The Standard Model of Particle Physics and

One-loop Renormalization in Quantum

Electrodynamics

a dissertation presented
by
Neha Zaidi
to
The Department of Physics

in partial fulfillment of the requirements

for the degree of
Bachelor of Science
in the subject of
Physics

Lahore University of Management Sciences

Lahore, Punjab
May 2021
 Contents

I A Primer 4
1 Understanding Gauge Invariances in Quantum Field Theories 5

II The Electroweak Interactions 22

2 Setting up the Preamble for Quantum Flavour Dynamics 23

3 Extending to Other Quarks and Leptons 41

4 The Lagrangian of the Standard Model of Particle Physics 46

III Renormalization at the One-loop Level 49

5 Renormalization of the propagator in the φ3 context 50

6 Renormalization of the vertex in the φ3 context 60

7 Renormalization in the context of QED 65

IV Appendices and References 76

Appendix A A Brief Overview of Group Theory 77

Appendix B A Brief Introduction to Quantum Electrodynamics 79

2
Appendix C A Brief Discussion of Quantum Chromodynamics 82

Appendix D Spontaneous Symmetry Breaking 85

Appendix E Momentum Integration in Loops 89

References 95

3
 Part I

A Primer

4
”…although the symmetries are hidden from us, we can
sense that they are latent in nature, governing everything
about us. That’s the most exciting idea I know: that
nature is much simpler than it looks.”

S. Weinberg

Understanding Gauge Invariances in

1
Quantum Field Theories

One curious aspect in physics is that various seemingly distinct physical systems share the
same mathematical description. This is because, no matter how different the setups might
appear, as long as the variables and constituents are the same, the systems are essentially the

5
same. One excellent example of this is the gravitational pull between two particles in a sys-
tem. As long as the separation of the particles and system conditions are kept constant, it
does not matter how you orient your system or where you perform your calculations - the
particles could be situated on the earth or on Neptune, the dynamics would be identical.
Another important detail is that it is always possible, at least in principle, to switch from
one physical setup to another by means of some transformation. But since these transfor-
mations connect physically equivalent systems, they should not contribute, in any way,
to observable quantities. This is a concept best illustrated in light of Maxwell’s equations.
Looking at the Maxwell-Faraday equation

∂B
∇×E= ,
∂t

we observe that we could add to the electric field a curl-less quantity - most generally, the
gradient of some scalar field - without affecting the observable magnetic field. This essen-
tially refers to connecting two equivalent systems with different reference points, and is an
example of something we call gauge invariance.
The remarkable thing about the Standard Model of Particle Physics is that it is a gauge
theory. This means that we can build its theoretical framework, almost entirely, using some
symmetry principles. This makes studying gauge invariances in the context of quantum
field theories important. In this chapter, we will begin by understanding and differentiating
between global Abelian and non-Abelian gauge transformations. We will then proceed to
localizing the transformations and understanding its consequences on the theory.

6
 Let us begin by considering the Dirac Lagrangian:

L (x) = ψ̄(x) (iγλ ∂λ − m)ψ(x) (1.1)

It is straightforward to see that this Lagrangian is invariant under transformations of the

form
ψ(x) −→ ψ(x)′ = ei β ψ(x), (1.2)

where β is a complex, purely numerical parameter, independent of x. Given the discussion

in Appendix C, we can conclude that these transformations form an Abelian group. The
group is referred to as the U(1) gauge group, where U denotes unitary. Verifying the uni-
tarity of the transformations is a trivial task that we leave as an exercise for the reader.
The leap from Abelian gauge theories to non-Abelian gauge theories comes from a sim-
ple step - one that has many ramifications in terms of the complexity of algebra 1 . Instead of
considering one wavefunction at one time, as in the case discussed previously, we allow our-
selves to speak of multiple wavefunctions. Let us consider a system S with the a set of states
Ψ1 (x), . . . , ΨN (x), where we have suppressed spinor indices. For now, we shall assume that
Ψ1 (x) and Ψ2 (x) are degenerate * , such that

H Ψ1 (x) = E Ψ1

H Ψ2 (x) = E Ψ2 ,

* We have chosen to work with only two wavefunctions here for purposes of simplicity. The discussion is
straightforwardly generalized to N wavefunctions as we shall see later.

7
 where H is the Hamiltonian that governs the dynamics of S and E is the energy of Ψ1 (x)andΨ2 (x).
However, laws of Quantum Mechanics posit that if Ψ1 (x)andΨ2 (x) are truly degenerate,
then any arbitrary linear combination of the two vectors would be perfectly equivalent.
This can be verified as follows

Ψ1 (x) −→ Ψ′1 (x) = α Ψ1 (x) + β Ψ2 (x)

Ψ2 (x) −→ Ψ′2 (x) = γ Ψ1 (x) + δ Ψ2 (x) (1.3)

which ultimately leads to

HΨ′1 (x) = E(α Ψ1 (x) + β Ψ2 (x)) = EΨ′1 (x)

HΨ′2 (x) = E(γ Ψ1 (x) + δ Ψ2 (x)) = EΨ′2 (x),

where α, β, δ, and γ are arbitrary complex coefficients.

We wish to explore this avenue further and understand the consequences it has on our
theory. The two-fold degeneracy of states Ψ1 (x) and Ψ2 (x) is reminiscent of the more fa-
miliar case of a spin-1/2 system in the absence of magnetic fields, where spin +1/2 states
are completely degenerate with spin -1/2 states. We shall exploit this similarity and build
our terminology and mathematical structure in complete analogy. We, therefore, combine
Ψ1 (x) and Ψ2 (x) in a 2 × 1 isospin matrix

8
      
 Ψ1 (x)  1 0
Ψ(x) =   = Ψ1 (x)   + Ψ2 (x)  
Ψ2 (x) 0 1

= Ψ1 (x) χ1 + Ψ2 (x) χ2 (1.4)

where

   
1 0
χ1 =   , χ2 =   ,
0 1

with χ1 referring to isospin ’up’ and χ2 referring to isospin ’down’ in complete analogy
with the spin up and spin down cases of true spin-1/2 systems.
Defining Ψ(x) as in Eq. 1.4 allows us to rewrite Eq. 1.3 in the following concise manner

 
α β
Ψ′ (x) = U Ψ(x), U= . (1.5)
γ δ

Of course the parallels being drawn here are limited to mathematical structure and ter-
minology and do not extend to the domain of physicality and degrees of freedom. For ex-
ample, the transformation, as given in Eq. 1.5, does not affect the true spin of Ψ1 and Ψ2
and a measure of spin of these states leaves their isospins untouched.
As can be seen in Eq. 1.5, four complex numbers, or eight real numbers, must be deter-
mined independently to pin the form of the transformation matrix U. However, we shall
reduce that number by imposing a set of restraints as demanded by the physicality of our
problem.

9
 Firstly, we note that U should preserve the normalization of the initial states. This brings
us to the following condition

Ψ′ † (x) Ψ′ (X) = Ψ† (x) U† U Ψ(x) = Ψ† (x)Ψ(x).

U† U = 12×2 , (1.6)

which suggests that U must be unitary. Using elementary determinant properties, we can
see that this condition results in constraints on the determinant of U:

detU† U = detU† . detU = |detU|2 = 1

detU = eiθ or detU = 1 (1.7)

where is a real number. If U is such that its determinant is given by the first possibility
given in Eq. 1.7, U takes the trivial form

U = eiθ I

which essentially means rotating Ψ1 (x) and Ψ2 (x) by equal amounts in the complex vec-
tor space spanned by the two vectors. We have already considered this case while discussing
Abelian gauge groups in Quantum Field theories. New mathematics arises when we con-

10
sider the second possibility in Eq. 1.7, i.e.,

det U = 1. (1.8)

Therefore, U, as defined in 1.5, are a set of 2 × 2 unitary matrices with unit determinant.
Such a set, under the binary operation of matrix multiplication, forms a group* . In particu-
lar, we have an instance of the SU(2) group at our hands.
7
The great thing about SU(2) groups is that they belong to a category of groups called
Lie Groups. This is useful because any transformation in this division can be built by aggre-
gating infinitesimal contributions. So, if, for example, we have Usmall as

Usmall = 12×2 + i ζ2×2

where ζ is a 2 × 2 matrix with infinitesimal entries so that Usmall only differs slightly from
identity, or put simply, a transformation of no-change, we can build a finite transformation
using the result†

Φ n
U = lim (Usmall )n = lim (1 + ζ) n = lim (1 + ) , = eΦ . (1.9)
n→∞ n→∞ n→∞ n

This tells us that if we somehow determine ζ, we will have essentially determined U. There-
fore, our next task would be to use the constraints we developed for U to see how they help
us get ζ.
* A brief overview of Group Theory can be found in Appendix C
†
It may also help to think of ζ as the infinitesimal matrix obtained when we break the physical transforma-
tion, Φ into n pieces, where n → ∞.

11
 We begin by imposing the condition unit determinant, as given in Eq. 1.8. Allowing ζ to
be  
a b 
ζ= , (1.10)
c d

where a,b,c,d are complex numbers, gives

 
1 + a b 
Usmall =  (1.11)
c 1+d

which, combined with restriction on the matrix’s determinant, yields the condition

a + d = 0 ⇒ Tr ζ = 0. (1.12)

We now consider the consequences of the unitarity of U upon the form of ζ. Since U is
unitary,

U†small Usmall = (12×2 − i ζ† )(12×2 + i ζ) = 12×2

⇒ ζ † = ζ, (1.13)

meaning that we are essentially looking for 2 × 2 hermitian matrices with zero trace. This
means that ζ must be of the form
 
 a b − ic 
′ ′ ′
 , (1.14)
b′ + i c ′ −a′

12
where a′ , b′ , c′ are real numbers. This shows that the number of variables to be indepen-
dently determined have been reduced to three from eight. We can cast ζ in a more sugges-
tive form by defining
ε ≡ (ε1 , ε2 , ε3 ) = (a′ , b′ , c′ ),

where ε is the infinitesimal magnitude of transformation. This which ultimately allows us

to write ζ as
ζ = ε · τ/2 (1.15)

where τ are our usual Pauli matrices and form a basis of the space of 2 × 2 Hermitian ma-
trices. These Pauli matrices can then be interpreted as the ’generators’ of transformations
U, much like they generate spin angular momentum transformations. ε gives us the mea-
sures of transformation corresponding to each Pauli matrix. Since Pauli matrices follow the
following commutation relation

[τi , τj ] = εijk τk ,

we know that any two transformations of this sort will not commute, hence giving us our
non-Abelian SU(2) group.
An arbitrary transformation U that allows us to go from one linear combination of Ψ1
and Psi2 to another, as in Eq. 1.3, and is sufficiently constrained by conditions in Eqs. 1.6
and 1.8 can be written using Eq. 1.9 as

τ
U = ei α · 2 g , (1.16)

13
where g is a constant we have introduced in foresight of its use later on and α is defined as

α ≡ (α1 , α2 , α3 )

and
α α1 α2 α3
lim = lim ( , , ) ≡ (ε1 , ε2 , ε3 ) = ε.
n→∞ n n→∞ n n n

Let us now pause here and discuss the progress so far. We started with two degenerate
state vectors and argued that according to Quantum Mechanics, any linear combination of
those vectors, that satisfies appropriate conditions, is equally valid. We labelled the trans-
formation that takes us from one combination of Ψ1 and Ψ2 U and derived its most general
form in Eq. 1.16. Now, we shall view the same problem from a geometric lens, for it proves
instructive.

Consider the two-dimensional complex vector space spanned by Ψ1 and Ψ2 U. Much

like the x- and the y- axes spanning the two-dimensional real, Cartesian space, we can as-
sume Ψ1 and Ψ2 U to be the two spanning axes. The span of the two vectors is indepen-
dent of their orientation, so we have the liberty to rotate the axes. Rotating the vectors here
essentially means that we have the liberty to define and re-define our axes. In light of the
underlying degeneracy of the problem, this can be interpreted as having the freedom of la-
belling axes with Ψ1 and Ψ2 labels, for both labels are entirely equivalent. This discussion
of rotations brings us back to rotations in the complex vector space, like we previously en-
countered in the case of the U(1) gauge group. However, what differs in this case is that
we are transforming two vectors simultaneously, meaning that upon rotation we would
have something like Eq. 1.3. Then the working that follows from that point is an attempt

14
to build transformations that rotate one linearly independent combination‡ of Ψ1 and Ψ2
to another linearly independent combination. The generators of transformations, τ, are
then generators of rotation in three independent directions, with α being the measures of
rotations.
Now that we have seen the effects of degeneracy for the case of two states, we are pre-
pared to generalize the discussion for M states. Suppose that our system is governed by a
Hamiltonian, H, and Ψ1 , . . . , ΨM are the degenerate eigenstates of H. Then, Eq. 1.3 be-
comes

Ψ1 −→ Ψ1′ (x) = a11 Ψ1 (x) + . . . a1M ΨM (x)

..
.

ΨM −→ Ψ′M (x) = aM1 Ψ1 (x) + . . . aMM ΨM (x),

which leads to Eq. 1.5 taking the form

 
 a11 . . . a1M 
 . .. .. 
Ψ′ (x) = U Ψ(x), U=
 .
. . . 
.
 
aM1 . . . aMM

Since the requirement of preservation of normalization of initial and final states persists,
U must be unitary, much in the spirit of Eq. 1.6. We again choose U to have unit determi-
nant, which leads us to the conclusion that U are special, unitary matrices of order M. Since
they form the generalized Lie group, SU(M), we can build them bit by bit, combining in-
‡
Linear independence guarantees the invariance of the span of the vectors so that we remain in the same
space.

15
finitesimal contributions. Therefore, U can be written as

U = lim (Usmall )n = lim (1M + i ζM )n

n→∞ n→∞

Usmall = 1M + i ζM (1.17)

where the matrices involved are now M-dimensional. Imposing conditions of unitarity and
unit determinant on Usmall in Eq. 1.17 tells us that ζ must be Hermitian and traceless, ex-
actly like our conclusion in the two-dimensional case. Therefore, we must find a set of lin-
early independent, Hermitian, and traceless matrices of order M, such that we can combine
them linearly to write any U. Let T be such a set and α be associated magnitudes. Then, a
transformation in the M-dimensional degenerate vector space can be written as

T
U = ei α · 2 g , (1.18)

which works as

T
Ψ′ (x) = ei α · 2 g Ψ(x)
 
T  Ψ1 (x) 
= ei α · 2 g  
Ψ2 (x).

With this, we have completed building transformations that mix degenerate Hamilto-
nian eigenstates. Now, we shall proceed to discuss the implications this has on our theory.
When we began our discussion of degenerate states, we argued that Quantum Physics does
not distinguish does not distinguish between two different sets of linear combinations of

16
states, provided each set has linearly independent vectors. What effect does this have on
the Physics of the system? For one, it means that it does not matter which linear combi-
nation we choose to work with because the equations of motion governing the system are
independent of this choice. This means that the Lagrangian of such a system with degen-
eracies should remain invariant under transformations that mix degenerate states. This can
be explicitly seen. To do that, let’s assume that in a system of N non-interacting massless
spin-1/2 Fermions, M are degenerate, so they enter the Lagrangian in a symmetric way like
follows
 
( )  Ψ1 
 . 
L= Ψ̄1 . . . Ψ̄M i ∂/  . 
 .  + Ψ̄M+1 i ∂/ΨM+1 + . . . + Ψ̄N i ∂/ΨN ,
 
ΨM

where we have combined the degenerate states in an isospinor to emphasize on their equiva-
lence. Now, if we transform the isospinor using the transformation in Eq. 1.18, we see that
the Lagrangian remains unchanged. It is also important to note that these transformations
are continuous, for they can be built bit by bit, using infinitesimal contributions. This ul-
timately means that Noether’s theorem is applicable here and hints at the conservation of
some quantity. However, we shall not explore this arena further here, and leave it to be dis-
cussed in future chapters.
Till now, we have explored the avenue of global gauge transformations. This means that
the amount of rotation, α, is independent of the space coordinate x, which results in the
same transformation being applicable in all parts of the universe. To put this another way,
although the process of assigning wavefunctions labels initially is completely arbitrary, but

17
once the decision is made at some space-time point, everyone at all other locations and
times must prescribe to the same convention. This seems non-sensical, since the transfor-
mations we are concerned with are in the non-physical, state spaces. Therefore, we should
have the liberty to make arbitrary definitions at different points. This brings the concept of
local transformations into the picture. We allow α to have an x dependency, meaning that
we can perform different transformations (random definitions of axes) at each space-time
point independently. We postulate the invariance of the Lagrangian under such transfor-
mations as well. However, we have run into a problem now. Where previously, the La-
grangian remained unaffected by the transformations due to x independence, such is not
the case anymore. Instead of a case like this

∂μ Ψ(x) −→ ∂μ Ψ′ (x) = ∂μ (Ψ(x)) U + Ψ(x) ∂μ (U) = U (∂μ Ψ(x)),

we now have a non-zero contribution from the derivative of the transformation matrix.
However, all hope is not lost. We can rescue our theory from this seemingly fatal flaw
by following in the footsteps of Yang and Mills. and attempting to rewrite the Lagrangian
such that it is capable of handling local transformations while remaining invariant. Since
we have already extended our discussion of non-Abelian transformations to N-dimensions,
we see it fit to continue with the same generalizations. The discussion that proceeds with-
out any specific reference to particular dimensions or Lie groups, and can therefore be spe-
cialized to different cases, as we shall do in later parts of this paper.
We start by replacing the partial derivative with the covariant derivative Dμ

∂μ −→ Dμ = ∂μ + i g Mμ (x), (1.19)

18
where Mμ are real vector fields that serve as the gauge connection between two space-time
points. The introduction of the covariant derivative here is reminiscent of the covariant
derivative ubiquitous in General Relativity, for they serve the same purpose - construct a
quantity that remains invariant under transformations. The vector fields Mμ are defined as

T
Mμ (x) = Mμ (x) · , (1.20)
2

which means that we have as many linearly independent vector fields as the number of gen-
erators of the transformation group. These fields, which we shall call gauge fields hence-
forth, must also transform when spinors transform under Eq. 1.18. These transformation
requirements exhibit themselves when we demand

Dμ Ψ(x) −→ D′μ Ψ′ (x) = U D′μ Ψ(x).

We perform some explicit computation

D′μ Ψ′ (x) = (∂μ + igB′μ ) Ψ′

= (∂μ + igB′μ ) U Ψ(x)

= (∂μ Ψ(x))U + Ψ(x)(∂μ U) + (i g B′μ U Ψ(x))

≡ U(∂μ + i gBμ )Ψ(x),

where to make the leap from the second-last equation to the third, we must have 5

i g B′μ U Ψ(x) = i g UBμ Ψ(x) − (∂μ G)∂μ Ψ(x),

19
which, we can massage into the following form, assuming the transformation law to be an
operator equation

i
B′μ (x) = UBμ (x)U−1 + U U−1 (∂μ U) U−1 . (1.21)
g

With this recipe, we can ensure that our Lagrangian is now finally gauge covaraint, meaning
that it preserves its shape under gauge transformations.
We can give these gauge fields a life of their own by introducing a gauge-specific kinetic
term into the Lagrangian. However, we must appreciate our Lagrangian’s gauge covari-
ance, and therefore must construct a quantity which respects this. In general, the following
prescription always yields invariant field strength tensors

Mμν (x) = [Dμ , Dν ]

= ∂μ Maν (x) − ∂ν Maμ (x) + ig[Mμ (x), Mν (x)]. (1.22)

This equation illuminates one of the differences we expect to see in Abelian and non-
Abelian gauge theories. The commutator we see in the second line of the equation is zero
iff the generators of the transformation group commute. Using these tensors, we write the
kinetic term as
1
Tr ( Mμν (x) Mμν (x) ).
2

20
With this, we can now write the complete, non-interacting, N-particle Lagrangian as

1
L(x) = Tr ( Mμν (x) Mμν (x) )
2  
( )  Ψ1 
 . 
+ Ψ̄1 . . . Ψ̄M /
iD . 
 . 
 
ΨM

+ Ψ̄M+1 i ∂/ΨM+1 + . . . + Ψ̄N i ∂/ΨN . (1.23)

21
 Part II

The Electroweak Interactions

22
 Setting up the Preamble for Quantum
2
Flavour Dynamics

4
Let us begin our discussion by considering the electron-neutrino family of leptons. Since
experimental evidence suggests that nature discriminates between left-handed and right-
handed neutrinos in beta decay, with only the former being observed, we shall distinguish

23
between different helicities in electrons as well. We therefore write

e(x) = eL (x) + eR (x) (2.1)

where e(x) is the Dirac field operator associated with the electron, and eL andeR represent
field operators for the left-handed and right-handed electron species, respectively. These
single helicity operators are given by

1
eL (x) = (1 − γ5 ) e(x)
2
1
eR (x) = (1 + γ5 )e (x). (2.2)
2

However, it is important to note that operators, as defined in Eq. 2.2 are not Dirac field
operators, for they fail to satisfy the Dirac equation. The problem arises due to the non-
zero mass term in the equation. However, since electron mass is negligible when compared
to the orders of energy considered in High Energy Physics, we take the leap and assume
electrons to be massless. This renders eL andeR Dirac field operators and valid entities to be
included in the Dirac Lagrangian. For now, we shall also assume that both electromagnetic
and weak interactions have been ’switched off’ to write the non-interacting, two particle
Lagrangian

 
( )( )
νeL (x) λ
L (x) = ν¯eL (x) e¯L (x) iγλ ∂λ   + e¯R (x) i γ ∂λ eR (x), (2.3)
eL (x)

where we have combined the left-handed electron and neutrino in an isodoublet to

24
emphasize that the two enter the theory symmetrically. This is warranted by our assump-
tions, which make it impossible to physically distinguish between a left-handed electron
and a neutrino. Mathematically, this translates to the Lagrangian being invariant under
global SU(2) transformations in the state space spanned by νeL and eL , as explained in Ch. 1.
Therefore, when
     
νeL (x) νeL (x) ν (x)
′
i α · τ g  eL
  −→  =e 2  
eL (x) e′L (x) eL (x)

eR (x) −→ eR (x)

then
L ′ (x) = L (x).

However, as argued in Ch. 1, we are more interested in local gauge transformations. There-
fore, we allow α to have an x dependence and borrow the prescription prepared in our dis-
cussion on gauge invariances in quantum field theories, specialized to the case of SU(2)
symmetry. The resulting Lagrangian, in spirit of Eq. 1.23, is then

1
L (x) = Tr(Wλρ (x) Wλρ (x))
2  
( )
νeL (x)
+ ν¯eL (x) e¯L (x) iγλ (∂λ + igWλ (x))  
eL (x)

+ e¯R (x) i γλ ∂λ eR (x), (2.4)

25
where we have introduced three linearly independent gauge fields, Wλ , such that

τα
Wλ (x) = Wαλ ,
2

which transform according to the general rule as developed in Eq. 1.21

i
W′λ (x) = U(x)Wλ (x)U−1 (x) − U(x)∂λ U−1 (x), (2.5)
g

where U(x) is the local SU(2) transformation. We will be referring to this group as the weak
isospin group. Again, it must be emphasized that it has nothing to do with the true spin of
the spinors and is merely used to highlight similarity between the mathematics of the two
formulations.
Let us now inspect Eq. 2.4. We see that introducing vector fields in the Lagrangian has
generated interaction terms of the sort

 
( )
λ νeL (x)
ν¯eL (x) e¯L (x) iγ igWλ (x)  .
eL (x)

We wish to bare the nature of this interaction term and assess if it aligns with experimen-
tal predictions and the propositions we made while phenomenologically discussing elec-
troweak interactions. We define

1 1 2
W±
λ ≡ √ (Wλ ∓ i Wλ ),
2

−
where W+
λ and Wλ are hermitian conjugates of each other. This definition allows us to

26
rewrite the interaction term as
 
( )
λ a τa νeL (x)
Lint. = ν¯eL (x) e¯L (x) iγ igWλ (x)  
2 e (x)
L

g[ 3
=− Wλ ( ν¯eL γλ νeL − eL γ¯λ L )
2
√ λ
+ 2W+
λ ν¯eL γ eL
√ λ
+ 2W−
λ e¯L γ νeL ]. (2.6)

A very interesting thing has happened. Massaging the interaction term has brought out two
different types of currents in the Lagrangian, i.e., charged and neutral, which mediate two
−
different types of processes. It is the processes involving W+
λ and Wλ that interest us the

most. Inspecting their terms closely, we observe that the processes involved involve changes
in charge in the initial and final states. Conservation of charge then necessitates that the
gauge boson mediating the process must be charged, so that it either takes away charge or
imparts charge. More specifically, if we consider the term

√ λ
2W+
λ e¯L γ νeL

we see that W−
λ must bring charge into the system. In the language of Feynman diagrams,

this is understandable, because the process involves the destruction of a neutral neutrino
and the creation of a charged electron. Since the process is mediated by W−
λ , it must carry

the necessary charge. This seems promising so far. We started with a non-interacting theory
involving electrons and neutrinos, and using symmetry principles, we ended up with weak
interactions. The theory also gave us charged and neutral bosons, both of which were ex-

27
perimentally observed. However, we must verify if the interaction has the correct γλ (1 − γ5 )
or the (V-A) structure. To do so, we write the interaction term

√
i 2 λ
− g W−
λ e¯L γ νeL .
2

and manipulate this expression using definitions in Eq. 2.2 and Clifford algebra. Following
are a few computations explicitly done. Note that we have suppressed constants for clarity
purposes and will reintroduce them towards the end.

λ 1 − γ5 1 − γ5
W− −
λ e¯L γ νeL = Wλ ( e) γλ ( ν)
2 2
1 − γ5 † 0 λ 1 − γ5
= W−λ ( e) γ γ ( ν)
2 2
(1 + γ5 )(γλ − γλ γ5 )
= W−λ ē ν
4
1
= W− ēγλ (1 − γ5 )ν.
2 λ

Using Feynman rules, we can then derive the vertex term

1
−ig √ γλ (1 − γ5 ),
2 2

which indeed has the required form. So, everything is looking great so far, indicating that
we may be on the right track. However, there still are a few things absent from our formu-
lation. First of all, although the form of the coupling is accurate, it still is lacking in that it
does not describe point-like interactions, which is how we have observed processes happen-
ing. Secondly, the theory has no mass terms in it. This means that the universe it is describ-
ing is a massless one, which, as we know from experience, is not true. If we find a way to

28
introduce mass into the system, we will be in a position to resolve the first issue as well by
requiring the W bosons to be so massive that their couplings are point-like. At this point,
one may consider adding mass terms into the Lagrangian by hand. However, all terms in a
Lagrangian should respect all symmetries and invariances of the physical system. Careful
inspection of terms of the typical mass-term form

(Wλ )2

and
ē(x) e(x)

shows that these terms are not invariant under gauge transformations. Therefore, they can-
not be used. We must come up with some other technique to bring mass into the picture.
But before we do that, we will embark on a little detour. We observe that our Lagrangian
does not account for electromagnetic interactions. This is concerning because of two rea-
sons - first, we want to unify all fundamental forces in our framework; second, the pres-
ence of charged entities, i.e., the electron and the positron, in our system means that we will
physically observe electromagnetic effects. Therefore, our Lagrangian must be equipped
with tools to cater to electromagnetism. If we recall quantum electrodynamics (for refer-
ence, check Appendix ??), we realize that electrodynamics arises from invariance under
U(1) transformations. Since we have not yet built invariance under these transformations
into our model, we shall broach this task. And as we shall see, it will yield us electrodynam-
ics.

29
 Consider U(1) transformations of the form
   
νeL (x) ν (x)
iφ′  eL
  −→ e  
eL (x) eL (x).

eR (x) −→ eiφ eR (x).

While the Lagrangian does in fact remain invariant under these transformations, we shall
not pursue them exactly further. This is because they further yield two massless photon-
like bosons, which contradicts experimental observation. Therefore, we shall modify the
transformations slightly. Instead of considering two independent U(1) transformations, we
consider a single U(1) group generated by a hypercharge operator, Y, such that

 
 yL 0 0 
 
Y= 
 0 yL 0  .
 
0 0 yR

This transformation then acts on Ψ(x), which is a spinnor consisting all Fermions in the
system, like    
νeL (x) νeL (x)
   
Ψ(x) =  
 L  −→ e
e (x)
i g′ Y  e (x)  .
 L  (2.7)
   
eR (x) eR (x)

We now gauge this symmetry into the Lagrangian using the same protocols we defined ear-
lier. We introduce a set of real gauge fields Bλ , which transform in accordance to Eq. 1.22.

30
However, since U(1) is an Abelian group, the fields commute and we are left with

Bλρ (x) = ∂λ Bρ − ∂ρ Bλ .

Since the theory is invariant under two different gauge groups, we will combine them in a
⊗
total SU(2) U(1) group. Y and Ta then form a representation of the group, where we
have defined Ta as  
1
 2 τa 0
Ta =  .
0 0

These generators then follow the following commutation rules:

[Ta , Tb ] = i εabc Tc

[Ta , Y] = 0

[Y, Y] = 0.

With these new descriptions in place, we can rewrite the Lagrangian in a more compact
form

1 1
L (x) = − Tr (Wλρ (x)Wλρ (x)) − Tr (Bλρ (x)Bλρ (x))
2 2
+ Ψ̄(x)iγλ Dλ Ψ(x) (2.8)

where
Dλ = ∂λ + igWaλ (x)Ta + ig′ Bλ (x)Y.

31
Let us now extract the interaction term in Eq. 2.8 and inspect it closely. The term is

L ′ (x) = Ψ̄(x)iγλ (∂λ + igBaλ (x)Ta + ig′ Bλ (x)Y)Ψ(x)

g λ − λ
= − √ (W+λ ν̄L γ eL + Wλ ēL γ νL )
2
1
− (gW3λ + 2yL g′ Bλ )νL γλ νL
2
1
+ (gW3λ − 2yL g′ Bλ )ēL γλ eL − yR g′ Bλ ēR γλ eR . (2.9)
2

The first two terms in the expression are precisely the ones we encountered when we gauged
the SU(2) symmetry into the Lagrangian. It is the last two lines that interest us. We notice
that vector fields W3λ and Bλ enter the equation in a symmetric manner. This is expected
since both vector fields are massless and chargeless, and hence equivalent. As we discussed
in Ch. 1, such a setup means that any orthogonal linear combination of W3λ and Bλ would
be perfectly equivalent. Knowing this, we introduce Aλ and Zλ , such that

1
Zλ = √ 2 (gW3λ − g′ Bλ )
g + g′2
1
Aλ = √ 2 (g′ W3λ + gBλ ),
g +g ′2

where we have structured the new fields such that one of them couples to the neutral neu-
trino current, while the other couples to the electron current. If we define the weak angle to
be such that
g′ g
sinθw = √ 2 and cosθw = √ ,
g + g′2 g2 + g′2

32
we can rewrite the Aλ and Zλ fields as

Zλ = cosθw W3λ − sinθw Bλ

Aλ = sinθw W3λ + cosθw Bλ ,

where it is easy to verify that the two new fields are indeed orthogonal to one another. We
can now make these redefinitions in Eq. 2.9 to obtain

g λ − λ
L ′ (x) = − √ (W+ λ ν̄eL γ eL + Wλ ēL γ νeL )
2
√ 1 1
− g2 + g′ 2 Zλ { ν̄eL γλ νeL − ēL γλ eL
2 2
− sin2 θw (−ēL γλ eL + yR ēR γλ eR )}
g g′
−√ 2 Aλ (−ēL γλ eL + yR ēR γλ eR )
g + g′ 2

where we have defined

1
yL = −
2

for reasons that will be clear soon. If we further define

g g′ √
yR = 1 and e = √ 2 = g2 + g′ 2 sinθw cosθw ,
g + g′ 2

33
we can write Eq. 2.9 in a very concise and instructive form

λ 1 λ
L ′ = −e [ Aλ Jem ++ Zλ JNC
sinθw cosθw
1 λ − λ
+√ λ ν̄L γ eL + Wλ ēL γ νL ) ],
(W+ (2.10)
2 sinθw

where

λ
Jem = −ēL γλ eL − −ēR γλ eR = −ē γλ e
λ 1 1
JNC = ν̄L γλ νL − ēL γλ eL − sin2 θw Jem
λ
.
2 2

A few words about this result are in order. Looking at the first term in Eq. 2.10, we see that
⊗
our scheme of gauging SU(2) U(1) symmetry in the Lagrangian has rightly worked. We
now have separate vector fields that mediate processes of electrodynamics and weak inter-
actions. The nature of their couplings and interactions is also precisely as predicted by ex-
periment. We have an additional neutral boson as well, whose presence is not unwelcome.
For one, the existence of weak, neutral currents is well-known and documented. Secondly,
the unification of a charged electromagnetic current of the pure vector structure and weak
current of the (V-A) structure necessitates its existence.
Now we proceed to the issue of masslessness. To be consistent with experiment, we must
have mass terms for the W- and Z- bosons, in addition to the electron. However, as dis-
cussed earlier, we cannot add such terms into the Lagrangian by hand. What we will instead
be following is the Higgs Mechanism, based on ideas of symmetry breaking, as suggested by
Salam and Weinberg. An introduction to the process of symmetry breaking can be found

34
in Appendix D. In our discussion on symmetry breaking in the aforementioned section,
we considered a complex scalar field φ. We will borrow the description of the field and its
properties identically from our previous discussion and extend its characteristics to suit our
purposes.
Let φ(x) be called the Higgs field. The field permeates all space and interacts with other
quantum fields, including the ones we have been working with so far. Since the coupling
⊗
should respect the constraints our theory adheres to, i.e., gauge invariance under SU(2) U(1)
symmetry group, and should be renormalizable, we postulate that the coupling should have
the following form:

 
 ν eL 
Lyuk = −ce ēR φ†   + hermitian conjugate
eL

= −ce (φ†1 ēR νeL + φ†2 ēR eL ) + hermitian conjugate. (2.11)

Here ce is a real coupling constant. As it is evident, the coupling is invariant under SU(2)
transformations. To have it be invariant under hypercharge, or U(1), transformations as
well, we consider the following elementary process:

eL −→ eR + φ2 . (2.12)

If the coupling indeed exhibits U(1) symmetry, it must preserve hypercharge, which brings
us to the condition
1
yH = yL − yR = , (2.13)
2

35
where yH is the hypercharge value for the Higgs field.
We now introduce a kinetic term governing the dynamics of the Higgs field into our
Lagrangian to complete its description in our system. With this addition, the overall La-
grangian can be written as

 
( )
1 1  ν eL 
L = − Tr(Wλρ Wλρ ) − Tr(Bλρ Bλρ ) + ν̄eL e¯L i γλ Dλ  
2 2 e L
   
 ν eL   ν eL 
+ e¯R iγλ Dλ eL − ce e¯L φ†   − c∗e   φeR
eL eL

+ (Dλ φ)† (Dλ φ) − V(φ), (2.14)

where
( τa )
Dλ φ = ∂λ + igWaλ ′
+ ig Bλ yH φ, (2.15)
2

so that the kinetic term is consistent with the symmetry of our system. Given the Lagrangian
in Eq. ??, it is simple to see how we were right to assume ce to be real. Since we have the lib-
erty to redefine the right-handed electron field as

eR −→ eiΦ eR , (2.16)

which is equivalent to replacing the coupling constant by ce exp(−iΦ), we can conclude

that the phase does not matter for ce . Therefore, without any loss of generality, we can de-
mand ce to be real and positive definite.

36
 Let us now move to the implications of introducing the Higgs field into our system.
Since the field is invariant under SU(2) transformations, there always exists a unitary trans-
formation such that  
 0 
U(x)φ(x) =   (2.17)
√1 ρ(x)
2

where ρ(x) ≥ 0. Since the discussion in Appendix D can be trivially extended and applied
to the case of the Higgs field, its vacuum expectation value, to first degree of approximation
is given by

 
 0 
⟨0| φ(x) |0⟩ =  
⟨0| ρ(x) |0⟩
 
0
=  . (2.18)
ρ0

However, as it is evident, this expectation value is not invariant under the symmetry group
we have been working with so far. It is instead invariant under a U(1) subgroup generated
by T3 + Y, as is shown below

eiχ(x)(T+Y) φ = eiχ(x)((τ3 /2)+yH ) φ

1
≈ 1̂φ + χ (τ3 + yH )φ
2
= φ. (2.19)

Since the vacuum expectation value of the Higgs field is non-zero, we cannot expand around

37
the state as it is given in Eq. 2.17, for it is unstable, and hence, unfit for perturbation theory.
Therefore, we define a new field φ′ , such that

ρ′ (x) = ρ(x) − ρ0

⟨0| ρ′ (x) |0⟩ = 0. (2.20)

In defining this new field, we have essentially shifted the origin such that it coincides with
one of the stable ground state solutions, making it possible to integrate it in our scheme
of perturbation theory. This redefinition, in spirit, is nothing but choosing one of the in-
finitely many ground states available, whose consequences are observed when we shift the
entire Lagrangian by ρ0 and rewrite the Lagrangian in terms of ρ′ . A few terms in the result-
ing Lagrangian interest us the most. From the covariant derivative of the Higgs field, we
obtain the term

( )( )
† a τa ′ a τa ′
⟨0| φ |0⟩ − igWλ − ig B† yH igWλ ig B† yH ⟨0| φ |0⟩
2 2
g2 ρ20 − λ + (g2 + g′ 2 ρ20
= Wλ W Zλ Zλ . (2.21)
4 g

Something remarkable has happened here. The result of these terms signify that the W−
and Z− bosons are now massive. Since their masses appear to depend on the vacuum ex-
pectation value of the Higgs field, in the case that it is large enough, the bosons can be-
come arbitrarily heavy, and in accordance to the experiment. The boson masses can be

38
re-expressed in terms of observable quantities as

g2 ρ2o e2 ρ20
m2W = =
4 4sin2 θw
g2 + g′ 2 ρ20 e2 ρ20
mm2Z = = . (2.22)
4 4sin2 θw cos2 θw

Now we turn to coupling term between the fermions and the Higgs field. Massaging the
expression gives

 
( )
[ †  ν eL  ]
−ce ēR ⟨0| φ |0⟩   + ν¯eL ēL ēR ⟨0| φ |0⟩ eR
eL
ρ
= −ce √0 (ēR eL + ēL eR )
2
ρ
= −ce √0 ēe, (2.23)
2

where we see how although electrons acquire mass, neutrinos are neatly spared* . We can
read off the electron mass to be
ρ
me = ce √0 . (2.24)
2

With this, we conclude setting up the preamble for quantum flavour dynamics. The com-
plete Lagrangian governing electrodynamics and weak interactions for the electron-neutrino
* This was considered a remarkable feat when first discovered, since the neutrinos were then regarded as
massless entities

39
family of leptons is given as

1 1
L = − Tr(Wλρ (x)Wλρ (x)) − Bλρ (x)Bλρ (x)
2 4
1
+ ν̄L iγλ ∂λ νl + ēiγλ ∂λ e + ∂λ ρ′ ∂ λ ρ′
2
( ρ ′ )2
1 ( ρ ′ )2 ( ρ′ )
− + 2 λ 2
+ W λ W λ mW 1 + + Z λ Z mz 1 + − me ēe 1 +
ρo 2 ρo ρo
1 ′
[ ρ ′ (
1 ρ ′ )2 ]
− m2ρ′ (ρ 2 ) 1 + +
2 ρo 4 ρo

+ L ′,

where L ′ refers to the coupling terms in Eq. 2.10.

We now dissect this Lagrangian. Our scheme has yielded us a massless vector boson Aλ
that mediates electrodynamics; three massive bosons, W± and Z; a massless neutrino; and,
a massive electron. Although we begun our discussion by requiring the electron and the
neutrino to enter the Lagrangian in a symmetric manner, our final Lagrangian discrimi-
nates between the two rightly, as predicted by experiment. Furthermore, we see that our
formulation predicts the existence of another particle, ρ′ , with mass

m2ρ′ = 2λ ρ20 . (2.25)

This particle is referred to as the Higgs particle. For decades, its search continued before it
was finally observed in 2012 at the Large Hadron Collider.

40
 Extending to Other Quarks and Leptons
3
4
In the previous chapter, we set up the preamble for quantum flavour dynamics consid-
ering only one family of leptons, i.e., the electron-neutrino pair. We will now extend the
discussion to other Fermions.

41
 All elementary Fermions are combined into a total spinor, ψ.

 
ν eL
 
 
 eL 
 
 
ψ= 
 eR  (3.1)
 
 .. 
 . 
 
bR

Here, all left-handed fermions transform as isodoublets under SU(2) transformations,

while their right-handed counterparts transform as singlets. Hypercharge values corre-
sponding to each fermion are chosen such that when added to the third-component of
their respective weak isospin, they yield experimentally observed electrical charge values. By
minimal substitution, the covariant derivative pertaining to this total spinnor is written as

Dλ ψ = (∂λ + igWaλ Ta + ig′ Bλ Y)ψ. (3.2)

However, unlike the previous case, the Yukawa coupling, or the coupling between the
fermions and the Higgs field, is not trivial. We instead have

Lyuk = ψ̄(x)φi (x)Ci ψ(x) + hermitian conjugate, (3.3)

where Ci , with i ∈ 1, 2, are complex matrices that act on the flavour index of ψ. Our re-
⊗
quirement that the coupling respect the SU(2) U(1) symmetry of the Lagrangian re-
stricts the form of these matrices.
We generalize the Lagrange density in Eq. ?? to cater to the Fermion spinnor as given in

42
Eq. 3.1. The result is

1 1
L = − Tr(Wλρ Wλρ ) − Tr(Bλρ Bλρ )ψ̄iγλ Dλ ψ + Lyuk
2 2
+ (Dλ φ)† (Dλ φ) − V(φ). (3.4)

Note that this expression exhibits all the symmetries we have discussed in the previous sec-
tion. We now again choose one of the infinitely many possible ground states of the Higgs
field by shifting the origin by the vacuum expectation value of the field. Re-expressing the
field in terms of ρ′ yields

1 1
LQFD = − Tr(Wλρ Wλρ ) − Tr(Bλρ Bλρ )
2 2
( ρ ′ )2
1 ( ρ ′ )2
+ 2 λ 2
+ W− λ W λ m W 1 + + Z λ Z m z 1 +
ρo 2 ρo
( ρ ′ )2
+ ψ̄iγλ Dλ ψ − ψ̄Mψ 1 +
ρ0
1 1 [ ρ′ 1 ( ρ′ )]
+ ∂λ ρ′ ∂ λ ρ′ − m2ρ′ ρ′ 2 1 + + , (3.5)
2 2 ρ0 4 ρ0

where
ρ
M = −(C2 + C†2 ) √0 . (3.6)
2

We now study the interaction terms that are given by the covariant derivative.

43
 Lint = −ψ̄γλ (gWaλ Ta + g′ Bλ Y)ψ
λ 1 λ
= −e [ Aλ Jem ++ Zλ JNC (3.7)
sinθw cosθw
1 λ − λ†
+√ λ JCC + Wλ JCC ) ],
(W+ (3.8)
2 sinθw

where

λ
Jem = ψ̄γλ (T3 + Y)ψ
λ
JNC = ψ̄γλ (T3 − sin2 θw (T3 + Y))ψ = ψγλ T3 ψ − sin2 θw Jem
λ

λ ¯ λ (T1 + iT2 )ψ.

JCC = psiγ (3.9)

λ
As can be seen from the electromagnetic current term Jem , the electric charge of the fermions
is given by
Q = T3 + Y, (3.10)

where specific values of T3 and Y for each species yields specific charges.

44
Figure 3.1: The table represents the internal boson lines in quantum flavour dynamics. As can be seen, in the limit
mW ≥ q and mZ ≥ q, the propagator is reduced to a constant value.

Let us now consider the bosons involved in our theory. It has been experimentally ob-
served that processes mediated by W- and Z- are essentially point interactions, while those
mediated by the photon may occur at a distance. Given the analytical expressions of the
propagators in Fig. 3.1, it is straightforward to see that W- and Z- bosons having acquired
mass can be reduced to constant functions in the limits q2 << m2w and q2 << m2Z .
Constant functions in the momentum space correspond to δ-like functions in the position
space, indicating that it is indeed possible for the two to mediate point interactions.

45
 The Lagrangian of the Standard Model of
4
Particle Physics

4
So far, we have only been working with weak and electromagnetic interactions, ignor-
ing strong forces altogether* . In this chapter, we will combine all three of these forces into
* A brief introduction to strong forces can be found in Appendix B

46
one Standard Model Lagrangian. We begin our task by noting that since the leptons only
participate in weak and electromagnetic interactions, they will be ignored by SU(3) trans-
formations. Mathematically, this means that all leptons behave like singlets under SU(3)
transformations, while quarks behave like triplets. Therefore, we use the following repre-
sentation of transformations

F = SU(3) × SU(2) × U(1), (4.1)

which acts on Ψ, the spinnor that combines all quarks and gluons.
Given this description, we have now the complete, unbroken Lagrangian

1 1 1
L = − Tr(Gλρ Gλρ ) Tr(Wλρ Wλρ ) − Tr(Bλρ Bλρ )
2 2 2
+ ψ̄iγλ Dλ ψ + Lyuk + (Dλ φ)† (Dλ φ) − V(φ). (4.2)

Here
Dλ ψ = (∂λ + iga Gaλ Fa + igWaλ Ta + ig′ Bλ Y)Ψ, (4.3)

where Ga is the gluon field introduced in Appendix B.

In the exact same manner as before, we shift the ρ field by its expectation value, which

47
yields

1 1 1
L = − Tr(Gλρ Gλρ ) Tr(Wλρ Wλρ ) − Tr(Bλρ Bλρ )
2 2 2
( ρ ′ )2
1 ( ρ ′ )2
+ 2 λ 2
+ W− λ W λ m W 1 + + Z λ Z m z 1 +
ρo 2 ρo
∑{ [ ( ρ ′ )]}
+ ν̄lL iγλ ∂λ νlL + l̄ iγλ ∂λ − ml 1 +
ρ0
l
∑ [ ( λa ) ( ρ′ )]
+ q̄ iγλ ∂λ + igs Gaλ − mq 1 + q
q
2 ρ0
1 1 [ ρ′ 1 ( ρ′ )]
+ Lint + ∂λ ρ′ ∂ λ ρ′ − m2ρ′ ρ′ 2 1 + + . (4.4)
2 2 ρ0 4 ρ0

Here the dummy variables l and q represent leptons and quarks respectively. The interac-
tion term, L , carries the same form as before, with only the electromagnetic current having
changed to
∑ ∑
λ
Jem = −l̄ γλ l + Qq q̄ γλ q (4.5)
l q

in order to account for quark charges.

The particle content in this Lagrangian is straightforward to interpret, as it has been bor-
rowed exactly from previous sections. The symmetry group of the Lagrangian is, however,
worth a comment. Owing to the process of spontaneous symmetry breaking, the SU(2)
× U(1) symmetry has been broken to U(1), which we previously saw, corresponds to the
observed electromagnetic interactions. The symmetry breaking process does not touch the
SU(3) group, therefore leaving us with

F = SU(3) × SU(2) × U(1) −→ SU(3) × Uem (1). (4.6)

48
 Part III

Renormalization at the One-loop Level

49
 Renormalization of the propagator in the
5
φ3 context

6
Before we proceed with the renormalization scheme, let us discuss why the need for it
emerges after all. The approach to quantum field theories we have been considering so far
relies on the perturbative expansion of the S-matrix, which gives us the probability of an

50
Figure 5.1: Higher‐order contributions to a tree‐level process in complex scalar theory.

initial state ending in some final state. Although the lowest-order term in the expansion is a
very good approximation of the experimentally observed amplitude, it is an approximation
nonetheless. We would like theoretically predicted answers to be more accurate and more
precise. Therefore, we come to consider the higher-order terms in the S-matrix expansion.
Some of the possible terms are sketched in Fig. 5.1. Let us consider the first of the four fig-
ures, which is reproduced in Fig. 5.2 after amputating its external legs. Using Feynman
rules, we can assign the propagator lines momenta

1~ 2 1 1 [ ]
2 1
Δ(k ) = Δ̃(k ) + Δ̃(k ) iΠ(k ) Δ̃(k2 ) + O(g4 )
2 2
(5.1)
i i i i

where Δ̃(k2 ) is the free field operator and is defined as

1
Δ̃(k2 ) = . (5.2)
k2 + m2 − iε

Π(k2 ) here refers to the self-energy of the particle, and therefore has a magnitude given by

51
Figure 5.2: The O(g2 ) corrections to the propagator.

∫
1 1 dd l
iΠ(k ) = (ig)2 ( )2
2
Δ̃((l + k)2 )Δ̃(l2 )
2 i (2π)d
− i (Ak2 + Bm2 ) + O(g4 ). (5.3)

Here, the half factor in front of the integral is to account for symmetry under exchange of
the upper and lower lobes in the loop diagram in Fig. 5.2.
Let’s have a closer look at the integral. We perform a rough dimensional analysis of the
term

∫ ∫
dd l dd l 1 1
Δ̃((l + k)2 )Δ̃(l2 ) = · 2
(2π)d (2π)d (l + k)2 + m − iε l + m2 − iε
2
∫
dd l 1 1
∝ ·
(2π)d l2 l2
∫
dd l 1
∝ .
(2π)d l4

Since the momentum l is not fixed by any conservation law or experiment, the integral
over it runs from 0 to infinity. This means that the integral diverges as l goes to infinity
for d ≥ 4. This brings us to the need to renormalization. To put briefly, renormalization is
a scheme which allows us to regularize infinite contributions to amplitudes. There are mul-

52
tiple methods that help us accomplish this, but for the purposes of this paper, we will make
use of counterterms.
The scheme of using counterterms to regularize integrals relies on the introduction of
coefficients in the Lagrangian, which cancel divergeces term by term. We will put this idea
into action. For that, let’s consider the interacting scalar field theory. The Lagrangian gov-
erning the dynamics of this field is given as

1
L1 = Zg g φ3 + Lct
6
1 1
Lct = − (Zφ − 1) ∂μ φ ∂ μ φ − (Zm − 1)m2 φ2 + Yφ
2 2
1 1 1
=⇒ L = − (Zφ − 1) ∂μ φ ∂ μ φ − (Zm − 1)m2 φ2 + Zg g φ3 + Yφ (5.4)
2 2 6

where m and g are both fixed constants, their value determined by experiment. Here Zg ,
Zφ , and Zm are the counterterm coefficients we ought to determine.
The first step in our scheme is based on casting the integral in Eq. 5.3 in a more tractable
form. To do so, we use the Feynman’s trick or Feynman’s formula to combine denomina-
tors. Given A1 . . . An ,

∫
1
= dFn (x1 A1 + . . . xn An )−n .
A1 . . . An

Here the integration measure dFn is over Feynman parameters, defined as

∫ ∫ 1
dFn = (n − 1)! dx1 . . . dxn δ(x1 + . . . + xn − 1).
0

53
This measure is normalized such as

∫
dFn 1 = 1.

Using Feynman’s formula, the propagators can be re-expressed as

1 1
Δ̃((l + k)2 )Δ̃(l2 ) =
((l + k) + m ) (l + m2 )
2 2 2
∫ 1 [ ]−2
= dx x((l + k)2 + m2 ) + (1 − x)(l2 + m2 )
0
∫ 1 [ ]−2
2 2 2
= dx (l + xk) + x(1 − x)k + m
0
∫ 1 [ ]−2
= dx q2 + D , (5.5)
0

where
q ≡ l + xk,

and
D ≡ x(1 − x)k2 + m2 .

Using this, Eq. 5.3 can be written as is now in the form

∫ ∫
1 2 1
dd l [ 2 ]−2
g dx q + D − Ak2 − Bm2 + O(g4 ), (5.6)
2 0 (2π)d

where we must now evaluate the momentum integral to fix A and B. If we now change
the momentum integration variable from l to q, the integral is cast in the form considered
in the discussion in Appendix E. We borrow from the discussion there and evaluate the

54
integral in terms of Gamma functions. We also allow for the substitution

g −→ gμ̄ε/2 ,

where we have contained all dimension related information in μ̄ as to render g dimension-

less. ε is defined as
ε ≡ 6 − d,

The purpose behind the introduction of ε is to move from a general discussion applicable
for arbitrary d to the one that focusses on a specific value, 6 in this case. However, for now,
we will continue working with undefined d, and only take the limit ε −→ 0 towards the
end. Together with this newly introduced parameter, our evolved integral has taken the
following shape:

∫
Γ(−1 + 2ε ) 1
4π ε/2
dx D ( ) − Ak2 − Bm2 + O(g4 ). (5.7)
(4π)3 0 D

If we now define
g2
α≡ ,
(4π)3

we get

∫ 1
2 1 ε 4πμ̄2 ε/2
Π(k ) = α Γ(−1 + ) dx D ( )
2 2 0 D
− Ak2 − Bm2 + O(g4 ),

55
which combined with the fact that

ε
Qε/2 = 1 + ln(Q) + O(ε2 )
2

in the limit ε −→ 0, the definition

√
μ≡ 4π e−γ/2 μ̃,

and
∫ 1
1
dx D = k2 + m2
0 6

reduces the integral to

∫
2 1
Π(k ) = − α dxDln(D/m2 )
2
1 1 1
= −{ α[ + ln(μ/m) + ] + A}k2
6 ε 2
1 1
− {α[ + ln(μ/m) + ] + B}m2 + O(α2 ). (5.8)
ε 2

We can see that our scheme has confined all divergences in the last two parenthesis, ren-
dering the integral in the first line finite. If we now allow

1 1 1
A = − α[ + ln(μ/m) + + κA ] + O(α2 ) (5.9)
6 ε 2
1 1
B = −{α[ + ln(μ/m) + + κB ] + O(α2 ), (5.10)
ε 2

56
the final result of our regularization procedure is

∫ 1
2 1 1
Π(k ) = α dx D ln(D/m2 ) + α( κA k2 + κB m2 ) + O(α2 ). (5.11)
2 0 6

We see that our choice of A and B has rendered the integral, and hence the amplitude, fi-
nite. It is true that both A and B are formally infinite, but it is okay, since they are not ob-
servables and do not show up in anything measurable.
κA and κB are numerical constants to be determined. For this purpose, we need two re-
straints, which we shall obtain in terms of conditions on Π(k2 ).
Consider the exact propagator of the theory

1
Δ(x1 − x2 ) ≡ ⟨0| Tφ(x1 )φ(x2 ) |0⟩ . (5.12)
i

The propagator gives us information regarding the probability of a particle traveling from
one point in space to another. While it is perfectly permissible for it to do in the fashion
depicted at the tree-level (which is also the most dominant contribution), other mecha-
nisms may exist too. We, therefore, define the exact propagator as a sum of all one-particle
irreducible diagrams, as shown in Fig.??. Luckily, this appears as a geometric series

1~ 2 1 1 1
Δ(k ) = Δ̃(k2 ) + Δ̃(k2 )[iΠ(k2 )] Δ̃(k2 )
i i i i
1 1 1
+ Δ̃(k2 )[iΠ(k2 )] Δ̃(k2 )[iΠ(k2 )] Δ̃(k2 )
i i i
+ ...,

57
which sums up to
~ 2) = 1
Δ(k . (5.13)
k2 + m2 − iε − Π(k2 )

Since we know that the exact propagator has a pole at k2 = −m2 with residue 1, we must
have

Π(−m2 ) = 0 (5.14)

Π′ (−m2 ) = 0 (5.15)

for Eq. 5.13 to be consistent with Eq. 5.12. This gives us the two conditions we needed to
fix κA and κB . Using these conditions gives

∫ 1
′ 2 1 1
Π (−m ) = 0 =⇒ α dx[(1 − x)x + (1 − x)xln[1 − x(1 − x)] + α( κA ) = 0
2 0 6
√
7 π 3
κA = − + (5.16)
3 2

and

∫ 1
2 1 [1 − x(1 − x)]m2
2
Π(−m ) = 0 =⇒ α dx [1 − x(1 − x)]m ln{ }+
2 0 m2
1
α (− κA m2 + κB m2 ) + O(α2 ) = 0
6
1 π 19 1
=⇒ ( √ − ) + (− κA + κB ) = 0
2 2 3 18 6
1 1 π 19
=⇒ − κA + κB = − ( √ − )
6 2 2 3 18
1 √
=⇒ κB = (−11 + π 3). (5.17)
12

58
To summarize, with our dimensional regularization scheme, we attempted to resolve the
issue of divergent contributions to correction terms. We did this by following the following
steps:

1 Use Feynman parameters to re-express amplitude of Feynman propagators.

2 Perform a Wick rotation to obtain a d-dimensional Euclidean vector.

3 Evaluate the integral in d-dimensions using spherical polar coordinates.

4 Introduce a series of parameters and take the limit ε −→ 0, where ε fixes dimensions.

5 From the obtained expression, read off what values counter-term constants must
take to cancel divergences.

This procedure is standard and we will be actively making use of it in the following chap-
ters.

59
 6
Renormalization of the vertex in the φ3
context

60
Figure 6.1: The O(g3 ) corrections to the vertex.

Let us consider the total amplitude contribution from i Zg g and Fig. 6.1 6 .

∫
1 dd l
iV3 (k1 , k2 , k3 ) = iZg g + (ig) ( )33
Δ̃((l − k1 )2 )Δ̃((l + k2 )2 )Δ̃(l2 )
i (2π)d
+ O(g5 ) (6.1)

Here k1 , k2 , and k3 correspond to the momenta of the three external lines, with the sign of
the k0i term fixed by their incoming/outgoing nature. Furthermore, momentum conserva-
tion demands that
k1 + k2 + k3 = 0.

We can perform a dimensional analysis to conclude that the integral in Eq. 6.1 diverges
for d ≥ 6. The method detailed in the previous chapter is equally applicable in this case as
well, and we shall be using it to regularize divergences. We begin by Feynman parameteriza-

61
tion of the propagators.

Δ̃((l − k1 )2 )Δ̃((l + k2 )2 )Δ̃(l2 )

∫
= dF3 {x1 [(l − k1 )2 + m2 ] + x2 [(l + k2 )2 + m2 ] + x3 [l2 + m2 ]}−3
∫ [ ]−3
2 2 2 2
= dF3 l − 2l · (x1 k1 − x2 k2 ) + x1 k1 + x2 k2 + m
∫ [ ]−3
= dF3 (l − x1 k1 + x2 k2 )2 + x1 (1 − x1 )k21 + x2 (1 − x2 )k22 + 2x1 x2 k12 + m2
∫ [ ]−3
= dF3 q2 + D . (6.2)

Here
q ≡ l − x1 k1 + x2 k2 (6.3)

and

D ≡ x1 (1 − x1 )k21 + x2 (1 − x2 )k22 + 2x1 x2 k12 + m2 . (6.4)

Again, after changing the integration variable from l to q, we we perform a Wick rota-
tion over the q0 contour, which yields us the following integral.

∫ ∫
2 dd q̄ 1
V3 (k1 , k2 , k3 )/g = Zg + g dF3 + O(g4 ) (6.5)
(2π) (q̄ + D)
d 2 3

This integral can be evaluated using the technique elaborated in Appendix E. The result is

∫
dd q̄ 1 Γ(3 − 21 d) −(3−d/2)
= D . (6.6)
(2π)d (q̄2 + D)3 2(4π)d/2

62
Substituting this result in Eq. 6.5 along with a few familiar substitutions allows us to write

∫ ( 4πμ̄2 )
1 ε
V3 (k1 , k2 , k3 )/g = Zg + αΓ( ) dF3 + O(α2 )
2 2 D
[ ∫ ]
1 2 ( 4πμ̄2 )
= Zg + α + dF3 ln γ + O(α2 ) (6.7)
2 ε eD

where we have defined α = g2 /(4π)3 . The last equation was obtained under the limit
ε −→ 0 and the normalization condition in Eq. ?? was used. Given that we allow

μ2 = 4πe−γ μ̃2 , (6.8)

and set
Zg = 1 + C

we can rewrite Eq. 6.7 as

1
V3 (k1 , k2 , k3 )/g = 1 + {α[ + ln(μ/m)] + C}
∫ ε
1
− α dF3 ln(D/m2 )
2
+ O(α2 ), (6.9)

where all divergent terms have been restricted to the paranthesis in the first line. To for-
mally cancel these contributions, we allow

1
C = −α[ + ln(μ/m) + κC ] + O(α2 ), (6.10)
ε

63
where κ again is a purely numerical constant. To determine it, we again require conditions
akin to those in Eq. 5.15. However, such a constraint does not exist. We notice, however,
that different values of the constant are contingent upon different definitions of g, which
is typically calculated using experimentally obtained cross-section data. Therefore, we have
the liberty to decide upon some value of κC , after which we must unanimously agree on it.
The simplest choice in this case is
κc = 0,

which means
V3 (0, 0, 0) = 1.

64
 Renormalization in the context of QED
7
So far we have looked at the method of dimensional regularization in the context of the
interacting scalar field theory. In this chapter, we will perform an on-shell renormalization
of the theory using the same methods as discussed earlier. This means that we will demand
the poles of the renormalized propagator to coincide with the physical masses of the species
involved.

65
 The bare QED Lagrangian is given by

1 μν
L (x) = iΨ̄DΨ
/ − mΨ̄Ψ − F Fμν
4

/ = γμ (∂μ − ieAμ ). In spirit of our previous discussion, we add to this Lagrangian

where D
a counter-term Lagrangian that will serve to cancel divergent quantities from our loop inte-
grals. Therefore, the Lagrangian we are to work with is

L1 = Z1 eΨ̄AΨ
/ + Lct (7.1)
1
Lct = i(Z2 − 1)Ψ̄∂/Ψ − (Zm − 1)mΨ̄Ψ − (Z3 − 1) Fμν Fμν (7.2)
4

The interaction term in Eq. 7.1 shows that a QED vertex can entertain both photonic and
fermionic propagators, both of which contain infinite loop integrals. Therefore, renormal-
ization in the context of QED involves addressing both cases separately. We will begin with
the case of the photonic propagator.
Let’s begin by considering the exact photon propagator in the momentum space.

~ μν (k) = Δ̃μν (k) + Δ̃μρ (k) Πρσ (k)Δ̃σν (k) + . . .

Δ (7.3)

Here Πμν (k) is the sum of 1P-I diagrams and

1 [ kμ k ν ]
Δ̃μν (k) = g μν − (1 − ζ) , (7.4)
k2 − iε k2

where ζ is the parameter that allows us to switch between gauges. Since different gauges

66
merely correspond to different descriptions of the same physical setup, observables, such as
amplitude and cross-sections, should not depend on ζ. This is only achievable if we demand
Πμν (k) to be transverse to the propagator, or

kμ Πμν (k) = kν Πμν (k) = 0. (7.5)

This constraints the form of Πμν (k). Quite generally, we can say that

( )
Πμν (k) = Π(k2 ) k2 gμν − kμ kν

= k2 Π(k2 )Pμν (k) (7.6)

where Π(k2 ) is a scalar function. Using Eqs. 7.4 and 7.6 in Eq. 7.3, and summing the series
gives us
~ μν (k) = Pμν (k) kμ kν /k2
Δ + ζ . (7.7)
k2 [1 − Π(k2 )] − iε k2 − iε

Since the exact propagator must have a pole at k2 = 0 with residue Pμν (k)/[1 − Π(0)], we
must demand
Π(0) = 0 (7.8)

for Eq. 7.7 to be consistent.

With the conditions on Π(k2 ) and Πμν (k) in place, we now move to dealing with diver-
gent integrals directly. The amplitude of the photon propagator at the one-loop level with

67
Figure 7.1: One‐loop and counterterm corrections to the photon propagator

counter-term contributions, as shown in Fig. ??, is given as follows

∫ [ ]
1 d4 l
iΠ (k) = (−1)(iZ1 e) ( )2
μν 2
Tr S̃(/l + k
/ )γ μ / ν
S̃( l )γ
i (2π)4
− i(Z3 − 1)(k2 gμν − kμ kν ) + O(e4 ) (7.9)

. Here S̃(/p) is the fermion propagator given by

−/p + m
S̃(/p) = .
p2 + m2 − iε

The denominators in Eq. 7.9 can be combined following the usual Feynman prescription.

[ ] ∫ 1
4Nμν
μ ν
Tr S̃(/l + /k)γ S̃(/l )γ = dx , (7.10)
0 (q2 + D)2

where
D = x(1 − x)k2 + m2 − iε (7.11)

68
and

[ ]
4Nμν ≡ Tr S̃(/l + /k)γμ S̃(/l )γν
{ }
= 4 (l + k)μ lν + lμ (l + k)ν − [l(l + k) + m2 ]gμν . (7.12)

A closer look at this expression tells that it diverges for d ≥ 4, which is why we analytically
continue to d dimensions. To keep e dimensionless throughout, we replace it with eμ̃ε/2 ,
where ε = d − 4, similar to the cases discussed in the previous chapter. In the concluding
steps of our problem, we will take the limit ε −→ 0 to specialize our discussion for d = 4.
Next, we must evaluate the momentum integrals in Eq. 7.9. To do so, we first set l =
q − xk and use

∫
dd q qμ f(q2 ) = 0
∫ ∫
d μ ν 2 gμν
d qq q f(q ) = dd qq2 f(q2 ) (7.13)
d

on the resulting expression. The first of the two equations here is valid for it is odd under
parity change, while the second one can be proved by contracting both sides of the equa-
tion with gμν . Using them together allows us to drop terms linear in qμ because they eventu-
ally integrate to zero, and replacing all qμ qν terms with gμν /d, as shown below:

[( ) ]
μν μ ν 2
N −→ −2x(1 − x)k k + 2 2 2
− 1 q + x(1 − x)k − m gμν . (7.14)
d

69
To further simplify this expression, we note that

(1 )∫ dd q q2
∫
dd q 1
d−1 =D , (7.15)
2 (2π)d (q2 + D)2 (2π) (q + D)2
d 2

which can be shown using these Gamma function properties

( 1 ) ( 1 ) ( 1 )
1− d Γ 1− d =Γ 2− d and Γ(1 + z) = zΓ(z).
2 2 2
( )
1
Since the integrals in Eq. 7.15 are essentially equal, we can replace 2
d − 1 with D in
the expression of Nμν in foresight. This reduces its expression to

Nμν −→ 2x(1 − x)(k2 gμν − kμ kν ) (7.16)

If we look closely at this expression, it is precisely in line with the form we predicted Πμν (k)
to have in Eq. 7.6, hinting that we are on the right track.
Now, performing all the usual tricks we have been working with so far, we conclude that

∫ 1 [1 1 ]
2 e2
Π(k ) = − 2 dx x(1 − x) − ln(D/μ ) − (Z3 − 1) + O(e4 ).
2
(7.17)
π 0 ε 2

where we have replaced Z1 with 1 + O(e2 ). Imposing the condition in Eq. 7.5, we fix

e2 [ 1 ]
Z3 = 1 − − ln(m/μ) + O(e4 ), (7.18)
6π ε
2

70
Figure 7.2: One‐loop and counterterm contributions to the fermion propagator

which leaves behind a finite Π(k2 ), as wanted.

∫ 1
2 e2
Π(k ) = 2 dx x(1 − x) ln(D/m2 ) + O(e4 ) (7.19)
2π 0

We now turn to the case of the Fermion propagator at the one-loop level. We define the
exact propagator S̃(/p) to all orders of magnitude and sum up the resulting geometric series
to obtain
1
S̃(/p) = , (7.20)
/p + m − iε − Σ(/p)

where again to be consistent with the pole at /p = −m and residue 1 requirement, we must
have
Σ(−m) = 0 and Σ′ (−m) = 0. (7.21)

Using Fig. 7.2 in conjunction with Feynman rules, we write down the amplitude for one-
loop and counter-term contributions to the Fermion propagator.

∫
1 d4 l [ ν ]
iΣ(/p) = (iZ1 eμ̃ ) ( )2
ε 2 μ
γ S̃(/p + /l )γ Δ̃μν (l)
i (2π)4
− i(Z2 − 1)/p − i(Zm − 1)m + O(e4 ) (7.22)

71
where the inclusion of μ̃ε absorbs the dimensions of e. For the purposes of this calculation,
it is simplest to work in Feynman gauge. Therefore, we set ζ = 0, so that

gμν
Δ̃(l) = .
l2 + m2γ − iε

Combining denominators using the Feynman parametrization yileds

∫ 1 ∫
2 ε dd q N
iΣ(/p) = e μ̃ dx
0 (2π) (q + D)2
d 2

− i(Z2 − 1)/p − i(Zm − 1)m + O(e4 ) (7.23)

where

q = l + xp (7.24)

D = x(1 − x)p2 + xm2 + (1 − x)m2γ (7.25)

N = γμ (−/p − /l + m)

= −(d − 2)[/q + (1 − x)/p] − dm. (7.26)

Here we have used γμ γμ = 4 and γμ /p γμ = (d − 2)/p. Results summarized in Eq. 7.13

have been used to drop terms linear in qμ . Delving into our usual bag of tricks, we evaluate
the momentum integrals, which gives us

∫ 1 ( ) [1 1 ]
e2
Σ(/p) = − 2 dx (2 − ε)(1 − x)/p + (4 − ε)m − ln(D/μ2 )
8π 0 ε 2
− (Z2 − 1)/p − (Zm − 1)m + O(e4 ). (7.27)

72
Figure 7.3: The one‐loop corrections to the photon‐fermion‐fermion vertex.

Now, to cancel infinite terms in Σ(/p), we must have

e2 ( 1 )
Z2 = 1 − + finite + O(e4 ) (7.28)
8π2 ε
e2 ( 1 )
Zm = 1 − 2 + finite + O(e4 ). (7.29)
2π ε

To fix the condition Σ(−m) = 0 we write

∫
e2 [ 1 ( ) ]
Σ(/p) = 2 dx (1 − x)/p + 2m ln(D/Do ) + κ2 (/p + m) + O(e4 ) (7.30)
8π 0

where κ2 is fixed using the condition Σ′ (−m) = 0

κ2 = −2ln(mγ ) + 1 (7.31)

We now turn to our final task - renormalization of the QED vertex. If we define the vertex
function iVμ (p′ , p) to be the sum of all one-particle irreducible diagrams with incoming
momentum p and outgoing momentum p′ , the vertex function at the one-loop level would

73
be then
μ
iVμ (p′ , p) = iZ1 eγμ + iV1−loop (p′ , p) + O(e5 ) (7.32)

μ
where V1−loop in light of Fig. 7.3 is given by

∫
μ 3
( 1 )3 d4 l [ ρ ′ μ ν
]
iV1−loop = (ie) γ S̃(/p + /l )γ S̃(/p + /l )γ Δ̃νρ (l). (7.33)
i (2π)4

If we now perform Feynman parameterization and take

q ≡ l + x1 p + x2 p′ , (7.34)

D ≡ x1 (1 − x1 )p2 + x2 (1 − x2 )p′ 2 − 2x1 x2 p·′ (7.35)

+ (x1 + x2 )m2 + x3 m2γ , (7.36)

Nμ ≡ γν (−/p′ − /l + m)γμ (−/p − /l + m)γν

= γν /q γμ /q γν + Ñμ + (linear in q) (7.37)

where

Ñμ = γν [x1 /p − (1 − x2 )/p′ + m]γm u[−(1 − x1 )/p + x2 /p′ + m]γν , (7.38)

we can re-express the one-loop vertex function as

∫ ∫
μ 3 d4 q Nμ
iV1−loop =e dF3 . (7.39)
(2π)4 (q2 + D)3

As it can be seen, only the first term in Nμ diverges for d ≥ 4. We generalize the integral
to d dimension, and make the replacements

74
 1
γν /q γμ /q γν −→ q2 γν γρ γμ γρ γν
d
γμ γμ γρ = (d − 2)γμ . (7.40)

These reduce the integral to

iVμ (p′ , p) = iZ1 eγμ

∫ ∫
e3 [( 1 1 2
)
μ 1 Ñμ ]
+ 2 −1− dF3 ln(D/μ ) γ + dF3 , (7.41)
8π ε 2 4 D

which requires Z1 to take the form

e2 ( 1 )
Z1 = 1 − 2 + finite + O(e4 ) (7.42)
8π ε

to render the vertex function finite 6 .

75
 Part IV

Appendices and References

76
”When a clergyman asked J.B.S. Haldane1 what he had
learned about the Creator after a lifetime of studying
Nature, he answered, “an inordinate fondness for bee-
tles.” If the clergyman were to ask a theoretical physicist
like me a similar question, I would have answered, an
inordinate fondness for group theory.
Group theory governs the universe. Literally.”

Anthony Zee

A Brief Overview of Group Theory

A
Consider a set G with elements {gα }. Then, G is a group if it is equipped with a binary
compositional operation, ⋆, under which it is closed. This means that if gα and gβ ∈ G,
then gα ⋆ gβ = gγ ∈ G. For G to be identified as a group, its elements and binary operation
must adhere to the following axioms:

1. Composition must be associative, meaning that (gα ⋆ gβ ) ⋆ gγ = gα ⋆ (gβ ⋆ gγ )

77
 2. There must exist an identity element, I, with respect to the operation, such that I ⋆
gα = gα ⋆ I

3. Each element gα must have a unique inverse g−1

α such that gα ⋆ gα = gα ⋆ gα = I
−1 −1

Note that there is no necessity for the composition of two elements to commute. In case
composition is commutative, G is known as an Abelian group. In the other case, it is known
as non-Abelian. We shall now consider a few examples of groups:

1. The set of all real numbers, R, is a group under the operations ⋆ ≡ + and ⋆ ≡ ×.
This is also an example of an Abelian group.

2. The set of all 2 × 2 matrices, GL2 (R), under matrix multiplication forms a group.
Since matrix multiplication is not generally commutative, this group is non-Abelian. 7

78
 A Brief Introduction to Quantum
B
Electrodynamics

3
Consider the following non-interacting Dirac Lagranian governing the dynamics of n
Leptons.
∑
Li = Ψ̄i (i ∂/ − m) Ψi (B.1)
i

79
Here, Ψi represents the spinnor of the ith Lepton. Borrowing from the discussion in Ch. 1
and using the principle of minimal substitution, we see that this Lagrangian remains under
transformations of the form

Ψ1 −→ Ψ′1 = Ψ ei g1 θ

Ψ2 −→ Ψ′2 = Ψ ei g2 θ
..
.

Ψn −→ Ψ′n = Ψ ei gn θ , (B.2)

where θ1 , . . . , θn are complex constants independent of spatial coordinates. Much in the

spirit of Eq. 2.7, we combine all the spinnors in one spinnor, Ψ and define a diagonal charge
matrix Q, such that its entries correspond to the observed Lepton charges.

 
 0 0 . . . 0 
 
0 −1 . . . 0 
 
Q = . .  (B.3)
 .. ... . 
 . 
 
0 0 . . . − 31

If we now turn to make these transformations local and demand the Lagrangian to remain
invariant under their action, we must have transform the derivative according to Eq. 1.19.
We, therefore, now have a massless vector boson, Aμ , which transforms according to Eq. ??.
For the U(1) Abelian case, this translates into transformation of the form

Aμ −→ A′μ = Aμ + ∂μ θ(x). (B.4)

80
This boson, also known as the photon in this case, mediates interactions whose form is
given by the covariant derivative
i q Ψ̄ A
/μ Ψ. (B.5)

81
 A Brief Discussion of Quantum
C
Chromodynamics

Quantum Chromodynamics is a non-Abelian gauge theory, based on the principle of

flavour universality of the strong forces, i.e., strong forces do not distinguish between quarks
of different flavours, and equality of the masses of u-, d-, and s- quarks. This contends that

82
these quarks are equivalent and it is reasonable to work with any of their orthogonal linear
combinations. Transformations between different sets of these linearly independent com-
binations correspond to SU(3) transformations in the state-space. The generators of this
group are the 3 × 3 linearly independent and Hermitian matrices with trace 0. They are
provided as follows:

     
0 1 0 0 −i 0 1 0 0
     
λ1 = 
1 0 0, λ2 = 
 i 0 0  , λ3 = 0
  −1 0,
     
0 0 0 0 0 0 0 0 0
     
0 0 1 0 0 −i 0 0 0
     
λ4 = 
0 0 0, λ5 = 
 0 0 0  , λ 6 = 0
  0 1,
     
1 0 0 i 0 0 0 1 0
   
0 0 0 1 0 0 
  1  
λ7 = 
0 0 −i, λ8 = √  0 1 0 . (C.1)
  3


0 i 0 0 0 −2

Corresponding to these eight generators of rotations, we have eight vector fields, Gλ that
mediate strong interactions. Therefore, the Dirac Lagrangian, as given in Eq. B.1, but

83
suited to deal with local SU(3) transformations in the complex state space is given by

f
1 ∑
Lq (x) = − Tr(Gλρ (x)Gλρ (x)) + q̄ j (x) (iγλ Dλ − mj )q j (x)
2 j=1

1 ∑ f [ ( λa ) ]
= − Tr(Gλρ (x)Gλρ (x)) + q̄ j (x) iγλ ∂λ + igq Gaλ (x) − mj q j 4 .
2 j=1
2

(C.2)

84
 Spontaneous Symmetry Breaking
D
Spontaneous symmetry breaking offers an explanation to the phenomenon where a ground
state fails to demonstrate the symmetry of the system. We begin by considering two scalar
fields φ1 (x) and φ2 (x). Their dynamics are governed by the following Lagrangian

Lφ (x) = ∂μ φ†1 (x) ∂ μ φ1 (x) + ∂μ φ†2 (x) ∂ μ φ2 (x) + V(φ1 , φ2 ), (D.1)

85
where
V(φ1 , φ2 ) = −kφ† (x)φ(x) − λ(φ† (x) φ(x))2 . (D.2)

Here κ = −μ2 < 0 and λ > 0. A plot visualizing this potential can be found in Fig. D.1.
As can be seen in Fig. D.1, the system exhibits SU(2) invariance. Therefore, we demand

Figure D.1: A visualization of the potential V(φ1 , φ2 ), also known as the Mexican hat potential. Notice how the poten‐
tial does not have a unique minimum, rather a continuous circle of minima.

the fields to be such that they form a doublet under SU(2) rotations, i.e., the Lagrangian in
Eq. D.1 is does not change under SU(2) transformations . Therefore, it is illuminating to

86
combine the two fields as  
 φ1 (x) 
Φ(x) =  , (D.3)
φ2 (x)

so the Lagrangian looks like

Lφ (x) = ∂μ φ† (x) ∂ μ φ(x) + V(φ), (D.4)

where
V(φ) = −kφ† (x)φ(x) − λ(φ† (x) φ(x))2 . (D.5)

The total energy of this system is given by Hamiltonian, H , where

∫
H (x) = d3 x[φ̇† φ̇ + ∇φ† ∇φ + V(φ)]. (D.6)

Since the derivative terms in this Hamiltonian are positive definite, the energy of the system
must have a lower bound given by the minimum of the potential function. Let φ(x) ≡ φ
√
be the constant potential that minimizes the potential. If we define ρ/2 to be the length
of this constant potential, i.e.,
√
ρ
φ† φ = √ , (D.7)
2

the minimum of the potential can be written as

1 1
V(φ) = − μ2 ρ2 + λρ4 . (D.8)
2 4

87
The value of ρ that minimizes this function is given by

√
μ2
ρ = ρo = . (D.9)
λ

Since φ is invariant under SU(2) rotations, we can rotate the field in any one of the in-
finitely many possible directions. The only restraint the field suffers from is the length as
found in D.9. Therefore, we can rewrite φ as given in Eq. D.3 as

 
 0 
Φ= . (D.10)
√1 ρ
2 o

As is evident from this result, picking a particular ground state (or orientation), sponta-
neously breaks the symmetry of the problem. The φ as in Eq. D.10 is no longer invariant
under SU(2) rotations 4 .

88
 Momentum Integration in Loops
E
Following the introduction of Feynman parameters, xi , loop integrals over momenta take
the form 2
∫
dd q (q2 )r
Isr (D, d) ≡ . (E.1)
(2π)d (q2 + D)s

89
Our job is to obtain a general solution for this integral. Let us inspect what we have. q is a
d-dimensional vector in the Minkowski space, such that

q = (q0 , q1 , . . . , qd ) and q2 = −q20 + q21 + . . . + q2d . (E.2)

Let us think of the integral over q0 from −∞ to +∞ as a contour integral in the complex
q0 space. Since the poles of Feynman propagators typically exist slightly above the negative
real axis and slightly below the positive real axis, a π/2counterclockwise rotation of the
contour does not pass through any poles. This transformation is therefore permissible and
does not change the result of integration. However, it does another, more useful thing for
us. Rotating the contour as such means that the integration now runs over the imaginary
q0 axis. We, therefore, define a new vector q̄ such that 6

q̄0 = −i q0 and q̄i = qi (E.3)

so that although q2 = q̄2 , but now

q̄2 = q̄20 + q̄21 + . . . + q̄2d , (E.4)

meaning that we now have a Euclidean vector in hand. This transformation is known as
the Wick rotation and serves to take integrals from the Minkowski space to the Euclidean
space, which allows us to manipulate integrals using the bag of tricks familiar to us. Given

90
this scheme, we write in general

∫ ∫
d 2
d q f(q − iε) = i dd q̄ f(q̄2 ),

and the integral in Eq. E.1 can be most generally expressed as

∫
dd q̄ (q̄2 )r
Isr (D, d) ≡ i (−1) r+2
. (E.5)
(2π)d (q̄2 + D)s

Integration over q̄ is now, in fact, an integration in a d-dimensional Euclidean space. It is

simplest to evaluate this expression using spherical polar coordinates. Our scheme will be as
follows: we shall break the integral into two parts, i.e., polar and radial. We shall then evalu-
ate both parts independently and obtain general results for them, which we shall ultimately
combine to write a solution for Eq. E.5.
The shift from Cartesian coordinates to spherical polar in a d-dimensional vector space
can be written as
∫ ∫ ∞
d
d x f(r) = Ωd dr rd−1 f(r),
0

where r is the radial vector such that

N
∑
q̄2i = r2 ,
i=1

and Ω is the polar part of the integral and the part we shall approach first.
Consider the integral

∫
2
Id ≡ dd r e−r̄ . (E.6)

91
This is a standard expression, easily solvable using both Cartesian and spherical polar coor-
dinates. We will first evaluate using the latter coordinate system. We have

∫ ∞
2
Id = Ω d dr rd−1 e−r
0
∫ ∞
1
= Ωd du ud/2−1 e−u
2 0
1 1
= Ωd Γ( d), ??
2 2

where in the second last line we introduced a reparameterization of the form u = −r2 and
left the angular part untouched. The radial part of the integral is in a standard form, with
its solution expressed in terms of Gamma functions, which are defined as

Γ(n + 1) = n!
1 (2n)! √
Γ(n + ) = π
2 n!2n
(−1)n [ 1
n
∑
Γ(−n + x) = −γ+ k−1 + O(x)].
n! x
k=1

We now take a few steps back and evaluate the integral in Cartesian coordinates, xi , which
gives

d ∫
∏ +∞
2
Id = dxi e−xi
i=1 −∞
√
= ( π)d

= πd/2 .

92
Since both modes of calculation are essentially equivalent, we can equate results in Eqs. ??
and ??. This allows us to write
2πd/2
Ωd = ).
Γ( 21 d

With the angular part of the integral determined, it remains to calculate the result of the
radial part. First we rewrite Eq. E.5 as

i (−1)r+s 1
Irs (D, d) = Rrs (d), (E.7)
(4π) Γ(d/2) D
d/2 s−r−d/2

2
where D has been extracted from within the primary integral to cast the radial integral,
Rrs (d), in the following form

∫ ∞ 1
y r−1+ 2 d
Rrs (d) = dy . (E.8)
0 (y + 1)s

Here y = q̄2 /D. This integral can be re-expressed as a standard definition upon the intro-
duction of the following substitution

1
z= .
y+1

Given this substitution,

y dy
1−z= , dz = − ,
y+1 (y + 1)2

which we can use in Eq. E.8 to simplify it as

93
 ∫ ∞
dy ( y )r−1+ 21 d ( 1 )s−r−1− 21 d
Rrs (d) =
0 (y + 1)2 y + 1 y+1
∫ 1
1 1
= dz (1 − z)r−1+ 2 d zs−r−1− 2 d
0
1 1
= B(r + d, s − r − d)
2 2
Γ(r + 21 d)Γ(s − r − 21 d)
= (E.9)
Γ(s)

where in the second-last step we have introduced the beta-function, B, which in terms of
Gamma functions is defined as

Γ(z)Γ(z′ )
B(z, z′ ) =
Γ(z + z′ )

Now that the radial integral in arbitrary dimensions has been evaluated, we can sub-
stitute Eq. E.9 in Eq. E.7 to get the solution of a general momentum integral as given in
Eq. E.1. We finally obtain

i (−1)r+s 1 Γ(r + 21 d)Γ(s − r − 21 d)

Isr (D, d) = (E.10)
(4π)d/2 Ds−r− 21 d Γ( 21 d)Γ(s)

This is an important result that we shall be referring to often while working out renormal-
ization at the one-loop level.

94
 References

[1] Aitchison, I. J. R. (1982). Gauge Theories in Particle Physics. Boca Raton, Florida:
CRC Press, 2012.

[2] Lahiri, A. & Pal, P. B. (2004). A First Book of Quantum Field Theory. Boca Raton,
Florida: CRC Press.

[3] Mandl, F. & Shaw, G. (1986). Quantum Field Theory. Hoboken, New Jersey: John
Wiley and Sons.

[4] Nachtmann, O. (1989). Elementary Particle Physics: Concepts and Phenomena.

Midtown Manhattan, New York City: Springer.

[5] Quigg, C. (2013). Gauge Theories of the Strong, Weak, and Electromagnetic Interac-
tions. 41 William Street, Princeton, New Jersey: Princeton University Press.

[6] Srednicki, M. (2007). Quantum Field Theory. The Edinburgh Building, Cambridge:
Cambridge University Press.

[7] Zee, A. (2016). Group Theory in a Nutshell for Physicists. 41 William Street, Prince-
ton, New Jersey: Princeton University Press.

95