Phil Anderson And Gauge Symmetry Breaking

Edward Witten
Institute for Advanced Study
Princeton, NJ 08540

In this lecture, I am going to talk primarily about just one of the dozens of important papers
that Phil Anderson has written. This paper was written just a little over 51 years ago, in 1962.
I will set it in context by telling a little about the work that Julian Schwinger had done shortly
before, and also we will recall Phil’s earlier work on superconductivity that helped to set the stage.
Then we will take a trip through later developments that occurred through the rest of the 1960’s
and beyond, and we will conclude with some observations about the present. (I won’t talk about
some early precursors such as Stueckelberg in 1938, but later we will get to the model introduced
by Landau and Ginzburg in 1950.)
The title page of Phil’s paper, which is called “Plasmons, Gauge Invariance, and Mass,” and was
received on November 8, 1962, can be found in fig. 1. As one can see, Phil starts out by citing the
work of Julian Schwinger. The reference is to two very short papers that Schwinger had written,
also published in 1962. To understand Phil’s work, we should first take a look at Schwinger’s
contributions.
In the first paper (fig. 2), Schwinger argues somewhat abstractly that – in contrast to what we
are familiar with in the case of electromagnetism – gauge invariance does not imply the existence
of a massless spin one particle. A couple of things are worth noting here, apart from the fact that
Schwinger was forward-thinking to even ask the question. One is that Schwinger was motivated
by the strong interactions (there is no mention of weak interactions in the paper). The question
he asks in the first paragraph is whether the conservation law of baryon number could be a gauge
symmetry. There is an obvious problem, which is that we do not see a massless spin 1 particle
coupled to baryon number. So Schwinger asked whether it is possible for baryon number (or
something like baryon number) to be conserved because of a gauge symmetry, without the gauge
symmetry producing a massless spin 1 particle.
The answer that Schwinger proposes – even in the first sentence of the abstract – is that gauge
invariance does not necessarily imply the existence of a massless spin 1 particle if the coupling
is large. His idea is that there is no massless particle if there is a pole in the current-current
correlation function hJµ (q)J⌫ ( q)i. The role of strong coupling is supposed to be to generate this
pole. Thus, QED is weakly coupled, and has no such pole; Schwinger’s point of view is that in a
suitable gauge-invariant theory, strong coupling e↵ects might produce such a pole.

1
 One gets or
k 0k. o-=m' 4&„(+A) =-,' Tr(a "Aa, At).
We have also
(k. a)"' is the Hermitian square root of k a. ~ Tr(a "o.r) =g, ",
The relation between the unimodular matrices and O. „~=CO.„C ' or 0„=Co.„~C '.
the restricted Lorentz transformations is given by
2 For any 2 by 2 matrix M the relation CM~C '
Acr„A~ = A„"fr„, =M ' detM is an identity.

PH YSI CAL REVIEW VOLUME 130, NUMBER 1 1 A P R IL 1963

Plasmons, Gauge Invariance, and Mass

p. W. ANDERsoN
BdI TelePhoee Laboratories, MNrray IIN, ¹mJersey
(Received 8 November 1962)

Schwinger has pointed out that the Yang-Mills vector boson implied by associating a generalized gauge
transformation with a conservation law (of baryonic charge, for instance) does not necessarily have zero
mass, if a certain criterion on the vacuum fluctuations of the generalized current is satisfied. %'e show that
the theory of plasma oscillations is a simple nonrelativistic example exhibiting all of the features of Schwin-
ger's idea. It is also shown that Schwinger's criterion that the vector field m&0 implies that the matter
spectrum before including the Yang-Mills interaction contains m=0, but that the example of supercon-
ductivity illustrates that the physical spectrum need not. Some comments on the relationship between these
ideas and the zero-mass difhculty in theories with broken symmetries are given.

ECKXTLY, Schwinger' has given an argument is equivalent to the mass, while the 6nite density of
strongly suggesting that associating a gauge electrons leading to divergent "vacuum" current
transformation with a local conservation law does not fluctuations resembles the strong renormalized coupling
necessarily require the existence of a zero-mass vector of Schwinger's theory. In spite of the absence of
boson. For instance, it had previously seemed impossible low-frequency photons, gauge invariance and particle
to describe the conservation of baryons in such a conservation are clearly satisfied in the plasma.
manner because of the absence of a zero-mass boson In fact, one can draw a direct parallel between the
and of the accompanying long-range forces. ' The dielectric constant treatment of plasmon theory4 and
problem of the mass of the bosons represents the major Schwinger's argument. Schwinger comments that the
stumbling block in Sakurai's attempt to treat the commutation relations for the gauge 6eld A give us
dynamics of strongly interacting particles in terms of one sum rule for the vacuum fluctuations of A, while
the Yang-Mills gauge fields which seem to be required those for the matter field give a completely independent
to accompany the known conserved currents of baryon value for the Auctuations of matter current Since j. j
number and hypercharge. ' (We use the term "Yang- is the source for A and the two are connected by 6eld
Mills" in Sakurai's sense, to denote any generalized equations, the two sum rules are normally incompatible
gauge field accompanying a local conservation law. ) unless there is a contribution to the A rule from a free,
The purpose of this article is to point out that the homogeneous, weakly interacting, massless solution of
familiar plasmon theory of the free-electron gas ex- the 6eld equations. If, however, the source term is
emplifies Schwinger's theory in a very straightforward large enough, there can be no such contribution and
manner. In the plasma, transverse electromagnetic the massless solutions cannot exist.
waves do not propagate below the "plasma frequency, " The usual theory of the plasmon does not treat the
which is usually thought of as the frequency of long- electromagnetic field quantum-mechanically or discuss
wavelength longitudinal oscillation of the electron gas. vacuum Quctuations; yet there is a close relationship
At and above this frequency, three modes exist, in between the two arguments, and we, therefore, show
close analogy (except for problems of Galilean invari- that the quantum nature of the gauge field is irrelevant.
ance implied by the inequivalent dispersion of longi- Our argument is as follows:
tudinal and transverse modes) with the massive vector The equation for the electromagnetic 6eld is
boson mentioned by Schwinger. The plasma frequency
' J. Schwinger, p'A„= (k' —(o')A„(k, ai) = 4~ j„(k,~d).
Rev. 125, 397 (1962).
' T. D. Lee and Phys.
C. N. Yang, Phys. Rev. 98, 1501 (1955).
3
J. J. Sakurai, Ann. Phys. (N. Y.) 11, 1 (1961}. ' P. Nozieres and D. Pines, Phys. Rev. 109, 741 (1958).

Figure 1. The first page of Phil Anderson’s paper on gauge symmetry breaking, received
November 8, 1962.

In Schwinger’s second paper (fig. 3), he gives a concrete example of gauge invariance not implying
the existence of a massless spin 1 particle. The example is based on a remarkable exact solution,
not assuming weak coupling.
The model Schwinger solved was simply 1 + 1-dimensional Quantum Electrodynamics, with
electrons of zero bare mass. The action is
Z Z
1
I= d 2
x Fµ⌫ F µ⌫
+ d2 x ¯ iD
/ .
4e2
Nowadays this model – which is known as the Schwinger model – is usually solved simply and
understandably (but surprisingly) by “bosonization,” which converts it to a free theory (of the
gauge field Aµ and a scalar field ) with all the properties that Schwinger claimed. Schwinger’s
approach to solving it was more axiomatic.
The model is actually considered an important example that illustrates quite a few things, and
not only what Schwinger had in mind. For instance, it is also used to illustrate the physics of a
 3

PHYSICAL REVIEW VOLUME 125, NUMBER ANUARY 1, 1962

JUL&AN ScawxNGER
Harvard Urtiversity, Cambridge, 3dassachlsetts, and Urtiversity of Catiforrtia, los Angeles, California
(Received July 20, 1961)

It is argued that the gauge invariance of a vector field does not necessarily imply zero mass for an associ-
ated particle if the current vector coupling is sufficiently strong. This situation may permit a deeper under-
standing of nucleonic charge conservation as a manifestation of a gauge invariance, ~vithout the obvious
confIict ~ith experience that a massless particle entails.

&~OES the requirement of gauge invariance for a. Green's functions of other gauges have more compli-
vector Geld coupled to a dynamical current imply cated operator realizations, however, and will generally
the existence of a corresponding particle with zero lack the positiveness properties of the radiation gauge.
mass? Although the answer to this question is invari- Let us consider the simplest Green's function associ-
ably given in the affirmative, ' the author has become ated with the field A „(x), which can be derived from the
convinced that there is no such necessary implication, unordered product
once the assumption of weak coupling is removed. Thus
the path to an understanding of nucleonic (baryonic) (A„(x)A„(x'))
charge conservation as an aspect of a gauge invariance,
in strict analogy with electric charge, ' may be open for
(dP) .
a'vt* "&dm-s st+(p)b(p'+m')A„, (p),
the Grst time. (2or)s
One potential source of error should be recognized at
the outset. A gauge-invariant system is not the con- where the factor +st(p)8(p'+ m) enforces the spectral
tinuous limit of one that fails to admit such an arbitrary restriction to states with mass m& 0 and positive energy.
function transformation group. The discontinuous The requirement of non-negativeness for the matrix
change of invariance properties produces a correspond- A„„(p) is satisfied by the structure associated with the
ing discontinuity of the dynamical degrees of freedom radiation gauge, in virtue of the gauge-dependent asym-
Figure 2.
and of the operatorThe first ofrelations.
commutation Schwinger’s
No reliable two
metrypapers
between on
spacegauge
and time symmetry
(the time axis isbreaking.
specified
conclusions about the mass spectrum of a gauge- by the unit vector rt„):
invariant system can be drawn from the properties of
an apparently neighboring system, with a smaller in- —(P.~.+P.~,) (~P)+P.P
A„P(P)=B(m') g„.
variance group. Indeed, if one considers a vector Geld P'+(&P)'
coupled to a divergenceless current, where gauge
invariance is destroyed by a so-called mass term with Here B (m') is a real non-negative number. It obeys the
parameter mt, it is easily shown' that the mass spectrum sum rule
must extend below mp. The lowest mass value will
1= dm' B(m')
therefore
pl-IVSI CAL becomeREVIE arbitrarily small as mo OLUME approachesf28, -CEM»" y962
zero. Nevertheless, if m, o is exactly zero the commutation
relations, or equivalent properties, G,upon which this
conclusion is based become entirely different and the
IIIge Ig, VQ, I'laII&ewhich
~

dM«s
~~ is a full expression
&&
time commutation relations.
of all the fundamental equal-

argument fails. The Geld equations supply the analogous construction

SC~ING~R
If invariance under arbitrary gaugeIJ+rwr j
Ugg't)ers& ]3 s Cg~br
transformations
~

for Ce, ass(J&husettsexpectation value of current products

the vacuum
is asserted, one should distinguish sharply between (Received J&4' (j„(x)96)
9/2
j„(x')), in terms of the non-negative matrix
numerical gauge functions and operator gauge func-
tions, for the The
same quantum
.

possibility
various that gauges are e not
operator
so]ution of a one-djmensjonal l mo
footing. In each coordinate frame there
f eld on 11nP y a nonhero mass "(P)
can the =m'B(
p article j
ls 'll trated g"P').
')(P»Pby the e"act
&
—
The factor m' has the derisive consequence that m=0
is a unique operator gauge, characterized by three-
is not contained in the current vector's spectrum of
dimensional transversality k. d' that(radiation
the gauge gauge), for which
invariance o a vacuum unless ~=The latter determines
an d y~gOfluctuations. jn the B(m')
spect for
in a vector .
one vechas the standard e t necessariconstruction
operator
y requi Thus jt
ns&0, butis necessary
leaves unspeciGed a possible h f tà delta
p is tofunction
appear
space of positive norm, with
1 p article. a physical
n t is n probability as anisoa1 t datmassvaluein th e p h;cal
ysi spectum
interpretation. When
p y
the theory is formulated with the
contribution m=0,
d comments an give it is also necessary that
aid of vacuum expectation ution alarms valuesis of time-ordered B (m') = Bob(m')+Bi(m')
operator products, he the Green'sif unwor
functions, the freedom ~ ~
s(0) =0,
of '
h sical,
formal gauge transformation ' '
can be restored. '
The The that
such non-negative constant 80 is then Gxed by the sum
electrodynamics P t 1 di h
dm
' For example,
' gJ.Dilac 6eld Phys.
as 110 Rev.
no 75, 651 (1949).
associated masss rule,
' T. D. Lee
Schwinger,
C. N. isYang,
s(m'4)( ~,
Phys. Rev. 98, 1501 (1955).
' K. Johnson,andNuclear
P Phys. rather uniq(1961).
25, 435 1=Be+ 0 5$ dms Bi(m').
' J. Schwinger,
model for whic ich there
Phys. e 115,
Rev. is '121 exac
an (1959).
we have a pole at p '=0
0
solution. ' for only th en doo w =0,
GENERAL DISCUSSION '-O: b(p)-B,/(p' —z.), 0&B,&1.
Figure The 3. Green's
Schwinger’s
functionn ofsecond e' paper
an Abelian vectoron gauge
gauge symmetry
Under breaking, in which he introduced
these conditions,
field has the structure
and solved what is now called the Schwinger model. B(m') = B,S(m')+B, (m'),
= zr„„—
g„,(x,,x')')= ( zza)g—
( H)b(— x x'), — where
— d5$ 1

8 0 = 1+
0
where'„„ (p) is agaug-
a auge-dependent proj
ro ection matrix and s(m'))
0 tlS
aild B(m')
gauge theory ✓-angle, and '(')= afterd52 a small perturbation to give the electron a bare mass, it becomes
p+--' 0 BI (m') = [S(mz)/me j
)

a model of confinement
which is subject of charged
to the sum rule particles. However, although I was not in physics at the time, I
s(m")
suspect that Schwinger’s extremely dm' B (m').
short paper mystified many 1+I'of dm" his contemporaries.
—tS +[zrs(m His way of
0

solving the model probably was a little abstract, and the whole thingof probably seemed to revolve

e—
er retation ss(mz)
nz derives from
An alternative form of g(P Is '
Green's' function t o the vacuum trans-
thme relation of the Gre
around peculiarities of 1 + 1 dimensions. s(mz) formation function in the presence o f-
' i e+ (p' —z e
g (p) = p'+)). — J5$ anciently wea k ex e
xternal currents J „(x),
p'+me i
and the constant 'A' are non-
s
'
m'& (0~ 0)s= exp ', z (dx) -x'
x (dx') J~(x) g„„(x,x') J "(x')
d d' with the under-
=Oof
standing that the pole at 2'= o h
= exp lz (dp)~"(p)*B(p)~.(p) ,
——+ m — dies
s
0 m S 0 ' ' '
~ ~ ~
4

Schwinger’s concept was summarized in the last sentence of his first paper: “the essential point
is embodied in the view that the observed physical world is the outcome of the dynamical play
among underlying primary fields, and the relationship between these fundamental fields and the
phenomenological particles can be comparatively remote, in contrast to the immediate correlation
that is commonly assumed.” In other words, in general, there need be no simple relationship
between particles and fields – or in condensed matter physics, between bare electrons and nuclei
and the emergent quasi-particles that give a more useful description at long distances.
Schwinger is saying that the situation that prevails in QED – in which the electron field corre-
sponds to electrons, and the photon field corresponds to photons – results from the fact that this
theory is weakly coupled. In a strongly coupled theory, there might be no simple correspondence
between fields and particles. This was actually a very wise remark, probably putting Schwinger way
ahead of his contemporaries in particle physics. And it is at the core of the way we now understand
the strong interactions.
But Phil Anderson showed that Schwinger was actually not entirely correct about the specific
question he was writing about – how to have gauge invariance without a massless spin one particle.
To be more precise, everything that Schwinger said about strong coupling is true, but it is not the
whole story. As Anderson showed, a weakly coupled vector meson might also acquire a mass, and
here the essence of the matter is not strong coupling but symmetry breaking – the properties of
the vacuum. In fact, Phil expressed a point of view that is quite opposite to Schwinger’s, showing
that not just strong coupling but even quantum mechanics is irrelevant to the problem of how to
have gauge invariance without a massless spin one particle.
Phil’s paper is largely devoted to two examples from well-established physics. He begins by
saying that “the familiar plasmon theory of the free electron gas exemplifies Schwinger’s theory in
a very straightforward manner,” with the plasma frequency, below which electromagnetic waves do
not propagate, playing the role of the vector meson mass for Schwinger. He shows that the usual
analysis of screening in a plasma can be put in close parallel with what Schwinger had said in the
relativistic case. He also observes that the problem of screening in a plasma is usually understood
classically, without invoking quantum mechanics, and deduces that “the quantum nature of the
gauge field is irrelevant” to the question of how to have gauge invariance without a massless vector
particle.
The second part of Phil’s 1962 paper deals with an example that is even more incisive – su-
perconductivity. The background to this was provided by a series of three papers that Phil had
written in 1958 (shown in fig. 4 is the title page of the first of the three papers – which incidentally
is the one he referred to in 1962). In these papers, Phil had analyzed gauge invariance and the
fate of the “Goldstone” boson (the term is ahistorical as Goldstone had not yet formulated his
relativistic theorem) in the BCS theory of superconductivity, showing that this mode combines
with ordinary photons to become a gapped state of spin 1. Thus the electronic state of an ordinary
BCS superconductor is truly gapless.
In the 1962 paper, Phil explains cogently how superconductivity illustrates the phenomenon
described by Schwinger, in a context in which the gauge field is weakly coupled and the physics is
well understood. Much of this paper reads just like what one would explain to a student today.
 5

PH YSI CAL REVIEW VOLUME 110, NUM HER 4 MAY 15, 1958

Coherent Excited States in the Theory of Superconductivity: Gauge

Invariance and the Meissner Effect
P. W. ANDERSoN
Bell Telephone Laboratories, Murruy Hill, Sew Jersey
(Received January 27, 1958)

We discuss the coherent states generated in the Bardeen, Cooper, and Schrieffer theory of supercon-
ductivity by the momentum displacement operator pu= Z exp(iQ r„). Without taking into account plasma
e8ects, these states are like bound Cooper pairs with momentum AQ and energies lying in the gap, and they
play a central role in the explanation of the gauge invariance of the Meissner effect. Long-range Coulomb
forces recombine them into plasmons with equations of motion unaffected by the gap. Central to the argu-
ment is the proof that the non-gauge-invariant terms in the Hamiltonian of Bardeen, Cooper, and SchrieEer
have an effect on these states which vanishes in the weak-coupling limit.

I. INTRODUCTION instead of zero. The total operator applied to +, leads

V CKINGHAM' has questioned whether an energy- to a linear combination of such states, which can be
as that of thought of as equivalent to a Cooper bound state' of a
gap model of superconductivity„such
Bardeen, Cooper, and Schrieffer, ' can explain the pair of electrons with nonzero momentum, superimposed
Meissner effect without violating a certain identity on the B.C.S. ground state.
derived by Schafroth' on the basis of gauge invariance, Our discussion of these problems is based on the
and by Buckingham using essentially an f-sum rule. following physical picture: any transverse excitation
This identity is what causes the insulator, which also involves breaking up the phase coherence over the whole
has an energy gap, to fail to show a Meissner effect; Fermi surface of at least one pair in the superconducting
thus, Buckingham and Schafroth4 argue, a proof of ground state, and so involves a loss of attractive binding
gauge invariance lies at the core of the problem of energy.
This causes the Meissner eGect. Longitudinal
superconductivity, especially since the Hamiltonian excitations, however, such as those generated by pQ, do
used in B.C.S. is not gauge-invariant. not break up phase coherence in the superconducting
Bardeen' argues that the matrix elements and energy state, and so their energies involve only kinetic energy,
states involved in the gauge problem bring in coherent or electromagnetic energy when plasma effects are
Figure 4. The first of Anderson’s three papers of 1958
included. Thus on gauge invariance
longitudinal in the
and transverse BCS
excitations
excitations which will be strongly coupled to the plasma
model of superconductivity.
modes, a high-frequency phenomenon presumably un- are different in the superconductor, in a sense in which
altered by superconductivity. Vnfortunately, while we they are not in either the metal or the insulator, and it
6nd that this is indeed exactly the situation, the insula- turns out to be this difference which allows a gauge-
tor also often“the the invariant explanation of the Meissner effect.
For example: way isplasma
has normal nowmodes.
openThus,
for awhile
degenerate We
vacuum
proceed
theory of the Nambu type without
further in two stages. First, we discuss
B.C.S. calculation in the London gauge is probably
any difficulties involving
entirely correct, eitheronzero-mass
and justi6able Yang-Mills
physical grounds, it the gauge bosons or
Gctitious problem zero-mass
in which the onlyGoldstone
plasma effect bosons.
throws little light on the basic differences between the is the screening of the long-range repulsion. In this stage
Thesethree
two cases —
typesinsulator,
of bosons seem capable of ‘canceling
metal, and superconductor. gaugeeach other requires,
invariance out’ and andleaving
we indeedfinite mass
find, that thebosons
only.” He goesnoticed
We also on: “It
that is
the not at all
operator clear
which that inthestates
is central way for a Sakurai theory [with baryon number
+Q=PQ+0
the gauge problem as well as the plasma theory, (3)
as a gauge symmetry] is equally uncluttered. The only mechanism suggested by the present work
pg — P„exp(iQ r„) have energies in the energy gap and proportional to Q'.
(of course, we have notk, discussed tr Ck+Q,
tr Ck,a&
non-Abelian gauge perfectly
In a groups) for giving the
gauge-invariant gauge field
theory, their energymass is
would be just the kinetic energy
the degenerate vacuum property:
has another interesting type of itstheory,
separate in which the original symmetry is not manifest in the
compo-
nents ck+Q, *ck, , when applied to the B.C.S. ground- Eo = (O'Q'/2m) (2eg/3m. es), (4)
observable domain. Therefore, it needs
state wave function 4 „create excited pairs of electrons
to be demonstrated that the necessary conservation laws
but we And a small correction going to zero in the weak-
can bek~, maintained.”
ks with momentum In other words, as one would
pairing say limit.
coupling today, if baryon
Equation number
(4) follows is same
from the f and
gauged
sum rule which leads to gauge invariance.
kr+ks=
spontaneously broken, then baryon number will notThere
Q (2) be conserved in nature.
is a fundamental difference, which we demon-
' M. J. Buckingham, Nuovo cimento 5, 1763 (1957). the
And let' us look
Bardeen, Cooper, at
and the endPhys.
Schrie6er, of the paper
Rev. 106,
strate, between
(fig. 5): “I should like to close
162 (1957);
ways in which with one finalandremark
superconducting
108, 1175 (1957). The latter we call B.C.S., and we shall follow its normal substances satisfy this sum rule. Both normal
on thenotation
' Goldstone theorem. This theorem was initially
as far as possible. metals and conjectured,
insulators (leavingone out presumes, because of
the rather confusing
M. R. Schafroth, Helv. Phys. Acta 24, 645 (1951). effects of long-range Coulomb forces which can be
4 M. R. Schafroth
the solid-state I to
analogs, via the work of Nambu and Anderson” (the reference here is to Nambu’s
(private communication). am indebted
G. Wentzel for an elegant presentation of these questions in a studied later) satisfy this rule with ordinary excitations
series of discussions, to which M. Lax and C. Herring also con- in such more familiar optical sum rule—
a way that the paper
work on spontaneously broken chiral symmetry and
tributed. to Anderson’s whose title page appears
' J. Bardeen, Nuovo cimento 5, 1765 (1957). L. N. Cooper, Phys. Rev. 104, 1189 (1956).
in fig. 4). He goes on to give various examples, both old (spin waves and phonons) and new
(superconductors and superfluids). And then he writes, “It is noteworthy that in most of these
cases, upon closer examination, the Goldstone bosons do indeed become tangled up with Yang-Mills
gauge bosons and do not in any true sense really have zero mass. Superconductivity is a familiar
example, but a similar phenomenon happens with phonons; when the phonon frequency is as low
 but not for the special
a" type of vacuum. En vacuum theory of the Nambu type' without any
emains localized in the diS.culties involving either zero-mass Yang-Mills gauge
all of the charge is bosons or zero-mass Goldstone bosons. These two
ding to the fact, well types of bosons seem capable of "canceling each other
s, that all the charge out" and leaving 6nite mass bosons only. It is not at
a plasma is actually 6on all clear that the way for a Sakurai' theory is equally
ervation of particles, if uncluttered. The only mechanism suggested by the
maintained. Note that present work (of course, we have not discussed non-
the case of infinite" Abelian
PHYSICALgauge groups) for giving the gauge GeldVOLUME
REVIEW mass 130, NUMBER 1 1 APRIL 1963
e the infinity in the is the degenerate vacuum type of theory, in which the
occur at p'=0. original symmetry is not manifest in the observableP. W. ANDERSON
larizable vacuum or of domain. Therefore, it needs to be demonstrated that
nergy experiment, and the necessary conservation laws can be maintained.
one, seems capable of I should like to close with one 6nal remark on the
e vacuum polarizability Goldstone theorem. This theorem was initially con-
we conclude that the jectured, one presumes, because of the solid-state
demonstrating Schwing- 442442 via the work of Nambu" and of P.
analogs, P.%.%.ANANDERSON
Anderson. DERSON "
ome circumstances the The theorem states, essentially, that if the Lagrangian
on need not have zero possessesa continuous
possesses a continuoussymmetrysymmetrygroup groupunder
underwhichwhich mass. ""
mass.Utiyama"
Utiyama"andand Feynmanhave
Feynman have pointedoutout
pointed thatthat
he short range of forces the' the
J.ground
groundor vacuum
Goldstone, Nuovo
or state
Cimento
vacuum state19,not
is is154 invariant,thatthat
(1961).
invariant,
not state gravityis also
state gravity is also a Yang-Mills
a Yang-Millsfield. It Itis isan anamusing
field. amusing
might interpret without J.
is, is,
Goldstone, A. Salam, and S. Weinberg, Phys. Rev. 127,
therefore, degeneratewithwith
therefore,degenerate other other groundstates.
ground This
states. observationthatthat
This observation three
the the threephonons
phononsplusplus twotwo gravitons
gravitons
965 (1962).
viewpoint of W. Thirring, 9 Y.
implies
implies
Nambua zero-mass
aandzero-mass boson.
boson.Thus,
G. Jona-Lasinio, Phys.Thus, the the
Rev. solid
122, 345 crystal are are
solidcrystal
(1961). justjust
enough
enough componentsto make
components to make the the
up up appropriate
appropriate
cs (Academic Press Inc. , ' Y. Nambu, Phys. Rev. and117, 648 (1960). invariance,
rotational a finite-mass
violates
»violates translational
translational and rotational invariance,andand tensor tensorparticle which
particle whichwouldwouldbe be requiredfor for
required a finite-mass
P. W. Anderson, Phys. Rev. 110, 827 (1958).
possesses
possessesphonons; liquidhelium
phonons;liquid heliumviolates a certain
violates(in (in a certain graviton.
graviton.
sensesenseonly, of ofcourse)
course)gauge invariance,
invariance,andand possesses Spin waves are are
alsoalso knownto interact
known to interactstronglystronglywithwith
a longitudinal
only,
a longitudinal phonon;
gauge
phonon;ferro-magnetism
ferro-magnetism violates
possesses Spin
violatesspinspin magnetostatic
waves
magnetostatic forces forcesat at wavelengths, for for
veryverylonglongwavelengths, ""
rotation symmetry,andandpossesses
rotationsymmetry, possessesspinspin super- rather
waves;super-
waves; rathermore moreobscure
obscureandand satisfactory
lessless satisfactoryreasons.reasons.WeWe
conductivityviolates
conductivity violatesgauge
gauge invariance,
invariance,andand would
would havehave conclude,
conclude,then, thatthat
then, the the Goldstonezero-mass
Goldstone zero-massdifFiculty
difFiculty
a zero-mass
a zero-masscollective mode
collective mode in the
in the absence
absence of long-range is not
of long-range a serious
is not a serious because
one,one, because we wecancan probablycancel
probably cancelit it
Coulomb forces. against
off off against an anequal
equal Yang-Mills
Yang-Mills zero-mass
zero-mass problem.
problem.
Coulomb forces.
WhatWhat is not clear
is not clear on on
yet,yet, other
the the hand,
other hand, is whether
is whether it isit is
lt lt is noteworthythatthatin in
is noteworthy most mostof of these
thesecases,
cases,uponupon possible to describe a
possible to describe truly a trulystrongstrongconservation
conservationlawlaw
closer examination, the theGoldstone
closerexamination, Goldstonebosons bosonsdo do indeed
indeed suchsuchas as thatthatof of baryonswithwitha gauge
baryons a gaugegroup groupandanda a
become
become tangled
tangledup upwithwithYang-Mills
Yang-Millsgauge gaugebosons
bosons and,and, Yang-Mills
Yang-Millsfieldfield havingfinite
having mass.
finite mass.
thus,thus,do donotnotin in
anyanytruetruesensesense reallyhave
really havezerozero mass. I should
mass. I shouldlikeliketo tothank thankDr.Dr. JohnJohnR. R. Klauderfor for
Klauder
Superconductivity is ais familiar
Superconductivity a familiarexample,example,butbut a similar
a similar valuable
valuableconversations
conversationsand,and, particularly,for for
particularly, correcting
correcting
phenomenon
phenomenonhappens happenswithwith phonons;
phonons;when when the the phonon some
phonon someserious misapprehensions on on
seriousmisapprehensions my my part, part,andandDr.Dr.
frequencyis as
frequency lowlow
is as as asthe thegravitational
gravitational plasmaplasmafre-fre- John John Taylor
C'. C'. Taylor for for
calling
callingmy my attention
attentionto to Schwinger's
Schwinger's
quency, (4~Gp)"-
quency, (wavelength 10'10'km kmin innormal
(4~Gp)"-(wavelength normal work. work.
matter)
matter)there
thereis ais phonon-graviton
a phonon-graviton interaction: in that "J."J.
interaction:in that H. Jeans, Phil.Phil.
H. Jeans, Trans. Roy.
Trans. Soc.Soc.
Roy. London 101,101,
London (1903).
157157 (1903).
~ R.~ R. Utiyama,Phys.
Utiyama, Phys.Rev.Rev.
101,101,
1597 (1956);
1597 R. R.
P. Feynman
P. Feynman
case,
case, becauseof of
because the the
peculiar of of
peculiarsignsign the the
gravitational
gravitational (unpublished). (1956);
(unpublished).
interaction, leadingto to
interaction, leading instability rather
instability 6nite "L."L.
ratherthanthan6nite R. Walker,
R. Walker,Phys. Rev.Rev.
Phys. 390390
105,105, (1957).
(1957). !
!
Figure
YS
PH PH ICAL
YS 5. The concluding
ICALREVIEW
REVIEW partVOLUME
ofVOLUME
Anderson’s
130,
130, 1962
NUM paper
BER1 1 on gauge symmetry
BER
NUM A P ARbreaking.
1963
LIL
PI R 1963

Constructionof of
Construction Invariant
InvariantScattering Amylitudes forfor
ScatteringAmylitudes Arbitrary
Arbitraryspina
spinaandand
as the gravitational plasma Analytic
Analytic Continuation
frequency,
Continuation (4⇡G) in in 1/2
TotalTotal AngularMomentum*
(wavelength
Angular ⇠ 104 km in normal matter) there
Momentum*
is a phonon-graviton interaction: Asim O. O.
Asim BARUT,in t that
BARUT, IvAN case,
t IvANMUzrNzcH,
MUzrNzcH, ) Awo
because) AwoDAvm ofN. the
DAvm N.
WiLLiAMS peculiar sign of the gravitational
WiLLiAMS
Lawrence
Lawrence Radiation
Radiation Laboratory,
Laboratory,University
University of California,
of California,Berkeley,
Berkeley,California
California
interaction, leading to instability rather than9 November
(Received
(Received finite
9 November mass. 1962)Utiyama and Feynman have pointed out
1962)

that gravity is From alsoFrom agroup-theoretical

Yang-Mills
group-theoretical field. invariant
considerations,
considerations, Itinvariant
is an amusing
scattering
scattering amplitudes observation
amplitudes for for
two-body
two-body thatofthree
reactions
reactions particles
of particlesphonons plus two
arbitrary
withwith spinsspins
arbitrary nonzero
and and nonzeromasses are are
masses constructed
constructed in various
in variousforms, including
forms, includinghelicity amplitudes
helicity amplitudes and and
gravitons are amplitudes just enough
amplitudes free free ofcomponents
of kinematical
kinematicalsingularities. to make
They
singularities. Theyareup linear
are the appropriate
combinations
linear combinationsof spin of spintensor
basis particle
functions
basis functions withwith which would be
scalar coefficients.
scalar coefficients.In the
In the
process
processof construction
of construction the the
PauliPauli
spinspinmatrices are are
matrices generalized
generalizedfor for
arbitrary
arbitraryspin.spin.
required for a finite-mass
On On the basis of a graviton.”
the basis of a Mandelstam So the for
representation
Mandelstam representation for answer
the scalar
the scalar tocoef6cients,
the the
coef6cients, question
the
unique of
analytic
unique who
analytic first tried to discuss a
continuation
continuation
of the
of the amplitudesin total
amplitudes angular
in total angular momentumis obtained.
momentum is obtained. Possible kinematical
Possible singularitiesof the
kinematicalsingularities scalar
of the scalar
gravitational analog of gauge
amplitudes
amplitudes at the
at boundary
the symmetry
boundary of the
of the
physical breaking
region
physical are is that Anderson did.
discussed.
region are discussed.

What happenedI. next? In 1964, Peter Higgs wrote two papers on gauge invariance with massive
I. INTRODUCTION
INTRODUCTION structstructthe the invariant
invariantamplitudes
amplitudesin in termstermsof of the theirre-irre-
vector particles
HEHE basic
basic in relativistic
quantities
quantities of ofS-matrixphysics.
S-matrix theory
theory Like theAnderson,
are are the ducible but
ducibleunitaryunitaryunlike Schwinger,
representations
representations of ofthe the his starting point is
inhomogeneous
inhomogeneous
(5 (5aproper properI.orentzI.orentzgroup, based
basedon ona atwo-component
spontaneous Lorentz-invariant
Lorentz-invariant scattering
breaking of scattering
symmetry. matrix
matrixelements
Thiselements
was few formalism
spinor years after ' 'group,Goldstone’s two-component
theorem and particle
functions),
functions),which whichdependdependon on the thespinsspinsandandtypes typesof of spinor formalism
Although
Although thethe invariant
invariant scalar
scalaramplitudes for
physicists were more familiar with this concept. the
incoming
incoming and andoutgoing
outgoing particles
particles and andon the
on mass
the massshell
shell In the first paper, he explains somewhatwhich amplitudes for
which abstractly,
valuesof of
values theirtheirfour-momenta.
four-momenta. From From the the S functions, theMandelstam
S functions, Mandelstamrepresentation
representationis expectedis expected to tobe be
validvalid
in a invariant
language similar
invariant scattering toamplitudes
scatteringamplitudes Schwinger’s, (M(M functions)why that
functions) Goldstone’s
that have have been theorem is not
beenknown
known for some
for some valid
time in
time in inthesimpler
the thesimpler
casecases of a gauge
cases
such as those of the
have simplertransformation
havesimpler transformation properties propertiesandandthatthatare are such as those of the pion-nucleon'
pion-nucleon'andand nucleon-nucleon4
nucleon-nucleon4
symmetry.
expected Higgs’s
expectedto to be be paper
freefreeof of that really
kinematical
had greater
singularitiescancan
kinematicalsingularities be be
impact was the second one in which he described
defined. ' A' general procedure has been to con- Theory
Theory $W.$W.A. Benjamin,
A. Benjamin, Inc.Inc.
, New, NewYork York{to {to
be published)
be published) j. j.
a concrete (and weakly coupled) model that everyone
defined. A general procedure has been given
given to con- ~ ~ A. O. O. could
A. Barut, Phys.
Barut, understand
Phys.Rev.Rev. 321321
127,127, and
(1962).
(1962).
also introduced the
*Work*Work donedone
underunderthe the
auspices
auspices of the
of theU. U.S. Atomic
S. Atomic Energy
Energy
3 G.3 G.F. Chew,
F. Chew, M. M.L. Goldberger,
L. Goldberger,F. E. F. E. Y.
Low, Low, and Y.
and Nambu,
Higgs Commission.
particle. Let us take a look at this paper Phys.
Commission. (fig.
Phys. 6).
Rev.Rev. 1337
106,106, {1957).
1337 {1957).
Nambu,
t Present
t Presentaddress:
address: Universityof Colorado,
University of Colorado, Boulder,
Boulder, Colorado. 'M.'M.
Colorado. L. L. Goldberger,M. M.
Goldberger, T. T. Grisaru,S. W.
Grisaru, S. W.MacDowell,
MacDowell,and and
Higgsf explains at the outset that the phenomenon
Present
f address:
Present address: University
University of Washington,
of Washington, Seattle, Washing-
Seattle, Washing- D. Y.
D. Y.
Wong,ofD.a gauge120,boson
Wong, Phys. Rev.
Phys. Rev. 2250
120, 2250 acquiring
(1960),
(1960),
(referred toahereafter
(referred tomass
hereaftervia sym-
ton.ton. as GGMW);
as D. E. E. B.
' H.' P. P. Stapp,
H.Stapp, Phys. Rev.Rev. 2139
125,125, (1962); Lectlres on S-Matrix
GGMW); Amati,
Amati, Leader,
Leader,and and B.
Vitale, Nuovo
Vitale, Nuovo Cimento
Cimento
Phys. 2139 (1962); Lectlres on S-Matrix 1?, 1?, 68 68 (1960).
(1960).
metry breaking “is just the relativistic analog of the plasmon phenomenon to which Anderson has
 7

VoLUME 1$, NUMBER 16 PHYSICAL REVIEW LETTERS 19 OcTQBER 1964

BROKEN SYMMETRIES AND THE MASSES OF GAUGE BOSONS

Peter W. Higgs
Tait Institute of Mathematical Physics, University of Edinburgh, Edinburgh, Scotland
(Received 31 August 1964)

In a recent note' it was shown that the Gold- about the "vacuum" solution y, (x) =0, y, (x) = y, :
stone theorem, ' that Lorentz-covaria. nt field
theories in which spontaneous breakdown of s "(s (np 1 )-ep 0A ) =0, (2a)
symmetry under an internal Lie group occurs
contain zero-mass particles, fails if and only if
the conserved currents associated with the in-
(&'-4e, 'V" (y, ')f(&y, ) = 0, (2b)
ternal group are coupled to gauge fields. The
purpose of the present note is to report that, s
V
r"'=eq (s"(c,p, )
0 1
ep 0A -t. (2c)
p,
as a consequence of this coupling, the spin-one
quanta of some of the gauge fields acquire mass; Equation (2b) describes waves whose quanta have
the longitudinal degrees of freedom of these par- (bare) mass 2po(V"(yo'))'"; Eqs. (2a) and (2c)
ticles (which would be absent if their mass were may be transformed, by the introduction of new
zero) go over into the Goldstone bosons when the var iables
coupling tends to zero. This phenomenon is just fl =A -(ey 0) '8 (n, (p 1'),
the relativistic analog of the plasmon phenome- p. p,
non to which Anderson' has drawn attention:
that the scalar zero-mass excitations of a super-
G
IL(. V
=8 B
p. V
-BBp, =F
V LL(V

conducting neutral Fermi gas become longitudi-

into the form
nal plasmon modes of finite mass when the gas
is charged. t" Pv +e 2 y 2 8 =0.
The simplest theory which exhibits this be-
8 B =0, 8 (4)
v 0
havior is a gauge-invariant version of a model
used by Goldstone' himself: Two real' scalar Equation (4) describes vector waves whose quanta
fields y„y, and a real vector field A interact have (bare) mass ey, . In the absence of the gauge
through the Lagrangian density field coupling (e =0) the situation is quite differ-
ent: Equations (2a) and (2c) describe zero-mass
2 2 scalar and vector bosons, respectively. In pass-
L =-&(&v ) -@'7v )
1 2 ing, we note that the right-hand side of (2c) is
—V(rp 2 + y 2 ) ~ -P' V JL(, just the linear approximation to the conserved
1 2 P, v current: It is linear in the vector potential,
gauge invariance being maintained by the pres-
ence of the gradient term. '
where
Figure 6. Peter
V p Higgs’s
=~ p -eA second2'
paper on gaugeWhen
symmetry breaking,
one considers in which
theoretical models he
in introduced
1 1
jL(, p,
in eqn. (1) what particle physicists know as thewhich abelian Higgs
spontaneous model.of symmetry under
breakdown
a semisimple group occurs, one encounters a
p2 +eA {p1' variety of possible situations corresponding to
the various distinct irreducible representations
drawn attention: that the scalar zero-mass excitations of a superconducting neutral Fermi gas be-
to which the scalar fields may belong; the gauge
F =8 A -BA field always belongs to the adjoint representa-
come longitudinal plasmon modes of finite mass when the gas is charged.” He then goes on, in the
PV P, V V
tion. ' The model of the most immediate inter-
e is a dimensionless coupling constant, and the
next paragraph (bottom left in fig. 6), to write down his model. est is that in which the scalar fields form an
metric is taken as -+++. I. is invariant under octet under SU(3): Here one finds the possibil-
simultaneous gauge transformations of the first ity of two nonvanishing vacuum expectation val-
Higgs’s model was simply a relativistic version of the model that Landau and Ginzburg had
kind on y, + iy, and of the second kind on A ues, which may be chosen to be the two Y=0,
introduced to describe superconductivity.) (Neither Higgs nor any of the authors I have mentioned
Let us suppose that V'(cpa') 0, V"(&p, ') 0; then
= I3=0 members of the octet. There are two
spontaneous breakdown of U(1) symmetry occurs. massive scalar bosons with just these quantum
cited Landau and Ginzburg. Higgs describes his model as a gauge-invariant version of a model of
Consider the equations [derived from (1) by numbers; the remaining six components of the
treating ~y„ay„and A & as small quantities] scalar octet combine with the corresponding
Goldstone.) The Landau-Ginzburg model can be deduced from the action
governing the propagation of small oscillations components of the gauge-field octet to describe
!
508 Z X D 2
1 D 1
3
I = d x dt Fµ⌫ F + i ¯
µ⌫
(¯ 2 2
a ) .
4 Dt 2m Dxi
i
Higgs’s model, which particle physicists call the abelian Higgs model, is the same thing (apart from
minor rescalings) with the kinetic energy of the scalar field made relativistic
Z ✓ ◆
3 1 µ⌫ ¯ µ ¯ 2 2
I = d x dt Fµ⌫ F + Dµ D ( a ) .
4
The models are the same except that the abelian Higgs model is quadratic rather than linear in
time derivatives.
When Dirac (for spin 1/2) or Klein and Gordon (for spin 0) made the Schrodinger equation
relativistic, they introduced an extra degree of freedom (the antiparticle). Something somewhat
8

similar happens in making the Landau-Ginzburg model relativistic. There is an extra degree of
freedom because and ¯ become independent rather than being canonically conjugate, as they
are in Landau-Ginzburg theory. In Landau-Ginzburg theory, if we ignore the gauge fields, there is
one spin 0 particle, the Goldstone boson, but if we include the gauge fields, then – as Anderson
had explained in the more sophisticated context of BCS theory – it becomes part of a massive spin
1 particle. In the abelian Higgs model, there is a second and massive spin 0 mode – this is the
fluctuation in the magnitude of , which is now usually called the Higgs particle.
Actually, although there is not quite a Higgs particle in usual models of superconductivity – or
in superconducting phenomenology – there is a close cousin. In a superconductor, there are two
characteristic lengths, the penetration depth and the coherence length. They are described in the
Landau-Ginzburg and BCS models of superconductivity and are measured experimentally. (The
di↵erence between a Type I and Type II superconductor has to do with which is bigger.) These are
the analogs of the gauge boson mass and the Higgs boson mass in particle physics. Relativistically,
the rate at which the field decays in space is related to a particle mass, but nonrelativistically
there is no reason for this to happen, and in the Landau-Ginzburg model it doesn’t, in the case of
the correlation length. The Landau-Ginzburg model and the abelian Higgs model are completely
equivalent for static phenomena since they coincide once one drops the time derivatives.
Similar ideas were developed by others at roughly the same time as Higgs. We will just take
a quick look. The paper of Englert and Brout is in fig. 7. This paper is notable for considering
symmetry breaking in non-abelian gauge theory, while previous authors had considered the abelian
case, sometimes saying that this was for simplicity. “The importance of this problem,” they say,
“lies in the possibility that strong-interaction physics originates from massive gauge fields coupled
to a system of conserved currents,” for which they refer to Sakurai. Soon after was the paper of
Guralnik, Hagen, and Kibble (fig. 8), followed by Migdal and Polyakov (fig. 9). The title of Migdal
and Polyakov, “Spontaneous Breakdown of Strong Interaction Symmetry and Absence of Massless
Particles,” shows that they, too, were thinking of the strong interactions as the arena in which
gauge symmetry breaking might play a role.
From here, let us move forward to Kibble in 1966 (fig. 10). After mentioning the example
of superconductivity, Kibble writes “The first indication of a similar e↵ect in relativistic theories
was provided by the work of Anderson, who showed that the introduction of a long-range field,
like the electromagnetic field, might serve to eliminate massless particles from the theory. More
recently, Higgs has exhibited a model which shows explicitly how the massless Goldstone bosons
are eliminated by coupling the current associated with the broken symmetry to a gauge field.” He
then goes on to discuss some important details of symmetry breaking in nonabelian gauge theory.
He explains how it it is possible to have partial breaking of nonabelian gauge symmetry, with some
gauge mesons remaining massless. Like Higgs and some of the others, he does not really say what
the physical application is supposed to be, but he does remark that nature has only one massless
vector particle – the photon – but various (in some cases approximate) global symmetries. At least
this was on the right track.
 9

Figure 7. Brout and Englert on gauge symmetry breaking.

Figure 8. Guralnik, Hagen, and Kibble on gauge symmetry breaking.

The next milestone, of course, was that in 1967-8, Weinberg and Salam actually found what
spontaneous gauge symmetry breaking is good for in particle physics. (Their model was a gauge-
invariant refinement of an earlier model by Glashow. That model had W and Z mesons, but lacked
the relationship between their masses and couplings that follows from the spontaneous symmetry
breaking mechanism introduced by Weinberg and Salam.) However, since we have already looked
at quite a few original papers, let us jump ahead to Weinberg’s Nobel Prize address in 1979.
In the passage copied in fig. 11, Weinberg explains quite vividly how – like everyone else in
the 1960’s, it seems – he started by assuming that gauge symmetry breaking was supposed to be
applied to the strong interactions. His detailed explanation actually makes interesting reading. To
10

!
!
Figure 9. Migdal and Polyakov on gauge symmetry breaking.

help the reader understand this passage, I will make the following remarks. If baryon number is
a gauge symmetry, what is the gauge meson? The lightest hadronic particle of spin 1 with the
appropriate quantum numbers is the ! meson, or the meson if one includes strange particles. So
one might think of one of those as a gauge meson. But if baryon number is a gauge symmetry,
perhaps isospin symmetry is a gauge symmetry also. In this case, the lightest candidates for the
massive gauge particles are the ⇢ mesons. But bearing in mind that isospin symmetry is part of a
spontaneously broken SU (2) ⇥ SU (2) chiral symmetry, perhaps there is also an axial vector triplet
of massive gauge mesons; the A1 is the lightest candidate. All this is quite alien to present-day
thinking, and as Weinberg explains, there were a lot of problems: massless ⇢ mesons, or no pions, or
explicit (rather than spontaneous) breaking of gauge invariance and therefore no renormalizability,
depending on what assumptions he made.
Then enlightenment dawns. Weinberg explains that “At some point in the fall of 1967, I think
while driving to my office at M.I.T., it occurred to me that I had been applying the right ideas to
the wrong problem. It is not the ⇢ mesons that is massless: it is the photon. And its partner is not
the A1 , but the massive intermediate boson, which since the time of Yukawa had been suspected to
be the mediator of the weak interactions. The weak and electromagnetic interactions could then be
described in a unified way in terms of an exact but spontaneously broken gauge symmetry.... And
 11

PHVSI CAL R EVIEAV VOLUM'E 155, NUM B ER 5 2 5 M A R C I-I 1 967

Symmetry Breaking in Non-Abelian Gauge Theories*

T. W. B. KIBBLE
Department of Physics, Imperial College, London, England
iReceived 24 October 1966l

According to the Goldstone theorem, any manifestly covariant broken-symmetry theory must exhibit
massless particles. However, it is known from previous work that such particles need not appear in a rela-
tivistic theory such as radiation-gauge electrodynamics, which lacks manifest covariance. Higgs has shown
how the massless Goldstone particles may be eliminated from a theory with broken U(1) symmetry by
coupling in the electromagnetic field. The primary purpose of this paper is to discuss the analogous problem
for the case of broken non-Abelian gauge symmetries. In particular, a model is exhibited which shows how
the number of massless particles in a theory of this type is determined, and the possibility of having a
broken non-Abelian gauge symmetry with no massless particles whatever- is established. A secondary
purpose is to investigate the relationship between the radiation-gauge and Lorentz-gauge formalisms.
The Abelian-gauge case is reexamined, in order to show that, contrary to some previous assertions, the
Lorentz-gauge formalism, properly handled, is perfectly consistent, and leads to physical conclusions
identical with those reached using the radiation gauge.

L INTRODUCTION In either case the theorem is inapplicable in the presence

of long-range forces, essentially because the continuity
'HEORIES with spontaneous symmetry breaking
equation B„j&=0 no longer implies the time indepen-
(in which the Hamiltonian but not the ground dence of expressions like J'd'x [j'(x),g(0)], since the
state is symmetric) have played an important role in
relevant surface integrals do not vanish in the limit of
our understanding of nonrelativistic phenomena like
infinite volume. (In the relativistic case, the theorem
superconductivity and ferromagnetism. Many authors,
does apply if we use the Lorentz gauge; but then it tells
beginning with Nambu, ' have discussed the possibility
us nothing about whether the massless particles are
that some at least of the observed approximate sym-
physical. ) It should be noted that the extension or
metries of relativistic particle physics might be inter-
corollary of the Goldstone theorem discussed by
preted in a similar way. The most serious obstacle has Streater' also fails in precisely this case. If long-range
been the appearance in such theories of unwanted mass-
less particles, as predicted by the Goldstone theorem. '
fields are introduced, spontaneous symmetry breaking
can lead to mass splitting.
In nonrelativistic theories such as the BCS model, the As has been emphasized recently by Higgs, ' it thus
corresponding zero-energy-gap excitation modes may
appears that the only way of reconciling spontaneous
be eliminated by the introduction of long-range forces.
symmetry breaking in relativistic theories with the
The first indication of a similar effect in relativistic
theories was provided by the work of Anderson, ' who
absence of massless particles is to couple in gauge fields.
Another possibility is that Goldstone bosons may turn
showed that the introduction of a long-range field, like
out to be completely uncoupled and therefore physically
the electromagnetic field, might serve to eliminate irrelevant. In this case, however, the Hilbert space
massless particles from the theory. More recently,
decomposes into the direct product of a physical Hilbert
Higgs' has exhibited a model which shows explicitly
space and a free-particle Fock space for the Goldstone
how the massless Goldstone bosons are eliminated by
bosons. The broken symmetry appears only in the
coupling the current associated with the broken sym-
Figure 10.
metry to a gauge field. The Kibble
reasons on symmetry
for the breakdown breaking
latter, and no in trace
nonabelian
of it remainsgauge
in anytheory.
physical
quantities. In most simple cases, the symmetry trans-
of the Goldstone theorem in this case have been formations leave the physical Hilbert space completely
'
analyzed by Guralnik. , Hagen, and Kibble. The situ-
invariant; and in any case they act unitarily on it. Such
ation is identical with that in the nonrelativistic domain.
theories cannot therefore explain observed approximate
this theory would be renormalizable like quantum electrodynamics because it is gauge invariant
*The research reported in this document has been sponsored symmetries. This decoupling of Goldstone modes does
in part by the U. S. Air Force Office of Scientific Research OAR occur in the I orentz-gauge treatment of models like
like quantum electrodynamics.”
through the European Office Aerospace Research, U. S. Air Force. that discussed by Higgs, in which in fact no trace of the
"j|. Nambu, Phys. Rev. Letters 4, 380 (1960). Y. Nambu and
~

Jona-Lasinio, Phys. Rev. 122, 345 {1961);M. Baker and S. L. original U(1) symmetry remains in the physical states.
I have been G.
asked
Glashow, whether
ibid. Weinberg
128, 2462 (1962); S. L. Glashow, and Salam
ioid. 130, 2132 were the first to use gauge symmetry breaking to
However it does not occur in corresponding non-Abelian
(1962). gauge theories, to which the conventional (i.e., Gupta-
' J. Goldstone, Nuovo Cimento 19, 154 (1961); J. Goldstone,
give masses to A.particles other than gauge bosons. They were the first to generate masses for leptons
Salam, and S. Weinberg, Phys. Rev. 127, 965 (1962). Bleuler) Lorentz-gauge formation is inapplicable.
' P. W. Anderson, Phys. Rev. 130, 439 (1963). It has been suggested by Fuchs' that in the case of
in this way. For' strong
P. W.
4
interactions,
Higgs, Phys. matters are more complicated. The modern understanding is
Letters 12, 132 (1964). non-Abelian gauges the massless particles may persist
G. S. Guralnik, C. R. Hagen, and T. W. B. Kibble, Phys. Rev.
Letters 13, 585 (1964). See also T. W. B. Kibble, in Proceedings of
that hadron masses come partly
the Oxford International Conference onfrom
Elementarydynamical
Particles, 1965 e↵ects of Phys.
R. F. Streater,
6
QCD and15,partly
Rev. Letters 475 (1965). from the bare masses
P. W. Higgs, Phys. Rev. 145, 1156 (1966).
7
{Rutherford High Energy Laboratory, Harwell, England, 1966),
19. ' N. Fuchs, Phys. Rev. 140, B911, (1965).
of quarks and p.leptons. (The masses of protons, 1554 neutrons, and pions come mostly from the QCD
e↵ects while heavier hadrons containing charm or bottom quarks get mass mostly from the quark
bare masses.) In the modern understanding, it is the quark bare masses, not the part of the hadron
masses coming from QCD e↵ects, that result from gauge symmetry breaking and the coupling to
the Higgs particle. Such a clear statement was however not possible until QCD was put in its
modern form in 1973, enabling the full formulation of the Standard Model. From a modern point
of view, earlier attempts to connect hadron masses to gauge symmetry breaking (as opposed to
spontaneous breaking of global chiral symmetries) mostly did not focus on the right part of the
problem.
The emergence of the Standard Model brings us to the modern era. I will conclude this talk
by sketching briefly a few of the subsequent developments. First we will talk about the strong
12

CONCEPTUAL FOUNDATIONS OF THE UNIFIED

THEORY OF WEAK AND ELECTROMAGNETIC
INTERACTIONS
Nobel Lecture, December 8, 1979
by STEVEN WEINBERG
Lyman Laboratory of Physics Harvard University and Harvard-
Smithsonian Center for Astrophysics, Cambridge, MASS., USA.
tions.
Now, back to 1967. I had been considering the implications of the
broken SU(2) x SU(2) symmetry of the strong interactions, and I thought
of trying out the idea that perhaps the SU(2) x SU(2) symmetry was a
“local,” not merely a “global,” symmetry. That is, the strong interactions
might be described by something like a Yang-Mills theory, but in addition
to the vector mesons of the Yang-Mills theory, there would also be axial
vector Al mesons. To give the meson a mass, it was necessary to insert a
common and Al mass term in the Lagrangian, and the spontaneous
!
breakdown of the SU(2) x SU(2) symmetry would then split the and Al
by something like the Higgs mechanism, but since the theory would not be
gauge invariant the pions would remain as physical Goldstone bosons.
This theory gave an intriguing result, that the mass ratio should be
and in trying to understand this result without relying on perturbation
theory, I discovered certain sum rules, the “spectral function sum rules,”
[23] which turned out to have variety of other uses. But the SU(2) x SU(2)
theory was not gauge invariant, and hence it could not be renormalizable,
[24] so I was not too enthusiastic about it. [25] Of course, if I did not insert
the mass term in the Lagrangian, then the theory would be gauge
invariant and renormalizable, and the Al would be massive. But then
there would be no pions and the mesons would be massless, in obvious
contradiction (to say the least) with observation.
At some point in the fall of 1967, I think while driving to my office at

Figure 11. A passage from Steve Weinberg’s Nobel Prize Lecture in 1979.

interactions. Since the discovery of asymptotic freedom in 1973, we describe the strong interactions
via an unbroken non-abelian gauge theory with gauge group SU (3), coupled to quarks. The SU (3)
gauge symmetry is definitely unbroken, so at first sight it looks like spontaneous breaking of gauge
symmetry turned out to be the wrong idea for the strong interactions.
There is a mystery, however, in QCD: why don’t we see the quarks? Experiment !and computer
simulations! both seem to show that the quarks are “confined,” that the energy grows indefinitely
if one tries to separate a quark from an antiquark. Confinement is quite a surprise and I would say
that we still do not fully understand it today. However it was realized in the 1970’s that supercon-
ductivity comes to the rescue again, giving an understandable explanation of how confinement can
happen. An isolated magnetic monopole would have infinite energy in a superconductor because
of the Meissner e↵ect. As sketched in fig. 12, a monopole is a source of magnetic flux, but the
Meissner e↵ect would cause this flux to be compressed into an Abrikosov-Gorkov flux tube, with
an energy proportional to its length.
 13

Flux Tube

Figure 12. In a superconductor, a magnetic monopole (small ball) is the source of a flux
tube. As a result, its energy grows linearly with the size of the system.

The best qualitative understanding that we have of confinement today is to say that it involves a
“dual” to the Meissner e↵ect, where this “duality” somehow generalizes to nonabelian gauge theory
the symmetry of Maxwell’s equations that exchanges electric and magnetic fields. Electric charges
– quarks – are then confined by a “dual Meissner e↵ect.” We do not fully understand this in the
case of QCD, but by now we know various situations in four-dimensional gauge theories in which
something like this happens.
Now we come to the weak interactions. The original Weinberg-Salam model was based on a
weakly-coupled picture with an elementary Higgs field – an elaboration of the Landau-Ginzburg
and Higgs models to include nonabelian gauge symmetry and leptons (and later quarks). But many
physicists for decades have wondered if the analogy with superconductivity is even stronger – if the
breakdown of the electroweak gauge symmetry involves something more like the BCS mechanism
of superconductivity.
There have been numerous motivations, and of course di↵erent physicists have had di↵erent
motivations at di↵erent times. Some simply suspected that the analogy between the weak inter-
actions and superconductivity would turn out to be even closer. Some considered the model with
an elementary scalar field to be arbitrary and inelegant. Another motivation for some was the
fact that the Standard Model is not predictive for lepton and quark masses (and “mixing angles”).
Each mass is a free parameter, determined by the strength of the coupling of the Higgs field to
a given quark or lepton. Maybe a model of “dynamical symmetry breaking,” more like the BCS
mechanism, would give a more predictive model.
Perhaps the most compelling motivation came from the “hierarchy problem.” Although the
electroweak gauge theory with a Higgs field is renormalizable, there is a puzzle about it. In the
action describing the Higgs field
Z
d4 x D µ ¯ D µ (| |2 a2 )2 ,

the parameter a2 , which determines the mass scale of weak interactions, is a “relevant” parameter
in the renormalization group sense. Generic ideas of renormalization theory suggest that a2 should
be in order of magnitude as large as the largest mass scale of the theory – probably the mass scale
of gravity or of grand unification of some sort, but anyway much bigger than the mass scale of weak
interactions. By analogy, in condensed matter physics, unless one tunes a parameter – such as the
temperature – one does not see a correlation length much longer than the lattice spacing. Why the
electroweak length scale is so much bigger than the particle physics analog of the lattice spacing is
the “hierarchy problem.”
There is no problem writing down a model that replaces the Higgs field with a pairing mecha-
nism (involving a new “technicolor” gauge symmetry with“techniquarks”) and solves the hierarchy
14

problem. There is even an immediate success: such a model can easily reproduce a relationship
between the W and Z masses and the weak mixing angle that was one of the early triumphs of the
Standard Model. However, a serious problem was well-recognized in the late 1970’s: one can argue
that the way the Standard Model gives quark and lepton masses is inelegant and unpredictive, but
at least it works. Simple models of “dynamical electroweak gauge symmetry breaking” have serious
problems giving realistic quark and lepton masses. Of course, people found clever fixes but it never
looked like a match made in heaven.
Experiment began to weigh in seriously in the 1990’s. Neither the Higgs particle nor the new
particles required by “dynamical” models were discovered. But tests of the Standard Model –
especially in e+ e annihilation – became precise enough that it was possible to say that the original
version of the Standard Model with a simple Higgs field is a better fit than more sophisticated
“dynamical” models. There certainly were still fixes, but people had to work harder to find them.
Probably we all know where this story has reached, at least for now. A particle with properties
a lot like the Higgs particle of the Standard Model was found a year ago with a mass around
125 GeV. It looks like the electroweak scale is weakly coupled, as is possible in part because of
Anderson’s insights about gauge symmetry breaking in 1962. But the hierarchy problem is still
with us. “Dynamical” models that tried to solve it have not been confirmed, and weakly coupled
models – notably based on supersymmetry – that tried to solve it have also not yet been confirmed.
I will just end with a question: When the LHC gets to higher energies in 2015, will this situation
persist or will it be resolved?