Aft 2010 PDF

Advanced Field Theory
ETH Zurich and University of Zurich
FS 2010
Prof. Dr. Babis Anastasiou [ETH Zurich]

Prof. Dr. Thomas Gehrmann [University of Zurich]
Typeset by Peter Reiter
June 28, 2010

2
Please feel free to send feedback or error reports to [email protected]

Contents
1 Dimension Regularization and the Axial Anomaly 1

1.1 Basics of Dimensional Regularization . . . . . . . . . . . . . . . . . . . . . 2
1.2 Clifford Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Chiral Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 QED Axial Anomaly in Two Dimensions . . . . . . . . . . . . . . . . . . . 6
1.5 QED Axial Anomaly in Four Dimensions . . . . . . . . . . . . . . . . . . . 11
2 Chiral Symmetry in the Strong Interaction 15

2.1 Chiral Symmetry of QCD . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 -Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Chiral Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 The ⇥-Vacuum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 The Axion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.6 Instantons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3 Topological Aspects of Field Theory 43

3.1 Topological Objects in Field Theory . . . . . . . . . . . . . . . . . . . . . . 43
3.2 The U (1) Higgs Model in 2 + 1 Dimensions . . . . . . . . . . . . . . . . . . 48
3.3 The Dirac Monopole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4 The ’t-Hooft-Polyakov Monopole . . . . . . . . . . . . . . . . . . . . . . . 54
3.5 Phase Transitions in the Early Universe . . . . . . . . . . . . . . . . . . . . 59
3.6 Baryon and Lepton Number Violation . . . . . . . . . . . . . . . . . . . . 66
3.7 Sphalerons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.8 Baryogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
i
ii CONTENTS
4 (Quantum) Fields in Curved Spacetime 73

4.1 Evolution of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3 Quantum theory of the inflaton . . . . . . . . . . . . . . . . . . . . . . . . 78
Chapter 1
Dimension Regularization and the

Axial Anomaly
From quantum field theory we know that field theories contain loop diagrams with ultra-
violet (k 2 ! 1) or the infrared (k 2 ! 0) divergences.
One example of a UV-divergent integral is,

Z Z
d4 k 1 k!1 d4 k 1
! . (1.1)
(2⇡)4 (k 2 m2 )[(k q)2 m2 ] (2⇡)4 k 4
Regulating this integral will yield a logarithmic divergence which can be removed using
renormalization techniques.
In the low energy limit of a massless theory, consider for example the vertex diagram
corresponding to the integral
Z
d4 k 1
, (1.2)
(2⇡)4 k 2 (k + p1 )2 (k p2 )2
where we identify in the vertex diagram k ! p1 , p2 , p21 = p22 = 0, k = p1 + p2 . We observe

that due to the missing mass in the propagator, we have a divergence for k ! 0. In the
IR limit we can distinguish between two limits:
1. The soft limit is defined as the limit of k ! k, ! 0:

Z Z
d4 k 1 k!0 d4 k 1
4 2
! , (1.3)
(2⇡) k (2k · p1 )( 2k · p1 ) (2⇡)4 k 4
which is obviously logarithmically divergent.
1
2 CHAPTER 1. DIMENSION REGULARIZATION AND THE AXIAL ANOMALY
2. In the collinear limit we have k 2 = 0 and k k p1 , k k p2 , use the Sudakov parametriza-

tion (also known as light cone parametrization) by choosing a frame where p1 and
p2 move in opposite directions, i.e.
0 1 0 1
1 1
1 B 0C 1 B C
eµ+ = p B C , eµ = p B 0 C , pµ1 = p1 eµ+ . (1.4)
2 @0A 2@ 0 A
1 1
We can thus decompose

01
k+ + k
1
k µ = k+ eµ+ + k eµ + k? = p @ k? A , (1.5)
2 k k
+
and then using k2? = k12 + k22 continue with
k 2 = k02 k12 k22 k32 = (k0 k3 )(k0 + k3 ) k2? = 2k+ k k2? (1.6)
to change our measure,
d4 k = 2dk+ dk d2 kT = dk+ dk dk 2 . (1.7)
We can rewrite our loop integral, and take, as usual, the infrared limit,
Z Z
dk 2 dk+ dk k!0 dk 2 dk+ dk
! , (1.8)
k 2 (k 2 + 4p1 k ) (. . . ) k 2 4p1 k . . .
which again is logarithmically divergent.
1.1 Basics of Dimensional Regularization

Let d be a complex number. We wish to define an operation that we may regard as
integration over a d-dimensional space:
Z
dd pf (p). (1.9)
Here f (p) is any given function of a vector p, which is in the d-dimensional space. We will
suppose that the space is Euclidean. (Minkowski space is regarded as a one-dimensional
time together witha (d 1)-dimensional Euclidian space.)
What properties must we impose on a functional of f in order to regard it as d-
dimensional integration? The following properties or axioms are natural and are necessary
in applications to Feynman graphs:
1.2. CLIFFORD ALGEBRA 3
1. Linearity: For any a, b 2 :

Z Z Z
d p (af (p) + bg(p)) = a d p f (p) + b dd p g(p).
d d
(1.10)
2. Scaling: For any number s
Z Z
d
d p f (sp) = s d
dd p f (p). (1.11)
3. Translation invariance: For any vector q:

Z Z
d p f (p + q) = dd p f (p).
d
(1.12)
We will also require rotational covariance of our results.

Linearity is true of any integration, while translation and rotation invariance are basic
properties of Euclidian space, and the scaling property embodies the d-dimensionality.
Not only are the above three axioms necessary, but they also ensure that integration is
unique, aside from an overall normalization. In fact, they determine the usual integration
measure in an integer-dimensional space (again up to normalization).
1.2 Clifford Algebra

The Dirac matrices satisfy the following properties:
1. Anticommutation relation:
{ µ
, ⌫
} = 2g µ⌫ (1.13)
2. Hermiticity:
⇢ µ
if µ = 0,
µ†
= µ = (1.14)
µ
if µ 1.
When we use dimensional regularization, the Lorentz indices range over an infinite set
of values, so we need infinite-dimensional matrices to represent the algebra. We will also
need a trace operation:
tr 1 = f (d),
so that the representation behaves as if its dimension were f (d). We must require f (d0 )
to be the usual value at the physical space-time dimension, d = d0 . Usually this means
f (4) = 4.
The trace is a linear operation on the matrices which we will define later. In an even
integer dimension d = 2!, the standard representation of the µ ’s has dimension 2! .
However, it is not necessary to choose f (d) = 2d/2 . The variation f (d) f (d0 ) is only
relevant for a divergent graph. It is usually convenient to set f (d) = f (d0 ) for all d.
In four dimensions, 5 ⌘ i 0 1 2 3 and " µ⌫ is a totally antisymmetric Lorentz-
invariant tensor with "0123 = 1. We need 5 , for example, to define the axial current
¯ µ 5 . The "-tensor comes in because 5 = i" µ⌫  µ ⌫
/4!, and we have the trace
formula:
tr 5  µ ⌫
= i" µ⌫ tr1 = i" µ⌫ tr1.
The appropriate definition changes when we go to two dimensions: Instead of 5 we

have ˆ(1) = 0 1 , and instead of " µ⌫ we have "µ⌫ , for which "01 = "10 = 1, "00 = "11 = 0.
To continue dimensionally, we might expect 5 to satisfy
µ
{ 5, } = 0,
just as in four dimensions. But then, the only consistent result for 5 is that it has zero
trace when multiplied by any string of µ ’s. Thus we do not have a regularization involving
the usual 5 .
A consistent definition is obtained by writing
5
=i 0 1 2 3=i  µ ⌫
" µ⌫ /4! (1.15)
" µ⌫ = ( µ⌫), (1.16)
where ( µ⌫) gives the sign of the permutation of  µ⌫ with respect to (0123). This
definition is not Lorentz invariant on the full space, but only on the first four dimensions.
We have
{ 5 , µ } = 0, if µ = 0, 1, 2, 3, (1.17)
[ 5 , µ ] = 0, otherwise, (1.18)
( 5 )2 = 1, (1.19)
†
5 = 5. (1.20)
The lack of full Lorentz invariance is a nuisance, but it does give the correct axial anomaly.
Lorentz and Dirac-Algebra We require gµµ = d, so a vector Aµ has d 2 physical

degrees of freedom. We don’t have to generalize the gamma matrices themselves, just the
traces. So we have to discuss the Lorentz indices and the -matrix in { µ , ⌫ } = 2g µ⌫ .
1.3. CHIRAL SYMMETRY 5
In dimensional regularization we only care about limd!4 tr = 4. The consequence is

that if we have spinors, then they do not have additional degrees of freedom than in 4
dimension, : 2 degrees of freedom.
So we are only left with 5 , "µ⌫⇢ .
• introduce (d 4)-dimensional metric ĝµ⌫ = ĝ⌫µ , which acts as projector on general-

ized gamma matrices, with:
– ĝµ⌫ g µ ⇢ = ĝ⇢⌫ = ĝµ⌫ ĝ µ ⇢ (projects on (d 4) subspaces)

– ĝ µ µ = (d 4)
– ĝ µ ⌫ p⌫ = p̂µ , ĝ µ ⌫ ⌫
= ˆ µ : (d 4)-dim components of pµ , µ
– { µ , ˆ⌫ } = {ˆµ , ˆ⌫ } = 2ĝµ⌫
• define product of "-tensors (4 dim!)
X 4
Y
" µ1 µ2 µ3 µ 4 " ⌫ 1 ⌫ 2 ⌫ 3 ⌫ 4 = sgn(⇡) (gµi ⌫⇡(i) ĝµi ⌫⇡(i) ) (1.21)
⇡2S4 i=1
such that "µ1 µ2 µ3 µ4 = sgn(⇡)"µ⇡(1) "µ⇡(2) "µ⇡(3) "µ⇡(4) antisymmetric and "µ⌫⇢ ĝ µ↵ = 0
we can think of µ in d-dimensions, but " in 4 dimensions. this is just a prescription
with a definition of 5 such that it does not spoil symmetries.
1.3 Chiral Symmetry
In massless QED (also QCD), the left-handed and right-handed fermions decouple in the
Lagrangian:
1
L = ¯(iD) / Fµ⌫ F µ⌫ , @/ = µ @µ , Dµ = @µ + ieAµ
4
which we can rewrite it using
¯ µ
= ¯L µ
L + ¯R µ R
with R,L = 1±
2
5
.
Question How is m ¯ =?
Noether currents and conserved charges
• vector current J µ = ¯ µ R is conserved. @µ J µ = 0 (even for m 6= 0). This implies

a conserved number N = d3 xJ 0 (x), which is constant dN dt
= 0. (Fermion number
conservation).
• Axial vector current J5 µ = ¯ µ

5 = ¯R µ
R
¯L µ
L is conserved, @µ J5 µ = 0
for m = 0.
Question What is the symmetry for J5 µ ?

This means that the difference N = NR NL is conserved as well.
) NR , NL both conserved independently.
1.4 QED Axial Anomaly in Two Dimensions
Eventually, we will want to analyze the current conservation equation for the axial current
in massless QCD. However, this discussion will involve some technical complication, so
we will first study the physics that violates axial current conservation in a context in
which the calculations are relatively simple. A particularly simple model problem is that
of two-dimensional massless QED.
The Lagrangian of the massless two-dimensional QED is
1
L = ¯(iD)
/ (Fµ⌫ )2 , (1.22)
4
with µ, ⌫ = 0, 1 and Dµ ⌘ @µ + ieAµ . The Dirac matrices must be chosen to satisfy the
Dirac algebra
{ µ
, ⌫
} = 2g µ⌫ . (1.23)
In two dimensions, this set of relations can be represented by 2 ⇥ 2 matrices; we choose

✓ ◆ ✓ ◆
0 i 0 i
0
= , 1
= . (1.24)
i 0 i 0
The Dirac spinors will be two-component fields.

The product of the Dirac matrices, which anticommutes with each of the µ
, is
✓ ◆
1 0
5
= 0 1
= . (1.25)
0 1
1.4. QED AXIAL ANOMALY IN TWO DIMENSIONS 7
Then, just as in four dimensions, there are two possible currents,
jµ = ¯ µ
, j µ5 = ¯ µ 5
, (1.26)
and both are conserved if there is no mass term in the Lagrangian.

To make the conservation laws quite explicit, we label the components of the fermion
field in this spinor basis as
✓ ◆
+
= . (1.27)
The subscript indicates the 5 eigenvalue. Then, using the explicit representations of
0
, 1 , we can rewrite the fermionic part as
† †
L= + i(D0 + D1 ) + + i(D0 D1 ) . (1.28)
In the free theory, the field equation of +, would be
i(@0 + @1 ) + = i(@0 @1 ) = 0; (1.29)
the solutions to this equation are waves that move to the right in the one dimensional
space at the speed of light. We will thus refer to the particles associated with + as
right-moving fermions. The quanta associated with are, similarly, left-moving. This
distinction is analogous to the distinction between left- and right-handed particles which
gives the physical interpretation of 5 in four dimensions. Since the Lagrangian contains
no terms that mix left- and right-moving fields, it seems obvious that the number currents
for these fields are seperately conserved. Thus,
✓ ✓ 5
◆ ◆ ✓ ✓ 5
◆ ◆
1 1 +
@µ ¯ µ
= 0, @µ ¯ µ
= 0. (1.30)
2 2
It is a curious property of two-dimensional spacetime that the vector and axial vector
fermionic currents are not independent of each other. Let "µ⌫ be the totally antisymmetric
symbol in two dimensions, with "01 = +1. Then the two-dimensional Dirac matrices obey
the identity
µ 5
= "µ⌫ ⌫. (1.31)
The currents j µ5 , j µ have the same relation. Thus we can study the properties of the axial
vector current by using results that we have already defined for the vector current.
Once we have an explicit expression for the vacuum polarization, we can find the
expectation value of the current induced by a background electromagnetic field. This
quantity is generated by the diagram of Fig. 1.1, which gives
Z ✓ ◆
i qµq⌫ e
2 iq·x µ µ⌫
d xe hj (x)i = (i⇧ (x))A⌫ (q) = g µ⌫
2
· A⌫ (q), (1.32)
e q ⇡
Figure 1.1: Computation of hj µ i in a background electromagnetic field.
where A⌫ (q) is the Fourier transform of the background field. This quantity manifestly
satiesfies the current conservation relation qµ hj µ (q)i = 0.
The identity µ 5 = "µ⌫ ⌫ between the vector and axial vector currents allows us to
derive from the preceding integral the corresponding expectation value of j µ5 . We find
hj µ5 (q)i = "µ⌫ hj⌫ (q)i (1.33)

✓ ◆
µ⌫ e q⌫ q
=" A⌫ (q) A (q) . (1.34)
⇡ q2
If the axial vector current were conserved, this object would satisfy the Ward identity.
Instead, we find
e µ⌫
qµ hj µ5 (q)i = " qµ A⌫ (q). (1.35)
⇡
This is the Fourier transform of the field equation
e µ⌫
@µ j µ5 = " Fµ⌫ . (1.36)
2⇡
Apparently, the axial vector current is not conserved in the presence of electromagnetic
fields, as the result of an anomalous behaviour of its vacuum polarization diagram. Thus
we observe that the seperate conservation of left and right moving particles is violated by
quantum interactions with the background field. The symmetries of the classical theory
are broken by quantum effects.
To complete our discussion of the two-dimensional axial vector current, we will show
that the nonconservation equation also has global aspect. In free fermion theory, the
integral of the axial current conservation law gives
Z Z
µ d
d x @µ j5 (x) = d⌧ (NR NL ) = (NR NL )|t=+1 (NR NL )|t= 1 . (1.37)
2
d⌧
This relation implies the difference in the number of right-moving and left-moving fermions
cannot be changed in any possible process. Combining this with the conservation law for
the vector current, we conclude that the number of each type of fermion is seperately
conserved. We might conclude that these seperate conservation laws hold also in two-
dimensional QED. However, we have already found that we must be careful in making
statements about the axial current.
1.4. QED AXIAL ANOMALY IN TWO DIMENSIONS 9
In two-dimensional QED, the conservation equation for the axial current is replaced by
the anomalous nonconservation equation. If the right-hand side of this equation were the
total derivative of a quantity falling off sufficiently rapidly at infinity, its integral would
vanish and we would still retain the global conservation law. In fact, "µ⌫ Fµ⌫ is a total
derivative:
"µ⌫ Fµ⌫ = 2@µ ("µ⌫ A⌫ ). (1.38)
However, it is easy to imagine examples where the integral of this quantity does not
vanish, for example, a world with a constant background electric field. In such a world,
the conservation law must be violated. But how can this happen?
Let us analyse this problem by thinking about fermions in one space dimension in a
background A1 field that is constant in space and has a very slow time dependence. We will
assume that the system has a finite length L, with periodic boundary conditions. Notice
that the constant A1 field cannot be removed by a gauge transformation that satisfies the
periodic boundary conditions.
Following the derivation of the three-dimensional Hamiltonian, we find that the Hamil-
tonian of this one-dimensional system is
Z
H = dx † ( i 0 1 D1 ) (1.39)
Z n o
†
= dx i + (@1 ieA1 ) + i † (@1 ieA1 ) . (1.40)
For a constant A1 field, it is easy to diagonalize this Hamiltonian. The eigenstates of the
covariant derivatives are wavefunctions
2⇡n
eikn x , with kn = ,n 2 , (1.41)
L
to satisfy the periodic boundary conditions. Then the single-particle eigenstates of H have
energies
+ : En = +(kn eA1 ), (1.42)

: En = (kn eA1 ). (1.43)
Each type of fermion has an infinite tower of equally spaced levels. To find the ground
state of H, we fill the negative energy levels and interpret holes created among these filled
states as antiparticles.
Now, adiabatically change the value of A1 ,
2⇡
A1 = . (1.44)
eL
The fermion energy levels slowly shift in accord with the previously stated relation back
to its original value, the spectrum of H returns to its original form. In this process, each
Figure 1.2: Effect on the vacuum state of the Hamiltonian H of one-dimensional QEd due
to an adiabatic change in the background A1 field.
level of + moves down to the next position, and each level of moves up to the next
position, as shown in Fig. 1.2. The occupation numbers of levels should be maintained in
this adiabatic process. Thus, remarkably, one right-moving fermion disappears from the
vacuum and one extra left-moving fermion appears. At the same time,
Z ⇣e ⌘ Z
e
2
dx µ⌫
" Fµ⌫ = dtdx @0 A1 (1.45)
⇡ ⇡
e
= L( A1 ) (1.46)
⇡
= 2. (1.47)
Thus the integrated form of the anomalous nonconservation equation is indeed satisfied:
Z ⇣ e ⌘
NR NL = d2 x "µ⌫ Fµ⌫ . (1.48)
2⇡
Even in this simple example, we see that it is not possible to escape the question of
ultraviolet regularization in analysing the chiral conservation law. Right-moving fermions
are lost and left-moving fermions appear from the depths of the fermionic spectrum,
E ! 1. In computing the changes in the seperate fermion numbers, we have assumed
that the vacuum cannot change the charge it contains at large negative energies. This
prescription is gauge invariant, but it leads to the nonconservation of the axial vector
current.
1.5. QED AXIAL ANOMALY IN FOUR DIMENSIONS 11
Figure 1.3: Diagrams contributing to the two-photon matrix element of the divergence of
the axial vector current.
1.5 QED Axial Anomaly in Four Dimensions

We can confirm the Adler-Bell-Jackiw relation by checking, in standard perturbation
theory, that the divergence of the axial vector current has a nonzero matrix element to
create two photons. To do this, we must analyse the matrix element
Z
d4 xe iq·x hp, k|j µ5 (x)|0i = (2⇡)4 (4) (p + k q)"⇤⌫ (p)"⇤ (k)Mµ⌫ (p, k). (1.49)
The leading-order diagram contributing to Mµ⌫ are shown in Fig. 1.3. The first diagram
gives the contribution
Z " #
d4
l i(/l /
k ) i/l i(/l + p
/ )
= ( 1)( ie)2 tr µ 5 ⌫
, (1.50)
(2⇡)4 (l k)2 l2 (l + p)2
and the second diagram gives an identical contribution with (p, ⌫) and (k, ) interchanged.
It is easy to give a formal argument that the matrix element of the divergence of the
axial vector current vanishes at this order. Taking the divergence of the axial current is
equivalent to dotting this quantity with iqµ . Now we operate on the right-hand side as we
do to prove a Ward identity. Replace
qµ µ 5
= (/l + p/ /l + k/) 5
= (/l + p/) 5
+ 5
(/l k/). (1.51)
Each momentum factor combines with the numerator adjacent to it to cancel the corre-
sponding denominator. Thus we get
Z " #
d4
l (/l /
k ) /
l /l (/l + p
/ )
iqµ · [triangle] = e2 tr 5 ⌫
+ 5 2 ⌫ . (1.52)
(2⇡)4 (l k)2 l2 l (l + p)2
Now pass ⌫ through 5 in the second term and shift the integral over the first term
according to l ! (l + k):
Z " #
d4
l /
l (/l + /
k ) /l (/l + p
/ )
iqµ · [triangle] = e2 tr 5 2 ⌫ 5 ⌫
. (1.53)
(2⇡)2 l (l + k)2 l2 (l + p)2
This expression is manifestly antisymmetric under the interchange of (p, ⌫) and (k, ), so
the contribution of the second diagram in Fig. 1.3 precisely cancels.
However, because this derivation involves a shift of the integration variable, we should
look closely whether this shift is allowed by the regularization. We see that the integral
that must be shifted is divergent. If the diagram is regulated with a simple cutoff, or even
with Pauli-Villars regularization, it turns out that the shift leaves oer a finite, nonzero
term. The analysis of the axial vector current, even dimensional regularization has an
extra subtlety, because 5 is an intrinsically four-dimensional object. In their original
paper on dimensional regularization, ’t-Hooft and Veltman suggested using the definition
5
=i 0 1 2 3
(1.54)
in d dimensions. This definition has the consequence that 5
anticommutes with µ
for
µ = 0, 1, 2, 3 but commutes with µ for other values of µ.
In the evaluation, the external indices and the momenta p, k, q all live in the physical
four dimensions, but the loop momentum l has components in all dimensions. Write
l = l + ˆl (1.55)
where the first term has nonzero components in dimensions 0,1,2,3 and the second term
has nonzero components in the other d 4 dimensions. Because 5 commutes with the µ
in these extra dimensions, the identity is modified to
qµ µ 5
= (/l + k/) 5
+ 5
(/l p/) 2 5 /̂l . (1.56)
The first two terms cancel according to the argument given above; the shift is justified by
the dimensional regularization. However, the third term gives an additional contribution:
Z " #
d4
l (/l /
k ) /
l (/l + p
/ )
iqµ · [triangle] = e2 tr 2 5 /̂l ⌫
. (1.57)
(2⇡)4 (l k)2 l2 (l + p)2
To evaluate this contribution, combine denominators in the standard way, and shift the
integration variable l ! l + P , where P = xk yp. In expanding the numerator, we must
retain one factor each of ⌫ , , p/, k/ to give a nonzero trace with 5 . This leaves over one
factor of /̂l and one factor of /l which must also be evaluated with components in extra
dimensions in order to give a nonzero integral. The factors /̂l anticommute with the other
Dirac matrices in the problem and thus can be moved to adjacent positions. Then we
must evaluate the integral
Z /̂l /̂l
d4 l
, (1.58)
(2⇡)4 (l2 )3
where is a function of k, p and the Feynman parameters. Using
d 4
(/̂l )2 = ˆl2 ! l2 (1.59)
d
lieu % →
1.5. QED AXIAL ANOMALY IN FOUR DIMENSIONS 13
under the symmetrical integration, we can evaluate

Z /̂l /̂l d
d4 l i d 4 (2 )
4 2 3
,= d/2
2
2 d/2
(1.60)
(2⇡) (l ) (4⇡) 2 (3)
d!4 i
! . (1.61)
2(4⇡)2
Notice the behaviour in which a logarithmically divergent integral contributes a factor

(d 4) in the denominator and allows an anomalous term, formally proportional to (d 4),
to give a finite contribution. The remainder of the algebra in the evaluation of the integral
is straightforward. The terms involving the momentum shift P cancel, and we find
✓ ◆
i ⇥ 2 ⇤
iqµ · [triangle] = e2
tr 2 ( /
k p
/
⌫
(1.62)
2(4⇡)2
e2
= 2 "↵ ⌫ k↵ p . (1.63)
4⇡
This term is symmetric under the interchange of (p, ⌫) with (k, ), so the second diagram
of Fig. 1.3 gives and equal contribution. Thus,
e2 ↵⌫
hp, k|@µ j µ5 (0)|0i = " ( ip↵ )"⇤⌫ (p)( ik )"⇤ (k) (1.64)
2⇡ 2
e2
= 2
hp, k|"↵⌫ F↵⌫ F (0)|0i, (1.65)
16⇡
as we would expect from the Adler-Bell-Jackiw anomaly equation, which is nothing else
~ ·B
then the term 4c E ~ in classical electrodynamics.
Chapter 2
Chiral Symmetry in the Strong

Interaction
• Peskin-Schroder, Chapter 19
• Yadurain: Theory of Quark and Gluon Interactions, Ch. 7
2.1 Chiral Symmetry of QCD

The Adler-Bell-Jackiw anomaly has a number of important implications for QCD. To
describe these, we must first discuss the chiral symmetries of QCD systematically. In this
discussion, we will ignore all but the lightest quarks u, d. In many analyses of the low-
energy structure of the strong interactions, one also treats the s quark as light; this gives
results that naturally generalize the ones we will find beloew.
The fermionic part of the QCD Lagrangian is
¯ Dd
/ + di
L =ūiDu / mu ūu md dd ¯ (2.1)
/ L + d¯L iDd
=ūL iDu / r + d¯R iDd
/ L + ūR iDu / R (2.2)
+ mu (ūR uL + ūL uR ) + md (d¯R dL + d¯L dR ). (2.3)
If the u, d quarks are very light, the last two terms are small and can be neglected. Let
us study the implications of making this approximation. If we ignore the u, d masses, the
Lagrangian of course has isospin symmetry, the symmetry of an SU (2) unitary trans-
formation mixing the u, d fields. However, because the classical Lagrangian for massless
fermions contains no coupling between left- and right-handed quarks, this Lagrangian
actually is symmetric under the seperate unitary transformations
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
u u u u
! UL , ! UR . (2.4)
d L d L d R d R
15
16 CHAPTER 2. CHIRAL SYMMETRY IN THE STRONG INTERACTION
Is is useful to seperate the U (1) and SU (2) parts of these transformations; then the
symmetry group of the classical, massless QCD Lagrangian is SU (2)L ⇥SU (2)R ⇥U (1)L ⇥
U (1)R or, alternatively, SU (2)V ⇥ SU (2)A ⇥ U (1)V ⇥ U (1)A . Let Q denote the quark
doublet, with chiral components
5
✓ ◆ ✓ ◆
1 u 1+ 5 u
QL = , QR = . (2.5)
2 d 2 d
Then we can write the currents associated with these symmetries as
jLµ = Q̄L µ QL , jRµ = Q̄R µ QR , (2.6)

jLµa = Q̄L µ ⌧ a QL , jRµa = Q̄R µ ⌧ a QR , (2.7)
where ⌧ a = a /2 represent the generators of SU (2). The sums of left- and right-handed
currents give the baryon number and isospin currents
j µ = Q̄ µ
Q, j µa = Q̄ µ a
⌧ Q. (2.8)
The corresponding symmetries are the transformations given above with UL = UR . The
difference of the currents give the corresponding axial vector currents j µ5 , j µ5a :
j µ5 = Q̄ µ 5
Q, j µ5a = Q̄ µ 5 a
⌧ Q. (2.9)
In the discussion to follow, we will derive conclusions about the strong interactions by
assuming that the classical conservation laws for these currents are not spoiled by anoma-
lies. We will show below that this assumption is correct for the isotriplet currents j µ5a
but not for j µ5 .
The vector SU (2) ⇥ U (1) transformations are manifest symmetries of the strong in-
teractions, and the associated currents lead to familiar conservation laws. What about
the orthogonal, axial vector, transformations? These do not correspond to any obvious
symmetry of the strong interactions. In 1960, Nambu and Jona-Lasinio hypothesized that
these are accurate symmetries of the strong interactions that are spontaneously broken.
This idea has led to a correct and surprisingly detailed description of the properties of
the strong interactions at low energy.
Spontaneous Breaking of Chiral Symmetry Before we describe the consequence

of spontaneoulsy broken chiral symmetry, let us ask why we might expect the chiral sym-
metries to be spontaneously broken in the first place. In the theory of superconductivity,
a small electron-electron attraction leads to the appearance of a condensate of electron
pairs in the ground state of a metal. In QCD, quarks and antiquarks have strong attrac-
tive interactions, and, if these quarks are massless, the energy cost of creating an extra
quark-antiquark pair is small. Thus we expect that the vacuum of QCD will contain a
condensate of quark-antiquark pairs. These fermion pairs must have zero total momentum
and angular momentum. Thus, as Fig XX shows, they must contain net chiral charge,
2.1. CHIRAL SYMMETRY OF QCD 17
pairing left-handed quarks with antiparticles of right-handed quarks. The vacuum state
with a quark pari condensate is characterized by a nonzero vacuum expectation value for
the scalar operator
h0|Q̄Q|0i = h0|Q̄L QR + Q̄R QL |0i =
6 0, (2.10)
which transforms under UL , UR with UL 6= UR . The expectation value signals that the
vacuum mixes the two quark helicities. This allows the u, d quarks to acquire effective
masses as they move through the vacuum. Inside quark-antiquark bound states, the u, d
quarks would appear to move as if they had a sizeable effective mass, even if they had
zero mass in the original QCD Lagrangian.
The vacuum expectation value h0|Q̄Q|0i signals the spontaneous breaking of the full
symmetry group down to subgroup of vector symmetries with UL = UR . Thus there are
four spontaneously broken continuous symmetries, associated with the four axial vector
currents. The Goldstone theorem states that every spontaneously broken continuous sym-
metry of a quantum field theory leads to a massless particle with the quantum numbers
of the global symmetry rotation. This means that, in QCD with massless u, d quarks, we
should find four spin-zero particles with the correct quantum numbers to be created by
the four axial vector currents.
The real strong interactions do not contain any massless particles, but they do contain
an isospin triplet of relatively light mesons, the pions. These particles are known to have
odd parity (as we expect if they are quark-antiquark bound states). Thus, they can be
created by the axial isospin currents. We can parametrize the matrix element of j µ5a
between the vacuum and an on-shell pion by writing
h0|j µ5a (x)|⇡ b (p)i = ipµ f⇡ ab
e ip·x
, (2.11)
where a, b are isospin indices and f⇡ is a constant with dimensions of (mass)1 . We show
in an exercise that the value of f⇡ can be determined from the rate of ⇡ + decay through
the weak interaction; one finds f⇡ = 98 MeV. For this reason, f⇡ is often called the pion
decay constant. If we contract it with pµ and use the conservation of the axial currents,
we find that an on-shell pion must satisfy p2 = 0, that is, it must be massless, as required
by Goldstone’s theorem.
If we now restore the quark mass terms, the axial currents are no longer exactly
conserved. The equation of motion of the quark field is now
/ = mQ,
iDQ iDµ Q̄ µ
= Q̄m, (2.12)
where
✓ ◆
mu 0
m= (2.13)
0 md
is the quark mass matrix. Then one can readily compute
@µ j µ5a = iQ̄{m, ⌧ a }Q. (2.14)
Using this equation with the on-shell parametrization, we find
h0|@µ j µ5a (0)|⇡ b (p)i = p2 f⇡ ab

= h0|iQ̄{m, ⌧ a } 5 Q|⇡ b (p)i. (2.15)
The last expression is an invariant quantity times

1
tr[{m, ⌧ a }⌧ b ] = ab
(mu + md ). (2.16)
2
Thus, the quark mass terms give the pions masses of the form
M2 mu + md
m2⇡ = (mu + md ) = h0|Q̄Q|0i. (2.17)
f⇡ f⇡2
The mass parameter M has been estimated to be of order 400 MeV. Thus, to give the
observed pion mass of 140 MeV, one needs only (mu + md ) ⇡ 10 MeV, which then yields
h0|Q̄Q|0i = (260 MeV)3 . This is a small perturbation on the strong interactions.
This argument has an interesting implication for the nature of the isospin symmetry
of the strong interactions. In the limit in which the u, d quarks have zero mass in the
Lagrangian, these quarks acquire large, equal effective masses from the vacuum with
spontaneously broken chiral symmetry. As long as the masses mu , md in the Lagrangian
are small compared to the effective mass, the u, d quarks will behave inside the hadrons
as though they are approximately degenerate. Thus the isospin symmetry of the strong
interactions need have nothing to do with a fundamental symmetry linking u, d; it follows
for any arbitray relation between mu , md , provided that both of these parameters are much
less than 300 MeV. Similarly, the approixmate SU (3) symmetry of the strong interactions
follows if the fundamental mass of the s quark is also small compared to the strong
interaction scale. The best current estimates of the mass ratios mu : md : ms are in
fact 1 : 2 : 40, so that the fundamental Lagrangian of the strong interactions shows no
sign of flavour symmetry among the quark masses.
The Leptonic decay of charged pions is described by axial vector currents,
X µa X
j5 = Q̄ µ 5 ⌧ a Q = ū µ 5 d + d¯ µ 5 u + ū µ 5 u d¯ µ 5 d, (2.18)
a a
is probed by weak decays (exercise)

GF
LF = p µ̄ (1 5 )⌫µ ū (1 5 )d + h.c. + (µ $ e). (2.19)
2
The identification of the pion as the Goldstone boson of spontaneously broken chi-
ral symmetry leads to numerous other predictions for current matrix elements and pion
scattering amplitudes. In particular, the leading term of the pion-pion and pion-nucleon
scattering amplitude at low energy can be computed directly in terms of f⇡ by arguments
similar to one just given.
Figure 2.1: Diagrams that lead to an axial vector anomaly for a chiral current in QCD.
Anomalies of Chiral Currents Up to this point, we have discussed the chiral sym-
metries of QCD according to the classical current conservation equations. We must now
ask whether these equations are affected by the Adler-Bell-Jackiw anomaly, and what the
consequences of that modifications are.
To begin, we study the modification of the chiral conservation laws due to the coupling
of the quark currents to the gluon fields of QCD. The arguments given in the previous
section go through equally well in the case of massless fermions coupling to non-Abelian
gauge field, so we expect that an axial vector current will receive an anomalous contribu-
tion from the diagrams shown in Fig. 2.1. The anomaly equation should be the Abelian
result, supplemented by an appropriate group theory factor. In addition, since the axial
current is gauge invariant, the anomaly must also be gauge invariant. That is, it must
contain the full non-Abelian field strength, including its nonlinear terms. These terms are
actually included in the functional derivative of the anomaly.
For the axial currents of QCD, we can read the group theory factors for the Adler-Bell-
Jackiw anomaly from the diagrams in Fig. 2.1. for the axial isospin isotriplett currents,
g2 ↵
@µ j µ5a = " µ⌫
F↵c Fµ⌫
d
· tr[⌧ a tc td ], (2.20)
16⇡ 2
where Fµ⌫c
is a gluon field strength, ⌧ a is an isospin matrix, tc is a color matrix, and the
trace is taken over colors and flavours. In this case, we find
tr[⌧ a tc td ] = tr[⌧ a ]tr[tc td ] = 0, (2.21)
since the trace of a single ⌧ a vanishes. Thus the conservation of the axial isospin currents is
unaffected by the Adler-Bell-Jackiw anomaly of QCD. However, in the case of the isospin
singlet axial current, the matrix ⌧ a is replaced by the matrix 1 on flavours, and we find
g 2 nf ↵
@µ j µ5 = " µ⌫
F↵c Fµ⌫
c
6= 0, (2.22)
32⇡ 2
where nf is the number of flavours; nf = 2 in our current model.
Thus, the isospin singlet axial current is not in fact conserved in QCD. The divergence
of this current is equal to a gluon operator with nontrivial matrix elements between
hadron states. Some subtle questions remain concerning the effects of this operator. In
particular, it can be shown, as we saw for the two-dimensional axial anomaly, that the
right-hand side is a total divergence. Nevertheless, again in accord with our experience
in two dimensions, there are physically reasonable field configurations in which the four-
dimensional integral of this term takes a nonzero value. This topic is discussed further in
chapter 3. In any event, the last equation indeed implies that QCD has no isosinglet axial
symmetry and no associated Goldstone boson. This equation explains why the strong
interaction contain no light isosinglet pseudoscalar meson with mass comparable to that
of the pions, m⌘0 = 960 MeV m⇡ .
Though the axial isospin currents have no axial anomaly from QCD interactions, they
do have an anomaly associated with the coupling of quarks to electromagnetism. Again,
referring to the diagrams of Fig. 2.1, we see that the electromagnetic anomaly of the axial
isospin currents is given by
e2 ↵
@µ j µ5a = " µ⌫
F↵ Fµ⌫ · tr[⌧a Q2 ], (2.23)
16⇡ 2
where Fµ⌫ is the electromagnetic field strength, Q is the matrix of quark electric charges,
✓2 ◆
0
Q= 3 1 , (2.24)
0 3
and the trace again runs over flavours and colors. Since the matrices in the trace do not
depend on color, the color sum simply gives a factor of 3. The flavor trace is nonzero only
for a = 3,
✓4 ◆ ✓ ◆
1 a3 0 1 a3 1 1
a 2
tr[⌧ Q ] = tr 9
1 = , ,⌧ =3
; (2.25)
2 0 9 6 2 1
in that case, the electromagnetic anomaly is
e2 ↵
@µ j µ53
= " µ⌫
F↵ Fµ⌫ . (2.26)
32⇡ 2
Because the current j µ53 annihilates a ⇡ 0 meson, the last equation indicates that the axial
vector anomaly contributes to the matrix element for the decay ⇡ 0 ! 2 . We will show
that, in fact, it gives the leading contribution to this amplitude. Again, we work in the
limit of massless u, d quarks, so that the chiral symmetries are exact up to the effects of
the anomaly.
Consider the matrix element of the axial current between the vacuum and a two-photon
state:
hp, k|j µ53 |0i = "⇤⌫ "⇤ Mµ⌫ (p, k). (2.27)
This is the same matrix element that we studied in QED perturbation theory. Now,
however, we will study the general properties of this matrix element by expanding it in
Figure 2.2: Contribution that leads to a pole in the axial vector current form factor M1 .
form factors. In general, the amplitude can be decomposed by writing all possible tensor
structures and applying the restrictions that follow from symmetry under the interchange
of (p, ⌫) and (k, ) and the QED Ward identities. This leaves three possible structures:
Mµ⌫ =q µ "⌫ ↵
p↵ k M1 + ("µ⌫↵ k "µ ↵
p⌫ )k↵ p M2 (2.28)
+ [(" µ⌫↵
p " µ ↵ ⌫
k )k↵ p " µ⌫
(p k) p · k]M3 . (2.29)
The second term satisfies p⌫ Mµ⌫ = k Mµ⌫ = 0 by virtue of the on-shell conditions
p2 = k 2 = 0.
Now contract with (iqµ ) to take the divergence of the axial vector current. We find
iqµ Mµ⌫ = iq 2 "⌫ ↵

p↵ k M1 i"µ⌫ qµ (p k) p · kM3 ; (2.30)
the other terms automatically give zero. Using q = p + k, q 2 = 2p · k, we can simplify this
to
iqµ Mµ⌫ = iq 2 "⌫ ↵

p↵ k (M1 + M2 ). (2.31)
The whole quantity is proportional to q 2 and apparently vanishes in the limit q 2 ! 0.

This contrasts with the prediction of the axial vector anomaly. Taking the matrix element
of the right-hand side of @µ j µ53 , find
e2 ⌫
iqµ Mµ⌫ = " ↵
p↵ k . (2.32)
4⇡ 2
The conflicts can be resolved if one of the form factors contains a pole in q 2 . Such a
pole can arise through the process shown in Fig. 2.2, in which the current creates a ⇡ 0
meson with subsequently decays to two photons. The amplitude for the current to create
the meson is given by h0|j µ5a (x)|⇡ b (x)i. Let us parametrize the pion decay amplitude as
iM(⇡ 0 ! 2 ) = iA"⇤⌫ "⇤ "⌫ ↵

p↵ k , (2.33)
where A is a constant to be determined. Then the contribution of this process for Fig. 2.2
to the amplitude Mµ⌫ defined by hp, k||j µ53 (q)|0i is
i
(iq µ f⇡ ) (iA"µ ↵
p↵ k ). (2.34)
q2
This is a contribution to the form factor M1 ,

i
M1 = f⇡ · A, (2.35)
q2
plus terms regular at q 2 = 0. Now, we can determine A in terms of the coeffecient of the
anomaly:
e2 1
A= . (2.36)
4⇡ 2 f⇡
From the decay matrix element iM(⇡ 0 ! 2 ), it is straightforward to work out the
decay rate of ⇡ 0 . Note that, though we have worked out the decay matrix element in the
limit of a massless ⇡ 0 , we must supply the physically correct kinematics which depends
on the ⇡ 0 mass. Including a factor 1/2 for the phase space of identical particles, we find
1 1 1 X
(⇡ 0 ! 2 ) = |M(⇡ 0 ! 2 )|2 (2.37)
2m⇡ 8⇡ 2
pols.
1
= · A2 · 2(p · k)2 (2.38)
32⇡m⇡
m3
= A2 · ⇡ . (2.39)
64⇡
Thus, finally,
↵2 m3⇡
(⇡ 0 ! 2 ) = . (2.40)
64⇡ 3 f⇡2
This relation, which provies a direct measurem of the coefficient of the Adler-Bell-Jackiw
anomaly, is satisfied experimentally to an accuracy of a few percent.
2.2 -Model
It is instructive to see the abstract quantities of chiral symmetry emerging naturally in

a specific example. The simplest such example is the linear -model. It is a seeming
counterexample to effective theories because it is as renormalizable quantum field theory
describing the spontaneous breaking of chiral symmetry. Rewriting it in the form of a
non-decoupling effective field theory will bring the ingredients of spontaneously broken
chiral symmetry to the surface. Although it has the right symmetries by construction, the
linear -model is not general enough to describe the real world. It serves the purpose of
a toy model, but it should not be mistaken for the effective field theory of QCD at low
energies.
2.2. -MODEL 23
We rewrite the -model Lagrangian for the pion-nucleon system,

1
L = (@µ @ µ + @µ~⇡ @ µ~⇡ ) ( 2
+ ~⇡ 2 v 2 )2 + ¯i@/ g ¯( + i~⌧ ~⇡ 5 ) (2.41)
2 4
with the nucleon consisting of
✓ ◆
p
= (2.42)
n
in the form
1 2
L = tr(@µ ⌃@ µ ⌃) tr(⌃† ⌃) 2v 2 + ¯L i@/ L + ¯R i@/ R g ¯R ⌃ L g ¯L ⌃† R
4 16
(2.43)
using
1
⌃= i~⌧ ~⇡ , ( 2
+ ~⇡ 2 ) = tr(⌃† ⌃) (2.44)
2
to exhibit the chiral symmetry G = SU (2)L ⇥ SU (2)R :
G G
A ! gA A, gA 2 SU (2)A (A = L, R), ⌃ ! gR ⌃gL 1 .
For v 2 > 0, the chiral symmetry is spontaneously broken and the "physical" fields are the
massive field ˆ = v and the Goldstone fields ~⇡ . The Lagrangian with its non-derivative
couplings for the fields ~⇡ seems to be at variance with the Goldstone theorem predicting
a vanishing amplitude whenever the momentum of a Goldstone boson goes to zero.
As an example, we will now state the linear and non-linear representation of L :
1. Linear representation:
Given a linear reparametrization for our field, = v + ˜ , and with Goldstone
bosons ⇡, we arrive at the spontaneously broken Lagrangian given by
1 1
L = ((@µ ˜ )(@ µ ˜ ) 2µ2 ˜ 2 ) + (@µ~⇡ )(@ µ~⇡ ) v˜ (˜ 2 + ~⇡ 2 ) (2.45)
2 2
(˜ 2 + ~⇡ 2 )2 + ¯(i@/ gv) g ¯(˜ + i~⌧ ~⇡ 5 ) . (2.46)
4
We see the effect of spontaneous symmetry breaking, the nucleon and the ˜ fields
have gained mass,
m = gv, m ˜ = 2µ2 = 2 v, (2.47)
while the Goldstone bosons ~⇡ have remained massless which we interpret as SU (2)
gauge bosons. Please note that and g need to be small parameters.
Figure 2.3: Contributions to ⇡ + ⇡ 0 elastic scattering.
2. Non-linear representation:
In order to make the Goldstone theorem manifest in the Lagrangian using a non-
linear representation, we perform a field transformation from the original fields
, , ~⇡ to a new set , S, ~ through a polar decomposition of the matrix field ⌃
where we have a small deviation of ~ = ~⇡ + . . . :
~⌧ ~
U ( ) = exp i , ⌃ = (v + S)U ( ), (2.48)
v
S † = S, U † = U 1
, det U = 1, u2 = U, L =u L, R = u† R, U ! gR U gL 1
(2.49)
In the new fields, the model takes the form:
1 (v + S)2
L = [(@µ S)2 2µ2 S 2 ] + tr(@µ U @ µ U † ) (2.50)
2 4
vS 3 S 4 + ¯ i@/ g(v + s)(( ¯ L U R ) + ( ¯ R U † L )). (2.51)
4
As expected, we only have derivative couplings of Goldstone bosons (⇡).
Representation independence We have introduced two sets of interactions with very

different appearances. They are all nonlinearly related. In each of these forms the free
particle sector, found by looking at terms bilinear in the field variables, has the same
masses and normalizations. To compare their dynamical content, let us calculate the
scattering of the Goldstone bosons of the theory, specifically ⇡ + ⇡ 0 ! ⇡ + ⇡ 0 . The diagrams
that enter at tree level are displayed in Fig. 2.3. The relevant terms in the Lagrangians
and their tree-level amplitudes are as follows.
1. Linear representation: Given an interaction by
Lint = (~⇡ 2 )2 v˜~⇡ 2 , (2.52)

4
2.2. -MODEL 25
we calculate an amplitude given by

 ✓ ◆
i 2 v2 iq 2 q4
M= 2i + ( 2i v) 2
= 2i 1 + = +O , (2.53)
q2 m2˜ q2 2 v2 v2 v4
where q = p0+ p+ = p0 p00 and the relation m2˜ = 2 v 2 = 2µ2 has been used.
The contributions of Fig. 2.3. are seen to cancel at q 2 = 0. Thus, to leading order,
the amplitude is momentum-dependent even though the interaction contains no
derivatives. The vanishing of the amplitudes at zero momentum is universal in the
limit of exact chiral symmetry.
2. Exponential representation: Given an interaction by
(v + S)2 1 h i
Lint = tr(@µ U @ µ U † ) = 2 ( ~ @µ ~ )2 ~ 2 (@µ ~ @ µ ~ ) + g(S), (2.54)
4 6v
where, again, Fig. 2.3 (b) has a higher order O(p4 ) contribution, leaving only Fig.
2.3 (a),
i(p0+ p+ )2
M= + .... (2.55)
v2
The lesson to be learned is that both representations give the same answer despite very
different forms and even different Feynman diagrams. A similar conclusion would follow
for any other observable that one might wish to calculate.
The above analysis demonstrates a powerful field theoretic theorem, proved first by R.
Haag, on representation independence. It states that if two fields are related nonlinearly,
e.g. = F ( ) with F (0) = 1, then the same experimental observables result if one
calculates with the field using L( ) or instead with using L( F ( )). The proof
consists basically of demonstrating that two S-matrices are equivalent if they have the
same single particle singularities, and since F (0) = 1, and have the same free field
behaviour and single particle singularities. This result can be made plausible if we think of
the scattering in non-mathematical terms. If the free particles are isolated they have the
same mass and charge, and experiment cannot tell the particle from the particle. At
this level they are in fact the same particles, due to F (0) = 1. The scattering experiment
is then performed by colliding the particles. The results cannot depend on whether a
theorist has chosen to calculate the amplitude using the or the names. That is, the
physics cannot depend on a labeling convention.
This result is quite useful as it lets us employ nonlinear representations in situations
where they can simplify the calculation. The linear sigma model is a good example. We
have seen that the amplitude of this theory are momentum-dependent. Such behaviour
is obtained naturally when one uses the nonlinear representations, whereas for the linear
representation more complicated calculations involving assorted cancelations of constant
terms are required to produce the correct momentum dependence. In addition, the non-
linear representations allow one to display the low energy results of the theory without
explicitly including the massive ˜ (or S) and fields.
2.3 Chiral Perturbation Theory
Now we want to couple our purely strongly interacting theory with the electroweak interac-
tions. We know that pions, protons and neutrons interact both strongly and electroweakly.
The QCD Lagrangian with Nf (Nf = 2 or 3) massless quarks q = (u, d, . . . )
✓ ◆
↵ ↵ 1 ↵ ↵µ⌫
0
LQCD = q̄i µ
@µ + igs Gµ q G G (2.56)
2 4 µ⌫
1 ↵ ↵µ⌫
/ L + q̄R iDq
= q̄L iDq / R G G (2.57)
4 µ⌫
1
qR,L = (1 ± 5 )q (2.58)
2
has a global symmetry
SU (Nf )L ⇥ SU (Nf )R ⇥U (1)V ⇥ U (1)A .

| {z }
chiral group G
At the effective hadronic level, the quark number symmetry U (1)V is realized as baryon
number. The axial U (1)A is not a symmetry at the quantum level due to the Abelian
anomaly. The Noether currents of the chiral group G are
a
JAaµ = q̄A µ
qA (A = L, R; ↵ = 1, . . . , Nf2 1). (2.59)
2
A classical symmetry can be realized in quantum field theory in two different ways
depending on how the vacuum responds to a symmetry transformation. All theoretical
and phenomenological evidence suggests that the chiral group G is spontaneously broken
to the vectorial subgroup SU (Nf )V . The axial generators of G are non-linearly realized
and there are (Nf2 1) massles pseudoscalar Goldstone bosons. There is a well-known
procedure how to realize a spontaneously broken symmetry on quantum fields. In the
special case of chiral symmetry with its parity transformation, the Goldstone fields can
be collected in an unitary matrix field U ( ) transforming as
G
U ( ) ! gR U ( )gL 1 , (gL , gR ) 2 G (2.60)
under chiral rotations. There are different parametrizations of U ( ) corresponding to

different choices of coordinates for the chiral coset space SU (Nf )L ⇥ SU (Nf )R /SU (Nf )V .
In the case of non-linear exponential parametrization, we can write
!
~t · ~
U ( ) = exp i (2.61)
v
2.3. CHIRAL PERTURBATION THEORY 27
where ~ is a vector with the 3 components of the pion field and where the product ~t · ~
describes a given transformation.
For Nf = 2 with ~t = ~⌧ we have
!
⇡0
1 p ⇡+
p ~⌧ · ~ = 2
⇡0 . (2.62)
2 ⇡ p
2
For Nf = 3 with ~t = ~ we have

0 ⇡0 ⌘8
1
p
2
+ p
6
⇡+ K+
1 B ⌘8 C
p ~·~=@ ⇡ ⇡0
p
2
+ p
6
K0 A . (2.63)
2 0 2⌘ 8
K K̄ p
6
The Lagrangian of the Standard Model is not chiral invariant. The chiral symmetry of
the strong interactions is broken by the electroweak interactions generating in particular
non-zero quark masses. The basic assumptions of PT is that the chiral limit consti-
tutes a realistic starting point for a systematic expansion in chiral symmetry breaking
interactions.
How do we include the other forces? Since all these fields are given by vector fields
in the background we can parametrize by introducing external fields. We have separate
left- and right-handed couling because the weak coupling is chiral. To incorporate this
chiral structure, we write an effective field theory by absorbing the projector of a specific
chirality state in the field definition. In this way we can fill our space with external fields
for the electromagnetic and weak forces as vector and axial vector fields.
We extend the chiral invariant QCD Lagrangian by coupling to the external hermitian
matrix fields vµ , aµ , s, p (vector, axial vector, scalar and pseudo-scalar):
L = L0QCD + q̄ µ
(vµ + aµ 5 )q q̄(s ip 5 )q. (2.64)
The external field method has two major advantages:
1. External photons and W boson fields are among the gauge fields vµ , aµ (Nf = 3):
rµ = vµ + aµ = eQAext
µ (2.65)
e
lµ = vµ aµ = eQAextµ p Wµext,+ T+ + h.c. (2.66)
2 sin ✓W
0 1
0 Vud Vus
1
Q = diag(2, 1, 1), T+ = @0 0 0 A. (2.67)
3
0 0 0
Q is the quark charge matrix, the Vij are the Kobayashi-Maskawa mixing matrix
elements, the only allowed charge modifying couplings are d ! u, s ! u and T+
is the raising operator in isospin state. Green functions for electromagnetic and
semileptonic weak currents can be obtained as functional derivatives of a generating
functional Z[v, a, s, p] with respect to external photon and W boson fields. This
procedure is valid for all fields ⇡, K, just K 0 needs the Z boson field as well. Z
boson fields would fit in the same structure, although it would look a bit more
complicated.
2. The scalar and pseudoscalar fields s, p give rise to Green functions of (pseudo)scalar
quark currents, but they also provide a very convenient way of incorporating ex-
plicit chiral symmetry breaking through the quark masses. The physically interesting
Green functions are functional derivatives of the generating functional Z[v, a, s, p]
at
v µ = aµ = p = 0
and
s = Mq = diag(mu , md , . . . ). (2.68)
The practical advantage is that Z[v, a, s, p] can be calculated in a manifestly chiral

invariant way. The actual Green functions with broken chiral symmetry are then
obtained by taking appropriate functional derivatives. Loosely speaking, we can use
s, p to reintroduce masses as external fields which are constant over all space.
Inclusion of external fields promotes the global chiral symmetry G to a local one:
G 1 1
q ! gR (1 + 5 )q + gL (1 5 )q (2.69)
2 2
G 1
rµ ! gR rµ gR + igR @µ gR 1 (2.70)
G
lµ ! gL lµ gL 1 + igL @µ gL 1 (2.71)
G
s + ip ! gR (s + ip)gL 1 . (2.72)
The local nature of G requires the introduction of a covariant derivative

G
Dµ U = @µ U irµ U + iU lµ , Dµ ! gR Dµ U gL 1 , (2.73)
and of associated non-Abelian field strength tensors
FLµ⌫ = @ µ l⌫ @ ⌫ lµ i[lµ , l⌫ ] (2.74)

FRµ⌫ = @ µ lr⌫ @ ⌫ rµ i[rµ , r⌫ ]. (2.75)
External fields do not have kinetic parts. Consequently, the external fields are not
affected by the spontaneous breakdown of G. On a fundamental level, the electroweak
2.3. CHIRAL PERTURBATION THEORY 29
gauge symmetry is broken in two steps, at the Fermi scale and at the chiral symmetry
breaking scale. Thus, there is in principle a small mixing between q̄q and whichever fields
are responsible for the electroweak breaking at the Fermi scale (Higgs, technicolour, . . . ).
via the Higgs-Kibble mechanism, three of those states become the longitudinal components
of W and Z bosons. Here, we are only interested in the light orthogonal states, the
pseudoscalar pseudo-Goldstone bosons.
We consider interaction processes in which the structure of mesons and baryons is not
resolved by the strong interaction.
PT is the low-energy effective field theory of the Standard Model. The chiral La-
grangians are organized in a derivative expansion based on the chiral counting rules
U O(p0 ) (2.76)
Dµ U, vµ , aµ O(p1 ) (2.77)
µ⌫
FL,R O(p2 ). (2.78)
PT is also an expansion in quark masses around the chiral limit. In principle, one can
formulate PT as an independent expansion in both derivatives and quark masses. It is
convenient, however, to combine these two expansions in a single one by making use of
the relations between meson and quark masses. Standard PT is defined by the simplest
choice corresponding to the counting rule
s, p O(p2 ) (2.79)
for the scalar and pseudoscalar external fields.
The locally chiral invariant Lagrangian of lowest order describing strong, electromag-
netic and semileptonic weak interactions of mesons is given by
v2
L2 = tr Dµ U Dµ U † + U † + †
U , = 2B(s + ip). (2.80)
4
The two low energy constants (LECs) of O(p2 ) are related to the pion decay constant and
to the quark condensate in the chiral limit:
v = f⇡ (1 + O(mq )) = 92.4 MeVh0|ūu|0i = v 2 B. (2.81)
This theory is nonrenormalizable, meaning that at higher orders divergences show up

which cannot be absorbed into parameter redefinitions. One can, however, consider the
extension to O(p4 ), then the divergences of the one-loop diagrams of the theory given by
L2 can be absorbed into the parameters of L2 + L4 , resulting in a consistent theory with
considerably more parameters.
Baryons can be described in PT with
✓ ◆
i
1/2
u = U exp ~⌧ · ~⇡ for SU (2) (2.82)
2v
and ~ · ~ for SU (3) correspondingly. With this we can define a vector field including the
external gauge fields
⇥ ⇤
uµ = i u† (@µ irµ )u u(@µ ilµ )u† (2.83)
which allows us to construct a chiral meson baryon Lagrangian of order O(p).

The chiral Lagrangians at O(p) for the Fermi theory of ⇡N and for M B are given by
(1) gA
L⇡N = N̄ (i@/ M + U / 5 )N (Nf = 2) (2.84)
 2
(1) d f
LM B = tr B̄(i@/ M )B + B̄ µ {uµ , B} + B̄ µ
5 [uµ , B] (Nf = 3) (2.85)
2 2
(1)
The Lagrangian L⇡N is of the form expected from the discussion of the linear -model.
B is the octet given by
0 ⌃0 ⇤
1
p + p
2 6
⌃+ p
B ⌃0 C
B=@ ⌃ p
2
+ p⇤6 n A. (2.86)
0 2⇤
⌅ ⌅ p
6
At O(p), there are two (three) LECs for Nf = 2(3) : M is the nucleon (baryon) mass
and gA is the nucleon axial-vector coupling constant in the chiral limit (experimentally
determined, gA = 1.25, by pure weak processes, i.e. neutron decay). The axial SU (3)
coupling constants f, d are related to gA via
gA = f + d. (2.87)
For nf = 3 the baryons are in an octet representation of SU (3) whereas for nf = 2,

the baryones (p, n) transform as a doublet under SU (2).
Note that we have created mass spontaneously by chiral symmetry breaking, thus the
mass-matrix is computable in terms of -models extended to SU (3). Also recall that we
are doing perturbation theory expanding quark masses in (small) momentum, and not in
the coupling constant as in the usual way.
2.4 The ⇥-Vacuum

Typically we can represent a finite gauge transformation by the exponential of an in-
finitesimal gauge transformations (see Lie Algebras and Lie Groups), which means that
the gauge transformation is related to the identity continuously.
The action of such a finite gauge transformation is, generated by parameter ↵,
! U (↵) (2.88)
2.4. THE ⇥-VACUUM 31
Dµ = (@µ + igAµ ) (2.89)
with Aµ ! U Aµ U 1
+ gi (@µ U )U 1
.
For continuous transformations, we have U = exp [ i↵a (x)T a ] with T a generators.

However, for non-Abelian groups there exists a different class of gauge transforma-
tions, finite discrete transformations, i.e. parity transformation. It cannot be continuously
related to identity in odd dimensions, hence discrete:
P = diag( 1, 1, 1), P 2 O(3), P 2

/ SO(3). (2.90)
The ✓ vacuum One is used to consider the effect on gluon fields of ’small’ gauge
transformations, i.e. those which are connected to the identity operator in a continuous
manner. There also exists ’large’ gauge transformations which change the color gauge
fields in a more drastic fashion. For example the gauge transformation generated by
x2 d2 2id⌧ · x
⇤1 (x) = + , (2.91)
x2 + d2 x2 + d2
where d is an arbitrary parameter and ⌧ is an SU (2) Pauli matrix in any SU (2) subgroup
of SU (3), e.g.
0 1
0 1 0
⌧1 = 1 = @1 0 0A , ⌧2 = 2 , ⌧3 = 3 ,
0 0 0
(other embeddings possible, but for simplicity we choose this, other’s are just a linear
cominbation of other ’s), transforms the null potential A(x) = 0 into
(1) i
Aj (x) = (rj ⇤1 (x))⇤1 1 (x) (2.92)
g
2d ⇥ ⇤ (1)
= 2 2 2
⌧j (d2 x2 ) + 2xj (⌧ · x) 2d(x ⇥ ⌧ )j , A0 (x) = 0. (2.93)
g(x + d )
Note that for d ! 0 this gauge transformation does not vanish, in fact, it becomes singular
at the origin. It cannot be deformed into unity continuously. Here, we are using the matrix
notation
a
Aµ = Aaµ . (2.94)
2
This potential lies in an SU (2) subgroup of the full color SU (3) group, and is ’large’
in the sense that it cannot be deformed continuously into the identity. The ⌧ · x factor
couples the internal color indices to the spatial position such that a path in coordinate
space implies a corresponding path in the SU (2) color subspace. We can associate with
a gauge potential A a topological charge called winding number, invariant under small
gauge transformations,
Z
ig33
n= d3 xtr [Ai (x)Aj (x)Ak (x)] "ijk . (2.95)
24⇡ 2
(1)
As can be demonstrated by direct substitution, the gauge field of Aj corresponds to
the value n = 1. Fields with integer vlaue of the winding number n can be obtained by
repeated application of ⇤1 (x),
⇤n (x) = [⇤1 (x)]n . (2.96)
All gauge potentials can be classified into disjoint sectors labeled by their winding number.
The existence of these distinct classes has interesting consequences. For example, con-
sider a configuration of the gluon field that starts off at t = 1 as the zero potential
A(x) = 0, has some interpolating A(x, t) for intermediate times, and ends up at t = +1
lying in the gauge equivalent configuration A(x) = A(1) (x). In other words, we discusss
an adiabatic change. Then the following integral can be shown to be nonvanishing:
Z
g33 1
2
d4 x Fµ⌫
a
F̃ aµ⌫ (F̃ aµ⌫ = "µ⌫↵ F↵a ). (2.97)
32⇡ 2
This is surprising because the integrand is a total divergence. As noted in electromag-
netism, F F̃ can be rewritten as
1
a
Fµ⌫ F̃ aµ⌫ = @µ K µ , K µ = 2"µ⌫ (Aa⌫ F a + gfabc Aa⌫ Ab Ac ), (2.98)
3
and thus the integral can be written as a surface integral at t = ±1. For the field
configuration under consideration, this reduces to the winding number integral
Z Z
g2 g2
4 a
d x Fµ⌫ F̃ aµ⌫
= d4 x @µ K µ (2.99)
32⇡ 2 32⇡ 2
Z
g2
= d3 x K0 |t=1
t= 1 (2.100)
32⇡ 2
Z h i
g2 (1) (1) (1)
= i d3
x " ijk
tr Ai (x)Aj (x)Ak (x) (2.101)
24⇡ 2
= 1. (2.102)
More generally, the integral of F F̃ gives the change in the winding number
Z Z
g2 g2
2
4 a
d x Fµ⌫ F̃ aµ⌫
= 2
d3 x K0 |t=1
t= 1 = n+ n (2.103)
32⇡ 32⇡
between asymptotic gauge field configurations.
2.4. THE ⇥-VACUUM 33
Thus, the vacuum state vector will be characterized by configurations of gluon fields
which fall into classes labeled by the winding number. Moreover, there will exist a cor-
respondence between the gauge transformations {⇤n } and unitary operators {Un } which
transform the state vectors. For example, a vacuum state dominated by a field config-
uration in the zero winding class (’near’ to Aµ = 0) would be transformed by U1 into
configurations with a dominance of n = 1 configurations, or more generally,
U1 |ni = |n + 1i. (2.104)
This implies that a gauge-invariant vacuum state requires contributions from all classes,
such as the coherent superposition
X
|✓i = e in✓ |ni, (2.105)
n
where ✓ is an arbitrary parameter. It follows from U1 |ni = |n + 1i that this ✓-vacuum is

gauge-invariant up to an overall phase
U1 |✓i = ei✓ |✓i. (2.106)
The QCD vacuum must contain contributions from all topological classes.
The ✓-term Given this nontrivial vacuum structure, one requires three ingredients to
completely specify QCD: the QCD Lagrangian, the coupling constant (i.e. ⇤QCD ), and
the vacuum label ✓. How can we account for the different vacua corresponding to choices
of ✓? In a path integral representation, the ✓ = 0 vacuum would imply generic transition
elements of the form
Z X
out h✓ = 0|X|✓ = 0iin = [dAµ ][d ][d ¯]XeiSQCD
= out hm|X|niin . (2.107)
n,m
The presence of a nonzero ✓ leads to an extra phase,

X
out h✓|x|✓iin = ei(m n)✓ out hm|x|niin . (2.108)
n,m
However, this phase can be accounted for in the path integral by the addition of a new
term to SQCD . In particular we have, through the use of the winding number,
Z 2
g3 R 4 a aµ⌫
out h✓|X|✓iin = [dAµ ][d ][d ¯]XeiSQCD +i 64⇡2 ✓ d x Fµ⌫ F̃ (2.109)
X
= ei(m n) out hm|X|niout , (2.110)
n,m
where X is some operator. We see that the quantity (m n) given by the winding number
difference of the fields contributing to the path integral is equivalent to a new exponential
factor containing Fµ⌫

a
F̃ aµ⌫ . Thus a correct procedure for doing calculations involving the
✓-vacua is to follow the ordinary path integral methods but with the QCD Lagrangian
containing the new term,
(✓=0) g32 a aµ⌫

LQCD = LQCD + ✓ F F̃ . (2.111)
64⇡ 2 µ⌫
The parameter ✓ is to be considered a coupling constant. Since the operator F F̃ is P -odd
and T -odd, a nonzero ✓ can induce measurable T violation. Later, we shall show how to
connect ✓ to physical observables. There is an important distinction between the various ✓
vacua of QCD and the many possible vacuum states of a spontaneously broken symmetry
such as the Higgs sector of the electroweak theory. In the latter case, the various possible
vacuum expectation values of the Higgs field label different states within the same theory.
In contrast, each value of ✓ corresponds to a different theory, just as each value of ⇤QCD
would label a different theory. Specifying ✓ and ⇤QCD then specifies the content of the
version of QCD used by nature. One of the experimental consequences is that one can
have electrical dipole moments of elementary particles, but the strongest constraint comes
from the neutron (dexp
n < 8 · 10
26
e · cm), in exercises we conclude that |✓| / 10 9 .
Connection with the chiral rotations There is a connection between the axial
anomaly and the presence of a ✓-vacuum. It involves the matrix element of F F̃ as follows.
Consider the limit of Nf massless quarks. The U (1) axial current
Nf
X
j5µ = ¯j µ 5 j (2.112)
j=1
is not conserved due to the anomaly,

Nf ↵s a aµ⌫
@ µ j5µ = F F̃ . (2.113)
8⇡ µ⌫
However, because of the fact that F F̃ is a total divergence, one can define a new conserved
current
Nf ↵s
j̃5µ = j5µ Kµ . (2.114)
8⇡
While j̃5µ does form a conserved charge,
Z
Q̃5 = d3 x j̃5µ (x), (2.115)
neither Q̃5 nor j̃5µ is gauge-invariant. In fact, under the gauge transformation of ⇤1 it
follows that that the operator Q̃5 changes by a c-number integer,
U1 Q̃5 U1 1 = Q̃5 2Nf . (2.116)

2.5. THE AXION 35
This tells us that in the world of massless quarks, the different ✓-vacua are related by a
chiral U (1) transformation,
U1 ei↵Q̃5 |✓i = U1 expi↵Q̃5 U1 1 U1 |✓i = ei(✓ 2Nf ↵) i↵Q̃5

e |✓i, (2.117)
or,
ei↵Q̃5 |✓i = |✓ 2Nf ↵i, (2.118)
where ↵ is a constant. Therefore, in the limit of massles quarks, when Q̃5 is a conserved
quantity, all of the ✓-vacua are equivalent and one can transform away the ✓ dependence
by a chiral U (1) transformation. The same is not true if quarks have mass, as the mass
terms in LQCD are not invariant under a chiral transformation.
To summarize, one finds that the existence of topologically nontrivial gauge transfor-
mations, and of field configurations which make transitions between the different topolog-
ical sectors of the theory, leads to the existence of nonvanishing effects from a new term
in the QCD action. Chiral rotations can change the value of ✓, allowing it to be rotated
away if any of the quarks are massless. However for massive quarks, the net effect is a
measurable CP violating term in the QCD Lagrangian.
2.5 The Axion

The strong CP problem: If instanton effects necessarily contribute an extra parameter to
QCD, then why is ✓ so small?
The simplest suggested solution to why ✓ is so small is to invoke yet another U (1)
symmetry, the Peccei-Quinn symmetry. The presence of this additional U (1) symmetry
would be sufficient to keep ✓ = 0.
To see how this axion hypothesis works, consider the possibilitiy of CP violation
in QCD caused by introducing a complex, nondiagonal mass matrix M for the quarks
transforming non-trivially under U (1)A ,
M = SL† M 0 SR , L = SL† 0
L, R = SR† 0
R, (2.119)
with SL,R 2 U (Nf ). We can now factor out the U (1) transformation by using
SL,R = ei L,R
S̄L,R (2.120)
with S̄L,R 2 SU (Nf ), which then yields a U (1)A transformation
SA = exp i( R L) (2.121)
such that ✓ transforms under chiral rotation as
✓!✓ 2Nf ( R L ). (2.122)
Discussing the physical phase difference L R we require to obtain a real diagonal M ,

which is related to the CKM phases, we find:
arg det M = arg(det SL† det M 0 det SR ) (2.123)

= arg det SL† + arg det SR + arg det M 0 (2.124)
= 2Nf ( R
0
L ) + arg det M , (2.125)
where the det M 0 is the determinant of the masses, not of the CKM matrix.
Now we can state that the effective and measured CP-violating angle ✓¯ in strong
interactions is given by
✓¯ = ✓ + arg det M 0 . (2.126)
However, from experiments we get constraints which then imply that ✓¯ < 10 9 , which
then raises the question: why is ✓¯ so small? If, however, we had massless quarks, we would
get det M 0 = 0. Thus, a big question being investigated for more than two decades is
whether the up-quark is in fact massless.
A dynamical explanation to eliminate this effective ✓¯ term is given by adding a new
field to the QCD action given by:
1
Laxion = ¯(M e iv
) + @µ @ µ (2.127)
2
where is the axion field, which couples to the quark mass term via the phase factor.
The axion arises as a Nambu-Goldstone boson of the new broken U (1) symmetry of the
quark and the Higgs sector.
Now perform another axial U (1) transformation on the quark fields that eliminates
the F F̃ term entirely and puts all CP violating terms in the mass matrix. We then find
that the mass term in QCD is multiplied by
⇣ ⇣ ⌘⌘
exp i ✓ + arg det M 0
(2.128)
v
which can be absorbed by shifting to
! + v(✓ + arg det M 0 ). (2.129)
Since the axion is massless, the kinetic term is invariant under this shift, so the shift is
sufficient to absorb all CP violating terms that appear exclusively in the mass matrix.
In this way, the introduction of a massless axion field, to lowest order, can absorb all
strong CP violating effects by a shift. The motivation is again the minimization of the
vacuum energy (which corresponds to the minimum of vacuum energy) because any term
of the form
g 2 a aµ⌫
F F̃
32⇡ 2 µ⌫
2.5. THE AXION 37
only increase the vacuum energy.

To recapitulate, we have taken the ✓ into the Lagrangian and we have made it dynam-
ical. By introducing the axion field, we have added two new parameters, v and M .
Although the axion gives us a way in which the strong CP problem might be solved,
experimentally the situation is still unclear. Experimental searches for the axion have
been unsuccessful. In fact, the naive axion theory that we have presented can actually
be experimentally ruled out. However, it is still possible to revive the axion theory if we
assume that it is very light and weakly coupled. Experimentally, this "invisible axion", if
it exists, should have a mass between 10 6 and 10 3 eV. The invisible axion would then
be within the bounds of experiments.
We now quickly present an overview of common features of different realizations of
axion models:
1. Axion is the Goldstone boson of a new, broken chiral symmetry (Peccei-Quinn

Symmetry)
2. Because the chiral symmetry is not exact, the axion has mass (typically of order
m⇡ fv⇡ )
3. The effective low energy Lagrangian with Nf = 2 up to 1

v
is given by
1 g ⇣ ⌘ ifu ifd
L = @µ @ µ + ✓ a
Fµ⌫ F̃ aµ⌫ @µ ū 5
µ
u @µ d¯ 5
µ
d.
2 32⇡ 2 v v v
(2.130)
Again, we can then eliminate the Fµ⌫

a
F̃ aµ⌫ term by chiral rotation of the quarks
given by
u ! ei 5 ↵u
u, d ! ei 5 ↵d
d. (2.131)
By choosing
⇣ ⌘c
f
↵f = ✓ , cu + c d = 1 (2.132)
v 2
we get
✓ ! ✓ + 2↵u + 2↵d . (2.133)
Again, we have successfully eliminated the Fµ⌫

↵
F̃ ↵µ⌫ term. We have, however, explicit
phase factors appearing in the fermionic terms,
icu (✓ ) 5u ¯ icd (✓ ) 5 d.
Lm = mu ūe v md de v (2.134)
Since the axion field is not constant, = (x), we obtain a derivative coupling as
well from the quark kinetic term,
cu cd ¯
L qq =i ū µ 5
u @µ + i d µ 5
d @µ . (2.135)
2v 2v
It is also of advantage to redefine our constants,
cu,d
0
fu,d = fu,d . (2.136)
2
4. From axion pion mixing, through the neutral ⇡ 0 , in the chiral QCD Lagrangian we
have the relevent parts, using 1 = h i= v✓, c = h0|q q̄|0i,
✓ 0◆
1 1 1 0 1 2 ⇡
0 µ 0
L = @µ ⇡ @ ⇡ + @µ @ 1 µ 1
⇡ M0 1 (2.137)
2 2 2
with the non-diagonal mass matrix M (similar to the pion mass matrix)
mu +md mu cu +md cd
!
f⇡2 c f⇡ v
c
M02 = mu cu +md cd mu c2u +md c2d . (2.138)
f⇡ v
c v 2 c
Assuming v f⇡ , we obtain the Eigenvalues

md + mu
m2⇡ = c (2.139)
f⇡2
c mu md f 2 md mu
m2 = 2 = ⇡2 m2 0 . (2.140)
v mu + md v (md + mu )2 ⇡
We thus see that the axion is indeed a very light particle,
13 MeV
m ⇡ , (2.141)
v
where we remark that v is indeed not very small.
We can also use this formalism to say something about the interactions of axions with
hadrons. The eigenvector of M02 with eigenvalue m2 has a component along the original ⇡ 0
direction equal to (mu cu md cd )f⇡ /(mu + md )v. As mentioned earlier, because of the one-
pion pole this is the dominant axion-hadron coupling. We see that the ratio of the axion
and pion production interaction amplitudes will typically be of order f⇡ /v. The fact that
axions are not observed in such collisions indicates that v > 3 TeV, in contradiction with
the original expectation that the anomalous U (1) symmetry is spontaneously broken by
the same scalar vacuum expectation values of order 0.3 TeV that break the electroweak
SU (2) ⇥ U (1) symmetry. It is possible to explain why axions are not found in reactor
or accelerator experiments by taking v as independent parameter, much larger than the
electroweak breaking scale, but there are still astrophysical limitations. Limits on the
2.6. INSTANTONS 39
rate of cooling of red giant stars give v > 107 GeV, while observations of the supernova
SN1987A indicate that v > 1010 GeV. For v > 107 GeV the axion mass would be less than
about 1 eV, so that stars are hot enough to produce axions. The ratio of the axion and
⇡ 0 decay rates into two photons is expected to be of order (f⇡ /v)2 times a phase space
ratio of order (m /m⇡ )3 , or
✓ ◆2 ✓ ◆3 ✓ ◆ 5
( !2 ) f⇡ m f⇡
= ⇡ . (2.142)
(⇡ 0 ! 2 ) v m⇡ v
Hence for v > 107 GeV the axion lifetime is expected to be longer is expected to be longer
than about 1024 s, which is ample time for the axion to travel even cosmological distances
before decaying. Cosmological arguments suggest an upper bound v < 1012 GeV, leaving
an open but narrow window of allowed axion parameters.
2.6 Instantons
In ordinary quantum mechanics, the WKB approximation is obtained by expanding in
powers of Planck’s constant, ~. To zero order we have the classical trajectory; higher order
yield the quantum fluctuations around this trajectory. The path integral formulation lends
itself particularly well to the extension of the method to the field-theoretic case. The
usefulness of the method lies in the fact that, to each order, the functional integral is of
Gaussian type and can therefore be evaluated.
It is known that there are quantum mechanical situations for which no classical trajec-
tory exists. This occurs when there is tunnelling through a potential barrier. However, one
can still adapt the WKB method to cope with this situation. We will exemplify this with
the typical case of a particle in one dimensions, subject to a potential V (x). To leading
order in the WKB approximation, the wave function is
(x) = CeiAcl ,
where Acl is now the action calculated along the classical trajectory
1
mẍ + V (x) = E.
2
Take a potential with two minima, both corresponding to V = 0, and located at x =
x0 , x = x1 , as in in figure. If E > max V , the motion from x0 to x1 is possible, and (x)
yields the "transition" or "diffusion" amplitude.
However, if E < max V , the correct WKB analysis gives a result in which the transition
amplitude
hx1 |x0 i = CeiAcl (x1 ,x0 ) ,
is to be replaced by the tunnelling amplitude,
A(x1 ,x0 )
hx1 |x0 i = Ce ,
where the Euclidian action A is not calculated along the solution of the previous equation
of motion, but for
1
mẍ + V (x) = E.
2
We see that to obtain a tunnelling amplitude we can use the same formula as that for
a transition, making only the formal replacement of t by it, both in the expression for the
action,
Z x(⇠1 )
A= dtL ! iA,
t(⇠0 )
with ⇠i the turning points, and in the equations of motions.

The tunnelling amplitude and the wave function do not give the normalization, which
may, however, be disposed of by dividing by hx0 |x0 i. We thus infer that, in quantum field
theory, the leading tunnelling amplitude will be
✓ Z ◆
4
h 1 , t = +1| 0 , t = 1i ⇡ C exp d xL( cl ) ,
where cl is the classical solution to the Euclidian equations of motions, i.e., with x0
replaced by ±ix4 , x4 real. (The sign ± depends on the boundary conditions; the reason
for the name Euclidian is that, under the transformation x0 ! ix4 ,the Minkowski metric
becomes Euclidian, up to a global sign.)
According to the discussion at the beginning of this section, we may consider this to
be the leading order of the exact expression,
Z ✓ Z ◆
4
h 1 , t = +1| 0 , t = 1i = N exp D d xL( ) ,
when expanding the field in powers of ~ around cl

.
An important property of the states of a system in a situation when tunnelling is
possible is that the stationary states (in particular, the ground state, to be identified with
the vacuum in quantum field theory) are not those in which the system is localized in one
minimum of the potential, but is shared by all minima. The situation is familiar in solid
state theory, where the potentials are periodic.
Search: Solution of YM equation in 4d Euclidian space. Consider the energy-
momentum tensor of the pure Yang-Mills QCD, leaving quarks aside, as they are irrelevant
for the considerations of this. We can write it as
1 X 1 X 1
⇥µ⌫ = g↵ Faµ↵ Fa⌫ g↵ F̃aµ↵ F̃a⌫ , F̃ µ⌫ = "µ⌫↵ F↵, .
2 a
2 a
2
It follows that ⇥00 is positive for real gluon fields:

1X
⇥00 = {(Fa0k )2 + (F̃a0k )2 }.
2 k,a
2.6. INSTANTONS 41
Therefore ⇥µ⌫ = 0 requires F ⌘ 0, and thus only the zero-field configurations may
be identified with the vacuum. However, ⇥00 no longer has a definite sign if we allow
for complex F µ⌫ . Particularly important is the case where a complex Minkowskian F µ⌫
corresponds to a real F µ⌫ in Euclidian space; for this will indicate a tunnelling situation.
This is the rationale for seeking solutions to the QCD equations in Euclidian space.
Another point is that in Minkowski space,
F̃ãµ⌫ = Faµ⌫ ,
so only the trivial F = 0 may be dual,
F̃ = ±F.
(If the sign is (+) we say F is self-dual, if ( ) anti-dual.) However, in Euclidian space,
F̃˜ µ⌫ µ⌫
a = Fa ,
so nontrivial dual values of F may, and indeed do, exist. In addition, Euclidian dual F
automatically satisfy the equation of motion.
This last property comes about as follows: the equations of motion for F read
X
Dµ Faµ⌫ ⌘ @µ Faµ⌫ + g fabc Bbµ Fcµ⌫ = 0;
the condition
Dµ F̃aµ⌫ = 0
is the Bianchi identitiy, identically satisfied by any F = D ⇥B whether or not B solves the
equations of motion. However, if F is dual, then the Bianchi identity implies Dµ Faµ⌫ = 0.
The connection with the problem of the vacuum occurs because, in the Euclidian case,
the energy-momentum tensor ⇥µ⌫ is replaced by
1X
⇥µ⌫ = {F µ F ⌫ F̃ µ F̃ ⌫ },
2
so for dual fields ⇥µ⌫ = 0: dual F may represent nontrivial vacuum states. Another
property of dual fields has to do with a condition of minimum of the Euclidian action A.
We can write
Z X
1
A= d4 x F µ⌫ F µ⌫
4
Z Z
1X 4 1 2 1 X
= d x{ (F µ⌫ ± F̃ µ⌫ ) ⌥ F µ⌫ F̃ µ⌫ } | d4 x F F̃ |.
4 2 4
Thus the action is positive-definite and reaches its minimum for dual fields, where one
has the equality Z Z
1 X 1 X
A= | d x 4
F F̃ | = d4 x (F µ⌫ 2
a ) .
4 4 a,µ⌫
Now, and at least in situations where the semi-classical approximation WKB holds, we
know that the tunnelling amplitude is given by exp( A), so the leading tunnelling effect,
if it exists, will be provided by dual configurations.
We have been talking about "nontrivial vacuum states". It is not difficult to see that
nonzero values of B exist for which G = 0. In fact, the general form of such B is what is
called a pure gauge, and my be obtained from B = 0 by a gauge transformation. Nontrivial
solutions of the equations will be such that G 6= 0.
Chapter 3
Topological Aspects of Field Theory
• Ryder, Quantum Field Theory
• Srednicki, Quantum Field Theory
3.1 Topological Objects in Field Theory
It turns out that non-linear classical field theories possess extended solutions, commonly
known as solitons, which represent stable configurations with a well-defined energy which
is nowhere singular. May this be of relevance to particle physics? Since non-Abelian gauge
theories are non-linear, it may well be, and the last ten years have seen the discovery of
vortices, magnetic monopoles and ’instantons’, which are soliton solutions to the gauge-
field equations in two space dimensions (i.e. a ’string’ in 3-dimensional space), three space
dimensions (localised in space but not in time) and 4-dimensional space-time (localised
in space and time). If gauge theories are taken seriously then so must these solutions be.
It will be seen that they do give rise to new physics and there is even the hope that they
may solve the problem of quark confinement.
Not the least interesting feature of this subject is the branch of mathematics which it
involves; for the stability of these solitons arises from the fact that the boundary condi-
tions fall into distinct classes, of which the vacuum belongs only to one. These boundary
conditions are characterised by a particular correspondence (mapping) between the group
space and co-ordinate space, and because these mappings are not continuously deformable
into one another they are topologically distinct. The relevant notions in topology will be
developed as we go along. We begin our survey with the ’sine-Gordon’ equation which
has no relevance to particle physics but whose soliton solutions are quite well understood,
and therefore form a good introduction to the subject.
43
44 CHAPTER 3. TOPOLOGICAL ASPECTS OF FIELD THEORY
Figure 3.1: A solitary wave (soliton
1. The sine-Gordon kink The sine-Gordon equation

@2 @2 1
2 2
+ 2 sin(b ) = 0 (3.1)
@t @x b
describes a scalar field in one space and one time dimension. It possesses moving, as well
as stationary, solutions. To find moving solutions, we want a field of the form
(x, t) = f (x vt) = f (⇠). (3.2)
It is easy to check that

4
f (⇠) = atan exp ± p ⇠ (3.3)
b b
is a solution, where = (1 v 2 ) 1/2 . The appearance of this wave is shown in Fig. 3.1
It is a solitary wave, which moves without changing shape or size, and therefore without
dissipation, in strong contrast to the waves set up when, for instance, a stone is thrown
into a pond. These waves spread out and the energy is dissipated. Solitary waves (solitons)
have been observed, for example, moving along canals. In this case they are solutions of
the Korteweg de Vries equation.
Since solitons are solutions of non-linear wave equations the superposition principle
is not obeyed. This means that when two solitons meet the resultant wave form is a
complicated one, but the surprising thing is that, asymptotically, the solitons separate
out again - they ’pass through’ one another. This property is, of course, of interest to
particle physics, though we shall not develop it any further here. Another consequence of
the fact that the superposition principle does not hold is that the quantisation of solitons
becomes non-trivial. We shall not follow this matter any further either. Instead, we turn
to the stationary solutions of the sine-Gordon equation, which possess an interest of a
different type.
It is clear that the sine-Gordon equation possesses an infinite number of constant
solutions (which, as we shall see in a moment, have zero energy):
2⇡n
= , n2 ; (3.4)
b
3.1. TOPOLOGICAL OBJECTS IN FIELD THEORY 45
Figure 3.2: The sine-Gordon potential V ( ).
that is, the sine-Gordon equation possesses a degenerate vacuum (’Vacuum’ here does
not, of course mean the state in Hilbert space, but simply a classical field configuration
of zero energy). The Lagrangian for the sine-Gordon equation is
✓ ◆2 ✓ ◆2
1 @ 1 @
L= V( ) (3.5)
2 @t 2 @x
with
1
V( )= [1 cos(b )], (3.6)
b2
where the constant has been chosen so that the infinite constant solutions have V = 0.
They therefore have zero energy since the energy density of the field configuration is
✓ ◆2 ✓ ◆2
1 @ 1 @
H= + + V ( ). (3.7)
2 @t 2 @x
Note that we may write
1 b2
V( )= 2 4
+ ..., (3.8)
2 4!
or, with = b2 and unit mass m
m2
V( )= 2 4
+ ..., (3.9)
2 4!
and m stands for the ’particle’ mass and for the self-interaction coupling.
The potential V is shown in Fig. 3.2 with the (zero energy) ground state given by
= 2⇡n
b
. Now construct the following configuration. Let approach one of the zeros of V
(say n = 0) as x ! 1, but a different zero (say n = 1) as x ! 1. Between these two
there is clearly a region where
2⇡n @
6= , 6= 0, (3.10)
b @x
and therefore, from the Hamiltonian H, where there is a positive energy density. We
assume the configuration is static, so @@t = 0. Because of the boundary conditions on ,
we expect the total energy to be finite. Let us find what it is. For a stationary solution
to the sine-Gordon equation we have
@2
@x 2
=
@V
@
dd ¥ faff -
-
du
(3.11)
-
which gives on integration
1
✓
@
◆2
= V ( ),
.LI#Y--dv
de
IIa
(3.12)
2 @x -
•
the integration constant being zero. Then the energy of the stationary soliton is
Z
E = Hdx
Z " ✓ ◆2
! Fy IT
#
1 @ dy
= + V ( ) dx ,
2 @x
EE KIT
Z dos
= 2V ( )dx
Z 2⇡/b p
= 2V ( )d (3.13)
0
where we have put in the integration limits given by = 2⇡n

b
between n = 0 and n = 1.
This integral is now easily performed. We have
p Z 2⇡/b
2 p
E= 1 cos(b )d
b 0
p Z 2⇡
2 p
= 2 1 cos ↵d↵
b 0
8
= 2
b
8m3
= (3.14)
where in the last step we have used the substitution given by the potential. So this soliton
has a finite energy, with the interesting property that the energy is inversely proportional
to the coupling constant. This may indeed be a useful property for particle physics.
ceavijas
There is a simple model which makes this soliton easy to visualize. consider an infinite
horizontal string with pegs attached to it at equally spaced intervals, and connect each
peg to its neighbour with a small spring (the ’coupling’). Each peg is also acted on by
gravity. The ground state corresponds to every peg hanging vertically. The soliton we have
found, with n = 0 ! 1, corresponds to the situation in Fig. 3.3. This soliton - and others
of tihs type (see below) - is called a kink. It should be clear from the peg model that the
3.1. TOPOLOGICAL OBJECTS IN FIELD THEORY 47
Figure 3.3: Pegs on a line representing the kink (soliton) solution to the sine-Gordon
equation.
kink is stable and cannot decay into the ground state with E = 0. This would involve a
(semi-)infinite number of pegs turning over, which would need a (semi-) infinite amount of
energy. But what is the mathematical reason for the stability of the kink? It is to be found
in the boundary conditions. ’Space’ in this model is an infinite line, whose boundary is
two points (the end-points). At these two points the 1-kink solution has n = 0 and n = 1,
and this is not continuously deformable into n = 0 and n = 0 (the ground state). The
kink, then, is a ’topological’ object. Its existence depends on the topological properties of
the space (in particular, its boundary, which in this case is a discrete set). This conclusion
is a general one; that is to say, the stability of soliton solutions in non-linear field theories
is a consequence of topology.
Finally, the stability of this soliton (kink) obviously signals a conservation law; there
must be a conserved charge Q, equal to an integer N (the difference between the two
integers in = 2⇡n b
), and a corresponding divergenceless current j µ (µ = 0, 1). They are
easy to construct. With
b µ⌫
jµ = " @⌫ , (3.15)
2⇡
with the antisymmetric tensor "µ⌫ normalized with "01 = 1, we have the identity
@µ j µ = 0, (3.16)
and the charge is

Z 1
Q= j 0 dx
1
Z 1
b @
= dx
2⇡ 1 @x
b
= [ (1) ( 1)] = N. (3.17)
2⇡
The interesting thing is that the current j µ does not follow the invariance of L under any
symmetry transformation. It is therefore not a Noether current. Its divergencelessness
follows independently of the equations of motion.
We consider, in the following sections, examples of solitons in gauge theories, beginning

with one in two space dimensions - the vortex.
2. The Vortex lines Now consider a scalar field in 2-dimensional space. The ’boundary’
of this space is the circle at infinity, denoted S 1 . We construct a field whose value on the
boundary is
= aein✓ (r ! 1) (3.18)
where r and ✓ are polar coordinates in the plane, a is a constant, and, to make single-
valued, n is an integer. We propose this form, rather than simply = a, because it is a
generalisation to two dimensions of the solution of the sine-Gordon equation. Taking the
gradient, we have
1 ˆ
r = inaein✓ ✓. (3.19)
r
The Lagrangian and Hamiltonian functions are
✓ ◆2 ✓ ◆2
1 @ 1 2 1 @ 1
L= |r | V ( ), H = + |r |2 + V ( ). (3.20)
2 @t 2 2 @t 2
Now let us consider a static configuration with, for example,

⇥ ⇤2
V ( ) = a2 ⇤
(3.21)
so that V = 0 on the boundary. Then as r ! 1
1 n 2 a2
H= |r |2 = (3.22)
2 2r2
and the energy (mass) of the static configuration is
Z 1 Z 1
1
E= Hrdrd✓ = ⇡n a 2 2
dr. (3.23)
r
This is logarithmically divergent; the kink, as it stands, cannot be generalised to two
dimensions - nor to more than two, for it turns out that in all these cases the energy is
divergent.
Our next goal is to change the theory in order to have such a configuration by intro-
ducing gauge bosons to cancel the bad behaviour of the scalars at infinity.
3.2 The U (1) Higgs Model in 2 + 1 Dimensions

To be a little more systematic, let us start from the Higgs Lagrangian:
1
L= Fµ⌫ F µ⌫ + |(@µ + ieAµ ) |2 m2 k |2 | |4 . (3.24)
4
3.2. THE U (1) HIGGS MODEL IN 2 + 1 DIMENSIONS 49
Spontaneous symmetry breaking is signalled by m2 < 0, and the vacuum is then given by
r
m2
| |vac = . (3.25)
2
The equations of motion are
Dµ (Dµ ) = m2 2 | |2 (3.26)
@ ⌫ Fµ⌫ = ie( @µ ⇤ ⇤
@µ ) + 2e2 Aµ | |2 (3.27)
Once again we have degenerate vacuum configurations with
| |=v ) = vei✓ . (3.28)
There is an extended field configuration with a finite energy which is not trivial by choosing
Aµ of the form
(r) ! vei✓ (3.29)

1
Aµ (r) = r✓ (r ! 1) (3.30)
e
with the components for r ! 1 Not
°
extraHR tons Ar ! 0 (3.31)
pogge us &
for cost
needof
2e 1
see ooo A✓ ! . (3.32)
z
.
won er
) r
For a static configuration H = L we have H ! 0 as r ! 1, making possible a field
.
prog d-word
configuration of finite energy. We shall now see that the effect of adding the gauge field
H
is to give the soliton magnetic flux. Consider the integral A · d~l round the circle S 1 at
~ mayimpost
R
infinity. By Stoke’s theorem, this is B~ · dS
~ = , the flux enclosed, hence
↳
I
~ ~
I
2⇡
(3.33)
pond
ninety
= A · dl = A✓ rd✓ = , go
e
and the flux is quantised. So we have, after all, constructed a 2-dimensional field config- e.
.(the electromagnetic field!).
uration, consisting of a charged scalar field and a gauge field
It carries magnetic flux, and since Dµ ! 0 and Fµ⌫ ! 0 on the boundary at infinity,
it appears to have finite energy.
. It is clear that by adding a third dimension (the z axis)
on which the fields have no dependence, this configuration becomes a vortex line. Apart
from the presence of the scalar field, it is the same as the solenoid discussed under the
Bohm-Aharonov effect; and just as that effect was attributable to the topology of the
gauge group U (1), so here also we shall se that this is this same topology which ensures
stability of the vortex.
Why are these solutions stable? As with the kink, the reason is topological. The La-
grangian is invariant under a symmetry group - in this case U (1), the electromagnetic
gauge group. The field (with boundary value given by parametrization = rei✓ with
r ! 1) is a representation of U (1). The group space of U (1) is a circle S 1 , since an
element of U (1) may be written exp(i✓) = exp[i(✓ + 2⇡)], so the space of all values of ✓ is
a line with ✓ = 0 identified with ✓ = 2⇡, and the line becomes a circle S 1 . The field is a
representation of basis of U (1), but it is the boundary value of the field in a 2-dimensional
space. This boundary is clearly a circle S 1 (the circle r ! 1, ✓ = (0 ! 2⇡)). Hence phi
defines a mapping of the boundary S 1 in physical space onto the group space S 1 :
: S 1 ! S 1, (3.34)
the mapping being specified by the integer n. Now a solution characterised by one value
of n is stable since it cannot be continuously deformed into a solution with a different
value of n (a rubber band which fits twice round a circle cannot be continuously deformed
into one which goes once round the circle). This is to say that the first homotopy group
of S 1 , the group space of U (1), is not trivial:
⇡1 (S 1 ) = . (3.35)
is the additive group of integers.
The status of a topological argument like this is that it provides a very general condition
which must be fulfilled in order that solitons exist in a particular model. If, as in the
model above, the topological argument indicates that soliton solutions are possible in
principle then one goes to the equations of motion to find them. Topology therefore
provides existence arguments.
3.3 The Dirac Monopole

Consider a magnetic monopole of strength g at the origin. The magnetic field is radial
and is given by a Coulomb-type law
g 1
B= r= gr( ) (3.36)
r3 r
(we are using Gaussian units). Since r2 ( 1r ) = 4⇡ (3)
(r), we have
r · B = 4⇡g (3)
(r) (3.37)
corresponding to a point magnetic charge, as desired. Since B is radial, the total flux
through a sphere surrounding the origin is
= 4⇡r2 B = 4⇡g. (3.38)
Consider a particle with electric charge e in the field of this monopole. The wave function
for a free particle is

i
= | | exp (p · r Et) . (3.39)
~
3.3. THE DIRAC MONOPOLE 51
In the presence of an electromagnetic field, p ! p ec A, so

✓ ◆
ie
! exp A·r ; (3.40)
~c
or the phase ↵ changes by
e
↵!↵ A · r. (3.41)
~c
Consider a closed path at fixed r, ✓, with ranging from 0 to 2⇡. The total change in
phase is
I
e
↵= A · dL (3.42)
~c
Z
e
= curlA · dS (3.43)
~c
Z
e
= B · dS (3.44)
~c
e
= (r, ✓); (3.45)
~c
(r, ✓) is the flux through the cap defined by a particular r and ✓, as shown by the shaded
area in the figure. As ✓ is varied the flux through the cap varies. As ✓ ! 0 the loop shrinks
to a point and the flux passing through the cap approaches zero:
(r, 0) = 0.
As the loop is lowered over the sphere the cap encloses more and more flux until, eventually,
at ✓ ! ⇡ we should have
(r, ⇡) = 4⇡g. (3.46)
However, as ✓ ! ⇡ the loop as again shrunk to a point so the requirement that (r, ⇡) is
finite entails that A is singular at ✓ = ⇡. This argument holds for all spheres of all possible
radii, so it follows that A is singular along the entire negative z axis. This is known as the
Dirac string. It is clear that by a suitable choice of coordinates the string may be chosen
to be along any direction, and, in fact, need not be straight, but must be continuous.
The singularity in A gives rise to the so-called Dirac veto - that the wave function
vanish along hte negative z axis. Its phase is therefore indeterminate there and there is
no necessity that ✓ ! ⇡, ↵ ! 0. We must have ↵ = 2⇡n, however, in order for to
be single-valued. We then have
e
2⇡n = 4⇡g, (3.47)
~c
1
eg = n~c. (3.48)
2
Figure 3.4:
This is the Dirac quantisation condition. It implies that the product of any electric with
any magnetic charge is given by the above. Then, in principle, if there exists a magnetic
charge anywhere in the universe all electric charges will be quantised:
~c
e=n .
2g
This is a possible explanation for the observed quantisation of electric charge, though
nowadays this is more commonly ascribed to the existence of quarks and non-Abelian sym-
metry groups. Note, however, that the quantisation condition has an explicit dependence
on Planck’s constant, and therefore on the quantum theory. In natural units, ~ = c = 1,
it becomes
1
eg = n.
2
Let us now derive an expression for the vector potential Aµ . As seen above, it is singular.
This much is clear, for if B = curlA and A is regular divB = 0, and no magnetic charges
may exist. From the argument above, A is constructed by considering the pole as the
end-point of a string of magnetic dipoles whose other end is at infinity. This gives
y x
Ax = g , Ay = g , Az = 0 (3.49)
r(r + z) r(r + z)
or
g 1 + cos ✓
Ar = A✓ = 0, A = . (3.50)
r sin ✓
A is clearly singular along r = z. If, on the other hand, the Dirac string were chosen to
be along r = z, we should have
g 1 + cos ✓
Ar = A✓ = 0, A = . (3.51)
r sin ✓
The rationale for writing the alternative expression is that the Dirac string singularity
is clearly unphysical, and in these expressions it is in different places. The only physical
singularity in A is at the origin, where divB = divcurlA is singular. Since it is obviously
desirable to get rid of unphysical singularities, this suggests the following construction.
3.3. THE DIRAC MONOPOLE 53
Figure 3.5: Ra and Rb are overlapping domains on the sphere. Ra excludes the S pole, Rb
the N pole.
Divide the space surrounding the monopole, - the sphere, essentially - into two overlapping
regions Ra and Rb as shown in the figure. Ra excludes the negative z axis (S pole) and Rb
excluded the positive z axis (N pole). In each region A is defined differently:
g 1 cos ✓
Aar = Aa✓ = 0, Aa = , (3.52)
r sin ✓
g 1 + cos ✓
Abr = Ab✓ = 0, Ab = . (3.53)
r sin ✓
It is clear that Aa and Ab are both finite in their own domain. In the region of overlap,
however, they are not the same, but are related by a gauge transformation (~ = c = 1):
2g i
Ab = Aa = Aa Sr S 1
(3.54)
r sin ✓ e
with
S = exp(2ige ). (3.55)
The covariant form is

i
Abµ = Aaµ S@µ S 1
. (3.56)
e
The requirement that the gauge transform function S be single-valued as ! + 2⇡
is clearly the Dirac quantisation condition. To check that they really do represent a
monopole, we calculate the total magnetic flux through a sphere surrounding the ori-
gin.
Z
= Fµ⌫ dxµ⌫
I
= curlA · dS
Z Z
= curlA · dS + curlA · dS.
Ra Rb
Here we take Ra and Rb as not actually overlapping, but having a common boundary,
which for convenience is taken to be the equator ✓ = ⇡/2. Since Ra , Rb have boundaries
Stokes’ theorem is applicable, and since the equator bounds Ra in a positive orientation
and Rb in a negative one we have
I I
= a
A · dl a
Ab · dlb (3.57)
✓=⇡/2 ✓=⇡.2
I
i d
= (log S 1 )d (3.58)
e d
= 4⇡g. (3.59)
This construction is due to Wu and Yang, and is, in essence, a fibre bundle formulation
of the magnetic monopole. The base space (3-dimensional space 3 minus the origin
⇠ S 2 ⇥ ) is parameterised in two independent ways, corresponding to two overlapping but
not identical regions. In each region the vector potential is given by a different expression.
Readers familiar with the Moebius strip will recognize a similarity here. there is no unique
parameterisation of the Moebius strip; locally it is the direct product of an interval (0, 1)
and a circle, but globally the circle has to be divided into two distinct overlapping regions,
with a different parameterisationof the strip in each region.
There is thus a fibre-bundle formulation of the Dirac monopole. The base space is
essentially S 2 (the sphere surrounding the monopole) and the group space is S 1 (since the
gauge group is U (1)). The fibre bundle is not S 2 ⇥ S 1 but S 3 , whih is locally the same as
S 2 ⇥ S 1 but is globally distinct.
3.4 The ’t-Hooft-Polyakov Monopole
The previous discussion of magnetic monopoles, although interesting, was not compelling,
because ordinary electrodynamics does not require that monopoles exist. Electrodynamics
withouth monopoles is perfectly consistent. However, in certain gauge theories, we will
find that spontaneous symmetry breaking is intimately connected with the existence of
monopole solutions. Hence, monopoles must exist for these theories as a consequence of
broken gauge symmetry.
It can be shown that pure gauge theory does not, by itself, possess any static non-
singular monopole configurations. However, a more general case, such as a gauge theory
coupled to scalar fields, does possess monopole solutions.
We now want to look for a finite-energy solution of the classical field equations with
nonzero winding number, but we already know that these will not exist unless we intro-
duce gauge fields. We therefore take a to be in the adjoint representation of an SU (2)
gauge group (we can also say that it is in the fundamental representation of O(3)). The
3.4. THE ’T-HOOFT-POLYAKOV MONOPOLE 55
Lagrangian is given by
1 1 a ↵
L = (Dµ )a (Dµ )a F F V ( ), a = 1, 2, 3, (3.60)
2 4 µ⌫ µ⌫
(Dµ )a = @µ ↵ + e"abc Abµ c , (3.61)
a
Fµ⌫ = @µ Aa⌫ @⌫ Aaµ + e" abc
Abµ Ac⌫ , (3.62)
with potential V ( )
1
V( = ( a a
v 2 )2 . (3.63)
8
The potential has its minimum at | | = v, thus the gauge symmetry is spontaneously
broken to U (1). The system will choose on of the possible minima we denote by for
h a i = t (0, 0, v)
h ai = v a3
. (3.64)
We can now expand around the vacuum,
0 1 1
x
a
= @ x2 A = h a i + xa . (3.65)
3
x +v
and find that the A3µ field remains massless (interpreted as electromagnetic field). Taking
a linear combination of the other vactor fields, we can define the complex vector fields
Aaµ ± iA2µ
Wµ± = p (3.66)
2
with mass mW = ev and electrical charge ±e. This theory, known as Georgi-Glashow
model, was once considered as an alternative to the Standard Model of electroweak inter-
actions, but it has been ruled out due to a missing Z 0 boson.
We discuss this model because the mass is proportional to the charge as well as to the
vacuum expectation value, mW = ev. In other words, the mass depends on the coupling
constant. If one had a way to measure the vacuum expectation value, the mass would
be completely determined by the coupling. In this model we would have to measure
the Higgs boson x to measure the vacuum expectation value. In the Standard Model of
electro-weak interaction (Glashow-Salam-Weinberg model), however, we can calculate the
vacuum expectation value without knowing the Higgs mass.
Returning to topological phenomena, we try to use the massless property of the A3µ
field to construct a field strength tensor. We can write down a gauge-invariant expression
that reduces to the electromagnetic field strength tensor Fµ⌫ = @µ A3⌫ @⌫ A3µ when setting
a
= v a3 :
1 abc
Fµ⌫ = ˆFµ⌫
a
" (Dµ ˆ)b (D⌫ ˆ)c , (3.67)
e
with
a
â = (3.68)
| |
. In this procedure we have only used the first term and ignored the second term. It
is properly derived in Voloshin (1982). We can now use this as the definition of the
electromagnetic field strength at any spacetime point where | | = 6 0. (If | | = 0, the
SU (2) symmetry is unbroken, and there is no gauge-invariant way to pick out a particular
component of the non-Abelian field strength Fµ⌫ a
.) Using repeatedly â â = 1, we can
rewrite our electromagnetic field strength tensor as
1 abc â ˆb ˆc
Fµ⌫ = @µ ( â Aa⌫ ) @⌫ ( â Aaµ ) " @µ @⌫ . (3.69)
e
This allows us to write the magnetic field components easily as
1
B i = "ijk Fjk (3.70)
2
1 ijk abc â ˆb ˆc
= "ijk @j ( â Aa )
k " " @j @k . (3.71)
2e
Let us consider the magnetic flux through a sphere at spatial infinity; this is given by
Z
= B · dS (3.72)
Z
1 ijk abc â ˆb ˆc
= dSi "ijk @j ( â Aak ) " " @j @k (3.73)
2e
Z
1
= dSi "ijk "abc â @j ˆb @k ˆc , (3.74)
2e
where the first term is a curl and has thus zero divergence. The second term, however,
yields with the winding number n
4⇡n
= . (3.75)
e
This flux implies that any soliton with nonzero winding number is a magnetic monopole
with magnetic charge QM = .
If we add a field in the fundamental representation of SU (2), then the component fields
have electric charges ± 12 e. This is the smallest electric charge we can get, and all possible
electric charges are integer multiples of it. All magnetic charges are integer multiples of
4⇡/e. Thus the possible electric and magnetic charges obey the Dirac charge quantization
condition, which is
QE QM = 2⇡k, (3.76)
3.4. THE ’T-HOOFT-POLYAKOV MONOPOLE 57
where k is an integer. This condition can be derived from general considerations of the
quantum properties of monopoles.
Now let us turn to the explicit construction of a soliton solution. This simplest case to
consider is provided by the identity map (which has winding number n = 1); the soliton
we will find is the ’t-Hooft-Polyakov monopole. We will follow the procedure from our
earlier discussions on vortex solutions, where we had a scalar field getting its vacuum
expectation value at infinity and a magnetic field falling on the boundary. In other words,
we need a gauge field A which is a total derivative.
low *
°
our to
The boundary condition on the scalar field is µ
orgy i.
xa or
lim a (x) = v . (3.77)
r
,
r!1
We can find the appropriate boundary condition on the gauge field by requiring (Dµ )a =
0 for large r which yields
✓ a◆
x xc
@i + e"abc Abi = 0. (3.78)
r r
Solving this equation by writing out the first derivative explicitly and multiplying by waxed
rxj "jda , we will find imoo
xj
doin
Adi = "dij 2 . we (3.79)
If
to
er
We finally find the static solution given by
xa
y
a
(x) = vf (r) , (3.80)
r here
1 xj score
Aai (x) = a(r)"aiy 2 , → (3.81)
e r
with f (1) = a(1) = 1 (so that Aai and a have the desired asymptotic limits) and
f (0) = a(0) = 0 (so that Aai and a are well defined at r = 0).
The total energy of the soliton (which we call M because it is the mass of the monopole)
is given by
Z
M = d3 x H (3.82)
Z 
1 a a 1
= dx 3
B B + (Di )a (Di )a + V ( ) (3.83)
2 i i 2
Z 
1 2
= dx 3
Bia + (Di )2 Bia (Di )a V ( ) (3.84)
2
where we completed the square in the last line. Using the distribution rule for covariant
derivatives,
Bia (Di )a = @i (Bia a
) (Di Bi )2 2
(3.85)
Eyes :c
er
I.2j
@ VICE ijhE"Tile 2%5 .i
VI 25A" " VI ZV
Eijr
- '
④s LHS Eijh
= - = -
[email protected]
.in#EEEsEEEEEE
÷oog÷÷÷÷
'EEe¥fE 58 CHAPTER 3. TOPOLOGICAL ASPECTS OF FIELD THEORY
.
and the implication of the Bianchi identity of
(Di Bi )a = 0, (3.86)
we can write
Z
1 a
 Z
t÷÷÷÷÷s
M= (B + (Di )a )2 + V ( )
d3 x d3 x@i (Bia a ) (3.87)
2 i
⇥ ⇤
and observe that the squared bracket, 12 (Bia + (Di )a )2 + V ( ) , is positive.
This means that the mass is constrained by
Z Z
M d3 x@i (Bia a ) = dSi Bia a
(3.88)
S1
where S 1 denotes the sphere at infinity and we used Stoke’s theorem. At space infinity,
however, we have
"
A
viii.
mais
a
= vx̂ a ri -
-
air (3.89)
and
A ,
Bia a
= vBi , ⑦ ④ (3.90)
where vBi is the generalized magnetic field form Fµ⌫ in non-Abelian gauge theory. More
loosly writing,
Z
M v · flux = v dS · B, (3.91)
S1
and see that the lowest value is related to the flux times the vacuum expactation value.
With fixed B from the boundary conditions on the sphere, we find find the flux
Z
4⇡
dS · B = (3.92)
e
and the limit for the magnetic monopole,
4⇡v 4⇡(ev) mW
M = 2
= , (3.93)
e e ↵
e2
using mW = ev, ↵ = 4⇡
. Since ↵ ⌧ 1, the monopole is much heavier than the W boson.
Alas, the Georgi-Glashow model, which has monopole solutions, is not in accord with
nature, while the Standard Model, which is in accord with nature, does not have monopole
solutions. This is because the Standard Model, electric charge is a linear combination
of an SU (2) generator and the U (1) hypercharge generator. Nothing prevents us from
introducing an SU (2) singlet field with an arbitrarily small hypercharge. Such a field
3.5. PHASE TRANSITIONS IN THE EARLY UNIVERSE 59
would have an arbitrary small electric charge (in units of e), and then the Dirac charge
quantization condition would preclude the existence of magnetic monopoles.
This disappointing situation is remedied in unified theories, where the gauge group has
a single non-Abelian factor like SU (5). In unified theories, the monopole mass is of order
Mx /↵, where mX is the mass of a superheavy vector boson; typically mX ⇠ 1015 GeV.
In other words, the physics here is that we have finite static energy solutions which
correspond to massive objects. We have a radial magnetic field, a magnetic monopole. The
mass of the magnetic monopole is related to the mass of the gauge boson that came out of
the SSB and the coupling constant. We saw that with current accelerators it is impossible
to have enough energy to produce it. However, in the early universe enough energy was
present, so magnetic monopoles would possibly exist given a correct gauge group. In a
phase transition where spontaneous symmetry breaking occurs magnetic monopoles can
be created with the correct boundary conditions.
In a cooling universe, it is possible to create an extended field configuration with a
different vacuum expectation value for for different space regions. The second ingredient
is having a correct gauge group with possibilities for degenerate vacua. Here we discussed
O(3) ⇠= SU (2) ! U (1) breaking, another possible scenario would be SU (5) ! SU (3) ⇥
SU (2) ⇥ U (1).
Thus, if you embed the Standard Model in a greater theory unifying all forces, it is
possible to have magnetic monopoles.
3.5 Phase Transitions in the Early Universe
Elementary particle theory possesses gauge symmetries that are spontaneously broken
by scalar fields belonging to non-trivial representations of the gauge group when these
fields develop non-zero expectation values at the minimum of the effective potential. In
particular, the SU (2)L ⇥U (1)Y gauge group of electroweak theory is spontaneously broken
to the U (1)em of electromagnetism by the electroweak Higgs scalar expectation value. If
grand unification to a group larger than the SU (3)C ⇥ SU (2)L ⇥ U (1)Y of the Standard
Model, e.g. to SU (5), occurs at some energy scale, then the grand unified gauge group
breaks spontaneously to the Standard Model gauge group before this gauge group in turn
breaks to the U (1)em gauge group. Things may be more complicated than this, with a
sequence of spontaneous symmetry breakings to subgroups of the original grand unified
group.
As we shall see later, finite temperature effects may result in some other minimum of
the effective potential being deeoper than the absolute minimum of the zero-temperature
theory. Then, as the universe cools, it may undergo a series of first- and second-order
phase transitions between different minima of the effective potential. Such transitions will
occur at temperatures corresponding to the scales of energy associated with the various
spontaneous symmetry breakings. In the case of electroweak phase transition to the phase
with only U (1)em unbroken, the scale of temperature for the phase transition will be or
order 102 103 GeV and, in the case of a grand unified phase transition to a phase
with SU (3)C ⇥ SU (2)L ⇥ U (1)Y in stages through a sequence of phase transitions, the
additional phase transitions will occur at intermediate scales.
It should be noted that the phase transitions can have profound effect on the history of
the universe through a number of different processesses. For example, topologically stable
objects such as domain walls, cosmic strings and magnetic monopoles can be formed when
the ’alignment’ of the spontaneous symmetry breaking expectation value is different in
adjacent causal domains. These can make substational contributions to the energy density
of the universe. Moreoever, if supercooling occurs before the phase transition is completed,
the reheating that takes place when the phase transition occurs can greatly modify pre-
existing particle densities. In addition, if the universe spends some time with positive
vacuum energy (cosmological constant) before relaxing to a minimum with zero vacuum
energy, then rapid expandion can occur. Such an ’inflationary’ stage in the history of
the universe may explain the extreme isotropy, homogeneity and flatness of the present
day observed universe. For all of these reasons it is important to understand any phase
transitions that may have occured as the universe cooled.
After a short review of the partition function of quantum statistical mechanics and
the effective potential, we calculate the effective potential with nonzero temperature and
discuss different phase transition phenomena.
Summary of Thermal Field Theory In quantum statistical mechanics, the partition

function is the trace,
X
Z = tre H = hn|e Ĥ |ni (3.94)
n
over a complete set of states |ni. It is basically the trace multiplied by a probability factor
given by the exponential. Because we want to have an equilibrium we have to start from
|ni and return back to it, hn|. After you make a small transition in temperature, you
see that it corresponds just to time. So, if you really want to see how it evolves in time,
where we do not have exp it as in quantum mechanics, but ⌘ kBLT = T1 , we see the
probability of the state return back to it.
Z X
Z= d3 xd3 y hn|xihx|e Ĥ
|yihy|ni, (3.95)
n
where we have inserted twice the unity, 1 = |yihy|, to get with the delta function hx|yi =
(x y)
Z
= d3 xd3 y hx|yihx|e Ĥ |yi (3.96)
Z
Z= d3 xhx|e Ĥ
|xi. (3.97)
We know something nice about those transitions: we can formulate the pathintegral. Let’s
introduce time:
Z 3
Z= d3 xhx, t = i |x, t = 0i (3.98)
So, the partition function, is nothing else than the integral over all transition amplitudes
of starting from a point back to the same point, after an imaginary time i has elapsed
(Wick-rotated).
Only two things are different from path integral: first, we have evolution to imaginary
time, the second is that it returns to the same point.
In the path integral formulation we have the same partition function written as
Z R
Z= Dq(z) e 0 d⌧ L(q) , (3.99)
periodic boundary conditions
with ⌧ = i . If you are surprised by this, remember that we have to Wick-rotate in field
theory as well. In fact, if you go to infinity, you go exactly back to field theory.
However, here we have periodic boundary conditions, because we start from x and
want to return to the exact same point x, q(0) = q( ).
In statistical quantum field theory we have the partition function
Z R R 3
Z = tre Ĥ
= D e 0 d⌧ d xL[ ] (3.100)
P BC
where we have to take the trace over all states and Hamiltonian (which is dependent on
the field). Otherwise, nothing else has changed.
The periodic boundary conditions (PBC) are given by (x, ⌧ = 0) = (x, ⌧ = ). So,
statistical field theory is nothing more than restricting your time interval.
In the zero T limit, ! 1, we recover the standard Wick-rotated field theory. With
the partition function we can compute many things: temperature, entropy, etc.
Here, however, we want to compute something else, the vacuum expectation value.
How do you compute the vacuum expectation value in field theory with given Z. We take
the functional derivative in respect to sources, in other words, from the effective potential.
So nothing changes if we want to compute the vacuum expectation value.
Effective Potential At Zero Temperature Given a scalar field theory defined by

Z
Z=e iW [J]
= D ei[S( )+J ] , (3.101)
R
with J = d4 xJ(x) (x). Then the vacuum expectation value is given by
R
D ei[S+J ]
h0iJ = R . (3.102)
D ei[S+J ]
For a physical theory we would later put J = 0. In the presence of sources we have
W
h (x)iJ = h0| |0iJ = . (3.103)
J(x)
This means that we have reduced the task of computing the vacuum expectation value to
computing the path integral, which, unfortunately, is rather hard.
Given a functional W of J we can perform a Legendre transform to obtain a functional
of h i. Legendre transform is just the fancy term for the simple relation
Z
[h i] = W [J] d4 xJ(x)h (x)i (3.104)
which implies the simple equation

[h i]
= J(x) (3.105)
h i(x)
and for J = 0,
[h i]
= 0. (3.106)
h i(x)
Expanding the functional [h i] gives
Z
[h i] = d4 x[ Vef f (h i) + Z(h i)(@h i)2 + . . . ] (3.107)
which yields
Vef0 f (h i) = J (3.108)
and without external sources
Vef0 f (h i) = 0. (3.109)
In other words, the vacuum expectation value of h i in the absence of an external source
is determined by minimizing Vef f (h i).
All of these formal manipulations are not worth much R if we cannot evaluate W [J].
In fact, in most cases we can only evaluate exp iW [J] = D exp i[S[h i] + Jh i] in the
steepest descent approximation, namely the solution of
R
[S( ) + d4 yJ(y) (y)]
| s =0 (3.110)
(x)
or more explicitly,
@ 2 s (x) + V 0 [ x (x)] = J(x). (3.111)
We write the dummy integration variable as = s + ˜ and expand to quadratic order
in ˜ to obtain
Z
i i
Z = exp W [J] = D exp [S[ ] + J ] (3.112)
~ ~
Z R 4 1
i i ˜2 00 ˜2
⇡e ~
[S[[phis ]+J s ]]
D ˜e ~ d x 2 [(@ ) V ( s ) ] (3.113)
✓ ◆
i 1
= exp [S[ s ] + J s ] tr log[@ 2 + V 00 ( s )] . (3.114)
~ 2
Now we have determined
i~
W [J] = [S[ s ] + J s] + tr log[@ 2 + V 00 ( s )] + O(~2 ), (3.115)
2
and need to calculate:
W [S[ s ] + J s] s
h i= = + s + O(~) = s + O(~). (3.116)
J s J
To leading order in ~, h i is equal to s. Thus we obtain
i~
[ ] = S[ ] + tr log[@ 2 + V 00 ( )] + O(~2 ). (3.117)
2
Nice though this formula looks, in practice it is impossible to evaluate the trace for
arbitrary (x): We have t ofind all the eigenvalues of the operator @ 2 + V 00 ( ), take their
log, and sum. Our task simplifies drastically if we are content with studying [ ] for
independent of x, in which case V 00 ( ) is a constant and the operator @ 2 + V 00 ( ) is
translation invariant and easily treated in momentum space,
Z
tr log[@ + V ( )] = d4 xhx| log[@ 2 + V 00 ( )]|xi
2 00
(3.118)
Z Z
d4 k
= dx 4
hx|kihk| log[@ 2 + V 00 ( )]|kihk|xi (3.119)
(2⇡)4
Z Z
d4 l
= dx 4
log[ k 2 + V 00 ( )], (3.120)
(2⇡)4
where we took the Fourier transformation and used the plane wave hk|xi = eikx . The result
is not divergent, but still infinite. We can do renormalization and regulate the integral,
impose counter-terms to cancel the divergences, but a finite piece will be remaining. Since
we are talking about the effective potential we can always add constants (global shifts)
to the effective potential,
Z  2
i~ d4 k k V 00 ( )
Vef f ( ) = V ( ) log + O(~2 ), (3.121)
2 (2⇡)4 k2
and get the result known as the Coleman-Weinberg effective potential.
The effective potential at nonzero temperature In quantum field theory at zero

temperature, the expectation value c of a scalar field (also referred to as the classical
field) is determined by minimizing the effective potential V ( c ). The effective potential
contains a tree-level potential term, which can be read off from the Hamiltonian, and
quantum corrections form various loop orders. The one-loop quantum correction is calcu-
lated from various loop orders. The one-loop quantum correction is calculated by shifting
the fields by their expectation values c and isolating the terms Lquad ( c , ˜) in the
Lagrangian which are quadratic in the shifted fields ˜. If we write
V ( c ) = V0 ( c ) + V1 ( c ) (3.122)
where V0 is the tree-level contribution and V1 is the one-loop quantum correction then,
for a single scalar field,
✓ Z ◆ Z ✓ Z ◆
exp 4 ˜ 4
i d xV1 ( c ) = D exp i d xLquad ( c , ) ˜ (3.123)
R
where D ˜ denotes a path integral.
At finite temperature, scalar fields (t, x) are replaced by fields (⌧, x) periodic in ⌧
with period , where is given by T 1 . We now write the finite-temperature effective
potential Ṽ ( c ) as
Ṽ ( c ) = Ṽ0 ( c ) + Ṽ1 ( c ) (3.124)
where Ṽ0 and Ṽ1 are the tree-level and one-loop terms and the expectation value c is now
a thermal average. Then we get
✓ Z Z ◆ Z ✓Z Z ◆
exp 3
d⌧ d xṼ1 ( c ) = ˜
D exp 3 ˜
d⌧ d xLquad ( c , ) . (3.125)
0 periodic 0
If gauge fields and fermion fields are included (but with only scalar fields being given
expectation values to avoid breaking Lorentz invariance), then path integrals over gauge
fields and their associated Fadeev-Popov ghosts and over antiperiodic fermion fields are
included as well.
The finite-temperature Lagrangian of the U (1) Higgs model is given by
1 1
L = (Dµ )(Dµ )⇤ m2 ⇤
Fµ⌫ F µ⌫ (@µ Aµ )2 + @µ ⌘ ⇤ @ µ ⌘ + V ( ), (3.126)
4 2f
and the potential of the usual Higgs model,
V( )= (| |2 v 2 ). (3.127)
16
We then get for the effective potential to one-loop order,
4⇡ 2 T 4 1 2 CT
Vef f = + m (T ) 2
c
3
c + 4
(3.128)
90 2 3 16
Figure 3.6: Td > Tc > Tb > Ta
with e4 and
+ 3e2 2
2
m (T ) = m + 2
T (3.129)
12
and
✓ ◆ 32 ✓ ◆ 32
3
4⇡C ⇡ + + 3e3 . (3.130)
4 4
We observe that m2 (T ) can be positive or negative at zero energy. However, at high energy
it will be much smaller.
On the figure, we observe on Fig. 3.6 that we have phase transitions, jumping from
one minimum to another, and perhaps, finally, to the global one, as the vacuum relaxes.
These are transitions from symmetric system down to system with spontaneously broken
symmetries.
This is a viable way of breaking down the models of the unvierse. It cools down
and dynamical symmetry breaking occurs. The consequences is that it can give rise to
topological objects and extended field configurations. We found earlier that they exist and
have finite energy. Recall that we studied kinks (domain walls), vortices (cosmic strings)
and magnetic monopoles earlier.
The kink was an interpolation between two points with h i = v and h i = v (see
subsection 3.1). In a three-dimensional geometry, the kink is interpreted as domain wall.
It is of interest that computing the energy over the line is equivalent to computing the
energy density on the surface of the domain wall, and we remark that the energy of
solitons can be computed. We have an upper constraint by the size of the universe too.
Knowing the energy and the limit of the size, we can estimate the energy density due to
the domain walls. However, it turns out that this rough estimate is too much for domain
walls, vortices or magnetic monopoles.
One possible solution might be that those topological defects would not have formed,
although topology tells us that they do, or that they might have existed at one point in
the universe. During an inflation, however, all those densities were diluted, we then can
compute the gravitational effects and discuss phenomenological implications. This small
density also suggests that we had an inflationary period in our universe.
3.6 Baryon and Lepton Number Violation

Baryon and Lepton number are considered, in a generalised sense, to be charges, which
are space integrals of densities. So we start with the definition of baryonic and leptonic
currents:
1X X
jBµ = (q̄L µ qL + ūR µ uR + d¯R µ dR )jLµ = (¯l µ l + ēR µ eR + ⌫¯R µ ⌫R ) (3.131)
3
with l doublett and fr singletts, such that
Z
Baryon Number : B = d3 xjB0 (3.132)
Z
Lepton Number :L = d3 xjL0 . (3.133)
We know that electroweak only couples to the left-handed components, not to the right-
handed ones. With this, it contains an axial vector coupling. This may lead to a chiral
anomaly. In fact, both currents are non-conserved because of the electroweak (SU (2)L )
anomaly. We are coupling here to the left-handed field strength tensor. Analogously for
leptons.
The anomalies are given by,
Nf 2 a µ⌫,a
@µ jBµ = g Fµ⌫ F̃ , (3.134)
16⇡ 2
Nf 2 a µ⌫,a
@µ jLµ = g Fµ⌫ F̃ , (3.135)
16⇡ 2
Fµ⌫a
= @µ W⌫a @⌫ Wµa + g"abc Wµb W⌫c (3.136)
1
a
F̃µ⌫ = "µ⌫⇢ F ⇢ ,r (3.137)
2
3.7. SPHALERONS 67
We observe, however, that B L is conserved while B + L is violated,

B L : @µ (jBµ jLµ ) = 0 (3.138)
Nf
B + L :@µ (jBµ + jLµ ) = 2 gF F̃ . (3.139)
8⇡
This statement is that B +L is violated, as seen in the context of instantons in QCD, while
B L has to be conserved. Then we can construct field configurations where instantons
might exist. So, there might be electroweak instantons mediating this anomaly.
However, if we look at transition rates of the B + L violation with (B + L) = 2Nf n
with n topological charge, we get
16⇡ 2
⇠ exp i 2 ⇠ l 170
(3.140)
g
which is an negligible value. This rate is far too small. However, we have oversimplified
our discussion in an important point: the nature of electroweak interactions. We held the
discussion analogously to the discussion of unbroken gauge theories, however electroweak
theory is a spontaneously broken gauge theory with the Higgs mechanism present. The
Higgs does not charge B or L, but has SU (2) charge which indicates that there may exist
other, not like instanton, vacuum configurations which might be baryon number violating
which, we hope, will provide a stronger violation.
3.7 Sphalerons
We will now discuss vacuum configurations of an SU (2) model with Higgs field,
1 a aµ⌫
L= F F + |Dµ |2 V( ) (3.141)
4 µ⌫
with
ig a a
Dµ = (@µ ⌧ Wµ ) (3.142)
2
1 2 2
V( )= ( †
v ), > 0. (3.143)
2
How do these vacuum configurations look like? This discussion is different from the
usual treatment of the Higgs mechanism. We are interested in more general solution: In
the presence of the Higgs potential, do new extended field configurations exist for the W
field?
Proceeding in the usual manner, we use the Hamiltontian approach: Choose gauge
Wµa = 0
1 1 a aij
L = (Ẇia )2 F F |Di |2 V( ) (3.144)
2 4 ij
and canonical variables,

L
⇡Wi = = Ẇi (3.145)
Wi
⇡ = ˙⇤ (3.146)
⇡ ⇤= ˙ (3.147)
which yields the Hamiltonian density (which is positive definite)
H = ⇡Wi Ẇi + ⇡ ˙ + ⇡ ⇤ ˙ ⇤ L (3.148)
1 1
= (Ẇi )2 + | ˙ |2 + Fija F aij + |Di |2 + V ( ) (3.149)
2 4
and we get the vacuum solution
Ẇia = 0, ˙ = 0, F 2 = 0, Di = 0, V ( ) = 0. (3.150)
ij
This solution exists and is realized by a pure gauge configuration, where a hatted operator
acts in SU (2) space, is:
2i ˆ ˆ 1 (x),
Ŵi (x) = Wia ⌧ a = (@i ⇤(x))⇤ (3.151)
g
✓ ◆
v ˆ 0
(x) = p ⇤(x) , (3.152)
2 1
where the gauge function ⇤ ˆ variies in space. Ŵi (x) would not be a solution alone. By in-
troducing the Higgs field (x) we get a stable topologically non-trivial gauge field solution
by the interaction of those two.
How do we classify the solutions? They are, again, classified by topology and we use
the same definition of the charge as in QCD. Thus, the topology of field configurations is
classified by topological charge Qk . We have time-independent solutions of the Yang-Mills
equations of SU (2) changing topological charge. It shifts B + L number by 2Nf units (the
same situation as in the axial anomaly in QED).
In other words, sphalerons are static solutions of field equations connecting B + L = 0
state at r ! 0 with B + L = 2Nf state at r ! 1. Recall that for instantons we had
t ! 1 instead of the space variable r.
The next question is how do we excite these field configurations. Do they have them-
selves a contribution to the energy? How much energy does it take to generate such a
configuration? It is not clear that they can be generated out of the void. Perhaps there
are barriers like in the case of instantons, in other words, an energy threshold.
In computing the energy, we start with the field equations,
i
(Di Fija ) = g( † ⌧ a (Dj ) (Dj † a
⌧ )) (3.153)
2
v2
2
Di = 2 ( †
) (3.154)
2
3.7. SPHALERONS 69
Figure 3.7: Fermion energy levels for two states with different topological numbers, show-
ing that the state with Q = 1 has unit baryon number.
Figure 3.8: The functions f (r) and h(r) for the sphaleron.
and calculate the solutions,

2i 1
Wia = f (r)(i"iaj xj 2 ) (3.155)
g r
✓ ◆
v ~⌧ · ~x 0
= p h(r)i (3.156)
2 r 1
by connecting a group index with a space index, and we need
h(r ! 0) = 0, f (r ! 0) = 1, h(r ! 1) = 1, f (r ! 1) = 1 (3.157)
Then for a specified field configuration, we can calculate the topological charge,
1
, (3.158)
2
and the mass,
Z
8⇡
Msph = d3 xH = cmW 2 ⇠ 2 TeV2 (3.159)
g
where c is O(1), and c = 1.6 for ! 0 and c = 2.7 for ! 1. It mediates transitions
across a potential barrier between B + L = 0 and B + L = 2Nf vacua with the height of
this potential barrier given by the mass, ⇠ Msph . The transition rate is proportional to the
Msph
exponential, ⇠ exp T
. This means that they become significant for high temperature,
T MW . This means that these sphyerons may be created at the LHC which are not
point-like particles, but are intermediate field configurations mediating between different
states. It would be possible, with sufficient energy, to produce a field configuration to
mediate the violation. The signature would not be a resonance, but a multi-particle state
that violates the B + L conservation. One may imagine a tb̄ state which violates for
all generations. They are, however, hard to detect and they are dominant in the sense
that they are equivalent to electroweak at higher energy, where one might see substantial
sphyerons processes at higher energies. They are hard to detect because we have to look
at a global property of the final state.
How does this translate into practice for baryon symmetry.
3.8 Baryogenesis
We have observed a baryon asymmetry,
nB nB̄
⌘= ⇠ 5 · 10 10
(3.160)
n
where n are the photons in the cosmic microwave background where the number of
photon and baryon-antibaryon is assumed to be equivalent in the early universe. If we
start with a baryon-antibaryon symmetric universe, we would expect it to be roughly
equal to the number of the photon. However, we observe, that most baryon-antibaryon
have violated. We observe, however, a slight left-over.
We try to formulate necessary conditions for generation of this ⌘ (Sakharov). We find
the condition for our field theory Lagrangian:
• baryon number violating interactions
• C and CP violation (otherwise B excess = B̄ excess = 0)
• universe out of thermal equilibrium at least at one point in its history. Thermal
equilibrium means that globally particle numbers are conserved while locally re-
action rate to creation and annihilation are the same. Otherwise we would have
6B 6B
(X1 X2 ! Y1 Y2 ) = (Y1 Y2 ! X1 X2 ). We find in the standard model for the
SU (2)L
– C violation (only eL and ēR couple)

– CP violation (CKM for quarks)
3.8. BARYOGENESIS 71
– B violation through sphaleron processes, non-perturbative tunneling between

nf
different vacuum states, equivalent to V1 ! V2 = V1 ± ⇧i=1 |uLi dLi dLi ⌫i i. They
change B, L but conserve all other charges. Charges |uL dL dL ⌫i given by
charge uL dL dL ⌫ total
1
T3 2
0.5 .5 .5 0
Y 1/3 1/3 1/3 1 0
Q 2/3 1/3 1/3 0 0 (3.161)
B 1/3 1/3 1/3 0 1
L 0 0 0 1 1
B L 1/3 1/3 1/3 1 0
where we observe that B L conserved and B +L violated. sphaleron processes in thermal

equilibrium for T MW . If in thermal equilibrium, statistical rate of lowering and raising
are the same and they are thus not able to generate a B, L violation. So we have to start
with a non-vanishing B L at high energy in order to reshuffle by thermal processes. So,
we ned either of:
• need mechanism to generate B L anomaly at high temperatures
• baryon asymmetry generated below T ⇠ MW (electroweak baryogenesis)
Sphaleron processes in thermal equilibrium do not change the global number. However,
if we start with a B L asymmetry at high temperature, the sphareons would redistribute
that asymmetry and we could use it, e.g., to redistribute a leptonic to baryonic. Only for
energies below mW these asymmetry can be generated.
We know the physics up to the scale MW pretty well and have constraints by C, CP
violation as well. In other words, we have to look into electroweak baryogenesis, worked
out in detail with standard model constraints (with Higgs) and has been ruled out because
it was too low for the baryon asymmetry by the order of 102 .
This leaves us with baryogenesis from leptogenesis. Now, if we have
hBi = 0, hLi =
6 0 ) hLi = hB Li at high T (3.162)
Then sphaleron processes redistribute in thermal equilibrium hB Li =

6 0 between baryons
and leptons.
The constraints on baryon number and CP violaton in the baryonic sector are much
stronger than those in the leptonic sector. At present, the CKM mechanism is the only
we have for CP violations. We have much weaker constraints on the leptonic sector. In
fact, CP violation has not been observed in the leptonic sector. One can postulate by
ultra heavy Majorana-neutrinos.
6 0 from CP violation in decay of ultra-heavy Majorana-neutrinos (N = N̄ ).
hLi =
Then we get the total reaction rate,
tot
+
= (Ni ! H + lL ) + (Ni ! H + lR ) ⇠ (hh† )ii Mi , (3.163)
where the reactions are out of equilibrium for a temperature below the mass of the
Majorana-Neutrino, T ⌧ Mi . Then we have
• C violated
• CP violated if imh 6= 0 for a Higgs (not necessarily standard model Higgs)

Chapter 4
(Quantum) Fields in Curved Spacetime
Literature:
• L. Bergstrom, A. Goober: Cosmology & Particle Astrophysics
We have a number of unexplained phenomena in the standard model of cosmology

observed by experiments. They include:
• Absence of GUT monopoles and topological defects:

The production rate of GUT monopoles and its flux can be computed. However, the
computed quantity is of several orders of magnitudes larger than current observation.
• Flatness of the universe:

Measured by red-shift of distant supernovae, size of angular defects in cosmic mi-
crowave background, etc.
• Homogeneity of the universe:

From the cosmic microwave background we have measured a homogeneity of the
order TT ⇡ 10 5 .
These phenomena suggest an epoch of exponential expansion of the universe induced

by vacuum energy, also known as ’inflation’. One of the results of inflation is the diluted
density of monopoles as well as the homogeneity of the universe. It is assumed that during
the inflation epoch a small thermally connected region is being expanded exponentially
which yields in flat universe after the expansion.
73
74 CHAPTER 4. (QUANTUM) FIELDS IN CURVED SPACETIME
4.1 Evolution of the Universe
We discuss the evolution of the universe by applying Einstein’s field equations (we have
absorbed the cosmological constant in the matter density Tµ⌫ ),
1
Rµ⌫ gµ⌫ R = 8⇡GTµ⌫ , (4.1)
2
using the Robertson-Walker metric,
✓ ◆
dr2
2
ds = dt 2 2
a (t) 2
+ r2 d✓2 + r2 sin2 ✓d 2
, (4.2)
1 kr
to an homogeneous, isotropic medium (due to Friedman-Robertson-Lemaitre-Walker,

FRLW). The matter density for such an homogeneous, isotropic medium is given by,
Tµ⌫ = (p + ⇢)uµ u⌫ p gµ⌫ , (4.3)

dxµ
with p pressure, ⇢ density, uµ = d⌧
flow-velocity and uµ = t (1, ~0) in comoving frame.
We discuss different media categorized by
• vacuum with p⇤ = ⇢⇤ and

⇤
Tµ⌫ = diag(⇢⇤ , ⇢⇤ , ⇢⇤ , ⇢⇤ ), (4.4)
• (non-relativistic) diluted matter, pm = 0, given by

m
Tµ⌫ = diag(⇢m , 0). (4.5)
• radiation with pr = ⇢r
3
and
⇢r ⇢r ⇢r
r
Tµ⌫ = diag(⇢r , , , ). (4.6)
3 3 3
We arrive at the Friedman equations with the only non-zero components given by
✓ ◆
ȧ 3K
for µ = ⌫ = 0 3 + 2 = 8⇡G⇢, (4.7)
a a
✓ ◆2
ä ȧ K
µ = ⌫ = (1, 2, 3)2 + + 2 = 8⇡Gp. (4.8)
a a a
We now proceed by discussing several types of different universes:

4.1. EVOLUTION OF THE UNIVERSE 75
Flat and vacuum dominant universe In a flat, K = 0, and vacuum dominant,

⇢ = ⇢⇤ , universe we get a Hubble Parameter H(t) := ȧa characterised by the differential
equation
✓ ◆2
ȧ ⇤
= = [H(t)]2 , (4.9)
a 3
which corresponds to
dH(t)
= 0, (4.10)
dt
which has the solution of an exponential growth
a(t) = a0 exp Ht. (4.11)
Flat universe with radiation or matter In a flat universe (k = 0) with radiation

or matter we have the following differential equation expressing energy conservation in a
given volume dV ⇡ a3 ,
dE = pdV (4.12)
d d
(⇢a3 ) = p a3 . (4.13)
dt dt
We find for the universe with radiation,
ȧ ⇢0
⇢˙ + 4⇢ = 0 ) ⇢ = 4 , (4.14)
a a
and matter,
ȧ ⇢0
⇢˙ + 3⇢ = 0 ) ⇢ = 3 . (4.15)
a a
Applying these results in the Friedmann equation,
8⇡G⇢(t) 3
ȧ2 = a, (4.16)
3
yields for the radiation dominated universe,
✓ ◆ 12
t
a(t) = (4.17)
t0
⇢0 t2
⇢(t) = 3 0 (4.18)
t
and for the matter dominated,
✓ ◆2/3
t
a(t) = (4.19)
t0
⇢0 t20
⇢(t) = 2 . (4.20)
t
Universe with curvature For a universe characterised by curvature k = ±1, the

equation
✓ ◆2
ȧ 8⇡G⇢ k
2
H (t) = = (4.21)
a 3 a2
3H 2 (t)
has a fix point solution. When defining the cirtical density ⇢c = 8⇡G
and normalizing
up to ⌦i = ⇢⇢ci we find
⇢ ⇢c 3k
= . (4.22)
⇢ 8⇡G⇢a2
From present day observation, however, we have
⇢ ⇢c
⇡ O(1). (4.23)
⇢
When summarizing the evolutiong of the different epochs for ⇢ ! ⇢c ,

⇢ ⇢c
vacuum dominated: ⇡e 2Ht
, (4.24)
⇢
⇢ ⇢c
radiation dominated: ⇡ t, (4.25)
⇢
⇢ ⇢c
matter dominated: ⇡ t2/3 , (4.26)
⇢
we note that in the radiation and matter dominated epochs any deviation from ⇢ = ⇢c is
amplifed by time.
4.2 Inflation
We will first discuss this epoch in a classical field theoretic framework. Later, we modify
it to discuss it in a quantum field theory picture.
The field theoretic model for an inflationary epoch in the early universe (chaotic in-
flation, A. Linde).
• at t ⇠ 3 ⇠
= tpl 5.4 · 10 44 s, have an area of size lpl = (1.63 · 10 35 m)3 filled with a
homogeneous scalar field (inflaton) with potential V ( ). The action is given by
Z ✓ ◆
p 1 µ⌫
S= 4
gd x g @µ @ ⌫
V( ) (4.27)
2
4.2. INFLATION 77
and homogeneity is expressed by ⌧ V ( ). This means that we can apply FRLW

metric and find
Z Z
3 1 ˙2
S = dta ( V ( )) = dtL. (4.28)
2
This Lagrangian allows us to find classical equations of motions for ,
d @L @L
=0 (4.29)
dt @ ˙
@
ȧ @V
) ¨+3 ˙ = (4.30)
a @
is a classical motion in potential V , with a time-dependent friction term 3 ȧa = 3H.
We calculate the content of the Friedman equations with k = 0,

1
⇢ = T00 = ˙ 2 + V ( ), (4.31)
2
✓ ◆2
ȧ 8⇡ 1 ˙ 2
= ( + V ( )), (4.32)
a 3Mpl2 2
where ⇢ behaves like a vacuum energy density, then varies only slowly with time; i.e.
ȧ ˙ ¨ and V ( ) ˙ 2 (slow-roll conditions).
a
With these slow-roll conditions our equations of motions simplify to
ȧ @V
3 ˙= (4.33)
a @
✓ ◆2
ȧ 8⇡
= V( ) (4.34)
a 3Mpl2
and we have for large V ( ) a large H and thus the slow-roll condition. Looking at the
dynamical evolution of the field we observe that for some time there is a violation. How
large?
1 Mpl2
˙ 2 ⇠ (@ V )2 ⇠ (@ V )2
. (4.35)
(ȧ/a)2 V
We then have for the slow-roll condition

V
@ V ⌧ (4.36)
Mpl2
V
@ V ⇠ for any polynomial potential (4.37)
and finally
Mpl (4.38)
We see that our slow-roll conditions are fulfilled when the field is much larger than the
Planck-Mass.
However, it does not mean that the mass energy density beyond the Planck-density
⇢pl ⇠ Mpl2 , since V = 4
V ( ⇠ Mpl ) Mpl4 ⌧ Mpl4 for ⌧ 1. (4.39)
In other words, the start of inflation V ( in ) ⇠ Mpl4

1/4
in ⇠ Mpl Mpl
and our of inflation for out ⇠ Mpl yields the end of slow-roll and we have a damped
oscillations around = 0 (reheating).
The moment where it ends it sets the initial conditions for our universe, i.e. the curva-
ture. In late-time epoch of inflation we want annihilation of inflaton and creation of SM
particles.
4.3 Quantum theory of the inflaton

The free Lagrangian of the massive scalar inflaton field in curved spacetime is given by,
1p
L= g(g µ⌫ (x)@µ @⌫ m2 2
) (4.40)
2
p
where we use the FRLW metric gµ⌫ = diag(1, a2 , a2 , a2 ) such that g = a3 . We
use the Euler-Lagrange equations,

@ @L @L
µ µ
= 0, (4.41)
@x @(@ /@x ) @
to obtain the equation of motion,
[@g2 + m2 ] (x) = 0 (4.42)
where
1 p
@g2 (x) = g µ⌫ Dµ D⌫ (x) = p @µ [ gg µ⌫ (x)@⌫ (x)]. (4.43)
g
Rewritten in the FRLW metric,
¨ + 3 ȧ ˙ + m2 (x) = 0. (4.44)
a a2
4.3. QUANTUM THEORY OF THE INFLATON 79
We now evaluate how the field evolves for different behaviour of the scale factor a(t).
For an generic inflationary period with a nearly constant Hubble parameter, H = ȧ/a ⇠
const, and the exponential scale factor,
a(t) = a0 exp H(t t0 ). (4.45)
Using conformal time,

Z t
dt 1 1
⌘= = +c= , (4.46)
a(t) a0 H exp H(t t0 ) a(⌘)H
we can rewrite our metric as being proportional to the Minkowski metric,
ds2 = a2 (⌘)[d⌘ 2 d~r2 ]. (4.47)
To quantize the field, we calculate the canonical momentum,
@L @
⇡ = @
= a2 (⌘) , (4.48)
@( @⌘ ) @⌘
and impose the canonical quantization condition at equal conformal time,
@
[ , ⇡]⌘ = a2 (⌘)[ (⌘, ~r), (⌘, ~r0 )] = i (3)
(~r ~r0 ). (4.49)
@⌘
We then compute the mode expansion in terms of momentum eigenfunctions,
Z
d3 k ~ ~
(⌘, ~r) = 3/2
[a~k ~k (⌘)eik·~r + a~†k ~⇤k (⌘)e ik·~r ] (4.50)
(2⇡)
to find that it fulfills the quantization condition if

⇤
@ ~k @ ~k
2
a (⌘)( ~k ~k ) = i. (4.51)
@⌘ @⌘
Here we have the problem that ⌘ is proportional changing with time. So, the dif-
ferential equations determing our eigenfunctions are much more complicated than the
Klein-Gordon equations because we have an explicit dependence on the parameter ⌘. The
problem is finding momentum eigenfunctions, taking into account explicit dependence on
⌘.
To find the momentum eigenfunctions we use the ansatz
1
~k (⌘) = ~k (⌘) (4.52)
a(⌘)
in the equation of motions,

✓ ◆
m2 ä
¨~k + ~k =0 (4.53)
H 2⌘2 a
with ä
a
= 2
⌘2
. This differential equation has a solution in terms of Hankel functions,
p
~k = ⌘ c1 H⌫(1) ( k⌘) + c2 H⌫(2) ( k⌘) (4.54)
(1) ⇤ (2) m2
with H⌫ = H⌫ and ⌫ 2 = 9
4 H2
.
Looking at asymptotic behavior at early times we find,
r
2
~k (k⌘ ! 1) ! (c1 e ik⌘ + c2 eik⌘ ) (4.55)
⇡k
which tells us what positive and negative frequency modes at early times were. In ordinary
field theory one would then define the asymptotic frequencies as in and out states for any
given times. Here we cannot use this any more. However, we can use that they do not
mix at asympotically early or late
p times and thus derive the inflaton field operator, with
positive frequency modes c1 = 2⇡ , c2 = 0 and vice-versa. It then yields the inflaton field
operator,
p Z
⇡⌘ d3 k ~ ~
= 3/2
[a~k H⌫(1) ( k⌘)eik·~r + a~†k H⌫(2) ( k⌘)e ik·~r ] (4.56)
2a(⌘) (2⇡)
This is a new feature of fields on curved spacetime because it yields the creation of the
standard model particles. However, it does not allow to determine feedback on the curved
background; in other words, it is basically statically curved.
Bibliography
[1] Anastasiou, Quantum Field Theory 2, Lecture Notes at ETH Zurich 2008.
[2] Bailin, Cosmology In Gauge Field Theory And String Theory, IOP Publishing 2004.
[3] Bergstrom, Cosmology and Particle Astrophysics, Springer 2004.
[4] Bertlmann, Anomalies in Quantum Field Theory, Oxford University Press 1996.
[5] Coleman, Aspects Of Symmetry, Cambridge University Press 1985.
[6] Collins, Renormalization, Cambridge University Press 1984.
[7] Donoghue, Dynamics Of The Standard Model, Cambridge University Press 1992.
[8] Ecker, Chiral Perturbation Theory, arXiv:hep-ph/9501357
[9] Frampton, Gauge Field Theories, John Wiley & Sons 2000.
[10] Fukugita, Physics of Neutrinos and Application to Astrophysics, Springer 2003.
[11] Huang, Quarks Leptons & Gauge Fields, World Scientific Publishing 1982.
[12] Kaku, Quantum Field Theory, Oxford University Press 1993.
[13] Peskin, Introduction To Quantum Field Theory, Perseus Book Publishing 1995.
[14] Ryder, Quantum Field Theory, Cambridge University Press 1996.
[15] Srednicki, Quantum Field Theory, Cambridge University Press 2007.
[16] Weinberg, The Quantum Theory Of Fields II: Modern Applications, Cambridge Uni-
versity Press 1996.
[17] Yndurain, The Theory Of Quark And Gluon Interaction, Springer 2006.
[18] Zee, Quantum Field Theory In A Nutshell, Princeton University Press 2003.
81

Aft 2010 PDF

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Aft 2010 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aft 2010 PDF

Uploaded by

Copyright:

Available Formats

Advanced Field Theory

ETH Zurich and University of Zurich

Prof. Dr. Babis Anastasiou [ETH Zurich]

June 28, 2010

Please feel free to send feedback or error reports to [email protected]

1 Dimension Regularization and the Axial Anomaly 1

2 Chiral Symmetry in the Strong Interaction 15

3 Topological Aspects of Field Theory 43

4 (Quantum) Fields in Curved Spacetime 73

Dimension Regularization and the

One example of a UV-divergent integral is,

where we identify in the vertex diagram k ! p1 , p2 , p21 = p22 = 0, k = p1 + p2 . We observe

1. The soft limit is defined as the limit of k ! k, ! 0:

which is obviously logarithmically divergent.

2. In the collinear limit we have k 2 = 0 and k k p1 , k k p2 , use the Sudakov parametriza-

We can thus decompose

and then using k2? = k12 + k22 continue with

to change our measure,

d4 k = 2dk+ dk d2 kT = dk+ dk dk 2 . (1.7)

which again is logarithmically divergent.

1.1 Basics of Dimensional Regularization

1. Linearity: For any a, b 2 :

2. Scaling: For any number s

3. Translation invariance: For any vector q:

We will also require rotational covariance of our results.

1.2 Cliﬀord Algebra

The appropriate definition changes when we go to two dimensions: Instead of 5 we

Lorentz and Dirac-Algebra We require gµµ = d, so a vector Aµ has d 2 physical

In dimensional regularization we only care about limd!4 tr = 4. The consequence is

• introduce (d 4)-dimensional metric ĝµ⌫ = ĝ⌫µ , which acts as projector on general-

– ĝµ⌫ g µ ⇢ = ĝ⇢⌫ = ĝµ⌫ ĝ µ ⇢ (projects on (d 4) subspaces)

• define product of "-tensors (4 dim!)

1.3 Chiral Symmetry

Noether currents and conserved charges

• vector current J µ = ¯ µ R is conserved. @µ J µ = 0 (even for m 6= 0). This implies

• Axial vector current J5 µ = ¯ µ

Question What is the symmetry for J5 µ ?

1.4 QED Axial Anomaly in Two Dimensions

In two dimensions, this set of relations can be represented by 2 ⇥ 2 matrices; we choose

The Dirac spinors will be two-component fields.

Then, just as in four dimensions, there are two possible currents,

and both are conserved if there is no mass term in the Lagrangian.

In the free theory, the field equation of +, would be

i(@0 + @1 ) + = i(@0 @1 ) = 0; (1.29)

Figure 1.1: Computation of hj µ i in a background electromagnetic field.

hj µ5 (q)i = "µ⌫ hj⌫ (q)i (1.33)

"µ⌫ Fµ⌫ = 2@µ ("µ⌫ A⌫ ). (1.38)

+ : En = +(kn eA1 ), (1.42)

1.5 QED Axial Anomaly in Four Dimensions

under the symmetrical integration, we can evaluate

Notice the behaviour in which a logarithmically divergent integral contributes a factor

Chiral Symmetry in the Strong

2.1 Chiral Symmetry of QCD

Then we can write the currents associated with these symmetries as

jLµ = Q̄L µ QL , jRµ = Q̄R µ QR , (2.6)

Spontaneous Breaking of Chiral Symmetry Before we describe the consequence

Using this equation with the on-shell parametrization, we find

h0|@µ j µ5a (0)|⇡ b (p)i = p2 f⇡ ab