction modification of the pair creation

by an extended photon source. These processes can be classified into two
types. Those in which all modifications of D+ into D+ take place and then
the emission of tht; final pair occurs in one act, and all others which cannot
be so analyzed. The separation of these two types of processes are represented
pictorially in Fig. 2.11. Correspondingly, we write the vacuum amplitude
as two terms, for weak sources,
if (dx)(1/2)1/f(x)·;°y'1eql/f(x)Aµ(x)
- f (dx)(dx')(dfj(1/2)1/f(X)y eqKµ(xx', ~)l/f(X')Aµ(fj,
0
(2.65)
where A is the vector potential constructed from the photon source J with
the aid of the modified propagation function D+. The momentum space
form of the second term is
-
1
-
2
I--
4
(dk) dwPdwp.17(-p)y°(m-yp)eqKµ(p, p', k)
(2n)
x (-m-yp')11(-p')Aµ(k). (2.66)
Translational invariance requires the following structure for Kµ:

1
- Kµ(p, p', k) = ~(p+p' -k)(2n) 4 Kµ(p-p', k). (2.67)
2n
Consider the general structure of Kµ(p- p', k ). Now
-k2=M2,
k(p-p') = p2-p'2 = O, (-p2 = -p'2 =m2)
(p-p')2+k2 = -4m2'
therefore, M 2 is the only scalar that can be constructed from the momentum
vectors p, p' and k. The possible vectors are JI', kµ and (p- p')µ. Suppose
we use Lorentz gauge, so that
Furthermore,
(p-p')µ = - (1/2){yµ' yp-yp'}.
In virtue of the projection matrices (m-yp) and (-m-yp') in (2.66), when
yp appears as a left factor (denoted as Y]3), it can be replaced by -m, and
similarly a right factor of yp' (denoted as yp') can be replaced by + m. Then
yp = yk - yp' ~ yk - m,
"YP' = yk - yP ~ yk + m.
Therefore, the only possible structure for Kµ(p-p', k) is
or
(2.68)
Since there is no individual reference top and p', (2.66) can be put back in
coordinate space in the form
f
i(1/2) (dx)l/f(X)y 0 eqyl'l/f(X)eikxdM 2K 1(M)idwkAµ(k)
+ i(1/2) f (dX)l/f(X)y 0 eq(1/2)aµvl/f(X)eikxdM 2

x K 2(M)idwkFµv(k), (2.69)
where
f
Aµ(k) = (dx')e-ikx'Aµ(x'), (2.70)
and
Fµv(k) = i[kµAv(k)-k"Aµ(k)] (2.71)
is the field strength tensor.

48 J. SCHWINGER
f 0
We are discussing an addition to (dx)(1/2)1/fy eqyµl/fAµ which contains the
primitive interaction and a description of how the gauge variance of A (or A)
is tied to l/f. Therefore, one requires all additional effects to be gauge
invariant. The second term in (2.69) is gauge invariant by itself, and the
first term can also be so written [kµAµ(k) =OJ,
(2.72)
The causal situation under consideration is clear cut. The source produces a
virtual photon which converts into an electron-positron pair close to the
source. Then various interactions take place at a later time. Thus > x 0 ', ;o
and again we can introduce Lf+(x-x', M 2 ) as the appropriate space-time
iJ
generalization of dwkeik(x-x').
The complete structure of (2.65) is then
if (dx)(dx')(1/2)1/f(x)y eqyµl/f(x)F (x-x')Aµ(x')

0
1
+if (dx)(dx')(1/2)1/f(X)y !!!...(1/2)aµvl/f(X)µ'

2m
0
(2.73)
where, using Fourier transforms,
F1 (k) = 1-k2f dM </J1(M)

• , F1 (0) = 1, (2.74)
k 2 +M 2 -ze
</J2(M)
Fi(k) = f dM k 2
+M 2 -1e
• ,Fi(O) = 1,
and
Here F1(k) and F2 (k) are, respectively, the electric and magnetic form factor.
And µ' in (2. 73) is the anomalous magnetic moment.
Let us now carry out a calculation of a specific mechanism. Consider the
effect of the Coulomb interaction of the electron-positron pair emitted by
an extended photon source. As we have discussed [see (2.54) in particular],
the primitive interaction describes an extended photon source as an effective
electron-positron source. Inserting this effective two-particle source into
the electron-positron scattering description, one gets a contribution to the
vacuum amplitude given by
f
(1/2) (dx)(dx')(d()ip(x)JJ°eqyµG+(x-1;)
x eqyA(()G+(l;-x')eqyµip(x')D +(x-x'). (2.75)
FIGURE 2.12
The situation is described in Fig. 2.12. The causal arrangement is such that
x 0 > <; 0 , and x 0 ' > <; 0 • The evaluation is performed without difficulty in the
center-of-mass frame of the two-particle system. Only one special point
should be mentioned. It is the integration over all scattering angles. The
long-range Coulomb interaction gives infinity in the forward direction. This
difficulty can be removed by using a screened Coulomb potential, i.e.
(1/r)e-µr, which is equivalent to the unphysical use of a small photon massµ.
It is clear that this is simply a version of the infra-red problem. This problem
is easily resolved by asking the proper physical question, that is, soft photon
emission processes must also be considered. The result is expressed in the
general framework of (2.74) as
(2m)2 1
2
i/! (M)=A{ Ji-(2:;)" M>2m,
- 23+ (2m)
2
M ] , M>2m. (2.76)
It should be noticed that all the integrals exist.

50 J. SCHWINGER
The result that we have just obtained describes the electromagnetic properties
of a free electron. For the Lamb-shift calculation, on the other hand, one has
to deal with a bound electron with p 2 + m 2 ¥= 0. In this case one can use J
to represent the charge distribution of the nucleus, which produces its
effect through A . The basic problem is then the multiple scattering of an
FIGURE 2.13
electron in a Coulomb field. To obtain the radiative interaction modifi-

cation one considers the scattering problem using extended electron sources
to produce a well-defined causal situation. The extended electron source
1'/i emits a virtual electron of momentum Pi which cannot last very long and
converts into a real photon of momentum k and a real electron of momentum
Pi - k. The electron moves to the nuclear potential and is scattered. The
final electron of momentum p 1 -k and the photon recombine into a virtual
electron of momentum p 1 which again cannot last very long and is detected
by the extended electron source 17 1 . The causal sequence is schematically
represented in Fig. 2.13. The calculational advantage of this process is that
the integration over all possible photon momenta k is limited by kinematics,
k2 = 0, (p2-k)2 +m2 = 0, (P1 -k)2 +m2 = 0, (2.77)
and only one parameter is left, which appears as an angle in a certain

°
coordinate system (say, p 2 = O,p 2 =Mi)· The result of this calculation is a
contribution to the vacuum amplitude of the form
.J (dp1) (dp2) 0
z ( n)3 ( n)3 lf/1(-P1)Y (-eq)Aµ{p1 -pi)
2 2
X [gd'+giiO'µv(P1 -pi)v]lfli(Pi), (2.78)

where g 1 and g 2 are invariant functions of the three scalars, -p 12 = M 12 ,

-P22 = M22' and {p1 -p2)2.
We now write
lf/1(-P1) = f (dx1)1f11(X1)eipixi,
f
lf/2(P2) = (dx2)e-ip2x21f1i{X2),
f
Aµ(p1 -p2) = (d()e- i(p1 - p2>~ Aµ((). (2.79)
°
The well-defined causal arrangement guarantees that x 1 > c; 0 , and c;0 > x 2 °.
On writing
(dp1) 2 (dp2) 2
(2n)3 = dmP1dM1 ' (2n)3 = dwP2dM2 ' (2.80)
we find that the typical structure in space-time representation is
f (dx1Xdx2)(d;)dM1 2dM/V11(X1)A+(X1 -e, Mi 2

)
x A(()A+(e-x2, M2 2)1f12(X2), (2.81)
in which A+(c;-x 2 , M 22) and A+(x 1 -c;, M 12) describe, respectively, the
transmission of mass M 2 from the extended source 17 2 to the nuclear potential
A(e), and transmission of mass M 1 from A(() to the extended source 1'/t •
After the space-time generalization, we return to the momentum-space
representation,
-z
·f (dp1) (dp 2 ) 0[
( n)4 ( n)41f11(-p1)Y K1yi(-eq)Aµ(p1 -pz)
2 2
+g2eq(1/2)aµvpµv(p1 -P2)}/f2{p2)
1
f
x dM 12 dM 2 2 [ •
P1 2 +M1 2 -ze
. 1
2+M 2 • f(<P1 -pz) 2 , Mi 2 , Mz 2 )] , (2.82)
P2 2 -ze
in whichp 1 2 andp/ are not restricted to -p 1 2 = -p 2 2 = m 2 • The additional
term/(<p 1 -p 2 ) 2 , M 12 , M/) appears here because the discussion by means
of the well-defined causal situation effectively studies the dependence upon
p 1 2 and p 2 2 , but leaves undetermined any contribution that involves only
the third invariant {p 1-p 2 )2. The/term can be removed by subtracting the
already known result for - p 1 2 = - p 2 2 = m 2 • Despite the way (2.82) has
been written, this is to be performed separately for the yA and the aFterms,
and turns out to be unnecessary for the latter.
3. STRONG AND WEAK INTERACTIONS
3.1 Low-Energy 7r + N System
In electrodynamics the very intuitive idea that an accelerated

charge radiates enables us to elaborate the whole dynamics of electrons and
photons, beginning with the primitive interaction. Nothing so simple is
available in strong and weak dynamics. We shall exhibit another facet of
the source formalism by using it to search for empirical regularities, with a
minimum use of speculative hypotheses.
The essential contribution of the source theory is the possibility of presenting
the phenomenological description of a physical system in the general form
(3.1)
where x are the fields corresponding to the sources S, and .!l'(x) is the
Lagrange function of the system. The latter is not a local function of the
fields, but may be so represented under limited physical circumstances.
Consider for illustration the n + N system at low energy. The pion is an
isotopic triplet (T= 1), and the nucleon is an isotopic doublet (T=(1/2)).
One also has to distinguish a nucleon from an antinucleon. This distinction
is conveyed by the use of a complex source. We will use the particle symbols
to represent the corresponding fields. For the n+N system we have then
(3.2)
53
54 J. SCHWINGER
where
N = N*y 0 , (3.3)
and
(3.4)
Ii' 11: = -(1/2)[ (oµn) 2 +m11: 2 n 2 ], (3.5)
.ZN= -N(yµ(1/i)oµ+M)N. (3.6)
Some significant interaction terms for the low-energy n + N system are
f -
!t'11:N =-N~iy 5 r:N · oµn
m11:
2
/ 0 ) _
+ ( m11: N ~r: N · oµn x n. (3.7)
If we consider the exchange of a virtual 1t, the first term gives the long-range
behavior of nuclear forces. If the exchange of a virtual nucleon is considered,
it describes the low-energy limit of p-wave n-N scattering. The value off
obtained in this way is
f = 1.01±0.01, (3.8)
12
or ,...., 0.08. The second term is introduced phenomenologically to account
4n
for the low-energy s-wave n-N scattering. The experimental data are
m11:(a(1f2)-a(3f2)) = 0.292±0.02,
m11:(2a(3!2) + a(lf2)) = -0.035 ± 0.012, (3.9)
where a(l/l) and a< 312 > are, respectively, the scattering amplitudes for the
n+N system in the T= (1/2) and T= (3/2) states. The/0 2 term alone gives
2a(3f2>+ac 112 >= 0. But the/term also gives a smalls-wave contribution
(3.10)
m 1
where 2 ~ - . The result is consistent with experiment. The relation
M 6.9
(3.11)
gives
lo = 0.84 ± 0.03. (3.12)
The two parameters I and lo are the information extracted from the low-
energy strong interaction phenomenology of the n + N system.
56 J. SCHWINGER
3.2 Partial Symmetry. Chiral Invariance
The Lagrange function (3.4) is invariant under the isotopic

rotations,
<5n = - <5w x n,
<5N = (i/2)r · <5wN, (3.13)
which form a group 0 3 or SU2 • This invariance is actually only approximate,

since
Mn=l=Mp.
Isotopic spin invariance is only a partial symmetry, abstracted by dis-
regarding certain terms in the Lagrange function, specificall:Y the mass
differences in the present case. This is a very general feature, empirically, of
strong interaction dynamics. Another significant example of partial sym-
metry is obtained upon the recognition that the pion mass is the smallest of
all meson masses. If it were zero, then the Lagrange function
would have a new symmetry. It is invariant under the displacement of the

n field by a constant,
n-+-n+q.1.
Does this symmetry have any significance for the nN interaction? Observe
that it is also true for the f interaction term of (3. 7), since this term involves
only derivatives of the n· field. However, it is not valid for the / 0 term.
Considering an infinitesimal displacement, we obtain
2
/ 0 ) _
<5.!l'nN = (
mn N'/r:N. oµ(n x <5q.1). (3.14)
This structure can be compensated by introducing an isotopic-rotation like

transformation on the nucleon fields
i( ~:)'
JN= r · n xc5tpN,
- (/o)
<5N = - i mn
2
-
N r · n x <5q.1. (3.15)
The transformation (3.15) also introduces changes in the f interaction term.

But they are at least cubic in the n field and its variation, and therefore lie
outside the limited framework being considered. Our attention is concen-
trated on only those processes involving one or two n particles.
Given a set of transformations, it is essential to examine its group structure.
A simple way of testing the closure property is to consider a sequence of
infinitesimal transformations, say (12), and to compare with the transfor-
mations performed in opposite order, that is, (21)- 1 • The group property
demands that the resultant transformation, symbolized as (12)(21)- 1 , be
a third transformation of the general kind. Such a procedure has commonly
been used to derive the angular-momentum commutation relations. Now
consider the sequence of transformations
N' =[ l+i(~:)' t · nxO,~] N,

N" = [ !+i(~:)' t · n' x 02 ~ ]N', (n' = n+O~)
=[ !+i(~:)' r · (n+0 1 ~) x 02 ~]
x [1+;(~:)' r · nx0 1 +. (3.16)
On comparing with the transformations in opposite order, and keeping only

lowest powers of the n field, to be consistent with our restricted starting
point, the final transformation connecting the two sequences of operations
is an ordinary isotopic rotation
(3.17)
or
(3.18)
where
(3.19)
The corresponding group property of the isotopic rotations is

~c12fV = ~ 1 w x ~ 2 w. (3.20)
The relabelling
(3.21)
(3.22)
58 J. SCHWINGER
enables us to recognize the group structure of 0 4 (the four-dimensional

Euclidean rotation group with six parameters <5w 12 , '5w 34 , etc.). It is well-
known that the group 0 4 can be factored
(3.23)
if the parameters <5w 12 ± <5w 34 , etc. are introduced. Thus, we have now
two independent isotopic rotation groups with parameters
<5w ± = <5w ± M<p . (3.24)
Under a spatial reflection, '5w~ '5w, but '5<p~ - <5<p, therefore '5w ++-+'5w _ .
The combinations '5w+ and '5w_ thus correspond to opposite handness, or
chirality, which are interchanged under a spatial reflection. We shall call the
0 4 transformations in general and '5<p transformations in particular, chiral
transformations.
Thus far, we have abstracted the chiral group from the phenomenology of
the low-energy n + N system. Let's see if we can apply the result to deduce
new relations. The actual response of the Lagrange function to a constant
'5w and <5<p transformation is
(3.25)
Following a classic Lagrangian technique for deriving currents, we now

let '5w and '5<p be arbitrary space-time functions. This gives
(3.26)
where
_rT = oµn x n+Nyµ(1/2)rN + ... , (isotopic current)
(axial current) (3.27)
in which we have retained only the simplest contributions. The stationary

property of ft' (at points far away from sources) then gives
aµrr=O,
ak µ = m1t 2 n = ( m1t) m 211: (3.28)
µ A. 2/o 1t '
which state the conservation of isotopic current and partial conservation of
axial current, respectively. Alternatively
(3.29)
where
(3.30)
(3.31)
The known chiral structure of lepton currents in weak interaction suggests

the use of j + µ as the corresponding n + N weakly interacting current. This
single hypothesis puts together in one package a number of hypotheses
accumulated over the years. These include: the Feynman-Gell-Mann
hypothesis connecting pion P-decay to nucleon vector P-decay coupling;
the Goldberger-Treiman relation between pion instability and the nucleon
axial vector coupling; and the nucleonic axial-vector-vector ratio
GA f 1.01
- - =- =- = 1.20+0.03 (3.32)
Gv fo 0.84 -
(Experimental value: 1.18 ± 0.027)

which is the Adler-Weisberger relation, as interpreted by Tomozawa and
Weinberg. Later we will discuss another approach, which gives a pure
number for the strong- and weak-interaction ratios. It is
GA (5/3)
- - =--=- = 1.18. (3.33)
Gv ,J2
60 J. SCHWINGER
3.3 Non-Abelian Vector Gauge Particles, p and A 1
As the experimental energy increases, the assumption of locality

for the Lagrange function becomes invalid. But we also begin to recognize
new particles, in terms of which we can hope to regain some of the simplicity
of a local description. Consider the lightest 1- particle, the p meson, T = 1. It
will be natural to build a theory in analogy to the photon, with the isotopic
current replacing the electric current. But the p meson itself must contribute
to the isotopic current, while the photon is electrically neutral. Also the
non-zero mass of the p meson is a fundamental difference. The attitude we
want to take is another example of partial symmetry. We now approach this
problem with a gauge invariant requirement that would be exact in the
absence of the p mass term.
Since the p meson is coupled to itself, the gauge invariance here is a more
complicated kind. Consider the p + n system for illustration. Take the
Lagrange function as
ft'= -(1/2)(Dµn) 2 -(1/2)m/n 2
-(1/4)(oµPv-OvPµ+gpµ xpv}2-(1/2)m/(Pµ) 2 • (3.34)

where
(3.35)
is the analogue of oµ-ieqAµ in electrodynamics. Without the p mass term,

ft' is invariant under the isotopic-spin gauge transformation (non-Abelian
gauge transformation),
~n = -&v(x) x n,
~Pµ = -&o(x) xpµ+(1/g)oµ~w(x) = (1/g)Dµ~w(x). (3.36)
The first term in the p transformation expresses the fact that the p meson is
an isotopic-spin-1 particle, while the second term is characteristic of a
gauge field. The general response of ft' is
(3.37)
which implies, in particular,
(3.38)
When the nucleon is included, the relevant structure in ft' is
-Ny"'((1/i)oµ-g(1/2)r · pµ)N+mass term, (3.39)

where N responds to the gauge transformation as
<5N = i(r/2) · <5w(x)N. (3.40)
We can find the connection with the previous approach through the p field
equation
=gjµT· (3.41)
Under the previous low-energy circumstances, the p meson cannot be
produced as a real particle, i.e., iJ2«mp2 , and (3.41) reduces to
m2
p •
--pµ'::::!.)µT' (3.42)
g
so (3.37) gives the previous result. It indicates that the introduction of the
p meson is a consistent generalization.
We can also make contact between the p and chiral transformations. Under
the displacement
<5n = <5rp (constant), (3A3)
we have
f5jµT = 0µ11: X f5rp+ .. .
= <Vn x <>rp)+ ... . (3.44)
The low-energy connection (3.42) gives
. (3.45)
and we identify the gauge parameter <5w,
Oro= ( !J 7tX Jtp. (3.46)
The corresponding nucleon transformation (3.40) now gives
<5N= -i ( - g )2 r · nx<5rpN. (3.47)

2 mP
But we also know the appropriate chiral transformation for the nucleon; it is
2
<5N = i ( fo ) r · n x <5rpN. (3.48)
li'ln
62 J. SCHWINGER
The comparison of (3.47) with (3.48) implies the relation
(3.49)
From the known values oflo , one finds

g2
4
n = 3.4±0.2. (3.50)
Thus, the universal coupling constant g is fixed by the low-energy pheno-

menology.
Observe that the s-wave nN coupling term (the lo 2 term) is now completely
accounted for by the p exchange mechanism. The relevant coupling terms
in !l' are
(3.51)
Each coupling term describes an effective p source, and the consideration of
p exchange between the two sources gives the contribution to the action,
f (dx)(dx')gN )""( 1/2)rN(x) ( g., - m> a.a,) A+ (x- x'),

xgoµ'n x n(x'). (3.52)
For small momentum exchanges, p 2 «m/, we have
Ll+(x-x) =
I I (dp)
- -4 -
eip(x-x')
---
P (2n) p +m/-ie
2
1 I
-+- - 2
c5(x-x), (3.53)
mp
and (3.52) becomes
g2
f
- 2 (dx)Nyµ7:N · oµnxn
2mP
=f (dx) ( ~:r N)""rN • a•" x ", (3.54)
in which the connection (3.49) has been used. Notice that (3.54) is just the
s-wave coupling term in (3. 7).
Since the coupling constant g is known, it would seem that the interaction
term Pµ · oµn x n could be used to describe the instability of the p meson,
the decay p-+-2n, i.e., the p width could be calculated. However, the p decay
is a high-energy process, and a more elaborate high-energy theory is needed.
What is missing here is a treatment of the <5rp transformations analogous to

the <5w transformations. Recall that <5w and J.<5rp are the six parameters
associated with the group 0 4 • So far we have just let the parameters <5w be
space-time dependent to introduce the p meson. Obviously, we should
proceed analogously with the <5rp transformations. These considerations
suggest the introduction of a 1+ , T = 1 particle and the corresponding use
of non-Abelian gauge transformations. This particle is A 1 , with mass
m = 1080 MeV. (From the experimental side, there are repeated attempts to
show that it is only a kinematical effect.) Let's begin with the n field and
consider its response to a '5rp(x) transformation,
<>[ (oµ + gpµ x )n] = oirp+gpµ x <5rp, (3.55)
which suggests the introduction of a compensating term - (1/2)mAdµ,

where the constant mA is introduced for dimensional reasons. The trans-
formation law is
2g 2
<5.91µ = - -<5rpxp
mA µ+-o
mA µ<5rp-<5wx d,
µ (3.56)
where we have included the response of dµ to an isotopic rotation <5w.

Also we have
(3.57)
where <5<ppµ is the response of Pµ to a chiral transformation, and is to be

determined. Consider the combinations <5(pµ± dµ),
(3.58)
Since the combinations Ow±(~) Otp appear everywhere except the last
terms, the requirement of a group structure obviously demands
(3.59)
And we have
(3.60)
with
2g
<5w ± = <5w ± - <5rp. (3.61)
mA
64 J. SCHWINGER
This must be the same group as before, where we had
2/o
<5w ± = <5w ± - <5rp;
m1t
therefore
g Jo g
-=-=-=--,
mA m1t .J2mP
and
(3.62)
which is only a mathematical statement, not yet a physical assertion.

We now introduce the quantities Pµv and dµv defined by
Pµv±dµv = oµ(pv±dµ)-ov(pµ±dµ)
(3.63)
or
Pµv = OµPv-OvPµ+g(pµ x Pv+dµ x dv),
dµv = (oµ+gpµ x)dv-(ov+gpv x )dµ
= Dµdv-Dvdµ- (3.64)
A natural choice for p and d contributions to the Lagrange function !£ is
!l'P~ = -(1/4)( (pµv)2 +(dµv) 2 ]-(1/2)mp 2 [ (pµ) 2 +(.91µ) 2 ]. (3.65)
Only the combinations (p + .91) and (p- .91) can occur in !l'P~ in order
2 2
to satisfy the chiral invariance, yet these two structures must occur with the
same weight since parity is conserved in strong interactions. The partial
symmetry response of !£ is
or
m2
<)ft'= - _P_pµ • Oµf5W-mA.9/µ • oµ<5rp. (3.66)
g
Comparison with the low-energy behavior (3.26) indicates the necessary
correspondences for low-energy behavior,
(3.67)
2
If we now include the n term [ (c\+gpµ x )n-(1/2)mAdµ] , it is
immediately seen that n and dµ are mixed objects. We must carry out a
diagonization, which will also change the normalization of the n term. In
anticipation of this, let us write
!l'np~ = -a[Dµn-(1/2)mAdµ] 2 -(1/2)mn 2 n 2
-(1/4)((pµv)2+(dµv)2]-(1/2)m/(Pµ 2 +d/). {3.68)

In order to identify the physical particle fields, it is sufficient to consider
only the quadratic structure,
-a[ oµn-(1/2)mAdµ] 2 -(1/2)m1t2 n 2 -(1/2)mp 2 dµ 2

-(1/4)(oµdv-ovdµ) 2 •
Now, the transformation
1
dµ = A 1 µ+-oµn, (3.69)
mA
or, more generally,
1
dµ = A 1 µ+-Dµn, (3.70)
mA
gives
-(a/4)[ oµn-mAA 1 µ] 2 -(1/4)[ oµn+mAA 1 µ] 2

-(1/2)m/n 2 -(1/4)(oµA 1 v-ovA 1 µ) 2 •
Diagonization is achieved by the choice
a= 1.
And we get
!l' = -(1/2)[ (oµn) 2 +m/n 2 ]-(1/2)[ (1/2)(oµA 1 v-ovA 1 µ) 2
+m/A 1 /J,
i.e. m A is the mass of the A 1 particle. One compares the theoretical prediction
mA =-J2 = 1.41, (3.71)

mp
with the experimental value
mA 1080
- = - - = 1.42.
mp 160
which are in striking agreement. The relation (3.71) was first derived by
Weinberg from a different approach.
66 J. SCHWINGER
We now return to the necessary low-energy correspondence (3.67). We have

already shown that Pµ does have the correct low-energy behavior. So we
now consider its counterpart, which can be expressed in the following
form:
(3.72)
But
mAdµ = oµn+mAA 1 µ,
which shows that the n term is correct. We are left with the necessary
correspondence
f-
mAA1µ"' - -Nyµiy 5 rN, (3.73)
mn
which anticipates a NA 1 coupling of the form
(3.74)
Let us check if this is indeed the case.

Now, the nucleon contribution
-Nyi((1/i)8µ-g(1/2)r · p,JN+ mass term (3.75)
is invariant under the isotopic gauge transformation
<JN= i( 1/2)r · <5wN,

<5pµ = (1/g)Dµ<5w. (3.76)
But it is not invariant under a chiral transformation oqi(x), to which the
nucleon field responds as
oN = i2 ( -
g )21- r · n x JqiN, (3.77)
mA 2
which is a nonlinear gauge transformation depending on the n field. This
suggests that one should replace pµ in the Np coupling term by a new
field Pµ':
I 2g
Pµ = Pµ + -2 n x (mAdµ-Dµn). (3.78)
mA
Under a oqi(x) transformation, we have
I 2g
<5pµ = - 2 Dµ(n x <5qi), (3.79)
mA
which, in conjunction with (3. 77), correctly gives chiral invariance. In
arriving at (3.79), we have neglected the variation of Pµ in Dµ which is outside
the limited framework being considered. So far, the f term in nN coupling is
missing. However, we always have the possibility of adding a phenomeno-
logical coupling which is a chiral invariant by itself. The same situation
occurs also in electrodynamics at a phenomenological level. A magnetic
coupling, which is gauge invariant, has to be introduced to describe the
anomalous magnetic moment. In order to account for the low-energy
p-wave nN scattering, one must introduce the coupling
(3.80)
wlµch is chiral invariant if one disregards more complicated couplings. The

structure 2Dµn-mA.9'µ is the appropriate gauge-invariant combination
when orp is space-time dependent. Now the complete pN and nN couplings
are
which indeed predicts a NA 1 coupling of the form (3. 74) demanded by

consistency requirement.
If we turn to high-energy processes, the vector current will appear through
the intermediary of the p meson, and the axial-vector current will appear
through the intermediary of the A 1 particle. Thus we learn that nucleon
vector and axial-vector form factors are governed by the p and A 1 masses,
respectively.
68 J. SCHWINGER
3.4 Widths of p and A 1

We can now use our results to predict the observed properties of
p and A 1 , in particular the p width and A 1 width. The Lagrange function
describing n, p, and A 1 is
ft'= -(Dµn-(1/2)mAdµ) 2 -(1/2)m/n 2
-(1/4)[ (Pµv) 2 +(dµv) 2 ]-(1/2)mp 2 ( (pµ) 2 +(dµ) 2 ], (3.82)

with
Dµ = oµ+gpµ x,
Pµv = OµPv-fJvPµ+g(pµ xpv+ dµ x dv),
dµv = Dµdv-Dvdµ• (3.83)

When we introduce
1
.9'µ = A 1 µ+-Dµn,
mA
this becomes
ft'= -(1/2)( (Dµn) 2 +m1/n 2 ]-(1/4)(pµJ 2 -(1/2)m/(Pµ) 2
-(1/4)(dµv) 2 -(1/2)mA 2 (A 1µ) 2 , (3.84)

where
and
(3.86)
which is obtained with the aid of the relation
[Dµ, Dv] =gPµvX. (3.87)

We now pick out the pnn and A 1pn coupling terms. These are
ft'p1r.1r. = gpµ. a1C x n- -41 -pµv.

µ
g
2
a 1C x fJ n
µ V'
mp
(3.88)
These expressions can be simplified,
(3.89)
When these couplings are applied to real p particles, further simplifications

can be introduced through the use of the field equation
a pµv = m 2pv
µ p '
Thus,
!l'p1Cn-+(3/4)gpP • oµn x n, (3.90)
fi7Aipn~-(1f2)gmA~ • A 1 µxn. (3.91)

Note the factor (3/4) in the pnn coupling. A straightforward calculation
gives the widths for p and A 1 :
3 g2
p3
I ' = - - -2 (3.92)
p 8 4n m p '
g2 [1+-1 (p- )2] p,

I'A 1 =1- - (3.93)
44n 3 mP
where p is the momentum of the decay products in the center-of-mass
frame. For p~2n, p == 350 MeV, and for A1 ~pn, p = 245 MeV. The pre-
dicted values are
~~100MeV (3.94)
[without the factor (3/4), the result would be rp"' 180 MeV] and
rA1~200 MeV. (3.95)
Experimentally, the average width of p is 125 MeV (116 MeV for p 0 and
132 MeV for p±), and the width of A 1 is 130±40 MeV. The theoretical
results are certainly qualitatively right, but indicate that the predicted I'P is
too small and I'A 1 is too big. However, the qualitative consistency lends
support to the underlying theoretical ideas.
Of course, we have used only the simplest possible theory in which no
adjustable parameter appears. As in the discussion of nN coupling, it is
always possible to add individually invariant terms with arbitrary coeffi-
cients. A relevant example here is
(3.96)
70 J. SCHWINGER
where
(3.97)
and
(3.98)
are chiral invariant combinations as far as our problem is concerned. The
contributions to pnn and A 1pn couplings are
~ ,,_!_pµv.
1 2
aµ nx an-
v
yg //,..µv. Ai µ XO v
n. (3.99)
2 mA mA
The first term corrects the effective pnn coupling constant,
pnn: (3/4)g~{(3/4)+ (1/4)y)g. (3.109)

We now consider an approximate evaluation of the A 1pn additional coup-
ling, based on the nonrelativistic behavior of the p meson in the decay
A 1 ~ p + n [(p/mp) 2 ,..., (1/10)]. For a particular decay, A1° = 0 in the rest
frame of A 1 , and
(3.101)
Here
PnvPv = (Pv-Ppv)Pv = -mAp0 , (3.102)
where Pv is the total energy-momentum in the decay process. But
{3.103)
and therefore
v p·p
PnvP ~-mA- (3.104)
mP
Also
-ppp1C = -(1/2)(P 2 -p/-p/)

= (1/2)(mA 2 -mP 2 -m1C
2
),
or
-ppPn~(1/2)mp 2 , (3.105)
neglecting the pion mass. Thus
or
(3.106)
This gives the additional A 1pn coupling

+ y(1/4)gmAfY' · A 1µ x n, (3.107)
and the effective A 1pn coupling constant becomes
A 1pn: g~g(1-(1/2)y). (3.108)
If y is chosen to be positive, it will increase I;, and decrease rA 1 • The choice
y = (2/5) (3.109)
predicts
I'P = 125 MeV,
rA1 = 130 MeV' (3.110)
in complete accordance with the experimental numbers.
IL, J. SCHWINGER
3.5 Low-Energy 7r7r Interactions
Our theory of n, N, and p chiral transformations considered so far

is mathematically (but not physically) inconsistent. The nucleon chiral
transformation, for example,
(JN= ;(:J 2
t · 11: x &pN, (3.111)
is non-Abelian, while the n transformation

(3.112)
is Abelian. This is because we have not considered nn interactions. Recall
that the nucleon interaction properties were used to identify the chiral g_roup
structure. If one considers a n system, in which the nucleon does not appear,
the physical evidence of the non-Abelian chiral group can only be revealed
through the interactions of the n mesons. We now extend the n transfor-
mation law and introduce nn interactions by insisting on the group structure
established for the n + N system under limited physical circumstances.
Since the transformation (3.112) through the p coupling induces an isotopic
rotation with ow= 2(gfmA) 2 n x orp, any isotopic-spin bearing field must
n
respond universally. Thus,
;Jn= - 2 (:J (11: x (J(P) x n+;J(P [ 1+(:A 11: (3.113)
in which we anticipate that a scale dilation of orp is also required for the
complete group property. The group property for the transformation law
(3.113) can be checked, as before, by comparing successive transformations
in one order
(:J (11:'
11:" = 11:' + (j2(P [ ! + (:A 11:')']- 2 X(j2(P) X11:',
7!' = 7!+(J1(P [ l+(:A n-2 (:J

1! (7! X(J,q>) X1!, (3.114)
and in opposite order. To illustrate the procedure, consider the simplest

terms, up to cubic in n and b1 rp, o2 rp.
11:"=11:+J++(:A 11:)'J+(J2(P[l+(:A 11:)']

+2 (:J (j2(P7!. ;J,qi-2 (:J (7! x (j2(P) x ;J,qi
-2 (:J 2
(;J, (P x (j2(P) x 1!' (3.115)
or
11:''= 11:-2 (:A)' (01q> 02'1') x 11:

X
+(0 q>+02q>)[ 1+(:J 1 :

2
1 ]
+2 (:J 2
11:0,q>. 02'1'· (3.116)
Upon comparing with the transformations in reversed order, all terms

symmetrical in 0 1 <p and 02 <p disappear, and we get exactly
(3.117)
with
(3.118)
since all higher order terms cancel completely. The nucleon transformation
law (3.111) is now also exact. The n transformation law (3.113) can also be
exhibited in the following form
011: = Oq>+ (:J 2

[211:0q> • 11:-0q>11: 2). (3.119)
We have exhibited a mathematical object which is transformed very simply

according to (3.119). But it is not necessarily the n field. It is always possible
to redefine the n field in such a way that the original physical identification
in terms of noninteracting particles is not changed. Nevertheless, the re-
definition may change interactions. Only experiment can determine which is
the correct identification. Let's recognize this ambiguity, and write generally,
in an infinite series
(3.120)
This general n field has the transformation property
(3.121)
We now return to 2 ir and ask for its chiral response. It is
(3.122)
74 1. SCHWINGER
Substitution of (3.121) for on gives
02.+m!n · O<p = o [~(2-P) (~J (O.n 2 ) 2 -P (:J

x (oµn) 2 n 2 -p(_!_) ~4 (oµn
mA
2
2 2
) + ~ m/(2-3fi)(n 2) 2
4
J+ .... (3.123)
It is important to recognize that we have the option of insisting that

(3.124)
still holds, if necessary by a redefinition of n produced by a suitable choice
of p, etc. This is a convention for the n field, not a hypothesis. Then the
combination
with
2 .. = _ ~( :J'[ (1-p)(a•"2)2 _ 2p(O•"),n'
+m/(1-~p )(n 2 2
) l (3.125)
is partially chiral invariant, broken only by the pion mass term in accordance
with (3.124). For the application to n-n scattering, the n field satisfies the
equation
(3.126)
and
(oµn 2 ) 2 + 2(oµn) 2 n 2 = 20µn. oµ(nn 2 )
-+ - 2o 2 n · nn 2
(3.127)
Consequently,
2,,-+- ~(:J[<a.n2)2 +m!( 1+ ~P )<n')'l

(3.128)
All sensitivity to the model parameter pis in the pion mass term.
The physical background of these considerations should be kept in mind.
Chiral invariance is recognized from the strong-interaction phenomenology
of the low-energy re+ N syste~ The conclusions are therefore mainly
about low-energy .processes. The possibility of generalizing to high-energy
processes is of course a tempting one, and in fact we have done so in the
discussion of A 1prc dynamics. But as a local interaction, (3.128) is expected
to be only applicable in the low-energy region. Thus we apply it to the
n-n scattering at the threshold. The implied scattering amplitudes for
T = 0, 2 states are
(3.129)
A model-independent combination is
(3.130)
However, no direct experimental comparison is available.

76 J. SCHWINGER
3.6 Spectrum of the Decay Process 11*(960)~11+27r
There is a closely related strong-interaction process for which some

experimental information is available. It is the decay
17*(960)-+11+2n, (3.131)
and the accompanying deviation of the 11 energy distribution from purely

phase-space considerations. This is at the moment the only experimentally
known strong-interaction process which involves just spin-0 particles.
Both 11and11* are T= 0, o- particles.
We must first consider the unitary generalization of our chiral transfor-
mation. Let </> be a n x n Hermitian matrix and consider the infinitesimal
unitary transformation (ow is an infinitesimal Hermitian matrix),
o<f> = i[ ow, ¢]. (3.132)
The group property is
Oc12Jw = (1/i)[ 0 1 w, 02w]. (3.133)
For an SU2 matrix, we have
</> = (1/2)r · n, (3.134)
and
ow= ( 1/2)r · bw, (3.135)
where the bw on the right-hand side is a vector, while the ow on the left-
hand side is a matrix. Now (3.132) simply reduces to the ordinary isotopic
rotation of the n field.
The general chiral transformation on </> is
(3.136)
The exact group property
(3.137)
with
(3.138)
can be readily verified. For SU2 , &p = (1/2)r · &p, and (3.136) gives the
previous chiral transformation of the n field, since
7: • n 7: • &p 7: • n = 7: • [ 2n'5<p · n - O<pn 2 J. (3.139)
The complete group structure of the transformations (3.132) and (3.136) is

Un x Un if the parameters bw±M<p are introduced.
For the present purpose, we are interested in the generalization of SU2 to
U2 , and the field matrix ¢ is
¢ = (1/2)(17 2 +r · n), (3.140)
where
17 2 = cos 917 + sin 917*, (3.141)
is the projection of a full U3 object into the U2 subspace, and is a mixture of

the physical particles 11 and 11*. There will be another combination q3 ,
orthogonal to 172, to represent the projection along the third axis. We shall
not consider chiral transformations associated with q2 • The n chiral trans-
formation is now represented by
'5<p = (1/2)r · '5<p, (3.142)
and (3.136) implies the differential transformation laws for n and 17 2 ,

'5n = O<p + ((1/2).1) 2 [ 2n'5<p • n + O<p(17 2 2 - n 2 )],
(3.143)
Incidentally, the combination 17 2 / [ 1 + ((1/2)A.n) 2] is a chiral invariant,
(3.144)
if terms cubic in 17 2 are ignored. Again let's recognize the possibility of

redefining fields and write more generally
n-+n[ 1 + (1-fi)((1/2)A.n) 2-(1-P 1 )((1/2)A.17 2) 2 + ... ] ,
112-+112[ 1 +(1-P2)((1/2)A.n) 2
-(1-fi3)((1/2)A.112) 2 + ... ].
Then
on= orp[ 1-(2-fi)((1/2)A.n) 2+ (2-fi 1 )((1/2)A.71 2) 2]
+ (1/2)fiA.2 n'5<p • n,
'5112 = (1/2)fi2A- 2orp · n112. (3.145)
We now return to !l'n and investigate the 17 2 -dependent part of its

response (the 17 2 -independent part will simply lead to !l'nir which we have
already discussed). It is
('5!l'1t)112 = -oµn · orp(2-p 1)((1/2)A.) 2oµ172 2
-m/n · O<p(2-fi 1 )((1/2)A.17 2)2.
J. SCHWINGER
The additional coupling which will provide a compensating term is there-

fore
_
It'1t1/2 -
fo ) 2 ( _
mn
~
1 2 /31
)
[ .
aµ 7t 2 0µ112 2 +mn 2 n 2r/2 2] ' (3.146)
(
in which n2 is treated as a chiral invariant. Now (3.146) implies the n*nn
(
coupling
.P, ••• = ( ~:r 1- ~ p1) sin 28[a•n a.<1111*) 2
(3.147)
using (3.141). It is interesting to notice that all the unknown parameters 9

and /31 are lumped together as a multiplicative factor.
Analogous considerations on Ii',,+ z,,.
will give new couplings, but they
do not contribute to the decay process n* -+1J + 2n. This can be seen from the
following. From (3.144) and a similar statement about 17 3 it follows that
substitutions of the form
'f]-n7+a11*n2'
rJ*-+rJ* +b'f/7t2'
will make z,, + 2,,. chiral invariant as far as the present applicatio,n is
concerned. The additional terms so obtained which contribute to the decay
'f/*-+'f/+ 2n are
- aµ'f/oµ( a'f/ *n 2 ) - m,, 217Q'f] *n 2

-0µ11*oµ(b'fJn 2 )-m,,. 2'f/*b'f/n 2
-+(8 2 -m/)rJ. an*n 2 +(o 2 -m,,. 2 )rJ*. bnn 2 =0, (3.148)
since 1J and n* are being used as fields of real particles. Therefore, there is
no contribution from 2 11 + 2 11 •• This is a consequence of the stationary
action principle.
The structureless coupling n 2 r/'f/* simply introduces an invariant phase-
space factor. For the decay
'f/*(P)-+'fJ(p) + 2n, (3.149)
as described by (3.147) there is an additional factor
-(P-p) 2 -mn 2 = m,,. 2 +m,, 2 -mn 2 -2m11 .+(m,,+ T,,)

= (m,,.-m,,) -mn 2-2m .T ,
2
11 11
(3.150)
where T,, is the kinetic energy of 11. It is very useful to note that
(3.151)
within experimental error. Experimental analysis is expressed in terms of
the Dalitz parameter
m,,+2mn T,, 8
y= -1, -1<y< - ' (3.152)
mn m,,.-m,,-2mn 7
and (3.150) becomes
7
1--y (3.153)
17 '
apart from a multiplicative factor. Thus, the spectrum will deviate from the
pure phase-space prediction by an asymmetry factor
( 1- 177 y y. (3.154)
Comparison with experiment (102 events - 21 background) is listed as

follows.
y= -0.8, -0.4, 0, 0.4, 0.8,
Experiment 26±8, 17±5, 15±5, 11±3, 12±4,
151(1-2.
3
17
y )' 27, 21, 15, 11, 7,
The normalization is adjusted to the total number of events after removing

the background. The theoretical predictions are in reasonable agreement with
experiment.
ov J. SCHWINGER
3.7 Pion Electromagnetic Mass
We shall now discuss the superposition of electromagnetic effects

on strong interactions through the requirement of gauge invariance. Let's
begin with the isotopic-spin non-Abelian gauge invariance, broken only by
the p mass term. The p transformation and the response of the Lagrange
function are
1
opµ = - <>w x Pµ + - aµow,
g
m2
Oft' = - _P_ pµ ' 0µJW. (3.155)
g
The latter implies, incidentally, that
oµpµ = 0.
Suppose we pick out one axis, say the third axis, and consider ow directed
along this axis; we find
(3.156)
which is an Abelian gauge transformation. We now recognize the possibility
of maintaining invariance under this transformation through the compen-
sating effect of the electromagnetic gauge transformation
oeAµ = oµow, (3.157)
\
and we have
o(gpµ 3 -eAµ) = 0. (3.158)
Thus we propose to incorporate the electromagnetic effects by the generali-
zation of the p mass term to
2
- -21 mp 2 (p µ,1 2 ) 2 - -21 mp 2 ( pµ 3 - -ge A µ ) (3.159)
Gauge invariance now appears as a combined property. When ow is directed

along the third axis, Pµi and Pµi rotate in a way characteristic of charged
fields, leaving the first term in (3.159) unchanged, while the inhomogeneous
transformation of pµ 3 , (3.156), is compensated by the gauge transformation
of Aµ, (3.157). One must verify that the constant e in (3.159) is indeed the
physical charge. For this, imagine a strongly interacting system is subject to
an external potential, which acts as a driving term. The major effect is to
displace the p field,
(3.160)
Thus, the nucleon coupling, for example, becomes
gNyµ(1/2)rN · Pµ~gNyµ(1/2)rN · Pµ
+ N yµe( 1/2)r3 AµN, (3.161)
which reproduces correctly the isotopic-spin-dependent part of electrical

charge. The isotopic-spin scalar part is obtained by a similar treatment for
the other neutral 1- particles, w and ~.
The coupling between the photon and strongly interacting particles is
2 e
mp -plAw (3.162)
g
Through the identification of a photon source, (3.162) implies a coupling in
the action, through photon exchange,
(3.163)
Thus, the introduction of the photon modifies the strong-interaction

phenomenology. Among other things, it produces a mass splitting between
electrically neutral and charged particles. As a simplest example, we apply
(3.163) to the pion electromagnetic mass splitting by using chiral trans-
formations. The idea is as follows. The pion mass term breaks chiral sym-
metry. Now the action we will produce an additional violation which can
be identified as a change of the pion mass term. Under a chiral transfor-
mation we have
(3.164)
where dµ is the axial-vector gauge field associated with chiral transfor-

mations.
Since chiral symmetry is broken, the divergenceless condition on ~ is
violated under a chiral transformation, (3.164). But, as a photon source, it
must be conserved. In order to guarantee that only the conserved part
contributes, we insert a projection operator in (3.163), i.e.,
(3.165)
where
(3.166)
82 J. SCHWINGER
In momentum space, Dµv has the form
(3.167)
Now, the chiral response of the action we is
OqiWe=
m
_P_
( g
2
e) --f
2
2g
(dx)(dx')
mA
(3.168)
and
m
o/we = ( ~
2e) 2(mA
2g)2 f (dx)(dx')Dµv(x-x')
x {(dµ x ocp)3(x). (dv x ocp)3(x')
+ [ (pµ x ocp) x bcp ]3(x)p/(x')}. (3.169)
Let us emphasize that we are now considering processes that are not included
in the strong-interaction phenomenology. A typical term in (3.169) contains
two fields. dP(x)dv(x'), or pµ(x)pv(x'), which operate at two different
FIGURE 3.1
or
FIGURE 3.2
space-time points, and describe the emission or absorption of a particle,

as shown in Fig. 3.1. Included among all these processes is the possibility
that one of the two sources absorbs the particle emitted by the other (Fig.
3.2). That is, any two-particle term also includes the virtual exchange of
one particle. All this is contained in the general vacuum amplitude for
noninteracting particles
(O+ Io_?= exp[ (i/2) f (dx)(dx')S(x)G+(x-x')S(x')]. (3.170)
Let
S-+S1 +s2 +S,
and note that those terms which depend on S 1 and S 2 are
exp[ if (dx)(dx')S 1 (x) G+ (x- x')S2 (x') +if (dx)S 1 (x)x(x)
+if (dx)x(x)Si(x) J, (3.171)
where the field x(x) is

x(x) = f (dx')G+(x-x')S(x'). (3.172)
For weak sources, the terms bilinear in S 1 and S 2 are

- f (dx)(dx')S (x)[x(x)x(x')-iG+(x-x') ]S (x').
1 2 (3.173)
Accordingly, the virtual exchange process is produced from the process

involving two real particles by the substitution
x(x)x(x')-+ -iG+(x-x'). (3.174)
Isotopic-spin invariance demands

G+(x-x')ab = <>abG+(x-x'), a, b = 1, 2, 3, (3.175)
where a and bare isotopic-spin indices. The substitution (3.174) gives

<5/we = 2e 2 m/if (dx)(dx')Dµv(x-x')[ ~"(x-x')P
- Gµ"(x-x')a](o<fJch) 2 , (3.176)
where O<fJch refers to charged components. In momentum-space representa-

tion, we have
f
1 (dp)
<5<1' 2 We= -2e 2 mp 2J (dx)(O<fJch) 2 --:-- -
z
(
2n)
" [
4 Dµ (p) Gµv(P)p
- Gµv(P)a]. (3.177)
This is identified as the chiral response of the additional charged pion mass
term
(3.178)
and gives
(3.179)
84 J. SCHWINGER
Now
(3.180)
1
dµ=A 1µ+-oµn, (3.181)
mA
we also have
PµPv 1
gµv+--2 --2 PµPv
mA mA
Gµv(P)a =
p
2
+ mA 2 - ie
+ p 2 + mn 2 - is
• (3.182)
All the longitudinal terms are ineffective in virtue of the propagation

function Dµv(P)[pµDµv(P) =OJ, and therefore
(3.183)
But
(3.184)
so that
om
n
2 =6e2m
P
2 f!__ (dp) _1_r ___1 __
i (2n) 4 p 2 -it:l_P 2 +m/-ie
- p 2 +m~ 2 -ie] · (3.185)
We now use Euclidean spherical coordinates ((1/i)dp 0 = dp 4 )

(1/i)(dp)4n 2p 2 dp 2 '
and get
om 2 = 6e2mp 2 [''° dp2 ( 1 - 1 ) . (3.186)
n 16n2 Jo p2+m/ p2+m}
The result is
om n
2
=[ 3
1X log 2Jm
2n P
2
'
(3.187)
where the relation mA 2 = 2m/ has been used. This gives

omn = 5.0 MeV,
to be compared with the experimental value 4.6 MeV. The chirality cal-
culation has several features. It is restricted to the pion. Implicit in this calcu-
lation is the neglect of the pion mass relative to those of other particles.
The method seems to suggest that the particle A 1 is essential to a pion
electromagnetic mass calculation, as indicated by the cancellation of the
two propagation functions, Gµv(P)p and Gµv(P)a·
We shall now describe briefly a more general approach to show that although
the consideration of A 1 is significant, it is not fundamental. A not unsatis-
factory result is obtained from the re+ p system alone. This method is not
restricted to the re calculation. It recognizes that the action we produces an
electromagnetic modification of the p0 propagation function. The modifi-
cation introduces new processes associated with p0 exchange. Thus, for n
mesons we have the couplings pnn and nA 1 p. Two re sources can exchange
pre or A 1p, as indicated in Fig. 3.3. These processes are already included in the
FIGURE 3.3
strong interaction phenomenology, but the electromagnetic modification

effects are not. They produce new phenomena that can be evaluated indi-
vidually. The results (for mn«mp) are
2 3a 2
(omn ),mp= -mP , (omn)n1tp = 3.6 MeV. (3.188)
4n
(3.189)
But in this approach we can also consider the mn 2 correction to these pro-
cesses. For the most important one (nnp) we get
2
(omn )nnp = -3a mp [ 1 + -
2 2 m/ mP + -1 )]
- ( log-- (3.190)
4n mp 2 mn 2 2
in which the second term gives a 13°/0 correction, and (om1t)1C1tP becomes
(3.191)
The result is still qualitatively satisfactory, but the quantitative problem

remains.
86 J. SCHWINGER
3.8 U4 x U4 Partial Symmetry
We shall now discuss an entirely different example of partial

symmetry. Another aspect of the source formalism has not yet been men-
tioned. It is connected with the classification scheme of strongly interacting
particles, using the unitary group U3 to label the particle multiplets. We use
the sources, or associated fields, to carry the indices. Let us designate the
two types of three-valued unitary indices by a and a*. Then a source like
Sab• can be decomposed into a unitary octuplet and a unitary singlet, 3 x 3
= 8+1. Similarly, the source Sabe which is totally symmetrical in abc
describes a unitary decuplet (10), and Sabe with an antisymmetrical pair be
contains a unitary octuplet and a unitary singlet, since an antisymmetrical
pair of similar unitary indices is equivalent to a complex conjugate index.
If we employ the uniform representation of all spins by multispinors, the
complete structure of the source now takes the form
Particularly interesting is the subset with n = nf, which seems to describe

the best known particles. The source s, 1a, , 2b• describes spin-0 and spin-1
meson multiplets, each containing a unitary octuplet and singlet. This
source can be applied to the well-established o- and 1- mesons. And the
source s,1a1. {2a2. {3a3 which is totally symmetrical in the three pairs of
indices contains both spin (1/2) and (3/2), forming a unitary \octuplet and
a decuplet, respectively. This source is applicable to the known system of
(1/2)+ baryons and (3/2)+ baryon resonances. These structures appear as
generalized sources which unite sources of individual particles with various
spins and a common parity, but with no implication about mass
degeneracy.
Let's consider the generalized meson source and for the moment omit the
unitary indices. The structure s, 1 , 2 is a 4 x 4 matrix. This number of com-
ponents is more than necessary to describe the spin multiplicities. The
physically significant components are selected by the use of projection
matrices. However, we can view this from a more general standpoint which
will give a connection with the field description we have adopted for spin-0
and spin-1 particles. The indices of the source s, 1 , 2 can be classified by the
eigenvalues of y0 and the spin component <J along a certain direction, i.e.,
' = y0 , <J. Then we can decompose s, 1, 2 into four sets, in accordance with
eigenvalues of y1° and y2 °. This is a parity classification. The even-parity
components have y1 Of= + 1, y2 °f = + 1, or y1°f = -1, y2 Of = ~ 1. The
odd-parity components have y1 °f=+1, y2 ° = -1, or y1 Of= -1, y2 Of=+1.
1
The ( + +) components form a matrix Sa,a'(+) and similarly (- - ) com-

ponents form another matrix S<1,<1'<->. In the Lagrange function describing
the pion (spin 0, odd parity) and the p meson (spin 1, odd parity), the pion
is represented by a pseudoscalar field n, and the p meson by a four-vector pµ
in which the spatial components pk have odd parity, but the time component
p 0 has even parity. In order to establish the correspondence between the
multispinor description and the tensor field description, we use the equivalent
first-order form for the Lagrange functions of n and p,
(3.192)
and
2
fi'P = -(1/2)pµv · (aµPv-avPµ)+(1/4)(Pµv)
-( 1/2)mp 2 (pµ) 2 • (3.193)
The principle of stationary action implies the connections
(3.194)
(3.195)
and fi'n and fi'P restate the previous structures. But now the pion is des-
cribed by five fields (n and nµ), and the p meson by ten fields (Pµ and Pµv).
Thus, on supplying one more scalar, we have a one to one correspondence
between the multispinor and the tensor field description, each having
16 components.
Let's now consider a situation in which the meson fields carry arbitrary small
spatial momenta,
vx=O, (3.196)
where x is any meson field. Then we have

fi'1t-+ -n°o 0 n-(1/2)(n°) 2 -(1/2)m1/n 2 odd-parity fields
+(1/2)(nk) 2 , even-parity fields (3.197)

and
fi'P-+ - pokOoPk-(1/2)(pok)2
-(1/2)mp i(pk)2 odd-parity fields
+ (1/2)Hp 2 + (1/2)mp 2 (p 0 ) 2 , even-parity fields (3.198)
where HP 1 = p23 , etc. It is noted that the dynamical connections between

odd-parity and even-parity fields are completely severed, in the absence of
the spatial derivative terms, and new symmetries appear. If we restrict our
attention to the U2 :!iubspace, the particles involved are n, 17 2 (11, 17*), p and w.
88 J, SCHWINGER
The correspondence between the multispinor notation and the tensor field
description is established by introducing the 4 x 4 matrices
M< ±) = (TC± Hp) . O"'l" + (112 ± H (1)) • (J
-(±S1 +mPp 0 ) • r-(±S0 +mrow0 ). (3.199)
These constructions are such that the even-parity structures in the Lagrange
function are reproduced,
(1/16) Tr [(M<+>) 2 +(M<->) 2 ]
=(1/2)[(rc) 2 +H/+m/(p 0 ) 2 +Hro 2 +m 0 /(w 02 )+ ... ] . (3.200)
In this form we recognize that this structure is invariant under unitary

transformations on M<+> and M<->, independently. That is, we have a
partial symmetry. It is the group U4 x U4 •
Next, consider the baryons in their rest frame. Since the momentum transfer
is arbitrary small, the baryons retain a definite parity, and only even-parity
meson field components contribute to the interaction. If one regards all
the physical particles to be the outcome of some unknown dynamics, then
even the symmetry or partial symmetry revealed by noninteracting particles
has dynamical significance. We therefore make the partial symmetry
hypothesis-that the meson-baryon coupling also possesses the U4 sym-
metry. In the rest frame, the baryon field P,1 a 1 ,,202 ,,303 reduces effectively to
1 1
°1
°
'PA 1 A 2 A 3 with A =u, r, and y1 =y 2 ° =y 3 = + 1. The latter is totally
symmetrical in the four-valued indices A 1 , A 2 , and A 3 • Such a coupling is
3
2~p 'P* IX~l M,}+)tp, (3.201)
where each Ma<+> acts only on the corresponding indices. The scale factor is
fixed by the requirement that p0 couples universally to the isotopic-spin
3
density 'ff'* L (1/2)ra 'P giving the form
IX=l
-gp 0 • 'P*,L(1/2)ra'P. (3.202)
The nucleon part of the interaction (3.201) is
__§_ N* [_?_(n+H) · ur+H · u-m p 0 • r-3m w 0 ] N (3.203)

2m(J 3 P ro P ro '
where we have made use of the reductions for the nucleon
,L(1/2)ra~C1/2)r,
IX
L(1f2)aa~(1/2)u,
IX
(3.204)
where the first two are simply the assertion that ( 1/2)r and ( 1/2)a are the
isotopic spin and ordinary spin of the nucleon, respectively. The third one
is obtained by a simple calculation. The nN coupling
(3.205)
becomes, in the rest frame of the nucleon,
(3.206)
Comparison of (3.203) with (3.206) implies the connection
I (5/3)
lo= .J2
1.01 +0.01
Exp, strong interaction: - = 1.20±0.03
= 1.18 0.84±0.03
(
Exp, weak interaction: 1.18±0.03, (3.207)
since
g 1 lo
(3.208)
2mP = .J2 mn ·
One also notices that the substitutions
(3.209)
reproduce the correct electrical charges of the baryons. Thus (3.203) pre-
dicts the total magnetic moments for the nucleon
5 e 1 e
µ=---7:3+---. (3.210)
3 2mP 3 2mw
If the mass difference between w and p0 is ignored, we recover the famous
ratio
_ µP =~
µn 2
An assumed mass difference of 25 MeV gives the ratio 1.48. The experi-
mental value is 1.46. The absolute moment predictions are about fifteen
percent too small.
90 J. SCHWINGER
It is interesting to note the dynamical significance of the combination

1t +HPfor nuclear forces. Consider a nonrelativistic two-nucleon system.
The nuclear potential from these couplings is proportional to
When the two particles are close to each other, the n contribution alone
involves a term --(1/r 3 ), which is physically unacceptable. The particular
combination 1t +HP removes the overly singular potential. The nucleon not
only interacts with n but also with A 1 in the same manner. The much shorter
ranged A 1 coupling tends to increase the tensor force of the n coupling. This
suggests qualitively that the HP coupling could be stronger than indicated.
If so, the predictions about the absolute magnetic moments for the nucleon
will then be improved.
3.9 Concluding Remarks
As a phenomenological description the source theory has re-

produced and even improved, without unnecessary speculative assumptions,
the successful results of other formalisms. As a computational tool it has
also simplified many of the calculations. I hope in these lectures I have given
you enough of the general view, the method, and the spirit of the source
theory that you will think seriously of accepting the source theory for
what it is-a self-contained comprehensive approach to particle physics,
which is intended to replace the present competing methods by a simpler
synthesis.
These notes record the state of the art of Sourcery in July, 1967. For
more details and later developments, go to the Source papers published in
the Physical Review. They are:
"Particles and Sources": 152, 1219 (1966);

"Sources and Electrodynamics": 158, 1391 (1967);
"Gauge Fields, Sources and Electromagnetic Masses": 165~ 1714
(1968); 167, 1546 (1968);
"Sources and Gravitons": 173, 1264 (1968);
"Sources and Magnetic Charge": 173, 1536 (1968).
See also:
"Chiral Transformations": 167, 1432 (1968)
and these Physical Review Letters:
"Partial Symmetry": 18, 923 (1967);
"Photons, Mesons and Form Factors": 19, 1154 (1967);
"Radiative Corrections in fJ Decay": 19, 1501 (1967).

(Documents On Modern Physics) Schwinger Julian - Particles and Sources-Gordon and Breach, Science Publishers (1969) PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

(Documents On Modern Physics) Schwinger Julian - Particles and Sources-Gordon and Breach, Science Publishers (1969) PDF

Hochgeladen von

Copyright:

Verfügbare Formate

DOCUMENTS ON MODERN PHYSICS

ELLIOTT w. MONTROLL, University of Rochester

A. ABRAGAM, L'Effet Mossbauer

Additional Volumes in preparation

Notes by Tung-mow Yan

GORDON AND BREACH, SCIENCE PUBLISHERS

Library of Congress Catalog Card Number: 69-19390

Editorial office for Great Britain:

Gordon and Breach, Science Publishers Ltd.

Editorial office for France:

Gordon & Breach

Distributed in Canada by:

The Ryerson Press

Printed in Switzerland by City-Druck AG, Zurich

2.1 Combined System of Noninteracting Electrons and Photons. 29

3. STRONG AND WEAK INTERACTIONS 53

3.1 Low-Energy re+ N System 53

Research in particle physics presently falls in two main classi-

S(p) = f (dx)e-ipxS(x), (1.1)

and units h = c = 1 are used.

Unstable particles eventually decay and the decay process is a detection

1.3 Spinless Particles

To make the concept the basis of a quantitative theory, we consider

dw = (dp) _1_ po= +,Jp2+m2. (1.2)

We now define the source K in terms of the creation and annihilation

(1PIO_ )K =.J dwPiK(p), (1.3)

(O+ I 1P)K =.J dwPiK(-p), (1.4)

1~1(0+10_)Kl 2 + .l:1(0+11p)Kl 2 • (1.6)

(O+ 1o_y;;(O+ 1o_)K (0+ 1o_)K + _L(O+ 11pY (1PIO-Y

(O+ Io_ y = 1+(i/2) f (dx)(dx')K(x)L1+(x-x')K(x'), (1.8)

L1+(x-x') = Lf +(x' -x), (1.10)

This function is familiar. It is the solution of the differential equation

(m 2 - a2 )L1 + (x- x') = t5(x- x') (1.12)

= f dwPI K(p)l 2' (1.16)

I <o+ Io_ >K 12 = 1- LI <1pIo_>K1 2 , (1.17)

which is equivalent to (1.5) or (1.6). It can now be recognized that the

(O+ IO_)K =TI [t+(i/2)f (dx)(dx')Ka(x)Ll+(x-x')Ka(x')], (1.18)

a=l a=2 a=3

Ka= (K1 +K2)a'

f (dx)(dx')Ka(x)L1+(x-x')Kp(x')~O for r:t.=Ffi, (1.20)

the appropriate expression is

We have produced, under special circumstances, a situation in

in which we have defined

(O+ IO_)K = }2(0+ I {n})K1({n} 1o_)K2. (1.25)

This is accomplished by writing

<{n} Io_ > =

which clearly describe multiparticle states with nP = 0, 1, 2, · · · particles in

= 1(0+ IO_)Kl exp(}:IKPl 2 ).

1.4 An Application. Stimulated Emission

The original definition of a creation source refers to the creation

x exp[_L(iK1 P *iKP+ iKP *iK2 p)]. (1.32)

If K is a weak source, only terms linear in K are significant, and

x [ 1+ _L(iK1 P*iKP + iKP *iK2 p)]. (1.33)

(O+ I {n} )KiKP * = (O+ I {n} + 1Py.Jnp+ 1, (1.35)

III(O+ I {n} + 1P)K .Jnp+ 1iKP({n}10_ y

(O+ 10_ )CK>= I(O+ l{n}Y 1 ({n} I {n'}Y( {n'} IO-Y2 •

On comparison, we infer the matrix elements

( 1.36) can be rewritten as

({n}+ 1pl {n})K =.Jnp+ 1(1PIO_)K (1.38)

which gives the characteristic BE connection of spontaneous and stimulated

1.5 Spin-1 Particles. The Photon

The function Ll+(x-x') is an invariant one which depends only on

(O+ IO_)K = exp[(i/2)f (dx)(dx')K(x)Ll+(x-x')K(x')]

x exp[_L(iK1 P iKP+ iKP iK2 p)]. (1.32)

x [ 1+ _L(iK1 PiKP + iKP iK2 p)]. (1.33)

1 = l(O+ Io_),, 12 .L ITT<11p;. tp;.[ITT<11p;. tp).]*.