Peskin Problems

A Complete Solution to Problems in
An Introduction to Quantum Field Theory

by Peskin and Schroeder
Zhong-Zhi Xianyu
Harvard University
May 2016
ii
Preface
In this note I provide solutions to all problems and final projects in the book An Intro-
duction to Quantum Field Theory by M. E. Peskin and D. V. Schroeder [1], which I worked
out and typed into TEX during the first two years of my PhD study at Tsinghua University.
I once posted a draft version of them on my personal webpage using a server provided by
Tsinghua, which was however closed unfortunately after I graduated. Since then I received
quite a number of emails asking for the solutions, so I decided to put them on arXiv.
Nothing much has been updated in this note compared with the previous draft due to
the lack of time, except for some editorial work, as well as a few newly added references.
In particular, I dont have enough time to proofread and therefore I cannot guarantee the
correctness of them, though I expect that most of them are correct. With that said, any
feedback via email about errors, either physical or typographical, is much appreciated.
I would not claim any novelty or originality of this note, since almost all of problems in
the book belong to standard material of quantum field theory. Occasionally, I learned the
answer to a problem or the strategy for solving it before I started to work it out. But still,
I believe that the problem set in the book will always remain a treasure to any beginner of
this subject, and I feel it worthy to write up the solutions.
The contraction macro provided by the authors of the book has been used in this note.
I would like to express my gratitude to Prof. Qing Wang and Prof. Hong-Jian He for
their wonderful courses of quantum field theory and their great help in my early days of
learning this subject. I would also like to thank Prof. Michael Peskin in particular, for his
generous permission and kind encouragement to letting me publish this note.
Comments on notations. All notations and conventions are the same with the book.
The book will be cited in the main text as P&S for short. The +i prescription for
Feynman propagators is always assumed and is usually hidden.

The submission, however, was rejected by one of arXiv volunteer moderators based on the reason that
arXiv does not allow submissions containing solutions to problems in physics textbooks, and that (the)
moderators consider that this type of submissions are harmful for students and instructors. Insofar as I
can see, however, the solution can only do harm to those who are willing to do harm to themselves.

xianyuzhongzhi@gmail.com

http://physics.weber.edu/schroeder/qftbook.html
iii
iv Preface
Contents
Preface iii
2 The Klein-Gordon Field 1

2.1 Classical electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2.2 The complex scalar field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.3 The spacelike correlation function . . . . . . . . . . . . . . . . . . . . . . . . 6
3 The Dirac Field 7

3.1 Lorentz group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 The Gordon identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 The spinor products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Majorana fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.5 Supersymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.6 Fierz transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.7 The discrete symmetries P , C and T . . . . . . . . . . . . . . . . . . . . . . 19
3.8 Bound states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Interacting Fields and Feynman Diagrams 23

4.1 Scalar field with a classical source . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Decay of a scalar particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Linear sigma model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4 Rutherford scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5 Elementary Processes of Quantum Electrodynamics 33

5.1 Coulomb scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Bhabha scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.3 The spinor products (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.4 Positronium lifetimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.5 Physics of a massive vector boson . . . . . . . . . . . . . . . . . . . . . . . . 45
5.6 The spinor products (3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6 Radiative Corrections: Introduction 51

6.1 Rosenbluth formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
v
vi CONTENTS
6.2 Equivalent photon approximation . . . . . . . . . . . . . . . . . . . . . . . . 53

6.3 Exotic contributions to g 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7 Radiative Corrections: Some Formal Developments 59

7.1 Optical theorem in 4 theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
7.2 Alternative regulators in QED . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.3 Radiative corrections in QED with Yukawa interaction . . . . . . . . . . . . 63
Final Project I. Radiation of Gluon Jets 67
9 Functional Methods 73
9.1 Scalar QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9.2 Statistical field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
10 Systematics of Renormalization 79
10.1 One-Loop structure of QED . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
10.2 Renormalization of Yukawa theory . . . . . . . . . . . . . . . . . . . . . . . 80
10.3 Field-strength renormalization in 4 theory . . . . . . . . . . . . . . . . . . . 82
10.4 Asymptotic behavior of diagrams in 4 theory . . . . . . . . . . . . . . . . . 83
11 Renormalization and Symmetry 87

11.1 Spin-wave theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
11.2 A zeroth-order natural relation . . . . . . . . . . . . . . . . . . . . . . . . . 88
11.3 The Gross-Neveu model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
12 The Renormalization Group 93

12.1 Beta Function in Yukawa Theory . . . . . . . . . . . . . . . . . . . . . . . . 93
12.2 Beta Function of the Gross-Neveu Model . . . . . . . . . . . . . . . . . . . . 93
12.3 Asymptotic Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
13 Critical Exponents and Scalar Field Theory 99

13.1 Correlation-to-scaling exponent . . . . . . . . . . . . . . . . . . . . . . . . . 99
13.2 The exponent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
13.3 The CP N model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Final Project II. The Coleman-Weinberg Potential 103
15 Non-Abelian Gauge Invariance 113

15.1 Brute-force computations in SU (3) . . . . . . . . . . . . . . . . . . . . . . . 113
15.2 Adjoint representation of SU (2) . . . . . . . . . . . . . . . . . . . . . . . . . 113
15.3 Coulomb potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
15.4 Scalar propagator in a gauge theory . . . . . . . . . . . . . . . . . . . . . . . 115
15.5 Casimir operator computations . . . . . . . . . . . . . . . . . . . . . . . . . 117
CONTENTS vii
16 Quantization of Non-Abelian Gauge Theories 121

16.1 Arnowitt-Fickler gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
16.2 Scalar field with non-Abelian charge . . . . . . . . . . . . . . . . . . . . . . 122
16.3 Counterterm relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
17 Quantum Chromodynamics 133

17.1 Two-Loop renormalization group relations . . . . . . . . . . . . . . . . . . . 133
17.2 A Direct test of the spin of the gluon . . . . . . . . . . . . . . . . . . . . . . 134
17.3 Quark-gluon and gluon-gluon scattering . . . . . . . . . . . . . . . . . . . . . 135
17.4 The gluon splitting function . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
17.5 Photoproduction of heavy quarks . . . . . . . . . . . . . . . . . . . . . . . . 141
17.6 Behavior of parton distribution functions at small x . . . . . . . . . . . . . . 142
18 Operator Products and Effective Vertices 145

18.1 Matrix element for proton decay . . . . . . . . . . . . . . . . . . . . . . . . . 145
18.2 Parity-violating deep inelastic form factor . . . . . . . . . . . . . . . . . . . 147
18.3 Anomalous dimensions of gluon twist-2 operators . . . . . . . . . . . . . . . 151
18.4 Deep inelastic scattering from a photon . . . . . . . . . . . . . . . . . . . . . 156
19 Perturbation Theory Anomalies 159

19.1 Fermion number nonconservation in parallel E and B fields . . . . . . . . . . 159
19.2 Weak decay of the pion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
19.3 Computation of anomaly coefficients . . . . . . . . . . . . . . . . . . . . . . 162
19.4 Large fermion mass limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
20 Gauge Theories with Spontaneous Symmetry Breaking 167

20.1 Spontaneous breaking of SU (5) . . . . . . . . . . . . . . . . . . . . . . . . . 167
20.2 Decay modes of the W and Z bosons . . . . . . . . . . . . . . . . . . . . . . 168
20.3 e+ e hadrons with photon-Z 0 interference . . . . . . . . . . . . . . . . . . 170
20.4 Neutral-current deep inelastic scattering . . . . . . . . . . . . . . . . . . . . 173
20.5 A model with two Higgs fields . . . . . . . . . . . . . . . . . . . . . . . . . . 174
21 Quantization of Spontaneously Broken Gauge Theories 177

21.1 Weak-interaction contributions to the muon g 2 . . . . . . . . . . . . . . . 177
21.2 Complete analysis of e+ e W + W . . . . . . . . . . . . . . . . . . . . . . 180
21.3 Cross section for du W . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
21.4 Dependence of radiative corrections on the Higgs boson mass . . . . . . . . . 185
Final Project III. Decays of the Higgs Boson 189
Bibliography 201
viii CONTENTS
Chapter 2
The Klein-Gordon Field
2.1 Classical electromagnetism

In this problem we derive the field equations and energy-momentum tensor from the
following action of classical electrodynamics,
Z
1
S= d4 x F F , with F = A A . (2.1)
4
(a) Maxwells equations To take variation of the classical action with respect to the
field A , we note,
F F
= , = 0. (2.2)
( A ) A
Then from the first equality we get:

F F = 4F .

(2.3)
( A )
Now substitute this into Euler-Lagrange equation, we have,

L L
0 = = F . (2.4)
( A ) A
This is sometimes called the second pair of Maxwells equations. The so-called first pair
follows directly from the definition of F = A A , and reads
F + F + F = 0. (2.5)
The familiar electric and magnetic field strengths can be written as E i = F 0i and ijk B k =
F ij , respectively. From this we deduce the Maxwells equations in terms of E i and B i :
i E i = 0, ijk j B k 0 E i = 0, ijk j E k = 0, i B i = 0. (2.6)
1
2 Chapter 2. The Klein-Gordon Field
(b) The energy-momentum tensor The energy-momentum tensor can be defined to be

the Nother current of the space-time translational symmetry. Under space-time translation
the vector A transforms as,
A = A . (2.7)
Thus
L 1
T = A L = F A + F F . (2.8)
( A ) 4
Obviously, this tensor is not symmetric. We can add an additional term K to T with
K antisymmetric with its first two indices. Its easy to see that this term does not affect
the conservation of T . So if we choose K = F A , then,
1
T = T + K = F F + F F . (2.9)
4
Now this tensor is symmetric and is sometimes called the Belinfante tensor in literature.
We can also rewrite it in terms of E i and B i ,
1 i i
T 00 = (E E + B i B i ), T i0 = T 0i = ijk E j B k , etc. (2.10)
2
2.2 The complex scalar field

The Lagrangian is given by,
L = m2 . (2.11)
(a) The conjugate momenta of and :

L L
= = , = = = . (2.12)

The canonical commutation relations:
[(x), (y)] = [ (x), (y)] = i(x y), (2.13)
The rest of commutators are all zero.

The Hamiltonian:
Z Z
H = d x + L = d3 x + + m2 .
3

(2.14)
(b) Now we Fourier transform the field as:

d3 p
Z
1 ipx ipx

(x) = p a p e + b p e , (2.15)
(2)3 2Ep
thus:
d3 p
Z
1 ipx ipx

(x) = p b p e + a p e . (2.16)
(2)3 2Ep
2.2. The complex scalar field 3
Substitute the mode expansion into the Hamiltonian:

Z
H = d3 x + + m2

d3 p d3 q
Z Z
3
= dx p p
(2)3 2Ep (2)3 2Eq

Ep Eq ap eipx bp eipx aq eiqx bq eiqx

+ p q ap eipx bp eipx aq eiqx bq eiqx

2 ipx ipx iqx iqx
+ m ap e + b p e aq e + bq e
d3 p d3 q
Z Z
3
= dx p p
(2)3 2Ep (2)3 2Eq

2 i(pq)x i(pq)x
(Ep Eq + p q + m ) ap aq e + bp bq e

i(p+q)x
2
(Ep Eq + p q m ) bq aq e + ap bq ei(p+q)x
d3 p d3 q
Z
= p p
(2)3 2Ep (2)3 2Eq

2 i(Ep Eq )t i(Ep Eq )t
(Ep Eq + p q + m ) ap aq e + bp bq e (2)3 (3) (p q)

i(Ep +Eq )t
2
(Ep Eq + p q m ) bq aq e + ap bq ei(Ep +Eq )t 3 (3)
(2) (p + q)
Ep2 + p2 + m2
Z
d3 x ap ap + bp bp

=
2Ep
Z
= d3 x Ep ap ap + bp bp + [bp , bp ] ,

(2.17)
p
where we have used the mass-shell condition Ep = m2 + p2 . Note that the last term
contributes an infinite constant, which can be interpreted as the vacuum energy and can
be dropped, for instance, by the prescription of normal ordering. Then we get a finite
Hamiltonian, Z
H = d3 x Ep ap ap + bp bp ,

(2.18)
Hence we get two sets of particles with the same mass m.
(c) The theory is invariant under the global transformation: ei , ei . The

corresponding conserved charge is:
Z
Q = i d3 x .

(2.19)
Rewrite this in terms of the creation and annihilation operators:

Z
Q = i d3 x

d3 p d3 q
Z Z
ipx

3 ipx iqx iqx
=i dx p p b p e + ap e a q e + b q e
(2)3 2Ep (2)3 2Eq t
ipx
ipx iqx iqx
bp e + ap e aq e + bq e
t
d3 p d3 q
Z Z
3 ipx ipx iqx iqx
= dx p p E q b p e + a p e a q e b q e
(2)3 2Ep (2)3 2Eq

ipx ipx iqx iqx
E p bp e ap e aq e + bq e
d3 p d3 q
Z Z
3 i(p+q)x i(p+q)x
= dx p p (E q E p ) b a
p q e a b
p q e
(2)3 2Ep (2)3 2Eq

i(pq)x i(pq)x
+ (Eq + Ep ) ap aq e bp bq e
d3 p d3 q
Z
= p p
(2)3 2Ep (2)3 2Eq

(Eq Ep ) bp aq ei(Ep +Eq )t ap bq ei(Ep +Eq t) (2)3 (3) (p + q)

i(Ep Eq )t i(Ep Eq )t 3 (3)
+ (Eq + Ep ) ap aq e bp bq e (2) (p q)
d3 p
Z
= 2Ep (ap ap bp bp )
(2)3 2Ep
d3 p
Z

= a p a p b p b p , (2.20)
(2)3
where the last equal sign holds up to an infinitely large constant term, as we did when
calculating the Hamiltonian in (b). Then the commutators follow straightforwardly:
[Q, a ] = a , [Q, b ] = b . (2.21)
We see that the particle a carries one unit of positive charge, and b carries one unit of
negative charge.
(d) Now we consider the case with two complex scalars of same mass. In this case the
Lagrangian is given by
L = i i m2 i i , (2.22)
where i with i = 1, 2 is a two-component complex scalar. Then it is straightforward to

see that the Lagrangian is invariant under the U (2) transformation i Uij j with Uij a
matrix in fundamental representation of U (2) group. The U (2) group, locally isomorphic to
SU (2) U (1), is generated by 4 independent generators 1 and 21 a , with a Pauli matrices.
Then 4 independent Nother currents are associated, which are given by,
L L
j =
i
i = ( i )(ii ) ( i )(ii ),
( i ) ( i )
2.2. The complex scalar field 5
L L a ih
i
ja = a
i i = ( i ) ij j ( i ) ij j . (2.23)
( i ) ( i ) 2
The overall sign is chosen such that the particle carry positive charge, as will be seen in the
following. Then the corresponding Nother charges are given by,
Z
Q = i d3 x i i i i ,

Z
i
a
d3 x i ( a )ij j i ( a )ij j .

Q = (2.24)
2
Repeating the derivations above, we can also rewrite these charges in terms of creation and
annihilation operators, as,
d3 p
Z

Q= a a ip b ip ip ,
b
(2)3 ip
d3 p a
Z
a 1 a

Q = a a ip b jp ij jp .
b (2.25)
2 (2)3 ip ij
The generalization to N -component complex scalar is straightforward. In this case we

only need to replace the generators a /2 of SU (2) group to the generators ta in the funda-
mental representation of SU (N ) group with commutation relation [ta , tb ] = if abc tc .
Then we are ready to calculate the commutators among all these Nother charges and
the Hamiltonian. Firstly we show that all charges of the U (N ) group commute with the
Hamiltonian. For the U (1) generator, we have
d3 p d3 q
Z h

i
[Q, H] = Eq aip aip bip bip , ajq ajq + bjq bjq
(2)3 (2)3
d3 p d3 q
Z

= Eq a ip [a ip , ajq ]a jq + a [a
jq ip , ajq ]a ip + (a b)
(2)3 (2)3
d3 p d3 q
Z

= Eq a a
ip iq a a
iq ip + (a b) (2)3 (3) (p q)
(2)3 (2)3
= 0. (2.26)
Similar calculation gives [Qa , H] = 0. Then we consider the commutation among internal
U (N ) charges:
d3 p d3 q h a
Z
a b a
b b
i
[Q , Q ] = aip tij ajp bip tij bjp , akq tk` a`q bkq tk` b`q
(2)3 (2)3
d3 p d3 q a b
Z
b a
= a t t a`q akq tk` t`j ajp + (a b) (2)3 (3) (p q)
(2)3 (2)3 ip ij j`
d3 p c
Z
abc c
= if a t a jp b t b jp
(2)3 ip ij ip ij
= if abc Qc , (2.27)
and similarly, [Q, Q] = [Qa , Q] = 0.

2.3 The spacelike correlation function

We evaluate the correlation function of a scalar field at two points,
D(x y) = h0|(x)(y)|0i, (2.28)
with x y being spacelike. Since any spacelike interval x y can be transformed to a form
such that x0 y 0 = 0, thus we will simply take:
x0 y 0 = 0, and |x y|2 = r2 > 0. (2.29)
Now:
d3 p 1 ip(xy) d3 p
Z Z
1
D(x y) = e = p eip(xy)
(2)3 2Ep (2)3 2 m2 + p2
Z 2 Z 1 Z
1 p2
= d d cos dp p eipr cos
(2)3 0 1 0
2
2 m +p 2
Z ipr
i pe
= 2
dp p (2.30)
2(2) r m2 + p2
Now we make the path deformation on p-complex plane, as is shown in Figure 2.3 of P&S.
Then the integral becomes,
Z
1 er m
D(x y) = 2 d p = 2 K1 (mr). (2.31)
4 r m 2 m2 4 r
Chapter 3
The Dirac Field
3.1 Lorentz group

The generators of Lorentz group satisfy the following commutation relation,
[J , J ] = i(g J g J g nu J + g J ). (3.1)
(a) Let us redefine the generators as Li = 12 ijk J jk (All Latin indices denote spatial
components), with Li generate rotations, and K i generate boosts. The commutators of
them can be derived straightforwardly to be,
[Li , Lj ] = iijk Lk , [K i , K j ] = iijk Lk . (3.2)
1
If we further define Ji = 2
(Li iK i ), then the commutators become,
[Ji , Jj ] = iijk Jk , [J+i , Jj ] = 0. (3.3)
Thus we see that the algebra of the Lorentz group is a direct sum of two identical algebra
su(2).
(b) It follows that we can classify the finite dimensional representations of the Lorentz
group by a pair (j+ , j ), where j = 0, 1/2, 1, 3/2, 2, are labels of irreducible representa-
tions of SU (2).
We study two specific cases.
1. ( 21 , 0). Following the definition, we have J+i represented by 1

2
i and Ji represented by
0. This implies
Li = (J+i + Ji ) = 1
2
i, K i = i(J+i Ji ) = 2i i . (3.4)
Hence a field under this representation transforms as:

i i i i
ei /2 /2 . (3.5)
7
8 Chapter 3. The Dirac Field
2. ( 12 , 0). In this case, J+i 0, Ji 1

2
i . Then
Li = (J+i + Ji ) = 1
2
i, K i = i(J+i Ji ) = 2i i . (3.6)
Hence a field under this representation transforms as:

i i i i
ei /2+ /2 . (3.7)
We see that a field under the representation ( 12 , 0) and (0, 12 ) are precisely the left-handed
spinor L and right-handed spinor R , respectively.
(c) Let us consider the case of ( 12 , 21 ). To put the field associated with this representation
into a familiar form, we note that a left-handed spinor can also be rewritten as row, which
transforms under the Lorentz transformation as:
LT 2 LT 2 1 + 2i i i + 1
ii .

2
(3.8)
Then the field under the representation ( 12 , 21 ) can be written as a tensor with spinor indices:
!
V 0 + V 3 V 1 iV 2
R LT 2
V = . (3.9)
V 1 + iV 2 V 0 V 3
In what follows we will prove that V is in fact a Lorentz vector.

A quantity V is called a Lorentz vector, if it satisfies the following transformation law:
V V , (3.10)
where = 2i (J ) in its infinitesimal form. We further note that:
(J ) = i( ). (3.11)
and also, ij = ijk k , 0i = i0 = i , then the combination V = V i i + V 0 transforms

according to
i i i
V i i ji mn (J mn )i j V j i + 0n (J 0n )i 0 n0 (J n0 )0 i V 0 i
2 2
2
= ji 2i mnk k (i)(im jn jm in ) V j i + i i (i)(in ) V 0 i

=V i i ijk V i j k + V 0 i i ,
i i
V 0 V 0 + 0n (J 0n )0i n0 (J n0 )0i V i
2 2
= V + i (ii ) V = V 0 + i V i .
0 i n
i
In total, we have
V i ijk j k + i V i + (1 + i i )V 0 .

(3.12)
3.2. The Gordon identity 9
If we can reach the same conclusion by treating the combination V a matrix transforming
under the representation ( 12 , 21 ), then our original statement will be proved. In fact:
i 1 i 1
V 1 j j + j j V 1 + j j + j j
2 2 2 2
i 1
= i + j [ i , j ] + j { i , j } V i + (1 + i i )V 0
2 2i
= + V + (1 + i i )V 0 ,
i ijk j k i
(3.13)
as expected. Hence we proved that V is a Lorentz vector.
3.2 The Gordon identity

In this problem we derive the Gordon identity,
p0 + p i (p0 p )
u(p0 ) u(p) = u(p0 ) + u(p). (3.14)
2m 2m
Let us start from the right hand side,
p0 + p i (p0 p )
0
u(p ) + u(p)
2m 2m
1 0

0 0

= u(p ) (p + p ) + i (p p ) u(p)
2m
1 0

0 1 0
= u(p ) (p + p ) [ , ](p p ) u(p)
2m 2
1 0
1
0 1 0
= u(p ) { , }(p + p ) [ , ](p p ) u(p)
2m 2 2
1 0

0

0
= u(p ) p/ + p/ u(p) = u(p ) u(p),
2m
where we have used the commutator and anti-commutators of gamma matrices, as well as
the Dirac equation.
3.3 The spinor products

In this problem, together with the Problems 5.3 and 5.6, we will develop a formalism
that can be used to calculating scattering amplitudes involving massless fermions or vector
particles. This method can profoundly simplify the calculations, especially in the calcula-
tions of QCD. Here we will derive the basic fact that the spinor products can be treated as
the square root of the inner product of lightlike Lorentz vectors. Then, in Problem 5.3 and
5.6, this relation will be put in use in calculating the amplitudes with external spinors and
external photons, respectively.
To begin with, let k0 and k1 be fixed four-vectors satisfying k02 = 0, k12 = 1 and
k0 k1 = 0. With these two reference momenta, we define the following spinors:
1. Let uL0 be left-handed spinor with momentum k0 ;

2. Let uR0 = k/1 uL0 ;
3. For any lightlike momentum p (p2 = 0), define:

1 1
uL (p) = p/uR0 , uR (p) = p/uL0 . (3.15)
2p k0 2p k
(a) We show that k/0 uR0 = 0 and p/uL (p) = p/uR (p) = 0 for any lightlike p. That is, uR0 is a
massless spinor with momentum k0 , and uL (p), uR (p) are massless spinors with momentum
p. This is quite straightforward,
k/0 uR0 = k/0 k/1 uL0 = (2g )k0 k1 uL0 = 2k0 k1 uL0 k/1 k/0 uL0 = 0, (3.16)
and, by definition,
1 1
p/uL (p) = p/p/uR0 = p2 uR0 = 0. (3.17)
2p k0 2p k0
In the same way, we can show that p/uR (p) = 0.
(b) Now we choose k0 = (E, 0, 0, E) and k1 = (0, 1, 0, 0). Then in the Weyl represen-
tation, we have:
0 0 0 0
0 0 0 2E
k/0 uL0 = 0 uL0 = 0. (3.18)

2E 0 0 0
0 0 0 0

Thus uL0 can be chosen to be (0, 2E, 0, 0)T , and:

0 0 0 1 0
0 0 1 0 0
uR0 = k/1 uL0 = uL0 = . (3.19)

0 1 0 0 2E
1 0 0 0 0
Let p = (p0 , p1 , p2 , p3 ), then:

1
uL (p) = p/uR0
2p k0

0 0 p0 + p3 p1 ip2
1 0 0 p1 + ip2 p0 p3
= p uR0

2E(p0 + p3 ) p0 p3 p1 + ip2

0 0
p1 ip2 p0 + p3 0 0

(p0 + p3 )
1 (p1 + ip2 )

= . (3.20)
p0 + p3

0
0
3.4. Majorana fermions 11
In the same way, we get:

0
1 0
uR (p) = . (3.21)

p0 + p3 p1 + ip2

p0 + p3
(c) We construct explicitly the spinor product s(p, q) and t(p, q).
(p1 + ip2 )(q0 + q3 ) (q1 + iq2 )(p0 + p3 )
s(p, q) = uR (p)uL (q) = p ; (3.22)
(p0 + p3 )(q0 + q3 )
(q1 iq2 )(p0 + p3 ) (p1 ip2 )(q0 + q3 )

t(p, q) = uL (p)uR (q) = p . (3.23)
(p0 + p3 )(q0 + q3 )
It can be easily seen that s(p, q) = s(q, p) and t(p, q) = (s(q, p)) .
Now we calculate the quantity |s(p, q)|2 :
2 2
2 p1 (q0 + q3 ) q1 (p0 + p3 ) + p2 (q0 + q3 ) q2 (p0 + p3 )
|s(p, q)| =
(p0 + p3 )(q0 + q3 )
q 0 + q 3 p0 + p 3
=(p21 + p22 ) + (q12 + q22 ) 2(p1 q1 + p2 q2 )
p 0 + p3 q0 + q 3
=2(p0 q0 p1 q1 p2 q2 p3 q3 ) = 2p q. (3.24)
Where we have used the lightlike properties p2 = q 2 = 0. Thus we see that the spinor
product can be regarded as the square root of the 4-vector dot product for lightlike vectors.
3.4 Majorana fermions

(a) We at first study a two-component massive spinor lying in ( 21 , 0) representation,
transforming according to UL (). It satisfies the following equation of motion:
i im 2 = 0. (3.25)
To show this equation is indeed an admissible equation, we need to justify: 1) It is relativis-

tically covariant; 2) It is consistent with the mass-shell condition (namely the Klein-Gordon
equation).
To show the condition 1) is satisfied, we note that is invariant under the simultaneous
transformations of its Lorentz indices and spinor indices. That is U () U (1 ) = .
This implies
UR () UL (1 ) = ,
as can be easily seen in chiral basis. Then, the combination transforms as
UR () UL (1 ). As a result, the first term of the equation of motion transforms as
i iUR () UL (1 )UL () = UR () i .

(3.26)
To show the full equation of motion is covariant, we also need to show that the second term
i 2 transforms in the same way. To see this, we note that in the infinitesimal form,
UL = 1 ii i /2 i i /2, UR = 1 ii i /2 + i i /2.
Then, under an infinitesimal Lorentz transformation, transforms as:
(1 ii i /2 i i /2), (1 + ii i /2 i i /2)
2 2 (1 + ii ( )i /2 i ( )i /2) = (1 ii i /2 + i i /2) 2 .
That is to say, 2 is a right-handed spinor that transforms as 2 UR () 2 . Thus
we see the the two terms in the equation of motion transform in the same way under the
Lorentz transformation. In other words, this equation is Lorentz covariant.
To show the condition 2) also holds, we take the complex conjugation of the equation:
i( ) im 2 = 0.
Combining this and the original equation to eliminate , we get
( 2 + m2 ) = 0, (3.27)
which has the same form with the Klein-Gordon equation.
(b) Now we show that the equation of motion above for the spinor can be derived from
the following action through the variation principle:
Z
4 im T 2 2
S = d x i + ( ) . (3.28)
2
Firstly, let us check that this action is real, namely S = S. In fact,
Z
4 im 2 T 2
S = d x ( i ) ( ) ,
2
where the first term ( i ) = i( )i is identical to the original kinetic term upon
integration by parts. Thus we see that S = S.
Now we vary the action with respect to , that gives
S im
0 = = i 2 2 = 0, (3.29)
2
which is exactly the Majorana equation.
(c) Let us rewrite the Dirac Lagrangian in terms of two-component spinors:

L = (i/ m)
! ! !

0 1 m i 1
= 1 iT2 2
1 0
i m i 2 2
= i1 1 + iT2 2 im T2 2 1 1 2 2

= i1 1 + i2 2 im T2 2 1 1 2 2 ,

(3.30)
where the equality should be understood to hold up to a total derivative term.
3.4. Majorana fermions 13
(d) The familiar global U (1) symmetry of the Dirac Lagrangian ei now becomes
1 ei 1 , 2 ei 2 . The associated Nother current is
J = = 1 1 2 2 . (3.31)
To show its divergence J vanishes, we make use of the equations of motion:
i 1 im 2 2 = 0,
i 2 im 2 1 = 0,
i( 1 ) imT2 2 = 0,
i( 2 ) imT1 2 = 0.
Then we have
J = ( 1 ) 1 + 1 1 ( 2 ) 2 2 2
= m T2 2 1 + 1 2 2 T1 2 2 2 2 1 = 0.

(3.32)
In a similar way, one can also show that the Nother currents associated with the global
symmetries of Majorana fields have vanishing divergence.
(e) To quantize the Majorana theory, we introduce the canonical anticommutation rela-
tion,
a (x), b (y) = ab (3) (x y),

and also expand the Majorana field into modes. To motivate the mode expansion, we
note that the Majorana Langrangian can be obtained by replacing the spinor 2 in the
Dirac Lagrangian (3.30) with 1 . Then, according to our experience in Dirac theory, it can
be found that
d3 p p Xh
Z r i
ipx 2 ipx
(x) = a aa (p)e + (i )a aa (p)e . (3.33)
(2)3 2Ep a
Then with the canonical anticommutation relation above, we can find the anticommutators
between annihilation and creation operators:
{aa (p), ab (q)} = ab (3) (p q), {aa (p), ab (q)} = {aa (p), ab (q)} = 0. (3.34)
On the other hand, the Hamiltonian of the theory can be obtained by Legendre transforming
the Lagrangian:
Z Z
3 L 3 im 2 T 2

H= dx L = d x i + . (3.35)
2
Then we can also represent the Hamiltonian H in terms of modes:
d3 pd3 q X
Z Z
3 ipx T 2 ipx
H= dx p a
a a (p)e + a (i )a a (p)e
(2)6 2Ep 2Eq a,b

( p ) (q ) q b ab (q)eiqx (i 2 )b ab (q)eiqx
im
+ a aa (p)eipx + aT (i 2 )aa (p)eipx
2
2

iqx 2 iqx

( p ) ( q ) b ab (q)e + (i )b ab (q)e
im T
ipx 2 ipx

a aa (p)e + a (i )aa (p)e
2

T 2

iqx 2 iqx
( p ) q b ab (q)e + (i )b ab (q)e
d3 pd3 q X

Z Z
3
= dx p a a (p)a b (q)a ( p ) (q ) q
(2)6 2Ep 2Eq a,b

im 2 im 2 T 2
+ ( p ) ( q ) (i ) 2
(i )( p ) q b ei(pq)x
2 2

im
+ aa (p)ab (q)a ( p ) (q ) q (i 2 ) +

( p ) 2 ( q )
2

im
(i 2 )( p )T 2 q (i 2 ) b ei(p+q)x
2

im
+ aa (p)ab (q)a (i 2 )( p ) (q ) q +
T
(i 2 )( p ) 2 ( q ) (i 2 )
2

im
( p )T 2 q b ei(p+q)x
2

im
+ aa (p)ab (q)a (i 2 )( p ) (q ) q (i 2 ) +
T
(i 2 )( p ) 2 ( q )
2

im
( p )T 2 q (i 2 ) b ei(pq)x
2
3

Z
dp X
= 3
aa (p)ab (p)a ( p ) (p ) p
(2) 2Ep a,b

im 2 2 im 2 T 2
+ ( p ) ( p ) (i ) (i )( p ) p b
2 2

p im p
+ aa (p)ab (p)a ( p ) (p ) p (i 2 ) +

( p ) 2 ( p )
2

im p
(i 2 )( p )T 2 p (i 2 ) b
2

p im p
+ aa (p)ab (p)a (i 2 )( p ) (p ) p +
T
(i 2 )( p ) 2 ( p ) (i 2 )
2

im p
( p )T 2 p b
2

im
+ aa (p)ab (p)a (i 2 )( p ) (p ) p (i 2 ) +
T
(i 2 )( p ) 2 ( p )
2

im
( p )T 2 p (i 2 ) b
2
3.5. Supersymmetry 15
d3 p X 1
Z
2 2 2
h T
i
= E + |p| + m a (p)a b (p)a b a a (p)a (p)
(2)3 2Ep a,b 2 p a b a b
d3 p Ep X h
Z i

= aa (p)a a (p) a a (p)a a (p)
(2)3 2 a
d3 p
Z X
= Ep aa (p)aa (p). (3.36)
(2)3 a
In the calculation above, each step goes as follows in turn: (1) Substituting the mode ex-
pansion for into the Hamiltonian. (2) Collecting the terms into four groups, characterized
by a a, a a , aa and aa . (3) Integrating over d3 x to produce a delta function, with which
one can further finish the integration over d3 q. (4) Using the following relations to simplify
the spinor matrices:
(p )2 = (p )2 = Ep2 + |p|2 , (p )(p ) = p2 = m2 , p = 1
2
(p p ).
In this step, the a a and aa terms vanish, while the aa and a a terms remain. (5)
Using the normalization a b = ab to eliminate spinors. (6) Using the anticommutator
{aa (p), aa (p)} = (3) (0) to further simplify the expression. In this step we have throw away
a constant term 12 Ep (3) (0) in the integrand. The minus sign of this term indicates that
the vacuum energy contributed by Majorana field is negative. With these steps done, we
find the desired result, as shown above.
3.5 Supersymmetry
(a) In this problem we briefly study the Wess-Zumino model, which may be the simplest
supersymmetric field theory in 4 dimensional spacetime. Firstly let us consider the massless
case, in which the Lagrangian is given by
L = + i + F F, (3.37)
where is a complex scalar field, is a Weyl fermion, and F is a complex auxiliary scalar
field. By auxiliary we mean a field with no kinetic term in the Lagrangian and thus it does
not propagate, or equivalently, it has no particle excitation. However, in the following, we
will see that it is crucial to maintain the off-shell supersymmtry of the theory.
The supersymmetry transformation in its infinitesimal form is given by:
= iT 2 , (3.38a)
2
= F + ( ) , (3.38b)

F = i , (3.38c)
where is a 2-component Grassmann variable. Now let us show that the Lagrangian is
invariant (up to a total divergence) under this supersymmetric transformation. This can be
checked term by term, as follows:
( ) = i 2 + ( ) iT 2 ,

( i ) = F + T 2 i + i F + 2

= iF 2 + i T 2 ( ) iT 2 ( )

+ i F + i 2
= iF 2 + i T 2 ( ) iT 2 ( 2 )

+ i F + i 2 2 ,
(F F ) = i( ) F iF ,
where we have used = 2 . Now summing the three terms above, we get:
h i
L = i 2 + F + T 2 , (3.39)
which is indeed a total derivative.
(b) Now let us add the mass term in to the original massless Lagrangian:
L = mF + 12 imT 2 + c.c.

(3.40)
Let us show that this mass term is also invariant under the supersymmetry transformation,
up to a total derivative:
(L) = imT 2 F im + 1

2
im[T F + ( 2 )T ( )T ] 2
+ 1
2
imT 2 [F + ( ) 2 ] + c.c.
= 1
2
imF (T 2 T 2 ) im
1
2
im( ) + 1
2
im( )T ( )T + c.c.
= 1
2
imF (T 2 T 2 ) im ( )
+ 1
2
im( )[ + T ( )T ] + c.c
= im ( ) + c.c (3.41)
where we have used the following relations:
( 2 )T = 2 , 2 ( )T 2 = , T 2 = T 2 , = T ( )T .
Now let us write down the Lagrangian with the mass term:
L = + i + F F + mF + 1
imT 2 + c.c. .

2
(3.42)
Varying the Lagrangian with respect to F , we get the corresponding equation of motion:
F = m . (3.43)
Substitute this algebraic equation back into the Lagrangian to eliminate the field F , we get
L = m2 + i + 21 imT 2 + c.c. .

(3.44)
Thus we see that the scalar field and the spinor field have the same mass.
3.5. Supersymmetry 17
(c) We can also include interactions into this model. Generally, we can write a Lagrangian
with nontrivial interactions containing fields i , i and Fi (i = 1, , n), as
i 2 W [] T 2

W []
L = i i + i i i + Fi Fi + Fi + j + c.c. , (3.45)
i 2 i j i
where W [] is an arbitrary function of i .
To see this Lagrangian is supersymmetry invariant, we only need to check the interactions
terms in the square bracket:
i 2 W [] T 2

W []
Fi + j + c.c.
i 2 i j i
W 2W i 3W
= i ( i ) + Fi (iT 2 j ) + (iT 2 k )Ti 2 j
i i j 2 i j k
i 2W h T i
Fi + ( 2 )T ( )T i 2 j + Ti 2 Fj + j 2 + c.c..

+
2 i j
The term proportional to 3 W/3 vanishes. To see this, we note that the partial derivatives
with respect to i are commutable, hence 3 W/i j k is totally symmetric on i, j, k.
However, we also have the following identity:
(T 2 k )(Ti 2 j ) + (T 2 i )(Tj 2 k ) + (T 2 j )(Tk 2 i ) = 0, (3.46)
which can be directly checked by brute force. Then it can be easily seen that the 3 W/3
term vanishes indeed. On the other hand, the terms containing F also sum to zero, which
is also straightforward to justify. Hence the terms left now are
W 2W 2 T T
i ( i ) +i ( ) ( ) ( i ) 2 j
i i j
W 2W 2W
= i i + i i j i ( i )j
i i j i j
W
= i i , (3.47)
i
which is a total derivative. Thus we conclude that the Lagrangian (3.45) is supersymmetri-
cally invariant up to a total derivative.
Let us end up with a explicit example, in which we choose n = 1 and W [] = g3 /3.
Then the Lagrangian (3.45) becomes

L = + i + F F + gF 2 + iT 2 + c.c. . (3.48)
We can eliminate F by solving it from its field equation,
F + g( )2 = 0. (3.49)
Substituting this back into the Lagrangian, we get
L = + i g 2 ( )2 + ig(T 2 2 ). (3.50)
This is a Lagrangian of massless complex scalar and a Weyl spinor, with 4 and Yukawa
interactions. The field equations can be easily got by the variation.
3.6 Fierz transformations

In this problem, we derive the generalized Fierz transformation, with which one can
express (u1 A u2 )(u3 B u4 ) as a linear combination of (u1 C u4 )(u3 D u2 ), where A is any
normalized Dirac matrices from the following set:
1, , = 2i [ , ], 5 , 5 = i 0 1 2 3 .

(a) The Dirac matrices A are normalized according to

tr (A B ) = 4 AB . (3.51)
For instance, the unit element 1 is already normalized, since tr (11) = 4. For Dirac matrices
containing one , we calculate the trace in Weyl representation without loss of generality.
Then the representation of !

0
=
0
gives tr ( ) = 2 tr ( ) (no sum on ). For = 0, we have tr ( 0 0 ) = 2 tr (122 ) = 4,
and for = i = 1, 2, 3, we have tr ( i i ) = 2 tr ( i i ) = 2 tr (122 ) = 4 (no sum on i).
Thus the normalized gamma matrices are 0 and i i .
In the same way, we can work out the rest of the normalized Dirac matrices, as:
tr ( 0i 0i ) = 2 tr ( i i ) = 4, (no sum on i)
tr ( ij ij ) = 2 tr ( k k ) = 4, (no sum on i, j, k)
tr ( 5 5 ) = 4,
tr ( 5 0 5 0 ) = 4, tr ( 5 i 5 i ) = 4.
Thus the 16 normalized elements are:
1, 0 , i i , i 0i , ij , 5 , i 5 0 , 5 i .

(3.52)
(b) Now we derive the desired Fierz identity, which can be written as:
X
(u1 A u2 )(u3 B u4 ) = C AB CD (u1 C u4 )(u3 D u2 ). (3.53)
C,D
Left-multiplying the equality by (u2 F u3 )(u4 E u1 ), we get:

X
(u2 F u3 )(u4 E u1 )(u1 A u2 )(u3 B u4 ) = C AB CD tr (E C ) tr (F D ). (3.54)
CD
The left hand side:
(u2 F u3 )(u4 E u1 )(u1 A u2 )(u3 B u4 ) = u4 E A F B u4 = tr (E A F B );
the right hand side:
X X
C AB CD tr (E C ) tr (F D ) = C AB CD 4 EC 4 F D = 16C AB EF ,
C,D C,D
thus we conclude:
C AB CD = 1
16
tr (C A D B ). (3.55)
3.7. The discrete symmetries P , C and T 19
(c) Now we derive two Fierz identities as particular cases of the results above. The first
one is:
X tr (C D )
(u1 u2 )(u3 u4 ) = (u1 C u4 )(u3 D u2 ). (3.56)
C,D
16
The traces on the right hand side do not vanish only when C = D, thus we get:
X
(u1 u2 )(u3 u4 ) = 1
4
(u1 C u4 )(u3 C u2 )
C
h
= 1
4
(u1 u4 )(u3 u2 ) + (u1 u4 )(u3 u2 ) + 1
2
(u1 u4 )(u3 u2 )
i
(u1 5 u4 )(u3 5 u2 ) + (u1 5 u4 )(u3 5 u2 ) . (3.57)
The second example is:

X tr (C D )
(u1 u2 )(u3 u4 ) = (u1 C u4 )(u3 D u2 ). (3.58)
C,D
16
Again, the traces vanish if C 6= C D with C a commuting number, which implies

that C = D . That is,
X tr (C C )
(u1 u2 )(u3 u4 ) = (u1 C u4 )(u3 C u2 )
C
16
h
= 4 4(u1 u4 )(u3 u2 ) 2(u1 u4 )(u3 u2 )
1
i
5 5 5 5
2(u1 u4 )(u3 u2 ) 4(u1 u4 )(u3 u2 ) . (3.59)
We note that the normalization of Dirac matrices has been properly taken into account by
raising or lowering of Lorentz indices.
3.7 The discrete symmetries P , C and T

(a) In this problem, we will work out the C, P and T transformations of the bilinear
, with = 2i [ , ]. Firstly,
P (t, x) (t, x)P = 2i (t, x) 0 [ , ] 0 (t, x).
With the relations 0 [ 0 , i ] 0 = [ 0 , i ] and 0 [ i , j ] 0 = [ i , j ], we get:

(
(t, x) 0i (t, x);
P (t, x) (t, x)P = (3.60)
(t, x) ij (t, x).
Secondly,
T (t, x) (t, x)T = 2i (t, x)( 1 3 )[ , ] ( 1 3 )(t, x).

Note that gamma matrices keep invariant under transposition, except 2 , which changes the
sign. Thus we have:
(

(t, x) 0i (t, x);
T (t, x) (t, x)T = (3.61)
(t, x) ij (t, x).
Thirdly,
C (t, x) (t, x)C = 2i (i 0 2 )T (i 0 2 )T = 0 2 ( )T 0 2 .
Note that 0 and 2 are symmetric while 1 and 3 are antisymmetric, we have
C (t, x) (t, x)C = (t, x) (t, x). (3.62)
(b) Now we work out the C, P and T transformation properties of a scalar field . Our
starting point is
P ap P = ap , T ap T = ap , Cap C = bp .
Then, for a complex scalar field
d3 k 1 h ikx
Z i
ikx
(x) = ak e + bk e , (3.63)
(2)3 2k 0
we have
d3 k 1 h
Z i
i(k0 tkx) i(k0 tkx)
P (t, x)P = a k e + b k e = (t, x). (3.64a)
(2)3 2k 0
d3 k 1 h
Z i
i(k0 tkx) i(k0 tkx)
T (t, x)T = a k e + b k e = (t, x). (3.64b)
(2)3 2k 0
d3 k 1 h i(k0 tkx)
Z i
i(k0 tkx)
C(t, x)C = 3
b k e + a k e = (t, x). (3.64c)
(2) 2k 0
As a consequence, we can deduce the C, P , and T transformation properties of the current

J = i ( ) , as follows:

P J (t, x)P = (1)s() i (t, x) (t, x) (t, x) (t, x)

= (1)s() J (t, x), (3.65a)
where s() is the label for space-time indices that equals to 0 when = 0 and 1 when
= 1, 2, 3. In the similar way, we have
T J (t, x)T = (1)s() J (t, x); (3.65b)

CJ (t, x)C = J (t, x). (3.65c)
One should be careful when playing with T it is antihermitian rather than hermitian,

and anticommutes, rather than commutes, with 1.
3.8. Bound states 21
(c) Any Lorentz-scalar hermitian local operator O(x) constructed from (x) and (x)
can be decomposed into groups, each of which is a Lorentz-tensor hermitian operator and
contains either (x) or (x) only. Thus to prove that O(x) is an operator of CP T = +1, it
is enough to show that all Lorentz-tensor hermitian operators constructed from either (x)
or (x) have correct CPT value. For operators constructed from (x), this has been done
as listed in Table on Page 71 of Peskin & Schroeder; and for operators constructed from
(x), we note that all such operators can be decomposed further into a product (including
Lorentz inner product) of operators of the form
(1 m )(1 n ) + c.c
together with the metric tensor . But it is easy to show that any operator of this form
has the correct CP T value, namely, has the same CP T value as a Lorentz tensor of rank
(m+n). Therefore we conclude that any Lorentz-scalar hermitian local operator constructed
from and has CP T = +1.
3.8 Bound states

(a) A positronium bound state with orbital angular momentum L and total spin S can
be build by linear superposition of an electron state and a positron state, with the spatial
wave function L (k) as the amplitude. Symbolically we have
X
|L, Si L (k)a (k, s)b (k, s0 )|0i.
k
Then, apply the space-inversion operator P , we get

X X
P |L, Si = L (k)a b a (k, s)b (k, s0 )|0i = (1)L a b L (k)a (k, s)b (k, s0 )|0i.
k k
(3.66)
Note that b = a , L+1
we conclude that P |L, Si = () |L, Si. Similarly,
X X
C|L, Si = L (k)b (k, s)a (k, s0 )|0i = (1)L+S L (k)b (k, s0 )a (k, s)|0i. (3.67)
k k
That is, C|L, Si = (1)L+S |L, Si. Then its easy to find the P and C eigenvalues of various
states, listed as follows:
S 1 3 1 3 1 3
L S S P P D D
P + +
C + + +
(b) We know that a photon has parity eigenvalue 1 and C-eigenvalue 1. Thus we see
that the decay into 2 photons are allowed for 1 S state but forbidden for 3 S state due to
C-violation. That is, 3 S has to decay into at least 3 photons.
Chapter 4
Interacting Fields and Feynman

Diagrams
4.1 Scalar field with a classical source

In this problem we consider the theory with the following Hamiltonian:
Z
H = H0 d3 j(t, x)(x), (4.1)
where H0 is the Hamiltonian for free Klein-Gordon field , and j is a classical source.
(a) We calculate the probability that the source creates no particles. The corresponding
amplitude is given by the inner product between the in-state and the out-state, both of
which are vacuum in our case. Therefore,
2 2
P (0) =out h0|0iin = lim h0|ei2Ht |0i

t(1i)
n Z o 2 n Z o 2
=h0|T exp i d x Hint |0i = h0|T exp i d4 x j(x)I (x) |0i .
4
(4.2)

(b) Now we expand this probability P (0) to j 2 . The amplitude reads,

n Z o 1
Z
4
h0|T exp i d x j(x)I (x) |0i =1 d4 x d4 y j(x)h0|T I (x)I (y)|0ij(y) + O(j 4 )
2
d3 p 1
Z Z
1 4 4
=1 d x d y j(x)j(y) + O(j 4 )
2 (2)3 2Ep
d3 p 1
Z
1
=1 |j(p)|2 + O(j 4 ). (4.3)
2 (2)3 2Ep
Thus the probability is given by,
1
P (0) = |1 + O(j 4 )|2 = 1 + O(j 4 ), (4.4)
2
where
d3 p 1
Z
|j(p)|2 . (4.5)
(2)3 2Ep
23
24 Chapter 4. Interacting Fields and Feynman Diagrams
(c) We can calculate the probability P (0) exactly, by working out the j 2n term of the
expansion as,
i2n
Z
d4 x1 d4 x2n j(x1 ) j(x2n )h0|T (x1 ) (x2n )|0i
(2n)!
i2n (2n 1)(2n 3) 3 1
Z
= d4 x1 d4 x2n j(x1 ) j(x2n )
(2n)!
Z 3
d p1 d3 pn 1
eip1 (x1 x2 ) eipn (x2n1 x2n )
(2)3n 2n Ep1 Epn
(1)n d3 p |j(p)|2 n (/2)2
Z
= n = . (4.6)
2 n! (2)3 2Ep n!
Then,

X (/2)n 2
P (0) = = e . (4.7)
n=0
n!
(d) Now we calculate the probability that the source creates one particle with momentum
k, which is given by,
n Z o 2
4
P (k) = hk|T exp i d x j(x)I (x) |0i (4.8)

Expanding the amplitude to the first order in j, we get:

d3 p eipx
Z Z 2
4 2
P (k) =hk|0i + i d x j(x) hk|a |0i + O(j )

p p
(2)3 2Ep

Z d3 p j(p) p 2
3
=i 2E (2) (p k) = |j(k)|2 + O(j 3 ). (4.9)

p p
(2)3 2Ep
If we go on to work out all the terms, we get,

X i(2n + 1)(2n + 1)(2n 1) 3 1 2
P (k) = j n+1 (k) = |j(k)|2 e|j(k)| . (4.10)

n
(2n + 1)!
(e) To calculate the probability that the source creates n particles, we write down the
relevant amplitude,
d3 k1 d3 kn
Z n Z o
p hk1 kn |T exp i d4 x j(x)I (x) |0i. (4.11)
(2)3n 2n Ek1 Ekn
Expanding this amplitude in terms of j, we find that the first nonvanishing term is the one
of nth order in j. Repeat the similar calculations above, we can find that the amplitude is:
in d3 k1 d3 kn
Z Z
p d4 x1 d4 xn j1 jn hk1 kn |1 n |0i + O(j n+2 )
n! (2) 3n n
2 Ek1 Ekn
n Z
d k1 d3 kn j n (k) X (1)n
3
d3 p |j(p)|2 n
Z
i
= p . (4.12)
n! (2)3n 2n Ek1 Ekn n=0 2n n! (2)3 2Ep
4.2. Decay of a scalar particle 25
Then we see the probability is given by,

n
P (n) = e , (4.13)
n!
which is a Poisson distribution.
(f ) Its easy to check that the Poisson distribution P (n) satisfies the following identities:
X
P (n) = 1. (4.14)
n
X
hN i = nP (n) = . (4.15)
n
d
The first one is almost trivial, and the second one can be obtained by acting d to both
d
sides of the first identity. If we apply d again to the second identity, we get:
h(N hN i)2 i = hN 2 i hN i2 = . (4.16)
4.2 Decay of a scalar particle

This problem is based on the following Lagrangian,
1 1 1 1
L= ( )2 M 2 2 + ( )2 m2 2 . (4.17)
2 2 2 2
When M > 2m, a particle can decay into two particles. We want to calculate the
lifetime of the particle to lowest order in .
The two-body decay rate is given in (4.86) of P&S,
Z Z 3 3
1 d p1 d p2 1 M((0) (p1 )(p2 ))2 (2)4 (4) (p p1 p2 ). (4.18)

d = 6
2M (2) 4Ep1 Ep2
To lowest order in , the amplitude M is given by,
iM = 2i. (4.19)
The delta function in our case reads,
(4) (p p1 p2 ) = (M Ep1 Ep2 ) (3) (p1 + p2 ), (4.20)
thus,
1 22 d3 p1 d3 p2
Z
1
= 6
(2)4 (M Ep1 Ep2 ) (3) (p1 + p2 ), (4.21)
2 M (2) 4Ep1 Ep2
where an additional factor of 1/2 takes account of two identical s in final state. Further-
more, there are two mass-shell constraints,
m2 + p2i = Ep2 i . (i = 1, 2) (4.22)

Hence,
2 d3 p1 1 2 4m2 1/2
Z
= (2)(M 2Ep1 ) = 1 . (4.23)
M (2)3 4Ep2 1 8M M2
Then the lifetime of is,
8M 4m2 1/2
= 1 = 1 . (4.24)
2 M2
4.3 Linear sigma model

In this problem, we study the linear sigma model described by the following Lagrangian,
L= 1
2
i i 1
2
m2 i i 1
4
(i i )2 . (4.25)
Where is an N -component scalar.
(a) We firstly compute the differential cross sections to the leading order in for the
following three processes,
1 2 1 2 , 1 1 2 2 , 1 1 1 1 . (4.26)
Since the masses of all incoming and outgoing particles are identical, the cross section is
simply given by
d |M|2
= , (4.27)
d CM 64 2 s
where s is the square of center-of-mass energy, and M is the scattering amplitude. From
the Feynman rules its easy to get,
M(1 2 1 2 ) = M(1 1 2 2 ) = 2i,

M(1 1 1 1 ) = 6i. (4.28)
It follows immediately that

2
(1 2 1 2 ) = M (1 1 2 2 ) = ,
16 2 s
92
(1 1 1 1 ) = . (4.29)
16 2 s
(b) Now we study the symmetry broken case, that is, m2 = 2 < 0. Then, the scalar
multiplet can be parameterized as
= ( 1 , , N 1 , + v)T , (4.30)
p
where v is the VEV of ||, and equals to 2 / at tree level.
Substitute this into the Lagrangian, we get

L = 21 ( k )2 + 12 ( )2 12 (22 ) 2 3 k k
4.3. Linear sigma model 27

4
4
2
2 ( k k )
4
( k k )2 . (4.31)
Then its easy to read the Feynman rules from this expression:
k i
= ; (4.32a)
k2 22
k i ij
= ; (4.32b)
k2
= 6iv; (4.32c)
= 2iv ij ; (4.32d)
i j
= 6i; (4.32e)
= 2i ij ; (4.32f)
i j
i j
= 2i( ij k` + ik j` + i` jk ). (4.32g)
` k
(c) With the Feynman rules derived in (b), we can compute the amplitude
M i (p1 ) j (p2 ) k (p3 ) ` (p4 ) ,

as:
h i i i i
M = (2iv)2 ij k`
+ ik j`
+ i` jk

s 22 t 22 u 22
2i( ij k` + ik j` + i` jk ), (4.33)
where s, t, u are Mandelstam variables (See Section 5.4 of P&S). Then, at the threshold
pi = 0, we have s = t = u = 0, and M vanishes.
On the other hand, if N = 2, then there is only one component in , thus the amplitude
reduces to
h 22 22 22 i
M = 2i + + +3
s 22 t 22 u 22
hs+t+u i
4
= 2i + O(p ) . (4.34)
22
In the second line we perform the Taylor expansion on s, t and u, which are of order O(p2 ).
Note that s + t + u = 4m2 = 0, thus we see that O(p2 ) terms are also canceled out.
(d) We minimize the potential with a small symmetry breaking term:
V = 2 i i +
4
(i i )2 aN , (4.35)
which yields the following equation that determines the VEV:
2 + i i i = a iN .

(4.36)
Thus, up to linear order in a, the VEV hi i = (0, , 0, v) is

r
2 a
v= + 2. (4.37)
2
Now we repeat the derivation in (b) with this new VEV, and write the Lagrangian in terms
of new field variable i and , as

L= 1
2
( k )2 + 1
2
( )2 1
2
a

k k 1
2
(22 ) 2
v 3 v k k 1
4
4
2
2 ( k k )
4
( k k )2 . (4.38)
The i j k ` amplitude is still given by
i 2
h
ij k` i ik j` i i` jk
i
M = (2iv) + +
s 22 t 22 u 22
2i( ij k` + ik j` + i` jk ). (4.39)

However this amplitude does not vanish at the threshold. Since the vertices 6=
exactly even at tree level, and also s, t and u are not exactly zero in this case due to nonzero
mass of i . Both deviations are proportional to a, thus we conclude that the amplitude M
is also proportional to a.
4.4 Rutherford scattering

The Rutherford scattering is the scattering of an election by the coulomb field of a
nucleus. In this problem, we calculate the cross section by treating the electromagnetic field
as fixed classical background given by potential A (x). Then the interaction Hamiltonian
is, Z
HI = d3 x e A . (4.40)
4.4. Rutherford scattering 29
(a) We first calculate the T -matrix to lowest order,

Z Z
0 0
out hp |piin =hp |T exp(i d x HI )|pi = hp |pi ie d4 x A (x)hp0 | |pi + O(e2 )
4 0
Z
0
=hp |pi ie d4 x A (x)u(p0 ) u(p)ei(p p)x + O(e2 )
0
=(2)4 (4) (p p0 ) ieu(p0 ) u(p)A (p0 p) + O(e2 ). (4.41)
On the other hand,

0
out hp |piin = hp0 |S|pi = hp0 |pi + hp0 |iT |pi. (4.42)
Thus to the first order of e, we get,
hp0 |iT |pi = ieu(p0 ) u(p)A (p0 p). (4.43)
(b) Now we calculate the cross section d in terms of the matrix elements iM.
The incident wave packet |i is defined to be:
d3 k eibk
Z
|i = (k)|ki, (4.44)
(2)3 2Ek
where b is the impact parameter.
The probability that a scattered electron will be found within an infinitesimal element
3
d p centered at p is,
d3 p 1 2
P = hp|i

3 out in
(2) 2Ep

d3 p 1 d3 kd3 k 0
Z
0 0 ib(kk0 )
= (k) (k ) out hp|ki in out hp|k i in e
(2)3 2Ep (2)6 2Ek 2Ek0
d3 p 1 d3 kd3 k 0
Z 0
= 3 6
0 0
(k) (k ) hp|iT |ki hp|iT |k i eib(kk ) . (4.45)
(2) 2Ep (2) 2Ek 2Ek0
In the last equality we throw away the trivial scattering part from the S-matrix. Note that,
hp0 |iT |pi = iM(2)(Ep0 Ep ), (4.46)
so we have,
d3 p 1 d3 kd3 k 0
Z
0
P = (k) (k0 )|iM|2 (2)2 (Ep Ek )(Ep Ek0 )eib(kk ) .
(2)3 2Ep 6
(2) 2Ek 2Ek0
(4.47)
The cross section d is given by:
Z
d = d2 b P (b), (4.48)
thus the integration over b gives a delta function:

Z
0
d2 b eib(kk ) = (2)2 (2) (k k0 ). (4.49)
The other two delta functions in the integrand can be modified as follows,
Ek 1
(Ek Ek0 ) = (kk kk0 ) = (kk kk0 ), (4.50)
kk v
where we have used |v| = v = vk . Taking all these delta functions into account, we get,
d3 p 1 d3 k
Z
1
d = (k) (k)|iM|2 (2)(Ep Ek ). (4.51)
(2)3 2Ep (2)3 2Ek v
Since the momentum of the wave packet should be localized around its central value, we
can pull out the quantities involving energy Ek outside the integral,
d3 p 1 1 1 d3 k
Z
d = (2)|M| 2
(Ep Ek ) (k) (k). (4.52)
(2)3 2Ep 2Ek v (2)3
Recall the normalization of the wave packet,
d3 k
Z
(k) (k) = 1, (4.53)
(2)3
then,
d3 p 1 1 1
d = |M(k p)|2 (2)(Ep Ek ). (4.54)
(2)3 2Ep 2Ek v
We can further integrate over |p| to get the differential cross section d/d,
dp p2 1 1 1
Z
d
= |M(k p)|2 (2)(Ep Ek )
d (2)3 2Ep 2Ek v
dp p2 1 1 1
Z
Ek
= 3
|M(k p)|2 (2) (p k)
(2) 2Ep 2Ek v k
1
= |M(k, )|2 . (4.55)
(4)2
In the last line we work out the integral by virtue of delta function, which constrains the
outgoing momentum |p| = |k| but leave the angle between p and k arbitrary. Thus the
amplitude M(k, ) is a function of momentum |k| and angle .
(c) We work directly for the relativistic case. Firstly the Coulomb potential A0 = Ze/4r
in momentum space is
Ze
A0 (q) = . (4.56)
|q|2
This can be easily worked out by Fourier transformation, with a regulator emr inserted:
Z
Ze Ze
A (q, m) d3 x eipx emr
0
= . (4.57)
4r |q|2 + m2
This is simply Yukawa potential, and Coulomb potential is a limiting case when m 0.
The amplitude is given by
iM(k, ) = ieu(p) A (q)u(p) with q = p k. (4.58)

4.4. Rutherford scattering 31
Then we have the squared amplitude with initial spin averaged and final spin summed (See
5.1 of P&S for details), as,
X X
2 1 2
1
2
|iM(k, )| = 2
e A (q)A (q) u(p) u(k)u(k) u(p)
spin spin
= e A (q)A (q) tr (p/ + m) (k/ + m)

1 2

2
=2e2 2(p A)(k A) + m2 (k p) A2 .

(4.59)
Note that
Ze Ze
A0 (q) = = , (4.60)
|p k|2 4|k|2 sin2 (/2)
thus
Z 2 e4 1 v 2 sin2 2

X
1 2
2
|iM(k, )| = 4 v 2 sin4 (/2)
, (4.61)
spin
4|k|
and
Z 2 2 1 v 2 sin2 2

d
= (4.62)
d 4|k|2 v 2 sin4 (/2)
. In non-relativistic case, this formula reduces to
d Z 2 2
= (4.63)
d 4m2 v 4 sin4 (/2)
Chapter 5
Elementary Processes of Quantum

Electrodynamics
5.1 Coulomb scattering

In this problem we continue our study of the Coulomb scattering in Problem 4.4. Here
we consider the relativistic case. Lets first recall some main points considered before. The
Coulomb potential A0 = Ze/4r in momentum space is
Ze
A0 (q) = . (5.1)
|q|2
Then the scattering amplitude is given by
iM(k, ) = ieu(p) A (q)u(p) with q = p k. (5.2)
Then we can derive the squared amplitude with initial spin averaged and final spin summed,
as:
1 X 1 X
|iM(k, )|2 = e2 A (q)A (q) u(p) u(k)u(k) u(p)
2 spin 2 spin
1 h i
= e2 A (q)A (q) tr (p/ + m) (k/ + m)
2 h
2 2
2i
= 2e 2(p A)(k A) + m (k p) A . (5.3)
Note that
Ze Ze
A0 (q) = = , (5.4)
|p k|2 4|k| sin2 (/2)
2
thus
Z 2 e4 1 v 2 sin2 2

1 X 2
|iM(k, )| = , (5.5)
2 spin 4|k|4 v 2 sin4 (/2)
Now, from the result of Problem 4.4(b), we know that
2
2 2 2

d 1 1 X Z 1 v sin
= |M(k, )|2 = 2
. (5.6)
d (4)2 2 spin 4|k|2 v 2 sin4 (/2)
33
34 Chapter 5. Elementary Processes of Quantum Electrodynamics
This is the formula for relativistic electron scatted by Coulomb potential, and is called Mott
formula.
Now we give an alternative derivation of the Mott formula, by considering the cross
section of e Z e Z . When the mass of goes to infinity and the charge of is
taken to be Ze, this cross section will reduces to Mott formula.
e Z
- p1 p2 %
% k1 k2 -
e Z
Figure 5.1: The scattering of an electron by a charged heavy particle Z . All initial
momenta go inward and all final momenta go outward.
The corresponding Feynman diagram is shown in Figure 5.1, which reads,
i
iM = Z(ie)2 u(p1 ) u(k1 ) U (p2 ) U (k2 ), (5.7)
t
where u is the spinor for electron and U is the spinor for muon, t = (k1 p1 )2 is one of three
Mandelstam variables. Then the squared amplitude with initial spin averaged and final spin
summed is
X Z 2 e4 h i h i
1
4
|iM|2 = tr ( /
k 1 + m)
( p
/ + m) tr (k/ 2 + M ) (p/ + M )
spin
t2 1 2
Z 2 e4 h
= 16m2 M 2 8M 2 (k1 p1 ) + 8(k1 p2 )(k2 p1 )
t2 i
8m2 (k2 p2 ) + 8(k1 k2 )(p1 p2 ) . (5.8)
Note that the cross section is given by

d 1 |p1 | X
1 2

= |M| . (5.9)
d CM 2Ee 2E |vk1 vk2 | (2)2 4ECM 4
When the mass of goes to infinity, we have E ' ECM ' M , vk2 ' 0, and |p1 | ' |k1 |.
Then the expression above can be simplified to
d 1 X
1 2
= |M| . (5.10)
d CM 16(2)2 M 2 4
When M , only terms proportional to M 2 are relevant in |M|2 . To evaluate this

squared amplitude further, we assign each momentum a specific value in CM frame,
k1 = (E, 0, 0, k), p1 ' (E, sin , 0, k cos ),

k2 ' (M, 0, 0, k), p2 ' (M, k sin , 0, k cos ), (5.11)
5.2. Bhabha scattering 35
then t = (k1 p1 )2 = 4k 2 sin2 2 , and,
1
X
2 Z 2 e4 (1 v 2 sin2 2 ) 2
4
|iM| = M + O(M ). (5.12)
k 2 v 2 sin2 2
Substituting this into the cross section, and sending M , we reach the Mott formula
again,
Z 2 2 1 v 2 sin2 2

d
= . (5.13)
d 4|k|2 v 2 sin4 (/2)
5.2 Bhabha scattering

The Bhabha scattering is the process e+ e e+ e . At the tree level, it consists of two
diagrams, as shown in Figure 5.2.
e e+
e e+
p1 p2 p1 p2

k1 k2 k1 k2

e e+
e e+
Figure 5.2: Bhabha scattering at tree level. All initial momenta go inward and all final
momenta go outward.
The minus sign before the t-channel diagram comes from the exchange of two fermion field
operators when contracting with in and out states. In fact, the s- and t-channel diagrams
correspond to the following two ways of contraction, respectively,
/ A|k
hp1 p2 | A / 1 k2 i, / A|k
hp1 p2 | A / 1 k2 i. (5.14)
In the high energy limit, we can omit the mass of electrons, then the amplitude for the
whole scattering process is,

2 i i
iM = (ie) v(k2 ) u(k1 ) u(p1 ) v(p2 ) u(p1 ) u(k1 ) v(k2 ) v(p2 ) , (5.15)
s t
where we have used the Mandelstam variables s, t and u. They are defined as,
s = (k1 + k2 )2 , t = (p1 k1 )2 , u = (p2 k1 )2 . (5.16)
In the massless case, k12 = k22 = p21 = p22 = 0, thus we have,
s = 2k1 k2 = 2p1 p2 , t = 2p1 k1 = 2p2 k2 , u = 2p2 k1 = 2p1 k2 . (5.17)
We want to get the unpolarized cross section, thus we must average the ingoing spins and
sum over outgoing spins. That is,
1 X e4 X 2
|M|2 = 2 v(k2 ) u(k1 )u(p1 ) v(p2 )

4 spin 4s
e4 X
2
+ u(p ) u(k )v(k ) v(p )

1 1 2 2
4t2

e4 X h
i
v(p2 ) u(p1 )u(k1 ) v(k2 )u(p1 ) u(k1 )v(k2 ) v(p2 ) + c.c.
4st
e4 e4
= 2 tr (k/1 k/2 ) tr (p/2 p/1 ) + 2 tr (k/1 p/1 ) tr (p/2 k/2 )
4s 4t
4 h
e
i
/ /
tr (k 1 k 2 p/2 p/1 ) + c.c.
4st
2e4 (u2 + t2 ) 2e4 (u2 + s2 ) 4e4 u2
= 2
+ +
s2 t2 st
2 2
t s 1 1
= 2e4 2 + 2 + u2 + . (5.18)
s t s t
In the center-of-mass frame, we have k10 = k20 k 0 , and k1 = k2 , thus the total energy
2
ECM = (k10 + k20 )2 = 4k 2 = s. According to the formula for the cross section in the four
identical particles case (Eq.4.85):
d 1 X
1 2
= |M| , (5.19)
d CM 64 2 ECM 4
thus
2 t2 s2
d
2 1 1 2

= + 2 +u + , (5.20)
d CM 2s s2 t s t
where = e2 /4 is the fine structure constant. We integrate this over the angle to get:
2 t2 s2
d
2 1 1 2

= + 2 +u + . (5.21)
d cos CM s s2 t s t
5.3 The spinor products (2)

In this problem we continue our study of spinor product method in last chapter. The
formulae needed in the following are:
1 1
uL (p) = p/uR0 , uR (p) = p/uL0 . (5.22)
2p k0 2p k0
s(p1 , p2 ) = uR (p1 )uL (p2 ), t(p1 , p2 ) = uL (p1 )uR (p2 ). (5.23)
For detailed explanation for these relations, see Problem 3.3.
(a) Firstly, we prove the following relation,
|s(p1 , p2 )|2 = 2p1 p2 . (5.24)
We make use of the another two relations,

1 5 1 + 5
uL0 uL0 = k/0 , uR0 uR0 k/0 , (5.25)
2 2
5.3. The spinor products (2) 37
P
which are direct consequences of the familiar spin-sum formula u0 u0 = k/0 . We now
generalize this to,
1 5 1 + 5
uL (p)uL (p) = p/, uR (p)uR (p) = p/. (5.26)
2 2
We prove the first one:
1 1 1 + 5
uL (p)uL (p) = p/uR0 uR0 p/ = p/ k/0 p/
2p k0 2p k0 2
1 1 5 1 1 5
= p/k/0 p/ = (2p k k/0 p/)p/
2p k0 2 2p k0 2
1 5 1 1 5 1 5
= p/ k/0 p2 = p/. (5.27)
2 2p k0 2 2
The last equality holds because p is lightlike. Then we get,

|s(p1 , p2 )|2 =|uR (p1 )uL (p2 )|2 = tr uL (p2 )uL (p2 )uR (p1 )uR (p1 )
1
= tr (1 5 )p/2 (1 5 )p/1 = 2p1 p2 .

(5.28)
4
(b) Now we prove the relation,
tr ( 1 2 n ) = tr ( n 2 1 ), (5.29)
where i = 0, 1, 2, 3, 5.
To make things easier, let us perform the proof in Weyl representation, without loss of
generality. Then its easy to check that
(
T
, = 0, 2, 5;
( ) = (5.30)
, = 1, 3.
Then, we define M = 1 3 , and it can be easily shown that M 1 M = ( )T , and M 1 M =

1. Then we have,
tr ( 1 2 n ) = tr (M 1 1 M M 1 2 M M 1 n M )
= tr ( 1 )T 2 )T ( n )T = tr ( n 2 1 )T

= tr ( n 2 1 ). (5.31)
With this formula in hand, we can derive the equality,
uL (p1 ) uL (p2 ) = uR (p2 ) uR (p1 ), (5.32)
as follows,
LHS = C uR0 p/1 p/2 uR0 = C tr (p/1 p/2 )

= C tr (p/2 p/1 ) = C uL0 p/2 p/1 uL0 = RHS,
p 1
in which C 2 (p1 k0 )(p2 k0 ) .
(c) The way of proving the Fierz identity
uL (p1 ) uL (p2 )[ ]ab = 2 uL (p2 )uL (p1 ) + uR (p1 )uR (p2 ) ab

(5.33)
has been indicated in P&S. The right hand side of this identity, as a Dirac matrix, which we
denoted by M , can be written as a linear combination of 16 matrices listed in Problem
3.6. In addition, it is easy to check directly that M = M 5 . Thus M must have the
form
1 5 1 + 5
M= V + W .
2 2
Each of the coefficients V and W can be determined by projecting out the other one with
the aid of trace technology, that is,
1 1 5
V = tr M = uL (p1 ) uL (p2 ), (5.34)
2 2
1 1 + 5
W = tr M = uR (p2 ) uR (p1 ) = uL (p1 ) uL (p2 ). (5.35)
2 2
The last equality follows from (5.32). Substituting V and W back, we finally get the left
hand side of the Fierz identity, which completes the proof.
(d) The amplitude for the process at leading order in is given by,
i
iM = (ie2 )uR (k2 ) uR (k1 ) vR (p1 ) vR (p2 ). (5.36)
s
To make use of the Fierz identity, we multiply (5.33), with the momenta variables changed

to p1 k1 and p2 k2 , by vR (p1 ) a and vR (p2 ) b , and also take account of (5.32), which
leads to,
uR (k2 ) uR (k1 )vR (p1 ) vR (p2 )

= 2 vR (p1 )uL (k2 )uL (k1 )vR (p2 ) + vR (p1 )uR (k1 )uR (k2 )vR (p2 )
= 2s(p1 , k2 )t(k1 , p2 ). (5.37)
Then,
4e4 16e4
|iM|2 = |s(p ,
1 2k )|2
|t(k ,
1 2p )|2
= (p1 k2 )(k1 p2 ) = e4 (1 + cos )2 , (5.38)
s2 s2
and
d + + |iM|2 2
(eL eR L R ) = 2
= (1 + cos )2 . (5.39)
d 64 Ecm 4Ecm
It is straightforward to work out the differential cross section for other polarized processes
in similar ways. For instance,
d + e4 |t(p1 , k1 )|2 |s(k2 , p2 )|2 2

(eL eR +
R L ) = = (1 cos )2 . (5.40)
d 64 2 Ecm 4Ecm
5.4. Positronium lifetimes 39
(e) Now we recalculate the Bhabha scattering studied in Problem 5.2, by evaluating all
the polarized amplitudes. For instance,
+
iM(e+ L eR eL eR )

i
= (ie)2 uR (k2 ) uR (k1 ) vR (p1 ) vR (p2 )
s

i
uR (p1 ) uR (k1 ) vR (k2 ) vR (p2 )
t

2 s(p1 , k2 )t(k1 , p2 ) s(k2 , p1 )t(k1 , p2 )
= 2ie . (5.41)
s t
Similarly,
+ t(p1 , k1 )s(k2 , p2 )
iM(e+
L eR eR eL ) = 2ie
2
, (5.42)
s
+ s(p1 , k1 )t(k2 , p2 )
iM(e+
R eL eL eR ) = 2ie2 , (5.43)
s
+ 2 t(p1 , k2 )s(k1 , p2 ) t(k2 , p1 )s(k1 , p2 )
iM(e+
R eL eR eL ) = 2ie , (5.44)
s t
+ t(k2 , k1 )s(p1 , p2 )
iM(e+
R eR eR eR ) = 2ie2 , (5.45)
t
+ s(k2 , k1 )t(p1 , p2 )
iM(e+
L eL eL eL ) = 2ie2 . (5.46)
t
Squaring the amplitudes and including the kinematic factors, we find the polarized differ-
ential cross sections as,
d + d + 2 u2 1 1 2
(eL eR e+
L eR ) = (eR eL e+
R eL ) = + , (5.47)
d d 2s s t
d + d + 2 t2
(e e e+ R eL ) = (e e e+ L eR ) = , (5.48)
d L R d R L 2s s2
d + d + 2 s2
(e e e+ R eR ) = (e e e+ L eL ) = . (5.49)
d R R d L L 2s t2
Therefore we recover the result obtained in Problem 5.2,
2 t2 s2

d + + 2 1
1 2
(e e e e ) = + 2 +u + . (5.50)
d 2s s2 t s t
5.4 Positronium lifetimes

In this problem we study the decay of positronium (Ps) in its S and P states. To
begin with, we recall the formalism developed for bound states with nonrelativistic quantum
mechanics in P&S. The positronium state |Psi, as a bound state of an electron-positron pair,
can be represented in terms of electron and positrons state vectors, as,
d3 k
Z
p 1 1
|Psi = 2MP 3
(k)C ab |ea (k)i |e+
b (k)i, (5.51)
(2) 2m 2m
where m is the electrons mass, MP is the mass of the positronium, which can be taken to be
2m as a good approximation, a and b are spin labels, the coefficient Cab depends on the spin
configuration of |Psi, and (k) is the momentum space wave function for the positronium
in nonrelativistic quantum mechanics. In real space, we have,
r
(mr )3
100 (r) = exp(mr r), (5.52)
r
(mr /2)5 i
21i (r) = x exp(mr r/2). (5.53)

where mr = m/2 is the reduced mass. Then the amplitude of the decay process Ps 2
can be represented in terms of the amplitude for the process e+ e 2 as,
d3 k
Z
1 c e (k)e+ (k) 2 .

M(Ps 2) = 3
(k)Cab M a b (5.54)
m (2)
We put a hat on the amplitude of e+ e 2. In the following we always use a hat to

denote the amplitude of this process.
(a) In this part we study the decay of the S-state positronium. As stated above, we have
to know the amplitude of the process e+ e 2, which is illustrated in Figure 5.3 with the
B replaced with , and is given by,
c = (ie)2 (p1 ) (p2 )

iM
i(k/1 p/1 + m) i(k/1 p/2 + m)

v(k2 ) + u(k1 ), (5.55)
(k1 p1 )2 m2 (k1 p2 )2 m2
where the spinors can be written in terms of two-component spinors and 0 in the chiral
representation as,

k2 0

k1
u(k1 ) = , v(k2 ) = . (5.56)
k1 k2 0
We also write as, !
0
= ,
0
where = (1, i ) and = (1, i ) with i the three Pauli matrices. Then the amplitude
can be brought into the following form,

2 0 t u
iM = ie (p1 ) (p2 )
c + , (5.57)
(k1 p1 )2 m2 (k1 p2 )2 m2
with
p p p p

t = k2 k1 k2 k1 m
p p p p
+ k2 k1 k2 k1 (k1 p1 ) ,
p p p p

u = k2 k1 k2 k1 m
p p p p

+ k2 k1 k2 k1 (k1 p2 ) .
In the rest of the part (a), we take the nonrelativistic limit, with the momenta chosen to be
k1 = k2 = (m, 0, 0, 0), p1 = (m, 0, 0, m), p2 = (m, 0, 0, m). (5.58)
Accordingly, we can assign the polarization vectors for final photons to be,
(p1 ) = 1 (0, 1, i, 0),

2
(p2 ) = 1 (0, 1, i, 0).
2
(5.59)

Now substituting the momenta (5.58) into (5.57), noticing that ki = ki = m (i =
1, 2), and (k1 p1 )2 = (k1 p2 )2 = m2 , and also using the trick that one can freely make
the substitution since the temporal component of the polarization vectors
always vanishes, we get a much more simplified expression,
c = ie2 (p1 ) (p2 ) 0 3 3 .

iM (5.60)
The positronium can lie in spin-0 (singlet) state or spin-1 (triplit) state. In the former
case, we specify the polarizations of final photons in all possible ways, and also make the
substitution 0 12 (See (5.49) of P&S), which leads to,

cs = iM
iM cs = i2 2e2 , cs = iM
iM cs = 0, (5.61)
++ + +
where the subscripts denote final photons polarizations and s means singlet. We show the
mid-step for calculating iMs++ as an example,
ie2 h 1 i
cs =
iM ++ tr ( + i 2 ) 3 ( 1 + i 2 ) ( 1 + i 2 ) 3 ( 1 + i 2 ) 0 = i2 2e2 .
2
In the same way, we can calculate the case of triplet initial state. This time, we make the

substitution 0 n / 2, with n = (x iy)/ 2 or n = z, corresponding to three
independent polarizations. But it is straightforward to show that the amplitudes with these
initial polarizations all vanish, which is consistent with our earlier results by using symmetry
arguments in Problem 3.8.
Therefore it is enough to consider the singlet state only. The amplitude for the decay of
a positronium in its 1 S0 state into 2 then follows directly from (5.54), as
(x = 0) cs
M (1 S0 2) = M , (5.62)
m
p
where (x = 0) = (m/2)3 / according to (5.52). Then the squared amplitude with final
photons polarizations summed is
2
M(1 S0 2)2 = |(0)| |Ms++ |2 + |Ms |2 = 165 m2 .
X
(5.63)
spin
2m
Finally we find the decay width of the process Ps(1 S0 ) 2, to be

d3 p1 d3 p2 P
Z
1 1 1 1
2
(2)4 (4) (pPs p1 p2 )
( S0 2) = M( S0 2)
2 4m (2)6 2E1 2E2
d3 p1 P
Z
1 1 1
2
= M( S0 2) (2)(m E1 )
2 4m (2)3 4m2
1
= 5 m, (5.64)
2
where an additional factor of 1/2 follows from the fact that the two photons in the final
state are identical particles.
(b) To study the decay of P state (l = 1) positronium, we should keep one power of
3-momenta of initial electron and positron. Thus we set the momenta of initial and final
particles, and also the polarization vectors of the latter, in e e+ 2, to be
k1 = (E, 0, 0, k), k2 = (E, 0, 0, k),

p1 = (E, E sin , 0, E cos ), p2 = (E, E sin , 0, E cos ),
(p1 ) = 1 (0, cos , i, sin ),
2
(p2 ) = 1 (0, cos , i, sin ).
2
(5.65)
Here we have the approximate expression up to linear order in k:

p p k
k1 = k2 = m 3 + O(k 2 ),
2 m
p p k
k2 = k1 = m + 3 + O(k 2 ),
2 m
1 1 k cos
2 2
= 2 + O(k 2 ),
(k1 p1 ) m 2m 2m3
1 1 k cos
2 2
= 2 + + O(k 2 ).
(k1 p2 ) m 2m 2m3
Consequently,
2 1 3 3 3 3
+ O(k 2 ),

t = 2m ( s + c ) mk + + 2
2 1 3 3 3 3
+ O(k 2 ),

u = 2m ( s + c ) mk + + 2
where we use the shorthand notation s = sin and c = cos . We can use these expansion
to find the terms in the amplitude iM c of linear order in k, to be
2 k 0 h
iMc
O(k)
= ie (p )
1 2 (p ) 2c ( 1 s + 3 c ) 2c ( 1 s + 3 c )
2m
3 3 3
3 3 3
i
+ + + 2 + + + 2 , (5.66)
Feeding in the polarization vectors of photons, and also make the substitution 0 n

/ 2 or 1/ 2 for triplet and singlet positronium, respectively, as done in last part, we get
c
iM |O(k) = 0,
c
iM 2
|O(k) = i2s (1 + c )e k/m,
+
2 2
iM
c |O(k) = i2 2e2 k/m, c+
iM |O(k) = i2 2s e k/m,
c
iM |O(k) = 0,
c
iM 2
|O(k) = i2s (1 + c )e k/m,
c
iM |O(k) = 0,
c
iM |O(k) = 0. (5.67)
The vanishing results in the last line indicate that S = 0 state of P -wave positronium cannot
decay to two photons.
(c) Now we prove that the state,
d3 p
Z
i (p)ap+k/2 i bp+k/2 |0i
p
|B(k)i = 2MP 3
(5.68)
(2)
, is a properly normalized state for the P -wave positronium. In fact,
d3 p0 d3 p 0
Z
hB(k)|B(k)i = 2MP (p )i (p)
(2)3 (2)3 j
h0|bp0 +k/2 j ap0 +k/2 ap+k/2 i bp0 +k/2 |0i
d3 p0 d3 p 0
Z
= 2MP (p )i (p)
(2)3 (2)3 j
h0|bp0 +k/2 j i bp0 +k/2 |0i(2)3 (3) (p0 p)
d3 p
Z
= 2MP 3
j (p)i (p)h0|bp+k/2 j i bp+k/2 |0i
(2)
d3 p
Z
= 2MP 3
j (p)i (p)h0| tr (j i )|0i(2)3 (3) (0)
(2)
= 2MP (2)3 (3) (0), (5.69)
which is precisely the needed normalization of a state. In this calculation we have used the
anticommutation relations of creation and annihilation operators, as well as the normaliza-
tion of the wave function and the matrices.
(d) Now we evaluate the partial decay rate of the S = 1 P -wave positronium of definite
J into two photons. The states for the positronium is presented in (c), with the matrices
chosen as
1 i

, J = 0,
6

1 ijk j k

= n , J = 1, (5.70)

2
1 ij j

h , J = 2,

3
and the wave function given by (5.53).
Firstly, consider the J = 0 state, in which case we have,
d3 k
Z
3 1 1
c e +
i iM

iM( P0 ) = 3
i (k) a (k)eb (k) , (5.71)
m (2) 6 ab
where , = + or are labels of photons polarizations and a, b = or are spinor indices.

For amplitude iM,
c we only need the terms linear in k, as listed in (5.67). Let us rewrite
this as,
c e (k)e+ (k) = F ab k i .

iM a b ,i
In the same way, the wave function can also be put into the form of i (x) = xi f (r), with
r = |x|. Then the integration above can be carried out to be,

3 i i ab i
iM( P0 ) = ab F,j i j (x)
6m x x=0
i i ab
= ab F,i f (0). (5.72)
6m
On the other hand, we have chose the direction of k to be in the x3 -axis, then F,1
ab
=
ab
F,2 = 0 as a consequence. Therefore,
r
i 7
iM(3 P0 ) = f (0) F,3

F,3 = m sin . (5.73)
6m 24
Square these amplitudes, sum over the photons polarizations, and finish the phase space
integration in the same way as what we did in (a), we finally get the partial decay rate of
the J = 0 P -wave positronium into two photons to be,
1 7
(3 P0 ) = m. (5.74)
576
The positronium in 3 P1 state, namely the case J = 1, cannot decay into two photons
by the conservation of the angular momentum, since the total angular momentum of two
physical photons cannot be 1. Therefore let us turn to the case of J = 2. In this case we
should average over the initial polarizations of the positronium, which can be represented
by the symmetric and traceless polarization tensors hijn , with n = 1, 2, , 5 the labeled of
5 independent polarizations. Let us choose these tensors to be,
hij
1 =
1 ( i2 j3
2
+ i3 j2 ), hij
2 =
1 ( i1 j3
2
+ i3 j1 ),
hij
3 =
( i1 j2
1
2
+ i2 j1 ), hij
4 =
( i1 j1
1
2
i2 j2 ),
hij
5 =
1 ( i1 j1
2
i3 j3 ). (5.75)
Then the decay amplitude for a specific polarization of J = 2 Ps can be represented as,
d3 k
Z
3 1 1
ij j

c e +

iMn ( P2 ) = i (k) hn i M a (k)e b (k)
m (2)3 3 ab
1 j
= hij ab
n ab F,i f (0). (5.76)
3m
Now substituting all stuffs in, we find the nonvanishing components of the decay amplitude
to be,
r
7
iM2 (3 P2 ) = im,
48
5.5. Physics of a massive vector boson 45
r
7
iM2 (3 P2 ) = im sin2 ,
r48
7
iM5 (3 P2 ) = 2 im sin2 . (5.77)
48
Squaring these amplitudes, summing over photons polarizations and averaging the initial
polarization of the positronium (by dividing the squared and summed amplitude by 5), we
get,
1 X 2 7 m2
Mn (3 P2 2) = (1 + sin2 + 4 sin4 ). (5.78)
5 spin 120
Finally, we finish the phase space integration and get the partial decay rate of 3 P2 positro-
nium into 2 photons to be
19 7
(3 P2 ) = m. (5.79)
19200
5.5 Physics of a massive vector boson

In this problem, the mass of electron is always set to zero.
(a) We firstly compute the cross section (e+ e B) and the decay rate (B e+ e ).
For the cross section, the squared amplitude can be easily found to be
1 X 1 X (i) 0 2
|iM|2 = ig v(p ) u(p) = 2g 2 (p p0 ). (5.80)

4 spin 4 spin
Note that we have set the mass of electrons to be zero. Then the cross section can be
deduced from (4.79). Lets take the initial momenta to be,
p= 1
2
(E, 0, 0, E), p0 = 1
2
(E, 0, 0, E), (5.81)
with E being the center-of-mass energy. Then its easy to get,

g2 g2
= (2)(MB E) = (2)2MB (MB2 s) = g 2 (MB2 s), (5.82)
4E 4E
where s = E 2 .
To deduce the decay rate, we should average polarizations of massive vector B instead
of two electrons. Thus the squared amplitude in this case reads,
1 X 8
|iM|2 = g 2 (p p0 ). (5.83)
3 spin 3
The decay rate can be found from (4.86),
d3 p d3 p0 1
Z
1 1 1P
= |M|2
(2)4 (4) (pB p p0 )
2MB (2)3 2 3 2Ep 2Ep0 3
d3 p 1 16 2 2
Z
1
= g Ep (2)(MB 2Ep )
2MB (2)3 4Ep2 3
g 2 MB
Z
4 4 2 2 1 1
= dp 3
g Ep2 ( 2
MB Ep ) = . (5.84)
(2)2 2MB 12
We see the cross section and the decay rate satisfy the following relation, as expected,
12 2
(e+ e B) = (B e+ e )(s M 2 ). (5.85)
MB
(b) Now we calculate the cross section (e e+ + B) in COM frame. The related
diagrams are shown in Figure 5.3.
B B
p1 p2 p1 p2
+
k1 k2 k1 k2
e e+ e e+
Figure 5.3: The tree diagrams of the process e e+ + B. All initial momenta go inward
and all final momenta go outward.
The amplitude reads,

h i i i
iM = (ie)(ig) (p1 )e (p2 )v(k2 ) + u(k1 ), (5.86)
k/1 p/1 k/1 p/2
where is the polarization of photon while e is the polarization for B. Now we square
this amplitude,

1 X 2 1 2 2 (k/1 p/1 ) (p/1 k/2 )
|iM| = e g g g tr + k/1
4 spin 4 t u
(k/ p/ ) (p/1 k/2 )

1 1
+ k/2
t u

2 2 (k1 p1 )(k2 p1 ) (k1 p1 )(k2 p1 )
= 8e g 2
+
t u2

2(k1 k2 )(k1 k2 k1 p1 k2 p1 )
+
tu

2 2 u t 2s(s + t + u)
= 2e g + +
t u tu
2

2 2 u t 2sMB
= 2e g + + . (5.87)
t u tu
Then the cross section can be evaluated as,

d 1 |p1 | P
1 2

= |M|
d CM 2Ek1 2Ek2 |vk1 vk2 | (2)2 4ECM 4
e2 g 2 MB2 u 2sMB2

t
= 1 + + . (5.88)
32 2 s s t u tu
We can also write this differential cross section in terms of squared COM energy s and
scattering angle . To do this, we note,
2
s = ECM , 2
t = (MB2 ECM ) sin2 2 , 2
u = (MB2 ECM ) cos2 2 . (5.89)
Then we have,
e2 g 2 (1 MB2 /s) 4sMB2
d
2
= 1 + cos + , (5.90)
d CM 16 2 s sin2 (s MB2 )2
and,
g 2 (1 MB2 /s) 4sMB2
d
2
= 1 + cos + . (5.91)
d cos CM 2s sin2 (s MB2 )2
(c) The differential cross obtained in (b) diverges when 0 or . Now let us study
the former case, namely 0.
If we cut of the integral from c2 ' m2e /s, then we have,
Z 1m2e /s
g 2 (1 MB2 /s) 4sMB2
Z
d dt
sin d ' 2+ 2 2
c d cos CM 2s (s MB ) 1 t2
g 2 (1 MB2 /s) 4sMB2
s
' 2+ log
4s (s MB2 )2 m2e
g 2 1 + MB4 /s2 s
= log . (5.92)
2 s MB2 m2e
Now we calculate the following expression,
Z 1
dx f (x)(e+ e B)E =(1x)s

CM
Z0 1 2 s
1 + (1 x) 2 2

= dx log g MB (1 x)s
0 2 x m2e
2 4 2
g 1 + MB /s s
= 2
log . (5.93)
2 s MB m2e
5.6 The spinor products (3)

This problem generalize the spinor product formalism to the processes involving external
photons.
(a) Firstly we can represent photons polarization vectors in terms of spinors of definite
helicity. Let the momentum of the photon be k, and p be a lightlike momentum such that
p k 6= 0. Then, the polarization vector (k) of the photon can be taken to be,
1 1
+ (k) = uR (k) uR (p), (k) = uL (k) uL (p), (5.94)
4p k 4p k
where the spinors uL,R (k) have been introduced in Problems 3.3 and 5.3. Now we use this
choice to calculate the polarization sum,
+ (k)
+ (k) + (k) (k)
1 h i
= uR (k) uR (p)uR (p) uR (k) + uL (k) uL (p)uL (p) uL (k)
4p k
1 p k + p k
tr p/ k/ = g +

= . (5.95)
4p k pk
When dotted into an amplitude with external photon, the second term of the result vanishes.
This justifies the definitions above for photons polarization vectors.
(b) Now we apply the formalism to the process e+ e 2 in the massless limit. The
relevant diagrams are similar to those in Figure 5.3, except that one should replace the
label B by . To simplify expressions, we introduce the standard shorthand notations as
follows:
pi = uR (p), p] = uL (p), hp = uL (p), [p = uR (p). (5.96)
Then the spin products become s(p1 , p2 ) = [p1 p2 ] and t(p1 , p2 ) = hp1 p2 i. Various expres-
sions get simplified with this notation. For example, the Fierz identity (5.37) now reads
[k2 k1 i[p1 p2 i = 2[p1 k2 ]hk1 p2 i. Similarly, we also have hk1 k2 ]hp1 p2 ] = 2hk1 p1 i[p2 k2 ].

Now we write down the expression for tree amplitude of e+ R eL R L . For illustration,
we still keep the original expression as well as all explicit mid-steps. The auxiliary lightlike
momenta used in the polarization vectors are arbitrarily chosen such that the calculation
can be mostly simplified.

iM(e+
R eL L R )

i i
= (ie)2 (p1 )+ (p2 )uL (k2 )
+
uL (k1 )
k/1 p/1 k/1 p/1
hk2 (k/1 p/1 ) k1 ] hk2 (k/1 p/2 ) k1 ]

hk2 p1 ][k1 p2 i
2
= ie p +
4 (k2 p1 )(k1 p2 ) t u
hk2 p1 ][k1 p2 i hk2 k1 ]hk1 k1 ] hk2 p1 ]hp1 k1 ]

= ie2
2u t
hk2 k1 ]hk1 k1 ] hk2 p2 ]hp2 k1 ]

+
u
2

2ie hk1 k2 i[p1 k1 ]hk2 p2 i[k1 k1 ] hk2 p1 i[k1 p1 ]hk2 p2 i[k1 p1 ]
=
u t

hk2 k2 i[k1 p1 ]hk1 p2 i[k1 k1 ] hk2 k2 i[p2 p1 ]hp2 p2 i[k1 k1 ]
+
u
hk2 p1 i[k1 p1 ]hk2 p2 i[k1 p1 ]
= 2ie2 , (5.97)
tu
where we have used the spin sum identity p/ = p]hp + pi[p in the third equality, and also the
Fierz transformations. Note that all spinor products like hppi and [pp], or hp ki and [p k]
vanish. Square this amplitude, we get
iM(e+ e L R )2 = 4e4 t .

R L (5.98)
u
In the same way, we calculate other polarized amplitudes,

iM(e+
R eL R L )
[k2 (k/1 p/1 ) k1 i [k2 (k/1 p/2 ) k1 i

2[k1 p1 ihk2 p2 ]
= ie p +
4 (k1 p1 )(k2 p2 ) t u
hk2 p1 i[k1 p2 ]hk2 p2 i[k1 p2 ]
= 2ie2 . (5.99)
tu
Note that we have used a different set of auxiliary momenta in photons polarizations. After
evaluating the rest two nonvanishing amplitudes, we get the squared polarized amplitudes,
as follows:
M(e+ e L R )2 = M(e+ e R L )2 = 4e4 t ,

R L L R (5.100)
u
M(e+ e L R )2 = M(e+ e L R )2 = 4e4 u .

L R L R (5.101)
t
Then the differential cross section follows straightforwardly,
22 t
X
d 1 1 2 u
= |iM| = + , (5.102)
d cos 16s 4 spin s u t
which is in accordance with (5.107) of P&S.

Chapter 6
Radiative Corrections: Introduction
6.1 Rosenbluth formula

In this problem we derive the differential cross section for the electron-proton scattering
in the lab frame, assuming that the scattering energy is much higher than electrons mass,
and taking account of the form factors of the proton. The result is known as Rosenbluth
formula. The relevant diagram is shown in Figure 6.1.
e p
p1 p2
k1 k2
e p
Figure 6.1: The electron-proton scattering. The blob denotes form factors that includes
the effect of strong interaction. All initial momenta go inward and all final momenta go
outward.
Let us firstly work out the kinematics. In the lab frame, the momenta can be parame-
terized as
k1 = (E, 0, 0, E), p1 = (E 0 , E 0 sin , 0, E 0 cos ), k2 = (M, 0, 0, 0), (6.1)
and p2 can be found by momentum conservation, k1 + k2 = p1 + p2 . With the on-shell

condition p22 = M 2 , we find that
ME
E0 = . (6.2)
M + 2E sin2 2
We also use q = k1 p1 to denote the momentum transfer and t = q 2 its square. Note that
we have set the electron mass to zero.
Now we write down the amplitude M.
i q

2 2 2 i
iM = (ie) U (p2 ) F1 (q ) + F2 (q ) U (k2 ) u(p1 ) u(k1 ), (6.3)
2M t
51
52 Chapter 6. Radiative Corrections: Introduction
where U is the spinor for the proton and u is for the electron, M is the mass of the proton. At
this stage, we convert this expression into a more convenient form by means of the Gordon
identity (see Problem 3.2),
(p2 + k2 )

2 i
iM = (ie) U (p2 ) (F1 + F2 ) F2 U (k2 ) u(p1 ) u(k1 ). (6.4)
2M t
Now, the squared amplitude with initial spins averaged and final spins summed is,
e4 (p2 + k2 )

1 X 2
|M| = 4 tr (F1 + F2 ) F2 (k/2 + M )
4 4q 2M
(p2 + k2 )

(F1 + F2 ) F2 (p/2 + M ) tr k/1 p/1
2M
4 2

4e M
2E 2 + 2E 02 + q 2 (F1 + F2 )2

=
q4

2 q2
0 2 2 q2

2F1 F2 + F2 1 + 4M 2 (E + E ) + q 1 4M 2 . (6.5)
There are two terms in the square bracket in the last expression. We rewrite the first factor
in the second term as

q2 q2
2F1 F2 + F2 1 + 4M 2 = (F1 + F2 )2 F12 + 4M
2 2
2 F2 ,
and combine the (F1 + F2 )2 part into the first term, which leads to,
4e4 M 2 q4

1 X 2 2 2 q2 2
0 2
|M| = 2M 2
(F1 + F2 ) + 4 F1 4M 2 F2 EE cos 2 ,
4 q4
where we have used the following two relations which can be easily justified,
q2
E0 E = 2m
, (6.6)
q 2 = 4E 0 E sin2 2 . (6.7)
Now we can put the squared amplitude into its final form,
1 X 16e4 E 2 M 3
|M|2 = 4
q M + 2E sin2 2

4
q2

q2
F12 4M 2
2
2 F2 cos
2
(F1 + F2 )2 sin2
2
. (6.8)
2M 2
On the other hand, we can derive the A + B 1 + 2 differential cross section in the lab
frame as
d 3 p1 d 3 p2
Z
1
dL = |M|2 (2)4 (4) (p1 + p2 pA pB ). (6.9)
2EA 2EB |vA vB | (2)6 2E1 2E2
In our case, EA = E, EB = M , E1 = E 0 , and |vA vB | ' 1, thus,
d3 p1 d3 p2
Z
1
dL = 6
|M|2 (2)4 (4) (p1 + p2 pA pB )
4EM (2) 2E1 2E2
6.2. Equivalent photon approximation 53
E 02 dE 0 d cos d
Z
1 2 0 0

= |M| (2) E + E2 (E ) E M
4EM (2)3 2E 0 2E2
1
E 02 dE 0 d cos d E 0 E cos
Z
1 2 0 ME
= |M| 1 + E
4EM (2)2 2E 0 2E2 E2 (E 0 ) M + 2E sin2 2
E0
Z
1 d cos
= |M|2 ,
4EM 8 M + 2E sin2 2
where we use the notation E2 = E2 (E 0 ) to emphasize that E2 is a function of E 0 . That is,

E2 = M 2 + E 2 + E 02 2E 0 E cos .
Then,
d 1 2
= 2 |M| . (6.10)
d cos L 32 M + 2E sin2
2
So finally we get the differential cross section, the Rosenbluth formula,
d 2
=
2E 2 1 + 2E sin2 2 sin4 2

d cos L M
q2

2 q2 2
2 2 2
F1 4M 2 F2 cos 2 (F1 + F2 ) sin 2 . (6.11)
2M 2
6.2 Equivalent photon approximation

In this problem we study the scattering of a very high energy electron from a target in
the forward scattering limit. The relevant matrix element is,
ig c
M = (ie)u(p0 ) u(p) M (q). (6.12)
q2
(a) First, the spinor product in the expression above can be expanded as,
u(p0 ) u(p) = A q + B q + C 1 + D 2 . (6.13)
Now, using the fact that q u(p0 ) u(p) = 0, we have,
0 = Aq 2 + Bq q ' 4AEE 0 sin2

2
+ Bq q B 2 . (6.14)
(b) It is easy to find that
1 = N (0, p0 cos p, 0, p0 sin ), 2 = (0, 0, 1, 0),
where N = (E 2 + E 02 2EE 0 cos )1/2 is the normalization constant. Then, for the right-

handed electron with spinor u+ (p) = 2E(0, 0, 1, 0)T and left-handed electron with u (p) =
2E(0, 1, 0, 0)T , it is straightforward to show that

u+ (p0 ) = 2E 0 (0, 0, cos 2 , sin 2 )T , u (p0 ) = 2E 0 ( sin 2 , cos 2 , 0, 0), (6.15)
and,
E + E0
u (p0 ) 1 u (p) ' EE 0 , (6.16)
|E E 0 |

u (p0 ) 2 u (p) ' i EE 0 , (6.17)
0 0
u (p ) 1 u (p) = u (p ) 2 u (p) = 0. (6.18)
That is to say, we have,

E + E0
C = EE 0 , D = i EE 0 . (6.19)
|E E 0 |
(c) The squared amplitude is given by,
e2 c c (q) C 1 + D 2 C + D .
|M |2 =

M (q) M 1 2 (6.20)
(q 2 )2
Averaging and summing over the initial and final spins of the electron respectively, we get,
e2 c

1 X 2 c (q) |C+ |2 + |C |2 1 + |D+ |2 + |D |2 2

|M| = 2 2
M (q)M 1 2
2 2(q )

1 2 + C+ D+ + C D 2

+ C+ D+ + C D 1
e2 c
E + E 0 2
= 2 2 M (q)M (q)EE
c 0 2
0
1
1 + 2 2 . (6.21)
(q ) EE
Then the cross section reads,
d3 p0 d3 pt
Z Z X
1 1
|M| (2)4 (4)
2
P
d = 3 0 3
pi
2E2Mt (2) 2E (2 )2Et 2
e2 d3 p0 EE 0 2 E + E 0 2
Z
= +1
2E2Mt (2)3 2E 0 3(q 2 )2 E E0
d 3 pt
Z
c (q)|2 (2)4 (4) Ppi

|M
(2 3 )2Et
1
Z h 2 x 2 i Z 2 sin
= dx 1 + d
2E2Mt 2 x 0 4(1 cos )2
d3 pt
Z
c (q)|2 (2)4 (4) Ppi .

3
|M (6.22)
(2 )2Et
where we have used the trick described in the final project of Part I (radiation of gluon jets)
to separate the contractions of Lorentz indices, and x (E E 0 )/E. Now let us focus on
the integral over the scattering angle in the last expression, which is contributed from the
following factor, Z
2 sin
Z
d
d 4
. (6.23)
0 4(1 cos ) 0
which is logarithmically divergent as 0.

6.3. Exotic contributions to g 2 55
(d) We reintroduce the mass of the electron into the denominator to cut off the divergence,
namely, let q 2 = 2(EE 0 pp0 cos ) + 2m2 . Then we can expand q 2 , treating m2 and as
small quantities, as,
x2
q 2 ' (1 x)E 2 2 m2 .
1x
Then the polar angle integration near = 0 becomes,
2
x2 m2 E2
Z
3 2 1
d + log . (6.24)
0 (1 x)2 E 2 2 m2
(e) Combining the results above, the cross section can be expressed as,
2
x2 m2
Z Z 2 x 2 i Z
1 h
3 2
d = dx 1 + d +
2E2Mt 2 x 0 (1 x)2 E 2
d3 pt
Z
c (q)|2 (2)4 (4) Ppi

|M
(2 3 )2Et
1 + (1 x)2 E2
Z
1
= dx log 2
2E2Mt 2 x2 m
d3 pt
Z
c (q)|2 (2)4 (4) Ppi .

|M (6.25)
(2 3 )2Et
6.3 Exotic contributions to g 2

(a) The 1-loop vertex correction from Higgs boson is,
i 2 Z dd k i i i
0
u(p ) u(p) = d 2 2
u(p0 ) u(p)
2 (2) (k p) mh k/ + /q m k/ m
i2 1
Z 1x Z
dd k 0 2u(p0 )N u(p)
Z
= dx dy , (6.26)
2 0 0 (2)d (k 02 )3
with
N = (k/ + /q + m) (k/ + m), (6.27)

0
k = k xp + yq, (6.28)
2
= (1 x)m + xm2h 2 2
x(1 x)p y(1 y)q + 2xyp q. (6.29)
To put this correction into the following form,

i q
= F1 (q) + F2 (q), (6.30)
2m
we first rewrite N as,
N = A + B(p0 + p) + C(p0 p) , (6.31)
where term proportional to (p0 p) can be thrown away by Ward identity q (q) = 0. This
can be done by gamma matrix calculations, leading to the following result,
h i
N = d2 1 k 02 + (3 + 2x x2 )m2 + (y xy y 2 )q 2 + (x2 1)m(p0 + p) . (6.32)

Then, using Gordon identity, we find,

h i i q
N = 2
1 k 02 + (x + 1)2 m2 + (y y 2 xy)q 2 + 2m2 (1 x2 ).

d
(6.33)
2m
Comparing this with (6.30), we see that
1 1x
d4 k 0 1 x 2
Z Z Z
2 2
F2 (q = 0) = 2i m dx dy
0 0 (2)4 (k 02 )3
1
2 (1 x)2 (1 + x)
Z
= dx . (6.34)
(4)2 0 (1 x)2 + x(mh /m)2
To carry out the integration over x, we use the approximation that mh m. Then,
Z 1
2 1 + x x2

1
F2 (q = 0) ' dx
(4)2 0 1 + x(mh /m)2 (mh /m)2
2 h
2 2 7
i
' log(m h /m ) 6
. (6.35)
(4)2 (mh /m)2
(b) According to (a), the limits on and mh is given by,
2 h i
F2 (q = 0) = log(m2h /m2 ) 7
6
< 1 1010 . (6.36)
(4)2 (mh /m)2
For electron, ' 3 106 , m ' 0.511MeV, and with mh 60 GeV, we have F2 (q = 0)
1022 1010 . For realistic case mh ' 125GeV the effect is even smaller. On the other
hand, for muon, we have = 6 104 , m ' 106MeV, and with the input mh ' 125GeV,
we have F2 (q = 0) 1014 . At present the experimentally measured muons anomalous
magnetic moment is a bit different from the prediction of Standard Model, and the difference
is of order 109 , a not decisive but still noteworthy anomaly. More can be found in [3].
(c) The 1-loop correction from the axion is given by,

2 Z dd k i i i
0
u(p ) u(p) = d 2 2
u(p0 ) 5 5 u(p)
2 (2) (k p) mh k/ + /q m k/ m
i2 1
Z 1x Z
dd k 0 2u(p0 )N u(p)
Z
= dx dy , (6.37)
2 0 0 (2)d (k 02 )3
in which k 0 and are still defined as in (a) except the replacement mh ma , while N is
now given by,
N = 5 (k/ + /q + m) (k/ + m) 5 = (k/ + /q m) (k/ m). (6.38)
Repeating the same derivation as was done in (a), we get,
N = 2
1 k 02 (1 x y)yq 2 + (1 x)2 m2 (1 x)2 m(p0 + p)2 .

d
(6.39)
6.3. Exotic contributions to g 2 57
Again, using Gordon identity, we get,

1 1x
d4 k 0 (1 x)2
Z Z Z
2 2
F2 (q = 0) = 2i m dx dy
0 0 (2)4 (k 02 )3
2 1
(1 x)3
Z

= dx
(4)2 0 (1 x)2 + xm2a /m2
1
2 3 3x + x2
Z
1
' dx
(4)2 0 1 + xm2a /m2 m2a /m2
2 h
2 2 11
i
= log(m a /m ) 6
. (6.40)
(4)2 (m2a /m2 )
For order-of-magnitude estimation, its easy to see that m/ma & 105 is excluded.
Chapter 7
Radiative Corrections:
Some Formal Developments
7.1 Optical theorem in 4 theory

In this problem we check the optical theorem in phi-4 theory to order 2 . Firstly, the
total cross section tot at this order receives contributions from tree level only. The squared
amplitude is simply 2 . Then its easy to get the total cross section by complementing
kinematic factors. That is,
2
tot = , (7.1)
16s
2
where s = ECM and ECM is the COM energy. Then, consider the imaginary part of the
scattering amplitude. The contribution comes from 1-loop diagram in s-channel this time.
Lets evaluate this amplitude directly,
Z 1
dd k 2 dd k 0
Z Z
1 2 i i 1
iM = (i) d 2 2 2 2
= d
dx 02
2 (2) k m (k p) m 2 (2) 0 (k )2
1
i2 h 2
Z
2
i
= + log 4 dx log m x(1 x)s . (7.2)
2(4)2 0
Therefore,
1
2
Z h
2
i
ImM = dx Im log m x(1 x)s . (7.3)
2(4)2 0
The argument in the logarithm is real, thus the imaginary part of the logarithm equals to 0
or depending on the argument is positive or negative. (Strictly speaking the imaginary
part is but not due to our i prescription.) Then we see this logarithm contributes an
constant imaginary part , only when
p p
1 1 4m2 /s 1 + 1 4m2 /s
<x< .
2 2
Thus we have
2 p 2 pCM
ImM = 1 4m2 /s = . (7.4)
32 16ECM
59
60 Chapter 7. Radiative Corrections: Some Formal Developments
7.2 Alternative regulators in QED

In this problem we compute the first order corrections to Z1 and Z2 in QED, using cut-off
regularization and dimensional regularization respectively. By definition, we have,
(q = 0) = Z11 , (7.5)
d
Z21 = 1 . (7.6)
dp/ p/=m

Lets begin with dimensional regularization instead of momentum cut-off.
(b) Dimensional Regularization We firstly calculate F1 (0):
u(p0 ) (p, p0 )u(p)

dd k ig
Z
i i
=(ie) 2
d 2 2
u(p0 ) u(p)
(2) (k p) k/ + /q m k/ m
0
d u(p ) (k/ + /q + m) (k/ + m) u(p)
Z
2 d k
= ie
(2)d (k p)2 2 (k + q)2 m2 k 2 m2
Z 1 Z 1x
dd k 2u(p0 )N u(p)
Z
2
= ie dx dy , (7.7)
(2)d 0 0 (k 02 )3
in which we define
k 0 = k xp + yq,
= (1 x)m2 + x2 x(1 x)p2 y(1 y)q 2 + 2xyp q,
N = (k/ + /q + m) (k/ + m) .
The next step is to put N into the needed form. The calculation is basically in parallel
with Peskins Sec.6.3. Let me show some details. The first step is to finish the summation
over dummy Lorentz indices. Note that we are using dimensional regularization, thus we
should use Peskins Eq.(A.55). The result is:
h i
2
N = 2k (k + /q) + 4m(2k + q) (d 2)m + (4 d) (k + /q) k m(k + /q) m k .
/ / / / / /
Its worth noting here that d will be sent to 4 at the end of the calculation. Thus in the
square bracket in this expression, only the combination k/ k/ contributes to the final result.
Thus we simply have
N = 2k/ (k/ + /q) + 4m(2k + q) (d 2)m2 + (4 d)k/ k/. (7.8)
Here and following, we are free to drop off terms in N which contribute nothing to final
results. The equal sign should be understood in this way. The next step is to rewrite N in
terms of k 0 instead of k:
0 0
N = (2 d)k/ k/ 2 xp/ y /q xp/ + (1 y)/q + 4m 2xp + (1 2y)q 2m2 . (7.9)

7.2. Alternative regulators in QED 61
Terms linear in k 0 has been dropped since they integrate to zero. The third step is to put N
into a linear combination of , (p+p0 ) . Terms proportional to (pp0 ) will be dropped due
to Ward identity. The basic strategy is using substitution q = (p0 p) , on shell condition
u(p0 )p/0 = u(p0 )m and p/u(p) = mu(p). Here we show the detailed steps for the second term
above:
2 xp/ y /q xp/ + (1 y)/q

= 2 x2 p/ p/ y(1 y)/q /q xy /q p/ + x(1 y)p/ /q

= 2 x2 (2p p/)p/ y(1 y)(2q /q)/q xy /q p/ + x(1 y)(p/0 /q) /q

= 2 2x2 mp x2 m2 + y(1 y)q 2 2xymq + xm /q x(1 y)/q /q

= 2 x(x + 2)m2 + (x + y)(1 y)q 2 + 2x2 mp + 2xmp0 .

Combining this with other terms, and also make the momentum symmetrization (Peskins
Eq.(7.87)), which amounts to make the substitution:
0 0
k/ k/ 2
1 k 02 ,

d
we get
(2d)2 0 0
N = k/ k/ + 2(x2 + 2x 1)m2 2(x + y)(1 y)q 2 + 2x(1 x)m(p0 + p) . (7.10)

d
Now we employ Gordons identity

h p0 + p i q i
0 0
u(p ) u(p) = u(p ) + u(p),
2m 2m
to put N into a linear combination of and :
(2d)2
N = k 02 2(x2 4x + 1)m2 2(x + y)(1 y)q 2 2x(1 x)mi q . (7.11)

d
Now we have put the vertex into the following form:

i q
= F1 (q) + F2 (q). (7.12)
2m
We are interest in F1 (q), which is related to Z1 by Z1 = F1 (q = 0). Finishing
momentum integral:
Z 1 Z 1x Z
2 dd k 0 1 (2d)2 02 2 2

F1 (0) = 2ie dx dy d
k 2(x 4x + 1)m
0 0 (2)d (k 02 )3
2e2 1 1x h (2 d)2 (2 d ) d i
2 (3 2 )
Z Z
2 2
= dx dy + (x 4x + 1)m , (7.13)
(4)d/2 0 0 42d/2 3d/2
and sending d = 4 4:
1
2e2
Z h
2
F1 (0) = dx(1 x)
+ log 4
(4)2 0
(x2 4x + 1)m2 i
log (1 x)2 m2 + x2 2 +

, (7.14)
(1 x)2 m2 + x2
we reach the needed result Z1 = F1 (0). Now let us turn to Z2 . The correction of first

order is given by Z2 = (d/dp/)p/=m . Therefore wed better evaluate (p/) using dimensional
regularization.
dd k i i
Z
2
i(p/) =(ie) d

(2) k m (p k)2 2
/
Z 1
dd k (2 d)k/ + dm dd k 0 (2 d)xp/ + dm
Z Z
2 2
=e = e dx
d 2 2
(2) (k m ) (p k) 2 2 (2) 0d (k 02 )2
(2 d2 )
Z 1
ie2
= d/2
dx 2d/2
(2 d)xp/ + dm , (7.15)
(4) 0
where
k 0 = k xp;
= (1 x)m2 + x2 x(1 x)p2 .
Then we can compute
(2 d2 ) (2 d2 )(2 d2 ) d
Z 1 h
d(p/) e2 i
= dx (2 d)x (2 d)xp/ + dm
dp/ (4)d/2 0 2d/2 3d/2 dp/
(2 d2 ) (3 d2 )
Z 1 h
e2 i
= dx (2 d)x + 2x(1 x)p/ (2 d)x p
/ + dm .
(4)d/2 0 2d/2 3d/2
(7.16)
Then, setting p/ = m and d = 4 with 0, we get
d(p/) 2e2 1
Z h
2
= dx x
+ + log 4
dp/ p/=m (4)2 0

2(1 x)(2 x)m2 i

log (1 x)2 m2 + x2 1

(7.17)
(1 x)2 m2 + x2
Now its still not easy to see Z1 = Z2 immediately. To make this transparent, we need
some more steps. Lets focus on logarithm term:
Z 1 Z 1
2 2 2
dx(1 2x + x) log (1 x)2 m2 + x2

dx(1 x) log (1 x) m + x =
Z 10 h 0
(1 x)(1 x2 )m2 2 2 2
i
= dx (1 x) x log (1 x) m + x (7.18)
0 (1 x2 )m2 + x2
R R
Combining this with other terms, and also using the fact xdx = (1 x)dx, we get
Z 1
2e2 h
F1 (0) = 2
dxx 2 + log 4
(4) 0
2(1 x)(2 x)m2 i
log (1 x)2 m2 + x2 1

. (7.19)
(1 x)2 m2 + x2
Now its clear that Z1 = Z2 . Thus Z1 = Z2 keeps unaffected at this order when dimensional
regularization is used.
7.3. Radiative corrections in QED with Yukawa interaction 63
(a) Momentum Cut-off Now we repeat the calculation above using momentum cut-off.
Now we can directly borrow some results above. All we need to do is setting d 4 and
adding a UV momentum cut-off , as well as the following integral formulae:
Z 4
dk 1 i h 2
2
i
= log 1 +
2 +
,
(2)4 (k 2 )2 16 2
Z 4
dk k2 i h 2
(42 +3) 3
i
= log 1 + + 2(2 +)2 2 ,
(2)4 (k 2 )3 16 2
Z 4
dk 1 i 4
= .
(2)4 (k 2 )3 32 2 (2 + )2
We begin with (7.13):

Z 1 Z 1x Z 4 0
2 dk 1 02
k 2(x2 4x + 1)m2

F1 (0) = 2ie dx dy 4 02 3
0 0 (2) (k )
2 Z 1
2 (x2 4x + 1)m2

e 3
= 2 dx (1 x) log 1 + + , (7.20)
8 0 2
In the same way, we get
1
d4 k 0 xp/ 2m
Z Z
2
i(p/) = 2e dx , (7.21)
(2)4 0 (k 02 )2
and
d(p/) e2 1 2 2x(1 x)(x 2)m2
Z
= dx x log 1 + x+ . (7.22)
dp/ p/=m 8 2 0

This shows that Z1 6= Z2 with cut-off regularization.
7.3 Radiative corrections in QED with Yukawa inter-

action
(a) Let us calculate the first order corrections to Z1 and Z2 , as was done in Problem 7.2.
Firstly, we calculate , which is similar to the corresponding QED correction:
u(p0 ) (p, p0 )u(p)

2
2 Z dd k i 0 i i
=(i / 2) 2
u(p ) u(p)
(2)d (k p)2 m k/ + /q m k/ m
i2 1
Z 1x Z
dd k 2u(p0 )N u(p)
Z
= dx dy , (7.23)
2 0 0 (2)d (k 02 )3
where
k 0 = k xp + yq,
= (1 x)m2 + xm2 x(1 x)p2 y(1 y)q 2 + 2xyp q,

N = (k/ + /q + m) (k/ + m).
Then we put this correction into the following form, in parallel with steps of Problem 7.2.
That is: (1) replace k by k 0 in N :
0 0
N = k/ k/ + xp/ + (1 y)/q + m (xp/ y /q + m);

(2) rewrite the numerator N by gamma matrix relations and equations of motion, as
h i
N = d2 1 k 02 + (3 + 2x x2 )m2 + (y xy y 2 )q 2 + (x2 1)m(p0 + p) ;

(3) use Gordon identity to further transform N into:

h i i
N = d2 1 k 02 + (x + 1)2 m2 + (y y 2 xy)q 2 + 2m2 (1 x2 ).

2m
Then, we can read off F1 from the coefficient of , as:
Z 1 Z 1x Z
2 dd k 1 h
2
02 2 2
i
F1 (0) = i dx dy 1 k + (x + 1) m
0 0 (2)d (k 02 )3 d
1
2
Z
= dx (1 x) 2 + log 4
2(4)2 0
(x + 1)2 m2

2 2 2

log (1 x) m + xm 1 + . (7.24)
(1 x)2 m2 + xm2
Using the trick identity (7.18) again, we finally get

Z 1
2

F1 (0) = 2
dx x 2 + log 4
2(4) 0
2(1 x2 )m2

log (1 x)2 m2 + xm2 +

. (7.25)
(1 x)2 m2 + xm2
Now we calculate (p/).
2 Z dd k i i 2
Z
dd k 0
Z 1
xp/ + m
i(p/) = (i/ 2) d 2 2
= d
dx 02
(2) k/ m (p k) 2 (2) 0 (k )2
1
i2
Z h i
2
= dx
+ log 4 log (xp/ + m), (7.26)
2(4)2 0
where k 0 = k xp and = (1 x)m2 + x2 x(1 x)p2 . Then we have

Z 1
d(p/) 2 h
2
= dx x
+ log 4
dp/ p/=m 2(4)2 0

2(1 x2 )m2 i
log (1 x)2 m2 + xm2 +

. (7.27)
(1 x)2 m2 + xm2
Thus we have proved that Z1 = Z2 holds for 1-loop scalar corrections.

7.3. Radiative corrections in QED with Yukawa interaction 65
(b) Now consider the 1-loop corrections to Yukawa vertex. We focus on the divergent part
only. The equalities below should be understood to be hold up to a finite part. Then, for
vertex correction from photon, we have
dd k i
Z
0
2 i i
(p, p ) photon =(ie) d 2 2

(2) (k p) k + /q m k m
/ /
Z 1 Z 1x Z
2 dd k 2dk 02
= ie dx dy
0 0 (2)d (k 02 )3
(2 d2 )
Z 1 Z 1x
d2 e2 4e2 2
= dx dy = (7.28)
2(4)d/2 0 0 2d/2 (4)2
In the same way,

i 2 Z dd k i i i
0

(p, p ) scalar
= d 2 2 /
2 (2) (k p) m k + /q m k/ m
i2 1 1x
dd k 2k 02 2 2
Z Z Z
= dx dy = (7.29)
2 0 0 (2)d (k 02 )3 2(4)2
On the other hand, the 1-loop corrections for electrons self-energy also come from two
parts: one is the photon correction, which has been evaluated in Problem 7.1,
ie2 1
(2 d2 ) ie2 (p/ 4m) 2
Z

i(p/)photon = dx (2 d)x p
/ + dm = , (7.30)
(4)d/2 0 2d/2 (4)2
and the other is the scalar correction:

i 2 Z dd k i i i2 (p/ + 2m) 2
i(p/) scalar =
= , (7.31)
2 (2)d k/ m (p k)2 m2 4(4)2
To sum up, we have got the total vertex correction:
4e2 2 /2 2
(p, p0 ) = (p, p0 )photon + (p, p0 )scalar =

, (7.32)
(4)2
and also:

d(p/) d (p/)photon + (p/)scalar e2 + 2 /4 2
= = . (7.33)
dp/ p/=m dp/ (4)2

p
/=m
Final Project I
Radiation of Gluon Jets
In this final project we do a basic exercise about cancellation of infrared divergence. An

excellent pedagogical treatment of infrared divergence can be found in [2]
(a) First we calculate the 1-loop vertex correction to M(e+ e q q) from virtual gluon.
The amplitude is given by
dd k i i
Z
2 2 i i
i1 M = Qf (ie) (ig) u(p1 ) d
2 2
v(p2 ) 2 v(k2 ) u(k1 ).
(2) k/ k/ /q (k p1 ) q
(7.34)
Now we simplify the loop integral in the standard way, as was done in Problem 7.2. The
result is
(2d)2 02
Z 1 Z 1x
2
d
Z
d k 2 k 2(1 x)(x + y)q
i1 M = ig 2 dx dy d
iM0
(2)d 0 0 (k 02 )3
Z 1 Z 1x h
2g 2 (2 d)2 d (1 x)(x + y)q 2 d
i
= dx dy (2 2
) + (3 2
) iM0
(4)d/2 0 0 42d/2 3d/2
Z 1 Z 1x h
2g 2 2 (1 x)(x + y)q 2 i
= dx dy + log 4 log 2 + iM0 ,
(4)2 0 0
(7.35)
where
1
iM0 = Qf (ie)2 u(p1 ) v(p2 ) v(k2 ) u(k1 ) (7.36)
q2
is the tree amplitude, and
k 0 = k xq yp1 , = x(1 x y)q 2 y(1 y)p21 + y2 . (7.37)
With the external legs amputated, the result is,

1 1x
2g 2 y2 (1 x)(x + y)q 2 i
Z Z h
i1 M = dx dy log + 2 iM0 ,
(4)2 0 0 y2 x(1 x y)q 2 y x(1 x y)q 2
(7.38)
67
68 Final Project I. Radiation of Gluon Jets
Then the cross section is given by
42
(e+ e q q) = 3|F1 (q 2 = s)|2 , (7.39)
3s
with
Q2f g 1 1x
y2
Z Z h
2
F1 (q = s) = Q2f + dx dy log
2 0 0 y2 x(1 x y)s
(1 x)(x + y)s i
+ 2 . (7.40)
y x(1 x y)s
We will carry out the Feynman integration in (e).
(b) Now we simplify the 3-body phase space integral
d3 k1 d3 k2 d3 k3 (4)
Z Z
d3 = (q k1 k2 k3 ) (7.41)
(2)9 2E1 2E2 2E3
in the center of mass frame. It is convenient to introduce a new set of variables xi = 2ki q/q 2 ,
(i = 1, 2, 3). In COM frame, xi = 2Ei /Eq Then one can show that all Lorentz scalars
involving final states only can be represented in terms of xi and particles masses. In fact,
we only need to check (k1 + k2 )2 , (k2 + k3 )2 and (k3 + k1 )2 . From instance,
(k1 + k2 )2 = (q k3 )2 = q 2 + m23 2q k3 = s(1 x3 ) + m23 . (7.42)
Similarly,
(k2 + k3 )2 = s(1 x1 ) + m21 , (k3 + k1 )2 = s(1 x2 ) + m22 . (7.43)
To simply the phase integral, we first integrate out k3 with spatial delta function that
restricts k3 = k1 + k2 :
d3 k1 d3 k2
Z Z
d3 = (2)(Eq E1 E2 E3 ). (7.44)
(2)6 2E1 2E2 2E3
The integral measure can be rewritten as
d3 k1 d3 k2 = k12 k22 dk1 dk2 d1 d12 , (7.45)
where d1 is the spherical integral measure associated with d3 k1 , and d12 is the spherical
integral of relative angles between k1 and k2 . The former spherical integral can be directly
carried out and results in a factor 4. To finish the integral with d12 = d cos 12 d12 , we
make use of the remaining delta function, which can be rewritten as
E3 E 2 k12 k22 2
(Eq E1 E2 E3 ) = cos 12 3 , (7.46)
k1 k2 2k1 k2
p
by means of E3 = k12 + k22 + 2k1 k2 cos + 2 . Thus
8 2 E3
Z
d1 d12 (Eq E1 E2 E3 ) = .
k1 k2
Now using k1 dk1 = E1 dE1 and k2 dk2 = E2 dE2 , we have
dk1 dk2 k12 k22 8 2 E3

Z Z Z Z
1 s
d3 = = dE1 dE2 = dx1 dx2 . (7.47)
8(2)5 E1 E2 E3 k1 k2 32 3 128 3
To determine the integral region for m1 = m2 = 0 and m3 = , we note that there are two
extremal cases: k1 and k2 are parallel or antiparallel. In the former case, we have
p
Eq = E1 + E2 + E3 = E1 + E2 + (E1 + E2 )2 + 2 , (7.48)
which yields
2Eq (E1 + E2 ) = Eq2 2 , (7.49)
while in the latter case,
p
Eq = E1 + E2 + (E1 E2 )2 + 2 , (7.50)
which gives
(Eq 2E1 )(Eq 2E2 ) = 2 . (7.51)
These two boundary cases can be represented by xi variables as
2
x1 + x2 = 1 ; (7.52)
s
2
(1 x1 )(1 x2 ) = . (7.53)
s
The integral thus goes over the region bounded by these two curves.
(c) Now we calculate the differential cross section for the process e+ e qgg to lowest
order in and g . First, the amplitude is

2 i i i
iM = Qf (ie) (ig) (k3 )u(k1 ) v(k2 ) 2 v(p2 ) u(p1 ).
k/1 + k/3 k/2 + k/3 q
(7.54)
Then, the squared amplitude is
1 X 2
Q2f g 2 e4
|iM| = (g ) tr ( p/1 p/2 )
4 4s2
1 1 1 1

tr k/2 k/1
k/1 + k/3 k/2 + k/3 k/1 + k/3 k/2 + k/3
4Q2f g 2 e4

4(k1 k2 )(k1 k2 + q k3 )
= 8p1 p2
3s2 (k1 + k3 )2 (k2 + k3 )2

1 1
2

+ + 2(k1 k3 )(k2 k3 ) (k1 k2 ) . (7.55)
(k1 + k3 )4 (k2 + k3 )4
We have used the trick described in Peskins book (P261) when getting through the last
equal sign. Now rewrite the quantities of final-state kinematics in terms of xi , and set 0,
we obtain
2Q2f g 2 e4

1 X 2
2(1 x3 ) 1 x1 1 x2
|iM| = 8p1 p2 + +
4 3s2 (1 x1 )(1 x2 ) 1 x2 1 x1
2 2 4 2 2
8Qf g e x1 + x 2
= . (7.56)
3s (1 x1 )(1 x2 )
Thus the differential cross section, with 3 colors counted, reads
d 1 s 1P 2

= |M|
dx1 dx2 COM 2Ep1 2Ep2 |vp1 vp2 | 128 3 4

42 g x21 + x22
= 3Q2f , (7.57)
3s 2 (1 x1 )(1 x2 )
where we have used the fact that the initial electron and positron are massless, which implies

that 2Ep1 = 2Ep2 = s and |vp1 vp2 | = 2 in COM frame.
(d) Now we reevaluate the averaged squared amplitude, with kept nonzero in (7.55).
The result is
1 X 2
8Q2f g 2 e4
|iM| = F (x1 , x2 , 2 /s), (7.58)
4 3s
where
2 2
2
2(x1 + x2 1 + s )(1 + s )
F x1 , x2 , s =
(1 x1 )(1 x2 )
h 1 1 i
2

+ + (1 x1 )(1 x2 ) s . (7.59)
(1 x1 )2 (1 x2 )2
The cross section, then, can be got by integrating over dx1 dx2 :
Z
+ 1 s P
1 2

(e e q qg) = dx 1 dx 2 4 |M|
2Ep1 2Ep2 |vp1 vp2 | 128 3
Z 1 2 Z 1 t
42 2 g s s(1x1 ) 2

= 3Qf dx1 dx 2 F x 1 , x 2 , s
3s 2 0 1x1 s
2
42 g h 2 2 2 i
= 3Q2f log + 3 log + 5 13 2 + O(2 ) . (7.60)
3s 2 s s
(e) It is straightforward to finish the integration over Feynman parameters in (a), yielding
Q2f g 2
2 2

2

2 2 7 1 2 2
F1 (q = s) = Qf log + 3 log + 2 3 i 2 log + 7 + O( ) .
4 s s s
(7.61)
Then the cross section, to the order of g , is given by
42 g h 2 2 2
i
+ 2 7 1 2 2
(e e q q) = 3Qf 1 log + 3 log + 2
3
+ O( ) . (7.62)
3s 2 s s
(f ) Combining the results in (d) and (e), we reach the final result:
42

+ 2 3g
(e e q q + q qg) = 3Qf 1 + . (7.63)
3s 4
It is worth noting that all divergent terms as 0 cancel out in this expression.
Chapter 9
Functional Methods
9.1 Scalar QED

The Lagrangian for scalar QED reads,
L = 14 F F + (D ) (D ) m2 , (9.1)
with
F = A A , D = ( + ieA ). (9.2)
(a) Expanding the covariant derivative, its easy to find the corresponding Feynmans
rules:
(A -A - - interaction) = 2ie2 ,
(A - (p1 )-(p2 ) interaction) = ie(p1 p2 ) ,
with all momenta pointing inwards.

The propagators are standard. We will work in the Feynman gauge and set = 1, then
the propagator for photon is simply
i
,
p2 + i
and the propagator for scalar is
i
.
p m2 + i
2
(b) Now we calculate the spin-averaged differential cross section for the process e+ e
. The scattering amplitude is given by
i
iM = (ie)2 v(k2 ) u(k1 ) (p1 p2 ) . (9.3)
s
Then the spin-averaged and squared amplitude is
1 X e4
|M|2 = 2 tr (p/1 p/2 )k/1 (p/1 p/2 )k/2

4 spins 4s
73
74 Chapter 9. Functional Methods
e4 2

= 8(k1 p 1 k1 p 2 )(k2 p 1 k2 p 2 ) 4(k1 k 2 )(p 1 p 2 ) . (9.4)
4s2
We may parameterize the momenta as
k1 = (E, 0, 0, E), p1 = (E, p sin , 0, p cos ),

k2 = (E, 0, 0, E), p2 = (E, p sin , 0, p cos ),

with p = E 2 m2 . Then we have
1 X e 4 p2
|M|2 = 2
sin2 . (9.5)
4 spins 2E
Thus the differential cross section is:

d 1 p P
1 2
2 m2 3/2 2
= |M| = 1 sin . (9.6)
d CM 2(2E)2 8(2)2 E 4 8s E2
(c)
dd k dd k (p 2k) (p 2k)
Z Z
2 i 2
= 2ie d 2 2
(ie)
(2) k m (2) (k m2 )((p k)2 m2 )
d 2

dd k 2 (p k)2 m2 (p 2k) (p 2k)
Z
2
=e
(2)d (k 2 m2 ) (p k)2 m2
02
+ (1 2x)2 p p + 4k 0 k 0
2 2 2

Z d 0 Z 1 2 k + (1 x) p m
d k
= e2 dx
(2)d 0 (k 02 )2
2 k 02 (1 d2 ) + 2 (1 x)2 p2 m2 (1 2x)2 p p
1

dd k 0
Z Z
2
.=e dx
(2)d 0 (k 02 )2
ie2
Z 1 h (1 d )(1 d )2
2 2
= d/2
dx 2d/2
(4) 0
d
(2 2 ) 2 2 2
2
i
+ 2 (1 x) p m (1 2x) p p

2d/2
ie2 1
(2 d2 )
Z
2 (1 x)2 x(1 x) p2 (1 2x)2 p p .

= d/2
dx 2d/2
(9.7)
(4) 0
We can symmetrize the integrand as (1 x)2 12 (1 x)2 + x2 , then we get

ie2 1
(2 d2 )
Z
2 2

= dx (1 2x) (p p p . (9.8)
(4)d/2 0 2d/2
9.2 Statistical field theory

In this problem we study the path integral formulation in statistical mechanics. The
theory can be described by the partition function:
Z = tr eH , (9.9)
9.2. Statistical field theory 75
where H is the Hamiltonian of the system. It is a function of the generalized coordinates

q and the corresponding conjugate momentum p. In this problem, we simply assume the
Hamiltonian has the following form:
p2
H= + V (q). (9.10)
2m
We assume the dimension of the configuration space is d, then both q and p have d compo-
nents. Then we assume the eigenstates of both q and p form a complete orthonormal basis
of the Hilbert space:
dd p
Z Z
d
1 = d q |qihq|; 1= |pihp|. (9.11)
(2)d
Then the partition function can be written as
Z
H
Z = tr e = dd q hq|eH |qi. (9.12)
(a) Now we derive a path integral expression for the partition function. Following the
same way of deriving path integral in a quantum field theory, we separate the quantity eH
into N factors:
eH = eH eH , (N factors),
then inserting a complete basis between each pair of adjacent factors, as
Z
H
e = dd q1 dd qN 1 hq|eH |qN 1 ihqN 1 |eH |qN 2 i hq1 |eH |qi.
Now we focus on one factors:

1 2

2
hqi+1 |e H
|qi i = hqi+1 |e 2m
p +V (q)
|qi i = eV (qi ) hqi+1 |e 2m p |qi i,
and
dd pi+1 dd pi
Z
2 2
p
hqi+1 |e 2m |qi i = d d
hqi+1 |pi+1 ihpi+1 |ep /2m |pi ihpi |qi i
(2) (2)
Z d
d p ip(qi+1 qi ) p2 /2m m d/2 m(qi+1 qi )2 /2
= e e = 2 e .
(2)d
Inserting all this into the partition function, we get:

h m iN d/2 YN Z h m(q q )2 i
i+1 i
Z= dd qi exp V (qi ) , (9.13)
2 i=0
2
with qN +1 = q0 .
Now let N , then we have
Z h I i
Z = Dq exp d LE ( ) , (9.14)
where the integral measure is defined by

N
h m iN d/2 Y d
Dq = lim d qi , (9.15)
N 2(N ) i=0
and LE ( ) is a Lagrangian in Euclidean form:
m dq 2
LE ( ) = + V (q( )). (9.16)
2 d
Note that the periodic integral on comes from the trace in the partition function.
(b) Now we study an explicit example, a simple harmonic oscillator, which can by defined
by the Lagrangian
LE = 21 q 2 + 12 2 q 2 . (9.17)
Our task is to complete the path integral to find a expression for the partition function of
harmonic oscillator. This can be easily done by a Fourier transformation of the coordinates
q( ) with respect to . Since the time direction is periodic, the Fourier spectrum of q is
discrete. That is, X
q( ) = 1/2 e2in / qn , (9.18)
n
Then we have:
Z Z
1 X h 2i 2 i
d LE ( ) = d mn + 2 qm qn e2i(m+n) /
2 m,n
1 X h 2i 2 i 1 X h 2 2 2 i
= mn + 2 qm qn m,n = n + 2 qn qn
2 m,n 2 nZ
1 X h 2 2 2 i
= n + 2 |qn |2 . (9.19)
2 nZ
Then the path integral can be written as,

Z Z Y h 4 2 n2 i
2 q02 2 2
Z = C dq0 e dReqn dImqn exp + |qn |
n>0
2 2
C Y h 4 2 n2 2
i1 C Yh 1 2 i1
= + = 1 +
n>0 2 n>0 (n)2 2
X
= C sinh1 (/2) = C exp (n + 12 ) .

(9.20)
n0
(c) From now on we will consider the statistics of fields. We study the statistical properties
of boson system, fermion system, and photon system.
For a scalar field, the Lagrangian is given by,
Z
3 1h 2 2 2 2
i
LE ( ) = d x (, x) + (, x) + m (, x) . (9.21)
2
9.2. Statistical field theory 77
Following the method we used to deal with the simple harmonic oscillator, here we decom-
pose the scalar field (, x) into eigenmodes in momentum space:
d3 k ikx
X Z
1/2 2in /
(, x) = e 3
e n,k . (9.22)
n
(2)
Then the Lagrangian can also be rewritten in terms of modes, as,
Z Z X Z d3 kd3 k 0 1 2i 2
3 0 0 2
d LE ( ) = d d x nnk k+m
n,n0
(2)6 2
0 0
n,k n0 ,k0 ei2(n +n) /+i(k+k )x
d3 k 2 2 2
Z
1 X
= 3
n + k + m |n,k |2
2 2
2 n (2)
d3 k 1 2
Z X 2 2
2 2 2 2
= |0,k | + n + k |n,k | , (9.23)
(2)3 2 k n>0

where k2 = k2 + m2 . Then the partition function, as a path integral over the field configu-
rations can be represented by
Z Y
2 2 2 2 2
Z=C Ren,k Imn,k exp n + k |n,k | . (9.24)
n>0,k

By the calculation similar to that in (b), we get

Y Y 4 2 n2 1 h
2
Y
1
i
Z=C k + k = C exp k n + 2
. (9.25)
k n>0
2 k
1/2
This product gives the meaning to the formal expression det( 2 + m2 ) with proper
regularization.
(d) Then consider the fermionic oscillator. The action is given by,
Z Z
S = d LE ( ) = d ( )( ) + ( )( ) . (9.26)
The antiperiodic boundary condition ( + ) = ( ) is crucial to expanding the fermion

into modes: X
( ) = 1/2 e2i / n . (9.27)
nZ+1/2
Then the partition function can be evaluated to be

Z Y
X 2in
Z= dn dn n + n
n

nZ+1/2
1 2
Y 2in Y 4 2 (n + )
= C() + = C() 2
+ 2
n=0
2
nZ+1/2

1
= C() e/2 + e/2 ,

= C() cosh 2
(9.28)
with the form of a two-level system, as expected.
(e) Finally we consider the system of photons. The partition function is given by
Z Z
3 1 2 2

Z = DA DbDc exp d d x 2 A A b c
4(1/2)
det( 2 ),

= C() det() (9.29)
where the first determinant comes from the integral over the vector field A while the second
one comes from the integral over the ghost fields. Therefore,
2(1/2)
Z = C() det( 2 )

, (9.30)
which shows the contributions from the two physical polarizations of a photon. Here we see
the effect of the ghost fields of eliminating the additional two unphysical polarizations of a
vector field.
Chapter 10
Systematics of Renormalization
10.1 One-Loop structure of QED

(a) In this problem we show that any photon n-point amplitude with n an odd number
vanishes.
Now we evaluate explicitly the one-point photon amplitude and three-point photon am-
plitude at 1-loop level to check Furrys theorem. The one-point amplitude at 1-loop level is
simply given by,
dd k i tr [ (k/ + m)]
Z
(1)
i = (ie) = 0, (10.1)
(2)d k 2 m2
and the three-point amplitude consists of two diagrams,
dd k
Z h
(3) 3 i i i i
i = (ie) (1) tr
(2)d k/ m k/ + p/1 m k/ + p/1 + p/2 m

h
i i i i
+ tr . (10.2)
k/ + p/1 + p/2 m k/ + p/1 m k/ m
(b) Next we will show that the potential logarithmic divergences in photon four-point
diagrams cancel with each other. Since the divergence in this case does not depend on
external momenta, we will set all external momenta to be zero for simplicity. For the same
reason we will also set the fermions mass to be zero. Then the six diagrams contributing
the four-point amplitude can be evaluated as,
(Divergent part of i )
dd k 1 h
Z
= tr [ k/ k/ k/ k/] + tr [ k/ k/ k/ k/] + tr [ k/ k/ k/ k/]
(2)d (k 2 )4
i
+ tr [ k/ k/ k/ k/] + tr [ k/ k/ k/ k/] + tr [ k/ k/ k/ k/] . (10.3)
Now lets focus on the first trace, which can be worked out explicitly, to be
tr [ k/ k/ k/ k/] = 32k k k k 8k 2 k k g + k k g + k k g + k k g

79
80 Chapter 10. Systematics of Renormalization
+ 4(k 2 )2 (g g g g + g g ). (10.4)
Then, we symmetrize the momentum factors according to k k k 2 g /4 and k k k k
(k 2 )2 (g g + g g + g g )/24. (Since the divergence can be at most logarithmic, so it
is safe to set spacetime dimension d = 4 at this stage.) Then the first trace term reduces to,
4 2 2
tr [ k/ k/ k/ k/]
(k ) (g g 2g g + g g ). (10.5)
3
The other five terms can be easily got by permuting indices. Then it is straightforward to
see that the six terms sum to zero.
10.2 Renormalization of Yukawa theory

In this problem we study the pseudoscalar Yukawa Lagrangian,
L= 1
2
( )2 1
2
m2 2 + (i/ M ) ig 5 , (10.6)
where is a real scalar and is a Dirac Fermion.
(a) Lets figure out how the superficial degree of divergence D depends on the number of
external lines. From power counting, its easy to see that D can be represented by
D = 4L Pf 2Ps , (10.7)
where L is the no. of loops, Pf is the no. of internal fermion lines, and Ps is the no. of
internal scalar lines. We also note the following simple relations:
L = Pf + Ps V + 1 ,
2V = 2Pf + Nf ,
V = 2Ps + Ns .
Then we can deduce
3
D = 4L Pf 2Ps = 4(Pf + Ps V + 1) Pf 2Ps = 4 2
Nf Ns . (10.8)
Guided by this result, we can find all divergent amplitudes as follows.
D=2 D=1
D=0 D=0
We note that we have ignored the vacuum diagram, which simply contributes an infinitely
large constant, the potentially divergent diagrams with odd number of external scalars are
also ignored, since they actually vanish. This result shows that the original theory cannot
be renormalized unless we including a new 4 interaction, as
L = 4! 4 . (10.9)
10.2. Renormalization of Yukawa theory 81
(b) Now let us evaluate the divergent parts of all 1-loop diagrams of Yukawa theory. First
we consider the two point function of scalar. The one-loop contribution to this amplitude
is shown as follows.
+ +
The d = 4 pole of first two loop diagrams can be determined as
i dd k im2 1
Z
i
= . (10.10)
2 (2)d k 2 m2 (4)2
dd k 4ig 2 (p2 2M 2 ) 1
Z h i
2 5 i 5
i
= (ig) tr .
(2)d k/ M (k/ p/) M (4)2
(10.11)
From this we find the divergent part of the counterterm to be
(m2 8g 2 M 2 ) 1 4g 2 1
m , = . (10.12)
(4)2 (4)2
Then we come to the two point function of fermion, the 1-loop correction of which is given
by the following two diagrams.
From the pole of the loop diagram
dd k 5 i ig 2 (p/ 2M ) 1
Z
2 5 i
=g , (10.13)
(2)d k/ M (k p)2 m2 (4)2
we find the following counterterms:
2g 2 M 1 g 2 1
M , . (10.14)
(4)2 (4)2
The following two diagrams contribute to 1-loop corrections to Yukawa coupling and 4
coupling, respectively.
Since the divergent part of diagram is independent of external momenta, we can set all these
momenta to be zero. Then the loop diagram is
dd k 5 i g3 5 2
Z
3 5 i 5 i
=g (10.15)
(2)d k/ M k/ M k 2 m2 (4)2
(i)2 dd k i2 1
Z
i 2
= . (10.16)
2 (2)d k 2 m2 (4)2
dd k 8ig 4 1
Z
4
h
5 i 4 i
= (1)g tr . (10.17)
(2)d k/ M (4)2
Note that there are 3 permutations for the first diagram and 6 permutations for the second
diagram. Then we can determine the divergent part of counterterm to be
2g 3 1 32 48g 4 1
g , . (10.18)
(4)2 (4)2
10.3 Field-strength renormalization in 4 theory

In this problem we evaluate the two-loop corrections to scalars two-point function in 4
theory in the massless limit. There are three diagrams contribute in total.
The first diagram reads
(i)2 dd k dd q
Z
i i i
=
6 (2)d (2)d k 2 m2 q 2 m2 (p k q)2 m2
i2 dd kE dd qE
Z
i i i
= 2 2
6 (2) (2) kE + m qE + m (pE kE qE )2 + m2
d d 2 2
(2 d2 )
Z 1
i2 dd kE
Z
1 1
= dx
6 (2)d kE2 + m2 (4)d/2 0 [m2 + x(1 x)(pE kE )2 ]2d/2
i2 (2 d2 ) 1
Z
= dx dy
6(4)d/2 0
[x(1 x)]d/22 (1 y)1d/2 (3 d2 )/(2 d2 )
Z d
d kE
3d/2
(2)d (kE ypE )2 + y(1 y)p2 + 1 y + y

E x(1x)
m2
Z 1
i2 (3 d)[x(1 x)]d/22 (1 y)1d/2
= dx dy 3d . (10.19)
6(4)d 0 y

y(1 y)p2E + 1 y + x(1x) m2
10.4. Asymptotic behavior of diagrams in 4 theory 83
Now we take m2 = 0 and d = 4 4. Then we have

i2 2 1 i2 2 1
h
2
i
= (1 + )(p E ) + = p log(p ) +
12(4)4 12(4)4 E E
i2 2 1
h
2
i
= p log(p ) + . (10.20)
12(4)4
The second diagram actually vanishes in m 0 limit. In fact,
i dd k i (1 d2 )
Z
i
= = m 0. (10.21)
2 (2)d k 2 m2 2(4)d/2 m1d/2
The third diagram reads ip2 Z . Therefore we can choose the counterterm Z , under the M S
scheme, to be
2

1 2
Z = log M . (10.22)
12(4)4
Thus the field strength counterterm receives a nonzero contribution at this order. In the
massless limit, it is
i2 M2
2 (2) = p 2
log . (10.23)
12(4)4 p2
10.4 Asymptotic behavior of diagrams in 4 theory
In this problem we calculate the four point amplitude in 4 theory to 2-loop order in
s , t fixed, limit. The tree level result is simply i, and the 1-loop result can be easily
evaluated to be
(i)2 dd k
Z
i h i i i i
i1 M = + + i
2 (2)d k 2 m2 (ps k)2 m2 (pt k)2 m2 (pu k)2 m2
i2 h 2 i
' 3
+ log 4 log s log t log u i
2(4)2
i2 i2
= log s + log t + log u log s (10.24)
2(4)2 (4)2
In the last step we take the limit s . In this limit t can be ignored and u ' s. We
see the divergent part of the counterterm coefficient at 1-loop order is
32 1
. (10.25)
(4)2
Now we consider the two-loop correction.
2 2
(i)3 dd k i i3 1
(2 d2 )
Z Z
i
= = dx
4 (2)d k 2 (ps k)2 4(4)d 0 [x(1 x)s]2d/2
i3 1 1 1 2

log s + log s . (10.26)
(4)4 2 2
In the last line we only keep the divergent terms as 0 and s .
(i)3
Z d d
d kd q i i i i
=
2 (2) k (ps k) q (k p3 q)2
2d 2 2 2
(2 d2 )
Z 1
i3 dd k
Z
1 i
= dx
2 (2)d k 2 (ps k)2 (4)d/2 0 [x(1 x)(kE p3E )2 ]2d/2
(2 d2 )
Z 1
i3
Z d
d kE 1
= d/2
dx 2d/2 2
2(4) 0 [x(1 x)] (2) kE (psE kE ) [(kE p3E )2 ]2d/2
d 2
1
(2 d2 ) (4 d2 )
Z 1 Z 1y
i3 dd kE z 1d/2
Z Z
= dx dy dz
2(4)d/2 0 [x(1 x)]2d/2 (2)d 0 0 (kE2 + )4d/2 (2 d2 )
i3 z 1d/2 (4 d)
Z
= d
dxdydz 2d/2
, (10.27)
2(4) [x(1 x)] 4d
where = ys + zp23E (ypsE + zp3E )2 .

Then we find
i3 1 1 1 2

log s + log s . (10.28)
(4)4 2 2
The same result for the third diagram. Then we have
i3 3 3 3 2

+ + log s + log s . (10.29)
(4)4 2 2
Now we come to the counterterm. The fourth diagram reads
dd k i
Z
(i)(i ) i
=
2 (2)d k 2 (ps k)2
33 1 i 2 2 2

1 log s + log s +
2(4)2 (4)2 2 8
3
3i 1 1 1
log s + log2 s (10.30)
(4)4 2 2 8
The same result for the fifth diagram. Then we have
i3 3 3 2

+ + + + log s . (10.31)
(4)4 2 4
10.4. Asymptotic behavior of diagrams in 4 theory 85
So much for the s-channel. The t and u-channel results can be obtained by replacing s with
t and u respectively. In the limit s and t-fixed, we can simply ignore t and treating
u s, then the total 2-loop correction in this limit is
3i3
i2 M log2 s. (10.32)
2(4)4
The double pole 1/2 has been absorbed by .

In summary, we have the following asymptotic expression for the 4-point amplitude to
2-loop order in the s and t-fixed limit:
i2 3i3
iM = i log s log2 s + . (10.33)
(4)2 2(4)4
Chapter 11
Renormalization and Symmetry
11.1 Spin-wave theory

(a) Firstly we prove the following formula:

i(x) i(0)
Te e = e[D(x)D(0)] . (11.1)
Where D(x) = hT (x)(0)i is the time-ordered correlation of two scalars. The left hand
side of this equation can be represented by path integral, as
Z h Z
1 i(x) i(0)
i
D e e exp i dd xdd y 21 (x)D1 (x y)(y) . (11.2)
Z[0]
This expression precisely has the form Z[J]/Z[0], with J(y) = (y x) (0). Thus we
have
Z
1
dd xdd y J(x)D(x y)J(y) = exp D(x) D(0) ,

Z[J]/Z[0] = (11.3)
2
which is just the right hand side of the formula.
(b) The operator being translational invariant O[(x)] = O[(x) ] can depend on
only through . And the only relevant/marginal Lorentz-invariant operator satisfying
this condition is 21 ()2 .
(c) From now on we use bold x to denote coordinate and italic x to denote its length,
x |x|. We can use the result in (a) to evaluate hs(x)s (0)i, as
hs(x)s (0)i = A2 hei(x) ei(0) i = A2 eD(x)D(0) . (11.4)
Note that the correlation function

dd kE 1 ikE x
Z
1
D(x) = e (11.5)
(2)d kE2
87
88 Chapter 11. Renormalization and Symmetry
is the solution to the following equation:
2 D(x y) = (d) (x y). (11.6)
Since D(x) is a function of the length only, namely D(x) = D(x), thus we have

d1 (1 + d2 ) (x)
d1 x D(x) = . (11.7)
x x x d d/2 xd1
Then its easy to find
(1 + d2 )

1
, for d 6= 2,

d/2
d(d 2) x d2

D(x) = (11.8)
1

log x, for d = 2.
2
Then we have
Dimension d d=1 d=2 d=3 d=4

1 1 1 1
D(x) x log x
2 2 4x 4 x2
2
2
hss i e x
1/x 2
e1/x e1/x
Since 0 when d 2, the correlation function hss i in this case is independent of length
x.
11.2 A zeroth-order natural relation

We study N = 2 linear sigma model coupled to fermions:
L= 1
2
i i + 1
2
2 i i 1
4
(i i )2 + (i/ ) g (1 + i 5 2 ), (11.9)
with i a two-component field, i = 1, 2.
(a) Now, under the following transformation:

2 /2
1 1 cos 2 sin ; 2 1 sin + 2 cos ; ei , (11.10)
the first three terms involving i only keep invariant. The fourth term, as the kinetic term
of a chiral fermion, is also unaffected by this transformation. Thus, to show the whole
Lagrangian is invariant, we only need to check the last term, and this is really the case:
g (1 + i 5 2 )
5 5
g ei /2 (1 cos 2 sin ) + i 5 (1 sin + 2 cos ) ei /2

5 /2 5 5 /2
= g ei ei (1 + i 5 2 )ei = g (1 + i 5 2 ). (11.11)
11.2. A zeroth-order natural relation 89
p
(b) Now let acquire a vacuum expectation value v, which equals to 2 / classically.
Then, in terms of new variables = (v + (x), (x)), the Lagrangian reads
L= 1
2
( )2 + 1
2
( )2 2 2 1
4
( 4 + 4 )
1
2
2 2 v 3 v 2 + (i/ gv) g ( + i 5 ). (11.12)
That is, the fermion acquire a mass mf = gv.
(c) Now we calculate the radiative corrections to the mass relation mf = gv. The renor-
malization conditions we need are as follows.
p0
= g 5 at q 2 = 0, p2 = p02 = m2f . (11.13)
q p
= 0. (11.14)

These two conditions fixed g and v so that they receive no radiative corrections. Then we
want to show that the mass of the fermion mf receives finite radiative correction at 1-loop.
Since the tadpole diagrams of sum to zero by the renormalization condition above, the
fermions self-energy receive nonzero contributions from the following three diagrams:
The first two 1-loop diagrams can be evaluated as

Z 1
dd k dd k xp/ + mf
Z Z
2 i i 2
(e) = (ig) d 2 2
=g d
dx 02
(2) k/ mf (k p) 2 (2) 0 (k 1 )2
(2 d2 )
Z 1
ig 2
= dx 2d/2
(xp/ + mf )
(4)d/2 0 1
Z 1
ig 2 h
2
i
= dx (xp/ + mf ) + log 4 log 1 (11.15)
(4)2 0
Z 1
dd k 5 dd k xp/ mf
Z Z
2 i 5 i 2
(f) = g = g dx
(2)d k/ mf (k p)2 (2)d 0 (k 02 2 )2
ig 2 1
(2 d2 )
Z
= dx 2d/2
(xp/ mf )
(4)d/2 0 2
Z 1
ig 2 h
2
i
= dx (x p
/ m f )
+ log 4 log 2 . (11.16)
(4)2 0
This leads to
1
ig 2
Z
h i 2
(e) + (f) = dx 2xp/ 2 + log 4 1
2
log(1 2 ) + mf log (11.17)
(4)2 0 1
We see that the correction to the fermions mass mf from these two diagrams is finite. Be-
sides, the third diagram, namely the counterterm, contributes the mass correction through
g v. The the total correction to mf is finite only when g is finite. Let us check this by
means of the first renormalization condition (11.13) stated above. The 1-loop contributions
to (11.13) are as follows.
dd k
Z
2 i i i
(a) = (ig) g 5
(2) k/ mf k/ mf (k p)2 22
d
dd k (k/ + mf ) 5 (k/ + mf ) dd k
Z Z
3 3 5 1
= ig = ig
(2)d (k 2 m2f )2 (k p)2 22 (2)d (k 2 m2f ) (k p)2 22
Z 1 Z 1 d
dd k 0 g3 5 (2 )
Z
3 5 1 2
= ig dx 02 = dx 2d/2
(2)d 0 (k 1 )2 (4)d/2 0 1
g3 5 1 h 2
Z i
= dx + log 4 log 1 (11.18)
(4)2 0
dd k 5
Z
3 i i i
(b) = g d
5 5
(2) k/ mf k/ mf (k p)2 22
dd k 5 (k/ + mf ) 5 (k/ + mf ) 5 dd k
Z Z
3 3 5 1
= ig d 2 2 2 2 2
= ig d 2

(2) (k mf ) (k p) 2 (2) (k mf ) (k p)2 22
2
(2 d2 )
Z 1
dd k 0 g 3 5 1
Z Z
3 5 1
= ig dx = dx
(2)d 0 (k 02 2 )2 (4)d/2 0 2
2d/2
g 3 5 1 h 2
Z i
= dx
+ log 4 log 2 (11.19)
(4)2 0
dd k 5
Z
i i i
(c) = (ig)g(2iv) d

(2) k/ mf (k p) 2 (k p)2
2 2
Z 1 Z 1x
dd k 0 (x + y)p/ + mf
Z
2 5
= 4ig v dx dy
(2)d 0 0 (k 02 3 )3
2g 2 v 5 1 1x (x + y)p/ + mf
Z Z
= 2
dx dy (11.20)
(4) 0 0 3
11.3. The Gross-Neveu model 91
dd k
Z
i 5 i i
(d) = (ig)g(2iv)
(2)d k/ mf (k p)2 22 (k p)2
Z 1 Z 1x
dd k 0 (x + y)p/ + mf
Z
2 5
= 4ig v dx dy
(2) 0d
0 (k 02 3 )3
2g 2 v 5 1 1x (x + y)p/ + mf
Z Z
= 2
dx dy (11.21)
(4) 0 0 3
Thus,
g 5 1 1x m2f i
Z Z
h
2 2
(a) + (b) + (c) + (d) = dx g log + 4 dy . (11.22)
(4)2 0 1 0 3
11.3 The Gross-Neveu model

The Gross-Neveu Model is a theory of fermions in 1 + 1 dimensional spacetime:
L = i i/ i + 1
2
g 2 (i i )2 , (11.23)
with i = 1, , N . The gamma matrices are taken as 0 = 2 , 1 = i 1 , where i is the

familiar Pauli matrices. We also define 5 = 0 1 = 3 .
(a) The theory is invariant under the transformation i 5 i . It is straightforward to

check this. We note that:
i = i 0 i 5 0 = i 5 , (11.24)
thus:
L i 5 i/ 5 i + 1
2
g 2 (i 5 5 i )2
= i i/ i + 1
2
g 2 (i i )2 . (11.25)
However, a mass term will transform as mi i i mi i i , thus a theory respecting this

chiral symmetry does not allow such a mass term.
(b) The superficial renormalizability of the theory (by power counting) is obvious since
[g] = 0.
(c) The model can be phrased in another equivalent way:

Z Z
2 1 2

Z = DDD exp i d x i i/ i 2g2 i i . (11.26)
This can be justified by integrating out ,

Z Z Z
g2
D exp i d x ( 2g2 i i ) = N exp i d2 x
2 1 2
2
(i i ) . 2
(11.27)
which recovers the following path integral:

Z Z
Z = DD exp i d2 x i i/ i + 1 2 2

2
g (i i ) . (11.28)
(d) We can also integrate out the fermions i to get the effective potential for the auxiliary
field :
Z Z
N N
2
DD exp i d x i i/ i i i = det(i/ ) = det( 2 + 2 )

d2 k
Z
2 2
= exp N log(k + ) . (11.29)
(2)2
The integral is divergent, which should be regularized. We use the dimensional regulariza-
tion:
Z d Z d
d kE 2 2 d kE 1
N log(kE + ) = N
(2)d (2)d kE2 + 2 =0
(d/2)( 2 )d/2
= iN . (11.30)
(4)d/2
Now we set d = 2 and send 0,

Z d
iN 2 2

d kE 2 2 2
N log(kE + ) = + log 4 log + 1 . (11.31)
(2)d 4
Thus the effective potential is
2

1 2 N 2
Veff () = 2 + log 2 1 (11.32)
2g 4
by modified minimal subtraction.
(e) Now we minimize the effective potential:
Veff 1 N 2
0= = 2 + log 2 , (11.33)
g 2
2
and find nonzero vacuum expectation values hi = e/g N . The dependence of this result
on the renormalization condition is totally in the dependence on the subtraction point .
(f ) It is well-known that the loop expansion is equivalent to the expansion in powers

of ~ in generic perturbation theory around a classical vacuum. This is true because the
integrand of the partition function can be put into the form of eiS/~ . That is, ~ appears as
an overall coefficient of the action. In our case, we see that the overall factor N plays the
same role. Thus by the same argument, we conclude that the loop expansion is equivalent
to the expansion in powers of 1/N . More details can be found in Section III.3 of [4] and
Chapter 8 1/N of [5].
Chapter 12
The Renormalization Group
12.1 Beta Function in Yukawa Theory

In this problem we calculate the 1-loop beta functions in Yukawa theory. All needed
ingredients have been given in Problem 10.2 Here we list the needed counterterms:
g2

2 2
= log M ; (12.1)
2(4 2 )
2g 2

2 2
= log M ; (12.2)
(4 2 )
g3

2 2
g = log M ; (12.3)
(4)2
32 48g 4 2

2
= log M . (12.4)
2(4)2
Here is the UV cutoff and M is the renormalization scale. Then, the beta functions to
lowest order are given by
1
5g 3
g = M g + 2 g0 + g0 = ; (12.5)
M (4)2
32 + 8g 2 48g 4
= M + 20 = . (12.6)
M (4)2
12.2 Beta Function of the Gross-Neveu Model

We evaluate the function of the 2-dimensional Gross-Neveu model with the Lagrangian
L = i (i/ )i + 1
2
g 2 (i i )2 , (i = 1, , N ) (12.7)
to 1-loop order. The Feynman rules can be easily worked out to be

i
i j= ij
k k/
93
94 Chapter 12. The Renormalization Group
k l
= ig 2 (ij k` + i` jk )
i j
(2) (2)
Now consider the two-point function ij (p). The one-loop correction to ij (p) comes from
the following two diagrams:
i j + i j
R 1
It is easy to see the loop diagram contains a factor of d2 k tr [k/ ], which is zero un-
der dimensional regularization. Thus the wave function renormalization factor receives no
contribution at 1-loop level, namely = 0.
(4)
Then we turn to the 4-point function ijk` . There are three diagrams in total, namely,
k `
k ` ` k
0 0 n n
m n + +
0 0 m m
i j i j
i j
(a) (b) (c)
We calculate them in turn. The first one:
dd k
Z i
2 2
(a) = (ig ) (mn k` + n` mk )
0 0 0 0
(2)d k/ 0 0
i
(ij mn 0 0 + in jm 0 0 )
k/ 0 0
Z dd k 1
4 1
= g (2N + 2)ij k` + 2 i` jn ( ) ( ) (12.8)
(2)d k 2
The second diagram reads:
dd k
Z
1 i
(b) = (ig 2 )2 (
mj n` 0 0 + 0
m` nj 0 )
2 (2)d k/ 0 0
i
(im kn 0 0 + in km 0 0 )
k/ 0 0
4 Z dd k 1
g
= ij k` ( ) ( ) + i` jn ( ) ( ) (12.9)
2 (2)d k 2
The third diagram:
dd k
Z i
2 2
(c) = (ig ) ( mj nk 0 0 + mn jk 0 0 )
(2)d k/ 0 0
i
(im `n 0 0 + i` mn 0 0 )
k/ 0 0
12.3. Asymptotic Symmetry 95
4
Z dd k 1
=g 1
( ) ( )
2 ij k`
+ (2 2N )i` jk (12.10)
(2)d k 2
Summing up the three diagrams and using dimensional regularization with d = 2 , we

get
dd k 1
Z
4
2g (N 1)(ij k` + i` jk )
(2)d k 2
4
2(N 1)ig 2
(ij k` + i` jk ). (12.11)
4
Only the divergent terms are kept in the last expression, from which we can read the
counterterm
(N 1)g 4 2
g = log M 2 . (12.12)
2
Thus the function is
(N 1)(g 2 )2
(g 2 ) = M (g ) = , (12.13)
M
and
(N 1)g 3
(g) = . (12.14)
2
It is interesting to see that the 1-loop function vanishes for N = 1. This is because we
have the Fierz identity 2()() = ( )( ), and the Gross-Neveu model in this
case is equivalent to massless Thirring model, which is known to have vanishing function.
12.3 Asymptotic Symmetry

In this problem we study a bi-scalar model, given by the following Lagrangian:
2 2
1
( 1 )2 + ( 2 )2
(41 + 42 )

L= 2 4!
.
12 1 2
(12.15)
(a) First, we calculate the 1-loop beta functions and . The relevant 1-loop diagrams
for calculating are:
The relevant diagrams for calculating are:

Here the single line represents 1 and double line represents 2 . Since the divergent parts
of these diagrams are all independent of external momenta, we can therefore simply ignore
them. Then its easy to evaluate them, as follows.
(i)2 d4 k i i i2 2
Z
= , (12.16)
2 (2)4 k 2 k 2 2(4)2
(i/3)2 d4 k i i i2 2
Z
= . (12.17)
2 (2)4 k 2 k 2 18(4)2
The t-channel and u-channel give the same result. Thus we can determine to be
92 + 2 2
. (12.18)
6(4)2
On the other hand,
d4 k i i
Z
(i)(i/3) i 2
= = , (12.19)
2 (2)4 k 2 k 2 6(4)2
d4 k i i i2 2
Z
2
= = (i/3) . (12.20)
(2)4 k 2 k 2 9(4)2
Then we have
3 + 22 2
. (12.21)
3(4 2 )
Its easy to see that field strengths for both 1 and 2 receives no contributions from 1-loop
diagrams. Thus the 1-loop beta functions can be evaluated as
d 32 + 2 /3
= = ; (12.22)
d (4)2
d 2 + 42 /3
= = . (12.23)
d (4)2
(b) Now we derive the renormalization equation for /:

d 1 h 2
i
= 2 = (/) + 4(/) 3 . (12.24)
d 3(4)2
Then it is easy to see that / = 1 is an IR fixed point.
(c) In 4 dimensions, the functions for and are shifted as

32 + 2 /3
= + ; (12.25)
(4)2
2 + 42 /3
= + . (12.26)
(4)2
But it is easy to show that the terms containing cancel out in the function for /, and
the result is the same as (12.24). This is true because / still remains dimensionless in
4 dimensions. Therefore we conclude that there are three fixed points of the RG flow for
/ at 0, 1, and 3. We illustrate this in the diagram of RG flow in the - plane, with the
deviation of dimension = 0.01, in Figure 12.1.
12.3. Asymptotic Symmetry 97
1.0
0.8
0.6

0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0

Figure 12.1: The RG flow of the theory (12.15) in 4 dimensions with = 0.01. Three
nontrivial fixed points are shown by blue dots.
Chapter 13
Critical Exponents and Scalar Field

Theory
13.1 Correlation-to-scaling exponent

In this problem we consider the effect of the deviation of the coupling from its fixed
point to the two-point correlation function G(M, t) in d = 4 dimensions. Symbolically,
we can always write
G(M, t)
G(M, t) = G (M, t) + , (13.1)

=
where is the running coupling, defined to be the solution of the following renormalization
group equation:
d 2 ()
= . (13.2)
d log d 2 + 2()
As the first step, let us expand the function of around the fixed point, as

d()
() = ( ) + ( ) + O(( )2 )
d =

= ( ) + O(( )2 ). (13.3)
Then the renormalization group equation reads

d 2( )
' = ( ), (13.4)
d log d 2 + 2( )
where and on the right hand side are critical exponents, which in our case are defined
to be
d 2 + 2( ) 1
= , = .
d(2 ( )) 2 2 ( )
Dont confuse the critical exponent with the function. Now, from this equation we can
solve the running coupling to be
/
= + (0 ) . (13.5)
0
99
100 Chapter 13. Critical Exponents and Scalar Field Theory
Now let 0 be the scale at which the bare coupling is defined. Then we get
( )/ . (13.6)
13.2 The exponent

We have found the counterterm Z to O(2 ) with M S scheme in Problem 10.3, to be
2

1 2
Z = log M . (13.7)
12(4)4
Then the anomalous dimension to O(2 ) is given by
1 2
= M Z = . (13.8)
2 M 12(4)4
This result can be easily generalized to the O(N )-symmetric 4 theory, by replacing the
Feynman rule of the 4 coupling i with
2i( ij k` + ik j` + i` jk ),
which is equivalent to multiplying the diagram (10.19) by the following factor:
4 ik `m + i` km + im k` jk `m + j` km + jm k` = 12(N + 2) ij ,

(13.9)
and the anomalous dimension (13.8) obtained above should be multiplied by 12(N + 2),
which leads to
2
= (N + 2) , (13.10)
(4)4
which is the same as (13.47) of Peskin&Schroeder.
13.3 The CP N model

(a) The Lagrangian of the CP N model can be written as
2
1 X 2
X

L= 2 | zj | zj zj , (13.11)

g j j
with zj (j = 1, N + 1) the components of a vector in (N + 1) dimensional complex space,

subject to the constraint
X
|zj |2 = 1 (13.12)
j
and the identification

(ei z1 , , ei zN +1 ) (z1 , , zN +1 ). (13.13)
13.3. The CP N model 101
Now we prove that the Lagrangian given above is invariant under the following local trans-
formation:
zj (x) ei(x) zj (x), (13.14)
as,
2 2
g 2 L (ei zj ) + ei zj (ei zj )

= | zj |2 + | |2 + 2Re i( )zj zj

|zj zj |2 + | |2 + 2Re i( )zi zi zj zj
= g 2 L. (13.15)
Then we show that the nonlinear model with n = 3 is equivalent to the CP N model
with N = 1. To see this, we substitute ni = z i z into the Lagrangian of the nonlinear
sigma model, L = 2g12 | ni |2 , to get
1 i i
2
L = 2 ( z ) z + z z

2g
1 h i
= 2 i i 2( z )( z)z z + ( z)2 z 2 + ( z )2 z 2
2g
1 h 2 i
= 2 i i 2( z )( z) + z z + z z 2(z z)(z z )
2g
1 h i
= 2 i i 2( z )( z) + (z z)]2 2(z z)(z z ) .

(13.16)
2g
Then after a proper normalization of the field z, it is straightforward to see that the La-
grangian above reduces to
1
L = 2 | z|2 2|z z|2 , (13.17)
g
which is indeed the CP 1 model.
(b) The Lagrangian (13.11) can be obtained by the following Lagrangian with a gauge field
A and a Lagrange multiplier which expresses the local gauge symmetry and the constraint
explicitly:
1 2 2

L = 2 |D zj | |zj | 1 , (13.18)
g
with D = + iA . Now let us verify this by functionally integrating out the gauge field
A as well as the Lagrange multiplier to get
Z Z
2 i 2

2 2

Z = D zi DA D exp 2 d x |D zj | |zj | 1
g
Z Z
2 2
i 2 2
= D zi DA |zj | 1 exp 2 d x |D zj |
g
Z Z
2 2
i 2

2
= D zi DA |zj | 1 exp 2 d x A A + 2iA ( zj )zj + | zj |
g
Z Z 2
2 2
i 2

2
=N D zi |zj | 1 exp 2 d x | zj | |zj zj | . (13.19)
g
102 Chapter 13. Critical Exponents and Scalar Field Theory
(c) On the other hand one can also integrate out zi field in the Lagrangian (13.18), as
Z Z
i 2

2 2

Z = Dzi DA D exp 2 d x |D zj | |zj | 1
g
Z Z
2 i 2
= DA D exp N tr log(D ) + 2 d x (13.20)
g
We assume that the expectation values for A and are constants. Then the exponent can
be evaluated by means of dimensional regularization, as
Z
i
2
iS = N tr log(D ) + 2 d2 x
g
d
Z
d k i
log k + A A + 2 V (2)
2

= N d
(2) g
2

N M 1
i log + 1 ( A ) + 2 V (2) ,
2
(13.21)
4 A2 g
R
where V (2) = d2 x, and we have used the M S scheme to subtract the divergence. Now we
can minimize the quantity in the square bracket in the last line to get
2
4
A = 0, = M exp . (13.22)
gN 2
(d) The meaning of the effective action S is most easily seen from its diagrammatic rep-
resentations. For instance, at the 1-loop level, we know that the logarithmic terms in the
effective action is simply the sum of a series of 1-loop diagrams with n 0 external legs,
where the number of external legs n is simply the power of corresponding fields in the expan-
sion of S. Therefore, to the second order in A and in , the effective action is represented
precisely by the following set of diagrams,
where the dashed lines represent , curved lines represent A , and the internal loop are
z field. Then it is straightforward to see that the correct kinetic terms for and A are
generated from these diagrams. That is, the gauge field A becomes dynamical due to
quantum corrections. The gauge invariance of the resulted kinetic term F F can also be
justified by explicit calculation as was done in Problem 9.1.
Final Project II
The Coleman-Weinberg Potential
In this final project, we work out some properties of Coleman-Weinberg model, illustrat-
ing basic techniques of the renormalization group. The original paper [6] by S. Coleman
and E. Weinberg is always a good read, while a recent and very insightful treatment of the
model can be found in [7].
Simply put, the Coleman-Weinberg model is a theory of scalar electrodynamics, described
by the Lagrangian,
L = 12 F F + (D ) (D ) m2
6
( )2 , (13.23)
with a complex scalar and D = ( + ieA ).
(a) Consider the case of spontaneous breaking of the U (1) gauge symmetry (x)
ei(x) (x), caused by a negative squared mass, namely m2 = 2 < 0. The scalar then
p
acquires a nonzero vacuum expectation value (VEV) 0 = h||2 i. We split this VEV out
of the scalar field, namely,
1
= 0 + (x) + i(x) , (13.24)
2
p
with the new field (x) and (x) being real. At the tree level, it is easy to find 0 = 32 /

by minimize the scalar potential V () = 2 + 6 ( )2 . We also introduce v = 20
for convenience. Then, rewrite the Lagrangian in terms of these new field variables, we get,
L= 1
4
(F )2 + 12 ( )2 + 21 ( )2 + 21 e2 v 2 A A 12 (22 ) 2

24
( 4 + 4 + 2 2 2 + 4v 2 + 4v 3 ) + evA
+ eA ( ) + 12 e2 A A ( 2 + 2 + 2v). (13.25)
Then we see that the vector field A acquires a mass, equal to mA = ev at the classical
level.
(b) Now we calculate the 1-loop effective potential of the model. We know that 1-loop
correction of the effective Lagrangian is given by,
2L

i
L = log det + L, (13.26)
2 =0
103
104 Final Project II. The Coleman-Weinberg Potential
where is the fluctuating fields and L denotes counterterms.

Let the background value of the complex scalar be cl . By the assumption of Poincare
symmetry, cl must be a constant. For the same reason, the background value of the vector
field A must vanish. In addition, we can set cl to be real without loss of generality. Then
we have,
(x) = cl + 1 (x) + i2 (x),
where 1 (x), 2 (x), together with A (x), now serve as fluctuating fields. Expanding the
Lagrangian around the background fields and keeping terms quadratic in fluctuating fields
only, we get,
2
L = 12 F F + ( + ieA )(cl + 1 + i2 )
2 4
m2 cl + 1 + i2 6 cl + 1 + i2
= 21 A g ( 2 + 2e2 2cl ) A + 21 1 2 m2 2cl 1

+ 12 2 2 m2 3 2cl 2 2ecl A 2 + ,

(13.27)
where denotes terms other than being quadratic in fluctuating fields. Now we impose
the Landau gauge condition A = 0 to the Lagrangian, which removes the off-diagonal
term 2ecl A 2 . Then, according to (13.26), the 1-loop effective Lagrangian can be
evaluated as,
2L

i i
log det ( 2 + 2e2 2cl ) +

log det =
2 =0 2

2 2 2 2 2 2

+ log det + m + cl + log det + m + 3 cl
dd k
Z
i
= d
tr log(k 2 + 2e2 2cl )3
2 (2)

2 2 2 2 2 2
+ tr log(k + m + cl ) + tr log(k + m + 3 cl )
( d2 )

2 2 d/2 2 2 d/2 2 2 d/2
= 3(2e cl ) + (m + cl ) + (m + 3 cl ) . (13.28)
2(4)d/2
In the second equality we use the following identity,
det(I + AB) = n1 ( + BA), (13.29)
where A and B are matrices of n 1 and 1 n, respectively, is an arbitrary complex

number and I is the n n identity matrix. In our case, this gives,
det ( 2 + 2e2 2cl ) + = 2e2 2cl ( 2 + 2e2 2cl )3 .

(13.30)
Then the second equality follows up to an irrelevant constant term. The third equality
makes use of the trick in (11.72) of P&S. Then, for d = 4 and 0, we have,
2L

i 1 h
3(2e2 2cl )2 log(2e2 2cl )

log det = 2
2 =0 4(4)
+ (m2 + 2cl )2 log(m2 + 2cl )

2
i
+ (m +
3
2cl )2 2
log(m +
3
2cl ) , (13.31)
where we define 2 + log 4 + 32 for brevity.

Now, with M S scheme, we can determine the counterterms in (13.26) to be

1 2 2 2 2 2 2 2 2 2 2 2

L = + log 4 log M 3(2e cl ) + (m + cl ) + (m + 3 cl ) .
4(4)2
(13.32)
where M is the renormalization scale. Now the effective potential follows directly from
(13.26), (13.31) and (13.32),
M2

2 2 4 1 2 2 2
3
Veff [cl ] = m cl + cl 3(2e cl ) log +
6 4(4)2 2e2 2cl 2
2
M2

2 2 2
M 3
2 2 2
3
+ (m + cl ) log 2 + + (m + 3 cl ) log 2 2 + . (13.33)
m + 2cl 2 m + 3 cl 2
(c) Now taking the mass parameter 2 = m2 = 0, then the effective potential (13.33)
becomes
2e2 2cl 2cl

4 1 4 4
3 10 2 4 3
Veff [cl ] = cl + 12e cl log + cl log
6 4(4)2 M2 2 9 M2 2
4 4 2 2
3e cl 2e cl 3
' 4cl + 2
log 2
. (13.34)
6 (4) M 2
In the second line we use the fact that is of the order e4 to drop the 2 term. Then the
minimal point of this effective potential can be easily worked out to be,
M2 8 2
2cl = exp 1 . (13.35)
2e2 9e4
As e4 , we see that cl is of the same order with e1 M . Thus the effective potential
remains valid at this level of perturbation theory.
(d) We plot the effective potential as a function of cl in Figure (13.1). The purple curve
with m2 = 5 107 M 2 corresponds the case with no spontaneous symmetry breaking. The
blue curve shows that as m2 goes to 0 from above, new local minima is formed. Finally, the
orange and red curves correspond to broken symmetry, and in the case of the orange curve
with m2 = 0, the symmetry is dynamically broken.
(e) Now we calculate functions of the Coleman-Weinberg model to 1-loop level at high
energies, where we can send the mass parameter m2 to zero. It is convenient to work in the
Feynman gauge = 1. Then the relevant Feynman rules can be read from the Lagrangian
(13.25) to be,
0.02
0.00
Veff @cl DH10-2 M L

-0.02
-0.04
-0.06
0 1 2 3 4 5 6 7
-2
cl H10 M L
Figure 13.1: The effective potential Veff as a function of cl , with different values of m2 /M 2 =
5 107 , 2.4 107 , 0 and 1 107 from top to bottom, respectively.
i i i
= , = , = ,
k2 k2 k2
i
= i, = i, = ,
3
= 2ie2 , = 2ie2 = e(k1 k2 )
We first find the 1-loop wave function renormalization. For field, there is only one
diagram with nonzero contribution,
pk

which reads,
dd k i i 2ie2 p2 2
Z
2
e (p + k) (p k) . (13.36)
(2)d k 2 (p k)2 (4)2
Then we have,
2e2 2 2

= M , (13.37)
(4)2
and it is straightforward to see that = . For photons wave function renormalization

(vacuum polarization), we need to evaluate the following three diagrams,
The sum of the three diagrams is,
dd k i dd k i (p k)2
Z Z
2 i 1 2
e (2k p) (2k p) + 2 2ie
(2)d k 2 (p k)2 2 (2)d k 2 (p k)2
ie2 2 2
(p p p ), (13.38)
3(4)2
which gives,
e2 2 2

A = log M . (13.39)
3(4)2
Then we turn to the 1-loop corrections to couplings. For scalar self-coupling , we
consider the 1-loop corrections to 4 term in the Lagrangian. There are six types of diagrams
contributing, listed as follows, and we label them by (a) to (f) from left to right,
For each type there are several different permutations of internal lines giving identical result,
or more concretely, 3 permutations for each of the first three types, and 6 permutations for
each of the last three types. Now we evaluate them in turn. We set all external momenta
to zero to simplify the calculation. Then,
(i)2 dd k i 2 i2 2
Z
(a) = , (13.40)
2 (2)d k 2 2(4)2
(i/3)2 dd k i 2 i2 2
Z
(b) = , (13.41)
2 (2)d k 2 18(4)2
(2ie2 )2 dd k i 2 8ie4 2
Z

(c) = , (13.42)
2 (2)d k 2 (4)2
ie2 dd k i i 2 ie2 2
Z

(d) = (k k ) , (13.43)
3 (2)d k 2 k 2 3(4)2
dd k i i 2 2ie4 2
Z
2 2
(e) = (2ie )e (k k ) , (13.44)
(2)d k 2 k 2 (4)2
dd k i 2 i 2 ie4 2
Z
4 2
(f) = e (k k ) , (13.45)
(2)d k 2 k2 (4)2
Then multiplying (a)(c) by 3 and (d)(f) by 6, we find
52 /3 2e2 + 18e4 2
= . (13.46)
(4)2
Finally we consider the 1-loop corrections to e. For this purpose we calculate 1-loop diagrams
with three external lines with 1 A , 1 and 1 respectively, shown as follows, labeled again
by (a) to (d) from left to right,
Now we calculate them in turn.
dd k i
Z
3 i 2
(a) = e (2p k)2 (2k 2p)
(2)d k 2 (p k)2
Z 1
dd k 0 4x[(2 x)p k 0 ]2 [k 0 (1 x)p]
Z
3
= ie dx
(2)d 0 [k 02 + x(1 x)p2 ]3
1
dd k 0 k 02 (1 x)p 2(2 x)(p k 0 )k 0
Z Z
ie3 dx
(2)d 0 [k 02 + x(1 x)p2 ]3
dd k 0 1
[(1 x) d2 (2 x)]k 02 p
Z Z
= ie3 dx
(2)d 0 [k 02 + x(1 x)p2 ]3
2e3 2
p , (13.47)
(4)2
i Z dd k i 2
(b) = e d 2
2k = 0, (13.48)
3 (2) k
d
i 3e3 2
Z
d k i
(c) = (d) = e(2ie2 ) (k + p)
p . (13.49)
(2)d k 2 (p k)2 (4)2
Summing the four diagrams, we find that
2e3 2 2

e = log M . (13.50)
(4)2
Now we are ready to calculate functions,
1 e3
e = M e + (A + + ) = , (13.51)
M 2 48 2
52 18e2 + 54e4
= M + 2 = . (13.52)
M 24 2
The trajectory of renormalization group flows generated from these functions are shown
in Figure 13.2.
0.5
0.4
0.3
e2
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5

Figure 13.2: The renormalization group flow of Coleman-Weinberg model.
(f ) The effective potential obtained in (c) is not a solution to the renormalization group
equation, since it is only a first order result in perturbation expansion. However, it is possible
to find an effective potential as a solution to the RG equation, with the result in (c) serving
as a sort of initial condition. The effective potential obtained in this way is said to be RG
improved.
The Callan-Symansik equation for the effective potential reads

M + + e cl Veff (cl , , e; M ) = 0. (13.53)
M e cl
The solution to this equation is well known, that is, the dependence of the sliding energy
scale M is described totally by running parameters,
Veff (cl , , e; M ) = Veff cl (M 0 ), (M 0 ), e(M 0 ); M 0 ,

(13.54)
where barred quantities satisfy
e cl
M = (, e), M = e (, e), M = (, e)cl . (13.55)
M M M
The RG-improved effective potential should be such that when expanded in terms of coupling
constants and e, it will recover the result in (c) at the given order. For simplicity here
we work under the assumption that e4 , so that all terms of higher orders of coupling
constants than and e4 can be ignored. In this case, the perturbative calculation in (c)
gives
4 3e4 4cl 2e2 2cl 3

Veff = cl + log . (13.56)
6 (4)2 M2 2
Now we claim that the RG-improved edition of this result reads
4 3e4 4cl 2 3
Veff = cl + log 2e . (13.57)
6 (4)2 2
To see this, we firstly solve the renormalization group equations (13.55),
M0

0 4 9
(M ) = e + log , (13.58)
e4 4 2 M
e2
e2 (M 0 ) = , (13.59)
1 (e2 /24 2 ) log(M 0 /M )
M 0 2e2 /(4)2
cl (M 0 ) = cl , (13.60)
M
where the unbarred quantities , e and cl are evaluated at scale M . Now we substitute
these results back into the RG-improved effective potential (13.57) and expand in terms of
coupling constants. Then it is straightforward to see that the result recovers (13.56). To see
the spontaneous symmetry breaking still occurs, we note that the running coupling (M 0 )
flows to negative value rapidly for small M 0 = cl , while e(M 0 ) changes mildly along the
cl scale, as can be seen directly from Figure 13.2. Therefore the the coefficient before 4cl
is negative for small cl and positive for large cl . As a consequence, the minimum of this
effective potential should be away from cl = 0, namely the U (1) symmetry is spontaneously
broken.
To find the scalar mass m in this case (with = 0), we calculate the second derivative
of the effective potential Veff with respect to cl . Since the renormalization scale M can
be arbitrarily chosen, we set it to be M 2 = 2e2 h2cl i to simplify the calculation. Then the
vanishing of the first derivative of Veff at cl = hcl i implies that = 9e4 /8 2 . Insert this
back to Veff in (13.56), we find that
3e4 4cl 2cl

1
Veff = log 2 . (13.61)
16 2 hcl i 2
Then, taking the second derivative of this expression with respect to cl , we get the scalar
mass m2 = 3e4 h2cl i/4 2 = 3e4 v 2 /8 2 . Recall that the gauge bosons mass mA is given by
mA = e2 v 2 at the leading order, thus we conclude that m2 /m2A = 3e2 /8 2 at the leading
order in e2 .
(g) Now we consider the effect of finite mass, by adding a positive quadratic term into the
effective potential (13.56). For simplicity we still work with = 9e4 /8 2 , which is always
attainable without tuning. Then the effective potential reads,
3e4 4cl 2e2 2cl 1

Veff = m2r 2cl + log , (13.62)
(4)2 M2 2
in which the mass mr is not identical to the bare mass parameter appeared in the classical
Lagrangian, but has include 1-loop correction. As mr increases from zero, the energy of
symmetry breaking vacuum at hcl i = 6 0 also increases, until it reaches zero when m =
mc > 0, and has the same vacuum energy with the symmetric vacuum hcl i = 0. Then for
m > mc , there will be no stable symmetry breaking vacuum.
Using the effective potential above, we can find the position of the symmetry breaking

vacuum by solving the equation Veff /cl =v = 0, with the following solution,
cl
M2 (4)2 m2

r
2v = 2
exp W 2 2
, (13.63)
2e 3e M
in which W(z) is the Lambert W function, defined as the solution of z = W(z)eW(z) . Then
we can use the condition Veff (v ) = 0 to determine mc to be,
3e2 M 2
m2c = , (13.64)
32 2 e1/2
where the e in denominator is 2.718... and should not be confused with electric charge e.
00
Now we can evaluate the mass ratio m2 /m2A = 21 Veff (v )/(ev )2 , as a function of mass
parameter mr , to be,
3e2 (4)2 m2

2 2 r
m /mA = 2
1+W 2 2
. (13.65)
8 3e M
This is a monotonically decreasing function of mr , and when mr = 0, it recovers the previous
result 3e2 /8 2 . On the other hand, when mr = mc , the mass ratio m2 /m2A reaches its
3e2 1 2 2
minimum value, given by 8 2 [1 + W( 2e )] = 3e /16 , which is one half of the massless
case.
(h) When the spacetime dimension is shifted from 4 as d = 4 , the functions e and
are also shifted to be
e3 52 18e2 + 54e4
e = e + , = + . (13.66)
48 2 24 2
We plot the corresponding RG flow diagrams for several choice of in Figure 13.3, where
we also extrapolate the result to = 1.
0.5 0.5
0.4 0.4
0.3 0.3
e2
e2
0.2 0.2
0.1 0.1
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5

0.5 0.5
0.4 0.4
0.3 0.3
e2
e2
0.2 0.2
0.1 0.1
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5

Figure 13.3: The renormalization group flows of Coleman-Weinberg model in d = 4

spacetime dimensions, with = 0.005, 0.01, 0.1 and 1 in the upper-left, upper-right, lower-
left and lower-right diagram, respectively.
Chapter 15
Non-Abelian Gauge Invariance
15.1 Brute-force computations in SU (3)

(a) The dimension of SU (N ) group is d = N 2 1, when N = 3 we get d = 8.
(b) Its easy to see that t1 , t2 , t3 generate a SU (2) subgroup of SU (3). Thus we have
f ijk = ijk for i, j, k = 1, 2, 3. Just take another example, lets check [t6 , t7 ]:

[t6 , t7 ] = i( 21 t3 + 2
3 8
t ),
thus we get
f 678 = 2
3
, f 673 = 21 .
Then what about f 376 ?
[t3 , t7 ] = 2i t6 f 376 = 1
2
= f 673 .
1
(c) C(F ) = 2
. Here F represents fundamental representation.
4
(d) C2 (F ) = 3
, d(F ) = 3, d(G) = 8, thus we see that d(F )C2 (F ) = d(G)C(F ).
15.2 Adjoint representation of SU (2)

The structure constants for SU (2) is f abc = abc , thus we can write down the represen-
tation matrices for its generators directly from
(tbG )ac = if abc = iabc .
More explicitly,

0 0 0 0 0 i 0 i 0
t1G = 0 0 i , t2G = 0 0 0 , t3G = i 0 0 , (15.1)

0 i 0 i 0 0 0 0 0
113
114 Chapter 15. Non-Abelian Gauge Invariance
Then,
C(G) = tr (t1G t1G ) = tr (t2G t2G ) = tr (t3G t3G ) = 2,
C2 (G)I3 = t1G t1G + t2G t2G + t3G t3G = 2I3 C2 (G) = 2.
Here I3 is the 3 3 unit matrix.
15.3 Coulomb potential

(a) We calculate vacuum expectation value for Wilson loop UP (z, z), defined by
I

UP (z, z) = exp ie dx A (x) . (15.2)
P
By definition, we have
Z I

hUP (z, z)i = DA exp iS[A ] ie dx A (x) , (15.3)
P
where Z h i
S[A ] = d4 x 1
4
F F 1
2
( A )2 . (15.4)
Thus hUP (z, z)i is simply a Gaussian integral, and can be worked out directly, as
d4 k ig ik(xy)
I I Z
1

hUP (z, z)i = exp ie dx ie dy e (15.5)
2 P P (2)4 k 2 + i
Here we have set 0 to simplify the calculation. Working out the momentum integral,
we get
e2
I I
g
hUP (z, z)i = exp 2 dx dy . (15.6)
8 P P (x y)2
The momentum integration goes as follows
d4 k eik(xy) d kE eikE (xy)

Z Z 4
=i
(2)4 k 2 + i (2)4 kE2
Z 2 Z Z Z ikE |xy| cos
i 2 3 e
= d d sin d sin dk E kE
(2)4 0 0 0 0 kE2
Z Z
i
= 3 dkE kE d sin2 eikE |xy| cos
4 0 0
Z
i J1 kE |x y| i
= 2 dkE kE = 2 . (15.7)
4 0 kE |x y| 4 (x y)2
R
Where J1 (x) is Bessel function and we use the fact that 0
dx J1 (x) = 1.
15.4. Scalar propagator in a gauge theory 115
(b) Now taking a narrow rectangular Wilson loop P with width R in x1 direction (0 <
x1 < 1) and length T in x0 direction (0 < x0 < T ) and evaluate hUP i. When the integral
over dx and dy go independently over the loop, divergence will occur as |x y|2 0. But
what we want to show is the dependence of hUP i on the geometry of the loop, namely the
width R and length T , which should be divergence free. Therefore, when T R, the
integral in Wilson loop is mainly contributed by time direction and can be expressed as
T 0
2e2
Z Z
0 0 1
hUP (z, z)i ' exp dx dy 0 , (15.8)
8 2 0 T (x y 0 )2 R2 i
and we have add a small imaginary part to the denominator for the reason that will be clear.
Carry out the integration, we find
Z T Z 0
0 1 T R 2T
T i T
dx dy 0 0 0 2 2

arctanh = .
0 T (x y ) R i R R + i R
Therefore,
ie2
hUP i = exp T = eiV (R)T , (15.9)
4R
which gives the familiar result V (R) = e2 /4R.
(c) For the Wilson loop of a non-Abelian gauge group, we have

h I i

UP (z, z) = tr P exp ig dx Aa (x)tar , (15.10)
P
where tar is the matrices of the group generators in representation r. We expand this expres-
sion to the order of g 2 ,
I I
2
UP (z, z) = tr (1) g dx dy Aa (x)Ab (y) tr (tar tbr ) + O(g 3 )
P
IP I
2
= tr (1) 1 g C2 (r) dx dy A (x)A (y) + O(g 3 ).
a b
(15.11)
P P
Compared with the Abelian case, we see that to order g 2 , the non-Abelian result is given by
making the replacement e2 g 2 C2 (r). Therefore we conclude that V (R) = g 2 C2 (r)/4R
in non-Abelian case.
15.4 Scalar propagator in a gauge theory

In this problem we study very briefly the heat kernel representation of Green func-
tions/propagator of a scalar field living within a gauge field background.
(a) To begin with, we consider the simplest case, in which the background gauge field
vanishes. Then we can represent the Green function DF (x, y) of the Klein-Gordon equation,
defined to be
( 2 + m2 )DF (x, y) = i (4) (x y) (15.12)
with proper boundary conditions, by the following integral over the heat kernel function
D(x, y, T ): Z
DF (x, y) = dT D(x, y, T ). (15.13)
0
The heat kernel satisfies the following Schrodinger equation:

h i
i ( 2 + m2 ) D(x, y, T ) = i(T ) (4) (x y). (15.14)
T
The solution to this equation can be represented by
d4 k d4 k 0
Z
iHT
D(x, y, T ) = hx|e |yi = hx|kihk|eiHT |k 0 ihk 0 |yi
(2)4 (2)4
d4 k d4 k 0 i(k2 +m2 )T ikx+ik0 y
Z
= e e (2)4 (4) (k k 0 )
(2)4 (2)4
d4 k i(k2 m2 )T ik(xy)
Z
= e e , (15.15)
(2)4
with H = 2 + m2 . Integrating this result over T , with the +i prescription, we recover the
Feynman propagator for a scalar field:
Z
d4 k ik(xy)
Z Z
2 2
dT D(x, y, T ) = 4
e dT ei(k m +i)T
0 (2) 0
d4 k ieik(xy)
Z
= . (15.16)
(2)4 k 2 m2 + i
(b) Now let us turn on a background Abelian gauge field A (x). The corresponding
Schrodinger equation then becomes

2 2
i ieA (x) + m D(x, y, T ) = i(T ) (4) (x y), (15.17)
T
the solution of which, hx|eiHT |yi, can also be expressed as a path integral,
N
Z Y n
iHT

2 2
o
hx|e |yi = lim dxi xi exp it ieA (x) + m xi1 , (15.18)
N
i=1
where we have identify x = xN , y = x0 , and t = T /N . Then,

2 2
hxi |eit[( ieA (x)) +m ] |xi1 i
d4 ki
Z
2 2 2 2
= 4
hxi |eit[ ieA (x) +m ] |ki ihki |eit[ie A (x)e A (x)] |xi1 i
(2)
15.5. Casimir operator computations 117
d4 ki
Z
it[ki2 +eA (xi )ki +m2 ] it[eki A (xi1 )e2 A2 (xi1 )]
= hx i |e |ki ihk i |e |xi1 i
(2)4
d4 ki it[ki2 +eki (A(xi )+A(xi1 ))e2 A2 (xi1 )+m2 i] iki (xi xi1 )
Z
= e e
(2)4

it xi xi1 2
2 2 2
= C exp + eA(xi ) + eA(xi1 ) it(m e A (xi1 ))
4 t

it dx 2 dx 2
C exp iteA(x) itm . (15.19)
4 dt dt
In the last line we take the continuum limit, and C is an irrelevant normalization constant.
Then we get
Z Z T Z T
dx 2 2
D(x, y, T ) = Dx exp i dt + m ie dx(t) A(x(t)) . (15.20)
0 dt 0
15.5 Casimir operator computations

(a) In the language of angular momentum theory, we can take common eigenfunctions of
J 2 = a T a T a and Jz = T 3 to be the representation basis. Then the representation matrix
P
for T 3 is diagonal:
t3j = diag (j, j + 1, , j 1, j).
Thus
j
X
tr (t3j t3j ) = m2 = 1
3
j(j + 1)(2j + 1).
m=j
Then we have
X X
tr (t3r t3r ) = tr (t3ji t3ji ) = 1
3
ji (ji + 1)(2ji + 1) = C(r),
i i
which implies that X

3C(r) = ji (ji + 1)(2ji + 1). (15.21)
i
(b) Let the SU (2) subgroup be spanned by T 1 , T 1 and T 3 . Then in fundamental repre-
sentation, the representation matrices for SU (2) subgroup of SU (N ) can be taken as
!
i /2 0 2(N 2)
tiN = . (15.22)
0(N 2)2 0(N 2)(N 2)
Where i (i = 1, 2, 3) are Pauli matrices. We see that the representation matrices for SU (2)
decomposes into a doublet and (N 2) singlet. Then its easy to find that
1 1
( 12 + 1)(2 1 1

C(N ) = 3 2 2
+ 1) = 2
, (15.23)
by formula in (a).
In adjoint representation, the representation matrices (ti )ab = if aib (a, b = 1, , N 2

1, i = 1, 2, 3). Thus we need to know some information about structure constants. Here
we give a handwaving illustration by analyzing the structure of fundamental representation
matrices a little bit more. Note that therere three types of representation matrices, listed
as follows. For convenience, lets call them tA , tB and tC :
!
A22 02(N 2)
tA = . (15.24)
0(N 2)2 0(N 2)(N 2)
!
A22 B2(N 2)
tB = . (15.25)
B(N 2)2 0(N 2)(N 2)
!
21 tr (C)I22 02(N 2)
tC = . (15.26)
0(N 2)2 C(N 2)(N 2)
In which, tA is just the representation matrices for SU (2) subgroup. Thus we see that there
are 3 tA , 2(N 2) tB and (N 2)2 tC in total. Its also obvious that there is no way to generate
a tA from commutators between two tC or between a tB and tC , the only way to generate
tA are commutators between two tA or between to tB . Then, tA commutators correspond to
the triplet representation os SU (2) subgroup, and 2(N 2)-tB commutators correspond to
the doublet representation of SU (2). In this way we see that adjoint representation matrices
for SU (2) subgroup decompose into 1 triplet, 2(N 1) doublets and (N 2)2 singlets.
Then we can calculate C(G), again, by using formula in (a), as:
C(G) = 31 1(1 + 1)(2 1 + 1) + 2(N 2) 12 ( 12 + 1)(2 12 + 1) = N.

(15.27)
(c) Let U SU (N ) be N N unitary matrix, S be a symmetric N N matrix, and A

be an antisymmetric N N matrix. Then we can use S and A to build two representations
for SU (N ) respectively, as
S U SU T , A U AU T .
Its easy to verify that they are indeed representations. Lets denote these two representation
by s and a. Its also obvious to see that the dimensions of s and a are d(s) = N (N + 1)/2
and d(a) = N (N 1)/2 respectively.
Accordingly, the generator T a acts on S and A as:
S T a S + S(T a )T , A T a A + A(T a )T . (15.28)
To get C2 (s) and C2 (a), we can make use of the formula
d(r)C2 (r) = d(G)C(r). (15.29)
Thus we need to calculate C(r) and C(a). By formula in (a), we can take an generator in
SU (2) subgroup to simplify the calculation. Lets take
t3N = 1
2
diag(1, 1, 0, , 0),
15.5. Casimir operator computations 119
Then we have:

2S11 0 S13 S1n

S11 S1n 0 2S22 S23 S2n

.. .. .. 3 3 T 1
S= . . . tN S + S(tN ) = S31 S32
0 0
2 . . .. ... ..

Sn1 Snn .. .. . .
Sn1 Sn2 0 0

0 0 A13 A1n
0 A12 A1n
0 0 A23 A2n

. .. ..
A21 0 . 3 3 T 1
A=
..
tN A+A(tN ) = A31 A32 0 0
.. .. 2
. . .. .. .. .. ..

. An1,n .
. . . .
An1 An,n1 0
An1 An2 0 0
Thus we see that the representation matrices for T 3 , in both s representation and a repre-
sentation, are diagonal. They are:
t3s = diag(1, 0, 1
, 1 , 1, 1 , , 1 , 0, , 0 ); (15.30)
|2 {z 2} |2 {z 2} | {z }
N 2 N 2 (N 2)(N 1)/2
t3a = diag(0, 12 , , 21 , 0, , 0 ). (15.31)

| {z } | {z }
2(N 2) (N 2)(N 3)/2
Here we have rearrange the upper triangular elements of S and A by line.

Then we get
C(s) = tr (t3s )2 = 1
2
(N + 2); (15.32)
C(a) = tr (t3a )2 = 1
2
(N 2). (15.33)
Then,
d(G)C(s) (N 2 1)(N + 2)/2 (N 1)(N + 2)

C2 (s) = = = ; (15.34)
d(s) N (N + 1)/2 N
d(G)C(a) (N 2 1)(N 2)/2 (N + 1)(N 2)
C2 (a) = = = . (15.35)
d(a) N (N 1)/2 N
At last lets check the formula implied by (15.100) and (15.101):

X
C2 (r1 ) + C2 (r2 ) d(r1 )d(r2 ) = C2 (ri )d(ri ), (15.36)
in which the tensor product representation r1 r2 decomposes into a direct sum of irreducible
representations ri . In our case, the direct sum of representation s and a is equivalent to the
tensor product representation of two copies of N . That is,
N N
= s + a.
Thus, we have,
h N2 1 N2 1 i 2
N = N (N 2 1);

C2 (N ) + C2 (N ) d(N )d(N ) = +
2N 2N
and
C2 (s)d(s) + C2 (a)d(a) = C(s) + C(a) d(G) = N (N 2 1).

Thus formula (15.36) indeed holds in our case.

Chapter 16
Quantization of Non-Abelian Gauge

Theories
16.1 Arnowitt-Fickler gauge

In this problem we perform the Faddeev-Popov quantization of Yang-Mills theory in
Arnowitt-Fickler gauge (also called axial gauge), namely A3a = 0. More generally, we may
write the gauge condition as n Aa = 0 with n an arbitrary space-like vector of unit norm
(n2 = 1). The condition A3a = 0 corresponds simply to the choice n = g 3 . This gauge
has the advantage that the Faddeev-Popov ghosts do not propagate and do not couple to
gauge fields, as we will show below.
Our starting point, the partition function, reads
Z (n Aa )

Z = DA (n Aa )eiS[A ] det , (16.1)
b
with S = 14 d4 x (F
R a 2
) the classical action for the gauge field, and the Faddeev-Popov
determinant is given by
(n Aa ) 1
ab abc c
det = det n f n A
b g
Z Z
4 a ab abc c b

= DbDc exp i d x b n f n A c . (16.2)
When multiplied by the delta function (n Aa ), the second term in the exponent above
vanishes, which implies that the ghost and antighost do not interact with gauge field. Mean-
while, they do not propagator either, since there does not exist a canonical kinetic term for
them. Therefore we can safely treat the Faddeev-Popov determinant as an overall normal-
ization of the partition function and ignore it. Then, the partition function reduces to
Z Z 1
4 a 2 1 a 2
Z = lim DA exp i d x (F ) (n A )
0 4 2
Z Z
= lim DA exp i d4 x 12 Aa g 2 1 n n Ab

0
121
122 Chapter 16. Quantization of Non-Abelian Gauge Theories

abc
gf ( Aa )Ab Ac 1
4
2 eab ecd
g f f Aa Ab Ac Ad . (16.3)
where we have convert the delta function (n Aa ) into a limit of Gaussian function. Then
we see that the three-point or four-point gauge boson vertices share the same Feynman rules
with the ones in covariant gauge. The only difference arises from the propagator. Let us
parameterize the propagator in momentum space as
D (k) = Ak 2 g + Bk k + C(k n + n k ) + Dn n . (16.4)
Then, the equation of motion satisfied by the propagator,
g k 2 k k 1
n n D (k) = g

(16.5)
gives
i 1 + k 2 1
A= , B= C, C= A, D = 0. (16.6)
k2 kn kn
Note that the gauge fixing parameter should be sent to 0. Therefore the propagator reads
kk k n + k n

i
D(k) = 2 g + . (16.7)
k (k n)2 kn
16.2 Scalar field with non-Abelian charge

(a) Firstly we write down the Lagrangian for the Yang-Mills theory with charged scalar
field, as
1 a 2
L = (F ) + (D ) (D ), (16.8)
4

where the covariant derivative D = + igAa tar with tar the matrices of gauge group
generators in representation r. For simplicity we ignore the possible mass term for the scalar.
Then, it is straightforward to derive the Feynman rules for this theory by expanding this La-
grangian. The rules for the propagator and self-interactions of gauge boson are independent
of matter content and are the same with the ones given in Figure 16.1 in Peskin&Schroeder.
The only new ingredients here are the gauge boson-scalar field interactions, which generate
the following Feynman rules:
, a , b
= ig 2 (tar tbr + tbr tar )g ,
, a
p1 % - p=
2
igtar (p1 p2 ) ,
16.3. Counterterm relations 123
(b) To compute the function of coupling g, we introduce some additional Feynman rules
involving counterterms:
a
= ig1 tar (p1 p2 ) ,

p1 % - p2
= ip2 2 im ,
p
a b
= i ab 3 (p2 g p p ).
p
Then the function is given by

1
= gM 1 + 2 + 3 . (16.9)
M 2
To determine the counterterms, we evaluate the following relevant 1-loop diagrams. But the
calculations can be simplified a lot if we observe that the combination 1 2 is determined
by pure gauge sector, and is independent of matter content. This may be most easily seen
from the counterterm relation 1 2 = 1c 2c , where the right hand side comes from ghost
contribution which is a pure gauge quantity. We will demonstrate this counterterm relation
explicitly in the next problem for fermionic matter. Therefore, we can borrow directly the
result of Peskin & Schroeder, or from the result of Problem 16.3(a),
g2 2
2

1 2 = C 2 (G) log M . (16.10)
(4)2
On the other hand, 3 can be found by evaluating the loop corrections to the gauge bosons
self-energy. The contributions from the gauge boson loop and ghost loop have already been
given in eq.(16.71) in Peskin&Schroeder, while the rest of the contributions is from the
scalar-loop, and is simply the result we have found in Problem 9.1(c), multiplied by the
gauge factor tr (tar tbr ) = C(r) and the number of scalar ns . Combining these two parts gives
the divergent part of 3 :
g2 5 1 2 2

3 = C2 (G) ns C(r) log M . (16.11)
(4)2 3 3
Then it is straightforward to see that
g 3 11 1
= C 2 (G) n s C(r) . (16.12)
(4)2 3 3
16.3 Counterterm relations

In this problem we calculate the divergent parts of counterterms in Yang-Mills theory
with Dirac spinors at 1-loop level, to verify the counterterm relations, which is a set of
constraints set by gauge invariance. To begin with, let us rewrite the Lagrangian in its
renormalized form, with counterterms separated, as
L= 1
4
( Aa Aa )2 + (i/ m) ca 2 ca
+ gAa ta gf abc ( Aa )Ab Ac
1 2
4
g (f eab Aa Ab )(f ecd Ac Ad ) gca f abc (Ab cc )
1
( Aa Aa )2 + (i2 / m ) 2c ca 2 ca
4 3
+ g1 Aa g13g f abc ( Aa )Ab Ac
1 4g
(f eab Aa Ab )(f ecd Ac Ad ) g1c ca f abc (Ab cc ).
4 1
(16.13)
Then the counterterm relations we will verify are
1 2 = 13g 3 = 1
2
(14g 3 ) = 1c 2c . (16.14)
Note that 1 and 2 have been given in (16.84) and (16.77) in Peskin&Schroeder. Here we
simply quote the results:
g2 2 2

1 = C 2 (r) + C 2 (G) log M , (16.15)
(4)2
g2 2
2

2 = C2 (r) log M . (16.16)
(4)2
Therefore,
g2 2
2

1 2 = C 2 (G) log M . (16.17)
(4)2
(a) Firstly let us check the equality 1 2 = 1c 2c . The 1-loop contributions to 1c come
from the following three diagrams:
b
p1 p2
k
c a
The first diagram reads
dd k i
Z
3 ade ebf f dc i i
(g) f f f p (p2 k) (p1 k)
(2) k (p1 k) (p2 k)2 2
d 2 2
dd k k (p2 k) dd k 1
Z Z
3 ade ebf f dc 1 3 ade ebf f dc
g f f f = g f f f p 2
(2)d k6 4 (2)d k 4
i i 2
g 3 f ade f ebf f f dc p2 . (16.18)
4 (4)2
The second diagram reads
dd k i i i
Z
2 ade ebf f dc
(g) gf f f
(2) k (p1 k) (p2 k)2
d 2 2
p2 k g (k p2 q) + g (q p1 + k) + g (p1 + p2 2k)

dd k 1 2
Z
3 ade ebf f dc
p2 k + k (k p2 ) 2k (k p2 )

ig f f f d 6
(2) k
dd k 1
Z
3 3 ade ebf f dc 3 3 ade ebf f dc i 2
ig f f f p2 ig f f f p 2 . (16.19)
4 (2)d k 4 4 (4)2
To simplify the structure constant product, we make use of the Jacobi identity,
0 = f ebf (f abd f dce + f bcd f dae + f cad f dbe ) = 2f abd f dce f ebf f caf C2 (G),
then we have
1
f ade f ebf f f dc = f abc C2 (G). (16.20)
2
Note that the third diagram reads g1c f abc p2 , thus we see that to make the sum of these
three diagrams finite, the counterterm coefficient 1c should be
g 2 C2 (G) 2
1c log M 2
. (16.21)
2(4)2
Then consider 2c . This coefficient should absorb the divergence from the following dia-
gram:
a b
p k
This diagram reads
dd k i i
Z
2 bcd dca
(g) f f (p k)
(2) k (p k)2
d 2
Z 1
dd k 0 p (k 0 + xp)
Z
2 ab
= g C2 (G) dx
(2)d 0 (k 02 )2
g2 i 2
C2 (G) ab p2 + terms indep. of p2 . (16.22)
2 (4)2
The corresponding counterterm contributes i2c p2 , therefore we have
g 2 C2 (G) 2
2c log M 2
. (16.23)
2(4)2
Combining (16.21), (16.23) and (16.17), we see that the equality 1 2 = 1c 2c is satisfied.
(b) Now lets verify the equality 1 2 = 13g 3 . In this case the calculation turns
out to be more cumbersome, though. The coefficient 3 has been given by (16.74) in Pe-
skin&Schroeder. The result is
g2 h 5 4 i 2
2

3 = C 2 (G) n f C(r) log M . (16.24)
(4)2 3 3
Thus we only need to calculate 13g . The relevant loop diagrams are listed as follows.
a
p
b .p c
For simplicity, we have set the external momenta to be p, p and 0 for the three external
gauge boson lines labeled with (a), (b) and (c). Then the contribution of the counterterm
to this vertex is given by
g13g f abc (2g p g p g p ). (16.25)
To extract the divergent part from 13g , we have to evaluate the loop diagrams shown above.
Let us calculate them now in turn. The first diagram reads
1
g(ig 2 )f ade f def f bcf (g g g g )

= 2
+ f dbf f ecf (g g g g ) + f dcf f ebf (g g g g )

dd k i i
Z

g (p + k) + g (2k + p) + g (k 2p)
(2)d k 2 (k p)2
= ig 3 f abc C2 (G) 23 (g g g g )
1
2
dd k 1
Z
1
g (p + k) + g (2k + p) + g (k 2p)

d 2 2
(2) k (k p)
dd k 1
Z
1 3 abc 9
2 ig f C2 (G) 2 (g p g p )
(2)d k 4
9 i 2
ig 3 f abc C2 (G)(g p g p ) . (16.26)
4 (4)2
There are two additional diagrams associated with this diagram by the two cyclic permuta-
tions of the three external momenta. One gives
9 3 abc i 2
ig f C2 (G)(g p g p ) ,
4 (4)2
while the other yields zero. Therefore the sum of these three diagrams gives:
9 3 abc i 2
ig f C2 (G)(2g p g p g p ) , (16.27)
4 (4)2
Then we come to the second diagram, which reads
dd k i i i
Z
3 adf bed cf e
=g f f f
(2) k k (p + k)2
d 2 2
g (2p + k) + g (p 2k) + g (k p)

g (p + k) + g (2k p) + g (2p + k)

g k 2g k + g k

Z 1
dd k 0 2(1 x)k 02
Z
3
1 abc
ig 2 f C2 (G) dx
(2)d 0 (k 02 )3
1h i
2(8 + 15x)g p + (30x 23) g p + g p
4
dd k 0 1
Z
3
1 abc
13

ig 2 f C2 (G) 2g p g p g p
4 (2)d k 04
13 3 abc i 2
ig f C2 (G) 2g p g p g p

. (16.28)
8 (4)2
The third diagram reads
dd k
Z i 2
3 daf ebd f ce i
= (g) f f f d
(1) 2 2
(p + k) k k
(2) k (k + p)
Z 1
dd k 0 2(1 x)k 02
Z
3 1 abc
= ig 2 f C2 (G) dx
(2)d 0 (k 02 )3
d1 xg p xg p + (1 x)g p

1 3 abc i 2
ig f C2 (G)(g p + g p 2g p ) . (16.29)
24 (4)2
There is again a similar diagram with ghost loop running reversely, which gives
1 3 abc i 2
ig f C2 (G)(g p 2g p + g p ) .
24 (4)2
Then these two diagrams with ghost loops sum to
1 3 abc i 2
ig f C2 (G)(2g p g p g p ) . (16.30)
24 (4)2
Finally we consider the fourth diagram with fermion loop. There are also two copies with
fermions running in opposite directions. One (shown in the figure) gives
dd k
Z
3 a c b i i i
= nf (ig) tr (t t t ) (1) tr
(2)d k/ k/ k/ + p/
4 i 2
nf g 3 tr (ta tc tb )(2g p g p g p ) , (16.31)
3 (4)2
while the other gives
4 i 2
nf g 3 tr (ta tb tc )(2g p g p g p ) .
3 (4)2
Thus they sum to

4 i 2
nf g 3 tr ta [tc , tb ] (2g p g p g p )

3 (4)2
4 i 2
= inf g 3 C(r)f abc (2g p g p g p ) . (16.32)
3 (4)2
Now, sum up the four groups of diagrams, we get
g 3 2 abc 9

13 1 4
2
f + C2 (G) + nf C(r) (2g p g p g p ), (16.33)
(4) 4 8 24 3
and consequently,
g2 h 2 4 i 2
13g = C 2 (G) n f C(r) log M 2
. (16.34)
(4)2 3 3
Thus,
g2 2
13g 3 = C 2 (G) log M 2
, (16.35)
(4)2
which equals to 1 2 (16.17), as expected.
(c) Now lets move to the relation 1 2 = 21 (14g 3 ). This time we have to evaluate
14g , which is determined by the divergent part of the following five types of diagrams:
a b
c d
Firstly the counterterm itself contributes to the 1-loop corrections with
14g f abe f cde (g g g g ) + f ace f bde (g g g g )

+ f ade f bce (g g g g ) .

(16.36)
To evaluate the loop diagrams, we set all external momenta to zero for simplicity. The first
diagram then reads
1 h
= (ig ) f abg f ef g (g g g g ) + f aeg f bf g (g g g g )
2 2
2 ih
+ f af g f beg (g g g g ) f ef h f cdh (g g g g )
i Z dd k i 2
ech f dh edh f ch
+ f f (g g g g ) + f f (g g g g )
(2)d k 2
ig 4 2 h abg ef g ef h cdh
2
f f f f (2g g 2g g )
2(4)
+ f abg f ef g f ech f f dh (g g g g ) + f abg f ef g f edh f f ch (g g g g )

+ f aeg f bf g f ef h f cdh (g g g g ) + f af g f beg f ef h f cdh (g g g g )
+ tr (ta tb td tc )(2g g + g g ) + tr (ta tb tc td )(2g g + g g )
i
+ tr (ta tb tc td )(2g g + g g ) + tr (ta tb td tc )(2g g + g g )
ig 4 2 h
= 2
tr (ta tb tc td )(4g g 8g g + 10g g )
2(4)
i
+ tr (ta tb td tc )(4g g + 10g g 8g g ) , (16.37)
where we have used (16.20) and f abc = i(ta )bc with ta the generators in adjoint representation.
There are two additional diagrams similar to this one, which can be obtained by exchange
of labels as (b c) and (b d). Therefore the total contribution from these three
diagrams is
ig 4 2 h
(4)2
+ tr (ta tb td tc )(7g g + 7g g 8g g )
i
a c b d
+ tr (t t t t )(8g g + 7g g + 7g g ) . (16.38)
The second diagram has five additional counterparts. The one displayed in the figure
reads
= (ig 2 )g 2 f aeg f bgf f ef h f cdh (g g g g )

+ f ech f f dh (g g g g ) + f edh f f ch (g g g g )

dd k i 3
Z
d 2
(g k 2k + g k )(g k 2g k + g k )
(2) k
4 aeg bgf
ef h cdh
=g f f f f (g g g g )
+ f ech f f dh (g g g g ) + f edh f f ch (g g g g )

dd k 1 2
Z

2g g
+ 5g g 4g g

(2)d k 2
g 4 i 2 h aeg bgf ef h cdh
f f f f (6g g g g )
4 (4)2
+ f aeg f bgf f ech f f dh (13g g + 4g g 2g g )
i
+ f aeg f bgf f edh f f ch (13g g 2g g + 4g g )
g 4 i 2 h cdh
= if tr (ta tb th )(6g g g g )
4 (4)2
tr (ta tb td tc )(13g g + 4g g 2g g )
i
g4 i 2 h
= 2
4 (4)
i
+ tr (ta tb td tc )(13g g + 10g g 8g g ) , (16.39)

where we have used the fact that f aeg f bgf f ef h f cdh = if cdh tr (ta tb th ) = tr ta tb [tc , td ] , and
f aeg f bgf f ech f f dh = tr (ta tb td tc ), etc. There are additional five diagrams associated with
this one, namely,
c d c b a d d b a c
a b a d c b c a b d
(a) (b) (c) (d) (e)
In the diagrams above, (a) gives the identical result to the one have just evaluated, while (b)
and (c) give identical expression, so do (d) and (e). We can find (b) from the result above
by the exchange (a c), and (d) by the exchange (a d). Then we sum up all six
diagrams, which is equivalent to summing the original one with (b) and (d) and multiplying
the result by 2:
ig 4 2 h
2 tr (ta tb tc td )(13g g 8g g + 10g g )
4(4)2
+ tr (tc tb ta td )(13g g 8g g + 10g g )
+ tr (tc tb td ta )(13g g 8g g + 10g g )
+ tr (td tb tc ta )(13g g 8g g + 10g g )
i
+ tr (td tb ta tc )(13g g 8g g + 10g g )
ig 4 2 h
= tr (ta tb tc td )(23g g 16g g + 23g g )
2(4)2
i
a c b d
+ tr (t t t t )(16g g + 23g g + 23g g ) , (16.40)
where we use the cyclic symmetry of trace and also the relation tr (ta tb tc td ) = tr (td tb tc ta ).
The third diagram reads
dd k i 4
Z
4 aeh bhg cf e dgf
=g f f f f (g k 2g k + g k )
(2)d k 2
(g k 2g k + g k )(g k 2g k + g k )
(g k 2g k + g k )
dd k 1
Z
4 a b d c
= g tr (t t t t ) 34k k k k + k 4 (g g + g g )
(2)d (k 2 )4
+ 3k 2 (g k k + g k k + g k k + g k k )

dd k 1 h 34
Z
4 a b d c
= g tr (t t t t ) d 2 2
(g g + g g + g g )
(2) (k ) 24
3 i
+ (g g + g g ) + (g g + g g )
2
ig 4 2
= tr (ta tb td tc )(47g g + 47g g + 17g g ). (16.41)
12(4)2
Combined with the other two similar diagrams, we get
ig 4 2
tr (ta tb tc td )(47g g + 17g g + 47g g )
12(4)2
+ tr (ta tb td tc )(47g g + 47g g + 17g g )
+ tr (ta tc tb td )(17g g + 47g g + 47g g )

(16.42)
The fourth diagram with ghost loop is given by
dd k
Z i 4
4 eah hbg gdf f ce
= (g) f f f f (1) kk kk
(2)d k2
ig 4 2
tr (ta tb td tc )(g g + g g + g g ). (16.43)
24(4)2
There are six distinct diagrams with ghost loops, with different permutations of external
labels (Lorentz and gauge). They sum to
ig 4 2
tr (ta tb tc td ) + tr (ta tb td tc ) + tr (ta tc tb td )

2
12(4)
(g g + g g + g g ). (16.44)
Finally the diagram with fermion loop reads
dd k
Z
4 i i i i
= (ig) nf tr (tar tbr tdr tcr ) d
() tr
Z (2) k/ k/ k/ k/
d
d k 1
= g 4 nf tr (tar tbr tdr tcr ) 4(g g + g g g g )(k 2 )2

(2)d (k 2 )4
8(g k k + g k k + g k k + g k k )k 2 + 32k k k k

dd k 1
Z
4 a b d c
= g nf tr (tr tr tr tr ) 4(g g + g g g g )
(2)d (k 2 )2
4(g g + g g ) + 43 (g g + g g + g g )

4ig 4 nf 2
= tr (tar tbr tdr tcr )(g g + g g 2g g ). (16.45)
3(4)2
Combined with the similar diagrams with different permutations, we get
8ig 4 nf 2
tr (tar tbr tcr tdr )(g g 2g g + g g )
3(4)2
+ tr (tar tbr tdr tcr )(g g + g g 2g g )
+ tr (tar tcr tbr tdr )(2g g + g g + g g ). (16.46)
Now, we sum up the first four types of diagrams, namely, (16.38), (16.40), (16.42), and
(16.44), and find the result to be
2ig 4 2 h
tr (ta tb tc td )(g g + 2g g g g )
3(4)2
+ tr (ta tb td tc )(g g g g + 2g g )
i
+ tr (ta tc tb td )(2g g g g g g )
2ig 4 2 h
g g 2 tr (ta tc tb td ) tr (ta tb tc td ) tr (ta tb td tc )

= 2
3(4)
+ g g 2 tr (ta tb tc td ) tr (ta tb td tc ) tr (ta tc tb td )

a b d c a b c d a c b d
i
+ g g 2 tr (t t t t ) tr (t t t t ) tr (t t t t )
ig 4 2
= C 2 (G) g g (f ade f bce f ace f bde )
3(4)2
+ g g (f ade f bce f abe f cde ) + g g (f ace f bde + f abe f cde )

ig 4 2
C2 (G) f abe f cde (g g g g )

= 2
3(4)
+ f ace f bde (g g g g ) + f ade f bce (g g g g ) .

(16.47)
Similar manipulations on (16.46) gives
4ig 4 2 abe cde

nf C 2 (r) f f (g g g g )
3(4)2
+ f ace f bde (g g g g ) + f ade f bce (g g g g ) .

(16.48)
Therefore, we finally find 14g to be
g2 2 2

C2 (G) + 4nf C2 (r) log M , (16.49)
3(4)2
and it is straightforward to see that 14g 3 = 2(1 2 ).

Chapter 17
Quantum Chromodynamics
17.1 Two-Loop renormalization group relations

(a) In this problem we study the higher orders of QCD function. Formally, we have
b0 3 b1 5 b2 7
(g) = 2
g 4
g g + . (17.1)
(4) (4) (4)6
The we can deduce the corresponding function for s g 2 /(4), namely,
s 2b0 2 2b1 3 2b2 4
= s s + . (17.2)
4 (4) 2 (4)3 s
Integrate this equation, we get
Z Q Z s (Q2 ) 1
d b0 2 b1 3 b2 4
= ds + + + . (17.3)
2 4 s (4)2 s (4)3 s
The integral can be carried out approximately, as
s (Q2 )

2 4 1 b1
log(Q/) = + log b1
+ . (17.4)
b0 s (Q2 ) 4b0 1 + 4b s (Q2 )
0
Then the running coupling s (Q2 ) can be solved iteratively, to be,

b1 log log(Q/)2

2 4 1
s (Q ) = 2 + . (17.5)
b0 log(Q/)2 b0 [log(Q/)2 ]2
(b) Now we substitute (17.5) into the e+ e annihilation cross section, we get
(e+ e hadrons)
X
2 s 2
s 3
= 0 3 Qf 1 + + a2 + O(s )
f

4b1 log log(Q/)2
X
2 4 1
= 0 3 Qf 1 + 2
3 2 ]2
+ . (17.6)
f
b 0 log(Q/) b 0 [log(Q/)
Since the expression for the cross section is independent of renormalization scheme to the or-
der showed above, we conclude that the function coefficients b0 and b1 are also independent
of the renormalization scheme.
133
134 Chapter 17. Quantum Chromodynamics
17.2 A Direct test of the spin of the gluon

(a) We repeat the calculations in Part (c) of the Final Project I, with the gluon-quark
vertex replaced by a Yukawa vertex.

2 i i i
iM = Qq (ie) (ig)u(k1 ) v(k2 ) 2 v(p2 ) u(p1 ) (17.7)
k/1 + k/3 k/2 + k/3 q
Then, use the trick described in Final Project I, we have
1 X Q2q g 2 e4
|iM|2 = 2
tr ( p/1 p/2 )
4 4s
1 1 1 1

tr k/ k/1
k/1 + k/3 k/2 + k/3 2 k/1 + k/3 k/2 + k/3
2
32Q2q g 2 e4

1 1
= (p1 p2 )(k1 k3 )(k2 k3 ) + . (17.8)
3s2 (k1 + k3 )2 (k2 + k3 )2
Rewrite this in terms of xq , xq and x3 , we get
2
4Q2q g 2 e4

1 X 2 1 1
|iM| = (1 xq )(1 xq ) +
4 3s2 1 xq 1 xq
4Q2q g 2 e4 x23
= . (17.9)
3s2 (1 xq )(1 xq )
Note the phase space integral for 3-body final state is deduced in Final Project 1 to be
Z Z
s
d3 = dxq dxq ,
128 3
thus the differential cross section is given by
d2 s 42 Q2q g x23
(e+ e q qS) = 1
4
2
|M| = . (17.10)
dx1 dx2 128 3 3s 4 (1 xq )(1 xq )
(b) Now let xa > xb > xc . Then there are six ways to associated the original three
variables xq , xq and x3 to these three ordered ones. Note that the integral measure dxa dxb
does not change for different possibilities since the change of integral variables (xq , xq )
(xq , x3 ) or (xq , x3 ) generate an Jacobian whose absolute value is 1, due to the constraint
xq + xq + x3 = 2. Therefore, summing up all 6 possibilities, we get
d2
(e+ e q qS)
dxa dxb
x2c x2b x2a
+ + , (17.11)
(1 xa )(1 xb ) (1 xc )(1 xa ) (1 xb )(1 xc )
for q qS final state, and
d2
(e+ e q qS)
dxa dxb
x2a + x2b x2b + x2c x2c + x2a
+ + , (17.12)
(1 xa )(1 xb ) (1 xb )(1 xc ) (1 xc )(1 xa )
We plot these two distributions on the xa xb plain with the range xa > xb > xc , as shown
in Figure
17.3. Quark-gluon and gluon-gluon scattering 135
1.0 1.0
0.8 0.8
0.6 0.6
xb
xb
0.4 0.4
0.2 0.2
0 1 0 1
0.0 0.0
0.6 0.7 0.8 0.9 1.0 0.6 0.7 0.8 0.9 1.0
xa xa
Figure 17.1: The differential cross sections of e+ e q qg as a function of x1 and x2 ,

assuming gluon is a vector/scalar particle in left/right diagram.
17.3 Quark-gluon and gluon-gluon scattering

In this problem we evaluate the cross sections for two processes: (a) q q gg, (b)
gg gg.
(a) There are three diagrams contributing the process q(k1 )q(k2 ) g(p1 )g(p2 ) at the tree
level, as shown in Fig. 17.11 of Peskin&Schroeder. The amplitudes associated with these
diagrams are listed as follows:
i(k/1 p/1 )
iM1 = (ig)2 v(k2 )/ (p2 ) / (p1 )u(k1 )tb ta , (17.13a)
(k1 p1 )2
i(k/1 p/2 )
iM2 = (ig)2 v(k2 )/ (p1 ) / (p2 )u(k1 )ta tb , (17.13b)
(k1 p2 )2
iM3 = (ig)gf abc g (p2 p1 ) g (2p2 + p1 ) + g (p2 + 2p1 )

i
v(k2 ) u(k1 ) (p1 ) (p2 )tc . (17.13c)
(k1 + k2 )2
It is convenient to evaluate these diagrams with initial and final states of definite helicities.
By P and CP symmetry of QCD, there are only two independent processes, namely qL qR
gR gR and qL qR gR gL , that could be nonzero. Lets evaluate them in turn for the three
diagrams. To begin with, we set up the kinematics:
k1 = (E, 0, 0, E), p1 = (E, E sin , 0, E cos ),

k2 = (E, 0, 0, E), p2 = (E, E sin , 0, E cos ). (17.14)
Then,

uL (k1 ) = 2E(0, 1, 0, 0), vL (k2 ) = 2E(1, 0, 0, 0).
L (p1 ) = 1 (0, cos , i, sin ),

2
R (p1 ) = 1 (0, cos , i, sin ),
2
L (p2 ) = 1
(0, cos , i, sin ),
2
R (p2 ) = 1
(0, cos , i, sin ).
2
(17.15)
Now we begin the calculation. (In the following we use s sin and c = cos .)

s 1 + c
ig 2 E 2 tb ta 1 + c s
iM1 (qL qR gR gR ) = (0, 0, 1, 0)

t 1 c

s
1 c s

1 c s s 1 c 0
s 1 + c 1 c s 1

1 + c s 1 + c

s 0
s 1 c 1 + c s 0
2E 2
= ig 2 tb ta (1 cos ) sin = ig 2 tb ta sin . (17.16)
t

s 1 c
ig 2 E 2 ta tb 1 c s
iM2 (qL qR gR gR ) = (0, 0, 1, 0)

u s 1 + c

1 + c s

1 + c s s 1 + c 0
s 1 c 1 + c s 1

1 c s 1 c

s 0
s 1 + c 1 c s 0
2E 2
= ig 2 tb ta (1 + cos ) sin = ig 2 ta tb sin . (17.17)
u

c s 0
g 2 f abc tc E 2 s c 1

iM3 (qL qR gR gR ) = (0, 0, 1, 0) 4

s c s

0
s c 0
= g 2 f abc tc sin = ig 2 [ta , tb ] sin . (17.18)
Thus we find that

iM(qL qR gR gR ) = iM1 + iM2 + iM3 (qL qR gR gR ) = 0. (17.19)
In the same manner, we calculate the amplitude for qL qR gR gL . This time, we find:
iM1 (qL qR gR gL ) = ig 2 tb ta sin , (17.20a)

t
iM2 (qL qR gR gL ) = ig 2 ta tb sin , (17.20b)
u
17.3. Quark-gluon and gluon-gluon scattering 137
iM3 (qL qR gR gL ) = 0, (17.20c)
Therefore,
2

b a a bt
iM(qL qR gR gL ) = ig t t +t t sin , (17.21)
u
and by crossing symmetry,
u
iM(qL qR gL gR ) = ig 2 tb ta + ta tb sin . (17.22)
t
There are two more nonzero amplitudes with qR qL initial states, which are identical to the
amplitudes above. Then we find the spin- and color-summed/averaged squared amplitude
to be
1 1 X
2
2 |M|2
3 2 spin,color
2

1 4 2 b a a b b a b a t a b b a t
= 2 g sin tr (t t t t ) + 2 tr (t t t t ) + tr (t t t t ) 2 + (t u)
36 u u
2 2 2

8 s 16 t 4t
= (1 cos2 ) 1+ 2 + (t u)
9 3 u 3u
512 2 s2 t 9(t2 + u2 )

u
= + . (17.23)
27 u t 4s2
Therefore the differential cross section is given by
32s2 t 9(t2 + u2 )

d u
= + . (17.24)
dt 27s2 u t 4s2
(b) Now consider the process g(k1 )g(k2 ) g(p1 )g(p2 ). The four tree level diagrams are
shown in Fig. 17.12 of Peskin&Schroeder. Their amplitudes are given by:
i
iM1 = g 2 f abc f cde g (k1 k2 ) + g (k1 + 2k2 ) g (2k1 + k2 )

s
g (p2 p1 ) g (p1 + 2p2 ) + g (p1 + 2p2 ) (k1 ) (k2 ) (p1 ) (p2 ),

(17.25a)
i
iM2 = g 2 f ace f bde g (k1 + p1 ) g (2p1 k1 ) g (2k1 p1 )

t
g (k2 + p2 ) g (2p2 k2 ) + g (p2 2k2 ) (k1 ) (k2 ) (p1 ) (p2 ),

(17.25b)
i
iM3 = g 2 f ade f bce g (k1 + p2 ) g (2p2 k1 ) g (2k1 p2 )

u
g (k2 + p1 ) g (2p1 k2 ) + g (p1 2k2 ) (k1 ) (k2 ) (p1 ) (p2 ),

(17.25c)
iM4 = ig 2 f abe f cde (k1 ) (p1 )(k2 ) (p2 ) (k1 ) (p2 )(k2 ) (p1 )

+ f ace f bde (k1 ) (k2 ) (k1 ) (p2 ) (k1 ) (p2 )(k2 ) (p1 )

+ f ade f bce (k1 ) (k2 ) (p1 ) (p2 ) (k1 ) (p1 )(k2 ) (p2 ) .

(17.25d)
The choice for all external momenta and final states polarizations are the same with that in
(a). Now to evaluate the amplitude gR gR gR gR , we also need the initial states polarization
vectors for right-handed gluons with momenta k1 and k2 , which are given by
R (k1 ) = 1 (0, 1, i, 0),

2
R (k2 ) = 1 (0, 1, , 0).
2
(17.26)
Then after some calculations, we find,
iM1 = ig 2 f abe f cde cos , (17.27a)

19 + 7 cos 11 cos2 + cos3
iM2 = ig 2 f ace f bde ; (17.27b)
4(1 cos )
19 7 cos 11 cos2 cos3
iM3 = ig 2 f ade f bce ; (17.27c)
4(1 + cos )
h
iM4 = ig 2 f abe f cde cos + 1
4
f ace f bde (3 + 2 cos cos2 )
i
+ 41 f ade f bce (3 2 cos cos2 ) . (17.27d)
The sum of these four amplitudes is

h 2
iM(gR gR gR gR ) = 2ig 2 f abe f cde cos f ace f bde + cos
1 cos
2 i
f ade f bce cos
1 + cos
h 1 1 i
= 4ig 2 f ace f bde + f ade f bce
1 cos i 1 + cos
h s s
= 2ig 2 f ace f bde + f ade f bce . (17.28)
t u
We can also obtain the amplitudes for gL gR gL gR and gL gR gR gL from the result above
by crossing symmetry, namely the change of variables (s, b) (u, d) and (s, b) (t, c),
which gives
h u ui
iM(gL gR gL gR ) = 2ig 2 f ace f bde + f abe f cde , (17.29)
t s
h t ti
iM(gL gR gR gL ) = 2ig 2 f abe f cde f ade f bce . (17.30)
s u
The amplitudes for gL gL gL gL , gR gL gR gL and gR gL gL gR are identical to the
amplitudes for gR gR gR gR , gL gR gL gR and gL gR gR gL , respectively, due to parity
conservation of QCD. It can be shown by the conservation of angular momentum that other
helicity amplitudes all vanish. Therefore we have found all required amplitude. To get the
cross section, we take the square of these results.
X
|M(gR gR gR gR )|2
2 2 2 i
ace bde acf bdf s ade bce adf bcf s ace bde adf bcf s
h
4
= 4g f f f f +f f f f + 2f f f f
t2 u2 tu
h s2 s 2
s 2 i
= 4g 4 tr (ta ta tb tb ) 2 + 2 + 2 tr (ta tb ta tb )
t u tu
17.4. The gluon splitting function 139
s2 s2 s2
= 288g 4 + + , (17.31)
t2 u2 tu
where ta is the generator of SU (3) group in adjoint representation which is related to the
structure constants by f abc = i(ta )bc . Thus tr (ta ta tb tb ) = (C2 (G))2 d(G) = 72, and
tr (ta tb ta tb ) = tr (ta tb [ta , tb ]) + tr (ta ta tb tb ) = 1

2
if abc tr ([ta , tb ]tc ) + (C2 (G))2 d(G)
= 1
2
f abc f abd tr (tc td ) + (C2 (G))2 d(G) = 1
2
(C2 (G))2 d(G),
which is 36 for SU (3). Similarly, we can work out the square of other amplitudes, to be
X u2 u2 u2
|M(gL gR gL gR )|2 = 288g 4 + , + (17.32)
t2 s2 st
X t2 t2 t2
|M(gL gR gR gL )|2 = 288g 4 2 + 2 + . (17.33)
s u su
Therefore, the spin-averaged and squared amplitudes is
1 1 X 2 1 4
2tu 2us 2st
|M | = 2 288g 6
82 22 82 22 s2 t2 u2
tu us st
= 72 2 s2 3 2 2 2 . (17.34)
s t u
Thus the differential cross section is
d 9s2 tu us st
(gg gg) = 3 . (17.35)
dt 2s2 s2 t2 u2
17.4 The gluon splitting function

In this problem we calculate the gluon splitting function Pgg (z) by evaluating the
amplitude of the virtual process g gg, as shown in Fig. 17.2.
Figure 17.2: The Gluon splitting process.
The momenta of initial and final states are taken to be the same with that of Fig. 17.16
of Peskin&Schroeder. That is, we have
p = (p, 0, 0, p), q = (zp, p , 0, zp), k = ((1 z)p, p , 0, (1 z)p), (17.36)
and the polarization vectors associated with gluons are,
iL (p) = 1 (1, i, 0),

2
iR (p) = 1 (1, i, 0),
2
p p
iL (q) = 1 (1, i,
2
), iR (q) = 1 (1, i,
2
),
zp zp
p p
iL (k) = 1 (1, i, ), iR (k) = 1 (1, i, ). (17.37)
2 (1 z)p 2 (1 z)p
Then we can evaluate the amplitude for the process g gg directly, which is given by
iMabc = gf abc (q) (p) (p + q) (k) + (q) (k) (k q) (p)

(k) (p) (p + k) (q) .

(17.38)
We evaluate the amplitudes with definite initial and final polarizations in turn:
1 1 abc
iMabc (gL (p) gL (q)gL (k)) = 2 + gf p , (17.39a)
1 z z
abc 2z abc
iM (gL (p) gL (q)gR (k)) = gf p , (17.39b)
1 z
2(1 z) abc
iMabc (gL (p) gR (q)gL (k)) = gf p , (17.39c)
z
iMabc (gL (p) gR (q)gR (k)) = 0. (17.39d)
By parity invariance, the amplitudes with right-handed initial gluon are dictated by the
results above. Note further that f abc f abc = 24, thus we have
X
1
2
1
8
|M|2 = 12 18 2 24 2g 2 p2
spin,color
h 1 1 2 z2 (1 z)2 i
+ + +
1z z (1 z)2 z2
12g 2 p2 h 1 z z i
= + + z(1 z)
z(1 z) z 1z
2 2
2e p (0)
= Pgg (z), (17.40)
z(1 z)
where the superscript (1) represents the part of the splitting function contributed from the
diagram calculated above, in parallel with the notation of Peskin&Schroeder. (See 17.100.
for instance.) Therefore we get
h1z z i
(1)
Pgg =6 + + z(1 z) . (17.41)
z 1z
Besides, there should be a term proportional to (1z) in Pgg , which comes from the zeroth
order, as well as the corrections from Pqg and Pgg , where Pqg (z) = 21 (z 2 + (1 z)2 ).
Now lets take it to be A(1 z), then the coefficient A can be determined by the following
normalization condition (namely the momentum conservation):
Z 1
(1)

1= dz z 2nf Pqg (z) + Pgg (z) + A(1 z) , (17.42)
0
where nf is the number of fermion types, and the coefficient 2 is from contributions of both
1 1
quarks and anti-quarks. To carry out the integral, we use the prescription 1z (1z) +
,
11 1
then it is straightforward to find that A = 2 3 nf . Therefore,
h1z z i 11 nf
Pgg = 6 + + z(1 z) + (1 z). (17.43)
z (1 z)+ 2 3
17.5. Photoproduction of heavy quarks 141
17.5 Photoproduction of heavy quarks

In this problem we study the production of a pair of heavy quark-antiquark by the
scattering of a photon off a proton. At the leading order at the parton level, the process is
contributed from the photon-gluon scattering, as shown in Figure 17.3.
Figure 17.3: Tree diagrams for the photoproduction of heavy quarks at the parton level.
The corresponding amplitude can be read from a similar process e+ e in QED.

From (5.105) of Peskin & Schroeder, we have the amplitude for e+ e 2, which reads
(adapted to our notation for external momenta)

1 X 4 k1 p2 k1 p1 1 1
+ 2
|M(e e 2)| = 2e + + 2m2 +
4 k1 p1 k1 p2 k p k1 p2
1 1 1
1 2
m4 + . (17.44)
k1 p1 k1 p2
Then the amplitude M(g QQ) can be obtained by making the exchange (k1 , k2 )
(p1 , p2), replacing e4 by e2 g 2 , and also including the factor 81 Q2q tr (ta ta ) = 12 Q2q taking
account of the color average, the electric charge of quarks, and the summation of color
indices, respectively. Then the amplitude in the present case is

1 X 2 2 2 p1 k2 p1 k1 1 1
2
|M(g QQ)| = e g Qq + + 2m2 +
48 p1 k1 p1 k2 p1 k1 p1 k2
1
1 2
m4 + . (17.45)
p1 k1 p1 k2
In partons center-of-mass frame, we have k1 = (E, 0, 0, E), k2 = (E, 0, 0, E), p1 =

(E, p sin , 0, p cos ) and p2 = (E, p sin , 0, p cos ), with p2 = E 2 m2 . Then p1 k1 =
E(E p cos ) and p1 k2 = E(E + p cos ). Then the differential cross section is
s Q2q p E 2 + p2 cos2 2m2 2m4

d
= 2 . (17.46)
d cos 16 E3 E 2 p2 cos2 (E p2 cos2 )2
Then the cross section for photon and proton initial state is given by
Z
((k1 ) + p(k2 ) QQ) = dx fg (x)((k1 ) + g(xk2 ) QQ). (17.47)
17.6 Behavior of parton distribution functions at small

x
(a) In this problem we study the solution of A-P equations at small x with certain ap-
proximations. Firstly, we show that the A-P equations,
s (Q2 ) 1 dz
Z X x x
d
fg (x, Q) = Pgq (z) ff , Q + ff ,Q
d log Q x z f
z z
x
+ Pgg (z)fg ,Q , (17.48)
z
s (Q2 ) 1 dz
Z x
d x
ff (x, Q) = Pqq (z)ff , Q + Pqg (z)fg ,Q , (17.49)
d log Q x z z z
s (Q2 ) 1 dz
Z x
d x
f (x, Q) = Pqq (z)ff , Q + Pqg (z)fg ,Q , (17.50)
d log Q f x z z z
can be rewritten as a differential equation with variable = log log(Q2 /2 ). To see this,
we note that d/d log Q = 2e d/d, and to 1-loop order, s (Q) = 2/ b0 log(Q/) =

(4/b0 )e , so we have
2 1 dz
Z X x x
d
fg (x, ) = Pgq (z) ff , + ff ,
d b0 x z f
z z
x
+ Pgg (z)fg , , (17.51)
z
2 1 dz
Z x
d x
ff (x, ) = Pqq (z)ff , + Pqg (z)fg , , (17.52)
d b0 x z z z
2 1 dz
Z x
d x
f (x, ) = Pqq (z)ff , + Pqg (z)fg , . (17.53)
d f b0 x z z z
(b) Now we apply the approximation that 1) gluon PDF dominates the integrand in the
A-P equations and 2) the function g(x, Q) = xfg (x, Q) is slowly varying with x when x is
small. Then, define w = log(1/x), which gives d/dw = xd/dx, we can calculate
2

d
g(x, ) = x x fg (x, Q)
w dx
d 2x 1 dz
Z x
'x Pgg (z)fg ,Q
dx b0 x z z
2x 1
Z
2 d h x x i
= xPgg (x)fg (x, Q) dzPgg (z) fg ,Q
b0 b0 x dx z z
2
' xPgg (x)fg (x, Q). (17.54)
b0
From the result of Problem 17.4 we know that xPgg (x) = 6 as x 0. Therefore the A-P
equation for fg becomes
2 12
g(x, ) = g(x, ). (17.55)
w b0
17.6. Behavior of parton distribution functions at small x 143
Then we verify that

h i1/2
2 48
g = K(Q ) exp w( 0 ) (17.56)
b0
is an approximation solution to the differential equation above when w 1, where K(Q2 )

is an initial condition. We apply 2 /w on this expression, to get
s
2
h i1/2
1 48 48
g(w, ) = exp w( 0 )
w 4 b0 w( 0 ) b0
r
K(Q2 )

48 2
2( 0 ) + 1+ w( 0 ) K(Q ) . (17.57)
b0
In the limit w 1, the square root term in the last line dominates, thus
2
h i1/2
b0 2 48 12
g(w, ) ' K(Q ) exp w( 0 ) = g(w, ). (17.58)
w 12 b0 b0
(c) Then we consider the A-P equation for quarks. If we adopt the approximation in (b)
again, namely, the gluon PDF dominates and the function q(x, ) = xff (x, Q) is slowly
varying, then we have
2x 1 dz
Z
x
q(x, ) = x ff (x, ) = Pqg (z)fg ,
b0 x z z
2 1 1 1
Z x Z
x
= dzPqg (z)g , = dz z 2 + (1 z)2 g ,
b0 x z b0 x z

2 1 x 1 2
3 2
' (2z 2z + 3z)g , ' g(x, ), (17.59)
b0 6 z x 3b0
where we have used x 1 and q(x, )/x ' 0. Then, we verify that
r h i1/2
0 2 48
q = K(Q ) exp w( 0 ) (17.60)
27bo w b0
is again an approximate solution to the equation derived above, in the limit w 1. In

fact,
h i1/2 2
48
q(x, ) = exp w( 0 ) K(Q2 )
b0 3b0
s
K(Q2 )

1 3
2
+ K(Q ) + 2( 0 )
18 b0 w( 0 )
h
2 48 i 1/2 2
' K(Q2 ) exp w( 0 ) = g(x, ). (17.61)
3b0 b0 3b0
(d) We use the fitted formula of K(Q2 ) to plot the PDFs of gluon and quarks in Figure.
3.0
2.5
xfg Hx,Q2 L
2.0
1.5
xf f Hx,Q2 L
1.0
0.5
0.0
0.00 0.05 0.10 0.15 0.20
x
Figure 17.4: Approximate parton distribution functions at small x with Q = 500GeV.

Chapter 18
Operator Products and Effective

Vertices
18.1 Matrix element for proton decay

(a) We estimate the order of magnitude of the proton lifetime, through the decay p
e+ 0 , based on the following operator,
2
OX = eR uRi uLj dLk ,
2 ijk
(18.1)
mX
where mX is the scale of this higher dimensional operator, whose typical value is around
1016 GeV, and i, j, k are color indices for quarks, , , are spinor indices. Then, the
amplitude M of this decay process should be proportional to m2 X . Note that the amplitude
2
M has mass dimension 1, thus we should have M m3p mX with mp the proton mass. Now
take mX 1016 GeV and mp 1GeV, we have the decay width
1 1 1 m5p
|M|2 1065 GeV 1033 yr1 . (18.2)
8 2mp 16 m4X
(b) Now we consider the first order QCD correction to the estimation above. The correc-
tion comes from virtual gluon exchange among three quarks in the operator. To evaluate
these 1-loop diagrams, we firstly fixed the renormalization condition of OX to be
uRi
= iijk . (18.3)

uLj dLk
The 1-loop diagrams are shown in Figure 18.1. The Feynman rules can be written in two-
component spinor notations. The left-handed spinors propagator reads i(p)/p2 , the right-
handed spionrs propagator is i(p )/p2 , the QCD interaction between quark and gluon is

i[Li (ta )ij Lj + Ri (ta )ij Rj ], and the vertex corresponding OX reads iijk . Then
145
146 Chapter 18. Operator Products and Effective Vertices
Figure 18.1: 1-loop QCD correction to the effective operator of proton decay.
the first diagram reads

dd q i

iq iq
Z
2 a a 0 0
(a) = i(ig) imn (t )mj (t )nk
(2)d q 2 q2 0 q2 0
2 Z d
0 0 d q q q
= g2 ijk ( ) 0 ( )0
3 (2)d q 6
2 i 2
= g2 ijk 16 2
3 4(4)
2
8g 2
= iijk , (18.4)
3(4)2
where the Pauli matrices is simplified as follows,
0 0
( ) 0 ( )0 = (T T ) = (T )
= 4(T ) = 4( ) = 16 , (18.5)
in which we used the fact that = i 2 and T 2 = 2 . In the computation of this

diagram, we also used imn (ta )mj (ta )nk = (2/3)ijk . The coefficient of this equality can
be easily justified by contracting both sides with ijk . Similarly, we compute the second
diagram, as follows,
dd q i iq

iq
Z
2 a a 0 0
(b) = i(ig) imn (t )mj (t )nk
(2)d q 2 q2 0 q2 0
2 i 2
= g2 ijk 4 2
3 4(4)
2
2g 2
= iijk , (18.6)
3(4)2
where we used the identity ( ) ( ) = 2 . The third diagram gives the same result
as the second one. Therefore, we get the counterterm for the operator OX in M S scheme
to be
4g 2

2 2
OX = log M , (18.7)
(4)2
where M 2 is the renormalization scale. We further recall that the field strength renormal-
ization counterterm for quarks in QCD is given by
4g 2

2 2
2 = log M , (18.8)
3(4)2
18.2. Parity-violating deep inelastic form factor 147
then the anomalous dimension of operator OX is given by
3 4g 2
=M O X + 2 = . (18.9)
M 2 (4)2
Therefore this QCD correction will enhance the operator strength by a factor of
a0 /2b0
log(m2X /2 )

, (18.10)
log(m2p /2 )
where ' 200GeV, a0 = 4 is the coefficient from anomalous dimension, and b0 = 11

(2/3)nf = 7 is the 1-loop coefficient of QCD function. Taking mX = 1016 GeV and
mp = 1GeV, this factor is about 2.5. Then the decay rate of proton is enhanced by a factor
of 2.52 ' 6.3.
18.2 Parity-violating deep inelastic form factor

(a) We firstly compute the amplitude of the neutrino deep inelastic scattering through
charged current interaction, which reads
ig 2 0
1
5
iM(p X) = 2
u(k ) u(k)
2mW 2
Z
d4 x eiqx hX| J+ (x) + J (x) |P i.

(18.11)
Then the squared amplitude with initial protons spins averaged and final state X summed
is
1 g4 X
1
1 X 2 0 5
1+
5 0
|M| = u(k ) u(k)u(k) u(k )
2 2 4m4W spin 2 2
XZ
dX hP | J+ (x) + J (x) |XihX| J+ (0) + J (0) |P i.

(18.12)
X
The trace factor can be straightforwardly worked out to be

X 1
5
1+
5

0 0
L u(k ) u(k)u(k) u(k )
spin
2 2
h 1 1+ i
5 5 0
= tr k/ k/
2 2
= 2 k k + k k g k k 0 + i k k0 .
0 0

(18.13)
Then, use the optical theorem, we have

XZ
dX hP | J+ (x) + J (x) |XihX| J+ (0) + J (0) |P i

L
X
= 2 Im L W () ,

(18.14)
with Z
()
d4 x eiqx hP |T J (x)J+ (0) |P i,

W = 2i (18.15)
Therefore, the cross section is

d3 k 0 1
Z
1 1 X
(p X) = |M|2
2s (2)3 2k 0 2
g4
Z
1 ys ()

= dxdy Im L W , (18.16)
2s (4)2 4m4W
and the differential cross section is thus given by
d2 yG2F
(p X) = Im k k0 + k k0 g k k 0 + i k k0 W () .

2
(18.17)
dxdy 2
(b) The lepton momentum tensor obtained in (a) is
L = 2(k k0 + k k0 g k k 0 + i k k0 ). (18.18)
Then it is straightforward to see that q L = (k k 0 ) L = 0 and q L = 0. As a

consequence, any term in W () proportional to q or q is irrelevant. Therefore we can
()
rewrite the tensor W () in terms of three form factors Wi (i = 1, 2, 3). That is,
() () ()
W () = g W1 + P P W2 + i P q W3 + . (18.19)
Then the deep inelastic scattering cross section becomes

d2 yG2F h () ()
(p X) = 2
2(k k 0 ) Im W1 + 2(P k)(P k 0 ) Im W2
dxdy 2
i
()
4 (P k)(q k 0 ) (q k)(P k 0 ) Im W3 . (18.20)
()
(c) Now we evaluate Im W1,2,3 in the parton model. Firstly, W () can be written as
Z Z 1 X 1
qf (P )T {J (x)J+ (0)}qf (P ) ,
() 4 iqx

W = 2i d xe d ff () (18.21)
0 f

and be evaluated in terms of Feynman diagrams displayed in Fig. 18.10 of Peskin & Schroeder.
For the first diagram, we have
Z 1 1 5 1 5
1 i
2i d fd () u(p) u(p)
0 2 p/ + /q + i 2
5 5

1 1 i 1

+fu () u(p) u(p) , (18.22)
2 p/ + /q + i 2
where p = P . Then, averaging/summing over initial/final spin states gives

Z 1
1 1 1 5 1 5
2 d fd () tr p/ (p/ + /q)
0 2 2 2
18.2. Parity-violating deep inelastic form factor 149
1 1 1 5 5

1 1
+ fu () tr p/ (p/ + /q)
2 2 2 2p q + q 2 + i
Z 1
d
fd () + fu () 4 2 P P 2P qg

0

1
+ fd () fu () 2i P q , (18.23)
2p q + q 2 + i
where we have dropped terms containing q or q in the last line. Then it is easy to read
from this expression that
Z 1
() 1
Im W1 = 2P q d fd () + fu () Im , (18.24)
0 2p q + q 2 + i
Z 1
() 1
Im W2 = d 4 fd () + fu () Im , (18.25)
0 2p q + q 2 + i
Z 1
() 1
Im W3 = 2 d fd () fu () Im , (18.26)
0 2p q + q 2 + i
where 2P q = ys, and
1
Im 2
= ( x). (18.27)
2p q + q + i ys
Note that the second diagram in Fig. 18.10 of Peskin & Schroeder does not contributes, as
explained in the book. Therefore we conclude that
()
Im W1 = fd (x) + fu (x) , (18.28)
() 4x
Im W2 = fd (x) + fu (x) , (18.29)
ys
() 2
Im W3 = fd (x) fu (x) . (18.30)
ys
(d) The analysis above can be easily repeated for the left-handed current JfL of single
flavor f , defined by JfL = f PL f where PL (1 5 )/2. Then, define
Z
Wf L = 2i d4 x eiqx P T {JfL (x)JfL (0)}P ,

(18.31)
and its decomposition,
WfL = g W1f L + P P W2f L + i P q W3f L + . (18.32)
We see that it amounts to the replacement in the final result that d f and u f.
Therefore,
()
Im W1f L = ff (x) + ff(x) , (18.33)
() 4x
Im W2f L = ff (x) + ff(x) , (18.34)
ys
() 2
Im W3f L = ff (x) ff(x) . (18.35)
ys
(e) Now we perform OPE on WfL . Firstly,

Z Z
4
d xe iqx
JfL (x)JfL (0) ' 4
d xe iqx
q PL q(x)q PL q(0) + q PL q(x)q PL q(0)
i(i/ + /q)
= q PL P L q + , q q . (18.36)
(i + q)2
Then, the first term in the last line can be written as
i(i/ + /q) i(i/ + /q) i(i/ + /q) 5

1
q PL 2
PL q = q q q q
(i + q) 2 (i + q)2 (i + q)2
i
= q 2 ( (i ) ) g /q i (i + q)

2

1 X 2iq n
2 q, (18.37)
Q n=0 Q2
where we have symmetrize the indices for the first two terms in the square bracket and
antisymmetrize the indices for the third term, by using the equalities 12 ( + ) =
g + g g and 12 ( 5 5 ) = i , and terms proportional to
q or q have also been dropped. The (anti)symmetrization can be understood by looking
at (18.32), where the terms with no 5 are symmetric on while the term involving 5 is
antisymmetric on . Therefore, when including the second term in (18.36), we should keep
terms of even powers in q for symmetric indices and of odd powers in q for antisymmetric
.
Now, with these understood, and using the definition of twist-2, spin-n operator,
(n)1 n
Of = qf (1 (iD2 ) (iDn ) )qf traces, (18.38)
we have,
(2q1 ) (2qn2 ) (n)1 n2

Z X
i d4 x eiqx JfL (x)JfL (0) = 2 Of
n>0, even
(Q2 )n1
1 X (2q1 ) (2qn ) (n)1 n
g 2 )n
Of
2 n>0, even
(Q
X (2q1 ) (2qn1 ) (n)1 n1
i q Of . (18.39)
n>0 odd
(Q2 )n
(n)1 n
Then, using hP |Of |P i = 2Anf P 1 P n , we can get WfL to be
X (2q P )n2 n X (2q P )n

WfL = 8P P A f 2g
Anf
n>0, even
(Q2 )n1 n>0, even
(Q 2 )n

X (2q P )n1 n
+ 4i P q Af . (18.40)
n>0, odd
(Q2 )n
18.3. Anomalous dimensions of gluon twist-2 operators 151
So we can read out

X (2q P )n n
W1f L = 2 Af , (18.41)
n>0, even
(Q2 )n
X (2q P )n2 n
W2f L = 8 Af , (18.42)
n>0, even
(Q2 )n1
X (2q P )n1 n
W3f L = 4 Af . (18.43)
n>0, odd
(Q2 )n
(f ) Now we use W3f L obtained above to derive a sum rule for parton distribution ff ,
defined by
ys
ff (x, Q2 ) = Im W3f L (x, Q2 ), (18.44)
2
where x = Q2 / and = 2P q = ys. The the analytic behavior of W3f L on the v-complex
plane is shown in Fig. 18.11 of Peskin & Schroeder. Thus we can define the contour integral
Z
d 1
I3n = W3f L (, Q2 ), (18.45)
2i n
where the contour is a small circle around the origin = 0. This integral picks up the
coefficient of (n1) term, namely, I3n = 4Anf /(Q2 )n . On the other hand, the contour can be
deformed as shown in Fig. 18.12 of Peskin & Schroeder. Then the integral can be evaluated
as Z Z 1
d 1 4
I3n = 2 n
2
(2i) Im W3f L (, Q ) = 2 )n
dx xn1 ff (x, Q2 ). (18.46)
Qf2 2i (Q 0
Therefore we get the sum rule,

Z 1
dx xn1 ff (x, Q2 ) = Anf . (18.47)
0
18.3 Anomalous dimensions of gluon twist-2 operators

In this problem we finish evaluating anomalous dimension matrix n in (18.180) of Peskin
& Schroeder, given by !
2 n n
g a a
n = ff fg
(18.48)
(4)2 angf angg
where anff has already been evaluated explicitly in the book. Here we evaluate the remaining
(n) (n)
three elements. The needed Feynman rules involving operators Of and Og are listed as
follows:
/ ( k)n1 ,
=
k
h i
= 2 g ( k)n + k 2 ( k)n2 2k ( ) ( k)n1 ab ,
%k
a b
n
X
abc
= 2igf g ( k1 )j1 ( k2 )nj
k1 % - k3 j=1
k2
a c + (cyclic permutations on ak1 , bk2 , ck3 ) + .
b
In the last expression, we list only terms containing a metric tensor g , and the ignored
terms (marked by ) are irrelevant in the following calculations. To be clear, we have
(n) (n)
introduced a source J (n) to these operators, namely, we write L = J1 n Of,g 1 n , with
(n)
J1 n = 1 n , and 2 = 0. As can be easily seen, this source automatically projects
(n)
the operator Of,g to its symmetric and traceless part.
(a) Firstly, we consider anfg , which can be got by evaluating the following two diagrams.
p%
With the Feynman rules listed above, the first diagram reads,
d4 k
Z
b i / i a i i
h
2
(ig) 4
(1) tr t t ( k)n1
(2) /
k k / k p/
/
4 0 Z 1
2(1 x)
Z
dk
= ig 2 tr [ta tb ] 4
dx 02 ( k)n1 tr [ k/
/ k/ (k/ p/)]. (18.49)
(2) 0 (k )3
We need to extract terms of proportional to g ( p)n and of logarithmical divergence.

This needs some manipulations on the numerator of the integrand. We firstly evaluate
the gamma trace, keep terms containing at least two powers of k, and shift the variable
k k 0 = k xp . Then we pick up terms containing two k 0 , which contributes to
logarithmical divergence. At last we symmetrize the indices according to k 0 k 0 k 02 g /4.
The detailed steps are given as follows.
( k)n1 tr [ k// k/ (k/ p/)]

h i h i h i
16( k)n k k 4( k)n (k 2p) kg 4( k 0 )n1 ( p)k 2 g
h i h
16xn ( p)n k 0 k 0 4nxn (k 0 p)( k 0 )( p)n1 g
i
n1 0 0 n1 n n 02
+ 4n(x 2)x (k p)( k )( p) g + 4x ( p) k g
h i
8(n 1)xn1 (k 0 p)( k 0 )( p)n1 g + 4xn1 ( p)n k 02 g
h i h i
4xn ( p)n k 02 g nxn + n(x 2)xn1 + 4xn ( p)n k 02 g
h i
2(n 1)xn1 + 4xn1 ( p)n k 02 g
= (2nxn + 2xn1 )( p)n k 02 g . (18.50)
Then it is straightforward to finish the loop integral,

Z 1
d4 k 0 k 02
Z
2 a b n n n1
ig tr [t t ]( p) g dx 2(1 x)(2nx + 2x )
0 (2)4 (k 02 )3
g 2 2(n2 + n + 2) (2 d2 )
= ( p)n ab g . (18.51)
(4)2 n(n + 1)(n + 2) 2d/2
The second diagram contributes an identical term for n even. The two diagrams sum to
g 2 2(n2 + n + 2) (2 d2 ) n ab

2( p) g . (18.52)
(4)2 n(n + 1)(n + 2) 2d/2
Therefore the corresponding counterterm reads
g 2 2(n2 + n + 2) (2 d2 )
f g = , (18.53)
(4)2 n(n + 1)(n + 2) (M 2 )2d/2
and the anomalous dimension element reads

g 2 4(n2 + n + 2)
fng = M f g = , (18.54)
M (4)2 n(n + 1)(n + 2)
and thus,
4(n2 + n + 2)
anfg = . (18.55)
n(n + 1)(n + 2)
(b) Then we consider angf and angg . This time we need to evaluate the following four
diagrams.
The first diagram contributes to angf , which reads
d4 k a i a i 2
Z
2
2(ig) t t
(2)4 p/ k/ k2
h i
g ( k)n + k 2 ( k)n2 2k( ) ( k)n1
Z 1
d4 k 0 2(1 x) h
Z
2
= 2ig C2 (N ) 4
dx 02 3
(p/ k/) ( k)n
(2) 0 (k )
i
+ / ( k)n2 k 2
/ (p/ k/) / ( k)n1
/ (p/ k/)k/ + k/(p/ k/) (18.56)
To find the pieces proportional to / (p)n1 and of logarithmical divergence, we manipulate

on the expression in the square bracket, shifting the variable k k 0 = k xp , extracting
terms with two factors of k 0 , symmetrizing the integrand with k 0 k 0 k 02 g /4, and
throwing away terms proportional to 2 (= 0). This gives
h i h i
n / / n2 2
(p/ k/) ( k) + (p/ k/)( k) k
h i
/ (p/ k/)k/ + k/(p/ k/)/ ( k)n1
h i h i
= 2(p/ k/)( k)n + 2 / (p k) ( k)n2 k 2

h i
n1

2 (p k) k/ + (p k) k / (p/ k/)( k) ( k)
h i h
0
2nxn1 k/ ( k 0 )( p)n1 + 2xn1 (k 0 p)( k 0 )( p)n2
+ 2(n 2)(1 x)xn2 (k 0 p)( k 0 )( p)n2
i h
0
+ (1 x)x ( p) k + 2xn1 k/ ( k 0 )( p)n1
n2 n1 02

2(n 1)(1 x)xn2 ( k 0 )( p)n1 2 / xn1 k 02 ( p)n1
(n 1)xn1 (k 0 p)( k 0 )( p)n2

i
0 0 n1 0
n2
+ (n 1)(1 x)x (k p)( k )( p) n2
2nx k/ ( k 0 )( p)n1
hn i h i
xn1 / k 02 ( p)n1 + xn1 + n(1 x)xn2 / k 02 ( p)n1
2 h i
+ (n 1)(1 x)xn2 + 2xn1 / k 02 ( p)n1
hn i
= xn1 + xn2 / k 02 ( p)n1 . (18.57)
2
Then we have
1 Z d4 k 0 k 02
Z n
2 / ( p)n1 n1 n2
2ig C2 (N ) dx 2(1 x) x +x
0 2 (2)4 (k 02 )3
g 2 C2 (N ) 2(n2 + n + 2) (2 d2 )
= / ( p)n1 ,
(18.58)
(4)2 n(n2 1) 2d/2
which gives the counterterm coefficient,
n g 2 C2 (N ) 2(n2 + n + 2) (2 d2 )
gf = (18.59)
(4)2 n(n2 1) (M 2 )2d/2
Then, in a similar way as in (a), we get
n g 2 C2 (N ) 4(n2 + n + 2)
gf = , (18.60)
(4)2 n(n2 1)
and for N = 3, C2 (N ) = 4/3, we get
16 (n2 + n + 2)
angf = . (18.61)
3 n(n2 1)
The second to fourth diagrams contribute to angg . Now we evaluate them in turn. The
second one reads,
d4 k i 2 i
Z
2 ace bde cd
2g f f
(2)4 k 2 (p k)2
h i
g (p + k) + g (p 2k) + g (k 2p)
h i

g (p + k) + g (2k p) + g (2p k)
h i
g ( k)n + k 2 ( k)n2 2k( ) ( k)n1
Z 1
d4 k 0 2(1 x) h
Z
2 ab n
2ig C2 (G) dx 8( k) k k ( k)n k 2
(2)4 0 (k 02 )3
i
+ 2( k) (k p) 8( k) ( p)(k p) + 4( k) ( p) k g
n n1 n2 2 2
Z 1
2 ab n
h n n 1 i
2ig C2 (G)g ( p) dx 2(1 x) 3 + x nxn1 2xn2
0 2 2
Z 4 0 02
dk k
02
(2) (k )3
4
g 2 C2 (G) (2 d2 )

4 6 4 4
= + (2)g ab ( p)n . (18.62)
(4)2 n+2 n+1 n n1 2d/2
The third diagram reads (where an additional 1/2 is the symmetry factor),
d4 k i i
Z
1 2 acd bcd
2ig f f
2 (2) k (p k)2
4 2
h i
g (p + k) + g (p 2k) + g (k 2p)
n h
X i
g ( (p k)) j1
( p) nj
g ( j1
p) ( k) nj
j=1
n Z
X 1 h i
2 ab n
ig C2 (G)g ( p) dx (1 + x)xnj (x 2)(1 x)j1
j=1 0
Z 4 0
dk 1

(2)4 (k 02 )2
g 2 C2 (G) ab (2 d2 )
= g ( p)n
(4)2 2d/2
n
X 1 1 1 1
+ + +
j=1
j j+1 nj+1 nj+2
n
g 2 C2 (G) X 1 (2 d2 )

1
2
2 + +1 2d/2
(2)g ( p)n . (18.63)
(4) j=2
j n+1
The contribution from fourth diagram is identical to the one from the third diagram. Sum-
ming the last three diagram together, we get
n
g 2 C2 (G) (2 d2 )

4 4 X 1
+ 4 2 (2)g ab ( p)n . (18.64)
(4)2 (n + 1)(n + 2) n(n + 1) j=2
j 2d/2
Thus the corresponding counterterm is

n
g 2 C2 (G) (2 d2 )

4 4 X 1
g = + 4 2 . (18.65)
(4)2 (n + 1)(n + 2) n(n + 1) j=2
j (M 2 )2d/2
As a result,
n
gg =M (g + 3 )
M
n
2g 2

4 4 X 1 1 4
= + 4 C2 (G) nf C(N ) , (18.66)
(4)2 (n + 1)(n + 2) n(n + 1) j=2
j 3 3
therefore, for N = 3, C2 (N ) = 4/3 and C(N ) = 1/2, we have,

n
4 4 X 1 1 2
angg =6 + 4 nf . (18.67)
(n + 1)(n + 2) n(n + 1) j=2
j 3 9
18.4 Deep inelastic scattering from a photon

(a) The A-P equation for parton distributions in the photon can be easily written down by
using the QED splitting functions listed in (17.121) of Peskin & Schroeder. Taking account
of quarks electric charge properly, we have,
3Q2q 1 dz
Z x
d x
fq (x, Q) = Pee (z)fq , Q + Pe (z)f ,Q , (18.68)
d log Q x z z z
3Q2q 1 dz
Z x
d x
fq (x, Q) = Pee (z)fq , Q + Pe (z)f ,Q , (18.69)
d log Q x z z z
d X 3Q2q Z 1 dz
x x
f (x, Q) = Pe (z) fq , Q + fq ,Q
d log Q q
x z z z
x
+ P (z)f ,Q , (18.70)
z
where the splitting functions are
1 + z2 3
Pee (z) = + (1 z), (18.71)
(1 z)+ 2
2
1 + (1 z)
Pe (z) = , (18.72)
z
Pe (z) = z 2 + (1 z)2 , (18.73)
2
P (z) = (1 z). (18.74)
3
We take q = u, d, c, s, and Qu,c = +2/3, Qd,s = 1/3. The factor 3 in the A-P equations
above takes account of 3 colors. Since no more leptons appear in final states other than
original e+ e , they are not included in the photon structure. With the initial condition
18.4. Deep inelastic scattering from a photon 157
f (x, Q0 ) = (1 x) and fq,q (x, Q0 ) = 0 where Q0 = 0.5GeV, these distribution functions

can be solved from the equations above to the first order in , to be
3Q2q Q2
log 2 x2 + (1 x)2 ,

fq (x, Q) = fq (x, Q) = (18.75)
2 Q0
2
X Qq Q2

f (x, Q) = 1 log 2 (1 x). (18.76)
q
Q0
(b) The formulation of deep inelastic scattering from a photon is similar to the one for the
proton, as described in Peskin & Schroeder. The process can be formulated as a two-photon
scattering, with one photon being hard and the other one play the role of proton, which
has the internal structure as shown in (a). Therefore, we can write down the corresponding
current product as Z
W = i d4 x eiqx h|T{J (x)J (0)}|i, (18.77)
which can be again expanded in terms of scalar form factors,

qq P q P q
W = g + 2 W1 + P q 2 P q 2 W2 . (18.78)
q q q
After operator product expansion, the form factor W2 can be expressed as
X X 8 (2q P )n2
W2 = 3 Q2q 2 2 )n2
Anq (Q2 ), (18.79)
q n
Q (Q
and Anq (Q2 ) is a scale-dependent quantity, whose scaling behavior is dictated by the anoma-
lous dimension matrix . This matrix can be evaluated again from the diagrams in Fig. 18.13
in Peskin & Schroeder and in figures of last problem. The only difference is that we should
replace the gluon field with photon field. Therefore it is straightforward to see that an = 0.
For anqq and anq , we should take away the group factor C2 (N ) = 4/3, while for anq , we should
take away the factor tr (ta tb ) = ab /2. In addition, we should also include the factor Q2q
corresponding to electric charge of each quark. Then we have,
n
h X 1 2 i
anqq = 2Q2f 1+4 , (18.80)
j=2
j n(n + 1)
8Q2f (n2 + n + 2)
anq = , (18.81)
n(n + 1)(n + 2)
4Q2f (n2 + n + 2)
anq = , (18.82)
n(n2 1)
an = 0. (18.83)
(c) The n = 2 moment photon structure function can be worked out through the moment
sum rules (18.154) in Peskin & Schroeder, where the matrix elements Anq in our case is
a scale-dependent quantity. This dependence can be found by evaluating the anomalous
(2)
dimension matrix of operator Oq as is done below (18.185) of Peskin & Schroeder, but
with different entries, given by

64 32
a2uu 0 3 2a2u 27 0 27
16 8
= 0 a2dd 3 2a2u = 0 27 . (18.84)

4 4 27
2 2 2 64 16
au ad a 27 27
0
(d) As can be inferred from (a), the photon structure function f (x, Q) is originally peaked
at x = 1 for Q = Q0 , and the peak shifts toward smaller x and the peak goes lower and
broader, as Q goes large from Q0 .
Chapter 19
Perturbation Theory Anomalies
19.1 Fermion number nonconservation in parallel E

and B fields
(a) In this problem we investigate the effect of chiral anomaly on the (non)conservation
of fermion number with definite chirality. Let us begin with the Adler-Bell-Jackiw anomaly
equation,
e2
j 5 = F F . (19.1)
16 2
Integrating the left hand side over the whole spacetime, we get the difference between the
numbers of right-handed fermions NR and of left-handed fermions NL , namely,
Z Z Z t2

4 5 4
d x j = d x (jR jL ) = d3 x (jR0 jL0 ) = NR NL , (19.2)

t1
where we assume that the integral region for time is [t1 , t2 ] and that i j i integrates to zero
with suitable boundary conditions (i.e. vanishing at spatial infinity or periodic boundary
condition). On the other hand,
F F = 40ijk F0i Fjk = 8F0i 1

F
2 ijk jk
= 8E B. (19.3)
Therefore, the ABJ anomaly equation gives,
e2
Z
NR NL = 2 d4 x E B. (19.4)
2
(b) The Hamiltonian for massless charged fermions with background electromagnetic field
is given by
Z Z
H = d x D0 L = d3 x i i Di ,
3

(19.5)
where = i is the canonical conjugate momentum of , L = i D

/ is the Lagrangian for
the fermion, and D = +ieA is the covariant derivative. Now we expand the Hamiltonian
159
160 Chapter 19. Perturbation Theory Anomalies
in the chiral basis,

! !
Z i D 0 L
H = d3 x L R
0 i D R
Z h i
= d3 x L (i D)L R (i D)R . (19.6)
(c) Now we focus on the eigenvalue problem of the right-handed fermion R , namely the
equation i DR = ER . To be definite, we set the background electromagnetic potential
to be A = (0, 0, Bx1 , A) with B and A two constants. To seek the eigenfunction of the
T 2 3
form R = 1 (x1 ), 2 (x1 ) ei(k2 x +k3 x ) , we substitute it into the equation above and get
01 = (k2 eBx1 )1 + i(E + k3 eA)2 , (19.7a)

02 = i(E k3 + eA)1 (k2 eBx1 )2 . (19.7b)
Eliminating 2 from these two equations, we get a single differential equation in the form of
the harmonic oscillator,
h k2 2 i
001 e2 B 2 x1 E 2 + (k3 eA)2 eB 1 = 0. (19.8)
eB
(d) Now we specify the spatial boundary condition to be the box of length L in each side
and periodic boundary condition. Then the condition R (x1 , x2 , x3 ) = R (x1 , x2 + L, x3 ) =
R (x1 , x2 , x3 + L) implies that k2 and k3 are quantized according to ki = 2ni /L (i = 2, 3).
On the other hand, k2 also has an upper bound since (19.8) shows that the center of the
oscillator would be out of the box if k2 is too large. This condition implies that k2 /eB < L,
which further gives the maximum value of n2 to be (n2 )max = eBL2 /2. Note also that
the energy eigenvalue does not depend on k2 , thus each energy level consists of eBL2 /2
degenerate states. Furthermore, we can also write down explicitly the energy eigenvalue
associated with the state labeled by (n1 , n3 ):
1/2
2n3 2
3
E= eA (n1 + 2
)eB . (19.9)
L
(e) Now we consider the case with n1 = 0 for simplicity. Then the spectrum reads E =
2n3 /L eA. Suppose that the background potential changes by A = 2/eL. Then it is
easy to see that all state with energy marked by n3 will turn to states with energy marked by
n3 1. Note that each energy eigenvalue is eBL2 /2-degenerate, thus the net change of right-
handed fermion number is eBL2 /2. Similar analysis shows that the left-handed fermion
number get changed by eBL2 /2. Therefore the total change is NR NL = eBL2 /.
19.2. Weak decay of the pion 161
19.2 Weak decay of the pion

(a) In this problem we study the decay of charged pion. So let us work out the amplitude
for + `, with the effective four-fermion interaction
4GF
L = (`L L )(uL dL ) + h.c. (19.10)
2
and the relation
h0|j 5a (x)| b (p)i = ip f ab eipx (19.11)
as inputs. Firstly we recall that
j a = QL a QL + QR a QR , (19.12a)
j 5a = QL a QL + QR a QR , (19.12b)
where QL = (uL , dL )T and QR = (uR , dR )T . Thus,

1 1
(j + ij 2 j 51 ij 52 ) = QL ( 1 + i 2 )QL = uL dL . (19.13)
2

Then we find the decay amplitude M + (p) `+ (k)(q) to be
4iGF 1 1
5
iM = u(q) v(k) f ip = GF f u(q)p/(1 5 )v(k). (19.14)
2 2 2
(b) Now let us calculate the decay rate of the charged pion. We note that the amplitude
above can be further simplified to
iM = GF f u(q)(/q + k/)(1 5 )v(k) = GF f m` u(q)(1 + 5 )v(k). (19.15)
Therefore we have
X
|M|2 = G2F f2 m2` tr /q(1 + 5 )(k/ m` )(1 5 ) = 8G2F f2 m2` q k,

(19.16)
where the summation goes over all final spins. We choose the momenta to be
p = (m , 0, 0, 0), k = (Ek , 0, 0, k), q = (Eq , 0, 0, k). (19.17)
Then the kinematics can be easily worked out to be

m2 + m2` m2 m2`
Ek = , Eq = k = (19.18)
2m2 2m
The decay rate then follows straightforwardly,
d k 2 k
Z
1 k 1
= + 8G2F f2 m2` (q k)
2m 16 2 Ek Eq Ek Eq
G2 f 2 m` 2 2
= F (m m2` )2 , (19.19)
4m m
and we have the ratio between two decay channels,
( + e+ e ) m2e (m2 m2e )2

= ' 104 . (19.20)
( + + ) m2 (m2 m2 )2
Thus to determine the pion decay constant f , we can consider the channel + only as
a good approximation. With the lifetime of charged pion = 2.6 108 sec as well as m
and m , we find that
s
4m m 2
f = (m m2 )1 ' 90.6MeV. (19.21)
G2F m
19.3 Computation of anomaly coefficients

(a) By definition, Aabc = tr [ta , {tb , tc }]. Then for the product representation r1 r2 , we
have
h i
abc a a
b b c c
A (r1 r2 ) = tr r1 r2 t 1 + 1 t , t 1 + 1 t , t 1 + 1 t
h i
= tr r1 r2 ta 1 + 1 ta , {tb , tc } 1 + tb tc + tc tb + 1 {tb , tc }

= tr r1 r2 [ta {tb , tc }] 1 + [ta , tb ] tc + [ta , tc ] tb

+ tb [ta , tc ] + tb [ta , tc ] + 1 [ta , {tb , tc }]
= tr r1 [ta , {tb , tc }] tr r2 (1) + tr r2 [ta , {tb , tc }] tr r1 (1)
= Aabc (r1 )d(r2 ) + Aabc (r2 )d(r1 ). (19.22)
On the other hand, as we decompose the representation r1 r2 into a direct product of

P
irreducible representations ri , we have
i
P h X n X X oi XXX
abc a b c a b c
A ri = tr r ti , tj , tk = tr r [ti , {tj , tk }]
i i j k i j k
X X
= tr ri [tai , {tbi , tci }] = Aabc (ri ) (19.23)
i i
Note that Aabc (r) = 21 A(r)dabc where dabc is the unique symmetric gauge invariant. Then
equating the two expressions above, we get
X
d(r2 )A(r1 ) + d(r1 )A(r2 ) = A(ri ). (19.24)
i
(b) In this part we show that the representation (3 3)a of SU (3) is equivalent to 3.
Let i be the base vectors of 3 representation. Then, a set of base vectors of (3 3)a
can be chosen to be ijk j k . From the transformation rule i Uij j , we know that
the (3 3)a base vectors transform according to ijk j k imn Umj Unk j k . Now, it
19.3. Computation of anomaly coefficients 163
is easy to show that `mn Uì Umj Unk is totally antisymmetric, and thus is proportional to
ijk . Let us write `mn Uì Umj Unk = Cijk , then taking U = I shows that C = 1. Now
we multiply both sides of this equality by (U )ip . Since U is unitary, (U )ip = (U 1 )ip , so
we get pmn Umj Unk = ijk (U )ip . That is, the base vector ijk Uj Uk transforms according to
ijk Uj Uk (U )ì `jk j k = (U )i` `jk j k , which is exactly the transformation rule of 3.
Now from A(3) = 1, it follows that A(3) = 1. Therefore A((3 3)a ) = 1, and by
using the equation derived in (a), we have A((3 3)s ) = 6 (1) = 7.
(c) Now we compute the anomaly coefficients for a and s representations of the SU (N )
group. As indicated in Peskin & Schroeder, it is enough to consider an SU (3) subgroup
of SU (N ). Then the fundamental representation N is decomposed into a direct sum of
irreducible representations when restricted to SU (3), that is, N = 3 + (N 3)1. This
decomposition is easily justified by considering the upper-left 3 3 block of a matrix in
fundamental representation of SU (N ). When this block is treated as a transformation of
SU (3), the first three components of the vector on which the matrix acts form a fundamental
representation vector of SU (3), while the other (N 3) components of the column vector
are obviously invariant. With this known, we have,
N N = 3 + (N 3)1 3 + (N 3)1 = 3 3 + 2(N 3)3 + (N 3)2 1. (19.25)

On the other hand, we know that N = s + a while s and a are irreducible. Then we have, by
(a), 2N A(N) = A(s)+A(a). But we already know that A(N) = 1. Thus A(s)+A(a) = 2N .
Now, to compute A(a), we make use of the SU (3) restriction,
1
(N N)a = (3 3)a + (N 3)3 + 2
(N 3)(N 4)1. (19.26)
Then,

A(a) = A (3 3)a + (N 3)A(3) = A(3) + (N 3)A(3) = N 4, (19.27)
and A(s) = 2N A(a) = N + 4.

Now consider totally antisymmetric rank-j tensor representation. Again we decompose
the fundamental representation as N = 3 + (n 3)1. Then the rank-j totally antisymmetric
tensor can be decomposed as
(N j) (N j + 1) (N 3) (N j)
(Nj )a = (3 3 3)a + (3 3)a
(j 3)! (j 2)!
(N 3) (N j 1)
+ 3 + 1s. (19.28)
(j 1)!
Therefore,
(N 3) (N j) (N 3) (N j 1)
A(a) = +
(j 2)! (j 1)!
(N 3) (N j)(N 2j)
= . (19.29)
(j 1)!
19.4 Large fermion mass limits

In this problem we study the chiral anomaly and the trace anomaly in triangle diagrams
with Pauli-Villars regularization.
(a) and (c) Firstly we evaluate the expectation value of the divergence of the chiral current
j 5 between the vacuum and the two-photon state, namely the matrix element hp, k|j 5 |0i.
This matrix element receives contributions at 1-loop level from the following two diagrams:
In momentum space, the divergence of the first diagram reads

d4 `
Z
2 5 i i i
iq M1 = (1)(ie) tr /q
(2)4 / k/ / / + p/

5 i i i
tr /q (19.30)
/ k/ M / M / + p/ M
The integral is finite, thus we are allowed to shift the integral variable. For the first trace
and the second trace above, we rewrite the /q 5 factors, respectively, as follows,
5 5
/q = (/ + p/ / + k/) = (/ + p/) + (/ k/),
5 5 5
/q = (/ + p/ M / + k/ + M ) = (/ + p/ M ) + (/ k/ M ) + 2M .
Then, the loop integral becomes

d4 `
Z
2 5 1 1 5 1 1
iq M1 = e tr +
(2)4 / k/ / / / + p/

5 1 1 5 1 1
tr +
/ k/ M / M / M / + p/ M

5 1 1 1
+ 2M tr (19.31)
/ k/ M / M / + p/ M
In the expression above, the first and the second lines are canceled by the corresponding
terms from the second diagram with (k, p, ), while the third line is doubled. Therefore
the sum of two diagrams gives
d4 `
Z
2 5 1 1 1
iq M = 4e M tr
(2)4 / k/ M / + M / + p/ M
Z 1 Z 1x
d4 `
Z
2 2N1
= 4e M 4
dx dy 3
(2) 0 0 (` xk + yp)2
4ie2 M N1 1
Z Z 1x
1
= dx dy (19.32)
(4)2 0 0
19.4. Large fermion mass limits 165
with
N1 = tr 5 (/ k/ + M ) (/ M ) (/ + p/ + M ) = 4iM k p ,

= M 2 x(1 x)k 2 y(1 y)p2 2xyk p.

Then the integral can be carried out directly in the M 2 limit, to be
e2
iq M = k p , (19.33)
2 2
as expected.
(b) and (d) For scale anomaly, the diagrams are the same. Now the relevant matrix
element is given by hp, k|M |0i. Then the first diagram reads
d4 `
Z
2 1 1 1
iM1 (p) (k) = ie M / (k) / (p)
(2)4 / k/ k/ / + p/

1 1 1
tr / (k) / (p) (19.34)
/ k/ M / M / + p/ M
The first trace vanishes upon regularization, then,
Z 1 Z 1x
d4 `0
Z
2 2N2
iM1 (p) (k) = ie M 4
dx dy 02 , (19.35)
(2) 0 0 (` )3
where `0 = ` xk + yp, = M 2 2xyk p, and the trace in the numerator is
N2 = tr (/ k + M )/ (k)(/ + M )/ (p)(/ + p/ + M )

= 4M M 2 (k) (p) + (k) p (p) k (k) (p) k p

+ 4 (k) ` (p) ` (k) (p) `2

= 4M M 2 (k) (p) + (1 4xy) (k) p (p) k

(1 2xy) (k) (p) (k p) + d4 1 (k) (p) `02 ,

where we used the transverse condition k (k) = p (p) = 0, and in the last equality,
the substitution `0 `0 d1 g `02 . We also dropped all terms linear in `0 in the last equality.
The integral is then divergent, and we regularize it by dimensional regularization. Then
after carrying out the loop integral, we get
e2

iM1 (p) (k) = (k) (p) (k p) (k) p (p) k
4 2Z
1 Z 1x
(1 4xy)M 2
dx dy 2 (19.36)
0 0 M 2xyk p
Then, taking M 2 limit, we find
e2

iM1 (p) (k) = (k) (p) (k p) (k) p (p) k (19.37)
12 2
The second diagram is obtained, again, by the exchange (k, p, ), which gives the
identical result. Therefore we finally get
e2
iM (p) (k) = 2 (k) (p) (k p) (k) p (p) k .

(19.38)
6
Chapter 20
Gauge Theories with Spontaneous

Symmetry Breaking
20.1 Spontaneous breaking of SU (5)

We consider two patterns of spontaneous breaking of SU (5) gauge symmetry, with an
adjoint-representation scalar field picking up vacuum expectation values
hi = Adiag(1, 1, 1, 1, 4), hi = Bdiag(2, 2, 2, 3, 3), (20.1)
respectively. The kinetic term of the scalar field in the Lagrangian is

Lkin. = tr (D ) (D ) = tr + g[A , ] + g[A , ] . (20.2)
Then the mass term of gauge bosons after symmetry breaking is given by

L = g 2 tr [A , ] [A , ] = g 2 Aa Ab tr [T a , hi][T b , hi] . (20.3)
To analyze the gauge bosons spectrum, we note that there are 24 independent generators
for SU (5) group, each of which can be represented as a 5 5 traceless hermitian matrix.
Then, for the first choice of hi = diag(1, 1, 1, 1, 4), we see that for the generators of the
form
!
(4)
T
T = and T = 2110 diag(1, 1, 1, 1, 4),
0
where T (4) is a 4 4 matrix being any generator of SU (4) group, the commutators vanish.
That is, a subgroup SU (4) U (1) remains unbroken in this case. Then, for the rest of the
generators, namely

0 0 0 0 1 0 0 0 0 i 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
1
1
1
0 0 0 0 0 , 0 0 0 0 0 , 0 0 0 0 0 ,

2 2 2
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 0 0 0 i 0 0 0 0 0 1 0 0 0
167
168 Chapter 20. Gauge Theories with Spontaneous Symmetry Breaking
etc, it is easy to calculate the commutators to get the trace equal to 25A2 /2. Thus the
corresponding components of gauge bosons acquire mass MA = 5gA. In the same way, we
can also analyze the case of hi = diag(2, 2, 2, 3, 3). This time the unbroken subgroup is
SU (3) SU (2) U (1), and the remaining 12 components of gauge bosons acquire a mass
equal to MA = 5gB, as can be found by evaluating the corresponding commutators.
20.2 Decay modes of the W and Z bosons

(a) The relevant interaction term in the Lagrangian reads
X X
1 + c c
L = 2 gW iL eiL + ujL djL , (20.4)
i j,c
where the sum on i goes over all three generations of leptons, the sum on j goes over the
first two generations of quarks, since mt > mW , and the sum on c is due to 3 colors.
Now consider the decay of W + boson. The amplitude of the decay into a pair of fermions
is
5
ig 1

iM = (k)u(p1 ) v(p2 ), (20.5)
2 2
where is the polarization vector for W+ , and the labels for momenta are shown in Fig.
20.1. Thus the squared amplitude with initial polarizations averaged is
f f f f
p1 p2 p1 p2
k k
W+ Z0
Figure 20.1: The decay of W+
and Z0
into fermion-antifermion pairs. All initial momenta go
inward and all final momenta go outward.
g2 5 5

1 X k k 1 1

2
|iM| = g + 2 tr p/2 p/1
3 spin 6 mW 2 2
g2

k p1 k p2
= p1 p2 + 2 . (20.6)
3 m2W
The momenta in the center-of-mass frame can be taken to be
k = (mW , 0, 0, 0), p1 = (p, 0, 0, p), p2 = (p, 0, 0, p), (20.7)
and energy conservation requires that p = mW /2. Thus we get

1 X 1
|iM|2 = g 2 m2W , (20.8)
3 spin 3
20.2. Decay modes of the W and Z bosons 169
and the decay rate
d 3 p1 d 3 p2
Z Z
1 1 2 2 mW
d = g mW (2)4 (4) (k p1 p2 ) = , (20.9)
2mW 6
(2) 2E1 2E2 3 12 sin2 w
where we have used g = e/ sin w and = e2 /4. For each quark final state we multiply
the result by a QCD correction factor 1 + s . Then, taking account of 3 generations of

leptons and 2 generations of quarks with 3 colors, we get the partial decay rate of W + into
fermions,
mW
(W + e+
i i ) = ' 0.23GeV; (20.10)
12 sin2 w
+ mW s
(W uj dj ) = 1+ ' 0.70GeV; (20.11)
4 sin2 w
+ mW s
(W fermions) = 9+6 ' 2.08GeV. (20.12)
12 sin2 w
and also the branching ratios BR(W + e+ i i ) = 0.11%, and BR(W

+
uj dj ) = 0.34%.
Note that the fine structure constant at mW is (mW ) ' 1/129.
(b) In the same way, we can also calculate the decay rate of Z fermions. The relevant
term in the Lagrangian is
g X
fi Ii3 sin2 w Qi fi ,

L = Z (20.13)
cos w i
where the sum goes over all left- and right-handed fermions, including 3 generations of
leptons, and the first two generations of quarks with 3 colors, while Ii3 and Qi are associated
3-component of the weak isospin and the electric charge, respectively.
Then we can write down the amplitudes of the decay of Z 0 into a pair of fermions f f
with specific I 3 and Q, as illustrated in Fig. 20.3,
1 5 1 + 5

ig 3 2 2
iM = (k)u(p1 ) I sin w Q sin w Q v(p2 )
cos w 2 2
5

ig 3 1 2
= (k)u(p1 ) I sin w Q v(p2 ), (20.14)
cos w 2
the squared matrix elements,
1 X g2 k k
|iM|2 = g +
3 spin 3 cos2 w m2Z

1 3 5 2 1 3 5 2
tr p/2 2 I (1 ) sin w Q p/1 2 I (1 ) sin w Q
4g 2 h 1 3

2
2 i
1 3 2 2k p1 k p2
= I sin w Q + 2 I p1 p2 +
3 cos w2 2 m2Z
4g 2 m2Z h 1 3 2
2 i
1 3 2
= I sin w Q + I , (20.15)
3 cos w2 2 2
and the partial decay rate,

mZ h 2 3 2
i
(Z f f) =
0 1 3 2 1

I sin w Q + I . (20.16)
3 sin2 w cos2 w 2 2
We should also multiply the result for quarks by the QCD factor 1 + s . Now we list

the numerical results of partial width and the branching ratios for various decay products
as follows.
f f (f f)/GeV BR(f f)
e e , , 0.17 6.7%
e e , + , +
+
0.08 3.4%
uu, cc 0.30 11.9%

dd, ss, bb 0.39 15.4%
All fermions 2.51 100%
20.3 e+e hadrons with photon-Z 0 interference

(a) It is easier to work with amplitudes between initial and final fermions with definite
chirality. In this case the relevant amplitude is given by
(Ie3 + s2w )(If3 s2w Qf ) q2

2 i
iM = (ie) v(k2 ) u(k1 ) 2 u(p1 ) Qf v(p2 ) Qf + ,
q s2w c2w q 2 m2Z
(20.17)
where Ie3 = 1/2 or 0 when the initial electron is left-handed or right-handed, so as If3 to the
final fermion. The momenta is labeled as shown in Fig. 20.2. Then we can find associated
f f
p1 p2
q
k1 k2
e e+
Figure 20.2: The process of e+ e f f via the exchange of a photon/Z 0 in s-channel. The
directions of ki s and pi s are inward and outward, respectively.
differential cross section to be

d 2
(e+ e fR fL ) = (1 + cos )2 FLL (f ), (20.18a)
d cos R L 2s
d 2
(e+
R e L fL fR ) = (1 cos )2 FLR (f ), (20.18b)
d cos 2s
d 2
(e+ e fR fL ) = (1 cos )2 FRL (f ), (20.18c)
d cos L R 2s

In this problem we simplify the notation by sw sin w and cw cos w .
20.3. e+ e hadrons with photon-Z 0 interference 171
2
d L fR ) = (1 + cos )2 FRR (f ),
(e+ e f (20.18d)
d cos L R 2s
in which is the fine structure constant, s = q 2 is the center-of-mass energy, and the F
factors are defined as follows:
2
( 12 s2w )(If3 s2w Qf )

s
FLL (f ) = Qf + 2 2 2
, (20.19)
sw cw s mZ + imZ Z
2
( 21 s2w )Qf

s
FLR (f ) = Qf , (20.20)
c2w s m2Z + imZ Z
2
(If3 s2w Qf )

s
FRL (f ) = Qf , (20.21)
c2w s m2Z + imZ Z
2
s2w Qf

s
FRR (f ) = Qf + 2 2
, (20.22)
c w s m + imZ Z
Z
where we have added the correction from resonance by using the Breit-Wigner formula.
Summing up the four expressions in (20.18), averaging the initial spins, and integrating over
the angle , we get finally the unpolarized cross section
2 h i
(f f) = FLL (f ) + FLR (f ) + FRL (f ) + FRR (f ) . (20.23)
3s
When the final state particle f is a quark, one should multiply the result by 3 1+ s where

3 is the color factor, and 1 + s is the 1-loop QCD correction.

For the final fermion being muon (If3 = 1/2, Qf = 1), up quark (If3 = 1/2, Q = 2/3),
and down quark (If3 = 1/2, Qf = 1/3), we plot the corresponding cross section as a

function of center-of-mass energy ECM = s in Fig. 20.3.
104
1000
pb
100
10
50 100 150 200

ECM GeV
Figure 20.3: The cross section (e+ e f f) as a function of center-of-mass energy ECM . The
black, blue, and red curves correspond to f f = + , uu and dd,
respectively.
(b) Now we calculate the forward-backward asymmetry AfF B , defined to be

R1 R0
F B d cos (d/d cos )
f
AF B = = R01 R1 0 . (20.24)
F + B + d cos (d/d cos )
0 1
Then from (20.18), we find

2 h i
F = 7FLL (f ) + FLR (f ) + FRL (f ) + 7FRR (f ) , (20.25)
24s
2 h i
B = FLL (f ) + 7FLR (f ) + 7FRL (f ) + FRR (f ) . (20.26)
24s
Thus
3 FLL (f ) FLR (f ) FRL (f ) + FRR (f )
AfF B = . (20.27)
4 FLL (f ) + FLR (f ) + FRL (f ) + FRR (f )
Again, we plot AfF B , as a function of ECM , for f = , u, d, in Fig. 20.4.
0.6
0.4
0.2
AFB
0.0
-0.2
-0.4
-0.6
MZ
40 60 80 100 120 140
ECM GeV
Figure 20.4: The forward-backward asymmetry AfF B as a function of center-of-mass energy ECM .
The black, blue, and red curves correspond to f f = + , uu and dd,
respectively.
(c) Recall the definition of F s, we find, on the Z 0 resonance (s = mZ ),

1 2
3 2
2 1 2
2
2
s w If s w Q f mZ 2
s w Qf mZ
FLL (f ) ' , FLR (f ) ' ,
s2w c2w Z c2w Z
If s2w Qf mZ 2
3 2 2
sw Qf mZ
FRL (f ) ' , FRR (f ) ' ,
c2w Z c2w Z
therefore,
1
2 2
3 s2w s4w If3 s2w Qf (s2w Qf )2 3 e f
AfF B = 2
= A A . (20.28)
4 LR LR
2 2
4 1

2
s2w + s4w If3 s2w Qf + (s2w Qf )2
(d)
2 1 m2Z h 1 2 2
4
ih
3 2
2 2 2
i
peak = s w + s w I f s w Qf + (s w Qf )
3m2Z s4w c4w 2Z 2
i
12 mZ h 1 2 2
4
i mZ h 3 2
2 2 2
= 2 2 sw + sw If sw Qf + (sw Qf )
mZ Z 6s2w c2w 2 6s2w c2w
12 (Z 0 e+ e )(Z 0 f f)
= . (20.29)
m2Z 2Z
20.4. Neutral-current deep inelastic scattering 173
20.4 Neutral-current deep inelastic scattering

(a) In this problem we study the neutral-current deep inelastic scattering. The process is
mediated by Z 0 boson. Assuming mZ is much larger than the energy scale of the scattering
process, we can write down the corresponding effective operators, from the neutral-current
Feynman rules in electroweak theory,
g2
h
4 2
4 2

L = PL u 1 3 sw PL 3 sw PR u
4m2W
i
(1 2 s2 )PL 2 s2 PR d + h.c.,
+ d (20.30)
3 w 3 w
where PL = (1 5 )/2 and PR = (1+ 5 )/2 are left- and right-handed projectors, respectively.
Compare the effective operator with the charged-operator in (17.31) of Peskin & Schroeder,
we can write down directly the differential cross section by modifying (17.35) in Peskin &
Schroeder properly, as
d2 G2 sx nh 2 i
(p X) = F 1 34 s2w + 16 s
9 w
4
(1 y 2
) fu (x)
dxdy 4
h 2 i
+ 1 32 s2w + 49 s4w (1 y 2 ) fd (x)
h i
16 4 4 2 2 2

+ 9 sw + 1 3 sw (1 y ) fu (x)
h 2 i o
+ 49 s4w + 1 32 s2w (1 y 2 ) fd(x) , (20.31)
d2 G2 sx nh 16 4 4 2 2
i
(p X) = F 2

s
9 w
+ 1 s
3 w
(1 y ) fu (x)
dxdy 4
h i
4 4 2 2 2 2

+ 9 sw + 1 3 sw (1 y ) fd (x)
h 2 i
+ 1 34 s2w + 16 s
9 w
4
(1 y 2
) fu (x)
h 2 i o
+ 1 32 s2w + 49 s4w (1 y 2 ) fd(x) . (20.32)
(b) For the neutrino scattering from a nucleus A with equal numbers of protons and
neutrons, we have fu = fd and fu = fd. Then the differential cross sections reads
d2 G2 sx n 1
(A X) = F s2w + 59 s44 + 95 s44 (1 y 2 ) fu (x)

2
dxdy
o
+ 95 s4w + 21 s2w + 59 s4w (1 y 2 ) fu (x) ,

(20.33)
d2 G2F sx n 5 4 1 2 5 4
2

(p X) = 9
s w + 2
s w + 9
s w (1 y ) fu (x)
dxdy
o
2 5 4 5 4 2
1
+ 2 sw + 9 s4 + 9 s4 (1 y ) fu (x) . (20.34)
Recall that for charged-current neutrino deep inelastic scattering, the differential cross sec-
tions are given by (17.35) in Peskin & Schroeder. Thus it is easy to find that
d2 /dxdy(A X) fu (x)(1 y 2 ) + fu (x)

1 2 5 4
R = = sw + sw 1 + , (20.35)
d /dxdy(A X) 2 9 fu (x) + fu (x)(1 y)2
1.0
0.8
R
0.6
0.4
sin2 w = 0.23
0.2
0.0 0.1 0.2 0.3 0.4 0.5
R
Figure 20.5: Weinbergs nose with r = 0.4. See problem 21.4.
d2 /dxdy(A X) fu (x) + fu (x)(1 y)2

u 1 2 5 4
R = = sw + sw 1 + , (20.36)
d /dxdy(A + X) 2 9 fu (x)(1 y 2 ) + fu (x)
where
fu (x)(1 y 2 ) + fu (x)
= r. (20.37)
fu (x) + fu (x)(1 y)2
(c) The plot Weinbergs Nose with r = 0.4 is shown in Figure 20.5.
20.5 A model with two Higgs fields

(a) The gauge boson mass matrix comes from the kinetic term of scalar fields,
(D 1 ) (D 1 ) + (D 2 ) (D 2 ),
with D 1,2 = 2i gAa a 2i g 0 B 1,2 . After 1,2 acquire the vacuum expectation value

1 0

2 v1,2
, we observe that each of the kinetic terms gives rise to mass terms for gauge bosons
similar to the ones in the standard electroweak theory. Thus it is straightforward that the
masses of gauge bosons in this model is given by the replacement v 2 v12 + v22 .
0
(b) The statement that the configuration 12 v1,2

is a locally stable minimum, is equivalent
to that all particle excitations generated above this solution have positive squared mass m2 .
0
Thus we investigate the mass spectrum of the theory with the vacuum chosen to be 12 v1,2

.
Firstly, we parameterize two scalar doublets as
i+ ,

i = 1 , (i = 1, 2) (20.38)
(vi + hi + i 0 )
2 i
and substitute this parameterization into the potential,
V = 21 1 1 22 2 2 + 1 (1 )2 + 2 (2 2 )2
+ 3 (1 1 )(2 2 ) + 4 (1 2 )(2 1 ) + 5 (1 2 )2 + h.c. .

(20.39)
20.5. A model with two Higgs fields 175
Then the mass term of various scalar components can be extracted, as follows.
! !
+
v /v
2 1 1
Lmass = (4 + 25 )v1 v2 1 2 1
1 v1 /v2 +2
! !
v /v
2 1 1 01
+ 25 v1 v2 10 20
1 v1 /v2 02
! !
1 (v1 /v2 ) 3 + 4 + 25 h1
v1 v2 h1 h2 . (20.40)
3 + 4 + 25 2 (v2 /v1 ) h2
The eigenvalues of these matrices are easy to be found. For charged components, there
is a zero mode corresponding two broken directions in SU (2), and the mass of the other
charged scalar is given by m2c = (4 + 25 )(v12 + v22 ). For pseudoscalar components, there
is also a zero mode corresponding to the rest one direction of broken SU (2), and the mass
of the other pseudoscalar is m2p = 45 (v12 + v22 ). Finally, for neutral scalars, the two mass
eigenvalues are given by the roots of following equation,
m2n (1 v12 + 2 v22 )m2n + 1 2 (3 + 4 + 25 )2 = 0.

(20.41)
Therefore, to make m2c > 0, m2p > 0 and m2n > 0, it is sufficient that
4 + 25 < 0, 5 < 0, 1 , 2 > 0, 1 2 > (3 + 4 + 25 )2 . (20.42)
(c) From the mass terms in (b) we can diagonalize the charged scalar mass matrix with
the rotation matrix ! ! !
+ cos sin 1+
= , (20.43)
+ sin cos 2+
where + is the Goldstone mode and + is a physical charged scalar. Given that + to get
the physical mass, it is easy to see that the rotation angle can be chosen to be tan = v2 /v1 .
(d) Assuming that the Yukawa interactions between quarks and scalars take the following
form,
" ! ! #
+ 1 v2

Lm = uL dL d 1 1
dR + u 2

uR + h.c., (20.44)
v1
2
where we have suppressed flavor indices and neglected neutral scalar components. We
focus on charged component only. Then, with Peskin & Schroeders notation, we make the
replacement uL Uu uL , dL Ud dL , uR Wu uR , and dR Wd dR . Then, together with
d = Ud Dd Wd and u = Uu Du Wu where Dd and Du are diagonal matrix, we have
1
Lm = v1 dL Dd dR + v2 uL Du uR

2

uVCKM Dd dR 1+ + dL VCKM Du uR 2 + h.c. (20.45)
From the first line we see that the diagonal mass matrix for quarks are given by mu =
p
(v1 / 2)Du and md = (v2 / 2)Dd . We further define v = v12 + v22 and note that 1+ =
+ sin , 2+ = + cos + , then the Yukawa interactions between charged boson
and quarks can be written as

2

Lm uL VCKM md dR 1+ + dL VCKM mu uR 2 + h.c.
v
1
2

uL VCKM md dR + tan + dL VCKM mu uR cot + h.c.. (20.46)
v
Chapter 21
Quantization of Spontaneously
Broken Gauge Theories
21.1 Weak-interaction contributions to the muon g 2

In this problem we study the weak-interaction corrections to the muons anomalous
magnetic moment (AMM). The relevant contributions come from the W -neutrino loop and
Z-muon loop, together with the diagrams with the gauge bosons replaced with the cor-
responding Goldstone bosons. Here we will evaluate the W -neutrino loop diagram with
Feynman-t Hooft gauge and general R gauge in part (a) and part (b) respectively, and
Z-muon diagram in part (c).
(a) Now we come to the W -neutrino loop diagram and the corresponding Goldstone boson
diagrams, shown in Fig. 21.1.
+ + + +
k.
q q q q
q+k &

(a) (b) (c) (d)
Figure 21.1: The weak-interaction contributions to muons EM vertex. These four diagrams
contain neutrino internal lines in the loops.
The Fig. 21.1(a) with W -neutrino loop reads

(ig)2 d4 k
Z
(a)

(q) = g (2k + q) + g (2q k) + g (q k)
2 (2)4
5 5
ig ig 0 1
i 1

2 u(p ) u(p)
k m2W (q + k)2 m2W 2 p/0 + k/ 2
Z 1 Z 1x
ig 2 d4 k 0
Z
2 0
h
= 4
dx dy 02 3
u(p ) (2k + q) (p/0 + k/)
2 (2) 0 0 (k )
177
178 Chapter 21. Quantization of Spontaneously Broken Gauge Theories
i 1 5
+ (k/ 2/q)(p/0 + k/) + (p/0 + k/)(/q k/) u(p), (21.1)
2
where
k 0 = k + xq + yp0 ,
= (1 y)m2W x(1 x)q 2 y(1 y)p02 + 2xyq p0 .
To extract the form factor F2 (q 2 ), recall that the total diagram can be written as a linear
combination of (p0 + p) , q , , and parity-violating terms containing 5 . Only the (p0 + p)
terms contribute to F2 (q 2 ) through the Gordon identity. With this in mind, now we try to
simplify the expression in the square bracket in (21.1), during which we will drops terms
proportional to q or freely, and totally ignore the 5 terms.
h i h i h i
0 0 0
(2k + q) (p/ + k ) + (k 2/q)(p/ + k ) + (p/ + k )(/q k )
/ / / / /
h 0 i
= 2 2k 0 + (1 2x)q 2yp0 k/ x/q + (1 y)p/0
h i
0 0
+ k/ (2 x)/q + y p/0 k/ x/q + (1 y)p/0
h i
0 0
0 0
+ k/ x/q + (1 y)p/ k/ + (1 + x)/q + y p/
h i h i h i
4y(1 y)mp0 + 2(x + 2y 2)mp + 2(1 x + y)mp0
(1 y)(3 2y)m(p0 + p)
i q
2(1 y)(3 2y)m2 .
2m
The steps of this calculation is basically in parallel with the one of Problem 7.2. Here we have
written the mass of muon as m instead of m to avoid confusions. Thus the contribution to
the muons AMM from Fig. 21.1(a) is
Z 1 Z 1x
ig 2 d4 k
Z
2 1
4
dx dy 02 3
2(1 y)(3 2y)m2
2 (2) 0 0 (k ) 2
2 2 2
7 g m 7 GF m
' 2 2
= , (21.2)
3 64 mW 3 8 2 2
where we have used the approximation mW m, and set q 2 = 0 in the second line. The

Fermi constant GF / 2 = g 2 /8m2W .
Fig. 21.1(b) and 21.1(c) read

ig mW i 2gm d4 k ig
Z
(b) i
(q)= g
2 2 mW (2) 4 k mW (q + k)2 m2W
2 2
1 5 i 5
1

0
u(p ) u(p). (21.3)
2 p/0 + k/ 2

ig mW i 2gm d4 k ig
Z
(c) i
(q) = g
2 2 mW (2)4 k 2 mW (q + k)2 m2W
2
21.1. Weak-interaction contributions to the muon g 2 179
1 5 i 1 + 5
u(p0 ) u(p). (21.4)
2 p/0 + k/ 2
Through the calculation similar to that of Fig. 21.1(a), it is easy to show that these two
diagrams contribute the same to the AMM, which reads
1 GF m2
. (21.5)
2 8 2 2
Finally, Fig. 21.1(d) reads

i2gm 2 Z dd k i i
(d)
(q) = (2k q) 2
mW (2) d k mW (q + k) m2W
2 2
1 5 i 1 + 5
u(p0 ) u(p). (21.6)
2 p/0 + k/ 2
But it is not difficult to see that the contribution to the muons AMM from this diagram
is proportional to (m/mW )4 , which can be omitted in the limit mW m, compared with
the other three diagrams. Therefore we conclude that the AMM of the muon contributed
by W -neutrino and corresponding Goldstone bosons 1-loop diagrams is

7 1 1 m2 G m2 10 GF m2
F
a () = + + +O ' . (21.7)
3 2 2 m2W 8 2 2 3 8 2 2
(c) Now we come to the second set of diagrams as shown in Fig. 21.2.
+ +
q q

(a) (b)
Figure 21.2: The weak-interaction contributions to muons EM vertex. These two diagrams
contains no neutrino internal lines.
Firstly the Fig. 21.2(a) reads

ig 2 Z dd k ig
(a)
Z (q) =
4cw (2) (p + k)2 m2Z
d 0
i i
u(p0 ) 4s2w 1 5 4s2w 1 5 u(p)

k/ m /q k/ m
1 1x
ig 2 dd k 0
Z Z Z
2 0
h
2 d
dx dy 02 3
u(p ) (k/ + m) (k/ + /q + m)
16cw (2) 0 0 (k )
i
+ (4s2w 1)2 (k/ m) (k/ + /q m) u(p), (21.8)
where we have omitted terms proportional to 5 , as indicated by sign, and
k 0 = k + xq + yp0 ,
= (1 y)m2 + ym2Z x(1 x)q 2 y(1 y)p02 + 2xyq p0 .
We will again focus only on the terms proportional to (p0 + p) . Then the spinor part can
be reduced to
h i
u(p0 ) (k/ + m) (k/ + /q + m) + (3s2w c2w )2 (k/ m) (k/ + /q m) u(p)
h i i q
2y(3 + y) (4s2w 1)2 2y(1 y) 2m2 u(p0 ) u(p).
2m
Thus the AMM contribute by this diagram is
Z 1 Z 1x
ig 2 d4 k 0 2 2m2 [2y(3 + y) (4s2w 1)2 2y(1 y)]
Z
dx dy
16c2w (2)4 0 0 (k 02 )3
2
GF m 1 h i
= (4s2w 1)2 5 . (21.9)
8 2 3
2
On the other hand, the Fig. 21.2(b) only contributes terms of order m4 /m4W that can be
omitted, as can be seen from the coupling between the Goldstone boson and the muon.
Thus we conclude that the total contribution to a (Z) from the two diagrams in Fig. 21.2
at the leading order is given by (21.9).
21.2 Complete analysis of e+e W +W

In this problem we calculate the amplitude for the process e+ e W + W at tree level
in standard electroweak theory. There are 3 diagrams contributing in total, as shown in
Figure 21.3.
W+ W W+ W
W+ W
p1 p2
Z0
k1 k2
e e+
e e+ e e+
Figure 21.3: The process e e+ W + W at tree level. All initial momenta go inward and all
final momenta go outward.
We will evaluate these diagrams for definite helicities for initial electrons as well as
definite polarizations for final W bosons. The initial and final momenta can be parameterized
as
k1 = (E, 0, 0, E), p1 = (E, p sin , 0, p cos ),

k2 = (E, 0, 0, E), p2 = (E, p sin , 0, p cos ), (21.10)
21.2. Complete analysis of e+ e W + W 181
with E 2 = p2 +m2W , and electron mass ignored. For initial electron and positron, the spinors
with definite helicities can be chosen to be

uL (k1 ) = 2E(0, 1, 0, 0)T , vL (k2 ) = 2E(1, 0, 0, 0)T ,

uR (k1 ) = 2E(0, 0, 1, 0)T , vR (k2 ) = 2E(0, 0, 0, 1)T . (21.11)
For final W bosons, the polarization vectors are
(p1 ) = 1 (0, cos , i, sin ),

2
(p2 ) = 1 (0, cos , i, sin ),
2
+ (p1 ) = 1
(0, cos , i, sin ),
2
+ (p2 ) = 1
(0, cos , i, sin ),
2
L (p1 ) = 1
mW
(p, E sin , 0, E cos ), L (p2 ) = 1
mW
(p, E sin , 0, E cos ). (21.12)
It is easy to see that for initial electron-positron pair, only two helicity states e +
L eR
and e +
R eL contribute nonzero amplitudes. This is because the first two diagrams with s-
channel gauge bosons vanish for the other two possibilities e + +
L eL and eR eR due to angular
momentum conservation, while the third diagram vanishes since the weak coupling vanishes
for right-handed electron and left-handed positron. With this known, we can write down
the amplitudes for e + +
L eR and eR eL initial states, as follows. Generally the amplitude reads
iM(e +
L eR W W )
+
ie( 12 + s2w )

i i
= (ie) (ie) + (igcw )
(k1 + k2 )2 cw s w (k1 + k2 )2 m2Z
vL (k2 ) uL (k1 ) (p2 p1 ) + (p1 2p2 ) + (2p1 + p2 )

ig 2
i
+ vL (k2 )
uL (k1 ) (p1 ) (p2 )

2 k/1 p/2
m2Z

1 1
= ie2
vL (k2 ) (p1 ) (p2 )(p/2 p/1 )
s(s m2Z ) 2s2w s MZ2

(p1 + 2p2 ) (p1 )/ (p2 ) + (2p1 + p2 ) (p2 )/ (p1 ) uL (k1 )
ie2 1
vL (k2 )/ (p1 )(k/1 p/2 )/ (p2 )uL (k1 ), (21.13)
2s2w u
and,
iM(e +
R eL W W )
+
ies2w

i i
= (ie) (ie) + (igcw ) vR (k2 ) uR (k1 )
(k1 + k2 )2 cw sw (k1 + k2 )2 m2Z
(p2 p1 ) + (p1 2p2 ) + (2p1 + p2 ) (p1 ) (p2 )

m2Z
= ie 2
2
vR (k2 ) (p1 ) (p2 )(p/2 p/1 ) (p1 + 2p2 ) (p1 )/ (p2 )
s(s mZ )

+ (2p1 + p2 ) (p2 )/ (p1 ) uR (k1 ), (21.14)
In what follows we need the inner products among some of these vectors, as listed below.
2Ep
p1 p 2 = E 2 + p2 p1 0 (p2 ) = p2 0 (p1 ) = ,
mW
E 2 + p2
+ (p1 ) + (p2 ) = (p1 ) (p2 ) = 1, 0 (p1 ) 0 (p2 ) = . (21.15)
m2W
We also need
vL (k2 )p/1 uL (k1 ) = vL (k2 )p/2 uL (k1 ) = 2Ep sin , (21.16)

1 + cos
vL (k2 )/ (p1 )uL (k1 ) = 2E , (21.17)
2
1 cos
vL (k2 )/ (p2 )uL (k1 ) = 2E , (21.18)
2
2E 2 sin
vL (k2 )/0 (p1 )uL (k1 ) = vL (k2 )/0 (p2 )u(k1 ) = , (21.19)
mW
vR (k2 )p/1 uR (k1 ) = vR (k2 )p/2 uR (k1 ) = 2Ep sin , (21.20)
1 + cos
vL (k2 )/ (p1 )uL (k1 ) = 2E , (21.21)
2
1 cos
vL (k2 )/ (p2 )uL (k1 ) = 2E , (21.22)
2
2E 2 sin
vL (k2 )/0 (p1 )uL (k1 ) = vL (k2 )/0 (p2 )u(k1 ) = . (21.23)
mW

We first consider e e
L
+
R W +
W
. In this case we take u(k 1 ) = u L (k 1 ) = 2E(0, 1, 0, 0)T
and v(k2 ) = vL (k2 ) = 2E(0, 0, 1, 0). Then each of the final W particle can have polar-
ization (+, , 0), which gives 9 possible combinations for (W + , W ). Now we evaluate the
corresponding amplitudes in turn.
iM(e + +
L eR W(0) W(0) )
m2Z 4Ep(E 2 + p2 ) 16E 3 p

2 1 1
= ie + sin
s(s m2Z ) 2s2w s MZ2 m2W m2W
ie2 1 2E(3E 2 p + p3 2E 3 cos ) sin
+ 2
2sw u m2W
m2Z

s
= ie2 2 2
(3 2 )
4m s mZ
W
1 2 s 2 4 cos
2 (3 ) + sin (21.24)
2sw 1 + 2 + 2 cos s m2Z 1 + 2 + 2 cos
iM(e + + + +
L eR W(0) W() ) = iM(eL eR W() W(0) )

m2Z 8E 2 p 1 + cos

2 1 1
= ie
s(s m2Z ) 2s2w s MZ2 mW 2
ie2 1 2E 1 + cos
2 (E 2 (2 cos 1) + 2Ep p2 )
2sw u mW 2
21.3. Cross section for du W 183

m2Z 1 2 cos 2 2

2 1 s s 1 + cos
= ie 2
2 + (21.25)
s mZ 2sW s m2Z 1 + 2 + 2 cos mW 2
iM(e + +
L eR W() W() )
m2Z ie2 1

2 1 1
= ie 4Ep sin + 2E(p + E cos ) sin
s(s m2Z ) 2s2w s MZ2 2s2w u
m2Z

2 1 s 2( + cos )
= ie + 2 sin (21.26)
(s m2Z ) 2sw s MZ2 (1 + 2 + 2 cos )
iM(e + +
L eR W() W() )
ie2 1 2 ie2 2(1 cos ) sin

= 2E (1 + cos ) sin = . (21.27)
2s2w u 2s2w (1 + 2 + 2 cos )
Though not manifest, these expressions have correct high energy behavior. To see this,
we note that ' 1 2m2W /s when s m2W . Then, for instance, the amplitude for two
longitudinal W final state becomes
m2Z

+ + 2 s
iM(eL eR W(0) W(0) ) = ie (3 2 )
4m2W s m2Z

1 2 s 2 4 cos
2 (3 ) + sin
2sw 1 + 2 + 2 cos s m2Z 1 + 2 + 2 cos
ie2 (1 + 2 cos ) sin
= 2 + O(1/s). (21.28)
2sW 1 + cos
Then we can plot the azimuthal distribution of the corresponding differential cross section
at s = (1000GeV)2 , as shown in Figure 21.4.
Next we consider the other case with e +
R eL initial state. Now there is no contribution
from u-channel neutrino exchange. The amplitudes for various polarizations of final W pairs
can be worked out to be
s m2Z
iM(e + +
R eL W(0) W(0) ) = ie
2
2 2
( 2 3) sin , (21.29)
s mZ 4mW
iM(e + + + +
R eL W(0) W() ) = iM(eR eL W() W(0) )
2

2 mZ s 1 cos
= ie 2
(21.30)
s mZ mW 2
2
2 mZ
iM(e + +
R eL W() W() ) = ie sin (21.31)
s m2Z
iM(e + +
R eL W() W() ) = 0. (21.32)
21.3 Cross section for du W

In this problem we compute the tree amplitude of du W at high energies so that
the quark masses can be ignored. In this case the left-handed and right-handed spinors
100
10
total
ddcos
H-,+L
1
H0,0L
H+,-L
0.1
H+,0L+H0,-L
H-,0L+H0,+L
0.01
-1.0 -0.5 0.0 0.5 1.0
cos
Figure 21.4: The differential cross section of e e+ W + W with definite helicity as a

function of azimuthal angle at s = (1000GeV)2 .
W
W W
p1 p2 p 1 p 2 p 1 p 2
k1 k2
k1 k2 k1 k2
d u d u
d u
Figure 21.5: The process du W at tree level. All initial momenta go inward and all final
momenta go outward.
decouple and only the amplitudes with dL uR initial state do not vanish. To calculate it, we
firstly work out the kinematics as follows.
k1 = (E, 0, 0, E), p1 = (p, p sin , 0, p cos ),
k2 = (E, 0, 0, E), p2 = (EW , p sin , 0, p cos ), (21.33)
where p = E m2W /4E and EW = E + m2W /4E. The initial spinors of definite helicities are
given by

uL (k1 ) = 2E(0, 1, 0, 0)T , vL (k2 ) = 2E(1, 0, 0, 0)T , (21.34)
while the polarization vectors for final photon and W read
(p1 ) = 1 (0, cos , i, sin ),
2
(p2 ) = 1 (0, cos , i, sin ),
2
L (p2 ) = 1
mW
(p, EW sin , 0, EW cos ). (21.35)
Then the amplitude is given by
ie2 ie2

Ns Nt 2Nu
iM(dL uR W ) = + , (21.36)
2sw s m2W 3 2sw t u
21.4. Dependence of radiative corrections on the Higgs boson mass 185
where
Ns = vL (k2 ) (p1 ) (p2 )(p/1 p/2 ) + (p1 + 2p2 ) (p1 )/ (p2 )

(2p1 + p2 ) (p2 )/ (p1 ) uL (k1 ),

(21.37)

Nt = vL (k2 )/ (p2 )(k/1 p/1 )/ (p1 )uL (k1 ), (21.38)
Nu = vL (k2 )/ (p1 )(k/1 p/2 )/ (p2 )uL (k1 ). (21.39)
Now, using the physical conditions (pi ) pi = 0, k/1 uL (k1 ) = 0 and vL (k2 )k/2 = 0, we can
show that Ns = Nt Nu . In fact,
Ns = vL (k2 ) 2 (p1 ) (p2 )p/1 + 2p2 (p1 )/ (p2 ) 2p1 (p2 )/ (p1 ) uL (k1 ),

Nt = vL (k2 ) 2k1 1 /2 + 2 (p1 ) (p2 )p/1 / (p1 )/ (p2 )p/1 uL (k1 ),

Nu = vL (k2 ) 2k2 1 /2 + 2p1 (p2 )/ (p1 ) / (p1 )/ (p2 )p/1 uL (k1 ).

Then Ns = Nt Nu is manifest. Note further that s m2W = (t + u), we have
ie2

Nt Nu Nt 2Nu
iM(dL uR W ) = +
2sw t+u 3t 3u
ie2 (2t u) Nt Nu
= +
2sw 3(t + u) t u
2
ie Nt Nu
= (1 3 cos ) + . (21.40)
6 2sw t u
One can see clearly from this expression that all helicity amplitudes vanish at cos = 1/3.
(Note that the definition of scattering angle is different from the one in Peskin & Schroeder,
which, in our notation, is .) Then, by including all helicity combinations (6 in total),
we find the differential cross section, as a function of s and , to be
d 2 1 cos 2 x3 + 18x2 + 9x + 24 (x3 14x2 + 9x 8) cos 2

= , (21.41)
d cos 32s2w sin 36(s m2W )
where x m2W /s.
21.4 Dependence of radiative corrections on the Higgs

boson mass
(a) We first analyze the radiative corrections to decay process at 1-loop level with the
Higgs boson in the loop. It is easy to see that if the internal Higgs boson line is attached to
one of the external fermions, the resulted vertex will contribute a factor of mf /v which can
be ignored. Therefore only the vacuum polarization diagrams are relevant, and they should
sum to a gauge invariant result.
(b) Now we compute the vacuum polarization amplitudes of W , Z 0 and photon with
Higgs contribution. We will only consider the pieces proportional to g , namely W W (q 2 ),
ZZ (q 2 ), (q 2 ) and Z (q 2 ). It is easy to show that (q 2 ) and ZZ (q 2 ) receive no
contribution from Higgs boson at 1-loop level, while W W (q 2 ) and ZZ (q 2 ) can be found by
computing the following three diagrams:
h
h h
W+ W+
Z0 Z0
0
Now we compute these three diagrams in turn for W + . The first diagram reads
d4 k i
Z
2 i
(igmW ) g
(2) k mh (q k)2 m2W
4 2 2
(2 d2 )
Z 1
i 2 2
= g mW g dx 2d/2 2 2
(4)2 0 (mW , q )
Z 1
M2

i 2 2
g mW g E + dx log , (21.42)
(4)2 0 (m2W , q 2 )
where (m2W , q 2 ) = xm2W + (1 x)m2h x(1 x)q 2 , E = 2/ + log 4 log M 2 , and M 2
is the subtraction scale. The second one reads
d4 k i
Z
2 i
(ig/2) (2k p) (2k p)
(2) k mh (q k)2 m2W
4 2 2
Z 1
g 2 d4 k 0 (4/d)k 02
Z
g dx
4 (2)4 0 (k 02 (m2W , q 2 ))2
i g 2 1 M2
Z
2 2
g dx 2(mW , q ) E + 1 + log , (21.43)
(4)2 4 0 (m2W , q 2 )
in which we have ignored terms proportional to q q . Then, the last diagram reads
d4 k g 2 d4 k (q k)2 m2W
Z Z
1 2 i 1
(ig /2)g = g
2 (2)4 k 2 m2h 4 (2)4 k 2 m2h (q k)2 m2W
1
g 2 d4 k k 02 + (1 x)2 q 2 m2W
Z Z
= g dx
4 (2)4 0 (k 02 (m2W , q 2 ))2
i g 2 1
Z
2 2 2 2 2
g dx 2(mW , q ) mW + (1 x) q E
(4)2 4 0
M2

2 2 2 2 2 2 2
+ 2(mW , q ) mW + (1 x) q log + (mW , q ) . (21.44)
(m2W , q 2 )
Thus we have, when the three diagrams above are taken into account only,
g2

2 2 1 2
W W (q ) = 3m W + q E
4(4)2 3
Z 1
M2

2 2
2 2 2
+ dx (mW , q ) 3mW + (1 x) q ] log . (21.45)
0 (m2W , q 2 )
21.4. Dependence of radiative corrections on the Higgs boson mass 187
Now we extract Higgs mass contribution from this expression in the large Higgs limit, and
also fix the subtraction point at M 2 = m2W . In this limit we may take (m2W , q 2 ) ' xm2h ,
and log(M 2 /) ' log(m2h /m2W ). We also throw divergent terms with E, which should be
canceled out in the final expression of zeroth order natural relation after including completely
loop diagrams with W , Z, and would-be Goldstone boson internal lines. Then we have
g2 m2h

2 1 2 2 1 2
W W (q ) = m + 3mW + q log 2 . (21.46)
4(4)2 2 h 3 mW
Similarly, we have, for ZZ (q 2 ),
g2 m2h

2 1 2 2 1 2
ZZ (q ) = m + 3mZ + q log 2 . (21.47)
4(4)2 cos2 w 2 h 3 mZ
(c) Now, we derive the zeroth order natural relation given in (21.134) of Peskin & Schroeder,
in the large Higgs mass limit. Note that = Z = 0. Thus,
sin2 w cos2 w ZZ (m2Z )

2 2 2 W W (0)
s sin 0 =
cos2 w sin2 w m2Z m2W
1 + 9 sin2 w m2h
= log , (21.48)
48 cos2 w sin2 w m2W
W W (m2W ) m2W ZZ (m2Z ) 5 m2h
s2W s2 = + = log . (21.49)
m2Z m2Z m2Z 24 m2W
Final Project III
Decays of the Higgs Boson
In this final project, we calculate partial widths of various decay channels of the standard
model Higgs boson. Although a standard-model-Higgs-like boson has been found at the LHC
with mass around 125GeV, it is still instructive to treat the mass of the Higgs boson as a
free parameter in the following calculation.
The main decay modes of Higgs boson include h0 f f with f the standard model
fermions, h0 W + W , h0 Z 0 Z 0 , h0 gg and h0 . The former three processes
appear at the tree level, while the leading order contributions to the latter two processes
are at one-loop level. We will work out the decay widths of these processes in the following.
In this problem we only consider the two-body final states. The calculation of decay
width needs the integral over the phase space of the two-body final states. By momentum
conservation and rotational symmetry, we can always parameterize the momenta of two final
particles in CM frame to be p1 = (E, 0, 0, p) and p2 = (E, 0, 0, p), where E = 12 mh by
energy conservation. Then the amplitude M will have no angular dependence. Then the
phase space integral reads Z
1 p
d2 |M|2 = |M|2 . (21.50)
4 mh
Then the decay width is given by
Z
1 1 p
= d2 |M|2 = |M|2 . (21.51)
2mh 8 m2h
In part (d) of this problem, we will also be dealing with the production of the Higgs
boson from two-gluon initial state, thus we also write down the formula here for the cross
section of the one-body final state from two identical initial particle. This time, the two
ingoing particles have momenta k1 = (E, 0, 0, k) and k2 = (E, 0, 0, k), with E 2 = k 2 + m2i
and 2E = mf where mi and mf are masses of initial particles and final particle, respectively.
The final particle has momentum p = (mf , 0, 0, 0). Then, the cross section is given by
d3 p 1
Z
1
= 3
|M|2 (2)4 (4) (p k1 k2 )
2s (2) 2Ep
1
= |M|2 (2)(2k mf ) = 2
|M|2 (s m2f ), (21.52)
4mf s mf
189
190 Final Project III. Decays of the Higgs Boson
p
where = 1 (4mi /mf )2 is the magnitude of the velocity of the initial particle in the
center-of-mass frame.
(a) The easiest calculation of above processes is h0 f f, where f represents all quarks
and charged leptons. The tree level contribution to this process involves a single Yukawa
vertex only. The corresponding amplitude is given by
imf
iM(h0 f f) = u (p1 )v(p2 ). (21.53)
v
Then it is straightforward to get the squared amplitude with final spins summed to be
X m2f 2m2f
|M(h f f)|2 =
0
(m2h 4m2f ).

tr (p/1 + mf )(p/2 mf ) = (21.54)
v2 v2
In CM frame, the final states momenta can be taken to be p1 = (E, 0, 0, p) and p2 =
(E, 0, 0, p), with E = 21 mh and p2 = E 2 m2f . Then the decay width is given by
mh m2f 4m2f 3/2

1 p
(h f f) =
0 2
|M| = 1 . (21.55)
8 m2h 8v 2 m2h
This expression can be expressed in terms of the fine structure constant , the mass of W
boson mw and Weinberg angle sin w , as
mh m2f 4m2f 3/2

0
(h f f ) = 1 . (21.56)
8 sin2 w m2W m2h
(b) Next we consider the decay of h0 to massive vector bosons W + W and Z 0 Z 0 . The
amplitude for the process h0 W + W is given by
ig g 2 v
iM(h0 W + W ) = (p1 ) (p2 ). (21.57)
2
Then the squared amplitude with final polarizations summed reads
g4v2 p2 p2

X
2 p1 p1
|M| = g g 2
4 m2W mW
4 2
12m4W

mh 4mW
= 1 + . (21.58)
sin2 w m2W m2h m4h
Therefore the decay width is
1 p1 m3h 1 2 1 1/2
(h0 W + W ) = |M|2
= (1 4W + 12W )(1 4W ) , (21.59)
8 m2h 16m2W sin2 w
where we have defined W (mh /mW )2 for brevity. For h0 Z 0 Z 0 process, it can be easily
checked that nothing gets changed in the calculation except that all mW should be replaced
with mZ , while an additional factor 1/2 is needed to account for the identical particles in
final state. Therefore we have
m3h
(h0 Z 0 Z 0 ) = (1 4Z1 + 12Z2 )(1 4Z1 )1/2 , (21.60)
32m2Z sin2 w
where Z (mh /mZ )2 .

The calculation above considered on-shell decay only, while for realistic 125GeV boson,
the off-shell decay turns out to be very important. That is, although h0 W + W and
h0 Z 0 Z 0 are kinetically forbidden when mh = 125GeV according to above results, the
produced W or Z pair can subsequently decay into lighter fermions, and the process like
h0 W (f f)W contributes considerable amount of Higgs decay, where W (f f) means an
off-shell W decaying to a pair of fermions. More details can be found in [8]
(c) Now we come to the process h0 gg. The leading order contribution comes from
diagrams with one quark loop.
The amplitude reads

imq
iM(h0 gg) = (igs )2 (p1 ) (p2 ) tr (ta tb )
Zv d
d q i i i
(1) tr
(2)d /q mq /q + p/2 mq /q p/1 mq

i i i
+ (1) tr (21.61)
/q mq /q + p/1 mq /q p/2 mq
The first trace in the integrand can be simplified through standard procedure,

i i i
tr
/q mq /q + p/2 mq /q p/1 mq

i tr (/q + mq )(/q + p/2 mq )(/q p/1 mq )
= 2
(q m2q ) (q + p2 )2 m2q (q p1 )2 m2q
Z 1 Z 1x
N
= 2i dx dy 02 , (21.62)
0 0 (q )3
where
q0 = q xp1 + yp2 , (21.63)

= m2q x(1 x)p21 y(1 y)p22 2xyp1 p2 = m2q xym2h ,
N = 4mq p1 p2 p1 p2 + 2p2 q 2p1 q + 4q q + (m2q p1 p2 q 2 )

(21.64)
Then we can reexpress N in terms of q 0 , p1 and p2 and drop off all terms linear in q which
integrates to zero. It is most easy to work with definite helicity states for final gluons. Then
the result gets simplified if we dot N with polarization vectors as N (p1 ) (p2 ). Note
that (pi )pj = 0 with i, j = 1, 2. Note also the on-shell condition p21 = p22 = 0, p1 p2 = 12 m2h .
then
h 02 i
N (p1 ) (p2 ) = 4mq mq + xy 2 mh + d 1 q (p1 ) (p2 ).
2 1
2 4
(21.65)
The same calculation shows that the second trace in the integrand of (21.61) gives identical
result with the first trace. To check the gauge invariance of this result, one can simply
replace (p1 ) with p1 in the expression above, then it is straightforward to find that
N p1 (p2 ) = 0. Similarly, it can also be checked that N (p1 )p2 = 0.
Then the amplitude (21.61) now reads
2g 2 mq ab 1 1x
dd q 0 N (p1 ) (p2 )
Z Z Z
iM(h gg) = s
0
dx dy , (21.66)
v 0 0 (2)d (q 02 )3
where the relation tr (ta tb ) = 21 ab in fundamental representation is also used. The momen-
tum integration is finite as d 4 under dimensional regularization, and can now be carried
out directly to be
2igs2 m2q ab 1 1x
(1 4xy)m2h
Z Z
0
iM(h gg) = (p1 ) (p2 ) dx dy
(4)2 v 0 0 m2q xym2h
is m2h
= ab (p1 ) (p2 )If (q ), (21.67)
6v
where q (mh /mq )2 , and

1 1x
1 4xy
Z Z
If (q ) 3 dx dy
0 0 1 xyq
Note that the inner product between two polarization vectors is nonzero only for +
and + . Therefore the squared amplitude with final states polarizations, color indices
summed ( ab ab = 8) is,
4s2 m4h
|M(h0 gg)|2 = |M+ (h0 gg)|2 + |M+ (h0 gg)|2 = |If (q )|2 , (21.68)
9 2 v 2
and the decay width is
m m2 2
h
(h0 gg) = 2h s2 |If (q )|2 , (21.69)
8 sin2 w mw 9
where an additional factor 1/2 should be included in (21.51) when calculating (h0 gg)
because the two gluons in final states are identical particles. This result is easily generalized
for Nq copies of quarks to be
m m2 2 X 2
h h s
(h0 gg) = If q ,
( ) (21.70)

8 sin2 w m2W 9 2

q
(d) Now we calculate the cross section for the Higgs production via gluon fusion at the
leading order. The amplitude is simply given by the result in (c), namely (21.67). When
we take the square of this amplitude, an additional factor ( 18 12 )2 should be included, to
average over helicities and color indices of initial gluons. Then, comparing (21.52) with
(21.51), we find that
2
(gg h0 ) = (s m2h )(h0 gg), (21.71)
8mh
where the hatted variable s is the parton level center-of-mass energy. We note again that
the correct formula is obtained by including an factor of ( 18 12 )2 in (gg h0 ) to average
over the initial degrees of freedom of two gluons, and an factor of 1/2 in (h0 gg) to
count the identical particles in the final state. Then, from (21.70), it is straightforward to
find
0 s2 m2h X 2
(gg h ) = 2 If (q ) (s m2h ). (21.72)

2
576 sin w mW q
Then the proton-level cross section of Higgs boson production via gluon-gluon fusion is given
by
GGF p(P1 )p(P2 ) h0 )

Z 1 Z 1
dx2 fg (x1 )fg (x2 ) g(x1 P1 )g(x2 P2 ) h0

= dx1
Z0 0
2
(x1 , x2 )
fg (x1 )fg (x2 ) g(x1 P1 )g(x2 P2 ) h0

= dM dY
(M 2 , Y )
Z
1
= dM 2 dY 0

x 1 f g (x 1 )x 2 f g (x 2 ) g(x 1 P 1 )g(x 2 P 2 ) h , (21.73)
M2
where M 2 = x1 x2 s is the center-of-mass energy of two initial gluons, while s is the center-of-
p
mass energy of two initial protons, and Y , given by exp Y = x1 /x2 , is the rapidity of the
produced Higgs boson relative to the center-of-mass frame of the proton system. (Note that
in our case M 2 = m2h .) The relations between M 2 , Y and the momentum fractions x1 , x2 can

be inverted to give x1 = (M/ s)eY and x2 = (M/ s)eY . Furthermore, fg is the parton
distribution function of the gluon in a proton, which we will take to be fg = 8(1 x)7 /x in
the following calculations. Then the cross section can be evaluated to be
GGF p(P1 )p(P2 ) h0 )

s2 1 X 2 Z Y0 mh Y 7 mh Y 7
= I ( ) dY 1 e 1 e , (21.74)

f q
9 sin2 w m2W q s s

Y0

where Y0 , given by cosh Y0 = s/2mh is the largest possible rapidity of a produced Higgs

boson. We plot this cross section as a function of the center-of-mass energy s of the pp
pair, with the Higgs bosons mass taken to be mh = 30GeV and mh = 125GeV, respectively,
in Figure
50
20
Cross Sectionpb
mh =30 GeV
10
5 mh =125 GeV
1
1 2 5 10 20 50
s GeV
(e) Next we consider the process h0 2. The contribution to this decay channel at the
leading (1-loop) level is from two types of diagrams, one with a fermion loop and the other
with a W boson (and related would-be Goldstone boson) loop. The former contribution is
easy to find by virtue of the result in (c) for h0 gg. The calculation here is in fully parallel,
except that we should include the factor for the electric charges of internal fermions Qf , take
away the color factor tr (ta tb ), change the strong coupling gs by the electromagnetic coupling
e, and sum over all charged fermions. Note that the color factor enters the expression of
the decay width as | tr (ta tb )|2 = 12 ab 21 ab = 2, then it is straightforward to write down the
fermion contribution to the h0 2 to be
m m2 s2 X 2 2
h h
iM(h0 2)f = Q N (f )If f ,
( ) (21.75)

2 2 2 f c
mw 18

8 sin w f
where Nc (f ) is the color factor, equal to 3 for quarks and 1 for charged leptons.
(f ) Now we come to the W -loop contributions to h0 2. In Feynman-t Hooft gauge, we

should also include the corresponding Goldstone loop diagrams. Then there are 13 diagrams
in total. We compute them as follows,
(a) (b)
1 ig g 2 v
iM(a) = (ie2 )(2 ) (p1 ) (p2 )
2 Z 2
dd q
DW (q)DW (k q)
(2)d
2i e2 m2W
= (p1 ) (p2 )(d 1)(2 d2 )
(4)d/2 v
Z 1
dx
2 2 2d/2
, (21.76)
0 [mW x(1 x)mh ]
dd q
Z
1 2
iM(b) = (2iv)(2ie ) (p1 ) (p2 ) Ds (q)Ds (k q)
2 (2)d
i e2 m2h
= (p1 ) (p2 )(2 d2 )
(4)d/2 v
Z 1
dx
2 2 2d/2
. (21.77)
0 [mW x(1 x)mh ]
(c) (d)
ig 2 sin w ig 2 v sin w dd q
Z
(c) (d)
iM = iM = (p1 ) (p2 ) Ds (q)DW (p2 q)
2 2 (2)d
i e2 m2W d 1
= (p 1 ) (p 2 )(2 2
) 2
. (21.78)
(4)d/2 v (mW )2d/2
(e) (f) (g)
(h) (i) (j)
(k) (l) (m)

ig 2 v dd q
Z
(e) 2
iM = (ie) (p1 ) (p2 ) DW (q)DW (q p1 )DW (q + p2 )
2 (2)d
(2q p1 ) + (2p1 q) (p1 + q)

(2q + p2 ) (q p2 ) (2p2 + q)

e2 m2W (5 x y + 4xy)m2h
Z
i
= (p 1 ) (p 2 ) dxdy
(4)d/2 v m2W xym2h
Z
d dxdy
+ 6(d 1)(2 2 ) , (21.79)
(m2W xym2h )2d/2
dd q
Z
2
iM(f) = (2i)(ie) (p1 ) (p2 ) (2q p1 ) (2q + p2 )
(2)d
Ds (q)Ds (q p1 )Ds (q + p2 )
e2 m2h
Z
i d 2dxdy
= (p1 ) (p2 )(2 2 ) , (21.80)
(4) d/2 v (mW xym2h )2d/2
2
im2 Z
dd q
W 2
iM(g) = (ie) (p1 ) (p2 ) (1)(q p1 ) q
v (2)d
Ds (q)Ds (q p1 )Ds (q + p2 )
e2 m2W
Z
i dxdy
= (p1 ) (p2 )(2 d
) , (21.81)
(4)d/2 v 2
(m2W xym2h )2d/2
ig ig g 2 v sin w dd q
Z
iM (h)
= iM (i)
= (ie) (p1 ) (p2 ) (q p1 k)
2 2 (2)d
(2q + p2 ) (q p2 ) (2p2 + q)

DW (q)Ds (q p1 )DW (q + p2 )
e2 m2W (1 x)(1 + y)m2h
Z
i
= (p 1 ) (p 2 ) dxdy
(4)d/2 v m2W xym2h
Z
1 d dxdy
(d 1)(2 2 ) , (21.82)
2 (m2W xym2h )2d/2
ig 2 v ig 2 v sin w 2
iM(j) = (p1 ) (p2 )
2Z 2
dd q
Ds (q)DW (q p1 )DW (q + p2 )
(2)d
e2 m2W 2m2W
Z
i
= (p 1 ) (p 2 ) dxdy , (21.83)
(4)d/2 v m2W xym2h
ig ig 2 v sin w
iM(k) = iM(l) = (ie) (p1 ) (p2 )
2 2
dd q
Z
(p1 + 2p2 + q) (2q + p2 ) Ds (q)DW (q p1 )Ds (q + p2 )
(2)d
e2 m2W
Z
i d dxdy
= (p 1 ) (p 2 )(2 ) , (21.84)
(4)d/2 v 2
(mW xym2h )2d/2
2
ig 2 v sin 2
w
iM(m) = (2i) (p1 ) (p2 )
2
dd q
Z
DW (q)Ds (q p1 )Ds (q + p2 )
(2)d
e2 m2h m2W
Z
i
= (p 1 ) (p 2 ) dxdy . (21.85)
(4)d/2 v m2W xym2h
The results can be summarized as,
i e2 m2W
h i
iM(X) = (p 1 ) (p 2 A (2
) d
2
)+B , (X = a, b, , m) (21.86)
(4)d/2 v
with the coefficients A and B for each diagram listed in Table.
Diagrams A B
(a) 2(d 1)J1 0
(c)+(d) 2(m2W )d/22 0
(e) 6(d 1)J2 J3
(g) J2 0
(h)+(i) (d 1)J2 2J4
(j) 0 2(mW /mh )2 J5
(k)+(l) 2J2 0
(m) 0 J5
(b) (mh /mW )2 J1 0
(f) 2(mh /mW )2 J2 0
where
Z 1
1
J1 = dx
0 [m2W x(1 x)m2h ]2d/2
Z 1

= 1 dx log m2W x(1 x)m2h + O(2 ), (21.87)
2
Z 1 Z 01x
1
J2 = dx dy 2
0 0 (mW xym2h )2d/2
1
Z Z 1x
1
= dx dy log mW xymh + O(2 ),
2 2
(21.88)
2 2 0 0
Z 1 Z 1x
(5 x y + 4xy)m2h
J3 = dx dy , (21.89)
0 0 m2W xym2h
Z 1 Z 1x
(1 x)(1 + y)m2h
J4 = dx dy , (21.90)
0 0 m2W xym2h
Z 1 Z 1x
m2h
J5 = dx dy 2 . (21.91)
0 0 mW xym2h
To see that the divergences of all diagrams cancel among themselves, it just needs to show
that sum of all A-coefficients is of order . This is straightforward by noting that J1 =
1 + O() and J2 = 1/2 + O().
Before reaching the complete result, let us first find out the W -loop contribution in the
limit m2h m2W , although it seems unlikely to be true within our current knowledge. To
find the amplitude in this limit, we expand the five integrals J1 , , J5 in terms of mh /mW ,
m2h 1 m2h
J1 ' 1 log m2W + , J2 ' log m2W + ,
2 12 m2W 2 4 48 m2W
7 m2h 11 m2h 1 m2h 1 m2h 2
J3 ' , J4 ' , J5 ' + .
3 m2W 24 m2W 2 m2W 24 m2W
Then the amplitude can be recast into
ie2 m2W m2h

2
2
iM = (p1 ) (p2 ) C + log 4 + D log mW + E + F 2 (21.92)
(4)2 v mW
Diagrams C D E F
(a) 6 3 4 1
(c)+(d) 2 1 0 0
(e) 9 9/2 6 37/12
(g) 1/2 1/4 0 1/24
(h)+(i) 3/2 3/4 1 19/24
(j) 0 0 1 1/12
(k)+(l) 1 1/2 0 1/12
(m) 0 0 0 1/2
(b) (mh /mW )2 (mh /mW )2 /2 0 0
(f) (mh /mW )2 (mh /mW )2 /2 0 0
sum 0 0 0 7/2
Therefore, the amplitude in the limit m2h m2W is given by

7 im2h
iM(h0 2)W = 2 (p1 ) (p2 ), (21.93)
2 4v
where the factor 2 counts the identical contributions from the diagrams with two final
photons changed. Now we sum up the fermion-loop contribution found in (e) and the result
here to get the h0 2 amplitude in the light Higgs limit,
im2h X 2

21
iM = Qf Nc (f ) (p1 ) (p2 ). (21.94)
3v f
4
Then the corresponding partial width is given by

m m2 2
2
X
h h
21
(h0 2) = Q2f Nc (f ) , (21.95)
8 sin2 w m2w 18 2 f
4
Now we retain mh as a free variable. Then the various diagrams sum into the following
full expression for the W -loop contribution to h0 2,
im2h
iM(h0 2)W = (p1 ) (p2 )IW (W ), (21.96)
2v
where the factor IW (W ), as a function of W (mh /mW )2 , is given by

1
IW (W ) = 6I1 (W ) 8I2 (W ) + W I1 (W ) I2 (W ) + I3 (W ) , (21.97)
W
where
Z 1
I1 (W ) dx log 1 x(1 x)W , (21.98)
0
Z 1 Z 1x

I2 (W ) 2 dx dy log 1 xyW , (21.99)
Z 10 Z 1x0
(8 3x + y + 4xy)W
I3 (W ) dx dy . (21.100)
0 0 1 xyW
Then the full expression for the partial width of h0 2 at one-loop is

m m2 2
2 X 2

0 h h

(h 2) = Q f Nc (f )If ( f ) IW ( W ), (21.101)
8 sin2 w m2w 18 2 f
(h) Collecting all results above (expect the channel, which is quite small ), we plot
the total width and decay branching fractions of the Higgs boson in Figures 21.6 and 21.7,
respectively.
10
1
Total WidthGeV
0.1
0.01
0.001
10-4
100 200 300 400 500
mh GeV
Figure 21.6: The total width of the Higgs boson as a function of its mass.

but very important!
1.000
bb W +W -
0.500
+ - Z 0Z 0
0.100 cc
Branching Ratios
0.050
tt
0.010
0.005
gg
0.001
100 200 300 400 500
mh HGeVL
Figure 21.7: The Higgs decay branching fractions of tt, bb, cc, + , W W , ZZ and gg
channels, as functions of Higgs mass.
Bibliography
[1] M. E. Peskin and D. V. Schroeder, An Introduction to Quantum Field Theory, West-

view Press, 1995.
[2] M. D. Schwartz, Quantum Field Theory and the Standard Model , Cambridge Uni-
versity Press, 2014.
[3] K. A. Olive et al. (Particle Data Group), Chin. Phys. C, 38 (2014) 090001.
[4] A. Zee, Quantum Field Theory in a Nutshell , 2nd edition, Princeton University Press,
2010.
[5] S. Coleman, Aspects of Symmetry, Cambridge University Press, 1985.
[6] S. R. Coleman and E. J. Weinberg, Radiative Corrections as the Origin of Spontaneous

Symmetry Breaking, Phys. Rev. D 7 (1973) 1888.
[7] A. Andreassen, W. Frost, and M. D. Schwartz, Consistent Use of Effective Potentials,

Phys. Rev. D 91 (2015) 016009 [arXiv:1408.0287].
[8] LHC Higgs Cross Section Working Group Collaboration (S Heinemeyer (ed.)), Hand-
book of LHC Higgs Cross Sections: 3. Higgs Properties, arXiv:1307.1347.
201

Peskin Problems

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Peskin Problems

Загружено:

Авторское право:

Доступные форматы

A Complete Solution to Problems in

An Introduction to Quantum Field Theory

2 The Klein-Gordon Field 1

3 The Dirac Field 7

4 Interacting Fields and Feynman Diagrams 23

5 Elementary Processes of Quantum Electrodynamics 33

6 Radiative Corrections: Introduction 51

6.2 Equivalent photon approximation . . . . . . . . . . . . . . . . . . . . . . . . 53

7 Radiative Corrections: Some Formal Developments 59

Final Project I. Radiation of Gluon Jets 67

11 Renormalization and Symmetry 87

12 The Renormalization Group 93

13 Critical Exponents and Scalar Field Theory 99

Final Project II. The Coleman-Weinberg Potential 103

15 Non-Abelian Gauge Invariance 113

16 Quantization of Non-Abelian Gauge Theories 121

17 Quantum Chromodynamics 133

18 Operator Products and Effective Vertices 145

19 Perturbation Theory Anomalies 159

20 Gauge Theories with Spontaneous Symmetry Breaking 167

21 Quantization of Spontaneously Broken Gauge Theories 177

Final Project III. Decays of the Higgs Boson 189

The Klein-Gordon Field

2.1 Classical electromagnetism

Then from the first equality we get:

Now substitute this into Euler-Lagrange equation, we have,

i E i = 0, ijk j B k 0 E i = 0, ijk j E k = 0, i B i = 0. (2.6)

(b) The energy-momentum tensor The energy-momentum tensor can be defined to be

2.2 The complex scalar field

(a) The conjugate momenta of and :

The canonical commutation relations:

[(x), (y)] = [ (x), (y)] = i(x y), (2.13)

The rest of commutators are all zero.

(b) Now we Fourier transform the field as:

Substitute the mode expansion into the Hamiltonian:

(c) The theory is invariant under the global transformation: ei , ei . The

Rewrite this in terms of the creation and annihilation operators:

[Q, a ] = a , [Q, b ] = b . (2.21)

where i with i = 1, 2 is a two-component complex scalar. Then it is straightforward to

The generalization to N -component complex scalar is straightforward. In this case we

and similarly, [Q, Q] = [Qa , Q] = 0.

2.3 The spacelike correlation function

D(x y) = h0|(x)(y)|0i, (2.28)

x0 y 0 = 0, and |x y|2 = r2 > 0. (2.29)

The Dirac Field

3.1 Lorentz group

[Li , Lj ] = iijk Lk , [K i , K j ] = iijk Lk . (3.2)

[Ji , Jj ] = iijk Jk , [J+i , Jj ] = 0. (3.3)

1. ( 21 , 0). Following the definition, we have J+i represented by 1

Hence a field under this representation transforms as:

2. ( 12 , 0). In this case, J+i 0, Ji 1

Hence a field under this representation transforms as:

In what follows we will prove that V is in fact a Lorentz vector.

where = 2i (J ) in its infinitesimal form. We further note that:

and also, ij = ijk k , 0i = i0 = i , then the combination V = V i i + V 0 transforms

as expected. Hence we proved that V is a Lorentz vector.

3.2 The Gordon identity

3.3 The spinor products

1. Let uL0 be left-handed spinor with momentum k0 ;

2. Let uR0 = k/1 uL0 ;

3. For any lightlike momentum p (p2 = 0), define:

Let p = (p0 , p1 , p2 , p3 ), then:

In the same way, we get:

i E i = 0, ijk j B k 0 E i = 0, ijk j E k = 0, i B i = 0. (2.6)

[Li , Lj ] = iijk Lk , [K i , K j ] = iijk Lk . (3.2)

[Ji , Jj ] = iijk Jk , [J+i , Jj ] = 0. (3.3)

and also, ij = ijk k , 0i = i0 = i , then the combination V = V i i + V 0 transforms

(L) = imT 2 F im + 1