Gauge theory field

1
1
Gauge Invariance
1.1
Introduction
Gauge field theories have revolutionized our understanding of elementary particle
interactions during the second half of the twentieth century. There is now in place
a satisfactory theory of strong and electroweak interactions of quarks and leptons
at energies accessible to particle accelerators at least prior to LHC.
All research in particle phenomenology must build on this framework. The pur-
pose of this book is to help any aspiring physicist acquire the knowledge necessary
to explore extensions of the standard model and make predictions motivated by
shortcomings of the theory, such as the large number of arbitrary parameters, and
testable by future experiments.
Here we introduce some of the basic ideas of gauge field theories, as a starting
point for later discussions. After outlining the relationship between symmetries of
the Lagrangian and conservation laws, we first introduce global gauge symmetries
and then local gauge symmetries. In particular, the general method of extending
global to local gauge invariance is explained.
For global gauge invariance, spontaneous symmetry breaking gives rise to mass-
less scalar Nambu–Goldstone bosons. With local gauge invariance, these unwanted
particles are avoided, and some or all of the gauge particles acquire mass. The sim-
plest way of inducing spontaneous breakdown is to introduce scalar Higgs fields
by hand into the Lagrangian.
1.2
Symmetries and Conservation Laws
A quantum field theory is conveniently expressed in a Lagrangian formulation. The
Lagrangian, unlike the Hamiltonian, is a Lorentz scalar. Further, important conser-
vation laws follow easily from the symmetries of the Lagrangian density, through
the use of Noether’s theorem, which is our first topic. (An account of Noether’s
theorem can be found in textbooks on quantum field theory, e.g., Refs. [1] and [2].)
Gauge Field Theories. Paul H. Frampton
Copyright © 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ISBN: 978-3-527-40835-1

2 1 Gauge Invariance
Later we shall become aware of certain subtleties concerning the straightforward
treatment given here. We begin with a Lagrangian density
L φk(x), ∂μφk(x) (1.1)
where φk(x) represents genetically all the local ﬁelds in the theory that may be of
arbitrary spin. The Lagrangian L(t) and the action S are given, respectively, by
L(t) = d3
xL φk(x), ∂kφk(x) (1.2)
and
s
t2
t1
dtL(t) (1.3)
The equations of motion follow from the Hamiltonian principle of stationary
action,
δS = δ
t2
t1
dt d3
xL φk(x), ∂μφk(x) (1.4)
= 0 (1.5)
where the ﬁeld variations vanish at times t1 and t2 which may themselves be chosen
arbitrarily.
It follows that (with repeated indices summed)
0 =
t2
t1
dt d3
x
∂L
∂φk
δφk +
∂L
∂(∂μφk)
δ(∂μφk) (1.6)
=
t2
t1
dt d3
x
∂L
∂φk
− ∂μ
∂L
∂(∂μφk)
δφk +
∂L
∂(∂μφk)
δφk
t=t2
t=t1
(1.7)
and hence
∂L
∂φk
= ∂μ
∂L
∂(∂μφk)
(1.8)
which are the Euler–Lagrange equations of motion. These equations are Lorentz
invariant if and only if the Lagrangian density L is a Lorentz scalar.
The statement of Noether’s theorem is that to every continuous symmetry of
the Lagrangian there corresponds a conservation law. Before discussing internal
symmetries we recall the treatment of symmetry under translations and rota-
tions.

1.2 Symmetries and Conservation Laws 3
Since L has no explicit dependence on the space–time coordinate [only an im-
plicit dependence through φk(x)], it follows that there is invariance under the trans-
lation
xμ → xμ = xμ + aμ (1.9)
where aμ is a four-vector. The corresponding variations in L and φk(x) are
δL = aμ∂μL (1.10)
δφk(x) = aμ∂μφk(x) (1.11)
Using the equations of motion, one ﬁnds that
aμ∂μL =
∂L
∂φk
δφk +
∂L
∂(∂μφk)
δ(∂μφk) (1.12)
= ∂μ
∂L
∂(∂μφk)
δφk (1.13)
= aν∂μ
∂L
∂(∂μφk)
∂νφk (1.14)
If we deﬁne the tensor
Tμν = −gμνL +
∂L
∂(∂μφk)
∂νφk (1.15)
it follows that
∂μTμν = 0 (1.16)
This enables us to identify the four-momentum density as
Pμ = T0μ (1.17)
The integrated quantity is given by
Pμ = d3
xPμ (1.18)
= d3
x(−g0μL + πk∂μφk) (1.19)
where πk = ∂L /∂φk is the momentum conjugate to φk. Notice that the time
component is
P0 = πk∂0φk − L (1.20)
= H (1.21)

where H is the Hamiltonian density. Conservation of linear momentum follows
since
∂
∂t
Pμ = 0 (1.22)
This follows from Pi = J0i and ∂
∂t J0i becomes a divergence that vanishes after
integration d3x.
Next we consider an infinitesimal Lorentz transformation
xμ → xμ = xμ + μνxν (1.23)
where μν = − νμ. Under this transformation the fields that may have nonzero
spin will transform as
φk(x) → δkl −
1
2
μν
μν
kl φl(x ) (1.24)
Here
μν
kl is the spin transformation matrix, which is zero for a scalar field. The
factor 1
2 simplifies the final form of the spin angular momentum density.
The variation in L is, for this case,
δL = μνxν∂μL (1.25)
= ∂μ( μνxνL ) (1.26)
since μν∂μxν = μνδμν = 0 by antisymmetry.
We know, however, from an earlier result that
δL = ∂μ
∂L
∂(∂μφk)
δφk (1.27)
= ∂μ
∂L
∂(∂μφk)
λνxν∂λφk −
1
2
λν
kl λνφl (1.28)
It follows by subtracting the two expressions for δL that if we define
Mλνν
= xν
gλμ
− xμ
gλν
L +
∂
∂(∂λφk)
xμ
∂ν − xν
∂μ φk +
μν
kl φl (1.29)
= xμ
T λν
− xν
T λμ
+
∂L
∂(∂λπk)
μν
kl φl (1.30)
then
∂λMλμν
= 0 (1.31)

1.2 Symmetries and Conservation Laws 5
The Lorentz generator densities may be identified as
Mμν
= M0μν
(1.32)
Their space integrals are
Mμν
= d3
xMμν
(1.33)
= d3
x xμ
Pν
− xν
Pμ
+ πk
μν
kl φl (1.34)
and satisfy
∂
∂t
Mμν
= 0 (1.35)
The components Mij (i, j = 1, 2, 3) are the generators of rotations and yield
conservation of angular momentum. It can be seen from the expression above that
the contribution from orbital angular momentum adds to a spin angular momen-
tum part involving
μν
kl .
The components M0i generate boosts, and the associated conservation law [3]
tells us that for a field confined within a finite region of space, the “average” or
center of mass coordinate moves with the uniform velocity appropriate to the result
of the boost transformation (see, in particular, Hill [4]). This then completes the
construction of the 10 Poincaré group generators from the Lagrangian density by
use of Noether’s theorem.
Now we may consider internal symmetries, that is, symmetries that are not re-
lated to space–time transformations. The first topic is global gauge invariance; in
Section 1.3 we consider the generalization to local gauge invariance.
The simplest example is perhaps provided by electric charge conservation. Let
the finite gauge transformation be
φk(x) → φk(x) = e−iqk φk(x) (1.36)
where qk is the electric charge associated with the field φk(x). Then every term in
the Lagrangian density will contain a certain number m of terms
φk1 (x)φk2 (x) · · · φkm (x) (1.37)
which is such that
m
i=1
qki = 0 (1.38)
and hence is invariant under the gauge transformation. Thus the invariance im-
plies that the Lagrangian is electrically neutral and all interactions conserve elec-
tric charge. The symmetry group is that of unitary transformations in one dimen-

sion, U(1). Quantum electrodynamics possesses this invariance: The uncharged
photon has qk = 0, while the electron field and its conjugate transform, respec-
tively, according to
ψ → e−iqθ
ψ (1.39)
¯ψ → e+iqθ ¯ψ (1.40)
where q is the electronic charge.
The infinitesimal form of a global gauge transformation is
φk(x) → φk(x) − i i
λi
klφl(x) (1.41)
where we have allowed a nontrivial matrix group generated by λi
kl. Applying
Noether’s theorem, one then observes that
δL = ∂μ
∂L
∂(∂μφk)
δφk (1.42)
= −i i
∂μ
∂L
∂(∂μφk)
λi
klφl (1.43)
The currents conserved are therefore
Ji
μ = −i
∂L
∂(∂μφk)
λi
klφl (1.44)
and the charges conserved are
Qi
= d3
xji
0 (1.45)
= −i d3
xπkλi
klφl (1.46)
satisfying
∂
∂t
Qi
= 0 (1.47)
The global gauge group has infinitesimal generators Qi; in the simplest case, as in
quantum electrodynamics, where the gauge group is U(1), there is only one such
generator Q of which the electric charges qk are the eigenvalues.

1.3 Local Gauge Invariance 7
1.3
Local Gauge Invariance
In common usage, the term gauge field theory refers to a field theory that possesses
a local gauge invariance. The simplest example is provided by quantum electrody-
namics, where the Lagrangian is
L = ¯ψ(i/∂ − e/A − m)ψ −
1
4
FμνFμν (1.48)
Fμν = ∂μAν − ∂νAν (1.49)
Here the slash notation denotes contraction with a Dirac gamma matrix: /A ≡
γμAμ. The Lagrangian may also be written
L = ¯ψ(i/D − m)ψ −
1
4
FμνFμν (1.50)
where Dμψ is the covariant derivative (this terminology will be explained shortly)
Dμψ = ∂μψ + ieAμψ (1.51)
The global gauge invariance of quantum electrodynamics follows from the fact
that L is invariant under the replacement
ψ → ψ = eiθ
ψ (1.52)
¯ψ → ¯ψ = e−iθ ¯ψ (1.53)
where θ is a constant; this implies electric charge conservation. Note that the pho-
ton field, being electrically neutral, remains unchanged here.
The crucial point is that the Lagrangian L is invariant under a much larger
group of local gauge transformations, given by
ψ → ψ = eiθ(x)
ψ (1.54)
¯ψ → ¯ψ = e−iθ(x) ¯ψ (1.55)
Aμ → Aμ = Aμ −
1
e
∂μθ(x) (1.56)
Here the gauge function θ(x) is an arbitrary function of x. Under the transfor-
mation, Fμν is invariant, and it is easy to check that
¯ψ (i/∂ − e/A )ψ = ¯ψ(i/∂ − e/A)ψ (1.57)

so that ¯ψ/Dψ is invariant also. Note that the presence of the photon field is essential
since the derivative is invariant only because of the compensating transformation
of Aμ. By contrast, in global transformations where θ is constant, the derivative
terms are not problematic.
Note that the introduction of a photon mass term −m2AμAμ into the Lagrangian
would lead to a violation of local gauge invariance; in this sense we may say that
physically the local gauge invariance corresponds to the fact that the photon is
precisely massless.
It is important to realize, however, that the requirement of local gauge invari-
ance does not imply the existence of the spin-1 photon, since we may equally well
introduce a derivative
Aμ = ∂μ (1.58)
where the scalar transforms according to
→ = −
1
e
θ (1.59)
Thus to arrive at the correct L for quantum electrodynamics, an additional as-
sumption, such as renormalizability, is necessary.
The local gauge group in quantum electrodynamics is a trivial Abelian U(1)
group. In a classic paper, Yang and Mills [5] demonstrated how to construct a field
theory locally invariant under a non-Abelian gauge group, and that is our next topic.
Let the transformation of the fields φk(x) be given by
δφk(x) = −iθi
(x)λi
klφl(x) (1.60)
so that
φk(x) → φk(x) = klφl (1.61)
with
kl = δkl − iθi
(x)λi
kl (1.62)
where the constant matrices λi
kl satisfy a Lie algebra (i, j, k = 1, 2, . . . , n)
λi
, λj
= icijkλk
(1.63)
and where the θi(x) are arbitrary functions of x.
Since depends on x, a derivative transforms as
∂μφk → kl(∂μφl) + (∂μ kl)φl (1.64)
We now wish to construct a covariant derivative Dμφk that transforms according to
Dμφk → kl(Dμφl) (1.65)

To this end we introduce n gauge fields Ai
μ and write
Dμφk = (∂μ − igAμ)φk (1.66)
where
Aμ = Ai
μλi
(1.67)
The required transformation property follows provided that
(∂μ )φ − igAμ φ = −ig( Aμ)φ (1.68)
Thus the gauge field must transform according to
Aμ → Aμ = Aμ
−1
−
i
g
(∂μ ) −1
(1.69)
Before discussing the kinetic term for Ai
μ it is useful to find explicitly the infinites-
imal transformation. Using
kl = δkl − iλi
klθi
(1.70)
−1
kl = δkl + iλi
klθi
(1.71)
one finds that
λi
klA i
μ = kmλi
mnAi
μ
−1
−
i
g
(∂μ km) −1
ml
(1.72)
so that (for small θi)
λi
klδAi
μ = iθj
λi
, λj
kl
Ai
μ −
1
g
λi
kl∂μθi
(1.73)
= −
1
g
λi
kl∂μθi
− cijmθj
Ai
μλm
kl (1.74)
This implies that
δAi
μ = −
1
g
∂μθi
+ cijkθj
Ak
μ (1.75)
For the kinetic term in Ai
μ it is inappropriate to take simply the four-dimensional
curl since
δ ∂μAi
μ − ∂νAi
μ = cijkθi
∂μAk
ν − ∂νAk
μ
+ cijk ∂μθj
Ak
ν − ∂νθj
Ak
μ (1.76)

whereas the transformation property required is
δFi
μν = cijkθj
Fk
μν (1.77)
Thus Fi
μν must contain an additional piece and the appropriate choice turns out to
be
Fi
μν = ∂μAi
ν − ∂νAi
μ + gcijkAj
μAk
ν (1.78)
To conﬁrm this choice, one needs to evaluate
gcijkδ Aj
μAk
ν = −cijk ∂μθj
Ak
ν − ∂νθj
Ak
μ
+ g cijkcjlmθl
Am
μ Ak
ν + cijkAj
μcklmθl
Am
ν (1.79)
The term in parentheses on the right-hand side may be simpliﬁed by noting that
an n × n matrix representation of the gauge algebra is provided, in terms of the
structure constants, by
λi
jk
= −icijk (1.80)
Using this, we may rewrite the last term as
gAm
μ An
νθj
(cipncpjm + cimpcpjn) = gAm
μ An
νθj
λi
, λj
mn
(1.81)
= igAm
μ An
νθj
cijkλk
mn (1.82)
= gAm
μ An
νθj
cijkckmn (1.83)
Collecting these results, we deduce that
δFi
μν = δ ∂μAi
ν − ∂νAi
μ + gcijkAj
μAk
ν (1.84)
= cijkθj
∂μAk
ν − ∂νAk
μ + gcklmAl
μAm
ν (1.85)
= cijkθj
Fk
μν (1.86)
as required. From this it follows that
δ Fi
μνFi
μν = 2cijkFi
μνθj
Fk
μν (1.87)
= 0 (1.88)
so we may use −1
4 Fi
μνFi
μν as the kinetic term.

To summarize these results for construction of a Yang–Mills Lagrangian: Start
with a globally gauge-invariant Lagrangian
L (φk, ∂μφk) (1.89)
then introduce Ai
μ (i = 1, . . . , n, where the gauge group has n generators). Define
Dμφk = ∂μ − igAi
μλi
φk (1.90)
Fi
μν = ∂μAi
ν − ∂νAi
μ + gcijkAj
μAk
ν (1.91)
The transformation properties are (Aμ = Ai
μλi)
φ = φ (1.92)
Aμ = Aμ
−1
−
i
g
(∂μ ) −1
(1.93)
The required Lagrangian is
L (φk, Dμφk) −
1
4
Fi
μνFi
μν (1.94)
When the gauge group is a direct product of two or more subgroups, a different
coupling constant g may be associated with each subgroup. For example, in the
simplest renormalizable model for weak interactions, the Weinberg–Salam model,
the gauge group is SU(2)×U(1) and there are two independent coupling constants,
as discussed later.
Before proceeding further, we give a more systematic derivation of the locally
gauge invariant L , following the analysis of Utiyama [6] (see also Glashow and
Gell-Mann [7]). In what follows we shall, first, deduce the forms of Dμφk and Fi
μν
(merely written down above), and second, establish a formalism that could be ex-
tended beyond quantum electrodynamics and Yang–Mills theory to general relativ-
ity.
The questions to consider are, given a Lagrangian
L (φk, ∂μφk) (1.95)
invariant globally under a group G with n independent constant parameters θi,
then, to extend the invariance to a group G dependent on local parameters θi(x):
1. What new (gauge) fields Ap(x) must be introduced?
2. How does Ap(x) transform under G ?
3. What is the form of the interaction?
4. What is the new Lagrangian?

We are given the global invariance under
δφk = −iT i
klθi
φl (1.96)
with i = 1, 2, . . . , n and T i satisfying
T i
, T j
= icijkT k
(1.97)
where
cijk = −cjik (1.98)
and
cijlclkm + cjklclim + ckilcljm = 0 (1.99)
Using Noether’s theorem, one finds the n conserved currents
Ji
μ =
∂L
∂φk
T i
kl∂μφl (1.100)
∂μJi
μ = 0 (1.101)
These conservation laws provide a necessary and sufficient condition for the invari-
ance of L .
Now consider
δφk = −iT i
klθi
(x)φl(x) (1.102)
This local transformation does not leave < J invariant:
δL = −i
∂L
∂(∂μφk)
T i
klφl∂μ∂i (1.103)
= 0 (1.104)
Hence it is necessary to add new fields A p (p = 1, . . . , M) in the Lagrangian,
which we write as
L (φk, ∂μφk) → L φk, ∂μφk, A p
(1.105)
Let the transformation of A p be
δA p
= Ui
pqθi
A q
+
1
g
Cjp
μ ∂μφj
(1.106)
Then the requirement is

δL = −i
∂L
∂φk
T
j
klφl − i
∂L
∂(∂μφk)
T
j
kl∂μφl +
∂L
∂A p
U
j
pqA p
θj
+ −i
∂L
∂(∂μφk)
T
j
klφl +
1
g
∂L
∂A p
Cpj
μ ∂μθj
(1.107)
= 0 (1.108)
Since θj and ∂μθj are independent, the coefficients must vanish separately. For the
coefficient of ∂μθi, this gives 4n equations involving A p and hence to determine
the A dependence uniquely, one needs 4n components. Further, the matrix C
pj
μ
must be nonsingular and possess an inverse
Cjp
μ C−1jq
μ = δpq (1.109)
Cjp
μ C−1j p
μ = gμνδjj (1.110)
Now we define
Aj
μ = −gC−1jp
μ A p
(1.111)
Then
i
∂L
∂(∂μφk)
T i
klφl +
∂L
∂Ai
μ
= 0 (1.112)
so only the combination
Dμφk = ∂μφk − iT i
klφlAi
μ (1.113)
occurs in the Lagrangian
L φk, ∂μφk, A p
= L (φk, Dμφk) (1.114)
It follows from this equality of L and L that
∂L
∂φk Dμφ
− i
∂L
∂(Dμφl) φ
T i
klAi
μ =
∂L
∂φk
(1.115)
∂L
∂(Dμφk) φ
=
∂L
∂(Dμφk)
(1.116)
ig
∂L
∂(Dμφk) φ
T a
klφlC−1ap
μ =
∂L
∂A p
(1.117)
These relations may be substituted into the vanishing coefficient of θj occurring
in δL (above). The result is

0 = −i
∂L
∂φk Dμφ
T i
klφl +
∂L
∂(Dμφk) φ
T i
klDμφl
+ i
∂L
∂φk φ
φlAa
ν i T a
, T i
kl
λμν + Sba,j
μν = 0 (1.118)
where
Sba,j
μν = C−1ap
μ U
j
pqCbq
ν (1.119)
is defined such that
δAa
μ = gδ −C−1ap
μ A p
(1.120)
= Sba,j
μν Ab
νθj
−
1
g
∂μθa
(1.121)
Now the term in the first set of brackets in Eq. (1.118) vanishes if we make the
identification
L (φk, Dμφk) = L (φk, Dμφk) (1.122)
The vanishing of the final term in parentheses in Eq. (1.118) then enables us to
identify
Sba,j
μν = −cajbgμν (1.123)
It follows that
δAa
μ = cabcθb
Ac
μ −
1
g
∂μθa
(1.124)
From the transformations δAA
μ and δφk, one can show that
δ(Dμφk) = δ ∂μφk − iT a
klAa
μφl (1.125)
= −iT i
klθi
(Dμφl) (1.126)
This shows that Dμφk transforms covariantly.
Let the Lagrangian density for the free Aa
μ field be
L0 Aa
μ, ∂νAa
μ (1.127)
Using
δAa
μ = cabcθb
Ac
μ −
1
g
∂μθa
(1.128)

one ﬁnds (from δL = 0)
∂L0
∂Aa
μ
cabcAc
μ +
∂L0
∂(∂νAa
μ)
cabc∂νAc
μ = 0 (1.129)
−
∂L0
∂Aa
μ
+
∂L0
∂(∂μAb
ν)
cabcAc
ν = 0 (1.130)
∂L0
∂(∂νAa
μ)
+
∂L0
∂(∂μAa
μ)
+
∂L0
∂(∂μAa
ν)
= 0 (1.131)
From the last of these three it follows that ∂μAa
μ occurs only in the antisymmetric
combination
Aa
μν = ∂μAa
ν − ∂νAa
μ (1.132)
Using the preceding equation then gives
∂L0
∂Aa
μ
=
∂L0
∂Ab
μν
cabcAc
ν (1.133)
so the only combination occurring is
Fa
μν = ∂μAa
ν − ∂νAa
μ + gcabcAb
μAc
ν (1.134)
Thus, we may put
L0 Aa
μ, ∂νAa
μ = L0 Aa
μ, Fa
μν (1.135)
Then
∂L0
∂(∂νAa
μ) A
=
∂L0
∂Fa
μν A
(1.136)
∂L0
∂Aa
μ ∂μA
=
∂L0
∂Aa
μ F
+
∂L0
∂(∂Fb
μν) A
cabcAc
ν (1.137)
But one already knows that
∂L0
∂Aa
μ ∂μA
=
∂L0
∂Fb
μν
cabcAc
ν (1.138)
and it follows that L0 does not depend explicitly on Aa
μ.
L0(Aμ, ∂νAμ) = L0 Fa
μν (1.139)
Bearing in mind both the analogy with quantum electrodynamics and renormaliz-
ability we write

L0 Fa
μν = −
1
4
Fa
μνFa
μν (1.140)
When all structure constants vanish, this then reduces to the usual Abelian case.
The final Lagrangian is therefore
L (φk, Dμφk) −
1
4
Fa
μνFa
μν (1.141)
Defining matrices Mi in the adjoint representation by
Mi
ab = −iciab (1.142)
the transformation properties are
δφk = −iT i
klθi
φl (1.143)
δAa
μ = −iMi
abθi
Ab
μ −
1
g
∂μθa
(1.144)
δ(Dμφk) = −iTklθi
(Dμφl) (1.145)
δFa
μν = −Mi
abθi
Fb
μν (1.146)
Clearly, the Yang–Mills theory is most elegant when the matter fields are in the
adjoint representation like the gauge fields because then the transformation prop-
erties of φk, Dμφk and Fa
μν all coincide. But in theories of physical interest for
strong and weak interactions, the matter fields will often, instead, be put into the
fundamental representation of the gauge group.
Let us give briefly three examples, the first Abelian and the next two non-Abelian.
Example 1 (Quantum Electrodynamics). For free fermions
L ¯ψ(i/∂ − m)ψ (1.147)
the covariant derivative is
Dμψ = ∂μψ + ieAμψ (1.148)
This leads to
L (ψ, Dμψ) −
1
4
FμνFμν = ¯ψ(i/∂ − e/A − m)ψ −
1
4
FμνFμν (1.149)
Example 2 (Scalar φ4
Theory with φa
in Adjoint Representation). The globally in-
variant Lagrangian is
L =
1
2
∂μφa
∂ − μφa
−
1
2
μ2
φa
φa
−
1
4
λ φa
φa 2
(1.150)

1.4 Nambu–Goldstone Conjecture 17
One introduces
Dμφa
= ∂μφa
− gcabcAb
μφc
(1.151)
Fa
μν = ∂μAa
ν − ∂νAa
μ + gcacAb
μAc
ν (1.152)
and the appropriate Yang–Mills Lagrangian is then
L =
1
2
Dμφa
Dμφa
−
1
4
Fa
μνFa
μν −
1
2
μa
φa
φa
−
1
4
φa
φa 2
(1.153)
Example 3 (Quantum Chromodynamics). Here the quarks ψk are in the fun-
damental (three-dimensional) representation of SU(3). The Lagrangian for free
quarks is
L ¯ψk(i/∂ − m)ψk (1.154)
We now introduce
Dμψk = ∂μψk −
1
2
gλi
klAi
μψl (1.155)
Fa
μν = ∂μAa
ν − ∂νAa
μ + gfabcAb
μAc
ν (1.156)
and the appropriate Yang–Mills Lagrangian is
L ¯ψ(i/D − m)ψ −
1
4
Fa
μνFa
μν (1.157)
If a ﬂavor group (which is not gauged) is introduced, the quarks carry an additional
label ψa
k , and the mass term becomes a diagonal matrix m → −Maδab.
The advantage of this Utiyama procedure is that it may be generalized to include
general relativity (see Utiyama [6], Kibble [8], and more recent works [9–12]).
Finally, note that any mass term of the form +m2
i Ai
μAi
μ will violate the local
gauge invariance of the Lagrangian density L . From what we have stated so far,
the theory must contain n massless vector particles, where n is the number of gen-
erators of the gauge group; at least, this is true as long as the local gauge symmetry
is unbroken.
1.4
Nambu–Goldstone Conjecture
We have seen that the imposition of a non-Abelian local gauge invariance appears
to require the existence of a number of massless gauge vector bosons equal to
the number of generators of the gauge group; this follows from the fact that a
mass term +1
2
2
Ai
μAi
μ in L breaks the local invariance. Since in nature only one

massless spin-1 particle—the photon—is known, it follows that if we are to exploit
a local gauge group less trivial than U(1), the symmetry must be broken.
Let us therefore recall two distinct ways in which a symmetry may be broken. If
there is exact symmetry, this means that under the transformations of the group
the Lagrangian is invariant:
δL = 0 (1.158)
Further, the vacuum is left invariant under the action of the group generators
(charges) Qi:
Qi|0 = 0 (1.159)
From this, it follows that all the Qi commute with the Hamiltonian
Qi
, H = 0 (1.160)
and that the particle multiplets must be mass degenerate.
The ﬁrst mechanism to be considered is explicit symmetry breaking, where one
adds to the symmetric Lagrangian (L0) a piece (L1) that is noninvariant under
the full symmetry group G, although L1 may be invariant under some subgroup
G of G. Then
L = L0 = cL1 (1.161)
and under the group transformation,
δL0 = 0 (1.162)
δL1 = 0 (1.163)
while
Qi|0 → 0 as c → 0 (1.164)
The explicit breaking is used traditionally for the breaking of ﬂavor groups SU(3)
and SU(4) in hadron physics.
The second mechanism is spontaneous symmetry breaking (perhaps more ap-
propriately called hidden symmetry). In this case the Lagrangian is symmetric,
δL = 0 (1.165)
but the vacuum is not:
Qi|0 = 0 (1.166)
This is because as a consequence of the dynamics the vacuum state is degenerate,
and the choice of one as the physical vacuum breaks the symmetry. This leads to
nondegenerate particle multiplets.

1.4 Nambu–Goldstone Conjecture 19
It is possible that both explicit and spontaneous symmetry breaking be present.
One then has
L = L0 + cL1 (1.167)
δL = 0 (1.168)
δL1 = 0 (1.169)
but
Qi
|0 = 0 as c → 0 (1.170)
An example that illustrates all of these possibilities is the infinite ferromag-
net, where the symmetry in question is rotational invariance. In the paramagnetic
phase at temperature T > Tc there is exact symmetry; in the ferromagnetic phase,
T < Tc, there is spontaneous symmetry breaking. When an external magnetic field
is applied, this gives explicit symmetry breaking for both T > Tc and T < Tc.
Here we are concerned with Nambu and Goldstone’s well-known conjecture [13–
15] that when there is spontaneous breaking of a continuous symmetry in a quan-
tum field theory, there must exist massless spin-0 particles. If this conjecture were
always correct, the situation would be hopeless. Fortunately, although the Nambu–
Goldstone conjecture applies to global symmetries as considered here, the conjec-
ture fails for local gauge theories because of the Higgs mechanism described in
Section 1.5.
It is worth remarking that in the presence of spontaneous breakdown of sym-
metry the usual argument of Noether’s theorem that leads to a conserved charge
breaks down. Suppose that the global symmetry is
φk → φk − iT i
klφlθi
(1.171)
Then
∂μji
μ = 0 (1.172)
Ji
μ = −i
∂L
∂(∂μφk)
T i
klφl (1.173)
but the corresponding charge,
Qi
= d3
xji
0 (1.174)
will not exist because the current does not fall off sufficiently fast with distance to
make the integral convergent.

Figure 1.1 Potential function V (φ).
The simplest model field theory [14] to exhibit spontaneous symmetry breaking
is the one with Lagrangian
L =
1
2
∂μφ∂μφ − m2
0φ2
−
λ0
24
φ4
(1.175)
For m2
0 > 0, one can apply the usual quantization procedures, but for m2
0 < 0, the
potential function
V (φ) =
1
2
m2
0φ2
+
λ0
24
φ4
(1.176)
has the shape depicted in Fig. 1.1. The ground state occurs where V (φa) = 0,
corresponding to
φ0 = ±χ = ±
−6m2
0
λ0
(1.177)
Taking the positive root, it is necessary to define a shifted field φ by
φ = φ + χ (1.178)
Inserting this into the Lagrangian L leads to
L =
1
2
∂μφ ∂μφ + 2m2
0φ 2
−
1
6
λ0χφ 3
−
λ0
24
φ 4
+
3m4
0
λ0
(1.179)
The (mass)2 of the φ field is seen to be −2m2
0 < 0, and this Lagrangian may now be
treated by canonical methods. The symmetry φ → −φ of the original Lagrangian
has disappeared. We may choose either of the vacuum states φ = ±χ as the phys-
ical vacuum without affecting the theory, but once a choice of vacuum is made,
the reflection symmetry is lost. Note that the Fock spaces built on the two possi-

Gauge theory field

More Related Content

What's hot (19)

Viewers also liked (19)

Similar to Gauge theory field (20)

Recently uploaded (20)

Gauge theory field