Você está na página 1de 91

Anomalies in Quantum Field Theories

Andrew James Bruce

October 6, 2005
Preface
The main purpose of this thesis is to introduce the various anomalies that arise in quantum
field theories; in particular their connection with topology and geometry. A theory is said to
be anomalous when a classically conserved current is no longer conserved upon quantisation
of the theory. More precisely, if there exists no regularisation procedure that preserves all
the classical symmetries then the theory is said to have an anomaly. A better name for such
anomalies would be quantum mechanical symmetry breaking as all anomalies arise as one
loop corrections to the classical conservation laws.

Anomalies shed much light on the deep nature of quantum field theories. In particular,
anomalies were first calculated within perturbation theory but later shown to be related to
the global topology of the theory via index theorems.

This thesis uses the modern formulation of differential geometry and topology to describe
the various anomalies that can arise in a quantum field theory. Much emphasis is placed on
the methods initiated by Stora, Zumino and Wess [74, 77] who found an algebraic formula-
tion known as the descent equations. We set up the descent equations in Gauge theory via
geometric and topological arguments in particular the Atiyah-Singer index theorem and the
BRST transformations.

It is then explained how the descent equations are related to BRST cohomology and how
one can approach the problem of finding the anomalies in this set up.

The reader is expected to have a good grasp on the basics of quantum field theory, espe-
cially functional methods and perturbation theory. Knowledge of differential geometry and
algebraic topology is assumed. A familiarity with general relativity would also be useful.

The following books were found useful in preparing this thesis; [20, 34, 44, 45, 58, 53, 54,
55, 60].

i
Acknowledgements
I would like to thank Dr. P. M. Saffin for support and direction in the undertaking of this
thesis. I would also like to thank Drs. M. B. Hindmarsh and A. Tranberg who have both
enhanced my understanding of quantum field theory. I also wish to thank all the members
of the particle theory group at the University of Sussex all of whom have made my studies
enjoyable.

Andrew James Bruce

Sussex.

ii
Conventions and Notation
Throughout we use natural units where c (speed of light) = ~ (Planck’s constant) = 1, un-
less otherwise explicitly stated. Einstein’s summation convention is employed: if the index
appears twice, once as a superscript and once as a subscript, then the index is summed over
all possible values. For example if µ runs over 0 to m, we have,

m
X
Aµ Bµ = Aµ Bµ .
µ=0

The Minkowski metric is given by gµν = ηµν = diag(1, −1..., −1), while the Euclidean metric
is gµν = δµν = diag(+1, +1..., +1). Lower case Greek letters will generically denote indices
running from 0 to m and lowercase Latin letters will denote indices running from 1 to m,
where m is the dimension of the manifold

With a Euclidean metric the Dirac Matrices satisfy

γ µ† = γ µ ,
{γ µ , γ ν } = 2δ µν .

Assuming the space-time to be even-dimensional (dimM = m = 2l) we also define

γ m+1 = (i)l γ 1 ...γ m ,


γ m+1† = γ m+1 ,
(γ m+1 )2 = I.

Unless otherwise stated, the metric will be taken to be Euclidean. Under some assumptions
it has been shown that within perturbation theory the results of the Euclidian theory imply
the results of the corresponding Minkowskian theory. The Minkowskian theory is formally
obtained via the analytic continuation in the complex plane. This is equivalent to replacing
p0 of the 4-momentum by −ip0 . This is known as a Wick rotation.

iii
CONTENTS

1 Introduction 1

2 The Anomalies of Gauge Theory 6


2.1 The ABJ-Anomaly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 The Singlet Anomaly via Fujikawa’s Method . . . . . . . . . . . . . . . . . . 12
2.3 The Non-Abelian Anomaly via Fujikawa’s Method . . . . . . . . . . . . . . . 18
2.4 The Wess-Zumino Consistency Conditions and the Descent Equations . . . . 22
2.5 Epitome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3 BRST Cohomology and the Descent Equations 26


3.1 Local Cohomology and the Cohomology of d . . . . . . . . . . . . . . . . . . 26
3.2 The BRST Operator and its Cohomology . . . . . . . . . . . . . . . . . . . . 28
3.3 The Cohomology of s mod d . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Some Solutions to the Descent Equations . . . . . . . . . . . . . . . . . . . . 32
3.5 General Solutions to the Decent Equations and the Russian Formula . . . . . 34
3.6 The Bottom Up Approach to Solving the Descent Equations . . . . . . . . . 38
3.7 BRST Cohomology and Physics . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.8 Descent Equations for Nontrivial Gauge Bundles . . . . . . . . . . . . . . . . 40
3.9 Epitome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

4 The Anomalies of Gravity 42


4.1 Chiral U (1) Gravitational Anomalies . . . . . . . . . . . . . . . . . . . . . . 43
4.2 The BRST Algebra for Gravitation . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Pure Lorentz Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4 The Equivalence of Einstein and Lorentz Anomalies . . . . . . . . . . . . . . 48
4.5 The Descent Equations for Gravity via the Russian Formula . . . . . . . . . 49
4.6 Mixed Anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.7 Epitome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

5 The Atiyah-Singer Index Theorem and Anomaly Polynomials 53


5.1 Twisted spin complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 The Rarita-Schwinger spin complex . . . . . . . . . . . . . . . . . . . . . . . 55

iv
5.3 The signature complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.4 The Invariant Polynomials in Various Dimensions . . . . . . . . . . . . . . . 56
5.5 Anomalies in Supergravity Theories . . . . . . . . . . . . . . . . . . . . . . . 57
5.6 Epitome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6 Conclusion 60

Appendices 61

A Non-Abelian Gauge Theory and BRST Symmetry 62


A.1 Classical Non-Abelian Gauge Theories . . . . . . . . . . . . . . . . . . . . . 62
A.2 Instantons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.3 Faddeev-Popov Ghosts and Gauge Fixing . . . . . . . . . . . . . . . . . . . . 66
A.4 BRST Quantisation of Gauge theory . . . . . . . . . . . . . . . . . . . . . . 69
A.5 The Gribov Ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

B Basics of Lie Algebra Cohomology 74


B.1 Lie Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
B.2 Lie Algebra Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
B.3 Chevalley-Eilenberg Cohomology . . . . . . . . . . . . . . . . . . . . . . . . 76
B.4 Lie Algebra Cohomology and BRST Cohomology . . . . . . . . . . . . . . . 76

C The Atiyah-Singer Index Theorem and some Characteristic Classes 78


C.1 Elliptic Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
C.2 Statement of The Atiyah-Singer Index Theorem . . . . . . . . . . . . . . . . 79
C.3 Some Characteristic Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Bibliography 81

v
CHAPTER 1
Introduction

Symmetries and their conservation laws play a huge role in the formulation of modern physics.
This is most acute in the application of quantum field theory to describe the fundamental
forces of nature. However, it is not guaranteed that a classical symmetry will be respected
in the quantum theory.

Indeed, there is little reason to expect the classical symmetry to be a symmetry of the cor-
responding quantum system. Consider the fact that the classical system is on mass-shell
as where the quantum system is necessarily off mass-shell. Hence it is not obvious what
properties of the classical system are shared by the quantum system. A simple example
is that of the vacuum energy of the harmonic oscillator in quantum mechanics. Quantum
mechanically the lowest energy state is greater than zero as where classical it is zero. This
simple and well known fact cannot be seen from the classical Lagrangian and is a quantum
effect.

However, classical symmetries do play an important role in quantum field theories. The most
important classical symmetry in modern quantum field theory is that of gauge symmetry. It
forms the basic backbone of the standard model of particle physics. If one insists on gauge
symmetry in the quantum field theory one is lead to non-trivial constraints on the generating
functionals; known as the Ward-Takahashi identities [70, 67] for QED or the Slavnov-Taylor
identities [63, 68] for Yang-Mills. These identities and hence the gauge symmetry are nec-
essary for the renormalisation of the theory. Without gauge symmetry, renormalisability of
the theory is lost and the theory is inconsistent.

Not all anomalies are as disastrous for the theory as these local gauge anomalies. Generically,
any local symmetry that becomes anomalous makes the theory inconsistent, as where global
symmetries do not place any constraints on the quantum theory. In fact global anomalies
can be phenomenologically welcome.

For example, take the decay of a π meson into two photons. Early calculations by Steinberger
[65] within the pion-nucleon model lead to decay rates that did not agree with experimental
observation.

1
The Feynman graph which represents the decay is the triangle graph in which two photons
couple to an internal fermion loop via two currents. We have the so called V-V-A triangle
which is now known to lead to an anomaly. Bell and Jackiw [19] correctly calculated the
decay rate by taking into account the axial anomaly. Adler [1] discovered a similar result in
QED.

The triangle diagram containing one axial and two vector currents leads to the anomalous
conservation law for the axial current

∂ µ Jµ5 (x) = A(x), (1.1)


where A is the Adler-Bell-Jackiw (ABJ) anomaly

e2 µναβ
A(x) =  Fµν Fαβ . (1.2)
16 π 2
A similar result is found for Yang-Mills fields; Aµ = Aaµ T a and Fµν = Fµν
a
T a . This is the so
called singlet anomaly
1 µναβ
A(x) = ∂ µ jµ5 =  TrFµν Fαβ . (1.3)
16π 2
Bardeen [14] discovered that chiral fermions can lead to the break down of gauge symmetry
for non-Abelian gauge fields.

  
µ 1 1
Ga [Aµ ] = Dµ hj ia = ± 2
Tr Ta κλµν ∂ Aλ ∂µ Aν + Aλ Aµ Aν . (1.4)
24π 2
Where ± corresponds to right and left handed fermions.

It should be noted that the anomalies are not just artifacts of perturbation theory. It was
noticed that the singlet anomaly can be given purely in topological terms by using the
Atiyah-Singer (AS) [7, 8, 9, 10, 11] index theorem. This theorem loosely states that the
index of an operator can be expressed in terms of characteristic classes. The singlet anomaly
is given by
Z
dx A(x) ∝ v+ − v− = ind D+ . (1.5)
The AS index theorem then states that the index of the Dirac operator is given completely
in terms of the Chern character, in 4 dimensions we explicitly have
1
ind D+ = − TrF ∧ F. (1.6)
8π 2
Thus, the singlet anomaly is completely determined on topological grounds only.

Fujikawa [35, 36] noticed that the anomaly would appear as the Jacobian of the path inte-
gral measure. This formulation is non-perturbative in nature and shows most directly the
anomaly’s connection with the global properties of the theory.

Modern differential geometry and algebraic topology can also be used to describe the anoma-
lies. Written as differential forms the singlet and non-abelian anomaly in 4 dimensions are

2
 
1 2
A = 2
d Tr AdA + A3 (1.7)
4π 3
 
1 1 3
Ga [A] = ± TrTa d AdA + A . (1.8)
24π 2 2
The non-abelian anomaly can be used to define the Wess-Zumino consistency condition.
Defining
Z
G[η, A] = η a Ga [A], (1.9)
allows the Wess-Zumino consistency condition to be written as

sG[η, A] = 0. (1.10)
Where we have the Faddeev-Popov ghost η = Ta η a and the BRST operator s which is
nilpotent, s2 = 0. The gauge anomaly is understood as a non-trivial solution to the Wess-
Zumino consistency condition. That is a solution not of the form

G[η, A] = sĜ[A]. (1.11)


Thus, the consistency condition defines the cohomology of s. The local form of the anomaly
can be determined via local cohomology. That is the cohomology defined on the space of
local functionals. The local variables are η, A and F . This local cohomology can be solved
via the so called descent equations

dQ02n+2 = P2n+2
sQ02n+1 + dQ12n = 0
sQ12n + dQ22n−1 = 0
..
.
sQ2n
1 + dQ0
2n+1
= 0
2n+1
sQ0 = 0. (1.12)

P2n+2 is a symmetric gauge invariant polynomial in F . The lower index refers to the form
degree and the upper the power in η known as the ghost degree. Starting from a symmetric
polynomial all the terms in the descent equation can be explicitly calculated.

The term Q12n is identified as the local form of the gauge anomaly
Z
G[η, A] = N Q12n (η, A). (1.13)
M2n

The normalisation N is not fixed by the descent equations. The normalisation has to be
fixed via perturbation theory or careful topological analysis such as k-theory.

The important question is what symmetric invariant polynomial should be used? This de-
pends on the theory and is directly related to the Atiyah-Singer index theorem for the ellipti-
cal operator in question. For example, if one chooses the symmetrised trace as the invariant

3
polynomial which gives the singlet anomaly, the non-abelian anomaly is the result. Thus, the
singlet anomaly in 2n+2 dimensions is related to the non-abelian anomaly in 2n dimensions.

The normalisation of the non-abelian anomaly requires the use of k-theory. Alvarez-Gaumé
and Ginsparg [4, 5] demonstrated how the anomaly in 2n dimensions is related to the AS
theorem in a (2n + 2) dimensional space. Their method avoids the use of k-theory, but in
some sense it closely resembles it. They showed by using two-parameter families of gauge
potentials that the normalisation of the non-abelian anomaly is

in+1
N = −2πi . (1.14)
(2π)n+1 (n + 1)!
The situation is very similar in gravity. Gravitation can be considered as a gauge theory of
diffeomorphisms and Lorentz transformations. Thus, with minor modification the machinery
developed for gauge theory can be applied to gravitation. It turns out that the two anoma-
lies, diffeomorphic and Lorentz are in fact not independent, but are really two expressions
of the same phenomenon.

Outline

The thesis is arranged as follows:

In chapter 2 we explicitly calculate the ABJ anomaly via evaluating the triangle diagram
and Pauli-Villars regularization. Then we examine the singlet and non-abelian anomaly us-
ing Fujikawa’s method and set up the descent equations via the Wess-Zumino consistency
conditions.

We then put the descent equations on strong mathematical ground in chapter 3. We develop
the idea of local cohomology and use it to re-derive the descent equations. Solutions to
the descent equations are presented using several methods. Explicit examples are given for
the Dirac field. Finally in this chapter we set up the descent equations for non-trivial fibre
bundles.

Now armed with the necessary tools to solve the descent equation in general, the question
of gravitational anomalies is tackled in chapter 4. We set up the correct BRST algebra for
gravitational theories with out torsion. It is shown how this resembles the BRST algebra for
gauge theory and so solutions to the descent equations are readily available. Mixed anoma-
lies are briefly introduced in this chapter.

In chapter 5 we address the question of how to pick the invariant symmetric polynomials,
which are also called anomaly polynomials in the literature. We construct the invariant poly-
nomials for the spin-1/2 Dirac, spin-3/2 Rarita-Schwinger and (anti-)self-dual antisymmetric
form fields via the AS index theorem for the relevant complexes. We introduce the necessary
complexes and apply the AS index theorem. Using the so called anomaly polynomials we
show that 10-d supergravity theories are anomaly free.

Chapter 6 we round up the discussion and make our concluding remarks. Several appendixes
have also been included. In appendix A we quickly review Yang-Mills theory and show how

4
the BRST operator arises here. Appendix B outlines the basics of Lie algebra cohomology
and shows how this is related to BRST cohomology. The Atiyah-Singer index theorem and
some characteristic classes are presented in appendix C. These results have been used ex-
tensively throughout the thesis. The bibliography is extensive but by no means exhaustive.
The study of anomalies is huge and as such it would be impossible to list all references on
the subject.

Omissions

Due to time and length restrictions several important topics have been omitted from this
thesis. These include;

1. any detailed discussion about k-theory and the normalisation of anomalies. See [54]
for an accessible account. We do not discuss in any detail the Alvarez-Gaumé and
Ginsparg index procedure. For a nice account see [20].

2. anomalies in the Batalin-Vilkovisky (BV) formulism [16, 17]. See [13, 37].

3. gravity theories with torsion. See [30, 51, 52, 75].

4. anomalies in non-commutative gauge theories. See [25, 47, 50] for a discussion of
anomalies. For an introduction to non-commutative geometry see [27, 48]. For a short
introduction to non-commutative field theory see [12].

5. branes, M-theory and anomaly inflow. See [21, 22, 26, 33].

5
CHAPTER 2
The Anomalies of Gauge Theory

Here we consider the anomalies that can arise in gauge theory. There are two kinds of
anomaly; global chiral and local gauge. The global chiral anomaly is the quantum break-
down of the chiral symmetry (or a modification to it in the massive case). As it is a global
symmetry is does not affect the renormalisation and consistency of the field theory. It is
indeed welcomed phenomenologically.

When we have a chiral theory, that is the left and right handed fermions do not couple to
the gauge fields in the same way we have the possibility of a gauge symmetry being broken
quantum mechanically. Thus, the local gauge anomaly is catastrophic for a consistent quan-
tum field theory and must be cancelled at all costs.

In this chapter we explicitly calculate the so called ABJ chiral anomaly in QED [1, 19] via
examining the triangle graph and use Fujikawa’s method [36, 35] to determine the generali-
sation of this to Yang-Mills theory and to calculate the non-abelian anomaly in chiral gauge
theory.

Fujikawa’s method of determining the anomalies is the most direct way to see their geomet-
ric and topological nature. The anomalies can be seen as the failure of the regularisation
scheme to preserve the classical symmetry. Fujikawa’s method explicitly shows that this
failure occurs in the path integral measure. This method also directly shows the topological
nature of anomalies.

The non-abelian anomaly is also explained as the lack of BRST invariance of the effective
action. This leads to the so called Wess-Zumino consistency condition. Solutions to this
condition can be found via the descent equations.

2.1 The ABJ-Anomaly


Here we examine the Adler-Bell-Jackiw (ABJ) anomaly in QED via direct computation of
the so called V-V-A triangle graph (see figure 1.1). We employ the Pauli-Villars regulisation
[57], which as we explicitly add a massive fermion to the theory directly breaks chiral sym-

6
metry. It should be noted that the anomaly exists for all other regulisation schemes. For
example, when one uses dimensional regulisation we extend the theory to a higher dimen-
sional complex space. As there is no generalistion of γ 5 in these spaces there is no notion of
chirality and so chiral symmetry is spoiled.

We explicitly calculate the chiral anomaly in the simplest gauge theory, QED. Here we
assume a Minkowski metric of signature (1, 1, 1, −1). Consider the fermionic part of the
Lagrangian for QED.

L = ψ(i∂/ − m + eA)ψ.
/ (2.1)
Where we have used the “slashed” notation for ∂/ = γ µ ∂µ and A
/ = γ µ Aµ , with Aµ being the
U (1) gauge field.

We also define three bilinear currents

jµ (x) = ψ(x)γµ ψ(x) − vector


jµ5 (x) = ψ(x)γµ γ5 ψ(x) − axial
P (x) = ψ(x)γ5 ψ(x) − pseudoscalar. (2.2)

Using the Dirac equation we can derive conservation laws



∂ µ jµ = ψ ∂/ ψ + ψ ∂/ψ
/ + iψ(−m + eA)ψ
= iψ(m − eA)ψ /
= 0 (2.3)



∂ µ jµ5 = ψ ∂/ γ5 ψ + ψ ∂/γ5 ψ


= ψ ∂/ γ5 ψ − ψγ5 ∂/γ5 ψ
= iψ(m − eA)γ/ 5 ψ − iψγ5 (−m + eA)ψ
/
= 2imP. (2.4)

Equation (2.4) vanishes in the massless limit and then we classically have both vector cur-
rent and axial current symmetry. This is not the case when we quantise the theory as the
divergence of the axial current is non-zero even in the massless case. One needs to examine
the V-V-A triangle graph which breaks the axial symmetry.

For this we consider a Lagrangian with both vector and axial gauge fields Vµ and Aµ respec-
tively.

L = ψ(i∂/ + V/ + Aγ
/ 5 − m)ψ. (2.5)
The lagrangian (2.5) is invariant under local vector gauge transformations

7
ψ → eiθ(x) ψ
ψ → ψe−iθ(x)
Vµ → Vµ + ∂µ θ(x). (2.6)
There is also a local axial gauge transformation which leaves (2.5) invariant

ψ → eiα(x)γ5 ψ
ψ → ψe−iα(x)γ5
Aµ → Aµ + ∂µ α(x). (2.7)
All together the Lagrangian (2.5) is said to have a local UV (1) × UA (1) symmetry.

q q

p−q p p−q p

k2 p − k1 k1 p − k1 k1
k2

ν µ ν µ

FIGURE 1.1: The V-V-A and V-V-P triangle graphs that lead to the anomaly.

Now, examining the V-V-A and V-V-P triangle graphs we see that the corresponding matrix
elements are
Z
Tµνρ (k1 , k2 , q) = d4 xe−ik1 x1 −ik2 x2 +iqx3 h0|T [jµ (x1 )jν (x2 )jρ5 (x3 )]|0i. (2.8)
Z
Tµν (k1 , k2 ) = d4 xe−ik1 x1 −ik2 x2 +iqx3 h0|T [jµ (x1 )jν (x2 )P (x3 )]|0i. (2.9)
Where we have defined d4 x = d4 x1 d4 x2 d4 x3 . Equations (2.8) and (2.9) are essentially multi-
dimensional Fourier transformations. By taking derivatives we derive the important relations

Z 
q ρ Tµνρ = ∂xρ3 d4 x e−ik1 x1 −ik2 x2 +iqx3 h0|T [jµ (x1 )jν (x2 )jρ5 (x3 )]|0i
Z
= d4 x h0|T [jµ (x1 )jν (x2 )∂xρ3 jρ5 (x3 )]|0ie−ik1 x1 −ik2 x2 +iqx3
Z
= d4 x [−2mh0|T [jµ (x1 )jν (x2 )P (x3 )]|0i] e−ik1 x1 −ik2 x2 +iqx3
= 2mTµν . (2.10)

8
In order to preserve gauge symmetry we insist that ∂µ j µ = 0 which means that

k µ Tµνρ = k ν Tµνρ = 0. (2.11)


Equation (2.10) is known as the axial Ward identity and (2.11) is the vector Ward identity.
Equation (2.11) can be relaxed, leading to an anomaly in the vector current. However, by
insisting that we have gauge symmetry forces the anomaly into the axial current.

Calculating the amplitudes Tµνρ and Tµν directly using the Feynman rules one sees that this
does not agree with equation (2.10) and (2.11). The amplitudes are

Z
d4 p i i i
Tµνλ = −i tr γλ γ5 γν γµ
(2π) 4 p/ − m p/ − /q − m p/ − k/1 − m
!
k1 → k2
+ . (2.12)
µ→ν

Z
d4 p i i i
Tµν = −i tr γ5 γν γµ
(2π) 4 p/ − m p/ − /q − m p/ − k/1 − m
!
k1 → k2
+ . (2.13)
µ→ν

With q = k1 + k2 .

The Pauli-Villars regulated amplitude is defined as the difference between the unregulated
amplitude and the amplitude with mass M .
reg
Tµνλ = Tµνλ (m) − Tµνλ (M ). (2.14)
phys
The physical amplitude Tµνλ is then understood as the limit of the regulated amplitude as
the regulator mass tends to infinity, that is
phys reg
Tµνλ = lim Tµνλ . (2.15)
M →∞
phys
The same applies to Tµν , but since Tµν (M ) ∼ 1/M we see that Tµν is convergent and does
not need regularisation

phys
Tµν = lim [Tµν (m) − Tµν (M )] = Tµν (m). (2.16)
M →∞

Then by using equation (2.10) we see that the axial conservation law becomes
phys
q λ Tµνλ = 2mTµν (m) − lim 2M Tµν (M ). (2.17)
M →∞

Then the AJB anomaly is completely described by the second term in equation (2.4).

Aµν = − lim 2M Tµν (M ). (2.18)


M →∞

In order to evaluate equation(2.18) we need to write equation (2.9) in a more useful form.
First we need to introduce the Feynman parameter integral

9
Z 1 Z 1−x1
1 1
=2 dx1 dx2 . (2.19)
abc 0 0 [ax1 + b(1 − x1 − x2 ) + cx1 ]3
Then we can write

Z " #
d4 p i(p/ + m) i(p/ − /q + m i(p/ − k/1 + m)
Tµν = −i 4
tr 2 2
γ5 2 2
γµ γν
(2π) p −m (p − q) − m (p − k1 )2 − m2
Z
d4 p Z 1 Z 1−x1
= −2i dx 1 dx2
(2π)4 0 0

tr[i(p/ + m)γ5 i(p/ − /q + m)γµ i(p/ − k/1 + m)γν ]


× 2
[(p − m2 )x2 + ((p − q)2 − m2 )(1 − x1 − x2 ) + ((p − k1 )2 − m2 )x3 ]3
!
k1 → k2
+ . (2.20)
µ→ν

Now we need to explicitly evaluate the trace. Expanding the trace term it is easy to see that
this term reduces to

tr[i(p/ + m)γ5 i(p/ − /q + m)γµ i(p/ − k/1 + m)γν ] = −imtr[γ5 /qγν k/1 γµ ]. (2.21)
Then we have

Z
d4 p Z 1 Z 1−x1
Tµν = 2i dx1
(2π)4 0 0

im tr[γ5 /qγν k/1 γµ ]


×
[(p2 − m2 )x2 + ((p − q)2 − m2 )(1 − x1 − x2 ) + ((p − k1 )2 − m2 )x3 ]3
!
k1 → k2
+ . (2.22)
µ→ν

Now we explicitly evaluate the trace.

tr[γ5 /qγµ k/1 γν ] = tr[γ5 (k/1 + k/2 )γµ k/1 γν ]


= tr[γ5 γα k1α γµ γβ k1β γν ]
+tr[γ5 γα k2α γµ γβ k2β γν ]
= tr[γ5 γα γµ γβ γν ]k1α k1β
+tr[γ5 γα γµ γβ γν ]k2α k1β
= 4iαµβν k2α k1β . (2.23)

Then we have
Z
d4 p Z 1 Z 1−x1
im4iαµβν k2α k1β
Tνµ (m) = 2i dx 1 dx 2 2 . (2.24)
(2π)4 0 0 [p3 − 2pk − m 2 ]3
In the above we have included a factor of 2 which arises from k1 → k2 and µ → ν. We have
also defined

10
k = q(1 − x1 − x2 ) + k1 x1 , (2.25)

m 2 = m2 − q 2 (1 − x1 − x2 ). (2.26)
We now use the ’t Hooft-Veltman integral formula [66]
Z
dn p 1−2α n Γ(α − n2 ) 1
2 α
=i π 2 n . (2.27)
2
(p − 2pk − m ) Γ(α) (k + m 2 )α− 2
2

Here we have α = 3 and n = 4. Thus,

Γ(3 − 4/2) 1 1 Γ(1)


i1−6 π 2 = i −5 2
π
Γ(3) (k 2 + m 2 )3−2 (k 2 + m 2 ) Γ(3)
π2 1
= −i . (2.28)
2 (k + m 2 )
2

For large M the integral clearly behaves as 1/M 2 . Therefore,

2i π 2 M 2
lim 2M Tµν (M ) = lim 2 2i 4iµναβ k1α k2β
M →∞ M →∞ (2π)4 2i M 2
23 π 2
= 4
4µναβ k1α k2β
(2π)
1
= µναβ k1α k2β . (2.29)
2π 2
Which is the momentum space expression for the anomaly. Hence even in the massless case
(m = 0), a local chiral rotation is not a symmetry of the quantum action.

phys 1
q λ Tµνλ µναβ k1α k2β .
= 2mTµν − (2.30)
2π 2
Then in x space the anomalous conservation law can be written as
1
∂ µ jµ5 = +2miP + µναβ F αβ F µν . (2.31)
16π 2
In the massless case we have
1
A(x) = ∂ µ jµ5 =
2
µναβ F αβ F µν . (2.32)
16π
We call the term A(x) the singlet (or abelian)anomaly.

Expressions (2.30) and (2.31) are equivalent, this can be shown by considering the expecta-
tion value with two external photons with momentum k1 , k2 and polarisation vectors 1 , 2 .
See [2] for details.

So far, we have calculated the singlet anomaly in QED via a one-loop perturbation calcu-
lation. It is important to realise that higher order contributions only renormalise the fields

11
and their couplings, they do not effect the anomaly. Thus all the structure of the singlet
anomaly is contained within the triangle graph. This is known as the Adler-Bardeen theorem
[2]. This theorem has been proved for QED and QCD. However, it is not at all clear if this
theorem holds true in general gauge theories.

The singlet anomaly (2.32) is independent of the regularisation employed. The ’t Hooft-
Veltman dimensional regularisation will yield the same result.

The anomaly also appears in x-space (rather that momentum space) when one considers
current operators. As the product of quantum fields at at the same point in space is infinite,
the axial current

jµ5 (x) = ψ(x)γµ γ5 ψ(x), (2.33)


is formally infinite. This current requires regularisation which induces the singlet anomaly.
See [42] for a further discussion.

2.2 The Singlet Anomaly via Fujikawa’s Method


Here we work on a 4-dimensional manifold (M ) with Euclidean signature metric. As we
are dealing with spin bundles we need the manifold to have trivial first ω1 (M ) and second
ω2 (M ) Stiefel-Whitney classes. Such a manifold will be called a spin manifold. This is not
a severe restriction and is satisfied for example by spheres.

The first Stiefel-Whitney class ω1 (M ) is an obstruction to the orientability of a manifold.


Thus, to be orientable ω1 (M ) must be trivial.

The second Stiefel-Whitney class ω2 (M ) can be thought of as an obstruction to the existence


of a spin bundle. Spin bundles over orientiable manifolds only exist if ω2 (M ) is trivial.
b
In this section we have also restricted our attention to manifolds that have trivial A-genus.

Let ψ be a massless Dirac Field interacting with a non-abelian gauge field Aµ = Aαµ Tα .
Where {Tα } are the generators of some non-abelian group G. The Lagrangian is given by

L = iψγ µ (∂µ − Aµ )ψ. (2.34)


This Lagrangian is invariant under a local gauge transformation

ψ(x) → g −1 ψ(x), Aµ (x) → g −1 [A(x) + ∂µ ]g. (2.35)

There is also a global symmetry

ψ(x) → eiγ5 α ψ(x), ψ(x) → ψ(x)eiγ5 α . (2.36)

This is the global chiral symmetry which generates the Noether chiral current j5µ .

12
j5µ (x) = ψ(x)γ µ γ5 ψ(x). (2.37)
Classically we have ∂µ j5µ = 0, however when one considers the quantum effective action the
above conservation law does not hold. This is seen as a non-trivial transformation of the
path integral measure under a chiral rotation. The effective action is defined as
Z R
−W [A] − /
dxψiDψ
e = DψDψe . (2.38)
/ = iγ µ (∂µ + Aµ ).
Where we have used the Feynman slash notation to define iD

The chiral current j5µ can be calculated by considering an infinitesimal chiral transformation.
Now we perform a local chiral rotation of the spinor fields

ψ(x) → eiγ5 α(x) ψ(x) ≈ ψ + iα(x)γ5 ψ (2.39)


ψ(x) → ψ(x)eiγ5 α(x) ≈ ψ + iψα(x)γ5 .
Thinking of the local chiral rotations as a coordinate transformation, it is clear that these
transformations will be symmetries of the path integral. This is in analogy to standard
integrals. If we insist that
Z Z
n
d x f (x) = dn x0 f (x0 ), (2.40)
then we need

∂x0
n 0 n
d x = d x. (2.41)
∂x
Thus in changing variables, the Jacobian factor is important in the evaluation of integrals.

Under a local chiral rotation, the classical action transforms under this rotation as
Z Z
/
dxψiDψ → dx(ψ + iψαγ 5 )iD(ψ
/ + iαγ 5 ψ)
Z Z
= / +i
dxψiDψ dx[αψγ 5 iDψ
/ + ψiD(αγ
/ 5
ψ)]
Z Z
= / −
dxψiDψ dx[αψγ 5 γ µ (∂µ + Aµ )ψ
+ψγ µ (∂µ + Aµ )(αγ 5 ψ)]
Z Z
= / +
dxψiDψ dxα(x)∂µ j5µ . (2.42)

However, as stated earlier not only does the classical action change under the chiral ro-
tations also the quantum measure changes. To see this define the chiral (infinitesimally)
rotated fields as

X
ψ 0 = ψ + iαγ5 ψ = a0i ψ (2.43)
0 X 0
ψ = ψ + iψαγ5 = bi ψi† . (2.44)

13
Where a,b are anti-commuting Grassmann variables and ψi are eigenspinors of the Dirac
operator

/ i = λi ψi .
iDψ (2.45)
The eigenspinors are normalised such that
Z
hψi |ψj i = dxψi† (x)ψj (x) = δij , (2.46)
which can be achieved provided M is compact, which we assume. This allows us to write
the fermionic quantum measure in terms of the Grassmann parameters a and b. We define
the change in this measure under a chiral rotation as
Z Y Z Y
0
dai dbi → da0i dbi . (2.47)
i i

We now want an expression for a0i and b0i .

a0i = hψi |ψ 0 i
= hψi |(1 + iαγ5 )|ψi
X
= hψi |(1 + iαγ5 )aj |ψj i
j
X X
= hψi |ψj i + iαaj hψi |γ5 |ψj i
j j
X X
= δij + iαaj hψi |γ5 |ψj i
j j
X
= Cij aj (2.48)
j

Then the Jacobian is given by

Y Y
da0j = (det[Cij ])−1 dai
Y
= exp[−tr ln Cij ] dai
Y
= exp[−tr ln(1 + iαhψi |γ5 |ψj i)] dai
Y
≈ exp[−tr iαhψi |γ5 |ψj i] dai
" #
X Y
= exp −iα hψi |γ5 |ψi i dai . (2.49)
i

It is the inverse of the determinant that is used in the above as we are dealing with Grassmann
Q 0 Q
numbers. The contribution for bi is identical to a0i . Hence the measure changes overall
as

" #
Y Y X
dai dbi → da0i dbi 0 exp −2α hψn |γ5 |ψn i
i i n
" Z #
Y X
= da0i dbi 0 exp −2i dxα(x) ψn† (x)γ5 ψn (x) . (2.50)
i n

14
Then we have two expressions for the effective action

Z Y Z
−W [A] /
e = dai dbi exp[− ψiDψdx]
i
Z Y Z
= da0i dbi 0 exp[− / − α(x)∂µ j5µ
dx(ψiDψ
i
X
−2iα(x) ψn† γ5 ψn )]. (2.51)
n

Since α(x) is arbitrary we require the chiral conservation law


X
∂µ j5µ = −2i ψn† γ5 ψn . (2.52)
n

The term on the right up to factor of −2i is known as the Abelian anomaly
X
A(x) = ψn† γ5 ψn . (2.53)
n

The above expression for the anomaly is not well defined and requires regularisation. Here
we use the heat kernel method.
X
A(x) = lim ψn† (x)γ5 exp[(iD/M
/ )2 ]ψn (x). (2.54)
M →∞ n

We change basis to momentum space as


Z
d4 k ik.x
ψn (x) = e ψn (k). (2.55)
(2π)2
Using this the trace of any matrix becomes
Z
d4 k ik.x
tr M (x) = 4
e M (x)e−ik.x . (2.56)
(2π)
Then the anomaly can be expressed as
Z
d4 k ik.x −(iD/M 2
A(x) = lim tr 4
e γ5 e / ) e−ik.x . (2.57)
M →∞ (2π)
Now we expand the square of the Dirac operator as
1
/ 2 = −Dµ Dµ − [γ µ , γ ν ]Fµν .
(iD) (2.58)
4
Where we have made use of the definition Fµν = [Dµ , Dν ]. Then we have

Z " #
d4 k ik.x ∇µ ∇ν + 41 [γ µ , γ ν ]Fµν −ik.x
A(x) = lim tr e γ 5 exp e
M →∞ (2π)4 M2
Z " #
d4 k ik.x −k 2 + 14 [γ µ , γ ν ]Fµν −ik.x
= lim tr e γ5 exp e
M →∞ (2π)4 M2
   Z
1 µ ν d4 k − k22
= lim tr γ5 exp [γ , γ ]Fµν e M . (2.59)
M →∞ 4M 2 (2π)4

15
Expanding the first exponential

 
1 1 1 1
exp [γ µ
, γ ν
]Fµν ≈ 1 + [γ µ
, γ ν
]Fµν + ([γ µ , γ ν ]Fµν )2 . (2.60)
4M 2 4M 2 2 (4M 2 )2

Then using

trγ5 = trγ5 γ µ γ ν = 0, (2.61)


we arrive at

" #Z
1 1 µ ν 2 d4 k − k22
A(x) = lim tr γ5 ([γ , γ ]F µν ) e M
M →∞ 2 (4M 2 )2 (2π)4
" #
1 1 µ ν 2 M 4π2
= lim tr γ5 ([γ , γ ]F µν )
M →∞ 2 (4M 2 )2 (2π)4
1 1 1 h i
= tr γ5 ([γ µ , γ ν ]Fµν )2
2 162 π 2
4 h i
κ λ µ ν
= tr γ 5 γ γ γ γ F κλ F µν . (2.62)
2 162 π 2
Now use

tr [γ5 γ µ γ ν γ ρ γ σ ] = −4µνρσ . (2.63)


Then we obtain
1
trκλµν Fκλ Fµν .
A(x) = − (2.64)
32π 2
Then the chiral conservation law becomes

1
∂µ j5µ = trκλµν Fκλ Fµν
16π 2  
1 κλµν 2
= tr  ∂κ (Aλ ∂µ Aν + Aλ Aµ Aν . (2.65)
4π 2 3
Which can be regarded as the local version of the Atiyah-Singer index theorem. Notice the
Chern-Simons form that appears in the local form of the anomaly. By local, we mean that
we are not integrating over the manifold as one does in the standard Atiyah-Singer index
theorem.

If one integrates over the manifold we make direct contact with the Atiyah-Singer index
theorem.

Z Z X h i
dxA(x) = dx ψn† γ5 ψn exp −λ2n /M 2 |M →∞ ,
n
X h i
= /
hψn |γ5 exp −(iD/M )2 |ψn i|M →∞ . (2.66)
n

16
Now, suppose we have iD|ψ / n i = λn |ψn i with λn non-vanishing. Then we have another spinor
/ n i = −λn |φn i. To see this we note that
|φn i ≡ γ5 |ψn i such that iD|φ

/ n i = iDγ
iD|φ / 5 |ψn i
= −λn γ5 |ψn i
= −λn |φn i. (2.67)

Then using the above it is clear that we must have

hψn |φn i = hψn |γ5 |ψn i = 0, (2.68)


due to the orthogonality of the eigenspinors. Then contributions to the anomaly by the
spinors with non-vanishing eigenvalues is
  2  h i
/
hψn |γ5 exp − iD/M |ψn i = hψn |γ5 |ψn i exp − (λn /M )2 = 0. (2.69)

Then the only possible contribution to the anomaly must be from the zero modes. This
observation directly links the anomaly to an index theorem.

Let |0, ii be zero modes of the Dirac operator, remembering that the Dirac operator is elliptic
on a compact manifold and hence has a finite dimensional kernel and co-kernel. These can
be classified by their eigenvalue of γ 5 as they are not an irreducible representation of the
spin algebra. Then,

γ 5 |0, ii± = ±|0, ii± . (2.70)


Then we have,

Z X
dxA(x) = hψi |γ 5 exp[−(iD/M
/ )2 ]|ψi i|M →∞
X X
= h0, i|0, ii+ − h0, i|0, ii−
i+ i−
/ +.
= υ+ − υ− = indiD (2.71)

Where ν± is the number of positive and negative chirality zero modes. This is precisely the
Atiyah- Singer index theorem.

Writing F = 21 Fµν dxµ ∧ dxν , we see that


Z Z
ν+ − ν− = dx∂µ j5µ = ch2 (F ). (2.72)
M M
Which is the AS index theorem for a twisted Dirac complex on a compact 4-manifold with
trivial A-roof genus.

In the language of differential forms the anomalous conservation law can be written as
1
d∗J =− tr(F ∧ F ) (2.73)
4π 2

17
Where the current one-form is defined as J = jµ5 dxµ , d is the exterior derivative and ∗ is the
Hodge dual.

This result can be generalised to any compact spin-manifold with trivial A-roof genus of
arbitrary (but even) dimension Dim M = 2l as
Z Z
ν+ − ν− = dx∂µ j5µ = chl (F ), (2.74)
M M
which is the AS-index theorem for twisted spin complexes on compact spin-manifolds with
trivial A-roof genus without boundary.

It should be noted that the anomaly is independent of the regularisation used. Instead of
using the exponential function, we could have used any function f which is smooth and
decreases rapidly enough at infinity
!
λn
f , (2.75)
M2
with

f (∞) = f 0 (∞) = f 00 (∞) = . . . = 0


f (0) = 1. (2.76)

For example, we could use


!
λn 1
f = (2.77)
M2 1 + λ2n /M 2
as the regularisation function, which corresponds to Pauli-Villars regularisation.

2.3 The Non-Abelian Anomaly via Fujikawa’s Method


Here we examine the case where we have a chiral theory and so the left and right-handed
Weyl fermions ψ do not couple equally to the gauge field Aµ . The result is a theory in which
the effective action is not gauge (or more precisely BRST) invariant, whereas the classical
action is. In what follows we ignore all possible gauge fixing and FP ghost terms. This is a
justified restriction as the anomaly only arises in the fermionic sector of the theory. Again,
we assume the manifold is a compact spin-manifold with trivial geometry and Euclidean
metric. We use the techniques developed for the abelian anomaly, but pointing out the main
differences and complications as they arise.

We construct an effective action in which the gauge field A couples only to the left-handed
Weyl fermion. ψ also transforms under a complex representation of the gauge group G. The
effective action is
Z  Z 
−W [A] / +ψ .
e = DψDψ exp − dxψiD (2.78)

18
Where we have defined
1
/ + = iDP
iD / + P± = (1 ± γ 5 ). (2.79)
2
The gauge current is given by

jαµ = iψγ µ Tα P+ ψ. (2.80)


Now we consider an infinitesimal gauge transformation g = 1 − v, with v = v α Tα . Then we
have

Aµ → (1 + v)(Aµ + d)(1 − v) = Aµ − Dµ v, (2.81)


where the covariant derivative is defined as Dµ v ≡ ∂µ + [Aµ , v]. The effective action then
transforms as

W [A] → W [A − Dv]
Z !
δ
= W [A] − dx tr Dv W [A]
δA
Z   δ
= W [A] − dx tr ∂µ v α + fαβγ Aβµ v γ W [A]
δAαµ
Z !
δ
α
= W [A] + dx tr v D W [A]α . (2.82)
δA

We can then write


Z
W [A − Dv] − W [A] = dx tr(vα Dµ hj µ iα ). (2.83)
Where we have used
δ 1
α
W [A] = hiψγ µ Tα (1 + γ 5 )ψiA = hjαµ i. (2.84)
δAµ 2
One would now like to preceded by writing the effective action as the determinant of the
Q
/ = λi , with λ being eigenvalues of iD.
Dirac operator det(iD) / However, this construction
has no meaning as is not a well posed eigenvalue problem. The difficulty arises due to the
/ + maps sections of S+ ⊗ E to S− ⊗ E. Where S± are the spin bundles of positive
fact that iD
and negative chirality and E is the associated vector bundle to the principle G-bundle.

We formally overcome this complication by introducing a Dirac spinor ψ and defining the
effective action as
Z Z 
−W [A]
e = DψDψ exp dxψiD̂ψ . (2.85)

Where the operator iD̂ is defined as


!
µ 0 i∂/
iD̂ ≡ iγ (∂µ + iAµ P+ ) = . (2.86)
/+ 0
iD

19
In this construction it is clear that the gauge field only couples to the Dirac spinor of positive
chirality. The eigenvalue problem iD̂ψi = λi ψi is well posed. However, one must note that
iD̂ is not Hermitian and as such λi will be a complex number in general. As such, one needs
to introduce right and left eigenfunctions.

iD̂ψi = λi (2.87)

←−̂
χ†i i D = λi χ†i , (iD̂)† χi = λi χi . (2.88)
As these define a complete set of eigenvectors we can use them to define an orthonormal
basis,
Z
χ†i ψj = δij . (2.89)
However, the eigenvalues λi are not gauge invariant. This is clearly seen as the operator
iD̂ is in fact not gauge invariant. This is not really a problem as we are interested in the
product of all the eigenvalues and this is gauge invariant. We see this using

det(iD̂) det((iD̂)† ) = det(iD̂(iD̂)† )


!
(i∂/− )(i∂/+ ) 0
= det
0 (iD/ − )(iD/ +)
= det(i∂/− i∂/+ ) det(iD/ − iD
/ + ). (2.90)

With i∂/+ = (i∂/− )† and iD


/ − = (iD/ + )† . This is up to a numerical factor just the Dirac
determinant which is gauge invariant. In the above we have implicitly used a gauge invariant
regularisation scheme such as the Pauli-Villars. Explicitly we have
!
2 / − )(iD
(iD / +) 0
/
(det(iD)) = = (det(iD / − ))2 .
/ + iD (2.91)
0 / + )(iD
(iD / −)

As the Dirac determinant is gauge invariant so is | det(iD̂)|. Now if we examine the real part
of the effective action Re W [A] we see that it is gauge invariant

exp(−W [A]) exp(−W [A]) = det(iD̂) det((iD̂)† ) ∝ det(iD


/ + iD
/ − ). (2.92)
This is a crucial observation, only the imaginary part of the effective action, that is the phase
of det(iD̂), may be anomalous. This means that in order to have gauge anomalies the gauge
group must have a complex representation.

We can now evaluate the anomaly by examining the Jacobian of the fermionic measure, as
we did for the abelian anomaly.

We write Weyl fermions in terms of the Dirac spinors which are eigenspinors of D̂.

X
ψ= ai ψi (2.93)
i
X
ψ= bi χ i (2.94)
i

20
Then consider infinitesimal gauge transformations,

A → A − Dv ψ → ψ + vψ+ ψ → ψ − ψ − v. (2.95)
We can thus define the gauge transformed fields as

X
ψ → ψ + v(x)P+ ψ = a0i ψi
i
X
ψ → ψ − v(x)P− ψ = b 0i χi . (2.96)
i

Then we can form the Jacobian in the same way as for the abelian anomaly. We first have
to calculate the transformation properties of the ai and bi .
X
a0i = hψi |ψ 0 i = hψi |1 + v(x)P+ |ψi = hψi |1 + v(x)P+ |ψj iaj . (2.97)
j

Then using the same arguments as before the Jacobian for dai is given by

Y Y
da0j = exp[−tr ln(1 + v(x)hψi |P+ |ψj i)] dai
Y
≈ exp[−tr v(x)hψi |P+ |ψi i)] daj . (2.98)

In a similar way we calculated the Jacobian for dbi .


X
b 0i = hψ 0 |χi i = hψ|1 − P− v(x)|χi i = bj hψj |1 − P− v(x)|χi i. (2.99)
j

Then we have
Y Y
db 0j ≈ exp[tr v(x)hψi |P− |χi i] dbj . (2.100)
Then the Jacobian factor in the effective action is given by
Z X
dxtr v(x) hχn |γ 5 |ψn i (2.101)
n

We then pick a representation of the spinors such that hx|ni = ψn (x) and (n|xi = χ†n (x).
Notice that (n| is not the Hermitian conjugate of |ni. This integral is not well defined and
requires regulisation. Again, we use the heat kernel regulisation,

Z X 2
dx lim tr v(x) (x|yiγ 5 hx|e−(iD̂)/M |ni
M →∞ n
Z
2
= dx lim tr v(x)γ 5 e−(iD̂)/M δ(x − y). (2.102)
M →∞

Where we have used the completeness relation


X
|ni(n| = I. (2.103)
n

Then as W [A] is gauge invariant, its gauge variation (2.83) must be cancelled by the Jacobian
factor. Thus, we are able to write

21
Z Z !
α δ 2
dx v (x)Dµ α
W [A] = dx lim tr[ v(x)γ 5 e−(iD̂x )/M δ(x − y)]. (2.104)
δAµ M →∞

We now “split” the trace into two pieces depending on their chirality.

2 /M 2 2
tr[v(x)γ 5 e−(iD̂x ) ] = tr[v(x)(P+ − P− )e−(i∂/− iD/ + )−(iD/ − i∂/+ )/M (2.105)
/ D)/M
(i∂i / 2 /
/ ∂)/M
(iDi 2
= tr[v(x)P+ e ] − tr[v(x)P− e ].

This can then be evaluated in the plane wave basis, see [40] for details.

  
1 Z α κλµν 1
W [A − Dv] − W [A] = dx tr v Tα  ∂κ Aλ ∂µ Aν + Aλ Aµ Aν
24π 2 2
  
1 Z 1 3
= tr v d AdA + A . (2.106)
24π 2 2
Then the anomalous conservation law can be written as
  
1 1
Gα [A] = Dµ hj µ iα = tr Tα  κλµν
∂ κ Aλ ∂ µ Aν + Aλ Aµ Aν . (2.107)
24π 2 2
Notice the difference in the factors in front of the A3 between the abelian and non-abelian
anomaly. This difference has a deep topological origin which will be explained in the next
section and proceeding chapter.

We have calculated the non-abelian anomaly for left chiral spin- 1/2 fermions coupled to
an external Yang-Mills field. The whole calculation could be carried out for right handed
fermions. The final result is that the non-abelian anomaly for the right handed fermions is
identical to the left handed fermions apart from an overall minus sign.

  
1 1
Gright
α [A] = −Glef t
α [A] =− 2
tr Tα κλµν ∂κ Aλ ∂µ Aν + Aλ Aµ Aν . (2.108)
24π 2
Thus, in a non-chiral theory the two sectors cancel resulting in an anomaly free theory.

2.4 The Wess-Zumino Consistency Conditions and the


Descent Equations
As shown earlier the effective action W [A] of a Weyl fermion in a complex representation of
a gauge group G transforms under an infinitesimal gauge transformation as
Z Z
δ
δν W [A] = − (Dµ ν)α α
W [A] = ν α Dµ hj ν iα . (2.109)
δAµ
Where δν A = −Dν. However, the correct statement in quantum field theory is that the
gauge fixed effective action should be BRST invariant (see Appendix A). That is the BRST
operator should annihilate the effective action. The presence of the gauge anomaly transpires

22
to be non-invariance under the BRST transformations.

Define Gα [A] as the local form of the non-abelian anomaly

Gα = Dµ hj µ iα , (2.110)
Then we define the global form of the non-abelian anomaly in terms of the BRST transfor-
mations and ghost fields
Z
sW [A] = G[η, A] = dxη α Gα [A] 6= 0, (2.111)
if the effective action is no longer BRST invariant. As s is a nilpotent we have

sG[η, A] = s2 W [A] = 0. (2.112)


Equation (2.112) is known as the Wess-Zumino consistency condition and is strong enough
to determine the gauge anomaly up to normalisation.

Now in order to solve the consistency equation one defines a new operator

dˆ = d + s (2.113)
which itself is also a nilpotent and anticommutes with the exterior derivative.

dˆ2 = d2 + ds + sd + s2 = 0. (2.114)
We also define

A ≡ g −1 (A + d)g. (2.115)
From this we define the field strength

F ≡ dA + A2 = g −1 F g. (2.116)
The Faddeev-Popov ghost is identified with

η ≡ g −1 sg. (2.117)
We also define

ˆ = A + η,
A ≡ g −1 (A + d)g (2.118)
ˆ + A2 = F.
F ≡ dA (2.119)

Stora and Zumino constructed solutions to the Wess-Zumino consistency conditions via the
so called decent equations. The Abelian anomaly in (2l + 1) dimensions is given by the
following Chern-Simons form as we have seen
!l+1
1 iF
chl+1 (F) = tr . (2.120)
(l + 1)! 2π
Let Q2l+1 (A, F) be the Chern-Simons form

23
chl+1 (F) = dQ2l+1 (A, F). (2.121)
ˆ A, F) is identical of that of (d, A, F) we have
As the algebraic structure of (d,

ˆ 2l+1 (A, F) = dQ
chl+1 (F) = dQ ˆ 2l+1 (A + η, F). (2.122)
Expanding the above in powers of η

Q2l+1 (A, F) = Q02l+1 (A, F) + Q12l (η, A, F) + Q22l−1 (η, A, F)


+ ... + Q2l+1
0 (η, A, F). (2.123)
Where Qsr is s-th order in η and r + s = 2l + 1.

We see that

Q̂2l+1 (A, F) = dQ2l+1 (A, F), (2.124)


as we have chl+1 (F) = chl+1 (F). Then we have

(d + s)[Q02l+1 (A, F) + Q12l (η, A, F)


+... + Q2l+1
0 (η, A, F)] = dQ02l+1 (A, F). (2.125)
Then collecting terms order by order in η we arrive at the decent equations

dQ02n+1 = chl+1 (F)


sQ02l+1 (A, F) + dQ12l (η, A, F) = 0
sQ12l (ηA, F) + dQ22l−1 (η, A, F) = 0
..
.
2l 2l+1
sQ1 (η, A, F) + dQ0 (η, A, F) = 0
sQ2l+1
0 (η, A, F) = 0. (2.126)
Notice that if we take the 2l-form Q12l and define
Z
G[η, A, F] = Q12l (η, A, F), (2.127)
M
then G[η, A, F] automatically satisfies the Wess-Zumino consistency condition.
Z Z
sG[η, A, F] = s Q12l (η, A, F) =− dQ22l−1 (η, A, F)
M M
Z
= − Q22l−1 (η, A, F) = 0. (2.128)
∂M

Where we have explicitly assumed that the manifold has no boundary. Hence, finding the
anomaly becomes a case of finding Q12l . This is further discussed in the next chapter. How-
ever, it should be noted that that this approach to calculating the gauge anomaly has some
unanswered questions: i)the consistency condition does not fix the normalisation of the
anomaly. Hence, the anomaly is not unique. ii) it is not clear why we start form the abelian
anomaly in (m + 2) dimensions.

24
2.5 Epitome
In gauge theory there are two types of anomaly (in 4-dimensions)

1. Singlet anomaly
 
1 3
A(x) = 2 Trd AdA + A3 (2.129)
4π 2
2. Non-abelian anomlay
 
1 1
Gα [A] = 2
Trd AdA + A3 . (2.130)
24π 2
These were calculated via Fujikawa’s method in which these anomalies show up as Jacobians
for the path integral measure under a chiral rotation and a gauge transformation for the
singlet and non-abelian anomaly respectively.

In arbitrary, but even dimensions dimM = 2l the singlet anomaly can be calculated via the
Atiyah-Singer index theorem. Considering only manifolds with trivial Â-genus we have
Z Z
/+ =
dxA(x) = ind iD chl (F ). (2.131)
M
The non-abelian anomaly is completely characterised by the Wess-Zumino consistency con-
ditions

sG[η, A] = 0, (2.132)
which directly lead to the descent equations (2.126). The local form non-abelian anomaly
is given by the Q12l term in the descent equations, at least up to normalisation. Thus the
non-abelian anomaly becomes a case of solving the descent equations.

25
CHAPTER 3
BRST Cohomology and the Descent Equations

As shown in Chapter 3 the Wess-Zumino consistency conditions lead directly to a BRST


cohomology problem. The consistency equations take a very simple form

sG[η, A] = 0, G[η, A] 6= sF [A], s2 = 0, (3.1)


where the anomaly G is a functional
Z
G[η, A] = η α Gα [A]. (3.2)
The first equation defines the BRST cohomology in the space of local functions and we
will denote this H(s). The consistency condition may also be written in terms of the local
functionals,

sQ12l + dQ22l−1 = 0, Q12l 6= sQ̂02l + dQ̂12l−1 . (3.3)


Thus the relevant cohomology is the BRST cohomology in the space of local functionals
which we denote as H(s|d). The consistency condition may also be also be studied for ghost
number other than one; which corresponds to the gauge anomalies. For example for zero
ghost number the cohomology controls the possible counterterms in the renormalistion of
Yang-Mills.

In this section we are using differential forms and as such the wedge product is understood
when we multiply forms.

3.1 Local Cohomology and the Cohomology of d


Local forms are defined as the differential forms Q whose coefficients Qµ1 ...µp (x) are local
functionals. That is the coefficients are polynomials in the fields and their derivatives taken
at the same point x. The integral of a local form is also called a local functional.

Let V(A, dA, η, ζ = dη) be the space of polynomials of differential forms. We define the
one-form gauge connection as A = Ta Aaµ dxµ and the zero-form ghost as η = Ta η a , which

26
can be identified as the Maurer-Cartan form of the gauge group, see [61, 62]. We wish to
compute the various local cohomologies on V(A, dA, η, ζ), first we consider the cohomology
of the exterior derivative.

Local cohomolgies are in general difficult to compute. Introducing a filtration operator N


and a simpler exterior derivative operator d0 will prove to be useful in calculating the local
cohomology. The filtration operator N : V → V is defined as
∂ ∂ ∂ ∂
N =ζ +η +A +F . (3.4)
∂ζ ∂η ∂A ∂F
The integral of which counts the number of fields in a given monomial. Direct computation
shows that

NA = A , NF = F
N η = η , N ζ = ζ. (3.5)

Acting on V the exterior derivative acts as an ordinary differential operator

∂ ∂ ∂
d = dA + dη + dF
∂A ∂η ∂F
∂ ∂ ∂
= (F − A2 ) +ζ + [F, A] . (3.6)
∂A ∂η ∂F
Where we have used the Bianchi identity (A.16) and the definition of the Yang-Mills curva-
ture (A.14).

Now we suppose that the forms Qqp can be expanded in terms of eigenvalues of the filtration
operator

Qqp = Qqp(0) + Qqp(1) , N Qqp(n) = nQqp(n) . (3.7)

where n = 0, 1.

From (3.6) it is clear that the exterior derivative decomposes as

d = d0 + d1 , [N, dn ] = ndn , (3.8)

Where
∂ ∂
d0 = F +ζ . (3.9)
∂A ∂η
We have that

d0 F = 0,
d0 ζ = 0. (3.10)

27
Also it is easy to see that

d0 A = F,
d0 η = ζ. (3.11)

We now use the following general results from homological algebra, see for example [41, 59];

1. d0 is a coboundary operator.

2. The cohomology of d is isomorphic to a subspace of the cohomology of d0 .

Then from (3.10) and (3.11) it is clear that the cohomology of d0 is trivial. As such, the
cohomology of d on V(A, dA, η, ζ) must be also trivial. This result will be important later
in deriving the descent equations.

3.2 The BRST Operator and its Cohomology


The classical BRST transformations (A.52) on a gauge field Aµ , a ghost field η = Ta η a and
the field strength Fµν are given by

sAµ = Dµ η,
1
sη = − [η, η] = −η 2
2
sFµν = [Fµν , η]. (3.12)

The exterior algebra and the BRST algebra (3.12) can be combined into a graded algebra.
This algebra is graded by the total degree Q which is defined as the sum of the form degree
p and the ghost degree q. On this graded algebra both d and s act as antiderivatives and
anticommute

s2 = d2 = {s, d} = 0. (3.13)
These two coboundary operators act on the differential forms Ωqp with their degrees being
(q, p). The action of s increases q by one unit and the action d increases p by one. The
differential forms are given by
1 q
Ωqp = Ω dxµ1 ...dxµp . (3.14)
p! µ1 ...µp
These forms are partitioned into even and odd forms according to the parity of the total
degree (q + p). The coefficients Ωqµ1 ...µp (x) are local polynomials in the fields and their deriva-
tives.

The definition of the exterior derivative is chosen to be


1 ν
dΩqp = dx ∂ν Ωqµ1 ...µp dxµ1 ...dxµp . (3.15)
p!

28
.. .. ..
. . .

d- n d- n
··· Ωnm Ωm+1 Ωm+2 ···

s s s
? ? ?
d - n+1 d - n+1
··· Ωn+1
m Ωm+1 Ωm+2 ···

.. .. ..
. . .

Figure 3.1: The BRST bicomplex whose local cohomology is important in the calculation of
possible anomalies in gauge theories.

Then the BRST transformations (3.12) are written in terms of the differential forms A and
F

sA = −(dη + [A, η])


1
sη = − [η, η]
2
sF = [F, η]. (3.16)

In the BRST transformations (3.16), [, ] is the graded commutator,

[Ω, Λ] = ΩΛ − (−1)degΩ.degΛ ΛΩ. (3.17)


In particular we have

[A, η] = Aη + ηA
[F, η] = F η − ηF
[η, η] = ηη + ηη = 2η 2 . (3.18)

The BRST algebra (3.16) is independent of any Lagrangian. All that is needed is a principle
gauge bundle over a manifold. The BRST algebra is a fundamental symmetry that any
gauge theory must obey. Thus, in this section and following sections the physical theory is
not fixed by a particular Lagrangian. Only BRST symmetry need be assumed.

We briefly comment that an anti-BRST operator s can also be constructed in a similar way
to the BRST operator s. Here s decreases the the ghost number by one. The anit-ghost η
had ghost number −1. The graded anti-BRST algebra is defined as

sA = −Dη
sF = [F, η]

29
1
sη = − [η, η]
2
s2 = sd + ds = 0. (3.19)

The anti-BRST transformations play no role in the following discussions on the anomalies
and as such will not feature in subsequent sections of this thesis.

As the BRST operator is a nilpotent it can be used to define a cohomology. The BRST
cocycles A are forms that are BRST-closed and hence in the kernel of s,

sΩ = 0. (3.20)
The BRST coboundaries are defined as forms that are BRST-exact (in the image of s) and
as such are also BRST-closed

Ω = sΩ̂. (3.21)
The BRST-cohomology is defined as the quotient space Kers/Ims,
Kers
H(s) = . (3.22)
Ims
An element of H(s) defines an equivalence class of BRST cocycles, where the cocycles are
equivalent if they differ by a BRST coboundary. Note the similarity here with de Rham
cohomology.

The cohomology of s on V(A, dA, η, ζ) [24, 64] is spanned by polynomials in F and η,


generated by
!
η 2m+1
Tr P2n+2 (F ), (3.23)
(2m + 1)!
where n, m = 1, 2... and P2n+1 (F ) is an invariant polynomial of degree 2n + 1. One such
example is

P2n+2 (F ) = Tr(F n+1 ), (3.24)


which will turn out to be the relevant monomial for evaluating the non-abelian anomaly for
a spin-1/2 fermion.

The invariant polynomial (3.24) is certainly not the most general expression possible. The
exact form of the invariant polynomial depends on the physical theory in question. Generi-
cally the invariant polynomial will be the sum of products of traces.

3.3 The Cohomology of s mod d


The cohomology of the exterior derivative on V(A, dA, η, dη) is vanishing, [15, 24].

From the Bianchi identity it is clear that P2n+2 (F ) is d-closed,

dP2n+2 (F ) = 0. (3.25)

30
As the cohomology is trivial we see that

P2n+2 = dQ02n+1 . (3.26)


Using equation (3.14) the descent equations (2.126) can easily be derived. Acting on (3.26)
with s gives

sdQ02n+1 = sP2n+1 = 0. (3.27)


Where we have used the fact that P2n+1 is gauge invariant and hence it must also be BRST
invariant. Using (3.14) we see that

d sQ02n+1 = 0. (3.28)
As the cohomology of d on V(A, dA, η, ζ) is trivial, every closed form is exact. This allows
us to write

sQ02n+1 + dQ12n = 0, (3.29)


which we recognise as the first line in the descent equations. Now apply s to (3.29) we get

sdQ12n = 0. (3.30)
Then by interchanging s and d, and using the triviality of the cohomology of d as before we
arrive at

sQ12n + dQ22n−1 = 0. (3.31)


Continuing this iteratively until one reaches a differential form with negative degree. The
descent equations must end here and the final line must be given by

sQ2l+1
0 = 0. (3.32)
The descent equations are thus given by

dQ02n+2 = P2n+1
sQ02n+1 + dQ12n = 0
sQ12n + dQ22n−1 = 0
..
.
2n 2n+1
sQ1 + dQ0 = 0
2n+1
sQ0 = 0. (3.33)

We define a s-ladder Q as the set of differential forms that satisfy the descent equations.

Thus the problem of solving the descent equations (3.33) is the problem of solving the
cohomology of s mod d. The elements of this cohomology are the solutions of the descent
equations which are not trivial, that is not of the form

31
Qqp = sQ̂q−1
p + dQ̂qp−1 , for p 6= 0,
Q02n+1 = sQ̂2n
0 . (3.34)

Where Q̂ is some ladder.

Two ladders, Q and Q0 are said to be equivalent if their difference is a trivial latter (3.34).
Elements of the cohomology of s mod d define the corresponding equivalence classes.

3.4 Some Solutions to the Descent Equations


We present solutions to the Descent equations as first solved by Stora and Zumino. The
solution relies on the triviality of the cohomology of d on V(A, dA, η, ζ) which means that
the invariant polynomials are always exact.

P2n+2 (F ) = dQ2n+1 (A, F ), (3.35)


where Q2n+1 (A, F ) is a Lie algebra valued 2n + 1 form known as the Chern-Simons form.
This form may be written as
 n+1 Z
1 i 1
Q2n+1 (A, F ) = dt str(A, Ftn ). (3.36)
(n)! 2π 0

Where we have

Ft = tF + (t2 − t)A2 . (3.37)


Some examples that will come in useful later are;

i Z1 i
Q1 (A, F ) = dt A = Tr(A). (3.38)
2π 0 2π
 
i 2Z 1
Q3 (A, F ) = dt str(A, tdA + t2 A2 )
2π 0
   
1 i 2 2 3
= Tr AdA + A . (3.39)
2 2π 3
 
1 i 3Z 1
Q5 (A, F ) = dt str(A, (tdA + t2 A2 )2 )
2 2π 0
   
1 i 3 3 3
= Tr A(dA)2 + A3 dA + A5 . (3.40)
6 2π 2 5
The method is then to “feed in” the solution for Q02n+1 (A, F ) and then “descend” the equa-
tions.

Lets consider the descent equations for n = 1 as an example. Explicitly we have

32
sQ03 + dQ12 = 0 (3.41)
sQ12 + dQ21 = 0 (3.42)
sQ21 + dQ30 = 0 (3.43)
sQ30 = 0. (3.44)

Also we have
   
2 1
= Tr AdA + A3 = Tr AF − A3 .
Q03 (3.45)
3 3
Here we have omitted the normalisation of the Chern-Simons form as it is arbitrary as the
descent equations are concerned. Then it is straight forward to verify that

sQ03 = Tr (sAF − AsF − sAA2 )


= −Tr (d(ηF ) − d(ηA2 ))
= −dTr η(F − A2 ). (3.46)

Then consulting the descent equations we see

sQ03 = −dQ12 , (3.47)


gives

Q12 = Tr η(F − A2 ) = Tr ηdA. (3.48)


Notice that up to a factor this is the non-abelian anomaly in 2 dimensions.
Now we apply s to Q12 to give the next term in the chain.

s Q12 = d Tr η 2 A. (3.49)
Then

Q21 = −Tr η 2 A. (3.50)


Again looking at the chain we see that
1
Q30 = −Tr η 3 . (3.51)
3
So the full s-ladder is

   
2 1
Q03 = Tr AdA + A3 = Tr AF − A3 ,
3 3
Q12 = 2
Tr η(F − A ) = Tr ηdA,
Q21 = −Tr η 2 A,
1
Q30 = −Tr η 3 . (3.52)
3

33
Now we consider the n = 2 case which produces the anomaly in 4 dimensions. Again
starting from the Chern-Simons form in 5 dimensions for Q05 , direct computation of the
decent equations gives the ladder

 
1 1
Q05 = 2
Tr AF − A3 F + A5 ,
2 10
 
1 3
Q14 = Tr η d AdA + A ,
2
1  2 
Q23 = − Tr (η A + ηAη + Aη 2 )dA + η 2 A3 ,
2
1  3 
Q32 = Tr −η dA + AηAη 2 ,
2
1
Q41 = Tr η 4 A,
2
1
Q50 = Tr η 5 . (3.53)
10
Notice that dQ05 is the U (1) anomaly in 6 dimensions and that Q14 is the non-Abelian anomaly
in 4 dimensions, up to normalisation.

The non-abelian anomaly is given by


Z
in+1
G[η, A] = (±)(−2πi) Q1 (η, A). (3.54)
(2π)n+1 (n + 1)! M 2n
The normalisation can be calculated via k-theory or perturbation theory.

This method can be applied to any value of n. However, it quickly becomes lengthy and
complicated. A general expression for arbitrary n is developed in the next section.

3.5 General Solutions to the Decent Equations and the


Russian Formula
A general solution (up to elements in the local cohomology of s) can be obtained for the
descent equations using a generalised transgression formula known as the Russian formula.
Again we define a generalised exterior derivative

dˆ = d + s, (3.55)
which is nilpotent. We also introduce a family of gauge one-forms and there curvatures

At = tA + η, t ∈ [0, 1], (3.56)

where

A1 = A + η, A0 = η. (3.57)

34
We define the associated curvature as

ˆ t + A2
F̂t = dA t
= Ft + (1 − t)dη. (3.58)

Where we have defined Ft = tF + (t2 − t)A2 .

The BRST transformations are equivalent to the horizontality condition

ˆ 1 + A2 = dA + A2 = F (A),
F̂1 (A1 ) = dA (3.59)
1

which is known as the “Russian Formula”. We also have that

F̂0 = dη. (3.60)


Given a G-invariant symmetric polynomial P we consider a generalised transgression formula
with l = n + 1
l
P(Fˆ1 ) − P((dη)l ) = dˆQ̂2l−1 , (3.61)
where
Z 1
Q̂2l−1 = l dt P(A, (F̂t )l−1 ). (3.62)
0

From (3.62) the first part of the descent equations can be derived by expanding Q̂2l−1 in
powers of dη and using the transgression formula (3.59).

The symmetric polynomials can be expanded as

l−1
l−1
X (l − 1)!
P(A, (Ft + (1 − t)dη) )= (1 − t)k P((dη)k , A, (Ft )l−1−k ), (3.63)
k=0 k!(l − k − 1)!

and then the writing

Q̂2l−1 = Q02l−1 + Q12l−1 + ... + Ql−1


l , (3.64)
the first part of the descent equations are obtained

P(F l ) − dQ02l−1 = 0
sQ02l−1 + dQ12l−2 = 0
..
.
sQl−1
l + P((dη)l ) = 0. (3.65)

Then we can write


Z 1
(l − 1)!
Qk2l−1−k = dt (1 − t)k P((dη)k , A, (Ft )l−1−k ). (3.66)
k!(l − k − 1)! 0

35
With 0 ≤ k ≤ l − 1. This formula gives the terms in the descent equations up to k ≤ l − 1
in the power of dη.

For k = 0 we have the definition of the Chern-Simons form Q02l−1 . For k = 1 we have an
explicit solution for the anomaly
Z 1
Q12l−2 = l(l − 1) dt(1 − t)P(dη, A, Ftl−2 ). (3.67)
0
To express the anomaly in terms of η rather than dη we can shift the derivative from η to A
and Ft . This picks up a d-exact term which is irrelevant. Thus the anomaly can be expressed
as
Z 1
Q12l−2 = (l − 1) dt(1 − t)P(η, dA, d(Ft )l−2 ). (3.68)
0
For the higher terms one needs to consider fields of a different homotopy type.

At = tη. (3.69)
In this case we have

A1 = η, A0 = 0. (3.70)

Then the curvature is

ˆ t + A2 = t(d + s)η + t2 η 2 = tdη + (t2 − t)η 2 ,


F̂t = dA (3.71)
t

thus

F̂1 = dη, F̂0 = 0. (3.72)

Now we apply the transgression formula

ˆ l−1 ,
P((F̂1 )l ) − P((F̂0 )l ) = dQ (3.73)
with
Z 1
Ql−1 = n dt P(η, (tdη + (t2 − t)η 2 )l−1 ). (3.74)
0
The invariant polynomials in this case can be expanded as
l−1
X (l − 1)!
P(η, F̂tl−1 ) = tl−1−k (t2 − t)k P((dη)l−1−k , η, (η 2 )k ). (3.75)
k=0 k!(n − k − 1)!
Now we expand Ql−1 as

Ql−1 = Qll−1 + Ql+1 2l−1


l−2 + · · · + Q0 . (3.76)
Then (3.73) can be expanded as

P((dη)l ) = (s + d)Qll−1 + (s + d)Ql+1 2l−1


l−2 + · · · + (s + d)Q0 . (3.77)

36
Collecting terms in the same order of dη produces the second part of the descent equations

P((dη)l ) − dQll−1 = 0
sQll−1 + dQl+1
l−2 = 0
..
.
2l−2 2l−1
sQ1 + dQ0 = 0
2l−1
sQ0 = 0. (3.78)

From (3.75), after evaluating the integrals in terms of beta functions the terms in chain
(3.78) can be written as

l!(l − 1)!
Ql+k
l−1−k = (−1)
k
P((dη)l−1−k , η, (η 2 )k ), (3.79)
(l − 1 − k)!(l + k)!
with 0 ≤ k ≤ l − 1.

Notice how these terms do not depend on A or F , but just the ghost form η. Although the
chain terms are of different homotopy they do indeed belong to the same chain.

Using equation (3.79) the last term in the chain can be calculated,

l!(l − 1)!
Q2l−1
0 (η) = (−1)l−1 P(η 2l−1 ). (3.80)
(2l − 1)!
In order to compare these results to Section 3.4 the symmetric polynomial P is taken to be
the symmetrised trace. We calculate the ladder for l = 2 explicitly and then compare them
with (3.52). For k = 0 and using (3.66) we immediately get the Chern-Simons form

Z 1
Q03 = 2! dtStr(A, Ft )
0
Z 1
= dtTr(A, tF + (t2 − t)A2 )
0
Z 1  
= dtTr tAF + (t2 − t)A3
0
 
1 1
= Tr AF − A3
2 3
 
1 2
= Tr AdA + A3 , (3.81)
2 3
which up to normalisation agrees with (3.52). Using (3.68) it is clear that the 2-dimensional
anomaly is

1Z 1
Q12 = (1 − t)Tr(ηdA)
2 0
1
= Tr(ηdA). (3.82)
4

37
Which clearly agrees with the previous analysis again up to normalisation. Now we need to
use formula (3.79) to calculate terms lower in the ladder. It is easy to see that using k = 0
and k = 1

Q21 = Tr dηη, (3.83)


and
1
Q30 = − Trη 3 . (3.84)
3
Notice that the Schwinger term (3.83) does not not appear to be the same as (3.50). However,
it is straightforward to see that the difference between the two expressions is BRST-exact,
hence they are BRST cohomologous.

(3.83) − (3.50) = Tr(dηη + η 2 A)


= Trη(dη + Aη − ηA + ηA)
= Trη(dη + [A, η] + ηA)
= −Trη(sA − Aη)
= Trs(ηA). (3.85)

3.6 The Bottom Up Approach to Solving the Descent


Equations
So far we have presented the “top-down” approach to solving the descent equations using
homotopic arguments such as the Russian formula. There also exists a “bottom-up” approach
which starts by solving the bottom descent equation. We follow [58].

sQ02n+1 = 0. (3.86)
The solution to the above is clearly

Q02n+1 = A02n+1 + sQ̂2n


0 . (3.87)
Where A02n+1 is a representative of the cohomology of s. Now inserting (3.87) into the next
line of the descent equations, i.e. p = 1 gives

s(Q2n 2n 2n+1
1 − dQ̂0 ) + dA0 =0 (3.88)
To solve (3.88)one assumes that a special solution Â2n
1 to the descent equations exists

sÂ2n 2n+1
1 + dA0 = 0. (3.89)
Then we can write
 
s Q2n 2n 2n
1 − dQ̂0 − Â0 = 0. (3.90)
Thus we are solving for the cohomology of s again. Hence we have,

38
Q2n 2n 2n 2n 2n−1
1 = Â1 + A1 + dQ̂0 + sQ̂1 , (3.91)
where A2n1 is another representative of the cohomology of s. We can continue this procedure
in an iterative manner

Q2n+1−p
p = Â2n+1−p
p + A2n+1−p
p + dQ̂2n+1−p
p−1 + sQ̂2n+1−(p+1)
p , (3.92)
for 1 ≤ p ≤ D − 1, where D is the space-time dimensions. A2n+1−p
p is a representative of the
cohomlogy of s in the sector of forms of degrees (2n + 1 − p, p) and Â2n+1−p
p depends on the
cohomology of s in the sectors of degree less that p − 1. Then the next line up in the descent
equations is
 
sQ2n+1−(p+1) + d Â2n+1−p
p + A2n+1−p
p − sdQ̂2n+1−(p+1)
p = 0. (3.93)
2n+1−(p+1)
Now assume as before the existence of a special solution Âp+1
 
2n+1−(p+1)
sÂp+1 + d Â2n+1−p
p + A2n+1−p
p = 0. (3.94)
Then applying (3.92) with p → p + 1

2n+1−(p+1) 2n+1−(p+1) 2n+1−(p+1) 2n+1−(p+2)


Qp+1 = Âp+1 + Ap+1 + dQ̂2n+1−p+1
p + sQ̂p+1 , (3.95)

here A2n+1−(p+1) is a representative of the chomology of s in the sector of forms with de-
grees (2n+1−(p+1), p+1). Equation (3.95) is the general solution to the descent equations.

It is worth noting that the condition (3.94) may not be satisfied for all p.

3.7 BRST Cohomology and Physics


So far we have concentrated on the singlet and gauge anomalies that arise in gauge theories.
Other elements in the chain have a physical interpretation.

1. Abelian Anomaly - TrF n+1


The trace of the curvatures gives the abelian anomaly in in (2n + 1) dimensions. The
correct normalisation needs to be calculated via perturbation theory or by the index
theorem. See chapter 2.

2. Topological Field Theory - Q02n+1


The Chern-Simons form is the starting point of all topological field theories.

3. Gauge Anomalies - Q12n


This term is the non-abelian anomaly in 2n dimensions. Again the correct normal-
isation is calculated via pertubation theory or via a generalised index theorem. See
chapter 2.

4. Schwinger terms - Q22n−1


This term to order η 2 has been identified with the Schwinger term in an equal time
commutatior of Gauss-law operators for Yang-Mills theory [32].

39
5. Magnetic monopoles and the Jacobi identity - Q32n−2
This term is related to the breakdown of the Jacobi identity for velocity operators in
the presence of a magnetic monopole [43].

However, higher order terms in the descent equations have so far lacked any physical inter-
pretation.

3.8 Descent Equations for Nontrivial Gauge Bundles


So far in deriving the descent equations the gauge theories considered had trivial fibre-bundle
structure. That is the principle bundles P (M, G) with group structure G over manifold M
were globally M × G. The structure of the anomaly and the descent equations can be gen-
eralised to nontrivial gauge bundles via the introduction of reference connection A0 . This
construction was first proposed by Manes, Stora and Zumino [49]

The reference potential does not transform under a BRST transformation

s A0 = 0 (3.96)
The anomalous Ward identity becomes
Z
sW [A, A0 ] = G[η, A, A0 ] = η α Gα [A, A0 ]. (3.97)
M
The nilpotency of s produces the Wess-Zumino consistency condition

sG[η, A, A0 ] = 0. (3.98)
Following [49] we a transgression formula for non-trivial bundles

P(F l (A)) − P(F l (A0 )) = dQ2l−1 (A, A0 ) (3.99)


Z 1
= ld dtP(A − A0 , Ftl−1 ), (3.100)
0

where we have defined

Ft = dAt + A2t , At = tA + (1 − t)A0 . (3.101)


Now we shift the gauge field A and define a generalised exterior derivative as before

 = A + η
dˆ = d + s. (3.102)

Applying the “Russian Formula” (3.59) to the “shifted transgression” formula

ˆ 2l−1 (A + η, A0 )
dQ2l−1 (A, A0 ) = dQ
Z 1
= ldˆ dtP(A + η − A0 , F̂tl−1 ), (3.103)
0

40
where we have the homotopies

F̂t = dˆÂt + Â2t ,


Ât = t(A + η) + (1 − t)A0 . (3.104)

Expanding the Chern-Simons form in powers of η as before

Q2l−1 (A + η, A0 ) = Q02l−1 (A, η, A0 ) + Q12l−2 (A, η, A0 ) + · · · + Q2l−1


0 (η), (3.105)

and then collecting terms of the same order produces the descent equations for non-trivial
bundles

P (F n (A)) − P (F n (A0 )) − dQ02n−1 (A, A0 ) = 0


sQp2n−1−p (η, A, A0 ) + dQp+1
2n−2−p (η, A, A0 ) = 0
sQ02n−1 (η) = 0 (3.106)

with p = 0, 1, ..., 2n − 2.

Such a generalisation to non-trivial bundles would be needed if magnetic monopoles are


present. However, all fibre bundles are locally trivial and hence our discussion so far is
locally valid provided we are not actually at a monopole.

3.9 Epitome
The local cohomology of the exterior derivative d on the space of form polynomials V(A, F, η, ζ)
is trivial. This allows the descent equations to be derived in a mathematically sound manner.
The descent equations in fact define the local cohomology of s mod d.

Solutions to the descent equations can be obtained by picking an invariant polynomial P2n+2 ,
applying the graded BRST algebra and consulting the descent equations. A “bottom up”
approach also exists.

Calculating solutions in arbitrary dimensions becomes difficult. General solutions can be


obtained via transgression formula and the “Russian Formula”. The non-abelian anomaly
is given by
Z 1
Q12n = n dt(1 − t)P(η, dA, d(Ft )n−1 ), (3.107)
0

where Ft = tF + (t2 − t)A2 .

The remaining question is how to pick the symmetric polynomial P2n+1 ? If the symmetrised
trace is used then one recovers the non-abelian anomaly in 2n dimensions starting from the
singlet anomaly in (2n + 2) dimensions.

41
CHAPTER 4
The Anomalies of Gravity

Here we investigate the anomalies that are present when matter fields are coupled to gravity
as calculated by Alvarez-Gaumé and Witten [3, 6].//
We show that there is a generalisation of the global U (1) chiral anomaly for standard gauge
theory. This is shown to be related to the Dirac genus of the underlying manifold, which
must be of even dimensions. Fujikawa’s method is used to explicitly calculate this anomaly.

There are three kinds of gravitational anomaly each related to the three symmetries of
gravity; Einstein (diffeomorphisms), Lorentz and Weyl. These anomalies destroy the usual
properties of the energy-momentum tensor.

1. The anomaly in local Lorentz transformations is equivalent to the existence of an


antisymmetric part of the energy momentum tensor

hT ab i − hT ba i =
6 0. (4.1)

2. The Einstein anomaly is understood as the nonconservation of the energy-momentum


tensor

∇µ hT µν i =
6 0. (4.2)

3. The Weyl anomaly is equivelent the nonvanishing of the trace of the energy-momentum
tensor and is thus also know as the trace anomaly

hT µµ i =
6 0. (4.3)

We consider only the local Lorentz and Einstein anomalies here. We set up the correct ghost
structure for gravitational theories and construct the correct BRST algebra. Then using the
same arguments as for the gauge theory case the Descent equations are presented and solved.
Using this construction the equivalence of the Lorentz and Einstein anomalies is shown. We
assume that the connection is metric compatible and torsionless.

42
4.1 Chiral U (1) Gravitational Anomalies
We consider a Dirac field coupled to gravity via the Lagrangian
 
√ i
g ψ iγ α eµα (∂µ − ωµab σab ) ψ.
L= (4.4)
4
The Dirac operator, which is hermitian is defined as
i
iD/ = iγ α eµα (∂µ − ωµab σab ), (4.5)
4
where have the spin connection ωµab and vielbein eµα ∈ GL(m,R). We have also defined
i
σa,b = [γa , γb ], (4.6)
2
A complete orthonormal set of eigenspinors is defined as
Z

/ = λn ψn ,
D dm x gψn† (x)ψ(x) = δn,m . (4.7)
In exactly the same way as for the situation in gauge theory, we expand the path integral
measure in terms of these eigenspinors we obtain the following relation

Z
DψDψ exp[S] (4.8)
Z X Z ∞
!
0 √ X
= da0i dbi exp[S + dm x gα(x) Dµ J5µ (x) − 2i lim ψl† (x)γm+1 ψl (x) ].
N →∞
i l=1

Thus we identify the anomalous conservation law as

Dµ J5µ = 2iA5 (x). (4.9)


With J5µ = ψ(x)eµα γ α γm+1 ψ(x) and

X
A5 (x) = lim ψl† (x)γm+1 ψl (x). (4.10)
N →∞
l=1

The above expression is not well defined and needs to be regulated. As before, we pick the
heat kernel method.

Z N
√ X
dm x gA5 (x) = lim ψl† (x)γm+1 ψ(x)
N →∞
l=1
Z ∞ h i
√ X
= lim dm x g ψl† (x)γm+1 exp −λ2l /M 2 ψ(x)
M →∞
l=1
Z ∞ h i
√ X 2
= lim dm x g ψl† (x)γm+1 exp −D
/ /M 2 ψ(x)
M →∞
l=1
h 2
i
/ /M 2 |ψl i|M →∞ .
= Trhψl |γm+1 exp −D (4.11)

Using the othogonality condition of the eigenspinors the only contribution to the anomaly
comes from the zero modes and hence

43
Z h i
2
dm xA5 (x) = Trhψl |γm+1 exp −D
/ /M 2 |ψl i|M →∞
= v+ − v− .

Where v± are the number of normalisable zero modes of the Dirac operator with positive
and negative chirality.

Now we apply the AS index theorem for a spin complex,


Z
v+ − v− = Â(T M )|vol . (4.12)
m

Where Â(T M ) is the Dirac or A-roof genus which is defined as


k
Y xj /2
Â(T M ) = . (4.13)
j=1 sinh(xj /2)

The above can be expressed in terms of the Pontrjagin classes and hence in terms of the
curvature. For example in 4 dimensions the local form of the anomaly is
i
A5 = Dµ J5µ (x) = − µν
αβλρ Rαβ Rλρµν . (4.14)
348 π 2

4.2 The BRST Algebra for Gravitation


In analogy to gauge theory ghosts are introduced for gravity [56, 71]. We have the following
ghost structure (we consider connections with zero torsion and only trivial bundles, also
ignoring any complications due to the Gribov ambiguity). We follow [73]. Generalisations to
gravity theories including torsion have been constructed, see [52]. For an overview of gravity
theories with torsion and Poincare gravity see [23].

ξ α (x) − Einstein Ghost


αab (x) − Lorentz Ghost
σ(x) − Weyl Ghost. (4.15)

There are three BRST operators associated with GR each corresponding to a classical sym-
metry,

sE − Einstein general coordinate transformation


sL − Local Lorentz transformation
sW − Weyl transformation. (4.16)

Consider just the Lorentz BRST transformation. We define the spin connection one form ω
and the zero form Lorentz ghost α

44
1 ab
ω = L ωab , (4.17)
2
1 ab
α = L αab . (4.18)
2
Both ωab and αab are antisymmetric in their indices. {Lab } are the hermitian generators of
SO(k) in some representation, k is the dimension of the Euclidean space.

Let V(ω, dω, α, ρ = dα) be the space of form polynomials. The local space V has a natural
grading given by the sum of the form degree and the ghost number. ω has ghost number
zero where α has ghost number one. In general any p-form with ghost number q is denoted
Ωqp .

The (Lorentz) BRST transformations on the spin connection and Lorentz ghost are

sL ω = −(dα + [ω, α]),


sL α = −α2 . (4.19)

The BRST operator is a nilpotent, s2L = 0. It is worth comparing these transformations with
the gauge theory case. They are almost identical, which is what we would expect as we are
treating gravity as a gauge theory of the Lorentz group.

The Riemann curvature two-form is defined as

R(ω) = dω + ω 2 . (4.20)
Under the Lorentz BRST the curvature two-form transforms as

sL R(ω) = [R(ω), α]. (4.21)


Also the Bianchi identity holds

dR = [R, ω]. (4.22)


Now consider the Einstein BRST transformations. We define F(Γ, dΓξ, v = dξ) as the space
of polynomials in (Γ, dΓ, ξ, v). Here we have defined

Γ = e−1 de + e−1 ωe, (4.23)


to be the Christoffel connection one-from, and we have defined the vielbein e = eµ dxµ with
eµ being a coordinate basis for the tangent bundle.

The Einstein BRST transformations act on the connection and ghost as

sE Γ = Lξ Γ − ∇v
1
sE ξ = [ξ, ξ]. (4.24)
2

45
Here ∇ = d + [Γ, ] and [ξ, ξ] is the graded Lie bracket. Lξ is the Lie derivative taken with
respect to the ghost parameter ξ.

It is convenient to shift this BRST operator by the Lie derivative and define a new BRST
operator

sξ = sE − Lξ . (4.25)
It is then also useful to redefine the basic variables as F(Γ, R(Γ), ξ, v), where R(Γ) = dΓ +
Γ2 . Under the new shifted transformations we obtain the “Yang-Mills like” Einstein BRST
transformations

sξ v = −v 2
sξ Γ = −∇v
sξ R(Γ) = [R(Γ), v]. (4.26)

Notice that both the Lorentz and Einstein BRST transformations are now cast in the same
form as the BRST transformations for gauge theory. This means that we can apply the
powerful machinery developed in section 3 to the gravitational case.

4.3 Pure Lorentz Anomalies


We define a local Lorentz anomaly to be the non-vanishing action of sL on the quantum
action

sL W [ω] = GL [ω, α]. (4.27)


This leads to the Lorentz anomaly consistency condition

sL GL [ω, α] = 0. (4.28)
As discussed in chapter 3, we seek local functionals that are linear in the ghost fields which
are s-closed but not s-exact. The solution can be calculated via the descent equations.

The cohomology of d on V(ω, dω, α, ρ) is important in deriving and solving the descent equa-
tions. As for gauge theory, we use a filtration operator to reduce the calculation to that of a
simpler exterior derivative d0 . It will be continent to change the local variables to (ω, R, α, ρ)

In analogy to the gauge theory case, we define the exterior derivative and a filtration operator
on V(ω, R, α, ρ)

∂ ∂ ∂
d = dω + dα + dR
∂ω ∂α ∂R
∂ ∂ ∂ ∂
N = ρ +α +ω +R . (4.29)
∂ρ ∂α ∂ω ∂R
The exterior derivative decomposes as

46
d = d0 + d1 , (4.30)
with

[N, dn ] = ndn n = 0, 1. (4.31)


d0 d0 = 0. (4.32)

It is easy to see that

d0 ω = R; d0 α = ρ
d0 R = 0; d0 ρ = 0. (4.33)

Thus (4.33) shows that d0 has vanishing cohomology and thus d must also have vanishing
cohomology as the cohomology of d is isomorphic to a subspace of the cohomology of d0 .

We start from with an invariant polynomial of degree (2p + 2). All such polynomials are
expressed in terms of sums and products of traces in the curvature R(ω)

P2p+2 (R) ∼ TrRp+1 . (4.34)


However, such invariant monomials vanish unless p = (2n − 1). Thus as we shall see, pure
Lorentz anomalies can only occur in 4n − 2 space-time dimensions.

Via the Bianchi identity P4n is d-closed,

dP4n = 0. (4.35)
As the cohomology of d is trivial, P4n must also be d-exact

P4n (R) = dQ04n−1 . (4.36)


Then using the fact that the cohomology of d is trival and the anticommuting properties of
d and s, the gravitational descent equations can be derived

sL Q04n−1 + dQ14n−2 = 0
sL Q14n−2 + dQ24n−3 = 0
..
.
4n−2 4n−1
sL Q1 + dQ = 0
4n−1
sL Q0 = 0. (4.37)

We recognise Q14n−2 as the Lorentz SO(4n − 2) anomaly.

As an explicit example, lets consider the n = 1 case. The descent equations are

47
sL Q03 + dQ12 = 0 (4.38)
sL Q12 + dQ21 = 0 (4.39)
sL Q21 + dQ30 = 0 (4.40)
sL Q30 = 0. (4.41)

Lets consider the case for the Dirac field. In this instance we have that

P4n = Â|vol . (4.42)


In two dimensions Â|vol is given by the first Pontrjagin class p1 ∝ TrR2 . Hence, we start
with the the SO(3) Chern-Simons form
 
1
Q03 = Tr ωR − ω 3 , (4.43)
3
the SO(2) Lorentz anomaly is easily calculated via (4.38). We see that

sL Q03 = −dQ12 (4.44)


Acting of the Chern-Simons form with sL gives

sL Q03 = −dTrα(R − ω 2 ) = −dTrαdω. (4.45)


Then the Lorentz anomaly (up to normalisation) is given by

Q12 = Trαdω (4.46)


Other terms in the ladder can be calculated in this way. For completeness we quote them
here. Notice that the calculation is identical to the gauge theory example in Chapter 3.

   
2 1
Q03 = Tr ωdω + ω 3 = Tr ωR − ω 3 ,
3 3
1 2
Q2 = Tr α(R − ω ) = Tr αdω,
Q21 = −Tr α2 ω,
1
Q30 = −Tr α3 . (4.47)
3
Higher dimensional examples can be calculated in this way, however it clearly becomes
clumsy and long winded. A general solution can be obtained via the “Russian Formula”
which is presented in a later section.

Again, we stress that the normalisation of the anomaly is not fixed by the descent equations.

4.4 The Equivalence of Einstein and Lorentz Anoma-


lies
Einstein anomalies are defined in a similar fashion to the Lorentz anomalies,

48
sE W [e] = GE [ξ, Γ]. (4.48)
This leads to the Einstein consistency condition

sE GE [ξ, Γ] = 0. (4.49)
As for the Lorentz anomaly we want to set up the descent equations and provide a systematic
method of solving them. Notice that the shifted Einstein BRST algebra (4.26) is identical to
the Lorentz BRST algebra (4.19),(4.21) with Γ and v playing the role of the spin-connection
ω and the Lorentz ghost α. Thus the cohomological arguments in section 4.3 also hold on
F(Γ, R(Γ), ξ, v). Thus the structure of the descent equations is identical to (4.37), except
now the invariant polynomials are given in terms of the Gl(2n) curvature.

So, the term in the descent equations relevant for the Einstein anomaly is

sξ Q14n−2 + dQ24n−3 = 0. (4.50)


To show that (4.50) does indeed lead back to the consistency condition (4.49) we first assume
that the manifold has no boundary. Thus via Stokes theorem
Z
sξ Q14n−2 (v, Γ) = 0. (4.51)
The Lie derivative Lξ acting on a form of maximal space-time degree gives a d-exact term

Lξ Qqmax = dQq+1
max−1 , (4.52)
thus we recover the consistency condition for the Einstein anomaly.
Z
sE Q12n−2 (v, Γ) = 0. (4.53)
In order to calculate the Einstein anomaly it is sufficient to calculate the Lorentz anomaly
and then replace the spin-connection ω with the Christoffel connection Γ and the Lorentz
ghost α with the ghost variable v. As an explicit example, the two dimensional Einstein
anomaly for the Dirac field viz (4.47) is given by

Q12 (v, Γ) = TrvdΓ. (4.54)

4.5 The Descent Equations for Gravity via the Russian


Formula
We consider only trivial fibre-bundles in this section. The “Russian Formula” equally holds
for both the Lorentz and Einstein sectors

R̂(Γ̂) = ∆Γ̂ + Γ̂2 = dΓ + Γ2 = R(Γ)


R̂(ω̂) = ∆ω̂ + ω̂ 2 = dω + ω 2 = R(ω). (4.55)

Where we have defined

49
Γ̂ = Γ + v, ω̂ = ω + α, dˆ = d + sξ . (4.56)
We also notice that dˆ is a nilpotent

dˆ2 = 0. (4.57)
Again we start with a symmetric invariant polynomial and use the transgression formula.
Consider for the moment just the Einstein sector.

P(Rl ) = dQ2l−1 (Γ, R). (4.58)


Here Q2l−1 is the GL(2l) Chern-Simons form. The transgression formula also holds for the
shifted fields

ˆ 2l−1 (Γ̂, R̂).


P(R̂l ) = dQ (4.59)
The “Russian Formula” then implies

ˆ 2l−1 (Γ + v, R) = dQ2l−1 (Γ, R).


dQ (4.60)
Then expanding the Chern-Simons form in powers of the ghost v we obtain the gravitational
descent equations

P(Rl ) − dQ02l−1 = 0
sξ Q02l−1 + dQ12l−2 = 0
..
.
2l−1
sξ Q0 = 0. (4.61)

The anomaly is given by

Z 1
Q12l−1 (v, Γ) = l(l − 1) dt(1 − t)P(v, dΓ, dRtl−2 ) (4.62)
0
Rt = tR(Γ) + (t2 − t)Γ

Z 1
Q12l−1 (α, ω) = l(l − 1) dt(1 − t)P(α, dω, dRtl−2 ) (4.63)
0
Rt = tR(ω) + (t2 − t)ω.

4.6 Mixed Anomalies


Consider a theory in which the Dirac field is coupled to the gravitational field and a Yang-
Mills field. The corresponding Dirac operator is

/ = γ a eµa (∂µ + Aµ + ωµ ).
D (4.64)

50
The above operator will in general generate gauge, gravitational and mixed anomalies. These
anomalies are best tackled via the index theorem, which is used to generate the necessary
invariant polynomials needed at the top of the descent equations.
Z
/+ =
ind D ch(F ) ∧ Â(T M )|vol . (4.65)
M

/ +.
Where D+ = iDP

The symmetric polynomials for the first few dimensions are

i Z
2 − dim ind D+ = TrF
2π M2
 
1 Z 1 2 r 2
4 − dim ind D+ = − TrF + TrR (4.66)
(2π)2 M4 2 48
Z  
1 i 3 i 2
6 − dim ind D+ = − TrF + TrF TrR .
(2π)3 M6 6 48
Where r is the dimensions of the gauge group representations. The various anomalies can
then be calculated from these invariant polynomials using the methods described earlier.

For other fields coupled to gravity and gauge fields, the correct invariant polynomials can
be constructed (including the correct normalisation) from the Atiyah-Singer index theorem
for the relevant elliptical operator. This is described in the next chapter.

4.7 Epitome
The Dirac field coupled to gravity has a chiral anomaly which is the analogue of the singlet
anomaly in gauge theory. Via Fujikawa’s method and appealing to the AS index theorem
this anomaly is
Z Z
dxA5 = Â(T M )|vol . (4.67)
m
The BRST algebra for both the Lorentz transformations and Einstein differomorpisms have
the same structure as the BRST algebra for gauge theory. This enables us to apply the
machinery of chapter 3 to gravity. As the theory is assumed to be torsion free the curvatures
defined by either the Christoffel connection one-form or the spin-connection are identical.
This allows the Lorentz and Einstein anomalies to be identified.

The “Russian Formula” and transgression formula give an explicit form for the Lorentz
anomaly
Z 1
Q12n (α, ω) =n dt(1 − t)P(α, dω, d(Rt )n−1 ) (4.68)
0
The Einstein anomaly is given by the replacement of ω with Γ and α with v.

If the invariant polynomial P4n is given by the Â-genus then the in complete analogy to the
gauge theory case we recover the gravitational anomaly in terms of the chiral U (1) anomaly

51
for the Dirac field coupled to gravity.

52
CHAPTER 5
The Atiyah-Singer Index Theorem and
Anomaly Polynomials

In the previous chapters the invariant polynomial P2n+2 have been left undetermined except
when explicit calculations were preformed for the Dirac field. In this case we used the chi-
ral anomaly in (2n + 2) as calculated by the Atiyah-Singer index theorem as the invariant
polynomial (ignoring at that stage normalisation).

It is well-known that the only possible fields that can give rise to anomalies are the spin-1/2
Dirac field, the spin-3/2 Rarita-Schwinger field and the (anti-)self-dual antisymmetric tensor
field. The question is what invariant polynomial should be used?

Although the anomaly is not uniquely determined by the descent equations (due to topologi-
cally trivial terms), we propose that the invariant polynomials are fixed including the correct
normalistion if one appeals to the Atiyah-Singer index theorem for the relevant operator in
question. Thus, the anomaly including normalisation is determined.

The mathematical proof of the above statement requires k-theory. For a very accessible
introduction to k-theory (both differential and algebraic) see [72]. For the application of
k-theory to anomalies see [54]. Alvarez-Gaumé and Ginsparg [4, 5] demonstrated how the
anomaly in 2n dimensions is related to the AS theorem in a (2n+2) dimensional space. Their
method avoids the use of k-theory, but in some sense it closely resembles it. See [20, 53] for
a modern account.
In this chapter we present the necessary complexes associated with the Dirac, Rarita-
Schwinger and self-dual tensor fields. We then apply the AS-theorem to generate the invari-
ant polynomials.

We then present the symmetric polynomials for the spin-1/2, spin-3/2 and (anti-)self-dual
tensor field in four, six and ten dimensions. Then the anomaly content of effective super-
garvity theories in ten dimensions are examined. These supergravity theories are of partic-
ular interest as they are the low energy effective theories of the known five consistent string
theories.

53
5.1 Twisted spin complexes
Consider a spin bundle S(M ) over an m-dimensional orientable manifold M . The sections of
this bundle are denoted ∆(M ) = Γ(M, S(M )). A Dirac spinor ψ ∈ ∆(M ) is an irreducible
representation of the Clifford algebra, but is not an irreducible representation of the SPIN(m)
group generated by the Dirac matrices. ∆(M ) can be decomposed in to eigenvectors of γ m+1 ,
with the eigenvalues ±1 being the chirality.
M
∆(M ) = ∆+ (M ) ∆− (M ), (5.1)
The irreducible representations of SPIN(m) are the spinors in ∆± , with

ψ + ∈ ∆+ (M ), ψ − ∈ ∆− (M ). (5.2)
Where we have γ m+1 ψ ± = ±ψ ± . The Dirac operator is defined as

/ = iγ a ea µ (∂µ + ωµ ).
iD (5.3)
ab
ωµ is the spin connection; ωµ = −i/4 ωmu σab . Defining the projection operators as
1 1
P + = (I + γ m+1 ), P − = (I − γ m+1 ), (5.4)
2 2
allows us to define

/ + : ∆+ (M ) → ∆− (M )
D = iDP
/ − : ∆− (M ) → ∆+ (M ).
D† = iDP (5.5)

Hence we have the two term complex


D
∆+ (M )  - ∆− (M ) (5.6)
D†
which is known as the spin complex. The analytical index of this complex gives the difference
in the number of zero-modes with different chirality

v+ − v− = ind D = dim kerD − dim ker D† . (5.7)


Applying the AS-index theorem to this spin complex gives

Z
Td(T M C )
v+ − v− = ch(∆+ (M ) − ∆− (M )) |vol
M e(T M )
Z
= Â(T M )|vol . (5.8)
M

Here  is the Dirac or A-roof genus.

Next we consider the twisted spin complex. A spinor that belongs to a representation of
N
a Lie group G is a section of the bundle S(M ) E, where is the associated vector bundle
P (M, G) in a given representation. The Dirac operator is defined as

54
DE = iγ α eαµ (∂µ + ωµ + Aµ )P + . (5.9)
Where Aµ is the gauge field on E. The AS index theorem for the twisted spin complex is
Z
v+ − v− = Â(T M )ch(E)|vol . (5.10)
M

5.2 The Rarita-Schwinger spin complex


The full construction of the Rarita-Schwinger spin complex requires the use of K-theory.
The usual method is to construct Rarita-Schwinger chiral bundles ∆±
3/2 (M ) with local coor-
dinated (xµ , ψν ) as virtual bundles.

The twisted Rarita-Schwinger complex is defined in analogy to the spin complex. The spinors
N
are sections of the bundle ∆±
3/2 E. The AS index theorem for the twisted Rarita-Schwinger
complex gives the difference of normalisable zero-modes of the twisted Rarita-Schwinger
operator DRS
Z
3/2 3/2
v+ − v− = ind DRS = [Â(T M )(Tr exp(iR/2π) − 1)ch(E)]|vol . (5.11)
M

5.3 The signature complex


The antisymmetric tensor fields with self-dual (+) or anti-self-dual field strengths form bun-
dles, Λ± (M ) over the manifold M . The signature complex is the correct mathematical
description of these fields.

In order to define the signature complex we need to introduce an operator that maps Ωp (M )
into Ωn−p ;τ . Here Ωp denotes the space of complex p-forms over M and n is the dimension
of the manifold and is taken to be even. τ is in fact related to the Hodge star ∗,

τ = ip(p−1)+n/2 ∗ . (5.12)
From the definition of τ and using the fact that ∗2 = (−1)p implies that τ 2 = 1 and has
eigenvalues ∓1. The exterior algebra can thus be decomposed into the eigenspaces of τ
M
Ω∗ (M ) = Ωp (M ) = Ω+ ⊕ Ω− . (5.13)
p

Notice that d + d∗ maps Ω+ into Ω− and vice versa. We define the operator d+ as the
restriction of d + d∗ to Ω+ .

d+ : Ω+ → Ω− . (5.14)
It is easy to see that d− is indeed the adjoint of d+ and is the restriction of d + d∗ to Ω− .
We can thus define the two term signature complex

d+
Ω+ (M )  - Ω− (M ) (5.15)
d−

55
The index of d+ is known as the sign of M , Sign(M ). See [54] for a deeper discussion on the
sign of M .

Applying the AS index theorem to the signature complex gives


Z n
xi
Y
ind d+ = |vol . (5.16)
M i=1 tanh xi

It is customary to define the Hirzebruch L-polynomial,


n
Y xi
L(T M ) = , (5.17)
i=1 tanh xi
and write (5.16) as
Z
ind d+ = L(T M )|vol . (5.18)
M
Now the signature complex can be twisted by taking the coefficients to belong to an arbitrary
bundle. The AS index theorem for the twisted signature complex is
Z
ind dE
+ = ch(E) ∧ L(T M ). (5.19)
M

5.4 The Invariant Polynomials in Various Dimensions


The so called “master” equation for the anomalies is the BRST variation of the quantum
action
Z
sW = −2πi G12n (5.20)
M
1
The descent equations then allow the anomaly G2n to an invariant polynomial P2n+1 .

dQ02n+1 = P2n+2
sQ02n+1 + dQ12n = 0. (5.21)
For the three kinds of potentially anomalous matter we have the following symmetric poly-
nomials 1

h i
Â(T M )ch(E)
(1/2)
P2n+2 = (5.22)
vol
h i
Â(T M ) (Tr exp(i/2π R) − 1) ch(E)
3/2
P2n+2 = (5.23)
vol
 
A 11
P2n+1 = − L(T M ) , (5.24)
24 vol

which correspond to the Dirac field, Rarita-Schwinger and self-dual form field. In the latter
a factor of −1/2 is needed from the duality conditions.

The invariant polynomials can be expanded in terms of Pontrjagin classes. We have in two,
six and ten dimensions the following polynomials
1
String theorists use Iˆ to denote the symmetric polynomials.

56
• Anomaly Polynomials in four dimensions
Only the chiral spin-1/2 fermion can contribute to the anomaly
 
1 i i
P 1/2 = − TrF 3
+ TrF TrR2 . (5.25)
(2π)3 6 48
• Anomaly Polynomials in six dimensions
All three kinds of fields generate potential anomalies.

   
1/2 1 4 1 1 1
P = TrF + r TrR4 + (TrR2 )2 − TrR2 TrF 2
4!(2π)4 246 192 4
 
1 49 43
P 3/2 = TrR4 − (TrR2 )2
4!(2π)4 48 192
 
1 7 1
PA = TrR4 − (TrR2 )2 . (5.26)
4!(2π)4 60 24

• Anomaly Polynomials in ten dimensions


Again all three fields contribute to potential anomalies.

  
1 1 1 5
P 1/2 = 6
−TrF 6
+ r TrR 6
+ TrR 4
TrR 2
+ (TrR2 )3
6!(2π) 504 384 4608

1 5 5
− TrR4 TrF 2 + (TrR2 )2 TrF 2 − TrR2 TrF 4
16 64 8
 
1 55 75 35
P 3/2 = − TrR 6
+ TrR 4
TrR 3
− (TrR 2 3
)
6!(2π)6 56 128 512
 
1 496 7 5
PA = TrR 6
− TrR 4
TrR 2
+ (TrR 2 3
) . (5.27)
6!(2π)6 504 12 72

5.5 Anomalies in Supergravity Theories


Supergravity theories are known to be the low energy limit of of the five known string
theories. It is interesting to note that they are in fact anomaly free. We can use the
invariant polynomials as defined in the previous section to prove this statement. We discuss
the various supergravity theories individually

• Type IIA Supergravity


This theory is non-chiral and thus is free of (local) anomalies.

• Type IIB Supergravity


In 10 dimensions type IIB supergravity has a self-dual five-form field, a pair of chiral
spin-3/2 Majorana-Weyl gravitinos and a pair of anti-chiral Majorana-Weyl spin-1/2
fermions. The total anomaly polynomial is thus

1 1
P(R) = P A (R) + 2 P 1/2 (R) − P 3/2 (R). (5.28)
2 2

57
Where we have used just the gravitational part of the anomaly for the spin-1/2 fields.
The factor of 1/2 is included as we have Majorana-Weyl spinors and not Dirac spinors.
Thus, P = 0 when we add up all the terms. Type IIB supergravity has no anomalies.

• Type I Supergravity coupled to super Yang-Mills


Type I supergravity theory is chiral and in general will posses anomalies. These anoma-
lies vanish if type I supergravity is coupled to E8 × E8 or SO(32) super-Yang-Mills,
this is the well known Green-Schwartz mechanism [38].

Type I supergravity in ten dimensions has a chiral Majorana-Weyl spin-3/2 gravitino


and a anti-chiral Majorana-Weyl spin-1/2 dilatino. The super-Yang-Mills theory con-
tains a chiral Majorana-Weyl spin-1/2 gauginos transforming under the adjoint rep-
resentation of some as of yet unspecified Lie group G. The anomaly polynomial is
thus

1
P(R, F ) = (P(R)3/2 − P(R)1/2 + P(R, F )1/2 ). (5.29)
2
Here I 1/2 (R) is the term from the pure gravitational anomaly of the gravitino and
I 1/2 (R, F ) is the anomaly polynomial for the super-Yang-Mills gaugino which contains
a gravitational, gauge and mixed part. Using the explicit formula (5.27) we have


−1 496 − r 224 + n 5
P(R, F ) = 6
TrR6 − TrR4 TrR2 + (64 − r)(TrR2 )3
2(2π) 6! 504 384 4608

1 4 2 5 2 2 2 5 2 4 6
+ TrR TrF + (TrR ) TrF − TrR TrF + TrF (5.30)
16 64 8

In order for the anomaly to be cancelled by a local counter term the above anomaly
polynomial has to factorise into a four-form and an eight-form. The term TrR6 does
not allow this factorisation and so it must vanish. Thus we are led to the first condition
on the gauge group

r = 496. (5.31)

To factorise the remaining anomaly polynomial we require that

1 1
TrF 6 = TrF 4 TrF 2 − (TrF 2 )3 . (5.32)
48 14400
The 496-dimensional Lie groups SO(32) and E8 × E8 have the above property. They
also hold for E8 × U (1)248 and U (1)496 , however no such string theories are known. The
anomaly polynomial reads
 
1 1 1
P=− 2
TrF 2 − TrR2 X̂, (5.33)
2 (2π) 2! 30
where we have defined

58
 
1 1 1 1 1 1
X = 4
TrR4 + (TrR2 )2 + TrR2 TrF 2 TrF 4 − (TrF 2 )2 . (5.34)
(2π) 4! 8 32 240 24 7200

5.6 Epitome
The non-abelian anomaly in 2n dimensions G[η, A], is given via the descent equations in
terms of a symmetric invariant polynomial in 2n + 2 dimensions P2n+2 . The exact form
including the normalisation of the invariant polynomial is given by the Atiyah-Singer index
theorem in 2n + 2 dimensions. Thus, classical index theorems completely determine the
quantum anomalies in chiral theories.

The invariant polynomials are calculated by applying the Atiyah-Singer index theorem to
the relevant complexes. We have for spin-1/2 Dirac fermions, spin-1/2 Rarita-Schwinger
fermions and the (anti-)self-dual form field respectively

h i
Â(T M )ch(E)
(1/2)
P2n+2 = (5.35)
vol
h i
Â(T M ) (Tr exp(i/2π R) − 1) ch(E)
3/2
P2n+2 = (5.36)
vol
 
A 11
P2n+2 = − L(T M ) . (5.37)
24 vol

Using these invariant polynomials one can examine the anomalies in various theories without
knowing the details of the theory. The matter content is all that is needed. For example, it
was shown that the supergravity theories that are the low energy limit of the five consistent
superstring theories are anomaly free after applying the Green-Schwartz mechanism.

59
CHAPTER 6
Conclusion

This thesis addressed the calculation of potential anomalies in quantum field theories. The
main emphasis was on the calculation of consistent anomalies via the Wess-Zumino consis-
tency conditions and the descent equations. This approach was adopted as the graded BRST
algebra and the descent equations do not depend explicitly on a specific Lagrangian, but
are a generic feature of gauge (gravity) theories. The only drawback to this approach is the
normalisation is not fixed. This requires topological analysis or perturbation theory.

This approach is powerful enough to calculate the anomalies that arise when Dirac fermions,
Rarita-Schwinger fermions or (anti-)self-dual form fields are coupled to gauge and/or gravity
fields. Only the symmetric polynomial at the “top” of the descent equations fixes the rest of
the s-ladder. The symmetric polynomial is selected via the Atiyah-Singer index theorem for
the relevant elliptical operator. Amazingly, using the index theorem in (2n + 2) dimensions
gives the correct normalisation of the anomaly in 2n dimensions. Thus, the anomalies can
be calculated with relative ease.

The Alvarez-Gaumé and Ginsparg index procedure relates the singlet anomaly in (2n + 2)
dimensions to the non-abelian anomaly in 2n dimensions, it also gives the correct normalisa-
tion. k-theory is needed to show that the Atiyah-Singer index theorem in 2n + 2 dimensions
gives the correct anomaly in 2n dimensions in general. Due to time and space restrictions
this was not elaborated on in this thesis.

As the anomalies are calculated independently of any lagrangian, only the field content is
need in order to examine the anomalies. This leads means that the potential anomalies in
string theories, supergravity theories etc. can be examined without detailed analysis of the
theory.

60
Appendices

61
APPENDIX A
Non-Abelian Gauge Theory and BRST
Symmetry

A.1 Classical Non-Abelian Gauge Theories


Yang and Mills [76] generalised electromagnetism to include non-abelian gauge groups. Here
we briefly review the classical (pure) Yang-Mills theory needed for later sections. Yang-Mills
theory is a generalisation for Maxwell’s electromagnetism from an abelian gauge group U (1)
to a nonabelian group, G. The essence of gauge theory is to extend the notion of a global
symmetry to one that is now local, i.e. it varies with space-time. This seeming simple
extension leads to highly nontrivial and nonlinear constraints on the quantum field theory.
First some basic facts about Lie groups and Lie algebras are needed. A representation of a
Lie algebra is a set of N anti-Hermitian matrices T a , a = 1, 2...N obeying

[T a , T b ] = f abc T c . (A.1)
Where the f’s are the structure constants of the Lie group G.

For SU (2), the f abc will be equal to abc . Thus in the isospin representation we have
iσ a
Ta = − . (A.2)
2
Where σ a are the Pauli spin matrices.

The gauge potentials are vector fields Aaµ (x). It is often convenient to define matrix valued
vector field Aµ known as the Yang-Mills connection.

Aµ = Aaµ Ta . (A.3)
Where g is the gauge coupling constant.

Let the field φi (which could be fermionic or bosonic) transform under some representation
of G.

62
φi (x) → Λij (x)φj (x). (A.4)
Where Λij ∈ G. We do not restrict ourselves to the fundamental representation. The group
element, which is a function of space-time can be parameterised as

Λij (x) = (exp[−iθa (x)T a ])ij . (A.5)


Where the parameters θa (x) are local variables and T a is defined in the representation under
consideration.

Using the gauge potential it is now possible to construct the covariant derivative Dµ .

Dµ = ∂µ − igAµ . (A.6)
The important point is that the covariant derivative of a field φ is gauge covariant. That is

(Dµ φ)0 = ∂φ0 − igA0µ φ0


= Λ∂µ φ + (∂µ Λ)φ − igA0µ Λφ
= ΛDµ φ. (A.7)

This can be achieved if the term ∂µ Λ is cancelled by the variation of Aµ


i
A0µ = − [∂µ Λ(x)]Λ−1 (x) + Λ(x)Aµ (x)Λ−1 (x). (A.8)
g
Infinitesimal variations are

1
δAaµ = − ∂µ θa + f abc θb Acµ
g
δφ = −igθa T a φ. (A.9)

From the covariant derivative the Yang-Mills field strength tensor is constructed as the
curvature with respect to the Yang-Mills connection. The Yang-Mills field strength tensor
Fµν is defined as

i
Fµν = [Dµ , Dν ]
g
= ∂µ Aν − ∂ν Aµ + g[Aµ , Aν ]
= (∂µ Aaν − ∂ν Aaµ + gf abc Abµ Acν )T a
a
= Fνµ T a. (A.10)

We wish to construct an invariant action out of the field strength tensor. As Dµ is covariant
we have

Fµν = ΛFµν Λ−1 . (A.11)


Where Λ ∈ G.

63
The unique action with only two derivatives is constructed from the trace. However, it
possible to define an action using other functionals. The problem with these are that they
will contain higher derivatives and hence ghosts which spoil the unitary properties needed
to be physical. So we use the functional

tr(ΛFµν Λ−1 ΛF µν Λ−1 ) = tr(Fµν F µν ), (A.12)


which is clearly gauge invariant and it is also Lorentz invariant. Then we construct the
Yang-Mills action as
1Z
S= tr(Fµν F µν )dx4 . (A.13)
4
All of the above can be set in the language of differential forms by defining

1 a
F = F Ta dxµ ∧ dxν
2 µν
1
= Fµν dxµ ∧ dxν
2
= dA + A ∧ A. (A.14)

Where we have defined

A = Aaµ Ta dxµ
= Aµ dxµ . (A.15)

The Bianchi identity expressed in terms of the exterior derivative d, becomes

DF = dF + [A, F ] = 0. (A.16)
Then in this notation the Yang-Mills action becomes
1Z
S=− T r(F ∧ ∗F ) (A.17)
2 m
Where ∗ is the Hodge dual, which on a Euclidean manifold in local coordinates can be
defined as
1
∗Fµν = µναβ F αβ (A.18)
2
Variation of the action with respect to Aµ produced the field equation

Dµ F µν = 0. (A.19)
Or in form notation

D ∗ F = 0. (A.20)

64
A.2 Instantons
Instantons are self-dual solutions of Yang-Mills theory in a Euclidean space, for which the
field strength tensor is maximal. By dual we mean ∗F = ±F , self-dual is + and anti-self-dual
is −. As instantons mediate between different vacua they are very important in understand-
ing the vacua in gauge theories.

In order to solve the self or anti-self dual equations, one needs to consider the homotopy of
the gauge group G. The boundary conditions are asymptotic conditions which describe how
A behaves at infinity in <4 . The requirement that the action must be finite means that Fµν
must be square integrable.

Fµν (x) → 0,
|x| → ∞. (A.21)

Which also imply

i
Aµ (x) → ∂µ Λ(x)Λ−1 (x),
g
|x| → ∞. (A.22)

Λ(x) can be thought of as being defined by a sphere at infinity in <4 , i.e. on S 3 . Choose
G = SU (2), then for each x, Λ(x) ∈ SU (2) gives a continuous map

Λ : S 3 → SU (2). (A.23)
Such maps fall into homotopy classes and are elements of π3 (SU (2)). But SU (2) is topolog-
ically S 3

π3 (SU (2)) = π3 (S 3 ) = Z. (A.24)


The mappings λ : S 3 → S 3 are characterised by integers, that is the points of one S 3 can
be mapped smoothly to another S 3 such that the winding around S 3 an integer number of
times k.

Using the language and formalism of principle fiber bundles with group SU (N ), it can be
shown that the instanton winding number is equal to the 2nd Chern class. As k must be an
integer we have 3 possibilities

1. If k = 0 then S = 0, so we have a flat connection Aν (x),

2. If k > 0 then F = ∗F , we have self dual solutions and S = 8π 2 n,

3. If k < 0 then F = − ∗ F and we have anti-self dual solutions.

Remembering that G = SU (N ), the Chern characters are

1. ch1 = 0 = c1 ,

65
2. ch2 = − 8π1 2 T r(F ∧ F ).
1 R
Then the winding number can be expressed in terms of the Chern classes, c2 = 8π 2
T r(F ∧
F ) = −n, therefore
1 Z
n = − 2 T r(F ∧ F ) (A.25)

Using the definitions of the forms in local coordinates, it is straight forward to show
1 Z
n=− T r(Fνµ ∗ Fνµ ) (A.26)
16π 2

A.3 Faddeev-Popov Ghosts and Gauge Fixing


Now we consider how to quantise pure Yang-Mills theory via path integrals. There are sev-
eral complications that do not arise in abelian gauge theory. These originate primarily from
the gauge fixing, which effects the path integration measure in a non-trivial way. We briefly
review the issues of gauge fixing which leads to the Faddeev-Popov determinant [31], which
produces the correct measure. This determinant can be reexpressed in terms of the now
famous Faddeev-Popov ghost fields.

It was noticed that the gauge fixed Lagrangian possess a new global symmetry that rotates
the gauge fields into ghosts. This is known as the BRST transformation [15],[69]. This sym-
metry can be used in a similar way to the Gupta-Bleuler constraint to remove unphysical
states from the Fock space. However, there is one potential source of difficulties. We demon-
strate the Gribov ambiguity [39] in the Coulomb gauge means that much of the formalism
developed may not exist outside of perturbation theory.

Feynman in 1962 showed that using standard quantisation methods available at the time,
Yang-Mills theory was not unitary. Feynman also showed that counter terms could be added
that removed the nonunitary parts. These terms are now known as Faddeev-Popov Ghosts.

In gauge theories like Maxwell’s theory or Yang-Mills the path integral is ill-defined due to
the gauge degree of freedom. In these gauge theories both the path integral measure DA
and the action S are gauge invariant and so when we functionally integrate over DA we over
count the degrees of freedom. To see this consider QED. Maxwell’s theory is invariant under
gauge transformations of the form

AΛµ = Aµ + ∂µ Λ, (A.27)
so that when one functionally integrates over both AΛµ and Aµ the integrand is infinity over
counted.
Z
DA eiS = ∞. (A.28)
The problem is that when one starts with a particular configuration Aµ and then consider
all possible gauge equivalent configurations AΛµ we over count the degrees of freedom. Thus
we are sweeping out an orbit in the space of all gauge connections. As Λ changes along the

66
orbits we inevitably over count the path integral an infinite number of times.

The solution to this is to fix the gauge. Traditionally one adds the term
1
LGF = − (∂µ Aµ )2 , (A.29)

to the action. However for the following argument we will fix the gauge by inserting an
arbitrary function of the gauge fields into the integration

δ(F (Aµ )) (A.30)


which forces the gauge condition to be F (Aµ ) = 0.

However, by inserting a delta function into the path integral we change the functional inte-
gration measure DA. Inserting delta functions always produces an ambiguity in the measure.
To remove this ambiguity we insert 1 into the integral.
Z
1 = ∆F P DΛ δ(F (AΛµ )), (A.31)
here ∆F P is the Faddeev-Popov determinant which gives the correct measure. DΛ =
Q
x dΛ(x) is the invariant group measure. Inserting this into the functional integration we
obtain
Z  Z   Z 
DA ∆F P DΛ δ(F (AΛµ ) exp i 4
d x L(A) . (A.32)
It is important to notice that the Faddeev-Popov determinant is independent of the gauge.
This is clear as by construction we integrate over all gauge factors Λ.

∆F P (Aµ ) = ∆F P (AΛµ ). (A.33)


Now we make a gauge the transformation Aµ → AµΛ . The measure DA, the Faddeev-
−1

Popov determinant ∆F P and the action S are all gauge invariant. The only part that is not
is F (AΛµ ). The path integral becomes
Z Z  Z 
4
DΛ DA ∆F P δ(F (Aµ )) exp i d x L(A) . (A.34)
R
Now the infinite part of the integration has been isolated, DΛ = ∞ which is a volume
factor. Hence, simply dividing by this factor renders the integral finite. The gauge condition
is enforced by the delta function δ(F (Aµ )). However, the most important result is that
the Faddeev-Popov determinant ∆F P gives the correct integration measure. The finite path
integral is thus
Z  Z 
4
DA ∆F P δ(F (Aµ )) exp i d x L(A) . (A.35)
The problem of gauge fixing then becomes one of finding an explicit form of the determi-
nant.The method is to use equation (A.31) and to change variables in the integration from
Λ to F . Thus, the Jacobian is calculated
" #
δF
DF = det DΛ. (A.36)
δΛ

67
Then the Faddeev-Popov determinant can be written as

Z " #
δΛ
∆−1
FP = DF det δF
δF
" #
δΛ
= det . (A.37)
δF F =0

Where hence we can write


" #
δF
∆F P = det . (A.38)
δΛ F =0
In order to do calculations it will be convenient to write the determinant in such a way as
new Feynman rules can be derived. We power expand the function F (AΛµ ) for small group
parameter θ.
Z
F (AΛµ (x)) = F (Aµ (x)) + d4 yM (x, y)θ(y) + O(θ2 ). (A.39)
When one takes the derivative only the matrix M survives. Then the Faddeev-Popov deter-
minant can be written as the determinant of the matrix M . The matrix determinant can
then be converted to a Gaussian integral over Grassmann fields η and eta† .
Z  Z 
∆F P = det M = Dη Dη † exp i d4 xd4 y η † (x)M (x, y)η(y) . (A.40)

The Grassmann fields η and η † are ghost fields. That is they do not obey the spin-statistics
theorem, in fact they are scalar fields that obey Fermi-Dirac statistics. These fields are
known as the Faddeev-Popov ghosts.

Now we explicitly calculate the Fadeev-Popov ghosts for QED. We choose the Lorentz gauge
condition

F (Aµ ) = ∂ µ Aµ = 0. (A.41)
Clearly under a infinitesimal gauge transformation we have

F (AΛµ ) = ∂ µ Aµ + ∂ µ ∂µ θ. (A.42)
Then the matrix M is given by

M (x, y) = [∂µ ∂ µ ]x,y . (A.43)


Writing the determinant in terms of the ghost fields means we must include the following
term in the action
Z
d4 xd4 y η † (x)∂ µ ∂µ η(y). (A.44)
This determinant does not couple to any of the gauge fields or fermions of QED and hence
decouples from the theory giving only a multiplicative factor which can be removed at will.
This is why in QED the gauge fixing can be done without any worry about the effect on the

68
measure, the Faddeev-Popov determinant is trivial.

The same cannot be said of Yang-Mills theory. Here the Faddeev-Popov ghosts give non-
trivial corrections. This is partly why Yang-Mills theory took so long to quantise. Again we
pick the Lorentz condition

∂ µ Aaµ = 0. (A.45)
Then when placing this condition into the integration we must also include the term

∆F P = det(M )x,y;a,b . (A.46)


Here x and y are discretised space-time variables and a and b are gauge indices. The matrix
M is easily calculated

!
δ −1
Ma,b (x, y) = a
∂µ (∂µ θb (y) + gf bcd Acµ (y)θd (y))
δθ (x) g
1  ab µ 4 
= −δ ∂ ∂µ δ (x − y) + gf bcd ∂ µ Acµ δ ad δ 4 (x − y)
g
1  ab µ 
= −δ ∂ ∂µ + gf abc ∂ µ Acµ δ 4 (x − y). (A.47)
g x,y

After rescaling and preforming the y integrations we see that the term added to the action
is
Z  
d4 x η † a (x) δ ab ∂ µ ∂µ − gf abc ∂ µ Acµ (x) η b (x). (A.48)
The ghost fields have some unusual properties:

1. They violate the spin-statistics theorem. They transform as scalars under the Lorentz
group, but have fermionic statistics.

2. They couple only to the gauge field for Yang-Mills theory. They do not appear as
external states and as such cannot be present in tree diagrams. They only appear as
in loop diagrams.

3. They decouple from the physical states and should be viewed as a mathematical trick
of quantisation.

A.4 BRST Quantisation of Gauge theory


Once the gauge has been fixed the theory is no longer gauge invariant. The gauge fixing term
removes the local degrees of freedom (up to Gribov ambiguity discussed later). However,
the theory now has a new global symmetry involving the rotation of gauge fields into ghosts
and anti-ghost fields. This symmetry is named after it’s discoverers Becchi, Rouet, Stora
[18] and independently Tyupin [69], thus BRST symmetry. As this is a global symmetry no
new degrees of freedom can be eliminated.

69
We write the gauge fixed action including the Ghost term as
1 1
L = − tr (Fµν F µν ) − (∂.A)2 − η a ∂ µ Dµ η a . (A.49)
2 2α
Before gauge fixing the action was invariant under gauge transformations of the form
1
δAaµ = ∂µ Θa + f abc Abµ Θc . (A.50)
g
Redefining the gauge parameter as

Θa = −η a λ, (A.51)
here η a and λ are Grassmann variables and λ is constant. The gauge fixed action equation
(A.49) is invariant under the global symmetry

1
δAaµ = − (Dµ η a )λ
g
1
δη a = − f abc η b η c λ
2
1
δη a = − (∂ µ Aaµ )λ. (A.52)
αg

It is straight forward to see that the Lagrangian (A.49) is indeed BRST invariant. The
tr(Fµν F µν ) term is clearly BRST invariant as it is just a reparametrisation of the gauge
transformation. Next consider the ghost term

δLF P = −(δη a )∂ µ Dµ η a − η a ∂ µ δ(Dµ η a ). (A.53)


Now consider the gauge fixed part

1
δLGF = − (∂ µ Aaµ )∂ ν δAaν
α
1 µ a ν
= (∂ Aµ )(∂ Dν η a )λ. (A.54)
αg
Adding these two contributions together

1 µ a ν
δL = (∂ Aµ )(∂ Dν η a )λ − (δη a )∂ µ Dµ η a − η a ∂ µ δ(Dµ η a )
αg
1 µ a ν 1 µ a
= (∂ Aµ )(∂ Dν η a )λ + (∂ Aµ )λ(∂ ν Dν η a )
αg αg
−η a ∂ µ δ(Dµ η a )
= −η a ∂ µ δ(Dµ η a ). (A.55)

Remembering that λ and η being Grassmann variables anti-commute. Provided δ(Dµ η a ) = 0


then the theory will be BRST invariant. We now proceed to prove this

70
δ(Dµ η a ) = δ[∂µ η a + g f abc Aµ η c ]
= ∂µ δη a + g f abc (δAbµ )η c + g f abc Abµ (δη c )
1
= − f abc ∂µ (η b η c )λ − f abc (Dµ η b )λη c
2
g
− f abc f cmn Abµ η m η n λ. (A.56)
2
Now consider the derivative in the first term. It is easy to show that

f abc ∂µ (η b η c ) = 2f abc (∂µ η b )η c . (A.57)


Then we obtain
δ(Dµ η a ) = −f abc (∂µ η b )η c λ − f abc (∂µ η b )λη c
g abc cnm b m n
+gf abc f bmn Am n c
µη η λ− f f Aµ η η λ

2 
abc bmn m n c 1 abc cnm b m n
= g f f Aµ η η − f f Aµ η η λ. (A.58)
2
Then using the Jacobi identity

f abc f cnm + f amc f cnb + f anc f cbm = 0. (A.59)


We obtain
δL = g(f abc f bmn Am n c
µη η
1 amc cnb b m n 1 anc cbm b m n
f f Aµ η η + f f Aµ η η )λ. (A.60)
2 2
Interchanging n and m in the last term shows that this term is identical to the middle term.
Then we obtain

δL = g(f abc f bmn + f anb f bcm )Am n c


µη η λ

= g(−f abn f bmc + f anb f bcm )Am n c


µη η λ
= 0. (A.61)
Hence we have proved that gauge fixed Yang-Mills is indeed invariant under the BRST trans-
formations.

It is straight forward to see that the BRST variational is a nilpotent

2
δBRST = 0. (A.62)
From this symmetry a Noether current can be constructed

XδL δBRST φi
Jµ = µ
i δ∂ φi δλ
 
a ν a g a abc b c
= −Fµν D η − ∂µ η f η η . (A.63)
2

71
Then in the usual way the BRST charge is defined
Z
QBRST = d3 xJ0 , (A.64)
which itself is a nilpotent

Q2BRST = 0. (A.65)
The Fock space of the theory is greatly increased due to the presence of the ghost and anti-
ghost fields. The ghost and anti-ghost stated are unphysical and have to be removed from
the theory. This can be done in a similar way to the Gupta-Bleuler condition in QED by

QBRST |Ψi = 0. (A.66)


States that satisfy the above are the physical states.

A.5 The Gribov Ambiguity


One has to be very careful if choosing a gauge condition completely removes the gauge de-
grees of freedom. In fact for the Coulomb gauge condition ∇i Aai = 0 does not remove the
gauge degree of freedom. This leads one to question the existence of the Yang-Mills propa-
gator ∆ab (x − y; A) and the over counting problem stated earlier.

Given ∇i Aai = 0 it is always possible to find a A0i which is gauge equivalent to Ai (here we
suppress gauge indices) that also satisfies

∇i A0i = 0. (A.67)
This can easily be seen if one explicitly preforms the gauge transformation
i
A0i = ΛAi Λ−1 − (∂i Λ)Λ−1 . (A.68)
g
Notice that the second term is proportional to g −1 . This means that perturbation theory
will never pick up this factor. That is, provided we stay within perturbation theory the
Coulomb gauge will in effect remove the gauge degrees of freedom. Picking SU (2) as a
concrete example, and take Ai to be gauge equivalent to 0.
i
Ai = − (∂i Λ)Λ−1 , (A.69)
g
where we have chosen the field such that ∇i Ai = 0. We parameterise the gauge using radial
coordinates

Λ = cos(ω(r)/2) + iσ i ni sin(ω(r)/2). (A.70)


Where ni ni = 1 and ni = xi /r. Substituting this into the Coulomb condition we obtain

d2 ω dω
+ − sin(2ω) = 0. (A.71)
dt2 dt

72
Where t = log(r) This is simply the equation for a damped harmonic oscillator in a constant
gravitational field. The solution ω = 0 leads to Ai = 0, which is the solution stated earlier.
However, it is clear that there are many other solutions to the equation with ω 6= 0 and
hence the Coulomb gauge condition is not unique. For non singular solutions all we need is
ω = 0, 2π, 4π, ... when t = −∞. For non-trivial solutions at t = ∞ we have the asymptotic
condition

σ i xi
Λ → ±i . (A.72)
r
As the Coulomb gauge does not fix the gauge completely there is a infinite sequence of fields
which are gauge equivalent that satisfy the Coulomb condition. These copies are known as
Gribov copies [39].

This ambiguity can also be explained by the fact that the matrix M ab (x, y; A) as defined in
equation (??) possess zero modes. Suppose we have non-zero eigenvalues λn

M ab ψnb (x) = λn ψna (x). (A.73)


However as stated earlier in general we also have the zero modes

M ab ϕbm (x) = 0. (A.74)


Now consider the Faddeev-Popov determinant.
Y Y
det M ab = λn 0 = 0. (A.75)
n m

As this determinant vanishes the matrix is non-invertible, thus the formulas stated earlier are
incorrect for the Coulomb gauge. As such the zero modes spoil the possibility of quantising
the system canonically, at least technically. However as long as one stays within perturbation
theory these zero modes will not make any contribution to the quantisation procedure. This
is due to the fact that the zero modes do not couple to the physical states.

Feynman rules are by their very nature pertubative. Perturbation theory is really just the
analytic part of the generating functional as g → 0, thus it is restricted to the first Gribov
copy as all other copies have a factor of 1/g associated with them.

73
APPENDIX B
Basics of Lie Algebra Cohomology

Here we present the basics of Lie algebra cohomology needed to understand the construction
of the BRST operator and it’s cohomology. We follow the conventions of [29], [28] and also
recommend [46] for a mathematical review of Lie Algebra cohomology. For a overview of Lie
algebras from a geometry point of view see [53].

B.1 Lie Algebras


Let a and g be elements of a Lie group G. The left-translation La : G → G of g by a are
defined as

La g = ag. (B.1)
The left-translation La is defined in a similar way. As La is a diffeomorphism from G to G.
This induces a diffeomorphism La? : Tg G → Tag G.

On a Lie group there exists a special class of vector fields defined by their invariance under
group translations. Let X be a vector field a vector field on a Lie group G. X is a left-
invariant vector field if La? X|ag = X|ag .

A vector V ∈ Te G defines a unique left-invariant vector field XV everywhere defined on G


via

XV |g = Lg? = Lg? V, (B.2)


with g ∈ G. Furthermore, a left-invariant vector field X defines a unique vector V = X|e ∈
Te G. The set of left-invariant vector fields on G is denoted by g.

Let X = X µ ∂/∂xµ and Y = Y µ ∂/∂xµ be vector fields on G. The Lie bracket [X, Y ] is
defined as

[X, Y ]f = X[Y [f ]] − Y [X[f ]], (B.3)

74
where f ∈ F (G) and F (G) is the space of smooth functions on G. It is straight forward to
see that [X, Y ] is another vector field. The Lie bracket satisfies the following

1. Bilinearity
[X, c1 Y1 + c2 Y2 ] = c1 [X, Y1 ] + c2 [X, Y2 ] (B.4)

[c1 X1 + c2 X2 , Y ] = c1 [X1 , Y ] + c2 [X2 , Y ]. (B.5)

For any constants c1 and c2 .

2. Skew-symmetry

[X, Y ] = −[Y, X]. (B.6)

3. The Jacobi identity

[[X, Y ], Z] + [[Z, X], Y ] + [[Y, Z], X] = 0. (B.7)

If one takes two points g and ag = Lag in G and apply La? to the Lie bracket [X, Y ] of
X, Y ∈ g, we see that

La? [X, Y ]|g = [La? X|g , La? Y |g ] = [X, Y ]|ag , (B.8)


and hence [X, Y ] ∈ g and thus g is closed under the Lie bracket.

The set of left-invariant vector fields g with the Lie bracket [, ] : g × g → g is the Lie algebra
of the Lie group G.

B.2 Lie Algebra Cohomology


Let V be a vector space. A V -valued n-dimensional cochain Ωn on g is a skew-symmetric
n-linear mapping

Ωn : g ∧ ..n . ∧ g → V, (B.9)
1 A
ΩA
n = Ωi1 ...in ω i1 ∧ ... ∧ ω in , (B.10)
n!
{ω (i) } is a basis of g and the upper index A labels the components in V . We denote the
abelian group of all n-cochains as C n (g, V ).

Thinking of V as a left ρ(g)-module, where ρ is a representation of the Lie algebra g,


ρ(Xi )A C A C A n
C ρ(Xj )B − ρ(Xj )C ρ(Xi )B = ρ([Xi , Xj ])B . The coboundary operator δ : C (g, V ) →
n+1
C (g, V ) is defined as

75
n+1
X
A
(δΩn ) (X1 , ..., Xn+1 ) = (−1)i+1 ρ(X1 )A B
B (Ωn (X1 , ..., X̂i , ..., Xn+1 )) (B.11)
i=1
n+1
X
+ (−1)j+k ΩA
n ([Xj , Xk ], X1 , ..., X̂j , ..., X̂k , ...Xn+1 ).
j,k=1 j<k

From the definition of δ it is straight forward to see that it is indeed a nilpotent, δ 2 = 0.

A n-cochain is a cocycle when δΩn = 0. The group of all n-cocycles is denoted Zρn (g, V ).
If a cocycle can be written as Ωn = δΩ0n−1 then Ωn is a coboundary. The group of all
n-coboundaries is denoted Bρn (g, V ). The n-th Lie algebra cohomology group is defined as

Hρn (g, V ) = Zρn (g, V )/Bρn (g, V ). (B.12)

B.3 Chevalley-Eilenberg Cohomology


Let V be R and the representation ρ be trivial. In equation (B.11)the first term is not
present and hence on left-invariant one-forms δ and d (the exterior derivative) act the same.
As there is a one-to-one mapping between n-antisymmetric maps on g and left-invariant
n-forms on G a cochain Ωn ∈ C n (g, R) can be written as a left-invariant n-form on G
1
Ωi ...i ω (i1 ) (g) ∧ ... ∧ ω in (g),
Ωn (g) = (B.13)
n! 1 n
where g ∈ G and the cobounary operator is now the exterior derivative d.

However, it should be noticed that the Chevalley-Eilenberg Lie algebra cohomology is in


general different to the de Rham cohomology. For instance, a differential form α on G may
be exact, α = dβ, but β may not be a left-invariant form and hence not a cochain.

B.4 Lie Algebra Cohomology and BRST Cohomology


Let η i be anticommuting fields, known in physics a ghosts:

{η i , η j } = 0, (B.14)
where i, j = 1, ..., dim g. The BRST operator s is defined by
1 ∂
s = η i ρ(Xi ) + fiji η i η j k . (B.15)
2 ∂η
From this definition it is straight forward to show that it is indeed a nilpotent.

In the trivial representation the BRST operator becomes


1 ∂
s = fijk η i η j k . (B.16)
2 ∂η

76
Here the BRST operator s acts on the ghosts as the exterior derivative d acts on the left
invariant one-forms.

The BRST cochains are given by


1 A
ΩA
n = Ω η i1 η in . (B.17)
n! i1 ...in
Notice that the action of s (in a given representation ρ) is the same as δ and hence the BRST
cohomology is identical the the standard Lie algebra cohomology.

77
APPENDIX C
The Atiyah-Singer Index Theorem and some
Characteristic Classes

In this appendix we outline the Atiyah-Singer index theorem for elliptical complexes on com-
pact manifolds without boundary. We follow [53] and also recommend [55].

A differential operator, D on a manifold M is a map between sections of vector bundles over


the manifold

D : Γ(M, E) → Γ(M, F ), (C.1)


where E and F are some vector bundles. If inner products are defined on E and F , as is
usually the case in physics the adjoint of D can be defined,

D† : Γ(M, F ) → Γ(M, E). (C.2)


The kernel of D and D† are defined as

ker D ≡ {s ∈ Γ(M, E)|Ds = 0}


ker D ≡ {s ∈ Γ(M, F )|D† s = 0}. (C.3)

The analytical index is defined as

ind D = dim kerD − dim kerD† . (C.4)


The Atiyah-Singer index theorem states that the analytical index is a topological invariant
and can be expressed in terms of integrals over characteristic classes. Thus the index of an
operator completely determined by the topology of the manifold.

78
C.1 Elliptic Complexes
Consider a sequence of elliptic operators,

Di−1 Di Di+1 -
··· - Γ(M, Ei−1 ) - Γ(M, Ei ) - Γ(M, Ei+1 ) ··· (C.5)
Here {Ei } is a sequence of vector bundles over M . The sequence (Ei , Di ) is called an elliptic
complex if Di is nilpotent.

C.2 Statement of The Atiyah-Singer Index Theorem


Let (E, D) be an elliptic complex over an m-dimensional compact manifold without bound-
ary. The topological index of this complex is given by
Z
T d(T M C )
ind(E, D) = (−1)m(m+1)/2 ch (⊕r (−1)r Er ) (C.6)
M e(T M ) vol
The Index theorem states that the analytical and topological index are the same.

If we have a two term complex then index theorem simplifies slightly. Let

D-
Γ(M, E) Γ(M, F ) (C.7)
be a two-term elliptic complex. The index of D is given by
Z
T d(T M C )
ind D = (−1)m(m+1)/2 (chE − chF ) . (C.8)
M e(T M ) vol
For the exact definitions of the various characteristic classes stated in (C.6) and (C.8) see
[53]

C.3 Some Characteristic Classes


In applying the Atiyah-Singer index theorem to the spin, Rarita-Schwinger complexes we
encountered the Â-genus, the total Chern character and the Hirzebruch L-polynomial. Here
we present some basic facts about these topological objects.

Consider a principle fibre bundle P (M, G) over an even dimensional compact manifold M
of dimension 2n without boundary. The group G is taken to be a Lie group. Let E be the
associated vector bundle with dim E = k. Let the curvature be denoted by F .
The total Chern character is defined by
   j
iF X 1 iF
ch(F ) ≡ Tr exp = Tr . (C.9)
2π j=1 j! 2π
The jth Chern character chj (F ) is
 j
1 iF
chj (F ) = Tr . (C.10)
j! 2π

79
The Â-genus is defined as
n
Y xj /2
Â(R) = . (C.11)
j=1 sinh(xj /2)

Where xj is defined via the Pontrjagin class


[k/2]
F X
p(F ) = det(I + )= (1 + x2i ) = 1 + p1 (F ) + p2 (F ) + · · · , (C.12)
2π i=1

Here
(
k/2 if k is even;
[k/2] = (C.13)
(k − 1)/2 if k is odd.
The xj are the eigenvalues of F/2π once it has been diagonalised.

The Â(R)-genus and the Chern character ch(F ) can be expanded in term of traces of cur-
vatures as

 
1 1 1 1 1
Â(R) = 1 + 2
TrR2 + 4
(TrR2 )2 + TrR4
(4π) 12 (4π) 288 360
 
1 1 2 3 1 2 4 1 6
+ (TrR ) + TrR R + TrR + ··· (C.14)
(4π)6 10368 4320 5670
 2
i 1 i
ch(F ) = k + TrF + + ···. (C.15)
2π 2 2π
The Hirzebruch L-polynomial is defined as
n
Y xj /2
L(M ) = . (C.16)
j=1 tanh(xj /2)

This can also be expanded in terms of traces

 
1 1 1 1 7
L(M ) = 1 − 2
TrR2 + 4
(TrR2 )2 − TrR4
(2π) 6 (2π) 72 180
 
1 1 2 3 7 2 4 31 6
+ − (TrR ) + TrR TrR − TrR + ···. (C.17)
(2π)6 1296 1080 2835

80
BIBLIOGRAPHY

[1] Stephen L. Adler. Axial vector vertex in spinor electrodynamics. Phys. Rev., 177:2426–
2438, 1969.

[2] Stephen L. Adler and William A. Bardeen. Absence of higher order corrections in the
anomalous axial vector divergence equation. Phys. Rev., 182:1517–1536, 1969.

[3] Luis Alvarez-Gaume. An introduction to gravitational anomalies. In *Cargese 1983,


Proceedings, Progress In Gauge Field Theory*, 1-22.

[4] Luis Alvarez-Gaume and Paul H. Ginsparg. The topological meaning of nonabelian
anomalies. Nucl. Phys., B243:449, 1984.

[5] Luis Alvarez-Gaume and Paul H. Ginsparg. The structure of gauge and gravitational
anomalies. Ann. Phys., 161:423, 1985.

[6] Luis Alvarez-Gaume and Edward Witten. Gravitational anomalies. Nucl. Phys.,
B234:269, 1984.

[7] M. F. Atiyah and I. M. Singer. The index of elliptic operators. 1. Annals Math.,
87:484–530, 1968.

[8] M. F. Atiyah and I. M. Singer. The index of elliptic operators. 3. Annals Math.,
87:546–604, 1968.

[9] M. F. Atiyah and I. M. Singer. The index of elliptic operators. 4. Annals Math.,
93:119–138, 1971.

[10] M. F. Atiyah and I. M. Singer. The index of elliptic operators. 5. Annals Math.,
93:139–149, 1971.

[11] M. F. Atiyah and I. M. Singer. Dirac operators coupled to vector potentials. Proc. Nat.
Acad. Sci., 81:2597–2600, 1984.

[12] J. L. F. Barbon. Introduction to noncommutative field theory. Prepared for ICTP


Summer School in Particle Physics, Trieste, Italy, 18 Jun - 6 Jul 2001.

81
[13] J. Barcelos-Neto. Axial and gauge anomalies in the bv formalism. Int. J. Mod. Phys.,
A12:5053–5066, 1997.

[14] William A. Bardeen. Anomalous ward identities in spinor field theories. Phys. Rev.,
184:1848–1857, 1969.

[15] Glenn Barnich, Friedemann Brandt, and Marc Henneaux. Local brst cohomology in
gauge theories. Phys. Rept., 338:439–569, 2000.

[16] I. A. Batalin and G. A. Vilkovisky. Gauge algebra and quantization. Phys. Lett.,
B102:27–31, 1981.

[17] I. A. Batalin and G. a. Vilkovisky. Feynman rules for reducible gauge theories. Phys.
Lett., B120:166–170, 1983.

[18] C. Becchi, A. Rouet, and R. Stora. Renormalization of gauge theories. Annals Phys.,
98:287–321, 1976.

[19] J. S. Bell and R. Jackiw. A pcac puzzle: pi0 → gamma gamma in the sigma model.
Nuovo Cim., A60:47–61, 1969.

[20] R. A. Bertlmann. Anomalies in quantum field theory. Oxford, UK: Clarendon (1996)
566 p. (International series of monographs on physics: 91).

[21] Adel Bilal. Anomaly cancellations on lower-dimensional hypersurfaces by inflow from


the bulk. 2004.

[22] Adel Bilal and Steffen Metzger. Anomaly cancellation in m-theory: A critical review.
Nucl. Phys., B675:416–446, 2003.

[23] M. Blagojevic. Gravitation and Gauge Symmetries. Bristol, UK: IOP (2002) 522 p.

[24] L. Bonora and P. Cotta-Ramusino. Some remarks on brs transformations, anomalies


and the cohomology of the lie algebra of the group of gauge transformations. Commun.
Math. Phys., 87:589, 1983.

[25] L. Bonora, M. Schnabl, and Alessandro Tomasiello. A note on consistent anomalies in


noncommutative ym theories. Phys. Lett., B485:311–313, 2000.

[26] Jr. Callan, Curtis G. and Jeffrey A. Harvey. Anomalies and fermion zero modes on
strings and domain walls. Nucl. Phys., B250:427, 1985.

[27] A. Connes. Noncommutative geometry. Accademic Press, page 661, 1994.

[28] J. A. de Azcarraga, J. M. Izquierdo, and J. C. Perez Bueno. An introduction to some


novel applications of lie algebra cohomology in mathematics and physics. Rev. R. Acad.
Cien. Exactas Fis. Nat. Ser. A Mat., 95:225–248, 2001.

[29] Jose A. de Azcarraga and Jose M. Izquierdo. Lie groups, lie algebras, cohomology and
some applications in physics.

[30] Antonio Dobado and Antonio L. Maroto. Standard model anomalies in curved space-
time with torsion. Phys. Rev., D54:5185–5194, 1996.

82
[31] L. D. Faddeev and V. N. Popov. Feynman diagrams for the yang-mills field. Phys. Lett.,
B25:29–30, 1967.

[32] L. D. Faddeev and Samson L. Shatashvili. Realization of the schwinger term in the
gauss law and the possibility of correct quantization of a theory with anomalies. Phys.
Lett., B167:225–228, 1986.

[33] Dan Freed, Jeffrey A. Harvey, Ruben Minasian, and Gregory W. Moore. Gravitational
anomaly cancellation for m-theory fivebranes. Adv. Theor. Math. Phys., 2:601–618,
1998.

[34] K. Fujikawa and H. Suzuki. Path integrals and quantum anomalies. Oxford, UK:
Clarendon (2004) 284 p.

[35] Kazuo Fujikawa. Anomalous ward identities and path integration. Presented at 20th
Int. Conf. on High Energy Physics, Madison, Wis., Jul 17-23, 1980.

[36] Kazuo Fujikawa. Path integral measure for gauge invariant fermion theories. Phys. Rev.
Lett., 42:1195, 1979.

[37] J. Gomis and J. Paris. Anomalous gauge theories within bv framework. Prepared for
International Europhysics Conference on High- energy Physics, Marseille, France, 22-28
Jul 1993.

[38] Michael B. Green and John H. Schwarz. Anomaly cancellation in supersymmetric d=10
gauge theory and superstring theory. Phys. Lett., B149:117–122, 1984.

[39] V. N. Gribov. Quantization of non-abelian gauge theories. Nucl. Phys., B139:1, 1978.

[40] David J. Gross and R. Jackiw. Effect of anomalies on quasirenormalizable theories.


Phys. Rev., D6:477–493, 1972.

[41] Vaisman I. Cohomology and differential forms. M. Dekker Inc.(1973).

[42] R. Jackiw. Field theoretic investigations in current algebra. In *Treiman, S.b. ( Ed.)
Et Al.: Current Algebra and Anomalies*, 81-210.

[43] R. Jackiw. Magnetic sources and three cocycles (comment). Phys. Lett., B154:303–304,
1985.

[44] M. Kaku. Introduction to superstrings. NEW YORK, USA: SPRINGER (1988) 568 P.
(GRADUATE TEXTES IN CONTEMPORARY PHYSICS).

[45] M. Kaku. Quantum Field Theory: A Modern Introduction. New York, USA: Oxford
Univ. Pr. (1993) 785 p.

[46] Anthony W. Knapp. Lie groups, lie algebras and cohomology.

[47] Edwin Langmann. Descent equations of yang-mills anomalies in noncommutative ge-


ometry. J. Geom. Phys., 22:259–279, 1997.

[48] J. Madore. An introduction to noncommutative differential geometry and physical


applications. Lond. Math. Soc. Lect. Note Ser., 257:1–371, 2000.

83
[49] Juan Manes, Raymond Stora, and Bruno Zumino. Algebraic study of chiral anomalies.
Commun. Math. Phys., 102:157, 1985.

[50] C. P. Martin. Chiral gauge anomalies on noncommutative minkowski space- time. Mod.
Phys. Lett., A16:311–320, 2001.

[51] O. Moritsch, M. Schweda, and T. Sommer. Yang-mills gauge anomalies in the presence
of gravity with torsion. Class. Quant. Grav., 12:2059–2070, 1995.

[52] Otmar Moritsch, Manfred Schweda, and Silvio P. Sorella. Algebraic structure of gravity
with torsion. Class. Quant. Grav., 11:1225–1242, 1994.

[53] M. Nakahara. Geometry, Topology and Physics. Bristol, UK: Hilger (1990) 505 p.
(Graduate student series in physics).

[54] C. Nash. Differential topology and quantum field theory. London, UK: Academic (1991)
386 p.

[55] C. Nash and S. Sen. Topology and Geometry for Physicists. Academic Press, New
York.

[56] Kazuhiko Nishijima and Masanori Okawa. The becchi-rouet-stora transformation for
the gravitational field. Prog. Theor. Phys., 60:272, 1978.

[57] W. Pauli and F. Villars. On the invariant regularization in relativistic quantum theory.
Rev. Mod. Phys., 21:434–444, 1949.

[58] O. Piguet and S. P. Sorella. Algebraic renormalization: Perturbative renormalization,


symmetries and anomalies. Lect. Notes Phys., M28:1–134, 1995.

[59] Mosher R. and Tangora M. Cohomology operations and applications in homotopy


theory. Harper and Row (1968). (Harper’s Series in Modern Mathematics).

[60] Lewis H. Ryder. Quantum Field Theory. Cambridge University Press UK, (1996)
487p.

[61] R. Schmid. A few brst bicomplexes. In *Tianjin 1992, Proceedings, Differential geo-
metric methods in theoretical physics* 375-378.

[62] R. Schmid. Local cohomology in gauge theories, brst transformations and anomalies.
Differ. Geom. Appl., 4:107, 1994.

[63] A. A. Slavnov. Ward identities in gauge theories. Theor. Math. Phys., 10:99–107, 1972.

[64] Silvio P. Sorella. Algebraic characterization of the wess-zumino consistency conditions


in gauge theories. Commun. Math. Phys., 157:231–243, 1993.

[65] Jack Steinberger. Phys. Rev, 76:1180.

[66] Gerard ’t Hooft and M. J. G. Veltman. Regularization and renormalization of gauge


fields. Nucl. Phys., B44:189–213, 1972.

[67] Y. Takahashi. On the generalized ward identity. Nuovo Cim., 6:371, 1957.

84
[68] J. C. Taylor. Ward identities and charge renormalization of the yang- mills field. Nucl.
Phys., B33:436–444, 1971.

[69] I. V. Tyutin. Gauge invariance in field theory and statistical physics in operator for-
malism. LEBEDEV-75-39.

[70] J. C. Ward. An identity in quantum electrodynamics. Phys. Rev., 78:182, 1950.

[71] Satoshi Watamura. The brs transformation and the consistent gravitational anomalies
in poincare gravity. RIFP-565.

[72] N. E. Wegge-Olsen. K-theory and c*-algebras a friendly approach. Oxford University


Press: (1993) 336 p.

[73] M. Werneck de Oliveira and S. P. Sorella. Algebraic structure of lorentz and diffeomor-
phism anomalies. Int. J. Mod. Phys., A9:2979–2996, 1994.

[74] J. Wess and B. Zumino. Consequences of anomalous ward identities. Phys. Lett., B37:95,
1971.

[75] Satoshi Yajima. Mixed anomalies in four-dimensional and six-dimensional space with
torsion. Prog. Theor. Phys., 79:535, 1988.

[76] Chen-Ning Yang and R. L. Mills. Conservation of isotopic spin and isotopic gauge
invariance. Phys. Rev., 96:191–195, 1954.

[77] Bruno Zumino, Yong-Shi Wu, and Anthony Zee. Chiral anomalies, higher dimensions,
and differential geometry. Nucl. Phys., B239:477–507, 1984.

85

Você também pode gostar