Download as pdf or txt
Download as pdf or txt
You are on page 1of 117

Theoretical Physics

Riccardo Fecchio

June 7, 2021
Contents

1 Relativistic wave equations 4


1.1 Rewiew of Quantum Mechanic . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Continuity equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Free Shrödinger equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Review of Special Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Physical meaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.3 KG equation coupled with an EM field . . . . . . . . . . . . . . . . . . . . 11
1.3.4 Prediction of antimatter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.5 The non-relativistic limit . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.6 Klein paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Dirac wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.1 The spinorial space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4.2 Representation of gamma matricies . . . . . . . . . . . . . . . . . . . . . . 19
1.4.3 Other representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.4 Covariance of Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.5 Dirac conjugate spinor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.6 Continuity Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.4.7 Bilinear forms and Lorentz transformations . . . . . . . . . . . . . . . . . 23
1.5 General solution of Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.5.1 Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.5.2 The helicity operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.5.3 Chirality operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.5.4 Dirac equation coupled to EM external field . . . . . . . . . . . . . . . . . 30

2 Lagrangian & Hamiltonian formalism 33


2.1 Finite system: Lagrangian formalism . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Finite system: Hamiltonian formulation . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Continuos system: Lagrangian formulation . . . . . . . . . . . . . . . . . . . . . . 36
2.3.1 Interlude on functional and functional derivative . . . . . . . . . . . . . . 37
2.4 Continuos system: Hamiltonian formulation . . . . . . . . . . . . . . . . . . . . . 38
2.5 Continuos symmetries & Noether Theorem . . . . . . . . . . . . . . . . . . . . . 39
2.5.1 Global continuous symmetry . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5.2 Application: invariance under Poincarè group . . . . . . . . . . . . . . . . 42
2.6 Scalar field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.6.1 U(1) invariance for a complex field theory . . . . . . . . . . . . . . . . . . 46
2.6.2 Application of free scalar field . . . . . . . . . . . . . . . . . . . . . . . . . 46

3 Canonical quantization 48
3.1 Canonical quantization of a real scalar field . . . . . . . . . . . . . . . . . . . . . 49
3.1.1 Normal ordering (for bosons) . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Fock space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

1
CONTENTS 2

3.2.1 Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.2 Consistency with spin-statistic theorem . . . . . . . . . . . . . . . . . . . 52
3.2.3 Normalization in Fock space . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Identification of the field operators . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4 Covariant commutators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4.1 Covariance and microcausality . . . . . . . . . . . . . . . . . . . . . . . . 54
3.5 Canonical quantization of a complex scalar field . . . . . . . . . . . . . . . . . . . 54
3.5.1 Fock space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.5.2 Covariant commutator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.6 Canonical quantization of a Dirac field theory . . . . . . . . . . . . . . . . . . . . 56
3.6.1 Hamiltonian formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.6.2 Momentum space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.6.3 Canonical quantization with anticommutators . . . . . . . . . . . . . . . . 60
3.6.4 Normal ordering (for fermions) . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6.5 Fock space for fermion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6.6 Covariant anticommutator . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.7 Quantization of vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.8 Massive vector boson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.8.1 Proca equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.8.2 General solution of Proca equation . . . . . . . . . . . . . . . . . . . . . . 64
3.8.3 Hamiltonian formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.9 Massless vector boson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.9.1 Gauge invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.9.2 Fixing gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.9.3 The gauge fixing Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.9.4 Covariant quantization of EM field . . . . . . . . . . . . . . . . . . . . . . 71
3.9.5 Fock space & Gupta-Bleur condition . . . . . . . . . . . . . . . . . . . . . 72
3.9.6 Pseudo-photon state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.9.7 Covariant commutators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4 Interacting field theories 77


4.1 Types of interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.1.1 Scalar field self-interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.1.2 Dirac field self-interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1.3 Scalar and Dirac field interactions . . . . . . . . . . . . . . . . . . . . . . 78
4.1.4 Vector and Dirac field interactions . . . . . . . . . . . . . . . . . . . . . . 78
4.2 Canonical quantization of interacting scalar field . . . . . . . . . . . . . . . . . . 79
4.3 Interacting picture for perturbative theory . . . . . . . . . . . . . . . . . . . . . . 80
4.4 S-matrix operator & T-order operator . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5 Wick theorem & Feynman propagator . . . . . . . . . . . . . . . . . . . . . . . . 85
4.5.1 Generalization of Feynman propagator . . . . . . . . . . . . . . . . . . . . 86
4.5.2 Generalization of Wick theorem . . . . . . . . . . . . . . . . . . . . . . . . 87
4.6 Feynman propagator: physical interpretation . . . . . . . . . . . . . . . . . . . . 88
4.6.1 For scalar field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.6.2 For real vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.6.3 For Dirac field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.7 Uncontracted fields: physical interpretation . . . . . . . . . . . . . . . . . . . . . 91

5 Quantum electro dynamic 94


5.1 S-matrix expansion in coordinate space . . . . . . . . . . . . . . . . . . . . . . . 94
5.2 S-matrix expectation value in momentum space . . . . . . . . . . . . . . . . . . . 100
5.2.1 S-matrix expansion in momentum space . . . . . . . . . . . . . . . . . . . 101
5.3 Feynman rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.4 Observable result: the cross section . . . . . . . . . . . . . . . . . . . . . . . . . . 105
CONTENTS 3

5.5 QED process at lowest order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106


5.5.1 Electron-positron to muon-antimuon . . . . . . . . . . . . . . . . . . . . . 107
5.5.2 Compton scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.6 Ward identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

Introduction
Prof. Stefano Rigolin - office 368 via Marzolo. PDF version of the lectures are uploded on Lecture
Moodle. The exam will be oral. On moodle we can find the suggested textbook. This course is 1/3/21
in the middel of relativistic QM and Field Quantization. We will start from a relativistic version
of the wave function that approaches to inconsistency, so we will leave this road and approach
the field quantization. Thsi course will concentrate on QED, the first QFT. We will do just a
couple of explicit calculation.
The program is:

1. Hystory approach: non relativistic wave equations, so extend the Shrödinger non-relativistic
approach to relativistic theory. There are two results, the Klein-Gordon and the Dirac one.
This is a failed approach, we will se why. It is unconsistent for a relativistic reason.

2. Classical field theory: we will write a Lagrangian and the equation of motions (Eulero-
Lagrangian equation). Then we will discuss the Noether theorem and conserved quantities.

3. Canonical quantization in order to connect free fields to QFT: we will see scalar, Dirac
and vector fields (spin 0, 1/2, 1 respectivelly). Notice that till now GR is a calssical Field
theory since there is no proof of the graviton as a particle with some spin.

4. We will se the interaction field: S-matrix, Feynman rules and Feyman diagrams.

5. A couple of applications in QED (they are just an example, we will not do the calculation
at the exam).

We must to know the notation: lagrangians and equations of motions and from a process to
write down the lagrangian. Then we must know the idea to calculate easy processes in QED (at
the exam it will not ask to re-do the all calculation but we must know the order of processes -
like exponential and so on). We have to know the general idea of where a process starts, which
are the important step and where it arrives.
There will be few list of exercices on moodle, tipically they are ordered from simple to
complicate. Mabye the last are to much advanced. They are no officially homework, just a check
point during the course.

Notation
We will adopt the following notation: relativistic (R), non-relativistic (NR), quantum mechanics
(QM), classical (C) where classical signifies non quantize.
Chapter 1

Relativistic wave equations

What we want to do in this course is to extend QM results to relativistic particles. Before to


start we have to fix notation and convention.

Notation & convention In relativistic + quantistic theories two consants show up: ~ =
h/2π = 6.582 × 10−12 M eV /s and c = 2.997 × 108 m/s the speed of light. So it is useful to use
natural unit. It is a convention such that ~ = c = 1 adimensional. To come back to the usual
KMS units we have to know the true dimension of the quantity and up back the dimensional
constants.
The first think to notice is that quantities have strange dimension:

[L]
[c] = = 1 =⇒ [L] = [T ]
[T ]
[M ][L]2
[~] = [S] = = 1 =⇒ [L] = [T ] = [M ]−1
[T ]

where ~ is energy times second, so it is an action S. In natural unit a common representation for
these equal quantities is [M ].
The most relevant consequence of natural units are

[E] = [M ][V ]2 =⇒ [E] = [M ]


[P ] = [M ][V ] =⇒ [P ] = [M ]

Usually in relativistic theories we travel around GeV of energy.


After the choice of natural units, there is a second convention to build up a relativitic theory:
we denote a generic vector as a controvarian vector, namely with upper index, for example:
v µ = (v0 , ~v ), xµ ≡ (x0 , ~x), pµ ≡ (E, p~).
To go from a controvariant vector to a covariant vector we have to use the metric to down
the idicies. There are two opposite way to choice the matrix: the one with the mostly positive or
mostly negative sign. Our second convention is to choice the mostly negative sign, it is the most
diffusive way in particle physisc

η = η µν = diag(1, −1, −1, −1) = ηµν = η −1

So covariant vector can be obtained from controvariant ones by using η and so it comes out
a vector with opposite sign of spatial component, for example: xµ = ηµν xν = (x0 , −~x), pµ =
ηµν pν = (E, −~p).
There is an exeption, the partial derivative is naturally defined as a covariant vector since it
is the derivation respect to a controvariant vector.

∂µ ≡ ~
= (∂0 , ∇) and ~
∂ µ = η µν ∂ν = (∂0 , −∇)
∂xµ

4
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 5

1.1 Rewiew of Quantum Mechanic


Let’s briefly rewiew the principles behind the Schrödinger derivation of the NR wavefunction in
QM. They are two: the correspondence principle and the bohr condition.
The first one is the correspondence principle: consider a CNR particle (classical so no
quantized) of mass M in a potential field V (x), its energy is

p2
E= + V (x)
2M
The corresponding principle tells us that to obtain the quantum description of the system we
have to replace functions with operator:

E → i∂t (1.1)
~
p~ → −i∇
~x → ~x

and the energy equation of the system becomes the Schrödinger eq: the quantistic wavefunction
equation !
d ∇~2
i ψ(~x, t) = − + V (~x) ψ(~x, t) ≡ Hψ(~x, t)
dt 2M
where we introduced the Hamiltonian operator H.
The second important principle is the Bohr condition and it gives physical meaning to
wavefunction: under the correct normalization, its square is the probability density of finding
the particle in a specific region. So we introduce the density of probability

ρ(~x, t) ≡ |ψ(~x, t)|2 ≥ 0

The normalization is such that the probability of finding the particle at time t0 in the total space
is Z
P(t0 ) ≡ dx3 ρ(~x, t) = 1
R3

1.1.1 Continuity equation


The continuity eq is a consequence of principles. It arises from the question: is the probability
constant in time? Since we consider a conservative system the particle can not disappear so the
"total-probability" should be alway one. We demonstrate it computing the time derivative:

dψ ∗
Z Z  
dP(t) 3 d 2 3 ∗ dψ
= dx |ψ(~x, t)| = dx ψ + ψ the state is a
dt R3 dt R3 dt dt complex function
Z  ∗   
ψ  ψ  ~2
= dx 3 ~ 2
−∇ + 2M V ψ − −∇ + 2M V ψ ∗ for a CNR particle
R3 i2M i2M

To develope the multiplication be careful that we deal with operators so commutations are not
obvious, however V is a symmetric operator (it has real eigenvalues) so potential terms simplify:
Z
1 h ∗~ 2 ~ 2ψ∗
i
= dx3 −ψ ∇ ψ + ψ ∇
3 i2M
ZR
i ~ h ∗~ ~ ∗
i
= d3 x ∇ ψ ∇ψ − ψ ∇ψ collect −∇
R3 2M

If we define the integrand as the three dimensional current:

~j(~x, t) ≡ − i ψ ∗ (∇ψ)
h i
~ ~ ∗ )ψ
− (∇ψ
2M
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 6

minus sign is a convention. we can arrange the time-derivative of the probability as


Z I
dP(t)
=− 3 ~ ~
d x ∇ · j(~x, t) = − dσ ~j · ~nσ = 0 (1.2) divergence theorem
dt R3 ∂R3

where ~nσ is a versor orthogonal to the surface of integration, however the result is zero since ~j
computed at infinite borders of R3 is zero, it holds from ψ ∈ `2 . So, probability is conserved in
time and since we normalidet it to 1 at a certain time, it keeps being equal to one.
If we recall the definition of the probability, we can arrange eq (1.2) as

d|ψ(~x, t)|2 ~ ~
Z  
3 d ~
dx |ψ(~x, t)| + ∇ · ~j = 0 =⇒
2
+ ∇ · J = 0,
R3 dt dt
It is the conservation law of probability, called continuity equation, In more general, if a theory
has a continuity eq then there is an associated quantity conserved in time, in this time it is the
probability.

1.1.2 Free Shrödinger equation


We are able to solve the free Schrödinger eq, namely without potential:

d ~2

i ψ(~x, t) = − ψ(~x, t) ≡ Hψ(~x, t)
dt 2M
We will solve it using separation of variables: we suppose that exist a solution ψ(~x, t) ≡ χ(t)ϕ(~x).
In this way we end up with an separable variable ode

dχ(t) χ(t) ~ 2 i dχ(t) ~ 2 ϕ(x)



i ϕ(~x) = − ∇ ϕ(~x) =⇒ =−
dt 2M χ(t) dt ϕ(x)2M
Since there is an equality between two hand-sides depending on different variables, the only
possibility is that each hand-side is constant. Let us call the constant value as E. So
i dχ(t)
= E =⇒ χ(t) = χ(0) exp(−iEt)
χ(t) dt
∇~ 2 ϕ(x)  
− = E =⇒ ϕ(~x) = ϕ(0) exp i~k · ~x
ϕ(x)2M

where we introduced |~k|2 = 2M Ek , it is called dispersion relation. Let us define the energy
related to the wave vector ~k:
|~k|2
ωk ≡
2M
So our solution of the Schrodinger eq is of the form
h  i
ψ(~x, t) = χ(t)ϕ(~x) ∝ exp −i ωk t − ~k · x

k|2
|~
ωk = 2M

it is monochromatic solution, namely a monoenergetic wave. A general solution of the free


Schrödinger eq is the superposition of different waves:
Z
1 3
h  i
~k) exp −i ωk t − ~k · ~x
ψ(~x, t) = dk ϕ̃( k|2
|~
(2π)3/2 ωk = 2M

where we wrote ϕ(~x) as the Fourier transformation of ϕ̃(~k) that is the state written in the
momentum space: Z
1 3
h
~k) exp −i~k · ~x
i
ϕ(~x) = dk ϕ̃(
(2π)3/2
Do note that 1/(2π)3/2 and exp{−i} are conventions. Compiled
25/4/21
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 7

1.2 Review of Special Relativity


SR is the theory invariant under the Lorentz transformation. The Lorentz group in a mathematical 2/3/21
way can be defined as the isometry group of the Minkowski space, namely the group that preserve
distances. A Lorentz transformation is

x → x0 = Λx x0µ = Λµ
νx
ν

where Λ ∈ SO(1, 3) with 1

η = ΛT ηΛ Λµν Λρσ ηµρ = ηνσ


The relation between η and Λ comes out by the constraint to be an isometric group into the
Minkowski space, namely the scalar product is conserved and what follows is
µ µ ν T
x0 · y 0 = x0 y 0 ν = x0 y 0 ηµν = x0 ηy 0 = xT ΛT ηΛy = xT ηy = x · y =⇒ η = ΛT ηΛ

The introduction of Lorentz transformation instead of Galileian one comes from the idea that
they are the real symmetry of the Universe, namely physical laws are the same between frames
connected by boosts. This idea is represented by the special relativity principle:

the laws of physics must be the same in form in any inertial reference frame and they
are connected by Lorentz transformations.

This principle must be respected if we want to write a theory in SR.2 Mathematical results
show that the principle is automatically satisfied if a theory is written by tensors since they are
covariant objects under a change of coordinates between inertial frames.

1.3 Klein-Gordon equation


We are going to see that the "canonical" QM is not a covariant theory and so we have to modify
it if we want a RQM.
Consider the Schrödinger eq:

∂ ~2

i ψ(~x, t) = − ψ(~x, t)
∂t 2M
if it was possible to write it using the covarinat derivative ∂µ = (∂0 , ∇)
~ it would be a relativistic
theory, but it is not possible and QM does not satisfy the special relativistic principle. Compiled
What is the wright eq? The first example as relativistic wave eq is the Klein-Gordon equation 25/4/21
for a free particle. Essentially it derives from same argument using to obtain the Schrödinger eq
but using the relativistic energy. It is the zeroth component of the 4momentum of a relativistic
particle pµ = (E, p~). Explicitly

pµ pµ = M 2 uµ uµ = M 2 = E 2 − |~
p|2 =⇒ E 2 = M 2 + |~
p|2

By using the corrispondence principle (1.1) we can write the differential eq becribes the a
qunatum-relativistic free particle
 2 
∂ ~ 2 2
− ∇ + M ϕ(~x, t) = 0
∂t2
∂2
here arises nice covariant object  ≡ −∇
∂t2
~ 2 and the eq reads

 + M 2 ϕ(x) = 0
 
use the 4-dim
notation x = (~
x, t)
1
Note 3-dim rotation group is SO(3). The term (1, 3) indicates that there is one term with a sign and 3 with
the opposite sign.
2
In classical mechanic there is the simplified formulation: the Galilean principle.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 8

It is the free KG eq. It is covariant since the mass is a scalar quantity3 (from pµ pµ ≡ M 2 ), the
derivative  ≡ ∂ µ ∂µ is a scalar too, and what about the wavefunction? We havo to see how it
transforms under a Lorentz transformation.
If we assume ϕ(x) is a scalar function, then by definition

ϕ(x) → ϕ0 (x0 ) = ϕ(x)

And we are sure that the free KG eq for a scalar field is Lorentz invariant and so satisfies the
special relativity principle.
2
( + M 2 )ϕ(x) = 0 → (0 + M 0 )ϕ0 (x0 ) = ( + M 2 )ϕ(x) = 0

1.3.1 Physical meaning


We saw the eq is mathematical consistency, but what is its physical meaning? Like in QM, we
expect that the probability density is equal to |ϕ|. It is possible if KG eq has a continuity eq
for a conserved quantity that could be interpreted as a probability. Recall a continuity eq is
∂µ J µ (x) = 0 where J µ (x) = (ρ, ~j)(x) and the conserved qunatity Q(t) is the space integral of
the zeroth component since
Z Z Z
dQ(t) ~ · ~j(~x, t) = 0
Q(t) ≡ d3 x ρ(~x, t) =⇒ = d3 x∂0 ρ(~x, t) = − d3 x∇ using the
R3 dt divergence th

if it exists and Q is a probability we are done.


In order to find the continuity eq for KG eq, let us do the same we did for Schrödinger eq:
multiply KG at left by ϕ∗ , multiply the conjugate of KG at left by ϕ

ϕ∗ (x)( + M 2 )ϕ(x) = 0
ϕ(x)( + M 2 )ϕ∗ (x) = 0

take the difference:

0 =ϕ∗ (x)(ϕ(x)) − ϕ(x)(ϕ∗ (x)) mass terms

∗ ∗ ∗ ∗ simplify since M 2
=∂0 [ϕ (∂0 ϕ) − (∂0 ϕ )ϕ] + (∂0 ϕ )(∂0 ϕ) − (∂0 ϕ )(∂0 ϕ) is a scalar

~ ∗ (∇ϕ)
− ∇[ϕ ~ ~ ∗ )ϕ] + (∇ϕ
− (∇ϕ ~ ∗ )(∇ϕ)
~ ~ ∗ )(∇ϕ)
− (∇ϕ ~
=∂0 [ϕ∗ (∂0 ϕ) − (∂0 ϕ∗ )ϕ] − ∇[ϕ
~ ∗ (∇ϕ)
~ ~ ∗ )ϕ]
− (∇ϕ

if we define the following quantities


i i ↔
ρ(x) ≡ [ϕ∗ (∂0 ϕ) − (∂0 ϕ∗ )ϕ] = ϕ∗ ∂0 ϕ
2 2

~j(x) ≡ − i ϕ∗ (∇ϕ) ~ ∗ )ϕ = − i ϕ∗ ∇ϕ
h i
~ − (∇ϕ
2 2
we can rewrite the eq above using ρ and the 3dim current ~j:
 
~ ~ µ µ ρ(x)
∂0 ρ + ∇ · j = ∂µ J = 0 J (x) = ~
j(x)

it is the Klein-Gordon continuity eq. Till here things are good, the problem now is that the
charge Q is not positive:
Z
i
Q(t) = d3 x[ϕ∗ (∂0 ϕ) − (∂0 ϕ∗ )ϕ] ≶ 0
2
namely it can be positive ore negative. So it can not be associated to a probability. This is one
of the point where the KG approach fails.
3
Do note a scalar quantity is more than covariant, it is invariant.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 9

1.3.2 Solution
For the moment forget about the failure and let us solve the free KG eq. In order to do it we
will use the Fourier transformation tool. Recall a general scalar function can be written in terms
of 4dim Fourier transformation as
Z
1
ϕ(x) = d4 k exp(−ikx)ϕ̃(k)
(2π)2
Notice
 theFourier transformation ϕ̃(k) is a scalar function too since the monochromatic wave
~
exp ik · ~x and the measure d k are Lorentz scalars.4
R 4

Just for completness let us write the inverse Fourier transformation


Z
1
ϕ̃(k) = d4 x exp(ikx)ϕ(x)
(2π)2
To pass from the Fourier and the antiFourier transformation we need the 4dim Dirac δ, we define
it in the momentum and coordinate space:
Z
4 1
δ (k − p) ≡ d4 x exp(−i(k − p)x)
(2π)4
Z
4 1
δ (x − y) ≡ d4 k exp(−i(x − y)k)
(2π)4
Now, let us apply the Fourier transformation to free KG eq. We will impose that ϕ is a
solution of KG eq and use the Fourier-formula

d4 k
Z
2
exp(−ikx) −k 2 + M 2 ϕ̃(k) = 0 =⇒ (−k 2 + M 2 )ϕ̃(k) = 0
 
( + M )ϕ(x) = 2
(2π)
A general solution of the latter eq is a scalar function ϕ̃(k) such that can be non vanishing only
at k 2 = M 2 , namely something like

δ(k 2 − M 2 )f˜(k) ≡ ϕ̃(k)

Recall the property of the Dirac delta: if xi are the zeroth of f , namely f (xi ) = 0, then
N
X δ(x − xi )
δ(f (x)) =
|f 0 (xi )|
i=1

So, if we denote ωk = (M 2 + |~k|2 )1/2 , we can write


δ(k0 − ωk ) + δ(k0 + ωk )
δ(k 2 − M 2 ) = δ(k02 − ωk ) =
2ωk
Do note the relation with ωk is common in physics and it is called the dispersion equation.
We can rewrite the general solution of KG eq 8/3/21
Z 4
1 d k
ϕ(x) = (δ(k0 − ωk ) + δ(k0 + ωk )) exp(−ikx)f˜(k)
(2π)2 2ωk
Z 3
1 d k  h
~k · ~x exp(−iωk x0 )f˜(ωk , ~k) + exp(+iωk x0 )f˜(−ωk , ~k)
i
= exp i integrate the delta
(2π)2 2ωk k0 =ωk over dk0
Z 3 h
1 d k i
= 2
exp(−ikx)f˜(−k) + exp(ikx)f˜(k)
(2π) 2ωk
4
About the Lorentz invariance of the meause:

d4 k → d4 k0 = | det Λ|d4 k = d4 k

since det Λ = ±1 is a property of the Lorentz group.


CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 10

notice we are using 4dim notation even if we have only three indetpendet variables. Let us define
the following quantities:

f˜(k) f˜(−k)
a(k) ≡ √ √ b∗ (k) ≡ √ √
2π 2ωk 2π 2ωk

where b∗ is just a convention, don’t focus on the conjunction. Now, the general KG solution reads

d3 k
Z
1
ϕ(x) = √ [a(k) exp(−ikx) + b∗ (k) exp(ikx)]k0 =ωk ≡ ϕ+ (x) + ϕ− (x)
(2π)3/2 2ωk

where ϕ+ (x) is the positive energy solution, and ϕ− (x) is the negative one.5 In the Schrödinger
eq we have only one wave function associated to a positive energy. In the KG the solution is sum
of two waves: there is a double dof that comes from the quadratic relativistic energy so positive
and negative energy solution arise. This is the main difference between Schrödinger and KG eq.
Recall the solution is covariant scalar functioneven if it does not look like a covariant object.

Real solution
In general a, b are complex functions. However the differential KG eq is a real eq and so it admits
real solutions. So let us focus on the real part of the general complex solution ϕ. To find real
solution we have to impose
ϕ(x) = ϕ(x)∗ =⇒ b(k) = a(k)
that is the only solution to the constraint.

Comment (About complex solutions in physics). Complex sol has double dof wrt a real sol. In
our case we are dealing with a scalar field and the complex solution ϕ has two dof: a, b. Instead
the real sol has 1 dof.

Exercise. Fix some results:

1. For a real KG solution show


d3 x
Z
1
a(k) = √ (ωk ϕ(x) + i∂0 ϕ(x)) exp(ikx)

(2π)3/2 2ωk k0 =ωk

it is the antiFourier transformation.

2. For a complex KG solution show that

d3 x
Z
1
a(k) = √ (ωk ϕ(x) + i∂0 ϕ(x)) exp(ikx)

(2π)3/2 2ωk k0 =ωk

d3 x
Z
1
b∗ (k) = √ (ωk ϕ(x) + i∂0 ϕ(x)) exp(ikx)

(2π)3/2 2ωk k0 =ωk

3. (connection between real and complex solutions) given 2 real solutions of the KG eq ϕ1,2
one can always unit a complex solution

ϕ1 (x) + iϕ2 (x)


ϕ(x) = √
2

This was everything about the introduction of free KG eq.


5
Check just appling the energy operator, eg for ϕ+ (x):

i∂0 (exp(−ikx)a(k)) = ωk (exp(−ikx)a(k)) =⇒ Ek = ωk


CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 11

1.3.3 KG equation coupled with an EM field


Now, to understand what are the two dof, the best way is to coulple the KG with a classical
external EM field (the only quantistic object is the KG eq). The field is 4dim vector potential
which contains electric and magnetic fields:
~ = −∇A~ 0 − ∂0 A~
 
A0 E
µ
A (x) = ~ (x) such that ~ ~ ~
A B =∇×A

where A0 = V is the electric potential. How can we couple the KG eq with this external field? We
will use the minimum coupling, it is a standard way to couple a quantistic particle with an EM
field. "Minimal" means it is the simplest way to couple. The principle requires the introduction
of covariant derivative6
Dµ ≡ ∂µ + iqAµ
where q is the charge of our particle, and the substitution of partial derivatives in KG eq with
covariant ones
∂µ → Dµ
So the coupled KG eq with external EM field is

(Dµ Dµ + M 2 )ϕ(x) = 0

Do note this prescription can be generalized for any quantum particle.


Now, let us study more in detail the KG coupled eq:

(Dµ Dµ + M 2 )ϕ(x) = 0
 µ
(∂ + iqAµ )(∂µ + iqAµ ) + M 2 ϕ(x) = 0


 + 2iqAµ (x)∂µ + iq(∂µ Aµ (x)) − q 2 A2 (x) + M 2 ϕ(x) = 0


 

Do note the term iq(∂µ Aµ (x)) many time we lost in calcoulus. We must remember the order
of application: anything on the right acts on anything on the left, so it is ∂ µ [iqAµ ϕ(x)] =
(iq∂ µ Aµ + iqAµ ∂ µ )ϕ(x).
Comment (Minimal coupling in nonrelativistic QM). This is exactly what we did with the
Schrödinger eq: minimal coupling corresponds to substitute
   
ωp ωp − qA0
pµ = → pµ − qAµ = ~
p~ p~ − q A

[this is just the Fourier transform of the covariant derivative, in the momentum space.] into
Schrödinger eq to have the coupled with an EM external field.
" #
∂ψ P2 ∂ψ (~ ~ 2)
p2 − q A
i = ψ =⇒ i = + qA0 ψ
∂t 2M ∂t 2M

Lorentz gauge The 4-potential is not physical: if we change it by Aµ → A0µ = Aµ + ∂µ Λ


for some scalar function Λ it does not change the resulting measurable electric and magnetic
fields. Therefore, while still retaining full generality in our description of physical systems we
can impose certain conditions, such as ∂µ Aµ = 0. This condition, known as the Coulomb gauge,
does not actually fix all of the gauge freedom, that is, even by imposing this we still do not have
a one-to-one correspondence between the physical fields and the 4-potentials: after imposing it,
we can still perform gauge transformations where the function Λ is harmonic, that is, it satisfies
Λ = 0. This condition is convenient since it allows us to get rid of a term in the KG equation,
so we impose it. Compiled
6
26/4/21
This is different to covariant derivative of GR since we are in a Minkowski space and not in a curved space.
Now it is related to gauge group of QED.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 12

1.3.4 Prediction of antimatter


A monochromoatic wave solution with 4momentum ∓k µ is

ϕmw
± (x) ∼ exp(∓ikx)

In order to better understand what is the physical meaning of the ± sign, let us spell out thec
oupled KG eq for monochromatic waves.

(Dµ Dµ + M 2 )ϕmw
± (x) = 0

The square covariant derivative can be written as Dµ Dµ = (D0 )2 − (D) ~ 2 . Inside each term there
is a partial derivative which brings down the momentum. If we compute separately the plus and
minus solution we have
h i
−Dµ Dµ ϕmw (x) = (ωk − qA 0 )2
− (~k − q A)
~ 2 ϕmw = M 2 ϕmw
+ + +

−Dµ Dµ ϕmw 2 ~ ~ 2 mw 2 mw
− (x) = [(ωk + qA0 ) − (k + q A) ]ϕ− = M ϕ−

Notice the negative solution ϕmw − is coupled into KG eq with a charge −q, while the positive
one ϕmw+ is coupled with a charge +q. However the two equations can be transformed into
each other by swapping the charge, q → −q. So we can say the coupled KG eq describes two
dof that are a particle with charge q and a particle with the same mass but charge −q, called
antiparticle. So relativistic theory, like the KG one, predicts automaticaly the existence of particle
and antiparticle.
Do note we are able to see the prediction of antimatter only for particles coupled with EM
field since the charge comes to play. The mathematical reason is that to couple the free KG eq
we upgrade the partial derivatives to covariant ones making a jump from real eq to complex eq;
so solutions become complex too, doubling the dof.
The solution of free KG eq is real and describes neutral particles in the sense there is no
discrimination for charges.

1.3.5 The non-relativistic limit


The relativistic KG eq must be compatible with the non-relativistic Schrödinger when the velocity
goes down. Let us investigate it for KG coupled EM.
In the non-relativistic limit the mass is the dominant term of energy:7
|p|2
 4
p
2 2
|~
p|
E = M + |~ p| ' M + +O |p|  M
2M M4
We can factorize the wavefunction expliciting the energy term of the complex exponent: it is the
zeroth component of the momentum:

ϕ(~x, t) ≡ exp(−iM t)ϕ0 (~x, t)

Let us evaluate the coupled KG eq for it:


h i
~ − iq A)
(Dµ Dµ + M 2 )ϕ = (∂0 + iqA0 )2 − (∇ ~ 2 + M2 ϕ = 0 D µ Dµ = D0
2 ~2
−D

now we can move the momentum terms to the right and keep the mass and energy on the left, be
carefull with the time derivative since it applies to the product exp(−itM )ϕ0 (~x, t) and so there
are mixed terms:
h i
(∂0 + iqA0 )2 + M 2 exp(−iM t)ϕ0 = exp(−iM t) (∇ ~ 2 ϕ0
~ − iq A)
 
h i
exp(−iM t)[∂02 − M 2 − 2iM ∂0 + 2iqA0 ∂0 + 2iq(∂0 A0 ) + 2qM A0 − q 2 A20 + M 2 ]ϕ0 = exp(−iM t) (∇ ~ 2 ϕ0
~ − iq A)

7
Photons are relativistic particles since they are massles, so their mometum can never be smaller than the mass.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 13

Till now we did not do any approximation, now is time to impose the non-relativistic limit:
that all sources of energy in my sistem are smaller than the mass, namely
∂ 0 ϕ0
∂ A
'EM, 0 0

ϕ0 |qA 0 |, |M
A0

So, some terms in the KG eq are negligible , and the first order terms are
 
1 ~
i∂0 ϕ0 (x, t) = − ~ 2 + aA0 ϕ0 (~x, t)
(∇ − iq A)
2M
It is the Schrödinger eq. So, our relativistic theory is perfectly compatible with the non-relativistic
theory described by the Schrodinger eq.
When we start to see positrons in our expreriments signify that we are not more in the
non-relativistic regime: energy in our machine is at least double the mass of an electron. To
describe it we need to build a relativistic theory instead of Schrödinger eq.
9/3/21
Exercise. Derive the continuity eq for the EM coupled KG eq.
The solusion is obtained following the procedure done for the free KG eq. We will find
  

ρEM = 2 ϕ ∂ 0 ϕ − qA0 ϕ∗ ϕ
1 ∗


~ ~
∂ρEM + ∇ · jEM = 0  

~jEM = − ϕ ∇ϕ − q Aϕ

 1 ∗ ~ ∗ϕ
2

1.3.6 Klein paradox


KG eq suffers some problem: it does do not have conserved quantity interpretable as probability;
it contains two pieces interpreted as particle and antiparticle (we understood it coupling it with
EM).
These differences lead to an inconsistency called the Klein paradox. The inconsistency
arises thrught the following theoretical experiment: consider a relativistic particle in a EM
step-potential, how does the particle behave?
In classical mechanic, if the particle enregy is lower than the barrer it cannot escape. In
quantum mechanic if its energy is low than the potential-step there is however a small probability
of the particle to pass the barrer. What about relativistic QM? Consider 1dim lab frame with an
EM potential
(
x < 0 region (1)
 
V (x) 0
Aµ (x) ≡ ~0 where V (x) =
V0 x > 0 region (2)
Consider a monochromatic wave describing the relativistic particle. When it interacts with the
step, we expect that a part of the wave is reflected with wave-vector k and a part transmessed
with k 0 . Let us denote with ϕ1 and ϕ2 the general wavefunctions in region (1) and (2), they are
monocromatic waves:
ϕ1 (x, t) = exp[−iωt]χ1 (x) χ1 (x) = exp[ikx] + r exp[−ikx]
χ2 (x) = t exp ik 0 x
 
ϕ2 (x, t) = exp[−iωt]χ2 (x)
where the spatial part χi (x) must contain the incoming and reflected wave in (1) and the
transmitted wave in (2), they are weighted by coefficients r, t.
Impose the condition that in region (1) the wave satisfies the free KG equation, while in
region (2) the wave satisfies the coupled KG eq. What we obtain are dispersion relations:
region (1) ( + M 2 )ϕ1 (x, t) = 0 =⇒ ω 2 − k 2 − M 2 = 0
1/2
=⇒ k = ω 2 − M 2


2
region (2) (Dµ Dµ + M 2 )ϕ2 (x, t) = 0 =⇒ (ω − V0 )2 − k 0 − M 2 = 0
1/2
=⇒ k 0 = (ω − V0 )2 − M 2

CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 14

Now we have to find the real solution for the total space (1)+(2): it is a solution continuous at
x = 0 and with the first derivative continuous too. Let us impose them:

ϕ1 (0, t) = exp[−iωt](1 + r) = exp[−iωt]t = ϕ2 (0, t)


∂x ϕ1 (0, t) = exp[−iωt](1 − r)ik = exp[−iωt]tik 0 = ∂x ϕ2 (0, t)

We obtain condition on r and t:


0
( (
1+r =t r = k−k
k+k0
0 =⇒
1 − r = kk t 2k
t = k+k 0

Up to this point it is exacly like non-relativistic QM. Now, write down continuity eq for region
(1) and (2) (recall we already saw continuity eq for free and coupled KG eq).
  

i ∗
ρ1 = 2 ϕ ∂ 0 ϕ


region (1) ~ · ~j1 = 0 =⇒
∂0 ρ1 + ∇ 


~ i
j1 = − 2 ϕ ∇ϕ

 ∗

  

ρ2 = 2 ϕ ∂ 0 ϕ − V0 ϕ∗ ϕ
i ∗


region (2) ~ · ~j2 = 0 =⇒
∂0 ρ2 + ∇ 



 ~ i ∗
j2 = − 2 ϕ ∇ϕ −  ~
q Aϕ∗
ϕ ~ = 0 in our
A
experiment

Now, we plug inside the density and current the expression of the wavefunctions ϕ1,2 , we find
the following expressions:
(
ρ1 = ω|χ1 |2
region (1)
j1 = k(1 − |r|2 )
(
ρ2 = (ω − V0 )|t|2 exp i(k 0 − k 0 ∗ )x
 
region (2) 0 0∗
2 exp i(k 0 − k 0 ∗ )x
j2 = k +k
 
2 |t|

Notice, k 0 can be complex, since it depends on energy: if the particle energy is bigger than the
step k 0 is real, overwise it is imaginary. We do not use vector notation for the current density,
however any k and j is a vector.
Let us define the reflection and trasmission coefficient as done in QM:
jin − j1 j2
R= T=
jin jin
where jin is the initial current, namely the current associated to a wave propagating into region
(1) without reflection. Such a wave is

ϕin (x, t) = exp[−iωt]χin (x) = exp[−iωt] exp[ikx]

and the relative density current comes from the definition:


 
i ↔ i i
jin ≡ − ϕ ∇ϕ = − (ϕ∗ ikϕ − c.c.) = − (2ik) = k
∗ ↔
ϕ∗ ∇ϕ =
2 2 2 ϕ ∇ϕ − (∇ϕ∗ )ϕ

c.c states for complex conjugate. What changes with respect to QM is the consequence to have
quadratic particle energy: there are three different situations depending on the value of the
energy of the particle ω.
1. ω ≥ V0 + M : let us called it "small V0 ", it is very similar of the non-relativistic case,

2. V0 + M > ω > V0 − M : called "intermediate V0 ";

3. ω ≤ V0 − M : called "large V0 ".


CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 15

First scenario ω > V0 + M . This scenario imposes k, k 0 reals, in fact from dispersion relations:

ω2 − M 2 = k2 > 0
2
(ω − V0 )2 − M 2 = k 0 > 0

Since k and k 0 are real, there is a transmitted and reflected wave. Now let us calculate densities
and currents in region (1) and (2):

ρ1 = ω|χ1 |2 > 0 ρ2 = (ω − k)|t|2 > 0 in region (2) the


conjugate k0 ∗ = k0
j1 = k(1 − |r|2 ) j2 = k 0 |t|2 since it is real

From densities we can compute the reflection and transition coefficients:

k − k0 2 k0 2 4kk 0
 
R = |r|2 = < 1 T = |t| = <1
k + k0 k (k + k 0 )2
and R + T = 1, so these coefficients are interpretable as probability like we did in . In case 1
everything is as expeced in non-relativistic QM. In case 1 everything is as expected in NRQM.

Second scenario V0 + M > ω > V0 − M . Now we are not sure that ω is larger than the
potential step and k 0 can be imaginary, in fact from dispersion relations

ω 2 − M 2 = k 2 > 0 =⇒ k is real
2
(ω − V0 )2 − M 2 = k 0 < 0 =⇒ k 0 is imaginary
Let us compute densities and currents:

ρ1 = ω|χ1 |2 > 0 ρ2 = (ω − V0 )|t|2 exp −2|k 0 |x ≶ 0




j1 = k(1 − |r|2 ) j2 = 0 k0 + k0 ∗ = 0 since


k0 is only
imaginary
but now we are not sure about the sign of (ω − V0 ) and this is a problem for the probabilistic
interpretation of ρ. In addiction j2 = 0 implies no transmission. Compute the probability of
transmission and reflection:
R = |r|2 = 1 T=0 since k is real and
k0 imaginary, |r| is
So, in this it seems the particle is certainly reflected but we missed the probabilistic interpretation as follow:
|α−iβ|
|n| = |α+iβ| =
of ρ1,2 since the latter is not positive. Consistency problems start. α2 +β 2
=1
α2 +β 2

Third scenario ω ≤ V0 − M . Here the particle energy is always smaller that the large step.
from dispersion relations

ω 2 − M 2 = k 2 > 0 =⇒ k is real
2
(ω − V0 )2 − M 2 = k 0 > 0 =⇒ k 0 is real

We exprect that we have a transmitted wave in region (2). Compute the densitied and currents:

ρ1 = ω|χ1 |2 > 0 ρ2 = (ω − V0 )|t|2 < 0


j1 = k(1 − |r|2 ) j2 = k 0 |t|2

Since ρ2 < 0 the probability interpretation is not valid, even more strange is the fact that we
have a current propagates in region (2) since k 0 is real.
The saddle point is that k 0 2 is positive, but we do not know which sign k 0 is, in order to find
it we can look to the group velocity vg of a monochromatic wave (it is connected with the wave
number):
∂ω ∂ h 2 0 2 1/2
i k0
vg ≡ = V 0 + (M + k ) =
∂k 0 ∂k 0 ω − V0
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 16

If we consider a packet propagates forward we need vg > 0, so relative to the sign of (ω − V0 ) we


descover the sign of k 0 .
Do note the possibility to have a negative wavenumber k 0 is a big problem. If is is negative,
the reflected coefficient is
k − k0 k + |k 0 |
r= = >1
k + k0 k − |k 0 |
and reflection and transmission probability are
−4k|k 0 |
R = |r|2 > 1 T= <0
(k − |k 0 |)2
They are something not interpretable as probability. Fully consistency problem! However
R + T = 1.

This is the Klein paradox. At the end this is the main arguments that make people change
way from the direct extension of NR-wavefunction to R-wavefunction.
How to interpretate this result? Recall the NR limit V0  M , so since scenario 3 has V0 > M ,
it is truly relativistic (while scenario 1 is the NR approximation).
How to interpretate the physics? In R case, energy is equal to mass. When the particle comes
near the potential, it interacts with V0 and from the energy of the potential a couple of electron
and positron is generated (transforming energy into masses): electron travels to the left while
positron travels to the right because of the potential. So the interaction with the EM field, in
relativistic scenario, generates an electron-positron pair where all the electrons generated are
sent to left, all positrons are sent to the right. Now, we understand that the quantity conserved
in the continuity eq is the charge!
In RQM problems arise since the wavefunction approach is not appropriate to describe
systems where the number of particles changes. Now, we have to find a new paradigm, it will
be the quantum field theory (QFT) where the number of particles allows to change. Before
introduce it, we will see another idea to describe RQM from Dirac. We will see that Dirac failed
too, but he found a new particle: fermion.

1.4 Dirac wave equation


It is an important part of the course. We will retake Dirac formalism in QFT. Recall the first 15/3/21
attempt to build a RQM theory is the KG wave eq, but it suffers two main problems:
1. it is a second order eq in time and space derivatives, since it arises from the relativistic
energy:
1/2
E 2 = M 2 + |~
p|2 =⇒ E = ±ωp = ± M 2 + |~ p|2
so, it seems there are positive and negative energy solution. It is a problem since from
Schrödinger eq we got used to have just a positive soltion.
2. there is a continuity eq, but the conserved quantity is not positive defined:
Z
Q ≡ d3 x ρ(~x, t) ≶ 0

since ρ ≶ 0, so the charge Q has not a probabilistic interpretation.


These problems claim the Klein paradox: it tells us that the KG eq is unconsistency to describe
a physical particle.
They are the starting point of Dirac work. He wanted to derive a consistent relativistic
quantum wave eq. Dirac thought the main problem of KG eq was that it was a second derivative
eq (in fact Scrhödinger is first order and it works). Dirac initially used a generic linear differentail
eq
∂  
~ + βM ψ(~x, t) ≡ HD ψ(~x, t)
i ψ(~x, t) = −i~ α·∇
∂t
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 17

where α~ is a parameter vector and β is a constant parameter (in space and time).
Dirac ansatz on the enrgy of the relativistic paricle is based on the correspondence principle:
it could be
E=α ~ · p~ + βM
To determine the parameters we need to impose two conditions:
1. the Dirac hamiltonian HD should be an observable: namely it should be Hermitian;

2. the Dirac linear eq should be compatible with the KG eq since it want to be a relativistic
eq: namely the Dirac eq should be a "case" of the KG eq.
Let us impose these constraint. Starting from the first: imposing the hermiticity of the Dirac
operator8
hψ|HD ψi = hHD ψ|ψi
and we obtain that α
~ and β have to satisfy

αi = αi∗ β = β∗

Exercise. Prove it as an exercice.


Now the second: requiring the consistency with the KG eq means if we derive it in time the
Dirac eq we want to recover the KG one

∂2 
~ + βM ∂ψ

ψ(~
x , t) = −i −i~α · ∇ derivatives
∂t2 ∂t commute and βM

= −(−i~α·∇ ~ + βM )(−i~ α·∇ ~ + βM )ψ(~x, t) is a Lorentz scalar

= αi αj ∂i ∂j + i(αi β + βαi )M ∂i − β 2 M 2 ψ(~x, t)


 

Notice we leave parameters αi β in general form with no use any commutation property. When
we impose the derivative-Dirac eq equal to the KG eq it comes out commutation properties for
parameters:
h i
αi αj ∂i ∂j + i(αi β + βαi )M ∂i − β 2 M 2 ψ(~x, t) = ∇~ 2 − M 2 ψ(~x, t)
 

implies
1
{αi , αj } = δij
2
{αi , β} = 0
β2 = I

The first condition could seem peculiar but it is based on the identity
1 1
αi αj = [αi , αj ] + {αi , αj }
2 2
So consistency between Dirac and KG eq implies that α and β cannot be real numers because
they must have a non trivial anticommutation, while real number commute. The simplest non
trivial anticommutation objects are matricies: αi and β are n × n matricies.
This is the starting point to understand the Dirac eq: since αi , β are matricies, ψ(~x, t) cannot
be a scalar function but it must be a ndim vector. Matricies act in a vector space of dimension
n called spinorial space where vectors ψ, called spinor, live.
In conclusion if we want to write relativistic linear differential wave eq we must assume
the wave function describes a vector which lives in a new space. The dimension n has to be
determined.
8 †
Recall that it is not just the requirement HD = HD since it is a differential operator, so the requirement is
hψ|HD ψi = hHD ψ|ψi.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 18

1.4.1 The spinorial space


The next step is to try to understand the existence and dimensionality of matricies αi , β. Recall
the constraints introducted:

(a) αi† = αi β† = β imposing hermiticy

(b) αi2 = In 2
β = In
1
(c) {αi , αj } = δij {αi , β} = 0
2
(do note αi , β with i = 1, 2, 3 form a group of 4 n × n matricies act in the spinorial space, the
ndim vector space of ψ) notice condition (b) derives from (c) when indicies are equal. If we
write αi and β in diagonal form, which we can do since they are Hermitian, we get their real
eigenvalues on the diagonal. If we square them we get the identity, therefore they must all be ±1.
From condition (b) and (c), we obtain that all these matricies are traceless: Tr[αi ] = 0 = Tr[β].
Since the trace is the sum of the eigenvalues, which are ±1, the dimension n must be even.

Exercise. Exercice: prove Tr[αi ] = 0 = Tr[β].

Tr[αi ] = Tr αi β 2 = Tr[βαi β] = − Tr αi β 2 = − Tr[αi ]


   

2dim spinorial space


Let us better understand which is the dimension of the spinorial space. The lowest posibility is
n = 2, it this space we have to find 4 hermitian and traceless anticommutig 2 × 2 matrices which
are αi and β.
Recall any hermitian 2 × 2 matrix can be written as a combination of the identity and the
Pauli matricies (for generators {Ii , σi }), namely

H2×2 = c0 I + ci σi

however we want traceless matrixies and so we must impose c0 = 0. This means a 2dim spinorial
space has only 3 generators. The dof that we lost is the parameter β and it can happen pysicaly
if we are considering massless M = 0 spinors. So in n = 2 we cannot describe massive Dirac eq.
Notice the Pauli matrices respect all the conditions.
1
{σi , σj } = δij σi2 = I, σi† = σi Tr[σi ] = 0
2

4dim spinorial space


Let us pass to the next lowest possibility: n = 4. It can be shown that the following 4 matricies
satisfy all the requirement conditions
   
0 σi I 0
αi = β=
σi 0 0 −I

where each block is a 2 × 2 matrix. Since we preserve the β dof (Ppysicaly the mass dof), we can
say the lowest spinorial dimension in which we are able to describe massive Dirac equation is
n = 4.

Exercise. Prove that αi , β satisfy all properties.

Compiled
28/4/21
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 19

1.4.2 Representation of gamma matricies


Now, let us use a modern notation for our results: it is conventional to define the following 4 × 4
matricies
0 σi
      
0 I 0 i I 0 0 σi
γ ≡β= γ ≡ βαi = =
0 −I 0 −I σi 0 −σi 0

They are the γ-matricies in the Dirac/Pauli representation. In a compact way, we put all the
matricies in the object:
γ µ = (γ 0 , γ i )
This is not a Lorents vector, the index µ runs from 0 to 3 but it does not represents the
components of a Lorentz vectror.

Properties From the properties of αi , β, the γ matricies have the following properties:

1. γ 0 = (γ 0 )† , γ i = −(γ i )†

2. (γ 0 )2 = I4 , (γ i )2 = −I4

3. {γ µ , γ ν } = 2η µν

Dirac equation Now, we can write the Dirac eq in terms of γ µ : recall it

i∂0 ψ = (−iαi ∂i + βM )ψ

We just have to multiply the eq for β

iβ∂0 ψ = (−iβαi ∂i + β 2 M )ψ
iγ 0 ∂0 = (−iγ i ∂i + M )ψ
(iγ µ ∂µ − M )ψ = 0

That is the Dirac eq using the gamma-matricies in vectorial notation. Let us write down
explicitely all indicies in order to well understand what object we have:
µ
(iγαβ ∂µ − M δαβ )ψα = 0

here γ µ is not a Lorentz index, however it is "coupled" with the Lorentz vector ∂µ; while α, β are
spinorial indicies of dimension 4.
In conclusion the Dirac eq is a linear differential eq in time and space in a 4 dimensional
vector space (that is not the Minkowski space, it is the spinorial one), so it is a set of 4 eqs
labeled by β and since gamma-matricies are not diagonal, components of the unknown function
ψα are mixed in the system.
If we use gamma-matricies, we can introduce a further notation: the slashed one. We will
denote any Lorentz vector pµ multiplied by gamma-matrix γ µ as "slashed":
µ 0 i
/ ≡ γ pµ = γ p0 + γ pi
p

In this way, the Dirac eq becomes more compact:

(i∂/ − M )ψ(x) = 0
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 20

1.4.3 Other representations


Since gamma-matrices properties do not depend on the specific representation, so the Dirac/Pauli
it is not the only possible choice. Given a unitary matrix C, we can define a new represenation
of the gamma matricies as
γ̃ µ ≡ C −1 γ µ C
In this course we will use the Dirac/Pauli representation. However let us just write another
diffused representation of gamma matricies: the relativistic/Weyl one
   
0 0 I2 i 0 σi
γWeyl = γWeyl =
I2 0 −σi 0

Exercise. Show how to recover the KG eq from the Dirac eq written with the slashed-notation.
Let us "square" the complex Dirac eq in order to have second differential terms.
 2
∂ + M2 ψ = 
 2
∂ + M2 ψ
 
0 = −(i ∂ − M )ψ = 
∂ + M )(i ∂ + iM ∂ − iM 

opening the operator:


0 = (γ µ γ ν ∂µ ∂ν + M 2 )ψ = ( + M 2 )ψ
where we used:
1 1
γ µ γ ν = [γ µ , γ ν ] + {γ µ , γ ν }
2 2
Notice, while the Dirac eq is not diagonal in the spinorial space, the KG eq is diagonal. In
fact  and the mass have not spinorial indicies so they are just proportional to the idenity of the
spinorial space: ( + M 2 )Iαβ ψα = 0, so each component of the eq solve the ith KG equation:
( + M 2 )ψi = 0.
Why does Dirac impose to recover the KG eq from its eq? Dirac is trying to identify one of
the problem of KG eq: the quadratic form (that is not present in the well-working Schrödinger
eq). But he fails since if we want to deal with relaticistic we must use the square formula of the
Energy.

1.4.4 Covariance of Dirac equation


Let us see if this eq is compatible with SR. Suppose to apply a Lorentz transformation Λ ∈ SO(1, 3) 16/3/21
connecting two inertial reference frame O → O0 , so coordinates and partial derivative transform
as:
µ
x0 = Λ µ ν xν ∂ 0 µ = Λµ ν ∂ν
Recall also Λµ ν Λµ ρ = δν ρ . Now, how does transform the spinorial field ψ(x)?
We know two examples about fields transformation:

1. In the KG eq we introduce the concept of a scalar field that is invariant: ϕ0 (x0 ) = ϕ(x).

2. From Maxwell eqs we saw the trasformation of the electomagnetic potential, that is a
vectror field: A0 µ (x0 ) = Λµ ν Aν (x)

However the spinorial field is a 4dim vector that does not have Lorentz indicies, so how to deal
with this object? We have to impose a new transformation property: we can think the spinoral
field is rotating in the spinorial space when we apply a Lorentz transformation. It is an analogy
with the rotation in Minkowski space done by Lorentz group. The spinorial rotation S must be a
function of Λ, so a Lorentz transformation (of Minkowski coordinates) translates into a spinorial
rotation:
ψ 0 (x0 ) = S(Λ)ψ(x)
where S(Λ) is a 4 × 4 matrix and Λ ∈ SO(1, 3). This is our proposal.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 21

The question now is: which constraint S(Λ) has? It has to make covariant the Dirac eq,
namely it holds in a new frame as well as in the old:
0
(i∂/ − M )ψ 0 (x0 ) = S(Λ)[i∂/ − M ]ψ(x) = 0
where Λ is the Lorentz transformation connecting reference frame O and O0 . Step by step: write
any term on O0 as functions of O terms
0
(i∂/ − M )ψ 0 (x0 ) = (iγ µ Λνµ ∂ν − M )S(Λ)ψ(x)
= [iΛνµ S(Λ)S(Λ)−1 γ µ ∂ν S(Λ) − M S(Λ)]ψ(x)
= S(Λ) iΛνµ S −1 (Λ)γ µ S(Λ)∂ν − M ψ(x)
 

where we used the following proprties:


1. Λ commute with γ since the former is constant in the spinorial space (same for S(Λ));
matricies γ µ and S(Λ) do not commute a priori so we introduced S −1 S = I.
2. S commute with ∂ν since the former is constant with respect to the spatial coordinates
(the Lorentz transformation is fixed); M is a scalar; we take out S(Λ).
Now, if Λνµ S −1 (Λ)γ µ S(Λ) = γ ν then Dirac eq is covariant. This is the requirement that defines
the spinorial tranformation S(Λ): it must respect
S −1 (Λ)γ µ S(Λ) = Λµν γ ν (1.3)
Definition (Spinorial transformation). S(Λ) is the spinorial representation of Lorentz group. It
apply to a spinor ψ(x) as follow (in spinorial components):
ψ 0 α (x0 ) = S(Λ)αβ ψβ (x)
And it must respect the following relation:
S −1 (Λ)γ µ S(Λ) = Λµν γ ν
in order to make covariant Dirac eq.
We have to find an explicit expression of the matrix S(Λ). Tipically one does an infinitesimal
Lorentz transformation in the neightbor of the identity and then write an infinitesimal spinorial
transformation imposing (1.3). The infinitesiaml Lorentz is
Λµν = ηνµ − ωνµ
where ω is an antisymmetric tensor (ωµν = −ωνµ ) with six independet parameter: 3 parameter
are spatial rotations and 3 the boosts. A general inifinetimal spinor looks like
i
S(Λ) = I4 − ωµν Σµν
2
where Σ is the set of the generators of the spinorial representation of the Lorentz group: 6
µν

independet 4 × 4 matrices in the spinorial space (for each element of the Lorentz we are finding
a transformation matrix: indicies µν are contracted in the sense it is a sum of generators Σαβ
multiplied by terms of ωµν , here αβ are spinorial indicies).
Now we plug the infinitesimal form of S(Λ) in (1.3), using the fact that to first order
S −1 = 1 +  is the inverse of S = 1 − :
   
i ρσ µ i λρ
I + ωρσ Σ γ I − ωλρ Σ = γ µ + ωνµ γ ν
2 2
i i
γ µ − γ µ ωλϕ Σλϕ + γ µ ωρσ Σρσ + O(ω 2 ) = γ µ + ωλν η µλ γ ν
2 2
i h i 1 
− ωλρ γ µ , Σλρ = ωλν η µλ γ ν − η µν γ λ
2 h i  2 
γ µ , Σλρ = i η µλ γ ρ − η µρ γ λ
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 22

where we neglected O(ω 2 ) terms since we deal with infinitesimal tranformations, and we anti-
symmetrized η µλ γ ν in λν since it is contracted with the antisymmetric tensor ωλν (namely we
took only the antisymmetric part of the sum since symmetric one would not contribute to the
equation).
We can then prove that it satisfied:
i
Σµν ≡ [γ µ , γ ν ]
4
Exercise. Prove that the above expression of Σµν satisfies the relation γ µ , Σλσ .
 

The finite transformation S(Λ) is given by


 
i
S(Λ) = exp − ωµν Σµν
2

and the inverse  


−1 i
S (Λ) = exp ωµν Σµν
2
Why to exponential S? Because if we expand the S we found we obtain the infinitesimal for of S.
Summarizing: we started requiring the covariant of Dirac eq, so we found a spinorial
tranformation S(Λ) for the Dirac field (it is the spinorial representatin of Lorentz group) such
that, to make Dirac covarian must respect S −1 (Λ)γ µ S(Λ) = Λµν γ ν . From this constraint we
found the S(Λ) = exp − 2 ωµν Σµν where Σµν ≡ 4i [γ µ , γ ν ].
 i 

1.4.5 Dirac conjugate spinor


Consider the hermitian conjugate of the Dirac eq

∂ − M )ψ]† = −ψ † (iγ µ† ∂ µ + M ) = 0
[(i

where the result follows from the fact that the adjoint of a product is the product of the reverse-
ordered adjoints (since there are vectors and matrices we have to consider the conjugate and the

hermitian), while the notation ∂ µ means that the derivative operator is acting on what is on its
left.
Now, we simplify the eq using the following properties:

1. γ 0 γ 0 = I

2. γ 0 γ µ† γ 0 = γ µ

So

0 = −ψ † γ 0 γ 0 (iγ µ† γ 0 γ 0 ∂ µ + M )

= −(ψ † γ 0 )(iγ µ ∂ µ γ 0 + γ 0 M )

= −(ψ † γ 0 )(iγ µ ∂ µ + M )γ 0

where we can move along the derivative γ 0 since it is constant. If we define the Dirac conjugate
spinor field
ψ̄(x) ≡ ψ † (x)γ 0
and we remove the γ 0 at the right since the equation is equal to zero, the conjugate Dirac eq
reads ←
ψ̄(x)(i 
∂ + M) = 0

Exercise. Prove the identity γ 0 γ µ† γ 0 = γ µ .


CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 23

The conjugate dirac spinor transform under a Loretz tranformation in the following way:

ψ̄ 0 (x0 ) = (ψ 0 (x0 ))† γ 0 = ψ † (x)S † (Λ)γ 0 = ψ † (x)γ 0 S −1 (Λ)γ 0 γ 0 = ψ̄(x)S −1 (Λ)

Exercise. Prove γ 0 S † (Λ)γ 0 = S −1 (x), we have to use the identity γ 0 γ µ† γ 0 = γ µ .

Recall when we have to deal wirh the Dirac (conjugate) eq we deal with matrices and vectors:
ψ is a column vector, ψ̄ is a row vector, and (i
∂ ± M ) it is a matrix. In fact the order of them
changes: for Dirac eq (i
∂ − M )ψ = 0, for Dirac conjugate eq ψ̄(i
∂ − M ) = 0.The result is always
a vector of equations.

1.4.6 Continuity Dirac equation


Why do we introduce the conjugate Dirac eq? Because we want to find the continuity eq for
the Dirac eq in order to find a conserved quantity hoping that it could have the property of a
probability.
Now, we refear to Dirac conjugate and not eqs as Dirac eqs. As for the KG case, let us
multiply the Dirac eqs by the respected conjugate field:

∂ − M )ψ = 0
ψ̄(i 

ψ̄(i 
∂ + M )ψ = 0

sum the eqs together:


→ ← → ←
∂ )ψ = i ψ̄γ µ (∂µ ψ) + γ µ (∂µ ψ̄)ψ
 
∂ − M + i
0 = ψ̄(i  ∂ + M )ψ = ψ̄(i 
∂ + i

It is a total derivative of the following quantity:

ψ†ψ
   
µ µ ρ
j (x) ≡ ψ̄(x)γ ψ(x) = = i (x)
ψ̄γ i ψ j

Let us call it 4current. The conserved quantity is


Z Z
Q = d3 x ρ(~x, t) = d3 xψ † (x)ψ(x) > 0

and it is a positive quantity since ρ = ψ † ψ > 0. It seems we found a possible probability, however
it is just apparent positive defined, in fact the Klein paradox will be the same also for the Dirac
eq.

1.4.7 Bilinear forms and Lorentz transformations


Now, we can understend completely why gamma matrices have lorentz indices also if they are
constant in eqch reference frame: it si the Dirac bilinears and their relations with Lorentz
transformation
In our world we do not see spinorial indicies, so we assume that each observable must be
a scalar-spinorial quantity called bilinear, namely spinoral indicies are saturated. For example
throught product of ψ̄ and ψ. An object like

ψ̄(. . . )ψ

is a bilinear in the spinorial point of view. However bilinears are Lorentz tensor and so they
carry Lorentz indicies.
For example let us see two examples.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 24

Exercise. Show ψ̄Iψ is a scalar.


Recall ψ 0 (x0 ) = S(Λ)ψ(x) and ψ̄ 0 (x0 ) = ψ̄(x)S −1 (Λ), so

ψ̄(x)Iψ(x) → ψ̄ 0 (x0 )Iψ 0 (x0 ) = S −1 (Λ)IS(Λ) = I

Here there are no Lorentz indicies and in fact such bilinear qunatity is a scalar in the Minkowski
space (from a Lorentz transformation point of view).

Exercise. Show that ψ̄γ µ ψ is a vector.


Let us see if it transform as a vectror:

ψ̄ 0 (x0 )γ µ ψ 0 (x0 ) = ψ̄(x)S −1 (Λ)γ µ S(Λ)ψ(x) = Λµ ν ψ̄(x)γ ν ψ(x)

There is one Lorentz index and the bilinear object is a Lorentz vector because it transforms like
that.
From this example, we learn that even if the gamma matrix is constant in each reference
frame, the index µ is a real Lorentz index because the bilinear object transform like a vector.

( In these lectures µν are Lorentz indicies and αβ are spinorial indicies)

1.5 General solution of Dirac equation


Remember we imposed that Dirac eq is consistent with KG eq. So a solution of Dirac eq is aslo 22/3/21
a solution of KG eq. So let us start using what we know about KG eq: it contains two terms,
one for the positive energy solution and one for the negative one. In general also a solution of
Dirac eq can be written as a a sum of two terms related to the sign of the energy:

ψ(x) ≡ ψ+ (x) + ψ− (x) ' [exp(−ikx)u(k) + exp(ikx)v(x)]|k0 =ω0

where u(k) and v(k) are Dirac spinors (4dim vector). Now we have to indentify u, v in order to
do it let us apply the Dirac eq on ψ(x)± :

(i∂/ − M )ψ+ (x) ∼ exp(−ikx)(k/ − M )u(k) = 0 =⇒ (k/ − M )u(k) = 0


(i∂/ − M )ψ− (x) ∼ − exp(ikx)(k/ + M )v(k) = 0 =⇒ (k/ + M )v(k) = 0

Recall k/ = γ µ kµ . We found conditions in the momentum space: these are called the Dirac eq
in the momentum space. Notice we are assuming our particles have M 6= 0, so let us look for
solution u(k), v(k) in the rest frame of the particles (it can be done only if the particle is massive,
however it will move at the speed of ligth and do not exist a boost). In the RF: k µ = (M, ~0):

(k/ − M )u(k) = 0 =⇒ M (γ 0 − I4 )u(M ) = 0


(k/ + M )v(k) = 0 =⇒ M (γ 0 + I4 )v(M ) = 0

Now u(M ), v(M ) are the spinors in the particle rest frame.
Using the Dirac-Pauli representation for the gamma-matricies:
     
I2 0 I2 0 ξ
M − u(M ) = 0 =⇒ u(M ) = c
0 −I2 0 I2 0
     
I2 0 I 0 0
M + 2 v(M ) = 0 =⇒ v(M ) = c
0 −I2 0 I2 ξ

where ξ is a generic 2dim vector, since some 2 × 2 blocks deletes together, and c is a normalization
constant. Notice this results depend on the choicen representation of the gamma matricies.
So, we obtained that Dirac eq has 4 independent solution: two solution u, v for positive and
negative energy and each solution is a 2dim vector with 2 dof: in total 4 dof. It is intuitively
correct since Dirac eq is a 4dim eq.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 25

A possible basis are the following 4 vectors


√ √
   
ξ 0
ur (M ) = 2M r vr (M ) = 2M
0 ξr
where r = 1, 2 and ξ1 = (1, 0) and ξ2 = (0, 1).
Note. Do note that with the chosen normalization we have

ūr (M )us (M ) = 2M δrs


v̄r (M )vs (M ) = −2M δrs
ūr (M )vs (M ) = 0 = v̄s (M )ur (M )

where we defined ū = u† γ 0 and v̄ = v † γ 0 .


To find a generic solution in a generic frame, we have to apply a Lorentz boost from the
rest frame solution: u(k) = S(Λk )u(M ). However there is a trick equivalent of performing the
Lorentz boost. We will found a Dirac solution in the momentum space using the fact that

(k/ − M )(k/ + M ) = (k/ + M )(k/ − M ) = k 2 − M 2 |k0 =ωk = 0

for a particle of mass M . So a generic solution u(k), v(k) is something like

u(k) = C(k/ + M )u(M )


v(k) = C(k/ − M )v(M )

since they solve the Dirac eq in momentum space:

(k/ − M )u(k) = 0
(k/ + M )v(k) = 0

C is the normalization constant. We fix it requiring that the general solution must be equal to
the rest frame solution in a rest frame, so:

ūr (k)us (k) = 2M δrs


v̄r (k)vs (k) = −2M δrs
ūr (k)vs (k) = 0 = v̄s (k)ur (k)

then C = 1/[2M (ωk + M )]1/2 . Let us write explicitelly the expression for the Dirac spinors in
the Dirac-Pauli representation of gamma-matrix:
√ 
ωk + M ξr
(k/ + M )  for E > 0
ur (k) = p ur (M ) =  ~k · ~σ
2M (ωk + M ) √ ξr
ωk + M
~k · ~σ
 
(−k/ + M ) √ ξ
vr (k) = p vr (M ) =  ωk + M r  for E < 0
2M (ωk + M ) √
ωk + M ξr

Recall r = 1, 2 represents the two independent dof for E > 0 and E < 0. Notice ~k = (k1 , k2P , k3 ) is
vector of numbers while ~σ = (σ1 , σ2 , σ3 ) is a vector of 2 × 2 complex matricies, so ~k · ~σ = ki σi
is a 2 × 2 complex matrix. We found the Dirac general solutions in the momentum space!
Exercise. Show that Dirac conjugate ū(k), v̄(k) have the following expression:
!
k+M p ~k · ~σ
T T
= ξr ωk + M , −ξr √

ūr (k) = ūr (M ) p
2M (ωk + M ) ωk + M
!
k+M ~k · ~σ p
= ξrT √ , −ξrT ωk + M

v̄r (k) = v̄r (M ) p
2M (ωk + M ) ωk + M
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 26

Return to our goal: find ψ(x). We found a solution in the momentum space, using Fourier
tranform we write it in the coordinate space:
2
d3 k X
Z
1
ψ(x) = √ (cr (k)ur (k) exp(−ikx) + d∗r (k)vr (k) exp(ikx))k0 =ωk
(2π)3/2 2ωk r=1

we used the same normalization of KG solution. Where cr (k) and d∗r (k) are the weight of the
positive and negative solution: they are analogous objects to a(k), b∗ (k) for the KG eq. Now we
have r = 1, 2 so two independent solution for each positive and negative energy solution. The
conjugate solution is
2
d2 k X
Z
1
ψ̄(x) = √ (dr (k)v̄r (k) exp(−ikx) + c∗r (k)ūr (k) exp(ikx))k0 =ωk
(2π)3/2 2ωk r=1

1.5.1 Interpretation
What is the meaning of these independent 4 solution ur , vr with r = 1, 2 and the related dof?
We know that two solutions are the same of KG eq, in order to recover them let us define the
energy projectors Λ± : they project solutions over the KG ones.

Definition. The following 4dim matricies are projectors over the positive/negative energy
solution:
±k/ + M
Λ± ≡
2M
Exercise. Show that Λ± (k) are projectors, namely they respect the following properties:

1. Λ2± (k) = Λ± (k)

2. Λ+ (k) + Λ− (k) = I

3. Λ+ (k)Λ− (k) = 0

4. tr[Λ± (k)] = 2

Exercise. Show that Λ+ (k)ur (k) = (+1)ur (k) and Λ− (k)ur (k) = 0. While Λ+ (k)vr (k) = 0 and
Λ− (k)vr (k) = +vr (k).

There is an alternative expression for Λ± (k):


2
X ur (k)ūr (k)
Λ+ (k) =
2M
r=1
2
X vr (k)v̄r (u)
Λ− (k) = −
2M
r=1

Notice that since Λ is a matrix we need first a raw vector on the left (ie ur ) and a column vector
on the right (ie ūr ).

Exercise. Prove that the two definitions are equivalent.

So 2 dof are used for describe positive energy solution: ur , and negative energy solution: vr .
But there are still 2 dof given by the index r for each energy solution. Then why the dof were
duplicated? The reason is that this object describes particles with spin 1/2, so there are two
possibilities for each solution: to have spin +1/2 or −1/2. Let us formaly prove it.
Recall Σµν is a 6 independent matrices in 4×4 space and represents generators of the spinorial
representation of Lorentz group.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 27

Let us define explicitely generators for boosts and rotations: the 3 boost generators are
 
0i i 0 σi
Ki ≡ Σ = −
2 σi 0

notice they use time and space indices and a boost mix these two. While the 3 rotations
generators are  
1 jk 1 σi 0
Σi ≡ ijk Σ =
2 2 0 σi
there are only spatial indicies since rotations only mix spatial indicies.
So, if we make a boost in the spinorial space the solution changes as
 
0 0 i
ψ (x ) = S(ΛK )ψ(x) with S(ΛK ) = exp − ω0i Ki
2

While if we make a 3d rotation


 
0 0 i
ψ (x ) = S(ΛΣ )ψ(x) with S(ΛΣ ) = exp − θk Σk
2

where θk = ijk ω ij is the 3dim angle rotation.


So, in analogy with the coordinate space, the generator of the rotations ΣK represents the
angular momentum of the particle in the spinorial space. Σk are called spin operators and they
result in an internal rotation of the spinor corresponding to a rotation of our coordinates. So
what is the spin of particles describe by ψ? Consider the reference frame of the particle of
mass M 6= 0, the spinors are u1,2 (M ), v1,2 (M ), so if we apply the spin operator along the third
direction:
1 σ3 0 √
   
ξ 1
Σ3 u1 (M ) = 2M 1 = u1 (M )
2 0 σ 3 0 2

   
1 σ3 0 ξ 1
Σ3 u2 (M ) = 2M 2 = − u2 (M )
2 0 σ3 0 2

u1 , u2 are both energy positive solution of Λ+ , but they have respectivelly +1/2 and −1/2 of Σ3 .

Exercise. Verify the same relation for the negative energy solution:
1 1
Σ3 v1 (M ) = + v1 (M ) Σ3 v2 (M ) = − v2 (M )
2 2
Resuming, the massive Dirac spinor describes two states of energy (positive and negative)
which are interpretated as particle and antiparticle, and each of them are particle of spin 1/2
with 2 more dof ±1/2, so 4 dof in total.

1.5.2 The helicity operator


Be carefull that spin is not a good relativistic invariant while it is a good quantum number in 23/3/21
the non-relativistic particles: it is defined in the rest frame but it is not invariant with Lorentz
transformation.
The observable quantity connected with the spin is the helicity: the spin projected along the
direction of motion. It is the good relativistic quantum number.

Definition (Helicity operator). The helicity is the projection of the spin of the particle along
the direction of motion. It is obtained from the helicity operator

~ · p~
Σ
Σp ≡ = p̂i Σi
|~
p|
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 28

But why the spin is not more a good qunatum number? A good quantum number is a constant
and so must commute with the Hamiltonian; helicity operator communtes [H, Σp ] = 0, while the
spin spin operator don’t [H, Σ3 ] 6= 0. The commutation implies the quantity is conserved during
the evolution of the system.
Exercise. Prove that the commutator [H, Σ3 ] 6= 0.
Solution: recall the Dirac Hamiltonian:

H ≡ (−i~ ~ + βM ) = (~
α·∇ α · P~ + βM )

where we introduced the momentum operator. Let us explicelly compute the commutator: the
complication is that H is a differential and matricial operator.
X
α · P~ + βM, Σ3 ] =
[~ Pi [αi , Σ3 ] + M [β, Σ3 ]
i

from direct calcoulus: [β, Σ3 ] = 0. Compute


       
1 0 σi σ3 0 σ3 0 0 σi 1 0 [σ3 , σi ]
[αi , Σ3 ] = − =
2 σi 0 0 σ3 0 σ3 σi 0 2 [σ3 , σi ] 0

where [σ3 , σi ] = 2i3ik σk . So finally we obtain [H, Σ3 ] = ikj3 αk Pj 6= 0.


Exercise. Prove [H, Σp ] = 0.
Notice the definition of helicity is not covariant since it depends on 3dim vectors, it is better
to define a covariant object: the Pauli-Lubanski vector and helicity.
Definition. The Pauli-Lubanski vector is a covarinat (so 4dim) vector. It is defined as:
1
ω µ ≡ µνρσ Jνρ pσ
2
where Jνρ = Lνρ + Σνρ is the total angular momentum and Lνρ ≡ xν pρ − xρ pν . Notice any terms
in this definition is a well defined Lorentz tensor, and so the Pauli-Lubanski vector.
There are two terms in the expression for ωµ , but the external angular momentum one vanishes
since Lνρ pσ is symmetric in ρσ (double product o 4momentum), so due to the antisimmetry of
µνρσ , it vanishies. We can then write:
1
ω µ = µνρσ Σνρ pσ
2
Let us compute this object in the rest frame of the particle, where pµ = (M, ~0):
 0  
µ 1 µνρ0 ω 0
(ω )RF = M  Σνρ =⇒ =
2 ωi M Σi

since if we fix to zero an index of Levi-civita tensor, all other nonvanishing indicies run from 1 to
3. So the spatial component of Pauli-Lubanski vector is proportional to the spin of the particle.
Let us compute the square: it is a scalar so we can compute it in the RF

ω µ ωµ
   
i i 1 i σi σi 0 3
= Σ Σi = −Σ Σ = − = − I4
M2 RF 4 0 σi σi 4

This is consistent with the expression we know from QM, that is, the eigenvalue of s2 being
s(s + 1), and for particles with s = 1/2 it is exactly s(s + 1) = 3/4. The negative sign has no
particular physical meaning: it is due to the fact that our metric signature gives spacelike vectors
negative norms.
So, ω µ is an invariant quantity that for the Dirac spinor is associated to s = 1/2. More in
general, Pauli-Lubanski vector is a covariant way to define the spin of a particle.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 29

What is the connection between Pauli-Lubanski vector and helicity? Let us work out
the spin projection along a generic direction ~n such that it is a 4dim versor and it is space-like
orthogonal to the 4momentum of the particle, namely
(
nµ nµ = −1
 0
n
µ
n ≡ s.t.
~n nµ pµ = 0

Let us compute the contraction with Pauli-Lubanski vector, it is a scalar so we compute it in the
RF:  µ 
ω nµ ~ · ~u
= −Σ
M RF
Now, we can find the helicity considering a specific versor with the spatial part directed as the
momentum
1

p~
  µ 
ω nµ ~ · p~
Σ
µ
up ≡ |~
p|, ωp =⇒ = −Σp = −
M |~
p| M RF |~
p|

1.5.3 Chirality operator


For complete the description of Dirac spinors we treat the chirality operator and chirality
projection.
In n = 4 spinorial representation we can define a matrix
i
γ 5 ≡ − µνρσ γ µ γ ν γ ρ γ σ = iγ 0 γ 1 γ 2 γ 3
4
taking the definition of γ µ we can show that γ 5 satisfies the following properties

1.(γ 5 )† = γ 5
2.(γ 5 )2 = I4
3. γ µ , γ 5 = 0


Exercise. Prove the properties 1. 2. 3.

Notice in the Pauli/Dirac γ representation


 
5 I2 0
γ ≡
0 I2

Since property 3. we can show


[γ 5 , Σµν ] = 0
When two matricies commute, it implies that one can diagonalize simultaneosly both matricies,
in our case γ 5 , Σµν . This allows us to classify the eigenvectors by the eigenvalue f γ 5 instead of
the spin.

Definition (Chirality operator and projectors). γ 5 is called the chirality operator, and given it
we can define the right/left-hand chirality projectors PL and PR :

I − γ5
PL ≡
2
I + γ5
PR ≡
2
Exercise. Prove that PR,L are projectors, namely have to satisfy the follofing properties:
2 †
PR,L = PR,L , PR,L = PR,L , PL PR = 0 = PR PL , PL + PR = I.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 30

Appling the projectors, any Dirac spinor can be decomposed in L and R chiral components:

ψL ≡ PL ψ
ψR ≡ PR ψ

and so ψ = (PL + PR )ψ = ψL + ψR . Here chiral spinors ψL,R are eigenevector of γ 5 with


eingenvalues ∓1 respectively. This comes from direct computation:
5
 
5 5 I−γ
γ ψL = γ ψ = −ψL
2
5
 
5 5 I+γ
γ ψR = γ ψ = +ψR
2

Now, we can define also the conjugate of chiral spinors

ψ̄L ≡ (ψL† )γ 0 = ψ † PL γ 0 = ψ̄PR



ψ̄R ≡ (ψR )γ 0 = ψ † PR γ 0 = ψ̄PL

where we used γ 5 = γ 5 and the anticommutation property γ µ , γ 5 = 0. Pay attention in some


books the definitions are opposite.


Why we introduce chiral division of spinorial components? In QED all interactions (fermions
and photons) are equal for left and right components, in more general (in SM), for weak interaction,
there is a different between chiralities.

1.5.4 Dirac equation coupled to EM external field


In the same way we coupled KG eq with an EM external field, we will do the same with the
Dirac eq. We will follow the same prescription used for KG: the minimal coupling ansatz

∂µ → Dµ ≡ ∂µ + iqAµ

Using this prescription, the coupled Dirac eq reads

/ − M )ψ = (i∂/ − q A
(iD / − M )ψ = 0

using the 3dim notation we have


h i
~ + iq A)
α(−∇
i∂0 ψ = i~ ~ + βM + qA0 ψ

where Aµ ≡ (A0 , A)
~ and ∂µ ≡ (∂0 , ∇),
~ also recall γ 0 ≡ β and γ i ≡ βαi where
   
I2 0 0 σi
β= αi =
0 −I2 σi 0

Now we ask the same thing: the non-relativistic limit of the eq must be the Schrödinger eq.
It is useful to define
ψ(~x, t) ≡ exp(−iM t)ψ 0 (~x, t)
The idea of this definition comes from the NR limit where the energy of the particle is E =
M + p2 /2M +terms, so we can do a rescale in phase factorizing the mass dependence since we
are interested in the kinetic energy.9 Plug it in the differential eq:
h i
 0  ~ ~ 0
exp(−iM
   t)(i∂ 0 + M )ψ = exp(−iM
   t) i~
α (−∇ + iq A) + βM + qA 0 ψ

9
Given a wave, we can factorize it in plane wave exp(i·) and it is what we did with the mass.
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 31

Now it is explicit that to study the behavior of a particle in the NR limit it is not necessary the
mass term: we transformed the eq on ψ in an eq on ψ 0 .
h i
i∂0 ψ 0 = i~ ~ + (β − I)M + qA0 ψ 0
~ + iq A)
α(−∇

Let us define  0
ϕ
ψ0 ≡
χ0
where ϕ0 , χ0 are 2dim vectors. So spelling out the Dirac eq:
 0 " !  # 
~ + iq A)
~ ϕ0
  
ϕ 0 i~σ · (−∇ 0 0 I2 0
i∂0 0 = + M + qA 0
χ ~ + iq A)
i~σ · (−∇ ~ 0 0 −2I2 0 I2 χ0

we find that the original Dirac eq can be written as two 2nd (Weyl) equation

i∂0 ϕ0 = qA0 ϕ0 + i~σ (−∇ ~ 0


~ + iq A)χ
i∂0 χ0 = (qA0 − 2M )χ0 + i~σ (−∇ ~ 0
~ + iq A)ϕ

Notice there is an asimmetry: the mass term only appears in the ∂0 χ0 eq. This is an important
observation.
Till now we did not do any approximation, just rescaled the solution factorizing the mass.
Now let us apply the NR approximation, namely the mass term is the bigger contribution to
energy:
∂0 χ0

χ0  M |qA0 |  M

The first approximation regards the "kinetic energy" of the field, the second one regards the
electric potential.
In the first eq we cannot neglet anything since there are no mass term, but in the second eq
we can do something; the NR Dirac eqs reads:

i∂0 ϕ0 = qA0 ϕ0 + i~σ (−∇ ~ 0


~ + iq A)χ
i
χ0 = ~ 0
~ + iq A)ϕ
~σ (−∇
2M
The second eq is a constraint, it tells us how we can write χ0 in function of ϕ0 and derivative.
Do note since χ0 ∼ diff. function for(ϕ0 /M ), it is smaller that ϕ0 , but since we do not know how
small is ϕ0 we cannot neglet χ0 . We just say:

ϕ0 = large 2dim component


χ0 = small 2dim component

29/3/21
Comment. Chirality is esier that helicity since the dirac spinor is easily decomposed in chirality
than helicity. And also the SM couples terms with chirality.
Let us combine the eqs: insert the 2nd into the 1st
1  2
i∂0 ϕ0 = qA0 ϕ0 − ~ ϕ0
~ + iq~σ · A
−~σ · ∇
2M
This is the Pauli equation: it is the generalization of the NR Schrödinger eq coupled to an
external EM field for a spin 1/2 charged particle.
To write Pauli eq in the standard mode we open the square differential operator:
 2
~ ϕ0 = (∂i + iqAi )(∂j + iqAj )σi σj ϕ0
~ + iq~σ · A
−~σ · ∇
= ∂i ∂j + iq(∂i Aj + Aj ∂i ) + iqAi ∂j − q 2 Ai Aj σi σj ϕ0
 
CHAPTER 1. RELATIVISTIC WAVE EQUATIONS 32

where we are in 3dim and the position of the indicies is irrelevant. Be careful, as we already found,
derivatives also apply to ϕ0 , so the correct way to calculate the square requires the following
passages: iq∂i (Aj σi σj ϕ0 ) = iq(∂i Aj + Aj ∂i )σi σj ϕ0 .
Now we notice that all terms, except for ∂i Aj , are symmetric in ij, so we split the product
σi σj into its symmetric and antisymmetric parts. It is
1 1
σi σj ≡ {σi , σj } + [σi , σj ] = δij + iijk σk
2 2
where the first term is a symmetric tensor while the second term is antisymmetric. So we find
 
 2
−~σ · ∇ ~ ϕ0 = 
~ + iq~σ · A ∇~ 2 − q2A
~ 2 − 2iq A
~·∇~ − iq ∇~ · A]
~ + ~
q~σ · B
 0
ϕ
| {z } | {z }
from symmetric tensor from antisymmetric tensor

where the antisymmetric term comes out from


~ × A)
iq(∂i Aj )iijk σk = −qijk (∂i Aj )σk = −q(∇ ~ k σk = −q~σ · B
~

We can chose che Coulomb gauge: ∇ ~ = 0. So, the Pauli eq (NR Dirac eq for ϕ0 ) reads
~ ·A
 
∂ 0 1  ~ 2
~ + qA0 − q ~ ϕ0
i ϕ = − −∇ + iq A ~σ · B
∂t 2M 2M
it is the Schrodinger eq for the 2dim ϕ0 . Notice the first terms are similar to KG eq, while the
last term is new (denotes the coupled of particle spin with the external magnetic field): the spin
1/2 eq differs from the spin 0 eq by the presence of the dipole term
q ~ = −~ ~
HDip ≡ − ~σ · B µs · B
2M
where µ ~ s is the intrinsic magnetic moment
q q ~ (2)
~s ≡
µ ~σ = Σ
2M M
where Σ ~ (2) = ~σ /2 is the spin operator in 2dim (it is a 2 dim vector, not the Σ defined in 4dim).
Notice in NRQM we introduced the dipole term by hand, here it is a well defined qunatity
expressed in term of q, M . The Dirac eq also predicts an electron gyromagnetic factor ge = 2.
Let us see how.
Recall the definition of the magnetic dipole moment of a particle with external angular
momentum L, ~ charge q and mass m is
q ~ |~
µL | q
~L ≡
µ L =⇒ =
2M ~
|L| 2M
For the internal (spin) magnetic dipole moment we used the same definition:

~ (2) =⇒ |~µs |
 q   q 
µ
~s = ge Σ = ge
2M ~ (2) |
|Σ 2M

If we compare it with the theoretical prediction µ ~ (2) we have


~ s = (q/M )Σ
ge = 2
So it is a prediction that comes automatically from the NR Dirac eq. Experimental meauseres
were consistent with Dirac prediction (in Schrödinger eq the factor 2 is just an experimental
prediction, in Dirac eq it is derived).

Problem of Dirac eq
Dirac eq suffers by Klein paradox (as the KG eq). So we need a new approach to describe
particles: use the field theory! We change the wave function field to operator field, and then we
have to introduce the right space.
Now we finished the first part of the course.
Chapter 2

Lagrangian & Hamiltonian formalism

We will deal with the field theory description of a system. In order to do it we need to introduce
the Lagrangian and Hamiltonian formalism for the field theory.
Firstly we will see the Lagrangian/Hamiltonina formalism for a classical system with finite
dof. Then we will pass to a system with infinite (continuos) dof namely a field.
Consider a classical system with N dof. For example N particles in 1dim with mass mi ,
position qi (t) and velocity q̇i (t). Consider an external force linked to an external potential V (qi ).
One can describe the mechanical properties of the system by solving N Newton eq (second order
diff eq):
∂V
mi q̈i (t) = − i = 1, . . . , N
∂qi
The physical trajectories q(t) are the solution of these diff eq.

2.1 Finite system: Lagrangian formalism


An alternative description of the mechanical properties of the system can be obtained by defining
the object
L (qi (t), q̇i (t), t) ≡ L (t)
In conservative system the lagrangian does not depend explicity on time t. We will consider
conservative system:
L = L (qi (t), q̇i (t))
A typical lagrangian for 1 particle system is:
1
L (q(t), q̇(t)) = mq̇ 2 − V (q) = T − V
2
where T = Ekinetic . How to obtain the mechanical properties (physical trajectory) of a classical
particle from the lagrangian?
Consider a particle starting from q(tin ) and finishing at q(tfin ). In principle the particle can
follow infinite trajectory just respecting the boundary (initial) condition. We are interested in
which is the physical trajectory, to obtain it we have to introduce the action.
Definition (Action). The action of the system is
Z tfin
S[γ, tin , tfin ] ≡ dt L (qγ (t), q̇γ (t)) (?)
tin

so given a trajectory γ we obtain a number S that does not depend on time: the action depends
only on the specific trajectory. The action is a functional: it takes a function f (the trajectory)
and gives a number.

S :f →R
γ → S(γ)

33
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 34

The physical trajectory is the one that satisfies the least action principle.
Definition (Least action principle). The physical trajectory between the choosen boundary
condition (q(tin ), q(tfin )) is the one that satisfies:
δS = 0
This definition can also be said as the physical trajectory is the one that minimize the action (the
above condition is just a stationary condition, however the right solution is the minimum one).
Let us calculate the variation of the action along trajectories that satisfy the boundary
condition. Denote γ and γ 0 two near trajectories, namely
γ → γ 0 = γ + δ0 γ
This translates to the coordinates as:
qγ (t) → qγ 0 (t) = qγ (t) + δ0 qγ (t)
with the condition δ0 qγ (tin ) = δ0 qγ (tfin ) = 0. From now we simplify the notation neglecting the
γ.
Definition. We introduced the variation δ0 called synchronous variation:it connects the position
of two particles at the same time t, in fact it is defined as
δ0 q(t) = q 0 (t) − q(t)
The synchronous variation of the action is
δ0 S[γ] ≡ S[γ 0 ] − S[γ]
Z tfin Z tfin
= dt L (q + δ0 q, q̇ + δ0 q̇) − dt L (q, q̇)
tin tin
Z tfin
= dt δ0 L (q, q̇)
tin
Z tfin  
∂L ∂L
= dt δ0 q + δ0 q̇
tin ∂q ∂ q̇
What is the synchronous variation of the velocity? Starting from the definition of δ0 :
d 0 d
δ0 q̇ ≡ q̇ 0 (t) − q̇(t) =

q (t) − q(t) = δ0 q
dt dt
So the variation of the action reads:
Z tfin   Z tfin  
∂L d ∂L d  ∂L


δ0 S = dt + δ0 q + dt

δ0 q
tin ∂q dt ∂ q̇ t
in
dt ∂ q̇
where the second terms vanishing since the integral is proprtional to the variation δ0 q evaluated
in the boundaries where it values zero.
Recall the least action principle states that the physical trajectory has to satisfy
Z tfin  
∂L d ∂L
δ0 S = dt + δ0 q = 0
tin ∂q dt ∂ q̇
and this must hold for any δ0 q with fixed boundary condition, this means the square brackets
must be zero. They are the so called Eulero-Lagrange eq:
∂L d ∂L
+ =0
∂q dt ∂ q̇
Do note physical quantity is the Eulero-Lagrange eq and not the Lagrangian itself, in fact there
could be different lagrangian with the same EL eq.
Exercise. Let f (q) be a generic function of q and not of q̇, show that the two lagrangian L(q, q̇)
and L0 (q, q̇) = L(q, q̇) + df
dt are equivalent, namely they give the same eq of motion.
These results extend easily to systems of n particles just using vectorial notation: q, p ∈ Rn .
It holds for Lagrangian and hamiltonian formulation.
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 35

2.2 Finite system: Hamiltonian formulation


Recall in the lagrangian formulation we can obtain all the mechanical information (the trajectories) 30/3/21
definig the lagrangian, the action and using the least action principle.
Let us see the Hamiltonian formulation, another method to derive the mechanical property of
the system. It is obtained from the Lagrangian formulation. Befor to introduce the hamiltonian,
let us introduce the canonical momentum p conjugates to the coordinate q:
∂L
p≡
∂ q̇
The hamiltonian of the system is given by the Legendre transform

H(p, q) ≡ p · q̇ − L (q, q̇)

where q̇ ≡ q̇(p, q) from solving the definition of p.


To not confuse the notation remember that in the lagrangian formalism the independent
variable are q, q̇ and so L = L (q, q̇), while in the Hamiltoninan formalism the independent
variable are p, q and so H = H(p, q).
In general the hamiltonian is a function of time: H(p(t), q(t), t). For our course we assume
that the systems are conservative and so the Hamilonian does not depend explictly from time.
We want to obtain the eq of motion: they come by differentiate the Hamiltonian and plug
the definition of H in term of L :
   
∂H ∂H
dH = dp + dq
∂p ∂q
   
∂p ∂ q̇ ∂L ∂q ∂L ∂ q̇ ∂p ∂ q̇ ∂L ∂q ∂L ∂ q̇
= q̇ + p − − dp + +p − − dq
∂p ∂p ∂q ∂p ∂ q̇ ∂p ∂q ∂q ∂q ∂q ∂ q̇ ∂q
∂L
= q̇dp + dq EL eq
∂q
= q̇dp − ṗdq

if the system is not conservative there is an additional term +(∂H/∂t)dt.


The eq of motion in the Hamiltonian formalism, called Hamiltonian eqs are
∂H
q̇ =
∂p
∂H
ṗ = −
∂q
Recall here qunatities are vectrors. Do note Hamiltonian eq are equivalent to EL eq, in fact to
obain the former we used the latter.
There is an alterntive way to write the hamiltonian eq and is using the Poisson brackets. Let
us introduce them for a finite n dof system.
Definition. Given f (p, q) and g(p, q), the Poisson brackets at a given time t are defined as
 
∂f ∂g ∂f ∂g
{f, q}t ≡ −
∂p ∂q ∂q ∂p t
recall p, q are vector and this is a condensed notation: there is sum from 0 to N .
Using Poisson parentesis, Hamilton eq can be witten as

q̇(t) = {q, H}t


ṗ(t) = {p, H}t

Exercise. Show the equivalence of the last formula.


CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 36

In term of Poisson bracket the following relation holds:

{p, p}t = 0 = {q, q}t


{q, p}t = 1 = −{p, q}t

Using the index notation:

{pi , pj }t = 0 = {qi , qj }t
{qi , pj }t = δij = −{pj , qi }t

in future we will use similar properties.

2.3 Continuos system: Lagrangian formulation


Now, apply the same formalism to describe classical system with inifinite dof (so continuos dof:
a calssica field system). Let us states the parallellism between the finite and contiuous systems
(dof):
• For the finite: the cooridnates are
P qi (t) and when we want a total quantity (ie the energy)
we sum the dof together with N i=1 , then in the Lagrangian formalism we describe the
system using L (qi , q̇i ).

• For the continuos: the variable is a funtion of time per each coordinate ϕ(t) ≡ ϕ(x, t) with
x ∈ R3 , so we are labelling the dof with a continuos coordinate x; when we want total
qunatity we use the continuos limit of the sum, namely the integral in the volume of the
coordinates: Z
d3 x
V
then the system is described by a quantity called Lagrangian density L(ϕ, ∂0 ϕ, ∂i ϕ).
This is the parallelism to pass from a finite dof system to an infinite dof system. Let us define
the lagrangian theory in field theory. Consider a conservative system.
Definition. In field theory is more useful to introduce the Lagrangian density:

L(ϕ(x, t), ∂0 ϕ(x, t), ∂i ϕ(x, t)) = L(x, t)

and the lagrangian, that is just a function of time, is obtained as


Z
L (t) ≡ d3 xL(x, t)
V

Now, in analogy with the dicrete case, let us define the action as
Z tf in Z
S[ϕ, D] ≡ dtL (t) = d3 xdtL(x, t)
tin D

where D are the boundary condition of a 4dim volume.


The action is a functional and an observable, so it takes functions and gives real numbers:

S : ϕ → S[ϕ] ∈ R

Notation. Do note, as in the interested to describe relativistic theory we will use from now a
compact notation where x = (~x, t) identifies coordinates in the Minkowski spacetime. So, for
example, the lagrangian density reads

L(x) = L(ϕ(x), ∂µ ϕ(x))

Tipically D ≡ M4 is the Minkowski space with some boundary conditions at infinity: we assume
that the fields and their derivatives vanish at infinity.
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 37

Now, as we did for the finite case, let us introduce the minimum least principle: we want
identify the right physical configuration ϕ(x) (field) of our system.
Definition. The physical field configuration ϕ(x) for a chosen boundary condition ∂D is the
one that minimize the action. So in particular the necessary condition is that the field must be a
stationary point of the action, namely
δS = 0
Similar with the discrete case, we assume a synchronous variation for the field: δ0 ϕ(x) such
as
ϕ(x) → ϕ0 (x) = ϕ(x) + δ0 ϕ(x)
and we impose that the boundary conditions are conserved, namely δ0 ϕ(∂D) = 0. This variation
induces a variation on the action:

δ0 S[ϕ] ≡ S[ϕ0 ] − S[ϕ]


Z Z
= d4 xL(ϕ + δ0 ϕ, ∂µ ϕ + δ0 ∂µ ϕ) − d4 xL(ϕ, ∂µ ϕ)
ZD D
4
= d xδ0 L(ϕ, ∂µ ϕ)
ZD  
4 ∂L ∂L
= d x δ0 ϕ + δ 0 ∂µ ϕ synchronuous
D ∂ϕ ∂∂µ ϕ commute with the
Z   Z   partial derivative
∂L ∂L ∂L

4 4 
= d x − ∂µ δ0 ϕ + d x∂
µ ∂∂ ϕ δ0 ϕ
 
D ∂ϕ ∂∂µ ϕ D µ

where we variation of the lagrangian is just its partial derivative. The last term cancels out for
the boundary condition (as in the finite case). We are left with the same type of expression of
the finite case: since the result is valid for any δ0 ϕ we have
∂L ∂L
ϕ − ∂µ =0
∂ϕ ∂∂µ ϕ
this are the Eulero-Lagrange eq for a field theory. Notice the structure is the same but it is an eq
for the lagrangian density. As before, the physical quantity is not the lagrangian density, because
it is not unique for a field theory, but it is the EL eqs.
Exercise. Consider k µ (ϕ) depending only on ϕ (and not for ∂µ ϕ). Show that the two lagrangian
L, L0 are equivalente:

L(ϕ, ∂µ ϕ) L0 (ϕ, ∂µ ϕ) = L(ϕ, ∂µ ϕ) + ∂µ k µ (ϕ)

ie they give the same EL eq (eq of motion).

2.3.1 Interlude on functional and functional derivative


That was the basic description of a lagrangian field theory. To inroduce the hamiltonian formalism
we need a deeper interlude about functionals and functional derivative.
Definition (Functional). A functional is a map between a manifold of function C (tipically
continuos or differentiables) and a number R or C

F : C → R or C
f 7→ F [f ]

for example

F : M4 → R
x 7→ f (x)
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 38

Example. The action is a functional


Z
S[ϕ, ∂D] ≡ d4 xL(x, t)
D

Example. The lagrangian is a function of time, but it is also a functional of fileds at fixed time:
consider a field ϕ(x, t), suppose to fix a time t and let us use the notation ϕ(x, t) ≡ ϕt (x) for our
field (namely we estrapolate the time dependence since we fixed the time), now

ϕt : R 3 → R
x 7→ ϕt (x)

So the lagrangian is a function of time L (t) ≡ d3 xL(x, t), however when we fixed the time,
R

the lagragian "becomes" a functional

Lt : Ct → R
ϕt 7→ Lt [ϕt ]

Now let us introduce the derivative of a functional with respect to the function.

Definition (Variation of the functional). Let g, δg be function (with δg  1), then the variation
of a functional is
δF [ϕ] ≡ F [g + δg] − F [g]

Definition (Functional derivative). The functional derivative δF/ deg g is defined by the relation
Z  
δF [g]
δF [ϕ] ≡ δg(x)
D δg(x)

where δF/δg(x) is the functional derivative of the functional F to respect the function g(x). This
object is a well defined derivative: it respects linearity, Leibniz rule and so on.

Proposition. Do note any function can be written using the identity functional:
Z
g(y) = dxδ(x − y)g(x) ≡ Fid [g]

For it
δFid [g] δg(y)
= = δ(x − y)
δg(x) δg(x)
Example.
δq(t) δϕ(y) δϕt (y) δϕ(y, t)
= δ(t − t0 ) = δ 4 (x − y) = = δ 3 (x − y)
δq(t0 ) δϕ(x) δϕt (x) δϕ(x, t)
so when we integrate a function with the delta we obtain the fuction: it is the functional identity.

Now we are ready to pass to the hamiltonian formulation.

2.4 Continuos system: Hamiltonian formulation


As for the finite case, we introduce the canonical conjugate momentum field:
∂L δL (ϕ)
π(x, t) ≡ =
∂∂0 ϕ(x, t) δ ϕ̇t (x)
then we can define the hamiltonian density of the system as the Legendre transformation of the
lagrangia density
H ≡ π(x)∂0 ϕ(x) − L
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 39

while the Hamiltonian operator is


Z
H(t) ≡ d3 xH(x, t)

Notice the Hamiltonian (as for Lagrangian) is a function of time because we integrate in the
spatial coordinates, but at a fixed time t, it is a functional of the field ϕt (x) and the conjugate
field πt (x):
H(t) ≡ H[ϕt , πt ]
Now we can obtain the Hamilton eq of motion by differentiating the Hamiltonian H:
Z  
3 δH δH
dH = d x δπt (x) + δϕt (x) using the
δπt (x) δϕt (x) definitions and the
Z Legendre
transformation
= d3 x(ϕ̇t (x)δπt (x) − π̇t (x)δϕt (x))

where dot inidcates time derivative. The hamilton eq are


δH
ϕ̇(x, t) =
δπt (x)
δH
π̇(x, t) = −
δϕt (x)

They are Hamilton eqs for field. Now, we can define poisson bracket also for the field theory,
just replacing function with functional.

Definition. Given two functional F = F [ϕt , πt ] and G = G[ϕt , πt ], one can define the Poisson
brackets as Z  
3 δF δG δF δG
{F, G}t = d x −
δϕt (x) δπt (x) δπt (x) δϕt (x) t
Using Poisson brackets, Hamilton eq can be written as

ϕ̇(x, t) = {ϕ, H}t


π̇(x, t) = {π, H}t

In terms of Poisson bracket we have also the following relation

{ϕ, ϕ}t = 0 = {π, π}t {ϕ, π}t = δ 3 (x − y) = −{π, ϕ}t

In this course we will pass to Hamiltonian formalism since the quantization process holds
better in Hamiltoninan formalism.

2.5 Continuos symmetries & Noether Theorem


Let us see general properties about a field theory. Starting from the Lagrangian of the theory, 12/4/21
the Noether theorem associates conserved quantities to its continuos symmetries. Lagrangian
simmetries are respected also by the field theory. However do note there are symmetries of the
theory that are not symmetries of the lagrangian density: for example transformations that hold
δL = ∂µ k µ gives δS = 0, so even if it is not a symmetry of the lagrangian it is a symmetry of the
theory.

Definition (Symmetry). A symmetry of a theory is a transformation that act in general on


both fields and coordinates and leaves the action invariant: δS = 0. Namely a symmetry
transformation leaves invariant the eq of motion.

Let us classify the type of symmetries:


CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 40

Discrete symmetry associated to a discrete group of description. For example em is invariant


under parity where
Parity = {1, −1}

Continuous symmetry associated to continuos group transformation. For example SO(2):


rotations in 2dim  
cos θ sin θ
SO(2) =
− sin θ cos θ

Each group represents a symmetry, so we will use group and symmetry as synomimous. For
the rest of the class we will focus on continuous groups.
Continuous symmetries can be global or local, and each of them can be interest only fields,
or both coordinates and field. Let us build up a table with some examples:

Continuos symmetries Global αi = const Local αi = αi (x)


Internal transformation Leptonic and hadronic gauge symmetries
(only fileds transform) number, flavor symmetry like EM and QED
Spacetime trasformation Lorentz, Poincarè, general relativity
both fields and coordinates Galileian group
transform

2.5.1 Global continuous symmetry


Fundamental shape of the global continuos symmetries is the Noether theorem.
Theorem (Noether theorem). To every continuous global symmetry of the action is associated
µ
a conserved current and a conserverd charge: namely the continuity eq ∂µ j(a) = 0 holds. The
conserved charge is Z
Q(a) ≡ d3 xj(a)
0

where a = 1, . . . n is the number of independent parameters of the group of symmetry.


The conservation of the charge derives directly by the continuity eq:
Z Z t2 Z   Z t2 Z Z t2
4 µ 3 0 i 3 0

0= d x∂µ j(a) = dt d x ∂0 j(a) + 
∂i 
j(a) = dt∂0 d xj(a) = dt∂ Q(a) = Q(a) (t1 )−Q(a) (t2 )
M4 t1 R3 t1 R3 t1

where we set the integral of the divergence of j(a)


i to zero because of we assume field at infinity are

zero: ϕ(t, ∞) = 0 (necessary to have finite quantities). Explicitely, we used the Gauss theorem
Z I
3 i
d ∂i j(a) = dΣ ~u · ~j = 0
R3 ∂R3

since ~j(t, ∞) = j(ϕ(t, ∞)) = j(0) = 0.


We disciminate global continuous internal between internal trasform (touch only field) and
coordinate transform (touch coordinate and consequently field).

Internal global symmetries


An internal global symmetry transforms filed and not coordinates, so an infinitesimal tranforma-
tion acts as
x0µ = xµ δxµ = 0
=⇒
ϕ0 (x) = ϕ(x) + δ0 ϕ(x) δ0 ϕ = (a) X(a) (ϕ)
where the contranction in (a) index is the sum between generators X(a) of the symmetry.
The variation of the action is:
Z Z    
4 4 ∂L ∂L ∂L
δS = d xδ0 L = d x − ∂µ δ0 ϕ + ∂ µ δ0 ϕ
∂ϕ ∂∂µ ϕ ∂∂µ ϕ
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 41

Unlike what we usually do when we derive the EL equations, we cannot neglect the boundary
term: since we are integrating in an arbitrary volume, there is no reason why the variation of
the field should vanish at the boundary. Notice the first term (the one multiplied by δ0 ϕ) is the
EL eq and it is zero. While in order for the variation in the action to be zero, we must ask the
extra term to vanish: this means
Z   Z    
4 ∂L (a) 4 ∂L ∂L
d x∂µ δ0 ϕ =  d x∂µ X (ϕ) = 0 =⇒ ∂µ X (ϕ) = 0
∂∂µ ϕ ∂∂µ ϕ (a) ∂∂µ ϕ (a)

since it must hold for any integration region and for any chice of (a) .
µ
This is a conservation eq just like ∂µ j(a) = 0, so we found the conserved current

µ ∂L
j(a) ≡ X (ϕ)
∂∂µ ϕ (a)

and now we are done, the conserved charges follow by definition:


Z
∂L
Q(a) ≡ d3 x X (ϕ)
∂∂0 ϕ (a)
Recall the index a = 1, . . . n counts the number of independent parameters of the symmetry
group (ie the number of generators X(a) ).

Spacetime global symmetry


They are transformations applied to fields and coordinates:
µ
x0µ = xµ + δxµ δxµ ≡ (a) Y(a)
=⇒
ϕ0 (x0 ) = ϕ(x) + δϕ(x) δϕ ≡ (a) X(a)

Notice there are two important difference compared to internal symmetry:

1. Regarding the field transform δϕ, we can write it as

δϕ ≡ ϕ0 (x0 ) − ϕ(x) = ϕ0 (x0 ) − ϕ(x0 ) + ϕ(x0 ) − ϕ(x) = δ0 ϕ(x) + (∂µ ϕ)δxµ


 

where the first term is the syncronous variation (how field changes at the same coordinate),
while the second term is called transport term.

2. Regarding the coordinate transformation: in general under a coordinate transformation


also the volume of integration changes, to take into account it we have to introduce the
jacobian
d4 x0 ≡ |Jac|d4 x = (1 + ∂µ δxµ )d4 x
so δd4 x = d4 x0 − d4 x = ∂µ δxµ d4 x.

Under a spacetime global symmetry, the variation of the action reads


Z Z
δS = (δd x)L + d4 xδL
4

Z Z
= d x(∂µ δx )L + d4 x(δ0 L + (∂µ L)δxµ )
4 µ

Z     
4 µ ∂L ∂L ∂L
= d x ∂µ (Lδx ) + − ∂µ δ0 ϕ + ∂ µ δ0 ϕ
 

∂ϕ
 ∂∂µ ϕ ∂∂µ ϕ

where we canceled the EL eq. In order to have the variation of the action equal to zero, the
integrand must be zero since we are integrating in a generic region. Do note it is a divergence
and so we found a conserved current! It is, however, still written in terms of δ0 ϕ, while the
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 42

variation of the field is not the synchronous one but it is δϕ. So let us substitute the expression
for δ0 ϕ in terms of δϕ. This yields
Z   
4 µ ∂L ρ ∂L
δS = d x∂µ Lηρ − ∂ρ ϕ δx + δϕ δϕ =
∂∂µ ϕ ∂∂µ ϕ δ0 ϕ + (∂µ ϕ)δxµ
Z   
(a) 4 µ ∂L ρ ∂L
= d x∂µ Lηρ − ∂ρ ϕ Y(a) + X =0
∂∂µ ϕ ∂∂µ ϕ (a)

this holds ∀(a) , so the conserved current and conserved charge are
 
µ µ ∂L ρ ∂L
j(a) ≡ −Lηρ + ∂ρ ϕ Y(a) − X
∂∂µ ϕ ∂∂µ ϕ (a)
Z
Q(a) ≡ d3 xj(a)
0

So we found the conserved current and charges for the case of a global symmetry theory
for internal and spacetime transformation. Each time these transform hold, these objects are
conserved.

2.5.2 Application: invariance under Poincarè group


If we want to build a natural theory, we must impose the principle of special relativity, namely
the action must be a Lorentz scalar. Poincaré transforms are the most general ones which respect
the principle of special relativity, they consist in combinations of translations (4 generators),
Lorentz boosts (3 generators) and Lorentz rotations (3 generators).

Translation
Consider a theory and impose the symmetry under translation. Translationa are spacetime
transforms which change coordinates but not fields. So an infinitesimal translation holds

x0µ = xµ + µ δxµ = µ = ν ηνµ = (ν) Y(ν)


µ
=⇒
ϕ0 (x0 ) = ϕ(x) δϕ = 0

so for a translation we have no generators for the field transform (namely X = 0), while
µ
coordinates have generators Y(ν) = ηνµ (namely for each generator there is a shift ν along such
direction). Notice ν indicates the number of independent parameters.

Comment (Motivation of X = 0). Suppose to be in a system O and fix another system O0 , now
µ is the vector connected the two systems. Take a point x in O, the field seen by an observer at
O is ϕ(x). While an observed in O0 , will measure ϕ0 (x). The observer in O0 will measure the
same field-value if he adjustes its coordinates, namely if he applies a translation x0 = x +  he
will measure ϕ0 (x0 ) = ϕ(x).

From the general expressions for spacetime global symmetry, the conserved current is

µ ∂L
j(ν) ≡ ∂ν ϕ − Lηνµ ≡ T̃νµ
∂∂µ ϕ

and it is called canonical energy momentum tensor (little different from the covariant energy
momentum tensor since for some lagrangian they differs but they describes the same physical
thing). The conserved charge is
Z Z
Q(ν) ≡ d3 xT̃ν0 ≡ d3 xPν ≡ pν
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 43

where we identify pν as the 4momentum and so Pν is the density of 4momentum. Why this
identification? Running the index ν we find quantities which are dimensional energy and
momenta:
Z Z  
3 3 ∂L
Q(0) ≡ p0 = d P0 = d ∂0 ϕ − L = H
∂∂0 ϕ
Z
Q(i) ≡ pi = d3 xπ(∂i ϕ)

Boost
Let us do the same we did for traslation imposing invariance under Lorentz transform (boosts). 13/4/21
We will obtain another set of current and charge conserved. Boost are spatial-rotations and
spatial(veloctity)-boosts. Consider an infinitesimal boost applied to a vector field: recall Lorentz
transforms both coordinates and fields
µ
x0µ = xµ + ωνµ xν δxµ = 12 ω ρσ Y(ρσ)
=⇒ µ
ϕ0 (x0 ) = ϕ(x) − 2i ω µν Σµν ϕ(x) δϕ = 12 ω ρσ X(ρσ)

where ω is an antisimmetric tensor (6 dof), and so we included a 1/2 in the coordinate trans-
formations since we are summing over two antisymmetric indices, so we will have two copies of
every term. Generators of the transforms are such that
1 µ
ωνµ xν = ω ρσ Y(ρσ)
2
i 1
− ω µν Σµν ϕ(x) = ω ρσ X(ρσ)
2 2
which is solved by
µ
= ηρµ ηνσ − ησµ ηνρ xν

Y(ρσ)
X(ρσ) = −iΣρσ

here (ρσ) account for the generators of the Lorentz transform, which are the free indices of the
Lorentz transformation.
Comment (Generator form). The generator X(ρσ) depends on the specific representation of the
field considered: if ϕ is a scalar field (KG eq), it does not transform under Lorentz and we have
δϕ = 0 and so X = 0 (and so the representation of the generators of Lorentz transformations for
our field: Σ); if ϕ is a spinor field it transforms under Lorentz with δϕ and the representation of
the transform in the spinorial space is Σρσ = i[γ µ , γ ν ]/4; if ϕ is a vector field we will see.
We can directly plug in X and Y µ in the general formula of conserved quantities to obtain:

µ
  ∂L
jρσ = xρ T̃σµ − xσ T̃ρµ − i Σρσ ϕ
∂∂µ ϕ
Z
Q(ρσ) = d3 x (xρ Pσ − xρ Pρ ) + Sρσ 0
 
= Lρσ + Sρσ ≡ Jρσ

Do note the conserved charge containts two pieces: the first is independent on how field trans-
forms, the second term depends on the way the field trasforms (Σ). Here J is the total angular
momentum, L is the external angular momentum, while we identify S as an intrinsic angular
momentum: the one associated on how field transforms. So a field theory predicts an angular
momentum which depends on how the field transforms, the spin.

Putting together resuts from translations and boosts, we proved that a relativistic theory,
namely invariant under Poincarè group, has in total 10 conserved quantities: 4 related to
traslations and 6 related to roto-boosts (3 dimensional rotation and 3 spatial-boosts).
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 44

Resuming: we defined a field theory, for any field theory we defined the related lagrangian
density, then we obtained the EL equation. We saw how to look the eq of motion in the
Hamiltonian representation. We saw that if we have global continuos symmetry of our theory,
with Noether theorem we associate a conserved quantity. If the theory is relativistic there are
10 conserved quantities related to Poincarè symmetry. 6 of them compound the total angular
momentum, it is a sum of an external angular momentum and an internal angular momentum
related to how the field transforms called spin.

2.6 Scalar field theory


We will discuss the easiest field theory under the Lagrangian/Hamiltonian formalism: the scalar
field. We assume a relativistic theory (invariant under Poincarè group). Let us start with general
requirements for the lagrangian density:

1. since the action is an observable (hermitian operator), we impose it is a real functional (it
gives real number), and so the lagrangian density must be a real function;

2. since the action is adimensional in natural unit (since h is adim), the lagrangian density
must have [L] = [M ]4 in natural unit.

3. since we treat relativistic theory, the action must be a Poincarè scalar (scalar under Poincarè
transformation), the lagrangian density must be a scalar

4. (it is an doc-requirement) since in physics eq of motion are 2nd order differential eq, we
require the lagrangian is function only on field and first derivative: L(x) ≡ L(ϕ, ∂µ ϕ).

Real scalar field Now, let us start to discuss a real scalar field theory. For a while we only
consider free field (without interactions). We have to impose some conditions in order to have a
real scalar field:

1. real: ϕ(x) = ϕ∗ (x)

2. Lorentz scalar: ϕ0 (x0 ) = ϕ(x)

In analogy to classical theory (finite system), we choose the Lagrangian density


1 1
L(x) = (∂µ ϕ)(∂ µ ϕ) − m2 ϕ2
2 2
The first term is the kinetic term and the second is the mass term. This is the lagrangian for a
theory of free field. Why do we talk about mass term? Since [∂µ ] = [L]−1 = [M ], then [ϕ] = [M ]
and we can also see that [m] = [M ] (in support to the.
Now, let us calculate the EL eq of motion:
∂L ∂L
− ∂µ = −m2 ϕ − ∂µ ∂ µ ϕ = 0 =⇒ ( + m2 )ϕ = 0
∂ϕ ∂∂µ ϕ

this is the KG eq, we already know it. Recall its solution, now we can say it is the physical field
configuration of a real scalar field theory. Solution is:

d3 k
Z
1
ϕ(x) = √ (a(k) exp(−ikx) + a∗ (k) exp(ikx))k0 =ωk
(2π)3/2 2ωk
The hamiltonian formulation of the theory is obtained throught the conjugate field:
∂L
π(x) ≡ = ∂0 ϕ
∂∂0 ϕ
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 45

The hamiltonian density is obtained via Legendre transformation of the Lagrangian density
1 1 ~ 2 1 2 2
H(x) = π∂0 ϕ − L = π 2 + (∇ϕ) + ωk ϕ
2 2 2
Notice the energy of the field (the hamiltonian) is the integral of a positive fuction and so it is
also positive: Z
H = d3 xH(x) ≥ 0

Now, we can write the Hamilton eq of motion:


δH
∂0 ϕ = = {ϕ(x), H}t = π(x)
δπ(x)
δH
∂0 π = − = {π(x), H}t = ∂02 ϕ(x)
δϕ(x)

where the index t indicates that the time is fixed for the parenthesis. Since our theory is a
relativistc theory (simmetric under Poincarè transformation), there are 10 conserved quantities
from Noether theorem, they are the 4momentum and the total angular momentum:

Q(ν) = p(ν)
Z
J(µν) = d3 x[xρ Pσ − xρ Pρ ] = L(µν)

where J is the total angular momentum.

Complex scalar field Let us complicate the problem a little bit: consider complex scalar
field (free). In this case we have ϕ(x) 6= ϕ∗ (x) but we impose again invariant under Poincarè:
ϕ0 (x0 ) = ϕ(x). The lagrangian density is similar, however we combine ϕ∗ ϕ in order to have a
real lagrangian:
L(x) = (∂µ ϕ∗ )(∂ µ ϕ) − m2 ϕ∗ ϕ
There are again the kinetic and mass energy term. Notice there is not the factor 1/2 since a
complex field describes two independent dof: one for the field and one for the conjugate. Do
note we can always write a complex field with two real scalar field with

ϕ1 (x) + iϕ2 (x) ϕ1 (x) − iϕ2 (x)


ϕ(x) = √ ϕ∗ (x) = √
2 2
the only connection between the two field is that they have the same value of the parameter m.
Let us obtain the EL eq: deriving for ϕ∗ we obtain the eq for ϕ and viceversa
∂L ∂L

− ∂µ = −( + m2 )ϕ = 0
∂ϕ ∂∂µ ϕ∗
∂L ∂L
− ∂µ = −( + m2 )ϕ∗ = 0
∂ϕ ∂∂µ ϕ

however note
( + m2 )ϕ = 0 = ( + m2 )ϕ∗
Recall the general complex solution of KG eq:

d3 k
Z
1
ϕ(x) = √ (a(k) exp(−ikx) + b∗ (k) exp(ikx))k0 =ωk
(2π)3/2 2ωk
The two dof of the field are represented by a and b in the solution.
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 46

Now, develope the hamiltonian formulation. Since we have two fileds, we have two independent
conjugate fields:
∂L
π(x) ≡ = ∂0 ϕ∗
∂∂0 ϕ
∂L
π ∗ (x) ≡ = ∂0 ϕ
∂∂0 ϕ∗

For a complex scalar field, we have two pair of conjugate fields: (ϕ, π) and (ϕ∗ , π ∗ ). The
hamiltoninan density comes from the legendre transformation. From now, let us use a peculiar
notation: " · ", it means the sum off its argument over all dof. In our case it is the sum over the
normal field and the conjugate one

H(x) = ”π∂0 ϕ − L” = π∂0 ϕ + π ∗ ∂0 ϕ∗ − L = π ∗ π + (∇ϕ


~ ∗ )(∇ϕ)
~ + m2 ϕ∗ ϕ

Notice again this quantity is automatically positive. The Hamiltonian is always H = d3 xH(x) ≥
R

0.

Exercise. The conserved quantities associated to the Poincarè invariance by the Noether theorem
are for exercice.

2.6.1 U(1) invariance for a complex field theory


A complex scalar field is symmetric under the a global internal U (1) symmetry. Consider the
following U (1) internal (do not touch the coordinate) transformation:

ϕ0 (x) = exp(iα)ϕ(x)

ϕ0 (x) = exp(−iα)ϕ∗ (x)

where α ∈ R constant is the only generator. It is symmetry for the lagrangian since from direc
calculous:
∗ ∗
L0 (x) = (∂µ ϕ0 )(∂ µ ϕ0 ) − m2 ϕ0 ϕ0 = L(x)
So from Noether theorem there must be a conserved quantity. The infinitesimal form of such
transformation is
x0µ = xµ δxµ = 0
0 =⇒
ϕ (x) = (1 + iα)ϕ(x) δ0 ϕ = iαϕ = αX
and respectively for the conjugate, just take the conjugate. So we found generators Xϕ = iϕ and
Xϕ∗ = −iϕ∗ , while Y = 0 (notice there is a generator for each dof). Take the general expression
for conserved current and charge:
∂L ∂L ∂L ↔
µ ∗ ∗ ∗
j(”) =” X” = Xϕ + Xϕ∗ = i(∂µ ϕ )ϕ − iϕ (∂µ ϕ) = iϕ ∂µ ϕ
∂∂µ ϕ ∂∂µ ϕ ∂∂µ ϕ∗

Z Z
∂L
Q(”) = d3 x” X” = d3 x(iϕ∗ ∂µ ϕ)
∂∂0 ϕ

this is the conserved current associated to the invariance of such U (1) transformation.

2.6.2 Application of free scalar field


Real scalar field
Exercise. Consider a free real scalar field and write the expression of conserved quantities H, Pi
(componnets of 4moemntum) as function of a(k). Essentially do a Fourier transformation to
move to the momentum space.
CHAPTER 2. LAGRANGIAN & HAMILTONIAN FORMALISM 47

Solution: Recall teh general solution of the EL eq for a free real scalar field theory
d3 k
Z
1
ϕ(x) = √ (a(k) exp(−ikx) + a(∗ (k) exp(ikx))k0 =ωk
(2π)3/2 2ωk
The hamiltonian is Z
1 
~ 2 + m2 ϕ2

H= d3 x (∂0 ϕ)2 + (∇ϕ)
2
By substituing in H, one has
d3 x d3 kd3 p
Z Z
1
H= × (sum of terms)
(2π)3
p
2 4ωk ωp
where k relates to the first term and p to the second. Terms:
(−ωk )(−iωp ) ak ap exp(−i(k + p)x) + a∗k a∗p exp(−i(k + p)x) − ak a∗p exp(−i(k − p)x) − a∗k ap exp(−i(k − p)x)
 

p) ak ap exp(−i(k + p)x) + a∗k a∗p exp(i(k + p)x) − ak a∗p exp(−i(k − p)x) − a∗k ap exp(i(k − p)x)
(−~k)(−i~
 

m2 ak ap exp(−i(k + p)x) + a∗k a∗p exp(i(k + p)x) + ak a∗p exp(−i(k − p)x) + a∗k ap exp(i(k − p)x) k0 =ωk
 
p0 =ωp
Now we integrate, the result is
Z
1
H= d3 k ωk (a(k)a∗ (k) + a∗ (k)a(k))k0 =ωk
2
If we do the same steps for the free momentum we find
Z
1
Pi = d3 k ki (a(k)a∗ (k) + a∗ (k)a(k))k0 =ωk
2
it is a general result. If now we assume the fact that a(k), a∗ (k) are function, and so commute:
Z
H = d3 k ωk [a∗ (k)a(k)]
Z
Pi = d3 k ki (a∗ (k)a(k))

So we are in Hamiltonian representation.

Exercise 2
Consider a complex scalar field. 19/4/21
Solution: the solution is
d3 k
Z
1
ϕ(x) = √ (a(k) exp(−ikx) + b∗ (k) exp(ikx))k0 =ωk
(2π)3/2 2ωk
So we substitute a∗ = b∗ (?) of the above exercise results and we obtain
Z
1
H= d3 kωk [(a∗k ak + ak a∗k ) + (b∗k bk + bk b∗k )]
2
Z
q
Q(1) = d3 k[a∗k ak − b∗k bk ]
2
Assume the components commute, we find
Z
H = d3 kωk [a∗k ak + b∗k bk ]
Z
Q(1) = d3 k[(q)a∗k ak + (−q)b∗k bk ]

so Q is the sum of total particle charges and antiparticle charges of our system.
Recall two real scalar field with the same masses is equivalent with a complex sacalr field
with the same mass.
Chapter 3

Canonical quantization

We want to quantize the classical field theory. Till now fields were functions and classical objects.
We will face the canonical quantization: the most intuitive way to quantize a field theory. Let us
start with a preliminary discussion.

Discrete system
Consider a classical system with a finite number of dof. Recall we can describe the theory
in Hamiltonian formulation and summarise dynamic in term of Poisson bracket forming the
hamilton eq of motion:

q̇(t) = {q, H}t


ṗ(t) = {p, H}t

Recall the conjutate variables relations:

{q, p}t = 1 {q, q}t = 0 = {p, p}t

Note that these Poisson brackets are to be calculated at a fixed time.


Now, consider a quantum system with finite number of dof. We can quantize this system
requiring the following replacements:
1. From variables to operators:

(q, p) → (X, P )

where (q, p) are classical variables and (X, P ) are quantistic operators.

2. From Poisson brackets to commutator of operators:

{·, ·} → −i[·, ·]

These two "passages" are the canonical quantization. Using it, in a quantistic theory the Hamilton
eq of motion become the eq of motion of the quantistic operators:
dX
= −i[X, H]t
dt
dP
= −i[P, H]t
dt
notice these are the evolution eq of operators X and P in Heisenberg picture. The conjugate
variables relations becomes

[X, P ] = i [X, X] = 0 = [P, P ]

Notice we are using natural units with ~ = 1.

48
CHAPTER 3. CANONICAL QUANTIZATION 49

Field system
Now, canonical quantization of a field theroy proceeds along the same line: we start from a
classical field theory where Poisson brackets descibe the evolution of the field theory:

ϕ̇(x, t) = {ϕ, H}t π̇(x, t) = {π, H}t

these are the Hamilton eq for the field. The conjugate relations are

{ϕ, ϕ}t = 0 = {π, π}t {ϕ, π}t = δ 3 (x − y) = −{π, ϕ}t

The caninical quantization of a classical field theory comes taking the field and doing the following
replacements:

(ϕ, π) → (ϕ̂, π̂)

where (ϕ, π) are calssical fields and (ϕ̂, π̂) are quantistic fields operators. The Poisson brakets
are replaced as the same of a system with a finite dof

{·, ·} → −i[·, ·]

From the Poisson brakets, we obtain in QFT the following evolution of quantistic fields (namely
field of operators):
dϕ̂
= −i[ϕ̂, H]t
dt
dπ̂
= −i[π̂, H]t
dt
and conjugate relations are

[ϕ̂, π̂]t = iδ 3 (x − y) [ϕ̂, ϕ̂]t = 0 = [π̂, π̂]t (3.1)

Till now we used a redondant notation, in future we leave the hat, for example [ϕ̂, ϕ̂]t ≡ [ϕ, ϕ]t .
Recall computations are at fixed time t.

3.1 Canonical quantization of a real scalar field


We start from classical scalar fields, substitute functions with operators but maintain the form,
namely field and the conjugate read (we do not use more the hat)

d3 k h
Z
1 †
i
ϕ(x) = √ a(k) exp(−ikx) + a (k) exp(ikx)
(2π)3/2 2ωk k0 =ωk

1
Z 3
d k h i

π(x) = √ (−iω k ) a(k) exp(−ikx) − a (k) exp(ikx)
(2π)3/2 2ωk k0 =ωk

where ϕ(x), π(x) are operators in the coordinate space, therefore a(k), a† (k) must also be: they
are operators in momentum space. If we transform these in the coordinates space we have:

d3 x
Z
1
a(k) = √ [ω ϕ(x) + iπ(x)] exp(ikx)

k
(2π)3/2

2ωk k0 =ωk

1
Z 3
d x
a† (k) = √ [ωk ϕ(x) − iπ(x)] exp(−ikx)

(2π) 3/2 2ωk k0 =ωk

Recall (ϕ, π) satisfy the quantization condition (3.1), then we obtain the following quantization
conditions for a, a† :

[a(p), a† (k)] = iδ 3 (k − p) [a(p), a(k)] = 0 = [a† (p), a† (k)]


CHAPTER 3. CANONICAL QUANTIZATION 50

Do note for a system with discrete number of dof, we can rewrite the relations for any dof
[ap , a†k ] = δpk [ap , ak ] = 0 = [a†p , a†k ]
These are the relations of two uncopled armonical oscillator. So we understand that the
quantization condition for a real scalar field corresponds to describe the system as a continuos
number of free (or decoupled) harmonic oscillators. Recalling the quantum HO, operators
a† (k), a(k) are the creation and annihilation operators of momentum k. This similitude allow us
to extend what we know about quantum HO to canonical quantization field.

Number and number density operators Following what we know for harmonic oscillators,
we denote the number density operator
N (k) ≡ a† (k)a(k)
which tells how much oscillators there are in the system with momentum k. And the number
operator Z
N ≡ d3 kN (k)

which says the total number of oscillators of the system.


The number density satisfyes the properties:
N † (k) = N (k)
[N (k), a(p)] = −a(p)δ 3 (p − k)
[N (k), a† (p)] = +a† (p)δ 3 (p − k)
We can do same calcoulus for the number operator, it satisfies
N† = N
[N (k), a(p)] = −a(p)
[N (k), a† (p)] = +a† (p)
Another consistency comes from the fact that from these commutator relations we recover the
usual properies of the number operator given in QM. When we have commutator relations we
know all about the operator.

Hamiltonian operator Recall the result found in sec 2.6.2 about the formulation of the
hamiltonian (classical, not canonical quantizatized):
Z Z
1
3
H = d xH(x) = d3 k ωk (a(k)a∗ (k) + a∗ (k)a(k))k0 =ωk
2
To proceed with the canonical quantization we have to substitute functions with operators. We
can write better the hamiltonian using commutator rules:
Z   Z Z
3 1 † 3 1
H = d kωk N (k) + [a(k), a (k)] = d kωk N (k) + d3 kωk δ 3 (0)
2 2
The second term is proportional to ω0 which diverges, what to do with it? Firstly let us
understrand what it is in practice. Let us write this expression for a system with discrete dof:
X 1X
H= ωi Ni + ωi
2
i i

here the second term is not an operator, it Pis a constant, and since only differences of energy
have physical meaning (namely only ∆H = ωi Ni ), we can interpret the constant term as the
energy of the vacuum (namely the energy of the dof but withput HO, so with Ni = 0) and we
can take it to zero with no physical problems. In field, the term is infiite since we are summing
to infinite dof of energy, however we recognize it as the energy of the vacuum.
CHAPTER 3. CANONICAL QUANTIZATION 51

Moment operator Using tricks as for the hamiltonian, it is


Z Z Z
3 3 1
Pi = d xπ(x)∂i ϕ = d kki N (k) + d3 kki δ 3 (0)
2
namely we derived and we imposed the commutator relations. Again two terms: the second one
tells the free momentum of the vacuum state, it is no measurable so we can discart it.

3.1.1 Normal ordering (for bosons)


Note every time it comes out a constant vacuum state we can remove it automatically in our
theory since it is unphysical. Notice we are allowed to do it only in a flat space theory where we
are not link with the energy of the vacuum state. In order to do it we use not simple operator
but the so called normal ordered operators. For example for the energy it is something defined as
Z
N [H] ≡ d3 kωk N (k)

How to generalize this idea? We use the normal ordered product.

Definition. The normal ordered product between two or more operators is a product in which
alll the creation operators a† are on the left and all the annihilation operators a are on the right.
For example
N [a† (k)a(k) + a(k)a† (k)] = 2a† (k)a(k)
without using the commutator operator which gives the deltas and the infinities qunatities. So
normal order product approximates that inside the square bracket the operators commute.

Let us resume what we did. We introduced the procedure for canonical quantization of a real 21/4/21
free scalar theory: take the Lagrangin of the theory, the hamiltonian and impose the quantization
condition, namely fields become operators and poisson brackets become commutators. Doing so
we have a new theory with new objects: quantum field theory and we have to interpretate it. A
way to do it is to go to the momentum space. We did it and we realized that operators a, a†
behave equal to creation and destruption operators of harmonical oscillator. Now, what does it
mean? What are states and what particles in QFT?

3.2 Fock space


In QFT states live in a space called Fock space. How is it form? Let us discover it starting
from a real scalar field theory. We use the interpretation of our system as an infinte ensemble of
harmonic oscillators. So we assume that exists a state that is the vacuum state |0i, we define it
as
a(k) |0i = 0
It is equivalent to N (k) |0i = 0. Then we assume we create other states applying the creator
operator a† (k) on |0i:
|n1 , . . . , nm i ∝ (a† (k1 ))n1 . . . (a† (km ))nm |0i
This is a generalization of what we did in QM when we generated a space of HO states.

3.2.1 Identification
Now, Fock space is nice since we can match its states with particles. Let us start from the
vacuum state and apply the number operator N : by definition of vacuum space the result is

N |0i = 0
CHAPTER 3. CANONICAL QUANTIZATION 52

Since the hamiltonian and the momentum are proportional to the number operator we have

H |0i = 0
P~ |0i = 0

So we can say the vacuum state is a state with zero number of particle, and so also zero energy
and momentum.
We pass to the next state appling the creation operator in the momentum space:

|1(p)i ∝ a† (p) |0i

To understand which particle it represents let us apply the number, energy and momentum
operators:

N |1(p)i = +1 |1(p)i
H |1(p)i = ωp |1(p)i
P |1(p)i = p |1(p)i

(here k, p are momenta) namely this is the state of 1 particle with enenrgy ωp and free momentum
p. Notice here P = (P1 , P2 , P3 ).
We can generalized to a state |n(p)i:

|n(p)i ∝ (a† (p))n |0i

Here

N |n(p)i = +n |1(p)i
H |n(p)i = nωp |1(p)i
P |n(p)i = np |1(p)i

This is a state of n identical particle with energy ωp and momentum p.


Recalling the definition of state in Fock space, we understand how is composed a state of
particles.

3.2.2 Consistency with spin-statistic theorem


As done in statistic QM, we impose the commutator relation and we end up with a theory that
have the Bose-Einstein statistic. Let us see why: we found scalar fields can describe states
with identical particles and a system of identical particles must follow Bose-Einsteins statistic.
In addiction recall particles with spin zero obey to Bose-Einsten statistic. We discussed that
the scalar field has only externa angular momentum, so scalar field descibes spin 0 particle.
Everything is consist.

3.2.3 Normalization in Fock space


All the previous properties are not affected to the choice of normalization, however let us fix it.
Recall in HO we fixed the vacuum state as the state with zero energy, while in QFT we fix the
vacuum normalization as
h0|0i = 1
Doing the same, the easiest choice would be to follow the non-relativisic QM normalization of an
HO system: |1(p)i = a† (p) |0i. But there is a problem: this normalization is not covariant. So it
is preferable to choose a covariant normalization. It is not covariant because when we compute

h1(k)|1(p)i = h0| a(k)a† (p) |0i = δ 3 (k − p)

and three dimensional vector are not covariant.


CHAPTER 3. CANONICAL QUANTIZATION 53

A better choice of normalization is the following


|1(p)i ≡ (2π)3/2 2ωp a† (p) |0i
p

where the important thing is 2ωp . Now if we compute


p

h1(k)|1(p)i = (2π)3 2ωp δ 3 (k − p)


it looks three dimensional but it is Lorentz invariant (ωp is the zeroth component of k µ ).
Exercise. Show that (2π)3 2ωp δ 3 (k−p) is Lorentz invariant under a proper lorentz transformation
(recall solution of KG eq).

3.3 Identification of the field operators


Till now we discussed just in momentum space. Recall the solution of the KG eq, it is a field
operator
d3 k h
Z
1 †
i
ϕ(x) = √ a(k) exp(−ikx) + a (k) exp(ikx) = ϕ+ (x) + ϕ− (x)
(2π)3/2 2ωk k0 =ωk

where ϕ+ ∝ a(k) and ϕ− ∝ a† (k).


Consider a state with a particle with momentum p and apply the operator ϕ+ , compare it
with vacuum state:  
h0| ϕ+ (x) |1(p)i = exp(−ipx)
this means we create a state proportional to the vacuum state: the operator ϕ+ (x) destroies the
particle state of momentum p in the point x; and it is valid for any momentum.
Similarly  
h1(p)| ϕ− (x) |0i = exp(ipx)
so ϕ− (x) creates from the vacuum a particle state of any momentum p in the x position.
Notice ϕ+ ∝ a(k) so destroy and ϕ− ∝ a† (k) so creates. Namely a(k), a† (k) destroy and
create particles of momentum k, while ϕ+ , ϕ− do the same for any momentum (since there is an
integral). This is the particle interpatation of these operators in coordinates space.

3.4 Covariant commutators


It is an operation between field. Recall when we quantize the theory we imposed equal time
relation, eq (3.1):
[ϕ, π]t = iδ 3 (x − y) [ϕ, ϕ]t = 0 = [π, π]t (3.2)
But are these conditions at fixed time consistent with covariant? To understand it let us calculate
the commutator at two generic Minkowski points (at different times) and look if it is covariant.
If it is so, then also commutator at fixed time is covariant:
[ϕ(x), ϕ(y)] = [ϕ+ (x) + ϕ− (x), ϕ+ (y) + ϕ. (y)]
= [ϕ+ (x), ϕ− (y)] + [ϕ− (x), ϕ+ (y)]
= D+ (x − y) + D− (x − y) ≡ D(x − y)

where we recalled the proportial to a, a† of ϕ+ , ϕ− to vanish other terms. We can calculate the
objects expliciting the expression of ϕ+ , ϕ− finding
d3 kd3 p d3 k
Z Z
1 h

i 1
D+ (x − y) = exp(−ikx) exp(ipy) a(k), a (p) = √ exp(−ik(x − y))

(2π)3 (2π)3
p
4ωk ωp 2ωk k0 =ωk

d3 kd3 p d3 k
Z Z
1 h i 1
D− (x − y) = exp(ikx) exp(−ipy) a(k)† , a(p) = − √ exp(ik(x − y))

3 3
p
(2π) 4ωk ωp (2π) 2ωk k0 =ωk

= −D+ (y − x)
CHAPTER 3. CANONICAL QUANTIZATION 54

where we integrate the delta arises from the commutation a, a† , so an integral goes away. Notice
D+ (x − y) and D− (x − y) are functions in Fock space and not operators. Putting them together:
Z 3
−i d k
D(x − y) = sin k(x − y)

2π ωk k0 =ωk

At first view, expressions D± (x − y) look like not autmatically covariant. To demostrate they
are covariant we have to use complex analisys. We can write

d4 k exp(−ik(x − y))
Z
D+ (x − y) ≡ i 4
C+ (2π) k 2 − m2

We have to perform this integral in the complex space since our physical particle has k 2 = m2
and so there are holes along the integral path. Path C+ moves around ωk . In the same way

d4 k exp(−ik(x − y))
Z
D− (x − y) ≡ i 4
C− (2π) k 2 − m2

where the integral is around a path which contains −ωk .


So, even if we quantize the theory at a fixed time, commutators at generic time are consistent
with the covariance.

3.4.1 Covariance and microcausality


We know that [ϕ(x), ϕ(y)] is a covariant function by the covariance of D(x − y). In fact we know

[ϕ(x), ϕ(y)] = D(x − y)


[ϕ(~x, t), ϕ(~y , t)] = 0

where the last eq is at fixed time. However, when we calculate it at fixed time, then we get
back the equal-time commutator, which is equal to zero. Due tu covarinace it has to vanish for
any spacelike interval since any spacelike interval can be mapped onto another, and D must be
covariant:
D(x − y) = 0 ∀ (x − y)2 < 0
It is because the quantization condition. This property is called microcausality. The name comes
from the fact that our theory must preserve causality: the fields are not observable so it is
not strictly speaking causality, however all observers are products of fields, if microcausality is
satisfied then causality is satisfied.
For example if a particle is created at Minkowski point x and another at y, if they are
separated by a space-like distance, particles cannot interfear (recall commuator of observables
mean we can measure them symultaneously). Creation or annihilation of particles at point x, y
cannot communicate (be influenced) if (x − y)2 < 0.

3.5 Canonical quantization of a complex scalar field


We have two types of objects and when we go to momentum space there are a, b so it is more
complicated. Shortly let us identify the main complications. Canonical quantization proceed as
before but now we have a couple of conjugate variables. Recall the caninical quatization: at the
left we have classical quantities and to the right the quantistic one

(ϕ, π) (ϕ̂, π̂)


∗ ∗
(ϕ , π ) (ϕ̂† , π̂ † )
{·} − i[ , ]
CHAPTER 3. CANONICAL QUANTIZATION 55

The complete set of commutators of the theory, at equal time of commutation is:

[ϕ(~x, t), π(~y , t)] = iδ 3 (~x − ~y ) = [ϕ† (~x, t), π † (~y , t)]
[ϕ(~x, t), ϕ(~y , t)] = 0 = [ϕ† (~x, t), ϕ† (~y , t)] = . . .

Recall the general complex solution of KG eq:

d3 k h
Z
1 †
i
ϕ(x) = √ a(k) exp(−ikx) + b (k) exp(ikx)
(2π)3/2 2ωk k0 =ωk

double dof so two objects in momentum space: a, b. The commutation relations in momentum
space are:

[a(k), a† (p)] = δ 3 (~k − p~) = [b(k), b† (p)]


[a(k), a(p)] = 0 = [b(k), b(p)] = . . .

here we double the number of HO system. The particle interpretation of complex field is that we
are dealing with particles and antiparticles. In fact if we do all the definition as real scalar field,
we have double definitions: for number density and number operators

Na (k) ≡ a† (k)a(k) Nb (k) ≡ b† (k)b(k)


Z Z
Na = d3 kNa (k) Nb = d3 kNb (k)

Let us introduce
N = Na + Nb
Now, we can define hamiltonian and momentum of our free system. The hamiltonian operator
reads: Z
H = d3 kωk (Na (k) + Nb (k)) ≡ N [H]

where we are using the normalization removing the divergences.


To find more differences about particle and antiparticle, let us compute the U (1) conserved
charge. It comes out Z
Q(1) = d3 k(qNa (k) + (−q)Nb (k))

The only difference between particles a and b is the opposite charge.

3.5.1 Fock space


Now, we define the Fock space for complex scalar field: assume the existence of the vacuum
space such as a(k) |0i = b(k) |0i = 0, so

Na (k) |0i = Nb |bi = 0

Assume all particle states can be obtained by applying a† (k), b† (k) on the vacuum. For example
a generic state is obtained as

|n(p), n̄(k)i ∝ (a† (p))n (b† (k))n̄ |0i

The particle interpretation is trivial: recall |1(p)i ∝ a† (p) |0i and |1̄(p)i ∝ b† (p) |0i, so

N |1(p)i = 1 |1(p)i N |1̄(p)i = 1 |1̄(p)i


H |1(p)i = ωp |1(p)i H |1̄(p)i = ωp |1̄(p)i
Q |1(p)i = +q |1(p)i Q |1̄(p)i = −q |1̄(p)i

where we introduced the charge operator. Now we can say particle and antiparticle differ only by
the sign of the charge Q.
CHAPTER 3. CANONICAL QUANTIZATION 56

3.5.2 Covariant commutator


Last comment is for covariant commutators.We found we can generalize to Minkowski space any
commutator even if we start at same time. For example let us compute

[ϕ(x), ϕ† (y)] = [ϕ+ (x), ϕ†− (y)] + [ϕ− (x), ϕ†+ (y)] = D+ (x − y) + D− (x − y) = D(x − y)

We generalize the commutator relation saying it is invariant and this fact is called microcausality.
This conclude our discussion of the quantization of a scalar field. We descussed in detailed
the real filed case and we mentioned the difference from it to a complex scalar field.

3.6 Canonical quantization of a Dirac field theory


We will talk about the quantization of a free Dirac theory. Last time we discussed free scalar 26/4/21
field (real and complex). Now, we will see a Dirac field. We start from a classical theory, write
down the Lagrangian, pass to Hamiltonian, make the canonical quantization assumption: change
Poisson brackets and so on. We will see that something peculiar happens and we have to modify
our approach.
The starting point is the classical (free) dirac field theory. Arguments are similar to KG
theory, the difference is that Dirac eq is a linear eq and so we guess a Lagrangian that is linear in
derivative (while KG eq is quadratic and so the relative Lagrangian is quadratic in derivatives).
Recall the lagrangian must be a scalar, so the lagrangian density shoud be something like
i µ
L = ψ̄γ (∂µ ψ) − (∂µ ψ̄)γ µ ψ − mψ̄ψ

2
It is a scalar: invariant under Lorentz transformation. Notice the first piece is the kinetic one,
and the second is the mass one. This is not the usual way to write the Dirac lagrangian, but it is
the most symmetric.

Comment. Note the following Lagrangian density is equivalent:

L 0 = ψ̄(i~∂/ − m)ψ

Exercise. Prove the equivalence for exercise.

Let us talk about dimensions of terms in Lagrangian: recall [L ] = [L]4 and recalling the
dimensions of derivative it must be
[ψ] = [M ]3/2
Now, derive the EL eq for fields ψ, ψ̄:
∂L ∂L i i
− ∂µ = ~∂/ ψ − mψ + ~∂/ ψ = (i~∂/ − m)ψ = 0
∂ ψ̄ ∂∂µ ψ̄ 2 2
← ←
∂L ∂L i i
− ∂µ = ψ̄ ∂/ − mψ̄ + ψ̄ ~∂/ = −ψ̄(i ∂/ + m) = 0
∂ψ ∂∂µ ψ 2 2

Note the difference of the arrow up to ∂ since it was matrical calcolus: ψ̄ is a row vector.
We know the solutions of Dirac eq:
2
d3 k X
Z
1
ψ(x) = √ [cr (k)ur (k) exp(−ikx) + d∗r (k)vr (k) exp(ikx)]k0 =ωk
(2π)3/2 2ωk r=1

Dirac eq is a complex eq so two terms, and we have 2 dof from spin so we sum over r = 1, 2.
Here c and d are coefficients, u and v are unit vectors in spinor space.
We can use the Noether theorem to find the conserved charges.
CHAPTER 3. CANONICAL QUANTIZATION 57

Translation invariance
µ ∂L ∂L  ∼ i ψ̄γ µ ∂ ψ − (∂ ψ̄)γ µ ψ 
j(ν) = T̃νµ = + (∂ν ψ̄) L
ηνµ
− ν ν
∂∂µ ψ ∂∂µ ψ̄ 2
The lagrangian vanishing on the physical Dirac configuration, namely fields must respect Dirac
eq and so L = 0 and it is explicit looking at L 0 . Note T̃ µν is not symmetric in the exchange
µ ↔ ν. A symmetric object is the following

T̃ µν + T̃ νµ
Θµν ≡
2
and can be interpreted as the symmetric energy-momentum tensor of our theory from the
canonical energy-moemntum tensor. Notice there is no problem to use Θ if it derives from the
Noether T̃ .

Exercise. Show ∂µ T̃ µν = 0 = ∂ν T̃ µν from the eq of motion. Then show ∂µ Θµν = 0.

The conserved charges are



Z Z
i
Pµ ≡ d3 xT̃µ0 = d3 xψ † ∂µ ψ
2

Lorentz invariance
Recall how things change under an infinitesimal Lorentz transformation:
µ
x0µ = xµ + ωνµ xν δxµ = 12 ω ρσ Y(ρσ)
=⇒ µ
ψ 0 µ (x0µ ) = I − 2i ω ρσ Σρσ ψ(x) δϕ = 12 ω ρσ X(ρσ)


where
µ
= ηρµ ηνσ − ησµ ηνρ xν

Y(ρσ)
X(ρσ) = −iΣρσ ϕ

We can conjugateX to get an expression for the generators of the variation of the conjugate
spinor:
X̄(ρσ) = iψ̄Σρσ
Recall ρσ indicate the number of 4-currents we have. So the currents are
 
µ
j(ρσ) = xρ T̃σµ − xσ T̃ρµ + ψ̄γ µ Σρσ ψ

the last term is depends on how field transforms. The conserved charges are
Z Z
Qρσ ≡ d xj(ρσ) = d3 x (xρ Pσ − xσ Pρ ) + ψ̄γ µ Σρσ ψ = Lρσ + Sρσ
3 0
 

notice what is conserved is the total angular momentum. Indicies are Lorentz: the charge is an
antisimmetric tensor with 6 independent component, 3 boost and 3 rotation.
The fact that we can distinguish a regular angular momentum part as well as a spin part
means that Dirac Field is a spin 1/2 field.
CHAPTER 3. CANONICAL QUANTIZATION 58

Global U (1) invariance


It is an internal symmetry
x0µ = xµ δxµ = 0
=⇒
ϕ0 (x) = (1 + iα)ϕ(x) δ0 ϕ = iαϕ = X
The conserved current and charge are
µ ∂L ∂L
jν(1) ≡ Xψ + X̄ψ = ψ̄γ µ ψ
∂∂µ ϕ ∂∂µ ϕ̄
Z Z Z
Qν(1) ≡ d xjν(1) = d x ψ γ γ ψ = d3 x ψ † ψ
3 0 3 † 0 0

Notice dQ/dt = 0, that is the Dirac continuity eq.

3.6.1 Hamiltonian formulation


Now, we know everything about lagrangian description. Now we want the Hamiltonian description.
Recall we have complex field, so we have to define the conjugate of ψ and ψ † :
∂L i
π≡ = ψ†
∂∂0 ψ 2
∂L i
π† ≡ †
= ψ
∂∂0 ψ 2
then the hamiltonian density is
i ↔ i ↔
H ≡ π(∂0 ψ) + (∂0 ψ † )π † − L = − ψ̄γ i ∂i ψ + mψ̄ψ = ψ † ∂0 ψ
2 2
last step uses the Dirac equation (i∂/ − m)ψ = 0. The hamiltonian is given by
Z Z
H = d3 xH = i d3 x ψ † ∂0 ψ

It is not manifestly positive definite. Now, hamilton eq and Poisson brackets


δH
ψ̇α = = {ψα , H}t
δπα (x)
δH
π̇α = − = {πα , H}t
δψα (x)
The conjugate fields brackets reads
{ψα (~x, t), πβ (~y , t)}t = δαβ δ 3 (~x − ~y ) {ψα (~x, t), ψβ (~y , t)}t = 0 = {πα (~x, t), πβ (~y , t)}t
and similarly for ψ † , π † . Here αβ are spinorial indicies.
Do note that the Hamiltonian description is redundant: between the fields ψ, ψ † and π, π †
there are actually only two degrees of freedom.

3.6.2 Momentum space


Let us start from "classical" objects in the sense that they are not yet quantize. We are in
the momentum space to study relations between observables. The Dirac fields ψ, ψ † in the
momentum space read
2
d3 k X
Z
1
ψ(x) = √ [cr (k)ur (k) exp(−ikx) + d∗r (k)vr (k)) exp(ikx)]k0 =ωk
(2π)3/2 2ωk r=1
2
d3 k X h
Z
1 i
ψ † (x) = √ d r (k)v †
r (k) exp(−ikx) + c∗
r (k)u†
r (k)) exp(ikx)
(2π)3/2 2ωk r=1 k0 =ωk
CHAPTER 3. CANONICAL QUANTIZATION 59

The expression for cr (k) and dr (k) are:

d3 x h †
Z
1 i
cr (k) = √ ur (k)ψ(~x, t) exp(ikx)
(2π)3/2 2ωk k0 =ωk

1
Z 3
d x h i
dr (k) = √ ψ † (~x, t)vr (k) exp(ikx)
(2π)3/2 2ωk k0 =ωk

Recall what takes track of the spinoral are u, v and not the functions c, d.
Now we have to take expression of observables in function of c, d. The hamiltonian and the
charge read
Z 2
X
H= 3
d k ωk [c∗r (k)cr (k) − dr (k)d∗ (k)]
r=1

where ∗ stays for classical function, while it turns to † in a quantum field theory where functions
becomes operators. The energy becomes the sum over all momentum. The charge:
Z 2
X
Qν(1) = q d3 k [c∗r (k)cr (k) + dr (k)d∗ (k)]
r=1

Notice charges could be a positive quantity but energy could be negative and this is a problem!
However let us continue with the canonical quantization with commutators. Following the
canonical quantization ansatz we do the following substitutions:

(ψ, π) (ψ̂, π̂)


∗ ∗
(ψ , π ) → (ψ̂ † , π̂ † )
{·} −i[ , ]

namely we interpretate fields as quantistic operators. In this way the Hamilton eq of motion
becomes the Heisenberg eq:

ψ̇α = −i[ψα (~x, t), H]t


π̇α = −i[πα (~x, t), H]t

The quantization conditions (commutators between different fields and between fields them-
selves) read

[ψα (~x, t), πβ (~y , t)] = iδαβ δ 3 (~x − ~y ) [ψα (~x, t), ψβ (~y , t)]t = 0 = [πα (~x, t), πβ (~y , t)]t

and similarly for the conjugate fields. Then, since π ∝ ψ † we also have

[ψα (~x, t), ψβ† (~y , t)] = δαβ δ 3 (~x − ~y ) (3.3)

In the momentum space we obtain the quantization condition using the Fourier transform (to
find relations between c, d and ψ, ψ † ) and using the quantization condition on coordinate space
(since it informs us about commutators between ψ, ψ † ). The result is

[cr (k), c†s (p)] = δrs δ 3 (~k − p~) = −[dr (k), d†s (p)] [cr (k), cs (p)] = 0 = [dr (k), ds (p)] = . . .

notice it arises a bad minus from relation (3.3).


The Ladder commutative relations read
Z
Nc(r) (k) ≡ c†r (k)cr (k) Nc(r) = d3 xNc(r) (k)
Z
(r) (r) (r)
Nd (k) ≡ dr (k)d†r (k) Nd = d3 xNd (k)
CHAPTER 3. CANONICAL QUANTIZATION 60

Commutators are:

[Nc(r) , c(†) (†)


s (k)] = ±cs (k)δrs
(r)
[Nd , d(†) (†)
s (k)] = ∓ds (k)δrs

where the dagger in parentheses means that the relations hold both with it and without it. Do
note for the d-type operator the definition of N is opposite in sign to the c-type operator. Notice
the quantization we did till now shows some prossible problems arise from the different signs
that come out: we can say canonical quantization of the Dirac theory with commutators [ , ] is
inconsistency, namely it is wrong. Let us fix the inconsistencies arose:

1. The first inconsistency is related to energy: hamilonian H is not positive, for example if
we write the hamiltonian using the number operator
Z 2 h i
(r)
X
H= d3 kωk Nc(r) (k) − Nd (k) ≶ 0
r=1

Even if the conserved charges are positive, the problem of the energy is heavy.

2. The second inconsistency relates the spin-statistic theorem which is violated: the theorem
tells us we cannot have states with identical particles. While from our quantization we can
form a state with identical particle of momentum p with spin r:
h ik
|n(p)i ∝ c†r (p) |0i

We have to find another way to quantize the Dirac theory in order to solve these problems.

3.6.3 Canonical quantization with anticommutators


Last time we studied the Dirac lagrangian in field theory. We quantize our theory using canonical 27/4/21
quantization, as done the scalar theory. So classical fields become quantum operators and Poisson
brackets become commutators. However in momentum space we found stranger signs. Doing the
same step as done with scalar field we have two inconsistency: the hamiltonian is not positive
defined while the charge (that can be positive or negative) becomes only positive, and in Fock
space we can have states with identical particles even if it is experimentally wrong.
Now let us see if there is a Dirac consistent field theory. We are going to modify the canonical
quantization prescription for Dirac fields: change poisson brackets with anticommutator.
Notice in a classical theory we have two types of poisson brackets: the first regarding the
hamiltonian eq of motion
ϕ̇ = {ϕ, H} → ϕ̇ = −i[ϕ, H]
and we will do not touch it; the second type regards Poisson brakets between fields, we will
change it with the anticommutator

Poisson bracket {ϕ, π} → −i{ϕ, π} anticommutator

To the right we have anticommutator, it has the same symbol as Poisson brackets in classical
theory, but now we are in quantum theory and { , } denotes an anticommutator. So the new
qunatization procedure for Dirac field substitute Poisson brackets between fields with −i times
anticommutator.
The new quantization conditions are:

{ψα (~x, t), πp (~y , t)} = iδαβ δ 3 (~x − ~y ) {ψα (~x, t), ψβ (~y , t)} = 0 = {πα (~x, t), πβ (~y , t)}

that is equivalent to n o
ψα (~x, t), ψβ† (~y , t) = δαβ δ 3 (~x − ~y )
CHAPTER 3. CANONICAL QUANTIZATION 61

In the momentum space we have the following quantization conditions:


n o n o
cr (k), c†s (p) = iδrs δ 3 (~k − p~) = dr (k), d†s (p) {cr (k), cs (p)} = 0 = {dr (k), ds (p)} = . . .

Using these properties we will study the relations with the density number operator and number
operator:
Z
(r) †
Nc (k) ≡ cr (k)cr (k) Nc = Nc(r) (k)d3 k
(r)

Z
(r) †
Nc (k) ≡ dr (k)dr (k) Nc = Nc(r) (k)d3 k
(r)

Use the commutator rules, we find


n o
[Nc(r) (k), cs (p)] = c†r (k){cr (k), cs (p)} − c†r (k), cs (p) cr (k)
= −cs (k)δrs δ 3 (~k − p~)
[N (r) (k), c† (p)] = +c† (k)δrs δ 3 (~k − p~)
c s s

We computed explicitely the first eq to convince ourselves: the first term is zero by properties
and we substitute the value of the anticommutator. The second eq just follows from the first.
Notice there is an inverse sign in the eqs.
These relations are what defines the spectrum of number operator N , as in QM for HO, they
are the Ladder relations. They are the same as for scalar field where we used the commutators
even if now we deal with anticommutators.
Similar for d:
(r)
[Nd (k), ds (p)] = −ds (k)δrs δ 3 (~k − p~)
(r)
[Nd (k), d†s (p)] = d†s (k)δrs δ 3 (~k − p~)
Resuming, thanks to Ladder commutator relations we know everithing about the number operator
Nc , Nd . It comes out they are like the scalar field and so they have the good properties we are
looking for.
However let us verify explicely that problems are solved. The hamiltonian operator is
Z X2 h i
H = d3 k ωk c†r (k)cr (k) − dr (k)d†r (k)
r=1
Z 2 h i Z 2
(r)
X X
= 3
d k ωk Nc(r) (k) + Nd (k) + 3
d k ωk δ 3 (~k − p~) ≥ 0
r=1 r=1

We have to do a renormalization as before since the second term is an infinite constant term: we
interpretate it as the vacuum energy state and we set it to zero. The result is a positive define
hamiltonian.
The U (1) charge operator with anticommuator property becomes
Z 2 h
X i
Qν(1) = q d k 3
c†r (k)cr (k) + dr (k)d† (k)
r=1
Z 2 h i Z 2
(r)
X X
=q 3
d k Nc(r) (k) − Nd (k) +q d3 k ωk δ 3 (~k − p~) ≷ 0
r=1 r=1

Again the second term is an infinite and we identified it as the charge of the vacuum state and
we set it to zero with a renormalization.
As for scalar theory we defined a normal order product to remove automaticaly the infinite
terms from all the observable. However the normal order product in scalar field works with
commutators, here we introduced anticommutators, so now we have to count the number of
permutations since if it is even we have a plus sign while if it is odd we have a minus sign. So let
us implement the normal order product for fermions.
CHAPTER 3. CANONICAL QUANTIZATION 62

3.6.4 Normal ordering (for fermions)


The normal order product between fermionic operators is a product in which all creation operators
(c† , d† ) are on the left and all annihilation operators (c, d) are on the right (till now equal to
normal ordering for bosons) , but we have to take into account a global sign that depends on the
number of permutations np . For example

N [cc . . . c† c† ] = (−1)np c† c† . . . cc

So energy now is well defined. Now let us introduce the Fock space to see if the statistical
theorem is respected.

3.6.5 Fock space for fermion


We proceed as the HO. We assume it exists a vacuum state with the property that annihilations
operators give zero:

cr (k) |0i = 0 =⇒ Ncn (k) |0i = 0


dr (k) |0i = 0 =⇒ Ndn (k) |0i = 0

namely the vacuum state is the state with eingenvalue zero for the number operator.
The easiest state is 1-particle state

|1r (p)i ∝ c†r (p) |0i |1̄r (p)i ∝ d†r (p) |0i

To study its properties let us apply the hamiltonian and number operators, results are

Nc |1r (p)i = +1 |1r (p)i Nd |1̄r (p)i = +1 |1̄r (p)i


H |1r (p)i = ωp |1r (p)i H |1̄r (p)i = ωp |1̄r (p)i
Q |1r (p)i = q |1r (p)i Q |1̄r (p)i = −q |1̄r (p)i

these are the same relations as the scalar field, so we interpretate this state as a particle/antipar-
ticle with energy ωp and charge q/-q.
Now, let us verify the spin statistic theorem: if we try to built a state with two identical
particles
 2
|2r (p)i ∝ c†r (p) |0i

then we can write the operator with anticommutator and thanks to quantization condition:
 2 1n † o
|2r (p)i ∝ c†r (p) |0i = cr (p), c†r (p) |0i = 0
2
So in this Fock space we can’t build a state with identical particles and it follows from the
anticommutation imposition. Now, Dirac field theory obeys Fermi-Dirac statistic.
Do note in a relativistic field theory we do not have the freedom to choose in which way
quantize the theory if we want a consistent QFT: there is only one consistent way to choose
the quantization condition between ([, ], {, }) for a consistent relativistic QFT. If we try to
quantize the KG theory with anticommutators we will find a non positive defined hamiltonian
(inconsistency). Viceversa for the Dirac theory if we use commutators. The satisfaction of the
spin statisic theorem verify the right choice of the quantization procedure.

Comment (Schrödinger quantum field theory). It is the non-relativisic theory of KG and Dirac
theory. We will see we can quantize equivalently with commutators and anticommutators, so in
order to be consistent with the spin statictic theorem we have to choice the right quantization
procedure wrt if we are dealing with bosons or fermions.
CHAPTER 3. CANONICAL QUANTIZATION 63

3.6.6 Covariant anticommutator


Let us consider the following anticommutator computed on two spinorial fields at different points
+ −
  
Sαβ (x − y) ≡ ψα (x), ψ̄β (y) = ψ+ (x), ψ̄− (y) αβ + ψ− (x), ψ̄+ (y) αβ = Sαβ (x − y) − Sαβ (x − y)

where the lower αβ mean we take the alpha component of ψ+ and the beta components of ψ− .
Where
Z 3
+ 1 d k
Sαβ (x − y) = ( /
k + m) exp(−ik(x − y))

αβ
(2π)3 2ωk

k0 =ωk
Z 3
− 1 d k
Sαβ (x − y) = (k/ − m)αβ exp(ik(x − y))

(2π)3 2ωk k0 =ωk

These are functions componets of the 4dim matrix S. If indicies are not present we refers to
matrix and vectors.
With these exprections we can replace the (k/ + m) and (k/ − m) inside the integral and insert
the covariant anticommutator:

S+ (x − y) = (i∂/ + m)D+ (x − y)
S− (x − y) = (i∂/ + m)D− (x − y)

Since we already discuss covariant and microcausality for D+ , D− , then they are also valid for
S+ , S− .
The sum of the terms forms the condense formula

S(x − y) = (i
∂ + m)D(x − y)

These commutators are covariant and so Dirac theory has microcausality too (as KG theory).
Now we are able to quantize scalar (zero spin particle) and Dirac (1/2 spin particle) for free
theories. Next step is to quantize a 1 spin particle theory: namely a theory for a vector field.

3.7 Quantization of vector field


We will discuss relativistic free vector field theory in order to quantize particles with spin 1.
We want to build a classical relativistic field theory for a vector field V µ (x). Firstly, recall a
vector field is a field that transforms like a vector under Lorentz transformation:
µ
V 0 (x0 ) = Λµ ν V ν (x)

where Λ ∈ SO(1, 3).


Notice there is a problem: a vector field contains 4 dof that are one scalar and one 3dim
vector so V µ = (V 0 , V
~ ). A state of spin 1 (a massive boson) has only 3 dof: (1, 0, −1) component
of the angular momentum in a specific direction. This is a problem, we want to describe a
particle with spin 1 with 3 dof starting from an object that has 4 dof. There is another problem:
the photon (massless boson) has only two independent polarization, so a massless particle with
spin 1 has only (1, −1) polarization and not the 0. So even if we want to describe a massive or
massless boson starting with a nice Lorentz covariant object that is the vector, we start with an
object with a redundant number of dof. This will force us to complicate a little our relativistic
QFT.

3.8 Massive vector boson


We shall start with the massive theory since it is easier to start from 4 and go to 3 dof that to 2
dof.
CHAPTER 3. CANONICAL QUANTIZATION 64

The first task to solve is to find a nice lagrangian. Let us guess the lagrangian density: we
want a second order differential eq because of EM (photons from EM are massless bosons with
2 dof, we guess massive bosons come from a lagrangian similar to the EM one), and so the
lagrangian must have a second order derivative, so we want to use V µ (x) and partial derivatives
∂ µ V ν (x). The guess we can do is
1 1
L = − V µν Vµν + M 2 V µ Vµ
4 2
where
Vµν ≡ ∂µ Vν − ∂ν Vµ
This is the best guess for the vector field theory. is the field strenght. The first term contains
derivatives and it is the kinetic term, while the second is the mass term.
Let us derive the EL eq:
∂L ∂L
− ∂ρ = M 2 V σ + ∂ρ ∂ ρ V σ − ∂ σ ∂ρ V ρ = ( + M 2 )V σ − ∂ σ (∂ρ V ρ ) = 0
∂Vσ ∂∂ρ Vσ
This is the Proca equation.
Example (Calculous power). Before to talk about it, let us spell out one term of the EL eq just
for calculous knowledge, for example
 
∂ 1 2 µ ∂Vµ
M V Vµ = M 2 V µ = M 2 δµσ = M 2 V σ
∂Vσ 2 ∂Vσ
A second example is

(∂µ Vν ) = δρµ δσν
∂∂ρ Vσ

3.8.1 Proca equation


It seems a complicate eq, however it can be decompose in two equations. Let us derive the Proca
eq with respect ∂σ :
∂σ ( + M 2 )V σ − ∂ σ (∂ρ V ρ ) = ( + M 2 )∂σ V σ − (∂ρ V ρ ) = M 2 ∂σ V σ = 0 =⇒ ∂σ V σ = 0
 

where derivative and scalar commutes with derivative so ∂σ passes ( + M 2 ). We are developing
a massive theory where M 6= 0 so deriving the Proca eq we found a new condition. They are
together
( + M 2 )V σ = 0
∂σ V σ = 0
The first is the KG eq, when we will find its solution we have to impose the second eq that is a
constraint. All together are a different way to write the Proca equation.

3.8.2 General solution of Proca equation


We are describing a lagrangian theroy for a massive vector field. We guessed the Lagrangian form 3/5/21
hypotesis of hermitian, Lorentz invariant and so. Then we saw EL eqs are equivalent to the eqs
( + M 2 )V σ = 0
∂σ V σ = 0
Next step is to find a general solution of them. It is simple, we look solution of KG eq and then
we impose the constraint. Recall the solution for a real scalar field, we generalize it to a complex
field and from a scalar to a vector, namely
d4 k µ
Z
µ
V (x) = [f (k) exp(−ikx) + f µ∗ exp(+ikx)]
(2π)4
CHAPTER 3. CANONICAL QUANTIZATION 65

we can consider f, f ∗ as the positive and negative energy solutions. Then we impose that this is
a solution of the KG equation, so

d4 k
Z
0 = ( + M 2 )V µ = (−k 2 + M 2 )[f µ (k) exp(−ikx) + f µ∗ exp(+ikx)]
(2π)4

notice all mixed derivative terms cancel out. Following the scalar case we can define
3
√ X µ
f µ (k) = (2π)5/2 ωk δ(k 2 − M 2 ) λ (k)aλ (k)
λ=0

Since we are describing a 4dim vector, we introduced the vector µλ (k) called polarization; notice
there are 4 polarization, labeled by λ, they describe the independent degrees of freedom in
momentum space. We are summing over the 4 dof of the field.
The general solution reads:
3
d3 k X  µ
Z
1 ∗
µ
λ (k)ak (k) exp(−ikx) + µλ (k)a∗k (k) exp(+ikx) k0 =ω

V (x) = √
(2π)3/2 ωk
λ=0
k

To be solution of the Proca eq, we have to impose ∂µ V µ (x) = 0, namely

kµ µ(λ) (k) = 0

This constraint removes one of the degrees of freedom for the polarization vector, leaving three,
which is consistent with the fact that we have a massive spin-1 particle.
So for a spinor we have dof in spinorial space, for a vector the dof are described by the
polarization in momentum (and so coordinate) space

Exercise. Given k µ = (ωk , 0, 0, k), show that the following polarization vectors describe the
independent dof of Proca solution:
 
µ µ µ k ωk
(1) = (0, 1, 0, 0) (2) = (0, 0, 1, 0) (3) = , 0, 0,
M M
(λ)
Solution: show that k µ µ = 0

Exercise. Show that µλ satisfy the following properties: (where λ, λ0 = 1, 2, 3)

1. orthogonality relations:
0
µ(λ) (λ
µ
)
= −δλλ0 = ηλλ0

2. Completness relations:
3
kµ kν
 
µ(λ) ν(λ)
X
µν
=− η −
M2
λ=1

Solution: take k µ = (ωk , 0, 0, k) and with the previous exercise we can prove the properties.

Exercise. Obtain the explicit computation of the conserved currents and charges for the
translational invariant.

Conserved charges from Lorentz symmetry


The conserved charges associated to Lorentz transformation is an interesting computation to do
in order to understand the dof. Let us do it. Use an infinitesimal Lorentz transformation:
µ
x0µ = xµ + ωνµ xν δxµ = 12 ω ρσ Y(ρσ)
=⇒ µ
V 0 µ (x0µ ) = V µ (x) + ων V ν (x) δV µ = 12 ω ρσ X(ρσ)
CHAPTER 3. CANONICAL QUANTIZATION 66

where
µ
Y(ρσ) = −i(τ µν )ρσ xν
µ
X(ρσ) = −i(τ µν )ρσ Vν
(τ )µν µ ν µ ν

ρσ ≡ i ηρ ησ − ησ ηρ

where τ is the generator of the Lorentz group in the vectorial space (the equivalent of Σ for the
spinorial space).
The conserved currents read
µ
j(ρσ) = xρ T̃σµ − xσ T̃ρµ + (τνλ )(ρσ) V µν (x)V λ (x)

where recall V µν is the field strenght. The two first terms are common to all fields, while the
µ
third one is specific on how filed transforms (∝ X(ρσ) ).
The conserved charges are
Z
J(ρσ) = d3 x j(ρσ)
0
= L(ρσ) + S(ρσ)

Resuming what we found: scalar field has no angular monetum, in spinorial field we introduced
Σ and there is an internal angular momentum proportional to it, in vector field we introduced τ
and the intrinsic angular moentum is proportional to it.
To be sure we are describing a 1-spin vector, let us compute the Pauli-Lubanski vector.

Exercise. Calculate ωRF2 (in the rest frame) for the V µ (x).
Solution: notice ω 2 is a scalar so it is easy to compute it if we switch to the RF. Recall
µ
kRF = (M, 0, 0, 0). Recall the definition

1
ω µ = µνρσ (τ )νρ kσ
2
In the RF only the zeroth component is vanishing since there is µνρσ , so
0
ωRF =0
i M ijk
ωRF =  τjk
2
If we use the same spin definition as for Dirac field:

(ν) 1
Σi ≡ ijk τ jk
2
and so
i
ωRF = M Σi(ν)
It corresponds to the result of Dirac field apart from the mass, but here we have a massive vector
field. So the square is
   
0  2 0
2 i 2 2~ ~
 1
2
 ω  1 
ωRF = −(ωRF ) = −M Σ · Σ = −2M   =⇒ = −2  
1  M 2 RF  1 
1 1

where first row and column are for V 0 while the rest block of the matrix is for V i . Here 2 is the
eingenvalue of s2 and from QM: 2 = s(s + 1) that is s = 1 for V i .

We end here since we do not know how to quantize particle with spin 1.
CHAPTER 3. CANONICAL QUANTIZATION 67

3.8.3 Hamiltonian formalism


The dof of our vector field are the three spatial component (dynamical dof) and not the time
one. To convince ourself about it we switch to Hamiltonian formalism. Firstly we have to define
the conjugate field
∂L π0 = 0
π µ (x) ≡ = −V 0µ =⇒
∂∂0 Vµ π i = −V 0i
again V µν is the strength field. The result π 0 = 0 comes from direct computation of V 00 . Because
one of Hamiltons equations is ∂0 V0 = {π0 , H} we have that V 0 is a constant in time: V 0 (x) is
not a dynamical variable. One can take any value for it, and we will assume V 0 = 0. So we
throught out V 0 from the possible dof.
The hamiltonian density is

1 1 M2 2
H = π µ (∂0 Vµ ) − L = πi2 + Vij2 + V + V02 + ∂( ·

)
2 4 2 i
where the divergence is not important since when we compute the Hamiltonian we use Gauss
theorem and the divergence term goes to zero: no contribution, while V0 = 0 for us. So the
Hamiltonian is Z Z  
3 1 3 2 1 2 2 2
H = d xH = d x πi + Vij + M Vi ≥ 0
2 2
where i, j = 1, 2, 3. So we have 3 dof, the right number to describe an object with 3 polarizations.
Let us go to brackets relations: eqs of motion

V̇ i (~x, t) = V i (~x, t), H t




π̇ i (~x, t) = π i (~x, t), H t



 i
V (~x, t), π j (~y , t) t = 0 = π i (~x, t), π j (~y , t) r = V i (~x, t), V j (~y , t) r
 

notice we focus only on spatial components since we saw the time component is not a dynamical
variable.

Canonical quantization
Now we can use the canonial quantization ansatz and quantize the field. However there is a
problem: Poisson brackets have only ij terms, so for sure we can quantize the theory loosing
explicitely the covariant (since ij are not covarinat). So we can canonical quantize but we obtain
a theory that is not covarinat. We broke the covariance since V 0 = 0, and so π 0 = 0, namely we
are using only spatial (dynamical) dof and not the time one. This is the result of the fact we
start with a 4 dof object but then we lost one and we finish with 3 spatial dof.

3.9 Massless vector boson


Now, we wil describe a massless vector field theory, like the electromagnetic field. The lagrangian
of the theory is the Proca lagrangian (the one for massive vector field) setting the mass to zero:
1
LEM = − F µν Fµν
4
where now Fµν = ∂µ Aν − ∂ν Aµ is the strenght field.
The EL eq are
∂L ∂L
− ∂ρ = ∂ρ F ρσ = 0
∂Aσ ∂∂ρ Aσ

Opening the derivative: Aσ + ∂ σ ∂ρ Aρ = 0. They are the free Maxwell eqs.
CHAPTER 3. CANONICAL QUANTIZATION 68

By apply the partial derivative ∂σ , as done for the Proca eq, on Maxwell eqs we obtain an
identity:
(∂σ Aσ ) − (∂σ Aσ ) = 0
an identity does not give any additional condition. So in EM theory we have 4 dof and no
additional condition to decrease the dof. However our theory has the gauge invariance, thing not
present in the massive case.

3.9.1 Gauge invariance


In addition of the general Poincarre invarinace, our theory has an internal invarinace called
gauge invariance. It arises when we set M = 0 to a vector theory. A gauge transformation on
the vectro field is a transformation such that
µ
Aµ (x) → A0 (x) ≡ Aµ (x) + ∂ µ α(x)
where α is a function. It is called local (gauge) transformation since for any point the function
α(x) gives a differnet contribution. Notice it cannot be a global transformation since if α = 0
then ∂α = 0. It is a local U (1) group of transformation (number 1 states there is only one
independent parameter that is α).
The lagrangian density is invariant under this gauge transformation since in L enter only
field strenght and it transform under the local transformation as
µν ν µ
F µν (x) → F 0 (x) ≡ ∂ µ A0 (x) − ∂ ν A0 (x) = F µν + 
∂ µ
∂ να − 
∂ ν
∂ µα = F µν
 

then
0
LEM → LEM = LEM
So the massless vector field theory is more siymmetric thst the massive one since it has an
additional symmetry that is U (1).
Notice the SM of particle theory is a theory with a gauge group of symmetry GSM ≡
SU (3)c ⊗ SU (2)L ⊗ U (1)Y .
Since our EM theory is local invariance, we know that whatever we do is indepenendent over
a choice of a gauge fixing.

3.9.2 Fixing gauge


As the theory is invariant under U (1) local transformation of Aµ (x), we can choice a smart
formula for the field, it is not a problem since the physical results are independent on our choice.
For NR theory, usually it is used the Coulomb gauge:
~ ·A
∇ ~=0

notice it is a non-covariant choice: we will lost the covariance of the theory, it is the reason why
it is used for NR theories.
We want a R theory so we choice the Lorenz gauge
∂µ Aµ = 0
There are nice properties regarding it:
1. one can always find a gauge transformation that takes to the Lorenz gauge. Let us find
what is the formula for it: we apply a gauge transformation
µ
Aµ (x) → A0 (x) = Aµ (x) + ∂ µ α(x)
such that it takes us to the Lorenz gauge where
µ
∂µ A0 (x) = ∂µ Aµ + α(x) = 0 =⇒ α(x) = −∂µ Aµ (x)
So this is the kind of transformation we can always apply to reduce ourself to the Lorenz
gauge.
CHAPTER 3. CANONICAL QUANTIZATION 69

2. The Lorenz gauge does not completely fix the gauge freedom. To prove this statement take
another gauge transformation from a Lorenz gauge system A0 µ :
µ µ µ
A0 (x) → A00 (x) = A0 (x) + ∂ µ β(x)
To verify if we are again in the Lorenz gauge compute
µ µ
∂µ A00 (x) =  A0 + β(x)
∂µ


where ∂µ A00 µ (x) = 0 if β(x) = 0, this is a residual gauge coming from the Lorenz gauge.
Why all this is important? Consider Maxwell eqs in a generic gauge, they are
Aσ − ∂ σ (∂ρ Aρ ) = 0
Now we can fix the Lorenz gauge, we obtain
Aσ = 0
∂ρ Aρ = 0
They are the Proca eq with M = 0.

Hamiltonian formalism
To better understand the dof, let us go to Hamiltonian formalism:
∂L π0 = 0
π µ (x) ≡ = −F 0µ =⇒
∂∂0 Aµ π i = −F 0i
But, wait a minute, we did not take Coulomb gauge in order to preservare covariant, so we took
Lorenz gauge. We pass to Hamiltonian description and we find again the time component lost
its dof! if we write the commutators they will be functions only of ij spatial components, so if
we can quantize the theory we will lose the covariace. We have to try a difference quantization
to save covariance!
Resuming: we guessed the EM lagrangian and we found what we knew about em field theory 4/5/21
namely we obtained Maxwell eq of motion. We saw lagrangian has the nice property to be
invariant under local gauge symmetry U (1). We faced the problem to calculate observable in a
gauge theroy: we have to fix a gauge (the symmetry). The two typical choice are the Coulomb
gauge (typical for NR theory, it is non covarinat) and the Lorenz gauge (here eqs of motion are
covariant). We adopted the second one and the result was a system with a vector field that must
satisfy the massless KG eq and the Lorenz gauge fixing, they are the same of Proca eq with
M = 0. In Hamiltonian formutation things are the same (just massless M = 0), namely π 0 = 0
so it is not dynamical, while π i = −F 0i , spatial field is the only dynamical field. If we use the
canonical quantization procedure, we wil find it is a procedure non-covarinat since we have only
spatial component of the conjugate field.
We have to find an alternative way to have a covariant quantization. Fermi had the idea: the
EM lagrangian reproduce a good classical theory but cannot be canonical quantize in a covariant
way, so let us write another lagrangian asking at the end we recover em theory.

3.9.3 The gauge fixing Lagrangian


Consider the following density lagrangian:
1 1
L = LEM + LGF = − F µν Fµν − (∂µ Aµ )2
4 2ξ
where GF states for gauge fixing and ξ is a generic real parameter. The new piece at the classical
level vanishing when we go to Lorenz gauge and the theory is independent under the parameter,
so we recover EM theory at classical level (techinally what was added is a lagrangian multiplier).
Additing this term the lagrangian maintains the good properties of hermitian, real and so on.
However this lagrangian is not invariant under U (1) gauge transformation.
CHAPTER 3. CANONICAL QUANTIZATION 70

Exercise. Show that this lagrangian is not invariant under U (1) transform. And show that it is
equivalent to the lagrangian
 
0 1 1 ξ−1
L = − (∂µ Aν )(∂ A ) +
µ ν
(∂µ Aµ )2
2 2 ξ
namely they differ only by a total derivative.
Let us compute the EL eq from the equivalent L 0 :
 
∂L ∂L ξ−1 ρ
− ∂ρ = Aσ − ∂ (∂µ Aµ ) = 0
∂Aσ ∂∂ρ Aσ ξ
Since the ξ parameter is arbitrary, a simple choice for it is
ξ=1
this freedom to the parameter value choice is called Feynman gauge. In this gauge eqs of motion
are given by the massless KG eq:
Aσ = 0
However recall they are not the Maxwell eq of EM theory, they will be the same if we impose
the Lorenz gauge.
We already know the general solution of KG eq, they are the same of our new eq of motion
in the gauge ξ = 1:
3
d3 k X µ
Z
µ 1 


A (x) =  (k) aλ (k) exp(−ikx) + aλ (k) exp(ikx)
(2π)3/2 (2ωk )1/2 λ=0 (λ) k0 =ωk

where we do not have any constraint so we sum over all possible polarization: λ = 0, . . . 3.
For simplicity we can think about real field and so also real polarization vector µ(λ) and not
the general circular complex polarization (if it was complex polarization, then we should have
, ∗ ). So we are describing 4 independent dof via polarization. What could be a base of dof, a
polarization base vector? It can be chosen with the following properties:
1. basis vectors are orthogonal: orthogonality condition
0
µ(λ) (k)(λ )
µ (k) = η(λλ0 )

2. if we sum over all polarization, the result is "normalized": completness condition


3
µ(λ) (k)ν(λ) (k)η λλ = η µν
X

λ=0

We want a covariant way to write explicitely this polarization basis. A convenient choice is
vectors
k µ = (|~k|, ~k)
nµ = (1, 0)
with these vectors we can write the 4 independent polarization vectors:
µ(0) (k) = nµ scalar polarization

− (n · k)nµ
µ(3) (k) = longitudinal polarization
(n · k)
µ
(1) (k) = (0,~1 ) transverse polarization
µ(2) (k) = (0,~2 ) transverse polarization
where we choose transverse polarizations such that
~k · ~(1,2) = 0

[but this does not fix orthogonality from polarization 1 and 2?]
CHAPTER 3. CANONICAL QUANTIZATION 71

Exercise. Prove that these vectors satisfy orthogonality and completness relations.
Recall massive case have 3 dof of polarization: it corresponds to the case with µ(0) = 0. Now,
go on with the theory. Pass to hamiltonian formulation
Exercise. Show hamiltonian formulation for L and L 0 is the same.
So let take L 0 , with the gauge ξ = 1. Compute the conjugate momenta
µ ∂L 0
π 0 (x) ≡ = −∂0 Aµ
∂∂0 Aµ
We solved the problem of momenta conjugate since all terms π 0 µ 6= 0.
Write the hamiltonian density
µ 1h 2 2
i
H0 = π 0 (∂0 Aµ ) − L 0 = π 0 i + (∂i Aj )2 − π 0 0 − (∂i A0 )2
2
We solved the problem of the explicit covarinat since we include the zeroth dynamical component
(that is nonvanishing now). However notice that another problem arises, the hamiltonian density
has a negative contributions. If we compute the hamiltonian
Z Z
0 1 h
2 2
i
H = d xH =3
d3 x π 0 i + (∂i Aγ )2 − π 0 0 −)∂i A0 )2 ≷ 0
2
(notice the hamiltoninan is independent from the choice of the lagrangian, so there is no more
the prime, thanks to Gauss theorem)
With the hamiltonian we can pass to Poisson brackets:
Ȧµ (~x, t) = {Aµ (~x, t), H}t
µ  µ
π̇ 0 (~x, t) = π 0 (~x, t), H t

ν
A (~x, t), π 0 (~y , t) t = η µν δ 3 (~x, ~y )
 µ
 µ ν
{Aµ (~x, t), Aν (~y , t)}t = 0 = π 0 (~x, t), π 0 (~y , t) t

Exercise. Prove these relations.


Exercise. Prove these relations using L instead of L 0 .

3.9.4 Covariant quantization of EM field


So start to treat a new theory with a lagrangian L = LEM + LGF . We choose the Feynman
gauge condition: ξ = 1. For this theory components of momentum conjugate are different from
zero and so it covariant problem is not present (while it is present in the EM theory). So let
us quantize the theory using the canonical procedure: map functions to operators and Poisson
brackets to commutators. Then we have to verify the statistical principle and that the energy is
non-negative.
Let us quantize commutators:
ν
A (~x, t), π 0 (~y , t) t = iη µν δ 3 (~x, ~y )
 µ 
 µ ν
[Aµ (~x, t), Aν (~y , t)]t = 0 = π 0 (~x, t), π 0 (~y , t) t


We can rewrite the commutators between Aµ and π 0 ν in a simplest way recalling π 0 µ = −∂0 Aµ (x);
let us do it splitting time and space componets
 0
A (~x, t), ∂0 A0 (~y , t) t = −iδ 3 (~x, ~y )

 i
A (~x, t), ∂0 Ai (~y , t) t = +iδ 3 (~x, ~y )


since η µν = diag(−1, ~1). Notice if we think to our vector field as a collection of 4 scalar field,
each spatial component has the same sign than the scalar field, while the time component has the
inverse sign. So we expect time component of vector field satisfies opposite commuator relation
than the one we expecte for the scalar field. Again, the time component creates confusion. We
expect some troubles when we go to momentum space.
CHAPTER 3. CANONICAL QUANTIZATION 72

Commutator relations in momentum space


Firstly let us write the creation and annihilation operators in momentum space.
Exercise. Show that the creation and annihilation operators for the massless vector field, aλ , a†λ ,
can be written as follows; starting from aλ
−1 d3 x
Z
µ
a0 (k) =  (k)(ωk Aµ (x) + i∂0 Aµ (x)) exp(ikx)

(2π)3/2 (2ωk )1/2 (0) k0 =|~k|

1
Z 3
d x
ai (k) = µ (k)(ωk Aµ (x) + i∂i Aµ (x)) exp(ikx)

(2π) 3/2 (2ωk )1/2 (i) k0 =|~k|

and consequently for a† . Notice the sign difference between time and space polarization.
Now we can obtain the commutation relations for the canonical quantization condition in the
momentum space
h i
aλ (k), a†λ0 (p) = −ηλλ0 δ 3 (~k − p~)
h i
[aλ (k), aλ0 (p)] = 0 = a†λ (k), a†λ0 (p)

again notice the zeroth component polarization (λ = 0) has the wrong sign wrt the scalar case.
Let us define the number density operator and the number operator, we expect something
wrong for the zeroth component wrt spatial components, so let us introduce in advance a minus
sign in order to make things work:
Z

N0 (k) = −a0 (k)a0 (k) → N0 = d3 N0 (k)
Z
Ni (k) = +a†i (k)ai (k) → Ni = d3 N0 (k)

In this way the Ladder commutator relations are the ones expected:

[Nµ , a(†) (†)


µ (k)] = ∓aµ (k)

where we have the + if the dagger is present, - if it is not present.


Given the number operator we can define the hamiltonian using it:
Z Z 3
X
3 0
H= d xH = d3 k ωk Nλ (k)
λ=0

recall ωk = |~k|. While for the moment operator


Z Z 3
X
3
Pi = d x Pi = d3 k ki Nλ (k)
λ=0

where all operators are normal: H ≡ N [H] and Pi ≡ N [Pi ], so divergence is neglected.
To verify if the theory is physical we have to undertand if states could be interpretate as
particles, so let us give a look to the Fock space.

3.9.5 Fock space & Gupta-Bleur condition


It is the space of states. We assume the exixtence of the vacuum state and we define it such that

N (k) |0i = 0 aλ (k) |0i = 0

Let us start from the spatial polarization: λ = 1, 2, 3. Define one particle state:

|1i (k)i ∝ a†i (k) |0i


CHAPTER 3. CANONICAL QUANTIZATION 73

with i = 1, 2, 3. We obtain the same properties as the scalar field, so we have no problems for
the spatial polarization. Now, let us see the time component: one particle state is

|10 (k)i ∝ a†0 (k) |0i

but then we can see that the norm of this state is negative:

h10 (k)|10 (k)i ∝ h0| a0 (k)a†0 (k) |0i = h0| [a0 (k), a†0 (k)] |0i = −δ 3 (0) < 0

This is an unphysical property: λ = 0 state has a negative norm. This implies we deal with an
Hilbert space that has some states with positive norm and other with negative norm. This fact
is a complication if we want to identify states with particles.
Other problem is the negative eingenvalue for the energy: consider for example a 1-state

h10 (k)| N0 |10 (k)i = −1 =⇒ h10 (k)| H |10 (k)i = −ωk

So we were not able to canonical quantize the EM theory since covariance missed, we
introduced a "like"-EM theory adding LGF . However this new lagrangian reproduced the
massless KG eq in the Feynman gauge. We quantized the massless vector field in a covarinat
way but then we had problems with the zeroth polarization component. However this theory is
not yet the EM theory, it is a generic QFT. Recall to have the Maxwell eq as eq of motion we
must fix the Lorentz gauge.
We can do it brutaly just imposing ∂µ Aµ = 0, but in this way we break the covariance. Or
we can impose a soft condition: the expectation value of ∂µ Aµ = 0 is zero only for some states,
namely
hϕ| ∂µ Aµ |ϕi = 0
for a subset of |ϕi of the Fock Hilbert space.
What we mean is that the Hilbert space of our theory is larger than what we physical need: it
contains also zeroth negative norm states and three spatial states when we just want to describe
photons which need only two states. The idea to focus just on a subset of states is a smart choice.
This procedure is called Gupta-Bleur condition (GB).
Definition (Gupta-Bleur condition). The physical states |ϕphys i are the elements of the Fock
space that satisfy the condition

0
ϕphys ∂µ Aµ |ϕphys i = 0

namely the set of physical state is {|ϕphys i} ⊂ Fock space. We can write the GB condition in an
equivalent way:
∂µ Aµ+ (x) |ϕphys i = 0 ←→ hϕphys | ∂µ Aµ− = 0
We saw the GB condition in the coordinate space, let us see how it is in the momentum space
(namely in terms of creation and annihilation operators). We have to FT the operator
3
−i d3 x
Z
∂µ Aµ+ (x) µ
X
= k  (k)a (k) exp(−ikx)

µ (λ) λ
(2π)3/2 1/2 k0 =|~k|

(2ωk ) λ=0

So we identify the operator


3
µ(λ) (k)aλ (k)
X
L(k) ≡ kµ
λ=0
as the translation of ∂µ in the momentum space. So we can give the following definition of

the GB condition in the momentum space:
Definition (In momentum space). The physical states are the elements of the Fock space that
satisfy the condition
L(k) |ϕphys i = 0 ←→ hϕphys | L+ (k) = 0
CHAPTER 3. CANONICAL QUANTIZATION 74

Recall the polarization basis and use it to write a new formula for L(k):
  µ  
µ k µ
L(k) = kµ n a0 + − n a3 = (n · k)[a0 (k) − a3 (k)]
n·k

notice transverse polarizations (a1,2 ) disappear because of ortoghonality. Now we start to see the
physics beyond the math: the GB condition tells us that

L(k) |ϕphys i = 0 =⇒ a0 (k) |ϕphys i = a3 |ϕphys i (3.4)

Let us see what this means (namely the GB condition) in terms of actual measurable properties
of these physical states:

1. Compute the expectation value of the number of states with zero and three polarization in
a physical state:

L(k)
hϕphys | N0 +N3 |ϕphys i = hϕphys | a†3 (k)a3 (k)−a†0 a0 (k) |ϕphys i = − hϕphys | a†3 (k) |ϕphys i = 0
n·k
where we used eq (3.4). So the sum of the unphysical polarization state in the physical
states is zero. Very good!

2. Compute the total number of phyisical states


3
X
hϕphys | Nλ |ϕphys i = hϕphys | N1 (k) + N2 (k) |ϕphys i = n1 (k) + n2 (k)
λ=0

where ni (k) indicates the numebr of states with moment k and polarization i = 1, 2.

3. Compute the hamiltonian


Z 3
X Z
hϕphys | H |ϕphys i = d3 k ωk hϕphys | Nλ |ϕphys i = d3 k ωk (n1 (k) + n2 (k)) > 0
λ=0

so with physical states we have positive defined hamiltonian.

So, if we require the Gupta-Blauer condition the observables do not depend on the 0 and 3
polarizations; so we can only consider the two transverse ones and be left with a theory that
now corresponds to a true quantization of classical electromagnetism, since we have the three
conditions

1. Aµ = 0;

2. ∂µ Aµ = 0 for physical states;

3. two polarizations.

Resuming: we added a piece to the EM lagrangian in order to be able to quantize covariantly 10/5/21
the new theory (notice this piece breaks the U (1) symmetry). With this modification this picture
work in the Hamiltonian formalism. Then we started to quantize. We saw there are some
problems with the quatization: the "wrong sign" with the zeroth component that translated to
negative norm for zeroth state and negative energy expectation value. However we have a theory
larger than the EM theroy since we have 4 dof. To go back to EM we impose the Feynman gauge
finding the massless KG eq we "impose" the Lorentz gauge via GB condition. In this way we
recover a physical interpretation for the states in the Fock space.
CHAPTER 3. CANONICAL QUANTIZATION 75

3.9.6 Pseudo-photon state


We started to study the properties of physical states. Now focus to a particular class of physical
state: the pseudo-photon state. It is the state obtained from a physical state by appling L†
operator, for example applying it to vacuum state, or to a general state:

|ψ1 i ∝ L† |0i |ψi ∝ L† |ϕphys i

which are their properties?

1. Pseudo-photon states has zero norm:

hψ|ψi = 0

2. it is ortoghonl to |ϕphys i
hψ|ϕphys i = 0

3. Zero energy/momentum
hψ| H |ψi = 0

Due to these properties, any state that is obtained by a linear combination of a physical state
plus a pseudo-photon state is indistinguishable from the physical state, namely
0
phys = |ϕphys i + |ψi ' |ϕphys i
ϕ

in fact due to pseudo-photon state properties: the normalization is the same



0
ϕ phys ϕ0 phys = hϕphys |ϕphys i

and also energy is the same



0 0
ϕphys H ϕphys = hϕphys | H |ϕphys i

So we can define equivalence class: we define it as the ensemble of physical states obtained
appling a pseudo-photon state to a starting physical state:

{|ϕT i} = {|ϕphys i = |ϕT i + |ψi}

where |ϕT i ∝ (a†i )n (a†2 )m |0i.


The birth of equivalence class is related to the residual gauge: the Lorentz gauge does not fix
complitely the dof of the theory, the result is taht we can have an infinite amout of gauge choices
that are still Lorentz gauge. This translates to the possibility to have an equivalence classes.

3.9.7 Covariant commutators


Recall we use the Feynman gauge ξ = 1. The covariant commutators are

Dµν (x − y) ≡ [Aµ (x), Aν (y)]


= A+ µ (x), Aν− (y) + Aµ− (x), Aν+ (y)
   
µν µν
= D+ (x − y) − D− (x − y)

One can show

Aµ+ (x), Aν− (y) = −η µν D+ (x − y)


 
 µ
A− (x), Aν+ (y) = −η µν D− (x − y)


so their sum is
Dµν (x − y) = −η µν D(x − y)
CHAPTER 3. CANONICAL QUANTIZATION 76

and all the good results from the scalar case generalize.
Resuming what we did till now: we quantize scalar real and complex fields, Dirac field (where
we cannot use commutator but anticommutator), vector field.
We can quantize electromagnetism in a non-covariant way in either the Coulomb or Lorentz
gauge. In order to quantize it covariantly we need a different theory: we quantize a theory with
an additional gauge-fixing term in the Lagrangian. This theorys Fock space includes unphysical
states, with negative norm and negative energy. If we select a subset of the Fock space with
the Gupta-Bleuer condition we recover electromagnetism. Different choices for the gauge-fixing
Lagrangian give rise to different gauge choices. One may ask: are the observables actually
independent of the choice of ξ in the gauge-fixing Lagrangian? In fact they are, but this is not
trivial to show; it depends on the Ward identities.
Chapter 4

Interacting field theories

We will see interaction in classical and quantum field theory. Recall we have scalar field, Dirac
field and vector field. Which type of interacting lagragian we can write with these fields? And
how to discriminate between interaction?

4.1 Types of interaction


4.1.1 Scalar field self-interaction
Let us start from the real scalar field with self interaction. Which kind of interaction can we
introduce with this field? We can write the lagrangian of such a system as
1 1
L = (∂µ ϕ)(∂ µ ϕ) − m2 ϕ2 − V (ϕ) = L0 + Lint
2 2
where we introduced all possible self-interaction into Lint , it is something like

V (ϕ) = kϕ3 + λϕ4 + µϕ5 . . .

where the power of the term indicates the vertex of the interaction. How to discriminates the
interaction? From the dimension of the couplings. Let us study them. Recall in natural units
the action is adimensional and so

[L ] = M 4 ←→ [ϕ] = M 1

and so

[k] = M 2
[λ] = M 0
[µ] = M −1
...

Definition. A theory is callled renormalizable if it contains only coupling g that have a mass
dimension that is non-negative: [g] = M α with α ≥ 0. For example in our case, the theory is
renormalizable if it contains only kϕ3 + λϕ4 .

For now, only renormalizable theory can be a theory consistence for any range of energy,
overwise there is an energy at which non-physical predictions come out.

Exercise. Describe the renormalizable interaction for a complex scalar field.

77
CHAPTER 4. INTERACTING FIELD THEORIES 78

4.1.2 Dirac field self-interaction


Recall the free lagrangian and let us add the self-interaction term
L = ψ̄ i∂/ − m ψ + GF ψ̄Γψ ψ̄Γψ = L0 + Lint
  

recall we alway must have ψ̄ on the left and ψ on the right such that spinor indicies contract.
This type of lagrangian was firstly introduced by Fermi to describe the beta (neutron) decay:
the Fermi lagrangian is
GF
Lint = √ ψ̄p γ µ PL ψn ψ̄e γµ PL ψνe
 
2
where the first term is a barion current and the other one is the lepton current. What is the
dimension of the Fermi constant? Recall [ψ] = M 3/2 and so [GF ] = M −2 . Using our definition,
the Fermi interaction is not renormalizable, in fact we find trouble if we want describe the world
using Fermi theory, however at his time its theory worked.

4.1.3 Scalar and Dirac field interactions


No more only self-interaction. Consider a real scalar field for simplicity, the lagrangian is
1 1
L = (∂µ ϕ)(∂ µ ϕ) − m2 ϕ2 + ψ̄ i∂/ − m ψ + yS ϕψ̄ψ + yP ϕψ̄γ5 ψ + ω1 ϕ2 ψ̄ψ + . . .

2 2
as always first terms are the one of free fields. Then there are coupling terms, y states for
Yukawa. ys is a scalar, yp is a pseudoscalar (because of the presence of γ5 , it changes under
parity transformation.).
Given a lagrangian, we must do the count of the dimensions of coupling to see if the theory
is renormalizable:
[yS ] = [yP ] = M 0
[ω1 ] = M −1
So only Yukawa interactions are renormalizable. This is nice since we can build interaction
between Higgs field and fermions in SM.

4.1.4 Vector and Dirac field interactions


The lagrangian is
1
L = − F µν Fµν + ψ̄ i∂/ − m ψ + gV (ψ̄γ µ ψ)Aµ + gA (ψ̄γ µ γ5 ψ)Aµ

4
+ cV (ψ̄ψ)Aµ Aµ + icA (ψ̄γ5 ψ)Aµ Aµ + dV (ψ̄σ µν ψ)Fµν + idA (ψ̄σ µν γ5 ψ)Fµν . . .
where index V indicates vector interaction (interaction between fermions and photons, it is QED),
while index A idicates axial iteraction (since ψ̄γ µ γ5 ψ transforms like an axial).
What are the dimension of the coupling?
[gV ] = [gA ] = M 0
[cV ] = [cA ] = [dV ] = [dA ] = M −1
So only gV , gA coupling lead to renormalizable theory. Recall vector and axial can be written
using left and right (just notation: vector and axial or left and right):
gR = gV + gA g = (gR + gL )/2
←→ V
gL = gV − gA gA = (gR − gL )/2
So renormalizable interaction can be written as
gV (ψ̄γ µ ψ)Aµ + gA (ψ̄γ µ γ5 ψ)Aµ = gL (ψ̄γ µ PL ψ)Aµ + gR (ψ̄γ µ PR γ5 ψ)Aµ
We say vector-spinor interaction is vector-like when gA = 0, namely gL = gR (like QED), while
we say vector-spinor interaction is chiral when gA 6= 0, namely gL 6= gR (like Weak interaction).
CHAPTER 4. INTERACTING FIELD THEORIES 79

Exercise (QED and gauge invariance). Consider the Lagrangian:


1
LQED = − F µν Fµν + ψ̄ i∂/ − m ψ − q(ψ̄γ µ ψ)Aµ

4
Show the QED lagrangian has the following U (1) gauge symmetry:
µ
Aµ (x) → A0 (x) = Aµ (x) + ∂µ α(x)

ψ(x) → ψ 0 (x) = exp[iqα(x)]ψ(x)


notice it is a local phase transformation: α = α(x).

So we can build many interaction between fields, but if we want renormalizable interaction,
then we reduce a lot the possibilities of interaction. We impose renormalization since we want to
introduce interactions that are realizable in Nature.

4.2 Canonical quantization of interacting scalar field


We saw how to introduce interaction in a lagrangian classical field theory for particles with spin 11/5/21
0, 1/2 and 1. Now we want to quantize a field theory with interaction. We will see which are the
main obstacles with an example: quantize scalar field. Then we will see how to exceed them.
Consider a real self-interacting scalar field with interacting potential V (ϕ) = 4! ϕ , so the
λ 4

lagrangian is
1 1 λ
L = (∂µ ϕ)(∂ µ ϕ) − m2 ϕ2 − ϕ4 = L0 + Lint
2 2 4!
Let us go over the standard procedure to quantize a scalar field theory. At the end we will see
problems. First of all we want the EL eq of motion, doing the computation they are
λ
 + m2 ϕ = − ϕ3

3!
the right-hand side is the first derivative of the potential. We skip the Noether quantities and we
go to the hamiltonian description. The conjugate field is
∂L
π≡ = ∂0 ϕ
∂∂0 ϕ
and
λ 4
H = π(∂0 ϕ) − L = H0 + ϕ = H0 + Hint
4!
so H0 is the hamiltonian density we found for a free real scalar field, now we have also a part
regarding the interaction which describes the evolution of the interaction. Now the Poisson
brackets:

ϕ̇(~x, t) = {ϕ(~x, t), H}


π̇(~x, t) = {π(~x, t), H}
{ϕ(~x, t), π(~y , t)} = δ 3 (~x − ~y )
{ϕ(~x, t), ϕ(~y , t)} = 0 = {π(~x, t), π(~y , t)}

Now, let us apply the canonical quantization ansatz:

[ϕ(~x, t), π(~y , t)] = iδ 3 (~x − ~y )


[ϕ(~x, t), ϕ(~y , t)] = 0 = [π(~x, t), π(~y , t)]

and they are the same conditions as for the free scalar real field theory.
CHAPTER 4. INTERACTING FIELD THEORIES 80

Problem to find a solution


Next step is to find an explicit solution of the eq of motion in terms of operators (creation and
annihilation). The free KG solution is no more valid since the eq of motion has the interaction
term. In general, interacting theory has no (close) solution, namely the field cannot be written
in terms of operators like

ϕ ∼ a(k) exp(−ikx) + a† (k) exp(ikx)

this implies no link with HO and the Fock space interpretation. In free theory, appling a† we
find eigenstate of hamiltonian and we link them to particles. Here we are not able to find the
eigenbasis of the Hamiltonian and the energy associated to a state, so we are not able to link it
to particle: we lost the particle interpretation.
When we are not able to finde a close solution, we have to find an approximate one. How
to do? Perturbative theory. We will expand the theory with a parameter, and we compare
the expansion with experimental results that have error band. Experimental results will tell us
how much we have to expand the free solution. If the experimental collider is better than the
theoretic approximation, we have to improve an higer order of precision (expansion).
Notice not for all interaction we are able to know a good expansion parametr (with em and
weak interactions we can, for strong interaction at low energy we are not able since the parameter
becomes bigger than 1).

4.3 Interacting picture for perturbative theory


To proceed we need a new picture: the interaction picture (picture like Heisenberg and
Schrödinger). Recall
1. In Schrödinger picture states evolve while operators not.

|ψ(t)iS = U (t, t0 ) |ψ(t0 )iS XS (t) = XS (t0 )

2. In Heisenberg picture states do not evolve while operators do (it is the opposite).

|ψ(t)iH = |ψ(t0 )iH XH (t) = U † (t, t0 )XH (t0 )U (t, t0 )

However the evolution operator is the one satsfing the following differential eq
d
i U (t, t0 ) = HU (t, t0 )
dt
with initial condition U (t0 , t0 ) = I. the solution of the diff eq is the evolution operator

U (t, t0 ) = exp[−iH(t − t0 )]

Now, let us introduce the interaction picture: it is in the middle of Schrödinger and Heisenberg
pictures, namely both states and operatos evolve. Let us write the hamiltonian of our interactiong
system:
H = H0 + Hint
We can write two evolution operator: the free evolution operator and the full evolution operator

free U0 (t, t0 ) = exp[−iH0 (t − t0 )]


full U (t, t0 ) = exp[−iH(t − t0 )]

The interaction picture (wrt the Schrödinger one) is defined as the picture where states and
operators evolve as follow:

|ψ(t)iI = U0† (t, 0) |ψ(t)iS


XI (t) = U0† (t, 0)XS U0 (t, 0)
CHAPTER 4. INTERACTING FIELD THEORIES 81

Notice here t = 0 6= t0 . t0 is just the initial time we decided to evolve the system, while t = 0
is the time when we decided to compare the interaction and Schrödinger picture. We have
arbitrarity to choose this time and so we choose 0. Notice we wrote XS without dependece on t
because in Schrödinge operators do not evolve.

Evolution of operators in interaction picture


What is the meaning of these definitions? Let us study the evolution of operators. Let us
differentiate it:
d d † 
i XI (t) = i U0 (t, 0)XS U0 (t, 0) = U0† (t, 0)[XS , H0 ]U0 (t, 0) = [XI (t), H0 ]
dt dt
last passage is valid because U0 commutes with H0 . We found the eq of motion for our interacting
operator XI . It evolves like a non-interaction theory in the Heisenberg picture, since it works
with H0 . This is the first step to understand why to pass to interacting field: things valid for
free theory are valid also for interacting theory, namely filed operators are the solutions of the
free KG eq, so in the interaction picture

ϕI (x) ∼ a(k) exp(−ikx) + a† (k) exp(ikx)

and it is a magnificient reason to use interaction picture.


Also note that the expectation values on the three pictures are the same, and this is a must
for a picture!

Evolution of states in interaction picture


How does evolve our state? Let us take the definition of the state in the interaction picture and
differentiate it:
d d † 
i |ψ(t)iI = i U0 (t, 0)U (t, t0 ) |ψ(t0 )iS = U0† (t, 0)Hint U0 (t, 0) |ψ(t)iI
dt dt
where we substitute also the definition of the state in the Schrödinger picture. Here there are H
and H0 and this is a difference wrt to free theory. What is U0† (t, 0)Hint U0 (t, 0)? By definition, it
is the interacting hamiltonian in interacting picture:
I
Hint (t) ≡ U0† (t, 0)Hint U0 (t, 0)

So the state in interaction picture solves the differential eq


d I
i |ψ(t)iI = Hint (t) |ψ(t)iI
dt
This look like a Schrödinger eq, but with an hamiltonian that depends on time. Now commuting
objects may lost this property, for example [HintI , H ] 6= 0. We can write the evolution of the
0
state using the evolution operator in the interaction picture:

|ψ(t)iI = U0† (t, 0)U (t, t0 ) |ψ(t0 )iS = U0† (t, 0)U (t, t0 )U0 (t, 0) |ψ(t0 )iI = UI (t, t0 ) |ψ(t0 )iI

so the evolution operator is

UI (t, t0 ) ≡ U0† (t, 0)U (t, t0 )U0 (t, 0)

and by definition it is an operator in the interaction picture.


CHAPTER 4. INTERACTING FIELD THEORIES 82

Differential equation for evolution operator


Now we can ask: which is the differential eq for the evolution of the evolution operator in the
interaction picture? Recall the differential eq for operators and consider UI :
d
i UI (t, t0 ) = U0† (t, 0)(H − H0 )U (t, t0 )U0 (t0 , 0)
dt
= U0† (t, 0)(H − H0 )(U0 U0† )U (t, t0 )U0 (t0 , 0) = Hint
I
(t)UI (t, t0 )

where we multiply by I = (U0 U0† ) and we recognized H − H0 = Hint in the interaction picture,
namely U0† Hint U0 = Hint
I , and also the evolution operator in the interacting picture that is

U0 U U0 = UI .
Now we have to solve this differential eq. However we have two problem:
1. the hamiltonian is not a constant but depends on time,

2. the hamiltonian is an operator that does not commute with the free and full hamiltonian
at different time.
The easiest way to deal with such differential eqs is to write them as an integral eq:
Z t
I
UI (t, t0 ) = I − i dτ Hint (τ )UI (τ, t0 )
t0

notice the above integral eq is equivanet to the differential one, just take its derivative in time.
This is the easiest expression to apply perturbative theory.
Recall we want to describe an interacting field theory but we are not able to solve the diff eq.
So we introduced an interaction picture (a mixed picture), it is useful since operators evolve as
in free theory, so what we knew about free theory is still valid (namely the explicit expression of
the operator field). But states evolve in a complicate way throught the interaction hamiltonian
(it depends on time and it does not commute with anything else). The evolutive operator in
interaction picture turns out to solve a complicate diff eq (and so integral) too. We well use
perturbative tecniques to solve it. Let us assume that the interaction is formed by a small
parameter λ. We look to "expanded" solutions such that can be written as

X
UI (t, t0 ) ≡ Cn (t, t0 )
n=0

where n gives the order of expansion: O(λn ). If λ is small, we can approximate the complete
solution up to order N (generic):
N
 X
UI (t, t0 ) = UIN (t, t0 ) + O λN +1 = Cn (t, t0 ) + O λN +1

n=0

With N we control the theoretical error we do wrt the complete solution.


Let us plug in the integral eq the approximate solution:
Z t
N I
UI (t, t0 ) = I − i dτ Hint (τ )UIN (τ, t0 )
t0

on the left side th object is the solution up to order N , while on the right side we have solution
with order λN and an interaction hamiltonian of first order in λ, so if we want solution up to
order N we can write Z t
UIN (t, t0 ) = I − i I
dτ Hint (τ )UIN −1 (τ, t0 )
t0

it is an iterative procedure: once we know UIN −1 we can find UIN : starting to order 0 we can
reach any order.
CHAPTER 4. INTERACTING FIELD THEORIES 83

Zeroth order
(0)
Let us start from zeroth order: UI . It is when we put to zero the interaction: λ = 0. Then the
interaction hamiltonian goes to zero and the integral eq states
UI0 (t, t0 ) = I = C0 (t, t0 )
If we have no interaction we have no evolution: zeroth order approximation is the free theory. In
fact
(0)
|ψ(t)iI = I |ψ(t0 )iI
d
i XI (t) = [XI (t), H0 ]
dt
this is the Heisenberg evolution for a free theory, this is a consistency check.

First order & second order


(1)
Let us pass to order 1: UI . As above
Z t
1 I
UI (t, t0 ) = I − i dτ Hint (τ )I = C0 (t, t0 ) + C1 (t, t0 )
t0
where Z t
I
C1 (t, t0 ) = −i dτ Hint (τ )I
t0
(2)
Iterating also for order 2: UI .
Z t Z t Z t Z τ1
(1)
UI2 (t, t0 ) = I − i I
dτ Hint (τ )UI (t, t0 ) = I − i I
dτ Hint (τ ) + (−i)2 dτ1 I
dτ2 Hint I
(τ1 )Hint (τ2 )
t0 t0 t0 t0
where Z t Z τ1
2 I I
C2 (t, t0 ) = (−i) dτ1 dτ2 Hint (τ1 )Hint (τ2 )
t0 t0
Be carefull that interaction hamiltonian does not commute evaluated at different times.

Generalization to Nth order


Generalize the solution to N th order: UIN .
N
(N )
X
UI = Cn (t, t0 )
n=0
where Z t Z τ1 Z τn−1
Cn (t, t0 ) = (−i)n dτ1 dτ2 · · · I
dτn Hint I
(τ1 ) . . . Hint (τn )
t0 t0 t0
where the product of n interaction hamiltonians are a term of order λn . The exact solution is
recovered formaly with the sum of n from 0 to ∞.
By algebraic manipulations we can write
(−i)n
Z Z
 I I

Cn (t, t0 ) = dτ1 · · · dτn T Hint (τ1 ) . . . Hint (τn )
n!
where T = [ · ] is the time ordered product, also called T-product.
If we recall the expansion of the exponential, we see that the complete solution can be written
as
∞  Z t
(−i)n
X Z Z  
 I I
 I
UI (t, t0 ) = dτ1 · · · dτn T Hint (τ1 ) . . . Hint (τn ) = T exp −i dτ Hint (τ )
n! t0
n=0
So allcomplications are inside the time order product, it arises from the non-commutation of the
interaction hamiltonian.
CHAPTER 4. INTERACTING FIELD THEORIES 84

4.4 S-matrix operator & T-order operator


We are describing how to treat QFT with interatcion. We defined and worked in the interaction 17/5/21
picture. We obtained two results: fields evolve like free fields,
d
i φI (~x, t) = [φI (~x, t), H]
dt
To calculate the evolutive operator in the interaction picture we used perturbative expansion
since overwise we are not able to solve the eq. However this procedure holds only if the coupling
of the theory is small. The expansion of elotive eq is

(−i)n t
X Z Z t   Z t 
 I I
 I
UI (t, t0 ) = dτ1 · · · dτn T Hint (τ1 ) . . . Hint (τn ) = T exp −i dτ Hint (τ )
n! t0 t0 t0
n=0

where we introduced the T-operator. Recall it comes out since the interacting hamilonian does
not commute in time. Let us see better what it is.
Definition (Time order product). Consider two operator A, B, then the time order product is
defined as (
A(t1 )B(t2 ) if t1 > t2
T [A(t1 ), B(t2 )] =
±B(t2 )A(t1 ) if t1 < t2
where + is for bosonic operator while − is for fermionic operator (it is because we want T
consistence with commutator and anticommutator results).
There is also another way to write it using the Heaviside Θ function.
T [A(t1 ), B(t2 )] = Θ(t1 − t2 )A(t1 )B(t2 ) ± Θ(t2 − t1 )B(t2 )A(t1 )
For a generic number of fields we can generalize the time product to:
T [A(t1 )B(t2 ) . . . Z(tz )] = (−1)p B(t2 )A(t1 ) . . . Z(tz )
for t2 > t1 > · · · > tz and p is the number of fermionic operator swaps.
Note (T-product to interacting hamiltonian). Consider the Hint of the theory, it is a bosonic
operator, namely if we apply the time product:
(
H(t1 )H(t2 ) if t1 > t2
T [H(t1 ), H(t2 )] =
+H(t2 )H(t1 ) if t1 < t2
since the hamiltonian is an "ensemble" of fields that must be alway real and positive, so any
physical hamiltonian contains fermions each time in couple. So for hamiltonian we can forget the
minus sign.
Another important operator we will see today is the S-matrix.
Definition (S-matrix operator). The S-matrix operator is defined as
  Z ∞ 
I
S ≡ UI (∞, −∞) = T exp −i dτ Hint (τ )
−∞
  Z 
= T exp −i d4 xHint I
(τ )

where we used the hamiltonian density. This object describes all the interaction that can happen
over all of time.
The temporal infinity is asymptotically the same as the particles flying away and not
interacting anymore, so this operator is useful to connect theoretical results with observables
since it describes the interacting happening in scattering processes. This operator containt
information about evolution of fields, so we expect it is "massive" presenets in formuals.
CHAPTER 4. INTERACTING FIELD THEORIES 85

4.5 Wick theorem & Feynman propagator


We have to develope a formalism to open up the time order product. The idea is inside the Wick
theorem. As always let us start from the easiest problem: derive the Wick problem for a real
scalar field and then we will extend to other fields.
First object we need is the Feynman propagator: for a real scalar field it is

DF (x − y) = h0| T [ϕ(x)ϕ(y)] |0i = ϕ(x)ϕ(y)

Notice this is the definition for a real scalar field, for example for a complex scalar field the
argument of the T operator changes.
Let us compute the calulation opening the T product and using theta function:

DF (x − y) = Θ(x0 − y0 ) h0| ϕ(x)ϕ(y) |0i + Θ(y0 − x0 ) h0| ϕ(x)ϕ(y) |0i


= Θ(x0 − y0 ) h0| ϕ+ (x)ϕ− (y) |0i + Θ(y0 − x0 ) h0| ϕ+ (x)ϕ− (y) |0i
= Θ(x0 − y0 ) h0| [ϕ+ (x), ϕ− (y)] |0i + Θ(y0 − x0 ) h0| − [ϕ− (x), ϕ+ (y)] |0i
= Θ(x0 − y0 )D+ (x − y) − Θ(y0 − x0 )D− (x − y)

where we open interaction fields in the positive and negative energy component and we neglect
fields applied to vacuum state which brings a zero result: recall ϕ+ |0i = 0 and h0| ϕ− = 0, as
ϕ+ ∼ a and ϕ− ∼ a† . Here D± are the components of the covariant commutator.
So the Fenyman propagator is equal to a correlation function in QFT.
Then, the crucial relation holds:

T [ϕ(x)ϕ(y)] = N [ϕ(x)ϕ(y)] + DF (x − y)

where it compares the Feynman propagator. This is the base of Wick theorem.
Exercise. Prove this relation.

Proof. we will prove this relation for the real scalar field. The idea si to prove

ϕ(x)ϕ(y) = T [ϕ(x)ϕ(y)] − N [ϕ(x)ϕ(y)]

To do it we open things:

T [ϕ(x)ϕ(y)] − N [ϕ(x)ϕ(y)] = + Θ(x0 − y0 )[ϕ− (x)ϕ− (y) + ϕ+ (x)ϕ+ (y) + ϕ− (x)ϕ+ (y) + ϕ+ (x)ϕ− (y)]
+ Θ(y0 − x0 )[ϕ− (y)ϕ− (x) + ϕ+ (y)ϕ+ (x) + ϕ− (y)ϕ+ (x) + ϕ+ (y)ϕ− (x)]
− Θ(x0 − y0 )[ϕ− (x)ϕ− (y) + ϕ+ (x)ϕ+ (y) + ϕ− (x)ϕ+ (y) + ϕ− (y)ϕ+ (x)]
− Θ(y0 − x0 )[ϕ− (x)ϕ− (y) + ϕ+ (x)ϕ+ (y) + ϕ− (x)ϕ+ (y) + ϕ− (y)ϕ+ (x)]

Now just simplify equal terms. The result is

T [ϕ(x)ϕ(y)] − N [ϕ(x)ϕ(y)] = Θ(x0 − y0 )[ϕ+ (x), ϕ− (y)] − Θ(y0 − x0 )[ϕ− (x), ϕ+ (y)] = DF (x − y)

Wick theorem is the generalization of this relation for n real scalar field
Theorem (Wick theorem). The T-product of n real scalar field can be writte in the following
way
X
T [ϕ(x1 )ϕ(x2 ) . . . ϕ(xn )] = N [ϕ(x1 )ϕ(x2 ) . . . ϕ(xn )] + N [ϕ(x1 ) . . . ϕ(xi )ϕ(xj ) . . . ϕ(xn )]
1 contraction
X
+ N [ϕ(x1 ) . . . ϕ(xc )ϕ(xd ) . . . ϕ(xi )ϕ(xj ) . . . ϕ(xn )]
2 contraction
+ ...
+ sum of all possible contractions.
CHAPTER 4. INTERACTING FIELD THEORIES 86

The proof is on textbook, for two fields it is the computation did in the exercice.
Another important result is a corollary of Wick theorem.
Corollary (of Wick theorem). Contractions between fileds evaluated at same time in N [ · ] do
not contribute to T-product. We can recall this fact using the following notation
T [N [ϕ(x)ϕ(x)]ϕ(xi ) . . . ϕ(xn )] = T [ϕ2 (x)ϕ(x1 ) . . . ϕ(xn )]NCET
where NCET = No Contractions at Equal Time. So when we compute the time-ordered product
using Wicks theorem we ignore the contractions we would have to compute at equal time.
This implies we will never see a Feymann propagator of fields calculated at the same point.

Proof. To prove the corollary of Wick theorem use the Wick theorem at the same point
T [ϕ(x)ϕ(x)] = N [ϕ(x)ϕ(x)] + DF (x − x)
where T [ϕ(x)ϕ(x)] = ϕ(x)ϕ(x) by definition of T-product. So, if we multiply by fields at other
times and take the time-ordering we find
T [N [ϕ(x)ϕ(x)]ϕ(xi ) . . . ϕ(xn )] = T [ϕ2 (x)ϕ(x1 ) . . . ϕ(xn )] − DF (x − x)T [ϕ(x1 ) . . . ϕ(xn )]

Notice DF (x − y) = ϕ(x)ϕ(y) is a number (not an operator), so for scalar field we can put it
inside and outside the T-product.

Now we can extend this results to complex scalar field and vector fields.

4.5.1 Generalization of Feynman propagator


Definition (Complex scalar field). The Feynman propagator for a complex scalar field is defined
in the following way
DF (x − y) = h0| T [ϕ(x)ϕ† (y)] |0i = ϕ(x)ϕ† (y)
Later we will understand why we used the same symbols. By opening the fields in component ±,
we see
DF (x − y) = Θ(x0 − y0 )D+ (x − y) − Θ(y0 − x0 )D− (x − y)
and no other terms are present since contraction of ϕ(x)ϕ(y) and ϕ† (x)ϕ† (y) with vacuum state
vanishies, so only combination of ϕϕ† is non vanishing. Here D± are same functions of the real
case but they represent two different propagators (one for a real scalar field and one for a complex
scalar field).
Definition (Vector field). The Feynman propagator for real Aµ and complex W µ vector fields is
defined as
µν µ ν µ ν
DF, real (x − y) = h0| T [A (x)A (y)] |0i = A (x)A (y)
µν µ ν† µ ν†
DF, complex (x − y) = h0| T [W (x)W (y)] |0i = W (x)W (y)

where
DFµν (x − y) = Θ(x0 − y0 )D+
µν µν
(x − y) − Θ(y0 − x0 )D− (x − y)
= −η µν DF (x − y)
where we used the Feynman gauge ξ = 1.
Exercise. Prove fundamental relations:
T [ϕ(x)ϕ† (v)] = N [ϕ(x)ϕ† (y)]
T [Aµ (x)Aν (y)] = N [Aµ (x)Aν (y)] + Aµ (x)Aν (x)
T [W µ (x)W ν† (y)] = N [W µ (x)W ν† (y)] + W µ (x)W ν† (y)
CHAPTER 4. INTERACTING FIELD THEORIES 87

Feyman propagator for fermions


Recall with fermions things are trinky since we have spinorial indicies (fermions are Dirac vector
fields) and it arises minus sign from anticommutation. So let us explicit the Feynman propagator
for a Dirac field:
F
Sαβ (x − y) = h0| T [ψα (x)ψ̄β (y)] |0i = ψ α (x)ψ̄β (y) = −ψ̄β (y)ψ α (x)

where αβ are spinorial indicies. Notice anticommutation is ok. Now let us use the compact
notation without the spinorial indicies, one can easy verify

SF (x − y) = Θ(x0 − y0 )S+ (x − y) + Θ(y0 − x0 )S− (x − y) = (i∂/ + m)DF (x − y)

This is a way to find the Feynman propagator for Dirac field from DF (x − y).

Exercise. Prove fundamental relations for complex scalar field and Dirac field:

T [ϕ(x)ϕ† (v)] = N [ϕ(x)ϕ† (y)]

T [ψα (x)ψ̄β (y)] = N [ψα (x)ψ̄β (y)] + ψ α (x)ψ̄β (y)

Solution: recall the following definition of T [ · ] and N [ · ].


α β α β α β β α
N [ψα (x)ψ̄β (y)] = ψ+ (x)ψ̄+ (y) + ψ− (x)ψ̄− (y) + ψ− (x)ψ̄+ (y) − ψ̄− (y)ψ+ (x)
T [ψα (x)ψ̄β (y)] = Θ(x0 − y0 )ψα (x)ψ̄β (y) − Θ(y0 − x0 )ψ̄β (y)ψα (x)

4.5.2 Generalization of Wick theorem


Now, the general formulation of Wick theorem. We saw it for scalar, dirac and vector field. 18/5/21

Theorem (General formulation of Wick theorem). Denote a generic field Bi (xi ):

Bi (x) = {ϕR , ϕC , Aµ , W µ , ψα , . . . }

(for example: ϕR real, ϕC complex scalar fields, Aµ , W µ vector fields and ψα Dirac field). Then
the T-product of n fields Bi can be written in the follow way

T [B1 (x1 ), B2 (x2 ), . . . , Bn (xn )] = N [B1 (x1 ), B2 (x2 ), . . . , Bn (xn )]


X h i
+ (−1)Pij B i (xi )B j (xj )N B1 (x1 ), . . . , B̂i (xi ), B̂j (xj ), . . . , Bn (xn )
ij
X
+ (−1)Pij +Pkl B i (xi )B j (xj )B k (xk )B l (xl )
ij,kl
h i
N B1 (x1 ), . . . , B̂i (xi ), B̂j (xj ), B̂k (xk ), B̂l (xl ), . . . , Bn (xn )
..
.
X
+ (−1)Pij +Pkl +··· . . . B i (xi )B j (xj ) . . . B k (xk )B l (xl ) . . . N [. . .ˆ. . .ˆ. . . ]
max

Pij denotes the number of permutation, the hat ˆ· denotes fields removed form the N-product.

The T-product can be expressed as N-product of felds and sum of Feynman propagators.
Pay attenction if we trick fermions we can put the propagator outside the sum only if we take
component, overwise we will break the matricial structure (recall SF (x − y)αβ ).
Now, recall that some contractions vanish, we have to find the non-zero. The only non-
vanishing contractions are the ones of fields associated to the same particle, namely:
CHAPTER 4. INTERACTING FIELD THEORIES 88

1. contraction for two real or complex scalar fields

ϕR (x)ϕR (y) = DF (x − y) = ϕC (x)ϕ†C (y)

2. contraction for a real and complex vector fields

Aµ (x)Aν (y) = DFµν (x − y) = W µ (x)W ν † (y)

3. contraction for Dirac spinorial fields

ψα (x)ψ̄β (y) = SF (x − y)αβ

Recall the Wick corollary: N-operator calculated at the same spacetime does not contribute to
T-product, so we can remove the normal order of fields calculated at the same point

T [N [B1 (x1 ), B1 (x1 )]B2 (x2 ), . . . Bn (xn )] = T [B1 (x1 )2 B2 (x2 ) . . . Bn (xn )]N CET

This is the case of the Hamiltonian: it is the normal order product of the hamiltonian.

4.6 Feynman propagator: physical interpretation


Let us start an intuitive way to interpret the Feynman propagator and the T-operator. If we
think about the contraction of two fields, mathematically it is the vacuum expectation value of
the product of two fields, but what is it physically?

4.6.1 For scalar field


Let us see it for scalar field and then it is the same for other fields. Recall the definition of
Feynman propagator for scalar field:

DF (x − y) = h0| T [ϕ(x)ϕ(y)] |0i = ϕ(x)ϕ(y)

ad using the Theta function

DF (x − y) = Θ(x0 − y0 ) h0| ϕ(x)ϕ(y) |0i + Θ(y0 − x0 ) h0| ϕ(x)ϕ(y) |0i


= Θ(x0 − y0 ) h0| ϕ+ (x)ϕ− (y) |0i + Θ(y0 − x0 ) h0| ϕ+ (x)ϕ− (y) |0i
= Θ(x0 − y0 ) h0| [ϕ+ (x), ϕ− (y)] |0i + Θ(y0 − x0 ) h0| − [ϕ− (x), ϕ+ (y)] |0i
= Θ(x0 − y0 )D+ (x − y) − Θ(y0 − x0 )D− (x − y)

To interprete this object consider the case where x0 > y0 , then only the first term contributes

DF (x − y) = h0| ϕ+ (x)ϕ− (y) |0i

Recall ϕ+ ∼ a and ϕ− ∼ a† and we know their physical interpretation: desruption and creation
respectively. In the Feynman propagator we start from a state that is zero and in which it is
applied ϕ− (y) at time y0 , then the state evolves and at time x0 it is applied ϕ+ (x) and we go
to the final state that is the vacuum one, the same of the beginning. DF (x − y) for y0 > x0 for
real scalar field can be drawn with a dashed connecting x, y and with an arrow pointing to the
arrival point x as shown in fig 4.1.
In the case when y0 > x0 , the Feynman propagator is

DF (x − y) = h0| [ϕ+ (y)ϕ− (x)] |0i

so the operator ϕ− creates a particle in x (at time x0 ), it propagates to point y where it is


desrupted by ϕ+ at time y0 . Grafically it is again the dashed line (scalar propagator) but now
the arrow points to x: see fig 4.2.
CHAPTER 4. INTERACTING FIELD THEORIES 89

y y0

x x0

Figure 4.1: Virtual scalar particle. The diagram shows the virtual scalar particle being created
at y and then destroyed at x.

x x0

y y0

Figure 4.2: Virtual scalar particle. The diagram shows the virtual scalar particle being created
at x and then destroyed at y.

However notice only the sum of the two terms is Lorentz invariant and not only a piece (eg
x0 > y0 ), so we will draw it with the dashed line and two opposite arrows, since both cases are
contained in the propagator:
a
DF (x − y) = y x
a
So Feynman propagator describes the motion of virtual particle and not the one of real
particle: the propagator does not represent a physical path of the particle. We have two physical
fields at points x, y which do not physical move from one point to the other, however we used
the propagator just as a picture way to interpretate the correlation between two point y, x.
Example (Potential barrier). A physical example of this is a particle propagating through a
potential barrier. We start from a particle at the left of a potential barrer, it hits the barrier.
We do not know what happen but after a certan time we find a particle at the other side of the
barrier. Graphycally we can interpret it with a virtual propagation (with a dashed line) since
it is unphysical to have a particle below a potential barrier. Just like a quantum particle can
tunnel through a potential barrier higher than its energy, the propagator describes a probabilistic
physical process.

Explicit calculation of the scalar propagator


Let us spell out the expression
DF (x − y) = Θ(x0 − y0 )D+ (x − y) − Θ(y0 − x0 )D− (x − y)
Z 3
1 d k
= 3
[Θ(x0 − y0 ) exp(−ik(x − y)) + Θ(y0 − x0 ) exp(ik(x − y))]k0 =ωk
(2π) 2ωk
If we add the complex k0 plane we have a 4dim integration: use the Cauchy theorem
d4 k exp(−ik(x − y))
Z
DF (x − y) = i 4
CF (2π) k 2 − m2
d4 k
Z
= 4
exp(−ik(x − y))D̃F (k)
CF (2π)

where we are integrating in a circuit that is in the imaginary k0 plane: we have two zeros at
k0 = ±ωk , so we pass below the negative pole and above the positive pole. This shows we must
have k 2 6= m2 , otherwise the propagator diverges. It is the condition to have a physical paticle.
CHAPTER 4. INTERACTING FIELD THEORIES 90

Here DF (x − y) is the Feynman propagator in the coordinate space and D̃F (k) is the Feynman
propagator in the momentum space. Explicitely
i
D̃F (k) ≡
k2
− m2
Exercise. Discuss the physical interpretation of a complex scalar field.
Solution. Recall the Feynman propagator for a complex scalar field

DF (x − y) = h0| T [ϕ(x)ϕ† (y)] |0i = ϕ(x)ϕ† (y)


= Θ(x0 − y0 ) h0| ϕ+ (x)ϕ†− (y) |0i + Θ(y0 − x0 ) h0| ϕ†+ (y)ϕ− (x) |0i
For x0 > y0 the Feynman propagator is just
DF (x − y) = h0| ϕ+ (x)ϕ†− (y) |0i
Recall ϕ+ ∼ a and ϕ†− ∼ a† , so at y we create (virtualy) a particle of type a that is annihilated
at point x. Notice when we have complex field we put also an arrow in the slashed line which
connect the two points x, y; the direction of the arrow indicates the charge flow of the virtual
particle, in our case from y to x.
For y0 > x0 , we have the second piece:
DF (x − y) = h0| ϕ†+ (y)ϕ− (x) |0i
where ϕ− ∼ b† and ϕ†+ ∼ b. So at point x we create a particle of type b that is annihilated at
point y, recall type b is the antiparticle of type a. In this case the arrow in the dashed line
points always from y to x since the charge flow of an antiparticle is equal to the inverse flow of a
particle (since antiparticle has the opposite charge of the particle).
To denote the timespace flow we add an arrow for particle type a (points to y) and an
arrow from particle type b (points to x). So the Feynman propagator for complex scalar field is
graphycally represents by
a
DF (x − y) = y x
b

4.6.2 For real vector field


Notice in real scalar field we have only one type of particle (no flow of charge), while in complex
scalar field we have floe of charge since the propagators contains the flow of a particle or the
opposite flow of an antiparticle. Given scalar fields, we can extend to vector and Dirac fields.
The Feynman propagator for a real vector Aµ is

DFµν (x − y) = h0| T [Aµ (x)Aν (y)] |0i = Aµ (x)Aν (y)

= y, ν x, µ

The interpretation is similar. We have a real field so only one type of particle: no charge flow
and so no arrow in the propagation line. For a vector field we used wider line, connected points
y and x. Now we have indicies µ, ν, so we had to specify also this: at point y we create a vector
boson of index ν and in point x we destroy a vector boson of index µ.
The explicit expression of the vector boson propagator is
d4 k
Z
µν
D = 4
exp(−ik(x − y))D̃Fµν (k)
CF (2π)

If we use the ξ = 1 gauge, then


−iη µν
DFµν (x − y) = −η µν DF (x − y) D̃Fµν (k) =
k2
Notice these expressions are compatible with photons with m = 0.
CHAPTER 4. INTERACTING FIELD THEORIES 91

4.6.3 For Dirac field


The last propagator useful is the Feynman propagator for Dirac fermions. Its form is

SFαβ (x − y) = h0| T [ψα (x)ψ̄β (y)] |0i = ψ α (x)ψ̄β (y)


α β β α
= Θ(x0 > y0 ) h0| ψ+ (x)ψ̄− (y) |0i − Θ(x0 < y0 ) h0| ψ̄+ (y)ψ− (x) |0i

Recall ψ is a complex field, so the interpretation will be similar to the complex scalar field. One
piece of the propagator describes a virtual particle of type c flows from y to x: created (ψ̄− ∼ c†
in y) and then annihilated (ψ+ ∼ c in x). While the other piece describes the a virtual particle
of type d (antiparticle) flows from x to y: created (ψ− ∼ d† in x) and then annihilated (ψ̄+ ∼ d
in y).
Let us graphycally denote the Fermion propagator with a continouos line between y and x.
We put the charge flow arrow from y to x (as for the complex scalar field, a particle will flow
in the same verse, while an antiparticle will flow in the opposite verse). So, the total (Lorentz-
invariant) propagator is graphycally represented as

e− = c
SFαβ (x − y) = y, β x, α
e+ = d

where we considered an electron.


The explicit expression for the Dirac field propagator is

d4 k
Z

SF (x − y) = 4
exp(−ik(x − y))S̃F (k) = i∂/ + m DF (x − y)
CF (2π)

where
i k/ + m
S̃F (k) = =i 2
k/ − m k − m2
Notice in the last formulas we considerred the matrix object and not the components, so we did
not write spinorial indicies αβ. The first expression of S̃F (k) indicates something divided by a
4 × 4 matrix, we have to interpretate its meaning, that is the second expressione written.

4.7 Uncontracted fields: physical interpretation


Resuming: we start from the T-product, we use the Wick theorem, then we have two term,
contractions of fields (propagators) and normal-ordered products of uncontracted fields. We
already gave a physical interpretation on the Feynman propagator, now let us give one to the
N-product. So let us give a look to uncontracted fields physical interpretation.

Real scalar field


As always let us start from uncontracted real scalar field. They arise when we open the T-product.
Recall real scalar field is the sum of positive and negative solution, proportional to annihilation
and creation operators:
ϕ(x) = ϕ+ (x) + ϕ− (x) ∼ a + a†
If we use the same graphic notation used for propagators, then

1
Z
d3 k a
ϕ+ (x) = √ exp(−ikx)a(k) = x
(2π)3/2 2ωk

namely ϕ+ (x) ∼ a means a particle comes from the left and disappear at x.
CHAPTER 4. INTERACTING FIELD THEORIES 92

While
a
d3 k
Z
1
ϕ− (x) = √ exp(ikx)a† (k) = x
(2π)3/2 2ωk

ϕ− (x) ∼ a† means a particle is created in the point x and propagates to the right.
This interpretation can be extended to any other fields.

Complex scalar field


Consider a complex scalar field. Recall the general solution

d3 k h
Z
1 †
i
ϕ(x) = √ a(k) exp(−ikx) + b (k) exp(ikx) = ϕ+ (x) + ϕ− (x)
(2π)3/2 2ωk k0 =ωk

So there are 4 components: a, b† from ϕ and a† , b from ϕ† .


The interpretation is:

• ϕ+ (x) ∼ a(k) is the annihilation of a charged particle at x

a
ϕ+ (x) = x

• ϕ†− (x) ∼ a† (k) is the creation of a charged particle at x

a
ϕ†− (x) = x

• ϕ†+ (x) ∼ b(k) is the annihilation of an antiparticle at x

b
ϕ†+ (x) = x

• ϕ− (x) ∼ b† (k) is the creation of an antipaticle at x

b
ϕ− (x) = x

Real vector field


Recall the general solution

d3 k X µ
Z
µ 1 


A (x) = √  (λ) (k) a(k) exp(−ikx) + a (k) exp(ikx)
(2π)3/2 2ωk
λ
k0 =ωk

We can divide the two components in Aµ+ which annihilates a vector particle in point x, and Aµ−
which creates a vector particle.
Recall with vector fields we have associated a Lorentz index to the spacetime point. The
graphycal interpretation is the following:

Aµ+ (x) = x, µ

Aµ− (x) = x, µ
CHAPTER 4. INTERACTING FIELD THEORIES 93

Dirac fields
The general solution is
2
d3 k X 
Z
1 
ψ(x) = √ cr (k)ur (k) exp(−ikx) + d†r (k)vr (k) exp(ikx)
(2π)3/2 2ωk r=1 k0 =ωk

It is a complex field so we have 4 different operators: c, d† , c† , d. So we can identify the operators:

• ψ+ ∼ c(k) is the annihilation of type c at point x

c
ψ+ (x) = x

• ψ̄− (x) ∼ c† (k) is the creation of type c at point x

c
ψ − (x) = x

• ψ̄+ (x) ∼ d(k) is the annihilation of type d (antiparticle) at x

d
ψ + (x) = x

• ψ− (x) ∼ d† (k) is the creation of type d at x

d
ψ− (x) = x

To summarize: we took a classical theory with interaction, we quantized it, then we described
the evolution of the interacting QFT using the interacting picture. Fields operator evolves
like free fields, so solutions are the same as found at the beginning of the interaction picture.
The complication is contained in the evolution of states: complicate differential equation with
hamiltonians calculated at different times. To solve it, we use a perturbative way, we introduced
a compact notation with the T-product. Then Wick theorem and corollary told us how to deal
with T-product: use N-product and contraction (where we defined the Feynman propagator).
We introduced a physical interpretation for contracted fields (virtiual particles) and uncontracted
fields (propagation till/from a point).
Now, we will apply there results in a specific field theoy: QED. We will open the T-product
and calculate some simple events.
Chapter 5

Quantum electro dynamic

We will see the QED S-matrix expansion. We saw it in general QFT and now we will specify
defining QED theory. This theory is an interaction-theory between fermions and photons
The lagrangian of the theory is
1
LQED = − F µν Fµν + LGF + ψ̄(iD / − m)ψ
4
The first pieceis the free propagation of photons. To quantize the theory we need to fix the gauge:
in LGF there are the gauge fixed terms. The third term is the free kinetic of Dirac field and
the interactions with photons. We wrote them in a compact way using the covariant derivative,
recall our convention Dµ = ∂µ + iqAµ .
We are interested in the interactions, namely in the interacting term of the Lagrangian,
explicit it is
Lint = −q ψ̄γ µ ψAµ = −q ψ̄ Aψ
/ = −Hint
Notice that the interaction Lagrangian and Hamiltonian are opposite of each other: this holds in
general, as long as there are no derivative terms.
There are two fermions and a photon: this is the minimum we need to have an interaction in
QED.
Do note all fields are in the interaction picture (we will omit the index I), so we already
know their expression: the usual one (in interaction picture fields evolve as free ones). The next
piece of information we must know is the evolution operator U (t) in order to know the states
evolution. Recall the S-matrix in QED:
  Z 
4 /
SQED = T exp −iq d xN [ψ̄(x)A(x)ψ(x)]

Let us expand the exponential:



(−iq)n
X Z
SQED = d4 x1 . . . d4 xn T [N [ψ̄ Aψ]
/ x1 . . . N [ψ̄ Aψ]
/ xn ]
n!
n=0

all physics of the interaction is in this matrix.

5.1 S-matrix expansion in coordinate space


Now we will apply Wick theorem and corollary. First thing is the S-matrix expansion, for the
moment in coordinate space.

0th order
The 0th order is when we put to zero the coupling:
S (0) = I
namely the theory is a free fields theory and we already know everything.

94
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 95

1st order
The 1st order term of the S-matrix expansion is (notice it is the 1st order term of the expansion
and not the S-matrix up to 1st order, since the latter is the sum of the 0th and 1st order)
Z
S (1) = −iq d4 xT [N [ψ̄(x)A(x)ψ(x)]]
/

Notice S (1) = O(q). Now, let us use the Wick theorem and the corollary since we are computing
fields at the same time: Z
S = −iq d4 xN [ψ̄(x)A(x)ψ(x)]
(1) /

We can split ψ̄, A


/ and ψ into their + and - components: N-product we can identify 8 different
contributions

/+ + A
N [(ψ̄+ + ψ̄− )(A / − )(ψ+ + ψ− )](x)

which we can represent using Feynman diagrams. All of these diagram only have one vertex at x,
and they have a fermion, an antifermion and a photon being created/annihilated there. As an
example, if we take the operator with only + terms:

/ + ψ+
ψ̄+ A

where ψ+ is the annihilation of a fermion in x (like an electron), A+ is the annihilation of a


photon in xµ , and ψ̄+ is the annihilation of an antiparticle in point x.
A second example is
e−

/ − ψ+ =
ψ̄+ A x γ

e+
namely there are an electron and a positron which annihilate at point x (+ operator) and a
photon is created in x (- operator). Notice there are no time consequence, all is done at the same
point x.
From these example we see flow of the fermion arrows is continuous in the diagram we wrote,
this can be stated as a general rule: arrows never have a crash (this is associated to conservation
of charge). If we have a diagram with crashed arrow we do not conserve charge. In addivction,
generally particles which are created (so, with subscript ) go rightward, particles which are
annihilated (with subscript +) come from the left (it is explicit in the second example).
Do note, all these 8 diagrams represent unphysical processes: due to the fact that the photon
is massless, relativistic kinematics forbids them. For example electron-positron annihilation in a
photon is not a physical process since it does not conserve the 4momentum. The physical process
requires 2 photons. However a 1st order interaction can be physical for massive vector bosons.

2nd order
Recall S-matrix is the evolution operator computed from −∞ to ∞. All evolution is inside 24/5/21
S-matrix. First ored of S-matrix in QED is not physical possible since kinematic does not allow
them.
We will see physical processes start form 2nd order. It contains two interaction hamiltonian
/
Hint = q ψ̄ Aψ:
(−iq)2
Z
(2)
S = d4 xd4 yT [N (ψ̄ Aψ)
/ x N (ψ̄ Aψ)
/ y]
2!
It is a second order since it is O(q 2 ) and q is our expansion parameter.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 96

By the corollary of Wick’s theorem, we can write

/ x N (ψ̄ Aψ)
T [N (ψ̄ Aψ) / y ] = T [(ψ̄ Aψ)
/ x (ψ̄ Aψ)
/ y ]NCET

so when we compute the contractions we can only connect fields computed at x with ones
computed at y (NCET). Also, contractions between different types of fields vanish, as do
contractions between a complex-valued field and itself. So, the T-product of the 2nd order term
of S-matrix expansion is the sum of the following non-vanishing terms:

1. the 0 contraction term


h   i
N /
ψ Aψ x
/
ψ Aψ y
,
| {z }
A

2. the 1 contraction term


     
/ x (ψ Aψ)
N (ψ Aψ) / y + N (ψ Aψ)
/ x (ψ Aψ)
/ y + N (ψ Aψ)
/ x (ψ Aψ)
/ y ,
| {z } | {z }
B1 B2

3. the 2 contraction term


     
/ x (ψ Aψ)
N (ψ Aψ) / y + N (ψ Aψ)
/ x (ψ Aψ)
/ y + N (ψ Aψ)
/ x (ψ Aψ)
/ y ,
| {z } | {z }
C1 C2

4. and the 3 contraction term


 
/ x (ψ Aψ)
N (ψ Aψ) / y .
| {z }
D

For any of these we can write a specific (2nd order) propagator, whose expression is

(−iq)2
Z
(2)
SX = d4 x d4 y X
2!

where X is any of the terms we have just written.

No propagators

Let us start from the A term: processes with zero contraction (namely zero propagator):

−iq)2
Z
(2)
SA = d4 xd4 yN [(ψ̄ Aψ)
/ x (ψ̄ Aψ)
/ y]
2!
The diagram looks like the product of two disconnected first-order diagrams. Since they were
not physically allowed, this is not allowed either. If our vector were massive, like the Z boson,
this would describe two separate decay processes.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 97

One fermion propagator

Now, consider processes B1 : contraction between ψ ψ̄, so 1 fermion propagator. The S-matrix is

(−iq)2
Z     
(2)
SB1 = d4 xd4 y / x (ψ Aψ)
N (ψ Aψ) / y + N (ψ Aψ)
/ x (ψ Aψ)
/ y
2!

if we recall that fermion operators anticommute, we can swap (ψ̄ Aψ) / y and (ψ̄ Aψ)
/ x inside the
N-product (for example let us do it for the first N-product), we move an even number of fermions
(4) and so the additional sign due to anticommutation is (−1)4 . Then we can change x ↔ y
without problems since they are both integrated away. The result is that the two contributions
are equal. So
Z  
(2) 2 4 4
SB1 = −q d xd y N (ψ Aψ) / x (ψ Aψ)
/ y
Z
= −q 2 d4 xd4 y N ψ(x)A(x)S
 
/ /
F (x − y)A(y)ψ(y)

where SF (x − y) = ψ(x)ψ̄(y).
There are 24 = 16 possible contributions once the N-product is written in terms of ψ± , Aµ± .
However in QED only processes 2 particles to 2 particles are kinematic allowed, so let us
concentarte only on these physical processes: the ones with 2- and 2+ field operators.
For example consider the process of the form

γ γ
/ + (y)
A / − (x)
A
SF (x − y)
/ − )x SF (x − y)(ψ+ A
(ψ − A / + )y = ,
ψ+ (y) ψ − (x)

e− e−

from (A / + ψ+ )y an electron and a photon annihilate at y, then SF is a virtual propagation, then


/ − )x is the generation of an electron and a photon: this process contributes to the Compton
(ψ̄− A
scattering. The idea is that something happened in y is related to something happend in y and
the correlation is reppresented by the Feynman propagator.
(2)
Exercise. Derive all possible 2 → 2 processes that can arise from the operator SB1 .

One photon propagator

Now, B2 processes: contraction between photons, so one photon progatator:

(−iq)2
Z  
(2)
SB2 = d4 xd4 y N (ψ Aψ)
/ x (ψ Aψ)
/ y
2!
(−iq)2
Z
= d4 xd4 yDFµν (x − y)N [(ψ̄γµ ψ)x (ψ̄γν ψ)y ]
2!

Notice for one fermion propagator we cannot bring out SF (x − y) from the remain N-product
/ y ] since it is a matrix; while we can do it if we consider spinorial components,
/ x (Aψ)
N [(ψ̄ A)
namely if we write SFαβ (x − y)N [(ψ̄α A) / β )y ]. In the case of a photon propagator, we can
/ x (Aψ
µν
bring out the propagator DF (x − y) since photons do not have spinorial indicies.
As before, we have 16 possible combinations of + and - when we open the normal-ordered
product, but only the ones with two destructions and two creations are kinematically allowed.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 98

A typical physical process is e+ e− → e+ e− . One contribution to this process comes from the
term
DFµν (x − y)(ψ̄− γµ ψ− )x (ψ̄+ γν ψ+ )y
annihilation of electron and positron at point y, then virtual photon, then creation of and at x
creation of positron and electron at point x (the description of the formula is readible from right
to left).
(2)
Exercise. Derive all possible 2 → 2 processes that can arise from the operator SB2 .

Exercise. Derive the S-matrix elements at 2nd order contributing to the following scattering

e+ e− → e+ e−

Solution: in the S-matrix we need ψ+ and ψ̄+ (corresponding to the annihilation of e− , e+ )


and ψ̄− and ψ− (corresponding to the creation of e− , e+ ). All other combination of 2 + and 2-
fields do not contribute to the scattering process. The operator must have contracted photons
since there are no photons in initial or final states. So the S-matrix is:

−q 2
Z 
(2)
Se+ e− = d4 xd4 y DFµν (x − y) N [(ψ̄− γµ ψ− )x (ψ̄+ γν ψ+ )y ]
2
+N [(ψ̄− γµ ψ+ )x (ψ̄+ γν ψ− )y ]
+N [(ψ̄+ γµ ψ+ )x (ψ̄− γν ψ− )y ]

+N [(ψ̄+ γµ ψ− )x (ψ̄− γν ψ+ )y ]

Notice term 3 is equal to term 1 and term 4 is equal to term 2, this is because DFµν is symmetric
in the exchaneg of µν and by properties of ψ̄γψ. So
Z  
(2)
Se+ e− = −q 2 d4 xd4 yDFµν (x − y) N [(ψ̄− γµ ψ− )x (ψ̄+ γν ψ+ )y ] + N [(ψ̄− γµ ψ+ )x (ψ̄+ γν ψ− )y ]

Namely there are only two topological diagram: two difference physical processes. The first is
S-channel contribution, in which the fermions annihilate into a virtual photon which propagates
and then annihilates into fermions again. The second is the T-channel contribution, in which
the electron and positron exchange a virtual photon but are never annihilated. These names are
related to Mandelstam variables s, t, u.

e− e−

S-channel DFµν (x − y)(ψ̄− γµ ψ− )x (ψ̄+ γν ψ+ )y =

e+ e+
e− e−

T-channel DFµν (x − y)(ψ̄− γµ ψ+ )x (ψ̄+ γν ψ− )y =

e+ e+
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 99

One photon and one fermion propagator

Now, C1 processes with two contractions: 1 bosonic and 1 fermionic propagators

q2
Z     
(2) 4 4
SC1 =− / / / /
d x d y N (ψ Aψ)x (ψ Aψ)y + N (ψ Aψ)x (ψ Aψ)y
2!
Z
2
d4 x d4 y DFµν (x − y)N ψ(x)γµ SF (x − y)γν ψ(x)
 
= −q

since the two components can be cast into each other by renaming x ↔ y (which can always be
done since they are integrated away), and permuting 4 fermion operators.
One can have 22 = 4 different terms and only 1 → 1 are kinetic allowed (the 0 → 2 and 2 → 0
processes are not allowed). So, we have two possibilities: either an electron having a loop and
coming back to itself, or a positron doing the same. Graphic loops are in fig 5.1, they are called
loop diagrams. For example the electron loop term is

ψ̄− (x)γµ SF (x − y)γν ψ+ (y)DFµν

Figure 5.1: Electron loop, positron loop.

The problem is that there are infinite loops and this is a problem for renormalization.

Two fermion propagators

Now, C2 processes: two contractions between Dirac fields, so 2 fermions propagators

−q 2
Z  
(2) 4 4
SC2 = d xd y N (ψ Aψ) / x (ψ Aψ)/ y
2
−q 2
Z
µ
= d4 xd4 y(ψ̄α γαβ ν
ψβ )x (ψ̄γ γγρ ψρ )y N [Aµ Aν ]
2
q2
Z
= d4 xd4 yTr[SF (x − y)γ µ SF (x − y)γ ν ]N [Aµ Aν ]
2
where we used permutation of the spinors:
 
µ ν
N (ψ α γαβ ψβ )x (ψ γ γγδ ψδ )y
µ ν
= (−)3 ψδ ψ α ψβ ψ γ γαβ γγδ
µ ν
= −SF (y − x)δα γαβ SF (x − y)βγ γγδ
= − Tr(SF (y − x)γ µ SF (x − y)γ ν ) ,

where we are taking the trace since all the indices are contracted, including the first and last.
Also here one can have 22 = 4 different terms and only 1 → 1 are kinetic allowed. For example

Aµ+ (y)Aν− (x) =

this is again a loop diagram called photon self-energy. The only other valid one is the same
except for x ↔ y.
From this part what we must know for the exam is to interpretate and draw diagrams and to
open the T-product to find propagators.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 100

Three propagators

Now, process D : with 3 propagators

−q 2
Z  
(2) 4 4
SD = / /
d xd y N (ψ Aψ)x (ψ Aψ)y
2
q2
Z
= d4 xd4 yTr[SF (y − x)γµ SF (y − x)γν ]DFµν (x − y)
2
where all is contracted, so we do not have initial or final particle, namely this is the vacuum
energy. So the vacuum energy of an interacting theory is different from the one of a free theory.
This becomes relevant when QED is coupled to gravity.
The graphic of the propagators is in fig 5.2.

Figure 5.2: Vacuum diagram.

5.2 S-matrix expectation value in momentum space


Let us consider two Fock states: an initial |ii , and a final |f i

|ii = |ψ(t = −∞)i


|f i = |ψ(t = +∞)i

The initial state is We want to see the probability that the inital state evolves to the final state.
This probability is related to the expectation value of the S-matrix:

Sf i ≡ hf | S |ii

This is a transition amplitude i → f . The probability density will be the modulus square:

probability density i → f = |Sf i |2

Recall in QM the probability is related to the wave function. Here, in QFT, the solution
of diff eq are fields (namely operators) and not functions, while states are Fock states. The
quantity that connects us to numbers (accesible physics information) is the modulus square of
the expectation value of S-matrix. This is the observable we want measure.
So, let us focus on how to contract S-matrix with initial and final states. S-matrix is
proportional to T-product, we can open it using Wick theorem so we have contracted and
uncontracted fields, now we want to applied these fields to initial and final states to have numbers,
so measurable quantities.

Contraction with initial state


Consider to make a contraction with initial state |ii, for example an electron with spin s and
momentum p: −
es (p) = (2π)3/2 2ωp cs (p) |0i
p
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 101

Let us calculate all passible contraction with annihilation operators ψ+ , ψ̄+ :


Z r
− 2ωp −ikx X
ψ+ (x) es (p) = d k 3
e ur (k)cr (k)c†s (p) |0i
2ωk r
Z r
2ωp −ikx X n o 
= d3 k e ur (k) cr (k), c†s (p) − c†s (p)cr (k) |0i
2ωk r
Z r
2ωp −ikx X
= d3 k e p − ~k) |0i
ur (k)δrs δ (3) (~
2ωk r
= e−ipx us (p) |0i .

p − ~k). The result is a spinoral vector but


where the anticommutator gives a delta Dirac δrs δ (3) (~
just a function and not an operator.
If we do it with the field ψ̄+ (x):
Z  1/2
2ωp X
ψ̄+ (x) e− 3
ur (k)cr (k)c†s (p) |0i = 0

s (p) = d k exp(−ikx)
2ωk r
n o
since dr , c†s = 0. he same holds for any field operator which is unrelated to the state (like a
photon annihilation operator).
Generalizing this, all non-vanishing contractions between fields and initial states are the
following:

ψ+ (x) e− electron

s (p) = exp(−ipx)us (p) |0i
positron
+
ψ̄+ (x) es (p) = exp(−ipx)v̄s (p) |0i

Aµ+ (x) |γλ (p)i = exp(−ipx)(λ) |0i photon

So we found a plane wave times an objects that describes the polarization dof of the physical
quantity.

Contraction with final state


Doing as above, all non-vanishing contractions with final state |f i:


e (p) ψ̄− (x) = h0| exp(ipx)ūs (p) electron

s+
es (p) ψ− (x) = h0| exp(ipx)vs (p) positron
hγλ (p)| Aµ− (x) = h0| exp(ipx)µ(λ) (p) photon

5.2.1 S-matrix expansion in momentum space


We saw which are the nonvanishing contracting fields. These are the only way to annihilate or 25/5/21
create a particle.

0th order
Now let us calculate the 0th order of S-matrix expectation value

hf | S (0) |ii = hf | I |ii = hf |ii

and the resut is 1 if initial and final states are equal, overwise it is 0.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 102

1st order
Now, 1st order term: we have to choose intial and final states, for example we can select the
following process:
e−
s0
p0
p
e−
s

k0
γ(λ0 )

it is an electron with spin s and momentum p which radiates a photon, so the final states are an
electron with s0 and p0 and a photon with polarization λ0 and momentum k 0 .
So the initial state is
|ii = e−

s (p)
and the final one is
|f i = e− 0 0

s0 (p ), γ(λ0 ) (k )
Let us calculate the brakets of 1st order S-matrix: recall the complete expression
Z
S = −iq d4 xN [ψ̄(x)A(x)ψ(x)]
(1) /

however the only term that contributes in the expectation value of the process e− → e− γ is the
one with an electron annihilated and a photon and electron created. So
Z
(1)
Se− →e− γ = −iq d4 xN [ψ̄− (x)A(x)ψ(x)
/ +]

where ψ+ annihilates an electron and ψ̄− A


/ creates an electron and a photon.
Let us compute the expectation value:
(1) (1)
Sf i ≡ hf | Se− →e− γ |ii
Z
= −iq d4 x e− 0 0


/
s0 (p ), γ(λ0 ) (k ) (ψ̄− Aψ+ )x es (p)

now we select the only nonvanishing contaction: initial electron with ψ+ , photon with A / and
final electron with ψ̄− (this comes from the fact that any operator has a specific state to acting
on, if it acts on a state which is not its own the contribution vanishes as seen in relations wrote
before). So recalling the results of each contractions:
Z
(1) µ
Sf i = −iqλ0 (k )ūs0 (p )γµ us (p) d4 x exp −i(p − p0 − k 0 )x h0|0i
0 0
 

where h0|0i = 1 and the integral is a delta that implies the conservation of the 4momentum. All
the physic is contained in the remians elements that we enclose in the term Mf i called Feynman
amplitude, so the result is
(1)
Sf i = (2π)4 δ 4 (p − p0 − k 0 )Mf i
where
Mf i = −iqūs0 (p0 )(−iqγ µ )us (p)λ0 (k 0 ) = ūs0 (p0 )/λ0 (k 0 )us (p)
However recall 1st order S-matrix processes are not physical since the kinematic is not respected.
Let us stress two points:
• all the physics of the interaction is in the Feynman amplitude Mf i ; in it we have spinors
for electrons and polarizations for photons while the term −iqγ µ says the specific way in
which the photon is interacting with fermions
• the kinematic is on the delta
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 103

5.3 Feynman rules


Now, we want to automatize the computation of Feynman amplitude Mf i . Again consider the
above non-physical process e− → e− γ as example, following the so called Feynman rules we can
automatize the writing of M :

1. go to the end of the fermionic line and then read the diagram backward (from final state
to initial one)

2. find the final electron e−


s0 (p), so we associate ūs (p )
0
0

3. running the fermionic line reach the vertex of QED, it is −iqγ µ so write it

4. following the fermionic line we end to the initial electron e−


s (p), so we associate us (p)
0
5. include the photon polarization since there is a photon in the process, so write λµ (k 0 )

Recall we are observing an unphysical process, however the basic process of QED are "valid",
however we will never observe this since all is gone to zero by the kinematic delta. With different
interction we have to change the vertex.
Using the Feynman rule we do not have to go throught all computations but just write the
final physical terms. Recall all this is valid in the perturbation picture.
To verify the Feynman rule let us calculate explicitely the 2nd order of Feynman amplitude
expectation value and compare it with the result from Feynman rule.

2nd order
We pass to second order to study real physical process. Let us consider the Compton scattering:

e− γ → e− γ

let us denote the final state with the prime symbol. So initial and final states are

|ii = e−

s (p), γλ (k)
|f i = e− 0 0

s0 (p ), γλ (k )
0

The S-matrix of Compton scattering was given as exercise last time:


Z  
(2) 2
S = (−iq) d4 xd4 y N [(ψ̄− A
/ − )x SF (x − y)(A
/ + ψ+ )y ] + N [(ψ̄− A
/ + )x SF (x − y)(A
/ − ψ+ )y ]
| {z } | {z }
I II

The graphical interpretation is shown in fig 5.3.

γ γ γ γ

e− e−
s s0 e− e−

Figure 5.3: The two contribution of Compton scattering: first term on the left and second term
on the right.

We want to compute
(2) (2)
Sf i = hf | SCompt |ii
Let us do it separately for each term.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 104

I term
(2)
Let us calculate the I contribution to the expectation value Sf i :
Z
d4 xd4 y e− 0 0

SfI i = −q 2


/ /
s0 (p ), γλ (k ) N [(ψ̄− A− )x SF (x − y)(A+ ψ+ )y ] es (p), γλ (k)
0

d4 q
Z Z Z
2 0 0 4 0 0
d4 y exp(i(q − p − k)y) h0|0i

= −q ū(p )/(k )S̃F (q)/(k)u(p) × d x exp −i(q − p − k )x
(2π)4

where the nonvanishing contraction are 4: final electron with ψ̄− , final photon with A
/ − , initial
electron with ψ+ and initial photon with A
/ + . Note that we have substituted

d4 q
Z
SF (x − y) = exp(−iq(x − y))S̃F (q)
(2π)4

(the term −q 2 is the charge and is no related with the momentum dq). We didn’t write the spin
index just to relax the notation.
We found at each vertex we have a delta conservation:
Z
d4 x exp −i(q − p0 − k 0 )x = (2π)4 δ 4 (q − p0 − k 0 )


Z
d4 y exp(i(q − p − k)y) = (2π)4 δ 4 (q − p − k)

Now, integrating on q:
SfI i = (2π)4 δ 4 (p + k − p0 − k 0 )MfIi
where
MfIi = −q 2 ūs0 (p0 )/λ0 (k 0 )S̃F (p + k)/λ (k)us (p)
Let us compare this result with the Feynman rules. The digram of I is the one on the left
in fig 5.3.

1. go to the end of fermionic line

2. final electron, so ū(p0 )

3. QED vertex, so −iqγ ν (Lorentz index is chosen arbitrarily)

4. electron propagator, so S̃F (p + k) (or quivalent (p0 + k 0 ), in the propagator flows the total
4momentum)

5. QED vertes, so −iqγ µ

6. initial electron, so u(p)

7. two photons, so polarization ν (k 0 ), µ (k) (assuming real polarization)

If we put all contribution together we recast the Feynman amplitude calculated. Notice photon
polarization can be putted whatever we want, while for fermions we have to follow a precise
direction starting from the end of the fermionic line (because of the spinorial structure: spinors
are vectors and the relative propagatos are matricies).

II term

Now, let us focus the type II . Recall what we know from the theory:

(2)
SfIIi ≡ hf | SII |ii = (2π)4 δ 4 (p + k − p0 − k 0 )MfIIi
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 105

The type II diagram is the one on the right in fig 5.3. It is topologically different from the
one before: here the outgoing photon is created in the first vertex while the incoming photon is
destroyed in the second vertex. Again the two QED vertecies has lorentz indicies, for exapmle
let us denote −iqγ µ for the second (final) and −iqγ ν for the first (initial). Now we have to fix
momenta in each vertex: in the first one electron momentum p enters and photon momentum k 0
exits, so the net propagation is q = p − k 0 . If we follow Feynman rules we find the amplitude
MfIIi :
MfIIi = ū(p0 )(−iqγ µ )S̃F (p − k 0 )(−iqγ ν )u(p)µ (k)ν (k 0 )
Notice if there are positron we will use v, v † .
Finally, the total S-matrix element for the Compton scattering (at O(q 2 )) is
(2)
Sf i = (2π)4 δ 4 (p + k − p0 − k 0 )(MfIi + MfIIi )

The only thing to think inside Feynman rule is about the sign between the two diagram which
depends on the number of permutation we did with the fermionic fields. Notice fermionic line is
the same for diagram I and diagram II , and this assures the sign is plus, in fact what we
switch is where the bosonic operators are applied. If the fermionic line changes, we have to check
the permutations.

Exercise. It is an example with antiparticle. Compute the Feynman amplitude of

e+ γ → e+ γ

Solution: we know there are two diagrams that are summed, the only difference is the opposite
direction for the fermions since we deal with antifermions, so the end of the diagram is reversed.
For example for the first diagram

MA = v̄(p)(−iqγ µ )S̃F (p + k)(−iqγ ν )v(p0 )µ (k)ν (k 0 )

When we have a loop we have complication since we have two propagation, one can be
"absorbe" by the delta, but the other remains and general it diverges. This is a problem faced in
the renormalization. So higher than 2nd order there are problems following these procedures.

5.4 Observable result: the cross section


Now, we will focus and apply our knowledge to compute explicit results. What we experimentaly
measure is the cross section, so we want to see how to obtain it. We will see an intuitive definition
and how it is related to S-matrix. Notice another observable is the decay rate, but in QED there
is no decay.
So let us work out the cross section. Consider two particles interacting. At time t = −∞ they
are at distance ∆x = ∞, namely there is zero interacion between them, so the initial state is a
free particles initial state. At the finital state we assume an equal situation but just at t = +∞.
During the interaction the two particles modify their momentum from (p, q) to (p0 , q 0 ).
The easiest way to define the cross section is in the LAB frame. Suppose to have a target
that is a particle denoted with 2, and a beam of particles denoted with 1. When they meet they
scatter and what we measure is the energy-momentum for particles 10 and 20 using detectors,
particles will deviate by an angle θ wrt the line of the beam.
We want to measure the number of scattering event Ns per unit of volume V and time T .
It will be proportional to the number of particles in the beam and their velocity, namely the
density n01 and v10 , and to the density of the target n02 , at the LAB frame the target is at rest, so
v20 = 0. Let us denote the proportional parameter with σ, it is the cross section.

Ns
= σ(n01 v10 )n02
VT
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 106

The cross section has the dimension of an area:


[σ] = L2 = M −2
The physical meaning of the cross section is the effective area of interaction (how large the beam
sees the target).
Now, how is related the cross section to what we seen in the course?

Relation between S-matrix element and cross section


The S-matrix element is the object Sf i , it is defined by initial momenta p, q and p0 , q 0 . We can
prepare states with certain momenta but what we measure is not exactly q 0 , p0 since there is an
uncertaintly: we will end with an infinitesimal volume in the momentum space around the final
momenta, namely the final state is such that pf ∈ (pf , pf + δpf ). The transition rate between
initial to final state is
nf  3 
2
Y d pj V
dWf i ≡ |Sf i |
(2π)3
j=1
So in the LAB frame we can define
dWf i 1
dσ =
V T n01 v10 n02
this is how we connect the quantity we computed with the cross section.
Let us jump to the final explicit formula without focus on the details. Consider a process
where m1 = m01 and m2 = m02 (for example Compton scattering e− γ → e− γ), we call this process
of type A. In the center of mass frame, the differential cross section per solid angle is
1 |Mf i |2
 

=
dΩ CM 64π 2 s
where s is the total energy squared.
Other example is process where m1 = m2 and m01 = m02 (for example e+ e− → µ+ µ− ), this is
type B. The muon mass is larger that the electron and so we can approximate m1,2 = me ∼ 0
and m01,2 = mµ = M . In the center of mass frame, the differential cross section is
1/2
4M 2 |Mf i |2
  
dσ 1
= 1 −
dΩ CM 64π 2 s s
Notice the kinematic term (1 − 4M 2 /s)1/2 tells us we cannot produce massive states like muons
if we do not have enought energy (higher than 2 times the mass of the muon, it comes from the
delta conservations and so on).
The message is that whatever observable we want calculate there is a kinematic part and
a part depending on the theory we are calculating (it is in the Feynman amplitude). We can
combine them to obtain real observables like the cross section.

5.5 QED process at lowest order


We will now study tree level processes in QED: this means that we only consider QED processes 31/5/21
which have no loops. This usually corresponds to considering the leading order of S-matrix. We
will focus on physical process, the ones that kinematic allows. With higher order of S-matrix
expansion we have corrections of interactions. The perturbative approach cannot give a precise
result so we have to see how precise is the experimental measure and so we decide at which order
to stop the expansion.
We will look at two processes explicitly:
1. e+ e− → µ+ µ− scattering
2. e− γ → e− γ (Compton) scattering
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 107

5.5.1 Electron-positron to muon-antimuon


This is the simplest process to see. If we would to analize e+ e− → e+ e− we will have two channel:
t and s.
Let us write the QED lagrangian for more fermions (in our case electron and muon but the
formula is general):
1 1
LQED = − F µν Fµν − (∂µ Aµ )2 + ψ̄e (iD
/ e − me )ψe + ψ̄µ (iD
/ µ − mµ )ψµ
4 2ξ
where we recall for a generic fermion labeled for f :

Dfµ ≡ ∂ µ + iqf Aµ

Notice electron , muon and tau are three different lepton flavor of a fermion with the same charge
and spin, so qe = qf = −1. Since there are two types of fermions, we have to be careful about
the vertex that linked particles: the QED vertex between f and f¯ is −iqf γ µ . The lagrangian we
wrote there aren’t mixed terms between ψµ and ψe , so there is only one channel for the process
e+ e− → µ+ µ− at leading order (for example there are no operators to destroy electron and
create muon at the same point, so no t channel). Let us denote the spin of electrons with labels
r, s and momenta p, q, while for final muons let us use spin labels r0 , s0 and momenta p0 , q 0 . The
electron vertex is −iqe γ µ and the muon vertex is −iqµ γ ν where the Lorentz indicies are general.
The Feynman diagram at lowest order is the shown in fig 5.4. We use the conservation of the
momentum to find the momentum flowing in the propagator, it is

k = p + q = p0 + q 0

Notice the mass of the muon is bigger than the electron one, so let us denote

me = m
mµ = M

e−
r fr−0

p~ p~0

~k = p~ + ~q

−iqe γµ −iqf γν

~q ~q0

e+
s fs+0

Figure 5.4: Tree level diagram for electron-fermion interaction.

Let us compute the Feynman amplitude of the process using Feynman rules

Mf i = (iqe )(iqf ) ūr0 (p0 )γ ν vs0 (q 0 ) (v̄s (q)γ µ ur (p))D̃µν (k)




where D̃µν (k) is the photon propagator in the momentum space, recall it is
 
−i kµ kν
D̃µν (k) = 2 ηµν − (1 − ξ) 2
k k
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 108

here we can use the Feynman gauge ξ = 1. Sometimes the gauge is not fixed in order to see if
the final result is independent on the xi, if it is the case then things work nice.
Recall the Mandelstam variable s = (p + q)2 = (p0 + q 0 )2 , the final amplitude is
iqe qf
ūr0 (p0 )γ µ vs0 (q 0 ) (v̄s (q)γµ ur (p))

Mf i =
s
since k = p + q and k 2 = s. However in the cross section what is present is the modulus square
of the amplitude, so let us compute it. In general the amplitude can be a complex number, so its
square modulus is

|Mf i |2 = Mf i Mf∗i
 q q 2 ∗
e f
ū(p0 )γ µ v(q 0 ) ū(p0 )γ ν v(q 0 ) (v̄(q)γµ u(q))(v̄(q)γµ u(q))∗

=
s
Let us spell out each term separately to understand what they are, it is convenient to write also
the spinorial indicies. Start from the first conjugate term:
∗ ∗
ū(p0 )γν v(q 0 ) = u∗α (p0 )(γ 0 γ ν )αβ vβ (q 0 )
= uα (p0 )(γ0∗ γ ν ∗ )αβ vβ∗ (q 0 )
= uα (p0 )(γ0 γν∗ γ0 γ0 )αβ vβ∗ (q 0 )

Recall the properties of γ0 :

γ0 = γ0† = γ0∗
γ02 = I
γ0 㵆 γ0 = γµ

using them, we can rewrite γ0∗ :


∗  
ū(p0 )γν v(q 0 ) = vβ∗ (q 0 ) γν† γ0 uα (p0 )
αβ
= vβ∗ (q 0 )(γ0 γν )βα uα (p0 )
= v̄(q 0 )γν u(p0 )

So, the final result is an inversion of spinor fields position. Generalizing this result to each
conjugate term, the modulus square of the amplitide is
 q q 2 h i h i
e f
|Mf i |2 = ūr0 (p0 )γ µ v(q 0 )v̄s0 (q 0 )γ ν ur0 (p0 ) v̄s (q)γµ ur (p)ū(p)γν vs (q)
s f e

Unpolarized Feynman amplitude


To simplify the task let us switch from the computation of Feynman amplitude, that is teh one
we found and that is difficult to handle, to the unpolarized Feynman aplitude |M ¯ |2 . It is used
when we have a mixing spin of the initial beam, this is often the case in experiments and we
will assume it for our electron-muon scattering process. To compute the unpolarized Feynman
amplitude we have to average over the initial polarizations and we have to sum over the final
polarizations (if we have a particle with spin 1, we have three possible polarization, that are
-1,0,+1, and we have to average them). In our case we have electron and positron as initial state,
each with spin 1/2, so each particle has two possibilty (±1/2) and since we have two types of
particle we have to average over 4: precisely it comes out from (2se− + 1)(2se+ + 1). So
2 1 XX ¯ 2
|M¯f i | = |Mf i |r,s,r0 ,s0
4 r,s 0 0
r ,s
 q q 2 X   X 
e f
= ūr0 (p0 )γ µ v(q 0 )v̄s0 (q 0 )γ ν ur0 (p) × vs (q)γµ ur (p)ū(p)γν vs (q)
2s 0 0
f
r,s
e
r ,s
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 109

To simplify the expression let us put back the spinorial indicies, in this way we can move the
components whenever we want; precisely let us sum together each spin term:
  !
 q q 2 X 0 0 0
2 e f 0
|M¯f i | = ūαr0 (γ µ )α0 β 0 vsβ0 v̄sγ0 (γ ν )γ 0 δ0 uδr0  ×
X
 ūγr (γν )αβ vsβ v̄sγ (γµ )γδ uδr
2s
r0 ,s0 rs e
f
! ! ! !
 q q 2 X 0 0 X β0 γ0
e f
X X
= uδr0 ūαr0 vs0 v̄s0 (γ µ )α0 β 0 (γ ν )γ 0 δ0 × uδr ūαr vsβ v̄sγ (γ µ )αβ (γ ν )γδ
2s 0 0 r s
r s

Now we have to remember what are the sum of spinors:


X
ur (p)ūr (p) ≡ 2mΛ+ = p
/+m
r
X
vs (p)v̄s (p) ≡ −2mΛ− = p
/−m
s

they are the energy projectors. So the unpolarized Feynman amplitude is


 q q 2
2 e f
|M¯f i | = 0 0 µ ν
  
/ + M δ0 α0 /q − M β 0 γ 0 (γ )α0 β 0 (γ )γ 0 δ0 × p
p / + m δα (/q − m)βγ (γν )αβ (γµ )γδ
2s
Notice we "closed" the indicies for both first and secon term, namely it is like a trace calculation
for each term. In fact we can write our combination of spinors as
 q q 2 
2 e f
|M¯f i | = 0 µ 0 ν
  
Tr (p / + M )γ (/q − M )γ × Tr (p / + m)γν (/q − m)γµ
2s
This makes sense: there are no free spinorial indices in the probability, so we must trace them all
away. This trace result can be automitized in order to do not need to open each term as we done.
Now we have to compute the traces, let us do it as an exercise. Let us denote the first trace
as B and the second as A.
Exercise. Calculate traces A and B.
Solution: we do the A trace (so we do not have to write many primes), the other one is
precisely analogous.
2
       
A = Tr (p/ + m)γν (/q − m)γµ = Tr p/γν /qγν + m Tr γν /qγµ − Tr p/γν γµ − m Tr[γν γµ ]
Now, we have terms with two, three and four gamma matrices. This procedure is useful to
identify the number of gamma matricies present in each trace term since the following property
holds: the trace of an odd number of gamma matricies vanishies
Tr[ odd γ] = 0
Proof. Consider
Tr[γ µ ] = Tr[γ µ γ5 γ5 ] = − Tr[γ5 γ µ γ5 ] = − Tr[γ µ γ5 γ5 ] = − Tr[γ µ ] = 0
since γ5 γ5 = I and {γ5 , γµ } = 0 and we used the cyclicity of the trace. This idea can be used for
each trace of odd number of gamma matricies, and so the property is demostarted.

With this, we can throw away the terms with an odd number of γ matrices: so, we are left
with
2
(5.1)
   
Tr γµ (/q − m)γν (p / − m Tr[γµ γν ] ,
/ + m) = Tr γµ /qγν p
and we can evaluate the two pieces separately. Let us compute the term
Tr[γµ γν ] = Tr[{γµ , γν }] − Tr[γν γµ ]
= 2ηµν Tr[I] − Tr[γµ γν ] Cyclicity of
2 Tr[γµ γν ] = 2ηµν × 4 the trace.
Tr[γµ γν ] = 4ηµν .
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 110

where we used {γ µ , γ ν } = 2η µν I and the cyclicity of the trace. Then we have to compute also
the term
Tr[γ µ γ ν γ ρ γ σ ] = 4(η µν η ρσ − η µρ η νσ + η µσ η νρ )
Combining these, we get:
2 ν σ 2
 
/ − m Tr[γµ γρ ] = q p Tr[γµ γν γρ γσ ] − m Tr[γµ γρ ]
Tr γµ /qγρ p
= 4q ν pσ (ηµν ηρσ + ηµσ ηνρ − ηµρ ηνσ ) − 4m2 ηµρ
= 4 qµ pρ + pµ qρ − ηµρ p · q + m2 .


Finaly we can prove that

A = 4 (pµ q ν + pν q µ ) − (m2 + p · q)η µν


 
 µ ν µ ν
B = 4 p0 q 0 + q 0 p0 − M 2 p0 q 0 η µν
 

Now we have to contract these two terms. We can use the following approximation: M 2  m2 ,
so we can neglect some terms in the contraction and the result is
!
2 8qe2 qf2 
¯
|Mf i | = (p · p0 )(q · q 0 ) + (p · q 0 )(q · p0 ) + M 2 (p · q)

s 2

Everything we have derived so far is Lorentz-invariant, and the probability is a Lorentz scalar
with all the scalar possible combinations of 4momentum in our theory. However, we can specify
these results to a specific frame in order to make them more concretely useful. We choose the
center of mass frame.

COM frame
Let us calculate the unpolarized Feynman amplitude in the center of mass frame, as an example.
In CM frame we have electron and positron coming with spatial momentum p~, ~q and muon and
antimuon arises with spatial moentum p~0 , ~q0 with an angle θ wrt the initial spatial momentum.
Spatial momenta are equal in modulus and inverse in versus. Energies are equals since the masses
are the same of the incoming and outcoming particles: ωp = ωq = ω and ωp0 = ωq0 = ω 0 . Recall
in the CM the total mometum is zero and the total energy is the square root of the Mandelstam
variable s: √
(q + p) = (q 0 + p0 ) = ( s, 0)
Computing explicitely

(p + q)2 = 2m2 + 2pq = 4ω 2


2
(p0 + q 0 ) = 2M 2 + 2p0 q 0 = 4ω 0

but these two terms are equal to s and so

ω = ω0

To find the angle θ let us spell out the scalar products, use the approximation m ∼ 0 wrt M :

pp0 = ω 2 − |~p||p~0 | cos θ ' ω(ω − |p~0 | cos θ) = qq 0


pq 0 = ω 2 + |~
p||p~0 | cos θ ' ω(ω + |p~0 | cos θ) = qp0 |~
p| ' ω

2 2 2
pq = ω + |~
p| ' 2ω
p0 q 0 = ω 2 + |~
p0 |2 ' 2ω 2 − M 2 |p~0 |2 ' ω 2 − M 2

and so
M2
 
0 2 2
|~
p| =ω 1− 2
ω
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 111

Finaly, we can substitute the kinematic in the unpolarized Feynman amplitude, the result is
valid for an experiment done in the CM farme:
!
2 8qe2 qf2 h 4 i
|M¯f i |CM ' 2ω + 2ω 2 ~0 2
|p | cos 2
θ + 2M 2 2
ω
s2
4M 2 4M 2
    
4 2
=e 1+ + 1− cos θ
s s

So we can predict a probability of the angular direction of scattering. Notice we have an angular
depencence since we are scattering objects with spin (with scalar there would be not, each
direction will be equipropable).
The unpolarized cross section comes from the definition we already gave:
1/2 
1 |p~0 | ¯ 2 α2 4M 2 4M 2 4M 2
      

= 2 |M |CM = 1− 1+ + 1− cos θ
dΩ CM 6π |~p| 4s s s s

where we have introduced α = e2 /(4π 2 ). In this formula there are 4 important things:

1. kinematic: an observable cannot be complex so the square root must be positive, namely

we can produce only particles such that s > 2M

2. typical order of magnitude of the process: depends on the interaction, in QED it is governed
by α since it is the strenght of the coupling (vertex). The cross section is proportional to
α2 , since it is proportional to the square of the amplitude.

3. angular dependence: comes from the spin

4. higher energy corresponds smaller cross section (think particles as wavelength)

5.5.2 Compton scattering


Consider a process with an initial electron and photon. The process is the compton scattering: 1/6/21

e− γ → e− γ

It is similiar to the above one, there is just the substitution of the spinor fermion ψ µ with the
vector boson Aµ . There are two topological different Feynman diagrams of Compton scattering,
they are

γ γ γ γ

e− e− e− e−

The first is the s-chanell since the momentum flowing in the propagator is (p + k) = s and
we call it A, we fix the Lorentz indicies of vertecies as ν on left and µ on right. The second

diagram is the u-channel since the momentum flowing in the propagator is (p − k 0 ) = u and
we call it B, here Lorentz index µ on left and ν on right. This is the better choice of Lorentz
indicies, we will see in computation why.
Recall when we have two diagram we have to fix the relative sign between them, if the
fermionic line is the same the sign is a plus, while when it is inverse the sign is a minus. In
Compton scattering the sign is plus.
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 112

Now we compute the Feynman amplitude for A and B:


0
h i
MfAi = −qe2 λµ (k 0 )λν (k) ūr0 (p0 )γ µ S̃F (p + k)γ ν ur (p)
0
h i
MfBi = −qe2 λµ (k 0 )λν (k) ūr0 (p0 )γ ν S̃F (p − k 0 )γ µ ur (p)

where the Feynman propagator for fermions in momentum space is

i (/q + me )
S̃F (q) = =i 2
/q − me q − m2e

We can simplify the expression recalling the following properties (from now me = m to simplify
notation):

(p + k)2 − m2 = 2p · k
(p − k 0 )2 − m2 = −2p · k 0
ν ν
(p
/ + m)γ u(p) = 2p u(p)

where in the last formula we used the anticommutation properties of gamma matrix.
So the total amplitude is
" #
0
tot 2 λ0 0 λ 0 γ µ (k/γ ν + 2pν ) γ ν (k/ γ µ − 2pµ )
Mf i = −iqe µ (k )ν (k)ūr0 (p ) + ur (p)
2p · k 2p · k 0
0
≡ λµ (k 0 )λν (k)Lµν 0 0
r,r0 (k, k , p, p )

where we introduced the tensor Lµν which contains all the spinorial dependence. It is t important
since the Ward identities holds.
Theorem (Ward identities). When we contract the L tensor with the momentum of initial or
final photon the result is zero

kµ0 Lµν = 0 and kν Lµν = 0 .

Exercise. Prove the Ward identities.

5.6 Ward identities


The Ward identities are consequence of the gauge invariance of the theory. The presence of a
symmetry, as the local U (1) gauge symmetry, implies the arising of a lot of identientites.
Proposition. In whatever QED process with one or more external photons it is always possible
to factorize the Feynman amplitude Mf i as

Mf i = µ (k)Lµ (k1 , . . . )

where the Ward identity holds:


kµ Lµ (k1 , . . . ) = 0

Connection to gauge invariance


Let us see explicitly how the Ward identity is connected to gauge invariance. Recall the gauge
transformation
A0µ (x) = Aµ (x) + ∂µ α(x)
Let us go to the momentum space
Z
Aµ (x) ∼ d3 k exp(−ikx)µ (k)
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 113

with Z
α(x) = d3 k exp(−ikx)α̃(k)

so in momentum space the gauge transformation reads

0µ (k) = µ (k) − ikµ α̃(k)

namely it is the Fourier transform of the gauge invariance on coordinate space.


Recall observables are invariant under gauge transformation and the square of the Feynman
amplitude is proportional to observable (cross section), so also the amplitude must be invariant
under gauge transformtion. Suppose to have an external photon, so a Feynman amplitude
Mf i = µ (k)Lµ (k1 , . . . ); under a gauge transformation it reads

M 0 = 0µ (k)Lµ (k1 , . . . ) = µ (k)Lµ (k1 , . . . ) − iα̃(k)kµ Lµ (k1 , . . . )

and since it must be M 0 = M then the additional term is zero:

iα̃(k)kµ Lµ (k1 , . . . ) = 0

Sum over photon polarizations


From a practical point of view, the Ward identity allows to sum over the photon polarization.
Consider again the simple case with one external photon with momentum k and real polarization.
The Feynman amplitude is
M = λµ (k)Lµ (k1 , . . . )
While the square unpolarized amplitude is (recall to average over polarization, so multiply by
1/2 since photon has two polarization)
1X λ
|M̄ |2 = |µ (k)Lµ (k)|2
2
λ
1X λ
= µ (k)λν (k)Lµ (k1 , . . . )Lν † (k1 , . . . )
2
λ

Let us assume k µ = (k, 0, 0, k) then the physical polarizations are

µ(1) = (0, 1, 0, 0)
µ(2) = (0, 0, 1, 0)

(i)
since k µ µ = 0. So the sum over polarization becomes
2
µ(λ) ν(λ) Lµ L†ν = |L1 |2 + |L2 |2
X

λ=1

and using the Ward identity

kµ Lµ = |k|(L0 − L3 ) = 0 =⇒ L0 = L3
(i)
again since k µ µ = 0. This simplify our life since
2
µ(λ) ν(λ) Lµ L†ν = −|L0 |2 + |L1 |2 + |L2 |2 + |L3 |2 = −η µν Lµ L†ν
X

λ=1

this is how to sum over polarization.


CHAPTER 5. QUANTUM ELECTRO DYNAMIC 114

Return to the unpolarized Feynman amplitude of Compton scattering: now we have to


average for initial photon and electron, so times 1/4
1 X (λ0 )∗ 0 (λ0 ) 0 X (λ) X 
†ρσ

|M̄ |2 = µ (k )ρ (k ) ν (k)(λ)∗
σ (k) (Lµν
) rr 0 L
4 0 rr0
λ λ rr 0
| {z }| {z }
−ηµρ −ηνσ
1X  
= (Lµν )rr0 L†µν
4 rr0
rr0

In detail
" #
µ (k µ + 2pν ) ν (k 0 µ µ)
γ / γ γ / γ − ep
Lµν
rr0 = −iqe2 ūr0 (p0 ) + ur (p)
2pk 2pk 0
" #
νk ν )γ µ µk 0 µ )γ ν
µν (γ / + 2p γ / − 2p
L† rr0 = iqe2 ūr (p) + u†r0 (p0 )
2pk 2pk 0

Notice it comes out a trace computation as the first example, however now we have just a
complicate expression since we are considering two diagrams. So
" ! !#
4 µk ν + 2γ µ pν νk 0 µ µγν 0
q 0 γ / γ γ / γ − 2p γ ν /
k γ µ + 2γ µ p ν γ µ /
k γ ν − 2γ ν p µ
|M̄f i |2 = e Tr (p/ + m) + × (p
/ + m) +
4 2pk 2pk 0 2pk 2pk
q 4 TAA
 
TAB + TBA TBB
= e + +
4 (2pk)2 (2pk)(2pk 0 ) (2pk 0 )2

where

A = γ µ k/γ ν + 2γ µ pν
0
B = γ ν k/ γ µ − 2γ ν pµ

here TAB means Tr[AB]. We must find a sum of traces since observable does not have spinorial
indicies and, given a matrix, the trace does not have indicies.
Now, we have to compute all the traces, this is done as an exercise.

Exercise. Calculate TAA , TBB , TAB and TBA .


Solution: notice there are useful properties:

TBB = TAA (k → k 0 ) and TAB = TBA

Let us explicit calculate the trace TAA :


 0 µ ν µ ν

TAA = Tr (p / + m)(γ k/γ + 2γ p )(p
/ + m)(γν k/γµ + 2γµ pν )

First thing to do is to classify components of the trace wrt the number of gamma matricies since
the terms with an odd number are zero, so opening the terms:
 0 µ ν 
1st = + Tr p
/ γ kγ p/γν k/γµ
 0 µ ν   0 µ ν 
2nd = +2pν Tr p / γ k/γ p/γµ + tr p
/γ p /γ k/γµ
3rd = +m2 Tr[γ µ k/γ ν γν k/γµ ]
 0 µ
4th = +4p2 Tr p

/γ p / γν
5th = +2m2 pν (Tr[γ µ k/γ ν γµ ] + Tr[γ µ γ ν k/γµ ])
6th = +4m2 p2 Tr[γ µ γµ ]
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 115

Notice there is always gamma matricies and something in the middle, so one can show (using
the anticommutation properties of gamma matrix) the following relations

γ µ γµ = 4
γµp
/γµ = −2p
/
γµp
//qγµ = 4pq
γµp
//qk/ = −2k//qp
/

apply these relations with also the kinematic relations p2 = p0 2 = m2 and k 2 = k 0 2 = 0 we obtain

1st = 32(pk)(p0 k)
2nd = −16m2 (p0 k)
3rd = 0
4th = −32m2 (pp0 )
5th = 64m2 (pk)
6th = 64m4

At the end, putting all together:

TAA = 16 2(pk)(p0 k) + 2m2 2(pk) − (pp0 ) − (kp0 ) + 4m2


  

The term TBB is equal to TAA just sending k → k 0 , so explictely:

TBB = 16 2(pk 0 )(p0 k 0 ) + 2m2 2(pk 0 ) − (pp0 ) − (k 0 p0 ) + 4m2


  

Remaining traces are

TAB = TBA = −16 2m4 + m2 ((pk) − (pk 0 ))


 

Now we are ready to calculate the unpolarized Feynman amplitude:


" 2 #
0
   
pk pk 1 1 1 1
|M̄ |2 = 2qe4 + + 2m2 − + m4 −
pk 0 pk pk pk 0 pk pk 0

To evaluate this quantity there are two principal frame: the LAB and CM frame. Last time
we saw the process in the CM frame, so now let us see Compton scattering in the LAB frame.
Consider a NR limit. In the LAB frame we have an incoming photon with momentum k, an
electron target, an outgoing photon with momenttum k 0 and electron with p0 , the angle of the
outgoing photon wrt the incoming direction is θ. So for the incoming particles

p = (m, 0)
k = (ωγ , 0, 0, ωγ )

while for the outgoing particles

p0 = (E 0 , p~0 )
k 0 = (ω 0 , ~k 0 )
γ

The energy of the outgoing photon ωγ0 can be related with other quantities using the kinematic:

(p0 )2 = (p + k − k 0 )2 = m2 + 2pk − 2pk 0 − 2kk 0 = m2 =⇒ kk 0 = p(k − k 0 )

using the explicit expression of the 4momentum:


ωγ
ωγ ωγ0 (1 − cos θ) = m(ωγ − ωγ0 ) =⇒ ωγ0 = (5.2)
1 + ωγ (1 − cos θ)/m
CHAPTER 5. QUANTUM ELECTRO DYNAMIC 116

So
ωγ0
  
ωγ
|M̄f i |2LAB = 2qe4 + 0 2
− sin θ
ωγ ωγ

the amplitude depends only on the energy of photons and the angle θ.
Now we can put it in the formula of teh cross section in the LAB frame:
2
ωγ0
  
dσ 1 1
= |M̄f i |2LAB
dΩ LAB 64π 2 m2e ωγ
2 
qe4 1 ωγ0 ωγ0
  
ωγ
= + 0 − sin2 θ
32π 2 m2e ωγ ωγ ωγ

integrating in the azimutal angle ϕ gives the Klein-Nishima formula (1929):

πα2 ωγ0
  0
ωγ
   
dσ ωγ
= 2 + 0 − sin2 θ
d cos θ LAB m ωγ ωγ ωγ

This is a complete general formula, we did not make any assumptions about the masses (apart
the fact it is a perturbative result up to O(q 2 )).
In the low energy (LE) limit ωγ  me and from eq (5.2) we have

ωγ0 ' ωγ

neglecting the mass term, namely the scattering is elastic. This also implies ωγ0  me . So in the
low energy limit the cross section is

πα2
 

= 2 1 + cos2 θ

d cos θ LAB in LE limit m

and this is the formula of the Thomson scattering. Amazing result: after builing QED, in the
low energy limit we found the known Thomson scattering. So our new QED theory is consistent
with the low energy physics. Recall we did theoretical calculation with a precision of α2 . If we
have to be more precise then we will add higher order and we add loop diagrams.

We are seeing life as a perturbative approach.


Stefano Rigolin

You might also like